&EPA
United States
Environmental Protection
Agency
Review of the National Ambient Air Quality
Standards for Particulate Matter:
Policy Assessment of Scientific
and Technical Information
OAQPS Staff Paper
-------
-------
EPA-452/R-05-005
June 2005
Review of the National Ambient Air Quality
Standards for Particulate Matter:
Policy Assessment of Scientific
and Technical Information
OAQPS Staff Paper
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
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 Staff Paper contains the conclusions and recommendations of the staff of the OAQPS
and does not necessarily represent those of the EPA. Mention of trade names or commercial
products is not intended to constitute endorsement or recommendation for use.
-------
ACKNOWLEDGMENTS
This Staff Paper is the product of the Office of Air Quality Planning and Standards
(OAQPS). For the section on PM-related health effects and primary standards, the principal
authors include Dr. Mary Ross, who also led the PM NAAQS review team, Mr. Harvey
Richmond, and Dr. Karen Martin, who also managed the project. For the section on PM-related
welfare effects and secondary standards, the principal authors include Mr. Rich Damberg, Ms.
Amy Vasu, Mr. Mark Schmidt, Ms. Vicki Sandiford, and Dr. Karen Martin. The principal
authors of the chapter on the characterization of ambient PM include Mr. Mark Schmidt, Mr.
John Langstaff, and Mr. Scott Mathias. The authors would also like to acknowledge Mr.
Venkatesh Rao, Mr. David Mintz, Mr. Lance McCluney, Mr. Neil Frank, Mr. Tim Hanley, and
Mr. Lewis Weinstock for providing input and analyses on specific topics. Staff from other EPA
offices, including the Office of Research and Development, the Office of General Counsel, and
the Office of Transportation and Air Quality, also provided valuable comments. Special
acknowledgment is given to Mr. John Bachmann for his leadership in past PM NAAQS reviews
and for his expert advice and valuable input in this review.
Earlier drafts of this document were formally reviewed by the Clean Air Scientific
Advisory Committee (CASAC) and made available for public comment. This document has
been informed by the expert advice and comments received from CASAC, as well as by public
comments submitted by a number of independent scientists, officials from State and local air
pollution organizations, environmental groups, and industrial groups and companies.
-------
-------
Table of Contents
List of Tables viii
List of Figures xi
Abbreviations and Acronyms xvi
1 INTRODUCTION 1-1
1.1 PURPOSE 1-1
1.2 BACKGROUND 1-2
1.2.1 Legislative Requirements 1-2
1.2.2 History of PM NAAQS Reviews 1-4
1.2.3 Litigation Related to 1997 PM Standards 1-5
1.2.4 Current PM NAAQS Review 1-7
1.3 GENERAL APPROACH AND ORGANIZATION OF DOCUMENT 1-8
REFERENCES 1-11
2. CHARACTERIZATION OF AMBIENT PM 2-1
2.1 INTRODUCTION 2-1
2.2 PROPERTIES OF AMBIENT PM 2-1
2.2.1 Particle Size Distributions 2-1
2.2.1.1 Modes 2-2
2.2.1.2 Sampler Cut Points 2-4
2.2.2 Sources and Formation Processes 2-6
2.2.3 Chemical Composition 2-8
2.2.4 Fate and Transport 2-9
2.2.5 Optical Properties of Particles 2-10
2.2.6 Other Radiative Properties of Particles 2-12
2.3 AMBIENT PM MEASUREMENT METHODS 2-13
2.3.1 Particle Mass Measurement Methods 2-13
2.3.2 Indirect Optical Methods 2-15
2.3.3 Size-Differentiated Particle Number Concentration Measurement
Methods 2-16
2.3.4 Chemical Composition Measurement Methods 2-16
2.3.5 Measurement Issues 2-17
2.4 PM CONCENTRATIONS, TRENDS, AND SPATIAL PATTERNS 2-18
2.4.1 PM25 2-20
2.4.2 PM10 2-30
2.4.3 PM10.25 2-33
2.4.4 Ultrafme Particles 2-40
2.4.5 Components of PM 2-40
2.4.6 Relationships Among PM2 5, PM10, and PM10_2 5 2-46
-------
2.5 PM TEMPORAL PATTERNS 2-49
2.5.1 PM25 and PM10.25 Patterns 2-49
2.5.2 Ultrafme Patterns 2-63
2.6 PM BACKGROUND LEVELS 2-63
2.7 RELATIONSHIP BETWEEN AMBIENT PM MEASUREMENTS AND
HUMAN EXPOSURE 2-67
2.7.1 Definitions 2-68
2.7.2 Centrally Monitored PM Concentration as a Surrogate for Particle
Exposure 2-69
2.8 RELATIONSHIP BETWEEN AMBIENT PM AND VISIBILITY 2-73
2.8.1 Particle Mass and Light Extinction 2-74
2.8.2 Other Measures of Visibility 2-76
2.8.3 Visibility at PM Background Conditions 2-76
REFERENCES 2-79
PM-RELATED HEALTH EFFECTS
AND PRIMARY PM NAAQS
3. POLICY-RELEVANT ASSESSMENT OF HEALTH EFFECTS EVIDENCE .. 3-1
3.1 INTRODUCTION 3-1
3.2 MECHANISMS 3-4
3.3 NATURE OF EFFECTS 3-10
3.3.1 Premature Mortality 3-11
3.3.1.1 Mortality and Short-term PM Exposure 3-12
3.3.1.2 Mortality and Long-term PM Exposure 3-17
3.3.2 Morbidity 3-19
3.3.2.1 Hospitalization and Medical Visits 3-19
3.3.2.2 Effects on the Respiratory System from Short-term Exposures
3-22
3.3.2.3 Effects on the Respiratory System from Long-term Exposures
3-23
3.3.2.4 Effects on the Cardiovascular System 3-23
3.3.3 Developmental effects 3-24
3.3.4 Summary 3-24
3.4 INTEGRATIVE ASSESSMENT OF HEALTH EVIDENCE 3-25
3.4.1 Strength of Associations 3-25
3.4.2 Robustness of Associations 3-27
3.4.3 Consistency 3-28
3.4.4 Temporality 3-30
3.4.5 Nature of concentration-response relationships 3-30
3.4.6 Natural Experiment Studies 3-31
3.4.7 Coherence and Plausibility 3-31
3.4.8 Summary 3-34
11
-------
3.5 PM-RELATED IMPACTS ON PUBLIC HEALTH 3-35
3.5.1 Potentially Susceptible and Vulnerable Subpopulations 3-35
3.5.2 Potential Public Health Impact 3-36
3.6 ISSUES RELATED TO QUANTITATIVE ASSESSMENT OF
EPIDEMIOLOGIC EVIDENCE 3-37
3.6.1 Air Quality Data in Epidemiologic Studies 3-38
3.6.2 Measurement Error and Exposure Error 3-39
3.6.3 Alternative Model Specifications 3-42
3.6.3.1 Time-series epidemiologic studies 3-42
3.6.3.2 Prospective cohort epidemiologic studies 3-43
3.6.4 Co-pollutant Confounding and Effect Modification 3-44
3.6.4.1 Co-pollutant Confounding 3-44
3.6.4.2 Effect Modification 3-46
3.6.5 Issues Related to Alternative Exposure Periods in Epidemiologic
Studies 3-47
3.6.5.1 Lag Structure in Short-term Exposure Studies 3-47
3.6.5.2 Seasonal Differences in Time-Series Epidemiologic Results . 3-51
3.6.5.3 Health Effects Related to Different Short-term Exposure
Time Periods 3-52
3.6.5.4 Exposure Periods Used in Prospective Cohort Studies 3-53
3.6.6 Concentration-Response Relationships and Potential Thresholds ... 3-55
3.7 SUMMARY AND CONCLUSIONS 3-57
REFERENCES 3-60
4. CHARACTERIZATION OF HEALTH RISKS 4-1
4.1 INTRODUCTION 4-1
4.1.1 Overview of Risk Assessment From Last Review 4-1
4.1.2 Development of Approach for Current Risk Assessment 4-2
4.2 SCOPE OF PM RISK ASSESSMENT 4-3
4.2.1 Selection of Health Endpoint Categories 4-4
4.2.2 Selection of Study Areas 4-5
4.3 COMPONENTS OF THE RISK MODEL 4-7
4.3.1 Air Quality Considerations 4-15
4.3.1.1 Estimating PM Background Levels 4-15
4.3.1.2 Simulating PM Levels That Just Meet Specified Standards . . 4-17
4.3.2 Concentration-Response Functions 4-19
4.3.2.1 Linear and Nonlinear Models 4-20
4.3.2.2 Single and Multi-City Models 4-23
4.3.2.3 Single and Multi-Pollutant Models 4-23
4.3.2.4 Single, Multiple, and Distributed Lag Functions 4-24
4.3.2.5 Alternative Short-Term Exposure Model Specifications 4-25
4.3.2.6 Long-term Exposure Models 4-25
4.3.3 Baseline Health Effects Incidence Rates and Population Estimates .. 4-26
4.3.4 Characterizing Uncertainty and Variability 4-26
in
-------
4.4 PM2 5 RISK ESTIMATES 4-35
4.4.1 Recent Air Quality 4-35
4.4.1.1 Base Case Risk Estimates Above Initial Cutpoint 4-35
4.4.1.2 Base Case Risk Estimates Above Various Cutpoints 4-42
4.4.1.3 Risk Estimates from Sensitivity Analyses 4-43
4.4.2 Just Meeting Current PM2 5 Standards 4-49
4.4.2.1 Base Case Risk Estimates Above Initial Cutpoint 4-49
4.4.2.2 Base Case Risk Estimates Above Various Cutpoints 4-52
4.4.2.3 Risk Estimates from Sensitivity Analyses 4-52
4.4.3 Just Meeting Alternative PM2 5 Standards 4-56
4.4.3.1 Base Case Risk Estimates 4-56
4.4.3.2 Risk Estimates from Sensitivity Analyses 4-62
4.4.4 Key Observations 4-66
4.5 PM10.2 5 RISK ESTIMATES 4-70
4.5.1. Recent Air Quality 4-70
4.5.1.1 Base Case Risk Estimates 4-70
4.5.1.2 Risk Estimates from Sensitivity Analyses 4-71
4.5.2 Just Meeting Alternative PM10_2 5 Standards 4-71
4.5.3 Key Observations 4-76
REFERENCES 4-78
5. STAFF CONCLUSIONS AND RECOMMENDATIONS ON PRIMARY PM
NAAQS 5-1
5.1 INTRODUCTION 5-1
5.2 APPROACH 5-1
5.3 FINE PARTICLE STANDARDS 5-4
5.3.1 Adequacy of Current PM25 Standards 5-4
5.3.1.1 Evidence-based Considerations 5-7
5.3.1.2 Risk-based Considerations 5-10
5.3.1.3 Summary 5-14
5.3.2 Indicators 5-15
5.3.3 Averaging Times 5-19
5.3.4 Alternative PM25 Standards to Address Health Effects Related to
Long-term Exposure 5-21
5.3.4.1 Evidence-based Considerations 5-22
5.3.4.2 Risk-based Considerations 5-23
5.3.4.3 Summary 5-28
5.3.5 Alternative PM25 Standards to Address Health Effects Related to
Short-term Exposure 5-29
5.3.5.1 Evidence-based Considerations 5-29
5.3.5.2 Risk-based Considerations 5-33
5.3.5.3 Summary 5-39
5.3.6 Alternative Forms for Annual and 24-Hour PM25 Standards 5-40
5.3.6.1 Form of Annual Standard 5-40
5.3.6.2 Form of 24-Hour Standard 5-42
5.3.7 Summary of Staff Recommendations on Primary PM2 5 NAAQS .... 5-45
iv
-------
5.4 THORACIC COARSE PARTICLE STANDARDS 5-47
5.4.1 Adequacy of Current PM10 Standards 5-47
5.4.2 Indicators 5-53
5.4.2.1 Evidence Related to Urban and Rural Thoracic Coarse
Particles 5-54
5.4.2.2 Related Requirements for PM10_25 Monitors and Monitoring
Network Design 5-58
5.4.3 Averaging Times 5-61
5.4.4 Alternative Standards to Address Health Effects Related to
Short-term Exposure 5-61
5.4.4.1 Evidence-based Considerations 5-62
5.4.4.2 Risk-based Considerations 5-69
5.4.4.3 Summary 5-69
5.4.5 Summary of Staff Recommendations on Primary PM10.2 5 NAAQS .. 5-70
5.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
RECOMMENDATIONS RELATED TO SETTING PRIMARY PM
STANDARDS 5-71
REFERENCES 5-75
PM-RELATED WELFARE EFFECTS
AND SECONDARY PM NAAQS
6. POLICY-RELEVANT ASSESSMENT OF PM-RELATED WELFARE
EFFECTS 6-1
6.1 INTRODUCTION 6-1
6.2 EFFECTS ON VISIBILITY 6-1
6.2.1 Overview of Visibility Impairment 6-2
6.2.2 Visibility Trends and Current Conditions in Class I and Non-Urban
Areas 6-3
6.2.3 Visibility Conditions in Urban Areas 6-5
6.2.3.1 ASOS Airport Visibility Monitoring Network 6-5
6.2.3.2 Correlation between Urban Visibility and PM25 Mass 6-6
6.2.4 Economic and Societal Value of Improving Visual Air Quality 6-13
6.2.5 Programs and Goals for Improving Visual Air Quality 6-18
6.2.5.1 Regional Protection 6-18
6.2.5.2 Local, State, and International Goals and Programs 6-19
6.2.6 Approaches to Evaluating Public Perceptions and Attitudes 6-20
6.2.6.1 Photographic Representations of Visual Air Quality 6-20
6.2.6.2 Survey Methods 6-22
6.2.7 Summary 6-24
6.3 EFFECTS ON VEGETATION AND ECOSYSTEMS 6-25
6.3.1 Major Ecosystem Stressors in PM 6-26
6.3.2 Direct Vegetation Effects of Particulate Nitrate and Sulfate
Deposition 6-28
-------
6.3.3 Ecosystem Effects of Chronic Inputs of Reactive Nitrogen and
Acidifying Compounds From PM Deposition and Other Sources . . . 6-29
6.3.3.1 Environmental Effects of Reactive Nitrogen (Nr) Deposition 6-29
6.3.3.2 Environmental Effects of PM-Related Acidic and Acidifying
Deposition 6-38
6.3.4 Characteristics and Location of Sensitive Ecosystems in the U.S. ... 6-44
6.3.5 Ecosystem Exposures to PM Deposition 6-46
6.3.6 Critical Loads 6-48
6.3.7 Summary and Conclusions 6-51
6.4 EFFECTS ON MATERIALS 6-53
6.4.1 Materials Damage Effects 6-54
6.4.2 Soiling Effects 6-55
6.4.3 Summary and Conclusions 6-56
6.5 EFFECTS ON CLIMATE CHANGE AND SOLAR RADIATION 6-56
6.5.1 Climate Change and Potential Human Health and Environmental
Impacts 6-57
6.5.2 Alterations in Solar UV-B Radiation and Potential Human Health
and Environmental Impacts 6-58
6.5.3 Summary and Conclusions 6-60
REFERENCES 6-62
7. STAFF CONCLUSIONS AND RECOMMENDATIONS ON
SECONDARY PM NAAQS 7-1
7.1 INTRODUCTION 7-1
7.2 APPROACH 7-1
7.3 STANDARDS TO ADDRESS VISIBILITY IMPAIRMENT 7-3
7.3.1 Adequacy of Current PM25 Standards 7-4
7.3.2 Indicators 7-6
7.3.3 Averaging Times 7-6
7.3.4 Alternative PM25 Standards to Address Visibility Impairment 7-7
7.3.5 Alternative Forms of a Short-term PM25 Standard 7-11
7.3.6 Summary of Staff Recommendations 7-12
7.4 STANDARDS TO ADDRESS OTHER PM-RELATED WELFARE
EFFECTS 7-13
7.4.1 Vegetation and Ecosystems 7-14
7.4.2 Materials Damage and Soiling 7-16
7.4.3 Climate Change and Solar Radiation 7-17
7.4.4 Summary of Staff Recommendations 7-17
7.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
RECOMMENDATIONS RELATED TO SETTING SECONDARY
PM STANDARDS 7-18
REFERENCES 7-20
VI
-------
CHAPTER APPENDICES:
APPENDIX 3 A: Mortality and Morbidity Effect Estimates and PM
Concentrations from U.S. and Canadian Studies for Short-term Exposures to
PM10, PM25, and PM10.25 3A-1
APPENDIX 3B: Mortality and Morbidity Effect Estimates and PM
Concentrations from U.S. and Canadian Studies for Long-Term Exposures to
PM10, PM25, and PM10.25 3B-1
APPENDIX 4A: Study-specific information for PM25 and PM10_25 studies 4A-1
APPENDIX 4B: Estimated annual mortality associated with short- and long-term
exposure to PM2 5 when alternative standards are met for Los Angeles, Philadelphia,
Pittsburgh, and St. Louis 4B-1
APPENDIX 5A: Reductions in PM2 5-related mortality risk for alternative
standards (99thpercentileform) 5A-1
APPENDIX 5B: Predicted percent of counties not likely to meet alternative
standards 5B-1
APPENDIX 6A: Images of Visual Air Quality in Selected Urban Areas in the U.S.
available at: http://www.epa. gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html
APPENDIX 7A:
Figure 7A-1 Estimated exceedances (%) of various PM2 5 levels for 12 pm - 4 pm
based on daily county maximum, 2001-2003 7A-1
Table 7A-1. Predicted percent of counties with monitors (and percent of population
in counties with monitors) not likely to meet alternative 4-hour (12 pm - 4 pm)
PM2 5 secondary standards 7A-2
ATTACHMENT A: Clean Air Scientific Advisory Committee Letter
vn
-------
List of Tables
Number Page
2-1 Particle Size Fraction Terminology Used in Staff Paper 2-7
2-2 Comparison of Ambient Fine Particles (Ultrafine plus Accumulation-Mode) and
Coarse Particles 2-11
2-3 Summary of PM25 FRM Data Analyses in 49 Metropolitan Areas, 2001-2003 2-29
2-4 Summary of Estimated PM10_25 Analyses in 21 Metropolitan Areas, 2001-2003 .... 2-39
2-5 PM25 Composition on High Mass Days in Select Urban Areas, 2003 2-57
2-6 IMPROVE Sites Selected for Estimates of Regional Background 2-66
2-7 Estimates of Long-Term Means, Daily Standard Deviations and 99th Percentiles
of PM25 Background Concentrations (|ig/m3) 2-67
4-1 Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the
PM2 5 Risk Assessment 4-8
4-2 Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in the
PM2 5 Risk Assessment 4-9
4-3 Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in the
PM10_2 5 Risk Assessment 4-9
4-4 Sensitivity Analyses 4-12
4-5 Summary of PM Ambient Air Quality Data for Risk Assessment Study Areas 4-16
4-6 Relevant Population Sizes for PM Risk Assessment Locations 4-27
4-7 Baseline Mortality Rates for 2001 for PM2 5 Risk Assessment Locations 4-28
4-8 Baseline Hospitalization Rates for PM Risk Assessment Locations 4-30
4-9 Estimated Annual Health Risks of Short-Term Exposure Mortality Associated with
Recent PM2 5 Concentrations Assuming Various Cutpoint Levels 4-44
4-10 Estimated Annual Health Risks of Long-Term Exposure Mortality Associated with
Recent PM2 5 Concentrations Assuming Various Cutpoint Levels 4-46
viii
-------
4-11 Air Quality Adjustments Required to Just Meet the Current Annual PM2 5 Standard
of 15 |ig/m3 Using the Maximum vs. the Average of Monitor-Specific Averages . . . 4-50
4-12 Estimated Annual Mortality Associated with Short-Term Exposure to PM2 5 When
the Current Annual Standard of 15 |ig/m3 and the Current Daily Standard of 65 |ig/m3
Are Just Met, Assuming Various Cutpoint Levels 4-54
4-13 Estimated Annual Mortality Associated with Long-Term Exposure to PM2 5 When the
Current Annual Standard of 15 |ig/m3 and the Current Daily Standard of 65 |ig/m3 Are
Just Met, Assuming Various Cutpoint Levels 4-55
4-14 Sensitivity Analysis Comparing the Use of Multi-City vs. Single-City Concentration-
Response Relationships on Estimates of Short-Term Exposure Mortality Associated with
Just Meeting the Current PM2 5 Standards 4-57
4-15 Alternative Sets of PM25 Standards Considered in the PM25 Risk Assessment 4-59
4-16 Estimated Design Values for Annual and 98th and 99th Percentile Daily PM25 Standards
Based on 2001-2003 Air Quality Data 4-59
4-17 Estimated Annual Mortality Associated with Short-Term Exposure to PM2 5 When
Alternative Standards Are Just Met, Assuming Various Cutpoint Levels for
Detroit, MI 4-60
4-18 Estimated Annual Mortality Associated with Long-Term Exposure to PM25 When
Alternative Standards Are Just Met, Assuming Various Cutpoint Levels for
Detroit, MI 4-63
4-19 Alternative PM10_2 5 Standards Considered in the PM10_2 5 Risk Assessment 4-74
4-20 Estimated Design Values for 98th and 99th Percentile Daily PM10.2 5 Standards Based
on 2001-2003 Air Quality Data 4-74
4-21 Estimated Annual Hospital Admissions for Ischemic Heart Disease Associated with
Short-Term PM10_2 5 Exposures When Alternative Standards Are Just Met, Assuming
Various Cutpoint Levels for Detroit, MI 4-75
5-1 (a) Estimated PM25-related annual total mortality associated with long-term exposure
when current PM25 standards are met 5-12
5-l(b) Estimated PM25-related annual mortality associated with short-term exposure
when current PM25 standards are met 5-13
IX
-------
5-2 Estimated PM10_2 5-related annual incidence of hospital admissions and cough in children
associated with short-term exposure with 2003 air quality 5-51
-------
List of Figures
Number Page
2-1 Distribution of coarse, accumulation, and nucleation-mode particles by three
characteristics 2-3
2-2 An idealized distribution of ambient PM showing fine and coarse particles and
the fractions collected by size-selective samplers 2-5
2-3 Regions used in PM Staff Paper in data analysis summaries 2-19
2-4 Distribution of annual mean PM2 5 and estimated annual mean PM10_2 5
concentrations by region, 2001-2003 2-21
2-5 Distribution of 98th percentile 24-hour average PM2 5 and estimated PM10_2 5
concentrations by region, 2001-2003 2-22
2-6 County-level maximum annual mean PM25 concentrations, 2001-2003 2-23
2-7 County-level maximum 98th percentile 24-hour average PM2 5 concentrations,
2001-2003 ' 2-24
2-8 Regional trends in annual average PM2 5 concentrations in the EPA FRM network,
1999-2003 2-26
2-9 Average annual mean trend in PM2 5 mass, ammonium sulfate, ammonium nitrate,
total carbonaceous mass, and crustal material at IMPROVE sites, 1993-2003 2-28
2-10 County-level maximum PM10 annual mean concentrations, 2001-2003 2-31
2-11 County-level maximum 24-hour PM10 "design value" concentrations, 2001-2003 .. 2-32
2-12 Estimated county-level maximum annual mean PM10_25 concentrations, 2001-
2003 ' 2-34
2-13 Estimated county-level maximum 98th percentile 24-hour average PM10_2 5
concentrations, 2001-2003 2-35
2-14 Average measured annual average PM10_25 concentration trend at IMPROVE
sites, 1993-2003 2-37
2-15 Urban versus rural estimated PM10_25 concentrations in select areas 2-38
2-16 Annual average composition of PM25 by region, 2003 2-42
xi
-------
2-17 Average PM10_2 5 PM25, and PM0A (ultrafme) chemical composition at the USC
EPA 'supersite' monitor in Los Angeles, CA, 10/2001 to 9/2002 2-43
2-18 Average PM10_2 5 composition for Los Angeles and two eastern urban-rural pairs . . . 2-45
2-19 Distribution of ratios of PM25 to PM10 by region, 2001-2003 2-47
2-20 Regional average correlations of 24-hour average PM by size fraction 2-48
2-21 Urban 24-hour average PM2 5 concentration distributions by region and month,
2001-2003 2-50
2-22 Urban 24-hour average estimated PM10_2 5 concentration distributions by region
and month, 2001-2003 2-51
2-23 Seasonal average composition of urban PM25 by region, 2003 2-52
2-24 Seasonal average composition of rural PM25 by region, 2003 2-53
2-25 Distribution of annual mean vs. 98th percentile 24-hour average PM2 5
concentrations, 2001-2003 2-55
2-26 Distribution of estimated annual mean vs. 98th percentile 24-hour average PM10_25
concentrations, 2001-2003 2-56
2-27 Hourly average PM2 5 and PM10_2 5 concentrations at a Greensboro, NC monitoring
site, 2001-2003 2-59
2-28 Seasonal hourly average PM2 5 and PM10_2 5 concentrations at a Greensboro, NC
monitoring site, 2001-2003 2-60
2-29 Hourly average PM_2 5 and PM10_2 5 concentrations at a Denver, CO monitoring
site, 2001-2003 . . . .' ' 2-61
2-30 Hourly PM2 5 and PM10_2 5 concentrations at an El Paso, TX monitoring site, April
26, 2002-April 27, 2002 2-62
2-31 Relationship between light extinction, deciviews, and visual range 2-77
3-1 Excess risk estimates for total nonaccidental, cardiovascular, and respiratory
mortality with short-term exposure to PM in single-pollutant models for U.S. and
Canadian studies 3-13
xn
-------
3-2 Excess risk estimates for hospital admissions and emergency department visits for
cardiovascular and respiratory diseases in single-pollutant models from U.S. and
Canadian studies 3-20
3-3 Associations between PM25 and total mortality from U.S. studies, plotted against
gaseous pollutant concentrations from the same locations 3-48
3-4 Natural logarithm of relative risk for total and cause-specific mortality with long-
term exposure to PM2 5 3-57
4-1 Maj or components of particulate matter health risk assessment.
4-11
4-2 Relationship between estimated log-linear concentration-response relationship and
hockeystick model with cutpoint C 4-22
4-3 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of total (non-accidental) mortality associated with short-term exposure to PM25
above background (and 95 percent confidence intervals): single-pollutant, single-city
models 4-37
4-4 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of total (non-accidental) mortality associated with short-term exposure to PM25
above background (and 95 percent confidence intervals): single-city versus multi-city
models 4-38
4-5 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of health effects associated with short-term exposure to PM25 above background
(and 95 percent confidence intervals): single-pollutant versus multi-pollutant models.4-39
4-6 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of total (non-accidental) mortality associated with long-term exposure to PM2 5
above 7.5 |ig/m3 (and 95 percent confidence intervals): single-pollutant models .... 4-40
4-7 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of total (non-accidental) mortality associated with long-term exposure to PM2 5
above 7.5 |ig/m3 (and 95 percent confidence intervals): multi-pollutant models .... 4-41
4-8 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of total (non-accidental) mortality associated with short-term exposure to PM25
above background (and 95 percent confidence intervals) for air quality just meeting the
current PM2 5 standards 4-51
Xlll
-------
4-9 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of total mortality associated with long-term exposure to PM2 5 above 7.5 |ig/m3
(and 95 percent confidence intervals) for air quality just meeting the current PM2 5 standards.
'. . . 4-53
4-10 Distribution of average daily PM25 concentrations in Detroit (2003 air quality data) (top
panel), estimated non-accidental mortality per day in Detroit associated with exposure to
daily PM2 5 concentrations (middle panel), and estimated non-accidental mortality in
Detroit associated with exposure to daily PM2 5 concentrations over the course of a year
(bottom panel) 4-68
4-11 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of hospital admissions associated with short-term exposure to PM10_25 (and 95
percent confidence intervals) 4-72
4-12 Estimated annual percent (top panel) and cases per 100,000 general population (bottom
panel) of respiratory symptoms associated with short-term exposure to PM10_25 (and 95
percent confidence intervals) 4-73
5-1 (a) Estimated percent reduction in PM2 5-related long-term mortality risk for alternative
standards (98thpercentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint of 7.5 |ig/m3) 5-25
5-l(b) Estimated percent reduction in PM2 5-related long-term mortality risk for alternative
standards (98thpercentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint of 10 |ig/m3) 5-26
5-l(c) Estimated percent reduction in PM2 5-related long-term mortality risk for alternative
standards (98th percentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint of 12 |ig/m3) 5-27
5-2(a) Estimated percent reduction in PM2 5-related short-term mortality risk for alternative
standards (98th percentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint equal to policy-relevant background) 5-34
5-2(b) Estimated percent reduction in PM2 5-related short-term mortality risk for alternative
standards (98th percentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint of 10 |ig/m3) 5-35
5-2(c) Estimated percent reduction in PM2 5-related short-term mortality risk for alternative
standards (98th percentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint of 15 |ig/m3) 5-36
xiv
-------
5-2(d) Estimated percent reduction in PM2 5-related short-term mortality risk for alternative
standards (98thpercentile form) relative to risk associated with meeting current
standards (based on assumed cutpoint of 20 |ig/m3) 5-37
6-1 PM2 5 concentration differences between eastern and western areas and between
rural and urban areas for 2003 6-7
6-2 Distribution of hourly and 24-hour average relative humidity at eastern and
western U.S. National Weather Service sites, 2003 6-8
6-3 Relationship between reconstructed light extinction (RE) and 24-hour average
PM25, 2003 6-10
6-4 Model slope for relationship between reconstructed light extinction (RE) and
hourly PM25 (increase in RE due to incremental increase in PM2 5), 2003 6-11
6-5 Relationship between reconstructed light extinction (RE) and 12 p.m. - 4 p.m.
average PM25, 2003 6-13
6-6 Illustration of the nitrogen cascade 6-29
7-1 Distributions of PM25 concentrations for 12 p.m. - 4 p.m. corresponding to visual
ranges of 25 km (panel a), 30 km (panel b), and 35 km (panel c) - by region 7-9
xv
-------
Abbreviations and Acronyms
AC Automated colorimetry
ACS American Cancer Society
AHSMOG Adventist Health and Smoke Study
AIRS Aerometric Information Retrieval System
ANC Acid neutralizing capacity
APHEA Air Pollution and Health, a European Approach
AQCD Air Quality Criteria Document
AQS Air Quality System
ARS Air Resource Specialists, Inc.
ASOS Automated Surface Observing System
BC Black carbon
BS British or black smoke
CAA Clean Air Act
CAMM Continuous Ambient Mass Monitor
CAP Concentrated ambient particles
CASAC Clean Air Scientific Advisory Committee
CASTNet Clean Air Status and Trends Network
CB Base cation
CBS A Core Based Statistical Area
CD Criteria Document
CDC Centers for Disease Control
CDPHE Colorado Department of Public Health and Environment
CFR Code of Federal Regulations
CL Critical loads
C:N Carbon-to-nitrogen ratio
CO Carbon monoxide
COH Coefficient of haze
COPD Chronic obstructive pulmonary disease
CPSC Consumer Product Safety Commission
C-R Concentration-response
CSA Combined Statistical Area
CSS Coastal sage scrub community
CV Contingent valuation
EC Elemental carbon
ECG Electrocardiogram
ED Emergency department
EEA Essential Ecological Attribute
EMAP Environmental Monitoring and Assessment Program
EPA Environmental Protection Agency
EPEC Ecological Processes and Effects Committee
ERP Episodic Response Project
FDMS Filter Dynamics Measurement System
FLM Federal Land Manager
xvi
-------
FRM Federal reference method
GAM Generalized additive models
GCVTC Grand Canyon Visibility Transport Commission
GLM Generalized linear models
HAPs Hazardous air pollutants
HEI Health Effects Institute
HF Heart failure
hosp. adm. Hospital admissions
1C Ion chromatography
IFS Integrated Forest Study
IHD Ischemic heart disease
IMPROVE Interagency Monitoring of Protected Visual Environments
LML Lowest measured level
LPC Laser particle counter
LRS Lower respiratory symptoms
mort. Mortality
NAAQS National ambient air quality standards
NADP National Atmospheric Deposition Program
NAPAP National Acid Precipitation Assessment Program
NCEA National Center for Environmental Assessment
NDDN National Dry Deposition Network
NEG/ECP New England Governors/Eastern Canadian Premiers
NMMAPS National Mortality and Morbidity Air Pollution Study
N2 Nonreactive, molecular nitrogen
NO2 Nitrogen dioxide
non-accid
mort Non-accidental mortality
Nr Reactive nitrogen
NSMPS Nano-scanning mobility particle sizer
NuCM Nutrient cycling model
NWS National Weather Service
O3 Ozone
OAQPS Office of Air Quality Planning and Standards
OAR Office of Air and Radiation
OC Organic carbon
ORD Office of Research and Development
OSHA Occupational Safety and Health Administration
PAHs Polynuclear aromatic hydrocarbons
pneum. Pneumonia
PTEAM EPA's Particle Total Exposure Assessment Methodology
PCBs Polychlorinated biphenyls
PCDD/F Polychlorinated dibenzo-p-dioxins/dibenzofurans
PM Particulate matter
PMio-2.5 Particles less than or equal to 10 jim in diameter and greater than 2.5 jim in
diameter
xvn
-------
PM25 Particles less than or equal to 2.5 jim in diameter
PM10 Particles less than or equal to 10 jim in diameter
PnET-BGC A forest net productivity model (PnET) linked to a soil model (BGC)
POPs Persistent organic pollutants
PRB Policy relevant background
REVEAL Regional Visibility Experimental Assessment in the Lower Fraser Valley
RR Relative risk
SAB Science Advisory Board
SMPS Standard scanning mobility particle sizer
SO2 Sulfur dioxide
SO4 Sulfate
SOCs Semivolatile organic compounds
STN PM2 5 Chemical Speciation Trends Network
SP Staff Paper
TEOM Tapered Element Oscillating Microbalance sensor
TEVIE/LTM Temporally Integrated Monitoring of Ecosystems/Long-Term Monitoring Project
TL Target load
TMO Thermal manganese oxidation method
TOR Thermal/optical reflectance method
TOT Thermal/optical transmission method
TSD Technical support document
TSP Total suspended particulates
|ig micrograms
l-ig/m3 micrograms per cubic meter
UNEP United Nations Environmental Program
UPM10_2 5 Thoracic coarse urban particulate matter
URS Upper respiratory symptoms
U.S. United States
UV Ultraviolet
UV-B Ultraviolet-B
Vd Deposition velocity
VOCs Volatile organic compounds
WMO World Meteorological Organization
XRF X-ray fluorescence
xvin
-------
1. INTRODUCTION
1.1 PURPOSE
This Staff Paper, prepared by staff in the U.S. Environmental Protection Agency's (EPA)
Office of Air Quality Planning and Standards (OAQPS), evaluates the policy implications of the
key studies and scientific information contained in the document, Air Quality Criteria for
Particulate Matter (EPA, 2004; henceforth referred to as the Criteria Document (CD) and cited
as CD), prepared by EPA's National Center for Environmental Assessment (NCEA). This Staff
Paper also presents and interprets results from staff analyses (e.g., air quality analyses, human
health risk assessments, and visibility analyses) that staff believes should be considered in EPA's
current review of the national ambient air quality standards (NAAQS) for particulate matter
(PM). Finally, this Staff Paper presents staff conclusions and recommendations as to potential
revisions of the primary (health-based) and secondary (welfare-based) PM NAAQS, based on
consideration of the available scientific information and analyses and related limitations and
uncertainties.
The policy assessment presented in this document is intended to help "bridge the gap"
between the scientific review contained in the CD and the judgments required of the EPA
Administrator in determining whether, and if so, how, it is appropriate to revise the NAAQS for
PM. This assessment focuses on the basic elements of PM air quality standards: indicators,
averaging times, forms1, and levels. These elements, which serve to define each standard within
the suite of PM NAAQS, must be considered collectively in evaluating the health and welfare
protection afforded by the standards.
While this Staff Paper should be of use to all parties interested in the PM NAAQS
review, it is written for those decision makers, scientists, and staff who have some familiarity
with the technical discussions contained in the CD.
1.2 BACKGROUND
1.2.1 Legislative Requirements
Two sections of the Clean Air Act (Act) govern the establishment and revision of the
NAAQS. Section 108 (42 U.S.C. 7408) directs the Administrator to identify and list "air
pollutants" that "in his judgment, may reasonably be anticipated to endanger public health and
welfare" and whose "presence ... in the ambient air results from numerous or diverse mobile or
stationary sources" and to issue air quality criteria for those that are listed. Air quality criteria
are intended to "accurately reflect the latest scientific knowledge useful in indicating the kind
1 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
-------
and extent of identifiable effects on public health or welfare which may be expected from the
presence of [a] pollutant in ambient air ... ."
Section 109 (42 U.S.C. 7409) directs the Administrator to propose and promulgate
"primary" and "secondary" NAAQS for pollutants listed under section 108. 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."2 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."3
In setting standards that are "requisite" to protect public health and welfare, 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).
The requirement that primary standards include 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. 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 U.S. 1034 (1982). Both kinds of uncertainties are components of the risk associated
with pollution at levels below those at which human health effects can be said to occur with
reasonable scientific certainty. Thus, in selecting primary standards that include an adequate
margin of safety, the Administrator is seeking not only to prevent pollution levels that have been
demonstrated to be harmful but also to prevent lower pollutant levels that may pose an
unacceptable risk of harm, even if the risk is not precisely identified as to nature or degree.
In selecting a margin of safety, the EPA considers such factors as the nature and severity
of the health effects involved, the size of the sensitive population(s) at risk, and the kind and
degree of the uncertainties that must be addressed. The selection of any particular approach to
2 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)].
3 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
-------
providing an adequate margin of safety is a policy choice left specifically to the Administrator's
judgment. Lead Industries Association v. EPA, supra. 647 F.2d at 1161-62.
Section 109(d)(l) of the Act 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 (CAS AC) of EPA's Science Advisory Board.
1.2.2 History of PM NAAQS Reviews
Particulate 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. Particles originate from a variety of anthropogenic stationary and mobile sources as well
as natural sources. Particles may be emitted directly or formed in the atmosphere by
transformations of gaseous emissions such as sulfur oxides, nitrogen oxides, and volatile organic
compounds. The chemical and physical properties of PM vary greatly with time, region,
meteorology, and source category, thus complicating the assessment of health and welfare
effects.
EPA first established national ambient air quality standards for PM in 1971, based on the
original criteria document (DHEW, 1969). 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 (44 FR 56731), 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 particles from
TSP to PM10, the latter including particles with a mean aerodynamic diameter4 less than or equal
4 The more precise term is 50 percent cut point or 50 percent diameter (D50). This is the aerodynamic
particle diameter for which the efficiency of particle collection is 50 percent. Larger particles are not excluded
altogether, but are collected with substantially decreasing efficiency and smaller particles are collected with
increasing (up to 100 percent) efficiency.
1-3
-------
to 10 jim, which delineates that subset of inhalable particles small enough to penetrate to the
thoracic region (including the tracheobronchial and alveolar regions) of the respiratory tract
(referred to as thoracic particles). EPA also revised the level and form of the primary standards
by: (1) replacing the 24-hour TSP standard with a 24-hour PM10 standard of 150 |ig/m3 with no
more than one expected exceedance per year; and (2) replacing the annual TSP standard with a
PM10 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 PM10 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. Administrator, 902
F. 2d 962 (D.C. Cir. 1990), cert, denied. 498 U.S. 1082 (1991).
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). In that decision, EPA revised the PM NAAQS in several respects. While it was
determined that the PM NAAQS should continue to focus on particles less than or equal to 10
jim in diameter, it was also determined that the fine and coarse fractions of PM10 should be
considered separately. New standards were added, using PM2 5, referring to particles with a
nominal mean aerodynamic diameter less than or equal to 2.5 jim, as the indicator for fine
particles, with PM10 standards retained for the purpose of regulating the coarse fraction of PM10
(referred to as thoracic coarse particles or coarse-fraction particles; generally including particles
with a nominal mean aerodynamic diameter greater than 2.5 jim and less than or equal to 10 jim,
or PM10_2 5). EPA established two new PM25 standards: an annual standard of 15 |ig/m3, based
on the 3-year average of annual arithmetic mean PM25 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. A new reference method for the measurement of PM25 in the ambient air was also
established, as were rules for determining attainment of the new standards. To continue to
address thoracic coarse particles, the annual PM10 standard was retained, while the 24-hour PM10
standard was revised to be based on the 99th percentile of 24-hour PM10 concentrations at each
monitor in an area. EPA revised the secondary standards by making them identical in all
respects to the primary standards.
1.2.3 Litigation Related to the 1997 PM Standards
Following promulgation of the revised PM NAAQS, 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
1-4
-------
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 PM10 standards, concluding in
part that PM10 is a "poorly matched indicator for coarse paniculate pollution" because it includes
fine particles. Id. at 1053-55. Pursuant to the court's decision, EPA removed the vacated 1997
PM10 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 PM10 standards to the 1997 PM10 standards (65 FR 80776, December 22,
2000). The pre-existing 1987 PM10 standards remained in place. Id. at 80777. In the current
review, EPA is addressing thoracic coarse particles in part by considering standards based on a
more narrowly defined indicator.
More generally, the Panel held (with one dissenting opinion) that EPA's approach to
establishing the level of the standards in 1997, both for PM and for ozone 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, the Panel also
reaffirmed prior rulings holding that in setting NAAQS EPA is "not permitted to consider the
cost of implementing those standards." Id. at 1040-41.
Both sides filed cross appeals on these issues to the United States Supreme Court, and
the Court 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 guided 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 PM25 standards were reasonably supported by the administrative record
and were not "arbitrary and capricious." American Trucking Associations v. EPA, 283 F. 3d 355,
369-72 (D.C. Cir. 2002).
1-5
-------
1.2.4 Current PM NAAQS Review
In October 1997, EPA published its plans for the current periodic review of the PM
criteria and NAAQS (62 FR 55201, October 23, 1997), including the 1997 PM2 5 standards and
the 1987 PM10 standards. As part of the process of preparing the PM CD, NCEA hosted a peer
review workshop in April 1999 on drafts of key CD chapters. The first external review draft CD
was reviewed by CAS AC and the public at a meeting held in December 1999. Based on CAS AC
and public comment, NCEA revised the draft CD and released a second draft in March 2001 for
review by CAS AC and the public at a meeting held in July 2001. A preliminary draft Staff
Paper (EPA, 2001) was released in June 2001 for public comment and for consultation with
CASAC at the same public meeting. Taking into account CASAC and public comments, a third
draft CD was released in May 2002 for review at a meeting held in July 2002.
Shortly after EPA released the third draft CD, the Health Effects Institute (HEI)5
announced that researchers at Johns Hopkins University had discovered problems with
applications of statistical software used in a number of important epidemiological studies that
had been discussed in that draft CD. In response to this significant issue, EPA took steps in
consultation with CASAC to encourage researchers to reanalyze affected studies and to submit
them expeditiously for peer review by a special expert panel convened at EPA's request by HEI.
EPA subsequently incorporated the results of this reanalysis and peer-review process into a
fourth draft CD, which was released in June 2003 and reviewed by CASAC and the public at a
meeting held in August 2003.
The first draft Staff Paper, based on the fourth draft CD, was released at the end of
August 2003, and was reviewed by CASAC and the public at a meeting held in November 2003.
During that meeting, EPA also consulted with CASAC on a new framework for the final chapter
(integrative synthesis) of the CD and on ongoing revisions to other CD chapters to address
previous CASAC comments. EPA held additional consultations with CASAC at public meetings
held in February, July, and September 2004, leading to publication of the final CD in October
2004. The second draft Staff Paper, based on the final CD, was released at the end of January
2005, and was reviewed by CASAC and the public at a meeting held in April 2005. The
CASAC's advice and recommendations to the Administrator, based on their review of the
second draft Staff Paper, were further discussed during a public teleconference held in May 2005
and are provided in a letter to the Administrator (Henderson, 2005), which is reproduced in
Attachment 2. This final Staff Paper takes into account the advice and recommendations from
CASAC and public comments received on the earlier drafts of this document. Any subsequent
5 HEI is an independent research institute, jointly sponsored by EPA and a group of U.S.
manufacturers/marketers of motor vehicle and engines, that conducts health effects research on major air pollutants
related to motor vehicle emissions.
1-6
-------
advice and recommendations received from CAS AC related to this final Staff Paper will also be
considered by the Administrator.
The schedule for completion of this review is now governed by a consent decree
resolving a lawsuit filed in March 2003 by a group of plaintiffs representing national
environmental organizations. The lawsuit alleged that EPA had failed to perform its mandatory
duty, under section 109(d)(l), of completing the current review within the period provided by
statute. American Lung Association v. Whitman (No. 1:03CV00778, D.D.C. 2003). An initial
consent decree was entered by the court in July 2003 after an opportunity for public comment.
The consent decree, as modified by the court, provides that EPA will sign for publication notices
of proposed and final rulemaking concerning its review of the PM NAAQS no later than
December 20, 2005 and September 27, 2006, respectively.
1.3 GENERAL APPROACH AND ORGANIZATION OF DOCUMENT
This policy assessment is based on staff evaluation of the policy implications of the
scientific evidence contained in the CD and the results of quantitative analyses based on that
evidence, which taken together help inform staff conclusions and recommendations on the
elements of the PM standards under review. While the CD focuses on new scientific information
available since the last criteria review, it appropriately integrates that information with scientific
criteria from previous reviews. The quantitative analyses presented herein (and described in
more detail in a number of technical support documents) are based on the most recently available
air quality information, so as to provide current characterizations of PM air quality patterns,
estimated human health risks related to exposure to ambient PM, and PM-related visibility
impairment.
Partly as a consequence of EPA's decision in the last review to consider fine particles and
thoracic coarse particles separately, much new information is now available on PM air quality
and human health effects directly in terms of PM25 and, to a much more limited degree, PM10_25.
This information adds to the body of evidence on PM10 that has continued to grow since the
introduction of that indicator in the first PM NAAQS review. Since the purpose of this review is
to evaluate the adequacy of the current standards that separately address fine and thoracic coarse
particles, staff has focused this policy assessment and associated quantitative analyses primarily
on the evidence related directly to PM25 and PM10_25. In so doing, staff has considered PM10-
related evidence primarily to help inform our understanding of key issues and to help interpret
and provide context for the more limited PM2 5 and PM10_2 5 evidence.
Following this introductory chapter, this Staff Paper is organized into three main parts:
the characterization of ambient PM; PM-related health effects and primary PM NAAQS; and
PM-related welfare effects and secondary PM NAAQS. The characterization of ambient PM is
presented in Chapter 2, which focuses on properties of ambient PM, measurement methods,
spatial and temporal patterns in ambient PM concentrations, PM background levels, and ambient
1-7
-------
PM relationships with human exposure and with visibility impairment. Thus, Chapter 2 provides
information relevant to both the health and welfare assessments in the other two main parts of
this document.
Chapters 3 through 5 comprise the second main part of this Staff Paper dealing with
human health and primary standards. Chapter 3 presents a policy-relevant assessment of PM
health effects evidence, including an overview of the evidence, key human health-related
conclusions from the CD, and an examination of issues related to the quantitative assessment of
the epidemiologic health evidence. Chapter 4 presents a quantitative assessment of PM-related
health risks, including risk estimates for current air quality levels as well as those associated with
just meeting the current NAAQS and various alternative standards that might be considered in
this review. Chapter 5 presents the staff review of the current primary standards for fine and
thoracic coarse particles. This chapter begins with a discussion of the broader approach used by
staff in this review of the primary PM NAAQS than in the last review, generally reflecting both
evidence-based and quantitative risk-based considerations. This review includes consideration
of the adequacy of the current standards, conclusions as to alternative indicators, averaging
times, levels and forms, and staff recommendations on ranges of alternative primary standards
for consideration by the Administrator.
Chapters 6 and 7 comprise the third main part of this Staff Paper dealing with welfare
effects and secondary standards. Chapter 6 presents a policy-relevant assessment of PM welfare
effects evidence, including evidence related to visibility impairment as well as to effects on
vegetation and ecosystems, climate change processes, and man-made materials. This chapter's
emphasis is on visibility impairment, reflecting the availability of a significant amount of policy-
relevant information and staff analyses which serve as the basis for staff consideration of a
secondary standard specifically for protection of visual air quality. Chapter 7 presents the staff
review of the current secondary standards, beginning with a discussion of the approach used by
staff in this review of the secondary PM NAAQS. This review includes consideration of the
adequacy of the current standards, conclusions as to alternative indicators, averaging times,
levels and forms, and staff recommendations on ranges of alternative secondary standards for
consideration by the Administrator.
-------
REFERENCES
Environmental Protection Agency. (2001) Review of the National Ambient Air Quality Standards for Paniculate
Matter: Policy Assessment of Scientific and Technical Information - Preliminary Draft OAQPS Staff
Paper. June.
Environmental Protection Agency. (2003) Review of the National Ambient Air Quality Standards for Paniculate
Matter: Policy Assessment of Scientific and Technical Information - First Draft OAQPS Staff Paper.
August.
Environmental Protection Agency. (2004) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
Office of Research and Development; report no. EPA/600/P-99/002a,bF. October.
Henderson, R. (2005) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee, to
Administrator Stephen L. Johnson. Re: Clean Air Scientific Advisory Committee (CASAC) Paniculate
Matter (PM) Review Panel's Peer Review of the Agency's Review of the National Ambient Air Quality
Standards for P'articulate Matter: Policy Assessment of Scientific and Technical Information (Second Draft
PM Staff Paper, January 2005); and Particulate Matter Health Risk Assessment for Selected Urban Areas:
Second Draft Report (Second Draft PM Risk Assessment, January 2005). June 6, 2005.
U.S. Department of Health, Education and Welfare (DEHW). (1969) Air Quality Criteria for Particulate Matter.
U.S. Government Printing Office, Washington DC, AP-49.
1-9
-------
-------
2. CHARACTERIZATION OF AMBIENT PM
2.1 INTRODUCTION
This chapter generally characterizes various classes of ambient PM in terms of physical
and chemical properties, measurement methods, recent concentrations and trends, and
relationships with human exposure and visibility impairment. This information is useful for
interpreting the available health and welfare effects information, and for making
recommendations on appropriate indicators for primary and secondary PM standards. The
information presented in this chapter was drawn from the CD and additional analyses of data
from various PM monitoring networks.
Section 2.2 presents information on the basic physical and chemical properties of classes
of PM. Section 2.3 presents information on the methods used to measure ambient PM and some
important considerations in the design of these methods. Section 2.4 presents data on PM
concentrations, trends, and spatial patterns in the U.S. Section 2.5 provides information on the
temporal variability of PM. Much of the information in Sections 2.4 and 2.5 is derived from
analyses of data collected by the nationwide networks of PM25 and PM10 monitors through 2003.
Section 2.6 defines and discusses background levels of ambient PM. Section 2.7 addresses the
relationships between ambient PM levels and human exposure to PM. Section 2.8 addresses the
relationship between ambient PM2 5 levels and visibility impairment.
2.2 PROPERTIES OF AMBIENT PM
PM represents a broad class of chemically and physically diverse substances that exist as
discrete particles in the condensed (liquid or solid) phase. Particles can be characterized by size,
formation mechanism, origin, chemical composition, and atmospheric behavior. This section
generally focuses on size since classes of particles have historically been characterized largely in
that manner. Fine particles and coarse particles, which are defined in Section 2.2.1.1, are
relatively distinct entities with fundamentally different sources and formation processes,
chemical composition, atmospheric residence times and behaviors, transport distances, and
optical and radiative properties. The CD concludes that these differences justify consideration of
fine and coarse particles as separate subclasses of PM pollution (CD, pp. 2-111 and 9-21).
2.2.1 Particle Size Distributions
Particle properties and their associated health and welfare effects differ by particle size.
The diameters of atmospheric particles span 5 orders of magnitude, ranging from
2-1
-------
0.001 micrometers to 100 micrometers (nm).1 The size and associated composition of particles
determine their behavior in the respiratory system, including how far the particles are able to
penetrate, where they deposit, and how effective the body's clearance mechanisms are in
removing them. Furthermore, particle size is one of the most important parameters in
determining the residence time and spatial distribution of particles in ambient air, key
considerations in assessing exposure. Particle size is also a major determinant of visibility
impairment, a welfare effect linked to ambient particles. Particle surface area, number, chemical
composition, and water solubility all vary with particle size, and are also influenced by the
formation processes and emissions sources.
Common conventions for classifying particles by size include: (1) modes, based on
observed particle size distributions and formation mechanisms; and (2) "cut points," based on the
inlet characteristics of specific PM sampling devices. The terminology used in this Staff Paper
for describing these classifications is summarized in Table 2-1 and discussed in the following
subsections.
2.2.1.1 Modes
Based on extensive examinations of particle size distributions in several U.S. locations in
the 1970's, Whitby (1978) found that particles display a consistent multi-modal distribution over
several physical metrics, such as mass or volume (CD, p. 2-7). These modes are apparent in
Figure 2-1, which shows average ambient distributions of particle number, surface area, and
volume by particle size.2 Panel (a) illustrates that by far, the largest number of ambient particles
in a typical distribution are very small, below 0.1 jim in diameter, while panel (c) indicates most
of the particle volume, and therefore most of the mass, is found in particles with diameters larger
than 0.1 |im.3 Most of the surface area (panel b) is between 0.1 and 1.0 jim. The surface area
distribution in panel (b) peaks around 0.2 |im. Distributions may vary across locations,
conditions, and time due to differences in sources, atmospheric conditions, topography, and the
age of the aerosol.
1 In this Staff Paper, particle size or diameter refers to a normalized measure called aerodynamic diameter
unless otherwise noted. Most ambient particles are irregularly shaped rather than spherical. The aerodynamic
diameter of any irregular shaped particle is defined as the diameter of a spherical particle with a material density of 1
g/cm3 and the same settling velocity as the irregular shaped particle. Particles with the same physical size and shape
but different densities will have different aerodynamic diameters (CD, p. 2-4).
2 Particle size distributions, such as those in Figure 2-1, are often expressed in terms of the logarithm of the
particle diameter (Dp) on the X-axis and the measured concentration difference per logarithmic increment in particle
diameter on the Y-axis. When the Y-axis concentration difference is plotted on a linear scale, the number of
particles, the particle surface area, and the particle volume (per cm3 air) having diameters in the size range from log
Dp to log(Dp + ADp) are proportional to the area under that part of the size distribution curve.
3 Mass is proportional to volume times density.
2-2
-------
Tt
O
15-
1_ -
o E
J2 O
Ez
3 •*"
D)
O
03
o
n
E
o
II
D)
O
10-
5-
600-
< E
-------
As illustrated in panel (c) of Figure 2-1, volume distributions typically measured in
ambient air in the U.S. are found to be bimodal, with overlapping tails, and an intermodal
minimum between 1 and 3 jim (CD, p. 2-25). The distribution of particles that are mostly larger
than this minimum make up the coarse mode and are called "coarse particles," and the
distribution of particles that are mostly smaller than the minimum are called "fine particles."
Fine particles can be subcategorized into smaller modes: "nucleation mode," "Aitken mode,"
and "accumulation mode." Together, nucleation-mode and Aitken-mode particles make up
"ultrafine particles."4 Ultrafine particles are apparent as the largest peak in the number
distribution in panel (a), and are also visible in the surface area distribution in panel (b).
Nucleation-mode and Aitken-mode particles have relatively low mass and grow rapidly into
accumulation-mode particles, so they are not commonly observed as a separate mode in volume
or mass distributions. The accumulation mode is apparent as the leftmost peak in the volume
distribution in panel (c) and the largest peak in the surface area distribution in panel (b).
2.2.1.2 Sampler Cut Points
Another set of particle size classifications is derived from the characteristics of ambient
particle samplers. Particle samplers typically use size-selective inlets that are defined by their 50
percent cut point, which is the particle aerodynamic diameter at which 50 percent of particles of
a specified diameter are captured by the inlet, and their penetration efficiency as a function of
particle size. The usual notation for these classifications is "PMX", where x refers to
measurements with a 50 percent cut point of x jim aerodynamic diameter. Because of the
overlap in the size distributions of fine and coarse-mode ambient particles, and the fact that inlets
do not have perfectly sharp cut points, no single sampler can completely separate them. Given a
specific size cut, the smaller the particles the greater the percentage of particles that are captured.
The objective of size-selective sampling is usually to measure particle size fractions that provide
a relationship to human health impacts, visibility impairment, or emissions sources.
Since 1987, the EPA has defined indicators of PM for NAAQS using cut points of
interest. Figure 2-2 presents an idealized distribution of ambient PM showing the fractions
collected by size-selective samplers. Prior to 1987, the indicator for the PM NAAQS was total
suspended particulate matter (TSP), and was defined by the design of the High Volume Sampler
(Hi Vol).5 As illustrated in Figure 2-2, TSP typically includes particles with diameters less than
about 40 |im, but the upper size cut varies substantially with placement and wind velocity .
When EPA established new PM standards in 1987, the selection of PM10 as an indicator was
4 Whitby (1978) did not identify multiple ultrafine particle modes between 0.01 and 0.1 um, and therefore
separate nucleation and Aitken modes are not illustrated in Figure 2-1. See CD Figure 2-6 for a depiction of all
particle modes.
5 40 CFR Part 50, Appendix B, Reference Method for the Determination of Suspended Paniculate Matter
in the Atmosphere (High-Volume Method).
2-4
-------
70
60
50
0
Accumulation Mode
Coarse Mode
0.1 0.2
0.5 1
Particle Diameter, Dp (pm)
Total Suspended Particles (TSP) —
PM-
'10
PM
2.5
PM
TSP
HiVol
WRAC
100
10-2.5
Figure 2-2. An idealized distribution of ambient PM showing fine and coarse particles and
the fractions collected by size-selective samplers. (WRAC is the Wide Range Aerosol
Classifier which collects the entire coarse mode).
Source: Adapted from Wilson and Suh (1997) and Whitby (1978); CD page 2-18
2-5
-------
intended to focus regulatory attention on particles small enough to be inhaled and to penetrate
into the thoracic region of the human respiratory tract. In 1997, EPA established standards for
fine particles measured as PM2 5 (i.e., the fine fraction of PM10). The dashed lines in Figure 2-2
illustrate the distribution of particles captured by the PM10 Federal Reference Method (FRM)
sampler6, including all fine and some coarse particles, and the distribution captured by the PM2 5
FRM sampler7, including generally all fine particles and potentially capturing a small subset of
coarse particles.
The EPA is now considering establishing standards for another PM indicator identified in
Table 2-1 as PM10_25, which represents the subset of coarse particles small enough to be inhaled
and to penetrate into the thoracic region of the respiratory tract (i.e., the coarse fraction of PM10,
or thoracic coarse particles). Section 2.3 discusses measurement methods for this indicator.
2.2.2 Sources and Formation Processes
In most locations, a variety of activities contribute to ambient PM concentrations. Fine
and coarse particles generally have distinct sources and formation mechanisms, although there is
some overlap (CD, p. 3-60). Coarse particles are generally primary particles, meaning they are
emitted from their source directly as particles. Most coarse particles result from mechanical
disruption of large particles by crushing or grinding, from evaporation of sprays, or from dust
resuspension. Specific sources include industrial process emissions, fugitive emissions from
storage piles, traffic related emissions including tire and paving materials and grinding and
resuspension of crustal, biological, industrial, and combustion materials that have settled on or
near roadways, construction and demolition activities, agriculture, mining and mineral
processing, sea spray, and wind-blown dust and biological materials. The amount of energy
required to break down primary particles into smaller particles normally limits coarse particle
sizes to greater than 1.0 jam diameter (EPA 1996a, p. 13-7). Some combustion-generated
particles, such as fly ash, are also found as coarse particles.
By contrast, a significant amount of fine particles are produced through combustion
processes and atmospheric transformation processes of precursor gases. Common directly
emitted fine particles include unburned carbon particles from combustion, and nucleation-mode
particles emitted as combustion-related vapors that condense within seconds of being exhausted
to ambient air. Fossil-fuel combustion sources include motor vehicles and off-highway
equipment, power generation facilities, industrial facilities, residential wood burning,
agricultural burning, and forest fires.
6 40 CFR Part 50, Appendix J, Reference Method for the Determination of Paniculate Matter as PM10 in
the Atmosphere.
7 40 CFR Part 50, Appendix L, Reference Method for the Determination of Fine Paniculate Matter as PM2 5
in the Atmosphere.
2-6
-------
Table 2-1. Particle Size Fraction Terminology Used in Staff Paper
Term
Description
Size Distribution Modes
Coarse Particles
Thoracic Coarse Particles
Fine Particles
Accumulation-Mode Particles
Ultrafine Particles
Aitken-Mode Particles
Nucleation-Mode Particles
The distribution of particles that are mostly larger than the intermodal
minimum in volume or mass distributions; also referred to as coarse-mode
particles. This intermodal minimum generally occurs between 1 and 3 um.
A subset of coarse particles that includes particles that can be inhaled and
penetrate to the thoracic region (i.e., the tracheobronchial and the
gas-exchange regions) of the lung. This subset includes the smaller coarse
particles, ranging in size up to those with a nominal aerodynamic diameter less
than or equal to 10 microns.
The distribution of particles that are mostly smaller than the intermodal
minimum in volume or mass distributions; this minimum generally occurs
between 1 and 3 um. This includes particles in the nucleation, Aitken, and
accumulation modes.
A subset of fine particles with diameters above about 0.1 um. Ultrafine
particles grow by coagulation or condensation and "accumulate" in this size
range.
A subset of fine particles with diameters below about 0.1 um, encompassing
the Aitken and nucleation modes.
A subset of ultrafine particles with diameters between about 0.01 and 0.1 um.
Freshly formed particles with diameters below about 0.01 um.
Sampling Measurements
Total Suspended Particles (TSP)
PM,,
PM25
PM.o.,5
Particles measured by a high volume sampler as described in 40 CFR Part 50,
Appendix B. This sampler has a cut point of aerodynamic diameters that
varies between 25 and 40 um depending on wind speed and direction.
Particles measured by a sampler that contains a size fractionator (classifier)
designed with an effective cut point (50% collection efficiency) of 10 um
aerodynamic diameter. This measurement includes the fine particles and a
subset of coarse particles, and is an indicator for particles that can be inhaled
and penetrate to the thoracic region of the lung; also referrred to as thoracic
particles.
Particles measured by a sampler that contains a size fractionator (classifier)
designed with an effective cut point (50% collection efficiency) of 2.5 um
aerodynamic diameter. This measurement, which generally includes all fine
particles, is an indiator for fine particles; also referred to as fine-fraction
particles. A small portion of coarse particles may be included depending on
the sharpness of the sampler efficiency curve.
Particles measured directly using a dichotomous sampler or by subtraction of
particles measured by a PM2 5 sampler from those measured by a PM10
sampler. This measurement is an indicator for the coarse fraction of thoracic
particles; also referred to as thoracic coarse particles or coarse-fraction
particles.
2-7
-------
The formation and growth of fine particles are influenced by several processes including:
(1) nucleation (i.e., gas molecules coming together to form a new particle); (2) condensation of
gases onto existing particles; (3) coagulation of particles, the weak bonding of two or more
particles into one larger particle; (4) uptake of water by hygroscopic components; and (5) gas
phase reactions which form secondary PM. Gas phase material condenses preferentially on
smaller particles since they have the greatest surface area, and the efficiency of coagulation for
two particles decreases as the particle size increases. Thus, ultrafine particles grow into the
accumulation mode, but accumulation-mode particles do not normally grow into coarse particles
(CD, p. 2-29).
Secondary formation processes can result in either new particles or the addition of PM to
pre-existing particles. Examples of secondary particle formation include: (1) the conversion of
sulfur dioxide (SO2) to sulfuric acid (H2SO4) droplets that further react with gaseous ammonia
(NH3) to form various sulfate particles (e.g., ammonium sulfate (NH4)2SO4 or ammonium
bisulfate NH4HSO4); (2) the conversion of nitrogen dioxide (NO2) to nitric acid (HNO3) vapor
that reacts further with ammonia to form ammonium nitrate (NH4NO3) particles; and (3)
reactions involving gaseous volatile organic compounds (VOC) yielding organic compounds
with low ambient temperature (saturation) vapor pressures that nucleate or condense on existing
particles to form secondary organic aerosol particles (CD, p. 3-65 to 3-71). In most of the
ambient monitoring data displays shown later in this chapter, the first two types of secondary PM
are generally labeled plurally as 'sulfates' and 'nitrates' (respectively), which implies that the
ammonium content is encompassed. The third type of secondary PM may be lumped with the
directly emitted elemental or organic carbon particles and labeled 'total carbonaceous mass,' or
the two types of carbonaceous PM may be reported separately as elemental carbon (EC) and
organic carbon (OC).
2.2.3 Chemical Composition
Based on studies conducted in most parts of the U.S., the CD reports that a number of
chemical components of ambient PM are found predominately in fine particles including:
sulfate, ammonium, and hydrogen ions; elemental carbon8, secondary organic compounds, and
8 Also called light absorbing carbon and black carbon. The terms elemental carbon and black carbon are
often used interchangeably, but may be defined differently by different users. Black carbon is most often used in
discussions of optical properties and elemental carbon is most often used when referring to chemical composition.
In many cases, there is little difference between the two, but care must be taken when comparing data from studies
with different purposes. In addition, the term soot is also used in many instances to refer to either EC or BC. The
differences between soot and either EC or BC can be significant, as soot refers to elemental carbon formed from gas
phase hydrocarbons in the combustion process, and tends to be in the submicron fraction and often in the fraction of
particles that are smaller than 0.10 microns in aerodynamic diameter. EC and BC both include carbonaceous
particles formed from incomplete burnout of solid carbonaceous fuels; these particles have distinctly different
physical characteristics compared to char. As an additive to automotive tires, commercially produced 'carbon black'
and associated contaminants can also be found in resuspended urban road dust.
2-8
-------
primary organic species from cooking and combustion; and certain metals, primarily from
combustion processes. Chemical components found predominately in coarse particles include:
crustal-related materials such as calcium, aluminum, silicon, magnesium, and iron; and primary
organic materials such as pollen, spores, and plant and animal debris (CD, p. 2-38).
Some components, such as nitrate and potassium, may be found in both fine and coarse
particles. Nitrate in fine particles comes mainly from the reaction of gas-phase nitric acid with
gas-phase ammonia to form ammonium nitrate particles. Nitrate in coarse particles comes
primarily from the reaction of gas-phase nitric acid with pre-existing coarse particles (CD, p. 2-
38). Potassium in coarse particles comes primarily from soil, with additional contributions from
sea salt in coastal areas. Potassium in fine particles, generally not a significant contributor to
overall mass, comes mainly from emissions of burning wood, with infrequent but large
contributions from fireworks, as well as significant proportions from the tail of the distribution
of coarse soil particles (i.e., < 2.5 jim in diameter ) in areas with high soil concentrations.
Many ambient particles also contain water (i.e., particle-bound water) as a result of an
equilibrium between water vapor and hygroscopic PM (CD, p. 2-40). Particle-bound water
influences the size of particles and in turn their aerodynamic and light scattering properties
(discussed in section 2.2.5). Particle-bound water can also act as a carrier to convey dissolved
gases or reactive species into the lungs which, in turn, may cause heath consequences. (CD, p.
2-112). The amount of particle-bound water in ambient particulate matter will vary with the
particle composition and the ambient relative humidity. Sulfates, nitrates, and some secondary
organic compounds are much more hygroscopic than elemental carbon (EC), primary organic
carbon (OC), and crustal material.
2.2.4 Fate and Transport
Fine and coarse particles typically exhibit different behaviors in the atmosphere. These
differences may affect several exposure-related considerations, including the representativeness
of central-site monitored values and the penetration of particles formed outdoors into indoor
spaces. The ambient residence time of atmospheric particles varies with size. Ultrafme particles
have a very short life, on the order of minutes to hours, since they are more likely to reach the
accumulation mode. However, their chemical content persists in the accumulation mode.
Ultrafme particles are also small enough to be removed through diffusion to falling rain drops.
Accumulation-mode particles remain suspended longer (i.e. accumulate ) in the atmosphere
because they are too large to diffuse rapidly to surfaces or to other particles and too small to
settle out or impact on stationary objects. They can be transported thousands of kilometers and
remain in the atmosphere for days to weeks. Accumulation-mode particles serve as
condensation nuclei for cloud droplet formation and are eventually removed from the
atmosphere in falling rain drops. Accumulation-mode particles that are not involved in cloud
2-9
-------
processes are eventually removed from the atmosphere by gravitational settling and impaction on
surfaces.
By contrast, coarse particles can settle rapidly from the atmosphere with lifetimes
ranging from minutes to days depending on their size, atmospheric conditions, and altitude.
Larger coarse particles are not readily transported across urban or broader areas, because they
are generally too large to follow air streams, and they tend to be easily removed by gravitational
settling and by impaction on surfaces. Smaller coarse particles extending into the tail of the
distribution can have longer lifetimes and travel longer distances, especially in extreme
circumstances. For example, dust storms in desert areas of Africa and Asia lift coarse particles
to high elevations and these 'dust clouds' are readily observed to undergo intercontinental
transport to North America (CD, p. 2-49). Coarse particles also are readily removed by falling
rain drops (CD, p. 2-50).
The characteristics of ultrafme, accumulation-mode, and coarse-mode particles that were
discussed in the preceding sections are summarized in Table 2-2.
2.2.5 Optical Properties of Particles
Particles and gases in the atmosphere scatter and absorb light and, thus, affect visibility.
As discussed in section 4.3 of the CD, the efficiency of particles in causing visibility impairment
depends on particle size, shape, and composition. Accumulation-mode particles are more
efficient per unit mass than coarse particles in causing visibility impairment. The accumulation-
mode particle components principally responsible for visibility impairment are sulfates, nitrates,
organic matter, and elemental carbon. Soil dust, particularly in the fine tail of the coarse particle
distribution, can also impair visibility. All of these particles scatter light to some degree, but, of
these, elemental carbon plays the most significant role in light absorption. Since elemental
carbon, which is a product of incomplete combustion from activities such as the burning of wood
or diesel fuel, is a relatively small component of PM in most areas, visibility impairment is
generally dominated by light scattering rather than by light absorption.
Because humidity causes hygroscopic particles to grow in size, humidity plays a
significant role in particle-related visibility impairment. The amount of increase in particle size
with increasing relative humidity depends on particle composition. Humidity-related particle
growth is a more important factor in the eastern U.S., where annual average relative humidity
levels are 70 to 80 percent compared to 50 to 60 percent in the western U.S. Due to relative
humidity differences, aerosols of a given mass, dry particle size distribution, and composition
would likely cause greater visibility impairment in an eastern versus a western location. The
relationship between ambient PM and visibility impairment is discussed below in Section 2.8.
2-10
-------
Table 2-2. Comparison of Ambient Fine Particles
(Ultrafine plus Accumulation-Mode) and Coarse Particles
Fine
Ultrafine
Accumulation
Coarse
Formation
Processes:
Formed by:
Composed
of:
Solubility:
Sources:
Combustion, high-temperature
processes, and atmospheric reactions
Nucleation
Condensation
Coagulation
Sulfate
Elemental carbon
Metal compounds
Organic compounds
with very low
saturation vapor
pressure at ambient
temperature
Probably less soluble
than accumulation
mode
Combustion
Atmospheric
transformation of
SO2 and some
organic compounds
High temperature
processes
Atmospheric Minutes to hours
half-life:
Removal
Processes:
Travel
distance:
Grows into
accumulation mode
Diffuses to raindrops
-------
2.2.6 Other Radiative Properties of Particles
In addition to the optical properties related to visibility summarized above, ambient
particles scatter and absorb radiation across the full electromagnetic spectrum, including
ultraviolet, visible, and infrared wavelengths, affecting climate processes and the amount of
ultraviolet radiation that reaches the earth. As discussed in section 4.5 of the CD, the effects of
ambient particles on the transmission of these segments of the electromagnetic spectrum depend
on the radiative properties of the particles, which in turn are dependent on the size and shape of
the particles, their composition, the distribution of components within individual particles, and
their vertical and horizontal distribution in the lower atmosphere.
The effects of PM on the transfer of radiation in the visible and infrared spectral regions
play a role in global and regional climate. Direct effects of particles on climatic processes are
the result of the same processes responsible for visibility degradation, namely radiative
scattering and absorption. However, while visibility impairment is caused by particle scattering
in all directions, climate effects result mainly from scattering light away from the earth and into
space. This reflection of solar radiation back to space decreases the transmission of visible
radiation to the surface 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 elemental carbon, results in an increase in the heating
rate of the lower atmosphere.
The relative proportions of scattering and absorption by ambient particles are highly
dependent on their composition and optical properties and on the wavelength of the radiation.
For example, sulfate and nitrate particles effectively scatter solar radiation, and they weakly
absorb infrared, but not visible, radiation. The effects of mineral dust particles are complex;
depending on particle size and refractive index, mineral aerosol can reflect or absorb radiation.
Dark minerals absorb across the solar and infrared spectral regions leading to warming of the
atmosphere. Light-colored mineral particles in the appropriate size range can scatter visible
radiation, reducing radiation received at the earth's surface. Organic carbon particles mainly
reflect radiation, whereas elemental carbon particles strongly absorb radiation; however, the
optical properties of carbonaceous particles are modified if they become coated with water or
sulfuric acid. Upon being deposited onto surfaces, particles can also either absorb or reflect
radiation depending in part on the relative reflectivity of the particles and the surfaces on which
they are deposited.
The transmission of solar radiation in the ultraviolet (UV) range through the earth's
atmosphere is affected by ozone and clouds as well as by particles. The effect of particles on
radiation in the ultraviolet-B (UV-B) range, which has been associated with various biological
effects, is of particular interest. Relative to ozone, the effects of ambient particles on the
transmission of UV-B radiation are more complex. The CD notes that even the sign of the effect
can reverse as the composition of the particle mix in an air mass changes from scattering to
2-12
-------
absorbing types (e.g., from sulfate to elemental carbon), and that there is an interaction in the
radiative effects of scattering particles and absorbing molecules, such as ozone, in the lower
atmosphere.
2.3 AMBIENT PM MEASUREMENT METHODS
The methods used to measure ambient PM are important to understanding population
exposure to PM, evaluating health and welfare risks, and developing and evaluating the
effectiveness of risk management strategies. Because PM is a complex mixture of substances
with differing physical and chemical properties , measuring and characterizing particles
suspended in the atmosphere is a significant challenge.9 Ambient measurements include particle
mass, composition, and particle number. Most instruments collect PM by drawing a controlled
volume of ambient air through a size-selective inlet, usually defined by the inlet's 50 percent cut
point. Measurable indicators of fine particles include PM2 5, PMLO, British or black smoke (BS),
coefficient of haze (COH), and PM10 (in areas dominated by fine particles). Measurable
indicators of coarse-mode particles include PM10_2 5, PM15.2 5, and PM10 (in areas dominated by
coarse-mode particles).
2.3.1 Particle Mass Measurement Methods
Ambient PM mass can be measured directly, by gravimetric methods, or indirectly, using
methods that rely on the physical properties of particles. Methods can also be segregated as
either discrete or continuous according to whether samples require laboratory analysis or the data
are available in real-time. Discrete methods provide time integrated data points (typically over a
24- hour period) that allow for post-sampling gravimetric analyses in the laboratory. These
methods are typically directly linked to the historical data sets that have been used in health
studies that provide the underlying basis for having a NAAQS. Continuous methods can provide
time resolution on the order of minutes and automated operation up to several weeks, facilitating
the cost-effective collection of greater amounts of data compared with discrete methods.
The most common direct measurement methods include filter-based methods where
ambient aerosols are collected for a specified period of time (e.g., 24 hours) on filters that are
weighed before and after collection to determine mass by difference. Examples include the FRM
monitors for PM2 5 and PM10. Dichotomous samplers contain a separator that splits the air stream
from a PM10 inlet into two streams so that both fine- and coarse-fraction particles can be
collected on separate filters. These gravimetric methods require weighing the filters after they
9 Refer to CD Chapter 2 for more comprehensive assessments of particle measurement methods. A recent
summary of PM measurement methods is also given inFehsenfeld et al. (2003). Significant improvements and
understanding of routine and advanced measurement methods is occurring through EPA's PM Supersites Program
(see www.epa.gov/ttn/amtic/supersites.html').
2-13
-------
are subjected to specific equilibrium conditions (i.e., 20 - 23° C and 30 - 40 percent RH in most
cases).
Discrete, gravimetric methodologies have been refined over the past 20 years as PM
monitoring networks have evolved from sampling based on the high volume TSP and PM10
method to the PM2 5 FRM. The inclusion of such measures as size-selective inlets and
separators, highly specific filter media performance criteria, active flow control to account for
ambient changes in temperature and pressure, and highly prescriptive filter weighing criteria
have reduced levels of measurement uncertainty, compared with earlier methods.
National quality assurance data analyzed by EPA between 1999-2001 indicate that the
PM25 FRM has been a robust indicator of ambient levels by meeting the data quality objectives
(DQO) established at the beginning of the monitoring program. Three-year average estimates
from reporting organizations aggregated on a national basis for collocated sampler precision (7.2
percent), flow rate accuracy (0.18 percent), and method bias (-2.06 percent, from the
Performance Evaluation Program)10 are well within their respective goals of+10 percent, +4
percent, and +10 percent.
There are a number of continuous PM measurement techniques. A commonly used
method is the Tapered Element Oscillating Microbalance (TEOM®) sensor, consisting of a
replaceable filter mounted on the narrow end of a hollow tapered quartz tube. The air flow
passes through the filter, and the aerosol mass collected on the filter causes the characteristic
oscillation frequency of the tapered tube to change in direct relation to particle mass. This
approach allows mass measurements to be recorded on a near-continuous basis (i.e., every few
minutes).
The next generation of the TEOM® is the Filter Dynamics Measurement System
(FDMS®) monitor. This method is based upon the differential TEOM that is described in the
CD (CD, p. 2-78). The FDMS method employs an equilibration system integrated with a
TEOM® having alternating measurements of ambient air and filtered air. This self-referencing
approach allows the method to determine the amount of volatile PM that is evaporating from the
TEOM sensor for 6 of every 12 minutes of operation. An hourly measurement of the total
aerosol mass concentration, including non-volatile and volatile PM, is calculated and reported
every 6 minutes.
Other methods that produce near-continuous PM mass measurements include the beta
attenuation sampler and the Continuous Ambient Mass Monitor (CAMM). A beta attenuation
10 The Performance Evaluation Program (PEP) is designed to determine total bias for the PM2 5 sample
collection and laboratory analysis processes. Federally referenced audit samplers are collocated adjacent to a
monitoring site's routine sampler and run for a 24-hour period. The concentrations are then determined
independently by EPA laboratories and compared in order to assess bias. The performance evaluations are
conducted four times per year (once per quarter) at one-fourth (25 percent) of the sampling sites in a reporting
organization.
2-14
-------
(or beta gauge) sampler determines the mass of particles deposited on a filter by measuring the
absorption of electrons generated by a radioactive isotope, where the absorption is closely
related to the mass of the particles. The CAMM measures the pressure drop increase that occurs
in relation to particle loading on a membrane filter. Both methods (beta-attenuation and
CAMM) require calibration against standard mass measurements as neither measures PM mass
directly by gravimetric analysis.
The number of continuous PM25 monitors across the U.S. has increased from 300 to over
500 between 2003 and 2005. Although a subset of these monitors were required by regulation to
be placed in metropolitan areas of greater than 1 million population, a higher percentage were
installed to provide improved temporal resolution for daily air quality index reporting and PM2 5
forecast verifications through EPA's AIRNOW program. Some of the continuous PM25 data
reported through the AIRNOW program are adjusted to better match FRM results.11 The
continuous data used in the analyses in this chapter were obtained from EPA's Air Quality
System (AQS); some of these AQS data are adjusted and some are not. There is currently an
effort underway to better characterize this facet of the continuous data in AQS. Still, the AQS
continuous data utilized in analyses here do show excellent correlation with collocated FRM
measurements; over 95 percent of the continuous/FRM site pairs had a correlation coefficient of
over 0.72, and almost 75 percent had a correlation of 0.9 or higher (Schmidt et al., 2005).
Work also continues on the development of national approval criteria for determining
regional and national equivalency for continuous PM2 5 monitors. Once promulgated, these
criteria would provide the regulatory basis for approving appropriate continuous methods as
equivalent to FRMs, and permit the assessment of NAAQS attainment status with continuous
PM25 data, reducing the number of manually-operated FRM monitors that need to continue
operating.
2.3.2 Indirect Optical Methods
PM has also been characterized in the U.S. and elsewhere by indirect optical methods
that rely on the light scattering or absorbing properties of either suspended PM or PM collected
11 When data are sent to the AIRNOW website, they are assumed to be "FRM like" which means that their
values are highly correlated (R2 > 0.8) with actual FRM concentrations so that values can be compared not only to
the FRM measurements but also across State boundaries. Statistical adjustments to the raw continuous data are
necessary because some of the sampling methodologies, such as the TEOM monitors, have inlets heated from 30°C
to 50°C which causes semi-volatile fine paniculate matter including nitrates to be vaporized and never measured.
The result of this vaporization is a lower measured TEOM concentration when compared to the FRM. Adjustments
have been accomplished on a seasonal basis as well as using meteorologic variables (e.g., ambient temperature) with
linear and non-linear regression techniques. The need to adjust the continuous data can depend on several factors
including the type of method, the location of the site in the country and the composition of the ambient paniculate
matter being measured.
2-15
-------
on a filter.12 These include BS, COH, and estimates derived from visibility measurements. In
locations where they are calibrated to standard mass units, these indirect measurements can be
useful surrogates for particle mass. The BS method typically involves collecting samples from a
4.5 |im inlet onto white filter paper where blackness of the stain is measured by light absorption.
Smoke particles composed primarily of elemental carbon (EC), including black carbon (BC),
typically make the largest contribution to stain darkness. COH is determined using a light
transmittance method. This involves collecting samples from a 5.0 jim inlet onto filter tape
where the opacity of the resulting stain is determined. This technique is somewhat more
responsive to non-carbon particles than the BS method. Nephelometers measure the light
scattered by ambient aerosols in order to calculate light extinction. This method results in
measurements that can correlate well with the mass of fine particles below 2 jim diameter. Since
the mix of ambient particles varies widely by location and time of year, the correlation between
BS, COH, and nephelometer measurements and PM mass is highly site- and time-specific. The
optical methods described here, as well as the particle counters described below, are based on the
measurement of properties such as light scattering and electric mobility, which are inherently
different than previous methods described based on aerodynamic diameter.
2.3.3 Size-Differentiated Particle Number Concentration Measurement Methods
Recently there has been increasing interest in examining the relationship between the
particle number concentration by size and health effects. Several instruments are needed to
provide size distribution measurements (number and size) over the 5 orders of magnitude of
particle diameters of interest. A nano-scanning mobility particle sizer (NSMPS) counts particles
in the 0.003 to 0.15 jim range. A standard scanning mobility particle sizer (SMPS) counts
particles in the 0.01 to 1 |im range, and a laser particle counter (LPC) counts particles in the 0.1
to 2 |im range. An aerodynamic particle sizer measures particles in the 0.7 to 10 |im range.
These techniques, while widely used in aerosol research, have not yet been widely used in health
effects studies.
2.3.4 Chemical Composition Measurement Methods
There are a variety of methods used to identify and describe the characteristic
components of ambient PM.13 X-ray fluorescence (XRF) is a commonly used laboratory
technique for analyzing the elemental composition of primary particles deposited on filters. Wet
chemical analysis methods, such as ion chromatography (1C) and automated colorimetry (AC)
12 See Section 2.2.5 of this chapter for a discussion of the optical properties of PM.
13 The reader is referred to Chapter 2, section 2.2, of the CD for a more thorough discussion of sampling
and analytical techniques for measuring PM. Methods used in EPA's National PM2 5 Speciation Trends Network
and other special monitoring programs are summarized in Solomon et al. (2001).
2-16
-------
are used to measure ions such as nitrate (NO3"), sulfate (SO4), chloride (Cl"), ammonium (NH4+),
sodium (Na+), organic cations (such as acetate), and phosphate (PO43").
There are several methods for separating organic carbon (OC) and elemental carbon (EC)
or black carbon (BC) in ambient filter samples. Thermal optical reflectance (TOR), thermal
manganese oxidation (TMO), and thermal optical transmittance (TOT) have been commonly
applied in aerosol studies in the United States. The thermal optical transmission (TOT) method,
used in the EPA speciation program, uses a different temperature profile than TOR, which is
used in the Interagency Monitoring of Protected Visual Environments (IMPROVE) visibility
monitoring program. The two methods yield comparable estimates of total carbon, but give a
different split between OC and EC.
Commercial instruments are now available to measure carbon (OC, EC, TC), nitrate, and
sulfate on a near-continuous basis. These instruments provide time-resolved measurements from
a few minutes to a few hours. The semi-continuous methods involved a variety of techniques
that include thermal reduction; wet impaction and flash vaporization; and thermal oxidation with
non-dispersive infrared (NDIR) detection. They have been field tested and compared through
the EPA's Environmental Technology Verification (ETV) program and the Supersites program
and proven to be good candidates for additional testing (EPA, 2004a). Data are now becoming
available from regional planning and multi-state organizations and the EPA to understand the
comparison with filter-based methods and the potential limitations of these technologies.
The U.S. EPA is coordinating a pilot study of semi-continuous speciation monitors at five
Speciation Trends Network (STN) sites. The pilot study began in 2002. The goals of the pilot
study are to assess the operational characteristics and performance of continuous carbon, nitrate,
and sulfate monitors for routine application at STN sites; work with the pilot participants and the
vendors to improve the measurement technologies used; and evaluate the use of an automated
data collection and processing system for real time display and reporting. After the pilot
monitoring and data evaluation phase, proven semi-continuous monitors will become the
framework for a long-term network of up to 12 STN sites equipped with semi-continuous sulfate,
nitrate, and carbon monitors.
2.3.5 Measurement Issues
There is no perfect PM sampler under all conditions, so there are uncertainties between
the mass and composition collected and measured by a sampler and the mass and composition of
material that exists as suspended PM in ambient air (Fehsenfeld et al., 2003). To date, few
standard reference materials exist to estimate the accuracy of measured PM mass and chemical
composition relative to what is found in air. At best, uncertainty is estimated based on
collocated precision and comparability or equivalency to other similar methods, which
themselves have unknown uncertainty, or to the FRM, which is defined for regulatory purposes
but is not a standard in the classical sense. There are a number of measurement-related issues
2-17
-------
that can result in positive or negative measurement artifacts which could affect the associations
that epidemiologic researchers find between ambient particles and health effects.
The semi-volatile components of PM can create both positive and negative measurement
artifacts. Negative artifacts arise from evaporation of the semi-volatile components of PM
during or after collection, which is caused by changes in temperature, relative humidity, or
aerosol composition, or due to the pressure drop as collected air moves across the filter. Nitrate
losses due to evaporation may represent as much as 10-20 percent of total PM25mass, as shown
in southern California studies (CD, p. 2-68). Positive artifacts arise when gas-phase compounds
absorb onto or react with filter media or already collected PM, or when particle-bound water is
not removed. The chemical interaction of gases being collected with particles already on the
filter and conversion of PM components to gas-phase chemicals can also result in negative
artifacts. These interactions depend on the compounds contained in collected particles and in the
gas phase, and also depend on both location and time.
Particle-bound water can represent a significant fraction of ambient PM mass under
conditions where relative humidity is more than 60 percent (CD; p. 2-63, p. 2-109). It can also
represent a substantial fraction of gravimetric mass at normal equilibrium conditions (i.e., 22° C,
35 percent RH) when the aerosol has high sulfate content. The amount of particle-bound water
will vary with the composition of particles, as discussed in section 2.2.3. The use of heated
inlets to remove particle-bound water (e.g. TEOM at 50° C) can result in loss of semi-volatile
compounds unless corrective techniques are applied, although the newer generation TEOM's use
diffusion dryers rather than heating to reduce the relative humidity (CD, p. 2-100, Table 2-7).
In areas with significant amounts of dust, high wind conditions resulting in blowing dust
can interfere with accurate separation of fine- and coarse-fraction particles. In these unique
conditions a significant amount of coarse-fraction material can be found in the inter-modal
region between 1 and 3 jam, thus overstating the mass of fine-fraction particles. The addition of
a PMj 0 measurement in these circumstances can provide greater insights into the magnitude of
this problem (CD, p. 9-12).
2.4 PM CONCENTRATIONS, TRENDS, AND SPATIAL PATTERNS
This section provides analysis of the latest available PM air quality data, including PM
levels, composition, and spatial patterns. The EPA and the States have been using a national
network to measure and collect PM10 concentrations since 1987, and PM2 5 concentrations since
1999. Summaries through the end of 2003, based on data publicly available from EPA's Air
Quality System (AQS) as of August 2004, are presented here. PM25 data from the IMPROVE
network are also presented. Many data summaries are presented by region, as shown in
Figure 2-3. These regions are the same as those defined in the CD and have proven useful for
understanding potential differences in the characteristics of PM in different parts of the U.S..
2-18
-------
Southern
California
Figure 2-3. Regions used in PM Staff Paper in data analysis summaries.
2-19
-------
As is the case with all surface-based ambient monitoring data, these data can be considered
representative of exposures in typical breathing zones in the lowest 15 meters of the atmosphere.
2.4.1 PM25
Following the establishment of new standards for PM2 5 in 1997, the EPA led a national
effort to deploy and operate over 1000 PM25 monitors. Over 90 percent of the monitors are
located in urban areas. These monitors use the PM2 5 FRM which, when its procedures are
followed, assures that PM data are collected using standard equipment, operating procedures,
and filter handling techniques.14 Most of these FRM monitors began operation in 1999. The
EPA has analyzed the available data collected by this network from 2001-2003. Data from the
monitors were screened for completeness with the purpose of avoiding seasonal bias. To be
included in these analyses, a monitoring site needed all 12 quarters (2001-2003), each with 11 or
more observations. A total of 827 FRM sites in the U.S. met these criteria.15
The 3-year average annual PM2 5 mean concentrations range from about 4 to 28 |ig/m3,
with a median of about 13 |ig/m3. The 3-year average annual 98th percentiles of the 24-hour
average concentrations range from about 9 to 76 |ig/m3, with a median of about 32 |ig/m3.
Figures 2-4 and 2-5 depict the regional distribution of site-specific 3-year average annual mean
and 3-year average 98th percentile 24-hour average PM25 (and PM10_25, discussed in section
2.4.3) concentrations, respectively, by geographic region (excluding Alaska, Hawaii, Puerto
Rico, and the Virgin Islands). In general, with the exception of southern California, PM2 5 annual
average mass is greater in the eastern regions than in the western regions, whereas PM10_2 5
annual average mass is greater in the western regions. Figures 2-6 and 2-7 are national maps that
depict county-level 3-year average annual mean and 3-year average annual 98th percentile 24-
hour average PM25 concentrations, respectively, from the FRM network.16 The site with the
highest concentration in each monitored county is used to represent the value in that county. The
map and box plots show that many locations in the eastern U.S. and in California had annual
mean PM25 concentrations above 15 |ig/m3. Mean PM25 concentrations were above 18 |ig/m3 in
several urban areas throughout the eastern U.S., including Chicago, Cleveland, Detroit,
14 See 40 CFR Parts 50 and 58 for monitoring program requirements.
15 810 of the 827 monitors are located in the contiguous continental U.S. covered by the regions shown in
Figure 2-3. The remainder are located in Alaska, Hawaii, and U.S. territories.
16 No conclusions should be drawn from these data summaries regarding the potential attainment status of
any area. EPA regulations, in 40 CFR Part 50, Appendix N, require 3 consecutive years of monitoring data and
specify minimum data completeness requirements for data used to make decisions regarding attainment status.
Although 11 samples per quarter, as required in these analyses, is sufficient to show nonattainment, additional data
capture (at least 75 percent per quarter) is required to show attainment of the standards. Not all of the PM federal
reference method (FRM) sites that contributed data to the summaries presented here recorded 75 percent data
capture for all four calendar quarters for each of the 3 years.
2-20
-------
CO
E
cc
(U
(C
75:
65
55
45
35
15
5
Northeast
*
^
X
k
-^-
Southeast
*
_,
T
.
r r r
PM25 PM10.25 PM25 PM10.2
= 121 63 216 97
strial Midwest
E
*
^
>l
k
4-
pper Midwest
Z)
^
^
¥•
^
r r r
5 PM25 PM10.25 PM25
217 97 71
K
-i— •
00
(U
^
-1— •
0
CO
4-
PM10.25 PM25
41 33
Northwest
£
*
lern California
j^_
0
CO
i
*
F~
r r
PM10.25 PM25 PM10.25 PM25 PM10.25
32 110 108 42 37
Figure 2-4. Distribution of annual mean PM2 5 and estimated annual mean PM10_2 5 concentrations by
region, 2001-2003. Box depicts interquartile range and median; whiskers depict 5th and 95th percentiles;
asterisks depict minima and maxima. N = number of sites.
Source: Schmidt et al. (2005)
2-21
-------
PM
2.5
PM10-2.5 PM2.5
PM,
PM,
PM
PM
2.5
'10-2.5
'10-2.5
"2.5
PM
10-2.5
'10-2.5 ' 1U12.5 ' '"'10-2.5
N= 121 63 216 97 217 97 71 41 33 32 110 108 42 37
Figure 2-5. Distribution of 98th percentile 24-hour average PM2 5 and estimated PM10_2 5 concentrations by
region, 2001-2003. Box depicts interquartile range and median; whiskers depict 5th and 95th percentiles;
asterisks depict minima and maxima. N = number of sites.
Source: Schmidt et al. (2005)
-------
PM2 5 Concentration (|ig/m3) p——} x <= 12 i 1 12 < x <= 15
562 counties fflmffl 15 < x <= 18 ^^H x > 18
Figure 2-6. County-level maximum annual mean PM25 concentrations, 2001-2003.
Source: Schmidt et al. (2005) 2-23
-------
PM25 Concentration (jig/m3) p——] x <= 30 I 1 30 < x <= 40
562 counties HffiHffl 40 < x <= 65 ^^H x > 65
Figure 2-7. County-level maximum 98th percentile 24-hour average PM25 concentrations, 2001-2003.
Source: Schmidt et al. (2005) 2-24
-------
Indianapolis, Pittsburgh, and St. Louis. Los Angeles and the central valley of California also
were above 18 |ig/m3. Sites in the upper midwest, southwest, and northwest regions had
generally lower 3-year average annual mean PM2 5 concentrations, most below 12 |ig/m3. Three-
year average annual 98th percentile 24-hour average PM2 5 concentrations above 65 |ig/m3 appear
only in California. Values in the 40 to 65 |ig/m3 range were more common in the eastern U.S.
and on the west coast, mostly in or near urban areas, but relatively rare in the upper midwest and
southwest regions. In these regions, the 3-year average 98th percentile PM25 concentrations were
more typically below 40 |ig/m3, with many below 25 |ig/m3.
The PM maps shown in this chapter encompass all valid data, including days that were
flagged for episodic events, either natural or anthropogenic. Examples of such events include
biomass burning, construction/demolition activities, dust storms, and volcanic and seismic
activity. PM concentrations can increase dramatically with these 'natural' or 'exceptional'
events. Although these events are rare (e.g., affecting less than 1 percent of reported PM25
concentrations between 2001 and 2003), they can affect people's short-term PM exposure,
briefly pushing daily PM levels into the unhealthy ranges of the Air Quality Index (AQI).
Analyses of 2001-2003 PM25 data found that over 9 percent of the days above (site-based)
98th percentile 24-hour concentrations were flagged for events. The events, in fact, were found
to cause the 98th percentiles to inflate by up to 18 |ig/m3, with an average increase of 0.8 |ig/m3.
Natural and exceptional events, however, rarely have a significant effect on annual or longer
averages of PM. In the afore-mentioned analyses of 2001-2003 PM25 data, the average effect of
natural and exceptional events on 3-year annual means was less than 0.1 |ig/m3 (Schmidt, et al.,
2005). Episodic event-flagged data are often excluded from trends-type analyses and are
addressed for the purpose of determining compliance with the NAAQS by EPA's national and
exceptional events policies, as described below in section 2.6.
PM2 5 short-term trends were recently evaluated by EPA in The Particle Pollution Report
(EPA, 2004b, p. 14). In the EPA FRM network, PM2 5 annual average concentrations decreased
10 percent nationally from 1999 to 2003. The northeast, where moderate concentrations are
found, was the only region that did not show a decline between these years; annual
concentrations in that region were somewhat flat or rose slightly (about 1 percent) over the 5-
year period. Except in the northeast, PM2 5 generally decreased the most in the regions with the
highest concentrations - the southeast (20 percent), southern California (16 percent), and the
industrial midwest (9 percent) from 1999 to 2003. The remaining regions with lower
concentrations (the upper midwest, the southwest, and the northwest) posted modest declines in
PM25; see Figure 2-8 (EPA, 2004b, p. 15).
The IMPROVE monitoring network, which consists of sites located primarily in national
parks and wilderness areas throughout the U.S., generally provides data for long-term PM25
2-25
-------
Annual Average PM2 5 Concentrations, 1999-2003
Northwest
11.4 10.7
Industrial
Midwest
Northeast
_
-
132 134
T1%
Southern
California
20.0
16.9
National Standard: 15 ug/m
Regional Trend
Figure 2-8. Regional trends in annual average PM2.5 concentrations in the EPA network, 1999-2003.
Source: EPA (2004b)
2-26
-------
trends in rural areas.17 Figure 2-9 shows the composite long-term trend at 8 eastern sites, 17
western sites, and one urban site in Washington, D.C. The 4 westernmost U.S. subregions
(northwest, southern California, upper midwest, and southwest) are considered 'western sites'
and the 3 eastern subregions (northeast, southeast, and industrial midwest) are considered
'eastern sites.' At the eastern rural sites, measured PM2 5 mass decreased about 23 percent from
1993 to 2003. At the western rural sites, PM25 mass decreased about 21 percent from 1993 to
2003. At the Washington, D.C. site, the annual average PM2 5 concentration in 2003 was about
31 percent lower than the value in 1993.
The relative spatial homogeneity of the ambient air across a specified area can be
assessed by examining the values at multiple sites using several indicators, including: (1) site
pair correlations, (2) differences in long-term (e.g., annual and multi-year) average
concentrations, and (3) differences in short-term (e.g., daily) average concentrations. An
analysis of these indicators for site pairs in 27 Metropolitan Statistical Areas (MS As) using PM25
FRM monitoring data from 1999-2001 is included in the CD (CD, Appendix 3 A). A similar
analysis, for 49 urban areas defined as either Core Based Statistical Areas (CBSAs) or Combined
Statistical Areas (CSAs), was conducted on PM25 FRM monitoring data from 2001-2003
(Schmidt etal.,2005).18
An analysis of site pairs from each of the 49 urban areas indicates that multiple sites in
these areas were highly correlated throughout the period. About 83 percent (1901 out of 2290)
of the between-site correlation coefficients in all 49 areas were greater than or equal to 0.80, and
more than 48 percent (1113 out of 2290) of the correlations were greater than or equal to 0.90.
Further, every area had at least one monitor pair with a correlation coefficient greater than or
equal to 0.82.
A summary of the analyses of long-term and short-term concentration differences for the
49 urban areas is shown in Table 2-3. The difference in 3-year average annual mean PM25
concentrations between monitor pairs in the 49 cities ranged from less than 1 |ig/m3 in four
areas to about 18 |ig/m3 in Los Angeles. Large differences in mean concentrations across a
metropolitan area may be due to differences in emissions sources, meteorology, or topography.
Small differences may be due only to measurement imprecision (CD, p. 3-46). Most sites in the
49 areas had annual means within 15 percent of the area spatial average; the largest percent
17 IMPROVE monitoring instalments and protocols (defined at http://vista.cira.colostate.edu/improve/) are
not identical to FRM monitors.
18 Metropolitan areas for use in federal statistical activities, as defined by the Office of Management and
Budget, include core-based statistical areas (CBSA) that are comprised of "metropolitan" and "micropolitan" areas,
and combined statistical areas (CSA) that are comprised of two or more core-based statistical areas. Counties are the
geographic building blocks for defining CBSA's. The analysis described here, and other analyses throughout this and
subsequent chapters, utilize the latest area definitions which are available at:
http://www.wMtehouse.gov/omb^lletiiis/fv05/b05-02.html.
2-27
-------
CO
E
c
03
CD
CD
D
C
C
CO
E
CD
CD
CD
D
C
CO
E
c
CD
CD
CD
n
12
10
8
6
4
2
0
25
20
15
10
WESTERN UNITED STATES
EASTERN UNITED STATES
WASHINGTON, DC
1993
1994
1995
1996 1997 1998 1999 2000 2001 2002 2003
i- Amm. Sulfate ™~ TCM -*- Crustal ^^ Amm. Nitrate
Figure 2-9. Average annual mean trend in PM2 5 mass, ammonium
sulfate, ammonium nitrate, total carbonaceous mass, and
crustal material at IMPROVE sites, 1993-2003.
Source: Schmidt et al. (2005)
2-28
-------
Table 2-3. Summary of PM2.5 FRM Data Analyses in 49 Metropolitan Areas, 2001-2003
Area *
Albuquerque, NM
Atlanta-Sandy Springs-Gainesville, GA
Bakersfield, CA
Baton Rouge- Pierre Part, LA
Birmingham-Hoover-Cullman, AL
Charlotte-Gastonia-Salisbury, NC-SC
Chicago-Naperville-Michigan City, IL-IN-WI
Cincinnati-Middletown- Wilmington, OH-KY-IN
Cleveland- Akron-Elyria, OH
Dallas-Fort Worth, TX
Denver-Aurora-Boulder, CO
Detroit- Warren-Flint, MI
Eugene-Springfield, OR
Grand Rapids-Muskegon-Holland, MI
Greensboro— Winston-Salem— High Point, NC
Houston-Baytown-Huntsville, TX
Indianapolis-Anderson-Columbus, IN
Kansas City-Overland Park-Kansas City, MO-KS
Knoxville-Sevierville-La Follette, TN
Las Vegas-Paradise-Pahrump, NV
Lexington-Fayette— Frankfort —Richmond, KY
Little Rock-North Little Rock-Pine Bluff, AR
Los Angeles-Long Beach-Riverside, CA
Louisville-Elizabethtown-Scottsburg, KY-IN
Memphis, TN-MS-AR
Miami-Fort Lauderdale-Miami Beach, FL
Milwaukee-Racine- Waukesha, WI
Minneapolis-St. Paul-St. Cloud, MN-WI
New Orleans-Metairie-Bogalusa, LA
New York-Newark-Bridgeport, NY-NJ-CT-PA
Omaha-Council Bluffs-Fremont, NE-IA
Philadelphia-Camden-Vineland, PA-NJ-DE-MD
Phoenix-Mesa-Scottsdale, AZ
Pittsburgh-New Castle, PA
Portland-Vancouver-Beaverton, OR-WA
Provo-Orem, UT
Raleigh-Durham-Cary, NC
Richmond, VA
Sacramento— Arden- Arcade— Truckee, CA-NV
Salt Lake City-Ogden-Clearfield, UT
San Diego-Carlsbad-San Marcos, CA
San Jose-San Francisco-Oakland, CA
San Juan-Caguas-Fajardo, PR
Seattle-Tacoma-Olympia, WA
St. Louis-St. Charles-Farmington, MO-IL
Virginia Beach-Norfolk-Newport News, VA-NC
Washington-Baltimore-Northern Virginia, DC-MD-VA-WV
Weirton-Steubenville, WV-OH
Wichita- Winfield, KS
N Sites
4
8
5
5
8
5
28
12
13
7
6
14
4
4
4
6
6
10
5
5
4
5
22
6
6
6
6
12
4
29
7
14
5
13
6
4
5
5
5
7
5
9
5
10
12
5
20
4
4
3 -year Average Annual Mean
Levels (f^g/m )
Area
Avg
7.0
15.9
15.3
12.3
14.8
14.3
14.7
16.0
15.5
12.8
8.7
15.2
9.4
13.0
14.6
11.7
15.3
12.0
15.3
7.1
14.4
13.0
19.0
15.6
13.1
8.2
13.1
10.5
11.5
13.5
10.4
14.9
9.3
15.8
8.2
9.8
13.3
13.4
9.9
11.4
15.0
10.8
7.2
9.4
15.0
12.5
14.5
17.1
10.9
Max
Site
10.2
18.0
21.8
13.1
18.0
14.9
17.7
17.8
18.3
13.9
10.8
19.5
13.4
13.8
15.8
14.2
16.7
13.9
16.7
11.0
15.7
14.1
27.8
16.9
14.0
9.5
13.2
12.0
12.2
16.4
10.7
16.4
11.4
21.2
9.5
10.9
13.9
14.0
12.5
14.0
15.9
11.8
9.3
11.1
17.5
13.0
16.7
17.8
11.1
Min
Site
5.0
14.1
6.7
10.8
12.6
14.0
11.7
14.5
13.4
11.7
4.5
12.6
6.6
12.3
14.0
9.6
13.6
10.8
14.2
4.0
13.5
11.9
9.9
14.1
11.7
7.4
12.5
9.7
10.4
11.2
9.8
13.8
6.3
13.2
6.1
8.8
12.2
12.8
7.6
9.0
12.8
8.4
5.1
5.3
14.0
11.9
12.2
16.2
10.2
Percent Difference
Largest
diff, any
site versus
Area Avg
31%
12%
56%
12%
18%
4%
20%
10%
15%
9%
48%
22%
30%
6%
8%
18%
11%
14%
8%
44%
8%
8%
48%
10%
11%
14%
5%
13%
10%
18%
6%
9%
32%
25%
26%
10%
8%
4%
23%
21%
15%
22%
29%
44%
14%
5%
16%
5%
6%
Max site
versus Min
site
51%
22%
69%
18%
30%
6%
34%
19%
27%
16%
58%
35%
51%
11%
11%
32%
19%
22%
15%
64%
14%
16%
64%
17%
16%
22%
5%
19%
15%
32%
8%
16%
45%
38%
36%
19%
12%
9%
39%
36%
19%
29%
45%
52%
20%
8%
27%
9%
8%
r
(Max site
versus Min
site)
0.42
0.71
0.00
0.85
0.78
0.94
0.77
0.95
0.87
0.92
0.40
0.85
0.57
0.91
0.94
0.78
0.93
0.76
0.86
0.03
0.86
0.79
0.50
0.85
0.86
0.73
0.96
0.79
0.91
0.85
0.86
0.94
0.22
0.75
0.84
0.88
0.93
0.88
0.37
0.92
0.89
0.67
0.71
0.30
0.82
0.93
0.82
0.87
0.96
24-Hour P90(ng/m3)**
Max
Pair
10.9
9.4
44.8
7.7
12.7
4.1
13.6
7.0
11.4
5.2
11.4
14.1
19.3
5.8
5.5
8.9
6.8
9.1
6.2
17.6
5.9
7.6
39.6
8.2
6.3
5.5
4.1
8.0
4.0
12.5
5.2
7.6
14.0
21.8
9.5
6.5
5.7
5.8
16.0
11.0
10.6
13.5
6.8
19.1
10.3
4.6
9.7
8.3
2.9
Min
Pair
2.6
3.5
6.0
2.4
3.5
1.7
2.2
2.4
3.2
2.3
4.0
3.2
4.8
3.2
2.5
6.2
2.0
1.4
2.7
2.5
3.3
5.1
5.3
3.9
2.2
1.7
2.2
2.6
2.8
2.0
2.1
3.1
4.2
3.2
3.0
3.0
2.4
3.2
6.0
3.8
4.6
4.7
1.7
2.9
2.2
2.7
2.6
6.1
1.3
r
(Max
Pair)
0.42
0.71
0.16
0.62
0.78
0.92
0.73
0.95
0.87
0.92
0.42
0.85
0.57
0.90
0.93
0.64
0.93
0.76
0.86
-0.03
0.86
0.78
0.50
0.85
0.82
0.73
0.93
0.79
0.90
0.84
0.78
0.94
0.22
0.69
0.76
0.92
0.88
0.88
0.21
0.92
0.69
0.67
0.71
0.30
0.76
0.90
0.82
0.86
0.91
1 'Area' is the larger of a Combined Statistical Area (CSA) or a Core Based Statistical Area (CBSA). See http://www.whitehouse.gov/omb/bulletins/fy05/b05-02.html.
:* 'P90' is the 90th percent'! e of the distribution of differences in 24-hour averages between two sites in the same urban area.
2-29
-------
difference between any site in an area and the area's spatial average ranged from 4 to 56 percent
with a median of 14 percent. In most urban areas (39 of the 49), the site pair with the maximum
and minimum annual mean concentration was highly correlated (r(maxmin) >0.70); there are,
however, some notable exceptions (i.e., 8 areas had r(maXjmin)<_0.50).
The spatial analysis also examined differences in 24-hour average concentrations
between the urban site pairs. Small differences throughout the distribution would indicate
relatively homogeneous concentration levels between the sites. Table 2-3 presents a summary of
the 90th percentile of the distribution (P90) of daily site pair differences in each urban area. The
site pairs with the largest difference (max pair) and the smallest difference (min pair) are shown.
The P90 values for the 2290 monitor pairs in the 49 urban areas ranged from about 1 to 45 |ig/m3.
Often the site pair with the maximum P90 value in each city was also the pair with the largest
annual mean difference. The site pair with the highest P90 values in each city was generally
highly correlated (rmax>0.70), and in some cases was more highly correlated than the sites with
the largest annual mean differences.
2.4.2 PM10
For the purpose of comparison to PM2 5 and PM10_2 5 concentrations, PM10 data from
2001-2003 are presented in Figures 2-10 and 2-11. Figure 2-10 shows the PM10 annual mean
concentrations and Figure 2-11 shows the concentration-based 24-hour average 'design value'
type metric.19'20 As in the earlier PM2 5 maps, the monitor with the highest value in each
monitored county is used to represent the value in each county. Most areas of the country had
concentrations below the level of the annual PM10 standard of 50 |ig/m3. Exceptions include six
counties in central and southern California. Most areas of the country also had concentrations
below the level of the 24-hour standard of 150 |ig/m3, with exceptions concentrated in the
southwestern U.S. and isolated counties scattered across the east.
EPA recently examined national and regional PM10 trends from 1988 to 2003 (EPA,
2004b, p. 13). The EPA found a national average decline in annual average concentrations of
approximately 31 percent over the 16-year period, with regional average declines ranging from
16 to 3 9 percent.
19 These figures do not depict officially designated PM10 nonattainment areas. As of January 1, 2005, there
were a total of 58 areas classified as moderate or serious nonattainment areas, mostly in the western U.S. See
designated nonattainment areas at www.epa.gov/oar/oaqps/greenbk/pnc.html. Further, note that these maps (like the
other PM ones in this Chapter) do not exclude event-flagged data (natural or exceptional). Data flagged for events
are sometimes excluded from regulatory design value calculations.
20 The form of the 1987 PM10 24-hour standard is based on the number of exceedances; the metric used for
this map, "concentration-based 24-hour 'design value' type metric" is almost always calculated to be 150 ng/m3 or
higher when the monitoring site violates the explicit exceedance-based NAAQS. Utilization of the concentration-
based metric permits delineation of gradients and facilitates comparisons with PM25 and PM10_25.
2-30
-------
PM10 Concentration (|ig/m3) [
585 counties
D <= 20 I I 20 < x <= 30
Figure 2-10. County-level maximum PM10 annual mean concentrations, 2001-2003.
Source: Schmidt et al. (2005)
2-31
-------
PM10 Concentration (|ig/m3)
585 counties
^^^ I i 50 < x <=75
] <= 50
Figure 2-11. County-level maximum 24-hour PM10 'design value' concentrations, 2001-2003.
Source: Schmidt et al. (2005)
2-32
-------
2.4.3 PM1025
PM10_2 5 is a measure of the coarse-mode fraction of PM10 being considered in this review.
It can be directly measured by a dichotomous sampler, or by using a difference method with
collocated PM10 and PM2 5 monitors. For the latter, collocated PM10 and PM2 5 monitors using
identical inlets, sampling flow rates, and analysis protocols produce the most precise results. A
nationwide network of samplers with the specific intent to consistently and accurately measure
PMio-2.5 does not currently exist. The EPA is currently evaluating a variety of monitoring
platforms, including alternative continuous methods, to permit establishment of reference and
equivalent methods for PM10_2 5. These could be used in the future to design a national network
of monitors to measure coarse-fraction particles. Until such a network is established, estimates
of PM10_25 can be generated for a limited number of locations using a difference method on
same-day data. For this review, PM measurements collected from collocated PM10 and PM2 5
FRM monitors are utilized. Since the protocol for each monitor is not usually identical, the
consistency of these PM10_2 5 measurements is relatively uncertain, and they are referred to as
"estimates" in this Staff Paper.21
The 98th percentile 24-hour average PM10_2 5 concentrations range from about 5 to 208
|ig/m3, with a median of about 28 |ig/m3. The box plots in Figures 2-4 and 2-5 (introduced in
section 2.4.1) depict the regional distribution of site-specific estimated annual mean and 98th
percentile 24-hour average PM10_2 5 concentrations, respectively, by geographic region (excluding
Alaska, Hawaii, Puerto Rico, and the Virgin Islands). Figures 2-12 and 2-13 are national maps
that depict estimated county-level annual mean PM10_2 5 concentrations and 98th percentile 24-
hour average concentrations, respectively. To construct the maps, the site with the highest
concentration in each monitored county is used to represent the value in that county. The annual
mean PM10_2 5 concentrations are generally estimated to be below 40 |ig/m3, with one maximum
value as high as 64 |ig/m3 (see Figure 2-4), and with a median of about 10-11 |ig/m3. Compared
to annual mean PM2 5 concentrations, annual mean PM10_2 5 estimates are more variable, with
more distinct regional differences. As shown in Figure 2-4, eastern U.S. estimated annual mean
PM10_2 5 levels tend to be lower than annual mean PM25 levels, and in the western U.S. estimated
PM10_2 5 levels tend to be higher than PM2 5 levels. The highest estimated annual mean PM10_2 5
concentrations appear in the southwest region and southern California. The estimated 98th
21 Note that the urban PM10_2 5 estimates derived in this review, labeled '2001-2003', actually represent
either the entire 12-quarter period or the most recent consecutive 4- or 8-quarter period (from that 3-year period)
with 11 or more samples each. This technique was used to maximize the number of usable sites (and not introduce
seasonal bias). Of the 489 total sites, 230 had 12 complete quarters, 122 sites had 8 quarters, and 137 had 4. Similar
to PM2 5 and PM10 processing, 'annual' means and 'annual' 98th percentiles were first constructed from 4-quarter
periods, albeit for PM10_2 5, not all necessarily from the same calender year. The 4-quarter statistics were then
averaged together for the 8- and 12-quarter sites. Hence there is some temporal variability intrinsic in 2001-2003
estimates. The 1-, 2-, or 3-year averages of the 'annual' statistics are subsequently referred to simply as 'annual
means' or '98thpercentiles'.
-------
PM10.2 5 Concentration (|ug/m3) ] x <= 10 i CX 10 < x <= 15
351 counties 15 < x <= 25 ^^m x > 25
Figure 2-12. Estimated county-level maximum annual mean PMi0_25 concentrations, 2001-2003.
Source: Schmidt et al. (2005) 2-34
-------
PM10.2 5 Concentration (|ug/m3) ] <= 25 I I 25 < x <= 45
351 counties ^™ 45 < x <= 75 ^H x > 75
Figure 2-13. Estimated county-level maximum 98th percentile 24-hour average PM10_25 concentrations,
2001-2003.
Source: Schmidt et al. (2005) 2-35
-------
percentile 24-hour average PM10_2 5 concentrations are generally highest in the southwest,
southern California, and upper midwest, where a few sites have estimated concentrations well
above 100 |ig/m3 (see Figure 2-5). As noted before, these maps include days that were flagged
for natural or exceptional episodic events. Episodic events can affect PM10_2 5 98th percentiles
even more than for PM25. An evaluation of 2001-2003 PM10_2 5 data found that such events
caused 98th percentile values to be elevated by an average of 2.5 |ig/m3 (Schmidt et al., 2005).
The IMPROVE monitoring network generally provides long-term PM10_2 5 trends for rural
areas. Figure 2-14 presents the composite long-term trend at 7 eastern sites, 17 western sites,
and one urban site in Washington, D.C. At the eastern rural sites, measured PM10_2 5 in 2003 was
about 33 percent lower then the corresponding value in 1993. At the western rural sites,
measured PM10_25 was about 17 percent higher in 2003 than the corresponding value in 1993. At
the Washington, D.C. site, the annual average PM10_2 5 concentration in 2003 was about 25
percent lower than the 10-year peak in 1994, but nearly 2 |ig/m3 (over 40%) higher than the 1998
low point.
The long-term PM10_2 5 levels in the relatively remote non-urban IMPROVE sites shown
in Figure 2-14 are notably lower than those found in most urban areas. While PM10_2 5
concentrations in rural areas affected by sources such as windblown dry lake beds, unpaved
roads, or agricultural activities can be quite high, comparison of paired urban and nearby rural
sites suggest that PM10_2 5 levels are generally higher in urban areas. Figure 2-15 shows urban
and corresponding rural PM10_2 5 concentrations for several large metropolitan areas in the eastern
and western U.S.. The urban data represent inter-city or suburban monitoring sites located in
densely populated regions of the metro areas, and the rural data typically represent one or more
sites situated on the outskirts of the areas where population density is low. In all the metro areas
shown, the urban PM10_2 5 concentrations exceed those in the nearby rural locations.
The CD contains an analysis of 1999-2001 PM10_25 estimates in 17 MS As that is useful
for assessing the spatial homogeneity of PM10_25 across the urban areas (CD, Appendix 3 A). A
similar analysis, for 21 urban areas defined as either Core Based Statistical Areas (CBSAs) or
Combined Statistical Areas (CSAs), was conducted on PM10_25 estimates from 2001-2003
(Schmidt et al., 2005). These analyses are similar to the 49-city analysis for PM2 5 discussed in
section 2.4.1 and summarized earlier in Table 2-3. However, since there were fewer site
pairings, fewer urban areas covered, and because of higher uncertainty in daily concentration
estimates, the PM10_2 5 results are not as robust as the PM2 5 results. The PM10_2 5 analysis is
summarized in Table 2-4. The analysis reveals generally lower correlations for PM10_2 5
compared to the PM25 correlations in the same city. Of the 200 monitor pairs analyzed, only 17
(9%) had correlation coefficients greater than or equal to 0.80, in contrast to around 83 percent
(1901 of 2290) of the pairs for PM 25.
The difference in estimated annual mean PM10_25 between site pairs in the 21 areas also
covered a greater range than was seen for PM2 5, with differences up to almost 31 |ig/m3 in Los
2-36
-------
8
7
CO
5
4
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
-•- 7 Eastern Sites o 17 Western Sites -*• Washington, D.C. Site
Figure 2-14. Average measured annual average PM1025 concentration trend at IMPROVE
sites, 1993-2003.
Source: Schmidt et al. (2005)
2-37
-------
ou
7n
^C" en
j3 ou
"op
5 *;n
a
o
•-C3 jn
a 4U
gn
u on
a ou
o
o on
£u
•in
IU
S
Vi-i
on
OK
£O
^g
oo £U
o -IK
.x ID
1
S in
g 1U
o
O c
O
-S
N=
,c
?
i
^
/
\*'
/
107
1 £
West
5
West
9
2
TTM
-s
s
I, „
^
f
^
?
3/2
Estimated 98th percentile 24-hour average PM™ _2.5 concentrations
East
1 Urban (average of urban sites)
! !
^H
1
1
' / / -" / ^ / /
Estimated annual mean PMi02.5 concentrations
East
EH Urban (average of urban sites)
EEI Rural (averag of rural sites)
| |
' / / / / *** / /
7/2 5/2 3/1 2/1 1/1 12/1 8/1
average PM10.25 concentrations shown in top panel and estimated annual mean PM10.25 concentrations shown in
bottom panel. Urban bar (left) is average of urban sites in area, rural bar (right) is average of nearby rural sites.
Urban / rural designation from AQS. N= number of sites (urban / rural).
Schmidt et al. (2005)
2-38
-------
Table 2-4. Summary of Estimated PMiQ-2.5 Analyses in 21 Metropolitan Areas, 2001-2003
Area*
Anchorage, AK
Birmingham-Hoover-Cullman, AL
Cleveland- Akron-Elyria, OH
Denver- Aurora-Boulder, CO
Detroit- Warren-Flint, MI
El Paso, TX
Las Vegas-Paradise-Pahrump, NV
Los Angeles-Long Beach-Riverside, CA
Miami-Fort Lauderdale-Miami Beach, FL
Minneapolis-St. Paul-St. Cloud, MN-WI
New York-Newark-Bridgeport, NY-NJ-CT-PA
Orlando-The Villages, FL
Philadelphia-Camden-Vineland, PA-NJ-DE-MD
Pittsburgh-New Castle, PA
Sacramento— Arden- Arcade— Truckee, CA-NV
Salt Lake City-Ogden-Clearfield, UT
San Jose-San Francisco-Oakland, CA
San Juan-Caguas-Fajardo, PR
Virginia Beach-Norfolk-Newport News, VA-NC
Weirton-Steubenville, WV-OH
Wichita- Winfield, KS
N Sites
3
5
8
3
3
4
5
11
4
3
5
3
3
6
3
3
7
3
3
4
3
3-vear Average Annual Mean
Levels Cue
Area
Avg
14.8
7.0
11.6
15.5
15.3
23.2
23.2
21.6
10.2
19.1
8.7
9.5
5.5
6.4
10.4
17.9
10.8
24.4
4.2
12.4
11.9
Max
Site
23.7
9.0
16.3
22.1
18.7
28.3
33.3
44.5
15.3
23.6
22.3
10.2
6.4
8.5
12.0
24.1
13.4
30.2
4.5
13.8
13.7
\n}
Min
Site
9.6
5.6
5.6
7.7
8.8
13.9
9.0
13.7
8.4
15.5
2.9
8.5
4.3
3.5
8.2
14.4
7.8
18.0
4.0
10.7
10.3
Fercent uinerence
Largest
diff., any
site versus
Area Avg
38%
22%
52%
50%
42%
40%
61%
51%
33%
19%
67%
11%
22%
45%
21%
26%
28%
26%
7%
14%
13%
Max site
versus Min
site
59%
38%
66%
65%
53%
51%
73%
69%
45%
34%
87%
17%
33%
59%
32%
40%
42%
40%
11%
22%
25%
r
(Max site
versus Min
site)
0.13
0.76
0.55
0.54
0.60
0.89
0.65
0.38
0.63
0.62
0.21
0.71
0.48
0.67
0.38
0.72
0.69
0.64
0.54
0.53
0.81
24-Hour Pon (ug/m3)**
Max
Pair
52.3
10.0
26.0
29.3
30.5
31.0
40.0
57.5
14.0
23.0
35.3
6.0
10.0
13.0
17.5
24.0
13.5
22.0
5.0
15.0
11.0
Min
Pair
22.5
3.0
8.0
14.5
25.0
15.0
17.0
8.5
3.0
19.5
6.5
4.0
6.0
5.0
6.5
9.0
4.5
17.0
3.0
11.5
5.0
r
(Max
Pair)
0.13
0.55
0.64
0.54
0.32
0.92
0.65
0.03
0.63
0.38
0.21
0.71
0.48
0.46
0.25
0.72
0.53
0.64
0.54
0.43
0.69
* 'Area' is the larger of a Combined Statistical Area (CSA) or a Core Based Statistical Area (CBSA). See http://www.whitehouse.gov/omb/bulletins/fy05/b05-02.html.
** 'P9o' is the 90th percentile of the distribution of differences in 24-hour averages between two sites in the same urban area.
2-39
-------
Angeles, CA. The largest percent difference between any site's annual mean and it's
corresponding area spatial average ranged from 7 to 67 percent with a median of 28 percent,
which is about double the median from the corresponding PM25 analyses. Of the 18 common
metropolitan areas analyzed for both PM2 5 and PM10_2 5, only 2 areas (Sacramento, CA and San
Juan, PR) had higher values for this indicator (largest percent difference for any site in area
verus area spatial average) for PM2 5 compared to PM10_2 5. The P90 values (described in section
2.4.1) for the 200 PM10_25 site pairs ranged from about 3 |ig/m3 to about 58 |ig/m3, which is wider
than the range of about 1 to 45 |ig/m3 observed for PM25.
These analyses indicate that spatial distribution of PM10_25 is more heterogeneous than
PM2 5 in many locations but may be similar in other areas. Any conclusions should be tempered
by the inherent uncertainty in the PM10_25 estimation method (discussed at the beginning of this
section), and the relatively small sample size for PM10_2 5 relative to PM2 5.
2.4.4 Ultrafine Particles
There are no nationwide monitoring networks for ultrafine particles (i.e., those with
diameters < 0.1 |im), and only a few recently published studies of ultrafine particle counts in the
U.S. At an urban site in Atlanta, GA, particles in three size classes were measured on a
continuous basis between August 1998 and August 1999 (CD, p. 2B-21). The classes included
ultrafine particles in two size ranges, 0.003 to 0.01 |im and 0.01 to 0.1 jim, and a subset of
accumulation-mode particles in the range of 0.1 to 2 jim. In Atlanta, the vast majority (89
percent) of the number of particles were in the ultrafine mode (smaller than 0.1 jim), but 83
percent of the particle volume was in the subset of accumulation-mode particles. The
researchers found that for particles with diameters up to 2 jim, there was little evidence of any
correlation between number concentration and either volume or surface area. Similarly poor
correlations between PM25 mass and number of ultrafine particles were confirmed for sites in
Los Angeles and nearby Riverside, CA (Kim et al., 2002). This suggests that PM2 5 cannot be
used as a surrogate for ultrafine mass or number, so ultrafine particles need to be measured
independently.
Studies of near-roadway particle number and size distributions have shown sharp
gradients in ultrafine concentrations around Los Angeles roadways (CD, p. 2-35 to 2-36).
Ultrafine PM concentrations were found to decrease exponentially with distance from the
roadway source, and were equal to the upwind "background" location at 300 m downwind.
2.4.5 Components of PM
Atmospheric PM is comprised of many different chemical components that vary by
location, time of day, and time of year. Further, as discussed in section 2.2, fine and coarse
particles have fundamentally different sources and composition. Recent data from the rural
IMPROVE network and from the EPA urban speciation network provide indications of regional
2-40
-------
composition differences for fine particles. Although both programs provide detailed estimates of
specific PM chemical components (individual metals, ions, etc.), only gross-level speciation
breakouts are shown here. Figure 2-16 shows urban and rural 2003 annual average PM25 mass
apportionment among chemical components averaged over several sites within each of the U.S.
regions. In general:
• While PM2 5 mass and all component concentrations are higher in urban areas than
in IMPROVE sites, in general, nitrates and carbonaceous components appear to
have a greater urban/rural enhancement as compared to sulfates.
• PM25 in the eastern U.S. regions is dominated by sulfates and carbonaceous mass.
PM2 5 in the western U.S. urban sites has a greater proportion of carbonaceous
mass.
Trends concentrations of fine particle components from the IMPROVE network from
1993 to 2003 are shown in Figure 2-9 for rural areas and for urban Washington, D.C. (section
2.4.1 above). The top two panels of this figure aggregate rural IMPROVE sites in the eastern
and western U.S. The bottom panel shows a comparable period for the Washington, D.C. urban
IMPROVE site. Consistent with more recent data in Figure 2-15, levels of rural annual average
PM2 5 mass are significantly higher in the east than in the west, but are trending downward in
both regions. Annual levels of sulfates have decreased the most (and contributed the most to the
reductions in PM2 5 mass) both in eastern and western rural areas. At the Washington, D.C.,
IMPROVE site, mass has decreased 31 percent from 1993-2003. Total carbonaceous mass (34
percent reduction) and sulfates (down 29 percent) are the biggest contributors to the mass
reduction over the past 10 years. Both total carbonaceous mass and sulfates dropped
significantly at this site in 1995, but have not shown significant improvements since then. All
other components in all areas have shown small changes over the 10-year period.
Though most of the speciation data currently available are for PM25, there are a limited
amount of recent data available on speciation profiles for the coarse fraction, and still less for
ultrafine particles. The EPA "Supersite" program addresses a number of scientific issues
associated with PM.22 A Supersite location in the Los Angeles metropolitan area (USC site)
provides a unique comparison of the composition of ultrafine, fine, and coarse particles (Sardar
et al., 2005). Based on the reported measurement data, ultrafine, fine, and coarse PM have
distinctly different compositions at this site (Figure 2-17). Increasing in size from ultrafine to
fine to coarse, the relative fraction of organic carbon (dominant for ultrafine) drops, and the
crustal element portion goes from a minor component (ultrafine, fine) to the dominant fraction
22 More information can be found at http://www.epa.gov/ttn/amtic/supersites.html.
2-41
-------
CO
o
W— •
03
(U
o
c
o
o
LO
c\i
tn
(0
re
a)
a)
T3
(0
3
T3
C
a)
a)
a.
a
to
a)
+j
3
o
(O
a)
(0
E
(0
O
c
a)
^
+j
o
(O
Figure 2-16. Annual average composition of PM25 by region, 2003. Rural data
(top panel) from IMPROVE network, urban data (bottom panel) from
EPA Speciation Network. Components (from top to bottom) are
crustal material, total carbonaceous mass (TCM), ammonium
nitrate, and ammonium sulfate.
Source: Schmidt et al. (2005)
2-42
-------
PMio-2.5 ('coarse')
0.7%
12.9%^
5.8%
59.5%
Nitrate
Sulfate
Crustal Elements
0c
EC
PM25('fme')
4.4%
30.0%
13.0%
22.5%
Nitrate
Sulfate
Crustal Elements
oc
EC
PM0_j ('ultrafme')
5.7%
13.0%
8.9%
71.2%
Nitrate
Sulfate
Crustal Elements
0c
EC
Figure 2-17. Average PM10_25, PM25, and PM01 (ultrafine) chemical composition at the
USC EPA 'supersite' monitor in Los Angeles, CA, 10/2001 to 9/2002.
components represent measured ions, carbon and crustal elements including
trace metals and are shown in clockwise order (starting with nitrate) as listed
in legend from top to bottom.
Source: Sardaretal. (2005)
2-43
-------
(coarse). The ultrafine results are consistent with other work in Southern California (CD, p.
3-39). The large crustal fraction in the coarse mode is typical of earlier work on western sites
reported in the 1996 Criteria Document (EPA, 1996a, p. 6-165 to 6-167, Figure 6-85a-c), as well
as for more recent work in Phoenix (CD, p. 3-36).
Other recent work on ambient PM10_2 5 particle composition comes from the SouthEastern
Aerosol Research and Characterization (SEARCH) Study.23 This study examined two urban
sites (Birmingham, AL and Atlanta, GA) and nearby rural sites in the southeast. Figure 2-18
presents the results of this work together with the Los Angeles results. In this graphic, the
measured chemical components are presented in terms of their estimated coarse particle mass, as
derived from the reported measurement data.24
Although the scope of these results are limited, staff notes the following:
Consistent with the mass-based comparisons in Section 2.4.3, the western site has
more coarse mass than any of the 4 eastern locations, and the urban
concentrations are clearly higher than nearby non-urban sites.
• The larger absolute and relative crustal and nitrate contributions in LA appear to
be the main source of the higher mass. In contrast, the carbonaceous fraction is
more significant at the eastern sites. While this may be due in part to a greater
contribution of biological materials in the southeast, such materials would not
explain the larger elemental carbon contribution, particularly in Birmingham.
• The higher urban concentrations of PM10_2 5 in the southeast appear to be due to
higher crustal and carbonaceous components than are found in nearby rural areas,
suggesting urban sources make a substantial contribution to both components.
These recent studies have focused more on the indicators of the major categories of
coarse particles - crustal, carbonaceous, and inorganic anions, and less so on trace elements and
specific organic constituents. The CD notes that the concentrations of a number of trace
elements in the coarse fraction can be comparable or higher than that for fine particles (e.g. Cr,
Ni Zn, Pb, Cu), while the crustal elements (Al, Si, K, Ca, Fe) are, of course, much higher (CD,
p3-37-38). While urban sources apparently increase total crustal materials, the relative
proportions of some crustal elements may be enriched by urban sources relative to the proportion
23 See http://www.atmospheric-research.com/ for information on SEARCH.
24 Inorganic nitrate and sulfate concentrations were assumed to be solely associated with their ammonium
salts, the crustal component reflect the measured elements plus their common oxides and organic carbon mass was
estimated by multiplying measured organic carbon by a factor of 2.5 to account for the mass of H, O, and other
elements in the coarse particle organic compounds. For the SEARCH sites, the total carbonaceous mass is estimated
as the difference between measured coarse particle mass and its inorganic constituents. The OC-EC split is derived
from a special carbon measurement study during 2000, 2001, 2003 and 2004.
2-44
-------
o
1)
u
o
U
o
o
U
D unknown
• EC
DOOM
• Crustal
D Sulfates
• Nitrates
use
BHM CTR
ATL YRK
100%
80%
60% -
40%
20%
D unknown
• EC
DOOM
• Crustal
D Sulfates
• Nitrates
use
BHM CTR
ATL YRK
Figure 2-18. Average PMio-i.s composition for Los Angeles and two eastern urban-
rural pairs. Based on USC Supersite data (10/2002 to 9/2003), and
Birmingham, AL (BHM, urban), Centerville, AL (CTR, rural), Atlanta, GA
(ATL, urban) and Yorkville, GA (YRK, rural) monitoring sites in the
Southeastern Aerosol Research and Characterization (SEARCH) Study, 4/2003-
12/2003. The top panel shows mass concentration in |j,g/m3 and the bottom panel
shows composition as percent of measured mass.
Source: USC site data (Sardar et al., 2005); eastern data (SEARCH website) adjusted as described in
Schmidt et al. (2005)
2-45
-------
found in soils. For example, urban industrial (e.g. steel) and automobiles (rust) can increase the
relative amount of iron in urban coarse particles.
The CD review (Appendix 3C) lists no recent studies that speciated any substantial
portion of organic components of PM10_2 5, but some inferences can be drawn from analyses of
the composition of road dusts. The CD reports on two California studies that found organic
substances consistent with particles of biologic origin, tire and brake wear, asphalt, and
combustion in fine fraction samples of resuspended road dust particles (CD, p. 3D-3 to 5). One
of these (Rogge et al., 1993) suggests that the action of automobile traffic on leaves and other
vegetative debris on roads may serve to elevate their atmospheric concentrations and decrease
particle size as compared what might be found in more natural settings. The findings regarding
road dust as well as trace elements are buttressed by a very recent report of similar work
comparing urban and rural road dust in and near Pittsburgh (Robinson et al., 2005). These
authors found that most of the over 100 organic species examined were "significantly enriched"
in urban as compared to rural resuspended road dust samples. Marker substances suggested both
vegetative debris and non-biological sources. Comparing trace elements, the authors found that
urban road dust in Pittsburgh was enriched in metals associated with anthropogenic sources,
notably Fe, Zn, Cu, Pb, Cr, Ni, Mo, and Sb. Ca and Mn were more prevalent in the rural road
dust sample.
2.4.6 Relationships Among PM2 5, PM10, and PM10_2 5
In this section, information on the relationships among PM indicators in different regions
is presented based on data from the nationwide PM FRM monitoring networks.25 Figure 2-19
shows the distribution of ratios of annual mean PM25 to PM10 at sites in different geographic
regions for 2001-2003. The ratios are highest in the eastern U.S. regions with median ratios of
about 0.6 to 0.65, and lowest in the Southwest region, with a median ratio near 0.3. These data
are generally consistent with earlier findings reported in the 1996 CD from a more limited set of
sites. Ratios greater than one are an artifact of the uncertainty in the independent PM10 and PM25
measurement methods.
Correlations among pollutant indicators can provide insights into how well one indicator
can represent the variability in another indicator. Figure 2-20 shows the results of a nationwide
analysis of correlations among PM size fractions using 24-hour average data from the FRM
monitoring networks for 2001-2003. PM25 and PM10 measured on the same days at collocated
monitors are fairly well correlated, on average, in the eastern regions, and not as well correlated
in the western regions, particularly in the upper midwest. PM10 is fairly well correlated with
25 In this section's analyses, information was gleaned from the 489 site (4-, 8-, 12-quarter) PM10_25 database
for all 3 sizes in order to get seasonally unbiased estimates of their statistical relationships (i.e., to ensure a minimum
number of data pairs each quarter for 4-, 8-, or 12 quarters).
2-46
-------
O
O
1.25-
1,00
i= = =*=
0.75-
0.50'
o.ool
*
Northeast Southeast
•—#-•
*
'5"
*
Industrial
Midwest
97
Upper
Midwest
41
Southwest
Northwest
Southern
California
37
N = 63 97 97 41 32 108
Figure 2-19. Distribution of ratios of PM2 5 to PM10 by region, 2001-2003. Box depicts interquartile range and
median; whiskers depict 5th and 95th percentiles; asterisks depict minima and maxima. N = number
of sites.
Source: Schmidt et al. (2005)
-------
1 .0-
OQ .
.O
. )
br» n £:
d) U .O
'o
it
fli n c; -
o °'5
O
^ n /i .
.Q
^
fl) n o -
JJ; O.o
O
o
U .z
n -i -
U . 1
o.o-
4
4
n A
A A
4 n n
D
*
• • *
* • •
Northeast Southeast Industrial Upper Southwest Northwest Southern
Midwest Midwest California
• PM25 vs. estimated PM10_25 ^ PM25 vs. PM10 ^ PM10 vs. estimated PM10.25
Figure 2-20. Regional average correlations of 24-hour average PM by size fraction.
Source: Schmidt et al. (2005)
2-48
-------
estimated PM10_2 5 in most regions, with the highest average correlations in the southwest, upper
midwest, and southern California regions. These data suggest that PM10 might be a suitable
indicator for either fine or coarse particles, depending upon location-specific factors. However,
in all locations estimated PM10_2 5 and PM2 5 are very poorly correlated, which should be expected
due to their differences in origin, composition, and atmospheric behavior.
2.5 PM TEMPORAL PATTERNS
2.5.1 PM25 and PM10_25 Patterns
Data from the PM FRM networks from 2001-2003 generally show distinct seasonal
variations in PM2 5 and estimated PM10_2 5 concentrations. Although distinct, the seasonal
fluctuations are generally not as sharp as those seen for ozone concentrations. Figure 2-21
shows the monthly distribution of 24-hour average urban PM25 concentrations in different
geographic regions. The months with peak urban PM2 5 concentrations vary by region. The
urban areas in the northeast, industrial midwest, and upper midwest regions all exhibit peaks in
both the winter and summer months. In the northeast and industrial midwest regions, the
summer peak is slightly more pronounced than the winter peak, and in the upper midwest region
the winter peak is slightly more pronounced than the summer peak. In the southeast, a single
peak period in the summer is evident. In western regions, peaks occur in the late fall and winter
months.
Figure 2-22 shows the distributions of estimated 24-hour average urban PM10_25
concentrations by U.S. geographic region. The lowest concentrations generally occur in the
winter months. Elevated levels are apparent in the easternmost regions in April. In the upper
midwest, northwest, and southern California regions, the highest levels occur in the mid- to late-
summer to mid-fall. The southwest region exhibits the greatest range of variability throughout
the year. Elevated levels are apparent in the spring, consistent with winds that contribute to
windblown dust. In the southwest and southern California, highly elevated levels in the fall,
especially October, were caused by forest fires in the vicinity of the monitoring sites.
The chemical components of fine particles also exhibit seasonal patterns. Figures 2-23
and 2-24 show seasonal 2003 urban and rural patterns for each of the U.S. regions. Seasonal
patterns are shown by calendar quarter. In general:
PM2 5 values in the east are typically higher in the third calendar quarter (July-
September) when sulfates are more readily formed from SO2 emissions from
power plants predominantly located there and sulfate formation is supported by
increased photochemical activity.
• Urban PM2 5 values tend to be higher in the first (January-March) and fourth
(October-December) calendar quarters in many areas of the western U.S., in part
because more carbon is produced when woodstoves and fireplaces are used and
2-49
-------
Northeast
Figure 2-21. Urban 24-hour average
PM2 5 concentration distributions by
region and month, 2001-2003. Box
depicts interquartile range and median; line
connects monthly means. Counts above
boxes indicate number of 24-hour
observations
Source: Schmidt et al. (2005)
Southeast
w>
8
>-> 15
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Upper Midwest
E?
-------
Northeast
Figure 2-22. Urban 24-hour average
PM10_2 5 concentration distributions by
region and month, 2001-2003. Box
depicts interquartile range and median;
line connects monthly means. Counts
above boxes indicate number of 24-hour
observations.
Source: Schmidt et al. (2005)
Southeast
U
0 15
1714 1564
50 -|
f -
=L 40 -
bo ;
s
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Southwest
483 519 481
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Southern California
$
O 15 -
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
2-51
-------
W)
gp
5-H
25
20
15
10
10
TCM
I Amm.
Mitrate
Amm.
Sulfate
s
•#"
^
Figure 2-23. Seasonal (calendar quarter) average composition of urban PM25 by
region, 2003. Data from EPA Speciation Network. Components (from
top to bottom) are crustal material, total carbonaceous mass (TCM),
ammonium nitrate, and ammonium sulfate.
Source: Schmidt et al. (2005)
2-52
-------
W)
gp
5-H
12
10
8
6
4
2
0
4th Quarter
*£
M^
Amm.
Sulfate
Figure 2-24. Seasonal (calendar quarter) average composition of rural PM2 5 by
region, 2003. Data from IMPROVE Network. Components (from top to
bottom) are crustal material, total carbonaceous mass (TCM), ammonium
nitrate, and ammonium sulfate.
Source: Schmidt et al. (2005)
2-53
-------
particulate nitrates are more readily formed in cooler weather. In addition, the
effective mixing depth is restricted due to enhanced thermal stability in the
planetary boundary layer during the cooler seasons.
Urban concentrations of PM25 are seen to be generally higher than rural
concentrations in all four quarters, though in the west the difference seems to be
greatest in the cooler months.
The relationship between the annual mean at a site and the shorter-term 24-hour average
peaks is useful for examining the relationships between short- and long-term air quality
standards. The box plots in Figures 2-25 and 2-26 show the relationships for PM2 5 and
estimated PM10_2 5, respectively, between annual mean PM concentrations and peak daily
concentrations as represented by the 98th percentile of the distribution of daily average
concentrations at FRM sites across the U.S. Although there is a clear monotonic relationship
between 98th percentiles and annual means, there is considerable variability in peak daily values
for sites with similar annual means. For annual mean PM25 values between 10 and 15 |ig/m3, the
interquartile range of 98th percentiles spans about 5 to 6 |ig/m3 for each 1 |ig/m3 interval. The
range between the 5th and 95th percentile values for each interval varies substantially. For all
sites with an annual mean less than or equal 15 |ig/m3, the corresponding 98th percentile value is
less than 65 |ig/m3. Estimated PM10_2 5 generally exhibits greater variability in 98th percentile
values for sites with similar annual means than seen for PM2 5 The maximum estimated PM10_2 5
values are quite high relative to the rest of the distribution for annual mean intervals above
20 |ig/m3.
Staff evaluated speciated PM2 5 data for 2003 from the urban EPA network in order to
compare the component profiles on high PM2 5 mass days to annual average profiles (Schmidt et
al., 2005). Table 2-5 shows the analysis results for 8 different sites in large metropolitan areas
(in the east: Birmingham, AL; Atlanta, GA; New York City, NY; Cleveland, OH; Chicago, IL;
and St. Louis, MO; in the west: Salt Lake City, UT; and Fresno, CA). Mass is proportioned
into four categories: sulfates, nitrates, crustal, and total carbonaceous mass (TCM, the sum of EC
and OCM). For each site, the table shows the 2003 annual average speciation pattern, the profile
for the five highest PM2.5 mass days in that year — both individually and averaged together —
and corresponding FRM mass values (annual average, five highest days, and average of five
highest). The table shows some notable differences in the percentage contribution of each of the
species to total mass when looking at the high end of the distribution versus the annual average.
In all of the eastern city sites, the percentage of sulfates is somewhat higher on the five high days
as compared to the annual averages. In the two western cities, the percentage of nitrates is
higher on the five high days as compared to the annual averages. TCM appears somewhat lower
percentage on the five high days compared to the annual averages in most cities. It is of note
2-54
-------
CO
E
(U
(U
Q_
_c
-§->
00
CO
80:
70 —
50:-
30:-
20--
Annual mean <6 6-8 8-9 9-10 10-11 11-12 12-13 13-14 14-15 15-16 16-17 >17
(ng/m3) N= 24 64 38 57 7g 88 1Q1 12Q 1Q5 65 4Q 46
Figure 2-25. Distribution of annual mean vs. 98th percentile 24-hour average PM25 concentrations, 2001-2003.
Box depicts interquartile range and median; whiskers depict 5th and 95th percentiles; asterisks depict
minima and maxima. N= number of sites.
Source: Schmidt et al. (2005)
£*~3 3
-------
max=152
140
max=208 95th=152
1 30
Annualmean<6 6-8 8-10 10-12 12-14 14-16 16-20 20-30 >30
^9m N= 84 66 83 66 58 37 49 32 14
Figure 2-26. Distribution of estimated annual mean vs. 98th percentile 24-hour average PM10_2 5
concentrations, 2001-2003. Box depicts interquartile range and median; whiskers depict 5th and
95th percentiles; asterisks depict minima and maxima. N= number of sites.
Source: Schmidt et al. (2005)
2-56
-------
Table 2-5. PM2.5 composition on high mass days in select urban areas, 2003
Urban Area
Birmingham, AL
Atlanta, GA
New York City,
NY
Cleveland, OH
Chicago, IL
St. Louis, MO
Salt Lake City,
UT
Fresno, CA
Statistic*
• Annual average
• Average of 5 highest PM25 mass days
• Hrghest PM2 , mass day
• 2 highest PM2 5 mass day
• 3r highest PM25 mass day
• 4th Hrghest PM25 mass day
• 5 Hrshest PM, , mass dav
• Annual average
• Average of 5 highest PM2|5 mass days
• Highest PM2 5 mass day
• 2 highest PM2, mass day
• 3 highest PM2 , mass day
• 4th Hrghest PM25 mass day
. 5th Hrghest PM,, mass day
• Annual average
• Average of 5 highest PM2 , mass days
• Highest PM2|5 mass day
• T highest PM25 mass day
• 3 highest PM2, mass day
• 4 Hrghest PM2 5 mass day
. 5th Hrehest PM,, mass dav
• Annual average
• Average of 5 highest PM2, mass days
• Hrghest PM2 , mass day
• T highest PM25 mass day
• 3r highest PM25 mass day
• 4 Hrghest PM2, mass day
• 5 Hrghest PM, , mass day
• Annual average
• Average of 5 highest PM25 mass days
• Hrghest PM2 , mass day
• 2 highest PM2 5 mass day
• 3r highest PM25 mass day
• 4th Hrghest PM25 mass day
• 5 Hrshest PM, , mass dav
• Annual average
• Average of 5 highest PM2|5 mass days
• Highest PM2 5 mass day
• 2 highest PM2, mass day
• 3 highest PM2 , mass day
• 4th Hrghest PM25 mass day
. 5th Hrghest PM,, mass day
• Annual average
• Average of 5 highest PM2 , mass days
• Highest PM2|5 mass day
• T highest PM25 mass day
• 3rd highest PM2 5 mass day
• 4 Hrghest PM2 , mass day
• 5th Highest PM, , mass day
• Annual average
• Average of 5 highest PM2|5 mass days
• Hrghest PM2 , mass day
• 2" highest PM2 , mass day
• 3r highest PM25 mass day
• 4th Hrghest PM25 mass day
• 5 Hrghest PM, , mass day
Composition Percents (%)
Amm.
Nitrate
8.5
3.8
1.9
4.2
15.3
2.7
2.6
8.1
2.6
2.0
2.0
2.4
3.2
3.6
20.2
11.6
3.6
5.0
27.8
5.1
9.7
22.3
21.4
32.7
25.1
4.8
8.8
31.4
28.0
41.2
46.0
49.2
51.8
5.6
47.8
20.0
12.2
6.2
5.0
6.4
5.0
40.2
28.3
46.3
50.6
43.5
42.4
48.2
45.4
35.5
42.4
55.2
58.4
17.5
35.1
44.6
Amm.
Sulfate
35.6
40.0
55.1
26.9
15.7
51.1
34.6
42.8
60.1
70.5
47.8
67.6
50.8
67.5
38.3
57.9
58.3
69.0
42.1
59.4
62.2
38.3
42.5
43.2
41.5
64.4
37.5
20.5
31.8
34.0
30.7
36.4
27.7
61.7
16.1
36.0
61.9
69.1
67.0
69.2
58.9
42.3
12.2
10.8
6.3
11.9
13.5
5.9
20.2
10.2
4.7
4.6
8.5
1.5
5.3
3.7
Crustal
7.6
7.8
5.5
11.0
10.7
7.4
6.4
4.0
2.3
1.9
2.5
2.1
2.9
1.9
5.1
3.0
5.5
1.4
3.1
4.6
2.0
7.4
6.3
2.3
4.0
8.7
14.7
4.0
4.6
2.3
1.2
0.8
1.2
3.8
5.3
5.6
3.9
3.6
2.0
3.2
8.2
2.7
8.5
2.9
2.5
2.6
3.7
4.7
1.5
3.6
1.3
2.1
0.9
1.3
1.0
1.3
TCM
48.3
48.3
37.4
57.9
58.4
38.7
56.3
45.0
34.3
25.6
47.8
27.9
43.1
27.0
36.4
27.4
32.6
24.6
27.0
30.9
26.1
32.1
30.0
21.7
29.3
22.1
39.0
44.0
35.6
22.4
22.1
13.6
19.3
28.9
30.8
38.4
22.0
21.0
26.0
21.3
28.1
14.7
51.1
40.0
40.5
42.0
40.4
41.3
32.8
50.7
51.6
38.2
32.2
79.7
58.6
50.3
PM2.5
mass**
(Hg/m3)
17.9
40.7
46.6
40.4
39.2
39.1
38.3
15.2
35.2
37.8
37.1
36.8
35.0
29.3
13.1
40.5
45.9
45.8
38.2
36.4
36.0
17.6
44.1
57.9
46.4
45.5
35.7
35.0
15.2
34.4
38.3
35.3
35.1
32.5
30.7
14.5
35.9
50.6
36.0
33.1
30.8
28.9
10.0
40.6
59.5
52.1
34.2
28.7
28.4
18.0
54.2
59.0
56.3
54.4
52.6
50.0
Annual
average
CD
\
/
/''"T^x
I L— "A
\^f j
^~~~^/
/^ "X
/ \
.^r~\
~^r /
^^__^/
/H^X
( \
,/^
\* 7
^^__^y
/"^T~"x
/ \
//""i
V A )
4^_^y
\
1
\ \ /
^~^L^^
/^~^^^
( \ ]
\ 1^7
^^J^^
Average of 5
highest days
(T>
viy
7\
\
j
^^^^/
./^ ^^\
/ v\
^
\ /
\ /
v^^ ./
/^ ^,
/ ^^
1 ^^^i
"^r I
~-~^^^y
/^~~\~~\
/ \
\====^\
\ \ J
-~~~—^
/^~~\^-\
1 ^r\
r^=^ j
\ J
\^^^^^/
/^ r^x
/ \
/•
\ // ^1 /
\^^ \ ./
/ \
/ \
I
\ \\\/
VJJ^
* The 5 highest days shown (and aggregated) for each site actually represent the 5 highest days (based on
collocated FRM mass; see next bullet) that the speciation monitor sampled. FRM monitors at different
locations in the metropolitan area and/or collocated FRM measurements on days that the speciation sampler did
not record valid data may have had higher values than some or all of the 5 high values shown. Event-flagged
data were omitted from this analyses. TCM
** 'PM25 mass' concentration represents the collocated (w/ speciation monitor) same-day FRM measurement
unless not available, in which case the speciation monitor gravimetric mass was substituted. i^-
2-57
-------
that event-flagged data were excluded from this analyses; the carbonaceous fraction of mass
would be significantly higher on sites where peak days are affected by smoke from wildfires.
Monitors that provide near-continuous measurements can provide insights into short-term
(e.g., hourly average) patterns in PM, which could be important to understanding associations
between elevated PM levels and adverse health and welfare effects. Examples of average hourly
profiles for PM25 and PM10_2 5 from 2001-2003 are shown in Figures 2-27 and 2-28 for a
monitoring site in the Greensboro, NC, metropolitan area. As with most eastern urban sites, the
PM2 5 concentrations are significantly higher than those for PM10_2 5. Profiles, for both PM2 5 and
PM10_2 5, in Figure 2-27 indicate that elevated hourly average levels occurred most often between
the hours of 6:00 am and 9:00 am, corresponding to the typical morning rush of automobile
traffic. An evening peak starting about 5:00 pm is also evident for both size indicators. The
95th percentile concentrations during peak hours can be as high as three to four times the median
level for the same hour. As indicated in Figure 2-28 the lowest seasonal levels for both size
fractions occur in the winter. For PM2 5, the summer concentrations are considerably higher than
the other season. These profiles of hourly average PM25 and PM10_25 levels are typical of many,
but not all, eastern U.S. urban areas.
Figure 2-29 shows hourly average PM2 5 and PM10_2 5 concentrations for a monitoring site
in the Denver metropolitan area from 2001-2003. Like many western U.S. sites, the PM10_25
concentrations are higher than the PM25 levels for all hours of the day. Similar to the eastern
example site, this western one also shows a morning and afternoon rush hour traffic signal.
Some western monitoring sites, located in areas subject to routine episodes of windblown dust,
can have unusual diurnal concentration distributions (e.g., 95th percentile concentrations for
some hours more than ten times the median levels; and hourly means significantly higher than
the medians and even 75th percentiles) (Schmidt et al., 2005). Figure 2-30 highlights how
continuous data can be used to pinpoint an unusual or episodic source, in this case a short but
significant dust storm in El Paso, Texas. On April 26, 2002, this dust storm caused large
increases in both PM2 5 and PM10_2 5 concentrations. As might be expected, the dust had a greater
impact on the PM10_2 5 concentrations than the PM2 5. (Note that the PM10_2 5 scale is about 6 times
as large as the PM2 5 scale.) Hourly PM10_2 5 levels approaching 3000 |ig/m3 were recorded this
day.
The hourly ranges shown in Figures 2-27 and 2-29 suggest that hour-to-hour changes in
PM2 5 concentrations encompass several |ig/m3; however, extreme values for hour-to-hour
variations can be much larger. An analysis of the distribution of increases in hour-to-hour
concentrations at multiple sites across the U.S. for 2001-2003 found site-level median hourly
increases ranging up to 6 |ig/m3 (maximum), with an average median increase of about
1.8 |ig/m3.
2-58
-------
4cr
3CT
20"
c
O
(U
O
c
O
O
3CT
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
PM
10-2.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour
Figure 2-27. Hourly average PM2 5 and PM10 2 5 concentrations at a Greensboro,
NC monitoring site, 2001-2003. Upper panel shows the distribution of
PM2 5 concentrations and the lower panel shows the distribution of PM10_2 5
concentrations. (Box plots of interquartile ranges, means, medians, 5th and
95th percentiles.)
Source: Schmidt et al. (2005)
2-59
-------
40
CO
o
"(0
(U
O
c
o
O
30
20
10
0
40
30
20
10
Spring (Mar-May)
Summer (Jun-Aug)
Fall (Sep-Noy)
Winter (Dec-Feb)
PM
-2.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
PM
10-2.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour
Figure 2-28. Seasonal hourly average PM2 5 and PM10_2 5 concentrations at a
Greensboro, NC monitoring site, 2001-2003. Upper panel shows the
PM2 5 concentrations and the lower panel shows the PM10_2 5
concentrations.
Source: Schmidt et al. (2005)
2-60
-------
CO
E
.0
U— •
03
(U
o
c
o
o
60 -4
50 -
40 -
30-
20-
10 -
PM
604
50 -
40 -
30 -
20-
10-
0 -
2.5
0 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23
PM
10-2.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour
Figure 2-29. Hourly average PM2 5 and PM10_2 5 concentrations at a Denver, CO
monitoring site, 2001-2003. Upper panel shows the distribution of PM2 5
concentrations and the lower panel shows the distribution of PM10_2 5
concentrations. (Box plots of interquartile ranges, means, medians, 5th
and 95th percentiles.)
Source: Schmidt et al. (2005)
2-61
-------
500 -\
o
I
§
c
o
O
3000 -f
PM
2.5
April 26, 2002
April 27, 2002
PM
10-2.5
April 26, 2002
Hour April 27, 2002
Figure 2-30. Hourly PM2 5 and PM10 2 5 concentrations at a El Paso, TX monitoring
site, April 26, 2002-April 27, 2002. Upper panel shows the hourly PM2 5
concentrations and the lower panel shows the hourly PM10_2 5
concentrations. Note the different scales.
Source: Schmidt et al. (2005)
2-62
-------
2.5.2 Ultrafine Patterns
Diurnal or seasonal patterns for ultrafine particles have been studied in relatively few
areas of the U.S. A study done at the most extensively studied urban location in the U.S.,
Atlanta, GA, is discussed in the CD (p.3-32). In this study, (CD, p. 3-32 to 3-33) ultrafine
particle number concentrations were found to be higher in the winter than in the summer.
Concentrations of particles in the range of 0.01 to 0.1 jim were higher at night than during the
daytime, and tended to reach their highest values during the morning period when motor vehicle
traffic is heaviest. Smaller particles in the range of 0.004 to 0.01 jim were elevated during the
peak traffic period, most notably in cooler temperatures, below 50°F. .
2.6 PM BACKGROUND LEVELS
For the purposes of this document, background PM is defined as the distribution of PM
concentrations that would be observed in the U.S. in the absence of anthropogenic (man-made)
emissions of primary PM and precursor emissions (e.g., VOC, NOX, SO2, and NH3) in the U.S.,
Canada, and Mexico. Background levels so defined are referred to policy-relevant background,
since this definition of background facilitates separating pollution levels that can be controlled
by U.S. regulations (or through international agreements with neighboring countries) from levels
that are generally uncontrollable by the U.S.. As defined here, background includes PM from
natural sources in the U.S. and transport of PM from both natural and man-made sources outside
of the U.S. and its neighboring countries.
Section 3.3.3 of the CD discusses annual average background PM levels, and states that
"[estimates of annually averaged PRB concentrations or their range have not changed from the
1996 PM AQCD" (CD, p. 3-105). Annual average background estimates for PM10 range from 4
to 8 jig/m3 in the western U.S. and 5 to 11 i-ig/m3 in the eastern U.S.; for PM2 5, estimates range
from 1 to 4 |ig/m3 in the west and 2 to 5 |ig/m3 in the east. The lower bounds of these ranges are
based on estimates of "natural" background midrange concentrations. The upper bounds are
derived from the multi-year annual averages of the remote monitoring sites in the IMPROVE
network (EPA, 1996a, p. 6-44). Ranges presented in the CD for background PM10_25 levels were
derived from the PM10 and PM2 5 ranges by subtraction, resulting in relatively wide ranges with
mid-point estimates of 3.5 |ig/m3 in the west and 4.5 |ig/m3 in the east (CD, p. 3-83). Since the
IMPROVE data unavoidably reflect some contributions from the effects of anthropogenic
emissions from within the U.S., Canada, and Mexico, as well as background, they likely
overestimate the U.S. background concentrations as defined here.
There is a distinct geographic difference in background levels, with lower levels in the
western U.S. and higher levels in the eastern U.S. The eastern U.S. is estimated to have more
natural organic fine particles and more water associated with hygroscopic fine particles than the
western U.S. due to generally higher humidity levels.
2-63
-------
Background levels of PM vary by geographic location and season, and have a natural
component and an anthropogenic component. The natural background arises from: (1) physical
processes of the atmosphere that entrain coarse particles (e.g., windblown crustal material, sea
salt spray); (2) volcanic eruptions (e.g., sulfates); (3) natural combustion such as wildfires (e.g.,
elemental and organic carbon, and inorganic and organic PM precursors); and (4) biogenic
sources such as vegetation, microorganisms, and wildlife (e.g., organic PM, inorganic and
organic PM precursors). The exact magnitude of the natural portion of background PM for a
given geographic location cannot be precisely determined because it is difficult to distinguish
local sources of PM from the long-range transport of anthropogenic particles and precursors.
PM can be transported long distances from natural or quasi-natural events occurring
outside the continental U.S. (CD, p. 3-82). The occurrence and location of these long-range
transport events are highly variable and their impacts on the U.S. are equally variable. The
contributions to background from sources outside of the U.S., Canada, and Mexico can be
significant on an episodic, but probably not on an annual basis (CD, p. 3-91). Several studies
have focused on identifying the origin, sources, and impacts of recent trans-national transport
events from Canada, Mexico, and extra-continental sources.
• The transport of PM from biomass burning in Central America and southern
Mexico in 1998 has been shown to contribute to elevated PM levels in southern
Texas and throughout the entire central and southeastern United States (CD, p.
3-86).
Wildfires in the boreal forests of northwestern Canada may impact large portions
of the eastern United States. The CD estimates that a July 1995 Canadian wildfire
episode resulted in excess PM2 5 concentrations ranging from 5 |ig/m3 in the
southeast, to nearly 100 |ig/m3 in the northern plains states (CD, p. 3-87).
• Windblown dust from dust storms in the North African Sahara desert has been
observed in satellite images as plumes crossing the Atlantic Ocean and reaching
the southeast coast of the U.S., primarily Florida; North African dust has also
been tracked as far as Illinois and Maine. These events have been estimated to
contribute 6 to 11 |ig/m3 to 24-hour average PM2 5 levels in affected areas during
the events (CD, p. 3-84).
• Dust transport from the deserts of Asia (e.g., Gobi, Taklimakan) across the Pacific
Ocean to the northwestern U.S. also occurs. Husar et al. (2001) report that the
average PM10 level at over 150 reporting stations throughout the northwestern
U.S. was 65 |ig/m3 during an episode in the last week in April 1998, compared to
an average of about 20 |ig/m3 during the rest of April and May (CD, p. 3-84).
Background concentrations of PM25, PM10_25, and PM10 may be conceptually viewed as
comprised of baseline and episodic components. The baseline component is the contribution
2-64
-------
from natural sources within the U.S., Canada, and Mexico and from transport of natural and
anthropogenic sources outside of the U.S., Canada, and Mexico that is reasonably well
characterized by a consistent pattern of daily values each year, although they may vary by region
and season.
In addition to this baseline contribution to background concentrations, a second
component consists of more rare episodic high-concentration events over shorter periods of time
(e.g., days or weeks) both within the U.S., Canada, and Mexico (e.g., volcanic eruptions, large
forest fires) and from outside of the U.S., Canada, and Mexico (e.g., transport related to dust
storms from deserts in North Africa and Asia). Over shorter periods of time (e.g., days or
weeks), the range of background concentrations is much broader than the annual averages.
Specific natural events such as wildfires, volcanic eruptions, and dust storms, both of U.S. and
international origin, can lead to very high levels of PM comparable to, or greater than, those
driven by man-made emissions in polluted urban atmospheres. Because such excursions can be
essentially uncontrollable, EPA has in place policies that can remove consideration of them,
where appropriate, from attainment decisions.26
Disregarding such large and unique events, an estimate of the range of "typical"
background on a daily basis can be obtained from reviewing multi-year data at remote locations.
Estimates of background concentrations for time scales shorter than daily averages are not
feasible at this time, since almost all of the rural measurements of speciated PM are 24-hour
averages. EPA staff have conducted an analysis of daily PM25 measurements from 1990 to 2002
at IMPROVE sites across the U.S., focused on the non-sulfate components of PM25 (Langstaff,
2005). Ambient sulfate concentrations are almost entirely due to anthropogenic sources (with
the exception of sulfates from volcanic eruptions), so while non-sulfate PM25 is partly of
anthropogenic origin, it captures almost all of the background.
Based on regional differences in geography and land use, the U.S. is divided into a
number of regions for estimating regional background levels. The "eastern U.S." region extends
west to include Minnesota, Iowa, Missouri, Arkansas, and Louisiana. The "central west" region
is comprised of states west of the eastern U.S. region and east of Washington, Oregon, and
California. Washington, Oregon, and northern California make up the "north west coast," and
26 There are two policies which allow PM data to be flagged for special consideration due to natural events:
the Exceptional Events Guideline (EPA, 1986) and the PM10 Natural Events Policy (Nichols, 1996). Under these
policies, EPA will exercise its discretion not to designate areas as nonattainment and/or to discount data in
circumstances where an area would attain but for exceedances that result from uncontrollable natural events. Three
categories of natural PM10 events are specified in the natural events policy: volcanic or seismic activity, wildland
fires, and high wind dust events. The exceptional events policy covers natural and other events not expected to recur
at a given location and applies to all criteria pollutants. Categories of events covered in the exceptional events
guidance include, but are not limited to, high winds, volcanic eruptions, forest fires, and high pollen counts. EPA is
drafting further guidance concerning how to handle data affected by natural events related to the PM standards.
2-65
-------
southern California (south of about 40 degrees latitude) makes up the "south west coast"
regions.27
To arrive at estimates of background we use the averaged non-sulfate PM25 values28 at
IMPROVE sites in these regions. The Eastern U.S. region is heavily impacted by anthropogenic
emissions and we selected sites in northern states, which we judge to be affected to a lesser
extent by anthropogenic pollution, to derive estimates of background concentrations, using all
IMPROVE sites in the selected states. In all of the other regions we include all of the
IMPROVE sites. Table 2-6 describes the IMPROVE sites selected to represent these different
regions of the U.S. We recognize that these estimates will likely be biased high, as they include
an anthropogenic component, some sites more than others.
The 99th percentile concentrations at each of these sites were calculated to assess high
values measured at these sites, while avoiding excursions that potentially reflect exceptional
natural events. Standard deviations were also calculated for characterization of the daily
variation of background concentrations. Table 2-7 presents the results of this analysis as means
and ranges of individual site statistics within each of the background regions.
Table 2-6. IMPROVE sites selected for estimates of regional background
Region
Eastern
Central West
North West Coast
South West Coast
Alaska
Hawaii
IMPROVE Sites
All sites in Maine, New Hampshire, Vermont, Minnesota, and Michigan
All sites in this region (sites in ID, MT, WY, ND, SD, CO, UT, NV, AZ)
All sites in this region (all Washington and Oregon sites, and the northern
California sites REDW and LAVO)
All sites in this region (all California sites except the northern sites
REDW and LAVO)
All sites in Alaska
All sites in Hawaii
27 The 'eastern' region roughly equates to the combined southeast, northeast, industrial midwest, and
eastern portion (MN, I A, & MO) of the upper midwest regions as defined previously in this chapter (Figure 2-4).
The 'central west' region roughly corresponds to the western portion of the upper midwest region and the eastern
two thirds (ID, MT, CO, UT, NV) of the northwest region. The 'north west coast' approximates the remaining one
third (northern CA, OR, and WA) of the northwest region. The 'south west coast' area is similar to the southern
California region.
28 Non-sulfate PM25 is defined as measured PM25 minus reported ammonuim sulfate.
2-66
-------
Table 2-7. Estimates of long-term means, daily standard deviations and 99th percentiles of
PM2 5 background concentrations (jig/m3)
Region
Eastern U.S.
Central West
North West Coast
South West Coast
Alaska
Hawaii
# Sites
7
37
8
8
1
3
Means
3.0(2.5-3.6)
2.5(1.6-4.6)
3.4(2.2-6.6)
5.2 (2.6-8.6)
1.2
1.1 (0.7-1.8)
St Devs
2.5(2.1-2.8)
1.9(1.3-3.7)
2.8(2.1-4.2)
3.7(1.8-6.8)
1.5
0.9(0.8-1.0)
99th %iles
13(11-15)
10 (6-17)
14(10-21)
20 (9-33)
9
4 (4-5)
Notes:
1) Some of these estimates likely contain a significant North American anthropogenic component.
2) The "Means" column has the mean of the long-term averages of the sites representing the region followed by the
minimum and maximum of the long-term averages of these sites in parentheses. Similarly for the "St Devs" column,
which presents standard deviations of the daily concentrations about the annual means, and the "99th %iles" column,
which presents the 99th percentiles of the daily concentrations over the 23-year period.
Considering these factors, the distributions of daily PM25 concentrations at these sites
provide an indication of the ranges for the daily variability of PM25 background concentrations,
and the 99th percentiles of these distributions are an estimate of the highest daily background
concentrations. Staff notes that these recent findings are generally consistent with those from the
last review, which suggested a range of about 15 to 20 |ig/m3 as the upper end of the distribution
of daily PM25 background concentrations in the U.S. (EPA, 1996b).
2.7 RELATIONSHIP BETWEEN AMBIENT PM MEASUREMENTS AND HUMAN
EXPOSURE
The statutory focus of the primary NAAQS for PM is protection of public health from
the adverse effects associated with the exposure to ambient PM - that is, the focus is on particles
in the outdoor atmosphere that are either emitted directly by sources or formed in the atmosphere
from precursor emissions. We refer to PM in the ambient air as ambient PM. An understanding
of human exposure to ambient PM helps inform the evaluation of underlying assumptions and
interpretation of results of epidemiologic studies that characterize relationships between
monitored ambient PM concentrations and observed health effects (discussed in Chapter 3).
An important exposure-related issue for this review is the characterization of the
relationships between ambient PM concentrations measured at one or more centrally located
monitors and personal exposure to ambient PM, as characterized by particle size, composition,
source origin, and other factors. Information on the type and strength of these relationships,
discussed below, is relevant to the evaluation and interpretation of associations found in
epidemiologic studies that use measurements of PM concentrations at centrally located monitors
2-67
-------
as a surrogate for exposure to ambient PM.29 The focus here is on particle size distinctions; the
CD (CD, Section 5.4) also discusses exposure relationships related to compositional differences.
2.7.1 Definitions
Exposure to a contaminant is defined as contact at a boundary between a human and the
environment (e.g., the breathing zone) at a specific contaminant concentration for a specific
interval of time; it is measured in units of concentration(s) multiplied by time (or time interval)
(National Research Council, 1991). An individual's total personal exposure to PM results from
breathing air containing PM in different types of environments (e.g., outdoors near home,
outdoors away from home, indoors at home, indoors at office or school, commuting, restaurants,
malls, other public places). These environments may have different concentrations of PM with
particles originating from a wide variety of sources.
Ambient PM is comprised of particles emitted by anthropogenic and natural sources and
particles formed in the atmosphere from emissions of gaseous precursors. This includes
emissions not only from outdoor sources such as smokestacks, industrial sources, and
automobiles, but also from sources located indoors with emissions vented outdoors, such as
fireplaces, wood stoves, and some cooking appliances. Exposure to ambient PM can occur both
outdoors and indoors to the extent that ambient PM penetrates into indoor environments - we
use the term PM of ambient origin to refer to both outdoor and indoor concentrations of ambient
PM. We use the term nonambient PM to refer to concentrations of PM that are only due to
indoor sources of particles that are not vented outdoors such as smoking, cooking, other non-
vented sources of combustion, cleaning, mechanical processes, and chemical interactions
producing particles. In characterizing human exposure to PM concentrations relevant to setting
standards for ambient air quality, the CD conceptually separates an individual's total personal
exposure to PM into exposure to PM of ambient origin and exposure to all other sources of PM
(i.e., nonambient PM exposure}.
Outdoor concentrations of PM are affected by emissions, meteorology, topography,
atmospheric chemistry, and removal processes. Indoor concentrations of PM are affected by
several factors, including outdoor concentrations, processes that result in infiltration of ambient
PM into buildings, indoor sources of PM, aerosol dynamics and indoor chemistry, resuspension
of particles, and removal mechanisms such as particle deposition, ventilation, and air-
conditioning and air cleaning devices (CD, p. 5-122). Concentrations of PM inside vehicles are
subject to essentially the same factors as concentrations of PM inside buildings. Personal
exposure to PM also includes a component which results specifically from the activities of an
29 Consideration of exposure measurement error and the effects of exposure misclassification on the
interpretation of the epidemiologic studies are addressed in Chapter 3.
2-68
-------
individual that typically generate particles affecting only the individual or a small localized area
surrounding the person, such as walking on a carpet, referred to as the personal cloud.
Epidemiologic studies generally use measurements from central monitors to represent the
ambient concentrations in an urban or rural area. We use the term central site to mean the site of
a PM monitor centrally located with respect to the area being studied. In many cases,
epidemiologic studies combine the measurements from more than one monitor to obtain a
broader representation of area-wide PM concentrations than a single monitor provides.
2.7.2 Centrally Monitored PM Concentration as a Surrogate for Particle Exposure
The 1996 Criteria Document (EPA, 1996a) presented a thorough review of PM exposure-
related studies up to that time. The 1996 Staff Paper (EPA, 1996b) drew upon the studies,
analyses, and conclusions presented in the 1996 Criteria Document and discussed two
interconnected PM exposure issues: (1) the ability of central fixed-site PM monitors to represent
population exposure to ambient PM and (2) how differences between fine and coarse particles
affect population exposures. Distinctions between PM size classes and components were found
to be important considerations in addressing the representativeness of central monitors. For
example, fine particles have a longer residence time and generally exhibit less variability in the
atmosphere than coarse fraction particles. As discussed in the 1996 Staff Paper, the 1996
Criteria Document concluded that measurements of daily variations of PM have a plausible
linkage to daily variations of human exposures to PM of ambient origin for the populations
represented by the nearby ambient monitoring stations, and that this linkage is stronger for fine
particles than for PM10 or the coarse fraction of PM10. The 1996 Criteria Document further
concluded that central monitoring can be a useful, if imprecise, index for representing the
average exposure of people in a community to PM of ambient origin (EPA, 1996b, p. IV-15, 16).
Exposure studies published since 1996 and reanalyses of studies that appeared in the
1996 Criteria Document are reviewed in the current CD, and provide additional support for these
findings. The CD discusses two classes of fine particles: ultrafine and accumulation-mode
particles (see Chapter 2). Ultrafine, accumulation-mode, and coarse particles have different
chemical and physical properties which affect personal exposures in different ways (CD, Table
9-2, p. 9-17).
An individual's total personal exposure to PM may differ from the ambient concentration
measured at the central site monitor because: (1) spatial differences in ambient PM
concentrations exist across a city or region; (2) generally only a fraction of the ambient PM is
present in indoor or in-vehicle environments, whereas individuals generally spend a large
percentage of time indoors; and (3) a variety of indoor sources of PM contribute to total personal
exposure. Thus, the amount of time spent outdoors, indoors, and in vehicles and the types of
activities engaged in (e.g., smoking, cooking, vacuuming) also will heavily influence personal
2-69
-------
exposure to PM. The first two factors are important for determining the strength of the
relationship between ambient PM and ambient personal exposure.
With regard to the first factor that influences the relationship between total personal
exposure and concentrations measured at central sites, the spatial variability of PM plays a large
role. As discussed in Section 2.4, for many areas PM2 5 concentrations are fairly uniform
spatially, with higher concentrations near roadways and other direct sources of PM25. Analyses
of PM2 5 data for 27 urban areas indicate that differences in annual mean concentrations between
monitoring sites in an urban area range from less than 1 |ig/m3 to as much as 8 |ig/m3. However,
the correlations of daily PM25 between sites are typically greater than 0.80. Daily mean PM25
concentrations exhibit much higher spatial variability than annual means, even when the daily
concentrations at sites are highly correlated. Although the spatial variability of PM25 varies for
different urban areas, overall, some degree of uniformity results from the widespread formation
and long lifetime of the high regional background of secondary PM2 5. In summarizing the key
findings related to spatial variability in PM25 concentrations, the CD states (p. 3-101):
Differences in annual mean PM2 5 concentrations between monitoring sites in
urban areas examined are typically less than 6 or 7 |ig/m3. However, on
individual days, differences in 24-h average PM2 5 concentrations can be much
larger. Some sites in metropolitan areas are highly correlated with each other but
not with others, due to the presence of local sources, topographic barriers, etc.
Although PM2 5 concentrations at sites within a MSA can be highly correlated,
significant differences in their concentrations can occur on any given day.
Consequently, additional measures should be used to characterize the spatial
variability of PM2 5 concentrations. The degree of spatial uniformity in PM2 5
concentrations in urban areas varies across the country. These factors should be
considered in using data obtained by the PM2 5 FRM network to estimate
community-scale human exposure, and caution should be exercised in
extrapolating conclusions obtained in one urban area to another. PM2 5 to PM10
ratios were generally higher in the east than in the west, and values for this ratio
are consistent with those found in numerous earlier studies presented in the 1996
PM AQCD.
Relative to fine particles, coarse and ultrafme particles are likely to be more variable
across urban scales. Daily mean PM10_2 5 concentrations tend to be more variable and have lower
inter-site correlations than PM2 5, possibly due to their shorter atmospheric lifetime (travel
distances < 1 to 10s of km) and the more sporadic nature of PM10_25 sources (CD, Section 3.2.5).
Ultrafme particles also have shorter atmospheric lifetimes (travel distances < 1 to 10s of km,
compared with 100s to 1000s of km for PM25) and spatially variable sources. High
concentrations of ultrafme particles have been measured near roadways, but with concentrations
falling off rapidly with increasing distance from the roadway. Both coarse and ultrafme particles
also have reduced concentrations indoors compared to PM2 5, due to lower infiltration rates,
2-70
-------
greater deposition rates, and coagulation of ultrafme particles into larger particles. These
differences make it more difficult to find a relationship between ambient concentrations and
personal exposures to these size fractions than for PM2 5.
The second factor influencing the relationship between ambient PM concentrations
measured at central sites and total personal exposure to PM is the extent to which ambient PM
penetrates indoors and remains suspended in the air. If the flow of ambient PM into the home
from the outdoors is very restricted, the relationship between ambient PM concentrations
measured at a central site and total exposure to PM will tend to be weaker than in a situation
where ambient PM flows more readily into the home and is a greater part of the overall indoor
PM concentrations. This is heavily dependent on the building air exchange rate, and also on
penetration efficiency and deposition or removal rate, both of which vary with particle
aerodynamic size. Air exchange rates (the rates at which the indoor air in a building is replaced
by outdoor air) are influenced by building structure, the use of air conditioning and heating,
opening and closing of doors and windows, and meteorological factors (e.g., difference in
temperature between indoors and outdoors). Based on physical mass-balance considerations,
usually the higher the air exchange rate the greater the fraction of PM of ambient origin found in
the indoor and in-vehicle environments. Higher air exchange rates also dilute the concentration
of indoor- generated PM. Rates of infiltration of outdoor PM into homes through cracks and
crevices are higher for PM2 5 than for PM10, PM10_2 5, or ultrafme particles (CD, p. 5-123). Since
PM10_2 5 and ultrafme particles penetrate indoors less readily than PM2 5 and deposit to surfaces
more rapidly than PM25, a greater proportion of PM25 of ambient origin is found indoors than
PM10_2 5 and ultrafme particles, relative to their outdoor concentrations. Thus, the particle size
distribution influences the amounts of PM of ambient origin found indoors.
Since people typically spend a large part of their time indoors at home, the air exchange
rate of the home has a large impact on exposures to ambient pollution. Homes with low air
exchange rates are more protected from outdoor sources, and vice-versa. Homes in regions with
moderate climate tend to be better ventilated and have higher air exchange rates than areas which
have very cold or very hot climates. Thus, climate plays an important role in regional population
exposure to ambient pollution.
The third factor influencing the relationship between ambient concentrations measured at
central sites and total personal exposure is the contribution of indoor sources to total personal
exposure. On average, individuals spend nearly 90 percent of their time indoors. The
contribution of indoor sources to indoor concentrations of PM is significant, and can be quite
variable on different days and between individuals. Indoor sources such as combustion devices
(e.g., stoves and kerosene heaters) generate predominantly fine particles; cooking produces both
fine and coarse particles; and resuspension (e.g., dusting, vacuuming, and walking on rugs)
generates predominantly coarse particles (CD, p. 5-82). This factor, however, does not influence
exposure to PM of ambient origin.
2-71
-------
These three factors related to total personal exposure can give rise to measurement error
in estimating exposures to fine and coarse PM (CD, Section 5.5.3), thus making the
quantification of relationships between concentrations measured at central site monitors and
health effects more difficult due to reduction in statistical power. Moreover, exposure
measurement errors can also affect the magnitude and the precision of the health effects
estimates. However, as discussed in the CD and below in Chapter 3, exposure measurement
errors under most ordinary circumstances are not expected to influence the overall interpretation
of findings from either the long-term exposure or time-series epidemiologic studies that have
used ambient concentration data (CD, p. 5-121). They will more likely affect the magnitude of
the effects found from these studies and result in higher effects estimates, since exposure
measurement errors tend to bias towards the null hypothesis.
The CD discusses the finding by some researchers that some epidemiologic studies yield
statistically significant associations between ambient concentrations measured at a central site
and health effects even though there is a very small correlation between ambient concentrations
measured at a central site and total personal exposures. The explanation of this finding is that
total personal exposure includes both ambient and nonambient generated components, and while
the nonambient portion of personal exposure is not generally correlated with ambient
concentrations, the exposure to concentrations of ambient origin is correlated with ambient
concentrations. Thus, it is not surprising that health effects might correlate with central site PM
concentrations, because exposure to PM of ambient origin correlates with these concentrations,
and the lack of correlation of total exposure with central site PM concentrations does not
statistically alter that relationship. By their statistical design, time-series epidemiologic studies
of this type only address the ambient component of exposure, since the impact of day-to-day
fluctuations in ambient PM on acute health effects is examined.
In looking more specifically at the relationship between personal exposure to PM of
ambient origin and concentrations measured at central site monitors, an analysis of data from the
PTEAM study30 provides important findings, as discussed in the CD (p. 5-63 to 5-66 and 5-125
to 5-126). The PTEAM study demonstrated that central site ambient PM10 concentrations are
well correlated with personal exposure to PM10 of ambient origin, while such concentrations are
only weakly correlated with total personal exposure. This study also found that estimated
exposure to nonambient PM10 is effectively independent of PM10 concentrations at central site
monitors, and that nonambient exposures are highly variable due to differences in indoor sources
across the study homes.
30 EPA's Particle Total Exposure Assessment Methodology (PTEAM) field study (Clayton et al, 1993;
Ozkaynak et al., 1996a;b) is a large-scale probability sample based field study. The study measured indoor, outdoor,
and personal PM10, the air exchange rate for each home, and time spent in various indoor residential and outdoor
environments for 147 subjects/households, 12-hr time periods in Riverside, California.
2-72
-------
When indoor sources only have minor contributions to personal exposures, total exposure
is mostly from PM of ambient origin. In these cases high correlations are generally found
between total personal exposure and ambient PM measured at a central site (CD, p. 5-54). For
example, measurements of ambient sulfate, which is mostly in the fine fraction, have been found
to be highly correlated with total personal exposure to sulfate (CD, p. 5-124). Since in these
studies there were minimal indoor sources of sulfate, the relationship between ambient
concentrations and total personal exposure to sulfate was not weakened by possible presence of
small indoor-generated sulfates in some environments.
It is recognized that existing PM exposure measurement errors or uncertainties most
likely will reduce the statistical power of PM health effects analyses, thus making it more
difficult to detect a true underlying association between the exposure metric and the health
outcome of interest. However, the use of ambient PM concentrations as a surrogate for personal
ambient exposures is not expected to change the principal conclusions from PM epidemiological
studies that use community average health and pollution data (CD, p. 5-121). Based on these
considerations and on the review of the available exposure-related studies, the CD concludes that
for epidemiologic studies, ambient PM2 5 concentration as measured at central site monitors is a
useful surrogate for exposure to PM2 5 of ambient origin. However, for coarse and ultrafine PM,
such ambient concentrations are not likely to be as good a surrogate for personal ambient
exposure. While nonambient PM may also be responsible for health effects, since the ambient
and nonambient components of personal exposure are independent, the health effects due to
nonambient PM exposures generally will not bias the risk estimated for ambient PM exposures
(CD, p. 9-17).
2.8 RELATIONSHIP BETWEEN AMBIENT PM AND VISIBILITY
The effect of ambient particles on visibility is dependent upon particle size and
composition, atmospheric illumination, the optical properties of the atmosphere, and the optical
properties of the target being viewed. The optical properties of particles, discussed in section
2.2.5, can be well characterized in terms of a light extinction coefficient. For a given distribution
of particle sizes and compositions, the light extinction coefficient is strictly proportional to the
particle mass concentration. Light extinction is a measure of visibility impairment, and, as such,
provides a linkage between ambient PM and visibility, as discussed below in section 2.8.1.
Other measures directly related to the light extinction coefficient are also used to characterize
visibility impairment, including visual range and deciviews, as discussed below in section 2.8.2.
Light extinction associated with background levels of PM is also discussed below in section
2.8.3.
2-73
-------
2.8.1 Particle Mass and Light Extinction
Fine particle mass concentrations can be used as a general surrogate for visibility
impairment. However, as described in many reviews of the science of visibility, the different
constituents of PM2 5 have variable effects on visibility impairment. For example, sulfates and
nitrates contribute substantially more to light scattering per unit mass than other constituents,
especially as relative humidity levels exceed 70 percent. Thus, while higher PM2 5 mass
concentrations generally indicate higher levels of visibility impairment, it is not as precise a
metric as the light extinction coefficient. By using historic averages, regional estimates, or
actual day-specific measurements of the component-specific percentage of total mass, however,
one can develop reasonable estimates of light extinction from PM mass concentrations (see
section 6.2.2 for further discussion).
The light extinction coefficient has been widely used in the U.S. for many years as a
metric to describe the effect of concentrations of particles and gases on visibility. It can be
defined as the fraction of light lost or redirected per unit distance through interactions with gases
and suspended particles in the atmosphere. The light extinction coefficient represents the
summation of light scattering and light absorption due to particles and gases in the atmosphere.
Both anthropogenic and non-anthropogenic sources contribute to light extinction. The light
extinction coefficient (6ext) is represented by the following equation (CD, 4-155):
*ex, = *aP + bag + *sg+ bsp (5-1)
where &ap = light absorption by particles
6ag = light absorption by gases
bsg = light scattering by gases (also known as Rayleigh scattering)
bsp = light scattering by particles.
Light extinction is commonly expressed in terms of inverse kilometers (km"1) or inverse
megameters (Mm"1), where increasing values indicate increasing impairment.
Total light extinction can be measured directly by a transmissometer or it can be
estimated from ambient pollutant concentrations. Transmissometers measure the light
transmitted through the atmosphere over a distance of 1 to 15 kilometers. The light transmitted
between the light source (transmitter) and the light-monitoring component (receiver) is
converted to the path-averaged light extinction coefficient. Transmissometers operate
continuously, and data are often reported in terms of hourly averages.
Direct relationships exist between measured ambient pollutant concentrations and their
contributions to the extinction coefficient. The contribution of each aerosol constituent to total
light extinction is derived by multiplying the aerosol concentration by the extinction efficiency
for that aerosol constituent. Extinction efficiencies vary by type of aerosol constituent and have
2-74
-------
been obtained for typical atmospheric aerosols by a combination of empirical approaches and
theoretical calculations. For certain aerosol constituents, extinction efficiencies increase
significantly with increases in relative humidity.
EPA guidance for tracking progress under the regional haze rule specifies an algorithm
for calculating total light extinction as the sum of aerosol light extinction for each of the five
major fine particle components and for the coarse fraction mass, plus 10 Mm"1 for light
extinction due to Rayleigh scattering, discussed below. This algorithm is represented by the
following equation (CD, 4-169):
bat = (3)f(RH) [SULFATE]
+ (3)f(RH) [NITRATE]
+ (4) [ORGANIC CARBON]
+ (10) [LIGHTABSORBING CARBON] (5-2)
+ (1) [SOIL]
+ (0.6) [COARSEPM]
+ 10 (for Rayleigh scattering by gases)
The estimated mass for each component is multiplied by its dry extinction efficiency and,
in the case of sulfate and nitrate, by a relative humidity adjustment factor, f(RH), to account for
their hygroscopic behavior (CD, p. 4-169). The relative humidity adjustment factor increases
significantly with higher humidity, ranging from about 2 at 70 percent, to 4 at 90 percent, and
over 7 at 95 percent relative humidity (CD, p. 4-170, Figure 4-38).
Rayleigh scattering represents the degree of natural light scattering found in a particle-
free atmosphere, caused by the gas molecules that make up "blue sky" (e.g., N2, O2). The
magnitude of Rayleigh scattering depends on the wavelength or color of the light being
scattered, as well as on the density of gas in the atmosphere, and varies by site elevation,
generally from 9 to 11 Mm"1 for green light at about 550 nm (CD, p. 4-156 to 4-157). A standard
value of 10 Mm-1 is often used to simplify comparisons of light extinction values across a
number of sites with varying elevations (Malm, 2000; CD, p. 4-157). The concept of Rayleigh
scattering can be used to establish a theoretical maximum horizontal visual range in the earth's
atmosphere. At sea level, this maximum visual range is approximately 330 kilometers ignoring
the Earth's curvature. Since certain meteorological conditions can lead to visibility conditions
that are close to "Rayleigh," it is analogous to a baseline or boundary condition against which
other extinction components can be compared.
The light extinction coefficient integrates the effects of aerosols on visibility, yet is not
dependent on scene-specific characteristics. It measures the changes in visibility linked to
emissions of gases and particles. By apportioning the light extinction coefficient to different
2-75
-------
aerosol constituents, one can estimate changes in visibility due to changes in constituent
concentrations (Pitchford and Malm, 1994).
2.8.2 Other Measures of Visibility
Visual range is a measure of visibility that is inversely related to the extinction
coefficient. Visual range can be defined as the maximum distance at which one can identify a
large black object against the horizon sky. The colors and fine detail of many objects will be lost
at a distance much less than the visual range, however. Visual range has been widely used in air
transportation and military operations in addition to its use in characterizing air quality.
Conversion from the extinction coefficient to visual range can be made with the following
equation (NAPAP, 1991):
Visual Range (km) = 39\2/bext(Mml) (5-3)
Another important visibility metric is the deciview, a unitless metric which describes
changes in uniform atmospheric extinction that can be perceived by a human observer. It is
designed to be linear with respect to perceived visual changes over its entire range in a way that
is analogous to the decibel scale for sound (Pitchford and Malm, 1994). Neither visual range nor
the extinction coefficient has this property. For example, a 5 km change in visual range or 0.01
km"1 change in extinction coefficient can result in a change that is either imperceptible or very
apparent depending on baseline visibility conditions. Deciview allows one to more effectively
express perceptible changes in visibility, regardless of baseline conditions. A one deciview
change is a small but perceptible scenic change under many conditions, approximately equal to a
10 percent change in the extinction coefficient (Pitchford and Malm, 1994). Deciview can be
calculated from the light extinction coefficient (bext) by the equation:
Haziness (dv) = 10 \n(bJW Mm'1) (5-4)
Figure 2-31 graphically illustrates the relationships among light extinction, visual range, and
deciview.
2.8.3 Visibility at PM Background Conditions
Light extinction caused by PM from natural sources can vary significantly from day to
day and location to location due to natural events such as wildfire, dust storms, and volcanic
eruptions. It is useful to consider estimates of natural background concentrations of PM on an
annual average basis, however, when evaluating the relative contributions of anthropogenic
(man-made) and non-anthropogenic sources to total light extinction. Background PM is defined
and discussed in detail in section 2.6, and Table 2-65 provides the annual average regional
background PM2 5 mass ranges for the eastern and western U.S..
2-76
-------
The National Acid Precipitation Assessment Program report (NAPAP, 1991) provides
estimates of extinction contributions from background levels of fine and coarse particles, plus
Rayleigh scattering. In the absence of anthropogenic emissions of visibility-impairing particles,
these estimates are 26 + 7 Mm"1 in the east, and 17 + 2.5 Mm"1 in the west. These equate to a
naturally-occurring visual range in the east of 150 + 45 km, and 230 + 35 km in the west.
Excluding light extinction due to Rayleigh scattering, annual average background levels of fine
and coarse particles are estimated to account for approximately 14 Mm"1 in the east and about 6
Mm"1 in the west. The primary non-anthropogenic substances responsible for natural levels of
visibility impairment are naturally-occurring organics, suspended dust (including coarse
particles), and water associated with hygroscopic particles. At the ranges of fine particle
concentrations associated with background conditions, discussed above in section 2.6, small
changes in fine particle mass have a large effect on total light extinction. Thus, higher levels of
background fine particles and associated average humidity levels in the east result in a fairly
significant difference between naturally occurring visual range in the rural east as compared to
Extinction (Mm"1) 10 20 30 40 50 70 10° 20° 30° 40° 50° 70° 100°
Deciviews (dv)
1
0
1
1 1
7 11
I I
I
14
I
I
16
I
I I I
19
I II
II
23
II
I
30
I
I
34
I
I
37
I
I
39
I
I I I
42
I I I
II
46
II
Visual Range (km) 40o 2oo 130 100 so eo 40 20 13 10 a s 4
Figure 2-31. Relationship between light extinction, deciviews, and visual
range.
Source: Malm (1999)
the rural west. This issue is discussed further in Chapter 6, section 6.2.
Fine particles originate from both natural and anthropogenic, or man-made, sources.
Background concentrations of fine particles are those originating from natural sources. On an
annual average basis, concentrations of background fine particles are generally small when
compared with concentrations of fine particles from anthropogenic sources (NRC, 1993). The
same relationship holds true when one compares annual average light extinction due to
background fine particles with light extinction due to background plus anthropogenic sources.
Table VIII-4 in the 1996 Staff Paper makes this comparison for several locations across the
country by using background estimates from Table VIII-2 and light extinction values derived
from monitored data from the IMPROVE network. These data indicate that anthropogenic
emissions make a significant contribution to average light extinction in most parts of the country,
as compared to the contribution from background fine particle levels. Anthropogenic
contributions account for about one-third of the average extinction coefficient in the rural west
and more than 80 percent in the rural east (NAPAP, 1991).
-------
It is important to note that, even in areas with relatively low concentrations of
anthropogenic fine particles, such as the Colorado plateau, small increases in anthropogenic fine
particle concentrations can lead to significant decreases in visual range. As discussed in the CD,
visibility in an area with lower concentrations of air pollutants (such as many western Class I
areas) will be more sensitive to a given increase in fine particle concentration than visibility in a
more polluted atmosphere. Conversely, to achieve a given amount of visibility improvement, a
larger reduction in fine particle concentration is required in areas with higher existing
concentrations, such as the east, than would be required in areas with lower concentrations. This
relationship between changes in fine particle concentrations and changes in visibility (in
deciviews) also illustrates the relative importance of the overall extinction efficiency of the
pollutant mix at particular locations. At a given ambient concentration, areas having higher
average extinction efficiencies, due to the mix of pollutants, would have higher levels of
impairment. In the east, the combination of higher humidity levels and a greater percentage of
sulfate as compared to the west causes the average extinction efficiency for fine particles to be
almost twice that for sites on the Colorado Plateau.
2-78
-------
REFERENCES
Clayton, C. A.; Perritt, R. L.; Pellizzari, E. D.; Thomas, K. W.; Whitmore, R. W.; Wallace, L. A.; Ozkaynak, H.;
Spengler, J. D. (1993). Particle total exposure assessment methodology (PTEAM) study: distributions of
aerosol and elemental concentrations in personal, indoor, and outdoor air samples in a southern California
community. J. Exposure Anal. Environ. Epidemiol. 3: 227-250.
Environmental Protection Agency (1986). Guideline on the Identification and Use of Air Quality Data Affected by
Exceptional Events. EPA-450/4-86-007.
Environmental Protection Agency (1996a). Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
Office of Research and Development; report no. EPA/600/P-95/001aF-cF.3v
Environmental Protection Agency (1996b). 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: Office of Air Quality Planning and Standards; report no. EPA-452/R-96-013.
Environmental Protection Agency (2004a). EPA's Environmental Technology Verification Program. Research
Triangle Park, NC: Office of Research and Development; report no. EPA/600F-04/064.
Environmental Protection Agency (2004b). The Particle Pollution Report. Current Understanding of Air Quality
and Emissions through 2003. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office
of Air Quality Planning and Standards; report no. EPA-454-R-04-002.
Fehsenfeld, F.; D. Hastie; C. Chow; and P.A. Solomon (2003). "Gas and Particle Measurements, Chapter 5." In
NARSTO Paniculate Matter Science Assessment. McMurray, P., Shepherd, M., and Vickery, J. eds.
NARSTO, U.S. Environmental Protection Agency, Research Triangle Park, NC.
Husar, R. B.; Tratt, D. M.; Schichtel, B. A.; Falke, S. R.; Li, F.; Jaffe, D.; Gasso, S.; Gill, T.; Laulainen, N. S.;
Lu, F.; Reheis, M. C.; Chun, Y.; Westphal, D.; Holben, B. N.; Gueymard, C.; McKendry, L; Kuring, N.;
Feldman, G. C.; McClain, C.; Frouin, R. J.; Merrill, J.; DuBois, D.; Vignola, F.; Murayama, T.;
Nickovic, S.; Wilson, W. E.; Sassen, K.; Sugimoto, N.; Malm, W. C. (2001) Asian dust events of
April 1998. J. Geophys. Res. (Atmos.) 106: 18,317-18,330.
Kim, S.; Shen, S.; Sioutas, C. (2002). Size distribution and diurnal and seasonal trends of ultrafme particles in
source and receptor sites of the Los Angeles basin. J. Air. Waste Manag. Assoc.., Mar; 52(3):297-307.
Langstaff, J. E. (2005). OAQPS Staff Memorandum to PM NAAQS Review Docket (OAR-2001-0017). Subject:
Estimation of Policy-Relevant Background Concentrations of Particulate Matter. [January 27, 2005].
Malm, W.C. (2000) Spatial and seasonal patterns and temporal variability of haze and its constituents in the United
States. Report III. Colorado State University, Cooperative Institute for Research in the Atmosphere. Fort
Collins, CO. Available: http://vista.cira.colostate.edu/improve/Publications/Reports/2000/2000.htm [22
March, 2002].
Malm, W.C.; Day, D.E.; Kreidenweis, S.M. (1999). Light scattering characteristics of aerosols as a function of
relative humidity: a comparison of measured scattering and aerosol concentrations using the theoretical
models. J. Air Waste Manage. Assoc. 50: 686-709
National Acid Precipitation Assessment Program (NAPAP), (1991). Office of the Director, Acid Deposition: State
of Science and Technology. Report 24, Visibility: Existing and Historical Conditions - Causes and Effects.
Washington, D.C.
National Research Council (1991). Human exposure assessment for airborne pollutants: advances and opportunities.
Washington, DC: National Academy of Sciences.
2-79
-------
National Research Council. 1993. Protecting Visibility in National Parks and Wilderness Areas. National
Academy Press, Washington, D.C.
Nichols, M. D. (1996) Memorandum to EPA Air Division Directors regarding Areas Affected by PM-10 Natural
Events. May 30, 1996.
Ozkaynak, H.; Xue, I; Spengler, I; Wallace, L.; Pellizzari, E.; Jenkins, P. (1996a). Personal exposure to airborne
particles and metals: results from the particle TEAM study in Riverside, California. J. Exp. Anal. Environ.
Epidemiol. 6: 57-78.
Ozkaynak, H.; Xue, J.; Weker, R.; Bulter, D.; Koutrakis, P.; Spengler, J. (1996b). The particle TEAM (PTEAM)
study: analysis of the data: final report, volume III. Research Triangle Park, NC: U.S. Environmental
Protection Agency, Atmospheric Research and Exposure Assessment Laboratory; report no. EPA/600/R-
95/098. Available from: NTIS, Springfield, VA; PB97-102495.
Pitchford, M.; Malm, W. (1994) Development and Applications of a Standard Visual Index. Atmospheric
Environment. Vol. 28, no. 5, pp. 1049-1054.
Robinson, A.L.; Lipsky, E.M.; Pekney, N., Lucas, L., Wynne, D., Rogge, W.F., Bernado-Bricker, A. and Sevimoglu,
O. (2005). "Fine Particle Emission Profile For Road Dust in Pittsburgh, Pennsylvania." 17PB-1. Presented
at AAAR Specialty Conference: Paniculate Matter, Supersites Program and Related Studies , February 7-
11,2005, Atlanta GA.
http://www.netl.doe.gov/coal/E&WR/pubs/AAAR/robinson. road, dust.aaar.ss.poster.pdf
Rogge, W.F.; Hildemann, L.M.; Mazurek, M.A.; Cass, G.R.; Simoneit, B.R.T. (1993). Sources of fine organic
aerosol. 3. Road dust, tire debris, and organometallic brake lining dust: roads as sources and sinks.
Environ. Sci. Technol. 27:1982-1904.
Sardar. S.B.; Fine, P.M., and Sioutas, C. (2005). "Seasonal and spatial variability of the size-resolved chemical
composition of paniculate matter (PM10) in the Los Angeles Basin." Journal of Geophysical Research, V.
110.
Schmidt, M.; D. Mintz; V Rao; L. McCluney; Frank, N. (2005). U.S. EPA Memorandum to File. Subject:
Analyses of 2001-2003 PM Data for the PM NAAQS Review. June 30, 2005. Available:
www.epa.gov/oar/oaqps/pm25/docs.html.
Solomon, P. A.; M.P. Tolocka; G. Norris; and M. Landis (2001). "Chemical Analysis Methods for Atmospheric
Aerosol Components." In Aerosol Measurement: Principles, Techniques, and Application, Second Edition,
Eds. P. Barren and K. Willeke. John Wiley & Sons, Inc., New York, NY.
Whitby, K. T. (1978). The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
Wilson, W. E.; Suh, H.H. (1997) Fine particles and coarse particles: concentration relationships relevant to
epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
2-80
-------
3. POLICY-RELEVANT ASSESSMENT OF HEALTH EFFECTS EVIDENCE
3.1 INTRODUCTION
This chapter assesses key policy-relevant information on the known and potential health
effects associated with exposure to ambient PM, alone and in combination with other pollutants
that are routinely present in ambient air. More specifically, this assessment focuses on health
effects associated with exposures to ambient fine particles and to thoracic coarse particles,
consistent with EPA's decision in the last review to establish new standards for fine particles
separate from those intended to address effects related to thoracic coarse particles. The
presentation here first summarizes the qualitative assessment of health evidence contained in the
CD, as a basis for the evidence-based assessment of primary standards for PM presented in
Chapter 5. Secondly, this assessment addresses key issues relevant to quantitative assessment of
the epidemiologic health evidence available in this review so as to provide a foundation for the
quantitative health risk assessment discussed in Chapter 4 and used in the risk-based assessment
of primary standards for PM presented in Chapter 5.
In the last review of the PM NAAQS, a variety of health effects had been associated with
ambient PM at concentrations extending from those elevated levels found in the historic London
episodes down to levels below the 1987 PM10 standards. The epidemiologic evidence for PM-
related effects was found to be strong, suggesting a "likely causal role" of ambient PM in
contributing to a range of health effects (62 FR 38657). Of special importance in the last review
were the conclusions that (1) ambient particles smaller than 10 jim that penetrate into the
thoracic region of the respiratory tract remained of greatest concern to health, (2) the fine and
coarse fractions of PM10 should be considered separately for the purposes of setting ambient air
quality standards, and (3) the consistency and coherence of the health effects evidence greatly
added to the strength and plausibility of the observed PM associations. Important uncertainties
remained, however, such as issues related to interpreting the role of gaseous co-pollutants in PM
associations with health effects, and the lack of demonstrated biologic mechanisms that could
explain observed effects.
EPA's conclusion in the last review that fine and thoracic coarse particles should be
considered as separate pollutants was based on differences in physical and chemical properties,
sources, atmospheric formation and transport, relationships with human exposure, and evidence
of health effects (62 FR 38667). In this review, the CD has evaluated the newly available
evidence related to the physics and chemistry of particulate matter, exposure relationships, and
particle dosimetry. The CD notes that the chemical and physical distinctions between fine and
coarse particles recognized in the last review remain generally unchanged; recent studies
continue to show that fine and coarse particles generally have different sources and composition
and different formation processes (see Table 2-2 herein). Recent exposure research finds that
accumulation-mode fine particles can infiltrate into buildings more readily than can thoracic
3-1
-------
coarse particles, and that ambient concentrations of PM10_25 are less well correlated and less
uniform across a community than ambient concentrations of PM25(CD, p. 9-21). The CD also
concludes that the new evidence from dosimetry studies continues to reinforce distinctions
between fine and coarse particles, and submodes within fine particles, with regard to deposition
patterns in the respiratory tract, though there is significant overlap between particle size classes
(CD, p. 9-21 to 9-22). Based on these considerations, the CD concludes that it remains
appropriate to consider fine and thoracic coarse particles as separate subclasses of PM (CD, p.
9-22).
The assessment of health evidence in this chapter therefore focuses on health effects
associated with fine and thoracic coarse particles, drawing from the CD's evaluation and
conclusions on the full body of evidence from health studies, summarized in Chapters 6 through
9 of the CD, with particular emphasis on the integrative synthesis presented in Chapter 9. That
integrative synthesis focuses on integrating newly available scientific information with that
available from the last review and integrated from various disciplines, so as to address a set of
issues central to EPA's assessment of scientific information upon which this review of the PM
NAAQS is to be based. It is intended to provide a coherent framework for assessment of human
health effects elicited by ambient PM in the U.S., and to facilitate consideration of the key
policy-related issues to be addressed in this Staff Paper, including recommendations as to
appropriate indicators, averaging times, levels, and forms for PM NAAQS.
As summarized in Chapters 6 through 9 of the CD, a large number of new studies
containing further evidence of serious health effects have been published since the last review,
with important new information coming from epidemiologic, toxicologic, controlled human
exposure, and dosimetric studies. As was true in the last review, evidence from epidemiologic
studies plays a key role in the CD's evaluation of the scientific evidence. As discussed further in
section 3.3, some highlights of the new evidence include:
New multi-city studies that use uniform methodologies to investigate the effects
of various indicators of PM on health with data from multiple locations with
varying climate and air pollution mixes, contributing to increased understanding
of the role of various potential confounders, including gaseous co-pollutants, on
observed associations. These studies provide more precise estimates of the
magnitude of an effect of exposure to PM than most smaller-scale individual city
studies.
More studies of various health endpoints evaluating independent associations
between effects and fine and thoracic coarse particles, as well as ultrafme
particles or specific components (e.g., sulfates, metals).
• Numerous new studies of cardiovascular endpoints, with particular emphasis on
assessment of cardiovascular risk factors or physiological changes.
O O
3-2
-------
• Studies relating population exposure to PM and other pollutants measured at
centrally located monitors to estimates of exposure to ambient pollutants at the
individual level have lead to a better understanding of the relationship between
ambient PM levels and personal exposures to ambient PM.
New analyses and approaches to addressing issues related to potential
confounding by gaseous co-pollutants, possible thresholds for effects, and
measurement error and exposure misclassification.
• Preliminary attempts to evaluate the effects of air pollutant combinations or
mixtures including PM components using factor analysis or source apportionment
methods to link effects with different PM source types (e.g., combustion, crustal1
sources).
• Several "intervention studies" have reported improvements in health in areas
where policy, economic or regulatory changes resulted in reduced air pollutant
concentrations (section 8.2.3.4 in the CD).
In addition, the body of evidence on PM-related effects has greatly expanded with
findings from studies that help inform mechanisms of action, including important new
dosimetric, toxicologic and controlled human exposure studies.
• Animal and controlled human exposure studies using concentrated ambient
particles (CAPs), new indicators of response (e.g., C-reactive protein and
cytokine levels, heart rate variability), and animal models simulating sensitive
subpopulations, that are relevant to demonstrating plausibility of the
epidemiologic evidence and provide insights into potential mechanisms for PM-
related effects.
• Dosimetric studies using new modeling methods that provide increased
understanding of the dosimetry of different particle size classes and in members
of potentially sensitive subpopulations, such as people with chronic respiratory
disease.
In presenting that evidence and conclusions based on it, this chapter first summarizes
information from the CD's evaluation of health evidence from the different disciplines. Sections
3.2 and 3.3 provide overviews of the CD's findings on the evidence of potential mechanisms for
PM-related effects and on the nature of effects associated with PM exposures, respectively.
Drawing from the integration of evidence in Chapter 9 of the CD, the chapter summarizes the
1 "Crastal" is used here to describe particles of geologic origin, which can be found in both fine- and
coarse-fraction PM.
O O
3-3
-------
CD's integrative findings and conclusions regarding causality in section 3.4, with a particular
focus on results for fine and thoracic coarse particles. Section 3.5 also draws from the CD's
integrative synthesis to characterize potential at-risk subpopulations and potential public health
impacts of exposure to ambient PM. Finally, section 3.6 addresses several key issues relevant to
the staffs interpretation and quantitative assessment of the health evidence, including:
(1) considerations related to air quality measurements and data used in the health studies;
(2) measurement error and exposure error in fine and thoracic coarse particle studies;
(3) specification of models used in epidemiologic studies; (4) approaches to evaluating the role
of co-pollutants and potential confounding in PM-effects associations; (5) questions related to
exposure time periods used in associations between air quality and health effects, including lag
periods used in short-term exposure studies and the selection of time periods used to represent
exposures in long-term exposures studies; and (6) questions related to the form of concentration-
response relationships and potential threshold levels. In this final section, staff builds upon the
CD's detailed evaluation and integration of the scientific evidence on these issues to reach
conclusions regarding the use of the health study results in quantitative evaluation and the PM
risk assessment discussed in Chapter 4.
3.2 MECHANISMS
This section provides an overview of evidence presented in the CD on potential
mechanisms by which exposure to PM may result in effects, drawing from Chapters 6 and 7 of
the CD. Evidence from dosimetric studies has played a key role in previous PM NAAQS
reviews, especially in the decision to revise the indicator from total suspended particulates (TSP)
to PM10 to focus on thoracic particles (52 FR 24634, July 1, 1987). In contrast, in previous
reviews of the PM NAAQS there has been little available evidence on potential biologic
mechanisms by which deposited particles could affect the lungs or heart.
An evaluation of the ways by which inhaled particles might ultimately affect human
health must take account of patterns of deposition and clearance in the respiratory tract (CD,
p. 6-1). Briefly, the human respiratory tract can be divided into three main regions: (1)
extrathoracic, (2) tracheobronchial, and (3) alveolar (CD, Figure 6-1). The regions differ
markedly in structure, function, size, mechanisms of deposition and removal, and sensitivity or
reactivity to deposited particles. Overall, the health concerns related to ambient particles are
greater for the two lower regions.
Fine particles, including accumulation mode and ultrafme particles, and thoracic coarse
particles can all penetrate into and be deposited in the tracheobronchial and alveolar regions of
the respiratory tract, though (as noted above) there are differences among these size fractions.
Penetration into the tracheobronchial and alveolar regions is greater for accumulation mode
particles than coarse or ultrafme particles, since coarse and ultrafme particles are more
5-4
-------
efficiently removed from the air in the extrathoracic region than are accumulation-mode fine
particles (CD, 6-105).
Once past the extrathoracic region, deposition fraction in the tracheobronchial and
alveolar regions varies with different exertion levels or breathing patterns, and whether breathing
is through the mouth or nose. As shown in Figures 6-16 and 6-17 of the CD, deposition fractions
in these regions are largest for particles in the coarse fraction and ultrafme modes. More
specifically, the CD concludes that fractional deposition in the alveolar region of the respiratory
system for healthy individuals is greatest for particles in the size ranges of approximately 2.5 to
5 |im and 0.02 to 0.03 jim, and fractional deposition to the tracheobronchial region is greatest for
particles in the size range of approximately 4 to 6 |im (CD, p. 6-109).
Respiratory disease status can also affect regional particle deposition patterns. Studies
have indicated that, in general, enhanced deposition of particles occurs at airway bifurcations
(CD, p. 6-20). New evidence confirms that people with chronic obstructive lung diseases can
have increased total lung deposition and can also show increases in local deposition ("hot spots")
due to uneven airflow in diseased lungs (CD, p. 6-34). In such cases, the respiratory condition
can enhance sensitivity to inhaled particles by increasing the delivered dose overall as well as
increased doses to localized regions. Such dosimetry studies are of obvious relevance to
identifying sensitive populations (see section 3.5).
The potential effects of deposited particles are influenced by the rate and nature of
removal. The predominant clearance and translocation mechanisms vary across the three regions
of the respiratory system. For example, dissolution or absorption of particles or particle
constituents and endocytosis by cells such as macrophages are two primary mechanisms
operating in the alveolar region. These mechanisms also occur in the tracheobronchial region,
but the primary mechanisms for particle clearance or translocation from the tracheobronchial
region are mucociliary transport and coughing (CD, 6-44, Table 6-2). Soluble components of
particles may also move into the circulatory system and thus throughout the body. Recent
studies also suggest that ultrafme particles or their soluble constituents may move directly from
the lungs into the systemic circulation, providing a pathway by which ambient PM exposure
could affect extrapulmonary organs (CD, p. 6-55).
In summary, new evidence from dosimetry studies has advanced our understanding of the
complex and different patterns of particle deposition and clearance in the respiratory tract
exhibited by fine particles in the accumulation mode, ultrafme particles, and thoracic coarse
particles. The evidence shows that all three size fractions can enter the tracheobronchial or
alveolar regions of the respiratory system and potentially cause effects.
A major research need identified in the last review was the need to understand the
potential biologic mechanisms by which deposited particles could result in the varying effects
observed in epidemiologic studies with PM exposure. New evidence from toxicologic and
controlled human exposure studies has helped to identify and provide support for a number of
3-5
-------
potential pathways by which particles could have biologic effects, as discussed in Chapter 7 of
the CD. Fully defining the mechanisms of action for PM would involve description of the
pathogenesis or origin and development of any related diseases or processes resulting in
premature mortality. While the evidence summarized in the CD has provided important insights
that contribute to the plausibility of effects observed in community health studies, this more
ambitious goal of fully understanding fundamental mechanisms has not yet been attained. Some
of the more important findings presented in chapter 7 of the CD, including those related to the
cardiovascular system, may be more accurately described as intermediate responses potentially
caused by PM exposure rather than complete mechanisms. It appears unlikely that the complex
mixes of particles that are present in ambient air would act alone through any single pathway of
response. Accordingly, it is plausible that several physiological responses might occur in
concert to produce reported health endpoints.
By way of illustration, Mauderly et al. (1998) discussed particle components or
characteristics hypothesized to contribute to PM health, producing an illustrative list of 11
components or characteristics of interest for which some evidence existed. The list included:
1) PM mass concentration, 2) PM particle size/surface area, 3) ultrafine PM, 4) metals, 5) acids,
6) organic compounds, 7) biogenic particles, 8) sulfate and nitrate salts, 9) peroxides, 10) soot,
and 11) co-factors, including effects modification or confounding by co-occurring gases and
meteorology. The authors stress that this list is neither definitive nor exhaustive, and note that
"it is generally accepted as most likely that multiple toxic species act by several mechanistic
pathways to cause the range of health effects that have been observed" (Mauderly et al., 1998).
In assessing the more recent animal, controlled human, and epidemiologic information,
the CD developed a summary of current thinking on pathophysiological mechanisms for the
effects related to PM exposure. Section 7.10.1 of the CD discusses a series of potential
mechanisms or potential general pathways for effects on the heart and lung. The CD's
conclusions on the evidence supporting different types of effects is briefly summarized below.
The relative support for these potential mechanisms/intermediate effects and their relevance to
real world inhalation of ambient particles varies significantly. Moreover, the CD highlights the
variability of results that exist among different approaches, investigators, animal models, and
even day-to-day within studies. Nonetheless, the CD states that "[fjindings since 1996 have
provided evidence supporting many hypotheses regarding induction of PM effects; and this body
of evidence has grown substantially." (CD, p. 7-205). For the most part, the evidence from
toxicologic and controlled human exposure studies discussed below reflects the effects of fine
particles or fine particle constituents.
Direct Pulmonary Effects. Potential pathways for direct pulmonary effects include:
lung injury and inflammation, increased airway reactivity and asthma exacerbation, and
increased susceptibility to respiratory infections. The CD finds "particularly compelling"
evidence that PM exposure causes lung injury and inflammation. Evidence that supports
3-6
-------
hypotheses on direct pulmonary effects includes toxicologic and controlled human exposure
studies using both sources of ambient particles and combustion-related particles. Toxicologic
studies using inhalation or intratracheal instillation of ambient particles from various locations
have shown that ambient particles can cause lung inflammation and injury (CD, Tables 7-4 and
7-5). Several studies using filter extracts from Utah Valley ambient samples collected before,
during and after the shut-down of a major particle-emitting facility have reported effects such as
increases in oxidant generation, release of cytokines such as IL-8, and evidence of pulmonary
injury such as increased levels of lactate dehydrogenase (CD, p 7-46, 7-47). Administration of
residual oil fly ash has been shown to produce acute lung injury and severe inflammation, with
effects including recruitment of neutrophils, eosinophils and monocytes into the airway (CD, p.
7-60). New toxicologic or controlled human exposure studies using exposure to CAPs have
reported some evidence of inflammatory responses in animals, as well as increased susceptibility
to infections, though the results of this group of studies are more equivocal (CD, p. 7-85). In
vitro studies, summarized in section 7.4.2 of the CD, also report evidence of lung injury,
inflammation, or altered host defenses with exposure to ambient particles or particle constituents.
Some toxicologic evidence also indicates that PM can aggravate asthmatic symptoms or increase
airway reactivity, especially in studies of the effects of diesel exhaust particles (CD, section
7.3.5). Finally, some new evidence suggests that particles can initiate neurogenic responses in
the respiratory system. For example, several studies have indicated that some particles can
activate sensory nerve receptors in the airways, leading to inflammatory responses such as
cytokine release (CD, section 7.4.4.4)
Systemic Effects Secondary to Lung Injury. Adding to the list of direct pulmonary
effects, these pathways include: impairment of lung function leading to cardiac effects,
pulmonary inflammation and cytokine production leading to systemic hemodynamic effects,
lung inflammation leading to increased blood coagulability, and lung inflammation leading to
hematopoiesis effects. While more limited than for direct pulmonary effects, some new evidence
from toxicologic studies suggests that injury or inflammation in the respiratory system can lead
to changes in heart rhythm, reduced oxygenation of the blood, changes in blood cell counts, or
changes in the blood that can increase the risk of blood clot formation, a risk factor for heart
attacks or strokes (CD, pp. 7-209 to 7-212).
Effects on the Heart In addition, potential pathways for effects on the heart include:
effects related to uptake of particles or particle constituents in the blood, and effects on the
autonomic control of the heart and circulatory system. In the last review, little or no evidence
was available on potential cardiovascular effects from toxicologic studies. More recent studies
have provided some initial evidence that particles can have direct cardiovascular effects. As
shown in Figure 7-1 of the CD, there are several pathways by which particle deposition in the
respiratory system could lead to cardiovascular effects, such as PM-induced pulmonary reflexes
resulting in changes in the autonomic nervous system that then could affect heart rhythm (CD, p.
3-7
-------
7-8). Also, inhaled PM could affect the heart or other organs if particles or particle constituents
are released into the circulatory system from the lungs; some new evidence indicates that the
smaller ultrafme particles or their soluble constituents can move directly from the lungs into the
systemic circulation (CD, p. 6-55). The CD concludes that the data remain limited but provide
some new insights into mechanisms by which particles, primarily fine particles, could affect the
cardiovascular system (CD, 7-35, 7-212).
The above list of potential mechanisms and/or general pathways for effects was
developed mainly in reference to effects from short-term rather than long-term exposure to PM.
Repeated occurrences of some short-term insults, such as inflammation, might contribute to
long-term effects, but wholly different mechanisms might also be important in the development
of chronic responses. Some mechanistic evidence is available, however, for potential
carcinogenic or genotoxic effects of particles. Section 7.10.1 of the CD also includes a
discussion of the evidence for mutagenic or genotoxic effects of particles or particle constituents,
concluding that "both ambient PM and combustion products of coal, wood, diesel, and gasoline
are mutagenic/genotoxic." (CD, p.7-215).
While some new studies have exposed animals or humans to ambient fine particles, many
toxicologic and controlled human exposure studies have used exposures to fine particle
constituents or emission-related particles, such as fly ash or diesel exhaust particles. The
evidence related to fine particle types or components is summarized in section 7.10.2 of the CD.
Overall, the findings indicate that different health responses are linked with different particle
characteristics and that both individual components and complex particle mixtures appear to be
responsible for many biologic responses relevant to fine particle exposures (CD, p. 7-206).
In addition to the evidence discussed above that related primarily to fine particles, there
is some limited evidence from toxicologic studies on PM10_2 5, for either acute or chronic
exposures (CD, p. 9-55). The CD includes results from several in vitro toxicologic studies that
provide some insight into potential effects of thoracic coarse particles, particularly related to
inflammatory or allergic effects. Two recent studies report inflammatory responses in cells
exposed to extracts of water-soluble and water-insoluble materials from thoracic coarse particles
and fine particles collected in Chapel Hill, NC (CD, p. 7-83, Monn and Becker, 1999; and CD, p.
7-101 and 7-102, Soukup and Becker, 2001). One study focused on water-soluble materials, and
reported significant cytotoxicity and cytokine production with water-soluble extracts of ambient
PM10_2 5, in contrast to the lack of effects observed with extracts from ambient PM2 5 as well as
indoor-collected PM10_2 5 and PM2 5. The authors report that endotoxin appeared to have a role in
inflammatory effects, while metals appeared to have a role in the cytoxocity of thoracic coarse
particle materials (CD, p. 7-83, Monn and Becker, 1999). Soukup and Becker (2001) used both
soluble and insoluble components of thoracic coarse particles and fine particles, and report that
the insoluble materials from thoracic coarse particles resulted in cytokine production, decreased
phagocytic ability and oxidant generation (CD, p. 7-101 and 7-102). In this extract of thoracic
-------
coarse particles, endotoxin appeared to be the most pro-inflammatory component, but "other
moieties" (not endotoxin or metals) appeared to contribute to oxidant generation (CD, p. 7-102).
Using particles collected in two urban areas in the Netherlands, Becker et al. (2003) reported that
thoracic coarse particles, but not fine or ultrafine particles, resulted in effects related to
inflammation and decreased pulmonary defenses (CD, p. 7-106). This small group of studies
thus suggests that exposure to thoracic coarse particles may cause pro-inflammatory effects, as
well as cytotoxicity and oxidant generation.
In addition, Diociaiuti et al. (2001) reported greater hemolytic effects with fine particles
than with thoracic coarse particles when exposing blood cell cultures to extracts of particles
collected in an urban area of Rome; increased hemolysis was seen with only the highest PM10_25
dose (CD, p. 7-102). Hornberg et al. (1998) reported evidence of genotoxic activity in human
bronchoepithelial cells exposed to both PM2 5 and PM10_2 5, with stronger evidence for
genotoxicity in fine particles (CD, p. 7-171). These two studies suggest only limited hemolytic
or carcinogenic effects of thoracic coarse particle exposures.
Road dust is a common source of thoracic coarse particles and can be considered as a PM
sample that is more representative of thoracic coarse particles than fine particles. In the 1996
Staff Paper, results from one key toxicologic study were highlighted in which immunological
and cellular toxicity was observed in rats with exposure to road dust. Higher concentrations of
road dust were needed to cause effects, compared with exposures to fine particle components
(e.g., sulfates, nitrates), but it was observed that some of the apparent differential toxicity was
due to differential penetration efficiencies of particles in the rat (EPA, 1996b, p. V-70). A recent
study reported that road tunnel dust particles had greater adjuvant activity in two animal models
of allergy than several other particle samples, including residual oil fly ash and diesel exhaust
particles (Steerenberg et al., 2003; CD, p. 7-136 to 7-137). In contrast, a number of studies have
reported that Mt. St. Helens volcanic ash, which is generally in the size range of thoracic coarse
particles, has very little toxicity in animal or in vitro toxicologic studies (CD, p. 7-216).
Many of the newer studies use relatively high doses (in mg or hundreds of jig), though
some have used doses that are close to ambient concentrations. A key consideration for
evaluating the results of animal toxicologic studies is the relation between effects reported in
animals with high dose exposures to effects that would be expected in human populations with
ambient exposures. The CD presents an illustrative set of analyses evaluating the doses and
responses reported in human and animal studies in Appendix 7A of the CD. In the analyses,
dosimetric models were used to predict doses of deposited and retained particles in various
regions of the respiratory system for humans and rats. In this series of analyses, the dose ratios
for humans to rats were quite variable across dose metrics and respiratory system regions. For
example, even when humans and rats are similarly exposed (i.e., exposed at rest to the same
aerosol for 6 hours) the equivalent exposure ratios can range from 0.09 to 33 (CD, p. 7A-30,
Table 7A-7a). The CD also evaluated relative dose levels using data from two sets of studies in
3-9
-------
which toxicologic and controlled human exposure studies used the same type of ambient
particles (Utah Valley dust and concentrated ambient particles). Based on these data, deposited
and retained doses in the alveolar and tracheobronchial regions were estimated for three studies
using concentrated ambient particles, and doses were 40- to 67-fold higher in rats than in humans
from these inhalation exposure studies (CD, pp. 7A-61). However, the CD observed that similar
and/or lesser inflammatory responses were reported for rats than for humans, suggesting that rats
may be less susceptible to effects of concentrated ambient particles than healthy humans (CD, p.
7A-61). Recognizing the limitations of this small set of illustrative analyses, the CD concludes
that larger doses in rats may be dosimetrically equivalent to lower doses in humans, given the
faster particle clearance rates in rats (CD, p. 7A-62). However, the CD also observed that the
prediction of dose levels depends on a number of factors, and estimated equivalent exposure
ratios for rats and humans vary substantially (CD, 7-163).
The CD also observes that particles may help carry other airborne substances into the
respiratory tract (CD, section 7.9). For example, hygroscopic particles can take up moisture and
grow in the humid atmosphere of the respiratory tract, thus potentially altering the deposition
and clearance patterns of the particles. Water-soluble gases can be carried into the lung on
particles, and delivery of reactive gases such as SO2 and formaldehyde to the lower respiratory
regions can be increased when carried on particles since these gases would otherwise be more
likely trapped in the upper airways. Particles can also carry reactive oxygen species, such as
hydrogen peroxide, and other toxic compounds such as polynuclear aromatic hydrocarbons or
allergens, into the lower respiratory regions (CD, pp, 7-203, 7-204).
In summary, while investigation of potential mechanisms for the effects of particles
remains an important research question, new mechanistic studies provide evidence to support a
number of hypothesized mechanisms of action for ambient PM, primarily for fine PM. In
evaluating this new body of evidence, the CD states: "Thus, there appear to be multiple biologic
mechanisms that may be responsible for observed morbidity/mortality due to exposure to
ambient PM. It also appears that many biologic responses are produced by PM whether it is
composed of a single component or a complex mixture" (CD, p. 7-206).
3.3 NATURE OF EFFECTS
An extensive body of new epidemiologic studies has been published since completion of
the 1996 PM CD. In the last review, epidemiologic evidence indicated that exposure to PM
(using various indicators) was associated with increased risk for various cardiopulmonary
effects, including mortality and a range of indices of morbidity associated with respiratory and
cardiovascular disease such as hospital admissions and emergency room visits, school absences,
work loss days, restricted activity days, effects on lung function and symptoms, morphological
changes, and altered host defense mechanisms. The CD finds that recent epidemiologic studies
have continued to report associations between various indicators of ambient PM and effects such
3-10
-------
as premature mortality, hospital admissions or emergency department visits for respiratory and
cardiovascular disease, and effects on lung function and symptoms (CD, p. 9-23). In addition,
recent studies now identify several new types of health outcomes reported to be associated with
exposure to various indicators of PM, including physicians' office or clinic visits, cardiovascular
health indicators such as heart rate variability or increased C-reactive protein levels, and
developmental effects such as low birth weight, and infant mortality (CD, p. 9-23, 9-24).
The discussions that follow draw primarily from epidemiologic evidence evaluated in
Chapter 8 of the CD as well as the CD's integration of evidence across disciplines (section 9.2).
The CD evaluates evidence from the full body of epidemiologic studies conducted world-wide
and summarizes results of all such mortality and morbidity studies in Appendices 8A and 8B,
respectively, in the CD. For the purposes of this Staff Paper, staff draws from the CD's
qualitative evaluation of all studies, but focuses on those conducted in the U.S. and Canada for
quantitative assessments.2 Effect estimates for mortality and morbidity effects associated with
increments of PM10, PM25, and PM10_25 from multi-city and single-city U.S. and Canadian studies
are summarized in Appendices 3 A and 3B to this chapter for short-term and long-term exposure
studies, respectively, as a consolidated reference for the following discussions.3
A number of the new time-series epidemiologic studies have used generalized additive
models (GAM) in their analyses, and issues have been found with the convergence criteria and
the method for determining standard errors when using GAM, as discussed in section 3.6.3 more
fully and in section 8.4.2 of the CD. In Appendix 3A, results are presented from those short-
term exposure studies that have been reanalyzed to address issues related to GAM, or that did
not use GAM in their analyses. In presenting study results in figures in this section, for studies
in which multiple reanalysis results were presented, staff has selected effect estimates based on
the authors' stated judgments, where offered, or selected results from models using generalized
linear models (GLM).4
2 Findings of U.S. and Canadian studies are more directly applicable for quantitative considerations in this
review, since studies conducted in other countries may well reflect quite different population and air pollution
characteristics.
3 For consistency across studies, the effect estimates summarized in Appendices 3 A and 3B, and the results
presented in figures in this section, are from single-pollutant models. Results of multi-pollutant models are
discussed in the text. As presented in the CD, effect estimates are presented using standardized PM increments to
allow for comparison across studies. For short-term exposures studies, increments of 50 ug/m3 for PM10 and 25
ug/m3 for PM2 5 and PM10_2 5 were used; for long-term exposures studies, increments of 20 ug/m3 for PM10 and 10
ug/m3 for PM2 5 and PM10_2 5 were used (CD, p. 8-4).
4 For studies that include results for GLM analyses using several methods to adjust for temporal or weather
variables, if no judgment is offered by the authors on model selection, staff has presented results from the models
using adjustment methods most closely matching those of the initial study.
3-11
-------
3.3.1 Premature Mortality
This section includes an overview of the CD's findings on (1) mortality associations with
short-term PM exposure, with emphasis on results from newly available multi-city analyses; and
(2) mortality associations with long-term PM exposure.
3.3.1.1 Mortality and Short-term PM Exposure
Historical reports of dramatic pollution episodes have provided clear evidence of
mortality associated with high levels of PM and other pollutants, as summarized in the 1996 CD
(EPA, 1996a, pp. 12-28 to 12-31). More recently, associations between increased daily mortality
and various indicators of PM have been reported at much lower concentrations in a large number
of areas with differing climates, PM composition, and levels of gaseous co-pollutants. Since the
last review, a large number of new time-series studies of the relationship between short-term
exposure to various indicators of PM and mortality have been published, including several multi-
city studies that are responsive to the recommendations from the last review (CD, p. 8-24).
Included in the PM CD are results from numerous studies that have been conducted in single
cities or locations in the U.S. or Canada, as well as locations in Europe, Mexico City, South
America, Asia and Australia (Table 8A in the CD). As was observed based on the more limited
studies available in the last review, the associations reported in the recent studies on short-term
exposure to PM10 and mortality are largely positive, and frequently statistically significant. Staff
have focused on the results of studies conducted in the U.S. and Canada in this assessment;
effect estimates from U.S. and Canadian multi-city and single-city studies are presented in
Figure 3-1 for associations between PM10, PM25 and PM10_25 and mortality.5
In this review, the CD has emphasized the results of the multi-city studies as being of
particular relevance. The multi-city studies combine data from a number of cities that may vary
in climate, air pollutant sources or concentrations, and other potential risk factors. The
advantages of multi-city analyses include: (1) evaluation of associations in larger data sets can
provide more precise effect estimates than pooling results from separate studies; (2) consistency
in data handling and model specification can eliminate variation due to study design; (3) effect
modification or confounding by co-pollutants can be evaluated by combining data from areas
with differing air pollutant combinations; (4) regional or geographical variation in effects can be
evaluated; and (5) "publication bias" or exclusion of reporting of negative or nonsignificant
findings can be avoided (CD, p. 8-30).
The National Morbidity, Mortality and Air Pollution Study (NMMAPS) is the largest
available multi-city analysis, and included analyses of PM10 effects on mortality in 90 U.S. cities
(Samet et al., 2000a,b; Dominici et al., 2003a). Additional, more detailed, analyses were
conducted in a subset of the 20 largest U.S. cities (Samet et al., 2000b). The NMMAPS study
5 The effect estimates in Figure 3-1 (for mortality effects) and in Figure 3-2 (for morbidity effects;
discussed below in section 3.3.2) have been plotted in order of decreasing study power, using as an indicator the
natural log of the product of the number of study days and number of health events per day.
3-12
-------
so
£ «*
o
.E ..H
£L
«
re i o
£
UJ
-10 -
-ZQ '•
PM
1f,
PM™
(92)
•> /•
35
•— (fl
25
' * ' '
26
Total Mortality
Respiratory Total
CV Mortality Mortality Mortality
CV Respiratory Total CV Resp.
Jy^ojtaNty^Jy^ojiajity^^Mj^^
Figure 3-1. Excess risk estimates for total nonaccidental, cardiovascular, and respiratory mortality in multi-pollutant (in
bold font below) and single-pollutant models for U.S. and Canadian studies. PM increments: 50 ug/m3 for PM10
and 25 ug/m3 for VM2S and PM10_25. Results presented from time-series studies that did not use GAM or were
reanalyzed using GLM.
1. Domlnici et al. (2003a), 90 U.S. cities 14.
2. Schwartz (2003b), 10 U.S. cities 15.
3. Klemm and Mason (2003), 6 U.S. cities 16.
4. Burnett and Goldberg (2003), 8 Canadian cities 17.
5. Moolgavkar (2003), Cook County 18.
6. Kinneyetal. (1995), Los Angeles 19.
7. Schwartz (2003b), Chicago 20.
8. Ito and Thurston (1996), Cook County 21.
9. Schwartz (2003b), Pittsburgh 22.
10. Styer et al. (1995), Cook County 23.
11. Schwartz (2003b), Detroit 24.
12. Moolgavkar (2003), Los Angeles 25.
13. Schwartz (2003b), Seattle 26.
Schwartz (2003b), Minneapolis
Klemm and Mason (2003), St. Louis
Klemm and Mason (2003), Boston
Schwartz (2003b), Birmingham
Schwartz (2003b), New Haven
Chock et al. (2000), Pittsburgh (< 75 y.o.)
Chock et al. (2000), Pittsburgh (75+ y.o.)
Klemm and Mason (2003), Kingston-Harriman
Klemm and Mason (2003), Portage
Schwartz (2003b), Canton
Schwartz (2003b), Spokane
Ito (2003), Detroit
Fairley (2003), Santa Clara County
27. Schwartz (2003b), Colorado Springs
28. Klemm and Mason (2003), Topeka
29. Tsai et al. (2000), Newark
30. Klemm and Mason (2003), Steubenville
31. Pope et al. (1992), Utah Valley
32. Tsai et al. (2000), Elizabeth
33. Tsai et al (2000), Camden
34. Lipfert et al. (2000), Philadelphia
35. Mar et al. (2003), Phoenix
36. Ostro et al. (2003), Coachella Valley
37. Klemm and Mason (2000), Atlanta
38. Ostro et al. (1995), Southern California
3-13
-------
was designed to use a multi-city approach such as that recommended following an earlier report
of time-series study reanalyses that recommended investigating the role of co-pollutants in PM-
health outcome relationships by conducting multi-city studies, using consistent analytical
approaches across cities (HEI, 1997, p. 38; Samet et al., 2000c, p. 1). The NMMAPS used a
uniform methodology to evaluate the relationship between mortality and PM10 for the different
cities, and the results were synthesized to provide a combined estimate of effects across the
cities. The authors reported associations between total and cardiorespiratory mortality and PM10
that were robust to different modeling approaches and to adjustment for gaseous co-pollutants.
For total mortality, the overall risk estimate for all cities is a statistically significant increase of
1.4% (using more stringent GAM) or 1.1% (using GLM) per 50 |ig/m3 PM10 (Dominici et al.,
2003a; CD, p. 8-33). Key components of the NMMAPS analyses include assessment of the
potential heterogeneity in effects and effects of co-pollutants, as discussed below in sections
3.4.3 and 3.6.4, respectively.
Another major multi-city study used data from 10 U.S. cities that were selected from
NMMAPS cities where daily PM10 monitoring data were available (in many areas, monitoring is
done on a l-in-3 or l-in-6 day basis) (Schwartz, 2003b). The authors reported a statistically
significant association between PM10 and total mortality, with an effect estimate of an increase of
3.4% per 50 |ig/m3 PM10 (in reanalyzed GAM results) or 2.8% per 50 |ig/m3 PM10 (using GLM)
(Schwartz, 2003b; CD, p. 8-38). The CD observes that the effect estimates from this study are
larger than those reported in NMMAPS, and suggests that the availability of more frequent
monitoring data may partly account for the differences (CD, p. 8-39).
In the previous review, results for one key multi-city study were available, in which
associations were assessed between daily mortality and PM10, PM2 5, and PM10_2 5 measurements
from six U.S. cities (the "Six Cities" study) (Schwartz, et al., 1996). The authors reported
significant associations for total mortality with PM2 5 and PM10, but not with PM10_2 5. Reanalyses
of Six Cities data have reported results consistent with the findings of the original study, with
statistically significant increases in total mortality ranging from 2% to over 3% reported for
results from more stringent GAM or GLM analyses using either PM2 5 (per 25 |ig/m3 increment)
or PM10 (per 50 |ig/m3 increment), whereas PM10_2 5 was only significantly associated with
mortality in one of the six cities (Steubenville) (Schwartz, 2003 a; Klemm and Mason, 2003; CD,
p. 8-40 to 8-41).
Using data for the eight largest Canadian cities, mortality was associated with PM2 5,
PM10, and PM10_2 5 and the effect estimates were of similar magnitude for each PM indicator
(Burnett et al., 2000; Burnett and Goldberg, 2003). Using either more stringent GAM or GLM,
the authors reported increases ranging from 2% to 3% in total mortality for each PM indicator.
The association between mortality and PM2 5 generally remained statistically significant in a
number of analyses when gaseous co-pollutants and 0- and 1-day lags were included in the
models, although in a few instances the effect estimates were reduced and lost statistical
3-14
-------
significance. Associations with PM10, and PM10_2 5 did not reach statistical significance, though
the effect estimates were similar in magnitude to those for PM2 5. While the associations
reported with PM10_2 5 were somewhat increased in magnitude in reanalyses, they did not reach
statistical significance. The CD concludes that it is difficult to compare the relative significance
of associations with PM25 and PM10_25, but for this study, "overall, they do not appear to be
markedly different" (Burnett and Goldberg, 2003; CD, p. 8-42).
The CD also highlights results of analyses from a major European multi-city study, the
Air Pollution and Health: A European Approach (APHEA) study, that evaluated associations
between mortality and various PM measures (CD, section 8.2.2.3.3). In the analyses that
included data from 29 European cities, overall effect estimates of 2 to 3% increased risk of
mortality per 50 |ig/m3 PM10 were reported; reanalysis resulted in reduced effect estimate size,
though the authors conclude that their findings are robust to the application of alternative
modeling strategies (Katsouyanni et al., 2003; CD, p. 8-47). Taken together, the CD concludes
that multi-city studies in the U.S., Canada, and Europe reported statistically significant
associations with effect estimates ranging from -1.0 to 3.5% increased risk of total mortality per
50 |ig/m3 PM10 (CD, p. 8-50).
In considering the results from single-city analyses, Figure 3-1 shows that almost all
effect estimates for PM2 5 are positive and a number are statistically significant, particularly
when focusing on the results of studies with greater precision. As summarized in the CD, effect
estimates for total mortality from the multi-city studies range from ~1 to 3.5% per 25 |ig/m3
PM2 5. For the relatively more precise single-city studies, effect estimates range from
approximately 2 to 6% per 25 |ig/m3 PM25 (CD, p. 9-28). Figure 3-1 also shows effect estimates
for PM10_2 5 that are generally positive and similar in magnitude to those for PM2 5 and PM10 but
for total mortality, none reach statistical significance. Staff notes that on a unit mass basis, the
effect estimates for both PM2 5 and PM10_2 5 are generally larger than those for PM10, which is
consistent with PM2 5 and PM10_2 5 having independent effects (CD, p. 9-25).
In general, effect estimates are somewhat larger for respiratory and cardiovascular
mortality than for total mortality. In the NMMAPS analyses using data from the 20 largest U.S.
cities, the effect estimates for deaths from cardiorespiratory causes were somewhat larger than
those for deaths from all causes (1.6% versus 1.1% increased risk per 50 |ig/m3 PM10, using
GLM) (Dominici, et al., 2003a; CD, p. 8-78). In Figure 3-1, for all three PM indicators, it can be
seen that not only is the effect estimate size generally larger for cardiovascular mortality, but the
effect estimates are also more likely to reach statistical significance. This is particularly true for
PM10_25, where two of the five effect estimates for cardiovascular mortality shown are positive
and statistically significant (Mar et al., 2003; Ostro et al., 2003). For respiratory mortality,
effect estimates are often larger than those for either total or cardiovascular mortality, but they
are often less precise, which would be expected since respiratory deaths comprise a small
proportion of total deaths. The CD concludes that effect estimates fall in the range of 3 to 7%
3-15
-------
per 25 |ig/m3 PM2 5 for cardiovascular or cardiorespiratory mortality, and 2 to 7% per 25 |ig/m3
PM2 5 for respiratory mortality in U.S. and Canadian cities. The magnitude of the effect
estimates for PM10_2 5 are similar to those for PM2 5, generally falling in the range of 3 to 8% for
cardiovascular mortality and 3 to 16% for respiratory mortality per 25 |ig/m3 PM10_25 (CD, p.
8-306).
While some of the studies conducted in Europe, Mexico or South America use
gravimetric PM measurements (e.g., PM10, PM25, PM10_25), many of the non-North American
studies use PM indicators such as TSP, black smoke (BS) or coefficient of haze (COH), and the
Australian studies used nephelometric measures of PM. While effect estimates for different PM
indicators may not be quantitatively comparable, the CD observes that "many of the newly
reported analyses continue to show statistically significant associations between short-term
(24-hr) PM exposures indexed by a variety of ambient PM measurements and increases in daily
mortality in numerous U.S. and Canadian cities, as well as elsewhere around the world" (CD,
p. 8-24). These effect estimates are generally within (but toward the lower end of) the range of
PM10 estimates previously reported in the 1996 PM AQCD.
As discussed in section 8.2.2.5 of the CD, associations have been reported between
mortality and short-term exposure to a number of PM components, especially fine particle
components. Three recent studies have used PM25 speciation data to evaluate the effects of air
pollutant combinations or mixtures using factor analysis or source apportionment methods to
link effects with different PM2 5 source types. These studies reported that fine particles from
combustion sources, including motor vehicle emissions, coal combustion, oil burning and
vegetative burning, were associated with increased mortality. No significant increase in
mortality was reported with a source factor representing crustal material in fine particles (CD, p.
8-85). These studies indicate that exposure fine particles from combustion sources, but not
crustal material, is associated with mortality.
The findings of these studies, while providing some insight into what sources of fine
particles might be associated with mortality, are not directly relevant to evaluating effects of
thoracic coarse particles from different sources. Combustion sources are a major contributor to
PM2 5 emissions, but not PM10_25, while crustal material is an important component of PM10_2 5 but
only a small portion of PM25. Staff observes that no epidemiologic evidence is available to
evaluate effects of different components or sources of thoracic coarse particles. One study that
does have some relevance to considering the effects of PM10_25 from different sources assessed
the contribution of dust storms to PM10-related mortality. The authors focused on days when
dust storms or high wind events occurred, during which thoracic coarse particles are the
dominant fraction of PM10, in Spokane. No evidence was reported of increased mortality on
days with high PM10 levels related to dust storms (average PM10 level was 221 |ig/m3 higher on
3-16
-------
dust storm days than on other study days) (Schwartz, et al., 1999), suggesting that PM10_2 5 from
wind-blown dust is also not likely associated with mortality.6
3.3.1.2 Mortality and Long-term PM Exposure
In the 1996 PM CD, results were presented for three prospective cohort studies of adult
populations (i.e., the Six Cities, American Cancer Society (ACS), and California Seventh Day
Adventist (AHSMOG) studies). The 1996 CD concluded that the chronic exposure studies,
taken together, suggested associations between increases in mortality and long-term exposure to
PM (EPA, 1996a, p. 13-34). New studies discussed in the CD (section 8.2.3) include a
comprehensive reanalysis of data from the Six Cities and ACS studies, new analyses using
updated data from the AHSMOG and ACS studies, and a new analysis using data from a cohort
of veterans. Effect estimates from all four of these studies are provided in Appendix 3B.
The reanalysis of the Six Cities and ACS studies included two major components, a
replication and validation study, and a sensitivity analysis, where alternative risk models and
analytic approaches were used to test the robustness of the original analyses. The reanalysis
investigators replicated the original results, confirming the original investigators' findings of
associations with both total and cardiorespiratory mortality (Krewski et al., 2000; CD, p. 8-95).
In single-pollutant models, none of the gaseous co-pollutants was significantly associated with
mortality except SO2. Further reanalyses of the ACS study included multi-pollutant models with
the gaseous pollutants, and the associations between mortality and both fine particles and
sulfates were unchanged in these models, except for those including SO2. While recognizing that
increased mortality may be attributable to more than one component of ambient air pollution, the
authors report that the reanalysis confirmed the association between mortality and fine particle
and sulfate exposures (Krewski et al., 2000; CD, p. 8-95).
The extended analyses for the ACS cohort study included follow-up health data and air
quality data from the new fine particle monitoring network for 1999-2000, and reported
significant associations between long-term exposure to fine particles (using various averaging
periods for air quality concentrations) and premature mortality from all causes, cardiopulmonary
diseases, and lung cancer (Pope et al., 2002; CD p. 8-102). This extended analysis included the
use of data on gaseous pollutant concentrations, more recent data on fine particle concentrations,
and evaluated further the influence of other covariates (e.g., dietary intake data, occupational
exposure) and model specification for the PM-mortality relationship (e.g., new methods for
6In addition, studies conducted in several areas in the western U.S. have reported that associations between
PM10 and mortality or morbidity remained unchanged or became larger and more precise when days indicative of
wind-blown dust or high PM10 concentration days were excluded from the analyses (Pope et al., 1999; Schwartz,
1997; Chen et al., 2000; Hefflin et al., 1994). This group of studies does not provide conclusive evidence, however,
of any effects or lack of effects associated with wind-blown dust or high concentration days, but does indicate that
associations between PM10 and health outcomes in these western areas are not overly influenced or "driven by" such
days.
3-17
-------
spatial smoothing and random effects models in the Cox proportional hazards model) (CD, p. 8-
97). The investigators reported that the associations found with sulfate and fine particle
concentrations were robust to the inclusion of many covariates for socioeconomic factors or
personal health variables (e.g., dietary factors, alcohol consumption, body mass index); however,
as was found in the reanalysis of the original ACS study, education level was found to be an
effect modifier, in that associations were stronger for those with lower education levels (Pope et
al., 2002; CD, p. 8-104). In both the reanalyses and extended analyses of the ACS cohort study,
long-term exposure to PM10_25 was not significantly associated with mortality (CD, p. 8-105;
Krewski et al., 2000; Pope et al., 2002).
There are also new analyses using updated data from the AHSMOG cohort. These
include more recent air quality data for PM10 and estimated PM2 5 concentrations from visibility
data, along with new health information from continued follow-up of the Seventh Day Adventist
cohort (CD, pp. 8-105, 8-110; Abbey et al., 1999; McDonnell et al., 2000). In contrast to the
original study in which no statistically significant results were reported with TSP, a significant
association was reported between total mortality and PM10 for males, but not for females (CD, p.
8-106). Additional analyses were conducted using data from males only and estimated PM2 5 and
PM10_2 5 concentrations; larger effect estimates were reported for mortality with PM2 5 than with
PM10_2 5, but the estimates were generally not statistically significant (McDonnell et al., 2000;
CD, pp. 8-110 and 8-117).
In the VA cohort study, analyses were done using subsets of PM exposure and mortality
time periods, and the investigators report inconsistent and largely nonsignificant associations
between PM exposure (including, depending on availability, TSP, PM10, PM2 5, PM15 and
PM15.2 5) and mortality (CD, pp. 8-110 to 8-111; Lipfert et al., 2000b).
Based on an evaluation of all the available long-term exposure studies, the CD places
greatest weight on the results of the Six Cities and ACS studies. In so doing, the CD notes that
the Six Cities and ACS studies (including reanalyses and extended analyses) included measured
PM data (in contrast with AHSMOG PM estimates based on TSP or visibility measurements),
have study populations more similar to the general population than the VA study cohort, and
have been validated through an exhaustive reanalysis (CD, pp. 8-116 and 8-118; 9-33).
One new effect reported in the extended analysis of the ACS study was a statistically
significant association between fine particle and sulfate concentrations and lung cancer
mortality, with a 13% increased risk of lung cancer mortality per 10 |ig/m3 PM25, using air
quality data averaged across all available years (CD, p. 8-99). This effect estimate is little
changed and remains significant with adjustment for covariates, random effects modeling and
spatial smoothing methods (CD, Figure 8-8). Also, in new analyses using updated data from the
AHSMOG cohort, positive associations were reported between long-term PM10 exposure lung
cancer mortality that were statistically significant for males, but not females (CD, p. 8-108 and
8-109).
3-18
-------
Thus, emphasizing the results from the Six Cities and ACS cohorts, the CD finds that
there are significant associations for mortality with long-term exposure to PM2 5. Based on these
studies, effect estimates for deaths from all causes fall in a range of 6 to 13% increased risk per
10 |ig/m3 PM25, while effect estimates for deaths from cardiopulmonary causes fall in a range of
6 to 19% per 10 |ig/m3 PM2 5. For lung cancer mortality, the effect estimate was a 13% increase
per 10 |ig/m3 PM25 in the results of the extended analysis from the ACS cohort (Pope et al.,
2002; CD, Table 8-12). In addition, based on evidence from reanalyses and extended analyses
using ACS cohort data, the CD concludes that the long-term exposure studies find no
associations between long-term exposure to thoracic coarse particles and mortality (CD, p.
8-307).
3.3.2 Morbidity
The epidemiologic evidence also includes associations between various indicators of PM
and a wide range of endpoints reflecting both respiratory- and cardiovascular-related morbidity
effects. The following sections summarize the CD's findings on PM-related morbidity effects,
beginning with hospital admissions and medical visits for respiratory and cardiovascular
diseases. Subsequent sections provide overviews of the CD's evaluation of evidence for effects
on the respiratory and cardiovascular systems. Effect estimates for associations between short-
term exposure to PM10, PM25 and PM10_25 with hospitalization and medical visits from U.S. and
Canadian studies are presented below in Figure 3-2. Appendix 3A includes effect estimates for
associations with hospitalization and medical visits, as well as those for respiratory symptoms
and lung function and physiological cardiovascular effects, with short-term exposures to PM10,
PM25 or PM10_25 from U.S. and Canadian studies. The results for all new cardiovascular and
respiratory admissions/visits studies, including those using nongravimetric PM measurements
and studies from non-North American locations, are summarized in the CD in section 8.3, and a
more complete discussion of all studies is available in Appendix 8B of the CD.
3.3.2.1 Hospitalization and Medical Visits
Numerous recent studies have continued to report significant associations between short-
term exposures to PM and hospital admissions or emergency department visits for respiratory or
cardiovascular diseases. The new studies have included multi-city analyses, numerous
assessments using cardiovascular admissions/visits, and evaluation of the effects of fine and
thoracic coarse particles.
The NMMAPS multi-city analysis included analyses of associations with hospital
admissions among the elderly, and reported statistically significant associations between PM10
and hospital admissions in the elderly for cardiovascular diseases, pneumonia and chronic
obstructive pulmonary disease (COPD) in 14 cities (Samet et al., 2000; Schwartz et al., 2003).
Increases of 5% in hospital admissions for cardiovascular disease and 8% and 6% in hospital
admissions for COPD or pneumonia, respectively, per 50 |ig/m3 PM10 were reported in the
3-19
-------
O
*-•
CO
70 -
60 -
50 -
40 -
UJ 30 -
(A
(A
(A
ro
o
IE!
ro
O
ro
t
j
f>
n
O
"
3
O
ro i_ t_ *°
zj *— ro ro is
1 | > > .2
-a > o o -a
co — "o "o ro t ^
1 50 OS s 5
i 2 3 4 5 i I
^HM
r
c
3_
- f
8
•
1
4
= ]
io
!
4
Hospital
ro <
,c
>.
j i_
?°
13 r
j
4
Admissions
i
j
e-
o
ro
'o.
t/)
_ a:
TO
= 1
Is
=3 ™
Q) ._
C: CL <
^1
1 2
T+
0 0
oil
' Q . — O
O ,4 4
1 T 1.
*
u
<
1
1 1 i
EO
Visit
i i
CV Diseases
!
11
4
3
3
0<
O
o
8 '
4
Hospital
Admissions
^
c
ro
E
* s.
f
'EL
. en
9
4
1 l_
!
i
<
1
4
ro
c
L
a>
Of
ro
E
x
; ' ' O ^
o £r
^ 2
ro £
Q.
9 ,2 S
ra
f=
O £
™ TEE ^
3 ro J^
>
"O
O
4 "
f
i i
^ a> >*
08 | 8 '
8
•
4
c
0
8
4
Hospital
4
Visit Admissions
Resp. Diseases
i i
CV Dis
<= ir
o
>, | 13
m c £
°" E 8 2
dJ .C Q •£_
7 < o S
CL 11 8 9
O
o < h
4 "•" <
o ro
~ro O ro
•— 7 >, c: Q
CL ' i- O Q_
(/) O g: Q
^ ro ™ S °
18 I '«. = 8
en •*— Ot
1s ltro8
9 I- 7 1
^ ' — w >, £
8X8° 2
8 8 *
4 > 4 ,
<
'
H .. I III I
4
' •« '
if i i i i i t i
ED Hospital ED Hospital Hospital
Visit Admissions Visit Admissions Admissions
|
ii ii ii i
?ases Resp. Diseases CV Diseases Resp. Dis.
-10 -
-20 -
Figure 3-2. Excess risk estimates for hospital admissions and emergency department visits for cardiovascular
and respiratory diseases in single-pollutant models from U.S. and Canadian studies, including aggregate results
from one multicity study (as denoted in bold below). PM increments: 50 ug/m3 for PM10 and 25 ug/m3 for PM2 5
and PM10_2 5. Results presented from time-series studies that did not use GAM or were reanalyzed using GLM.
PM effect size estimate (± 95% confidence intervals) are depicted for the studies listed below. (Source: CD
Figure 9-5)
1. Zanobetti and Schwartz (2003)
U.S. 14 cities
1. Linn et al. (2000), Los Angeles
3. Moolgavkar (2003), Cook County
4. Moolgavkar (2003), Los Angeles
5. Schwartz and Morris (1995), Detroit
6. Morris and Naumova (1998), Chicago
7. Burnett et al. (1997), Toronto
8. Ito (2003), Detroit
9. Stieb et al. (2000), St. John
10. Schwartz (1994), Detroit
11. Sheppard (2003), Seattle
12. Nauenberg and Basu (1999), Los Angeles
13. Thurston et al. (1994), Toronto
14. Tolbert et al. (2000), Atlanta
15. Lipsett et al. (1997), Santa Clara County
16. Choudhury et al. (1997), Montreal
17. Delfmo et al. (1997), Montreal
18. Delfmo et al. (1998), Montreal
3-20
-------
NMMAPS. Effect estimates with PM10 were not correlated with city-specific correlations
between PM10 and co-pollutant levels, which the authors conclude indicates a lack of
confounding by co-pollutants, although the CD recognizes that further evaluation is needed on
this method to assess potential confounding (CD, p. 8-146, 8-175).
Numerous single-city studies have been published that report associations between short-
term PM10 exposure and hospitalization or medical visits for cardiovascular or respiratory
diseases. Overall, the CD reports that the more precise effect estimates for hospitalization range
from 2 to 6% per 50 |ig/m3 PM10 for cardiovascular diseases and 2 to 12% per 50 |ig/m3 PM10 for
respiratory diseases. The CD also observes that new studies reporting associations between
PM10 and medical (e.g., physicians' office) visits for respiratory diseases offer a link between the
more severe endpoints, such as increased mortality and hospital admissions or emergency room
visits for respiratory diseases, and less serious effects such as respiratory symptoms and
decreased lung function. These new studies also indicate the potentially more widespread public
health impact of exposure to PM (CD, p. 8-194).
As shown in Figure 3-2, associations between PM25 and hospitalization or emergency
room visits for the general category of respiratory diseases that are all positive and statistically
significant, while the results for individual disease categories (COPD, pneumonia, and asthma)
are less consistent, perhaps due to smaller sample sizes for the specific categories. Associations
with the general category of cardiovascular diseases are also all positive and statistically
significant or nearly so, but again the results for specific diseases (ischemic heart disease,
dysrhythmia, congestive heart disease or heart failure, and stroke) are positive but often not
statistically significant. Similarly, associations between hospital admissions for respiratory and
cardiovascular diseases and PM10_2 5 are generally positive and the more precise estimates are
statistically significant. Overall, the CD finds that excess risks for cardiovascular admissions
range from about 1 to 10% per 25 |ig/m3 PM2 5 or PM10_2 5 (CD, p. 8-310). For total respiratory or
COPD admissions, risk estimates tend to fall in the range of 5 to 15% per 25 |ig/m3 PM2 5 or
PM10.2.5 (CD, p. 8-193).
Many studies using PM10 or other PM indicators have been conducted in areas where fine
particles are the dominant fraction of PM10; results of these studies would likely be reflective of
associations with fine particles. In the last review, staff recognized that information about the
effects of thoracic coarse particles can also come from studies linking health effects with PM10 in
areas where thoracic coarse particles are the dominant fraction of PM10. Evidence available at
that time suggested that aggravation of asthma and respiratory infections and symptoms were
associated with PM10 in areas where thoracic coarse particles were dominant, such as Anchorage,
AK, and southeast Washington (62 FR 38679). Staff observes that several recent studies have
also been conducted in urban areas where thoracic coarse particles are the dominant fraction of
PM10 such as Reno, NV; Tucson, AZ; and Anchorage, AK, and these findings support the
evidence from the limited group of studies that have found associations between measured
3-21
-------
PM10_2 5 and morbidity. In these areas, most of which have levels exceeding the PM10 standards,
statistically significant associations have been reported between PM10 and increased
hospitalization for cardiovascular diseases (Schwartz, 1997), hospitalization for respiratory
diseases (Chen et al., 2000) and medical visits for asthma (Choudhury et al., 1997).
3.3.2.2 Effects on the Respiratory System from Short-term Exposures
As was found in the last review, some significant associations have been reported
between increased respiratory symptoms and decreased lung function and short-term exposures
to PM (section 8.3.3 in the CD). For asthmatic subjects, associations were reported between
PM10 and PM25 and decreases in lung function measures (e.g., decreased peak expiratory flow
rate); some but not all of the associations reached statistical significance. In addition, positive
associations were reported between PM10 and PM2 5 and one or more of a range of respiratory
symptoms (e.g., cough, wheeze, shortness of breath), but the findings were less consistent than
those for lung function (CD, p. 8-199). In studies of nonasthmatic subjects, while inconsistent
results were reported for changes in lung function, there were generally positive associations for
respiratory symptoms that often were not statistically significant. Generally similar results were
found for both PM10 and PM2 5 (CD, p. 8-206).
Few studies of respiratory symptoms and lung function have included both PM25 and
PM10_2 5 data. The CD summarizes findings from a Six Cities study analysis (Schwartz and Neas,
2000), a study in Philadelphia (Neas et al., 1999) and a study in Kupio, Finland (Tiittanen et al.,
1999). The findings of these studies suggest roles for both fine and thoracic coarse PM in
reduced lung function and increased respiratory symptoms (CD, p. 8-313). For example, in the
Six Cities study, lower respiratory symptoms were found to be significantly increased for
children with PM2 5 but not with PM10_2 5, while the reverse was true for cough. When both PM2 5
and PM10_2 5 were included in models, the effect estimates were reduced for each, but PM2 5
retained significance in the association with lower respiratory symptoms and PM10_2 5 retained
significance in the association with cough (Schwartz and Neas, 2000). The new epidemiologic
studies continue to show effects of short-term exposure to PM10 and PM2 5 and offer additional
evidence for associations between PM10_25 and respiratory morbidity (CD, p. 8-312).
The CD finds that the recent epidemiologic findings are consistent with those of the
previous review in showing associations with both respiratory symptom incidence and decreased
lung function (CD, p. 9-70). PM10 and PM2 5 were associated with small decreases in lung
function and increases in respiratory symptoms, though the associations were not always
statistically significant, and a few new studies reported associations between PM10_2 5 and
respiratory morbidity. The findings from studies of physicians' office visits for respiratory
diseases offer new evidence of acute respiratory effects with exposure to ambient PM that is
coherent with evidence of increased respiratory symptoms and admissions/visits to the hospital
or emergency room for respiratory disease.
3-22
-------
3.3.2.3 Effects on the Respiratory System from Long-term Exposures
In the last review, several studies had reported that long-term PM exposure was linked
with increased respiratory disease and decreased lung function. One study, using data from 24
U.S. and Canadian cities ("24 Cities" study), reported associations with these effects and long-
term exposure to fine particles or acidic particles, but not with PM10 exposure (Dockery et al.,
1996; Raizenne et al., 1996). The 1996 Staff Paper included further staff evaluation of the data
from this study that suggested that lung function decrements were not associated with long-term
exposure to thoracic coarse particles (EPA, 1996b, p. V-67a).
Several new epidemiologic analyses have been conducted on long-term pollutant
exposure effects on respiratory symptoms or lung function in the U.S.; numerous new European,
Asian, and Australian studies have also been published. In the U.S., studies have been based on
data from two cohort studies, cohorts of schoolchildren in 12 Southern California Communities
and an adult cohort of Seventh Day Adventists (AHSMOG). Results for the new studies,
together with the findings available in the last review, are presented in Appendix 3B.
In general, these studies have indicated that long-term exposure to PM2 5 is associated
with reduced lung function growth and increased risk of developing chronic respiratory illness
(CD, p. 8-313). In section 8.3.3.2.2, the CD describes results from Southern California cohorts,
where significant decreases in lung function growth were associated with increasing exposure to
PM10, PM25 and PM10_25 in one analysis (Gauderman et al., 2000), while in a second cohort of
children recruited in this study there were decreases in lung function growth with long-term
exposure to PM10 and PM2 5 (PM10_2 5 data were not included in this study) but the results were
generally not statistically significant (Gauderman et al., 2002). In an analysis of cohort
participants who moved during the course of the study, those who moved to areas with lower PM
concentrations (using PM10 as the indicator) showed increased lung function growth, whereas
lung function growth decreased in the group of children who moved to areas with high pollution
levels (Avol et al., 2001; CD, p. 8-213). A number of long-term studies of respiratory effects
also have been conducted in non-North American countries, and many report significant
associations between indicators of long-term PM exposure and either decreases in lung function
or increased respiratory disease prevalence (Table 8-B8 of the CD).
Considered together, the CD finds that the long-term exposure studies on respiratory
morbidity reported positive and statistically associations between fine particles or fine particle
components and lung function decrements or chronic respiratory diseases, such as chronic
bronchitis (CD pp. 8-313, 8-314). The CD observes that little evidence is available on potential
effects of long-term to exposure to PM10_25 (CD pp. 8-313, 8-314).
3.3.2.4 Effects on the Cardiovascular System
In contrast with the limited information available in the previous review, the CD observes
that new epidemiologic studies provide much more evidence of effects on the cardiovascular
system with short-term exposures to PM, particularly PM10 and PM2 5 (CD, p. 9-67).
3-23
-------
Epidemiologic studies have reported associations between short-term exposures to ambient PM
(often using PM10) and measures of changes in cardiac function such as arrhythmia, alterations in
electrocardiogram (ECG) patterns, heart rate or heart rate variability changes, though the CD
urges caution in drawing conclusions regarding the effects of PM on heart rhythm (CD, p.
8-166). Recent studies have also reported increases in blood components or biomarkers such as
increased levels of C-reactive protein and fibrinogen (CD, p. 8-169). In addition, one new study
reported an association between PM25 and onset of myocardial infarction, though another study
reported no significant associations between PM10 and sudden cardiac death (CD, pp. 8-165 to
8-166). Several of these studies report significant associations between various cardiovascular
endpoints and short-term PM25 exposures; only one of the new set of studies included PM10_25, in
which significant associations were reported between onset of myocardial infarction and short-
term PM25 exposures but not with PM10_25 exposures (CD, p. 8-165; Peters et al., 2001). These
new epidemiologic findings can provide some insight into potential biologic mechanisms that
underlie associations between short-term PM exposure and cardiovascular mortality and
hospitalization that have been reported previously (see Section 3.2).
3.3.3 Developmental effects
Some new evidence is available that is suggestive of adverse effects of exposure to PM
and gaseous co-pollutants on prenatal development, including both mortality and morbidity
effects. Several recent studies have shown significant associations between PM10 concentration
averaged over a month or a trimester of gestation and risk of intrauterine growth reduction
(AEGIR) and low birth weight. In addition, several new studies have suggested that infant
mortality may be associated with exposure to PM and gaseous co-pollutants during gestation.
The CD concludes that these effects are emerging as potentially more important than was
appreciated in the 1996 CD, but the evidence is still preliminary regarding these effects (CD, pp.
8-347).
3.3.4 Summary
In summary, exposure to various PM indicators is associated with a broad range of
cardiovascular and respiratory health endpoints. Newer studies report associations between
short-term exposure to various indicators of PM and cardiopulmonary mortality, hospitalization
or emergency department visits, and respiratory symptoms. In addition, there is now evidence
for associations with cardiovascular health outcomes, such as myocardial infarction or
physiological changes such as C-reactive protein increases. There are also a broader range of
respiratory health effects associated with exposure to various indicators of PM than those
previously documented. These effects include visits to physicians or clinics for treatment of
respiratory illnesses (CD, p. 9-23).
3-24
-------
More specifically, the epidemiologic evidence includes associations between short-term
exposure to PM2 5 and cardiorespiratory mortality, hospitalization and emergency department
visits for respiratory diseases, respiratory symptoms and decreased lung function, as well as
effects on the cardiovascular system, including changes in physiological indicators or biomarkers
for cardiovascular health (CD, pp. 8-338, 8-342). New studies also build upon previous
evidence for associations between long-term exposure to fine particles and cardiopulmonary
mortality or respiratory morbidity, with new evidence suggesting that long-term exposure to fine
particles is associated with lung cancer mortality (CD, p. 8-345).
Epidemiologic studies have linked short-term exposure to PM10_2 5 with respiratory
morbidity, such as hospitalization or respiratory symptoms, with suggested associations with
mortality in some areas. Available studies have not supported a link between long-term
exposure to PM10_2 5 and mortality or morbidity.
3.4 INTEGRATIVE ASSESSMENT OF HEALTH EVIDENCE
In Chapter 9, the CD assesses the new health evidence, integrating findings from
epidemiologic studies with experimental (e.g., dosimetric and toxicologic) studies, to make
judgments about the extent to which causal inferences can be made about observed associations
between health endpoints and various indicators or constituents of ambient PM, acting alone
and/or in combination with other pollutants. In evaluating the evidence from epidemiologic
studies in section 9.2.2, the CD focuses on well-recognized criteria, including (1) the strength of
reported associations; (2) the robustness of reported associations to the use of alternative model
specifications, potential confounding by co-pollutants, and exposure misclassification related to
measurement error; (3) the consistency of findings in multiple studies of adequate power, and in
different persons, places, circumstances and times; (4) temporality between exposure and
observed effects; (5) the nature of concentration-response relationships; and (6) information
from so-called natural experiments or intervention studies (CD, p. 9-23). Integrating more
broadly across epidemiologic and experimental evidence in section 9.2.3, the CD also focuses on
the coherence and plausibility of observed PM-related health effects to reach judgments about
causality. The following discussion summarizes the conclusions and judgments from the CD's
integrative assessment.
3.4.1 Strength of Associations
The strength of associations most directly refers to the magnitude of the reported relative
risk estimates. Taking a broader view, the CD draws upon the criteria summarized in a recent
report from the U.S. Surgeon General, which define strength of an association as "the magnitude
of the association and its statistical strength" which includes assessment of both effect estimate
size and precision, which is related to the statistical power of the study (CD, p. 9-6; CDC, 2004).
In general, when associations are strong in terms of yielding large relative risk estimates, it is
3-25
-------
less likely that the association could be completely accounted for by a potential confounder or
some other source of bias (CDC, 2004). With associations that yield small relative risk estimates
it is especially important to consider potential confounding and other factors in assessing
causality.
As observed in the previous PM NAAQS review, in historical air pollution episodes with
very high concentrations, reported relative risks were quite large.7 In more recent studies with
much lower ambient concentrations, the CD observes that the associations reported between
health effects and PM yield much smaller relative risk estimates (CD, p. 9-24). Focusing on the
results from more precise mortality studies done in the U.S. and Canada, the CD reports that
associations with short-term exposure, expressed as relative risks, are in the range of about 1.02
to 1.06 per 25 |ig/m3 PM25 or PM10_25 (CD, p. 9-28), while relative risks associated with long-
term exposure range upward to about 1.2 per 25 |ig/m3 PM25 for cardiovascular mortality (CD, p.
8-117). Regarding the size of relative risk estimates, the CD states: "In contrast with the marked
increase in health effects observed during historic episodes of very high air pollution levels,
relatively small effect estimates would generally be expected with current ambient PM
concentrations in the United States. The etiology of most air pollution-related health outcomes
is multifactorial, and the impact of ambient air pollution exposure on these outcomes may be
small in comparison to that of other risk factors." (CD, p. 9-24). Thus, while the associations
reported in the more recent body of epidemiologic studies are appropriately characterized as
being weak in terms of the magnitude of the relative risk estimates, such weak associations are
generally coherent with outcomes that may reasonably be expected.
In considering both the magnitude and statistical strength of the associations, the CD
observes a pattern of positive and often statistically significant associations for cardiovascular
and respiratory health outcomes with short-term exposure to PM10 and PM2 5 with estimates of
similar magnitude but less precision with PM10_25 (CD, p. 9-32). Of particular note are several
multi-city studies that have yielded relative risk estimates for associations between short-term
exposure to various indices of PM and mortality or morbidity that, while small in size, have great
precision due to the statistical power of the studies. Such associations are strong relative to the
precision of the studies; that is, the associations were strong enough to have been reliably
measured by the studies such that many of the associations can be distinguished from the null
hypothesis with statistical confidence.
In considering the strength of the associations between long-term exposure to fine
particles and mortality or morbidity, the CD concludes that the magnitude and precision of
associations with long-term exposure to PM2 5 constitute "strong evidence" for associations with
7 For example, in the week of the well-documented episode that occurred in London in 1952, when PM
concentrations exceeded 500 ug/m3, the relative risk of all-cause mortality was 2.6, and the relative risk for
bronchitis mortality was 9.3 (62 FR 38659).
3-26
-------
mortality and "fairly strong evidence" for associations with respiratory morbidity. However, the
little evidence available for PM10_2 5 provide no evidence for associations with mortality and
allow no conclusions to be drawn regarding associations with morbidity (CD, p. 9-34).
3.4.2 Robustness of Associations
In section 9.2.2.2, the CD evaluates the robustness of epidemiologic associations in part
by considering the effect of differences in statistical model specification, potential confounding
by co-pollutants and exposure error on PM-health associations. The 1996 CD included an
assessment of evidence then available on these issues, and concluded that the effects observed in
epidemiologic studies "cannot be wholly attributed to" issues such as confounding by co-
pollutants, differing model specifications, or measurement error (EPA, 1996a, p. 13-92). These
issues have been further evaluated in many new studies available in this review.
As discussed below in section 3.6.3, the CD assesses the findings of studies that
evaluated alternative modeling strategies, with a particular focus on the recent set of analyses to
address issues related to the use of GAM in time-series epidemiologic studies. The reanalyses
included the use of alternative statistical models and methods of control for time-varying effects,
such as weather or season. In the results of these reanalyses, some studies showed little change
in effect estimates, while others reported reduced effect estimate size, though the CD observes
that the reductions were often not substantial (CD, p. 9-35). Overall, the CD concludes that
associations between short-term exposure to PM and various health outcomes are generally
robust to the use of alternative modeling strategies, though further evaluation of alternative
modeling strategies is warranted (id.). The CD also notes that the results of reanalyses indicated
that effect estimates were more sensitive to the modeling approach used to account for temporal
effects and weather variables than to the GAM specifications, and recommended further
exploration of alternative modeling approaches for time-series analyses (CD, pp. 8-236 to
8-237).
In addition, the reanalysis and extended analyses of data from prospective cohort studies
have shown that reported associations between mortality and long-term exposure to fine particles
are robust to alternative modeling strategies. As stated in the reanalysis report, "The risk
estimates reported by the Original Investigators were remarkably robust to alternative
specifications of the underlying risk models, thereby strengthening confidence in the original
findings" (Krewski et al., 2000, p. 232).
The CD also included extensive evaluation of the sensitivity of PM-health responses to
confounding by gaseous co-pollutants, as discussed in detail in section 8.4.3 of the CD, and more
briefly below in section 3.6.4. In the new multi-city studies, as well as in many of the single-city
studies, health outcome associations with short-term exposures to PM10 PM2 5 and PM10_2 5 are
little changed in multi-pollutant models including one or more of the gaseous co-pollutants (CD,
p. 8-253). However, in some single-city analyses, PM-health outcome associations were
3-27
-------
attenuated in multi-pollutant models; the CD observes that collinearity between co-pollutants can
make interpretation of multi-pollutant models difficult (id.). Similarly, in the prospective cohort
studies, associations between long-term exposure to PM2 5 and mortality were generally not
sensitive to inclusion of co-pollutants, except for SO2, which was also associated with mortality
(CD, p. 8-136). Overall, the CD concludes that these studies indicate that effect estimates for
associations between mortality and morbidity and various PM indices are robust to confounding
by co-pollutants, while recognizing that disentangling the effects attributable to various
pollutants within an air pollution mixture is challenging (CD, p. 9-37).
Finally, as discussed in section 3.6.2, a number of recent studies have evaluated the
influence of exposure error on PM-health associations. This includes both consideration of error
in measurements of PM, and the degree to which measurements from an individual monitor
reflect exposures to the surrounding community. As further discussed in section 3.6.2, several
studies have shown that fairly extreme conditions (e.g., very high correlation between pollutants
and no measurement error in the "false" pollutant) are needed for complete "transfer of
causality" of effects from one pollutant to another (CD, p. 9-38). In comparing fine and thoracic
coarse particles, the CD observes that exposure error is likely to be more important for
associations with PM10_2 5 than with PM2 5, since there is generally greater error in PM10_2 5
measurements, PM10_2 5 concentrations are less evenly distributed across a community, and less
likely to penetrate into buildings (CD, p. 9-38). Therefore, while the CD concludes that
associations reported with PM10 PM2 5 and PM10_2 5 are generally robust, the CD recognizes that
factors related to exposure error may result in reduced precision for epidemiologic associations
with PM10.2 5 (CD, p. 9-46).
3.4.3 Consistency
Consistency refers to the persistent finding of an association between exposure and
outcome in multiple studies of adequate power in different persons, places, circumstances and
times (CDC, 2004). The 1996 CD reported associations between short-term PM exposure and
mortality or morbidity from studies conducted in locations across the U.S. as well as in other
countries, and concluded that the epidemiologic data base had "general internal consistency"
(EPA, 1996a, p. 13-30). This epidemiologic data base has been greatly expanded with numerous
studies conducted in single locations, as well as several key multi-city studies. As described
above, the CD finds that the epidemiologic studies generally report positive and often
statistically significant associations with various cardiorespiratory health outcomes. The larger
body of evidence also has shown more variability in effect estimate size for a given health
outcome than was apparent in the last review.
New multi-city studies have allowed evaluation of consistency in effect estimates across
geographic locations, using uniform statistical modeling approaches. In the NMMAPS results,
effect estimates for many individual cities exhibited wide confidence ranges, with varied effect
3-28
-------
estimate sizes, that suggested potentially more heterogeneity in effect estimates across cities than
had been seen with single-city studies in the last review. However, the authors observed that
there was no statistically significant heterogeneity across the effect estimates in the NMMAPS
analyses (Samet et al., 2000; Dominici et al., 2003a). The Canadian multi-city study also
reported some limited evidence suggesting heterogeneity in responses for PM2 5 and PM10_2 5 in
the reanalysis to address GAM questions, whereas there had been no evidence of heterogeneity
in initial study findings (Burnett and Goldberg, 2003; CD, p. 9-39). Finally, in the European
multi-city study, there were differences seen between effect estimates from eastern and western
European cities in initial analyses, but these differences were less clear with reanalysis to address
GAM issues (CD, pp. 8-46 to 8-47; Katsouyanni et al., 2003). Overall, the new multi-city study
results suggest that effect estimates differ from one location to another, but the extent of
heterogeneity is not clear.
The CD discusses a number of factors that would be likely to cause variation in PM-
health outcomes in different populations and geographic areas in section 9.2.2.3. The CD
recognizes that differences might well be expected in effects across locations, and discusses
investigation of a number of factors that appeared to be associated with variation in effect
estimates, including indicators of exposure to traffic-related pollution and climate-related
increases in exposure to ambient pollution (CD, p. 9-39). Other factors might also be expected
to cause variation in observed effects between locations, including population characteristics that
affect susceptibility or exposure differences, distribution of PM sources, or geographic features
that would affect the spatial distribution of PM (CD, p. 9-41).
In addition, the CD observes that NMMAPS, while advantageous in including data from
many different locations with different climates and pollutant mixes, included many locations for
which the sample size (i.e., population size and PM10 data) was inherently smaller for a given
study period than that used in most single-city studies (CD, p. 9-40). The Canadian 8-city study,
as well, used PM data from a monitoring network that operated primarily on a l-in-6 day
collection schedule, although the data were available for a long time period. In general, the CD
observes that the use of data collected on every sixth day results in reduced statistical power,
resulting in less precision for estimated effect estimates for the individual cities and increased
potential variability in results (CD, p. 9-40).
Overall, the CD finds that "[fjocusing on the studies with the most precision, it can be
concluded that there is much consistency in epidemiologic evidence regarding associations
between short-term and long-term exposures to fine particles and cardiopulmonary mortality and
morbidity." (CD, p. 9-47). For short-term exposure to thoracic coarse particles, the CD
concludes that there is some consistency in effect estimates for hospitalization for cardiovascular
and respiratory causes, though fewer studies are available on which to make such an assessment
(CD, p. 9-47).
3-29
-------
3.4.4 Temporality
Temporality refers to the occurrence of a cause before its purported effect, and is most
relevant to studies of diseases that develop over time. This factor is difficult to investigate in
situations where the pollutant concentrations are correlated over time as is the case to some
degree in PM time series studies and to a greater degree in the cohort studies. The short-term
exposure studies evaluate associations between acute health outcomes and PM measured on a
daily or hourly basis. In many studies, associations have been reported between health events
and pollutants measured contemporaneously. For example, in studies of total and cardiovascular
mortality, the CD observes that effects have been most clearly linked with PM measured on the
same day or the preceding day (CD, p. 8-273). This would be expected for acute health effects,
however, it is difficult to characterize these associations in terms of temporality. Issues related
to the evaluation and selection of lag periods among studies are further discussed in section
3.6.5.
The studies of effects related to long-term PM exposures have generally used air quality
levels averaged over months or years as exposure indicators. It is important to recognize that
these studies do not test specifically for latency in an exposure-effect relationship. Instead, the
average PM levels are used to represent long-term exposure to ambient PM, and the exposure
comparisons are basically cross-sectional in nature (CD, p. 9-42). Thus, the long-term exposure
studies do not allow an assessment of the temporal relationship between exposure and health
outcome. Taken together, it is difficult to assess temporality in the available studies of both
short-term and especially long-term exposures to PM, given that PM concentrations are
generally correlated over time in any given area.
3.4.5 Nature of concentration-response relationships
This is an assessment of whether increases in the potential causal factor result in
increased effects, also referred to as a biologic gradient. In epidemiologic time-series analyses,
the results have consistently shown positive associations, indicating that increases in various PM
indicators are associated with increases in health outcomes (CD, pages 9-28 to 9-29). The
prospective cohort studies have also generally reported positive associations between long-term
exposure to PM, primarily PM2 5, and increases in mortality or morbidity (CD, pp. 8-344 to
8-345). The available toxicologic studies have generally not been designed to quantify dose-
response relationships (CD, p. 7-2). Among the studies reviewed in chapter 7 are some that
report no evidence of a dose-response gradient (CD, p. 7-152), while some do (CD, p. 7-155),
and the CD draws no overall conclusions regarding dose-response relationships from toxicologic
studies. Therefore, while epidemiologic studies provide clear indication of increasing response
with increasing concentration, no conclusions can be drawn from toxicologic evidence.
3-30
-------
3.4.6 Natural Experiment Studies
Few studies are available that assess the extent to which reductions in ambient PM
actually lead to reductions in health effects attributable to PM. As discussed in section 9.2.2.6 of
the CD, and in somewhat more detail in section 3.4.7 below, one set of studies in the Utah
Valley were conducted over a time period when a major source of PM was closed, resulting in
markedly decreased PM10 concentrations. An epidemiologic study reported that respiratory
hospital admissions decreased during the plant closure time period (CD, p. 8-131; Pope et al.,
1989). Newly available controlled human exposure and animal toxicology studies, using
particles extracted from stored PM10 sampling filters from the Utah Valley, have shown
inflammatory responses that are greater with extracts of particles collected during the time
period of source operation than when the source was closed (CD, p. 9-73). Epidemiologic
studies in Dublin, Ireland and Hong Kong also provides evidence for reduced relative risks for
mortality when PM (measured as BS or PM10) and/or SO2 were reduced as the result of
interventions aimed at reducing air pollution (CD, pp. 8-131 to 8-135). From this small group of
new studies, the CD concludes:
By providing evidence for improvement in community health following reduction in air
pollutant emissions, these studies add further support to the results of the hundreds of
other epidemiologic studies linking ambient PM exposure to an array of health effects.
Such studies showing improvements in health with reductions in emissions of ambient
PM and/or gaseous co-pollutants provide strong evidence that reducing emissions of PM
and gaseous pollutants has beneficial public health impacts. (CD, p. 9-45 to 9-46).
3.4.7 Coherence and Plausibility
Section 9.2.3 of the CD integrates and evaluates evidence from the different health
disciplines to draw conclusions regarding the coherence of effects observed in the cardiovascular
and respiratory systems, as well as evidence for biologic plausibility of these effects. The CD
finds that progress has been made in substantiating and expanding epidemiologic findings on
cardiovascular- and respiratory-related effects of PM, and in obtaining evidence bearing on the
biologic plausibility of observed effects and potential mechanisms of action for particles (CD, p.
9-49).
As was concluded in the previous review, in considering evidence from epidemiologic
studies using PM10 and other PM indicators, the CD finds coherence for effects on the
cardiovascular and respiratory systems. Figures 8-24 through 8-28 of the CD show effect
estimates for associations between short-term exposures to PM10 and a range of cardiovascular
and respiratory health endpoints from within the same geographic location. In addition, the CD
finds that epidemiologic studies report associations for PM2 5 with a broad range of effects on the
cardiovascular and respiratory systems, primarily from short-term exposure studies, but also
3-31
-------
supported by associations reported for long-term fine particle exposure with cardiovascular
mortality (CD, pp. 9-67).
As described briefly in section 3.2 above, and in more depth in Chapter 7 of the CD, the
findings of new toxicologic and controlled human exposure studies, while still limited, support a
number of potential biologic mechanisms or pathways for PM-related effects, and this evidence
is largely from studies of fine particles or fine particle components.
Focusing first on effects related to the cardiovascular system, in section 9.2.3.2.1, the CD
summarizes evidence from both epidemiologic and toxicologic studies on subtle changes in
cardiovascular health. These changes include increased blood fibrinogen and fibrin formation,
certain ECG parameters (e.g., heart rate variability or HRV), and vascular inflammation. The
CD notes that vascular inflammation induces release of C-reactive proteins and cytokines that
may cause further inflammatory responses which, on a chronic basis, could lead to
atherosclerosis.
Where a series of studies have been conducted in the same location, these studies can
provide evidence for coherence of effects, linking results from different study types for exposure
to PM in the same airshed. As discussed in the CD, in Boston, epidemiologic associations were
reported between PM25 and incidence of myocardial infarction, increases in recorded discharges
from implanted cardiovertex defibrillators, and decreases in HRV measures. Toxicologic studies
in Boston, using PM2 5 CAPs exposures in dogs, also suggested changes in cardiac rhythm with
PM2 5 mass and changes in blood parameters with certain PM2 5 components (CD, p. 9-68, 9-69).
The CD observes: "While many research questions remain, the convergence of evidence related
to cardiac health from epidemiologic and toxicologic studies indicates both coherence and
plausibility in this body of evidence." (CD, p. 9-78).
In the last review, evidence was available suggesting coherence of effects on the
respiratory system, and the CD finds that new epidemiologic and toxicologic studies expand
upon that knowledge (CD, p. 9-74). In locations where epidemiologic studies have been
conducted, toxicologic or controlled human exposure studies using exposures to concentrated
ambient particles have shown effects related to lung inflammation, though minimal effects on
lung function have been reported (CD, p. 9-72). As discussed in section 3.2, toxicologic and
controlled human exposure studies have provided substantial evidence that particles can cause
lung injury and inflammatory responses.
Interesting new evidence that links toxicologic and epidemiologic findings is available
from some "intervention studies" in the Utah Valley area. Epidemiologic studies in the Utah
Valley area observed that respiratory hospital admissions decreased during a period when a
major source of PM10 (a steel mill) was closed. More recent toxicologic and controlled human
exposure studies have used particles collected from this locale during the same time period, and
reported increased inflammatory responses with particles collected while the PM source was
operating than when it was closed. Several in vitro studies have also reported evidence of
3-32
-------
increased oxidative stress in lung cell cultures exposed to particles collected in Utah Valley. In
some toxicologic studies, the transition metal content of the particles appeared to be more
closely linked to reported effects than the quantity of particles (CD pp. 7-46 to 7-48). While
urging caution in interpreting the findings of the toxicologic studies where higher doses were
used, the CD concludes that "[t]he fact that instillation of ambient PM collected from different
geographical areas has been shown to cause pulmonary inflammation and injury tends to support
epidemiologic studies that report increased PM-associated respiratory effects living in some of
the same geographical areas" (CD, p. 7-48). Staff observes that, in contrast with most evidence
discussed here, this group of studies may well implicate thoracic coarse particles, since such
particles generally dominate PM10 concentrations in the Utah Valley area.
As was true in the last review, there is some coherence in the epidemiologic evidence
linking long-term exposure to fine particles with mortality and effects on the respiratory system.
Available toxicologic studies have generally not studied cardiopulmonary effects of long-term or
chronic exposures to ambient air pollution mixtures, so for the most part, no conclusions can be
drawn regarding biologic plausibility of observed effects with long-term PM25 exposures and
mortality from heart and lung diseases (CD, p. 9-69). However, for lung cancer, the CD
summarizes evidence that supports coherence and plausibility in the epidemiologic associations
reported between long-term exposures to fine particles and lung cancer mortality. The CD
discusses toxicologic evidence on mutagenic or genotoxic potential of ambient PM, particles
from wood and coal combustion, and particles from diesel and gasoline engine emissions (CD
Section 7.8). These toxicologic studies have provided evidence of mutagenicity or genotoxicity
with exposure to combustion-related particles or to ambient particles collected in Los Angeles,
Germany and the Netherlands (CD, p. 9-76). In addition, the Health Assessment Document for
diesel engine exhaust concludes that diesel engine exhaust, one source of PM emissions, is a
likely human carcinogen (EPA, 2002). On the results of the new epidemiologic studies, the CD
concluded "[o]verall, these new cohort studies confirm and strengthen the published older
ecological and case-control evidence indicating that living in an area that has experienced higher
PM exposures can cause a significant increase in RR of lung cancer incidence and associated
mortality" (CD, p. 8-318). A number of toxicologic studies, summarized in section 7.10.1 of the
CD, report evidence of genotoxicity or mutagenicity with particles. The CD also finds that the
evidence indicates that fine particles may be more mutagenic than thoracic coarse particles (CD,
p. 7-214), which is consistent with the evidence from epidemiologic studies. Considered with
the results of toxicologic studies, the CD finds that this new evidence supports the plausibility of
a relationship between fine particles and lung cancer mortality (CD, p. 9-78).
Less information is available to allow conclusions to be drawn about coherence or
plausibility for associations with PM10_2 5 Based on the epidemiologic evidence discussed
previously, the CD concludes that the results are suggestive of associations between short-term
exposure to PM10_2 5 and morbidity effects, especially effects on the respiratory system (CD, p.
3-33
-------
9-80). From the limited number of toxicologic studies using PM10_25, as noted in section 3.2,
there is some evidence supporting effects such as inflammation or oxidative stress. In addition,
allergic adjuvant effects were linked with road dust exposures, but coarse particle sample of
geologic origin, Mt. St. Helens ash, has not been linked with effects in toxicologic studies. As
discussed above, fractional deposition to the tracheobronchial region is greatest for thoracic
coarse particles in the size range of 4 to 6 jim (CD, p. 6-109). This would be consistent with
epidemiologic evidence linking PM10_2 5 with respiratory morbidity, such as increased respiratory
symptoms or risk of hospitalization for asthma.
3.4.8 Summary
The new evidence from epidemiologic studies builds upon the conclusions of the last
review regarding the strength, robustness and consistency of the evidence. While uncertainties
remain and the new studies raise some new questions, the CD concludes:
In conclusion, the epidemiological evidence continues to support likely causal
associations between PM2 5 and PM10 and both mortality and morbidity from
cardiovascular and respiratory diseases, based on an assessment of strength, robustness,
and consistency in results. For PM10_2 5, less evidence is available, but the studies using
short-term exposures have reported results that are of the same magnitude as those for
PM10 and PM2 5, though less often statistically significant and thus having less strength,
and the associations are generally robust to alternative modeling strategies or
consideration of potential confounding by co-pollutants. (CD, p. 9-48).
Much more evidence is now available related to the coherence and plausibility of effects
than in the last review. For short-term exposures, the CD finds that the integration of evidence
from epidemiologic and toxicologic studies indicates both coherence and plausibility of effects
on the cardiovascular and respiratory systems, particularly for fine particles (CD, p. 9-78). Also,
there is evidence supporting coherence and plausibility for the observed associations between
long-term exposures to fine particles and lung cancer mortality (CD, p. 9-79). The smaller body
of evidence on thoracic coarse particles, especially the limited evidence from toxicologic studies,
provides only limited evidence of coherence for effects of thoracic coarse particles.
Epidemiologic and dosimetric evidence, along with limited support from toxicologic studies,
support associations between PM10_2 5 and the respiratory system, with less evidence available on
cardiovascular effects.
Finally, the evaluation of these criteria leads the CD to draw conclusions regarding
causality of effects seen with fine or with thoracic coarse particles. Overall, the CD concludes
that the available evidence supports the general conclusion that PM2 5 or fine particle components
are "likely causally related to cardiovascular and respiratory mortality and morbidity" (CD, p.
9-79). For PM10_25, the "much more limited body of evidence is suggestive of associations
3-34
-------
between short-term (but not long-term) exposures . . . and various mortality and morbidity
effects observed at times in some locations." (CD, p. 9-79).
3.5 PM-RELATED IMPACTS ON PUBLIC HEALTH
The following discussion draws from sections 9.2.4 and 9.2.5 of the CD to characterize
subpopulations potentially at risk for PM-related effects and potential public health impacts
associated with exposure to ambient PM. In particular, the potential magnitude of at-risk
population groups is discussed, along with other key considerations related to impacts on public
health, such as the concept of "mortality displacement" or "harvesting."
3.5.1 Potentially Susceptible and Vulnerable Subpopulations
The CD summarizes information on potentially susceptible or vulnerable groups in
section 9.2.4. As described there, the term susceptibility refers to innate (e.g., genetic or
developmental) or acquired (e.g., personal risk factors, age) factors that make individuals more
likely to experience effects with exposure to pollutants. A number of population subgroups
have been identified as potentially susceptible to health effects as a result of PM exposure,
including people with existing heart and lung diseases, including diabetes, and older adults and
children. In addition, new attention has been paid to the concept of some population groups
having increased vulnerability to pollution-related effects due to factors including socioeconomic
status (e.g., reduced access to health care or low socioeconomic status) or particularly elevated
exposure levels, such as residence near sources such as roadways (CD, p. 9-81). Most available
studies have used PM10 or other measures of thoracic particles, with little specific evidence on
potential susceptibility to effects of PM25 or PM10_2 5.
A good deal of evidence indicates that people with existing heart or lung diseases are
more susceptible to PM-related effects. In addition, new studies have suggested that people with
diabetes, who are at risk for cardiovascular disease, may have increased susceptibility to PM
exposures. This body of evidence includes findings from epidemiologic studies that associations
with mortality or morbidity are greater in those with preexisting conditions, as well as evidence
from toxicologic studies using animal models of cardiopulmonary disease (CD, section 9.2.4.1).
In addition, as described previously in section 3.2, dosimetric evidence indicates that deposition
of particles is increased, and can be focused in "hot spots" in the respiratory tract, in people with
chronic respiratory diseases.
Two age groups, older adults and the very young, are also potentially at greater risk for
PM-related effects. Epidemiologic studies have generally not shown striking differences
between adult age groups. However, some epidemiologic studies have suggested that serious
health effects, such as premature mortality, are greater among older populations (CD, p. 8-328).
In addition, preexisting respiratory or cardiovascular conditions are more prevalent in older
3-35
-------
adults than younger age groups; thus there is some overlap between potentially susceptible
groups of older adults and people with heart or lung diseases.
Epidemiologic evidence has reported associations with emergency hospital admissions
for respiratory illness and asthma-related symptoms in children (CD, p. 8-328). The CD also
observes that several factors may make children susceptible to PM-related effects, including the
greater ventilation per kilogram body weight in children and the fact that children are more
likely to be active outdoors and thus have greater exposures (CD, p. 9-84). In addition, the CD
describes a limited body of new evidence from epidemiologic studies for potential PM-related
health effects in infants, using various PM indicators. Results from this body of evidence,
though mixed, are suggestive of possible effects; more research is needed to further elucidate the
potential risks of PM exposure for these health outcomes (CD, p. 8-222).
The CD also discusses other potentially susceptible groups for which less evidence is
available. Gender is a potential factor, and there are suggested differences in epidemiologic
study results, but the findings are not always consistent (CD, section 9.2.4.4). There is some new
suggestive evidence on genetic susceptibility to air pollution, but no conclusions can be drawn at
this time (CD section 9.2.4.3).
In considering populations groups that might be more vulnerable to PM-related effects,
there is some new evidence from epidemiologic studies that people from lower socioeconomic
strata, or who have greater exposure to sources such as roadways, may be more vulnerable to PM
exposure. Such population groups would be considered to be more vulnerable to potential
effects on the basis of socioeonomic status or exposure conditions, as distinguished from
susceptibility due to biologic or individual health characteristics (CD, section 9.2.4.5).
In summary, there are several population groups that may be susceptible or vulnerable to
PM-related effects. These groups include those with preexisting heart and lung diseases, older
adults and children. Emerging evidence indicates that people from lower socioeconomic strata
or who have particularly elevated exposures may be more vulnerable to PM-related effects. The
available evidence does not generally allow distinctions to be drawn between the PM indicators,
in terms of which groups might have greater susceptibility or vulnerability to PM2 5 and/or
PM,
L10-2.5
3.5.2 Potential Public Health Impact
As summarized above, there are several populations groups that may be susceptible or
vulnerable to effects from exposure to PM. The CD provides estimates of the size of population
subgroups, such as young children or older adults, and people with pre-existing heart or lung
diseases (CD, section 9.2.5.1) that are the subpopulations considered to be likely susceptible to
the effects of PM exposure. As shown in Table 9-4 of the CD, approximately 22 million people,
or 11% of the U.S. population, have received a diagnosis of heart disease, about 20% of the
population have hypertension and about 9% of adults and 11% of children in the U.S. have been
3-36
-------
diagnosed with asthma. In addition, about 26% of the U.S. population are under 18 years of age,
and about 12% are 65 years of age or older (CD, p. 9-89). The CD concludes that combining
fairly small risk estimates and small changes in PM concentration with large groups of the U.S.
population would result in large public health impacts (CD, p. 9-93).
These health statistics also generally illustrate increasing frequency of less serious health
outcomes that would be expected in a "pyramid of effects." Along the spectrum of severity, it is
expected that incidence or frequency of health endpoints would be larger for the less severe
effects, such as respiratory symptoms or the more subtle measures of cardiovascular health such
as levels of C-reactive protein. In contrast, with more severe health outcomes, such as
hospitalization or mortality, lower incidence would be expected.
One issue that is important for interpreting the public health implications of the
associations reported between mortality and short-term exposure to PM is whether mortality is
occurring only in very frail individuals (sometimes referred to as "harvesting"), resulting in loss
of just a few days of life expectancy. A number of new analyses are discussed in the CD
(section 8.4.10.1) that assess the likelihood of such "harvesting" occurring in the short-term
exposure studies. Overall, the CD concludes from the time-series studies that there appears to be
no strong evidence to suggest that short-term exposure to PM is only shortening life by a few
days (CD, p. 8-334).
In addition to the evidence from short-term exposure studies discussed above, one new
report used the mortality risk estimates from the ACS prospective cohort study to estimate
potential loss of life expectancy from PM-related mortality in a population. The authors
estimated that the loss of population life expectancy associated with long-term exposure to PM25
was substantial, on the order of a year or so (CD, p. 9-94). Taken together, these results suggest
that exposure to ambient PM, especially PM2 5, can have substantial public health impacts (CD,
p. 9-93). Furthermore, in the ACS cohort, the strongest associations between PM2 5 and
mortality were among the less educated participants who form a relatively small portion of the
total study cohort. If the education distribution were adjusted to reflect the education distribution
in the general U.S. population, the summary effect estimate would increase.
3.6 ISSUES RELATED TO QUANTITATIVE ASSESSMENT OF EPIDEMIOLOGIC
EVIDENCE
The 1996 CD included extensive discussions of methodological issues for epidemiologic
studies, including questions about model specification or selection, co-pollutant confounding,
measurement error in pollutant measurements, and exposure misclassification. Based on
information available in the last review, the 1996 PM CD concluded that PM-health effects
associations reported in epidemiologic studies were not likely an artifact of model specification,
since analyses or reanalyses of data using different modeling strategies reported similar results
3-37
-------
(EPA 1996a, p. 13-92). Little information was available at that time to allow for evaluation of
these and other related methodological issues.
A large number of studies now available in this review have provided new insights on
these and other issues as evaluated in Chapters 8 and 9 of the CD. The following discussion
builds upon the CD's evaluation of key methodological issues related to epidemiologic studies as
a basis for staff judgments specifically regarding the use of epidemiologic evidence in
quantitative assessments, as discussed in Chapters 4 and 5.
This section addresses a number of key methodological issues. Section 3.6.1 discusses
issues related to air quality data used in epidemiologic studies, and section 3.6.2 discusses the
potential impact of measurement error and exposure error, related to the use of ambient air
concentrations as indicators of population exposures, on epidemiologic studies. Section 3.6.3
addresses statistical modeling and model specifications used in epidemiologic studies. Section
3.6.4 addresses the issue of potential confounding by co-pollutants, as it relates to staff
conclusions about the use of specific study results in quantitative assessments. Section 3.6.5
includes discussion of several topics related to the exposure periods used in epidemiologic
studies, with an emphasis on the question of lag periods. In section 3.6.6, the form of
concentration-response relationships in both short-term and long-term exposure studies is
discussed, as is evidence related to the potential existence of population threshold levels for
effects.
3.6.1 Air Quality Data in Epidemiologic Studies
In general, epidemiologic studies use ambient measurements to represent population
exposures to PM of ambient origin. This section discusses some considerations with regard to
the ambient PM measurements. First, staff observes that PM measurements from several
different monitoring methods were used in epidemiologic studies. Many studies have used PM2 5
and PM10_2 5 measurements from dichotomous samplers or Harvard impactors, as well as PM2 5
and PM10 measurements from co-located TEOMs or BAMs, and other methods (see Chapter 2
for more detailed descriptions of monitors). In reviewing results from studies using various
monitoring methods for PM2 5 and PM10_2 5, staff finds that there appear to be no systematic
differences in the effect estimates related to the use of differing monitoring methods.
In considering the frequency of PM data collection, staff observes that it can have a
systematic effect on the results reported from epidemiologic analyses. The CD discusses the use
of less-than-everyday monitoring data as a source of uncertainty for time-series analyses (CD, p.
8-296). Many such studies were conducted in areas where PM was monitored on a daily basis;
in fact, the availability of every-day monitoring is cited as a basis for study location in a number
of reports. This is particularly true for panel studies on respiratory or cardiovascular symptoms,
all of which use daily PM monitoring data, though generally for shorter time periods. However,
staff observes that a small number of the recent studies have been based on less frequently
3-38
-------
collected data. Data collection frequency is one component of statistical power for time-series
studies, and missing data would result in increased uncertainty and reduced precision in study
results. In addition, for either PM2 5 or PM10_2 5, one would expect that a substantial proportion of
missing data may complicate time-series analyses (CD, p. 9-41). As illustrated in the CD, effect
estimates for PM10 and mortality varied in size and statistical significance in a series of analyses
of data collected on a l-in-6 day schedule (CD, p. 8-297). The CD presents results from a study
in Chicago, IL, where a significant association was reported between PM10 and mortality using
data collected on a daily basis (Ito et al., 1996). However, when the data set was divided into 6
subsets representing l-in-6 day monitoring frequency, the effect estimates for the PM10-mortality
association were quite variable in size and more uncertain. Consistent with the CD's observation
that uncertainty is increased in studies using infrequently collected PM data, staff judges that
greater weight should be placed on those studies with daily or near-daily PM data collection in
drawing quantitative conclusions (CD, p. 9-41).
3.6.2 Measurement Error and Exposure Error
Measurement error, or uncertainty in the air quality measurements can be an important
source of uncertainty in epidemiologic associations with PM10_2 5 or PM25. The CD summarizes
the findings of several new analyses that show the potential influence of differential
measurement error on epidemiologic analysis results, for either PM with gaseous pollutants, or
PM10_25 and PM25 as separate pollutants (section 8.4.5). Several studies used simulation analyses
of a "causal" pollutant and a "confounder" with differing degrees of measurement error and
collinearity between the pollutants. These studies found that, in some circumstances, a causal
variable measured with error may be overlooked and its significance transferred to a surrogate.
However, for "transfer of apparent causality" from the causal pollutant to the confounder to
occur, there must be high levels of both measurement error in the causal variable and collinearity
between the two variables (CD, p. 8-282, 8-283). The conditions required for the error to
substantially influence the epidemiologic findings are severe and unlikely to exist in current
studies. Thus, while the potential remains for differential error in pollutant measurements to
influence the results of epidemiologic studies, it is unlikely that the levels of measurement error
and correlation between pollutants reported in existing studies would result in transfer of
apparent causality from one pollutant to another (CD, p. 9-38).
One analysis applied measurement error models to data from the Harvard Six Cities
study, specifically testing relationships between mortality and either fine or thoracic coarse
particles (Carrothers and Evans, 2000). The authors identified several variables that could result
in biased effect estimates for fine- or coarse-fraction particles: the true correlation of fine- and
coarse-fraction particles, measurement errors for both, and the underlying true ratio of the
toxicity of fine- and coarse-fraction particles. The existence of measurement error and
collinearity between pollutants could result in underestimation of the effects of the less well-
3-39
-------
measured pollutant. However, the authors conclude "it is inadequate to state that differences in
measurement error among fine and coarse particles will lead to false negative findings for coarse
particles. If the underlying true ratio of the fine and coarse particle toxicities is large (i.e.,
greater than 3:1), fine particle exposure must be measured significantly more precisely in order
not to underestimate the ratio of fine particle toxicity versus coarse particle toxicity" (Carrothers
and Evans, 2000, p. 72; CD, p. 8-286). These analyses, using data from a study in which
significant associations were reported for mortality with PM2 5, but not with PM10_2 5, indicate that
it is unlikely that measurement error in one PM measurement will result in "false negative"
results for coarse particles or "false positive" results for fine particles (CD, p. 8-286). Thus, for
either PM2 5 or PM10_2 5 measurement error is not likely to be falsely attributing effects from one
pollutant to another pollutant in the existing epidemiologic studies.
However, it must be recognized that measurement error is a larger issue for PM10_2 5 than
for fine particles, especially when PM10_2 5 concentrations are calculated as the difference
between PM10 and PM2 5 measurements (see section 2.4.3). It is likely that measurement error
would increase the uncertainty of an epidemiologic association. With increased error in PM10_2 5
monitoring methods, any reported epidemiologic associations would be less likely to reach
statistical significance (CD, p. 5-126). Thus, a set of positive but generally not statistically
significant associations between PM10_2 5 and a health outcome could be reflecting a true
association that is measured with error. Decreases in study precision would also occur even if
gravimetric PM10_25 were perfectly measured, but the sources and relative composition of the
coarse particles were highly variable. In evaluating the implications of the epidemiologic studies
showing effects of PM10_25, therefore, staff places more emphasis on the pattern of results in a
series of studies than on the statistical significance of any single effect estimate.
Exposure error is an issue that is closely linked with the preceding discussion of PM air
quality monitoring. Concentrations measured at ambient monitoring stations are generally used
to represent a community's exposure to ambient PM. For time-series studies, the emphasis is on
the temporal (usually daily) changes in ambient PM. In cohort or cross-sectional studies, air
quality data averaged over a period of months to years are used as indicators of a community's
long-term exposure to ambient PM and other pollutants.
As discussed in section 2.7, one component of exposure error is how evenly distributed
PM is across a community, as indicated by levels at different monitoring sites; another
component is how well particles penetrate from ambient air into indoor environments. Several
factors affect how readily particles can move into buildings and remain suspended in indoor air.
In general, fine particles move indoors and remain suspended more easily than do thoracic
coarse particles. In time-series analyses, measurements of PM25 made at a central site are found
to be better correlated with indoor measurements than are measurements of PM10_25 (see section
2.7.2). A number of recent studies have evaluated the effect of this type of exposure error on
epidemiologic study results. The results of these studies, primarily focused on fine particles,
3-40
-------
indicate that exposure error related to the use of PM data from central monitoring sites is likely
to result in underestimation of the effect of PM exposure on health (CD, p. 8-288).
Analyses of site-to-site variability for PM2 5 measurements, including time-series
correlations of measurements across monitors and differences in mean concentrations between
monitors, are presented in Table 2-3. The temporal correlation coefficients between monitors
are high, often exceeding 0.80, indicating good correlation between time-series PM2 5
measurements. However, a few areas, such as Los Angeles and Seattle, had lower temporal
correlation coefficients, in the range of 0.60. As observed in the CD, western areas are less
influenced by regional sources of fine particles (CD, p. 8-293), and geographic or topographic
features may make PM2 5 levels less homogeneous. Even where there is good temporal
correlation between monitors, there may be a spatial gradient in PM2 5 across the area. As
discussed in the CD (Table 8-40), some areas had strong correlation coefficients (on the order of
0.90) but substantial differences in annual means were found between some monitor pairs. For
example, correlation coefficients averaged about 0.90 between PM2 5 monitor pairs in Detroit,
but annual mean differences of up to 6 |ig/m3 were found between monitor pairs.
This same type of analysis was done using available data for PM10_25, as discussed in
section 2.4.3. Table 2-4 shows that there are greater differences in concentrations between
paired PM10_2 5 monitors than were seen in data from paired PM2 5 monitors. Differences in
annual mean values of over 20 |ig/m3 are shown between some paired PM10_2 5 monitors,
representing differences of 60-70% in some cases. Correlations between the monitoring sites
were also somewhat lower than those for PM25, ranging from about 0.3 to 0.8. In some cities,
for example Cleveland, OH and Detroit, MI, the PM10_2 5 measurements at paired monitors show
both a large difference in magnitude as well as poor correlation in day-to-day changes; for both
cities, the values are 60-70% different between the monitor pairs, and the correlation coefficient
is about 0.4. However, for a number of the cities shown in Table 2-4, the correlation coefficients
between data from paired monitors are in the range of 0.7 to 0.8, indicating that the data are
fairly well correlated temporally, but there remain substantial differences in annual mean
concentrations between the monitors. In interpreting the results of epidemiologic associations
with PM10_2 5, the data from the central monitoring sites may be characterizing day-to-day
changes in PM10_25 concentrations adequately, but staff observes that it is difficult to determine
how well such concentrations characterize the magnitude of population exposures to PM10_25.
In summary, there are some key exposure-related distinctions between PM2 5 and PM10_2 5.
In section 9.2.1, the CD concludes that PM25 concentrations are frequently evenly distributed
across cities, and frequently have high site-to-site correlations; as summarized above, there can
be differences in some locations. In contrast, the CD concludes that PM10_2 5 is "seldom" evenly
distributed across cities and that there are "frequently low" site-to-site correlations. In such
situations, while the epidemiologic associations may be illustrating true time-series relationships
between PM and a health outcome, it is more difficult to draw inferences about the population
3-41
-------
exposure levels at which those effects are seen. From studies in which significant associations
are reported with PM10_25, the distribution of ambient monitoring data available for the study may
reflect levels that are higher or lower than those experienced by neighborhoods in other parts of
the community.
3.6.3 Alternative Model Specifications
As observed earlier, statistical modeling issues for epidemiologic studies were discussed
in great detail in the 1996 PM CD (EPA, 1996a, sections 12.6.2 and 12.6.3). This evaluation led
to the conclusion that PM-related effects observed in epidemiologic studies were unlikely to be
seriously biased by inadequate statistical modeling or confounded by weather (CD, p. 8-22).
Statistical modeling issues have re-emerged in this review, however, and much attention has
been given to further investigations of approaches to model specification for epidemiologic
analyses. The following discussions draw from the CD's evaluation of model specification
issues for both short-term and long-term exposure studies.
3.6.3.1 Time-series epidemiologic studies
In 2002, questions were raised about the default convergence criteria and standard error
calculations made using GAM, which have been commonly used in recent time-series
epidemiologic studies. As discussed more completely in the CD (section 8.4.2), a number of
time-series studies were reanalyzed using alternative methods, typically GAM with more
stringent convergence criteria and alternative models such as GLM with natural smoothing
splines. The results of the reanalyses have been compiled and reviewed in an HEI publication
(HEI, 2003a). Reanalyzed PM10 mortality study results are presented in Figure 8-15 in the CD,
where it can be seen that the reanalyses generally did not substantially change the findings of the
original analyses, and the changes in effect estimates with alternative analysis methods were
much smaller than the variation in effects across studies. In the HEI reanalyses, the CD finds
that mortality effect estimates were often, but not always, reduced with the use of GAM with
more stringent convergence criteria; however, the extent of these changes was not substantial in
most cases (CD, p. 8-232). Further, for morbidity studies, the CD finds that the impact of the
reanalyses was relatively small and the basic conclusions regarding the significance of PM-
related hospital admissions remained unchanged when more stringent GAM criteria were used
(CD, p. 8-235).
These reanalyses also investigated alternative model specifications to control for
potential weather effects and temporal trends. As shown in Figures 8-20 and 8-21 in the CD, the
magnitude of the effect estimate for PM can decrease with increasing control for weather and
temporal trend, though it generally stabilizes at some point. The CD observes that there is no
clear consensus at this time as to what constitutes appropriate control for such variables, while
recognizing that no single approach is likely to be most appropriate in all cases (CD, p. 8-340).
If the model does not adequately address daily changes in weather variables, then some effects of
3-42
-------
temperature on health would be falsely ascribed to the pollution variable. Conversely, if the
model overcontrols for weather, such that the temperature-health relationship is more "wiggly"
than the true dose-response function, then the result will be a much less efficient estimate of the
pollutant effect (CD, p. 8-236). This would result in incorrectly ascribing some of the true
pollution effect to the temperature variable, which would make it difficult to detect a real but
small pollution effect. The CD concludes that the available studies appear to demonstrate that
there are PM-related effects independent of weather influences, but that further evaluation is
needed on how to best characterize possible combined effects of air pollution and weather (CD,
p. 8-340).
In summary, the reanalyses generally support the findings of the original studies, while
raising questions for further research. For quantitative assessment, staff considers it appropriate
to use results from short-term exposure studies that did not use GAM initially, or that used either
more stringent GAM or GLM analyses. As recognized in the CD, there is no one correct
approach for model specification or covariate adjustment (CD, p. 9-35). An advantage to the use
of GAM is that the model is "data-driven" and selects the degree of smoothing or adjustment for
covariates that best fits the data. The GLM approach is advantageous in allowing more accurate
calculation of standard errors.
3.6.3.2 Prospective cohort epidemiologic studies
Data from the ACS and Six Cities prospective cohort studies were used in a major
reanalysis study that evaluated a number of issues that had been raised for the long-term
exposure studies. These issues included whether the results were sensitive to alternative
modeling strategies. The reanalysis included two major components, a replication and validation
study, and a sensitivity analysis, where alternative risk models and analytic approaches were
used to test the robustness of the original analyses. In the first phase, the data from the two
studies were found to be of generally high quality, and the original results were replicated,
confirming the original investigators' findings of associations with both total and
cardiorespiratory mortality (Krewski et al., 2000; CD, p. 8-91). In the second phase, the
sensitivity analyses generally reported that the use of alternative models, including variables that
had not been used in the original analyses (e.g., physical activity, lung function, marital status),
did not alter the original findings. Data were also obtained for additional city-level variables that
were not available in the original data sets (e.g., population change, measures of income,
maximum temperature, number of hospital beds, water hardness) and reanalysis investigators
included these data in the models. The associations between fine particles and mortality were
generally unchanged in these new analyses, with the exception of population change, which did
somewhat reduce the size of the associations with fine particles or sulfates (CD, p. 8-92).
3-43
-------
3.6.4 Co-pollutant Confounding and Effect Modification
Confounding occurs when a health effect that is caused by one risk factor is attributed to
another variable that is correlated with the causal risk factor; epidemiologic analyses attempt to
adjust or control for potential confounders. A gaseous copollutant (e.g., O3, CO, SO2 and NO2)
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 (CD, p.
8-10). Effect modifiers include variables that may influence the health response to the pollutant
exposure (e.g., co-pollutants, individual susceptibility, smoking or age). Both are important
considerations for evaluating effects in a mixture of pollutants, but for confounding, the
emphasis is on controlling or adjusting for potential confounders in estimating the effects of one
pollutant, while the emphasis for effect modification is on identifying and assessing the level of
effect modification (CD, p. 8-12).
In addition to acting as confounders or effect modifiers, the CD recognizes that pollutants
may act together in an ambient pollution mixture, potentially having additive or synergistic
effects. For example, recent animal toxicologic studies have tested effects of exposure to PM
(e.g., urban PM, carbon particles, acid aerosols) in combination with O3 and suggested that co-
exposure to O3 and urban particles resulted in greater effects than those reported with exposure
to O3 alone, while mixed results were reported from studies using combinations of acid aerosols
and O3 (CD, Table 7-13).
3.6.4.1 Co-pollutant Confounding
Potential confounding by gaseous copollutants has been most commonly assessed by
using multi-pollutant models. As discussed in the CD (section 8.4.3.2), there are statistical
issues to be considered with multi-pollutant models, such as possibly creating mis-fitting models
by forcing all pollutants to fit the same lag structure, by adding correlated but non-causal
variables, or by omitting important variables. There are issues relating to potential copollutant
confounding that multi-pollutant models may not be able to address. Inclusion of pollutants in a
multi-pollutant model that are highly correlated with one another can lead to misleading
conclusions in identifying a specific causal pollutant. Collinearity between pollutants may occur
if the gaseous pollutants and PM come from the same sources, if PM constituents are derived
from gaseous pollutants (e.g., sulfates from SO2), or if meteorological conditions contribute to
the formation of both PM and gaseous pollutants (CD, p. 8-12). These situations certainly occur.
For example, sources of fine particle constituents include combustion of various fuels, gasoline
or diesel engine exhaust, and some industrial processes (CD, Table 9-1); these sources also emit
gaseous pollutants. In addition, SO2 and PM2 5 are often derived from the same sources in an
area (e.g., coal-fired power plants) and thus simultaneous inclusion in models may result in
diminished effects for one or both pollutants, which can be misleading (CD, p. 8-14).
3-44
-------
In the NMMAPS multi-city analyses, one key objective was to characterize the effects of
PM10 and the gaseous co-pollutants, alone and in combination. Multi-pollutant modeling was
used in the NMMAPS mortality analyses for 20 and 90 U.S. cities, in which the authors added
first O3, then O3 and another co-pollutant (e.g., CO, NO2 or SO2) to the models (CD, p. 8-35).
The relationship between PM10 and mortality was little changed in models including control for
O3 and other gaseous pollutants (CD, Figure 8-4, p. 8-35). The authors concluded that the
PM10-mortality relationship was not confounded by co-pollutant concentrations across 90 U.S.
cities (Samet et al., 2000a,b; Dominici, 2003). Single- and multi-pollutant model results for a
range of health outcomes with PM10, PM25 and PM10_25 from multi- and single-city studies are
presented in Figures 8-16 through 8-19 of the CD. For the most part, the addition of gaseous co-
pollutants had little influence on PM associations, although substantial reduction in associations
with PM could be seen in some cases when gaseous pollutants are added to the model.
Using an alternative approach, investigators in the NMMAPS morbidity analyses for 14
U.S. cities tested for relationships between the coefficients for the PM10-admissions with PM10-
co-pollutant correlations for each city. No such relationships were found between the PM10
effect estimates for cardiovascular or respiratory hospitalization and PM10-co-pollutant
correlations (CD, pp. 8-146, 8-175). The authors concluded that associations with PM10 were not
dependent on the correlation between PM10 and the gaseous copollutants, though as mentioned
previously, the CD highlights the need for additional evaluation of this type of analysis (CD, pp.
8-146).
In the long-term exposure studies, multi-pollutant models have been tested in some
analyses. The reanalysis of data from the ACS cohort indicated that associations between
mortality and PM2 5 or sulfates were reduced in size in co-pollutant models including SO2 but not
with the other gaseous pollutants. Since SO2 is a precursor for fine particle sulfates, it is
inherently difficult to distinguish effects from the precursor SO2 and fine particles (CD, p. 9-37).
Some recent exposure studies have collected personal and ambient monitoring data,
collected at a single central site, for PM2 5 and gaseous pollutants (e.g., O3, SO2 and NO2), and
assessed the degree of day-to-day correlation between the different measures of personal and
ambient concentrations. The investigators reported that the personal and ambient PM2 5
measurements were correlated, as were personal exposure to PM2 5 and ambient concentrations
of the gaseous pollutants. However, the personal and ambient concentrations of each of the
gaseous pollutants were not well correlated. These findings suggest that associations reported
with ambient PM2 5 are truly reflecting associations with fine particles and that fine particles are
unlikely to be simply acting as surrogates for other gaseous pollutants (Sarnat et al., 2000, 2001;
CD, p. 5-90).
In summary, where various indicators of PM and the other pollutants are correlated, it
can be difficult to distinguish effects of the various pollutants in multi-pollutant models.
However, a number of research groups have found the effects of various indicators of PM and
3-45
-------
gases to be independent of one another, as illustrated in Figures 8-16 through 8-19 of the CD. In
addition, new evidence on exposure considerations suggests that it is unlikely that a relationship
found between a health endpoint and ambient PM2 5 concentrations is actually representing
relationships with other pollutants.
Taking into consideration the findings of single- and multi-city studies and other
evaluations of potential confounding by gaseous co-pollutants described in preceding sections,
the CD concludes that while research questions remain, in general, "associations for various PM
indices with mortality or morbidity are robust to confounding by co-pollutants." (CD, p. 9-37).
As shown in figures 8-16 through 8-19 of the CD, effect estimates for PM10, PM2 5 and PM10_2 5
were little changed in multi-pollutant models, as compared with single-pollutant models. This
indicates that effect estimates from single-pollutant models can be used to represent the
magnitude of a concentration-response relationship, though there will remain uncertainty with
regard to potential contributions from other pollutants. For quantitative assessment, staff
concludes that single-pollutant model results provide reasonable indicators of the magnitude of
PM-related effects, supported by analyses including multi-pollutant model results as available.
3.6.4.2 Effect Modification
One approach to evaluate the effect of co-pollutants on associations reported with PM2 5
is illustrated in Figure 3-3. As discussed in the 1996 Staff Paper, if PM is acting independently,
then a consistent association should be observed in a variety of locations of differing levels of
co-pollutants. Effect estimates for PM10-mortality associations were plotted against
concentrations of gaseous pollutants in the study area, and there was no evidence that
associations reported between PM10 and mortality were correlated with copollutant
concentrations. (EPA, 1996b, Figure V-3a,b). Similarly, Figure 3-3 shows the reported effect
estimates for PM25 and mortality (from single-pollutant models) from U.S. and Canadian studies
relative to the levels of O3, NO2, SO2, and CO present in the study locations. As was seen in the
last review for PM10, the magnitude and statistical significance of the associations reported
between PM25 and mortality in these studies show no trends with the levels of any of the four
gaseous co-pollutants. While not definitive, these consistent patterns indicate that it is more
likely that the effect of PM25 is not appreciably modified by differing levels of the gaseous
pollutants.
An alternative approach would be assessment within a single study of whether the effect
of time-varying PM is modified by time-varying concentrations of the gases in the time-series
models, through addition of interaction terms, for example. However, such studies have not
been conducted. While potential effect modification between various indicators of PM and the
gaseous pollutants has been little studied, the limited available evidence indicates that the gases
do not have a major role as effect modifiers for PM-related health outcomes.
3-46
-------
3.6.5 Issues Related to Alternative Exposure Periods in Epidemiologic Studies
3.6.5.1 Lag Structure in Short-term Exposure Studies
In the short-term exposure epidemiologic studies, many investigators have tested
associations for a range of lag periods between the health outcome and PM concentration (see
CD, sections 8.4.4 and 9.2.2.4). As discussed in the CD, it is important to consider the pattern of
results that is seen across the series of lag periods. If there is an apparent pattern of results
across the different lags, such as that shown in Figure 8-22 of the CD, then selecting the single-
day lag with the largest effect from a series of positive associations is likely to underestimate the
overall effect size, since single-day lag effect estimates do not fully capture the risk that may be
distributed over adjacent or other days (CD, p. 8-270). Where effects are found for a series of
lag periods, a distributed lag model will more accurately characterize the effect estimate size.
However, if there is no apparent pattern or reported effects vary across lag days, any result for a
single day may well be biased (CD, p. 9-42). Staff also observes that the high degree of
autocorrelation in PM measurements complicates the assessment of various lag periods.
For selecting effect estimates from studies for use in quantitative risk assessment, or for
evaluation of potential revisions to the standards, staff considered patterns of results for PM25 or
PM10_2 5 across lag periods from U.S. and Canadian studies. As discussed below, most of the
studies included in Appendix 3 A evaluated results for a range of lag periods, with many authors
reporting effect estimates for one lag period based on this evaluation. However, a few
researchers selected lag periods a priori. Examples of studies that used a priori selection of lag
periods includes Liao et al. (1999), in which the 24-hour PM25 average preceding measurement
of cardiac function was used, and Schwartz et al. (1996), in which an average of 0-day and 1-day
lagged PM10, PM2 5 and PM10_2 5 measurements was used in analyses of associations with
mortality.
Most authors report testing associations across a range of lag periods, and in many cases
the authors reported a pattern of positive associations across several lag periods. Figure 8-22 in
the CD presents associations for PM25 levels over both a series of days and a series of hours
preceding myocardial infarction incidence, and positive associations can be seen over several
adjacent lag periods (CD, p. 8-270; Peters et al., 2001). In an analysis using hospitalization for
asthma, researchers report testing associations for lags to 3-days and beyond, and reported
consistent patterns across lags for associations between asthma hospitalization and PM10, PM2 5
or PM10_2 5 (CD, p. 8-270; Sheppard et al.,1999; 2003). Results for the strongest associations are
presented in this study, with the authors observing:
When considering single (vs. distributed) lag estimates, it is important to put the estimate
in the context of the pattern of lags nearby and to recognize that effect estimates contain
information from adjacent days owing to serial correlation of the pollutant series. The
pollutant effects given for asthma are larger than and consistent with estimates obtained
for adjacent lags. In contrast, adjacent lags to the same-day PM and SO2 effects on
3-47
-------
0
CM '
1 i , 5
o
CO
o .
_f
.
I" 1
-L 1
1
r B
a
•
-L i
1
_
16
It '
17
1
i
,1 ^
0.08 0.10 0.12
Ozone (ppm), 4th highest 8-hour daily maximum
0.14
0.16
0.05
0.10 0.15
NO2 (ppm), 2nd highest 1 -hour daily maximum
0.20
CO
O
CD CM
CD
O
CD
| 0_|
i
16
17
13
o
co •
0
CM
0 .
j
_
4
i
_l_ t
<
1
}; •
' t
«
D
2
«
r
<
JJ-i
[*J ! t
0.02 0.04 0.06
SO2 (ppm), 2nd highest 24-hour maximum
0.08
0.10
CO (ppm), 2nd highest 8-hour daily maximum
10
Figure 3-3. Associations between PM2 5 and total mortality from U.S. studies, plotted against gaseous pollutant
concentrations from the same locations. Air quality data obtained from the Air Quality System (AQS) for each
study time period: (A) mean of 4th highest 8-hour ozone concentration; (B) mean of 2nd highest 1-hour NO2
concentration; (C) mean of 2nd highest 24-hour SO2 concentration; (D) mean of 2nd highest 8-hour CO
concentration. Study locations are identified below:
1. Chock et al., 2000, Pittsburgh, PA
2. Fairley, 2003, Santa Clara County, CA
3. Ito, 2003, Detroit, MI
4. Klemm and Mason, 2000, Atlanta, GA
5. Lipfert et al., 2000a, Philadelphia, PA
6. Mar et al., 2003, Phoenix, AZ
7. Moolgavkar, 2003, Los Angeles, CA
8. Ostro et al., 2003, Coachella Valley, CA
9. Ostro et al., 1995, Southern California
10. Schwartz, 2003a, Boston, MA
11. Schwartz, 2003a, Knoxvffle, TN
12. Schwartz, 2003a, Portage, WI
13. Schwartz, 2003a, St. Louis, MO
14. Schwartz, 2003a, Steubenville, OH
15. Schwartz, 2003a, Topeka, KS
16. Tsai et al., 2000, Camden NJ
17. Tsai et al., 2000, Elizabeth NJ
18. Tsai et al., 2000, Newark N
3-48
-------
o
CO"
0
in o_
T3
c
ro
CD
"ro
•^ °-
"55 *~
LJJ
t>
LJJ
0-
B
6
,
J
4
i
f '
I J
A-4
i'i
E
16
10
i T
17
1 "
|
i
18 13
•l I
T 12
5 1
[13J3
I
17
10
18
il
0.005 0.010 0.015 0.020
SO2 (ppm), annual mean
0.025
0.01 0.02 0.03
NO2 (ppm), annual mean
0.04
Figure 3-3 (continued). Associations between PM2 5 and total mortality from U.S. studies, plotted against gaseous pollutant
concentrations from the same locations. Air quality data obtained from the Air Quality System
(AQS) for each study time period: (E) annual mean SO2 concentration; (F) annual mean NO2
concentration. Study locations are identified below:
1. Chock et al., 2000, Pittsburgh, PA
2. Fairley, 2003, Santa Clara County, CA
3. Ito, 2003, Detroit, MI
4. Klemm and Mason, 2000, Atlanta, GA
5. Lipfert et al., 2000a, Philadelphia, PA
6. Mar et al., 2003, Phoenix, AZ
7. Moolgavkar, 2003, Los Angeles, CA
8. Ostro et al., 2003, Coachella Valley, CA
9. Ostro et al., 1995, Southern California
10. Schwartz, 2003a, Boston, MA
11. Schwartz, 2003a, Knoxvffle, TN
12. Schwartz, 2003a, Portage, WI
13. Schwartz, 2003a, St. Louis, MO
14. Schwartz, 2003a, Steubenville, OH
15. Schwartz, 2003a, Topeka, KS
16. Tsai et al., 2000, Camden NJ
17. Tsai et al., 2000, Elizabeth NJ
18. Tsai et al., 2000, Newark NJ
3-49
-------
appendicitis change much more abruptly, and the overall pattern is unstable. (Sheppard
etal., 1999, p. 27)
In a study of mortality in Phoenix, positive associations are reported with PM10, PM25 and
PM10_25 out to a lag period of 4 days, but effect estimates were larger and more often statistically
significant for the 0- or 1-day lag periods (Mar et al., 1999; 2003). In an analysis of mortality
and hospitalization in Detroit, results for models including individual lag days and moving
average lags periods for PM10, PM2 5 and PM10_2 5 are presented in appendices and the results of
the most significant lag results are presented in the body of the report, with the observation that
significant associations often occurred at multiple lags (Lippmann et al., 2000, p. 24). Among
U.S. and Canadian studies, the CD observes that many authors report finding a pattern of PM-
related effects across adjacent lags (CD, p. 8-279).
An example of results that do not follow a consistent pattern across lags can be found in
results for the association between PM2 5 and mortality in Coachella Valley (Ostro et al., 2000;
2003). In this study, the pattern of results across a series of lag periods was not consistent in
associations between PM2 5 and total or respiratory mortality8. Based on the greater uncertainty
on the effect estimate size for the PM2 5-mortality association from this study, staff concludes
that it would not be appropriate to use these PM2 5 results for quantitative assessments.9 In
addition, a series of studies in Cook County, IL and Los Angeles County, CA, include effect
estimates for 0- to 5-day lag periods and, for most health endpoints, the results follow a pattern.
However, the pattern of results specifically for COPD mortality with PM25 was quite
inconsistent (Moolgavkar, 2000a,b,c; Moolgavkar, 2003, p. 191).10 Based on the considerations
described above, staff concludes that it would not be appropriate to use the results for COPD
mortality from this study for quantitative assessment.
The CD concludes that it is likely that the most appropriate lag period for a study will
vary, depending on the health outcome and the specific pollutant under study. Some general
observations can be made about lag periods for different health outcomes. For total and
cardiovascular mortality, it appears that the greatest effect size is generally reported for the 0-day
lag and 1-day lag, generally tapering off for longer lag periods (CD, p. 8-279). This is true also
for hospitalization for cardiovascular diseases. For cardiovascular effects such as myocardial
8 Staff observes that the results for a series of lags show fairly consistent patterns for associations between
PM10 and PM10_2 5 and cardiovascular mortality in this analysis.
9The air quality measurements available for PM2 5 and PM10_2 5 may also contribute to the more uncertain
findings for PM2 5 in this study. For PM10_2 5, a 10-year series of concentrations was modeled from a 2 1A year series
of ambient measurements at co-located beta attenuation monitors, while predictive models for PM2 5 concentrations
were not reported to be adequate, so only the 2 !/2 year series of measurements were used in PM2 5 analyses.
10 That only l-in-6 day PM measurements were available in Los Angeles County is likely to be an important
factor contributing to less consistent findings there.
3-50
-------
infarction or HRV change, there appears to be a pattern of larger effects with shorter lag periods,
such as 1- to 4-hours. For respiratory symptoms, many studies report effects over a series of
lags, with larger effect estimates for moving average or distributed lag models. Similarly, for
asthma hospitalization, there appear to be larger effects over longer average time periods, out to
5- to 7-day average lags (CD, p. 8-279).
A number of recent studies that have investigated associations with distributed lags
provide effect estimates for health responses that persist over a period of time (days to weeks)
after the exposure period. The available studies have generally used PM10 or other PM
indicators, but not PM2 5 or PM10_2 5. Effect estimates from distributed lag models are often, but
not always, larger in size that those for single-day lag periods (CD, p. 8-281). For example, in
multi-city analyses of data from 10 U.S. cities, the effect estimates for total mortality from
distributed lag models are about twice those from 0-1 day average lag models (Schwartz, 2003b).
In the 14-city NMMAPS analysis of hospitalization in the elderly, the combined city effect
estimate for COPD hospitalization is larger (about doubled) in results of distributed lag models
than in 0-1 day average lag models, while the CVD hospitalization effect estimate is only
increased by a small amount, and the effect estimate for pneumonia hospitalization is somewhat
smaller in distributed lag models, compared with a 0-1 day average lag (Schwartz, et al., 2003).
In summary, the CD concludes that distributed lag results would likely provide more
accurate effect estimates for quantitative assessment than an effect estimate for a single lag
period (CD, p. 9-42). However, at this time, studies using PM2 5 and PM10_2 5 have not included
distributed lag models Most U.S. and Canadian studies have reported consistent patterns in
results for different lags; for these studies, an effect estimate for a single-day lag period is likely
to underestimate the effect. In quantitative assessments for PM2 5 and PM10_2 5, since results are
generally not available for distributed lag models, staff concludes that it is appropriate to use
single-day lag period results, recognizing that this is likely to underestimate the effect. For
quantitative assessment, staff concludes that it is appropriate to use results from lag period
analyses consistent with those reported in the CD, focusing on shorter lag periods for
cardiovascular effects and lag periods of several days for respiratory effects, depending on
availability of results. For the few studies that show inconsistent patterns, the use of single-day
lag results are not appropriate for quantitative assessment.
3.6.5.2 Seasonal Differences in Time-Series Epidemiologic Results
As discussed in section 3.5.3, time-series epidemiologic studies generally use some
temporal or seasonal terms in the models to control for seasonal changes in health outcomes. In
addition, a few epidemiologic studies have also evaluated PM-health associations across seasons
by doing analyses on data subdivided into different seasons, thus evaluating differences in
effects across the season rather than trying to control for seasonal influences. The CD observes
that there can be seasonal differences in correlations between PM and other pollutants, or in PM
levels across seasons (CD, p. 8-57).
3-51
-------
The CD presents results for seasonal analyses for individual studies in Chapter 8 and the
Appendices to Chapter 8. In 10 U.S. cities, the relationship between PM10 and mortality was the
same in analyses for data divided into summer and winter seasons (Schwartz, et al., 2000). In
Pittsburgh, relationships between PM10_2 5 and PM2 5 and mortality were "unstable" when
stratified by season, and there was evidence of differing multi-collinearity between seasons
(Chock et al., 2000). In Coachella Valley, associations between mortality and several PM
indicators were stronger in the winter season (October-May) than in the summer season (Ostro et
al., 2000). However, an earlier analysis in two Southern California counties reported significant
associations between estimated PM2 5 and mortality in the summer (April-September) quarter
only (Ostro et al., 1995). Seasonal analyses were done for the mortality-PM25 relationship in
San Jose, and there were no significant differences between the four seasons (Fairley, 2003). In
Phoenix, the association between PM10_2 5 and mortality was reported to be highest in spring and
summer, when PM10_2 5 concentrations were lowest (Mar et al., 2003). Associations between
PM10 and hospitalization for cardiovascular diseases in Los Angeles yielded larger effect
estimates in the winter and fall seasons than in spring or summer (Linn et al., 2000). Asthma
hospitalization was significantly associated with PM10 for both "wet" and "dry" seasons in Los
Angeles, but the effect estimates were larger during the wet season (January-March) (Nauenberg
and Basu, 1999). In Seattle, associations between PM10, PM2 5 and PM10_2 5 and asthma
hospitalization were positive in all seasons, but effect estimates were larger in spring and fall
(Sheppard et al., 2003).
Staff observes that these few studies show no general pattern in results across seasons.
The largest of these studies showed no seasonal differences in the results combining data from
10 U.S. cities (Schwartz et al., 2000). Most of the studies listed above show generally positive
results across all seasons tested, with some reporting larger effect estimates in one or more
season(s), but the differences were not statistically significant. Staff concludes that the available
evidence does not support quantitative assessment of seasonal differences in relationships
between PM and health outcomes at this time.
3.6.5.3 Health effects related to different short-term exposure time periods
While most time-series epidemiologic studies use 24-hour average PM measurements,
several new studies have used ambient PM concentrations averaged over shorter time intervals,
such as 1- or 4-hour averages. Many such studies have evaluated associations with
cardiovascular health biomarkers or physiological changes. Section 8.3.1.3.4 of the CD
describes several epidemiologic studies that report statistically significant associations between
2- to 4-hour PM10 or PM2 5 concentrations and cardiovascular health endpoints, including
myocardial infarction incidence and heart rate variability (CD, pp. 8-162 to 8-165). One study
reported effect estimates for myocardial infarction incidence with PM2 5 averaged over 2- and 24
hours that are quite similar in magnitude, and both are statistically significant (Peters et al.,
2001; CD, p. 8-165).
3-52
-------
For respiratory health outcomes, two panel studies of symptoms in asthmatic subjects are
summarized in the CD (section 8.3.3.1.1). One study in a small Southern California community,
reported larger effect estimates for 1- or 8-hour concentrations than for 24-hour PM10
concentrations (Delfmo et al., 1998), while the other, in Los Angeles, reported larger effect
estimates for 24-hour PM10 concentrations (Ostro et al., 2001; CD, p. 8-206). However, several
studies of hospital admissions or medical visits for respiratory diseases reported the strongest
associations with several-day average PM concentrations (CD, p. 8-279).
Evidence of health effects associations with different exposure time periods can inform
staff conclusions and recommendations regarding potential NAAQS averaging times. Staff
observes that the very limited information available in the CD suggests that cardiovascular
effects may be associated with acute exposure time periods on the order of an hour or so.
3.6.5.4 Exposure periods used in prospective cohort studies
The prospective cohort studies have used air quality measurements averaged over long
periods of time, such as several years, to characterize the long-term ambient levels in the
community. The exposure comparisons are basically cross-sectional in nature, and do not
provide evidence concerning any temporal relationship between exposure and effect (CD, p.
9-42). As discussed in the CD, it is not easy to differentiate the role of historic exposures from
more recent exposures, leading to potential exposure measurement error (CD, p. 5-118). This
potential misclassification of exposure is increased if average PM concentrations change over
time differentially between areas.
Several new studies have used different air quality periods for estimating long-term
exposure and tested associations with mortality for the different exposure periods. In the
extended analysis of the ACS study, Pope et al. (2002) reported associations between mortality
and PM2 5 using the original air quality data (1979-1983), data from the new fine particle
monitoring network (1999-2000), and the average PM25 concentrations from both time periods.
The authors reported that the PM2 5 concentrations for the different time periods were well
correlated, indicating that the ordering of the cities from low to high pollution levels had
changed little. When using average PM2 5 levels from all years, the associations for total,
cardiopulmonary and lung cancer were slightly larger in size, though not significantly so, than
for either individual air quality data set.
A new analysis of the Six Cities data has evaluated mortality risk with different estimates
of long-term PM25 exposure. The original study (Dockery et al., 1993) averaged PM25 and PM10
concentrations over a period of years (1979 to 1986) to represent long-term PM exposure
estimates, while the new analysis includes PM2 5 data from more recent years and evaluates
associations with PM25 averaged over a range of time periods, such as 2 or 3-5 years preceding
the individual's death (Villeneuve et al., 2002). The authors reported that effect estimates for
mortality were lower with time-dependent PM2 5 exposure indicators (e.g., 2 years before
individual's death), than with the longer-term average concentrations. They postulate that this is
3-53
-------
likely due to the "influence of city-specific variations in mortality rates and decreasing levels of
air pollution that occurred during follow-up" (CD, p. 8-97). This might be expected, if the most
polluted cities had the greatest decline in pollutant levels as controls were applied (CD, p. 8-93).
The authors observe that the fixed average concentration window may be more representative of
cumulative exposures, and thus a more important predictor of mortality, than a shorter time
period just preceding death (Villeneuve et al., 2002, p. 574).
Using essentially the same air quality data set as that used in the original ACS analyses,
Lipfert et al. (2000b) investigated associations between mortality and PM (using several PM
indicators) over numerous averaging periods. When using methods similar to those of the other
prospective cohort studies, the authors report finding similar associations between fine particles
and mortality (CD, p. 8-115). However, in analyses using mortality and PM data in different
time segments, the results were varied, with some statistically significant negative associations
reported. The authors report that the strongest positive associations were found with air quality
data from the earliest time periods, as well as the average across all data.
All three analyses indicate that averaging PM concentrations over a longer time period
results in stronger associations; as the Six Cities study authors observe, the longer series of data
is likely a better indicator of cumulative exposure. In these studies, spatial variation in the PM
concentrations is the key exposure indicator, and one key question is the extent to which
concentrations change over time, particularly whether there are differential changes across cities.
As observed above, the order of cities from high to low pollution levels changed little across
time periods in the cities used in the ACS analyses. Where lower effect estimates are reported
with data collected in more recent years, the CD observes: "This is likely indicative of the
effectiveness of control measures in reducing source emissions importantly contributing to the
toxicity of ambient particles in cities where PM levels were substantially decreased over time"
(CD, p. 9-43). The CD concludes that further study is warranted on the importance of different
time windows for exposure indicators in studies of effects of chronic PM exposure.
For use in quantitative assessments, staff concludes that it appropriate to use results from
analyses that are based on averaging PM levels over longer time periods, since the recent studies
indicate that this provides a better indicator of long-term PM exposure. Thus, as described in
Chapter 4, the results from the extended ACS analyses using average PM25 concentrations from
both the original and more recent time periods are used in the PM risk assessment. Staff notes
that this is consistent with the advice to EPA from the Health Effects Subcommittee (HES) of the
SAB's Clean Air Act Compliance Council (SAB, 2004), in their review of methods used for
EPA's health benefits assessments. The HES recommended using the results of ACS cohort
analyses that used air quality data averaged over the full study time period, indicating that this
represented the best period to use in order to reduce measurement error.
3-54
-------
3.6.6 Concentration-Response Relationships and Potential Thresholds
In assessing or interpreting public health risk associated with exposure to PM, the form
of the concentration-response function is a critical component. The CD recognized that it is
reasonable to expect that there likely are biologic thresholds for different health effects in
individuals or groups of individuals with similar innate characteristics and health status.
Individual thresholds would presumably vary substantially from person to person due to
individual differences in genetic-level susceptibility and pre-existing disease conditions (and
could even vary from one time to another for a given person). Thus, it would be difficult to
detect a distinct threshold at the population level, below which no individual would experience a
given effect, especially if some members of a population are unusually sensitive even down to
very low concentrations. The person-to-person difference in the relationship between personal
exposure to PM of ambient origin and the concentration observed at a monitor may also add to
the variability in observed concentration-response relationships, further obscuring potential
population thresholds within the range of observed concentrations (CD, p. 9-43, 9-44).
The 1996 CD evaluated evidence from epidemiologic studies regarding both linear and
nonlinear forms of concentration-response relationships and whether any effect thresholds could
be identified. Based on the few available studies, the 1996 CD concluded that linear model
results "appear adequate for assessments of PM10 and PM25 effects" (EPA, 1996a, p. 13-91).
Among the new epidemiologic studies of short-term PM exposure are several that use different
modeling methods to investigate alternative forms of concentration-response functions and
potential threshold levels.
Several time-series studies have evaluated potential threshold levels for associations
between mortality and short-term PM exposures. In plots of concentration-response curves from
multi-city analyses, using the NMMAPS data, it is difficult to discern any evident threshold for
relationships between PM10 and total or cardiorespiratory mortality. The authors also present
posterior probabilities for the existence of thresholds at different levels of PM10 showing that if
there is a threshold in the relationships between PM10 and total or cardiorespiratory mortality, the
likelihood of the threshold being above about 25 |ig/m3 is essentially zero (Dominici et al.,
2003b; CD, pp. 8-320, 8-321). In one single-city analysis, various statistical methods were used
to test for thresholds in simulated data sets that were created with assumed threshold levels
ranging from 12.8 to 34.4 |ig/m3 for the relationship between PM10 and mortality. The authors of
this analysis concluded that, in the data for this city, it was highly likely that standard statistical
methods could detect a threshold level, if one existed (Cakmak et al., 1999; CD, p. 8-319). Thus,
a number of studies have thus been unable to detect threshold levels in the PM-mortality
relationship, and in fact one single-city analysis suggests that statistical methods would allow
detection of a threshold in the epidemiologic data if a clear threshold existed.
However, a few analyses in individual cities have provided suggestions of some potential
threshold levels. One single-city study used PM2 5 and PM10_2 5 measurements in Phoenix and
3-55
-------
reported that there was no indication of a threshold in the association between PM10_2 5 and
mortality, but that there was suggestive evidence of a threshold for the mortality association with
short-term exposure to PM25 up to levels of about 20-25 |ig/m3 (Smith et al., 2000; CD, 8-322).
In addition, single-city analyses in Birmingham and Chicago suggested that the concentration-
response functions for PM10 and mortality changed to show increasing effects at levels of 80 to
100 |ig/m3 PM10, but "not to an extent that statistically significant distinctions were
demonstrated" (CD, p. 8-322).
For long-term exposure to PM and mortality, the shape of the concentration-response
function was evaluated using data from the ACS cohort. The concentration-response
relationships for associations between PM2 5 and all-cause, cardiopulmonary and lung cancer
mortality are shown in Figure 3-4. The authors reported that the associations for all-cause,
cardiovascular and lung cancer mortality "were not significantly different from linear
associations" (Pope, et al., 2002). It is apparent in this figure that the confidence intervals
around each of the estimated concentration-response functions expand significantly as one looks
below around 12-13 i-ig/m3, indicating greater uncertainty in the shape of the concentration-
response relationship at concentration ranges below this level. In addition, for lung cancer, the
relationship appears to have a steeper linear slope at lower concentrations, with a flatter linear
slope at PM2 5 concentrations that exceed about 13 |ig/m3 (CD, p.8-98).
In summary, while staff recognizes that there likely are individual biologic thresholds for
specific health responses, existing studies do not support or refute the existence of thresholds in
PM-mortality relationships at the population level, for either long-term or short-term PM
exposures within the range of air quality observed in the studies (CD, p. 9-44). While
epidemiologic analyses have not identified thresholds in observed associations in the range of air
quality concentrations in the studies, it is possible that such thresholds exist toward the lower
end of these ranges (or below these ranges) but cannot be detected due to variability in
susceptibility across a population. Even in those few studies with suggestive evidence of such
thresholds, the potential thresholds are at fairly low concentrations (CD, p. 9-45).
Based on the above considerations, staff concludes in part that it is appropriate to use the
linear or log-linear concentration-response models reported in epidemiologic studies in the
quantitative risk assessment. Staff also recognizes, however, the possibility that thresholds may
exist in reported associations at fairly low levels within the range of air quality observed in the
studies, though no specific threshold levels have been clearly identified. While the biologic
plausibility of the existence of individual thresholds supports the potential that concentration-
response relationships may be non-linear at the lower end of the range of observed
concentrations, statistical evaluations comparing linear and non-linear concentration-models
have been unable to resolve this question. Therefore, the staff also concludes that the
implications of assuming a non-linear concentration-response relationship also should be
3-56
-------
JC
ir
S
0.2
0.1
0.0
en
-0.2-
-0.3-
-0.4-
0.2
0.1
0.0
s -°'1
a- -0.2
§> -03
-0.4-
All Cause
10 15 20
Concentration (|jg/m3)
Lung Cancer
10 15 20
Concentration (jjg/m3)
in
I
0)
en
o
en
01
03
0)
0.2-
0.1 -
-0.1 -
-0.2-
-0.3-
-0.4-
^-"
^^_T.
X
X
Cardiopulmonary
10 15 20
Concentration (M9/ni3)
0.2-
0.1 -
0.0 -
-0.1 -
-0.2-
-0.3-
-0.4-
~ -"'--- ~~ — ~~
All Other
10 15 20
Concentration ((jg/rn3)
Figure 3-4. Natural logarithm of relative risk for total and cause-specific mortality per
10 ug/m3 PM2 5 (approximately the excess relative risk as a fraction), with
smoothed concentration-response functions. Based on Pope et al. (2002) mean
curve (solid line) with pointwise 95% confidence intervals (dashed lines).
(Source: CD, Figure 8-7).
included in the quantitative risk assessment. In the absence of published concentration-response
models reflecting typical sigmoidal or "hockey-stick" shaped relationships, staff has included in
the quantitative risk assessment (described in greater detail in Chapter 4) analyses incorporating
a modified linear slope with an imposed cut point. This approach is used as a surrogate for a
non-linear, sigmoidal-shaped function, in which the cut point is intended to reflect an inflection
point at the lower end of the relationship, below which it is assumed that there is little or no
population response.
3.7 SUMMARY AND CONCLUSIONS
Based on the available evidence and the evaluation of that evidence in the CD,
summarized briefly above, staff concludes that the body of evidence supports an inference of
causality for associations between PM25 and a broad range of health effects. Short-term
3-57
-------
exposure to PM2 5 is likely causally associated with mortality from cardiopulmonary diseases,
hospitalization and emergency department visits for cardiopulmonary diseases, increased
respiratory symptoms, decreased lung function, and physiological changes or biomarkers for
cardiac changes. Long-term exposure to PM2 5 is likely causally associated with mortality from
cardiopulmonary diseases and lung cancer, and effects on the respiratory system such as
decreased lung function or the development of chronic respiratory disease. Staff concludes that
there is less strength, but suggestive evidence of causality for short-term exposure to PM10_25 and
indicators of morbidity, including hospitalization for cardiopulmonary diseases, increased
respiratory symptoms and decreased lung function. Staff concludes that it is appropriate to
consider including the health outcomes listed above in quantitative assessments for PM2 5 and
PM10_25. Further, staff notes that more equivocal evidence is available for other PM-health
responses, such as associations between short-term exposure to PM10_2 5 and mortality, and
between PM and effects on infants. Staff believes that less certain evidence, while not
appropriate for quantitative assessment, can inform more general assessments of the evidence.
Several issues that are relevant to the interpretation of health evidence for quantitative
assessment of PM-related effects are discussed above. Measurement error and exposure error
are issues that are distinctly more important for interpretation of results for PM10_25 than PM25.
For PM10_2 5, there is greater uncertainty in the relationship between ambient PM measured at
central monitors and individuals' exposure to ambient PM, based on both variability in PM10_2 5
concentrations across an area and decreased ability for coarse particles to penetrate into
buildings. This uncertainty is likely to broaden the confidence intervals around effect estimates.
In interpreting results of associations with PM10_25, staff places greater emphasis on evaluating
results from the pattern of findings in multiple studies than on statistical significance of any
individual result.
In the evaluation of different epidemiologic model specifications, as described above,
some effect estimates differ upon reanalysis to address issues associated with the use of the
default GAM procedures, but many are little affected. Recognizing that there is no single
"correct" analytical approach, staff concludes that it is appropriate for quantitative assessment to
use results from short-term exposure studies that were reanalyzed with more stringent GAM
criteria or with other approaches such as GLM, or that did not use GAM in the original analysis.
Regarding potential confounding by co-pollutants, the CD concludes that the evidence
supports the existence of independent effects of PM, while recognizing the difficulties in
distinguishing effects from mixtures of correlated pollutants. For quantitative assessment, staff
concludes that single-pollutant model effect estimates can be used as reasonable indicators of the
magnitudes of effect sizes, especially for comparing results across studies. Additional analyses
using multi-pollutant model results, where available, can allow assessment of risks related to PM
exposure with adjustment for co-pollutants.
3-58
-------
The CD concludes that distributed lag periods may provide the most representative
quantitative estimates of effect for some health outcomes, such as mortality. Recognizing that
distributed lags have not been used in the available studies of PM25 and PM10_25, staff concludes
that a reasonable approach to selection of effect estimates for use in quantitative assessment is to
evaluate the pattern of lag results available from studies. If the data show a reasonable pattern of
results, then selecting a single lag period is appropriate, recognizing that this result is likely to
underestimate effects. Conversely, if the pattern of results across lag periods is unstable, staff
concludes that it is inappropriate to use such results for quantitative assessment since the "best"
lag day result may be biased upward.
For the long-term exposure studies, recent studies indicate that long-term PM exposure is
likely to be better estimated from air quality data averaged over longer time periods (e.g.,
multiple years of data). Staff concludes that effect estimates based on PM data averaged over
longer times periods are more representative of population health responses for use in risk
assessment. Specifically, for the results from the extended analysis of the ACS study, staff
concludes that it is most appropriate to use the concentration-response functions from the models
using averaged air quality data over the full study time period for quantitative assessment.
Finally, evaluation of the health effects data summarized in the CD provides no evidence
to support selecting any particular population threshold for PM2 5 or PM10_2 5, recognizing that it
is reasonable to expect that, for individuals, there may be thresholds for specific health
responses. Based on the above considerations, staff concludes in part that it is appropriate to use
the linear or log-linear concentration-response models reported in epidemiologic studies in the
quantitative risk assessment. Staff also recognizes, however, the possibility that thresholds may
exist in reported associations at fairly low levels within the range of air quality observed in the
studies, though no specific threshold levels have been clearly identified. Therefore, the staff also
concludes that the implications of assuming a non-linear concentration-response relationship also
should be included in the quantitative risk assessment. To do so, alternative cutpoints can be
used as surrogate for a non-linear, sigmoidal-shaped function, to reflect an inflection point at the
lower end of the relationship, below which it is assumed that there is little or no population
response.
3-59
-------
REFERENCES
Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
ambient concentrations of particulates and oxidants and development of chronic disease in a cohort of
nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
paniculate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
7: 19-34.
Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that
2.5 microns in aerodynamic diameter (PM2 5) from airport visibility data in California. J. Exposure Anal.
Environ. Epidemiol. 5: 161-180.
Abbey, D. E.; Burchette, R. J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med. 158:
289-298.
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, L.; Yang, J. X. (1999) Long-
term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
Care Med. 159:373-382.
Avol, E. L.; Gauderman, W. J.; Tan, S. M.; London, S. J.; Peters, J. M. (2001) Respiratory effects of relocating to
areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
Becker, S.; Soukup, J. M.; Sioutas, C.; Cassee, F. R. (2003) Response of human alveolar macrophages to ultrafme,
fine and coarse urban air pollution particles. Exp. Lung Res. 29: 29-44.
Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105:614-620.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between paniculate- and gas-phase components of urban air pollution and daily
mortality in eight Canadian cities. Inhalation Toxicol. 12(suppl. 4): 15-39.
Burnett, R. T.; Goldberg, M. S. (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].
Cakmak, S.; Burnett, R. T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
concentrations for air pollution-related mortality with exposure measurement error. Risk Anal. 19:487-
496.
Carrothers, T. J.; Evans, J. S. (2000) Assessing the impact of differential measurement error on estimates of fine
particle mortality. J. Air Waste Manage. Assoc. 50:65-74.
Centers for Disease Control and Prevention. (2004) The health consequences of smoking: a report of the Surgeon
General. Atlanta, GA: U.S. Department of Health and Human Services, National Center for Chronic
Disease Prevention and Health Promotion, Office on Smoking and Health. Available:
http://www.cdc.gov/tobacco/sgr/sgr_2004/chapters.htm (18 August, 2004).
3-60
-------
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12:281-298
Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52:113-117.
Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
Arizona. Seattle, WA: University of Washington, National Research Center for Statistics and the
Environment; NRCSE technical report series, NRCSE-TRS no. 040. Available:
http://www.nrcse.washington.edu/pdf/trs40_pm.pdf [18 October, 2004].
Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
155: 568-576.
Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. G. (1998) Symptoms in pediatric asthmatic and air pollution:
differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging
time. Environ. Health Perspect. 106:751-761.
Diociaiuti, M.; Balduzzi, M.; De Berardis, B.; Cattani, G.; Stacchini, G.; Ziemacki, G.; Marconi, A.; Paoletti, L.
(2001) The two PM25 (fine) andPM25-10 (coarse) fractions: evidence of different biological activity.
Environ. Res. A 86: 254-262.
Dockery, D. W.; Pope, C. A., Ill; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
(1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
329: 1753-1759.
Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
symptoms. Environ. Health Perspect. 104: 500-505.
Dominici, F.; McDermott, A.; Daniels, M.; Zeger, S. L.; Samet, J. M. (2003a) Mortality among residents of 90 cities.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 9-24. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Dominici, F.; Daniels, M.; McDermott, A.; Zeger, S. L.; Samet, J. M. (2003b) Shape of the exposure-response
relation and mortality displacement in the NMMAPS database. In: Revised analyses of time-series studies
of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 91-96. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
EPA. (1996a) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC: National Center for
Environmental Assessment-RTF Office; report no. EPA/600/P-95/00laF-cF. 3v
EPA. (1996b) 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-452YR-96-013.
3-61
-------
EPA. (2002) Health assessment document for diesel engine exhaust. Washington, DC: Office of Research and
Development, National Center for Environmental Assessment; report no. EPA/600/8-90/057F. Available:
http://cfpub.epa.gov/ncea/ [22 May, 2003].
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. Boston, MA: Health Effects Institute; pp.
97-106. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdftl8 October, 2004].
Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98
Gauderman, W. J.; McConnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; 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, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
Margolis, H. G.; 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.
Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, L.; Allen, G.; Verrier, M.; Cherry, R.;
Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101:1267-1273.
Goldberg, M. S.; Bailar, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible
to short-term increases in paniculate air pollution: a time-series studi in Montreal, Quebec. Cambridge,
MA: Health Effects Institute; research report 97.
Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
HEI. (1997) Paniculate air pollution and daily mortality: analyses of the effects of weather and multiple air
pollutants (The phase I.E. report of the Particle Epidemiology Evaluation Project). Health Effects Institute.
Cambridge, MA.
HEI. (2003c) Commentary on revised analyses of selected studies. In: Revised analyses of time-series studies of air
pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 255-290. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Hornberg, C.; Maciuleviciute, L.; Seemayer, N. H.; Kainka, E. (1998) Induction of sister chromatid exchanges
(SCE) in human trachea! epithelial cells by the fractions PM-10 and PM-2.5 of airborne particulates.
Toxicol. Lett. 96/97: 215-220.
Ito, K.; Kinney, P.; Thurston, G. D. (1995) Variations in PM-10 concentrations within two metropolitan areas and
their implications for health effects analyses. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the
colloquium on paniculate air pollution and human mortality and morbidity, part II; January 1994; Irvine,
CA. Inhalation Toxicol. 7: 735-745.
Ito, K.; Thurston, G. D. (1996) Daily PM10/mortality associations: an investigation of at-risk subpopulations. J.
Exposure Anal. Environ. Epidemiol.6:79-95.
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. Boston,
3-62
-------
MA: Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[12 May, 2004].
Jacobs, I; Kreutzer, R.; Smith, D. (1997) Rice burning and asthma hospitalizations, Butte County, California,
1983-1992. Environ. HealthPerspect. 105: 980-985.
Katsouyanni, K.; Touloumi, G.; Samoli, E.; Petasakis, Y.; Analitis, A.; Le Tertre, A.; Rossi, G.; Zmirou, D.;
Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojtyniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
Schindler, C.; Schwartz, J. (2003) Sensitivity analysis of various models of short-term effects of ambient
particles on total mortality in 29 cities in APHEA2. In: Revised analyses of time-series studies of air
pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 157-164. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM-10 associations in Los Angeles.
Inhalation Toxicol. 7:59-69.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 165-172. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J,; Jerrett, M.; Abrahamowicz, M.; White,
W. H. (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.
Liao, D.; Creason, J.; Shy, C.; Williams, R.; Watts, R.; Zweidinger, R. (1999) Daily variation of paniculate air
pollution and poor cardiac autonomic control in the elderly. Environ. Health Perspect. 107: 521-525.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and size-
classified paniculate matter. J. Air Waste Manage. Assoc. 50:1501-1513.
Lipfert, J. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) the Washington
University-EPRI veteran's cohort mortality study: preliminary results. Inhalation Toxicol. 12(Suppl. 4):41-
73.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report 95.
Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
California. Environ. Health Perspect. 105: 216-222.
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108:347-353.
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 [18 October, 2004].
3-63
-------
Mauderly, I; Neas, L.; Schlesinger, R. (1998) PM monitoring needs related to health effects. In: Atmospheric
observations: helping build the scientific basis for decisions related to airborne paniculate matter; Report of
the PM measurements research workshop, July 22-23, 1998. Available from "PM Measurements Report",
Health Effects Institute, 955 Massachusetts Ave., Cambridge MA 02139.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. I; Vora, H.; Avol, E.; Gauderman, W. I; Margolis, H. G.;
Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
California children with asthma. Environ. Health Perspect. 107: 757-760.
McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of
mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers. J.
Exposure Anal. Environ. Epidemiol. 10:427-436.
Monn, C.; Becker, S. (1999) Cytotoxicity and induction of proinflammatory cytokines from human monocytes
exposed to fine (PM2 5) and coarse particles (PM10_2 5) in outdoor and indoor air. Toxicol. Appl. Pharmacol.
155: 245-252.
Moolgavkar, S. H. (2000a) Air pollution and mortality in three U.S. counties. Environ. Health Perspect. 108:777-
784.
Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J. Air Waste Manage. Assoc. 50:271-280.
Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
metropolitan areas of the United States. Inhalation Toxicol. 12(Suppl. 4):75-90.
Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
heart failure among elderly people in seven large US cities. Am. J. Public Health 85: 1361-1365.
Naeher, L. P.; Holford, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B.; Leaderer, B. P. (1999)
Health women's PEF variations with ambient summer concentrations of PM10, PM25, SO42", H+, and O3.
Am. J. Respir. Crit. Care Med. 160: 117-125.
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
and exposure to air pollution. Public Health Rep. 114: 135-148.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
141: 111-122.
Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Speizer, F. E. (1999) Fine particles and peak flow in children: acidity
versus mass. Epidemiology 10:550-553.
3-64
-------
Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
70: 98-104.
Ostro, B. D.; Lipsett, M. I; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols.
Am. J. Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
exacerbations among African-American children in Los Angeles. Inhalation Toxicol. 7:711-722.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10:412-419.
Ostro, B.; Lipsett, M.; Mann, J.; Braxtron-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
African-American children in Los Angeles. Epidemiology 12: 200-208.
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
[18 October, 2004].
Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities
with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit.
Care Med. 159: 760-767.
Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
Monahan, K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology
11: 11-17.
Peters, A.; Dockery, D. W.; Muller, J. E.; Mittleman, M. A. (2001) Increased paniculate air pollution and the
triggering of myocardial infarction. Circulation 103:2810-2815.
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah Valley. Arch.
Environ. Health 47:211-217.
Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Paniculate 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, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999a) Paniculate air pollution and daily mortality on Utah's Wasatch
Front. Environ. Health Perspect. 107: 567-573.
Pope, C. A., Ill; Dockery, D. W.; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999b) Oxygen saturation, pulse rate
and paniculate pollution: a daily time-series panel study. Am. J. Respir. Crit. Care Med. 159: 365-372.
Pope, C. A., Ill; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
G. M.; Gold, D. R.; Dockery, D. W. (1999c) Heart rate variability associated with paniculate air pollution.
Am. Heart J. 138:890-899.
3-65
-------
Pope, C. A., Ill; Thun, M. I; Namboodiri, M. M; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
Respir. Crit. CareMed. 151:669-674.
Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. J. Am. Med. Assoc.
287:1132-1141.
Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer, F.
E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
Perspect. 104: 506-514.
SAB. (2004) Advisory for plans on health effects analysis in the analytical plan for EPA's second prospective
analysis - benefits and costs of the clean air act, 1990-2000. Advisory by the Health Effects Subcommittee
of the Advisory Council for Clean Air Compliance Analysis. EPA SAB Council - ADV-04-002. March,
2004. Available: http://www.epa.gov/science 1/pdf/council adv 04002.pdf.
Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1996) Weather, air pollution and mortality in
Philadelphia, 1973-1980, report to the Health Effects Institute on phase IB, Particle Epidemiology
Evaluation Project. Cambridge, MA: Health Effects Institute; review draft.
Samet, J. M.; Zeger, S. L.; Domenici, F.; Curriero, F.; Coursac, L; Dockery, D.W.; Schwartz, J.; Zanobetti, A.
(2000a) The national morbidity, mortality, and air pollution study. Part I: methods and methodological
issues. Cambridge, MA: Health Effects Institute: research report no. 94.
Samet, J. M.; Zeger, S. L.; Domenici, F.; Curriero, F.; Coursac, L; Dockery, D.W.; Schwartz, J.; Zanobetti, A.
(2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute: research report no. 94.
Samet, J. M.; Domenici, F.; Curriero, F.; Coursac, L; Zeger, S. L. (2000c) Fine paniculate air pollution and
mortality in20 U.S. cities, 1987-1994. N. Engl. J. Med. 343:1742-9.
Sarnat, J. A.; Koutrakis, P.; Suh, H. H. (2000) Assessing the relationship between personal paniculate and gaseous
exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50: 1184-1198.
Sarnat, J. A.; Schwartz, J.; Catalano, P. J.; Suh. H. H. (2001) Gaseous pollutants in paniculate matter epidemiology:
confounders or surrogates? Environ. Health Perspect. 109:1053-1061.
Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137:1136-1147.
Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
8: 371-377.
Schwartz, J. (2000) Assessing confounding, effect modification, and thresholds in the association between ambient
particles and daily deaths. Environ. Health Perspect. 108: 563-568.
Schwartz, J. (2003a) Daily deaths associated with air pollution in six US cities and short-term mortality displacement
in Boston. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 219-226. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Schwartz, J. (2003b) Airborne particles and daily deaths in 10 US cities. In: Revised analyses of time-series studies
of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 211-218. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
3-66
-------
Schwartz, I; Dockery, D. W.; Neas, L. M; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis, P.; Speizer, F. E.;
Ferris, E.G., Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
Am. J. Respir. Crit. Care Med. 150:1234-1242.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
Michigan. Am. J. Epidemiol 142:23-35.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J.
Air Waste Manage. Assoc. 46:927-939.
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.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology 11:6-10.
Schwartz, J. (2001) Air pollution and blood markers of cardiovascular risk. Environ. Health Perspect.
Suppl. 109(3): 405-409.
Schwartz, J.; Zanobetti, A.; Bateson, T. (2003) Morbidity and mortality among elderly residents of cities with daily
PM measurements. In: Revised analyses of time-series studies of air pollution and health. Special report.
Boston, MA: Health Effects Institute; pp. 25-58. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and
coarse particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
Soukup, J. M.; Becker, S. (2001) Human alveolar macrophage responses to air pollution particulates are associated
with insoluble components of coarse material, including paniculate endotoxin. Toxicol. Appl. Pharmacol.
171: 20-26.
Steerenberg, P. A.; Withagen, C. E.; Dormans, J. A. M. A.; VanDalen, W. J.; VanLoveren, H.; Casee, F. R. (2003)
Adjuvant activity of various diesel exhaust and ambient particle in two allergic models. J. Toxicol. Environ.
Health A 66: 1421-1439.
Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in
Saint John, Canada. J. Exposure Anal. Environ. Epidemiol.: 10: 461-477.
Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne paniculate matter on daily
death counts. Environ Health Perspect. 103:490-497.
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
summertime haze air pollution in Toronto, Ontario: Consideration of the role of acid aerosols. Environ. Res.
65:271-290.
3-67
-------
Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine paniculate air pollution,
resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M;
Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000) Air quality and
pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
Villeneuve, P. J.; Goldberg, M. S.; Krewski, D.; Burnett, R. T.; Chen, Y. (2002) Fine paniculate air pollution and
all-cause mortality within the Harvard six-cities study: variations in risk by period of exposure. Ann.
Epidemiol. 12: 568-576.
Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
Inhalation Toxicol. 9: 15-29.
Zanobetti, A.; Schwartz, J. (2003a) Airborne particles and hospital admissions for heart and lung disease. In:
Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 241-248. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October,
2004].
3-68
-------
4. CHARACTERIZATION OF HEALTH RISKS
4.1 INTRODUCTION
This chapter presents information regarding the results from an updated PM health risk
assessment that builds upon the methodology used in the more limited assessment conducted as
part of the last PM NAAQS review. This updated assessment includes estimates of (1) risks of
mortality, morbidity, and symptoms associated with recent ambient PM2 5 and PM10_2 5 levels; (2)
risk reductions associated with just meeting the current suite of PM25 NAAQS; and (3) risk
reductions associated with just meeting various alternative PM25 and PM10_25 standards. This
risk assessment is more fully described and presented in a technical support document,
Paniculate Matter Health Risk Assessment for Selected Urban Areas (Abt Associates, 2005b;
henceforth referred to as the Technical Support Document and cited as TSD).
The goals of this PM risk assessment are: (1) to provide estimates of the potential
magnitude of mortality and morbidity effects associated with current PM25 and PM10_25 levels,
and with meeting the current suite of PM25 NAAQS and alternative PM25and PM10_25 standards,
in specific urban areas;1 (2) to develop a better understanding of the influence of various inputs
and assumptions on the risk estimates; and (3) to gain insights into the distribution of risks and
patterns of risk reductions associated with meeting alternative suites of PM standards. Staff
recognizes that there are many sources of uncertainty and variability inherent in the inputs to this
assessment and that there is a high degree of uncertainty in the resulting PM risk estimates.
While some of these uncertainties have been addressed quantitatively in the form of estimated
confidence ranges around central risk estimates, other uncertainties and the variability in key
inputs are not reflected in these confidence ranges, but rather are addressed through separate
sensitivity analyses or characterized qualitatively.
Following this introductory section, this chapter discusses the scope of the risk
assessment, including selection of urban areas and health endpoints; components of the risk
model; characterization of uncertainty and variability associated with the risk estimates; and key
results from the assessment for both PM2 5 and PM10_2 5. The TSD provides a more detailed
discussion of the risk assessment methodology and includes additional risk estimates beyond
those summarized herein.
4.1.1 Overview of Risk Assessment From Last Review
In the last review, PM-associated risks were estimated for two urban areas: Philadelphia
and Los Angeles counties (Abt Associates, 1996). The PM health risk model used in the last
lrTo provide a broader perspective on health risks associated with ambient PM, risk estimates associated
with current PM10 levels also have been included in an appendix to the TSD for those urban areas where PM2 5 risks
have been estimated.
4-1
-------
assessment combined information about daily PM air quality for these two areas with estimated
concentration-response functions derived from epidemiologic studies and baseline health
incidence data for specific health endpoints to derive estimates of the annual incidence of
specific health effects associated with recent air quality levels (termed "as is" air quality in both
the previous and current TSD). Since site-specific relative risks were not available for all
endpoints in both locations (and in the absence of more information concerning which individual
studies might best characterize the health risk in a given location), a type of meta-analysis
(referred to as a "pooled analysis") was conducted that combined the results of those studies that
met specified criteria. The assessment also examined the reduction in estimated incidence that
would result from just meeting the existing PM10 standards and various alternative PM25
standards. In addition, the assessment included sensitivity analyses and integrated uncertainty
analyses to better understand the influence of various inputs and assumptions on the risk
estimates. The methodological approach followed in conducting the last risk assessment and risk
estimates are described in Chapter 6 of the 1996 Staff Paper (EPA, 1996b) and in several
technical reports (Abt Associates, 1996; Abt Associates, 1997a,b) and publications (Post et al.,
2000; Deck etal., 2001).
In the 1997 review of the PM NAAQS, EPA placed greater weight on the overall
qualitative conclusions derived from the health effect studies - that ambient PM is likely causing
or contributing to significant adverse effects at levels below those permitted by the existing PM10
standards - than on the specific concentration-response functions and quantitative risk estimates
derived from them. Nevertheless, EPA judged that the assessment provided reasonable
estimates as to the possible extent of risk for those effects given the available information (62 FR
at 38656).
4.1.2 Development of Approach for Current Risk Assessment
The scope and methodology for this updated PM risk assessment have been developed
over the last few years. In June 2001, OAQPS released a draft document, PM NAAQS Risk
Analysis Scoping Plan (EPA, 200 Ic), for CAS AC consultation and public comment, which
described staffs general plan for this assessment. In January 2002, OAQPS released a more
detailed draft document, ProposedMethodology for P articulate Matter Risk Analyses for
Selected Urban Areas (Abt Associates, 2002), for CAS AC review and public comment, which
described staffs plans to assess (a) PM25-related risks for several health endpoints, including
mortality, hospital admissions, and respiratory symptoms and (b) PM10_2 5-related risks for
hospital admissions and respiratory symptoms. During a February 2002 teleconference, CASAC
discussed this draft document and public comments were made; CASAC sent an advisory letter
to the Administrator documenting its advice in May 2002 (Hopke, 2002). In its advisory letter,
CASAC "concluded that the general methodology as described in the report is appropriate . . .
4-2
-------
Thus, the general framework of the approach is the sensible approach to this risk analysis"
(Hopke, 2002).
In response to a request from CASAC to provide additional details about the planned
scope of the PM10_25 and PM10 components of the assessment, in April 2003 OAQPS released a
draft memorandum (Abt, 2003a) for CASAC consultation and public comment, addressing these
topics. In August 2003, OAQPS released a draft PM risk assessment report (Abt Associates,
2003b) in conjunction with the first draft PM Staff Paper. The CASAC provided its comments
on the draft PM risk assessment in a letter to the Administrator (Hopke, 2004). A second draft
PM risk assessment report (Abt Associates, 2005a) was released in January 2005 in conjunction
with the second draft PM Staff Paper, and was reviewed by CASAC and the public at a meeting
held in April 2005. The final PM risk assessment report (Abt Associates, 2005b) takes into
account the advice from CASAC and public comments received on the earlier drafts of that
document.
4.2 SCOPE OF PM RISK ASSESSMENT
This risk assessment estimates risks of various health effects associated with exposure to
ambient PM25 and PM10_25 in a number of urban areas selected to illustrate the public health
impacts of these pollutants. The health endpoints selected for the PM2 5 assessment, discussed in
section 4.2.1, include those related to short- and long-term exposure for which the CD concludes
that the association with PM2 5 (or one or more PM2 5 components), acting alone and/or in
combination with gaseous co-pollutants, is likely causal (CD, p. 9-79). The health endpoints
selected for the PM10_2 5 assessment, also discussed in section 4.2.1, include those related to
short-term exposure for which the CD concludes that the scientific evidence is suggestive of an
association that the staff judges to be likely causal. This assessment includes risk estimates for
nine urban areas for PM25 and three urban areas for PM10_2 5 The basis for selection of these
areas is discussed below (section 4.2.2). This assessment is intended to estimate risks
attributable to anthropogenic sources and activities (i.e., risk associated with concentrations
above policy-relevant background or above various higher cutpoints that reflect possible
population thresholds).
This assessment uses concentration-response functions from epidemiologic studies of
short- and long-term exposures to ambient PM based on PM concentrations measured at fixed-
site, community-oriented, ambient monitors. As discussed in Chapter 2 (section 2.7) and
Chapter 3 (section 3.6.2), measurements of daily variations of ambient PM concentrations, as
used in the time-series studies, have a plausible linkage to the daily variations of exposure to
ambient PM2 5 and PM10_2 5 for the populations represented by ambient monitoring stations. The
CD concludes that "at this time, the use of ambient PM concentrations as a surrogate for
exposures is not expected to change the principal conclusions from PM epidemiologic studies
that use community average health and pollution data" (CD, p. 5-121). The possible impact of
4-3
-------
exposure misclassification on the estimated concentration-response relationships derived from
the community epidemiologic studies is discussed above in Chapter 3 (section 3.6.2). Since the
currently available epidemiologic health effects evidence relates ambient PM concentrations, not
exposures, to health effects, this assessment does not include a quantitative exposure analysis.
While quantitative estimates of personal or population exposure do not enter into this risk
assessment, an understanding of the nature of the relationships between ambient PM
concentrations and its various components and human exposure underlies the conceptual basis
for this assessment.2
While the NAAQS are intended to provide protection from health effects associated with
exposure to ambient PM, EPA recognizes that exposures to PM from other sources (i.e., non-
ambient PM) also have the potential to affect health. The EPA's Office of Radiation and Indoor
Air and other Federal Agencies, such as the Consumer Product Safety Commission (CPSC) and
the Occupational Safety and Health Administration (OSHA), address potential health effects
related to indoor, occupational, environmental tobacco smoke, and other non-ambient sources of
PM exposure. As with the prior PM risk assessment, contributions to health risk from non-
ambient sources are beyond the scope of the risk assessment for this NAAQS review.
4.2.1 Selection of Health Endpoint Categories
As discussed in Chapter 3, OAQPS staff carefully reviewed the health effects evidence
evaluated in the CD to identify potential health effect categories for inclusion in this assessment.
Given the large number of endpoints and studies addressing PM-related effects, staff included
only the more severe and better understood (in terms of health consequences) health endpoint
categories. In addition, the staff included only those health endpoints for which the overall
weight of the evidence from the collective body of studies supports the CD conclusion that there
is likely to be a causal relationship or that the scientific evidence is sufficiently suggestive of a
causal relationship that staff judges the effects to be likely causal between PM and the health
effects category. Finally, for the three PM indicators (PM2 5, PM10, PM10_2 5), staff considered
only those endpoint categories which provided concentration-response relationships based on
U.S. and Canadian studies that used PM concentrations obtained by one of the following
approaches: (1) directly measuring fine particles using PM2 5 or PM2 b (2) estimating the
concentration of fine particles using nepholometry data, and (3) estimating PM10_25
concentrations based on co-located PM10 and PM2 5 monitors or based on measurements using
dichotomous samplers.
2As discussed in Chapter 5 of the CD, EPA and the exposure analysis community are working to improve
exposure models designed specifically to address PM and to collect new information in PM exposure measurement
field studies that will improve the scientific bases for exposure analyses that may be considered in future reviews.
4-4
-------
Based on a review of the evidence evaluated in the CD and discussed in Chapter 3, as
well as the criteria discussed above, staff included the following broad categories of health
endpoints in the risk assessment for PM2 5 and PM10_2 5:
Related to short-term PM2 5 exposure:
total (non-accidental), cardiovascular, and respiratory mortality;
hospital admissions for cardiovascular and respiratory causes;
respiratory symptoms not requiring hospitalization
Related to long-term PM2 5 exposure:
total, cardiopulmonary, and lung cancer mortality.
Related to short-term PM10_2^5 exposure:
• hospital admissions for cardiovascular and respiratory causes;
respiratory symptoms.
As discussed in Chapter 3 (sections 3.4 and 3.7), the available evidence for other health
responses, such as associations between short-term exposure to PM10_2 5 and mortality, is more
equivocal. Staff believe that these health endpoints, which are based on less certain evidence,
are not appropriate for inclusion in the quantitative risk assessment.
4.2.2 Selection of Study Areas
A primary goal of the current PM risk assessment has been to identify and include urban
areas in the U.S. for which epidemiologic studies are available that estimate concentration-
response relationships for those locations. This goal is in large part motivated by the evaluation
contained in the CD and staff assessment in Chapter 3 that suggests there may be geographic
variability in concentration-response relationships across different urban areas in the U.S. The
selection of urban areas to include in the PM risk assessment was based on the following criteria:
• An area is the same as or close to the location where at least one concentration-
response function, for one of the selected health endpoints, has been estimated by
an epidemiologic study that satisfies the study selection criteria (see below).
An area had relatively recent area-specific baseline incidence data available for
those locations with epidemiologic studies reporting PM-related hospital
admissions.
An area is one in which epidemiologic studies exist that had relatively greater
precision, as discussed below.
4-5
-------
• An area had sufficient air quality data for a recent year (1999 or later). Sufficient
PM2 5 data are defined as having at least one PM monitor at which there are at
least 11 observations per quarter for a one year period.3 Sufficient air quality data
for PM10_2 5 are defined as a one year period with at least 11 daily values per
quarter based on data from co-located PM2 5 and PM10 monitors.4
For the PM2 5 risk assessment, staff focused on selecting urban areas in which studies
reported total and/or cardiovascular mortality associated with short-term exposure to PM2 5
concentrations, since this was the largest data base in terms of number of studies in different
locations. Staff then supplemented this by consideration of other morbidity endpoints (i.e.,
hospital admissions). Based on a review of studies listed in Tables 8A and 8B of the CD (see
also Appendices 3 A and 3B of this Staff Paper), a candidate pool of 17 urban locations was
initially suggested based on short-term exposure mortality studies (16 of the candidate
locations); Seattle was added based on a hospital admissions study.
Staff next considered an indicator of study precision for the urban areas associated with
the short-term exposure mortality studies identified in the first step. As discussed above in
Chapter 3 (section 3.3.1.1) and in Chapter 8 of the CD (pp. 8-324 to 8-325), the natural logarithm
of the mortality-days (a product of each city's daily mortality rate and the number of days for
which PM data were available) can be used as a rough indicator of the degree of precision of
effect estimates; studies with larger values for this indicator should be accorded relatively greater
study weight. While there was no bright line for selecting any particular cutoff, staff chose to
consider only those urban areas in which studies with relatively greater precision were
conducted, specifically including studies that have a natural log of mortality-days greater than or
equal to 9.0 (i.e., approximately 8,000 deaths) for total non-accidental mortality.5 As a result of
applying this criterion, six urban areas were excluded as potential study areas (Camden, NJ;
Coachella Valley, CA; Elizabeth, NJ; Newark, NJ; Steubenville, OH; and Topeka, KS).
Finally, staff considered which of the remaining potential study locations identified from
steps one and two above also had sufficient PM2 5 ambient monitoring data consistent with the
above criterion. This final criterion excluded two of the remaining potential study areas
(Knoxville, TN and Portage, WI), leaving nine urban areas (i.e., Boston, MA; Detroit, MI; Los
3For PM2 5 an additional requirement was that a city had to have at least 122 days of data (i.e., equivalent to
1 in 3 day monitoring) for a recent year of air quality to be included.
4The criterion of at least 11 observations per quarter is based on EPA guidance on measuring attainment of
the daily and annual PM standards and is contained in Appendix N of the July 18, 1997 Federal Register notice.
5Most of the epidemiologic studies reporting total non-accidental mortality also report on one or more cause
specific mortality categories. In such studies, the natural log of mortality days is often less than 9.0 because there
are fewer deaths from a specific cause. We included cause-specific mortality concentration-response functions from
such studies, as long as the natural log of total mortality-days was greater than or equal to 9.0.
4-6
-------
Angeles, CA; Philadelphia, PA; Phoenix, AZ; Pittsburgh, PA; San Jose, CA; Seattle, WA; and
St. Louis, MO) in which epidemiologic studies reported concentration-response relationships for
PM2 5 and mortality or hospital admissions and which had sufficient air quality data in a recent
year. The PM2 5 risk assessment for long-term exposure mortality also was conducted for these
same nine urban areas.
Most of the short-term morbidity and respiratory symptom studies reporting PM25-related
effects were conducted in the same set of locations as the short-term exposure mortality studies.
In considering these other health endpoints, staff applied the same criteria, focusing on locations
which had studies with relatively greater precision, had adequate PM2 5 ambient air quality data,
and, for the hospital admissions effect category, had the necessary baseline incidence data.
The selection of urban areas to include for the PM10_2 5 risk assessment was based on
examining the pool of epidemiologic studies reporting associations for PM10_25 with the
morbidity endpoints (hospital admissions and respiratory symptoms) in any of the urban areas
already selected for the PM2 5 risk assessment. As noted earlier, the PM10_2 5 risk assessment is
more limited because of the more limited air quality data as well as the smaller number of health
endpoints and studies. Based on the available data, EPA has included in the PM10_2 5 risk
assessment the following health endpoints and locations: increased hospital admissions in
Detroit and Seattle, and increased respiratory symptoms in St. Louis.
The health endpoints and urban locations selected for the PM2 5 risk assessment are
summarized in Tables 4-1 and 4-2, for mortality and morbidity endpoints, respectively;
endpoints and locations for the PM10_25 risk assessment are summarized in Table 4-3. These
tables also list the specific studies that provided the estimated concentration-response functions
used in the PM2 5 and PM10_2 5 risk assessment. More detailed information on the studies selected
can be found in Appendices 3A, 3B, and 4A of this Staff Paper and Appendix C of the TSD.
4.3 COMPONENTS OF THE RISK MODEL
In order to estimate the incidence of a particular health effect associated with recent
conditions in a specific county or set of counties attributable to ambient PM2 5 or PM10_2 5
exposures in excess of background or various cutpoints, as well as the change in incidence of the
health effect in that county or set of counties corresponding to a given change in PM2 5 or PM10_2 5
levels resulting from just meeting a specified set of PM25 or PM10_25 standards, the following
three elements are required:
air quality information including: (1) recent air quality data for PM2 5 and PM10_2 5
from ambient monitors for the selected location, (2) estimates of background
PM2 5 and PM10_2 5 concentrations appropriate for that location, and (3) a method
for adjusting the recent data to reflect patterns of air quality estimated to occur
when the area just meets a given set of PM25 (or PM10_25) standards;
4-7
-------
Table 4-1. Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the PM2 5 Risk Assessment
Urban Location
Boston, MA
Detroit, MI
Los Angeles,
CA
Philadelphia, PA
Phoenix, AZ
Pittsburgh, PA
San Jose, CA
Seattle, WA
St. Louis, MO
Mortality Associated with Short-Term Exposure
Total (non-accidental)
Schwartz et al. (1996)A*
Lippmann et al. (2000)°
Moolgavkar (2000a)D
Lipfert et al. (2000)
Chock et al. (2000)
Fairley (1999)F
Schwartz etal. (1996)A
Cardiovascular
Klemm et al. (2000)B
- ischemic heart
disease *
Moolgavkar (2000a)D
Lipfert et al. (2000) *
Maretal. (2000)E
Fairley (1999)F
Klemm et al. (2000)B
- ischemic heart
disease *
Circulatory
Lippmann et al.
(2000)°
Respiratory
Klemm et al. (2000)B -
COPD *,
pneumonia *
Lippmann et al. (2000)°
Fairley (1999)F
Klemm et al. (2000)B -
COPD *,
pneumonia *
Mortality Associated with Long-
Term Exposure0
Krewski et al. (2000)-6cities
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
Krewski et al. (2000)-6cities
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
* Includes a multi-city or multi-county concentration-response function
A Reanalyzed in Schwartz (2003a)
B Reanalyzed in Klemm and Mason (2003)
c Reanalyzed in Ito (2003)
D Reanalyzed in Moolgavkar (2003)
E Reanalyzed in Mar et al. (2003)
F Reanalyzed in Fairley (2003)
GKrewski et al. (2000)-6 cities and -ACS provide total
and cardiopulmonary mortality and Pope et al. (2002)-
ACS extended provide total, cardiopulmonary, and lung
cancer mortality coefficients
4-8
-------
Table 4-2. Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in the PM2 5 Risk
Assessment
Urban Location
Boston, MA
Detroit, MI
Los Angeles, CA
Seattle, WA
St. Louis, MO
Cardiovascular Hospital Admissions
Lippmann et al. (2000)A - ischemic heart
disease, congestive heart failure,
dysrhythmias
Moolgavkar (2000b)B
Respiratory Hospital Admissions
Lippmann et al. (2000)A - pneumonia,
COPD+
Moolgavkar (2000c)B - COPD+
Sheppard et al. (1999)° - asthma
Respiratory Symptoms
Schwartz and Neas (2000)* - cough, lower
respiratory symptoms (LRS)
Schwartz and Neas (2000)* - cough, LRS
Includes multi-city concentration-response function
A Reanalyzed in Ito (2003); COPD+ is indicated here because the authors included asthma in their definition of COPD.
B Reanalyzed in Moolgavkar (2003); COPD+ is indicated here because the authors included asthma in their definition of COPD.
c Reanalyzed in Sheppard (2003)
Table 4-3. Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in the PM10_2 5 Risk Assessment
Urban Location
Detroit, MI
Seattle, WA
St. Louis, MO
Cardiovascular Hospital Admissions
Lippmann et al. (2000)A -
Congestive heart disease,
Ischemic heart disease
Dysrhythmias
Respiratory Hospital Admissions
Lippmann et al. (2000)A - Pneumonia,
COPD+
Sheppard et al. (1999)B - asthma
Respiratory Symptoms
Schwartz and Neas (2000) - LRS, cough
*Includes multi-city concentration-response function
A Reanalyzed in Ito (2003); COPD+ is indicated here because the authors included asthma in their definition of COPD.
B Reanalyzed in Sheppard (2003)
4-9
-------
• relative risk-based concentration-response functions (preferably derived in the
assessment location) which provide an estimate of the relationship between the
health endpoints of interest and ambient PM concentrations; and
annual or seasonal baseline health effects incidence rates and population data
which are needed to provide an estimate of the annual or seasonal baseline
incidence of health effects in an area before any changes in PM air quality.
Figure 4-1 provides a broad schematic depicting the role of these components in the risk
assessment. EPA evaluated several base case scenarios, using various alternative cutpoints (see
section 4.3.4.2). Those points where EPA has conducted analyses of alternative assumptions,
procedures, or data across the various locations are indicated by a diamond with Sx in it. A
summary description of the sensitivity analyses performed is included in Table 4-4.6 Each of the
key components (i.e., air quality information, estimated concentration-response functions, and
baseline incidence and population data) is discussed below, highlighting those points at which
judgments have been made.
The concentration-response relationships used in the PM risk assessment are empirically
estimated relationships between average ambient PM concentrations and the health endpoints of
interest reported by epidemiologic studies for specific urban areas. Most epidemiologic studies
estimating relationships between PM and health effects used a method referred to as "Poisson
regression" to estimate exponential (or log-linear) concentration-response functions.7 In this
model,
y = B e^ (Equation 4-1)
where y is the incidence of the health endpoint of interest associated with ambient PM level x, p
is the coefficient of ambient PM concentration, and B is the incidence of the health endpoint
when there is no ambient PM2 5 or PM10_2 5 The difference in health effects incidence, Ay = y0 - y,
from y0 to the baseline incidence rate, y, that corresponds to a given difference in ambient PM2 5
(or PM10_2 5) levels, Sx = x0 - x, is then
6Two additional sensitivity analyses were carried out in single locations: one addressing the impact of an
exceptional event episode in Boston and one examining the effect of different model specifications on annual health
risks associated with recent air quality levels in Los Angeles.
7For some studies on respiratory hospital admissions used in the risk assessment, a linear concentration-
response function was estimated.
4-10
-------
Air Quality
Ambient Population-
oriented Monitoring
and Estimated
Background Levels for
Selected Cities
Air Quality Adjustment
Procedures
Alternative Proposed
Standards
Recent Air
Quality Analysis
Concentration-Response
Human Epidemiological
Studies (various health
endpoints)
Estimates of City-specific
Baseline Health Effects
Incidence Rates
(various health
endpoints) and
Population Data
Changes in
Distribution
of PM Air
Quality
Concentration
Response
Relationships
Health
Risk
Model
Risk Estimates:
• Recent Air
Quality
• Alternative
Scenarios
= kth Sensitivity Analysis (See Table 4-4): Analysis of effects of alternative assumptions, procedures or data
occurs at these points.
Figure 4-1. Major components of particulate matter health risk assessment.
4-11
-------
Table 4-4. Sensitivity Analyses
Analysis
Number
(Figure 4-1)
1
2
3
4
5
6
7
PM
Indicator
PM25,
PM10.2.5
PM25
PM25
PM25
PM25
PM25
PM25
Component of
the Risk
assessment
Air Quality
Air Quality
Air Quality
Air Quality
Concentration-
Response
Concentration-
Response
Concentration-
Response
Sensitivity Analyses or Comparisons
Sensitivity analyses of the effect of assuming different
(constant) background PM levels
Sensitivity analyses of the effect of assuming a
constant background PM level versus a distribution of
daily background levels
Sensitivity analyses of the effect of just meeting the
current and alternative annual PM2 5 standards using
the maximum versus the average of monitor-specific
averages
Sensitivity analyses of the effect of an alternative air
quality adjustment procedure on the estimated risk
reductions resulting from just meeting the current 24-
hr and annual PM2 5 standards
Sensitivity analyses using an approach to estimate the
possible impact of using a distributed lag
concentration-response function
Sensitivity analyses of the impact on mortality
associated with long-term exposure of different
assumptions about the role of historical air quality
concentrations in contributing to the reported effects
Sensitivity analysis of the impact on mortality
associated with short-term exposure of using a multi-
city concentration-response function compared to
location-specific concentration-response functions
from single-city studies
Source: Abt Associates (2005b)
4-12
-------
Ay = y[e^ - 1] (Equation 4-2)
or, alternatively,
Ay = y(RR^ - 1) (Equation 4-3)
where RRAx is the relative risk associated with the change in ambient PM2 5 (or PM10.2 5) levels,
Ax. Equations 4-2 and 4-3 are simply alternative ways of expressing the relationship between a
given difference in ambient PM2 5 (or PM10_2 5) levels and the corresponding difference in health
effects. These are the key equations that combine air quality information, concentration-
response information, and baseline health effects incidence and population information to
estimate ambient PM2 5 and PM10_2 5 health risks.
For the first part of the risk assessment that characterizes risks associated with recent
ambient PM concentrations, Ax is the difference between the recent ambient PM concentrations
(on each day for the short-term exposure [i.e, daily or 24-hour] endpoints or the annual average
for the long-term exposure [i.e., annual average or longer] endpoints) and either the estimated
policy-relevant background concentration or alternative cutpoints for short-term exposure
endpoints or 7.5 |ig/m3 or the alternative cutpoints for long-term exposure mortality.8 For the
second part of the risk assessment, characterizing the reduction in health effects incidence
associated with alternative PM standards, Ax is the difference between ambient PM
concentrations when the current PM standards are just met (on each day for the short-term
exposure endpoints or the annual average for the long-term exposure endpoints) and ambient PM
concentrations associated with just meeting the specified alternative standards.9
For short-term exposure health endpoints, the risk assessment first calculated the daily
changes in incidence. Since most areas had at least some days for which no ambient PM
concentration data were available, the estimated annual incidence was summed up for each
quarter of the year and adjusted by using the ratio of the total number of days in each quarter to
As indicated previously, staff judges that the most relevant risk estimates are for those PM levels in excess
of an estimated policy-relevant background and various cutpoints well above this background. As discussed more
fully in Section 4.3.2.6, risk estimates for long-term exposure mortality are calculated in excess of the minimum of
the lowest measured levels for the long-term exposure studies included in the risk assessment and in excess of two
alternative cutpoint levels.
9For those areas already meeting the current PM2 5 standards, Ax is the difference between the recent
ambient PM concentrations and ambient PM concentrations associated with just meeting the specified standards.
4-13
-------
the number of days in the quarter for which air quality data was available.10 This simple
adjustment assumes that missing air quality data occur randomly with respect to level within a
quarter and that the distribution of PM concentrations on the days with missing data is
essentially the same as the distribution on days for which there are PM data. The quarterly
incidence estimates were then summed to derive an annual estimate.
The daily time-series epidemiologic studies used models estimating concentration-
response functions in which the PM-related incidence on a given day depends only on some
specified lagged PM concentration measure (e.g., 0-day lag, 1-day lag, 2-day lag, average of 0-
and 1-day lag). As discussed in Chapter 3 (section 3.6.5.1), such models necessarily assume that
the longer pattern of PM levels preceding the PM concentration on a given day does not affect
mortality on that day. To the extent that PM-related mortality on a given day is affected by PM
concentrations over a longer period of time, then these models would be mis-specified; and this
mis-specification would affect the predictions of daily incidence based on the model. The extent
to which longer-term (i.e., weekly, monthly, seasonal, or annual) PM2 5 exposures affect the
relationship observed in the daily time-series studies is unknown. However, there is some
evidence, based on analyses of PM10 data, that mortality on a given day is influenced by prior
PM exposures up to more than a month before the date of death (Schwartz, 2000a, reanalyzed in
Schwartz, 2003b). As indicated in section 3.6.5.2, our use of single day lag models which ignore
longer-term influences may result in the risk being underestimated. Currently, there is
insufficient information to adjust for the impact of longer-term exposure (on the order of weeks
or months) on mortality associated with short-term PM2 5 exposures, and this is an important
uncertainty that should be kept in mind as one considers the results from the short-term exposure
PM2 5 risk assessment.
The estimated PM2 5-related mortality associated with long-term exposure studies is likely
to include mortality related to short-term exposures as well as mortality related to longer-term
exposures. As just discussed, estimates of daily mortality based on the time-series studies also
are likely to be affected by prior exposures. Therefore, the estimated annual incidences of
mortality calculated based on the short- and long-term exposure studies are not likely to be
completely independent and should not be added together.
The statistical uncertainty surrounding the estimated PM2 5 and PM10_2 5 coefficients in the
reported concentration-response functions is reflected in the confidence intervals provided for
the risk estimates in sections 4.4 and 4.5. As discussed in greater detail in section 4.3.2.1, due to
the significant uncertainty associated with whether or not the concentration-response
relationships are approximately linear down to policy-relevant background, additional base case
risk estimates are presented using alternative cutpoints for both short- and long-term exposure
10Adjustment was done on a quarterly basis to reduce possible bias that would be introduced where missing
data are not uniformly distributed throughout the year.
4-14
-------
mortality associated with PM2 5 concentrations. As summarized in Table 4-4, a number of
sensitivity analyses (St through S7 in Figure 4-1) were also conducted. The results of these
sensitivity analyses are discussed in sections 4.4 and 4.5.
4.3.1 Air Quality Considerations
As illustrated in Figure 4-1, and noted earlier, air quality information required to conduct
the PM risk assessment includes: (1) recent air quality data for PM25 and PM10_25 from suitable
monitors for each selected location, (2) estimates of background PM25 and PM10_25
concentrations appropriate for each location, and (3) air quality adjustment procedures to
modify the recent data to reflect changes in the distribution of PM air quality estimated to occur
when an area just meets a given set of PM25 (or PM10_25) standards. OAQPS retrieved ambient
air quality data for PM2 5 and PM10 for the potential study areas for the years 1999 through 2003
from EPA's Air Quality System (AQS). Staff calculated PM10_25 concentrations from co-located
PM2 5 and PM10 monitors that met the minimum observation cutoff criterion. Generally, the most
recent year of PM data were used for each study area and PM indicator subject to meeting this
requirement.
A composite monitor data set was created for each assessment location based on
averaging the 24-hour values from all monitors eligible for comparison with the standards for
each day. The resulting composite monitor data set provides a single series of daily
concentrations for the urban area which serves as the surrogate index of exposure for the urban
area. The use of a composite monitor value to represent ambient PM air quality most closely
matches the approach taken in the epidemiology studies that serve as the source of the
concentration-response relationships used in the risk assessment. Table 4-5 provides a summary
of the PM2 5 and PM10_2 5 ambient air quality data for the urban study areas, including the range of
annual and 24-hr average statistics across monitors in each study area and the composite monitor
values used in the risk assessment. Additional tables providing more detailed information on PM
ambient concentrations for these locations, including the number of observations available on a
quarterly and annual basis for each monitor, can be found in Appendix A of the TSD.
4.3.1.1 Estimating PM Background Levels
Background PM concentrations used in the PM risk assessment are defined above in
Chapter 2 as the PM concentrations that would be observed in the U.S. in the absence of
anthropogenic emissions of PM and its precursors in the U.S., Canada, and Mexico. For the
initial base case risk estimates, the midpoint of the appropriate ranges of annual average
estimates for PM25 background presented in section 2.6 were used (i.e., eastern values were used
for eastern study locations and western values were used for western study locations). For
PM10_2 5 analyses, the approximate mid-points of the annual average estimates for PM10_25
background presented in section 2.6 were used for the initial base risk estimates.
4-15
-------
Table 4-5. Summary of PM Ambient Air Quality Data for Risk Assessment Study Areas*
Area
Boston, MAa
Detroit, MIb
Los Angeles County, CA°
Philadelphia County, PAd
Phoenix, AZe
Pittsburgh, PAf
San Jose, CAB
Seattle, WAh
St. Louis1
Population
(millions)
2.8
2.1
9.5
1.5
3.1
1.3
1.7
1.7
2.5
PM,;*(ug/m3)
Annual Average
Range
Across
Monitors
11.4-13.6
14.1-19.1
9.4-22.1
13.2-16.1
9.2-10.9
12.0-20.2
10.1-11.7
7.8-10.8
13.0-17.5
Composite
Monitor
12.1
15.7
19.1
14.3
10.4
16.9
11.1
8.3
14.0
24-hr Average, 98ft%ile
Range
Across
Monitors
30.6-41.3
33.6-46.2
17.0-61.3
35.6-42.3
22.7-35.3
30.7-66.6
36.9-40.1
10.9-28.4
30.5-40.8
Composite
Monitor
34.1
41.5
55.0
38.4
28.9
43.9
37.6
21.7
30.6
PM10.7;*(ug/m3)
Annual Average
Range
Across
Monitors
10.9-25.0
10.0-12.6
10.1-14.9
Composite
21.7
11.4
12.0
24-hr , gsToile
Range
Across
Monitors
40.2-105.9
25.4-30.3
24.2-33.3
Composite
Monitor
105.9
26.2
24.1
'Based on air quality data for the year 2003, unless otherwise noted in footnotes below.
"Summary statistics for a "composite monitor" based on average of 24-hour values at the different monitors in urban area that reported on each day.
"Includes Middlesex, Norfolk, and Suffolk Counties.
blncludes Wayne County.
Includes Los Angeles County.
Includes Hennepin and Ramsey Counties.
Includes Philadelphia County.
Includes Maricopa County; PM2 5 air quality data are for 2001.
Includes Allegheny County
includes Santa Clara County
"includes King County
'Includes St. Louis, Franklin, Jefferson, St. Charles Counties in MO, Clinton, Madison, Monroe, and St. Clair Counties in IL and St. Louis City.
June 2005
4-16
Draft - Do Not Quote or Cite
-------
In sensitivity analyses, we examine the impact of assuming 1) a constant background set
at the lower and upper ends of the range of estimated background levels for the eastern and
western United States, depending on the assessment location (St in Figure 4-1), and 2) a variable
daily PM2 5 background, using distributions whose means are equal to the values used in the base
case analysis and whose distributions are based on an analysis of PM25 data from relatively
remote sites with the sulfate component removed (S2 in Figure 4-1) (see Langstaff, 2005).
4.3.1.2 Simulating PM Levels That Just Meet Specified Standards
To estimate the health risks associated with just meeting the current PM25 standards and
alternative PM25 and PM10_2 5 standards, it is necessary to estimate the distribution(s) of PM
concentrations that would occur under each specified standard (or sets of standards). Since
compliance with the standards is based on a 3-year average, air quality data from 2001 to 2003
have been used to determine the amount of reduction in PM25 concentrations required to meet
the current or alternative suites of standards. Estimated design values11 (see Table 4-13 later in
this Chapter), based on the highest community-oriented monitor within each study area, are used
to determine the percent adjustment necessary to just meet annual, 98th percentile daily, and 99th
percentile daily standards. The amount of control has then been applied to a single year of data
(2003, unless otherwise specified) to estimate risks for a single year.
Under the current annual PM2 5 standard, urban areas may (under certain circumstances)
use the average of the annual averages of several monitors within an urban area to determine
compliance, commonly referred to as the "spatial averaging approach." Therefore, a sensitivity
analysis (S3 in Figure 4-1) has been conducted for three urban areas which satisfy the criteria for
use of spatial averaging to allow comparison of the estimated incidence and percent reduction in
incidence associated with using either the highest monitor or the spatial average for determining
the percent adjustment necessary to just meet the current and alternative annual standards.
The percent adjustment to simulate just meeting alternative standards is applied to the
composite monitor for the urban area. The composite monitor is used because it is the best
surrogate indicator of exposure that matches the type of exposure measure used in the original
epidemiologic studies. When assessing the risks associated with long-term exposures, which use
concentration-response functions from epidemiologic studies that are specified in terms of long-
term average concentrations, the annual mean is simply set equal to the standard level. In
contrast, when assessing the risks associated with short-term exposures, which use
concentration-response functions from epidemiologic studies that consider the sequence of daily
average concentrations, the distribution of 24-hour values that would occur upon just meeting a
given 24-hour and/or annual PM standard has to be simulated.
11A design value is a statistic that describes the air quality status of a given area relative to the level of the
NAAQS. Design values are often based on multiple years of data, consistent with the specification of the NAAQS in
Part 50 of the CFR. For example, for the base case analyses for the current PM2 5 NAAQS, the 3-year averages (of
annual means or 98th percentiles) based on the maximum monitor within an urban area are the design values.
4-17
-------
There are many possible ways to create an alternative distribution of daily concentrations
that just meets a specified set of PM standards. Both the 1996 assessment (see Abt Associates,
1996, section 8.2) and a more recent analysis of historical air quality data (see Appendix B in the
TSD) have found that PM25 levels in excess of estimated background concentrations in general
have historically decreased in a roughly proportional manner (i.e., concentrations at different
points in the distribution of 24-hour PM2 5 values in excess of an estimated background
concentration have decreased by approximately the same percentage). This suggests that, in the
absence of detailed air quality modeling, a reasonable method to simulate PM25 reductions that
would result from just meeting a set of standards is to use a proportional adjustment (i.e., to
decrease non-background PM levels on all days by the same percentage) for all concentrations
exceeding the background level.12 We are using that approach in the base case here. The
assessment also includes a sensitivity analysis (S4 in Figure 4-1) to examine the impact on the
PM25 risk estimates of an alternative air quality adjustment procedure (e.g., a method that
reduces the top 10% of daily PM25 concentrations more than the lower 90%).
Because the PM10_25 historical air quality data are substantially more sparse, there were
insufficient data to carry out the type of evaluation of historical data that was done for PM2 5 to
see whether the shape of the distribution of daily values has changed over time. In the absence
of a clearly preferable alternative, the same proportional rollback approach used for PM2 5 has
been used for the PM10_2 5 assessment. This increases the uncertainty about the PM10_2 5 risk
estimates associated with meeting alternative PM10_2 5 standards.
Where sets of standards are considered, as is the case for PM2 5 where both an annual and
a daily standard are specified, the percent reduction is determined by the "controlling standard."
The "controlling standard" is defined as the standard which would require the greatest reduction
in PM levels to just meet the standard. For example, for the current suite of PM2 5 standards, the
existing annual standard of 15 |ig/m3 is the controlling standard for the five urban study areas
(i.e., Detroit, Los Angeles, Philadelphia, Pittsburgh, and St. Louis) that do not meet the current
standards based on design values.13 In four of these five urban areas, suites of annual standards
within the range of 12 to 15 |ig/m3 combined with the current daily standard of 65 |ig/m3, using a
98th percentile form, requires the same reduction as when these annual standards are combined
with a daily standard of 40 |ig/m3, using the same daily form. Therefore, the risk assessment
only includes the 14 |ig/m3 annual standard combined with the current daily standard for one
12 The portion of the distribution below the estimated background concentration is not rolled back, since air
quality strategies adopted to meet the standards will not reduce the background contribution to PM concentrations.
13See www.epa.gov/airtrends/pdfs/ for a discussion of how design values are calculated, noting in
particular that concentrations flagged as natural events (e.g, high winds, wildfires, volcanic eruptions) or exceptional
events (e.g., construction, prescribed burning) are not included in these calculations and that no regulatory decisions
on attainment status have been made at this time based on these data.
4-18
-------
location (i.e., Philadelphia) where there was a difference in the reduction required between daily
standards of 40 and 65 |ig/m3.
4.3.2 Concentration-Response Functions
As indicated in Figure 4-1, another key component in the risk model is the set of
concentration-response functions which provide estimates of the relationship between each
health endpoint of interest and ambient PM concentrations. As discussed above, the health
endpoints that have been included in the PM2 5 risk assessment for short-term exposure include
mortality, hospital admissions, and respiratory symptoms not requiring hospitalization; long-term
exposure mortality is also estimated. The health endpoints that have been included in the PM10_2 5
risk assessment for short-term exposure include hospital admissions and respiratory symptoms
not requiring hospitalization. Once it had been determined that a health endpoint was to be
included in the assessment, the assessment includes all estimates of response magnitude from
studies judged suitable for inclusion in this assessment, including those which are not
statistically significant. As discussed in section 4.2.2 above, one of the criteria for inclusion of
studies in the risk assessment is that studies have enough sample size to provide a sufficient
degree of precision. Effect estimates that are not statistically significant are used from studies
judged suitable for inclusion in this assessment to avoid introducing bias into the estimate of the
magnitude of the effect. Both single-pollutant and, where available, multi-pollutant,
concentration-response functions are used from the studies listed in Tables 8 A and 8B of the CD
(see also Appendices 3 A and 3B of this Staff Paper).
As discussed in the CD (section 8.4.2) and Chapter 3 (section 3.6.3), questions were
raised in 2002 about the default convergence criteria (which impact the mean estimate) and
standard error calculations (which result in understated standard errors) used in many of the
short-term PM time-series studies employing generalized additive models (GAMs) in a
commonly used statistical software package. To address these concerns, many of the study
authors performed reanalyses of certain of the studies using alternative statistical estimation
approaches (e.g., GLM with different degrees of freedom and different types of splines), in
addition to using GAMs with a more stringent convergence criterion. To avoid producing a
prohibitively large set of results, the PM risk assessment included concentration-response
functions using only GAM with the more stringent convergence criterion, denoted "GAM
(stringent)," for all urban locations, except Los Angeles.14 It should be noted that the GAM
(stringent) concentration-response functions do not address the issue of understated standard
errors of the coefficient estimates. Thus, the confidence intervals included in the risk assessment
involving use of the GAM (stringent) concentration-response functions are somewhat
14PM2 5 risk estimates for various combinations of statistical estimation approaches (GAM and GLM with
varying degrees of freedom) have been included for Los Angeles as a sensitivity analysis to illustrate the impact of
alternative model specification choices.
4-19
-------
understated. As indicated in the CD, "the extent of downward bias in standard error reported in
these data (a few percent to -15%) also appears not to be very substantial, especially when
compared to the range of standard errors across studies due to differences in population size and
number of days available" (CD, p. 9-35).
More detailed information about the concentration-response relationships used in the PM
risk assessment is provided in Appendix 4A of this Staff Paper. This information includes
population characteristics (e.g., age and disease status), form of the model (e.g., log-linear,
logistic), whether other pollutants were included in the model, lags used, observed minimum and
maximum ambient PM concentrations, and PM coefficients along with lower and upper 5th and
95th confidence intervals.
4.3.2.1 Linear and Nonlinear Models
In assessing or interpreting public health risk associated with exposure to PM, the form of
the concentration-response relationships is a critical component. As discussed in Chapter 3
(section 3.6.6), staff recognizes that while there are likely biological thresholds in individuals for
specific health responses, the available epidemiologic studies do not support or refute the
existence of thresholds at the population level for either long-term or short-term PM exposures
within the range of air quality observed in the studies. Thus, staff has concluded that it is
appropriate to consider health risks estimated not only with the reported linear or log-linear
concentration-response functions, but also with modified functions that incorporate alternative
assumed cutpoints as surrogates for potential population thresholds.
For short-term exposure mortality and morbidity outcomes associated with PM2 5 and
PM10_2 5, the initial base case uses linear or log-linear concentration-response models reported in
the epidemiology studies. These concentration-response relationships are applied down to the
estimated policy-relevant background concentration level. Generally, the lowest measured
concentrations in the short-term exposure studies were relatively near or below the estimated
policy-relevant background levels such that little or no extrapolation was required beyond the
range of data in the studies. In the case of the long-term exposure mortality studies for PM25that
have been included in the risk assessment, the lowest measured long-term levels were in the
range 7.5 to 11 |ig/m3. Staff concludes that the initial base case scenario for this endpoint should
include the reported linear models applied down to 7.5 |ig/m3, which is the lowest of the lowest
measured levels in these long-term studies. Going down to an estimated policy-relevant
background level for short-term exposure studies and to 7.5 |ig/m3 for long-term studies provides
a consistent framework which facilitates comparison of risk estimates across urban locations
within each group of studies and avoids significant extrapolation beyond the range of
concentrations included in these studies.
Additional base case scenarios involved the use of alternative concentration-response
functions. The approach used to develop the alternative functions incorporates a modified linear
slope with an imposed cutpoint (i.e., an assumed threshold) that is intended to reflect an
4-20
-------
inflection point in a typical non-linear, "hockeystick" shaped function, below which there is little
or no population response. This approach also is a surrogate for a non-linear, sigmoidal-shaped
function, in which the cut point is intended to reflect the inflection point at the lower end of the
relationship, below which there is assumed to be little or no population response.
The staff recognizes that the alternative cutpoint analyses assume such a hockeystick
shaped relationship, and it is appropriate to adjust the slope of the upper part of the hockeystick
to be consistent with this assumption. If the data in the original study actually supported a
hockeystick model better than a log-linear model, then the slope of the log-linear fitted
relationship reported by the study would have understated the degree to which PM is associated
with mortality or morbidity above the cutpoint, as shown in Figure 4-2. This rationale applies
equally in the case of long- and short-term exposure mortality and morbidity. Therefore, the
slope of the upward-sloping portion of the hockeystick should not use the slope reported for the
concentration-response relationship but should be adjusted upward.
For the base case scenarios involving alternative cutpoints, the slope of the
concentration-response relationship has been adjusted assuming that the upward-sloping portion
of the hockeystick would be the slope estimated in the original epidemiologic study adjusted by
the inverse of the proportion of the range of PM levels observed in the study that was above the
cutpoint. Staff believes that this simple slope adjustment approach represents a reasonable
approach to illustrate the potential impact of possible non-linear concentration-response
relationships. A more definitive evaluation of the effect of alternative cutpoints and non-linear
models is a subject that should be explored in much needed further research.
Based on the staff evaluation contained in section 3.6.6, a cutpoint of 20 |ig/m3 was
selected as the highest value for inclusion in base case scenarios for short-term exposure
mortality for PM2 5 and short-term exposure morbidity for PM10_2 5. Two additional alternative
cutpoints, 10 and 15 |ig/m3, also were selected to be included in base case scenarios for these
short-term exposure health outcomes, so as to span the range between the initial cutpoint (i.e.,
estimated policy-relevant background) and the upper cutpoint value at roughly 5 |ig/m3 intervals.
With regard to long-term exposure mortality associated with PM2 5 exposures, staff selected
12 |ig/m3 as the highest value for an alternative cutpoint based on the following two
considerations: 1) the confidence intervals in the ACS extended study (Pope et al., 2002) begin
to expand significantly starting around 12 to 13 |ig/m3(see Figure 3-4) indicating greater
uncertainty about the shape of the reported concentration-response relationship at and below this
level and 2) it is unlikely that the relationship is non-linear near the reported mean concentration
levels in the long-term exposure studies (e.g., 14 |ig/m3 in the ACS extended study). An
additional alternative cutpoint of 10 |ig/m3 has been included as a base case scenarios for long-
term exposure mortality, representing an approximate midpoint value between the cutpoints
already selected Results of these analyses are discussed below in section 4.4 and 4.5.
4-21
-------
(0
r
o
."True"
hockeystick
model
• Estimated C-R
function
Low est
Measured Level
PM
Highest Measured
Level
Figure 4-2. Relationship between estimated log-linear concentration-response relationship and hockeystick model with
cutpoint C.
Source: Abt Associates (2005b)
4-22
-------
4.3.2.2 Single and Multi-City Models
As described in section 4.2, staff have selected urban areas based on where epidemiologic
studies have estimated concentration-response relationships. This approach avoids uncertainties
associated with estimating health risks for an area based on a relationship developed for a
different location. Staff has included both single-city and multi-city concentration-response
functions in the current assessment. As discussed in section 3.3.1.1 and in the CD, there are a
number of advantages of using concentration-response relationships obtained from multi-city
studies which combine data from a number of locations that may vary in climate, sources and
concentrations, and other potential risk factors. These advantages include, but are not limited to:
(1) more precise effect estimates due to larger data sets, (2) greater consistency in data handling
and model specification that can eliminate city-to-city variation due to study design, and (3) less
likelihood of publication bias or exclusion of reporting of negative or nonsignificant findings.
However, at this time very few multi-city studies have been carried out in the U.S. that report
concentration-response relationships for PM2 5 and/or PM10_2 5. In the one instance where both
single- and multi-city concentration-response relationships are available for the locations
included in the risk assessment (e.g., the Six Cities study), risk estimates have been developed
using both the single- and multi-city concentration-response relationships. In addition, a
sensitivity analysis (S7 in Figure 4-1) has been conducted to examine the potential impact on
short-term exposure mortality of using a single multi-city concentration-response function from
the Six Cities study (Schwartz, 2003b) across five of the PM25 locations included in the risk
assessment compared to use of location-specific concentration-response functions from single-
city studies. The results of this sensitivity analysis are presented in section 4.4.3.2.
4.3.2.3 Single and Multi-Pollutant Models
For several of the epidemiologic studies from which concentration-response relationships
for the PM risk assessment were obtained, concentration-response functions are reported both for
the case where only PM levels were entered into the health effects model (i.e., single-pollutant
models) and where PM and one or more other measured gaseous co-pollutants (i.e., ozone,
nitrogen dioxide, sulfur dioxide, carbon monoxide) were entered into the health effects model
(i.e., multi-pollutant models). To the extent that any of the co-pollutants present in the ambient
air may have contributed to the health effects attributed to PM in single-pollutant models, risks
attributed to PM might be overestimated where concentration-response functions are based on
single-pollutant models. However, the CD finds that associations for various PM indices with
mortality or morbidity are robust to confounding by co-pollutants (CD, p.9-37). Given that
single and multi-pollutant models each have both potential advantages and disadvantages, with
neither type clearly preferable over the other in all cases, risk estimates based on both single and
multi-pollutant models have been developed for the assessment.
4-23
-------
4.3.2.4 Single, Multiple, and Distributed Lag Functions
The question of lags and the problem of correctly specifying the lag structure in a model
are discussed extensively in the CD (section 8.4.4) and in section 3.6.5 of this Staff Paper. As
noted in those discussions, it is important to consider the pattern of results that is seen across the
series of lag periods. In staff s judgment, observation of a consistent pattern of results across
adjacent lags in a study supports use of a study in the risk assessment. In contrast, where an
inconsistent pattern of results has been observed, staff judges that it would be inappropriate to
include results from such studies in the risk assessment.
As noted in section 3.6.5.1, staff concludes that it is appropriate to use single-day lag
period results for the risk assessment. When a study reports several single lag models, unless the
study authors identify a "best lag," the following lag models were included in the risk assessment
based on the assessment in the CD and in section 3.6.5.1:
both 0- and 1-day lag models for mortality (both total and cause specific),
both 0- and 1-day lag models for cardiovascular and respiratory hospital
admissions, and
0-, 1-, and 2-day lag models (if all three were available) for COPD hospital
admissions.
When there is an observed pattern showing effects across different lags, use of any
single-day lag with the largest effect, while reasonable, is likely to underestimate the overall
effect size (since the largest single-lag day results do not fully capture the risk also distributed
over adjacent or other days) (CD, p. 8-270). As discussed in section 3.6.5.1, there is recent
evidence (Schwartz, 2000b, reanalyzed in Schwartz, 2003b), that the relationship between PM
and health effects may best be described by a distributed lag (i.e., the incidence of the health
effect on day n is influenced by PM concentrations on day n, day -1, day -2 and so on). If this is
the case, a model that includes only a single lag (e.g., a 0-day lag or a 1-day lag) is likely to
understate the total impact of PM. Because of this, a distributed lag model may be preferable to
a single lag model. However, distributed lag models have been used in only a few cases and
only for PM10.
The risk assessment includes a sensitivity analysis (S5 in Figure 4-1) which examines the
potential impact of using a distributed lag approach for short-term exposure mortality associated
with PM25 based on the distributed lag analysis of PM10 and mortality (Schwartz, 2000b,
reanalyzed in Schwartz, 2003b). This sensitivity analysis has been included to provide a very
rough sense of the possible underestimation of risk due to use of single-day lags models.
4-24
-------
4.3.2.5 Alternative Short-Term Exposure Model Specifications
As discussed in section 3.6.3, time series studies investigating health effects associated
with PM have used a range of alternative model specifications. For the risk assessment base case
analyses, only the concentration-response functions using the more stringent convergence
criterion, denoted "GAM (stringent), have been included to provide a consistent basis for
comparison across studies and locations. In fact, most studies use a limited number of modeling
approaches, in part to avoid producing an unwieldy and confusing number of different estimates.
One study that used a wide variety of modeling approaches is the Moolgavkar (2003) study
which is the basis for the short-term PM2 5 exposure mortality and morbidity estimates for Los
Angeles. This study included two different versions of the "GAM stringent" approach, one with
30 degrees of freedom and the other with 100 degrees of freedom, as well as models using GLM.
The risk assessment includes a sensitivity analysis to examine the potential impact of alternative
model specifications on estimates of short-term exposure morbidity and mortality in Los Angeles
based on the results from Moolgavkar (2003).
4.3.2.6 Long-term Exposure Models
The available long-term exposure mortality concentration-response functions are all
based on cohort studies, in which a cohort of individuals is followed over time. As discussed in
section 3.3.1.2, based on the evaluation contained in the CD and the staffs assessment of the
complete data base addressing mortality associated with long-term exposure to PM2 5, staff have
concluded that two cohorts that have been studied are particularly relevant for the PM2 5 risk
assessment. These include (a) the Six Cities study cohort, referred to here as Krewski et al.
(2000) - Six Cities, and (b) the American Cancer Society (ACS) cohort, referred to as Krewski et
al. (2000) - ACS, containing a larger sample of individuals from many more cities. In addition,
Pope et al. (2002) extended the follow-up period for the ACS cohort to sixteen years and
published findings on the relation of long-term exposure to PM25 and all-cause mortality as well
as cardiopulmonary and lung cancer mortality (referred to here as Pope et al. (2002) - ACS
extended). EPA's use of these particular cohort studies to estimate health risks associated with
long-term exposure to PM25 is consistent with the views expressed in the NAS (2002) report,
"Estimating the Public Health Benefits of Proposed Air Pollution Regulations," and the SAB
Clean Air Act Compliance Council review of the proposed methodology to estimate the health
benefits associated with the Clean Air Act (SAB, 2004).
As explained in section 3.6.5.4, three different indicators of long-term PM25 exposure
were considered in the extended ACS study; and staff have selected the concentration-response
function associated with an average of the 1979-1983 and 1999-2000 PM25 ambient
concentrations to use in the current risk assessment. The assessment includes a sensitivity
analysis (S6 in Figure 4-1) which examines the potential impact on mortality associated with
long-term exposure of different assumptions about the role of historical air quality
concentrations in contributing to the reported effects.
4-25
-------
4.3.3 Baseline Health Effects Incidence Rates and Population Estimates
As illustrated in Figure 4-1, the most common health risk model expresses the reduction
in health risk (Ay) associated with a given reduction in PM concentrations (Sx) as a percentage
of the baseline incidence (y). To accurately assess the impact of PM air quality on health risk in
the selected urban study locations, information on the baseline incidence of health effects (i.e.,
the incidence under recent air quality conditions) and population size in each location is
therefore needed. Population sizes, for both total population and various age ranges used in the
PM risk assessment were obtained for the year 2000 from the 2000 U.S. Census data15 and are
summarized in Table 4-6. Where possible, county-specific incidence or incidence rates have
been used. County-specific mortality incidences were available for the year 2001 from CDC
Wonder (CDC, 2001), an interface for public health data dissemination provided by the Centers
for Disease Control (CDC). The baseline mortality rates for each risk assessment location are
provided in Table 4-7 and are expressed as a rate per 100,000 general population.16
County-specific rates for cardiovascular and respiratory hospital discharges, and various
subcategories (e.g., pneumonia, asthma), have been obtained, where possible, from state, local,
and regional health departments and hospital planning commissions for each of the risk
assessment locations.17 Baseline hospitalization rates used in each PM25 and PM10_25 risk
assessment location are summarized in Table 4-8 and are expressed a rate per 100,000 general
population. For respiratory symptoms in children, the only available estimates of baseline
incidence rates were from the studies that estimated the concentration-response relationships for
those endpoints. However, because the risk assessment locations for these endpoints were
selected partly on the basis of where studies were carried out, baseline incidence rates reported in
these studies should be appropriate for the risk assessment locations to which they were applied.
4.3.4 Characterizing Uncertainty and Variability
An important issue associated with any population health risk assessment is the
characterization of uncertainty and variability. Uncertainty refers to the lack of knowledge
regarding both the actual values of model input variables (parameter uncertainty) and the
physical systems or relationships (model uncertainty - e.g., the shapes of concentration-response
15 See http ://factfinder. census, gov/.
16Since the baseline incidence rates are expressed in terms of cases per 100,000 general population, the
general population estimates have been used in combination with these rates to generate the baseline incidence in
each location for various effects in calculating the risk estimates.
17The data were annual hospital discharge data, which were used as a proxy for hospital admissions.
Hospital discharges are issued to all people who are admitted to the hospital, including those who die in the hospital.
Use of the annual discharge rate is based on the assumption that admissions at the end of the year that carry over to
the beginning of the next year, and are therefore not included in the discharge data, are offset by the admissions in
the previous year that carry over to the beginning of the current year.
4-26
-------
Table 4-6. Relevant Population Sizes for PM Risk Assessment Locations
City
Boston1
Detroit2
Los Angeles3
Philadelphia4
Phoenix5
Pittsburg6
San Jose7
Seattle8
St. Louis9
Populationa
Total
2806000
2061000
9519000
1518000
3072000
1282000
1683000
1737000
2518000
Ages 7-14
283,000 (10%)
308,000 (12%)
Ages 525
1,903,000 (68%)
—
—
—
—
—
—
—
1,637,000 (65%)
Ages 5 30
1,673,000 (60%)
1,153,000 (56%)
5,092,000 (53%)
852,000 (56%)
1,684,000 (55%)
814,000 (64%)
965,000 (57%)
1,044,000 (60%)
1,475,000 (59%)
Ages <65
—
—
—
—
—
—
—
1,555,000
(90%)
—
Ages 5 65
249,000 (12%)
927,000 (10%)
—
359,000 (12%)
—
—
—
—
Ages <75
—
—
—
—
—
1,166,000
(91%)
—
—
—
Ages 5 75
—
—
—
—
—
116,000(9%)
—
—
—
a Total population and age-specific population estimates taken from the CDC Wonder website are based on 2000 U.S. Census data. See
http://factfmder.census.gov/. Populations are rounded to the nearest thousand. The urban areas given in this table are those considered in the studies used in the
PM2 5 risk assessment. The percentages in parentheses indicate the percentage of the total population in the specific age category.
1 Middlesex, Norfolk, and Suffolk Counties. 2 Wayne County. 3 Los Angeles County. 4 Philadelphia County.
5 Maricopa County. 6 Allegheny County. 7 Santa Clara County. 8 King County.
9 St. Louis, Franklin, Jefferson, St. Charles, Clinton (IL), Madison (JL), Monroe (IL), and St. Clair (IL) Counties and St. Louis City.
4-27
-------
Table 4-7. Baseline Mortality Rates for 2001 for PM2 5 Risk Assessment Locations
Health Effect
Boston1
Detroit2
Los
Angeles3
Philadelphia4
Phoenix5
Pittsburgh6
San
Jose7
St.
Louis8
Seattle9
National
Average
A. Mortality Rates Used in Risk Analysis for Short-Term Exposure Studiesa'b (deaths per 100,000 general population/year)
Non-accidental (all ages):
ICD-9 codes < 800
Non-accidental (75+):
ICD-9 codes < 800
Non-accidental (<75):
ICD-9 codes < 800
Cardiovascular (all ages):
ICD-9 codes: 390-459
Cardiovascular (all ages):
ICD-9 codes: 390-448
Cardiovascular (65+):
ICD-9 codes: 390-448
Cardiovascular (all ages):
ICD-9 codes: 390-429
Ischemic Heart Disease (all
ages): ICD-9 codes: 410-
414
Respiratory (all ages):
ICD-9 codes: 11, 35, 472-
519, 710.0, 710.2, 710.4
Respiratory (all ages):
ICD-9 codes: 460-519
776
—
—
—
—
—
—
122
—
—
916
—
—
416
—
—
—
—
—
72
581
—
—
—
—
—
207
—
—
—
1070
—
—
—
418
—
—
—
—
—
—
—
—
—
—
211
—
—
—
—
—
761
399
—
—
—
—
—
—
—
494
—
—
206
—
—
—
—
51
—
869
—
—
—
—
—
—
206
—
—
—
—
—
—
—
—
—
—
—
—
791
469
322
328
324
273
252
152
80
79
4-28
-------
Health Effect
COPD without Asthma (all
ages): ICD-9 codes: 490-
492, 494-496
Pneumonia (all ages):
ICD-9 codes: 480-487
Boston1
36
26
Detroit2
—
—
Los
Angeles3
—
—
Philadelphia4
—
—
Phoenix5
—
—
Pittsburgh6
—
—
San
Jose7
—
—
St.
Louis8
39
27
Seattle9
—
—
National
Average
42
22
B. Mortality Rates Used in Risk Analysis for Long-term Exposure Studiesa'b (deaths per 100,000 general population/year)
Total mortality (25+):
ICD-9 codes: all
Total mortality (30+):
ICD-9 codes: all
Cardiopulmonary
Mortality (25+): ICD-9
codes: 400-440, 485-495
Cardiopulmonary
Mortality (30+): ICD-9
codes: 401-440, 460-519
Lung Cancer Mortality
(30+): ICD-9 code: 162
803
797
297
347
55
—
937
—
468
64
—
591
—
313
33
—
1100
—
489
72
—
676
—
313
42
—
1189
—
573
78
—
499
—
247
30
905
897
391
439
61
—
637
—
287
44
822
814
341
391
55
*The epidemiologic studies used in the risk assessment reported causes of mortality using the 9th revision of the International Classification of Diseases (ICD-9)
codes. However, the 10th revision has since come out, and baseline mortality incidence rates for 2001 shown in this table use ICD-10 codes. The groupings of
ICD-9 codes used in the epidemiologic studies and the corresponding ICD-10 codes used to calculate year 2001 baseline incidence rates is given in Exhibit 5.4 of
the TSD (Abt Associates, 2005b).
a Mortality figures were obtained from CDC Wonder for 2001. See http://wonder.cdc.gov/.
b Mortality rates are presented only for the locations in which the concentration-response functions were estimated. All incidence rates are rounded to the nearest
unit. Mortality rates for St. Louis may be slightly underestimated because some of the mortality counts in the smaller counties were reported as missing in CDC
Wonder.
1 Middlesex, Norfolk, and Suffolk Counties. 2 Wayne County. 3 Los Angeles County. 4 Phil. County. 5 Maricopa County. 6 Allegheny County. 7 Santa Clara
County. 8 St. Louis, Franklin, Jefferson, St. Charles, Clinton (IL), Madison (IL), Monroe (IL), and St. Clair (IL) Counties and St. Louis City. 9 King County.
4-29
-------
Table 4-8. Baseline Hospitalization Rates for PM Risk Assessment
Locations*
Health Effect
Detroit1
Los Angeles2
Seattle3
Hospital Admissions (per 100,000 general population/year)
Pneumonia admissions (65 and over): ICD codes 480-486
COPD and asthma admissions (all ages): ICD codes 490-
496
COPD and asthma admissions (65 and over): ICD codes
490-496
Asthma (<65): ICD code 493
Cardiovascular admissions (65 and over): ICD codes:
390-429
Ischemic heart disease (65 and over): ICD codes 410-414
Dysrhythmias (65 and over): ICD code 427
Congestive heart failure (65 and over): ICD code 428
250
—
192
—
—
487
161
341
—
318
—
—
728
—
—
—
—
—
—
92
—
—
—
—
* Hospitalization rates are presented only for the locations in which the concentration-response functions were
estimated. For each location, the number of discharges was divided by the location's population from the 2000 U.S.
Census estimates to obtain rates. All incidence rates are rounded to the nearest unit.
'Wayne County. Year 2000 hospitalization data were obtained from the Michigan Health and Hospital Association.
2Los Angeles County. Year 1999 hospitalization data were obtained from California's Office of Statewide Health
Planning and Development - Health Care Information Resource Center.
3King County. Year 2000 hospitalization data were obtained from the State of Washington Department of Health,
Center for Health Statistics, Office of Hospital and Patient Data Systems.
4-30
-------
functions). In any risk assessment uncertainty is, ideally, reduced to the maximum extent
possible, but significant uncertainty often remains. It can be reduced by improved measurement
and improved model formulation. In addition, the degree of uncertainty can be characterized,
sometimes quantitatively. For example, the statistical uncertainty surrounding the estimated
PM2 5 and PM10_2 5 coefficients in the reported concentration-response functions is reflected in the
confidence intervals provided for the risk estimates in this chapter and in the TSD. Additional
uncertainties are addressed quantitatively through sensitivity analyses and/or qualitatively and
have been discussed throughout section 4.3.
As noted above, the updated risk assessment presents qualitative and quantitative
considerations of uncertainty, including sensitivity analyses of key individual uncertainties.
Given the existing data gaps in the scientific evidence and associated uncertainties, a more
comprehensive integrated assessment of uncertainties, would be desirable, but in the staffs
judgment would require use of techniques involving elicitation of probabilistic judgments from
health scientists. While the Agency is currently developing these approaches, such
comprehensive assessments of uncertainty are not available for the current risk assessment for
this PM NAAQS review.
Variability refers to the heterogeneity in a population or variable of interest that is
inherent and cannot be reduced through further research. For example, there may be variability
among concentration-response functions describing the relation between PM2 5 and mortality
across urban areas. This variability may be due to differences in population (e.g., age
distribution), population activities that affect exposure to PM (e.g., use of air conditioning),
levels and composition of PM and/or co-pollutants, and/or other factors that vary across urban
areas.
The current risk assessment incorporates some of the variability in key inputs to the
assessment by using location-specific inputs (e.g., location-specific concentration-response
functions, baseline incidence rates, and air quality data). Although spatial variability in these
key inputs across all U.S. locations has not been fully characterized, variability across the
selected locations is imbedded in the assessment by using, to the extent possible, inputs specific
to each urban area. Temporal variability is more difficult to address, because the risk reduction
portions of the risk assessment (i.e., estimated risk reduction associated with just meeting
specified standards) focus on some unspecified time in the future when specified PM standards
are just met. To minimize the degree to which values of inputs to the assessment may be
different from the values of those inputs at that unspecified time, we have used the most current
inputs available (i.e., year 2003 air quality data for most locations and the most recent available
mortality baseline incidence rates (from 2001)). However, we have not tried to predict future
changes in inputs (e.g., future population levels or possible changes in baseline incidence rates).
4-31
-------
Key Uncertainties and Assumptions
The following briefly summarizes the major sources of uncertainties and variability and
how they are dealt with in the risk assessment:
Causality. There is uncertainty about whether each of the estimated associations
between the two PM indicators (PM2 5 and PM10_2 5) and the various health
endpoints included in this risk assessment actually reflect a causal relationship.
There are varying degrees of uncertainty associated with the various PM
indicators and health endpoints related to differences in the weight of evidence
supporting judgments about whether an observed association truly reflects a
causal relationship. For example, there is much greater uncertainty associated
with the morbidity effects associated with PM10_2 5 exposures than for PM2 5 due to
the much smaller health effects data base and greater concerns about exposure
measurement error. Chapter 3 presents a more detailed discussion of the staffs
qualitative assessment of the varying weight of evidence associated with the
effects included in the risk assessment.
• Empirically estimated concentration-response relationships. In estimating the
concentration-response relationships, there are uncertainties: (1) surrounding
estimates of PM coefficients in concentration-response functions used in the
assessment, (2) concerning the specification of the concentration-response model
(including the shape of the concentration-response relationship) and whether or
not a population threshold or non-linear relationship exists within the range of
concentrations examined in the studies, and (3) related to the extent to which
concentration-response functions derived from studies in a given location and
time when PM concentrations were higher provide accurate representations of the
concentration-response relationships for the same location with lower annual and
daily PM concentrations. For the few instances where multi-city PM
concentration-response functions are included in the base case analyses (e.g., use
of the Six-Cities study function for respiratory symptoms associated with short-
term exposures to PM2 5 applied in Boston and St. Louis), there also is uncertainty
related to the transferability of PM concentration-response functions from
multiple locations to the specific location selected for the risk assessment.18
Statistical uncertainty, based on the standard errors reported in the epidemiologic
studies, is incorporated in the risk assessment and is discussed below. Base case
risk estimates incorporating various cutpoints have been included in the risk
assessment to reflect the uncertainty about whether or not population thresholds
1 8
A concentration-response function derived from a multi-cities study may not provide an accurate
representation of the concentration-response relationship in a specific assessment location because of (1) variations
in PM composition across cities, (2) the possible role of associated co-pollutants in influencing PM risk, (3)
variations in the relation of total ambient exposure (both outdoor exposure and ambient contributions to indoor
exposure) to ambient monitoring in different locations (e.g, due to differences in air conditioning use in different
regions of the U.S.), (4) differences in population characteristics (e.g., the proportions of members of sensitive
subpopulations) and population behavior patterns across locations.
4-32
-------
or non-linear concentration-response relationships might exist at the lower range
of ambient PM25 and PM10_25 concentrations. As discussed previously (see
section 4.3) several sensitivity analyses have been presented related to
uncertainties in the concentration-response relationships.
Adequacy of ambient PM monitors as surrogate for population exposure. The
extent to which there are differences in the relationship between spatial variation
in ambient PM2 5 or PM10_2 5 concentrations and ambient exposures in the original
epidemiology studies compared to more recent ambient PM2 5 or PM10_2 5 data
introduces additional uncertainty in the risk estimates. This is expected to be
more of a concern for PM10_2 5 where greater spatial variability in ambient
monitoring data within urban areas has been observed.
Adjustment of air quality distributions to simulate just meeting specified
standards. The shape of the daily distribution of PM25 and PM10_25 ambient
concentrations that would result upon meeting alternative PM standards is
unknown. Based on an analysis of historical data, staff believes it is a reasonable
assumption that PM2 5 concentrations would be reduced by roughly the same
percentage. However, there is much greater uncertainty associated with the use of
this same approach for meeting PM10_2 5 standards given the lack of sufficient data
to evaluate the reasonableness of this assumption.
Background concentrations. Since one of the base case scenarios includes
estimating risks in excess of estimated policy-relevant background, uncertainty
about background concentrations contributes to uncertainty about the risk
estimates. As discussed previously, the assessment includes sensitivity analyses
examining the impact of alternative constant and varying daily background levels
on the risk estimates.
Baseline incidence rates and population data. There are uncertainties related to:
(1) the extent to which baseline incidence rates, age distribution, and other
relevant demographic variables that impact the risk estimates vary for the year(s)
when the actual epidemiology studies were conducted, the recent year of air
quality used in the assessment, and some unspecified future year when air quality
is adjusted to simulate just meeting the current or alternative standards; (2) the
use of annual incidence rate data to develop daily health effects incidence data;
and (3) related to the use of an overall combined incidence rate for six cities for
the respiratory symptoms endpoint which is applied to individual cities (i.e.,
Boston and St. Louis). Spatial variability in baseline incidence and population
data is taken into account by use of city-specific data in most cases.
The uncertainties from some of these sources — in particular, the statistical uncertainty
surrounding estimates of the PM coefficients in concentration-response functions — are
characterized quantitatively in the PM risk assessment. It is possible, for example, to calculate
4-33
-------
confidence intervals around risk estimates based on the uncertainty associated with the estimates
of PM coefficients used in the risk assessment. These confidence intervals express the range
within which the risks are likely to fall if the sampling error uncertainty surrounding PM
coefficient estimates were the only uncertainty in the assessment.19 In situations where the point
estimate for a concentration-response function is positive, but the lower confidence limit
estimate is less than 1.0, the lower confidence limit of the risk estimate is a negative value.
Based on the overall body of evidence on the relationships between PM and health effects, the
staff believes that these negative estimates should not be interpreted as implying that increasing
PM levels will result in reduced risks, but rather that the negative risk estimates are simply a
result of statistical uncertainty in the reported concentration-response relationships in the
epidemiologic studies.
Steps also have been taken to minimize some of the uncertainties noted above. For
example, the current PM risk assessment includes only health endpoints for which the CD
evaluation or staff assessment (see Chapter 3) find that the overall weight of the evidence
supports the conclusion that PM25 is likely causally related. Also, for most of the health
endpoints and locations included in the risk assessment, this assessment uses the concentration-
response functions derived from epidemiologic studies carried out in those same locations. This
serves to minimize the uncertainties, such as differences in composition and differences in
factors affecting human exposure associated with applying concentration-response functions
developed in one location to a different location. However, the use of functions from single-city
studies does suffer the disadvantage of introducing possible publication bias and single-city
studies generally have lower precision than larger multi-city studies.
In summary, the key assumptions on which the current PM risk assessment is based
include the following:
• The relationship between PM2 5 and PM10_25 and the health endpoints included in
the assessment are causal;
• Concentration-response models are appropriately specified (including the
functional form and lag structure);
Baseline incidence rates and population size and age distributions have not
changed appreciably from those used in the assessment;
• For short-term endpoints, that obtaining average daily incidence rates from annual
baseline incidence rates and using them to estimate daily incidences associated
with exposure to PM does not bias the estimates;
19However, as discussed earlier in section 4.2.6, for the short-term concentration-response functions based
on reanalyzed GAM (stringent) models the confidence intervals are somewhat understated.
4-34
-------
• The distribution of PM concentrations on missing days is essentially the same as
the distribution on days for which we have PM data;
The estimated policy-relevant background concentrations for PM2 5 and PM10_2 5
are appropriate for each urban area included in the analysis;
A single year of air quality data is appropriate to characterize risks associated
with recent air quality levels and just meeting specified standards;
• Proportional rollback of concentrations over an estimated policy-relevant
background appropriately reflects the air quality distribution when specified
standards would just be met.
4.4 PM2 5 RISK ESTIMATES
Several "base case" analyses for PM2 5 are presented in this section and include risk
estimates associated with a recent year of air quality (generally, 2003), air quality adjusted to just
meet the current PM25 standards, and air quality adjusted to simulate just meeting alternative
PM2 5 standards. The initial base case analyses for the recent air quality and just meeting the
current PM2 5 standards scenarios include concentration-response models that extend down to
estimated policy-relevant background for short-term exposure health outcomes (i.e., equivalent
to setting a cutpoint at estimated policy-relevant background) and extend down to 7.5 |ig/m3 for
long-term exposure mortality (equivalent to setting a cutpoint at 7.5 |ig/m3). For this initial set
of base case analyses, the slope of the concentration-response function is based on that obtained
directly from the published studies. A variety of models (single city with different lags, single
city vs. multi-city, single pollutant vs. multi-pollutant) and health outcomes (mortality, hospital
admissions, respiratory symptoms) are included using this set of initial base case analyses.
Following this initial set of base case analyses, risk estimates, additional base case
estimates are developed only for non-accidental short-term exposure mortality (or if not
available, cardiovascular mortality) and for all-cause mortality with long-term exposure for each
study area employing the same cutpoints indicated above, as well as several additional
alternative cutpoints. For the additional alternative cutpoints, the slope of the concentration-
response function has been modified based on a simple hockeystick model (see discussion in
section 4.3.2.1). Staff considers the initial set of base case analyses, as well as the analyses
estimating health risks associated with alternative cutpoints, as being part of the complete set of
base case analyses.
4.4.1 Recent Air Quality
4.4.1.1 Base Case Risk Estimates Above Initial Cutpoint
The base case risk estimates associated with recent PM2 5 concentrations in excess of
policy-relevant background levels for short-term exposure outcomes and in excess of 7.5 |ig/m3
4-35
-------
for long-term exposure mortality are presented in a series of figures in this section. The risk
estimates are expressed both in terms of percent of total incidence (the top panel in each figure)
and cases per hundred thousand general population (the bottom panel in each figure). The
percent of total incidence provides information about what portion of total incidence for a given
health outcome is estimated to be due to exposure to ambient PM2 5 levels. Expressing risk in
terms of cases per hundred thousand general population provides a metric that takes into account
the variation in population size for each of the urban areas. For each series of estimates, a point
estimate is provided along with 95% confidence intervals.20 Additional detailed tables which
present the estimated incidence (both as the number of effects and as a percentage of total
incidence) for each risk assessment location are included in the TSD. Risk estimates in a given
assessment location are presented only for those health endpoints for which there is at least one
acceptable concentration-response function reported for that location. Therefore, the set of
health effects shown in the figures varies for the different locations.
Figures 4-3 through 4-7 present the PM2 5 risk estimates across the various assessment
locations associated with recent concentrations in excess of an initial cutpoint. For short-term
exposure outcomes, this initial cutpoint is the estimated policy-relevant background. For long-
term exposure mortality the initial cutpoint is 7.5 |ig/m3, the lowest of the lowest measured levels
used in the long-term exposure studies included in the risk assessment. Figure 4-3 compares risk
estimates for mortality associated with short-term (i.e., 24-hour) exposure to PM2 5 above policy-
relevant background using single-pollutant, single-city models. The point estimates are in the
range from about 0.8 to 2.5% of total non-accidental mortality incidence. In terms of cases per
hundred thousand general population, the point estimates range from about 4 to 13. The
differences in estimates across locations is due to several factors including differences in recent
air quality levels, use of different concentration-response functions from various single-city
studies, and variation in baseline incidence rates. Differences in concentration-response
functions across the various single-city studies may reflect methodological differences between
studies and/or real differences due to differences in the population and extent of population
exposure to ambient PM2 5 concentrations. In addition, there are significant differences in
baseline incidence rates among the cities which also contribute to city-to-city variation. For
example, as shown in Table 4-7, the baseline mortality rate for non-accidental mortality (all
ages) is nearly twice as large in Philadelphia as in Los Angeles.
Figure 4-4 compares risk estimates for non-accidental and cause-specific mortality
associated with short-term exposure to PM2 5 above policy-relevant background based on single
20As noted above, in some cases, where the lower confidence limit of the concentration-response function is
less than 1.0, the resulting lower confidence limit of the risk estimate is a negative value. The staffs interpretation
of these negative values is that while they indicate statistical uncertainty about the concentration-response
relationships, they do not at all suggest that risk reductions would be associated with an increase in PM levels.
4-36
-------
6 '
5 '
O
ID
T3
I 3 '
CD
0
C 2
, ro" °> -
ro o J5
6 II S I
- " » j?
T3
.
O ZIs
I
Mooigavkar
J
(2003)
0
^ o
£
n
X!
CO
1 j?
^ ro
iC I
S o
0
CO
0 CO
>- 03
_OJ "O
•— o
1 1 £
Chock et al. (2000)
0-day lag
I | III I | , | | | |
Boston
Detroit
Los Angeles
Pittsburgh
San Jose
St. Louis
50 -
45 •
40 -
35 '
30 -
2£
^
20 -
1 5 •
1 0 -
5 -
0
-5 -
-1 0 '
A C
2
(r
c
c
I"
>
; a> ^ en
Chock et al. (2000)
0-day lag
I 1 CO
0 CD —
T Q m js
^i S& le ^^ S
12 ra c_i. tn >^ nj (— j CD
^T3 ^ o>
)
I
"i
5 IE
| LO
^ 1*^-
I) V
™ ™ a>
\
T
>> >
o n:
K, O5
.m_
3 "O
S, C-
ijL~ ±: T3
% C3
-C
C/)
o
In
O)
T3
C
_JO_
C
o
^2
Q.
O
CL
"co
OJ
-------
1 U
9-
8
O
o 7 '
a*"1
10
a. 6 -
^"
•2- *>
•» =
CD t/i --fe ^— CD
CsJ__ ^> CO "ro CO t±
rti C TO eg 'g CO O e~-"'1 E
Jvs "^"" TO £_j e "&3 ."^ PS csi iQ *^ W
gP "^ 'E OJ '^ -""^ S S "~~" Cl- S OJ
fc ~^§tK ^^eOra.«
^s.-g^ i -g g Sg-^^^^ j
isjciy !=_^5,^ o_^^: ^^ 'c
-I-1 «J •" ^T-, c^m ffi Sp «
OJT=>0> E „ .." ^^^
J "8 ^ ^ 1 S « _^ w | -o -9
en \ f <£> ^ "o ¥ -,—! ^? m "01
C -is "V CO c= '55 n£G
>^ CCJ H1 c >-%
o « 1 ^ V _^
^o> sj ""S -y> "^ '^
"§ I ,1 § i 's
1 to ' I -
\ 1 T * $ T ' (
' crT
o ^^
C3 -^
-e
So
ro E
S -?S
I J T3J S
Klemm and Mason (2003) § §
COPD mortality | |_ ,
1
^
^
_i
^ °" Klemm and Mason (2003)
pneumonia mortality
i 11 i
Boston St. Louis
2 4
O 22 •
3?
g 20 '
T3
c ig-
-2-
.2 1 6 '
'ro
g. " •
O
Q- 12 .
CO
g 10 •
m
*1*
S 8 "
o
0 A -
IT Schwartz (2003b)
•§> non-accidental mortality _>,
1 1 _ CO
-^ vj co t:
••g -9J <=> 0
1 ^ 5 ti
sr "="S S S" ^ 's
l§ g^f§,^ ^^;^
"£i>,^oE^15 rojH Ss
i! ill SI |_f li U
-OQE^IO E°i -^ 2 ™
EJ~^ "^ ^ G) CI " O ' C r-
C_> , , •= Si CT U_ ^.^ m (r
r^ , i 8 f J i 1
, , = . 1 Jf , z f S I -
o"
o
CD
CL
en
CD
O
Boston
St. Louis
Figure 4-4. Estimated annual percent (top panel) and cases per 100,000 general
population (bottom panel) of total (non-accidental) mortality associated with
short-term exposure to PM2 5 above background (and 95 percent confidence
intervals): single-city versus multi-city models.
Source: Abt Associates (2005b)
4-38
-------
1 0 •
9 -
-— -s
0 8
^ 7 '
in
01
-a 6 •
c
-5- 5 -
8 4 '
C 4
0)
•5 3 '
c
— 2 •
S
|2 1 -
H_
o n
j-j 0
c
m
0 -1 -
0)
Q- .2 -
-3 •
-4 '
10 '
9 '
" — •
O
^ 7 •
LO
S 6 •
d
— . * 5
§ 4 •
0)
•5 3 '
c
1 2 '
j_ 1
*° n
*- 0
c
, 1 S
1 -day lag § J3
I j cardiovascular S S.
y.. mortality S
S 0-day lag o
iO y
2 "g 8
^ 8
I '*
Moolgavkar (2003)
•£•
O
if>
i
S
D- o -3?-
S csf
°- Q
c-
o
to
i
_|
age <75
I ,
S 8 °
OJ Q_
O
z
o
CO
"
i
c
5
rj'
D
Fairley (2003)
non-accidental mortality
0-day lag
1
age 75+
i i \ ' ' !
Los Angeles
Pittsburgh
San Jose
Figure 4-5. Estimated annual percent (top panel) and cases per 100,000 general
population (bottom panel) of health effects associated with short-term
exposure to PM2 5 above background (and 95 percent confidence intervals):
single-pollutant versus multi-pollutant models.
Source: Abt Associates (2005b)
4-39
-------
16
1 5
1 4
O 13
5? 12
LO
O) 1 1
8
g
TO
0
-^. 0 Q)
§ r |
r S 1
2&
a> o
~ Q_
2C
1
CO *, c
-< 8 §§
< .re
II 1 II
15 o O
i 5s a.
1 TSo
I •£ °-
' t }
Boston
Detroit
L. A. Philadelphia Phoenix Pittsburgh San Jose Seattle
St. Louis
0>
O
x
CO
J-,
s
~
"
"
ia
1
? ?
i Is
"° CO S c
| Ss!
Q> ' ? W
- 8 - §1*
•> s| J. 8 §1 §•
! I 1 '
U X 0 'i
2 <» S <
g S3 £i <
5 ra ^ _g c
g 5
£
n "x "5
L "* j I
CD
< <
C3 g
" — ' ~~~^
lu ro
<5 "S
2 S
-g a.
1 °"
Krewski et at. (2000)
Six Cities
"O
1
(D
"x "O
CO 5 1
XJ O 75 --
ki et al. (2000) - ACS
- Pope et al. (2002)
ACS extended
u
1
ic
ig o "S ro
% °- •= »
i
1 1
1/3 S6
| £
» 1
}
>
o
"O
n
CD
O
-30
Boston
Detroit
L. A. Philadelphia Phoenix Pittsburgh San Jose Seattle
St. Louis
Figure 4-6. Estimated annual percent (top panel) and cases per 100,000 general
population (bottom panel) of total (non-accidental) mortality associated with
long-term exposure to PM2 5 above 7.5 |ig/m3 (and 95 percent confidence
intervals): single-pollutant models. Source: Abt Associates (2005b)
4-40
-------
16 -
1 5 -
1 4 -
O u "
<$ 12 -
S 1 1 •
"g 10 -
CO „ .
•• — ' y
CD
o 8 -
CD 7 -
'P 6 '
c
_ 5 _
|O 4
•5 3 •
c 2 -
E 1 -
CD .
Q. 0
-1 -
-2 -
-3 '
o
i
C
"L
o
i
L
5
>
2
C
C,
O
o
S
J I
o
C"
C
CM:
O
Z
o
O
§
3
O
«
1 1
Boston Detroit
160"
5140-
•*p
s^
in
01 1 20 H
"O
c
as
c 1 00 '
o
f 80 '
o
Q.
"ro 60 '
c
0
CD 40 '
o
o
o" 20 '
o
05
co
CD
ro -20 •
O
> A .
"~i
"
,O^^ ' ' "
IE Q O O
o w
^l' S ,
(1 ™ |oOgs
+ T t f ,
1 1 1 1 1 1 1
L. A. Philadelphia Phoenix Pittsburgh San Jose Seattle St. Louis
o
C
O
CO
c
c
D
•
L
CM
O
~Z.
3
J
(•
c
O
3
O
Z
o
o
5
S
D-
O
o
0
2
o
u
.>-
0
a
O
ca
* 0
8Z8
2 1 1 . j
nti
Si
3
5
L
CO
O
O
cs*
O O 01
£O ^O O
c
o
"c O
o O z ro
g° j O^ ^ ^
i 1 f 1 °> 5o9«^
|tH, i"^°g
I «»**+
I \ I \ \ I \ J \ j I J I I I I I I
Boston Detroit L. A. Philadelphia Phoenix Pittsburgh San Jose Seattle St. Louis
Figure 4-7. Estimated annual percent (top panel) and cases per 100,000 population
(bottom panel) of total (non-accidental) mortality associated with long-term
exposure to PM2 5 above 7.5 ^ig/m3 (and 95 percent confidence intervals):
single-pollutant and multi-pollutant models (based on Krewski et al. (2000) -
ACS study). Source: Abt Associates (2005b)
4-41
-------
city versus multi-city models. Generally, the estimated incidence for the single- and multi-city
models are roughly comparable, with somewhat lower risk estimates seen in Boston for the
multi-city models compared to the single-city models and the reverse being observed in St.
Louis.
Figure 4-5 compares risk estimates based on single-pollutant versus multi-pollutant
concentration-response models provided in the epidemiologic studies for PM2 5 short-term
exposure health endpoints above policy relevant background. As noted earlier, the multi-
pollutant models reflect where PM2 5 and one or more other measured gaseous co-pollutants (i.e.,
O3, NO2, SO2, CO) were entered into the health effects model. In two cases there is relatively
little difference in the risk estimates between the single-pollutant and multi-pollutant models
(i.e., Pittsburgh and San Jose), while in the third case (Los Angeles) there are larger differences
when either CO or NO2 are added to the model along with PM2 5.
Figures 4-6 and 4-7 show risk estimates for mortality related to long-term (i.e., annual
average) exposure to PM25 levels above 7.5 |ig/m3 based on single- and multi-pollutant models,
respectively. The point estimates for the single-pollutant models, based on the ACS-extended
study (Pope et al., 2002), range from about 0.5% in Seattle to as high as about 6.5% of total
mortality in Los Angeles, with most point estimates falling in the 2 to 5% range. The point
estimates based on the original ACS study (Krewski et al., 2000) are somewhat lower in all of
the study areas (ranging from about 0.2 to 5% in terms of percent of total incidence). For Boston
and St. Louis, the point estimates based on the Six Cities study (Krewski et al., 2000) are more
than twice as large as the estimates based on the ACS extended study. As noted in Chapter 3
(section 3.5.2), the strongest associations between PM25 and mortality in the ACS study were
among the less educated participants who form a relatively small portion of the total study
cohort. If the education distribution were adjusted to reflect the education distribution in the
general U.S. population, the summary effect estimate would increase and this would narrow the
difference observed in the risk estimates between the ACS and Six Cities studies. As shown in
Figure 4-7, the risk estimates based on multi-pollutant models, involving addition of different
single co-pollutants in the ACS study, show generally greater risk associated with PM25 when
CO, NO2, or O3 were added to the models and lower risk associated with PM25 when SO2 was
added.21
4.4.1.2 Base Case Risk Estimates Above Various Cutpoints
As discussed above, the assessment includes additional base case annual short- and long-
term mortality risk estimates associated with recent air quality levels assuming not only the
21 The addition of a second pollutant reduced the number of cities available for estimating the
concentration-response function from 50 for PM2 5 alone to 44 with addition of CO, to 33 with addition of NO2, to 45
with addition of O3 and to 3 8 with addition of SO2. The effect of the reduction in the number of cities available for
each analysis is to increase the size of the confidence intervals.
4-42
-------
initial outpoint included in the previous section but the alternative cutpoint levels as well. For
short-term exposure mortality, a single non-accidental mortality function has been included,
except for Philadelphia and Phoenix where cardiovascular mortality has been used since a
suitable non-accidental mortality concentration-response function is not available. For long-term
exposure mortality, a single all cause mortality concentration-response function has been
included based on the ACS-extended study (Pope et al., 2002). Tables 4-9 and 4-10 present the
annual health risks for short- and long-term exposure mortality, respectively. Both tables present
the risk estimates expressed in terms of incidence (i.e., cases), cases per 100,000 general
population, and percent of total incidence, along with 95 percent confidence intervals for each of
these risk metrics.
4.4.1.3 Risk Estimates from Sensitivity Analyses
As discussed previously, several sensitivity analyses have been carried out to provide
some perspective on the impact of various assumptions and uncertainties on the health risk
estimates (see Table 4-4 above for a summary of different types of sensitivity analyses). Most
of these sensitivity analyses have been conducted in each of the study areas and use the initial set
of cutpoints (i.e., policy relevant background for short-term exposure outcomes and 7.5 |ig/m3
for long-term exposure mortality). The complete results of the sensitivity analyses are included
in the TSD. In some cases, sensitivity analyses have been conducted only in one location due to
data constraints (for example Los Angeles is the only city where the sensitivity analysis uses
alternative concentration-response model specifications since, as explained in section 4.3.2.5
above, it is the only study that presents results for a wide range of alternative model
specifications).
Alternative Background Levels
As explained earlier, for purposes of informing decisions about the PM NAAQS, we are
interested in PM-related risks due to concentrations over policy-relevant background levels,
where background excludes anthropogenic emissions of PM and its precursors in the U.S.,
Canada, and Mexico (discussed in section 2.6). One set of sensitivity analyses has examined the
impact of using the lower and upper end of the range of estimated background concentrations
provided in section 2.6. In the nine locations, using the upper- and lower-end of the range of
estimated background generally has a small to modest impact, on the order of roughly +/- 10-
20% change in short-term exposure health endpoint risk estimates compared to use of the
midpoint of the estimated range of background levels. Alternative estimated PM25 background
levels have no impact on long-term exposure mortality in any of the PM25 locations, because the
range of alternative policy-relevant background levels is lower than the lowest cutpoint of
7.5 |ig/m3 used for these analyses.
A sensitivity analysis also has been conducted that focuses on the impact of using a
varying estimated PM25 background concentration instead of the fixed level used in each study
area in the base case assessment. Staff developed a Monte Carlo simulation approach to
4-43
-------
Table 4-9. Estimated Annual Health Risks of Short-Term Exposure Mortality
Associated with Recent PM2.5 Concentrations Assuming Various Outpoint Levels*
Urban Area
Boston
Detroit
Los Angeles
Philadelphia
Phoenix
Study
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Lipfert et al. (2000) - 7
counties
Mar (2003) [reanalysis of
Mar (2000)]
Type
Non-accidental
Non-accidental
Non-accidental
Cardiovascular
Cardiovascular
Ages
all
all
all
all
65+
Lag
mean of lag 0 & 1
day
3 day
Oday
1 day
1 day
Incidence Associated with PM25 Assuming Various Outpoint Levels
(95% Confidence Interval)
Incidence per 100,000 General Population
(95% Confidence Interval)
Percent of Total Incidence
(95% Confidence Interval)
Policy Relevant
Background**
=2.5 or 3.5 ug/m3
390
(265-514)
14
(9-18)
1.8%
(1 .2% - 2.4%)
170
(-170-501)
8
(-8 - 24)
0.9%
(-0.9% - 2.7%)
494
(-62-1038)
5
(-1-11)
0.9%
(-0.1% -1.9%)
412
(197-628)
27
(13-41)
2.5%
(1 .2% - 3.9%)
323
(97 - 536)
11
(3-17)
5.0%
(1 .5% - 8.3%)
Cutpoint***
=10 ug/m3
173
(118-228)
6
(4-8)
0.8%
(0.5% -1.1%)
99
(-99 - 293)
5
(-5-14)
0.5%
(-0.5% - 1 .6%)
308
(-38 - 647)
3
(0-7)
0.6%
(-0.1% -1.2%)
231
(110-352)
15
(7 - 23)
1.4%
(0.7% - 2.2%)
86
(26-143)
3
(1-5)
1.3%
(0.4% - 2.2%)
Cutpoint***
=15 ug/m3
82
(56-109)
3
(2-4)
0.4%
(0.3% - 0.5%)
62
(-62-184)
3
(-3 - 9)
0.3%
(-0.3% - 1 .0%)
212
(-26 - 445)
2
(0-5)
0.4%
(-0.1% -0.8%)
141
(67-215)
9
(4-14)
0.9%
(0.4% - 1 .3%)
56
(17-93)
2
(1-3)
0.9%
(0.3% - 1 .4%)
Cutpoint***
=20 ug/rn3
41
(28 - 53)
1
(1-2)
0.2%
(0.1% -0.2%)
37
(-38-110)
2
(-2 - 5)
0.2%
(-0.2% - 0.6%)
146
(-18-306)
2
(0-3)
0.3%
(0.0% - 0.6%)
83
(40-127)
5
(3-8)
0.5%
(0.2% - 0.8%)
39
(12-63)
1
(0-2)
0.6%
(0.2% - 1 .0%)
4-44
-------
Urban Area
Pittsburgh
San Jose
St. Louis
Study
Chock etal. (2000)
Fairley (2003) [reanalysis
of Fairley (1 999)]
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Type
Non-accidental
Non-accidental
Non-accidental
Ages
75+
all
all
Lag
Oday
Oday
mean of lag 0& 1
day
Incidence Associated with PM25 Assuming Various Outpoint Levels
(95% Confidence Interval)
Incidence per 100,000 General Population
(95% Confidence Interval)
Percent of Total Incidence
(95% Confidence Interval)
Policy Relevant
Background**
=2.5 or 3.5 ug/m3
77
(-166-311)
6
(-13-24)
0.8%
(-1 .7% - 3.2%)
218
(45 - 387)
13
(3 - 23)
2.6%
(0.5% - 4.7%)
233
(86 - 379)
9
(3-15)
1.1%
(0.4% - 1 .7%)
Cutpoint***
=10 ug/m3
48
(-103-193)
4
(-8-15)
0.5%
(-1.1% -2.0%)
80
(17-141)
5
(1-8)
1.0%
(0.2% - 1 .7%)
114
(42-185)
5
(2-7)
0.5%
(0.2% - 0.8%)
Cutpoint***
=15 ug/m3
31
(-67-125)
2
(-5-10)
0.3%
(-0.7% - 1 .3%)
44
(9 - 77)
3
(1-5)
0.5%
(0.1% -0.9%)
55
(20 - 89)
2
(1-4)
0.3%
(0.1% -0.4%)
Cutpoint***
=20 ug/m3
20
(-43 - 80)
2
(-3 - 6)
0.2%
(-0.4% - 0.8%)
28
(6 - 50)
2
(0-3)
0.3%
(0.1% -0.6%)
23
(8 - 38)
1
(0-1)
0.1%
(0.0% - 0.2%)
All results are for single pollutant, non-accidental mortality models, unless otherwise specified.
'Policy relevant background is 2.5 ug/m3 in the West (Los Angeles, Phoenix, and San Jose) am
"For these alternative outpoints the slope of the concentration-response relationship has been modified based on a simple hockeystick model (see discussion in section 4.3.2.1).
"Policy relevant background is 2.5 ug/m3 in the West (Los Angeles, Phoenix, and San Jose) and 3.5 ug/m3 in the East (Boston, Detroit, Philadelphia, Pittsburgh, and St. Louis).
4-45
-------
Table 4-10. Estimated Annual Health Risks of Long-Term Exposure Mortality
Associated with Recent PM2.5 Concentrations Assuming Various Outpoint Levels*
Urban Areas
Boston
Detroit
Los Angeles
Philadelphia
Phoenix
Pittsburgh
San Jose
Seattle
St. Louis
Incidence Associated with PM2 5 Assuming Various Outpoint
Levels
(95% Confidence Interval)
Incidence per 100,000 General Population
(95% Confidence Interval)
Percent of Total Incidence
(95% Confidence Interval)
Cutpoint
= 7.5 ua/m3
594
(204-1053)
21
(7 - 38)
2.7%
(0.9% - 4.7%)
906
(313-1592)
44
(15-77)
4.7%
(1 .6% - 8.2%)
3684
(1280-6426)
39
(13-68)
6.6%
(2.3% - 1 1 .4%)
650
(224-1146)
43
(15-76)
3.9%
(1 .3% - 6.9%)
349
(119-620)
11
(4 - 20)
1.7%
(0.6% - 3.0%)
816
(282-1430)
64
(22-112)
5.4%
(1 .9% - 9.4%)
172
(59 - 306)
10
(4-18)
2.1%
(0.7% - 3.6%)
50
(1 7 - 89)
3
(1-5)
0.5%
(0.2% - 0.8%)
842
(290-1486)
33
(12-59)
3.7%
(1 .3% - 6.6%)
Cutpoint**
=10 ug/m°
309
(106-551)
11
(4 - 20)
1.4%
(0.5% - 2.5%)
713
(245-1259)
35
(12-61)
3.7%
(1 .3% - 6.5%)
3267
(1132-5715)
34
(12-60)
5.8%
(2.0% -10.2%)
466
(160-825)
31
(1 1 - 54)
2.8%
(1 .0% - 4.9%)
55
(19-98)
2
(1-3)
0.3%
(0.1% -0.5%)
678
(234-1193)
53
(18-93)
4.5%
(1 .5% - 7.8%)
58
(20-104)
3
(1-6)
0.7%
(0.2% -1.2%)
0
(0-0)
0
(0-0)
0.0%
(0.0% - 0.0%)
587
(201 - 1041)
23
(8-41)
2.6%
(0.9% - 4.6%)
Cutpoint**
=12 ug/m°
20
(7 - 36)
1
(0-1)
0.1%
(0.0% - 0.2%)
519
(178-920)
25
(9 - 45)
2.7%
(0.9% - 4.8%)
2846
(984 - 4994)
30
(10-52)
5.1%
(1 .8% - 8.9%)
280
(96 - 497)
18
(6 - 33)
1.7%
(0.6% - 3.0%)
0
(0-0)
0
(0-0)
0.0%
(0.0% - 0.0%)
539
(185-951)
42
(14-74)
3.5%
(1 .2% - 6.2%)
0
(0-0)
0
(0-0)
0.0%
(0.0% - 0.0%)
0
(0-0)
0
(0-0)
0.0%
(0.0% - 0.0%)
330
(113-587)
13
(4 - 23)
1.5%
(0.5% - 2.6%)
"Based on Pope et al. (2002) — ACS extended, all cause mortality among ages 30 and older.
**For these alternative cutpoints the slope of the concentration-response relationship has been modified
based on a simple hockeystick model (see discussion in section 4.3.2.1).
4-46
-------
generate a year long series of daily PM2 5 background concentrations for specific urban areas
based on using available distributional information for the observed and background
concentrations to estimate their joint distribution, which yields the distribution of the background
concentrations conditioned on the level of the observed concentrations (see Langstaff, 2004 for
additional details describing the approach). This approach involves assigning a background
value to an observed concentration by randomly selecting a value from the conditional
distribution corresponding to the observed value. The analysis has been done both without any
correlation assumed and with a 0.4 correlation between background and observed concentrations.
To implement this approach, the mean of the background distribution is assumed to be the mid-
point estimate of PM25 background discussed in section 2.6. Estimates of the variation in
background concentrations for different regions of the United States have been obtained by an
analysis of daily data from IMPROVE sites with the sulfate component removed (Langstaff,
2005). It is important to recognize that all IMPROVE sites measure some PM25 from
anthropogenic sources, and that removing sulfate from the PM2 5 component considered does not
completely remove all anthropogenic contributions to the observed concentrations.
The sensitivity analysis examining varying daily background has been carried out in
Detroit and St. Louis using recent air quality levels for short-term exposure non-accidental
mortality associated with PM2 5. As shown in Exhibit 7.9 in the TSD, the difference between the
risk estimates based on a constant versus a varying daily background are very small in Detroit
(i.e., 0.8 percent of total incidence with varying daily background vs. 0.9 percent with assumed
constant background). The difference is even smaller in St. Louis in both the no correlation and
0.4 correlation cases, with essentially no difference in risk estimates between the constant and
varying daily background cases (see section 7.2 in the TSD).
It should be noted that the estimated distributions for background may not fully reflect
peak 24-h average natural background concentrations which can be substantially higher than the
annual or seasonal average background concentrations within areas affected by wildfires and
dust storms and long range transport from outside the United States, Canada, and Mexico (see
section 2.6). While the current PM2 5 base case risk estimates do not capture these unusual
events, it should be noted that there are provisions to exclude such events for purposes of
judging whether an area is meeting the current NAAQS (as noted above in section 2.6). The
PM2 5 risk assessment also includes a sensitivity analysis which used 2002 air quality data for
Boston to examine the impact of an extreme example (i.e., the Quebec fire episode in July 2002)
of this type of natural episodic event on short- and long-term exposure mortality (see Exhibit
7.11 in the TSD). This sensitivity analysis shows that there is hardly any difference (i.e.,
differences ranged from 0 to 0.1% of total incidence) in estimated short-term exposure mortality
4-47
-------
associated with PM2 5 including or excluding this fairly extreme, but short-term episode.22 This
same sensitivity analysis shows a difference of about 0.2% in total long-term exposure mortality
incidence associated with PM25 with and without inclusion of the Quebec fire episode days.
Alternative Concentration-Response Models
Another sensitivity analysis illustrates the impact on the risk estimates if the
concentration-response functions used for short-term exposure mortality had used distributed lag
models instead of single lag models. Schwartz (2000b) has shown in a study of short-term
exposure mortality in 10 cities using PM10 as the indicator that a distributed lag model predicted
the same relative risk that a single lag model would have predicted if the coefficient was
approximately two times what it was estimated to be. To simulate the possible impact of using a
distributed lag model, the PM2 5 coefficients were multiplied by two in this sensitivity analysis.
As would be expected, the risk estimates are almost doubled using the distributed lag
approximation (see Appendix D in TSD).
The influence of using different periods of exposure on the risks estimated in long-term
exposure mortality studies also has been examined in a sensitivity analysis. Two alternatives are
examined in the assessment: assuming the relevant PM2 5 ambient concentrations were
respectively 50% higher than and twice as high as the PM2 5 ambient concentrations used in the
original epidemiologic study. These levels have been picked by staff to give a very rough
indication of the possible impact of previous higher ambient PM2 5 levels. Assuming that the
relevant PM2 5 concentrations were 50% higher than and twice as high as the levels reported in
the original studies reduces long-term exposure mortality risk estimates by about one-third and
one-half, respectively.
As noted earlier, while few studies have reported PM2 5 concentration-response functions
using a wide variety of alternative model specifications (e.g., GAM vs. GLM, different degrees
of freedom, alternative lags), Moolgavkar (2003) did for his study in Los Angeles. Exhibits
7.12.a and 7.12.b in the TSD show the results as a sensitivity analysis for different models that
employed either the more stringent GAM approach or GLM, with either 30 or 100 degrees of
freedom, and included both single and multi-pollutant models. For this particular study, use of
GLM instead of GAM in single-pollutant models tended to either have no impact or to lower by
a small amount the estimated percent incidence of mortality in single pollutant models (e.g.,
changing the estimate from 0.9 to 0.7% of total incidence for 0-day lag with 30 degrees of
freedom). For multi-pollutant models, use of GLM instead of GAM tended to either increase
by 0.2 to 0.3% total incidence for cause-specific mortality and hospital admission estimates for
0-day lag with 100 degrees of freedom. Generally, but not always, the confidence intervals were
22This extreme episode included 2 days with PM2 5 levels above 30 ug /m3 and 1 day above 50 ug/m3.
4-48
-------
a little wider when GLM functions were used compared to GAM functions. Also, the use of a
greater number of degrees of freedom tended to reduce the estimated incidence for both mortality
and hospital admissions.
4.4.2 Just Meeting Current PM2 5 Standards
The second part of the PM2 5 risk assessment estimates the risk reductions that would
result if the current suite of PM25 standards (15 |ig/m3 annual average and 65 |ig/m3 daily
average) were just met in the assessment locations. This part of the risk assessment only
considers those locations that do not meet the current standards based on 2001-2003 air quality
data (i.e., Detroit, Philadelphia, Pittsburgh, Los Angeles, and St. Louis). As noted previously,
the 15 |ig/m3 annual average standard is the controlling standard in all five study areas.
Consequently, just meeting this standard also results in each of these areas meeting the 24-hour
standard (65 |ig/m3).
The percent rollback necessary to just meet the annual standards depends on whether the
maximum or the spatial average of the monitor-specific annual averages is used. For the risk
assessment, the approach used to simulate just meeting the current annual average standard for
the base case risk estimates used the maximum of the monitor-specific annual averages as shown
in Table 4-11. Since an area could potentially use the spatial average of the community-oriented
monitors to determine whether or not it met the annual average standard, Table 4-11 also
presents the percent rollbacks and annual average design values that would have resulted from
using this alternative approach in each urban study area which does not meet the current annual
standard and which meets the minimum criteria for use of spatial averaging. A sensitivity
analysis examining the impact of using design values based on spatial averaging is discussed in
section 4.4.3.2 for both the current and alternative PM25 annual average standards.
4.4.2.1 Base Case Risk Estimates Above Initial Cutpoint
Similar to the presentation of risk estimates in section 4.4.1 associated with recent air
quality levels, this section presents risk estimates for PM2 5 exposures after PM2 5 levels are
reduced to levels associated with just meeting the current set of standards, using the initial
cutpoint. For short-term exposure outcomes, the initial cutpoint is the estimated policy-relevant
background level. For long-term exposure mortality the initial cutpoint is 7.5 |ig/m3. Risks are
expressed both in terms of percent of total incidence and cases per hundred thousand general
population. Figure 4-8 shows the estimates for four of the risk assessment study areas that do
not meet the current PM2 5 standards when their air quality is adjusted to simulate meeting the
current standards.23 The point estimates are in the range of about 0.5 to 1 percent of total
incidence or 5 to 10 cases per hundred thousand general population across the four study areas.
23Short-term exposure non-accidental mortality estimates were not included for Philadelphia because the
concentration-response function did not include confidence limits for this endpoint.
4-49
-------
Table 4-11. Air Quality Adjustments Required to Just Meet the Current
Annual PM2 5 Standard of 15 |ng/m3 Using the Maximum vs. the
Average of Monitor-Specific Averages
Assessment
Location
Detroit
Los Angeles*
Philadelphia
Pittsburgh
St. Louis
Percent Rollback Necessary to
Just Meet the Current Annual
PM2 5 Standard
Using
Maximum of
Monitor-
Specific
Annual
Averages
28.1%
59.2%
10.9%
35.0%
17.9%
Using Average
of Monitor-
Specific Annual
Averages
11.5%
—
-0.9%
22.8%
13.5%
Design Value Based on 2001-2003
Data
Annual
Based on
Maximum
Monitor
19.5
23.6
16.4
21.2
17.5
Annual Based on
Average of
Monitor-Specific
Annual Averages
16.5
—
14.9
18.4
16.8
*Los Angeles does not meet the minimum requirements for use of spatial averaging.
Source: Abt Associates (2005b)
4-50
-------
o
LO
5-
8
c
CD
'o
s
O)
CL
Chocket al. (2000)
0-day lag
C3 TO
"O
CM T
t
-2
Detroit
Los Angeles
Pittsburgh
St. Louis
2 a -
,-, 26 -
""i 24 '
o~-
£ 22 '
•§ 20 '
fT3 . n .
• — - 1 8
o 16"
| 14 -
8- 12 -
D.
o5 8 •
<0 c
O °
O 4 -
O
°- 2 -
o
^ 0
Q-
co -4 -
S e -
CO '"
0 -8 -
i r\
CD
CD
CD
—
Ci ™
~o
-»— CO
1
CO
03
_
o
s
CT> *C
JH ^
Chock et al. (2000)
0-day lag
i * Schwartz (2003b)
^ mean of 0&1-day lag
m ' '
Cn 'J3^ ^-,
— • 'v .K>
^> ••«
I V
5 1
«* S'
3 2
.?
-------
Similarly, Figure 4-9 displays the annual risk estimates in terms of percent of total
incidence and cases per hundred thousand general population for all cause mortality associated
with long-term exposure to PM25 concentrations above a cutpoint of 7.5 |ig/m3 after air quality is
adjusted to simulate just meeting the current standards in the five areas that do not meet the
current PM2 5 standards. The point estimates generally are in the range of about 2 to 5 percent of
total incidence or 12 to 45 cases per hundred thousand general population across the five study
areas.
4.4.2.2 Base Case Risk Estimates Above Various Cutpoints
In the same manner as the risk estimates for recent air quality levels, additional base case
short- and long-term exposure annual mortality risk estimates have been developed associated
with air quality levels just meeting the current standards including both the initial cutpoint used
in the previous section and the same alternative cutpoints discussed previously. For short-term
exposure mortality, a single non-accidental mortality function has been included, except for
Philadelphia and Phoenix where cardiovascular mortality has been used since a suitable non-
accidental mortality concentration-response function is not available. For long-term exposure
mortality, a single all cause mortality concentration-response function has been included based
on the ACS-extended study (Pope et al., 2002). Tables 4-12 and 4-13 present the annual health
risks for short- and long-term exposure mortality, respectively, in terms of incidence (i.e., cases),
cases per 100,000 general population, and percent of total incidence, along with 95 percent
confidence intervals for each of these risk metrics.
4.4.2.3 Risk Estimates from Sensitivity Analyses
Three sensitivity analyses have been conducted associated with the air quality scenario
of just meeting the current PM2 5 standards. Two of these sensitivity analyses are discussed in
this section. The third one examines the impact of using a spatial average of annual average
monitor values versus the use of the maximum of annual average monitor values to determine
the design value. The design value then determines the amount of adjustment required to just
meet a specified set of standards. This third sensitivity analysis is presented in section 4.4.3.2
for both the current and alternative standards.
The first sensitivity analysis examines the impact of alternative approaches to simulating
air quality levels that just meet the current standards. The base case risk analyses use a
proportional rollback approach to adjust air quality distributions to simulate the pattern that
would occur in an area improving its air quality so that it just meets the current annual average
PM2 5 standard. The support for this approach is briefly discussed in section 4.3.1.2 and in more
detail in Appendix B of the TSD. While the available data suggest that this is a reasonable
approach, other patterns of change are possible. In a sensitivity analysis an alternative air
quality adjustment approach has been used which reduces the top 10 percent of the distribution
of PM2 5 concentrations by 1.6 times as much as the lower 90 percent of concentrations. The
4-52
-------
1 4 '
1 3 '
f ^
_ 12 ,
•*»P
S- 1 1 •
IO
-o 1 0 •
c
CD .
— 9 •
CD
O
c 8 '
CD
'o 7 •
c
j3 ®
1- 5 '
**_
O
*-< 4
C '
CD
i- 3
D.
2 '
1 •
A
O
o
o
CM,
: C/)
CO .CD
CD 'o
._ ^
w CM" CD
8 ~ 8 « 1 « _ *
< CM <^ jj^
2 1 ^ B
^
< g
1 o
o £i
o
o ~^
Q ™
a
>
"CD CD
*^« Q.
CD O
l^i
t
V)
^A
?
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
f ^
CO CM
o g a
8
o
TB .CD
"5 '15
32 o" ~T
o m £S CD S ™
'^^. rsj C/j 0° ^^ - -
o £i {5 — P
8 - < ™ Q
CM_ ™ fj CD
* o" 9 ^
2 §. la i
CD O CM CU
._ a. "T ra i
ir> "CB "St
GL a
_: a
CD C
"53 Q
._
en
CD
J
J
1
>
CO
O — -
< S
J~. °
o ~~r
CM ^
"53
03 CD
4-< Q.
CD O
35
^
CD
^
^ 1 2(
O
in
O)
-o 1 i
90
80
70
60
50
40
30
20
o
Q.
O
Q-
CD
CD
O
O
o
o
o
o
0)
Q.
(/)
CD
en
CD
O
110-
Detroit
Los Angeles
Philadelphia
Pittsburgh
St Louis
Figure 4-9. Estimated annual percent (top panel) and cases per 100,000 general
population (bottom panel) of total mortality associated with long-term
exposure to PM7«above 7.5 us/in3 (and 95 percent confidence intervals) for
air quality just meeting the current PM2 s standards.
Source: Abt Associates (2005b)
4-53
-------
Table 4-12. Estimated Annual Mortality Associated with Short-Term Exposure to PM2.5 When the Current Annual Standard
of 15 ug/m3 and the Current Daily Standard of 65 ug/m3 Are Just Met, Assuming Various Cutpoint Levels*
Urban Area
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
Study
Ito (2003) [reanalysis
of Lippmann et al.
(2000)]
Moolgavkar (2003)
[reanalysis of
Moolgavkar (2000a)l
Lipfert et al. (2000) -- 7
counties
Chock et al. (2000)
Schwartz (2003b)
[reanalysis of
Schwa rtzetal. (1996)]
Type
Non-accidental
Non-accidental
Cardiovascular
Non-accidental
Non-accidental
Ages
all
all
all
75+
all
Lag
3 day
Oday
1 day
Oday
mean of
lag 0 & 1
Incidence Associated with PM2.5 Assuming Various Cutpoint Levels
(95% Confidence Interval)
Incidence per 100,000 Population
(95% Confidence Interval)
Percent of Total Incidence
(95% Confidence Interval)
Policy Relevant Background**
=2.5 or 3.5 ug/m3
122
(-123-358)
6
(-6-17)
0.7%
(-0.7% - 1 .9%)
292
(-37-612)
3
(0-6)
0.5%
(-0.1% -1.1%)
367
(175-560)
24
(12-37)
5.8%
(2.8% - 8.8%)
50
(-108-200)
4
(-8- 16)
0.5%
(-1.1% -2.1%)
191
(70-311)
8
(3-12)
0.9%
(0.3% - 1 .4%)
Cutpoint***
=10 ug/m3
54
(-55- 159)
3
(-3 - 8)
0.3%
(-0.3% - 0.8%)
115
(-14-240)
1
(0-3)
0.2%
(0.0% - 0.4%)
189
(90 - 288)
12
(6-19)
3.0%
(1 .4% - 4.5%)
22
(-48 - 87)
2
(-4 - 7)
0.2%
(-0.5% - 0.9%)
75
(28- 122)
3
(1-5)
0.3%
(0.1% -0.6%)
Cutpoint***
=15 ug/m3
26
(-27 - 77)
1
(-1 - 4)
0.1%
(-0.1% -0.4%)
58
(-7-121)
1
(0-1)
0.1%
(0.0% - 0.2%)
106
(51 - 162)
7
(3-11)
1 .7%
(0.8% - 2.6%)
10
(-23-41)
1
(-2 - 3)
0.1%
(-0.2% - 0.4%)
29
(1 1 - 46)
1
(0-2)
0.1%
(0.1% -0.2%)
Cutpoint***
=20 ug/m3
12
(-12-35)
1
(-1 - 2)
0.1%
(-0.1% -0.2%)
29
(-4-61)
0
(0-1)
0.1%
(0.0% -0.1%)
57
(27 - 87)
4
(2-6)
0.9%
(0.4% - 1 .4%)
5
(-11 -18)
0
(-1 - 1)
0.1%
(-0.1% -0.2%)
9
(3-14)
0
(0-1)
0.0%
(0.0% -0.1%)
*AII results are for single pollutant, non-accidental mortality models, unless otherwise specified.
"Policy relevant background is 2.5 |jg/m3 in the West (Los Angeles) and 3.5 |jg/m3 in the East (Detroit, Philadelphia, Pittsburgh, and St. Louis).
***For these alternative cutpoints the slope of the concentration-response relationship has been modified based on a simple hockeystick model (see discussion in section 4.3.2.1).
4-54
-------
Table 4-13. Estimated Annual Mortality Associated with Long-Term Exposure to
PM2.s When the Current Annual Standard of 15 ug/m3and the Current Daily
Standard of 65 ug/m3 Are Just Met, Assuming Various Cutpoint Levels*
Urban Areas
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
Incidence Associated with PM25 Assuming Various Cutpoint Levels
(95% Confidence Interval)
Incidence per 100, 000 Population
(95% Confidence Interval)
Percent of Total Incidence
(95% Confidence Interval)
Cutpoint
= 7.5 ug/m3
522
(181 -910)
25
(9 - 44)
2.7%
(0.9% - 4.7%)
1507
(531 - 2587)
16
(6 - 27)
2.7%
(0.9% - 4.6%)
536
(1 85 - 943)
35
(1 2 - 62)
3.2%
(1.1% -5.7%)
403
(1 41 - 699)
31
(1 1 - 55)
2.7%
(0.9% - 4.6%)
596
(206-1047)
24
(8 - 42)
2.6%
(0.9% - 4.6%)
Cutpoint**
=10 ug/m3
282
(98 - 494)
14
(5 - 24)
1 .5%
(0.5% - 2.6%)
823
(290-1415)
9
(3-15)
1 .5%
(0.5% - 2.5%)
338
(116-597)
22
(8 - 39)
2.0%
(0.7% - 3.6%)
215
(75 - 373)
17
(6 - 29)
1 .4%
(0.5% - 2.5%)
311
(107-548)
12
(4 - 22)
1 .4%
(0.5% - 2.4%)
Cutpoint**
=12 ug/m3
41
(14-72)
2
(1 -3)
0.2%
(0.1% -0.4%)
138
(48 - 237)
1
(1 -2)
0.2%
(0.1% -0.4%)
137
(47 - 244)
9
(3-16)
0.8%
(0.3% - 1 .5%)
25
(9 - 43)
2
(1-3)
0.2%
(0.1% -0.3%)
23
(8 - 40)
1
(0-2)
0.1%
(0.0% - 0.2%)
'Based on Pope et al. (2002) — ACS extended, all cause mortality among ages 30 and older.
"For these alternative outpoints the slope of the concentration-response relatonship has been modified based on a
simple hockeystick model (see discussion in section 4.3.2.1).
4-55
-------
result of this alternative hypothetical adjustment which reduces the highest days more than the
rest of the distribution shows only a small difference (less than 1%) in the percent change in
PM-associated incidence (see Exhibit 8.9 and Appendix E, Exhibits E.33 to E.36, in the TSD).
The second sensitivity analysis explores the potential impact on the short-term exposure
non-accidental mortality risk estimates if the same multi-city concentration-response
relationship is used in five risk assessment locations compared to the single-city concentration-
response relationships used in the base case analysis. As noted earlier, the multi-city
concentration-response relationship used in this sensitivity analysis is from the Six-Cities study
(Schwartz, 2003b), the only U.S. multi-city study on PM25 short-term exposure mortality that is
currently available. Table 4-14 shows the results of this sensitivity analysis, including the
results from the base case analysis which used the single-city concentration-response
relationships. As expected, given the generally larger relative risk reported in the Six Cities
study, the estimated incidence and deaths per 100,000 general population are somewhat larger in
four of the five locations using the Six Cities study function. The range of risk estimates across
the five locations also considerably narrows when the same concentration-response relationship
is used in all five locations. For example, using the risk metric that normalizes across locations
with different population sizes, the range goes from 4 to 24 deaths per 100,000 general
population using the single-city functions to 8 to 14 deaths per 100,000 general population using
the Six Cities study function. As noted previously, there are a number of possible sources for
the differences observed in risk estimates based on single-city studies. These include known
differences in baseline mortality incidence rates, possible differences in study methodology,
increased statistical uncertainty due to smaller sample sizes in some single-city studies,
differences in degree and patterns of exposure to ambient PM2 5, differences in sources or
components that might impact the toxicity, and differences in co-pollutants or other unidentified
factors that may play a role in modifying the concentration-response relationship.
4.4.3 Just Meeting Alternative PM2 5 Standards
4.4.3.1 Base Case Risk Estimates
The third part of the PM25 risk assessment estimates the risk associated with just meeting
alternative suites of annual and daily PM25 standards, along with the risk reduction associated
with going to these levels from the current standards. For the five urban areas that exceed the
current PM2 5 suite of standards (i.e., Detroit, Los Angeles, Philadelphia, Pittsburgh, and St.
Louis), the estimated risk reductions are those associated with a further reduction in PM2 5
concentrations from just meeting the current standards to just meeting various suites of
alternative PM2 5 standards. For the four urban areas that meet the current PM2 5 standards based
on our analysis of 2001-2003 levels (i.e., Boston, Phoenix, San Jose, and Seattle), the estimated
risk reductions are those associated with a reduction from recent air quality levels to just
meeting various suites of alternative PM25 standards.
4-56
-------
Table 4-14. Sensitivity Analysis Comparing the Use of Multi-City vs. Single-City
Concentration-Response Relationships On Estimates of Short-Term Exposure
Mortality Associated with Just Meeting the Current PM2.5 Standards
Urban Area
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
Study
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Lipfert et al. (2000) -- 7
counties
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Chock etal. (2000)
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Type of
Mortality and
Single vs. Multi
City
Concentration-
Response
Relationship*
Non-accidental
Single City
Non-accidental
Multi-City
Non-accidental
Single City
Non-accidental
Multi-City
Cardiovascular
Single City
Non-accidental
Multi-City
Non-accidental
Single City
Non-accidental
Multi-City
Non-accidental
Single City
Non-accidental
Multi-City
Ages
all
all
all
all
all
all
75+
all
all
all
Lag
3 day
mean of lag 0
&1 day
Oday
mean of lag 0
&1 day
1 day
mean of lag 0
&1 day
Oday
mean of lag 0
&1 day
mean of lag 0
&1 day
mean of lag 0
&1 day
Incidence Associated with PM2.s Assuming Various Cutpoint Levels
(95% Confidence Interval)
Incidence per 100,000 General Population
(95% Confidence Interval)
Percent of Total Incidence
(95% Confidence Interval)
Policy Relevant
Background**
=2.5or3.5ug/m3
122
(-123-358)
6
(-6-17)
0.7%
(-0.7% -1.9%)
224
(160-286)
11
(8-14)
1.2%
(0.9% -1.5%)
292
(-37-612)
3
(0-6)
0.5%
(-0.1% -1.1%)
731
(526 - 935)
8
(6-10)
1 'V/
(1.0%' -1.7%)
367
(175-560)
24
(12-37)
5.8%
(2.8% - 8.8%)
213
(153-273)
14
(10-18)
1.3%
(0.9% -1.7%)
50
(-108-200)
4
(-8-16)
0.5%
(-1.1% -2.1%)
174
(125-223)
14
(10-17)
1.2%
(0.8% -1.5%)
191
(70-311)
8
(3-12)
0.9%
(0.3% -1.4%)
256
(183-328)
10
(7-13)
1.2%
(0.8% -1.5%)
Cutpoint***
=10ug/m3
54
(-55-159)
3
(-3 - 8)
0.3%
(-0.3% - 0.8%)
87
(62-111)
4
(3-5)
0.5%
(0.3% -0.6%)
115
(-14-240)
1
(0-3)
0.2%
(0.0% -0.4%)
270
(194-344)
3
(2-4)
0.5%
(0.4% -0.6%)
189
(90 - 288)
12
(6-19)
3.0%
(1.4% -4.5%)
103
(74-132)
7
(5-9)
0.6%
(0.5% -0.8%)
22
(-48 - 87)
2
(-4 - 7)
0.2%
(-0.5% - 0.9%)
68
(49 - 87)
5
(4-7)
0.5%
(0.3% -0.6%)
75
(28-122)
3
(1 -5)
0.3%
(0.1% -0.6%)
97
(69-124)
4
(3-5)
0.4%
(0.3% -0.6%)
Cutpoint***
=15ug/m3
26
(-27 - 77)
1
(-1 - 4)
0. 1 %
(-0.1% -0.4%)
34
(25 - 44)
2
(1 -2)
0.2%
(0.1% -0.2%)
58
(-7-121)
1
(0-1)
0. 1 %
(0.0% - 0.2%)
123
(89-157)
1
(1 -2)
0.2%
(0.2% - 0.3%)
106
(51 -162)
7
(3-11)
1.7%
(0.8% - 2.6%)
50
(36 - 65)
3
(2-4)
0.3%
(0.2% - 0.4%)
10
(-23-41)
1
(-2 - 3)
0. 1 %
(-0.2% -0.4%)
27
(1 9 - 34)
2
(1 -3)
0.2%
(0.1% -0.2%)
29
(1 1 - 46)
1
(0-2)
0. 1 %
(0.1% -0.2%)
36
(25 - 46)
1
(1 -2)
0.2%
(0.1% -0.2%)
Cutpoint***
=20 ug/m3
12
(-12-35)
1
(-1 - 2)
0.1%
(-0.1% -0.2%)
15
(11-19)
1
(1 -1)
0.1%
(0.1% -0.1%)
29
(-4-61)
0
(0-1)
0.1%
(0.0% -0.1%)
55
(40 - 70)
1
(0-1)
0.1%
(0.1% -0.1%)
57
(27 - 87)
4
(2-6)
0.9%
(0.4% -1.4%)
24
(17-31)
2
(1-2)
0.2%
(0.1% -0.2%)
5
(-11 -18)
0
(-1-1)
0.1%
(-0.1% -0.2%)
11
(8-14)
1
(1 -1)
0.1%
(0.1% -0.1%)
9
(3-14)
0
(0-1)
0.0%
(0.0% -0.1%)
10
(7-13)
0
(0-1)
0.1%
(0.0% -0.1%)
*AII results are for single pollutant models.
"Policy relevant background is 2.5 |jg/m3 in the West (Los Angeles) and 3.5 |jg/m3 in the East (Detroit, Philadelphia, Pittsburgh, and St. Louis).
"Tor these alternative cutpoints the slope of the concentration-response function has been modified based on a simple hockeystick model (see discussion in section
4.3.2.1).
4-57
-------
The selection of the suites of alternative annual and daily standards included in the risk
assessment has been based, in part, on consideration of CASAC and public comments, and is
consistent with the staff recommendations described in Chapter 5. Annual standards of 15, 14,
13, and 12 |ig/m3 are each combined with 98th percentile daily standards of 40, 35, 30, and
25 |ig/m3, and 99th percentile daily standards at the same levels.24 In addition, an annual
standard of 15 |ig/m3 has been combined with a 99th percentile daily standard of 65 |ig/m3. The
combinations of annual and daily alternative standards used in the PM25 risk assessment are
summarized in Table 4-15. The same proportional adjustment approach used to simulate air
quality just meeting the current standards, described previously in section 4.3.1.2 and in section
2.3 of the TSD, has been used to simulate air quality just meeting the various alternative suites
of standards. Table 4-16 provides the design values for the annual and 98th and 99th percentile
daily standards for all of the PM25 risk assessment study areas based on air quality data from
2001-2003 for the base case risk estimates.
The base case analyses examining alternative PM2 5 standards include non-accidental
mortality (or cause-specific if there was no suitable function for non-accidental mortality
available) associated with short-term exposure to PM2 5 above policy-relevant background and
several alternative cutpoint levels. In addition, the base case analyses include estimates of risk
for all cause mortality, cardiopulmonary mortality, and lung cancer mortality associated with
long-term exposure to PM2 5 above 7.5 |ig/m3 and two alternative cutpoint levels based on Pope
et al. (2002) - ACS extended. Since the patterns observed were identical, only the all cause
long-term exposure mortality results are presented in this Staff Paper (see the TSD for the cause-
specific mortality estimates). As in the earlier base case analyses, in addition to having a
cutpoint set equal to policy-relevant background, cutpoints of 10, 15, and 20 |ig/m3 have been
included for health endpoints associated with short-term exposures. For long-term exposure
mortality, cutpoints set equal to 7.5 |ig/m3, the lowest measured level in the ACS-extended
study, and alternative cutpoints of 10 and 12 |ig/m3 have been included in the base case analysis.
The base case analysis results for alternative annual standards combined with 98th and
99th percentile daily standards, respectively, are given in Tables 4-17 for Detroit for mortality
associated with short-term exposure. Short-term exposure mortality risk estimates for the other
four urban locations (Los Angeles, Philadelphia, Pittsburgh, and St. Louis) are provided in
Tables 4B-1 through 4B-4 in Appendix 4B. Similarly, the estimated risk reduction in total all
cause mortality associated with long-term PM2 5 exposures for these same alternative standards
24In four of the five urban areas that do not meet the current suite of PM2 5 standards, annual standards
within the range of 12 to 15 ug/m3 combined with the current daily standard of 65 ug/m3, using a 98th percentile
form, require the same reduction as when these annual standards are combined with a daily standard of 40 ug/m3,
using the same daily form. Therefore, the risk assessment only included the 14 ug/m3 annual standard combined
with the current daily standard for the one location (i.e., Philadelphia) and annual standard scenario where there was
a difference in the reduction required between daily standards of 40 and 65 ug/m3.
4-58
-------
Table 4-15. Alternative Sets of PM25 Standards Considered in the PM25 Risk
Assessment*
Annual
Standard
15
14
13
12
98th Percentile Daily Standard
65
x**
40
X
X
X
X
35
X
X
X
X
30
X
X
X
X
25
X
X
X
X
99th Percentile Daily Standard
65
X
40
X
X
X
X
35
X
X
X
X
30
X
X
X
X
25
X
X
X
X
*A11 standards are in \ig/m3.
**Only in Philadelphia.
Table 4-16. Estimated Design Values for Annual and 98th and 99th Percentile
Daily PM25 Standards Based on 2001-2003 Air Quality Data*
Location
Boston
Detroit
Los Angeles
Philadelphia
Phoenix
Pittsburgh
St. Louis
San Jose
Seattle
Annual
14.4
19.5
23.6
16.4
11.5
21.2
17.5
14.6
11.1
98th Percentile Daily
44
44
62
51
35
63
42
47
41
99th Percentile Daily
60
48
96
89
41
70
46
53
48
*The calculation of design values is explained in Schmidt (2005). All design values are in ng/m3. The design
values summarized here for the alternative standards are based on use of the maximum monitor in each urban area.
4-59
-------
Table 4-17. Estimated Annual Mortality Associated with Short-Term Exposure to
PM2.5 When Alternative Standards Are Just Met, Assuming Various Cutpoinl
Levels for Detroit, Ml*
Alternative Standards
Annual ((jg/m3)
15
14
Daily (|jg/m3)
65, 98th percentile value**"
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM25 Assuming Various Outpoint Levels
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
122
(-123-358)
0.0%
122
(-123-358)
0.0%
122
(-123-358)
0.0%
111
(-1 1 2 - 325)
9.0%
90
(-91 - 263)
26.2%
122
(-123-358)
0.0%
122
(-123-358)
0.0%
120
(-121 -352)
1 .6%
101
(-102-296)
17.2%
82
(-83 - 239)
32.8%
111
(-1 1 2 - 326)
9.0%
111
(-1 1 2 - 326)
9.0%
111
(-1 1 2 - 325)
9.0%
90
(-91 - 263)
26.2%
111
(-1 1 2 - 326)
9.0%
111
(-1 1 2 - 326)
9.0%
101
(-102-296)
17.2%
82
(-83 - 239)
32.8%
Cutpoint**
=10 ug/m3
54
(-55 - 1 59)
0.0%
54
(-55 - 1 59)
0.0%
54
(-55 - 1 59)
0.0%
45
(-45 -131)
16.7%
28
(-29 - 82)
48.1%
54
(-55 - 1 59)
0.0%
54
(-55 - 1 59)
0.0%
53
(-53 - 1 54)
1 .9%
37
(-37 - 1 07)
31 .5%
22
(-23 - 65)
59.3%
45
(-46 - 1 32)
16.7%
45
(-46 - 1 32)
16.7%
45
(-45 -131)
16.7%
28
(-29 - 82)
48.1%
45
(-46 - 1 32)
16.7%
45
(-46 - 1 32)
16.7%
37
(-37 - 1 07)
31 .5%
22
(-23 - 65)
59.3%
Cutpoint**
=15 ug/m3
26
(-27 - 77)
0.0%
26
(-27 - 77)
0.0%
26
(-27 - 77)
0.0%
20
(-20 - 58)
23.1%
10
(-1 0 - 28)
61 .5%
26
(-27 - 77)
0.0%
26
(-27 - 77)
0.0%
25
(-26 - 74)
3.8%
15
(-15-42)
42.3%
7
(-7-19)
73.1%
20
(-20 - 58)
23.1%
20
(-20 - 58)
23.1%
20
(-20 - 58)
23.1%
10
(-1 0 - 28)
61 .5%
20
(-20 - 58)
23.1%
20
(-20 - 58)
23.1%
15
(-15-42)
42.3%
7
(-7-19)
73.1%
Cutpoint**
=20 ug/m3
12
(-1 2 - 35)
0.0%
12
(-1 2 - 35)
0.0%
12
(-1 2 - 35)
0.0%
8
(-9 - 24)
33.3%
3
(-4-10)
75.0%
12
(-1 2 - 35)
0.0%
12
(-1 2 - 35)
0.0%
11
(-1 2 - 33)
8.3%
6
(-6-16)
50.0%
2
(-2-6)
83.3%
8
(-9 - 24)
33.3%
8
(-9 - 24)
33.3%
8
(-9 - 24)
33.3%
3
(-4-10)
75.0%
8
(-9 - 24)
33.3%
8
(-9 - 24)
33.3%
6
(-6-16)
50.0%
2
(-2-6)
83.3%
4-60
-------
Alternative Standards
Annual ((jg/m3)
13
12
Daily ((jg/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM25 Assuming Various Outpoint Levels
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 |jg/m3
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
90
(-91 - 263)
26.2%
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
82
(-83 - 239)
32.8%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 263)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
82
(-83 - 239)
32.8%
Cutpoint**
=10|jg/m3
36
(-37 - 1 06)
33.3%
36
(-37 - 1 06)
33.3%
36
(-37 - 1 06)
33.3%
28
(-29 - 82)
48.1%
36
(-37 - 1 06)
33.3%
36
(-37 - 1 06)
33.3%
36
(-37 - 1 06)
33.3%
22
(-23 - 65)
59.3%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
22
(-23 - 65)
59.3%
Cutpoint**
=15 ug/m3
14
(-15-42)
46.2%
14
(-15-42)
46.2%
14
(-15-42)
46.2%
10
(-1 0 - 28)
61 .5%
14
(-15-42)
46.2%
14
(-15-42)
46.2%
14
(-15-42)
46.2%
7
(-7-19)
73.1%
10
(-1 0 - 28)
61 .5%
10
(-1 0 - 28)
61 .5%
10
(-1 0 - 28)
61 .5%
10
(-1 0 - 28)
61 .5%
10
(-1 0 - 28)
61 .5%
10
(-1 0 - 28)
61 .5%
10
(-1 0 - 28)
61 .5%
7
(-7-19)
73.1%
Cutpoint**
=20 |jg/m3
6
(-6-16)
50.0%
6
(-6-16)
50.0%
6
(-6-16)
50.0%
3
(-4-10)
75.0%
6
(-6-16)
50.0%
6
(-6-16)
50.0%
6
(-6-16)
50.0%
2
(-2 - 6)
83.3%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
2
(-2-6)
83.3%
*This analysis was performed using Ito (2003).
**For these alternative outpoints the slope of the concentration-response function has been modified based on a simple hockeystick
model (see discussion in section 4.3.2.1).
"""Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth
4-61
-------
are given in Table 4-18 for Detroit and in Appendix 4B (see Tables 4B-5 through 4B-8) for the
other four urban areas.
Not surprisingly, estimated PM-related incidences varied substantially with both
alternative cutpoint levels and with alternative standards. In Detroit, for example, the estimated
number of cases of non-accidental mortality associated with short-term exposure to PM25 when
the current standards are just met decreases from 115 (when the cutpoint is set equal to policy-
relevant background) or to 54, 26, and 12 under alternative cutpoints of 10, 15, and 20 |ig/m3,
respectively. Because meeting increasingly lower level standards removes estimated cases at the
higher concentrations and considering higher alternative cutpoint increasingly removes
estimated cases at concentrations between background and the cutpoint, one would expect to see
an increase in the percent reduction associated with just meeting alternative standards for higher
cutpoints. This is exactly what is found. For example, as seen in Table 4-17, going from just
meeting the current standards (15 |ig/m3 annual and 65 |ig/m3 daily 98th percentile value) to just
meeting the lowest set of standards considered (12 |ig/m3 annual and 25 |ig/m3 daily 99th
percentile value) results in a reduction in short-term exposure mortality incidence of (115 -
75)/l 15 = 34.8 percent when the cutpoint equals policy-relevant background; but, with a
cutpoint equal to 10 |ig/m3, it results in a reduction of (54 - 22)/54 = 59 percent.
As shown in Table 4-18 for all-cause mortality associated with long-term exposure in
Detroit, the reduction in mortality incidence is even more dramatic when alternative cutpoint
levels are considered. Going from just meeting the current standards to just meeting the lowest
set of standards considered (12 |ig/m3 annual and 25 |ig/m3 daily 99th percentile value) results in
a reduction in long-term exposure mortality incidence of (522-207)7522= 60% with a cutpoint
equal to 7.5 |ig/m3; but, with the cutpoint set equal to 10 |ig/m3, it results in a reduction of (282 -
0)/282 =100 percent. The same general patterns can be seen in all locations and for all health
endpoints considered.
4.4.3.2 Risk Estimates from Sensitivity Analyses
Spatial Averaging Versus Maximum Community Monitor
As discussed previously in section 4.2.3.2, under the current annual PM2 5 standard urban
areas may, under certain circumstances, use the "spatial averaging approach" to determine
compliance with the annual standard. This involves using the average of the annual averages of
several monitors within the urban area. Four of the five urban areas included in the PM25 risk
assessment that do not meet the current annual standard based on the maximum community-
oriented monitor meet the minimum requirements to allow use of spatial averaging. The design
values and percent rollback required to meet the current annual standard for these four areas are
shown in Table 4-11. Tables 4B-9 and 4B-10 in Appendix 4B present the PM-related mortality
risk estimates associated with short- and long-term exposure, respectively, in Detroit using the
maximum versus the average of monitor-specific averages to determine the design value for the
annual standards. Risk estimates for alternative suites of standards are expressed in terms of
4-62
-------
Table 4-18. Estimated Annual Mortality Associated with Long-Term Exposure to PM25 When
Alternative Standards Are Just Met, Assuming Various Outpoint Levels for Detroit, Ml*
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5 Assuming Various Outpoint Levels
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
522
(181 -910)
0.0%
522
(181 -910)
0.0%
522
(181 -910)
0.0%
435
(151 -757)
16.7%
270
(94 - 468)
48.3%
522
(181 -910)
0.0%
522
(181 -910)
0.0%
507
(176-884)
2.9%
356
(124-619)
31 .8%
207
(72 - 358)
60.3%
438
(152-762)
16.1%
438
(152-762)
16.1%
435
(151 -757)
16.7%
270
(94 - 468)
48.3%
438
(152-762)
16.1%
438
(152-762)
16.1%
356
(124-619)
31 .8%
207
(72 - 358)
60.3%
Cutpoint**
=10 ug/m3
282
(98 - 494)
0.0%
282
(98 - 494)
0.0%
282
(98 - 494)
0.0%
185
(64 - 323)
34.4%
0
(0-0)
100.0%
282
(98 - 494)
0.0%
282
(98 - 494)
0.0%
266
(92 - 465)
5.7%
97
(34-168)
65.6%
0
(0-0)
100.0%
188
(65 - 328)
33.3%
188
(65 - 328)
33.3%
185
(64 - 323)
34.4%
0
(0-0)
100.0%
188
(65 - 328)
33.3%
188
(65 - 328)
33.3%
97
(34-168)
65.6%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
41
(14-72)
0.0%
41
(14-72)
0.0%
41
(14-72)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
41
(14-72)
0.0%
41
(14-72)
0.0%
23
(8-40)
43.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
4-63
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.s Assuming Various Outpoint Levels
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
354
(123-615)
32.2%
354
(123-615)
32.2%
354
(123-615)
32.2%
270
(94 - 468)
48.3%
354
(123-615)
32.2%
354
(123-615)
32.2%
354
(123-615)
32.2%
207
(72 - 358)
60.3%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
270
(94 - 468)
48.3%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
207
(72 - 358)
60.3%
Cutpoint**
=10 ug/m3
94
(33-164)
66.7%
94
(33-164)
66.7%
94
(33-164)
66.7%
0
(0-0)
100.0%
94
(33-164)
66.7%
94
(33-164)
66.7%
94
(33-164)
66.7%
0
(0-0)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-0)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
"This analysis was performed using Pope et al. (2002) - ACS extended.
"For these alternative outpoints the slope of the concentration-response function has been modified based on a simple hockeystick model (see discussion
in section 4.3.2.1).
••"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4-64
-------
estimated mortality incidence and percent reduction in incidence from just meeting the current
standards using the initial cutpoint and assuming alternative cutpoints. Similar tables for
Pittsburgh and St. Louis (the other two locations that do not meet the current standards and for
which both approaches result in positive percent rollbacks) are given in Exhibits E.37 to E.40 in
the TSD. Alternative suites of annual and daily PM25 standards, where the daily standard is the
controlling standard under both design value approaches, have not been included in this
sensitivity analysis, since there is no change in the risk estimates.
For those cases where the annual standard is the controlling standard under both design
value approaches, use of spatial averaging requires less reduction in PM2 5, thus higher mortality
incidence and less reduction in risk are associated with the current and alternative annual
standards compared to use of the maximum monitor based approach. There are also cases where
the annual standard is the controlling standard under the maximum monitor based approach, but
the daily standard becomes controlling when the same annual standard is considered using the
spatial averaging approach. When this occurs, the estimated incidence reduction associated with
the spatially averaged annual standard combined with the daily standard is determined by the
daily standard. In this case, the incidence reduction will be less than that associated with
meeting the annual standard using the maximum-monitor based approach but greater than the
incidence reduction associated with meeting the annual standard using the spatial averaging
approach.
Based on the risk estimates for the three example urban areas (Detroit, Pittsburgh, and St.
Louis) using the initial cutpoint, the estimated mortality incidence associated with long-term
exposure is about 10 to over 40% higher for the current suite of standards where compliance with
the annual standard is based on spatial averaging than the estimated incidence where compliance
is based on the highest population-oriented monitor. The estimated mortality incidence
associated with short-term exposure using the initial cutpoint ranges from about 5 to 25% higher
when the spatial averaging approach is used for the current standards in these three example
urban areas.
As noted above, the use of spatial averaging for alternative suites of standards only has an
impact on risk estimates compared to the maximum-monitor based approach where the annual
standard is controlling for at least one of these approaches. For such cases in the three example
urban areas, the estimated mortality incidence associated with long-term exposure using the
initial cutpoint in most cases ranges from about 10 to 60% higher when spatial averaging is used
to determine compliance with the annual average. In these three example urban areas, the
estimated mortality incidence associated with short-term exposure using the intial cutpoint in
most cases ranges from about 5 to 25% higher when spatial averaging is used.
Changing from a maximum-monitor based approach to the spatial average approach
impacts the estimated risks associated with just meeting both the current and lower alternative
standards. Comparing the estimated percent reductions in mortality incidence associated with
4-65
-------
going from just meeting the current standard to alternative lower standards between the two
design value approaches for the three example urban areas (Detroit, Pittsburgh, and St. Louis),
there does not seem to be any clear pattern.
4.4.4 Key Observations
Recent PM Air Quality Levels
Sections 4.4.1.1, 4.4.1.2, and 4.4.1.3 have presented the PM25 health risk estimates and
sensitivity analyses associated with recent PM air quality levels. Presented below are key
observations resulting from this part of the risk assessment:
• A fairly wide range of risk estimates are observed for PM25-related morbidity and
mortality incidence across the urban areas analyzed associated with recent air
quality.
Most of the point estimates for PM25 for the base case analysis are in the range 0.8
to 3% for short-term exposure total non-accidental mortality when the cutpoint
equals estimated policy-relevant background. Generally, the point estimates for
the single- and multi-city models are roughly similar in most of the urban areas
analyzed. The impact of adding additional co-pollutants to the models was
variable; sometimes there was relatively little difference, while in other cases
there were larger differences.
The point estimates for long-term exposure mortality associated with PM2 5 range
from about 0.5% to as high as 6.6% with most estimates falling in the 2 to 5%
range for single-pollutant models (based on the ACS-extended study) when the
cutpoint equals 7.5 |ig/m3. Addition of a single co-pollutant resulted in higher risk
estimates when CO, NO2, or O3 were added to the models for the ACS study and
lower risk estimates when SO2 was added.
• The single most important factor influencing the risk estimates is the
consideration of which of the alternative concentration-response functions
included in this assessment best represents the unknown "true" concentration-
response relationships.
The wide variability in risk estimates associated with a recent year of air quality is to be
expected given the wide range of PM25 levels across the urban areas analyzed and the variation
observed in the concentration-response relationships obtained from the original epidemiologic
studies. Among other factors, this variability may reflect differences in populations, exposure
considerations (e.g., degree of air conditioning use), differences in co-pollutants and/or other
stressors, differences in study design, and differences related to exposure and monitor
measurement error.
4-66
-------
Based on the results of the sensitivity analyses, the following key observation is made:
• The following uncertainties have a moderate impact on the risk estimates in some
or all of the cities: choice of an alternative estimated constant background level,
use of a distributed lag model, and alternative assumptions about the relevant air
quality for estimating exposure levels for long-term exposure mortality. Use of a
distribution of daily background concentrations had very little impact on the risk
estimates.
Staff was interested in obtaining insight into the overall pattern of risk associated with
short-term PM25 exposures across the distribution of PM25 air quality, as typically observed in
urban areas. Figure 4-10 illustrates the relative contribution of different portions of a typical
urban ambient PM25 concentration distribution to mortality risk associated with short-term PM25
exposures. The top panel in Figure 4-10 shows the annual distribution of 24-hour PM25
concentrations in Detroit. The middle panel shows the estimated incidence expressed in terms of
deaths per day for the upper bound of each 5 |ig/m3 increment based on the short-term exposure
epidemiology study included in the current PM2 5 risk assessment.25 The bottom panel shows the
corresponding distribution of estimated mortality incidence (for PM25) for each 5 |ig/m3
increment taking into account the number of days in each interval and the concentration-response
relationship. Not surprisingly, the middle panel shows that higher 24-hour PM2 5 concentrations
pose greater risk in terms of deaths per day. However, as shown in the bottom panel, on an
annual basis, the very highest days contribute less to the total annual health risk associated with
short-term exposures than the middle of the distribution (i.e., in the range of about 10 to 35
|ig/m3 in this example), due to the much greater number of days that occur in this part of the air
quality distribution. As shown in the prior review (61 FR at 65652, December 13, 1996), a
similar, if somewhat scaled-back pattern, was observed when concentration-response
relationships were used that assumed a cutpoint (or hypothetical threshold).
Meeting the Current PM25 Standards
Sections 4.4.2.1, 4.4.2.2, and 4.4.2.3 have presented the PM health risk estimates and
sensitivity analyses associated with just meeting the current PM25 standards. Presented below
are key observations resulting from this part of the risk assessment:
There is a wide range of PM2 5-related incidence of short-term exposure mortality
and morbidity remaining across the five urban areas analyzed. This is likely due,
in large part, to differences in concentration-response relationships among
25The Detroit PM2 s example uses the concentration-response function for non-accidental mortality from
Lippmann et al. (2000), reanalyzed in Ito (2003).
4-67
-------
Upper Bounds of 5 ug/m Intervals of Daily Average PM2.s
Concentration (ug/m3)
S1
—•— Mean Estimate of Deaths per Day
• 2.5th% Estimate of Deaths per Day
A 97.5th% Estimate of Deaths per Day
30 35 40 45 50
Upper Bounds of 5 ug/m3 Intervals of Average Daily PM2 5 Concentration (ug/m3)
mated Non-Accidental
Mortality
WCOWCOWCOWCOWCOWCOWCOWCOWCOWCO
« 1
UJ .,1
-17
-22
-27
A A
4 — *— Mean Estimate of # of Deaths
• 2.5th% Estimate of # of Deaths
A
A 97.5th% Estimate of # of
A Deaths
A
^^ *" 1
1 5 10 15 20 25 30 35 H • •< 55 t
'
m
Upper Bounds of 5 ug/m3 Intervals of Average Daily PM2.5 Concentration (ug/m3)
I
)
Figure 4-10. Distribution of average daily PMi.s concentrations in Detroit (2003 air
quality data) (top panel), estimated non-accidental mortality per day
in Detroit associated with exposure to daily PMi.s concentrations
(middle panel), and estimated non-accidental mortality in Detroit
associated with exposure to daily PM2.s concentrations over the course
of a year (bottom panel). Source: Abt Associates (2005b)
4-68
-------
single-location short-term exposure studies, differences in baseline incidence
rates, and varying population sizes.
• Results of a sensitivity analysis which applied one multi-city concentration-
response function to all five urban areas analyzed narrowed considerably the
range of risk estimates when a risk metric was used that normalized for different
population sizes. However, it is still unknown whether the wider range of
estimates observed using single-city concentration-response functions reflect
methodological differences between studies and/or real city-to-city differences
related to exposure, population, composition of the particles, or other factors.
• The single most important factor influencing the risk estimates is the
consideration of which of the alternative concentration-response functions
included in this assessment best represents the unknown "true" concentration-
response relationships.
The risk estimates associated with just meeting the current PM2 5 standards
incorporate several additional sources of uncertainty, including: (1) uncertainty in
the pattern of air quality concentration reductions that would be observed across
the distribution of PM concentrations in areas meeting the standards ("rollback
uncertainty") and (2) uncertainty concerning the degree to which current PM risk
coefficients may reflect contributions from other pollutants, or the particular
contribution of certain constituents of PM25, and whether such constituents would
be reduced in similar proportion to the reduction in PM2 5 as a whole.
At least one alternative approach to rolling back the distribution of daily PM2 5
concentrations, in which the upper end of the distribution of concentrations was
reduced by a greater amount than the rest of the distribution, had little impact on
the risk estimates.
Meeting Alternative PM25 Standards
Section 4.4.3.1 presented the base case PM25-related incidence associated with meeting
alternative PM2 5 standards and the percent reduction in incidence from the current PM2 5
standards. Presented below are key observations resulting from this part of the risk assessment:
The most important factor influencing the base case risk estimates for both short-
and long-term exposure mortality associated with PM25 concentrations just
meeting alternative standards is the consideration of which of the alternative
concentration-response functions included in this assessment best represents the
unknown "true" concentration-response relationships.
For short-term exposure mortality, there is a significant decrease in the incidence
remaining as one considers alternative higher cutpoints. There also is a
4-69
-------
significant increase observed in the percent reduction in PM-associated incidence
upon just meeting alternative standards with higher alternative cutpoints. The
reduction in incidence and increase in percent reduction in PM-associated
incidence are even more dramatic for long-term exposure mortality as higher
alternative cutpoint levels are considered.
Section 4.4.3.2 presented the results of a sensitivity analysis considering the impact on
risk estimates associated with just meeting the current and alternative standards when the spatial
averaging approach is used to determine compliance with the annual standard. A key
observation resulting from this part of the risk assessment follows:
• There is an increase in estimated short-term and long-term exposure mortality
incidence associated with PM2 5 when a spatial averaging approach is used to
determine compliance with the current annual standard or alternative suites of
standards where the daily standard is not the controlling standard.
4.5 PM10 2 5 RISK ESTIMATES
A similar approach has been taken for PM10_2 5 risk estimates, with initial base case risk
estimates for recent air quality using estimated policy-relevant background as the initial cutpoint
and, then, additional base case estimates for recent air quality and alternative PM10_2 5 standards
including the initial and alternative cutpoints. For the alternative cutpoints, the slope of the
concentration-response relationship has been modified based on the same simple hockeystick
model approach used for PM2 5.
4.5.1. Recent Air Quality
4.5.1.1 Base Case Risk Estimates
Figure 4-11 shows risk estimates for hospital admissions associated with short-term
exposure to PM10_2 5 for Detroit and Seattle, and Figure 4-12 shows risk estimates associated with
respiratory symptoms for St. Louis associated with recent PM10_2 5 air quality levels. For Detroit
risk estimates are provided for several categories of cardiovascular and respiratory-related
hospital admissions and show point estimates ranging from about 2 to 7% of cause-specific
admissions being associated with as is short-term exposures to PM10_2 5. The point estimate for
asthma hospital admissions associated with PM10_2 5 exposures for Seattle, an area with lower
PM10_2 5 ambient concentrations than either Detroit or St. Louis, is about 1%. Point estimates for
lower respiratory symptoms and cough in St. Louis are about 12 and 15%, respectively. These
estimates use estimated policy-relevant background as the cutpoint. Table 4-21, discussed
below, provides risk estimates associated with recent PM10_2 5 air quality levels using policy-
relevant background and higher alternative cutpoints.
4-70
-------
4.5.1.2 Risk Estimates from Sensitivity Analyses
For PM10_2 5, the sensitivity analysis examining the effects of using the lower- and upper-
end of the range of estimated policy-relevant background levels shows about a 16% increase in
the risk estimates for various respiratory and cardiovascular-related short-term exposure hospital
admissions in Detroit between the base case (which used a value of 4.5 |ig/m3 for background)
and the lower end where background was estimated to be 1 |ig/m3. At the upper end, where
background was estimated to be 9 |ig/m3, the short-term exposure hospital admission risk
estimates are reduced by about 19% (see Exhibit 9.5 in the TSD). The effect of different
background concentrations for the other two PM10_2 5 locations is about + 30% for asthma
hospital admissions in Seattle (see Exhibit F.7 in the TSD) and about + 50% for respiratory
symptoms in St. Louis (see Exhibit F.8 in the TSD).
4.5.2 Just Meeting Alternative PM10_2 5 Standards
The second part of the PM10_25 risk assessment estimates the risk associated with just
meeting alternative daily PM10_2 5 standards for the three locations examined earlier (Detroit, St.
Louis, and Seattle), as well as the risk reductions associated with going to these levels from the
current air quality levels. Staff notes that the locations used in this part of the risk assessment are
not representative of urban locations in the U.S. that experience the most significant elevated 24-
hour PM10_2 5 ambient concentrations. Thus, observations regarding risk reductions associated
with alternative standards in these three urban areas may not be fully relevant to the areas
expected to have the greatest health risks associated with peak daily ambient PM10_2 5
concentrations.
Estimated reductions in risk were developed for going from recent air quality levels
(based on 2003 air quality) to just meeting alternative PM10_2 5 standards. Staff selected the daily
standards to be included in the risk assessment based on the preliminary staff recommendations
described in Chapter 5 of the draft 2005 Staff Paper (EPA, 2005) and consideration of public and
CAS AC comments. Table 4-19 summarizes the sets of 98th and 99th percentile daily standards
that were included in the PM10_2 5 risk assessment. The estimated design values which were used
to determine the air quality adjustment to be used in simulating just meeting alternative PM10_25
standards are shown in Table 4-20.
The estimated number of hospital admissions for ischemic heart disease associated with
short-term PM10.2 5 exposures for alternative 98th and 99th percentile daily standards, respectively,
are given in Table 4-21 for Detroit. This table includes risk estimates which are based on the
cutpoint being policy-relevant background as well as three higher alternative cutpoints. Daily
PM10.25 standards set at 80 (for 98th percentile form) and 100 or 80 (for 99th percentile form)
result in no reduction in risk in Detroit. The reason why no estimated risk reductions are
observed with these alternative standards is that the percent reduction of PM10_25 concentrations
at the composite monitor to just meet a standard is determined by comparing the alternative
4-71
-------
o
^
m
o>
73
C
-2-
8
c
0)
1 9 '
1 7 -
1 3
1 1
9
1
c
r 3
ro
o
Q.
ag
s S |
8 »
Detroit
Seattle
70
60
50
O
10
01
-a
c
o
oj 40
Q.
o
Q. 30
0)
O
o
o
o
o"
o
CD
Q.
tn
0)
tn
O
20
1 0
-10
Ito {2003} -
ischemic heart disease
hospital admissions
m
0 -p! us
ti
-L=l
n TO 's
CD as £
O -C T3
Ci a? ^
O -s -^
~ |t
-20 '
Detroit
Seattle
Figure 4-11. Estimated annual percent (top panel) and cases per 100,000 general
population (bottom panel) of hospital admissions associated with short-term
exposure to PM10_2 5 above background for recent air quality (and 95 percent
confidence intervals). Source: Abt Associates (2005b)
4-72
-------
o
^
in
CD
73
cc
-------
Table 4-19. Alternative PM10_25 Standards Considered in the PM10_25
Risk Assessment*
Daily
Standards Based on the 98th Percentile
Value
80
65
50
30
25
Daily Standards Based on the 99th Percentile
Value
100
80
60
35
30
*A11 standards are in ug/m3.
Table 4-20. Estimated Design Values for 98th and 99th Percentile Daily PM
Standards Based on 2001-2003 Air Quality Data*
10-2.5
Location
Detroit
St. Louis
Seattle
98th Percentile Daily
70
33
31
99th Percentile Daily
77
47
39
*The calculation of design values is explained in Schmidt (2005). All design values are in ug/m3.
4-74
-------
Table 4-21. Estimated Annual Hospital Admissions for Ischemic Heart Disease Associated with Short-Term
Exposure to PM10_25 When Alternative Standards Are Just Met, Assuming Various Outpoint Levels*
Detroit, Ml, 2003
(Recent Air Quality Levels = 21.7 ug/m3 Annual Average; 105.9 ug/m3, 98th Percentile Daily Value)
Recent PM10.25 Air Quality Levels and
Alternative Daily Standards (ug/m3)
Recent PM10_2.5 air quality levels
80 |jg/m3 daily 98th percentile value
65 |jg/m3 daily 98th percentile value
50 |jg/m3 daily 98th percentile value
30 |jg/m3 daily 98th percentile value
25 |jg/m3 daily 98th percentile value
100 |jg/m3 daily 99th percentile value
80 |jg/m3 daily 99th percentile value
60 |jg/m3 daily 99th percentile value
35 |jg/m3 daily 99th percentile value
30 |jg/m3 daily 99th percentile value
Incidence Associated with PM10_25 Assuming Various Outpoint Levels
(95% Confidence Interval)
Percent Reduction in Incidence from Recent PM10.2.s Air Quality Levels
Policy Relevant
Background
=4.5 ug/m3
654
(169-1083)
0.0%
654
(169-1083)
0.0%
600
(156-989)
8.3%
443
(117-719)
32.3%
242
(65 - 386)
63.0%
193
(52 - 307)
70.5%
654
(169- 1083)
0.0%
654
(169-1083)
0.0%
491
(129-801)
24.9%
262
(70-419)
59.9%
218
(59 - 347)
66.7%
Outpoint**
=10 ug/m3
569
(149-934)
0.0%
569
(149-934)
0.0%
508
(134-829)
10.7%
334
(90 - 532)
41 .3%
125
(36- 190)
78.0%
81
(24-120)
85.8%
569
(149-934)
0.0%
569
(149-934)
0.0%
387
(104-621)
32.0%
144
(41 -221)
74.7%
103
(30- 154)
81 .9%
Outpoint**
=15 ug/m3
489
(129-794)
0.0%
489
(129-794)
0.0%
425
(114-683)
13.1%
248
(69 - 384)
49.3%
65
(20-91)
86.7%
39
(13-52)
92.0%
489
(129-794)
0.0%
489
(129-794)
0.0%
301
(83 - 472)
38.4%
79
(24-113)
83.8%
51
(16-70)
89.6%
Outpoint**
=20 ug/m3
426
(115-682)
0.0%
426
(115-682)
0.0%
360
(99 - 567)
15.5%
183
(54-271)
57.0%
44
(15-57)
89.7%
25
(9 - 30)
94.1%
426
(115-682)
0.0%
426
(115-682)
0.0%
233
(67 - 353)
45.3%
53
(18-68)
87.6%
34
(12-43)
92.0%
"This analysis was performed using Ito (2003).
"For these alternative cutpoints the slope of the concentration-response function has been modified based on a simple hockeystick model (see
discussion in section 4.3.2.1).
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4-75
-------
standard level with the design value for that location based on 2001-2003 air quality data. In
Detroit, the design value for the 98th percentile daily PM10_2 5 standards is 70 |ig/m3 whereas the
98th percentile daily value in 2003 is 105.9 |ig/m3. Because the design value is lower than
80 |ig/m3, the highest 98th percentile daily PM10_2 5 standard considered in the assessment, zero
risk reductions were estimated to result from this standard, even though the 98th percentile daily
value at the composite monitor in 2003, 105.9 |ig/m3, is well above the standard level. Similarly,
the design value for the 99th percentile daily PM10_2 5 standards is 77 |ig/m3 for Detroit, whereas
the 99th percentile daily value at the composite monitor in Detroit in 2003 is substantially greater
than 100 |ig/m3, the highest 99th percentile daily PM10_25 standard considered. Thus, zero risk
reductions were estimated to result from both 100 and 80 |ig/m3 standards. In general, estimated
risk reductions increase and the confidence intervals around the estimates widen as lower daily
standards are considered.
As expected, the maximum reduction in risk, for the set of alternative standards included
in the analysis, is achieved with the 98th percentile 25 |ig/m3 standard and 99th percentile
30 |ig/m3 standard. The point estimate is that about a 4% reduction in hospital admissions for
ischemic reductions associated with just meeting daily 98th percentile PM10_25 standards of
80 |ig/m3 in Detroit, and 80, 65, and 50 |ig/m3 in St. Louis or Seattle. Similarly, there are no risk
reductions associated with just meeting daily 99th percentile PM10_25 standards of 100 or 80 |ig/m3
in Detroit, and 100, 80, or 60 |ig/m3 in St. Louis or Seattle.
4.5.3 Key Observations
Sections 4.5.1.1 and 4.5.1.2 presented the PM10_25 health risk estimates and sensitivity
analyses associated with recent PM10_2 5 air quality levels. Presented below are key observations
resulting from this part of the risk assessment:
• Various respiratory and cardiovascular cause-specific hospital admission point
estimates associated with short-term exposure to PM10_2 5 range from 1 to 7%,
depending on location and type of admission. Point estimates for lower
respiratory symptoms and cough were about 12 and 15% of total incidence for
recent air quality levels in a single urban area (St. Louis)
Results of a sensitivity analysis examining the impact of assuming different
values for policy relevant background showed moderate changes in short-term
morbidity risk estimates ranging from + 16 to +50% depending on the health
endpoint and location considered.
Section 4.5.2 presented the base case PM10_25-related incidence associated with meeting
alternative PM2 5 standards and the percent reduction in incidence from recent air quality levels.
Presented below are key observations resulting from this part of the risk assessment:
4-76
-------
For short-term exposure morbidity, there is a significant decrease in the remaining
estimated incidence associated with PM10_2 5 as one considers higher alternative
cutpoints for all of the standards that require reductions in recent PM10_25 air
quality levels. There also is a significant increase observed in the percent
reduction in PM-associated incidence upon just meeting these same alternative
standards with higher alternative cutpoints compared to recent PM10_2 5 air quality
levels.
Based on the point estimates, there are no risk reductions associated with just
meeting daily 98th percentile PM10_25 standards of 80 |ig/m3 in Detroit, and 80, 65,
and 50 |ig/m3 in St. Louis or Seattle. Similarly, there are no risk reductions
associated with just meeting daily 99th percentile PM10_25 standards of 100 or
80 |ig/m3 in Detroit, and 100, 80, or 60 |ig/m3 in St. Louis or Seattle.
4-77
-------
REFERENCES
Most Chapter 4 references are available at the end of Chapter 3. References not listed at the end
of Chapter 3 are listed here.
Abt Associates Inc. (1996). "A Paniculate Matter Risk Assessment for Philadelphia and Los Angeles." Bethesda,
MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, Contract No. 68-W4-0029. July 3 (revised November). Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm pr td.html.
Abt Associates Inc. (1997a). Abt Associates Memorandum to U.S. EPA. Subject: Revision of Mortality Incidence
Estimates Based on Pope et al. (1995) in the Abt Paniculate Matter Risk Assessment Report. June 5,
1997.
Abt Associates Inc. (1997b). Abt Associates Memorandum to U.S. EPA. Subject: Revision of Mortality Incidence
Estimates Based on Pope et al. (1995) in the December 1996 Supplement to the Abt Paniculate Matter
Risk Assessment Report. June 6, 1997.
Abt Associates Inc. (2002). Proposed Methodology for Paniculate Matter Risk Analyses for Selected Urban Areas:
Draft Report. Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Contract No. 68-D-03-002. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html.
Abt Associates Inc. (2003a). Abt Associates Memorandum to U.S. EPA. Subject: Preliminary Recommended
Methodology for PM10 and PM10_2 5 Risk Analyses in Light of Reanalyzed Study Results. April 8, 2003.
Available: http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html.
Abt Associates Inc. (2003b). Paniculate Matter Health Risk Assessment for Selected Urban Areas: Draft Report.
Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Contract No. 68-D-03-002. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html.
Abt Associates Inc. (2005a). Paniculate Matter Health Risk Assessment for Selected Urban Areas. Draft Report.
Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Contract No. 68-D-03-002. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html.
Abt Associates Inc. (2005b). Paniculate Matter Health Risk Assessment for Selected Urban Areas. Draft Report.
Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Contract No. 68-D-03-002. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm cr td.html.
Center for Disease Control (2001). CDC Wonder. Available: http://wonder.cdc.gov/.
Deck, L. B.; Post, E.S.; Smith, E.; Wiener, M.; Cunningham, K.; Richmond, H. (2001). Estimates of the health risk
reductions associated with attainment of alternative paniculate matter standards in two U.S. cities. Risk
Anal. 21(5): 821-835.
Environmental Protection Agency (2001). Paniculate Matter NAAQS Risk Analysis Scoping Plan, Draft. Research
Triangle Park, NC: Office of Air Quality Planning and Standards. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html.
4-78
-------
Hopke, P. (2002). Letter from Dr. Phil Hopke, Chair, Clean Air Scientific Advisory Committee (CASAC) to
Honorable Christine Todd Whitman, Administrator, U.S. EPA. Final advisory review report by the
CASAC Paniculate Matter Review Panel on the proposed paniculate matter risk assessment. May 23,
2002. Available: http://www.epa.gov/sab/pdf/casacadv02002.pdf.
Langstaff, J. (2004). OAQPS Staff Memorandum to PM NAAQS Review Docket (OAR-2001-0017). Subject: A
Methodology for Incorporating Short-termVariable Background Concentrations in Risk Assessments.
December 17, 2004. Available: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr td.html.
Langstaff, J. (2005). OAQPS Staff Memorandum to PM NAAQS Review Docket (OAR-2001-0017). Subject:
Estimation of Policy-Relevant Background Concentrations of Paniculate Matter. January 27, 2005.
Available: http://www.epa.gov/ttn/naaqs/standards/pm/sjm cr td.html.
National Academy of Sciences (2002). Estimating the Public Health Benefits of Proposed Air Pollution Regulations.
Washington, D.C.: The National Academy Press. Available:
http://www.nap.edu/books/0309086094/html/.
Post, E.; Deck, L.; Larntz, K.; Hoaglin. D. (2001). An application of an empirical Bayes estimation technique to the
estimation of mortality related to short-term exposure to paniculate matter. Risk Anal. 21(5): 837-842.
Schmidt, M.: Mintz, D.; Rao, V.; McCluney, L. (2005). U.S. EPA Memorandum to File. Subject: Draft Analyses
of 2001-2003 PM Data for the PM NAAQS Review. January 31, 2005. Available:
http://www.epa.gov/oar/oaqps/pm25/docs.html.
Science Advisory Board (2004). Advisory on Plans for Health Effects Analysis in the Analytical Plan for EPA's
Second Prospective Analysis - Benefits and Costs of the Clean Air Act, 1990-2000. Advisory by the
Health Effects Subcommittee of the Advisory Council for Clean Air Compliance Analysis. EPA SAB
Council - ADV-04-002. March. Available: http://www.epa.gov/sciencel/pdf/council adv 04002.pdf.
4-79
-------
-------
5. STAFF CONCLUSIONS AND RECOMMENDATIONS ON PRIMARY PM NAAQS
5.1 INTRODUCTION
This chapter presents staff conclusions and recommendations for the Administrator to
consider in deciding whether the existing primary PM standards should be revised and, if so,
what revised standards are appropriate. The existing suite of primary PM standards includes
annual and 24-hour PM2 5 standards, to protect public health from exposure to fine particles, and
annual and 24-hour PM10 standards, to protect public health from exposure to thoracic coarse
particles. Each of these standards is defined in terms of four basic elements: indicator,
averaging time, level and form. Staff conclusions and recommendations on these standards are
based on the assessment and integrative synthesis of information presented in the CD and on
staff analyses and evaluations presented in Chapters 2 through 4 herein.
In recommending a range of primary standard options for the Administrator to consider,
staff notes that the final decision is largely a public health policy judgment. A final decision
must draw upon scientific information and analyses about health effects and risks, as well as
judgments about how to deal with the range of uncertainties that are inherent in the scientific
evidence and analyses. The staffs approach to informing these judgments, discussed more fully
below, is based on a recognition that the available health effects evidence generally 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. 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 primary standards that, in the
Administrator's judgment, are requisite to protect public health with an adequate margin of
safety. 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 primary standards be set
at a zero-risk level, but rather at a level that avoids unacceptable risks to public health.
5.2 APPROACH
As an initial matter, PM standards for fine particles and for thoracic coarse particles are
addressed separately, consistent with the decision made by EPA in the last review and with the
conclusion in the CD that fine and thoracic coarse particles should continue to be considered as
separate subclasses of PM pollution. As discussed in Chapter 3, section 3.2.3, this conclusion is
based in part on long-established information on the differences in sources, properties, and
atmospheric behavior between fine and coarse particles; and it is reinforced by new information
that advances our understanding of differences in human exposure relationships and dosimetric
patterns characteristic of these two subclasses of PM pollution, as well as the apparent
independence of health effects that have been associated with them in epidemiologic studies.
5-1
-------
In general, in evaluating whether the current primary standards are adequate or whether
revisions are appropriate, and in developing recommendations on the elements of possible
alternative standards for consideration, staffs approach in this review builds upon and broadens
the general approach used by EPA in the last review. In setting PM25 standards in 1997, the
Agency mainly used an evidence-based approach that placed primary emphasis on epidemiologic
evidence from short-term exposure studies of fine particles, judged to be the strongest evidence
at that time, in reaching decisions to set an annual PM2 5 standard that was generally controlling,
and to set a 24-hour PM2 5 standard to provide supplemental protection. The risk assessment
conducted in the last review provided qualitative insights, but was judged to be too limited to
serve as a quantitative basis for decisions on the standards. In this review, the more extensive
and stronger body of evidence now available on health effects related to both short- and long-
term exposure to PM2 5, together with the availability of much more extensive PM2 5 air quality
data, have facilitated a more comprehensive risk assessment for PM25. As a result, staff has used
a broader approach in this review of the PM25 standards that takes into account both evidence-
based and quantitative risk-based considerations, placing greater emphasis on evidence from
long-term exposure studies and quantitative risk assessment results for fine particles than was
done in the last review. Staff has applied this approach to a more limited degree in reviewing the
PM10 standards, reflecting the far more limited nature of the health effects evidence and air
quality data available for thoracic coarse particles.
In reviewing the PM25 standards, for example, staff has taken into account evidence-
based considerations primarily by assessing the epidemiologic evidence of associations with
health endpoints that the CD has judged to be likely causal based on an integrative synthesis of
the entire body of evidence. Less weight is given to evidence of associations that are judged to
be only suggestive of possible causal relationships, taking this information into account as part
of margin of safety considerations. In so doing, staff has placed greater weight on U.S. and
Canadian studies reporting statistically significant associations, providing relatively more precise
effects estimates, using relatively more reliable air quality data, and reporting associations that
are generally robust to alternative model specifications and the inclusion of potentially
confounding co-pollutants. By considering the ambient particle levels present during specific
studies, staff has reached conclusions as to the degree to which alternative standards could be
expected to protect against the observed health effects, while being mindful of the inherent
limitations and uncertainties in such evidence.
Staff has also taken into account quantitative risk-based considerations, drawn from the
results of the risk assessment conducted in several example urban areas (discussed in Chapter 4).
More specifically, staff has considered estimates of the magnitude of PM-related risks associated
with current air quality levels, as well as the risk reductions likely to be associated with attaining
the current or alternative standards. In so doing, staff recognizes the considerable uncertainties
inherent in such risk estimates, and has taken such uncertainties into account by considering the
5-2
-------
sensitivity of the risk estimates to alternative assumptions likely to have substantial impact on
the estimates.
More specifically, in this review a series of questions frames staffs approach to reaching
conclusions and recommendations, based on the available evidence and information, as to
whether consideration should be given to retaining or revising the current primary PM standards.
Staffs review of the adequacy of the current standards begins by considering whether the
currently available body of evidence assessed in the CD suggests that revision of any of the basic
elements of the standards would be appropriate. This evaluation of the adequacy of the current
standards involves addressing questions such as the following:
• To what extent does newly available information reinforce or call into question
evidence of associations with effects identified in the last review?
• To what extent does newly available information reinforce or call into question
any of the basic elements of the current standards?
To what extent have important uncertainties identified in the last review been
reduced and have new uncertainties emerged?
To the extent that the evidence suggests that revision of the current standards would be
appropriate, staff then considers whether the currently available body of evidence supports
consideration of standards that are either more or less protective by addressing the following
questions:
Is there evidence that associations, especially likely causal associations, extend to
air quality levels that are as low as or lower than had previously been observed,
and what are the important uncertainties associated with that evidence?
• Are health risks estimated to occur in areas that meet the current standards; are
they important from a public health perspective; and what are the important
uncertainties associated with the estimated risks?
To the extent that there is support for consideration of revised standards, staff then identifies
ranges of standards (in terms of indicators, averaging times, levels and forms) that would reflect
a range of alternative public health policy judgments, based on the currently available evidence,
as to the degree of protection that is requisite to protect public health with an adequate margin of
safety. In so doing, staff addresses the following questions:
• Does the evidence provide support for considering different PM indicators?
• Does the evidence provide support for considering different averaging times?
What ranges of levels and forms of alternative standards are supported by the
evidence, and what are the uncertainties and limitations in that evidence?
5O
-------
• To what extent do specific levels and forms of alternative standards reduce the
estimated risks attributable to PM, and what are the uncertainties in the estimated
risk reductions?
Based on the evidence, estimated risk reductions, and related uncertainties, staff makes
recommendations as to ranges of alternative standards for the Administrator's consideration in
reaching decisions as to whether to retain or revise the primary PM NAAQS.
Standards for fine particles are addressed in section 5.3 below, beginning with staffs
consideration of the adequacy of the current primary PM25 standards. Subsequent subsections
address each of the major elements that define specific PM standards: pollutant indicator,
averaging time, level and form. Staff has evaluated separately the protection that a suite of PM25
standards would likely provide against effects associated with long-term exposures (section
5.3.4) and those associated with short-term exposures (section 5.3.5). These separate evaluations
provide the basis for integrated recommendations on alternative suites of standards that would
protect against effects associated with both long- and short-term exposures, based on considering
how a suite of standards would operate together to protect public health. In a similar manner,
standards for thoracic coarse particles are addressed in section 5.4 below. This chapter
concludes with a summary of key uncertainties associated with establishing primary PM
standards and with related staff research recommendations in section 5.5.
5.3 FINE PARTICLE STANDARDS
5.3.1 Adequacy of Current PM2 5 Standards
In considering the adequacy of the current PM25 standards, staff has first considered the
extent to which newly available information reinforces or calls into question evidence of
associations with effects identified in the last review, as well as the extent to which important
uncertainties have been reduced or have resurfaced as being more important than previously
understood. In looking across the extensive epidemiologic evidence available in this review, the
CD addresses these questions by concluding that "the available findings demonstrate well that
human health outcomes are associated with ambient PM" (CD, p. 9-24) and, more specifically,
that there is now "strong epidemiological evidence" for PM25 linking short-term exposures with
cardiovascular and respiratory mortality and morbidity, and long-term exposures with
cardiovascular and lung cancer mortality and respiratory morbidity (CD, p. 9-46). This latter
conclusion reflects greater strength in the epidemiologic evidence specifically linking PM2 5 and
various health endpoints than was observed in the last review, when the 1996 CD concluded that
the epidemiologic evidence for PM-related effects was "fairly strong," noting that the studies
"nonetheless provide ample reason to be concerned" about health effects attributable to PM at
levels below the then-current PM NAAQS (EPA, 1996, p. 13-92).
5-4
-------
As discussed in Chapter 3 (section 3.5) and the CD (section 9.2.2), the CD concludes that
the extensive body of epidemiologic evidence now available continues to support likely causal
associations between PM2 5 and the above health outcomes based on an assessment of the
strength of the evidence, including the strength and robustness of reported associations and the
consistency of the results. The CD recognizes that while the relative risk estimates are generally
small in magnitude, a number of new studies provide relatively precise estimates that are
generally positive and often statistically significant. Overall, the CD finds that the new evidence
substantiates that the associations are generally robust to confounding by co-pollutants, noting
that much progress has been made in sorting out contributions to observed health effects of PM
and its components relative to other co-pollutants. On the other hand, the CD notes that effect
estimates are generally more sensitive than previously recognized to different modeling
strategies to adjust for temporal trends and weather variables. While some studies showed little
sensitivity, different modeling strategies altered conclusions in other studies.
Although greater variability in effects estimates across study locations is seen in the
much larger set of studies now available, especially in the new multi-city studies, the CD finds
much consistency in the epidemiologic evidence, particularly in studies with the most precision.
There also are persuasive reasons why variation in associations in different locations could be
expected. Further, the CD concludes that new source apportionment studies and "found
experiments," showing improvements in community health resulting from reductions in PM and
other air pollutants, lend additional support to the results of other studies that focused
specifically on PM2 5.
Looking more broadly to integrate epidemiologic evidence with that from exposure-
related, dosimetric and toxicologic studies, the CD (section 9.2.3) considered the coherence of
the evidence and the extent to which the new evidence provides insights into mechanisms by
which PM, especially fine particles, may be affecting human health. Progress made in gaining
insights into mechanisms lends support to the biologic plausibility of results observed in
epidemiologic studies. For cardiovascular effects, the CD finds that the convergence of
important new epidemiologic and toxicologic evidence builds support for the plausibility of
associations especially between fine particles and physiological endpoints indicative of increased
risk of cardiovascular disease and changes in cardiac rhythm. This finding is supported by new
cardiovascular effects research focused on fine particles that has notably advanced our
understanding of potential mechanisms by which PM25 exposure, especially in susceptible
individuals, could result in changes in cardiac function or blood parameters that are risk factors
for cardiovascular disease. For respiratory effects, the CD finds that toxicologic studies have
provided evidence that supports plausible biologic pathways for fine particles, including
inflammatory responses, increased airway responsiveness, or altered responses to infectious
agents. Further, the CD finds coherence across a broad range of cardiovascular and respiratory
health outcomes from epidemiologic and toxicologic studies done in the same location,
5-5
-------
particularly noting, for example, the series of studies conducted in or evaluating ambient PM
from Boston and the Utah Valley. The CD also finds that toxicologic evidence examining
combustion-related particles supports the plausibility of the observed relationship between fine
particles and lung cancer mortality. With regard to PM-related infant mortality and
developmental effects, the CD finds this to be an emerging area of concern, but notes that current
information is still very limited in support of the plausibility of potential ambient PM
relationships.
Based on the above considerations and findings from the CD, staff concludes that the
newly available information generally reinforces the associations between PM2 5 and mortality
and morbidity effects observed in the last review. Staff recognizes that important uncertainties
and research questions remain, notably including questions regarding modeling strategies to
adjust for temporal trends and weather variables in time-series epidemiologic studies.
Nonetheless, staff notes that progress 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, alone and in combination with other pollutants, is causally
linked with cardiovascular, respiratory, and lung cancer associations observed in epidemiologic
studies. Thus, staff finds clear support in the available evidence, as assessed in the CD, for fine
particle standards that are at least as protective as the current PM2 5 standards.
Having reached this initial conclusion, staff also has addressed the question of whether
the available evidence supports consideration of standards that are more protective than the
current PM25 standards. In so doing, staff has considered first whether there is evidence that
health effects associations with short- and long-term exposures to fine particles extend to lower
air quality levels than had previously been observed, or to levels below the current standards. In
addressing this question, staff first recognizes that there are likely biologic threshold levels in
individuals for specific health responses. Staff notes, however, that the available epidemiologic
evidence neither supports nor refutes the existence of thresholds at the population level for the
effects of PM25 on mortality across the range of concentrations in the studies, for either long-
term or short-term PM25 exposures, as discussed in Chapter 3 (section 3.6.6) and the CD (section
9.2.2.5). Further, the CD notes that in the multi-city studies and most single-city studies,
statistical tests comparing linear and various nonlinear or threshold models have not shown
statistically significant distinctions between them (CD, p. 9-44). Even in those few studies with
suggestive evidence for thresholds, the potential thresholds are at fairly low concentrations (CD,
p. 9-45). While acknowledging that for some health endpoints, such as total nonaccidental
mortality, it is likely to be extremely difficult to detect thresholds, the CD concludes that
"epidemiologic studies suggest no evidence for clear thresholds in PM-mortality relationships
within the range of ambient PM concentrations observed in these studies." (CD, p. 9-48).
5-6
-------
5.3.1.1 Evidence-based Considerations
In considering the available epidemiologic evidence (summarized in Chapter 3, section
3.3 and Appendices 3 A and 3B), staff has focused on specific epidemiologic studies that show
statistically significant associations between PM25 and health effects for which the CD judges
associations with PM25 to be likely causal. Many more U.S. and Canadian studies are now
available that provide evidence of associations between PM2 5 and serious health effects in areas
with air quality at and above the level of the current annual PM25 standard (15 |ig/m3), which
was set to provide protection against health effects related to both short- and long-term
exposures to fine particles. Notably, a few of the newly available short-term exposure mortality
studies provide evidence of statistically significant associations with PM25 in areas with long-
term average air quality below the level of the current annual PM25 standard (summarized in
Appendix 3 A). In considering these studies, staff has focused on those studies that include
adequate gravimetric PM2 5 mass measurements, and where the associations are generally robust
to alternative model specification and to the inclusion of potentially confounding co-pollutants.
Three such studies conducted in Phoenix (Mar et al., 1999, 2003), Santa Clara County, CA
(Fairley, 1999, 2003) and eight Canadian cities (Burnett et al., 2000; Burnett and Goldberg,
2003) report statistically significant associations between short-term PM2 5 exposure and total
and cardiovascular mortality in areas in which long-term average PM2 5 concentrations ranged
between 13 and 14 |ig/m3. These studies were reanalyzed to address questions about the use of
GAM with default convergence criteria, and the study results from Phoenix and Santa Clara
County were little changed in alternative models (Mar et al., 2003; Fairley, 2003), although
Burnett and Goldberg (2003) reported that their results were sensitive to using different temporal
smoothing methods.
Beyond these mortality studies, other studies reported statistically significant associations
between short-term PM25 exposure and morbidity in such areas. Three studies of emergency
department visits were conducted in areas where the mean PM2 5 concentrations were
approximately 12 |ig/m3 or below, although these studies either had not been reanalyzed to
address the default convergence criteria problem with GAM, did not assess the potential for
confounding by co-pollutants, were not robust to the inclusion of co-pollutants, or were done
only during a single season. Another new study reported statistically significant associations
with incidence of myocardial infarction where the mean PM25 concentration was just above
12 |ig/m3; however, the CD urges caution in interpreting the results of the new body of evidence
related to such cardiovascular effects (CD, p. 8-166). Thus, these studies provide no clear
evidence of statistically significant associations with PM25 at such low concentrations.
New evidence is also available from U.S. and Canadian studies of long-term exposure to
fine particles (summarized in Appendix 3B). In evaluating this evidence (CD, section 9.2.3), the
CD notes that new studies have built upon studies available in the last review and that these
studies have confirmed and strengthened the evidence of associations for both mortality and
5-7
-------
respiratory morbidity. For mortality, the CD places greatest weight on the reanalyses and
extensions of the Six Cities and the ACS studies, finding that these studies provide "strong
evidence" for associations with fine particles (CD, p. 9-34), notwithstanding the lack of
consistent results in other long-term exposure studies. For morbidity, the CD finds that new
studies of a cohort of children in Southern California have built upon earlier limited evidence to
provide "fairly strong" evidence that long-term exposure to fine particles is associated with
development of chronic respiratory disease and reduced lung function growth (CD, p. 9-34).
As discussed in the CD and in Chapter 3 above, mortality studies of the Six Cities and
ACS cohorts available in the last review had aggregate long-term mean PM25 concentrations of
18 |ig/m3 (ranging from approximately 11 to 30 |ig/m3 across cities) and 21 |ig/m3 (ranging from
approximately 9 to 34 |ig/m3 across cities), respectively. Reanalyses of data from these cohorts
continued to report significant associations with PM2 5, using essentially the same air quality
distributions. The extended analyses using the ACS cohort also continued to report statistically
significant associations with PM25 with the inclusion of more recent PM25 air quality data, with
an average range across the old and new time periods from about 7.5 to 30 |ig/m3 (from figure 1,
Pope et al., 2002) and a long-term mean of approximately 17.7 |ig/m3 (Pope et al., 2002). As
with the earlier cohort studies, no evidence of a threshold was observed in the relationships with
total, cardiovascular, and lung cancer mortality reported in this extended study. In the morbidity
studies of the Southern California children's cohort, the means of 2-week average PM25
concentrations ranged from approximately 7 to 32 |ig/m3, with an across-city average of
approximately 15 |ig/m3 (Peters et al., 1999). Staff notes that in figures depicting relationships
between lung function growth and average PM concentration, no apparent threshold is evident in
this study (Gauderman et al., 2000, 2002).
Beyond the epidemiologic studies using PM25 as an indicator of fine particles, a large
body of newly available evidence from studies that used PM10, as well as other indicators or
components of fine particles (e.g., sulfates, combustion-related components), provides additional
support for the conclusions reached in the last review as to the likely causal role of ambient PM,
and the likely importance of fine particles in contributing to observed health effects. Such
studies notably include new multi-city studies, intervention studies (that relate reductions in
ambient PM to observed improvements in respiratory or cardiovascular health), and source-
oriented studies (e.g., suggesting associations with combustion- and vehicle-related sources of
fine particles). Further, the CD concludes that new epidemiologic studies of ambient PM
associations with potential PM-related infant mortality and/or developmental effects are very
limited. However, if these findings were further substantiated by future research, estimates of
the extent of life shortening due to PM-related premature mortality would likely significantly
increase (CD, p. 9-94). The CD also notes that new epidemiologic studies of asthma-related
increased physicians visits and symptoms, as well as new studies of cardiac-related risk factors,
-------
suggest likely much larger public health impacts due to ambient fine particles than just those
indexed by the mortality and morbidity effects considered in the last review (CD, p. 9-94).
Staff recognizes, however, that important limitations and uncertainties associated with
this expanded body of evidence for PM25 and other indicators or components of fine particles, as
discussed in Chapter 3 herein and section 9.2.2 of the CD, need to be carefully considered in
determining the weight to be placed on the studies available in this review. For example, the CD
notes that while PM-effects associations continue to be observed across most new studies, the
newer findings do not fully resolve the extent to which the associations are properly attributed to
PM acting alone or in combination with other gaseous co-pollutants, or to the gaseous co-
pollutants themselves. The CD notes that available statistical methods for assessing potential
confounding by gaseous co-pollutants may not yet be fully adequate, although the various
approaches that have now been used to evaluate this issue tend to substantiate that associations
for various PM indicators with mortality and morbidity are robust to confounding by co-
pollutants (CD, p. 9-37).
Another issue of particular importance is the sensitivity of various statistical models to
the approach used to address potential confounding by weather- and time-related variables in
time-series epidemiological studies. As discussed in section 3.5.3 herein and in section 9.2.2 of
the CD, this issue resurfaced in the course of reanalyses of a number of the newer studies that
were being conducted to address a more narrow issue related to problems associated with the use
of commonly used statistical software. These reanalyses suggest that weather continues to be a
potential confounder of concern and highlight that no one model is likely to be most appropriate
in all cases. The HEI Review Panel, in reviewing these reanalyses, concluded that this
awareness introduces a degree of uncertainty in evaluating the findings from time-series
epidemiologic studies that had heretofore not been widely appreciated.
In looking beyond PM mass indicators, a number of newly available studies highlight the
issue of the extent to which observed health effects may be associated with various specific
chemical components within the mix of fine particles. The potential for various fine particle
components to have differing relative toxicities with regard to the various health endpoints being
considered adds complexity to the interpretation of study results. The CD recognizes that more
research is needed to address uncertainties about the extent to which various components may be
relatively more or less toxic than other components, or than undifferentiated PM2 5 mass, across
the range of health endpoints studied.
While the limitations and uncertainties in the available evidence suggest caution in
interpreting the epidemiologic studies at the lower levels of air quality observed in the studies,
staff concludes that the evidence now available provides strong support for considering fine
particle standards that would provide increased protection beyond that afforded by the current
PM2 5 standards. More protective standards would reflect the generally stronger and broader
body of evidence of associations with mortality and morbidity now available in this review, both
5-9
-------
at lower levels of air quality and at levels below the current standards, and with more
understanding of possible underlying mechanisms.
5.3.1.2 Risk-based Considerations
In addition to this evidence-based evaluation, staff has also considered the extent to
which health risks estimated to occur upon attainment of the current PM2 5 standards may be
judged to be important from a public health perspective, taking into account key uncertainties
associated with the estimated risks. In so doing, staff first notes that the risk assessment
discussed in Chapter 4 addresses a number of key uncertainties through various base case
analyses, as well as through several sensitivity analyses. Most importantly, a series of base case
analyses were conducted to characterize the uncertainty associated with the form of the
concentration-response functions drawn from the studies used in the assessment, which had by
far the greatest impact on estimated risks. Other uncertainties, including the use of single-
versus multi-pollutant models, single- versus multi-city models, use of a distributed lag model,
alternative assumptions about the relevant air quality for long-term exposure mortality, and
alternative constant or varying background levels, have a more moderate and often variable
impact on the risk estimates in some or all of the cities.
In considering the health risks estimated to occur upon attainment of the current PM25
standards, staff focused in particular on base case risk estimates, while recognizing that the
confidence ranges in the selected base case estimates do not reflect all the identified
uncertainties. These risks were estimated using not only the linear or log-linear functions
reported in the studies,1 but also using a series of alternative modified linear functions as
surrogates for assumed non-linear functions that would reflect the possibility that thresholds may
exist in the reported associations within the range of air quality observed in the studies. The
approach used to develop the alternative functions, discussed more fully in Chapter 4 (section
4.3.2.1), incorporates a modified linear slope with an imposed cutpoint (i.e., an assumed
threshold) that is intended to reflect an inflection point in a typical non-linear, "hockey-stick"
shaped function, below which there is little or no population response. As discussed in Chapter
3 (section 3.6.6), staff recognizes that while there are likely biological thresholds in individuals
for specific health responses, the available epidemiologic studies do not support or refute the
existence of thresholds at the population level for either long-term or short-term PM exposures
within the range of air quality observed in the studies (CD, p. 9-44). Thus, staff has concluded
that it is appropriate to consider health risks estimated not only with the reported linear or log-
1 As discussed in Chapter 4, the reported linear or log-linear functions were applied down to 7.5 ug/m3 in
estimating risk associated with long-term exposure (i.e., the lowest measured level in the extended ACS study), and
down to the estimated policy-relevant background level in estimating risk associated with short-term exposure (i.e.,
3.5 ug/m3 for eastern urban areas and 2.5 ug/m3 for western urban areas).
5-10
-------
linear concentration-response functions, but also with modified functions that incorporate
alternative assumed cutpoints as surrogates for potential population thresholds.
Tables 5-1 (a) and (b) summarize the estimated PM2 5-related annual incidence and
incidence rate (in terms of incidence per 100,000 general population) of total mortality
associated with long-term and short-term exposures, respectively, assuming various cutpoint
levels in the example urban areas included in the risk assessment.2 In first looking at the annual
incidence of PM25-related mortality estimated to occur upon attainment of the current PM2 5
standards in the five study areas that do not meet the current standards based on 2001-2003 air
quality data (Detroit, Los Angeles, Philadelphia, Pittsburgh, and St. Louis), staff notes that there
is a fairly wide range of estimated incidence across the areas for both long- and short-term
exposures. Such variation would be expected considering, for example, differences in total
population, demographics, baseline mortality rates, exposure considerations (e.g., degree of air
conditioning use), presence of co-pollutants and other environmental stressors, and exposure
measurement error across urban areas. The somewhat greater variation in the estimated
incidence associated with short-term exposure than with long-term exposure would also be
expected, since the assessment uses the same long-term exposure concentration-response
function in all areas, whereas the assessment used different short-term exposure functions (for
different mortality endpoints in some cases) from studies conducted in each area. Staff also
recognizes that there are uncertainties associated with the procedure used to simulate air quality
that would just attain the current standards and in the degree to which various components of the
fine particle mix would likely be reduced in similar proportion to the simulated reduction in
PM2 5 as a whole (as discussed in Chapter 2, section 2.5.1).
In the five study areas that do not meet the current standards, staff observes for long-term
exposure that point estimates of annual incidence of total PM2 5-related mortality associated with
just meeting the current PM25 standards, based on the lowest cutpoint of 7.5 |ig/m3, range from
approximately 400 to 600 in four areas (from roughly 25 to 35 deaths per 100,000 general
population in these areas) to over 1500 annual deaths in Los Angeles (roughly 16 deaths per
100,000 general population) associated with long-term exposure. These estimated incidences
associated with long-term exposure represent 2.6 to 3.2 percent of total mortality incidence due
to all causes. In the same five areas, the annual incidence associated with short-term exposure,
based on a cutpoint equal to policy-relevant background, ranges from less than 20 % to over
50% of the estimated incidence associated with long-term exposure. In some areas, the 95%
confidence ranges associated with the estimates of total annual mortality incidence related to
short-term exposure (but not long-term exposure) extend to below zero, reflecting appreciably
more uncertainty in estimates based on positive but not statistically significant associations.
These tables include risk estimates drawn from Tables 4-9, 4-10, 4-12, and 4-13 in Chapter 4.
5-11
-------
Table 5-l(a) Estimated PM2 5-related annual total mortality associated with long-term
exposure when current PM2 5 standards are met*
Annual Incidence of All-Cause Mortality
and 95% Cl
(deaths/yr)
7.5 ua/m3
Cutpoints
10 ua/m3
12 ua/m3
Annual Incidence Rate of All-Cause Mortality
and 95% Cl
(deaths/yr/1 00,000 general population)
7.5 ua/m3
Cutpoints
10 ua/m3
12 ua/m3
Risks associated with just meeting current PM25 standards
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
520
180-910
1,510
530 - 2,590
540
190-940
400
140-0700
600
210-1,050
280
100-490
820
290-1420
340
120-0600
220
80-o 370
310
110-550
40
10-70
140
50-o 240
140
50 - 240
30
10-40
20
10-40
25
9-44
16
6-27
35
12-62
31
11-55
24
8-42
14
5-24
9
3-15
22
8-39
17
6-29
12
4-22
2
1-3
1
1-2
9
3-16
2
1-3
1
0-2
Risks associated with "as is " air quality (in areas that meet current PM25 standards)
Boston
Phoenix
San Jose
Seattle
590
200-1050
350
120-620
170
60-310
50
20-90
310
110-550
80
30-140
60
20-o 100
0
20
10-40
0
0
0
21
7-38
11
4-20
10
4-18
3
1-5
11
4-20
2
1-3
3
1-6
0
1
0-1
0
0
0
* These estimates are based on using the maximum monitor in an area to calculate the percent rollback needed to
just attain the current PM2 5 annual standard, and applying that percent rollback to the composite monitor in the area,
as described in Chapter 4, section 4.2.3. Estimates of annual mortality incidence based on using a spatially averaged
concentration to calculate the percent rollback needed to just attain the current standard, where this is allowed,
would be higher than the estimates shown here.
5-12
-------
Table 5-l(b) Estimated PM2 5-related annual mortality associated with short-term
exposure when current PM2 5 standards are met*
Annual Incidence of Non-Accidental Mortality
and 95% Cl
(except as noted)
(deaths/yr)
Cutpoints
Policy-
relevant
Background**
10 |jg/m3
15 |jg/m3
20 |jg/m3
Annual Incidence Rate of Non-Accidental
Mortality
and 95% Cl
(except as noted)
(deaths/yr/1 00,000 general population)
Cutpoints
Policy-
relevant
Background**
10 |jg/m3
15 |jg/m3
20 |jg/m3
Risks associated with just meeting current PM25 standards
Detroit
Los Angeles
Philadelphia
cardiovascular mortality
Pittsburgh
over age 74
St. Louis
120
-120-360
290
-40 - 610
370
180-560
50
-110-200
190
70-310
50
-60-160
120
-10-240
190
90 - 290
20
-50 - 90
80
30-120
30
-30 - 80
60
-10-120
110
50-160
10
-20 - 40
30
10-50
10
-10 to 40
30
-4 to 60
60
30 to907
5
-10 to 20
9
3 to 14
6
-6-17
3
0-6
24
12-37
4
-8-16
8
3-12
3
-3-8
1
0-3
12
6-19
2
-4-7
3
1-5
1
-1-4
1
0-1
7
3-11
1
-2-3
1
0-2
1
-1-2
0
0-1
4
2-6
0
-1-1
0
0-1
Risks associated with "as is " air quality (in areas that meet current PM25 standards)
Boston
Phoenix
cardiovascular mortality
over age 64
San Jose
390
270-510
320
100-540
220
50 - 390
170
120 - 230
90
30-140
80
20-140
80
60-110
60
20-90
40
10-80
40
30- 50
40
10-60
30
10-50
14
9-18
11
3-17
13
3-23
6
4-8
3
1-5
5
1-8
3
2-4
2
1-3
3
1-5
1
1-2
1
0-2
2
0-3
* These estimates are based on using the maximum monitor in an area to calculate the percent rollback needed to
just attain the current PM2 5 annual standard, and applying that percent rollback to the composite monitor in the area,
as described in Chapter 4, section 4.2.3. Estimates of annual mortality incidence based on using a spatially averaged
concentration to calculate the percent rollback needed to just attain the current standard, where this is allowed,
would be higher than the estimates shown here.
** Estimated policy-relevant background levels are 3.5 ug/m3 for eastern urban areas and 2.5 ug/m3 for western
urban areas.
5-13
-------
In the other four areas that meet the current standards based on recent air quality data (Boston,
Phoenix, San Jose, and Seattle), point estimates of annual incidence of total PM2 5-related
mortality associated with long-term exposure range from about 50 deaths in Seattle (roughly 3
deaths per 100,000 general population) to almost 600 deaths in Boston (roughly 21 deaths per
100,000 general population). Estimated incidence associated with short-term exposure in these
four areas generally falls within the range of the estimates associated with long-term exposure.
In considering the estimated incidences associated with long-term exposure based on
assumed cutpoint of 10 |ig/m3, staff observes that these estimates are roughly about half as large
as the estimates based on a cutpoint of 7.5 |ig/m3. Under this assumption, point estimates of
annual incidence of total PM2 5-related mortality associated with just meeting the current PM25
standards range from about 200 to over 300 in four of the areas that do not meet the current
standards (from roughly 12 to 22 deaths per 100,000 general population in these areas) to over
800 annual deaths in Los Angeles (roughly 9 deaths per 100,000 general population) associated
with long-term exposure. In considering an assumed cutpoint as high as 12 |ig/m3, point
estimates associated with long-term exposure in these five areas are roughly 5 to 20% of the
estimates based on the lowest cutpoint. A similar pattern is seen when considering the impact of
alternative assumed cutpoints in the range of 10 to 20 |ig/m3 on risks associated with short-term
exposure.
5.3.1.3 Summary
In considering these estimates of PM25-related mortality upon meeting the current
standards in a number of example urban areas, together with the uncertainties in these estimates,
staff concludes that they are indicative of risks that can reasonably be judged to be important
from a public health perspective and that they provide support for consideration of standards that
would provide increased protection beyond that afforded by the current PM2 5 standards. In the
absence of evidence of clear thresholds, staff believes it is appropriate to consider all the
estimates associated with the range of assumed cutpoints used in the risk assessment. Staff
believes that a relatively more precautionary approach to interpreting this evidence would give
more weight to the estimates based on the lowest cutpoints considered. Staff also takes note of
the view expressed by the CAS AC PM Panel which "favored the primary use of an assumed
threshold of 10 jig/m3." (Henderson, 2005). Regardless of the relative weight placed on the
estimates associated with either an assumed cutpoint of 10 |ig/m3 or the lowest cutpoints
considered, the risk assessment indicates the likelihood that thousands of premature deaths per
year would occur in urban areas across the U.S. even upon attainment of the current PM25
standards. Beyond the estimated incidences of mortality discussed above, staff also recognizes
that similarly substantial numbers of incidences of hospital admissions, emergency room visits,
aggravation of asthma and other respiratory symptoms, and increased cardiac-related risk are
also likely in many urban areas, based on risk assessment results presented in Chapter 4 and on
the discussion related to the pyramid of effects drawn from section 9.2.5 of the CD. Staff also
5-14
-------
believes it is important to recognize how highly dependent any specific risk estimates are on the
shape of the underlying concentration-response functions. In so doing, staff nonetheless
reiterates that based on even the highest assumed cutpoints used in the risk assessment, estimated
mortality risks are not completely eliminated when current PM2 5 standards are met in a number
of example urban areas, including all such areas that do not meet the standards based on recent
air quality data.
Staff also well recognizes that as the body of available evidence has expanded, it has
added greatly both to our knowledge of PM-related effects, as well as to the complexity inherent
in interpreting the evidence in a policy-relevant context as a basis for setting appropriate
standards. In considering available evidence, risk estimates, and related limitations and
uncertainties, staff concludes 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 PM2 5 standards to provide increased public health
protection. Staff conclusions and recommendations for indicators, averaging times, and levels
and forms of alternative, more protective primary standards for fine particles are discussed in the
following sections.
5.3.2 Indicators
In 1997, EPA established PM25 as the indicator for fine particles. In reaching this
decision, the Agency first considered whether the indicator should be based on the mass of a
size-differentiated sample of fine particles or on one or more components within the mix of fine
particles. Secondly, in establishing a size-based indicator, a size cut needed to be selected that
would appropriately distinguish fine particles from particles in the coarse mode.
In addressing the first question in the last review, EPA determined that it was more
appropriate to control fine particles as a group, as opposed to singling out any particular
component or class of fine particles. Community health studies had found significant
associations between various indicators of fine particles (including PM25 or PM10 in areas
dominated by fine particles) and health effects in areas with significant mass contributions of
differing components or sources of fine particles, including sulfates, wood smoke, nitrates,
secondary organic compounds and acid sulfate aerosols. In addition, a number of animal
toxicologic and controlled human exposure studies had reported health effects associations with
high concentrations of numerous fine particle components (e.g., sulfates, nitrates, transition
metals, organic compounds), although such associations were not consistently observed. It also
was not possible to rule out any component within the mix of fine particles as not contributing to
the fine particle effects found in epidemiologic studies. For these reasons, EPA concluded that
total mass of fine particles was the most appropriate indicator for fine particle standards rather
than an indicator based on PM composition (62 FR 38667, July 18, 1997).
5-15
-------
Having selected a size-based indicator for fine particles, the Agency then based its
selection of a specific size cut on a number of considerations. In focusing on a size cut within
the size range of 1 to 3 |im (i.e., the intermodal range between fine and coarse mode particles),
EPA recognized that the choice of any specific sampling size cut within this range was largely a
policy judgment. In making this judgment, the Agency noted that the available epidemiologic
studies of fine particles were based largely on PM25; only very limited use of PMX monitors had
been made. While it was recognized that using PMX as an indicator of fine particles would
exclude the tail of the coarse mode in some locations, in other locations it would miss a portion
of the fine PM, especially under high humidity conditions, which would result in falsely low fine
PM measurements on days with some of the highest fine PM concentrations. The selection of a
2.5 |im size cut reflected the regulatory importance that was placed on defining an indicator for
fine particle standards that would more completely capture fine particles under all conditions
likely to be encountered across the U.S., especially when fine particle concentrations are likely
to be high, while recognizing that some small coarse particles would also be captured by PM2 5
monitoring.3 Thus, EPA's selection of 2.5 |im as the size cut for the fine particle indicator was
based on considerations of consistency with the epidemiologic studies, the regulatory importance
of more completely capturing fine particles under all conditions, and the potential for limited
intrusion of coarse particles in some areas; it also took into account the general availability of
monitoring technology (62 FR 38668).
In this current review, the same considerations continue to apply for selection of an
appropriate indicator for fine particles. As an initial matter, the available epidemiologic studies
linking mortality and morbidity effects with short- and long-term exposures to fine particles
continue to be largely indexed by PM2 5. Some epidemiologic studies also have continued to
implicate various PM components (e.g., sulfates, nitrates, carbon, organic compounds, and
metals) as being associated with adverse effects; effects have been reported with a broad range
of PM components, as summarized in Table 9-3 of the CD (p. 9-31). Animal toxicologic and
controlled human exposure studies, evaluated in Chapter 7 of the CD, have continued to link a
variety of PM components or particle types (e.g., sulfates or acid aerosols, metals, organic
constituents, bioaerosols, diesel particles) with health effects, though often at high
concentrations (CD section 7.10.2). In addition, some recent studies have suggested that the
ultrafme subset of fine particles may also be associated with adverse effects (CD, pp. 8-67 and 8-
68, 8-199).
Staff recognizes that, for a given health response, some PM components are likely to be
more closely linked with that response than others (CD, p. 9-30). That different PM constituents
3 In reaching this decision, EPA indicated that it might be appropriate to address undue intrusion of coarse
mode particles resulting in violations of PM2 5 standards in the context of policies established to implement such
standards (62 FR 38668).
5-16
-------
may have differing biological responses is an important source of uncertainty in interpreting
epidemiologic evidence. For specific effects there may be stronger correlation with individual
PM components than with particle mass. For example, in some toxicologic studies of
cardiovascular effects (such as changes in heart rate, electrocardiogram measures, or increases in
arrhythmia), PM exposures of equal mass did not produce the same effects, indicating that PM
composition was important (CD, p. 7-30). In addition, section 9.2.3.1.3 of the CD indicates that
particles, or particle-bound water, can act as carriers to deliver other toxic agents into the
respiratory tract, suggesting that exposure to particles may elicit effects that are linked with a
mixture of components more than with any individual PM component.
Thus, epidemiologic and toxicologic studies summarized above and discussed in the CD
have provided evidence for effects associated with various fine particle components or size-
differentiated subsets of fine particles. The CD concludes: "These studies suggest that many
different chemical components of fine particles and a variety of different types of source
categories are all associated with, and probably contribute to, mortality, either independently or
in combinations" (CD, p. 9-31). Conversely, the CD provides no basis to conclude that any
individual fine particle component cannot be associated with adverse health effects. There is not
sufficient evidence that would lead toward the selection of one or more PM components as being
primarily responsible for effects associated with fine particles, nor is there any component that
can be eliminated from consideration. Staff continues to recognize the importance of an
indicator that not only captures all of the most harmful components of fine PM (i.e., an effective
indicator), but also places greater emphasis for control on those constituents or fractions,
including sulfates, transition metals, and organics that have been associated with health effects in
epidemiologic and/or toxicologic studies, are most likely to result in the largest risk reduction
(i.e., an efficient indicator). Taking into account the above considerations, staff concludes that it
remains appropriate to control fine particles as a group; i.e., that total mass of fine particles is the
most appropriate indicator for fine particle standards.
With regard to an appropriate size cut for a size-based indicator of total fine particle
mass, the CD most generally concludes that advances in our understanding of the characteristics
of fine particles continue to support the use of particle size as an appropriate basis for
distinguishing between these subclasses, and that a nominal size cut of 2.5 |im remains
appropriate (CD, p. 9-22). This conclusion follows from a recognition that within the intermodal
range of 1 to 3 |im there is no unambiguous definition of an appropriate size cut for the
separation of the overlapping fine and coarse particle modes (CD, p. 9-8). Within this range,
staff considered size cuts of both 1 |im and 2.5 |im. Consideration of these two size cuts took
into account that there is generally very little mass in this intermodal range, although in some
circumstances (e.g., windy, dusty areas) the coarse mode can extend down to and below 1 |im,
whereas in other circumstances (e.g., high humidity conditions, usually associated with very high
fine particle concentrations) the fine mode can extend up to and above 2.5 |im. The same
5-17
-------
considerations that led to the selection of a 2.5 |im size cut in the last review - that the
epidemiologic evidence was largely based on PM2 5 and that it was more important from a
regulatory perspective to more completely capture fine particles under all conditions likely to be
encountered across the U.S. (especially when fine particle concentrations are likely to be high)
than to avoid some coarse-mode intrusion into the fine fraction in some areas - lead to the same
conclusion in this review. In addition, section 9.2.1.2.3. of the CD discusses the potential health
significance of particles as carriers of water, oxidative compounds, and other components into
the respiratory system. This consideration adds to the importance of ensuring that larger
accumulation-mode particles are included in the fine particle size cut. Therefore, as observed
previously in section 3.1.2, the scientific evidence leads the CD to conclude that 2.5 jim remains
an appropriate upper size cut for a fine particle mass indicator.
Consistent with that conclusion, staff recommends that PM25 be retained as the indicator
for fine particles. Staff further concludes that currently available studies do not provide a
sufficient basis for supplementing mass-based fine particle standards with standards for any
specific fine particle component or subset of fine particles, or for eliminating any individual
component or subset of components from fine particle mass standards.
Further, staff notes that since the last review an extensive PM25 monitoring network has
been deployed and operated in cooperative efforts with State, local and Tribal agencies and with
instrument manufacturers. At the same time, EPA has been working on the development of
strategies and programs to implement the 1997 PM2 5 standards, based on the federal reference
method (FRM) sampler for PM2 5. The new monitoring network has provided substantial new air
quality information, in terms of PM2 5, that has been and is being used in ongoing PM research
and air quality analyses that inform this review. EPA also has conducted studies to evaluate
options for improvements to the FRM. As a result of continuing evaluation of the monitoring
network, staff is considering changes to the PM25 FRM to improve performance and minimize
the burden on agencies conducting the monitoring.4 Staff is also considering the addition of
federal equivalent method (FEM) designation criteria for continuous fine particle monitors.5
Continuous monitoring is advantageous in providing additional data for many purposes,
including compliance monitoring, health studies, and air quality forecasting, and it can also ease
the burden of data collection for regulatory agencies.
4 Changes to the PM2 s FRM being considered by staff are discussed in Hanley (2005).
This work is being done in consultation with the CAS AC Subcommittee on Ambient Air Monitoring and
Methods (AAMM).
5-18
-------
5.3.3 Averaging Times
In the last review, EPA established two PM2 5 standards, based on annual and 24-hour
averaging times (62 FR at 38,668-70). This decision was based in part on evidence of health
effects related to both short-term (from less than 1 day to up to several days) and long-term
(from a year to several years) measures of PM. EPA noted that the large majority of community
epidemiologic studies reported associations based on 24-hour averaging times or on multiple-day
averages. Further, EPA noted that a 24-hour standard could also effectively protect against
episodes lasting several days, as well as providing some degree of protection from potential
effects associated with shorter duration exposures. EPA also recognized that an annual standard
would provide effective protection against both annual and multi-year, cumulative exposures that
had been associated with an array of health effects, and that a much longer averaging time would
complicate and unnecessarily delay control strategies and attainment decisions. The possibility
of seasonal effects also was considered, although the very limited available evidence of such
effects and the seasonal variability of sources of fine particle emissions across the country did
not provide a satisfactory basis for establishing a seasonal averaging time.
In considering whether the information available in this review supports consideration of
different averaging times for PM25 standards, staff notes that the available information is
generally consistent with and supportive of the conclusions reached in the last review to set
PM2 5 standards with both annual and 24-hour averaging times. In considering the new
information, staff makes the following observations:
There is a growing body of studies that provide additional evidence of effects
associated with exposure periods shorter than 24-hours (e.g., one to several
hours), as discussed in Chapter 3 (section 3.5.5.1). While staff concludes that this
information remains too limited to serve as a basis for establishing a shorter-than-
24-hour fine particle primary standard at this time, staff believes that it gives
added weight to the importance of a standard with a 24-hour averaging time.
Staff recognizes shorter-than-24-hour exposures as an important area of research
that could provide a basis for the consideration of a shorter-term standard in the
future.
• As discussed in Chapter 3 (section 3.5.5), some recent PM10 studies have used a
distributed lag over several days to weeks preceding the health event, although
this modeling approach has not been extended to studies of fine particles. While
such studies continue to suggest consideration of a multiple day averaging time,
staff notes that limiting 24-hour concentrations of fine particles will also protect
against effects found to be associated with PM averaged over many days in health
studies. Consistent with the conclusion reached in the last review, staff again
concludes that a multiple-day averaging time would add complexity but would
not provide more effective protection than a 24-hour average.
5-19
-------
• While some newer studies have investigated seasonal effects, as noted in Chapter
3 (section 3.5.5.3), staff concludes that currently available evidence of such
effects is still too limited to serve as a basis for considering seasonal standards.
Based on the above considerations, staff concludes that the currently available
information supports keeping, and provides no adequate basis for changing, the averaging times of
the current PM25 standards. Staff notes that study of shorter-term averaging times, on the order
of one or more hours, is an important research priority, with a particular focus on associations
between exposure to fine particles and fine-particle constituents and indicators of cardiac-related
risk factors. Thus, a shorter-term averaging time may be an important consideration in the next
review of the PM NAAQS. Staff also notes that at present EPA has in place a significant harm
level program (40 CFR Part 51) and a widely disseminated Air Quality Index that could
potentially be adapted to provide information to the public based on episodic very short-term
peak fine particle levels that may be of public health concern.
In the last review, having decided to set both annual and 24-hour PM2 5 standards, EPA
also made judgments as to the most effective and efficient approach to establishing a suite of
standards that, taken together, would appropriately protect against effects associated with both
long- and short-term exposures. At that time, EPA selected an approach that was based on
treating the annual standard as the generally controlling standard for lowering the entire
distribution of PM25 concentrations, with the 24-hour standard providing additional protection
against the occurrence of peak 24-hour concentrations. The 24-hour standard was intended to
address in particular those peaks that result in localized or seasonal exposures of concern in areas
where the highest 24-hour-to-annual mean PM2 5 ratios are appreciably above the national
average. This approach was supported by results of the PM risk assessment from the last review
which indicated that peak 24-hour PM2 5 concentrations contribute a relatively small amount to
total health risk, such that much if not most of the aggregated annual 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. Further, no evidence suggested that risks associated
with long-term exposures are likely to be disproportionately driven by peak 24-hour
concentrations. Thus, a generally controlling annual standard was judged to reduce risks
associated with both short- and long-term exposures effectively and with more certainty than a
24-hour standard. Further, an annual standard was seen to be more stable over time, likely
resulting in the development of more consistent risk reduction strategies, since an area's
attainment status would be less likely to change due solely to year-to-year variations in
meteorological conditions that affect the atmospheric formation of fine particles.
In this review, some key considerations that led to establishing a generally controlling
annual standard in the last review are still valid. In particular,
EPA's updated risk assessment supports the previous conclusion that peak 24-
hour PM2 5 concentrations contribute a relatively small amount to the total health
5-20
-------
risk associated with short-term exposures on an annual basis, such that much if
not most of the aggregated annual risk results from the large number of days
during which the 24-hour average concentrations are in the low- to mid-range, as
discussed in Chapter 4 (section 4.4.4). Support for this conclusion is also found
in studies in which health effect associations remain when high-concentration
days are removed from the analysis (Schwartz et al., 1996; Ostro et al., 1999,
2000).
• It continues to be the case, as discussed in section 4.3.6.6, that available short-
term exposure studies do not provide evidence of clear population thresholds, but
rather reflect relationships between health effects and ambient PM across a wide
distribution of PM concentrations. Thus, as in the last review, staff recognizes
that these studies do not provide a basis for identifying a lowest-observed-effect
level that would clearly translate into a 24-hour standard that would protect
against all effects related to short-term exposures.
Nonetheless, staff believes that the greatly expanded body of epidemiologic evidence and
air quality data provide the basis for considering alternative approaches to establishing a suite of
PM25 standards. Thus, staff has not focused a priori on an annual standard as the generally
controlling standard for protection against effects associated with both long- and short-term
exposures. Rather, staff has broadened its view to consider both evidence-based and risk-based
approaches to evaluating the protection that a suite of PM25 standards can provide against effects
associated with long-term exposures and against effects associated with short-term exposures.
These evaluations, discussed in the next two sections, provide the basis for integrated
recommendations on ranges of alternative suites of standards that, when considered together,
protect against effects associated with both long- and short-term exposures.
5.3.4 Alternative PM2 5 Standards to Address Health Effects Related to Long-term
Exposure
In considering alternative PM2 5 standards that would provide protection against health
effects related to long-term exposures, staff has taken into account both evidence-based and risk-
based considerations. As discussed below in this section, staff has first evaluated the available
evidence from long-term exposure studies, as well as the uncertainties and limitations in that
evidence, to assess the degree to which alternative annual PM2 5 standards can be expected to
provide protection against effects related to long-term exposures. Secondly, staff has considered
the quantitative risk estimates for long-term exposure effects, discussed in Chapter 4, to assess
the extent to which alternative annual and/or 24-hour standards can be expected to reduce the
estimated risks attributable to long-term exposure to PM2 5. Staff conclusions as to ranges of
alternative annual and/or 24-hour standards that would provide protection against health effects
related to long-term exposures are summarized at the end of this section. The integrated staff
recommendations presented in section 5.3.7 are based in part on the conclusions from this
5-21
-------
section and in part on staff conclusions from the next section, in which alternative PM2 5
standards to address health effects related to short-term exposures are assessed.
5.3.4.1 Evidence-based Considerations
In taking into account evidence-based considerations, staff has focused on long-term
exposure studies of fine particles in the U.S. As discussed above, staff notes that the reanalyses
and extensions of earlier studies have confirmed and strengthened the evidence of long-term
associations for both mortality and morbidity effects. The assessment in the CD of these
mortality studies, taking into account study design, the strength of the study (in terms of
statistical significance and precision of result), and the consistency and robustness of results,
concluded that it was appropriate to give the greatest weight to the reanalyses of the Six Cities
study and the ACS study, and in particular to the results of the extended ACS study (CD, p.
9-33). The assessment in the CD of the relevant morbidity studies noted in particular the results
of the new studies of the children's cohort in Southern California as providing evidence of
respiratory morbidity with long-term PM exposures (CD, pp. 9-33 to 9-34).
Staff believes it is appropriate to consider a level for an annual PM25 standard that is
somewhat below the averages of the long-term concentrations across the cities in each of these
long-term exposure studies, recognizing that the evidence of an association in any such 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, 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. Staff also believes it
is appropriate to consider the long-term average concentration at the point where the confidence
interval becomes notably wider, suggestive of a concentration below which the association
becomes appreciably more uncertain and the possibility that an effects threshold may exist
becomes more likely. Staff further notes that in considering a level for a standard that is to
provide protection with an adequate margin of safety, it is appropriate to take into account
evidence of effects for which the reported associations provide only suggestive evidence of a
potentially causal association.
In looking first at the long-term exposure mortality studies, staff notes that the long-term
mean PM25 concentration in the Six Cities study was 18 |ig/m3, within an overall range of 11 to
30 |ig/m3. In the studies using the ACS cohort, the long-term mean PM25 concentration across
the cities was 21 |ig/m3 in the initial study and in the reanalysis of that study, within an overall
range of 9 to 34 |ig/m3. In the extended ACS study, the mean for the more recent time period
used in the analysis (from 1999 to 2000) was 14 |ig/m3; in looking at the association based on the
air quality averaged over both time periods (which was the basis for the concentration-response
functions from this study used in the risk assessment, as explained in Chapter 4), the long-term
mean PM25 concentration was 17.7 |ig/m3, with a standard deviation of + 4, ranging down to
7.5 |ig/m3. The CD notes that the confidence intervals around the relative risk functions in this
5-22
-------
extended study, as in the initial ACS study, start to become appreciably wider below
approximately 12 to 13 |ig/m3. In considering the Southern California children's cohort study
showing evidence of decreased lung function growth, staff notes that the long-term mean PM25
concentration was 15 |ig/m3, ranging from 7 to 32 |ig/m3 across the cities. This is approximately
equal to the long-term mean PM2l concentration in the earlier 24 City study, showing effects on
children's lung function, in which the long-term mean concentration was 14.5 |ig/m3, ranging
from 9 to 17 |ig/m3 across the cities.
In considering this evidence, staff concludes that these studies provide a basis for
considering an annual PM2 5 standard somewhat below 15 |ig/m3, down to about 12 |ig/m3. A
standard of 14 |ig/m3 would reflect some consideration of the more recent long-term exposure
studies that show associations over a somewhat lower range of air quality than had been
observed in the studies available in the last review. A standard of 13 |ig/m3 would be consistent
with a judgment that appreciable weight should be accorded these long-term exposure studies,
particularly taking into account the most recent extended ACS mortality study and the Southern
California children's cohort morbidity study. A standard level of 13 |ig/m3 would be well below
the long-term mean in the Six Cities mortality study and approximately one standard deviation
below the extended ACS mortality study mean, while being somewhat closer to the long-term
means in the morbidity studies discussed above. A standard of 12 |ig/m3 would be consistent
with a judgment that a more precautionary standard was warranted, potentially reflecting
consideration of the seriousness of the mortality effects, for which there is strong evidence of
likely causal relationships, and of the limited but suggestive evidence of possible links to effects
on fetal and infant development and mortality. As discussed in Chapter 1, these factors are
relevant to judgments about providing an adequate margin of safety to prevent pollution levels
that may pose an unacceptable risk of harm, even if the risk is not precisely identified as to
nature or degree. In staffs view, a standard set below this range would be highly precautionary,
giving little weight to the remaining uncertainties in the broader body of evidence, which
includes other long-term exposure studies that provide far more inconsistent results.
5.3.4.2 Risk-based Considerations
Beyond looking directly at the relevant epidemiologic evidence, staff also has considered
the extent to which specific levels and forms of alternative PM25 standards are likely to reduce
the estimated risks attributable to long-term exposure to PM2 5 and the uncertainties in the
estimated risk reductions. As discussed above (section 5.3.1), staff has based this evaluation on
the risk assessment results presented in Chapter 4, in which long-term exposure mortality risks,
based on the extended ACS study, were estimated using the reported concentration-response
function down to a level of 7.5 |ig/m3, the lowest measured level (LML) in that study, as well as
using modified concentration-response functions that incorporate alternative assumed cutpoints
as surrogates for potential population thresholds.
5-23
-------
Figures 5.1 (a), (b), and (c) show the estimated percentage reductions in mortality
attributable to long-term exposure to PM2 5 in going from meeting the current PM2 5 standards to
meeting alternative annual and 24-hour PM2 5 standards (with a 98th percentile form) in the five
example cities that do not meet the current standards (based on 2001-2003 air quality data),
based on assumed cutpoints of 7.5, 10, and 12 |ig/m3, respectively. To put the estimated
percentage reductions in perspective, these figures also include the estimated PM2 5-related
annual incidence rate (in terms of deaths/year/100,000 general population) and annual incidence
(in terms of deaths/year) of total mortality associated with long-term exposure associated with
just meeting the current PM25 standards. A similar series of figures is shown in Appendix 5 A for
meeting alternative 24-hour standards with a 99th percentile form. The alternative annual PM2 5
standards considered in these figures include a range of levels from 15 to 12 |ig/m3. Attainment
of the standards is simulated based on a percent rollback calculated using the highest monitor in
an area, as noted in Tables 5-l(a) and (b) and discussed in Chapter 4, section 4.2.2. The
alternative 24-hour PM25 standards considered in these figures include a range of levels from 65
to 25 |ig/m3. Further discussion of alternative forms of the annual and 24-hour standards is
presented below in section 5.3.6.
In considering the estimates based on a cutpoint of 7.5 |ig/m3 [Figures 5-l(a) and 5A-
l(a)], staff first examined the estimated reductions associated with lower levels of the annual
PM25 standard, without changing the 24-hour standard. Staff observes that alternative annual
standard levels of 14, 13, and 12 |ig/m3 result in generally consistent estimated risk reductions
from long-term exposure to PM25 of roughly 20, 30, and 50 percent, respectively, across all five
example cities. Thus, for this assumed cutpoint, estimated reductions in mortality associated
with long-term exposure to PM2 5 are no greater than 50 percent in any of the five example cities
with changes in the annual standard down to a level of 12 |ig/m3. Staff also examined the effect
on mortality reduction associated with alternative 24-hour standards, without changing the
annual standard. Staff first notes that the estimated reductions in long-term mortality risk
associated with changes to the 24-hour standard are much more variable across cities than with
changes in just the annual standard. Further, no combination of standards within the ranges that
staff has considered result in the elimination of all estimated long-term mortality risk in all
example cities. This assessment indicates that estimated reductions in long-term mortality risk
of approximately 50 percent or greater in the five example cities generally result from 24-hour
standards set at 30 to 25 |ig/m3, based on either the 98th or 99th percentile form of such a
standard, depending on the city.
Staff further considered the effects of various combinations of the annual and 24-hour
standard. Staff notes in particular that the estimates of long-term mortality risk reduction, based
on a cutpoint of 7.5 |ig/m3, associated with a 24-hour standard set at 25 |ig/m3 provides the same
degree of risk reduction regardless of the level of the annual standard within the range of 15 to
12 |ig/m3; a 24-hour standard set at 30 |ig/m3 provides the same degree of risk reduction in most
5-24
-------
Incidence Rate: 25 (9-44) deaths/yr/100,000
Incidence: 500 (180-910) deaths/yr
100'
Incidence Rate: 16 (6-27) deaths/yr/100,000
Incidence: 1500 (530-2600) deaths
100'
percent
24-hour
standard (ug/m3)
(ug/m3)
Los Angeles
Incidence Rate: 35 (12-62) deaths/yr/100,000
Incidence: 540 (190-940) deaths/yr
Incidence Rate: 31 (11-55) deaths/yr/100,000
Incidence: 400 (140-700) deaths/yr
100
(ug/m)
24-hour
standard (ug/m3)
Pittsburgh
Incidence Rate: 24 (8-42) deaths/yr/100,000
Incidence: 600 (210-1000) deaths/yr
100
24-hour
standard (ug/m3)
St. Louis
Figure 5-l(a) Estimated percent reduction in PM2.5-related long-term mortality risk for
alternative standards (9ffhpercentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 7.5 ng/m3). Risk
associated with meeting current PIVb.s standards, based on ACS extended study, is
shown in figures in terms of estimated annual incidence rate and annual incidence
(and 95% confidence ranges).
5-25
-------
Incidence rate: 14 (5-24) deaths/yr/100,000
Incidence: 280 (100-490) deaths/yr
100'
(ug/m3)
6S
Detroit
24-hour
standard (ug/m )
Incidence rate: 9 (3-15) deaths/yr/100,000
Incidence: 820 (290-1400) deaths/yr
100-r
80-
Estimated
percent Rn.
reduction
Annual *v>
(ug/m3)
24-hour
standard (ug/m )
Los Angeles
Incidence rate: 22 (8-39) deaths/yr/100,000
Incidence: 340 (120-600) deaths/yr
100'
Estimated
percent Kn±.
reduction
Philadelphia
Incidence rate: 17 (6-29) deaths/yr/100,000
Incidence: 220 (75-370) deaths/yr
Annual
standard
(ug/m )
24-hour
standard (ug/m )
Pittsburgh
Incidence rate: 12 (4-22) deaths/yr/100,000
Incidence: 310 (110-550) deaths/yr
80-
Estimated
percent fin±
reduction
St. Louis
Figure 5-l(b) Estimated percent reduction in PM2.5-related long-term mortality risk for
alternative standards (9ffh percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 10 fig/it/). Risk
associated with meeting current PM2.s standards, based on ACS extended study, is
shown in figures in terms of estimated annual incidence rate and annual incidence
(and 95% confidence ranges).
5-26
-------
Incidence rate: 2 (1-3) deaths/yr/100,000
Incidence: 40 (10-70) deaths/yr
100'
so-
Estimated
percent ^
reduction
24-hour
standard (ug/m3)
(ug/m3)
Detroit
Incidence rate: 1 (1-2) deaths/yr/100,000
Incidence: 140 (50-240) deaths/yr
Los Angeles
Incidence rate: 9 (3-16) deaths/yr/100,000
Incidence: 140 (50-240) deaths/yr]
100'
so-
Estimated
percent 60+
reduction
Annual
standard
(ug/m3)
24-hour
"tandard (ugtrnf
Philadelphia
Incidence rate: 2 (1-3) deaths/yr/100,000
Incidence: 30 (10-40) deaths/yr]
BO
s$
so
<£
b 6JJ
Pittsburgh
24-hour
standard (ug/m3)
Incidence rate: 1 (0-2) deaths/yr/100,000
Incidence: 20 (10-40) deaths/yr]
24-hour
standard (ug/m3)
(ug/m3)
St. Louis
Figure 5-l(c) Estimated percent reduction in PM2.5-related long-term mortality risk for
alternative standards (9&h percentileform) relative to risk associated with
meeting current standards (based on assumed cutpoint of 12 fig/it/). Risk
associated with meeting current PM2.5 standards, based on ACS extended study, is
shown in figures in terms of estimated annual incidence rate and annual incidence
(and 95% confidence ranges).
5-27
-------
but not all cases. That is, in the range of 30 to 25 |ig/m3, the 24-hour standard would be the
generally controlling standard in most cases relative to an annual standard in the range of 15 to
12 |ig/m3; and, in those cases, lowering the annual standard to as low as 12 |ig/m3 would result in
no additional estimated reductions in long-term mortality risks.
In considering the estimates of risk reductions based on an assumed cutpoint of 10 |ig/m3
[Figures 5-l(b) and 5A-l(b)], staff again notes that the estimates of mortality incidence and
incidence rate associated with meeting the current standards are roughly about half as large as
the estimates based on a cutpoint of 7.5 |ig/m3, as discussed above in section 5.3.1. Staff
observes that lowering the annual standard to alternative levels of 14, 13, and 12 |ig/m3 (without
changing the 24-hour standard) results in estimated risk reductions of roughly 30 to 40 percent,
50 to 70 percent, and 80 to 100 percent, respectively, across the five example cities. In
considering changes to the annual and/or 24-hour PM25 standards in this case, staff first notes
that mortality risk associated with long-term exposure is estimated to be reduced by 100 percent
in all five cities with a 24-hour standard set at 25 |ig/m3 (with either a 98th or 99th percentile
form), in combination with the current annual standard. For a 24-hour standard set at 30 |ig/m3
with a 98th percentile form, in combination with the current annual standard, estimated risk
reductions remain at or close to 100 percent in three of the cities, but are appreciably lower in the
other two cities. A 24-hour standard set at 35 |ig/m3 with a 98th percentile form results in
appreciable risk reductions in only two of the cities in conjunction with the current annual
standard, although appreciable risk reductions are observed with this 24-hour standard in
conjunction with a lower annual standard.
Further, in considering an assumed cutpoint of 12 |ig/m3 [Figures 5-l(c) and 5A-l(c)],
staff observes that lowering the annual standard to a level of 14 |ig/m3 (without changing the 24-
hour standard) results in estimated risk reductions of 100 percent in all five cities. In considering
changes to the 24-hour PM25 standard alone in this case, staff notes that long-term mortality risk
is estimated to be reduced by 100 percent in all five cities with a 24-hour standard set at
30 |ig/m3, 98th percentile form.
5.3.4.3 Summary
In considering the epidemiologic evidence, estimates of risk reductions associated with
alternative annual and/or 24-hour standards, and the related limitations and uncertainties, staff
concludes that there is clear support for considering revisions to the suite of current PM25
standards to provide additional protection against health effects associated with long-term
exposures. In looking specifically at the evidence of associations between long-term exposure to
PM2 5 and serious health effects, including total, cardiovascular, and lung cancer mortality, as
well as respiratory-related effects on children, staff concludes that it is appropriate to consider an
annual PM25 standard in the range of 15 down to 12 |ig/m3. In considering the results of the
quantitative risk assessment, staff believes that it is appropriate to consider all the estimates
associated with the range of assumed cutpoints used in the risk assessment. As discussed above
5-28
-------
in section 5.3.1.3, staff believes that a relatively more precautionary approach to interpreting this
evidence would give more weight to the estimates based on the lowest cutpoint considered,
while giving more weight to the estimates based on an assumed cutpoint of 10 |ig/m3 is
consistent with the view of the CASAC PM Panel. Taking into account the estimated risk
reductions based on the use of either cutpoint, staff finds further support for considering an
annual PM25 standard in the range of 14 to 12 |ig/m3. Alternatively, staff also finds support for a
revised 24-hour standard, in conjunction with retaining the current annual standard, in the range
of 35 to 25 |ig/m3, in conjunction with a 99th percentile form especially with a standard level in
the middle to upper end of this range or with a 98th percentile form with a standard level in the
middle to lower end of this range. Staff notes that a 24-hour standard at a level of 40 |ig/m3 is
estimated to provide no additional protection against the serious health effects associated with
long-term PM2 5 exposures in two or three of the five example cities (for a 99th or 98th percentile
form, respectively) relative to that afforded by the current annual PM2 5 standard, regardless of
the weight that is given to the alternative assumed cutpoints in the range considered by staff.
Staff believes that a suite of PM25 standards selected from the alternatives identified above could
provide an appropriate degree of protection against the mortality and morbidity effects
associated with long-term exposure to PM25 in studies in areas across the U.S..
5.3.5 Alternative PM2 5 Standards to Address Health Effects Related to Short-term
Exposure
In considering alternative PM2 5 standards that would provide protection against health
effects related to short-term exposures, staff has similarly taken into account both evidence-
based and risk-based considerations. As discussed below in this section, staff has first evaluated
the available evidence from short-term exposure studies, as well as the uncertainties and
limitations in that evidence, to assess the degree to which alternative 24-hour and/or annual
PM2 5 standards can be expected to provide protection against effects related to short-term
exposures. Secondly, staff has considered the quantitative risk estimates for short-term exposure
effects, discussed in Chapter 4, to assess the extent to which alternative annual and/or 24-hour
standards can be expected to reduce the estimated risks attributable to short-term exposure to
PM2 5. Staff conclusions as to ranges of alternative annual and/or 24-hour standards that would
provide protection against health effects related to short-term exposures are summarized at the
end of this section. As noted above, the integrated staff recommendations presented in section
5.3.7 are based in part on the conclusions from this section and in part on staff conclusions from
the previous section, in which alternative PM2 5 standards to address health effects related to
long-term exposures are assessed.
5.3.5.1 Evidence-based Considerations
In taking into account evidence-based considerations, staff has evaluated the available
evidence from short-term exposure studies, as well as the uncertainties and limitations in that
5-29
-------
evidence. In so doing, staff has focused on U.S. and Canadian short-term exposure studies of
fine particles (Appendix 3 A). We took into account reanalyses that addressed GAM-related
statistical issues and considered the extent to which the studies report statistically significant and
relatively precise relative risk estimates; the extent to which the reported associations are robust
to co-pollutant confounding and alternative modeling approaches; and the extent to which the
studies used relatively reliable air quality data. In particular, staff has focused on those specific
studies, identified above in section 5.3.1, that provide evidence of associations in areas that
would have met the current annual and 24-hour PM2 5 standards during the time of the study.
Staff believes that this body of evidence can serve as a basis for 24-hour and/or annual PM25
standards that would provide increased protection against effects related to short-term exposures.
As an initial matter, staff recognizes, as discussed above, that these short-term exposure
studies provide no evidence of clear thresholds, or lowest-observed-effects levels, in terms of 24-
hour average concentrations. Staff notes that of the two PM2 5 studies that explored potential
thresholds, one study in Phoenix provided some suggestive evidence of a threshold possibly as
high as 20 to 25 |ig/m3, whereas the other study provided evidence suggesting that if a threshold
existed, it would likely be appreciably below 25 |ig/m3. While there is no evidence for clear
thresholds within the range of air quality observed in the epidemiologic studies, for some health
endpoints (such as total nonaccidental mortality) it is likely to be extremely difficult to detect
threshold levels (CD, p. 9-45). As a consequence, this body of evidence is difficult to translate
directly into a specific 24-hour standard that would independently protect against all effects
associated with short-term exposures. Staff notes that the distributions of daily PM2 5
concentrations in these studies often extend down to or below typical background levels, such
that consideration of the likely range of policy-relevant background concentrations across the
U.S., as discussed in Chapter 2, section 2.6, becomes important in identifying a lower bound of a
range of 24-hour standards appropriate for consideration.
Being mindful of the difficulties posed by issues relating to threshold and background
levels, staff has first considered this short-term exposure epidemiologic evidence as a basis for
alternative 24-hour PM25 standards. In so doing, staff has focused on the upper end of the
distributions of daily PM25 concentrations, particularly in terms of the 98th and 99th percentile
values, reflecting the form of the current 24-hour standard and an alternative form considered in
the risk assessment, respectively. In looking at the specific studies identified in section 5.3.1 that
report statistically significant associations in areas that would have met the current PM2 5
standards, including studies in Phoenix (Mar et al., 1999, 2003), Santa Clara County, CA
(Fairley, 1999, 2003) and eight Canadian cities (Burnett et al., 2000 and Burnett and Goldberg,
2003), staff notes that the 98th percentile values range from approximately 32 to 39 |ig/m3 in
Phoenix and the eight Canadian cities, up to 59 |ig/m3 in Santa Clara Country; 99th percentile
values range from 34 to 45 |ig/m3 in Phoenix and the eight Canadian cities, up to 69 |ig/m3 in
Santa Clara Country. These ranges also encompass the 98th and 99th percentile values from the
5-30
-------
short-term exposure studies that reported positive PM-related effects and have long-term mean
PM25 concentrations at and somewhat above the current annual PM25 standard [up to 18 |ig/m3,
as summarized in Ross and Langstaff (2005)]. Based on this information, staff believes that the
range of alternative 24-hour PM2 5 standards appropriate for consideration should extend below
the ranges of 98th and 99th percentile values reported in the studies identified above, so as to
provide protection from the short-term exposure effects seen in these studies.
Since the available epidemiologic evidence provides no clear basis for identifying the
lower end of the range of consideration for a 24-hour standard level, staff has looked to the
information on background concentrations, recognizing that a standard intended to provide
protection from man-made pollution should be set above background levels. As discussed in
Chapter 2, section 2.6, staff notes that long-term average PM25 daily background levels are quite
low (ranging from 1 to 5 |ig/m3 across the U.S.), although the upper end (99th percentile values)
of daily distributions of background levels are estimated to extend from approximately 10 to
20 |ig/m3 in regions across the U.S, although such levels may include some undetermined
contribution from anthropogenic emissions (Langstaff, 2004). Even higher daily background
levels result from episodic occurrences of extreme natural events (e.g., wildfires, dust storms),
but levels related to such events are generally excluded from consideration under EPA's natural
events policy, as noted in section 2.6. Based on consideration of these background levels, staff
believes that 25 |ig/m3 is an appropriate lower end to the range of 24-hour PM25 standards for
consideration in this review. Thus, based on this evidence, staff concludes it is appropriate to
consider alternative 24-hour PM2 5 standards, with either a 98th or 99th percentile form, that range
down to as low as 25 |ig/m3 to provide protection from effects associated with short-term
exposures to PM2 5.
As in the last review, staff believes it is also appropriate to consider the evidence
discussed above as a basis for a alternative annual PM2 5 standards that would address risks
associated with short-term exposures. In the last review, annual standard levels were considered
at or somewhat below the long-term mean concentrations in short-term exposure studies
reporting statistically significant associations, recognizing that the evidence of an association in
such studies is strongest at and around this long-term mean, where the data in the study are most
concentrated. This approach follows from the observation that, when aggregated on an annual
basis, much of the risk related to daily exposures results from the large number of days during
which the 24-hour average concentrations are in the low- to mid-range, as discussed in Chapter 4
(section 4.4.4) and in section 5.3.3 above. Thus, to reduce the aggregate risk, it is necessary to
shift the bulk of the distribution to lower levels, not just to limit the concentrations on days when
the PM2 5 concentrations are relatively high. Shifting the distribution can be accomplished
through control strategies aimed at attaining either an annual or 24-hour standard.
Using this approach, the same short-term exposure studies identified above can be
considered as a basis for alternative levels of an annual standard that would provide additional
5-31
-------
protection from effects associated with short-term exposures. In particular, the multi-city
Canadian study (Burnett et al., 2000 and Burnett and Goldberg, 2003) reports statistically
significant associations between short-term PM2 5 exposure and total and cardiovascular
mortality across areas with an aggregate long-term mean PM25 concentration of 13.3 |ig/m3. The
other two studies, conducted in Phoenix (Mar et al., 1999, 2003) and Santa Clara County, CA
(Fairley, 1999, 2003), each had long-term mean PM25 concentrations of approximately 13 |ig/m3.
In considering this evidence, staff concludes that these studies provide a basis for considering an
annual PM25 standard within the range of 13 |ig/m3 to about 12 |ig/m3. An annual standard of
13 |ig/m3 would be consistent with a judgment that appreciable weight should be accorded these
studies as a basis for an annual standard that would protect against PM2 5-related mortality
associated with short-term exposure. An annual standard of 12 |ig/m3, somewhat below the
long-term means in these studies, would be consistent with a judgment that a more precautionary
standard was warranted. Such a standard could potentially reflect consideration of the
seriousness of the mortality effects, for which there is strong evidence of a likely causal
relationship, as well as the much more uncertain evidence of respiratory-related emergency
department visits, discussed above in section 5.3.1, in studies with long-term mean PM2 5
concentrations of approximately 12 |ig/m3 and below. As discussed in Chapter 1 and above in
section 5.3.4.1, these considerations are relevant to judgments about providing an adequate
margin of safety to prevent pollution levels that may pose an unacceptable risk of harm, even if
the risk is not precisely identified as to nature or degree. In staffs view, an annual standard set
below this range would be highly precautionary based on the evidence discussed above, giving
little weight to the remaining uncertainties in the broader body of short-term exposure evidence,
including the possibility of a threshold within the range of air quality in the studies and the
recognition that results may be sensitive to selection of statistical models beyond the range of
models examined in these particular studies.
Consistent with the conclusions reached in the last review (62 FR 38674-7), however,
staff continues to believe that an annual standard cannot be expected to offer an adequate margin
of safety against the effects of all short-term exposures, especially in areas with unusually high
peak-to-mean ratios of PM25 levels, possibly associated with strong local or seasonal sources, or
for potential PM2 5-related effects that may be associated with shorter-than-daily exposure
periods (noted above in section 5.3.3). As a result, in conjunction with an annual standard that
may be adopted in part to provide protection against effects associated with short-term
exposures, staff believes it is appropriate also to consider alternative 24-hour PM2 5 standards as
well. Such a 24-hour standard could reasonably be based on air quality information (from 2001
to 2003) in Chapter 2, Figure 2-23, that shows the distribution of 98th percentile values as a
function of annual means values in urban areas across the U.S. Based on this information, staff
concludes that a 24-hour standard in the range of approximately 40 to 35 |ig/m3 could limit peak
concentrations in areas with relatively high peak-to-mean ratios (i.e., generally in the upper
5-32
-------
quartile to the upper 5th percentile, respectively) and with annual mean concentrations in the
range of 12 to 13 |ig/m3.
5.3.5.2 Risk-based Considerations
Beyond looking directly at the relevant epidemiologic evidence, staff has also considered
the extent to which specific levels and forms of alternative 24-hour and annual PM2 5 standards
are likely to reduce the estimated risks attributable to short-term exposure to PM2 5, and the
uncertainties in the estimated risk reductions. As discussed above (section 5.3.1), staff has based
this evaluation on the risk assessment results presented in Chapter 4, in which short-term
exposure risks were estimated using reported city-specific concentration-response functions
down to policy-relevant background, as well as using modified concentration-response functions
that incorporate alternative assumed cutpoints as surrogates for potential population thresholds.
Figures 5-2(a), (b), (c), and (d) show the estimated percentage reductions in mortality
attributable to short-term exposure to PM2 5 in going from meeting the current PM2 5 standards to
meeting alternative annual and 24-hour PM2 5 standards (with a 98th percentile form) in the five
example cities that do not meet the current standards (based on 2001-2003 air quality data),
based on assumed cutpoints equal to estimated policy-relevant background and 10, 15, and
20 |ig/m3, respectively. To put the estimated percentage reductions in perspective, these figures
also include the estimated PM2 5-related annual incidence rate (in terms of deaths/year/100,000
general population) and annual incidence (in terms of deaths/year) of total mortality associated
with short-term exposure associated with just meeting the current PM25 standards. A similar
series of figures is shown in Appendix 5 A for meeting alternative 24-hour standards with a 99th
percentile form. As in the figures for long-term exposures discussed in section 5.3.4.2, the
alternative annual PM25 standards considered in these figures include a range of levels from 15
to 12 |ig/m3, and attainment of the standards is simulated based on a percent rollback calculated
using the highest monitor in an area, as noted in Tables 5-1 (a) and (b) and discussed in
Chapter 4, section 4.2.2. The alternative 24-hour PM2 5 standards considered in these figures
include a range of levels from 65 to 25 |ig/m3. Further discussion of alternative forms of the
annual and 24-hour standards is presented below in section 5.3.6.
In considering the estimates based on a cutpoint level equal to estimated policy-relevant
background [Figures 5-2(a) and 5A-2(a)], staff first examined the estimated reductions
associated with lower levels of the annual PM2 5 standard, without changing the 24-hour
standard. Staff observes that lowering the annual standard to alternative levels of 14, 13, and
12 |ig/m3 results in small but generally consistent estimated risk reductions of roughly 10 to 15
percent, 15 to 20 percent, and 25 to 30 percent, respectively, across all five example cities.
Thus, for this assumed cutpoint, estimated reductions in mortality associated with short-term
exposure to PM2 5 are no greater than 30 percent in any of the five example cities with changes in
the annual PM2 5 down to a level of 12 |ig/m3. In examining the effect of changes to the 24-hour
and/or annual PM25 standards in this case, staff first notes that the estimated reductions in
5-33
-------
Incidence Rate: 6 (-6 -17) deaths/yr/100,000
Incidence: 120 (-120 - 360) deaths/yr
100
Incidence Rate: 3 (0 -6) deaths/yr/100,000
Incidence: 290 (-40-610) deaths/yr
24-hour
standard (ug/m3)
(ug/m3)
Los Angeles
Incidence Rate: 24 (12 - 37) deaths/yr/100,000
Incidence: 370 (180-560) deaths/yr
100
24-hour
"*• -— J-rd(ug/m3)
Philadelphia
Incidence Rate: 4 (-8 -16) deaths/yr/100,000
Incidence: 50 (-110 -200) deaths/yr.
45
24-hour
65 standard (ug/m3)
Pittsburgh
Incidence Rate: 8 (3 -12) deaths/yr/100,000
Incidence: 190 (70 - 310) deaths/yr
100
Figure 5-2(a) Estimated percent reduction in PM2.s-related short-term mortality risk for
alternative standards (9tf percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint equal to policy-
relevant background). Risk associated with meeting current PM2.5 standards,
based on ACS extended study, is shown in figures in terms of estimated annual
incidence rate and annual incidence (and 95% confidence ranges). Estimated
policy-relevant background is 3.5 |ig/m in eastern cities and 2.5 |ig/m in
western cities.
5-34
-------
Incidence Rate: 3 (-3 - 8) deaths/yr/100,000
Incidence: 50 (-60-160)deaths/yr
100
Estimated
percent
reduction
Incidence Rate: 1 (0-3) deaths/yr/100,000
Incidence: 120 (-10-240)deaths/yr
100
Incidence Rate: 12 (6 -19) deaths/yr/100,000
Incidence: 190 (90 - 260) deaths/yr
100
Estimated
percent
reduction
Incidence Rate: 2 (-4 - 7) deaths/yr/100,000
Incidence: 20 (-50 - 90) deaths/yr
100
24-hour
standard (ug/m3)
Pittsburgh
Incidence Rate: 3 (1 - 5) deaths/yr/100,000
Incidence: 80 (30-120) deaths/yr
100
80
Estimated
percent 6Q
reduction
standard
(ug/m3)
24-hour
standard (ug/m3)
St. Louis
Figure 5-2(b) Estimated percent reduction in PMi.s-related short-term mortality risk for
alternative standards (9tf percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 10 ug/m ). Risk
associated with meeting current PM2.5 standards, based on ACS extended study,
is shown in figures in terms of estimated annual incidence rate and annual
incidence (and 95% confidence ranges).
5-35
-------
Incidence Rate: 1 (-1 - 4) deaths/yr/100,000
Incidence: 30 (-30 - 80) deaths/yr
100
Estimated
percent
reduction
Incidence Rate: 1 (0-1)deaths/yr/100,000
Incidence: 60 (-10-120) deaths/yr
100
40"
24-hour
standard (ug/m3)
Los Angeles
Incidence Rate: 7 (3 -11) deaths/yr/100,000
Incidence: 110 (50-160) deaths/yr
100
80
Estimated
percent 60
reduction
40
20
0
Annual
standard
(ug/m3)
Philadelphia
24-hour
iard (ug/m3)
Incidence Rate: 1 (-2-3) deaths/yr/100,000
Incidence: 10 (-20 - 40) deaths/yr
100
Pittsburgh
24-hour
ndard (ug/m3)
Incidence Rate: 1 (0 - 2) deaths/yr/100,000
Incidence: 30 (10-50) deaths/yr
percent 6Q|
reduction
24-hour
standard (ug/m3)
Figure 5-2(c) Estimated percent reduction in PM2.s-related short-term mortality risk for
alternative standards (9ffh percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 15 ug/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended study,
is shown in figures in terms of estimated annual incidence rate and annual
incidence (and 95% confidence ranges).
5-36
-------
Incidence Rate: 1 (-1 - 2) deaths/yr/100,000
Incidence: 10(-10-40)deaths/yr
standard
(ug/m3)
24-hour
standard (ug/m3)
Detroit
Incidence Rate: 0 (0 -1) deaths/yr/100,000
Incidence: 30 (-4 - 60) deaths/yr
Estimated
percent gQjr
reduction
Incidence Rate: 4 (2 - 6) deaths/yr/100,000
Incidence: 60 (30 - 90) deaths/yr
100
80 -"
Estimated
percent 60
reduction
40
20
0
Annual
standard
(ug/m3)
ft
Philadelphia
30
24-hour
standard (ug/m3)
Incidence Rate: 0 (-1 -1) deaths/yr/100,000
Incidence: 5 (-10-20) deaths/yr
100-1
percent 60
reduction
eo
ss
so
45
to
3S-
6.1
Pittsburgh
24-hour
standard (ug/m3)
Incidence Rate: 0 (0 -1) deaths/yr/100,000
Incidence: 9 (3-14) deaths/yr
100n
24-hour
standard (ug/m3)
(ug/m3)
St. Louis
Figure 5-2(d) Estimated percent reduction in PMi.s-related short-term mortality risk for
alternative standards (9tfh percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 20 ug/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended study,
is shown in figures in terms of estimated annual incidence rate and annual
incidence (and 95% confidence ranges).
5-37
-------
short-term mortality risk associated with changes to the 24-hour standard are generally larger
and much more variable across cities than with changes in just the annual standard. Further, no
combination of standards within the ranges that staff has considered results in the elimination of
all estimated mortality risk associated with short-term exposure in all example cities. More
specifically, a 24-hour standard of 25 |ig/m3 results in estimated reductions in short-term
mortality ranging from approximately 30 to 50 percent (98th percentile form) and 35 to 70
percent (99th percentile form) across the five cities in conjunction with any annual standard in the
range of 15 to 12 |ig/m3. A 24-hour standard of 30 |ig/m3 results in estimates of reductions in
short-term mortality ranging from approximately 25 to 35 percent (98th percentile form) and 25
to 65 percent (99th percentile form) across the five cities in conjunction with an annual standard
of 12 |ig/m3; the lower end, but not the upper end, of these ranges decreases somewhat in
conjunction with annual standards from 13 to 15 |ig/m3. As in the assessment of risk related to
long-term exposures discussed in section 5.3.4.2, this assessment indicates that 24-hour
standards of 30 to 25 |ig/m3 become generally controlling standards in most cases within this
range of annual standards.
In considering the estimates of risk reductions based on an assumed cutpoint of 10 |ig/m3
[Figures 5-2(b) and 5A-2(b)], staff observes that lowering the annual standard to alternative
levels of 14, 13, and 12 |ig/m3 (without changing the 24-hour standard) results in estimated risk
reductions of roughly 15 to 25 percent, 30 to 35 percent, and 45 to 55 percent, respectively,
across all five example cities. In considering changes to the 24-hour and/or annual PM2 5
standards in this case, staff notes that a 24-hour standard of 25 |ig/m3 results in estimates of
reductions in short-term mortality ranging from approximately 45 to 80 percent (98th percentile
form) and 60 to 95 percent (99th percentile form) across the five cities in conjunction with any
annual standard in the range of 15 to 12 |ig/m3. A 24-hour standard of 30 |ig/m3 results in
estimates of reductions in short-term mortality ranging from approximately 45 to 60 percent (98th
percentile form) and 50 to 95 percent (99th percentile form) across the five cities in conjunction
with an annual standard of 12 |ig/m3; as with the previous case (based on a cutpoint equal to
policy-relevant background), the lower end, but not the upper end, of these ranges decreases
appreciably in conjunction with annual standards from 13 to 15 |ig/m3. Thus, in this case, as in
the previous case, changes in the 24-hour standard, while retaining the current annual standard,
can result in generally larger but much more variable estimated reductions in risks associated
with short-term exposures across the five cities than with changes in just the annual standard.
Further, in considering assumed cutpoints of 15 or 20 |ig/m3, staff observes that lowering
the annual standard to alternative levels of 14, 13, and 12 |ig/m3 (without changing the 24-hour
standard) results in estimated risk reductions of roughly 20 to 45 percent, 40 to 65 percent, and
60 to 90 percent, respectively, across all five example cities. In considering changes to the 24-
hour and/or annual PM2 5 standards in these cases, staff notes that a 24-hour standard of 25 |ig/m3
results in estimates of reductions in short-term mortality ranging from approximately 60 to
5-38
-------
100 percent (98th percentile form) and 70 to 100 percent (99th percentile form) across the five
cities in conjunction with any annual standard in the range of 15 to 12 |ig/m3. A 24-hour
standard of 30 |ig/m3 results in estimates of reductions in short-term mortality ranging from
approximately 60 to 90 percent (98th percentile form) and 60 to 100 percent (99th percentile form)
across the five cities in conjunction with an annual standard of 12 |ig/m3; similarly, the lower
end, but not the upper end, of these ranges decreases appreciably in conjunction with annual
standards from 13 to 15 |ig/m3. Thus, in these cases as well, changes in the 24-hour standard,
while retaining the current annual standard, can result in generally larger but much more variable
estimated reductions in risks associated with short-term exposures across the five cities than with
changes in just the annual standard.
5.3.5.3 Summary
In considering the relevant epidemiologic evidence, estimates of risk reductions
associated with alternative annual and/or 24-hour standards, and the 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
short-term exposures. In looking specifically at the evidence of associations between short-term
exposure to PM2 5 and serious health effects, with a particular focus on mortality associations,
staff concludes that it is appropriate to consider a revised 24-hour standard in the range of 30 to
25 |ig/m3 in conjunction with retaining the current annual standard level of 15 |ig/m3.
Alternatively, staff also believes the evidence supports consideration of a revised annual
standard, in the range of 13 to 12 |ig/m3, in conjunction with a revised 24-hour standard in the
range of 40 to 35 |ig/m3.
In considering the results of the quantitative risk assessment, staff believes that it is
appropriate to consider all the estimates associated with the range of assumed cutpoints used in
the risk assessment. As discussed above in sections 5.3.1.3 and 5.3.4.3, staff believes that a
relatively more precautionary approach to interpreting this evidence would give more weight to
the estimates based on the lowest cutpoint considered, while giving more weight to the estimates
based on an assumed cutpoint of 10 |ig/m3 is consistent with the view of the CASAC PM Panel.
Taking into account the estimated risk reductions based on the use of either cutpoint, staff finds
additional support for considering a revised 24-hour standard in the range of 30 to 25 |ig/m3 in
conjunction with retaining an annual standard level of 15 |ig/m3. In either case, a 24-hour
standard at a level of 35 |ig/m3 is estimated to provide less than 30 percent reduction in mortality
incidence in two or three of the five example cities (for a 99th or 98th percentile form,
respectively) relative to that afforded by the current annual PM2 5 standard alone. Further, in
conjunction with a lower annual standard down to 12 |ig/m3, staff finds support for considering a
revised 24-hour standard in the range of 35 to 30 |ig/m3. Staff finds little support based on the
risk assessment for addressing short-term exposure effects solely with a revised annual standard
in a range down to 12 |ig/m3. Staff believes that a suite of PM25 standards selected from the
5-39
-------
alternatives identified above could provide an appropriate degree of protection against the
mortality and morbidity effects associated with short-term exposure to PM2 5 in studies 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 2001 to 2003 air quality) the
percentage of counties, and the population in those counties, that would not likely attain various
PM25 annual standards alone in comparison to the percentage of counties that would not likely
attain alternative combinations of annual and 24-hour PM2 5 standards. This assessment, shown
in Appendix 5B (Tables 5B-l(a) and (b), for 98th and 99th percentile forms of the 24-hour
standards, respectively), was not considered as a basis for the above staff conclusions.
5.3.6 Alternative Forms for Annual and 24-Hour PM2 5 Standards
5.3.6.1 Form of Annual Standard
In 1997 EPA established the form of the annual PM25 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
PM25 concentrations. The arithmetic mean serves to represent the broad distribution of daily air
quality values, and a 3-year average provides a more stable risk reduction target than a single-
year annual average. The current annual PM2 5 standard level is to be compared to measurements
made at the community-oriented monitoring site recording the highest level, or, if specific
constraints are met, measurements from multiple community-oriented monitoring sites may be
averaged (62 FR at 38,672). The constraints on allowing the use of spatially averaged
measurements were intended to limit averaging across poorly correlated or widely disparate air
quality values. This approach was judged to be consistent with the epidemiologic 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 this review, in conjunction with recommending that consideration be given to
alternative annual standard levels, staff is also reconsidering the appropriateness of continuing to
allow spatial averaging across monitors as part of the form of an annual standard. There now
exist much more PM2 5 air quality data than were available in the last review. Consideration of
the spatial variability across urban areas that is revealed by this new database (see Chapter 2,
section 2.4 above, and the CD Chapter 3, section 3.2.5) raises questions as to whether an annual
standard that allows for spatial averaging, within currently specified or alternative constraints,
would provide appropriate public health protection. In conducting analyses to assess these
questions, as discussed below, staff has taken into account both aggregate population risk across
an entire urban area and the potential for disproportionate impacts on potentially vulnerable
subpopulations within an area.
5-40
-------
The effect of allowing the use of spatial averaging on aggregate population risk was
considered as part of the sensitivity analyses included in the health risk assessment discussed in
Chapter 4. In particular, a sensitivity analysis was done in several example urban areas (Detroit,
Pittsburgh, and St. Louis) that compared estimated mortality risks (associated with both long-
and short-term exposures) based on calculating compliance using air quality values from the
highest community-oriented monitor in an area, with estimated risks based on using air quality
values averaged across all such monitors within the constraints allowed by the current standard.
As discussed in Chapter 4, section 4.2.2, the monitored air quality values were used to determine
the design value for the annual standard in each area, as applied to a "composite" monitor to
reflect area-wide exposures. Changing the basis of the annual standard design value from the
concentration at the highest monitor to the average concentration across all monitors reduces the
air quality adjustment needed to just meet the current or alternative annual standards. As
expected, the estimated risks remaining upon attainment of the current annual standard are
greater when spatial averaging is used than when the highest monitor is used (i.e., the estimated
reductions in risk associated with just attaining the current or alternative annual standards are
less when spatial averaging is used). Based on the results of this analysis in the three example
cities, estimated mortality incidence associated with long-term exposure based on the use of
spatial averaging is about 10 to over 40% higher than estimated incidence based on the use of the
highest monitor. For estimated mortality incidence associated with short-term exposure, the use
of spatial averaging results in risk estimates that range from about 5 to 25% higher. In
considering estimated risks remaining upon attainment of alternative suites of annual and 24-
hour PM2 5 standards, spatial averaging only has an impact in those cases where the annual
standard is the "controlling" standard. For such cases in the three example cities, the estimated
mortality incidence associated with long-term exposure in most cases ranges from about 10 to
60% higher when spatial averaging is used, and estimated mortality incidence associated with
short-term exposure in most cases ranges from about 5 to 25% higher.
In considering the potential for disproportionate impacts on potentially vulnerable
subpopulations, staff has assessed whether any such groups are more likely to live in census
tracts in which the monitors recording the highest air quality values in an area are located. Data
were obtained for demographic parameters measured at the census tract level, including
education level, income level, and percent minority. These data from the census tract in which
the highest air quality values were monitored were compared to area-wide average values
(Schmidt et al., 2005). Recognizing the limitations of such cross-sectional analyses, staff
observes that the results suggest 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 percentage minority levels. Staff notes that some
epidemiologic study results, most notably the associations between mortality and long-term
PM2 5 exposure in the ACS cohort, have shown larger effect estimates in the cohort subgroup
5-41
-------
with lower education levels (CD, p. 8-103). As discussed in Chapter 3, section 3.4, people with
lower socioeconomic status (e.g., lower education and income levels), or who have greater
exposure to sources such as roadways, may have increased vulnerability to the effects of PM
exposure. Combining evidence from health studies suggesting that people with lower
socioeconomic status may be considered a population more vulnerable to PM-related effects
with indications from air quality analyses showing that higher PM2 5 concentrations are measured
in local communities with lower socioeconomic status, staff finds that this is additional evidence
which supports a change from spatial averaging across PM2 5 monitors to provide appropriate
protection from public health risks associated with exposure to ambient PM2 5.
The allowance to use spatial averaging under certain constraints established in 1997 was
intended to provide for a relatively stable measure of air quality and to characterize area-wide
PM2 5 concentrations, while also precluding averaging across monitors that would leave a portion
of a metropolitan areas with substantially greater exposures than other areas (62 FR 38672).
Based on the PM2 5 air quality data now available, staff believes that the existing constraints on
spatial averaging may not be adequate to avoid substantially greater exposures in some areas,
potentially resulting in disproportionate impacts on potentially vulnerable subpopulations.
Thus, in considering whether alternative constraints on the use of spatial averaging may be
appropriate, staff has analyzed existing data on the correlations and differences between monitor
pairs in metropolitan areas (Schmidt et al., 2005). For all pairs of PM25 monitors, the median
correlation coefficient based on annual air quality data is approximately 0.9, which is
substantially higher than the current criterion for correlation of at least 0.6, which was met by
nearly all monitor pairs. Similarly, the current criterion that differences in mean air quality
values between monitors not exceed 20% was met for most monitor pairs, while the annual
median and mean differences for all monitor pairs are 5% and 8%, respectively. This analysis
also showed that in some areas with highly seasonal air quality patterns (e.g., due to seasonal
woodsmoke emissions), substantially lower seasonal correlations and larger seasonal differences
can occur relative to those observed on an annual basis.
In considering the results of the analyses discussed above, staff concludes that it is
appropriate to consider eliminating the provision that allows for spatial averaging from the form
of an annual PM25 standard. Further, staff concludes that if consideration is given to retaining an
allowance for spatial averaging, more restrictive criteria should be considered. Staff believes
that it would be appropriate to consider alternative criteria such as a correlation coefficient of at
least 0.9, determined on a seasonal basis, with differences between monitor values not to exceed
about 10%.
5.3.6.2 Form of 24-Hour Standard
In 1997 EPA established the form of the 24-hour PM25 standard as the 98th percentile of
24-hour concentrations at each population-oriented monitor within an area, averaged over three
years (62 FR at 38671-74). EPA selected such a concentration-based form because of its
5-42
-------
advantages over the previously used expected-exceedance form.6 A concentration-based form is
more reflective of the health risk posed by elevated PM25 concentrations because it 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 standard. Further, a
concentration-based form better compensates for missing data and less-than-every-day
monitoring; and, when averaged over 3 years, it has greater stability and, thus, facilitates the
development of more stable implementation programs. After considering a range of
concentration percentiles from the 95th to the 99th, EPA selected the 98th percentile as an
appropriate balance between adequately limiting the occurrence of peak concentrations and
providing increased stability and robustness. Further, by basing the form of the standard on
concentrations measured at population-oriented monitoring sites (as specified in 40 CFR part
58), EPA intended to provide protection for people residing in or near localized areas of elevated
concentrations.
In this review, in conjunction with recommending that consideration be given to
alternative 24-hour standard levels, staff is also considering the appropriateness of
recommending that the current 98th percentile form, averaged over 3 years, be retained or
revised. As an initial matter, staff believes that it is appropriate to retain a concentration-based
form that is defined in terms of a specific percentile of the distribution of 24-hour PM25
concentrations at each population-oriented monitor within an area, averaged over 3 years. Staff
bases this recommendation on the same reasons that were the basis for EPA's selection of this
type of form in the last review. As to the specific percentile value to be considered, staff has
narrowed the focus of this review to the 98th and 99th percentile forms. This focus is based on the
observation that the current 98th percentile form already allows the level of the standard to be
exceeded seven days per year, on average (with every-day monitoring), while potentially
allowing many more exceedance days in the worst year within the 3-year averaging period
(Schmidt et al., 2005). As a result, in areas that just attain the standards, EPA's communication
to the public through the Air Quality Index will on one hand indicate that the general level of air
quality is satisfactory (since the standards are being met), but on the other hand it may identify
many days throughout the year as being unhealthy, particularly for sensitive groups. Thus, staff
does not believe it would be appropriate to consider specifying the form in terms of an even
lower percentile value.
In considering differences between 98th and 99th percentile forms, staff believes it is
appropriate to take into consideration the relative risk reduction afforded by these alternative
forms at the same standard level. Based on the risk assessment results discussed in Chapter 4,
6 The form of the 1987 24-hour PM10 standard is based on the expected number of days per year (averaged
over 3 years) on which the level of the standard is exceeded; thus, attainment with the one-expected exceedance
form is determined by comparing the fourth-highest concentration in 3 years with the level of the standard.
5-43
-------
and the risk reductions associated with alternative levels and forms discussed above in sections
5.3.4 and 5.3.5, staff notes that the 99th percentile can, in some instances, result in appreciably
greater risk reductions in particular areas than that associated with a standard at the same level
but with a 98th percentile form. More specifically, staff considered the differences in risk
reductions associated with attaining alternative standards with 98th and 99th percentile forms in
five example urban areas which do not meet the current annual standard (Detroit, Los Angeles,
Philadelphia, Pittsburgh, and St. Louis). In looking at estimated risk reductions associated with
meeting a 24-hour standard of 30 |ig/m3, for example, estimated risk reductions for mortality
associated with long-term exposures were higher with the use of a 99th percentile form in some
areas by approximately 15%, ranging up to over 50% higher in Los Angeles. For estimated risk
reductions for mortality associated with short-term exposures, the use of a 99th percentile form
resulted in estimated reductions that were higher by less than 10% to over 30% across the five
urban areas.
Staff also analyzed the available air quality data from 2001 to 2003 to compare the 98th
and 99th percentile forms in terms of the numbers of days that would be expected to exceed the
level of the standard (on average over 3 years and in the worst year within a 3-year averaging
period) and by how much the standard would typically be exceeded on such days (Schmidt et al.,
2005). In so doing, as noted above, staff observes that the current 98th percentile form allows the
level of the standard to be exceeded seven days per year, on average (with every-day
monitoring), and finds that this form allows up to about 20 days in the worst year within the 3-
year averaging period. A 99th percentile form would allow the level of the standard to be
exceeded three days per year, on average (with every-day monitoring), while allowing up to
about 13 days in the worst year within the 3-year averaging period. Further, staff observes that
for either form, daily peak concentrations in the upper 1 to 2% of the annual air quality
distributions are within 5 |ig/m3 of the 98th or 99th percentile value somewhat more than half the
time and are almost always within 10 to 15 |ig/m3 above the 98th or 99th percentile values, with
very few excursions above this range.7
Based on these considerations, staff recommends either retaining the 98th percentile form
or revising it to be based on the 99th percentile air quality value. In selecting between these
alternative forms, staff believes primary consideration should be given to the degree of risk
reduction likely to result from the combination of the form and the level of a standard. Staff also
7 This analysis also looked at the number of days in which the reported air quality values were "flagged" as
being heavily influenced by natural events (including forest fires, dust storms) or exceptional events, for which the
Agency's natural and exceptional events policies would likely apply. While flagged days generally account for less
than 1% of all reported 24-hour average PM25 concentrations, they account for about 40% of the highest 100 days
across the country. In looking at the reported values that are above the 99th or 98th percentiles of the distribution of
values, approximately 3 to 6% of the highest 2% of days (above the 98th percentile) were flagged, and approximately
5 to 10% of the highest 1% of days (above the 99th percentile) were flagged.
5-44
-------
believes it is appropriate to take into account whether the 24-hour standard is set so as to add to
the protection afforded by a revised annual standard or is intended to be the primary basis for
providing protection against effects associated with short-term exposures. In choosing between
forms of alternative standards that provide generally equivalent levels of public health
protection, staff believes it is appropriate to consider the relatively stability of a standard with
either form as well as the implications from a public health communication perspective of the
extent to which either form allows different numbers of days in a year to be above the level of
the standard in areas that attain the standard. In particular, staff notes that the use of a 99th
percentile form would result in a more consistent public health message to the general public in
the context of the wide-spread use of the Air Quality Index.
5.3.7 Summary of Staff Recommendations on Primary PM2 5 NAAQS
Staff recommendations for the Administrator's consideration in making decisions on the
primary PM25 standards, together with supporting conclusions from sections 5.3.1 through 5.3.6,
are briefly summarized below. Staff recognizes 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 risk assessment. In
recommending these alternative suites of primary standards and ranges of levels for
consideration, staff is mindful that the 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 NAAQS be set at
zero-risk levels, but rather at levels that avoid unacceptable risks to public health.
( 1) Consideration should be given to revising the current PM2 5 primary 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 recommendation is based in general on the
evaluation in the CD of the newly available epidemiologic, toxicologic, dosimetric, and
exposure-related evidence, and more specifically on the evidence of mortality and
morbidity effects in areas where the current standards were met, together with judgments
as to the public health significance of the estimated incidence of effects upon just
attaining the current standards.
( 2) The indicator for fine particle standards should continue to be PM2 5. This
recommendation is based on the conclusion that the available evidence does not provide
a sufficient basis for replacing or supplementing a mass-based fine particle indicator with
an indicator for any specific fine particle component or subset of fine particles, nor does
it provide a basis for excluding any components; on the evaluation in the CD of air
quality within the intermodal particle size range of 1 to 3 jim; and on the policy judgment
5-45
-------
made in the last review to place regulatory importance on defining an indicator that
would more completely capture fine particles under all conditions likely to be
encountered across the U.S., while recognizing that some limited intrusion of small
coarse particles will occur in some circumstances. Consideration should be given to
modifying the FRM for PM2 5 based on instrumentation and operational improvements
that have been made since the PM25 monitoring network was deployed in 1999, and to
the adoption of FEMs for appropriate continuous measurement methods.
( 3) Averaging times for PM2 5 standards should continue to include annual and 24-hour
averages to protect against health effects associated with short-term (hours to days) and
long-term (seasons to years) exposure periods. Consideration of other averaging times,
especially on the order of one or more hours, was limited by a lack of adequate
information at this time.
( a) Consideration should be given to revising the form of the annual standard to one
based on the highest community-oriented monitor in an area or, alternatively, to
one with more constrained requirements for the use of spatial averaging across
community-oriented monitors.
( b) Consideration should be given to revising the form of the 24-hour standard to a
99th percentile form or, alternatively, to retaining the 98th percentile form, based in
part on considering the degree of risk reduction likely to result from the
combination of the form and the level of a standard.
(4) Consideration should be given to alternative suites of PM25 standards to provide
protection against effects associated with both long- and short-term exposures, taking
into account both evidence-based and risk-based considerations. Integrated
recommendations on ranges of alternative suites of standards that, when considered
together, protect against effects associated with both long- and short-term exposures
include:
( a) Staff recommends consideration of an annual PM25 standard at the current level
of 15 |ig/m3 together with a revised 24-hour PM25 standard in the range of 35 to
25 |ig/m3, based a 98th percentile form for a standard set at the middle to lower
end of this range, or a 99th percentile form for a standard set at the middle to upper
end of this range. Staff judges that such a suite of standards could provide an
appropriate degree of protection against serious mortality and morbidity effects
associated with long- and short-term exposures to fine particles.
(b) Alternatively, staff also recommends consideration of a revised annual PM2 5
standard, within the range of 14 to 12 |ig/m3, together with a revised 24-hour
PM2 5 standard in the range of 30 to 40 |ig/m3. Staff judges that a suite of
5-46
-------
standards that includes either the annual or the 24-hour standard, or both, set at
the middle to lower end of these ranges could provide an appropriate degree of
protection against serious mortality and morbidity effects associated with long-
and short-term exposures to fine particles.
5.4 THORACIC COARSE PARTICLE STANDARDS
5.4.1 Adequacy of Current PM10 Standards
In 1997, in conjunction with establishing new PM2 5 standards, EPA determined that the
new function of PM10 standards was to protect against potential effects associated with thoracic
coarse particles in the size range of 2.5 to 10 |im (62 FR 38,677). Although staff had given some
consideration to a more narrowly defined indicator that did not include fine particles (e.g.,
PM10_2 5), EPA decided to continue to use PM10 as the indicator for standards to control thoracic
coarse particles. 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
PM10 in areas where the coarse fraction was the dominant component of PM10, namely two
fugitive dust studies in areas that substantially exceeded the PM10 standards (62 FR 38,679). The
decision also reflected the fact that there were only very limited ambient air quality data then
available specifically on thoracic coarse particles, in contrast to the extensive monitoring
network already in place for PM10. In essence, EPA concluded at that time that it was
appropriate to continue to control thoracic coarse particles, but that the only information
available upon which to base such standards was indexed in terms of PM10.8
In the present review, staff has taken into account the information now available from a
growing, but still limited, body of evidence on health effects associated with thoracic coarse
particles from studies that use PM10_25 as a measure of thoracic coarse particles. In addition, staff
notes that there is now much more information available to characterize air quality in terms of
estimated PM10_2 5 than was available in the last review.9 In considering this information, staff
now finds that the major considerations that formed the basis for EPA's 1997 decision to retain
PM10 as the indicator for thoracic coarse particles, rather than a more narrowly defined indicator
that does not include fine particles, no longer apply. In particular, staff concludes that the
As discussed in Chapter 1, however, in subsequent litigation regarding the 1997 PM NAAQS revisions,
the court held in part that PM10 is a "poorly matched indicator" for thoracic coarse particles in the context of a rule
that also includes PM2 5 standards because PM10 includes PM2 5. American Trucking Associations v. EPA. 175 F. 3d.
at 1054. Although the court found "ample support" (id.) for EPA's decision to regulate thoracic coarse particles, it
vacated the 1997 revised PM10 standards for that reason.
9 As noted in section 2.5.3, coarse particle concentrations in EPA's monitoring network are currently
estimated, not measured directly, using a difference method in locations with same-day data from co-located PM10
and PM2 5 FRM monitors, resulting in air quality characterizations that are more uncertain than those available for
PM25orPM10.
5-47
-------
continued use of PM10 as an indicator for standards intended to protect against health effects
associated with thoracic coarse particles is no longer appropriate since information is now
available that supports the use of a more directly relevant indicator, PM10_2 5. Further, staff
concludes that continuing to rely principally on health effects evidence indexed by PM10 is no
longer appropriate since more directly relevant studies, indexed by PM10_2 5, are also now
available. Thus, quite aside from any legal considerations, staff finds that it is appropriate to
revise the current PM10 standards in part by revising the indicator for thoracic coarse particles,
and by basing any such revised standard principally on the currently available evidence and air
quality information indexed by PM10_2 5, but also considering evidence from studies using PM10
in locations where PM10_2 5 is the predominant fraction.
Staff has also considered whether the currently available evidence and information
support consideration of standards for thoracic coarse particles that afford either a similar or
notably different degree of public health protection compared to that afforded by the current
PM10 standards. In so doing, staff first focused on dosimetric and toxicologic evidence, then on
relevant findings from epidemiologic studies, followed by consideration of risk-based
information, as discussed below.
Dosimetric evidence formed the primary basis for initial development of the PM10
indicator. While considerable advances have been made, the available evidence continues to
support the basic conclusions reached in the 1987 and 1997 reviews of the standards regarding
penetration and deposition of size specific particles; an aerodynamic size of 10 jim remains a
reasonable separation point for particles that penetrate and potentially deposit in the thoracic
regions of the lungs, particularly for the more sensitive case of mouth breathing. As discussed in
Chapter 3, both fine and thoracic coarse particles penetrate to and deposit in the alveolar and
tracheobronchial regions. For a range of typical ambient size distributions, the total deposition
of thoracic coarse particles to the alveolar region can be comparable to or even larger than that
for fine particles. For areas with appreciable coarse particle concentrations, coarse particles
would tend to dominate particle deposition to the tracheobronchial region for mouth breathers
(CD, p. 6-16).
As noted in past reviews (EPA, 1981b, 1996b), deposition of a variety of particle types in
the tracheobronchial region, including resuspended urban dust and coarse-fraction organic
materials, has the potential to affect lung function and aggravate symptoms, particularly in
asthmatics. Of particular note are limited toxicologic studies that found urban road dust can
produce cellular and immunological effects (e.g., Kleinman, et al., 1995; Steerenberg et al.,
2003). In addition, the CD notes that some very limited in vitro toxicologic studies show some
evidence that coarse particles may elicit pro-inflammatory effects (CD, section 7.4.4), as
discussed further in section 3.4.2. The staff assessment of the physicochemical properties and
occurrence of ambient coarse particles (Chapter 2) suggests that both the chemical makeup and
5-48
-------
spatial distribution of coarse particles are likely to be more heterogeneous than fine particles. In
general, however, urban coarse particles can contain all of the components found in more rural
areas, but be contaminated by a number of additional materials, from motor vehicle-related
emissions to metals and transition elements associated with industrial operations. Taken
together, staff believes that the weight of the dosimetric, limited toxicologic, and atmospheric
science evidence lends support to the plausibility of the effects reported in urban epidemiologic
studies, as discussed in Chapter 3 (section 3.4), and provides support for retaining standards for
thoracic coarse particles so as to continue programs to protect public health from such PM10_2 5-
related effects.
Staff has considered also the available epidemiologic evidence of associations between
thoracic coarse particles, as indexed by PM10_2 5, and health endpoints, as well as evidence from
PM10 studies conducted in areas in which the coarse fraction is dominant. As summarized in
Chapter 3 (section 3.4 and Appendix 3-A), several U.S. and Canadian studies now provide
evidence of associations between short-term exposure to PM10_2 5 and various morbidity
endpoints. Three such studies conducted in Toronto (Burnett et al., 1997), Seattle (Sheppard et
al.,1999, 2003), and Detroit (Lippmann et al., 2000; Ito, 2003) report statistically significant
associations between short-term PM10_2 5 exposure and respiratory- and cardiac-related hospital
admissions, and a fourth study (Schwartz and Neas, 2000) conducted in six U.S. cities including
Boston, St. Louis, Knoxville, Topeka, Portage, and Steubenville reports statistically significant
associations with respiratory symptoms in children. The extent to which the results from these
studies are robust to the inclusion of co-pollutants varies depending on the various models used
and the number of co-pollutants included in the models. Staff observes that the morbidity
studies were done in areas in which PM2 5, rather than PM10_2 5, is the predominant fraction of
ambient PM10, such that they are not representative of areas with relatively high levels of
thoracic coarse particles.
The CD found that evidence from health studies on associations between short-term
exposure to PM10_2 5 and mortality was "not as strong" as evidence for associations with PM2 5 or
PM10 but nonetheless was suggestive of associations with mortality (CD, p. 9-32). As described
in Section 3.4, associations between PM10_25 and mortality are similar in magnitude, but less
precise, than those for PM2 5 or PM10. Statistically significant mortality associations were
reported in studies conducted in areas with relatively high PM10_2 5 concentrations, including
Phoenix (Mar et al., 2000, 2003), Coachella Valley, CA (Ostro et al., 2000, 2003), and
Steubenville (as part of the Harvard Six Cities study, Schwartz et al., 1996; Klemm et al., 2003).
In areas with lower PM10_2 5 concentrations, no statistically significant associations were reported
with mortality, though many were positive but not statistically significant.
In addition, some epidemiologic studies that used PM10 and were conducted in areas
where PM10 is typically dominated by coarse fraction particles can provide information relevant
to the evaluation of coarse fraction particles. Such studies include findings of associations with
5-49
-------
hospitalization for cardiovascular diseases in Tucson, AZ (Schwartz, 1997), hospitalization for
COPD in Reno/Sparks, NV (Chen et al., 2000), and medical visits for asthma or respiratory
diseases in Anchorage, AK (Gordian et al., 1996; Choudhury et al., 1997). In addition, a number
of epidemiologic studies have reported significant associations with mortality, respiratory
hospital admissions and respiratory symptoms in the Utah Valley area (e.g., Pope et al., 1989;
1991; 1992). This group of studies provides additional supportive evidence for associations
between coarse fraction particles and health effects, particularly morbidity effects, in areas with
concentrations generally not meeting the PM10 standard levels (all areas except Tucson).
Taken together, staff concludes that the health evidence, including dosimetric,
toxicologic and epidemiologic study findings, supports retaining standards to protect against
effects associated with short-term exposure to thoracic coarse particles. Staff believes that the
substantial uncertainties associated with this limited body of epidemiologic evidence on health
effects related to exposure to PM10_2 5, however, suggests a high degree of caution in interpreting
this evidence, especially at lower levels of ambient particle concentrations as observed in the
morbidity studies discussed above.
Beyond this evidence-based evaluation, staff has also considered the extent to which
PM10_2 5-related health risks estimated to occur at current levels of ambient air quality may be
judged to be important from a public health perspective, taking into account key uncertainties
associated with the estimated risks. Consistent with the approach used to address this issue for
PM2 5-related health risks, discussed above in section 5.3.1.2, staff has considered the results of a
series of base case analyses that reflect in part the uncertainty associated with the form of the
concentration-response functions drawn from the studies used in the assessment, as presented in
Chapter 4, section 4.5.10 Health risks were estimated in these analyses by using the reported
linear or log-linear concentration-response functions as well as modified functions that
incorporate alternative assumed cutpoints as surrogates for potential population thresholds. Such
estimates of risks attributable to short-term exposure to PM10_25 have been developed for Detroit,
Seattle, and St. Louis.11
Table 5-2 summarizes the estimated PM10_2 5-related annual incidence and incidence rates
(in terms of incidence per 100,000 general population) of hospital admissions and respiratory
symptoms (cough) in children associated with short-term exposure in these three example urban
areas. As an initial matter, staff observes that the range of estimates of cardiac-related hospital
admissions in Detroit is generally more than an order of magnitude greater than the range of
10 Uncertainties related to estimated policy-relevant background levels of PM10_2 5 were addressed in a
sensitivity analysis, which showed negligible impact on the risk estimates.
11 This table includes risk estimates for Detroit drawn from Table 4-20 in Chapter 4 and for Seattle and St.
Louis drawn from Exhibits E.33 and E.34, respectively, in the Technical Support Document.
5-50
-------
Table 5-2 Estimated PM10_2 5-related annual incidence of hospital admissions and cough in children associated with short-
term exposure with 2003 air quality
Detroit: hospital admissions for
ischemic heart disease
Seattle: hospital admissions for
asthma (age <65)
St. Louis: days of cough in
children
Annual Incidence
and 95% Cl
(events/yr)
Cutpoints
Policy-relevant
Background*
650
170-1,100
30
0-70
27,000
11,000-41,000
10 |jg/m3
570
150-930
10
0-20
12,000
4900-18,000
15|jg/m3
490
130-790
5
0-10
5,800
2,500 - 8,600
20 |jg/m3
430
120-680
2
0-4
2,900
1,300-4,000
Annual Incidence Rate
and 95% Cl
(events/yr/1 00,000 general population)
Cutpoints
Policy-relevant
Background*
32
8-53
2
0-4
1,070
440-1,600
10 |jg/m3
28
7-45
1
0-2
480
190-720
15|jg/m3
24
6-39
0
0-1
230
100-340
20 |jg/m3
21
6-33
0
0-0
120
50-160
* Estimated policy-relevant background levels are 4.5 |jg/m3 for eastern urban areas and 3.5 |jg/m3 for western urban areas.
5-51
-------
estimated asthma-related admissions in Seattle, which can be attributed in part to differences in
baseline risks related to cardiovascular- and respiratory-related health endpoints as well as to
differences in PM10_2 5 air quality levels in these two areas. To provide some context for
considering these risks estimates, staff notes that Detroit and St. Louis did not meet the current
24-hour PM10 standard of 150 |ig/m3 based on 2001 to 2003 air quality data (with 24-hour PM10
design values of 191 and 224 |ig/m3, respectively), whereas Seattle, with much lower daily
concentrations (with a 24-hour PM10 design value of 73 |ig/m3), meets the current PM10
standards.12 More specifically, in considering the risk estimates based on the lowest cutpoint
considered, the point estimate of annual incidence of PM10.25-related hospital admissions for
ischemic heart disease in Detroit is approximately 650 events per year (roughly 32 events per
year per 100,000 general population), and the estimate of days of cough in children is
approximately 27,000 days per year (over 1,000 days per year per 100,000 general population,
which would be roughly an order of magnitude higher in terms of days per year per 100,000
children). In considering the estimated incidences based on an assumed cutpoint of 10 |ig/m3,
staff observes that these estimates are about 15 percent lower in Detroit and over 50 percent
lower in St. Louis, whereas at the highest cutpoint considered, the estimates are about 35 percent
lower in Detroit and close to 90 percent lower in St. Louis.
Beyond the specific health endpoints presented in Table 5-2, staff notes that hundreds of
additional hospital admissions for other cardiac- and respiratory-related diseases are also
estimated in Detroit, based on risk assessment results presented in Chapter 4 (across the range of
cutpoints considered), as are thousands of additional days in which children are likely to
experience other lower respiratory tract symptoms in St. Louis. In considering these limited
ranges of estimates, staff concludes that they are indicative of risks that can reasonably be judged
to be important from a public health perspective, in contrast to the appreciably lower respiratory
morbidity risks estimated in Seattle.
In summary, staff recognizes that the substantial uncertainties associated with the limited
available epidemiologic evidence present inherent difficulties in interpreting the evidence for
purposes of setting appropriate standards for thoracic coarse particles. Nonetheless, in
considering the available evidence, the public health implications of estimated risks associated
with current levels of air quality, and the related limitations and uncertainties, staff concludes
that this information supports consideration of standards to provide public health protection from
morbidity effects and possibly mortality associated with current levels of short-term exposure to
thoracic coarse particles in some urban areas. Staff conclusions and recommendations for
12 See www.epa.gov/airtrends/pdfs/pmlO design values 2001-2003.pdf for a discussion of how these
design values are calculated, noting in particular that concentrations flagged as natural events (e.g, high winds,
wildfires, volcanic eruptions) or exceptional events (e.g., construction, prescribed burning) are not included in these
calculations and that no regulatory decisions on attainment status have been made at this time based on these data.
5-52
-------
indicators, averaging times, and levels and forms of alternative primary standards for thoracic
coarse particles are discussed in the following sections.
5.4.2 Indicators
In considering an appropriate indicator for a standard intended to afford protection from
health effects associated with exposure to thoracic coarse particles, staff makes the following
observations:
• The most obvious choice for a thoracic coarse particle standard is the size-
differentiated, mass-based indicator used in the epidemiologic studies that
provide the most direct evidence of such health effects, PM10_25.
The upper size cut of a PM10_2 5 indicator is consistent with the dosimetric
evidence discussed above that continues to reinforce the finding from past
reviews that an aerodynamic size of 10 jam is a reasonable separation point for
particles that penetrate to and potentially deposit in the thoracic regions of the
respiratory tract.
The lower size cut of such an indicator is consistent with the choice of 2.5 jim as
a reasonable separation point between fine and coarse fraction particles, based on
consideration of evidence from atmospheric sciences; it is also consistent with the
recommended continued use of PM25 as the indicator for fine particles (as
discussed above in section 5.3.2), while recognizing that it would exclude the tail
of the coarse mode in some locations.
Further, the limited available information is not sufficient to define an indicator
for thoracic coarse particles solely in terms of metrics other than size-
differentiated mass, such as specific chemical components.
• The available epidemiologic evidence for effects of PM10_25 exposure is quite
limited and is inherently characterized by large uncertainties, reflective in part of
the more heterogeneous nature of the spatial distribution and chemical
composition of thoracic coarse particles and the more limited and uncertain
measurement methods that have generally been used to characterize their ambient
concentrations.
In evaluating relevant information from atmospheric sciences, toxicology, and
epidemiology related to thoracic coarse particles, staff notes that there appears to be a clear
distinction between the character and nature of exposures and evidence concerning associated
health effects of coarse particles as found in urban as compared to those found in nonurban and,
more specifically, rural areas. As discussed more fully below, this evidence leads staff to
consider a more narrowly defined indicator for thoracic coarse particles that focuses on thoracic
coarse particles characteristic of urban areas. This is consistent with CASAC's recommendation
5-53
-------
to "qualify the PM10_2 5 standard . . . with a focus on urban areas" where thoracic coarse particles
are influenced by industrial or traffic-associated sources (Henderson, 2005, p. 8). The following
discussion briefly summarizes key observations from the available scientific and technical
information that are most relevant in comparing the potential health effects associated with
thoracic coarse particles in urban and rural settings.
5.4.2.1 Evidence Related to Urban and Rural Thoracic Coarse Particles
The atmospheric sciences and monitoring information in Chapter 2 indicates not only that
exposures to coarse particles tend to be higher in urban areas than in nearby rural locations, but
also that urban coarse particles are enriched by a number of contaminants not commonly found
in natural crustal materials that are typical of rural coarse particles. The elevation of urban
PM10_2 5 levels as compared to those at nearby rural sites indicates that sources located within
urban areas are generally the cause of elevated urban concentrations; conversely, PM10_2 5
concentrations in such urban areas are not largely composed of particles blown in from more
distant regions. Important sources of thoracic coarse particles in urban areas include dense
traffic that suspends significant quantities of road dust, as well as industrial and combustion
sources that contribute to ambient coarse particles both directly and through deposition to soils
and roads (Chapter 2, Table 2-2). It follows that thoracic coarse particles in urban areas would
differ in composition from those in rural areas, being enriched in components from urban
mobile, stationary, and area source emissions.
While detailed composition data are more limited for PM10_2 5 than for PM2 5, available
measurements from some areas as well as studies of road dust components do show a significant
influence of urban sources on urban thoracic coarse particle composition and mass. Although
crustal elements and natural biological materials represent a significant fraction of thoracic
coarse particles in urban areas, both their relative quantity and character may be altered by urban
sources. For example, in industrial cities, primary particle emissions from industrial sources
and resuspended road dust can increase the relative amount of iron, one of the metals that has
been noted as being of some interest in the studies of mechanisms of toxicity for PM, as well as
other industrial process-related and potentially toxic materials such as nickel, cadmium, and
chromium (CD, p. 9-63). Traffic-related activities can grind and resuspend vegetative materials
into forms not as common in more natural areas (Rogge et al., 1993). Studies of urban road
dusts find that levels of a variety of components are increased from traffic as well as from other
anthropogenic urban sources, including products of incomplete combustion (e.g. poly cyclic
aromatic hydrocarbons) from motor vehicle emissions and other sources, brake and tire wear,
rust, salt and biological materials (CD, p. 3D-3). As discussed in Chapter 2, limited ambient
coarse fraction composition data from various comparisons find that metals and sometimes
elemental carbon contribute a greater proportion of thoracic coarse particle mass in urban areas
than in nearby rural areas. In addition, while large uncertainties exist in emissions inventory
data, staff observes that major sources of PM10_25 emissions in the urban counties in which
5-54
-------
epidemiologic studies have been conducted are paved roads and "other" sources (largely
construction), and that such areas also have larger contributions from industrial emissions,
whereas unpaved roads and agriculture are the main sources of PM10_25 emissions nationwide.
Toxicologic studies, although quite limited, support the view that sources of coarse
particles common in urban areas are of greater concern than uncontaminated materials of
geologic origin. As noted above, one major source of urban coarse particles is paved road dust,
and the CD discusses results from a recent study in which road tunnel dust particles had greater
allergic adjuvant activity than several other particle samples, including residual oil fly ash and
diesel exhaust particles, in two animal models of allergy (Steerenberg et al., 2003; CD, p. 7-136-
137). This supports evidence available in the previous PM NAAQS review regarding potential
effects of road dust particles (EPA, 1996b, p. V-70). In contrast, a number of studies have
reported that Mt. St. Helens volcanic ash, an example of natural crustal material of geologic
origin, has very little toxicity in animal or in vitro toxicologic studies (CD, p. 7-216).
A few toxicologic studies have used ambient thoracic coarse particles from
urban/suburban locations (PM10_2 5), and the results suggest that effects can be linked with several
components of PM10_25. As described in more detail in sections 5.4.1 and 3.2, these in vitro
toxicologic studies linked coarse fraction particles with effects including cytotoxicity, oxidant
formation, and inflammatory effects. These studies suggest that several components (e.g.,
metals, endotoxin, other materials) may have roles in various health responses but do not suggest
a focus on any individual component.
Although largely focused on undifferentiated PM10, the series of epidemiologic
observations and toxicologic experiments related to the Utah Valley suggest that directly emitted
(fine and coarse) and resuspended (coarse) urban industrial emissions are of concern. Of
particular interest are area studies spanning a 13-month period when a major source of PM10 in
the area, a steel mill, was not operating. Observational studies found that respiratory hospital
admissions for children were lower when the plant was shut down (Pope et al., 1989). More
recently, a set of toxicologic and controlled human exposure studies have used particles
extracted from filters from ambient PM10 monitors from periods when the plant did and did not
operate. In both human volunteers and animals, greater lung inflammatory responses were
reported with particles collected when the source was operating, as compared to the period when
the plant was closed (CD, p. 9-73). In addition, in some studies it was suggested that the metal
content of the particles was most closely related to the effects reported (CD, p. 9-74). Staff
observes that while peak days in the Utah Valley occur in conditions that enhance fine particle
concentrations, over the long run, over half of the PM10 was in the coarse fraction. The
aggregation of particles collected on the filters during the study period reflect this long-term
composition. At a minimum, the filter-derived particles represent the kinds of industrial
components that would be incorporated in road dusts in the area.
5-55
-------
Taken together, the epidemiologic studies that examine exposures to thoracic coarse
particles generally found in urban environments and to natural crustal materials support the view
that urban thoracic coarse particles are of concern to public health, in contrast to uncontaminated
natural crustal dusts. With respect to the urban results, several recent studies have shown
associations between PM10_2 5 and health outcomes in a number of sites across the U.S.
Associations have been consistently reported with morbidity in urban areas, some of which had
relatively low PM10_2 5 concentrations. For mortality, statistically significant associations have
been reported only for urban areas that have notably higher ambient PM10_2 5 concentrations.
These associations are with short-term exposures to aggregated PM10_2 5 mass, and no
epidemiologic evidence is available on associations with different components or sources of
PM10_25. However, staff observes that the studies have all been conducted in urban areas of the
U.S., and thus reflect effects of thoracic coarse particles generally present in urban environments
from urban sources.
By contrast, recent evidence from epidemiologic studies has suggested that mortality and
possibly other health effects are not associated with thoracic coarse particles from dust storms or
other such wind-related events that result in suspension of natural crustal materials of geologic
origin. The clearest example is provided by a study in Spokane, WA, which specifically
assessed whether mortality was increased on dust-storm days using case-control analysis
methods. The average PM10 level was more than 200 |ig/m3 higher on dust storm days than on
control days, and the authors report no evidence of increased mortality on these specific days
(Schwartz et al., 1999). One caveat of note is the possibility that people may reduce their
exposure to ambient particles on the most dusty days (e.g., Gordian et al., 1996; Ostro et al.,
2000). Nevertheless, the Spokane study provides no suggestion of significant health effects from
uncontaminated natural crustal materials that would typically form a major fraction of coarse
particles in non-urban or rural areas.
Beyond the urban and rural distinctions discussed above, staff has also considered the
extent to which there is evidence of effects with exposure to ambient thoracic coarse particles in
communities predominantly influenced by agricultural or mining sources. For example, in the
last review, staff considered health evidence related to long-term silica exposures from mining
activities, but found that there was a lack of evidence that such emissions contribute to effects
linked with ambient PM exposures (EPA, 1996b, p. V-28). Similarly in this review, there is an
absence of evidence related to such community exposures. While dust generated from
agricultural activity can include biological material such as fungal or bacterial material, and
some occupational studies discussed in the CD report effects at occupational exposure levels
(Table7B-3, p. 7B-11), such studies do not provide relevant evidence for much lower levels of
community exposures. Further, it is unlikely that such sources contribute to the effects that have
been observed in the recent urban epidemiologic studies.
5-56
-------
The CD concludes its integrated assessment of the effects of natural crustal materials as
follows:
Certain classes of ambient particles appear to be distinctly less toxic than others
and are unlikely to exert human health effects at typical ambient exposure
concentrations (or perhaps only under special circumstances). For example,
particles of crustal origin, which are predominately in the coarse fraction, are
relatively non-toxic under most circumstances, compared to combustion-related
particles (such as from coal and oil combustion, wood burning, etc.) However,
under some conditions, crustal particles may become sufficiently toxic to cause
human health effects. For example, resuspended crustal particles may be
contaminated with toxic trace elements and other components from previously
deposited fine PM, e.g., metals from smelters (Phoenix) or steel mills
(Steubenville, Utah Valley), PAHs from automobile exhaust, or pesticides from
agricultural lands. (CD, p. 8-344)
This is consistent with CASAC's conclusion that the available evidence from health studies
suggests that the focus of an indicator for thoracic coarse particles should be on such particles
found in urban, not rural environments (Henderson, 2005).
The staff assessment of the available evidence relevant to the appropriate scope of an
indicator for coarse particles can be summarized as follows. Thoracic coarse particle
concentrations generally reflect contributions from local sources, and the limited information
available from speciation of thoracic coarse particles and emissions inventory data indicate that
the sources of urban thoracic coarse particles generally differ from those found in nonurban
areas. As a result, the kinds of thoracic coarse particles people are exposed to in urban areas can
be expected to differ significantly from the kinds found in non-urban or rural areas. Ambient
PM10_2 5 exposure is associated with health effects in studies conducted in urban areas, and the
limited available health evidence more strongly implicates coarse particles from industrial and
traffic-related sources than from uncontaminated soil or geologic sources. The limited evidence
does not support either the existence or the lack of causative associations for community
exposures to agricultural or mining industries. Given the apparent differences in composition
and in the epidemiologic evidence, it is not appropriate to conclude that evidence of associations
with health effects related to urban coarse particles would also apply to nonurban or rural coarse
particles.
Collectively, the evidence suggests that a more narrowly defined indicator for thoracic
coarse particles should be considered that would protect public health against effects linked with
thoracic coarse particles present in urban areas. Such an indicator would be principally based on
particle size, but also reflect a focus on those thoracic coarse particles that are generally present
in urban environments. Staff recommends consideration of thoracic coarse urban particulate
matter (UPM10_2 5) as an indicator for a thoracic coarse particle standard, referring to airborne
particles between 2.5 and 10 jim in diameter that are generally present in urban environments,
5-57
-------
which, as discussed above, are principally comprised of resuspended road dust typical of high
traffic-density areas and emissions from industrial sources. Staff considers that UPM10_25 would
more likely be an effective indicator for standards to protect against effects of thoracic coarse
particles than a more broadly focused PM10_25 indicator. Staff notes that this indicator would
also be consistent with an appropriately cautious interpretation of the epidemiologic evidence
that does not potentially over-generalize the results of the limited available studies.
5.4.2.2 Related Requirements for PM10_2 5 Monitors and Monitoring Network Design
Along with staffs recommendations on the definition for an indicator for urban thoracic
coarse particles, it is important to recognize that requirements for federal reference and
equivalent methods and monitoring network design are essential components in fully defining
and applying a PM indicator. While these efforts are not described in detail in this Staff Paper,
the discussions below highlight key components of these activities for an urban thoracic coarse
particle standard indicator.13
First, in conjunction with considering UPM10_2 5 as an indicator for standards to address
thoracic coarse particles, EPA is evaluating various ambient monitoring methods, including
continuous methods. This evaluation is being performed through field studies of commercially
ready and prototype methods to characterize the measurement of PM10_25 in urban areas.14 This
PM10_25 methods evaluation has resulted in characterizing the performance of multiple PM10_25
measurement technologies under a variety of aerosol and meteorological conditions typical of
urban areas in regions across the country. This characterization has demonstrated that the
majority of commercially available methods for the measurement of PM10_25 have good precision
and are well correlated with filter-based gravimetric methods, such as the difference method that
has primarily been used to date (i.e., operation of collocated PM10 and PM25 low volume FRMs
and calculating PM10_2 5 by difference). EPA is working with the instrument manufacturers to
address design issues that should reduce biases that have been observed among methods, in
preparation for another field study examining the performance of the methods.
EPA has also begun the process of examining data quality objectives for a potential 24-
hour UPM10_2 5 standard. On the basis of preliminary analyses, it is apparent that greater sampling
frequency will be important due to the high variability of PM10_2 5 in the atmosphere in urban
environments. Due to the resource intensive nature of filter sampling on a daily basis, staff
believes that it will be critical to include continuous monitoring in any network deployment
strategy for a possible UPM10_2 5 standard. In addition to providing high temporal resolution to
13 EPA plans to issue proposed and final revisions to related requirements for federal reference and
equivalent methods and monitoring network design concurrently with proposed and final decisions regarding
revisions to the PM NAAQS.
This work is being done in consultation with the CAS AC AAMM Subcommittee.
5-58
-------
PM10_25 data, continuous monitors would also support public reporting of UPM10_25 episodes and
inclusion of UPM10.25 in an air quality forecasting program.
In addition, EPA has also commenced the examination of monitoring network design
issues for a possible UPM10_2 5 standard. One key focus in monitoring network design is
consistency with information from the health studies and with staff conclusions regarding the
appropriateness of an indicator based on PM10_2 5 mass generally present in urban environments
and reflecting local urban sources. The available epidemiologic studies have reported
associations with PM10_2 5 mass that has been measured at sites that are located in or near heavily
populated areas, but not in close proximity to industrial or other specific sources, to reflect
community exposure levels. In the context of applying such a standard, staff has examined
various measures for designing an appropriate monitoring network for an urban PM10_2 5 standard,
including indicators of traffic density and locations of industrial sources, as well as total
population and population density. Staff observes that these traffic- and population-related
measures are very highly correlated with one another. On an initial basis, a focus on
CBSA/CSAs with populations of at least 100,000 would result in a base monitoring network that
would include approximately 350 areas. Further refinements in network size in terms of the
number of monitors to be placed within these areas can be accomplished through additional
criteria such as a graduated scale that increases monitoring requirements for more populous
CBSA/CSAs, as well as a hybrid strategy that combines population/traffic-related measures with
comparisons to historical PM10 or estimated PM10_2 5 design values to focus additional monitors in
areas with elevated concentrations associated with traffic-related and industrial sources, and
potentially reduce monitoring in CBSA/CSAs with lower concentrations levels of thoracic
coarse particles.
As noted elsewhere in this document, PM10_2 5 is more highly variable in the atmosphere
than PM2 5, such that the specific criteria regarding monitor placement will be a particularly
important consideration in the implementation of a PM10_2 5 monitoring network. Staff has
examined various options for targeting monitors within larger CBSA/CSAs to focus
measurements in locations of expected high concentrations of urban coarse thoracic particles that
also represent the population-oriented objective of monitors underlying key epidemiologic
studies. One useful metric for guiding monitor placement appears to be population density, for
example, the placement of monitors in U.S. Census Block Groups characterized by population
densities greater than 500 people per square mile. This type of criterion effectively guides
monitors toward the urban/center city and suburban location settings that likely represent areas
of high population exposure to elevated concentrations, as compared to the rural parts of
urbanized CBSA/CSAs. Staff notes, in addition, that the population density metric based on
Census Block Group has the potential to identify and de-emphasize monitoring in "population
holes" within the urbanized area, such as areas with only industrial development. For PM10,
some monitors have historically been located in such heavily industrialized areas away from
5-59
-------
significant population exposure, including some existing source-oriented monitors placed on
facility fence-lines to investigate specific emission-related complaints. Staff believes that it
would be inconsistent with the basis of a UPM10_2 5 indicator and, thus, inappropriate to include
any such monitors as part of a UPM10_2 5 monitoring network.
Consideration of specific measurement scale guidance for placement of PM10_25 monitors
is influenced by the rapid fallout of coarse fraction particles with increasing distance from
sources. Analysis of existing PM10_25 concentrations in cities such as Birmingham, Cleveland,
and Las Vegas has revealed large differences in three-year mean, 98th percentile design values,
as well as poor daily correlations, between monitors located within several kilometers of each
other. In many cases, these large differences can be explained by the immediate proximity of the
higher reading monitor to areas of industrial and high traffic activity, compared with lower
reading monitors that may be farther from coarse-particle generating or resuspending activities.
As a result, staff believes that to be consistent with the population-oriented objective of the
monitors underlying key epidemiologic studies, population-oriented monitors that represent
middle scale (i.e., 100 meters to 0.5 km) or neighborhood scale (i.e., 0.5 to 4.0 km) sized areas of
relatively uniform land use would be most appropriate for a UPM10_2 5 network, recognizing that
a significant spatial gradient in coarse thoracic particle concentrations may still exist across
middle scale-sized areas.15 Such preferred locations for PM10_2 5 monitors in a UPM10_2 5 network
would include densely populated communities located several hundred meters from significant
sources, such as industrial sources or heavily traveled roadways.16
Finally, with regard to elevations in thoracic coarse particle levels that may occur in
urban areas as a result of dust storms or other such events, staff observes that EPA has
historically used implementation policies to address such issues in the implementation of PM
standards. Examples of such policies include the "natural events" policy for implementation of
the PM10 standards. This policy includes guidance regarding exclusion of PM10 measurements
from natural events in making determinations on attainment or nonattainment of the standards.
The natural events discussed in this policy include volcanic and seismic activity, wildland fires
and high wind events. EPA is now in the process of revising policies for natural events and
exceptional events to address issues related to the PM2 5 NAAQS, as well as potential new
UPM,n_9 5 standards.17
15 Staff notes that these discussions of measurement scale for monitors are consistent with definitions in 40
CFR Part 58, Appendix D, http://www.access.gpo.gov/nara/cfr/waisidx 04/40cfr58 04.html.
16 Staff observes that PM10_2 5 monitors may also be sited to provide data for other purposes such as research
or trends assessment that would be outside the scope of a monitoring network for a UPM10_2 5 standard.
17 Current policy available at http://www.epa.gOv/ttn/oarpg/tl/memoranda/nepol.pdf.X EPA plans to issue
proposed and final revisions to the natural events policy concurrently with any proposed and final decisions
regarding revisions to the PM NAAQS.
5-60
-------
5.4.3 Averaging Times
In the last review, EPA retained both annual and 24-hour standards to provide protection
against the known and potential effects of short- and long-term exposures to thoracic coarse
particles (62 FR at 38,677-79). This decision was based in part on qualitative considerations
related to the expectation that deposition of thoracic coarse particles in the respiratory system
could aggravate effects in individuals with asthma. In addition, quantitative support came from
limited epidemiologic evidence suggesting that aggravation of asthma and respiratory infection
and symptoms may be associated with daily or episodic increases in PM10, where dominated by
thoracic coarse particles including fugitive dust. Further, potential build-up of insoluble thoracic
coarse particles in the lung after long-term exposures to high levels was also considered
plausible.
New information available in this review on thoracic coarse particles, as discussed above
in section 5.4.1, includes several epidemiologic studies that report statistically significant
associations between short-term (24-hour) exposure to PM10_2 5 and various morbidity effects and
mortality. With regard to long-term exposure studies, while one recent study reported a link
between reduced lung function growth and long-term exposure to PM10_2 5 and PM2 5, other such
studies reported no associations. The CD concludes that the evidence does not suggest an
association with long-term exposure to PM10_25 (CD, p. 9-79). Staff also notes that no evidence
is available to suggest associations between PM10_25 and very short exposure periods of one or
more hours.
Based on these considerations, staff concludes that the newly available evidence
continues to support a 24-hour averaging time for a standard intended to control thoracic coarse
particles, based primarily on evidence suggestive of associations between short-term (24-hour)
exposure and morbidity effects and, to a lessor degree, mortality. Noting the absence of
evidence judged to be even suggestive of an association with long-term exposures, staff
concludes that there is no evidence that directly supports an annual standard, while recognizing
that it could be appropriate to consider an annual standard to provide a margin of safety against
possible effects related to long-term exposure to thoracic coarse particles that future research
may reveal. Staff observes, however, that a 24-hour standard that would reduce 24-hour
exposures would also likely reduce long-term average exposures, thus providing some margin of
safety against the possibility of health effects associated with long-term exposures.
5.4.4 Alternative Standards to Address Health Effects Related to Short-term Exposure
As noted earlier, in the last review, EPA's decision to retain the level of the 24-hour PM10
standard of 150 |ig/m3 (with revision of the form of the standard) for protection against effects of
exposure to coarse fraction particles was based on two community studies of exposure to fugitive
dust that showed health effects in areas experiencing large exceedances of that standard (Gordian
et al., 1996; Hefflin et al., 1994), as well as on qualitative information regarding the potential for
5-61
-------
health effects related to short-term exposure to thoracic coarse particles. Because of the very
limited nature of this evidence, staff concluded that while it supported retention of a standard to
control thoracic coarse particles, it provided no basis for considering a more protective standard.
However, because of concerns about the expected-exceedance-based form of the 1987 PM10
standard, primarily related to the stability of the attainment status of an area over time and
complex data handling conventions needed in conjunction with less-than-every-day sampling,
EPA adopted a concentration-based form for the 24-hour standard, as was done for the 24-hour
PM25 standard, as discussed in section 5.3.6. In making this change, EPA selected a 99th
percentile form,18 in contrast to the 98th percentile form adopted for the 24-hour PM2 5 standard,
so as not to allow any relaxation in the level of protection that had been afforded by the previous
1-expected-exceedance form.
Since the last review, as discussed above in section 5.4.1, new evidence specific to
thoracic coarse particles has become available that reports associations between short-term
PM10_2 5 concentrations in some urban areas and various morbidity effects and, to a lesser degree,
mortality. In considering alternative standards that would provide protection against such health
effects, as discussed below, staff has taken into account evidence-based considerations and has
examined the extent to which risk-based considerations should also be taken into account.
5.4.4.1 Evidence-based Considerations
In considering the available evidence on associations between short-term PM10_2 5
concentrations and morbidity and mortality effects as a basis for setting a 24-hour standard for
thoracic coarse particles, staff has focused on relevant U.S. and Canadian studies (Appendix 3-
A). As discussed above in section 5.4.1 and in Chapter 3, staff has taken into account reanalyses
that addressed GAM-related statistical issues and has considered the extent to which the studies
report statistically significant and relatively precise relative risk estimates; the extent to which
the reported associations are generally robust to co-pollutant confounding and alternative
modeling approaches; and the extent to which the studies used relatively reliable air quality data.
As an initial matter, staff recognizes, as discussed in Chapter 3 (section 3.6.6), that these
short-term exposure studies provide no evidence of clear population thresholds, or lowest-
observed-effects levels, in terms of 24-hour average concentrations. Staff notes that in the one
study that explored a potential PM10_2 5 threshold, conducted in Phoenix, no evidence of a
threshold was observed for PM10_25, even though that study provided some suggestion of a
potential threshold for PM25. The CD concludes that while there is no evidence of a clear
18 As noted above, the court vacated the 1997 24-hour PM10 standard that had been revised to incorporate a
99th percentile form.
5-62
-------
threshold within the range of air quality observed in the studies,19 for some health endpoints it is
likely to be extremely difficult to detect threshold levels (CD, p. 9-45). As a consequence, this
body of evidence is difficult to translate directly into a specific 24-hour standard that would
protect against the range of effects that have been associated with short-term exposures.
In considering the evidence, staff is mindful of these uncertainties as well as the limited
nature of the available evidence. In examining the available evidence to identify a basis for a
range of standard levels that would be appropriate for consideration, staff has focused on the
upper end of the distributions of daily PM10_25 concentrations in the relevant studies, particularly
in terms of the 98th and 99th percentile values, consistent with the forms considered in section
5.3.6 above for a short-term PM25 standard. Staffs examination of the evidence is discussed
below, based on air quality information and analyses presented in Ross and Langstaff (2005) and
Ross (2005).
In looking first at the morbidity studies identified in section 5.4.1 that report statistically
significant associations with respiratory- and cardiac-related hospital admissions in Toronto
(Burnett et al., 1997), Seattle (Sheppard et al.,1999, 2003), and Detroit (Lippmann et al., 2000;
Ito, 2003), the reported 98th percentile values in the three areas range from approximately 30 to
36 |ig/m3, and the 99th percentile values range from 36 to 40 |ig/m3. To provide some
perspective on these PM10_2 5 levels, staff notes that the level of the 24-hour PM10 standard was
exceeded only on a few occasions during the time periods of the studies in Detroit and Seattle.20
Staff has also looked at the studies identified in section 5.4.1 that report statistically
significant and generally robust associations with mortality and short-term exposures to PM10_2 5.
Studies conducted in Phoenix (Mar et al., 2000, 2003) and Coachella Valley, CA (Ostro et al.,
2000, 2003) report 98th percentile PM10.25 values of approximately 70 and 107 |ig/m3, and 99th
percentile values of 75 and 134 |ig/m3, respectively. These studies were conducted in areas with
air quality levels that did not meet the current PM10 standards. In addition, a statistically
significant association was reported between PM10_25 and mortality in Steubenville as part of the
Harvard Six Cities analysis (Schwartz et al., 1996; Klemm et al., 2003). PM10_25 concentrations
were fairly high in this eastern city, with reported 98th and 99th percentile PM10_2 5 values of 53
and 61 |ig/m3, respectively. Staff notes that, in contrast with most areas, PM10_25 and PM25
concentrations were highly correlated (r=0.69) in Steubenville during the study period (Schwartz
19 Staff notes that the distributions of daily PM10_2 5 concentrations in these studies often extend down to or
below typical background levels, such that the likely range of background concentrations across the U.S., as
discussed in Chapter 2, section 2.6, could be a relevant consideration in this policy evaluation. Staff recognizes,
however, that there are insufficient data to estimate daily distributions of background PM10_2 5 levels (as was done for
background PM2 5 levels, as discussed in Chapter 2, section 2.6).
As shown in air quality data trends reports: for Seattle, 1997 Air Quality Annual Report for Washington
State, p. 17, at http://www.ecv.wa.gov/pubs/97208.pdf: for Detroit, Michigan's 2003 Annual Air Quality Report, p.
46, at http://www.deq.state.mi.us/documents/deq-aqd-air-reports-03AQReport.pdf.
5-63
-------
et al., 1996; Klemm et al., 2003). In contrast to the statistically significant mortality associations
with PM10_25 reported in these studies, staff notes that no such associations were reported in a
number of other studies, including in the five other cities that were part of the Harvard Six Cities
study (Boston, St. Louis, Knoxville, Topeka, and Portage), San Jose, Detroit, Philadelphia, and
Pittsburgh. With the exception of Pittsburgh, these cities had much lower 98th percentile PM10_25
values, ranging from 18 to 49 |ig/m3.
Based on the air quality information reported in these studies, staff makes the following
observations:
• In the morbidity studies that reported statistically significant associations with
respiratory or cardiovascular hospitalization in Detroit and Seattle, and with
respiratory symptoms in six U.S. cities, the reported 98th percentile PM10_25 values
ranged from approximately 30 up to 40 |ig/m3.
In the mortality studies that reported statistically significant associations in
Steubenville, Coachella Valley, and Phoenix, the reported 98th percentile PM10_2 5
values were all above 50 |ig/m3, ranging from 53 |ig/m3 up to 107 |ig/m3 in
Coachella Valley.
• In the mortality studies that reported no statistically significant associations in
five of the cities in the Harvard Six Cities study (Boston, St. Louis, Knoxville,
Topeka, and Portage), and in San Jose, Detroit, Philadelphia, and Pittsburgh, the
reported 98th percentile PM10_2 5 values were below 50 |ig/m3 for all cities except
Pittsburgh, ranging from 18 to 49 |ig/m3.
In looking more closely at air quality data used in the morbidity and mortality studies
discussed above, however, staff recognizes that the uncertainty related to exposure measurement
error can be potentially quite large. For example, in looking specifically at the Detroit study,
staff notes that the PM10_2 5 air quality values were based on air quality monitors located in
Windsor, Canada. The study authors determined that the air quality values from these monitors
were generally well correlated with air quality values monitored in Detroit, where the hospital
admissions data were gathered, and, thus concluded that these monitors were appropriate for use
in exploring the association between air quality and hospital admissions in Detroit. Staff has
observed, however, that the PM10_2 5 levels reported in this study are significantly lower than the
PM10_2 5 levels measured at some of the Detroit monitors in 2003 — an annual mean level of 13.3
|ig/m3 is reported in the study based on 1992 to 1994 data, as compared to an average annual
mean level of 21.7 |ig/m3 measured at two urban-center monitors in 2003 (which is used as the
basis for the risk assessment presented in Chapter 4). This observation prompted staff to further
explore the comparison between PM10_2 5 levels monitored at Detroit and Windsor sites. This
exploration has shown that in recent years, based on available Windsor and Detroit data from
1999 to 2003, the Windsor monitors used in this study typically have recorded PM10_2 5 levels that
5-64
-------
are generally less than half the levels recorded at urban-center Detroit monitors, though the
concentrations measured in Windsor are more similar to concentrations reported for suburban
areas well outside the city. These observations lead staff to conclude that the statistically
significant, generally robust association with hospital admissions in Detroit likely reflects
population exposures that may be appreciably higher in the central city area, but not necessarily
across the broader study area, than would be estimated using data from the Windsor monitors.
Staff also looked more specifically at the Coachella Valley mortality study (Ostro et al.,
2000; 2003), in which data were used from a single monitoring site in one city, Indio, within the
study area where daily measurements were available. The mean PM10_2 5 concentration during
the study period (1989-1998) was 30.5 |ig/m3 at this monitoring site. Consistently lower
concentrations were measured at another city in the Coachella Valley area, Palm Springs; during
the study period the mean value at this site was 17 |ig/m3. The authors report that the data for the
two sites were correlated, with correlation coefficients of about 0.6 for each of the three PM
indicators. Using 2001-2003 data, mean PM10_25 levels were reported to be 44 and 15 |ig/m3 at
the Indio and Palm Springs sites, respectively. Thus, in Coachella Valley, mortality was
significantly associated with PM10_2 5 measurements made at the Indio site, but a portion of the
study population would have been expected to experience appreciably lower ambient exposure
levels. In contrast to the Detroit study, air quality data used in the mortality study conducted in
Coachella Valley appear to represent concentrations on the high end of PM10_2 5 levels for
Coachella Valley communities.
A closer examination of the air quality data used in the other studies discussed above
generally shows less disparity between air quality levels at the monitoring sites used in the
studies and the broader pattern of air quality levels across the study areas than that described
above in the Detroit and Coachella Valley studies. More specifically, less variation across
monitoring sites was seen in air quality data in both Seattle and St. Louis. In Steubenville, the
PM10_2 5 concentrations measured at the centrally located monitor used in the study are somewhat
higher than those reported at other area monitors. In Phoenix, data from a larger network of
PM10 monitors across the area shows a large gradient in concentrations across the urban area,
with lower concentrations measured at sites on the north side to appreciably higher
concentrations on the south side; data from the one centrally located monitoring site used in the
study appears to fall in the mid-range of concentrations along the north-south gradient.
This closer examination of air quality information generally reinforces the view that
exposure measurement error is potentially quite large in these PM10_2 5 studies. As a
consequence, the air quality levels reported in these studies, as measured by ambient
concentrations at monitoring sites within the study areas, are not necessarily good surrogates for
population exposures that are likely associated with the observed effects in the study areas or
that would likely be associated in other urban areas across the country. The Detroit example
suggests that population exposures were probably appreciably underestimated in the Detroit
5-65
-------
morbidity study, such that the observed effects are likely associated with higher PM10_2 5 levels
than reported. In contrast, the Coachella Valley mortality study provides an example in which
population levels were probably appreciably overestimated, such that the observed effects may
well be associated with lower PM10_2 5 levels than reported. At relatively low levels of air
quality, population exposures implied by these studies as being associated with the observed
effects likely become more uncertain, suggesting an even higher degree of caution in interpreting
the group of morbidity studies as a basis for identifying a standard level that would protect
against the observed effects.
Taking into account this closer examination of the studies, staff concludes that this
evidence suggests consideration of a standard for urban thoracic coarse particles at a PM10_2 5
level at least down to 50 |ig/m3, in conjunction with a 98th percentile form (which would be
roughly equivalent to a level of 60 |ig/m3 in conjunction with a 99th percentile form). While
lower levels may be considered to provide a margin of safety against morbidity effects that may
possibly occur at such lower levels, staff believes that consideration of a standard below these
levels may not be warranted based on this evidence. Staffs view takes into account the
conclusion that this evidence is particularly uncertain as to population exposures, especially from
the morbidity studies reporting effects at relatively low concentrations, as well as the general
lack of evidence of associations from the group of mortality studies with reported concentrations
below these levels. A standard set at or somewhat above these levels could be expected to
provide protection against the potential mortality effects observed in studies that reported
ambient concentrations above this level, as well as morbidity effects that may occur above this
level.
An even more cautious or restrained approach to interpreting the limited body of PM10_25
epidemiologic evidence would be to judge that the uncertainties in this whole group of studies as
to population exposures that are associated with the observed effects are too large to use the
reported air quality levels directly as a basis for setting a specific standard level. Staff notes that
such a judgment would not be inconsistent with the conclusion reached above in section 5.4.1
that these studies, together with other dosimetric and toxicologic evidence, provide support for
retaining standards for thoracic coarse particles at some level to protect against the morbidity
and mortality effects observed in the studies, regardless of whether an associated population
exposure level can be clearly discerned from the studies.
Considering this more cautious interpretation, staff believes that it would be reasonable
to interpret the available epidemiologic evidence more generally, by considering whether it
provides a sufficient basis for a standard that would afford protection generally "equivalent" to
that afforded by the current PM10 standards. Considering the available evidence in this way leads
to the following observations:
5-66
-------
• The statistically significant mortality associations with short-term exposure to
PM10_2 5 reported in the Phoenix and Coachella Valley studies were observed in
areas that did not meet the current PM10 standards.
The statistically significant morbidity associations with short-term exposure to
PM10_2 5 reported in the Detroit and Seattle studies were observed in areas that
exceeded the level of the current 24-hour PM10 standard only on few occasions
during the time periods of the studies.
• All but one of the statistically significant morbidity and mortality associations
with short-term exposure to PM10 reported in areas dominated by coarse fraction
particles (including Reno/Sparks, NV, Tucson, AZ, Anchorage, AK, and the Utah
Valley area, as discussed above in section 5.4.1) were observed in areas that did
not meet the current PM10 standards.
Based on these considerations, staff finds little basis for concluding that the degree of
protection afforded by the current PM10 standards is greater than warranted, since potential
mortality effects have been associated with air quality levels not allowed by the current
standards, but have not been associated with air quality levels that would generally meet the
current standards. Further, staff finds little basis for concluding that a greater degree of
protection is warranted in light of the very high degree of uncertainty in the relevant population
exposures implied by the morbidity studies. Staff judges, therefore, that it is reasonable to
interpret the available evidence as supporting consideration of a short-term standard for urban
thoracic coarse particles set so as to provide generally "equivalent" protection to that afforded by
the current PM10 standards, recognizing of course that no one PM10_2 5 level will be strictly
equivalent to a specific PM10 level in all areas. Such a standard would likely provide protection
against morbidity effects especially in urban areas where, unlike the study areas, PM10 is
generally dominated by coarse-fraction rather than fine-fraction particles. Such a standard
would also likely provide protection against the more serious, but more uncertain, PM10_2 5-
related mortality effects generally observed at somewhat higher air quality levels.
To identify a range of levels for consideration for a short-term standard for urban thoracic
coarse particles set so as to afford generally "equivalent" protection as the current PM10
standards, staff has analyzed available data on PM10_2 5 and PM10 24-hour average concentrations
from monitors that would be included in the monitoring network design provisions discussed in
section 5.4.2.2.21 Based on a regression analysis of the 205 monitoring sites so identified
(Schmidt et al., 2005), staff finds that a UPM10_25 level of approximately 60 |ig/m3 in terms of a
21 Monitors included in this analysis are those in CBS As with at least 100,000 population and in census
block groups with a population density of at least 500, and that also had 3 years of complete data in each quarter for
bothPM10andPM10.25.
5-67
-------
98th percentile form (or approximately 70 |ig/m3 in terms of a 99th percentile form) would be
roughly equivalent on average across the U.S. to a PM10 level of 150 |ig/m3 in terms of a one-
expected-exceedance form.22 While noting appreciable variability in the estimated point of
equivalence across individual sites, these levels of approximate average equivalence are quite
consistent across each of the five regions in which all of the areas that do not meet the current
PM10 standards are located (including the southern California, southwest, northwest, upper mid-
west, and southeast regions). Notably different average equivalence levels were observed in the
other two regions, i.e., approximately 40 |ig/m3 in the northeast and over 70 |ig/m3 in the
industrial mid-west (in terms of 98th percentile forms).
Another approach to identifying a UPM10_2 5 standard that is generally "equivalent" to the
current PM10 standards is to compare the number of areas, and the population in those areas, that
would likely not meet a specific UPM10_2 5 standard, set at a given level and form, with the same
measures in areas that do not meet the current PM10 standards. Such an analysis, based on 2001
to 2003 data from monitors that would be included in the monitoring network design provisions
discussed in section 5.4.2.2 provides some rough indication of the breadth of protection
potentially afforded by alternative standards. The results of this analysis (Appendix 5B, Tables
5B-2(a), (b) and (c), for the 98th and 99th percentile forms of a 24-hour UPM10.25 standard and the
current PM10 standards, respectively) indicate that a UPM10_2 5 standard of about 70 or 65 i-ig/m3,
98th percentile form, (or approximately 85 or 80 i-ig/m3, 99th percentile form) would impact
approximately the same number of counties or number of people, respectively, as would the
current PM10 standards.23
Based on these alternative analyses of generally "equivalent" UPM10_25 standards, staff
concludes that it is reasonable to consider a 24-hour UPM10_2 5 standard in the range of
approximately 60 to 70 |ig/m3 with a 98th percentile form (approximately 70 to 85 |ig/m3 with a
99th percentile form). Considering standards within these ranges, somewhat above the levels
identified above based on an examination of the air quality concentrations in the relevant
epidemiologic studies (i.e., 50 i-ig/m3, 98th percentile form, or 60 i-ig/m3, 99th percentile form),
would reflect an even higher degree of caution in interpreting the epidemiologic evidence.
Consideration of a generally "equivalent" UPM10_25 standard would reflect a judgment that while
the epidemiologic evidence supports establishing a short-term standard for urban thoracic coarse
Across the U.S., the 95% confidence intervals around these point estimates are approximately + 3 i-ig/m3,
while region-specific intervals are approximately +10 ng/m3 in the five regions in which all of the areas that do not
meet the current PM10 standards are located.
23 As shown in Table 5B-2(c), staff notes that there are 585 counties with PM10 monitoring sites used in
determining compliance with the PM10 standards, whereas only 309 of those counties have monitor sites that would
be included in the monitoring network design provisions discussed in section 5.4.2.2. Of these 309 counties, 259
have PM10 and PM10_2 5 air quality data that meet the data completeness criteria defined for this analysis, which are
somewhat less restrictive that the criteria that were applied in the regression analysis described above.
5-68
-------
particles at such a generally "equivalent" level, the evidence concerning air quality levels of
thoracic coarse particles in the studies is not strong enough to provide a basis for revising the
level of protection generally afforded by the current PM10 standards.
5.4.4.2 Risk-based Considerations
Beyond looking directly at the relevant epidemiologic evidence and related air quality
information, staff has also considered the extent to which the PM10_25 risk assessment results
discussed in Chapter 4 can help inform consideration of alternative 24-hour PM10_25 standards.
While one of the goals of the PM10_2 5 risk assessment was to provide estimates of the risk
reductions associated with just meeting alternative PM10_25 standards, staff has concluded that the
nature and magnitude of the uncertainties and concerns associated with this portion of the risk
assessment weigh against use of these risk estimates as a basis for recommending specific
standard levels. These uncertainties and concerns include, but are not limited to the following:
as discussed above, concerns that the current PM10_2 5 levels measured at ambient
monitoring sites in recent years may be quite different from the levels used to
characterize exposure in the original epidemiologic studies based on monitoring
sites in different location, thus possibly over- or underestimating population risk
levels;
• greater uncertainty about the reasonableness of the use of proportional rollback to
simulate attainment of alternative PM10_2 5 daily standards in any urban area due to
the limited availability of PM10_25 air quality data over time;
concerns that the locations used in the risk assessment are not representative of
urban areas in the U.S. that experience the most significant 24-hour peak PM10_25
concentrations, and thus, observations about relative risk reductions associated
with alternative standards may not be relevant to the areas expected to have the
greatest health risks associated with elevated ambient PM10_2 5 levels; and
concerns about the much smaller health effects database that supplies the
concentration-response relationships used in the risk assessment, compared to that
available for PM25, which limits our ability to evaluate the robustness of the risk
estimates for the same health endpoints across different locations.
5.4.4.3 Summary
In considering the relevant dosimetric, toxicologic, and epidemiologic evidence,
associated air quality information, and related limitations and uncertainties, staff concludes that
there is support for considering a 24-hour UPM10_2 5 standard to replace the current PM10
standards to provide protection against health effects associated with short-term exposures to
thoracic coarse particles that are generally present in urban environments. In looking at the
evidence of associations between short-term exposure to PM10_2 5 and morbidity and mortality,
5-69
-------
both on its own and in the context of considering a standard set so as to be generally "equivalent"
to the current PM10 standards, staff concludes that it is appropriate to consider a 24-hour standard
in the range of 50 to 70 |ig/m3, with a 98th percentile form, or in the range of 60 to 85 |ig/m3, with
a 99th percentile form. A standard set within either of these ranges could be expected to provide
protection against the morbidity effects of PM10_25, especially those likely to occur in areas in
which PM10 is dominated by thoracic coarse particles, as well as to protect against the potential
mortality effects of PM10_25.
Staff recognizes, however, that the epidemiologic evidence on morbidity and mortality
effects related to PM10_2 5 exposure is very limited at this time. A key area of uncertainty in this
evidence is the potentially quite large uncertainty related to exposure measurement error for
PM10_2 5, as compared with fine particles. PM10_2 5 concentrations can vary substantially across a
metropolitan area and thoracic coarse particles are less able to penetrate into buildings than fine
particles; thus, the ambient concentrations reported in epidemiologic studies may not well
represent area-wide population exposure levels. Other key uncertainties include the very limited
information available on the composition of thoracic coarse particles and the effects of thoracic
coarse particles from various sources, and the lack of evidence on potential mechanisms for
effects of thoracic coarse particles. Staff believes that placing relatively more weight on these
uncertainties would focus consideration of standard levels toward the upper end of the ranges
identified above, whereas a more precautionary approach would focus consideration on the
lower end of these ranges.
5.4.5 Summary of Staff Recommendations on Primary PM10_2 5 NAAQS
Staff recommendations for the Administrator's consideration in making decisions on
standards for thoracic coarse particles, together with supporting conclusions from sections 5.4.1
through 5.4.4, are briefly summarized below. In making these recommendations, staff is mindful
that the Act requires standards to be set that, in the Administrator's judgment, 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 NAAQS be set at
zero-risk levels, but rather at levels that avoid unacceptable risks to public health.
( 1) The current primary PM10 standards should be revised in part by replacing the PM10
indicator with an indicator of urban thoracic coarse particles that does not generally
include fine particles. Any such revised standards should be based primarily on
available health effects evidence and air quality data generally indexed by PM10_2 5, to
provide public health protection more specifically directed toward effects related to
exposure to thoracic coarse particles in the ambient air, together with consideration of
evidence from studies using PM10 in locations where PM10_2 5 is the dominant fraction.
5-70
-------
(2) The indicator for a thoracic coarse particle standard should be UPM10.2 5, which is
consistent with the recommendation made in section 5.3.7 to retain PM25 as the indicator
for fine particle standards.
( a) As noted above, this recommended indicator is primarily based on particle size,
but with a more narrow focus on those thoracic coarse particles that are generally
present in urban environments. The available evidence from studies on
atmospheric chemistry and sources and health effects of thoracic coarse particles
indicates that a UPM10_2 5 indicator is more likely an effective indicator for
standards to protect against health effects of thoracic coarse particles than a more
broadly-focused PM10_2 5 indicator.
( b) In support of this recommendation, work should continue on the development of
an FRM for a UPM10_2 5 indicator based on the ongoing field program to evaluate
various types of PM10_25 monitors, and consideration should be given to the
adoption of FEMs for appropriate continuous measurement methods.
(3) A 24-hour averaging time should be retained for a UPM10_2 5 standard to protect against
health effects associated with short-term exposure periods, with consideration given to
the use of either a 98th or 99th percentile form. There is little basis for also retaining an
annual averaging time for protection against such health effects.
( 4) Consideration should be given to setting a 24-hour UPM10_2 5 standard with a level in the
range of approximately 50 to 70 |ig/m3, 98th percentile form, or approximately 60 to
85 |ig/m3, 99th percentile form. Staff believes that a more precautionary approach would
focus consideration on the lower end of these ranges, while consideration of a standard
set toward the upper end of these ranges would place relatively more weight on the
uncertainties inherent in the very limited epidemiologic evidence. Consideration of
UPM10_2 5 standards within the ranges recommended above, and design considerations for
an associated UPM10_2 5 monitoring network, should take into account the especially large
variability seen in currently available information on ambient concentrations and
composition of PM10_25 in urban areas.
5.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
RECOMMENDATIONS RELATED TO SETTING PRIMARY PM STANDARDS
Staff believes it is important to continue to highlight the unusually large uncertainties
associated with establishing standards for PM relative to other single component pollutants for
which NAAQS have been set. Key uncertainties and staff research recommendations on health-
related topics are outlined below. In some cases, research in these areas can go beyond aiding in
5-71
-------
standard setting to aiding in the development of more efficient and effective control strategies.
Staff notes, however, that a full set of research recommendations to meet standards
implementation and strategy development needs is beyond the scope of this discussion.
The 1996 PM Staff Paper included a discussion of uncertainties and research
recommendations (EPA, 1996b, pp. VII-41-44) that addressed the following issues related to
understanding health effects associated with exposure to PM:
• lack of demonstrated biological mechanisms for PM-related effects,
• potential influence of measurement error and exposure error,
• potential confounding by copollutants,
• evaluation of the effects of components or characteristics of particles,
the shape of concentration-response relationships,
methodological uncertainties in epidemiologic analyses,
the extent of life span shortening,
characterization of annual and daily background concentrations, and
• understanding of the effects of coarse fraction particles.
As has been discussed in depth in the CD, especially in Chapters 5 through 8, an
extensive body of new studies related to understanding health effects associated with exposure to
PM is now available that provides important information on many of the topics listed above. For
example, regarding the lack of demonstrated biological mechanisms, new evidence from
toxicologic and controlled human exposure studies has provided information on an array of
potential mechanisms for effects on the cardiac and respiratory systems, as discussed in Chapters
7 and 9 of the CD. Still, the CD emphasizes that much remains to be learned to fully understand
the pathways or mechanisms by which PM is linked with different health endpoints. For each of
the issues listed above, new evidence has become available that helps to reduce uncertainties,
although uncertainty has been reduced in some areas more than others. Staff has identified the
following key uncertainties and research questions that have been highlighted in this review of
the health-based primary standards:
(1) The body of evidence on effects of thoracic coarse particles has been expanded, but the
uncertainties regarding thoracic coarse particles are still much greater than those for fine
particles. As discussed in Chapter 2, the spatial variability of thoracic coarse particles is
generally greater than that for fine particles, which will increase uncertainty in the
associations between health effects and thoracic coarse particles measured at central site
monitors. Additional research is needed on such intra-city variability as well as on inter-
city variability and on temporal (e.g., seasonal) variability. Additional exposure research
is needed to understand the influence of measurement error and exposure error on
thoracic coarse particle epidemiology results. In addition, little is known about coarse
5-72
-------
particle composition, and less about the health effects associated with individual
components or sources of thoracic coarse particles, but it is possible that there are
components of thoracic coarse particles (e.g., crustal material in non-urban areas) that are
less likely to have adverse effects, at least at lower concentrations, than other
components.
(2) Identification of specific components, properties, and sources of fine particles that are
linked with health effects remains an important research need. Available evidence
provides no basis for expecting that any one component would be solely responsible for
PM2 5-related effects, but it is likely that some components are more closely linked with
specific effects than others. Continued source characterization, exposure, epidemiologic,
and toxicologic research is needed to help identify components, characteristics, or
sources of particles that may be more closely linked with various specific effects to aid in
our understanding of causal agents and in the development of efficient and effective
control strategies for reducing health risks. Conducting human exposure research in
parallel with such health studies will help reduce the uncertainty associated with
interpreting health studies and provide a stronger basis for drawing conclusions regarding
observed effects.
(3) An important aspect in characterizing risk and making decisions regarding air quality
standard levels is the shape of concentration-response functions for PM, including
identification of potential threshold levels. Recent studies continue to show no evidence
for a clear threshold level in relationships between various PM indicators and mortality,
within the range of concentrations observed in the studies, though some studies have
suggested potential levels.
(4) The relationship between PM and other air pollutants in causing health effects remains an
important question in reducing public health risk from air pollution. Numerous new
analyses have indicated that associations found between PM and adverse health effects
are not simply reflecting actual associations with some other pollutant. However, effects
have been found with the gaseous co-pollutants, and it is possible that pollutants may
interact or modify effects of one another. Further understanding of the sources,
exposures, and effects of PM and other air pollutants can assist in the design of effective
strategies for public health protection.
(5) Methodological issues in epidemiologic studies were discussed at length in the previous
review, and it appeared at the time that the epidemiologic study results were not greatly
affected by selection of differing statistical approaches or methods of controlling for
other variables, such as weather. However, investigation of recently discovered
questions on the use of generalized additive models in time-series epidemiologic studies
has again raised model specification issues. While reanalyses of studies using different
modeling approaches generally did not result in substantial differences in model results,
5-73
-------
some studies showed marked sensitivity of the PM effect estimate to different methods of
adjusting for weather variables. There remains a need for further study on the selection
of appropriate modeling strategies and appropriate methods to control for time-varying
factors, such as temperature.
(6) Selection of appropriate averaging times for PM air quality standards is important for
public health protection, and available information suggests that some effects, including
cardiac-related risk factors, may be linked to exposures of very short duration (e.g., one
or more hours). Data on effects linked with such peak exposures, such as those related to
wildfires, agricultural burning, or other episodic events, would be an important aid to
public health response and communication programs. Investigation into the PM exposure
time periods that are linked with effects will provide valuable information both for the
standard-setting process and for risk communication and management efforts.
(7) There remain significant uncertainties in the characterization of annual and daily
background concentrations for fine particles and especially for thoracic coarse particles.
Further analyses of air quality monitoring and modeling that improved these background
characterizations would help reduce uncertainties in estimating health risks relevant for
standard setting (i.e., those risks associated with exposure to PM in excess of background
levels) and would aid in the development and implementation of associated control
programs.
5-74
-------
REFERENCES
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105:614-620.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between paniculate- and gas-phase components of urban air pollution and daily
mortality in eight Canadian cities. Inhalation Toxicol. 12(suppl. 4): 15-39.
Burnett, R. T.; Goldberg, M. S. (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].
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12:281-298
Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
pollution. Arch. Environ. Health 52:113-117.
EPA (1996a) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC: National Center for
Environmental Assessment-RTF Office; report no. EPA/600/P-95/001aF-cF. 3v.
EPA (1996b) 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-452VR-96-013.
EPA (2004) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC: National Center for
Environmental Assessment-RTF Office; report no. EPA/600/P-99/002aD.
Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
Perspect. 107:637-641.
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. Boston, MA: Health Effects Institute; pp.
97-106. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Gauderman, W. J.; McConnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; 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, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
Margolis, H. G.; 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.
Gordian, M. E.; Ozkaynak, H.; Xue, J.; Morris, S. S.; Spengler, J. D. (1996) Paniculate air pollution and respiratory
disease in Anchorage, Alaska. Environ. Health Perspect. 104: 290-297.
Hanley, T. (2005) Memorandum to the PM NAAQS review docket, OAR-2001-0017. June 30, 2005.
Hefflin, B. J.; Jalaludin, B.; McClure, E.; Cobb, N.; Johnson, C. A.; Jecha, L.; Etzel, R. A. (1994) Surveillance for
dust storms and respiratory diseases in Washington State, 1991. Arch. Environ. Health 49: 170-174.
5-75
-------
Henderson, R. 2005. EPA's Review of the National Ambient Air Quality Standards for Paniculate Matter (Second
Draft PM Staff Paper, January 2005): A review by the Paniculate Matter Review Panel of the EPA Clean
Air Scientific Advisory Committee. June 6, 2005. Available:
http ://www. epa. gov/sab/pdf/casac-05 -007 .pdf.
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. Boston,
MA: Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[12 May, 2004].
Kleinman, M.T.; Bhalla, O.K.; Mautz, W.J.; Phalen, R.F. 1995. Cellular and immunological injury with PM-10
inhalation. Inhalation Toxicol. 7:589-602.
Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 165-172. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
Langstaff, J. (2004). Estimation of Policy-Relevant Background Concentrations of Paniculate Matter. Memorandum
to PM NAAQS review docket OAR-2001-0017. January 27, 2005.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report 95.
Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect. 108:347-353.
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 [18 October, 2004].
Ostro, B. D.; Hurley, S.; Lipsett, M. J. (1999) Air pollution and daily mortality in the Coachella Valley, California:
a study of PM10 dominated by coarse particles. Environ. Res. 81: 231-238.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10:412-419.
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
[18 October, 2004].
Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999) A study of twelve southern California communities
with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit.
Care Med. 159: 760-767.
Pope, C. A., III. (1989) Respiratory disease associated with community air pollution and a steel mill, Utah Valley.
Am. J. Public Health 79: 623-628.
Pope, C. A., III. (1991) Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and
Cache Valleys. Arch. Environ. Health 46: 90-97.
5-76
-------
Pope, C. A., Ill; Schwartz, I; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah Valley. Arch.
Environ. Health 47:211-217.
Pope, C. A., Ill; Burnett, R. T.; Thun, M. I; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. J. Am. Med. Assoc.
287:1132-1141.
Rogge, W.F.; Hildemann, L.M.; Mazurek, M.A.; Cass, G.R.; Simoneit, B.R.T. 1993. Sources of fine organic
aerosol. 3. Road dust, tire debris, and organometallic brake lining dust: roads as sources and sinks.
Environ. Sci. Technol. 27:1982-1904.
Ross, M. (2005) Updated information on air quality monitoring data for thoracic coarse particles used in
epidemiologic studies. Memorandum to the PM NAAQS review docket, OAR-2001-0017. June 30, 2005.
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.
Schmidt et al. (2005) Draft analysis of PM ambient air quality data for the PM NAAQS review. Memorandum to
PM NAAQS review docket OAR-2001-0017. January 31, 2005.
Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
8: 371-377.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J.
Air Waste Manage. Assoc. 46:927-939.
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.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology 11:6-10.
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
[18 October, 2004].
Steerenberg, P. A.; Withagen, C. E.; Dormans, J. A. M. A.; VanDalen, W. J.; VanLoveren, H.; Casee, F. R. (2003)
Adjuvant activity of various diesel exhaust and ambient particle in two allergic models. J. Toxicol. Environ.
Health A 66: 1421-1439.
5-77
-------
-------
6. POLICY-RELEVANT ASSESSMENT OF PM-RELATED WELFARE EFFECTS
6.1 INTRODUCTION
This chapter assesses key policy-relevant information on the known and potential effects
on public welfare associated with ambient PM, alone and in combination with other pollutants
commonly present in the ambient air. It draws upon the most relevant information contained in
the CD, as well as other significant reports referenced therein. The welfare effects to be
considered in this review of the secondary PM NAAQS include effects on visibility (section 6.2),
vegetation and ecosystems (section 6.3), materials (section 6.4), and climate change processes1
(section 6.5). For each category of effects, this chapter presents a summary of the relevant
scientific information and a staff assessment of whether the available information is sufficient to
be considered as the basis for secondary standards distinct from primary standards for PM. Staff
conclusions and recommendations related to secondary standards for PM are presented in
Chapter 7.
It is important to note that discussion of PM-related effects on visibility, vegetation and
ecosystems, and climate change processes in Chapters 4 and 9 of the CD builds upon and
includes by reference extensive information from several other significant scientific reviews of
these topics. Most notably, these reports include the Recommendations of the Grand Canyon
Visibility Transport Commission (1996), the National Research Council's Protecting Visibility
in National Parks and Wilderness Areas (1993), reports of the National Acid Precipitation
Assessment Program (1991, 1998), previous EPA Criteria Documents, including Air Quality
Criteria for Paniculate Matter and Sulfur Oxides (EPA, 1982) and Air Quality Criteria for
Oxides of Nitrogen (EPA, 1993), recent reports of the National Academy of Sciences (NAS,
2001) and the Intergovernmental Panel on Climate Change (IPCC, 1998, 2001a,b). In addition,
numerous other U.S. and international assessments of stratospheric ozone depletion and global
climate change carried out under U.S. Federal interagency programs (e.g., the U.S. Global
Climate Change Research Program), the World Meteorological Organization (WMO), and the
United Nations Environment Programme (UNEP) have been reviewed.
6.2 EFFECTS ON VISIBILITY
Visibility can be defined as the degree to which the atmosphere is transparent to visible
light (NRC, 1993; CD, 4-153). Visibility impairment is the most noticeable effect of fine
particles present in the atmosphere. Particle pollution degrades the visual appearance and
1 In assessing information on PM-related effects on climate change processes, consideration is given to
potential indirect impacts on human health and the environment that may be a consequence of changes in climate
and solar radiation attributable to changes in ambient PM.
6-1
-------
perceived color of distant objects to an observer and reduces the range at which they can be
distinguished from the background.
This section discusses the role of ambient PM in the impairment of visibility, drawing
upon the most relevant information contained in the CD (Chapters 4 and 9), as well as significant
reports on the science of visibility referenced therein, and building upon information presented in
section 2.8 of this document. In particular, this section includes new information on the
following topics:
• Summary findings of analyses of hourly PM2 5 measurements and reconstructed light
extinction coefficients for urban areas, for 2003, that demonstrate a significant
correlation between PM2 5 and light extinction across the U.S. during daylight hours.
An overview of visibility programs, goals, and methods for the evaluation of visibility
impairment as a basis for standard setting, in the U.S. and abroad, illustrating the
significant value placed on visual air quality, as demonstrated by efforts to improve
visibility in national parks and wilderness areas, as well as in urban areas.
This section summarizes available information as follows: (1) information on the general
types of visibility impairment; (2) trends and conditions in Class I and non-urban areas; (3)
visibility conditions in urban areas; (4) studies of the economic value of improving visual air
quality; (5) current policy approaches to addressing visibility impairment; and (6) approaches to
evaluating public perceptions of visibility impairment and judgments about the acceptability of
varying degrees of visibility impairment.
6.2.1 Overview of Visibility Impairment
Visibility impairment is manifested in two principal ways: as local visibility impairment
(e.g., localized plumes or "brown clouds") and as regional haze. In some cases, local-scale
visibility degradation is considered to be "reasonably attributable" to a single source or small
group of sources. Such impairment may take the form of a localized plume, a band or layer of
discoloration appearing well above the terrain that results from complex local meteorological
conditions. Localized plumes are composed of smoke, dust, or colored gas that obscures the sky
or horizon relatively near sources. Sources of locally visible plumes, such as the plume from an
industrial facility or a burning field, are often easy to identify. Historically, sources of visible
plumes were thought to be relatively minor contributors to visibility impairment in Class I areas
(i.e., 156 national parks, wilderness areas, and international parks identified for visibility
protection in section 162(a) of the Act). However, there have been a limited number of cases in
6-2
-------
which Federal land managers have certified the existence of visibility impairment in a Class I
area as being "reasonably attributable" to a particular source.2
In other cases, localized visibility impairment is manifested as an urban haze, sometimes
referred to as a "brown cloud." This type of impairment is predominantly caused by emissions
from multiple sources in the urban area and is not typically attributable to a single nearby source
or to long-range transport from more distant sources. Brown cloud conditions have been well-
documented in a number of western cities.
The second type of impairment, regional haze, generally results from pollutant emissions
from a multitude of sources located across a broad geographic region. Regional haze impairs
visibility in every direction over a large area, in some cases over multi-state regions. It also
masks objects on the horizon and reduces the contrast of nearby objects. The formation, extent,
and intensity of regional haze are functions of meteorological and chemical processes, which
sometimes cause fine particle loadings to remain suspended in the atmosphere for several days
and to be transported hundreds of kilometers from their sources (NRC, 1993). It is this second
type of visibility degradation, regional haze, that is principally responsible for impairment in
national parks and wilderness areas across the country (NRC, 1993).
While visibility impairment in urban areas at times may be dominated by local sources, it
often may be significantly affected by long-range transport of haze due to the multi-day
residence times of fine particles in the atmosphere. Fine particles transported from urban and
industrialized areas, in turn, may, in some cases, be significant contributors to regional-scale
impairment in Class I and other rural areas.
6.2.2 Visibility Trends and Current Conditions in Class I and Non-Urban Areas
In conjunction with the National Park Service, other Federal land managers, and State
organizations, EPA has supported visibility monitoring in national parks and wilderness areas
since 1988. The monitoring network was originally established at 20 sites, but it has now been
expanded to 110 sites that represent all but one (Bering Sea) of the 156 mandatory Federal Class
I areas across the country. This long-term visibility monitoring network is known as IMPROVE
(Interagency Monitoring of PROtected Visual Environments).
IMPROVE provides direct measurement of fine particles that contribute to visibility
impairment. The IMPROVE network employs aerosol measurements at all sites, and optical and
scene measurements at some of the sites. Aerosol measurements are taken for PM10 and PM25
mass, and for key constituents of PM2 5, such as sulfate, nitrate, organic and elemental carbon,
soil dust, and several other elements. Measurements for specific aerosol constituents are used to
Two of the most notable cases leading to emission controls involved the Navajo Generating Station in
Arizona and the Mohave power plant in Nevada. For both plants, it was found that sulfur dioxide emissions were
contributing to visibility impairment in Grand Canyon National Park.
6-3
-------
calculate "reconstructed" aerosol light extinction by multiplying the mass for each constituent by
its empirically-derived scattering and/or absorption efficiency, with adjustment for the relative
humidity. Knowledge of the main constituents of a site's light extinction "budget" is critical for
source apportionment and control strategy development. Optical measurements are used to
directly measure light extinction or its components. Such measurements are taken principally
with either a transmissometer, which measures total light extinction, or a nephelometer, which
measures particle scattering (the largest human-caused component of total extinction). Scene
characteristics are typically recorded 3 times daily with 35 millimeter photography and are used
to determine the quality of visibility conditions (such as effects on color and contrast) associated
with specific levels of light extinction as measured under both direct and aerosol-related
methods. Directly measured light extinction is used under the IMPROVE protocol to cross-
check that the aerosol-derived light extinction levels are reasonable in establishing current
visibility conditions. Aerosol-derived light extinction is used to document spatial and temporal
trends and to determine how proposed changes in atmospheric constituents would affect future
visibility conditions.
Annual average visibility conditions (reflecting light extinction due to both
anthropogenic and non-anthropogenic sources) vary regionally across the U.S. The rural East
generally has higher levels of impairment than remote sites in the West, with the exception of
urban-influenced sites such as San Gorgonio Wilderness (CA) and Point Reyes National
Seashore (CA), which have annual average levels comparable to certain sites in the Northeast.
Regional differences are illustrated by Figures 4-39a and 4-39b in the CD, which show that, for
Class I areas, visibility levels on the 20% haziest days in the West are about equal to levels on
the 20% best days in the East (CD, p. 4-179).
Higher visibility impairment levels in the East are due to generally higher concentrations
of anthropogenic fine particles, particularly sulfates, and higher average relative humidity levels.
In fact, sulfates account for 60-86% of the haziness in eastern sites (CD, p. 4-236). Aerosol light
extinction due to sulfate on the 20% haziest days is significantly larger in eastern Class I areas as
compared to western areas (CD, p. 4-182; Figures 4-40a and 4-40b). With the exception of
remote sites in the northwestern U.S., visibility is typically worse in the summer months. This is
particularly true in the Appalachian region, where average light extinction in the summer
exceeds the annual average by 40% (Sisler et al., 1996).
Regional trends in Class I area visibility are updated and presented in the EPA's National
Air Quality and Emissions Trends Report (EPA, 2001). Eastern trends for the 20% haziest days
from 1992-1999 showed a 1.5 deciview improvement, or about a 16% improvement. However,
visibility in the East remains significantly impaired, with an average visual range of
approximately 20 km on the 20% haziest days. In western Class I areas, aggregate trends
showed little change during 1990-1999 for the 20% haziest days, and modest improvements on
6-4
-------
the 20% mid-range and clearest days. Average visual range on the 20% haziest days in western
Class I areas is approximately 100 km.
6.2.3 Visibility Conditions in Urban Areas
Urban visibility impairment results from the combined effect of stationary, mobile, and
area source emissions. Complex local meteorological conditions may contribute to such
impairment as well. Localized or layered haze often results from emissions from many sources
located across an urban or metropolitan area. A common manifestation of this type of visibility
impairment is the "brown cloud" situation experienced in some cities particularly in the winter
months, when cooler temperatures limit vertical mixing of the atmosphere. The long-range
transport of emissions from sources outside the urban area may also contribute to urban haze
levels.
Visibility impairment has been studied in several major cities in the past decades because
of concerns about fine particles and their potentially significant impacts (e.g., health-related and
aesthetic) on the residents of large metropolitan areas (e.g., Middleton, 1993). Urban areas
generally have higher loadings of PM2 5 and, thus, higher visibility impairment than monitored
Class I areas. As discussed in Chapter 2, sections 2.4 and 2.5, annual mean levels of 24-hour
average PM2 5 levels are generally higher in urban areas than those found in the IMPROVE
database for rural Class I areas. Urban areas have higher concentrations of organic carbon,
elemental carbon, and particulate nitrate than rural areas due to a higher density of fuel
combustion and diesel emissions.
6.2.3.1 ASOS Airport Visibility Monitoring Network
For many years, urban visibility has been characterized using data describing airport
visibility conditions. Until the mid-1990's, airport visibility was typically reported on an hourly
basis by human observers. An extensive database of these assessments has been maintained and
analyzed to characterize visibility trends from the late-1940's to mid-1990's (Schichtel et al.,
2001).
In 1992, the National Weather Service (NWS), Federal Aviation Administration (FAA),
and Department of Defense began deployment of the Automated Surface Observing System
(ASOS). ASOS is now the largest instrument-based visibility monitoring network in the U.S.
(CD, p. 4-174). The ASOS visibility monitoring instrument is a forward scatter meter that has
been found to correlate well with light extinction measurements from the Optec transmissometer
(NWS, 1998). It is designed to provide consistent, real-time visibility and meteorological
measurements to assist with air traffic control operations. A total of 569 FAA-sponsored and
313 NWS-sponsored automated observing systems are installed at airports throughout the
country. ASOS visibility data are typically reported for aviation use in small increments up to a
maximum of 10 miles visibility. While these truncated data are not ideal for characterizing
actual visibility levels, the raw, non-truncated data from the 1-minute light extinction and
6-5
-------
meteorological readings are now archived and available for analysis for a subset of the ASOS
sites.3
6.2.3.2 Correlation between Urban Visibility and PM2 5 Mass
In an effort to better characterize urban visibility, staff has analyzed the extensive new
data now available on PM2 5 primarily in urban areas. This rapidly expanding national database,
including FRM measurements of PM25 mass, continuous measurements of hourly PM25 mass,
and PM2 5 chemical speciation measurements, now provides the opportunity to conduct such an
analysis. In this analysis, described below and documented in detail in Schmidt et al. (2005),
staff has sought to explore the factors that have historically complicated efforts to address
visibility impairment nationally, including regional differences related to levels of primarily fine
particles and relative humidity. Taking these factors into account, staff has compared
correlations between visibility, in terms of reconstructed light extinction (using the IMPROVE
methodology discussed in Chapter 2, section 2.8), with hourly PM2 5 concentrations in urban
areas across the U.S. and in eastern and western regions.
As an initial matter, staff has explored the factors contributing to the substantial
East/West differences that have been characterized primarily for Class I areas across the country,
as discussed above in section 6.2.2. In considering fine particle levels, staff notes that East/West
differences are substantially smaller in urban areas than in rural areas. As shown in Figure 6-1,
24-hour average PM2 5 concentrations in the East and West are much more similar in urban areas
than they are in rural areas. A significantly lower East/West ratio is observed in urban areas,
based on data from either the FRM or the EPA Speciation Network, than in rural areas, based on
data from the IMPROVE network.
In considering relative humidity levels, staff notes that, while the average daily relative
humidity levels are generally higher in eastern than western areas, in both regions relative
humidity levels are appreciably lower during daylight as compared to night time hours. These
differences can be seen in Figure 6-2, based on data from National Weather Service (NWS) sites.
As discussed in Chapter 2, section 2.8, the reconstructed light extinction coefficient, for a given
mass and concentration, increases sharply as relative humidity rises. Thus, visibility impacts
related to East/West differences in average relative humidity are minimized during daylight
hours, when relative humidity is generally lower.
Taking these factors into account, staff has considered both 24-hour and shorter-term
daylight hour averaging periods in evaluating correlations between PM2 5 concentrations in urban
areas and visibility, in terms of reconstructed light extinction (RE), in eastern and western areas,
3 A preliminary analysis of the archived data for 63 cities across the U.S. was presented in the first draft
Staff Paper (August 2003), but further analysis has not been conducted. While the preliminary analysis
demonstrated relatively well-characterized correlations between predicted PM2 5 concentrations (based on ASOS
extinction values) and measured PM2 5 concentrations in some urban areas, such correlations were not consistently
observed in urban areas across the country.
6-6
-------
o
O
O
o
14
12
10
8
6
4
2
Rural east/west ratio = 2.04
Urban east/west ratio = 1.29
Rural - IMPROVE (24-hr)
East
Urban - FRM (24-hr)
West
Note: Urban IMPROVE sites and rural FRM sites excluded.
Figure 6-1. PM2 5 concentration differences between eastern and western
areas and between rural and urban areas for 2003.
Source: Schmidt et al. (2005)
6-7
-------
1001 TJ^^^^^^
90
80
70
60 ^_ __ __ __
50 4- I X-
40
30
20
10 i
' East
13 0 1 2 3 4 5 6 7 8 91011121314151617181920212223 24-hr
^ avg.
(D 100
•^3 90
^3
(D 80
^- 70
60
50
40
30
20
10
0, West
01234567
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24-hr
Hour
avg.
Figure 6-2. Distribution of hourly and 24-hour average relative humidity at eastern and western U.S.
National Weather Service Sites, 2003. Box depicts interquartile range and median; whiskers
depict 5th and 95thpercentiles.
Source: Schmidt et al. (2005) 6-8
-------
as well as nationwide. Figure 6-3 shows clear and similarly strong correlations between RE and
24-hour average PM2 5 in eastern, western, and all urban areas. Figure 6-3 is based on data from
161 urban continuous PM25 mass monitoring sites across the country with co-located or nearby
24-hour PM2 5 speciation data. RE values were calculated based on a constructed hourly
speciated PM2 5 data set, hourly relative humidity data (either co-located or from nearby NWS
sites), and a coarse PM data set (estimated either by difference method from the continuous
PM2 5 and co-located continuous PM10 instruments, or based on regional ratios of PM fractions)
(Schmidt et al., 2005). In calculating RE, the relative humidity was capped at 95%, reflecting
the lack of accuracy in higher relative humidity values and their highly disproportionate impact
on RE.
For these analyses, staff has considered both 10 years of relative humidity data,
converted to 10-year average hourly f(RH)4 values (Figure 6-3, panel a), as well as actual hourly
relative humidity data for 2003, converted to f(RH) values (Figure 6-3, panel b). Of the two sets
of data, Staff recognizes that 10-year average hourly f(RH) data are more reflective of long-term
humidity patterns, and may provide a more appropriate basis for relating ambient PM2 5 levels to
visibility impairment in the context of consideration of a potential secondary standard to protect
against PM-related visibility impairment. On the other hand, since there can be significant day-
to-day variance in relative humidity that is not reflected in long-term average f(RH) data, actual
hourly f(RH) data were also included in the analyses, to reflect the potential ranges of high and
low relative humidity levels likely to occur over the course of a year.
In considering shorter-term daylight hour averaging periods, staff also evaluated the
slope and strength of the correlations between RE and PM25 concentrations on an hourly basis
(Schmidt et al., 2005). Figure 6-4 shows plots of the average slope of the correlation between
hourly RE and corresponding PM25 concentrations (i.e., the increase in RE due to the
incremental increase in PM2 5) by region, in eastern and western areas, and nationwide. The
slopes are all lower during daytime hours when the disproportionate effects of relative humidity
on the light extinction coefficients for fine particle sulfates and nitrates are diminished. Thus,
during daylight hours, the slope more closely represents the influence of PM2 5 mass on visibility
than the influence of relative humidity. In addition, Figure 6-4 shows that the slopes (and hence,
the relationships between RE and PM2 5) are more comparable across regions during daylight
hours. In considering the strength of these correlations, staff notes that the correlations between
RE and PM2 5, as indicated by the model R2 values, are strong for individual daylight hours,
similar to that for the 24-hour average (Schmidt et al., 2005). On a national basis, daytime (9
a.m. to 6 p.m.) hourly model R2 values are all above 0.6 for the RE's calculated with actual f(RH)
4 f(KH) is the relative humidity adjustment factor; it increases significantly with higher humidity. See
section 2.8.1 and Chapter 4 of the CD (CD, pp. 4-149 to 4-170) for further information.
6-9
-------
800:
600-
500-
300:
200-
a
East (circles): RE = 7.3 * PM2 5 + 8.5; R2=0.94
West (stars): RE = 6.8 *PM25 + 4.5; R2=0.83
A11:RE = 7.2*PM25 +7.8; R2=0.91
10
20
30
40
50
GO
70
80
90
100 110 120 130
140
b
East (circles): RE = 7.8 * PM2 5 + 8.5; R2=0.70
West (stars): RE = 6.9 * PM25+4.1; R2=0.72
All: RE = 7.6 * PM2 5 + 7.9; R2=0.70
100 110 120 130
PM2 5 (ug/m3)
Figure 6-3. Relationship between reconstructed light extinction
(RE) and 24-hour average PM2 5, 2003. RE in top panel
(a) computed with 10-year average f(RH)', RE in bottom
panel (b) computed using actual f(RH).
Source: Schmidt et al. (2005)
6-10
-------
11
10
0)
Q.
0
8
o
II 6
LU
0
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Northeast
Southeast
Industrial Midwest
Upper Midwest
Southwest
Northwest
Southern California " - ' West
East ^ U.S.
Figure 6-4. Model slope for relationship between reconstructed light extinction (RE) and hourly PM2 5
(increase in RE due to incremental increase in PM25), 2003. RE computed using 10-year
average/^?//,).
Source: Schmidt et al. (2005)
6-11
-------
values and above 0.8 for the RE's calculated with 10-year average f(RH) values (Schmidt et al.,
2005).
On the basis of lower slopes and more inter-region comparability, staff selected a number
of daylight time periods to consider in evaluating additional correlations between PM2 5
concentrations and RE in eastern and western regions, as well as nationwide. Evaluated time
periods included 7 a.m. to 7 p.m.; 9 a.m. to 5 p.m.; 10 a.m. to 6 p.m.; 10 a.m. to 4 p.m.; 12 p.m.
to 4 p.m.; and 8 a.m. to 4 p.m. With a focus on minimizing slope, minimizing regional and
East/West slope differences, maximizing R2 values, and considering other related factors, staff
selected the 12 p.m. to 4 p.m. time period for further analyses (Schmidt et al., 2005).
Using the same data as were used for Figure 6-3, Figure 6-5 shows examples of the
correlations between RE and PM25 concentrations averaged over a 4-hour time period, for 10-
year average hourly f(RH) data (panel a) and for actual hourly f(RH) data in 2003 (panel b). As
seen in this figure, the correlations between RE and PM2 5 concentrations during daylight hours
in urban areas are comparably strong (similar R2 values), yet more reflective of PM2 5 mass rather
than relative humidity effects (i.e., lower slopes), in comparison to the correlations based on a
24-hour averaging time. Further, these correlations in urban areas are generally similar in the
East and West, in sharp contrast to the East/West differences observed in rural areas.
6.2.4 Economic and Societal Value of Improving Visual Air Quality
Visibility is an air quality-related value having direct significance to people's enjoyment
of daily activities in all parts of the country. Survey research on public awareness of visual air
quality using direct questioning typically reveals that 80% or more of the respondents are aware
of poor visual air quality (Cohen et al., 1986). The importance of visual air quality to public
welfare across the country has been demonstrated by a number of studies designed to quantify
the benefits (or willingness to pay) associated with potential improvements in visibility
(Chestnut and Dennis, 1997; Chestnut and Rowe, 1991).
Individuals value good visibility for the sense of well-being it provides them directly,
both in the places where they live and work, and in the places where they enjoy recreational
opportunities. Millions of Americans appreciate the scenic vistas in national parks and
wilderness areas each year. Visitors consistently rate "clean, clear air" as one of the most
important features desired in visiting these areas (Department of Interior, 1998). A 1998 survey
of 590 representative households by researchers at Colorado State University found that 88% of
the respondents believed that "preserving America's most significant places for future
generations" is very important, and 87% of the respondents supported efforts to clean up air
pollution that impacts national parks (Hass and Wakefield, 1998).
Economists have performed many studies in an attempt to quantify the economic benefits
associated with improvements in current visibility conditions both in national parks and in urban
areas (Chestnut and Dennis, 1997). These economic benefits are often described by economists
6-12
-------
800:
500-
400"
200:
100-
East (circles): RE = 5.4 * PM
West (stars): RE = 5.7 * PM,, + 8.6; R2=0.82
10 20 30 40 50 GO 70 80 90 100 110 120 130 140
East (circles): RE = 6.0 * PM2 5 + 10.6; R2=0.66
West (stars): RE = 5.9 * PM25+ 8.9; R2=0.74
All: RE = 5.9 * PM9, + 10.3; R2=0.68
0 10 20 30 40 50 GO 70 BO 90 100 110 120 130 140
PM2 5 (ug/m3)
Figure 6-5. Relationship between reconstructed light extinction
(RE) and 12 p.m.- 4 p.m. average PM2 5, 2003. RE in
top panel (a) computed with 10-year aver age f(RH); RE
in bottom panel (b) computed using actual f(RH).
Source: Schmidt et al. (2005)
6-13
-------
as either use values or non-use values. Use values are those aspects of environmental quality
that directly affect an individual's welfare. These include improved aesthetics during daily
activities (e.g., driving or walking, looking out windows, daily recreations), for special activities
(e.g., visiting parks and scenic vistas, hiking, hunting), and for viewing scenic photography. Use
benefits of better visibility also include improved road and air safety.
Non-use values are those for which an individual is willing to pay for reasons that do not
relate to the direct use or enjoyment of any environmental benefit. The component of non-use
value that is related to the use of the resource by others in the future is referred to as the bequest
value. This value is typically thought of as altruistic in nature. Another potential component of
non-use value is the value that is related to preservation of the resource for its own sake, even if
there is no human use of the resource. This component of non-use value is sometimes referred to
as existence value or preservation value. Non-use values are not traded, directly or indirectly, in
markets. For this reason, the estimation of non-use values has proved to be more difficult than
the estimation of use values. Non-use values may be related to the desire that a clean
environment be available for the use of others now and in the future, or they may be related to
the desire to know that the resource is being preserved for its own sake, regardless of human use.
Non-use values may be a more important component of value for recreational areas, particularly
national parks and monuments, and for wilderness areas.
In addition, staff notes that the concept of option value is a key component of the
measured values. The option value represents the value that is tied to preserving improved
visibility in the event of a visit, even though a visit is not certain. This component is considered
by some as a use value and by others as a non-use value.
Tourism in the U.S. is a significant contributor to the economy. A 1998 Department of
Interior study found that travel-related expenditures by national park visitors alone average $14.5
billion annually (1996 dollars) and support 210,000 jobs (Peacock et al., 1998). A similar
estimate of economic benefits resulting from visitation to national forests and other public lands
could increase this estimate significantly.
It is well recognized in the U.S. and abroad that there is an important relationship
between good air quality and economic benefits due to tourism. McNeill and Roberge (2000)
studied the impact of poor visibility episodes on tourism revenues in Greater Vancouver and the
Lower Fraser Valley in British Columbia as part of the Georgia Basin Ecosystem Initiative of
Environment Canada. Through this analysis, a model was developed that predicts future tourist
revenue losses that would result from a single extreme visibility episode. They found that such
an episode would result in a $7.45 million loss in the Greater Vancouver area and $1.32 million
loss in the Fraser Valley.
The results of several valuation studies addressing both urban and rural visibility are
presented in the CD (CD, pp. 4-187 to 4-190), the 1996 Criteria Document (EPA, 1996a, p. 8-83,
Table 8-5; p. 8-85, Table 8-6) and in Chestnut and Rowe (1991) and Chestnut et al. (1994). Past
6-14
-------
studies by Schulze et al. (1983) and Chestnut and Rowe (1990) have estimated the preservation
values associated with improving the visibility in national parks in the Southwest to be in the
range of approximately $2-6 billion annually. An analysis of the residential visibility benefits in
the eastern U.S. due to reduced sulfur dioxide emissions under the acid rain program suggests an
annual value of $2.3 billion (in 1994 dollars) in the year 2010 (Chestnut and Dennis, 1997). The
authors suggest that these results could be as much as $1-2 billion more because the above
estimate does not include any value placed on eastern air quality improvements by households in
the western U.S.
Estimating benefits for improvements in visibility can be difficult because visibility is not
directly or indirectly valued in markets. Many of the studies cited above are based on a
valuation method known as contingent valuation (CV). Concerns have been identified about the
reliability of value estimates from contingent valuation studies because research has shown that
bias can be introduced easily into these studies if they are not carefully conducted. Accurately
estimating willingness-to-pay for avoided health and welfare losses depends on the reliability
and validity of the data collected. However, there is an extensive scientific literature and body of
practice on both the theory and technique of contingent valuation. EPA believes that well-
designed and well-executed CV studies are useful for estimating the benefits of environmental
effects such as improved visibility (EPA, 2000).
Some of the studies cited above used an alternative valuation method known as hedonic
pricing. Hedonic pricing is a technique used to measure components of property value (e.g.,
proximity to schools). It relies on the measurement of differentials in property values under
various environmental quality conditions, including air pollution and environmental amenities,
such as aesthetic views. This method works by analyzing the way that market prices change
with changes in environmental quality or amenity. EPA believes that well-designed and well-
executed hedonic valuation studies, in combination with public perception surveys, are useful for
estimating the benefits of environmental effects such as improved visibility.
Society also values visibility because of the significant role it plays in transportation
safety. Serious episodes of visibility impairment can increase the risk of unsafe air
transportation, particularly in urban areas with high air traffic levels (EPA, 1982). In some
cases, extreme haze episodes have led to flight delays or the shutdown of major airports,
resulting in economic impacts on air carriers, related businesses, and air travelers. For example,
on May 15, 1998 in St. Louis, Missouri, it was reported that a haze episode attributed to
wildfires in central America resulted in a reduction in landing rates and significant flight delays
at Lambert International Airport. The 24-hour PM2 5 levels reached 68 |ig/m3 during that
episode. In addition, the National Transportation Safety Board (NTSB) has concluded in
accident reports that high levels of pollution and haze, such as those experienced during the July
1999 air pollution episode in the northeastern U.S., have played a role in air transportation
6-15
-------
accidents and loss of life (NTSB, 2000). During this episode, 24-hour levels of PM25 ranged
from 35-52 |ig/m3 in the New England states.
6.2.5 Programs and Goals for Improving Visual Air Quality
Specific discussion is provided below on local, State, and international efforts to protect
visual air quality.
6.2.5.1 Regional Protection
Due to differences in visibility impairment levels (as a result of differences in chemical
composition of haze and in relative humidity levels) between the East and West, EPA, land
managers, and States have taken a regional approach, rather than a national approach, to
protecting visibility in non-urban areas in the U.S.. Protection against visibility impairment in
special areas is provided for in sections 169A, 169B, and 165 of the Act, in addition to that
provided by the secondary NAAQS. Section 169A, added by the 1977 CAA Amendments,
established a national visibility goal to "remedy existing impairment and prevent future
impairment" in 156 national parks and wilderness areas (Class I areas). The Amendments also
called for EPA to issue regulations requiring States to develop long-term strategies to make
"reasonable progress" toward the national goal. EPA issued initial regulations in 1980 focusing
on visibility problems that could be linked to a single source or small group of sources. Action
was deferred on regional haze until monitoring, modeling, and source apportionment methods
could be improved.
The 1990 CAA Amendments placed additional emphasis on regional haze issues through
the addition of section 169B. In accordance with this section, EPA established the Grand
Canyon Visibility Transport Commission (GCVTC) in 1991 to address adverse visibility impacts
on 16 Class I national parks and wilderness areas on the Colorado Plateau. The GCVTC was
comprised of the Governors of nine western states and leaders from a number of Tribal nations.
The GCVTC issued its recommendations to EPA in 1996, triggering a requirement in section
169B for EPA issuance of regional haze regulations.
EPA accordingly promulgated a final regional haze rule in 1999 (EPA, 1999; 65 FR
35713). Under the regional haze program, States are required to establish goals for improving
visibility on the 20% most impaired days in each Class I area, and for allowing no degradation
on the 20% least impaired days. Each state must also adopt emission reduction strategies which,
in combination with the strategies of contributing States, assure that Class I area visibility
improvement goals are met. The first State implementation plans are to be adopted in the 2003-
2008 time period, with the first implementation period extending until 2018. Five multistate
planning organizations are evaluating the sources of PM2 5 contributing to Class I area visibility
impairment to lay the technical foundation for developing strategies, coordinated among many
States, in order to make reasonable progress in Class I areas across the country.
6-16
-------
6.2.5.2 Local, State, and International Goals and Programs
A number of programs, goals, standards, and planning efforts have been established in
the U.S. and abroad to address visibility concerns in urban and non-urban areas. These
regulatory and planning activities are of particular interest because they are illustrative of the
significant value that the public places on improving visibility, and because they have made use
of developed methods for evaluating public perceptions and judgments about the acceptability of
varying degrees of visibility impairment.
Several state and local governments have developed programs to improve visual air
quality in specific urban areas, including Denver, CO; Phoenix, AZ; and, Lake Tahoe, CA. At
least two States have established statewide standards to protect visibility. In addition, visibility
protection efforts have been undertaken in other countries, including Australia, New Zealand,
and Canada. Examples of these efforts are highlighted below.
In 1990, the State of Colorado adopted a visibility standard for the city of Denver. The
Denver standard is a short-term standard that establishes a limit of a four-hour average light
extinction level of 76 Mm"1 (equivalent to a visual range of approximately 50 km) during the
hours between 8 a.m. and 4 p.m. (Ely et al., 1991). In 2003, the Arizona Department of
Environmental Quality created the Phoenix Region Visibility Index, which focuses on an
averaging time of 4 hours during actual daylight hours. This visiblity index establishes visual air
quality categories (i.e., excellent to very poor) and establishes the goals of moving days in the
poor/very poor categories up to the fair category, and moving days in the fair category up to the
good/excellent categories (Arizona Department of Environmental Quality, 2003). This approach
results in a focus on improving visibility to a visual range of approximately 48-36 km. In 1989,
the state of California revised the visibility standard for the Lake Tahoe Air Basin and
established an 8-hour visibility standard equal to a visual range of 30 miles (approximately 48
km) (California Code of Regulations).
California and Vermont each have standards to protect visibility, though they are based
on different measures. Since 1959, the state of California has had an air quality standard for
particle pollution where the "adverse" level was defined as the "level at which there will be ...
reduction in visibility or similar effects." California's general statewide visibility standard is a
visual range of 10 miles (approximately 16 km) (California Code of Regulations). In 1985,
Vermont established a state visibility standard that is expressed as a summer seasonal sulfate
concentration of 2 |ig/m3, that equates to a visual range of approximately 50 km. This standard
was established to represent "reasonable progress toward attaining the congressional visibility
goal for the Class 1 Lye Brook National Wilderness Area, and applies to this Class 1 area and to
all other areas of the state with elevations greater than 2500 ft.
Outside of the U.S., efforts have also been made to protect visibility. The Australian
state of Victoria has established a visibility objective (State Government of Victoria, 2000a and
2000b), and a visibility guideline is under consideration in New Zealand (New Zealand National
6-17
-------
Institute of Water & Atmospheric Research, 2000a and 2000b; New Zealand Ministry of
Environment, 2000). A survey was undertaken for the Lower Fraser Valley in British Columbia,
with responses from this pilot study being supportive of a standard in terms of a visual range of
approximately 40 km for the suburban township of Chilliwack and 60 km for the suburban
township of Abbotsford, although no visibility standard has been adopted for the Lower Fraser
Valley at this time.
6.2.6 Approaches to Evaluating Public Perceptions and Attitudes
New methods and tools have been developed to communicate and evaluate public
perceptions of varying visual effects associated with alternative levels of visibility impairment
relative to varying pollution levels and environmental conditions. New survey methods have
been applied and evaluated in various studies, such as those for Denver, Phoenix, and the Lower
Fraser Valley in British Columbia, and these studies are described below in more detail. These
methods are intended to assess public perceptions as to the acceptability of varying levels of
visual air quality, considered in these studies to be an appropriate basis for developing goals and
standards for visibility protection. For the Denver and British Columbia studies, actual slides
taken in the areas of interest, and matched with transmissometer and nephelometer readings,
respectively, were used to assess public perceptions about visual air quality. For the Phoenix
study, WinHaze, a newly available image modeling program, discussed below, was used for
simulating images. Staff finds that, even with variations in each study's approaches, the survey
methods used for the Denver, Phoenix, and British Columbia studies produced reasonably
consistent results from location to location, each with a majority of participants finding visual
ranges within about 40 to 60 km to be acceptable.
6.2.6.1 Photographic Representations of Visual Air Quality
In the past, the principal method for recording and describing visual air quality, for the
purpose of public perception surveys, has been through 35 millimeter photographs. Under the
IMPROVE program, EPA, federal land management agencies, and Air Resource Specialists, Inc.
(ARS) have developed an extensive archive of visual air quality photos for national parks and
wilderness areas. In comparison, we have only a limited archive of photos of urban areas.
The CD discusses some of the methods that are now available to represent different
levels of visual air quality (CD, p. 4-174). In particular, Molenar et al. (1994) describes a
sophisticated visual air quality simulation technique, incorporated into the WinHaze program
developed by ARS, which combined various modeling systems under development for the past
20 years. The technique relies on first obtaining an original base image slide of the scene of
interest. The slide should be of a cloudless sky under the cleanest air quality conditions possible.
The light extinction represented by the scene should be derived from aerosol and optical data
associated with the day the image was taken, or it should be estimated from contrast
measurements of features in the image. The image is then digitized to assign an optical density
6-18
-------
to each pixel. At this point, the radiance level for each pixel is estimated. Using a detailed
topographic map, technicians identify the specific location from which the photo was taken, and
they determine the distances to various landmarks and objects in the scene. With this
information, a specific distance and elevation is assigned to each pixel.
Using the digital imaging information, the system then computes the physical and optical
properties of an assumed aerosol mix. These properties are input into a radiative transfer model
in order to simulate the optical properties of varying pollutant concentrations on the scene.
WinHaze, an image modeling program for personal computers that employs simplified
algorithms based on the sophisticated modeling technique, is now available (Air Resource
Specialists, 2003).
The simulation technique has the advantage of being readily applicable to any location
as long as a very clear base photo is available for that location. In addition, the lack of clouds
and the consistent sun angle in all images, in effect, standardizes the perception of the images
and enables researchers to avoid potentially biased responses due to these factors. An alternative
to using simulated images is to obtain actual photographs of the site of interest at different
ambient pollution levels. However, long-term photo archives of this type exist for only a few
cities. In addition, studies have shown that observers will perceive an image with a cloud-filled
sky as having a higher degree of visibility impairment than one without clouds, even though the
PM concentration on both days is the same.
As part of a pilot study5 in Washington, D.C., both survey and photographic techniques
were applied (Abt Associates, 2001). In conjunction with this pilot project, images that illustrate
visual air quality in Washington, DC under a range of visibility conditions were prepared and are
available at http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html (labeled as Appendix
6A: Images of Visual Air Quality in Selected Urban Areas in the U.S.). Included as part of
Appendix 6A, this website also contains actual photographs of Chicago illustrating visibility
conditions associated with a range of PM25 concentrations, as well as simulated images for
Denver and Phoenix, as discussed below.
6.2.6.2 Survey Methods
Denver, Colorado: Visibility Standard
The process by which the Denver visibility standard was developed relied on citizen
judgments of acceptable and unacceptable levels of visual air quality (Ely et al., 1991).
Representatives from Colorado Department of Public Health and Environment (CDPHE)
conducted a series of meetings with 17 civic and community groups in which a total of 214
individuals were asked to rate slides having varying levels of visual air quality for a well-known
vista in Denver. The CDPHE representatives asked the participants to base their judgments on
5 A small pilot study for Washington, D.C. was conducted by EPA and was briefly discussed in the
preliminary draft staff paper (2001).
6-19
-------
three factors: 1) the standard was for an urban area, not a pristine national park area where the
standards might be more strict; 2) standard violations should be at visual air quality levels
considered to be unreasonable, objectionable, and unacceptable visually; and 3) judgments of
standards violations should be based on visual air quality only, not on health effects.
The participants were shown slides in 3 stages. First, they were shown seven warm-up
slides describing the range of conditions to be presented. Second, they rated 25 randomly-
ordered slides based on a scale of 1 (poor) to 7 (excellent), with 5 duplicates included. Third,
they were asked to judge whether the slide would violate what they would consider to be an
appropriate urban visibility standard (i.e., whether the level of impairment was "acceptable" or
"unacceptable").
The Denver visibility standard setting process produced the following findings:
• Individuals' judgments of a slide's visual air quality and whether the slide violated
a visibility standard are highly correlated (Pearson correlation coefficient greater
than 80%) with the group average.
• When participants judged duplicate slides, group averages of the first and second
ratings were highly correlated.
• Group averages of visual air quality ratings and "standard violations" were highly
correlated. The strong relationship of standard violation judgments with the
visual air quality ratings is cited as the best evidence available from this study for
the validity of standard violation judgments (Ely et al., 1991).
The CDPHE researchers sorted the ratings for each slide by increasing order of light
extinction and calculated the percent of participants that judged each slide to violate the
standard. The Denver visibility standard was then established based on a 50% acceptability
criterion. Under this approach, the standard was identified as the light extinction level that
divides the slides into two groups: those found to be acceptable and those found to be
unacceptable by a majority of study participants. The CDPHE researchers found this level to be
reasonable because, for the slides at this level and above, a majority of the study participants
judged the light extinction levels to be unacceptable. In fact, when researchers evaluated all
citizen judgments made on all slides at this level and above as a single group, more than 85% of
the participants found visibility impairment at and above the level of the selected standard to be
unacceptable.
Though images used in the Denver study were actual photographs, more recently,
WinHaze has been used to generate images that illustrate visual air quality in Denver under a
range of visibility conditions (generally corresponding to 10th, 20th, 30th, 40th, 50th, 60th 80th, and
90th percentile values), and these images are available in Attachment 6-A at
http ://www. epa. gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html.
6-20
-------
Phoenix, Arizona: Visibility Index
In 2002, the Arizona Department of Environmental Quality formed the Visibility Index
Oversight Committee. The Committee's goal was to coordinate the involvement of Phoenix-area
residents in the development of a visibility index. The Phoenix committee patterned its survey
process after the process used by Denver to develop its visibility standard.
The survey included 385 participants in 27 separate sessions. Participants were carefully
recruited to form a sample group that was demographically representative of the larger Phoenix
population. Three sessions were held in Spanish.
Participants were shown a series of 25 images of the same vista of downtown Phoenix,
with South Mountain in the background at a distance of about 40 km. Photographic slides of the
images were developed using the WinHaze program. The visibility impairment levels ranged
from 15 to 35 deciviews (87 to 12 km visual ranges). Participants first rated the randomly-
shown slides on a scale of 1 (unacceptable) to 7 (excellent). Next, the participants rated slides,
again shown in random order, as acceptable or unacceptable. This phase of the survey produced
the following findings:
• At least 90 percent of all participants found visible air quality acceptable between
15 deciviews (87 km visual range) and 20 deciviews (53 km);
At 24 deciviews (36 km), nearly half of all participants thought the visible air
quality was unacceptable; and
By 26 deciviews (29 km), almost three-quarters of participants said it was
unacceptable, with nearly all participants considering levels of 31 deciviews (18
km) and higher to be unacceptable.
The information developed in this survey informed the development of recommendations
by the Visibility Index Oversight Committee for a visibility index for the Phoenix Metropolitan
Area (Arizona Department of Environmental Quality, 2003). A final report of the survey
methods and results is available (BBC Research & Consulting, 2002). The Phoenix survey
demonstrates that the rating methodology developed for gathering citizen input for establishing
the Denver visibility standard can be reliably transferred to another city while relying on updated
imaging technology to simulate a range of visibility impairment levels.
Images used in this study were generated using WinHaze. Similar images, also generated
by WinHaze, which illustrate visual air quality in Phoenix under a range of visibility conditions,
are available in Appendix 6A at http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html.
British Columbia, Canada: Public Perception Survey
In 1993, the REVEAL (Regional Visibility Experimental Assessment in the Lower Fraser
Valley) field study was undertaken to characterize summertime visibility and ambient aerosol
loadings in southwestern British Columbia. In 1994, researchers at the University of British
6-21
-------
Columbia conducted a pilot study on the perception of acceptable visibility conditions in the
area, using photographs and optical measurements taken during the summer of 1993 (Pryor,
1996). The study was based on the methodology used in setting the Denver visibility standard
(Elyetal., 1991).
Participants in the study were shown slides of two suburban locations in British
Columbia: Chilliwack and Abbotsford. After using the same general protocol, Pryor found that
responses from this pilot study would indicate a standard in terms of visual range of
approximately 40 km for Chilliwack and 60 km for Abbotsford. Pryor (1996) discusses some
possible reasons for the variation in standard visibility judgments between the two locations.
Factors discussed include the relative complexity of the scenes, different levels of development
at each location, potential local source influence on site-specific nephelometer data, and
potential bias of the sample population since only students participated. The author expressed
the view that the pilot study reinforced the conclusion that the methodology originally developed
for the Denver standard-setting process is a sound and effective one for obtaining public
participation in a standard-setting process, and that it could be adapted for such use in another
geographic location with only minor modifications (Pryor, 1996).
6.2.7 Summary and Conclusions
The CD and other reports referenced in section 6.2 provide a significant body of
information documenting the effects of PM and its components on atmospheric visibility. Data
on visibility conditions indicate that urban areas generally have higher loadings of PM25 and,
thus, higher visibility impairment than monitored Class I areas.
Data analyses using extensive new monitoring data now available on PM2 5 primarily in
urban areas show a consistently high correlation between hourly PM2 5 data and RE coefficients
for urban areas across regions of the U.S. during daylight hours. These correlations in urban
areas are generally similar in the East and West, in sharp contrast to the East/West differences
observed in rural areas.
The importance of visual air quality to public welfare across the country has been
demonstrated by a number of studies designed to quantify the benefits (or willingness to pay)
associated with potential improvements in visibility. The value placed on protecting visual air
quality is further demonstrated by the existence of a number of programs, goals, standards, and
planning efforts that have been established in the U.S. and abroad to address visibility concerns
in urban and non-urban areas.
In some urban areas, poor visibility has led to more localized efforts to better
characterize, as well as improve, urban visibility conditions. The public perception survey
approach used in the Denver, Phoenix, and British Columbia studies yielded reasonably
consistent results, with each study indicating that a majority of citizens find value in protecting
local visibility to within a visual range of about 40 to 60 km. In the cases of Denver and
6-22
-------
Phoenix, these studies provided the basis for the establishment of their visibility standards and
goals.
Staff believes that the findings of the new data analyses, in combination with recognized
benefits to public welfare of improved visual air quality and an established approach for
determining acceptable visual range, provide a basis for considering revisions to the secondary
PM2 5 standards to protect against PM-related visibility effects in urban areas.
6.3 EFFECTS ON VEGETATION AND ECOSYSTEMS
Information and conclusions regarding what is currently known about the effects of
ambient PM on ecosystems and individual components of ecosystems such as vegetation, soils,
water, and wildlife are discussed in Chapters 4 and 9 of the CD. This section seeks to build upon
and focus this body of science using EPA's ecological risk paradigm in a manner that highlights
the usefulness and policy relevance of the scientific information. In doing so, staff has drawn
from EPA's Guidelines for Ecological Risk Assessment {Guidelines) (EPA, 1998), which
expanded upon the earlier document, Framework for Ecological Risk Assessment (EPA, 1992),
with the goal of improving the quality of ecological risk assessments and increasing the
consistency of assessments across the Agency.
According to the Guidelines document, the three main phases of ecological risk
assessment are problem formulation, analysis, and risk characterization. Problem formulation
includes the integration of the available information on ecosystem stressors (which can include
physical, chemical, and/or biological stressors), their sources, and the effects associated with
exposure of sensitive ecosystem components to each stressor.
During analysis, data are evaluated to determine how exposure to stressors is likely to
occur (exposure profile) and how stressor levels and ecological effects (stressor-response profile)
are related. These products provide the basis for risk characterization.
During risk characterization, the exposure and stressor-response profiles are integrated
through risk estimation. Risk characterization includes a summary of assumptions, scientific
uncertainties, and strengths and limitations of the analyses. The final product is a risk
description in which the results of the integration are presented, including an interpretation of
ecological adversity and description of uncertainty and lines of evidence.
Keeping these goals and guidelines in mind, the information is organized into the
following seven subsections: major ecosystem stressors in PM (6.3.1); direct vegetation effects
of parti culate nitrate and sulfate deposition (6.3.2); ecosystem effects associated with chronic
inputs of reactive nitrogen and acidifying compounds from PM and other sources (6.3.3);
characteristics and location of nutrient and acid sensitive ecosystems within the U.S. (6.3.4);
ecosystem exposures to PM deposition (6.3.5); consideration of critical loads as an approach for
effects characterization and/or as a management tool (6.3.6); and summary and conclusions
(6.3.7).
6-23
-------
This review will also consider and reference where applicable the extent to which PM
affects the essential ecological attributes (EEAs) outlined in the Framework for Assessing and
Reporting on Ecological Condition, recommended by the Ecological Processes and Effects
Committee (EPEC) of EPA's Science Advisory Board (hereafter EPEC Framework; SAB,
2002), as described in subsections 4.2.1 and 4.2.3 of the CD.
6.3.1 Major Ecosystem Stressors in PM
As previously discussed, PM is not a single pollutant, but a heterogeneous mixture of
particles differing in size, origin, and chemical composition. The heterogeneity of PM exists not
only within individual particles or samples from individual sites, but to an even greater extent,
between samples from different sites. Since vegetation and other ecosystem components are
affected more by paniculate 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. Though the chemical constitution of individual particles
can be strongly correlated with size, the relationship between particle size and particle
composition can also be quite complex, making it difficult in most cases to use particle size as a
surrogate for chemistry. Because PM size classes do not necessarily have specific differential
relevance for vegetation or ecosystem effects (Whitby, 1978; EPA, 1996a), it is the opinion of
the staff that an ecologically relevant indicator for PM should be based on one or more chemical
species found in ambient PM. At this time it remains to be determined as to what extent
NAAQS standards focused on a given size fraction would result in reductions of the ecologically
relevant constituents of PM for any given area.
A number of different chemical species found within ambient PM and their effects on
vegetation and ecosystems were discussed in chapter 4 of the PM CD. In particular, the CD
focused on nitrates and sulfates, concluding that these PM constituents are of greatest and most
widespread environmental significance (CD, p. 9-114). Other components of PM, such as dust,
trace metals, and organics, which can at high levels affect plants and other organisms, were also
discussed. However, some of these compounds, such as organics and some metals, are regulated
under separate statutory authorities, e.g., section 112 of the Clean Air Act. Further, 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
recommending a separate secondary NAAQS. Therefore, the remainder of this section will
focus on the effects of paniculate nitrates and sulfates, either individually, in combination,
and/or as contributors to total reactive nitrogen deposition and total deposition of acidifying
compounds, on sensitive ecosystem components and essential ecological attributes, which in
turn, affect overall ecosystem structure and function.
At the outset, it must be recognized that of paniculate nitrogen and sulfur as ecosystem
stressors with the recognition that nitrogen and sulfur in varying amounts are necessary and
6-24
-------
beneficial nutrients for most organisms that make up ecosystems. Optimal amounts of these
nutrients varies among organisms, populations, communities and ecosystems and across seasons
and time scales. An individual of a given species found in different ecosystems might have
different optimum nutrient requirements, and different species within a given ecosystem may be
stimulated, inhibited or unaffected by a given amount of inputs of these nutrients. Therefore, it
is impossible to generalize to all species in all circumstances as to the amount at which inputs of
these nutrients or acidifying compounds become stressors. On the other hand, while some
species are adapted to benefit from additions of nutrients in the short run, they may not continue
to benefit into the future (or may not benefit at the same rate), due to supplies of other nutrients
becoming short and limiting further growth. Further, species that benefit from nutrient additions
often do so to the detriment of their competitors, shifting the delicate balance that has evolved
under more nutrient-limited conditions.
The staff recognizes that the public welfare has benefitted from the use of nitrogen and
sulfur nutrients in fertilizers in managed agricultural and commercial forest settings. The focus
of this review is on identifying risks to sensitive species and ecosystems where unintentional
additions of these atmospherically derived nutrient and acidifying compounds may be forcing
unintended change on the nation's ecosystems and resulting in adverse impacts on essential
ecological attributes including, species shifts, loss of species richness and diversity, impacts on
threatened and endangered species, and alteration of native fire cycles. In these cases, deposited
particulate nitrate and sulfate are appropriately termed ecosystem "stressors".
6.3.2 Direct Vegetation Effects of Particulate Nitrate and Sulfate Deposition
Nitrogen is a critical limiting nutrient for plant growth. The process of photosynthesis
uses approximately 75% of the nitrogen in a plant leaf, and, thus, to a large extent, governs the
utilization of other nutrients such as phosphorus, potassium (CD, p. 4-95). Plants usually absorb
nitrogen (as NH4+ or NO3") through their roots. However, particle deposition of nitrate, together
with other nitrogen-containing gaseous and precipitation-derived sources, can represent a
substantial fraction of total nitrogen reaching vegetation. In nitrogen-limited ecosystems, this
influx of N can act as a fertilizer. Though it is known that foliar uptake of nitrate can occur, the
mechanism of foliar uptake is not well established, and it is not currently possible to distinguish
sources of chemicals deposited as gases or particles using foliar extraction. Since it has proven
difficult to quantify the percentage of nitrogen uptake by leaves that is contributed by ambient
particles, direct foliar effects of nitrogen-containing particles have not been documented. (CD,
pp. 4-69, 4-70).
Sulfur, similar to nitrogen, is an essential plant nutrient that can be deposited on
vegetation in the form of sulfate particles, or be taken up by plants in gaseous form. Greater than
90% of anthropogenic sulfur emissions are as sulfur dioxide (SO2), with most of the remaining
emissions in the form of sulfate. However, sulfur dioxide is rapidly transformed in the
6-25
-------
atmosphere to sulfate, which is approximately 30-fold less phytotoxic than SO2. Low dosages
of sulfur can also serve as a fertilizer, particularly for plants growing in sulfur-deficient soils.
There are only a few field demonstrations of foliar sulfate uptake, however, and the relative
importance of foliar leachate and prior dry-deposited sulfate particles remains difficult to
quantify. Though current levels of sulfate deposition reportedly exceed the capacity of most
vegetative canopies to immobilize the sulfur, sulfate additions in excess of needs do not typically
lead to plant injury (CD, pp. 4-71, 4-72).
Staff therefore concludes that at current ambient levels, risks to vegetation from short
term exposures to dry deposited particulate nitrate or sulfate are low. Additional studies are
needed, however, to determine possible effects of sulfate particles on physiological
characteristics of plants following chronic exposures (CD, p. 4-72).
Though dry deposition of nitrate and sulfate particles does not appear to induce foliar
injury at current ambient exposures, when found in acidifying deposition, such particles do have
the potential to cause direct foliar injury. This is especially true when the acidifying deposition
is in the form of fog and clouds, which may contain solute concentrations many times those
found in rain. In experiments on seedling and sapling trees, both coniferous and deciduous
species showed significant effects on leaf surface structures after exposure to simulated acid rain
or acid mist at pH 3.5, while some species have shown subtle effects at pH 4 and above.
Epicuticular waxes, which function to prevent water loss from plant leaves, can be destroyed by
acid rain in a few weeks, which suggests links between acid precipitation and aging. Due to
their longevity and evergreen foliage, the function of epicuticular wax is more crucial in
conifers. For example, red spruce seedlings, which have been extensively studied, appear to be
more sensitive to acid precipitation (mist and fog) when compared with other species (CD, pp. 4-
72, 4-73). In addition to accelerated weathering of leaf cuticular surfaces, other direct responses
of forest trees to acid precipitation include increased permeability of leaf surfaces to toxic
materials, water, and disease agents; increased leaching of nutrients from foliage; and altered
reproductive processes (CD, p. 4-86). All of these effects serve to weaken trees so that they are
more susceptible to other stresses (e.g., extreme weather, pests, pathogens).
Acid precipitation with levels of acidity associated with the foliar effects described above
are currently found in some locations in the U.S.. For example, in the eastern U.S., the mean
precipitation pH ranges from 4.3 (Pennsylvania and New York) to 4.8 (Maine)(EPA, 2003). It
can be assumed that occult (mist or fog) deposition impacting high elevations more frequently,
would contain even higher concentrations of acidity. Thus, staff concludes that the risks of foliar
injury occurring from acid precipitation is high. The contribution of particulate sulfates and
nitrates to the total acidity found in the acid precipitation impacting eastern vegetation is not
clear.
6-26
-------
6.3.3 Ecosystem Effects of Chronic Inputs of Reactive Nitrogen and Acidifying
Compounds From PM Deposition and Other Sources
Ecosystem-level responses related to PM occur when the effects of PM deposition on the
biological and physical components of ecosystems become sufficiently widespread as to impact
essential ecological attributes such as nutrient cycling and/or shifts in biodiversity. The most
significant PM-related ecosystem-level effects result from long-term cumulative deposition of a
given chemical species (e.g., nitrate) or mix (e.g., acidic and acidifying deposition) that exceeds
the natural buffering or storage capacity of the ecosystem and/or affects the nutrient status of the
ecosystem, usually by indirectly changing soil chemistry, populations of bacteria involved in
nutrient cycling, and/or populations of fungi involved in plant nutrient uptake (CD, pp. 4-90, 4-
91). To understand these effects, long-term, detailed ecosystem or site-specific data usually are
required. The availability of this type of long-term data is limited. The following discussion is
organized according to the speciated effects of PM on ecosystems.
6.3.3.1 Environmental Effects of Reactive Nitrogen (Nr) Deposition
In the environment, nitrogen may be divided into two types: nonreactive, molecular
nitrogen (N2) and reactive nitrogen (Nr). Molecular nitrogen is the most abundant element in the
atmosphere. However, it only becomes available to support the growth of plants and
microorganisms after it is converted into a reactive form. In nature, Nr creation is accomplished
by certain organisms that have developed the capability of converting N2 to biologically active
reduced forms (Galloway and Cowling, 2002; Hornung and Langan, 1999; EPA, 1993). By the
mid-1960's, however, Nr creation through natural terrestrial processes had been overtaken by Nr
creation as a result of human processes (CD, p. 4-95). The deposition of nitrogen in the U.S.
from human activity doubled between 1961 and 1997, with the largest increase occurring in the
1960s and 1970s (CD, p. 4-98). Reactive nitrogen is now accumulating in the environment on
all spatial scales - local, regional and global. The three main sources of anthropogenic Nr are:
(1) the Haber-Bosch process, which converts N2 to Nr to sustain food production and some
industrial activities; (2) widespread cultivation of legumes, rice and other crops that promote the
conversion of N2 to organic nitrogen through biological nitrogen fixation; and (3) combustion of
fossil fuels, which converts both atmospheric N2 and fossil nitrogen to reactive NOX (CD, pp. 4-
95, 4-96; Galloway and Cowling, 2002; Galloway et al., 2003). Though not currently regulated
under the NAAQS program, reduced forms of Nr from food production are estimated to be
approximately 2-4 times larger than emissions of oxidized forms of nitrogen produced during
combustion of fossil fuels. Specifically, Galloway and Cowling (2002) estimate that per capita
Nr creation for the world in the mid-1990s was 20 kg/N/person/year for food production and
only 3.9 kg/N/person/year for energy production (see Table 3).
Currently available forms of reactive nitrogen include inorganic reduced forms (e.g.,
ammonia [NH3] and ammonium [NH4+]), inorganic oxidized forms (e.g., nitrogen oxides [NOX],
6-27
-------
nitric acid [HNO3], nitrous oxide [N2O], and nitrate [NO3"]), and organic compounds (e.g., urea,
amine, proteins, and nucleic acids (CD, p. 4-95).
Emissions of nitrogen oxides from fuel burning increased exponentially froml940 until
the 1970s, leveled off after the passage of the 1970 amendments to the Clean Air Act, and
stabilized at approximately 7 Tg NOX /yr in the late 1990s. However, despite decreases in
emissions from fossil fuel burning industries, emissions from automobiles have increased
approximately 10% since 1970 due to greater total miles driven (Howarth et al., 2002). Some
NOX emissions are transformed into a portion of ambient air PM (particulate nitrate) and
deposited onto sensitive ecosystems.
The term "nitrogen cascade" refers to the sequential transfers and transformations of Nr
molecules as they move from one environmental system or reservoir (atmosphere, biosphere,
hydrosphere) to another, and the multiple linkages that develop among the different ecological
components, as shown in Figure 6-6. Because of these linkages, adding anthropogenic Nr alters
a wide range of biogeochemical processes and exchanges as the Nr moves among the different
environmental reservoirs, with the consequences accumulating through time (Galloway and
Cowling, 2002; Galloway et al., 2003). These changes in the nitrogen cycle are contributing to
both beneficial and detrimental effects to the health and welfare of humans and ecosystems
(Rabalais, 2002; van Egmond et al., 2002; Galloway, 1998).
Large uncertainties, still exist, however, concerning the rates of Nr accumulation in the
various environmental reservoirs. These uncertainties limit our ability to determine the temporal
and spatial distribution of environmental effects for a given input of Nr. These uncertainties are
of particular significance because of the sequential nature of Nr effects on environmental
processes. Reactive nitrogen does not cascade at the same rate through all environmental
systems. The only way to eliminate Nr accumulation and stop the cascade is to convert Nr back
to nonreactive N2 (Galloway et al., 2003).
Some of the more significant detrimental effects resulting from chronic increased inputs
of atmospheric Nr (e.g., ammonium and nitrate compounds) include: (1) decreased productivity,
increased mortality, and/or shifts in terrestrial plant community composition, often leading to
decreased biodiversity in many natural habitats wherever atmospheric Nr deposition increases
significantly and critical thresholds are exceeded (Aber et al., 1995); (2) leaching of excess
nitrate and associated base cations from terrestrial soils into streams, lakes and rivers and
mobilization of soil aluminum; (3) eutrophication, hypoxia, loss of biodiversity, and habitat
degradation in coastal ecosystems, now considered a major pollution problem in coastal waters
(Rabalais, 2002); (4) acidification and loss of aquatic flora and fauna biodiversity in lakes and
streams in many regions of the world when associated with sulfur deposition (Vitousek et al.,
1997); and (5) alteration of ecosystem processes such as nutrient and energy cycles through
6-28
-------
NOX
Energy
Production
S \
1
Ozi
Eff<
A
Atmosphere
Food
Production
People
(Food; Fiber)
Human Activities
NOV
NHV
NH3
Terrestrial
I Ecosystems
Agroecosystem Effects
Crop
t_
Animal
Soil
«-l
AAAAAAAAAAAA
The Nitrogen
Cascade
Indicates denitrification potential
NO,
N20
Figure 6-6 Illustration of the nitrogen cascade showing the movement of human-
produced reactive nitrogen (Nr) as it cycles through the various
environmental reservoirs in the atmosphere and in terrestrial and aquatic
ecosystems (Galloway et al., 2003; Figure 4-15, CD p. 4-97).
changes in the functioning and species composition of beneficial soil organisms (Galloway and
Cowling 2002).
Additional, indirect detrimental effects of excess Nr on societal values include: (1)
increases in fine PM resulting in regional hazes that decrease visibility at scenic rural and urban
vistas and airports (discussed above in section 6.2); (2) depletion of stratospheric ozone by N2O
emissions which can in turn affect ecosystems and human health; (3) global climate change
induced by emissions of N2O (Galloway et al., 2003); (4) formation of O3 and ozone-induced
injury to crops, forests, and natural ecosystems and the resulting predisposition to attack by
pathogens and insects, as well as human health related impacts (EPA, 1996); (5) decrease in
6-29
-------
quantity or quality of available critical habitat for threatened and endangered species (Fenn et al.,
2003); and (6) alteration of fire cycles in a variety of ecosystem types (Fenn et al., 2003).
A number of the more significant effects of chronic, long-term deposition of Nr on
terrestrial and aquatic ecosystems will be discussed below, specifically those effects which seem
to pose the greatest long-term risks to species or ecosystem health and sustainability or that
threaten ecosystem flows of goods and services important to human welfare.
Nitrogen Saturation of Terrestrial Ecosystems
Long-term, chronic additions of Nr (including ammonium and nitrate deposition from
ambient PM) to terrestrial ecosystems is resulting in numerous ecosystems shifting to a
detrimental ecological condition known as "nitrogen saturation." Nitrogen saturation does not
occur at a specific point in time, but is a set of gradually developing critical changes in
ecosystem processes which represent the integrated response of a system to increased Nr
availability over time (Aber, 1992). It occurs when Nr inputs exceed the capacity of plants and
soil microorganisms to utilize and retain the nitrogen (Aber et al., 1989, 1998; Garner, 1994;
EPA, 1993). Under conditions of nitrogen saturation, some other resource generally replaces
nitrogen in limiting biotic functions. The appearance of nitrate in soil solution (leaching) is an
early symptom of excess Nr accumulation.
Not all vegetation, organisms, or ecosystems react in the same manner to increased Nr
availability from atmospheric deposition. This is due in part to the variation both within and
across species in their inherent capacity to utilize additional Nr and the suite of other factors that
influence the range of community or ecosystem types possible at any given location. Such
factors can include the mineral composition of the underlying bedrock, the existing soil nutrient
pools, the local climatic conditions including weather extremes such as drought, high/low
temperatures, topography, elevations, natural/land use history, and fire regimes.
In U.S. ecosystems, the nutrient whose supply most often sets the limit of possible
primary productivity at a given site is biologically available nitrogen. However, in any given
ecosystem, not all plants are equally capable of utilizing extra nitrogen. Those plants that are
predisposed to capitalize on any increases in Nr availability gain an advantage over those that are
not as responsive to added nutrients. Over time, this shift in the competitive advantage may lead
to shifts in overall plant community composition. Whether or not this shift is considered adverse
would depend on the management context within which that ecosystem falls and the ripple
effects of this shift on other ecosystem components, essential ecological attributes (EEAs), and
ecosystems.
The effect of additions of Nr on plant community succession patterns and biodiversity
has been studied in several long-term nitrogen fertilization studies in both the U.S. and Europe.
These studies suggest that some forests receiving chronic inputs of Nr may decline in
productivity and experience greater mortality (Fenn et al. 1998). For example, fertilization and
nitrogen gradient experiments at Mount Ascutney, VT suggest that nitrogen saturation may lead
6-30
-------
to the replacement of slow-growing, slow nitrogen-cycling spruce-fir forest stands by fast-
growing deciduous forests that cycle nitrogen rapidly (Fenn et al. 1998). Similarly, experimental
studies of the effects of Nr deposition over a 12-year period on Minnesota grasslands dominated
by native warm-season grasses observed the shift to low-diversity mixtures dominated by cool-
season grasses at all but the lowest rates of Nr addition (Wedin and Tilman, 1996). The shift to
low-diversity mixtures was associated with the decrease in biomass carbon to N (C:N) ratios,
increased Nr mineralization, increased soil nitrate, high nitrogen losses, and low carbon storage.
Grasslands with high nitrogen retention and carbon storage rates were the most vulnerable to loss
of species and major shifts in nitrogen cycling. (Wedin and Tilman, 1996).
The carbon-to-nitrogen (C:N) ratio of the forest floor can be changed by nitrogen
deposition over time. In Europe, low C:N ratios coincide with high deposition regions. A strong
decrease in forest floor root biomass has also been observed with increased nitrogen availability,
and appears to occur when the ecosystem becomes nitrogen saturated. If root growth and
mycorrhizal formation are impaired by excessive nitrogen deposition, the stability of the forest
floor vegetation may be affected. The forest floor C:N ratio has been used as a rough indicator
of ecosystem nitrogen status in mature coniferous forests and the risk of nitrate leaching. Nitrate
leaching has been significantly correlated with forest floor nitrate status, but not with nitrate
deposition. Therefore, to predict the rate of changes in nitrate leaching, it is necessary to be able
to predict the rate of changes in the forest floor C:N ratio. Understanding the variability in forest
ecosystem response to nitrogen input is essential in assessing pollution risks (Gundersen et al.,
1998; CD, pp. 4-106, 4-107).
In the U.S., forests that are now showing severe symptoms of nitrogen saturation include:
the northern hardwoods and mixed conifer forests in the Adirondack and Catskill Mountains of
New York; the red spruce forests at Whitetop Mountain, Virginia, and Great Smoky Mountains
National Park, North Carolina; mixed hardwood watersheds at Fernow Experimental Forest in
West Virginia; American beech forests in Great Smoky Mountains National Park, Tennessee;
mixed conifer forests and chaparral watersheds in southern California and the southwestern
Sierra Nevada in Central California; the alpine tundra/subalpine conifer forests of the Colorado
Front Range; and red alder forests in the Cascade Mountains in Washington. All these systems
have been exposed to elevated nitrogen deposition, and nitrogen saturated watersheds have been
reported in the above-mentioned areas. Annual nitrogen additions through deposition in the
southwestern Sierra Nevada are similar in magnitude to nitrogen storage in vegetation growth
increments of western forests, suggesting that current nitrogen deposition rates may be near the
assimilation capacity of the overstory vegetation. Ongoing urban expansion will increase the
potential for nitrogen saturation of forests from urban sources (e.g., Salt Lake City, Seattle,
Tucson, Denver, central and southern California) unless there are improved emission controls
(Fennetal., 1998).
6-31
-------
The composition and structure of the plant community within an ecosystem in large part
determine the food supply and habitat types available for use by other organisms. In terrestrial
systems, plants serve as the integrators between above-ground and below-ground environments
and are influenced by and influence conditions in each. It is because of these linkages that
chronic excess Nr additions can lead to complex, dramatic, and severe ecosystem level/wide
changes/responses. Changes in soil Nr influence below ground communities as well. A
mutualistic relationship exists in the rhizosphere (plant root zone) between plant roots, fungi, and
microbes. Because the rhizosphere is an important region of nutrient dynamics, its function is
critical for the growth of the organisms involved. The plant roots provide shelter and carbon for
the symbionts, whereas the symbionts provide access to limiting nutrients such as nitrogen and
phosphorus for the plant. Bacteria make N, S, Ca, P, Mg, and K available for plant use while
fungi in association with plant roots form mycorrhizae that are essential in the uptake by plants
of mineral nutrients, such as N and P (Section 4.3.3; Wall and Moore, 1999; Rovira and Davy,
1974). Mycorrhizal fungal diversity is associated with above-ground plant biodiversity,
ecosystem variability, and productivity (Wall and Moore, 1999). Studies suggest that during
nitrogen saturation, soil microbial communities change from being predominately fungal, and
dominated by mycorrhizae, to being dominated by bacteria (Aber et al., 1998; CD, pp. 4-107, 4-
108), dramatically affecting both above- and below-ground ecosystems. These types of effects
have been observed in the field. For example, the coastal sage scrub (CSS) community in
California has been declining in land area and in drought deciduous shrub density over the past
60 years, and is being replaced in many areas by Mediterranean annual grasses. At the same
time, larger-spored below-ground fungal species (Scutellospora and Gigaspora), due to a failure
to sporulate, decreased in number with a concomitant proliferation of small-spored species of
Glomus aggregatum, G. leptotichum, and G. geospomm, indicating a strong selective pressure
for the smaller spored species of fungi (Edgerton-Warburton and Allen, 2000). These results
demonstrate that nitrogen enrichment of the soil significantly alters the arbuscular mycorrhizal
species composition and richness, and markedly decreases the overall diversity of the arbuscular
mycorrhizal community. The decline in the coastal sage scrub species can be directly linked to
the decline of the arbuscular mycorrhizal community (Edgerton-Warburton and Allen, 2000;
Allen et al., 1998; Padgett et al., 1999)(CD, pp. 4-108, 4-109).
Impacts on threatened and endangered species. In some rare and unique U.S.
ecosystems, chronic additions of atmospherically-derived nitrogen have already had some dire
and perhaps irreversible consequences. For example, California has many species that occur in
shrub, forb, and grasslands affected by N deposition, with up to 200 sensitive plant species in
southern California CSS alone (Skinner and Pavlik, 1994). Some 25 plant species are already
extinct in California, most of them annual and perennial forbs that occurred in sites now
experiencing conversion to annual grassland. As CSS converts more extensively to annual
grassland dominated by invasive species, loss of additional rare species may be inevitable.
6-32
-------
Though invasive species are often identified as the main threat to rare species, it is more likely
that invasive species combine with other factors, such as excess N deposition, to promote
increased productivity of invasive species and resulting species shifts.
Not surprisingly, as sensitive vegetation is lost, wildlife species that depend on these
plants are adversely affected. Included among these species are several threatened or
endangered species listed by the U.S. Fish and Wildlife Service, such as the desert tortoise and
checkerspot butterfly. A native to San Francisco Bay area, the bay checkerspot butterfly
(Euphydryas editha bayensis), has been declining steadily over the past decade, with local
extirpations in some reserves. This decline has been associated with the invasion of exotic
grasses replacing the native forbs on which the butterfly depends. In particular, the larval stage
is dependent on primarily one host plant, Plantago erecta, which is increasingly being out-
competed by exotic grasses.
Similarly, the desert tortoise has declined due to a number of co-occurring stresses,
including grazing, habitat destruction, drought, disease, and a declining food base. In the desert
shrub inter-spaces, sites where native forbs once flourished, invasive grasses now dominate,
reducing the nutritional quality of foods available to the tortoise (Fenn et al., 2003; Nagy et al.,
1998). Nitrogen deposition contributes to the productivity and density of N-fertilized grasses at
the expense of native forbs (Brooks, 2003). "Thus, protection of endangered species will
require increased exotic grass control, but local land management strategies to protect these
endangered species may not succeed unless they are accompanied by policy changes at the
regional or national level that reduce air pollution" (Fenn et al., 2003).
Community composition of epiphytic lichens is readily altered by small increases in
nitrogen deposition, an effect that seems to be widespread in the West (Fenn et al., 2003). Most
epiphytic lichens meet their nutritional requirements from atmospheric deposition and can store
N in excess of their nutritional needs (van Herk, 1999). In the San Bernardino Mountains, up to
50% of the lichen species that occurred in the region in the early 1900s have disappeared, with a
disproportionate number of the locally extinct species being (epiphytic) cyanolichens (Fenn et
al., 2003; Nash and Sigal, 1999). The Pacific Northwest, in contrast, still has widespread
populations of pollution-sensitive lichens (Fenn et al., 2003). However, in urban areas, intensive
agricultural zones and downwind of major urban and industrial centers, there is a sparsity of
sensitive lichen species and high levels of N concentrations have been measured in lichen tissue
(Fenn et al., 2003). Replacement of sensitive lichens by nitrophilous species has undesirable
ecological consequences. In late-successional, naturally N-limited forests of the Coast Range
and western Cascades, epiphytic cyanolichens make important contributions to mineral cycling
and soil fertility (Pike 1978; Sollins et al., 1980; Antoine, 2001), and together with other large,
pollution-sensitive macrolichens, are an integral part of the food web for large and small
mammals, insects and birds (McCune and Geiser, 1997).
6-33
-------
Alteration of native fire cycles. Several lines of evidence suggest that Nr deposition may
be contributing to greater fuel loads and thus altering the fire cycle in a variety of ecosystem
types, although further study is needed (Fenn et al., 2003). Invasive grasses promote a rapid fire
cycle in many locations (D'Antonio and Vitousek, 1992). The increased productivity of
flammable understory grasses increases the spread of fire and has been hypothesized as one
mechanism for the recent conversion of CSS to grassland (Minnich and Dezzani, 1998).
Thus, through its effect on habitat suitability, genetic diversity, community dynamics and
composition, nutrient status, energy and nutrient cycling, and frequency and intensity of natural
disturbance regimes (fire), excess Nr deposition is having profound and adverse impact on the
essential ecological attributes associated with terrestrial ecosystems. Strong correlation between
the stressor and adverse environmental response exists in many locations, and N-addition studies
have confirmed this relationship between stressor and response. Research efforts should be
made to elucidate what role particulate deposition is playing in contributing to these effects so as
to facilitate the mitigation of such effects.
Effects of Nitrogen Addition on Aquatic Habitats
Aquatic ecosystems (streams, rivers, lakes, estuaries or oceans) receive increased Nr
inputs either from direct atmospheric deposition (including nitrogen-containing particles),
surface runoff, or leaching from nitrogen saturated soils into ground or surface waters. The
primary pathways of Nr loss from forest ecosystems are hydrological transport beyond the
rooting zone into groundwater or stream water, or surface flows of organic nitrogen as nitrate
and Nr loss associated with soil erosion (Fenn et al., 1998). In the east, high nitrate
concentrations have been observed in streams draining nitrogen saturated watersheds in the
southern Appalachian Mountains (Fenn et al., 1998). The Great Smoky Mountains National
Park in Tennessee and North Carolina receives elevated amounts of total atmospheric deposition
of sulfur and nitrogen. A major portion of the atmospheric loading is from dry and occult
deposition. Nitrogen saturation of the watershed resulted in extremely high exports of nitrate
and promoted both chronic and episodic stream acidification in streams draining undisturbed
watersheds. Significant export of base cations was also observed (CD, pp. 4-110, 4-111; see also
section 6.3.3.2 on acidification from PM deposition).
In the west, the Los Angeles Air Basin exhibited the highest stream water NO3"
concentrations in wilderness areas of North America (Bytnerowicz and Fenn, 1996; Fenn et al.,
1998). Chronic N deposition in southern California, in the southwestern Sierra Nevada, and in
the Colorado Front Range leads to increased net N mineralization and nitrification rates in soil
and to elevated NO3" concentrations in lakes and streams. These symptoms occur in low- and
mid-elevation, high-deposition areas (>15 kg N/ha/yr) and in high elevation sites with relatively
low N deposition (4 to 8 kg N/ha/yr) but little capacity to assimilate and retain added N.
Estuaries are among the most intensely fertilized systems on Earth (Fenn et al., 1998).
They receive far greater nutrient inputs than other systems. In the Northeast, for example,
6-34
-------
nitrogen is the element most responsible for eutrophication in coastal waters of the region. Since
the early 1900s, there has been a 3- to 8-fold increase in nitrogen flux from 10 watersheds in the
northeast. These increases are associated with nitrogen oxide emissions from combustion which
have increased 5-fold. Riverine nitrogen fluxes have been correlated with atmospheric
deposition onto their landscapes and also with nitrogen oxides emissions into their airsheds.
Data from 10 benchmark watersheds with good historical records indicate that about 36-80% of
the riverine total nitrogen export, averaging approximately 64%, was derived directly or
indirectly from nitrogen oxide emissions (CD, pp. 4-109, 4-110).
The Pamlico Sound, NC estuarine complex, which serves as a key fisheries nursery
supporting an estimated 80% of commercial and recreational finfish and shellfish catches in the
southeastern U.S. Atlantic coastal region, has also been the subject of recent research (Paerl et
al., 2001) to characterize the effects of Nr deposition on the estuary. Direct atmospheric nitrogen
deposition onto waterways feeding into the Pamlico Sound or onto the Sound itself and indirect
nitrogen inputs via runoff from upstream watersheds contribute to conditions of severe water
oxygen depletion; formation of algae blooms in portions of the Pamlico Sound estuarine
complex; altered fish distributions, catches, and physiological states; and increases in the
incidence of disease. Especially under extreme rainfall events (e.g., hurricanes), massive
influxes of Nr (in combination with excess loadings of metals or other nutrients) into watersheds
and sounds can lead to dramatic decreases of oxygen in water and the creation of widespread
"dead zones" and/or increases in algae blooms that can cause extensive fish kills and damage to
commercial fish and sea food harvesting (Paerl et al., 2001; CD, pp. 4-109, 4-110).
6.3.3.2 Environmental Effects of PM-Related Acidic and Acidifying Deposition
Acid deposition has emerged over the past quarter century as a critical environmental
stress that affects diverse terrestrial and aquatic ecosystems in North America, Europe, and Asia
(Driscoll et al., 2001). In the eastern U.S. for example, the current acidity in precipitation is at
least twice as high as in pre-industrial times, with mean precipitation pH ranges from 4.3
(Pennsylvania and New York) to 4.8 (Maine) (EPA, 2003). Acid deposition is highly variable
across space and time, can originate from transboundary air pollution, can travel hundreds of
miles before being deposited, thereby affecting large geographic areas. It is composed of ions,
gases, and particles derived from the precursor gaseous emissions of SO2, NOX, NH3 and
particulate emissions of other acidifying compounds. Acid deposition disturbs forest and aquatic
ecosystems by giving rise to harmful chemical conditions (Driscoll et al., 2001).
Terrestrial Effects
Acid deposition has changed the chemical composition of soils by depleting the content
of available plant nutrient cations (e.g., Ca2+, Mg2+, K+) by increasing the mobility of Al, and by
increasing the S and N content (Driscoll et al., 2001). Soil leaching is often of major
importance in cation cycles, and many forest ecosystems show a net loss of base cations (CD, pp.
4-118). In acid sensitive soils, mineral weathering (the primary source of base cations in most
6-35
-------
watersheds) is insufficient to keep pace with leaching rates accelerated by acid deposition
(Driscolletal., 2001).
In the absence of acid deposition, cation leaching in northeastern forest soils is driven
largely by naturally occurring organic acids derived from the decomposition of organic matter.
Organic acids tend to mobilize Al through formation of organic-Al complexes, most of which are
deposited lower in the soil profile through adsorption to mineral surfaces. This process, termed
podzolization, results in surface waters with low concentrations of Al. Such concentrations are
primarily in a nontoxic, organic form (Driscoll et al., 1998). Acid deposition, however, has
altered podzolization by solubilizing Al with mobile inorganic anions, facilitating the transport
of inorganic Al into surface waters. In forest soils with base saturation values less than 20%,
acid deposition leads to increased Al mobilization and a shift in chemical speciation of Al from
organic to inorganic forms that are toxic to terrestrial and aquatic biota.
The toxic effect of Al on forest vegetation is attributed to its interference with plant
uptake of essential nutrients, such as Ca and Mg. Because Ca plays a major role in cell
membrane integrity and cell wall structure, reductions in Ca uptake suppress cambial growth,
reduce the rate of wood formation, decrease the amount of functional sapwood and live crown,
and predispose trees to disease and injury from stress agents when the functional sapwood
becomes less than 25% of cross sectional stem area (Smith, 1990). There are large variations in
Al sensitivity among ecotypes, between and within species, due to differences in nutritional
demands and physiological status, that are related to age and climate, which change over time
(CD, pp. 4-126).
Acid deposition has been firmly implicated as a causal factor in the northeastern high-
elevation decline of red spruce (DeHayes et al., 1999). Red spruce is common in Maine, where
it is an important commercial species. It is also common at high elevations in mountainous
regions throughout the Northeast, where it is valued for recreation and aesthetics, as well as for
providing a habitat for unique and endangered species. Dieback has been most severe at high
elevations in the Adirondack and Green Mountains, where more than 50% of the canopy trees
died during the 1970s and 1980s. In the White Mountains, about 25% of the canopy spruce died
during that same period (Craig and Friedland 1991). Dieback of red spruce trees has also been
observed in mixed hardwood-conifer stands at relatively low elevations in the western
Adirondack Mountains, areas that receive high inputs of acid deposition (Shortle et al., 1997).
Results of controlled exposure studies show that acidic mist or cloud water reduces the cold
tolerance of current-year red spruce needles by 3-10 degrees C (DeHayes et al., 1999). This
increased susceptibility to freezing occurs due to the loss of membrane-associated Ca2+ from
needles through leaching caused by the hydrogen ion. The increased frequency of winter injury
in the Adirondack and Green Mountains since 1955 coincides with increased exposure of red
spruce canopies to highly acidic and acidifying cloud water (Johnson et al., 1984). Recent
episodes of winter injury have been observed throughout much of the range of red spruce in the
6-36
-------
Northeast. (DeHayes et al., 1999). DeHayes et al. (1999) indicate that there is a significant
positive association between cold tolerance and foliar calcium in trees exhibiting deficiency in
foliar calcium, and further state that their studies raise the strong possibility that acid deposition
altering of foliar calcium is not unique to red spruce but has been demonstrated in many other
northern temperate forest tree species including yellow birch (Betula alleghaniensis)., white
spruce (Picea glaucus), red maple (Acer rubrum) eastern white pine (Pinus strobus), and sugar
maple (Acer saccharuni) (CD, p. 4-120).
Although less well established, there is also a strong possibility that low Ca to Al ratios
in soils may also be impacting northeastern red spruce. Cronan and Grigal (1995) concluded that
a Ca:Al ratio of less than 1.0 in soil water indicated a greater than 50% probability of impaired
growth in red spruce. They cite examples of studies from the northeast where soil solutions in
the field were found to exhibit Ca:Al ionic ratios less than 1.0.
Acid deposition may also be contributing to episodic dieback of sugar maple in the
Northeast through depletion of nutrient cations from marginal soils. Horsley et al. (1999) found
that dieback at 19 sites in northwestern and north-central Pennsylvania and south-western New
York was correlated with combined stress from defoliation and deficiencies of Mg and Ca.
Dieback occurred predominately on ridgetops and on upper slopes, where soil base availability
was much lower than at mid and low slopes of the landscape (Bailey et al., 1999). Because
multiple factors such as soil mineralogy and landscape position affect soil base status, the extent
to which sugar maple dieback can be attributed to acid deposition is not clear.
Less sensitive forests throughout the U.S. are experiencing gradual losses of base cation
nutrients, which in many cases will reduce the quality of forest nutrition over the long term
(National Science and Technology Council, 1998). In some cases, such effects may not even
take decades to occur because these forests have already been receiving S and N deposition for
many years.
In contrast to contributing to the adverse impacts of acid deposition, particles can also
provide a beneficial supply of base cations to sites with very low rates of supply from mineral
sources. In these areas, atmospheric inputs of bass cations can help ameliorate the acidifying
effects of acid particles. The Integrated Forest Study (IPS) (Johnson and Lindberg, 1992) has
characterized the complexity and variability of ecosystem responses to atmospheric inputs and
provided the most extensive data set available on the effects of atmospheric deposition, including
particle deposition, on the cycling of elements in forest ecosystems. This study showed that in
the US ecosystems, inputs of base cations have considerable significance, not only for base
cation status, but also for the potential of incoming precipitation to acidify or alkalize the soils.
The actual rates, directions, and magnitudes of changes that may occur in soils (if any), however,
will depend on rates of inputs from weathering and vegetation outputs, as well as deposition and
leaching. In other words, these net losses or gains of base cations must be placed in the context
of the existing soil pool size of exchangeable base cations (CD, p. 4-132). Given the wide
6-37
-------
ranges of particulate deposition for each base cation across the IPS sites, however, the unique
characteristics of various sites need to be better understood before assumptions are made about
the role particulate pollution plays in ecosystem impacts (CD, pp. 4-127, 4-128).
In a follow up study, Johnson et al. (1999) used the nutrient cycling model, NuCM, to
simulate the effects of reduced S, N, and base cation (CB) deposition on nutrient pools, fluxes,
soil, and soil solution chemistry in two contrasting southern Appalachian forest ecosystems. The
authors found that in an extremely acidic system, CB deposition can have a major effect on CB
leaching through time and S and N deposition had a major effect on Al leaching. At the less
acidic Coweeta site, CB deposition had only a minor effect on soils and soil solutions; whereas S
and N deposition had delayed but major effects on CB leaching (CD, pp. 4-136, 4-137).
Aquatic Effects
Inputs of acid deposition to regions with base-poor soils have resulted in the acidification
of soil waters, shallow ground waters, streams, and lakes in a number of locations within the
U.S. In addition, perched seepage lakes, which derive water largely from direct precipitation
inputs, are highly sensitive to acid deposition (Charles, 1991). These processes usually result in
lower pH and, for drainage lakes, higher concentrations of inorganic monomeric Al. Such
changes in chemical conditions are toxic to fish and other aquatic animals (Driscoll et al., 2001).
A recent report, Response of Surface Water Chemistry to the Clean Air Act of 1990
(EPA, 2003), analyzes data from 1990 through 2000 obtained from EPA's Long Term
Monitoring (LTM) and Temporally Integrated Monitoring of Ecosystems (TIME) projects, part
of EMAP (Environmental Monitoring and Assessment Program). The report assesses recent
changes in surface water chemistry in response to changes in deposition, in the northern and
eastern U.S., specifically in the acid sensitive regions defined as New England (Maine, New
Hampshire, Vermont and Massachusetts), the Adirondack Mountains of New York, the Northern
Appalachian Plateau (New York, Pennsylvania and West Virginia), the Ridge and Blue Ridge
Provinces of Virginia, and the Upper Midwest (Wisconsin and Michigan). Acidic waters are
defined as having acid neutralizing capacity (ANC) less than zero (i.e., no acid buffering
capacity in the water), corresponding to a pH of about 5.2. Increases in surface water ANC
values and/or pH would indicate improved buffering capacity and signal the beginning of
recovery (EPA, 2003).
Using National Atmospheric Deposition Program (NADP) data, trends in sulfate and N
(nitrate + ammonium) deposition were analyzed, along with CB deposition, sulfate and nitrate
concentrations in surface waters, ANC and pH levels. Over this timeframe, sulfate deposition
declined significantly across all regions, while N declined slightly in the Northeast and increased
slightly in the Upper Midwest. Base cation deposition showed no significant changes in the East
and increased slightly in the Upper Midwest. Concurrently, all regions except the Ridge/Blue
Ridge province in the mid-Atlantic showed significant declines in sulfate concentrations in
6-38
-------
surface waters, while nitrate concentrations decreased in two regions with the highest ambient
nitrate concentrations (Adirondacks, Northern Appalachian Plateau) but were relatively
unchanged in regions with low concentrations.
Given the declines in S and N deposition measured for these areas, one would expect to
find increasing values of ANC, pH or both in response. ANC values did increase in the
Adirondacks, Northern Appalachian Plateau and Upper Midwest, despite a decline in base
cations (Ca and Mg) in each region. The loss of base cations limited the extent of ANC and pH
increase. Toxic Al concentrations also declined slightly in the Adirondacks. In New England
and Ridge/Blue Ridge, however, regional surface water ANC did not change significantly (EPA,
2003).
Modest increases in ANC have reduced the number of acidic lakes and stream segments
in some regions. There are an estimated 150 Adirondack lakes with ANC less than 0, or 8.1% of
the population, compared to 13% (240 lakes) in the early 1990s. In the Upper Midwest, an
estimated 80 of 250 lakes that were acidic in mid-1980s are no longer acidic. TIME surveys of
streams in the Northern Appalachian Plateau region estimated that 8.5% (3,600 kilometers) of
streams remain acidic at the present time, compared to 12% (5,014 kilometers) of streams that
were acidic in 1993-94. In these three regions taken together, approximately one-fourth to one-
third of formerly acidic surface waters are no longer acidic, although still with very low ANC.
The report finds little evidence of regional change in the acidity status of New England or the
Ridge/Blue Ridge regions and infers that the numbers of acidic waters remain relatively
unchanged. Despite a general decline in base cations and a possible increase in natural organic
acidity, there is no evidence that the number of acidic waters have increased in any region (EPA,
2003).
Acidification has marked effects on the trophic structure of surface waters. Decreases in
pH and increases in Al concentrations contribute to declines in species richness and in the
abundance of zooplankton, macroinvertebrates, and fish (Schindler et al.,1985; Keller and Gunn
1995). Numerous studies have shown that fish species richness (the number offish species in a
water body) is positively correlated with pH and ANC values (Rago and Wiener, 1986; Kretser et
al., 1989). Decreases in pH result in decreases in species richness by eliminating acid-sensitive
species (Schindler et al., 1985). Of the 53 fish species recorded by the Adirondack Lakes Survey
Corporation, about half (26 species) are absent from lakes with pH below 6.0. Those 26 species
include important recreational fishes, such as Atlantic salmon, tiger trout, redbreast sunfish,
bluegill, tiger musky, walleye, alewife, and kokanee (Kretser et al., 1989), plus ecologically
important minnows that serve as forage for sport fishes.
A clear link exists between acidic water, which results from atmospheric deposition of
strong acids, and fish mortality. The Episodic Response Project (ERP) study showed that
streams with moderate to severe acid episodes had significantly higher fish mortality during
bioassays than nonacidic streams (Van Sickle et al., 1996). The concentration of inorganic
6-39
-------
monomeric Al was the chemical variable most strongly related to mortality in the four test
species (brook trout, mottled sculpin, slimy sculpin, and blacknose dace). The latter three
species are acid sensitive. In general, trout abundance was lower in ERP streams with median
episode pH less than 5.0 and inorganic monomeric Al concentrations greater than 3.7 - 7.4 mmol
L"1. Acid sensitive species were absent from streams with median episode pH less than 5.2 and
with a concentration of inorganic monomeric Al greater than 3.7 mmol L"1.
Given the significant decreases in sulfur emissions that have occurred in the U. S. and
Europe in recent decades, the findings of Driscoll et al. (1989, 2001) and Hedin et al. (1994) are
especially relevant. Driscoll et al. (1989, 2001) noted a decline in both SO4"2 and base cations in
both atmospheric deposition and stream water over the past two decades at Hubbard Brook
Watershed, NH. However, the decreases in SO2 emissions in Europe and North America in
recent years have not been accompanied by equivalent declines in net acidity related to sulfate in
precipitation, and may have, to varying degrees, been offset by steep declines in atmospheric
base cation concentrations over the past 10 to 20 years (Hedin et al., 1994).
Driscoll et al. (2001) envision a recovery process that will involve two phases. Initially,
a decrease in acid deposition following emissions controls will facilitate a phase of chemical
recovery in forest and aquatic ecosystems. Recovery time for this phase will vary widely across
ecosystems and will be a function of the following:
• the magnitude of decreases in atmospheric deposition
• the local depletion of exchangeable soil pools of base cations
• the local rate of mineral weathering and atmospheric inputs of base cations
the extent to which soil pools of S and N are released as SO42" or as NO3" to
drainage waters and the rate of such releases (Galloway et al., 1983).
In most cases, it seems likely that chemical recovery will require decades, even with additional
controls on emissions. The addition of base cations, e.g., through liming, could enhance
chemical recovery at some sites.
The second phase in ecosystem recovery is biological recovery, which can occur only if
chemical recovery is sufficient to allow survival and reproduction of plants and animals. The
time required for biological recovery is uncertain. For terrestrial ecosystems, it is likely to be at
least decades after soil chemistry is restored because of the long life of tree species and the
complex interactions of soil, roots, microbes, and soil biota. For aquatic systems, research
suggests that stream macroinvertebrate populations may recover relatively rapidly
(approximately 3 years), whereas lake populations of zooplankton are likely to recover more
slowly (approximately 10 years) (Gunn and Mills, 1998). Some fish populations may recover in
5 to 10 years after the recovery of zooplankton populations. Stocking could accelerate fish
population recovery (Driscoll et al., 2001)
6-40
-------
Projections made using an acidification model (PnET-BGC) indicate that full
implementation of the 1990 CAAA will not afford substantial chemical recovery at Hubbard
Brook EF and at many similar acid-sensitive locations (Driscoll et al., 2001) . Model
calculations indicate that the magnitude and rate of recovery from acid deposition in the
northeastern U.S. are directly proportional to the magnitude of emissions reductions. Model
evaluations of policy proposals calling for additional reductions in utility SO2 and NOX
emissions, year round emissions controls, and early implementation indicate greater success in
facilitating the recovery of sensitive ecosystems (Driscoll et al., 2001).
Indirect Radiation and Climate Condition Effects from Atmospheric PM
In addition to the direct and indirect effects of deposited PM, ambient atmospheric PM
can affect radiation and climate conditions that influence overall plant/ecosystem productivity.
The degree to which these effects occur in any given location will depend on the chemical and
physical composition and concentration of the ambient PM. Because plants are adapted to the
overall light and temperature environments in which they grow, any PM-related changes to these
conditions (see section 6.5 below) potentially alter the overall competitive success these plants
will have in that ecosystem.
With respect to radiation, the characteristics and net receipts of solar and terrestrial
radiation determine rates of both photosynthesis and the heat-driven process of water cycling.
Atmospheric turbidity (the degree of scattering occurring in the atmosphere due to particulate
loading) influences the light environment of vegetative canopy in two ways: through conversion
of direct to diffuse radiation and by scattering or reflecting incoming radiation back out into
space. Diffuse radiation increases canopy photosynthetic productivity by distributing radiation
more uniformly throughout the canopy so that it also reaches the lower leaves and improves the
canopy radiation use efficiency (RUE). Acting in the opposite direction, non-absorbing,
scattering aerosols present in PM reduce the overall amount of radiation reaching vegetative
surfaces, by scattering or reflecting it back into space. It appears that global albedo has been
increasing due to an increasing abundance of atmospheric particles. Using World
Meteorological Organization (WMO) data, Stanhill and Cohen (2001) have estimated that
average solar radiation receipts have declined globally by an average of 20 W m-2 since 1958.
The net effect of atmospheric particles on plant productivity is not clear, however, as the
enrichment in photosynthetically active radiation (PAR) present in diffuse radiation may offset a
portion of the effect of decreased solar radiation receipts in some instances (CD, pp. 4-92, 4-93).
Plant processes also are sensitive to temperature. Some atmospheric particles (most
notably black carbon) absorb short-wavelength solar radiation, leading to atmospheric heating
and reducing total radiation received at the surface. Canopy temperature and transpirational
water use by vegetation are particularly sensitive to long-wave, infrared radiation. Atmospheric
heating by particles can potentially reduce photosynthetic water uptake efficiency and vertical
temperature gradients, potentially reducing the intensity of atmospheric turbulent mixing.
6-41
-------
Stanhill and Cohen (2001) suggested that plant productivity is more affected by changes in
evapotranspiration induced by changes in the amount of solar radiation plants receive than by
changes in the amount of PAR plants receive (CD, p. 4-93).
6.3.4 Characteristics and Location of Sensitive Ecosystems in the U.S.
Ecosystems sensitive to anthropogenically derived nitrogen and/or acid deposition tend to
have similar characteristics. Some of these ecosystems and characteristics have already been
mentioned in earlier sections but are repeated here to provide a more comprehensive list that can
help ecological risk assessors/managers identify areas of known or potential concern. For
example, lower nitrogen and/or resource environments, such as those with infertile soils, shaded
understories, deserts, or tundras, are populated with organisms specifically adapted to survive
under those conditions. Plants adapted to these conditions have been observed to have similar
characteristics, including inherently slower growth rates, lower photosynthetic rates, and lower
capacity for nutrient uptake, and grow in soils with lower soil microbial activity. When N
becomes more readily available, such plants will be replaced by nitrophilic plants which are
better able to use increased amounts of Nr (Fenn et al., 1998).
Additionally, in some instances, there seem to be important regional distinctions in
exposure patterns, environmental stressors, and ecosystem characteristics between the eastern
and western U.S.. A seminal report describing these distinctive characteristics for the western
U.S. (11 contiguous states located entirely west of the 100th meridian) is Fenn et al., 2003.
In the western U.S., vast areas receive low amounts of atmospheric deposition,
interspersed with hotspots of elevated N deposition downwind of large, expanding metropolitan
centers or large agricultural operations. In other words, spatial patterns of urbanization largely
define the areas where air pollution impacts are most severe. The range of air pollution levels for
western wildlands is extreme, spanning from near-background to the highest exposures in all of
North America, with the possible exception of forests downwind of Mexico City. Over the same
geographic expanse, climatic conditions and ecosystem types vary widely. Some regions receive
more than 1000 millimeters of precipitation, namely the Pacific coastal areas, the Sierra Nevada,
the Colorado Rockies, and northern Idaho, while other regions are arid or semiarid, with more
than 300 clear days per year (Riebsame et al., 1997). In these latter regions, the contribution of
atmospheric dry deposition is likely to be most important. These characteristics which are
unique to the West require special consideration, and often make application of models and
ecological effects thresholds developed for other regions inappropriate.
In summary, sensitive or potentially sensitive ecosystems in the West include those that:
• are located downwind of large urban source areas; regions with a mix of
emissions sources that may include urban, mobile, agricultural, and industrial
sources; and/or sites near large point sources of N.
6-42
-------
• contain inherently N sensitive ecosystem components, such as lichens, diatoms, or
poorly buffered watersheds which produce high stream water NO3- levels. These
sensitive components can be affected by N deposition rates as low as 3-8
kg/ha/yr.
occur on top of siliclastic/crystalline bedrock with little potential for buffering
acidity.
• are naturally nitrogen limited. For example, the approximately 16,000 high
elevation western mountain lakes are generally oligotrophic and especially
sensitive to the effects of atmospheric deposition.
A seminal report describing key characteristics of sensitive ecosystems for the eastern
and in particular the northeastern U.S. is Driscoll et al. (2001). In the northeastern United States,
atmospheric deposition is largely a regional problem. Because S and N most often occur
together in the eastern atmosphere and deposit to the environment as acid deposition, acid
deposition is seen as a critical environmental stress.
Several critical chemical thresholds appear to coincide with the onset of deleterious
effects to biotic resources resulting from acid deposition. Thus, ecosystems sensitive to
additional acid inputs include those with the following characteristics:
a molar Ca: Al ratio of soil water that is less than 1.0;
• soil percentage base cation saturation less than 20%;
• surface water pH less than 6.0;
ANC less than 50 meq L-l; and
• concentrations of inorganic monomeric Al greater than 2 mmol L-l.
Knowledge of such indicators is necessary for restoring ecosystem structure and function.
6.3.5 Ecosystem Exposures to PM Deposition
In order for any specific chemical stressor present in ambient PM to impact ecosystems,
it must first be removed from the atmosphere through deposition. Deposition can occur in three
modes: wet (rain/frozen precipitation), dry, or occult (fog, mist or cloud). At the national scale,
all modes of deposition must be considered in determining potential impacts to vegetation and
ecosystems because each mode may dominate over specific intervals of time or space. (CD, p.
4-8 to 4-10). For example, in large parts of the western U.S. which are arid or semiarid, dry
deposition may be the source of most deposited PM (Fenn, et al., 2003). However, in coastal
areas or high elevation forests, where wet or occult deposition may predominate, deposition
amounts may greatly exceed PM amounts measured in the ambient air. Occult deposition is
particularly effective for delivery of dissolved and suspended materials to vegetation because:
(1) concentrations of ions are often many-fold higher in clouds or fog than in precipitation or
6-43
-------
ambient air (e.g., acidic cloud water, which is typically 5-20 times more acid than rainwater, can
increase pollutant deposition and exposure to vegetation and soils at high elevation sites by more
than 50% of wet and dry deposition levels); (2) PM is delivered in a hydrated and bioavailable
form to foliar surfaces and remains hydrated due to conditions of high relative humidity and low
radiation; and (3) the mechanisms of sedimentation and impaction for submicron particles that
would normally be low in ambient air are increased. High-elevation forests can be especially at
risk from depositional impacts because they receive larger particulate deposition loadings than
equivalent low-elevation sites, due to a number of orographic (mountain related) effects. These
orographic effects include higher wind speeds that enhance the rate of aerosol impaction,
enhanced rainfall intensity and composition, and increased duration of occult deposition.
Additionally, the needle-shaped leaves of the coniferous species often found growing in these
high elevation sites, enhance impaction and retention of PM delivered by all three deposition
modes (CD, pp. 4-29, 4-44).
In order to establish exposure-response profiles useful in ecological risk assessments, two
types of monitoring networks need to be in place. First, a deposition network is needed that can
track changes in deposition rates of PM stressors (nitrates/sulfates) occurring in sensitive or
symptomatic areas/ecosystems. Secondly, a network or system of networks should be
established that measures the response of key sensitive ecological indicators over time to
changes in atmospheric deposition of PM stressors.
Currently in the U.S., national deposition monitoring networks routinely measure total
wet or dry deposition of certain compounds. Atmospheric concentrations of dry particles began
to be routinely measured in 1986, with the establishment of EPA's National Dry Deposition
Network (NDDN). After new monitoring requirements were added in the 1990 CAAA, EPA, in
cooperation with the National Oceanic and Atmospheric Association, created the Clean Air
Status and Trends Network (CASTNet) from the NDDN. CASTNet comprises 85 sites and is
considered the nation's primary source for atmospheric data to estimate concentrations for
ground-level ozone and the chemical species that make up the dry deposition component of total
acid deposition (e.g., sulfate, nitrate, ammonium, sulfur dioxide, and nitric acid), as well as the
associated meteorology and site characteristics data that are needed to model dry deposition
velocities (CD, pg. 4-21; (http://www.epa.gov/castnet/).
To provide data on wet deposition amounts in the U.S., the National Atmospheric
Deposition Program (NADP) was initiated in the late 1970's as a cooperative program between
federal, state, and other public and private groups. By the mid-1980's, it had grown to nearly
200 sites, and it stands today as the longest running national atmospheric deposition monitoring
network (http: //nadp. sws .uiuc. edu/).
In addition to these deposition monitoring networks, other networks collect data on
ambient aerosol concentrations and chemical composition. Such networks include the
IMPROVE network, discussed above in section 2.5, and the newly implemented PM2 5 chemical
6-44
-------
Speciation Trends Network (STN) that consists of 54 core National Ambient Monitoring
Stations and approximately 250 State and Local Air Monitoring Stations.
Data from these deposition networks demonstrate that N and S compounds are being
deposited in amounts known to be sufficient to affect sensitive terrestrial and aquatic ecosystems
over time. Though the percentages of N and S containing compounds in PM vary spatially and
temporally, nitrates and sulfates make up a substantial portion of the chemical composition of
PM. In the future, speciated data from these networks may allow better understanding of the
specific components of total deposition that are most strongly influencing PM-related ecological
effects.
Unfortunately, at this time there are only a few sites where long-term monitoring of
sensitive indicators of ecosystem response to acidic and acidifying deposition is taking place
within the U.S.. Two examples are the Hubbard Brook Experimental Forest research site, that
provides the longest continuous record of precipitation and stream chemistry in the U.S. (Likens
and Bormann, 1995) and EPA's LTM and TIME projects which monitor changes in surface
water chemistry in the acid sensitive regions of the northern and eastern U.S.. Because the
complexities of ecosystem response make predictions of the magnitude and timing of chemical
and biotic recovery uncertain, it is strongly recommended that this type of long-term surface
water chemistry monitoring network be continued, and that a biological monitoring program be
added. Data from these long-term monitoring sites will be invaluable for the evaluation of the
response of forested watersheds and surface waters to a host of research and regulatory issues
related to acid deposition, including soil and surface water recovery, controls on N retention,
mechanisms of base cation depletion, forest health, sinks for S in watersheds, changes in
dissolved organic carbon and speciation of Al, and various factors related to climate change
(EPA, 2003).
6.3.6 Critical Loads
The critical load (CL) has been defined as a "quantitative estimate of an exposure to one
or more pollutants below which significant harmful effects on specified sensitive elements of the
environment do not occur according to present knowledge" (Lokke et al., 1996). The critical
load framework originated in Europe where the concept has generally been accepted as the basis
for abatement strategies to reduce or prevent injury to the functioning and vitality of forest
ecosystems caused by long-range transboundary chronic acid deposition. The concept is useful
for estimating the amounts of pollutants that sensitive ecosystems can absorb on a sustained
basis without experiencing measurable degradation. The estimation of ecosystem critical loads
requires an understanding of how an ecosystem will respond to different loading rates in the long
term and is a direct function of the level of sensitivity of the ecosystem to the pollutants in
question and its capability to ameliorate pollutant stress.
6-45
-------
Key to the establishment of a critical load is the selection of appropriate ecological
endpoints or indicators that are measurable characteristics related to the structure, composition,
or functioning of ecological systems (i.e., indicators of condition). In Europe, the elements used
in the critical load concept are a biological indicator, a chemical criterion, and a critical value
(CD, p. 4-125). The biological indicator is the organism used to indicate the status of the
receptor ecosystem; the chemical criterion is the parameter that results in harm to the biological
indicator; and the critical value is the value of the chemical criterion below which no significant
harmful response occurs to the biological indicator (Lokke et al., 1996).
A number of different types of indicators and chemical criteria for monitoring ecosystem
status have been proposed. Some examples for evaluating ecosystem nitrogen status include:
foliar nitrogen content, nutrient ratios (N:P, N:cation); foliar nitrate; foliar 615 N; arginine
concentration; soil C:N ratio; NO3" in soil extracts or increased and prolonged NO3" loss below
the main rooting zone and in stream water or in soil solution; and flux rates of nitrogenous trace
gases from soil (Fenn et al., 1998). Seasonal patterns of stream water nitrate concentrations are
especially good indicators of watershed N status. Biological indicators that have been suggested
for use in the critical load calculation in forest ecosystems include mycorrhizal fungi (Lokke
et al., 1996) and fine roots, since they are an extremely dynamic component of below-ground
ecosystems and can respond rapidly to stress. The physiology of carbon allocation has also been
suggested as an indicator of anthropogenic stress (Andersen and Rygiewicz, 1991). Lichen
community composition in terrestrial ecosystems or lichen N tissue levels are also fairly
responsive to changes in N deposition over time (Fenn et al., 2003). In aquatic systems, diatom
species composition can be a good indicator of changes in water chemistry (Fenn et al., 2003).
It should be kept in mind, however, that the response of a biological indicator is an integration of
a number of different stresses. Furthermore, there may be organisms more sensitive to the
pollutant(s) than the species selected (Lokke et al., 1996; National Science and Technology
Council, 1998) (CD, pp. 4-124 to 126).
Within North America, a number of different groups have recently begun to use or
develop critical loads. As discussed below, these groups include the U.S. Federal Land
Managers (FLMs), such as the National Park Service and the Forest Service, a binational group
known as New England Governors/Eastern Canadian Premiers (NEG/ECP), and several
Canadian Provinces.
Federal Land Managers have hosted a number of meetings over the last few years to
discuss how the CL concept might be used in helping them fulfill their mandate of providing
protection for the lands they manage. In trying to develop a consistent approach to using CL, a
number of issues and considerations have been identified. First, the distinction between critical
loads (which are based on modeled or measured dose-response data) and target loads (which can
be based on political, economic, spatial or temporal considerations in addition to scientific
information) needs to be recognized. When using either the critical or the target load (TL)
6-46
-------
approach, one must indicate the spatial (or geographic) scope, the temporal scope (timeframe to
ecological or ecosystem recovery), and a description of the sensitive receptors (or resource) to be
protected, the sensitive receptor indicators (physical, chemical biological, or social
characteristics of the receptor that can be measured), and the harmful effect on the receptor that
is of concern. Additionally, one would need to specify what is the "desired condition" that the
critical or target load is meant to achieve. For any given location, there may be a range or suite
of possible critical or target loads based on different sensitive receptors and/or receptor
indicators found at that site. Alternatively, one could focus on the most sensitive receptor and
select a single CL or TL for that receptor. Several aspects of the CL approach make it attractive
for use by the FLMs. Specifically, it can provide a quantitative, objective and consistent
approach for evaluating resource impacts in a given ecosystem. In an effort to progress the CL
approach, the Forest Service is testing the applicability of the European protocol to several U.S.
case study sites.
Under the auspices of the NEG/ECP, and other binational efforts, Canadian and U.S.
scientists are involved in joint forest mapping projects. A Forest Mapping Work Group has been
tasked with conducting a regional assessment of the sensitivity of northeastern North American
forests to current and projected sulfur and nitrogen emissions levels, identifying specific forested
areas most sensitive to continued deposition and estimating deposition rates required to maintain
forest health and productivity. They have completed the development of methods, models and
mapping techniques, and identification of data requirements. Some of these data requirements
include: pollution loading to forest landscapes; the interaction of pollutants with forest canopies;
plant nutrient requirements; and the ability of soils to buffer acid inputs and replenish nutrients
lost due to acidification.
In addition to CL measures, they have also defined a corresponding "deposition index"
for each CL value. The deposition index is the difference between the CL value for a given site
and the current deposition rates at that site. Positive values of the index at a particular forested
location reflect the capacity of that forest ecosystem to tolerate additional acid deposition.
Negative index values correspond to the reduction in S and N deposition required to eliminate or
deter the development of future nutrient limitations. This allows an assessor to identify areas
where the deposition problems are most severe, and which sites might be under the applicable
CL level currently but not far from reaching or exceeding that level should deposition levels
increase. Currently maps exist for Vermont and Newfoundland, though the goal is to develop
maps that will cover Quebec and the Atlantic provinces of Canada, along with the remaining
New England states. These maps show that 31% of Vermont forests and 23% of Newfoundland
forests are sensitive (e.g., current levels of S and N deposition are causing cation depletion).
Though these current activities hold promise for using the CL approach in environmental
assessments and in informing management decisions, widespread use of CLs in the U.S. is not
yet possible. The CL approach is very data-intensive, and, at the present time, there is a paucity
6-47
-------
of ecosystem- level data for most sites. However, for a limited number of areas which already
have a long-term record of ecosystem monitoring, (e.g., Rocky Mountain National Park in
Colorado and the Lye Brook Wilderness in Vermont), FLMs may be able to develop site-
specific CLs. Further, in areas already exceeding the applicable CL, it may be difficult to
determine what the management goals are/should be for each mapped area (e.g., what is the
"desired condition" or level of protection) without historic baseline data. More specifically, with
respect to PM deposition, there are insufficient data for the vast majority of U.S. ecosystems that
differentiate the PM contribution to total N or S deposition to allow for practical application of
this approach as a basis for developing national standards to protect sensitive U.S. ecosystems
from adverse effects related to PM deposition. Though atmospheric sources of Nr and acidifying
compounds, including ambient PM, are clearly contributing to the overall excess pollutant load
or burden entering ecosystems annually, insufficient data are available at this time to quantify
the contribution of ambient PM to total Nr or acid deposition as its role varies both temporally
and spatially along with a number of other factors. Thus, it is not clear whether a CL could be
developed just for the portion of the total N or S input that is contributed by PM.
6.3.7 Summary and Conclusions
The above discussions identify a group of ecosystems known to be sensitive to excess N
and S inputs and a list of characteristics that can be used to predict or locate other potentially
sensitive ecosystems within the U.S. Further, exposures of these sensitive ecosystems to
atmospherically derived pollutants (e.g., N and S) have been measured and documented, in some
cases for decades. Clear linkages between reduced atmospheric concentrations of these
pollutants and reduced deposition rates have been made. The mechanisms of environmental and
ecosystem responses to these inputs are increasingly understood, though very complex.
Fertilization and acidification studies have verified observed ecosystem responses to these
pollutants in the field. Ecosystem-level effects associated with excess N and S inputs are
profound, but in most cases potentially reversible. New assessment and management tools, such
as critical and target loads, are being developed to better characterize the relationship between
deposition loads and ecosystem response. The success of these tools will depend on the
availability of sufficient ecosystem response data, which is currently limited to a few long-term
monitoring networks/sites (e.g., TEVIE/LTM). The current risk to sensitive ecosystems and
especially sensitive species like the checkerspot butterfly, desert tortoise, epiphytic lichens,
native shrub and forb species, and aquatic diatom communities is high. The loss of species and
whole ecosystem types is adverse and should receive increased protection.
A number of ecosystem-level conditions (e.g., nitrogen saturation, terrestrial and aquatic
acidification, coastal eutrophication) have been associated with chronic, long-term exposure of
ecosystems to elevated inputs of compounds containing Nr, sulfur and/or associated hydrogen
ions. These ecosystem level changes have profound impacts on almost all of the EEAs
6-48
-------
identified in the EPEC Framework (SAB, 2002) and described in sections 4.2.1 and 4.2.3 of the
CD. These impacted EEAs include Landscape Condition, Biotic Condition, Chemical and
Physical Characteristics, Ecological Processes, and Natural Disturbance Regimes. Given that
humans, as well as other organisms, are dependent on the services ecosystems provide,
ecosystem changes of this magnitude are of concern and can lead to adverse impacts on human
health and welfare.
Based on the information included in the above discussions and Chapters 4 and 9 of the
CD, staff has reached the following conclusions:
• An ecologically-relevant indicator for PM should be based on one or multiple
chemical stressors found in ambient PM (e.g. N or S containing compounds).
PM, as a contributor to the chronic annual loads of total Nr and/or acidifying
compounds entering sensitive ecosystems, has been associated with numerous
effects on those ecosystems and their associated essential ecological attributes.
There is no bright line or threshold for these effects, but rather a "syndrome" of
complex changes over time. As levels of inputs of these pollutants are reduced,
ecosystem recovery can occur but may take decades, and may require controls
beyond those already established.
• Excess Nr or acidifying deposition acts in conjunction with other co-occurring
stresses (e.g., invasive species, reduced grazing pressure) that jointly determine
ecological outcomes. Therefore, these pollution-related stresses should not be
considered in isolation. Additionally, all forms of airborne nitrogen and
acidifying compounds need to be considered and managed in harmony.
Existing ambient air and deposition monitoring networks are not generally
sufficient to characterize the associated ecosystem response. Additional, long-
term, targeted ecosystem monitoring is needed (e.g., downwind of large urban
areas in the West).
Unfortunately, our ability to relate ambient concentrations of PM to ecosystem response
is hampered by a number of significant data gaps and uncertainties. First, U.S. monitoring
networks have only recently begun to measure speciated PM. Historically, measurements were
focused only on a particular size fraction such as PM10 and, more recently, PM2 5. An exception
to this is the IMPROVE network, which collects speciated measurements. Additionally, except
for the IMPROVE and some CASTNet sites, much of the PM monitoring effort has focused on
urban or near urban exposures, rather than on those in sensitive ecosystems. Thus, the lack of a
long-term, historic database of annual speciated PM deposition rates precludes establishing
relationships between PM deposition (exposure) and ecosystem response at this time. As a
result, while evidence of PM-related effects clearly exists, there is insufficient information
6-49
-------
available at this time to serve as a basis for a secondary national air quality standard for PM,
specifically selected to protect against adverse effects on vegetation and ecosystems.
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 ambient air concentration to a receptor response, an important factor in being able
to set a national ambient air quality standard. A multitude of factors (e.g., 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 at any given point
in time. Therefore, modeled deposition velocities, used in the absence of monitored data, can be
highly uncertain.
Third, each ecosystem has developed within a context framed by the topography,
underlying bedrock, soils, climate, meteorology, hydrologic regime, natural and land use history,
species associations that may co-occur only at that location (e.g., soil organisms, plants), that
make it unique from all others. 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, or to predict appropriate "critical loads"
for the vast majority of U.S. ecosystems.
As additional PM speciated air quality and deposition monitoring data become available,
there is much room for productive research into the areas of uncertainty identified above. At this
time, however, staff concludes that there is insufficient information available to recommend for
consideration an ecologically defined secondary standard that is specifically targeted for
protection of vegetation and ecosystems against the adverse effects potentially associated with
the levels of PM-related stressors of nitrate and sulfate found in the ambient air.
6.4 EFFECTS ON MATERIALS
The effects of the deposition of atmospheric pollution, including ambient PM, on
materials are related to both physical damage and impaired aesthetic qualities. The deposition of
PM (especially sulfates and nitrates) 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 concrete and limestone.
Particles contribute to these physical effects because of their electrolytic, hygroscopic and acidic
properties, and their ability to sorb corrosive gases (principally SO2). As noted in the last
review, only chemically active fine-mode or hygroscopic coarse-mode particles contribute to
these physical effects (EPA 1996b, p. VIII-16).
In addition, the deposition of ambient PM can reduce the aesthetic appeal of buildings
and culturally important articles through soiling. Particles consisting primarily of carbonaceous
compounds cause soiling of commonly used building materials and culturally important items
6-50
-------
such as statues and works of art (CD, p. 4-191). 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. Soiling can be remedied by cleaning or washing, and
depending on the soiled material, repainting (EPA, 1996b, p. VIII-19).
Building upon the information presented in the last Staff Paper (EPA, 1996b), and
including the limited new information presented in Chapter 4 (section 4.4) of the CD, the
following sections summarize the physical damage and aesthetic soiling effects of PM on
materials including metals, paint finishes, and stone and concrete.
6.4.1 Materials Damage Effects
Physical damage such as corrosion, degradation, and deterioration occurs in metals, paint
finishes, and building materials such as stone and concrete, respectively. Metals are affected by
natural weathering processes even in the absence of atmospheric pollutants. Atmospheric
pollutants, most notably SO2 and particulate sulfates, can have an additive effect, by promoting
and accelerating the corrosion of metals. The rate of metal corrosion depends on a number of
factors, including the deposition rate and nature of the pollutants; the influence of the protective
corrosion film that forms on metals, slowing corrosion; the amount of moisture present;
variability in electrochemical reactions; the presence and concentration of other surface
electrolytes; and the orientation of the metal surface. Historically, studies have shown that the
rate of metal corrosion decreases in the absence of moisture, since surface moisture facilitates
the deposition of pollutants and promotes corrosive electrochemical reactions on metals (CD, pp.
4-192 to 4-193).
The CD (p. 4-194, Table 4-18) summarizes the results of a number of studies
investigating the roles of particles and SO2 on the corrosion of metals. The CD concludes that
the role of particles in the corrosion of metals is not clear (CD, p. 4-193). While several studies
suggest that particles can promote the corrosion of metals, others have not demonstrated a
correlation between particle exposure and metal corrosion. Although the corrosive effects of
SO2 exposure in particular have received much study, there remains insufficient evidence to
relate corrosive effects to specific particulate sulfate levels or to establish a quantitative
relationship between ambient particulate sulfate and corrosion.
Similar to metals, paints also undergo natural weathering processes, mainly from
exposure to environmental factors such as sunlight, moisture, fungi, and varying temperatures.
Beyond these natural processes, atmospheric pollutants can affect the durability of paint finishes
by promoting discoloration, chalking, loss of gloss, erosion, blistering, and peeling. Historical
evidence indicates that particles can damage painted surfaces by serving as carriers of more
corrosive pollutants, most notably SO2, or by serving as concentration sites for other pollutants.
If sufficient damage to the paint occurs, pollutants may penetrate to the underlying surface. A
number of studies available in the last review showed some correlation between PM exposure
6-51
-------
and damage to automobile finishes. In particular, Wolff et al. (1990) concluded that damage to
automobile finishes resulted from calcium sulfate forming on painted surfaces by the reaction of
calcium from dust particles with sulfuric acid contained in rain or dew. In addition, paint films
permeable to water are also susceptible to penetration by acid-forming aerosols (EPA 1996b, p.
VIII-18). The erosion rate of oil-based house paint has reportedly been enhanced by exposure to
SO2 and humidity; several studies have suggested that this effect is caused by the reaction of SO2
with extender pigments such as calcium carbonate and zinc oxide, although Miller et al. (1992)
suggest that calcium carbonate acts to protect paint substrates (CD, p. 4-196).
With respect to damage to building stone, numerous studies discussed in the CD (pp.
4-196 to 4-202; Table 4-19) suggest that air pollutants, including sulfur-containing pollutants
and wet or dry deposition of atmospheric particles and dry deposition of gypsum particles, can
enhance natural weathering processes. Exposure-related damage to building stone results from
the formation of salts in the stone that are subsequently washed away by rain, leaving the surface
more susceptible to the effects of air pollutants. Dry deposition of sulfur-containing pollutants
and carbonaceous particles promotes the formation of gypsum (hydrated calcium sulfate) on the
stone's surface. Gypsum is a black crusty material that occupies a larger volume than the
original stone, causing the stone's surface to become cracked and pitted, leaving rough surfaces
that serve as sites for further deposition of airborne particles (CD, p. 4-200).
The rate of stone deterioration is determined by the pollutant mix and concentration, the
stone's permeability and moisture content, and the pollutant deposition velocity. Dry deposition
of SO2 between rain events has been reported to be a major causative factor in pollutant-related
erosion of calcareous stones (e.g., limestone, marble, and carbonated cement). While it is clear
from the available information that gaseous air pollutants, in particular SO2, will promote the
decay 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 (CD, p.
4-201 , 4-202).
6.4.2 Soiling Effects
Soiling affects the aesthetic appeal of painted surfaces. In addition to natural factors,
exposure to PM may give painted surfaces a dirty appearance. Early studies demonstrated an
association between particle exposure and increased frequency of cleaning painted surfaces.
More recently, Haynie and Lemmons (1990) conducted a study to determine how various
environmental factors contribute to the rate of soiling on white painted surfaces. They reported
that coarse-mode particles initially contribute more to soiling of horizontal and vertical surfaces
than do fine-mode particles, but are more easily removed by rain, leaving stains on the painted
surface. The authors concluded that the accumulation of fine-mode particles, rather than coarse-
mode particles, more likely promotes the need for cleaning of the painted surfaces (EPA 1996b,
p. VIII-21-22; CD, pp. 4-202 to 4-204). Haynie and Lemmons (1990) and Creighton et al.
6-52
-------
(1990) reported that horizontal surfaces soiled faster than vertical surfaces and that large
particles were primarily responsible for the soiling of horizontal surfaces not exposed to rainfall.
Additionally, a study was conducted to determine the potential soiling of artwork in five
Southern California museums (Ligocki, et al., 1993). Findings were that a significant fraction of
fine elemental carbon and soil dust particles in the ambient air penetrates to the indoor
environment and may constitute a soiling hazard to displayed artwork (EPA 1996b, p. VIII-22).
As for stone structures, the presence of gypsum is related to soiling of the stone surface
by providing sites for particles of dirt to concentrate. Lorusso et al. (1997) attributed the need
for frequent cleaning and restoration of historic monuments in Rome to exposure to total
suspended particles (TSP). Further, Davidson et al. (2000) evaluated the effects of air pollution
exposure on a limestone structure on the University of Pittsburgh campus using estimated
average TSP levels in the 1930s and 1940s and actual values for the years 1957 to 1997.
Monitored levels of SO2 were also available for the years 1980 to 1998. Based on the available
data on pollutant levels and photographs, the authors concluded that soiling began while the
structure was under construction. With decreasing levels of pollution, the soiled areas have been
slowly washed away, the process taking several decades, leaving a white, eroded surface (CD,
pp. 4-203).
6.4.3 Summary and Conclusions
Damage to building materials results from natural weathering processes that are
enhanced by exposure to airborne pollution, most notably sulfur-containing pollutants. Ambient
PM has been associated with contributing to pollution-related damage to materials, and can
cause significant detrimental effects by soiling painted surfaces and other building materials.
Available data indicate that particle-related soiling can result in increased cleaning frequency
and repainting, and may reduce the useful life of the soiled materials. However, to date, no
quantitative relationships between particle characteristics (e.g., concentrations, particle size, and
chemical composition) and the frequency of cleaning or repainting have been established. Thus,
staff concludes that PM effects on materials can play no quantitative role in considering whether
any revisions of the secondary PM NAAQS are appropriate at this time.
6.5 EFFECTS ON CLIMATE CHANGE AND SOLAR RADIATION
Atmospheric particles alter the amount of electromagnetic radiation transmitted through
the earth's atmosphere by both scattering and absorbing radiation. As discussed above in
Chapter 2 (section 2.2.6), most components of ambient PM (especially sulfates) scatter and
reflect incoming solar radiation back into space, thus tending to offset the "greenhouse effect" by
having a cooling effect on climate. In contrast, some components of ambient PM (especially
black carbon) absorb incoming solar radiation or outgoing terrestrial radiation, and are believed
to contribute to atmospheric warming. Lesser impacts of atmospheric particles are associated
6-53
-------
with their role in altering the amount of ultraviolet solar radiation (especially UV-B) penetrating
through the earth's atmosphere to ground level, where it can exert a variety of effects on human
health, plant and animal biota, and other environmental components (CD, p. 205). The extensive
research and assessment efforts into global climate change and stratospheric ozone depletion
provide evidence that atmospheric particles play important roles in these two types of
atmospheric processes, not only on a global scale, but also on regional and local scales as well.
Information on the role of atmospheric particles in these atmospheric processes and the
effects on human health and the environment associated with these atmospheric processes is
briefly summarized below, based on the information in section 4.5 of the CD and referenced
reports. These effects are discussed below in conjunction with consideration of the potential
indirect impacts on human health and the environment that may be a consequence of climatic
and radiative changes attributable to local and regional changes in ambient PM.
6.5.1 Climate Change and Potential Human Health and Environmental Impacts
As discussed in section 4.5.1 of the CD, particles can have both direct and indirect effects
on climatic processes. The direct effects are the result of the same processes responsible for
visibility degradation, namely radiative scattering and absorption. However, while visibility
impairment is caused by particle scattering in all directions, climate effects result mainly from
scattering light away from the earth and into space. This reflection of solar radiation back to
space decreases the transmission of visible radiation to the surface 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 black carbon,
results in an increase in the heating rate of the lower atmosphere.
In addition to these direct radiative effects, particles can also have a number of indirect
effects on climate related to their physical properties. For example, sulfate particles can serve as
condensation nuclei which alter the size distribution of cloud droplets by producing more
droplets with smaller sizes. Because the total surface area of the cloud droplets is increased, the
amount of solar radiation that clouds reflect back to space is increased. A further important
consequence of this effect on cloud properties is the suppression of rain and potentially major
disruption of hydrological cycles downwind of pollution sources, leading to a potentially
significant alteration of climate in the affected regions (CD, p. 4-218).
The overall radiative and physical effects of particles, both direct and indirect, are not the
simple sum of effects caused by individual classes of particles because of interactions between
particles and other atmospheric gases. As discussed in Section 4.5.1.2 of the CD, the effects of
sulfate particles have been the most widely considered, with globally averaged radiative effects
of sulfate particles generally estimated to have partially offset the warming effects caused by
increases in greenhouse gases. On the other hand, global-scale modeling of mineral dust
6-54
-------
particles suggests that even the sign as well as the magnitude of effects depends on the vertical
distribution and effective particle radius.
The CD makes clear that atmospheric particles play an important role in climatic
processes, but that their role at this time remains poorly quantified. In general, on a global scale,
the direct effect of radiative scattering by atmospheric particles is to likely exert an overall net
effect of cooling the atmosphere, while particle absorption may lead to warming. The net impact
of indirect effects on temperature and rainfall patterns remains difficult to generalize. However,
deviations from global mean values can be very large even on a regional scale, with any
estimation of more localized effects introducing even greater complexity (CD, p. 216). The CD
concludes that any effort to model the impacts of local alterations in particle concentrations on
projected global climate change or consequent local and regional weather patterns would be
subject to considerable uncertainty (CD, p. 4-240).
More specifically, the CD notes that while current climate models are successful in
simulating present annual mean climate and the seasonal cycle on continental scales, they are
less successful at regional scales (CD, p. 4-207). Findings from various referenced assessments
illustrate well the considerable uncertainties and difficulties in projecting likely climate change
impacts on regional or local scales. For example, uncertainties in calculating the direct radiative
effects of atmospheric particles arise from a lack of knowledge of their vertical and horizontal
variability, their size distribution, chemical composition, and the distribution of components
within individual particles. Any complete assessment of the radiative effects of PM would
require computationally intensive calculations that incorporate the spatial and temporal behavior
of particles of varying composition that have been emitted from, or formed by precursors emitted
from, different sources. In addition, calculations of indirect physical effects of particles on
climate (e.g., related to alteration of cloud properties and disruption of hydrological cycles) are
subject to much larger uncertainties than those related to the direct radiative effects of particles
(CD, p. 4-219). The CD concludes that at present impacts on human health and the environment
due to aerosol effects on the climate system can not be calculated with confidence, and notes that
the uncertainties associated with such aerosol-related effects will likely remain much larger than
those associated with greenhouse gases (CD, p. 4-219). Nevertheless, the CD concludes that
substantial qualitative information available from observational and modeling studies indicates
that different types of atmospheric aerosols (i.e., different components of PM) have both
warming and cooling effects on climate, both globally and regionally. Studies also suggest that
global and regional climate changes could potentially have both positive and negative effects on
human health, human welfare, and the environment.
6-55
-------
6.5.2 Alterations in Solar UV-B Radiation and Potential Human Health and
Environmental Impacts
As discussed in section 4.5.2 of the CD, the effects of particles in the lower atmosphere
on the transmission of solar UV-B radiation have been examined both by field measurements
and by radiative transfer model calculations. Several studies cited in the CD reinforce the idea
that particles can play an important role in modulating the attenuation of solar UV-B radiation,
although none included measurements of ambient PM concentrations, so that direct relationships
between PM levels and UV-B radiation transmission could not be determined. The available
studies, conducted in diverse locations around the world, demonstrate that relationships between
particles and solar UV-B radiation transmission can vary considerably over location, conditions,
and time. While ambient particles are generally expected to decrease the flux of solar UV-B
radiation reaching the surface, any comprehensive assessment of the radiative effects of particles
would be location-specific and complicated by the role of particles in photochemical activity in
the lower atmosphere. Whether the photochemical production of ozone is enhanced, remains the
same, or reduced by the presence of ambient particles will be location-specific and dependent on
particle composition. Also complicating any assessment of solar UV-B radiation penetration to
specific areas of the earth's surface are the influences of clouds, which in turn are affected by the
presence of ambient particles.
The main types of effects associated with exposure to UV-B radiation include direct
effects on human health and agricultural and ecological systems, indirect effects on human
health and ecosystems, and effects on materials (CD, p. 4-221). The study of these effects has
been driven by international concern over potentially serious increases in the amount of solar
UV-B radiation reaching the earth's surface due to the depletion of the stratospheric ozone layer
by the release of various man-made ozone-depleting substances. Extensive qualitative and
quantitative characterizations of these global effects attributable to projections of stratospheric
ozone depletion have been periodically assessed in studies carried out under WMO and UNEP
auspices, with the most recent projections being published in UNEP (1998, 2000) and WMO
(1999).
Direct human health effects of UV-B radiation exposure include: skin damage (sunburn)
leading to more rapid aging and increased incidence of skin cancer; effects on the eyes, including
retinal damage and increased cataract formation possibly leading to blindness; and suppression
of some immune system components, contributing to skin cancer induction and possibly
increasing susceptibility to certain infectious diseases. Direct environmental effects include
damage to terrestrial plants, leading to possible reduced yields of some major food crops and
commercially important tress, as well as to biodiversity shifts in natural terrestrial ecosystems;
and adverse effects on aquatic life, including reductions in important components of marine food
chains as well as other aquatic ecosystem shifts. Indirect health and environmental effects are
primarily those mediated through increased tropospheric ozone formation and consequent
6-56
-------
ground-level ozone-related health and environmental impacts. Effects on materials include
accelerated polymer weathering and other effects on man-made materials and cultural artifacts.
In addition, there are emerging complex issues regarding interactions and feedbacks between
climate change and changes in terrestrial and marine biogeochemical cycles due to increased
UV-B radiation penetration. (CD, p. 4-221, 4-222).
In contrast to these types of negative impacts associated with increased UV-B penetration
to the Earth's surface, the CD (p. 4-222, 4-223) summarizes research results that are suggestive
of possible beneficial effects of increased UV-B radiation penetration. For example, a number of
studies have focused on the protective effects of UV-B radiation with regard to non-skin cancer
incidence, which proved suggestive evidence that UV-B radiation, acting through the production
of vitamin D, may be a risk-reduction factor for mortality due to several types of cancer,
including cancer of the breast, colon, ovary, and prostate, as well as non-Hodgkin lymphoma.
The various assessments of these types of effects that have been conducted consistently
note that the modeled projections quantitatively relating changes in UV-B radiation (attributable
to stratospheric ozone depletion) to changes in health and environmental effects are subject to
considerable uncertainty, with the role of atmospheric particles being one of numerous
complicating factors. Taking into account the complex interactions between ambient particles
and UV-B radiation transmission through the lower atmosphere, the CD concludes that any
effort to quantify projected indirect effects of variations in atmospheric PM on human health or
the environment due to particle impacts on transmission of solar UV-B radiation would require
location-specific evaluations that take into account the composition, concentration, and internal
structure of the particles; temporal variations in atmospheric mixing heights and depths of layers
containing the particles; and the abundance of ozone and other absorbers within the planetary
boundary layer and the free troposphere (CD, 4-226).
At present, models are not available to take such complex factors into account, nor is
sufficient data available to characterize input variables that would be necessary for any such
modeling. The CD concludes, however, that the outcome of such modeling efforts would likely
vary from location to location, even as to the direction of changes in the levels of exposures to
UV-B radiation, due to location-specific changes in ambient PM concentrations and/or
composition (CD, p. 4-227). Beyond considering just average levels of exposures to UV-B
radiation in general, the CD notes that ambient PM can affect the directional characteristics of
UV-B radiation scattering at ground-level, and thus its biological effectiveness. Also, ambient
PM can affect not only biologically damaging UV-B radiation, but can also reduce the ground-
level ratio of photorepairing UV-A radiation to damaging UV-B radiation. Further, the CD notes
that ambient PM deposition is a major source of PAH in certain water bodies, which can enhance
the adverse effects of solar UV-B radiation on aquatic organisms, such that the net effect of
ambient PM in some locations may be to increase UV-B radiation-related biological damage to
certain aquatic and terrestrial organisms. (CD, p. 4-227).
6-57
-------
6.5.3 Summary and Conclusions
A number of assessments of the factors affecting global warming and climate change as
well as those affecting the penetration of solar UV-B radiation to the earth's surface clearly
recognize ambient PM as playing various roles in these processes. These assessments, however,
have focused on global- and regional-scale impacts, allowing for generalized assumptions to take
the place of specific, but unavailable, information on local-scale atmospheric parameters and
characteristics of the distribution of particles present in the ambient air. As such, the available
information provides no basis for estimating how localized changes in the temporal, spatial, and
composition patterns of ambient PM, likely to occur as a result of expected future emissions of
particles and their precursor gases across the U.S., would affect local, regional, or global changes
in climate or UV-B radiation penetration - even the direction of such effects on a local scale
remains uncertain. Moreover, similar concentrations of different particle components can
produce opposite net effects. It follows, therefore, that there is insufficient information available
to project the extent to which, or even whether, such location-specific changes in ambient PM
would indirectly affect human health or the environment secondary to potential changes in
climate and UV-B radiation.
Based on currently available information, staff concludes that the potential indirect
effects of ambient PM on public health and welfare, secondary to potential PM-related changes
in climate and UV-B radiation, can play no quantitative role in considering whether any
revisions of the primary or secondary PM NAAQS are appropriate at this time. Even
qualitatively, the available information is very limited in the extent to which it can help inform
an assessment of the overall weight of evidence in an assessment of the net health and
environmental effects of PM in the ambient air, considering both its direct effects (e.g.,
inhalation-related health effects) and indirect effects mediated by other routes of exposure and
environmental factors (e.g., dermal exposure to UV-B radiation).
6-58
-------
REFERENCES
Section 6.2 - Visibility Impairment
Abt Associates, Inc. (2001) Assessing Public Opinions on Visibility Impairment Due to Air Pollution: Summary
Report. Prepared for EPA Office of Air Quality Planning and Standards; funded under EPA Contract No.
68-D-98-001. Bethesda, Maryland. January 2001.
Air Resource Specialists, Inc. (2003) WinHaze Air Quality Modeler, version 2.9.0. Available from
http://www.air-resource.com/whatsnew.htm
Arizona Department of Environmental Quality. (2003) Visibility Index Oversight Committee Final Report:
Recommendation for a Phoenix Area Visibility Index. March 5, 2003.
http://www.phoenixvis.net/PDF/vis 031403final.pdf.
BBC Research & Consulting. (2002) Phoenix Area Visibility Survey. Draft Report. October 4, 2002.
http://www.bbcresearch.com/library/visibilitv draft report.pdf
California Code of Regulations. Title 17, Section 70200, Table of Standards.
Chestnut, L. G.; Rowe, R. D. (1990) Preservation values for visibility in the national parks. Washington, DC: U.S.
Environmental Protection Agency.
Chestnut, L. G.; Rowe, R. D. (1991) Economic valuation of changes invisibility: A state of the science assessment.
Sector B5 Report 27. In Acidic Depositions: State of Science and Technology Volume IV Control
Technologies, Future Emissions and Effects Valuation. P.M. Irving (ed.). The U.S. National Acid
Precipitation Assessment Program. GPO, Washington, D.C.
Chestnut, L.G.; Dennis, R. L.; Latimer, D. A. (1994) Economic benefits of improvements invisibility: acid rain
provisions of the 1990 clean air act amendments. Proceedings of Aerosols and Atmospheric Optics:
Radiative Balance and Visual Air Quality. Air & Waste Management Association International Specialty
Conference, pp. 791-802.
Chestnut, L. G.; Dennis, R. L. (1997) Economic benefits of improvements invisibility: acid rain provisions of the
1990 clean air act amendments. J. Air Waste Manage. Assoc. 47:395-402.
Cohen, S.; Evans, G.W.; Stokols, D.; Krantz, D.S. (1986) Behavior, Health, and Environmental Stress. Plenum
Press. New York, NY.
Department of Interior. (1998) Air Quality in the National Parks. Natural Resources Report 98-1. National Park
Service, Air Quality Division. Denver, Colorado.
Ely, D.W.; Leary, J.T.; Stewart, T.R.; Ross, D.M. (1991) The Establishment of the Denver Visibility Standard. For
presentation at the 84th Annual Meeting & Exhibition of the Air and Waste Management Association, June
16-21, 1991.
Environmental Protection Agency. (1982) Review of the National Ambient Air Quality Standards for Paniculate
Matter, Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park,
N.C.: Office of Air Quality Planning and Standards, Strategies and Air Standards Division. Report no.
EPA-450/5-82-001.
6-59
-------
Environmental Protection Agency. (1993) Air Quality Criteria for Oxides of Nitrogen. Research Triangle Park, NC:
Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office. Report
no. EPA-600/8-91/049F.
Environmental Protection Agency. (1996a) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
National Center for Environmental Assessment-RTF Office; report no. EPA/600/P-95/001aF-cF. 3v.
Environmental Protection Agency. (1996b) 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-452YR-96-013.
Environmental Protection Agency. (1999) Regional Haze Regulations. 40 CFR Part 51.300-309. 64 Federal
Register 35713.
Environmental Protection Agency. (2000) Guidelines for Preparing Economic Analyses. Washington, DC: Office of
the Administrator. EPA240-R-00-003.
Environmental Protection Agency. (2001) National Air Quality and Emissions Trends Report, 1999. Research
Triangle Park, NC: Office of Air Quality Planning and Standards. Report no. EPA/454/R-01-004. March.
Grand Canyon Visibility Transport Commission (1996) Report of the Grand Canyon Visibility Transport
Commission to the United States Environmental Protection Agency.
Hass, G. E.; Wakefield, T.J. (1998) National Parks and the American Public: A National Public Opinion Survey of
the National Park System. Colorado State University, Department of Natural Resource Recreation and
Tourism, College of Natural Resources, Fort Collins, CO. Report prepared for the National Parks and
Conservation Association. June 1998.
McNeill, R. and Roberge, A. (2000) The Impact of Visual Air Quality on Tourism Revenues in Greater Vancouver
and the Lower Fraser Valley. Environment Canada, Georgia Basin Ecosystem Initiative. GBEI report no.
EC/GB-00-028.
Middleton, P. (1993) Brown Cloud II: The Denver Air Quality Modeling Study, Final Summary Report. Metro
Denver Brown Cloud Study, Inc. Denver, CO.
Molenar, J.V.; Malm, W.C.; Johnson, C.E. (1994) Visual Air Quality Simulation Techniques. Atmospheric
Environment. Volume 28, Issue 5, 1055-1063.
Molenar, J.V. (2000) Visibility Science and Trends in the Lake Tahoe Basin: 1989-1998. Report by Air Resource
Specialists, Inc., to Tahoe Regional Planning Agency. February 15, 2000.
National Acid Precipitation Assessment Program (NAPAP) (1991) Acid Deposition: State of Science and
Technology. Report 24. Visibility: Existing and Historical Conditions - Causes and Effects. Washington,
DC.
National Acid Precipitation Assessment Program (NAPAP). (1998) Biennial Report to Congress: an
Integrated Assessment.
http://dwb.unl.edu/Teacher/NSF/C14/C14Links/www.nnic.noaa.gov/CENR/NAP AP/N APAP_96.htm
National Research Council. (1993) Protecting Visibility in National Parks and Wilderness Areas. National
Academy of Sciences Committee on Haze in National Parks and Wilderness Areas. National Academy
Press: Washington, DC.
6-60
-------
National Transportation Safety Board (NTSB). (2000) NTSB Report NYC99MA178, July 6, 2000. Report on July
16, 1999 fatal accident at Vineyard Haven, MA.
National Weather Service. (1998) Automated Surface Observing System (ASOS) User's Guide. ASOS Program
Office. Silver Spring, MD.
New Zealand Ministry for the Environment. (2000) Proposals for Revised and New
Ambient Air Quality Guidelines: Discussion Document. Air Quality Report No. 16. December.
New Zealand National Institute of Water & Atmospheric Research (NIWAR). (2000a) Visibility in New Zealand:
Amenity Value, Monitoring, Management and Potential Indicators. Air Quality Technical Report 17.
Prepared for New Zealand Ministry for the Environment. Draft report.
New Zealand National Institute of Water & Atmospheric Research (NIWAR). (2000b) Visibility in New Zealand:
National Risk Assessment. Air Quality Technical Report 18. Prepared for New Zealand Ministry for the
Environment. Draft report.
Peacock, B.; Killingsworth, C.; Simon, B. (1998) State and National Economic Impacts Associated with Travel
Related Expenditures by Recreational Visitors to Lands Managed by the U.S. Department of Interior. U.S.
Department of the Interior. January.
Pryor, S.C. (1996) Assessing Public Perception of Visibility for Standard Setting Exercises. Atmospheric
Environment, vol. 30, no. 15, pp. 2705-2716.
Schichtel, B.A., Husar, R.B., Falke, S.R., and Wilson, W.E. (2001) "Haze Trends over the United States,
1980-1995," Atmospheric Environment, vol. 35, no. 30, pp. 5205-5210.
Schmidt, S.M., Mintz, D., Rao, T., and McCluney, L. (2005) Draft analysis of PM ambient air quality data for the
PM NAAQS review. Memorandum to PM NAAQS review docket OAR-2001-0017. January 31, 2005.
Schulze, W. D.; Brookshire, D. S.; Walther, E. G.; MacFarland, K. K.; Thayer, M. A.; Whitworth, R. L.; Ben-Davis,
S.; Malm, W.; Molenar, Jr. (1983) The Economic Benefits of Preserving Visibility in the National
Parklands of the Southwest. Nat. Resour. J. 23: 149-173.
Sisler, J.; Malm, W.; Molenar, J.; Gebhardt, K. (1996) Spatial and Seasonal Patterns and Long Term Variability of
the Chemical Composition of Haze in the U.S.: An Analysis of Data from the IMPROVE Network. Fort
Collins, CO: Cooperative Institute for Research in the Atmosphere, Colorado State University.
State Government of Victoria, Australia. (2000a) Draft Variation to State Environment Protection Policy (Air
Quality Management) and State Environment Protection Policy (Ambient Air Quality) and Draft Policy
Impact Assessment. Environment Protection Authority. Publication 728. Southbank, Victoria.
State Government of Victoria, Australia. (2000b) Year in Review. Environment Protection Authority. Southbank,
Victoria.
Section 6.3 - Vegetation and Ecosystems
Aber, J. D.; Nadelhoffer, K. J.; Steudler, P.; Melillo, J. M. (1989) Nitrogen saturation in northern forest ecosystems:
excess nitrogen from fossil fuel combustion may stress the biosphere. Bioscience 39: 378-386.
Aber, J. D.; Magill, A.; McNulty, S. G.; Boone, R. D.; Nadelhoffer, K. J.; Downs, M.; Hallett, R. (1995) Forest
biogeochemistry and primary production altered by nitrogen saturation. Water Air Soil Pollut. 85:
1665-1670.
6-61
-------
Aber, I; McDowell, W.; Nadelhoffer, K.; Magill, A.; Berntson, G.; Kamakea, M.; McNulty, S.; Currie, W.;
Rustad, L.; Fernandez, I. (1998) Nitrogen saturation in temperate forest ecosystems. BioScience 48:
921-934.
Allen, E. B.; Padgett, P. E.; Bytnerowicz, A.; Minich, R. (1998) Nitrogen deposition effects on coastal sage
vegetation of southern California. USDA Forest Service Gen. Tech. Rep. PSW-GTR-166, pp. 131-139.
Andersen, C. P.; Rygiewicz, P. T. (1991) Stress interactions and mycorrhizal plant response: understanding carbon
allocation priorities. Environ. Pollut. 73: 217-244.
Antoine, M.E. (2001) Ecophysiology of the cyanolichen, Lobaria oregana. Master's thesis. Oregon State University,
Corvallis.
Bailey, S.W., Horsley, S.B., Long, R.P., Hallet, R. A. (1999) Influence of geologic and pedologic factors on health of
sugar maple on the Allegheny Plateau, U.S. In Horsley, S.B. and Long, R.P., eds. Sugar Maple Ecology
and Health: Proceedings of an International Symposium. Radnor, PA: U.S. Department of Agriculture,
Forest Service. General Technical Report NE-261. PP. 63-65.
Brooks, M.L. (2003) Effects of increased soil nitrogen on the dominance of alien annual plants in the Mojave Desert.
Journal of Applied Ecology. 40:344-353.
Bytnerowicz, A.; Fenn, M. E. (1996) Nitrogen deposition in California forests: a review. Environ. Pollut.
92: 127-146.
Charles, D.F., ed. (1991) Acidic Deposition and Aquatic Ecosystems. Regional Case Studies. New York: Springer-
Verlag.
Craig, B.W. and Friedland, A.J. (1991) Spatial patterns in forest composition and standing dead red spruce in
montane forests of the Adirondacks and northern Appalachians. Environmental Monitoring and
Assessment. 18:129-140.
Cronan, C. S.; Grigal, D. F. (1995) Use of calcium/aluminum ratios as indicators of stress in forest ecosystems.
J. Environ. Qual. 24: 209-226.
D'Antonio, C.M. and Vitousek, P.M. (1992) Biological invasions by exotic grasses: The grass-fire cycle and global
change. Annual Review of Ecology and Systematics. 23: 63-87.
DeHayes, D. H.; Schaberg, P. G.; Hawley, G. J.; Strimbeck, G. R. (1999) Acid rain impacts on calcium nutrition and
forest health. Bioscience 49: 789-800.
Driscoll, C. T.; Wyskowski, B. J.; DeStaffan, P.; Newton, R. M. (1989) Chemistry and transfer of aluminum in a
forested watershed in the Adirondack region of New York, USA. In: Lewis, T. E., ed. Environmental
chemistry and toxicology of aluminum. Chelsea, MI: Lewis Publishers, Inc.; pp. 83-105.
Driscoll, C.T., Likens, G.E., Church, M.R. (1998) Recovery of surface waters in the northeastern U.S. from
decreases in atmospheric deposition of sulfur. Water, Air and Soil Pollution. 105:319-329.
Driscoll, C. T.; Lawrence, G. B.; Bulger, A. J.; Butler, T. J.; Cronan, C. S.; Eagar, C.; Lambert, K. F.; Likens, G. E.;
Stoddard, J. L.; Weathers, K. C. (2001) Acidic deposition in the northeastern United States: sources and
inputs, ecosystem effects, and management strategies. BioScience 51: 180-198.
Edgerton-Warburton, L. M.; Allen, E. B. (2000) Shifts in arbuscular mycorrhizal communities along an
anthropogenic gradient nitrogen deposition gradient. Ecol. Appl. 10: 484-496.
6-62
-------
Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
Assessment Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA;
PB84-156777.
Environmental Protection Agency. (1992) Framework for Ecological Risk Assessment Washington, D.C.: Risk
Assessment Forum, U.S. Environmental Protection Agency. EPA/630/R-92/001.
Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park, NC:
Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525,
andPB95-124517.
Environmental Protection Agency. (1996a) Air quality criteria for paniculate matter. Research Triangle Park, NC:
National Center for Environmental Assessment-RTF Office; report nos. EPA/600/P-95/001aF-cF. 3v.
Environmental Protection Agency. (1997) Nitrogen oxides: impacts on public health and the environment.
Washington, DC: Office of Air and Radiation; August. Available:
www.epa.gov/ttncaaal/tl/reports/noxrept.pdf [1999, November 24].
Environmental Protection Agency. (1998) Guidelines for Ecological Risk Assessment. Washington, D.C: Risk
Assessment Forum, U.S. Environmental Protection Agency. EPA/630/R-95/002F.
Environmental Protection Agency. (2003) Response Of Surface Water Chemistry to the Clean Air Act Amendments
of 1990. National Health and Environmental Effects Research Laboratory, Office of Research and
Development, U.S. Environmental Protection Agency. Research Triangle Park, NC. EPA 620/R-03/001.
Fenn, M. E.; Poth, M. A.; Aber, J. D.; Baron, J. S.; Bormann, B. T.; Johnson, D. W.; Lemly, A. D.; McNulty, S. G.;
Ryan, D. F.; Stottlemyer, R. (1998) Nitrogen excess in North American ecosystems: predisposing factors,
ecosystem responses, and management strategies. Ecol. Appl. 8: 706-733.
Fenn, M. E.; Baron, J. S.; Allen, E. B.; Rueth, H. M.; Nydick, K. R.; Geiser, L.; Bowman, W. D.; Sickman, J. O.;
Meixner, T.; Johnson, D. W.; Neitlich, P. (2003) Ecological effects of nitrogen deposition in the western
United States. BioScience 53: 404-420.
Galloway, J.N, Norton, S.N., Church, M.R. (1983) Freshwater acidification from atmospheric deposition of sulfuric
acid: A conceptual model. Environmental Science and Technology. 17:541A-545A
Galloway, J. N. (1998) The global nitrogen cycle: changes and consequences. Environ. Pollut. 102(suppl. 1): 15-24.
Galloway, J. N.; Cowling, E. B. (2002) Reactive nitrogen and the world: 200 years of change. Ambio 31: 64-71.
Galloway, J. N.; Aber, J. D.; Erisman, J. W.; Seitzinger, S. P.; Howarth, R. W.; Cowling, E. B.; Cosby, B. J. (2003)
The nitrogen cascade. BioScience 53: 341-356.
Garner, J. H. B. (1994) Nitrogen oxides, plant metabolism, and forest ecosystem response. In: Alscher, R. G.;
Wellburn, A. R., eds. Plant responses to the gaseous environment: molecular, metabolic and physiological
aspects, [3rd international symposium on air pollutants and plant metabolism]; June 1992; Blacksburg, VA.
London, United Kingdom: Chapman & Hall; pp. 301-314.
Gunn, J.M. and Mills, K.H. (1998) The potential for restoration of acid-damaged lake trout lakes. Restoration
Ecology. 6:390-397.
6-63
-------
Gundersen, P.; Callesen, I.; De Vries, W. (1998) Nitrate leaching in forest ecosystems is related to forest floor C/N
ratios. Environ. Pollut. 102(suppl. 1): 403-407.
Hedin, L. O.; Granat, L.; Likens, G. E.; Buishand, T. A.; Galloway, J. N.; Butler, T. I; Rodhe, H. (1994)
Steep declines in atmospheric base cations in regions of Europe and North America. Nature (London)
367: 351-354.
Hornung, M; Langan, S. J. (1999) Nitrogen deposition: sources, impacts and responses in natural and semi-natural
ecosystems. In: Langan, S. J., ed. Impact of nitrogen deposition on natural ecosystems and semi-natural
ecosystems. Dordrecht, The Netherlands: Kluwer Academic Publishers; pp. 1-13. [Environmental Pollution,
Volume 3].
Horsley, S.B., Long, R.P., Bailey, S.W., Hallet, R.A., Hall, T.J. (1999) Factors contributing to sugar maple decline
along topographic gradients on the glaciated and unglaciated Allegheny Plateau. In Horsley, S.B. and
Long, R.P., eds. Sugar Maple Ecology and Health: Proceedings of an International Symposium. Radnor,
PA: U.S. Department of Agriculture, Forest Service. General Technical Report NE-261. PP. 60-62.
Howarth, R. W.; Boyer, E. W.; Pabich, W. J.; Galloway, J. N. (2002) Nitrogen use in the United States from
1961-2000 and potential future trends. Ambio 31: 88-96.
Johnson, A.H., Friedland, A.J., Dushoff, J.G. (1984) Recent and historic red spruce mortality: Evidence of climatic
influence. Water, Air and Soil Pollution. 30:319-330.
Johnson, D. W.; Lindberg, S. E., eds. (1992) Atmospheric deposition and forest nutrient cycling: a synthesis of the
integrated forest study. New York, NY: Springer-Verlag, Inc. (Billings, W. D.; Golley, F.; Lange, O. L.;
Olson, J. S.; Remmert, H., eds. Ecological studies: analysis and synthesis: v. 91).
Johnson, D.W.; Susfalk, R.B.; Brewer, P.,F.; Swank, W.T. (1999) Simulated effects of reduced sulfur, nitrogen, and
base cation deposition on soils and solutions in southern Appalachian forests. J. Environ. Qua! 28: 1336-1346.
Keller, W. and Gunn, J.M. (1995) Lake water quality improvements and recovering aquatic communities. In Gunn,
J.M. ed. Restoration and Recovery of an Industrial Region: Progress in Restoring the Smelter-damaged
Landscape Near Sudbury, Canada. New York: Springer-Verlag. PP. 67-80.
Kretser, W., Gallagher, J., Nicolette, J. (1989) Adirondack Lake Study. 1984-1987. An Evaluation of Fish
Communities and Water Chemistry. Ray Brook, New York: Adirondacks Lakes Survey Corporation.
Likens, G.E. and Bormann, F.H. (1995) Biogeochemistry of a Forested Ecosystem. 2nd ed., New York: Springer-
Verlag.
Lekke, H.; Bak, J.; Falkengren-Grerup, U.; Finlay, R. D.; Ilvesniemi, H.; Nygaard, P. H.; Starr, M. (1996)
Critical loads of acidic deposition for forest soils: is the current approach adequate. Ambio 25: 510-516.
McCune, B. and Geiser, L. (1997) Macrolichens of the Pacific Northwest. Corvallis: Oregon State University Press.
McDonnell, M. J.; Pickett, S. T. A.; Groffman, P.; Bohlen, P.; Pouyat, R. V; Zipperer, W. C.; Parmelee, R. W.;
Carreiro, M. M.; Medley, K. (1997) Ecosystem processes along an urban-to-rural gradient. Urban Ecosyst.
1:21-36.
Minnich, R.A. and Dezzani, R.J. (1998) Historical decline of coastal sage scrub in the Riverside-Ferris Plain,
California. Western Birds. 29:366-391.
Nagy, K.A., Henen, B.T., Vyas, D.B. (1998) Nutritional quality of native and introduced food plants of wild desert
tortoises. Journal of Herpetology 32: 260-267.
6-64
-------
Nash, T.H. and Sigal, L.L. (1999) Ephiphytic lichens in the San Bernardino mountains in relation to oxidant
gradients. In Miller, P.R. McBride, J.R., eds. Oxidant Air Pollution Impacts on the Montane Forests of
Southern California: A Case Study of the San Bernardino Mountains. Ecological Studies 134. New York:
Springer-Verlag. PP. 223-234.
National Science and Technology Council. (1998) National acid precipitation assessment program biennial report to
Congress: an integrated assessment; executive summary. Silver Spring, MD: U.S. Department of
Commerce, National Oceanic and Atmospheric Administration. Available:
www.nnic.noaa.gov/CENR/NAPAP/NAPAP_96.htm [24 November 1999].
Padgett, P. E.; Allen, E. B.; Bytnerowicz, A.; Minich, R. A. (1999) Changes in soil inorganic nitrogen as related to
atmospheric nitrogenous pollutants in southern California. Atmos. Environ. 33: 769-781.
Paerl, H. W.; Bales, J. D.; Ausley, L. W.; Buzzelli, C. P.; Crowder, L. B.; Eby, L. A.; Go, M.; Peierls, B. L.;
Richardson, T. L.; Ramus, J. S. (2001) Ecosystem impacts of three sequential hurricanes (Dennis, Floyd,
and Irene) on the United States' largest lagoonal estuary, Pamlico Sound, NC. Proc. Nat. Acad. Sci. U. S. A.
98:5655-5611.
Pike, L.H. (1978) The importance of epiphytic lichens in mineral cycling. Bryologist 81: 247-257.
Rabalais, N. N. (2002) Nitrogen in aquatic ecosystems. Ambio 31: 102-112.
Rago, P.J. and Wiener, J.G. (1986) Does pH affect fish species richness when lake area is considered? Transactions
of the American Fisheries Society. llb:438-447.
Riebsame, W.E., Robb, J.J., Gosnell, H., Theobald, D., Breding, P., Hanson, C., Rokoske, K. (1997) Atlas of the
New West: Portrait of a Changing Region. New York: W. W. Norton.
Rovira, A. D.; Davey, C. B. (1974) Biology of the rhizosphere. In: Carson, E. W., ed. The plant root and its
environment: proceedings of an institute; July, 1971; Blacksburg, VA. Charlottesville, VA: University
Press of Virginia; pp. 153-204.
Schindler, D.W., Mills, K.H., Malley, D.F., Findlay, S., Schearer, J.A., Davies, I.J., Turner, M.A., Lindsey, G.A.,
Cruikshank, D.R. (1985) Long-term ecosystem stress: Effects of years of experimental acidification.
Canadian Journal of Fisheries and Aqautic Science. 37:342-354.
Science Advisory Board. (2002) A framework for assessing and reporting on ecological condition: an SAB report.
Young, T. F.; Sanzone, S., eds. Washington, DC: U.S. Environmental Protection Agency, Ecological
Processes and Effects Committee; report no. EPA-SAB-EPEC-02-009. Available at:
http://www.epa.gov/sciencel/index.html (27 January 2003).
Shortle, W. C.; Smith, K. T.; Minocha, R.; Lawrence, G. B.; David, M. B. (1997) Acidic deposition, cation
mobilization, and biochemical indicators of stress in healthy red spruce. J. Environ. Qual. 26: 871-876.
Skinner, M.W.; Pavlik, B.M. (1994) CNPS Inventory of Rare and Endangered Vascular Plants of California.
Sacramento: California Native Plant Society.
Smith, W. H. (1990) Forests as sinks for air contaminants: soil compartment. In: Air pollution and forests:
interactions between air contaminants and forest ecosystems. 2nd ed. New York, NY: Springer-Verlag;
pp. 113-146. (Springer series on environmental management).
Sollins, P., Grier, C.C., McCorison, P.M., Cromack, K., Jr., Fogel, R., Fredriksen, R.L. (1980) The internal element
cycles of an old-growth Douglas-fir ecosystem in western Oregon. Ecological Monographs 50:261-285.
6-65
-------
Stanhill, G.; Cohen, S. (2001) Global dimming: a review of the evidence for a widespread and significant reduction
in global radiation with discussion of its probable causes and possible agricultural consequences. Agric.
For. Meteorol. 107: 255-278.
Van Egmond, K.; Bresser, T.; Bouwman, L. (2002) The European nitrogen case. Ambio 31: 72-78.
van Herk, C.M. (1999) Mapping of ammonia pollution with epiphytic lichens in the Netherlands. Lichenologist
31:9-20.
Van Sickle, I, Baker, J.P., Simonin, H.A., Baldigo, B.P., Kretser, W.A., Sharpe, W.F. (1996) Episodic acidification
of small streamsin the northeastern U.S.: Fish mortality in field bioassays. Ecological Applications. 6:408-
421.
Vitousek, P. M.; Mooney, H. A.; Lubchenco, J.; Melillo, J. M. (1997) Human domination of Earth's ecosystems.
Science (Washington, DC) 277: 494-499.
Wall, D. H.; Moore, J. C. (1999) Interactions underground: soil biodiversity, mutualism, and ecosystem processes.
Bioscience 49: 109-117.
Wedin, D. A.; Tilman, D. (1996) Influence of nitrogen loading and species composition on the carbon balance of
grasslands. Science 274: 1720-1723.
Wesely, M. L.; Hicks, B. B. (2000) A review of the current status of knowledge on dry deposition. Atmos. Environ.
34: 2261-2282.
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
World Health Organization. (1997) Nitrogen oxides. 2nded. Geneva, Switzerland: World Health Organization.
(Environmental health criteria 188).
Section 6.4 - Man-made Materials
Creighton, P. J.; Lioy, P. J.; Haynie, F. H.; Lemmons, T. J.; Miller, J. L.; Gerhart, J. (1990) Soiling by atmospheric
aerosols in an urban industrial area. J. Air Waste Manage. Assoc. 40: 1285-1289.
Davidson, C. L; Tang, W.; Finger, S.; Etyemezian, V.; Striegel, M. F.; Sherwood, S. I. (2000) Soiling patterns on a
tall limestone building: changes over 60 years. Environ. Sci. Technol. 34: 560-565.
Environmental Protection Agency. (1996b) 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-452YR-96-013.
Haynie, F.H.; Lemmons, T. J. (1990) Paniculate matter soiling of exterior paints at a rural site. Aerosol Sci. Technol.
13: 356-367.
Ligocki, M. P.: Salmon, L. G.; Fall, T.; Jones, M. C.; Nazaroff, W. W.; Cass, G. R. (1993) Characteristics of
airborne particles inside southern California museums. Atmos. Environ. Part A 27: 697-711.
Lorusso, S.; Marabelli, M.; Troili, M. (1997) Air pollution and the deterioration of historic monuments. J. Environ.
Pathol. Toxicol. Oncol. 16: 171-173.
Miller, W. C.; Fornes, R. E.; Gilbert, R. D.; Speer, A.; Spence, J. (1992) Removal of CaCO3 extender in residential
coatings by atmospheric acidic deposition. In: Measurement of toxic and related air pollutants: proceedings
6-66
-------
of the 1992 U. S. EPA/A&WMA international symposium. Pittsburgh, PA: Air & Waste Management
Association; pp. 129-134. (A&WMA publication VIP-25)
Wolff, G. T.; Collins, D. C.; Rodgers, W. R.; Verma, M. H.; Wong, C. A. (1990) Spotting of automotive finishes
from the interactions between dry deposition of crustal material and wet deposition of sulfate. J. Air Waste
Manage. Assoc. 40: 1638-1648.
Section 6.5 - Climate Change and Solar Radiation
Intergovernmental Panel on Climate Change (IPCC). (1998) The regional impacts of climate change: an assessment
of vulnerability. Cambridge, United Kingdom: Cambridge University Press.
Intergovernmental Panel on Climate Change (IPCC). (200la) Climate change 2001: the scientific basis.
Contribution of working group I to the third assessment report of the Intergovernmental Panel on Climate
Change. Cambridge, United Kingdom: Cambridge University Press.
Intergovernmental Panel on Climate Change (IPCC). (200 Ib) Climate change 2001: impacts, adaptation, and
vulnerability. Contribution of working group II to the third assessment report of the Intergovernmental
Panel on Climate Change. Cambridge, United Kingdom: Cambridge University Press.
National Academy of Sciences (NAS). (2001) Committee on the Science of Climate Change, National Research
Council. Climate Change Science: An Analysis of Some Key Questions, National Academy Press,
Washington, DC.
United Nations Environment Programme (UNEP). (1998) Environmental effects of ozone depletion: 1998
assessment. J. Photochem. Photobiol. B 46: 1-4.
United Nations Environment Programme (UNEP). (2000) Environmental effects of ozone depletion: interim
summary. Available at: http://www.gcrio.org/ozone/unep2000summary.html (9 April 2002).
World Meteorological Organization. (1999) Scientific assessment of ozone depletion: 1998. Geneva, Switzerland:
World Meteorological Organization, Global Ozone and Monitoring Project; report no. 44.
6-67
-------
-------
7. STAFF CONCLUSIONS AND RECOMMENDATIONS ON
SECONDARY PM NAAQS
7.1 INTRODUCTION
This chapter presents staff conclusions and recommendations for the Administrator to
consider in deciding whether the existing secondary PM standards should be revised and, if so,
what revised standards are appropriate. The existing suite of secondary PM standards, which is
identical to the suite of primary PM standards, includes annual and 24-hour PM2 5 standards and
annual and 24-hour PM10 standards. This existing suite of secondary PM standards is intended to
address visibility impairment associated with fine particles and materials damage and soiling
related to both fine and coarse particles. Each of these standards is defined in terms of four basic
elements: indicator, averaging time, level and form. Staff conclusions and recommendations on
these standards are based on the assessment and integrative synthesis of information related to
welfare effects presented in the CD and on staff analyses and evaluations presented in Chapters 2
and 6 herein.
In recommending a range of secondary standard options for the Administrator to
consider, staff notes that the final decision is largely a public policy judgment. A final decision
must draw upon scientific evidence and analyses about effects on public welfare, as well as
judgments about how to deal with the range of uncertainties that are inherent in the relevant
information. The NAAQS provisions of the Act require the Administrator to establish secondary
standards that, in the judgment of the Administrator, are requisite to protect public welfare1 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 to eliminate all risk of adverse welfare effects, but rather at a level requisite to protect
public welfare from those effects that are judged by the Administrator to be adverse.
7.2 APPROACH
Similar to the approach discussed in Chapter 5, section 5.2, for the review of the primary
NAAQS, staffs approach here can be framed by a series of questions that may be applicable for
each category of PM-related welfare effects identified in the CD as being associated with the
presence of the pollutant in the ambient air. Staffs review of the adequacy of the current PM
standards for each effects category involves addressing questions such as:
1 As noted in Chapter 1, 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."
7-1
-------
• To what extent does the available information demonstrate or suggest that PM-
related effects are occurring at current ambient conditions or at levels that would
meet the current standards?
• To what extent does the available information inform judgments as to whether
any observed or anticipated effects are adverse to public welfare?
• To what extent are the current secondary standards likely to be effective in
achieving protection against any identified adverse effects?
To the extent that the available information suggests that revision of the current
secondary standards would be appropriate for an effects category, staff then identifies ranges of
standards (in terms of indicators, averaging times, levels, and forms) that would reflect a range
of alternative policy judgments as to the degree of protection that is requisite to protect public
welfare from known or anticipated adverse effects. In so doing, staff addresses questions such
as:
• Does the available information provide support for considering different PM indicators?
Does the available information provide support for considering different averaging
times?
• What range of levels and forms of alternative standards is supported by the information,
and what are the uncertainties and limitations in that information?
To what extent would specific levels and forms of alternative standards reduce adverse
impacts attributable to PM, and what are the uncertainties in the estimated reductions?
Based on the available information, estimated reductions in adverse effects, and related
uncertainties, staff makes recommendations as to ranges of alternative standards for the
Administrator's consideration in reaching decisions as to whether to retain or revise the
secondary PM NAAQS.
In presenting this approach, staff recognizes that for some welfare effects the currently
available information falls short of what is considered sufficient to serve as a basis for a distinct
standard defined specifically in terms of the relationship between ambient PM and those effects.
In the case of visibility impairment, however, the available information may well provide a basis
for a distinctly defined standard. In either case, staff believes it is appropriate to consider the
extent to which the current or recommended primary standards may afford protection against the
identified welfare effects.
Staff first considers information related to the effects of ambient PM, especially fine
particles, on visibility impairment in section 7.3, and makes recommendations that consideration
7-2
-------
be given to a revised secondary PM25 standard defined in terms of that effect. Other PM-related
welfare effects, including effects on vegetation and ecosystems, materials, and global climate
change processes, are addressed in section 7.4. This chapter concludes with a summary of key
uncertainties associated with establishing secondary PM standards and related staff research
recommendations in section 7.5.
7.3 STANDARDS TO ADDRESS VISIBILITY IMPAIRMENT
In 1997, EPA decided to address the effects of PM on visibility by setting secondary
standards identical to the suite of PM25 primary standards, in conjunction with the future
establishment of a regional haze program under sections 169A and 169B of the Act (62 FR at
38679-83). In reaching this decision, EPA first concluded that PM, especially fine particles,
impairs visibility in various locations across the country, including multi-state regions, urban
areas, and remote Class I Federal areas (e.g., national parks and wilderness areas). EPA also
concluded that addressing visibility impairment solely through setting more stringent national
secondary standards would not be an appropriate means to protect the public welfare from
adverse impacts of PM on visibility in all parts of the country. As a consequence, EPA
determined that an approach that combined national secondary standards with a regional haze
program was the most appropriate and effective way to address visibility impairment.
In reaching these conclusions in 1997, EPA recognized, based on observations from
available monitoring data, primarily from rural sites in the IMPROVE monitoring network, that
the selection of an appropriate level for a national secondary standard to address visibility
protection was complicated by regional differences in visibility impairment. These differences
were due to several factors, including background and current levels of PM, the composition of
PM, and average relative humidity. As a result of these regional differences, EPA noted that a
national standard intended to maintain or improve visibility conditions in many parts of the West
would have to be set at or below natural background levels in the East; conversely, a national
standard that would improve visibility in the East would permit further degradation in the West.
Beyond such problems associated with regional variability, EPA also determined that there was
not sufficient information available to establish a standard level to protect against visibility
conditions generally considered to be adverse in all areas.
These considerations led EPA to assess whether the protection afforded by the
combination of the selected primary PM25 standards and a regional haze program would provide
appropriate protection against the effects of PM on visibility. Based on such an assessment,
EPA determined that attainment of the primary PM25 standards through the implementation of
regional control strategies would be expected to result in visibility improvements in the East at
both urban and regional scales, but little or no change in the West, except in and near certain
urban areas. Further, EPA determined that a regional haze program that would make significant
progress toward the national visibility goal in Class I areas would also be expected to improve
7-3
-------
visibility in many urban and non-Class I areas as well. EPA also noted, however, that the
combined effect of the PM NAAQS and regional haze programs may not address all situations in
which people living in certain urban areas may place a particularly high value on unique scenic
resources in or near these areas. EPA concluded that such situations were more appropriately
and effectively addressed by local visibility standards, such as those established by the city of
Denver, than by national standards and control programs.
As anticipated in the last review, EPA promulgated a regional haze program in 1999.
That program requires States to establish goals for improving visibility in Class I areas and to
adopt control strategies to achieve these goals. More specifically, States are required to establish
goals for improving visibility on the 20% most impaired days in each Class I area, and for
allowing no degradation on the 20% least impaired days. Since strategies to meet these goals are
to reflect a coordinated approach among States, multi-state regional planning organizations have
been formed and are now developing strategies, to be adopted over the next few years, that will
make reasonable progress in meeting these goals.
7.3.1 Adequacy of Current PM2 5 Standards
In considering the information now available in this review, as discussed in Chapters 2
and 6 (section 6.2), staff notes that, while new research has led to improved understanding of the
optical properties of particles and the effects of relative humidity on those properties, it has not
changed the fundamental characterization of the role of PM, especially fine particles, in visibility
impairment from the last review. However, extensive new information now available from
visibility and fine particle monitoring networks has allowed for updated characterizations of
visibility trends and current levels in urban areas, as well as Class I areas. These new data are a
critical component of the analysis presented in section 6.2.3 that better characterizes visibility
impairment in urban areas.
Based on this information, staff has first considered the extent to which available
information shows PM-related impairment of visibility at current ambient conditions in areas
across the U.S. Taking into account the most recent monitoring information and analyses, staff
makes the following observations:
In Class I areas, visibility levels on the 20% haziest days in the West are about equal to
levels on the 20% best days in the East. Despite improvement through the 1990's,
visibility in the rural East remains significantly impaired, with an average visual range of
approximately 20 km on the 20% haziest days (compared to the naturally occurring
visual range of about 150 + 45 km). In the rural West, the average visual range showed
little change over this period, with an average visual range of approximately 100 km on
the 20% haziest days (compared to the naturally occurring visual range of about 230 +
40 km).
7-4
-------
• In urban areas, visibility levels show far less difference between eastern and western
regions. For example, based on reconstructed light extinction values calculated from
24-hour average PM2 5 concentrations, the average visual ranges on the 20% haziest days
in eastern and western urban areas are approximately 20 km and 27 km, respectively.
Even more similarity is seen in considering 4-hour (12:00 to 4:00 pm) average PM25
concentrations, for which the average visual ranges on the 20% haziest days in eastern
and western urban areas are approximately 26 km and 31 km, respectively (Schmidt et
al., 2005).
Based on this information, and on the recognition that efforts are now underway to
address all human-caused visibility impairment in Class I areas through the regional haze
program implemented under sections 169A and 169B of the Act, as discussed above, staff has
focused in this review on visibility impairment primarily in urban areas. In so doing, staff has
considered whether information now available can inform judgments as to the extent to which
existing levels of visibility impairment in urban areas can be considered adverse to public
welfare. In so doing, staff has looked at studies in the U.S. and abroad that have provided the
basis for the establishment of standards and programs to address specific visibility concerns in
local areas, as discussed in section 6.2.5. These studies have produced new methods and tools to
communicate and evaluate public perceptions about varying visual effects associated with
alternative levels of visibility impairment relative to varying particle pollution levels and
environmental conditions. As discussed in section 6.2.6, methods involving the use of surveys to
elicit citizen judgments about the acceptability of varying levels of visual air quality in an urban
area have been developed by the State of Colorado, and they have been used to develop a
visibility standard for Denver. These methods have now been adapted and applied in other
areas, including Phoenix, AZ, and the province of British Columbia, Canada, producing
reasonably consistent results in terms of the visual ranges found to be generally acceptable by
the participants in the various studies, which ranged from approximately 40 to 60 km in visual
range.
Beyond the information available from such programs, staff believes it is appropriate to
make use directly of photographic representations of visibility impairment to help inform
judgments about the acceptability of varying levels of visual air quality in urban areas. As
discussed in section 6.2.6, photographic representations of varying levels of visual air quality
have been developed for several urban areas (Appendix 6A, available on EPA's website at
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html). In considering these images for
Washington, D.C., Chicago, and Phoenix (for which PM2 5 concentrations are reported), staff
makes the following observations:
At concentrations at or near the level of the current 24-hour PM2 5 standard, scenic views
(e.g., mountains, historic monuments), as depicted in these images around and within the
urban areas, are significantly obscured from view.
7-5
-------
• Appreciable improvement in the visual clarity of the scenic views depicted in these
images occurs at PM2 5 concentrations below 35 to 40 |ig/m3, which equates to visual
ranges generally above 20 km for the urban areas considered.
While being mindful of the limitations in using visual representations from a small
number of areas as a basis for considering national visibility-based secondary standards, staff
nonetheless concludes that the observations discussed above support consideration of revising
the current PM2 5 secondary standards to enhance visual air quality, particularly with a focus on
urban areas. Thus, in the sections that follow, staff evaluates information related to indicator,
averaging time, level and form to identify a range of alternative PM standards for consideration
that would protect visual air quality, primarily in urban areas.
7.3.2 Indicators
As discussed in Chapter 2, section 2.8, fine particles contribute to visibility impairment
directly in proportion to their concentration in the ambient air. Hygroscopic components of fine
particles, in particular sulfates and nitrates, contribute disproportionately to visibility impairment
under high humidity conditions, when such components can reach particle diameters up to and
even above 2.5 |im. Particles in the coarse mode generally contribute only marginally to
visibility impairment in urban areas. Thus, fine particles, as indexed by PM2 5, are an appropriate
indicator of PM pollution to consider for the purpose of standards intended to address visibility
impairment.
In analyzing how well PM2 5 concentrations correlate with visibility in urban locations
across the U.S., as discussed above in section 6.2.3 and in more detail in Schmidt et al. (2005),
staff concludes that the observed correlations are strong enough to support the use of PM25 as the
indicator for such standards. More specifically, clear correlations exist between 24-hour average
PM2 5 concentrations and reconstructed light extinction (RE), which is directly related to visual
range, and these correlations are similar in eastern and western regions. These correlations are
less influenced by relative humidity and more consistent across regions when PM25
concentrations are averaged over shorter, daylight time periods (e.g., 4 to 8 hours). Thus, staff
concludes that it is 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
of daylight hours.
7.3.3 Averaging Times
In considering appropriate averaging times for a standard to address visibility
impairment, staff has considered averaging times that range from 24 to 4 hours, as discussed in
section 6.2.3. Within this range, as noted above, correlations between PM25 concentrations and
RE are generally less influenced by relative humidity and more consistent across regions as the
7-6
-------
averaging time gets shorter. Based on the regional and national average statistics considered in
this analysis, staff observes that in the 4-hour time period between 12:00 and 4:00 p.m., the slope
of the correlation between PM2 5 concentrations and hourly RE is lowest and most consistent
across regions than for any other 4-hour or longer time period within a day (Chapter 6, Figure
6-4). Staff also recognizes that these advantages remain in looking at a somewhat wider time
period, from approximately 10:00 am to 6:00 pm. Staff concludes that an averaging time from 4
to 8 hours, generally within the time period from 10:00 am to 6:00 pm, should be considered for
a standard to address visibility impairment.
In reaching this conclusion, staff recognizes that the PM25 Federal Reference Method
(FRM) monitoring network provides 24-hour average concentrations, and, in some cases, on a
third- or sixth-day sample schedule, such that implementing a standard with a less-than-24-hour
averaging time would necessitate the use of continuous monitors that can provide hourly time
resolution. Given that the data used in the analysis discussed above are from commercially
available PM2 5 continuous monitors, such monitors clearly could provide the hourly data that
would be needed for comparison with a potential visibility standard with a less-than-24-hour
averaging time. Decisions as to which PM2 5 continuous monitors are providing data of
sufficient quality to be used in a visibility standard would follow protocols for approval of
federal equivalent methods (FEMs) that can provide data in at least hourly intervals.
Development of the criteria for approval of FEMs to support a visibility standard would be based
upon a data quality objective process that considers uncertainties associated with the
measurement system and the level and form of the standard under consideration.
7.3.4 Alternative PM2 5 Standards to Address Visibility Impairment
In considering alternative short-term (4- to 8-hour) PM2 5 standards that would provide
requisite protection against PM-related impairment of visibility primarily in urban areas, staff
has taken into account the results of public perception and attitude surveys 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. Staff believes that these
sources provide a basis for bounding a range of levels appropriate for consideration in setting a
national visibility standard primarily for urban areas.
As discussed above in section 6.2, public perception and attitude surveys conducted in
Denver, CO and Phoenix, AZ resulted in judgments reflecting the acceptability of a visual range
of approximately 50 and 40 km, respectively. A similar survey approach in the Fraser Valley in
British Columbia, Canada reflected the acceptability of a visual range of 40 to 60 km. Visibility
standards established for the Lake Tahoe area in California and for areas within Vermont are
both targeted at a visual range of approximately 50 km. Staff notes that, in contrast to this
convergence of standards and goals around a visual range from 40 to 60 km, California's long-
standing general state-wide visibility standard is a visual range of approximately 16 km. Staff
7-7
-------
believes that consideration should be given to national visibility standards for urban areas across
the U.S. that are somewhat less stringent than local standards and goals set to protect scenic
resources in and around certain urban areas that are particularly highly valued by people living in
those areas, suggesting an upper end of the range of consideration below 40 km.
Staff has also inspected the photographic representations of varying levels of visual air
quality that have been developed for Washington, D.C., Chicago, Phoenix, and Denver
(Appendix 6A, available on EPA's website at
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html). Staff observes that scenic views
(e.g., historic monuments, lake front and mountain vistas) depicted in these images (around and
within the three urban areas for which PM2 5 concentrations are reported) are significantly
obscured from view at PM25 concentrations of 35 to 40 |ig/m3 in Chicago, Washington, D.C.,
and Phoenix, corresponding to reported visual ranges in Washington, D.C. and Phoenix of 12 to
20 km, respectively. Staff also observes that visual air quality appears to be good in these areas
at PM2 5 concentrations generally below 20 |ig/m3, corresponding to reported visual ranges in
Washington, D.C. and Phoenix above approximately 25 to 35 km, respectively. In looking at the
images in Denver, staff observes that visual air quality appears to be generally good, specifically
in terms of the ability to view nearby mountain ranges, at a visual range above 52 km. These
observations are interpreted by staff as suggesting consideration of a national visibility standard
in the PM2 5 concentration range of 30 to 20 |ig/m3. The upper end of this range is below the
levels at which scenic views are significantly obscured, and the lower end is around the level at
which visual air quality generally appeared to be good in these areas. Staff recognizes that the
above observations about visual air quality in urban areas inherently take into account the nature
and location of scenic views that are notable within and around a given urban area, which has
implications for the appropriate design of an associated monitoring network.
Building upon the analysis discussed above in section 6.2.3, staff has characterized the
distributions of PM25 concentrations, 4-hour averages in the 12:00 to 4:00 pm time frame, by
region, that correspond to various visual range target levels. The results are shown in Figure 7-1,
panels (a) through (c), for visual range levels of 25, 30, and 35 km, respectively. This figure
shows notable consistency across regions in the median concentrations that correspond to the
target visual range level, with what more variation in regional mean values as well as notable
variation within each region. In focusing on the median values, staff observes that 4-hour
average PM25 concentrations of approximately 30, 25, and 20 |ig/m3 correspond to the target
visual range levels of 25, 30, and 35 km, respectively. Thus, a standard set within the range of
30 to 20 |ig/m3 can be expected to correspond generally to median visual range levels of
approximately 25 to 35 km in urban areas across the U.S.. Staff notes, however, that a standard
set at any specific PM2 5 concentration will necessarily result in visual ranges that vary somewhat
in urban areas across the country, reflecting in part the less-than-perfect correlation between
PM25 concentrations and reconstructed light extinction. Staff also notes that the range of PM25
7-8
-------
Visual range = 25km
|
i
=L
--S--T--A--"
PH 20
b
a:
tr
Visual range = 30km
T
Visual range = 35km
- - --
Northeast Southeas Industrial Upper
Midwest Midwest
Southwest Northwest Southern
California
Figure 7-1. Distributions of PM25 concentrations for 12 p.m. - 4 p.m.
corresponding to visual ranges of 25 km (panel a), 30 km (panel b),
and 35 km (panel c) - by region. Box depicts interquartile range and
median; whiskers depict 5th and 95th percentiles; star denotes mean.
Source: Schmidt et al. (2005)
7-9
-------
concentrations from 30 to 20 |ig/m3, suggested by staffs analysis and observations of
photographic representations, is generally consistent with a national target visual range below 40
km, the level suggested by the public perception surveys and the local visibility standards and
goals discussed above.
In considering a standard down to 20 |ig/m3, staff has again looked to information on
PM2 5 background concentrations, as was done in considering primary PM2 5 standard levels in
Chapter 5, section 5.3.5. In both instances, staff recognizes that a standard intended to provide
protection from man-made pollution should be set above background levels. In considering
background levels in conjunction with a primary standard, staff focused on the 99th percentile of
the distribution of estimated background levels, consistent with consideration of a 98th or 99th
percentile form for a primary standard, concluding in that case that 25 |ig/m3 was an appropriate
lower end to the range of 24-hour primary PM25 standards for consideration. For reasons
discussed below, staff believes that a range of percentile forms extending below the 98th
percentile would be appropriate to consider for a visibility standard, and thus staff has also
looked to lower percentiles in the distribution of estimated background levels as a basis for
comparison with the lower end of the range of short-term secondary PM2 5 standards for
consideration. As discussed in Chapter 2, section 2.6, staff notes that, while long-term average
daily PM25 background levels are quite low (ranging from 1 to 5 |ig/m3 across the U.S.), the
estimated 90th percentile values in distributions of daily background levels, for example, are
appreciably higher, but generally well below 15 |ig/m3, with levels below 10 |ig/m3 in most
areas, and these levels may include some undetermined contribution from anthropogenic
emissions (Langstaff, 2005). In addition, staff again notes that even higher daily background
levels result from episodic occurrences of extreme natural events (e.g., wildfires, global dust
storms), but levels related to such events are generally excluded from consideration under EPA's
natural events policy, as noted in section 2.6. Taking these considerations into account, staff
believes that 20 |ig/m3 is an appropriate lower end to the range of short-term PM2 5 standards for
visibility protection for consideration in this review if a percentile form ranging down to the 90th
percentile were to be considered. Alternatively, if a percentile form at about the 98th percentile
were considered, then, consistent with conclusions for the primary standard, staff believes that
25 |ig/m3 is an appropriate lower end to the range.
As in the last review, staff believes that a national visibility standard should be
considered 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. Staff recognizes that programs implemented to meet a national
standard focused primarily on 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 rule established for protection of visual air quality in Class I
areas. Staff further believes that the development of local programs continues to be an effective
7-10
-------
and appropriate approach to provide additional protection for unique scenic resources in and
around certain urban areas that are particularly highly valued by people living in those areas.
Based on these considerations, and taking into account the observations and analysis discussed
above, staff concludes that consideration should be given to a short-term (4- to 8-hour daylight
average) secondary PM25 standard in the range of 30 to 20 |ig/m3, depending in part on the form
of the standard, as discussed below, for protection of visual air quality primarily in urban areas.
7.3.5 Alternative Forms of a Short-term PM2 5 Standard
In considering an appropriate form for a short-term PM2 5 standard for visibility, staff has
taken into account the same general factors that were taken into account in considering an
appropriate form for the primary PM25 standard, as discussed above in Chapter 5, section 5.3.6.
In that case, as in the last review, staff has concluded that a concentration-based form should be
considered because of its advantages over the previously used expected-exceedance form2. One
such advantage is that a concentration-based form is more reflective of the impacts posed by
elevated PM2 5 concentrations because it 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 standard. Staff notes that the same advantage would apply for a visibility standard
as to a health-based standard, in that it would give proportionally greater weight to days when
PM-related visibility impairment is substantially higher than to days just above the standard.
Further, staff recognizes that a concentration-based form has greater stability and, thus,
facilitates the development of more stable implementation programs. Taking these factors into
account, staff concludes that consideration should be given to a percentile-based form for a
visibility standard.
To identify a range of concentration percentiles that would be appropriate for
consideration, staff first has taken into account similar considerations as were discussed in
Chapter 5, section 5.3.6.2, for the primary PM25 standard as a basis for identifying an
appropriate upper end to a range of percentile forms. More specifically, staff believes that the
upper end of the range of consideration should be the 98th to 99th percentiles, consistent with the
forms being considered for the 24-hour primary PM25 standard. Staff has also considered that
the regional haze program targets the 20% most impaired days for improvements in visual air
quality in Class I areas. If a similar target of the 20% most impaired days were 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 days in this range would be
improved by control strategies intended to attain the standard. A focus on improving the 20%
The form of the 1987 24-hour PM10 standard is based on the expected number of days per year (averaged
over 3 years) on which the level of the standard is exceeded; thus, attainment with the one-expected exceedance
form is determined by comparing the fourth-highest concentration in 3 years with the level of the standard.
7-11
-------
most impaired days suggests to staff that the 92nd percentile, which represents the mean of the
distribution of the 20% worst days (Schmidt et al., 2005), would be an appropriate lower end of
the range of forms for consideration.
Based on the factors discussed above, staff concludes that a percentile-based form should
be considered, based on a percentile within a range of the 92nd to about the 98th percentile. Staff
believes that a form selected from within this range could provide an appropriate balance between
adequately limiting the occurrence of peak concentrations and providing for a relatively stable
standard.
To assist in understanding the implications of alternative percentile forms, in conjunction
with alternative levels of a 4-hour secondary PM25 standard, staff assessed the percentage of days
estimated to exceed various PM25 concentrations in counties across the U.S., as shown in
Appendix 7 A, Figure 7A-1. This analysis is based on 2001 to 2003 air quality data from
continuous monitors, using the 4-hour average concentration from 12:00 to 4:00 pm at the
maximum monitor in each county. Staff also assessed (based on the same air quality database)
the percentage of counties, and the population in those counties, that would not likely attain
various PM25 secondary standards (Appendix 7A, Table 7A-1). These assessments are intended
to provide some rough indication of the breadth of coverage potentially afforded by various
combinations of alternative standards.
7.3.6 Summary of Staff Recommendations
Staff recommendations for the Administrator's consideration in making decisions on the
secondary PM2 5 standards to address PM-related visibility impairment, together with supporting
conclusions from sections 7.3.1 through 7.3.4, are briefly summarized below. Staff recognizes
that selecting from among alternative standards will necessarily reflect consideration of the
qualitative and quantitative uncertainties inherent in the relevant information. In making the
following recommendations, staff is mindful that the Act requires secondary standards to be set
that, in the judgment of the Administrator, are requisite to protect public welfare from those
known or anticipated effects that are judged by the Administrator to be adverse, such that the
standards are neither more nor less stringent than necessary. The Act does not require that
secondary standards be set to eliminate all risk of adverse welfare effects.
( 1) Consideration should be given to revising the current suite of secondary PM2 5 standards to
provide increased and more targeted protection primarily in urban areas from visibility
impairment related to fine particles. This recommendation reflects the recognition that
programs implemented to meet such a standard can be expected to improve visual air
quality in non-urban areas as well, just as programs now being developed to address the
requirements of the regional haze rule, for protection of visual air quality in Class I areas,
can also be expected to improve visual air quality in some urban areas.
7-12
-------
( 2) The indicator for a fine particle visibility standard should be PM2 5, reflecting the strong
correlation between short-term average PM25 in urban areas across the U.S. and light
extinction, which is a direct measure of visibility impairment.
( 3) Consideration should be given to a short-term averaging time for a PM2 5 standard, within
the range of 4 to 8 hours, within a daylight time period between approximately 10:00 am
to 6:00 pm. To facilitate implementation of such a standard, consideration should be
given to the adoption of FEMs for appropriate continuous methods for the measurement of
short-term average PM2 5 concentrations.
( 4) Recommendations on ranges of alternative levels and forms for alternative PM2 5 visibility
standards include:
( a) Staff recommends consideration of a 4- to 8-hour PM2 5 standard within the range
of 30 to 20 |ig/m3, depending in part on the form of the standard.
(b) Staff also recommends consideration of a percentile-based form for such a
standard, focusing on a range from the 92nd up to the 98th percentile of the annual
distribution of daily short-term PM25 concentrations, averaged over 3 years.
Staff judges that a standard within these ranges could provide an appropriate degree of
protection against visibility impairment, generally resulting in a visual range of
approximately 25 to 35 km, primarily in urban areas, as well as improved visual air
quality in surrounding non-urban areas.
7.4 STANDARDS TO ADDRESS OTHER PM-RELATED WELFARE EFFECTS
EPA's decision in 1997 to revise the suite of secondary PM standards took into account
not only visibility protection, but also materials damage and soiling, the other PM-related welfare
effect considered in the last review. Based on this broader consideration, EPA established
secondary standards for PM identical to the suite of primary standards, including both PM2 5 and
PM10 standards, to provide appropriate protection against the welfare effects associated with fine
and coarse particle pollution (62 FR at 38683). This decision was based on considering both
visibility effects associated with fine particles, as discussed above in section 7.3, and materials
damage and soiling effects associated with both fine and coarse particles. With regard to effects
on materials, EPA concluded that both fine and coarse particles can contribute to materials
damage and soiling effects. However, EPA also concluded that the available data did not provide
a sufficient basis for establishing a distinct secondary standard based on materials damage or
soiling alone. These considerations led EPA to consider whether the reductions in fine and coarse
particles likely to result from the suite of primary PM standards would provide appropriate
protection against the effects of PM on materials. Taking into account the available information
7-13
-------
and the limitations in that information, EPA judged that setting secondary standards identical to
the suite of PM2 5 and PM10 primary standards would provide increased protection against the
effects of fine particles and retain an appropriate degree of control on coarse particles.
In this review, in addition to addressing visibility impairment, the CD has broadened its
scope to include effects on ecosystems and vegetation, discussed in Chapter 6, section 6.3, and
also addresses PM-related effects on materials, discussed in section 6.4, and the role of ambient
PM in atmospheric processes associated with climate change and the transmission of solar
radiation, discussed in section 6.5. In considering the currently available evidence on each of
these types of PM-related welfare effects, staff notes that there is much information linking
ambient PM to potentially adverse effects on materials and ecosystems and vegetation, and on
characterizing the role of atmospheric particles in climatic and radiative processes. However, on
the basis of the evaluation of the information discussed in Chapter 6, which highlighted the
substantial limitations in the evidence, especially with regard to the lack of evidence linking
various effects to specific levels of ambient PM, staff concludes that the available evidence does
not provide a sufficient basis for establishing distinct secondary standards for PM based on any of
these effects alone. These considerations lead staff to address in the following sections whether
the reductions in PM likely to result from the current secondary PM standards, or from the range
of recommended revisions to the primary PM standards and from the recommended secondary
PM2 5 standard to address visibility impairment, would provide appropriate protection against
these other PM-related welfare effects.
7.4.1 Vegetation and Ecosystems
With regard to PM-related effects on ecosystems and vegetation, staff notes that the CD
presents evidence of such effects, particularly related to nitrate and acidic deposition, and
concludes that current PM levels in the U.S. "have the potential to alter ecosystem structure and
function in ways that may reduce their ability to meet societal needs" (CD, p. 4-153). Currently,
however, fundamental areas of uncertainty preclude establishing predictable relationships
between ambient concentrations of particulate nitrogen and sulfur compounds and associated
ecosystem effects. One source of uncertainty hampering the characterization of such
relationships is the extreme complexity and variability that exist in estimating particle deposition
rates. These rates are affected by numerous factors, including particle size and composition,
associated atmospheric conditions, and the properties of the surfaces being impacted. A related
source of uncertainty is establishing what portion of the total nitrogen and sulfur deposition
occurring at a given site is attributable to ambient PM. Though several national deposition
monitoring networks have been successfully measuring wet and dry deposition for several
decades, they often do not distinguish the form (e.g., particle, wet, and dry gaseous) in which a
given chemical species is deposited. Further, it is not clear how well data from monitoring sites
7-14
-------
may apply to non-monitored sites with different surface cover, meteorology, or other deposition-
related factors.
In addition, ecosystems have different sensitivities and capacities to buffer or assimilate
pollutants. Many of the documented ecosystem-level effects only became evident after long-
term, chronic exposures to total annual loads of Nr or acidifying compounds that eventually
exceeded the natural buffering or assimilative capacity of the system. In most cases, PM
deposition is not the only contributor to the total load of Nr or acidifying compounds entering the
affected system. Since it is difficult to predict the rate of PM deposition, and thus, the PM
contribution to total deposition at a given site, it is difficult to predict the ambient concentration
of PM that would likely lead to the observed adverse effects within any particular ecosystem.
Equally difficult is the prediction of recovery rates for areas already affected, if PM deposition
rates of various chemical species were to be reduced.
Despite these uncertainties, a number of significant environmental effects that either have
already occurred or are currently occurring have been linked to chronic deposition of chemical
constituents found in ambient PM. Staff notes, for example, that the following effects have been
linked with chronic additions of Nr and its accumulation in ecosystems:
• Productivity increases in forests and grasslands, followed by decreases in
productivity and possible decreases in biodiversity in many natural habitats
wherever atmospheric reactive nitrogen deposition increases significantly and
critical thresholds are exceeded;
• Acidification and loss of biodiversity in lakes and streams in many regions,
especially in conjunction with sulfate deposition; and
• Eutrophication, hypoxia, loss of biodiversity, and habitat degradation in coastal
ecosystems.
Staff notes that effects of acidic deposition have been extensively documented, as
discussed in the CD and other reports referenced therein. For example, effects on some species of
forest trees linked to acidic deposition include increased permeability of leaf surfaces to toxic
materials, water, and disease agents; increased leaching of nutrients from foliage; and altered
reproductive processes; all of which serve to weaken trees so that they are more susceptible to
other stresses (e.g., extreme weather, pests, pathogens). In particular, acidic deposition has been
implicated as a causal factor in the northeastern high-elevation decline of red spruce. Although
U.S. forest ecosystems other than the high-elevation spruce-fir forests are not currently
manifesting symptoms of injury directly attributable to acid deposition, less sensitive forests
throughout the U.S. are experiencing gradual losses of base cation nutrients, which in many cases
will reduce the quality of forest nutrition over the long term.
Taking into account the available evidence linking chemical constituents of both fine and
coarse PM to these types of known and potential adverse effects on ecosystems and vegetation,
7-15
-------
staff believes that further reductions in ambient PM would likely contribute to long-term recovery
and to the prevention of further degradation of sensitive ecosystems and vegetation. Staff
recognizes, however, that the available evidence does not provide any quantitative basis for
establishing distinct national standards for ambient PM to address these effects. Further, staff
recognizes that due to site-specific sensitivities to various components of ambient PM, differing
buffering and assimilative capacities, and local and regional differences in the percentage of total
deposition of Nr and acidifying compounds that is likely attributable to ambient PM, national
ambient air quality standards alone may not be an appropriate approach to protect against the
adverse impacts of total Nr and acidifying compounds, partially contributed by ambient PM, on
ecosystems and vegetation in all parts of the country. Nonetheless, staff believes that additional
reductions in fine particles and related precursor emissions likely to result from the current suite
of secondary PM standards, or the range of recommended revisions to the primary PM standards
and to the secondary PM2 5 standard to address visibility impairment, would contribute to
increased protection against PM-related effects on ecosystems and vegetation. Staff recommends
that the potential for increased protection of ecosystems and vegetation be taken into account in
considering whether to revise the current secondary PM standards. Further, staff believes that
any such increased protection should be considered in conjunction with protection afforded by
other programs intended to address various aspects of air pollution effects on ecosystems and
vegetation, such as the Acid Deposition Program and other regional approaches to reducing
pollutants linked to nitrate or acidic deposition.
7.4.2 Materials Damage and Soiling
With regard to PM-related effects on materials, staff notes that the available evidence
continues to support the following observations:
• Materials damage and soiling that occur through natural weathering processes are
enhanced by exposure to atmospheric pollutants, most notably SO2 and particulate
sulfates.
While ambient particles play a role in the corrosion of metals and in the
weathering of paints and building materials, no quantitative relationships between
ambient particle concentrations and rates of damage have been established.
• Similarly, while soiling associated with fine and coarse particles can result in
increased cleaning frequency and repainting of surfaces, no quantitative
relationships between particle characteristics (e.g., concentrations, particle size,
and chemical composition) and the frequency of cleaning or repainting have been
established.
Staff believes that these observations and the underlying available evidence continue to support
consideration of retaining an appropriate degree of control on both fine and coarse particles.
7-16
-------
Lacking any specific quantitative basis for establishing distinct standards to protect against PM-
related adverse effects on materials, staff believes that reductions in fine and coarse particles
likely to result from the current suite of secondary PM standards, or the range of recommended
revisions to the primary PM standards and to the secondary PM2 5 standard to address visibility
impairment, would contribute to protection against PM-related soiling and materials damage.
Staff recommends that the potential for such protection be taken into account in considering
whether to revise the current secondary PM standards.
7.4.3 Climate Change and Solar Radiation
With regard to the role of ambient PM in climate change processes and in altering the
penetration of solar UV-B radiation to the earth's surface, staff notes that information available in
this review derives primarily from broad-scale research and assessments related to the study of
global climate change and stratospheric ozone depletion. As such, this information is generally
focused on global- and regional-scale processes and impacts and provides essentially no basis for
characterizing how differing levels of ambient PM in areas across the U.S. would affect local,
regional, or global climatic changes or alter the penetration of UV-B radiation to the earth's
surface. As noted in section 6.5, even the direction of such effects on a local scale remains
uncertain. Moreover, similar concentrations of different particle components can produce
opposite net radiative effects. Thus, staff concludes that there is insufficient information
available to help inform consideration of whether any revisions of the current secondary PM
standards are appropriate at this time based on ambient PM's role in atmospheric processes
related to climate or the transmission of solar radiation.
7.4.4 Summary of Staff Recommendations
Taking into account the conclusions presented in sections 7.4.1 through 7.4.3 above, staff
makes the following recommendations with regard to PM-related effects on vegetation and
ecosystems and materials damage and soiling:
(1) Consideration should be given to setting secondary PM standards that at a minimum retain
the level of protection afforded by the current PM standards, so as to continue control of
ambient fine and coarse-fraction particles, especially long-term deposition of particles
such as particulate nitrates and sulfates, that contribute to adverse impacts on vegetation
and ecosystems and/or to materials damage and soiling. Any such standards should be
considered in conjunction with the protection afforded by other programs intended to
address various aspects of air pollution effects on ecosystems and vegetation, such as the
Acid Deposition Program and other regional approaches to reducing pollutants linked to
nitrate or acidic deposition.
7-17
-------
(2) While staff recognizes that PM-related impacts on vegetation and ecosystems and PM-
related soiling and materials damage are associated with chemical components in both fine
and coarse-fraction PM, staff also recognizes that sufficient information is not available at
this time to recommend consideration of either an ecologically based indicator or an
indicator based distinctly on soiling and materials damage, in terms of specific chemical
components of PM. Thus, for consistency with the primary standards, staff recommends
that consideration be given to basing secondary standards on the same indicators that are
used as the basis for the suite of primary PM standards.
In making these recommendations, staff has taken into account both the available
evidence linking fine and coarse particles with effects on vegetation and ecosystems and material
damage and soiling, as well as the limitations in the available evidence. In so doing, staff
recognizes that the available information does not provide a sufficient basis for the development
of distinct national secondary standards to protect against such effects beyond the protection
likely to be afforded by the suite of primary PM standards.
7.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH RECOMMENDATIONS
RELATED TO SETTING SECONDARY PM STANDARDS
Staff believes it is important to continue to highlight the unusually large uncertainties
associated with establishing standards for PM relative to other single component pollutants for
which NAAQS have been set. Key uncertainties and staff research recommendations welfare-
related topics are outlined below. In some cases, research in these areas can go beyond aiding in
standard setting to aiding in the development of more efficient and effective control strategies.
Staff notes, however, that a full set of research recommendations to meet standards
implementation and strategy development needs is beyond the scope of this discussion.
With regard to welfare-related effects, discussed in Chapter 4 of the CD and Chapter 6
herein, staff has identified the following key uncertainties and research questions that have been
highlighted in this review of the welfare-based secondary standards:
(1) Refinement and broader application of survey methods designed to elicit citizens'
judgments about the acceptability of varying levels of local visibility impairment could
help inform future reviews of a visibility-based secondary standard. Such research could
appropriately build upon the methodology developed by the State of Colorado and used as
a basis for setting a visibility standard for the city of Denver, which has been adapted and
applied in other areas in the U.S. and abroad.
(2) There remain significant uncertainties associated with the characterization and prediction
of particle deposition rates to natural surfaces, especially with respect to nitrogen and
7-18
-------
sulfur deposition. Reduction in these uncertainties, particularly in key acid- and Nr-
sensitive areas, will be important in developing the capability of quantitatively linking
ambient PM concentrations with environmental exposures. In order to better understand
the contribution of PM to cumulative long-term environmental impacts, more research
needs to be conducted on the percentage of total Nr and acidifying deposition contributed
by PM and where necessary, better monitoring methods and network designs should be
developed.
(3) Atmospheric sources of Nr and acidifying compounds, including ambient PM, are clearly
contributing to the total load of these pollutants entering U.S. ecosystems annually.
However, the immense variability in ecosystem response to total Nr and acidifying
deposition across the U.S., and the factors controlling ecosystem sensitivity and recovery,
have not been adequately characterized. Data should be collected on a greater variety of
ecosystems over longer time scales to determine how ecosystems respond to different
loading rates over time. Such research, in conjunction with the development of improved
predictive models, could help in future consideration within the U.S. of the "critical loads"
concept, and in determining how much of any given critical load is contributed by
different sources of pollutants.3
This recommendation is consistent with the views of the National Research Council (NRC) contained in
its recent review of air quality management in the U.S. (NRC, 2004). This report recognizes that for some resources
at risk from air pollutants, including soils, groundwaters, surface waters, and coastal ecosystems, a deposition-based
standard could be appropriate, and identifies "critical loads"as one potential approach for establishing such a
deposition-based standard.
7-19
-------
REFERENCES
Langstaff, John E. (2004). Estimation of Policy-Relevant Background Concentrations of Paniculate Matter.
Memorandum to PM NAAQS review docket OAR-2001-0017. January 27, 2005.
National Research Council (NRC) (2004). Air Quality Management in the United States. Committee on Air Quality
Management in the U.S., National Research Council of the National Academy of Science. The National
Academies Press, Washington, D.C. ISBN 0-309-08932-8.
Schmidt et al. (2005) Draft analysis of PM ambient air quality data for the PM NAAQS review. Memorandum to PM
NAAQS review docket OAR-2001-0017. January 31, 2005.
7-20
-------
Appendix 3A. Mortality and Morbidity Effect Estimates and PM Concentrations from U.S. and Canadian Studies
for Short-term Exposures to PM10, PM2 5, and PM10_2 5
Original study*
Study Location
Reanalysis study
Analysis % increase (95% CI)
Comments per
50 ug/m3 PM10
% increase (95% CI)
per
25 ug/m3 PM2.5
% increase (95% CI) PM10, PM2 5 and PM10.2 5
per Mean (Range) Levels
25 ug/m3 PM10_2.5 Reported"
MORTALITY:
Total (nonaccidental) Mortality
/to and Thurston, 1996
Chicago, IL
Kinneyetal, 1995
Los Angeles, CA
Pope et al, 1992
Utah Valley, UT
Schwartz, 1993
Birmingham, AL
Schwartz et al., 1996
Boston, MA
Schwartz, 2003a
Schwartz etal., 1996
Knoxville, TN
Schwartz, 2003a
Schwartz etal., 1996
St. Louis, MO
Schwartz, 2003a
GAM not used 2.47 (1. 26, 3. 69)
GAM not used 2.47 (-0. 1 7, 5. 18)
GAM not used 7. 63 (4. 41, 10. 95)
GAM not used 5.36(1.16,9. 73)
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
—
—
—
—
5.3(3.5, 7.1)
5.7 (3.7, 7.6)
5.0(3.1, 7.0)
4.5 (2.5, 6.5)
3.1 (0.0, 6.2)
3.0 (-0.3, 6.6)
2.8 (-0.5, 6.3)
2. 6 (-0.8, 6.1)
2.6(0.9,4.3)
2.4(0.6, 4.1)
2.6(0.9, 4.4)
2.3 (0.6, 4.1)
PM10 38 (max 128)
PM10 58(15, 177)
PM10 47 (11,297)
PM10 48 (21, 80)
( PM1024.5 (SD12.8)
PM2515.7 (SD9.2)
PM10.258.8(SD7.0)
0.7 (-1.9, 3.4)
PM10 32.0 (SD 14.5)
PM25 20.8 (SD 9.6)
PM10_25 11.2 (SD 7.4)
1.7 (-2.7, 6.3)
PM10 30.6 (SD 16.2)
PM2 5 18.7 (SD 10.5)
PM10_2511.9(SD8.5)
0.3 (-2.1, 2.7)
3A-1
-------
Original study*
Study Location
Reanalysis study
Schwartz etal, 1996
Steubenville, OH
Schwartz, 2003a
Schwartz etal., 1996
Portage, WI
Schwartz, 2003a
Schwartz etal., 1996
Topeka, KS
Schwartz, 2003a
Schwartz et al., 1996
6 Cities, Overall
Schwartz, 2003a
Styeretal., 1995
Chicago, IL
Samet et al., 2000a,b
90 Largest U.S. Cities
Dominici et al. (2003)
Schwartz, 2000a
10 U.S. cities
Schwartz, 2003b
Burnett et al., 2000
8 Canadian Cities
Burnett and Goldberg, 2003
Chock et al., 2000
Pittsburgh, PA
Analysis
Comments
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM Strict
GLMNS
GLMBS
GLMPS
GAM not used
GAM strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLM NS (6
knots/yr)
GAM not used
% increase (95% CI) % increase (95% CI) % increase (95% CI)
per per per
50 ug/m3 PM10 25 ug/m3 PM25 25 ug/m3 PM10.25
2.4 (-0.4, 5.3)
1.7 (-1.3 4.8)
1.5 (-1.5, 4.6)
1.8 (-1.2, 4.9) 5.2(0.0,10.7)
2.6 (-1.2, 6.6)
0.8 (-3. 3, 5.1)
1.5 (-2.7, 5.8)
1.1 (-3.1,5.4) 0.7 (-4.0, 5.6)
1.6 (-5.3, 9.0)
2.7 (-5.0, 10.9)
1.3 (-6.2, 9.3)
1.4 (-6.3, 9.6) -3.0 (-8.1, 2.3)
3.5 (2.5, 4.5)
3.3 (2.2, 4.3)
3.0(2.0, 4.0)
2.9(1.8,4.0)
4.08(0.08,8.24)
1.4 (0.9, 1.9)
1.1 (0.5, 1.7)
3.4(2.6,4.1)
2.8 (2.0, 3.6)
3.2(1.1,5.5) 2.8(1.2,4.4) 1.9 (-0.1, 3.9)
2.7 (-0.1, 5.5) 2.1(0.1,4.2) 1.8 (-0.6, 4.4)
<75 years 2.6 (-2.0, 7.7) <75 years 0.7 (-1.7, 3.)
>75 years 1.5 (-3.0, 6.3) >75 years 1.3 (-1.3, 3.8)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM10 45.6 (SD 32.3)
PM2529.6(SD21.9)
PM10.2516.1 (SD 13.0)
PM10 17.8 (SD 11.7)
PM2511.2(SD7.8)
PMW_256.6(SD6.8)
PM 10 26.7 (SD 16.1)
PM2512.2(SD 7.4)
PM10_25 14.5 (SD 12.2)
PM10 means 17.8-45.6
PM25 means 11.2-29.6
PM10_2 5 means 6. 6-16.1
PM1037 (4, 365)
PM10 mean range
15.3-52.0
PM10 mean range
27.1-40.6
PM10 25.9 (max 121)
PM25 13.3 (max 86)
PM10.25 12.9 (max 99)
PM2 5 20.5 (3. 0,86.0)
PM1(M521.6(0, 208.0)
3A-2
-------
Original study*
Study Location
Reanalysis study
Clyde et al., 2000
Phoenix, AZ
Fairley, 1999
Santa Clara County, CA
Fairley, 2003
Gamble, 1998
Dallas, TX
Goldberg etal., 2000
Montreal, CAN
Goldberg and Burnett, 2003
Klemm and Mason, 2000
Atlanta, GA
Klemm et al., 2000
Six City reanalysis - St. Louis
Klemm and Mason, 2003
Klemm et al., 2000
Six City reanalysis -
Steubenville
Klemm and Mason, 2003
Klemm et al., 2000
Six City reanalysis - Topeka
Klemm and Mason, 2003
Klemm et al., 2000
Six City reanalysis - Knoxville
Klemm and Mason, 2003
Klemm et al., 2000
Six City reanalysis - Boston
Klemm and Mason, 2003
Analysis
Comments
GAM not used
GAM Strict
GLMNS
GAM not used
GAM Strict
GLMNS
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
% increase (95% CI)
per
50 ug/m3 PM10
6(>0, 11)
7.8(2.8, 13.1)
8.3 (2.9, 13.9)
-3.56 (-12.73, 6.58)
"
8.7 (-5.2, 24.7)
2.0(0.0,4.1)
1.0 (-1.5, 3.6)
2.5 (-1.7, 7.0)
1.5 (-1.7, 4.9)
-3. 5 (-11.6, 5.4)
-5.4 (-14.3, 4.4)
6.1(1.5, 11.0)
5.1 (-0.2, 10.7)
6.1(3.6,8.8)
5.6 (2.8, 8.5)
% increase (95% CI)
per
25 ug/m3 PM2 5
—
8.1(1.6, 15.0)
7.0(1.4, 13.0)
—
4.2 (p<0.05)
1.5 (p>0.05)
4.8 (-3.2, 13.4)
2.0 (0.5, 3.5)
1.3 (-0.5, 3.0)
1.5 (-1.6, 4.7)
0.5 (-2.7, 3.8)
1.5 (-6.5, 10.2)
-0.5 (-9.5, 9.4)
4.3 (0.9, 7.8)
3.8 (-0.1, 7.8)
5.1(3.3,6.9)
4.0(1.9,6.2)
% increase (95% CI)
per
25 ug/m3 PM10.25
—
4.5 (-7.6,
3.3 (-5.3,
—
"
1.4 (-11.3,
0.0 (-2.2,
-0.5 (-3.0
4.6 (-0.7,
4.0 (-1.6,
-3.7 (-9.2
-4.7 (-10.*
3.5 (-1.0,
3.0 (-1.9,
1.3 (-1.1,
1.8 (-1.0,
18.1)
12.6)
15.9)
2.3)
,2.0)
10.1)
10.0)
,2.1)
5, 1.8)
8.2)
8.2)
3.7)
4.6)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM10 mean 45. 4
PM10 34 (6, 165)
PM25 13(2, 105)
PM10.2511(0,45)
PM1024.5(11, 86)
PM25 17.6 (4.6, 71.7)
PM25 19.9 (1.0, 54.8)
PM10.25 10. 1(0.2, 39.5)
PM10 30.6 (SD 16.2)
PM2518.7(SD 10.5)
PM10.2511.9(SD8.5)
PM10 45.6 (SD 32.3)
PM2529.6(SD21.9)
PM10.25 16.1 (SD 13.0)
PM1026.7(SD 16.1)
PM2512.2(SD7.4)
PM10.25 14.5 (SD 12.2)
PM10 32.0 (SD 14.5)
PM2520.8(SD9.6)
PM10.2511.2(SD7.4)
PM10 24.5 (SD 12.8)
PM2515.7(SD9.2)
PM10.258.8(SD7.0)
3A-3
-------
Original study*
Study Location
Reanalysis study
Klemm et al., 2000
Six City reanalysis - Madison
Klemm and Mason, 2003
Klemm et al., 2000
Six City reanalysis - overall
Klemm and Mason, 2003
Laden etal, 2000
Six City reanalysis
Schwartz, 2003a
Levy etal., 1998
King Co., WA
Lipfert et al., 2000
Philadelphia, PA
Lippmann et al., 2000
Detroit, MI
Ito, 2003
Moolgavkar, 2000a
Los Angeles, CA
Moolgavkar, 2003
Moolgavkar, 2000a
Cook Co., IL
Moolgavkar, 2003
Ostro, 1995
San Bernadino and Riverside
Counties, CA
Analysis
Comments
GAM Strict
GLMNS
GAM Strict
GLMNS
GLMPS
GAM not used
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
1.0 (-4.6, 7.0)
-1.5 (-7.7, 5.1)
3.5(2.0,5.1)
2.5 (0.8, 4.3)
7.2 (-6.3, 22.8)
5.99 (p>0.055)
3.3 (-2.0, 8.9)
3.1 (-2.2, 8.7)
2.4 (0.5, 4.2)
2.3(0.5,4.1)
2.4(1.4,3.5)
2.6(1.6,3.6)
"
% increase (95% CI)
per
25 ug/m3 PM2 5
1.5 (-2.7, 5.9)
-1.2 (-5.7, 3.5)
3.0(2.0,4.1)
2.0 (0.9, 3.2)
-5.1 (-13.9, 4.6) crustal
9.3 (4.0, 14.9) traffic
2.0 (-0.3, 4.4) coal
1.76 (-3.53, 7.34)
4.21 (p<0.055)
1.9 (-1.8, 5.7)
2.0 (-1.7, 5.8)
1.5 (0, 3.0)
1.4 (-0.4, 3.2)
0 (-1.4, 1.4)
% increase (95% CI)
per
25 ug/m3 PM10.25
0.0 (-4.8, 5.0)
-1.0 (-6.2, 4.5)
0.8 (-0.6, 2.1)
0.5(-1.0, 2.0)
—
5.07 (p>0.055)
3.2 (-1.9, 8.6)
2.8 (-2.2, 8.1)
"
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM10 17.8 (SD 11.7)
PM25 11.2(SD7.8)
PM10.25 6.6 (SD 6.8)
PM10 means 17.8-45.6
PM25means 11.2-29.6
PM10_25 means 6.6-16.1
PM2 5 same as
Schwartz et al., 1996
PM10 29.8 (6.0, 123.0)
PMj 28.7 (16.3, 92.2)
PM10 32.20 (7.0, 95.0)
PM2 5 17.28 (-0.6, 72.6)
PM10.25 6.80 (-20.0, 28.3)
PM1031(12, 105)
PM2518(6, 86)
PM10.2513(4,50)
mean (5%, 95%)
PM10 median 44 (7, 166)
PM2 5 22 (4, 86)
PM10 median 35 (3, 365)
PM2532.5(9.3, 190.1)
(estimated from
visibility)
3A-4
-------
Original study*
Study Location
Reanalysis study
Schwartz, 2000c
Boston, MA
Schwartz, 2003a
Schwartz, 2000
Chicago, IL
Schwartz, 2003b
Schwartz, 2000
Pittsburgh, PA
Schwartz, 2003b
Schwartz, 2000
Detroit, MN
Schwartz, 2003b
Schwartz, 2000
Seattle, WA
Schwartz, 2003b
Schwartz, 2000
Minneapolis, MN
Schwartz, 2003b
Schwartz, 2000
Birmingham, AL
Schwartz, 2003b
Schwartz, 2000
New Haven, CT
Schwartz, 2003b
Schwartz, 2000
Canton, OH
Schwartz, 2003b
Analysis % increase (95% CI) % increase (95% CI) % increase (95% CI) PM10, PM2 5 and PM10.2 5
Comments per per per Mean (Range) Levels
50 ug/m3 PM10 25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
GLMNS — 5.8 (4.5, 73) (15-day)
9.7(8.2, 11.2) (60-day)
Strict GAM 5.41(2.36,8.56)
(dist. lag)
StrictGAM 3.14(0.25,6.11)
(dist. lag)
StrictGAM 6.83(3.73,10.02)
(dist. lag)
StrictGAM 7.46(3.94,11.10)
(dist. lag)
StrictGAM 10.25(4.67,16.12)
(dist. lag)
StrictGAM 1.71 (-3.44, 7.13)
(dist. lag)
StrictGAM 9.17(1.04,17.96)
(dist. lag)
StrictGAM 8.79 (-4.69, 24.18)
(dist. lag)
PM25 15.6 (±9.2)
PM10 mean 36.5
PM10 mean 36.4
PM10 mean 36.9
PM10 mean 32.5
PM10 mean 27.5
PM10 mean 34.8
PM10 mean 28.6
PM10mean29.31
3A-5
-------
Original study*
Study Location
Reanalysis study
Schwartz, 2000
Spokane, WA
Schwartz, 2003b
Schwartz, 2000
Colorado Springs, CO
Schwartz, 2003b
Tsai et al., 2000
Newark, NJ
Tsai et al., 2000
Camden, NJ
Tsai et al., 2000
Elizabeth, NJ
Analysis
Comments
Strict GAM
(dist. lag)
Strict GAM
(dist. lag)
GAM not used
GAM not used
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
5.62 (-0.31, 11.91)
8.58 (-3.94, 22.73)
5.65 (4.62, 6.70)
11.07(0.70,22.51)
-4.88 (-17.88, 10.19)
% increase (95% CI) % increase (95% CI) PM10, PM2 5 and PM10.2 5
per per Mean (Range) Levels
25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
PM10 mean 40.6
PM10 mean 27.1
4.34 (2.82, 5.89) — PM15 55 (SD 6.5)
PM2 5 42.1(SD22.0)
5.65(0.11,11.51) — PM15 47.0 (SD 20.9)
PM25 39.9 (SD 18.0)
1.77 (-5.44, 9.53) — PM15 47.5 (SD 18.8)
PM2,37.1(SD 19.8)
Cause-Specific Mortality
Cardiorespiratory Mortality:
Samet et al., 2000a,b
90 Largest U.S. Cities
Dominici et al. (2002)
Tsai et al., 2000
Newark, NJ
Tsai et al., 2000
Camden, NJ
Tsai et al., 2000
Elizabeth, NJ
GLMNS
GAM not used
GAM not used
GAM not used
1.6 (0.8, 2.4)
7.79(3.65, 12.10)
15.03 (4.29, 26.87)
3.05 (-11.04, 19.36)
PM10 mean range
15.3-52.0
5.13(3.09,7.21) — PM15 55 (SD 6.5)
PM2 5 42. 1(80 22.0)
6.18(0.61,12.06) — PM15 47.0 (SD 20.9)
PM25 39.9 (SD 18.0)
2.28 (-4.97, 10.07) — PM1547.5 (SD 18.8)
PM2537.1(SD 19.8)
3A-6
-------
Original study*
Study Location
Reanalysis study
Total Cardiovascular Mortality
/to and Thurston, 1996
Chicago, IL
Pope et al, 1992
Utah Valley, UT
Fairley, 1999
Santa Clara County, CA
Fairley, 2003
Goldberg etal, 2000
Montreal, CAN
Goldberg and Burnett, 2003
Lipfert et al., 2000
Philadelphia, PA (7 -county
area)
Lippmann et al., 2000
Detroit, MI
Ito, 2003
Mar et al., 2000
Phoenix, AZ
Mar et al., 2003
Moolgavkar, 2000a
Los Angeles, CA
Moolgavkar, 2003
Moolgavkar, 2000a
Cook Co., IL
Moolgavkar, 2003
Analysis
Comments
GAM not used
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
% increase (95% CI)
per
50 ug/m3 PM10
1.49 (-0.72, 3.74)
9.36(1.91, 17.36)
8.5 (0.6, 17.0)
8.9(1.3, 17.0)
—
8.0(3.7, 12.3)
5.4 (-2.6, 14.0)
4.9 (-3.0, 13.5)
9.7(1.7, 18.3)
9.5 (0.6, 19.3)
4.5(1.6,7.5)
3.9 (0.6, 7.4)
2.2(0.3,4.1)
1.2 (-0.8, 3.1)
% increase (95% CI)
per
25 ug/m3 PM2 5
—
—
6.3 (-4.1. 17.9)
6.7 (-2.5, 16.7)
3.48 (-0.16, 7.26)
5.0 (2.4, 7.5)
2.2 (-3.2, 7.9)
2.0 (-3.4, 7.7)
18.0 (4.9, 32.6)
19.1(3.9,36.4)
2.6 (0.4, 4.9)
1.7 (-0.8, 4.3)
"
% increase (95% CI) PM10, PM2 5 and PM10.2 5
per Mean (Range) Levels
25 ug/m3 PM10.25 Reported"
PM10 38 (max 128)
PM10 47 (11,297)
5.0 (-13.3, 27.3) PM10 34 (6, 165)
PM25 13(2, 105)
PM10.2511(0,45)
PM25 17.6 (4.6, 71.7)
5.4 (-0.4, 11.2) PM10 32.20 (7.0, 95.0)
PM2 5 17.28 (-0.6, 72.6)
PM10.25 6.80 (-20.0, 28.3)
6.7 (-1.0, 15.0) PM10 31 (12, 105)
6.0 (-1.6, 14.3) PM2518(6, 86)
PM10.2513(4,50)
mean (10%, 90%)
6.4(1.3,11.7) PM1046.5(5, 213)
6.2 (0.8, 12.0) PM25 13.0 (0, 42)
PM10.25 33.5 (5, 187)
PM10 median 44 (7, 166)
PM2 5 median 22 (4, 86)
PM10 median 35 (3, 365)
3A-7
-------
Original study*
Study Location
Reanalysis study
Ostro et al., 2000
Coachella Valley, CA
Ostro et al., 2003
Ostro, 1995
San Bernadino and Riverside
Counties, CA
Total Respiratory Mortality:
/to and Thurston, 1996
Chicago, IL
Pope et al., 1992
Utah Valley, UT
Fairley, 1999
Santa Clara County, CA
Fairley, 2003
Lippmann et al., 2000
Detroit, MI
Ito, 2003
Ostro, 1995
San Bernadino and Riverside
Counties, CA
COPD Mortality:
Moolgavkar, 2000a
Cook Co., IL
Moolgavkar, 2003
Moolgavkar, 2000a
Los Angeles, CA
Moolgavkar, 2003
Analysis
Comments
GAM Strict
GLMNS
GAM not used
GAM not used
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
% increase (95% CI) % increase (95% CI)
per per
50 ug/m3 PM10 25 ug/m3 PM25
5.5 (1.6, 9.5) 9.8 (-5.7, 27.9)
5.1(1.2,9.1) 10.2 (-5.3, 28.3)
0.69 (-0.34, 1.74)
6.77(1.97,11.79)
19.78(3.51,38.61)
10.7 (-3.7, 27.2) 11.7 (-9.8, 38.3)
10.8 (-3.4, 27.1) 13.5 (-3.6, 33.7)
7.5 (-10.5, 29.2) 2.3 (-10.4, 16.7)
7.9 (-10.2, 29.7) 3.1 (-9.7, 17.7)
2.08 (-0.35, 4.51)
5.5(0.2,11.0)
4.5 (-1.6, 11.0)
4.4 (-3.2, 12.6) 1.0 (-5. 1,7.4)
6.2 (-3.4, 16.7) 0.5 (-6.8, 8.4)
% increase (95% CI) PM10, PM2 5 and PM10.2 5
per Mean (Range) Levels
25 ug/m3 PM10.25 Reported"
2.9 (0.7, 5.2) PM10 47.4 (3, 417)
2.7(0.4,5.1) PM2 5 16.8(5, 48)
PM10.25 17.9 (0, 149)
PM2 5 32.5 (9.3, 190.1)
(estimated from
visibility)
PM10 38 (max 128)
PM10 47 (11,297)
32.1 (-9.1, 92.2) PM10 34 (6, 165)
PM2513(2, 105)
PM10.2511(0,45)
7.0 (-9.5, 26.5) PM10 31(12, 105)
6.4 (-10.0, 25.7) PM2518(6, 86)
PM10.2513(4,50)
mean (10%, 90%)
PM2 5 32.5 (9.3, 190.1)
(estimated from
visibility)
PM10 median 35 (3, 365)
PM10 median 44 (7, 166)
PM2 5 22 (4, 86)
3A-8
-------
Original study* Analysis
Study Location Comments
Reanalysis study
% increase (95% CI)
per
50 ug/m3 PM10
% increase (95% CI)
per
25 ug/m3 PM2 5
% increase (95% CI) PM10, PM2 5 and PM10.2 5
per Mean (Range) Levels
25 ug/m3 PM10.25 Reported"
CARDIOVASCULAR MORBIDITY
Total Cardiovascular Hospital Admissions:
Samet et al., 2000 strict GAM
14 U.S. Cities (>65 years) strict GAM
Zanobetti and Schwartz (2003b) (dist lag)
GLMNS
GLMPS
Linn et al., 2000 GAM not used
Los Angeles, CA (>29 years)
Moolgavkar, 2000b strict GAMloodf
Cook Co., IL (allages) GLMNSloodf
Moolgavkar, 2003
Moolgavkar, 2000b GAM30df
Los Angeles, CA (all ages) GAMloodf
Moolgavkar, 2003 GLM NSloodf
Stieb et al., 2000 GAM not used
St. John, CAN (all ages)
Burnett et al., 1997 GAM not used
Toronto, CAN (all ages)
Ischemic Heart Disease Hospital Admissions:
Schwartz and Morris, 1995 GAM not used
Detroit (> 65 years)
Lippmann et al., 2000 Strict GAM
Detroit, MI (>65 years) GLM NS
4.95 (3.95 ,5.95))
5.73 (4.27, 7.20)
4.8(3.55,6.0)
5.0 (4.0, 5.95)
3.25 (2.04, 4.47)
4.05 (2.9, 5.2)
4.25 (3.0, 5.5)
3.35(1.2,5.5)
2.7 (0.6, 4.8)
2.75(0.1,5.4)
39.2 (5.0, 84.4)
12.07(1.43,23.81)
5.0(1.9,8.3)
8.0 (-0.3, 17.1)
6.2 (-2.0, 15.0)
—
3.95 (2.2, 5.7)
2.9(1.2,4.6)
3.15(1.1,5.2)
15.11(0.25,32.8)
7.18 (-0.61, 15.6)
—
3.65 (-2.05, 9.7)
3.0 (-2.7, 9.0)
PM10 means 24.4-45.3
PM10 45.5 (5, 132)
PM10 median 35 (3, 365)
PM10 median 44 (7, 166)
PM2 5 median 22 (4, 86)
summer 93
PM10 14.0 (max 70.3)
PM2 5 8.5 (max 53. 2)
20.46 (8.24, 34.06) PM10 28.4 (4, 102)
PM2 5 16.8 (1,66)
PM10.25 11.6 (1,56)
PM10 48 (22, 82)
mean (10%, 90%)
10.2 (2.4, 18.6) PM10 31 (max 105)
8.1(0.4,16.4) PM2518(6, 86)
Ito 2003
PM10.25 13 (4, 50)
3A-9
-------
Original study*
Study Location
Reanalysis study
Analysis
Comments
Dysrhythmias Hospital Admissions:
Lippmann et al., 2000 Strict GAM
Detroit, MI (>65 years) GLM NS
% increase (95% CI)
per
50 ug/m3 PM10
2.8 (-10.9-18.7)
2.0 (-11.7-17.7)
% increase (95% CI)
per
25 ug/m3 PM2 5
3.2 (-6.6-14.0)
2.6 (-7.1-13.3)
% increase (95% CI)
per
25 ug/m3 PM10.25
0.1 (-12.4-14.4)
0.0 (-12.5-14.3)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM1031(maxl05)
PM25 18 (6, 86)
Ito (2003)
Heart Failure/Congestive Heart Disease Hospital
Schwartz and Morris, 1995 GAM not used
Detroit (> 65 years)
Linn et al., 2000
Los Angeles, CA (>29 years)
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito, 2003
Morris and Naumova, 1998
Chicago, IL (>65 years)
Admissions:
2.8(0.7,5.0)
GAM not used 2.02 (-0.94, 5.06)
Strict GAM
GLMNS
GAM not used
Myocardial Infarction Hospital Admissions:
Linn et al., 2000 GAM not used
Los Angeles, CA (>29 years)
Cardiac arrhythmia Hospital Admissions:
Linn et al., 2000 GAM not used
Los Angeles, CA (>29 years)
Cerebrovascular Hospital Admissions:
Linn et al., 2000 GAM not used
Los Angeles, CA (>29 years)
Stroke Hospital Admissions:
Linn et al., 2000
Los Angeles, CA (>29 years)
GAM not used
9.2 (-0.3-19.6)
8.4 (-1.0-18.7)
3.92 (1.02, 6.90)
3.04(0.06,6.12)
1.01 (-1.93, 4.02)
0.30 (-2.13, 2.79)
6.72 (3.64, 9.90)
8.0(1.4-15.0)
6.8(0.3-13.8)
4.4 (-4.0-13.5)
4.9 (-3.55-14.1)
PM10.25 13 (4, 50)
PM10 48 (22, 82)
mean (10%, 90%)
PM10 45.5 (5, 132)
PM1031(maxl05)
PM2518(6, 86)
PM10.25 13 (4, 50)
PM1041(6, 117)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
3A-10
-------
Original study*
Study Location
Reanalysis study
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito, 2003
Other Cardiovascular Effects,
Gold et al.,
Boston, MA
Peters et al., 2000
Boston, MA
Peters etal., 2001
Boston, MA
Schwartz etal., 2001
U.S. population (NHANES)
Pope etal., 1999
Utah Valley, UT
Liao et al., 1999
Baltimore, MD
Levy etal., 2001
Seattle, WA
Analysis
Comments
Strict GAM
GLMNS
% increase (95% CI)
per
50 ug/m3 PM10
5.00 (-5.27, 16.38)
4.41 (-5.81, 15.74)
% increase (95% CI) % increase (95% CI)
per per
25 ug/m3 PM25 25 ug/m3 PM10.25
1.94 (-5.16, 9.57) 5.00 (-4.59, 15.56)
0.97 (-6.06, 8.52) 5.63 (-4.02, 16.25)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM1031(maxl05)
PM25 18(6,86)
PM1(M, 13 (4, 50)
Including Physiological Changes or Biomarkers
GAM
stringent
GAM not used
GAM not used
GAM not used
GAM not used
GAM not used
GAM not used
"
(cardiac arrhythmia, 10+
events)
144.6 (-2.8, 515.8)
(myocardial infarction)
132.7(18.7,356.3)
(fibrinogen)
25,7 (8.8, 42.6)
(heart rate)
34.5(3.1,65.9)
—
(cardiac arrest)
-30.3 (-53.4, 4.3)
(heart rate)
-2.3 (-4.2, -0.3)
(r-MSSD)
-6.3 (10.2, -2.3)
(cardiac arrhythmia, 10+
events)
75.4 (3.2, 198.2)
(myocardial infarction) (myocardial infarction)
82.8 (16.0, 188.1) 73.1 (-17.0, 261.1)
—
—
(heart rate variability)
-0.1 (-0.18, -0.03)
—
PM25(4-hr) 15.3(2.9,
48.6)
PM10 19.3 (max = 62.5)
PM25 12.7 (max = 53. 2)
PM10 19.4 (SD=9.4)
PM2512.1(SD=6.6)
PM10.2 5 7.4 (SD=4.4)
PM10 35.2 (SD=20.5)
PM10NR (15,145 from
figure)
PM25 16.1 (8.0,32.2)
PM1031.9(6.0, 178.0)
3A-11
-------
Original study*
Study Location
Reanalysis study
Analysis
Comments
% increase (95% CI)
per
50 ug/m3 PM10
% increase (95% CI)
per
25 ug/m3 PM2 5
% increase (95% CI)
per
25 ug/m3 PM10.25
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
RESPIRATORY MORBIDITY
Total Respiratory Hospital Admissions:
Thurston et al., 1994 GAM not used 23.26 (2.03, 44.49)
Toronto, Canada
Linn et al., 2000
Los Angeles, CA (>29 years)
Schwartz et al., 1996
Cleveland, OH (>65 years)
Burnett et al., 1997
Toronto, CAN (all ages)
Delfinoetal., 1997
Montreal, CAN (>64 years)
Delfinoetal., 1998
Montreal, CAN (>64 years)
Stieb et al., 2000
St. John, CAN (all ages)
GAM not used 2.89 (1.09, 4.72)
GAM not used
5.8 (0.5, 11.4)
GAM not used 10.93 (4.53, 17.72)
GAM not used 36.62 (10.02, 63.21)
GAM not used
GAM not used
Pneumonia Hospital Admissions:
Schwartz, 1995 GAM not used
Detroit (> 65 years)
8.8(1.8, 16.4)
5.9(1.9, 10.0)
15.00(1.97, 28.03)
8.61 (3.39, 14.08)
23.88 (4.94, 42.83)
13.17 (-0.22, 26.57)
5.69(0.61, 11.03)
22.25 (-9.53, 54.03)
12.71(5.33,20.74)
PM10 29.5-38.8 (max
96.0)
PM2515.8-22.3 (max
66.0)
PM10_2512.7-16.5 (max
33.0)
PM10 45.5 (5, 132)
PM10 43
PM1028.1(4, 102)
PM25 16.8(1,66)
PM10.25 11.6 (1,56)
summer 93
PM1021.7(max51)
PM2 5 12.2 (max 31)
PM25 18.6 (SD 9.3)
summer 93
PM10 14.0 (max 70.3)
PM2 5 8.5 (max 53. 2)
PM1048(22,82)
mean (10%, 90%)
3A-12
-------
Original study*
Study Location
Reanalysis study
Samet et al., 2000
14 U.S. Cities (>65 years)
Zanobetti and Schwartz (2003b)
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito 2003
COPD Hospital Admissions:
Schwartz, 1995
Detroit (> 65 years)
Samet et al., 2000
14 U.S. Cities (>65 years)
Zanobetti and Schwartz (2003b)
Linn et al., 2000
Los Angeles, CA (>29 years)
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito (2003)
Moolgavkar, 2000c
Cook Co., IL (all ages)
Moolgavkar 2003
Analysis
Comments
Strict GAM
Strict GAM
(dist. lag)
GLMNS
GLMPS
Strict GAM
GLMNS
GAM not used
Strict GAM
Strict GAM
(dist. lag)
GLMNS
GLMPS
GAM not used
Strict GAM
GLMNS
Strict GAM:
100 df
% increase (95% CI)
per
50 ug/m3 PM10
8.8(5.9, 11.8)
8.3 (4.9, 12.0)
2.9 (0.2, 5.6)
6.3 (2.5, 10.3)
18.1(5.3,32.5)
18.6(5.6,33.1)
10.6(4.4, 17.2)
8.8(4.8, 13.0)
13.3 (6.2, 20.9)
6.8 (2.8, 10.8)
8.0(4.3, 11.9)
1.5 (-0.5, 3.5)
6.5 (-7.8, 23.0)
4.6 (-9.4, 20.8)
3.24 (.03, 6.24)
% increase (95% CI) % increase (95% CI) PM10, PM2 5 and PM10.2 5
per per Mean (Range) Levels
25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
PM10 means 24.4-45.3
10.5(1.8,19.8) 9.9 (-0.1, 22.0) PM10 31 (max 105)
10.1(1.5,19.5) 11.2 (-0.02, 23.6) PM2518(6, 86)
PM10.25 13 (4, 50)
PM1048(22,82)
mean (10, 90)
PM10 means 24.4-45.3
PM10 45.5 (5, 132)
3.0(-6.9, 13.9) 8.7 (-4.8, 24.0) PM10 31 (max 105)
0.3(-9.3, 10.9) 10.8 (-3. 1,26.5) PM2518(6, 86)
PM10.25 13 (4, 50)
PM10 median 35 (3, 365)
3A-13
-------
Original study*
Study Location
Reanalysis study
Moolgavkar, 2000c
Los Angeles, CA (all ages)
Moolgavkar 2003
Asthma Hospital Admissions:
Choudburyetal., 1997
Anchorage, AK
Medical Visits (all ages)
Jacobs etal., 1997
Butte County, CA (all ages)
Linn et al., 2000
Los Angeles, CA (>29 years)
Lipsett et al., 1997
Santa Clara Co., CA (all ages)
Nauenberg and Basu, 1999
Los Angeles, CA (all ages)
Tolbert et al., 2000
Atlanta, GA (<17 years)
Sheppardetal., 1999
Seattle, WA (<65 years)
Sheppardetal., 2003
Analysis
Comments
Strict GAM:
30 df
Strict GAM:
100 df
GLMNS:
lOOdf
GAM not used
GAM not used
GAM not used
GAM not used
GAM not used
GAM not used
Strict GAM
GLMNS
% increase (95% CI)
per
50 ug/m3 PM10
7.78(4.30, 11.38)
5.52 (2.53-8.59)
5.00(1.22,8.91)
20.9(11.8,30.8)
6.11(p>0.05)
1.5 (-2.4, 5.6)
9.1(2.7, 15.9)
(below 40° F)
20.0 (5.3, 35)
13.2(1.2,26.7)
10.9 (2.8, 19.6)
8.1(0.1, 16.7)
% increase (95% CI) % increase (95% CI) PM10, PM2 5 and PM10.2 5
per per Mean (Range) Levels
25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
4.69 (2.06, 7.39) PM10 median 44 (7, 166)
PM2 5 median 22 (4, 86)
2.87 (0.53, 5.27)
2.59 (-0.29, 5.56)
PM10 42.5 (1, 565)
PM10 34.3 (6.6, 636)
PM10 45.5 (5, 132)
PM1061.2(9, 165)
44.8 (SE 17.23)
PM10 38.9 (9, 105)
8.7 (3.2, 14.4) 5.5 (0, 14.0) PM10 31.5 (90 55)
6.5(1.1,12.0) 5.5 (-2.7, 11.1) PM25 16.7 (90 32)
PM10.2 5 16.2 (90 29)
3A-14
-------
Original study*
Study Location
Reanalysis study
Analysis % increase (95% CI) % increase (95% CI) % increase (95% CI) PM10, PM2 s and PM10.2 5
Comments per per per Mean (Range) Levels
25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
50 ug/m3 PM10
Respiratory Symptoms
Odds Ratio (95% CI) for Odds Ratio (95% CI) for Odds Ratio (95% CI) for PM10.2.5 Mean (Range)
50 ug/m3 % increase in 25 ug/m3 % increase in 25 ug/m3 % increase in Levels Reported**
PM,n PM,, PMin.7,
Schwartz et al, 1994
6 U.S. cities
(children, cough)
Schwartz et al., 1994
6 U.S. cities
(children, lower respiratory
symptoms)
Neas et al, 1995
Uniontown, PA
(children, cough)
Ostroetal, 1991
Denver, CO
(adults, cough)
Pope et al., 1991
Utah Valley, UT
(lower respiratory symptoms,
schoolchildren)
Pope et al., 1991
Utah Valley, UT
(lower respiratory symptoms,
asthmatic patients)
GAM not used 1.39(1.05,1.85)
GAM not used 2.03 (1.36, 3.04)
GAM not used
GAM not used 1.09 (0.57, 2.10)
GAM not used 1.28 (1.06, 1.56)
GAM not used
1.01 (0.81, 1.27)
1.24(1.00, 1.54)
1.58(1.18,2.10)
2.45 (1.29, 4.64)
PM10 median 30.0 (max
117)
PM2 s median 18.0 (max
86)
PM10 median 30.0 (max
117)
PM2 5 median 18.0 (max
86)
PM25 24.5 (max 88.1)
PM1022(0.5,73)
PM1044 (11, 195)
PM1044(11, 195)
3A-15
-------
Original study*
Study Location
Reanalysis study
Neas et al., 1996
State College, PA
(children, cough)
Neas et al., 1996
State College, PA
(children, wheeze)
Neas et al., 1996
State College, PA
(children, cold)
Ostroetal., 1995
Los Angeles, CA
(children, asthma episode)
Analysis
Comments
GAM not used
GAM not used
GAM not used
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
NR
NR
NR
1.05 (0.64, 1.73)
% increase (95% CI)
per
25 ug/m3 PM2 5
1.48(1.17, 1.88) (1-d)
1.59 (0.93, 2.70) (1-d)
1.61 (1.21, 2. 17) (0-d)
—
% increase (95% CI) PM10, PM2 5 and PM10.2 5
per Mean (Range) Levels
25 ug/m3 PM10.25 Reported"
PM1031.9(max82.7)
PM2123.5(max85.8)
PM1031.9(max82.7)
PM2123.5(max85.8)
PM1031.9(max82.7)
PM2123.5(max85.8)
PM10 55.87 (19.63,
101.42)
Ostroetal., 1995
Los Angeles, CA
(children, shortness of breath)
Schwartz and Neas, 2000
Six Cities reanalysis
(children, cough)
Schwartz and Neas, 2000
Six Cities reanalysis
(children, lower respiratory
symptoms)
Vedaletal., 1998
Port Alberni, CAN
(children, cough)
GAMnotused 1.51(1.04,2.17)
GAM not used
GAM not used
1.28 (0.98, 1.67)
1.61(1.20,2.16)
GAM not used
1.40(1.14, 1.73)
PM10 55.87 (19.63,
101.42)
1.77 (1.23, 2.54) PM25 (same as Six Cities)
PM10.25 NR
1.51 (0.66, 3.43) PM25 (same as Six Cities)
PM10.25 NR
PM10 median 22.1(0.2,
159.0) (north site)
3A-16
-------
Original study*
Study Location
Reanalysis study
Vedaletal., 1998
Port Alberni, CAN
(children, phlegm)
Vedaletal., 1998
Port Alberni, CAN
(children, nose symptoms)
Vedaletal., 1998
Port Alberni, CAN
(children, sore throat)
Vedaletal., 1998
Port Alberni, CAN
(children, wheeze)
Vedaletal., 1998
Port Alberni, CAN
(children, chest tightness)
Analysis % increase (95% CI)
Comments per
50 ug/m3 PM10
GAM not used 1.40 (1.03, 1.90)
GAM not used 1.22 (1.00, 1.47)
GAM not used 1.34 (1.06, 1.69)
GAMnotused 1.16(0.82,1.63)
GAM not used 1.34 (0.86, 2.09)
% increase (95% CI) % increase (95% CI) PM10, PM25 and PM10.2 5
per per Mean (Range) Levels
25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
PM10 median 22. 1(0.2,
159.0) (north site)
Vedaletal., 1998
Port Alberni, CAN
(children, dyspnea)
Vedaletal., 1998
Port Alberni, CAN
(children, any symptom)
GAM not used 1.05 (0.74, 1.49)
GAM not used
1.16(1.00, 1.34)
PM10 median 22.1(0.2,
159.0) (north site)
PM10 median 22.1(0.2,
159.0) (north site)
3A-17
-------
Original study*
Study Location
Reanalysis study
Lung Function Changes
Neas et al, 1995
Uniontown, PA
(children)
Analysis % increase (95% CI)
Comments per
50 ug/m3 PM10
Lung Function change
(L/min) (95% CI) for 50
ug/m3 % increase in
PM,n
GAM not used
% increase (95% CI)
per
25 ug/m3 PM2 5
Lung Function change
(L/min) (95% CI) for 25
ug/m3 % increase in
PM,,
-2.58 (-5.33, +0.35)
% increase (95% CI)
per
25 ug/m3 PM10.25
Lung Function change
(L/min) (95% CI) for 25
ug/m3 % increase in
PM,n.,<
—
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM10_2.5 Mean (Range)
Levels Reported**
PM25 24.5 (max 88.1)
Thurstonetal, (1997)
Connecticut summer camp
(children)
Naeheretal., 1999
Southwest VA
(adult women)
Neas et al., 1996
State College, PA
(children)
Neas et al., 1999
Philadelphia, PA
(children)
Schwartz and Neas, 2000
Uniontown, PA (reanalysis)
(children)
Schwartz and Neas, 2000
State College PA (reanalysis)
(children)
GAM not used
GAM not used am PEFR -3.65 (-6.79,
0.51)
pm PEFR-1.8 (-5.03,
1.43)
GAM not used
GAMnotused am PEFR-8.17 (-14.81,
-1.56)
pm PEFR-1.44 (-7.33,
4.44)
GAM not used
GAM not used
PEFR-5.4 (-12.3, 1.5)
(15 ug/m3 S04=)
SO4=7.0 (1.1, 26.7)
am PEFR-1.83 (-3.44, - am PEFR-6.33 (-12.50, PM10 27.07 (4.89, 69.07)
0.21) -0.15) PM2 5 21.62 (3.48,59.65)
pmPEFR -1.05 (-2.77, pmPEFR -2.4 (-8.48, PM10.25 5.72 (0.00, 19.78)
0.67) 3.68)
pmPEFR-0.64 (-1.73,
0.44)
am PEFR-3.29 (-6.64,
0.07)
pmPEFR-0.91 (-4.04,
2.21)
pmPEFR-1.52, (-2.80,-
0.24)
pmPEFR-0.93 (-U
0.01)
am PEFR-4.31 (-11.44,
2.75)
pmPEFR 1.88 (-4.75,
8.44)
pmPEFR+1.73 (-2.2,
5.67)
pmPEFR-0.28 (-3.45,
2.87)
PM25 23.5 (max 85.8)
PM2522.2(IQR16.2)
PM10.259.5(IQR5.1)
PM2 5 24.5 (max 88.1)
PM10.25 NR
PM25 23.5 (max 85.8)
PM10.25 NR
3A-18
-------
Original study*
Study Location
Reanalysis study
Vedaletal., 1998
Port Alberni, CAN
(children)
Analysis
Comments
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
PEF -1.35 (-2.7, -0.05)
% increase (95% CI)
per
25 ug/m3 PM2 5
—
% increase (95% CI)
per
25 ug/m3 PM10.25
—
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM10 median 22. 1(0.2,
159.0) (north site)
* Studies in italics available in 1996 CD
** mean (minimum, maximum) 24-h PM level shown in parentheses unless otherwise noted.
3A-19
-------
Appendix 3B. Mortality and Morbidity Effect Estimates and PM Concentrations from
U.S. and Canadian Studies for Long-Term Exposures to PM10, PM2 5, and PM10_2 5
Study
Indicator (Increment)
Relative Risk (95% CI)
Study
Concentrations
(us/m3)
Increased Total Mortality in Adults
Six CityA
Six CityB
ACS Study0
(151 U.S. SMSA)
Six City ReanalysisD
ACS Study ReanalysisD
ACS Study Extended
AnalysesE
Southern CaliforniaF
Southern California11
Veterans Cohort0
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
SOI (15 ug/m3)
PM15.25 (10 ug/m3)
PM25 (10 ug/m3)
SOI (15 ug/m3)
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3) (dichot)
PM25 (10 ug/m3)
PM15.25 (10 ug/m3)
PM2 5 (10 ug/m3) (1979-83)
PM25 (10 ug/m3) (1999-00)
PM25 (10 ug/m3) (average)
PM10 (20 ug/m3)
PM10 (30 days/year>100 ug/m3)
PM10 (20 ug/m3)
PM10 (30 days/year>100 ug/m3)
PM25 (10 ug/m3)
PM10.25 (10 ug/m3)
PM25 (10 ug/m3) (1979-81)
1.18(1.06, 1.32)
1.13(1.04, 1.23)
1.54(1.15,2.07)
1.43 (0.83, 2.48)
1.07(1.04, 1.10)
1.11(1.06, 1.16)
1.19(1.06, 1.34)
1.14(1.05, 1.23)
1.04 (1.01, 1.07)
1.07(1.04, 1.10)
1.00 (0.99, 1.02)
1.04(1.01, 1.08)
1.06(1.02, 1.10)
1.06(1.02, 1.11)
1.09 (0.99, 1.21) (males)
1.08(1.01, 1.16) (males)
0.95 (0.87, 1.03) (females)
0.96 (0.90, 1.02) (females)
1.09 (0.98, 1.21) (males)
1.05 (0.92, 1.21) (males)
0.90 (0.85, 0.95) (males)
NR(18, 47)
MR (11, 30)
NR(5, 13)
18U (9, 34)
llu(4, 24)
NR(18, 47)
MR (11, 30)
59(34, 101)
20 (10, 38)
7.1(9,42)
21 (9, 34)
14 (5, 20)
18 (7.5, 30)
51(0,84)
51(0,84)
32 (17, 45)
27 (4, 44)
24 (6, 42)
Increased Cardiopulmonary Mortality in Adults
Six CityA
Six City ReanalysisD
ACS Study0
ACS Study ReanalysisD
ACS Study Extended
AnalysesE
Southern CaliforniaF
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
PM25 (10 ug/m3)
PM15/10 (20 ug/m3) (dichot)
PM25 (10 ug/m3)
PM15.25 (10 ug/m3)
PM25 (10 ug/m3) (1979-83)
PM25 (10 ug/m3) (1999-00)
PM25 (10 ug/m3) (average)
PM10 (20 ug/m3)
V
1.18(1.06, 1.32)
1.20(1.03, 1.41)
1.19(1.07, 1.33)
1.12(1.07, 1.17)
1.07(1.03, 1.12)
1.12(1.07, 1.17)
1.00 (0.98, 1.03)
1.06(1.02, 1.10)
1.08(1.02, 1.14)
1.09(1.03, 1.16)
1.01 (0.92, 1.10)
NR(18, 47)
MR (11, 30)
NR(18, 47)
MR (11, 30)
18U (9, 34)
59(34, 101)
20 (10, 38)
7.1(9,42)
21 (9, 34)
14 (5, 20)
18 (7.5, 30)
51(0,84)
3B-1
-------
Study Indicator (Increment)
Southern California11 PM2 5 (10 ug/m3)
PM10.25 (10 ug/m3)
Relative Risk (95% CI)
1.23 (0.97, 1.55) (males)
1.20 (0.87, 1.64) (males)
Study
Concentrations
(us/m3)
32 (17, 45)
27 (4, 44)
Increased Lung Cancer Mortality in Adults
Six CityA PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
Six City ReanalysisD PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
ACS Study0 PM25 (10 ug/m3)
ACS Study ReanalysisD PM15/10 (20 ug/m3) (dichot)
PM25 (10 ug/m3)
PM15.25 (10 ug/m3)
ACS Study Extended PM2 5 ( 10 ug/m3) ( 1 979-83)
Analyses13 PM25 (10 ug/m3) (1999-00)
PM25 (10 ug/m3) (average)
Southern CaliforniaF PM10 (20 ug/m3)
Southern California11 PM2 5 (10 ug/m3)
****
1.18(0.89, 1.57)
1.14(0.75, 1.74)
1.21 (0.92, 1.60)
1.01(0.91, 1.12)
1.01(0.91, 1.11)
1.01(0.91, 1.11)
0.99 (0.93, 1.05)
1.08(1.01, 1.16)
1.13 (1.04, 1.22)
1.14(1.05, 1.24)
1.81(1.14, 2.86) (males)
1.39 (0.79, 2.50) (males)
1.26 (0.62, 2.55) (males)
NR(18, 47)
MR (11, 30)
NR(18, 47)
NR (11,30)
18U (9, 34)
59(34, 101)
20 (10, 38)
7.1(9,42)
21 (9, 34)
14 (5, 20)
18 (7.5, 30)
51(0,84)
32 (17, 45)
27 (4, 44)
Increased Bronchitis in Children
Six City1 PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
24 CityJ SOI (15 ug/m3)
PM21(10ug/m3)
PM10 (20 ug/m3)
AHSMOGK SOJ (15 ug/m3)
12 Southern California PM10 (20 ug/m3)
communitiesL (1986-1990 data)
(all children)
12 Southern California PM10 (20 ug/m3)
communitiesM PM25 (10 ug/m3)
(children with asthma)
1.6(1.1,2.5)
1.3 (0.9, 2.0)
3.02 (1.28, 7.03)
1.31(0.94, 1.84)
1.60 (0.92, 2.78)
1.39 (0.99, 1.92)
0.95 (0.79, 1.15)
1.4(1.1, 1.8)
1.3 (0.9, 1.7)
NR (20, 59)
NR(12, 37)
4.7 (0.7, 7.4)
14.5 (5.8, 20.7)
23.8(15.4,32.7)
—
NR (28.0, 84.9)
34.8 (13.0, 70.7)
15.3 (6.7,31.5)
Increased Cough in Children
12 Southern California PM10 (20 ug/m3)
communitiesL (1986-1990 data)
(all children)
12 Southern California PM10 (20 ug/m3)
communitiesM PM25 (10 ug/m3)
(children with asthma)
1.05 (0.94, 1.16)
1.1 (0.7, 1.8)
1.2 (0.8, 1.8)
NR (28.0, 84.9)
13.0-70.7
6.7-31.5
Increased Airway Obstruction in Adults
AHSMOGK PM10 (20 ug/m3)
1.19(0.84, 1.68)
NR
3B-2
-------
Study
Indicator (Increment)
Relative Risk (95% CI)
Study
Concentrations
(us/m3)
Decreased Lung Function in Children
Six City1
24 CityJ
12 Southern California
communities1"
(all children)
12 Southern California
communities1"
(all children)
12 Southern California
communities'3
(4th grade cohort)
12 Southern California
communities'3
(4th grade cohort)
12 Southern California
communitiesR
(second 4th grade
cohort)
12 Southern California
communitiesR
(second 4th grade
cohort)
12 Southern California
communitiesR
(second 4th grade
cohort)
12 Southern California
communities8
12 Southern California
communities8
PM15/10 (50 ug/m3)
SOJ (15 ug/m3)
PM21(10ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
(1986-90 data)
PM10 (20 ug/m3)
(1986-1990 data)
PM10 (20 ug/m3)
PM25(10 ug/m3)
PM10.25 (10 ug/m3)
PM10 (20 ug/m3)
PM25(10ug/m3)
PM10.25 (10 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10 (20 ug/m3)
PM10 (20 ug/m3)
NS Changes
-6.56% (-9.64, -3.43) FVC
-2. 15% (-3.34, -0.95) FVC
-2.80% (-4.97, -0.59) FVC
-19.9 (-37.8, -2.6) FVC
-25.6 (-47.1, -5.1) MMEF
-0.23 (-0.44, -0.01) FVC %
growth
-0. 18 (-0.36, 0.0) FVC %
growth
-0.22 (-0.47, 0.02) FVC %
growth
-0.51 (-0.94, -0.08) MMEF %
growth
-0.4 (-0.75, -0.04) MMEF %
growth
-0.54 (-1.0, -0.06) MMEF %
growth
- 0.12 (-0.26, 0.24) FVC %
growth
-0.06 (-0.30, 0.18) FVC %
growth
-0.26 (-0.75, 0.23) MMEF %
growth
-0.42 (-0.84, 0.0) MMEF %
growth
-0.16 (-0.62, 0.30) PEFR %
growth
-0.20 (-0.64, 0.25) PEFR %
growth
-3.6 (-18, 11) FVC growth
-33 (-64, -2.2) MMEF growth
NR (20, 59)
4.7 (0.7, 7.4)
14.5 (5.8, 20.7)
23.8(15.4,32.7)
NR (28.0, 84.9)
NR (28.0, 84.9)
NR (15, 70)x
NR(10, 35)x
NR
NR (15, 70)x
NR(10, 35)x
NR
NR (10, 80)Y
NR (5, 30)Y
NR (10, 80)Y
NR (5, 30)Y
NR (10, 80)Y
NR (5, 30)Y
NR (15.0, 66.2)
NR (15.0, 66.2)
3B-2
-------
Study
Indicator (Increment)
Study
Relative Risk (95% CI) Concentrations
(ug/m3)
12 Southern California PM10 (20 ug/m3)
communities
s
-70 (-120, -20) PEFR growth NR (15.0, 66.2)
Lung Function Changes in Adults
AHSMOG1 (%
predicted FEVb
females)
AHSMOG1
(% predicted FEVb
males)
AHSMOG1
(% predicted FEVb
males whose parents
had asthma, bronchitis,
emphysema)
AHSMOG1
(% predicted FEVb
males)
PM10 (cutoff of 54.2 days/year
>100 ug/m3)
PM10 (cutoff of 54.2 days/year
>100 ug/m3)
PM10 (cutoff of 54.2 days/year
>100 ug/m3)
(1.6 ug/m3)
+0.9 % (-0.8, 2.5) FEVj 52.7 (21.3, 80.6)
+0.3 % (-2.2, 2.8) FEVj 54.1 (20.0, 80.6)
-7.2 % (-11.5,-2.7) FEVj
-1.5% (-2.9, -0.1)FEV!
54.1(20.0,80.6)
7.3(2.0, 10.1)
References:
ADockeryetal. (1993)
B EPA (1996a)
c Pope etal. (1995)
D Krewski et al. (2000)
E Pope et al. (2002)
F Abbey etal. (1999)
GLipfertetal. (2000b)
H McDonnell et al. (2000)
'Dockery etal. (1989)
JDockery etal. (1996)
K Abbey etal. (1995a,b,c)
L Peters etal. (1999a)
MMcConnell etal. (1999)
NBerglundetal. (1999)
0 Raizenne et al. (1996)
p Peters etal. (1999)
Q Gauderman et al. (2000)
R Gauderman et al. (2002)
sAvoletal. (2001)
T Abbey etal. (1998)
Note: Study concentrations are presented as mean (min, max), or mean (±SD); NS Changes = No significant changes
(no quantitative results reported); NR=not reported.
u Median
v Results only for smoking category subgroups.
x Estimated from Figure 1, Gauderman et al. (2000)
Y Estimated from figures available in online data supplement to Gauderman et al. (2002)
3B-4
-------
Table 4A-1. Study-Specific Information for Short-term Exposure PM2.5 Studies in Boston, MA
Study
Health Effect
ICD-9
Codes
Ages
Short-Term
Schwartz (2003b)
[reanalysis of Schwartz et
al. (1996)]
Schwartz (2003b)
preanalysis of Schwartz et
al. (1996)] -6 cities
Non-accidental
Non-accidental
<800
<800
all
all
Other
Model Pollutants
in Model
Exposure Total Mortality
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
Observed
Concentrations
min. max.
Lag
Exposure PM2.5
Metric Coeff.
Lower Upper
Bound Bound
•- Single Pollutant Models
0
0
yn R mean of
70'8 lag 0 & 1
mean of
Iag0&1
2-day avg 0.00206
2-day avg 0.00137
0.00139 0.00273
0.00098 0.00176
Short-Term Exposure Cause-Specific Mortality — Single Pollutant Models
Klemm and Mason (2003)
preanalysis of Klemm et al.
(2000)]
Klemm and Mason (2003)
'reanalysis of Klemm et al.
(2000)]
Klemm and Mason (2003)
preanalysis of Klemm et al.
(2000)]
Klemm and Mason (2003)
preanalysis of Klemm et al.
(2000)] -- 6 cities
Klemm and Mason (2003)
[reanalysis of Klemm et al.
(2000)] -- 6 cities
Klemm and Mason (2003)
[reanalysis of Klemm et al.
(2000)] -- 6 cities
COPD
Ischemic heart
disease
Pneumonia
COPD
Ischemic heart
disease
Pneumonia
490-492,
494-496
410-414
480-487
490-492,
494-496
410-414
480-487
all
all
all
all
all
all
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
Respiratory Symptoms and Illnesses**
Schwartz and Neas (2000) -
- 6 cities
Schwartz and Neas (2000) -
- 6 cities
Lower
respiratory
symptoms*
cough*
n/a
n/a
7-14
7-14
Respiratory
Schwartz and Neas (2000) -
- 6 cities
Schwartz and Neas (2000) -
- 6 cities
Lower
respiratory
symptoms*
cough*
n/a
n/a
7-14
7-14
logistic none
logistic none
Symptoms and Illnesses*
logistic PM10-2.5
logistic PM10-2.5
0
0
0
0
0
0
70.8
70.8
70.8
174
174
174
- Single Pollutant
N/A
N/A
N/A
N/A
0 day
Oday
0 day
0 day
Oday
0 day
Models
1 day
Oday
2-day avg 0.00276
2-day avg 0.00266
2-day avg 0.00573
2-day avg 0.00227
2-day avg 0.00178
2-day avg 0.00402
1 -day avg 0.01901
3-day avg 0.00989
-0.00131 0.00658
0.00149 0.00383
0.00257 0.00871
0.00010 0.00440
0.00109 0.00247
0.00188 0.00602
0.00696 0.03049
-0.00067 0.02050
* ~ Multi-Pollutant Models
N/A
N/A
N/A
N/A
1 day
Oday
1 -day avg 0.01698
3-day avg 0.00451
0.00388 0.03007
-0.00702 0.01541
*The C-R functions for lower respiratory symptoms and cough were calculated for the summer period April 1 through August 31
4A-1
-------
Table 4A-2. Study-Specific Information for Short-term Exposure PM2.5 Studies in Detroit, Ml
Study
Health Effect 'C°~9 Ages
Codes M
Short-Term
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Non-accidental <800
Short-Term
Circulatory 390-459
Respiratory 460-519
all
Other Observed
Model Pollutants Concentrations
in Model min. max.
Exposure Total Mortality - Single Pollutant
log-linear,
GAM
(stringent)
none 4
Exposure Cause-Specific Mortality -
all
all
log-linear,
GAM
(stringent)
log-linear,
GAM
(stringent)
Hospital Admissions
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Pneumonia 480-486
COPD 490-496
Ischemic heart 41Q_414
disease
Congestive heart
, .. 4zo
failure
Dysrhythmias 427
65+
65+
65+
65+
65+
log-linear
GAM
(stringent)
log-linear,
GAM
(stringent)
log-linear,
GAM
(stringent)
log-linear,
GAM
(stringent)
log-linear,
GAM
(stringent)
none 4
none 4
86
Lag
Models
3 day
Exposure PM2.5 Lower
Metric Coeff. Bound
1-dayavg 0.00074 -0.00073
Upper
Bound
0.00221
Single Pollutant Models
86
86
1 day
0 day
1-day avg 0.00087 -0.00131
1-dayavg 0.00090 -0.00438
0.00305
0.00618
- Single Pollutant Models
none 4
none 4
none 4
none 4
none 4
86
86
86
86
86
1 day
3 day
2 day
1 day
1 day
1-dayavg 0.00398 0.00074
1-dayavg 0.00117 -0.00287
1-dayavg 0.00143 -0.00082
1-dayavg 0.00307 0.00055
1-dayavg 0.00125 -0.00274
0.00725
0.00523
0.00371
0.00561
0.00523
4A-2
-------
Table 4A-3. Study-Specific Information for Short-term Exposure PM25 Studies in Los Angeles, CA
Study
Health Effect 'C°'9
Codes
Other
Ages Model Pollutants
In Model
Observed
Concentrations Lag
min. max.
E*P°SUre PM2.5Coeff. ^ower
Metric Bound
Upper
Bound
Short-Term Exposure Total Mortality - Single Pollutant Models
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
Non-accidental <800
.. log-linear, GAM
a" (stringent), 100 df none
.. log-linear, GAM
a" (stringent), 100 df none
,, log-linear, GAM
a" (stringent), 30 df
all l°g-linear, GLM, 30 none
df
„ log-linear, GAM
a ,f. „ „-- ,, none
(stringent), 100df
log-linear, GLM,
all 1Quodf
,, log-linear, GAM
all . f. ' „ ., none
(stringent), 30 df
„ log-linear, GLM, 30
all ,, none
df
log-linear, GAM
a" (stringent), 100 df n°ne
,, log-linear, GLM,
all 1Quodf
4 86 0 day
4 86 1 day
4 86 0 day
4 86 0 day
4 86 0 day
4 86 0 day
4 86 1 day
4 86 1 day
4 86 1 day
4 86 1 day
1-dayavg 0.00032 -0.00023
1-day avg 0.00010 -0.00046
1-dayavg 0.00054 -0.00007
1-dayavg 0.00040 -0.00034
1-dayavg 0.00032 -0.00023
1-dayavg 0.00030 -0.00043
1-dayavg 0.00059 0.00000
1-dayavg 0.00055 -0.00017
1-dayavg 0.00010 -0.00046
1-dayavg -0.00001 -0.00099
0.00086
0.00066
0.00114
0.00113
0.00086
0.00102
0.00117
0.00126
0.00066
0.00097
4A-3
-------
Study
Health Effect
'C°'9 Ages
Codes u
Other
Model Pollutants
in Model
Short-Term Exposure Cause-Specific Mortality
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
390-429
390-429
390-429
390-429
390-429
390-429
all
all
all
all
all
all
log-linear,
(stringent)
log-linear,
(stringent)
log-linear,
100df
log-linear,
(stringent)
log-linear,
(stringent)
log-linear,
100df
Tdf "-
Gioodf none
GLM,
none
GAM
30 df
GAM
ioodf none
GLM' none
Observed
Concentrations Lag
min. max.
Exposure
Metric
PM2.5 Coeff.
Lower
Bound
Upper
Bound
- Single Pollutant Models
4 86 0 day
4 86 0 day
4 86 0 day
4 86 1 day
4 86 1 day
4 86 1 day
1-day avg
1-day avg
1 -day avg
1 -day avg
1 -day avg
1 -day avg
0.00099
0.00097
0.00097
0.00103
0.00080
0.00069
0.00010
0.00014
-0.00002
0.00016
-0.00003
-0.00032
0.00187
0.00179
0.00195
0.00189
0.00162
0.00169
Short-Term Exposure Total Mortality - Multi-Pollutant Models
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)l
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000a)]
Non-accidental
Non-accidental
Non-accidental
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
<800
<800
<800
Short-Term
390-429
390-429
390-429
390-429
all
all
all
log-linear,
(stringent)
log-linear,
(stringent)
log-linear,
100df
GAM CO
30 df C°
GAM CO
100df OU
GLM' CO
Exposure Cause-Specific Mortality
all
all
all
all
log-linear,
(stringent)
log-linear,
100df
log-linear,
(stringent)
log-linear,
100df
GAM CO
100df C°
GLM' CO
GAM CO
100df °U
GLM' CO
4 86 1 day
4 86 1 day
4 86 1 day
1 -day avg
1 -day avg
1 -day avg
-0.00053
-0.00033
-0.00033
-0.00132
-0.00105
-0.00118
0.00025
0.00039
0.00051
- Multi-Pollutant Models
4 86 0 day
4 86 0 day
4 86 1 day
4 86 1 day
1 -day avg
1 -day avg
1 -day avg
1 -day avg
0.001 78
0.00188
0.00091
0.00091
0.00076
0.00068
-0.00012
-0.00034
0.00279
0.00306
0.00193
0.00215
4A-4
-------
Study
Health Effect
ICD-9
Codes
Ages
Model
Hospital Admissions —
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
Cardiovascular
COPD+
COPD+
COPD+
COPD+
COPD+
COPD+
COPD+
COPD+
COPD+
390-429
390-429
390-429
390-429
390-429
390-429
490-496
490-496
490-496
490-496
490-496
490-496
490-496
490-496
490-496
65+
65+
65+
65+
65+
65+
all
all
all
all
all
all
all
all
all
log-linear, GAM
(stringent), 30 df
log-linear, GAM
(stringent), 100 df
log-linear, GLM,
100df
log-linear, GAM
(stringent), 30 df
log-linear, GAM
(stringent), 100 df
log-linear, GLM,
100df
log-linear, GAM
(stringent), 30 df
log-linear, GAM
(stringent), 100 df
log-linear, GLM,
100df
log-linear, GAM
(stringent), 30 df
log-linear, GAM
(stringent), 100 df
log-linear, GLM,
100df
log-linear, GAM
(stringent), 30 df
log-linear, GAM
(stringent), 100 df
log-linear, GLM,
100df
Other
Pollutants
In Model
Observed
Concentrations
min. max.
Lag E*p°sure
u Metric
PM2.5 Coeff.
Lower
Bound
Upper
Bound
Single Pollutant Models
none
none
none
none
none
none
none
none
none
none
none
none
none
none
none
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
86
86
86
86
86
86
86
86
86
86
86
86
86
86
86
0 day 1 -day avg
Oday 1 -day avg
0 day 1 -day avg
1 day 1 -day avg
1 day 1 -day avg
1 day 1 -day avg
0 day 1 -day avg
0 day 1 -day avg
Oday 1 -day avg
1 day 1 -day avg
1 day 1 -day avg
1 day 1 -day avg
2 day 1 -day avg
2 day 1 -day avg
2 day 1 -day avg
0.00158
0.00116
0.00126
0.00139
0.00113
0.00120
0.00167
0.00138
0.00149
0.00119
0.00075
0.00077
0.00185
0.00114
0.00103
0.00091
0.00051
0.00045
0.00070
0.00047
0.00039
0.00069
0.00052
0.00042
0.00023
-0.00011
-0.00027
0.00084
0.00022
-0.0001 1
0.00224
0.00181
0.00206
0.00208
0.001 79
0.00200
0.00264
0.00223
0.00255
0.00214
0.00160
0.00180
0.00285
0.00205
0.00216
4A-5
-------
Study
ICD 9 °ther
Health Effect ' ™ Ages Model Pollutants
Codes in Model
Observed
Concentrations Lag
min. max.
E*P°SUre PM2.5Coeff. ^°wer
Metric Bound
Upper
Bound
Hospital Admissions - Single City, Multi-Pollutant Models
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000b)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)]
Moolgavkar (2003)
[reanalysis of Moolgavkar
(2000c)l
Cardiovascular 390-429 65+ '°?:llnear;
(stringent)
Cardiovascular 390-429 65+ f^'^631''
Cardiovascular 390-429 65+ '°?:llnear.;
(stringent)
Cardiovascular 390-429 65+ '°£;li"ear'
100 df
COPD+ 490-496 all '°9:lineart;
(stringent)
COPD+ 490-496 al, gjj
COPD+ 490-496 al, gjj
GAM CO
100df C°
GLM' CO
GAM CO
100df C°
GLM' CO
GAM N02
100df
GAM N02
100df N°2
GAM N02
100df N°2
4 86 0 day
4 86 0 day
4 86 1 day
4 86 1 day
4 86 0 day
4 86 1 day
4 86 2 day
1-dayavg 0.00039 -0.00044
1-dayavg 0.00058 -0.00041
1-dayavg 0.00024 -0.00065
1-dayavg 0.00027 -0.00075
1-dayavg 0.00042 -0.00091
1-dayavg -0.00004 -0.00162
1-dayavg 0.00035 -0.00103
0.00121
0.00156
0.00112
0.00128
0.001 73
0.00152
0.00171
4A-6
-------
Table 4A-4. Study-Specific Information for Short-term Exposure PM2.5 Studies in Philadelphia, PA
Lipfert et al
counties
Study*
(2000) - 7
Health Effect 'C°~9
Codes
Short-Term
Cardiovascular 390-448
Ages
Model
Other
Pollutants
in Model
Observed
Concentrations Lag
min. max.
Exposure Cause-Specific Mortality - Single
all
linear
none
-0.6 72.6
Pollutant
1 day
Exposure
.„ . . PM2.5 Coeff. Lower Bound
Metric
Models
1-dayavg 0.10440 0.04983
Upper
Bound
0.15897
The Lipfert et al. (2000) study does not provide the statistical uncertainties surrounding the PM2.5 non-accidental mortality coefficients and the cardiovascular mortality multi-pollutant coefficient.
Table 4A-5. Study-Specific Information for Short-term Exposure PM2.5 Studies in Phoenix, AZ
Study Health Effect
ICD-9 °ther
_ . Ages Model Pollutants
Codes . „„ . ,
in Model
Observed
Concentrations Lag ,?°SUre PM2.5 Coeff. Lower Bound ^Ppe'1
Metric Bound
mm. max.
Short-term Exposure Cause-Specific Mortality - Single Pollutant Models
Mar (2003) [reanalysis of Mar _ ,.
(2000)] Cardiovascular
Mar (2003) [reanalysis of Mar _ ,.
(2000)] Cardiovascular
oq0 log-linear,
4489 65+ GAM n°ne
(stringent)
390 log-linear,
65+ GAM none
(stringent)
0 42 Oday 1-dayavg 0.00371 -0.0010136 0.0084336
0 42 1 day 1-dayavg 0.00661 0.0019256 0.0112944
Table 4A-6. Study-Specific Information for Short-term Exposure PM2.5 Studies in Pittsburgh, PA
Study
Health Effect 'C°'9
Codes
Ages
Model
Other Observed
Pollutants Concentrations
Exposure
39 Metric
PM2.5 Coeff. Lower Bound ^Ppe!,
Bound
Short-term Exposure Total Mortality - Single Pollutant Models
Chock etal.
Chock etal.
(2000)
(2000)
Non-accidental <800
Non-accidental <800
<75
75+
log-linear
loq-linear
none
none
Short-term Exposure Total Mortality -
Chock etal.
Chock etal.
(2000)
(2000)
Non-accidental <800
Non-accidental <800
<75
75+
log-linear
log-linear
CO, O3,
SO2, NO2,
PM10-2.5
CO, O3,
SO2, NO2,
PM10-2.5
3
3
86
86
Multi-Pollutant
3
3
86
86
Oday 1-dayavg
Oday 1-dayavq
Models
Oday 1-dayavg
Oday 1-dayavg
0.00101 -0.00078964 0.00280964
0.00059 -0.00124556 0.00242556
0.0013 -0.00085932 0.00345932
0.0004 -0.00177778 0.00257778
4A-7
-------
Table 4A-7. Study-Specific Information for Short-term Exposure PM2.5 Studies in San Jose, CA
Study
Health Effect ICD-9 Codes
Other Observed
Ages Model Pollutants Concentrations
in Model min. max.
Lag
Exposure PM2.5 Lower
Metric Coeff. Bound
Upper
Bound
Short- lerm Exposure Total Mortality - single Pollutant Models
Fairley (2003) [reanalysis
of Fairley (1999)]
Fairley (2003) [reanalysis
of Fairley (1999)]
Non-accidental <800
Non-accidental <800
al
al
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
Short-Term Exposure Cause-Specific Mortality
Fairley (2003) [reanalysis
of Fairley (1999)]
Fairley (2003) [reanalysis
of Fairley (1999)]
11,35,472-
Respiratory 519,710.0,
710.2,710.4
Cardiovascular 390-459
al
al
Short-Term
Fairley (2003) [reanalysis
of Fairley (1999)]
Fairley (2003) [reanalysis
of Fairley (1999)]
Fairley (2003) [reanalysis
of Fairley (1999)]
Non-accidental <800
Non-accidental <800
Non-accidental <800
al
al
al
log-linear,
GAM none
(stringent)
log-linear,
GAM none
(stringent)
2 105
2 105
0 day
1 day
1-dayavg 0.00314 0.00064
1-day avg -0.00153 -0.00380
0.00567
0.00071
-Single Pollutant Models
2 105
2 105
Exposure Total Mortality - Multi-Pollutant
log-linear,
GAM N02
(stringent)
log-linear,
GAM CO
(stringent)
log-linear,
GAM O3 - 8hr
(stringent)
2 105
2 105
2 105
0 day
0 day
Models
0 day
0 day
0 day
1-dayavg 0.00446 -0.00416
1-dayavg 0.00248 -0.00168
1-dayavg 0.00402 0.00106
1-dayavg 0.00363 0.00085
1-dayavg 0.00340 0.00085
0.01307
0.00666
0.00698
0.00636
0.00594
Table 4A-8. Study-Specific Information for Short-term Exposure PM2.5 Studies in Seattle, WA
Study
Health Effect ICD-9 Codes
Ages
Model
Hospital Admissions
Sheppard (2003)
reanalysis of Sheppard
etal. (1999)]**
Asthma
"Sheppard (2003) [reanalysis of Sheppard et al
493
<65
log-linear,
GAM
(stringent)
Other
Pollutants
in Model
Observed
Concentrations
min. max.
Lag
Exposure
Metric
PM2.5
Coeff.
Lower
Bound
Upper
Bound
- Single Pollutant Models
none
2.5 96
1 day
1 -day avg 0
0033238
0.00084325
0.004938
(1999)] used daily PM2.5 values obtained from nephelometry measurements rather than from air quality monitors.
4A-8
-------
Table 4A-9. Study-Specific Information for Short-term Exposure PM2.5 Studies in St. Louis, MO
Study
Health Effect
ICD-9
Codes
Ages
Model
Other
Pollutants
in Model
Observed
Concentrations
min. max.
Lag
Exposure
Metric
PM2.5
Coeff.
Lower
Bound
Upper
Bound
Short-Term Exposure Total Mortality - Sinqle Pollutant Models
Schwartz (2003b)
reanalysis of Schwartz et
al. (1996)]
Schwartz (2003b)
reanalysis of Schwartz et
al. (1996)] -6 cities
Non-accidental
Non-accidental
<800
<800
all
all
log-linear,
GAM
(stringent)
log-linear,
GAM
(stringent)
none
none
0.9 88.9
0 174
mean of
lag 0 & 1
mean of
lag 0 & 1
2-day avg
2-day avg
0.00102
0.00137
0.00037
0.00098
0.00167
0.00176
Short-Term Exposure Cause-Specific Mortality - Single Pollutant Models
Klemm and Mason (2003)
reanalysis of Klemm et al.
(2000)]
Klemm and Mason (2003)
reanalysis of Klemm et al.
(2000)]
Klemm and Mason (2003)
reanalysis of Klemm et al.
(2000)]
Klemm and Mason (2003)
reanalysis of Klemm et al.
(2000)] - 6 cities
Klemm and Mason (2003)
reanalysis of Klemm et al.
(2000)] - 6 cities
Klemm and Mason (2003)
reanalysis of Klemm et al.
(2000)] - 6 cities
COPD
Ischemic heart
disease
Pneumonia
COPD
Ischemic heart
disease
Pneumonia
490-492,
494-496
41 0-41 4
480-487
490-492,
494-496
41 0-41 4
480-487
all
all
all
all
all
all
Log-linear,
GAM
(stringent)
Log-linear,
GAM
(stringent)
Log-linear,
GAM
(stringent)
Log-linear,
GAM
(stringent)
Log-linear,
GAM
(stringent)
Log-linear,
GAM
(stringent)
Respiratory Symptoms and
Schwartz and Neas (2000) -
- 6 cities
Schwartz and Neas (2000) -
- 6 cities
Lower respiratory
symptoms*
Cough*
n/a
n/a
7-14
7-14
logistic
logistic
Respiratory Symptoms and
Schwartz and Neas (2000) -
- 6 cities
Schwartz and Neas (2000) -
- 6 cities
Lower respiratory
symptoms*
Cough*
n/a
n/a
7-14
7-14
logistic
logistic
none
none
none
none
none
none
Illnesses**
none
none
Illnesses*
PM1 0-2.5
PM1 0-2.5
0.9 88.9
0.9 88.9
0.9 88.9
0 174
0 174
0 174
Oday
Oday
0 day
Oday
Oday
0 day
2-day avg
2-day avg
2-day avg
2-day avg
2-day avg
2-day avg
0.00060
0.00129
0.00109
0.00227
0.00178
0.00402
-0.00294
0.00030
-0.00253
0.00010
0.00109
0.00188
0.00411
0.00237
0.00459
0.00440
0.00247
0.00602
- Single Pollutant Models
N/A N/A
N/A N/A
' - Multi-Pollutant
N/A N/A
N/A N/A
1 day
0 day
Models
1 day
0 day
1 -day avg
3-day avg
1 -day avg
3-day avg
0.01901
0.00989
0.01698
0.00451
0.00696
-0.00067
0.00388
-0.00702
0.03049
0.02050
0.03007
0.01541
*The C-R functions for lower respiratory symptoms and cough were calculated for the summer period April 1 through August 31.
4A-9
-------
Table 4A-10. Study-Specific Information for Long-term Exposure PM2.5 Studies
Long-Term Exposure Mortality - Single Pollutant Models
Krewski et al. (2000)
ACS
Pope et al. (2002) -
ACS extended
Krewski et al. (2000)
ACS
Pope et al. (2002) -
ACS extended
Pope et al. (2002) -
ACS extended
All cause
All cause
Cardiopulmonary
Cardiopulmonary
Lung cancer
all
all
401-440,460-519
401-440,460-519
162
30+
30+
30+
30+
30+
Long-Term Exposure
Krewski et al. (2000)
ACS
Krewski et al. (2000)
ACS
Krewski et al. (2000)
ACS
Krewski et al. (2000)
ACS
All cause
All cause
All cause
All cause
all
all
all
all
30+
30+
30+
30+
log-linear
log-linear
log-linear
log-linear
log-linear
Mortality -
log-linear
log-linear
log-linear
log-linear
none
none
none
none
none
10
7.5
10
7.5
7.5
Multi-Pollutant
CO
N02
O3
SO2
10
10
10
10
38 n/a
30 n/a
38 n/a
30 n/a
30 n/a
Models
38 n/a
38 n/a
38 n/a
38 n/a
annual
mean
annual
mean
annual
mean
annual
mean
annual
mean
annual
mean
annual
mean
annual
mean
annual
mean
0.00463
0.00583
0.00943
0.00862
0.01310
0.00676
0.00812
0.00676
0.00121
0.00238
0.00198
0.00606
0.00296
0.00392
0.00389
0.00426
0.00389
-0.00209
0.00710
0.01044
0.01315
0.01484
0.02070
0.00976
0.01164
0.00976
0.00499
4A-10
-------
Table 4A-11. Study-Specific Information for PM10.2.5 Studies in Detroit, Ml
Study
Health Effect
ICD-9
Codes
Ages Model
Other Observed
Pollutants Concentrations
in Model min.
Hospital Admissions - Single
Ito (2003) [reanalysis
of Lippmann et al.
(2000)]
Ito (2003) [reanalysis
of Lippmann et al.
(2000)]
Ito (2003) [reanalysis
of Lippmann et al.
(2000)]
Ito (2003) [reanalysis
of Lippmann et al.
(2000)]
Ito (2003) [reanalysis
of Lippmann et al.
(2000)]
Pneumonia
COPD+
Ischemic
heart disease
Dysrhythmias
Congestive
heart failure
480-
486
490-
496
410-
414
427
428
log-linear,
65+ GAM
(stringent)
log-linear,
65+ GAM
(stringent)
log-linear,
65+ GAM
(stringent)
log-linear,
65+ GAM
(stringent)
log-linear,
65+ GAM
(stringent)
none
none
none
none
none
Lag
Exposure PM Coarse Lower
Metric Coefficient Bound
Upper
Bound
Pollutant Models
1
1
1
1
1
50
50
50
50
50
1 day
3 day
2 day
0 day
0 day
1-day avg 0.0037814 -0.0004188
1-day avg 0.0033223 -0.0019622
1-day avg 0.0038954 0.0009475
1-day avg 0.0000416 -0.0052791
1-day avg 0.0017142 -0.0016142
0.0079769
0.0085917
0.0068258
0.0053863
0.0050924
4A-11
-------
Table 4A-12. Study-Specific Information for PM10.25 Studies in Seattle, WA
Study
Health
Effect
ICD-9
Codes
Ages
Model
Other Observed
Pollutants Concentrations
in Model min.
Hospital Admissions -Single
Sheppard (2003)
(reanalysis of Sheppard
etal. (1999)*
Asthma
493
<65
log-linear,
GAM
(stringent)
none
Lag
Exposure
Metric
PM Coarse
Coefficient
Lower
Bound
Upper
Bound
Pollutant Models
N/A 88
1 day
1 -day avg
0.0021293
0.0000000
0.0052463
*Sheppard (2003) [reanalysis of Sheppard et al. (1999)] used daily PM2.5 values obtained from nephelometry measurements rather than from the difference between
PM2.5 and PM10 air quality monitors.
Table 4A-13. Study-Specific Information for Studies in St. Louis, MO
Study
Health
Effect
SSes A«es
Model
Respiratory Symptoms
Schwartz and Neas,
2000 - 6 cities
Schwartz and Neas,
2000 - 6 cities
Lower
respiratory
symptoms*
Cough*
N/A 7-14
N/A 7-14
logistic
logistic
Respiratory Symptoms
Schwartz and Neas,
2000 - 6 cities
Schwartz and Neas,
2000 - 6 cities
Lower
respiratory
symptoms*
Cough*
N/A 7-14
N/A 7-14
logistic
logistic
Other Observed
Pollutants Concentrations Lag ,'?°f!Jre
in Model min. Metnc
and Illnesses*
none
none
and Illnesses*
PM2.5
PM2.5
- Single Pollutant Models
0 121 0 day 3-day avg
0 121 Oday 3-day avg
- Multi-Pollutant Models
0 121 Oday 3-day avg
0 121 0 day 3-day avg
PM Coarse Lower Upper
Coefficient Bound Bound
0.0163785 -0.0025253 0.0633522
0.0227902 0.0084573 0.0375131
0.0060988 -0.0131701 0.0258768
0.0206893 0.0049026 0.0365837
*The C-R functions for lower respiratory symptoms and cough were calculated for the summer period April 1 through August 31.
4A-12
-------
Table 4B-1. Estimated Annual Mortality Associated with Short-Term Exposure to PlVfe.s When Alternative
Standards Are Just Met, Assuming Various Outpoint Levels*
Los Angeles, CA, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=2.5 ug/m3
292
(-37-612)
0.0%
292
(-37-612)
0.0%
269
(-34 - 564)
7.9%
228
(-28 - 476)
21 .9%
186
(-23 - 389)
36.3%
292
(-37-612)
0.0%
197
(-25-413)
32.5%
171
(-21 - 358)
41 .4%
145
(-18-302)
50.3%
118
(-15-247)
59.6%
269
(-34 - 562)
7.9%
269
(-34 - 562)
7.9%
228
(-28 - 476)
21 .9%
186
(-23 - 389)
36.3%
197
(-25-413)
32.5%
171
(-21 - 358)
41.4%
145
(-18-302)
50.3%
118
(-15-247)
59.6%
Cut point**
=10 ug/m3
115
(-14-240)
0.0%
115
(-14-240)
0.0%
96
(-12-200)
16.5%
65
(-8-135)
43.5%
39
(-5 - 80)
66.1%
115
(-14-240)
0.0%
45
(-6-94)
60.9%
30
(-4-63)
73.9%
18
(-2-37)
84.3%
9
(-1-18)
92.2%
96
(-12-199)
16.5%
96
(-12-199)
16.5%
65
(-8-135)
43.5%
39
(-5-80)
66.1%
45
(-6-94)
60.9%
30
(-4-63)
73.9%
18
(-2 - 37)
84.3%
9
(-1 - 18)
92.2%
Cutpoint**
=15 ug/m3
58
(-7-121)
0.0%
58
(-7-121)
0.0%
45
(-6-94)
22.4%
26
(-3 - 54)
55.2%
13
(-2 - 27)
77.6%
58
(-7-121)
0.0%
16
(-2 - 33)
72.4%
10
(-1 - 20)
82.8%
5
(-1 - 10)
91 .4%
2
(0-4)
96.6%
45
(-6 - 93)
22.4%
45
(-6 - 93)
22.4%
26
(-3 - 54)
55.2%
13
(-2-27)
77.6%
16
(-2 - 33)
72.4%
10
(-1 - 20)
82.8%
5
(-1-10)
91 .4%
2
(0-4)
96.6%
Cutpoint**
=20 ug/m3
29
(-4-61)
0.0%
29
(-4-61)
0.0%
22
(-3-46)
24.1%
12
(-2 - 25)
58.6%
5
(-1-11)
82.8%
29
(-4-61)
0.0%
7
(-1 - 14)
75.9%
3
(0-7)
89.7%
1
(0-3)
96.6%
0
(0-1)
100.0%
22
(-3-45)
24.1%
22
(-3 - 45)
24.1%
12
(-2 - 25)
58.6%
5
(-1-11)
82.8%
7
(-1 - 14)
75.9%
3
(0-7)
89.7%
1
(0-3)
96.6%
0
(0-1)
100.0%
4B-1
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=2.5 ug/m3
245
(-31 -513)
16.1%
245
(-31 -513)
16.1%
228
(-28 - 476)
21 .9%
186
(-23 - 389)
36.3%
197
(-25-413)
32.5%
171
(-21 - 358)
41 .4%
145
(-18-302)
50.3%
118
(-15-247)
59.6%
222
(-28 - 464)
24.0%
222
(-28 - 464)
24.0%
222
(-28 - 464)
24.0%
186
(-23 - 389)
36.3%
197
(-25-413)
32.5%
171
(-21 - 358)
41 .4%
145
(-18-302)
50.3%
118
(-15-247)
59.6%
Cut point**
=10 ug/m3
77
(-10-161)
33.0%
77
(-10-161)
33.0%
65
(-8-135)
43.5%
39
(-5-80)
66.1%
45
(-6-94)
60.9%
30
(-4-63)
73.9%
18
(-2-37)
84.3%
9
(-1-18)
92.2%
61
(-8-126)
47.0%
61
(-8-126)
47.0%
61
(-8-126)
47.0%
39
(-5-80)
66.1%
45
(-6-94)
60.9%
30
(-4-63)
73.9%
18
(-2-37)
84.3%
9
(-1-18)
92.2%
Cutpoint**
=15 ug/m3
34
(-4-69)
41.4%
34
(-4 - 69)
41 .4%
26
(-3 - 54)
55.2%
13
(-2 - 27)
77.6%
16
(-2 - 33)
72.4%
10
(-1 - 20)
82.8%
5
(-1 - 10)
91 .4%
2
(0-4)
96.6%
24
(-3-50)
58.6%
24
(-3 - 50)
58.6%
24
(-3 - 50)
58.6%
13
(-2-27)
77.6%
16
(-2 - 33)
72.4%
10
(-1 - 20)
82.8%
5
(-1 - 10)
91 .4%
2
(0-4)
96.6%
Cutpoint**
=20 ug/m3
16
(-2 - 33)
44.8%
16
(-2 - 33)
44.8%
12
(-2 - 25)
58.6%
5
(-1-11)
82.8%
7
(-1 - 14)
75.9%
3
(0-7)
89.7%
1
(0-3)
96.6%
0
(0-1)
100.0%
11
(-1 - 23)
62.1%
11
(-1 - 23)
62.1%
11
(-1 - 23)
62.1%
5
(-1-11)
82.8%
7
(-1 - 14)
75.9%
3
(0-7)
89.7%
1
(0-3)
96.6%
0
(0-1)
100.0%
"This analysis was performed using Moolgavkar (2003).
"For the outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model (see discussion in
section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percentsare rounded to the nearest tenth.
4B-2
-------
Table 4B-2. Estimated Annual Mortality Associated with Short-Term Exposure to PlVfe 5 When Alternative
Standards Are Just Met, Assuming Various Outpoint Levels*
Philadelphia, PA, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
367
(175-560)
0.0%
317
(151 -482)
13.6%
273
(130-416)
25.6%
230
(110-350)
37.3%
187
(89 - 284)
49.0%
297
(142-451)
19.1%
176
(84 - 268)
52.0%
152
(72-231)
58.6%
128
(61 -195)
65.1%
104
(49-158)
71.7%
336
(160-511)
8.4%
317
(151 -482)
13.6%
273
(130-416)
25.6%
230
(110-350)
37.3%
187
(89 - 284)
49.0%
176
(84 - 268)
52.0%
152
(72-231)
58.6%
128
(61 - 195)
65.1%
Cut point**
=10 ug/m3
189
(90 - 288)
0.0%
143
(68-218)
24.3%
106
(50-161)
43.9%
71
(34-108)
62.4%
41
(20 - 63)
78.3%
126
(60-191)
33.3%
35
(17-53)
81.5%
22
(11-34)
88.4%
12
(6-19)
93.7%
5
(2-8)
97.4%
160
(76 - 243)
15.3%
143
(68-218)
24.3%
106
(50-161)
43.9%
71
(34-108)
62.4%
41
(20 - 63)
78.3%
35
(17-53)
81.5%
22
(11-34)
88.4%
12
(6-19)
93.7%
Cutpoint**
=15 ug/m3
106
(51 -162)
0.0%
71
(34-107)
33.0%
45
(22 - 69)
57.5%
25
(12-38)
76.4%
11
(5-16)
89.6%
58
(28 - 89)
45.3%
8
(4-12)
92.5%
3
(2-5)
97.2%
1
(1-2)
99.1%
0
(0-0)
100.0%
83
(40-127)
21.7%
71
(34-107)
33.0%
45
(22 - 69)
57.5%
25
(12-38)
76.4%
11
(5-16)
89.6%
8
(4-12)
92.5%
3
(2-5)
97.2%
1
(1-2)
99.1%
Cutpoint**
=20 ug/m3
57
(27 - 87)
0.0%
34
(16-51)
40.4%
18
(9 - 28)
68.4%
7
(3-11)
87.7%
2
(1-3)
96.5%
26
(12-40)
54.4%
1
(1-2)
98.2%
0
(0-1)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
42
(20 - 63)
26.3%
34
(16-51)
40.4%
18
(9-28)
68.4%
7
(3-11)
87.7%
2
(1-3)
96.5%
1
(1-2)
98.2%
0
(0-1)
100.0%
0
(0-0)
100.0%
4B-2
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
104
(49-158)
71 .7%
304
(145-462)
17.2%
273
(130-416)
25.6%
230
(110-350)
37.3%
187
(89 - 284)
49.0%
176
(84 - 268)
52.0%
152
(72-231)
58.6%
128
(61 -195)
65.1%
104
(49-158)
71.7%
272
(130-414)
25.9%
272
(130-414)
25.9%
230
(110-350)
37.3%
187
(89 - 284)
49.0%
176
(84 - 268)
52.0%
152
(72-231)
58.6%
128
(61 - 195)
65.1%
Cut point**
=10 ug/m3
5
(2-8)
97.4%
132
(63 - 200)
30.2%
106
(50-161)
43.9%
71
(34-108)
62.4%
41
(20 - 63)
78.3%
35
(17-53)
81.5%
22
(11-34)
88.4%
12
(6-19)
93.7%
5
(2-8)
97.4%
104
(50-159)
45.0%
104
(50-159)
45.0%
71
(34-108)
62.4%
41
(20 - 63)
78.3%
35
(17-53)
81.5%
22
(11-34)
88.4%
12
(6-19)
93.7%
Cutpoint**
=15 ug/m3
0
(0-0)
100.0%
62
(30 - 95)
41 .5%
45
(22 - 69)
57.5%
25
(12-38)
76.4%
11
(5-16)
89.6%
8
(4-12)
92.5%
3
(2-5)
97.2%
1
(1-2)
99.1%
0
(0-0)
100.0%
44
(21 - 68)
58.5%
44
(21 - 68)
58.5%
25
(12-38)
76.4%
11
(5-16)
89.6%
8
(4-12)
92.5%
3
(2-5)
97.2%
1
(1-2)
99.1%
Cutpoint**
=20 ug/m3
0
(0-0)
100.0%
29
(14-44)
49.1%
18
(9-28)
68.4%
7
(3-11)
87.7%
2
(1-3)
96.5%
1
(1-2)
98.2%
0
(0-1)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
18
(9-27)
68.4%
18
(9-27)
68.4%
7
(3-11)
87.7%
2
(1-3)
96.5%
1
(1-2)
98.2%
0
(0-1)
100.0%
0
(0-0)
100.0%
"This analysis was performed using Lipfert et al. (2000).
**For the outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model (see discussion in
section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4B-4
-------
Table 4B-3. Estimated Annual Mortality Associated with Short-Term Exposure to PlVfe 5 When Alternative
Standards Are Just Met, Assuming Various Outpoint Levels*
Pittsburgh, PA, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
50
(-108-200)
0.0%
47
(-102-189)
6.0%
41
(-88-162)
18.0%
34
(-74-136)
32.0%
28
(-60-110)
44.0%
50
(-108-200)
0.0%
42
(-92-168)
16.0%
36
(-79-145)
28.0%
31
(-67-122)
38.0%
25
(-54 - 99)
50.0%
46
(-99-182)
8.0%
41
(-88-162)
18.0%
34
(-74-136)
32.0%
28
(-60-110)
44.0%
42
(-92 - 168)
16.0%
36
(-79-145)
28.0%
31
(-67-122)
38.0%
25
(-54 - 99)
50.0%
Cut point**
=10 ug/m3
22
(-48 - 87)
0.0%
19
(-43 - 77)
13.6%
14
(-31 - 56)
36.4%
9
(-21 - 37)
59.1%
5
(-12-20)
77.3%
22
(-48 - 87)
0.0%
15
(-34-61)
31.8%
11
(-24 - 43)
50.0%
7
(-15-27)
68.2%
4
(-8-14)
81.8%
18
(-40 - 72)
18.2%
14
(-31 - 56)
36.4%
9
(-21 - 37)
59.1%
5
(-12-20)
77.3%
15
(-34-61)
31.8%
11
(-24 - 43)
50.0%
7
(-15-27)
68.2%
4
(-8-14)
81.8%
Cutpoint**
=15 ug/m3
10
(-23-41)
0.0%
9
(-19-34)
10.0%
5
(-12-21)
50.0%
3
(-6-11)
70.0%
1
(-3 - 5)
90.0%
10
(-23-41)
0.0%
6
(-13-24)
40.0%
4
(-8-14)
60.0%
2
(-4 - 7)
80.0%
1
(-2 - 3)
90.0%
8
(-17-31)
20.0%
5
(-12-21)
50.0%
3
(-6-11)
70.0%
1
(-3 - 5)
90.0%
6
(-13-24)
40.0%
4
(-8-14)
60.0%
2
(-4 - 7)
80.0%
1
(-2 - 3)
90.0%
Cutpoint**
=20 ug/m3
5
(-11 -18)
0.0%
4
(-9-15)
20.0%
2
(-5-8)
60.0%
1
(-2-4)
80.0%
0
(-1 - 2)
100.0%
5
(-11-18)
0.0%
3
(-6-10)
40.0%
1
(-3-5)
80.0%
1
(-2 - 3)
80.0%
0
(-1-1)
100.0%
3
(-8-13)
40.0%
2
(-5-8)
60.0%
1
(-2 - 4)
80.0%
0
(-1 - 2)
100.0%
3
(-6-10)
40.0%
1
(-3-5)
80.0%
1
(-2 - 3)
80.0%
0
(-1-1)
100.0%
4B-5
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
41
(-90-165)
18.0%
41
(-88-162)
18.0%
34
(-74-136)
32.0%
28
(-60-110)
44.0%
41
(-90-165)
18.0%
36
(-79-145)
28.0%
31
(-67-122)
38.0%
25
(-54 - 99)
50.0%
37
(-80-147)
26.0%
37
(-80-147)
26.0%
34
(-74-136)
32.0%
28
(-60-110)
44.0%
37
(-80-147)
26.0%
36
(-79-145)
28.0%
31
(-67-122)
38.0%
25
(-54 - 99)
50.0%
Cut point**
=10 ug/m3
15
(-32 - 58)
31.8%
14
(-31 - 56)
36.4%
9
(-21 - 37)
59.1%
5
(-12-20)
77.3%
15
(-32 - 58)
31.8%
11
(-24 - 43)
50.0%
7
(-15-27)
68.2%
4
(-8-14)
81.8%
11
(-25 - 44)
50.0%
11
(-25 - 44)
50.0%
9
(-21 - 37)
59.1%
5
(-12-20)
77.3%
11
(-25 - 44)
50.0%
11
(-24 - 43)
50.0%
7
(-15-27)
68.2%
4
(-8-14)
81.8%
Cutpoint**
=15 ug/m3
6
(-13-22)
40.0%
5
(-12-21)
50.0%
3
(-6-11)
70.0%
1
(-3 - 5)
90.0%
6
(-13-22)
40.0%
4
(-8-14)
60.0%
2
(-4-7)
80.0%
1
(-2 - 3)
90.0%
4
(-8-15)
60.0%
4
(-8-15)
60.0%
3
(-6-11)
70.0%
1
(-3 - 5)
90.0%
4
(-8-15)
60.0%
4
(-8-14)
60.0%
2
(-4-7)
80.0%
1
(-2 - 3)
90.0%
Cutpoint**
=20 ug/m3
2
(-5 - 9)
60.0%
2
(-5-8)
60.0%
1
(-2-4)
80.0%
0
(-1 - 2)
100.0%
2
(-5 - 9)
60.0%
1
(-3-5)
80.0%
1
(-2-3)
80.0%
0
(-1-1)
100.0%
1
(-3 - 6)
80.0%
1
(-3-6)
80.0%
1
(-2-4)
80.0%
0
(-1 - 2)
100.0%
1
(-3 - 6)
80.0%
1
(-3-5)
80.0%
1
(-2-3)
80.0%
0
(-1-1)
100.0%
"This analysis was performed using Chock et al. (2000), age 75+ model.
"For the outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model (see discussion in
section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percentsare rounded to the nearest tenth.
4B-6
-------
Table 4B-4. Estimated Annual Mortality Associated with Short-Term Exposure to PlVy When Alternative
Standards Are Just Met, Assuming Various Cutpoint Levels*
St. Louis, MO, 2003
Alternative Standards
Annual (ug/m3)
15
14
Dally (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction In Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
191
(70-311)
0.0%
191
(70-311)
0.0%
190
(70-310)
0.5%
160
(59 - 260)
16.2%
130
(48-211)
31 .9%
191
(70-311)
0.0%
191
(70-311)
0.0%
172
(63 - 280)
9.9%
145
(53 - 235)
24.1%
118
(43-191)
38.2%
175
(64 - 284)
8.4%
175
(64 - 284)
8.4%
160
(59 - 260)
16.2%
130
(48-211)
31.9%
175
(64 - 284)
8.4%
172
(63 - 280)
9.9%
145
(53 - 235)
24.1%
118
(43-191)
38.2%
Cutpoint**
=10 ug/m3
75
(28-122)
0.0%
75
(28-122)
0.0%
75
(27-121)
0.0%
49
(18-80)
34.7%
28
(10-45)
62.7%
75
(28-122)
0.0%
75
(28- 122)
0.0%
59
(22 - 96)
21.3%
38
(14-62)
49.3%
20
(7 - 33)
73.3%
61
(22 - 99)
18.7%
61
(22 - 99)
18.7%
49
(18-80)
34.7%
28
(10-45)
62.7%
61
(22 - 99)
18.7%
59
(22 - 96)
21.3%
38
(14-62)
49.3%
20
(7 - 33)
73.3%
Cutpoint**
=15 ug/m3
29
(11 -46)
0.0%
29
(11 -46)
0.0%
28
(10-46)
3.4%
14
(5-23)
51 .7%
5
(2-8)
82.8%
29
(11 -46)
0.0%
29
(11-46)
0.0%
19
(7 - 31)
34.5%
9
(3-14)
69.0%
3
(1-4)
89.7%
20
(7 - 33)
31.0%
20
(7 - 33)
31 .0%
14
(5-23)
51.7%
5
(2-8)
82.8%
20
(7 - 33)
31.0%
19
(7-31)
34.5%
9
(3-14)
69.0%
3
(1-4)
89.7%
Cutpoint**
=20 ug/m3
9
(3-14)
0.0%
9
(3-14)
0.0%
8
(3-14)
11.1%
3
(1-4)
66.7%
1
(0-1)
88.9%
9
(3-14)
0.0%
9
(3-14)
0.0%
5
(2-7)
44.4%
2
(1-3)
77.8%
0
(0-1)
100.0%
5
(2-8)
44.4%
5
(2-8)
44.4%
3
(1-4)
66.7%
1
(0-1)
88.9%
5
(2-8)
44.4%
5
(2-7)
44.4%
2
(1-3)
77.8%
0
(0-1)
100.0%
4B-7
-------
Alternative Standards
Annual (ug/m3)
13
12
Dally (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction In Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m3
158
(58 - 256)
17.3%
158
(58 - 256)
17.3%
158
(58 - 256)
17.3%
130
(48-211)
31.9%
158
(58 - 256)
17.3%
158
(58 - 256)
1 7.3%
145
(53 - 235)
24.1%
118
(43-191)
38.2%
141
(52 - 229)
26.2%
141
(52 - 229)
26.2%
141
(52 - 229)
26.2%
130
(48-211)
31.9%
141
(52 - 229)
26.2%
141
(52 - 229)
26.2%
141
(52 - 229)
26.2%
118
(43-191)
38.2%
Cutpoint**
=10 ug/m3
47
(17-77)
37.3%
47
(17-77)
37.3%
47
(17-77)
37.3%
28
(10-45)
62.7%
47
(17-77)
37.3%
47
(17-77)
37.3%
38
(14-62)
49.3%
20
(7 - 33)
73.3%
35
(13-57)
53.3%
35
(13-57)
53.3%
35
(13-57)
53.3%
28
(10-45)
62.7%
35
(13-57)
53.3%
35
(13-57)
53.3%
35
(13-57)
53.3%
20
(7 - 33)
73.3%
Cutpoint**
=15 ug/m3
13
(5-21)
55.2%
13
(5-21)
55.2%
13
(5-21)
55.2%
5
(2-8)
82.8%
13
(5 - 21)
55.2%
13
(5 - 21)
55.2%
9
(3-14)
69.0%
3
(1-4)
89.7%
8
(3-12)
72.4%
8
(3-12)
72.4%
8
(3-12)
72.4%
5
(2-8)
82.8%
8
(3-12)
72.4%
8
(3-12)
72.4%
8
(3-12)
72.4%
3
(1-4)
89.7%
Cutpoint**
=20 ug/m3
3
(1-4)
66.7%
3
(1-4)
66.7%
3
(1-4)
66.7%
1
(0-1)
88.9%
3
(1-4)
66.7%
3
(1-4)
66.7%
2
(1-3)
77.8%
0
(0-1)
100.0%
1
(1-2)
88.9%
1
(1-2)
88.9%
1
(1-2)
88.9%
1
(0-1)
88.9%
1
(1-2)
88.9%
1
(1-2)
88.9%
1
(1-2)
88.9%
0
(0-1)
100.0%
*This analysis was performed using Schwartz (2003b).
**Forthe outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model (see discussion in section 4.3.2.1).
***Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4B-8
-------
Table 4B-5. Estimated Annual Mortality Associated with Long-Term Exposure to PM2.5 When
Alternative Standards Are Just Met, Assuming Various Outpoint Levels*
Los Angeles, CA, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
1507
(531 - 2587)
0.0%
1507
(531 - 2587)
0.0%
1265
(446 - 21 68)
16.1%
829
(293-1416)
45.0%
396
(140-675)
73.7%
1507
(531 - 2587)
0.0%
514
(182-876)
65.9%
240
(85 - 408)
84.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
1259
(444 - 21 58)
16.5%
1259
(444 - 21 58)
16.5%
829
(293-1416)
45.0%
396
(140-675)
73.7%
514
(182-876)
65.9%
240
(85 - 408)
84.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=10 ug/m3
823
(290-1415)
0.0%
823
(290-1415)
0.0%
553
(195-949)
32.8%
65
(23-111)
92.1%
0
(0-0)
100.0%
823
(290-1415)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
546
(192-937)
33.7%
546
(192-937)
33.7%
65
(23-111)
92.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
138
(48 - 237)
0.0%
138
(48 - 237)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
138
(48 - 237)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
4B-9
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2 5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
1013
(358-1732)
32.8%
1013
(358-1732)
32.8%
829
(293-1416)
45.0%
396
(140-675)
73.7%
514
(182-876)
65.9%
240
(85 - 408)
84.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
767
(271 -1310)
49.1%
767
(271 -1310)
49.1%
767
(271 -1310)
49.1%
396
(140-675)
73.7%
514
(182-876)
65.9%
240
(85 - 408)
84.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=10 ug/m3
270
(95 - 463)
67.2%
270
(95 - 463)
67.2%
65
(23-111)
92.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
This analysis was performed using Pope et al. (2002) - ACS extended.
**For the outpoints above policy relevant background, the slope of the C-R function has been modified based on a simple hockeystick model (see discussion in
section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4B-10
-------
Table 4B-6. Estimated Annual Mortality Associated with Long-Term Exposure to PM2.5 When
Alternative Standards Are Just Met, Assuming Various Outpoint Levels*
Philadelphia, PA, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
536
(185-943)
0.0%
408
(141 -716)
23.9%
299
(104-524)
44.2%
191
(67 - 334)
64.4%
84
(29-146)
84.3%
357
(124-626)
33.4%
58
(20-101)
89.2%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
456
(157-799)
14.9%
408
(141 -716)
23.9%
299
(104-524)
44.2%
191
(67 - 334)
64.4%
84
(29 - 1 46)
84.3%
58
(20-101)
89.2%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=10 ug/m3
338
(1 1 6 - 597)
0.0%
194
(67-341)
42.6%
72
(25-126)
78.7%
0
(0-0)
100.0%
0
(0-0)
100.0%
137
(47-241)
59.5%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
247
(85 - 435)
26.9%
194
(67-341)
42.6%
72
(25-126)
78.7%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
137
(47 - 244)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
37
(13-65)
73.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
4B-11
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2 5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
0
(0-0)
100.0%
375
(130-657)
30.0%
299
(104-524)
44.2%
191
(67 - 334)
64.4%
84
(29-146)
84.3%
58
(20-101)
89.2%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
295
(102-516)
45.0%
295
(102-516)
45.0%
191
(67 - 334)
64.4%
84
(29 - 1 46)
84.3%
58
(20-101)
89.2%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=10 ug/m3
0
(0-0)
100.0%
157
(54 - 276)
53.6%
72
(25-126)
78.7%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
67
(23 - 1 1 8)
80.2%
67
(23 - 1 1 8)
80.2%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
This analysis was performed using Pope et al. (2002) - ACS extended.
**For the outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model
(see discussion in section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4B-12
-------
Table 4B-7. Estimated Annual Mortality Associated with Long-Term Exposure to PM2.5 When
Alternative Standards Are Just Met, Assuming Various Outpoint Levels*
Pittsburgh, PA, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
403
(141 -699)
0.0%
361
(126-626)
10.4%
264
(93 - 456)
34.5%
168
(59 - 289)
58.3%
72
(25-124)
82.1%
403
(141 -699)
0.0%
287
(100-495)
28.8%
200
(70 - 345)
50.4%
114
(40 - 1 97)
71 .7%
29
(10-50)
92.8%
338
(118-585)
16.1%
264
(93 - 456)
34.5%
168
(59 - 289)
58.3%
72
(25 - 1 24)
82.1%
287
(1 00 - 495)
28.8%
200
(70 - 345)
50.4%
114
(40 - 1 97)
71 .7%
29
(10-50)
92.8%
Cutpoint**
=10 ug/m3
215
(75 - 373)
0.0%
168
(58-291)
21 .9%
59
(21 -102)
72.6%
0
(0-0)
100.0%
0
(0-0)
100.0%
215
(75 - 373)
0.0%
84
(29-145)
60.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
141
(49 - 245)
34.4%
59
(21 -102)
72.6%
0
(0-0)
100.0%
0
(0-0)
100.0%
84
(29 - 1 45)
60.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
25
(9 - 43)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
25
(9 - 43)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
4B-13
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2 5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
273
(96-471)
32.3%
264
(93 - 456)
34.5%
168
(59 - 289)
58.3%
72
(25-124)
82.1%
273
(96-471)
32.3%
200
(70 - 345)
50.4%
114
(40 - 1 97)
71 .7%
29
(10-50)
92.8%
208
(73 - 358)
48.4%
208
(73 - 358)
48.4%
168
(59 - 289)
58.3%
72
(25 - 1 24)
82.1%
208
(73 - 358)
48.4%
200
(70 - 345)
50.4%
114
(40 - 1 97)
71 .7%
29
(10-50)
92.8%
Cutpoint**
=10 ug/m3
68
(24 - 1 1 8)
68.4%
59
(21 -102)
72.6%
0
(0-0)
100.0%
0
(0-0)
100.0%
68
(24 - 1 1 8)
68.4%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
This analysis was performed using Pope et al. (2002) - ACS extended.
**For the outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model
(see discussion in section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4B-14
-------
Table 4B-8. Estimated Annual Mortality Associated with Long-Term Exposure to PM2.5 When
Alternative Standards Are Just Met, Assuming Various Outpoint Levels*
St. Louis, MO, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2.5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
596
(206-1047)
0.0%
596
(206-1047)
0.0%
592
(204-1039)
0.7%
414
(144-726)
30.5%
239
(83 - 41 7)
59.9%
596
(206-1047)
0.0%
596
(206-1047)
0.0%
486
(168-853)
18.5%
327
(113-571)
45.1%
168
(58 - 293)
71 .8%
498
(172-874)
16.4%
498
(172-874)
16.4%
414
(144-726)
30.5%
239
(83 - 41 7)
59.9%
498
(172-874)
16.4%
486
(168-853)
18.5%
327
(113-571)
45.1%
168
(58 - 293)
71 .8%
Cutpoint**
=10 ug/m3
311
(107-548)
0.0%
311
(107-548)
0.0%
306
(105-539)
1.6%
107
(37-188)
65.6%
0
(0-0)
100.0%
311
(107-548)
0.0%
311
(107-548)
0.0%
188
(65 - 330)
39.5%
8
(3-15)
97.4%
0
(0-0)
100.0%
201
(69 - 354)
35.4%
201
(69 - 354)
35.4%
107
(37-188)
65.6%
0
(0-0)
100.0%
201
(69 - 354)
35.4%
188
(65 - 330)
39.5%
8
(3-15)
97.4%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
23
(8 - 40)
0.0%
23
(8 - 40)
0.0%
17
(6 - 30)
26.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
23
(8 - 40)
0.0%
23
(8 - 40)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
4B-15
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM2 5
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
401
(139-702)
32.7%
401
(139-702)
32.7%
401
(139-702)
32.7%
239
(83 - 41 7)
59.9%
401
(139-702)
32.7%
401
(139-702)
32.7%
327
(113-571)
45.1%
168
(58 - 293)
71 .8%
304
(106-532)
49.0%
304
(106-532)
49.0%
304
(106-532)
49.0%
239
(83 - 41 7)
59.9%
304
(106-532)
49.0%
304
(106-532)
49.0%
304
(106-532)
49.0%
168
(58 - 293)
71 .8%
Cutpoint**
=10 ug/m3
92
(32-162)
70.4%
92
(32-162)
70.4%
92
(32-162)
70.4%
0
(0-0)
100.0%
92
(32-162)
70.4%
92
(32-162)
70.4%
8
(3-15)
97.4%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
This analysis was performed using Pope et al. (2002) - ACS extended.
**For the outpoints above policy relevant background, the slope of the concentration-response function has been modified based on a simple hockeystick model
(see discussion in section 4.3.2.1).
•"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tenth.
4B-16
-------
Table 4B-9. Sensitivity Analysis: Estimated Annual Mortality Associated with Short-Term Exposure to PIV2.5 When Alternative Standards Are Just Met, Assuming Various Outpoint Levels - Rollbacks
to Meet Annual Standards Using Design Values Based on Maximum vs. Average of Monitor-Specific Averages*
Detroit, Ml, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
65, 98th percentile value***
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM26 Using an Annual Design Value Based on the Maximum
of Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m°
122
(-123-358)
0.0%
122
(-123-358)
0.0%
122
(-123-358)
0.0%
111
(-112-325)
9.0%
90
(-91 - 263)
26.2%
122
(-123-358)
0.0%
122
(-123-358)
0.0%
120
(-121 -352)
1.6%
101
(-102-296)
17.2%
82
(-83 - 239)
32.8%
111
(-112-326)
9.0%
111
(-112-326)
9.0%
111
(-112-325)
9.0%
90
(-91 - 263)
26.2%
111
(-112-326)
9.0%
111
(-112-326)
9.0%
101
(-102-296)
17.2%
82
(-83 - 239)
32.8%
Cutpoint**
=10 ug/m°
54
(-55-159)
0.0%
54
(-55-159)
0.0%
54
(-55-159)
0.0%
45
(-45-131)
16.7%
28
(-29 - 82)
48.1%
54
(-55-159)
0.0%
54
(-55-159)
0.0%
53
(-53-154)
1.9%
37
(-37-107)
31.5%
22
(-23 - 65)
59.3%
45
(-46-132)
16.7%
45
(-46-132)
16.7%
45
(-45-131)
16.7%
28
(-29 - 82)
48.1%
45
(-46-132)
16.7%
45
(-46-132)
16.7%
37
(-37-107)
31.5%
22
(-23 - 65)
59.3%
Cutpoint**
=15 ug/m°
26
(-27 - 77)
0.0%
26
(-27 - 77)
0.0%
26
(-27 - 77)
0.0%
20
(-20 - 58)
23.1%
10
(-10-28)
61.5%
26
(-27 - 77)
0.0%
26
(-27 - 77)
0.0%
25
(-26 - 74)
3.8%
15
(-15-42)
42.3%
7
(-7-19)
73.1%
20
(-20 - 58)
23.1%
20
(-20 - 58)
23.1%
20
(-20 - 58)
23.1%
10
(-10-28)
61.5%
20
(-20 - 58)
23.1%
20
(-20 - 58)
23.1%
15
(-15-42)
42.3%
7
(-7-19)
73.1%
Cutpoint**
=20 ug/m°
12
(-12-35)
0.0%
12
(-12-35)
0.0%
12
(-12-35)
0.0%
8
(-9 - 24)
33.3%
3
(-4-10)
75.0%
12
(-12-35)
0.0%
12
(-12-35)
0.0%
11
(-12-33)
8.3%
6
(-6-16)
50.0%
2
(-2 - 6)
83.3%
8
(-9 - 24)
33.3%
8
(-9 - 24)
33.3%
8
(-9 - 24)
33.3%
3
(-4-10)
75.0%
8
(-9 - 24)
33.3%
8
(-9 - 24)
33.3%
6
(-6-16)
50.0%
2
(-2 - 6)
83.3%
Incidence Associated with PM26 Using an Annual Design Value Based on the Average of
Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/m°
150
(-151 - 442)
0.0%
150
(-151 - 442)
0.0%
132
(-133-388)
12.0%
111
(-112-325)
26.0%
90
(-91 - 263)
40.0%
150
(-151 - 442)
0.0%
139
(-140-409)
7.3%
120
(-121 - 352)
20.0%
101
(-102-296)
32.7%
82
(-83 - 239)
45.3%
137
(-138-403)
8.7%
132
(-133-388)
12.0%
111
(-112-325)
26.0%
90
(-91 - 263)
40.0%
137
(-138-403)
8.7%
120
(-121 - 352)
20.0%
101
(-102-296)
32.7%
82
(-83 - 239)
45.3%
Cutpoint**
=10 ug/m°
80
(-81 - 236)
0.0%
80
(-81 - 236)
0.0%
63
(-64- 186)
21.3%
45
(-45-131)
43.8%
28
(-29 - 82)
65.0%
80
(-81 - 236)
0.0%
70
(-70 - 206)
12.5%
53
(-53-154)
33.8%
37
(-37-107)
53.8%
22
(-23 - 65)
72.5%
68
(-68 - 200)
15.0%
63
(-64-186)
21.3%
45
(-45-131)
43.8%
28
(-29 - 82)
65.0%
68
(-68 - 200)
15.0%
53
(-53-154)
33.8%
37
(-37-107)
53.8%
22
(-23 - 65)
72.5%
Cutpoint**
=15 ug/m°
46
(-47-137)
0.0%
46
(-47- 137)
0.0%
33
(-33 - 97)
28.3%
20
(-20 - 58)
56.5%
10
(-10-28)
78.3%
46
(-47- 137)
0.0%
38
(-39- 112)
17.4%
25
(-26 - 74)
45.7%
15
(-15-42)
67.4%
7
(-7-19)
84.8%
37
(-37- 108)
19.6%
33
(-33 - 97)
28.3%
20
(-20 - 58)
56.5%
10
(-10-28)
78.3%
37
(-37- 108)
19.6%
25
(-26 - 74)
45.7%
15
(-15-42)
67.4%
7
(-7-19)
84.8%
Cutpoint**
=20 ug/m°
25
(-26 - 75)
0.0%
25
(-26 - 75)
0.0%
16
(-17-47)
36.0%
8
(-9-24)
68.0%
(-4-10)
88.0%
25
(-26 - 75)
0.0%
20
(-20 - 58)
20.0%
11
(-12-33)
56.0%
(-6-16)
76.0%
2
(-2 - 6)
92.0%
19
(-19-55)
24.0%
16
(-17-47)
36.0%
8
(-9-24)
68.0%
3
(-4-10)
88.0%
19
(-19-55)
24.0%
11
(-12-33)
56.0%
(-6-16)
76.0%
2
(-2 - 6)
92.0%
4B-17
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 9Bth percentile value
35, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM26 Using an Annual Design Value Based on the Maximum
of Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/mj
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
(-101 -295)
17.2%
90
(-91 - 263)
26.2%
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
101
(-101 -295)
17.2%
82
(-83 - 239)
32.8%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 263)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
90
(-91 - 264)
26.2%
82
(-83 - 239)
32.8%
Cutpoint**
=10 ug/m J
3B
(-37-106)
33.3%
36
(-37-106)
33.3%
(-37-106)
33.3%
28
(-29 - 82)
48.1%
36
(-37-106)
33.3%
36
(-37-106)
33.3%
36
(-37-106)
33.3%
22
(-23 - 65)
59.3%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
28
(-29 - 82)
48.1%
22
(-23 - 65)
59.3%
Cutpoint**
=15 ug/m J
14
(-15-42)
46.2%
14
(-15-42)
46.2%
(-15-42)
46.2%
10
(-10-28)
61.5%
14
(-15-42)
46.2%
14
(-15-42)
46.2%
14
(-15-42)
46.2%
7
(-7-19)
73.1%
10
(-10-28)
61.5%
10
(-10-28)
61.5%
10
(-10-28)
61.5%
10
(-10-28)
61.5%
10
(-10-28)
61.5%
10
(-10-28)
61.5%
10
(-10-28)
61.5%
(-7-19)
73.1%
Cutpoint**
=20 ug/mj
(-6-16)
50.0%
6
(-6-16)
50.0%
(-6-16)
50.0%
3
(-4-10)
75.0%
6
(-6-16)
50.0%
6
(-6-16)
50.0%
6
(-6-16)
50.0%
(-2 - 6)
83.3%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
3
(-4-10)
75.0%
(-2 - 6)
83.3%
Incidence Associated with PM26 Using an Annual Design Value Based on the Average of
Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Policy Relevant
Background
=3.5 ug/mj
124
(-125-364)
17.3%
124
(-125-364)
17.3%
(-112-325)
26.0%
90
(-91 - 263)
40.0%
124
(-125-364)
17.3%
120
(-121 - 352)
20.0%
101
(-102-296)
32.7%
82
(-83 - 239)
45.3%
111
(-112-325)
26.0%
111
(-112-325)
26.0%
111
(-112-325)
26.0%
90
(-91 - 263)
40.0%
111
(-112-325)
26.0%
111
(-112-325)
26.0%
101
(-102-296)
32.7%
82
(-83 - 239)
45.3%
Cutpoint**
=10 ug/m J
bB
(-57-165)
30.0%
56
(-57-165)
30.0%
(-45- 131)
43.8%
28
(-29 - 82)
65.0%
56
(-57-165)
30.0%
53
(-53- 154)
33.8%
37
(-37- 107)
53.8%
22
(-23 - 65)
72.5%
45
(-45-131)
43.8%
45
(-45- 131)
43.8%
45
(-45- 131)
43.8%
28
(-29 - 82)
65.0%
45
(-45-131)
43.8%
45
(-45- 131)
43.8%
37
(-37- 107)
53.8%
22
(-23 - 65)
72.5%
Cutpoint**
=15 ug/m J
2B
(-28-81)
39.1%
28
(-28-81)
39.1%
(-20 - 58)
56.5%
10
(-10-28)
78.3%
28
(-28-81)
39.1%
25
(-26 - 74)
45.7%
15
(-15-42)
67.4%
7
(-7-19)
84.8%
20
(-20 - 58)
56.5%
20
(-20 - 58)
56.5%
20
(-20 - 58)
56.5%
10
(-10-28)
78.3%
20
(-20 - 58)
56.5%
20
(-20 - 58)
56.5%
15
(-15-42)
67.4%
(-7-19)
84.8%
Cutpoint**
=20 ug/mj
13
(-13-38)
48.0%
13
(-13-38)
48.0%
(-9-24)
68.0%
3
(-4-10)
88.0%
13
(-13-38)
48.0%
11
(-12-33)
56.0%
6
(-6-16)
76.0%
2
(-2-6)
92.0%
8
(-9-24)
68.0%
8
(-9-24)
68.0%
8
(-9-24)
68.0%
3
(-4-10)
88.0%
8
(-9-24)
68.0%
8
(-9-24)
68.0%
6
(-6-16)
76.0%
(-2-6)
92.0%
'This analysis was performed using Ito (2003).
"For the cutpoints above policy relevant background, the slope ol
'"Current standards.
Note: Incidences are rounded to the nearest whole number; perci
if the concentration-response function has been modified based on a simple hockeystick model (see discussion in section 4.3.2.1).
mts are rounded to the nearest tenth.
4B-18
-------
Table 4B-10. Sensitivity Analysis: Estimated Annual Mortality Associated with Long-Term Exposure to PIV2.5 When Alternative Standards Are Just Met, Assuming Various Outpoint Levels - Rollbacks
to Meet Annual Standards Using Design Values Based on Maximum vs. Average of Monitor-Specific Averages*
Detroit, Ml, 2003
Alternative Standards
Annual (ug/m3)
15
14
Daily (ug/m3)
bb, 9Bth percentile value"""
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
65, 99th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM26 Using an Annual Design Value Based
on the Maximum of Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
b22
(181 -910)
0.0%
522
(181 -910)
0.0%
522
(181 -910)
0.0%
435
(151 -757)
16.7%
270
(94 - 468)
48.3%
522
(181 -910)
0.0%
522
(181 -910)
0.0%
507
(176-884)
2.9%
356
(124-619)
31.8%
207
(72 - 358)
60.3%
43B
(152-762)
16.1%
438
(152-762)
16.1%
435
(151 -757)
16.7%
270
(94 - 468)
48.3%
438
(152-762)
16.1%
438
(152-762)
16.1%
356
(124-619)
31.8%
207
(72 - 358)
60.3%
Cutpoint**
=10 ug/m3
282
(98 - 494)
0.0%
282
(98 - 494)
0.0%
282
(98 - 494)
0.0%
185
(64 - 323)
34.4%
0
(0-0)
100.0%
282
(98 - 494)
0.0%
282
(98 - 494)
0.0%
266
(92 - 465)
5.7%
97
(34-168)
65.6%
0
(0-0)
100.0%
188
(65 - 328)
33.3%
188
(65 - 328)
33.3%
185
(64 - 323)
34.4%
0
(0-0)
100.0%
188
(65 - 328)
33.3%
188
(65 - 328)
33.3%
97
(34-168)
65.6%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
41
(14-72)
0.0%
41
(14-72)
0.0%
41
(14-72)
0.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
41
(14-72)
0.0%
41
(14-72)
0.0%
23
(8 - 40)
43.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Incidence Associated with PM26 Using an Annual Design Value Based on the Average of
Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
/4/
(259-1309)
0.0%
747
(259-1309)
0.0%
602
(209-1051)
19.4%
435
(151 -757)
41.8%
270
(94 - 468)
63.9%
747
(259-1309)
0.0%
659
(229-1153)
11.8%
507
(176-884)
32.1%
356
(124-619)
52.3%
207
(72 - 358)
72.3%
B42
(223-1123)
14.1%
602
(209-1051)
19.4%
435
(151 -757)
41.8%
270
(94 - 468)
63.9%
642
(223-1123)
14.1%
507
(176-884)
32.1%
356
(124-619)
52.3%
207
(72 - 358)
72.3%
Cutpoint**
=10 ug/m3
b3b
(185-941)
0.0%
535
(185-941)
0.0%
372
(129-652)
30.5%
185
(64 - 323)
65.4%
0
(0-0)
100.0%
535
(185-941)
0.0%
437
(151 -766)
18.3%
266
(92 - 465)
50.3%
97
(34- 168)
81.9%
0
(0-0)
100.0%
41 B
(144-733)
21.9%
372
(129-652)
30.5%
185
(64 - 323)
65.4%
0
(0-0)
100.0%
418
(144-733)
21.9%
266
(92 - 465)
50.3%
97
(34- 168)
81.9%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
322
(1 1 1 - 568)
0.0%
322
(1 1 1 - 568)
0.0%
140
(48 - 247)
56.5%
0
(0-0)
100.0%
0
(0-0)
100.0%
322
(1 1 1 - 568)
0.0%
212
(73 - 374)
34.2%
23
(8 - 40)
92.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
191
(66 - 336)
40.7%
140
(48 - 247)
56.5%
0
(0-0)
100.0%
0
(0-0)
100.0%
191
(66 - 336)
40.7%
23
(8 - 40)
92.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
4B-19
-------
Alternative Standards
Annual (ug/m3)
13
12
Daily (ug/m3)
40, 98th percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
40, yath percentile value
35, 98th percentile value
30, 98th percentile value
25, 98th percentile value
40, 99th percentile value
35, 99th percentile value
30, 99th percentile value
25, 99th percentile value
Incidence Associated with PM26 Using an Annual Design Value Based
on the Maximum of Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
354
(123-615)
32.2%
354
(123-615)
32.2%
354
(123-615)
32.2%
270
(94 - 468)
48.3%
354
(123-615)
32.2%
354
(123-615)
32.2%
354
(123-615)
32.2%
207
(72 - 358)
60.3%
2/1
(94 - 469)
48.1%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
270
(94 - 468)
48.3%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
271
(94 - 469)
48.1%
207
(72 - 358)
60.3%
Cutpoint**
=10 ug/m3
94
(33-164)
66.7%
94
(33-164)
66.7%
94
(33-164)
66.7%
0
(0-0)
100.0%
94
(33-164)
66.7%
94
(33-164)
66.7%
94
(33-164)
66.7%
0
(0-0)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-0)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-1)
100.0%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
Incidence Associated with PM26 Using an Annual Design Value Based on the Average of
Monitor-Specific Averages**
(95% Confidence Interval)
Percent Reduction in Incidence from Current Standards
Cutpoint**
=7.5 ug/m3
538
(187-939)
28.0%
538
(187-939)
28.0%
435
(151 -757)
41.8%
270
(94 - 468)
63.9%
538
(187-939)
28.0%
507
(176-884)
32.1%
356
(124-619)
52.3%
207
(72 - 358)
72.3%
43b
(151 -756)
41.8%
435
(151 -756)
41.8%
435
(151 -756)
41.8%
270
(94 - 468)
63.9%
435
(151 -756)
41.8%
435
(151 -756)
41.8%
356
(124-619)
52.3%
207
(72 - 358)
72.3%
Cutpoint**
=10 ug/m3
301
(104-526)
43.7%
301
(104-526)
43.7%
185
(64 - 323)
65.4%
0
(0-0)
100.0%
301
(104-526)
43.7%
266
(92 - 465)
50.3%
97
(34- 168)
81.9%
0
(0-0)
100.0%
154
(64 - 322)
65.6%
184
(64 - 322)
65.6%
184
(64 - 322)
65.6%
0
(0-0)
100.0%
184
(64 - 322)
65.6%
184
(64 - 322)
65.6%
97
(34-168)
81.9%
0
(0-0)
100.0%
Cutpoint**
=12 ug/m3
61
(21 - 107)
81.1%
61
(21 - 107)
81.1%
0
(0-0)
100.0%
0
(0-0)
100.0%
61
(21 - 107)
81.1%
23
(8 - 40)
92.9%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
0
(0-0)
100.0%
'This analysis was performed using Pope et al. (2002) - ACS extended.
"For the cutpoints above policy relevant background, the slope of the concentration-response func
'"Current standards.
Note: Incidences are rounded to the nearest whole number; percents are rounded to the nearest tt
n has been modified basec
mple hockeystick model (se
4B-20
-------
Incidence Rate: 25 (9-44) deaths/yr/100,000
Incidence: 520(180-910)deaths/yr
Incidence Rate: 16 (6-27) deaths/yr/100,000
Incidence: 1500 (530-2600) deaths/yr
Incidence Rate: 35 (12-62) deaths/yr/100,000
Incidence: 540 (190-940) deaths/yr
24-hour
standard (ug/m )
(ug/m )
Philadelphia
Incidence Rate: 31 (11-55) deaths/yr/100,000
Incidence: 400 (140-700) deaths/yr
24-hour
standard (ug/m )
(ug/m )
Pittsburgh
Incidence Rate: 24(8-42)deaths/yr/100,000
Incidence: 600 (210-1100) deaths/yr
Figure 5A-l(a) Estimated percent reduction in PM2.5-related long-term mortality risk for
alternative standards (99th percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 7.5 ng/nf). Risk
associated with meeting current PM2.s standards, based on ACS extended
study, is shown in figures in terms of estimated annual incidence rate and
annual incidence (and 95% confidence ranges).
5A-1
-------
NOTE: Consistent with Figures 5-l(a), (b), and (c) and Figures 5-2(a), (b), (c),
and (d) that appear in Chapter 5, the figures in this Appendix are based on risk
reductions estimated for various combinations of alternative standards, including
annual standards of 15, 14, 13, and 12 |ig/m3 and 24-hour standards of 65, 40, 35,
30, and 25 |ig/m3. Since in most cases the estimated risk reductions were the
same or nearly so for alternative 24-hour standards of 65 and 40 i-ig/m3, the
interim points between these two levels were filled in so as to better depict the 3-
dimensional surface of risk reductions. In those cases involving the 99th
percentile form of the 24-hour standard where there were appreciable differences
in the estimated risk reductions between alternative 24-standards of 65 and 40
l-ig/m3, the interim points were not filled in, since the shape of the surface between
these points has not been calculated.
-------
Incidence rate: 14 (5-24) deaths/yr/100,000
Incidence: 280 (100-490) deaths/yr
10th
percent QQ|
reduction
(ug/m3)
6$
Detroit
24-hour
standard (ug/m )
Incidence rate: 9 (3-15) deaths/yr/100,000
Incidence: 820 (290-1400) deaths
10O,
Annual
standard
(ug/m3)
24-hour
standard (ug/m )
Los Angeles
Incidence rate: 22 (8-39) deaths/yr/100,000
Incidence: 340 (120-600) deaths/yr
100'
percent QQ.J.
reduction
standard
(ug/m )
24-hour
standard (ug/m )
Philadelphia
Incidence rate: 17 (6-29) deaths/yr/100,000
Incidence: 220 (80-370) deaths/a
24-hour
standard (ug/m )
(ug/m3)
Pittsburgh
Incidence rate: 12 (4-22) deaths/yr/100,000
Incidence: 310 (110-550) deaths/yr
100
24-hour
standard (ug/m)
(ug/m )
St. Louis
Figure 5A-l(b) Estimated percent reduction in PM2.5-related long-term mortality risk for
alternative standards (98?h percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 10 ng/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended
study, is shown in figures in terms of estimated annual incidence rate and
annual incidence (and 95% confidence ranges).
5A-2
-------
Incidence rate: 2 (1-3) deaths/yr/100,000
Incidence: 40 (10-70) deaths/yr
Incidence rate: 1 (1-2) deaths/yr/100,000
Incidence: 140 (50-240) deaths/yr
100
Los Angeles
Incidence rate: 9 (3-16) deaths/yr/100,000
Incidence: 140 (50-240) deaths/yr
24-hour
standard (ug/m )
(ug/m)
Philadelphia
Incidence rate: 2 (1-3) deaths/yr/100,000
Incidence: 30 (10-40) deaths/yr]
standard
(ug/m3)
24-hour
standard (ug/m )
Pittsburgh
Incidence rate: 1 (0-2) deaths/yr/100,000
Incidence: 20 (10-40) deaths/yr]
100
80-
Estimated
percent en. -
reduction
24-hour
standard (ug/m )
(ug/m )
St. Louis
Figure 5A-l(c) Estimated percent reduction in rM2.5-related long-term mortality risk for
alternative standards (9&h percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 12 ug/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended
study, is shown in figures in terms of estimated annual incidence rate and
annual incidence (and 95% confidence ranges).
5A-3
-------
Incidence Rate: 6 (-6 -17) deaths/yr/100,000
Incidence: 120 (-120 - 360) deaths/yr
100
standard (ug/m3)
(ug/m )
Detroit
Incidence Rate: 24 (12 - 37) deaths/yr/100,000
Incidence: 370 (180-560)deaths/\
80'
Estimated
percent 60
reduction
rd (ug/m )
(ug/m )
Philadelphia
Incidence Rate: 3 (0 -6) deaths/yr/100,000
Incidence: 290 (-40-610) deaths/yr
0
Annual K O ^ ~^ - 60
standard K ife °$
(ug/m3)
24-hour
standard (ug/m3)
Los Angeles
Incidence Rate: 4 (-8 -16) deaths/yr/100,000
Incidence: 50 (-110-200)deaths/yj
65
Pittsburgh
24-hour
standard (ug/m )
Incidence Rate: 8 (3 -12) deaths/yr/100,000
Incidence: 190 (70-310) deaths/yr
Figure 5-2(a) Estimated percent reduction in PM2.5-related short-term mortality risk for
alternative standards (9ff percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint equal to policy-
relevant background). Risk associated with meeting current PM2.5 standards,
based on ACS extended study, is shown in figures in terms of estimated annual
incidence rate and annual incidence (and 95% confidence ranges). Estimated
policy-relevant background is 3.5 |ig/m in eastern cities and 2.5 |ig/m in
western cities.
5A-4
-------
Incidence Rate: 3 (-3 - 8) deaths/yr/100,000
Incidence: 50 (-60 -160) deaths/yr
Incidence Rate: 1 (0 - 3) deaths/yr/100,000
Incidence: 120 (-10-240) deaths/yr
10Ch
Incidence Rate: 12 (6 -19) deaths/yr/100,000
Incidence: 190 (90-260) deaths/yr
100
Estimated
percent
reduction
Incidence Rate: 2 (-4 - 7) deaths/yr/100,000
Incidence: 20 (-50 - 90) deaths/yr
standard (ug/m )
Pittsburgh
Incidence Rate: 3 (1 - 5) deaths/yr/100,000
Incidence: 80 (30 -120) deaths/yr
Figure 5-2(b) Estimated percent reduction in PM2.5-related short-term mortality risk for
alternative standards (9tf percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 10 ug/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended study,
is shown in figures in terms of estimated annual incidence rate and annual
incidence (and 95% confidence ranges).
5A-5
-------
Incidence Rate: 1 (-1 -4) deaths/yr/100,000
Incidence: 30 (-30 - 80) deaths/yr
100
Incidence Rate: 1 (0 -1) deaths/yr/100,000
Incidence: 60 (-10-120) deaths/yr
Incidence Rate: 7 (3 -11) deaths/yr/100,000
Incidence: 110 (50-160) deaths/yr
Incidence Rate: 1 (-2 - 3) deaths/yr/100,000
Incidence: 10 (-20 - 40) deaths/yr
80
Estimated
percent 60'
reduction
40
Pittsburgh
Incidence Rate: 1 (0 - 2) deaths/yr/100,000
Incidence: 30 (10-50) deaths/yr
Estimated
percent 6Q,.
reduction
Figure 5-2(c) Estimated percent reduction in PM2.5-related short-term mortality risk for
alternative standards (98?h percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 15 ug/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended study,
is shown in figures in terms of estimated annual incidence rate and annual
incidence (and 95% confidence ranges).
5A-6
-------
Incidence Rate: 1 (-1 -2) deaths/yr/100,000
Incidence: 10 (-10 - 40) deaths/yr
100
80
Estimated
percent m. .
reduction
24-hour
standard (ug/m3)
(ug/m )
Detroit
Incidence Rate: 0 (0 -1) deaths/yr/100,000
Incidence: 30 (-4 - 60) deaths/yr
100
Incidence Rate: 4 (2 - 6) deaths/yr/100,000
Incidence: 60 (30 - 90) deaths/yr
Estimated
percent 60
reduction
40t
Annual
standard
(ug/m3)
Philadelphia
24-hour
standard (ug/m3)
Incidence Rate: 0 (-1 -1) deaths/yr/100,000
Incidence: 5 (-10 - 20) deaths/yr
Estimated
percent Rn4.
reduction
24-hour
standard (ug/m3)
Pittsburgh
Incidence Rate: 0 (0 -1) deaths/yr/100,000
Incidence: 9 (3-14) deaths/yr
Figure 5-2(d) Estimated percent reduction in PM^s-related short-term mortality risk for
alternative standards (98th percentile form) relative to risk associated with
meeting current standards (based on assumed cutpoint of 20 ug/m3). Risk
associated with meeting current PM2.5 standards, based on ACS extended study,
is shown in figures in terms of estimated annual incidence rate and annual
incidence (and 95% confidence ranges).
5A-7
-------
Table 5B-l(a)
Predicted percent of counties with monitors (and percent of population in counties with monitors) not
likely to meet alternative annual and 24-hour (98th percentile form) PM2 5 standards
Alternative Standards and Levels
(Ijg/rm)
No. of counties with monitors
(Population, in thousands)
Percent of counties, total and by region, (and total percent population) not likely to meet stated standard and level*
Total
counties (population)
562 (185,780)
Northeast
83
Southeast
168
Industrial
Midwest
130
Upper
Midwest
49
Southwest
21
Northwest
81
Southern
CA
15
Outside
Regions**
15
Annual standard only:
15
14
13
12
14 (30)
25 (41)
40 (55)
54 (66)
19
28
47
70
7
21
40
61
29
51
76
89
0
0
4
12
0
5
5
5
4
5
7
12
60
67
67
67
0
0
0
0
Combined annual /24-hour:
15/65
15/50
15/45
15/40
15/35
15/30
15/25
14/65
14/50
14/45
14/40
14/35
14/30
14/25
14 (30)
15 (31)
15 (33)
17 (35)
27 (48)
51 (72)
78 (86)
19
19
19
20
45
78
98
7
7
7
7
8
29
77
29
29
29
30
47
87
99
0
0
0
0
0
6
51
0
0
10
10
10
19
43
4
9
12
19
36
51
65
60
60
60
60
60
80
80
0
0
0
0
7
13
13
25 (41)
26 (43)
26 (44)
27 (46)
34 (55)
53 (72)
78 (86)
28
28
28
28
45
78
98
21
21
21
21
22
33
77
51
51
51
52
58
88
99
0
0
0
0
0
6
51
5
5
10
10
10
19
43
5
10
12
19
36
51
65
67
67
67
67
67
80
80
0
0
0
0
7
13
13
5B-1
-------
Alternative Standards and Levels
(Ijg/rm)
No. of counties with monitors
(Population, in thousands)
13/65
13/50
13/45
13/40
13/35
13/30
13/25
12/65
12/50
12/45
12/40
12/35
12/30
12/25
Percent of counties, total and by region, (and total percent population) not likely to meet stated standard and level*
Total
counties (population)
562 (185,780)
Northeast
83
Southeast
168
Industrial
Midwest
130
Upper
Midwest
49
Southwest
21
Northwest
81
Southern
CA
15
Outside
Regions**
15
40 (55)
40 (56)
41 (57)
42 (58)
45 (62)
57 (74)
78 (86)
47
47
47
47
53
78
98
40
40
40
40
40
43
77
76
76
76
76
77
90
99
4
4
4
4
4
8
51
5
5
10
10
10
19
43
7
10
12
19
36
51
65
67
67
67
67
67
80
80
0
0
0
0
7
13
13
54 (66)
54 (66)
54 (67)
55 (68)
58 (71)
64 (78)
79 (86)
70
70
70
70
70
84
98
61
61
61
61
61
62
78
89
89
89
89
89
94
99
12
12
12
12
12
14
51
5
5
10
10
10
19
43
12
12
14
20
36
51
65
67
67
67
67
67
80
80
0
0
0
0
7
13
13
* Based on 2001-2003 data for sites with at least 11 samples per quarter for all 12 quarters. As such, these estimates are not based on the same air quality data
that would be used to determine whether an area would attain a given standard or set of standards. These estimates can only approximate the number of
counties that are likely not to attain the given standards and should be interpreted with caution.
**
"Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
5B-2
-------
Table 5B-l(b)
Predicted percent of counties with monitors (and percent of population in counties with monitors) not
likely to meet alternative annual and 24-hour (99th percentile form) PM2 5 standards
Alternative Standards and Levels
(mg/ms)
No. of counties with monitors
(Population, in thousands)
Percent of counties, total and by region, (and total percent population) not likely to meet stated standards and levels*
Total
counties (population)
562 (185,780)
Northeast
83
Southeast
168
Industrial
Midwest
130
Upper
Midwest
49
Southwest
21
Northwest
81
Southern
CA
15
Outside
Regions**
15
Annual only:
15
14
13
12
14 (30)
25 (41)
40 (55)
54 (66)
19
28
47
70
7
21
40
61
29
51
76
89
0
0
4
12
0
5
5
5
4
5
7
12
60
67
67
67
0
0
0
0
Combined annual /24-hour:
15/65
15/50
15/45
15/40
15/35
15/30
15/25
14/65
14/50
14/45
14/40
14/35
14/30
14/25
14 (30)
16 (33)
18 (35)
27 (46)
44 (68)
68 (82)
85 (89)
19
19
24
47
72
96
100
7
7
7
9
17
54
86
29
29
32
42
77
97
99
0
0
0
0
0
35
69
0
10
10
10
19
38
48
5
15
21
36
51
59
73
60
60
60
67
80
80
87
0
0
0
7
13
13
13
25 (41)
27 (44)
28 (45)
35 (53)
47 (70)
68 (82)
85 (89)
28
28
30
48
72
96
100
21
21
21
23
27
54
86
51
51
52
57
78
97
99
0
0
0
0
0
35
69
5
10
10
10
19
38
48
6
15
21
36
51
59
73
67
67
67
73
80
80
87
0
0
0
7
13
13
13
5B-3
-------
Alternative Standards and Levels
(mg/ms)
No. of counties with monitors
(Population, in thousands)
13/65
13/50
13/45
13/40
13/35
13/30
13/25
12/65
12/50
12/45
12/40
12/35
12/30
12/25
Percent of counties, total and by region, (and total percent population) not likely to meet stated standards and levels*
Total
counties (population)
562 (185,780)
Northeast
83
Southeast
168
Industrial
Midwest
130
Upper
Midwest
49
Southwest
21
Northwest
81
Southern
CA
15
Outside
Regions**
15
40 (55)
41 (57)
42 (58)
47 (62)
54 (73)
70 (82)
85 (89)
47
47
49
59
75
96
100
40
40
40
40
40
58
86
76
76
76
77
85
97
99
4
4
4
4
4
35
69
5
10
10
10
19
38
48
9
15
21
36
51
59
73
67
67
67
73
80
80
87
0
0
0
7
13
13
13
54 (66)
55 (67)
56 (68)
59 (71)
63 (77)
73 (83)
85 (89)
70
70
71
75
80
96
100
61
61
61
62
62
68
86
89
89
89
89
92
98
99
12
12
12
12
12
35
69
5
10
10
10
19
38
48
12
16
22
36
51
59
73
67
67
67
73
80
80
87
0
0
0
7
13
13
13
* Based on 2001-2003 data for sites with at least 11 samples per quarter for all 12 quarters. As such, these estimates are not based on the same air quality data
that would be used to determine whether an area would attain a given standard or set of standards. These estimates can only approximate the number of
counties that are likely not to attain the given standards and should be interpreted with caution.
** "Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
5B-4
-------
Table 5B-2(a) Percent of counties with monitors (and percent of population in counties with monitors) not likely to meet
nth
alternative 24-hour (98 percentile form) UPMi0-2.s standards
Alternative Levels
Number of counties
with monitors
(Population, in
thousands)
70
65
60
55
50
Percent of counties, total and by region, (and total percent population) not likely to meet alternative 24-hour (98th percentile
form) PM10-2.s standards or current PM10 standards*
Total Counties
(population)
259 (141,859)
7(9)
9(11)
12 (16)
13 (18)
16 (27)
Northeast
44
2
2
2
5
5
Southeast
60
3
o
J
5
5
7
Industrial
Midwest
57
4
5
7
7
9
Upper
Midwest
18
0
6
6
17
22
Southwest
13
46
46
62
62
62
Northwest
45
4
9
13
13
16
Southern
CA
75
33
40
40
40
53
Outside
Regions**
7
29
29
43
43
57
* Based on 2001-2003 data for sites with 4, 8, or 12 consecutive quarters with at least 11 samples per quarter. As such, these estimates are not based on the
same air quality data that would be used to determine whether an area would attain a given standard or set of standards. These estimates can only approximate
the number of counties that are likely not to attain the given standards and should be interpreted with caution.
** "Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
5B-5
-------
Table 5B-2(b)Percent of counties with monitors (and percent of population in counties with monitors) not likely to meet
-.th
alternative 24-hour (99 percentile form) UPMi0-2.s standards
Alternative Levels
Number of counties
with monitors
(Population, in
thousands)
85
80
75
70
65
60
Percent of counties, total and by region, (and total percent population) not likely to meet alternative 24-hour
(99th percentile form) PM10.2.s standards or current PM10 standards*
Total Counties
(population)
259 (141,859)
8(10)
10(11)
12 (14)
13 (15)
16 (19)
19 (27)
Northeast
44
7
7
7
7
9
11
Southeast
60
2
o
J
5
7
8
10
Industrial
Midwest
57
4
4
5
5
7
7
Upper
Midwest
18
6
6
6
17
33
39
Southwest
13
46
54
54
54
54
62
Northwest
45
11
11
11
13
16
16
Southern
CA
75
20
20
40
40
40
60
Outside
Regions**
7
14
29
29
29
43
43
* Based on 2001-2003 data for sites with 4, 8, or 12 consecutive quarters with at least 11 samples per quarter. As such, these estimates are not based on the
same air quality data that would be used to determine whether an area would attain a given standard or set of standards. These estimates can only approximate
the number of counties that are likely not to attain the given standards and should be interpreted with caution.
** "Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
5B-6
-------
Table 5B-2(c) Percent of counties with monitors (and percentage of population in counties with monitors) not likely to meet
current PMio standards
Database
Percent of counties, total and by region, (and total percent population) not meeting the current PM10 standards
Total
Counties
(population)
Northeast
Southeast
Industrial
Midwest
Upper
Midwest
Southwest
Northwest
Southern
CA
Outside
Regions**
All PM'w sites :
[Number of counties
with monitors
(Population, in
thousands)]*
585(170,118) 84
120
115
52
33
142
18
21
Percent violating
(13)
27
10
61
10
rit sties that meet
'urban' criteria:
[Number of counties
with monitors
(Population, in
thousands)]
309 (153,546) 59
70
67
21
17
50
15
10
Percent violating 6 (12)
29
53
10
Urban PMrti sties,
alsoPMm-? *
[Number of counties
with monitor
(Population, in
thousands)]
259 (141,859) 44
60
57
18
13
45
sp
Percent violating
7 (11)
0
0
38
47
14
* Based on official EPA design values for 2001-2003; see http://epa.gov/airtrends/values.html.
** "Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
5B-7
-------
4>
—
M
=
U
H
o
0)
M)
U
100
90
80
70
60
50
40
30
20
10
0
blue = east counties*
md = west (-.
65
60
55
25
20
15
10
50 45 40 35 30
12 p.m. - 4 p.m. average PM2 5 i
Figure 7A-1. Estimated exceedances (%) of various PM25 levels for 12 p.m. - 4 p.m. based on daily
county maximum, 2001-2003.
* Figure 7A-1 can be seen in color at: http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr sp.html
Source: Schmidt et al. (2005)
7A-1
-------
Table 7A-1. Predicted percent of counties with monitors (and percent of population in counties with monitors) not likely to
meet alternative 4-hour (12 p.m. - 4 p.m.) PM2 5 secondary standards.
Alternative Forms and Levels of
Secondary PM2.5 Standard
Number of counties with moniton
(Population, in thousands)
Percent of counties, total and by region, (and total percent population) not likely to meet stated standards and levels*
Total Counties
(population)
168 (78,419)
Northeast
33
Southeast
45
Industrial
Midwest
30
Upper
Midwest
16
Southwest
14
Northwest
25
Southern CA
3
Outside
Regions**
2
PM2.5 standard levels and forms:
20 |jg/mj, 92nd percentile
25 |jg/mj, 92nd percentile
30 |jg/mj, 92nd percentile
51 (67)
27 (46)
8(17)
76
52
15
36
16
0
90
53
17
38
0
0
29
7
7
8
4
0
100
100
67
100
0
0
20 |jg/mj, 95th percentile
25 |jg/mj, 95th percentile
30 |jg/mj, 95th percentile
70 (83)
47 (67)
24 (43)
88
79
52
73
31
7
97
87
47
50
19
0
50
29
14
24
4
4
100
100
100
100
100
0
20 |jg/mj, 98th percentile
25 |jg/mj, 98th percentile
30 |jg/mj, 98th percentile
85 (96)
70(81)
56 (73)
100
94
85
100
62
38
100
100
90
63
50
19
57
57
57
48
28
24
100
100
100
100
100
100
* Based on 2001-2003 data for sites with at least 1 year of complete data. Completeness criteria (per year) = Minimum of 3 hours per day (in 4-hour 12-4pm
window), 275+ days per year. As such, these estimates are not based on the same air quality data that would be used to determine whether an area would attain a
given standard or set of standards. These estimates can only approximate the number of counties that are likely not to attain the given standards and should be
interpreted with caution.
** "Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
7A-2
-------
\
ATTACHMENT A
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
June 6, 2005
EPA-SAB-CASAC-05-007
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
Honorable Stephen L. Johnson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Subject: Clean Air Scientific Advisory Committee (CASAC) Particulate Matter (PM)
Review Panel's Peer Review of the Agency's Re view of the National Ambient
Air Quality Standards for Particulate Matter: Policy Assessment of Scientific
and Technical Information (Second Draft PM Staff Paper, January 2005); and
Particulate Matter Health Risk Assessment for Selected Urban Areas: Second
Draft Report (Second Draft PM Risk Assessment, January 2005)
Dear Administrator Johnson:
EPA's Clean Air Scientific Advisory Committee (CASAC), supplemented by subject-
matter-expert Panelists — collectively referred to as the CASAC Particulate Matter (PM)
Review Panel ("Panel") — met in a public meeting held in Durham, NC, on April 5-6, 2005, to
conduct a peer review of subject documents. The current Panel roster is found in Appendix A of
this report.
This meeting was a continuation of the CASAC PM Review Panel's peer review of the
Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment
of Scientific and Technical Information (First Draft PM Staff Paper, August 2003) and a related
draft technical report, Particulate Matter Health Risk Assessment for Selected Urban Areas (First
Draft PM Risk Assessment, August 2003). The previous draft of the PM Staff Paper was a
preliminary version since the Panel has not yet finished its review of the Air Quality Criteria
Document (AQCD) for PM (which was completed in October 2004). In addition, further risk
analyses and analyses of alternative forms of the PM standards were included in the Second
Draft PM Staff Paper and Second Draft PM Risk Assessment. The charge questions provided to
the Panel by EPA are found in Appendix B to this report. Panelists' individual review comments
are provided in Appendix C of this report.
In its peer review of the Second Draft of the PM Staff Paper, most of the members of the
CASAC PM Review Panel found the document was generally well-written and scientifically
-------
well-reasoned for all but the short term primary PMio-2.5 standard. A majority of the members of
the Panel were in agreement with the following: the primary PM2 5 24-hour and annual PM
national ambient air quality standards (NAAQS) should be modified to provide increased public
health protection. Although the evidence for a standard for coarse-mode particles was weaker
than for the PM2.5, the Panel agreed that a 24-hour NAAQS for PMio-2.5 was appropriate,
especially in urban areas, with caveats to make exceptions for those types of rural dusts thought
to have low toxicity. The Panel recommends that the Agency staff expand and strengthen the
discussion of the exposure index (size-range plus composition and/or source) and the monitoring
strategy to be used for the coarse-mode NAAQS, as well as the degree of public health
protection against thoracic coarse PM expected relative to the protection afforded by the current
PMio short-term NAAQS. As discussed below, the CASAC PM Review Panel will need to
review the final version of the PM Staff Paper before providing a final opinion to EPA on the
adequacy of a short-term PMio-2.5 NAAQS.
The approach used to set secondary NAAQS to protect the environment was considered
appropriate, but it was strongly recommended that, in the future, Agency staff also give serious
consideration to a shift to the European approach of critical loads to protect vegetation and
ecosystems in the U.S. In addition, most of the Panel supported Agency staff recommendations
regarding a standard to address the issue of urban visibility impairment.
1. Background
The CASAC, comprised of seven members appointed by the EPA Administrator, was
established under section 109(d)(2) of the Clean Air Act (CAA or "Act") (42 U.S.C. § 7409) as
an independent scientific advisory committee, in part to provide advice, information and
recommendations on the scientific and technical aspects of issues related to air quality criteria
and NAAQS under sections 108 and 109 of the Act. Section 109(d)(l) of the CAA requires that
EPA carry out a periodic review and revision, where appropriate, of the air quality criteria and
the NAAQS for "criteria" air pollutants such as PM. The CASAC, which is administratively
located under EPA's Science Advisory Board (SAB) Staff Office, is a Federal advisory
committee chartered under the Federal Advisory Committee Act (FACA), as amended, 5 U.S.C.,
App. The CASAC PM Review Panel is comprised of the seven members of the chartered
(statutory) Clean Air Scientific Advisory Committee, supplemented by fifteen technical experts.
Under section 108 of the CAA, the Agency is required to establish NAAQS for each
pollutant for which EPA has issued criteria, including PM. Section 109(d) of the Act
subsequently requires periodic review and, if appropriate, revision of existing air quality criteria
to reflect advances in scientific knowledge on the effects of the pollutant on public health and
welfare. EPA is also to revise the NAAQS, if appropriate, based on the revised criteria. The
purpose of the Second Draft PM Staff Paper is to evaluate the policy implications of the key
scientific and technical information contained in a related document, EPA's revised PM AQCD
(October 2004), and to identify critical elements that EPA believes should be considered in the
review of the PM NAAQS. The Staff Paper for PM is intended to "bridge the gap" between the
scientific review contained in the PM AQCD and the public health and welfare policy judgments
required of the Administrator in reviewing the PM NAAQS.
-------
This Second Draft PM Staff Paper is based on the information in the final PM AQCD,
which had been the subject of review by the CAS AC PM Review Panel since October 1999.
(The report from the Panel's final meeting to review the PM AQCD, dated October 4, 2004, is
posted on the SAB Web Site at: http://www.epa.gov/sab/pdf/casac05001.pdf The Agency
subsequently announced the availability of a final document, Air Quality Criteria for Paniculate
Matter (EPA/600/P-99/002aF, EPA/600/P-99/002bF) on October 29, 2004.) In addition, the
Second Draft PM Staff Paper builds upon the First Draft PM Staff Paper, which was the subject
of review by the CAS AC PM Review Panel held on November 12-13, 2003. The report from the
Panel's previous meeting to review these draft documents, dated February 18, 2004, is posted on
the SAB Web Site at: http://www.epa.gov/sab/pdf/casac 04004.pdf. The Second Draft PM Staff
Paper and the Second Draft PM Risk Assessment were made available for public review and
comment on January 31, 2005 by EPA's Office of Air Quality Planning and Standards
(OAQPS), within the Office of Air and Radiation (OAR). The Second Draft PM Risk
Assessment, which builds upon the Agency's First Draft PM Risk Assessment, describes the
methodology and presents the results from an updated PM health risk assessment for health risks
associated with exposure to fine and thoracic coarse particles in a number of U.S. cities.
2. CASAC PM Review Panel's Peer Review of the Second Draft PM Staff Paper and
Second Draft PM Risk Assessment
After reviewing the Second Draft PM Staff Paper and written comments from the public,
and after hearing public comments at the meeting, a majority of the members of the CASAC PM
Review Panel were in agreement with the following: the primary PM2.5 24-hour and annual
NAAQS should be modified to provide increased public health protection. The evidence for a
NAAQS for coarse mode particles is weaker than for PM2.5. The Panel agreed, however, that a
24-hour NAAQS for PMi0-2.5 is appropriate, especially in urban areas and with caveats to make
exceptions for those types of rural dusts thought to have low toxicity. Before the Panel renders
its final recommendation concerning a daily PMio-2.5 standard, the Panel recommends that the
Agency staff expand and strengthen the discussion of the exposure index (size-range plus
composition and/or source) and the monitoring strategy to be used for this standard, as well as
the degree of public health protection expected relative to the protection against thoracic coarse
PM afforded by the current PMio short-term NAAQS. Accordingly, after the Panel has reviewed
the Final Staff Paper and Risk Assessment for Particulate Matter following its issuance on June
30, 2005, the Panel will meet again this summer via a public teleconference to consider the final
Staff Paper's recommendations concerning the setting of a coarse PM standard. Subsequent to
the Panel's teleconference meeting, we will send you a separate letter providing the Panel's
recommendations concerning PMio-2.5 as an indicator together with our views on the averaging
time, statistical form, and level of any potential daily PMio-2.5 standard.
The approach used to set secondary standards to protect the environment was considered
appropriate, but it was strongly recommended that, in the future, Agency staff give serious
consideration to the European approach of critical loads to protect vegetation and ecosystems in
the U.S. In addition, most of the Panel supported Agency staff recommendations regarding a
standard to address the issue of urban visibility impairment.
In its peer review of the Second Draft of the PM Staff Paper, most of the members of the
CASAC PM Review Panel found the document was generally well-written and scientifically
-------
well-reasoned. The following represent summaries of advice and recommendations of the Panel
in response to the charge questions provided by EPA, which are found in Appendix B to this
report. More detailed responses are provided in the individual review comments of each member
of the Panel included in Appendix C to this report.
The CASAC PM Review Panel has reached agreement on the following synopsis of
advice and recommendations for the Agency:
AIR QUALITY
Chapter 2: Characterization of Ambient PM
Chapter 2 of the Second Draft PM Staff Paper was considered well-written, presenting an
accurate and concise summary of Chapters 2, 3, and 5 of the PM Air Quality Criteria Document.
The chapter was acceptable to the Panel reviewers as written, but some improvements were
suggested in two areas. In the area of measurement methods, the Panel thought there should be
more discussion of continuous PM monitoring methods in light of the recommended secondary
fine particle standard based on 4- to 8-hour concentration averages and the likely availability of a
continuous coarse particle monitor. A more quantitative characterization of PM mass
measurement errors could be presented, especially for PMio-2.s. Interest was expressed in a
discussion of alternative PM indicators for future NAAQS considerations related to the source of
the PM, especially for the potentially more toxic portion of coarse PM. In the area of health and
visibility assessments, concern was expressed that spatial gradients near major arterials and other
urban sources are not adequately addressed. It was suggested that spatial heterogeneity within a
city might better be characterized in terms of departures of individual sites from the metropolitan
average, in place of this draft's summary statistics of between-sites comparisons. Some
members of the Panel expressed concerns about the policy-relevant background (PRB) estimates.
The true background is not observable and is effectively unknowable. As indicated in the
summary of Chapter 5 comments, alternative standards should be analyzed in ways that are
insensitive to estimates of PRB.
HEALTH-BASED STANDARDS
Chapter 3: Policy-Relevant Assessment of Health Effects Evidence
Chapter 3 addresses each of the health effects issues relevant to PM NAAQS
reconsideration. Agency staff have adequately reviewed advances in understanding effects from
studies conducted subsequent to the 1997 NAAQS, as summarized in the latest PM AQCD.
Overall, EPA staff have done a reasonably good job of summarizing the health effects basis for
considering revised or new PM standards. However, there are instances where the summary of
findings and their interpretation are overstated (see individual Panel member review comments,
particularly on pages C-82 and C-83). Specifically, there was confusion over strength of
association versus strength of evidence, between confounding and effect modification, and
between temporality and lag structure. There are some areas where Agency staff have either
over-interpreted or overstated the extent to which the health data support a particular PM
indicator variable. These problems can be addressed if EPA staff give heed to the individual
-------
comments of the CAS AC PM Review Panel when revising the chapter. The discussion of the
effect of co-pollutants in interpreting the results of PM health studies would benefit from a
clearer discussion of EPA staffs approach to interpreting quantitative results from multi-
pollutant studies.
Numerous epidemiological studies that are reviewed in this chapter have shown
statistically significant associations between the concentrations of ambient air PM2.5 and PMio
(including levels that are lower than the current PM NAAQS) and excess mortality and
morbidity. Furthermore, the evidence presented indicates that the effects appear to be robust, in
that inclusion of other environmental variables in regression analyses has not been found to
materially affect the associations of the adverse health effects with ambient PM concentrations.
On the other hand, the scientific evidence included in the PM Air Quality Criteria Document and
draft Staff Paper provides substantially less data derived from controlled exposure studies in
humans or experimental animals; or from studies of biological mechanisms in animals in vivo or
cells and tissues in vitro, to support the biological plausibility of the effects of the relatively low
concentrations found in the human population studies. In the case of controlled human studies,
this appears to be due to the inherent limitations of such studies, which are largely confined to
young, healthy subjects exposed for short time periods and the examination of mild, reversible
effects. In the case of animal studies, it may be due to not having adequate animal models of
human disease processes and exposures to individual chemical agents rather than realistic
mixtures. Both types of studies may be inadequate to represent the real-world situation of
susceptible subpopulations of humans undergoing long-term exposures and occasional peak
levels of complex mixtures of PM, associated surface coatings of reactive chemicals, and
gaseous co-pollutants. This apparent discontinuity needs to be addressed in future research.
The current health effects data base for coarse mode particles (PMio-2.s) is relatively
weak. Few epidemiology studies have been conducted where PMio-2.5 was measured directly as
opposed to obtaining an estimate of this indicator variable by subtracting data from collocated
PM2.5 and PMio monitors. There is limited evidence that PMio-2.5 may be related to
cardiovascular mortality as well as to hospital admissions for respiratory diseases. The few
controlled human studies that have been conducted with concentrated ambient particles have not
shed any light on the morbidity findings from epidemiological studies. Moreover, animal
toxicological studies using coarse mode particles are virtually nonexistent; they are difficult to
perform because rodents are obligate nose breathers and thus few of these particles penetrate to
the lungs. A further complication with current epidemiological studies of the health effects of
PMio-2.5 is that most have been conducted in urban areas, and because coarse mode particles from
urban and rural areas may be markedly different, extrapolating these findings to rural settings
may be difficult. Considerably more research with PMio-2.5 is needed.
Chapter 4: Characterization of Health Risks
One major concern with the current version of the chapter is the clarity of presentation.
Readers need to struggle through dense prose and jargon-ridden text to identify key aspects of
the methods and findings. Key terms are sometimes used incorrectly or inconsistently across the
chapter. The chapter could be substantially shortened, and redundancies need to be addressed.
Figure 4-1 provides an overall framework for the risk assessment that could be used to shape the
-------
chapter. We suggest that the chapter refer to it repetitively as the risk assessment methods and
findings are described. Subheadings, such as "assumptions" and "sensitivity analyses," might be
more effectively used to guide the reader through the individual sections of the chapter.
A second concern is with methodological issues. The issue of the selection of
concentration-response (C-R) relationships based on locally-derived coefficients needs more
discussion. The Panel did not agree with EPA staff in calculating the burden of associated
incidence in their risk assessment using either the predicted background or the lowest measured
level (LML) in the utilized epidemiological analysis. The available epidemiological database on
daily mortality and morbidity does not establish either the presence or absence of threshold
concentrations for adverse health effects. Thus, in order to avoid emphasizing an approach that
assumes effects that extend to either predicted background concentrations or LML, and to
standardize the approach across cities, for the purpose of estimating public health impacts, the
Panel favored the primary use of an assumed threshold of 10 |ig/m3. The original approach of
using background or LML, as well as the other postulated thresholds, could still be used in a
sensitivity analysis of threshold assumptions.
The analyses in this chapter highlight the impact of assumptions regarding thresholds, or
lack of threshold, on the estimates of risk. The uncertainty associated with threshold or
nonlinear models needs more thorough discussion. A major research need is for more work to
determine the existence and level of any thresholds that may exist or the shape of nonlinear
concentration-response curves at low levels of exposure that may exist, and to reduce uncertainty
in estimated risks at the lowest PM concentrations.
Chapter 5: Staff Conclusions and Recommendations on Primary PM NAAQS
The Panel had the following advice and recommendations for the PM 2.5 standard:
The tack taken by EPA staff in recommending a suite of standards for PM2.5 by using
both an evidence-based and a risk-based approach, while necessarily ad hoc, was felt to be
reasonable. Most Panel members favored continued use of the 98th percentile form because it is
more robust than the 99th percentile form and therefore would provide more stability to prevent
areas from bouncing in and out of attainment from year to year. Some concern was expressed as
to whether EPA staff would exclude days on which natural phenomena such as forest fires distort
the distribution. The Panel felt that such days should be eliminated before standard compliance
is assessed. The link between the percentile form and the exposure level chosen is well-
illustrated in the type of three-dimensional figures created by Dr. Miller at the April meeting
(based on the data in Figure 5-2 in the 2nd draft PM Staff Paper), which were endorsed by the
Panel and later provided in expanded form by OAQPS staff. The Panel endorses the inclusion of
these types of figures in the Staff Paper. It would be helpful if reductions in risk associated with
different regulatory options were expressed in the form of absolute numbers normalized to a
fixed population size, in addition to those already expressed as percentage reductions.
In recommending revisions to the PM2.5 NAAQS, changes to either the annual or the 24-
hour standard, or both, could be recommended. Three arguments were made that support placing
more emphasis on lowering the 24-hour NAAQS. First, the vast majority of studies indicating
-------
effects of short-term PM2.5 exposures were carried out in settings in which PM2.5 concentrations
were largely below the current standard of 65 jug/m3. Second, the amount of evidence on short-
term effects, at least as reflected by the number of reported studies, is greater than for long-term
effects. Third, toxicological findings largely relate to effects of short-term PM2.s exposures.
There was a consensus among the Panel members in agreement with the EPA staff
recommendations that focused on decreasing PM2.5 concentrations through lowering of the 24-
hour PM standard, but the panel did not endorse the option of keeping the annual standard at its
present level of 15 jug/m3. It was appreciated that some cities have relatively high annual PM
concentrations, but without much variation in concentrations from day to day. Such cities would
only rarely exceed a 24-hour PM2.5 standard, even if set at levels below the current standard.
This observation indicates the desirability of lowering the level of the annual PM2.5 standard as
well.
Of the options presented by EPA staff for lowering the level of the PM standard, based
on the above considerations and the predicted reductions in health impacts derived from the risk
analyses, most Panel members favored the option of setting a 24-hour PM2.5 NAAQS at
concentrations in the range of 35 to 30 jug/m3 with the 98th percentile form, in concert with an
annual NAAQS in the range of 14 to 13 jug/m3. The justification for not moving to the lowest
staff-recommended levels within these ranges is that these were generally associated with only
small additional predicted reductions in risk. In addition, the uncertainties associated with
concentration-response relationships increase greatly below these ranges, as reflected in
substantial widening of the confidence limits for point estimates.
The Panel had the following advice and recommendation for the PM_m-2_5 standard:
It was acknowledged that the scientific basis supporting a causal role of PMio-2.5 in an
array of adverse health effects is weaker than that of PM2.5. Regardless, most of the Panel
members felt that the evidence that exists supports a causal role for health effects for PMio-2.5-
Moreover, setting this NAAQS would allow continuation and expansion of the PMio-2.5
monitoring network that would facilitate collection of data for future exposure assessment and
epidemiology studies. Because the evidence for the toxicity of PMio-2.5 comes from studies
conducted primarily in urban areas and is related, in large part, to the re-entrainment of urban
and suburban road dusts as well as primary combustion products, there is concern that the
associations of adverse effects with PMio-2.5 may not apply to rural areas where the PMio-2.5 is
largely composed of less-toxic components of windblown soil or products of agricultural
operations for which there is either no or limited evidence of health issues.
Further, although there is some evidence that short-term changes in concentrations of
PMio-2.5 are associated with changes in mortality, particularly cardiovascular mortality, the
evidence in support of effects on morbidity, especially respiratory morbidity, is stronger. Most
Panel members therefore favored not including short-term mortality effects in the health impact
predictions, in line with the approach taken by EPA staff. The Panel agreed with Agency staffs
approach of not setting an annual NAAQS for PMio-2.5 at this time.
-------
One of the major reservations expressed by the Panel in recommending a 24-hour
2.5 NAAQS related to the non-specificity of the PMi0-2.5 mass metric. Given that most evidence
indicates that the component of the coarse fraction in most rural areas has little or no toxicity at
environmental concentrations, it was felt important to qualify the PMi0-2.5 standard by somehow
allowing exceptions for regions where the coarse fraction was composed largely of material that
was not contaminated by industrial- or motor vehicle traffic-associated sources. Options
discussed by members of the Panel for attempting to achieve this approach included limiting the
standard to cover "all" urban areas, the judicious siting of monitors with a focus on urban areas,
or regulatory exceptions for regions where road dust is not an issue or where rural components
dominate the source. No single option was favored.
The panel also agreed that there was a need for more research on the health effects of
PMio-2.5- Such research will require the continuation and expansion of the PMio-2.5 monitoring
network in both rural and urban areas. The Panel recommends that the Agency staff expand and
strengthen the discussion of the exposure index (size-range plus composition and/or source) and
the monitoring strategy to be used for this NAAQS as well as the degree of public health
protection expected relative to the protection against thoracic coarse PM afforded by the current
short-term NAAQS.
WELFARE-BASED STANDARDS
Chapters 6 & 7: PM-Related Welfare Effects
Overall, these chapters are well done. Comments are provided below regarding
vegetation and ecosystem, materials soiling, and visibility.
Considering the effects of PM on vegetation and ecosystems, EPA staff are to be
commended for a well-written and concise reflection of the key science as presented in the final
PM AQCD. The ecological risk assessment is reasonable given the required "criteria pollutant"
approach. That being said, the criteria pollutant approach in this case (i.e.., PM) has serious
shortcomings when it comes to ensuring environmental protection of vegetation and ecosystems
in the U.S. This is illustrated in the following discussion.
Scientific evidence presented in the PM Staff Paper and the PM AQCD indicates that
forest ecosystems at a number of locations in the U.S. "are now showing severe symptoms of
nitrogen saturation." The Staff Paper makes the point that this is the result of chronic long-term
additions of reactive nitrogen (Nr) species that have been accumulating over time. The PM Staff
Paper also makes the point that the issue of forest-ecosystem deterioration is broader and more
complex than just Nr accumulation. The Staff Paper notes that, "The most significant PM-
related ecosystem-level effects result from long-term cumulative deposition of a given chemical
species (e.g., nitrate) or mix (e.g., acidic deposition) that exceeds the natural buffering or storage
capacity of the ecosystem and/or affects the nutrient status of the ecosystem." A key point
implied here and elaborated later in the PM Staff Paper text is that PM deposition is only
partially-responsible for the observed ecosystem-level effects and that the extent of the role of
PM deposition in these ecosystem-level effects needs to be determined. While this has scientific
merit, the question must be asked as to whether knowing the role of PM alone will improve the
-------
protection of vegetation and ecosystems in the U.S.? The answer to this question is critical
because forest ecosystems are responding to the cumulative total load that has resulted from the
chronic long-term deposition of both PM as well as gases and not to PM alone.
While EPA staff have done a commendable job within the context of the criteria pollutant
approach, it is strongly recommended that in the future that Agency staff give serious
consideration to a philosophical shift from the criteria pollutant approach to the European
approach of "critical loads" when it comes to ensuring protection of vegetation and ecosystems
in the U.S. The critical load is defined in the criteria document and is a quantitative estimate of
an exposure to one or more pollutants below which significant harmful effects on specified
sensitive elements of the environment do not occur according to present knowledge. The current
criteria pollutant approach is a significant limitation in the efforts of the Agency staff to address
the cumulative load of all the pollutant stressors to which ecosystems are responding.
Considering soiling and materials effects, several of the Panel members specifically
asked EPA to add some discussion of the welfare effects caused by soiling from coarse particles.
This may lead to consideration of a secondary PMio-2.5 standard intended to protect against
adverse welfare effects.
Considering visibility effects, most Panel members strongly supported the EPA staff
recommendation to establish a new, secondary PM2.5 standard to protect urban visibility.
Overall, the Second Draft Staff Paper visibility sections (Chapters 6 and 7 and the detailed
technical appendix by Schmidt et a/., 2005) are well-conceived and clearly-written. Agency staff
can also be commended for responsiveness to comments previously submitted by this Panel on
the PM AQCD and the First Draft PM Staff Paper. The recommended new standard was
considered by most Panel members to be a reasonable complement to the Regional Haze Rules
that protect Class I areas. The dissenting view is provided in one Panel member's individual
review comments (see pages C-101 and C-102).
The recommended range of secondary standards includes an indicator (PM2.5 mass),
averaging time (4 to 8 daylight hours), level (20 to 30 |ig/m3) and form (90th percentile "or
slightly higher"). The sub-daily averaging time is an innovative approach that strengthens the
quality of the PM2.5 indicator by targeting the driest part of the day. An indirect but important
benefit will come from the direct use of— and more intense scrutiny on the quality of—the
hourly data from the widely deployed continuous PM2.5 mass monitors. The net effect is a
"responsive" standard that (for the first time) would directly link public perception of air
pollution (predominantly due to visual effects of light scattering by fine particles in the ambient
air) to a routinely measured pollutant indicator (i.e., artificially-dried PM2.5 mass).
The recommended level and form of the standard are more difficult to specify. The draft
PM Staff Paper employs a "bounding" approach, suggesting a level that is below the "obviously
adverse" level of the current secondary standard — under which extreme short-term
concentrations exceeding 100 |ig/m3 have been observed on days when 24-hour concentrations
do not exceed 65 |ig/m3. Some members of the Panel felt the recommended level (and form) of
the standard were on the high side, but developing a more specific (and more protective) level in
future standards would require updated and refined public visibility valuation studies. Agency
-------
staff are strongly encouraged to support such studies prior to the next round of NAAQS review,
even as it moves forward with the currently-recommended standard.
Some felt the recommended 90th percentile form of the standard was the weakest element
of the EPA staff recommendation and the least well-justified. The visual effects of fine particle
pollution provide the most direct public perception of air pollution of any regulated (or
unregulated) pollutant, and the adversity of the effect is greatest on the haziest days that the 90th
percentile would discard. Some Panel members recommend considering a higher percentile
(92nd to 98th), accompanied by a level toward the upper-end of the recommended range, and/or
otherwise softened by an "exceptional events" policy to assure that secondary non-attainment is
not driven by natural source influences such as dust storms and wild fires.
To determine the degree of non-attainment that will result from a secondary standard,
Agency staff should include — for different combinations of 4-hour and 24-hour levels and
upper percentiles — estimates of concentrations and locations that would be expected to exceed
a recommended secondary standard. EPA staff should also add some discussion of estimated
"background" PM2j conditions for the 4-hour daylight period.
In conclusion, the CAS AC PM Review Panel encourages EPA in its efforts to protect the
public health and our environment from the adverse effects of ambient air PM in the most
effective manner possible. The Panel will continue to offer its advice and recommendations to
help the Agency in meeting the mandates of the Clean Air Act and will review the final version
of the staff paper with respect to EPA staffs approach to setting a PMio-zs standard. As always,
the CASAC PM Review Panel wishes the Agency well in this important endeavor.
Sincerely.,
Dr. Rogene Henderson, Chair
Clean Air Scientific Advisory Committee
Appendix A - Roster of the CASAC Particulate Matter Review Panel
Appendix B - Charge to the CASAC Particulate Matter Review Panel
Appendix C - Review Comments from Individual CASAC Particulate Matter Review Panelists
10
-------
Appendix A - Roster of the CASAC Particulate Matter Review Panel
U.S. Environmental Protection Agency
Science Advisory Board (SAB) Staff Office
Clean Air Scientific Advisory Committee (CASAC)
CASAC Particulate Matter Review Panel*
CHAIR
Dr. Rogene Henderson*, Scientist Emeritus, Lovelace Respiratory Research Institute,
Albuquerque, NM
MEMBERS
Dr. Ellis Cowling*, University Distinguished Professor-at-Large, North Carolina State
University, Colleges of Natural Resources and Agriculture and Life Sciences, North Carolina
State University, Raleigh, NC
Dr. James D. Crapo*, Professor, Department of Medicine, Biomedical Research and Patient
Care, National Jewish Medical and Research Center, Denver, CO
Dr. Philip Hopke**, Bayard D. Clarkson Distinguished Professor, Department of Chemical
Engineering, Clarkson University, Potsdam, NY
Dr. Jane Q. Koenig, Professor, Department of Environmental Health, School of Public Health
and Community Medicine, University of Washington, Seattle, WA
Dr. Petros Koutrakis, Professor of Environmental Science, Environmental Health , School of
Public Health, Harvard University (HSPH), Boston, MA
Dr. Allan Legge, President, Biosphere Solutions, Calgary, Alberta
Dr. Paul J. Lioy, Associate Director and Professor, Environmental and Occupational Health
Sciences Institute, UMDNJ - Robert Wood Johnson Medical School, NJ
Dr. Morton Lippmann, Professor, Nelson Institute of Environmental Medicine, New York
University School of Medicine, Tuxedo, NY
Dr. Joe Mauderly, Vice President, Senior Scientist, and Director, National Environmental
Respiratory Center, Lovelace Respiratory Research Institute, Albuquerque, NM
Dr. Roger O. McClellan, Consultant, Albuquerque, NM
Dr. Frederick J. Miller*, Consultant, Cary, NC
A-l
-------
Dr. Gunter Oberdorster, Professor of Toxicology, Department of Environmental Medicine,
School of Medicine and Dentistry, University of Rochester, Rochester, NY
Mr. Richard L. Poirot*, Environmental Analyst, Air Pollution Control Division, Department of
Environmental Conservation, Vermont Agency of Natural Resources, Waterbury, VT
Dr. Robert D. Rowe, President, Stratus Consulting, Inc., Boulder, CO
Dr. Jonathan M. Samet, Professor and Chair, Department of Epidemiology, Bloomberg School
of Public Health, Johns Hopkins University, Baltimore, MD
Dr. Frank Speizer*, Edward Kass Professor of Medicine, Channing Laboratory, Harvard
Medical School, Boston, MA
Dr. Sverre Vedal, Professor of Medicine, School of Public Health and Community Medicine
University of Washington, Seattle, WA
Mr. Ronald White, Research Scientist, Epidemiology, Bloomberg School of Public Health,
Johns Hopkins University, Baltimore, MD
Dr. Warren H. White, Visiting Professor, Crocker Nuclear Laboratory, University of California
- Davis, Davis, CA
Dr. George T. Wolff, Principal Scientist, General Motors Corporation, Detroit, MI
Dr. Barbara Zielinska*, Research Professor, Division of Atmospheric Science, Desert Research
Institute, Reno, NV
SCIENCE ADVISORY BOARD STAFF
Mr. Fred Butterfield, CASAC Designated Federal Officer, 1200 Pennsylvania Avenue, N.W.,
Washington, DC, 20460, Phone: 202-343-9994, Fax: 202-233-0643 (butterfield.fred@epa.gov)
(Physical/Courier/FedEx Address: Fred A. Butterfield, III, EPA Science Advisory Board Staff
Office (Mail Code 1400F), Woodies Building, 1025 F Street, N.W., Room 3604, Washington,
DC 20004, Telephone: 202-343-9994)
* Members of the statutory Clean Air Scientific Advisory Committee (CASAC) appointed by the EPA
Administrator
** Immediate past CASAC Chair
A-2
-------
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-452/D-05-005
Environmental Protection Air Quality Strategies and Standards Division June 2005
Agency Research Triangle Park, NC
------- |