Policy Assessment for the Review of the
Particulate Matter National Ambient Air
Quality Standards
-------
This page is intentionally blank.
-------
EPA452/R-11-003
April 2011
Policy Assessment
for the Review of the Particulate Matter
National Ambient Air Quality Standards
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Research Triangle Park, North Carolina
-------
DISCLAIMER
This document has been reviewed by the Office of Air Quality Planning and Standards
(OAQPS), U.S. Environmental Protection Agency (EPA), and approved for publication. This
OAQPS Policy Assessment contains conclusions of the staff of the OAQPS and does not
necessarily reflect the views of the Agency. Mention of trade names or commercial products is
not intended to constitute endorsement or recommendations for use.
-------
ACKNOWLEDGMENTS
This Policy Assessment is the product of the Office of Air Quality Planning and
Standards (OAQPS). It has been developed as part of the Environmental Protection Agency's
(EPA) ongoing review of the national ambient air quality standards (NAAQS) for particulate
matter (PM). The PM NAAQS review team has been led by Ms. Beth Hassett-Sipple. Dr. Karen
Martin has managed the project. For the chapter on health effects associated with fine particle
exposures and the primary PM2.5 standards, the principal authors include Ms. Beth Hassett-
Sipple, Dr. Pradeep Raj an, and Dr. Zach Pekar. For the chapter on health effects associated with
thoracic coarse particle exposures and the primary PMio standard, the principal author is Dr.
Scott Jenkins. For the chapter on visibility-related effects and the secondary PM2 5 standards, the
principal authors include Dr. Marc Pitchford (National Oceanic and Atmospheric
Administration, NOAA) and Mr. Phil Lorang. For the chapter on other welfare effects and
secondary PM standards, the principal author is Dr. Meredith Lassiter. The principal
contributors of ambient monitoring information, as presented in the introductory chapter,
Appendix B, and other analyses, are Mr. Tim Hanley and Ms. Joann Rice. The authors would
also like to acknowledge Mr. Mark Schmidt, Dr. Adam Reff, Dr. Michael Rizzo, and Mr. Mark
Evangelista for conducting analyses on specific topics. Staff from other EPA offices, including
the Office of Research and Development and the Office of General Counsel, also provided
valuable comments and contributions.
Earlier drafts of this document were reviewed by the Clean Air Scientific
Advisory Committee (CASAC) and made available for public comment. This final document
has been informed by the expert advice and comments received from CASAC, as well as by
public comments.
-------
This page left intentionally blank.
-------
TABLE OF CONTENTS
List of Tables iv
List of Figures v
List of Acronyms/Abbreviations vi
EXECUTIVE SUMMARY ES-1
1. INTRODUCTION 1-1
1.1 PURPOSE 1-1
1.2 BACKGROUND 1-2
1.2.1 Legislative Requirements 1-2
1.2.2 Previous PM NAAQS Reviews 1-4
1.2.3 Litigation Related to the 2006 PM Standards 1-9
1.2.4 Current PM NAAQS Review 1-10
1.3 CURRENT AMBIENT MONITORING NETWORKS FOR PM 1-12
1.3.1 PM2.5Mass 1-13
1.3.2 PM2.5 Speciation 1-14
1.3.3 PMioMass 1-14
1.3.4 PMio-2.5Mass 1-15
1.4 GENERAL APPROACH AND ORGANIZATION OF THIS DOCUMENT 1-15
1.5 REFERENCES 1-17
2. REVIEW OF THE PRIMARY STANDARDS FOR FINE PARTICLES 2-1
2.1 APPROACH 2-1
2.1.1 Approaches Used in Previous Reviews 2-2
2.1.1.1 Review Completed in 1997 2-2
2.1.1.2 Review Completed in 2006 2-4
2.1.2 Remand of Primary Annual PM2.5 Standard 2-6
2.1.3 General Approach Used in Current Review 2-8
2.2 ADEQUACY OF CURRENT STANDARDS 2-17
2.2.1 Evidence-based Considerations 2-17
2.2.2 Risk-based Considerations 2-36
2.2.3 CASAC Advice 2-47
2.2.4 Staff Conclusions on Adequacy of Current Standards 2-48
2.3 CONSIDERATION OF ALTERNATIVE STANDARDS 2-49
2.3.1 Indicator 2-49
2.3.2 Averaging Times 2-56
i
-------
2.3.3 Forms 2-58
2.3.3.1 Form of the Annual Standard 2-59
2.3.3.2 Form of the 24-Hour Standard 2-60
2.3.4 Alternative Levels 2-62
2.3.4.1 Evidence-based Considerations 2-63
2.3.4.2 Risk-based Considerations 2-92
2.3.4.3 CASAC Advice 2-100
2.3.4.4 Staff Conclusions on Alternative Standard Levels 2-101
2.4 SUMMARY OF STAFF CONCLUSIONS ON PRIMARY FINE PARTICLE
STANDARDS 2-104
2.5 KEY UNCERTAINTIES AND AREAS FOR FUTURE RESEARCH AND DATA
COLLECTION 2-106
2.6 REFERENCES 2-112
3. REVIEW OF THE PRIMARY STANDARD FOR THORACIC COARSE
PARTICLES 3-1
3.1 APPROACH 3-1
3.1.1 Approaches Used in Previous Reviews 3-2
3.1.1.1 Reviews Completed in 1987 and 1997 3-2
3.1.1.2 Review Completed in 2006 3-3
3.1.2 Litigation of 2006 Final Rule for Thoracic Coarse Particles 3-5
3.1.3 General Approach Used in Current Review 3-6
3.2 ADEQUACY OF THE CURRENT PMio STANDARD 3-9
3.2.1 Evidence-Based Considerations 3-9
3.2.2 CASAC Conclusions and Recommendations 3-23
3.2.3 Staff Conclusions on Adequacy of Current PMio Standard 3-24
3.3 CONSIDERATION OF POTENTIAL ALTERNATIVE STANDARDS 3-27
3.3.1 Indicator 3-27
3.3.2 Averaging Time 3-30
3.3.3 Form 3-31
3.3.4 Level 3-35
3.4 SUMMARY OF STAFF CONCLUSIONS ON PRIMARY THORACIC COARSE
PARTICLE STANDARD 3-50
3.5 KEY UNCERTAINTIES AND AREAS FOR FUTURE RESEARCH AND DATA
COLLECTION 3-51
3.6 REFERENCES 3-54
-------
4. REVIEW OF THE SECONDARY STANDARDS FOR VISIBILITY-RELATED
EFFECTS 4-1
4.1 APPROACH 4-1
4.1.1 Approaches Used in Previous Reviews 4-2
4.1.1.1 Review Completed in 1997 4-2
4.1.1.2 Review Completed in 2006 4-4
4.1.2 Remand of Secondary PM2.5 Standards 4-8
4.1.3 General Approach Used in Current Review 4-9
4.2 ADEQUACY OF CURRENT STANDARDS 4-12
4.2.1 Evidence-based and Impact-Based Considerations 4-12
4.2.2 CASAC Advice 4-39
4.2.3 Staff Conclusions on Adequacy of Current Standards 4-39
4.3 CONSIDERATION OF ALTERNATIVE STANDARDS 4-40
4.3.1 Indicator 4-41
4.3.1.1 Evidence-based and Impact-based Considerations 4-41
4.3.1.2 CASAC Advice on Indicator 4-51
4.3.1.3 Staff Conclusions on Indicator 4-51
4.3.2 Averaging Time and Related Considerations 4-52
4.3.2.1 Sub-daily Averaging Times 4-52
4.3.2.2 24-Hour Averaging Time 4-55
4.3.2.3 CASAC Advice on Averaging Time 4-56
4.3.2.4 Staff Conclusions on Averaging Time 4-57
4.3.3 Form 4-58
4.3.4 Level 4-60
4.4 SUMMARY OF STAFF CONCLUSIONS ON SECONDARY STANDARDS FOR
VISIBILITY-RELATED EFFECTS 4-63
4.5 KEY UNCERTAINTIES AND AREAS FOR FUTURE RESEARCH AND DATA
COLLECTION 4-66
4.6 REFERENCES 4-68
5. REVIEW OF THE SECONDARY STANDARDS FOR OTHER WELFARE
EFFECTS 5-1
5.1 APPROACH 5-1
5.1.1 Approaches Used in Previous Reviews 5-2
5.1.1.1 Review Completed in 1997 5-2
5.1.1.2 Review Completed in 2006 5-2
5.1.2 Scope of Current NAAQS Reviews 5-3
5.1.2.1 Scope of the Current Secondary PM NAAQS Review 5-3
5.1.2.2 Scope of the Current NOx/SOx Secondary NAAQS Review 5-5
5.1.3 General Approach Used in Current Review 5-5
iii
-------
5.2 CLIMATE 5-6
5.2.1 Scope 5-6
5.2.2 Adequacy of the Current Standards 5-7
5.2.3 Staff Conclusions 5-10
5.2.4 Key Uncertainties and Areas for Future Research and Data Collection 5-12
5.3 ECOLOGICAL EFFECTS 5-13
5.3.1 Scope 5-13
5.3.2 Adequacy of the Current Standards 5-16
5.3.3 Staff Conclusions 5-23
5.3.4 Key Uncertainties and Areas for Future Research and Data Collection 5-25
5.4 MATERIALS 5-25
5.4.1 Scope 5-25
5.4.2 Adequacy of the Current Standards 5-26
5.4.3 Staff Conclusions 5-28
5.4.4 Key Uncertainties and Areas for Future Research and Data Collection 5-29
5.5 REFERENCES 5-30
APPENDICES
A. Clean Air Scientific Advisory Committee Letter on Second Draft Policy Assessment,
September 10, 2010 A-l
B. Ambient PM Monitoring Networks B-l
C. Predicted Percent of Counties with Monitors Not Likely to Meet Current and Alternative
Primary Annual and 24-hour PM2.s Standards C-l
D. Predicted Percent of Counties with Monitors Not Likely to Meet Current and Alternative
Primary 24-hour PMi0 Standards D-l
E. Information Regarding a 1-hour PM2.5 Mass Indicator E-l
F. Two Simplified Approaches to Calculate Hourly PM2.s Light Extinction Values from
Hourly PM2.5 Mass and Relative Humidity Data Plus 24-hour Mean PM2.5 Composition
Data F-l
G. Calculated 24-hour Average PM2.5 Light Extinction and Adjusted Candidate Protection
Levels G-l
H. Predicted Percent of Counties with Monitors Not Likely to Meet Current and Alternative
Secondary PM25 Standards H-l
IV
-------
LIST OF TABLES
Table 1-1. Summary of National Ambient Air Quality Standards Promulgated for Particulate
Matter 1971-2006 1-5
Table 2-1. Summary of Causality Determinations for PM2.5 2-19
Table 2-2. Estimated Incidence and Percent of Total Annual Incidence Associated with
Long-term PM2.5 Exposure Based on Simulation of the Current Suite of Standards
(for fflD mortality based on 2007 PM2.5 Concentrations) 2-44
Table 2-3. Estimated Incidence and Percent of Total Annual Incidence Associated with Short-
Term PM2.5 Exposure Based on Simulation of the Current Suite of Standards (CV
mortality and hospital admissions based on 2007 PM25 concentrations) 2-45
Table 3-1. Summary of Causality Determinations for PMio-2.s 3-10
Table 3-2. Predicted Counts of Counties, and Population (x 1,000) within those Counties, Not
Likely to Meet the Current PMio Standard and Alterantive PMio Standards with 98th
Percentile Forms (based on air quality in 2007-2009) 3-44
Table 3-3. PMio Concentrations in Locations that Met Current PMio Standard and Where
Positive and Statistically Significant Associations with PMi0.2.5 Have Been
Reported 3-45
Table 4-1. Logit Model-estimated Visual Air Quality Values (in deciviews) Corresponding to
Various Percent Acceptability Values for the Four Cities 4-25
Table 4-2. Percentage of 2005-2007 Daily Maximum Daylight Hourly Values of PMio Light
Extinction Exceeding CPLs (excluding hours with relative humidity > 90%) 4-31
Table 4-3. Percentage of 2007-2009 Daily Maximum Daylight Hourly Values of PM2.5 Light
Extinction Exceeding CPLs (excluding hours with relative humidity > 90%) 4-32
Table 4-4. Percentage of Daily Maximum Daylight Hourly Values of PMio Light Extinction
Exceeding CPLs when Just Meeting the Current PM2 5 NAAQS (excluding hours
with relative humidity > 90%) 4-34
Table 4-5. Percentage of Daily Maximum Daylight 4-Hour Average Values of PMio Light
Extinction Exceeding CPLs when Just Meeting the Current PM2 5 NAAQS
(excluding hours with relative humidity > 90%) 4-36
Table 4-6. Percentage of Daily Maximum Daylight 1-Hour Average Values of PM2 5 Light
Extinction Exceeding CPLs when Just Meeting the Current PM2 5 NAAQS
(excluding hours with relative humidity > 90%) 4-37
Table 4-7. Percentage of Daily Maximum Daylight 4-Hour Average Values of PM2 5 Light
Extinction Exceeding CPLs when Just Meeting the Current PM2 5 NAAQS
(excluding hours with relative humidity > 90%) 4-38
Table 5-1. Scope of the Current Secondary PM NAAQS Review and Current NOx/SOx
secondary NAAQS review 5-4
-------
LIST OF FIGURES
Figure 2-1. Overview of Approach for Review of Primary PM2 5 Standards 2-16
Figure 2-2. Distribution of Site-level Variation in 98th and 99th Percentile Concentrations, as
Measured by Coefficient of Variation (SD/Mean) Computed by Site Across Years,
2000-2008 2-62
Figure 2-3. Confidence Intervals Around PM2.5 Concentration-Response Relationships -
Information from Multi-city Epidemiological Studies 2-67
Figure 2-4. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for
Multi-city, Long-term PM2.5 Exposure Studies of the General Population and
Older Adults 2-73
Figure 2-5. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for
Multi-City, Long-term PM2.5 Exposure Studies of Children 2-74
Figure 2-6. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for Multi-
City, Short-term PM2.5 Exposure Studies of the General Population and Older Adults
2-75
Figure 2-7. Distribution of Population-Level Data and Corresponding PM2.5 Concentrations for
Selected Multi-City Epidemiological Studies 2-76
Figure 2-8. Translating Epidemiological Evidence from Multi-City Exposure Studies into
Annual PM2.5 Standards 2-80
Figure 2-9. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for
Single-City, Short-term PM2.5 Exposure Studies 2-87
Figure 2-10. County-level 24-hour DVs versus Annual DVs, 2007-2009 2-90
Figure 2-11. Percent Reduction in Long-term Exposure-related Mortality Risk 2-95
Figure 2-12. Percent Reductions in Short-term Exposure-related Mortality and Morbidity
Risk 2-96
Figure 3-1. Overview of Approach for Review of Primary PMio Standard 3-8
Figure 3-2. PMio Air Quality and PMio-2.5 Effect Estimates in Locations of U.S. Single-City
PMio-2.5 Mortality Studies 3-19
Figure 3-3. PMio Air Quality and PMio-2.5 Effect Estimates in Locations of U.S. Single-City
PMio-2.5 Morbidity Studies 3-20
Figure 3-4. Site-Level Ratio of 24-Hour PMio-2.5 to 24-Hour PMio Concentrations 3-28
Figure 3-5. 98th Percentile PMio Concentrations in Locations of U.S. Single-City PMio-2.5
Mortality Studies 3-37
Figure 3-6. 98th Percentile PMio Concentrations in Locations of U.S. Single-City PMio-2.5
Morbidity Studies 3-38
Figure 3-7. Composite 3-year PMio 98th percentile 24-Hour Average Concentration Versus the
PMio Expected Exceedance Concentration-equivalent Design Value
(1988-2008) 3-41
Figure 3-8. Regional 3-year PMio 98th percentile 24-Hour Average Concentration Versus the
PMio Expected Exceedance Concentration-equivalent Design Value
(1988-2008) 3-42
VI
-------
Figure 4-1. Overview of Approach for Review of Secondary PM2.5 Standards 4-11
Figure 4-2. Summary of Results of Urban Visibility Studies in Four Cities, showing the
Identified Range of the 50% Acceptance Criteria 4-24
Figure 4-3. Distribution of Estimated Daily Maximum Daylight 1-hour PMio Light Extinction
Across the 2005-2007 Period, by Study Area (excluding hours with relative humidity
greater than 90%) 4-31
Figure 4-4. Distribution of Estimated Daily Maximum Daylight 1-hour PM2.5 Light Extinction
Across the 2007-2009 Period, by Study Area (excluding hours with relative humidity
greater than 90%) 4-32
Figure 4-5. Distribution of Daily Maximum Daylight 1-hour PMio Light Extinction when Rolled
Back to Just Meet Current PM2.5 NAAQS, by Study Area (excluding hours with
relative humidity greater than 90%) 4-34
Figure 4-6. Distribution of Daily Maximum Daylight 4-hour Average PMio Light Extinction
when Rolled Back to Just Meet Current PM2.5 NAAQS, by Study Area (excluding
hours with relative humidity greater than 90%) 4-36
Figure 4-7. Distribution of Daily Maximum Daylight 1-hour Average PM2.5 Light Extinction
when Rolled Back to Just Meet Current PM2 5 NAAQS, by Study Area (excluding
hours with relative humidity greater than 90%) 4-37
Figure 4-8. Distribution of Daily Maximum Daylight 4-hour Average PM2 5 Light Extinction
when Rolled Back to Just Meet Current PM2.5 NAAQS, by Study Area (excluding
hours with relative humidity greater than 90%) 4-38
Figure 4-9. Relationship between Daylight 1-hour PM2 5 Mass Concentration vs. Same-Hour
Calculated Light Extinction for!wo Cities 4-43
Vll
-------
LIST OF ACRONYMS/ABBREVIATIONS
AAMMS
ACS
AOD
ANPR
ANS
AQCD
AQS
AR4
BC
BMA
C
Ca
CAA
CAPs
CASAC
CBS A
CBVD
CCN
CCSP
Cd
CF
CFR
CHD
CHF
CHS
CH4
CO
CO2
COPD
CPL
Cr
C-R
Ambient Air Monitoring and Methods Subcommittee
American Cancer Society
Aerosol optical depth
Advance notice of proposed rulemaking
Autonomic nervous system
Air Quality Criteria Document
EPA's Air Quality System
Fourth Assessment Report of the Intergovernmental Panel on Climate Change
Black carbon
Bayesian model averaging
Carbon
Calcium
Clean Air Act
Concentrated ambient particles
Clean Air Scientific Advisory Committee
Consolidated Business Statistical Area
Cerebrovascular disease
Cloud Condensation Nuclei
US Climate Change Science Program
Cadmium
Cystic fibrosis
Code of Federal Regulations
Coronary heart disease
Congestive heart failure
Children's Health Study
Methane
Carbon monoxide
Carbon dioxide
Chronic obstructive pulmonary disease
Candidate protection level
Chromium
Concentration-response
Vlll
-------
CSN
Cu
cv
CVD
DE
DEP
DHEW
DLEE
dv
EC
ED
EPA
FEM
FEVi
FRM
GHG
HA
HEI
HF
Hg
IHD
IMPROVE
IPCC
IRP
ISA
IT
IUGR
Km
LC
LML
US
|ig/m3
MCAPS
Chemical Speciation Network
Copper
Cardiovascular
Cardiovascular disease
Diesel Exhaust
Diesel Exhaust Particles
US Department of Health, Education, and Welfare
Dry light extinction efficiency
deciview
Elemental carbon
Emergency department
Environmental Protection Agency
Federal Equivalent Method
Forced expiratory volume in one second, volume of air exhaled in first
second of exhalation
Federal Reference Method
Greenhouse gas
Hospital admissions
Health Effects Institute
Hygroscopic fraction
Mercury
Ischemic heart disease
Interagency Monitoring of Protected Visual Environment
Intergovernmental Panel on Climate Change
Integrated Review Plan
Integrated Science Assessment
Intratracheal
Intrauterine growth restriction, intrauterine growth retardation
Kilometer
Local conditions
Lowest measured level
microgram
micrometer, micron
micrograms per cubic meter
Medicare Air Pollution Study
IX
-------
MEA
MI
MLEE
Mm
Mm'1
MSA
N
NAAQS
NCEA
NCore
Ni
NMMAPS
NO
NO2
MV
NOx
NOAA
NRC
03
OAQPS
OAR
OC
OCM
OMB
ORD
PA
PAH
Pb
PBDES
PM
PMo.i
Millennium Ecosystem Assessment
Myocardial infarction
Moist light extinction efficiency
Megameter
Inverse megameters, l/(million meters)
Metropolitan Statistical Area
Nitrogen
National Ambient Air Quality Standards
National Center for Environmental Assessment
National Core Monitoring Network
Nickel
National Morbidity, Mortality, and Air Pollution Study
Nitric oxide
Nitrogen dioxide
Nitrate
Nitrogen oxides (NO+NO2)
National Oceanic and Atmospheric Administration
National Research Council
Ozone
Office of Air Quality Planning and Standards
Office of Air and Radiation
Organic carbon
Organic carbonaceous material
Office of Management and Budget
Office of Research and Development
Policy Assessment
Polyaromatic hydrocarbon
Lead
Polybrominated diphenyl ethers
Particulate matter
In general terms, particulate matter with a mobility diameter less than or
equal to 0.1 jim; a measurement of ultrafme particles; since no reference
method has been established, no 50% cut-point has been specified
-------
PM2.5 In general terms, particulate matter with an aerodynamic diameter less
than or equal to a nominal 2.5 um; a measurement of fine particles
In regulatory terms, particles with an upper 50% cut-point of 2.5 um
aerodynamic diameter (the 50% cut point diameter is the diameter at
which the sampler collects 50% of the particles and rejects 50% of the
particles) and a penetration curve as measured by a reference method
based on Appendix L of 40 CFR Part 50 and designated in accordance
with 40 CFR Part 53, by an equivalent method designated in accordance
with 40 CFR Part 53, or by an approved regional method designated in
accordance with Appendix C of 40 CFR Part 58
PMio In general terms, particulate matter with an aerodynamic diameter less
than or equal to a nominal 10 um; a measurement of thoracic particles
(i.e., that subset of inhalable particles thought small enough to penetrate
beyond the larynx into the thoracic region of the respiratory tract)
In regulatory terms, particles with an upper 50% cut-point of 10± 0.5 um
aerodynamic diameter (the 50% cut point diameter is the diameter at
which the sampler collects 50% of the particles and rejects 50% of the
particles) and a penetration curve as measured by a reference method
based on Appendix J of 40 CFR Part 50 and designated in accordance with
40 CFR Part 53 or by an equivalent method designated in accordance with
40 CFR Part 53
PMio-2.5 In general terms, particulate matter with an aerodynamic diameter less
than or equal to a nominal 10 um and greater than a nominal 2.5 um; a
measurement of thoracic coarse particulate matter or the coarse fraction of
PM10
In regulatory terms, particles with an upper 50% cut-point of 10 um
aerodynamic diameter and a lower 50% cut-point of 2.5 um aerodynamic
diameter (the 50% cut point diameter is the diameter at which the sampler
collects 50% of the particles and rejects 50% of the particles) as measured
by a reference method based on Appendix O of 40 CFR Part 50 and
designated in accordance with 40 CFR Part 53 or by an equivalent method
designated in accordance with 40 CFR Part 53
POM Particulate organic matter
POP Persistant organic pollutants
PRB Policy-relevant background
QAPP Quality Assurance Project Plan
RA Risk Assessment
xi
-------
REA
RF
RH
SANDWICH
SCAB
SD
SES
S
SIP
SLAMS
SO2
SOX
STN
STP
TB
TSP
UFPs
UFVA
V
VAQ
VOC
W/m2
WACAP
WHI
Zn
Risk and Exposure Assessment
Radiative forcing
Relative humidity
Sulfate, Adjusted Nitrate, Derived Water, Inferred Carbonaceous mass
approach
South Coast Air Basin (CA)
Standard deviation
Socioeconomic status
Sulfur
State implementation plan
State and local air monitoring stations
Sulfur dioxide
Sulfur oxides
Speciation Trends Network
Standard temperature and pressure
Tracheobronchial
Total suspended particulate
Ultrafine particles
Urban-Focused Visibility Assessment
Vanadium
Visual Air Quality
Volatile organic compounds
Watts per square meter
Western Airborne Contaminants Assessment Project
Women's Health Initiative
Zinc
Xll
-------
EXECUTIVE SUMMARY
This Policy Assessment (PA) has been prepared by staff in the Environmental
Protection Agency's (EPA) Office of Air Quality Planning and Standards (OAQPS) in
conjunction with the Agency's ongoing review of the national ambient air quality
standards (NAAQS) for particulate matter (PM), which include primary (health-based)
and secondary (welfare-based) standards. It presents staff conclusions regarding the
adequacy of the current suite of PM standards as well as potential alternative standards
for consideration in this review.
Staff conclusions are based on the scientific and technical information, as well as
uncertainties and limitations related to this information, assessed in other EPA
documents, including the Integrated Science Assessment for Particulate Matter (Final
Report) (ISA, US EPA, 2009a), the Quantitative Health Risk Assessment for Particulate
Matter (Final Report) (RA, US EPA, 2010a) and the Particulate Matter Urban-Focused
Visibility Assessment (Final Report) (UFVA, US EPA, 201 Ob). This PA is intended to
"bridge the gap" between the relevant scientific evidence and technical information and
the judgments required of the EPA Administrator in determining whether, and if so how,
to revise the PM NAAQS. The current and potential alternative PM standards are
considered in terms of the basic elements of the NAAQS: indicator, averaging time,
form, and level.
Primary Standards for Fine Particles (Chapter 2):
In assessing the adequacy of the current suite of annual and 24-hour PM2.5
standards meant to protect public health against long- and short-term exposures to fine
particles, staff concludes that the currently available information clearly calls into
question the adequacy of the current standards and that consideration should be given to
revising the suite of standards to provide increased public health protection. In
considering alternative PM2.5 standards, staff concludes that protection from both long-
and short-term PM2.5 exposures can most effectively and efficiently be provided by
relying primarily on the annual standard, with the 24-hour standard providing
supplemental protection for days with high peak concentrations.
Taking into account both evidence-based and risk-based considerations, staff
concludes that consideration should be given to revising the current annual PM2.5
standard level of 15 |ig/m3 to a level within the range of 13 to 11 |ig/m3. Staff further
concludes that the evidence most strongly supports consideration of an alternative annual
standard level in the range of 12 to 11 |ig/m3. In conjunction with consideration of an
annual standard in the range of 12 to 11 |ig/m3, staff concludes it is appropriate to
consider retaining the current 24-hour PM2.5 standard level at 35 |ig/m3. In conjunction
with consideration of an annual standard level of 13 |ig/m3, staff concludes there is
limited support to consider revising the 24-hour PM2 5 standard level to somewhat below
35 |ig/m3, such as down to 30 |ig/m3.
In reaching these conclusions, staff recognizes that uncertainties and limitations
remain in the currently available evidence and quantitative risk estimates. We note that
no discernible thresholds have been identified for any health effects associated with long-
ES-1
-------
or short-term PM2.5 exposures. Therefore, a primary focus of our consideration of
alternative standard levels has been related to identifying the broader range of ambient
PM2.5 concentrations which has the most influence on generating health effect estimates
in the epidemiological studies. We also recognize that there remain uncertainties in our
understanding of the relationship between differences in the ambient PM2.5 mixtures
and/or exposure-related factors and the heterogeneity in health effects observed in the
epidemiological evidence. Staff concludes that there is insufficient information at this
time to consider supplementing the mass-based PM2.5 indicator by considering a separate
indicator for ultrafme particles or for a specific PM2.5 component or group of components
associated with any source categories of fine particles, or for eliminating any individual
component or group of components from the mix of fine particles included in the PM2.5
mass-based indicator.
Primary Standard for Thoracic Coarse Particles (Chapter 3):
In assessing the adequacy of the current primary 24-hour PMi0 standard, which is
meant to protect public health against short-term exposures to thoracic coarse particles
(i.e., PMio-2.s), staff concludes that it would be appropriate to consider either retaining or
revising the current standard, depending on the relative weight placed on the evidence
supporting associations with PMi0-2.5 and the uncertainties and limitations in this
evidence. Important uncertainties and limitations include those associated with the air
quality estimates used in PMi0-2.5 epidemiologic studies; the extent to which PMi0-2.5 air
quality concentrations reflect exposures to PMio-2.s; the extent to which PMio-2.5 itself is
responsible for health effects reported in epidemiologic studies; and the extent to which
the chemical and/or biological composition of PMio-2.5 affects particle toxicity.
To the extent consideration is given to revising the current standard, which has a
one-expected-exceedance form and a level of 150 |ig/m3, staff concludes that
consideration should be given to revising both the form and level. In this case,
consideration should be given to a 98th percentile form and a level within the range of 85
|ig/m3 down to about 65 |ig/m3; in conjunction with retaining the PMio indicator and the
24-hour averaging time. Staff also concludes that standard levels in the upper part of this
range are supported by the strongest evidence.
Secondary Standards for PM-related Visibility Impairment (Chapter 4):
In assessing the adequacy of the current suite of secondary annual and 24-hour
PM2.5 standards (which are identical to the primary PM2.5 standards) meant to protect
against PM-related visibility impairment, staff concludes that the currently available
information clearly calls into question the adequacy of the current standards and that
consideration should be given to revising the suite of standards to provide increased
public welfare protection. Staff also concludes that the current PM2.5 mass indicator is
not appropriate for a national standard intended to protect against PM-related visibility
impairment since such a standard is inherently confounded by regional differences in
relative humidity and species composition of PM2.5, which are critical factors in the
relationship between ambient particles and associated visibility impairment.
Taking into account both evidence-based and impact assessment-based
considerations, staff concludes that consideration should be given to establishing a new
ES-2
-------
standard defined in terms of a calculated PM2.5 light extinction indicator. Such an
indicator would use speciated PM2 5 mass and relative humidity data to calculate PM2 5
light extinction, similar to how light extinction is now calculated in the Regional Haze
Program. In conjunction with a calculated PM2 5 light extinction indicator, staff
concludes that consideration should be given to a 24-hour averaging time and a level of
28 deciviews (dv) or somewhat below, down to 25 dv. Staff concludes that it would also
be appropriate to consider a multi-hour, sub-daily averaging period (e.g., 4 hours) to the
extent that data quality issues that have recently been raised about data from continuous
PM2.5 monitors classified as Federal Equivalent Methods (FEMs) can be appropriately
addressed. In conjunction with consideration of a standard with a 4-hour averaging time,
staff concludes that consideration should be given to a level of 30 dv or somewhat below,
down to 25 dv. In all cases, staff concludes that consideration should be given to a 90th
percentile form, averaged over three years.
Secondary Standards for Non-visibility Welfare Effects (Chapter 5):
In assessing the adequacy of the current suite of secondary PM standards (which
are identical to the primary PM2 5 and PMi0 standards) meant to protect against PM-
related effects other than visibility impairment, staff has considered PM-related effects on
climate, ecological effects, and effects on materials. Staff concludes that the currently
available information supports retaining control of both fine and coarse particles to
address PM-related effects on ecosystems and materials damage and soiling, but that
there is insufficient information to assess the adequacy of protection afforded by the
current standards. Staff also concludes that there is insufficient information at this time
to base a NAAQS on climate impacts associated with current ambient concentrations of
PM or its constituents.
ES-3
-------
This page left intentionally blank.
-------
1 INTRODUCTION
1.1 PURPOSE
The U.S. Environmental Protection Agency (EPA) is presently conducting a review of
the national ambient air quality standards (NAAQS) for particulate matter (PM). The plan and
schedule for this review were presented in the Integrated Review Plan for the National Ambient
Air Quality Standards for Particulate Matter (IRP; US EPA, 2008a). The IRP identified key
policy-relevant issues to be addressed in this review as a series of questions that frame our
consideration of whether the current NAAQS for PM should be retained or revised.
This Policy Assessment (PA), prepared by staff in the EPA's Office of Air Quality
Planning and Standards (OAQPS), is intended to help "bridge the gap" between the relevant
scientific information and assessments and the judgments required of the EPA Administrator in
determining whether, and if so how, it is appropriate to revise the NAAQS for PM.1 This PA
presents factors relevant to EPA's review of the primary (health-based) and secondary (welfare-
based) PM NAAQS. It focuses on both evidence- and risk-based information in evaluating the
adequacy of the current PM NAAQS and in identifying potential alternative standards for
consideration. In so doing, we are seeking to provide as broad an array of options as is
supportable by the available information, recognizing that the selection of a specific approach to
reaching final decisions on the primary and secondary PM standards will reflect the judgments of
the Administrator.
In this PA, we consider the scientific and technical information available in this review as
assessed in the Integrated Science Assessment for Particulate Matter (Final Report) (ISA, US
EPA, 2009a), the Quantitative Health Risk Assessment for Particulate Matter (Final Report)
(RA, US EPA, 2010a) and the Particulate Matter Urban-Focused Visibility Assessment (Final
Report) (UFVA, US EPA, 201 Ob). In so doing, we focus on information that is most pertinent to
evaluating the basic elements of NAAQS: indicator2, averaging time, form,3 and level. These
elements, which together serve to define each standard, must be considered collectively in
evaluating the health and welfare protection afforded by the PM standards.
Although this PA should be of use to all parties interested in this PM NAAQS review, it
is written with an expectation that the reader has familiarity with the technical discussions
1 Preparation of a PA by OAQPS staff reflects Administrator Jackson's decision to modify the NAAQS review
process that was presented in the IRP. See http://www.epa.gov/ttn/naaqs/review.html for more information on the
current NAAQS review process.
2 The "indicator" of a standard defines the chemical species or mixture that is to be measured in determining
whether an area attains the standard.
3 The "form" of a standard defines the air quality statistic that is to be compared to the level of the standard in
determining whether an area attains the standard.
1-1
-------
contained in the ISA (US EPA, 2009a) and in the quantitative health risk and visibility
assessment documents (US EPA, 2010a,b).
1.2 BACKGROUND
1.2.1 Legislative Requirements
Two sections of the Clean Air Act (CAA) govern the establishment and revision of the
NAAQS. Section 108 (42 U.S.C. section 7408) directs the Administrator to identify and list
certain air pollutants and then to issue air quality criteria for those pollutants. The Administrator
is to list those air pollutants that in her "judgment, cause or contribute to air pollution which may
reasonably be anticipated to endanger public health or welfare;" "the presence of which in the
ambient air results from numerous or diverse mobile or stationary sources;" and "for which . . .
[the Administrator] plans to issue air quality criteria..." Air quality criteria are intended to
"accurately reflect the latest scientific knowledge useful in indicating the kind and extent of all
identifiable effects on public health or welfare which may be expected from the presence of [a]
pollutant in the ambient air . . ." 42 U.S.C. § 7408(b). Section 109 (42 U.S.C. 7409) directs the
Administrator to propose and promulgate "primary" and "secondary" NAAQS for pollutants for
which air quality criteria are issued. Section 109(b)(l) defines a primary standard as one "the
attainment and maintenance of which in the judgment of the Administrator, based on such
criteria and allowing an adequate margin of safety, are requisite to protect the public health." 4 A
secondary standard, as defined in section 109(b)(2), must "specify a level of air quality the
attainment and maintenance of which, in the judgment of the Administrator, based on such
criteria, is requisite to protect the public welfare from any known or anticipated adverse effects
associated with the presence of [the] pollutant in the ambient air."5
The requirement that primary standards provide an adequate margin of safety was
intended to address uncertainties associated with inconclusive scientific and technical
information available at the time of standard setting. It was also intended to provide a reasonable
degree of protection against hazards that research has not yet identified. See Lead Industries
Association v. EPA, 647 F.2d 1130, 1154 (D.C. Cir 1980), cert, denied, 449 U.S. 1042 (1980);
American Petroleum Institute v. Costle, 665 F.2d 1176, 1186 (D.C. Cir. 1981), cert, denied, 455
4 The legislative history of section 109 indicates that a primary standard is to be set at "the maximum permissible
ambient air level.. . which will protect the health of any [sensitive] group of the population," and that for this
purpose "reference should be made to a representative sample of persons comprising the sensitive group rather than
to a single person in such a group" S. Rep. No. 91-1196, 91st Cong., 2d Sess. 10 (1970).
5 Welfare effects as defined in section 302(h) (42 U.S.C. § 7602(h)) 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."
1-2
-------
U.S. 1034 (1982); American Farm Bureau Federation v. EPA, 559 F. 3d 512, 533 (D.C. Cir.
2009); Association of Battery Recyclers v. EPA, 604 F. 3d 613, 617-18 (D.C. Cir. 2010). Both
kinds of uncertainties are components of the risk associated with pollution at levels below those
at which human health effects can be said to occur with reasonable scientific certainty. Thus, in
selecting primary standards that provide an adequate margin of safety, the Administrator is
seeking not only to prevent pollution levels that have been demonstrated to be harmful but also
to prevent lower pollutant levels that may pose an unacceptable risk of harm, even if the risk is
not precisely identified as to nature or degree. The CAA does not require the Administrator to
establish a primary NAAQS at a zero-risk level or at background concentration levels, see Lead
Industries v. EPA, 647 F.2d at 1156 n.51, but rather at a level that reduces risk sufficiently so as
to protect public health with an adequate margin of safety.
In addressing the requirement for an adequate margin of safety, the EPA considers such
factors as the nature and severity of the health effects involved, the size of sensitive population(s)
at risk, and the kind and degree of the uncertainties that must be addressed. The selection of any
particular approach to providing an adequate margin of safety is a policy choice left specifically
to the Administrator's judgment. See Lead Industries Association v. EPA, 647 F.2d at 1161-62;
Whitman v. American Trucking Associations, 531 U.S. 457, 495 (2001).
In setting primary and secondary standards that are "requisite" to protect public health
and welfare, respectively, as provided in section 109(b), EPA's task is to establish standards that
are neither more nor less stringent than necessary for these purposes. In so doing, EPA may not
consider the costs of implementing the standards. See generally, Whitman v. American Trucking
Associations, 531 U.S. 457, 465-472, 475-76 (2001). Likewise, "[attainability and
technological feasibility are not relevant considerations in the promulgation of national ambient
air quality standards." American Petroleum Institute v. Costle, 665 F. 2d at 1185.
Section 109(d)(l) requires that "not later than December 31, 1980, and at 5-year intervals
thereafter, the Administrator shall complete a thorough review of the criteria published under
section 108 and the national ambient air quality standards . . . and shall make such revisions in
such criteria and standards and promulgate such new standards as may be appropriate
Section 109(d)(2) requires that an independent scientific review committee "shall complete a
review of the criteria . . . and the national primary and secondary ambient air quality standards ...
and shall recommend to the Administrator any new . . . standards and revisions of existing
criteria and standards as may be appropriate . . . ." Since the early 1980's, this independent
review function has been performed by the Clean Air Scientific Advisory Committee (CASAC).6
6 Lists of CASAC members and of members of the CASAC PM Review Panel are available at:
http://vosemite.epa.gov/sab/sabproduct.nsfAVebCASAC/CommitteesandMembership7OpenDocument.
1-3
-------
1.2.2 Previous PM NAAQS Reviews
The EPA initially established NAAQS for PM under section 109 of the CAA in 1971.
Since then, the Agency has made a number of changes to these standards to reflect continually
expanding scientific information, particularly with respect to the selection of indicator7 and level.
Table 1-1 provides a summary of the PM NAAQS that have been promulgated to date. These
decisions are briefly discussed below.
In 1971, EPA established NAAQS for PM based on the original air quality criteria
document (DHEW, 1969; 36 FR 8186, April 30, 1971). The reference method specified for
determining attainment of the original standards was the high-volume sampler, which collects
PM up to a nominal size of 25 to 45 micrometers (|im) (referred to as total suspended particles or
TSP). The primary standards (measured by the indicator TSP) were 260 |ig/m3, 24-hour average,
not to be exceeded more than once per year, and 75 |ig/m3, annual geometric mean. The
secondary standard was 150 |ig/m3, 24-hour average, not to be exceeded more than once per
year.
In October 1979, EPA announced the first periodic review of the criteria and NAAQS for
PM, and significant revisions to the original standards were promulgated in 1987 (52 FR 24634,
July 1, 1987). In that decision, EPA changed the indicator for PM from TSP to PMio, the latter
including particles with an aerodynamic diameter less than or equal to a nominal 10 |im, which
delineates thoracic particles (i.e., that subset of inhalable particles thought small enough to
penetrate beyond the larynx into the thoracic region of the respiratory tract). The EPA also
revised the primary standards by: (1) replacing the 24-hour TSP standard with a 24-hour PMio
standard of 150 |ig/m3 with no more than one expected exceedance per year; and (2) replacing
the annual TSP standard with a PMio standard of 50 |ig/m3, annual arithmetic mean. The
secondary standard was revised by replacing it with 24-hour and annual standards identical in all
respects to the primary standards. The revisions also included a new reference method for the
measurement of PMio in the ambient air and rules for determining attainment of the new
standards. On judicial review, the revised standards were upheld in all respects. Natural
Resources Defense Council v. EPA, 902 F. 2d 962 (D.C. Cir. 1990), cert, denied, 498 U.S. 1082
(1991).
7 Paniculate matter is the generic term for a broad class of chemically and physically diverse substances that exist as
discrete particles (liquid droplets or solids) over a wide range of sizes, such that the indicator for a PM NAAQS has
historically been defined in terms of particle size ranges.
1-4
-------
Table 1-1. Summary of National Ambient Air Quality Standards Promulgated for
Particulate Matter 1971-20068
Final Rule
1971
36 FR 8186
April 30, 1971
1987
52 FR 24634
July 1, 1987
1997
62 FR 3 8652
July 18, 1997
2006
71 FR 61 144
October 17, 2006
Indicator
TSP
PMio
PM2.5
PM10
PM2.5
PMio
Averaging
Time
24-hour
Annual
24-hour
Annual
24-hour
Annual
24-hour
Annual
24-hour
Annual
24-hour
Level
260 |ig/m3
(primary)
150 |ig/m3
(secondary)
75 |ig/m3
(primary)
150 |ig/m3
50 |ig/m3
65 |ig/m3
15 |ig/m3
150 |ig/m3
50 |ig/m3
35 |ig/m3
15 |ig/m3
150 |ig/m3
Form
Not to be exceeded more than once per year
Annual average
Not to be exceeded more than once per year
on average over a 3 -year period
Annual arithmetic mean, averaged over 3
years
98th percentile, averaged over 3 years9
Annual arithmetic mean, averaged over 3
years10,11
Initially promulgated 99th percentile,
averaged over 3 years; when 1997 standards
were vacated, the form of 1987 standards
remained in place (not to be exceeded more
than once per year on average over a 3 -year
period)
Annual arithmetic mean, averaged over 3
years
98th percentile, averaged over 3 years9
Annual arithmetic mean, averaged over 3
yearsiu2
Not to be exceeded more than once per year
on average over a 3 -year period
8 When not specified, primary and secondary standards are identical.
9 The 24-hour standard NAAQS metric is defined as an integer (zero decimal places) as determined by rounding.
For example, a 3-year average 98th percentile concentration of 35.49 ug/m3would round to 35 ug/m3and thus meet
the 24-hour standard and a 3-year average of 35.50 ug/m3would round to 36 ug/m3 and hence, violate the 24-hour
standard (40 CFR part 50 Appendix N).
10 The annual standard NAAQS metric is defined to one decimal place (i.e., 15.0 ug/m3) as determined by rounding.
For example, a 3-year average annual mean of 15.04 ug/m3 would round to 15.0 ug/m3 and thus meet the annual
standard and a 3-year average of 15.05 ug/m3 would round to 15.1 ug/m3 and hence, violate the annual standard.
11 The level of the standard was to be compared to measurements made at sites that represent "community-wide air
quality" recording the highest level, or, if specific constraints were met, measurements from multiple community-
wide air quality monitoring sites could be averaged ("spatial averaging").
12 The constraints on the spatial averaging criteria were tightened by further limiting the conditions under which
some areas may average measurements from multiple community-oriented monitors to determine compliance (see
71FR 61165-61167).
1-5
-------
In April 1994, EPA announced its plans for the second periodic review of the criteria and
NAAQS for PM, and promulgated significant revisions to the NAAQS in 1997 (62 FR 38652,
July 18, 1997). Most significantly, EPA determined that although the PM NAAQS should
continue to focus on thoracic particles (PMio), the fine and coarse fractions of PMio should be
considered separately. New standards were added, using PM2.5 as the indicator for fine
particles.13 The PMio standards were retained for the purpose of regulating the coarse fraction of
PMio (referred to as thoracic coarse particles or PMio-2.s).14 The EPA established two new PM2.s
standards: an annual standard of 15 |ig/m3, based on the 3-year average of annual arithmetic
mean PM2.5 concentrations from single or multiple monitors sited to represent community-wide
air quality15; and a 24-hour standard of 65 |ig/m3, based on the 3-year average of the 98th
percentile of 24-hour PM2 5 concentrations at each population-oriented monitor16 within an area.
Also, EPA established a new reference method for the measurement of PM2.5 in the ambient air
and rules for determining attainment of the new standards. To continue to address thoracic
coarse particles, the annual PMio standard was retained, while the form, but not the level, of the
24-hour PMio standard was revised to be based on the 99th percentile of 24-hour PMio
concentrations at each monitor in an area. The EPA revised the secondary standards by making
them identical in all respects to the primary standards.
Following promulgation of the revised PM NAAQS in 1997, petitions for review were
filed by a large number of parties, addressing a broad range of issues. In May 1998, a three-
judge panel of the U.S. Court of Appeals for the District of Columbia Circuit issued an initial
decision that upheld EPA's decision to establish fine particle standards, holding that "the
growing empirical evidence demonstrating a relationship between fine particle pollution and
adverse health effects amply justifies establishment of new fine particle standards." American
Trucking Associations v. EPA , 175 F. 3d 1027, 1055-56 (D.C. Cir. 1999), rehearing granted in
part and denied in part, 195 F. 3d 4 (D.C. Cir. 1999), affirmed in part and reversed in part,
Whitman v. American Trucking Associations, 531 U.S. 457 (2001). The panel also found "ample
support" for EPA's decision to regulate coarse particle pollution, but vacated the 1997 PMio
standards, concluding, in part, that PMio is a "poorly matched indicator for coarse particulate
13 PM25 includes particles with an aerodynamic diameter less than or equal to a nominal 2.5 um.
14 See 40 CFR Parts 50, 53, and 58 for more information on reference and equivalent methods for measuring PM in
ambient air.
15 Monitoring stations sited to represent community-wide air quality will typically be at the neighborhood or urban-
scale; however, where a population-oriented micro or middle-scale PM2 5 monitoring station represents many such
locations throughout a metropolitan area, these smaller scales may also be considered to represent community-wide
air quality (40 CFR Part 58, Appendix D, 4.7. l(b)).
16 Population-oriented monitoring (or sites) means residential areas, commercial areas, recreational areas, industrial
areas where workers from more than one company are located, and other areas where a substantial number of people
may spend a significant fraction of their day. (40 CFR Part 58, §58.1)
1-6
-------
pollution" because it includes fine particles. Id. at 1053-55. Pursuant to the court's decision,
EPA removed the vacated 1997 PMio standards from the Code of Federal Regulations (CFR) (69
FR 45592, July 30, 2004) and deleted the regulatory provision [at 40 CFR section 50.6(d)] that
controlled the transition from the pre-existing 1987 PMio standards to the 1997 PMio standards.
The pre-existing 1987 PMio standards remained in place (65 FR 80776, December 22, 2000).
The court also upheld EPA's determination not to establish more stringent secondary standards
for fine particles to address effects on visibility (175 F. 3d at 1027).
More generally, the panel held (over a strong dissent) that EPA's approach to establishing
the level of the standards in 1997, both for the PM and for the ozone (63) NAAQS promulgated
on the same day, effected "an unconstitutional delegation of legislative authority." Id. at 1034-
40. Although the panel stated that "the factors EPA uses in determining the degree of public
health concern associated with different levels of ozone and PM are reasonable," it remanded the
rule to EPA, stating that when EPA considers these factors for potential non-threshold pollutants
"what EPA lacks is any determinate criterion for drawing lines" to determine where the
standards should be set. Consistent with EPA's long-standing interpretation and D.C. Circuit
precedent, the panel also reaffirmed its prior holdings that in setting NAAQS EPA is "not
permitted to consider the cost of implementing those standards" Id. at 1040-41.
On EPA's petition for rehearing, the panel adhered to its position on these points.
American Trucking Associations v. EPA, 195 F. 3d 4 (D.C. Cir. 1999). The full Court of
Appeals denied EPA's request for rehearing en bane, with five judges dissenting. Id. at 13. Both
sides filed cross appeals on these issues to the United States Supreme Court, which granted
certiorari. In February 2001, the Supreme Court issued a unanimous decision upholding EPA's
position on both the constitutional and cost issues. Whitman v. American Trucking Associations.,
531 U.S. 457, 464, 475-76. On the constitutional issue, the Court held that the statutory
requirement that NAAQS be "requisite" to protect public health with an adequate margin of
safety sufficiently cabined EPA's discretion, affirming EPA's approach of setting standards that
are neither more nor less stringent than necessary. The Supreme Court remanded the case to the
Court of Appeals for resolution of any remaining issues that had not been addressed in that
court's earlier rulings. Id. at 475-76. In March 2002, the Court of Appeals rejected all
remaining challenges to the standards, holding under the traditional standard of review that
EPA's PM2.5 standards were reasonably supported by the administrative record and were not
"arbitrary and capricious." American Trucking Associations v. EPA, 283 F. 3d 355, 369-72
(D.C. Cir. 2002).
In October 1997, EPA published its plans for the next periodic review of the air quality
criteria and NAAQS for PM (62 FR 55201, October 23, 1997). After CASAC and public review
of several drafts, EPA's National Center for Environmental Assessment (NCEA) finalized the
1-7
-------
Air Quality Criteria Document for P articulate Matter (henceforth, AQCD or the "Criteria
Document") in October 2004 (U.S. EPA, 2004) and OAQPS finalized an assessment document,
Paniculate Matter Health Risk Assessment for Selected Urban Areas (Abt Associates, 2005),
and a "Staff Paper," Review of the National Ambient Air Quality Standards for P articulate
Matter: Policy Assessment of Scientific and Technical Information, in December 2005 (U.S.
EPA, 2005). In conjunction with its review of the Staff Paper, CASAC provided advice to the
Administrator on revisions to the PM NAAQS (Henderson, 2005a). In particular, most CASAC
PM Panel members favored revising the level of the primary 24-hour PM2.5 standard in the range
of 35 to 30 |ig/m3 with a 98th percentile form, in concert with revising the level of the primary
annual PM2.5 standard in the range of 14 to 13 |ig/m3 (Henderson, 2005a, p.7). For thoracic
coarse particles, the Panel had reservations in recommending a primary 24-hour PMi0-2.5
standard, and agreed that there was a need for more research on the health effects of thoracic
coarse particles (Henderson, 2005b). With regard to secondary standards, most Panel members
strongly supported establishing a new, distinct secondary PM2.5 standard to protect urban
visibility (Henderson, 2005a, p. 9).
On January 17, 2006, EPA proposed to revise the primary and secondary NAAQS for PM
(71 FR 2620) and solicited comment on a broad range of options. Proposed revisions included:
revising the level of the primary 24-hour PM2.5 standard to 35 |ig/m3; revising the form, but not
the level, of the primary annual PM2 5 standard by tightening the constraints on the use of spatial
averaging; replacing the primary 24-hour PMio standard with a 24-hour standard defined in
terms of a new indicator, PMio-2.517 set at a level of 70 |ig/m3 based on the 3-year average of the
98th percentile of 24-hour PMi0-2.5 concentrations; revoking the primary annual PMio standard;
and revising the secondary standards by making them identical in all respects to the proposed
suite of primary standards for fine and coarse particles.18 Subsequent to the proposal, CASAC
provided additional advice to EPA in a letter to the Administrator requesting reconsideration of
CASAC's recommendations for both the primary and secondary PM2.5 standards as well as the
standards for thoracic coarse particles (Henderson, 2006a).
On October 17, 2006, EPA promulgated revisions to the PM NAAQS to provide
increased protection of public health and welfare (71 FR 61144). With regard to the primary and
secondary standards for fine particles, EPA revised the level of the primary 24-hour PM2.s
17 This proposed indicator was qualified so as to include any ambient mix of PM10.25 dominated by particles
generated by high-density traffic on paved roads, industrial sources, and construction sources, and to exclude any
ambient mix of particles dominated by rural windblown dust and soils and agricultural and mining sources (71 FR
2667 to 2668).
18 In recognition of an alternative view expressed by most members of the CASAC PM Panel, the Agency also
solicited comments on a subdaily (4- to 8-hour averaging time) secondary PM2 5 standard to address visibility
impairment, considering alternative standard levels within a range of 20 to 30 ug/m3 in conjunction with a form
within a range of the 92nd to 98th percentile (71 FR 2685).
1-8
-------
standard to 35 |ig/m3, retained the level of the primary annual PM2.5 standard at 15 |ig/m3, and
revised the form of the primary annual PM2.5 standard by adding further constraints on the
optional use of spatial averaging. The EPA revised the secondary standards for fine particles by
making them identical in all respects to the primary standards. With regard to the primary and
secondary standards for thoracic coarse particles, EPA retained the level and form of the 24-hour
PMio standard (such that the standard remained at a level of 150 |ig/m3 with a one-expected
exceedance form), and revoked the annual PMi0 standard. The EPA also established a new
Federal Reference Method (FRM) for the measurement of PMio-2.5 in the ambient air (71 FR
61212-13). Although the standards for thoracic coarse particles were not defined in terms of a
PMio-2.5 indicator, the new FRM for PMio-2.5 was established to provide a basis for approving
Federal Equivalent Methods (FEMs) and to promote gathering scientific data to support future
reviews of the PM NAAQS.
Following issuance of the final rule, CAS AC articulated its concern that "EPA's final
rule on the NAAQS for PM does not reflect several important aspects of the CASAC's advice"
(Henderson et al., 2006b). With regard to the primary PM2.s annual standard, CASAC expressed
serious concerns regarding the decision to retain the level of the standard at 15 |ig/m3. With
regard to EPA's final decision to retain the 24-hour PMio standard for thoracic coarse particles,
CASAC acknowledged concerns associated with retaining this standard while recognizing the
need to have a standard in place to protect against effects associated with short-term exposures to
thoracic coarse particles. With regard to EPA's final decision to revise the secondary PM2.s
standards to be identical in all respects to the revised primary PM2.5 standards, CASAC
expressed concerns that its advice to establish a distinct secondary standard for fine particles to
address visibility impairment was not followed.
1.2.3 Litigation Related to the 2006 PM Standards
Several parties filed petitions for review following promulgation of the revised PM
NAAQS in 2006. These petitions addressed the following issues: (1) selecting the level of the
primary annual PM2.5 standard; (2) retaining PMio as the indicator of a standard for thoracic
coarse particles, retaining the level and form of the 24-hour PMio standard, and revoking the
PMio annual standard; and (3) setting the secondary PM2.5 standards identical to the primary
standards. On February 24, 2009, the U.S. Court of Appeals for the District of Columbia Circuit
issued its opinion in the case American Farm Bureau Federation v. EPA, 559 F. 3d 512 (D.C.
Cir. 2009). The court remanded the primary annual PM2.s NAAQS to EPA because EPA failed
to adequately explain why the standard provided the requisite protection from both short- and
long-term exposures to fine particles, including protection for at-risk populations. American
Farm Bureau Federation v. EPA, 559 F. 3d 512, 520-27 (D.C. Cir. 2009). With regard to the
1-9
-------
standards for PMio, the court upheld EPA's decisions to retain the 24-hour PMio standard to
provide protection from thoracic coarse particle exposures and to revoke the annual PMio
standard. American Farm Bureau Federation, 559 F. 2d at 533-38. With regard to the
secondary PM2.5 standards, the court remanded the standards to EPA because the Agency failed
to adequately explain why setting the secondary PM standards identical to the primary standards
provided the required protection for public welfare, including protection from visibility
impairment. American Farm Bureau Federation, 559 F. 2d at 528-32.
The decisions of the court with regard to these three issues are discussed further in
chapters 2, 3 and 4, respectively (see sections 2.1.2, 3.1.2, and 4.1.2). The EPA is responding to
the court's remands as part of the current review of the PM NAAQS.
1.2.4 Current PM NAAQS Review
The EPA initiated the current review of the air quality criteria for PM in June 2007 with a
general call for information (72 FR 35462, June 28, 2007). In July 2007, EPA held two "kick-
off workshops on the primary and secondary PM NAAQS, respectively (72 FR 34003 and
34004, June 20, 2007).19 These workshops provided an opportunity for a public discussion of
the key policy-relevant issues around which EPA would structure this PM NAAQS review and
the most meaningful new science that would be available to inform our understanding of these
issues.
Based in part on the workshop discussions, EPA developed a draft IRP outlining the
schedule, process, and key policy-relevant questions that would guide the evaluation of the air
quality criteria for PM and the review of the primary and secondary PM NAAQS (US EPA,
2007). On November 30, 2007, EPA held a consultation with CASAC on the draft IRP (72 FR
63177, November 8, 2007), which included the opportunity for public comment. The final IRP
(US EPA, 2008a) incorporated comments from CASAC (Henderson, 2008) and the public on the
draft plan as well as input from senior Agency managers.20
As part of the process of preparing the PM ISA, NCEA hosted a peer review workshop in
June 2008 on preliminary drafts of key ISA chapters (73 FR 30391, May 27, 2008). The first
external review draft ISA (US EPA, 2008b; 73 FR 77686, December 19, 2008) was reviewed by
19 See workshop materials available at: http://www.regulations.gov/search/Regs/home.htmMhome Docket ID
numbers EPA-HQ-OAR-2007-0492-008; EPA-HQ-OAR-2007-0492-009; EPA-HQ-OAR-2007-0492-010; and
EPA-HQ-OAR-2007-0492-012.
20 The process followed in this review varies from the NAAQS review process described in section 1.1 of the IRP
(US EPA, 2008a). On May 21, 2009, EPA Administrator Jackson called for key changes to the NAAQS review
process including reinstating a policy assessment document that contains staff analyses of the scientific bases for
alternative policy options for consideration by senior Agency management prior to rulemaking. In conjunction with
this change, EPA will no longer issue a policy assessment in the form of an advance notice of proposed rulemaking
(ANPR) as discussed in the IRP. For more information on the overall process followed in this review including a
description of the major elements of the process for reviewing NAAQS see Jackson (2009).
1-10
-------
CASAC and the public at a meeting held in April 2009 (74 FR 2688, February 19, 2009). Based
on CASAC and public comments, NCEA prepared a second draft ISA (US EPA, 2009b; 74 FR
38185, July 31, 2009), which was reviewed by CASAC and the public at a meeting held on
October 5-6, 2009 (74 FR 46586, September 10, 2009). Based on CASAC and public
comments, NCEA prepared the final ISA (US EPA, 2009a; 74 FR 66353, December 15, 2009).
In preparing the Risk and Exposure Assessment (REA) documents that build on the
scientific evidence presented in the ISA, OAQPS released two planning documents: Particulate
Matter National Ambient Air Quality Standards: Scope and Methods Plan for Health Risk and
Exposure Assessment and Particulate Matter National Ambient Air Quality Standards: Scope
and Methods Plan for Urban Visibility Impact Assessment (henceforth, Scope and Methods
Plans, US EPA, 2009c,d; 74 FR 11580, March 18, 2009). These planning documents outlined
the scope and approaches that staff planned to use in conducting quantitative assessments as well
as key issues that would be addressed as part of the assessments. In designing and conducting
the initial health risk and visibility impact assessments, we considered CASAC comments
(Samet 2009a,b) on the Scope and Methods Plans made during an April 2009 consultation (74
FR 7688, February 19, 2009) as well as public comments. Two draft assessment documents,
Risk Assessment to Support the Review of the PM2.s Primary National Ambient Air Quality
Standards: External Review Draft - September 2009 (US EPA 2009e) and Paniculate Matter
Urban-Focused Visibility Assessment - External Review Draft - September 2009 (US EPA,
2009f) were reviewed by CASAC and the public at a meeting held on October 5 - 6, 2009 (74 FR
46586, September 10, 2009). Based on CASAC (Samet 2009c,d) and public comments, OAQPS
staff revised these draft documents and released second draft assessment documents (US EPA,
2010d,e) in January and February 2010 (75 FR 4067, January 26, 2010) for CASAC and public
review at a meeting held on March 10-11, 2010 (75 FR 8062, February 23, 2010). Based on
CASAC (Samet, 2010a,b) and public comments on the second draft assessment documents, we
revised these documents and released final assessment documents in June and July 2010 (US
EPA, 2010a,b; 75 FR 39252, July 8, 2010).
A preliminary draft PA (US EPA, 2009g) was released in September 2009 for
informational purposes and to facilitate discussion with CASAC at the October 5-6, 2009
meeting on the overall structure, areas of focus, and level of detail to be included in the PA.
CASAC's comments on the preliminary draft PA encouraged the development of a document
focused on the key policy-relevant issues that draws from and is not repetitive of information in
the ISA and REAs. These comments were considered in developing a first draft PA (US EPA,
2010c; 75 FR 4067, January 26, 2010) that built upon the information presented and assessed in
the final ISA and second draft risk and visibility assessment documents. The EPA presented an
overview of the first draft PA at a CASAC meeting on March 10, 2010 (75 FR 8062, February
1-11
-------
23, 2010). The first draft PA was reviewed by CASAC and the public and discussed during
public teleconferences on April 8 - 9, 2010 (75 FR 8062, February 23, 2010) and May 7, 2010
(75 FR 19971, April 16, 2010).
The CASAC (Samet, 2010c) and public comments on the first draft PA were considered
by EPA staff in developing a second draft PA (US EPA, 2010f; 75 FR 39253, July 8, 2010)
which was reviewed by CASAC at a meeting on July 26 - 27, 2010 (75 FR 32763, June 9, 2010).
CASAC (Samet, 2010d, see Appendix A) and public comments on the second draft PA were
considered by EPA staff in preparing this final PA.21 This document includes final staff
conclusions related to the adequacy of the current PM standards and the broadest range of
alternative standards that are supported by the currently available scientific evidence and
quantitative assessments.
1.3 CURRENT AMBIENT MONITORING NETWORKS FOR PM
In the U.S., state and local agencies operate the vast majority of PM monitors as part of
the State and local air monitoring stations (SLAMS) network. The SLAMS network supports
three major objectives: to provide air pollution data to the general public in a timely manner; to
support compliance with ambient air quality standards and emissions strategy development; and
to support air pollution research studies. PM monitors are deployed according to network design
criteria described in Appendix D to 40 CFR Part 58. For comparison to the annual standard,
PM2.s monitoring sites are required to represent community-wide air quality. For most urban
locations this will result in siting monitors at the neighborhood scale22 where PM2.5
concentrations are reasonably homogeneous throughout an entire urban sub-region. At least one
monitoring station representing community-wide air quality is also to be sited in a population-
oriented area of expected maximum concentration. Sites that represent relatively unique
population-oriented microscale,23 or localized hot-spot, or unique population-oriented middle
21 All written comments submitted to the Agency are available in the docket for this PM NAAQS review (EPA-HQ-
OAR-2007-0429). Transcripts of public meetings and teleconferences held in conjunction with CASAC's reviews
are also included in the docket.
22 Neighborhood scale for PM2.S: Measurements in this category would represent conditions throughout some
reasonably homogeneous urban sub-region with dimensions of a few kilometers and of generally more regular shape
than the middle scale. Homogeneity refers to the paniculate matter concentrations, as well as the land use and land
surface characteristics. Much of the PM2 5 exposures are expected to be associated with this scale of measurement.
In some cases, a location carefully chosen to provide neighborhood scale data would represent the immediate
neighborhood as well as neighborhoods of the same type in other parts of the city. See 40 CRF Part 58 Appendix D,
23 Microscale for PM2.5: This scale would typify areas such as downtown street canyons and traffic corridors where
the general public would be exposed to maximum concentrations from mobile sources. See 40 CFR Part 58
Appendix D, 4.7. l(c)(l).
1-12
-------
scale24 impact sites are only eligible for comparison to the 24-hour PM2.5 NAAQS. For PMio,
the network design criteria emphasize monitoring at middle25 and neighborhood26 scales to
effectively characterize the emissions from both mobile and stationary sources, although not
ruling out microscale monitoring in some instances (40 CFR Part 58 Appendix D, 4.6 (b)).
The sections below briefly summarize the monitoring networks for PM including PM2.5
mass, PM2.5 speciation, and PMio mass, as well as the new PMio-2.s mass network which will be
part of the forthcoming National Core Multipollutant Monitoring Network (NCore).27
Additional information and maps of monitoring sites are provided in Appendix B.
1.3.1 PM2.5 Mass
The PM2.5 monitoring requirements provide for monitors in Metropolitan Statistical
Areas28 (MSAs) based on a combination of population and design value29 (see Table D-5, 40
CFR Part 58) with higher populated locations having more polluted air required to have the most
monitors. Background and transport monitors30 are also required of each state with options for
utilizing IMPROVE31 and other PM2 5 data to provide for flexibility in meeting the monitoring
requirements.
24 Middle scale for PM2.S: People moving through downtown areas, or living near major roadways, encounter
particle concentrations that would be adequately characterized by this spatial scale. Thus, measurements of this type
would be appropriate for the evaluation of possible effects associated with short-term exposures to PM25. See 40
CFR Part 58 Appendix D, 4.7. l(c)(2).
25 Middle scale for PM10: Much of the short-term public exposure to coarse fraction particles (using PM10 as the
indicator) is characterized on this scale and on the neighborhood scale. People moving through downtown areas or
living near major roadways or stationary sources, may encounter paniculate matter that would be adequately
characterized by measurements of this spatial scale. See 40 CFR Part 58 Appendix D, 4.6(b)(2).
26 Neighborhood scale for PM10: Measurements in this category represent conditions throughout some reasonably
homogeneous urban subregion with dimensions of a few kilometers and of generally more regular shape than the
middle scale. Homogeneity refers to the paniculate matter concentrations, as well as the land use and land surface
characteristics. In some cases, a location carefully chosen to provide neighborhood scale data would represent not
only the immediate neighborhood but also neighborhoods of the same type in other parts of the city. See 40 CFR
Part 58 Appendix D, 4.6(b)(3).
27 The NCore network is a multi-pollutant network that includes measurements of particles, gases, and meteorology
(71 FR 61236, October 17, 2006). The network is intended to support integrated air program management needs.
The NCore monitoring network is expected to be fully operational by January 1, 2011.
28 Metropolitan and micropolitan statistical areas (metro and micro areas) are geographic entities defined by the U.S.
Office of Management and Budget (OMB) for use by Federal statistical agencies in collecting, tabulating, and
publishing Federal statistics. The term "Core Based Statistical Area" (CBSA) is a collective term for both metro and
micro areas. A metro area contains a core urban area of 50,000 or more population, and a micro area contains an
urban core of at least 10,000 (but less than 50,000) population. Each metro or micro area consists of one or more
counties and includes the counties containing the core urban area, as well as any adjacent counties that have a high
degree of social and economic integration (as measured by commuting to work) with the urban core.
29 Design values are the metrics (i.e., statistics) that are compared to the NAAQS levels to determine compliance.
For example, for PM2 5, design values are calculated as shown in section 4.0 of Appendix N to 40 CFR Part 50.
30 Background and transport sites are used to provide regional background and transport data. When evaluated in
combination with urban sites, these locations help determine the local urban contribution to PM.
31 The Interagency Monitoring of Protected Visual Environment (IMPROVE) program was established in 1985 to
aid the creation of federal and state implementation plans for the protection of visibility in Class 1 areas (i.e., 155
1-13
-------
The network of PM2.5 FRM monitors includes over 900 monitoring stations throughout
the country. The FRM is a manually operated sampler that is programmed to operate for a 24-
hour period covering midnight to midnight local standard time. Although approximately 150
FRM sites operate every day, most operate every third or sixth day according to a national
schedule provided by EPA.32 The network of PM2.5 continuous monitors has grown to
approximately 800 locations throughout the country of which about 115 locations have Class III
FEMs33. The number of PM2 5 FRM monitors may decrease over the coming years as PM2 5
continuous FEMs replace FRMs.
1.3.2 PM2.5 Speciation
As part of the PM2.5 NAAQS review completed in 1997, EPA established a PM2.5
Chemical Speciation Network (CSN) to conduct routine speciation monitoring in primarily urban
areas in the U.S. The PM2.5 CSN consists of about 50 Speciation Trends Network (STN) sites
and about 150 SLAMS supplemental sites. All STN sites operate on a one-in-three day sample
collection schedule. A majority of the SLAMS supplemental sites operate on a one-in-six day
sample collection schedule. These sites collect aerosol samples over 24 hours on filters that are
analyzed for PM2 5 mass, a number of trace elements, major ions (e.g., sulfate, nitrate,
ammonium), and organic and elemental carbon. Similar to the CSN, the IMPROVE program34
also provides PM2.5 mass and speciation data for organic and elemental carbon, major ions, and
trace elements; however, unlike CSN, IMPROVE also samples for PMio mass.
1.3.3 PM10 Mass
monitoring stations have an urban focus and are required in MSAs according to
Table D-4 of Appendix D to 40 CFR Part 58. Local considerations are a factor in determining
the actual required number of monitoring sites. More stations are required in larger MSAs and
areas with more evidence of poor air quality, while monitors are also required in "clean" MSAs
of certain size. The network currently includes over 800 monitoring stations throughout the
country with most metropolitan areas operating more PMio monitors than the minimum required
national parks and wilderness areas) as stipulated in the 1977 amendments to the CAA. The IMPROVE Network
also supports goals set forth in the 1999 Regional Haze Rule (64 FR 35714. July 1, 1999). Additional information is
available at http://www.epa.gov/visibilitv/program.html.
32 The national sampling schedule calendar is available on the web at: http://www.epa.gov/ttn/amtic/calendar.html.
33 Class III equivalent method means an equivalent method for PM2 5 that is an analyzer capable of providing
ambient air measurements representative of one hour or less integrated concentrations as well as 24-hour
measurements determined as, or equivalent to, the mean of 24 one-hour consecutive measurements.
34 IMPROVE is a cooperative measurement effort managed by a steering committee composed of representatives
from federal, regional, and state organizations. See Appendix B and http://vista.cira.colostate.edu/improve/ for more
information.
1-14
-------
by current monitoring rules. Many PMio monitoring stations operate FRMs on a mix of daily,
one-in-two day, one-in-three day, or one-in-six day sampling, based on the relative
concentrations measured at a specific monitoring site with respect to the 24-hour standard.
There are also FEMs that are operated continuously, providing hourly PMio measurements.
1.3.4 PMio-2.5 Mass
Ambient measurements of PMio-2.5 concentrations are not routinely measured and
reported at present (US EPA, 2009a, section 3.5.1.1). The EPA has required PMio-2.5 mass
monitoring as part of the NCore network which began January 1, 2011 at approximately 80
stations. Urban NCore stations are to be generally located at an urban scale35 or neighborhood
scale to provide representative concentrations of exposure expected throughout the metropolitan
area. Rural NCore stations are to be located, to the maximum extent practicable, at a regional36
or larger scale, away from any large local emission source, so that they represent ambient
concentrations over an extensive area.
1.4 GENERAL APPROACH AND ORGANIZATION OF THIS DOCUMENT
This PA includes staffs evaluation of the policy implications of the scientific assessment
of the evidence presented and assessed in the ISA and the results of quantitative assessments
based on that information presented and assessed in the REAs. Taken together, this information
informs staff conclusions and the identification of policy options for consideration in addressing
public health and welfare effects associated with exposure to ambient PM.
Since the last review, much new information is now available on PM air quality and
human health effects directly in terms of PM2.5 and, to a much more limited degree, PMio-2.5 and
ultrafme particles (UFPs).37 Since the purpose of this review is to evaluate the adequacy of the
current standards, which separately address fine and thoracic coarse particles, staff is focusing
this policy assessment and associated quantitative analyses primarily on the evidence related
directly to PM2.5 and PMio-2.5. In so doing, we are considering PMio-related evidence primarily
to help inform our understanding of key issues and to help interpret and provide context for
understanding the public health and welfare impacts of ambient fine and coarse particles. We are
also considering the currently available evidence related to UFPs as well as PM2.5 components to
35 Urban scale: This class of measurement would be used to characterize PM concentrations over an entire
metropolitan or rural area ranging in size from 4 to 50 kilometers. Such measurements would be useful for
assessing trends in area-wide air quality, and hence, the effectiveness of large scale air pollution control strategies.
Community-oriented PM25 sites may have this scale (see 40 CFR Part 58, Appendix D, section 4.7.1(c)(4)).
36 Regional scale: These measurements would characterize conditions over areas with dimensions as much as
hundreds of kilometers. Such measurements provide information about PM emissions and atmospheric losses and
transport (see 40 CFR Part 58, Appendix D, section 4.7.1(c)(5).
37 Ultrafine particles generally include particles with a nominal serodynamic diameter less than or equal to 0.1 um.
1-15
-------
aid in considering whether there is support to consider standards with a different size fraction
and/or distinct standards focused on regulating a specific PM2.5 component or group of
components associated with any source categories of fine particles.
Following this introductory chapter, this document is organized into two main parts:
review of the primary PM NAAQS (chapters 2 and 3) and review of the secondary PM NAAQS
(chapters 4 and 5). Chapters 2 and 3 present staff observations and conclusions related to review
of the primary standards for fine and thoracic coarse particles, respectively. Each chapter
includes background information on the rationale for previous reviews and the policy assessment
approaches followed in the current review, focusing on evidence-based considerations and, as
appropriate, quantitative risk-based considerations. Staff conclusions are presented with regard
to the adequacy of the current primary standards and potential alternative primary standards for
consideration, in terms of indicators, averaging times, forms, and levels. Chapter 4 focuses on
PM-related visibility impairment, and presents staff observations and conclusions with regard to
the adequacy of the current standards and potential distinct secondary standards for
consideration, in terms of alternative indicators, averaging times, forms, and levels. Chapter 5
focuses on other PM-related welfare effects, including effects on climate, ecological effects, and
effects on materials, and presents staff observations and conclusions with regard to the adequacy
of the current standards and the extent to which information is available to support consideration
of alternative standards.
1-16
-------
1.5 REFERENCES
Abt Associates (2005). Paniculate Matter Health Risk Assessment for Selected Urban Areas. Final Report.
Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Contract No. 68-D-03-002. EPA 452/R-05-007A. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/PMrisk20051220.pdf
Henderson R (2005a). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to
Honorable Stephen L. Johnson, Administrator, US EPA. CAS AC PM Review Panel's Peer Review of the
Agency's Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment
of Scientific and Technical Information (Second Draft PM Staff Paper, January 2005). June 6, 2005.
Available at: http://www.epa. gov/sab/pdf/casac-05-007.pdf.
Henderson, R (2005b). Clean Air Scientific Advisory Committee (CASAC) Review of the EPA Staff
Recommendations Concerning a Potential Thoracic Coarse PM Standard in the Review of the National
Ambient Air Quality Standards for Particulate Matter: Policy Assessment of Scientific and Technical
Information (Final PM OAQPS Staff Paper, EPA-452/R-05-005, June 2005). September 15, 2005.
Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/3562FF25F05133FC85257084000B lB77/$File/sab-casac-05-
012.pdf.
Henderson R (2006a). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Recommendations Concerning the Proposed National Ambient Air Quality Standards for Particulate
Matter. March 21, 2006. Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/CD706C976D AC62B3852571390081CC21/$File/casac-ltr-06-
002.pdf.
Henderson R, Cowling E, Crapo JD, Miller FJ, Poirot RL, Speizer F, Zielinska B (2006b). Letter from Clean Air
Scientific Advisory Committee to the Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air
Scientific Advisory Committee Recommendations Concerning the Final National Ambient Air Quality
Standards for Particulate Matter. September 29, 2006. Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/lC69E987731CB775852571FC00499A10/$File/casac-ltr-06-
003.pdf.
Henderson R (2008). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Particulate Matter Review Panel's Consultation on EPA's Draft Integrated Review Plan for the National
Ambient Air Quality Standards for Particulate Matter. January 3, 2008. Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/76D069B8191381DA8
52573 C500688E74/$File/EP A-CASAC-08-004-unsigned.pdf
Jackson L (2009). Memo from Administrator Lisa P. Jackson to Elizabeth Craig, Acting Assistant Administrator for
OAR and Lek Kadeli, Acting Assistant Administrator for ORD. Process for Reviewing the National
Ambient Air Quality Standards. May 21, 2009. Available at:
http://www.epa.gov/ttn/naaqs/pdfs/NAAQSReviewProcessMemo52109.pdf.
Samet J (2009a). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. Consultation on EPA's Particulate Matter National
Ambient Air Quality Standards: Scope and Methods Plan for Health Risk and Exposure Assessment. May
21,2009. Available at:
http://vosemite.epa. gov/sab/sabproduct.nsf/723FE644C5D758DF852575BD00763A32/$File/EPA-
CASAC-09-009-unsigned.pdf
1-17
-------
Samet, J (2009b). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. Consultation on EPA's Particulate Matter National
Ambient Air Quality Standards: Scope and Methods Plan for Urban Visibility Impact Assessment. May
21,2009. Available at:
http://voseniite.epa.gov/sab/sabproduct.nsf/OF63D7995F5850D5852575BD0077869C/$File/EPA-CASAC-
09-010-unsigned.pdf.
Samet J (2009c). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. Review of Risk Assessment to Support the Review of
the Particulate Matter (PM) Primary National Ambient Air Quality Standards - External Review Draft
(September 2009). November 24, 2009. Available at:
http://yosemite.epa. gov/sab/sabproduct.nsf/BClECC5D539EF72385257678006D5754/$File/EPA-
CAS AC-10-003 -unsignedodf
Samet J (2009d). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. Review of Particulate Matter Urban-Focused
Visibility Assessment (External Review Draft, September 2009). November 24, 2009. Available at:
http://YOsemite.epa.gov/sab/sabproduct.nsf/15872217938041F685257678006A26E3/$File/EPA-CASAC-
10-002-unsigned.pdf
Samet J (2010a). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CASAC Review of Quantitative Health Risk
Assessment for Particulate Matter - Second External Review Draft (February 2010). April 15, 2010.
Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/BC4F6E77B6385155852577070002F09F/$File/EPA-CASAC-
10-008-unsigned.pdf
Samet J (2010b). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CASAC Review of Particulate Matter Urban-Focused
Visibility Assessment - Second External Review Draft (January 2010). April 20, 2010. Available at:
http://vosemite.epa. gov/sab/sabproduct.nsf/OD5CB76AFE7FA77C8525770D004EED55/$File/EPA-
CASAC-10-009-unsigned.pdf
Samet J (2010c). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CASAC Review of Policy Assessment for the Review
of 'the PMNAAQS- First External Review Draft (March 2010). May 17,2010. Available at:
http://yosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/E504EE3276D87A9E8
525772700647 AFB/$File/EPA-CASAC-10-011-unsigned.pdf/.
Samet J (2010d). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CASAC Review of Policy Assessment for the Review
of the PM NAAQS - Second External Review Draft (June 2010). September 10, 2010. Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/CCF9F4C0500C500F8
525779D0073C593/$File/EP A-CASAC-10-015-unsigned.pdf.
US Department of Health, Education and Welfare (DHEW). (1969). Air Quality Criteria for Particulate Matter.
U.S. Government Printing Office, Washington DC, AP-49.
US EPA (2004). Air Quality Criteria for Particulate Matter. National Center for Environmental Assessment, Office
of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711;
report no. EPA/600/P-99/002aF and EPA/600/P-99/002bF. October 2004. Available at:
http://www.epa.gOv/ttn/naaas/standards/pm/s pm cr cd.html.
1-18
-------
US EPA (2005). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA EPA-452/R-05-005a. December 2005. Available
at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html.
US EPA (2007). Draft Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate
Matter. National Center for Environmental Assessment and Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/P-08-006.
October 2007. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_pd.html.
US EPA (2008a). Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate Matter.
National Center for Environmental Assessment-RTF Division and Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-
08-004. March 2008. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007jd.html.
US EPA (2008b). Integrated Science Assessment for Paniculate Matter: First External Review Draft. National
Center for Environmental Assessment-RTF Division, Office of Air Quality Planning and Standards,
Research Triangle Park, NC. EPA/600/R-08/139 and 139A. December 2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007 isa.html.
US EPA (2009a). Integrated Science Assessment for Paniculate Matter: Final Report. National Center for
Environmental Assessment-RTF Division, Office of Research and Development, Research Triangle Park,
NC. EPA/600/R-08/139F. December 2009. Available at:
http://www.epa.gOv/ttn/naaqs/standards/pm/s jm_2007_isa.html.
US EPA (2009b). Integrated Science Assessment for Paniculate Matter: Second External Review Draft. National
Center for Environmental Assessment-RTF Division, Office of Research and Development, Research
Triangle Park, NC. EPA/600/R-08/139B. July 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007 isa.html.
US EPA (2009c). Paniculate Matter National Ambient Air Quality Standards: Scope and Methods Plan for Health
Risk and Exposure Assessment. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA-452/P-09-002. February 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007_pd.html.
US EPA (2009d). Paniculate Matter National Ambient Air Quality Standards: Scope and Methods Plan for Urban
Visibility Impact Assessment. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA-452/P-09-001. February 2009. Available at:
http://www.epa. gov/ttn/naaqs/standards/pm/s_pm_2007_pd.html.
US EPA (2009e). Risk Assessment to Support the Review of the PM Primary National Ambient Air Quality
Standards - External Review Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA-452/P-09-006. September 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007 risk.html.
US EPA (20091). Paniculate Matter Urban-Focused Visibility Assessment - External Review Draft. Office of Air
Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA-452/P-09-005. September 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (2009g). Policy Assessment for the Review of the Paniculate Matter National Ambient Air Quality
Standards - Preliminary Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA-452/P-09-007. September 2009. Available at:
http://www.epa.gOv/ttn/naaas/standards/pm/s pm 2007 pa.html.
1-19
-------
US EPA (20 lOa). Quantitative Health Risk Assessment for Paniculate Matter - Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
10-005. June 2010. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (2010b). Paniculate Matter Urban-Focused Visibility Assessment - Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
10-004. July 2010. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (20 lOc). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards - First External Review Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA452/P-10-003. March 2010. Available at:
http://www.epa.gOv/ttn/naaqs/standards/pm/s jm_2007ja.html.
US EPA (2010d). Quantitative Risk Assessment for Particulate Matter - Second External Review Draft. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA-452/P-10-001. February 2010. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (2010e). Particulate Matter Urban-Focused Visibility Assessment - Second External Review Draft. Office
of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
NC. EPA-452/P-10-002. January 2010. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (2010f). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards - Second External Review Draft. Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. EPA 452/P-10-007. June 2010.
Available at: http://www.epa.gOv/ttn/naaas/standards/pm/s pm 2007 pa.html.
1-20
-------
2 REVIEW OF THE PRIMARY STANDARDS FOR FINE PARTICLES
This chapter presents staff conclusions with regard to the adequacy of the current suite of
primary PM2.5 standards and the alternative primary standards for fine particles that are
appropriate for consideration in this review. Our assessment of these issues is framed by a series
of key policy-relevant questions, which expand upon those presented at the outset of this review
in the IRP (US EPA, 2008a). Answers to these questions will inform decisions by the
Administrator on whether, and if so how, to revise the current suite of primary fine particle
standards.
Staff notes that final decisions regarding the primary standards must draw upon scientific
information and analyses about health effects and risks, as well as judgments made about how to
deal with the uncertainties that are inherent in the scientific evidence and analyses. Ultimately,
the final decisions are largely public health policy judgments. Our approach to informing these
judgments recognizes that the available health effects evidence generally reflects a continuum
consisting of ambient levels at which scientists generally agree that health effects occur through
lower levels at which the likelihood and magnitude of the response become appreciably more
uncertain.
Our approach for reviewing the primary standards for fine particles is presented in
section 2.1. Staff conclusions regarding the adequacy of the current suite of primary PM2.5
standards are presented in section 2.2, focusing on both evidence-based and quantitative risk-
based considerations. Section 2.3 presents our conclusions with respect to alternative fine
particle standards that are appropriate to consider, addressing each of the basic elements of the
standards: indicator (section 2.3.1), averaging time (section 2.3.2), form (section 2.3.3), and
level (section 2.3.4). Section 2.4 summarizes staff conclusions on the primary fine particle
standards. Key uncertainties and areas for future research and data collection efforts are
included in section 2.5.
2.1 APPROACH
Staffs general approach for reviewing the current primary PM2 5 standards, which
involves translating scientific and technical information into the basis for addressing key policy-
relevant questions, takes into consideration the approaches used in previous PM NAAQS
reviews (section 2.1.1) and the court's remand of the primary annual PM2.5 standard set in 2006
(section 2.1.2). The past and current approaches described below are all based most
fundamentally on using information from epidemiological studies to inform the selection of PM
standards that, in the Administrator's judgment, protect public health with an adequate margin of
safety. Such information can be in the form of air quality distributions over which health effect
2-1
-------
associations have been observed, or in the form of concentration-response (C-R) functions that
support quantitative risk assessment. However, evidence- and risk-based approaches using
information from epidemiological studies to inform decisions on PM standards are complicated
by the recognition that no population threshold, below which it can be concluded with
confidence that PM-related effects do not occur, can be discerned from the available evidence.
As a result, any general approach to reaching decisions on what standards are appropriate
necessarily requires judgments about how to translate the information available from the
epidemiological studies into a basis for appropriate standards. This includes consideration of
how to weigh the uncertainties in the reported associations across the distributions of PM
concentrations in the studies and the uncertainties in quantitative estimates of risk. Such
approaches are consistent with setting standards that are neither more nor less stringent than
necessary, recognizing that a zero-risk standard is not required by the CAA. Our current
approach for evaluating the primary PM2.5 standards using both evidence- and risk-based
considerations is outlined in section 2.1.3.
2.1.1 Approaches Used in Previous Reviews
2.1.1.1 Review Completed in 1997
In setting the 1997 primary PM2.5 annual and 24-hour standards, the Agency relied
primarily on an evidence-based approach that focused on epidemiological evidence, especially
from short-term exposure studies of fine particles judged to be the strongest evidence at that
time. The EPA did not place much weight on quantitative risk estimates from the very limited
risk assessment conducted, but did conclude that the risk assessment results confirmed the
general conclusions drawn from the epidemiological evidence that a serious public health
problem was associated with ambient PM levels allowed under the then current PMio standards
(62 FR 38665/1, July 18, 1997).
The EPA considered the epidemiological evidence and data on air quality relationships to
set an annual PM2 5 standard that was intended to be the "generally controlling" standard; i.e., the
primary means of lowering both long- and short-term ambient concentrations of PM2.5.1 In
conjunction with the annual standard, EPA also established a 24-hour PM2.5 standard to provide
supplemental protection against days with high peak concentrations, localized "hotspots," and
1 In so doing, EPA noted that because an annual standard would focus control programs on annual average PM2 5
concentrations, it would not only control long-term exposure levels, but would also generally control the overall
distribution of 24-hour exposure levels, resulting in fewer and lower 24-hour peak concentrations. Alternatively, a
24-hour standard that focused controls on peak concentrations could also result in lower annual average
concentrations. Thus, EPA recognized that either standard could provide some degree of protection from both short-
and long-term exposures, with the other standard serving to address situations where the daily peaks and annual
averages are not consistently correlated (62 FR 38669).
2-2
-------
risks arising from seasonal emissions that might not be well controlled by a national annual
standard (62 FR 38669/3). Recognizing that there are various ways to combine two standards to
achieve an appropriate degree of public health protection, EPA evaluated but did not use an
alternate approach that considered short- and long-term exposure evidence, analyses, and
standards independently. The EPA concluded that the selected approach based on a generally
controlling annual standard was the most effective and efficient approach. This conclusion was
based in part on a key observation from the quantitative risk assessment that most of the
aggregated annual risk associated with short-term PM2.5 exposures through a year results from
the large number of days during which the 24-hour average concentrations are in the low- to
mid-range, well below the peak 24-hour concentrations. As a result, lowering a wide range of
ambient 24-hour PM2 5 concentrations by means of a generally controlling annual standard, as
opposed to focusing on control of peak 24-hour concentrations, was determined to be the most
effective and efficient way to reduce total population risk (62 FR 38668 to 38671).
In setting the level of the annual standard in 1997, EPA first determined a level for the
annual standard based on the short-term exposure studies, and then considered whether the key
long-term exposure studies suggested the need for a lower level. While recognizing that health
effects may occur over the full range of concentrations observed in the studies, EPA concluded
that the strongest evidence for short-term PM2.5 exposure-related effects occurs at concentrations
near the long-term (e.g., annual) average in the short-term exposure studies. The EPA selected a
level for the annual standard at or below the long-term mean concentrations in studies providing
evidence of associations with short-term exposures, placing greatest weight on those short-term
exposure studies that reported clearly statistically significant associations with mortality and
morbidity effects (62 FR 38676/1). Further consideration of the average PM2.5 concentrations
across the cities in the key long-term exposure studies of mortality and respiratory effects in
children did not provide a basis for establishing a lower annual standard level. Because the
annual standard level selected was below the range of annual concentrations most strongly
associated with both short- and long-term exposure effects at that time, and because even small
changes in annual means in this concentration range could make a significant difference in
overall risk reduction and total population exposures, EPA concluded that this standard would
provide an adequate margin of safety against effects observed in these epidemiological studies
(62 FR 68676/3).
The selection of the level of the annual standard was done in conjunction with having
first selected the form of the annual standard to be based on the concentration measured at a
20
-3
-------
single monitor sited to represent community-wide air quality2, or a value resulting from an
average of measurements from multiple community-wide air quality monitoring sites that met
specific criteria and constraints (i.e., "spatial averaging"; 62 FR 38672). This decision
emphasized consistency with the types of air quality measurements that were used in the relevant
epidemiological studies. In reaching this decision, EPA recognized the importance of ensuring
that spatial averaging would not result in inequities in the level of protection provided by the
PM2.5 standards in some areas. Because the annual standard, defined in terms of single or
averaged community-wide air quality monitoring sites, could not be expected to offer an
adequate margin of safety against the effects of all potential short-term exposures in areas with
strong local or seasonal sources that could not be directly evaluated in the epidemiological
studies, EPA set the level of the 24-hour standard to supplement the control afforded by the
annual standard based on air quality relationships between annual and 24-hour concentrations.
This approach was intended to provide an adequate margin of safety against infrequent or
isolated peak concentrations that could occur in areas that attain the annual standard (62 FR
38677). The selection of the level of the 24-hour standard was done in conjunction with having
selected the form of the 24-hour standard to be based on the concentration measured at each
population-oriented monitor3 within an area (62 FR 38674).
2.1.1.2 Review Completed in 2006
In 2006, EPA used a different evidence-based approach to assess the appropriateness of
the levels of the 24-hour and annual PM2.5 standards. Based on an expanded body of
epidemiological evidence that was stronger and more robust, including both short- and long-term
exposure studies, the Administrator decided that using evidence of effects associated with
periods of exposure that were most closely matched to the averaging time of each standard was
the most appropriate public health policy approach for evaluating the scientific evidence to
inform selecting the level of each standard. Thus, the Administrator relied upon evidence from
the short-term exposure studies as the principal basis for selecting the level of the 24-hour PM2.5
standard that would protect against effects associated with short-term exposures. The
Administrator relied upon evidence from long-term exposure studies as the principal basis for
As outlined in section 1.3, community-wide monitoring sites are considered beyond the zone of influence of a
single source, and should have a neighborhood- to urban-scale zone of representation. The principal purpose of
community-oriented monitoring sites is to approximate the long- and short-term exposures of large numbers of
people where they live, work, and play. See 40 CFR part 58 for additional information.
3 As outlined in section 1.3, population-oriented monitoring sites represent residential areas, commercial areas,
recreational areas, industrial areas where workers from more than one company are located, and other areas where a
substantial number of people may spend a significant fraction of their day. See 40 CFR part 58, and especially the
definitions in section 58.1 and the provisions of section 58.30, for additional information.
2-4
-------
selecting the level of the annual PM2.5 standard that would protect against effects associated with
long-term exposures.
With respect to quantitative risk-based considerations, the Administrator determined that
the estimates of risks likely to remain upon attainment of the 1997 suite of PM2.5 standards were
indicative of risks that could be reasonably judged important from a public health perspective,
and, thus, supported revision of the standards. However, the Administrator judged that the
quantitative risk assessment had important limitations and did not provide an appropriate basis
for selecting the levels of the revised standards (71 FR 61174/1-2). The Administrator more
heavily weighed the implications of the uncertainties associated with the quantitative risk
assessment than CASAC did in their comments on the proposed rulemaking. Specifically, in
CASAC's request for reconsideration,4 they stated, "[w]hile the risk assessment is subject to
uncertainties, most of the PM Panel found EPA's risk assessment to be of sufficient quality to
inform its recommendations.. .The risk analyses indicated that the uncertainties would increase
rapidly below an annual level of 13 |ig/m3 - and that was the basis for the PM Panel's
recommendation of 13 |ig/m3 as the lower bound for the annual PM2.5 standard level"
(Henderson, 2006a, p.3).
With regard to the primary annual PM2.5 standard, the Administrator placed the greatest
weight on the long-term means of the concentrations associated with mortality effects in two key
long-term exposure studies in the record, the American Cancer Society (ACS) and Harvard Six
Cities studies (71 FR at 61172 to 61177). Important validation and reanalyses of the original
studies provided "evidence of generally robust associations and provide[d] a basis for greater
confidence in the reported associations than in the last review," and the extended ACS study
provided "new evidence of mortality related to lung cancer and further substantiate^] the
statistically significant associations with cardiorespiratory-related mortality observed in the
original studies" (71 FR 61172/1-2). The Administrator also recognized the availability of long-
term exposure studies that provided evidence of respiratory morbidity, including changes in lung
function measurements and decreased growth in lung function as reported in the 24-Cities study
and the Southern California Children's Health Study (CHS), respectively (Dockery et al. 1996,
Gauderman et al., 2002). In retaining the level of the annual standard at 15 |ig/m3, the
Administrator selected a level that was "appreciably below" the long-term average
concentrations reported in the long-term mortality studies the Administrator regarded as "key"
and "basically at the same level" as the long-term average concentrations in the long-term
respiratory morbidity studies. In the judgment of the Administrator, the two long-term
4 As summarized in section 1.2.2, subsequent to the January 17, 2006 proposed rule (71 FR 2620), CASAC provided
additional advice to EPA in a letter to the Administrator requesting reconsideration of CASAC's recommendations,
including its recommendation on the primary annual PM2 5 standard (Henderson, 2006a).
2-5
-------
respiratory morbidity studies of children did "not warrant setting a lower level for the annual
standard than the level warranted based on the key mortality studies" (71 FR 61176/3).
In considering the form of the primary annual PM2 5 standard, the Administrator
strengthened the standard by tightening the criteria for use of spatial averaging. Based on a
much larger set of PM2.5 air quality data than was available in the 1997 review, analyses were
conducted concerning the potential for disproportionate impacts on potentially vulnerable
populations. These analyses suggested that "the highest concentrations in an area tend to be
measured at monitors located in areas where the surrounding population [was] more likely to
have lower education and income levels, and higher percentages of minority populations" (71 FR
61166/2, see also US EPA, 2005, section 5.3.6.1; Schmidt et al., 2005, Attachment A/Analysis
7).5
In revising the level of the 24-hour PM2.5 standard from 65 |ig/m3 to 35 |ig/m3, the
Administrator placed greatest weight on the much expanded body of epidemiological evidence
from U.S. and Canadian short-term PM2.5 exposure studies with more reliable air quality data
that was reanalyzed to address statistical modeling issues. The Administrator recognized that
these studies provided no evidence of clear effect thresholds or lowest-observed effect levels.
Nonetheless, in focusing on the 98th percentile air quality values in these studies, the
Administrator sought to establish a standard level that would require improvements in air quality
generally in areas in which the distribution of daily short-term PM2 5 concentrations could
reasonably be expected to be associated with serious health effects. The Administrator
concluded that although future air quality improvement strategies in any particular area were not
yet defined, most such strategies were likely to move a broad distribution of PM2 5 air quality
values in an area lower, resulting in reductions in risk associated with exposures to PM2.5 levels
across a wide range of concentrations and not just at the 98th percentile concentrations (71 FR
61168/3).
2.1.2 Remand of Primary Annual PM2.s Standard
As noted above in section 1.2.3, several parties filed petitions for review following
promulgation of the revised PM NAAQS in 2006. These petitions challenged several aspects of
the final rule including the selection of the level of the primary PM2.5 annual standard. More
specifically, petitioners representing public health and environmental groups (American Lung
5 As summarized in footnote 29 at 71 FR 61166/2, the 2004 AQCD noted that some epidemiologic studies, most
notably the ACS study of associations between long-term PM2 5 exposure and mortality, reported larger effect
estimates in the cohort subgroup with lower education levels (US EPA, 2004, p 8-103). The 2004 AQCD also noted
that lower education level may be a marker for lower socioeconomic status (SES) that may be related to increased
vulnerability to the effects of fine particle exposures, for example, as a result of greater exposure from proximity to
sources such as roadways and industry, as well as other factors such as poorer health status and limited access to
health care (US EPA, 2004, section 9.2.4.5).
2-6
-------
Association, Environmental Defense, and the National Parks Conservation Association) and
several states and state agencies argued that the decision to retain the level of the annual PM2.5
standard at 15 |ig/m3 was "arbitrary, capricious, an abuse of discretion, or otherwise not in
accordance with law." 42 U.S.C. 7607(d)(9). The primary 24-hour PM2.5 standard was not
challenged by any of the litigants and, thus, not considered in the court's review and final
decision.
On judicial review, the Court of Appeals for the District of Columbia Circuit (D.C.
Circuit) remanded the primary annual PM2.5 NAAQS to EPA because the Agency failed to
adequately explain why the annual standard provided the requisite protection from both short-
and long-term exposures to fine particles including protection for susceptible populations.
American Farm Bureau Federation v. EPA, 559F.3d512(D.C. Cir. 2009). With respect to
human health protection from short-term PM2.5 exposures, the court considered the different
approaches used by EPA in the 1997 and 2006 PM NAAQS decisions, as summarized above.
The court found that EPA failed to adequately explain why a 24-hour PM2.5 standard by itself
would provide the protection needed from short-term exposures and remanded the primary
annual PM2.5 standard to EPA "for further consideration of whether it is set at a level requisite to
protect the public health while providing an adequate margin of safety from the risk of short-
term exposures to PM2.5." American Farm Bureau Federation, 559 F. 3d at 520-24. In so
holding, the court emphasized that the Administrator had failed to consider the short-term studies
when setting the annual standard, and failed to provide a reasoned explanation of why it had
done so. Id. at 521.
With respect to protection from long-term exposure to fine particles, the court found that
EPA failed to adequately explain how the primary annual PM2.5 standard provided an adequate
margin of safety for children and other susceptible populations. The court found that EPA did
not provide a reasonable explanation of why certain morbidity studies, including the study of
children in Southern California showing lung damage associated with long-term PM2.5 exposure
(Gauderman et.al, 2000) and a multi-city study (24-Cities Study) evaluating decreased lung
function in children associated with long-term PM2.5 exposures (Raizenne et al., 1996), did not
call for a more stringent annual PM2 5 standard. Id. at 522-23. Specifically, the court found that:
EPA was unreasonably confident that, even though it relied solely upon long-term
mortality studies, the revised standard would provide an adequate margin of safety with
respect to morbidity among children. Notably absent from the final rule, moreover, is
any indication of how the standard will adequately reduce risk to the elderly or to those
with certain heart or lung diseases despite (a) the EPA's determination in its proposed
rule that those subpopulations are at greater risk from exposure to fine particles and (b)
the evidence in the record supporting that determination. Id. at 525.
2-7
-------
In remanding the primary annual PM2.5 standard for reconsideration, the court did not
vacate the standard. Id. at 530.
2.1.3 General Approach Used in Current Review
This review is based on an assessment of a much expanded body of epidemiological
evidence, more extensive air quality data and analyses, and a more comprehensive quantitative
risk assessment relative to the information available in past reviews, as presented in the ISA and
RA. As a result, staffs general approach to reaching conclusions about the adequacy of the
current suite of PM2.5 standards and potential alternative standards that are appropriate to
consider is broader and more integrative than in past reviews. Our general approach also reflects
consideration of the issues raised by the court in its remand of the primary annual PM2.5
standard, since decisions made in this review, and the rationales for those decisions, will
comprise the Agency's response to the remand.
Our general approach takes into account both evidence-based and risk-based
considerations, and the uncertainties related to both types of information, as well as advice from
CAS AC (Samet, 2010c,d) and public comments on the first and second draft PAs (US EPA,
2010c,f). In so doing, we are seeking to provide as broad an array of policy options as is
supportable by the available 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 approaches and types of information
presented in this document.
We believe it is most appropriate to consider the protection against PM2.5-related
mortality and morbidity effects, associated with both long- and short-term exposures, afforded by
the annual and 24-hour standards taken together, as was done in the 1997 review, rather than to
consider each standard separately, as was done in the 2006 review.6 As EPA recognized in 1997,
there are various ways to combine two standards to achieve an appropriate degree of public
health protection. The extent to which these two standards are interrelated in any given area
depends in large part on the relative levels of the standards, the peak-to-mean ratios that
characterize air quality patterns in an area, and whether changes in air quality designed to meet a
given suite of standards are likely to be of a more regional or more localized nature.
6 By utilizing this approach, the Agency would also be responsive to the remand of the 2006 standard. As noted in
section 2.1.2 above, the D.C. Circuit, in remanding the 2006 primary annual PM25 standard, concluded that the
Administrator had failed to adequately explain why an annual standard was sufficiently protective in the absence of
consideration of the long-term mean PM25 concentrations in short-term exposure studies as well, and likewise had
failed to explain why a 24-hour standard was sufficiently protective in the absence of consideration of the effect of
an annual standard on reducing the overall distribution of 24-hour average PM2 5 concentrations. 559 F. 3d at 520-
24.
-------
In considering the combined effect of annual and 24-hour standards, we recognize that
changes in PM2.5 air quality designed to meet an annual standard would likely result not only in
lower annual average PM2 5 concentrations but also in fewer and lower peak 24-hour PM2 5
concentrations. We also recognize that changes designed to meet a 24-hour standard would
result not only in fewer and lower peak 24-hour PM2.5 concentrations but also in lower annual
average PM2.5 concentrations. Thus, either standard could be viewed as providing protection for
both short- and long-term exposures, with the other standard serving to address situations where
the daily peak and annual average concentrations are not consistently correlated.
With respect to the currently available evidence, we note the short-term exposure studies
are primarily drawn from epidemiological studies that associated variations in area-wide effects
with monitor(s) that gauged the variation in daily PM2 5 concentrations over the course of several
years. The strength of the associations in these data is demonstrably in the numerous "typical"
days within the air quality distribution, not on the peak days. Furthermore, based on the
quantitative risk assessments conducted for this and previous reviews, we recognize that much, if
not most of the aggregated risk associated with short-term exposures results from the large
number of days during which the 24-hour average concentrations are in the low-to mid-range,
below the peak 24-hour concentrations (see section 2.2.2; US EPA, 2010a, section 3.1.2.2). In
addition, there is no evidence suggesting that risks associated with long-term exposures are likely
to be disproportionately driven by peak 24-hour concentrations.7 For these reasons, strategies
that focus primarily on reducing peak days are less likely to achieve reductions in PM2.5
concentrations that are most strongly associated with the observed health effects compared to an
approach that concentrates on reducing the more typical part of the air quality distribution.
Furthermore, a policy approach that focuses on reducing peak exposures would most likely result
in more uneven public health protection across the U.S. by either providing inadequate
protection in some areas or overprotecting in other areas (US EPA, 2010a, section 5.2.3).
Staff concludes a policy goal of setting a "generally controlling" annual standard that will
lower a wide range of ambient 24-hour PM2 5 concentrations, as opposed to focusing on control
of peak 24-hour PM2.5 concentrations, is the most effective and efficient way to reduce total
population risk and so provide appropriate protection. This approach would likely reduce
aggregate risks associated with both long- and short-term exposures with more consistency than
a generally controlling 24-hour standard and would likely avoid setting national standards that
could result in relatively uneven protection across the country, due to setting standards that are
either more or less stringent than necessary in different geographical areas.
7 In confirmation, a number of studies that have presented analyses excluding higher PM concentration days
reported a limited effect on the magnitude of the effect estimates or statistical significance of the association (e.g.,
Dominici, 2006b; Schwartz et al, 1996; Pope and Dockery, 1992).
2-9
-------
However, we conclude that an annual standard intended to serve as the primary means for
providing protection for effects associated with both long- and short-term PM2.5 exposures
cannot be expected to offer an adequate margin of safety against the effects of all short-term
PM2.5 exposures. As a result, in conjunction with a generally controlling annual standard, we
conclude that it is appropriate to consider setting a 24-hour standard to provide supplemental
protection, particularly for areas with high peak-to-mean ratios possibly associated with strong
local or seasonal sources, or PM2.5-related effects that may be associated with shorter-than-daily
exposure periods.
Our consideration of the protection afforded by the current and alternative suites of
standards focuses on PM2.5-related health effects associated with long-term exposures, for which
the magnitude of quantitative estimates of risks to public health generated in the risk assessment
is appreciably larger in terms of overall incidence and percent of total mortality or morbidity
effects than for short-term PM2.s-related effects. We also consider effects and estimated risks
associated with short-term exposures. In both cases, we place greatest weight on health effects
that have been judged in the ISA to have a causal or likely causal relationship with PM2 5
exposures, while also considering health effects judged to be suggestive of a causal relationship
or that focus on specific susceptible populations. We focus on epidemiological studies
conducted in the U.S. and Canada, as studies in other countries reflect air quality and exposure
patterns that are not necessarily typical of the U.S.,8 and place relatively greater weight on
statistically significant associations that yield relatively more precise effect estimates and that are
judged to be robust to confounding by other air pollutants. In the case of short-term exposure
studies, we consider evidence from large multi-city studies, as well as single-city studies.
In translating information from epidemiological studies into the basis for reaching staff
conclusions on the adequacy of the current suite of standards (section 2.2.1), we consider a
number of factors. As an initial matter, we consider the extent to which the currently available
evidence and related uncertainties strengthens or calls into question conclusions from the last
review regarding associations between fine particle exposures and health effects. We also
consider evidence on susceptible populations and potential impacts on such populations.
Further, we explore the extent to which PM2.s-related health effects have been observed in areas
where air quality distributions extend to lower levels than previously reported or in areas that
would likely have met the current suite of standards.
In translating information from epidemiological studies into the basis for reaching staff
conclusions on alternative standard levels for consideration (section 2.3.4), we first recognize the
absence of discernible thresholds in the C-R functions from long- and short-term PM25 exposure
8 Nonetheless, we recognize the importance of all studies, including international studies, in the ISA's assessment of
the weight of the evidence that informs causality determinations.
2-10
-------
studies (US EPA, 2009a, section 2.4.3). In the absence of any discernible thresholds, our general
approach for identifying appropriate standard levels for consideration involves characterizing the
range of PM2 5 concentrations over which we have the most confidence in the associations
reported in epidemiological studies. In so doing, we recognize that there is no single factor or
criterion that comprises the "correct" approach, but rather there are various approaches that are
reasonable to consider for characterizing the confidence in the associations and the limitations
and uncertainties in the evidence. Identifying the implications of various approaches in reaching
conclusions on the range of alternative standard levels that is appropriate to consider can help
inform decisions that are made to either retain or revise the standards. Final decisions will
necessarily also take into account public health policy judgments as to the degree of health
protection that is to be achieved.
In reaching staff conclusions on the range of annual standard levels that is appropriate to
consider, we focus on identifying an annual standard that provides protection for effects
associated with both long- and short-term exposures, as discussed above. In so doing, we
explore different approaches for characterizing the range of PM2.5 concentrations over which our
confidence in associations for both long- and short-term exposures is greatest, as well as the
extent to which our confidence is reduced at lower PM2.5 concentrations.
The approach that most directly addresses this issue considers studies that present
confidence intervals around C-R relationships, and in particular, analyses that average across
multiple C-R models rather than considering a single C-R model.9 We explore the extent to
which such analyses have been published for studies of health effects associated with long- or
short-term PM2 5 exposures. Such analyses could potentially be used to characterize a
concentration below which uncertainty in a C-R relationship substantially increases or is judged
to be indicative of an unacceptable degree of uncertainty about the existence of a continuing C-R
relationship. We believe that identifying this area of uncertainty in the C-R relationship can be
used to inform identification of alternative standard levels that are appropriate to consider.
We also take into account the general approach used in previous PM reviews which
focused on consideration of alternative standard levels that were somewhat below the long-term
mean PM2 5 concentrations reported in epidemiological studies. This approach recognizes that
the strongest evidence of PM25-related associations occurs at concentrations near the long-term
(i.e., annual) mean. In using this approach, we place greatest weight on those long- and short-
9
This is distinct from confidence intervals around C-R relationships that are related to the magnitude of effect
estimates generated at specific PM25 concentrations (i.e., point-wise confidence intervals) and that are relevant to
the precision of the effect estimate across the air quality distribution, rather than to our confidence in the existence
of a continuing C-R relationship across the entire air quality distribution on which a reported association was based.
2-11
-------
term exposure studies that reported statistically significant associations with mortality and
morbidity effects.
In extending this approach, we also consider information beyond a single statistical
metric of PM2.5 concentrations (i.e., the mean) to the extent such information is available. In so
doing, we employ distributional statistics (i.e., statistical characterization of an entire distribution
of data) to identify the broader range of PM2.5 concentrations that were most influential in
generating health effect estimates in epidemiological studies. Thus, we consider the range of
PM2.5 concentrations where the data analyzed in the study (i.e., air quality and population-level
data, as discussed below) are most concentrated, specifically, the range of PM2.5 concentrations
around the long-term mean over which our confidence in the associations observed in the
epidemiological studies is greatest. We then focus on the lower part of this range and seek to
characterize where in the distributions the data become appreciably more sparse and, thus, where
our understanding of the associations correspondingly becomes more uncertain. We recognize
there is no one percentile value within a given distribution that is the most appropriate or
"correct" way to characterize where our confidence in the associations becomes appreciably
lower. We judge that the range from the 25th to 10th percentiles is a reasonable range to consider
as a region where we have appreciably less confidence in the associations.10
In considering distributional statistics from epidemiological studies, we first recognize
that there are two types of population-level metrics that are useful to consider in identifying the
PM2.5 concentrations most influential in generating the health effect estimates reported in the
epidemiological studies. The most relevant information is the distribution of health events (e.g.,
deaths, hospitalizations) occurring within a study population in relation to the distribution of
PM2.5 concentrations. However, in recognizing that access to health event data can be restricted,
we also consider the number of study participants within each study area as a reasonable
surrogate for health event data.11
In the absence of data on the number of health events or study participants, we consider
the distribution of PM2 5 concentrations across study areas in multi-city studies as representative
of the PM2.5 concentrations likely experienced by study participants, which are integral to the
generation of effect estimates. This approach is particularly relevant for identifying the range of
10 In the PM NAAQS review completed in 2006, staff recognized that the evidence of an association in any
epidemiological study is "strongest at and around the long-term average where the data in the study are most
concentrated. For example, the interquartile range of long-term average concentrations within a study [with a lower
bound of the 25th percentile] or a range within one standard deviation around the study mean, may reasonably be
used to characterize the range over which the evidence of association is strongest" (US EPA, 2005, p. 5-22). A
range of one standard deviation around the mean represents approximately 68% of normally distributed data, and,
below the mean falls between the 25th and 10th percentiles.
11 A limitation of applying population data as a substitute for the number of the health events occurring in a study
area is the recognition that baseline incidence rates of health events will vary across study areas.
2-12
-------
PM2.5 concentrations most influential in generating the effect estimates in short-term exposure
studies (e.g., time-series studies). In short-term exposure studies, a similar number of events are
likely occurring on a daily basis and, thus, information on the PM2 5 air quality distributions
provides a better approximation of a study participant's likely exposure in relation to the
observed health effects as compared to long-term exposure studies where individual study
participant's exposures change over the course of the study follow-up period (Samet, 2010d, p.
2).
We recognize that an approach considering analyses of confidence intervals around C-R
functions is intrinsically related to an approach that considers different distributional statistics.
Both of these approaches can be employed to identify the range of PM2.5 concentrations over
which we have the most confidence in the associations reported in epidemiological studies.
In applying these approaches, we consider PM2.5 concentrations from long- and short-
term PM2.5 exposure studies using composite monitor distributions12 For multi-city studies, this
distribution reflects concentrations averaged across ambient monitors within each area included
in a given study and then averaged across study areas for an overall study mean PM2.5
concentration. Beyond considering air quality concentrations based on composite monitor
distributions, in the second draft PA we also considered PM2.5 concentrations based on
measurements at the monitor within each area that records the highest concentration (i.e.,
maximum monitor} (US EPA, 2010f, sections 2.1.3 and 2.3.4.1).13 Although consideration of
alternative annual standard levels could be based on either composite or maximum monitor
distributions, we conclude that it is reasonable to place more weight on an approach based on
composite monitor distributions, which represent the PM2 5 concentrations typically presented
and used in epidemiological analyses. Such composite monitor distributions provide a direct link
between PM2 5 concentrations and the observed health effects reported in both long- and short-
term exposure studies.
In reaching staff conclusions on alternative standard levels that are appropriate to
consider, we have also included a broader consideration of the uncertainties related to the C-R
relationships from multi-city long- and short-term exposure studies. Most notably, these
uncertainties relate to our currently limited understanding of the heterogeneity of relative risk
estimates in areas across the country, which may be attributed, in part, to the potential for
12 Using the term "composite monitor" does not imply we can identify one monitor that represents the air quality
evaluated in a specific study area.
13 The maximum monitor distribution is relevant because it is generally used to determine whether a given standard
is met in an area and the extent to which ambient PM2 5 concentrations need to be reduced in order to bring an area
into attainment with the standard. However, maximum monitor distributions represent a far less robust metric than
composite monitor distributions for consideration of alternative annual standard levels because they are available for
only a few epidemiological studies.
2-13
-------
different components within the mix of ambient fine particles to differentially contribute to
health effects observed in the studies and to exposure-related factors. In addition, uncertainties
remain in more fully understanding the role of PM2 5 in relationship to the roles of gaseous co-
pollutants within complex ambient mixtures.
We recognize the level of protection afforded by the NAAQS relies both on the level and
the form of the standard. We conclude that a policy approach that uses data based on composite
monitor distributions to identify alternative standard levels, and then compares those levels to
concentrations at maximum monitors to determine if an area meets a given standard,
direct!onally has the potential to build in some margin of safety.14 This conclusion is consistent
with CAS AC's comments on the second draft PA, in which CAS AC expressed its preference for
focusing on an approach using composite monitor distributions "because of its stability, and for
the additional margin of safety it provides" when "compared to the maximum monitor
perspective" (Samet, et al., 2010d, pp. 2 to 3).
In reaching staff conclusions on alternative 24-hour standard levels that are appropriate
to consider for setting a 24-hour standard intended to supplement the protection afforded by a
generally controlling annual standard, we consider currently available short-term PM2.5 exposure
studies. The evidence from these studies informs our understanding of the protection afforded
by the suite of standards against effects associated with short-term exposures. In considering the
short-term exposure studies, we evaluate both the distributions of 24-hour PM2 5 concentrations,
with a focus on the 98th percentile concentrations to match the form of the current 24-hour PM2.5
standard, to the extent such data were available, as well as the long-term mean PM2.5
concentrations reported in these studies. In addition to considering the epidemiological
evidence, we also consider air quality information based on county-level 24-hour and annual
design values to understand the policy implications of the alternative standard levels supported
by the underlying science. In particular, we consider the extent to which different combinations
of alternative annual and 24-hour standards would support the policy goal of focusing on a
generally controlling annual standard in conjunction with a 24-hour standard that would provide
supplemental protection. Based on the evidence-based considerations outlined above, we then
develop integrated conclusions with regard to alternative suites of standards, including both
annual and 24-hour standards that we conclude are appropriate to consider in this review based
14 Statistical metrics (e.g., means) based on composite monitor distributions may be identical to or below the same
statistical metrics based on maximum monitor distributions. For example, some areas may have only one monitor,
in which case the composite and maximum monitor distributions will be identical in those areas. Other areas may
have multiple monitors that may be very close to the monitor measuring the highest concentrations, in which case
the composite and maximum monitor distributions could be similar in those areas. Still other areas may have
multiple monitors that may be separately impacted by local sources in which case the composite and maximum
monitor distributions could be quite different and the composite monitor distributions may be well below the
maximum monitor distributions.
2-14
-------
on the currently available evidence and air quality information. In so doing, we discuss the roles
that each standard might be expected to play in the protection afforded by alternative suites of
standards.
Beyond these evidence-based considerations, we also consider the quantitative risk
estimates and the key observations presented in the RA. This assessment included an evaluation
of 15 urban case study areas and estimated risk associated with a number of health endpoints
associated with long- and short-term PM2.5 exposures (US EPA, 2010a). As part of our risk-
based considerations, we consider estimates of the magnitude of PM2.s-related risks associated
with recent air quality levels and air quality simulated to just meet the current and alternative
suites of standards using alternative simulation approaches. We also characterize the risk
reductions, relative to the risks remaining upon just meeting the current standards, associated
with just meeting alternative suites of standards. In so doing, we recognize the uncertainties
inherent in such risk estimates, and take such uncertainties into account by considering the
sensitivity of the "core" risk estimates to alternative assumptions and methods likely to have
substantial impact on the estimates. In addition, we consider additional analyses characterizing
the representativeness of the urban study areas within a broader national context to understand
the roles that the annual and 24-hour standards may play in affording protection against effects
related to both long- and short-term PM2.5 exposures.
Staff conclusions related to the primary PM2 5 standards reflect our understanding of both
evidence-based and risk-based considerations to inform two overarching questions related to: (1)
the adequacy of the current suite of PM2.5 standards and (2) potential alternative standards, if
any, that are appropriate to consider in this review to protect against effects associated with both
long- and short-term exposures to fine particles. In addressing these broad questions, we have
organized the discussions below around a series of more specific questions reflecting different
aspects of each overarching question. When evaluating the health protection afforded by the
current or any alternative suites of standards considered, we have taken into account the four
basic elements of the NAAQS: the indicator, averaging time, form, and level.
Figure 2-1 provides an overview of the policy relevant questions that frame our review,
as discussed more fully below. We believe that this general approach provides a comprehensive
basis to help inform the judgments required of the Administrator in reaching decisions about the
current and potential alternative primary fine particle standards and in responding to the remand
of the 2006 primary annual PM2.5 standard.
2-15
-------
Figure 2-1. Overview of Approach for Review of Primary PMi.s Standards
Adequacy of Current Suite of PM26 Standards?
Evidence-based Considerations
f Does currently available evidence and related uncertainties
strengthen or call into question nature of associations?
r Expanded understanding of susceptible populations?
> Effects at lower concentrations then previously seen or in
areas that would have likely met current standards?
Risk-based Considerations |
'r Nature, magnitude, and uncertainties of estimated risks |
remaining upon just meeting the current suite of standards? |
r Roles of annual and 24-hour standards? {
> Importance of remaining risks from public health |
perspective? |
Does information call into
question adequacy of current
suite of PM,,. standards?
,' Support for
' retaining current
\ suite of PM2 b
v standards"
Alternative Suites of Standards Supported by Current Information?
Indicator
• PM:/S indicator?
• Other indicators?
• Ultrafine particles?
* Components?
Averaging Times
Annual and 24-hour?
• Other averaging times?
• Sub-daily exposures?
* Seasonal exposures?
Forms
^•Annual standard: spatial averaging?
^24-hour standard: 98th or 99th
percentiles'
Levels
'-For an annual standard providing primary
protection for long- and short-term exposures?
/'For a 24-hour standard providing supplemental
protection?
Evidence-based Considerations
'Epidemiological and air quality data
•Relative weight placed on information?
•Confidence bounds around C-R relationships?
• Different statistical meincs?
• Implications of alternative suites of standards?
Risk-based Considerations
^Nature, magnitude, and uncertainties of estimated risks
remaining upon just meeting alternative standards?
•Alternative annual standards?
•Alternative 24-hour standards?
/'Confidence in risk estimates?
Alternative suites of standards for consideration
2-16
-------
2.2 ADEQUACY OF CURRENT STANDARDS
In considering the adequacy of the current suite of PM25 standards, staff addresses the
following overarching question:
Does the currently available scientific evidence and risk-based information, as reflected in
the ISA and RA, support or call into question the adequacy of the protection afforded by
the current suite of fine particle standards?
To inform our consideration of this broad question, we address a series of more specific
questions to aid in considering the currently available scientific evidence (section 2.2.1) and the
results of recent quantitative risk analyses (section 2.2.2) in a policy-relevant context, 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 and the RA (US EPA, 2009a; US EPA, 2010a).
CASAC advice regarding the adequacy of the current suite of PM2 5 standards is summarized in
section 2.2.3. Staff conclusions regarding the adequacy of the current suite of PM2.5 standards
are presented in section 2.2.4.
2.2.1 Evidence-based Considerations
Our review of the adequacy of the current suite of primary PM2.5 standards begins by
considering the strength of the evidence, susceptible populations, and the air quality distributions
over which health effects associations have been reported.
• To what extent does the currently available scientific evidence and related uncertainties
strengthen or call into question evidence of associations between ambient fine particle
exposures and health effects?
In considering the strength of the associations between long- and short-term exposures to
PM2.s and health effects, we note that in the last review EPA concluded that there was "strong
epidemiological evidence" for linking long-term PM2 5 exposures with cardiovascular-related
and lung cancer mortality and respiratory-related morbidity and for linking short-term PM2.5
exposures with cardiovascular-related and respiratory-related mortality and morbidity (US EPA,
2004, p. 9-46; US EPA, 2005, p. 5-4). Overall, the epidemiological evidence supported "likely
causal associations" between PM2.5 and both mortality and morbidity from cardiovascular and
respiratory diseases, based on "an assessment of strength, robustness, and consistency in results"
(US EPA, 2004, p. 9-48).
In looking across the extensive new scientific evidence available in this review, our
overall understanding of health effects associated with fine particle exposures has been greatly
expanded. The currently available evidence is stronger in comparison to evidence available in
the last review because of its breadth and the substantiation of previously observed health effects
2-17
-------
(US EPA, 2009a, section 2.3.1). A number of large multi-city epidemiological studies have been
conducted throughout the U.S. including extended analyses of studies that were important to
inform decision making in the last review. These studies have reported consistent increases in
morbidity and/or mortality related to ambient PM2.5 concentrations, with the strongest evidence
reported for cardiovascular-related effects. In addition, the findings of new toxicological and
controlled human exposure studies greatly expand and provide stronger support for a number of
potential biologic mechanisms or pathways for cardiovascular and respiratory effects associated
with long- and short-term PM exposures (US EPA, 2009a, p 2-17; chapter 5; Figures 5-4 and 5-
5).
With regard to causal inferences described in the ISA, we note that since the last review
EPA has developed a more formal framework for reaching causal determinations that draws
upon the assessment and integration of evidence from across epidemiological, controlled human
exposure, and toxicological studies, and the related uncertainties, that ultimately influence our
understanding of the evidence (US EPA, 2009a, section 1.5). This framework employs a five-
level hierarchy that classifies the overall weight of evidence and causality using the following
categorizations: causal relationship, likely to be a causal relationship, suggestive of a causal
relationship, inadequate to infer a causal relationship, and not likely to be a causal relationship
(US EPA, 2009a, Table 1-3).15
Using this causal framework, the ISA concludes that the collective evidence is largely
consistent with past studies and substantially strengthens what was known in the last review to
reach the conclusion that a causal relationship exists between both long- and short-term
exposures to PM2 5 and mortality and cardiovascular effects including cardiovascular-related
mortality. The ISA also concludes that the collective evidence continues to support a likely
causal relationship between long- and short-term PM2.5 exposures and respiratory effects,
including respiratory-related mortality. Further, the ISA concludes that the currently available
evidence is suggestive of a causal relationship between long-term PM2 5 exposures and other
health effects, including developmental and reproductive effects (e.g., low birth weight) and
carcinogenic, mutagenic, and genotoxic effects (e.g., lung cancer mortality). Table 2-1
summarizes these causal determinations (US EPA, 2009a, sections 2.3.1 and 2.6).
15 Causal inferences, as discussed in the ISA, are based not only on the more expansive epidemiological evidence
available in this review but also reflect consideration of important progress that has been made to advance our
understanding of a number of potential biologic modes of action or pathways for PM-related cardiovascular and
respiratory effects (US EPA 2009a, chapter 5).
2-18
-------
Table 2-1. Summary of Causal Determinations for PM2.s
Exposure
Duration
Long-term
Short-term
Outcome
Mortality
Cardiovascular Effects
Respiratory Effects
Reproductive and Developmental Effects
Cancer, Mutagenicity, Genotoxicity Effects
Mortality
Cardiovascular Effects
Respiratory Effects
Central Nervous System Effects
Causality
Determination
Causal
Causal
Likely to be Causal
Suggestive
Suggestive
Causal
Causal
Likely to be Causal
Inadequate
Source: adapted from US EPA, 2009a, Table 2-6.
Health Effects Associated with Long-term PM^ Exposure
With regard to mortality, the ISA concludes that newly available evidence significantly
strengthens the link between long-term exposure to PM2 5 and mortality, while providing
indications that the magnitude of the PM2.5-mortality association may be larger than previously
estimated (US EPA, 2009a, sections 7.2.10, 7.2.1 1, 7.6.1; Figures 7-6 and 7-7). A number of
large U.S. cohort studies have been published since the last review, including extended analyses
of the American Cancer Society (ACS) and Harvard Six Cities studies (US EPA, 2009a, pp 7-84
to 7-85; Figure 7-6; Krewski et al., 2009; Pope et al., 2004; Jerrett et al., 2005; Laden et al.,
2006). In addition, new long-term PM2.5 exposure studies evaluating mortality impacts in
additional cohorts are now available (US EPA, 2009a, section 7.6). For example, the Women's
Health Initiative (WHI) study reported effects of PM25 on cardiovascular-related mortality in
post-menopausal women with no previous history of cardiac disease (Miller et al., 2007), based
on one year of air quality data and multiple years of health event data. Additionally, multiple
studies observed PM2.5-associated mortality among older adults using Medicare data (Eftim et
al., 2008; Zeger et al., 2007, 2008). Collectively, these new studies, along with evidence
available in the last review, provide us with consistent and stronger evidence of associations
between long-term exposure to PM2.5 and mortality (U.S. EPA, 2009a, sections 2.3.1 and 7.6).
The strength of the causal relationship between long-term PM2.s exposure and mortality is
also built upon new studies providing evidence of improvement in community health following
reductions in ambient fine particles. Pope et al. (2009) documented the population health
2-19
-------
benefits of reducing ambient air pollution by correlating past reductions in ambient PM2.5
concentrations with increased life expectancy. These investigators reported that reductions in
ambient fine particles during the 1980s and 1990s account for as much as 15 percent of the
overall improvement in life expectancy in 51 U.S. metropolitan areas, with the fine particle
reductions reported to be associated with an estimated increase in mean life expectancy across
the entire population of approximately 5 to 9 months (US EPA, 2009a, p. 7-95; Pope et al.,
2009). An extended analysis of the Harvard Six Cities study found that as cities cleaned up their
air, locations with the largest reductions in PM2.5 saw the largest improvements in reduced
mortality rates, while those with the smallest decreases in PM2.5 concentrations saw the smallest
improvements (Laden et al., 2006). Another extended follow-up to the Harvard Six Cities study
investigated the delay between changes in ambient PM2 5 concentrations and changes in mortality
(Schwartz et al., 2008) and reported that the effects of changes in PM2.5 were seen within the two
years prior to death (US EPA, 2009a, p. 7-92; Figure 7-9). Looking more broadly across studies,
the ISA concludes, "Generally, these results indicate a developing coherence of the air pollution
mortality literature, suggesting that the health benefits from reducing air pollution do not require
a long latency period and would be expected within a few years of intervention" (US EPA,
2009a, p. 7-95).
With regard to cardiovascular effects, several new studies have examined the association
between cardiovascular effects and long-term PM2 5 exposures in multi-city studies conducted in
the U.S. and Europe. The ISA concludes that the strongest evidence comes from recent studies
investigating cardiovascular-related mortality. This includes evidence from a number of large,
multi-city U.S. long-term cohort studies including extended follow-up analyses of the ACS and
Harvard Six Cities studies, as well as the WHI study (US EPA, 2009a, sections 7.2.10 and 7.6.1;
Krewski et al., 2009; Pope et al., 2004; Laden et al., 2006; Miller et al., 2007). Pope et al. (2004)
reported a positive association between mortality and long-term PM2.5 exposure for a number of
specific cardiovascular diseases, including ischemic heart disease (IHD), dysrhythmia, heart
failure, and cardiac arrest (US EPA, 2009a, p. 7-84; Figure 7-7). Krewski et al. (2009) provides
further evidence for IHD-related mortality associated with long-term PM2.5 exposure (US EPA,
2009a, p. 7-84, Figure 7-7).
With regard to cardiovascular-related morbidity associated with long-term PM2 5
exposures, studies were not available in the last review. Recent studies, however, have provided
new evidence linking long-term exposure to PM2 5 with cardiovascular outcomes that has
"expanded upon the continuum of effects ranging from the more subtle subclinical measures to
cardiopulmonary mortality" (US EPA, 2009a, p. 2-17). In the current review, studies are now
available that evaluated a number of endpoints ranging from subtle indicators of cardiovascular
health to serious clinical events associated with coronary heart disease (CHD) and cardiovascular
2-20
-------
disease (CVD) and cerebrovascular disease (CBVD), including myocardial infarction (MI),
coronary artery revascularization (e.g., bypass graft, angioplasty, stent, atherectomy), congestive
heart failure (CHF), and stroke. The most important new evidence comes from the WHI study
which provides evidence of nonfatal cardiovascular events including both coronary and
cerebrovascular events (Miller et al., 2007; US EPA, 2009a, sections 7.2.9 and 7.6.1).
Toxicological studies provide supportive evidence that the cardiovascular morbidity effects
observed in long-term exposure epidemiological studies are biologically plausible and coherent
with studies of cardiovascular-related mortality as well as with studies of cardiovascular-related
effects associated with short-term exposures to PM2.5, as described below (US EPA, 2009a, p 7-
19).
With regard to respiratory effects, the ISA concludes that extended analyses of studies
available in the last review as well as new epidemiological studies conducted in the U.S. and
abroad provide stronger evidence of respiratory-related morbidity associated with long-term
PM2.5 exposure. The strongest evidence for respiratory-related effects available in this review is
from studies that evaluated decrements in lung function growth, increased respiratory symptoms,
and asthma development (U.S. EPA, 2009a, sections 2.3.1.2, 7.3.1.1, and 7.3.2.1).16
Specifically, extended analyses of the Southern California CHS provided evidence that clinically
important deficits in lung function17 associated with children's long-term exposure to PM2.5
persisted into early adulthood (U.S., EPA, 2009a, p. 7-27; Gauderman et al., 2004). Additional
analyses of the CHS cohort reported an association between long-term PM2.5 exposure and
bronchitic symptoms (US EPA, 2009a, p. 7-24; McConnell et al., 2003) and a strong modifying
effect of PM2 5 on the association between lung function and asthma incidence (US EPA, 2009,
7-24; Islam et al., 2007). The outcomes observed in these more recent reports from the Southern
California CHS, including evaluation of a broader range of endpoints and longer follow-up
periods, were larger in magnitude and more precise than previously reported. Supporting these
results are new longitudinal cohort studies conducted by other researchers in varying locations
using different methods (U.S. EPA, 2009a, section 7.3.9.1). New evidence from a U.S. cohort of
cystic fibrosis (CF) patients provided evidence of association between long-term PM2.5
exposures and exacerbations of respiratory symptoms resulting in hospital admissions or use of
home intravenous antibiotics (US EPA, 2009a, p. 7-25; Goss et al., 2004).
16 Supporting evidence comes from studies "that observed associations between long-term exposure to PM10 and an
increase in respiratory symptoms and reductions in lung function growth in areas where PM10 is dominated by
PM25" (US EPA, 2009a, p. 2-12).
17 Clinical significance was defined as a FEVi below 80% of the predicted value, a criterion commonly used in
clinical settings to identify persons at increased risk for adverse respiratory conditions (US EPA, 2009a, pp. 7-29 to
7-30). The primary NAAQS for SO2 also includes this interpretation for FEV: (75 FR 35525, June 22, 2010).
2-21
-------
Toxicological studies provide coherence and biological plausibility for the respiratory
effects observed in epidemiological studies (US EPA, 2009a, p. 7-42). For example, pre- and
postnatal exposure to ambient levels of urban particles has been found to affect lung
development in an animal model (US EPA, 2009a, section 7.3.2.2; p. 7-43). This finding is
important because impaired lung development is one mechanism by which PM exposure may
decrease lung function growth in children (US EPA, 2009a, p. 2-12; section 7.3).
With regard to respiratory-related mortality associated with long-term PM2 5 exposure,
the ISA concludes that the evidence is "limited and inconclusive" (US EPA, 2009a, p. 7-41).
The extended follow-up of the Harvard Six Cities study reported a positive but statistically
nonsignificant association between long-term PM2.5 exposure and respiratory-related mortality
(Laden et al., 2006), whereas Pope et al. (2004) found no association in the ACS cohort (US
EPA, 2009a, p. 7-84). There is emerging but limited evidence for an association between long-
term PM2.5 exposure and respiratory mortality in post-neonatal infants where long-term exposure
was considered as approximately one month to one year (US EPA, 2009a, pp. 7-54 to 7-55).
Emerging evidence of short- and long-term exposure to PM2.5 and respiratory morbidity and
infant mortality provide some support for the weak respiratory-related mortality effects observed.
Beyond effects considered to have causal or likely causal relationships with long-term
PM2.5 exposure as discussed above, the following health outcomes are classified as having
evidence suggestive of a causal relationship with long-term PM2 5 exposure: (1) reproductive
and developmental effects and (2) cancer, mutagenicity, and genotoxicity effects (US EPA,
2009a, Table 2-6). With regard to reproductive and developmental effects, the ISA notes,
"[e]vidence is accumulating for PM25-related effects on low birth weight and infant mortality,
especially due to respiratory causes during the post-neonatal period" (US EPA, 2009a, section
2.3.1.2). New evidence available in this review reports a significant association between
exposure to PM2.5 during pregnancy and lower birth weight, pre-term birth, and intrauterine
growth restriction, and a significant association between postnatal exposure to PM2.5 and an
increased risk of infant mortality (US EPA, 2009a, section 7.4). The ISA further notes that
"[i]nfants and fetal development processes may be particularly vulnerable to PM exposure, and
although the physical mechanisms are not fully understood, several hypotheses have been
proposed involving direct effects on fetal health, altered placenta function, or indirect effects on
the mother's health" (US EPA, 2009a, section 7.4.1). While toxicological studies provide some
evidence that supports an association between long-term PM2 5 exposure and adverse
reproductive and developmental outcomes, there is "little mechanistic information or biological
plausibility for an association between long-term PM exposure and adverse birth outcomes (e.g.,
low birth weight, infant mortality)" (US EPA, 2009a, p. 2-13).
2-22
-------
With regard to cancer, mutagenic andgenotoxic effects, "[mjultiple epidemiologic
studies have shown a consistent positive association between PM2.5 and lung cancer mortality,
but studies have generally not reported associations between PM2 5 and lung cancer incidence"
(US EPA, 2009a, p. 2-13 and sections 2.3.1.2 and 7.5; Table 7-7; Figures 7-6 and 7-7). The
extended follow-up to the ACS study reported an association between long-term PM2.5 exposure
and lung cancer mortality (US EPA, 2009a, p. 7-71; Krewski et al., 2009) as did the extended
follow-up to the Harvard Six Cities study when considering the entire 25-year follow-up period
(Laden et al., 2006). There is some evidence, primarily from in vitro studies, providing
biological plausibility for the PM-lung cancer relationships observed in epidemiological studies
(US EPA, 2009a, p. 7-80), although toxi col ogical studies of carcinogenicity, mutagenicity, and
genotoxicity generally reported mixed results (US EPA, 2009a, section 7.5).
Health Effects Associated with Short-term PMg^ Exposure
In considering effects associated with short-term PM2.5 exposure, the body of currently
available scientific evidence has been expanded greatly by the publication of a number of new
multi-city time-series studies that have used uniform methodologies to investigate the effects of
short-term fine particle exposures on public health. This body of evidence provides a more
expansive data base and considers multiple locations representing varying regions and seasons
that provide evidence of the influence of local or regional climate and air pollution mixes on
PM2 5-associated health effects. These studies provide more precise estimates of the magnitude
of effects associated with short-term PM2.5 exposure than most smaller-scale single-city studies
that were more commonly available in the last review (U.S. EPA 2009a, chapter 6).
With regard to mortality, new U.S. and Canadian multi-city and single-city PM2 5
exposure studies have found generally consistent positive associations between short-term PM2.5
exposures and cardiovascular- and respiratory-related mortality as well as all-cause (non-
accidental) mortality (US EPA, 2009a, sections 2.3.1.1, 6.2.11 and 6.5.2.2; Figures 6-26, 6-27,
and 6-28). In an analysis of the National Morbidity, Mortality, and Air Pollution Study
(NMMAPS) data, Dominici et al. (2007a) reported associations between fine particle exposures
and all-cause and cardiopulmonary-related mortality (US EPA, 2009a, p. 6-175, Figure 6-26). In
a study of 1 12 U.S. cities, Zanobetti and Schwartz (2009) reported positive associations (in 99%
of the cities) and frequently statistically significant associations (in 55% of the cities) between
short-term PM2.5 exposure and total (non-accidental) mortality (US EPA, 2009a, pp 6-176 to 6-
179; Figures 6-23 and 6-24).18 A Canadian 12-city study (Burnett et al., 2004) is generally
18 Single-city Bayes-adjusted effect estimates for the 112 cities analyzed in Zanobetti and Schwartz (2009) were
provided by the study author (personal communication with Dr. Antonella Zanobetti, 2009; see also US EPA,
2009a, Figure 6-24).
2-23
-------
consistent with an earlier Canadian 8-city study (Burnett and Goldberg, 2003). Both studies
reported a positive and statistically significant association between short-term PM2.5 exposure
and nonaccidental mortality (US EPA, 2009a, p 6-182, Figure 2-1), although the influence of
NO2 and limited PM2.5 data for several years during the study period somewhat diminished the
findings reported in the 12-city study. Collectively, these studies provide generally consistent
and much stronger evidence for PM2.5-associated mortality than the evidence available in the last
review.
With regard to cardiovascular effects, new multi-city as well as single-city short-term
PM2.5 exposure studies conducted since the last review support a largely positive and frequently
statistically significant association between short-term exposure to PM2 5 and cardiovascular-
related morbidity and mortality, substantiating prior findings. For example, among a multi-city
cohort of older adults participating in the Medicare Air Pollution Study (MCAPS), investigators
reported evidence of a positive association between short-term PM2 5 exposures and hospital
admissions related to cardiovascular outcomes (US EPA, 2009a, pp. 6-57 to 58; Dominici et al,
2006a; Bell et al, 2008). The strongest evidence for cardiovascular effects has been observed
predominantly for hospital admissions and emergency department visits for IFID and CFIF and
cardiovascular-related mortality (US EPA, 2009a, Figure 2-1, p. 6-79, sections 6.2.10.3, 6.2.10.5,
and 6.2.11; Bell et al., 2008; Dominici et al., 2006a; Tolbert et al., 2007; Zanobetti and Schwartz,
2009). Furthermore, these findings are supported by a recent study of a multi-city cohort of
women participating in the WHI study that reported a positive but statistically nonsignificant
association between short-term exposure to PM2 5 and electrocardiogram (ECG) measures of
myocardial ischemia (Zhang et al., 2009).
In focusing on respiratory effects, the strongest evidence from short-term PM2 5 exposure
studies has been observed for respiratory-related emergency department visits and hospital
admissions for chronic obstructive pulmonary disease (COPD) and respiratory infections (U.S.
EPA, 2009a, sections 2.3.1.1 and 6.3.8.3; Figures 2-1 and 6-13; Dominici et al., 2006a).
Evidence for PM2.s-related respiratory effects has also been observed in panel studies, which
indicate associations with respiratory symptoms, pulmonary function, and pulmonary
inflammation among asthmatic children. Although not consistently observed, some controlled
human exposure studies have reported small decrements in various measures of pulmonary
function following controlled exposures to PM2 5 (US EPA, 2009a, p. 2-10). Furthermore, the
comparatively larger body of toxicological evidence since the last review is coherent with the
evidence from epidemiological and controlled human exposure studies that examined short-term
exposures to PM2.5 and respiratory effects (US EPA 2009a, section 6.3.10.1).
2-24
-------
Uncertainties in the Evidence
With respect to understanding the nature and magnitude of PM2.5-related risks, as
discussed above, we recognize that epidemiological studies evaluating health effects associated
with long- and short-term PM2.5 exposures have reported heterogeneity in responses both within
and between cities and geographic regions across the U.S. This heterogeneity may be attributed,
in part, to differences in the fine particle composition. However, the currently available evidence
and limited availability of city-specific PM2 5 speciation data does not allow conclusions to be
drawn regarding the relative toxicity of PM2.5 components, or group of components associated
with any source categories of fine particles in different locations. Overall, the ISA concludes
"that many constituents of PM2.5 can be linked with multiple health effects, and the evidence is
not yet sufficient to allow differentiation of those constituents or sources that are more closely
related to specific health outcomes" (US EPA, 2009a, p. 2-17).
Exposure measurement error is also an important source of uncertainty (US EPA, 2009a,
section 3.8.6). Variability in the associations observed across PM2 5 epidemiological studies may
be due in part to exposure error related to measurement-related issues, the use of central fixed-
site monitors to represent population exposure to PM2.5, models used in lieu of or to supplement
ambient measurements, and our limited understanding of factors that may influence exposures
(e.g., topography, the built environment, climate, source characteristics, ventilation usage,
personal activity patterns, photochemistry). As noted in the ISA, exposure measurement error
can introduce bias and increased uncertainty in associated health effect estimates (US EPA,
2009a, p. 2-17).
In particular, we note that there are challenges with interpreting differences in health
effects observed in the eastern versus western parts of the U.S, including evaluating effects
stratified by seasons. As noted in section 2.3.2 of the ISA, western U.S. counties tend to be
larger and more topographically diverse than eastern U.S. counties. These characteristics may
contribute to a higher likelihood of exposure measurement error influencing health effect
estimates. Seasonal differences in effects may be related to PM2.5 composition as well as
regional differences in climate and infrastructure that may affect time spent outdoors or indoors,
housing characteristics including air conditioning usage, and differences in baseline incidence
rates (US EPA, 2009a, p. 3-182). In consideration of these differences, however, the ISA notes
".. .the available evidence and the limited amount of city-specific speciated PM2.s data does not
allow conclusions to be drawn that specifically differentiate effects of PM in different locations"
(US EPA 2009a, p. 2-17). Therefore, we recognize that important uncertainties remain in this
review related to understanding the temporal and spatial variability in PM2 5 concentrations,
including PM2.5 components, and associated health impacts across different geographic areas and
seasons.
2-25
-------
In addition, where PM2.5 and other pollutants (e.g., 63, NO2, CO) are correlated, it can be
difficult to distinguish the effects of the various pollutants in the ambient mixture (i.e, co-
pollutant confounding).19 The use of multipollutant models is generally used in air pollution
epidemiological studies to provide insights into the potential for confounding or interaction
among pollutants (US EPA, 2009a, p. 1-16). A number of research groups have found the
effects of various indicators of PM, including PM2.5, and gaseous copollutants to be independent
of one another (US EPA, 2009a, Figures 6-9 and 6-15). Although many recent US multi-city
studies did not analyze multipollutant models, evidence from single-city studies available in the
last review suggest that gaseous copollutants do not confound the PM2.5-related mortality
association, which is further supported by studies that examined the PMio-mortality relationship
(US EPA, 2009a, p. 6-182 and 6-201).
Summary
In considering the extent to which newly available scientific evidence strengthens or calls
into question evidence of associations identified in the last review between ambient fine particle
exposures and health effects, we recognize that much progress has been made in assessing some
key uncertainties related to our understanding of health effects associated with long- and short-
term exposure to PM2.5. As briefly discussed above as well as in the more complete discussion
of the evidence as assessed in the ISA, we note that the newly available information combined
with information available in the last review provides substantially stronger confidence in a
causal relationship between long- and short-term exposures to PM2.5 and mortality and
cardiovascular effects. In addition, the newly available evidence reinforces and expands the
evidence supporting the likely causal relationship between long- and short-term exposure to
PM2.5 and respiratory effects. With respect to evidence suggestive of a causal relationship for a
broader range of effects, the body of scientific evidence is somewhat expanded but is still limited
with respect to associations between long-term PM2.5 exposures and developmental and
reproductive effects as well as cancer, mutagenic, and genotoxic effects. Thus, we conclude that
there is stronger and more consistent and coherent support for associations between long- and
short-term PM2.5 exposure and a broader range of health outcomes than was available in the last
review, providing the basis for fine particle standards at least as protective as the current PM2 5
standards.
Having reached this initial conclusion, we then consider how the new evidence informs
our understanding of susceptible populations by addressing the following question:
19 A copollutant meets the criteria for potential confounding in PM-health associations if: (1) it is a potential risk
factor for the health effect under study; (2) it is correlated with PM; and (3) it does not act as an intermediate step in
the pathway between PM exposure and the health effect under study (US EPA, 2004, p. 8-10).
2-26
-------
• To what extent does the currently available scientific evidence expand our
understanding of susceptible populations?
Specific groups within the general population, referred to as susceptible populations, are
at increased risk for experiencing adverse health effects related to PM exposures. These groups
could exhibit a greater risk of PM-related health effects than the general population for a number
of reasons including, being affected by lower concentrations of PM, experiencing a larger health
impact at a given PM concentration, and/or being exposed to higher PM concentrations than the
general population. Given the heterogeneity of individual responses to PM exposures, the
severity of the health effects experienced by a susceptible population may be much greater than
that experienced by the population at large.
As summarized below, the currently available epidemiological and controlled human
exposure evidence expands our understanding of previously identified susceptible populations
(i.e., children, older adults, and individuals with pre-existing heart and lung disease) and supports
the identification of additional susceptible populations (e.g., persons with lower socioeconomic
status (SES), genetic differences) (US EPA, 2009a, section 2.4.1, Table 8-2). In addition,
toxicological studies provide underlying support for the biological mechanisms that potentially
lead to increased susceptibility to PM-related health effects.
Lifestages: Children and Older Adults
Two different lifestages have been associated with increased susceptibility to PM-related
health effects: childhood (i.e., less than 18 years of age) and older adulthood (i.e., 65 years of
age and older). Childhood represents a lifestage where susceptibility to PM exposures may be
related to the following observations: children spend more time outdoors; children have greater
activity levels than adults; children have exposures resulting in higher doses per body weight and
lung surface area; and the developing lung is prone to damage, including irreversible effects,
from environmental pollutants as it continues to develop through adolescence (US EPA, 2009a,
section 8.1.1.2). Older adults represent a lifestage where susceptibility to PM-associated health
effects may be related to the higher prevalence of pre-existing cardiovascular and respiratory
diseases found in this age group compared to younger age groups as well as the gradual decline
in physiological processes that occur as part of the aging process (US EPA, 2009a, section
8.1.1.1).
With regard to mortality, recent epidemiological studies have continued to find that older
adults are at greater risk of all-cause (nonaccidental) mortality associated with short-term
exposure to both PM2.5 and PMio, providing consistent and stronger evidence of effects in this
susceptible population compared to the last review. Epidemiological studies that examined the
association between mortality and long-term exposure to PM2 5 that stratified the results by age
2-27
-------
(i.e., less than 65 years of age compared to aged 65 and older; different age groups within the
aged 65 and older population) reported results that are generally consistent with the findings of
these short-term exposure studies and also provided some evidence that the increased risk
declined with increasing age starting at age 60, such that there was no evidence of an association
among persons 85 years and older (US EPA, 2009a, Figure 7-7, section 8.1.1.1, Zeger et al.,
2008).
With regard to morbidity effects, currently available studies provide evidence that older
adults have heightened responses, especially for cardiovascular-related effects, and children have
heightened responses for respiratory-related effects associated with long- and short-term PM2.5
exposures. With regard to older adults, epidemiological studies provide evidence of increases in
PM2 s-related risk of MI, coronary revascularization,20 and their combination with CHD-related
death for participants free of CVD at baseline (Miller et al., 2007) as well as cardiovascular-
related hospitalization (Dominici et al., 2006a; Bell et al., 2008). Further, dosimetry studies have
shown a depression of fine and coarse PM clearance in all regions of the respiratory tract with
increasing age beyond young adulthood, suggesting that older adults are at greater risk of PM-
related respiratory health effects (US EPA, 2009a, section 8.1.1).
With regard to respiratory-related effects in children associated with long-term PM
exposures, our understanding of effects on lung development has been strengthened based on
newly available evidence that is consistent and coherent across different study designs, locations,
and research groups. For example, the strongest evidence comes from the extended follow-up
for the Southern California CHS which includes several new studies that report positive
associations between long-term exposure to PM2 5 and respiratory morbidity, particularly for
such endpoints as lung function growth persisting to young adulthood, respiratory symptoms
(e.g., bronchitic symptoms), and respiratory disease incidence (US EPA, 2009a, section 7.3;
McConnell et al, 2003; Gauderman et al., 2004; Islam et al., 2007). These analyses provide
evidence that PM2.5-related effects persist into early adulthood and are more robust and larger in
magnitude than previously reported.
With regard to respiratory effects in children associated with short-term exposures to PM,
currently available studies provide stronger evidence of respiratory-related hospitalizations with
larger effect estimates observed among children. In addition, reductions in lung function (i.e.,
FEVi) and increases in respiratory symptoms and medication use associated with PM exposures
have been reported among asthmatic children (US EPA, 2009a, sections 6.3.1, 6.3.2.1, 8.4.9).
In addition, accumulating evidence suggests that the developing fetus may also represent
an additional lifestage that is susceptible to PM exposures. The ISA notes that "[i]nfants and
20 Coronary revascularization includes percutaneous coronary interventions, such as angioplasty.
2-28
-------
fetal development processes may be particularly vulnerable to PM exposure, and although the
physical mechanisms are not fully understood, several hypotheses have been proposed involving
direct effects on fetal health, altered placenta function, or indirect effects on the mother's health"
(US EPA, 2009a, section 7.4.1). Evidence is accumulating for PM2.5-related effects on low
birthweight and infant mortality, especially due to respiratory causes during the post-neonatal
period (US EPA, 2009a, sections 2.3.1.2 and 7.4).
Pre-existing Diseases/Health Conditions
A number of health conditions have been found to put individuals at greater risk for
adverse effects following exposure to PM. The currently available evidence confirms and
strengthens evidence in the last review that individuals with underlying cardiovascular and
respiratory diseases are more susceptible to PM exposures.
The majority of the epidemiological studies that examined associations between short-
term PM exposures and cardiovascular outcomes focused on cardiovascular-related hospital
admissions and emergency department visits (US EPA, section 8.1.6.1). There is some new
evidence that individuals with pre-existing IHD are at greater risk of PM-associated hospital
admissions and emergency department visits related to cardiovascular effects. Additional studies
have focused on hypertension and on the effects of PM on cardiac function in individuals with
dysrhythmia with mixed results. One epidemiological study (US EPA, 2009a, section 7.2.9;
Zanobetti and Schwartz, 2007) investigated associations between long-term exposure to PMi0
and the progression of disease or reduced survival in a 21-city study of people discharged
following an acute MI, finding significant associations for mortality, CHF, and new
hospitalization for MI.
With regard to individuals with pre-existing respiratory illnesses (e.g., asthma, COPD),
the ISA presents and assesses a number of studies that evaluate a broad range of health outcomes
(e.g., mortality, asthma symptoms) in response to PM exposures (US EPA, 2009a, section
8.1.6.2). Evidence in asthmatics is stronger and more consistent than previously reported, while
studies of persons with COPD have reported mixed results. Epidemiological studies have
examined the effect of short-term exposure to PM in asthmatics finding an increase in
medication use, asthma attacks, and respiratory symptoms (i.e., cough, shortness of breath, chest
tightness). Controlled human exposure studies reported healthy and asthmatic subjects exposed
to fine, coarse, and ultrafine concentrated ambient particles (CAPs), exhibited similar respiratory
responses, although these studies excluded moderate and severe asthmatics that would be
expected to show increased susceptibility to PM exposure. Toxicological studies using diesel
exhaust particles (DEPs) provide mechanistic support that PM exposure results in allergic
sensitization, and individuals with allergic airway conditions are at greater risk of adverse effects
2-29
-------
upon exposure to PM2.5 (US EPA, 2009a, section 6.3.6.3). Further, there is emerging but limited
evidence which suggests that non-allergic respiratory morbidities may also increase the
susceptibility of an individual to PM-related respiratory effects (US EPA, 2009a, p. 8-12).
There is also emerging evidence that suggests the influence of additional pre-existing
diseases or health conditions, including diabetes and obesity, on the manifestation of PM-related
health effects. The ISA notes that additional research exploring the effect of PM exposures on
obese individuals and identifying the biological pathway(s) that could increase the susceptibility
of diabetic and obese individuals to PM could improve our understanding of these potentially
susceptible populations (US EPA 2009a, pp. 2-23-2-24).
Socioeconomic Status
Stronger evidence is available in this review indicating that people from lower
socioeconomic strata are a susceptible population relative to PM exposures (US EPA, 2009a,
section 8.1.7). Persons with lower socioeconomic status (SES)21 have been generally found to
have a higher prevalence of pre-existing diseases; limited access to medical treatment; and
increased nutritional deficiencies, which can increase this population's risk to PM-related effects.
Evidence available in the last review from the ACS and Harvard Six Cities cohort studies
indicated increased mortality risk with long-term exposure to PM2.5 in the cohort subgroups with
lower education levels (US EPA 2004, section 9.2.4.5). In this review, additional support is
available to identify persons with lower SES as a susceptible population. For example, Krewski
et al. (2009) found moderate evidence for increased lung cancer mortality in individuals with a
high school education or less in response to long-term exposure to PM2.5. However, IHD-related
mortality associated with long-term PM2 5 exposures was most strongly associated with
individuals with higher education levels (US EPA, 2009a, p. 8-15).
Genetic Factors
Investigation of potential genetic susceptibility has provided evidence that individuals
with null alleles or polymorphisms in genes that mediate the antioxidant response to oxidative
stress (e.g., GSTM1), regulate enzyme activity (i.e., MTHFR and cSHMT), or regulate levels of
procoagulants (i.e., fibrinogen) are more susceptible to PM-related effects. However, some
evidence suggests that polymorphisms in genes (e.g., HFE) may provide protection for PM-
related effects. Emerging evidence also suggests that PM exposure can impart epigenetic effects
(i.e., DNA methylation) (US EPA, 2009a, p. 8-16). More research is needed to better understand
the relationship between genetic effects and potential susceptibility to PM-related effects.
21 SES is a composite measure that usually consists of economic status, measured by income; social status measured
by education; and work status measured by occupation (US EPA, 2009a, p. 8-14).
2-30
-------
Summary
In summary, we conclude that there are several susceptible populations that are likely to
be at increased risk of PM-r elated effects, including the lifestages of childhood and older
adulthood, those with preexisting heart and lung diseases, and those of lower SES. Evidence for
PM-related effects in these susceptible populations has expanded and is stronger than previously
observed. We also recognize that there is emerging, though still limited evidence for additional
potentially susceptible populations, such as those with diabetes, people who are obese, or those
with specific genetic factors. We note that the available evidence does not generally allow
distinctions to be drawn between the PM indicators in terms of whether populations are more
susceptible to a particular size fraction (i.e., PM2.5 and PMio-2.s).
• To what extent does the currently available scientific evidence report associations that
extend to air quality concentrations that are lower than had previously been observed
or that are observed in areas that would likely meet the current suite of PM2.s
standards?
In focusing our attention on whether the available evidence supports consideration of
standards that are more protective than the current suite of PM2.5 standards, we first recognize
that the ISA concludes there is no evidence to support the existence of a discernible population
threshold below which effects would not occur (US EPA, 2009a, section 2.4.3). Next, we
consider whether the evidence provides information for health effects associated with air quality
concentrations that are lower than had previously been observed, or if epidemiological studies
have reported effects in areas that would likely have met the current suite of PM2.5 standards.
Associations with Long-term PM^ Exposure
Extended follow-up analyses of the ACS and Harvard Six Cities studies provide
consistent and stronger evidence of an association with mortality at lower air quality
distributions than had previously been observed. The original and reanalysis of the ACS study
reported positive and statistically significant effects associated with a long-term mean PM2.5
concentration of 18.2 |ig/m3 across 50 metropolitan areas for 1979-1983 (Pope et al., 1995;
Krewski et al., 2000). 22 In extended analyses, positive and statistically significant effects of
approximately similar magnitude were associated with declining PM2.5 concentrations, from an
aggregate long-term mean in 58 metropolitan areas of 21.2 |ig/m3 in the original monitoring
period (1979-1983) to 14.0 |ig/m3 for 116 metropolitan areas in the most recent years evaluated
(1999-2000), with an overall average across the two study periods in 51 metropolitan areas of
17.7 |ig/m3 (Pope et al., 2002; Krewski et al., 2009). With regard to the Harvard Six Cities
22 The study periods referred to in this document reflect the years of air quality data that were included in the
analyses, whereas the study periods identified in the ISA reflect the years of health status data that were included.
2-31
-------
Study, the original and reanalysis reported positive and statistically significant effects associated
with a long-term mean PM2 5 concentration of 18.0 ug/m3 for 1980-1985 (Dockery et al., 1993;
Krewski et al., 2000). In an extended follow-up of this study, the aggregate long-term mean
concentration across all years evaluated was 16.4 ug/m3 for 1980-198823 (Laden et al., 2006). In
an additional analysis of the extended follow-up of the Harvard Six Cities Study, investigators
reported the C-R relationship was linear and "clearly continuing below the level" of the current
annual standard (US EPA, 2009a, p. 7-92; Schwartz et al., 2008).
We then consider new cohort studies that provide evidence of mortality associated with
air quality distributions that are generally lower than those reported in the ACS and Harvard Six
Cities studies, with effect estimates that were similar or greater in magnitude. The WHI study
reported positive and most often statistically significant associations between long-term PM2 5
exposure and cardiovascular-related mortality, with much larger relative risk estimates than in
the ACS and Harvard Six Cities studies, as well as morbidity effects at an aggregate long-term
mean PM2.5 concentration of 12.9 ug/m3 for 2000 (Miller et al., 2007).24 Using the Medicare
cohort, Eftim et al. (2008) reported somewhat higher effect estimates than in the ACS and
Harvard Six Cities studies with aggregate long-term mean concentrations of 13.6 ug/m3 and 14.1
ug/m3, respectively, for 2000-2002. The MCAPS reported associations between long-term PM2 5
exposure and mortality for the eastern region of the U.S. at an aggregated long-term PM2 5
median concentration of 14.0 ug/m3, although no association was reported for the western region
with an aggregated long-term PM2.5 median concentration of 13.1 ug/m3 (US EPA, 2009a, p. 7-
88; Zeger et al., 2008).25 Premature mortality in children reported in a national infant mortality
study as well as mortality in a cystic fibrosis cohort including both children and adults reported
positive but statistically nonsignificant effects associated with long-term aggregate mean
23 Aggregate mean concentration provided by study author (personal communication from Dr. Francine Laden,
2009).
24 Miller et al. (2007) reported a long-term mean PM25 concentration across study areas of 13.5 ug/m3. This
concentration was presented in the ISA (US EPA, 2009a) and discussed in the second draft PA (US EPA, 2010f). In
response to a request from EPA for additional information on the air quality data used in selected epidemiological
studies (Hassett-Sipple and Stanek, 2009), study investigators provided updated air quality data for the study period.
The updated long-term mean PM2 5 concentration provided by the study authors was 12.9 ug/m3 (personal
communication from Cynthia Curl, 2009; Stanek et al., 2010). We note that this updated long-term mean
concentration matches the composite monitor approach annual mean calculated by EPA for the year of air quality
data (i.e., 2000) considered by the study investigators (Hassett-Sipple et al., 2010, Attachment A, p. 6). Staff
concludes it is most appropriate to include the updated air quality data in this final document.
25 Zeger et al. (2008) also reported positive and statistically significant effects for the central region, with an
aggregate long-term mean PM2 5 concentration of 10.7 ug/m3. However, in contrast to the eastern and western risk
estimates, the central risk estimate increased with adjustment for COPD (used as a proxy for smoking status). Due
to the potential for confounding bias influencing the risk estimate for the central region, we have not focused on the
results reported in the central region to inform the adequacy of the current suite of standards or alternative annual
standard levels discussed in section 2.3.4.1.
2-32
-------
concentrations of 14.8 ug/m3 and 13.7 ug/m3, respectively (Woodruff et al., 2008; Goss et al.,
2004).
With respect to respiratory morbidity effects associated with long-term PM2 5 exposure,
the across-city mean of 2-week average PM2.5 concentrations reported in the initial Southern
California CHS was approximately 15.1 |ig/m3 (Peters et al., 1999). These results were found to
be consistent with results of cross-sectional analyses of the 24-Cities Study (Dockery et al.,
1996; Raizenne et al., 1996), which reported a long-term cross-city mean PM2 5 concentration of
14.5 |ig/m3. In this review, extended analysis of the Southern California CHS provided stronger
evidence of PM2. 5 -related respiratory effects, at lower air quality concentrations than had
previously been reported, with a four-year aggregate mean concentration of 13.8 |ig/m3 across
the 12 study communities (McConnell et al., 2003; Gauderman et al., 2004, US EPA, 2009a,
Figure 7-4).
Broadening our consideration to effects for which evidence is suggestive of a causal
relationship, we note a limited number of birth outcome studies that reported positive and
statistically significant effects related to aggregate long-term mean PM2.5 concentrations of
approximately 12 |ig/m3 (US EPA, 2009a, Table 7-5; Bell et al., 2007; Liu et al., 2007; Parker et
al, 2005). In contrast, Parker and Woodruff (2008) reported no overall association with birth
weight with an aggregate long-term mean concentration of 13.5 |ig/m3 (US EPA, 2009a, section
7.4.1.1).
Collectively, the currently available evidence provides support for associations between
long-term PM2 5 exposure and mortality and morbidity effects that extend to air quality
concentrations that are lower than had previously been observed, with aggregate long-term mean
PM2.s concentrations extending to well below the level of the current annual standard. These
studies evaluated a broader range of health outcomes in the general population and in susceptible
populations than were considered in the last review, and include extended follow-up for
prospective epidemiological studies that were important in the last review as well as additional
evidence in important new cohorts.
Associations with Short-term PM^ Exposure
In considering long-term average ambient concentrations from multi-city, short-term
PM2 5 exposure studies, a 12-cities Canadian study (Burnett et al 2004; aggregate long-term
mean PM2.5 concentration of 12.8 |ig/m3) provided evidence of an association between short-
term PM2.s exposure and mortality at lower air quality distributions than had previously been
observed in an 8-cities Canadian study (Burnett and Goldberg, 2003; aggregate long-term mean
PM2.5 concentration of 13.3 |ig/m3). In a multi-city time-series analysis of 1 12 U.S. cities,
Zanobetti and Schwartz (2009) reported a positive and statistically significant association with
2-33
-------
all-cause, cardiovascular-related (e.g., MI, stroke), and respiratory-related mortality and short-
term PM2.5 exposure, in which the aggregate long-term mean PM2.5 concentration was 13.2
|ig/m3 (US EPA, 2009a, Figure 6-24). Furthermore, city-specific effect estimates in Zanobetti
and Schwartz (2009) indicate the association between short-term exposure to PM2.5 and total
mortality and cardiovascular- and respiratory-related mortality is consistently positive for an
overwhelming majority (99%) of the 112 cities across a wide range of air quality concentrations
(ranging from 6.6 |ig/m3to 24.7 |ig/m3; US EPA, 2009a, Figure 6-24, p. 6-178 to 179). We note
that for all-cause mortality, city-specific effect estimates were statistically significant for 55% of
the 112 cities, with long-term city-mean PM2.5 concentrations ranging from 7.8 |ig/m3 to 18.7
|ig/m3 and 24-hour PM2.5 city-mean 98th percentile concentrations ranging from 18.4 to 64.9
|ig/m3 (personal communication with Dr. Antonella Zanobetti, 2009).26
With regard to cardiovascular and respiratory morbidity effects, in the first analysis of the
MCAPS cohort conducted by Dominici et al. (2006a) across 204 U.S. counties, investigators
reported a statistically significant association with hospitalizations for cardiovascular and
respiratory diseases and short-term PM2.5 exposure, in which the aggregate long-term mean
PM2.5 concentration was 13.4 |ig/m3. Furthermore, a sub-analysis restricted to days with 24-hour
average concentrations of PM2.5 at or below 35 |ig/m3 indicated that, in spite of a reduced
statistical power from a smaller number of study days, statistically significant associations were
still observed between short-term exposure to PM2 5 and hospital admissions for cardiovascular
and respiratory diseases (Dominici, 2006b).27 In an extended analysis of the MCAPS study, Bell
et al. (2008) reported a positive and statistically significant increase in cardiovascular
hospitalizations associated with short-term PM2 5 exposure, in which the aggregate long-term
mean PM2.5 concentration was 12.9 |ig/m3. These results, along with the observation that
approximately 50% of the 204 county-specific mean 98th percentile PM2 5 concentrations in
Dominici et al. (2006a) aggregated across all years were below the 24-hour standard of 35
|ig/m3, suggests that the overall health effects observed across the U.S. are not primarily driven
by the higher end of the PM2 5 air quality distribution (Bell, 2009, personal communication from
Dr. Michelle Bell regarding air quality data for Bell et al., 2008 and Dominici et al., 2006a).
In considering single-city short-term PM2 5 exposure studies that were conducted in areas
that would likely have met the current suite of standards, the following studies reported positive
and statistically significant associations: three studies noted in the last review, Mar et al. (2003)
26 Single-city Bayes-adjusted effect estimates for the 112 cities analyzed in Zanobetti and Schwartz (2009) were
provided by the study authors (personal communication with Dr. Antonella Zanobetti, 2009; see also US EPA,
2009a, Figure 6-24).
27 This sub-analysis was not included in the original publication (Dominici et al., 2006a). Authors provided sub-
analysis results for the Administrator's consideration as a letter to the docket following publication of the proposed
rule in January 2006 (personal communication with Dr. Francesca Dominici, 2006b).
2-34
-------
reported a positive and statistically significant association for premature mortality in Phoenix and
Peters et al. (2001) and Delfmo et al. (1997) reported an association with short-term exposure to
PM2 5 and MI- and respiratory-related hospital admissions in Boston and Montreal, Canada,
respectively, and one more recent study, Mar et al. (2004) reported an association for short-term
PM2.5 exposures in relation to respiratory symptoms among children in Spokane. Single-city
studies that reported positive but statistically non-significant associations for cardiovascular and
respiratory endpoints include a number of studies conducted in Saint John (Steib et al., 2000),
Phoenix (Wilson et al., 2007), Denver (Rabinovitch et al., 2006), Edmonton (Villeneuve et al.,
2006), and Nueces County, TX (Lisabeth et al., 2008). Other single-city short-term PM2.5
exposure analyses reported null findings for cardiovascular and respiratory morbidity effects in
association with short-term exposure to PM2 5 in areas that would likely have met the current
suite of standards, including Spokane (i.e., respiratory symptoms in adults, Mar et al., 2004;
Slaughter et al. 2005), Denver (Rabinovitch et al., 2004), and Edmonton (Villeneuve et al.,
2006). In light of the mixed findings reported in single-city studies, particularly for studies
conducted in areas such as Phoenix, Denver, and Edmonton that report both positive and null
findings, we place comparatively greater weight on the results from multi-city studies in
considering the adequacy of the current suite of standards.
Collectively, the findings from multi-city and single-city short-term PM2.5 exposure
studies provide evidence of PM2 5-associated health effects occurring in areas that would likely
have met the current suite of PM2.5 standards. These findings are further bolstered by evidence
of statistically significant PM2 5 associated health effects occurring in analyses restricted to days
in which 24-hour average PM2 5 concentrations were below 35 |ig/m3 (Dominici, 2006b).
Summary
In evaluating the currently available scientific evidence, we conclude that the evidence
from long and short-term PM2.5 exposure studies clearly calls into question whether the current
suite of primary PM2.5 standards protects public health with an adequate margin of safety from
effects associated with long- and short-term exposures to PM2 5. We also conclude that this
evidence provides strong support for considering fine particle standards that would afford
increased protection beyond that afforded by the current annual and 24-hour PM2 5 standards.
More protective standards would reflect the substantially stronger and broader body of evidence
for mortality and cardiovascular-related and respiratory-related morbidity effects now available
in this review both at lower concentrations of air quality than had previously been observed and
at concentrations allowed by the current suite of PM2.5 standards.
2-35
-------
2.2.2 Risk-based Considerations
Looking beyond evidence-based considerations, staff also has considered the extent to
which health risks estimated to occur upon just meeting the current suite of PM2.5 standards may
be judged to be important from a public health perspective. For this review, we have estimated
risk for a set of health effect endpoints based on a number of selection criteria (US EPA, 2010a,
section 3.3.1). Specifically, we have estimated risks for (1) all-cause, IHD-related,
cardiopulmonary- and lung cancer-related mortality associated with long-term PM2.5 exposure,
(2) non-accidental, cardiovascular-related, and respiratory-related mortality associated with
short-term PM2.5 exposure, and (3) cardiovascular-related and respiratory-related hospital
admissions and asthma-related emergency department visits associated with short-term PM2.5
exposure. In the discussion below, we focus on cardiovascular-related endpoints, since the
causal relationship for these endpoints based on the currently available scientific evidence as
assessed in the ISA is the strongest of the endpoints considered. The estimated risks for the
broader set of health effect endpoints modeled are included in the RA (US EPA, 2010a).
As discussed below, three factors figure prominently in the interpretation of the risk
estimates associated with simulating just meeting the current suite of standards, including: (1)
the importance of changes in annual mean PM2.5 concentrations for a specific study area in
estimating changes in risks related to both long- and short-term exposures associated with recent
air quality conditions and air quality simulated to just meet the current suite of PM2 5 standards;
(2) the ratio of peak- to-mean ambient PM2.5 concentrations in a study area; and (3) the spatial
pattern of ambient PM2.5 reductions that result from using different approaches to simulate just
meeting the current standard levels (i.e., rollback approaches). The latter two factors are
interrelated and influence the degree of risk reduction estimated under the current suite of
standards.
The magnitude of both long- and short-term exposure-related risk estimated to remain
upon just meeting the current suite of standards is strongly associated with the simulated change
in annual mean PM2 5 concentrations. The role of annual mean PM2 5 concentrations in driving
long-term exposure-related risk estimates is intuitive given that risks are modeled using the
annual mean air quality metric.28 The fact that short-term exposure-related risk estimates are
also driven by changes in long-term mean PM2 5 concentrations is less intuitive, since changes in
As noted in section 3.2.1 of theRA (U.S.EPA, 2010a), estimates of long-term exposure-related mortality are
actually based on an annual mean PM2 5 concentration that is the average across monitors in a study area (i.e., based
on the composite monitor distribution). Therefore, in considering changes in long-term exposure-related mortality,
it is most appropriate to compare composite monitor estimates generated for a study area under each alternative suite
of standards considered. The annual mean at the highest reporting monitor (i.e., based on the maximum monitor
distribution) for a study area is the annual design value. The annual design value is used to determine the percent
reduction in PM2 5 concentrations required to meet a particular standard. Both types of air quality estimates are
provided in Table 3-4 of the RA and both are referenced in this discussion of core risk estimates, as appropriate.
2-36
-------
mean 24-hour PM2.5 concentrations are used to estimate changes in risk for this time period.29
Analyses show that short-term exposure-related risks are not primarily driven by the small
number of days with PM2 5 concentrations in the upper tail of the air quality distribution, but
rather by the large number of days with PM2.5 concentrations at and around the mean of the
distribution (US EPA, 2010a, section 3.1.2.2). Consequently, changes in annual mean PM2s
concentrations are related, to a large extent, to changes in short-term exposure-related risk.
Therefore, we focus on changes in annual mean PM2.5 concentrations to inform our
understanding of patterns of both long- and short-term exposure-related risk estimates across the
set of urban study areas evaluated in the quantitative risk assessment.
The ratio of peak-to-mean ambient PM2 5 concentrations within a study area informs the
type of rollback approach used to simulate just meeting the current suite of standards to
determine the magnitude of the reduction in annual mean PM2.5 concentrations for that study area
and consequently the degree of risk reduction.30 For example, study areas with relatively high
peak-to-mean ratios are likely to have greater estimated risk reductions for the current suite of
standards (depending on the combination of 24-hour and annual design values), and such
locations can be especially sensitive to the type of rollback approach used, with the proportional
rollback approach resulting in notably greater estimated risk reduction compared with the
locally-focused rollback approach. In contrast, study areas with lower peak-to-mean ratios
typically experience greater simulated risk reductions when simulating just meeting the current
annual-standard level than with simulating just meeting the current 24-hour standard level (again
depending on the combination of 24-hour and annual design values). In addition, the type of
rollback approach used will tend to have less of an impact on the magnitude of risk reductions
for study areas with lower peak-to-mean ratios. Rigorous consideration of these two factors,
allowed us to better understand the nature and pattern of estimated risk reductions and risk
remaining upon simulation of just meeting the current suite of standards across the urban study
areas (seeU.S.EPA, 2010a, section 5.2.1).
We have considered a series of questions to inform our understanding of the adequacy of
the current suite of fine particle standards based on insights obtained from the quantitative risk
assessment. We begin by considering the overall confidence associated with the quantitative risk
assessment and the degree to which the set of urban study areas analyzed is representative of
29 Estimates of short-term PM2 5 expo sure-related mortality and morbidity are based on composite monitor 24-hour
PM2 5 concentrations. However, similar to the case with long-term exposure-related mortality, under the current
rules, it is the 98th percentile 24-hour concentration estimated at the maximum monitor (the 24-hour design value)
that will determine the degree of reduction required to meet a given 24-hour standard.
30 The peak-to-mean ratio of ambient PM2 5 concentrations also has a direct bearing on whether the 24-hour or
annual standard will be the controlling standard for a particular study area, with higher peak-to-mean ratios
generally being associated with locations where the 24-hour standard is likely the controlling standard.
2-37
-------
urban areas across the U.S. We then consider the nature and magnitude of risk estimated to
remain based on simulating just meeting the current suite of standards.
• What is the level of confidence associated with risk estimates generated for simulating
just meeting the current suite of PM2.s standards?
A number of design elements were included in the quantitative risk assessment to
increase the overall confidence in the risk estimates generated for the 15 urban study areas.
These elements included: (1) use of a deliberative process for specifying components of the risk
model that reflects consideration of the latest research on PM2.5 exposure and risk (US EPA,
2010a, section 5.1.1), (2) integration of key sources of variability into the design as well as the
interpretation of risk estimates (U.S.EPA, 2010a, section 5.1.2), (3) assessment of the degree to
which the urban study areas are representative of areas in the U.S. experiencing higher PM2.5-
related risk (U.S.EPA, 2010a, section 5.1.3), and (4) identification and assessment of important
sources of uncertainty and the impact of these uncertainties on the core risk estimates (U.S.EPA,
2010a, section 5.1.4).31 Two additional analyses examined potential bias and overall confidence
in the risk estimates. The first analysis explored potential bias in the core risk estimates by
considering a set of alternative reasonable risk estimates generated as part of a sensitivity
analysis. The second analysis compared the annual mean PM2.5 concentrations associated with
simulating just meeting the current suite of standards with the air quality distribution used in
deriving the C-R functions applied in modeling mortality risk.32 Greater confidence is associated
with risk estimates based on simulated annual mean PM2.5 concentrations that are within the
region of the air quality distribution used in deriving the C-R functions where the bulk of the
data reside (e.g., within one standard deviation (SD) around the mean). Each of the design
elements listed above together with the two additional analyses is discussed below.
Staff used a deliberative process to specify each of the key analytical elements
comprising the core risk model, including: selection of urban study areas; selection of health
endpoints, including specification of the C-R functions to use in modeling those endpoints; and
choice of rollback approach used to simulate just meeting the current suite of standards. This
deliberative process involved rigorous review of the currently 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 quantitative risk assessment (US EPA,
2010a, section 5.1.1).33 The application of this deliberative process increases our overall
31 The "core" risk estimates produced in this assessment refer to those generated using the combination of modeling
elements and input data sets in which we had the highest confidence. Alternative modeling elements were included
as part of the sensitivity analyses.
32 This analysis also considered simulations of alternative standard levels as discussed in section 2.3.4.2.
33 In addition, as discussed in section 1.2.4, the quantitative risk assessment design reflects consideration of CASAC
and public comments on the Scope and Methods Plan and two draft assessment documents.
2-38
-------
confidence in the risk estimates by insuring 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 external experts on PM exposure and risk through CASAC and public reviews.
We considered key sources of variability that can impact the nature and magnitude of
risks associated with the current standard levels across the urban study areas. These sources of
uncertainty include those that contribute to differences in risk across urban study areas, but do
not directly affect the degree of risk reduction associated with the simulation of the current
standard levels (e.g., differences in baseline incidence rates, demographics and population
behavior). We also focused on factors that not only introduce variability into risk estimates
across study areas, but also play an important role in determining the magnitude of risk
reductions upon simulation of current standard levels (e.g., peak-to-mean ratios of ambient PM2 5
concentrations within individual urban study areas and the nature of the rollback approach used
to simulate just meeting the current standards - see earlier discussion).
Single and multi-factor sensitivity analyses were combined with a qualitative analysis to
assess the impact of potential sources of uncertainty on the core risk estimates. The qualitative
uncertainty analysis supplemented the quantitative sensitivity analyses by allowing coverage for
sources of uncertainty that could not be readily included in the sensitivity analysis (US EPA,
2010a, section 3.5.3). The quantitative sensitivity analyses informed our understanding of
sources of uncertainty that may have a moderate to large impact on the core risk estimates. With
respect to the long-term exposure-related mortality risk estimates, the most important sources of
uncertainty identified in the quantitative sensitivity analyses included: selection of C-R
functions;34 modeling risk down to policy-relevant background (PRB) versus lowest measured
level (LML); and the choice of rollback approach used. With regard to the qualitative analysis of
uncertainty, the following sources were identified as potentially having a large impact on core
risk estimates for the long-term exposure-related mortality: characterization of inter-urban
population exposures; impact of historical air quality; and potential variation in effect estimates
reflecting differences in PM2.5 composition. Together, the qualitative analysis of uncertainty and
quantitative sensitivity analyses provided us with a comprehensive understanding of which
sources of uncertainty could have a significant impact on the core risk estimates. This
34 In the case of long-term exposure-related mortality, we considered both alternative C-R functions from the
epidemiological study providing the C-R function used in the core analysis (i.e., alternative functions obtained from
the Krewski et al. (2009) study involving the ACS dataset) as well as alternative C-R functions identified in other
studies (i.e., C-R functions obtained from Krewski et al. (2000) based on a reanalysis of the Harvard Six Cities
study).
2-39
-------
information proved useful in interpreting core risk estimates and increased our overall
confidence in the analysis.35
In addition to identifying sources of uncertainty with a moderate to large impact on the
core risk estimates, the single and multi-element sensitivity analyses also produced a set of
reasonable alternative risk estimates that allowed us to place the results of the core analysis in
context with regard to uncertainty and potential bias.36 Most of the alternative model
specifications supported by the currently available scientific information produced risk estimates
that are higher (by up to a factor of 2 to 3) than the core risk estimates. This was not unexpected.
The C-R functions used in the core analysis for estimating mortality risks associated with long-
term PM2.5 exposures were selected from the extended analysis of the ACS study (Krewski et al.,
2009). The C-R functions used in the sensitivity analysis were from the reanalysis and validation
of the Harvard Six Cities study (Krewski et al., 2000). In generalizing the results of the extended
analyses of the ACS and Harvard Six Cities studies across the broader national population, we
recognize differences in the underlying populations enrolled in these long-term cohort studies,
specifically related to SES, a factor in considering impacts on susceptible populations. As noted
in the last review, the ACS study population has a higher SES status (e.g., educational status)
relative to the Harvard Six Cities study population (12% versus 28% of the cohort had less than a
high school education, respectively) (US EPA, 2004a, p. 8-118). The Harvard Six Cities cohort
may provide a more representative sample of the broader national population than the ACS
cohort.
As discussed above, lower SES groups have been identified as a susceptible population.
Therefore, use of effect estimates reported in the ACS study which does not provide
representative coverage for lower-SES groups, may result in risk estimates that are biased low.
In contrast, risk estimates developed in the sensitivity analysis based on the Harvard Six Cities
study data set provide better coverage for lower SES populations but give greater weight to
eastern and Midwestern populations and, therefore, result in higher risk estimates (US EPA,
2010a, section 5.1.5).
While being mindful that the use of C-R functions from Krewski et al. (2009) introduces
potential for low bias in the core risk estimates, we also recognize many strengths of this study
35 Given increased emphasis placed in this analysis on long-term exposure-related mortality, the uncertainty analyses
completed for this health endpoint category were more comprehensive than those conducted for analyses of short-
term exposure-related mortality and morbidity. This reflects, to some extent, limitations in the epidemiological data
available for addressing uncertainty in the latter categories (U.S.EPA 2010a, section 3.5.4.2).
36 The alternative set of reasonable risk estimates were based on alternative model specifications compared to the
core risk model. These alternative reasonable risk estimates were only generated for long-term exposure-related
mortality and not for any of the short-term exposure-related mortality or morbidity endpoints. Consequently,
consideration of overall confidence (and potential bias) in the core risk estimates based on consideration for these
alternative risk estimates is limited to estimates of mortality associated with long-term PM2 5 exposures.
2-40
-------
and reasons for its continued use for generating the core risk estimates, including: consideration
of a large number of metropolitan statistical areas (MS As), inclusion of two time periods for the
air quality data which allowed us to consider different exposure windows; and analysis of a wide
range of C-R function models. Therefore, we concluded that C-R functions obtained from this
study had the greatest overall support and should be used in the core risk model. Consideration
of the alternative set of reasonable risk estimates provided several observations relevant to the
interpretation of the core risk estimates including: (a) the core estimates are unlikely to
underestimate risk and (b) the degree of potential bias in the core risk estimates could range up to
at least a factor of 2-3 higher.37
In considering the overall confidence in the core risk estimates, we have compared the
PM2 5 concentrations simulated under the current suite of standard levels across the urban study
areas to the distribution of PM2.5 concentrations used in deriving the C-R functions used for
long-term exposure-related mortality (as presented in Krewski et al., 2009). Specifically, this
assessment compared the composite monitor annual mean PM2.5 concentrations used in modeling
long-term exposure-related mortality risk in the core analysis to the distribution of annual mean
PM2 5 concentrations from the 1999-2000 ACS exposure period.38 Generally, when composite
monitor annual mean concentrations were within one SD of the long-term mean concentration
for the most recent years of air quality considered in the ACS dataset (i.e., in the range of 14+/-3
|ig/m3), we had relatively high confidence in the risk estimates, since they were based on PM2 5
concentrations that roughly matched those used in deriving the C-R functions. However, as
composite monitor annual mean PM2 5 concentrations extend below this range, our confidence in
the risk estimates decreased, with our confidence being significantly reduced when composite
monitor annual mean concentrations approached the LML of the ACS data set (i.e., 5.8 |ig/m3).
37 We note that these findings regarding potential bias in the core risk estimates were based on modeling PM2 5-
attributable IHD and all-cause mortality associated with long-term PM2 5 exposure for the current suite of standards.
However, we would expect these observations regarding overall confidence in the core risk estimates to hold for
other long-term exposure-related mortality endpoints modeled in the quantitative risk assessment for both the
alternative annual and 24-hour standard levels considered in section 2.3.4.2. Furthermore, given increased emphasis
placed in this analysis on long-term exposure-related mortality, as noted earlier, the uncertainty analyses completed
for this health endpoint category are more comprehensive than those conducted for short-term exposure-related
mortality and morbidity effects, which to some extent reflects limitations in study data available for addressing
uncertainty in the latter category. Therefore, an alternative set of reasonable risk estimates was not generated to
supplement core risk estimates generated for short-term PM2 5 exposure.
38 As discussed in sections 3.3.3 and 4.0 of the RA (U.S.EPA, 2010a), each category of long-term exposure-related
mortality was estimated using separate C-R functions derived from the 1979-1983 and 1999-2000 ACS monitoring
periods. For purposes of comparing composite monitor annual mean PM2 5 concentrations to the ACS data sets used
in deriving the C-R functions, we focused on the later monitoring period (1999-2000), since ambient PM2 5
concentrations from this period more closely matched those associated with the study areas in our consideration of
recent air quality conditions (2005-2007). The 1999-2000 ACS monitoring period had a mean PM2 5 concentration
of 14.0 ug/m3, a SD of 3.0 ug/m3 and an LML of 5.8 ug/m3 (see Table 1 in Krewski et al., 2009).
2-41
-------
• How representative is the set of urban study areas for the broader set of urban areas in
the U.S. expected to experience elevated risk from ambient PM2.s exposure?
The goal in selecting urban study areas was to provide coverage for the range of larger
urban areas in the U.S. expected to experience relatively elevated risk due to ambient PM2.5
exposure and other factors associated with PM2.5-relatedrisk (e.g., elevated baseline incidence
rates for relevant health endpoints, relatively larger susceptible populations). As part of
considering our overall confidence in the quantitative risk assessment, we assessed the
representativeness of the 15 urban study areas in the broader national context. Three separate
analyses were used to explore representativeness:
• A comparison of PM25-risk-related attributes of the 15 urban study areas against
national distributions of these same attributes suggested that the urban study areas
likely reflect the distribution of risk for the nation, with the potential for better
characterization at the higher end of that distribution (US EPA, 2010a, section
4.4.1).39
• An analysis of the distribution of U.S. counties included in a national-scale mortality
analysis suggested that counties associated with the 15 urban study areas are likely to
experience elevated PM2.5-related risk (US EPA, 2010a, section 4.4.2).
• An evaluation of the mix of design values across the 15 urban study areas as
contrasted with design values for the broader set of urban study areas in the U.S. This
analysis suggested that (a) the 15 urban study areas reasonably captured the key
groupings of urban areas in the U.S. likely to experience elevated risk due to PM2.5
exposures and (b) we have included study areas likely to experience relatively greater
degrees of PM2.5-related risk (US EPA, 2010a, section 4.5.1).
Based on these analyses, we conclude that these study areas are generally representative
of urban areas in the U.S. likely to experience relatively elevated levels of risk related to ambient
PM2.5 exposure.
• What is the nature and magnitude of the long-term and short-term exposure-related
risks remaining upon just meeting the current suite of PMi.s standards?
In considering PM2.s-related risks likely to remain upon just meeting the current PM2.5
annual and 24-hour standards in the 15 urban study areas included in the quantitative RA, we
focus on the 13 areas that would likely not have met the current standards based on recent air
quality (2005-2007). These 13 areas have annual and/or 24-hour design values that are above the
39 This representativeness analysis also showed that the urban study areas do not capture areas with the highest
baseline morality risks or the oldest populations (both of which can result in higher PM2 5-related mortality
estimates). However, some of the areas with the highest values for these attributes have relatively low PM2 5
concentrations (e.g., urban areas in Florida) and consequently failure to include these areas in the set of urban study
areas is unlikely to exclude high PM2 5-risk locations.
2-42
-------
levels of the current standards (see Tables 2-2 and 2-3).40 Based on the core risk estimates for
these areas, using the proportional rollback approach, we make the following key observations
regarding the magnitude of risk remaining upon simulation of just meeting the current suite of
standards:
• Long-term exposure-related mortality risk remaining: IHD-related mortality
attributable to long-term PM2.5 exposure was estimated to range from less than 100 to
approximately 2,000 cases per year. The variability in these estimates reflects, to a
great extent, differences in the size of study area populations. These estimates
represent from 4 to 17% of all IHD-related mortality in a given year for the urban
study areas, representing a measure of risk that takes into account differences in
population size and baseline mortality rates (see Table 2-2).
• Short-term exposure-related mortality risk remaining: Cardiovascular (CV)-related
mortality associated with short-term PM2.5 exposure was estimated to range from less
than 10 to 500 cases per year. These estimates represent approximately 1 to 2% of
total CV-related mortality in a given year for the urban study areas (see Table 2-3).
• Short-term exposure-related morbidity risk remaining: CV-related hospitalizations
were estimated to range from approximately 10 to 800 cases per year across the study
areas, which is approximately equivalent to less than 1% of total CV-related
hospitalizations (see Table 2-3).
Although long- and short-term exposure-related mortality rates have similar patterns in
terms of the subset of urban study areas experiencing risk reductions for the current suite of
standard levels, the magnitude of risk remaining is substantially lower for short-term exposure-
related mortality. These findings were expected, since, as noted earlier, changes in annual mean
PM2.5 concentrations were expected to drive both long- and short-term exposure-related risk,
resulting in similar overall patterns in risk reduction for both exposure periods (in terms of the
subset of urban study areas experiencing risk reductions). We note, however, that the variability
in the effect estimates used to model short-term exposure-related health endpoints across urban
study areas introduced additional variation into the pattern of risk reduction across study areas.
40 Of the 15 study areas, only Dallas and Phoenix have both annual and 24-hour design values below the levels of
the current standards based on 2005-2007 air quality data.
2-43
-------
Table 2-2. Estimated Incidence and Percent of Total Annual Incidence Associated with
Long-term PM2.s Exposure Based on Simulation of the Current Suite of Standards (for
IHD Mortality based on 2007 PM2.5 Concentrations)
Urban Study Area
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit, MI
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, MO
Tacoma, WA
Incidence of Ischemic Heart Disease
Mortality Associated with Long-term
Exposure to PM2.53
Exposure Period:
1979-1983
220
(180 - 258)
297
(243 - 349)
131
(107 - 154)
195
(159-230)
377
(308 - 445)
77
(63 - 92)
344
(281-405)
860
(701 - 1018)
1755
(1435 - 2070)
261
(214 - 308)
317
(258 - 374)
256
(209 - 302)
15
(12-18)
446
(365 - 525)
38
(31-46)
Exposure Period:
1999-2000
277
(227 - 324)
374
(307 - 440)
165
(135 - 194)
247
(202-291)
478
(390 - 563)
98
(80-116)
434
(355-511)
1094
(890 - 1296)
2222
(1814-2620)
330
(270 - 389)
402
(327 - 476)
324
(264 - 382)
19
(16-23)
563
(461-662)
49
(40 - 58)
Percent of Incidence of Ischemic Heart
Disease Mortality Associated with Long-
term Exposure to PM2.53
Exposure Period:
1979-1983
13.2%
(10.9% - 15.5%)
11.7%
(9.6% -13.7%)
10.9%
(8.9% - 12.9%)
9%
(7.3% - 10.6%)
9.1%
(7.4% - 10.7%)
6.7%
(5.5% - 8%)
10.7%
(8.8% - 12.6%)
6.1%
(4.9% - 7.2%)
9.3%
(7.6% -11%)
10.5%
(8.6% - 12.3%)
6.7%
(5.5% - 7.9%)
9.3%
(7.6% -11%)
2.9%
(2.4% -3. 4%)
11.2%
(9.2% -13.2%)
3.7%
(3% - 4.4%)
Exposure Period:
1999-2000
16.7%
(13.7% -19.5%)
14.7%
(12.1% -17.3%)
13.8%
(11.3% -16.2%)
11.4%
(9.3% -13.4%)
11.5%
(9.4% -13.5%)
8.5%
(7% -10.1%)
13.6%
(11.1% -16%)
7.7%
(6.3% -9.1%)
11.8%
(9.6% -13.9%)
13.2%
(10.8% - 15.6%)
8.5%
(6.9% -10.1%)
11.8%
(9.6% -13.9%)
3.7%
(3% - 4.4%)
14.2%
(11.6% -16.7%)
4.7%
(3. 8% -5.6%)
1 The current primary PM25 standards include an annual standard set at 15 ug/m3 and a 24-hour standard set at
35 ug/m3.
2 Numbers rounded to the nearest whole number. Numbers in parentheses are 95% confidence or credible
intervals based on statistical uncertainty surrounding the PM coefficient.
3 Estimates based on Krewski et al. (2009), using ambient PM2 5 concentrations from 1979 - 1983 and from 1999-
2000, respectively. Estimated incidence is presented for 30 years of age and older within each urban study area.
2-44
-------
Table 2-3. Estimated Incidence and Percent of Total Annual Incidence Associated with
Short-term PM2.s Exposure Based on Simulation of the Current Suite of Standards (CV
mortality and hospital admissions based on 2007 PM2.s concentrations)
Urban Study Area
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit, MI
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, MO
Tacoma, WA
Incidence of
Cardiovascular
Mortality
Associated with
Short-term
Exposure to PM2.53
32
(-33 - 95)
62
(-4 - 126)
-1
(-42 - 40)
29
(-19-76)
60
(-8 - 127)
12
(-9 - 33)
46
(-31-122)
-30
(-132 - 72)
473
(276 - 668)
84
(22 - 145)
84
(-4 - 170)
43
(-9 - 93)
9
(-2 - 20)
106
(24 - 187)
11
(-6 - 27)
Incidence of
Cardiovascular
Hospitalizations
Associated with
Short-term
Exposure to PM2.54
41
(-27 - 109)
216
(159-273)
16
(-11-43)
28
(-18-73)
233
(171-295)
23
(0 - 46)
56
(-37 - 149)
258
(3-511)
752
(552-951)
203
(149 - 257)
108
(1-215)
140
(103 - 177)
9
(0 - 18)
178
(131-225)
19
(-46 - 82)
Percent of
Incidence of
Cardiovascular
Mortality
Associated with
Short-term
Exposure to PM2.53
0.8%
(-0.8% - 2.4%)
1.6%
(-0.1% -3. 2%)
0%
(-1.5% -1.5%)
0.8%
(-0.5% - 2.2%)
1%
(-0.1% -2.2%)
0.7%
(-0.5% - 2%)
0.9%
(-0.6% - 2.4%)
-0.2%
(-0.7% - 0.4%)
2.1%
(1.2% -3%)
2.1%
(0.5% - 3.6%)
1.3%
(-0.1% -2.7%)
1.1%
(-0.2% - 2.3%)
0.8%
(-0.2% -1.7%)
1.9%
(0.4% - 3.3%)
0.7%
(-0.4% -1.8%)
Percent of
Incidence of
Cardiovascular
Hospital
Admissions
Associated with
Short-term
Exposure to PM2.54
0.4%
(-0.2% - 1%)
1.3%
(1%-1.7%)
0.3%
(-0.2% - 0.9%)
0.3%
(-0.2% - 0.7%)
1.1%
(0.8% -1.4%)
0.5%
(0% - 0.9%)
0.3%
(-0.2% - 0.8%)
0.5%
(0% - 0.9%)
1.2%
(0.8% -1.5%)
1.3%
(0.9% -1.6%)
0.5%
(0% - 1%)
1.1%
(0.8% - 1.4%)
0.4%
(0% - 0.7%)
1.3%
(0.9% -1.6%)
0.5%
(-1.3% - 2.3%)
The current primary PM25 standards include an annual standard set at 15 ug/m3 and a 24-hour standard set at
35 ug/m3.
2 Percents rounded to the nearest tenth. Numbers in parentheses are 95% confidence or credible intervals based on
statistical uncertainty surrounding the PM coefficient.
3 Based on location-specific single pollutant concentration-response function estimates from Zanobetti and Schwartz
(2009) that have been "shrunken" towards the appropriate regional means. "Shrunken" coefficient estimates and
their standard errors were provided to EPA by the study authors (personal communication with Dr. Antonella
Zanobetti, 2009).
4 Incidence estimates were calculated using the appropriate regional concentration-response function estimates
reported in Table 2 of Bell et al. (2008). Location-specific C-R function estimates were not available from this
study.
2-45
-------
• Substantial variability exists in the magnitude of risk remaining across urban study
areas: Estimated risks remaining upon just meeting the current suite of standards
vary substantially across study areas, even when considering risks normalized for
differences in population size and baseline incidence rates. This variability is a
consequence of the substantial differences in the annual mean PM2.5 concentrations
across study areas that result from simulating just meeting the current standards. This
is important because, as discussed above, annual mean concentrations are highly
correlated with both long- and short-term exposure-related risk. The variability in
annual mean PM2.5 concentrations occurred primarily in those study areas in which
the 24-hour standard was the "controlling" standard. In such areas, the variability in
estimated risks across study areas was largest when regional patterns of reductions in
PM2.5 concentrations were simulated, using the proportional rollback approach, as
was done in the core analyses. Less variability was observed when more localized
patterns of PM2.5 reductions were simulated using the locally-focused rollback
approach, as was done in a sensitivity analysis. When simulations were done using
the locally-focused rollback approach, estimated risks remaining upon just meeting
the current suite of standards were appreciably larger than those estimated in the core
analysis (US EPA, 2010a, section 4.3.1.1).
• Simulation of just meeting the current suite of standards results in annual mean PM2.s
concentrations well below the current standard for some study areas: In simulating
just meeting the current suite of standards, the resulting composite monitor annual
mean PM2.5 concentrations ranged from about 15 |ig/m3 (for those study areas in
which the annual standard was controlling) down to as low as about 8 |ig/m3 (for
those study areas in which the 24-hour standard was controlling or the annual mean
concentration was well below 15 |ig/m3 based on recent air quality). As discussed
above, as the composite monitor annual mean PM2.5 concentrations used in generating
risk estimates extend below 11.0 |ig/m3 (one SD below the mean for the 1999-2000
ACS monitoring period, Krewski et al., 2009) we have increasingly less confidence in
the risk estimates, with confidence decreasing significantly as composite monitor
concentrations approach the LML for the ACS dataset (5.8 |ig/m3). Typically, for the
15 urban study areas assessed, the locations where the 24-hour standard was the
controlling standard were simulated to have the lowest composite monitor annual
mean PM2 5 concentrations. We observe that all four of the urban study areas with
simulated composite monitor annual mean PM2.5 concentrations below 11 |ig/m3 are
areas where the 24-hour standard is generally controlling (U.S.EPA, 2010a, Table 3-
4). While such locations often are estimated to have the greatest risk reductions, there
is also reduced confidence associated with these risk estimates.
• To what extent are the risks remaining upon simulation of the current suite of
standards important from a public health perspective?
Estimates of long-term exposure-related IHD mortality risk associated with simulating
just meeting the current suite of standards range from less than 100 deaths per year for the urban
study area with the lowest risk to approximately 2,000 deaths per year for the urban study areas
with the greatest risk. Estimates of risk for the urban areas included in the quantitative risk
2-46
-------
assessment suggest that IHD-related mortality associated with long-term PM2.5 exposure would
likely be in a range of thousands of deaths per year on a national scale. Based on these risk
estimates for IHD-related mortality alone, we conclude that risks estimated to remain upon
simulation of just meeting the current suite of standards are important from a public health
standpoint. This reflects consideration of both the severity of the effect and the magnitude of the
effect. We have also estimated risks for long-term exposure related mortality risk related to
cardiopulmonary effects and lung cancer, which increase the total annual incidence of mortality
attributable to long-term PM2.5 exposure (see U.S. EPA, 2010a, section 4.2.1).
In addition to long-term exposure-related mortality, we estimated cardiovascular-related
and respiratory-related mortality risk associated with short-term PM2.5 exposure. We note that
these mortality estimates are up to an order of magnitude smaller than estimates related to long-
term exposure-related mortality.41 As part of the quantitative risk assessment, we also estimated
respiratory and cardiovascular-related hospital admissions as well as asthma-related emergency
department visits associated with short-term exposure to PM2.5, with estimates of cardiovascular-
related and respiratory-related hospital admissions together ranging up to approximately 1,000
admissions per year across the urban study areas, with the estimated incidence of asthma-related
emergency department visits being several fold higher. Further, as discussed in section 2.2.1, we
recognize that the currently available scientific information includes evidence for a broader range
of health endpoints and susceptible populations beyond those included in the quantitative risk
assessment, including lung function growth and respiratory symptoms in children and
reproductive and developmental effects. Taken together, the set of quantitative risk estimates
related to long- and short-term PM2 5 exposure, together with consideration of the health
endpoints which could not be quantified, further strengthen the conclusion that risks estimated to
remain following simulation of just meeting the current suite of PM2.5 standards are important
from a public health perspective, both in terms of severity and magnitude.
2.2.3 CASAC Advice
In our consideration of the adequacy of the current suite of PM2.5 standards, in addition to
the evidence and risk-based information discussed above, we have also considered the advice of
CASAC, based on their review of drafts of the ISA, the RA, and this document, as well as
comments from the public on earlier drafts of this document and the RA. In their comments on
the second draft PA, CASAC stated agreement with EPA staffs conclusion that the "currently
41 Estimates of short-term exposure-related and long-term expo sure-related mortality should not be added because
there is the potential for overlap (i.e., the long-term exposure-related mortality estimate may include some of the
short-term exposure-related signal on a daily basis, aggregated over the year).
2-47
-------
available information clearly calls into question the adequacy of the current standards" (Samet,
2010d, p. i). Further, in an earlier draft of this document, CASAC noted:
With regard to the integration of evidence-based and risk-based considerations, CASAC
concurs with EPA's conclusion that the new data strengthens the evidence available on
associations previously considered in the last round of the assessment of the PM2 5 standard.
CASAC also agrees that there are significant public health consequences at the current levels
of the standard that justify consideration of lowering the PM2.5 NAAQS further (Samet,
2010c, p. 12).
2.2.4 Staff Conclusions on Adequacy of Current Standards
Collectively, taking into consideration the responses to specific questions focused on
different ways to address the adequacy of the current suite of PM2.5 standards, we revisit the
overarching policy question: does the currently available scientific evidence and risk-based
information support or call into question the adequacy of the protection afforded by the current
suite of fine particle standards?
With respect to evidence-based considerations, the currently available evidence provides
stronger evidence beyond what was available in the last review, that associations between short-
and long-term PM2 5 exposures and a broad range of adverse health effects exist. The currently
available information strengthens the associations between PM2 5 and mortality and
cardiovascular-related and respiratory-related morbidity effects observed in the last review. This
information also expands our understanding of a broader range of health outcomes as well as our
understanding of effects in susceptible populations. The currently available evidence provides
support for associations that extend to lower concentrations than what had been observed in the
last review, including at ambient concentrations below the levels of the current standards
providing the basis for consideration of alternative standards that would provide increased
protection beyond that afforded by the current suite of PM2 5 standards.
In relation to risk-based considerations for informing our understanding of the adequacy
of the current fine particle standards, we focus on the estimates of PM2.5-related mortality and
morbidity effects likely to remain upon simulations of just meeting the current standards in a
number of example urban areas. In considering the core risk estimates together with our
understanding of the uncertainties in these estimates, based upon extensive sensitivity analyses,
we conclude that the risks estimated to be associated with just meeting the current standards can
reasonably be judged to be important from a public health perspective. We further conclude that
these estimated risks provide strong support for consideration of alternative standards that would
provide increased protection beyond that afforded by the current PM2 5 standards.
We recognize that important uncertainties and research questions remain when
considering both evidence- and risk-based approaches. Nonetheless, we note that much progress
2-48
-------
has been made in reducing some key uncertainties since the last review, including important
progress in advancing our understanding of potential mechanisms by which ambient PM2.5 is
causally linked with mortality and cardiovascular-related and respiratory-related effects observed
in epidemiological and toxicological studies. Additional information continues to emerge for a
broader range of health effects including reproductive and development effects and more
information is available to understand effects in susceptible populations including children, older
adults, individuals with pre-existing cardiovascular and respiratory disease, persons at lower
SES, and persons with genetic susceptibility.
As was true in the last review, we recognize that as the body of available evidence has
expanded, it has added greatly both to our knowledge of health effects associated with fine
particle exposures, as well as to the complexities inherent in interpreting the evidence in a
policy-relevant context as a basis for setting appropriate standards. In evaluating both evidence-
based and risk-based considerations, along with associated limitations and uncertainties, we
reach the conclusion that the available information clearly calls into question the adequacy of the
current suite of PM2.5 standards and provides strong support for giving consideration to revising
the current suite of standards to provide increased public health protection.
2.3 CONSIDERATION OF ALTERNATIVE STANDARDS
Having reached the conclusion that the currently available scientific evidence calls into
question the adequacy of the current suite of PM2.5 standards, staff considers a second
overarching question:
What alternative suites of fine particle standards are supported by the currently available
scientific evidence and risk-based information, as reflected in the ISA and RA?
To address this overarching question, we have posed a series of more specific questions
to inform decisions regarding the basic elements of the NAAQS: indicator (section 2.3.1),
averaging time (section 2.3.2), form (section 2.3.3), and level (section 2.3.4). These elements are
considered collectively in evaluating the health protection afforded by alternative suites of
standards under consideration. In taking into account the currently available scientific and
technical information, we consider 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 and
RA prepared for this review (US EPA, 2009a; US EPA, 2010a).
2.3.1 Indicator
In initially setting standards for fine particles in 1997, EPA concluded it was more
appropriate to control fine particles as a group, rather than singling out any particular component
or class of fine particles for which only very limited evidence was available. In establishing a
2-49
-------
size-based indicator to distinguish fine particles from particles in the coarse mode, EPA noted
that the available epidemiological studies of fine particles were based largely on PM2.5 and also
considered monitoring technology that was generally available. The selection of a 2.5 jam size
cut 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., especially
when fine particle concentrations and humidity are likely to be high, while recognizing that some
small coarse particles would also be captured by current methods to monitor PM2.5 (62 FR 38666
to 38668, July, 18, 1997). In the last review, based on the same considerations, EPA again
recognized that the available information supported retaining the PM2.5 indicator and remained
too limited to support a distinct standard for any specific PM2.5 component or group of
components associated with any source categories of fine particles (71 FR 61162 to 61164,
October 17, 2006).
• Does the currently available information provide support for the continued use of a
PM2.s mass-based indicator for fine particles?
In this review, epidemiological studies linking cardiovascular and respiratory effects as
well as mortality with long- and short-term fine particle exposures continue to be largely indexed
by PM2.5. Based on the same considerations that informed the last two reviews, summarized
above in section 2.1.1, we again conclude that it is appropriate to retain a PM2.5 indicator to
provide protection associated with long- and short-term exposure to fine particles.
We also look to the expanded body of evidence available in this review to consider
whether there is sufficient evidence to support a separate standard for ultrafine particles (UFPs)42
and whether there is sufficient evidence to establish distinct standards focused on regulating
specific PM2 5 components or group of components associated with any source categories of fine
particles, as addressed below.
• To what extent does the currently available information provide support for
considering a separate indicator for UFPs?
A number of studies available in this review have evaluated potential health effects
associated with short-term exposures to UFPs. As noted in the ISA, the enormous number and
larger, collective surface area of UFPs are important considerations for focusing on this particle
size fraction in assessing potential public health impacts (US EPA, 2009a, p. 6-83).43 Per unit
mass, UFPs may have more opportunity to interact with cell surfaces due to their greater surface
42 Ultrafine particles, generally including particles with a mobility diameter less than or equal to 0.1 um, are emitted
directly to the atmosphere or are formed by nucleation of gaseous constituents in the atmosphere (US EPA, 2009a,
p. 3-3).
43 Particle number is most highly concentrated in the UFP fraction with volume (or mass) most concentrated in the
larger size fractions (US EPA, 2009a, p. 3-2, Figure 3-1).
2-50
-------
area and their greater particle number compared with larger particles (US EPA, 2009a, p. 5-3).
Greater surface area also increases the potential for soluble components (e.g., transition metals,
organics) to adsorb to UFPs and potentially cross cell membranes and epithelial barriers (US
EPA, 2009a, p. 6-83). In addition, evidence available in this review suggests that the ability of
particles to enhance allergic sensitization is associated more strongly with particle number and
surface area than with particle mass (US EPA, 2009a, p. 6-127).44
New evidence, primarily from controlled human exposure and toxicological studies,
expands our understanding of UFP-related cardiovascular and respiratory effects. However, this
evidence is still very limited and largely focused on exposure to diesel exhaust (DE), for which
the ISA concludes it is unclear if the effects observed are due to UFPs, larger particles within the
PM2.5 mixture, or the gaseous components of DE (US EPA, 2009a, p. 2-22). In addition, the ISA
notes uncertainties associated with the controlled human exposure studies as CAP systems have
been shown to modify the composition of UFPs (US EPA, 2009a, p. 2-22, see also section 1.5.3).
Relatively few epidemiological studies have examined potential cardiovascular and respiratory
effects associated with short-term exposures to UFPs. These studies have reported inconsistent
and mixed results (US EPA, 2009a, section 2.3.5).
Collectively, in considering the body of scientific evidence available in this review, the
ISA concludes that the currently available evidence is suggestive of a causal relationship
between short-term exposures to UFPs and cardiovascular and respiratory effects. Furthermore,
the ISA concludes that evidence is inadequate to infer a causal relationship between short-term
exposure to UFPs and mortality as well as long-term exposure to UFPs and all outcomes
evaluated (US EPA, 2009a, sections 2.3.5, 6.2.12.3, 6.3.10.3, 6.5.3.3, 7.2.11.3, 7.3.9, 7.4.3.3,
7.5.4.3, and 7.6.5.3; Table 2-6).
With respect to our understanding of ambient UFP concentrations, at present, there is no
national network of UFP samplers; thus, only episodic and/or site-specific data sets exist (US
EPA, 2009a, p. 2-2).45 Therefore, a national characterization of concentrations, temporal and
spatial patterns, and trends is not possible, and the availability of ambient UFP measurements to
support health studies is extremely limited. In general, measurements of UFPs are highly
dependent on monitor location and, therefore, more subject to exposure error than accumulation
mode particles (US EPA, 2009a, p. 2-22). In addition, UFPs are often monitored based on
numerical concentrations (i.e., particle counts), rather than mass concentrations, as UFP mass in
ambient air is typically very low (US EPA, 2009a, section 3.4.1.4). The UFP number
44 More information on possible modes of action for effects associated with UFP exposures is discussed in sections
5.1 and 5.4 of the ISA (US EPA, 2009a).
45 The ISA contains a review of the current scientific information related to measurements of UFPs (US EPA,
2009a, sections 3.5.1 and 3.5.2).
2-51
-------
concentrations fall off sharply downwind from sources, as UFPs may grow into the accumulation
mode by coagulation or condensation (US EPA, 2009a, p. 3-89). Limited studies of UFP
ambient measurements suggest these particles exhibit a high degree of spatial and temporal
heterogeneity driven primarily by differences in nearby source characteristics (US EPA, 2009a,
p. 3-84). Internal combustion engines and, therefore, roadways are a notable source of UFPs, so
concentrations of UFPs near roadways are generally expected to be elevated (US EPA, 2009a, p.
2-3). Concentrations of UFPs have been reported to drop off much more quickly with distance
from roadways than fine particles (US EPA, 2009a, p. 3-84).
In considering both the currently available health effects evidence and the air quality data
for UFPs, we conclude that this information is still too limited to provide support for
consideration of a distinct PM standard for UFPs.
• To what extent does the currently available information provide support for
considering a separate indicator for a specific PMi.5 component or group of components
associated with any source categories of fine particles? Conversely, to what extent does
the currently available information provide support for eliminating any component or
group of components associated with any source categories from the mix of fine
particles included in the PMi.s indicator?
In addressing the issue of particle composition, the ISA concludes that, "[f]rom a
mechanistic perspective, it is highly plausible that the chemical composition of PM would be a
better predictor of health effects than particle size" (US EPA, 2009a, p. 6-202). Heterogeneity of
ambient concentrations of PM2.5 constituents (e.g., elemental carbon (EC), organic carbon (OC),
sulfates, nitrates) observed in different geographical regions as well as regional heterogeneity in
PM2.5-related health effects reported in a number of epidemiological studies are consistent with
this hypothesis (US EPA, 2009a, section 6.6).
With respect to the availability of ambient measurement data for fine particle components
in this review, there are now more extensive ambient PM2.5 speciation measurement data
available through the CSN than in previous reviews (see section 1.3.2 and Appendix B, section
B. 1.3). Data from the CSN provide further evidence of spatial and seasonal variation in both
PM2.5 mass and composition among cities/regions (US EPA, 2009a, pp. 3-50 to 3-60; Figures 3-
12 to 3-18; Figure 3-47). Some of this variation may be related to regional differences in
meteorology, sources, and topography (US EPA, 2009a, p. 2-3).
The currently available epidemiological, lexicological, and controlled human exposure
studies have evaluated the health effects associated with ambient PM2.5 constituents and
categories of fine particle sources, using a variety of quantitative methods applied to a broad set
of PM2 5 constituents, rather than selecting a few constituents a priori (US EPA, 2009a, p. 2-26).
Epidemiological studies have used measured ambient PM2.5 speciation data, including
monitoring data from the CSN, while all of the controlled human exposure and most of the
2-52
-------
toxicological studies have used CAPs and analyzed the constituents therein (US EPA, 2009a, p.
6-203).46 The CSN provides PM2.5 speciation measurements generally on a one-in-three or one-
in-six day sampling schedule and, thus, do not capture data every day at most sites. To expand
our understanding of the role of specific PM2.5 components and sources with respect to the
observed health effects, researchers have expressed a strong interest in having access to PM2.5
speciation measurements collected more frequently.47
With respect to epidemiological studies evaluating short-term exposures to fine particle
constituents, several new multi-city studies are now available. These studies continue to show an
association between mortality and cardiovascular and/or respiratory morbidity effects and short-
term exposures to various PM2.5 components including nickel (Ni), vanadium (V), EC, OC, and
sulfates (US EPA, 2009a, sections 6.5.2.5 and 6.6). Lippmann et al. (2006) and Dominici et al.
(2007) evaluated the heterogeneity in the PMio-mortality association as evaluated in the
NMMAPS data by analyzing the PM2.5 speciation data. Nickel and V were identified as
significant predictors of variation in PMio-related mortality across cities, with Ni levels in New
York City being reported as particularly high (US EPA, 2009a, section 6.5.2.5; Figure 6-31).48
Bell et al. (2009) and Peng et al. (2009) conducted similar analyses focusing on the variation in
PM2.5-related cardiovascular and respiratory hospital admissions in older adults. Peng et al.
(2009) focused on the components that make up the majority of PM2.5 mass and using multi-
pollutant models reported only EC and OC were significantly associated with risk of
hospitalization for cardiovascular disease. Bell et al. (2009) used data from twenty PM2.5
components and found that EC, Ni, and V were most positively and significantly associated with
the risk of PM25-related hospitalizations suggesting that the observed associations between PM2 5
and hospitalizations may be primarily due to particles from oil combustion and traffic (US EPA,
46 Most studies considered between 7 to 20 ambient PM2 5 constituents, with EC, OC, sulfates, nitrates, and metals
most commonly measured. Many of the studies grouped the constituents with various factorization or source
apportionment techniques to examine the relationship between the grouped constituents and various health effects.
However, not all studies labeled the constituent groupings according to their presumed source and a small number of
controlled human exposure and toxicological studies did not use any constituent grouping. These differences across
studies substantially limit any integrative interpretation of these studies (US EPA, 2009a, p. 6-203).
47 As outlined in section 6.6.2.11 of the ISA, some investigators have circumvented the issue of less than daily
speciation data by using the PM2 5 chemical species data in a second stage regression to explain the heterogeneity in
PM10 or PM2.5 mortality risk estimates across cities and assuming that the relative contributions of PM2 5 have
remained the same over time (US EPA, 2009a, p. 6-206). In April 2008, EPA co-sponsored a workshop to discuss
modifications to the current ambient air quality monitoring networks that would advance our understanding of the
impacts of PM exposures on public health/welfare in the most meaningful way, including improving our
understanding of fine particle components. A summary of the workshop recommendations, including
recommendations for daily PM2 5 speciation measurements in large urban areas, is available at
www.epa.gov/ORD/npd/pdfs/FINAL-April-2008-AQ-Health-Research-Workshop-Summarv-Dec-2008.pdf.
48 However, as noted in the ISA, in a sensitivity analysis when selectively removing cities from the overall estimate,
the significant association between the PM10 mortality risk estimate and the PM2 5 Ni fraction was diminished upon
removing New York City from the analysis, which is consistent with the results presented by Dominici et al. (2007)
(US EPA, 2009a, section 6.5.2.5; Figure 6-32).
2-53
-------
2009a, section 6.2.10.1). In a study of 25 U.S. cities, Franklin et al. (2008) focused on atime-
series regression of mortality related to PM2.5 mass by season and also examined effect
modification due to various PM2 5 species. They concluded that Al, As, Ni, Si and sulfates were
significant effect modifiers of PM2.5 mortality risk estimates, and "simultaneously including Al,
Ni, and sulfates together or Al, Ni, and As together further increased explanatory power. Of the
species examined, Al and Ni explained the most residual heterogeneity" (US EPA, 2009a, p. 6-
194; Table 6-17).49 Furthermore, Ostro et al (2007) examined associations between PM2 5
components and mortality in six California counties and found an association between mortality,
especially cardiovascular-related mortality and several PM2.5 components including EC, OC,
nitrates, iron (Fe), potassium (K), and titanium (Ti) at various lags (US EPA, 2009a, p. 6-195).
Limited evidence is available to evaluate the health effects associated with long-term
exposures to PM2.5 components (US EPA, 2009a, section 7.6.2). The most significant new
evidence is provided by a study that evaluated multiple PM2 5 components and an indicator of
traffic density in an assessment of health effects related to long-term exposure to PM2.5 (Lipfert
et al., 2006). Using health data from a cohort of U.S. military veterans and PM2.5 measurement
data from the CSN, Lipfert et al. (2006) reported positive associations between mortality and
long-term exposures to nitrates, EC, Ni and V as well as traffic density and peak 63
concentrations. Additional evidence from a long-term exposure study conducted in a Dutch
cohort provides supportive evidence that long-term exposure to traffic-related particles is
associated with increased mortality (Beelen et al., 2008).
With respect to source categories of fine particles associated with a range of health
endpoints, the ISA reports that the currently available evidence suggests associations between
cardiovascular effects and a number of specific PM2.5-related source categories, specifically oil
combustion, wood or biomass burning, motor vehicle emissions, and crustal or road dust sources
(US EPA, 2009a, section 6.6; Table 6-18). In addition, a few studies have evaluated associations
between PM2.5-related source categories and mortality. These studies included a study that
reported an association between mortality and a PM2 5 coal combustion factor (Laden et al.,
2000), while other studies linked mortality to a secondary sulfate long-range transport PM2.5
source (Ito et al., 2006; Mar et al., 2006) (US EPA, 2009a, section 6.6.2.1). There is less
consistency in associations observed between sources of fine particles and respiratory health
effects, which may be partially due to the fact that fewer studies have evaluated respiratory-
related outcomes and measures. However, there is some evidence for PM2.5-related associations
with secondary sulfate and decrements in lung function in asthmatic and healthy adults (US
EPA, 2009a, p. 6-211; Gong et al., 2005; Lanki et al., 2006). Respiratory effects relating to the
49
We note that New York City was not included in the 25 cities examined by Franklin et al. (2008).
2-54
-------
crustal/soil/road dust and traffic sources of PM have been observed in asthmatic children and
adults (US EPA, 2009a, p. 6-205; Gent et al., 2009; Penttinen et al., 2006).
Recent studies have shown that source apportionment methods have the potential to add
useful insights into which sources and/or PM constituents may contribute to different health
effects. Of particular interest are several epidemiological studies that compared source
apportionment methods and reported consistent results across research groups (US EPA, 2009a,
p. 6-211; Hopke et al., 2006; Ito et al., 2006; Mar et al., 2006; Thurston et al., 2005). These
studies reported associations between total mortality and secondary sulfate in two cities for two
different lag times. The sulfate effect was stronger for total mortality in Washington D.C. and
for cardiovascular-related morality in Phoenix (US EPA, 2009a, p. 6-204). These studies also
found some evidence for associations with mortality and a number of source categories (e.g.,
biomass/wood combustion, traffic, copper smelter, coal combustion, sea salt) at various lag times
(US EPA, 2009a, p. 6-204). Sarnat et al. (2008) compared three different source apportionment
methods and reported consistent associations between emergency department visits for
cardiovascular diseases with mobile sources and biomass combustion as well as increased
respiratory-related emergency department visits associated with secondary sulfate (US EPA,
2009a, pp. 6-204 and 6-211).
Collectively, in considering the currently available evidence for health effects associated
with specific PM2 5 components or groups of components associated with any source categories
of fine particles as presented in the ISA, we conclude that additional information available in this
review continues to provide evidence that many different constituents of the fine particle mixture
as well as groups of components associated with specific source categories of fine particles are
linked to adverse health effects. However, as noted in the ISA, while "[t]here is some evidence
for trends and patterns that link particular ambient PM constituents or sources with specific
health outcomes.. .there is insufficient evidence to determine whether these patterns are
consistent or robust" (US EPA, 2009a, p. 6-210). Furthermore, the ISA concludes that "the
evidence is not yet sufficient to allow differentiation of those constituents or sources that are
more closely related to specific health outcomes" (US EPA, 2009a, pp. 2-26 and 6-212).
Therefore, we conclude that the currently available evidence is not sufficient to support
consideration of a separate indicator for a specific PM2 5 component or group of components
associated with any source category of fine particles. We also conclude that the evidence is not
sufficient to support eliminating any component or group of components associated with any
source categories of fine particles from the mix of fine particles included in the PM2.5 indicator.
2-55
-------
Summary
In considering whether the currently available evidence provides support for retaining,
revising, or supplementing the current PM2.5 mass-based indicator, we first conclude that it is
appropriate to consider retaining PM2 5 as the indicator for fine particles. Secondly, we conclude
that the currently available evidence does not provide a sufficient basis for supplementing the
mass-based PM2.5 indicator by considering a separate indicator for ultrafine particles. We also
conclude that the currently available evidence is too limited to provide support for considering a
separate indicator for a specific PM2.5 component or group of components associated with any
source categories of fine particles or for eliminating any individual component or group of
components associated with any source categories from the mix of fine particles included in the
PM2.5 mass-based indicator.
In their review of the first draft PA, CAS AC agreed with staffs conclusion that it is
appropriate to consider retaining PM2.5 as the indicator for fine particles and further asserted,
"There [is] insufficient peer-reviewed literature to support any other indicator at this time"
(Samet, 2010c, p. 12). CAS AC expressed a strong desire for EPA to "look ahead to future
review cycles and reinvigorate support for the development of evidence that might lead to newer
indicators that may correlate better with the health effects associated with ambient air
concentrations of PM ..." (Samet, 2010c, p 2). We further conclude that consideration of
alternative indicators (e.g., taking into account new evidence for UFP or PM2.5 composition) in
future reviews is desirable and could be informed by additional research and data collection
efforts, as described in section 2.5 below.
2.3.2 Averaging Times
In 1997, EPA initially set both an annual standard, to provide protection from health
effects associated with both long- and short-term exposures to PM2 5, and a 24-hour standard to
supplement the protection afforded by the annual standard (62 FR 38667 to 38668, July, 18,
1997). In the last review, EPA retained both annual and 24-hour averaging times (71 FR 61164,
October 17, 2006).
In this review, we consider whether the currently available information provides support
for maintaining standards with annual and 24-hour averaging times and whether there is
sufficient evidence to support setting standards with other averaging times to address sub-daily
or seasonal exposures.
• To what extent does the currently available information continue to provide support for
annual and 24-hour averaging times?
The overwhelming majority of studies conducted since the last review continue to utilize
annual (or multi-year) and 24-hour averaging times, reflecting the averaging times of the current
2-56
-------
PM2.5 standards. These studies continue to provide evidence that health effects are associated
with annual and 24-hour averaging times. Therefore, we conclude it is appropriate to retain the
current annual and 24-hour averaging times to provide protection for effects associated with both
long-term (from one year to several years) and short-term (from less than one day to up to
several days) PM2.5 exposures.
• To what extent does the currently available scientific evidence provide support for
considering a standard with an averaging time less than 24 hours to address health
effects associated with sub-daily fine particle exposures?
Relative to information available in the last review, recent studies provide additional
evidence for cardiovascular effects associated with sub-daily (e.g., one to several hours)
exposure to PM, especially effects related to cardiac ischemia, vasomotor function, and more
subtle changes in markers of systemic inflammation, hemostasis, thrombosis and coagulation
(US EPA, 2009a, section 6.2).50 Because these studies have used different indicators (e.g.,
PM2 5, PMio, PMio-2.5, UFPs), averaging times (e.g., 1, 2, and 4 hours), and health outcomes, it is
difficult to draw conclusions about cardiovascular effects associated specifically with sub-daily
exposures to PM2.5.
With regard to respiratory effects associated with sub-daily PM2.5 exposures, the
currently available evidence is much sparser than for cardiovascular effects and continues to be
very limited. The ISA concludes that for several studies of hospital admissions or medical visits
for respiratory diseases, the strongest associations were observed with 24-hour average or longer
exposures, not with less than 24-hour exposures (US EPA, 2009a, section 6.3).
We conclude that this information, when viewed as a whole, is too unclear, with respect
to the indicator, averaging time and health outcome, to serve as a basis for consideration of
establishing a primary PM2.5 standard with an averaging time shorter than 24-hours at this time.
• To what extent does the currently available scientific evidence provide support for
considering separate standards with distinct averaging times to address effects
associated with seasonal fine particle exposures?
With regard to health effects associated with PM2.5 exposure across varying seasons in
this review, Bell et al. (2008) reported higher PM2.5 risk estimates for hospitalization for
cardiovascular and respiratory diseases in the winter compared to other seasons. In comparison
to the winter season, smaller statistically significant associations were also reported between
PM2 5 and cardiovascular morbidity for spring and autumn, and a positive, but statistically non-
50 A limited number of additional studies have also provided evidence of reported ECG changes typically
representative of cardiac ischemia (S-T segment depression) or reported changes in heart rate variability (HRV) (US
EPA, 2009a, sections 6.2.1.2 and 6.2.12.1), however, these changes are often variable and difficult to interpret the
PM2 5 etiologically relevant mechanism underlying the observed effects.
2-57
-------
significant association was observed for the summer months. In the case of mortality, Zanobetti
and Schwartz (2009) reported a 4-fold higher effect estimate for PM2.5 associated mortality for
the spring as compared to the winter. Taken together, these results provide emerging evidence
that individuals may be at greater risk of dying from higher exposures to PM2.5 in the warmer
months and may be at greater risk of PM2.s-associated hospitalization for cardiovascular and
respiratory diseases during colder months of the year.
Overall, we observe that there are few studies presently available to deduce a general
pattern in PM2.5-relatedrisk across seasons. In addition, these studies utilized 24-hour exposure
periods within each season to assess the PM2.5 associated health effects, and do not provide
information on health effects associated with a season-long exposure to PM2.5. Due to these
limitations in the currently available evidence, we conclude that there is no basis to consider a
seasonal averaging time separate from a 24-hour averaging time.
Summary
We recognize that the currently available evidence informs our understanding of
exposure durations of concern and continues to provide strong support for standards that provide
protection for both long- and short- term exposures. In considering the possibility of effects
associated with sub-daily PM2.5 exposures (i.e., less than 24-hour exposures), we recognize that
there is additional evidence available in this review, primarily focused on cardiovascular effects
with more limited evidence for respiratory effects. However, because these studies have used
different indicators of PM exposure (e.g., PM2.5; PMio, UFPs), averaging times, and a broad
range of health outcomes, it is difficult to use this evidence to serve as a basis for establishing a
national standard with a shorter-than-24-hour averaging time. With respect to seasonal effects,
while we recognize there is some new evidence for PM2.5- related effects differentiated by
season, we conclude that this evidence is too limited to use as a basis for establishing a PM2.5
standard with a seasonal averaging time. Based on the above considerations, we conclude that
the currently available information provides strong support for consideration for retaining the
current annual and 24-hour averaging times but does not provide support for considering
alternative averaging times. CAS AC agreed with these conclusions and further noted, "[t]o the
extent that EPA supplements or replaces 24-hour fine PM samplers with continuous monitors, it
may become possible to conduct studies that demonstrate that other averaging times for acute
responses may be more appropriate than 24-hour averages, for at least some effects" (Samet,
2010c, p. 13).
2.3.3 Forms
The "form" of a standard defines the air quality statistic that is to be compared to the
level of the standard in determining whether an area attains the standard. In this review, we
2-58
-------
consider whether currently available information supports consideration of alternative forms for
the annual or 24-hour PM2.5 standards.
2.3.3.1 Form of the Annual Standard
In 1997, 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. This form
was intended to represent a relatively stable measure of air quality and to characterize area-wide
PM2.5 concentrations. The level of the standard was to be compared to measurements made at
the community-oriented monitoring site recording the highest concentration, or, if specific
constraints were met, measurements from multiple community-oriented monitoring sites could
be averaged51 (62 FR 38671 to 38672, July 18, 1997). The constraints were intended to ensure
that spatial averaging would not result in inequities in the level of protection provided by the
standard (62 FR 38672). This approach was consistent with the epidemiological studies on
which the PM2.5 standard was primarily based, in which air quality data were generally averaged
across multiple monitors in an area or were taken from a single monitor that was selected to
represent community-wide exposures, not localized "hot spots."
In the last review, EPA tightened the criteria for use of spatial averaging to provide
increased protection for vulnerable populations exposed to PM2.5.52 This change was based in
part on an analysis of the potential for disproportionate impacts on potentially vulnerable
populations, which found that the highest concentrations in an area tend to be measured at
monitors located in areas where the surrounding population is more likely to have lower
education and income levels, and higher percentages of minority populations (71 FR 61166/2;
US EPA, 2005, section 5.3.6.1).
In this review, we again consider the potential impact of allowing for spatial averaging,
noting that persons from lower socioeconomic strata have been identified as an additional
susceptible population (section 2.2.1).
• Does the currently available evidence provide support for the continued use of spatial
averaging as part of the form of the annual standard?
In considering the potential for disproportionate impacts on potentially susceptible
populations, we updated an air quality analysis conducted for the last review. This analysis
focused on determining if the spatial averaging provisions, as modified in 2006, could introduce
51 The original criteria for spatial averaging included: (1) the annual mean concentration at each site shall be within
20 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 0.6 for each calendar quarter (62 FR 38671 to 38672, July 18, 1997).
52 The current criteria for spatial averaging include: (1) the 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 61167/2-3, October 17, 2006).
2-59
-------
inequities in protection for susceptible populations exposed to PM2.5. Specifically, we evaluated
whether persons with a lower SES (a susceptible population discussed in US EPA,2009a, section
8.1.7) are more likely than the general population to live in areas in which the monitors recording
the highest air quality values in an area are located. Data used in this analysis included
demographic parameters measured at the Census Block or Census Block Group level, including
percent minority population, percent minority subgroup population, percent of persons living
below the poverty level, percent of persons 18 years of age or older, and percent of persons 65
years of age and older. In each candidate geographic area, data from the Census Block(s) or
Census Block Group(s) surrounding the location of the monitoring site (as delineated by radii
buffers of 0.5, 1.0, 2.0, and 3.0 miles) in which the highest air quality value was monitored were
compared to the area-wide average value in the area. This analysis looked beyond areas that
would meet the current spatial averaging criteria and considered all urban areas (i.e., CBSAs)
with at least two valid annual DV monitors (Schmidt et al., 201 la, Analysis A). Recognizing the
limitations of such cross-sectional analyses, we observe that the highest concentrations in an area
tend to be measured at monitors located in areas where the surrounding populations are more
likely to live below the poverty line and to have higher percentage of minorities.
Based upon the analysis described above, staff concludes that the existing constraints on
spatial averaging, as modified in 2006, may not be adequate to avoid substantially greater
exposures in some areas, potentially resulting in disproportionate impacts on susceptible
populations of persons with lower SES levels and minorities. Therefore, we conclude that it is
appropriate to consider revising the form of the annual PM2.5 standard such that it does not allow
for the use of spatial averaging across monitors. In doing so, the level of the annual PM2 5
standard would be compared to measurements made at the monitoring site that represents
community-wide air quality recording the highest PM2.5 concentrations. CASAC agreed with
staff conclusions that it is "reasonable" for EPA to eliminate the spatial averaging provisions
(Samet, 2010d, p. 2). Further, in CAS AC's comments on the first draft PA, they noted, "Given
mounting evidence showing that persons with lower SES levels are a susceptible group for PM-
related health risks, CASAC recommends that the provisions that allow for spatial averaging
across monitors be eliminated for the reasons cited in the (first draft) Policy Assessment" (Samet,
2010c, p. 13).
2.3.3.2 Form of the 24-Hour Standard
In 1997, EPA established the form of the 24-hour PM2.5 standard as the 98th percentile of
24-hour concentrations at each population-oriented monitor within an area, averaged over three
years (62 FR at 38671 to 38674, July 18, 1997). The Agency selected the 98th percentile as an
appropriate balance between adequately limiting the occurrence of peak concentrations and
2-60
-------
providing increased stability which, when averaged over 3 years, facilitated effective health
protection through the development of more stable implementation programs. By basing the
form of the standard on concentrations measured at population-oriented monitoring sites, EPA
intended to provide protection for people residing in or near localized areas of elevated
concentrations. In the last review, in conjunction with lowering the level of the 24-hour
standard, EPA retained this form based in part on a comparison with the 99th percentile form.53
In this review, we have again considered the relative stability of the 98th and 99th
percentile forms.
• Does the currently available evidence provide support for the continued use of the 98th
percentile form of the 24-hour standard?
We recognize that the selection of the appropriate form of the 24-hour standard includes
maintaining adequate protection against peak 24-hour concentrations while also providing a
stable target for risk management programs, which serves to provide for the most effective
public health protection in the long run.54 As in previous reviews, we recognize that a
concentration-based form, compared to an exceedance-based form, is more reflective of the
health risks posed by elevated pollutant concentrations because such a form gives proportionally
greater weight to days when concentrations are well above the level of the standard than to days
when the concentrations are just above the level of the standard. Further, staff concludes that a
concentration-based form, when averaged over three years, provides an appropriate balance
between limiting peak pollutant concentrations and providing a stable regulatory target, thus,
facilitating the development of more stable implementation programs.
In revisiting the stability of a 98th versus 99th percentile form for a 24-hour standard
intended to provide supplemental protection for a generally controlling annual standard, we
consider air quality data reported in 2000 to 2008 to update our understanding of the ratio
between peak-to-mean PM2.5 concentrations.55 As illustrated in Figure 2-2, the 98th percentile
value is a more stable metric than the 99th percentile.
On this basis, we conclude that it is appropriate to consider retaining the current 98th
percentile form of the 24-hour standard as it represents an appropriate balance between
53 In reaching this final decision, EPA recognized a technical problem associated with a potential bias in the method
used to calculate the 98th percentile concentration for this form. The EPA adjusted the sampling frequency
requirement in order to reduce this bias. Accordingly, the Agency modified the final monitoring requirements such
that areas that are within 5 percent of the standards are required to increase the sampling frequency to every day (71
FR61164 to 61165, October 17, 2006).
54 See A TA III, 283 F.3d at 374-375 which concludes it is legitimate for EPA to consider overall stability of the
standard and its resulting promotion of overall effectiveness of NAAQS implementation programs in setting a
standard that is requisite to protect the public health.
55 We consider a coefficient of variation instead of simply the standard deviation because the 99th percentile values
have higher concentration levels and dividing by the mean normalizes the data. In focusing on three years of recent
air quality (2006 to 2008), we see a similar pattern of peak-to-mean ratios (Schmidt et al., 2010, Analysis 3).
2-61
-------
adequately limiting the occurrence of peak concentrations and providing increased stability
relative to an alternative 99th percentile form. In addition, by basing the form of the standard on
concentrations measured at population-oriented monitoring sites, the standard would continue to
focus on providing protection for people residing in or near localized areas of elevated
concentrations.
Figure 2-2. Distribution of Site-level Variation in 98th and 99th Percentile Concentrations,
as Measured by Coefficient of Variation (SD/Mean) Computed by Site Across Years, 2000-
2008
Notes: The box indicates the
interquartile range of the
distribution, the line through the
box denotes the distribution
median, the whisker caps mark
the 5th and 95th percentile values,
and the star identifies the
distribution mean.
98th percentile 99th percentile
concentrations concentrations
Source: Schmidt et al., 2010, Analysis 3
2.3.4 Alternative Levels
In reaching staff conclusions for alternative standard levels that are appropriate to
consider, we take into account both evidence-based (section 2.3.4.1) and risk-based
considerations (section 2.3.4.2) as well as the related limitations and uncertainties associated
with this information as presented and discussed more fully in the ISA and RA (US EPA, 2009a;
US EPA, 2010a). CASAC advice on alternative standard levels is summarized in section 2.3.4.3.
Staff conclusions based on the integration of the evidence-based and risk-based approaches as
well as consideration of CASAC advice and public comments on the first and second draft PAs
are presented in section 2.3.4.4.
Alternative levels are discussed in conjunction with staff conclusions on other elements
of the standard presented above, notably, retaining PM2.5 as the indicator for fine particles
(section 2.3.1); retaining the current annual and 24-hour averaging times (section 2.3.2);
modifying the current form of the annual standard to eliminate spatial averaging (section 2.3.3.1)
2-62
-------
and retaining the current form of the 24-hour standard (section 2.3.3.2). Specifically, we address
the following overarching question:
What alternative standard levels are appropriate to consider to provide requisite
protection for health effects associated with long- and short-term PM2.s exposures?
2.3.4.1 Evidence-based Considerations
In translating information from epidemiological studies into the basis for reaching staff
conclusions on alternative standard levels that are appropriate for consideration, we apply the
policy framework outlined in section 2.1.3. In doing so, we focus on identifying levels for an
annual standard and a 24-hour standard that, in combination, provide protection for health effects
associated with both long- and short-term PM2.5 exposures. We also consider the extent to which
various combinations of annual and 24-hour standards reflect setting a generally controlling
annual standard with a 24-hour standard providing supplemental protection. We conclude this
policy goal results in the most efficient and effective way to provide appropriate public health
protection.
As discussed in section 2.1.3, we recognize that there is no single factor or criterion that
comprises the "correct" approach for reaching staff conclusions on alternative standard levels for
consideration, but rather there are various approaches that are reasonable to consider. In
reaching staff conclusions on the ranges of standard levels that are appropriate to consider, we
address a series of specific questions beginning with consideration of the relative weight to place
on different evidence.
In recognizing the absence of a discernible population threshold below which effects
would not occur, our general approach for identifying alternative annual standard levels that are
appropriate to consider focuses on characterizing the range of PM2.5 concentrations over which
we have the most confidence in the associations reported in the epidemiological studies, and
conversely where our confidence in the association becomes appreciably lower. The most direct
approach to address this issue is to consider epidemiological studies reporting confidence
intervals around C-R relationships. We also explore other approaches that consider different
statistical metrics to identify ranges of long-term mean PM2.5 concentrations that were most
influential in generating health effect estimates in long- and short-term epidemiological studies,
placing greatest weight on those studies that reported positive and statistically significant
associations.
With regard to identifying alternative 24-hour standard levels that are appropriate to
consider, we look at the distributions of 24-hour PM2 5 concentrations reported in short-term
exposure studies, focusing on the 98th percentile concentrations to match the form of the 24-hour
standard (as discussed in section 2.3.3.2). In recognizing that the annual and 24-hour standards
2-63
-------
work together to provide protection for effects associated with short-term PM2.5 exposures, we
also consider information on the long-term mean PM2.5 concentrations from these studies. In
addition to considering the epidemiological evidence, we also consider air quality information,
specifically peak-to-mean ratios using county-level 24-hour and annual design values, to
characterize air quality patterns in areas possibly associated with strong local or seasonal
sources. These patterns help us understand the extent to which different combinations of annual
and 24-hour standards would be consistent with the policy goal of setting a generally controlling
annual standard with a 24-hour standard that provides supplemental protection.
• What factors do we weigh in placing emphasis on epidemiological evidence to translate
this information into staff conclusions on alternative standard levels?
As discussed in section 2.1.3, we initially focus on long- and short-term PM2.5 exposure
studies conducted in the U.S. and Canada and place the greatest weight on health outcomes that
have been judged in the ISA as having evidence to support a causal or likely causal relationship.
We also consider the evidence for a broader range of health outcomes judged in the ISA to have
evidence suggestive of a causal relationship, specifically studies that focus on effects in
susceptible populations, to evaluate whether this evidence provides support for considering lower
alternative standard levels.
We take several factors into account in placing relative weight on the body of available
epidemiological studies, for example, study characteristics, including study design (e.g., time
period of air quality monitoring, control for potential confounders); strength of the study (in
terms of statistical significance and precision of results); and availability of population-level and
air quality distribution data. We place greatest weight on information from multi-city
epidemiological studies to inform staff conclusions regarding alternative annual standard levels.
These studies have a number of advantages compared to single-city studies56 that include
providing representation of ambient PM2.5 concentrations and potential health impacts across a
range of diverse locations providing spatial coverage for different regions across the country,
reflecting differences in PM2.5 sources, composition, and potentially other exposure-related
factors which might impact PM2.5-related risks; lack of 'publication bias' (US EPA, 2004, p. 8-
30); and consideration of larger study populations that afford the possibility of generalizing to
the broader national population and provide higher statistical power than single-city studies to
56 As discussed in section 2.2.1, we recognize that single-city studies provide ancillary evidence to multi-city studies
in support of calling into question the adequacy of the current suite of standards. However, in light of the mixed
findings reported in single-city short-term PM2 5 exposure studies, and the likelihood that these results are influenced
by localized events and not representative of air quality across the country, we place comparatively greater weight
on the results from multi-city studies in considering alternative annual standard levels.
2-64
-------
detect potentially statistically significant associations with relatively more precise effect
estimates.
In reaching staff conclusions regarding alternative 24-hour standard levels that are
appropriate to consider, we also take into account relevant information from single-city short-
term PM2.5 exposure studies. Although, as discussed above, multi-city studies have greater
power to detect associations and provide broader geographic coverage in comparison to single-
city studies, the extent to which effects reported in multi-city short-term PM2.5 exposure studies
are associated with the specific short-term air quality in any particular location is unclear,
especially when considering short-term concentrations at the upper end of the air quality
distribution (i.e., at the 98th percentile value) for a given study area. In contrast, single-city
studies are more limited in terms of power and geographic coverage but the link between
reported health effects and the air quality in a given study area is more straightforward to
establish. Therefore, we consider the results of both multi-city and single-city short-term
exposure studies to inform staff conclusions regarding alternative levels that are appropriate to
consider for a 24-hour standard that is intended to provide supplemental protection in areas
where the annual standard may not offer appropriate protection against the effects of all short-
term exposures.
• To what extent have confidence intervals around concentration-response relationships
reported in epidemiological studies been characterized and to what extent do they
inform the identification of alternative standard levels?
Based on a thorough search of the available evidence, we identified three long-term PM2.5
exposure studies reporting confidence intervals around C-R functions (i.e., Schwartz et al., 2008;
Pope et al., 2002; Miller et al., 2007; see Figure 2-3).57 In our assessment of these studies, we
placed greater weight on analyses that averaged across multiple C-R models since this approach
represents a more robust examination of the underlying C-R relationship than analyses
considering a single C-R model. Although epidemiological studies reporting C-R functions and
associated 95% confidence intervals provide information on the precision of the effect estimates
at specific concentrations in the air quality distribution (i.e., point-wise confidence intervals), the
concentration below which the confidence intervals for effect estimates becomes notably wider is
intrinsically related to data density (an issue that is explored in the subsequent question), and not
necessarily indicative of lower confidence in the underlying C-R relationship. Therefore, for the
purposes of this discussion, we are interested in additional information that confidence intervals
associated with C-R functions can provide, specifically as to the extent to which the confidence
57 Subsequent to a thorough exploration of the published evidence, we were unable to identify any short-term PM25
exposure studies that characterized confidence intervals around C-R relationships.
2-65
-------
intervals around non-parametrically estimated C-R functions indicate appreciably less
confidence that an association continues to exist at lower PM2.5 concentrations.
Based on our review of the evidence and on our discussions with the investigators of the
aforementioned three studies, we determined that these studies analyzed the C-R function
primarily to determine if a linear curve most appropriately represented the C-R relationship, and
not to characterize the confidence intervals associated with the underlying C-R relationship.
Nonetheless, we consider what additional information beyond the shape of the C-R relationship
these non-parametric curves can provide to inform our confidence in the association at lower
PM2.5 concentrations.
In the first study, Schwartz et al. (2008) used a variety of statistical methods to analyze the shape
of the C-R relationship associated with long-term PM2 5 exposure, and to investigate whether a
mortality effect threshold exists among participants of the Harvard Six Cities cohort. The
authors conducted a Bayesian Modeling Averaging (BMA) analysis that included 32 distinct
models, some with the option for a threshold at 10 |ig/m3 and higher. Based on the BMA
analysis, the C-R function reported by Schwartz et al. (2008) was found to be linear, with
"... .little evidence for a threshold in the association between exposure to fine particles and the
risk of death...". Moreover, based on the BMA analysis (Figure 2-3a), EPA staff observed a
widening of the confidence intervals around the C-R function for PM2.s-related mortality. This
widening of the confidence intervals occurred around the long-term mean concentration of
approximately 10 |ig/m3 with continued broadening of the confidence intervals around the C-R
function at lower PM2.5 concentrations. This broadening is likely due in part to the comparative
lack of data at the lower end of the air quality distribution considered in this analysis (i.e., 10.7
ug/m3 is one standard deviation below the long-term mean concentration and the lower bound of
the range of air quality considered was 8 ug/m3). Consequently, we observe from this analysis
that the widening of confidence intervals around C-R relationships at lower concentrations is
due, in part, to the density of air quality data, reinforcing the view that one has most confidence
in study results over the range of concentrations where the bulk of the data exist. In a separate
analysis, Schwartz et al. (2008) also included point-wise confidence intervals for the association
between PM2.5 and mortality in a smoothing plot of a single nonparametric model (Figure 2-3b).
However, since these point-wise confidence intervals relate specifically to the magnitude of
effect estimates generated at specific PM2.5 concentrations from a single model, we conclude that
this analysis is not informative for characterizing confidence intervals associated with the
underlying C-R relationship.
2-66
-------
Figure 2-3. Confidence Intervals Around PM2.s Concentration-Response Relationships
Information from Multi-city Epidemiological Studies
a.
0.6-
.2 0.4
5 0-2
N
n
20
30
10
Figure 2. The estimated concentration-response
relation between PM2.s and the risk of death in the
Six Cities Study, based on averaging the 32 possi-
ble models that were fit. Also shown are the point-
wise 95% CIs around that curve, based on jacknife
estimates.
b.
I
&
Figure 1. The estimated concentration—response relation between PMss and the risk of death in the Six
Cities Study, using a penalized spline witti IS knots. Also shown are the pointwise 95% CIs.
Source: Schwartz et al., 2008
2-67
-------
Figure 2-3. Confidence Intervals Around PM2.s Concentration-Response Relationships
from Multi-City Epidemiological Studies (cont.)
c.
Figure 2. N on parametric Smoothed Exposure Response Relationship
I A | AJI-Cau&g Mortality
E | CandfcipulTionary Mortally
.«.
DC
C-
-O.1-
-O.A-
| C | Lung Career Mortality
D | All Other Causa Mortality
G.2-
O.1-
O-
-0.4-
10
2O
2O
PM2 5,
Vertical lines along x-a*e$ Indicate mg or frequency plot of mean fine particulate pollution; PM.2 =, mean fine
pai"tlcles measuring less than 2.5 urn In diameter; RR, relative risk; and Clr confidence Interval.
Source: Pope et al., 2002
d.
B Between-City Effect
12-
; J
n
9 12 15 IS 21 24 27 30
C Within-City Effect
PM2
Figure 1. Level of Exposure to Fine Particulate Matter and the Risk of Death from Cardiovascular Causes in Women.
The graphs demonstrate the observed relationship between the risk of death from cardiovascular disease and the level of particulate
matter of less than 2.5 j/m in aerodynamic diameter (PM2.S), including both definite and possible deaths from coronary heart disease
or cerebrovascular disease. Panel A shows the overall relationship between the PM2-S level and death. Panel B the effects between met-
ropolitan areas, and Panel C the effects within metropolitan areas, with an indicator variable used to adjust for each city. These results
suggest a generally linear relationship between exposure and risk, though the 95% confidence intervals (shaded areas) are wide at the
extremes of exposure. Risk is depicted in comparison with a reference value of 11 jjg per cubic meter. The histogram in each panel illus-
trates the density of exposure distribution for air pollution. All estimates are adjusted for age, race or ethnic group, educational level,
household income, smoking status, systolic blood pressure, body-mass index, and presence or absence of a history of diabetes, hyper-
tension, or hypercholesterolemia.
Source: Miller et al., 2007
2-68
-------
In the second long-term PM2.5 exposure study, Pope et al. (2002) utilized a nonparametric
smoothing model to depict the shape of the C-R function and associated confidence intervals
using mortality data from the ACS cohort. This exploratory analysis was conducted primarily to
examine departures from linearity and to support the utilization of log-linear regression models.
Goodness-of-fit testing indicated these data supported a log-linear C-R relationship. We observe
(and observed in the last review) that there was an appreciable widening of point-wise
confidence intervals on the smoothing plot for all-cause mortality beginning at approximately 13
to 12 ug/m3 (Figure 2-3c; US EPA, 2005, p. 3-56; Figure 3-4) which is somewhat below where
the bulk of the air quality distribution in the study occurs (i.e., 14 ug/m3 is one standard deviation
below the long-term mean concentration of 17.7 ug/m3). However, similar to the single-model
analysis presented in the discussion of the Schwartz et al. (2008) study (Figure 2-3b), this
analysis was also generated from a single model, which resulted in point-wise confidence
intervals that relate to the magnitude of effect estimates generated at specific PM2.5
concentrations. Moreover, the widening of the confidence intervals in the smoothing plot is
likely a consequence of the comparative lack of air quality data at lower PM2.s concentrations in
this study, and does not suggest the possibility that an effects threshold may exist at lower PM2.5
concentrations. It also remains unclear how representative this single model is of the underlying
C-R relationship. Therefore, a potentially more robust approach to characterizing confidence
intervals associated with C-R relationships may involve the generation of multiple C-R functions
as was conducted by Schwartz et al. (2008), as just discussed.
The third long-term PM2.5 exposure study provides information on the shape of the C-R
function and associated confidence intervals using cardiovascular-related mortality and
morbidity data from the WHI cohort (Miller et al., 2007; Figure 2-3d). This analysis of
cardiovascular events in relation to long-term PM2.5 exposure is indicative of a continued
reduction in risk at lower concentrations of PM2.5, with no evidence of a discernible effects
threshold, and provides additional support for log-linear C-R relationships (US EPA, 2009a,
section 2.4.3). We also recognize that the relative number of events occurring at the lower
quintiles of exposure is similar to the number of events at the higher quintiles, indicating that
there is statistical power to detect an association between PM2.5 and mortality at the lower PM2.5
concentrations. However, since the reference value for the C-R function in this analysis is
1 Iug/m3; and effect estimates across the entire distribution are being compared to effect
estimates at this concentration, this analysis does not identify a PM2.5 concentration where there
is an appreciable widening of confidence intervals and reduced confidence in the underlying C-R
relationship in this study.
In summary, we recognize there are important differences in the statistical approaches
utilized to create the C-R functions and associated confidence intervals reported in Schwartz et
2-69
-------
al. (2008), Pope et al. (2002), and Miller et al. (2007) that inform our interpretation of these
studies. Although these analyses of long-term exposure to PM2.5 provide information on the lack
of any discernible population threshold, only Schwartz et al. (2008) conducted a multi-model
analysis to characterize confidence intervals around the estimated C-R relationship that can help
inform at what PM2.5 concentrations we have appreciably less confidence in the nature of the
underlying C-R relationship. Although analyses of confidence intervals associated with C-R
relationships can help inform consideration of alternative standard levels, we conclude that the
single relevant analysis now available is too limited to serve as the principal basis for identifying
alternative standard levels in this review.
• How do we consider different statistical metrics for identifying alternative levels that
are appropriate to consider for an annual standard?
As outlined above, we recognize that health effects may occur over the full range of
concentrations observed in the long- and short-term epidemiological studies and that the ISA
concluded no discernible population threshold for any effects can be identified based on the
currently available evidence (US EPA 2009a, section 2.4.3). In identifying alternative standard
levels that are appropriate to consider, we first take into account the statistical metric used in
previous reviews. This approach recognizes that the strongest evidence of associations occurs at
concentrations around the long-term mean concentration. Thus, in earlier reviews, we focused
on identifying standard levels that were somewhat below the long-term mean concentrations
reported in PM2.5 exposure studies. As outlined in section 2.1.3, the long-term mean
concentrations represent air quality data typically used in epidemiological analyses and provide a
direct link between PM2.5 concentrations and the observed health effects. Further, these data are
available for all long- and short-term exposure studies analyzed and, therefore, represent the data
set available for the broadest set of epidemiological studies.
Consistent with CASAC's comments on the second draft PA, we also explore ways to
take into account additional information from epidemiological studies, when available (Samet,
2010d, p. 2). We do this by evaluating different statistical metrics, beyond the long-term mean
concentration, to characterize the range of PM2.5 concentrations down through which we
continue to have confidence in the associations observed in epidemiological studies and below
which there is a comparative lack of data such that our confidence in the relationship is
appreciably less. This would also be the range of PM2.5 concentrations which have the most
influence on generating the health effect estimates reported in epidemiological studies. As
discussed in section 2.1.3, we recognize there is no one percentile value within a given
distribution that is the most appropriate or "correct" way to characterize where our confidence in
the associations becomes appreciably lower. We conclude that focusing on concentrations
within the lower quartile of a distribution, such as the range from the 25th to the 10th percentile, is
2-70
-------
reasonable to consider as a region within which we begin to have appreciably less confidence in
the associations observed in epidemiological studies.58 In staffs view, considering lower PM2.5
concentrations, down to the lowest concentration observed in a study, would be a highly
uncertain basis for selecting alternative standard levels.
As outlined in 2.1.3, we recognize that there are two types of population-level
information to consider in identifying the range of PM2.5 concentrations which have the most
influence on generating the health effect estimates reported in epidemiological studies. The most
relevant information to consider is the number of health events (e.g., deaths, hospitalizations)
occurring within a study population in relation to the distribution of PM2.5 concentrations likely
experienced by study participants. Although not as directly relevant, the number of participants
within each study area can be used as a surrogate for health event data in relation to the
distribution of PM2.5 concentrations.
In applying this approach, we focus on identifying the broader range of PM2.5
concentrations which had the most influence on generating health effect estimates in
epidemiological studies. As discussed below, in working with study investigators, we were able
to obtain health event data for three large multi-city studies (Krewski et al., 2009; Zanobetti and
Schwartz, 2009; Bell et al., 2008)59 and population data for the same three studies and one
additional long-term PM2.5 exposure study (Miller et al., 2007); as documented in a staff
memorandum (Rajan et al., 2011). In the absence of health event and population data, the
distribution of PM2.5 concentrations across study areas could be used to represent the PM2.5
concentrations likely experienced by study participants. However, we were unable to determine
from the methodologies published for these studies how the air quality was statistically weighted
to account for variations in the availability of daily PM2.5 measurements by study area.
Consequently, we are unable to consider this type of information as part of our effort to identify
the broader range of PM2.5 concentrations that were most influential in generating the health
effect estimates in epidemiological studies.
Figures 2-4 through 2-7 summarize policy-relevant epidemiological evidence. Long- and
short-term exposure studies that evaluated endpoints classified in the ISA as having evidence of
58 In the last review, staff believed it was appropriate to consider a level for an annual PM25 standard that was
somewhat below the averages of the long-term concentrations across the cities in each of the key long-term
exposures studies, recognizing that the evidence of an association in any such study was strongest at and around the
long-term average where the data in the study are most concentrated. For example, the interquartile range of long-
term average concentrations within a study and a range within one standard deviation around the study mean were
considered reasonable approaches for characterizing the range over which the evidence of association is strongest
(US EPA, 2005, p. 5-22). In this review, we note the interrelatedness of the distributional statistics and a range of
one standard deviation around the mean which contains approximately 68% of normally distributed data, in that one
standard deviation below the mean falls between the 25th and 10th percentiles.
59 Effect estimates from Krewski et al. (2009), Zanobetti and Schwartz (2009), and Bell et al. (2008) were used to
generate the core quantitative risk assessment results presented in the RA (US EPA, 2010a).
2-71
-------
a causal or likely causal relationship (including studies with long-term mean PM2.5
concentrations below 17 |ig/m3) are summarized in Figures 2-4 through 2-6.60 Figures 2-4 and
2-6 provide summaries of long- and short-term exposure studies that evaluated mortality
(evidence of a causal relationship); cardiovascular effects (evidence of a causal relationship);
and respiratory effects (evidence of a likely causal relationship) in the general population, as
well as older adults, a susceptible population. Figure 2-5 provides a summary of studies that
evaluated respiratory effects (evidence of a likely causal relationship) in children, a susceptible
population, as well as long-term exposure studies of children that evaluated developmental
effects that are classified as having evidence suggestive of a causal relationship. Figure 2-7
presents cumulative frequency plots of the health events or number of study participants and
corresponding long-term mean PM2 5 concentrations for the four studies for which we were able
to obtain such data (Krewski et al., 2009; Zanobetti and Schwartz, 2009; Bell et al., 2008; Miller
et al., 2007). This figure highlights the range of PM2.5 concentrations between the 25th to 10th
percentiles of the distribution of health events or number of study participants. In general, we
note that the long-term mean PM2.5 concentrations based upon the distributions of health event
and study population data are similar, and provide support for the utilization of data on the
number of study participants as a surrogate for health event data (Rajan et al., 2011).
As outlined above in section 2.1.3, epidemiological studies typically report
concentrations based upon a composite monitor distribution; that is, concentrations averaged
across ambient monitors within each area included in a given study are averaged across study
areas to calculate an overall study mean concentration. As noted above, a policy approach that
uses data based on composite monitor distributions to identify alternative standard levels, and
then compares those levels to concentrations at appropriate maximum monitors to determine if
an area meets a given standard, directionally has the potential to build in some margin of
safety.61
60 We note that additional studies presented and assessed in the ISA report effects at higher long-term mean PM2 5
concentrations.
61 In analyses conducted by EPA staff based on selected long- and short-term exposure studies, we note that the
differences between the maximum and composite distributions were greater for studies with fewer years of air
quality data (i.e., 1 to 3 years) and smaller numbers of study areas (i.e., 36 to 51 study areas). The differences in the
maximum and composite monitor distribution were much smaller for studies with more years of air quality data (i.e.,
up to 6 years) and larger numbers of study areas (i.e., 112 to 204 study areas) (Hassett-Sipple et al., 2010; US EPA,
2010f, section 2.3.4.1). Therefore, any margin of safety that may be provided by a policy approach that uses data
based on composite monitor distributions to identify alternative standard levels, and then compares those levels to
concentrations at appropriate maximum monitors to determine if an area meets a given standard, will vary
depending upon the number of monitors and air quality distributions within a given area. See also footnote 14 in
section 2.1.3 and associated text.
2-72
-------
Figure 2-4. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for Multi-City, Long-term PM2.s
Exposure Studies of the General Population and Older Adults
Study
Cits
Geographic
Area
Years of Air
Quality
Data
Endpoint
Air Quality Data (Mg.'m3)
Au
Mean
her Rape
Mean-
1SD
rted Data
Range
Effect Estimate (95% Cl)
General Population
WHI
Cystic Fibrosis
ACS-Reanalysis II
VA
Harvard Six Cities
(SCSJ-Extended
MCAPS-Western US
Medicare-ACS
MCAPS€astern US
Medicare -SCS
Miller ctal. (2007)
Gosse:al.(2004)b
Krewski etal. (2009)
Upfert eta. (2006)
Laden etal. (2003)
Zogcr eta. (2008)
Qtim etal. (2008)
Zeger eta.. (2008)
Eftim etal. (2008)
36 US cities
6 US regions (NE.
SE.NCSC.NW'
' sw; '
116USMSAs
6 US cities
(Northeast/
Midwest)
62 US courocs
51 US MSAs
421 US counties
6 US cities
2000
2000
1999-2000
1999-2001
1979-1998
2000 2005
2000-2002
2000-2005
2000-2002
Mortality-CV
AIICVD
Incident Ml
Revascularizaton
Stroke
CBVD
Mortality-AII-cause
Pumonary
exacerbated
Mo1aity-all cause
Mortality-IHO
Mortalif/-CPD
Mortality-Lung canea
Mortality-all cause
Vkrtality-all cause
Mortality-CV
Mortality-Respirator)1
Mortality-Lung cancer
Mcrtalily all cause
Molality-all cause
Vfcrtality-all cause
Molaity-all cause
12.9"
137
U.O
14.3
16.4'
01
13.1"
13.6
14.0=
U.I
102
95
110
113
10,8
terAdul
108
-
110
3.4 28.3
11815.9
(IQR)
5.8-22
£.0-?
13-22
Is
10.4-18.5
(IQR)
6.D-25.1
12.3-15.3
(IQR)
9.6-19. 1
•
•
•
-~~-*~~-
"
•
~*~
•
-•—
•
•
1
•
•
-•-
s 1 1 i i
'Update of Milleret al. BODO JM;5 data mdudsd n Curl, 2039
'-'Cohort included persons with cystic ftbrasra age 6 and older, mear age:' 8.4 yrs
•Estimated from data provided by study author (Laden. 2009)
^Median (IQR. Infe^uattile syiiyyi, uveiali US repyiteJ rnedian iJQR) uf 13,2 (jy,^n3111.1-"4.S)
Q8
1.2 1.4 1.6 1.8 2 22 24
2-73
-------
Figure 2-5. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for Multi-City, Long-term PM2.s
Exposure Studies of Children
Children
Study
SCACHS
24-Cities
Cite
Belletal.
(2007)
Liuetal.
(2007)
Parker and
Woodruff
(2008)
McConnel! et
al.(2003)
Dockery et al.
(1996)
Woodruff etal.
(2008)
Geographic
Area
CT,MA
3 Canadian cities
Continental US
12 communities -
SCA
24 communities -
US, Canada
96 US counties
Years of Air Quality
Data
1998-2002
1985-1999
2000-2003
1996-1999
1988-1991
1999-2002
Endpoint
Low Birth Weight
lUGR-l™1 trimester
IUGR-2"d trimester
IUGR- 3rd trimester
Low Birth Weight
Bronchitic
Symptoms
Bronchitis
Infant mortality
Air Quality Data (MQ/m3)
Mean
11.9s
12.2
13.5a
13.8
14.5
14.9b
Mean-
1SD
10.3
-
-
6.1
10.3
-
Range
6.3-15
10.9-16.1
(IQR)
6-29
5.8-20.7
12.0-18.6
(IQR)
Effect Estimate (95% Cl)
— •
*
-^
-
^
-•-
'Gestations! mean
''Median for all cause mortality; median (IQR: interquartile range) for survivors = 14.8 (11.7-187) |jg/m3. Exposure period was first 2 months of life.
1 1.2 1,4 1.6 1.8 2
2-74
-------
Figure 2-6. Summary of Effect Estimates (per 10 ug/m3) and Air Quality Distributions for Multi-City, Short-term PM2.s
Exposure Studies of the General Population and Older Adults
Study/Cite
Geographic
Area
Years of
Air Quality
Data
Endpoint
Air Quality Data fug/m3)
Author Reported Data
Mean
Mean-
1SO
Range
»
percentile
Effect Estimate (95% Cl) |
Genera/ Population
Burnett etal. (2004)
Zanobetti & Schwartz (2009)
Burnett &. Goldberg (2003)
Harvard Six Cities/ Klemm and Mason
(2003)
Franklin etal. (2008)
Franklin el al. (2007)
MCAPS/Bell etal. (2008)
MCAPSDominici etal. 2006
O'Connor (2008)
12 Canadian Cities
112 US counties
8 Canadian Cities
8 US cites
(Northeast/
Midwest)
25 US
communities
27 US
communities
202 US counties
204 US counties
7 US Cities
1981-1999
1999-2005
1986-1996
1979-1988
2000-2005
1997-2TO2
1999-2005
1999-2002
1998-2001
Nonacciclentai
mortality
Nonaccidenta!
mortality
Nonaccidental
mortality
Nonacodenta!
mortality
Nenacodentai
mortality
Nonacodentai
mortality
CVOHA
Resp HA
IHDHA
CHFHA
Dysittythmia HA
CBVD HA
PVDHA
COPD HA
RTIHA
Wheeze/Cough
12.8
13.2>
13.3
14.7'
14*
15.&-
Old
12*
13.4-
14,0
10.3'
3*
er Adult.
10.2»
10.5-
-
6.6-24.7
9-23
(IQR)
9.9-27.4"
8.8-23.9
i/CWidren
4-20
4-23
38.0
34.3
38.9"
43.0
45.8
$
342
34.8
39.0s
— —
.
-—
m
— —
— . —
— .
— » —
m
0.97 0.98 0.89
1.01 1.02 1.03
Estimated from data provided by study author or published study
"Estimated from coefficient of variation reported in original study by Burnett et al. (2000)
'-Mean value not reported in study, median presented from original study by Schwartz etal. (1996)
dMCAPS cohort included adulls 5 65 yrs; O'Connor (2008'l cohort included children, mean age: 7.7 yrs
iQR: interquartile range
2-75
-------
Figure 2-7. Distribution of Population-Level Data and Corresponding PM2.s Concentrations for Selected Multi-City
Epidemiological Studies
100
95-
o 90-3-
•7= 8S
a.
O 80-
3J
O
rs
U
2
o
70
65-
60^
55
50
45
40
35-
30
254
20
15
10--
5-
0
4
Health Event Data
• Krewski et al.. 2009 (ACS) - deaths (116 study areas)
_ Zanobetti & Schwartz, 2009 - deaths (99 study areas)
n Bell cl al., 2008 (MCAPS) - cardiovascular-related
hospilalizalions (202 study areas)
Population Data
* Miller et al., 2007 (WHI, 36 study areas)
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
Corresponding Annual Mean (jag/m3) Source: Rajan ct al., 2011
2-76
-------
We initially consider distributional statistics for both long- and short-term exposure
studies that evaluated mortality (causal relationship), cardiovascular effects (causal relationship)
or respiratory effects (likely causal relationship). In looking first at the long-term mean
concentrations in the long-term exposure studies, we observe positive and often statistically
significant associations at long-term mean PM2.5 concentrations ranging from 16.4 to 12.9 |ig/m3
(Laden et al., 2006; Lipfert et al., 2006; Krewski et al., 2009; Goss et al., 2004; Miller et al.;
2007; Zeger et al., 2008; Eftim et al., 2008; Dockery et al., 1996; McConnell et al., 2003; see
Figures 2-4 and 2-5). In considering the one long-term PM2.5 exposure study for which we have
health event data (Krewski et al., 2009), we observe that the long-term mean PM2.5
concentrations corresponding with study areas contributing to the 25th and 10th percentiles of the
distribution of mortality data are 12.0 |ig/m3 and 10.2 |ig/m3, respectively. As identified above,
although less directly relevant than event data, the number of participants within each study area
can be used as a surrogate for health event data in relation to the distribution of PM2.5
concentrations. The long-term mean PM2.5 concentrations corresponding with study areas
contributing to the 25th and 10th percentiles of the distribution of study participants for Miller et
al. (2007) were 11.2 |ig/m3 and 9.7 |ig/m3! respectively (Rajan et al., 2011).
In then considering information from multi-city, short-term exposure studies reporting
positive and statistically significant associations with these same broad health effect categories,
we observe positive and statistically significant associations at long-term mean PM2 5
concentrations in a similar range of 15.6 to 12.8 |ig/m3 (Franklin et al., 2007, 2008; Klemm and
Mason, 2003; Burnett and Goldberg, 2003; Zanobetti and Schwartz, 2009; Burnett et al., 2005;
Bell et al., 2008; Dominici et al., 2006a; see Figure 2-6).62 In considering the two multi-city,
short-term PM2.5 exposure studies for which we have health event data, we observe that the long-
term mean PM2 5 concentrations corresponding with study areas contributing to the 25th and 10th
percentiles of the distribution of deaths and cardiovascular-related hospitalizations are 12.5
|ig/m3 and 10.3 |ig/m3, respectively, for Zanobetti and Schwartz (2009), and 11.5 |ig/m3and 9.8
|ig/m3, respectively, for Bell et al. (2008) (Rajan et al., 2011).
Taking into consideration additional studies of specific susceptible lifestages (i.e.,
childhood), we expand our evaluation of the long-term exposure studies to include a broader
range of health outcomes judged in the ISA to have evidence suggestive of a causal relationship.
This evidence is taken into account to evaluate whether it provides support for considering lower
alternative levels than if weight were only placed on studies for which health effects have been
62 When integrating evidence from short-term exposure studies reporting positive and statistically significant
associations with mortality, cardiovascular, and respiratory effects, the ISA concluded that these associations are
generally consistent and precise at long-term mean PM2 5 concentrations of 12.8 ug/m3 and above (US EPA, 2009a,
pp. 2-10 to 2-11)
2-77
-------
judged in the ISA to evidence supporting a causal or likely causal relationship. We make note of
a limited number of studies that provide emerging evidence for PM2.s-related low birth weight
and infant mortality, especially related to respiratory causes during the post-neonatal period.
This more limited body of evidence indicates positive and often statistically significant effects
associated with long-term PM2.5 mean concentrations in the range of 14.9 to 11.9 |ig/m3
(Woodruff et al., 2008; Liu et al., 2007; Bell et al., 2007; see Figure 2-5). As illustrated in
Figure 2-5, although Parker and Woodruff (2008) did not observe an association between
quarterly estimates of exposure to PM2.5 and low birth weight in a multi-city U.S. study, other
U.S. and Canadian studies did report positive and statistically significant associations between
PM2.5 and low birth weight at lower ambient concentrations (Bell et al., 2007; Liu et al., 2007).63
There remain significant limitations (e.g., identifying the etiologically relevant time period) in
the evaluation of evidence on the relationship between PM2.5 exposures and birth outcomes (US
EPA, 2009a, pp. 7-48 and 7-56) which should be taken into consideration in reaching judgments
about how to weigh these studies of potential impacts on specific susceptible populations in
considering alternative standard levels that provide protection with an appropriate margin of
safety.
With respect to carcinogenicity, mutagenicity, and genotoxicity (evidence suggestive of a
causal relationship), the strongest evidence currently available is from long-term prospective
cohort studies that report positive associations between PM2.5 and lung cancer mortality. At this
time, the PM2.5 concentrations reported in studies evaluating these effects generally included
ambient concentrations that are equal to or greater than ambient concentrations observed in
studies that reported mortality and cardiovascular and respiratory effects (US EPA, 2009a,
section 7.5). Therefore, in selecting alternative levels, we note that in providing protection for
mortality and cardiovascular and respiratory effects, it is reasonable to anticipate that protection
will also be provided for carcinogenicity, mutagenicity, and genotoxicity effects.
Summary of Evidence-based Considerations to Inform Annual Standard Level
In considering the currently available evidence and air quality information within the
context of identifying potential alternative annual standard levels for consideration, we again
note that the ISA concludes there is no evidence of a discernible population threshold below
which effects would not occur. Thus, health effects may occur over the full range of
concentrations observed in the epidemiological studies. In the absence of any discernible
thresholds, our general approach for identifying alternative standard levels that would provide
63 As noted in section 7.4 of the ISA, Parker et al. (2005) reported that over a 9 month exposure period (mean PM25
concentration of 15.4 ug/m3) a significant decrease in birth weight was associated with infants in the highest quartile
of PM25 exposure as compared to infants exposed in the lowest quartile.
2-78
-------
appropriate protection against effects observed in epidemiological studies has focused on the
central question of identifying the range of PM2.5 concentrations below the long-term mean
concentrations where we continue to have confidence in the associations observed in
epidemiological studies.
In considering the evidence, we recognize that NAAQS are standards set so as to provide
requisite protection, neither more nor less stringent than necessary to protect public health with
an adequate margin of safety. This judgment, ultimately made by the Administrator, involves
weighing the strength of the evidence and the inherent uncertainties and limitations of that
evidence. Therefore, depending on the weight placed on different aspects of the evidence and
inherent uncertainties, considerations of different alternative standard levels could be supported.
As discussed in section 2.1.3, by applying the general policy approach used in previous
reviews, one could focus on identifying alternative standard levels for the annual standard that
are somewhat below the long-term mean PM2.5 concentrations reported in the long- and short-
term exposure studies. By expanding this approach to consider additional population-level
information from epidemiological studies, one could also focus on the range of PM2.s
concentrations below the long-term mean concentrations over which we continue to have
confidence in the associations observed in epidemiological studies, as well as the extent to which
our confidence in the associations is appreciably less at lower concentrations, to identify
alternative annual standard levels. Figure 2-8 provides a summary of the long-term mean PM2 5
concentrations for the long- and short-term exposure studies presented in Figures 2-4 through 2-6
as well as the range of PM2.5 concentrations corresponding with the 25th to 10th percentiles of
health event or number of study participants data from the four multi-city studies, for which
distributional statistics are available and presented in Figure 2-7.
Given the currently available evidence and considering the various approaches discussed
above, as an initial matter we conclude it is appropriate to focus on an annual standard level
within a range of about 12 to 11 |ig/m3. As illustrated in Figure 2-8, a standard level of 12
|ig/m3, at the upper end of this range, is somewhat below the long-term mean PM2.5
concentrations reported in all the multi-city long- and short-term exposure studies that provide
evidence of positive and statistically significant associations with health effects classified as
having evidence of a causal or likely causal relationship, including premature mortality and
hospitalizations and emergency department visits for cardiovascular and respiratory effects as
well as respiratory effects in children. Further, a level of 12 |ig/m3 would reflect consideration
of additional population-level information from such epidemiological studies in that it is the
PM2.s concentration generally corresponding with the 25th percentile of the available
distributions of health events data, which we consider to be at the high end of the range of PM2.5
concentrations within which the data become appreciably more sparse and, thus, where our
2-79
-------
Figure 2-8. Translating Epidemiological Evidence from Multi-City Exposure Studies into Annual PM2.s Standards
Causal i Likely Causal Long-Term Exposure Studies
Miller et nl., 2007 (WHI, 36 cities)
Eflm eta. 2008 (MCAPS-AC5 SUSS. 1 10 MLPtles)
GOES et al.. 2004 (Cystic Fibrosis)
McCormsn et al. 2305; Cauderman et al., 20C4 (S CA CHS. 12 com-ranities)
Krewski et al,, 2009 (ACS-Reanalysls II, 116 MSAs)
2=sere: al. 2006 (WCAPS-Ea=l 421 coLiife)
E«m et al. 2D06 (MCAFS-Harvatd Six Cities sites)
Lipfett et al.. 2OOG (Veterans Stjdy>
Dockeiy e at. 1996. Raizwine « al. 1996 (2«Mies Sludy
Laften et : 2006 -:-.:.:: Els cities)
Causal i Likely Causal - Short-Term Exposure Studies
Burnett«l a .. 2004 12 Canadian Cities..
Bell et al,. 2008 (MCAPS. 202 counties)
Zanobetti & Schwartz, 2009 (112 cities)
Buroett& GoHberj. 20M •& Canadian cities)
Domiufci et al.. 2006 (MCAPS. 204 oouittas)
Klemin & Mascn, 2003 (Haivana six cities)
Frarklin etaL 2C08 (2E US cities)
Frarklin et al.. 2C07 (27 US aties)
Suggestive - Long-Tern Exposure SMdies
Be I it al.. 2007 (kj-.v birth.ve ghl>
Liu etal.. 2007(WGR)
et al.. 20G8 (nfant mortaliti')
•Additional studies report effects at higher long-term mean
concentrations
•More limited and mxed evidence is availablefron single-city
short-term exposure studies with long-term mean PV^ 5
concentrations below 15 pg/m1
•,_,• w
•
4
A
•
+
+
+
+
+
+
•
•
,
A
A
A
A
A
•
^ Level of cu
rrent annual PM25
standard
•
A
Long-term Mean
PM, 5 Concentrations
(50th percentile)
4 = long-term exposure
studies; causal or likely
causal relationship
A = short-term exposure
studies; causal or likely
causal relationship
• = long-term exposure
studies; suggestive of a
causal relationship
Distributional Statistics
of Health Event and/or
Population Data
25* percentile
° 10* percentile
12 13 14 15
2-80
-------
confidence in the associations observed in epidemiological studies starts to become appreciably
less. In addition, a level of 12 |ig/m3 would reflect some consideration of studies that provide
more limited evidence of reproductive and developmental effects, which are suggestive of a
causal relationship, in that it is about at the same level as the lowest long-term mean PM2.5
concentrations reported in such studies (see Figure 2-8).
Alternatively, an annual standard level of 11 |ig/m3, at the lower end of this range, is well
below the lowest long-term mean PM2 5 concentrations reported in all multi-city long- and short-
term exposure studies that provide evidence of positive and statistically significant associations
with health effects classified as having evidence of a causal or likely causal relationship. A level
of 11 |ig/m3 would reflect placing more weight on the distributions of health event and
population data, in that this level is within the range of PM2 5 concentrations corresponding to the
25th and 10th percentiles of all the available distributions of such data. In addition, a level of 11
|ig/m3 is somewhat below the lowest long-term mean PM2 5 concentrations reported in
reproductive and developmental effects studies that are suggestive of a causal relationship.
Thus, a level of 11 |ig/m3 would reflect an approach to translating the available evidence that
places relatively more emphasis on margin of safety considerations than would a standard set at a
higher level. Such a policy approach would tend to weigh uncertainties in the evidence in such a
way as to avoid potentially underestimating PM2.s-related risks to public health. Further,
recognizing the uncertainties inherent in identifying any particular point at which our confidence
in reported associations becomes appreciably less, we conclude that the available evidence does
not provide a sufficient basis to consider alternative annual standard levels below 11 |ig/m3.
We have also considered the extent to which the available evidence provides a basis for
considering alternative annual standard levels above 12 |ig/m3. As discussed below, we
conclude that it could be reasonable to consider a standard level up to 13 |ig/m3 based on a
policy approach that tends to weigh uncertainties in the evidence in such a way as to avoid
potentially overestimating PM2.5-related risks to public health, especially to the extent that
primary emphasis is placed on long-term exposure studies as a basis for an annual standard level.
A level of 13 |ig/m3 is somewhat below the long-term mean PM2 5 concentrations reported in all
but one of the long-term exposure studies providing evidence of positive and statistically
significant associations with PM25-related health effects classified as having a causal or likely
causal relationship. As shown in Figure 2-8, the one long-term exposure study with a long-term
mean PM2 5 concentration just below 13 |ig/m3 is the WHI study (Miller et al., 2007).64 We note
that in comparison to other long-term exposure studies, the WHI study was more limited in that
64 As noted in section 2.2.1 and Table 2-4, WHI study investigators provided EPA with updated air quality data
indicating the long-term mean PM25 concentration for this study was 12.9 ug/m3, not 13.5 ug/m3 as originally
reported (Curl, 2009).
2-81
-------
it was based on only one year of air quality data. Thus, to the extent that less weight is placed on
the WHI study than on other long-term exposure studies with more robust air quality data, a level
of 13 |ig/m3 could be considered as being protective of long-term exposure related effects
classified as having a causal or likely causal relationship. In also considering short-term
exposure studies, we note that a level of 13 |ig/m3 is below the long-term mean PM2.5
concentrations reported in most such studies, but is above the long-term means of 12.8 and 12.9
|ig/m3 reported in Burnett et al. (2004) and Bell et al. (2008), respectively. In considering these
studies, we find no basis to conclude that these two studies are any more limited or uncertain
than the other short-term studies shown in Figure 2-8, noting in particular that Bell et al. (2008)
is a large study of older adults, a susceptible population. On this basis, as discussed below, we
conclude that consideration of an annual standard level of 13 |ig/m3 would have implications for
the degree of protection that would need to be provided by the 24-hour standard, such that taken
together the suite of PM2.5 standards would provide appropriate protection from effects on public
health related to short-term exposure to PM2.5.
We also note that a standard level of 13 |ig/m3 would reflect a judgment that the
uncertainties in the epidemiological evidence in general, including uncertainties related to the
heterogeneity observed in the epidemiological studies in the eastern versus western parts of the
U.S., the relative toxicity of PM2.5 components, and the potential role of co-pollutants, are too
great to warrant placing any weight on distributions of health event and population data that
extend down below the mean into the lower quartile of the data. This level would also reflect a
judgment that the evidence from reproductive and developmental effects studies that are
suggestive of a causal relationship is too uncertain to support consideration of any lower level.
Taken together, staff concludes that consideration of alternative annual standard levels in
the range of 13 to 11 |ig/m3 may be appropriate, although we also conclude that the currently
available evidence most strongly supports consideration of an alternative annual standard level in
the range of 12 to 11 |ig/m3. We conclude that an alternative level within the range of 12 to 11
|ig/m3 would more fully take into consideration the available information from all long- and
short-term PM2.5 exposure studies, including studies of susceptible populations, than would a
higher level. This range would also reflect placing weight on information from studies that helps
to characterize the range of PM2 5 concentrations over which we continue to have confidence in
the associations observed in epidemiological studies, as well as the extent to which our
confidence in the associations is appreciably less at lower concentrations.
• What alternative standard levels are appropriate to consider for a 24-hour standard
intended to supplement the protection afforded by an annual standard?
As recognized in section 2.1.3, an annual standard intended to serve as the primary means
for providing protection for effects associated with both long- and short-term PM2.5 exposures is
2-82
-------
not expected to provide appropriate protection against the effects of all short-term PM2.5
exposures. Of particular concern are areas with high peak-to-mean ratios possibly associated
with strong local or seasonal sources, or PM2 5-related effects that may be associated with
shorter-than-daily exposure periods. As a result, we believe it is appropriate to consider
alternative 24-hour PM2.5 standard levels that would supplement the protection provided by an
annual standard.
As outlined in section 2.1.3, we consider the available evidence from short-term PM25
exposure studies, as well as the uncertainties and limitations in that evidence, to assess the
degree to which alternative annual and 24-hour PM2.5 standards can be expected to reduce the
estimated risks attributed to short-term fine particle exposures. In considering the available
epidemiological evidence, we take into account information from multi-city studies as well as
single-city studies. We look both at the distributions of 24-hour PM2.5 concentrations, focusing
on the 98th percentile air quality values to match the form of the 24-hour standard (section
2.3.3.2), to the extent such data are available, as well as the long-term mean PM2.5
concentrations.
In considering the information provided by the short-term exposure studies, we recognize
that to the extent these studies were conducted in areas that likely did not meet one or both of the
current standards, such studies do not help inform the characterization of the potential public
health improvements of alternative standards set at lower levels.65 Therefore, in considering the
short-term exposure studies to inform staff conclusions regarding levels of the 24-hour standard
that are appropriate to consider, we place greatest weight on studies conducted in areas that
likely met both the current annual and 24-hour standards.
With regard to multi-city studies that evaluated effects associated with short-term PM2 5
exposures, as summarized in Figure 2-6, we observe an overall pattern of positive and
statistically significant associations in studies with 98th percentile values averaged across study
areas in the range of 45.8 to 34.2 |ig/m3 (Burnett et al., 2004; Zanobetti and Schwartz, 2009; Bell
et al., 2008; Dominici et al., 2006a, Burnett and Goldberg, 2003; Franklin et al., 2008).66 We
65 We recognize that in considering studies conducted in areas that likely did not meet the current annual standard,
by reducing the long-term mean PM2 5 concentrations in such areas to just meet the current annual standard, one
could reasonably anticipate additional public health protection from short-term exposure (as well as long-term
exposure). Further, we recognize that in considering studies conducted in areas that likely did not meet the current
24-hour standard, by reducing the upper end of the air quality distributions in order to meet the current 24-hour
standard, one could similarly reasonably anticipate additional public health protection.
66 We note that Figure 2-6 also includes one multi-city study conducted in areas that likely did not meet either the
annual or 24-hour standards (Franklin et al., 2007). To the extent that changes in air quality designed to meet the
current annual and/or 24-hour standard are undertaken, one could reasonably anticipate additional public health
protection will occur in these study areas. Figure 2-6 also includes one multi-city study for which the ISA reports a
long-term mean PM2 5 concentration but does not report information on the 98th percentile value (Klemm and
Mason, 2003). We do not know whether this study was conducted in areas that likely would have met the current
2-83
-------
note that, to the extent air quality distributions were reduced to reflect just meeting the current
24-hour standard, additional protection would be anticipated for the effects observed in the three
multi-city studies with 98th percentile values greater than 35 |ig/m3 (Burnett et al., 2004; Burnett
and Goldberg, 2003; Franklin et al., 2008). In the three additional studies with 98th percentile
values below 35 |ig/m3, specifically 98th percentile concentrations of 34.2, 34.3, and 34.8 |ig/m3,
we note that these studies reported long-term mean PM2.5 concentrations of 12.9, 13.2, and 13.4
|ig/m3, respectively (Bell et al., 2008; Zanobetti and Schwartz, 2009; Dominici et al., 2006a). To
the extent that consideration is given to revising the level of the annual standard, as discussed
above in section 2.3.4.1, we recognize that potential changes associated with meeting such an
alternative annual standard would result in lowering risks associated with both long- and short-
term PM2 5 exposures. Consequently, in considering a 24-hour standard that would work in
conjunction with an annual standard to provide appropriate public health protection, we note that
to the extent that the level of the annual standard is revised to within a range of 13 to 11 |ig/m3,
in particular in the range of 12 to 11 |ig/m3, as discussed above, additional protection would be
provided for the effects observed in these multi-city studies.
Taken together, staff concludes that the multi-city, short-term exposure studies generally
provide support for retaining the 24-hour standard level at 35 |ig/m3, specifically, in conjunction
with an annual standard level revised to within a range of 12 to 11 |ig/m3. Alternatively, in
conjunction with an annual standard level of 13 |ig/m3, we conclude that the multi-city studies
provide limited support for revising the 24-hour standard level somewhat below 35 |ig/m3, such
as down to 30 |ig/m3, based on one study (Bell et al., 2008) that reported positive and statistically
significant effects with an overall 98th percentile value below the level of the current 24-hour
standard in conjunction with an overall long-term mean concentration slightly less than 13
|ig/m3.
In reaching staff conclusions regarding alternative 24-hour standard levels that are
appropriate to consider, we also take into account relevant information from single-city studies
that evaluated effects associated with short-term PM2.5 exposures. We recognize that these
studies may provide additional insights regarding impacts on susceptible populations and/or on
areas with isolated peak concentrations. Although, as discussed above, multi-city studies have
advantages over single-city studies in terms of statistical power to detect associations and
24-hour PM2 5 standard and, therefore, this study does not inform staff conclusions regarding alternative standard
levels that are appropriate to consider. Further, we note one study reported no association between PM2 5 and
respiratory symptoms in children in areas that likely would have met the current annual standard but not the current
24-hour standard (O'Connor et al., 2008). This study reported effects on lung function in asthmatic children
associated with 5-day average PM25 concentrations but not associated with 1-day average PM25 concentrations (US
EPA, 2009a, section 6.3.2.1).
2-84
-------
broader geographic coverage as well as other factors such as less likelihood of publication bias,
reflecting differences in PM2.5 sources, composition, and potentially other factors that could
impact PM2 5-related effects, multi-city studies often present overall effect estimates rather than
single-city effect estimates. Since short-term air quality can vary considerably across cities, the
extent to which effects reported in multi-city studies are associated with short-term air quality in
any particular location is uncertain, especially when considering short-term concentrations at the
upper end of the distribution of daily PM2.5 concentrations (i.e., at the 98th percentile value). In
contrast, single-city studies are more limited in terms of power and geographic coverage but the
link between reported health effects and the air quality in a given study area is more
straightforward to establish. Therefore, we also consider evidence from single-city, short-term
exposure studies to inform staff conclusions regarding alternative levels that are appropriate to
consider for a 24-hour standard that is intended to provide supplemental protection in areas
where the annual standard may not provide an adequate margin of safety against the effects of all
short-term PM2.5 exposures.
As discussed above for the multi-city studies, we look both at the 24-hour PM2.5
concentrations in the single-city studies, focusing on the 98th percentile air quality values, as well
as the long-term mean PM2.5 concentrations. We consider single-city studies conducted in areas
that would likely have met the current suite of PM2 5 standards as most useful for informing staff
conclusions related to the level of the 24-hour standard, as summarized in Figure 2-9.67 We note
that additional single-city studies summarized in Figure 2-9 were conducted in areas that would
likely have met one but not both of the current PM2.s standards,68 To the extent changes in air
quality designed to just meet the current suite of PM2 5 standards are undertaken, one could
reasonably anticipate additional public health protection will occur in these study areas.
67 We note there are additional single-city studies discussed in the ISA beyond the short-term exposure studies
discussed in this section that were conducted in areas that would likely not have met both the current annual and 24-
hour standards (e.g., Ito et al., 2007; Sheppard, 2003; Burnett, 1997). To the extent that changes in air quality
designed to just meet the current annual and/or 24-hour standard are undertaken, one could reasonably anticipate
additional public health protection will occur in these study areas For another group of studies, we have no
information on the distribution of air quality concentrations, specifically information on the upper end of the
distribution (i.e., the 98th percentile values) (e.g., NYDOH, 2006; Sullivan et al., 2005; Zanobetti and Schwartz,
2006; Pope et al., 2006; Dockery et al., 2005; Rich et al., 2005; Villeneuve et al, 2003; Fung et al., 2006; Chen et al,
2004; Chen et al., 2005 Slaughter et al, 2003; Chimonas and Gessner, 2007). Therefore, we do not know whether
these studies were conducted in areas that would likely have met the current 24-hour PM2 5 standard. Collectively,
staff concludes these two groups of studies are not helpful to inform staff conclusions regarding alternative standard
levels that are appropriate to consider.
68 This group of studies included two studies conducted in areas that would likely have met the current annual
standard but not the current 24-hour standard (Santa Clara, California and the Wasatch Front, UT) that reported
positive and statistically significant associations between short-term PM2 5 exposures and mortality and
cardiovascular-related hospitalizations, respectively (Fairley, 2003; Pope et al., 2008) and one study conducted in an
area that would likely have met the current 24-hour standard but not the current annual standard (Coachella Valley,
California) that reported no association between short-term PM2 5 exposures and cardiovascular -related mortality
(Ostro et al., 2003).
2-85
-------
Figure 2-9. Summary of Effect Estimates (per 10 ug/m ) and Air Quality Distributions for Single-City, Short-term PM2.s
Exposure Studies
SfudylCite
lisabeth e! at. 2038
Rabinovich etal. 2006
Stebetal.2000
Vilteneuveetal.2006
Mar etal. 2004
Rahnovitch el ai.2004
Slaughter etal. 2005
Pope elal. 2008
Schreudef etal .2006
Peters etal. 2001
Deffino etal. 1997
Wilson, etal. 2007
Mar etal. 2003
Farley 2003
Osto eta!. 2003
Geographic Area
Nueoes County. TX
Denver
St. John. Canada
Edmonion, Canada
Spokane
Denver
Spokane
Wasatch Front UTe
Spokane
Boston
Montreal
Phoenix-central
Phoenix-middle
Phoenix-outer
Phoenix
Santa Clara Co. CA
Coachella Valley. CA
Years of
Air
Quality
Ma
2001-2005
2 winters
2001-2003
1992-1996
1992-2002
1997-19&9
3 winters
1999-2002
1995-2001
1993-2006
1995-2002
1995-1996
Summer
1993
1995-1997
1995-1997
1989-1996
1996-1998
Endpoint
Isdiemic Stroke/TIA HA
Asthma Medication Use (children)
CV ED Visits
Respiratory ED Visits
Hem hrgc stroke HA
Isdiemic strcfe HA
TiA stroke HA
Respiratory Sympiany) adults
Respiratory Symp (any) children
Asthma exacerbation (children)
CVD HA and ED visits
Respiratory HA and EO Visits
CHFHA
Carctec HA
Respiratory ED
Ml HA
Respiratory HA
CV mortality
CV morfatiiy
Nonaccidental mortality
CV mortality
Air Quality Data
(jjg/m )
Mean
7.0"
74
8,5
8.5
9.8C
10.6d
10.8
10.8c
10*
12.1
12.1
13.0
13.5"
13.6
15.8*
98th
percentile"
13.6
17.2
27.3
240
25.8
29.3
29.6
44.5
296
28.2
31.2
31.6
32.2
59.0
33.8
Effect Estimate (95% Cl)
\ m J
' *
: • |
: "
• |
!
* ( |
' * i
• '
^»_
• :
, „. .^p, .. I
_»_
•
, j
•
— r-
:~
, .
• : I
0.7
1.-0
i")gde«, S:ilt I ak* C in* Ps
2-86
-------
Therefore, staff concludes that these studies are not helpful to inform staff conclusions regarding
alternative standard levels that are appropriate to consider.
With regard to single-city studies that were conducted in areas that would likely have met
both the current 24-hour and annual standards, we first look to studies that reported positive and
statistically significant associations with short-term PM2.5 exposures. In considering this group
of studies, we note Mar et al. (2003) reported a positive and statistically significant association
for premature mortality in Phoenix with a long-term mean concentration of 13.5 |ig/m3 in
conjunction with a 98th percentile value of 32.2 |ig/m3. To the extent that consideration is given
to revising the level of the annual standard, within a range of 13 to 11 |ig/m3, as discussed above,
additional protection would be provided for the effects observed in this study.
Four additional studies reported positive and statistically significant associations with
98th percentile values within a range of 31.2 to 25.8 |ig/m3 and long-term mean concentrations
within a range of 12.1 to 8.5 |ig/m3 (Delfmo et al., 1997; Peters et al., 2001; Stieb et al., 2000;
and Mar et al., 2004). Delfmo et al. (1997) reported statistically significant associations between
PM2.s and respiratory emergency department visits for older adults (greater than 64 years old) but
not young children (less than 2 years old), in one part of the study period (summer 1993) but not
the other (summer 1992). Peters et al. (2001) reported a positive and statistically significant
association between short-term exposure to PM2.5 (2-hour and 24-hour averaging times) and
onset of acute MI in Boston.69 Stieb et al. (2000) reported positive and statistically significant
associations with cardiovascular- and respiratory-related emergency department visits in Saint
John, Canada, in single pollutant models but not in multi-pollutant models (US EPA, 2004, pp 8-
154 and 8-252 to 8-253). Mar et al. (2004) reported a positive and statistically significant
association for short-term PM2.5 exposures in relation to respiratory symptoms among children
but not adults in Spokane, however, this study had very limited statistical power because of the
small number of children and adults evaluated.
We then consider short-term single-city PM2 5 exposure studies that reported positive but
nonstatistically significant associations for cardiovascular and respiratory endpoints in areas that
would likely have met both the current 24-hour and annual standards. The 98th percentile values
reported in these studies ranged from 31.6 |ig/m3 to 17.2 |ig/m3 and the long-term mean
concentrations ranged from 13.0 to 7.0 |ig/m3. These studies included consideration of
cardiovascular-related mortality effects in Phoenix (Wilson et al., 2007), asthma medication use
in children in Denver (Rabinovitch et al., 2006), hospital admissions for hemorrhagic and
69 A King County, WA (Seattle) study using similar methods to Peters et al. (2001) reported no association with
acute onset MI and PM25 concentration with a long-term mean concentration of 12.8 ug/m3 (Sullivan et al., 2005).
We do not have information on the 98th percentile value for the Seattle study. It is unclear whether the differences
observed in these two studies were due to regional differences in population characteristics and/or air pollution
sources/PM25 components (US EPA, 2009a, section 6.2.10.4).
2-87
-------
ischemic stroke in Edmonton, Canada (Villeneuve et al., 2006), and hospital admissions for
ischemic stroke/transient ischemic attack in Nueces County, TX (Lisabeth et al., 2008).
Lastly, we consider single-city studies conducted in areas that would likely have met both
the current 24-hour and annual standards that reported null findings. The 98th percentile values
reported in these studies ranged from 29.6 |ig/m3 to 24.0 |ig/m3 and the long-term mean
concentrations ranged from 10.8 to 8.5 |ig/m3. These studies reported no associations with short-
term PM2.s exposures and cardiovascular-related hospital admissions and respiratory-related
emergency department visits (Slaughter et al., 2005) and cardiovascular-related emergency
department visits (Schreuder et al., 2006) in Spokane; asthma exacerbation in children in Denver
(Rabinovitch et al., 2004); and hospital admissions for transient ischemic attack in Edmonton,
Canada (Villeneuve et al., 2006).
Viewing the evidence as a whole, we observe a limited number of single-city studies that
reported positive and statistically significant associations for a range of health endpoints related
to short-term PM2.5 concentrations in areas that would likely have met the current suite of PM2.5
standards. Many of these studies had significant limitations (e.g., limited statistical power,
limited exposure data) as briefly identified above and discussed in more detail in the ISA (US
EPA, 2009a, chapter 6). Other studies reported positive but not statistically significant results or
null associations also in areas that would likely have met the current suite of PM2 5 standards. In
addition, still other studies (e.g., those conducted in Phoenix, Denver, and Edmonton) reported
mixed results within the same study areas. Overall, the entire body of results from these single-
city studies is mixed, particularly as 24-hour 98th percentile concentrations go below 35 |ig/m3.
Although a number of single-city studies report effects at appreciably lower PM2 5
concentrations than multi-city short-term exposure studies, the uncertainties and limitations
associated with the single-city studies were greater and, thus, we have less confidence in using
these studies as a basis for setting the level of a standard. Therefore, we conclude that the multi-
city short-term exposure studies provide the strongest evidence to inform decisions on the level
of the 24-hour standard, and the single city studies do not warrant consideration of 24-hour levels
different from those supported by the multi-city studies.
In addition to considering the epidemiological evidence, we look to air quality
information based on county-level 24-hour and annual design values to understand the
implications of the alternative standard levels supported by the currently available scientific
evidence, as discussed above. As outlined in section 2.1.3, we recognize that changes in PM2 5
air quality designed to meet an annual standard would likely result not only in lower annual
mean concentrations but also in fewer and lower peak 24-hour concentrations. We also
recognize that changes designed to meet a 24-hour standard would result not only in fewer and
lower peak 24-hour concentrations (especially when coupled with a high percentile-based form,
2-88
-------
such as the 98th percentile) but also in lower annual mean concentrations. As discussed in
section 2.1.3, we conclude that a policy goal which includes setting the annual standard to be the
"generally controlling" standard in conjunction with setting the 24-hour standard to provide
supplemental protection to the extent that additional protection is warranted, is the most effective
and efficient way to reduce total population risk associated with both long- and short-term PM2.5
exposures, resulting in more uniform protection across the U.S than the alternative of setting the
24-hour standard to be the controlling standard. Therefore, we consider the extent to which
different combinations of alternative annual and 24-hour standard levels based on the evidence
would support this policy goal.
Figure 2-10 presents a scatter plot of annual and 24-hour design values in urban areas
across the U.S., color-coded by geographic region. This figure provides a visual perspective of
whether the annual or 24-hour standard is likely to be the controlling standard for various
combinations of standards. In Figure 2-10, the horizontal lines represent alternative 24-hour
standard levels (i.e., 35 or 30 |ig/m3) with a 98th percentile form, averaged over three years, while
the vertical lines represent alternative annual standard levels (i.e., 13, 12, or 11 |ig/m3), using an
annual arithmetic mean averaged over three years. To understand this figure, we note that there
are four quadrants that distinguish counties that are likely to have met or not met one or both of
the PM2.5 standards. The lower left quadrant characterizes counties that would likely meet both
the annual and 24-hour standards. The lower right quadrant characterizes counties that would
likely meet the 24-hour standard but not the annual standard, representing counties where the
annual standard would be the controlling standard. The upper left quadrant characterizes
counties that would likely exceed the 24-hour standard but not the annual standard, representing
counties where the 24-hour standard would be the controlling standard. Finally, the upper right
quadrant characterizes counties that would likely exceed both the annual and 24-hour standards.
In this quadrant, the diagonal lines that intercept the origin and the intersection of a suite of
alternative standard levels (e.g., the "11/35" line) represents the point of demarcation between
those counties for which the 24-hour standard would be controlling (to the left of the diagonal
line) and those for which the annual standard would be controlling (to the right of the diagonal
line). More specifically, for counties to the left of the diagonal line, 24-hour concentrations
would need to be reduced by a greater proportion to meet the 24-hour standard than the
proportional reduction needed in the annual average concentrations to meet the annual standard.
2-89
-------
Figure 2-10. County-level 24-hour DVs versus Annual DVs, 2007-2009
ro
CD
_D
5
CD
D
i—
13
O
I
70
65:
60 -
55 :
50 -
45
40 -
35
30
25
20
10:
5-
Counties in this area of the plot would
likely exceed the 24-hr standard only
_
Counties in this area of the plot
would likely exceed both the
annual and 24-hr standards
Annual level
Counties in this area of the
plot would likely meet both the
annual and 24-hour standards
Counties in this area of the
plot would likely exceed the
annual standard only
01234567
10 11 12 13 14 15 16 17 18 19 20 21 22 23
Annual Design Value (ug/m3)
Symbol color indentifies
geographic region:
Northeast
Southeast
Upper Midwest
Northwest
So. CA
Other (AK, HI. VI, PR)
Source: Schmidt, 201 la, Analysis B
2-90
-------
Using information on the relationship of the 24-hour and annual design values, we
examine the implications of three alternative suites of PM2.5 standards identified as appropriate
to consider based on the currently available scientific evidence, as discussed above. As seen in
Figure 2-10, an alternative suite of PM2.5 standards that would include an annual standard level
of 11 or 12 |ig/m3and a 24-hour standard with a level of 35 |ig/m3 (i.e., 11/35 or 12/35), would
result in the 24-hour standard as the generally controlling standard in the Northwest (i.e.,
counties identified as light blue dots in Figure 2-10) but not generally in other regions across the
country. These Northwest counties generally represent areas where the annual mean PM2.5
concentrations have historically been low but where relatively high 24-hour concentrations
occur, often related to seasonal wood smoke emissions. Alternatively, combining an alternative
annual standard of 13 |ig/m3 with a 24-hour standard of 30 |ig/m3 would result in many more
areas across the country in which the 24-hour standard would likely become the controlling
standard than if an alternative annual standard of 12 or 11 |ig/m3 were paired with the current
level of the 24-hour standard (i.e., 35 |ig/m3). This can be seen by comparing the area to the left
of the "13/30" line to the areas to the left of the "12/35" and "11/35" lines, in particular in the
upper right quadrant in Figure 2-10, which represents the regions on the graph in which the 24-
hour standard would likely be controlling.
Summary of Evidence-based Considerations to Inform the 24-Hour Standard Level
In considering the currently available scientific information and air quality information
within the context of identifying potential alternative standard levels for consideration we have
identified a range of alternative 24-hour standard levels that is appropriate to consider based on
the general approach outlined in section 2.1.3. Our general approach for identifying alternative
24-hour standard levels that would protect against effects observed in epidemiological studies
has focused on setting a level somewhat below the 98th percentile values where associations have
been observed in epidemiological studies, particularly in areas that would meet alternative
annual standard levels discussed above. In considering the available evidence, staff concludes it
is appropriate to consider either retaining the level of the current 24-hour standard at 35 |ig/m3 or
revising the level to somewhat below 35 |ig/m3, such as, down to 30 |ig/m3.
Consideration for retaining the level at 35 |ig/m3 would reflect placing greatest weight on
evidence from multi-city studies that reported positive and statistically significant associations
with health effects classified as having a causal or likely causal relationship. In conjunction
with lowering the annual standard level, especially within a range of 12 to 11 |ig/m3, this
alternative would recognize additional public health protection for effects associated with short-
term PM2.5 exposures would be provided by lowering the annual standard such that revision to
the 24-hour standard would not be warranted. In addition, such combinations of 24-hour and
2-91
-------
annual standards levels (i.e., 11/35 and 12/35) would be expected to result in the annual standard
being the generally controlling standard, except in areas with particularly high peak-to-mean
ratios (e.g., Northwest), with the annual standard providing protection against PM25-related
health effects associated with long- and short-term exposures, in conjunction with the 24-hour
standard providing appropriate supplemental protection. These combinations would likely result
in providing more uniform protection across the U.S than if the 24-hour standard were the
generally controlling standard.
An alternative approach to considering the evidence would recognize that the currently
available information provides some support for revising the level below 35 |ig/m3, perhaps as
low as 30 |ig/m3. This alternative 24-hour standard level would be more compatible with an
alternative annual standard of 13 |ig/m3 based on placing greater weight on one multi-city short-
term exposure study (Bell et al., 2008) that reported positive and statistically significant effects at
a 98th percentile value less than 35 |ig/m3 (i.e., 34.2 |ig/m3) in conjunction with a long-term mean
concentration less than 13 |ig/m3 (i.e., 12.9 |ig/m3). However, in considering the implications of
this combination (i.e., 13/30), staff recognizes that this option would likely result in the 24-hour
standard being the controlling standard in a large number of areas in many geographic regions,
potentially resulting in more uneven public health protection across the U.S. than if the annual
standard was the generally controlling standard.
Taken together, while we consider it appropriate to consider an alternative 24-hour
standard level within a range of 35 to 30 |ig/m3, staff concludes that the currently available
evidence most strongly supports consideration for retaining the current 24-hour standard level at
35 |ig/m3 in conjunction with lowering the level of the annual standard within a range of 12 to 11
|ig/m3.
2.3.4.2 Risk-based Considerations
Beyond looking directly at the relevant epidemiologic evidence, staff has also considered
the extent to which specific levels of alternative PM2.5 standards are likely to reduce both long-
term exposure-related mortality risk and short-term exposure-related mortality and morbidity
risk. In addition to considering the nature and magnitude of PM2.5-attributable risk remaining
under each set of alternative standards, we have also considered the nature and magnitude of risk
reductions under the alternative standards considered. These risk estimates for the set of
alternative standard levels considered are based on the same methodology used in estimating risk
for the current suite of standard levels as discussed above in section 2.2.2.
The quantitative risk assessment initially included analyses of alternative annual standard
levels of 14, 13, and 12 |ig/m3 paired with either the current 24-hour standard level of 35 |ig/m3
or with alternative 24-hour standard levels of 30 and 25 |ig/m3. The specific combinations of
2-92
-------
alternative standard levels assessed in the quantitative risk assessment included: (a) alternative
suites of standards focusing on alternative annual standard levels in conjunction with the current
24-hour standard including combinations denoted by 14/35, 13/25, and 12/35and (b) alternative
suites of standards reflecting combinations of alternative annual and 24-hour standard levels
including combinations denoted by 13/30 and 12/25. This set of alternative annual and 24-hour
standard levels was chosen prior to completion of the first draft RA and reflects consideration for
evidence related to potential PM2.5-related health effects as presented in the Integrated Science
Assessment for Paniculate Matter: Second External Review Draft (2nd draft ISA) released in
July, 2009 (US EPA, 2009b). The presentation of that evidence was refined in subsequent
iterations of the PM ISA, culminating in the Final ISA released in December 2009 (US EPA,
2009a). The range of alternative standard levels discussed in section 23.4.1 above (i.e., annual
standard levels within a range of 13 to 11 |ig/m3 and 24-hour standard levels within a range of 35
to 30 |ig/m3), reflects consideration of evidence as presented in the Final ISA and consequently
differs somewhat from the set of alternative standard levels originally selected for modeling in
the quantitative risk assessment. Subsequent to the release of the second draft RA (US EPA,
2010d), we expanded the range of alternative annual standard levels evaluated in the final RA to
include an alternative annual standard level of 10 |ig/m3 and developed risk estimates for two
additional combinations of alternative standards - 10/35 and 10/25 (US EPA, 2010a, Appendix
J).
In simulating ambient PM2.5 levels associated with these alternative standard levels, we
included a more regional spatial pattern of reductions (reflected in the use of a proportional
rollback approach) as well as more localized spatial patterns of reductions (reflected in the use of
a hybrid approach and to an even greater extent in the use of a locally focused rollback approach)
(see U.S. EPA, 2010a, section 3.2.3). While the proportional rollback approach was used in
generating the core risk estimates, the other two more localized rollback approaches were
considered as part of sensitivity analyses.
Results for the alternative suites of standards considered are presented in Figures 2-11 and 2-12,
which depict patterns in risk reduction for long-term exposure-related risk (Figure 2-11) and
short-term exposure-related risk (Figure 2-12) using different combinations of alternative
standard levels relative to the estimated risks related to simulating just meeting the current
standards. These figures include results for each of the 15 urban study areas, thereby allowing
patterns in risk reduction across alternative standard levels and urban study areas to be
considered together.70 The discussion below of the magnitude of risk remaining under simulated
70 Patterns of risk reduction across alternative annual standard levels (in terms of percent change relative to risk for
the current annual standard level) are similar for all health endpoints modeled for a particular exposure duration (i.e.,
patterns of percent risk reduction will be similar for long-term exposure related all-cause, IHD-related and
2-93
-------
attainment of the alternative standard levels is based on risk estimates presented in U.S.EPA,
2010a, section 4.2.2. The figures also depict the level of confidence, also discussed below,
associated with risk estimates generated for the current suite of standard levels as well as
alternative standard levels considered.
• What is the nature and magnitude of risk associated with just meeting the alternative
annual PM2.s standards considered?
In characterizing estimates of PM25-related risk associated with simulation of the
alternative annual standards combined with the current 24-hour standard level (i.e., 14/35, 13/35
and 12/35), we estimated both the magnitude of risk reductions (relative to risk remaining upon
just meeting the current suite of standards), as well as the risk estimated to remain upon just
meeting the alternative annual standard levels.71 We have greater overall confidence in our
estimates ofrisk reduction with simulation of just meeting an alternative standard level (relative
to risk associated with simulation of just meeting the current standards) than for estimates of
absolute risk remaining upon simulation of just meeting that alternative standard level. In the
context of long-term exposure-related mortality, this greater confidence primarily reflects the
fact that we modeled risk down to the LML of the epidemiological study providing C-R
functions (Krewski et al., 2009).72 While this introduces the potential for underestimating the
absolute risk remaining upon simulation of just meeting an alternative standard level (since the
portion or exposure associated with ambient PM2.5 concentrations below the LML is not
translated into risk), it has little to no effect on the simulation of risk reduction for that alternative
standard level, relative to risk estimated for the current standards (since this calculation involves
ambient PM2.5 concentrations that are typically above the LML and are therefore translated into
risk). In the context of short-term exposure-related mortality and morbidity, while we modeled
risk down to PRB, we still have greater uncertainty in estimating absolute risk remaining since
this requires extrapolation of the C-R function to cover lower ambient PM2.5 concentrations,
cardiopulmonary-related mortality). This reflects the fact that the C-R functions used in the quantitative risk
assessment are close to linear across the range of ambient air concentrations evaluated. However, estimated
incidence will vary by health endpoint.
71 Our analysis also included the assessment of risk associated with an alternative annual standard level of 10
ug/m3.US EPA, 2010a, Appendix J).
72 As discussed in section 3.1.1 of the RA (U.S. EPA, 2010a), we did not model long-term exposure-related risk
below the LML due to concerns over uncertainty in extrapolating the C-R functions below the range of PM2 5
concentrations reflected in the epidemiological study providing those C-R functions. However, as stated in the ISA,
there is no evidence for a discernible threshold (US EPA, 2009a, section 2.4.3) and our decision not to estimate risk
below the LML should not be construed as support for the existence of a threshold. Therefore, the decision not to
estimate risk below the LML, while reasonable given our desire to generate higher-confidence estimates of risk,
does introduce low-bias into the estimate of absolute risk remaining upon simulation of each of the alternative
standard levels considered, since evidence suggests that risk does extend to long-term PM2 5 concentrations below
the LML.
2-94
-------
Figure 2-11 Percent reduction in long-term exposure-related mortality risk (alternative standards relative to the current standard)
(Note: inset shows PM25 related incidence and percent of total incidence for IHD mortality under the current suite of standards*)
60%
40%
20%
Alanta, GA 277 (227 -324); 16.7% (13,7% -19.5%)
Baltimore,MD 374 (307-440); 14.71 (12.1%- 17.3%)
Biimingham. AL 165 (135-194); 13.8% (11.3% - 16.2%)
Dallas. TX 247 (202-291): 11.4% (9.3% -13.4%)
Detroit.Ml 478 (390-563); 11.5% (9.4% -13.5%)
Fresno. CA 98 (80 -116); 8.5% (7% • 10.1%)
Houston. TX 434(355-511); 13.6% (11.1%-16%)
Los Angeles. CA 1094 (890 • 1296); 7.7% (B.3% - 9,1%)
New York. NY 2222 (1814 • 2620); 11.8% (9.6% - 13.9%)
Philadelphia. PA 330 (270 - 389); 13.2% (10.8% - 15.6%)
Phoenix. AZ 402 (327 -476); 8.5% (5.9% • 10.1%)
Pittsburgh. PA 324 (264 • 382); 11.8% (9.6% • 13.9%)
Salt Lake City. UT 19 (16-23); 3.7% (3% - 4.4%)
St. Louis.MO 563 (461.662); 14.2% (11.6%-16.7%)
Tacoma. WA 49 (40 - 58): 4.7% (3.8% - 5.6%)
1&35"
14/35
13O5
1205
1380
C intent and AHaiMtlveStjmd«
-------
Figure 2-12. Percent Reduction in Short-term Exposure-related Mortality and Morbidity Risk (alternative standards relative to the
current standards) (Note: inset shows PM25 related incidence and percent of total incidence for CV mortality under the current suite of
standards*)
I
35%
30%
25%
20%
-•- MM.OA32 (-33-06): 9»\ (-OJ8V2.41)
-•- Mfciwn.MD 03 (-4. 120]; 1J« (-0.1%. 32*)
—- nnfctfnm.M. -1 (42 -«X Ot <-ljS* •
DMto.TY 30 (-19-79J 04% (-0« -32Y)
-»- Dwek.M 00 (4- 127J I* (-0.1 »-*.«)
— Ftam.CA 13 (
< Hounon.W * {41 • I32X 04% (04* -2.+*)
— • Uw Argthi . CA -30 (•!» -72); -0.2% (-07% -04«)
f.NY 473(278- Me* 2.1« (12H-3*)
.PA 84 (22- 1*5) 2.1% (P«V3JB%)
(4-170);
-^ 8* l*i OlyiUT 0 (-2-30) 0.91 (-02%
-•- 9t.LoUfl.MO 106 CM- I87J l.fltt
11 (4.37J 0.7% (-0^4% - UK)
15%
10%
Lmlof"
confidence In ilsk
Mlimatee
Highest
Somewhat lower
Substantially lower
15/35"
14OS
13/35
12O5
13/30
12OS
RKWitAirQuilHy.Cui
*Based on Zanobetti and Schwartz (2009). The legend contains, for each urban area, the incidence estimate (and 95% CI) and the estimate of percent of total incidence (and 95%
CI) under the current standards. While incidence and percent of total incidence estimates are provided specifically for CV-related mortality, the percent reduction plots provided in
the figure apply to all short-term exposure-related mortality and morbidity categories assessed - see text.
* *The current standards consist of an annual standard of 15 ng/m3 and a daily standard of 3 5 ng/m3. Combinations of an annual standard (n) and a 24-hour standard (m) are
denoted n/m in this figure.
*** Although short-term exposure-related risk estimates differ from long-term exposure-related risk estimates in that the former are estimated down to policy-relevant background,
general observations regarding the level of confidence in risk estimates related to the composite monitor PM25 levels involved in risk estimation (i.e., decreased confidence
associated with lower composite monitor values) generally apply in the case of short-term exposure-related mortality and morbidity estimates.
2-96
-------
relative to the estimation of risk reduction, which involves assessing risk for higher ambient
PM2 5 concentrations.
Because we have greater confidence in our estimates of the magnitude of risk reduction
associated with alternative standard levels, we emphasize this risk metric in discussing risk
estimates below. While we do present estimates of absolute risk remaining for each of the
alternative standard levels, this is done with the caveat that, particularly in the context of long-
term exposure-related mortality, these estimates may be biased low.
In discussing the estimated risks, we focus on the set of urban study areas experiencing
risk reductions under each alternative annual standard modeled. Key policy-relevant
observations associated with these risk estimates include:
• Magnitude of estimated reductions in long-term exposure-related mortality risk:
Upon simulation of just meeting the alternative annual standard levels considered (14,
13, and 12 |ig/m3) in conjunction with the current 24-hour standard (35 |ig/m3,
denoted as alternative suites of standards of 14/35, 13/35 and 12/35), the core analysis
estimated reductions in long-term exposure-related mortality for 12 of the 15 urban
study areas, with the degree of estimated risk reduction increasing incrementally
across the alternative standard levels (both in terms of the number of study areas
experiencing risk reduction and the magnitude of those reductions). For the
alternative annual standard level of 12 |ig/m3 (in conjunction with the current 24-hour
standard), the core analysis estimates that these study areas have reductions in risk
(relative to risk remaining upon just meeting the current suite of standards) ranging
from about 11 to 35%.
For some of those areas in which the 24-hour standard is the generally controlling
standard, larger risk reductions would have been estimated in this case (i.e., 12/35) if
the locally-focused rollback approach had been used to simulate just meeting the
current suite of standards. This result would be expected since the magnitude of risk
remaining upon just meeting the current suite of standards would have been higher
than that estimated based on the proportional rollback approach used in the core
analysis. Therefore, while the absolute risks would not change, the percentage
difference would have been greater if we had started with higher risks related to
simulation of just meeting the current annual standard.
• Long-term exposure-related mortality risk remaining: For an annual standard level
of 14 |ig/m3, the percent of total incidence of long-term exposure-related IHD
mortality attributable to PM2.5 (i.e., risk remaining) in the 5 urban study areas
experiencing risk reductions ranged from an estimate of 9 to 15%. For an alternative
annual standard of 12 |ig/m3, estimated risk remaining in the 12 urban study areas
experiencing risk reductions ranged from 6 to 11% in terms of PM2.5-attributable
long-term exposure-related mortality. This translates into estimates of between 90
and 300 cases per year attributable to long-term PM2 5 exposure for those study areas
experiencing the greatest reductions in risk under the lowest alternative annual
standard level simulated.
2-97
-------
• Short-term exposure-related mortality and morbidity risk: For the alternative annual
standard level of 12 |ig/m3 (in conjunction with the current 24-hour standard, 12/35),
the core analysis estimated that reductions in both short-term exposure-related
cardiovascular-related mortality and morbidity risk ranged from 5 to 23%.73 In terms
of risk remaining upon simulation of 12 |ig/m3 (in conjunction with the current 24-
hour standard), the urban study areas with the greatest percent reduction have CV-
related mortality estimates ranging from 25 to 50 deaths per year.
• Simulation of risks for an alternative annual standard level below 12 jug/m :
Simulation of risks for an alternative annual standard of 10 |ig/m3 suggested that
additional risk reductions could be expected with alternative annual standards below
12 |ig/m3. However, we recognize that there is potentially greater uncertainty
associated with these risk estimates compared with estimates generated for the higher
alternative annual standards considered in the quantitative risk assessment, since
these estimates require simulation of relatively greater reductions in ambient PM2 5
concentrations. As lower ambient PM2.5 concentrations are simulated (i.e., ambient
concentrations further from recent conditions), potential variability in such factors as
the spatial pattern of ambient PM2.5 reductions (rollback) increases, thereby
introducing greater uncertainty into the simulation of composite monitor annual mean
PM2.5 concentrations and, consequently, risk estimates (US EPA, 2010a, Appendix J).
• Substantial variability in magnitude of estimated risk reduction across urban study
areas: While there is a consistent pattern of estimated risk reduction across the
alternative annual standards with lower alternative standard levels resulting in more
urban study areas experiencing increasingly larger risk reductions, there is
considerable variability in the magnitude of these reductions across study areas for a
given alternative annual standard level. This variability reflects differing degrees of
reduction in annual mean concentrations across the study areas, which results, in part,
because the study areas began with varying annual mean PM2.5 concentrations after
simulating just meeting the current suite of standards. Therefore, even if study areas
have similar "ending" annual mean PM2.5 concentrations after simulation of just
meeting a specific alternative annual standard, because the starting point in the
calculation (the annual mean PM2 5 concentrations upon just meeting the current suite
of standards) can be variable, the overall reduction in annual mean PM2.5
concentrations across the standards can also be variable. This translates into variation
in reductions in long-term exposure-related risk upon just meeting alternative annual
standard levels across the study areas.
• The nature of the spatial pattern in PM2.s reductions (reflected in the rollback
approach used) can impact the magnitude of estimated risk reductions: The
sensitivity analysis involving application of the locally focused rollback approach
revealed that the pattern of reductions in ambient PM2.5 concentrations upon just
meeting the current suite of standards can impact the magnitude of additional risk
reductions estimated for just meeting alternative (lower) annual standard levels.
73 Because the same air quality metric (annual distributions of 24-hour PM2 5 concentrations) was used in generating
short-term exposure-related mortality and morbidity endpoints, patterns of risk reduction (as a percent of risk under
the current suite of standards) are similar for both sets of endpoints (see US EPA, 20 lOa, section 4.2.2).
2-98
-------
Specifically, for those study areas with higher peak-to-mean PM2.5 ratios, application
of the locally focused rollback approach resulted in higher annual mean PM2 5
concentrations remaining upon just meeting the current suite of standards. If a
proportional rollback approach was then used to simulate just meeting alternative
annual standard levels, a greater degree of reduction in composite monitor annual
mean PM2 5 concentrations will result, since the starting point for the calculation
(annual mean PM2 5 concentrations upon just meeting the current suite of standards)
would be higher. These findings highlight the important role played by variability in
the spatial pattern of ambient PM2.5 concentrations in influencing the magnitude of
risk reductions under alternative annual standard levels.
• Based on consideration of the composite monitor annual mean PM2.s concentrations
involved in estimating long-term exposure-related mortality, we have varying levels
of confidence in risk estimates generated for the alternative annual standard levels
considered: With the exception of one study area, those study areas estimated to have
risk reductions under the alternative annual standards of 14 and 13 |ig/m3 had
simulated composite monitor annual mean PM2 5 concentrations ranging from just
below 10.6 to over 13.3 |ig/m3 (see US EPA, 2010a, Table 3-4). In other words,
these composite monitor annual mean PM2 5 concentrations generally fell well within
the range of ambient PM2 5 concentrations considered in fitting the C-R functions
used (i.e., within one SD of the mean PM2 5 concentration from 1999-2000 ACS
dataset, Krewski et al., 2009). The urban study areas estimated to have risk
reductions under the lower alternative annual standard level of 12 |ig/m3 had lower
composite monitor annual mean values ranging from 9.0 to over 11.4 |ig/m3. These
values generally extend to below one SD of the mean of the ACS dataset and
therefore, we have somewhat lower confidence in these risk estimates, relative to
those generated for the higher alternative annual standards. By contrast, urban study
areas estimated to have risk reductions under the alternative standard level of 10
|ig/m3 (paired with the current 24-hour standard) had simulated composite monitor
annual estimates ranging from 7.6 to 8.9 |ig/m3 (see US EPA, 2010a, Appendix J).
These concentrations are towards the lower end of the range of ACS data used in
fitting the C-R functions (in some cases approaching the LML) and, therefore, we
have substantially less confidence in these risk estimates, compared with those for the
higher alternative annual standards assessed. The levels of confidence associated
with risk estimates generated for the suites of alternative standard levels considered in
the quantitative risk assessment are depicted in Figure 2-11. These confidence levels
are repeated in presenting risk estimates for short-term exposure-related mortality and
morbidity in Figure 2-12, since the general observation that confidence in risk
estimates decreases as we consider lower composite monitor annual mean values also
holds for the short-term exposure-related health endpoints.
• What is the nature and magnitude of risk associated with simulating different
combinations of alternative annual and 24-hour PM2.s standards?
In characterizing PM2 5-related risks associated with simulation of alternative annual
standards combined with alternative 24-hour standards (i.e., 13/30 and 12/25), we estimated both
the magnitude of risk remaining upon just meeting these alternative standards, as well as the
2-99
-------
magnitude of risk reductions (relative to risk remaining upon just meeting the current suite of
standards). While the alternative 24-hour standard levels considered did result in estimated risk
reductions, we have substantially lower confidence in these risk estimates because of the
relatively low annual-average PM2.5 concentrations associated with simulation of these
alternative standard levels.
Of the 11 urban study areas estimated to have risk reductions under the alternative 24-
hour standard of 30 |ig/m3 (with the 24-hour standard controlling - see US EPA 2010a, Table 3-
4), composite monitor annual mean PM2.5 concentrations ranged from 6.6 to 11.3 |ig/m3 with
most of the urban study areas having concentrations in the 8 to 10 |ig/m3 range. These
concentrations extend into the lower range of PM2.5 concentrations considered in the ACS study
to fit the C-R functions and therefore, we have substantially lower confidence in these estimates.
When we consider composite monitor concentrations for urban study areas assessed to
have risk reductions under the alternative 24-hour standard level of 25 |ig/m3 (again, where the
24-hour standard is controlling), we observed composite monitor annual mean PM2.5
concentrations that are even lower, ranging from 5.6 to 11.2 |ig/m3 with most study areas having
concentrations in the range of 7 to 9 |ig/m3. Because this range extends well into the lower range
of PM2.5 concentrations considered in the ACS study to fit the C-R functions (in some cases
extending below the LML), we have substantially lower confidence in these risk estimates.
Furthermore, we find that those urban study areas with the greatest degree of estimated risk
reduction under these alternative 24-hour standard levels also had the lowest composite monitor
annual average PM2.5 levels, and therefore we have the lowest overall confidence in these results.
2.3.4.3 CASAC Advice
Based on its review of the second draft PA, CASAC concluded that the levels presented
in that draft document (i.e., alternative annual standard levels within a range of 13 to 11 |ig/m3
and alternative 24-hour standard levels within a range of 35 to 30 |ig/m3) "are supported by the
epidemiological and toxicological evidence, as well as by the risk and air quality information
compiled" in the ISA, RA, and second draft PA. CASAC further noted that "[although there is
increasing uncertainty at lower levels, there is no evidence of a threshold (i.e., a level below
which there is no risk for adverse health effects)" (Samet, 2010d, p. ii).
Although CASAC supported the alternative standard level ranges presented in the second
draft PA, they encouraged EPA to develop a clearer rationale for staff conclusions regarding
annual and 24-hour standards that are appropriate to consider, including consideration of the
combination of these standards supported by the available information. Specifically, CASAC
encouraged staff to focus on information related to the concentrations that were most influential
in generating the health effect estimates in individual studies to inform alternative annual
2-100
-------
standard levels (Samet, 2010d, p. 2). CAS AC also commented that the approach presented in the
second draft PA to identify alternative 24-hour standard levels which focused on peak-to-mean
ratios was not relevant for informing the actual level (Samet 2010d, p. 4). Further, they
expressed the concern that the combinations of annual/24-hour standard levels discussed in the
second draft PA (i.e., in the range of 13 to 11 |ig/m3 for the annual standard, in conjunction with
retaining the current 24-hour PM2.5 standard level of 35 |ig/m3; alternatively, revising the level of
the 24-hour standard to 30 |ig/m3 in conjunction with an annual standard level of 11 |ig/m3)
"may not be adequately inclusive" and encouraged EPA to more clearly explain its rationale for
identifying the 24-hour/annual combinations that are appropriate for consideration (Samet
2010d, p. ii).
In considering CASAC's advice, we note that staff conclusions in this final PA regarding
alternative standard levels that are appropriate to consider differ somewhat from the alternative
standard levels discussed in the second draft PA (US EPA, 2010f, section 2.3.4), upon which
CAS AC based its advice. In commenting on draft staff conclusions in the second draft PA,
CASAC did not have an opportunity to review the analyses of distributional statistics conducted
by staff to identify the broader range of PM2.5 concentrations that were most influential in
generating health effect estimates in epidemiological studies (Rajan et al., 2011), as presented in
section 2.3.4.1 above. In addition, CASAC was not aware of the revised long-term mean
concentration in the WHI study as discussed in section 2.2.1, specifically, footnote 24.
2.3.4.4 Staff Conclusions on Alternative Standard Levels
In considering the epidemiological evidence, estimates of risk reductions associated with
just meeting alternative annual and/or 24-hour standards, air quality analyses, and related
limitations and uncertainties, staff concludes that there is clear support for considering revisions
to the suite of current PM2.5 standards to provide additional protection against health effects
associated with long- and short-term exposures. We recognize that health effects may occur over
the full range of concentrations observed in the long- and short-term epidemiological studies and
that no discernible threshold for any effects can be identified based on the currently available
evidence. In reaching staff conclusions regarding appropriate alternative standard levels to
consider, we have examined where the evidence of associations is strongest and, conversely,
where we have appreciably less confidence in the associations and in quantitative estimates of
risk.
Based upon the currently available evidence, we conclude alternative annual standard
levels in the range of 13 to 11 |ig/m3 are appropriate to consider. We further conclude that the
evidence most strongly supports consideration of an alternative annual standard level in the
range of 12 to 11 |ig/m3. An alternative level within the range of 12 to 11 |ig/m3 would more
2-101
-------
fully take into consideration the available information from all long- and short-term PM2.5
exposure studies, including studies of susceptible populations, than would a higher level. This
range would also reflect placing weight on information from studies that helps to characterize the
range of PM2.5 concentrations over which we continue to have confidence in the associations
observed in epidemiological studies, as well as the extent to which our confidence in the
associations is appreciably less at lower concentrations.
In considering how the annual and 24-hour standards work together to provide
appropriate public health protection, we conclude it is appropriate to consider retaining the
current 24-hour standard level at 35 |ig/m3 in conjunction with an annual standard in the range of
12 to 11 |ig/m3, as well as to consider an alternative 24-hour standard level of 30 |ig/m3,
particularly in conjunction with an annual standard level of 13 |ig/m3.
These conclusions reflect the much stronger body of scientific evidence available in this
review supporting a causal relationship between long- and short-term PM2.5 exposures and
mortality and cardiovascular effects and a likely causal relationship between long- and short-
term PM2.5 exposures and respiratory effects, as well as evidence that is suggestive of a causal
relationship with other health outcomes such as low reproductive and development effects (e.g.,
birth weight and infant mortality) and cancer, mutagenicity, and genotoxicity effects. In
addition, we reflect upon the broad range of long- and short-term exposure studies that reported
PM2 s-related effects in areas that would likely have met the current suite of PM2 5 standards, and
specifically consider available information on the range of concentrations that were most
influential in generating the health effect estimates in epidemiological studies.
Beyond evidence-based considerations, we have also considered the extent to which the
quantitative risk assessment supports consideration of these alternative standard levels or
provides support for lower levels. We first conclude that risks estimated to remain upon
simulation of just meeting the current suite of standards are important from a public health
perspective, considering both the severity and estimated magnitude of effects. In considering
simulations of just meeting alternative annual standard levels within the range of 13 to 11 |ig/m3
in conjunction with the current 24-hour standard level of 35 |ig/m3, we conclude that important
public health improvements are associated with risk reductions estimated for standard levels of
13 and 12 |ig/m3, noting that the level of 11 |ig/m3 was not included in the quantitative risk
assessment. Our overall confidence in the quantitative risk estimates is strongest for the
alternative annual standard level of 13 |ig/m3. We have somewhat lower confidence in risk
estimates for the alternative annual standard level of 12 |ig/m3. We also estimated risks likely to
remain upon just meeting an annual standard level of 10 |ig/m3, although we have substantially
lower confidence in those estimates. With regard to level of the 24-hour standard, our overall
confidence in the quantitative risk estimates is strongest for the current standard level of 35
2-102
-------
|ig/m3. We have somewhat lower confidence in our risk estimates for an alternative 24-hour
standard level of 30 |ig/m3, and substantially lower confidence in our estimates of risks for an
alternative 24-hour standard level of 25 |ig/m3.
Based on the above considerations, we conclude that the quantitative risk assessment
provides support for considering an alternative annual standard within a range of 13 to 11 |ig/m3,
in conjunction with a 24-hour standard of 35 or 30 |ig/m3, but does not provide strong support
for considering lower alternative levels.
In evaluating this range of alternative annual standard levels, staff has taken into
consideration the importance of balancing the strength of the currently available evidence and
risk-based information with the remaining uncertainties and limitations associated with this
information. The upper end of the range of alternative annual standard levels (13 |ig/m3) reflects
placing appreciably more weight on the uncertainties and limitations in the information which
would serve to reduce the potential to overestimate public health risks and protection likely to be
associated with just meeting a standard set at this level. This policy option would reflect placing
greater weight on the remaining uncertainties in the evidence, including uncertainties associated
with understanding the heterogeneity observed in the epidemiological studies such as those
associated with the role of specific components, sources, and subtractions (e.g., UFPs) within the
current PM2.5 mass-based indicator; the role of fine particles and co-pollutants within the broader
ambient mixture; and exposure-related factors that influence the magnitude and duration of fine
particle exposures. The lower end of this range (11 |ig/m3) reflects placing much less weight on
uncertainties and limitations in the information which would serve to reduce the potential to
underestimate public health risks and protection likely to be associated with just meeting a
standard set at this level. This policy option would reflect placing considerably more weight on
limited evidence of serious effects in susceptible populations such as potential developmental
effects, while recognizing that significant limitations remain in assessing the relationship
between PM2.5 exposures and these effects, specifically, understanding the nature of the
association and exposure windows of concern. We recognize that air quality changes designed
to meet alternative annual standards are an effective and efficient way to reduce not only long-
term exposure-related mortality, but also short-term exposure-related risk. Thus, we judge it is
appropriate to consider a policy goal of focusing on establishing a generally controlling annual
standard intended to serve as the primary means for providing protection for effects associated
with both long- and short-term PM2.5 exposures in conjunction with a 24-hour standard that
provides supplemental protection.
The alternative suites of PM2.5 standards supported by the currently available evidence
and quantitative risk assessment discussed above might reasonably be judged to provide
appropriate public health protection. However, some combinations of standards are more likely
2-103
-------
to provide more consistent protection, whereas other combinations are more likely to provide
less uniform protection. Combining a need to provide requisite protection with a policy goal of
providing such protection as uniformly as possible in areas across the U.S., we conclude it is
most appropriate to consider revising the annual standard level within a range of 12 to 11 |ig/m3
in conjunction with retaining the 24-hour standard level at 35 |ig/m3. Such combinations of 24-
hour and annual standards levels (i.e., 12/35 and 11/35) would be expected to result in the annual
standard being the generally controlling standard, except in areas with particular high peak-to-
mean ratios.
Alternatively, we also reach the conclusion that there is limited support for considering
revising the annual standard level to 13 |ig/m3 in conjunction with revising the 24-hour standard
level to somewhat below 35 |ig/m3, such as, down to 30 |ig/m3 (i.e., 13/30). In considering the
implication of this alternative suite of standards, we conclude this combination would result in a
large number of areas in many geographic areas where the 24-hour standard would likely
become the controlling standard. Staff judges that this approach would likely provide less
uniform protection in areas across the U.S.
To provide some perspective on the implications of applying various combinations of
alternative annual and 24-hour standards, staff assessed (based on 2007 to 2009 air quality data)
the percentage of counties, and the population in those counties, that would not likely attain
various alternative suites of PM2 5 standards (Schmidt, 201 la, Analysis C). This assessment,
shown in Appendix C, Table C-l, was not considered as a basis for the above staff conclusions.
2.4 SUMMARY OF STAFF CONCLUSIONS ON PRIMARY FINE PARTICLE
STANDARDS
In reaching conclusions on the adequacy of the current suite of PM2.5 standards and
potential alternative suites of standards to provide the appropriate protection for health effects
associated with long- and short-term fine particle exposures, staff has considered these standards
in terms of the basic elements of the NAAQS: indicator, averaging time, form, and level
(sections 2.3.1 to 2.3.4). In considering the scientific and technical information, we reflect upon
the information available in the last review integrated with information that is newly available as
assessed and presented in the ISA and RA (US EPA, 2009a; US EPA, 2010a) and as summarized
in sections 2.2 and 2.3. We also consider the issues raised by the court in its remand of the
primary annual PM2.5 standard as discussed in section 2.1.2.
As outlined in section 2.1.3, our approach to reaching conclusions about the adequacy of
the current suite of PM2.5 standards and potential alternative standards that are appropriate for
consideration is broader and more integrative than approaches used in past reviews. Our
approach integrates a much expanded body of health effects evidence, more extensive air quality
2-104
-------
data and analyses, and a more comprehensive quantitative risk assessment, and considers the
combined protection against PM2.5-related mortality and morbidity effects associated with both
long- and short-term exposures afforded by the suite of annual and 24-hour standards.
We recognize that selecting from among alternative suites of standards will necessarily
reflect consideration of the qualitative and quantitative uncertainties inherent in the relevant
evidence and in the assumptions that underlie the quantitative risk assessment. In reaching staff
conclusions on alternative suites of standards that are appropriate to consider, we are mindful
that the CAA requires primary standards to be set that are requisite to protect public health with
an adequate margin of safety, such that the standards are to be neither more nor less stringent
than necessary. Thus, the CAA does not require that the NAAQS be set at zero-risk levels, but
rather at levels that reduce risk sufficiently so as to protect public health with an adequate margin
of safety (section 1.2.1).
Based on the currently available scientific evidence and other information, staff reaches
the following conclusions regarding the primary fine particle standards:
(1) Consideration should be given to revising the current suite of primary PM2 5 standards to
provide increased public health protection from the effects of both long- and short-term
exposures to fine particles in the ambient air. This conclusion is based, in general, on the
evaluation in the ISA of the currently available epidemiological, lexicological, dosimetric,
and exposure-related evidence, and on air quality information and analyses related to the
epidemiological evidence, together with judgments as to the public health significance of the
estimated incidence of effects remaining upon just meeting the current suite of standards.
(2) It is appropriate to retain PM2.5 as the indicator for fine particles. Staff concludes that the
available evidence does not provide a sufficient basis for replacing or supplementing the
PM2.5 indicator with any other indicator(s) defined in terms alternative size fractions (i.e.,
UFPs) or for any specific fine particle component or group of components associated with
any source categories of fine particles, nor does it provide a basis for excluding any
component or group of components associated with any source categories from the mix of
particles included in the PM2.5 indicator.
(3) With regard to averaging times for the PM2.5 standards, it is appropriate to retain annual and
24-hour averaging times to provide protection against health effects associated with long-
term (seasons to years) and short-term (hours to days) exposure periods. The available
evidence does not provide a sufficient basis for consideration of other averaging times,
including an averaging time less than 24 hours to address health effects associated with sub-
daily exposures or an averaging time to address effects associated with seasonal exposures,
given the relatively small amount of relevant information available.
(4) It is appropriate to consider revising the form of the annual standard to one based on the
highest appropriate monitor in an area rather than a form that allows averaging across
monitors (i.e., spatial averaging) to provide increased protection for susceptible populations.
Further, it is appropriate to retain the 98th percentileform of the current 24-hour standard.
2-105
-------
(5) Consideration should be given to revising the suite of PM2.5 standards to provide increased
protection against effects associated with both long- and short-term exposures, taking into
account both evidence-based and risk-based considerations, with a particular focus on
revising the annual standard level to provide protection for effects associated with both
exposure periods. An emphasis on the annual standard would be consistent with the policy
approach of setting a "generally controlling" annual standard to provide protection for both
long- and short-term PM2.5 exposures in conjunction with a 24-hour standard that provides
supplemental protection against days with high peak concentrations. This would limit peak
concentrations in areas with high peak-to-mean ratios, possibly associated with strong local
or seasonal sources. This would also provide supplemental protection for potential PM2 5-
related effects that may be associated with shorter-than-daily exposure periods. Staff
concludes that this policy goal is the most effective and efficient way to reduce total
population risk associated with both long- and short-term exposures, and would provide
relatively more uniform protection in areas across the country.
(a) Taken together, staff concludes that the currently available evidence and
information from a quantitative risk assessment and air quality analyses provide
support for considering revision of the level of the annual standard to within a
range of 13 to 11 |ig/m3. Staff further concludes that the evidence most strongly
supports consideration of an alternative annual standard level in the range of 12 to
11 |ig/m3.
(b) In conjunction with consideration of an annual standard level in the range of 12 to
11 |ig/m3, staff concludes it is appropriate to consider retaining the current 24-
hour standard level at 35 |ig/m3.
(c) In conjunction with consideration of an annual standard level of 13 |ig/m3, staff
concludes that there is limited support to consider revising the 24-hour standard
level to somewhat below 35 |ig/m3, such as down to 30 |ig/m3.
2.5 KEY UNCERTAINTIES AND AREAS FOR FUTURE RESEARCH AND DATA
COLLECTION
The uncertainties and limitations that remain in the review of the primary fine particle
standards are primarily related to understanding the range of ambient concentrations over which
we continue to have confidence in the health effects observed in the epidemiological studies, as
well as the extent to which the heterogeneity observed in the epidemiological evidence is related
to differences in the ambient fine particle mixture and/or exposure-related factors. In addition,
uncertainties remain in more fully understanding the role of PM2 5 in relationship to the roles of
gaseous co-pollutants within complex ambient mixtures.
In this section, we highlight areas for future health-related research, model development,
and data collection activities to address these uncertainties and limitations in the current body of
scientific evidence. These efforts, if undertaken, could provide important evidence for informing
future PMNAAQS reviews and, in particular, consideration of possible alternative indicators,
averaging times, and/or levels. In some cases, research in these areas can go beyond aiding
2-106
-------
standard setting to informing the development of more efficient and effective control strategies.
We note, however, that a full set of research recommendations to meet standards implementation
and strategy development needs is beyond the scope of this discussion.
As has been presented and discussed in the PM ISA, particularly in Chapters 4 through 8,
the scientific body of evidence informing our understanding of health effects associated with
long- and short-term exposures to fine particles has been broadened and strengthened since the
last review. In reviewing the adequacy of the current suite of primary PM2.5 standards and in
evaluating alternative health-based fine particle standards appropriate for consideration, we
identify the following key uncertainties and areas for future research and data collection efforts
that have been highlighted in this review. We recognize that some research could be available to
inform the next PM NAAQS review, while other research may require longer-term efforts.
Interpretation of Epidemiological Evidence
Additional research focused on identifying the most important factors contributing to the
observed heterogeneity in the epidemiological evidence could provide insights for interpreting
these studies. We encourage research and data collection efforts directed at improving our
understanding of the nature of the exposures contributing to the observed health effects, for
example, the role of specific components, sources, and different size fractions (e.g., UFPs)
within the current PM2 5 mass-based indicator and the role of fine particles and co-pollutants
within the broader ambient mixture, as well as improving our understanding of exposure-related
factors that influence the magnitude and duration of fine particle exposures. Much of this
research may depend on the availability of increased monitoring data, as discussed below.
• Components/Sources. The currently available scientific evidence continues to be largely
indexed by aggregate PM2 5 mass-based concentrations which vary in composition both
regionally and seasonally. Source characterization, exposure, epidemiological, and
toxicological research could focus on improving our understanding of the relative toxicity of
different fine particle components, properties, and sources that may be more closely linked
with various health effects. Critical to this better understanding of the impacts of PM2 5
components and their associated sources are data that refines the temporal and spatial
variability of the fine particle mixture. This research would reduce the uncertainties in
estimating risks. It could also inform consideration of alternative indicators in future PM
NAAQS reviews as well as aid in the development of efficient and effective source control
strategies for reducing health risks.
• Ultrafine Particles (UFPs). Additional monitoring methods development work, health
research, and ambient monitoring data collection efforts are needed to expand the currently
available scientific data base for UFPs. UFP measurements should include surface area as
well as number, mass and composition. It would be most useful for an UFPs monitoring
network to be designed to inform our understanding of the spatial and temporal variability of
these particles, including in near-roadway environments. This information would improve
2-107
-------
our ability to explore consideration of a separate indicator for UFPs in future PM NAAQS
reviews.
• Co-pollutant Exposures. Research focused on furthering our understanding of the extent to
which an association between fine particles and specific health effects can be modified by
one or more co-pollutants would inform our ability to discern the role of PM in the complex
ambient mixture. For example, does the magnitude of a PM2.5-related effect estimate differ
on days when O?, concentrations are higher compared to days when O?, concentrations are
lower?
• Factors Influencing Exposures. Additional research and analyses would be useful to provide
insights on population exposures, specifically in improving our understanding of intra-city
and inter-city differences related to various PM2.5 components, source contributions and
personal and building-related factors that may enhance our interpretation of the
epidemiological evidence. This could include time-activity data to support probabilistic
scenario-based exposure models, such as additional activity diary data to incorporate into the
Consolidated Human Activity Database (CHAD); air conditioning use; residence near
roadways; and penetration rates to better characterize ambient PM2 5 impacts on indoor
microenvironments. This research could focus on different size fractions in PM2 5 (i.e.,
UFPs) as well as components. Coordination between exposure and health studies could
advance our understanding of exposure-related factors. For example, epidemiological panel
studies might use various exposure measurements to explore differences in personal
exposures related to (1) indoor generated fine particles, (2) fine particle exposures measured
by community monitors, and (3) fine particle exposures not captured by community monitors
(i.e., personal exposures during commuting).
Health Outcomes, Exposure Durations of Concern, and Susceptible Populations
New information available in this review reinforces and expands the evidence of
associations between long- and short-term PM2.5 exposures and mortality and a number of
cardiovascular and respiratory effects. Less evidence is available to understand other health
effects (e.g., developmental/reproductive effects; central nervous system effects). Additional
research could expand our understanding of the associations between PM2.5 and a broader range
of health outcomes; reduce uncertainties associated with our current understanding of
concentration-response relationships; improve our understanding of exposure durations of
concern; and improve our understanding of the potential public health impacts of fine particle
exposures in susceptible populations. Toxicological studies could provide additional evidence of
coherence and biological plausibility for the effects observed in epidemiological studies as well
as additional insights on possible mechanisms of action.
• Health Effects. Research on a broader range of cardiovascular and respiratory endpoints
could improve our understanding of the mechanisms by which these effects occur. In
addition, future research could expand the scientific data base for health effects that are
currently less understood including effects categorized within the ISA as having evidence
suggestive of a causal relationship or for which currently available evidence is inadequate to
support a quantitative risk analysis. To the extent that research supports a link between fine
2-108
-------
particles and adverse effects on the nervous system, reproduction, development, or other
endpoints, such effects could play an increased role for informing future PM NAAQS
reviews including expanding the health endpoints that could potentially be evaluated in
future quantitative risk assessments.
• Concentration-Response Relationships. Research focused on improving our understanding
of the shape of the C-R relationships, especially at lower ambient fine particle
concentrations, as well as the confidence intervals around these C-R relationships, could
reduce uncertainties associated with estimating and characterizing risks throughout the full
range of air quality distributions. As more information becomes available on fine particle
components and sources, it will be important to understand the C-R relationships for key
constituents of the fine particle mixture, as well.
• Exposure Durations of Concern. Research should be directed at broadening the scientific
data base to improve our understanding of health effects associated with short-term, peak
exposures, such as those related to traffic-related sources, wildfires, agricultural burning, or
other episodic events, as well as to improve our understanding of health effects associated
with seasonal-length exposures, such as those related to wintertime wood-burning emissions.
Additional quantitative measures of exposure might take into account factors including the
magnitude and duration of sub-daily and seasonal length PM2.5 exposures and the frequency
of health impacts associated with repeated peak exposures. More research is needed to better
understand effects that occur at longer lag times than have historically been studied (e.g., 0 to
2 day lags).
• Susceptible Populations. Improving our understanding of the populations that are more
likely to experience adverse health effects related to fine particle exposures and the
concentrations at which these effects may occur is important for informing future PM
NAAQS reviews and for developing programs to reduce related public health risks. This
evidence may also provide insights into the biologic modes of action for toxicity.
o Pre-existing Health Conditions. While currently identified susceptible populations
include persons with pre-existing cardiovascular and respiratory disease, evidence
continues to emerge related to additional health conditions that may increase
susceptibility to fine particle exposures (e.g., diabetes, obesity, neurological
disorders). Research to replicate or extend these findings would enhance our
understanding of these and other potentially susceptible populations.
o Children. Epidemiological and toxicological studies provide evidence that children
are more susceptible to PM exposures, primarily for respiratory-related effects.
Evidence of developmental effects associated with PM exposures continues to
emerge. Additional research exploring issues to better understand key windows of
development impacted by PM exposures could enhance our understanding of this
important susceptible lifestage.
o Genetic Susceptibility. Research to expand our understanding of genetic
susceptibility could inform our understanding of potentially susceptible populations
and provide additional information for identifying the specific pathways and
mechanisms of action by which PM initiates health effects.
2-109
-------
o Socioeconomic status (SES). Additional research is needed to identity what factors
(e.g., general health status, diet, medication, stress, unmeasured pollution) cause SES
differences in response to pollution measured in communities.
Data Collection Needs and Methods Development Activities
Additional research and data collection efforts focused on expanding current monitoring
methods and networks as well as continued development of exposure models to expand data
available for health studies could improve our understanding of potential alternative indicators,
averaging times, and levels to consider in future PM NAAQS reviews. In particular, staff
encourages work to enhance our understanding of the temporal and spatial variability of PM2.5,
PM2.5 components, and different size fractions (e.g., UFPs).
• Monitoring Measurements. In order to improve our understanding of the association
between fine particles and health effects, more frequent measurement data could be collected.
This would provide information that could inform our understanding of alternative lags.
o PM?^ Components. With respect to improving our understanding of the impacts of
PM2.5 components, enhancements to the CSN, including more frequent measurement
schedules and the development and deployment of continuous monitoring methods
for specific fine particle components (e.g., EC/OC, sulfates), could enhance our
understanding of the temporal and spatial variability of specific components.
Furthermore, identifying chemical species within the mix of organic aerosols would
improve our understanding of the artifacts associated with semi-volatile PM
components and aid in designing toxicological experiments.
o Ultrafme Particles. In order to improve our understanding of the public health
impacts of UFPs, consideration should be given to establishing an FRM for UFPs and
establishing a national UFP monitoring network.
o Source Apportionment. Composition data with better time resolution (e.g., 1 to 6
hour) and better size resolution (e.g., UFPs, accumulation mode particles, coarse
particles in PM2.5 and PMio-2.s) could provide more precise and accurate information
on sources of fine particles to inform health research as well as development of more
efficient and effective control strategies.
o Spatial Variability. Some portion of the required PM25 monitoring network could be
dedicated to improving our ability to characterize spatial variability across urban
areas including both at localized and area-wide scales.
• Model Development. Continuing work to improve models for estimating PM2 5 mass and
composition in areas with only every third or sixth day measurements, and by space where
measurements are not available could enhance our understanding of the temporal and spatial
variability of fine particles. Refinement of these models to finer spatial scales may improve
exposure estimates in epidemiological studies as well as in quantitative risk and exposure
assessments.
• Air Quality Distributions Reported in Epidemiological Studies. Most epidemiological
studies provide some information on the distribution of ambient measurement data evaluated,
however, published information is often generally limited in scope and the descriptive
2-110
-------
statistics reported vary from one study to another. Understanding the air quality distributions
at which effects have been observed is important for informing consideration of the adequacy
of the current NAAQS as well as potential alternative indicators, averaging times, and levels
to consider. Working with intramural and extramural research groups, we plan to encourage
a more comprehensive and more consistent reporting of population-level and air quality data.
2-111
-------
2.6 REFERENCES
Bell ML, Ebisu K, Belanger K (2007). Ambient air pollution and low birth weight in Connecticut and
Massachusetts. Environ Health Perspect, 115: 1118-24.
Bell ML, Ebisu K, Peng RD, Walker J, Samet JM, Zeger SL, 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:
1301-1310.
Bell ML, Ebisu K, Peng R, Samet J, Dominici F (2009). Hospital Admissions and Chemical Composition of Fine
Particle Air Pollution. Am J Respir Crit Care Med, 179: 1115-1120.
Bell ML (2009). Personal communication with Dr. Michelle Bell: Annual PM2 5 levels used in Dominici et al. 2006
and Bell et al. 2008. December 7, 2009. Docket No. EPA-HQ-ORD-2007-0517-0087.
Beelen R, Hoek G, van den Brandt PA, Goldbohm RA, Fischer P, Schouten LJ, Jerrett M, Hughes E, Armstrong B,
Brunekreef B (2008). Long-term effects of traffic-related air pollution on mortality in a Dutch cohort
(NLCS-AIR study). Environ Health Perspect, 116: 196-202.
Burnett RT, Goldberg MS (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian cities. In:
Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Burnett RT, Stieb D, Brook JR, Cakmak S, Dales R, Raizenne M, Vincent R, Dann T (2004). Associations between
short-term changes in nitrogen dioxide and mortality in Canadian cities. Arch Environ Occup Health, 59:
228-236.
Curl C (2009). Personal communication with Cynthia Curl, MESA Air Project Manager, University of Washington;
email to Beth Hassett-Sipple, US EPA, OAQPS regarding request for PM air quality data. August 10,
2009. Docket No. EPA-HQ-ORD-2007-0517-0113.
Delfino RJ, Murphy-Moulton AM, Burnett RT, Brook JR, Becklake MR (1997). Effects of air pollution on
emergency room visits for respiratory illnesses in Montreal, Quebec. Am J Respir Crit Care Med, 155: 568-
576.
Dockery DW, Pope CA III, Xu X, Spengler JD, Ware JH, Fay ME, Ferris BG Jr, Speizer FE (1993). An association
between air pollution and mortality in six US cities. NEJM 329: 1753-1759.
Dockery DW, Cunningham J, Damokosh Al, Neas LM, Spengler JD, Koutrakis P, Ware JH, Raizenne M, Speizer
FE (1996). Health effects of acid aerosols on North American children: respiratory symptoms. Environ
Health Perspect. 104(5):500-5.
Dominici F, Peng RD, Bell ML, Pham L, McDermott A, Zeger SL, Samet JM (2006a). Fine paniculate air pollution
and hospital admission for cardiovascular and respiratory diseases. JAMA, 295: 1127-1134.
Dominic F (2006b). Letter from Francesca Dominici, Associate Professor of Biostatistics, Bloomberg School of
Public Health, Johns Hopkins University, comments to the proposed rule. Docket No. OAR-2001-0017-
0988. March 21, 2006.
Dominici F, Peng RD, Zeger SL, White RH, Samet JM (2007a). Paniculate air pollution and mortality in the United
States: did the risks change from 1987 to 2000? Am J Epidemiol, 166: 880-8.
Dominici F, Peng RD, Ebisu K, Zeger SL, Samet JM, Bell ML (2007b). Does the effect of PM10 on mortality
depend on PM nickel and vanadium content? A reanalysis of the NMMAPS data. Environ Health Perspect,
115: 1701-1703.
2-112
-------
Eftim SE, Samet JM, Janes H, McDermott A, Dominici F (2008). Fine Particulate Matter and Mortality: A
Comparison of the Six Cities and American Cancer Society Cohorts With a Medicare Cohort.
Epidemiology, 19: 209- 216.
Fairley D (2003). Mortality and air pollution for Santa Clara County, California, 1989-1996, In: Revised analyses of
time-series studies of air pollution and health. Special report. Health Effects Institute. Boston, MA.
http://www.healtheffects.org/Pubs/TimeSeries.pdf.
Franklin M, Zeka A, Schwartz J (2007). Association between PM2 5 and all-cause and specific-cause mortality in 27
US communities. J Expo Sci Enviorn Epidemiol, 17:279-287.
Franklin M, Koutrakis P, Schwartz J (2008). The role of particle composition on the association between PM2 5 and
mortality. Epidemiology, 19:680-689.
Gauderman WJ., McConnell R, Gilliland F, London S, Thomas D, Avol E, Vora H., Berhane K., Rappaport EB.,
Lurmann F, Margolis HG, Peters J. (2000). Association between air pollution and lung function growth in
southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
Gauderman WJ, Gilliland GF, Vora H, Avol E, Stram D, McConnell R, Thomas D, Lurmann F, Margolis HG,
Rappaport, E. B., Berhane, K., Peters, J. M. (2002). Association between air pollution and lung function
growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care Med. 166:
76-84.
Gauderman WJ, Avo, E, Gilliland F, Vora H, Thomas D, Berhane K, McConnell R, Kuenzli N, Lurmann F,
Rappaport E, Margolis H, Bates D, Peters J (2004). The effect of air pollution on lung development from
10 to 18 years of age. NEJM 351:1057-67.
Gent JF, Koutrakis P, Belanger K, Triche E, Holford TR, Bracken MB, Leaderer BP (2009) Symptoms and
medication use in children with asthma and traffic-related sources of fine particle pollution. Environ
Health Perspect. 117:1168-74.
Gong H Jr, Linn WS, Clark KW, Anderson KR, Geller MD, Sioutas C (2005) Respiratory responses to exposures
with fine particulates and nitrogen dioxide in the elderly with and without COPD. Inhal Toxicol.
17(3): 123-32.
Goss CH, Newsom SA, Schildcrout JS, Sheppard L, Kaufman JD (2004). Effect of ambient air pollution on
pulmonary exacerbations and lung function in cystic fibrosis. Am J Respir Crit Care Med, 169: 816-821.
Hassett-Sipple B and Stanek L (2009). Email to study authors of recent U.S, and Canadian epidemiological studies
evaluating health effects associated with exposure to fine and thoracic coarse particles. May 2, 2009 and
October 20, 2009. Docket ID number EPA-HQ-ORD-2007-0517-0050 and EPA-HQ-ORD-2007-0517-
0104.
Hassett-Sipple, B, Rajan, P, Schmidt, M. (2010). Analyses of PM2 5 Data for the PM NAAQS Review.
Memorandum to PM NAAQS review docket. Docket ID number EPA-HQ-OAR-2007-0492-0077. March
29,2010. Available: http://epa.gov/ttn/naaqs/standards/pm/s_pm 2007 td.html.
Henderson R (2006a). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Recommendations Concerning the Proposed National Ambient Air Quality Standards for Particulate
Matter. March 21, 2006. Available:
http://vosemite.epa.gov/sab/sabproduct.nsf/CD706C976D AC62B3852571390081CC21/$File/casac-ltr-06-
002.pdf.
2-113
-------
Hopke PK, Ito K, Mar T, Christensen WF, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Larson TV, Liu H, Neas
L, Pinto J, Stolzel M, Suh H, Paatero P, Thurston GD (2006). PM source apportionment and health effects:
1 Intercomparison of source apportionment results. J Expo Sci Environ Epidemiol, 16: 275-286.
Islam T, Gauderman WJ, Berhane K, McConnell R, Avol E, Peters JM, Gilliland FD (2007). The relationship
between air pollution, lung function and asthma in adolescents. Thorax, 62: 957-963.
Ito K, Christensen WF, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Larson TV, Neas L, Hopke PK, Thurston
GD (2006). PM source apportionment and health effects: 2 An investigation of intermethod variability in
associations between source-apportioned fine particle mass and daily mortality in Washington, DC. J Expo
Sci Environ Epidemiol, 16: 300-310.
Jerrett M, Burnett RT, Ma R: Pope CA, Kreski D, Newbold KB, Thurston G, Shi Y, Finkelstein N, Calle N, Thun
MJ (2005) Spatial analysis of air pollution and mortality in Los Angeles. Epidemiology, 16: 727-36.
Krewski D, Burnett RT, Goldberg MS, Hoover K, Siemiatycki J, Jerrett M, Abrahamowicz M, White WH (2000).
Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of paniculate air
pollution and mortality. A special report of the Institute's particle epidemiology reanalysis project.
Cambridge, MA: Health Effects Institute. http://pubs.healtheffects.org/view.php?id=6
Krewski D, Jerrett M, Burnett RT, Ma R, Hughes E, Shi Y, Turner MC, Pope AC III, Thurston G, Calle EE, Thun
MJ (2009). Extended Follow-Up and Spatial Analysis of the American Cancer Society Study Linking
Paniculate Air Pollution and Mortality. HEI Research Report 140, Health Effects Institute, Boston, MA.
Available: http://pubs.healtheffects.org/view.php?id=315.
Laden F, Neas LM, Dockery DW, Schwartz J (2000). Association of fine paniculate matter from different sources
with daily mortality in six US cities. Environ Health Perspect, 108: 941-947.
Laden F, Schwartz J, Speizer FE, Dockery DW (2006). Reduction in fine paniculate air pollution and mortality:
extended follow-up of the Harvard Six Cities Study. Am. J. Respir. Crit. Care. Med. 173: 667-672.
Laden F (2009). Personal communication with Dr. Francine Laden: Annual PM2 5 levels used in the update of the
Harvard Six Cities Study. May 21, 2009. Docket No. EPA-HQ-OAR-2007-0492.
Lanki T, Pekkanen J, Aalto P, Elosua R, Berglind N, D'Ippoliti D, Kulmala M, Nyberg F, Peters A, Picciotto S,
Salomaa V, Sunyer J, Tiittanen P, Von Klot S, Forastiere (2006). Associations of traffic-related air
pollutants with hospitalization for first acute myocardial infarction: the HEAPSS study. Occup Environ
Med, 63:844-851.
Lipfert FW, Baty JD, Miller JP, Wyzga RE (2006). PM2 5 constituents and related air quality variables as predictors
of survival in a cohort of U.S. military veterans. Inhal Toxicol, 18: 645-657.
Lippmann M, Ito K, Hwang JS, Maciejczyk P, Chen LC (2006). Cardiovascular effects of nickel in ambient air.
Environ Health Perspect, 114: 1662-1669.
Lisabeth LD, Escobar JD, Dvonch JT, Sanchez BN, Majersik JJ, Brown DL, Smith MA, Morgenstern LB (2008).
Ambient air pollution and risk for ischemic stroke and transient ischemic attack. Ann Neurol, 64: 53-59.
Mar TF, Norris GA, Larson TV, Wilson WE, Koenig JQ (2003) Air pollution and cardiovascular mortality in
Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health. Special report.
Boston, MA: Health Effects Institute, pp. 177-182. Available: http://www.healtheffects.org/news.htm [16
May, 2003].
2-114
-------
Mar TF, Larson TV, Stier RA, Claiborn C, Koenig JQ (2004). An analysis of the association between respiratory
symptoms in subjects with asthma and daily air pollution in Spokane, Washington. Inhal Toxicol, 16: 809-
815.
Mar TF, Ito K, Koenig JQ, Larson TV, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Neas L, Stolzel M, Paatero
P, Hopke PK, Thurston GD (2006). PM source apportionment and health effects. 3. Investigation of inter-
method variations in associations between estimated source contributions of PM2 5 and daily mortality in
Phoenix, AZ. J Expo Sci Environ Epidemiol. 16:311-20.
McConnell R, Berhane K, Gilliland F, Molitor J, Thomas D, Lurmann F, Avol E, Gauderman WJ, Peters JM (2003).
Prospective study of air pollution and bronchitic symptoms in children with asthma. Am J Respir Crit Care
Med, 168: 790-797.
Miller KA, Siscovick DS, Sheppard L, Shepherd K, Sullivan JH, Anderson GL, Kaufman JD (2007). Long-term
exposure to air pollution and incidence of cardiovascular events in women. N Engl J Med, 356: 447-458.
Ostro BD, Broadwin R, Lipsett MJ (2003) Coarse particles and daily mortality in Coachella Valley, California. In:
Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Ostro B, Feng W, Broadwin R, Green S, Lipsett M (2007). The Effects of Components of Fine Paniculate Air
Pollution on Mortality in California: Results from CALFINE. Environ Health Perspect 115(1): 13-19.
Peng RD, Bell ML, Geyh AS, McDermott A, Zeger SL, Samet JM, Dominici F (2009). Emergency admissions for
cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environ
Health Perspect. 117:957-63.
Penttinen P, Vallius M, Tiittanen P, Ruuskanen J, Pekkanen J (2006). Source-specific fine particles in urban air and
respiratory function among adult asthmatics. Inhal Toxicol, 18: 191-198.
Peters A., Dockery DW, Muller JE, Mittleman MA (2001). Increased particulate air pollution and the triggering of
myocardial infarction. Circulation 103:2810-2815.
Peters J, Avol E, Gauderman WJ, Linn WS, Navidi W, London S, Margolis H, Rappaport E, Vora H, Gong H Jr,
Thomas DC (1999). A study of twelve southern California communities with differing levels and types of
air pollution II Effects on pulmonary function. Am J Respir Crit Care Med, 159: 768-775.
Pope CA 3rd, Dockery DW (1992). Acute health effects of PM10 pollution on symptomatic and asymptomatic
children. Am Rev Respir Dis. 145(5): 1123-8.Pope CA 3rd, Thun M.J., Namboodiri MM, Dockery DW,
Evans JS, Speizer FE, Heath CW (1995). Particulate air pollution as a predictor of mortality in a
prospective study of U.S. adults. Am J Respir Crit Care Med. 151:669-674.
Pope CA 3rd, Burnett RT, Thun MJ, Calle EE, Krewski D, Ito K, Thurston GD (2002). Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. J. Am. Med. Assoc.
287:1132-1141.
Pope CA 3rd, Burnett RT, Thurston GD, Thun MJ, Calle EE, Krewski D, Godleski JJ (2004). Cardiovascular
mortality and long-term exposure to particulate air pollution: epidemiological evidence of general
pathophysiological pathways of disease. Circulation, 109: 71-77.
Pope CA 3rd, Ezzati M, Dockery DW (2009). Fine-particulate air pollution and life expectancy in the United States.
N Engl J Med, 360:376-386.
2-115
-------
Rabinovitch N, Zhang LN, Murphy JR, Vedal S, Dutton SJ, Gelfand EW (2004). Effects of wintertime ambient air
pollutants on asthma exacerbations in urban minority children with moderate to severe disease. J Allergy
Clinlmmunol, 114: 1131-1137.
Rabinovitch N, Strand M, Gelfand EW (2006). Paniculate levels are associated with early asthma worsening in
children with persistent disease. Am J Respir Crit Care Med, 173: 1098-1105.
Raizenne M, Neas LM, Damokosh AI, Dockery DW, Spengler JD, Koutrakis P, Ware JH, Speizer FE (1996). Health
effects of acid aerosols on North American children: pulmonary function. Environ Health Perspect, 104:
506-514.
Rajan P, Schmidt M, Hassett-Sipple B (2010). PM2 5 Distributional Statistical Analyses. Memorandum to PM
NAAQS review docket EPA-HQ-OAR-2007-0492. April 7, 2011. Available:
http://www.epa. gov/ttn/naaqs/standards/pm/s_pm_2007_td.html.
Samet J (2010c). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CAS AC Review of Policy Assessmentfor the Review
of the PM NAAQS-First External Review Draft (March 2010). May 17,2010. Available:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/E504EE3276D87A9E8
525772700647 AFB/$File/EPA-CASAC-10-011-unsigned.pdf/.
Samet J (2010d). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CAS AC Review of Policy Assessment for the Review
of the PM NAAQS - Second External Review Draft (June 2010). September 10, 2010. Available:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/CCF9F4C0500C500F8
525779D0073C593/$File/EP A-CASAC-10-015-unsigned.pdf
Schmidt M, Frank N, Mintz D, Rao T, McCluney L (2005). Analyses of paniculate matter (PM) data for the PM
NAAQS review. Memorandum to PM NAAQS Review Docket EPA-HQ-OAR-2001-0017. June 30,
2005. Available: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr td.html.
Schmidt M, Hassett-Sipple B, Rajan P (2010). PM25 Air Quality Analyses. Memorandum to PM NAAQS Review
Docket EPA-HQ-OAR-2007-0492. July 22, 2010. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 td.html
Schmidt M (201 la). PM2 5 Air Quality Analyses - Update. Memorandum to the PM NAAQS Review Docket EPA-
HQ-OAR-2007-0492. April 15, 2011. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007 td.html
Schwartz J Dockery DW, Neas LM. (1996). Is daily mortality associated specifically with fine particles? J Air
Waste Manage Assoc 46: 927-939.
Schwartz J, Coull B, Laden F, Ryan L (2008). The effect of dose and timing of dose on the association between
airborne particles and survival. Environ Health Perspect, 116: 64-69.
Slaughter JC, Kim E, Sheppard L, Sullivan JH, Larson TV, Claiborn C (2005). Association between paniculate
matter and emergency room visits, hospital admissions and mortality in Spokane, Washington. J Expo Sci
Environ Epidemiol, 15: 153-159.
Stanek LW; Yang R, Hassett-Sipple B. (2010). Paniculate Matter Air Quality Data Requested from Epidemiologic
Study Authors. Memorandum to the PM NAAQS Review Dockets (EPA-HQ-ORD-2007-0517 and EPA-
HQ-OAR-2007-0492). July 22, 2010. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm 2007 td.html.
2-116
-------
Thurston G, Ito K, Mar T, Christensen WF, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Larson TV, Liu H,
Neas L, Pinto J, Stolzel M, Suh H, Hopke PK (2005). Results and implications of the workshop on the
source apportionment of PM health effects. Epidemiology, 16: S134-S135.
Tolbert PE, Klein M, Peel JL, Sarnat SE, Sarnat JA (2007). Multipollutant modeling issues in a study of ambient air
quality and emergency department visits in Atlanta. J Expo Sci Environ Epidemiol, 24: 938-945.
Tsai FC, Apte MG, Daisey JM (2000). An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Inhal Toxicol, 12: 121-135.
US EPA (2004). Air Quality Criteria for Paniculate Matter. National Center for Environmental Assessment, Office
of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711,
report no. EPA/600/P-99/002aF and EPA/600/P-99/002bF. October 2004. Available:
http: //www. epa. gov/ttn/naaqs/standards/pm/s_pm_cr_cd. html.
US EPA (2005). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards, report no. EPA EPA-452/R-05-005a. December 2005. Available:
http://www.epa.gov/ttn/naaqs/standards/prn/sjm crsp.html.
US EPA (2008a). Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate Matter.
National Center for Environmental Assessment and Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-08-004. March
2008. Available: http://www.epa.gov/ttn/naaqs/standards/prn/sjm 2007jd.html.
US EPA (2009a). Integrated Science Assessment for Paniculate Matter (Final Report). U.S. Environmental
Protection Agency, Washington, DC, EPA/600/R-08/139F, December 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_isa.html.
US EPA (2009b). Integrated Science Assessment for Paniculate Matter: Second External Review Draft. National
Center for Environmental Assessment-RTF Division, Office of Research and Development, Research
Triangle Park, NC. EPA/600/R-08/139B. July 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007 isa.html.
US EPA (2010a). Quantitative Risk Assessment for Paniculate Matter -Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
010-005. June 2010. Available: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (20 lOc). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards - First External Review Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA452/P-10-003. March 2010. Available:
http://www.epa. gov/ttn/naaqs/standards/pm/s_pm_2007_pa. html.
US EPA (2010d). Quantitative Risk Assessment for Particulate Matter - Second External Review Draft. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA-452/P-10-001. February 2010. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (20101). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards- Second External Review Draft. Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. EPA 452/P-10-007. June 2010.
Available: http://www.epa.gov/ttn/naaqs/standards/prn/sjm 2007ja.html.
Villeneuve PJ, Chen L, Stieb D, Rowe BH (2006). Associations between outdoor air pollution and emergency
department visits for stroke in Edmonton, Canada. Eur J Epidemiol, 21: 689-700.
2-117
-------
Woodruff TJ, Darrow LA, Parker JD (2008). Air pollution and postneonatal infant mortality in the United States,
1999-2002. Environ Health Perspect, 116: 110-115.
Zanobetti, A. (2009). Personal communication with Dr. Antonella Zanobetti; email to Jason Sacks, US EPA, NCEA.
June 1, 2009. Docket No. EPA-HQ-ORD-2007-0517-0064.
Zanobetti A, Schwartz J (2007). Paniculate air pollution, progression, and survival after myocardial infarction.
Environ Health Perspect, 115: 769-775.
Zanobetti A, Schwartz J (2009). The effect of fine and coarse paniculate air pollution on mortality: A national
analysis. Environ Health Perspect, 117: 1-40.
Zhang Z, Whitsel E, Quibrera P, Smith R, Liao D, Anderson G, Prineas R (2009). Ambient fine paniculate matter
exposure and myocardial ischemia in the Environmental Epidemiology of Arrhythmogenesis in the
Women's Health Initiative (EEAWHI) study. Environ Health Perspect, 117: 751-756.
Zeger S, McDermott A, Dominicit F, Samet J (2007). Mortality in the Medicare population and chronic exposure to
fine paniculate air pollution. Johns Hopkins University. Baltimore.
http://www.bepress.com/ihubiostat/paperl33.
Zeger S, Dominici F, McDermott A, Samet J (2008). Mortality in the Medicare population and chronic exposure to
fine paniculate air pollution in urban centers (2000-2005). Environ Health Perspect, 116: 1614.
2-118
-------
REVIEW OF THE PRIMARY STANDARD FOR THORACIC COARSE
PARTICLES
This chapter presents staff conclusions with regard to the adequacy of the current primary
io standard, which is intended to protect public health against exposures to thoracic coarse
particles (PMio-2.s), and potential alternative primary standards for consideration in this review.
Our assessment of these issues is framed by a series of key policy-relevant questions, which
expand upon those presented in the IRP (US EPA, 2008a). The answers to these questions will
inform decisions on whether, and if so how, to revise the current PMio standard.
Our approach for reviewing the primary PMio standard is presented in section 3.1. Our
considerations and conclusions regarding the adequacy of the current PMio standard are
presented in section 3.2. Section 3.3 presents our considerations and conclusions with respect to
potential alternative standards, focusing on each of the basic elements of the standards: pollutant
indicator (section 3.3.1), averaging time (section 3.3.2), form (section 3.3.3), and level (section
3.3.4). Section 3.4 summarizes staff conclusions on the current and potential alternative
standards. Section 3.5 discusses key uncertainties and suggested future research areas and data
collection efforts.
3.1 APPROACH
Staffs approach for reviewing the current primary PMio standard builds upon the
approaches used in previous PM NAAQS reviews. The past and current approaches described
below are all based most fundamentally on using information from epidemiological studies to
inform the selection of PM standards that, in the Administrator's judgment, protect public health
with an adequate margin of safety. Evidence-based approaches to using information from
epidemiological studies to inform decisions on PM standards are complicated by the recognition
that no population threshold, below which it can be concluded with confidence that PM-related
effects do not occur, can be discerned from the available evidence (US EPA, 2009a, section
2.4.3). As a result, any approach to reaching decisions on what standards are appropriate
requires judgments about how to translate the information available from the epidemiological
studies into a basis for appropriate standards, which includes consideration of how to weigh the
uncertainties in reported associations across the distributions of PM concentrations in the studies.
Our approach to informing these decisions, discussed more fully below, recognizes that the
available health effects evidence reflects a continuum consisting of ambient levels at which
scientists generally agree that health effects are likely to occur through lower levels at which the
likelihood and magnitude of the response become increasingly uncertain. Such an approach is
3-1
-------
consistent with setting standards that are neither more nor less stringent than necessary,
recognizing that a zero-risk standard is not required by the CAA.
3.1.1 Approaches Used in Previous Reviews
3.1.1.1 Reviews Completed in 1987 and 1997
The PM NAAQS have always included some type of a primary standard to protect
against effects associated with exposures to thoracic coarse particles. In 1987, when EPA first
revised the PM NAAQS, EPA changed the indicator for PM from Total Suspended Particles
(essentially applicable to particles smaller than 25-45 micrometers) to focus on inhalable
particles, those which can penetrate into the trachea, bronchi, and deep lungs (52 FR 24634).
The EPA changed the PM indicator to PMio based on evidence that the risk of adverse health
effects associated with particles with a nominal mean aerodynamic diameter less than or equal to
10 jim was significantly greater than risks associated with larger particles (52 FR at 24639).
In the 1997 review, in conjunction with establishing new fine particle (i.e., PM2 5)
standards (discussed above in sections 1.2.2, 2.1.1.1), EPA concluded that continued protection
was warranted against potential effects associated with thoracic coarse particles in the size range
of 2.5 to 10 |im. This conclusion was based on particle dosimetry, toxicological information,
and on limited epidemiological evidence from studies that measured PMio in areas where coarse
particles were likely to dominate the distribution (62 FR 38677, July 18, 1997). Thus, EPA
concluded that the existing PMio standards would provide requisite protection against effects
associated with particles in the size range of 2.5 to 10 jim. Although EPA considered a more
narrowly defined indicator for thoracic coarse particles in that review (i.e., PMio-2.s), EPA
concluded that it was more appropriate, based on existing evidence, to continue to use PMio as
the indicator. This decision was based, in part, on the recognition that the only studies of clear
quantitative relevance to health effects most likely associated with thoracic coarse particles used
PMio in areas where the coarse fraction was the dominant fraction of PMio, namely two studies
conducted in areas that substantially exceeded the 24-hour PMio standard (62 FR 38679). In
addition, there were only very limited ambient air quality data then available specifically for
PMio-2.5, in contrast to the extensive monitoring network already in place for PMio, Therefore, it
was judged more administratively feasible to use PMio as an indicator. The EPA also stated that
the PMio standards would work in conjunction with the PM2.5 standards by regulating the portion
of particulate pollution not regulated by the PM2.5 standards.
As explained in chapter 1, in May 1998, a three-judge panel of the U.S. Court of Appeals
for the District of Columbia Circuit found "ample support" for EPA's decision to regulate coarse
particle pollution, but vacated the 1997 PMio standards, concluding that EPA had failed to
adequately explain its choice of PMio as the indicator for thoracic coarse particles pointing to the
3-2
-------
lack of reasoned explanation for the variable level of allowable concentrations of thoracic coarse
particles (varying by levels of PM^.s) and the consequent double regulation of PM2.5. American
Trucking Associations v. EPA , 175 F. 3d 1027, 1054-56 (D.C. Cir. 1999). The court also
rejected considerations of administrative feasibility as justification for use of PMio as the
indicator for thoracic coarse PM, since NAAQS (and their elements) are to be based exclusively
on health and welfare considerations. Id at 1054. Pursuant to the court's decision, EPA
removed the vacated 1997 PMi0 standards from the Code of Federal Regulations (CFR) (69 FR
45592, July 30, 2004) and deleted the regulatory provision (at 40 CFR section 50.6(d)) that
controlled the transition from the pre-existing 1987 PMio standards to the 1997 PMio standards
(65 FR 80776, December 22, 2000). The pre-existing 1987 PMio standards remained in place. Id.
at 80777.
3.1.1.2 Review Completed in 2006
In the review of the PM NAAQS that concluded in 2006, EPA considered the growing,
but still limited, body of evidence supporting associations between health effects and thoracic
coarse particles measured as PMio-2.5.1 The new studies available in the 2006 review included
epidemiological studies that reported associations with health effects using direct measurements
of PMio-2.5, as well as dosimetric and toxicological studies. In considering this growing body of
PMio-2.5 evidence, as well as evidence from studies that measured PMio in locations where the
majority of PMio was in the PMio-2.5 fraction (US EPA, 2005, section 5.4.1), staff concluded that
that the level of protection afforded by the existing 1987 PMio standard remained appropriate
(US EPA, 2005, p. 5-67), but recommended that the indicator for the standard be revised.
Specifically, staff recommended replacing the PMio indicator with an indicator of urban thoracic
coarse particles in the size range of 10-2.5 |im (US EPA, 2005, pp. 5-70 to 5-71). The agency
proposed to retain a standard for a subset of thoracic coarse particles, proposing a qualified
PMio-2.5 indicator to focus on the mix of thoracic coarse particles generally present in urban
environments. More specifically, the proposed revised thoracic coarse particle standard would
have applied only to an ambient mix of PMio-2.5 dominated by resuspended dust from high-
density traffic on paved roads and/or by industrial and construction sources. The proposed
revised standard would not have applied to any ambient mix of PMio-2.5 dominated by rural
windblown dust and soils. In addition, agricultural sources, mining sources, and other similar
sources of crustal material would not have been subject to control in meeting the standard (71 FR
2667 to 2668, January 17, 2006).
JThe PM Staff Paper (US EPA, 2005) also presented results of a quantitative assessment of health risks for PM10.2.5
(see also Abt Associates, 2005). However, staff concluded that the nature and magnitude of the uncertainties and
concerns associated with this risk assessment weighed against its use as a basis for recommending specific levels for
a thoracic coarse particle standard (US EPA, 2005, p.5-69).
3-3
-------
The Agency received a large number of comments overwhelmingly opposed to the
proposed qualified PMio-2.5 indicator (71 FR 61188 to 61197). After careful consideration of the
scientific evidence and the recommendations contained in the 2005 Staff Paper, the advice and
recommendations from CASAC, the public comments received regarding the appropriate
indicator for coarse particles, and after extensive evaluation of the alternatives available to the
Agency, the Administrator decided it would not be appropriate to adopt a qualified PMio-2.5
indicator. Underlying this determination was the decision that it was requisite to provide
protection from exposure to all thoracic coarse PM, regardless of its origin, rejecting arguments
that there are no health effects from community-level exposures to coarse PM in non-urban areas
(71 FR 61189). The EPA concluded that dosimetric, lexicological, occupational and
epidemiological evidence supported retention of a primary standard for short-term exposures that
included all thoracic coarse particles (i.e., particles of both urban and non-urban origin),
consistent with the Act's requirement that primary NAAQS provide an adequate margin of
safety. At the same time, the Agency concluded that the standard should target protection
toward urban areas, where the evidence of health effects from exposure to PMio-2.5 was strongest
(71 FR at 61193, 61197). The proposed indicator was not suitable for that purpose. Not only did
it inappropriately provide no protection at all to many areas, but it failed to identify many areas
where the ambient mix was dominated by coarse particles contaminated with urban/industrial
types of coarse particles for which evidence of health effects was strongest (71 FR 61193).
The Agency ultimately concluded that the existing indicator, PMio, was most consistent
with the evidence. Although PMio includes both coarse and fine PM, the Agency concluded that
it remained an appropriate indicator for thoracic coarse particles because, as reported by Schmidt
et al. (2005), fine particle levels are generally higher in urban areas and, therefore, a PMio
standard set at a single unvarying level will generally result in lower allowable concentrations of
thoracic coarse particles in urban areas than in non-urban areas. The EPA considered this to be
an appropriate targeting of protection given that the strongest evidence for effects associated
with thoracic coarse particles came from epidemiological studies conducted in urban areas and
that elevated fine particle concentrations in urban areas could result in increased contamination
of coarse fraction particles by PM2.5, potentially increasing the toxicity of thoracic coarse
particles in urban areas (71 FR 61195-96). Given the evidence that the existing PMio standard
afforded requisite protection with an adequate margin of safety, the Agency retained the level
and form of the 24-hour PMio standard.2
The Agency also revoked the annual PMio standard, in light of the conclusion in the PM
Criteria Document (US EPA, 2004, p. 9-79) that the available evidence does not suggest an
2Thus, the standard is met when a 24-hour average PM10 concentration of 150 |J.g/m3 is not exceeded more than one
day per year, on average over a three-year period.
3-4
-------
association with long-term exposure to PMio-2.5 and the conclusion in the Staff Paper (US EPA,
2005, p. 5-61) that there is no quantitative evidence that directly supports an annual standard.
In the same rulemaking, EPA also included a new FRM for the measurement of PMi0-2.5
in the ambient air (71 FR 61212 to 61213). Although the standard for thoracic coarse particles
does not use a PMio-2.5 indicator, the new FRM for PMio-2.5 was established to provide a basis for
approving FEMs and to promote the gathering of scientific data to support future reviews of the
PMNAAQS (section 1.3.4).
3.1.2 Litigation of 2006 Final Rule for Thoracic Coarse Particles
A number of groups filed suit in response to the final decisions made in the 2006 review.
See American Farm Bureau Federation and the National Pork Producers Council v. EPA (DC
Cir. 2009). Among the petitions for review were challenges from industry groups on the
decision to retain the PMi0 indicator and the level of the PMi0 standard and from environmental
and public health groups on the decision to revoke the annual PMio standard. The court upheld
both the decision to retain the 24-hour PMio standard and the decision to revoke the annual
standard.
First, the court upheld EPA's decision for a standard to encompass all thoracic coarse
PM, both of urban and non-urban origin. The court rejected arguments that the evidence showed
there are no risks from exposure to non-urban coarse PM. The court further found that EPA had
a reasonable basis not to set separate standards for urban and non-urban coarse PM, namely the
inability to reasonably define what ambient mixes would be included under either 'urban' or
'non-urban;' and the evidence in the record that supported EPA's appropriately cautious decision
to provide "some protection from exposure to thoracic coarse particles... in all areas." 559 F. 3d
at 532-33. Specifically, the court stated,
Although the evidence of danger from coarse PM is, as EPA recognizes,
"inconclusive," (71 FR 61193, October 17, 2006), the agency need not wait for
conclusive findings before regulating a pollutant it reasonably believes may pose
a significant risk to public health. The evidence in the record supports the EPA's
cautious decision that "some protection from exposure to thoracic coarse particles
is warranted in all areas." Id. As the court has consistently reaffirmed, the CAA
permits the Administrator to "err on the side of caution" in setting NAAQS.
559 F. 3d at 533.
The court also upheld EPA's decision to retain the level of the standard at 150 |ig/m3 and
to use PMio as the indicator for thoracic coarse particles. In upholding the level of the standard,
the court referred to the conclusion in the Staff Paper that there is "little basis for concluding that
the degree of protection afforded by the current PMio standards in urban areas is greater than
3-5
-------
warranted, since potential mortality effects have been associated with air quality levels not
allowed by the current 24-hour standard, but have not been associated with air quality levels that
would generally meet that standard, and morbidity effects have been associated with air quality
levels that exceeded the current 24-hour standard only a few times." 559 F. 3d at 534. The court
also rejected arguments that a PMio standard established at an unvarying level will result in
arbitrarily varying levels of protection given that the level of coarse PM would vary based on the
amount of fine PM present. The court agreed that the variation in allowable coarse PM accorded
with the strength of the evidence: typically less coarse PM would be allowed in urban areas
(where levels of fine PM are typically higher), in accord with the strongest evidence of health
effects from coarse particles. 559 F. 3d at 535-36. In addition, such regulation would not
impermissibly double regulate fine particles, since any additional control of fine particles
(beyond that afforded by the primary PM2.5 standard) would be for a different purpose: to prevent
contamination of coarse particles by fine particles. 559 F. 3d at 535, 536. These same
explanations justified the choice of PMio as an indicator, and provided the reasoned explanation
for that choice lacking in the record for the 1997 standard. 559 F. 3d at 536.
With regard to the challenge from environmental and public health groups, the court
upheld EPA's decision to revoke the annual PMio standard. Specifically, the court stated the
following:
The EPA reasonably decided that an annual coarse PM standard is not necessary
because, as the Criteria Document and the Staff Paper make clear, the latest
scientific data do not indicate that long-term exposure to coarse particles poses a
health risk. The CASAC also agreed that an annual coarse PM standard is
unnecessary. 559 F. 3d at 538-39.
3.1.3 General Approach Used in Current Review
Our approach relies most heavily on the health evidence, primarily the epidemiological
evidence, assessed in the ISA (US EPA, 2009a) and on available PM air quality information. As
discussed in more detail in the Quantitative Health Risk Assessment for Paniculate Matter -
Final (RA, US EPA, 2010a), we have not conducted a quantitative assessment of health risks
associated with PMio-2.5- Staff concluded that limitations in the monitoring network and in the
health studies that rely on that monitoring network, which would be the basis for estimating
PMio-2.5 health risks, would introduce significant uncertainty into a PMio-2.5 risk assessment such
that the risk estimates generated would be of limited value in informing review of the standard.
Therefore, staff concluded in the RA that a quantitative risk assessment for PMio-2.5 is not
supportable at this time (US EPA, 2010a, p. 2-6).
For purposes of this policy assessment, we seek to provide as broad an array of options
for consideration as is supportable by the available evidence and air quality information,
3-6
-------
recognizing that the final decisions on the primary PMio standard will reflect the judgments of
the Administrator. In developing these options for consideration, we consider the available
evidence and air quality information that informs overarching questions related to: (1) the
adequacy of the current 24-hour PMio standard to protect against effects associated with
exposures to thoracic coarse particles and (2) what potential alternative standard(s), if any,
should be considered in this review. In addressing these broad questions, we have organized the
discussions below around a series of more specific questions reflecting different aspects of each
overarching question. When evaluating the health protection afforded by the current or potential
alternative standards, we have taken into account the four basic elements of the NAAQS:
indicator, averaging time, form, and level.
Figure 3-1 provides an overview of the policy-relevant questions that frame our review,
as discussed more fully below. We believe that this general approach provides a comprehensive
basis to help inform the judgments required of the Administrator in reaching decisions about the
current and potential alternative primary standards meant to protect public health against
exposures to thoracic coarse particles.
3-7
-------
Figure 3-1. Overview of Approach for Review of Primary PMi0 Standard
Adequacy of Current 24-Hour PM10 Standard?
I
Evidence-Based Considerations
'r Does available PM10.2 5 health evidence support or call into
question associations with mortality/morbidity?
/'Uncertainties in the evidence?
'r Evidence for associations in locations that would likely
meet current PM,n standard?
Does information call into
question adequacy of
current 24-hour PM
standard?
Consider retaining current
24-hour PM standard
Consider Potentia A ternative Standards?
Indicator
*-Support for retaining PM10?
^Support for alternative
indicator (e.g., PM10_25)?
Averaging Time
'r Support for continued use of 24-
hour averaging time?
'r Support for alternative averaging
time(s)?
r Support for retaining one-expected-
exceedance form?
'r Support for 98th percentile form or
other alternative forms?
Level
-Evidence-based considerations (U.S. epidemiologic studies)
•Relative weight placed on different studies?
•PM10 air quality in study locations?
•Uncertainties in the evidence?
-Air quality-based considerations
•Identify 98th percentile PM10 concentration(s) that are "generally equivalent" to
current standard level
Identify range of potential alternative standard levels for consideration
-------
3.2 ADEQUACY OF THE CURRENT PM10 STANDARD
In considering the adequacy of the current 24-hour PMi0 standard to protect against
effects associated with exposures to thoracic coarse particles, we address the following
overarching question:
Does the available scientific evidence, as reflected in the ISA, support or call into question
the adequacy of the protection afforded by the current 24-hour PMio standard against
effects associated with exposures to thoracic coarse particles?
To inform our consideration of this overarching question, we consider the scientific
evidence for associations between PMi0-2.5 and mortality and morbidity, evidence linking
PMio-2.5 toxicity to specific sources/locations, uncertainties in the evidence, and available PMio
air quality concentrations in PMio-2.5 study locations (section 3.2.1). Evidence for populations
that are particularly susceptible to PM exposures is discussed in detail in section 2.2.1 above, and
is not repeated here. Staff conclusions regarding the adequacy of the current standard are
presented in section 3.2.2.
3.2.1 Evidence-Based Considerations
In considering the currently available body of scientific evidence for health effects of
thoracic coarse particles, we consider the following question:
• To what extent does the currently available scientific evidence, including associated
uncertainties, strengthen or call into question evidence of associations between ambient
thoracic coarse particle exposures and adverse health effects?
Since the conclusion of the last review, the Agency has developed a more formal
framework for reaching causal inferences from the body of scientific evidence. As discussed
above in section 2.2.1, this framework uses a five-level hierarchy that classifies the overall
weight of evidence using the following categorizations: causal relationship, likely to be a causal
relationship, suggestive of a causal relationship, inadequate to infer a causal relationship, and not
likely to be a causal relationship (US EPA 2009a, section 1.5, Table 1-3). Applying this
framework to thoracic coarse particles, the ISA concludes that the existing evidence is suggestive
of a causal relationship between short-term PMio-2.5 exposures and mortality, cardiovascular
effects, and respiratory effects (US EPA, 2009a, section 2.3.3; see Table 3-1 below). In contrast,
the ISA concludes that available evidence is inadequate to infer a causal relationship between
long-term PMio-2.5 exposures and various health effects (US EPA, 2009a, section 2.3; Table 3-1
below). Similar to the judgment made in the AQCD regarding long-term exposures (US EPA,
2004), the ISA states, "To date, a sufficient amount of evidence does not exist in order to draw
conclusions regarding the health effects and outcomes associated with long-term exposure to
3-9
-------
PMio-2.s" (US EPA, 2009a, section 2.3.4). Given these weight of evidence conclusions in the
ISA, our evidence-based considerations regarding the adequacy of the current 24-hour
standard focus on effects that have been linked with short-term exposures to PMio-2.5.
Table 3-1. Summary of Causality Determinations for PMio_2.s
Exposure Duration
Short-term
Long-term
Outcome
Mortality
Cardiovascular Effects
Respiratory Effects
Central Nervous System Effects
Mortality
Cardiovascular Effects
Respiratory Effects
Reproductive and Developmental Effects
Cancer Mutagenicity, Genotoxicity Effects
Causal Determination
Suggestive
Suggestive
Suggestive
Inadequate
Inadequate
Inadequate
Inadequate
Inadequate
Inadequate
Source: adapted from US EPA, 2009a; Table 2-6
As noted above, in the last review of the PM NAAQS, PMio studies conducted in
locations where PMio is comprised predominantly of PMi0-2.5 were also considered (US EPA,
2005, pp. 5-49 to 5-50). However, PMio studies are difficult to interpret within the context of a
standard meant to protect against exposures to PMio-2.5 because PMio is comprised of both fine
and coarse particles, even in locations with the highest concentrations of PMio-2.5 (see below). In
light of the considerable uncertainty in the extent to which PMio effect estimates reflect
associations with PMio-2.5 versus PM2.s, together with the availability in this review of several
studies that evaluated associations with PMio-2.5 and the fact that the ISA weight of evidence
conclusions for thoracic coarse particles were based on studies of PMio-2.5, we focus in this PA
on studies that have specifically evaluated PMi0.2.5. The evidence supporting a link between
short-term thoracic coarse particle exposures and adverse health effects is discussed in detail in
the ISA (US EPA, 2009a, Chapter 6) and is summarized briefly below for mortality,
cardiovascular effects, and respiratory effects.
Short-Term PMui^and Mortality
The ISA assesses a number of multi-city and single-city epidemiological studies that have
evaluated associations between mortality and short-term PMio-2.5 concentrations (US EPA,
2009a, Figure 6-30 presents PMio-2.5 mortality studies assessed in the last review and the current
3-10
-------
review). Different studies have used different approaches to estimate ambient PMio-2.5- Some
studies have used the difference between PMio and PM2.5 mass, either measured at co-located
monitors (e.g., Lipfert et al., 2000; Mar et al., 2003; Ostro et al., 2003; Sheppard et al., 2003;
Wilson et al., 2007) or as the difference in county-wide average concentrations (Zanobetti and
Schwartz, 2009), while other studies have measured PMio-2.5 directly with dichotomous samplers
(e.g., Burnett and Goldberg, 2003; Fairley et al., 2003; Burnett et al., 2004; Klemm et al., 2004).
Despite differences in the approaches used to estimate ambient PMio-2.5 concentrations, the
majority of multi- and single-city studies have reported positive associations between PMio-2.5
and mortality, though most of these associations were not statistically significant (US EPA,
2009a, Figure 6-30). When considered as a whole, the ISA concluded that epidemiological
studies have reported consistent, positive associations between short-term PMio-2.5 and mortality
(US EPA, 2009a, section 6.5.2.3).
In considering specific mortality studies, we note that the U.S. multi-city study by
Zanobetti and Schwartz (2009) reported positive and statistically significant associations with
PMio-2.5 for all-cause, cardiovascular-related, and respiratory-related mortality (US EPA, 2009a,
section 6.5.2.3) while other multi-city studies have reported positive, but not statistically
significant, PMio-2.5 effect estimates for mortality (US EPA, 2009a, Figure 6-30, Burnett and
Goldberg, 2003; Klemm et al., 2003; Burnett et al., 2004). In the study by Zanobetti and
Schwartz, the effect estimates for all-cause and respiratory-related mortality remained
statistically significant in co-pollutant models that included PM2.5, while the effect estimate for
cardiovascular-related mortality remained positive but not statistically significant. When risk
estimates in this study were evaluated by climatic region (US EPA, 2009a, Figure 6-28), the "dry
continental" region, which included areas with relatively high PMio-2.5 concentrations such as
Salt Lake City, Provo, and Denver, showed the largest risk estimates (see US EPA, 2009a,
Figure 6-29; Schmidt and Jenkins, 2010; and discussion of regional differences in PMio-2.5
concentrations below). However, the "dry" region, which included Phoenix and Albuquerque,
two locations that also have relatively high PMio-2.5 concentrations, did not show positive
associations with all-cause or respiratory-related mortality and only a relatively small positive
association for cardiovascular-related mortality. In addition, the "Mediterranean" region (which
included cities in California, Oregon, and Washington) did not show positive associations while
the other three regions (i.e., "hot summer, continental," "warm summer, continental," and
"humid, subtropical and maritime"), which included cities that correspond to the mid-west,
northeast, and southeast geographic regions, all showed positive associations (US EPA, 2009a,
Figure 6-28).
The ISA also presents single-city empirical Bayes-adjusted effect estimates (Le Tertre et
al., 2005) for the 47 cities evaluated by Zanobetti and Schwartz (US EPA, 2009a, Figure 6-29).
3-11
-------
City-specific estimates were positive, though generally not statistically significant, for
cardiovascular-related mortality in all 47 cities evaluated. Effect estimates were positive for all-
cause and respiratory-related mortality in all cities except Los Angeles (negative association for
all-cause and respiratory-related mortality) and Phoenix (negative association for respiratory-
respiratory-related mortality) (US EPA, 2009a, Figure 6-29). In addition, positive and
statistically significant associations between mortality (all-cause, cardiovascular-related, and/or
respiratory-related) and PMi0-2.5 were reported for six locations (i.e., St. Louis, MO; Salt Lake
City, UT; Chicago, IL; Pittsburgh, PA; Detroit, MI; and Birmingham, AL).
In considering single-city PMio-2.5 mortality studies, we note that all of the studies
included in Figure 6-30 of the ISA (US EPA, 2009a) reported positive PMio-2.5 effect estimates,
with three single-city studies reporting effect estimates that were statistically significant (Mar et
al., 2003; Ostro et al., 2003; Wilson et al., 2007). One study reported a negative PMi0-2.5 effect
estimate for respiratory-related mortality (Villeneuve et al., 2003), though effect estimates for
all-cause and cardiovascular-related mortality were positive in this study (US EPA, 2009a,
Figure 6-30).
Short-Term PMi 0.2.5 and Cardiovascular Effects
The ISA assesses a number of studies that have evaluated the link between short-term
ambient concentrations of thoracic coarse particles and cardiovascular effects. In considering the
available epidemiological evidence, the ISA concludes that single- and multi-city
epidemiological studies generally report positive associations between short-term PMio-2.5
concentrations and hospital admissions or emergency department visits for cardiovascular causes
(US EPA, 2009a, section 2.3.3, 6.2.12.2). Some of these studies have reported positive and
statistically significant PMio-2.5 effect estimates in co-pollutant models while others report that
PMio-2.5 effect estimates remain positive, but not statistically significant (US EPA, 2009a, Figure
6-5).
These studies include a recent U.S. multi-city study evaluating hospital admissions and
emergency department visits for cardiovascular disease in Medicare patients (MCAPS, Peng et
al., 2008). In this study of older adults, the authors reported a positive and statistically
significant association between 24-hour PMio-2.5 concentrations and cardiovascular disease
hospitalizations in a single pollutant model using air quality data for 108 U.S. counties with co-
located PMio and PM2.s monitors. The magnitude of this effect estimate was larger in counties
with higher degrees of urbanization (Peng et al., 2008). The effect estimate was reduced only
slightly in a two-pollutant model that included PM2.5, but it was no longer statistically significant
(US EPA, 2009a, sections 2.3.3, 6.2.10.9). County-specific analyses were not reported for the
locations evaluated by Peng and, therefore, it is not possible to consider differences in PMio-2.5
3-12
-------
effect estimates in specific locations. Effect estimates for PMio-2.5 were larger in the eastern U.S.
than the western U.S., though this difference was not statistically significant (Peng et al., 2008).
In addition to this U.S. multi-city study, positive associations reported for short-term PMio-2.s and
cardiovascular-related morbidity reached statistical significance in a multi-city study in France
(Host et al., 2007) and single-city studies in Detroit (Ito, 2003) and Toronto (Burnett et al., 1999)
(US EPA, 2009a, Figures 6-2, 6-3). In contrast, associations were positive but not statistically
significant in single-city studies conducted in Atlanta (Metzger et al., 2004; Tolbert et al., 2007; )
and Boston (Peters et al., 2001) (and for some endpoints in Detroit) (US EPA, 2009a, Figures 6-1
to 6-3, 6-5).
The plausibility of the positive associations reported for PMio-2.5 and cardiovascular-
related hospital admissions and emergency department visits is supported by a small number of
controlled human exposure studies that have reported alterations in heart rate variability
following short-term exposure to PMio-2.5 (Gong et al., 2004; Graff et al., 2009); by short-term
PMio-2.5 epidemiological studies reporting positive associations with cardiovascular-related
mortality (see discussion above); by a small number of recent epidemiological studies that have
examined dust storm events and reported increases in cardiovascular-related emergency
department visits and hospital admissions (see below); and by associations with other
cardiovascular effects including heart rhythm disturbances and changes in heart rate variability
(US EPA, 2009a, sections 2.3.3, 6.2.12.2). The few toxicological studies that examined the
effect of PMio-2.5 on cardiovascular health effects used intratracheal instillation and, as a result,
provide only limited evidence on the biological plausibility of PMio-2.5 induced cardiovascular
effects (US EPA, 2009a, sections 2.3.3, 6.2.12.2).
Short-Term PMi 0.2.5 and Respiratory Effects
The ISA also assesses a number of studies that have evaluated the link between short-
term ambient concentrations of thoracic coarse particles and respiratory effects. This includes
recent studies conducted in the U.S., Canada, and France (US EPA, 2009a, section 6.3.8),
including the U.S. multi-city study of Medicare patients by Peng et al. (2009). As discussed
above, Peng estimated PMio-2.5 concentrations as the difference between PMio and PM2.5
concentrations measured by co-located monitors. The authors reported a positive, but not
statistically significant, PMio-2.5 effect estimate for respiratory-related hospital admissions.
Single-city studies have reported positive, and in some cases statistically significant, PMio-2.5
effect estimates for respiratory-related hospital admissions and emergency department visits (Lin
et al., 2002; Ito, 2003; Sheppard et al., 2003; Chen et al., 2004; Yang et al., 2004; Chen et al.,
2005; Lin et al., 2005; Peel et al., 2005; Slaughter et al., 2005; Fung et al., 2006; NYS DOH,
2006; Tolbert et al., 2007) (US EPA, 2009a, Figures 6-10 to 6-15). Some of these PMi0-2.5
3-13
-------
respiratory morbidity studies have reported positive and statistically significant PMio-2.5 effect
estimates in co-pollutant models that included gaseous pollutants while others reported that
PMio-2.5 effect estimates remain positive, but not statistically significant, in such co-pollutant
models (US EPA, 2009a, Figure 6-15).
A limited number of epidemiological studies have focused on specific respiratory
morbidity outcomes and reported both positive and negative, but generally not statistically
significant, associations between PMio-2.5 and lower respiratory symptoms, wheeze, and
medication use (US EPA, 2009a, sections 2.3.3.1 and 6.3.1.1; Figures 6-7, 6-8, 6-9). Although
controlled human exposure studies have not observed an effect on lung function or respiratory
symptoms in healthy or asthmatic adults in response to short-term exposure to PMio-2.5, healthy
volunteers have exhibited an increase in markers of pulmonary inflammation. Toxicological
studies using inhalation exposures are still lacking, but pulmonary injury and inflammation has
been reported in animals after intratracheal instillation exposure (US EPA, 2009a, section
6.3.5.3) and, in some cases, PMio-2.5 was found to be more potent than PM2.5.
PMin-7 s Toxicity: Impacts of Sources and Composition
As discussed above, positive, and in some cases statistically significant, associations
between short-term PMio-2.5 concentrations and mortality and morbidity have been reported in a
number of different locations. Little is known about how PMio-2.5 composition varies across
these locations and how that variation could affect particle toxicity (US EPA, 2009a, sections
2.3.3, 2.3.4, 2.4.4). However, the limited available evidence suggests that specific components
of thoracic coarse particles tend to comprise different fractions of PMio-2.5 mass in different
environments (e.g., urban versus rural environments) (US EPA, 2009a, section 3.5.1.1; Schmidt
et al., 2005; Edgerton et al., 2009). It is possible that such differences in particle composition
affect particle toxicity, though the ISA concludes that currently available evidence is insufficient
to draw distinctions in toxicity based on composition and notes that recent studies have reported
that PM (both PM2.5 and PMio-2.s) from different sources, including crustal sources, is associated
with adverse health effects (US EPA, 2009a, section 2.4.4). The evidence for associations with
particles originating from different types of sources and in different locations is discussed briefly
below.
As discussed above, most PMio-2.5 epidemiological studies have been conducted in urban
locations in the U.S., Canada, and Europe while a small number of studies have examined the
health impacts of dust storm events (US EPA, 2009a, sections 6.2.10.1, 6.5.2.3). Although these
dust storm studies do not link specific particle constituents to health effects, it is useful to
consider them within the context of the toxicity of particles of non-urban crustal origin. Several
3-14
-------
studies have reported positive and statistically significant associations between dust storm events
and morbidity or mortality, including the following:
• Middleton et al. (2008) reported that dust storms in Cyprus were associated with a
statistically significant increase in risk of hospitalization for all causes and a non-
significant increase in hospitalizations for cardiovascular disease.
• Chan et al. (2008) studied the effects of Asian dust storms on cardiovascular-related
hospital admissions in Taipei, Taiwan and reported a statistically significant increase
associated with 39 Asian dust events. Evaluating the same data, Bell et al. (2008) also
reported positive and statistically significant associations between hospitalization for
ischemic heart disease and PMio-2.5.
• Perez et al. (2008) tested the hypothesis that outbreaks of Saharan dust exacerbate the
effects of PMio-2.5 on daily mortality in Spain. During Saharan dust days, the PMio-2.5
effect estimate was larger than on non-dust days and it became statistically significant,
whereas it was not statistically significant on non-dust days.
In contrast to the studies noted above, some dust storm studies have reported associations
that were not statistically significant. Specifically, Bennett et al. (2006) reported on a dust storm
in the Gobi desert that transported PM across the Pacific Ocean, reaching western North America
in the spring of 1998. The authors reported no excess risk of cardiovascular-related or
respiratory-related hospital admissions associated with the dust storm in the population of British
Columbia's Lower Fraser Valley (Bennett et al., 2006). In addition, Yang et al. (2009) reported
that hospitalizations for congestive heart failure were elevated during or immediately following
54 Asian dust storm events, though effect estimates were not statistically significant. The
implications of these studies for the current review, specifically for consideration of potential
alternative indicators, are discussed below in section 3.3.1.
Next we consider uncertainties associated with the evidence by addressing the following
question:
• What are the important uncertainties associated with the currently available scientific
evidence that should be considered in evaluating the adequacy of the current PMi0
standard?
The majority of the health evidence supporting the link between short-term thoracic
coarse particle exposures and mortality and morbidity comes from epidemiological studies.
Although new studies have become available since the last review and have expanded our
understanding of the association between PMio-2.5 and adverse health effects (see above and U.S.
EPA, 2009a, Chapter 6), important uncertainties remain. These uncertainties, and their
implications for interpreting the scientific evidence, are discussed below.
3-15
-------
The ISA (sections 2.3.3, 2.3.4) concludes that an important uncertainty in the PMio-2.5
epidemiological literature is related to the air quality estimates used in these studies.
Specifically, the ISA concludes that there is greater error in estimating ambient exposures to
PMio-2.5 than to PM2.5 and that such uncertainty is a particularly relevant consideration when
interpreting PMio-2.5 epidemiological studies. Contributing to this uncertainty is the relatively
limited spatial coverage provided by the existing PMio-2.5 monitoring network (discussed in
section 1.3.4 above; US EPA, 2009a, sections 2.2.3, 2.3.3, 2.3.4, 3.5.1.1). As discussed above, a
national monitoring network for PMio-2.5 is not in place, limiting the spatial area over which
PMio-2.5 concentrations are measured. In addition, based on the limited available evidence, the
ISA concluded that "there is greater spatial variability in PMio-2.5 concentrations than PM2.5
concentrations, resulting in increased exposure error for the larger size fraction" (US EPA,
2009a, p. 2-8) and that available measurements do not provide sufficient information to
adequately characterize the spatial distribution of PMio-2.5 concentrations (US EPA, 2009a,
section 3.5.1.1). The net effect of these uncertainties on epidemiological studies of PMio-2.5 is to
bias the results of such studies toward the null hypothesis. That is, as noted in the ISA, these
limitations in estimates of ambient PMio-2.5 concentrations "would tend to increase uncertainty
and make it more difficult to detect effects of PMio-2.5 in epidemiologic studies" (US EPA 2009a,
p. 2-21).
Given these limitations in the available PMio-2.5 monitoring data, different
epidemiological studies have employed different approaches for estimating PMio-2.5
concentrations, further contributing to uncertainty in interpreting these studies. For example, as
discussed above, the multi-city study by Peng et al. (2008) estimated PMio-2.5 by taking the
difference between collocated PMio and PM2.5 monitors while the study by Zanobetti and
Schwartz (2009) used the difference between county average PMio and PM2.5 concentrations. A
small number of studies have directly measured PMio-2.5 concentrations with dichotomous
samplers (e.g., Burnett et al., 2004; Villeneuve et al., 2003; Klemm et al., 2004). It is not clear
how computed PMio-2.5 measurements, such as those used by Zanobetti and Schwartz, compare
with the PMio-2.5 concentrations obtained in other studies either by direct measurement with a
dichotomous sampler or by calculating the difference using co-located samplers (US EPA,
2009a, section 6.S.2.3).3 Given the use of these different approaches to estimating PMio-2.5
3In addition, several sources of uncertainty can be specifically associated with PM10-2.5 concentrations that are
estimated based on co-located monitors. For example, the potential for differences among operational flow rates
and temperatures for PM10 and PM2 5 monitors add to the potential for exposure misclassification. As discussed in
Appendix B, PM10 data are often reported at standard temperature and pressure (STP) while PM2 5 is reported at
local conditions (LC). In these cases, the PM10 data should be adjusted to local conditions when estimating PM10.2.5
concentrations. In many of the epidemiological studies that estimated PM10.25 concentrations based on co-located
3-16
-------
concentrations across studies, and their inherent limitations, the distributions of thoracic coarse
particle concentrations over which reported health outcomes occur remain highly uncertain.
The ISA also notes that the potential for confounding by co-occurring pollutants,
particularly PM2.5, has been addressed in only a relatively small number of PMio-2.5
epidemiological studies, introducing additional uncertainty into the interpretation of these studies
(US EPA, 2009a, section 2.3.3). This is a particularly important consideration given the
relatively limited body of experimental evidence available to support the plausibility of
associations between PMio-2.5 itself and health effects reported in epidemiological studies. As
discussed above, many epidemiological studies that have evaluated co-pollutant models have
reported that PMio-2.5 effect estimates remain positive, but lose precision and are not statistically
significant in these models (US EPA, 2009a, Figures 6-5, 6-9, 6-15). The net effect of this
limitation in the number of epidemiological studies that have evaluated co-pollutant models,
combined with the limited number of supporting experimental studies, is to increase the
uncertainty associated with estimates of the extent to which PMio-2.5 itself, rather than one or
more co-occurring pollutants, is responsible for the reported health effects.
Another uncertainty results from the relative lack of information on the chemical and
biological composition of PMio-2.5, and the effects associated with the various components (US
EPA, 2009a, section 2.3.4). As discussed above, a few recent studies have evaluated
associations between health effects and particles of non-urban, crustal origin by evaluating the
health impacts of dust storm events. Though these studies provide some information on the
health effects of particles that likely differ in composition from the particles of urban origin that
are typically studied, without more information on the chemical speciation of PMio-2.5, the
apparent variability in associations with health effects across locations is difficult to characterize
(US EPA, 2009a, section 6.5.2.3).
As discussed above, a 24-hour PMio standard is in place to protect the public health
against exposures to thoracic coarse particles. Therefore, in further considering the adequacy of
the current PMio standard, we ask the following question:
• To what extent does the available scientific evidence report associations between
PMio-i.s and morbidity and mortality in areas that would likely meet the current
standard?
4.
In addressing this question, we have used EPA's AQS to characterize PM
10
concentrations in U.S. locations where both single-city and multi-city PMio-2.5 studies have been
monitors, it is not made explicitly clear whether this adjustment was made, adding to the overall uncertainty in the
PM10-2 5 concentrations that are associated with health effects.
4 Accessible at http://www.epa.gov/ttn/airs/airsaqs/
3-17
-------
conducted (see U.S. EPA, 2009a, Figures 6-1 to 6-30 for studies). When compared to single-city
studies, we note that multi-city studies assess PMio-2.s-associated health effects among larger
study populations, providing enhanced power to detect PMi0-2.5-associated health effects. In
addition, multi-city studies often provide spatial coverage for different regions across the
country, reflecting differences in PMio-2.5 sources, composition, and potentially other factors that
could impact PMi0-2.5-related effects. These factors make multi-city studies particularly
important when drawing conclusions about health effect associations. However, multi-city
studies often present overall effect estimates rather than single-city effect estimates, while short-
term air quality can vary considerably across cities. Therefore, the extent to which effects
reported in multi-city studies are associated with the short-term air quality in any particular
location is uncertain, especially when considering short-term concentrations at the upper end of
the distribution of daily concentrations for pollutants with relatively heterogeneous spatial
distributions such as PMio-2.5 and PMio (US EPA, 2009a, section 2.1.1.2). In contrast, single-city
studies are more limited in terms of power and geographic coverage but the link between
reported health effects and the short-term air quality in a given city is more straightforward to
establish. As a result, in considering 24-hour PMio concentrations in locations of
epidemiological studies, we have focused below primarily on single-city studies (Figures 3-2 and
3-3) and single-city analyses of the locations evaluated in the multi-city study by Zanobetti and
Schwartz (US EPA, 2009a, Figure 6-29).
The current PMio standard has a form of one-expected-exceedance per year, averaged
over 3 years.5 In order to compare PMio concentrations in study locations to the level of the
current standard, we have identified the PMio 3-year expected exceedance concentration-
equivalent design value for each study period (labeled "DV" in Figures 3-2 and 3-3 below) using
the protocol specified in the PMio State Implementation Plan (SIP) Development Guidelines (US
EPA, 1987).6 Some studies (indicated by a * in Figure 3-3) covered time periods of less than
three years. For these study areas, to characterize ambient PMio concentrations relative to
concentrations allowed under the current PMio standard, we averaged the second highest 24-hour
5The one-expected-exceedance form implies that the standard level is not to be exceeded more than once per year,
on average over 3 years. Therefore, in areas that report 24-hour PM10 concentrations every day, the 4th highest 24-
hour PM10 concentration measured during a three year period is compared to the standard level. In contrast, in areas
that monitor PM10 every six days or every three days, the PM10 concentrations that are comparable to the standard
level are, respectively, the highest and 2nd highest 24-hour PM10 concentrations measured during a three year period.
6Specifically, the PM10 3-year expected exceedance concentration-equivalent design value is identified as the highest
24-hour average concentration (i.e., from a single monitor in the study area) over a 3-year period when there are 347
or fewer samples reported for that time-frame, the second highest 24-hour average concentration when there are 348
to 695 samples in the 3-year period, the third highest 24-hour average concentration when there are 696 to 1042
samples in the 3-year period, and the fourth highest 24-hour average concentration when there are 1043 or more
samples reported over the 3-year period. Concentration-equivalent design values were not identified for study
periods less than 3 years.
3-18
-------
Figure 3-2. PM10 Air Quality and PM10-2.s Effect Estimates in Locations of U.S. Single-City PM10-2.s Mortality Studies4
Study
Ktemm, 2004
Ito. 2003
Lipfert, 2000
Fairley, 2003
Chock, 2000
Mar, 2003
Wilson, 2007
Ostro, 2003
Location
Atlanta
Detroit
Philadelphia
Santa Clara
Pittsburgh
Phoenix
Phoenix (outer)
(central)
Coachella Valley
Mortality Endpoint
AH Cause
Respiratory
Cardiovascular
All Cause
Cardiovascular
Respiratory
All Cause
Cardiovascular
Cardiovascular
Cardiovascular
Time Period
1998-2000
1992-1994
1992-1995
1989-1996
1989-1991
1995-1997
1995-1997
1989-1998
Age
65+
65+
All
All
75+
<75
All
25+
All
PM10 DV (u.g/m3)
76
123
129
130
199
307
307
315
Effect Estimate (95% Cl)
_-
-P-
—
—
0.95 0 1.05 1.10 1.15
^Studies in Figures 3-2 and 3-3 are a combination of those assessed in the last review and those assessed in the ISA in the current review. Studies in these
figures are ordered by increasing PMi0 concentration-equivalent design values.
3-19
-------
Figure 3-3. PM10 Air Quality PM10-2.s Effect Estimates in Locations of U.S. Single-City PM10-2.s Morbidity Studies
Study
Peters, 2001
ISIYS DOH, 2006
Peel. 2005
Metzger. 2004
Tolbert, 2007
Slaughter, 2005
I to, 2003
Sheppard, 2003
Location
Boston
Bronx
Manhattan
Atlanta
Atlanta
Atlanta
Spokane
Detroit
Seattle
Morbidity Endpoint
Ml HA
Asthma HA
URI ED visits
Pneumonia ED visits
COPO ED visils
Asthma ED visits
RD ED visits
!HO ED visits
CHF ED visits
RD ED visits
CVD ED visits
COPD
Asthma
RD
Rl HA
COPD HA
IHD HA
CHF HA
Asthma HA
Time Period
1995-1996
1999-2000
1993-2000
1993-2000
1993 2004
1995-2GU1
1992-1994
1987-1994
Age
Mean 62
Ail
All
Ail
65+
AH
AH
Ail
All
65+
All
All
65+
All
PM10DV(ng/rr.3)
49-
49*
57*
78
78
102
110
123
136
Effect Estimate (95% Cl)
'•
I
\
i
|
i
" !
1
f
J
•i
|
i _ .
\ * *
i
I
:
!
>-
i
*-
:
1
I "
j _
i
1 *
j m
i
0.8 0.9
1.10 1.20
* Concentration-equivalent design values were not identified for study periods less than three years. For study periods of less than three years, we averaged the
second highest 24-hour PM10 concentrations for each year of the study (i.e., second highest concentration measured by the single monitor in the study area
recording the highest such concentration).
3-20
-------
concentrations for each year of the study (i.e., second highest concentration measured at
the single monitor in the study area recording the highest such concentration). The identification
of concentration-equivalent design values and second highest PMio concentrations for each study
area are described in more detail in Schmidt and Jenkins (2010) and Jenkins (2011).
In addition to the single-city studies included in Figures 3-2 and 3-3 above, multi-city
averages of the 3-year expected exceedance concentration-equivalent design values for U.S.
multi-city studies were 110 |ig/m3, for the locations evaluated by Zanobetti and Schwartz (2009)
(see Jenkins, 2011 for PMio air quality concentrations), and 100 |ig/m3, for the locations
evaluated by Peng et al. (2008) (see Schmidt and Jenkins, 2010 for PMio air quality
concentrations). As discussed above, the extent to which overall PMio-2.5 effect estimates
reported in multi-city studies are associated with the air quality in any particular location is
uncertain. However, the ISA also presents single-city Bayes-adjusted effect estimates for each
of the cities evaluated by Zanobetti and Schwartz (US EPA, 2009a, Figure 6-29), providing the
opportunity to consider associations between PMio-2.5 and mortality, and to consider the PMio air
quality, in each of the individual cities evaluated in this study.
As discussed above, in single-city analyses in the locations evaluated by Zanobetti and
Schwartz, PMio-2.5 effect estimates for mortality were generally positive but not statistically
significant, and most were similar in magnitude and precision, particularly for cardiovascular-
related mortality, across a wide range of estimated PMio-2.5 concentrations (US EPA, 2009a,
Figure 6-29). Three-year PMio expected exceedance concentration-equivalent design values in
these cities ranged from 50 |ig/m3 (Davie, FL) to 283 |ig/m3 (Salt Lake City, UT). In most of the
cities evaluated (37 of the 45 for which concentration-equivalent design values could be
identified), concentration-equivalent design values were below 150 |ig/m3 (Jenkins, 2011). In
the six cities where positive and statistically significant PMio-2.5 mortality effect estimates were
reported (see above), concentration-equivalent design values were as follows (Jenkins, 2011):
• Chicago: 113 |ig/m3
• Pittsburgh: 139 |ig/m3
• Birmingham7: 154 |ig/m3
• Detroit: 165 |ig/m3
• St. Louis: 165 |ig/m3
• Salt Lake City: 283 |ig/m3
Therefore, while PMio-2.5 effect estimates in single-city analyses were not statistically significant
for most locations evaluated by Zanobetti and Schwartz, including some locations with
'According to rounding convention for the PM10 standard, a 24-hour PM10 concentration of 154 |ag/m3 would round
to 150 |J.g/m3 (71 FR 61144). Therefore, based on the PM10 one-expected-exceedance concentration-equivalent
design value, Birmingham would have been expected to just meet the current PM10 standard during the study period.
3-21
-------
concentrations well above those allowed by the current 24-hour PMio standard, positive and
statistically significant PMio-2.5 effect estimates were reported in two locations (Chicago,
Pittsburgh) with concentration-equivalent design values below 150 |ig/m3.7
In considering PMio-2.5 epidemiological studies conducted in Canada and elsewhere
outside the U.S., we note that we generally do not have access to PMio air quality information
beyond that published by the study authors. Many of these studies report PM concentrations
averaged across monitors, rather than from the highest monitor in the study area, and/or report
only mean or median concentrations. Lin et al. (2002) reported positive and statistically
significant associations between PMio-2.5 and asthma hospital admissions in children in Toronto
(US EPA, 2009a; Figures 6-12, 6-15). The authors reported a maximum PMio concentration
measured at a single monitor in the study area of 116 |ig/m3, indicating that the PMio air quality
in Toronto during this study would have been allowed by the current 24-hour PMio standard. In
contrast Middleton et al. (2008), who reported that dust storms in Cyprus were associated with a
statistically significant increase in risk of hospitalization for all causes and a non-significant
increase in hospitalizations for cardiovascular diseases, reported a maximum 24-hour PMio
concentration of 1,371 |ig/m3. Thus, the dust storm-associated increases in hospitalizations
reported in this study occurred in an area with PMio concentrations that were likely well above
those allowed by the current standard. Other dust storm studies did not report maximum 24-hour
PMio concentrations from individual monitors, though the studies by Chan et al. (2008) and Bell
et al. (2008), which reported positive and statistically significant associations between dust storm
metrics and cardiovascular-related hospital admissions, reported that 24-hour PMio
concentrations, averaged across monitors, exceeded 200 |ig/m3. It is likely that peak
concentrations measured at individual monitors in these studies were much higher and, therefore,
24-hour PMio concentrations in these study areas were likely above those allowed by the current
standard.
Summary of Evidence-Based Considerations
New evidence supporting an association between PMio-2.5 and mortality and morbidity
has become available since the last review of the PM NAAQS. The available evidence was
judged in the ISA to be suggestive of a causal relationship between short-term PMio-2.5
exposures and mortality, cardiovascular effects, and respiratory effects while the evidence was
judged inadequate to infer a causal relationship with long-term PMio-2.5 exposures for these
same broad health effect categories as well as other effects considered in the ISA (US EPA,
2009a, section 2.3.3; see Table 3-1 above). The evidence supporting a link between short-term
thoracic coarse particle exposures and adverse health effects comes primarily from
epidemiological studies, with limited supporting evidence from controlled human exposure
3-22
-------
studies and, to a lesser extent, animal instillation studies. This evidence includes several recent
(i.e., published since the last review of the PM NAAQS) multi-city epidemiological studies
conducted in the U.S., Canada, and Europe and a small number of recent studies of particles of
non-urban origin. In general, epidemiological studies have reported positive, and in some cases
statistically significant, PMio-2.5 effect estimates. In the limited number of studies that have
evaluated co-pollutant models that include either gaseous pollutants or fine particles, PMio-2.5
effect estimates generally remained positive, and in a few cases statistically significant.
Positive associations between PMio-2.5 and mortality and morbidity have been reported in
a number of locations across the U.S. with a wide range of PMio-2.5 and PMio concentrations.
Among single-city analyses, PMio-2.5 effect estimates were positive and statistically significant in
a few U.S. cities and at least one Canadian city with ambient PMio concentrations that would be
allowed by the current 24-hour PMio standard. In addition, multi-city average PMio one-
expected-exceeded concentration-equivalent design values were below the level of the current
PMio standard when averaged across U.S. cities that have been evaluated in multi-city studies
reporting positive, and in some cases statistically significant, associations between PMio-2.5 and
mortality and morbidity. However, most PMio-2.5 effect estimates, even those reported in
locations with PMio concentrations above the concentrations allowed by the current standard,
were not statistically significant.
3.2.2 CASAC Conclusions and Recommendations
Following their review of the first and second draft PAs, CASAC provided advice and
recommendations regarding the current and potential alternative standards for thoracic coarse
particles (Samet, 2010c; Samet, 2010d). With regard to the current PMio standard, CASAC
concluded that "the current data, while limited, is sufficient to call into question the level of
protection afforded the American people by the current standard" (Samet, 2010d, p. 7). In
drawing this conclusion, CASAC noted the positive associations in multi-city and single-city
studies, including in locations with PMio concentrations below those allowed by the current
standard. In addition, CASAC gave "significant weight to studies that have generally reported
that PMio-2.5 effect estimates remain positive when evaluated in co-pollutant models" and
concluded that "controlled human exposure PMio-2.5 studies showing decreases in heart rate
variability and increases in markers of pulmonary inflammation are deemed adequate to support
the plausibility of the associations reported in epidemiologic studies" (Samet, 2010d, p. 7).
Given all of the above conclusions CASAC recommended that "the primary standard for PMio
should be revised" in order to increase public health protection (Samet, 2010d, p. ii and p. 7).
-------
3.2.3 Staff Conclusions on Adequacy of Current PMi0 Standard
In light of the available PMi0-2.5 health evidence and the PMi0 air quality concentrations
in study locations, as discussed above, we revisit the overarching question: Does the currently
available scientific evidence, as reflected in the ISA, and air quality information support or call
into question the adequacy of the protection afforded by the current 24-hour PMio standard
against effects associated with exposures to thoracic coarse particles?
In considering the evidence and information as they relate to the adequacy of the current
24-hour PMio standard we note that, as discussed above, this standard is meant to protect the
public health against effects associated with short-term exposures to PMio-2.5- In the last review,
it was judged appropriate to maintain such a standard given the "growing body of evidence
suggesting causal associations between short-term exposure to thoracic coarse particles and
morbidity effects, such as respiratory symptoms and hospital admissions for respiratory diseases,
and possibly mortality" (71 FR 61185, October 17, 2006). Given the expanded body of evidence
available in the current review, discussed in detail in the ISA (US EPA, 2009a, Chapter 6) and
summarized above, we conclude that the evidence continues to support the appropriateness of a
standard to protect the public health against effects associated with short-term exposures to
PMio-2.5- In addition, when considering the evidence for associations with PMi0-2.5 from different
types of sources and in different locations (e.g., thoracic coarse particles of urban/industrial
origin as well as windblown dust of non-urban origin), we conclude that it remains appropriate to
provide some measure of protection against exposures to all thoracic coarse particles.
In considering the evidence, we note that a decision on the adequacy of the public health
protection provided by the current PMio standard will be a public health policy judgment in
which the Administrator weighs that evidence and its inherent uncertainties. Therefore,
depending on the emphasis placed on different aspects of the evidence and uncertainties,
consideration of different conclusions on adequacy could be supported.
For example, one approach to considering the evidence and its associated uncertainties
would be to place emphasis on the following:
• While important uncertainties are associated with the health evidence, several multi-city
epidemiological studies conducted in the U.S., Canada, and Europe, as well as a number
of single-city studies, have reported generally positive, and in some cases statistically
significant, associations between short-term PMio-2.5 concentrations and adverse health
endpoints including mortality and cardiovascular-related and respiratory-related hospital
admissions and emergency department visits.
• Both single-city and multi-city analyses, using different approaches to estimate ambient
PMio-2.5 concentrations, have reported positive PMio-2.5 effect estimates in locations that
would likely have met the current 24-hour PMio standard. In some cases, these PMio-2.5
effect estimates were statistically significant.
3-24
-------
• While limited in number, studies that have evaluated co-pollutant models have generally
reported that PMi0-2.5 effect estimates remain positive, and in a few cases statistically
significant, when these models include gaseous pollutants or fine particles.
• Support for the plausibility of the associations reported in epidemiological studies is
provided by a small number of controlled human exposure studies reporting that short-
term (i.e., 2-hour) exposures to PMio-2.5 decrease heart rate variability and increase
markers of pulmonary inflammation.
Such an approach to considering the evidence would place substantial weight on the generally
positive PMio-2.5 effect estimates that have been reported for mortality and morbidity, even those
effect estimates that are not statistically significant. This could be judged appropriate given that
consistent results have been reported across multiple studies using different approaches to
estimate ambient PMio-2.5 concentrations and that exposure measurement error, which is likely to
be larger for PMio-2.5 than for PM2.5, tends to bias the results of epidemiological studies toward
the null hypothesis, making it less likely that associations will be detected. In contrast, such an
approach would place relatively little weight on the uncertainties in the evidence that resulted in
the ISA conclusions that the evidence is only "suggestive" of a causal relationship, rather than
"likely causal" or "causal." To the extent that a decision on the adequacy of the current 24-hour
PMio standard were to place emphasis on the considerations noted above, it could be judged that
the current 24-hour PMio standard does not protect public health with an adequate margin of
safely and that it should be revised in order to increase protection against effects associated with
short-term exposures to thoracic coarse particles.
Another approach to considering the evidence and its uncertainties would be to place
emphasis on the following:
• While most of PMio-2.5 effect estimates reported for mortality and morbidity were
positive, many were not statistically significant, even in single-pollutant models. This
includes effect estimates reported in study locations with PMio concentrations above
those allowed by the current 24-hour PMio standard.
• The number of epidemiological studies that have employed co-pollutant models to
address the potential for confounding, particularly by PM2.5, remains limited.
Therefore, the extent to which PMio-2.5 itself, rather than one or more co-pollutants,
contributes to reported health effects remains uncertain.
• Only a limited number of experimental studies provide support for the associations
reported in epidemiological studies, resulting in further uncertainty regarding the
plausibility of the associations between PMio-2.5 and mortality and morbidity reported
in epidemiological studies.
3-25
-------
• Limitations in PMio-2.5 monitoring data and the different approaches used to estimate
PMio-2.5 concentrations across epidemiological studies result in uncertainty in the
ambient PMio-2.5 concentrations at which the reported effects occur.
• The chemical and biological composition of PMio-2.5, and the effects associated with
the various components, remains uncertain. Without more information on the chemical
speciation of PMio-2.5, the apparent variability in associations across locations is
difficult to characterize.
• In considering the available evidence and its associated uncertainties, the ISA
concluded that the evidence is "suggestive" of a causal relationship between short-term
PMio-2.5 exposures and mortality, cardiovascular effects, and respiratory effects. These
weight-of-evidence conclusions contrast with those for the relationships between PM2.5
exposures and adverse health effects, which were judged in the ISA to be either
"causal" or "likely causal" for mortality, cardiovascular effects, and respiratory effects.
To the extent that a decision on the adequacy of the current 24-hour PMio standard were to place
emphasis on the considerations noted above, it could be judged that, while it remains appropriate
to maintain a standard to protect against short-term exposures to thoracic coarse particles, the
available evidence suggests that the current 24-hour PMio standard appropriately protects public
health and that it provides an adequate margin of safety against effects that have been associated
with PMio-2.5. While this approach to considering the evidence would recognize the positive, and
in some cases statistically significant, associations between PMio-2.5 and mortality and morbidity
by maintaining a standard to protect against exposures to thoracic coarse particles, it would place
relatively greater emphasis on the limitations and uncertainties noted above, which tend to
complicate the interpretation of that evidence.
Given all of the above, we conclude that it would be appropriate to consider either
retaining or revising the current 24-hour PMio standard, depending on the approach taken to
considering the available evidence and information. Therefore, we judge that it is appropriate in
this PA to consider what potential alternative standards, if any, could be supported by the
available scientific evidence in order to increase public health protection against exposures to
PMio-2.5.
3-26
-------
3.3 CONSIDERATION OF POTENTIAL ALTERNATIVE STANDARDS
Staff next considers the following overarching question:
What potential alternative standard(s) could be supported by the currently available
scientific evidence and air quality information?
In addressing this overarching question, we consider how the currently available
scientific evidence and air quality information could inform decisions regarding the basic
elements of the NAAQS: indicator (section 3.3.1), averaging time (section 3.3.2), form (section
3.3.3), and level (section 3.3.4). These elements are considered collectively in evaluating the
health protection afforded by potential alternative standards under consideration.
3.3.1 Indicator
As discussed above, PMi0 includes both PMi0-2.5 and PM2 5, with the relative contribution
of each to PMio mass varying across locations (see below). In the most recent review completed
in 2006, EPA concluded that the PMio indicator remained appropriate because a PMio standard
would be expected to provide appropriate protection against effects associated with exposures to
PMio-2.5- In particular, a PMio indicator would be expected to target protection to urban areas,
where the evidence of effects from exposure to coarse PM is the strongest (71 FR at 61196). In
considering potential alternative standards in the current review, we have considered the
following question with regard to indicator:
• To what extent does the available evidence and/or air quality information provide
support for retaining or revising the current PMio indicator?
In addressing this question, we focus on the following considerations:
• The extent to which PMio is comprised of PMi0-2.5
• The appropriateness of a standard that would be expected to allow lower PMio-2.5
concentrations in areas with higher fine particle concentrations (i.e., urban areas) than
areas with lower fine particle concentrations (i.e., rural areas)8
As an initial matter, we consider the proportion of PMio mass in different regions of the
U.S. that is PMio-2.5 (see US EPA, 2009a, section 3.5.1.1; Schmidt and Jenkins, 2010). Schmidt
and Jenkins (2010) divided the U.S. into climatic regions using the same approach as used in the
multi-city epidemiological study by Zanobetti and Schwartz (2009) (see above).9 Consistent
8For comparisons of PM2 5 mass in urban and rural areas see Schmidt (2005, outputs for Attachment D).
9The Mediterranean region includes CA, OR, WA. The dry region includes NM, AZ, NV. The dry continental
region includes MT, ID, WY, UT, CO. The hot summer continental region includes SD, ME, IA, IL, IN, OH. The
warm summer continental region includes ND, MN, WI, MI, PA, NY, CT, RI, MA, VT, NH, ME. The humid
3-27
-------
with the air quality analyses in the ISA (US EPA, 2009a, section 3.5.1.1) and the concentration
estimates of Zanobetti and Schwartz (2009), PMio-2.5 concentrations were higher in the dry,
mediterranean, and dry continental regions, with the highest concentrations in the dry region,
which included the southwestern U.S. (data not shown, from Schmidt and Jenkins, 2010). On
average, ratios of PMio-2.5 concentrations to PMio concentrations were also higher in these
regions than in the remainder of the U.S., with the dry region having the highest ratios.
Consistent results were reported in the ISA analyses of PM air quality (US EPA, 2009a, compare
Tables 3-9 and 3-10). While the same general pattern persisted when the analysis was restricted
to days with PMio concentrations at the high end of the distribution of 24-hour concentrations
(i.e., at or above the 95th percentile values), ratios of PMio-2.5 to PMio on these days tended to be
somewhat higher, on average, across most regions of the U.S. (see right-hand columns in Figure
3-4).
Figure 3-4. Site-Level Ratio of 24-Hour PM10-2.s to 24-Hour PM10 Concentrations*
1.0
0.9
5e0.8
'0.7-
20.6 :
i 0.4 -
I 0.3 -
*0.2 :
0.1 -_
0.0 -
AII r.s.
N-162
N-65
ra
Dn' Conlincnlal
^ |
-7x. •«,
N=I71
Hot Summer
N-159
Warai Summer
Conlincnlal
!
N-232
Humid Subtropicii
and Mainline
! .
'^ •?/,
*Blue stars represent mean concentrations, horizontal lines represent median concentrations, boxes
represent 75% confidence intervals, and error bars represent 95% confidence intervals. N values
equal the number of site years of monitoring data for each region.
Thus, on average across the U.S., PMio-2.5 comprises a larger portion of PMio on days with
relatively high PMio concentrations than on days with more typical PMio concentrations. Given
this, a PMio standard that focuses on the upper end of the distribution of daily
concentrations could effectively control ambient PMio-2.5 concentrations.
subtropical and maritime region includes FL, LA, TX, GA, AL, MS, AR, OK, KS, MO, TN, SC, NC, VA, WV, KY,
NJ, DE, DC, MD.
3-28
-------
Given the above conclusion, in further considering the issue of indicator we note that
most of the evidence for positive associations between PMio-2.5 and morbidity and mortality,
particularly evidence for these associations at relatively low concentrations of PMio-2.5, continues
to come from studies conducted in locations where the PMio-2.5 is expected to be largely of urban
origin. While some studies have reported positive associations between relatively high
concentrations of particles of non-urban origin (i.e., crustal material from windblown dust in
non-urban areas, see above) and mortality and morbidity, we note that the extent to which these
associations would remain at the lower particle concentrations more typical of U.S. and
Canadian urban study locations remains uncertain.10 Given these considerations, and given the
increased potential for coarse particles in urban areas to become contaminated by toxic
components of fine particles from urban/industrial sources (US EPA, 2004), we conclude that it
remains appropriate to maintain a standard that allows lower ambient thoracic coarse particle
concentrations in urban areas than in non-urban areas.
Given this conclusion, we note that it would be reasonable to consider an indicator that
targets control on areas with the types of ambient mixes generally present in urban areas. Such
an indicator would focus control on areas with ambient mixes known with greater certainty to be
associated with adverse health effects and, therefore, would provide public health benefits with
the greatest degree of certainty. As noted in the last review of the PM NAAQS, a PMi0 standard
would allow lower concentrations of PMio-2.5 in areas with higher fine particle concentrations,
which tend to be urban locations, than areas with lower fine particle concentrations, which tend
to be rural locations (section 3.1.1.2 above). Therefore, as in the last review, we reach the
conclusion that a PMi0 indicator would appropriately target protection to those locations where
the evidence is strongest for associations between adverse health effects and exposures to
thoracic coarse particles. In contrast, we note that a PMio-2.5 indicator, for a standard set at a
single unvarying level, would not achieve this targeting, given that allowable thoracic coarse
particle concentrations would be the same regardless of the location or the likely sources of PM.
Therefore, given the currently available evidence, one possible result of using a PMio-2.5 indicator
would be a standard that is overprotective in rural areas and/or underprotective in urban areas.
In addition, while we note that administrative feasibility is not an appropriate basis for
informing decisions on the NAAQS or its elements (see above), we also note that PMio-2.5
concentrations are not routinely measured and reported at present (US EPA, 2009a, section
10Other than the dust storm studies, we note that the study in Coachella Valley by Ostro et al. (2003) reported
statistically significant associations in a location where thoracic coarse particles are expected to be largely due to
windblown dust. Specifically, we note the CASAC conclusion in the last review that "studies from Ostro et al.
showed significant adverse health effects, primarily involving exposures to coarse-mode particles arising from
crustal sources" (Henderson, 2005b). In considering this study, we also note the relatively high PM10 concentrations
in the study area (see Figure 3-2 above).
3-29
-------
3.5.1.1). In the last review of the PMNAAQS, EPA required monitoring of PMio-2.5 mass and
we expect that approximately 80 stations will be reporting mass concentrations as part of the
National Core (NCore) network (http://www.epa.gov/ttnamtil/ncore/index.html) (section 1.3.3
and Appendix B, section B.4). However, data from those monitors are not available for locations
and time periods of existing PMio-2.5 health studies.
In their review of the second draft PA, CAS AC agreed with staffs conclusions that the
available evidence supports consideration in the current review of a PMi0 indicator for a standard
that protects against exposures to thoracic coarse particles. Specifically, CASAC concluded that
"[w]hile it would be preferable to use an indicator that reflects the coarse PM directly linked to
health risks (PMio-2.s), CASAC recognizes that there is not yet sufficient data to permit a change
in the indicator from PMi0 to one that directly measures thoracic coarse particles" (Samet,
2010d, p. ii). In addition, CASAC "vigorously recommends the implementation of plans for the
deployment of a network of PMio-2.5 sampling systems so that future epidemiological studies will
be able to more thoroughly explore the use of PMio-2.5 as a more appropriate indicator for
thoracic coarse particles" (Samet, 2010d, p. 7).
Given all of the above considerations, as in the draft Policy Assessment, we conclude that
the available evidence supports consideration in the current review of a PMio indicator for a
standard that protects against exposures to thoracic coarse particles. We further conclude that
consideration of alternative indicators (e.g., PMio-2.s) in future reviews is desirable and could be
informed by additional research, as described below (section 3.5).
3.3.2 Averaging Time
Based primarily on epidemiological studies that reported positive associations between
short-term (24-hour) PMio-2.5 concentrations and mortality and morbidity, the Administrator
concluded in the last review that the available evidence supported a 24-hour averaging time for a
standard intended to control thoracic coarse particles. In contrast, given the relative lack of
studies supporting a link between long-term exposures to thoracic coarse particles and morbidity
or mortality (US EPA, 2004a, Chapter 9), the Administrator further concluded that an annual
coarse particle standard was not warranted at that time (71 FR 61198-61199).
In the current review, we consider the extent to which the available evidence provides
information relevant for decisions on averaging time by considering the following question:
• To what extent does the available evidence continue to support a 24-hour averaging
time for a standard meant to protect against effects associated with exposures to
PM102.5?
With regard to this question, we note the conclusions from the ISA regarding the weight
of evidence for short-term and long-term PMio-2.5 exposures as well as the studies on which those
3-30
-------
conclusions are based. Specifically, as discussed above (see Table 3-1 above), the ISA
concludes that the existing evidence is suggestive of a causal relationship between short-term
PMio-2.5 exposures and mortality, cardiovascular effects, and respiratory effects (ISA, section
2.3.3). This conclusion is based largely on epidemiological studies which have primarily
evaluated associations between 24-hour PMio-2.5 concentrations and morbidity and mortality
(e.g., see ISA, Figure 2-3), though a small number of controlled human exposure studies have
reported effects following shorter exposures (i.e., 2-hours) to PMio-2.5 (e.g., see ISA, sections
6.2.1.2, 6.3.3.2). In contrast, with respect to long-term exposures, the ISA concludes that
available evidence is inadequate to infer a causal relationship with all health outcomes
evaluated (US EPA, 2009a, section 2.3). Specifically, the ISA states, "To date, a sufficient
amount of evidence does not exist in order to draw conclusions regarding the health effects and
outcomes associated with long-term exposure to PMio^.s" (US EPA, 2009a, section 2.3.4; see
Table 3-1 above).
In considering these weight of evidence determinations in the ISA, we conclude that, at a
minimum, they suggest the importance of maintaining a standard that protects against short-term
exposures to thoracic coarse particles. Given that the majority of the evidence supporting the
link between short-term PMio-2.5 and morbidity and mortality is based on 24-hour average
thoracic coarse particle concentrations, we conclude that the evidence available in this review
continues to support consideration of a 24-hour averaging time for a PMi0 standard meant to
protect against effects associated with short-term exposures to PMio-2.5- We further conclude that
the available evidence does not support consideration of an annual thoracic coarse particle
standard at this time. In reaching this conclusion, we also note that, to the extent a short-term
standard requires areas to reduce their 24-hour ambient particle concentrations, long-term
concentrations would also be expected to decrease. Therefore, a 24-hour standard meant to
protect against short-term exposures to thoracic coarse particles would also be expected to
provide some protection against any potential effects associated with long-term exposures to
ambient concentrations. CAS AC agreed with these conclusions (Samet, 2010c).
3.3.3 Form
The "form" of a standard defines the air quality statistic that is to be compared to the
level of the standard in determining whether an area attains that standard. In identifying a single
statistic for the form, we note that although future air quality improvement strategies in any
particular area are not defined until after a standard is promulgated, many such strategies are
likely to affect a broad distribution of PM air quality concentrations in an area. Therefore,
although the form of the standard defines a single statistic, any reductions in health risks that are
3-31
-------
likely to result from strategies designed to meet a specific standard are likely to occur across a
wide range of concentrations.
As discussed above, in the last review the Administrator retained the one-expected
exceedance form of the primary 24-hour PMio standard. This decision was linked to the overall
conclusion that "the level of protection from coarse particles provided by the current 24-hour
PMio standard remains requisite to protect public health with an adequate margin of safety" (71
FR 61202). Because revising either the level or the form of the standard would have altered the
protection provided, it was concluded that such changes "would not be appropriate based on the
scientific evidence available at this time" (71 FR 21202). Therefore, the decision in the last
review to retain the one-expected-exceedance form was part of the broader decision that the
existing 24-hour standard provided requisite public health protection.
In the current review, we are also considering the form of the standard within the context
of the overall decision on whether, and if so how, to revise the current 24-hour PMio standard.
Given the conclusions above regarding the appropriate indicator and averaging time for
consideration for potential alternative standards, we consider potential alternative forms for a 24-
hour PMio standard. To frame our consideration of this issue, we pose the following question:
• To what extent does available evidence and information support consideration of an
alternative form for a 24-hour PMio standard?
Although the selection of a specific form must be made within the context of decisions on
the other elements of the standard, EPA generally favors concentration-based forms for short-
term standards. In 1997 EPA established a 98th percentile form for the 24-hour PM2.5 standard
and in 2010 EPA established a 98th percentile form for the 1-hour NO2 standard (62 FR 38671;
75 FR 6474) and a 99th percentile form for the 1-hour SO2 standard (75 FR 35541). In making
these decisions, EPA noted that, compared to an exceedance-based form, a concentration-based
form is more reflective of the health risks posed by elevated pollutant concentrations because
such a form gives proportionally greater weight to days when concentrations are well above the
level of the standard than to days when the concentrations are just above the level of the
standard. In addition, when averaged over three years, these concentration-based forms were
judged to provide an appropriate balance between limiting peak pollutant concentrations and
providing a stable regulatory target, facilitating the development of stable implementation
programs.
These considerations are also relevant in the current review of the 24-hour PMio standard.
Specifically, we conclude that it is appropriate to consider concentration-based forms that would
provide a balance between limiting peak pollutant concentrations and providing a stable
regulatory target. To accomplish this, it would be appropriate to consider forms from the upper
3-32
-------
end of the annual distribution of 24-hour PMio concentrations.11 However, given the potential
for local sources to have important impacts on monitored PMio concentrations (US EPA, 2009a,
section 2.1.1.2), we also note that it would be appropriate to consider forms that, when averaged
over three years, would be expected to promote the stability of local implementation programs.12
In considering these issues in the most recent review of the NC>2 primary NAAQS, we note that a
98th percentile form was adopted, rather than a 99th percentile form, due to the potential for
"instability in the higher percentile concentrations" near local sources (75 FR 6493).13
In considering the potential appropriateness of a 98th percentile form in the current
review, we note that, compared to the current PMio standard, attainment status for a PMio
standard with a 98th percentile form would be based on a more stable air quality statistic and
would be expected to be less influenced by relatively rare events that can cause elevations in
PMio concentrations over short-periods of time (Schmidt, 201 Ib). Specifically, we note that in
areas that monitor PMio every six days, every three days, or every day the one-expected-
exceedance concentrations that are comparable to the current standard level are, respectively, the
highest, 2nd highest, or 4th highest 24-hour PMio concentrations measured during a three year
period. In contrast, for the same monitoring frequencies, the PMio concentrations that would be
comparable to the level of a standard with a 98th percentile form would be the three-year average
of the 2nd highest, 3rd highest, or 7th/8th highest 24-hour PMio concentrations measured during a
single year.
In further considering this issue we note that, compared to the current expected-
exceedance form, a concentration-based form specified as a percentile of the annual distribution
of PMio concentrations (e.g., such as a 98th percentile form) would be expected to better
compensate for missing data and less-than-daily monitoring (Davidson and Hopke, 1984). This
is a particularly important consideration in the case of PMio because, depending largely on
ambient concentrations, the frequency of PMio monitoring differs across locations (i.e., either
daily, 1 in 2 days, 1 in 3 days, or 1 in 6 days) (Section 1.3.10 above and Appendix B). With a
98th percentile form, attainment status would be determined based on PMio concentrations from
nWith regard to this conclusion, we also note that PM10_25 is likely to make a larger contribution to PM10 mass on
days with relatively high PM10 concentrations than on days with more typical PM10 concentrations (see above).
1 Stability of implementation programs has been held to be a legitimate consideration in determining a NAAQS
(American Trucking Assn's v. EPA. 283 F. 3d at 374-75).
13See also, ATA III. 283 F. 3d at 374-75 (upholding 98th percentile form since "otherwise States would have to
design their pollution control programs around single high exposure events that may be due to unusual
meteorological conditions alone, rendering the programs less stable - and hence, we assume, less effective - than
programs designed to address longer-term average conditions."). In contrast, in the recently completed review of the
SO2 primary NAAQS, a 99th percentile form was adopted. However, in the case of SO2, the standard was intended
to limit 5-minute exposures and a 99th percentile form was markedly more effective at doing so than a 98th percentile
form. 75 FR at 35540, 41.
-------
the same part of the annual distribution of 24-hour PMio concentrations, regardless of the
frequency of PMio monitoring.
In light of all of the above considerations, we conclude that, to the extent it is judged
appropriate to revise the current 24-hour PMio standard, it would be appropriate to consider
revising the form to the 3-year average of the 98th percentile of the annual distribution of 24-hour
concentrations.14 CAS AC agreed, noting that they "felt strongly that it is appropriate to
change the statistical form of the PMio standard to a 98th percentile form" (Samet, 2010d). In
reaching this conclusion, CAS AC noted that "[p]ublished work has shown that the percentile
form has greater power to identify non-attainment and a smaller probability of misclassification
relative to the expected exceedance form of the standard" (Samet, 2010d).
However, CASAC also noted that such a change in form "will lead to changes in levels
of stringency across the country" and recommended that this issue be explored further. In
considering this issue, we acknowledge that, given differences in PMio air quality distributions
across locations (US EPA, 2009a, Table 3-10), a revised standard with a 98th percentile form
would likely target public health protection to some different locations than does the current
standard with its one-expected-exceedance form. Given this, we note that a further consideration
with regard to the appropriateness of revising the form of the current PMio standard is the extent
to which, when compared with the current standard, a revised standard with a 98th percentile
form would be expected to target public health protection to areas where we have more
confidence that ambient PMio-2.s is associated with adverse health effects.
In giving initial consideration to this issue, we have used recent PMio air quality
concentrations (i.e., from 2007-2009) to identify counties that would meet, and counties that
would violate, the current PMio standard as well as potential alternative standards with 98th
percentile forms (Schmidt, 201 lb).15' 16 In some cases, counties that would violate the current
standard do so because of a small number of "outlier" days (e.g., as few as one such day in three
years) with PMio concentrations well-above more typical concentrations (Schmidt, 201 lb). Mean
and 98th percentile PMio and PMio-2.5 concentrations were higher in counties that met the current
standard, but would have violated a revised standard with a 98th percentile form,17 than in
14As noted above, local sources can have important impacts on monitored PM10 concentrations. In the recent review
of the NO2 primary NAAQS, where this was also an important consideration, a 98th percentile form was adopted,
rather than a 99th percentile form, due to the potential for "instability in the higher percentile concentrations" near
local sources (75 FR 6493). A similar conclusion in the current review has led us to focus on the 98th percentile
rather than the 99th percentile.
15Section 3.3.4 discusses potential alternative standard levels that would be appropriate to consider in conjunction
with a revised standard with a 98th percentile form.
16The memo by Schmidt (20lib) identifies specific counties that are expected to meet, and counties that are
expected to violate, the current standard as well as potential alternative standards with 98th percentile forms.
17This analysis considered a revised PM10 standard with a 98th percentile form and a level from the middle of the
range discussed in section 3.3.4 (i.e., 75 |J.g/m3).
3-34
-------
counties that violated the current standard, but would have met a revised standard with a 98th
percentile form (Schmidt, 201 Ib). This suggests that, to the extent a revised PMio standard with a
98th percentile form could target public health protection to different areas than the current
standard, those areas preferentially targeted by a revised standard generally have higher ambient
concentrations of thoracic coarse particles. The issue of targeting public health protection is
considered further in section 3.3.4, within the context of considering specific potential alternative
standard levels for a 24-hour PMio standard with a 98th percentile form.
3.3.4 Level
As noted above, to the extent it is judged in the current review that the 24-hour PMio
standard does not provide adequate public health protection against exposures to thoracic coarse
particles, potential alternative standard levels could be considered. Given the conclusions
described above for indicator, averaging time, and form, we conclude that it would be
appropriate to consider potential alternative levels for a 24-hour PMio standard with a 98th
percentile form. To inform our consideration of this issue, we have considered the following
question:
• To what extent does available evidence and air quality information support
consideration of alternative standard levels for a 24-hour PMio standard with a 98th
percentile form?
Evidence-based Considerations
In considering the evidence as it relates to potential alternative standard levels, we first
consider the relative weight to place on specific epidemiological studies, including the weight to
place on the uncertainties associated with those studies. We have considered several factors in
placing weight on specific epidemiological studies including the extent to which studies report
statistically significant associations with PMio-2.5 and the extent to which the reported
associations are robust to co-pollutant confounding.
In addition, we consider the extent to which associations with PMio-2.5 can be linked to
the air quality in a specific location. With regard to this, we place greatest weight on information
from single-city analyses. Although, as discussed above, multi-city studies have advantages in
terms of power to detect associations and geographic coverage, the extent to which effects
reported in multi-city studies are associated with the short-term air quality in any particular
location is highly uncertain, especially when considering short-term concentrations at the upper
end of the distribution of daily concentrations for pollutants with relatively heterogeneous spatial
distributions such as PMio (US EPA, 2009a, section 2.1.1.2). In contrast, single-city studies are
more limited in terms of power and geographic coverage but the link between reported health
effects and the air quality in a given city is more straightforward to establish (US EPA, 2009a,
3-35
-------
section 2.1.1.2). Given this, in considering PMio concentrations in locations of epidemiological
studies, we place the most weight on single-city studies (Figures 3-2 and 3-3) and single-city
analyses of the locations evaluated in the multi-city study by Zanobetti and Schwartz (US EPA,
2009a, Figure 6-29).
In considering PM air quality in study locations, we also note that the available evidence
does not support the existence of thresholds, or lowest-observed-effects levels, in terms of 24-
hour average concentrations (US EPA, 2009a, section 2.4.3). In the absence of an apparent
threshold, for purposes of identifying a range of standard levels potentially supported by the
health evidence, we focus on the range of PMio concentrations that have been measured in
locations where U.S. epidemiological studies have reported associations with PMio-2.5 (see U.S.
EPA, 2009a, Figures 6-1 to 6-30 for studies). In characterizing PMio air quality in PMi0-2.5 study
locations, we have used EPA's AQS to identify the highest 98th percentile 24-hour PMio
concentrations for each year in each study location (i.e., from the monitor in the study area
recording the highest 98th percentile concentration), as described in Schmidt and Jenkins (2010)
and Jenkins (2011). The 98th percentile concentrations from each study year were averaged
together and these averages are presented below in Figure 3-5 (PMio-2.5 mortality studies) and
Figure 3-6 (PMio-2.5 morbidity studies) for locations of single-city studies.
3-36
-------
»th
Figure 3-5. 98 Percentile PM10 Concentrations in Locations of U.S. Single-City PM10-2.s Mortality Studies*
Study
Klemm, 2004
Ito, 2003
Chock. 2000
Mar, 2003
Wilson, 2007
Ostro, 2003
Lipfert, 2000
Location
Atlanta
Detroit
Pittsburgh
Phoenix
Phoenix (outer)
(central)
Coachella Valley
Philadelphia
Mortality Endpoint
All Cause
All Cause
Respiratory
Cardiovascular
All Cause
All Cause
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
Time Period
1998-2000
1992-1994
1989-1991
1995-1997
1995-1997
1989-1998
1992-1995
Age
65+
65+
75+
<75
All
25+
All
All
98th PM10 (n9/m3)
74
QA
102
136
163
163
200
244
Effect Estimate (95% Cl)
— —
-*-
—
~
0.95 0 1.05 1.10 1.15
These studies are a combination of those assessed in the last review and those assessed in the ISA in the current review. Studies are ordered by increasing 98th
percentile PM10 concentrations. See 2005 Staff Paper (US EPA, 2005, pp. 5-65 to 5-66) describing the levels from the two separate study monitors used in Ostro
et al. (2003).
3-37
-------
»th
Figure 3-6. 98 Percentile PM10 Concentrations in Locations of U.S. Single-City PM10-2.s Morbidity Studies*
Study
NYS DOH, 2006
Peel. 2005
Tolbert, 2007
Slaughter. 2005
!to, 2003
Sheppard, 2003
Location
Bronx
Manhattan
Atlanta
Atlanta
Spokane
Delroit
Seattle
Morbidity Endpoint
Asthma HA
UR1 ED visits
Pneumonia ED visits
COPD ED visits
Asthma ED visits
RD ED visits
GHF ED visits
RD ED visits
CVD ED visits
COPD
Asthma
RD
Rl HA
IHD HA
CHF HA
Asthma HA
Time Period
1999-2000
1993-2000
1993-2004
1995-2001
1992-1984
1937-1994
Age
All
All
All
65+
All
All
All
65+
All
All
65+
All
98<" PM10 (ng/m3)
49
52
67
71
81
102
105
Effect Estimate (35% Clj
—
...
•*-
•*-
—
0.8 0.9 0 1.10 1.20
These studies are a combination of those assessed in the last review and those assessed in the ISA in the current review. Studies are ordered by increasing 98th
percentile PM10 concentrations. See 2005 Staff Paper (US EPA, 2005, pp. 5-63 to 5-66) describing measurement uncertainties associated with the reported PM10
levels in Ito (2003).
3-38
-------
,th .
In addition to the single-city study locations in Figures 3-5 and 3-6, 98 percentile PM
concentrations averaged across the study locations evaluated in the multi-city studies by
10
3
Zanobetti and Schwartz (2009) and Peng et al. (2008) were 77 ng/m (Jenkins, 2011) and 68
Hg/m3 (Schmidt and Jenkins, 2010), respectively. Multi-city effect estimates remained positive,
and in some cases (i.e., Zanobetti and Schwartz for all-cause and respiratory mortality)
statistically significant, in co-pollutant models that included fine particles.
Bayes-adjusted single-city effect estimates for the 47 cities evaluated by Zanobetti and
Schwartz (US EPA, 2009a, Figure 6-29), which were generally positive but statistically
significant in only six cities, can provide some additional insight into the PMio concentrations in
specific locations where associations between PMi0-2.5 and mortality have been reported. The
98th percentile PMio concentrations in these 47 cities ranged from 39 |J,g/m3 (Davie, FL) to 187
Hg/m3 (Phoenix, AZ), and in the 6 cities where positive and statistically significant PMi0-2.5
mortality effect estimates were reported, 98th percentile PMio concentrations were as follows
(Schmidt and Jenkins, 2010):
• Chicago: 91 ng/m3
• Salt Lake City: 98 ng/m3
• Detroit: 105 ng/m3
• Pittsburgh: 112 ng/m3
• Birmingham: 122 |J,g/m3
• St. Louis: 138 ng/m3
Thus, in the single-city mortality studies in Figure 3-5 above, as well as the Bayes-adjusted
single-city analyses of the locations evaluated by Zanobetti and Schwartz, positive and
statistically significant PMio-2.5 effect estimates were reported in some locations with 98th
percentile PMio concentrations ranging from 200 ng/m3 to 91 ng/m3 (i.e., locations evaluated by
Mar et al., 2003; Ostro et al., 2003; Wilson et al., 2007; and the cities listed above, US EPA
2009a, Figure 6-29). Among the U.S. morbidity studies, Ito (2003) reported a positive and
statistically significant PMio-2.5 effect estimate for hospital admissions for ischemic heart disease
in Detroit, where the 98th percentile PMio concentration (102 |J,g/m3) was also within this range.
PMio-2.5 effect estimates in this study remained positive, and in some cases statistically
significant, in co-pollutant models with gaseous pollutants (US EPA, 2009a, Figures 6-5 and 6-
15). Other morbidity studies generally did not report statistically significant PMio-2.5 effect
estimates.
3-39
-------
Air Quality-based Considerations
In addition to the evidence-based considerations described above, we have estimated the
level of a 24-hour PMi0 standard with a 98th percentile form that would approximate the degree
of protection, on average across the country, provided by the current 24-hour PMio standard with
its one-expected-exceedance form. Our initial approach to estimating this "generally equivalent"
98th percentile PMio concentration was to use EPA's AQS as the basis for regressing 98th
percentile PMio concentrations onto one-expected-exceedance concentration equivalent design
values (Schmidt and Jenkins, 2010). Based on this approach, and using monitoring data from
1988 to 2008, a 98th percentile PMio concentration of 87 ng/m3 is, on average, generally
equivalent to the current standard level (Figure 3-7 below and Schmidt and Jenkins, 2010).
However, as indicated in Figure 3-7, the range of equivalent concentrations varies considerably
across monitoring sites and over the time period evaluated (95% confidence interval ranges from
63 to 111 ng/m3) (Schmidt and Jenkins, 2010). As a consequence, we note that in some
locations a 98th percentile standard with a level of 87 ng/rn3 would likely be more protective than
the current standard while in other locations it would likely be less protective than the current
standard.
Regional differences in the relationship between 98th percentile PMio concentrations and
one-expected-exceedance concentration equivalent design values are illustrated in Figure 3-8,
based on air quality data from 1988 to 2008. The 98th percentile PMio concentrations that are, on
average, generally equivalent to the current standard level ranged from just below 87 ng/m3 in
the southeast, southwest, upper Midwest, and outlying areas (i.e., generally equivalent 98th
percentile PMio concentrations ranged from 82 to 85 |J,g/m3 in these regions) to just above 87
Hg/m3 in the northeast, industrial Midwest, and southern California (i.e., generally equivalent
98th percentile PMio concentrations ranged from 88 to 93 ng/m3 in these regions) (Schmidt,
201 Ib). However, even within these regions there is considerable variability in the "generally
equivalent" 98th percentile PMio concentration across monitoring sites (Figure 3-8).
3-40
-------
Figure 3-7. Composite 3-year PMi0 98 Percentile 24-Hour Average concentration versus
the PMio Expected Exceedance Concentration-equivalent Design Value (1988-2008)
t-1
u
H>
SB
Current PM10NAAQS
0 50 100 150 200 250 300
PMKI expected exeeedance eoncentration-equi\'alent design value
3-41
-------
Figure 3-8. Regional 3-year PMi0 98 Percentile 24-Hour Average Concentrations Versus
the PMio Expected Exceedance Concentration-equivalent Design Values (1988-2008)*
Industrial
Mid-west
SO 1 DO 150 K» 250 300
) •:
•-,
Iff
we
«
. ..,,
at
m
90
'•
R
H
K
H
I
Northwest
xx
•
v" s
Jf1. MF «i ••
s^
m
Southwest
•/• t^r
«^ ^
:-L
i j
at
Southern
CA
Outlying
(AK, HI,
VI, PR)
*Bold regression lines reflect region-specific relationships between 24-hour average 98th percentile PM10
concentrations and PM10 expected exceedance concentration-equivalent design values. Non-bolded regression lines
reflect the overall relationship across all regions, as illustrated above in Figure 3-7. In some areas (i.e., southeast,
upper Midwest, northwest, southwest, and southern California), the two regression lines are almost
indistinguishable. Vertical lines mark the level of the current standard and horizontal lines mark a 98th percentile
concentration of 87 |J.g/m3.
3-42
-------
Given the spatial and temporal variability in the relationship between one-expected-
exceedance concentration-equivalent design values and 98th percentile PMio concentrations, no
single 98th percentile PMio standard level would provide public health protection equivalent to
that provided by the current standard, consistently over time and across locations.18 Therefore,
to provide a broader perspective on the relationship between the current standard and potential
98th percentile standards, we have also compared the size of the populations living in counties
with PMio one-expected-exceedance concentration-equivalent design values greater than the
current standard level to the size of the populations living in counties with 98th percentile PMio
concentrations above different potential alternative standard levels (based on air quality data
from 2007 to 2009). Such comparisons can be considered as surrogates for comparisons of the
breadth of public health protection provided by the current and potential alternative standards, as
discussed below. The results are presented in Table 3-2 below.19
Based on comparisons of total population counts across all regions of the U.S. for the
years 2007 to 2009, a 98th percentile PMio standard with a level between 75 and 80 ng/rn3 would
be most closely equivalent to the current standard. However, as with the regression analysis
described above, there is considerable variability across locations in the "generally equivalent"
98th percentile PMio concentration (see Table 3-2).
18The "generally equivalent" concentration also differs depending on the years of monitoring data used. For
example, when this analysis was restricted to only the most recent years available (i.e., 2007 to 2009), the "generally
equivalent" 98th percentile PM10 concentration was 78 |J.g/m3. Given the temporal variability in the relationship
between the current standard level and 98th percentile PM10 concentrations, and the potential for the "generally
equivalent" 98th percentile concentration to vary year-to-year, staff concluded that it remains appropriate to consider
the regression analyses that use the broader range of available monitoring years (i.e., 1998-2008), as these analyses
are likely to be more robust than analyses based on a shorter period of time.
19Table 3-2 presents counts of counties that would not meet the current and potential alternative standards (see also
Appendix D). These county counts reflect the net number of counties in each region. However, some counties that
meet the current standard would violate one or more of the alternative standards while some counties that violate the
current standard would meet one or more of the alternatives. Therefore, the net county counts presented in Table 3-
2 do not reflect the total number of counties that could potentially change attainment status if the PM10 standard
were to be revised. The specific counties that would violate the current standard and/or potential alternative
standards with 98th percentile forms are listed in the memo by Schmidt (201 Ib).
-------
Table 3-2. Predicted Counts of Counties, and Population (x 1,000) within those Counties, Not Likely to Meet the Current
Standard and Potential Alternative PMio Standards with 98th Percentile Forms (based on air quality in 2007-2009)20
Region ;*
Total 3 of counties >
Total population >
Current Standard
> 87 flg/Hl3
> 35 fig/in*
> 80 ng/m3
> 75 ng/ni3
> 70 fig/in3
3 -year average 98th perc entile
> 65 fig/m3
?f comities
Population
(thousands)
JF coimties
population
£ counties
population
•F counties
population
# comities
population
7f counties
population
fF counties
population
All U.S.
307
120,090
41
32,835
37
20,515
39
21,887
45
24,535
55
35,703
71
43,823
87
49,394
Northeast
37
15,397
0
0
0
0
0
0
0
0
0
0
0
0
2
775
Southeast
57
27,181
3
4,626
7
4,063
2
4,063
2
4,063
3
4,626
4
4,644
4
4,644
Iudustri.il
Michvest
50
21,352
0
0
2
507
3
1,789
4
3,183
6
3,491
7
8,868
9
10,421
Upper
Midwest
40
5,917
1
14
2
552
T
552
3
599
5
637
7
881
10
1,029
Southivest
25
11,112
11
5,48S
11
5,924
12
6.014
13
6,131
13
6,131
13
6,131
14
7,507
Noi thive st
77
15,270
13
1,878
10
1,789
10
1,789
11
1,833
15
2,570
27
5,052
33
5,989
Southern
California
18
22,695
12
20,571
9
7,421
9
7,421
11
8,467
12
17,986
12
17,986
13
18,739
Outlying
Areas
3
1,167
1
260
1
260
1
260
1
260
1
260
1
260
2
290
20The information in Table 3-2 (and presented in more detail in Schmidt, 201 Ib and Appendix D) is based on monitoring data from the years 2007 to 2009. This
is an update to the information presented in the memo from Schmidt and Jenkins (2010), and presented in draft versions of this Policy Assessment, which was
based on the years 2006 to 2008.
3-44
-------
One consequence of this variability, as noted above in the discussion of the form of the
standard, would be that a 24-hour PMio standard with a 98th percentile form and a revised level
would likely target public health protection to some different locations than does the current
standard. Therefore, in further considering the appropriateness of revising the form and level of
the current PMio standard, we have considered the following question:
• To what extent, when compared with the current standard, would a revised
standard be expected to target public health protection to areas where we have more
confidence that PMi0-2.s is associated with adverse health effects?
To address this question, we considered the potential impact of revising the form and level of
the PMio standard in locations where health studies have reported associations with PMi0.2.5.
We initially considered U.S. study locations that would likely have met the current PMio
standard during the study period and where positive and statistically significant associations with
PMio-2.5 were reported (PMio air quality concentrations in study locations are presented above
and in Schmidt and Jenkins, 2010; Jenkins, 201 1). Only Birmingham, Chicago, Pittsburgh, and
Detroit21 met these criteria. As shown in Table 3-3, these areas where positive and statistically
significant associations with PMio-2.5 were reported would likely have met the current PMio
standard during the study periods. However, none of these areas would likely have met a 98th
percentile 24-hour PMio standard with a level at or below 87
Table 3-3. PMio Concentrations in Locations that Met Current PMio Standard and where
Positive and Statistically Significant Associations with PMio-2.5 have been Reported
Study Location
Birmingham22
Chicago
Pittsburgh
Detroit
PMio Concentration-
Equivalent Design Value
154 ng/m3
113 ng/m3
139 ng/m3
123 ng/m3
PMio 98th Percentile
122 ng/m3
91 ng/m3
112ng/m3
102 |ig/m3
We next broadened our consideration of study locations to include U.S. locations where
health studies have reported positive, though not necessarily statistically significant, associations
between PMio-2.5 and mortality or morbidity. Such positive associations were reported in 47
21Positive and statistically significant PM10.2.5 effect estimates for Birmingham, Chicago, and Pittsburgh are reported
in the ISA (U.S. EPA, 2009a, figure 6-29; from cities evaluated by Zanobetti and Schwartz, 2009). Effect estimates
for Detroit are reported by Ito et al. (2003)
22According to rounding convention for the PM10 standard, a 24-hour PM10 concentration of 154 |J.g/m3 would round
to 150 |J.g/m3 (71 FR 61144). Therefore, based on the PM10 one-expected-exceedance concentration-equivalent
design value, Birmingham would have been expected to just meet the current PM10 standard during the study period.
3-45
-------
23
locations that would likely have met the current PMio standard during the study period. Of
,th
these 47 locations, 13 would likely not have met a 98 percentile 24-hour PMio standard with a
level at 87 |J,g/m3, 20 would likely not have met a 98th percentile 24-hour PMio standard with a
level of 75 ng/m3, and 31 would likely not have met a 98th percentile 24-hour PMio standard with
a level of 65 ng/m3 (Schmidt and Jenkins, 2010; Jenkins, 2011; Figures 3-2, 3-3, 3-5, 3-6 above).
In addition to the above analyses, we have also considered locations where health studies
reported positive associations with PMio-2.5 and where ambient PMio concentrations were likely
to have exceeded those allowed under the current PMio standard during the study period. Nine
locations met these criteria.24 Of these locations, all would also likely have exceeded a 98th
percentile PMio standard with a level at or below 87 |J,g/m3 (Schmidt and Jenkins, 2010; Jenkins,
2011; Figures 3-2, 3-3, 3-5, 3-6 above).
Therefore, among U.S. study locations where PMi0-2.5-associated health effects have been
reported, some areas that met the current standard would likely not have met a 98th percentile
PMio standard with a level at or below 87 ng/rn3. In contrast, of the locations that did not meet
the current standard during the study period, none would likely have met a 98th percentile PMio
standard with a level at or below 87 ng/rn3. Given this, we conclude that, compared to the
current PMio standard, a 24-hour PMio standard with a 98th percentile form could potentially
better target public health protection to locations where we have more confidence that ambient
PMio-2.5 concentrations are associated with mortality and/or morbidity.
Integration of evidence- and air quality-based considerations
In considering the evidence and air quality information within the context of identifying
potential alternative standard levels for consideration, we first note the following:
• Linear regression analysis suggests that a 98th percentile 24-hour PMio concentration as
high as 87 |J,g/m3 could be considered "generally equivalent" to the current PMio
standard, over time and across the country.
• A 98th percentile 24-hour PMio standard with a level at or below 87 ng/rn3 would be
expected to maintain PMio and PMio-2.5 concentrations below those present in U.S.
locations where single-city studies have reported PMio-2.5 effect estimates that are
positive and statistically significant (lowest concentration in such a location was 91
|j,g/m3). While some single-city studies have reported positive PMio-2.5 effect estimates in
locations with 98th percentile PMio concentrations below 87 ng/m3, these effect estimates
were not statistically significant.
23Philadelphia (Lipfert et al., 2000), Detroit (Ito et al, 2003), Santa Clara (CA) (Fairley et al., 2003), Seattle
(Sheppard et al., 2003), Atlanta (Klemm et al., 2004), Spokane (Slaughter et al., 2005), Bronx and Manhattan (NYS
DOH, 2006), and 39 of the cities evaluated by Zanobetti and Schwartz (2009) (US EPA, 2009a, Figure 6-29)
24Pittsburgh (Chock et al., 2000), Coachella Valley (CA) (Ostro et al., 2003), Phoenix (Mar et al., 2007; Wilson et
al., 2007), and 6 of the cities evaluated by Zanobetti and Schwartz (2009) (US EPA, 2009a, Figure 6-29).
3-46
-------
• Multi-city average 98th percentile PMio concentrations were below 87 ng/m3 for U.S.
multi-city studies, which have reported positive and statistically significant PMio-2.5 effect
estimates. However, the extent to which effects reported in multi-city studies are
associated with the short-term air quality in any particular location is highly uncertain,
especially when considering short-term concentrations at the upper end of the distribution
of daily PMio concentrations.
• Epidemiological studies have reported positive, and in a few instances statistically
significant, associations with PMio-2.5 in some locations likely to have met the current
PMio standard but not a PMio standard with a 98th percentile form and a level at or below
87 ng/m3.
To the extent the above considerations are emphasized, we note that a standard level as
high as about 85 ng/m3, for a 24-hour PMio standard with a 98th percentile form, could be
supported. Such a standard level would be expected to maintain PMio and PMio-2.5
concentrations below those present in U.S. locations of single-city studies where PMio-2.5 effect
estimates have been reported to be positive and statistically significant and below those present
in some U.S. locations where single-city studies reported PMio-2.5 effect estimates that were
positive, but not statistically significant. These include some locations likely to have met the
current PMio standard during the study periods. Given this, when compared to the current
standard, a 24-hour PMio standard with a 98th percentile form and a level at or below 85 |J,g/m3
could have the effect of focusing public health protection on locations where we have more
confidence that PMio-2.5 is associated with mortality and/or morbidity.
Given the above, we conclude that a 98th percentile standard with a level as high as 85
Hg/m3 could be considered to the extent that more weight is placed on the appropriateness of
focusing public health protection in areas where positive and statistically significant associations
with PMio-2.5 have been reported, and to the extent less weight is placed on PMio-2.5 effect
estimates that are not statistically significant and/or that reflect estimates across multiple cities.
It could be judged appropriate to place less weight on PMio-2.5 effect estimates that are not
statistically significant given the relatively large amount of uncertainty that is associated with the
broader body of PMio-2.5 health evidence, including uncertainty in the extent to which health
effects evaluated in epidemiological studies result from exposures to PMio-2.5 itself, rather than
one or more co-occurring pollutants. This uncertainty, as well as other uncertainties discussed in
section 3.2.1 above, are reflected in the ISA conclusions that the evidence is "suggestive" of a
causal relationship (i.e., rather than "causal" or "likely causal") between short-term PMio-2.5 and
mortality, respiratory effects, and cardiovascular effects. In addition, it could be judged
appropriate to place less weight on 98th percentile PMio concentrations averaged across multiple
3-47
-------
cities, given the uncertainty in linking multi-city effect estimates with the air quality in any
particular location.
However, we also note that, overall across the U.S. based on recent air quality
information (i.e., 2007-2009), fewer people live in counties with 98th percentile 24-hour PMio
concentrations above 85 |J,g/m3 than in counties likely to exceed the current PMio standard.
These results could be interpreted to suggest that a 98th percentile standard with a level of 85
Hg/m3 might decrease overall public health protection compared to the current standard. Based
on this analysis, a 98th percentile 24-hour PMio standard with a level between 75 and 80 |J,g/m3
would provide a level of public health protection that is generally equivalent, on average across
the U.S., to that provided by the current standard. To the extent these population counts are
emphasized in comparing the public health protection provided by the current and potential
alternative standards, and to the extent it is judged appropriate to set a standard that provides at
least the level of public health protection that is provided by the current standard, it would be
appropriate to consider standard levels in the range of approximately 75 to 80 ng/m3.
Alternative approaches to considering the evidence could also lead to consideration of
standard levels below 75 ng/m3. For example, a number of single-city epidemiological studies
have reported positive, though not statistically significant, PMio-2.5 effect estimates in locations
with 98th percentile PMio concentrations below 75 |J,g/m3. Given that exposure error is
particularly important for PMio-2.5 epidemiological studies and can bias the results of these
studies toward the null hypothesis, it could be judged appropriate to place more weight on
positive associations reported in these epidemiological studies, even when those associations are
not statistically significant. In addition, the multi-city averages of 98th percentile PMio
concentrations in the locations evaluated by Zanobetti and Schwartz (2009) and Peng et al.
(2008) were 77 and 68 ng/m3, respectively. Both of these multi-city studies reported positive
and statistically significant PMio-2.5 effect estimates that remained positive in co-pollutant models
that included PM2.5, though only Zanobetti and Schwartz (2009) reported PMio-2.5 effect
estimates that remained statistically significant in such co-pollutant models. Despite
uncertainties in the extent to which effects reported in these multi-city studies are associated with
the short-term air quality in any particular location, emphasis could be placed on these multi-city
associations. We conclude that, to the extent more weight is placed on single-city studies
reporting positive, but not statistically significant, PMio-2.5 effect estimates and on multi-city
studies, it could be appropriate to consider standard levels as low as 65 ng/rn3. A standard level
of 65 ng/rn3 would be expected to provide a substantial margin of safety against health effects
that have been associated with PMio-2.5 and, as discussed above, could better focus (compared to
3-48
-------
the current standard) public health protection on areas where health studies have reported
associations with PMio-2.5.
In considering potential alternative standard levels below 65 |J,g/m3, we note that, as
discussed above, the overall body of PMio-2.5 health evidence is relatively uncertain, with
somewhat stronger support in U.S. studies for associations with PMio-2.5 in locations with 98th
percentile PMio concentrations above 85 ng/m3 than in locations with 98th percentile PMio
concentrations below 65 ng/m3. Specifically, we note the following:
• Epidemiological studies, either single-city or multi-city, have not reported positive and
statistically significant PMio-2.5 effect estimates in locations with 98th percentile PMio
concentrations (multi-city average 98th percentile concentrations in the case of multi-city
studies) at or below 65 ng/m3.
• Although some single-city morbidity studies have reported positive, but not statistically
significant, associations with PMio-2.5 in locations with 98th percentile PMio
concentrations below 65 |J,g/m3, the results of U.S. morbidity studies were generally less
consistent than those of mortality studies, with some PMio-2.5 effect estimates being
positive while others were negative (i.e., negative effect estimates were reported in
several studies conducted in Atlanta, where the 98th percentile PMio concentrations
ranged from 67 |J,g/m3 to 71 |j,g/m3).
• Although Bayes-adjusted single-city PMio-2.5 effect estimates were positive, but not
statistically significant, in some locations with PMio concentrations below 65 ng/m3,
these effect estimates were based on the difference between community-wide PMio and
PM2 5 concentrations. As discussed above, it is not clear how these estimates of PMio-2.5
concentrations compare to those more typically used in other studies to calculate PMio-2.5
effect estimates. At present, few corroborating studies are available that use other
approaches (i.e., co-located monitors, dichotomous samplers) to estimate/measure
PMio-2.5 in locations with 98th percentile PMio concentrations below 65 |J,g/m3.
In light of these limitations in the evidence for a relationship between PMio-2.5 and
adverse health effects in locations with relatively low PMio concentrations, along with the
overall uncertainties in the body of PMio-2.5 health evidence as described above and in the ISA,
we conclude that while it could be judged appropriate to consider standard levels as low as 65
Hg/m3, it is not appropriate, based on the currently available body of evidence, to consider
standard levels below 65 ng/m3.
In its review of the draft Policy Assessment, CAS AC concluded that "alternative standard
levels of 85 and 65 ng/m3 (based on consideration of 98th percentile PMio concentration) could
be justified" (Samet, 2010d). However, in considering the evidence and uncertainties, CASAC
recommended a standard level from the lower part of the range, recommending a level
"somewhere in the range of 75 - 65 |j,g/m3" (Samet, 2010d). In making this recommendation,
3-49
-------
CAS AC noted that the number of people living in counties with air quality not meeting the
current standard is approximately equal to the number living in counties that would not meet a
98th percentile standard with a level between 75 and 80 ng/rn3. CASAC used this information as
the basis for their conclusion that a 98th percentile standard between 75 and 80 |J,g/m3 would be
"comparable to the current standard" (Samet, 2010d).
3.4 SUMMARY OF STAFF CONCLUSIONS ON PRIMARY THORACIC COARSE
PARTICLE STANDARD
In reaching conclusions on the adequacy of the current PMi0 standard and potential
alternative standards to provide requisite protection for health effects associated with short-term
exposures to thoracic coarse particles, staff has considered the basic elements of the NAAQS:
indicator, averaging time, form, and level (section 3.3.1 to 3.3.4 above). In considering available
scientific evidence and air quality information, we reflect upon the evidence and information
available in the last review integrated with evidence and information that has become available
since that review as assessed and presented in the ISA (US EPA, 2009a) and summarized above
in sections 3.2 and 3.3.
We recognize that selecting from among potential alternative standards will necessarily
reflect consideration of the evidence as well as the uncertainties inherent in that evidence. In
considering the current PMio standard and identifying potential alternative primary standards for
consideration, we are mindful that the Clean Air Act requires standards to be set that are
requisite to protect public health with an adequate margin of safety, such that the standards are to
be neither more nor less stringent than necessary. Thus, the Act does not require that the
NAAQS be set at zero-risk levels, but rather at levels that avoid unacceptable risks to public
health.
Based on the currently available scientific evidence, staff reaches the following
conclusions regarding the primary PMio standard:
(1) It would be appropriate to consider either retaining or revising the current 24-hour
primary standard, depending on the relative weight placed on the evidence supporting
associations with PMio-2.5, the uncertainties associated with this evidence, and the ISA
conclusions that the evidence is only "suggestive" of a causal relationship (i.e., rather than
"causal" or "likely causal") between short-term PMio-2.5 and mortality, respiratory effects,
and cardiovascular effects.
(2) It is appropriate to retain PMio as the indicator for thoracic coarse particles. This
conclusion is based on our assessment of the evidence for effects related to particles of
urban and non-urban origins. We also conclude that research should be targeted so as to
inform consideration of different indicators in future reviews.
(3) It is appropriate to retain a 24-hour averaging time for a PMio standard meant to protect
against short-term exposures to thoracic coarse particles. This conclusion reflects the body
3-50
-------
of epidemiological studies, which are most often based on 24-hour average PMio-2.5
concentrations.
(4) To the extent consideration is given to revising the current standard:
(a) Consideration should be given to a 98th percentile form for a 24-hour PMi0
standard. This conclusion is based on consideration of providing a balance between
limiting peak concentrations and providing a stable regulatory target, compensating
for differences in monitoring frequency across locations, focusing the standard on
days when PMi0-2.5 is likely to make a relatively larger contribution to PMi0 mass,
and focusing public health protection to areas where we have greater confidence
that mortality and morbidity are associated with ambient PMio-2.5.
(b) In conjunction with considering a 98th percentile form, it is appropriate to consider
PMio standard levels in the range of 85 |J,g/m3 down to about 65 |J,g/m3. This range
of levels is based on consideration of 98th percentile PMio concentrations in U.S.
study locations where PMi0-2.5 epidemiological studies have been conducted and of
the relationship between the current one-expected-exceedance form and a potential
alternative 98l percentile form. Staff concludes that standard levels in the upper
part of this range are supported by the strongest evidence and would reflect greater
emphasis on positive and statistically significant PMio-2.5 effect estimates reported
in single-city studies. Standard levels in the lower part of this range would reflect
greater emphasis on PMio-2.5 effect estimates that are positive, though not
necessarily statistically significant, and on multi-city effect estimates.
3.5 KEY UNCERTAINTIES AND AREAS FOR FUTURE RESEARCH AND DATA
COLLECTION
As discussed above (see section 3.2), a number of key uncertainties and limitations in the
health evidence have been considered in this review. These include uncertainties and limitations
in the air quality estimates used in PMio-2.5 epidemiological studies; in the extent to which PMio-
2.5 air quality concentrations reflect exposures to PMio^.s; in the extent to which PMio-2.5 itself is
responsible for health effects reported in epidemiological studies; and in the extent to which the
chemical and/or biological composition of PMio-2.5 affects particle toxicity. In this section, we
highlight areas for future health-related research, model development, 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 PM NAAQS
reviews and, in particular, consideration of possible alternative indicators, averaging times,
forms, and/or levels. In some cases, research in these areas can go beyond aiding standard
setting to informing the development of more efficient and effective control strategies.
As an initial matter, we note that many of the research needs identified for fine particles
(see above, section 2.5) are also relevant for thoracic coarse particles. This includes research in
the following areas:
3-51
-------
• Sources and components of coarse particles, including source apportionment modeling;
monitoring of components; linking specific sources/components to health outcomes;
linking sources/components to intra- and inter-city differences in health effects; linking
sources/components to population exposures; and evaluating different size cut-points
• Understanding the extent to which an association between thoracic coarse particles and
specific health effects can be modified by co-pollutants
• Understanding associations with a broad range of cardiovascular and respiratory
endpoints as well as adverse effects in the nervous system, on reproduction, and/or on
development
• Understanding C-R relationships and the confidence bounds around these relationships,
especially at lower ambient thoracic coarse particle concentrations
• Understanding air quality distributions in locations of epidemiological studies
• Identifying populations susceptible to PMio-2.5-related health effects
• Modeling to estimate PMio-2.5 mass and composition in areas with less-than-daily
monitoring
These uncertainties and areas for future research are discussed above in section 2.5 and
that discussion will not be repeated here. In addition to the above, there are several areas for
future research that are particularly relevant for thoracic coarse particles. These include the
following:
• The body of experimental inhalation studies (e.g., controlled human exposure and animal
toxicology studies) is currently relatively sparse. Additional well-conducted experimental
studies could play an important role in weight of evidence judgments in future ISAs.
Therefore, experimental evaluation of effects (e.g., vasomotor function, airways
responsiveness, pulmonary function/inflammation) of concentrated ambient PMio-2.5 from
specific sources (e.g., traffic, industrial, non-industrial) would be useful, particularly if
exposure-response relationships are evaluated.
• Exposure error is of particular concern for thoracic coarse particles, given the relative lack of
monitoring and its less homogeneous atmospheric distribution compared to fine particles (US
EPA, 2009a, section 2.1.1.2). Therefore, short-term studies with well-characterized personal
exposures to PMio-2.5 (e.g., panel studies) would be useful. Such studies could examine
indicators of cardiovascular and respiratory morbidity (e.g., arrhythmia, ischemia, vasomotor
function, respiratory symptoms, pulmonary inflammation, pulmonary function, pulmonary
injury) and would be particularly useful if they evaluated concentration-response
relationships and/or effects of repeated peak exposures.
• Epidemiological studies currently use a variety of approaches to measure/estimate PMio-2.5
concentrations. It is important that we better understand the relationship between results
from studies that estimate PMio-2.5 concentrations using either (1) difference method of co-
located monitors, (2) difference method of county-wide averages of PMio and PM2.5, or (3)
direct measurement of PMio-2.5 using a dichotomous sampler. In addition, as described
3-52
-------
above, PMio-2.s monitoring will be required at NCORE sites by 2011. It could be useful for
future epidemiological studies to make use of these new PMi0-2.5 monitoring sites.
• Very little information is available to inform weight of evidence conclusions for endpoints
associated with long-term PMio-2.5 exposures. Epidemiological and animal toxicological
studies of long-term exposures (i.e., months to years) to PMi0-2.5 would be helpful, though
limitations in the extent to which coarse particles penetrate rodent respiratory systems could
add uncertainty to the interpretation of rodent inhalation studies. Long-term studies could
evaluate links with cardiovascular and respiratory morbidity, reproductive and developmental
outcomes, cancer, and mortality.
-------
3.6 REFERENCES
Abt Associates (2005). Paniculate Matter Health Risk Assessment for Selected Urban Areas. Final Report.
Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Contract No. 68-D-03-002. EPA 452/R-05-007A. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/PMrisk20051220.pdf
Bell ML; Ebisu K; Peng RD; Walker J; Samet JM; Zeger SL; 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:
1301-1310.
Bennett CM; McKendry IG; Kelly S; Denike K; Koch T (2006). Impact of the 1998 Gobi dust event on hospital
admissions in the Lower Fraser Valley, British Columbia. Sci Total Environ, 366: 918-925.
Burnett RT; Smith-Doiron M; Stieb D; Cakmak S; Brook JR. (1999). Effects of paniculate and gaseous air pollution
on cardiorespiratory hospitalizations. Arch Environ Occup Health, 54: 130-139.
Burnett RT; Goldberg MS. (2003). Size-fractionated paniculate mass and daily mortality in eight Canadian cities.
Health Effects Institute. Boston, MA
Burnett RT; Stieb D; Brook JR; Cakmak S; Dales R; Raizenne M; Vincent R; Dann T (2004). Associations between
short-term changes in nitrogen dioxide and mortality in Canadian cities. Arch Environ Occup Health, 59:
228-236.
Chan CC; Chuang KJ; Chen WJ; Chang WT; Lee CT; Peng CM (2008). Increasing cardiopulmonary emergency
visits by long-range transported Asian dust storms in Taiwan. Environ Res, 106: 393-400.
Chen L; Yang W; Jennison BL; Omaye ST (2000). Air paniculate pollution and hospital admissions for chronic
obstructive pulmonary disease in Reno, Nevada. Inhal Toxicol, 12: 281-298.
Chen Y; Yang Q; Krewski D; Shi Y; Burnett RT; McGrail K. (2004). Influence of relatively low level of particulate air pollution
on hospitalization for COPD in elderly people. Inhal Toxicol, 16: 21-25.
Chen Y; Yang Q; Krewski D; Burnett RT; Shi Y; McGrail KM (2005). The effect of coarse ambient particulate
matter on first, second, and overall hospital admissions for respiratory disease among the elderly. Inhal
Toxicol, 17: 649-655.
Chock DP; Winkler SL; Chen C. (2000). A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J Air Waste Manag Assoc, 50: 1481-1500.
Choudhury, AH; Gordian ME; Morris SS. (1997). Associations between respiratory illness and PM10 air pollution.
Arch. Environ. Health 52: 113-117.
Davidson, JE; Hopke PK. (1984). Implications of Incomplete Sampling on a Statistical Form of the Ambient Air
Quality Standard for Particulate Matter. Environ. Sci. Technol., 18: 571-580.
Edgerton, ES; Casucclo GS; Saylor RD; Lersch TL; Hartsell BE; Jansen JJ; Hansen DA. (2009). Measurements of
OC and EC in coarse particulate matter in the southeastern United States. J. Air Waste Manag. Assoc., 59:
78-90.
Fairley D (2003). Mortality and air pollution for Santa Clara County, California, 1989-1996, In: Revised analyses of
time-series studies of air pollution and health. Special report. Health Effects Institute. Boston,
MA.http://www.healtheffects.org/Pubs/TimeSeries.pdf.
Fung KY; Khan S; Krewski D; Chen Y. (2006). Association between air pollution and multiple respiratory
hospitalizations among the elderly in Vancouver, Canada. Inhal Toxicol, 18: 1005-1011.
3-54
-------
Gong H; Linn WS; Terrell SL; Clark KW; Geller MD; Anderson KR; Cascio WE; Sioutas C. (2004). Altered heart-
rate variability in asthmatic and healthy volunteers exposed to concentrated ambient coarse particles. Inhal
Toxicol, 16: 335-343.
Gordian ME; Ozkaynak H; Xue J; Morris SS; Spengler JD (1996). Paniculate air pollution and respiratory disease in
Anchorage, Alaska. Environ Health Perspect, 104: 290-297.
Graff D; Cascio W; Rappold A; Zhou H; Huang Y; Devlin R. (2009). Exposure to concentrated coarse air pollution
particles causes mild cardiopulmonary effects in healthy young adults . Environ Health Perspect, 117:
1089-1094.
Hefflin BJ; Jalaludin B; McClure E; Cobb N; Johnson CA; Jecha L; Etzel RA (1994). Surveillance for dust storms
and respiratory diseases in Washington State, 1991. Arch Environ Occup Health, 49: 170-174.
Henderson, R. (2005b). Clean Air Scientific Advisory Committee (CASAC) Review of the EPA Staff
Recommendations Concerning a Potential Thoracic Coarse PM Standard in the Review of the National
Ambient Air Quality Standards for Particulate Matter: Policy Assessment of Scientific and Technical
Information (Final PM OAQPS Staff Paper, EPA-452/R-05-005). September 15, 2005. Available:
http://www.epa.gov/sab/panels/casacpmpanel.html.
Host S; Larrieu S; Pascal L; Blanchard M; Declercq C; Fabre P; Jusot JF; Chardon B; Le Tertre A; Wagner V;
Prouvost H; Lefranc A (2007). Short-term Associations between Fine and Coarse Particles and
Cardiorespiratory Hospitalizations in Six French Cities. Occup Environ Med, 18: S107-S108.
Host S; Larrieu S; Pascal L; Blanchard M; Declercq C; Fabre P; Jusot JF; Chardon B; Le Tertre A; Wagner V;
Prouvost H; Lefranc A (2008). Short-term associations between fine and coarse particles and hospital
admissions for cardiorespiratory diseases in six French cities. Occup Environ Med, 65: 544-551.
Ito K (2003). Associations of paniculate matter components with daily mortality and morbidity in Detroit,
Michigan, In: Revised analyses of time-series studies of air pollution and health. Special report. Health
Effects Institute. Boston, MA. R828112. http: //www.healtheffects.org/Pubs/TimeSeries.pdf.
Jenkins S.M. (2011). Supplemental Analysis of PM10 Air Quality from Locations Evaluated by Zanobetti and
Schwartz (2009). Memorandum to PM NAAQS review docket EPA-HQ-OAR-2007-0492.
Klemm RJ; Mason R (2003). Replication of reanalysis of Harvard Six-City mortality study. In HEI Special Report:
Revised Analyses of Time-Series Studies of Air Pollution and Health, Part II (pp. 165-172). Boston, MA:
Health Effects Institute.
Klemm RJ; Lipfert FW; Wyzga RE; Gust C (2004). Daily mortality and air pollution in Atlanta: two years of data
from ARIES. Inhal Toxicol, 16 Suppl 1: 131-141.
Le Tertre A; Schwartz J; Touloumi G. (2005). Empirical Bayes and adjusted estimates approach to estimating the
relation of mortality to eposure of PM10. Risk Anal, 25: 711-718.
Lin M; Chen Y; Burnett RT; Villeneuve PJ; Krewski D (2002). The influence of ambient coarse paniculate matter
on asthma hospitalization in children: case-crossover and time-series analyses. Environ Health Perspect,
110:575-581.
Lin M; Stieb DM; Chen Y. (2005). Coarse paniculate matter and hospitalization for respiratory infections in
children younger than 15 years in Toronto: a case-crossover analysis. Pediatrics, 116: 235-240.
Lipfert FW; Morris SC; Wyzga RE. (2000). Daily mortality in the Philadelphia metropolitan area and size-classified
paniculate matter. J Air Waste Manag Assoc, 50: 1501-1513.
3-55
-------
Lisabeth LD; Escobar JD; Dvonch JT; Sanchez BN; Majersik JJ; Brown DL; Smith MA; Morgenstern LB. (2008).
Ambient air pollution and risk for ischemic stroke and transient ischemic attack. Ann Neurol, 64: 53-59.
Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003). Air pollution and cardiovascular
mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf. October 18, 2004.
Metzger KB; Tolbert PE; Klein M; Peel JL; Flanders WD; Todd KH; Mulholland JA; Ryan PB; Frumkin H (2004).
Ambient air pollution and cardiovascular emergency department visits. Epidemiology, 15: 46-56.
Middleton N; Yiallouros P; Kleanthous S; Kolokotroni O; Schwartz J; Dockery DW; Demokritou P; Koutrakis P
(2008). A 10-year time-series analysis of respiratory and cardiovascular morbidity in Nicosia, Cyprus: the
effect of short-term changes in air pollution and dust storms. Environ Health, 7:39.
NYS DOH. (2006). A study of ambient air contaminants and asthma in New York City, Final Report Part B: Air
contaminants and emergency department visits for asthma in the Bronx and Manhattan. Prepared for: The
U.S. Department of Health and Human Services, Agency for Toxic Substance and Disease Registry.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003). Coarse particles and daily mortality in Coachella Valley,
California. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf.
October 18, 2004.
Peel JL; Tolbert PE; Klein M; Metzger KB; Flanders WD; Knox T; Mulholland JA; Ryan PB; Frumkin H (2005).
Ambient air pollution and respiratory emergency department visits. Epidemiology, 16: 164-174.
Peng RD; Chang HH; Bell ML; McDermott A; Zeger SL; Samet JM; Dominici F. (2008). Coarse paniculate matter
air pollution and hospital admissions for cardiovascular and respiratory diseases among Medicare patients.
JAMA, 299: 2172-2179.
Peng R; Bell M; Geyh A; McDermott A; Zeger S; Samet J; Dominici F (2009). Emergency admissions for
cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environ
HealthPerspect, 117: 957-963.
Perez L; Tobias A; Querol X; Kunzli N; Pey J; Alastuey A; Viana M; Valero N; Gonzalez-Cabre M; Sunyer J.
(2008). Coarse particles from Saharan dust and daily mortality. Epidemiology, 19: 800-807.
Peters A; Dockery DW; Muller JE; Mittleman MA (2001). Increased paniculate air pollution and the triggering of
myocardial infarction. Circulation, 103: 2810-2815.
Pope CA III (1989). Respiratory disease associated with community air pollution and a steel mill, Utah Valley. Am J
Public Health, 79: 623-628.
Pope CA III; Dockery DW; Spengler JD; Raizenne ME (1991). Respiratory health and PM10 pollution: a daily time
series analysis. Am J Respir Crit Care Med, 144: 668-674.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992). Daily mortality andPMlO pollution in Utah valley. Arch.
Environ. Health 47: 211-217.
Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999). Paniculate air pollution and daily mortality on Utah's Wasatch
Front. Environ. Health Perspect.: 107: 567-573.
3-56
-------
Ross, M; Langstaff J. (2005). Updated statistical information on air quality data from epidemiologic studies.
Memorandum to PM NAAQS review docket OAR-2001-0017. January 31, 2005. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/data/ross_langstaff_2005.pdf
Samet, J. (20 lOc). Letter to EPA Administrator Lisa Jackson: "CASAC Review of Policy Assessment for the
Review of the PM NAAQS - First External Review Draft (March 2010)." EPA-CASAC-10-011, May 17,
2010.
Samet, J. (20 lOd). Letter to EPA Administrator Lisa Jackson: "CASAC Review of Policy Assessment for the
Review of the PM NAAQS - Second External Review Draft (June 2010)." EPA-CASAC-10-105,
September 10, 2010.
Schmidt, M. (2005). Analyses of Paniculate Matter (PM) Data for the PM NAAQS Review. Memorandum to PM
NAAQS review docket OAR-2001-0017. June 30, 2005.
Schmidt, M.; Jenkins S.M. (2010). Analyses of PMj0 and PM10.2 5 Data for the PM NAAQS Review. Memorandum
to PM NAAQS review docket EPA-HQ-OAR-2007-0492.
Schmidt, M. (201 Ib). PM10andPM10-25 Air Quality Analyses. Memorandum to PM NAAQS review docket EPA-
HQ-OAR-2007-0492.
Schwartz J.; Dockery, D. W.; Neas, L. M. (1996). Is daily mortality associated specifically with fine particles? J. Air
Waste Manage. Assoc. 46: 927-939.
Schwartz J (1997). Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology, 8:
371-377.
Schwartz J; Norris G; Larson T; Sheppard L; Claiborne C; Koenig J (1999). Episodes of high coarse particle
concentrations are not associated with increased mortality. Environ Health Perspect, 107: 339-342.
Sheppard L; Levy D; Norris G; Larson TV; Koenig JQ (2003). Effects of ambient air pollution and nonelderly
asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology, 10: 23-30.
Slaughter JC; Kim E; Sheppard L; Sullivan JH; Larson TV; Claiborn C. (2005). Association between paniculate
matter and emergency room visits, hospital admissions and mortality in Spokane, Washington. J Expo Sci
Environ Epidemiol, 15: 153-159.
Stanek L, Hassett-Sipple B, Yang R. (2010). Paniculate Matter Air Quality Data Requested from Epidemiologic
Study Authors. Memorandum to PM NAAQS review docket EPA-HQ-OAR-2007-0492.
Tolbert PE; Klein M; Peel JL; Sarnat SE; Sarnat JA (2007). Multipollutant modeling issues in a study of ambient air
quality and emergency department visits in Atlanta. J Expo Sci Environ Epidemiol, 17: S29-S35.
US EPA (1987). PM10 SIP Development Guideline. Office of Air Quality Planning and Standards, Office of Air
and Radiation, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711; report no. EPA-
450/2-86-001. June 1987. Available:
http://nepis.epa.gov/Exe/ZyNET.exe/P1006IKV.txt7Zy ActionD=ZyDocument&Client=EPA&Index=1986
%20Thru%201990&Docs=&Query=%28sampling%20interval%29%20OR%20FNAME%3D%22P 1006IK
V.txt%22%20AND%20FNAME%3D%22P1006IKV.txt%22&Time=&EndTime=&SearchMethod=l&Toc
Restrict=n&Toc=&TocEntry=&QField=&QFieldYear=&QFieldMonth=&QFieldDay=&UseQField=&Int
QFieldOp=l&ExtQFieldOp=l&XmlQuery=&File=D%3A%5CZYFILES%5CINDEX%20DATA%5C86T
HRU90%5CTXT%5C00000019%5CP1006IKV.txt&User=ANONYMOUS&Password=anonymous&Sort
Method=h%7C-
&MaximumDocuments=10&FuzzyDegree=0&ImageQuality=r75g8/r75g8/xl50yl50gl6/i425&Display=p
3-57
-------
%7Cf&DefSeekPage=x&SearchBack=ZyActionL&Back=ZyActionS&BackDesc=Results%20page&Maxi
mumPages= 1 &Zy Entry = 1 & SeekPage=x
US EPA (2004). Air Quality Criteria for Paniculate Matter. National Center for Environmental Assessment, Office
of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711;
report no. EPA/600/P-99/002aF and EPA/600/P-99/002bF. October 2004. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_cd.html
US EPA (2005). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA EPA-452/R-05-005a. December 2005. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html
US EPA (2008a). Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate Matter.
National Center for Environmental Assessment and Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-08-004. March
2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/2008_03_final_integrated_review_plan.pdf
US EPA (2009a). Integrated Science Assessment for Paniculate Matter (Final Report). U.S. Environmental
Protection Agency, Washington, DC, EPA/600/R-08/139F, December 2009. Available:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=216546
US EPA. (2009b). Paniculate Matter National Ambient Air Quality Standards: Scope and Methods Plan for Health
Risk and Exposure Assessment. U.S. Environmental Protection Agency, Research Triangle Park, NC,
EPA-452/P-09-002. February 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_pd.html
US EPA (20 lOa). Quantitative Risk Assessment for Paniculate Matter - Final. Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/P-10-001.
February 2010. Available: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_index.html.
Villeneuve PJ; Burnett RT; Shi Y; Krewski D; Goldberg MS; Hertzman C; Chen Y; Brook J. (2003). A time-series
study of air pollution, socioeconomic status, and mortality in Vancouver, Canada. J Expo Sci Environ
Epidemiol, 13:427-435.
Wilson WE; Mar TF; Koenig JQ. (2007). Influence of exposure error and effect modification by socioeconomic
status on the association of acute cardiovascular mortality with paniculate matter in Phoenix. J Expo Sci
Environ Epidemiol, 17: S11-S19.
Yang C-Y; Chen Y-S; Yang C-H; Ho S-C (2004). Relationship between ambient air pollution and hospital
admissions for cardiovascular diseases in Kaohsiung, Taiwan. J Toxicol Environ Health A Curr Iss, 67:
483-493.
Yang CY; Cheng MH; Chen CC (2009). Effects of Asian Dust Storm Events on Hospital Admissions for Congestive
Heart Failure in Taipei, Taiwan. J Toxicol Environ Health A Curr Iss, 72: 324-328.
Zanobetti A; Schwartz J. (2009). The effect of fine and coarse paniculate air pollution on mortality: A national
analysis. Environ Health Perspect, 117: 898-903.
3-58
-------
4 REVIEW OF THE SECONDARY STANDARDS FOR VISIBILITY-
RELATED EFFECTS
This chapter presents staff conclusions with regard to the adequacy of the current suite of
secondary PM2.5 standards to protect against PM-related visibility impairment as well as
alternative secondary PM standards that are appropriate for consideration in this review. Our
assessment of these issues is framed by a series of key policy-relevant questions, which expand
upon those presented at the outset of this review in the IRP (US EPA, 2008). The answers to
these questions will inform decisions on whether, and if so how, to revise the current suite of
secondary PM2.5 standards for the purpose of providing appropriate protection from PM-related
visibility impairment.
In presenting staff conclusions on a range of alternative secondary standards that are
appropriate for consideration, we note that the final decision is largely a public welfare policy
judgment. A final decision must draw upon scientific information and analyses about PM-
related visibility impairment and related impacts on public welfare, as well as taking into
consideration the range of uncertainties that are inherent in the scientific evidence and analyses.
Our approach to informing these judgments is discussed more fully below.
Information on the approaches used to set the secondary PM2 5 standards in past reviews
as well as our current approach for this review are presented in section 4.1. Our conclusions
regarding the adequacy of the current suite of secondary PM2.5 standards to protect against PM-
related visibility impairment are presented in section 4.2. Section 4.3 presents our conclusions
with respect to alternative PM2.5 standards by focusing on each of the basic elements of the
standards: pollutant indicator (section 4.3.1), averaging time (section 4.3.2), and level and form
(section 4.3.3). The performance of alternative standards, with a focus on the uniformity of
protection from visibility impairment afforded by the alternative standards, is evaluated in
section 4.3.4. Section 4.4 summarizes all staff conclusions on the secondary PM2.5 standards for
visibility protection. This chapter concludes with an overview of areas of key uncertainties and
suggested future research areas and data collection efforts (section 4.5).
4.1 APPROACH
Staffs approach for reviewing the current suite of secondary PM2.5 standards builds upon
and broadens the approaches used in previous PM NAAQS reviews. We first present a brief
summary of the approaches used to review and establish secondary PM standards in the last two
reviews of the PM NAAQS (section 4.1.1). Recent litigation on the 2006 standards has resulted
in the remand of the secondary annual and 24-hour PM2.5 NAAQS to EPA as discussed in
4-1
-------
section 4.1.2. Our current approach for evaluating the secondary PM2.5 standards using both
evidence- and impact assessment-based considerations is outlined in section 4.1.3.
4.1.1 Approaches Used in Previous Reviews
The original suite of secondary PM2.5 standards was established in 1997 and revisions to
those standards were made in 2006. The approaches used in making final decisions on
secondary standards in those reviews, as well as the current review, utilize different ways to
consider the underlying body of scientific evidence. They also reflect an evolution in our
understanding of the nature of the effect on public welfare from visibility impairment, from an
approach focusing only on Class I area visibility impacts to a more multifaceted approach that
also considers PM-related impacts on non-Class I area visibility, such as in urban areas. This
evolution has occurred in conjunction with the expansion of available PM data and information
from associated studies of public perception, valuation, and personal comfort and well-being.
4.1.1.1 Review Completed in 1997
In 1997, EPA revised the identical primary and secondary PM NAAQS in part by
establishing new identical primary and secondary PM2.5 standards. In revising the secondary
standards, EPA recognized that PM produces adverse effects on visibility and that impairment of
visibility was being experienced throughout the U.S., in multi-state regions, urban areas, and
remote mandatory Class I Federal areas alike. However, in considering an appropriate level for a
secondary standard to address adverse effects of PM2 5 on visibility, EPA concluded that the
determination of a single national level was complicated by regional differences. These
differences included several factors that influence visibility such as background and current
levels of PM25, composition of PM2 5, and average relative humidity. Variations in these factors
across regions could thus result in situations where attaining an appropriately protective
concentration of fine particles in one region might or might not provide adequate protection in a
different region. EPA also determined that there was insufficient information at that time to
establish a level for a national secondary standard that would represent a threshold above which
visibility conditions would always be adverse and below which visibility conditions would
always be acceptable.
Based on these considerations, EPA assessed potential visibility improvements in urban
areas and on a regional scale that would result from attainment of the new primary standards for
PM2.5. The agency concluded that the spatially averaged form of the annual PM2.5 standard was
well suited to the protection of visibility, which involves effects of PM2 5 throughout an extended
viewing distance across an urban area. Based on air quality data available at that time, many
urban areas in the Northeast, Midwest, and Southeast, as well as Los Angeles, were expected to
see perceptible improvement in visibility if the annual PM2 5 primary standard were attained.
4-2
-------
The EPA also concluded that attainment of the 24-hour PM2.5 standard in some areas would be
expected to reduce, to some degree, the number and intensity of "bad visibility" days, resulting
in improvement in the 20% of days having the greatest impairment over the course of a year.
Having concluded that attainment of the annual and 24-hour PM2.5 primary standards
would lead to visibility improvements in many eastern and some western urban areas, EPA also
considered whether these standards could provide potential improvements to visibility on a
regional scale. Based on information available at the time, EPA concluded that attainment of
PM2.5 secondary standards set identical to the primary standards would be expected to result in
visibility improvements in the eastern U.S. at both urban and regional scales, but little or no
change in the western U.S., except in and near certain urban areas.
The EPA then considered the potential effectiveness of a regional haze program, required
by sections 169A and 169B of the Act1 to address those effects of PM on visibility that would
not be addressed through attainment of the primary PM2.5 standards. The regional haze program
would be designed to address the widespread, regionally uniform type of haze caused by a
multitude of sources. The structure and requirements of sections 169 A and 169B of the Act
provide for visibility protection programs that can be more responsive to the factors contributing
to regional differences in visibility than can programs addressing a nationally applicable
secondary NAAQS. The regional haze visibility goal is more protective than a secondary
NAAQS since the goal addresses any anthropogenic impairment rather than just impairment at
levels determined to be adverse to public welfare. Thus, an important factor considered in the
1997 review was whether a regional haze program, in conjunction with secondary standards set
identical to the suite of PM2 5 primary standards, would provide appropriate protection for
visibility in non-Class I areas. The EPA concluded that the two programs and associated control
strategies should provide such protection due to the regional approaches needed to manage
emissions of pollutants that impair visibility in many of these areas.
For these reasons, EPA concluded that a national regional haze program, combined with
a nationally applicable level of protection achieved through secondary PM2.5 standards set
identical to the primary PM2.5 standards, would be more effective for addressing regional
variations in the adverse effects of PM2.5 on visibility than would be national secondary
standards for PM with levels lower than the primary PM2 5 standards. The EPA further
recognized that people living in certain urban areas may place a high value on unique scenic
resources in or near these areas, and as a result might experience visibility problems attributable
1 In 1977, Congress established as a national goal' 'the prevention of any future, and the remedying of any existing,
impairment of visibility in mandatory Class I Federal areas which impairment results from manmade air pollution",
section 169A(a)(l) of the Act. The EPA is required by section 169A(a)(4) of the Act to promulgate regulations to
ensure that "reasonable progress" is achieved toward meeting the national goal.
4-3
-------
to sources that would not necessarily be addressed by the combined effects of a regional haze
program and PM2.5 secondary standards. The EPA concluded that in such cases, state or local
regulatory approaches, such as past action in Colorado to establish a local visibility standard for
the City of Denver, would be more appropriate and effective in addressing these special
situations because of the localized and unique characteristics of the problems involved.
Visibility in an urban area located near a mandatory Class I Federal area could also be improved
through state implementation of the then-current visibility regulations, by which emission
limitations can be imposed on a source or group of sources found to be contributing to
"reasonably attributable" impairment in the mandatory Class I Federal area.
Based on these considerations, EPA set secondary PM2.5 standards identical to the
primary PM2 5 standards, in conjunction with a regional haze program under sections 169A and
169B of the Act, as the most appropriate and effective means of addressing the public welfare
effects associated with visibility impairment. Together, the two programs and associated control
strategies were expected to provide appropriate protection against PM-related visibility
impairment and enable all regions of the country to make reasonable progress toward the
national visibility goal.
4.1.1.2 Review Completed in 2006
In 2006, EPA revised the secondary PM2.5 standards to address visibility impairment by
making them identical to the revised primary standards. The EPA's decision regarding the need
to revise the secondary PM2.5 standards reflected a number of new developments and sources of
information that had occurred and/or become available following the 1997 review. First, EPA
promulgated a regional haze program in 1999 (65 FR 35713) which required states to establish
goals for improving visibility in Class I areas and to adopt control strategies to achieve these
goals. Second, extensive new information from visibility and fine particle monitoring networks
had become available, allowing for updated characterizations of visibility trends and PM levels
in urban areas, as well as Class I areas. These new data allowed EPA to better characterize
visibility impairment in urban areas and the relationship between visibility and PM2.5
concentrations. Finally, additional studies in the U.S. and abroad provided the basis for the
establishment of standards and programs to address specific visibility concerns in a number of
local areas. These studies (e.g., in Denver, Phoenix, British Columbia) utilized photographic
representations of visibility impairment and produced reasonably consistent results in terms of
the visual ranges found to be generally acceptable by study participants. The EPA considered
the information generated by these studies useful in characterizing the nature of particle-induced
haze and for informing judgments about the acceptability of various levels of visual air quality in
urban areas across the U.S. Based largely on this information, the Administrator concluded that
4-4
-------
it was appropriate to revise the secondary PM standards to provide increased protection from
visibility impairment principally in urban areas, in conjunction with the regional haze program
for protection of visual air quality in Class I areas.
In so doing, the Administrator recognized that PM-related visibility impairment is
principally related to fine particle levels and that perception of visibility impairment is most
directly related to short-term, nearly instantaneous levels of visual air quality. Thus, in
considering whether the current suite of secondary standards would provide the appropriate
degree of protection, he concluded that it was appropriate to focus on just the 24-hour secondary
PM2.5 standard to provide requisite protection.
The Administrator then considered whether PM2.5 remained the appropriate indicator for
a secondary standard to protect visibility, primarily in urban areas. The Administrator noted that
PM-related visibility impairment is principally related to fine particle levels. Hygroscopic
components of fine particles, in particular sulfates and nitrates, contribute disproportionately to
visibility impairment under high humidity conditions. Particles in the coarse mode generally
contribute only marginally to visibility impairment in urban areas. With the substantial addition
to the air quality and visibility data made possible by the national urban PM2.5 monitoring
networks, an analysis conducted for the 2006 review found that, in urban areas, visibility levels
showed far less difference between eastern and western regions on a 24-hour or shorter time
basis than implied by the largely non-urban data available in the 1997 review. In analyzing how
well PM2.5 concentrations correlated with visibility in urban locations across the U.S., the 2005
Staff Paper (US EPA, 2005) concluded that clear correlations existed between 24-hour average
PM2 5 concentrations and calculated (i.e., reconstructed) light extinction, which is directly related
to visual range. These correlations were similar in the eastern and western regions of the U.S.
These correlations were less influenced by relative humidity and more consistent across regions
when PM2.5 concentrations are averaged over shorter, daylight time periods (e.g., 4 to 8 hours)
when relative humidity was generally lower and less variable. The 2005 Staff Paper noted that a
standard set at any specific PM2.5 concentration would necessarily result in visual ranges that
vary somewhat in urban areas across the country, reflecting the variability in the correlations
between PM2.5 concentrations and light extinction. The 2005 Staff Paper concluded that it was
appropriate to use PM2 5 as an indicator for standards to address visibility impairment in urban
areas, especially when the indicator is defined for a relatively short period (e.g., 4 to 8 hours) of
daylight hours. Based on their review of the Staff Paper, most CAS AC Panel members also
endorsed such a PM2.5 indicator for a secondary standard to address visibility impairment
(Henderson, 2005a). Based on the above considerations, the Administrator concluded that PM2.5
should be retained as the indicator for fine particles as part of a secondary standard to address
visibility protection, in conjunction with averaging times from 4 to 24 hours.
4-5
-------
In considering what level of protection against PM-related visibility impairment would be
appropriate, the Administrator took into account the results of the public perception and attitude
surveys regarding the acceptability of various degrees of visibility impairment in the U.S. and
Canada, state and local visibility standards within the U.S., and visual inspection of photographic
representations of several urban areas across the U.S. In the Administrator's judgment, these
sources provided useful but still quite limited information on the range of levels appropriate for
consideration in setting a national visibility standard primarily for urban areas, given the
generally subjective nature of the public welfare effect involved. Based on photographic
representations of varying levels of visual air quality, public perception studies, and local and
state visibility standards, the 2005 Staff Paper had concluded that 30 to 20 ug/m3 PM2.5
represented a reasonable range for a national visibility standard primarily for urban areas, based
on a sub-daily averaging time. The upper end of this range was below the levels at which
illustrative scenic views are significantly obscured, and the lower end was around the level at
which visual air quality generally appeared to be good based on observation of the illustrative
views. This concentration range generally corresponded to median visual ranges in urban areas
within regions across the U.S. of approximately 25 to 35 km, a range that was bounded above by
the visual range targets selected in specific areas where state or local agencies placed particular
emphasis on protecting visual air quality. In considering a reasonable range of forms for a PM2.5
standard within this range of levels, the 2005 Staff Paper had concluded that a concentration-
based percentile form was appropriate, and that the upper end of the range of concentration
percentiles should be consistent with the 98th percentile used for the primary standard and that
the lower end of the range should be the 92nd percentile, which represented the mean of the
distribution of the 20% most impaired days, as targeted in the regional haze program. While
recognizing that it was difficult to select any specific level and form based on then-currently
available information (Henderson, 2005a), the CASAC Panel was generally in agreement with
the ranges of levels and forms presented in the 2005 Staff Paper.
The Administrator also considered the level of protection that would be afforded by the
proposed suite of primary PM2.5 standards (71 FR 2681), on the basis that although significantly
more information was available than in the 1997 review concerning the relationship between fine
PM levels and visibility across the country, there was still little available information for use in
making the relatively subjective value judgment needed in selecting the appropriate degree of
protection to be afforded by such a standard. In so doing, the Administrator compared the extent
to which the proposed suite of primary standards would require areas across the country to
improve visual air quality with the extent of increased protection likely to be afforded by a
standard based on a sub-daily averaging time. Based on such an analysis, the Administrator
observed that the predicted percent of counties with monitors not likely to meet the proposed
4-6
-------
suite of primary PM2.5 standards was actually somewhat greater than the predicted percent of
counties with monitors not likely to meet a sub-daily secondary standard with an averaging time
of 4 daylight hours, a level toward the upper end of the range recommended in the 2005 Staff
Paper, and a form within the recommended range. Based on this comparison, the Administrator
tentatively concluded that revising the secondary 24-hour PM2.5 standard to be identical to the
proposed revised primary PM2.5 standard (and retaining the current annual secondary PM2.5
standard) was a reasonable policy approach to addressing visibility protection primarily in urban
areas. In proposing this approach, the Administrator also solicited comment on a sub-daily (4- to
8-hour averaging time) secondary PM2.5 standard (71 FR 2675-2781).
In commenting on the proposed decision, the CASAC requested that a sub-daily standard
to protect visibility be favorably reconsidered (Henderson, 2006a). The CASAC noted three
cautions regarding the proposed reliance on a secondary PM2.5 standard identical to the proposed
24-hour primary PM2 5 standard: (1) PM2 5 mass measurement is a better indicator of visibility
impairment during daylight hours, when relative humidity is generally low; the sub-daily
standard more clearly matches the nature of visibility impairment, whose adverse effects are
most evident during the daylight hours; using a 24- hour PM2 5 standard as a proxy introduces
error and uncertainty in protecting visibility; and sub-daily standards are used for other NAAQS
and should be the focus for visibility; (2) CASAC and its monitoring subcommittees have
repeatedly commended EPA's initiatives promoting the introduction of continuous and near-
continuous PM monitoring, and expanded deployment of continuous PM2.5 monitors is consistent
with setting a sub-daily standard to protect visibility; and (3) The analysis showing a similarity
between percentages of counties not likely to meet what the CASAC Panel considered to be a
lenient 4- to 8-hour secondary standard and a secondary standard identical to the proposed 24-
hour primary standard was a numerical coincidence that was not indicative of any fundamental
relationship between visibility and health. The CASAC Panel further stated that "visual air
quality is substantially impaired at PM2.5 concentrations of 35 ug/m3" and that "it is not
reasonable to have the visibility standard tied to the health standard, which may change in ways
that make it even less appropriate for visibility concerns."
In reaching a final decision, the Administrator focused on the relative protection provided
by the proposed primary standards based on the above-mentioned similarities in percentages of
counties meeting alternative standards, and on the limitations in the information available
concerning studies of public perception and attitudes regarding the acceptability of various
degrees of visibility impairment in urban areas, as well as on the subjective nature of the
judgment required. In so doing, the Administrator concluded that caution was warranted in
establishing a distinct secondary standard for visibility impairment and that the available
information did not warrant adopting a secondary standard that would provide either more or less
4-7
-------
protection against visibility impairment in urban areas than would be provided by secondary
standards set equal to the proposed primary PM2.5 standards.
4.1.2 Remand of Secondary PMi.5 Standards
Several parties filed petitions for review challenging EPA's decision to set the secondary
NAAQS for fine PM at the same level as the primary NAAQS. On judicial review, the D.C.
Circuit remanded to EPA for reconsideration the secondary NAAQS for fine PM because the
Agency's decision was unreasonable and contrary to the requirements of section 109(b)(2).
American Farm Bureau Federation v. EPA, 559F.3d512(D.C. Cir., 2009).
The petitioners argued that EPA's decision lacked a reasoned basis. First, they asserted
that EPA never determined what level of visibility was "requisite to protect the public welfare".
They argued that EPA unreasonably rejected the target level of protection recommended by its
staff, while failing to provide a target level of its own. The court agreed, stating that "the EPA's
failure to identify such a level when deciding where to set the level of air quality required by the
revised secondary fine PM NAAQS is contrary to the statute and therefore unlawful.
Furthermore, the failure to set any target level of visibility protection deprived the EPA's
decision-making of a reasoned basis." 559 F. 3d at 530.
Second, the petitioners challenged EPA's method of comparing the protection expected
from potential standards. They contended that the EPA relied on a meaningless numerical
comparison, ignored the effect of humidity on the usefulness of a standard using a daily
averaging time, and unreasonably concluded that the primary standards would achieve a level of
visibility roughly equivalent to the level the EPA staff and CASAC deemed "requisite to protect
the public welfare." Again, the court found that EPA's equivalency analysis based on the
percentages of counties exceeding alternative standards "failed on its own terms". The same
table showing the percentages of counties exceeding alternative secondary standards, used for
comparison to the percentages of counties exceeding alternative primary standards to show
equivalency, also included six other alternative secondary standards within the recommended
CASAC range that would be more "protective" under EPA's definition than the adopted primary
standards. Two-thirds of the potential secondary standards within the CAS AC's recommended
range would be substantially more protective than the adopted primary standards. The court
found that EPA failed to explain why it looked only at one of the few potential secondary
standards that would be less protective and only slightly less so than the primary standards.
More fundamentally, however, the court found that EPA's equivalency analysis based on
percentages of counties demonstrated nothing about the relative protection offered by the
different standards, and that the tables offered no valid information about the relative visibility
protection provided by the standards. 559 F. 3d at 530-31.
-------
Finally, the Staff Paper had made clear that a visibility standard using PM2.5 mass as the
indicator in conjunction with a daily averaging time would be confounded by regional
differences in humidity. The court noted that EPA acknowledged this problem, yet did not
address this issue in concluding that the primary standards would be sufficiently protective of
visibility. 559 F. 3d at 530. Therefore, the court granted the petition for review and remanded
for reconsideration the secondary PM2.5 NAAQS.
4.1.3 General Approach Used in Current Review
The staffs approach in this review broadens the general approaches used in the last two
PM NAAQS reviews by utilizing, to the extent available, enhanced tools, methods, and data to
more comprehensively characterize visibility impacts. As such, we are taking into account
considerations based on both the scientific evidence ("evidence-based") and a quantitative
analysis of PM-related impacts on visibility ("impact-based") to inform our conclusions related
to the adequacy of the current secondary PM2.5 standards and alternative standards that are
appropriate for consideration in this review. As in past reviews, we are also considering
secondary NAAQS to address PM-related visibility impairment in conjunction with the Regional
Haze Program, such that the secondary NAAQS would focus on protection from visibility
impairment principally in urban areas in conjunction with the Regional Haze Program that is
focused on improving visibility in Class I areas. We again recognize that such an approach is the
most appropriate and effective means of addressing the public welfare effects associated with
visibility impairment in areas across the country. We are seeking to provide as broad an array of
options as is supportable by the available information, recognizing that the selection of a specific
approach to reaching final decisions on the secondary PM2 5 standards will reflect the judgments
of the Administrator.
In preparing this final PA, staff has drawn from the qualitative evaluation of all studies
discussed in the ISA (US EPA, 2009a). We consider the extensive new air quality and source
apportionment information available from the regional planning organizations, long-standing
evidence of PM effects on visibility, and public preference studies from four urban areas (ISA
chapter 9), as well as the integration of evidence across disciplines (ISA chapter 2). In addition,
limited information that has become available regarding the characterization of public
preferences in urban areas has provided some new perspectives on the usefulness of this
information in informing the selection of target levels of urban visibility protection. On these
bases, we are again focusing our assessments in this review on visibility conditions in urban
areas.
Our conclusions reflect our understanding of both evidence-based and impact-based
considerations to inform two overarching questions related to: (1) the adequacy of the current
4-9
-------
suite of PM2.5 standards and (2) what potential alternative standards, if any, should be considered
in this review to provide appropriate protection from PM-related visibility impairment. In
addressing these broad questions, we have organized the discussions below around a series of
more specific questions reflecting different aspects of each overarching question, as summarized
in Figure 4-1. When evaluating the visibility protection afforded by the current or any
alternative standards considered, we have taken into account the four basic elements of the
NAAQS: indicator, averaging time, level, and form.
4-10
-------
Figure 4-1. Overview of Approach for Review of Secondary PM2.s Standards
Adequacy of the Current Suite of PM25 Standards to Protect
Against PM-related Visibility Impairment?
Evidence-based Considerations
> Nature of relationship between PM and visibility
impairment?
> Understandng public perceptions of acceptability
of various levels of visual air quality?
> Characterizing population exposure durations?
Impact Assessment-based Considerations
> Nature, magnitude, and uncertainties related to current
levels of visual air quality?
> Levels of PM-related visibility impairment remaining
upon just meeting current suite of PM-,*, standards?
> Importance of remaining PM-related visibility
impairment from a public welfare perspective?
Does information call into question
adequacy of current suite of
PM.-,.- standards'?
Support for retaining i
the current suite of !
PM,r, standards I
Alternative Secondary Standards Supported by
Current Information?
Indicators
>PM,5 mass concentration?
> DirecBy measured PM,,,, light extinction?
> Calculated PM-,, light extinction?
Averaging and Applicable Times
' Sub-daily (1- hour and mull-hour) averaging times?
« Daylight hours only?
« Relative humidity screen?
- 24-hour average?
Forms
• Percenfile-based forms?
• Maximum daily?
• All daylight hours?
Levels
> Candidate Protection Levels from
visibility preference studies
> Broader public welfare considerations?
Alternative Secondary Standards for Consideration
4-11
-------
4.2 ADEQUACY OF CURRENT STANDARDS
In considering the adequacy of the current suite of PM25 standards, staff addresses the
following overarching question:
Does the currently available scientific evidence and visibility impact information, as
reflected in the ISA and UFVA, support or call into question the adequacy of the visibility
protection afforded by the current suite of fine particle standards?
To inform the answer to this overarching question, we have posed a series of more
specific questions to aid in considering the currently available scientific evidence and the results
of recent quantitative visibility impact analyses in a policy-relevant context, 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 and UFVA (US EPA, 2009a; US EPA, 201 Ob).
4.2.1 Evidence-based and Impact-Based Considerations
In reviewing the adequacy of the current suite of PM2.5 standards, we have taken into
account evidence-based considerations, primarily as presented in the ISA, and impact-based
considerations as presented in the UFVA, by considering causal inference, impacts on
susceptible populations, and the nature and degree of PM-related visibility effects that would be
expected to exist in urban areas when meeting the current standards.
• To what extent do the newly available scientific evidence and other information
strengthen or call into question evidence of associations between ambient fine particle
exposures and visibility effects?
New research conducted by regional planning organizations in support of the Regional
Haze Rule, as discussed in chapter 9 of the ISA, continues to support and refine our
understanding of the nature of the PM visibility effect and the source contributions to that effect
in rural and remote locations. Additional by-products of this research include new insights
regarding the regional source contributions to urban visibility and better characterization of the
increment in PM concentrations and visibility impairment that occur in many cities (i.e., the
urban excess) relative to conditions in the surrounding rural areas (i.e., regional background).
Ongoing urban PM2.5 speciated and mass monitoring has produced new information that has
allowed for updated characterization of current visibility levels in urban areas. Information from
both of these sources of PM data, while useful, has not however changed the fundamental and
long understood science characterizing the contribution of PM, especially fine particles, to
visibility impairment. This science, briefly summarized below, provides the basis for the ISA
designation of the relationship between PM and visibility impairment as causal.
4-12
-------
Visibility impairment is caused by the scattering and absorption of light by suspended
particles and gases in the atmosphere. The combined effect of light scattering and absorption by
both 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 I/distance, which is often expressed in the technical literature as 1/(1 million meters) or
inverse megameters (abbreviated Mm"1). When PM is present in the air, its contribution to light
extinction typically greatly exceeds that of gases.
The amount of light extinction contributed by PM depends on the particle size
distribution and composition, as well as its concentration. If details of the ambient particle size
distribution and composition (including the mixing of components) are known, Mie theory can
be used to accurately calculate PM light extinction (US EPA, 2009a, chapter 9). However,
routine monitoring rarely includes measurements of particle size and composition information
with sufficient detail for such calculations. To make estimation of light extinction more
practical, visibility scientists have developed a much simpler algorithm, known as the IMPROVE
algorithm,2 to estimate light extinction using routinely monitored fine particle (PM2.s) speciation
and coarse particle mass (PMio-2.s) data. In addition, relative humidity information is needed to
estimate the contribution by liquid water that is in solution with hygroscopic PM components
(US EPA, 2009a, section 9.2.2.2; US EPA, 201 Ob, chapter 3). There is both an original and a
revised version of the IMPROVE algorithm. The revised version was developed to address
observed biases in the predictions using the original algorithm under very low and very high
light extinction conditions.3 These IMPROVE algorithms are routinely used to calculate light
extinction levels on a 24-hour basis in Class I areas under the Regional Haze Program.
In either version of the IMPROVE algorithm, the concentration of each of the major
aerosol components is multiplied by a dry extinction efficiency value and, for the hygroscopic
components (i.e., ammoniated sulfate and ammonium nitrate), also multiplied by an additional
factor to account for the water growth to estimate these components' contribution to light
extinction. Both the dry extinction efficiency and water growth terms have been developed by a
combination of empirical assessment and theoretical calculation using typical particle size
distributions associated with each of the major aerosol components. They have been evaluated
by comparing the algorithm estimates of light extinction with coincident optical measurements.
The algorithm is referred to as the IMPROVE algorithm because it was developed specifically to use the aerosol
monitoring data generated at network sites and with equipment specifically designed to support the IMPROVE
program and was evaluated using IMPROVE optical measurements at the subset of sites that make those
measurements (Malmetal., 1994).
3 These biases were detected by comparing light extinction estimates generated from the IMPROVE algorithm to
direct optical measurements in a number of rural Class I areas.
4-13
-------
Summing the contribution of each component gives the estimate of total light extinction (denoted
as bexi), as shown below for the original IMPROVE algorithm.
bext - 3 xf(RH) x [Sulfate]
+ 3 xf(RH) x [Nitrate]
+ 4 x [Organic Mass]
+ 10 x [Elemental Carbon]
+ 1 x [Fine Soil]
+ 0.6 x [Coarse Mass]
+ 10
Light extinction (£ext) is m units of Mm"1, the mass concentrations of the components
indicated in brackets are in units of ug/m3, andf(RH) is the unitless water growth term that
depends on relative humidity. The final term of 10 Mm"1 is known as the Rayleigh scattering
term and accounts for light scattering by the natural gases in unpolluted air. The dry extinction
efficiency for particulate organic mass is larger than those for particulate sulfate and nitrate
principally because the density of the dry inorganic compounds is higher than that assumed for
the PM organic mass components. Since IMPROVE does not include ammonium ion
monitoring, the assumption is made that all sulfate is fully neutralized ammonium sulfate and all
nitrate is assumed to be ammonium nitrate. Though often reasonable, neither assumption is
always true (see US EPA, 2009a, section 9.2.3.1). In the eastern U.S. during the summer there is
insufficient ammonia in the atmosphere to neutralize the sulfate fully. Fine particle nitrates can
include sodium or calcium nitrate, which are the fine particle fraction of generally much coarser
particles due to nitric acid interactions with sea salt at near-coastal areas (sodium nitrate) or nitric
acid interactions with calcium carbonate in crustal aerosol (calcium nitrate). Despite the
simplicity of the algorithm, it performs reasonably well and permits the contributions to light
extinction from each of the major components (including the water associated with the sulfate
and nitrate compounds) to be separately approximated.
1]\Qf(RH) term reflects the increase in light scattering caused by particulate sulfate and
nitrate under conditions of high relative humidity. For relative humidity below 40% thef(RH)
value is 1, but it increases to 2 at -66%, 3 at -83%, 4 at -90%, 5 at -93%, and 6 at -95%
relative humidity. The result is that both particulate sulfate and nitrate are more efficient per unit
mass than any other aerosol component for relative humidity above -85% where its total light
extinction efficiency exceeds the 10 m2/g associated with elemental carbon (EC). Based on this
algorithm, particulate sulfate and nitrate are estimated to have comparable light extinction
efficiencies (i.e., the same dry extinction efficiency andf(RH) water growth terms), so on a per
unit mass concentration basis at any specific relative humidity they are treated as equally
effective contributors to visibility effects.
4-14
-------
Inspection of the PM component-specific terms in the simple original IMPROVE
algorithm shows that most of the PM2.5 components contribute 5 times or more light extinction
than a similar concentration of PMi0-2.5. We also know that particles with hygroscopic
components (e.g., particulate sulfate and nitrate) contribute more light extinction at higher
relative humidity than at lower relative humidity because they change size in the atmosphere in
response to ambient relative humidity conditions. PM containing elemental or black carbon
absorbs light as well as scattering it, making it the component with the greatest light extinction
contributions per unit of mass concentration, except for the hygroscopic components under high
relative humidity conditions.4
Subsequent to the development of the original IMPROVE algorithm, an alternative
algorithm (variously referred to as the "revised algorithm" or the "new algorithm" in the
literature) has been developed that employs a more complex split component mass extinction
efficiency to correct biases believed to be related to particle size distributions, a sea salt term that
can be important for remote coastal areas, a different multiplier for organic carbon for purposes
of estimating organic carbonaceous material, and site-specific Rayleigh light scattering terms in
place of a universal Rayleigh light scattering value. These features of the revised IMPROVE
algorithm are described in section 9.2.3.1 of the ISA, which also presents a comparison of the
estimates produced by the two algorithms for rural areas. Compared to the original algorithm,
the revised IMPROVE algorithm can yield higher estimates of current light extinction levels in
urban areas on days with relatively poor visibility (Pitchford, 2010). This difference is primarily
attributable to the split component mass extinction efficiency treatment in the revised algorithm
rather than to the inclusion of a sea salt term or the use of site-specific Rayleigh scattering
values.
As mentioned above, particles are not the only contributor to ambient visibility
conditions. Light scattering by gases also occurs in ambient air. Under pristine atmospheric
conditions, naturally occurring gases such as N2 and O2 cause what is known as Rayleigh
scattering. Rayleigh scattering depends on the density of air as a function primarily of the
elevation above sea level, and can be treated as a site-dependent constant. The Rayleigh
scattering contribution to light extinction is only significant under pristine conditions. The only
other commonly occurring atmospheric gas to appreciably absorb light in the visible spectrum is
NO2. NO2 forms in the atmosphere from NO emissions associated with combustion processes.
These combustion processes also emit PM at levels that generally contribute much higher light
extinction than the NO2 (i.e., NO2 absorption is generally less than approximately 5% of the light
extinction, except where emission controls remove most of the PM prior to releasing the
4 The IMPROVE algorithm does not explicitly separate the light-scattering and light-absorbing effects of elemental
carbon.
4-15
-------
remaining gases to the atmosphere). The remainder of this chapter focuses on the contribution of
PM, which is typically much greater than that of gases, to ambient light extinction, unless
otherwise specified.
In the following discussions, visual air quality is characterized both in terms of light
extinction, as discussed above, and deciviews (dv). Deciview refers to an alternative scale for
characterizing visibility that is defined directly in terms of light extinction (expressed in units of
Mm"1) by the following equation5:
Deciview (dv) = 10 In (bext / 10 Mm"1)
The deciview scale is frequently used in the scientific and regulatory literature on visibility, as
well as in the Regional Haze Program. In particular, the deciview scale is used in the public
perception studies that were considered in the past and current reviews to inform judgments
about an appropriate degree of protection to be provided by a secondary NAAQS.
• To what extent does the available evidence inform our understanding of the temporal
nature of the PM visibility effect, including relevant exposure periods, associated
atmospheric conditions, and diurnal patterns of exposure?
Temporal Variations of Light Extinction
PM concentrations and light extinction in urban environments vary from hour-to-hour
throughout the 24-hour day due to a combination of diurnal changes in meteorological conditions
and systematic changes in emissions activity (e.g., rush hour traffic). Generally, low mixing
heights at night and during the early morning hours tend to trap locally produced emissions,
which are diluted as the mixing height increases due to heating during the day. Low
temperatures and high relative humidity at night are conducive to the presence of ammonium
nitrate particles and water growth by hygroscopic particles compared with the generally higher
temperatures and lower relative humidity later in the day. These combine to make early morning
the most likely time for peak urban light extinction. Superimposed on such systematic time-of-
day variations are the effects of synoptic meteorology (i.e., those associated with changing
weather) and regional-scale air quality that can generate peak light extinction impacts any time
of day. The net effects of the systematic urban- and larger-scale variations are that peak daytime
PM light extinction levels can occur any time of day, although in many areas they most often
occur in early morning hours (USEPA, 2010b, sections 3.4.2 and 3.4.3; Figures 3-9, 3-10, and 3-
12).
5 As used in the Regional Haze Program, the termbext refers to light extinction due to PM25, PM10-2.5j and "clean"
atmospheric gases. In this chapter, in focusing on light extinction due to PM2 5, the deciview values include only the
effects of PM2 5 and the gases. The "Rayleigh" term associated with clean atmospheric gases is represented by the
constant value of 10 Mm"1. Omission of the Rayleigh term would create the possibility of a negative deciview
values when the PM2 5 concentration is very low.
4-16
-------
This temporal pattern in urban areas contrasts with the general lack of a strong diurnal
pattern in PM concentrations and light extinction in most Class I areas, reflective of a relative
lack of local sources as compared to urban areas. The use in the Regional Haze Program of 24-
hour average concentrations in the IMPROVE algorithm is consistent with this general lack of a
strong diurnal pattern in Class I areas.
Periods during the Day of Interest for Assessment of Visibility
Typically, we think of visibility associated with daytime periods because we are outside
more during the day than at night and there are more viewable scenes at a distance during the day
than at night. We recognize, however, that PM light extinction behaves the same at night as
during the day, enhancing the scattering of anthropogenic light, contributing to the "skyglow"
within and over populated areas, adding to the total sky brightness, and contributing to the
reduction in contrast of stars against the background. These effects produce the visual result of a
reduction in the number of visible stars and the disappearance of diffuse or subtle phenomena
such as the Milky Way. The extinction of starlight is a secondary and minor effect also caused
by increased PM scattering and absorption.
However, there are significant and important differences between daytime and nighttime
visual environments with regard to how light extinction per se relates to visual air quality (or
visibility) and public welfare. First, daytime visibility has dominated the attention of those who
have studied the visibility effects of air pollution, particularly in urban areas. As a result, little
research has been conducted on nighttime visibility and the state of the science is not yet
comparable to that associated with daytime visibility impairment. We are not aware of urban-
focused preference or valuation studies providing information on public preferences for
nighttime visual air quality. Second, in addition to air pollution, nighttime visibility is affected
by the addition of light into the sight path from numerous sources, including anthropogenic light
sources in urban environments such as artificial outdoor lighting, which varies dramatically
across space, and natural sources including the moon, planets, and stars. Light sources and
ambient conditions are typically five to seven orders of magnitude dimmer at night than in
sunlight. Moonlight, like sunlight, introduces light throughout an observer's sight path at a
constant angle. On the other hand, dim starlight emanates from all over the celestial hemisphere
while artificial lights are concentrated in cities and illuminate the atmosphere from below. These
different light sources will yield variable changes in visibility as compared to what has been
established for the daytime scenario, in which a single source, the sun, is by far the brightest
source of light. Third, the human psychophysical response (e.g., how the human eye sees and
processes visual stimuli) at night is expected to differ (US EPA, 2009a, section 9.2.2).
4-17
-------
Given the above, we do not believe that the science is available at this time to support
adequate characterization specifically of nighttime PM light extinction conditions and the related
effects on public welfare. Thus, we have focused our assessments of PM visibility impacts in
urban areas on daylight hours. For simplicity, and because perceptions and welfare effects from
light extinction-related visual effects during the minutes of actual sunrise and sunset have not
been explored, we have defined daylight hours as those hours entirely after the local sunrise time
and before the local sunset time.
In so doing, we note that the 24-hour averaging time used in the Regional Haze Program
includes nighttime conditions. We also note, however, that the goal of the Regional Haze
Program is to address any manmade impairment of visibility without regard to distinctions
between daylight and nighttime conditions. Moreover, because of the lack of strong diurnal
patterns, both nighttime and daylight visibility are strongly correlated with 24-hour average
visibility conditions, so a 24-hour averaging period is suitable for driving both daylight and
nighttime visibility towards their natural conditions. Also, the focus on 24-hour average
visibility allows the Regional Haze Program to make use of more practically obtained ambient
PM measurements of adequate accuracy than if a shorter averaging period were used, an
important consideration given the remoteness of many Class I area monitoring sites and given
the low PM concentrations that must be measured accurately in such areas.
In addition, when natural conditions such as fog and rain cause poor visibility, it can be
reasonably assumed that the light extinction properties of the air that are attributable to air
pollution are not important from a public welfare perspective. Thus, it is appropriate to give
special treatment to such periods when considering whether current PM2 5 standards adequately
protect public welfare from PM-related visibility impairment. In evaluating alternative sub-daily
standards, we have addressed this issue by screening out hours with particularly high relative
humidity. As discussed further below (section 4.3.2.1), we used a relative humidity screen of
90% on the basis that it serves as a reasonable surrogate for excluding hours affected by fog and
rain.
Exposure Durations of Interest
The roles that exposure duration and variations in visual air quality within any given
exposure period play in determining the acceptability or unacceptability of a given level of visual
air quality has not been investigated via preference studies. In the preferences studies available
for this review, subjects were simply asked to rate the acceptability or unacceptability of each
image of a haze-obscured scene, without being provided any suggestion of assumed duration or
of assumed conditions before or after the occurrence of the scene presented. We do know from
preference and/or valuation studies that atmospheric visibility conditions can be quickly assessed
4-18
-------
and preferences determined. These studies show that a momentary glance at an image of a scene
(i.e., less than a minute) is enough for study participants to judge the acceptability or
unacceptability of the viewed visual air quality conditions. Moreover, individual participants in
general consistently judge the acceptability of same-scene images that differed only with respect
to light extinction levels when these images were presented repeatedly for such short periods.
That is, individuals generally did not say that a higher-light extinction image was acceptable
while saying a lower-light extinction, same-scene image was unacceptable, even though they
could not compare images side-to-side. However, we do not have information about what
assumptions, if any, the participants may have made about the duration of exposure in
determining the acceptability of the images. We are unaware of any studies that characterize the
extent to which different frequencies and durations of exposure to visibility conditions contribute
to the degree of public welfare impact that occurs.
In the absence of such studies, we considered a variety of circumstances that are
commonly expected to occur in evaluating the potential impact of visibility impairment on the
public welfare based on the information we do have. In some circumstances, such as infrequent
visits to scenic vistas in natural or urban environments, people are motivated specifically to take
the opportunity to view a valued scene and are likely to do so for many minutes to hours to
appreciate various aspects of the vista they choose to view. In such circumstances, the viewer
may consciously evaluate how the visual air quality at that time either enhances or diminishes
the experience or view. However, the public also has many more opportunities to notice
visibility conditions on a daily basis in settings associated with performing daily routines (e.g.,
during commutes and while working, exercising, or recreating outdoors). These scenes, whether
iconic or generic, may not be consciously viewed for their scenic value and may not even be
noticed for periods comparable to what would be the case during purposeful visits to scenic
visits, but their visual air quality may still affect a person's sense of wellbeing. Research has
demonstrated that people are emotionally affected by low visual air quality, that perception of
pollution is correlated with stress, annoyance, and symptoms of depression, and that visual air
quality is deeply intertwined with a "sense of place," affecting people's sense of the desirability
of a neighborhood (US EPA, 2009a, section 9.2.4). Though we do not know the extent to which
these emotional effects are linked to different periods of exposure to poor visual air quality,
providing additional protection against short-term exposures to levels of visual air quality
considered unacceptable by subjects in the context of the preference studies would be expected
to provide some degree of protection against the risk of loss in the public's "sense of wellbeing."
Some people have mostly intermittent opportunities on a daily basis (e.g., during morning
and/or afternoon commutes) to experience ambient visibility conditions because they spend
much of their time indoors without access to windows. For such people a view of poor visual air
4-19
-------
quality during their morning commute may provide their perception of the day's visibility
conditions until the next time they venture outside during daylight hours later or perhaps the next
day. Other people have exposure to visibility conditions throughout the day, conditions that may
differ from hour to hour. A day with multiple hours of visibility impairment would likely be
judged as having a greater impact on their wellbeing than a day with just one such hour followed
by clearer conditions.
We have no information or studies on the fraction of the public that has only one or a few
opportunities to experience visibility during the day, or information or studies on the duration of
the effect on wellbeing from exposure to different durations of poor visual air quality conditions.
However, it is logical to conclude that people with limited opportunities to experience visibility
conditions on a daily basis would receive the entire impact of the day's visual air quality based
on the visibility conditions that occur during the short time period when they can see it. Since
this group could be affected on the basis of observing visual air quality conditions for periods as
short as one hour or less, and because during each daylight hour there are some people outdoors,
commuting, or near windows, we judge that it would be appropriate to use the maximum hourly
value of PM light extinction during daylight hours for each day for purposes of evaluating the
adequacy of the current suite of secondary standards. This approach would recognize that at
least some but not all of the population of an area will actually be exposed to this worst hour and
that some that people who are exposed to this worst hour may not have an opportunity to observe
clearer conditions in other hours if they were to occur. Moreover, because visibility conditions
and people's daily activities on work/school days both tend to follow the same diurnal pattern
day after day, those who are exposed only to the worst hour will tend to have this experience day
after day.
For another group of observers, those who have access to visibility conditions often or
continuously throughout the day, the impact of the day's visibility conditions on their welfare
may be based on the varying visibility conditions they observe throughout the day. For this
group, it might be that an hour with poor or "unacceptable" visibility can be offset by one or
more other hours with clearer conditions. Based on these considerations, we judge that it would
also be appropriate to use the maximum multi-hour daylight period for evaluating the adequacy
of the current suite of secondary standards.
The above discussion is based on what people see, which is determined by the extinction
of light along the paths between observers and various objects that are looked at. A related but
separate issue is what measurement period is relevant, if (as expected) what will be measured is
the light extinction property or the PM concentration of the local air at a fixed site. Light
extinction conditions at a fixed site can change quickly (i.e., in less than a minute). Sub-hourly
variations in light extinction determined at any point in the atmosphere are likely the result of
4-20
-------
small-scale spatial pollution features (i.e., high concentration plumes) just generated in the
immediate vicinity due to local sources or being transported by the wind across that point. These
small-scale pockets of air causing short periods of higher light extinction at the fixed site likely
do not determine the visual effect for scenes with longer sight paths. In contrast, atmospheric
sight path-averaged light extinction which is pertinent to visibility impacts generally changes
more slowly (i.e., tens of minutes generally), because a larger air mass must be affected by a
broader set of emission sources or the air mass must be replaced by a cleaner or dirtier air mass
due to the wind operating over time. At typical wind speeds found in U.S. cities, an hour
corresponds to a few tens of kilometers of air flowing past a point, which is similar to sight path
lengths of interest in urban areas. Based on the above considerations, hourly average light
extinction would generally be reasonably representative of the net visibility effect of the spatial
pattern of light extinction levels, especially along site paths that generally align with the wind
direction.
• Based on currently available information, what range of levels of visibility impairment
is reasonable to consider in reaching judgments about the adequacy of the current
NAAQS?
In order to identify levels of visibility impairment appropriate for consideration in setting
secondary PM NAAQS to protect the public welfare, we comprehensively examined information
that was available in this review regarding people's stated preferences for acceptable and
unacceptable visual air quality.
Light extinction is an atmospheric property that by itself does not directly translate into a
public welfare effect. Instead, light extinction becomes meaningful in the context of the impact
of differences in visibility on the human observer. This has been studied in terms of the
acceptability or unacceptability expressed for the visibility impact of a given level of light
extinction by a human observer. The perception of the visibility impact of a given level of light
extinction occurs in conjunction with the associated characteristics and lighting conditions of the
viewed scene.6 Thus, a given level of light extinction may be perceived differently by observers
looking at different scenes or the same scene with different lighting characteristics. Likewise,
different observers looking at the same scene with the same lighting may have different
preferences regarding the associated visual air quality. When scene and lighting characteristics
are held constant, the perceived appearance of a scene (i.e., how well the scenic features can be
seen and the amount of visible haze) depends only on changes in light extinction. This has been
6 By "characteristics of the scene" we mean the distance(s) between the viewer and the object(s) of interest, the
shapes and colors of the objects, the contrast between objects and the sky or other background, and the inherent
interest of the objects to the viewer. Distance is particularly important because at a given value of light extinction,
which is a property of air at a given point(s) in space, more light is actually absorbed and scattered when light passes
through more air between the object and the viewer.
4-21
-------
demonstrated using the WinHaze model (Molenar et al., 1994) that uses image processing
technology to apply user-specified changes in light extinction values to the same base
photograph with set scene and lighting characteristics.
Much of what we know about the acceptability of levels of visibility comes from survey
studies in which participants were asked questions about their preference or the value they place
on various visibility levels as displayed to them in scenic photographs and/or WinHaze images
with a range of known light extinction levels. Urban visibility preference studies for four urban
areas were reviewed in the UFVA (chapter 2) to assess the light extinction levels judged by the
participant to have acceptable visibility for those particular scenes. While the results differed
among the four urban areas, results from a rating exercise showed that within each preference
study, individual survey participants consistently distinguish between photos or images
representing different levels of light extinction, and that more participants rate as acceptable
images representing lower levels of light extinction than do images representing higher levels.
The reanalysis of urban preference studies conducted for this review included three
completed urban visibility preference survey studies plus a pair of smaller focus studies designed
to explore and further develop urban visibility survey instruments. The three western studies
included one in Denver, Colorado (Ely et al., 1991), one in the lower Fraser River valley near
Vancouver, British Columbia (BC), 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 Inc., 2001). In response to an EPA request for public comment on the Scope
and Methods Plan (74 FR 11580, March 18, 2009), we received comments (Smith, 2009) about
the results of a new Washington, DC focus group study that had been conducted using methods
and approaches similar to the method and approach employed in the EPA pilot study (Smith and
Howell, 2009). When taken together, these studies from the four different urban areas included a
total of 852 individuals, with each individual responding to a series of questions answered while
viewing a set of images of various urban visual air quality conditions.
The approaches used in the four studies are similar and are all derived from the method
first developed for the Denver urban visibility study. In particular, the studies all used a similar
group interview type of survey to investigate the level of visibility impairment that participants
described as "acceptable." While each study asked the basic question, "What level of visibility
degradation is acceptable?", the term "acceptable" was not defined, so that each person's
response was based on his/her own values and preferences for visual air quality. Given the
similarities in the approaches used, it is reasonable to compare the results to identify overall
trends in the study findings and that this comparison can usefully inform the selection of a range
of levels for use in further analyses. However, variations in the specific materials and methods
used in each study introduce uncertainties that should also be considered when interpreting the
4-22
-------
results of these comparisons. Key differences between the studies include: 1) scene
characteristics; 2) image presentation methods (e.g., projected slides of actual photos, projected
images generated using WinHaze (a significant technical advance in the method of presenting
visual air quality conditions), or use of a computer monitor screen; 3) number of participants in
each study; 4) participant representativeness of the general population of the relevant
metropolitan area; and 5) specific wording used to frame the questions used in the group
interview process.
In the UFVA, each study was evaluated separately and figures developed to display the
percentage of participants that rated the visual air quality depicted in each photograph as
"acceptable." Ely et al. (1991) introduced a "50% acceptability" criterion analysis of the Denver
preference study results. The 50% acceptability criterion is designed to identify the visual air
quality level (defined in terms of deciviews or light extinction) that best divides the photographs
into two groups: those with a visual air quality rated as acceptable by the majority of the
participants, and those rated not acceptable by the majority of participants. We adopted the
criterion as a useful index for comparison between studies. The results of each individual
analysis were then combined graphically to allow for visual comparison. Figure 4-2 (Figure 2-
16 in UFVA) presents the graphical summary of the results of the studies in the four cities and
draws on results previously presented in Figures 2-3, 2-5, 2-7, and 2-11 of chapter 2 in the
UFVA. Figure 4-2 also contains lines at 20 dv and 30 dv that generally identify a range where
the 50% acceptance criteria occur across all four of the urban preference studies. Out of the 114
data points shown in Figure 4-2, only one photograph (or image) with a visual air quality below
20 dv was rated as acceptable by less than 50% of the participants who rated that photograph.7
Similarly, only one image with a visual air quality above 30 dv was rated acceptable by more
than 50% of the participants who viewed it.8
7 Only 47% of the BC participants rated a 19.2 dv photograph as acceptable.
8 In the 2001 Washington, D.C. study, a 30.9 dv image was used as a repeated slide. The first time it was shown
56% of the participants rated it as acceptable, but only 11% rated it as acceptable the second time it was shown. The
same VAQ level was rated as acceptable by 42% of the participants in the 2009 study (Test 1). All three points are
shown in Figure 4-2.
4-23
-------
Figure 4-2. Summary of Results of Urban Visibility Studies in Four Cities, Showing the
Identified Range of the 50% Acceptance Criteria9
2O Mm-1
SO Mm-1 1OO Mm-1 2OO Mm-1 4OO Mrrr1 SOO Mm'1
100%
'*
50%
f
i
D%
20 25 30
Deciview
35
4C
45
a Denver * Phoenix * BC * Washington
DenverLcgit PhoenixLogit BC Logit DC Logit
As can be seen in the figure, each urban area has a separate and unique response curve
that appears to indicate that it is distinct from the others.10 These curves are the result of a
logistical regression analysis using a logit model of the greater than 19,000 ratings of haze
images as acceptable or unacceptable. The model results can be used to estimate the visual air
quality in terms of deciview values where the estimated response functions cross the 50%
acceptability level, as well as any alternative criteria levels. Selected examples of these are
shown in Table 4-1 (US EPA, 2010b, Table 2-4). These results show that the logit model results
also supports the upper and lower ends of the range of 50th percentile acceptability values (e.g.,
near 20 dv for Denver and near 30 dv for Washington, DC) already identified in Figure 4-2.
Top scale shows light extinction in inverse megameter units; bottom scale in deciviews. Logit analysis estimated
response functions are shown as the color-coded curved lines for each of the four urban areas.
10 Note that the Washington, DC results (black dots) shown in Figure 4-2 include values from two separate studies
individually, so there are at least two points at each haze level (x-axis). There is a third point for some haze levels
because an image was sometimes presented twice. Since one of the two studies had only 9 participants, the percent
of participants' acceptable ratings are limited to multiples of 1/9, so that a single participant rating can move a point
by about 11%. A composite of the results from the two studies (similar to US EPA, 2010b, Figure 2-15) would
show much less scatter among the points, but obscures the fact of it being data from differing studies.
4-24
-------
Table 4-1. Logit Model-estimated Visual Air Quality Values (in deciviews) Corresponding
to Various Percent Acceptability Values for the Four Cities
Acceptability
Criteria
90% Acceptability
75% Acceptability
50% Acceptability
25% Acceptability
10% Acceptability
Denver
14.2
17.1
19.9
22.7
25.6
British
Columbia
16.8
19.6
22.5
25.3
28.1
Phoenix
19.4
21.8
24.2
26.5
28.9
Washington, DC
23.0
26.0
29.0
32.0
35.0
-,th
Based on the composite results and the effective range of 50 percentile acceptability
across the four urban preference studies shown in Figure 4-2 and Table 4-1, benchmark levels of
(total) light extinction have been selected in a range from 20 dv to 30 dv (75 to 200 Mm"1)11 for
the purpose of provisionally assessing whether visibility conditions would be considered
acceptable (i.e., less than the low end of the range), unacceptable (i.e., greater than the high end
of the range), or potentially acceptable (within the range). A midpoint of 25 dv (120 Mm"1) was
also selected for use in the assessment. This level is also very near to the 50th percentile criterion
value from the Phoenix study (i.e., 24.2 dv), which is by far the best of the four studies in terms
of least noisy preference results and the most representative selection of participants. Based on
the currently available information, we conclude that the use of 25 dv to represent the middle of
the distribution of results seems well supported.
These three benchmark values provide a low, middle, and high set of light extinction
conditions that are used to provisionally define daylight hours with urban haze conditions that
have been judged unacceptable by the participants of these preference studies. As discussed
above, PM light extinction is taken to be (total) light extinction minus the Rayleigh scatter (i.e.,
light scattering by atmospheric gases which is on average about 10 Mm"1), so the low, middle,
and high levels correspond to PM light extinction levels of about 65 Mm"1, 110 Mm"1, and 190
Mm"1. In the UFVA, these three light extinction levels were called Candidate Protection Levels
1: These values were rounded from 74 Mm"1 and 201 Mm"1 to avoid an implication of greater precision than is
warranted. Note that the middle value of 25 dv when converted to light extinction is 122 Mm"1 is rounded to 120
Mm"1 for the same reason. Assessments conducted for the UFVA and the first and second draft PAs used the
unrounded values. The EPA staff considers the results of assessment using unrounded values to be sufficiently
representative of what would result if the rounded values were used that it was unnecessary to redo the assessments.
That is why some tables and figures in this document reflect the unrounded values.
4-25
-------
(CPLs). We continue to use this term in this document. However, it is important to note that the
degree of protection provided by a secondary NAAQS is not determined solely by any one
component of the standard but by all the components (i.e., indicator, averaging time, form, and
level) being applied together. Therefore, the reader should keep in mind that the term CPL is
meant only to indicate target levels of visibility within a range that we feel is appropriate for
consideration that could, in conjunction with other elements of the standard, including indicator,
averaging time, and form, provide an appropriate degree of visibility protection.
In characterizing our degree of confidence in each CPL and across the range, a number of
issues were considered. Looking first at the two studies that define the upper and lower bounds
of the range, we considered whether they represent a true regional distinction in preferences for
urban visibility conditions between western and eastern U.S. There is little information available
to help evaluate this, especially given that we have preference studies in only one eastern urban
area. Smith and Howell (2009) found little difference in preference response to Washington, DC
haze photographs between the study participants from Washington, DC and those from Houston,
TX.12 This provides some limited evidence that the value judgment of the public in different
areas of the country may not be an important factor in explaining the differences in these study
results.
In further considering what factors could explain the observed differences in preferences
across the four urban areas, we noted that the urban scenes used in each study had different
characteristics. For example, each of the western urban visibility preference study scenes
included mountains in the background while the single eastern urban study did not. It is also true
that each of the western scenes included objects at greater distances from the camera location
than in the Washington, DC study. There is no question that objects at a greater distance have a
greater sensitivity to perceived visibility changes as light extinction is changed compared to
otherwise similar scenes with objects at a shorter range. This alone might explain the difference
between the results of the Washington, DC study and those from the Western urban studies.
Also, it seems likely that people value the views of mountains in the background more than
generic distant buildings in the foreground of the western scenes; just as it seems likely that the
Capital Mall and Washington Monument were the likely objects of greatest interest for the
Washington, DC study base photograph. Having scenes with the object of greatest intrinsic
value nearer and hence less sensitive for Washington compared with more distant objects of
12 The first preference study using WinHaze images of a scenic vista from Washington, DC was conducted in 2001
using subjects who were residents of Washington, DC. More recently, Smith and Howell (2009) interviewed
additional subjects using the same images and interview procedure. The additional subjects included some residents
of the Washington, DC area and some residents of the Houston, TX area.
4-26
-------
greatest intrinsic value in the western urban areas could further explain the difference in
preference results.
Another question that we considered was whether the high CPL value that is based on the
Washington, DC preference results is likely to be generally representative of urban areas that do
not have associated mountains or other valued objects visible in the distant background. Such
areas would include the middle of the country and many areas in the eastern U.S., and possibly
some areas in the western U.S. as well. In order to examine this issue, an effort would have to be
made to see if scenes in such areas could be found that would be generally comparable to the
western scenes (e.g., scenes that contain valued scenic elements at more sensitive distances than
that used in the Washington, DC study). This is only one of a family of issues concerning how
exposure to urban scenes of varying sensitivity affects public perception for which no preference
study information is currently available. Based on the currently available information, we
conclude that the high end of the CPL range (30 dv) is an appropriate level to consider.
With respect to the low end of the range, we considered factors that might further refine
our understanding of the robustness of this level. We concluded that additional urban preference
studies, especially with a greater variety in types of scenes, could help evaluate whether the
lower CPL value of 20 dv is generally supportable. Further, the reason for the noisiness in data
points around the curves apparent in both the Denver and British Columbia results compared to
the smoother curve fit of Phoenix study results could be explored. One possible explanation that
we identified is that these older studies used photographs taken at different times of day and on
different days to capture the range of light extinction levels needed for the preference studies. In
contrast, the use of WinHaze in the Phoenix (and Washington, DC) study reduced variations in
scene appearance that affects preference rating and avoided the uncertainty inherent in using
ambient measurements to represent sight path-averaged light extinction values. Reducing these
sources of noisiness and uncertainty in the results of future studies of sensitive urban scenes
could provide more confidence in the selection of a low CPL value.
Based on the above considerations, and recognizing the limitations in the currently
available information, staff concludes that it is reasonable to consider a range of CPL values
including a high value of 30 dv, a mid-range value of 25 dv, and a low value of 20 dv. Based on
its review of the second draft PA, CAS AC supported this set of CPLs for consideration by the
EPA in this review. CASAC noted that these CPL values were based on all available visibility
preference data and that they bound the study results as represented by the 50% acceptability
criteria. CASAC concluded that this range of levels is "adequately supported by the evidence
presented" (Samet, 2010d, p. iii).
4-27
-------
• To what extent does the available information demonstrate or suggest that PM-related
visibility impairment within the range of CPLs is occurring at current ambient
conditions, or that such impairment would occur under ambient PMi.s levels that would
meet the current standards?
Current Visibility Levels
Current visibility conditions have been characterized in terms of PM light extinction
levels for the 15 urban areas13 that were selected for analysis in the UFVA. We have analyzed
hourly average PM-related light extinction in terms of both PMio and PM2.5 light extinction.14
These current visibility conditions were then compared to the CPLs identified above.
As an initial matter, as noted above, we recognize that visibility impairment occurs
during periods with fog or precipitation irrespective of the presence or absence of PM. While it
is a popular notion that areas with many foggy or rainy days are "dreary" places to live compared
to areas with more sunny days per year, we have no basis for taking into account how the
occurrence of such days might modify the effect of pollution-induced hazy days on public
welfare. It is logical that periods with naturally impaired visibility due to fog or precipitation
should not be treated as having PM-impaired visibility. Moreover, depending on the specific
indicator, averaging time, and measurement approach used for the NAAQS, foggy conditions
might result in measured or calculated indicator values that are higher than the light extinction
actually caused by PM.15 Therefore, in order to avoid precipitation and fog confounding
estimates of PM visibility impairment, and as advised by CASAC as part of its comments on the
first draft UFVA, we restricted our assessment of visibility conditions to daylight hours with
relative humidity less than or equal to 90% when evaluating sub-daily alternative standards (US
EPA, 2010b, section 3.3.5, and US EPA, 2010b, Appendix G). However, not all periods with
13 Comments on the second draft UFVA from those familiar with the monitoring sites in St. Louis indicated that the
site selected to provide continuous PM10 monitoring, although less than a mile from the site of the PM25 data, is not
representative of the urban area and resulted in unrealistically large PMi0-2.5 values. The EPA staff considers these
comments credible and has set aside the St. Louis assessment results for PMi0 light extinction. Thus, results and
statements in this PA regarding PM10 light extinction apply to only the other 14 areas. However, results regarding
PM2.5 light extinction in most cases apply to all 15 study areas because the St. Louis estimates for PM2 5 light
extinction were not affected by the PM10 monitoring issue.
14 PM-related light extinction is used here to refer to the light extinction caused by PM regardless of particle size;
PM10 light extinction refers to the contribution by particles sampled through an inlet with a particle size 50%
cutpoint of 10 |am diameter; and PM2 5 light extinction refers to the contribution by particles sampled through an
inlet with a particle size 50% cutpoint of 2.5 |am diameter.
15 One example of an indicator and measurement approach for which indicator values could be higher than true PM
light extinction as a result of fog would be a light extinction indicator measured in part by an unheated
nephelometer, which is an optical instrument for measuring PM light scattering from an air sample as it flows
through a measurement chamber. Rain drops would be removed by the initial size-selective inlet device, although
some particles associated with fog may be small enough that they might pass through the inlet and enter the
measurement chamber of the instrument. This would result in a reported scattering coefficient that does not
correspond to true PM light extinction. Direct measurement of light extinction using an open-path instrument would
be even more affected by both fog and precipitation.
4-28
-------
relative humidity above 90% have fog or precipitation. Removing those hours from application
of a secondary PM standard involves a tradeoff between the benefits of not including many of
the hours with meteorological causes of visibility impacts and the loss of public welfare
protection of not including some hours with high relative humidity without fog or precipitation,
where the growth of hygroscopic PM into large solution droplets results in enhanced PM
visibility impacts. For the 15 urban areas included in the assessment for which meteorological
data were obtained to allow an examination of the co-occurrence of high relative humidity and
fog or precipitation, a 90% relative humidity cutoff criterion is effective in that on average less
than 6% of the daylight hours are removed from consideration, yet those hours have on average
ten times the likelihood of rain, six time the likelihood of snow/sleet, and 34 times the likelihood
of fog compared with hours with 90% or lower relative humidity. Based on these findings, we
conclude that it is appropriate that a sub-daily standard intended to protect against PM-related
visibility impairment would be defined in such a way as to exclude hours with relative humidity
greater than approximately 90%, regardless of measured values of light extinction or PM.
The UFVA analyses were done in terms of PMi0 light extinction. Figure 4-3 (Figure 3-8
in UFVA) presents box-and-whisker plots to illustrate the distributions of the estimates of the
daily maximum daylight 1-hour calculated PMio light extinction levels in each area (excluding
St. Louis) based on data from the 2005-2007 time period. The horizontal dashed lines in the
plots represent the low, middle, and high CPLs for PMio light extinction of 65, 110, and 190
Mm"1, corresponding to the benchmark visual air quality values of 20, 25, and 30 dv, as
discussed above. Table 4-2 (Table 3-7 in UFVA) provides the percentages of days (across all of
2005-2007, not seasonally weighted) in which the daily maximum daylight 1-hour PMio light
extinction level was greater than each of the three CPLs (excluding hours with relative humidity
greater than 90%).
From these UFVA-based displays it can be seen that among these 14 urban areas, those in
the East and in California tend to have a higher frequency of visibility conditions estimated to be
above the high CPL compared with those in the western U.S. Both Figure 4-3 and Table 4-2
indicate that all 14 urban areas have daily maximum hourly PMio light extinctions that are
estimated to exceed even the highest CPL some of the days. Except for the two Texas areas and
the non-California western urban areas, all of the other urban areas are estimated to exceed the
high CPL from about 20% to over 60% of the days. We also note that all 14 of the urban areas
are estimated to exceed the low CPL from about 40% to over 90% of the days.
Since the completion of the UFVA, EPA staff has repeated the UFVA-type modeling
based on PM2.s light extinction and data from the 2007-2009 time period for the same 15 study
areas (including St. Louis), as described in Appendix F. Figure 4-4 and Table 4-3 present the
same type of information as do Figure 4-3 and Table 4-2, respectively, except for the PM2.5 vs.
4-29
-------
distinction and the use of data from a more recent time period. While the estimates of the
percentage of daily maximum hourly PM2.5 light extinction values exceeding the CPLs are
somewhat lower than for PMio light extinction, the patterns of these estimates across the study
areas are similar. More specifically, except for the two Texas and the non-California western
urban areas, all of the other urban areas are estimated to exceed the high CPL from about 10% up
to about 50% of the days based on PM2.5 light extinction, while all 15 areas are estimated to
exceed the low CPL from over 10% to over 90% of the days.
4-30
-------
Figure 4-3. Distribution of Estimated Daily Maximum Daylight 1-hour PMi0 Light
Extinction Across the 2005-2007 Period, by Study Area (excluding hours with relative
humidity >90%). (Adapted from US EPA, 2010b, Figure 3-8)*
g _
273 143 M9 279 HI Z77
§_
o
1
i
iil
§
o
e
1
TT
* In the box-and-whlskerplot, the box represents the 25th - 75thpercentile range; the whiskers represent the 10th and90th
percentiles; individual data points below the 10th percentile and above the 90* percentile are graphed as small circles. The
three dashed horizontal lines represent the three CPL levels of 65, 110, and 190 Mm'1 (i.e., 20, 25, and30 dv).
Table 4-2. Percentage of 2005-2007 Daily Maximum Daylight Hourly Values of PM10 Light
Extinction Exceeding CPLs (excluding hours with relative humidity >90%)
(Adapted from Table 3-7 in UFVA)
Study Area
T acorn a
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days with
Estimates
109
324
300
86
306
273
148
349
279
141
277
181
143
225
Candidate Protection Level
20 dv | 25 dv
30 dv
Percentage of Daily Maximum Hourly Values Exceeding CPL
52
75
90
42
44
80
79
89
91
87
85
80
86
83
22
52
83
7
17
41
45
65
75
68
57
50
64
59
4
30
62
1
8
10
11
34
31
43
26
23
31
28
4-31
-------
Figure 4-4. Distribution of Estimated Daily Maximum Daylight 1-hour PM2.s Light
Extinction Across the 2007-2009 Period, by Study Area (excluding hours with relative
humidity >90%).*
I-
* In the box-and-whiskerplat, the box represents the 25th - 75th percentile range; the whiskers represent the 10th and 90th
percentttes; individual data points below the ltfh percentile and above the9tfh percentile are graphed as small circles. The
three dashed horizontal lines represent the three CPL levels of 65,110, and 190 Mm~'(i.e., 20, 25, and 30 dv).
Table 4-3. Percentage of 2007-2009 Daily Maximum Daylight Hourly Values of PM2.5
Light Extinction Exceeding CPLs (excluding hours with relative humidity >90%)
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days
with Estimates
150
325
161
84
276
257
144
287
330
258
133
264
140
98
145
Candidate Protection Level
20 dv
25 dv
30 dv
Percentage of Daily Maximum Hourly Values Exceeding CPL
40
66
89
14
41
72
70
78
79
85
71
77
72
92
82
14
43
78
2
15
30
25
49
51
49
49
43
33
71
50
1
22
48
0
8
5
2
17
21
12
27
15
9
34
23
4-32
-------
Visibility Levels That Just Meet Current Standards
We have also conducted analyses to assess the likelihood that PM-related visibility
impairment would exceed CPLs for a scenario based on simulating just meeting the current suite
of PM2.5 secondary standards: 15 |ig/m3 annual average PM2.5 concentration and 35 |ig/m3 24-
hour average PM2.5 concentration with a 98th percentile form, averaged over three years. As
describe in the UFVA, The steps needed to model this scenario involve explicit consideration of
changes in PM2 5 components. First, we applied proportional rollback to all the PM2 5 monitoring
sites in each study area, taking into account policy-relevant background (PRB) PM2.5 mass, to
"just meet" the NAAQS scenario for the area as a whole, not just at the visibility assessment
study site. The quantitative health risk assessment document (US EPA, 2010a) describes this air
quality roll-back procedure in detail. The degree of rollback (i.e., the percentage reduction in
non-PRB PM2 5 mass) is controlled by the highest annual or 24-hour design value, which in most
study areas is from a site other than the site used in this visibility assessment. The relevant result
from this analysis is the percentage reduction in non-PRB PM2 5 mass needed to "just meet" the
NAAQS scenario, for each study area. These percentage reductions are shown in Table 4-4 of
the UFVA. Note that Phoenix and Salt Lake City meet the current PM2.5 NAAQS under current
conditions and require no reduction. PM2 5 levels in these two cities were not "rolled up."
Second, for each day and hour for each PM2.5 component, we subtracted the PRB concentration
from the current conditions concentration to determine the non-PRB portion of the current
conditions concentration. Third, we applied the same percentage reduction from the first step to
the non-PRB portion of each of the five PM2.5 components and added back the PRB portion of
the component. Finally, we applied the IMPROVE algorithm, using the reduced PM2.5
component concentrations, the current conditions PMio-2.s concentration for the day and hour,
and relative humidity for the day and hour to calculate the PMio light extinction.
In these analyses, we have estimated both PMio and PM2 5 light extinction in terms of
both daily maximum 1-hour average values and multi-hour (i.e., 4-hour) average values for
daylight hours. Figure 4-5 and Table 4-4 display the results of the rollback procedure as a box
and whisker plot of daily maximum daylight 1-hour PMio light extinction and the percentage of
daily maximum hourly PMio light extinction values estimated to exceed the CPLs when just
meeting the current suite of PM2 5 secondary standards (excluding hours with relative humidity
greater than 90%). These displays show that the daily maximum 1-hour average values in all of
the study areas other than the three western non-California areas are estimated to exceed the high
CPL from 10% up to about 40% of the days and the middle CPL from 30% up to about 70% of
the days, while all 14 areas are estimated to exceed the low CPL from over 20% to about 90% of
the days.
4-33
-------
Figure 4-5. Distribution of Daily Maximum Daylight 1-hour PMi0 Light Extinction when
Rolled Back to Just Meet Current PM2.s NAAQS, by Study Area * (excluding hours with
relative humidity >90%)
£ 700-
§
1 «<
B
I 5(
3
400-
109 324 300 98 306 273 158 349 279 141 277 181 143 225
I I ! ii I I I !i 1 !i I
Tac Fres LA Pho SLC Dal Hou Bir Atl Det Pt Bai Phi NYC
* In the box-and-whisker plot, the box represents the 25th - 75thpercentile range; the whiskers represent the ltfh and990%). (Adapted from Table 4-7 in UFVA)
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days
with Estimates
109
324
300
98
306
273
158
349
279
141
277
181
143
225
Candidate Protection Level
20 dv
25 dv
30 dv
Percentage of Daily Maximum Hourly Values Exceeding CPL
40
53
85
44
23
79
74
83
89
80
77
76
84
76
12
30
70
7
10
43
41
55
71
61
49
49
61
49
1
10
39
1
4
10
11
24
25
33
17
20
29
19
4-34
-------
Figure 4-6 and Table 4-5 show the same type of results for PMio based on daily
maximum 4-hour average values. While the estimates of the percentage of values exceeding the
CPLs are lower than for 1-hour average light extinction, the patterns of these estimates across the
study areas are similar. More specifically, the daily maximum 4-hour average values in all of the
study areas other than the three western non-California areas and the two areas in Texas are
estimated to exceed the high CPL from about 7% up to about 20% of the days and the middle
CPL from about 20% up to over 50% of the days, while all 14 areas are estimated to exceed the
low CPL from 15% to about 80% of the days.
A similar set of figures and tables have been developed in terms of PM2.5 light extinction
for all 15 areas considered in the UFVA (including St. Louis), as shown below. Figure 4-7 and
Table 4-6 present results for PM2 5 light extinction based on daily maximum 1-hour average
values, and Figure 4-8 and Table 4-7 present results based on daily maximum 4-hour average
values. These displays show that the daily maximum 1-hour average PM2.5 light extinction
values in all of the study areas other than the three western non-California areas are estimated to
exceed the high CPL from about 8% up to over 30% of the days and the middle CPL from about
30% up to about 70% of the days, while all areas except Phoenix are estimated to exceed the low
CPL from over 15% to about 90% of the days. Further, the daily maximum 4-hour average
PM2.5 light extinction values in all of the study areas other than the three western non-California
areas and the two areas in Texas are estimated to exceed the high CPL from about 4% up to over
15% of the days and the middle CPL from about 15% up to about 45% of the days, while all
areas except Phoenix are estimated to exceed the low CPL from over 10% to about 75% of the
days.
4-35
-------
Figure 4-6. Distribution of Daily Maximum Daylight 4-hour Average PMi0 Light
Extinction when Rolled Back to Just Meet Current PM2.5 NAAQS, by Study Area*
(excluding hours with relative humidity >90%)
i
I
•t
a
107 320 298 98 302 272 156 346 274 140
179 138 220
Tac Fres LA Pho SLC Dal Hou Bit
Area
Pit Bal Phi NYC
* In the box-and-whisker plot, the box represents the 25* - 75th percentile range; the whiskers represent the 10th and 90th
percentiles; individual data points below the Iffh percentile and above the 9(fh percentile are graphed as small circles. The
three dashed horizontal lines represent the three CPL levels of 65, 110, and 190 Mm'1 (i.e., 20, 25, and30 dv).
Table 4-5. Percentage of Daily Maximum Daylight 4-Hour Average Values of PMio Light
Extinction Exceeding CPLs when Just Meeting the Current PMi.s NAAQS (excluding
hours with relative humidity >90%)
Study Area
T acorn a
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days
with Estimates
107
320
298
98
302
272
156
346
274
140
273
179
138
220
Candidate Protection Level
20 dv
25 dv
30 dv
Percentage of Daily Maximum Hourly Values Exceeding CPL
22
41
77
30
15
59
55
68
80
76
63
67
73
63
3
19
53
2
6
22
15
35
50
51
34
36
46
35
1
7
20
0
2
4
1
11
7
14
8
13
12
13
4-36
-------
Figure 4-7. Distribution of Daily Maximum Daylight 1-hour Average PM2.5 Light
Extinction when Rolled Back to Just Meet Current PM2.5 NAAQS, by Study Area*
(excluding hours with relative humidity >90%)
Tae Fres LA Pho SLC Dal Hou stL Bir All Del Pit Bal Phi NYC
* In the box-and-whisker plot, the box represents the 25th - 75thpercentile range; the whiskers represent the ltfh and 9tfh
percentiles; individual data points below the 10th percentile and above the 90th percentile are graphed as small circles. The
three dashed horizontal lines represent the three CPL levels of 65, 110, and 190 Mm'1 (i.e., 20, 25, and30 dv).
Table 4-6. Percentage of Daily Maximum Daylight 1-Hour Average Values of PM2.s
Light Extinction Exceeding CPLs when Just Meeting the Current PM2.5 NAAQS
(excluding hours with relative humidity >90%)
Study Area
T acorn a
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days
with Estimates
107
320
298
98
302
272
156
285
346
274
140
273
179
138
220
Candidate Protection Level
20 dv
25 dv
30 dv
Percentage of Daily Maximum Hourly Values Exceeding CPL
15
37
72
2
13
54
45
61
62
76
65
61
62
67
59
6
28
61
1
8
38
35
48
48
68
53
45
45
57
48
1
9
33
0
3
8
8
17
17
20
26
14
18
25
19
4-37
-------
Figure 4-8. Distribution of Daily Maximum Daylight 4-hour Average PM2.s Light
Extinction when Rolled Back to Just Meet Current PM2.5 NAAQS, by Study Area*
(excluding hours with relative humidity >90%)
t^
I
1
I,
X
§
f
a
107 320 298
302 272 156 285
346
274 140 273 179 138 220
Tac Fres LA Pho SLC Dal Hoii StL Bir Atl Del Pit Bal Phi NYC
* In the box-and-whisker plot, the box represents the 25th - 75thpercentile range; the whiskers represent the ltfh and 9tfh
percentiles; individual data points below the ltfh percentile and above the 90th percentile are graphed as small circles. The
three dashed horizontal lines represent the three CPL levels of 65, 110, and 190 Mm'1 (i.e., 20, 25, and30 dv).
Table 4-7. Percentage of Daily Maximum Daylight 4-Hour Average Values of PM2.5 Light
Extinction Exceeding CPLs when Just Meeting the Current PM2.s NAAQS (excluding
hours with relative humidity >90%)
Study Area
T acorn a
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days
with Estimates
107
320
298
98
302
272
156
285
346
274
140
273
179
138
220
Candidate Protection Level
20 dv
25 dv
30 dv
Percentage of Daily Maximum Hourly Values Exceeding CPL
15
37
72
2
13
54
45
61
62
76
65
61
62
67
59
1
16
45
0
6
18
8
30
28
42
45
30
31
39
35
1
6
16
0
2
3
1
7
7
4
10
6
10
10
12
4-38
-------
4.2.2 CASAC Advice
In our consideration of the adequacy of the current suite of PM2.5 secondary standards, in
addition to the evidence- and impact-based information discussed above, we have also
considered the advice of CASAC, based on its review of drafts of the ISA and the UFVA, and
drafts of this document, as well as comments from the public on earlier drafts of this document
and the UFVA. In its comments on the second draft PA, CASAC stated agreement with EPA
staffs conclusion that the "currently available information clearly calls into question the
adequacy of the current standards and that consideration should be given to revising the suite of
standards to provide increased public welfare protection" (Samet, 2010d, p. iii). CASAC noted
that the detailed estimates of hourly PM light extinction associated with just meeting the current
standards "clearly demonstrate that current standards do not protect against levels of visual air
quality which have been judged to be unacceptable in all of the available urban visibility
preference studies." Further, CASAC stated, with respect to the current suite of PM2.5 secondary
standards, that "[T]he levels are too high, the averaging times are too long, and the PM2.5 mass
indicator could be improved to correspond more closely to the light scattering and absorption
properties of suspended particles in the ambient air" (Samet, 2010d, p. 9).
4.2.3 Staff Conclusions on Adequacy of Current Standards
Taking into account the above considerations, EPA staff concludes that the available
information in this review, as described above and in the UFVA and ISA, clearly calls into
question the adequacy of the current suite of PM2.5 standards in the context of public welfare
protection from visibility impairment, primarily in urban areas, and supports consideration of
alternative standards to provide appropriate protection.
This conclusion is based in part on the large percentage of days, in many urban areas, that
exceed the range of CPLs identified for consideration under simulations of conditions that would
just meet the current suite of PM2.5 secondary standards. In particular, for air quality that is
simulated to just meet the current standards, for 9 of the 14 or 15 urban areas greater than 10% of
the days are estimated to exceed the highest, least protective CPL of 30 dv in terms of PMio and
PM2.5 light extinction, respectively, based on 1-hour average values, and would thus likely fail to
meet a 90th percentile-based standard at that level. For these areas, the percent of days estimated
to exceed the highest CPL ranges from approximately 10% to 40%. Similarly, when the middle
CPL of 25 dv is considered, for 11 of the 14 or 15 urban areas greater than 30% up to over 70%
of the days are estimated to exceed that CPL in terms of PMio and PM2 5 light extinction,
respectively, based on 1-hour average values. Based on a 4-hour averaging time, 5 or 6 of the
areas were estimated to have at least 10% of the days exceeding the highest CPL in terms of
PM2.s and PMio light extinction, respectively, and 8 or 11 of the areas were estimated to have
4-39
-------
greater than 30% of the days exceeding the middle CPL in terms of PM2.5 and PMio light
extinction, respectively. For the lowest CPL of 20 dv, the percentages of days estimated to
exceed that CPL are even higher for all cases considered. Based on all of the above, we
conclude that PM light extinction estimated to be associated with just meeting the current suite
of PM2.5 secondary standards in many areas across the country exceeds levels and percentages of
days that could reasonably be considered to be important from a public welfare perspective.
Further, we conclude that use of the current indicator of PM2.5 mass, in conjunction with
the current 24-hour and annual averaging times, is clearly called into question for a national
standard intended to protect public welfare from PM-related visibility impairment. This is
because such a standard is inherently confounded by regional differences in relative humidity
and species composition of PM2 5, which are critical factors in the relationship between the mix
of fine particles in the ambient air and the associated impairment of visibility. We note that this
concern was one of the important elements in the court's decision to remand the PM2.5 secondary
standards set in 2006 to the Agency, as discussed above in section 4.1.2.
Thus, beyond concluding that the available information clearly calls into question the
adequacy of the protection against PM-related visibility impairment afforded by the current suite
of PM2.5 standards, we also conclude that it clearly calls into question the appropriateness of each
of the current standard elements: indicator, averaging time, form, and level. Section 4.3 below
discusses considerations related to each of these elements in its discussion of alternative
standards for consideration.
4.3 CONSIDERATION OF ALTERNATIVE STANDARDS
Having reached the conclusion that the available information clearly calls into question
the adequacy and appropriateness of the current suite of PM2.5 standards, this section considers a
second overarching question:
What alternative fine particle standards are supported by the currently available scientific
evidence and impact-based information, as reflected in the ISA and UFVA?
In addressing this question, we have posed a series of more specific questions to inform
decisions regarding the basic elements of the NAAQS: indicator (section 4.3.1), averaging time
(section 4.3.2), form (section 4.3.3), and level (section 4.3.4). These elements are considered
collectively in evaluating the welfare protection afforded by alternative standards under
consideration.
4-40
-------
4.3.1 Indicator
4.3.1.1 Evidence-based and Impact-based Considerations
• To what extent does currently available information provide support for considering a
different indicator for PM to replace the current PMi.5 mass indicator?
As described below, EPA staff has considered three indicators: the current PM2.5 mass
indicator and two alternative indicators, including directly measured PM2.5 light extinction and
calculated PM2.5 light extinction.16 Directly measured PM2.5 light extinction is a measurement
(or combination of measurements) of the light absorption and scattering caused by PM2 5 under
ambient conditions. Calculated PM2.5 light extinction uses the IMPROVE algorithm to calculate
PM2.s light extinction using measured PM2 5 mass and/or measured PM components and relative
humidity.17
We believe that consideration of the use of either directly measured PM2 5 light extinction
or calculated PM2.s light extinction as an indicator is justified because light extinction is a
physically meaningful measure of the ambient PM2.5 characteristic that is most relevant and
directly related to PM-related visibility effects. Further, as noted in section 4.2.1 above, PM2.5 is
the component of PM responsible for most of the visibility impairment in most urban areas. In
these areas, the contribution of PMi0-2.5 is a minor contributor to visibility impairment most of
the time, although at some locations (see UFVA Figure 3-13 for Phoenix) PMio-2.s can be a
major contributor to urban visibility effects. Few urban areas conduct continuous PMi0.2.5
monitoring. For example, among the 15 urban areas featured in the UFVA, only four areas had
collocated continuous PMio data allowing calculation of hourly PMio-2.s data for 2005-2007. In
the absence of PMio-2.s air quality information from a much larger number of urban areas across
the country, it is not possible at this time to know in how many urban areas PMi0-2.5 is a major
contributor to urban visibility effects, though it is reasonable to assume that other urban areas in
the desert southwestern region of the country may have conditions similar to the conditions
shown for Phoenix. PMio-2.s is generally less homogenous in urban areas than PM2 5, making it
more challenging to select sites that would adequately represent urban visibility conditions.
While it would be possible to include a PMi0-2.5 light extinction term in a calculated light
extinction indicator, as was done in the UFVA, there is insufficient information available at this
16 In the second draft PA, this indicator was referred to as speciated PM2 5 mass calculated light extinction.
17 In 2009, the D.C. Circuit remanded the secondary PM2 5 standards to the Agency in part because EPA did not
address the problem that a PM2 5 mass-based standard using a daily averaging time would be confounded by regional
differences in relative humidity, although EPA had acknowledged this problem (as discussed above in section 4.1.2).
We note that both of the light extinction indicators considered in this assessment explicitly take into account
differences in relative humidity in areas across the country.
4-41
-------
time to assess the impact and effectiveness of such a refinement in providing public welfare
protection in areas across the country.
The basis for considering each of these three indicators is discussed below. The
discussion also addresses monitoring data requirements for directly measured PM2.5 light
extinction and for calculated PM2.5 light extinction. The following discussion also takes into
consideration different averaging times since the combination of indicator and averaging time is
relevant to understanding the monitoring data requirements. Consideration of alternative
averaging times is addressed more specifically in the next section (section 4.3.2) on averaging
time.
PMg_5 Mass Indicator
PM2 5 mass monitoring methods are in widespread use, including the Federal Reference
Method (FRM) involving the collection of periodic (usually l-day-in-6 or l-day-in-3) 24-hour
filter samples. Blank and loaded filters are weighed to determine 24-hour PM2.5 mass.
Continuous PM2.5 monitoring produces hourly average mass concentrations and is conducted at
about 900 locations. About 180 of these locations employ newer model continuous instruments
that have been approved by EPA as federal equivalent methods (FEM), although we note that
FEM approval has been based only on 24-hour average, not hourly, PM2.5 mass. These routine
monitoring activities do not include measurement of the full water content of the ambient PM2.5
that contributes, often significantly, to visibility impacts.18 Further, the PM2 5 mass concentration
monitors do not provide information on the composition of the ambient PM2.5, which plays a
central role in the relationship between PM-related visibility impairment and ambient PM2.5 mass
concentrations.19
The overall performance of 1-hour average PM2.5 mass as a predictor of PM-related
visibility impairment as indicated by PMio light extinction can be seen in scatter plots shown in
Figure 4-9 for two illustrative urban areas, Pittsburgh and Philadelphia, PA. (Similar plots for all
14 urban areas are in Appendix D, Figure D-2 of the UFVA). These illustrative examples
demonstrate the large variations in hourly PMio light extinction corresponding to any specific
level of hourly PM2.5 mass concentration as well as differences in the statistical average
relationships (depicted as the best fit lines) between cities. This poor correlation between hourly
18 FRM filters are stabilized in a laboratory at fixed temperature and relative humidity levels, which alters whatever
water content was present on the filter when removed from the sampler. FEM instruments are designed to meet
performance criteria compared to FRM measurements, and accordingly typically manage temperature and/or
humidity at the point of measurement to levels 1
19
As discussed below, 24-hour average PM2 5 c
EPA/state/local Chemical Speciation Network.
humidity at the point of measurement to levels that are not the same as ambient conditions.
19
As discussed below, 24-hour average PM2 5 chemical component mass is measured at about 200 sites in the
4-42
-------
light extinction and hourly PM2.5 mass is not due to any great extent to the contribution of
PMio-2.5 to light extinction, but rather is principally due to the impact of the water content of the
particles on light extinction, which depends on both the composition of the PM2 5 and the
ambient relative humidity. Both composition and especially relative humidity vary during a
single day, as well as from day to day, at any site and time of year. This contributes to the
noisiness of the relationship at any site and time of year. Also, there are systematic regional and
seasonal differences in the distribution of ambient humidity and PM2.5 composition conditions
that make it impossible to select a PM2.5 concentration that generally would correspond to the
same PM-related light extinction levels across all areas of the nation.
Figure 4-9. Relationship between Daylight 1-hour PM2.s Mass Concentration vs. Same-
Hour PMio Calculated Light Extinction for Two Cities
(from UFVA Appendix D, Figure D2)
1000
~ 800 -
E
™ 600 -
Q
J 400 -
s
^ 200 -
^
n_ 0 -
Pittsburgh. PA
Philadelphia. PA
0
50
100
150 200 0
50
100
150
200
PMi5 Mass Concentration
As part of the UFVA, we conducted an assessment that estimated PMio light extinction
levels that may prevail if areas were simulated to just meet a range of alternative secondary
standards based on hourly PM2.5 mass as the indicator. Appendix E contains the results of this
rollback-based assessment. This assessment quantifies the projected uneven protection, noted
qualitatively above, that would result from the use of 1-hour average PM2 5 mass as the indicator.
Directly Measured
Light Extinction Indicator
PM light extinction is the major contributor to light extinction, which is the property of
the atmosphere that is most directly related to visibility effects. It differs from light extinction by
the nearly constant contributions for Rayleigh (or clean air) light scattering and the minor
contributions by NO2 light absorption. The net result is that PM light extinction has a nearly
4-43
-------
one-to-one relationship to light extinction, unlike PM2.5 mass concentration. As explained
above, PM2.5 is the component responsible for the large majority of PMio light extinction in most
places and times. PM2 5 light extinction can be directly measured. Direct measurement of PM2 5
light extinction can be accomplished using several instrumental methods, some of which have
been used for decades to routinely monitor the two components of PM2.5 light extinction (light
scattering and absorption) or to jointly measure both as total light extinction (from which
Rayleigh scattering is subtracted to get PM2.5 light extinction). There are a number of
advantages to direct measurements of light extinction for use in a secondary standard relative to
estimates of PM2.5 light extinction calculated using PM2.5 mass and speciation data, as was done
to generate hourly light extinction values for the UFVA and as would be done for the calculated
PM2 5 light extinction indicator approach discussed below. These include greater accuracy of
direct measurements with shorter averaging times and overall greater simplicity when compared
to the need for measurements of multiple parameters to calculate PM light extinction.
As part of the UFVA, we conducted an assessment that estimated PMio light extinction
levels that may prevail in 14 urban study areas if the areas were simulated to just meet a
secondary standard based on directly measured hourly PMio light extinction as the indicator (see
section 4.3 of the UFVA).20 As would be expected, this assessment indicated that a secondary
standard based on a directly measured PMio light extinction indicator would provide the same
percentage of days having indicator values above the level of the standard in each of the areas,
with the percentage being dependent on the statistical form of the standard. We consider this
assessment reasonably informative for a directly measured PM2 5 light extinction indicator as
well, because in most of the UFVA study areas PMio light extinction is dominated by PM2 5 light
extinction.
In evaluating whether direct measurement of PM2 5 or PMio light extinction is appropriate
to consider in the context of this PM NAAQS review, EPA produced a White Paper on
Particulate Matter (PM) Light Extinction Measurements (US EPA, 201 Oh), and solicited
comment on the White Paper from the Ambient Air Monitoring and Methods Subcommittee
(AAMMS) of CAS AC. In its review of the White Paper (Russell and Samet, 2010), the
AAMMS made the recommendation to EPA that consideration of direct measurement should be
limited to PM2 5 light extinction as this can be accomplished by a number of commercially
available instruments and because PM2.5 is generally responsible for most of the PM visibility
impairment in urban areas. The AAMMS indicated that it is technically more challenging at this
time to accurately measure the PMio-2.s component of light extinction.
' This assessment was conducted prior to staff's decision to focus on PM2 5 light extinction indicators in this PA.
4-44
-------
The AAMMS not only endorsed the concept of direct measurement of PM2.5 light
extinction, but also commented on the capabilities of currently available instruments, and
expressed optimism regarding the near-term development of even better instruments for such
measurement than are now commercially available. The AAMMS advised against choosing any
currently available commercial instrument, or even a general measurement approach, as a
Federal Reference Method because to do so could discourage development of other potentially
superior approaches. Instead, the AAMMS recommended that EPA develop performance-based
approval criteria for direct measurement methods in order to put all approaches on a level
playing field. Such criteria would necessarily include procedures and pass/fail requirements for
demonstrating that the performance criteria have been met. For example, instruments might be
required to demonstrate their performance in a wind tunnel, where the concentration of PM2 5
components, and thus of PM2.5 light extinction, could be controlled to known values. It might
also be possible to devise approval testing procedures based on operation in ambient air,
although knowing the true light extinction level (without in effect treating some particular
instrument as if it were the FRM) would be more challenging. At the present time, EPA has not
undertaken to develop and test such performance-base approval criteria. The EPA anticipates
that if an effort were begun it would take at least several years before such criteria would be
ready for regulatory use.
Calculated PM^ Light Extinction Indicator
As discussed above in section 4.2.1, PM2.5 light extinction can be calculated from PM2.5
mass speciation data plus relative humidity data, as is presently routinely done on a 24-hour
average basis under the Regional Haze Program using data from the rural IMPROVE monitoring
network. This same calculation procedure, using a 24-hour average basis, could also be used for
a NAAQS focused on protecting against PM-related visibility impairment primarily in urban
areas using the type of data that is routinely collected from the urban CSN.21 This calculation
procedure, using the light extinction equations presented above in section 4.2. 1 on a 24-hour
basis, does not require PM2 5 mass concentration measurements.
Alternatively, a conceptually similar approach could be applied in urban areas on an
hourly or multi-hour basis. Applying this conceptual approach on a sub-daily basis would
involve translating 24-hour speciation data into hourly estimates of species concentrations, using
24-hour average species concentrations in conjunction with hourly PM2.5 mass concentrations.
21 About 200 sites in the EPA/state/local CSN routinely measure 24-hour average PM2 5 chemical components using
filter-based samplers and chemical analysis in a laboratory, on either a l-day-in-3 or l-day-in-6 schedule, see
Appendix B, section B. 1.3..
4-45
-------
This translation can be made using more or less complex alternative approaches, as discussed
below.
The approach used to generate hourly PMi0 light extinction for the UFVA was a
relatively more complex method for implementing such a conceptual approach. It involved the
use of the original IMPROVE algorithm22 with estimates of hourly PM2.5 components derived
from day-specific 24-hour and hourly measurements of PM2.5 mass, 24-hour measurements of
PM2.5 composition, and (for some but not all study sites) hourly PMi0-2.5 mass, along with hourly
relative humidity information (UFVA, Section 3.3). The UFVA approach also involved the use
of output from a chemical transport modeling run to provide initial estimates of diurnal profiles
for PM2.5 components at particular sites. The UFVA approach entailed numerous and complex
data processing steps to generate hourly PM2 5 composition information from these less time-
resolved data, including application of a mass-closure approach, referred to as the SANDWICH
approach (Frank, 2006), to adjust for nitrate retention differences between FRM and CSN filters,
which is a required step for consistency with the IMPROVE algorithm and for estimating organic
carbonaceous material via mass balance.23 The EPA staff employed complex custom software to
do these data processing steps. While the complexity of the approach used in the UFVA was
reasonable for assessment purposes at 14 urban areas, we recognize that a relatively more simple
approach would be more straightforward and have greater transparency, and thus should be
considered for purposes of a national standard.24 Therefore, we evaluated the degree to which
simpler approaches would correlate with the results of the highly complex method used in the
UFVA. This evaluation of two specific simpler approaches (described briefly below and in more
detail in Appendix F, especially Table F-l) demonstrated that the PM2 5 portions of the PMi0
light extinction values developed for the UFVA can be well approximated using the same
IMPROVE algorithm applied to hourly PM2.5 composition values that were much more simply
generated than with the method used in the UFVA.
The simplified approaches we examined are aimed at calculating hourly PM2.s light
extinction using the original IMPROVE algorithm (see section 4.2.1 above) excluding the
Rayleigh term for light scattering by atmospheric gases and the term for PMio-2.5-25 Initially, this
22 The original IMPROVE algorithm was selected for the described analysis in the UFVA because of its simplicity
relative to the revised algorithm.
23 Daily temperature data were also used as part of the SANDWICH method.
24 The sheer size of the ambient air quality, meteorological, and chemical transport modeling data files involved with
the UFVA approach would make it very difficult for state agencies or any interested party to consistently apply such
an approach on a routine basis for the purpose of implementing a national standard defined in terms of the UFVA
approach.
25 The original IMPROVE algorithm was the basis for the approaches considered here to maintain comparability to
the estimates developed in the UFVA. This allowed the effects of other simplifications relative to the UFVA
approach to be better discerned.
4-46
-------
leaves five PM2.5 species concentration terms (i.e., sulfate, nitrate, organic carbonaceous mass,
elemental carbon, and fine soil/crustal), plus relative humidity that need to be determined on an
hourly basis from monitoring data. For a given time period, this algorithm with its five species
concentration terms can be algebraically rearranged to a two-term algorithm (Pitchford, 2010), as
shown in the equation below. The rearranged algorithm relates PM2.5 light extinction to the
product of PM2.5 mass concentration and the sum of dry light extinction efficiency (DLEE) plus
moist light extinction efficiency (MLEE). As its name suggests, DLEE accounts for light
extinction (per microgram per cubic meter of air) that would occur if relative humidity were very
low. MLEE accounts for the increment of light extinction associated with the water content of
the PM2.5 and thus is a function of both species composition and ambient relative humidity.
PM2.5 light extinction ~ PM2.5 X (DLEE + MLEE)
where DLEE is dependent only on the relative amounts of dry PM2.5 components as measured in
CSN-type sampling, and MLEE is 3 times the hygroscopic fraction (HF) of the PM2 5 (the sulfate
plus nitrate component concentrations divided by the sum of all components) times a non-linear
function of relative humidity. The non-linear function of relative humidity is simply f(RH)-l,
where f(RH) represents the humidity adjustment factor of the original IMPROVE algorithm.
MLEE = 3 X HF X [f(RH)-l]
MLEE is zero for relative humidity below 40% and increases with increasing relative humidity
to be comparable to or greater than DLEE as relative humidity approaches 90%, depending on
theHF.
Given this algorithm structure and the goal of estimating PM2.5 light extinction for as
many 1-hour periods as possible, we focused on two alternative approaches to determining
values for DLEE and FTP for 1-hour periods that are based on using hourly measured relative
humidity. One approach (designated as "T" in Appendix F) is based on calculating 24-hour
values of DLEE and HF for each day of speciated sampling, averaging these together within each
calendar month to allow for seasonal variability, and applying the results to each daylight hour of
the respective month for which hourly PM2 5 mass data are available. A feature of this approach
is that values of DLEE and HF are available for use with hourly PM2 5 mass measurements taken
on days that were not speciated sampling days. Thus, this approach would allow the full history
of daylight hourly PM2 5 mass concentrations and relatively humidity levels to be used in
calculating PM2 5-related light extinction.
A second approach (designated as "W" in Appendix F) is based on calculating 24-hour
values of DLEE and HF for each day of speciated sampling and applying the results only to
4-47
-------
daylight hours within the respective speciated sampling day. No estimates of hourly PM2.5 light
extinction would be generated on days without speciated sampling. By recognizing that DLEE
and HF can vary from day to day within a month, this approach may provide somewhat more
accurate estimates of hourly PM2.5 light extinction for those hours for which estimates are made,
although calculations of PM2.5-related light extinction would be limited to air quality data
gathered on only a fraction of the days in each year.
A more detailed description of the sources of the data and steps required to determine
calculated PM2.5 light extinction by approaches T and W is contained in Appendix F (see Table
F-2). Also, Table F-l of Appendix F compares/contrasts each of these approaches with the
UFVA approach and with each other.
The PM2 5 light extinction values generated by using either approach T or approach W are
comparable to those developed for use in the UFVA as indicated by the regression statistics for
scatter plots of the paired data (i.e., slopes of the regression equation and R2 values are near 1 as
show in Tables F-3 and F-4 in Appendix F). Because the UFVA had no way of providing
estimates of PM2.5 light extinction on days that were not CSN sampling days, we have not
assessed how well approach T estimates PM2 5 light extinction on such days. Also, Appendix F
notes that approaches T and W both underestimate PM2.5 light extinction on some days in a few
study areas, which we attribute to the occurrence of very high nitrate concentrations and the
failure of the FRM-correlated/adjusted FEM instrument to report the entire nitrate mass.
Nevertheless, we believe that each of these simplified approaches provides reasonably good
estimates of PM2 5 light extinction and each is appropriate to consider as the indicator for a new
secondary standard. In addition, there are variations of approaches T and W that may also be
appropriate to consider. For example, some variations that may improve the correlation with
actual ambient light extinction in certain areas of the country include the use of the split mass
term approach from the revised IMPROVE algorithm,26 the use of more refined value(s) for the
organic carbon multiplier (see Appendix F), and the use of the reconstructed 24-hour PM2.5 mass
(i.e., the sum of the five PM2 5 components from speciated monitoring) as a normalization value
for the hourly measurements from the PM2 5 instrument as a way of better reflecting ambient
nitrate concentrations. Other variations may serve to simplify the calculation of PM2 5 light
extinction values, such as those suggested by CASAC for consideration, including the use of
historical monthly or seasonal speciation averages as well as speciation estimates on a regional
basis (Samet, 2010d, p. 11).
26 If the revised IMPROVE algorithm were used to define the speciated PM2 5 mass-based indicator, it would not be
possible to algebraically reduce the revised algorithm to a two-factor version as described above and in Appendix F
for approaches T and W. Instead, five component fractions would be determined from each day of speciated
sampling, and then either applied to hourly measurements of PM2 5 mass on the same day (as in approach W) or
averaged across a month and then applied to measurements of PM25 mass on each day of the month.
4-48
-------
As mentioned above, as part of the UFVA, we conducted an assessment of PMio light
extinction levels that would prevail if areas met a standard based on directly measured hourly
PMio light extinction as the indicator. This assessment indicated that a standard based on a
directly measured PMio light extinction indicator would provide the same percentage of days
having indicator values above the level of the standard across areas, with the percentage being
dependent on the statistical form of the standard. This assessment was based on the more
complex UFVA approach to estimating PMio light extinction, rather than the simpler T and W
approaches for estimating PM2.5 light extinction. Nevertheless, the generally close
correspondence between UFVA-consistent design values for PM2.5 light extinction and design
values based on approaches T and W (see Figure F-5 in Appendix F) suggest that the findings
regarding the protection offered by alternative PMio light extinction standards using a directly
measured light extinction would also hold quite well for standards based on the T or W
indicators.27 Thus, we conclude that the use of a calculated PM2.5 light extinction indicator
would provide a much higher degree of uniformity in terms of the visibility levels across the
country than is possible using PM2.5 mass as the indicator. This is due to the fact that the PM2.5
mass indicator does not account for the effects of humidity and PM2.5 composition differences
between various regions, while a calculated PM2.5 light extinction indicator directly incorporates
those effects.
The inputs that would be necessary to use either approach T or W to calculate a sub-daily
PM2.5 light extinction indicator (e.g., 1- or 4-hour averaging time) include PM2.5 chemical
speciation, relative humidity, and hourly PM2 5 mass measurements. In defining a standard in
terms of such an indicator, the criteria for allowable protocols for these inputs would need to be
specified. It would be appropriate to base these criteria on the protocols utilized in the
IMPROVE28 and CSN networks, as well as sampling and analysis protocols for ambient relative
humidity sensors, and approved FEM mass monitors for PM2.5. Any approach to approving
methods for use in calculating a light extinction indicator should take advantage of the existing
inventory of monitoring and analysis methods.
The CSN measurements have a strong history of being reviewed by CASAC technical
committees, both during their initial deployment about ten years ago,29 and during the more
27 The degree of emission reduction needed to meet a standard is tightly tied to the degree to which the design value
exceeds the level of the standard.
28 Several monitoring agencies utilize IMPROVE in urban areas to meet their chemical speciation monitoring needs.
These sites are known as IMPROVE-protocol stations.
29 Notification of a Consultation on the PM2 5 Chemical Speciation Network & Supersites Plan, EPA-SAB-CASAC-
CON-99-007; Advisory on the PM2 5 Monitoring Program, EPA-SAB-CASAC-ADV-99-002; CASAC Advisory on
the PM2 5 Monitoring Network, EPA-SAB-CASAC-ADV-00-006
4-49
-------
recent transition to carbon sampling that is consistent with the IMPROVE protocols.30 Because
the methods for the CSN are well documented in a nationally implemented Quality Assurance
Project Plan (QAPP) and accompanying SOPs, are validated through independent performance
testing, and are used to meet multiple data objectives (e.g., source apportionment, trends, and as
an input to health studies), consideration should be given to an approach that utilizes the existing
methods as the basis for criteria for allowable sampling and analysis protocols for purposes of a
calculated light extinction indicator. Such an approach of basing criteria on the current CSN and
IMPROVE methods provides a nationally consistent way to provide the chemical species data
used in the light extinction calculation, while preserving the opportunity for improved methods
for measuring the chemical species. For relative humidity, consideration should be given to
simply using criteria based on available relative humidity sensors such as already utilized by the
National Oceanic and Atmospheric Administration (NOAA) at routine weather stations. These
relative humidity sensors are already widely used by a number of monitoring agencies and can
be easily compared to other relative humidity measurements. Finally, approaches T and W
depend on having values of hourly PM2 5 mass, as discussed below.
Since 2008, EPA has approved several PM2.5 continuous mass monitoring methods as
FEMs.31 These methods have several advantages over filter-based FRMs, such as producing
hourly data and the ability to report air quality information in near real-time. However, initial
assessments of the data quality as operated by State and local monitoring agencies have mixed
results. A recent assessment of continuous FEMs and collocated FRMs conducted by EPA staff
(Hanley and Reff, 2011) found some sites and continuous FEM instruments to have an
acceptable degree of comparability of 24-hour average PM25 mass values derived from
continuous FEMs and filter-based FRMs, while others had poor data quality that would not meet
current data quality objectives. The EPA is working closely with the monitoring committee of
the National Association of Clean Air Agencies, instrument manufacturers, and monitoring
agencies to document and communicate best practices on these methods to improve quality and
consistency of resulting data. It should be noted that performance testing submitted to EPA for
purposes of designating the PM2.5 continuous methods as FEMs, and the recent assessment of
collocated FRMs and continuous FEMs, are both based on 24-hour sample periods. Therefore,
we do not have similar performance data for continuous PM2 5 FEMs for 1-hour or 4-hour
averaging periods, nor is there an accepted practice to generate performance standards for these
time periods. Until issues regarding the comparability of 24-hour PM2 5 mass values derived
30 EPA's Final Draft National Ambient Air Monitoring Strategy - An Advisory by the Ambient Air Monitoring and
Methods Subcommittee of the EPA Clean Air Scientific Advisory Committee, EPA-SAB-CASAC-05-006
31 EPA maintains a list of designated Reference and Equivalent Methods on the web at:
http://www.epa.gov/ttn/amtic/files/ambient/criteria/reference-equivalent-methods-list.pdf
4-50
-------
from continuous FEMs and filter-based FRMs are resolved, there is reason to be cautious about
relying on a calculation procedure that uses hourly PM2.5 mass values reported by continuous
FEMs and speciated mass values from 24-hour filter-based samplers. Section 4.3.2.1 discusses
another reason for such caution, based on a preliminary assessment of hourly data from
continuous FEMs.
This section has addressed the types of measurements that would be necessary to support a
calculated PM2 5 light extinction indicator for either 24-hour or sub-daily (e.g., 1-hour and 4-
hour) averaging periods. Considerations related specifically to each of these alternative
averaging times, in conjunction with a standard defined in terms of a calculated PM2.5 light
extinction indicator, are discussed below in section 4.3.2.
4.3.1.2 CASAC Advice on Indicator
Based on its review of the second draft PA, CASAC stated that it" overwhelmingly . . .
would prefer the direct measurement of light extinction," recognizing it as the property of the
atmosphere that most directly relates to visibility effects (Samet, 2010d, p. iii). CASAC noted
that "[I]t has the advantage of relating directly to the demonstrated harmful welfare effect of
ambient PM on human visual perception." However, based on CASAC's understanding of the
time that would be required to develop an FRM for this indicator, CASAC agreed with the staff
preference presented in the second draft PA for a calculated PM2.5 light extinction indicator.
CASAC noted that "[I]ts reliance on procedures that have already been implemented in the CSN
and routinely collected continuous PM2.5 data suggest that it could be implemented much sooner
than a directly measured indicator" (Samet, 2010d, p. iii).
4.3.1.3 Staff Conclusions on Indicator
Taking the above considerations and CASAC's advice into account, EPA staff concludes
that consideration should be given to establishing a new calculated PM2.5 light extinction
indicator. This conclusion takes into consideration the available evidence that demonstrates a
strong correspondence between calculated PM2 5 light extinction and PM-related visibility
impairment, as well as the significant degree of variability in visibility protection across the U.S.
allowed by a PM2.5 mass indicator. While a secondary standard that uses a PM2.5 mass indicator
could be set to provide additional protection from PM2.s-related visibility impairment, we
conclude that the advantages of using a calculated PM2 5 light extinction indicator make it the
preferred choice. In addition, We recognize that while in the future it would be appropriate to
consider a direct measurement of PM2 5 light extinction, or the sum of separate measurements of
light scattering and light absorption, as the indicator for the secondary PM2.5 standard, we
conclude this is not an appropriate option in this review because a suitable specification of the
4-51
-------
equipment and associated performance verification procedures cannot be developed in the time
frame for this review.
Further, we conclude that consideration could be given to defining a calculated PM2 5
light extinction indicator on either a 24-hour or a sub-daily basis, as discussed more fully below
in section 4.3.2. In either case, it would be appropriate to base criteria for allowable monitoring
and analysis protocols to obtain PM2.5 speciation measurements on the protocols utilized in the
IMPROVE and CSN networks. Further, in the case of a calculated PM2.5 light extinction
indicator defined on a sub-daily basis, it would be appropriate to consider using approaches T or
W, or some variations on these approaches. In reaching this conclusion, as discussed above, we
note that while it is possible to utilize data from PM2 5 continuous FEMs on a 1-hour or multi-
hour (e.g., 4-hour) basis, the mixed results of data quality assessments on a 24-hour basis, as well
as the near absence of performance data for sub-daily averaging periods, increases the
uncertainty of utilizing continuous methods to support 1-hour or 4-hour PM2 5 mass
measurements as an input to the light extinction calculation.
4.3.2 Averaging Time and Related Considerations
Consideration of appropriate averaging times for use in conjunction with a calculated
PM2.s light extinction indicator was informed by information related to the nature of PM
visibility effects, as discussed above in section 4.2.1, and the nature of inputs to the calculation
of PM2 5 light extinction, as discussed above in section 4.3.1. Based on this information, we
have considered both sub-daily (1- and 4-hour averaging times) and 24-hour averaging times, as
discussed below. In considering sub-daily averaging times, we have also addressed what diurnal
periods and ambient relative humidity conditions would be appropriate to consider in
conjunction with such an averaging time.
4.3.2.1 Sub-daily Averaging Times
• To what extent does the available information provide support for alternative sub-daily
averaging times in conjunction with a calculated PM2.s light extinction indicator?
As an initial matter, in considering sub-daily averaging times, we took into account what
we know from available studies concerning how quickly people experience and judge visibility
conditions, the possibility that some fraction of the public experience infrequent or short periods
of exposure to ambient visibility conditions, and the typical rate of change of the path averaged
PM light extinction over urban areas. While perception of change in visibility can occur in less
than a minute, meaningful changes to path-averaged light extinction occur more slowly. As
discussed above in section 4.2.1, one hour is a short enough averaging period to result in
indicator values that are close to the maximum one- or few-minute visibility impact that an
observer could be exposed to within the hour. Further, a 1-hour averaging time could reasonably
4-52
-------
characterize the visibility effects experienced by the segment of the population that experiences
infrequent short-term exposures during peak visibility impairment periods in each area/site.
Based on the above considerations, the initial analyses we conducted as part of the UFVA to
support consideration of alternative standards focused on a 1-hour averaging time.
In its review of the first draft PA, CAS AC agreed that a 1-hour averaging time would be
appropriate to consider, noting that PM effects on visibility can vary widely and rapidly over the
course of a day and such changes are almost instantaneously perceptible to human observers
(Samet, 2010c, p. 19). We note that this view related specifically to a standard defined in terms
of a directly measured PM light extinction indicator, in that CAS AC also noted that a 1-hour
averaging time is well within the instrument response times of the various currently available and
developing optical monitoring methods. CASAC also advised that if a PM2 5 mass indicator
were to be used, it would be appropriate to consider "somewhat longer averaging times - 2 to 4
hours - to assure a more stable instrumental response" (Samet, 2010c, p. 19). In considering this
advice, we conclude that since a calculated PM2.5 light extinction indicator relies in part on
measured PM2.5 mass, as discussed above in section 4.3.1, it is also appropriate to consider a
multi-hour averaging time in conjunction with such an indicator.
Thus, in preparing this final PA, we have also considered multi-hour averaging times, on
the order of a few hours as illustrated by a 4-hour averaging time. Such averaging times might
reasonably characterize the visibility effects experienced by the segment of the population who
have access to visibility conditions often or continuously throughout the day. For this segment
of the population, it may be that their perception of visual air quality reflects some degree of
offsetting an hour with poor visual air quality with one or more hours of clearer visual
conditions. Further, we recognize that a multi-hour averaging time would have the effect of
averaging away peak hourly visibility impairment, which can change significantly from one hour
to the next (see UFVA Figure 3-12). In considering either 1-hour or multi-hour averaging times,
we recognize that no data are available with regard to how the duration and variation of time a
person spends outdoors during the daytime impacts his or her judgment of the acceptability of
different degrees of visibility impairment. As a consequence, it is not clear to what degree, if at
all, the protection levels found to be acceptable in the public preference studies would change for
a multi-hour averaging time as compared to a 1-hour averaging time. Thus, we conclude that it
is appropriate to consider a 1-hour or multi-hour (e.g., 4-hour) averaging time as the basis for a
sub-daily standard defined in terms of a calculated PM2 5 light extinction indicator.
Additionally, as part of the review of data from all continuous FEM PM2.5 instruments
operating at state/local monitoring sites, as discussed above in section 4.3.1, we have recently
become aware of the occurrence of questionable outliers in 1-hour data submitted to AQS from
continuous FEM PM2.5 instruments at some of these sites (Evangelista, 2011). Some of these
4-53
-------
outliers are questionable simply by virtue of their extreme magnitude, as high as 985 |ig/m3,
whereas other values are questionable because they are isolated to single hours with much lower
values before and after, a pattern that is much less plausible than if the high concentrations were
more sustained.32 The nature and frequency of questionable 1-hour FEM data collected in the
past two years are being investigated by EPA. At this time, we note that any current data quality
problems might be resolved in the normal course of monitoring program evolution as operators
become more adept at instrument operation and maintenance and data validation or by improving
the approval criteria and testing requirements for continuous instruments. Regardless, we note
that multi-hour averaging of FEM data could serve to reduce the effects of such outliers relative
to the use of a 1-hour averaging time.
• What is an appropriate diurnal period to consider in conjunction with a sub-daily
averaging time?
In considering an appropriate diurnal period for use in conjunction with a sub-daily
averaging time, we recognize that nighttime visibility impacts, described in the ISA (section
9.2.2) are significantly different from daytime impacts and are not sufficiently well understood to
be included at this time. As a result, consistent with CAS AC advice (Samet, 2010c, p. 4), we
conclude that it would be appropriate to define a sub-daily standard in terms of only daylight
hours at this time. In the UFVA, daylight hours were defined to be those morning hours having
no minutes prior to local sunrise and afternoon hours having no minutes after local sunset. This
definition ensures the exclusion of periods of time where the sun is not the primary outdoor
source of light to illuminate scenic features.
• What ambient relative humidity conditions are appropriate to consider in conjunction
with a sub-daily averaging time?
In considering the well-known interaction of PM with ambient relative humidity
conditions, we recognize that PM is not generally the primary source of visibility impairment
during periods with fog or precipitation. In order to reduce the probability that hours with a high
degree of visibility impairment caused by fog or precipitation are unintentionally used for
purposes of determining compliance with a standard, staff determined that a relative humidity
screen that excludes daylight hours with average relative humidity above approximately 90% is
appropriate (UFVA section 3.3.5 and UFVA Appendix G). For example, for the 15 urban
32 Similarly questionable hourly data were not observed in the 2005-2007 continuous PM2 5 data used in the UFVA,
all of which came from early-generation continuous instruments that had not been approved as FEMs. However,
only 15 sites and instruments were involved in the UFVA analyses, versus about 180 currently operating FEM
instruments submitting data to AQS. Therefore, there were more opportunities for very infrequent measurement
errors to be observed in the larger FEM data set. Also, the instruments providing data used in the UFVA were a mix
of measurement approaches (e.g., TEOM® or beta attenuation), while the currently operating FEMs are mostly of
one model (beta attenuation).
4-54
-------
areas33 included in the UFVA, a 90% relative humidity cutoff criterion proved effective in that
on average less than 6% of the daylight hours were removed from consideration, yet those same
hours had on average 10 times the likelihood of rain, 6 times the likelihood of snow/sleet, and 34
times the likelihood of fog compared with hours with 90% or lower relative humidity. However,
not all periods with relative humidity above 90% have fog or precipitation. We recognize that
removing those hours from consideration involves a tradeoff between the benefits of avoiding
many of the hours with meteorological causes of visibility impacts and not counting some hours
without fog or precipitation in which high humidity levels (> 90%) lead to the growth of
hygroscopic PM to large solution droplets resulting in enhanced PM visibility impacts.
4.3.2.2 24-Hour Averaging Time
• To what extent does the available information provide support for a 24-hour averaging
time in conjunction with a calculated PMi.s light extinction indicator?
As discussed in section 4.3.1 and below, there are significant reasons to consider using
PM2.5 light extinction calculated on a 24-hour basis to reduce the various data quality concerns
over relying on continuous PM2.5 monitoring data. However, we recognize that 24 hours is far
longer than the hourly or multi-hour time periods that might reasonably characterize the visibility
effects experienced by various segments of the population, including those who have access to
visibility conditions often or continuously throughout the day, as discussed above in section
4.3.2.1. Thus, consideration of a 24-hour averaging time depends upon the extent to which PM-
related light extinction calculated on a 24-hour average basis would be a reasonable and
appropriate surrogate for PM-related light extinction calculated on a sub-daily basis, as
discussed below in this section. Further, since a 24-hour averaging time combines daytime and
nighttime periods, we recognize that the public preference studies do not directly provide a basis
for identifying an appropriate level of protection, in terms of 24-hour average light extinction,
based on judgments of acceptable daytime visual air quality obtained in those studies. Thus,
consideration of a 24-hour averaging time also depends upon developing an approach to translate
the candidate levels of protection derived from the public preference studies, which we have
interpreted on an hourly or multi-hour basis, to a candidate level of protection defined in terms of
a 24-hour average calculated light extinction, as discussed below in section 4.3.4 on level.
To determine whether PM2.5 light extinction calculated on a 24-hour basis is a reasonable
and appropriate surrogate to PM2.5 light extinction calculated on a sub-daily basis, we have done
comparative analyses of 24-hour and 4-hour averaging times in conjunction with a calculated
33 The 90% relative humidity cap assessment was conducted as part of the UFVA on all 15 of the urban areas,
including St. Louis.
4-55
-------
PM2.5 indicator.34 These analyses are presented and discussed in Appendix G, section G.4. For
these analyses, 4-hour average PM2.5 light extinction was calculated based on using the UFVA
approach. The 24-hour average PM2 5 light extinction calculations used the original IMPROVE
algorithm and long-term (1988 - 1997) average relative humidity conditions, to calculate
monthly average values of the relative humidity term in the IMPROVE algorithm, consistent
with the approach used for the Regional Haze Program. Based on these analyses, scatter plots
comparing 24-hour and 4-hour calculated PM2 5 light extinction are shown for each of the 15
cities included in the UFVA and for all 15 cities pooled together (Figures G-4 and G-5,
respectively). It can be seen, as expected, that there is some scatter around the regression line for
each city, because the calculated 4-hour light extinction includes day-specific and hour-specific
influences that are not captured by the simpler 24-hour approach. We note that this scatter could
be reduced by the use of same-day hourly relative humidity data to calculate a 24-hour average
value of the relative humidity term in the IMPROVE algorithm. Scatter plots are also shown for
the annual 90th percentile values, based on data from 2007 - 2009, for 4-hour and 24-hour
calculated PM2 5 light extinction across all 15 cities (Figure G-7) and for the 3-year design values
across all 15 cities (Figure G-8). We judge that these analyses shown 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 0.635 to over 0.8.
4.3.2.3 CASAC Advice on Averaging Time
As noted above, in its review of the first draft PA, CASAC agreed with the staff
conclusion that PM effects on visibility can vary widely and rapidly over the course of a day and
such changes are almost instantaneously perceptible to human observers (Samet, 2010c, p. 19).
Based in part on this consideration, CASAC agreed that a 1-hour averaging time would be
appropriate to consider in conjunction with a directly measured PM light extinction indicator,
noting that a 1-hour averaging time is well within the instrument response times of the various
currently available and developing optical monitoring methods. At that time, CASAC also
advised that if a PM2 5 mass indicator were to be used, it would be appropriate to consider
"somewhat longer averaging times - 2 to 4 hours - to assure a more stable instrumental
response" (Samet, 2010c, p. 19). Thus, CASAC's advice on averaging times that would be
appropriate for consideration were predicated in part on the capabilities of monitoring methods
that were available for the alternative indicators discussed in the draft PA. CAS AC's views on a
multi-hour averaging time would also apply to the calculated PM2 5 light extinction indicator
These analyses are also based on the use of a 90 percentile form, averaged over 3 years, as discussed below in
section 4.3.3 on form.
35 We note that the R-squared value (0.44) for Houston was notably lower than for the other cities.
4-56
-------
since PM2.5 mass measurements are also required for this indicator when calculated on a sub-
daily basis.
In considering this advice, we first note that CAS AC did not have the benefit of EPA's
recent assessment of the data quality issues associated with the use of continuous FEMs as the
basis for hourly PM2.5 mass measurements. We also note that since earlier drafts of this PA did
not include discussion of a calculated PM2.5 indicator based on a 24-hour averaging time,
CASAC did not have a basis to offer advice regarding a 24-hour averaging time. In addition, the
24-hour averaging time is not based on consideration of 24-hours as a relevant exposure period,
but rather as a surrogate for a sub-daily period of 4 hours, which is consistent with CASAC's
advice concerning an averaging time associated with the use of a PM2.5 mass indicator.
4.3.2.4 Staff Conclusions on Averaging Time
Taking the above considerations and CASAC's advice into account, EPA staff concludes
that it is appropriate to consider in this review a 24-hour averaging time, in conjunction with a
calculated PM2.5 light extinction indicator and an appropriately specified standard level. This
conclusion reflects the judgment that PM2.5 light extinction calculated on a 24-hour basis is a
reasonable and appropriate surrogate for sub-daily PM2.5 light extinction calculated on a 4-hour
average basis. This conclusion is also predicated on consideration of a 24-hour average standard
level, as discussed below in section 4.3.4, that is appropriately translated from the CPLs derived
from the public preference studies, which we have interpreted as providing information on the
acceptability of daytime visual air quality over an hourly or multi-hour exposure period.
A 24-hour average calculated PM2.5 light extinction indicator would avoid data quality
uncertainties that have recently been associated with currently available instruments for
measurement of hourly PM2.5 mass. The particular 24-hour indicator considered by staff uses the
original IMPROVE algorithm and long-term relative humidity conditions to calculate PM2.5 light
extinction. By using site-specific daily data on PM2.5 composition and site-specific long-term
relative humidity conditions, this 24-hour average indicator would provide more consistent
protection from PM2.5-related visibility impairment than would a secondary PM2.5 NAAQS based
only on 24-hour or annual average PM2.5 mass. In particular, this approach would account for
the systematic difference in humidity conditions between most eastern states and most western
states. Further, we have identified for consideration possible variations in the method used to
calculate PM2.5 light extinction on a 24-hour average basis that would be expected to provide
somewhat more consistent protection in areas across the country, including the use of the revised
IMPROVE algorithm and the use of same-day relative humidity data, either at the visibility
monitoring site or another site in the same area.
4-57
-------
Further, staff concludes that it would also be appropriate to consider a multi-hour, sub-
daily averaging time, for example a period of 4 hours, in conjunction with a calculated PM2.5
light extinction indicator and with further consideration of the data quality issues that have been
raised by the recent EPA study of continuous FEMs. Such an averaging time, to the extent that
data quality issues can be appropriately addressed, would be more directly related to the short-
term nature of the perception of visibility impairment, short-term variability in PM-related visual
air quality, and the short-term nature (hourly to multiple hours) of relevant exposure periods for
segments of the viewing public. Such an averaging time would also result in an indicator that is
less sensitive than a 1-hour averaging time to short-term instrument variability with respect to
PM2.5 mass measurement. In conjunction with consideration of a multi-hour, sub-daily
averaging time, we conclude that consideration should be given to including daylight hours only
and to applying a relative humidity screen of approximately 90% to remove hours in which fog
or precipitation is much more likely to contribute to the observed visibility impairment.
Recognizing that a 1-hour averaging time would be even more sensitive to data quality issues,
including short-term variability in hourly data from currently available continuous monitoring
methods, we conclude that it would not be appropriate to consider a 1-hour averaging time in
conjunction with a calculated PM2.5 light extinction indicator in this review.
4.3.3 Form
The "form" of a standard defines the air quality statistic that is to be compared to the
level of the standard in determining whether the standard is achieved. The form of the current
24-hour PM2.5 NAAQS is such that the level of the standard is compared to the 3-year average of
the annual 98th percentile of the measured indicator. The purpose in averaging for three years is
to provide stability from the occasional effects of inter-annual meteorological variability that can
result in unusually high pollution levels for a particular year that is otherwise typical. The use of
a multi-year percentile form, among other things, makes the standard less subject to the
possibility of transient violations caused by statistically unusual indicator values, thereby
providing stability to the air quality management process that may enhance the practical
effectiveness of efforts to implement the NAAQS. Also, a percentile form can be used to take
into account the number of times an exposure might occur as part of the judgment on
protectiveness in setting a NAAQS. For all of these reasons, we conclude it is appropriate to
consider defining the form of a new secondary standard in terms of a 3-year average of a
specified percentile air quality statistic.
4-58
-------
• To what extent does available information support consideration of an alternative
percentile form for a secondary standard defined in terms of a calculated PM2.s light
extinction indicator?
The urban visibility preference studies that provided results leading to the range of CPLs
being considered in this document offer no information that addresses the frequency of time that
visibility levels should be below those values. Given this lack of information, and recognizing
that the nature of the public welfare effect is one of aesthetics and/or feelings of well-being, it is
our view that it would not be appropriate to consider eliminating all exposures above the level of
the standard and that allowing some number of hours/days with reduced visibility can reasonably
be considered. In the UFVA, 90th, 95th, and 98th percentile forms were assessed for alternative
PM light extinction standards (chapter 4, UFVA). In considering these alternative percentiles,
we note that the Regional Haze Program targets the 20% most impaired days for improvements
in visual air quality in Class I areas. If improvement in the 20% most impaired days were
similarly judged to be appropriate for protecting visual air quality in urban areas, a percentile
well above the 80th percentile would be appropriate to increase the likelihood that all days in this
range would be improved by control strategies intended to attain the standard. A focus on
improving the 20% most impaired days suggests to staff that the 90th percentile, which represents
the median of the distribution of the 20% worst days, would be an appropriate form to consider.
Strategies that are implemented so that 90% of days 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% most impaired days. Higher percentile values within the range assessed could have
the effect of limiting the occurrence of days with peak PM-related light extinction in urban areas
to a greater degree. In considering the limited information available from the public preference
studies, we find no basis to conclude that it would be appropriate to consider limiting the
occurrence of days with peak PM-related light extinction in urban areas to a greater degree.
Another aspect of the form that was considered in the UFVA for a sub-daily (i.e., 1-hour)
averaging time is whether to include all daylight hours or only the maximum daily daylight hour.
This consideration would also be relevant for a multi-hour (e.g., 4-hour) averaging time,
although such an analysis was not included in the UFVA. The maximum daily daylight 1-hour
or multi-hour form is most directly protective of the welfare of people who have limited,
infrequent or intermittent exposure to visibility during the day (e.g., during commutes), but
spend most of their time without an outdoor view. For such people a view of poor visibility
during their morning commute may represent their perception of the day's visibility conditions
until the next time they venture outside during daylight, which may be hours later or perhaps the
next day. Other people have exposure to visibility conditions throughout the day. For those
4-59
-------
people, it might be more appropriate to include every daylight hour in assessing compliance with
a standard, since it is more likely that each daylight hour could affect their welfare.
We do not have information regarding the fraction of the public that has only one or a
few opportunities to experience visibility during the day, nor do we have information on the role
the duration of the observed visibility conditions has on wellbeing effects associated with those
visibility conditions. However, it is logical to conclude that people with limited opportunities to
experience visibility conditions on a daily basis would experience the entire impact associated
with visibility based on their short-term exposure. The impact of visibility for those who have
access to visibility conditions often or continuously during the day may be based on varying
conditions throughout the day.
In light of these considerations, the UFVA assessment included both the maximum daily
hour form and the all daylight hours form. We observed a close correspondence between the
level of protection afforded for all 15 urban areas in the assessment by the maximum daily
daylight 1-hour approach using the 90th percentile form and the all daylight hours approach
combined with the 98th percentile form (UFVA section 4.1.4). On this basis, we note that the
reductions in visibility impairment required to meet either form of the standard would provide
protection to both fractions of the public (i.e., those with limited opportunities and those with
greater opportunities to view PM-related visibility conditions). We note that CASAC generally
supported consideration of both types of forms without expressing a preference based on its
review of information presented in the second draft PA (Samet, 2010d, pi 1).
Staff Conclusions on Form
In conjunction with a calculated PM2 5 light extinction indicator and alternative 24-hour
or sub-daily (e.g., 4-hour) averaging times, based on the above considerations, and given the lack
of information on and the high degree of uncertainty over the impact on public welfare of the
number of days with visibility impairment over a year, we conclude that it is appropriate to give
primary consideration to a 90th percentile form, averaged over three years. Further, in the case of
a multi-hour, sub-daily alternative standard, we conclude that it is appropriate to give primary
consideration to a form based on the maximum daily multi-hour period in conjunction with the
90th percentile form. This sub-daily form would be expected to provide appropriate protection
for various segments of the population, including those with limited opportunities during a day
and those with more extended opportunities over the daylight hours to experience PM-related
visual air quality.
4.3.4 Level
In considering alternative levels for a new standard that would provide requisite
protection against PM-related visibility impairment primarily in urban areas, staff has taken into
4-60
-------
account the evidence- and impact-based considerations discussed above in section 4.2.1, with a
focus on the results of public perception and attitude surveys related to the acceptability of
various levels of visual air quality and on the important limitations in the design and scope of
such available studies. We consider this information in the context of a standard defined in
terms of a calculated PM2.5 light extinction indicator, discussed above in section 4.3.1; with
alternative averaging times of 24-hours or multi-hour, sub-daily periods (e.g., 4-hours), discussed
above in section 4.3.2; and a 90th percentile-based form, discussed above in section 4.3.3.
As part of our assessment of the adequacy of the current standards (section 4.2.1), we
interpreted the results from the visibility preferences studies conducted in four urban areas to
define a range of low, middle, and high CPLs for a sub-daily standard (e.g., 1- to 4-hour
averaging time) of 20, 25, and 30 dv, which are approximately equivalent to PM2 5 light
extinction of values of 65, 110, and 190 Mm"1. We note that CASAC agreed that this was an
appropriate range of levels to consider for such a standard (Samet, 2010d, p. 11).36 We also
recognize that to define a range of alternative levels that would be appropriate to consider for a
24-hour calculated PM2.5 light extinction standard, some adjustment to these CPLs is appropriate
since these preference studies cannot be directly interpreted as applying to a 24-hour exposure
period (as noted above in section 4.3.1). Such adjustments are more specifically discussed
below.
As an initial matter, in considering alternative levels for a sub-daily standard based
directly on the four preference study results, we note that the individual low and high CPLs are
in fact reflective of the results from the Denver and Washington, DC studies respectively, and
the middle CPL is very near to the 50th percentile criteria result from the Phoenix study. As
discussed above in section 4.2.1, we note that the Phoenix study was by far the best of the
studies, providing somewhat more support for the middle CPL. In considering the results from
these studies, we recognize that the available studies are limited in that they were conducted in
only four areas, three in the U.S. and one in Canada. Further, we recognize that available studies
provide no information on how the duration and variation of time a person spends outdoors
during the daytime may impact their judgment of the acceptability of different degrees of
visibility impairment. As such, there is a relatively high degree of uncertainty associated with
using the results of these studies to inform consideration of a national standard. Nonetheless, we
conclude, as did CASAC, that these studies are appropriate to use for this purpose.
36 In 2009, the D.C. Circuit remanded the secondary PM2 5 standards to EPA in part because the Agency failed to
identify a target level of protection, even though EPA staff and CASAC had identified a range of target levels of
protection that were appropriate for consideration. The court determined that the Agency's failure to identify a
target level of protection as part of its final decision was contrary to the statute and therefore unlawful, and that it
deprived EPA's decision-making of a reasoned basis (as discussed above in section 4.1.2).
4-61
-------
In identifying alternative levels for a 24-hour standard, we have explored various
approaches to adjusting the CPLs derived directly from the preference studies, as presented and
discussed in Appendix G, especially section G-5. These regression analyses have focused on
determining adjusted CPLs for a 24-hour standard that would provide generally equivalent
protection as that provided by a 4-hour standard with CPLs of 20, 25, and 30 dv. Some of these
approaches focused on comparing 24-hour and 4-hour light extinction values in each of the 15
urban areas assessed in the UFVA, whereas other approaches focused on comparisons based on
using aggregated data across the urban areas. Two of these approaches, which used regressions
of city-specific annual 90th percentile light extinction values or 3-year light extinction design
values, gave nearly identical results and were considered most appropriate to use as the basis for
identifying generally equivalent 24-hour adjusted CPLs. These approaches (shown in Figures G-
7 and G-8) were preferred based on the high R-squared values of the regressions and because the
regressions were determined by data from days with PM2.5 light extinction conditions in the
range of 20 to 40 dv. This contrasted with the other approaches that were highly influenced by
PM2.5 light extinction conditions well below this range. Based on these analyses and staff
conclusions presented in Appendix G, we have identified adjusted 24-hour CPLs of 21, 25, and
28 dv as being generally equivalent to 4-hour CPLs of 20, 25, and 30 dv.37 To provide some
perspective in considering these results (shown in Table G-6), we note that 1 deciview is about
the amount that persons can distinguish when viewing scenic vistas, and that a difference of 1
deciview is equivalent to about a 10% difference in light extinction expressed in Mm"1.
In more broadly considering alternative standard levels that would be appropriate for a
nationally applicable secondary standard focused on protection from PM-related urban visibility
impairment based on either a 24-hour or multi-hour, sub-daily (e.g., 4-hour) averaging time, we
are mindful of the important limitations in the available evidence from public preference studies.
We are also mindful that the scenic vistas available on a daily basis in many urban areas across
the country generally do not have the inherent visual interest or the distance between viewer and
object of greatest intrinsic value as in the Denver and Phoenix preference studies.
As in past reviews, we are considering a national visibility standard in conjunction with
the Regional Haze Program as a means of achieving appropriate levels of protection against PM-
related visibility impairment in urban, non-urban, and Class I areas across the country. We
37.As discussed in more detail in Appendix G, some days have higher values for 24-hour average light extinction
than for daily maximum 4-hour daylight light extinction, and consequently an adjusted "equivalent" 24-hour CPL
can be greater than the original 4-hour CPL. This can happen for two reasons. First, the use of monthly average
historical PJ3 data will lead to cases in which the f(RH) values used for the calculation of 24-hour average light
extinction are higher than all or some of the four hourly values of f(RH) used to determine daily maximum 4-hour
daylight light extinction on the same day. Second, PM2 5 concentrations may be greater during non-daylight periods
than during daylight hours.
4-62
-------
recognize that programs implemented to meet a national standard focused primarily on the
visibility problems in urban areas can be expected to improve visual air quality in surrounding
non-urban areas as well, as would programs now being developed to address the requirements of
the Regional Haze Program established for protection of visual air quality in Class I areas. We
also believe that the development of local programs, such as those in Denver and Phoenix, can
continue to be an effective and appropriate approach to provide additional protection, beyond
that afforded by a national standard, for unique scenic resources in and around certain urban
areas that are particularly highly valued by people living in those areas.
Staff Conclusions on Level
Based on the above considerations, we conclude that it is appropriate to give primary
consideration to alternative standard levels toward the upper end of the ranges identified above
for 24-hour and sub-daily standards, respectively. Thus, we conclude it is appropriate to
consider the following alternative levels: a level of 28 dv or somewhat below, down to 25 dv,
for a standard defined in terms of a calculated PM2.5 light extinction indicator, a 90th percentile
form, and a 24-hour averaging time; and a standard level of 30 dv or somewhat below, down to
25 dv, for a similar standard but with a 4-hour averaging time. We judge that such standards
would provide appropriate protection against PM-related visibility impairment primarily in urban
areas. We note that support for consideration of the upper part of the range of the CPLs derived
from the public preference studies was expressed by some CASAC Panel members during the
public meeting on the second draft PA. We conclude that such a standard would be appropriate
in conjunction with the Regional Haze Program to achieve appropriate levels of protection
against PM-related visibility impairment in areas across the country.
To provide some perspective on the implications of alternative averaging times, forms,
and levels for a new secondary standard, staff assessed the percentage of counties that would not
likely meet alternative 24-hour standards with a 90th percentile form and a range of levels from
25 dv to 28 dv, as well as alternative 4-hour standards with a 90th percentile form and a range of
levels from 25 dv to 30 dv. This assessment, shown in Appendix H, Tables H-l and H-2, was
not considered as a basis for the above staff conclusions. It should be noted that the geographic
coverage of this assessment was much more constrained by the availability of suitable data than
the similar assessments for the primary PM standards in Appendices C and D.
4.4 SUMMARY OF STAFF CONCLUSIONS ON SECONDARY STANDARDS FOR
VISIBILITY-RELATED EFFECTS
In reaching conclusions on the adequacy of the current suite of PM2.5 secondary standards
and potential alternative standards to provide requisite protection of PM-related visibility
4-63
-------
impairment, staff has considered these standards in terms of the basic elements of the NAAQS:
indicator, averaging time, form, and level (sections 4.3.1 to 4.3.4 above). In considering the
scientific and technical information, we reflect on the information available in the last review
integrated with information that is newly available as assessed and presented in the ISA (US
EPA, 2009a) and UFVA (US EPA, 2010a) and as summarized above and in and Appendices E,F,
and G. We also consider issues raised by the court in its remand of the secondary PM2.5
standards as discussed above in section 4.1.2.
As outlined in section 4.1.3, we emphasize a policy approach that broadens the general
approaches used in the last two PM NAAQS reviews by utilizing, to the extent available,
enhanced tools, methods, and data to more comprehensively characterize visibility impacts. As
such, we have taken into account both evidence-based and impact assessment-based
considerations to inform our conclusions related to the adequacy of the current PM2.5 secondary
standards and alternative standards that are appropriate for consideration in this review.
We recognize that selecting from among alternative standards will necessarily reflect
consideration of the qualitative and quantitative uncertainties inherent in the relevant evidence
and in the assumptions that underlie the quantitative visibility impact assessment. In reaching
staff conclusions on alternative suites of standards and ranges of levels that are appropriate to
consider, we are mindful that the CAA requires secondary standards to be set that are requisite to
protect public welfare from known and anticipated adverse effects, such that the standards are to
be neither more nor less stringent than necessary.
Based on the currently available information, staff reaches the following conclusions
regarding the secondary PM2 5 standards for protecting against PM2.5-related visibility
impairment:
(1) Consideration should be given to revising the current suite of PM2 5 secondary standards to
provide increased public welfare protection from PM2.5-related visibility impairment,
primarily in urban areas. This conclusion is based in part on the relatively large number of
days in which PM-related light extinction is estimated to exceed levels that can reasonably be
judged to be important from a public welfare perspective under simulations of conditions
associated with just meeting the current suite of PM2 5 standards. This conclusion is also
based on information that indicates that the current PM2.5 mass indicator is not appropriate
for a national standard intended to protect public welfare from PM-related visibility
impairment since such a standard is inherently confounded by regional differences in relative
humidity and species composition of PM2 5, which are critical factors in the relationship
between the mix of fine particles in the ambient air and the associated impairment of
visibility.
(2) Consideration should be given to establishing a new calculated PM2.5 light extinction
indicator. This conclusion takes into consideration the available evidence that demonstrates
a strong correspondence between PM2 5 light extinction as calculated based on the
4-64
-------
IMPROVE algorithm used in the Regional Haze Program and PM-related visibility
impairment, as well as the significant degree of variability in visibility protection across the
U.S. allowed by a PM2.5 mass indicator that does not take into account relative humidity and
PM2.5 species composition.
(a) While a secondary standard that uses a PM 2.5 mass indicator could be set to provide
additional protection from PM2.5-related visibility impairment, we conclude that the
advantages of using a calculated PM2.5 light extinction indicator make it the preferred
choice.
(b) While in the future it may be appropriate to consider a direct measurement of PM-
related light extinction, or the sum of separate measurements of light scattering and
light absorption, as the indicator for the secondary PM2.5 standard, we conclude this is
not an appropriate option in this review because a suitable specification of the
equipment and associated performance verification procedures, or suitable alternative
to these, cannot be developed in the time frame for this review.
(3) Consideration should be given to a 24-hour averaging time., in conjunction with a calculated
PM2 5 light extinction indicator and an appropriately specified standard level. This
conclusion reflects the judgment that PM2.5 light extinction calculated on a 24-hour basis is a
reasonable and appropriate surrogate for light extinction calculated on a sub-daily basis (e.g.
4 hours) that is more directly related to public perception of visual air quality. Such a
standard would avoid data quality uncertainties that have recently been associated with
currently available instruments for the measurement of hourly PM2.5 mass.
It would also be appropriate to consider a multi-hour, sub-daily averaging period, for
example a period of four hours, to the extent that data quality issues about continuous FEMs
that have recently been raised can be appropriately addressed. Such a multi-hour averaging
period would be more directly related to the short-term nature of the perception of visibility
impairment, short-term variability in PM-related visual air quality, and the short-term nature
(hourly to multiple hours) of relevant exposure periods for segments of the viewing public. It
would also be less sensitive than a 1-hour averaging time to short-term variability in PM2.5
mass measurement. In conjunction with consideration of a multi-hour, sub-daily averaging
time, we conclude that consideration should be given to including daylight hours only and to
applying a relative humidity screen of approximately 90% to remove hours in which fog or
precipitation is much more likely to contribute to the observed visibility impairment.
(4) With regard to form, in conjunction with a calculated PM2.5 light extinction indicator and
alternative 24-hour or multi-hour, sub-daily (e.g., 4-hour) averaging times, primary
consideration should be given to a 90th percentile form, averaged over three years. This form
recognizes the high degree of uncertainty over the impact on public welfare of the number of
days with visibility impairment over a year. In the case of a sub-daily alternative standard,
we conclude that it is appropriate to give primary consideration to a maximum daily value
form.
(5) With regard to level, primary consideration should be given to alternative levels toward the
upper end of the ranges identified above for 24-hour and multi-hour, sub-daily standards,
respectively, including: a level of 28 dv or somewhat below, down to 25 dv, for a standard
defined in terms of a calculated PM2.5 light extinction indicator, a 90th percentile form, and a
24-hour averaging time; and a level of 30 dv or somewhat below, down to 25 dv, for a
4-65
-------
similar standard but with a 4-hour averaging time. Staff concludes that such a standard
would be appropriate in conjunction with the Regional Haze Program to achieve appropriate
levels of protection against PM-related visibility impairment in areas across the country.
4.5 KEY UNCERTAINTIES AND AREAS FOR FUTURE RESEARCH AND DATA
COLLECTION
This section will be organized into two overarching topic areas: refining current
understanding of visibility preferences and characterization of ambient urban visibility
conditions. The first deals principally with how the public reacts to and values visibility
conditions, while the second is more concerned with determining ambient visibility conditions
and the relationships between PM component concentrations and light extinction.
Visibility Preferences
• Levels: The results of the reanalysis of the four urban preference studies (UFVA Chapter 2)
demonstrated well defined though significant statistical differences in visibility impairment
levels that divide participant decisions on acceptable from unacceptable conditions across the
study areas. A number of hypotheses concerning why the results differed for each area are
discussed in chapter 2 of the UFVA, but the current state of knowledge does not support a
definitive explanation for the range of results. A better understanding of the reasons for the
differences in preference response among the studies of the four urban areas could influence
the design of future visibility preference survey studies and the interpretation of their results
ultimately leading to a better defined range of CPLs for the next PM NAAQS review.
• Averaging Times/Forms: Additional information would also be helpful in deciding among
the various forms and averaging times to develop an effective visibility-based secondary PM
NAAQS and to assess the overall benefits of visibility improvements. Our current
understanding of urban visibility effects does not provide insights concerning:
o relative importance of degree of visibility impairment (i.e., light extinction level)
versus frequency of visibility impairment;
o strength of preference for different distributions of visibility conditions; and
o public exposure patterns.
Future research to address these deficiencies should include designing and conducting
additional preference, valuation and exposure studies to:
o expand the number and geographic coverage of urban area preference results;
o evaluate the sensitivity of results to the differences in survey study methodology;
o apply consistent methodology at multiple urban areas to better understand reasons for
preference difference among results in different urban areas;
o develop information on the strength of preference and relative importance of intensity
versus frequency of visibility impairment;
o identify the types of scenic elements that are most influential for informing public
visibility impact awareness; and
4-66
-------
o provide insights concerning visibility impact exposure duration, intensity, and timing
and their relationship to the degree and longevity of public welfare effects.
As part of the planning for additional preference and valuation survey studies, a literature review
of recent social science literature could usefully be conducted to assess the state of knowledge of
view exposure mechanisms and the psychological and behavioral effects associated with viewed
stimuli.
Urban Visibility Conditions
In this review, the paucity of light extinction monitoring data for urban areas led to the
use of the original IMPROVE algorithm to calculate hourly light extinction from continuous PM
mass and 24-hour PM2.5 component and relative humidity data (UFVA Chapter 3). The steps
used to temporally apportion 24-hour PM2.5 components in order to calculate hourly-averaged
values used monthly-averaged diurnal PM2.5 component variations from chemical transport air
quality modeling. The mass balance method used to estimate organic carbonaceous material
concentration and the loss of nitrate is reasonable but not likely to be precise. The original
IMPROVE algorithm was originally developed for remote area application to estimate 24-hour
light extinction and it was not verified for use in generating urban hourly estimates. The revised
IMPROVE algorithm would notably increase estimates of PM2 5 light extinction at the high end
of the range of light extinction conditions. Nevertheless, the resulting hourly PM2.5 light
extinction data set is thought by EPA staff to be sufficiently representative of the hourly PM2.5
light extinction levels in the study areas to allow an assessment of whether the current suite of
secondary standards is adequately protective, although these data may be biased low at high light
extinction levels, and are certainly less accurate than would be data from direct measurements.
A pilot PM2.5 light extinction monitoring program could usefully be designed and
deployed at some number of locations selected to cover a range of PM2 5 air quality conditions
with emphasis given to locations with continuous PM2 5 mass and speciation monitoring as well
as 24-hour mass and speciation sampling. Information from such a pilot monitoring program
could be used to:
o evaluate the performance of PM2 5 light extinction monitoring methods that could
ultimately be use as an FRM;
o evaluate and refine approaches for apportioning 24-hour PM2 5 species to hourly
values (needed for sites without continuous PM2.5 speciation monitoring);
o evaluate and refine light extinction calculation algorithms for use in urban settings;
and
o conduct the visibility effects assessment for the next PM secondary NAAQS.
4-67
-------
4.6 REFERENCES
Abt Associates Inc. (2001). Assessing Public Opinions on Visibility Impairment Due to Air Pollution: Summary
Report.
BBC Research & Consulting. (2003). Phoenix Area Visibility Survey. Draft Report. Available:
http://www.azdeq.gov/environ/air/download/vis 021903f.pdf. Accessed 9/16/2008
Edgerton, E. et al. (2006). The Southeastern Aerosol Research and Characterization Study, Part 3: Continuous
Measurements of Fine Paniculate Matter Mass and Composition. J. Air Waste Manage. Assoc., 56, 1325-
1341.
Evangelista, M. (2011). Investigation of 1-hour PM25 Mass Concentration Data from EPA-Approved Continuous
Federal Equivalent Method Analyzers. Memorandum to PM NAAQS review docket EPA-HQ-OAR-2007-
0492. April 5, 2011. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_td.html.
Frank, N. (2006). Retained Nitrate, Hydrated Sulfates, and Carbonaceous Mass in Federal Reference Method Fine
Paniculate Matter for Six Eastern U.S. Cities. J. Air Waste Manage. Assoc., 56, 500-511.
Hanley, T.and Reff, A (2011). Assessment of PM2 5 FEMs compared to collocated FRMs. Memorandum to PM
NAAQS review docketEPA-HQ-OAR-2007-0492. April?, 2011. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007 td.html.
Henderson, R. (2005a). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to
Honorable Stephen L. Johnson, Administrator, US EPA. CAS AC PM Review Panel's Peer Review of the
Agency's Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment
of Scientific and Technical Information (Second Draft PM Staff Paper, January 2005). June 6, 2005.
Available at: http://www.epa.gov/sab/pdf/casac-05-007.pdf.
Henderson, R. (2006a). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Recommendations Concerning the Proposed National Ambient Air Quality Standards for Particulate
Matter. March 21, 2006. Available: http://www.epa.gov/sab/pdf/casac-ltr-06-002.pdf
Ely DW; Leary JT; Stewart TR; Ross DM (1991). The establishment of the Denver Visibility Standard. Presented at
Presented at: 84th annual meeting & exhibition of the Air & Waste Management Association; June;
Vancouver, British Columbia. Pittsburgh, PA: Air & Waste Management Association; paper no. 91-48.4.
Lorang, P. (2010). Simplified Approaches for Calculation of Hourly PM25 Light Extinction Values from Hourly
PM2 5 Mass and Relative Humidity Data and 24-hour PM2 5 Composition Data. . Memorandum to PM
NAAQS review docket EPA-HQ-OAR-2007-0492.November 17, 2010. Available at:
http://www.epa.gOv/ttn/naaqs/standards/pm/s jm_2007_td.html.
Malm, W.C., Sisler, J.F., Huffman, D., Eldred, R.A., Cahill, T.A. (1994) Spatial and Seasonal Trends in Particle
Concentration and Optical Extinction in the United States, Journal of Geophysical Research (Atmospheres)
99:1347-1370.
Molenar, J.V., Malm, W.C., Johnson, C.E. (1994) Visual air quality simulation techniques. Atmos Environ, 28(5):
1055-1063.
Pitchford , M. (2010). Assessment of the Use of Speciated PM2 5 Mass-Calculated Light Extinction as a Secondary
PM NAAQS Indicator of Visibility. . Memorandum to PM NAAQS review docket EPA-HQ-OAR-2007-
0492. November 17, 2010. Available at: http://www.epa.gOv/ttn/naaqs/standards/pm/s pm 2007 td.html.
4-68
-------
Pryor SC (1996). Assessing public perception of visibility for standard setting exercises. Atmos Environ, 30: 2705-
2716.
Russell, A; Samet, J.M. (2010). Letter from the Clean Air Science Advisory Committee (CASAC) Ambient Air
Monitoring and Methods Subcommittee (AAMMS) to the Honorable Lisa P, Jackson, Administrator, US
EPA. Review of the White Paper on Particulate Matter (PM) Light Extinction Measurements. April 29,
2010, Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/92C9F5AA09A76A938
52577150004A782/$File/EP A-CASAC-10-010-unsigned.pdf
Samet J (2010c). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CASAC Review of Policy Assessmentfor the Review
of 'the PMNAAQS- First External Revew Draft (March 2010). May 17,2010. Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/E504EE3276D87A9E8
525772700647 AFB/$File/EPA-CASAC-10-011-unsigned.pdf/.
Samet J (2010d). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to the
Honorable Lisa P. Jackson, Administrator, US EPA. CASAC Review of Policy Assessment for the Review
of the PMNAAQS- Second External Revew Draft (June 2010). September 10, 2010. Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/CCF9F4C0500C500F8
525779D0073C593/$File/EP A-CASAC-10-015-unsigned.pdf
Smith, A. (2009). Comments on the First External Review Draft of EPA's "Integrated Science Assessment for
Particulate Matter." CRA International. Washington, DC. March 30. Prepared for the Utility Air
Regulatory Group. Submitted as public comments to the public meeting of the EPA Clean Air Science
Advisory Council. April 2.
Smith AE; Howell S (2009). An assessment of the robustness of visual air quality preference study results. CRA
International. Washington, DC.
http://vosemite.epa.gov/sab/sabproduct.nsf/B5591 lDF9796E5E385257592006FB737/$File/CRA+VAQ+Pr
ef+Robustness+Studv+3+30+09+final.pdf
US EPA (2005). Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA EPA-452/R-05-005a. December 2005. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr sp.html
US EPA (2008). Integrated Review Plan for the National Ambient Air Quality Standards for Particulate Matter.
National Center for Environmental Assessment and Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-08-004. March
2008. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007_pd.html.
US EPA (2009a). Integrated Science Assessment for Particulate Matter: Final Report. National Center for
Environmental Assessment-RTF Division, Office of Air Quality Planning and Standards, Research
Triangle Park, NC. EPA/600/R-08/139F. December 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007 isa.html.
US EPA (2010 a). Quantitative Risk Assessment for Particulate Matter - Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
10-005. June 2010. Available: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_risk.html
US EPA (2010b). Particulate Matter Urban-Focused Visibility Assessment - Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
10-004. June 2010. Available: http://www.epa.gOv/ttn/naaas/standards/pm/s pm 2007 risk.html.
4-69
-------
US EPA (20 lOc). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards - First External Review Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA452/P-10-003. March 2010. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007_pa.html.
US EPA (2010f). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards - Second External Review Draft. Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. EPA 452/P-10-007. June 2010.
Available at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_pa.html.
US EPA (20 lOh). White Paper on PM Light Extinction Measurements. Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. January 2010. Available
at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/823a6c8842610e76852
5764900659b22!OpenDocument
4-70
-------
5 REVIEW OF THE SECONDARY STANDARDS FOR OTHER
WELFARE EFFECTS
This chapter presents staff conclusions with regard to the current suite of secondary PM
standards to protect against PM-related welfare effects other than visibility impairment.
Specifically, staff has assessed the relevant information related to effects of atmospheric PM on
the environment, including effects on climate, ecological effects, and effects on materials. Our
assessment is framed by a series of key policy-relevant questions, which expand upon those
presented in the Integrated Review Plan (TRP) (US EPA, 2008a, section 3.2). The answers to
these questions will inform decisions on whether to retain or revise the current suite of secondary
PM standards.
In presenting staff conclusions with regard to the current secondary standards relative to
PM-related effects on climate, ecological effects, and materials, we note that the final decision is
largely a public welfare policy judgment. A final decision must draw upon scientific information
and analyses about non-visibility PM-related effects and related impacts on public welfare, as
well as judgments about how to deal with the range of uncertainties that are inherent in the
scientific evidence and analyses. Our approach to informing these judgments is discussed more
fully below. This approach is consistent with the requirements of the NAAQS provisions of the
Act and with how EPA and the courts have historically interpreted the Act. These provisions
require the Administrator to establish secondary standards that, in the Administrator's judgment,
are requisite to protect public welfare from any known or anticipated adverse effects associated
with the presence of the pollutant in the ambient air. In so doing, the Administrator seeks to
establish standards that are neither more nor less stringent than necessary for this purpose. The
Act does not require that secondary standards be set at a zero-risk level, but rather at a level that
avoids unacceptable public welfare impacts.
Information on the approaches used to set the secondary PM standards in past reviews as
well as our current approach for this review are presented in section 5.1. A discussion of the
scope of the review as related to non-visibility welfare effects of PM is included in section 5.1.2.
This chapter considers each of the non-visibility welfare effects separately. The discussion of
PM-associated effects on climate (section 5.2), ecological effects (section 5.3), and materials
(section 5.4) are each followed by a consideration of key uncertainties and areas for future
research and data collection.
5.1 APPROACH
Background information on the approaches used to establish the PM secondary standards
in 1997 and revisions to those standards in 2006 are summarized below. This section also
5-1
-------
includes a discussion of the ongoing joint review of ecological effects of oxides of nitrogen and
sulfur (NOx/SOx secondary NAAQS review) for clarity, since deposit!onal effects of PM
components of NOx and SOx to ecosystems were historically considered as a component of the
PM secondary review. Lastly, there is a discussion of the current approach for evaluating the
effects of PM on climate, ecosystems, and materials using evidence-based considerations to
inform our understanding of the key policy-relevant issues.
5.1.1 Approaches Used in Previous Reviews
5.1.1.1 Review Completed in 1997
In the 1997 review, as discussed in section 2.1.1.1, EPA determined that for the primary
standard the fine and coarse fractions of PMio should be considered separately and added a suite
of new primary standards, using PM2.s, as the indicator for fine particles, and retaining PMio as
the indicator for regulating thoracic coarse particles. The EPA established two new PM2.5
standards: an annual standard of 15 |ig/m3, based on the 3-year average of annual arithmetic
mean PM2.5 concentrations from single or multiple community-oriented monitors; and a 24-hour
standard of 65 |ig/m3, based on the 3-year average of the 98th percentile of 24-hour PM25
concentrations at each population-oriented monitor within an area (62 FR 38652, July 18, 1997).
With respect to the secondary PM standards, EPA concluded in 1997 that the available
evidence on effects of PM on non-visibility welfare endpoints was not sufficient to warrant a
separate secondary standard. Therefore, the secondary standards were set equal to the primary
PM2.5 and PMio standards in the final rule to provide protection against effects on visibility as
well as materials damage and soiling effects related to fine and coarse particles (62 FR 38683).
5.1.1.2 Review Completed in 2006
In 2006, the Administrator concluded that there was insufficient information to consider a
distinct secondary standard based on PM-related impacts to ecosystems, materials damage and
soiling, and climatic and radiative processes (71 FR 61144, October 17, 2006). Specifically,
there was a lack of evidence linking various non-visibility welfare effects to specific levels of
ambient PM. To provide a level of protection for welfare-related effects, the secondary
standards were set equal to the revised primary standards to directionally improve the level of
protection afforded vegetation, ecosystems and materials (71 FR 61210).
In the last review, the 2004 AQCD concluded that regardless of size fraction, particles
containing nitrates and sulfates have the greatest potential for widespread environmental
significance (US EPA, 2004, sections 4.2.2 and 4.2.3.1). Considerable supporting evidence was
available that indicated a significant role of NOx, SOx, and transformation products in
5-2
-------
acidification and nutrient enrichment of terrestrial and aquatic ecosystems (71 FR 61209). The
recognition of these ecological effects, coupled with other considerations detailed below, led
EPA to initiate a joint review of the NC>2 and SC>2 secondary NAAQS that will consider the
gaseous and paniculate species of NOx and SOx with respect to the ecosystem-related welfare
effects that result from the deposition of these pollutants and transformation products.
5.1.2 Scope of Current NAAQS Reviews
Non-visibility welfare-based effects of oxides of nitrogen and sulfur are divided between
two NAAQS reviews; (1) the PM NAAQS review and, (2) the joint NOx/SOx secondary
NAAQS review. The scope of each document and the components of nitrogen (N) and sulfur (S)
considered in each review are detailed in this section and summarized in Table 5-1.
5.1.2.1 Scope of the Current Secondary PM NAAQS Review
In reviewing the current suite of secondary PM standards to address visibility impairment
(chapter 4), climate forcing effects (section 5.2), and other welfare-related effects (sections 5.3
and 5.4), all PM-related effects that are not being covered in the NOx/SOx review are
considered. With regard to the materials section (5.4), the discussion has been expanded to
include particles and gases that are associated with the presence of ambient NOx and SOx, as
well as NOy, NH3 and NHX for completeness. By excluding the effects associated with
deposited particulate matter components of NOx and SOx and their transformation products
which are addressed fully in the NOx/SOx secondary review, as outlined below, the discussion
of ecological effects of PM has been narrowed to focus on effects associated with the deposition
of metals and, to a lesser extent, organics (section 5.3).
5-3
-------
Table 5-1. Scope of the Current Secondary PM NAAQS Review and Current NOx/SOx Secondary NAAQS Review
Welfare
Effect
Documents
ISA
REA
PA
Components
NOx/SOx Secondary
Review
Acidifying
deposition,
nutrient
enrichment
NOx/SOx
NOx/SOx
NOx/SOx
Deposited
particulate and
gaseous forms
of oxides of
nitrogen and
sulfur and
related N and S
containing
compounds.
Direct effects
of gas-phase
NOx/SOx on
vegetation
NOx/SOx
NOx/SOx
NOx/SOx
Gaseous forms
of oxides of
nitrogen and
sulfur and
related N and S
containing
compounds in
the ambient air.
PM Secondary
Review
Visibility
impairment
PM
PM (Urban
focused
visibility
assessment)
PM
All particles
10 microns or
smaller in the
ambient air.
Climate
Forcing effects
PM
PM
Climate-related
particles
(aerosols) in
the ambient air.
Ecological
effects
PM
PM
Deposited
components
ofPM,
including
metals and
organics but
not N and S
containing
compounds.
Materials
Damage
PMand
NOx/SOx
Annex E
PM
Particles and
gases
associated with
ambient NOx
and SOx
including NOy,
NH3 and NHx.
Soiling
PM
PM
Deposited
particles
5-4
-------
5.1.2.2 Scope of the Current NOx/SOx Secondary NAAQS Review
This is the first time since the NAAQS were established in 1971 that a joint review of the
secondary NAAQS for NOx and SOx has been conducted. This review is being conducted
because the atmospheric chemistry and environmental effects of NOx, SOx, and their associated
transformation products are linked, and because the National Research Council (NRC) has
recommended that EPA consider multiple pollutants, as appropriate, in forming the scientific
basis for the NAAQS. The NOx/SOx secondary review focuses on the welfare effects associated
with exposures from deposited particulate and gaseous forms of oxides of nitrogen and sulfur
and related N and S containing compounds and transformation products on ecosystem receptors.
An assessment of the complex ecological effects associated with N deposition requires
consideration of multiple forms of N. These include evaluation of data on inorganic reduced
forms of N (e.g., ammonia [NHa] and ammonium ion [NH4+]), inorganic oxidized forms (e.g.,
NOx, nitric acid [HNOs], nitrous oxide [TSPzO], nitrate [NO3"]), and organic N compounds (e.g.,
urea, amines, proteins, nucleic acids). In addition to acidification and N-nutrient enrichment,
other welfare effects related to deposition of N-and S-containing compounds are discussed, such
as SOx interactions with mercury (Hg) methylation. In addition, the NOx/SOx secondary review
includes evidence related to direct ecological effects of gas-phase NOx and SOx since the direct
effects of gas-phase SOx on vegetation formed a primary basis for the initial establishment of the
secondary NAAQS for SO2.
Effects of acidifying deposition associated with parti culate N and S are covered in the
recent Integrated Science Assessment for Oxides of Nitrogen and Sulfur-Ecological Criteria
(Final Report, US EPA, 2008c). The Risk and Exposure Assessment for Review of the
Secondary National Ambient Air Quality Standards for Oxides of Nitrogen and Oxides of Sulfur
(Final)(NOx/SOx REA) (US EPA, 2009h) considers four main targeted ecosystem effects
considered in the review of secondary effects of NOx and SOx: (1) aquatic acidification due to N
and S, (2) terrestrial acidification due to N and S, (3) aquatic nutrient enrichment, including
eutrophi cation and (4) terrestrial nutrient enrichment. In the Policy Assessment for Review of the
Secondary National Ambient Air Quality Standards for Oxides of Nitrogen and Oxides of Sulfur
(US EPA, 2011) an acidification index is being considered. This index provides potential
ecosystem protection from deposition related to atmospheric concentrations.
5.1.3 General Approach Used in Current Review
The remainder of this chapter summarizes and highlights key aspects of the policy
relevant information from the ISA to help inform the Administrator's judgments regarding the
adequacy of the current suite of secondary PM NAAQS in relation to climate processes,
5-5
-------
ecological effects, and materials damage. The ISA uses a five-level hierarchy that classifies the
weight of evidence for causation, not just association, into a qualitative statement about the
overall weight of evidence and causality (US EPA, 2009a, section 1.5.5, Table 1-3): causal
relationship; likely to be a causal relationship; suggestive of a causal relationship; inadequate to
infer a causal relationship; not likely to be a causal relationship (see US EPA, 2009a, Table 1-3).
Staff is evaluating evidence-based considerations primarily by assessing the evidence of
associations identified in the ISA. All relationships between PM and climate, ecological effects,
and materials damage effects identified in the ISA are considered to be either "likely causal" or
"causal". The staffs approach in this review of non-visibility welfare effects of PM is to
consider information regarding particulate matter effects on climate, ecological endpoints and
materials. This includes new literature available since the last review as well as existing,
relevant information as presented in the ISA (US EPA 2009a).
5.2 CLIMATE
5.2.1 Scope
Information and conclusions about what is currently known about the role of PM in
climate is summarized in Chapter 9 of the PM ISA (US EPA, 2009a). The ISA concludes; "that
a causal relationship exists between PM and effects on climate, including both direct effects on
radiative forcing and indirect effects that involve cloud feedbacks that influence precipitation
formation and cloud lifetimes" (US EPA, 2009a, section 9.3.10). Material from the climate
section of the ISA is principally drawn from the U.S. Climate Change Science Program
Synthesis and Assessment Product 2.3, Atmospheric Aerosol Properties and Climate Impacts, by
Chin et al., (CCSP 2009) and Chapter 2, Changes in Atmospheric Constituents and in Radiative
Forcing,(ForstQr et al., 2007) in the comprehensive Working Group I report in the Fourth
Assessment Report (AR4) from the Intergovernmental Panel on Climate Change (IPCC), Climate
Change 2007: The Physical Science Basis. Sections 9.3.7 (Fire as a Special Source of PM
Welfare Effects), 9.3.9 (Other Special Sources and Effects), 9.3.9.1 (Glaciers and Snowpack)
and 9.3.9.3 (Effects on Local and Regional Climate) of the ISA were written by NCEA staff.
This section of the PA summarizes and synthesizes the policy-relevant science in the ISA for the
purpose of helping to inform consideration of climate aspects in the review of the secondary PM
NAAQS.
Atmospheric PM (referred to as aerosols1 in the remainder of this section to be consistent
with the ISA) affects multiple aspects of climate. These include absorbing and scattering of
1 In the sections of the ISA included from IPCC AR4 and CCSP SAP2.3, 'aerosols' is more frequently used than
"PM" and that word is retained.
5-6
-------
incoming solar radiation, alterations in terrestrial radiation, effects on the hydrological cycle, and
changes in cloud properties (US EPA, 2009a, section 9.3.1). Major aerosol components that
contribute to climate processes include black carbon (BC), organic carbon (OC), sulfates, nitrates
and mineral dusts. There is a considerable ongoing research effort focused on understanding
aerosol contributions to changes in global mean temperature and precipitation patterns. The
Climate Change Research Initiative identified research on atmospheric concentrations and effects
of aerosols as a high research priority (National Research Council, 2001) and the IPCC 2007
Summary for Policymakers states that anthropogenic contributions to aerosols remain the
dominant uncertainty in radiative forcing (IPCC 2007). The current state of the science of
climate alterations attributed to PM is in flux as a result of continually updated information.
5.2.2 Adequacy of the Current Standards
In considering the adequacy of the suite of secondary standards, staff addresses the
following overarching question:
Does currently available scientific information, as reflected in the ISA, support or call into
question the adequacy of the protection for climate effects afforded by the current suite of
secondary PM standards?
To inform the answer to this overarching question, staff has posed specific questions to
aid in assessing the available scientific evidence as related to climate effects attributed to
aerosols. In considering the currently available scientific and technical information, we included
both the information available from the last review and information that is newly available since
the last review synthesized in Chapter 9 of the ISA (US EPA, 2009a).
• What new techniques are available to improve our understanding of climate effects of
aerosols?
Global climate change has increasingly been the focus of intense international research
endeavors. Major efforts are underway to understand the complexities inherent in atmospheric
aerosol interactions and to decrease uncertainties associated with climate estimations. Two recent
reports, the US CCSP Product 2.3 and sections of the IPCC AR4 were combined to form the
climate discussion in the ISA (CCSP 2009; Forster et al., 2007). A review of the most recently
available techniques for assessing climate-aerosol relationships is presented in the ISA. Aerosol
measurement capabilities reviewed in the ISA include a discussion of the increasingly
sophisticated instrumentation and techniques available for quantifying aerosols, the enhanced
sensing capabilities of satellites, development of remote sensing networks and synergy of
5-7
-------
measurements with model simulations (US EPA, 2009a, section 9.3.2). Advances in measured
aerosol properties as related to modeling as well as outstanding issues remaining in these
measurement-based studies are elaborated in the ISA (US EPA, 2009a, sections 9.3.3 and 9.3.4).
Section 9.3.6 of the ISA, "Global Aerosol Modeling" considers the capabilities of climate
modeling that have developed over the last decade and limitations of the techniques currently in
use (US EPA, 2009a).
• To what extent does newly available evidence improve our understanding of the nature
and magnitude of climate responses to PM (aerosols)?
Aerosols have direct and indirect effects on climate processes. The direct effects of
aerosols on climate result mainly from particles scattering light away from earth into space,
directly altering the radiative balance of the Earth-atmosphere system. This reflection of solar
radiation back to space decreases the transmission of visible radiation to the surface of the earth
and results in a decrease in the heating rate of the surface and the lower atmosphere. At the same
time, absorption of either incoming solar radiation or outgoing terrestrial radiation by particles,
primarily BC, results in an increased heating rate in the lower atmosphere. Global estimates of
aerosol direct radiative forcing (RF) were recently summarized using a combined model-based
estimate (Forster et al., 2007). The overall, model-derived aerosol direct RF was estimated in the
IPCC AR4 as -0.5 (-0.9 to -0.1) watts per square meter (W/m2), with an overall level of scientific
understanding of this effect as "medium low" (Forster et al., 2007), indicating a net cooling
effect in contrast to greenhouse gases (GHGs) which have a warming effect.
The contribution of individual aerosol components to total aerosol direct radiative forcing
is more uncertain than the global average (US EPA, 2009a, section 9.3.6.6). The direct effect of
radiative scattering by atmospheric particles exerts an overall net cooling of the atmosphere,
while particle absorption of solar radiation leads to warming. For example, the presence of OC
and sulfates decrease warming from sunlight by scattering shortwave radiation back into space.
Such a perturbation of incoming radiation by anthropogenic aerosols is designated as aerosol
climate forcing, which is distinguished from the aerosol radiative effect of the total aerosol
(natural plus anthropogenic). The aerosol climate forcing and radiative effect are characterized
by large spatial and temporal heterogeneities due to the wide variety of aerosol sources, the
spatial non-uniformity and intermittency of these sources, the short atmospheric lifetime of
aerosols (relative to that of the greenhouse gases), and processing (chemical and microphysical)
that occurs in the atmosphere. For example, OC can be warming (positive forcer) when
deposited on or suspended over a highly reflective surface such as snow or ice but, on a global
average, is a negative forcer in the atmosphere.
More information has also become available on indirect effects of aerosols. Particles in
the atmosphere indirectly affect both cloud albedo (reflectivity) and cloud lifetime by modifying
5-8
-------
the cloud amount, and microphysical and radiative properties (US EPA, 2009a, section 9.3.6.4).
The RF due to these indirect effects (cloud albedo effect) of aerosols is estimated in the IPCC
AR4 to be -0.7(-1.8 to -0.3) W/m2 with the level of scientific understanding of this effect as
"low" (Forster et al., 2007). Aerosols act as cloud condensation nuclei (CCN) for cloud
formation. Increased particulates in the atmosphere available as CCN with no change in
moisture content of the clouds have resulted in an increase in the number and decrease in the size
of cloud droplets in certain clouds that can increase the albedo of the clouds (the Twomey
effect). Smaller particles slow the onset of precipitation and prolong cloud lifetime. This effect,
coupled with changes in cloud albedo, increases the reflection of solar radiation back into space.
The altitude of clouds also affects cloud radiative forcing. Low clouds reflect incoming sunlight
back to space but do not effectively trap outgoing radiation, thus cooling the planet, while higher
elevation clouds reflect some sunlight but more effectively can trap outgoing radiation and act to
warm the planet (US EPA, 2009a, section 9.3.3.5).
The total negative RF due to direct and indirect effects of aerosols computed from the top
of the atmosphere, on a global average, is estimated at -1.3 (-2.2 to -0.5) W/m2 in contrast to the
positive RF of+2.9 (+3.2 to +2.6) W/m2for anthropogenic GHGs (IPCC 2007, pg. 200).
The understanding of the magnitude of aerosol effects on climate has increased
substantially in the last decade. Data on the atmospheric transport and deposition of aerosols
indicate a significant role for PM components in multiple aspects of climate. Aerosols can
impact glaciers, snowpack, regional water supplies, precipitation and climate patterns (US EPA,
2009a, section 9.3.9). Aerosols deposited on ice or snow can lead to melting and subsequent
decrease of surface albedo (US EPA 2009a, section 9.3.9.2). Aerosols are potentially important
agents of climate warming in the Arctic and other locations (US EPA, 2009a, section 9.3.9).
Incidental fires and biomass burning are being recognized as having a significant impact on
PM2.5 concentrations and climate forcing. Intermittent fires can occur at large enough scales to
affect hemispheric aerosol concentrations (US EPA 2009a, section 9.3.7).
A series of studies available since the last review examine the role of aerosols on local
and regional scale climate processes (US EPA, 2009a, section 9.3.9.3). Studies on the South
Coast Air Basin (SCAB) in California indicate aerosols may reduce near-surface wind speeds,
which, in turn reduce evaporation rates and increase cloud lifetimes. The overall impact can be a
reduction in local precipitation (Jacobson and Kaufmann, 2006). Conditions in the SCAB impact
ecologically sensitive areas including the Sierra Nevadas. Precipitation suppression due to
aerosols in California (Givati and Rosenfield, 2004) and other similar studies in Utah and
Colorado found that orographic precipitation decreased by 15-30% downwind of pollution
sources. Evidence of regional-scale impacts of aerosols on meteorological conditions in other
regions of the U.S. is lacking.
5-9
-------
• To what extent does the currently available information provide evidence of association
between specific PM constituents (i.e. BC, OC, sulfates) and climate-related effects?
Advances in the understanding of aerosol components and how they contribute to climate
change have enabled refined global forcing estimates of individual PM constituents. The global
mean radiative effect from individual components of aerosols was estimated for the first time in
the IPCC AR4 where they were reported to be (all in W/m2 units): -0.4 (±0.2) for sulfate, -0.05
(+0.05) for fossil fuel-derived OC, +0.2 (+0.15) for fossil fuel derived BC, +0.03 (+0.12) for
biomass burning, -0.1 (+0.1) for nitrates, and -0.1 (+0.2) for mineral dust (US EPA, 2009a,
section 9.3.10). Sulfate and fossil fuel-derived OC cause negative forcing whereas BC causes
positive forcing because of its highly absorbing nature (US EPA, 2009a, 9.3.6.3). Although BC
comprises only a small fraction of anthropogenic aerosol mass load and aerosol optical depth
(AOD), its forcing efficiency (with respect to either AOD or mass) is an order of magnitude
stronger than sulfate and particulate organic matter (POM), so its positive shortwave forcing
largely offsets the negative direct forcing from sulfate and POM (IPCC, 2007; US EPA 2009a,
9.3.6.3). Global loadings for nitrates and anthropogenic dust remain very difficult to estimate,
making the radiative forcing estimates for these constituents particularly uncertain (US EPA,
2009a, section 9.3.7).
Improved estimates of anthropogenic emissions of some aerosols, especially BC and OC,
have promoted the development of improved global emissions inventories and source-specific
emissions factors useful in climate modeling (Bond et al. 2004). Recent data suggests that BC is
one of the largest individual warming agents after carbon dioxide (CO2) and perhaps methane
(CH4) (Jacobson 2000; Sato et al., 2003; Bond and Sun 2005). There are several studies
modeling BC effects on climate and/or considering emission reduction measures on
anthropogenic warming detailed in section 9.3.9 of the ISA. In the U.S., most of the warming
aerosols are emitted by biomass burning and internal engine combustion and much of the cooling
aerosols are formed in the atmosphere by oxidation of SO2 or VOC's. (US EPA, 2009a, section
3.3). Fires release large amounts of BC, CO2, CH4 and OC (US EPA, 2009a, section 9.3.7).
5.2.3 Staff Conclusions
Aerosols alter climate processes directly through radiative forcing and by indirect effects
on cloud brightness, changes in precipitation and possible changes in cloud lifetimes.
• Individual components of aerosols differ in their reflective properties, and direction of
climate forcing. Overall, based on current estimates of aerosol radiative forcing,
aerosols have a net climate cooling effect.
5-10
-------
• Most climate model simulations are based on global scale scenarios. These models
may fail to consider the local variations in climate forcing due to emissions sources
and local meteorological patterns.
• Most of the warming aerosols in the U.S. are emitted by biomass burning and internal
engine combustion. Much of the cooling aerosols are formed in the atmosphere by
oxidation of SO2 or VOC's. The relative mix and sources of warming and cooling
components will vary in areas across the U.S. and over time. Thus, a set of controls
to reduce warming PM would not necessarily reduce cooling PM and vice versa.
Collectively taking into consideration the responses to specific questions regarding the
adequacy of the current secondary PM standards for climate effects, we revisit the overarching
question: "does available scientific information, as reflected in the ISA, support or call into
question the adequacy of the protection for climate effects afforded by the current suite of
secondary PM standards?" As an initial matter, we considered the appropriateness of the current
secondary standard defined in terms of PM2.5 and PMio indicators, for providing protection
against potential climate effects of aerosols. Newly available scientific information on climate-
aerosol relationships has improved our understanding of direct and indirect effects of aerosols
and aerosol properties. The major aerosol components that contribute to climate processes
include BC, OC, sulfate, nitrate and mineral dusts. These components vary in their reflectivity,
forcing efficiencies and even in the direction of climate forcing. The current standards that are
defined in terms of aggregate size mass cannot be expected to appropriately target controls on
components of fine and coarse particles that are related to climate forcing effects. Thus, the
current mass-based PM2.5 and PMio secondary standards are not an appropriate or effective
means of focusing protection against PM-associated climate effects due to these differences in
components.
Overall, there is a net climate cooling associated with aerosols in the global atmosphere
(US EPA, 2009a, section 9.2.10). Staff recognizes that some individual aerosol components,
such as BC, are positive climate forcers, whereas others, such as OC and sulfates, are negative
climate forcers. The relative mix of components will vary in areas across the U.S. and over time.
Due to 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, it is not currently feasible to conduct a
quantitative analysis for the purpose of informing revisions of the current NAAQS PM standard
based on climate. Based on these considerations, we conclude that there is insufficient
5-11
-------
information at this time to base a national ambient standard on climate impacts associated with
current ambient concentrations of PM or its constituents.2
5.2.4 Key Uncertainties and Areas for Future Research and Data Collection
Although considerable progress is being made in estimating aerosol contributions to
radiative forcing and climate fluctuations, significant uncertainties remain that preclude
consideration of climate effects as a basis for establishing a separate NAAQS secondary
standard. Further research into the effects of aerosols on climate could provide important
information to reduce these uncertainties.
A major impediment at this time to establishing a secondary standard for PM based on
climate is the lack of accurate measurement of aerosol contributions, specifically quantification
of aerosol absorption and inability to separate the anthropogenic component from total aerosol
forcing. Section 9.3.4 of the ISA details the current limitations in aerosol measurement. Most
measurement studies focus on the sum of natural and anthropogenic contributions under clear
sky conditions, however, this scenario is simplistic when effects of cloud cover and differing
reflective properties of land and ocean are considered. Satellite measurements do not currently
have the capability to distinguish anthropogenic from natural aerosols. Due to a lack of data on
the vertical distribution of aerosols, above-cloud aerosols and profiles of atmospheric radiative
heating are poorly understood (US EPA, 2009a, section 9.3.4).
Another uncertainty in considering climate effects of PM in the NAAQS review is the
spatial and temporal heterogeneity of aerosols. In regions having high concentrations of
anthropogenic aerosols, aerosol forcing is greater than the global average, and can exceed
warming by GHGs, locally reversing the sign of the forcing (US EPA, 2009a, section 9.3.1).
The contributions of policy-relevant background (PRB) concentrations to aerosol climate forcing
are not sufficiently characterized (US EPA, 2009a, section 3.7). Emissions of carbonaceous
aerosols from intermittent fires and volcanic activity can further complicate regional climate
forcing estimates (US EPA, 2009a, sections 9.3.7 and 9.3.8). Individual components of aerosols
may either be positive or negative climate forcers. Airborne PM components may be directly
emitted or undergo a variety of physical and chemical interactions and transformations. These
result in changes in particle size, structure and composition which alter aerosol reflective
properties. Aerosols can grow in size in the atmosphere because ambient water vapor condenses
on individual particles, a phenomenon known as hygroscopic growth (US EPA, 2009a, section
2 Given the reasons discussed above, this conclusion would apply for both the secondary (welfare based) and the
primary (health based) standards.
5-12
-------
9.3.6.2). Atmospheric lifetimes of individual aerosol components vary greatly confounding
tracking source receptor relationships.
Improved representation of aerosols in climate models is essential to more accurately
predict the role of PM in climate forcing (US EPA, 2009a, section 9.3.6.7). The influence of
aerosols on climate is not yet adequately taken into account in computer predictions although
considerable progress in being made in this area. For example, PM components
underrepresented or missing from many models include nitrate aerosols and anthropogenic
secondary aerosols (US EPA, 2009a, section 9.3.6.7). The modeling of aerosol indirect effects
and absorption is difficult due to the high level of uncertainty associated with these climate
factors.
The interaction of PM with clouds remains a large source of uncertainty in climate
estimates. The interactions of aerosols with clouds and linkages between clouds and the overall
climate system are complex and limit the feasibility of conducting quantitative analysis for the
purpose of establishing a secondary PM standard based on welfare effects on climate processes.
There are uncertainties associated with the potential effects of the alternative standards
for visibility discussed in Chapter 4 on regional radiative forcing and climate. A secondary
standard for visibility based on light extinction would result in reduced emissions that affect PM
in areas where monitoring shows exceedance of the standard. The extinction budget work
conducted for the UFVA (Figure 3-13, U.S. EPA, 201 Ob) and second draft PA (US EPA, 201 Of,
Appendix B) indicates that most of the current visibility impact contributions on worst days
comes from light scattering particles (e.g., nitrates, sulfates) that are negative climate forcers, and
a smaller portion comes from absorbing aerosols (e.g., black carbon) that are positive climate
forcers. The relative proportions of scattering and absorbing particles vary by location and some
major contributing emission sources contribute to both scattering and absorbing PM, so it is
unclear how the ratio of scattering to absorption might change in response to a secondary
standard for visibility affects. However, since the prevailing mixture of aerosol is thought to
have a net cooling effect on regional climate, reducing PM and light scattering aerosols could
lead to increased radiative forcing and regional climate warming while having a beneficial effect
on visibility.
5.3 ECOLOGICAL EFFECTS
5.3.1 Scope
Information on what is currently known about ecological effects of PM is summarized in
Chapter 9 of the ISA (US EPA, 2009a). Four main categories of ecological effects are identified
in the ISA: direct effects, effects of PM-altered radiative flux, indirect effects of trace metals and
indirect effects of organics. Exposure to PM for direct effects occur via deposition (e.g. wet, dry
5-13
-------
or occult) to vegetation surfaces, while indirect effects occur via deposition to ecosystem soils or
surface waters where the deposited constituents of PM then interacts with biological organisms.
Both fine and coarse-mode particles may affect plants and other organisms; however, PM size
classes do not necessarily relate to ecological effects (U.S. EPA, 1996). More often the chemical
constituents drive the ecosystem response to PM (Grantz et al., 2003). The trace metal
constituents of PM considered in the ecological effects section of the ISA are cadmium (Cd),
copper (Cu), chromium (Cr), mercury (Hg), nickel (Ni) and zinc (Zn). Ecological effects of lead
(Pb) in particulate form are covered in the Air Quality Criteria Document for Lead (US EPA,
2006). The organics included in the ecological effects section of the ISA are persistent organic
pollutants (POPs), polyaromatic hydrocarbons (PAHs) and polybromiated diphenyl ethers
(PBDEs).
Ecological effects of PM include direct effects to metabolic processes of plant foliage;
contribution to total metal loading resulting in alteration of soil biogeochemistry and
microbiology, plant and animal growth and reproduction; and contribution to total organics
loading resulting in bioaccumulation and biomagnification across trophic levels. It is important
to emphasize that the metal and organic constituents of PM contribute to total metal and organic
loads in ecosystems.
The ISA states that overall, ecological evidence is sufficient to conclude that a causal
relationship is likely to exist between deposition of PM and a variety of effects on individual
organisms and ecosystems based on information from the previous review and limited new
findings in this review (US EPA, 2009a, sections 2.5.3 and 9.4.7). However the ISA also finds,
in many cases, it is difficult to characterize the nature and magnitude of effects and to quantify
relationships between ambient concentrations of PM and ecosystem response due to significant
data gaps and uncertainties as well as considerable variability that exists in the components of
PM and their various ecological effects.
Ecological effects of PM must then be evaluated to determine if they are known or
anticipated to have an adverse impact on public welfare. Characterizing a known or anticipated
adverse effect to public welfare is an important component of developing any secondary
NAAQS. The most recent secondary NAAQS reviews have assessed changes in ecosystem
structure or processes using a weight-of-evidence approach that uses both quantitative and
qualitative data. For example, the 2008 ozone (63) final rule and 2010 63 proposal conclude that
a determination of what constitutes an "adverse" welfare effect in the context of secondary
NAAQS review can appropriately occur by considering effects at higher ecological levels
(populations, communities, ecosystems) as supported by recent literature. In the 2008
rulemaking and current ozone proposal, the interpretation of what constitutes an adverse effect
on vegetation can vary depending on the location and intended use of the plant. The degree to
5-14
-------
which (Vrelated effects are considered adverse depends on the intended use of the vegetation
and its significance to public welfare (73 FR 16496, March 27, 2008; 75 FR 2938, January 19,
2010). Therefore, effects (e.g. biomass loss, foliar injury) may be judged to have a different
degree of impact on public welfare depending, for example, on whether that effect occurs in a
Class I area, a city park, commercial cropland or private land.
A paradigm useful in evaluating ecological adversity is the concept of ecosystem
services. Ecosystem services identify the varied and numerous ways that ecosystems are
important to human welfare. Ecosystems provide many goods and services that are of vital
importance for the functioning of the biosphere and provide the basis for the delivery of tangible
benefits to human society. An EPA initiative to consider how ecosystem structure and function
can be interpreted through an ecosystem services approach has resulted in the inclusion of
ecosystem services in the NOx/SOx REA (US EPA, 2009h). The Millennium Ecosystem
Assessment (MEA) defines these to include supporting, provisioning, regulating and cultural
services (Hassan et al., 2005):
• Supporting services are necessary for the production of all other ecosystem services.
Some examples include biomass production, production of atmospheric 62, soil
formation and retention, nutrient cycling, water cycling, and provisioning of habitat.
Biodiversity is a supporting service that is increasingly recognized to sustain many of
the goods and services that humans enjoy from ecosystems. These provide a basis for
three higher-level categories of services.
• Provisioning services, such as products (Gitay et al., 2001) i.e., food (including game,
roots, seeds, nuts, and other fruit, spices, fodder), fiber (including wood, textiles), and
medicinal and cosmetic products (including aromatic plants, pigments).
• Regulating services that are of paramount importance for human society such as (a) C
sequestration, (b) climate and water regulation, (c) protection from natural hazards
such as floods, avalanches, or rock-fall, (d) water and air purification, and (e) disease
and pest regulation.
• Cultural services that satisfy human spiritual and aesthetic appreciation of ecosystems
and their components.
An important consideration in evaluating biologically adverse effects of PM and linkages
to ecosystem services is that many of the MEA categories overlap and any one pollutant may
impact multiple services. For example, deposited PM may alter the composition of soil-
associated microbial communities, which may affect supporting services such as nutrient
cycling. Changes in available soil nutrients could result in alterations to provisioning services
such as timber yield and regulating services such as climate regulation. If enough information is
5-15
-------
available, these alterations can be quantified based upon economic approaches for estimating the
value of ecosystem services. Valuation may be important from a policy perspective because it
can be used to compare the benefits of altering versus maintaining an ecosystem. Knowledge
about the relationships linking ambient concentrations and ecosystem services can be used to
inform a policy judgment on a known or anticipated adverse public welfare effect.
This policy assessment seeks to build upon and focus this body of science using the
concept of ecosystem services to qualitatively evaluate linkages between biologically adverse
effects and particulate deposition. This approach is similar to that taken in the NOx/SOx REA in
which the relationship between air quality indicators, deposition of N and S, ecologically
relevant indicators and effects on sensitive receptors are linked to changes in ecosystem structure
and services (US EPA, 2009h). This approach considers the benefits received from the resources
and processes that are supplied by ecosystems. Ecosystem components (e.g. plants, soils, water,
wildlife) are impacted by PM air pollution, which may alter the services provided by the
ecosystems in question. The goals of this policy assessment are: (1) to identify ecological effects
associated with PM deposition that can be linked to ecosystem services and (2) to qualitatively
evaluate ecological endpoints when possible. Keeping these goals and guidelines in mind,
limited new data on PM effects on plants, soil and nutrient cycling, wildlife and water are
evaluated in the context of ecosystem services to qualitatively evaluate linkages between
biologically adverse effects and particulate deposition for the purpose of evaluating the adequacy
of the current standard.
5.3.2 Adequacy of the Current Standards
In considering the adequacy of the suite of secondary standards, staff addresses the
following overarching question:
Does available scientific information, as reflected in the ISA support or call into question
the adequacy of the protection afforded by the current suite of secondary PM standards for
vegetation and ecosystems from the effects of deposited particulate metals and organics?
To inform the answer to this overarching question, staff has posed specific questions to
aid in assessing the available scientific evidence as related to ecosystem effects attributed to PM
deposition as presented in the ISA (US EPA, 2009a).
• To what extent has key scientific evidence become available to improve our
understanding of the nature and magnitude of ecosystem responses, the variability
associated with these responses, and the impact of PM on ecosystem services?
Key scientific evidence regarding PM effects on plants, soil and nutrient cycling, wildlife
and water available since the last review is summarized below to evaluate how this information
has improved our understanding of ecosystem responses to PM.
5-16
-------
Plants
As primary producers, plants play a pivotal role in energy flow through ecosystems.
Ecosystem services derived from plants include all of the categories (supporting, provisioning,
regulating, cultural) identified in the MEA (Hassan et al., 2005). Vegetation supports other
ecosystem processes by cycling nutrients through food webs and serving as a source of organic
material for soil formation and enrichment. Trees and plants provide food, wood, fiber, and fuel
for human consumption. Flora help to regulate climate by sequestering CC>2, control flooding by
stabilizing soils and cycling water via uptake and evapotranspiration. Plants are significant in
aesthetic, spiritual and recreational aspects of human interactions.
Particulate matter can adversely impact plants and ecosystem services provided by plants
by deposition to vegetative surfaces (US EPA, 2009a, section 9.4.3). Particulates deposited on
the surfaces of leaves and needles can block light, altering the radiation received by the plant.
PM deposition can obstruct stomata limiting gas exchange, damage leaf cuticles and increase
plant temperatures. This level of PM accumulation is typically observed near sources of heavy
deposition such as smelters and mining operations (US EPA, 2009a, section 9.4.3). Plants
growing on roadsides exhibit impact damage from near-road PM deposition, having higher levels
of organics and heavy metals, and accumulate salt from road de-icing during winter months (US
EPA, 2009a, sections 9.4.3.1 and 9.4.5.7).
In addition to damage to plant surfaces, deposited PM can be taken up by plants from soil
or foliage. The ability of vegetation to take up heavy metals and organics is dependent upon the
amount, solubility and chemical composition of the deposited PM. Uptake of PM by plants from
soils and vegetative surfaces can disrupt photosynthesis, alter pigments and mineral content,
reduce plant vigor, decrease frost hardiness and impair root development. The ISA indicates that
there are little or no effects on foliar processes at ambient levels of PM (US EPA, 2009a,
sections 9.4.3 and 9.4.7) however, damage due to atmospheric pollution can occur near point-
sources or under conditions where plants are subjected to multiple stressors.
Though all heavy metals can be directly toxic at sufficiently high concentrations, only
Cu, Ni, and Zn have been documented as being frequently toxic to plants (U.S. EPA, 2004),
while toxicity due to Cd, Co, and Pb has been observed less frequently (Smith, 1990; US EPA
2009a, section 9.4.5.3). In general, plant growth is negatively correlated with trace metal and
heavy metal concentration in soils and plant tissue (Audet and Charest, 2007). Trace metals,
particularly heavy metals, can influence forest growth. Growth suppression of foliar microflora
has been shown to result from Fe, Al, and Zn. These three metals can also inhibit fungal spore
formation, as can Cd, Cr, Mg, and Ni (see Smith, 1990). Metals cause stress and decreased
photosynthesis (Kucera et al., 2008) and disrupt numerous enzymes and metabolic pathways
(Strydom et al., 2006). Excessive concentrations of metals result in phytotoxicity through: (i)
5-17
-------
changes in the permeability of the cell membrane; (ii) reactions of sulfydryl (-SH) groups with
cations; (iii) affinity for reacting with phosphate groups and active groups of ADP or ATP; and
(iv) replacement of essential ions (Patra et al., 2004).
New information since the last review provides additional evidence of plant uptake of
organics (US EPA, 2009a, section 9.4.6). An area of active study is the impact of PAHs on
provisioning ecosystem services due to the potential for human and other animal exposure via
food consumption (US EPA, 2009a, section 9.4.6 page 9-190). The uptake of PAHs depends on
the plant species, site of deposition, physical and chemical properties of the organic compound
and prevailing environmental conditions. It has been established that most bioaccumulation of
PAHs by plants occurs via leaf uptake, and to a lesser extent, through roots. Differences
between species in uptake of PAHs confound attempts to quantify impacts to ecosystem
provisioning services. For example, zucchini (Cucurbitapepo) accumulated significantly more
PAHs than related plant species (Parrish et al., 2006).
Plants as ecosystem regulators can serve as passive monitors of pollution (US EPA,
2009a, section 9.4.2.3). Lichens and mosses are sensitive to pollutants associated with PM and
have been used with limited success to show spatial and temporal patterns of atmospheric
deposition of metals (US EPA, 2009a, section 9.4.2.3). For example, the presence or absence of
a specific species of lichen can be used as a bioindicator of metal or organics contamination.
PBDEs detected in moss and lichens in Antarctica indicate long-range transport of PM
components (Yogui and Sericano 2008). In the U.S. Blue Ridge Mountains, a study linked metal
concentrations in mosses to elevation and tree canopy species at some sites but not with
concentrations of metals in the O horizon of soil (Schilling, 2002). A limitation to employing
mosses and lichens to detect for the presence of air pollutants is the difference in uptake
efficiencies of metals between species. The European Moss Biomonitoring Network has been
shown to be useful in Europe for estimating general trends in metal concentrations and
identification of some sources of trace contaminants. However, quantification of ecological
effects is not possible due to the variability of species responses (US EPA, 2009a, section
9.4.2.3).
A potentially important regulating ecosystem service of plants is their capacity to
sequester contaminants (US EPA, 2009a, section 9.4.5.3). Ongoing research on the application
of plants to environmental remediation efforts are yielding some success in removing heavy
metals and organics from contaminated sites (phytoremediation) with tolerant plants such as the
willow tree (Salix spp.) and members of the family Brassicaceae (US EPA, 2009a, section
9.4.5.3). Tree canopies can be used in urban locations to capture particulates and improve air
quality (Freer-Smith et al., 2004). Plant foliage is a sink for Hg and other metals and this
regulating ecosystem service may be impacted by atmospheric deposition of trace metals.
5-18
-------
An ecological endpoint (phytochelatin concentration) associated with presence of metals
in the environment has been correlated with the ecological effect of tree mortality (Grantz et al.,
2003). Metal stress may be contributing to tree injury and forest decline in the Northeastern U.S.
where red spruce populations are declining with increasing elevation. Quantitative assessment of
PM damage to forests potentially could be conducted by overlaying PM sampling data and
elevated phytochelatin levels. However, limited data on phytochelatin levels in other species
currently hinders use of this peptide as a general biomarker for PM.
The presence of PM in the atmosphere affects ambient radiation as discussed in the ISA
which can impact the amount of sunlight received by plants (US EPA, 2009a, section 9.4.4).
Atmospheric PM can change the radiation reaching leaf surfaces through attenuation and by
converting direct radiation to diffuse radiation. Diffuse radiation is more uniformly distributed
in a tree canopy, allowing radiation to reach lower leaves. The net effect of PM on
photosynthesis depends on the reduction of photosynthetically active radiation (PAR) and the
increase in the diffuse fraction of PAR. Decreases in crop yields (provisioning ecosystem
service) have been attributed to regional scale air pollution, however, global models suggest that
the diffuse light fraction of PAR can increase growth (US EPA, 2009a, section 9.4.4).
Soil and Nutrient Cycling
Many of the major indirect plant responses to PM deposition are chiefly soil-mediated
and depend on the chemical composition of individual components of deposited PM. Major
ecosystem services impacted by PM deposition to soils include support services such as nutrient
cycling, products such as crops and regulating flooding and water quality. Upon entering the soil
environment, PM pollutants can alter ecological processes of energy flow and nutrient cycling,
inhibit nutrient uptake to plants, change microbial community structure and, affect biodiversity.
Accumulation of heavy metals in soils depends on factors such as local soil characteristics,
geologic origin of parent soils, and metal bioavailability. It can be difficult to assess the extent
to which observed heavy metal concentrations in soil are of anthropogenic origin (US EPA,
2009a, section 9.4.5.1). Trace element concentrations are higher in some soils that are remote
from air pollution sources due to parent material and local geomorphology.
Heavy metals such as Zn, Cu, and Cd and some pesticides can interfere with
microorganisms that are responsible for decomposition of soil litter, an important regulating
ecosystem service that serves as a source of soil nutrients (US EPA, 2009a, sections 9.4.5.1 and
9.4.5.2). Surface litter decomposition is reduced in soils having high metal concentrations. Soil
communities have associated bacteria, fungi, and invertebrates that are essential to soil nutrient
cycling processes. Changes to the relative species abundance and community composition can
be quantified to measure impacts of deposited PM to soil biota. A mutualistic relationship exists
5-19
-------
in the rhizophere (plant root zone) between plant roots, fungi, and microbes. Fungi in
association with plant roots form mycorrhizae that are essential for nutrient uptake by plants.
The role of mychorrizal fungi in plant uptake of metals from soils and effects of deposited PM
on soil microbes is discussed in section 9.4.5.2 of the ISA.
Wildlife
Animals play a significant role in ecosystem function including nutrient cycling and crop
production (supporting ecosystem service), and as a source of food (provisioning ecosystem
service). Cultural ecosystem services provided by wildlife include bird and animal watching,
recreational hunting and fishing. Impacts on these services are dependent upon the
bioavailability of deposited metals and organics and their respective toxicities to ecosystem
receptors. Pathways of PM exposure to fauna include ingestion, absorption and trophic transfer.
Bioindicator species (known as sentinel organisms) can provide evidence of contamination due
to atmospheric pollutants. Use of sentinel species can be of particular value because chemical
constituents of deposited PM are difficult to characterize and have varying bioavailability (US
EPA, 2009a, section 9.4.5.5). Snails readily bioaccumulate contaminants, such as PAHs and
trace metals. These organisms have been deployed as biomonitors for urban pollution and have
quantifiable biomarkers of exposure including growth inhibition, impairment of reproduction,
peroxidomal proliferation and induction of metal detoxifying proteins (metallothioneins)
(Gomet-de Vaufleury, 2002; Regoli, 2006). Earthworms have also been used as sensitive
indicators of soil metal contamination.
Evidence of deposited PM effects on animals is limited (US EPA, 2009a, section 9.4.5.5).
Trophic transfer of pollutants of atmospheric origin has been demonstrated in limited studies.
PM may also be transferred between aquatic and terrestrial compartments. There is limited
evidence for biomagnifications of heavy metals up the food chain except for Hg which is well
known to move readily through environmental compartments (US EPA, 2009a, section 9.4.5.6).
Bioconcentration of POPs and PBDEs in the Arctic and deep-water oceanic food webs indicates
the global transport of particle-associated organics (US EPA, 2009a, section 9.4.6). Salmon
migrations are contributing to metal accumulation in inland aquatic systems, potentially
impacting the provisioning and cultural ecosystem service of fishing (US EPA, 2009a, section
9.4.6). Stable isotope analysis can be applied to establish linkages between PM exposure and
impacts to food webs, however, the use of this evaluation tool is limited for this ecological
endpoint due to the complexity of most trophic interactions (US EPA 2009a, section 9.4.5.6).
Foraging cattle have been used to assess atmospheric deposition and subsequent bioaccumulation
of Hg and trace metals and their impacts on provisioning services (US EPA, 2009a, section
9.4.2.3).
5-20
-------
Water
New limited information on impacts of deposited PM on receiving water bodies indicate
that the ecosystem services of primary production, provision of fresh water, regulation of climate
and floods, recreational fishing and water purification are adversely impacted by atmospheric
inputs of metals and organics (US EPA, 2009a, sections 9.4.2.3 and 9.4.5.4). Deposition of PM
to surfaces in urban settings increases the metal and organic component of storm water runoff
(US EPA, 2009a, sections 9.4.2.3). This atmospherically-associated pollutant burden can then be
toxic to aquatic biota.
Atmospheric deposition can be the primary source of some organics and metals to
watersheds. The contribution of atmospherically deposited PAHs to aquatic food webs was
demonstrated in high elevation mountain lakes with no other anthropogenic contaminant sources
(US EPA, 2009a, section 9.4.6). Metals associated with PM deposition limit phytoplankton
growth, impacting aquatic trophic structure. Long-range atmospheric transport of 47 pesticides
and degradation products to the snowpack in seven national parks in the Western U.S. was
recently quantified indicating PM-associated contaminant inputs to receiving waters during
spring snowmelt (Hageman et al., 2006).
• What new techniques are available to improve our understanding of ecosystem effects
associated with metal and organic components of PM?
The recently completed Western Airborne Contaminants Assessment Project (WACAP)
is the most comprehensive database on contaminant transport and PM depositional effects on
sensitive ecosystems in the U.S. In this project, the transport, fate, and ecological impacts of
anthropogenic contaminants from atmospheric sources were assessed from 2002 to 2007 in seven
ecosystem components (air, snow, water, sediment, lichen, conifer needles and fish) in eight core
national parks (Landers et al., 2008). The goals of the study were to identify where the
pollutants were accumulating, identify ecological indicators for those pollutants causing
ecological harm, and to determine the source of the air masses most likely to have transported
the contaminants to the parks (US EPA, 2009a, section 9.4.6). Collected data were analyzed to
identify probable local, regional and/or global sources of deposited PM components and their
concurrent effects on ecological receptors. The study concluded that bioaccumulation of semi-
volatile organic compounds (VOCs) was observed throughout park ecosystems (Landers et al.,
2008). Findings from this study included the observation of an elevational gradient in PM
deposition with greater accumulation at higher altitude areas of the parks. Furthermore, specific
ecological indicators were identified in the WACAP that can be useful in assessing
contamination on larger spatial scales. For example, quantification of concentrations of selected
5-21
-------
pesticides in second-year conifer needles served as a method for regional-scale comparison of
pollutant distribution (Landers et al., 2008).
In the WACAP study, bioaccumulation and biomagnification of airborne contaminants
were demonstrated on a regional scale in remote ecosystems in the Western United States.
Contaminants were shown to accumulate geographically based on proximity to individual
sources or source areas, primarily agriculture and industry (Landers et al., 2008). This finding
was counter to the original working hypothesis that most of the contaminants found in western
parks would originate from Eastern Europe and Asia (Landers et al., 2008 p 6-8). The WACAP
study represents an experimental design in which ecological effects could be correlated to
ambient pollutant levels on a regional scale. Although this assessment focuses on chemical
species that are components of PM, it does not specifically assess the effects of particulates
versus gas-phase forms; therefore, in most cases it is difficult to apply the results to this
assessment based on particulate concentration and size fraction (US EPA, 2009a, section 9.4.6).
There is a need for ecological modeling of PM components in different environmental
compartments to further elucidate links between PM and ecological indicators.
Europe and other countries are using the critical load approach to assess pollutant effects
at the level of the ecosystem. This type of assessment requires site-specific data and information
on individual species responses to PM. In respect to trace metals and organics, there are
insufficient data for the vast majority of U.S. ecosystems to calculate critical loads. However, a
methodology is being presented in the NOx/SOx Secondary REA (US EPA, 2009g) to calculate
atmospheric concentrations from deposition that may be applicable to other environmental
contaminants.
• Is there currently available information on ambient levels of PM that cause adverse
effects on ecosystem components?
As reviewed above, there is considerable data on impacts of PM on ecological receptors,
but few studies that link ambient PM levels to observed effect. This is due, in part, to the nature,
deposition, transport and fate of PM in ecosystems. PM is not a single pollutant, but a
heterogeneous mixture of particles differing in size, origin and chemical composition (US EPA,
2009a, section 9.4.1). The heterogeneity of PM exists not only within individual particles or
samples from individual sites, but to even a greater extent, between samples from different sites.
Since vegetation and other ecosystem components are affected more by particulate chemistry
than size fraction, exposure to a given mass concentration of airborne PM may lead to widely
differing plant or ecosystem responses, depending on the particular mix of deposited particles.
Many of the PM components bioaccumulate over time in organisms or plants making
correlations to ambient levels of PM difficult. For example, in the WACAP study, SOC
5-22
-------
accumulation in vegetation and air showed different patterns, possibly because each medium
absorbs different types of SOCs with varying efficiencies (Landers et al., 2008).
Bioindicator organisms demonstrated biological effects including growth inhibition,
metallothionein induction and reproductive impairment when exposed to complex mixtures of
ambient air pollutants (US EPA, 2009a, section 9.4.5.5). Other studies quantify uptake of metals
and organics by plants or animals. However, due to the difficulty in correlating individual PM
components to a specific physiological response, these studies are limited. Furthermore, there
may be differences in uptake between species such as differing responses to metal uptake
observed in mosses and lichens (US EPA 2009a, section 9.4.2.3). PM may also biomagnify
across trophic levels confounding efforts to link atmospheric concentrations to physiological
endpoints (US EPA, 2009a, section 9.4.5.6).
Evidence of PM effects that are linked to a specific ecological endpoint can be observed
when ambient levels are exceeded. Most direct ecosystem effects associated with particulate
pollution occur in severely polluted areas near industrial point sources (quarries, cement kilns,
metal smelting) (US EPA, 2009a, sections 9.4.3 and 9.4.5.7). Extensive research on biota near
point sources provide some of the best evidence of ecosystem function impacts and demonstrates
that deposited PM has the potential to alter species composition over long time scales.
Ecological field studies conducted in proximity to Cu-Ni smelter in Harjavalta, Finland indicated
ecological structure and community composition are altered in response to PM and these effects
decrease with increasing distance from the point source (US EPA, 2009a, section 9.4.5.7). The
ISA indicates at 4 km distance, species composition of vegetation, insects, birds, and soil
microbiota changed, and within 1 km only the most resistant organisms were surviving (US
EPA, 2009a, section 9.4.5.7). Heavy metal concentrations were quantified in understory plant
species growing at varying distance from the Harjavalta smelter (Salemaa et al., 2004). Heavy
metal concentrations were highest in bryophytes, followed by lichens and were lowest in
vascular plants. At the Harjavalta smelter there are clear links between PM deposition levels,
ecological endpoints and compromised ecosystem structure. However, these conditions are not
reflective of ambient concentrations of PM in the majority of US ecosystems (US EPA, 2009a,
section 9.4.7).
5.3.3 Staff Conclusions
• A number of significant environmental effects that either have already occurred or are
currently occurring are linked to deposition of chemical constituents found in ambient
PM.
• Ecosystem services can be adversely impacted by PM in the environment, including
supporting, provisioning, regulating and cultural services.
5-23
-------
• The lack of sufficient information to relate specific ambient concentrations of
particulate metals and organics to a degree of impairment of a specific ecological
endpoint hinders our ability to identify a range of appropriate indicators, levels, forms
and averaging times of a distinct secondary standard to protect against associated
effects.
• Data from regionally-based ecological studies can be used to establish probable local,
regional and/or global sources of deposited PM components and their concurrent
effects on ecological receptors.
Taking into consideration the responses to specific questions regarding the adequacy of
the current secondary PM standards for ecological effects, we revisit the overarching question:
"does available scientific information, as reflected in the ISA, support or call into question the
adequacy of the protection for ecosystems afforded by the current suite of secondary PM
standards?" Staff concludes that the available information is insufficient to assess the adequacy
of the protection for ecosystems afforded by the current suite of PM secondary standards.
Ecosystem effects linked to PM are difficult to determine because the changes may not be
observed until pollutant deposition has occurred for many decades. Because the high levels
necessary to cause injury occur only near a few limited point sources and/or on a very local
scale, protection against these effects alone may not provide sufficient basis for considering a
separate secondary NAAQS based on the ecological effects of parti culate metals and organics.
Data on ecological responses clearly linked with atmospheric PM is not abundant enough to
perform a quantitative analysis although the WACAP study may represent an opportunity for
quantification at a regional scale. At this time, we conclude that available evidence is not
sufficient for establishing a distinct national standard for ambient PM based on ecosystem effects
of particulates not addressed in the NOx/SOx secondary review (e.g. metals, organics).
Staff considered the appropriateness of continuing to use the PM2.5 and PMio size
fractions as the indicators for protection of ecological effects of PM. The chemical constitution
of individual particles can be strongly correlated with size, and the relationship between particle
size and particle composition can be quite complex, making it difficult in most cases to use
particle size as a surrogate for chemistry. At this time it remains to be determined as to what
extent PM secondary standards focused on a given size fraction would result in reductions of the
ecologically relevant constituents of PM for any given area. Nonetheless, in the absence of
information that provides a basis for specific standards in terms of particle composition,
observations continue to support retaining an appropriate degree of control on both fine and
coarse particles to help address effects to ecosystems and ecosystem components associated with
PM.
5-24
-------
5.3.4 Key Uncertainties and Areas for Future Research and Data Collection
The above discussions identify linkages between ecological effects of deposited PM and
potential impacts to ecosystem services. Unfortunately, our ability to relate ambient
concentrations of PM to ecosystem response is hampered by a number of significant data gaps
and uncertainties. These limitations include the presence of multiple ecological stressors
confounding attempts to link specific ecosystem responses to PM deposition. These stressors
can be anthropogenic (e.g. habitat destruction, eutrophication, other pollutants) or natural (e.g.
drought, fire, disease). Deposited PM interacts with other stressors to affect ecosystem patterns
and processes. Furthermore, the environmental effects of deposited PM are decoupled in space
and time from the point of emission confounding efforts to identify ecological perturbations
attributed to PM deposition.
A second source of uncertainty lies in predicting the amount of PM deposited to sensitive
receptors from measured concentrations of PM in the ambient air. This makes it difficult to
relate a given air concentration to a receptor response, an important factor in being able to set a
national ambient air quality standard. A multitude of factors such as the mode of deposition
(wet, dry and occult), wind speed, surface roughness or stickiness, elevation, particle
characteristics (e.g. size, shape, chemical composition), and relative humidity exert varying
degrees of influence on the deposition velocities for different PM components in any point in
time. Composition of ambient PM varies in time and space and the particulate mixture may have
synergistic, antagonistic or additive effects on ecological receptors depending upon the chemical
species present. Furthermore, presence of co-occurring pollutants make it difficult to attribute
observed effects to ecological receptors to PM alone or one component of deposited PM.
Third, each ecosystem has developed within a context framed by the topography,
underlying bedrock, soils, climate, meteorology, hydrologic regime, natural and land use history,
and species composition that make it unique from all others. Sensitivity of ecosystem response
is highly variable in space and time. Because of this variety and lack of sufficient baseline data
on each of these features for most ecosystems, it is currently not possible to extrapolate with
confidence any effect from one ecosystem to another. Further research is needed to decrease the
uncertainties associated with ambient PM effects on ecosystems and ecosystem components.
5.4 MATERIALS
5.4.1 Scope
Welfare effects on materials associated with deposition of PM include both physical
damage (materials damage effects) and impaired aesthetic qualities (soiling effects). Because the
effects of PM are exacerbated by the presence of acidic gases and can be additive or synergistic
due to the complex mixture of pollutants in the air and surface characteristics of the material, this
5-25
-------
discussion will also include those particles and gases that are associated with the presence of
ambient NOx and SOx, as well as NHa and NHx for completeness. Building upon the
information presented in the last Staff Paper (US EPA, 2005), and including the limited new
information presented in Chapter 9 of the PM ISA (US EPA, 2009a) and Annex E. Effects of
NOy, NHx, and SOx on Structures and Materials of the Integrated Science Assessment for
Oxides of Nitrogen and Sulfur-Ecological Criteria (NOx/SOx ISA) (US EPA, 2008c) the
following sections consider the policy-relevant aspects of physical damage and aesthetic soiling
effects of PM on materials including metal and stone.
The ISA concludes that evidence is sufficient to support a causal relationship between
PM and effects on materials (US EPA, 2009a, sections 2.5.4 and 9.5.4). The deposition of PM
can physically affect materials, adding to the effects of natural weathering processes, by
potentially promoting or accelerating the corrosion of metals, by degrading paints and by
deteriorating building materials such as stone, concrete and marble (US EPA, 2009a, section
9.5). Particles contribute to these physical effects because of their electrolytic, hygroscopic and
acidic properties, and their ability to sorb corrosive gases (principally SO2). In addition, the
deposition of ambient PM can reduce the aesthetic appeal of buildings and objects through
soiling. Particles consisting primarily of carbonaceous compounds cause soiling of commonly
used building materials and culturally important items such as statues and works of art. Soiling
is the deposition of particles on surfaces by impingement, and the accumulation of particles on
the surface of an exposed material results in degradation of its appearance (US EPA, 2009a,
section 9.5). Soiling can be remedied by cleaning or washing, and depending on the soiled
material, repainting.
5.4.2 Adequacy of the Current Standards
In considering the adequacy of the suite of secondary standards, staff addresses the
following overarching question:
Does available scientific information, as reflected in the ISA support or call into question
the adequacy of the protection for materials afforded by the current suite of secondary PM
standards?
To inform the answer to this overarching question, staff has posed a specific question to
aid in assessing the available scientific evidence as related to materials damage and soiling
attributed to PM deposition as presented in the ISA (US EPA, 2009a).
• What new evidence is available to improve our understanding of effects of PM on
materials and linking ambient concentrations to materials damage?
The majority of available new studies on materials effects of PM are from outside the
U.S., however, they provide limited new data for consideration of the secondary standard.
5-26
-------
Metal and stone are susceptible to damage by ambient PM. Considerable research has
been conducted on the effects of air pollutants on metal surfaces due to the economic importance
of these materials, especially steel, zinc, aluminum, and copper. Chapter 9 of the PM ISA and
Annex E of the NOx/SOx ISA summarize the results of a number of studies on the corrosion of
metals (US EPA, 2009a; US EPA, 2008c). Moisture is the single greatest factor promoting metal
corrosion, however, deposited PM can have additive, antagonistic or synergistic effects. In
general, SO2 is more corrosive than NOx although mixtures of NOx, SO2 and other particulate
matter corrode some metals at a faster rate than either pollutant alone (US EPA, 2008c, Annex
E.5.2). Information from both the PM ISA and NOx/SOx ISA suggest that the extent of damage
to metals due to ambient PM is variable and dependent upon the type of metal, prevailing
environmental conditions, rate of natural weathering and presence or absence of other pollutants.
The PM ISA and NOx/SOx ISA summarize the results of a number of studies on PM and
stone surfaces. While it is clear from the available information that gaseous air pollutants, in
particular SO2, will promote the deterioration of some types of stones under specific conditions,
carbonaceous particles (non-carbonate carbon) and particles containing metal oxides may help to
promote the decay process. Studies on metal and stone summarized in the ISA do not show an
association between particle size, chemical composition and frequency of repair.
A limited number of new studies available on materials damage effects of PM since the
last review consider the relationship between pollutants and biodeterioration of structures
associated with microbial communities that colonize monuments and buildings (US EPA 2009a,
section 9.5). Presence of air pollutants may synergistically enhance microbial deterioration
processes. The role of heterotrophic bacteria, fungi and cyanobacteria in biodeterioration varied
by local meterological conditions and pollutant components. In a comparative study of
biodeterioration processes on monuments in Latin America, limestone deterioration at the Mayan
site of Uxmal was enhanced by biosolubilization by metabolic acids from bacteria and fungi
while destruction of the Cathedral of La Plata was attributed primarily to atmospheric pollutants
(Herrera and Videla, 2004).
PM deposition onto surfaces such as metal, glass, stone and paint can lead to soiling.
Soiling results when PM accumulates on an object and alters the optical characteristics
(appearance). The reflectivity of a surface may be changed or presence of particulates may alter
light transmission. These effects can impact the aesthetic value of a structure or result in
reversible or irreversible damage to statues, artwork and architecturally or culturally significant
buildings. Due to soiling of building surfaces by PM, the frequency and duration of cleaning
may be increased. Soiling affects the aesthetic appeal of painted surfaces. In addition to natural
factors, exposure to PM may give painted surfaces a dirty appearance. Pigments in works of art
5-27
-------
can be degraded or discolored by atmospheric pollutants, especially sulfates (US EPA, 2008c,
Annex E-15).
Formation of black crusts due to carbonaceous compounds and buildup of microbial
biofilms results in discoloration of surfaces. Black crust includes a carbonate component derived
from building material and organic carbon (OC) and elemental carbon (EC). In limited new
studies quantifying the OC and EC contribution to soiling by black crust, OC predominated over
EC at almost all locations (Bonazza et al., 2005). Limited new studies suggest that traffic is the
major source of carbon associated with black crust formation (Putaud et al., 2004) and that
soiling of structures in Oxford, UK showed a relationship with traffic and NO2 concentrations
(Viles and Gorbushina, 2003). These findings attempt to link atmospheric concentrations of PM
to observed damage. However, no data on rates of damage are available and all studies were
conducted outside of the U.S.
5.4.3 Staff Conclusions
Available evidence in regards to materials damage and soiling supports the following
observations:
• Materials damage and soiling that occur through natural weathering processes are
enhanced by exposure to atmospheric pollutants, most notably SO2 and paniculate
sulfates.
• While ambient particles play a role in the corrosion of metals and in the weathering of
materials, no quantitative relationships between ambient particle concentrations and
rates of damage have been established.
• While soiling associated with fine and course particles can result in increased
cleaning frequency and repainting of surfaces, no quantitative relationships between
particle characteristics and the frequency of cleaning or repainting have been
established.
• Limited new data on the role of microbial colonizers in biodeterioration processes and
contributions of black crust to soiling are not sufficient for quantitative analysis.
• While several studies in the PM ISA and NOx/SOx ISA suggest that particles can
promote corrosion of metals there remains insufficient evidence to relate corrosive
effects to specific particulate levels or to establish a quantitative relationship between
ambient PM and metal degradation. With respect to damage to calcareous stone,
numerous studies suggest that wet or dry deposition of particles and dry deposition of
gypsum particles can enhance natural weathering processes.
5-28
-------
Revisiting the overarching policy question as to whether the available scientific evidence
supports or calls into question the adequacy of the protection for materials afforded by the
current suite of secondary PM standards, we conclude that no new evidence in this review calls
into question the adequacy of the protection for materials afforded by the current standard. PM
effects on materials can play no quantitative role in considering whether any revisions of the
secondary PM NAAQS are appropriate at this time. Nonetheless, in the absence of information
that provides a basis for establishing a different level of control, observations continue to support
retaining an appropriate degree of control on both fine and coarse particles to help address
materials damage and soiling associated with PM.
5.4.4 Key Uncertainties and Areas for Future Research and Data Collection
Quantitative relationships are needed between particle size, concentration, chemical
concentrations and frequency of repainting and repair. Deposition rates of airborne PM to
surfaces would provide an indication of rate and degree of damage to surfaces. There is
considerable uncertainty with regard to interaction of co-pollutants in regards to materials
damage and soiling processes.
5-29
-------
5.5 REFERENCES
Audet P; Charest C (2007). Heavy metal phytoremediation from a meta-analytical perspective. Environ Pollut, 147:
231-237.
Bonazza A; Sabbioni C; Ghedini N. (2005). Quantitative data on carbon fractions in interpretation of black crusts
and soiling on European built heritage. Atmos Environ, 39: 2607-2618.
Bond TC; Streets DG; Yarber KF; Nelson SM; Woo J-H; Klimont Z. (2004). A technology-based global inventory
of black and organic carbon emissions from combustion. J Geophys Res, 109.
Bond TC; Sun H. (2005). Can reducing black carbon emissions counteract global warming?. Environ Sci Technol,
39: 5921-5926.
CCSP (2009). Atmospheric Aerosol Properties and Climate Impacts, A Report by the U.S. Climate Change Science
Program and the Subcommittee on Global Change Research. [Mian Chin, Ralph A. Kahn, and Stephen E.
Schwartz (eds.)]. National Aeronautics and Space Administration, Washington, D.C., USA.
Forster P; Ramaswamy V; Artaxo P; Betss R; Fahey DW; Haywood J; Lean J; Lowe DC; Myhre G; Nganga J; Prinn
R; Raga G; Schultz M; Van Borland R. (2007). Changes in atmospheric constituents and in radiative
forcing. In Solomon, S.; Qin,D.; Manning, M.; Chen, Z.; Marquis, M.; Averyt, K. B.; Tignor, M.; Miller,
H. L. (Ed.),Climate Change 2007: the physical science basis. Contribution of Working Group I to the fourth
assessment report of the intergovernmental panel.
Freer-Smith PH; El-khatib A; Taylor G. (2004). Capture of paniculate pollution by trees: a comparison of species
typical of semi-arid areas (Ficus nitida and Eucalyptus globulus) with European and North American
species. Water Air Soil Pollut, 155: 173-187.
Gitay H; Brown S; Easterling W; Jallow B. (2001). Ecosystems and their goods and services. In Climate change
2001: impacts,adaptation and vulnerability. Contribution of working group II to the third assessment report
of the Intergovernmental Panel on Climate Change (pp. 237-342). Cambridge, United Kingdom:
Cambridge University Press.
Givati A; Rosenfeld D. (2004). Quantifying precipitation suppression due to air pollution. J Appl Meteorol, 43:
1038-1056.
Gomot-De Vaufleury A; Pihan F. (2002). Methods for toxicity assessment of contaminated soil by oral or dermal
uptake inland snails: Metalbioavailability andbioaccumulation. EnvironToxicol Chem, 21: 820-827.
Grantz DA; Garner JHB; Johnson DW. (2003). Ecological effects of paniculate matter. Environ Int, 29: 213-239.
Hageman KJ; Simonich SL; Campbell DH; Wilson GR; Landers DH. (2006). Atmospheric deposition of current-use
and historic-use pesticides in snow at national parks in the western United States. Environ Sci Technol, 40:
3174-3180
Hassan R; Scholes R; AshN. (2005). Ecosystems and human well-being: current state and trends, volume 1. United
Kingdom: Shearwater Books.
Herrera LK; Videla HA. (2004). The importance of atmospheric effects on biodeterioration of cultural heritage
constructional materials. Int Biodeterior Biodegradation, 54: 125-134.
IPCC (2007): Summary for Policymakers. In: Climate Change 2007: The Physical Science Basis. Contribution of
Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change
[Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)].
Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
5-30
-------
Jacobson MZ (2000). A physically-based treatment of elemental carbon optics: implications for global direct forcing
of aerosols. Geophys Res Lett, 27: 217-220.
Jacobson MZ; Kaufman YJ (2006). Wind reduction by aerosol particles. Geophys Res Lett, 33 ARLN 24814
Kucera T; Horakova H; Sonska A (2008). Toxic metal ions in photoautotrophic organisms. Photosynthetica, 46:
481-489.
Landers DH; Simonich SL; Jaffe DA; Geiser LH; Campbell DH; Schwindt AR; Schreck CB; Kent ML; Hafner WD;
Taylor HE; HagemanKJ; Usenko S; AckermanLK; Schrlau JE; Rose NL; Blett TF; Erway MM. (2008).
The Fate, Transport and Ecological Impacts of Airborne Contaminants in Western National Parks (USA).
U.S. Environmental Protection Agency, Office of Research and Development, NHEERL, Western Ecology
Division. Corvallis, Oregon. EPA/600/R-07/138
National Research Council (2001). Climate change science: an analysis of some key questions. National Research
Council. National Academy Press. Washington, DC.
Parrish ZD; White JC; Isleyen M; Gent MPN; lannucci-Berger W; Eitzer BD; Kelsey JW; Mattina Ml. (2006).
Accumulation of weathered polycyclic aromatic hydrocarbons (PAHs) by plant and earthworm species.
Chemosphere, 64: 609-618.
Patra M; Bhowmik N; Bandopadhyay B; Sharma A (2004). Comparison of mercury, lead and arsenic with respect to
genotoxic effects on plant systems and the development of genetic tolerance. Environ Exp Bot, 52: 199-
223.
Putaud J-P; Raes F; Van Dengenen R; Bruggemann E; Facchini M-C; Decesari S; Fuzzi S; Gehrig R; Huglin C; Laj
P; Lorbeer G; Maenhaut W; Mihalopoulos N; Muller K; Querol X; Rodriguez S; Schneider J; Spindler G;
ten Brink H; Torseth K; Wiedensohler A. (2004). A European aerosol phenomenology—2: chemical
characteristics of paniculate matter at kerbside, urban, rural and background sites in Europe. Atmos
Environ, 38: 2579-2595.
Regoli F; Gorbi S; Fattorini D; Tedesco S; Notti A; Machella N; Bocchetti R; Benedetti M; Piva F. (2006). Use of
the land snail Helix aspersa sentinel organism for monitoring ecotoxicologic effects of urban pollution: An
integrated approach. Comp Biochem Physiol A Mol Integr Physiol, 114:63-69.
Salemaa M; Derome J; Helmisaari HS; Nieminen T; Vanha-Majamaa I. (2004). Element accumulation in boreal
bryophytes,lichens and vascular plants exposed to heavy metal and sulfur deposition in Finland. Sci Total
Environ, 324: 141-160.
Sato M; Hansen J; Koch D; Lucis A; Ruedy R; Dubovik O; Holben B; Chin M; Novakov T. (2003). Global
atmospheric black carbon inferred from AAEONET. Presented at Proceedings of the National Academy of
Science.
Schilling JS; Lehman ME. (2002). Bioindication of atmospheric heavy metal deposition in the Southeastern US
using the moss Thuidium delicatulum. Atmos Environ, 36: 1611-1618.
Smith WH (1990). Forest nutrient cycling: toxic ions. In Air pollution and forests: interactions between air
contaminants and forest ecosystems. New York, NY: Springer-Verlag.
Strydom C; Robinson C; Pretorius E; Whitcutt JM; Marx J; Bornman MS (2006). The effect of selected metals on
the central metabolic pathways in biology: A review. WaterSA, 32: 543-554.
U.S. EPA (1996). Air Quality Criteria for Paniculate Matter. U.S. Environmental Protection Agency. Research
Triangle Park, NC.EPA/600/P-95/001aF-cF. November 2004. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm7s jm_cr_cd. html
5-31
-------
US EPA (2004). Air Quality Criteria for Paniculate Matter. National Center for Environmental Assessment, Office
of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711;
report no. EPA/600/P-99/002aF and EPA/600/P-99/002bF. October 2004. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr cd.html
US EPA (2005). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA EPA-452/R-05-005a. December 2005. Available
at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr sp.html
U.S. EPA. (2006). Air Quality Criteria for Lead Final Report. U.S. Environmental Protection Agency, Washington,
DC, EPA/600/R-05/144aF-bF, October 2006. Available at:
http://www.epa.gov/ttn/naaqs/standards/pb/s_pb_cr_cd.html
US EPA (2008a). Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate Matter.
National Center for Environmental Assessment and Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-08-004. March
2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/2008 03 final integrated review_plan.pdf
US EPA (2008c). U.S. EPA. Integrated Science Assessment (ISA) for Oxides of Nitrogen and Sulfur Ecological
Criteria (Final Report). U.S. Environmental Protection Agency, Washington, D.C., EPA/600/R-08/082F,
December 2008. Available at: http://www.epa.gov/ttn/naaqs/standards/no2so2sec/cr_isi.html
US EPA (2009a). U.S. EPA. Integrated Science Assessment for Paniculate Matter (Final Report). U.S.
Environmental Protection Agency, Washington, DC, EPA/600/R-08/139F, December 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_isa.html
US EPA (2009h) Risk and Exposure Assessment for Review of the Secondary National Ambient Air Quality
Standards for Oxides of Nitrogen and Oxides of Sulfur (Final Report). US Environmental Protection
Agency, Research Triangle Park, NC, EPA-452/R-09-008a.
US EPA (2010b). Paniculate Matter Urban-Focused Visibility Assessment - Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
10-004. June 2010. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (2011) Policy Assessment for the Review of the Secondary National Ambient Air Quality Standards for
Oxides of Nitrogen and Oxides of Sulfur. U.S. Environmental Protection Agency, Research Triangle Park,
NC, EPA-452/R-ll-005a,b. February 2011. Available at:
http://www.epa.gov/ttn/naaqs/standards/no2so2sec/cr_pa.html
Viles HA; Gorbushina AA (2003). Soiling and microbial colonisation on urban roadside limestone: A three year
study in Oxford, England. Building Environ, 38: 1217-1224.
Yogui G; Sericano J (2008). Polybrominated diphenyl ether flame retardants in lichens and mosses from King
George Island, maritime Antarctica. Chemosphere, 73: 1589-1593.
5-32
-------
APPENDICES
A. Clean Air Scientific Advisory Committee Letter on Second Draft Policy Assessment,
September 10, 2010 A-l
B. Ambient PM Monitoring Networks B-l
C. Predicted Percent of Counties with Monitors Not Likely to Meet Current and Alternative
Primary Annual and 24-hour PM2.5 Standards C-l
D. Predicted Percent of Counties with Monitors Not Likely to Meet Current and Alternative
Primary 24-hour PMio Standards D-l
E. Information Regarding a 1-hour PM25 Mass Indicator E-l
F. Two Simplified Approaches to Calculate Hourly PM2.5 Light Extinction Values from
Hourly PM2 5 Mass and Relative Humidity Data Plus 24-hour Mean PM2 5 Composition
Data F-l
G. Calculated 24-hour Average PM2 5 Light Extinction and Adjusted Candidate Protection
Levels G-l
H. Predicted Percent of Counties with Monitors Not Likely to Meet Current and Alternative
Secondary PM2.s Standards H-l
-------
This page left intentionally blank.
-------
APPENDIX A
Clean Air Scientific Advisory Committee (CASAC) Letter on
Second Draft Policy Assessment
September 10,2010
A-l
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON B.C. 20460
"^ tff" OFFICE OF THE ADMINISTRATOR
4 PR01 SCIENCE ADVISORY BOARD
September 10, 2010
EPA-CASAC-10-015
The Honorable Lisa P. Jackson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, D.C. 20460
Subj ect: CASAC Review of Policy Assessment for the Review of the PMNAAQS - Second
External Review Draft (June 2010)
Dear Administrator Jackson:
The Clean Air Scientific Advisory Committee (CASAC) Parti culate Matter (PM) Review
Panel met on July 26 - 27, 2010 and on August 25, 2010 in a public teleconference to review the
Policy Assessment for the Review of the PMNAAQS - Second External Review Draft (June 2010).
This letter highlights CASAC' s main comments on this document, followed by consensus
responses to the charge questions and comments of individual Panel members.
This review of the Second Draft Policy Assessment completes the first cycle through the
revised suite of NAAQS review documents and thus represents a major milestone. CASAC
commends EPA staff for developing an ordered and transparent basis for decision-making
throughout the NAAQS review process from the Integrated Science Assessment (ISA) to the
Quantitative Health Risk Assessment and Urban Focused Visibility Assessment and then to the
Policy Assessment. The Second Draft Policy Assessment was notably responsive to CASAC' s
comments on the first draft. At CASAC's request, the current draft sets out the underlying
decision-making algorithms, greatly enhancing the transparency and readability of the document.
EPA's approach to reviewing the standard is explicitly articulated throughout the document, as are
the key decision-making points and the evidence considered. CASAC's major concerns, as
expressed in our letter of May 17, 2010, have been addressed. EPA staff are to be congratulated
for building on CASAC's suggestions and developing an assessment that provides a scientifically
sound basis for making decisions on the primary and secondary PM standards.
Primary Standards for Fine Particles
CASAC supports the EPA staffs conclusion in the Second Draft Policy Assessment that
"currently available information clearly calls into question the adequacy of the current standards".
For PM2.5, the current 24-hour primary standard is 35 |ig/m3 and the annual standard is 15 |ig/m3.
EPA staff also conclude that consideration should be given to alternative annual PM2.5 standard
levels in the range of 13 - 1 1 |ig/m3, in conjunction with retaining the current 24-hour PM2.5
-------
standard level of 35 |ig/m3, and that consideration could also be given to an alternative 24-hour
PM2 5 standard level of 30 |ig/m3 in conjunction with an annual standard level of 1 l|ig/m3.
CASAC concludes that the levels under consideration are supported by the epidemiological and
toxicological evidence, as well as by the risk and air quality information compiled in the Integrated
Science Assessment (December 2009), Quantitative Health Risk Assessment for Particular Matter
(June 2010) and summarized in the Second Draft Policy Assessment. Although there is increasing
uncertainty at lower levels, there is no evidence of a threshold (i.e., a level below which there is no
risk for adverse health effects). In addition, these combinations of annual/daily levels may not be
adequately inclusive. It was not clear why, for example, a daily standard of 30 |ig/m3 should only
be considered in combination with an annual level of 11 |ig/m3. The rationale for the 24-
hour/annual combinations proposed for the Administrator's consideration (and the exclusion of
other combinations within the ranges contemplated) should be more clearly explained.
Primary Standard for Thoracic Coarse Particles
CASAC recommends that the primary standard for PMi0 should be revised downwards.
While current evidence is limited, it is sufficient to call into question the level of protection
afforded by the current standard (a 24-hour standard of 150 |ig/m3).
CASAC supports the EPA staff conclusion that it is appropriate to change the PMi0
standard to a 98th percentile form because of its higher rate of identifying areas in nonattainment
while reducing the rate of misclassification. We do not agree that the available scientific evidence
strongly supports the proposed upper bound standard level of 85 |ig/m3. The Second Draft Policy
Assessment demonstrates that a 98th percentile level of 85 |ig/m3 would be less stringent as
compared to the current standard, protecting a smaller fraction of the population. In fact, on a
population basis, results in the Second Draft Policy Assessment demonstrate that a 98th percentile
level between 75 and 80 |ig/m3 is comparable in the degree of protection afforded to the current
PMio standard. The change in form will lead to changes in levels of stringency across the country,
a topic needing further exploration. While recognizing scientific uncertainties, CASAC supports a
lower level to provide enhanced protection, somewhere in the range of 75 - 65 |ig/m3. We
recognize that the Administrator will need to apply the Clean Air Act's requirement for a "margin
of safety" in a context of uncertainty with respect to the health effects of thoracic coarse particles.
The Second Draft Policy Assessment concludes that PMio should continue to be the
indicator for thoracic coarse particles. While it would be preferable to use an indicator that reflects
the coarse PM directly linked to health risks (PMio-2.s), CASAC recognizes that there is not yet
sufficient data to permit a change in the indicator from PMio to one that directly measures thoracic
coarse particles. To improve EPA's scientific basis for the next NAAQS review, we recommend
the deployment of a network of PMio-2.5 sampling systems so that future studies will be able to
expand the evidence base on this indicator and facilitate assessment of whether PMio-2.5 should be
used as an appropriate indicator for thoracic coarse particles. In concluding this letter, we elaborate
further on the urgency of research on certain aspects of PM and health.
-------
Secondary Standard for PM-Related Visibility Impairment
CAS AC supports the EPA staff conclusion that "currently available information clearly
calls into question the adequacy of the current standards and that consideration should be given to
revising the suite of standards to provide increased public welfare protection." The current
secondary standards are identical to the current primary standards for fine and thoracic coarse
particles. The detailed estimates of hourly PM light extinction under current conditions (and for
assumed scenarios of meeting current standards) clearly demonstrate that current standards do not
protect against levels of visual air quality which have been judged to be unacceptable in all of the
available urban visibility preference studies. EPA staffs approach for translating and presenting
the technical evidence and assessment results is logically conceived and clearly presented. The 20-
30 deciview range of levels chosen by EPA staff as "Candidate Protection Levels" is adequately
supported by the evidence presented.
While the evidence shows that the current standard does not adequately protect visibility,
the choice of indicator for such protection was a subject of considerable discussion among CAS AC
panelists. The Second Draft Policy Assessment discusses three potential indicators: a PM2.5 Mass
Indicator, a Speciated PM2 5 Mass-calculated Light Extinction Indicator, and a Directly Measured
PM2.5 Light Extinction Indicator. Overwhelmingly, CASAC would prefer the direct measurement
of light extinction, the property of the atmosphere that most directly relates to visibility effects. It
has the advantage of relating directly to the demonstrated harmful welfare effect of ambient PM on
human visual perception. However, in discussing the Directly Measured PM2 5 Light Extinction
Indicator with EPA staff, we learned that the time required to develop an official Federal Reference
Method (FRM) for this indicator would postpone its implementation for years. Given the time lag
associated with implementing the Directly Measured Indicator, CASAC agrees with EPA staffs
preference for a Speciated PM2 5 Mass-calculated Light Extinction Indicator. Its reliance on
procedures that have already been implemented in the Chemical Speciation Network (CSN) and
routinely collected continuous PM2.5 data suggest that it could be implemented much sooner than a
directly measured indicator.
Areas for Future Research
The Second Draft Policy Assessment has identified scientific issues that will need to be
addressed in order to improve EPA's scientific basis for promulgating PM standards in the future.
As stated in our letter of May 17, 2010, CASAC urges the Agency to reinvigorate research that
might lead to new indicators that may be more directly linked to the health and welfare effects
associated with ambient concentrations of PM. CASAC also suggests the ongoing collection of
more comprehensive PM monitoring data, including expanding the range of sizes to provide
information in the ultrafme particle range, and adding measurements of numbers, chemistry,
species, and related emissions characteristics of particles. CASAC strongly urges EPA to pursue
research to develop a Federal Reference Method for a Directly Measured PM2.5 Light Extinction
Indicator and to develop baseline light extinction data so that it will be available for the next 5 year
review cycle. CASAC is available to provide advice on priorities for PM-related research.
in
-------
Thank you for the opportunity to comment on the Second Draft Policy Assessment. We
look forward to receiving your response.
Sincerely,
/Signed/
Dr. Jonathan M. Samet, Chair
Clean Air Scientific Advisory Committee
IV
-------
NOTICE
This report has been written as part of the activities of the EPA's Clean Air Scientific Advisory
Committee (CASAC), a federal advisory committee independently chartered to provide extramural
scientific information and advice to the Administrator and other officials of the EPA. CAS AC
provides balanced, expert assessment of scientific matters related to issues and problems facing the
Agency. This report has not been reviewed for approval by the Agency and, hence, the contents of
this report do not necessarily represent the views and policies of the EPA, nor of other agencies
within the Executive Branch of the federal government. In addition, any mention of trade names or
commercial products does not constitute a recommendation for use. CASAC reports are posted on
the EPA Web site at: http://www.epa.gov/casac.
-------
Clean Air Scientific Advisory Committee
Particulate Matter Review Panel
CHAIR
Dr. Jonathan M. Samet, Professor and Chair, Department of Preventive Medicine, University of
Southern California, Los Angeles, CA
CASAC MEMBERS
Dr. Joseph Brain, Philip Drinker Professor of Environmental Physiology, Department of
Environmental Health, Harvard School of Public Health, Harvard University, Boston, MA
Dr. H. Christopher Frey, Professor, Department of Civil, Construction and Environmental
Engineering, College of Engineering, North Carolina State University, Raleigh, NC
Dr. Donna Kenski,l Data Analysis Director, Lake Michigan Air Directors Consortium, Rosemont,
IL
Dr. Armistead (Ted) Russell, Professor, Department of Civil and Environmental Engineering,
Georgia Institute of Technology, Atlanta, GA
Dr. Helen Suh, Associate Professor, Harvard School of Public Health, Harvard University, Boston,
MA
Dr. Kathleen Weathers, Senior Scientist, Gary Institute of Ecosystem Studies, Millbrook, NY
CONSULTANTS
Dr. Lowell Ashbaugh, Associate Research Ecologist, Crocker Nuclear Lab, University of
California, Davis, Davis, CA
Prof. Ed Avol, Professor, Preventive Medicine, Keck School of Medicine, University of Southern
California, Los Angeles, CA
Dr. Wayne Cascio, Professor, Medicine, Cardiology, Brody School of Medicine at East Carolina
University, Greenville, NC
Dr. David Grantz, Director, Botany and Plant Sciences and Air Pollution Research Center,
Riverside Campus and Kearney Agricultural Center, University of California, Parlier, CA
Dr. Joseph Helble, Dean and Professor, Thayer School of Engineering, Dartmouth College,
Hanover, NH
Dr. Rogene Henderson, Senior Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
1 / Did not participate in this review.
VI
-------
Dr. Philip Hopke, Bayard D. Clarkson Distinguished Professor, Department of Chemical
Engineering, Clarkson University, Potsdam, NY
Dr. Morton Lippmann, Professor, Nelson Institute of Environmental Medicine, New York
University School of Medicine, Tuxedo, NY
Dr. William Malm, Research Physicist, National Park Service Air Resources Division,
Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO
Mr. Charles Thomas (Tom) Moore, Jr., Air Quality Program Manager, Western Governors'
Association, Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort
Collins, CO
Dr. Robert F. Phalen, Professor, Department of Community & Environmental Medicine; Director,
Air Pollution Health Effects Laboratory; Professor of Occupational & Environmental Health,
Center for Occupation & Environment Health, College of Medicine, University of California
Irvine, Irvine, CA
Dr. Kent Pinkerton, Professor, Regents of the University of California, Center for Health and the
Environment, University of California, Davis, CA
Mr. Richard L. Poirot, Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Waterbury, VT
Dr. Frank Speizer, Edward Kass Professor of Medicine, Channing Laboratory, Harvard Medical
School, Boston, MA
Dr. Sverre Vedal, Professor, Department of Environmental and Occupational Health Sciences,
School of Public Health and Community Medicine, University of Washington, Seattle, WA
SCIENCE ADVISORY BOARD STAFF
Dr. Holly Stallworth, Designated Federal Officer, EPA Science Advisory Board Staff Office
vn
-------
CASAC Responses to Charge Questions on the Second Draft Policy Assessment for the
Review of the Particulate Matter NAAQS
Primary Standards for Fine Particles
1. Current Approach (Section 2.1.3):
a. What are CASAC's views on the staffs approach to translating the available
epidemiological evidence, risk information, and air quality information into the
basis for reaching conclusions on the adequacy of the current standards and on
alternative standards for consideration?
CASAC agrees with the approach as described in Section 2.1.3 and appreciates the clarity with
which the approach is detailed. The overview of the approach presented in Figure 2-1 is well-
organized, logical, and clear. CASAC agrees that it is appropriate to return to the strategy used in
1997 that considers the annual and the short-term standards together, with the annual standard as
the controlling standard, and the short-term standard supplementing the protection afforded by the
annual standard. CASAC commends the expansion of the discussion on evidence of risk across life
stages as well as of specific susceptibility risk factors and the use of empirical evidence and risk
assessment findings together. CASAC considers it appropriate to place the greatest emphasis on
health effects judged to be causal or likely causal in the analysis presented in the ISA. Finally, the
statement that the data "call into question" the adequacy of the current standard could be more
forcefully stated by concluding that the current standard is not protective.
b. Has staff appropriately applied this approach in reviewing the adequacy of the
current standards (Section 2.2) and potential alternative standards (Section 2.3)?
The staff has carefully followed this approach in reviewing the adequacy of the current standards
and in considering potential alternative standards. The outline of the text of Section 2.3 follows the
outline presented in the overview of the approach given in Figure 2-1.
2. Form of the Annual Standard (Section 2.3.3.1):
a. What are CASAC's views on the additional analyses conducted to characterize the
potential for disproportionate impacts on susceptible populations, including low
income groups and minorities associated with spatial averaging allowed by the
current annual standard?
b. In light of these analyses, what are CASAC's views on staffs conclusion that the
form of the annual standard should be revised to eliminate spatial averaging?
CASAC found the additional analyses provided in the 2nd draft PA to be helpful in understanding
how spatial averaging differs relative to the highest average value from a single community site.
This latter approach helps to ensure adequate protection of populations living in lower
socioeconomic areas and contributes an additional margin of safely for other populations.
Although much of the epidemiological research has been conducted using community-wide
averages, several key studies reference the nearest measurement site, so that some risk estimates
are not necessarily biased by the averaging process. Further, the number of such studies is likely to
-------
expand in the future. CASAC concludes that it is reasonable for EPA to eliminate the spatial
averaging in the new PM2 5 annual average standard.
3. Alternative Levels (Section 2.3.4): What are CASAC's views on the following:
a. The insights that can be gained into potential alternative standard levels by
considering:
i. Confidence bounds on concentration-response relationships?
CASAC commends the progress made in attempting to use confidence bounds in considering
alternative levels of the standard, but also finds unresolved complexities. First, staff apparently
made a comprehensive effort to identify relevant studies for which bounds were reported on
concentration-response (C-R) relationships; this should be explicitly stated. Second, the statement
made in reference to what these bounds do not indicate ("these analyses do not provide evidence of
a concentration below which the confidence interval becomes notably wider and uncertainty in a
C-R relationship substantially increases" [p.2-57]) is contradictory to what they, in fact, do
indicate. The confidence bounds widen at lower concentrations because there are fewer data at
such concentrations, as acknowledged by staff. This widening is of interest in characterizing
precision of estimates as one source of uncertainty. Third, CASAC does not agree with the
conclusion that these bounds cannot be used in considering alternative levels of the PM NAAQS,
even with the limited C-R functions shown. EPA Staff should be encouraged to integrate the
information available on relevant C-R confidence bounds with that on study concentration
distributions in arriving at a range of levels for consideration.
For the future, findings of epidemiological studies might be used in several ways in considering a
range of levels for a NAAQS. It would be preferable to have information on the concentrations
that were most influential in generating the health effect estimates in individual studies. Less ideal,
but still useful, would be information on the distribution of concentrations experienced by
participants in the studies. For time-series studies, because of the similar number of events (e.g.,
deaths) per day, this is likely to be the same as the PM concentration distribution; the situation is
more complex for cohort studies in which exposures of individuals change over time. Least
preferable is using PM concentration distribution metrics, such as those used by EPA Staff in
arriving at a range of levels for consideration. An attempt should be made, to the extent possible,
to integrate this latter approach with aspects of the first two approaches, realizing that the reported
study findings and data needed to accomplish this goal may not be readily available, and that
interactions with investigators may be needed.
ii. Different statistical metrics that characterize air quality distributions from
multi-city epidemiological studies?
The Second Draft Policy Assessment provides two alternatives, referred to as the composite
monitor and the maximum monitor. On the top of page 2-61, the text appears to be stating that, for
the same air quality domain, the composite monitor concentrations are less than those based on the
maximum monitor approach, and an argument is made that an approach based on composite
monitors has a "margin of safety" compared to the maximum monitor perspective. However, a
judgment is made that data should be selected from the epidemiological studies for which the C-R
relationships are "strongest," and that concentrations not more than one standard deviation below
-------
the long-term mean concentration should be used. The judgment, while not unreasonable, is not
explained.
It is not clear why the lower bound to be considered is a range from the 10th to 25th percentiles, as
opposed to, say, the 10th percentile alone. In Figure 2-7, for long-term exposure studies, in the
~,th
upper panel, the 10 percentile annual mean concentrations range from approximately 9 to 11
|ig/m3. The population-weighted values are 10 to 13 |ig/m3. In both cases, the upper bounds c
these ranges are for the high site, and the lower bounds are for the composite monitor.
In summary, this section of the report lacks clarity and focus on the key consideration of
identifying ambient concentrations at which adverse effects are observed, in anticipation of
supporting a range of concentrations that take into account the statutory mandate to provide an
adequate margin of safety.
b. Potential alternative annual standard levels based on composite monitor
distributions versus maximum monitor distributions?
The composite monitor approach is preferable because of its stability, and for the additional margin
of safety it provides. The NAAQS should provide health protection for both long-term and short-
term health effects. It is not clear, for example, as to why the 24-hour level should be at least 2.5
times higher than the annual standard. Such a statement seems to be independent of consideration
of health effects. A statement is made on page 2-73, lines 26-27 that "based on this consideration"
consideration should be given to retaining the 35 |ig/m3 24-hr level in conjunction with annual
standards of 13 to 11 |ig/m3. Setting aside the math problem here (e.g., 11 *2.5 = 27.5, not 35), the
rationale for the 2.5 times factor appears arbitrary and not based on health considerations.
c. Use of risk information in informing staff conclusions on alternative annual and
24-hour standard levels, including approaches used to assess overall confidence
and potential bias in the risk estimates?
The risk information provides valuable insights, and should be used in drawing conclusions.
However, there is not symmetry between the evidence-based section and the risk-based section. .
The "evidence-based" section reaches the conclusion that alternative levels to be considered should
be 11 to 13 |ig/m3 for the annual standard and 35 ug/m3 for the 24-hour standard, and also a
combination of 11/30 |ig/m3 for the annual/24-hour levels. However, the risk-based analysis does
not systematically evaluate these combinations, omitting the 11/35 |ig/m3 and 11/30 |ig/m3
combinations. Furthermore, the text implies that a 10/35 ug/m3 case was analyzed, but no results
were reported. This difference between the ranges from the two sections reflects in part the
scenarios considered in the risk assessment. While the Administrator's consideration should not be
limited to those combinations that were analyzed quantitatively, the final policy assessment should
be systematic and emphatic about providing conclusions regarding combinations of annual and
daily levels that were not analyzed quantitatively but that are recommended for consideration.
The results of the risk assessments are presented mainly in terms of percentage risk reduction
compared to the current standard, in Figures 2-11 and 2-12 for long-term and short-term effects,
respectively. While this is useful information, it is not directly relevant to the setting of a NAAQS,
-------
given the goal of a NAAQS—to protect public health with an adequate margin of safety.
Additionally, the information on risk reduction might be better presented as the absolute numbers
of deaths avoided rather than the percentage reduction under the various scenarios. The text should
be rewritten to better reflect the utility and relevance of the information on reduction of disease
burden for determining the NAAQS.
This section should not only focus on the best estimate of risk, but the confidence intervals and
non-quantified sources of bias, such as the role of socio-economic status (SES). See also Page 2-
35, lines 10-12, which indicates that sensitivity analysis of model specification used in the risk
assessment produce risk estimates that are a factor of 2 to 3 higher than the core risk estimates.
d. Staffs conclusion that alternative annual standard levels in the range of 13 to 11
ug/m3 are most strongly supported by the available evidence and risk-based
information?
The rationale for the conclusion was well developed, but could use further justification, particularly
in regard to the pairing of the 24-hour and annual standards. The risk assessment did not explore all
the combinations considered in the Policy Assessment. While CAS AC agrees with the range of 13
to 11 ug/m3, it finds less justification for the pairings proposed.
e. Staffs approach of focusing on peak-to-mean ratios to inform the level of a 24-
hour standard that would provide supplemental protection to a generally
controlling annual standard?
The peak-to-mean ratio merits consideration in providing insight as to whether the annual or 24-
hour standard would be controlling in a particular area. It is not relevant to informing the actual
level to be selected for the 24-hour standard.
f. Staffs conclusion that consideration should be given to retaining the current 24-
hour standard level of 35 ug/m3 in conjunction with annual standard levels in the
range of 13 to 11 ug/m3, and that consideration could also be given to an
alternative 24-hour standard level of 30 ug/m3 particularly in conjunction with an
annual standard level of 11 ug/m3?
The conclusions are reasonable in relation to the criteria established by the Clean Air Act (CAA),
and those developed by the OAQPS Staff that have been endorsed by CASAC. The choices within
these options will need to be based on the Administrator's interpretation of the CAA's requirement
for an adequate margin-of-safety. In other words, in the absence of thresholds in the dose-response
relationships for the health outcomes of concern, how much public health impact resulting from
exposure to ambient air PM2.5 is acceptable under the CAA.
The least protective option (35-13 |ig/m3) would provide significant additional public health
benefits in most of the U.S., in comparison to the current limits (35-15 |ig/m3). The most protective
option (30-11 |ig/m3) would provide significant additional public health benefits to a larger part of
the U.S. population in comparison to the current limits (35-15 |ig/m3) and any of the intermediate
options, but would not prevent at least some adverse health effects among the most susceptible
-------
segments of the population, given our current understanding of dose-response relationships.
4. Key Uncertainties and Areas for Future Research and Data Collection (Section 2.5):
What are CASAC's views on the areas for future research and data collection outlined in
this section, on relative priorities for research in these areas, and on any other areas that
ought to be identified?
The key uncertainties and areas for future research and data collection are well summarized in
Section 2.5. The acknowledgement (at the top of page 2-87) that "Much of this research may
depend on the availability of increased monitoring data" is apt and appreciated. The opportunities
for epidemiological research to effectively address the knowledge gaps on the effects, and
concentration-response relationships, of PM components and source-related mixtures cannot be
achieved without additional monitoring data to provide PM speciation and better temporal and
spatial resolution. Only the EPA can provide the impetus and support for such an enhancement in
air quality monitoring.
The research needs to address uncertainties in health outcomes, exposure durations of concern, and
susceptible populations that are also very nicely outlined are well targeted, and can be effectively
studied in human populations. Such studies, to be most productive, will need the enhanced
monitoring data, as recognized by EPA staff.
This section, as written, has more to do with future research priorities than with uncertainties that
influence impending decisions on revisions to the PM2.5 NAAQS. The section outlines a very
broad and ambitious research agenda. It would help to begin this section with a prioritized review
of key uncertainties in order to help establish priorities among the suggested research topics.
Obviously the key uncertainty is the range of concentrations that are causing the observed health
effects in the epidemiological studies, and the degree of certainty in effects at the lower
concentrations along the C-R relationship. This uncertainty has necessitated using the
distributional measures of concentrations from the epidemiology studies in attempting to make the
link between the epidemiological findings and consideration of alternative concentrations for the
PM NAAQS. While this uncertainty is reflected in two (p.2-88 and 2-90) of the many
recommendations for future research that C-R functions include confidence bounds, this
uncertainty should be highlighted. We urge careful attention to priorities in relation to future
revisions of the PM NAAQS, rather than a lengthy list of research topics.
CASAC finds the list to be appropriate, but also suggests consideration of the following:
• Generating time-activity data to support probabilistic scenario-based exposure models,
such as additional activity diary data to incorporate into the Consolidated Human
Activity Database (CHAD).
• Characterizing indoor exposures to PM of ambient origin. For example, the penetration
of ambient PM2 5 and PMi0 into indoor microenvironments (home, work, school,
restaurant, bar, vehicle) should be better characterized, particularly taking into account
differences in penetration with respect to particle size and composition. Given the
greater amount of time we spend in indoor vs. outdoor environments, the need for these
data is compelling.
-------
• Addressing the bidirectional linkages between climate change and concentration, size
distribution and composition of PM in the PMio, PM2.5, and ultrafine particle (UFP)
fractions. This would include assessing the relative effects of climate cooling due to
aerosols (e.g., sulfate) vs. climate warming due to elemental carbon. Effects of
increased wildfires, windblown dust and pollen seasonality are also of interest.
• Continuing support of toxicological research in terms of chemical components, sources
and subfractions (to include UFP). Toxicological studies will address biological
plausibility and give insights as to possible mechanisms. Although C-R relationships
are a challenge to extrapolate from animal to human, animal studies do provide an
effective means to conduct controlled and well-characterized exposure scenarios to
examine C-R relationships.
Primary Standard for Coarse Particles
5. Current Approach (sections 3.1.4, 3.2, 3.3):
a. What are CASAC's views on the approach to translating the available evidence
and air quality information into the basis for reviewing the coarse particle
standard?
CASAC finds the second draft superior to the first draft reviewed earlier; it demonstrates
considerable progress and responsiveness to CASAC's suggestions. The document is grounded on
explicit data and clearly stated arguments. EPA staff has done its best to take the available
evidence relating to exposure and health effects and to use them as the basis for reviewing the
coarse particle standard.
There are inherent deficiencies which persist because of lack of data. Concentrations of the coarse
particle fraction-particles between 2.5 and 10 microns—are usually estimated by subtraction and
not measured directly. Moreover, given the limited data on coarse particles, much of the evidence
on health effects comes from interpreting studies using PMio and assessing the extent to which the
health effects observed relate to the entire size range collected [including PM^.s] or to only the
coarse particle fraction.
b. Has staff appropriately applied this approach in reviewing the adequacy of the
current standard (section 3.2) and potential alternative standards (section 3.3)?
CASAC responds affirmatively to this question. The staff have noted the limitations of the data and
used them in light of these limitations to address the question of whether current standards are
adequate. CASAC also finds that staff has adequately discussed alternative standards and the
consequences of applying them.
In toto, Chapter 3 reads well and is much improved. EPA staff has done its best to describe an
evidence-based approach for applying the limited amount of health effects evidence and air quality
information in different US regions as a basis for reviewing the adequacy of the current coarse
particle standard.
-------
6. Adequacy of the Current PMi0 Standard (section 3.2): What are CASAC's views on the
alternative approaches presented for considering the evidence and its uncertainties as
they relate to the adequacy of the current standard?
The general consensus of CASAC is that consideration should be given to revising the current 24-
hour PMio standard. The rationale for this recommendation emerges from the judgment that the
current data, while limited, is sufficient to call into question the level of protection afforded the
American people by the current standard. The opinion hinges on the strength of associations in
multi-city studies and positive trends in single city studies linking PMio exposure and health
endpoints, and moreover that these health effects can occur below the current standard. This
approach gives significant weight to studies that have generally reported that PMio-2.5 effect
estimates remain positive when evaluated in co-pollutant models. Likewise controlled human
exposure PMio-2.5 studies showing decreases in heart rate variability and increases in markers of
pulmonary inflammation are deemed adequate to support the plausibility of the associations
reported in epidemiologic studies.
7. Indicator (Section 3.3.1): What are CASAC's views on the approach taken to
considering standard indicator and on staffs conclusion that PMio remains an
appropriate indicator in this review?
The majority of CASAC determined that there was insufficient evidence currently available to
support a change in the indicator from PMio to PMio-2.5. However, CASAC vigorously
recommends the implementation of plans for the deployment of a network of PMio-2.5 sampling
systems so that future epidemiological studies will be able to more thoroughly explore the use of
PMio-2.5 as a more appropriate indicator for thoracic coarse particles.
If a PMio indicator is retained, the Agency should consider limiting the Federal Reference Method
to include only low volume PMio samplers, as high volume PMio samplers do not produce
comparable results.
8. Form (Section 3.3.3): What are CASAC's views on the approach taken to considering
the form of the standard and on staffs conclusion that revising the form to a 98th
percentile form would be appropriate for a 24-hour PMio standard meant to protect
against exposures to thoracic coarse particles?
CASAC felt strongly that it is appropriate to change the statistical form of the PMio standard to a
98th percentile form. Published work has shown that the percentile form has greater power to
identify non-attainment and a smaller probability of misclassification relative to the expected
exceedance form of the standard. This change in form will lead to changes in levels of stringency
across the country, a topic needing further exploration.
9. Level (Section 3.3.4): What are CASAC's views on the following:
a. The approach taken by staff to identify potential alternative PMio standard levels,
in conjunction with a 98th percentile form, including the weight placed on different
studies?
-------
b. Staffs conclusion that the evidence most strongly supports standard levels around
85 ug/m3?
c. The alternative approach to considering the evidence that could support standard
levels as low as 65 ug/m3?
CASAC concurs that the approach in identifying potential alternative PMi0 standard levels are
appropriate, with the discussion regarding the weight placed on different studies clearly and
cogently presented. CASAC also considered that the proposed alternative standard levels of 85
and 65 ug/m3 (based on consideration of 98th percentile PMio concentration) could be justified.
CASAC, however, does not agree that scientific evidence most strongly supports an upper bound
standard level of 85 |ig/m3. As stated in the Second Draft Policy Assessment, scientific evidence
supports the adoption of a standard at least as stringent as the current standard of 150 |ig/m3 based
on one expected exceedance. Table A3 suggests that a 98th percentile level of 85|ig/m3 is less
stringent as compared to the current standard, protecting a smaller fraction of the population.
Results instead point to a 98th percentile level between 75 and 80 |ig/m3 as comparable to the
current standard. CASAC further notes that setting new 24-hour PMio standard levels should also
consider the impact of corresponding changes in PM2 5 standards, which will likely result in lower
24-hour PM2.5 concentrations and lower measured PMio values. Thus, proportionately more coarse
particle mass could be airborne at the standard level. Absent corresponding reduction in the PMio
standard, these lower PM2.5 concentrations would lessen the level of protection provided by the
PMio standard for exposure to PMio-2.s
The Second Draft Policy Assessment does not adequately convey the possible rationale for
selecting the lower end of the proposed range of levels. Therefore, the considerations that might
lead to selecting a PMio standard level more stringent than afforded by the current standard should
be more clearly elaborated. These considerations focus on margin of safety, particularly as it
relates to the impact and weight given to suggestive findings of causality, to findings of positive
but statistically insignificant results, and to exposure measurement error and other sources of
uncertainty.
10. Key Uncertainties and Areas for Future Research and Data Collection (Section 3.5):
What are CASAC's views on the areas for future research and data collection outlined in
this section, on relative priorities for research in these areas, and on any other areas that
ought to be identified?
See comments on Chapter 2.
The key distinction for this chapter is the need to seriously focus on PMio-2.5 for both mass and
composition. CASAC looks forward to the planned implementation of monitors that measure
PMio-2.5, rather than PMi0. There is a critical need for national monitoring data on PMi0-2.5 in order
to provide a basis for epidemiological studies that focus on this size fraction. Furthermore, there is
a need for speciated data to support health effects research. Spatial and temporal variability in
coarse particle mass and composition need to be characterized. In addition, the national
monitoring data will support a baseline for ambient air quality in order to compare with health
effects data in order to assess whether there is a need for a more stringent standard.
8
-------
The research areas described in the draft Section 3.5 are reasonable, but there needs to be strong
emphasis on the critical need for coarse PM data, in order that the NAAQS can move beyond PMi0
as an indicator for thoracic coarse PM in a future NAAQS revision.
Another question to be considered is regarding what size cut-points are most appropriate, and also
regarding what specific components are of most interest or concern with respect to health effects.
There is a need for continuous monitoring of coarse PM (and of PM2.s) in order to support health
effects studies and to be able to assess alternative forms of possible future standards.
Other challenges for future research: (a) it may be difficult to get useful data from rodent
inhalation studies since they can breathe particles only up to about 2 to 3 microns into their lung
airways; (b) getting good chemical characterization of the particles will be a problem, since there
are primary biological materials of potential interest in the thoracic coarse size range.
Prioritization of the research topics is needed, such as via a separate meeting or workshop.
Secondary Standard for PM-related Visibility
11. Current Approach (Section 4.1.3):
a. What are CASAC's views regarding our approach for translating technical
evidence and assessment results into the basis for assessing current fine particle
standards and considering alternative standards to provide protection against
PM-related visibility impairment?
The translation of technical evidence and assessment results as a basis for reviewing and revising
the current secondary fine particle standard is logically conceived, clearly presented, and
responsive to previous CASAC recommendations. The combined evidence-based and impact-
based assessments effectively contrast and integrate the various combinations of metrics for
protecting urban visibility. While this approach is inherently complex, it is clearly explained in the
text and concisely summarized in Figure 4-1. The various tables and graphics in Chapter 4 and its
associated appendices are helpful in communicating the inherent complexity that results from the
evaluation of so many possible combinations of indicators, averaging times, levels and forms.
b. Has staff appropriately applied this approach in reviewing the adequacy of the
current standard (Section 4.2) and potential alternative standards (Section 4.3)?
The detailed estimates of hourly PM light extinction under current conditions and for "what if
scenarios of just meeting current standards clearly indicate that the current PM2.5 standards do not
protect against levels of visual air quality which have been judged to be unacceptable in all of the
available urban visibility preference studies. The levels are too high, the averaging times are too
long, and the PM2.5 mass indicator could be improved to correspond more closely to the light
scattering and absorption properties of suspended particles in the ambient air.
While not discussed in detail in the Second Draft Policy Assessment., direct measurements of light
extinction are the preferable indicator for an alternate standard to make an accurate assessment of
the PM effect on urban visibility. These measurements would provide timely and easy-to-
-------
understand results to address the protection of the public welfare from PM impacts, but without a
Federal Reference Method (FRM) adopted or in the development process - these data are not
currently available for most urban areas. Additional discussion of the timeline and process
anticipated by EPA to advance direct measurement of light extinction monitoring methods to FRM
status would be helpful.
Given this limitation, the detailed estimates of PM light extinction employed for 15 urban areas in
the UFVA, and used to evaluate alternative new indicators including hourly PM2.s mass and
"speciated PM2.5 mass-calculated light extinction" in the Second Draft Policy Assessment are
appropriate for the initial promulgation and first generation of regulatory air quality analysis and
planning; similar to the process for the Regional Haze Rule. The speciated PM2.5 mass-calculated
light extinction indicator produces hourly extinction values quite similar to those resulting from
more complex calculations, and it could be an appropriate indicator for a revised secondary
standard, if employed on an interim basis until methods for direct light extinction measurements
can be developed and deployed.
While the stated intent of the Second Draft Policy Assessment is "to provide as broad an array of
options as is supportable by the available information", the CASAC recommends providing
additional and more focused discussion of the policy implications that may be associated with
selecting and implementing specific combinations of indicators, levels and forms from within this
broad array of options. Some discussion should also be provided to indicate that reductions in light
scattering aerosols could decrease light extinction but increase radiative forcing, while reductions
in light absorbing aerosols would decrease both light extinction and radiative forcing. The
contributions of anthropogenic controllable "Short-Lived-Climate-Forcers" that contribute
significantly to urban visibility impairment would also be worthy of some attention in the analysis
of policy implications.
12. Nature of the Indicator (Section 4.3.1): What are CASAC's views on the following:
a. Staffs consideration of the three indicators identified in this section and our
conclusions on the appropriateness of these indicators for consideration in this
review?
b. The development and evaluation of a new approach that is based on using
speciated PM2.s mass and relative humidity to calculate PM2.s light extinction by
means of the IMPROVE algorithm?
c. The assessment approach and results comparing the PM components that
contribute to the hours selected in the top percentiles for PM2.s mass and PMio
light extinction?
As noted in past comments, CASAC strongly prefers directly measuring light extinction to using
estimates based on mass measurements (e.g., the other options provided in the Second Draft Policy
Assessment). In their recent review, the Ambient Air Monitoring and Methods Subcommittee
(AAMMS) noted that there are commercial instruments available that provide light extinction
measurements directly, and promising additional technologies may soon become available. The
AAMMS also encouraged the EPA to begin the process of developing performance standards for
PM light extinction measurements. However, a FRM for light extinction measurement does not yet
10
-------
exist, and as EPA does not view it as practical to develop an FRM in time for this rule making,
CASAC recognizes that alternative approaches need to be considered.
A current weakness of the Second Draft Policy Assessment is that it does not explicitly state the
reasons that EPA does not currently recommend using a direct measurement of light extinction. It
also does not provide any indication that the proposed mass-based indicators are intended for use
on an interim basis, to be replaced with direct light extinction-based measurements as those
methods are developed, tested and deployed. If staff consider it impractical to develop performance
standards for an FRM in time for this round of rule making, this should be clearly stated and a
schedule for developing such performance standards and evaluating candidate instruments should
be specified well in advance of the next PM NAAQS review.
Assuming it is currently impractical to develop a FRM for direct measurements of PM light
extinction in a sufficiently timely manner, CASAC agrees that for this rule making, a method to
estimate extinction based on measurements from continuous PM2.5 monitors, preferably adjusted
by PM2.s speciation and relative humidity (RH) data, is appropriate. The "speciated PM2 5 mass-
calculated light extinction" method described in the Second Draft Policy Assessment appears to be
a reasonable approach for estimating hourly light extinction. For purposes of "near real time"
visibility tracking, CASAC recommends considering a simpler calculation in which historical,
rather than concurrent, monthly or seasonal speciation averages would be used to estimate
speciation for combining with real-time continuous PM2.5 and RH data, even though the most
recent speciation data would be used for developing plans for improving visibility. CASAC also
recommends that the Agency consider developing the monthly or seasonal speciation estimates on
a regional basis as well as on a site-specific basis, as this would allow light extinction estimates at
all (>700) sites with continuous PM2.5 data, rather than just the relatively few sites with collocated
continuous PM2 5 and speciation monitors.
13. Alternative Levels and Forms (Section 4.3.3): What are Panel views on the following:
a. The performance assessment which focused on the Candidate Protection Levels of
64,112,191 Mm * for PM2.s light extinction and speciated PM2.s mass-calculated
light extinction, and alternative levels of 10, 20, and 30 ug/m3 for PMi.5 mass
concentration?
These are appropriate CPL and PM2 5 levels. The CPL values were based on all visibility
preference data that are available and bound the study results as represented by the 50%
acceptability criteria. However, the presentation could be improved by expanding some of the
tables to include 10 and 40 dv values, in that at 10 dv, no viewer found the scene to be
unacceptable, and at 40 dv, virtually all viewers found all scenes to be unacceptable. What would
these dv levels correspond to in the context of PM2 5 and the various percentile levels?
b. Use of three-year averaged 90th and 95th percentiles in conjunction with a 1-hour
daily maximum form and use of three-year averaged 98th percentile in
conjunction with the all daylight hours form?
While these levels may be appropriate, they are not well justified. A cursory argument was made
that the 90-95* percentiles in conjunction with the 1-hour daily maximum identified similar days
11
-------
and hours of non-compliance, as did the 98th percentile in conjunction with all daylight hours, and
this correspondence was a sufficient basis to pick these two approaches. It would be informative to
compare all, or at that least the same, percentiles for both all days and the daily hourly daily
maximum. These analyses should be informative as to whether one approach is preferred.
Whether different sources might be identified, depending on use of daily average or maximum
values has not been adequately addressed. For example, a significantly extended episode of low
visibility might be attributed to a single source, such as a large wildfire or prescribed fire, which
would result in the all hour, all day approach targeting only one large emission episode that
occurred for only one or a few time periods. For wintertime episodes in many cities of multi-day
poor urban visibility conditions, the events can cross the end of the calendar year, tracking the
highest daily hour for each day to form a full 3-year distribution of values (i.e., N = -1,095) for
which the compliance value is then compared to the percentile level selected by EPA.
c. Insights to be drawn by comparing the PM components for hours included among
the 10% highest for a 1-hour daily maximum form with the hours included among
the 2% highest for an all daylight hours form, for the various indicators
considered (Appendix C)?
See comments above. These two approaches appear to be similar; however, it would be helpful to
quantify the similarities as opposed to relying only on a qualitative discussion. A scatter plot might
be useful for the 14 sites that provides the average fractional contribution of a species in relation to
the time metric used. Additionally, comparisons should be shown for the specific days found in
non-compliance by metric.
14. Key Uncertainties and Areas for Future Research and Data Collection (Section 4.5):
What are CASAC's views on the areas for future research and data collection outlined in
this section, on relative priorities for research in these areas, and on any other areas that
ought to be identified?
The major areas of research and data collection needed to address key uncertainties related to a
visibility-based secondary standard are nicely captured in Section 4.5 of the Second Draft Policy
Assessment. The section appropriately identifies two major areas of need, one related to visibility
preference, and one related to methods of measurement.
In the first category, preference studies, the details noted by EPA all identify a strong need for
additional urban visibility preference studies conducted using consistent methodology. The range
of 50% acceptability values discussed as possible standards are based on just four studies (Figure
4-2), which, given the large spread in values, provide only limited confidence that the benchmark
candidate protection levels cover the appropriate range of preference values. Studies using a range
of urban scenes (including, but not limited to, iconic scenes - "valued scenic elements" such as
those in the Washington DC study), should also be considered.
In the second category related to methods of measurement, CAS AC supports the proposal to
conduct studies in several cities, pairing direct monitoring of light extinction with enhanced
monitoring of PM size and composition distributions (i.e., continuous PM speciation monitoring).
Additional work should also be conducted to understand the contribution of PMio-2.s in
12
-------
southwestern areas other than Phoenix, to address the lack of information for scattering associated
with this fraction of PMi0 as is noted on page 4-30.
Underlying this overall discussion is a clear need for better particle size - composition distribution
information (i.e., particle composition distributions as a function of particle size). These data gaps
are addressed in different ways in the discussion of future research needs elsewhere in the Second
Draft Policy Assessment (Sections 2.5 and 3.5). Moreover, the development of continuous
monitoring methods for specific PM components addressed in Section 2.5 is equally applicable
here. Improved understanding of size-dependent PM composition would also help address the
questions related to the role of scattering and absorbing aerosols in climate forcing that are raised
in Section 5.2.4.
Finally, a number of research and data collection topics overlap between the secondary PM
NAAQS, and the PM2.s and PMi0 primary PM NAAQS. For example, the fraction of combustion-
related primary carbon PM species can be an important indicator of harmful health effects,
visibility impairment and climate forcing.
With these characteristics, research to jointly quantify and reduce these primary PM carbon species
from combustion sources would advance the information available to the Administrator for her
judgment about the necessary level of protection to be provided by the future PM NAAQS, to be
assessed in the next review cycle.
CASAC suggests that EPA look for additional opportunities to align health and welfare
improvement strategies simultaneously for common indicators, such that the next reviews of the
PM and other NAAQS have not only the analyses of the effects of PM and other NAAQS
indicators on health and welfare, but also include metrics useful for measuring progress toward
attainment.
13
-------
This page left intentionally blank.
-------
APPENDIX B
Ambient PM Monitoring Networks
The measurement of ambient air pollution in the United States is provided through a
number of ambient air monitoring networks operated almost exclusively by state, local, and
tribal air monitoring programs. This section briefly describes the objectives for each of the PM
monitoring networks as well as the coverage for each network across the country.
The ambient air monitoring networks are designed to meet three basic monitoring
objectives. Each objective is important and must be considered individually. The objectives are:
• to support compliance with ambient air quality standards and emissions strategy
development, including comparison to the NAAQS, assess ambient exposures,
development of attainment and maintenance plans, evaluation of regional air quality
models used in developing emission strategies, and tracking trends in air pollution
abatement control measures' impact on improving air quality;
• to provide air pollution data to the general public in a timely manner such as reporting
the Air Quality Index (AQI) through AIRNow (www. airnow. gov), monitoring agency
web sites, conventional media outlets such as newspapers, radio and television news,
and emerging outlets such as social networking sites; and
• to support air pollution research studies, including atmospheric, health, and
epidemiological studies that are used to inform future reviews of the NAAQS.
The sections below briefly summarize the monitoring networks for PM2 5, including
PM2.5 mass, PM2.5 speciation, PMio mass, PMio-2.5 mass, and the forthcoming National Core
(NCore) multi-pollutant network.
B.I PM2.5
The PM2.5 design criteria require State and Local Air Monitoring Stations (SLAMS) in
Metropolitan Statistical Areas (MS As) based on a combination of population and design value
(Table D-5, 40 CFR Part 58) with higher populated locations having more polluted air required
to have the most stations. Background and transport sites are also required of each state with
options for utilizing Interagency Monitoring of Protected Visual Environments (IMPROVE)
stations and other PM2.5 data to provide for flexibility in meeting the requirement.
In urban areas, required PM2.5 monitoring stations are sited to represent community-wide
air quality. These monitoring stations will typically be at neighborhood or urban scale; however,
where a population-oriented micro- or middle-scale PM2.5 monitoring station represents many
such locations throughout a metropolitan area, these smaller scales can be approved by the
B-l
-------
applicable EPA Regional Office to also represent community-wide air quality. The EPA's
existing network design criteria for PM2.5 states: "(1) at least one monitoring station is to be
sited in a population-oriented area of expected maximum concentration and (2) for areas with
more than one required SLAMS, a monitoring station is to be sited in an area of poor air quality"
(40 CFR, PART 58, Appendix D, section 4.7). Since monitors sited for either of these network
design criteria must represent community-wide air quality, they are also representative of
population exposure. Most PM2 5 monitoring in urban areas should be representative of a
neighborhood scale.
In rural areas, the design of the network relies on IMPROVE, rural NCore stations, a
limited number of smaller cities, and partner monitoring agencies to provide for regional
characterization of PM2 5. Stations in these areas are typically sited to represent regional scale air
quality and are therefore located away from any local sources, should they exist.
B.1.1 PM2.5 Federal Reference Method (FRM) Network
The network of PM2 5 FRMs has been operational since 1999. This network includes
over 900 monitoring stations throughout the country. The FRMs are primarily used to determine
compliance with the NAAQS, but also serve other objectives. Since FRMs are filter-based
methods requiring post-sampling laboratory analysis, which is labor intensive, monitoring
programs have some flexibility in how often they must be operated. Approximately 150 FRMs
operate every day, 600 every third day, and another 150 every sixth day. Sample frequencies of
every third and sixth day are based on a national sample calendar provided by EPA. The number
of PM2 5 FRMs may decrease over the coming years as PM2 5 continuous FEMs1 are now
available and can replace FRMs for purposes of comparison to the NAAQS. Figure B-l
illustrates the locations of PM2 5 FRMs reporting to the Air Quality System (AQS).
B.I.2 PM2.s Continuous Monitor Network
Continuous PM2 5 monitors are required in MSAs at one half (rounded up) the number of
monitoring stations that are required to have an FRM/FEM monitor. Since most deployed PM2.5
continuous monitors are not approved as FEMs, many of these monitors are collocated at
monitoring locations with an FRM so that the availability of data from both instruments supports
each of the major monitoring objectives described earlier in this section. Collocation with PM2.5
FRMs and continuous monitors also ensures that reference data are available to validate the
performance of the continuous monitor. While PM2 5 continuous monitors primarily support
forecasting and reporting the AQI, they are also used in interpreting the diurnal characterization
1 A Federal Equivalent Method (FEM) means a method for measuring the concentration of an air pollutant in the
ambient air that has been designated as an equivalent method in accordance with 40 CFR Part 53.
B-2
-------
of PM2.5. The network of PM2.5 continuous monitors has grown to over 700 locations throughout
the country with approximately 160 monitors in operation as continuous FEMs. Continuous
FEMs are compared to the NAAQS when designated as the primary monitor at a station. A
recent assessment of continuous PM2.5 FEM data quality, as compared to collocated FRMs,
indicates that some FEMs are meeting the performance criteria used to approve these methods,
which are based on daily measurements, while others are not. The assessments and
recommendations for addressing data quality issues are detailed in Hanley and Reff, 2011.2
Additionally, as part of a review of 1-hour data from all continuous FEM PM2.5 instruments
operating at state/local monitoring sites, we have recently become aware of the occurrence of
questionable outliers in 1-hour data submitted to AQS from continuous FEM PM2.5 instruments
at some of these sites (Evangelista, 2011).3 Figure B-2 illustrates the locations of PM2 5
continuous monitors reporting to AQS, including those that are now approved as FEMs.
B.I.3 PM2.s Speciation
As part of the PM2 5 NAAQS review completed in 1997, EPA established a PM2 5
Chemical Speciation Network (CSN) consisting of 54 Speciation Trends Network (STN) sites.
The STN was established to conduct routine Speciation monitoring in primarily urban areas to
provide nationally consistent data for the assessment of trends and to provide a long-term record
of the chemical composition of PM2 5 in the United States. The initial STN monitoring began
with a pilot of 13 sites in February 2000. In addition to the STN, EPA also implemented a
network of about 200 supplemental Speciation sites for multiple monitoring objectives including
support for: the development of modeling tools and the application of source apportionment
modeling for control strategy development to implement the NAAQS; health effects and
exposure research studies; assessment of the effectiveness of emission reductions strategies
through the characterization of air quality; and development of state implementation plans
(SIPs). The STN and supplemental speciation monitoring sites together are referred to as the
CSN. The CSN sampling systems do not include any FRM/FEMs; therefore, data produced from
this network are not used for comparison to the NAAQS. However, FRMs are almost always
collocated with the CSN since these are among the most important PM2.5 sites in a network.
2 Hanley, T._and Reff, A (2011). Assessment of PM2 5 FEMs Compared to Collocated FRMs. Memorandum to PM
NAAQS Review Docket, EPA-HQ-OAR-2007-0492. April 7, 2011. Available:
http://www.epa.gov/ttn/naaqs/srandards/pm/sjm 2007 td.html.
3 Evangelista, M. (2011). Investigation of 1-hour PM25 Mass Concentration Data from EPA-Approved Continuous
Federal Equivalent Method Analyzers. Memorandum to PM NAAQS Review Docket (EPA-HQ-OAR-2007-0492).
Aprils, 2011. Available: http://www.epa.gOv/ttn/naaqs/standards/pm/sjm_2007_td.html.
B-3
-------
In 2005, EPA conducted an assessment specifically focused on the PM2.5 speciation
monitoring network. In consultation with State and local monitoring agencies, EPA evaluated
CSN sites to determine which ones might be shut down to provide resources for future
monitoring needs. The EPA ranked the sites according to their overall information value. The
ranking was based on several factors, including whether the site was in a non-attainment area and
B-4
-------
Figure B-l. PM2.5 FRMs Reporting to AQS
Legend
• PM2.5FRMS
£3> Lakes
State Boundaries
PM2.5 FRMs Reporting to AQS
B-5
-------
Figure B-2. PM2.s Continuous Monitors Reporting to
AQS
Legend
o Thermo BAM
* FDMS
* M1 BAM
* TEOM
« Nephelometer
A FEM-Thermo 5030 SHARP
FEM-Thermo FDMS 1405-DF
A FEM-Thermo FDMS 8500C
A FEM -M1 BAM 1020
£3 Lakes
State Boundaries
PM2.5 Continuous Monitors
Reporting to AQS
Alton Fr
-------
whether other sites were nearby. There was general agreement that some of the sites should be
shut down when FY 2005 funding ran out. Other sites were identified as high value sites,
particularly with regard to the PM2 5 NAAQS program. The EPA evaluated each of these sites
when FY 2006 regional funding allocations for continued operation and maintenance were
developed. In doing so, EPA balanced filter-based PM2.5 speciation against funding for other
PM2.5 measurements, such as FRM site operations, filter analysis, and startup of additional
precursor gas sites and continuous speciation sites.
As of 2010, the PM2.5 CSN includes about 50 STN sites and about 150 SLAMS
supplemental sites. All STN sites operate on a one-in-three day sample collection schedule. A
majority of the SLAMS supplemental sites operate on a one-in-six day sample collection
schedule. These sites collect aerosol samples over 24 hours on filters that are analyzed for PM2 5
mass, a number of trace elements, major ions (e.g., sulfate, nitrate, and ammonium), and organic
and elemental carbon.
The IMPROVE program was established in 1985 to aid the creation of federal and state
implementation plans for the protection of visibility in Class 1 areas (155 national parks and
wilderness areas) as stipulated in the 1977 amendments to the CAA and further supported in the
goals set forth in the 1999 Regional Haze Rule.4 Similar to the CSN, the IMPROVE program
provides PM2 5 mass and speciation data for organic and elemental carbon, major ions, and trace
elements; however, unlike CSN IMPROVE also samples for PMio mass. The IMPROVE
network is presently comprised of 110 regionally representative monitoring sites; 7 sites operated
collaboratively with the Clean Air Status and Trends Network (CASTNET); 30 sites operated by
State and local agencies referred to as "IMPROVE Protocol Stations"; several sites operated by
Tribal air monitoring programs; and about 9 sites operated by federal land managers. IMPROVE
samplers are also collocated at 5 CSN stations to assess on-going comparability of the programs.
Figure B-3 provides a map of all IMPROVE and CSN stations.
From May 2007 until October 2009, the approximately 200 CSN sites transitioned to a
new method of sampling and analysis for carbon measurements that is consistent with the
IMPROVE network methodology. This transition occurred in three phases: 56 CSN stations
began using the new method in the first phase which started May 1, 2007; 63 stations were in the
second phase beginning on April 1, 2009; and 78 stations were in the third phase beginning on
October 1, 2009.
While the network of approximately 200 CSN sites provide valuable data for
development and tracking of control strategies, its use for supporting studies of health and
welfare effects is somewhat limited. The CSN sites provide data on a one-in-three or one-in-six
4 Additional information is available at http://www.epa.gov/visibilitv/actions.html
B-7
-------
Figure B-3. IMPROVE and Chemical Speciation Network
Legend
A Trend sCSN
• IMPROVE and IMPROVE Protocol Stations
c Supplemental CSN
£> Lakes
State Boundaries
IMPROVE and Chemical Speciation Network
ill*r> Projection
• *nl(.J r.>]|.Jk.', -E*
i: |::=H Pa-allel: 35
B-8
-------
day schedule and do not capture data every day or everywhere. In April 2008, EPA co-
sponsored a workshop to discuss modifications to the current ambient air quality monitoring
networks that would advance our understanding of the impacts of PM2.5 exposures on public
health/welfare in the most meaningful way. This workshop was a major step in a series of
interactions to foster improved long-term communication between external stakeholders,
including air quality monitoring experts and health researchers. These researchers expressed a
strong interest in having access to PM2 5 speciation measurements collected more frequently.5
B.2 PMio
Measurements from PMio monitoring stations are primarily used for comparison to the
NAAQS; however, most serve multiple objectives. PMio monitoring stations generally have an
urban focus and are required in MSAs according to population and concentrations relative to the
NAAQS. Local considerations are also a factor in determining the actual required number of
monitoring sites. More stations are required in larger MSAs and MSAs with more evidence of
poor air quality, while monitors are also required in clean MSAs of certain size. The number of
monitors in areas where MSA populations exceed 1,000,000 must be in the range from 2 to 10
stations, while in low population urban areas no more than two stations are required (see Table
D-4 of Appendix D to CFRPart 58). Because sources of air pollutants and local control efforts
can vary from one part of the country to another, some flexibility is allowed in selecting the
actual number of PMio stations in any one locale.
The network of PMio monitors has been operational since 1987. The network currently
includes over 800 monitoring stations throughout the country with most metropolitan areas
operating more PMio monitors than required by current monitoring requirements. The PMio
monitoring stations operate FRMs using different sampling frequencies including: daily, one-in-
two day, one-in-three day, or one-in-six day sampling. The sampling frequency is based on the
relative concentration level of the site with respect to the 24-hour standard. There are also FEMs
that are operated continuously. PMio monitors operating across the country are almost
exclusively FRMs or FEMs. Figure B-4 illustrates the locations of the PMio FRMs and FEMs
reporting to AQS.
The PMio monitoring stations are currently required to collect and report monitoring data
under standard temperature and pressure (STP) conditions. PM2 5 and PMi0-2.5 are required to be
collected and reported at local conditions. Correction of the sampled aerosol volume to
"standard" conditions is typically small (e.g., less than a few percent) except in locations at
' A summary of the workshop including recommendations was published in December 2008 and is available at
www.epa.gov/ORD/npd/pdfs/FINAL-April-2008-AQ-Health-Research-Workshop-Summarv-Dec-2008.pdf.
B-9
-------
Figure B-4. PMi0 Monitoring Network Reporting to AQS
v-s-n yv i- xjZ-«^
• Sj—i ( f^LfST •*
IOWA« J .fc- T—V>4VMEMIISYLUAU
• V i • * Bt---;
•_ *7[» ', ,^e- ^^ji
Luiui:*J-««—\9 ,' v Uak'
Legend
PM10 dichot sampler
PM1 0 low-volume sampler
PM1 0 continuous monitor
PM1D hi-volume sampler
Lakes
State Boundaries
PM10 Monitoring Network Reporting to AQS
B-10
-------
higher altitudes and those with large diurnal or seasonal temperature changes (US EPA, 1996,
Volume I, Section 4.2.4).
B.3 PMio.2.5
The EPA is requiring approximately 80 PMio-2.5 monitoring stations as part of the
National Core (NCore) multiple pollutant network. This network is planned to be fully
operational by January 1, 2011.6 As State and local agencies measure PMio-2.5 mass as required
through the NCore program, we anticipate the number of stations will increase. In addition to
NCore, it is appropriate to calculate PMio-2.5 concentrations from collocated PM2.5 and PMio
methods of the same make and model. However, in many cases the PMio data will need to be
adjusted to local conditions since most PMio data are reported as STP, while PM2.5 is reported at
local conditions (LC). Note: only low-volume samplers are used in calculating PMio-2.5 as these
are the only approved reference and equivalent methods for this measurement.7
For PMio-2.5 speciation, we do not expect methods to be fully developed in time to meet
the January 1, 2011, start date for monitoring at NCore. The EPA has been working with the
CASAC Ambient Air Monitoring and Methods Subcommittee (AAMMS) on this issue and has
implemented a pilot program evaluating PMio-2.5 speciation methods at two locations in 2010.8
B.4 NATIONAL CORE (NCORE)
The NCore network is a multi-pollutant network that includes measurements of particles,
gases, and meteorology (71 FR 61236, October 17, 2006). The network is intended to support
integrated air program management needs. While measurements made at NCore stations will
often be used for comparison to one or more NAAQS as well as for public reporting of air
quality data, their enhanced value is in providing multi-pollutant data for validation of models,
trends, and as an input to research studies to determine the relative importance of collocated
pollutants. The NCore monitoring network began January 1, 2011 at approximately 80 stations.
NCore stations are intended to be long-term stations useful for a variety of applications. The
locations of these stations are specified to be placed away from direct emissions sources that
could substantially impact the ability to detect area-wide concentrations. The NCore network is
comprised of stations in both urban and rural areas. Urban NCore stations are to be generally
6 Monitoring agencies are bringing their PM10_25 methods online now and, as of April 2011, 21 NCore sites are
already reporting data to the AQS.
7 A list of approved FRMs and FEMs is available on EPA's web site at: http://www.epa.gov/ttn/amtic/criteria.html.
8 A February 2009 consultation with the CASAC AAMM subcommittee discussed issues related to coarse particle
speciation measures. For more information on this consultation, please see:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/3494de4dOccb394485257463006
4d4e4! OpenDocument
B-ll
-------
located at an urban or neighborhood scale to provide representative concentrations of exposure
expected throughout the metropolitan area. Rural NCore stations are to be located, to the
maximum extent practicable, at a regional or larger scale away from any large local emission
source, so that they represent ambient concentrations over an extensive area. States and where
applicable, local monitoring agencies submitted plans for meeting the NCore requirements
during the summer of 2009. The proposed NCore stations are collocated with several other well
leveraged networks such as CSN, Photochemical air monitoring stations (PAMS), National Air
Toxics Trends Stations (NATTS), IMPROVE, and CASTNET. The expected NCore locations
are identified in Figure B-5.
B-12
-------
Figure B-5. NCore Monitoring Network
/PartfJc i_
Ocemi^ ^
Rural NCors
Urban NCore
£5, Lakes
State Boundaries
National Core (NCore) Multipollutant Network
B-13
-------
This page left intentionally blank.
-------
APPENDIX C
Table C-l. Predicted Percent of Counties with Monitors (and percent of population in counties with monitors)
Not Likely to Meet Current and Alternative Annual and 24-hour PM2.s Standards
Region >
Total # of counties >
Total population (x 1,000)>
All U.S.
532
184,180
annual
Hg/m3
15
24-
hour
Hg/m3
35
Statistic
# counties
population
% # counties
% population
Northeast
93
44,345
Southeast
149
40,271
Industrial
Midwest
135
37,512
Upper
Midwest
45
7,694
Southwest
17
8,962
Northwest
69
20,821
Southern
California
17
22,663
Outlying
areas
7
1,913
Current Standards
Numbers of counties, populations, and percentages of total
35
28,801
7%
16%
0
0
0%
0%
1
662
1%
2%
5
3,683
4%
10%
0
0
0%
0%
1
180
6%
2%
19
6,615
28%
32%
8
17,579
47%
78%
1
83
14%
4%
Alternative Standards
annual
Hg/m3
12
11
13
24-
hour
Hg/m3
35
35
30
Statistic
# counties
population
% # counties
% population
# counties
population
% # counties
% population
# counties
population
% # counties
% population
Numbers of counties, populations, and percentages of total
149
76,579
28%
42%
263
107,447
49%
58%
130
78,286
24%
43%
21
14,936
23%
34%
39
22,952
42%
52%
21
13,688
23%
31%
30
10,318
20%
26%
82
21,224
55%
53%
10
6,152
7%
15%
66
23,998
49%
64%
102
31,346
76%
84%
50
23,692
37%
63%
0
0
0%
0%
5
1,491
11%
19%
2
1,627
4%
21%
2
218
12%
2%
o
6
3,290
18%
37%
o
6
393
18%
4%
20
6,634
29%
32%
22
6,668
32%
32%
32
12,210
46%
59%
9
20,393
53%
90%
9
20,393
53%
90%
10
20,411
59%
90%
1
83
14%
4%
1
83
14%
4%
2
114
29%
6%
C-l
-------
This page left intentionally blank.
-------
APPENDIX D
Table D-l. Predicted Percent of Counties with Monitors (and percent of population in counties with monitors)
Not Likely to Meet Current and Alternative 24-hour PMio Standards
Region>
Total # of counties >
Total population (xl,000)
>
All U.S.
307
120,090
Northeast
37
15,397
Southeast
57
27,181
Industrial
Midwest
50
21,352
Upper
Midwest
40
5,917
Southwest
25
11,112
Northwest
77
15,270
Southern
California
18
22,695
Outlying
areas
3
1,167
Current Standard: 3-Year Expected Exceedance Equivalent Design Value
Hg/m3
150
Statistic
# counties
population
% # counties
% population
Numbers of counties, populations, and percentages of total
41
32,835
13%
27%
0
0
0%
0%
o
6
4,626
5%
17%
0
0
0%
0%
1
14
3%
0%
11
5,485
44%
49%
13
1,878
17%
12%
12
20,571
67%
91%
1
260
33%
22%
Alternative Standards: 24-hour level (3-year average 98th percentile)
Hg/m3
87
85
80
75
70
65
Statistic
# counties
population
% # counties
% population
# counties
population
% # counties
% population
# counties
population
% # counties
% population
# counties
population
% # counties
% population
# counties
population
% # counties
% population
# counties
population
% # counties
% population
Numbers of counties, populations, and percentages of total
37
20,515
12%
17%
39
21,887
13%
18%
45
24,535
15%
20%
55
35,703
18%
30%
71
43,823
23%
36%
87
49,394
28%
41%
0
0
0%
0%
0
0
0%
0%
0
0
0%
0%
0
0
0%
0%
0
0
0%
0%
2
775
5%
5%
2
4,063
4%
15%
2
4,063
4%
15%
2
4,063
4%
15%
o
J
4,626
5%
17%
4
4,644
7%
17%
4
4,644
7%
17%
2
507
4%
2%
3
1,789
6%
8%
4
3,183
8%
15%
6
3,491
12%
16%
7
8,868
14%
42%
9
10,421
18%
49%
2
552
5%
9%
2
552
5%
9%
3
599
8%
10%
5
637
13%
11%
7
881
18%
15%
10
1,029
25%
17%
11
5,924
44%
53%
12
6,014
48%
54%
13
6,131
52%
55%
13
6,131
52%
55%
13
6,131
52%
55%
14
7,507
56%
68%
10
1,789
13%
12%
10
1,789
13%
12%
11
1,833
14%
12%
15
2,570
19%
17%
27
5,052
35%
33%
33
5,989
43%
39%
9
7,421
50%
33%
9
7,421
50%
33%
11
8,467
61%
37%
12
17,986
67%
79%
12
17,986
67%
79%
13
18,739
72%
83%
1
260
33%
22%
1
260
33%
22%
1
260
33%
22%
1
260
33%
22%
1
260
33%
22%
2
290
67%
25%
D-l
-------
This page left intentionally blank.
-------
APPENDIX E
Information Regarding a 1-hour PM2.s Mass Indicator
This Appendix presents information on 2005-2007 levels of 1-hour PM2.5 mass
concentrations in 14 urban study areas1 and on the "what if PMio light extinction conditions that
would exist if the study areas met each of 10 alternative secondary PM NAAQS scenarios based
on a 1-hour PM2.5 mass indicator. With respect to the latter subject, this Appendix is therefore
similar to Chapter 4 of the Paniculate Matter Urban-Focused Visibility Assessment (UFVA, US
EPA, 201 Ob)2, which presented similar information for 18 secondary PM NAAQS scenarios
based on PMio light extinction as the indicator, for the current annual and 24-hour PM2.5
NAAQS, and for a scenario with an annual NAAQS of 12 |ig/m3 and a 24-hour NAAQS of 25
|ig/m3.
In section B.I.2 of Appendix B, it was noted that recent investigations have identified
issues with data quality for measurements of both 24-hour concentrations and 1-hour
concentrations from continuous Federal Equivalent Methods for PM2.5. This appendix on a 1-
hour PM2.5 mass indicator is based only on 2005-2007 data from continuous instrument models
that pre-date the introduction of continuous Federal Equivalent Methods for PM2.5. While no
systematic investigation was performed on these 2005-2007 data similar to the investigations
described in Hanley and Reff, 2011 and Evangelista, 2011,3'4 no similar issues were noticed in
the course of using these 2005-2007 data.
E. 1 INDICATOR AND MONITORING METHOD
As in Chapter 4 of the UFVA, this Appendix excludes from all NAAQS scenarios and
results all non-daylight hours and all daylight hours with relative humidity greater than 90
percent. This applies to both the definition of 10 secondary NAAQS scenarios, and to graphics
and tables that characterize ambient conditions. While ambient humidity should not affect
1 The UFVA assessed light extinction conditions in 15 study areas, one of which was St. Louis. The PM10_25 values
for St. Louis in the UFVA were determined by difference between two nearby monitoring sites. Based on comments
received on the draft UFVA to the effect that the PM10 monitoring site used for St. Louis was not representative of
the St. Louis urban area, EPA does not consider the PM10_2.5 concentrations for St. Louis to be credible. The PM10
light extinction results for St. Louis have therefore been excluded from tables and figures in this appendix that
involve PM10 light extinction results.
2 Particulate Matter Urban-Focused Visibility Assessment Final Document, EPA 452/R-10-004, July 2010.
Available: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
3 Hanley, T and Reff, A (2011). Assessment of PM2 5 FEMs Compared to Collocated FRMs. Memorandum to PM
NAAQS Review Docket EPA-HQ-OAR-2007-0492^ April 7, 2011. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_td.html.
4 Evangelista, M. (2011). Investigation of 1-hour PM25 Mass Concentration Data from EPA-Approved Continuous
Federal Equivalent Method Analyzers. Memorandum to PM NAAQS Review Docket (EPA-HQ-OAR-2007-0492).
Aprils, 2011. Available: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 td.html.
E-l
-------
conventional measurement approaches for 1-hour PM2.5 mass, the issue of co-occurrence of high
humidity levels with light extinction due to natural conditions would still apply. See section
3.3.5 of the UFVA, and section 4.2.1 of this document at "Current Visibility Levels." The
assumed hours of daylight are the same as those used in the UFVA, as shown in Table 3-5 of the
UFVA.
All values for 1-hour PM2.5 mass concentration in this appendix come from the
continuous instruments at the 14 urban study sites, with no adjustment to make these values
consistent with the collocated 24-hour FRM measurement of PM2.5 mass. Appendix A of the
UFVA provides details on the type of continuous instrument at each study site. TEOMs were
used at all sites except for beta attenuation instruments in Fresno and Philadelphia, nephelometer
instruments in Tacoma and Phoenix, and an FDMS instrument in Salt Lake City.
For conciseness, only the daily maximum daylight 1-hour PM25 mass concentration
indicator is considered in this Appendix. It would also be possible to construct alternative
NAAQS scenarios of an all-hours type, which could be analyzed in the same manner as
presented in this Appendix.
E.2 CURRENT CONDITIONS OF 1-HOUR PM2.5 MASS
Figure E-l is a box plot of 2005-2007 daily maximum daylight 1-hour PM2.5 mass
concentrations for the 14 study areas, excluding hours with relative humidity greater than 90
percent, to give a sense of the range and central tendency of this parameter. The horizontal
reference lines are at 10, 20, 30, 40 and 60 |ig/m3. The relative positions of the 90th percentile
concentrations (indicated by the horizontal stroke at the top of the whisker) are generally
consistent with the relative ranking of these sites according to their design values for the 24-hour
PM2.5 NAAQS (see Table 3-2 of the UFVA); similarly, the relative positions of the median
concentrations are generally consistent with the annual PM2.5 design values. Table E-l, based on
the same data as Figure E-l, presents the percentage of days in 2005-2007 on which the daily
maximum daylight 1-hour PM2.5 concentration exceeded the reference levels represented by the
horizontal lines in Figure E-l.
E-2
-------
Figure E-l. 2005-2007 Daily Maximum Daylight 1-hour PM2.s Mass Concentrations
(ug/m3) for the 14 Study Areas (excluding hours with relative humidity > 90 %)
m cj
E S
C
o
o
U
27D 141 27?
143 225
T
&
Table E-l. Percentage of Days with Daily Maximum Daylight 1-hour PM2.s Mass
Concentration Exceeding Reference Levels in 2005-2007 (excluding hours with relative
humidity > 90 %)
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Number of Days
with Estimates
109
324
300
86
306
273
148
349
279
141
277
181
143
225
1-hour PM2.5 Mass Reference Level (|ig/m3)
10
50
88
92
60
64
75
80
92
86
92
94
90
99
75
20
11
62
81
8
20
25
42
60
56
72
57
46
84
43
30
1
37
67
1
11
5
14
37
28
52
28
22
63
25
40
0
20
46
1
7
0
5
23
10
36
15
8
45
13
60
0
8
20
1
2
0
0
8
1
13
3
1
20
3
E-3
-------
E.3 ALTERNATIVE NAAQS SCENARIOS BASED ON 1-HOUR PM2.5 MASS AS
THE INDICATOR
To ensure examination of a wide enough range of alternative standards based on 1-hour
PM2 5 mass to encompass the range of standards that might be considered as alternatives to the
PMio light extinction NAAQS scenarios examined in Chapter 4 of the UFVA, we considered
levels of 10, 20, 30, 40, and 60 |ig/m3. Only the daily maximum daylight hour form was
considered. Each level was combined with two statistical forms: the three-year average of the
annual 90th percentile value and the three-year average of the annual 95th percentile value. For
ease of reference, these scenarios are designated by letters from "aa" to "jj" and listed in Table
E-2. Looking somewhat ahead to results presented below, the scenarios are arranged in Table E-
2 in order of least to most stringent in terms of the reductions in ambient PM2.5 needed from
current levels to meet the current and alternative NAAQS levels and forms.
Table E-2. Alternative NAAQS Scenarios Based on Daily Maximum Daylight 1-Hour
PM2.s Mass, Averaged Over Three Years
(excluding hours with relative humidity > 90 %)
NAAQS Scenario
aa
bb
cc
dd
ee
ff
gg
hh
ii
jj
Level (|ig/m )
60
60
40
40
30
30
20
20
10
10
Statistical Form
3 -year average of 90th percentile
3 -year average of 95th percentile
3 -year average of 90th percentile
3 -year average of 95l percentile
3-year average of 90 percentile
3 -year average of 95th percentile
3-year average of 90th percentile
3 -year average of 95 percentile
3 -year average of 90l percentile
3 -year average of 95 percentile
E.4 APPROACH TO MODELING "WHAT IF" CONDITIONS OF PMio LIGHT
EXTINCTION FOR ALTERNATIVE SECONDARY NAAQS BASED ON 1-
HOURPM2.5MASS
Before modeling "what if conditions, we augmented the data set described in Table 3-4
of the UFVA in the same manner as described in section 4.1.4 of the UFVA, to achieve seasonal
balance despite the lack of monitoring data for one quarter in each of Houston and Phoenix. In
Tacoma and Phoenix, which had data only for two years in the 2005-2007 period, we averaged
the percentile values from the only two available years rather than the three years defined for the
statistical form of the NAAQS scenarios.
The modeling of daily maximum daylight 1-hour PM2 5 mass under each of the scenarios
listed in Table E-2 used a rollback approach that combined relevant concepts and steps from the
rollback methods described in sections 4.1.4 (for PMio light extinction scenarios) and 4.2.2 (for
scenarios based on annual average and 24-hour average PM^.s) of the UFVA. The following are
the steps in the modeling.
E-4
-------
1. Identify the 90th percentile daily maximum daylight 1-hour PM2.5 mass value in each
of 2005, 2006, and 2007 for a study area. Average these to determine the 3-year
average design value for that percentile form. Repeat for the 95th percentile form.
These design values are presented in Table E-3. They range from 22 to 81 |ig/m3,
indicating that some study areas meet some of the NAAQS scenarios under current
conditions. In such cases, PM2 5 concentrations were not adjusted, i.e., there was no
"roll up" for any area in any scenario.
2. Using the same days and hours as contributed by the three annual 90th percentile
values for actual 1-hour PM2 5 mass, find the three corresponding values of policy
relevant background (PRB) 1-hour PM2.5 mass. Average these three annual values of
PRB 1-hour PM2.5 to obtain the 3 year average PRB portion of the actual 1-hour
PM2.5 design value for the 90th percentile form. Repeat for the 95th percentile form.
In the modeling for the NAAQS scenarios examined in the UFVA, PRB for 1-hour
PM2.5 mass was not explicitly calculated because it was not needed in the rollback
modeling for the scenarios addressed in the UFVA. Therefore, it was necessary to
reconstruct this parameter by adding the values for the PRB concentrations of the five
components of PM2 5: nitrate, sulfate, elemental carbon, organic carbon material, and
soil. The method for estimating PRB for these five components is described in
Appendix C of the UFVA.
3. Subtract the value from step 2 from the value from step 1, to determine the non-PRB
portion of the 1-hour PM2.5 mass design value.
4. Calculate the percentage reduction required in non-PRB 1-hour PM2.5 mass in order
to reduce the design value to the level that defines the NAAQS scenario, using the
following equation:
Percent reduction required = 1 - (NAAQS level - PRB portion of the design value)
(non-PRB portion of the design value)
The percentage reductions determined in step 4 are shown in Table E-4. Note that for
some combinations of area and scenario no reduction is required because the 2005-
2007 design value already meets the NAAQS scenario. These cases with a required
percentage reduction of zero are shaded blue in Table E-4.
5. Turning to the entire set of day/hour-specific actual and PRB daylight 1-hour
concentrations of the five PM2.5 components for the three (or two) year period,
determine the non-PRB portion of each of the five components in an hour by
subtracting the PRB value from actual value, reduce it by the percentage determined
in step 4, and add back in the PRB 1-hour concentration of the component.
6. Finally, calculate PMio light extinction using the reduced values of the five
components, the original value of 1-hour PMio-2.5, and the 1-hour value of f(RH),
according to the following equation for PMio light extinction (see section 3.2.3 of the
UFVA for an explanation of the variables in this equation).
E-5
-------
= 3 x f(RH) x [Sulfate]
+ 3 xf(RH)x [Nitrate]
+ 4 x [Organic Mass]
+10 x [Elemental Carbon]
+ 1 x [Fine Soil]
+ 0.6 x [Coarse Mass]
These steps assume that in order to meet a PM NAAQS scenario based on 1-hour PM2.5
as the indicator, each component of PM2 5 is reduced by an equal percentage, across the five
components and across all hours. In actual implementation of such a NAAQS, each state would
develop an attainment strategy, which might result in unequal percentage reductions of the
components. If the strategy emphasized reductions in the fine soil component, for example, PM
light extinction levels would remain high relative to those estimated by these steps, because fine
soil is not efficient in terms of reducing visibility compared to the other four components on a
dry mass-to-mass basis. On the other hand, a strategy that involves relatively large reductions in
sulfate or nitrate would achieve greater reductions in PM light extinction than estimated by these
steps. The uncertainty in how the results of this rollback method compare to the results of actual
attainment strategies should be kept in mind when comparing the results of "what if scenarios
for NAAQS based on PM2.s mass as the indicator versus scenarios based on PMi0 light
extinction. Unlike the effect of humidity variation between areas, this source of uncertainty is
not reflected in any of the results presented in this Appendix and will not be apparent in
comparisons of results in this Appendix to results presented in the UFVA for NAAQS scenarios
based on PMi0 light extinction.
These steps also assume no change in PMi0-2.5 concentrations between current conditions
and "what if conditions. While reductions in PMio-2.5 would not be needed to meet a secondary
NAAQS based on 1-hour PM2.5 mass, it is possible that strategies to control PM2.5 concentrations
might also achieve reductions in PMio-2.5 concentrations because some sources emit both and
some control methods achieve some reductions in both. However, in most of the 14 study areas,
PMio-2.5 makes a small contribution to estimated PMio light extinction. For those areas for which
no local data on PMio-2.5 concentrations were available, a low contribution of PMio-2.5 to light
extinction was inevitable because of the method used to fill the PMio-2.5 data gap. That method
was to apply a long-term regional ratio of PMio-2.5 to PM2.5 to the local hourly PM2.5
concentration, which by its nature cannot produce estimates of hourly PMio-2.5 that are as high as
might actually be created by shorter-term PMi0.2.5 emissions episodes
E-6
-------
Table E-3. 2005-2007 Design Values for 1-Hour PM2.5 Mass (jig/m )
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Percentile Form
90th
22
55
72
20
32
26
33
55
40
64
46
37
67
44
95th
27
66
81
24
45
29
37
74
45
79
51
43
77
55
Table E-4. Percentage Reductions in non-PRB PM2.s Components Required to Meet
NAAQS Scenarios based on 1-Hour PM2.s Mass
Scenario
Level (|ig/m )
Percentile Form
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
aa
60
90
bb
60
95
cc
40
90
dd
40
95
ee
30
90
ff
30
95
gg
20
90
hh
20
95
ii
10
90
ii
10
95
Percentage Reduction
0
0
17
0
0
0
0
0
0
7
0
0
10
0
0
10
26
0
0
0
0
19
0
24
0
0
22
0
0
28
45
0
0
0
0
28
0
38
13
0
40
8
0
40
51
0
12
0
0
46
12
50
22
8
49
28
0
46
59
0
7
0
9
46
25
54
35
19
55
32
0
55
64
0
34
0
20
60
34
63
42
31
62
46
11
65
73
0
39
23
40
65
51
70
57
47
71
55
27
71
76
15
56
34
49
74
57
75
62
55
75
65
60
83
87
51
70
64
71
84
77
85
79
74
86
78
69
86
88
58
78
69
78
87
80
88
81
78
88
83
E-7
-------
E.5 1-HOUR PM2.5 MASS RESULTS FOR "JUST MEETING" ALTERNATIVE
SECONDARY NAAQS SCENARIOS BASED ON 1-HOUR PM2.5 MASS
As a check on the reasonableness of the rollback method described in section 4.0 and on
the accuracy of the code used to implement it, it is of interest to examine the distribution of the
levels of 1-hour PM2.5 that result from the method. Ideally, after rollback any area that had a
non-zero required reduction should have a post-rollback design value for 1-hour PM2.5 mass that
is exactly equal to the target design value. Also, there should be a progression of reductions in 1-
hour PM2.5 medians and other percentile points on the distribution as progressively more
stringent scenarios are modeled.
Table E-5 shows the post-rollback 1-hour PM2.5 mass design values for the scenarios,
with percentile forms matched. Design values for area-scenario combinations for which the
required reductions were zero have been omitted, because the current conditions design values
for these combinations would not be expected to reflect the target design value. It can be seen
that the design values progress as expected and are in the vicinity of the target design values, but
are not always exactly equal to the targets. EPA staff attributes this to the fact that PRB
concentrations of 1-hour PM2.5 mass vary from hour to hour. It is possible for the daily
maximum PM2.5 mass concentration on a certain day in 2005 with a percentile rank of, for
example, 96th to have a relatively small PRB portion and a large non-PRB portion compared to
the daily maximum concentration that ranks 95th. When an equal reduction is made to the non-
PRB portion of each total concentration, the two values may switch rank positions, and so a new
day and hour becomes the 2005 contributor to the rolled back three-year design value. Since this
day and hour was not used to determine the required percentage reduction, the resulting design
value will not exactly meet the target design value. It would be possible to iterate with higher
and lower percentage reductions until the rolled back design value exactly matched the target
design value, but EPA considered this degree of refinement to be unnecessary in order to meet
the objectives of the Policy Assessment, given other uncertainties in the underlying data and in
the assumptions used to estimate PMio light extinction values.
Staff also generated and examined box plots of daily maximum daylight 1-hour PM2.5
mass concentrations as a check for conceptual or programming errors, and found them to match
expectations. They are not included here, for conciseness.
E-8
-------
Table E-5. Post-rollback Design Values for Daily Maximum 1-Hour PM2.s Mass
Design values are shown only for combinations of study area and scenario for which the study area does not meet
the scenario under current conditions, such that reductions were made during the rollback modeling.
Scenario
Level (|ig/m3)
Statistical Form
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
aa
60
90th
bb
60
95th
cc
40
90th
dd
40
95th
ee
30
90th
ff
30
95th
gg
20
90th
hh
20
95th
ii
10
90th
ii
10
95th
Corresponding Design Value (|ig/m ) (same percentile form as the scenario)
53
52
46
63
53
58
59
44
40
35
42
34
33
31
42
42
35
38
39
36
39
33
38
30
40
30
26
29
29
32
28
26
24
31
23
32
31
26
28
27
29
27
29
25
28
22
30
20
20
18
19
23
19
21
19
17
16
21
16
21
21
21
18
19
19
23
18
20
18
20
17
19
15
20
11
10
9
10
10
12
10
11
10
9
8
10
8
11
12
10
9
10
10
11
9
10
10
10
9
10
8
10
E.6 PMio LIGHT EXTINCTION RESULTS FOR "JUST MEETING" ALTERNATIVE
SECONDARY NAAQS SCENARIOS BASED ON 1-HOUR PM2.5 MASS
The rollback steps described in section 4.0 resulted in estimates of PMio light extinction
for each day and hour in each study area, for each of the 10 NAAQS scenarios based on 1-hour
PM2.5 mass as the indicator. Two summaries of these conditions are presented here.
Figure E-2 presents a box plot of daily maximum daylight 1-hour PMio light extinction
for each NAAQS scenario based on 1-hour PM2.5 mass. These can be compared to Figure 3-8(a)
of the UFVA representing pre-rollback daily maximum PMio light extinction, and to the upper
panel of the figures in Appendix F of the UFVA representing the daily maximum PMio light
extinction levels resulting from the 20 NAAQS scenarios examined in the UFVA (18 scenarios
based on PMio light extinction as the indicator, the current annual and 24-hour PM2.5 NAAQS,
and a scenario with an annual NAAQS of 12 |ig/m3 and a 24-hour NAAQS of 25 |ig/m3). It can
be seen that the distribution of PM2.5 mass in a given study area shifts downward as the NAAQS
scenarios progress from least to most stringent (as indicated by the required percentage
reduction) and in most cases become more similar to other areas (once the progression of more
stringent scenarios begins to require reductions in a given area).
Table E-6 presents the percentage of days in 2005-2007 on which daily maximum 1-hour
PMio light extinction exceeded each of the CPLs, under each of the 10 secondary PM NAAQS
scenarios based on 1-hour PM2.5 mass. These percentages are necessarily based on the days for
E-9
-------
which data to estimate PMio light extinction were available, but are best estimates of the
percentage of all days in the year given that the days with data were well distributed across the
year on either a one-in-three or one-in-six sampling schedule. These percentages can be
compared to the same-basis percentages presented in Table 4-7 of the UFVA.
E-10
-------
Figure E-2. Distributions of Daily Maximum Daylight 1-Hour PMi0 Light Extinction
under "Just Meet" Conditions for NAAQS Scenarios Based on 1-Hour PM2.s Mass
(excluding hours with > 90% RH)
(aa) NAAQS Scenario: 60 ng/m3 and 90th percentile
PMRo!lbackDailyMaxNAAQS60Pctl90DVsFromdaily.max
s -
324 300 98 308 273 158 349 279
277 1B1
O O
-r "T
~? ? 0/ <*•' ?
y C' . j>-
X ,*# /
(bb) NAAQS Scenario: 60 ^g/m3 and 95th percentile
PMRollbackDailyyaxNAAQS60Pctl95DVsFromdaily.max
O O
T
"T
.f *»• ^
E-ll
-------
Figure E-2 (cont). Distributions of Daily Maximum Daylight 1-Hour PMi0 Light
Extinction under "Just Meet" Conditions for NAAQS Scenarios Based on 1-Hour
Mass (excluding hours with > 90% RH)
(cc) NAAQS Scenario: 40 ng/m3 and 90th percentile
PMRollbackDailyyaxNAAQS40Pctl90DVsFromdaily.max
S -
324 300 38 306 273 158 349 279
277 1B1
O O
O O
O
i ; i 4
y *' s / c/ / /
x
-------
Figure E-2 (cont). Distributions of Daily Maximum Daylight 1-Hour PMi0 Light
Extinction under "Just Meet" Conditions for NAAQS Scenarios Based on 1-Hour
Mass (excluding hours with > 90% RH)
(ee) NAAQS Scenario: 30 ng/m3 and 90th percentile
PMRollbackDailyMaxNAAQS30Pctl90DVsFromdaily.max
-4:-
Q 9
-1-4- , -*-
(ff) NAAQS Scenario: 30 Mg/m3 and 95th percentile
PMRoHbackDailyMaxNAAQS30Pc!!95DVsFromdaiiy.max
109 324 300 98 306 273 158 34i 279
27? 181 143 225
4-
T
"X"
-1---4'
E-13
-------
Figure E-2 (cont). Distributions of Daily Maximum Daylight 1-Hour PMi0 Light
Extinction under "Just Meet" Conditions for NAAQS Scenarios Based on 1-Hour
Mass (excluding hours with > 90% RH)
(gg) NAAQS Scenario: 20 ng/m3 and 90th percentile
PMRollbackDailyMaxNAAQS20Pctl90DVsFromdaily.max
2T3 158 349 279
143 225
•* y
s ^
^ J? ^
<^ ^
(hh) NAAQS Scenario: 20 ^g/m3 and 95th percentile
PMRollbackDailyMaxNAAQS20Pctl95DVsFromdaily.max
E-14
-------
Figure E-2 (cont). Distributions of Daily Maximum Daylight 1-Hour PMi0 Light
Extinction under "Just Meet" Conditions for NAAQS Scenarios Based on 1-Hour
Mass (excluding hours with > 90% RH)
(ii) NAAQS Scenario: Daily Max: 10 ng/m3 and 90th percentile
PMRollbackDailyMaxNAAQS10Pctl90DVsFromdaily.max
324 300 98 306 273 1SB 349
(jj) NAAQS Scenario: Daily Max: 10 Mg/m3 and 95th percentile
PMRollbackDai!yMaxNAAQS10Pctl95DVsFrotndaily.max
E-15
-------
Table E-6. Percentage of Days Across Three Years (two years in the case of Phoenix and Houston) with Maximum 1-Hour
Daylight PMio Light Extinction Above CPLs when "Just Meeting" NAAQS Scenarios Based on 1-Hour PM2.s Mass
(Blue shading indicates no reduction required from current conditions.)
Scenario
NAAQS Level
(Hg/m3)
NAAQS
Percentile
Form
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake
City
Dallas
Houston
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Days with max hour above
64 Mm"1
aa bb cc dd ee ff gg hh ii jj
60 60 40 40 30 30 20 20 10 10
90 95 90 95 90 95 90 95 90 95
Percentage of days
53 53 53 53 53 53 53 53 43 35
76 73 65 57 69 60 55 44 28 17
89 87 84 81 84 79 74 69 41 30
44 44 44 44 44 44 44 44 37 32
45 45 45 37 45 45 45 26 17 10
81 81 81 81 81 81 81 71 41 29
79 79 79 79 79 79 74 65 32 27
89 85 80 68 87 80 72 62 41 34
91 91 91 89 91 89 82 77 47 34
84 80 74 72 76 73 65 60 40 33
85 85 81 77 81 77 63 55 27 19
81 81 81 76 81 74 64 56 31 20
84 78 71 62 72 63 55 43 17 10
83 83 80 71 81 73 63 56 27 19
76 75 72 68 73 69 64 56 34 25
Days with max hour above
112 Mm -1
aa bb cc dd ee ff gg hh ii jj
60 60 40 40 30 30 20 20 10 10
90^95^90^95^90^95^90^95^90^95
Percentage of days
23 23 23 23 23 23 23 23 1 1 6
52 48 37 31 44 32 29 18 9 4
78 76 65 57 65 53 41 31 1 1 7
6666666666
17 17 17 15 17 17 17 11 85
41 41 41 41 41 41 41 32 8 5
44 44 44 44 44 44 35 28 6 3
65 56 51 36 58 51 40 30 15 12
75 75 75 68 74 66 51 35 3 3
67 57 51 43 53 48 34 21 96
57 57 51 45 52 44 29 22 3 0
51 51 51 45 51 44 31 23 4 3
60 54 33 29 37 31 16 10 3 3
60 60 56 39 56 40 32 22 6 3
50 48 43 37 45 38 31 23 7 4
Days with max hour above
191 Mm -1
aa bb cc dd ee ff gg hh ii jj
60 60 40 40 30 30 20 20 10 10
90^95^90^95^90^95^90^95^90^95
Percentage of days
4444444411
30 27 17 11 23 12 10 510
52 46 30 24 30 19 11 633
1 11 1111 111
8 88 7888 521
10 10 10 10 10 10 10 700
11 11 11 11 11 11 9 610
34 26 21 13 30 20 15 11 4 3
31 31 31 21 31 19 5 300
43 28 13 7 14 9 6 41 1
26 26 18 14 21 13 6 200
23 23 23 18 23 16 8 21 1
26 17 8 5850 000
29 29 25 16 25 17 9 500
23 21 15 11 17 12 8 4 11
E-16
-------
APPENDIX F
Two Simplified Approaches to Calculate Hourly PM2.s Light
Extinction Values from Hourly PM2.5 Mass and Relative Humidity
Data Plus 24-hour Mean PM2.s Composition Data
F.I OVERVIEW
The goal of this appendix is to describe two procedures for calculating PM2.5 light
extinction that are considerably simpler than the complex method used in the original 15-city
assessment (UFVA, Section 3.3).1 The possible benefits of moving to a simpler approach
include (1) more transparency in the required calculations, (2) less intensive data processing, (3)
an increase in the number of monitoring sites that could meet the data requirements of the
approach without adding new sampling equipment or additional laboratory analysis, and (4) an
increase in the number of days per year for which the calculation of PM light extinction could be
conducted.
The two simpler procedures addressed in this memo, denoted "T" and "W", are similar in
many respects to approaches "D" and "F" for which a draft analysis was presented in appendix
4B of the second external review draft of the Policy Assessment Document.2 The original
UFVA and the draft analysis of approaches D and F were based on 2005-2007 data. The
analysis of approaches T and W versus the UFVA approach presented here is based on 2007-
2009 data. The UFVA approach has been re-executed on this more recent data set to provide a
consistent point of comparison to approaches T and W.
In section B.I.2 of Appendix B, it was noted that recent investigations have identified
issues with data quality for measurements of both 24-hour concentrations and 1-hour
concentrations from continuous Federal Equivalent Methods for PM2.5. This appendix on
calculated PM2.5 light extinction values is based on 2007-2009 data, most of which came from
continuous instrument models that pre-date the introduction of continuous Federal Equivalent
Methods for PM2.5. Some of the 2009 data may have come from continuous FEM instruments.
While no systematic investigation was performed on the 2007-2009 non-FEM data similar to the
investigations described in Hanley and Reff, 2011 and Evangelista, 2011,3'4 no similar issues
1 Particulate Matter Urban-Focused Visibility Assessment (UFVA) EPA 452/R-10-004, July 2010. The UFVA
presented a method and results for PM10 light extinction. The assessment in this appendix focuses on PM2 5 light
extinction. The UFVA is available: http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
2 The analysis of approaches D and F has been finalized in the form of a technical memo so that it is available for
reference. Simplified Approaches for Calculation of Hourly PM2 5 Light Extinction Values from Hourly PM25 Mass
and Relative Humidity Data and 24-hour PM2 5 Composition Data, Phil Lorang, EPA, November 17, 2010.
Available at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007 td.html.
3 Hanley, T and Reff, A (2011). Assessment of PM2 5 FEMs compared to collocated FRMs. Memorandum to PM
NAAQS Review Docket EPA-HQ-OAR-2007-0492. April 7, 2011. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007 td.html.
4
Evangelista, M. (2011). Investigation of 1-hour PM2 5 Mass Concentration Data from EPA-Approved Continuous
Federal Equivalent Method Analyzers. Memorandum to PM NAAQS Review Docket (EPA-HQ-OAR-2007-0492).
Aprils, 2011. Available at: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007 td.html
F-l
-------
were noticed in the course of using these data. Moreover, all of the 1-hour data used in this
Appendix was normalized to 24-hour concentrations measured by filter-based Federal Reference
Method samplers on the same day at the same site.
F.2 COMPLEXITY OF THE UFVA APPROACH
The UFVA approach to estimation of hourly PM2.5 light extinction has the following
complex aspects:
1. The SANDWICH mass balance model is used to estimate 24-hour average PM2 5 sulfate,
nitrate, and organic carbonaceous material (OCM) mass loading on the FRM filter for
each Chemical Speciation Network (CSN) sample day.5 This requires information on
daily temperature and relative humidity. The SANDWICH-estimated FRM loadings are
initially used to compare with FRM mass. The sulfate and nitrate components are
initially derived from the relevant CSN filters, with adjustments to represent FRM mass.
FRM sulfate includes estimated particle bound water while the FRM nitrate loading may
under-represent the ambient concentration of nitrate. These are re-adjusted in a
subsequent step to represent ambient conditions prior to calculation of light extinction.
2. The estimates of 24-hour average PM2.5 elemental carbon and fine soil component
concentrations are determined from the analysis of the relevant CSN filter.
3. Monthly mean diurnal variations of each of the major PM2.5 components from CMAQ air
quality simulation modeling results for the location of each monitoring site are applied to
sample day-specific CSN 24-hour samples to create preliminary estimates of hourly
component concentrations. For the UFVA, available output from a 2004 CMAQ
modeling platform was used. This step can result, for example, in preliminary estimates
of concentrations of sulfate that are fairly uniform throughout a day while concentrations
of nitrate may show much more variation because of temperature effects on the
gas/particle partitioning of nitrate.
4. Estimates of hourly PM2.5 mass are developed by normalizing continuous PM2.5
measurements to the 24-hour FRM filter mass.6
5. The preliminary estimates of hourly components from step 3 above are scaled up or down
in equal proportion to reconcile their sum to the estimate of hourly PM2.5 mass from step
4.
5 Frank, N., Retained Nitrate, Hydrated Sulfates, and Carbonaceous Mass in Federal Reference Method Fine
Paniculate Matter for Six Eastern U.S. Cities, J. Air Waste Manage. Assoc., 56, 500-511, 2006.
6 The UFVA, Appendix 4B of the second external review draft of the Policy Assessment Document, and the analysis
in the technical memo referenced in footnote 2 were all based on 2005-2007 data. The continuous PM2 5 instruments
operating in this period were not EPA-approved as Federal Equivalent Methods (FEMs). The EPA-approved FEMs
were first used in state networks in significant numbers in 2009. The EPA staff envisions that in any future
monitoring of hourly PM2 5 mass to implement a visibility-based secondary PM NAAQS, EPA-approved continuous
FEMs will be required. For the UFVA, Appendix 4B of the second draft Policy Assessment, and the analysis in the
technical memo referenced in footnote 2, therefore, hourly PM2 5 mass values from continuous instruments were
adjusted day-by-day to match the 24-hour average PM2 5 mass reported by the collocated filter-based sampler. This
was intended to make the values of hourly PM2 5 mass more like the values that would be obtained by FEMs. The
analysis in this appendix is based on 2007-2009 data from continuous instruments, but only a fraction of the data are
from FEMs. Therefore, the approach of adjusting hourly PM2 5 mass values from continuous instruments day-by-
day to match the 24-hour average filter-based PM2 5 mass was maintained in the analysis presented in this appendix.
F-2
-------
6. The resulting hourly PM2.5 sulfate and nitrate component concentrations are adjusted to
reflect actual atmospheric concentration, which is assumed to be represented by the CSN
filter sulfate and nitrate measurements. (This step in effect un-does the estimated FRM
sulfate mass enhancement due to particle bound water and the FRM nitrate loss that were
both introduced by the SANDWICH mass balance model.)
7. The original IMPROVE algorithm is used to estimate hourly PM2 5 light extinction from
hourly PM2.5 component and hourly relative humidity values.
F.3 RESULTS OF THE RE-EXECUTION OF THE UFVA APPROACH FOR 2007-
2009
Table F-l shows the number of daily samples of suitable data that were available for the
re-execution of the UFVA approach for each quarter and study area for the 2007-2009 period.
Due to the discontinuation of CSN monitoring at AQS site 040137020 in the Phoenix area as of
the end of 2006, PM2.5FRM and speciation data from site 040139997 (JLG SUPERSITE) were
used in the re-execution of the UFVA method for 2007-2009. Continuous PM2.5 data were taken
from site 040139998. The needed combination of data was available from these two sites only
for 2009, for only 84 days. Results for Phoenix should be viewed in the light of this limited set
of data compared to other areas. All other study areas' monitoring sites were as documented in
the UFVA.
Figure F-l (which is also Figure 4-4 of this document) shows the box-and-whisker plots
of 2007-2009 daily maximum PM2.5 light extinction. Table F-2 shows the 90th and 95th
percentile design values for daily maximum PM2.5 light extinction.7 These displays can be
compared to 2005-2007 results presented in UFVA Table 3-4, the bottom panel of Figure 3-8
(also reproduced as Figure 4-3 of this document), and the top half of Table 4-2, respectively,
keeping in mind that the tables and figures in the UFVA included the estimated contribution of
PMio-2.5 to PMio light extinction. This comparison shows that many of the eastern areas' median
and 90th percentile 2007-2009 PM2.5 light extinction levels are notably lower than the
corresponding 2005-2007 PMio light extinction levels, even though PMio-2.5 in general did not
contribute much to the UFVA estimates of PMio light extinction in eastern areas. This is
directionally consistent with the known substantial improvement in PM2.5 air quality in the
eastern U.S. over the 2005-2009 period.
7 As in the UFVA, 2007-2009 design values have been calculated using whatever daily maximum values were
available in each year. In a future regulatory program there would presumably be a data completeness requirement
in order to find that an area meets or does not meet the NAAQS.
F-3
-------
Table F-l. Number of Days per Quarter in Each Study Area
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit-Ann Arbor
Pittsburgh
Baltimore
Philadelphia-Wilmington
New York
Year
Total
Number of
Days
150
325
161
84
276
257
144
287
330
258
133
264
140
98
145
2007
Qi
13
26
21
0
23
18
15
29
30
25
11
22
12
13
19
Q2
13
28
26
0
25
23
14
25
30
19
11
22
12
14
15
Q3
14
30
24
0
19
24
9
22
27
26
12
23
17
12
19
Q4
13
28
24
0
29
21
0
10
26
21
11
23
16
12
21
2008
Qi
14
29
29
0
28
22
15
24
31
22
7
23
0
9
20
Q2
14
27
28
0
12
25
13
22
27
25
12
24
0
11
14
Q3
14
30
9
0
27
26
12
27
30
24
13
24
0
14
13
Q4
12
29
0
0
15
20
14
28
29
21
13
25
0
13
24
2009
Qi
7
15
0
11
13
13
7
14
13
12
3
10
13
0
0
Q2
14
27
0
24
29
23
15
30
28
22
15
25
26
0
0
Q3
11
28
0
21
28
21
15
29
28
21
10
22
21
0
0
Q4
11
28
0
28
28
21
15
27
31
20
15
21
23
0
0
Figure F-l. Distributions of Estimated Daily Maximum Daylight 1-Hour PMi.s Light
Extinction Across the 2007-2009 Period, by Study Area (excluding hours with relative
humidity > 90%)
Daily Max
s -
cf" #'
«/ <<*
«> ^>
-------
Table F-2. 2007-2009 Daily Maximum PM2.5 Light Extinction Design Values for the Study
Areas
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit-Ann Arbor
Pittsburgh
Baltimore
Philadelphia-Wilmington
New York
Design Value for 90th Percentile
Form (Mm *)
121
312
434
71
152
162
145
224
269
189
275
222
183
280
295
Design Value for 95th Percentile Form
(Mm1)
132
394
503
97
285
204
177
267
326
212
436
256
212
320
365
F.4 SIMPLIFIED APPROACHES T AND W
This analysis examines the difference between calculated hourly PM2.5 light extinction
values used in the UFVA and values generated using two simpler approaches. Both of the
simpler approaches (designated by letters T and W) use the original IMPROVE algorithm
without the Rayleigh and PMi0-2.5 contributions to total light extinction.
Approach T is similar to the previous approach F. The difference is that while approach
F involved the calculation and use of monthly-averaged PM2.5 component percentages for each
of the five components, approach T uses a two-factor rearrangement of the IMPROVE algorithm
and accordingly involves the calculation and use of only two monthly-averaged parameters.
These parameters are the hygroscopic fraction (HF) and the dry light extinction efficiency
(DLEE). A technical memo explains in more detail the definition of these parameters and their
relationship to the IMPROVE algorithm.8 Looking to a hypothetical future program based on
approach T, it would be possible to develop estimates of PM2 5 light extinction for any day for
which hourly PM2.5 mass concentrations are available from a continuous FEM instrument,
including days that do not have PM2.5 speciation data, by applying the monthly-averaged HF and
DLEE values to each hourly PM2.5 measurement.
Approach W differs from approach T in that the HF and DLEE parameters are applied
day-by-day rather than averaged across a month. The inclusion of this approach in this analysis
allows a close examination of the effect of the monthly-averaging step in approach T. Approach
8 Pitchford, M (2010). Assessment of the Use of Speciated PM25 Mass-Calculated Light Extinction as a Secondary
PM NAAQS Indicator of Visibility. Memorandum to PM NAAQS Review Docket (EPA-HQ-OAR-2007-0492),
Marc Pitchford, NOAA, November 17, 2010. Available at:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm 2007 td.html.
F-5
-------
W can yield estimates of PM2.5 light extinction only for days on which PM2.5 speciation data are
available. The EPA staff notes that many PM2 5 and PMi0 monitoring sites currently operate on
l-day-in-3 or l-day-in-6 schedules, and are nevertheless considered to provide sufficiently robust
data for air quality management purposes.
Both approaches T and W assume that within a day, the percentage mix of PM2 5
components is the same hour-to-hour.
Table F-3 explicitly defines the calculation steps for approaches T and W. Table F-4
compares the UFVA approach, approach T, and approach W in conceptual terms.
F-6
-------
Table F-3. Calculation Steps for Approaches T and W
Approach T
Approach W
Comments
(i) For each CSN sampling day, subtract a sampler-
dependent estimate of the OC artifact from the OC
measurement, and multiply by 1.6 to estimate
organic carbonaceous material (OCM).
• OCM = (OC - artifact) * 1.6
Same as T.
The values for the OC artifact used for 2007-2009
ranged from 0.32 to 1.53 ug/m3, depending on CSN
sampler model. The artifact adjustment for the
URG 3000N sampler is of most interest
prospectively, because it is the single sampler now
in use for carbon sampling in CSN. The URG
3000N was used only at about one-half of the 15
study sites and only in the second half of 2007. For
those sites and days, an organic carbon artifact of
0.4 ug/m3 was assumed for the purposes of the
UFVA and this document, based on early
experience with this sampler. EPA staff is currently
exploring whether there is a better way to adjust for
organic carbon artifact based on a more recent,
larger field blank and back-up filter data set.
A section in the body of this appendix explains the
selection of 1.6 as the multiplier for OC.
(ii) For each CSN sampling day, calculate fine
soil/crustal PM2 5 (FS) from CSN measurements of
crustal elements AL, Si, Ca, Fe, and Ti, using the
formula
• FS= 2.20 x[Al]+ 2.49 x [Si]+ 1.63 x
[Cal+2.42 x [Pel+ 1.94 x [Til
Same as T.
This is the same equation as used in the IMPROVE
network and the Regional Haze program, and differs
from the corresponding step in approaches D and F.
(iii) For each CSN sampling day, multiply CSN
measurement of sulfate ion by 1.375, and multiply
CSN measurement of nitrate ion by 1.29, to reflect
associated ammonium under an assumption of full
neutralization.
• AS = Sulfate ion* 1.375
• AN = Nitrate ion * 1.29
Same as T.
(iv) For each CSN sampling day, sum the above
estimates of the 5 components of PM25:
• Sum=AS + AN + OCM + EC + FS
Same as T.
F-7
-------
Approach T
Approach W
Comments
(v) For each CSN sampling day, calculate HF and
DLEE:
• HFdaiiy= (AS+AN)/Sum
Same as T.
[3*(AS+AN)+4*OCM+10*EC+1.0*FS]
Sum
(vi) Average the values
from step (v) across the CSN sampling days of each
calendar month. This results in values for
T-TP
JTLT
This step is omitted. HF and BLEE are day-
specific:
• HF daily
monthly _ average
Note: An alternative approach to calculating
HFmonthiy average would be to define it as the monthly
average of the daily values of (AS + AN) divided by
the monthly average of the daily values of Sum. A
similar alterative definition is possible for
In the analysis described in this appendix, we
applied a minimum requirement of four samples per
month, which is usually 80% of the samples
scheduled per month at a site using one-in-six-days
sampling. If there were fewer samples in a month,
approach T estimates were not generated but
approach W estimates were. _
F-8
-------
Approach T
Approach W
Comments
(vii) For each daylight hour of each day of a month
(including days without CSN sampling), calculate
PM2 5 light extinction using the 1-hour PM2 5 mass
concentration from a continuous instrument, HF,
DLEE, and 1-hour relative humidity (RH):
• Hourly PM2 5 light extinction = PM2.shom-iy
* { 3*[f(RHhourly)-l] * [HFmontuy_average] +
-L'J-'LJ-'monthly average /
Estimates of PM2 5 light extinction in hours with
RH greater than 90% are not valid.
(vii) For each daylight hour of each day of a month
for which CSN speciation data is available,
calculate PM2 5 light extinction using the 1-hour
PM2 5 mass concentration from a continuous
instrument, HF, DLEE, and 1-hour relative
humidity (RH):
• Hourly PM2 5 light extinction = PM2 5houriy
* { 3*[f(RHhouriy)-l] * [HFdaiiy] +
iy }
Estimates of PM2 5 light extinction in hours with
RH greater than 90% are not valid.
Prospectively, this appendix assumes that only
continuous instruments approved as federal
equivalent methods (FEM) would be allowed for
purposes of estimating hourly PM2 5 light extinction.
Accordingly, the equations to the left are based on
the instrument-reported hourly PM2 5 mass
concentration. As explained in footnote 3, the
available hourly PM2 5 mass concentrations from
continuous instruments for this analysis were mostly
from non-FEM instruments. All the hourly PM2 5
concentrations were therefore adjusted on a day-by-
day basis to match the 24-hour concentration
reported by a collocated FRM/FEM filter-based
sampler, prior the step described on the left. This
was intended to better simulate the future scenario
of interest.
F-9
-------
Table F-4. Detailed Comparative Description of the UFVA Approach, Approach T, and Approach W for Estimating 1-Hour
PM2.s Light Extinction
UFVA Step*"
1
1, continued
1, continued
Aspect of Approach
Estimation of 24-hour organic
carbonaceous mass
Estimation of 24-hour
elemental carbon mass
Estimation of 24-hour
hydrated ammonium sulfate
mass
UFVA Approach
The SANDWICH method1"
is used to subdivide the 24-
hour PM2 5 mass reported by
the FRM for each day and
site into hydrated ammonium
sulfate, ammonium nitrate,
elemental carbon, organic
carbonaceous material
(OCM), fine soil, and water.
This is done using
information from the CSN
measurements, physical
models, and day-specific
temperatures and relative
humidity. OCM is estimated
as the residual needed to
reach mass closure after
estimation of the other
components.
CSN elemental carbon
concentration
CSN sulfate ion
concentration, with day-
specific SANDWICH
estimates of associated
ammonium and water.
Approach T
Organic carbonaceous mass is
assumed to equal the organic
carbon value reported from
CSN sampling, minus a
blank filter artifact correction
value that depends on PM
sampler model, times 1.6.
Same as UFVA
Sulfate ion measurement
from the CSN filter is
multiplied by 1.375 to
represent dry ammonium
surf ate.
Approach W
Same as T
Same as UFVA
Same as T
The numbering of steps follows that used to describe the UFVA approach in section 3.3.1 of the UFVA.
m Frank NH. (2006). Retained nitrate, hydrated sulfates, and carbonaceous mass in federal reference method fine paniculate matter for six eastern U.S. cities. J
Air Waste Manag Assoc, 56: 500-11
F-10
-------
UFVA Step
Aspect of Approach
UFVA Approach
Approach T
Approach W
1, continued
Estimation of 24-hour
ammonium nitrate mass
Nitrate ion on the FRM
Teflon filter is estimated by
SANDWICH, with day-
specific estimates of
associated ammonium and
water.
Nitrate ion measurement from
the CSN filter is multiplied
by 1.29 to represent dry
ammonium nitrate.
Same as T
1, continued
Estimation of 24-hour fine
soil/crustal mass
Calculated from 4 CSN
elements, not including Al (a
difference from the
IMPROVE approach)
IMPROVE approach, using 5
elements
Same as T
Diurnal pattern of PM25
components
The CMAQ-derived monthly
normalized diurnal profiles
for the sulfate, nitrate,
elemental carbon, organic
carbonaceous material and
fine soil/crustal components
(each of which averages to
1.0 across 24 hours) were
multiplied by the day-specific
SANDWICH-based estimates
of the 24-hour average
concentrations of the five
PM2 5 components, to get
intermediate day-specific
hourly estimates of the five
components (including
ammonium and water
associated with sulfate and
nitrate ion).
No diurnal profiles are used.
Same as T
F-ll
-------
UFVA Step
Aspect of Approach
UFVA Approach
Approach T
Approach W
Sum the 5 components
The hourly concentrations of
these five components
(including day-specific
ammonium and water
associated with sulfate and
nitrate ion when the FRM
Teflon filter is weighed) were
added together, to get a sum-
of-components estimate of
hourly PM2 5 mass for the day
of FRM/CSN sampling.
The following approach was
used instead of steps 3,5,7,
and 8.
• Calculate HF and
DLEE for each day.
• Average the daily
values of HF and
DLEE across the
available CSN
sampling days in the
month.
• Use the monthly-
averaged HF and
DLEE values in a
two-factor version of
the original
IMPROVE
algorithm.
The following approach was
used instead of steps 3,5,7,
and 8.
• Calculate HF and
DLEE for each day.
• Use the day-specific
HF and DLEE
values in a two-
factor version of the
original IMPROVE
algorithm.
Hourly PM25 concentration,
consistent with 24-hour FRM
concentration.
The hourly data from the
continuous PM2 5 instrument
on the day of the FRM
sampling were normalized by
their 24-hour average, to get a
normalized diurnal profile.
This profile was applied to
the 24-hour PM2 5 mass
reported by the FRM sampler.
This keeps the average of the
valid 1-hour PM2 5 values
equal to the 24-hour value
from the FRM sampler.
See the comment on this topic
in Table F-3.
See the comment on this topic
in Table F-3.
F-12
-------
UFVA Step
Aspect of Approach
UFVA Approach
Approach T
Approach W
5,6
Adjust preliminary estimates
of hourly PM2 5 component
concentrations (reflecting
CMAQ diurnal profiles and
24-hour measurements) to be
consistent with the estimate
of hourly PM2.5 mass.
The two estimates of hourly
PM2.5 mass from steps 3 and
4 were compared, hour-by-
hour. Within each hour, the
estimates of all five
components from step 3 were
increased or decreased by a
common percentage so that
the sum of the five
components after this
adjustment was equal to the
estimate of the hourly PM25
mass from step 4. The
adjustment percentage varied
from hour-to-hour.
Not applicable.
Not applicable.
Adjust the FRM-consistent
estimate of sulfate to the
CSN/IMPROVE-consistent
basis expected by the
IMPROVE algorithm.
Each hourly estimate of
sulfate concentration on the
FRM filter from step 6
(which includes estimates of
associated ammonium and
particle bound water) was
adjusted so that it excludes
water and reflects full
neutralization and therefore is
consistent with the reporting
practices of the IMPROVE
program and the IMPROVE
algorithm.
Not applicable.
Not applicable.
F-13
-------
UFVA Step
Aspect of Approach
UFVA Approach
Approach T
Approach W
Adjust the FRM-consistent
estimate of nitrate to the
CSN/IMPROVE-consistent
basis expected by the
IMPROVE algorithm.
A similar adjustment as in
step 7 (for sulfate) was made
to each hour's nitrate
concentration from step 6, so
that the estimate of hourly
nitrate would reflect actual
atmospheric conditions and
be consistent with the
IMPROVE algorithm.
This can result in the estimate
of nitrate used in the
IMPROVE algorithm being
higher than the FRM-
consistent estimate, for days
on which the SANDWICH
method predicts a loss of
nitrate from the FRM filter.
Not applicable.
Implication: On days when
the FRM filter has lost nitrate
mass, the estimates of hourly
PM2 5 will be lower than
actual atmospheric mass. All
hourly PM2 5 components will
be biased low relative to the
values that should be used in
the IMPROVE algorithm, by
the fraction that the lost
nitrate is of total PM25 mass.
In the opposite direction, any
water mass included in the
hourly PM2 5 mass reported
by the continuous PM2 5
instrument will in effect be
allocated among the five
components.
Same as T.
Not numbered in UFVA
Estimation of PM25 light
extinction from estimates of
hourly concentrations of
PM25 components.
Original IMPROVE
algorithm, including f (RH)
determined from hourly RH.
Hours with RH >90% were
excluded from design values
and from most graphical
displays of results.
Use a two-factor re-
arrangement of the original
IMPROVE algorithm, per
table F-3. Hours with RH
>90% were excluded from
design values and from all
graphical displays of results.
Same as T.
F-14
-------
F.5 COMPARATIVE PERFORMANCE OF SIMPLIFIED PM2.5 LIGHT
EXTINCTION APPROACHES T AND W
The performance assessment of simplified approaches T and W for calculated PM2.5 light
extinction was accomplished by comparing hourly values of PM2.s light extinction generated by
each approach to their corresponding paired values generated using the original UFVA method.
Data from 2007-2009 were used for the comparison. Annual box plots of the percentage
differences between the paired values, as well as annual and monthly scatter plots and regression
analysis of these paired data, were generated. As the results below show, approaches T and W
produce hourly PM2.5 light extinction values that in most cases are quite well correlated to the
hourly PM2.5 light extinction values generated by the original UFVA method, with a few notable
exceptions discussed below. Selected graphs and summary tables of regression statistics are
included and discussed below to show the degree of comparability. Note that approaches T and
W can be compared to the UFVA approach only for days with CSN data. Similarly, approaches
T and W can be compared to each other only for those days, as approach W does not produce
estimates of PM2.5 light extinction for other days.
The box plots of the percentage differences between calculated hourly PM2.5 light
extinction by approaches T and W versus the UFVA approach are shown in Figure F-2. In this
box plot, the percentage difference is calculated as follows:
Percentage difference = 100% * [(T or W estimate) - (UFVA estimate)]/(UFVAestimate)
The patterns of percentage bias for the 15 urban areas are notably similar in both the T
and W plots.11 Fresno and Los Angeles are noticeably biased low by approaches T and W, while
Houston and Philadelphia are biased high. However, these percent difference plots do not reveal
the absolute level of PM2.5 light extinction for the points corresponding to smaller or larger
percent differences.
Scatter plots of calculated hourly PM2.5 light extinction for all daylight hours comparing
approaches T and W to the UFVA approach for all 15 areas are shown in Figure F-3. The same
comparisons are made for daily maximum PM2.5 light extinction in Figure F-4. Note that the
vertical scale is not the same in all plots. The dashed red line is the 1:1 line, and regression
statistics appear at the top of each plot. These figures give a better visual sense of how the
deviations between the simpler approaches and the UFVA approach vary by area and, by
implication, with the nature of PM2.5 in each study area. For example, both simpler approaches
have an obvious negative bias in Fresno, Los Angeles, and New York across a wide range of
PM2.5 light extinction values, consistent with the percentage difference box plot.
The degree of comparability for paired hourly PM2.5 light extinction values between
approach T and the UFVA approach and between approach W and the UFVA approach, by
month and urban area, can also be displayed via regression statistics, as in Tables F-5 (approach
T) and F-6 (approach W). Note that the independent and dependent variables in Tables F-5 and
F-6 are reversed from the order used to provide regression statistics in the scatter plots of Figures
F-3 and F-4. Color shading is used to highlight cells (representing a given month across 2007-
2009) in which the slope of the regression falls outside the range of 0.9 to 1.1, with deeper
1J In the box plots, some extreme values of the percentage change are due to rounding effects when the values
involved were very small.
F-15
-------
shading indicating a larger deviation from a slope of 1.0. In both tables, for most eastern urban
areas and months the regression lines have slopes in the range of 0.9 to 1.1, R2 values near one,
and small intercepts implying that the values are highly comparable. The western urban areas
(Fresno, Houston, Los Angeles, Phoenix, Salt Lake City, and Tacoma) have intercepts, slopes,
and R2 values for some months that imply a bias and/or noisier relationship than the eastern
areas between values calculated by approaches T and W versus the UFVA approach. Los
Angeles and Fresno, and to a lesser degree other western areas, stand out as having regression
slopes that indicate that approaches T and W are biased low relative to the UFVA estimates for
most of the year.
A comprehensive assessment of the reasons for the differences between approaches T and
W versus the UFVA approach has not been conducted. However, some explanations are
suggested by the results themselves and by the previous draft analysis of similar approaches D
and E. The fact that the CMAQ-based hour-to-hour variations and the monitoring-based day-to-
day variations in the dry PM2.5 composition in the UFVA approach can in most areas be replaced
by either daily-averaged or monthly-averaged values without much loss of precision in
calculated hourly PM2 5 light extinction suggests that these shorter-term variations within a single
month at a single monitoring site are usually not very influential. This may imply that the
differences between approaches T and W and the UFVA approach are probably not mostly due
to the relative composition changes caused by use of the mass-closure SANDWICH model
versus the simpler multiplier model used in approaches T and W. This suggests that the
difference may be due to differences in the value of PM2.5 mass to which the HF and DLEE
factors are applied. Approaches T and W lack any step that would effectively reconcile the
implicit nitrate portion of the average of the 24 hourly values of PM2.5 mass to the CSN-
measured 24-hour average nitrate concentrations, to compensate for any loss of nitrate from the
continuous PM2.5 instrument.12 This explanation is consistent with the direction of the biases
seen for sites that are known to have high ammonium nitrate (e.g., Los Angeles and Fresno). For
Los Angeles and Fresno, a logical explanation is that the FRM-adjusted estimates of 1-hour
PM2.5 mass do not include the entire ambient nitrate that is present in particulate form. The
UFVA approach includes an adjustment to restore this mass in the estimation of light extinction
while approaches T and W do not. There are variations of approaches T and W that may
improve the correlation with actual ambient light extinction in certain areas of the country, such
as one that would retain some version of a mass-closure model to account for the negative
artifact for ammonium nitrate, but that otherwise has the remaining simplifications of approaches
TorW.
12 FRM sampling is well known to have a negative artifact for nitrate compared to ambient concentrations under
some conditions. This negative artifact presumably carries over to any continuous PM2 5 FEM instrument that is
well correlated with the FRM. A negative artifact would logically lead to an underestimate of PM2 5 light extinction.
F-16
-------
Figure F-2. Box and Whisker Plot of the Percent Difference in Calculated Hourly PM2.s
Light Extinction between Approach T and the UFVA Approach (top plot) and between
Approach W and the UFVA Approach (bottom plot) by Urban Study Area
Approach T
Uii
T
Approach W
-
T
F-17
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
1 i-
3
o
Tacoma. WA
y=0.97x + 0.4, R-squared = 0.97
50 100
150 200
Original Bext
250 300
Tacoma. WA
y=0.96x » 0.67, R-squared = 0.98
50 100 150 200 250 300
Original Bext
Fresno, CA
y=0.9x * -1.84, R-squared - 0.95
Fresno, CA
y=0.95x t -4.29, R-squared - 0.95
8 -
100 200 300 400
Original Bext
500 600
200 300 400
Original Bext
500 600
F-18
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
8 -
Los Angeles, CA
y=0.78x * 3.96, R-squaied = 0.93
100 200 300 400 500 600
Original Bext
Los Angeles, CA
y=0.85x * 0.49, R-squared = 0.92
100 200 300 400 600 600
Original Bext
8 -
Phoenix, AZ
y=0.81x * 4.13, R-squared = 0.91
Phoenix, AZ
y=0.84x * 2.87, R-squared = 0.96
8
20 40 60 80 100 120 140
Original Bext
20 40 60 80 100 120 140
Original Bext
F-19
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
§ -
Salt Lake City. UT
y=1.05x t -3.89, R-squared = C
9-'
200 tta
Original Bext
Salt Lake City, UT
y=1.08x t -4.85, R-squared = 0.99
400
Original Bext
s
8 -
S -
Dallas. TX
y=0.92x t 2.94, R-squared = 0.96
50 100 150 200 250 300
Original Bext
Dallas. TX
y=0.97x * 1.11, R-squared = 0.96
100 150 200 250
Original Bext
F-20
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
s
I S-
s -
Houston, TX
y=1.07x * -0.57, R-squared = 0.96
100
Original Bext
Houston, TX
y=1,06x * 0.17, R-squared = 0.98
• %
o. S
S
100
Original Bext
St. Louis. IL
y=0.97x t 2.54, R-squared - 0.95
200 300
Original Bext
St. Louis. IL
y=1x + 1.19, R-squared = 0.96
8 -
O O
200 300
Original Bext
F-21
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
Birmingham, AL
y=1.01x * 1.27, R squared = 0.98
400 600
Original Bext
Birmingham, AL
y=1 X + 2.51, R-squared = 0.98
8 -
400 600
Original Bext
8 -
Atlanta, GA
y=0.98x + 2.23, R-squared = 0.97
Atlanta, GA
y=0.99x * 2.08, R-squared = 0.97
8
200
Original Bext
200
Original Bext
F-22
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
Detroit. Ml
y=0.88x * 5.28, R-squared = 0.96
400
Original Bext
Detroit. Ml
y=0.93x * 2.93, R-squared = 0.97
400
Original Bext
8
Pittsburgh, PA
y=0.99x t 2.56, R-squared - 0.97
Pittsburgh, PA
y=1x + 2.36, R-squared = 0.97
200
Original Bext
200 300
Original Bext
F-23
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
H S-
I
8 -
S
Baltimore, MD
y=1.02* + 1.2, R-squared =0.96
100 150 200
Original Bext
250 300
Baltimore, MD
y=1.02x * 1.59, R-squared = 0.96
! s
8 -
50 100 150 200 250 300
Original Bext
Philadelphia. PA
y=1.06x + 0.63. R-squared = 0.97
Philadelphia, PA
y=1.06x + 0.75. R-squared = 0.97
50 100 150 200 250 300 350
Original Bext
F-24
-------
Figure F-3. Scatter Plots of Hourly PM2.5 Light Extinction (all daylight hours) Calculated
by Approach T (y, left plot) and W (y, right plot) versus the UFVA Approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in this figure.
8 -
New York, NY
y=0.92x * 3.25, R-squared = 0.96
200 300
Original Bext
8 -
New York, NY
y=0.94x » 1.48, R-squared = 0.98
200 300
Original Bext
F-25
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
Tacoma, WA
y=0.99x * -4.17, R-squared = 0.96
1-
1-
i
i a.
8 -
S -
0 -
o .
,' o
o
o P'
0 - '
? ••'
.- ° o
<&$ '*s°
%: 0°V°
£' °^° °
Jif^
J^Sf 0
/$i
0 50 100 150 200 250 300
Original Bext
Fresno, CA
y=0.9x •*• -9.71, R-squared = 0.93
§-
1-
1-
I
S 8-
jn 8
o.
g _
8 -
0 -
°
,'''J°°
**'$
°,.'o ° oo
o - ' ** o
,' o o
.^"
0 o '''JpSo ° °
®- o o
-^&) ^ °
afe^^0' °
j^:
0 100 200 300 400 500 600
Original Bext
Tacoma, WA
y=0.97x + -3.26, R-squared = 0.97
1-
1-
g_
1
!i s -
D. —
8 -
S -
o -
O,'
o,-'
/
0
•6
§-b
«'°
q^°^
'•*T! '^
^,_^ooo
^ °
Js^6 °
£v
0 50 100 150 200 250 300
Original Bext
Fresno, CA
y=0.98x * -16.66, R-squared = 0.93
8-
s-
8 -
1
c | -
S —
8 -
O —
§ °.-''
o
a ° '°
o °, '
o p
°0'' 0 °
O O . ' 0 O
,-cP
°-"
e'
,«g o
°^t> Q O O
•^^ a& o o
.«^^^oi
y^
0 TOO 200 300 400 500 600
Original Bext
F-26
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
8
§^
o
1
c § -
i
8-
8 -
o —
Los Angeles, CA
y=0.74x + -3.92, R-squared = 0.88
5>
o
0
•' 0 °
o
0 °0003J
,-' o °
/ o8°o °° o
0.8,9°° ° & °o
. ##*8° S"" *
0 100 200 300 400 500 600
Original Bent
8 .
s -
1 s
e
J2
D.
S -
s-
Phoenix, AZ
y=0.79x * 6.13, R-squared = 0.87
/ °
O
0
9-0 o o
• ' O
.00°
o ,'o
o ,'D o
O '' " O
o (5
J|l0^
^p
,'o
,0
20 40 60 80 100 120 140
Original Bext
8 -
1-
g
•
1 §^
£ s -
8 -
8 -
O -
Los Angeles, CA
y=0.85x + -15.77. R-squared = 0.85
o
O'
o
,-''
o0o °,''
o
,-' o o °
t?
X5 °
o
X \ ° °
° °'ei°o° ° Sti °o ° °
^•—
0 100 200 300 400 500 600
Original Bext
B -
8 -
s -
Jl
I s -
Q-
?_
s -
Phoenix, AZ
y=0.84x + 3.68. R-squared = 0.96
o
o
/
o
o
o
X =
' '0
,' O O
«/* °
4?*
/?
%*
,o
20 40 60 80 100 120 140
Original Bext
F-27
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
Salt Lake City. UT
y=1.06x * -8.32. R-squared = 0.98
400
Original Bext
Salt Lake City, UT
y=1.1x * -10.64, R-squared = 0.99
400
Original Bexl
8
S
Dallas. TX
y=0.86x + 5.4, R-squared = 0.91
Dallas. TX
y=0.95x + -1.74, R-squared - 0.93
S -
8 -
S -
100 150 200
Original Bext
250 300
100 150 200
Original Bext
250 300
F-28
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
8 -
8 -
S -
Houston, TX
y=1.09x + -2.18, R-squared = 0.93
0 O
100
Original Bext
Houston, TX
y=1.09x + -1.86, R-squared = 0.98
8 -
8 -
S -
100 150
Original Bext
8 -
St. Louis. IL
y=0.92x * 2.55, R-squared - 0.92
200 300
Original Bext
St. Louis, IL
y=0.97x * -1.7, R-squared = 0.95
8 -
0? 0
200 300
Original Bext
F-29
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
Birmingham, AL
y=1x+ -2.91, R-squared =0.97
•WO 600
Original Bext
Birmingham, AL
y=0.97x * 1.33, R-squared = 0.98
.'o
400
Original Bexl
8
Atlanta, GA
y=0.97x + -0.84, R-squared = 0.95
Atlanta, GA
y=0.98x t -1.68, R-squared - 0.96
8 -
200 300
Original Bext
200
Original Bext
F-30
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
1-
1
1
1-
0 -
Detroit, Ml
y=0.85x + 5.9, R-squared = 0.96
/'«
/'
O
o /
,' o
»4' *
.• ^
j^ o°°
JP
A
0 200 400 600 800
Original Bext
s -
1-
1-
1 1-
c
JS
D.
? '
8 -
S -
Pittsburgh, PA
y=0.97x + -0.78, R-squared = 0.96
o ,'
o
00 •' O °
o
•"'"
0 °°°°^
Vf°t°
all*''*
° C^^1 ^ *
3$H^
c^^^%
.y^
50 1X 150 200 250 300 350
Original Bext
1 -
1
| |-
g-
O -
Detroit, Ml
y=0.91x + -1.07, R-squared = 0.97
,' o
,' o
o ,'
.-''%
^*"
J*^0 °°
^
0 200 400 600 800
Original Bext
8
|.
c ri
E
S -
Pittsburgh, PA
y=0.99x t -2.93. R-squared = 0.96
o
a
0 .*' o
>o
O
O,''
o o °- '
9' o
o , • ' o
Jo.l'l
° ,°6 °
o °.'
ga^>0 o° o
.dr^5 °
«si^^*
wS^
50 100 150 200 250 300 350
Original Bext
F-31
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
Baltimore, MD
y=0.95x + 2.18, R-squared = 0.9
°
1-
8-
"S
1 1-
8 -
s -
o
o ,' °
o ,•'
o
A
0-'
% o-' oo
0 .' 0
0
,6 o o
OjP '.$# °
0*.$ °
0 o°<»f
*a •*"
*P^ 0%°
^gjSo^o
^
50 100 150 200 250 300
Original Bext
Philadelphia. PA
y=0.99x * 6.83, R-squared - 0.93
I-
i •
S .
(N
—
3
c 8 -
E
S -
8 -
S -
o
o
0
_9X o
0 0' 0 0
o „ •' °
o o
D^*0
^''°
o-°
O ^'' °
,8-8'°
ft, ' O
56 °
o-"'
tf
50 100 150 200 250 300 350
Original Bext
F-32
-------
Figure F-4. Scatter Plots of Daily Maximum Daylight Hourly PM2.s Light Extinction
Calculated by Approach T (y, left plot) and W (y, right plot) versus the UFVA approach (x)
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide
regression statistics in the scatter plots in this figure.
New York, NY
y=0.87x + 5.46, R-squated = 0.95
g
§
8 -
H
*
I I -
8 -
O -
• b
0
O ,''
-° 0
o
o
o
o
,-'6o o
o o o o o
O o
p
o o'o
o
° O -o °O -
lh°°°
d^*°°
_a6 o
,'•
0 100 200 300 400
Original Bext
New York, NY
y=0.9x * 1.89. R-squared = 0.97
0
°
8 -
I
™
1 8-
8 -
0 -
o'
o
o
-6 OQ o
,' 0
,' o
,°o' °
o
o
,' o
o ,-
,-'° *° o
#®°o °
°t^!° °
as'^°
f-pSr o ^5
^r
,o
0 100 200 300 400
Original Bext
F-33
-------
Table F-5. Linear regression equation and R values (Rsq) for relating hourly PM2.s light extinction values (all daylight hours)
calculated using approach T (x in the equation) to those using the UFVA approach (y in the equation) by month for 15 urban
areas. A slope >1 indicates that approach T is biased low relative to the UFVA approach.
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide regression statistics in the scatter plots of
Figures F-3 and F-4.
Slope =1.25 +
Slope = 1.10-1.24
0.90 < slope < 1.10
Slope = 0.76 - 0.9
Slope <= 0.75
Approach T is biased low.
Approach T is biased high.
Month
Tacoma,
WA
Fresno, CA
Los
Angeles,
CA
Phoenix,
AZ
Salt Lake
City, UT
Dallas, TX
Houston,
TX
1
y=1.01*x
+1.92;
Rsq=0.99
y=1.06*x
+6.65;
Rsq=0.95
y=1.34*x
+-6.07;
Rsq=0.98
NA
y=0.93*x
+6.67;
Rsq=0.99
y=1.14*x
+1.25;
Rsq=0.93
y=0.9*x+
0.54;
Rsq=0.95
2
y=0.9*x+7
.15;
Rsq=0.96
y=1.15*x+
-0.69;
Rsq=0.97
y=1.28*x+
-5.96;
Rsq=0.99
y=1.2*x+0
.58;
Rsq=0.98
y=1.02*x+
4.81;
Rsq=0.98
y=l.l*x+l
.86;
Rsq=0.96
y=0.94*x+
1.15;
Rsq=0.95
3
y=0.96*x+
4.53;
Rsq=0.96
y=1.42*x+
4.5;
Rsq=0.89
y=1.35*x+
-6.53;
Rsq=0.98
y=1.36*x+
-6.26;
Rsq=0.78
y=1.15*x+
2.05;
Rsq=0.96
y=1.27*x+
-6.95;
Rsq=0.96
y=0.8*x+5
.75;
Rsq=0.97
4
y=1.28*x
+-4.87;
Rsq=0.97
y=1.72*x
+-10.63;
Rsq=0.77
y=1.41*x
+-10.4;
Rsq=0.98
y=1.09*x
+-1.26;
Rsq=0.95
y=1.21*x
+-0.35;
Rsq=0.87
y=1.13*x
+-3.03;
Rsq=0.94
y=0.96*x
+1.77;
Rsq=0.93
5
y=1.16*x
+-1.53;
Rsq=0.92
y=1.08*x
+4.36;
Rsq=0.79
y=1.35*x
+-7.94;
Rsq=0.97
y=0.76*x
+4.92;
Rsq=0.72
y=1.05*x
+1.72;
Rsq=0.83
y=1.05*x
+-1.42;
Rsq=0.95
y=0.88*x
+2.6;
Rsq=0.96
6
y=0.99*x+
1.42;
Rsq=0.96
y=1.14*x+
0;
Rsq=0.95
y=1.56*x+
-12.42;
Rsq=0.94
y=1.3*x+-
4.25;
Rsq=0.84
y=1.02*x+
1.41;
Rsq=0.9
y=0.97*x+
1.62;
Rsq=0.9
y=0.71*x+
13.31;
Rsq=0.89
7
y=0.97*x+
2.99;
Rsq=0.97
y=1.02*x+
2.87;
Rsq=l
y=1.74*x+
-30.53;
Rsq=0.89
y=0.52*x+
11.94;
Rsq=0.79
y=1.07*x+
-0.69;
Rsq=0.99
y=0.91*x+
4.45;
Rsq=0.93
y=0.74*x+
11.79;
Rsq=0.91
8
y=1.04*x
+0.19;
Rsq=0.96
y=1.3*x+
-5.87;
Rsq=0.94
y=2.21*x
+-74.46;
Rsq=0.92
y=1.12*x
+-3.46;
Rsq=0.94
y=l*x+0.
81;
Rsq=0.96
y=0.98*x
+0.58;
Rsq=0.95
y=0.9*x+
1.86;
Rsq=0.95
9
y=l.ll*x
+-1.16;
Rsq=0.98
y=l.l*x+
6.77;
Rsq=0.93
y=1.83*x
+-27.22;
Rsq=0.93
y=1.02*x
+-0.69;
Rsq=0.94
y=1.07*x
+-0.45;
Rsq=0.93
y=l*x+-
2.67;
Rsq=0.98
y=0.75*x
+7.22;
Rsq=0.97
10
y=0.93*x
+7.08;
Rsq=0.98
y=1.52*x
+-6.34;
Rsq=0.92
y=1.39*x
+-3.97;
Rsq=0.88
y=1.38*x
+-6.57;
Rsq=0.97
y=1.06*x
+2.39;
Rsq=0.92
y=1.04*x
+-0.44;
Rsq=0.97
y=0.97*x
+-1.47;
Rsq=0.98
11
y=1.01*x+
1.91;
Rsq=0.99
y=1.01*x+
16.31;
Rsq=0.97
y=1.32*x+
-14.27;
Rsq=0.99
y=1.13*x+
-2.63;
Rsq=0.96
y=1.05*x+
2.44;
Rsq=0.99
y=1.12*x+
-2.81;
Rsq=0.97
y=0.91*x+
3.46;
Rsq=0.97
12
y=0.93*x
+5.78;
Rsq=l
y=0.99*x
+7.29;
Rsq=0.95
y=1.15*x
+6.65;
Rsq=0.94
y=l.l*x+
6.23;
Rsq=0.99
y=0.9*x+
5.51;
Rsq=0.96
y=1.01*x
+-0.06;
Rsq=0.95
y=0.88*x
+3.24;
Rsq=0.94
F-34
-------
Month
St Louis,
IL
Birmingham
AL
Atlanta,
GA
Detroit, MI
Pittsburgh,
PA
Baltimore,
MD
Philadelphia
PA
New York,
NY
1
y=0.99*x
+1.76;
Rsq=0.98
y=0.92*x
+7.51;
Rsq=0.97
y=0.97*x
+5.24;
Rsq=0.94
y=1.08*x
+-3.49;
Rsq=0.97
y=0.94*x
+7.2;
Rsq=0.9
y=0.98*x
+2.1;
Rsq=0.98
y=0.95*x
+3.19;
Rsq=0.99
y=1.08*x
+1.14;
Rsq=0.96
2
y=0.83*x+
13.19;
Rsq=0.98
y=0.98*x+
3.84;
Rsq=0.96
y=l*x+1.5
9;
Rsq=0.98
NAW*
y=0.97*x+
3.64;
Rsq=0.98
y=0.84*x+
8.13;
Rsq=0.98
NAm
y=0.99*x+
-1.41;
Rsq=0.99
3
y=1.12*x+
-3.82;
Rsq=0.98
y=0.95*x+
4.16;
Rsq=0.97
y=0.81*x+
13.95;
Rsq=0.94
y=1.05*x+
0.45;
Rsq=0.98
y=0.95*x+
3.97;
Rsq=0.98
y=0.99*x+
2.51;
Rsq=0.99
y=1.05*x+
-0.06;
Rsq=0.98
y=0.92*x+
10.01;
Rsq=0.98
4
y=1.08*x
+0.2;
Rsq=0.97
y=1.06*x
+-2.25;
Rsq=0.97
y=1.01*x
+-2.51;
Rsq=0.92
y=1.17*x
+-3.64;
Rsq=0.97
y=1.02*x
+1.71;
Rsq=0.94
y=1.12*x
+-1.54;
Rsq=0.94
y=l*x+l.
67;
Rsq=0.93
y=1.15*x
+-5.37;
Rsq=0.98
5
y=1.24*x
+-10.62;
Rsq=0.92
y=0.92*x
+4.45;
Rsq=0.98
y=0.9*x+
4.55;
Rsq=0.94
y=1.23*x
+-5.53;
Rsq=0.95
y=1.06*x
+-3. 12;
Rsq=0.91
y=1.06*x
+-0.39;
Rsq=0.93
y=1.03*x
+-2.46;
Rsq=0.97
y=1.2*x+
-5.67;
Rsq=0.96
6
y=1.25*x+
-16.39;
Rsq=0.9
y=0.98*x+
-2.15;
Rsq=0.99
y=0.92*x+
4.58;
Rsq=0.96
y=1.07*x+
-1.23;
Rsq=0.97
y=1.06*x+
-2.33;
Rsq=0.91
y=1.07*x+
-2.87;
Rsq=0.83
y=1.06*x+
-6.29;
Rsq=0.91
y=1.02*x+
-0.08;
Rsq=0.95
7
y=1.19*x+
-17.28;
Rsq=0.95
y=0.95*x+
-0.84;
Rsq=0.97
y=0.94*x+
0.95;
Rsq=0.95
y=1.12*x+
-7.1;
Rsq=0.97
y=1.09*x+
-9.79;
Rsq=0.97
y=1.19*x+
-13.9;
Rsq=0.92
y=0.89*x+
-0.11;
Rsq=0.98
y=1.17*x+
-19.48;
Rsq=0.93
8
y=0.99*x
+-4. 12;
Rsq=0.98
y=0.96*x
+-0.66;
Rsq=0.96
y=0.98*x
+-0.29;
Rsq=0.99
y=1.01*x
+0.72;
Rsq=0.92
y=1.08*x
+-11.34;
Rsq=0.98
y=l*x+-
7.75;
Rsq=0.94
y=0.86*x
+3.68;
Rsq=0.97
y=1.05*x
+-4.46;
Rsq=0.97
9
y=1.02*x
+-3.36;
Rsq=0.97
y=0.91*x
+1.63;
Rsq=0.95
y=1.02*x
+-4.52;
Rsq=0.96
y=1.13*x
+-3. 04;
Rsq=0.93
y=1.06*x
+-5.84;
Rsq=0.97
y=1.09*x
+-6. 14;
Rsq=0.97
y=0.89*x
+2.27;
Rsq=0.98
y=1.28*x
+-2.87;
Rsq=0.89
10
y=1.17*x
+-4.2;
Rsq=0.91
y=0.96*x
+2.56;
Rsq=0.97
y=1.02*x
+-0.64;
Rsq=0.93
y=1.29*x
+-7.99;
Rsq=0.97
y=1.04*x
+-0.78;
Rsq=0.95
y=1.14*x
+-6.88;
Rsq=0.94
y=1.07*x
+-0.74;
Rsq=0.97
y=1.3*x+
-8.98;
Rsq=0.97
11
y=l*X+4.4
9;
Rsq=0.97
y=0.96*x+
5.23;
Rsq=0.97
y=1.15*x+
-4.56;
Rsq=0.94
y=l.l*x+0
.33;
Rsq=0.98
y=1.03*x+
1.25;
Rsq=0.99
y=0.97*x+
3.86;
Rsq=l
y=1.04*x+
-1.94;
Rsq=l
y=1.12*x+
-0.2;
Rsq=0.98
12
y=1.01*x
+0.69;
Rsq=0.98
y=0.9*x+
7.7;
Rsq=0.95
y=l*x+l.
05;
Rsq=0.98
y=1.05*x
+-2.05;
Rsq=0.98
y=0.94*x
+6.27;
Rsq=0.98
y=l.l*x+
-0.93;
Rsq=0.97
y=1.23*x
+-7.68;
Rsq=0.98
y=0.92*x
+7.36;
Rsq=0.98
: Because approach T requires a minimum of four CSN samples in a calendar month to calculate valid values of HF and DLEE, estimates for February in
Detroit and Philadelphia are not available from approach T so no regression results are shown. These months do have regression results for approach W in Table
F-6.
F-35
-------
Table F-6. Linear regression equation and R values (Rsq) for relating hourly PM2.s light extinction values (all daylight hours)
calculated using approach W (x in the equation) to those using the UFVA approach (y in the equation) by month for 15 urban
areas. A positive intercept indicates that approach W is biased low relative to the UFVA approach.
Note that the independent and dependent variables in Tables F-5 and F-6 are reversed from the order used to provide regression statistics in the scatter plots of
Figures F-3 and F-4.
Slope =1.25 +
Slope = 1.10-1.24
0.90 < slope < 1.10
Slope = 0.76 - 0.9
Slope <= 0.75
Approach W is biased low.
Approach W is biased high.
Month
Tacoma,
WA
Fresno, CA
Los
Angeles,
CA
Phoenix,
AZ
Salt Lake
City, UT
1
y=1.07*x
+0.02;
Rsq=0.99
y=1.03*x
+6.82;
Rsq=0.97
y=1.04*x
+8.5;
Rsq=0.99
NA
y=0.91*x
+6.4;
Rsq=l
2
y=1.06*x+
1.07;
Rsq=0.99
y=1.06*x+
7.93;
Rsq=0.98
y=1.02*x+
7.16;
Rsq=0.99
y=1.25*x+
0.12;
Rsq=0.98
y=0.94*x+
8.75;
Rsq=0.98
3
y=1.15*x+
-1.13;
Rsq=0.99
y=1.32*x+
8.06;
Rsq=0.89
y=1.05*x+
15.49;
Rsq=0.98
y=1.31*x+
-4.86;
Rsq=0.95
y=1.07*x+
3.8;
Rsq=0.96
4
y=1.23*x
+-4.3;
Rsq=0.96
y=1.75*x
+-12.08;
Rsq=0.8
y=1.17*x
+4.22;
Rsq=0.98
y=1.07*x
+-0.71;
Rsq=0.98
y=1.17*x
+-0.09;
Rsq=0.92
5
y=1.26*x
+-4.68;
Rsq=0.95
y=0.94*x
+9.16;
Rsq=0.78
y=1.13*x
+8.45;
Rsq=0.97
y=0.93*x
+1.96;
Rsq=0.89
y=1.17*x
+-0.77;
Rsq=0.87
6
y=1.01*x
+1.04;
Rsq=0.96
y=1.07*x
+1.79;
Rsq=0.95
y=1.36*x
+5.73;
Rsq=0.92
y=1.2*x+
-2.49;
Rsq=0.83
y=1.02*x
+1.15;
Rsq=0.94
7
y=0.96*x
+3.1;
Rsq=0.96
y=0.98*x
+4.51;
Rsq=l
y=1.56*x
+-15.11;
Rsq=0.91
y=1.09*x
+-1.41;
Rsq=0.98
y=1.06*x
+-0.55;
Rsq=l
8
y=l.l*x+
-1.77;
Rsq=0.97
y=1.2*x+
-3.17;
Rsq=0.94
y=1.94*x
+-56.56;
Rsq=0.93
y=l.ll*x
+-2.87;
Rsq=0.97
y=1.01*x
+0.56;
Rsq=0.98
9
y=1.18*x
+-3. 19;
Rsq=0.98
y=1.17*x
+2.83;
Rsq=0.96
y=1.56*x
+-12.84;
Rsq=0.95
y=1.04*x
+-1.29;
Rsq=0.97
y=1.04*x
+0.16;
Rsq=0.98
10
y=1.03*x
+1.15;
Rsq=0.98
y=1.25*x
+4.14;
Rsq=0.94
y=1.25*x
+1.07;
Rsq=0.94
y=l.l*x+
-1.01;
Rsq=l
y=l.ll*x
+1.19;
Rsq=0.96
11
y=1.02*x+
0.72;
Rsq=l
y=0.83*x+
47.17;
Rsq=0.93
y=1.03*x+
13.84;
Rsq=0.99
y=1.08*x+
-0.91;
Rsq=0.98
y=0.93*x+
8.49;
Rsq=0.98
12
y=0.98*x+
3.2;
Rsq=l
y=0.91*x+
14.37;
Rsq=0.98
y=1.06*x+
6.99;
Rsq=0.98
y=1.06*x+
8.02;
Rsq=0.99
y=0.84*x+
10.5;
Rsq=0.97
F-36
-------
Month
Dallas, TX
Houston,
TX
St Louis,
IL
Birmingham
AL
Atlanta,
GA
Detroit, MI
Pittsburgh,
PA
Baltimore,
MD
Philadelphia,
PA
1
y=1.12*x
+0.63;
Rsq=0.95
y=0.88*x
+1.47;
Rsq=0.97
y=0.95*x
+3.53;
Rsq=0.98
y=l*x+l.
7;
Rsq=0.99
y=0.97*x
+4.46;
Rsq=0.96
y=1.03*x
+1.64;
Rsq=0.97
y=0.92*x
+8.37;
Rsq=0.92
y=1.06*x
+-2.44;
Rsq=0.99
y=1.01*x
+-0.42;
Rsq=l
2
y=1.09*x+
2.57;
Rsq=0.97
y=0.86*x+
2.75;
Rsq=0.99
y=0.85*x+
10.08;
Rsq=0.97
y=1.03*x+
0.47;
Rsq=0.99
y=0.97*x+
1.55;
Rsq=0.96
y=0.97*x+
1.05;
Rsq=l
y=1.03*x+
0.13;
Rsq=0.97
y=0.93*x+
3.3;
Rsq=0.99
y=1.03*x+
-3.16;
Rsq=0.99
3
y=1.18*x+
-4.36;
Rsq=0.96
y=0.86*x+
2.15;
Rsq=0.99
y=1.08*x+
-1.58;
Rsq=0.99
y=1.01*x+
-1.58;
Rsq=0.99
y=0.9*x+7
.22;
Rsq=0.97
y=0.98*x+
5.45;
Rsq=0.98
y=1.06*x+
-3.39;
Rsq=0.98
y=0.96*x+
3.44;
Rsq=0.99
y=0.99*x+
2.14;
Rsq=0.99
4
y=1.01*x
+1.32;
Rsq=0.94
y=0.94*x
+0.32;
Rsq=0.99
y=1.04*x
+1.61;
Rsq=0.98
y=1.03*x
+-1.5;
Rsq=0.99
y=l*x+-
2.29;
Rsq=0.95
y=1.17*x
+-3.56;
Rsq=0.99
y=1.07*x
+-1.33;
Rsq=0.93
y=l.l*x+
-1.37;
Rsq=0.95
y=0.98*x
+1.53;
Rsq=0.97
5
y=0.93*x
+3.16;
Rsq=0.96
y=0.85*x
+2.96;
Rsq=0.99
y=1.09*x
+-2.77;
Rsq=0.95
y=0.96*x
+-0.53;
Rsq=0.99
y=l*x+-
3.22;
Rsq=0.98
y=1.02*x
+3.5;
Rsq=0.95
y=1.05*x
+-3.5;
Rsq=0.93
y=0.98*x
+4.15;
Rsq=0.95
y=1.02*x
+-2. 14;
Rsq=0.97
6
y=l*x+0.
88;
Rsq=0.95
y=0.92*x
+2.05;
Rsq=0.96
y=1.16*x
+-12.62;
Rsq=0.93
y=0.98*x
+-1.93;
Rsq=0.99
y=0.94*x
+2.64;
Rsq=0.98
y=1.02*x
+0.66;
Rsq=0.98
y=1.01*x
+-1.36;
Rsq=0.95
y=1.08*x
+-4. 61;
Rsq=0.85
y=1.05*x
+-6.57;
Rsq=0.92
7
y=0.96*x
+3.17;
Rsq=0.93
y=0.92*x
+3.75;
Rsq=0.95
y=1.05*x
+-7.34;
Rsq=0.98
y=0.92*x
+2.04;
Rsq=0.99
y=0.88*x
+7.28;
Rsq=0.96
y=1.09*x
+-5.24;
Rsq=0.98
y=1.03*x
+-6.36;
Rsq=0.98
y=1.14*x
+-11.71;
Rsq=0.94
y=0.94*x
+-3. 14;
Rsq=0.98
8
y=0.96*x
+2.08;
Rsq=0.97
y=0.89*x
+3.22;
Rsq=0.98
y=0.95*x
+-1.47;
Rsq=0.99
y=0.91*x
+3.61;
Rsq=0.98
y=0.96*x
+1.44;
Rsq=0.99
y=1.04*x
+-0.25;
Rsq=0.93
y=0.97*x
+-1.25;
Rsq=0.99
y=0.97*x
+-5.3;
Rsq=0.97
y=0.89*x
+2.37;
Rsq=0.98
9
y=0.93*x
+0.44;
Rsq=0.98
y=0.89*x
+0.03;
Rsq=0.99
y=l*x+-
2.58;
Rsq=0.98
y=0.94*x
+-0.87;
Rsq=0.99
y=0.98*x
+-1.52;
Rsq=0.97
y=1.09*x
+-2. 11;
Rsq=0.96
y=1.04*x
+-4.35;
Rsq=0.98
y=1.07*x
+-5.52;
Rsq=0.96
y=0.89*x
+2.41;
Rsq=0.99
10
y=0.98*x
+0.71;
Rsq=0.97
y=0.91*x
+0.18;
Rsq=l
y=1.16*x
+-4.78;
Rsq=0.96
y=1.01*x
+-0.64;
Rsq=0.98
y=0.95*x
+4.24;
Rsq=0.98
y=1.15*x
+-2.29;
Rsq=0.99
y=1.04*x
+-1.42;
Rsq=0.97
y=l.l*x+
-4.68;
Rsq=0.96
y=1.05*x
+-0.49;
Rsq=0.97
11
y=1.09*x+
-2.7;
Rsq=0.97
y=0.98*x+
-5.56;
Rsq=0.99
y=1.01*x+
2.76;
Rsq=0.98
y=1.01*x+
0.23;
Rsq=0.98
y=1.03*x+
4.43;
Rsq=0.93
y=1.07*x+
1.83;
Rsq=0.99
y=1.02*x+
2.42;
Rsq=0.99
y=l*x+2.3
3; Rsq=l
y=0.99*x+
-0.3;
Rsq=l
12
y=0.9*x+4
.72;
Rsq=0.97
y=0.86*x+
2.2;
Rsq=0.98
y=0.97*x+
2.83;
Rsq=0.99
y=0.97*x+
2.37;
Rsq=0.98
y=1.01*x+
-0.82;
Rsq=0.99
y=1.03*x+
-1.05;
Rsq=l
y=1.02*x+
1.75;
Rsq=0.97
y=1.09*x+
-1.26;
Rsq=0.99
y=0.98*x+
2.66;
Rsq=0.98
F-37
-------
Month
New York,
NY
1
y=1.07*x
+1.42;
Rsq=0.99
2
y=1.06*x+
-3.48;
Rsq=0.99
3
y=1.01*x+
5.6;
Rsq=0.99
4
y=0.99*x
+0.63;
Rsq=0.98
5
y=1.01*x
+2.12;
Rsq=0.98
6
y=1.04*x
+-2.02;
Rsq=0.98
7
y=1.09*x
+-10.86;
Rsq=0.97
8
y=l*x+0.
3;
Rsq=0.99
9
y=1.25*x
+-1.39;
Rsq=0.89
10
y=1.17*x
+-2.29;
Rsq=0.98
11
y=l.l*x+l
.54;
Rsq=0.96
12
y=0.98*x+
4.74;
Rsq=0.98
F-38
-------
Figure F-5 compares the three-year, daily maximum design values for PM2.5 light
,th
extinction developed using approaches T and W versus the UFVA approach, for both the 90 and
95th percentile forms. In many cases, all three methods yield similar design values, suggesting
that the stringency of emission controls required to meet a given level standard may be about the
same regardless of which approach is used to estimate PM2.5 light extinction. The exceptions are
for the highest levels of light extinction, which occur in Los Angeles, Fresno, and Detroit. Table
F-7 provides the same information as Figure F-5 in tabular form.
Figure F-6 compares the 1-hour PM2 5 light extinction estimates generated by approaches
T and W to each other in scatter plot format, for both all daylight hours and daily maximum
daylight hours. The dashed red line is the 1:1 line, and regression statistics are shown with each
scatter plot. The agreement between the two approaches is very close in most cases. In Fresno,
Los Angeles and to some degree also Dallas, there is a tendency for approach T to be biased low
relative to approach W for the points at the high end of the range of PM2 5 light extinction. This
is consistent with the expected tendency of the monthly averaging of HF and DLEE that is part
of approach T to mute the extreme daily values of HF and DLEE, extreme values that are
directly used for their respective days under approach W.
Figure F-7 compares the 2007-2009 90th and 95th percentile PM2 5 light extinction
design values calculated based on approach T versus approach W. Despite some scatter between
the approaches in Figure F-6, the design values from the two approaches are nearly identical.
This is not surprising, since using monthly average values of HF and DLEE instead of day-
specific values will sometimes result in an overestimate and sometimes in an underestimate of
PM2 5 light extinction. Also, the bias in approach T for Los Angeles, Fresno, and Dallas noted in
the previous paragraph mostly affects value above the 90th and 95th percentile points. This
suggests that approach T's ability to produce estimates of light extinction on every day, thereby
allowing the full history of hourly PM2.5 concentrations to be taken into account, may make it the
preferable approach because using all days' PM2 5 concentration provides more stability to the
90th or 95th percentile value than using only one-day-in-three or one-day-in six data.
It is interesting to note that in the "all hours" panels of Figure F-6, there are instances of
several points falling on a straight line somewhat offset from the main data cloud, which itself
seems less organized. These data points can be interpreted as days on which the day-specific
values of HF and/or DLEE were markedly different from their monthly-averaged values, while
relative humidity was stable for several hours. When RH is stable within part of a day, approach
T (and W) results in estimates of hourly light extinction that are directly proportional to hourly
PM2.5 concentration, with the proportionality constant dependent on HF and DLEE for the day.
The HF and DLEE for the day are always different between approaches T and W. The scatter
plot for Phoenix appears to have three such groups of points, for example. Houston appears to
have two groups.
F-39
-------
Figure F-5. Comparison of 2007-2009 PM2.s light extinction design values based on
approaches T and W vs. the UFVA approach
Daily Maximum PM2.5 Light Extinction
Design Values, 2007-2009
«
_D
c 400
tp
n
01
Q
0
1-
.c
a.
a.
<
100
i
f
- ' *
• • •
m*i
*J&
A"
*•
0 100 200 iOO 400 500 600
UFVA Design Value
* Approach T- 90th Percentlle • Approach T - 95th Percentile
A Approach W- 90th Percentile Approach W- 95th Percentile
Table F-7. Comparison of 2007-2009 PMi.s light extinction design values based on
approaches T and W vs. the UFVA approach
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
90l Percentile Daily Maximum
PM2.5 Light Extinction (Mm"1)
UFVA
121
312
434
71
152
162
145
224
269
189
275
222
183
280
295
T
121
260
332
68
144
144
164
209
257
190
252
219
188
287
272
W
111
276
358
66
150
151
162
217
271
186
261
214
176
294
275
95l Percentile Daily Maximum
PM2.5 Light Extinction (Mm"1)
UFVA
132
394
503
97
285
204
177
267
326
212
436
256
212
320
365
T
145
370
402
74
301
170
192
254
336
212
446
261
213
306
315
W
127
403
461
75
310
194
197
254
337
213
433
273
218
315
335
F-40
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
S -
Tacoma, WA
y=0.98x + 0.47, R-squared = 0.99
Daily Max, Tacoma, WA
y=0.97x * 0.88, R-squared = 0.99
II)
In
g -
S -
50 100 150 200 250
plan T PM2 5 bexl
50 100
150 200
plan T PM2 5 bext
250 300
Fresno, CA
jr=1.05x t -2, R-squared =0.99
Daily Max, Fresno, CA
y=1 -07x + -4.96, R-squared = 0.98
100 200 300 400 500 600
plan T PM2 5 bexl
200 300 400
plan T PM2 5 bext
500 600
F-41
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
-
8 -
-
Los Angeles, CA
y=1.08x + -3.76, R-squared = 0.99
Daily Max, Los Angeles, CA
y=1.15x+ -13.05, R-squared = 0.99
8 -
S -
8 -
100 200 300 400
plan T PM2 5 bext
500 600
200 300 400
plan T PM2 5 bext
500 600
8 -
-
S -
8 -
Phoenix, AZ
y=0.98x + 0.15, R squared = 0.93
Daily Max. Phoenix, AZ
y=0.96x + 1.25, R-squared = 0.9
S -
§ -
S
40 60 30
plan T PM2 5 bext
40 60 80
plan T PM2 5 bext
F-42
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
Salt Lake City, UT
y=1.02x + -0.66, R-squared = 0.99
Daily Max, Salt Lake City, UT
y=1.04x + -1.48, R-squared = 0.9
o .'
9-'
f,*
200 400 600
plan T PM2 5 bext
400 600
plan T PM2 5 bext
8
Dallas. TX
y=1,04x + -1.4, R-squared = 0.97
Daily Max. Dallas, TX
y=1.08x t -4.87, R-squared = 0.95
S
100 150
plan T PM2 5 bext
200 250
100 150
plan T PM2 5 bext
F-43
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
s -
3 8 -
I '
Houston, TX
y=0.95x t 2.73, R-squared = 0.94
Daily Max, Houston. TX
y=0.93x * 7.35, R-squared = 0.91
o _ -DE
100 150
plan T PM2 5 text
100 150
plan T PM2 5 bexl
3
I
S -
St. Louis, IL
y=1.02x + -0.56, R-squared = 0.99
Daily Max, St. Louis, IL
y=1.03x * -1.41, R-squared = 0.97
200 300
plan T PM2 5 text
200 300
plan T PM2 5 bext
F-44
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
3
I
Birmingham, AL
y=0.98x t 2.01, R-squared = 0.98
Daily Max, Birmingham, AL
y=0.96x * 6.19, R-squared = 0.98
400 600
plan T PM2 5 pext
400 600
plan T PM2 5 bext
3
I
Atlanta, GA
y=1x + 0.56, R-squared = 0.98
100 200
plan T PM2 5 text
Daily Max. Atlanta, GA
y=0.99x * 1.78, R-squared = 0.97
8 -
V
,' O O
100 200
plan T PM2 5 bext
F-45
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
3
I
Detroit, Ml
y=1.03x + -1.1, R-squared = 0.99
Daily Max. Detroit. Ml
y=1.04x + -2.89, R-squared = 0.99
400
plan T PM2 5 text
200 400
plan T PM2 5 bext
3
I
8
Pittsburgh, PA
y=1x + 0.19. R-squared = 0.99
Daily Max, Pittsburgh, PA
y=1.01x + -0.79, R-squared =0.98
8
100 200
plan T PM2 5 text
50 100 150 200 250 300 350
plan T PM2 5 bext
F-46
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
3
I
8 -
8
8
Baltimore, MD
y=1x + 0.62, R-squared = 0.99
Daily Max, Baltimore. MD
y=0.98x * 1.96, R-squared = 0.99
8 -
3 8
I
8
S
50 100 150 200 250 300
plan T PM2 5 pext
50 100
150 200
plan T PM2 5 bext
250 300
3
I
S -
Philadelphia, PA
y=1x + 0.68, R-squared = 1
Daily Max. Philadelphia, PA
y=1x + 1.59. R-squared = 0.99
8 -
100 200
plan T PM2 5 text
50 100 150 200 250 300 350
plan T PM2 5 bext
F-47
-------
Figure F-6. Scatter plots of daily maximum daylight hourly PM2.s light extinction
calculated by approach T (x) versus approach W (y).
All Daylight Hours
Daily Maximum Daylight Hour
3
I
8 -
New York. NY
+ -0.84, R-squared = 0.99
200 300
plan T PM2 5 text
i -
8
Daily Max. New York. NY
y=1.01x + -0.9, R-squared = 0.98
200 300
plan T PM2 5 bexl
F-48
-------
Figure F-7. Comparison of 2007-2009 PMi.s light extinction design values based on
approach T vs. approach W.
Approach W
500
450
400
350
300
250
200
150
100
50
0
(
PM2.5 Light Extinction Design Values (Mm'1)
u
jfu
^T * 90th percentile daily max
if'' • 95th percentile daily max
vfr-
X 95th percentile all hours
f
A 90th percentile all hours
A "
V
*"'
) 100 200 300 400 500
Approach!
F.6 EXPLANATION OF THE SELECTION OF 1.6 AS THE MULTIPLIER FOR
OCM
The original IMPROVE algorithm uses a multiplier of 1 .4 to estimate organic
carbonaceous material (OCM) from the measurement of organic carbon (OC), after subtraction
of an estimate of the organic carbon blank filter artifact. This value has been extensively
questioned and investigated since its initial selection, and it is clear from the literature that the
actual ratio of OCM to OC varies by site, season, PM sampler characteristics, the approach used
the artifact correction, and the OC quantification method. Various researchers and organizations
have suggested that other, generally higher, values be used instead of 1.4. The revised
IMPROVE algorithm, for example, is based on a factor of 1.8. For simplified approaches T and
W, the issue of an appropriate single value for this multiplier was re-investigated. This
investigation focused on data only from samples taken in 2007-2009 from sites and months
where the URG 3000N sampler was in use. The investigation considered all such CSN sites, of
which there were 171 sites with an average of 77 sample days each. Because of the phased
introduction of the URG 3000N sampler, there was a wide range in the number of sample days
across sites, and not all sites were seasonally balanced. These samples were all analyzed for OC
by Desert Research Institute using the current IMPROVE TOR method.
At each site, the SANDWICH estimate of OCM was used as the dependent (y) variable in
a linear regression with the OC measurement (minus an assumed artifact of 0.4 |ig/m3) as the
independent (x) variable. The slope of the regression line is an estimate of the OC-to-OCM
F-49
-------
multiplier for that site, across the seasons represented in the monitoring data. In addition, the
data from all sites were pooled and a single regression was performed for the pooled data. The
following results were observed.
• The average value for R-squared from the 171 regressions was 0.68, with a wide
range from site-to-site.
• The slopes of the 171 regression lines ranged from 0.57 to 3.05 (not including one
negative value from a site with only nine samples), with a median of 1.73 and an
average of 1.72.
• The sample size-based median slope was 1.68 and the sample size-weighted
average slope was 1.65.
• The slope of the pooled regression was 1.6.
The value of 1.6 from the pooled regression was selected for use as the carbon multiplier
in approaches T and W. Note that this is midway between the factor of 1.4 used in the original
IMPROVE algorithm and the factor of 1.8 used in the revised IMPROVE algorithm.
F.7 VARIABILITY IN ESTIMATES OF MONTHLY-AVERAGED HF AND DLEE
AND SELECTION OF A MINIMUM SAMPLE SIZE
As part of the analysis of simplified methods for this appendix, EPA staff investigated the
issue of the variability of monthly-averaged values of HF and DLEE, in light of the one-day-in-
six sampling schedule at many CSN sites. A one-day-in-six CSN site will typically attempt to
collect five samples in a month, versus 10 or 11 at a one-day-in-three site.
The margin of error (MOE) for the 90 percent confidence range for monthly-average HF
and DLEE was calculated for each site-month of data from CSN sites for 2007-2009 (including
many site-months prior to conversion to the URG 3000N sampler). The MOE is effectively the
half-width of the confidence range. The MOE reflects both the similarity of the parameter of
interest from day-to-day and the sample size.14 Site-months were then binned according to the
number of samples in the month. Figure F-7 shows the box plots of the MOEs for HF and DLEE
(referred to as ADLEE in the plot), versus the number of samples available. The "N" along the
top of each box plot indicates the number of CSN site-months that had the indicated number of
sample days. Based on these box plots, EPA staff decided to incorporate a minimum
requirement of four sample days per month for the calculation of valid monthly-average values
of HF and DLEE via approach T, for the purposes of this appendix. It should be noted that based
on the "N" values in the box plots, 85 percent of monthly averages for HF and DLEE were
actually based on five or more samples and have a smaller margin of error than shown for the
four-sample case.
14 The formula used was MOE = (STD ERR)*t(n-l, 0.95), where STD ERR = (sample standard deviation)/sqrt(n)
and t(n-l, 0.95) is the 95th percentile t-distribution with n-1 degrees of freedom (effectively removing 5% from each
tail to get 90%).
F-50
-------
Figure F-7. Margin of errors in monthly-averaged values of hygroscopic fraction (HF) and
dry light extinction efficiency (ADLEE in this figure) versus the number of CSN samples in
a site-month.
Hygroscopic Fraction Sample Size vs. 90% Margin of Error
lf>
o
d
o
LJJ
o n
c d
O)
5
0 CM
o
o
p
d
N=113
x
N=323 N=844 N=129 N=96 N=208 N=245 N=211 N=52
x
x
X
x _ x
8
— |— x x
X
. X X ^
V X
— T — ¥ ?
_L t=d ^ y
45
678
Sample Size
10
11
ADLEE Sample Size vs. 90% Margin of Error
is
£ d
o
en o
CJ
d
N=113 N=323
x
N=844 N=129 N=96 N=208 N=245 N=211 N=52
6 7
Sample Size
10
11
F-51
-------
This page left intentionally blank.
-------
APPENDIX G
Calculated 24-hour Average PM2.s Light Extinction and Adjusted
Candidate Protection Levels
G.I OVERVIEW
The concept of considering and adopting a relatively long averaging period for aNAAQS
indicator in order to control shorter-term ambient conditions is not at all unfamiliar in the history
of NAAQS reviews. This appendix investigates the possibility of using calculated 24-hour
average PM2.5 light extinction as the indicator and averaging period for a secondary PM2.5
NAAQS that would be aimed at controlling daily maximum daylight 4-hour average light
extinction conditions. This appendix assumes that the form for both standards is based on a
three-year average of annual 90th percentile values. A calculated 24-hour average approach
would avoid data quality uncertainties that have recently been associated with currently available
instruments for direct measurement of hourly PM25 light extinction or for measurement of
hourly PM2.5 mass.
The calculation procedure for the 24-hour average indicator addressed in this appendix is
based on the general method used by the IMPROVE monitoring program and the Regional Haze
Program to characterize 24-hour light extinction levels based on 24-hour filter-based sampling
for PM2.5 composition.1 There are several variations of this IMPROVE method, distinguished by
the use of the original versus the revised IMPROVE algorithm for light extinction and by the use
of same-day relative humidity (RH) at the monitoring site versus long-term (climatological)
relative humidity conditions. This appendix reports the results of using the original IMPROVE
algorithm, for consistency with previous results reported in the Urban Focused Visibility
Assessment (UFVA)2 and in Chapter 4 and Appendices E and F of this Policy Assessment (PA).
For relative humidity, this appendix applies long-term conditions, which is the approach used in
the Regional Haze Program for assessing progress towards the Program's goal of achieving
natural conditions of visibility in Class I areas by 2064. By using site-specific daily data on both
PM2.5 mass and composition and site-specific long-term relative humidity conditions, it can be
expected that the 24-hour average PM2.5 light extinction indicator explored in this appendix
would provide more protection of visibility than would the current secondary PM2.5 NAAQS
based only on 24-hour average and annual average PM2.5 mass. In particular, the systematic
difference in humidity conditions between the eastern states and the western states would be
accounted for in large degree.3
1 The IMPROVE monitoring program and the Regional Haze Program also use filter-based data on PM2 5 and PM10
mass, in order to estimate light extinction due to PM10_2 s- PM2.5 and PM10 mass values are not used in the indicator
investigated in this appendix, because this indicator excludes light extinction due to PM10_2.5.
2 Paniculate Matter Uiban-Focused Visibility Assessment Final Document, EPA 452/R-10-004, July 2010. The
UFVA presented a method and results for PM10 light extinction. The assessment in this memo focuses on PM2 5
light extinction.
3 The use of long-term relative humidity conditions in defining the NAAQS indicator would avoid the need for on-
site relative humidity data and hence would avoid the need to establish a reference method for relative humidity,
establish siting and operational requirements for the relative humidity measurements, and invest in new relative
humidity instruments. On the other hand, the use of long-term relative humidity conditions can be expected to
weaken to some degree the correlation between the calculated values of 24-hour average PM2 5 light extinction and
G-l
-------
This appendix investigates approaches for calculating adjusted Candidate Protection
Levels (CPLs) for a 24-hour average PM2 5 light extinction indicator that are generally equivalent
to CPLs of 20, 25, and 30 deciviews (dv) for a daily maximum daylight 4-hour average PM2.5
light extinction indicator, on an aggregate or "central tendency" basis. It was initially expected
that the values of 24-hour average PM2.5 light extinction and daily maximum daylight 4-hour
average PM2 5 light extinction would differ on any given day, with the shorter term peak value
generally being larger. This would mean that, in concept, the level of a NAAQS based on a 24-
hour average indicator should include a downward adjustment compared to the level that would
be applied to a NAAQS based on a daily maximum daylight 4-hour average indicator. As
discussed in section G.5, this initial expectation was verified for the 30 dv level but not for the
other two levels (25, 20 dv).
In developing adjusted CPLs, comparisons are made between values of 24-hour average
PM2.5 light extinction calculated as described in section G.2 and values of daily maximum
daylight 4-hour average PM2.5 light extinction calculated from the hourly results of the UFVA
approach for modeling hourly PM2.5 light extinction, for the original 15 study areas using 2007-
2009 data. The UFVA approach and the results of re-executing it for 2007-2009 data are
described in Appendix F. Table G-l shows the size and temporal distribution of the available
2007-2009 data base for the 15 study areas. Because 24-hour average PM2.5 light extinction can
be calculated with fewer required measurements as input, it is possible to calculate it for more
days than shown in Table G-l. However, comparisons of 24-hour and 4-hour values for the
purpose of developing adjusted CPLs can only be made for these days.
Table G-l. Number of Days per Quarter in Each Study Area
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Total Number
of Days
150
325
161
84
276
257
144
287
330
258
133
264
140
98
145
2007
01
13
26
21
0
23
18
15
29
30
25
11
22
12
13
19
02
13
28
26
0
25
23
14
25
30
19
11
22
12
14
15
03
14
30
24
0
19
24
9
22
27
26
12
23
17
12
19
04
13
28
24
0
29
21
0
10
26
21
11
23
16
12
21
2008
01
14
29
29
0
28
22
15
24
31
22
7
23
0
9
20
02
14
27
28
0
12
25
13
22
27
25
12
24
0
11
14
03
14
30
9
0
27
26
12
27
30
24
13
24
0
14
13
04
12
29
0
0
15
20
14
28
29
21
13
25
0
13
24
2009
01
7
15
0
11
13
13
7
14
13
12
3
10
13
0
0
02
14
27
0
24
29
23
15
30
28
22
15
25
26
0
0
03
11
28
0
21
28
21
15
29
28
21
10
22
21
0
0
04
11
28
0
28
28
21
15
27
31
20
15
21
23
0
0
actual daily maximum daylight 4-hour PM2 5 light extinction, since the latter depends on actual relative humidity
conditions along the sight path in the specific 4-hour period. It can be expected that the variability illustrated in this
appendix would be somewhat less if the 24-hour indicator were based on same-day measurements of relative
humidity either at the PM2 5 site or another site in the same area.
G-2
-------
G.2 CALCULATED 24-HOUR PM2.5 LIGHT EXTINCTION
This section describes and implements an approach for calculating 24-hour average PM2 5
light extinction using on-site 24-hour average PM2.5 chemical speciation data but no other same-
day data. The PM2 5 light extinction values calculated by the method described here are based on
the original IMPROVE algorithm and on long-term relative humidity conditions. Possible
variations, not further analyzed here, include the use of the revised IMPROVE algorithm and the
use of same-day relative humidity data. When converting PM2 5 light extinction values in Mm"1
to deciviews, the Rayleigh term must be included to avoid the possibility of negative values.
Table G-2 explicitly defines the calculation steps for the 24-hour average indicator
approach. The 24-hour average PM2 5 component concentrations needed for the calculation are
derived from 24-hour filter measurements, including a blank correction for organic carbon
artifacts, a multiplier to estimate organic carbonaceous material from the blank-corrected organic
carbon measurement, and the IMPROVE formula for estimating fine soil from concentrations of
five crustal elements. Instead of using same-day on-site relative humidity data, climatological
data are used for the humidity adjustment factor, f(RH). A spatial interpolation analysis of 1988-
1997 relative humidity data completed in 2001 for the Regional Haze program provided
monthly-average f(RH) values across the United States. 4 For this analysis, the monthly values
from the grid point nearest to a given Chemical Speciation Network (CSN) monitoring site were
used in the calculation of daily 24-hour PM2.s light extinction values for 2007-2009.
Figure G-l presents monthly-average relative humidity values graphically. Table G-3
contains the month-specific 24-hour average values of f(RH)hOUriy for each of the 15 cities. The
values in Figure G-l and Table G-3, as expected, show that the more eastern of the 15 cities have
higher values of f(RH) with the notable and expected exception of Tacoma which has the most
humid conditions of all the areas. For those areas with strong seasonal patterns of relative
humidity, the seasonal patterns of the values in Figure G-l and Table G-3 are also as expected.
Note that while some individual hours in 1988-1997 had relative humidity above 90%, no
monthly average relative humidity exceeds 90%. The highest value of monthly average f(RH) is
5.05 for December in Tacoma, when the average relative humidity value is 89%. The average
f(RH) value of 5.05 for this month is higher than the value of f(89%) = 3.93 because of the non-
linear nature of the f(RH) function. The single hour value of relative humidity corresponding to
an f(RH) value of 5.05 would be between 92% and 93%.
Figure G-2 presents the results of applying the 24-hour average approach in the form of a
box-and-whisker plot of the daily 24-hour PM2 5 light extinction values for 2007-2009. For ease
of comparison to similar figures in this document that present estimates of other metrics for
PM2.s light extinction, for example Figure F-l, horizontal lines are drawn to represent 65, 100,
and 190 Mm"1, although we emphasize that these are unadjusted values meant to be compared to
short-term averages of light extinction rather than to 24-hour averages. As expected, the
estimated values for 24-hour PM2 5 light extinction generally are lower than for daily maximum
daylight 1-hour PM2 5 light extinction (Figure F-l), but the comparisons across cities are
generally the same.
4 Interpolating Relative Humidity Weighting Factors to Calculate Visibility Impairment and the Effects of
IMPROVE Monitor Outliers, Report by Science Applications International Corporation to EPA, August 30, 2001.
http://vista.cira.colostate.edu/improve/publications/guidancedocs/DraftReportSept20.pdf. Note that this reference
describes a procedure in which hourly f(RH) values were capped at f(98%). Subsequently, the IMPROVE Steering
Committee decided to cap f(RH) values at f(95%). The f(RH) values used in here are based on the 95% cap.
G-3
-------
Table G-2. Calculation Steps for the 24-hour Average PM2.5 Light Extinction
24-hour Average Approach
Comments
(i) Using historical data, determine a representative long-term
monthly average of hourly f(RH) at the monitoring site, for
each month of the year. There will be 12 such values for any
monitoring site.
A spatial interpolation analysis of 1988-1997 relative humidity completed in 2001 for the Regional Haze program provided monthly-
average f(RH) values for 15,000 '/4-degree grid points of latitude/longitude across the United States. For this analysis, the monthly
values from the grid point nearest to a given PM2.5 (CSN) monitoring site were used in the calculation of daily 24-hour PM2.5 light
extinction values for 2007-2009. If a 24-hour average indicator were to be adopted as part of a secondary PM NAAQS, these same
gridded data could be specified, or EPA could develop updated tables of gridded monthly average f(RH) values and a procedure for
determining the values to be applied to any visibility monitoring site. Alternatively, the NAAQS could specify the use of same-day
site-specific relative humidity data, or data from another monitoring site in the same area, for example the nearest National Weather
Service site.
(ii) For each CSN sampling day, subtract a sampler-dependent
estimate of the OC artifact from the OC measurement, and
multiply by 1.4 to estimate organic carbonaceous material
(OCM).
OCM = (OC - artifact) * 1.4
The IMPROVE blank filter correction values are not representative of conditions in urban areas, and cannot be used in this approach.
The values for the OC artifact used for this analysis ranged from 0.32 to 1.53 ug/m3, depending on CSN sampler model. The artifact
adjustment for the URG 3000N sampler is of most interest prospectively, because it is the single sampler now in use for carbon
sampling in CSN. For those sites and days, an organic carbon artifact of 0.4 ug/m3 was assumed for the purposes of the UFVA and this
document, based on early experience with this sampler. EPA staff is currently exploring whether there is a better way to adjust for
organic carbon artifact based on a more recent, larger field blank and back-up filter data set.
1.4 is used for the multiplier to maintain consistency with the original IMPROVE algorithm.
(iii) For each CSN sampling day, calculate fine soil/crustal
PM2.5 (FS) from CSN measurements of crustal elements AL,
Si, Ca, Fe, and Ti, using the formula
. FS= 2.20 x[Al]+ 2.49 x [Si] + 1.63 x [Ca] +
2.42 x [Fe] + 1.94 x [Ti]
This is the same equation as used in the IMPROVE network and the Regional Haze program.
(iv) For each CSN sampling day, multiply CSN measurement
of sulfate ion by 1.375, and multiply CSN measurement of
nitrate ion by 1.29, to reflect associated ammonium under an
assumption of full neutralization.
. AS = Sulfate ion* 1.375
• AN = Nitrate ion * 1.29
(v) Apply the original IMPROVE algorithm to calculate 24-
hour average PM2 5 light extinction.
PM25 bext (Mm4) =
3[AS]f(RH) + 3[AN]f(RH) + 4[OCM] + 1[FS] + 10 [EC]
Where f(RH) is the monthly-averaged value from step (i) for
the particular month.
G-4
-------
Figure G-l. Month-specific 24-Hour Average Values of Relative Humidity from 1988-1997
— Birmingham
-I'hocnix
-Phil.iriplphia
-AllanU
-Baltimore
-Drl rail
-M. Louis
-NfwYmk City
-Dallas
- Houston
-iallLaketily
I acoma
Table G-3. Month-specific 24-Hour Average Values of f(RH) from 1988-1997
Month
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Jan
4.85
3.47
2.48
1.86
3.07
2.81
3.56
3.26
3.30
3.26
3.06
3.07
2.85
2.87
2.71
Feb
4.35
2.97
2.40
1.67
2.77
2.59
3.38
2.93
3.03
2.95
2.81
2.86
2.60
2.59
2.49
Mar
3.98
2.65
2.39
1.50
2.09
2.42
3.31
2.66
2.80
2.68
2.71
2.83
2.68
2.70
2.58
Apr
3.81
2.09
2.16
1.21
1.96
2.42
3.28
2.52
2.76
2.54
2.47
2.55
2.47
2.49
2.51
May
3.49
1.88
2.16
1.13
1.89
2.70
3.45
2.73
3.09
2.81
2.44
2.92
2.84
2.84
2.76
June
3.26
1.67
2.10
1.06
1.55
2.52
3.43
2.73
o o o
3.33
3.13
2.62
3.02
2.89
2.83
2.72
July
3.02
1.62
2.11
1.25
1.29
2.25
3.27
2.87
3.47
3.31
2.71
3.16
3.04
3.03
2.85
Aug
3.17
1.66
2.17
1.41
1.32
2.25
3.30
3.05
3.38
3.40
3.04
3.36
3.20
3.16
2.96
Sept
3.68
1.77
2.20
1.33
1.53
2.42
3.32
2.98
3.43
3.31
3.07
3.50
3.27
3.24
2.99
Oct
4.56
1.97
2.18
1.30
1.85
2.50
3.33
2.73
3.28
3.16
2.88
3.11
3.11
3.16
2.93
Nov
5.04
2.53
2.08
1.49
2.56
2.60
3.41
2.93
3.13
3.08
2.93
2.94
2.77
2.78
2.65
Dec
5.05
3.26
2.23
1.85
2.88
2.82
3.60
3.20
3.29
3.26
3.13
3.16
2.88
2.84
2.68
G-5
-------
Figure G-2. Distributions of Estimated 24-Hour Average PM2.s Light Extinction Across the 2007-2009 Period, by Study Area
using the 24-Hour Average Approach
c
o
£Z
IK
ui
c-i
ro
800:
700:
600:
500:
400:
300:
200:
100:
o-
8
I I i i i i
Tac Fres LA Pho SLC Dal
i i
HOU StL
Area
l
Bir
Atl Det
i i
Pit Bal
Phi NYC
G-6
-------
G.3 CALCULATED DAILY MAXIMUM DAYLIGHT 4-HOUR PM2.5 LIGHT
EXTINCTION
Values for the daily maximum daylight 4-hour average PM2.5 light extinction in the 15
study cities were calculated using the results of the re-execution of the UFVA approach for
2007-2009, described in Appendix F. The UFVA-approach modeling results (rather than the
modeling results using simplified approach T reported in Appendix F) are used because these
results are based on modeling assumptions believed to give the best available representation of
actual hourly light extinction, in that they incorporate the SANDWICH method for estimating
organic carbonaceous material (rather than a simple multiplier as in approach T) and on
information on site-specific diurnal concentration patterns from a run of a 2004 CMAQ
modeling platform. Note that these 4-hour estimates are based on same-day hourly relative
humidity data from the nearest site able to provide such data, not on gridded long-term data as
are the 24-hour PM2.5 light extinction values to which they are compared.
The daily maximum daylight 4-hour average PM2.5 light extinction values were
calculated from hourly UFVA-approach results by applying a moving 4-hour window to average
the hourly light extinction estimates. Only 4-hour periods that were entirely within the daylight
window were considered. At least three of four hours were required to have a valid 1-hour light
extinction estimate, meaning that necessary hourly PM2.5 mass and relative humidity data were
available to calculate light extinction for the hour and the relative humidity was not greater than
90%. As such, the resulting estimates represent indicator values that could be considered
comparable to a NAAQS with a 4-hour averaging period with the same restrictions for indicator
value validity. The requirement for at least 3 out of 4 hourly PM2.5 light extinction values
resulted in the elimination of 39 of the 3052 site-days shown in Table G-l, because those days
had no 4-hour daylight periods with at least 3 hours that had hourly PM2.5 data and had relative
humidity of 90% or less. Figure G-3 presents the results of this screening and calculation
procedure in the form of a box and whisker plot of the daily maximum daylight 4-hour average
PM2.5 light extinction values for 2007-2009. As expected, the comparisons across cities are
generally the same as for daily maximum daylight 1-hr PM2.5 light extinction (Figure F-l).
G-7
-------
Figure G-3. Distributions of Estimated Daily Maximum Daylight 4-Hour Average PM2.s Light Extinction Across the 2007-
2009 Period, by Study Area, Based on the UFVA Modeling Approach
800:
700:
600:
c
g
« 500
o5
4—I
IZ
CO
400:
CN
S
D_
X
ra
300:
Q 200:.
100:
o-
O
i i i i i i
Tac Fres LA Pho SLC Dal
Hou
I
StL
Area
i
Bir
Atl
Det
I
Pit
Bal Phi
NYC
G-8
-------
G.4 COMPARISON OF CALCULATED DAILY MAXIMUM DAYLIGHT 4-HOUR
PM2.5 LIGHT EXTINCTION AND CALCULATED 24-HOUR PM2.5 LIGHT
EXTINCTION
The daily values of the 24-hour average PM2.5 light extinction from section G.2 and the
values of the daily maximum daylight 4-hour average PM2.5 light extinction from section G.3 are
compared here across days within each city, to gain a sense of how they compare and as a quality
control check for reasonableness. Before doing so, both sets of daily PM2.5 light extinction
values were converted to deciview units, using the following equation for the deciview metric.
Note that the Rayleigh term of 10 Mm"1 is included in order avoid the possibility of negative
deciview values, but no term is included for light extinction due to PMio-2.5-
Deciview (dv) = 10 In ((PM2.5 bext in Mm"1 + 10)710)
Figures G-4 and G-5 compare the magnitudes of the daily values of the 24-hour and daily
maximum daylight 4-hour average PM2.5 light extinction estimates expressed in deciviews.
Figure G-4 shows a scatter plot for each of the 15 cities including the best-fit linear regression
line for each city, with the 4-hour light extinction value treated as the independent (x-axis)
variable in the regression. It can be seen that there is scatter around the regression line for each
city, because the estimate of 4-hour light extinction includes day-specific and hour-specific
influences that are not captured by the simpler 24-hour approach. Figure G-5 shows a scatter
plot for the pooled data points from all 15 cities. In both figures, the linear regression line is
solid and the 1:1 line is dashed. In both figures, there are some data points - each representing a
site-day - for which the value of 24-hour average PM2.5 light extinction is higher than the value
of daily maximum daylight 4-hour PM2.5 light extinction. Directionally, the focus on the daily
maximum 4-hour average should tend to make the 4-hour average number higher than the 24-
hour average number because PM2.5 concentrations in urban areas generally peak during the day
rather than at night. The directionally opposite comparison seen for some data points can result
due to the use of 1988-1997 humidity values that in some cases may be higher than the actual
same-day conditions used to calculate the 4-hour light extinction values. Approximately one-
half of all site days can be expected to have same-day conditions that are more humid than the
long-term average conditions, if there is no average bias between the gridded 1988-1997 relative
humidity data and 2007-2009 site conditions. The use of same-day hourly relative humidity data
to calculate daily 24-hour average f(RH) would avoid this, and in theory result in data points that
lie closer to the regression line. It is also quite possible for 24-hour average values of f(RH) to
be higher than for hours during the daylight. A higher value for 24-hour PM2.5 light extinction
can also in theory result from the invalidation of one or more 4-hour PM2.5 light extinction values
on a day due to the screen for hourly relative humidity above 90% (resulting in a time period
with lower light extinction providing the valid daily daylight 4-hour maximum), or from a
diurnal pattern in which 4-hour average PM2.5 mass peaks during the night rather than during
daylight.
Table G-3 presents the city-specific regression coefficients for the scatter plots in Figure
G-4, the unweighted average of these coefficients across cities, and the regression coefficients
for the pooled 15-city data shown in Figure G-5. While in broad terms most of the individual
scatter plots seem to have similar variability about the regression line, some values of R-squared
are notably lower. Closer examination shows that the lower values of R-squared are for the
cities with the smaller range of light extinction conditions.
G-9
-------
Figure G-6 and Table G-4 describe the distributions of the daily ratio of the 24-hour and
daily maximum daylight 4-hour values of PM2 5 light extinction. For each day, the 4-hour value
was divided by the 24-hour value. The distribution of these ratios across 2007-2009 is shown in
Figure G-6 in box-and-whisker plot format. Table G-4 presents the minimum, 10th percentile,
25th percentile, mean, median (50th percentile), 75th percentile, 90th percentile, and maximum
values from the distribution of ratios in each city.
It is also of interest to compare the annual 90th percentile values and the 3-year design
values for 24-hour average and daily maximum daylight 4-hour average PM2 5 light extinction
across areas. Table G-5 presents the annual 90th percentile values and the 3-year 2007-2009
design values.5 There are only a few cases in which the 24-hour average annual percentile or 3-
year design value is notably larger than the corresponding daily maximum daylight 4-hour
average value, despite the occurrence of many such points in the scatter plots of daily values.
Figure G-7 is a scatter plot of the 4-hour versus the 24-hour 90th percentile values for individual
years of data. Figure G-8 is a scatter plot of the 4-hour versus the 24-hour design values. The
equation of the linear regression line for each scatter plot is shown below the graphic.
5 For the purpose of this analysis, the "3 -year" design values have been calculated with only 2 years of data for Los
Angeles, Baltimore, Philadelphia, and New York City and one year of data for Phoenix because the re-execution of
the UFVA approach for 2007-2009 created estimates only for those years. See Table G-l. Also, no data
completeness minimums have been applied. The smallest sample used to determine an annual 90th percentile was 37
days, for Houston in 2007.
G-10
-------
Figure G-4. Scatter Plots of 24-Hour Light Extinction (y-axis) and Daily Maximum Daylight 4-Hour Light Extinction (x-axis)
for 15 Individual Cities, in deciviews
."• .'X
Los Angeles
Salt Lake City
G-ll
-------
Figure G-4. Scatter Plots of 24-Hour Light Extinction (y-axis) and Daily Maximum Daylight 4-Hour Light Extinction (x-axis)
for 15 Individual Cities, in deciviews (cont.)
Birmingham
Pittsburgh
G-12
-------
Figure G-5. Scatter Plot of 24-Hour Light Extinction (y-axis) versus Daily Maximum
Daylight 4-Hour Light Extinction (x-axis) for All Cities Pooled Together, in deciviews
dv 24
~
40
All 15 cities
30-
25-
20-
15
10-
5-
Blue line = regression line
Black dashed line = 1:1 line
10
15
20 25
dv 4x
30
35
40
45
G-13
-------
Table G-3. Regression Results for 24-Hour Light Extinction versus Daily Maximum
Daylight 4-Hour Light Extinction for Individual Cities and for All Cities Pooled Together,
in deciviews (4-hour light extinction is the independent variable)
Study Area
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Average Across Cities
(equal weighting)
Regression Using
Pooled Data
Intercept
1.520
-0.726
2.749
3.360
-1.255
5.418
10.224
6.563
7.210
4.912
4.163
4.101
5.925
2.797
8.109
4.338
3.507
Slope
0.948
0.961
0.760
0.804
0.926
0.625
0.500
0.655
0.684
0.740
0.780
0.828
0.736
0.761
0.626
0.756
0.790
R-squared
0.586
0.823
0.814
0.616
0.795
0.634
0.423
0.608
0.641
0.641
0.816
0.797
0.701
0.564
0.645
0.674
0.715
G-14
-------
Figure G-6. Distributions of the Ratio of Daily Maximum Daylight 4-Hour Light Extinction, in deciviews (numerator), and
24-Hour Light Extinction, in deciviews (denominator)*
3.0-
2 5-
1.5-
1.0-
0,5
o.o-
I I I ! I I I I I
Tac Fees LA Pho SLC Dal Hou StL Bir
i i i i i i
Ati Oet Pit Bai Phi NYC
Area
* Whiskers are at 10th and 90th percentiles.
G-15
-------
Table G-4. Points on the Distributions of the Ratio of Daily Maximum Daylight 4-Hour
Light Extinction (numerator) and 24-Hour Light Extinction (denominator), in deciviews
Study Area
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia PA
New York, NY
Average Across Cities
(equal weighting)
Min
0.169
0.718
0.648
0.535
0.562
0.574
0.458
0.282
0.498
0.121
0.631
0.614
0.569
0.796
0.584
10th
%-tile
0.689
0.892
0.960
0.788
0.946
0.908
0.751
0.831
0.775
0.882
0.828
0.839
0.794
0.956
0.829
25th
%-tile
0.882
0.986
1.089
0.868
1.041
1.009
0.862
0.953
0.879
0.965
0.938
0.913
0.857
1.064
0.922
Mean
1.009
1.106
1.163
0.970
1.223
1.126
0.970
1.056
0.987
1.047
1.037
0.986
0.979
1.170
1.028
1.057
Median
1.015
1.097
1.172
0.962
1.182
1.114
0.972
1.057
0.986
1.044
1.032
0.987
0.996
1.150
1.018
1.052
75th
%-tile
.135
.192
.246
.039
.359
.224
.084
.163
.095
.137
.139
.052
.077
.262
.133
90th
%-tile
.254
.314
.340
.149
.566
.351
.205
.268
.191
.240
.220
.132
.154
.420
.241
Max
2.521
2.087
.638
.840
2.319
.902
.501
.600
.523
.541
.692
.348
.393
2.008
1.542
Table G-5. Annual 90th Percentile Values and 3-year 2007-2009 90th Percentile Design
Values for Daily Maximum Daylight 4-Hour PMi.s Light Extinction and 24-Hour
Light Extinction, in deciviews
Study Area
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia PA
New York, NY
4-hour light extinction
Design Value
22.35
31.71
34.64
18.08
24.80
26.61
24.59
29.03
28.23
27.35
30.38
28.87
27.94
31.30
32.51
2007
22.93
33.95
35.50
26.10
27.91
24.85
28.90
29.60
29.65
31.18
30.54
28.74
30.77
33.50
2008
21.40
32.96
33.78
26.10
26.10
24.68
29.55
29.12
27.15
29.23
29.55
31.82
31.53
2009
22.72
28.21
18.08
22.19
25.80
24.25
28.62
25.95
25.26
30.73
26.53
27.15
24-hour light extinction*
Design Value
24.29
31.45
29.95
20. 54
24.54
22.90
24.20
27.10
28.07
25.89
29.96
28.39
27.15
28.27
29.31
2007
26.74
34.78
31.22
26.46
23.22
24.90
28.27
30. 45
28.85
31.57
30.87
28.51
29.07
30.11
2008
23.61
31.66
28.68
25.88
23.51
24.20
27.60
27.97
25.80
29.34
27.85
27.47
28.51
2009
22.51
27.91
20. 54
21.28
21.97
23.60
25.42
25.80
23.03
28.96
26.46
25.80
Entries in italics indicate cases in which the 24-hour average value is larger than the corresponding 4-hour average
value.
G-16
-------
Figure G-7. Scatter Plot of 2007-2009 Annual 90 Percentile Values for Daily Maximum
Daylight 4-Hour PMi.s Light Extinction (x-axis) and 24-Hour Average PMi.s Light
Extinction (y-axis), in deciviews
All 15 cities
Annual24 pct90
45
40
35
30
25
20
15
10
5 10 15 20 25 30 35 40 45
Annual4x pct90
Model: Annual24_pct90 = 6.636 + 0.725 * Annual4x_pct90 (rsquare=0.734)
G-17
-------
Figure G-8. Scatter Plot of 90 Percentile 2007-2009 3-year Design Values for Daily
Maximum Daylight 4-Hour PMi.s Light Extinction (x-axis) and 24-Hour Average
Light Extinction (y-axis), in deciviews
DV24_pct90
45
40
35
30
25
20
15
10
All 15 cities
10
15
20 25
DV4x_pct90
30
35
40
45
Model: DV24_pct90 = 8.786+ 0.646 * DV4x_pct90 (rsquare=0.820)
G-18
-------
G.5 DEVELOPMENT OF ADJUSTED INDICATOR LEVELS APPROPRIATE FOR
COMPARISON TO CALCULATED 24-HOUR PM2.5 LIGHT EXTINCTION
This section explores several different approaches for calculating adjusted CPLs for a 24-
hour average PM2 5 light extinction indicator that may be considered to be generally equivalent
on an aggregate or "central tendency" basis to CPLs of 20, 25, and 30 deciviews (dv) for a daily
maximum daylight 4-hour average PM2.5 light extinction indicator.
Several possible approaches are investigated in order to examine how robust the adjusted
levels are to the choice of a specific approach. In some of the approaches explored in this
section, the "equivalent" levels for individual cities are determined and then averaged across
cities to get a single adjusted level. For these approaches, the range of city-specific adjusted
levels is provided to give a sense of the possible variability in protection resulting from the use
of the longer averaging period and long-term relative humidity data.
The following approaches were applied. The results are shown in Table G-6, rounded to
the nearest whole-number value in deciviews. While some deciview results in Table G-5 were
expressed with two decimal digits, EPA staff believes it would be most appropriate to set a
NAAQS level in whole deciviews, given the uncertainties in the chain of analyses that may be
used to determine that level.
• Approach A - The regression line based on city-specific 3-year design values
displayed in Figure G-8 was used to find the 24-hour equivalents of 4-hour levels
of 20, 25, and30dv.
• Approach B - The regression line based on city-specific annual 90th percentile
values displayed in Figure G-7 was used to find the 24-hour equivalents of 4-hour
levels of 20, 25, andSOdv.
• Approach C - The regression line using all days for each city (Figure G-4 and
Table G-3) was used to find 24-hour equivalents of 4-hour levels of 20, 25, and
30 dv. These results were then averaged across the 15 cities. Table G-7 shows
the intermediate adjusted levels for each of the cities.
• Approach D - The pooled regression line using all site-days (Figure G-5 and the
last row of Table G-3) was used to find the 24-hour equivalents of 4-hour levels
of 20, 25, and30dv.
• Approach E - In each city, the median ratio of the 24-hour PM2.5 light extinction
indicator to the 4-hour PM2.5 light extinction indicator was applied to 20 dv (and
also applied to 25 and 30 dv in turn). The median ratio was used to minimize the
influence of days with extreme ratios, although Table G-4 indicates that in most
cases the difference between the mean and median ratios is small. These city-
specific results were averaged across the 15 cities. Table G-8 shows the
intermediate adjusted levels for each of the cities.
Of these approaches, EPA staff considers Approaches A and B, which give nearly
identical results, to be the most appropriate for further consideration, given the high R-squared of
the regressions and the fact that the regression lines are determined by data from days with PM2.5
light extinction conditions in the range of 20 to 40 dv. In contrast, the city-specific regression
lines and the pooled data regression lines are influenced by PM2.5 light extinction conditions well
G-19
-------
below 20 dv because of the large amount of data in that range. Approaches A and B yield
identical results except for the case of a 4-hour level of 20 dv, for which the A and B results are
only 0.65 dv different before rounding but after rounding differ by 1 dv. Both A and B result in
a 24-hour level that is higher than 20 dv for this case. EPA staff recommends consideration of
the results from Approach B, which are adjusted CPLs of 21, 25, and 28 dv, as the 24-hour PM2.5
light extinction indicator levels. These levels can be considered to be generally equivalent in an
aggregate or central tendency sense to CPLs of 20, 25, and 30 dv applied to a daily maximum
daylight 4-hour PM2 5 light extinction indicator.
As seen above, the adjusted 24-hour CPL can be greater than an "equivalent" 4-hour
level. The regression-based Approach B is sensitive to specific cases where use of the monthly
average historical f(RH) values yields higher estimates of light extinction than seen when using
the actual same-day RH conditions. This appears to occur more frequently at the upper end of
the 24-hour distribution (i.e., design values or 90th percentiles). Figure G-7 shows three data
points, out of 39, in which the annual 90th percentile, 24-hour average PM2.5 light extinction
value are significantly greater than their daily maximum 4-hour daylight counterpart. These data
points represent Tacoma (2007), Tacoma (2008), and Phoenix (2009). The most extreme case is
Tacoma (2007). Because there are 51 observations for this site-year, the sixth high value is the
90th percentile value. For the daily maximum 4-hour case the sixth high occurs on 10/9/2007
(22.93 dv). The average of the four hourly RH values that went into that day's calculation of the
4-hour indicator was 78%. For the 24-hour average case, the sixth high occurs on 11/02/2007
(26.74 dv). The 24-hour average of the RH values that went into that calculation was 89%
(based on the monthly average RH estimates). On this day, the 4-hour value was only 17.23 dv
(24th ranked) and the average of the four hourly RH values that went into that day's calculation
of the 4-hour indicator was 59%. On the five higher 24-hour light extinction days in Tacoma in
2007, the RH used for the 24-hour calculation was, on average, 15% higher than the actual used
for the sub-daily maximum. As a quick sensitivity analysis, we determined that if one were to
remove these three data points (i.e., Tacoma (2007 and 2008) and Phoenix (2009)) from the
Approach B regression analysis, the resulting CPLs would be: 20, 24, and 28 dv. While some of
these RH differences are surely due to diurnal variations and are therefore inherent and
appropriate in the CPL adjustment process, some may also be due to trends in RH over the years
and/or site-related differences (the climatological values were based on spatial interpolation of
only NWS monitoring sites, while many of the RH values used to calculate the 4-hour indicator
were obtained from non-NWS instruments at the CSN site itself). Consideration should be given
to re-executing the analysis in this Appendix using same-day, same-site relative humidity data, to
eliminate the possibility that differences in humidity conditions between 1988-1997 and 2007-
2009 and/or differences in RH sites have substantially affected the final equivalent levels.
In considering the results shown in Table G-6, it can be kept in mind that 1 deciview is
about the amount that a person can distinguish when viewing scenic vistas, and that a difference
of 1 deciview is equivalent to about a 10% difference in light extinction expressed in Mm"1.
Examining Table G-5, there are three cases in which comparing the original CPLs of 20, 25, and
30 dv to calculated 4-hour PM2.5 light extinction produces a different outcome than comparing
the adjusted CPLs of 21, 25, and 28 dv to calculated 24-hour PM2.5 light extinction. Dallas is
estimated to fail a 4-hour CPL of 25 dv, but to pass a 24-hour CPL of 25 dv, using an appropriate
rounding convention. Philadelphia is estimated to fail a 4-hour CPL of 30 dv, but to pass a 24-
hour CPL of 28 dv. Detroit is estimated to pass a 4-hr CPL of 30 dv, but to fail a 24-hr CPL of
28 dv. In each case, only a small difference in design value is responsible.
G-20
-------
Table G-6. "Equivalent" Levels for Calculated 24-Hour PM2.5 Light Extinction Using Five
Approaches
Approach
A
B
(Preferred)
C
D
E
Description
3 -year 90th percentile design
values regression
Annual 90th percentile values
regression
All-days city-specific
regressions, then averaged
All-days pooled regression
Median ratios, then averaged
24-hour level
equivalent to
20 dv
for 4-hour
(range among
15 cities)
22 dv
21 dv
19 dv
(17-20)
19 dv
19 dv
(16-20)
24-hour level
equivalent to
25 dv
for 4-hour
(range among
15 cities)
25 dv
25 dv
23 dv
(20-24)
23 dv
24 dv
(20-25)
24-hour level
equivalent to
30 dv
for 4-hour
(range among
15 cities)
28 dv
28 dv
27 dv
(24-29)
27 dv
29 dv
(26-30)
Table G-7. Intermediate Results for Approach C
Study Area
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia PA
New York, NY
Average
Minimum
Maximum
Intercept
(Table G-3)
1.520
-0.726
2.749
3.360
-1.255
5.418
10.224
6.563
7.210
4.912
4.163
4.101
5.925
2.797
8.109
Coefficient
(Table G-3)
0.948
0.961
0.760
0.804
0.926
0.625
0.500
0.655
0.684
0.740
0.780
0.828
0.736
0.761
0.626
20 dv equiv.
20.47
18.49
17.96
19.44
17.27
17.92
20.22
19.67
20.89
19.72
19.77
20.66
20.65
18.02
20.62
19.45
17.27
20.89
25 dv equiv.
25.21
23.29
21.76
23.46
21.91
21.04
22.72
22.95
24.31
23.42
23.67
24.80
24.33
21.82
23.75
23.23
21.04
25.21
30 dv equiv.
29.95
28.09
25.56
27.47
26.54
24.17
25.22
26.22
27.73
27.13
27.57
28.94
28.02
25.63
26.88
27.01
24.17
29.95
G-21
-------
Table G-8. Intermediate Results for Approach E
Study Area
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St Louis, IL
Birmingham AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia PA
New York, NY
Average
Minimum
Maximum
Median Ratio
4-hr : 24-hour
(Table G-4)
1.015
1.097
1.172
0.962
1.182
1.114
0.972
1.057
0.986
1.044
1.032
0.987
0.996
1.150
1.018
20 dv equiv.
19.70
18.23
17.06
20.80
16.92
17.95
20.58
18.92
20.29
19.16
19.37
20.27
20.08
17.39
19.65
19.09
16.92
20.80
25 dv equiv.
24.63
22.79
21.32
25.99
21.15
22.44
25.73
23.64
25.37
23.95
24.21
25.33
25.10
21.73
24.56
23.86
21.15
25.99
30 dv equiv.
29.55
27.35
25.59
31.19
25.38
26.92
30.87
28.37
30.44
28.74
29.06
30.40
30.12
26.08
29.47
28.64
25.38
31.19
G.6 SUMMARY
EPA staff has investigated the possibility of using a particular calculated 24-hour average
PM2.5 light extinction as the indicator and averaging period for a secondary PM2.5 NAAQS that
would be aimed at controlling daily maximum daylight 4-hour average light extinction
conditions, focusing on a form for both standards that is based on a three-year average of annual
90th percentile values. A calculated 24-hour average approach would avoid data quality
uncertainties that have recently been associated with currently available instruments for direct
measurement of hourly PM2.5 light extinction or for measurement of hourly PM2.5 mass. The
particular 24-hour indicator considered by EPA staff uses the original IMPROVE algorithm and
long-term (1988-1997) relative humidity conditions to calculate light extinction due to PM2.5.
By using site-specific daily data on PM2.5 composition and long-term relative humidity
conditions interpolated to each specific site, this 24-hour average indicator would provide more
consistent protection of visibility than would a secondary PM2.5 NAAQS based only on 24-hour
or annual average PM2.5 mass. In particular, this approach would account for the systematic
difference in humidity conditions between most eastern states and most western states. Possible
variations include the use of the revised IMPROVE algorithm and the use of same-day
measurements of relative humidity either at the visibility monitoring site or another site in the
same area. These variations in the method used to calculate PM2.5 light extinction on a 24-hour
G-22
-------
basis would be expected to provide somewhat more consistent protection in areas across the
country.
In concept, in order to provide generally equivalent protection, the level of a NAAQS
based on a 24-hour average indicator should include an adjustment compared to the level that
would be applied to a NAAQS based on a daily maximum daylight 4-hour average indicator.
Using 15 study sites, EPA staff investigated five approaches to making this adjustment, for 4-
hour indicator NAAQS levels of 20, 25, and 30 dv. An approach (B) thought by EPA staff to be
more appropriate for further consideration yielded adjusted NAAQS levels of 21, 25, and 28 dv
as the 24-hour PM2.5 light extinction indicator levels that are generally equivalent in an aggregate
or central tendency sense to levels of 20, 25, and 30 dv applied to a daily maximum daylight 4-
hour PM2.5 light extinction indicator.
G-23
-------
This page left intentionally blank.
-------
APPENDIX H
Table H-1. Predicted Number of Counties Not Likely to Meet Current Secondary PM2.s Standards and Potential
Alternative Secondary Standards Based on a 4-Hour Calculated PM2.s Light Extinction Indicator1
Region >
Total # of counties with suitable data to
calculate the indicator >
Standard
Statistic
All
U.S.
114
Northeast
24
Southeast
28
Industrial
Midwest
31
Upper
Midwest
8
Southwest
2
Northwest
15
Southern
California
5
Other
Areas
1
Number of counties and percentage of counties with suitable data2
Current Standards - PM2 5 mass indicator
Annual
15 ug/m3
24-Hour
35 ug/m3
# counties
% counties
15
13%
0
0%
1
4%
3
10%
0
0%
0
0%
6
40%
4
80%
1
100%
Alternative Standards - Calculated PM2 5 Light Extinction Indicator
T7 2
Form
90th
90th
90th
90th
90th
90th
Level
(Mm"1)
25
26
27
28
29
30
# counties
% counties
# counties
% counties
# counties
% counties
# counties
% counties
# counties
% counties
# counties
% counties
96
84%
79
69%
58
51%
47
41%
32
28%
18
16%
23
96%
20
83%
17
71%
12
50%
10
42%
6
25%
22
79%
14
50%
6
21%
3
11%
0
0%
0
0%
31
100%
29
94%
27
87%
26
84%
16
52%
7
23%
6
75%
5
63%
1
13%
1
13%
1
13%
0
0%
1
50%
1
50%
0
0%
0
0%
0
0%
0
0%
7
47%
4
27%
1
7%
1
7%
1
7%
1
7%
5
100%
5
100%
5
100%
4
80%
4
80%
4
80%
1
100%
1
100%
1
100%
0
0%
0
0%
0
0%
1 3-year average of annual 90th percentile daily maximum 4-hour daylight PM2 5 light extinction, excluding hours with relative humidity >90%.
2 Design values for comparison with the level of the standard were calculated based on approach T, using 2007-2009 monitoring data, if at least 2500 hours of
2005-2007 data were available. (See Appendix F for the description of approach T). Actual future outcomes may differ from these estimates due to changes in
instrumentation, siting, and/or the specific procedure for calculating the indicator.
H-1
-------
Table H-2. Predicted Number of Counties Not Likely to Meet Current Secondary PM2.s Standards and Potential Alternative
Secondary Standards Based on a 24-Hour Average Calculated PM2.s Light Extinction Indicator3
Region >
Total # of counties with suitable data
to calculate the indicator >
Standard
Statistic
All
U.S.
187
Northeast
33
Southeast
47
Industrial
Midwest
53
Upper
Midwest
10
Southwest
9
Northwest
26
Southern
California
7
Other
Areas
2
Number of counties and percentage of counties with suitable data2
Current Standards - PM2 5 mass indicator
Annual
15 ug/m3
24-Hour
35 ug/m3
# counties
% counties
21
11%
0
0%
1
2%
3
6%
0
0%
0
0%
11
42%
5
71%
1
50%
Alternative Standards - Calculated PM2 5 Light Extinction Indicator
Form
90th
90th
90th
90th
Level
(Mm'1)
25
26
27
28
# counties
% counties
# counties
% counties
# counties
% counties
# counties
% counties
122
65%
95
51%
72
39%
44
24%
26
79%
22
67%
16
48%
6
18%
28
60%
14
30%
5
11%
2
4%
49
92%
48
91%
42
79%
27
51%
4
40%
1
10%
0
0%
0
0%
0
0%
0
0%
0
0%
0
0%
8
31%
4
15%
o
6
12%
3
12%
6
86%
5
71%
5
71%
5
71%
1
50%
1
50%
1
50%
1
50%
3 3-year average of annual 90th percentile 24-hour average PM25 light extinction.
4 Design values for comparison with the level of the standard were calculated based on the approach specified in Table G-2 using 2007-2009 monitoring data.
Actual future outcomes may differ from these estimates due to changes in instrumentation, siting, and/or the specific procedure for calculating the indicator.
H-2
-------
This page left intentionally blank.
-------
United States Office of Air Quality Planning and Standards Publication No. EP A-452/R-11-003
Environmental Protection Health and Environmental Impacts Division April 2011
Agency Research Triangle Park, NC
------- |