&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 - First Draft
-------
-------
EPA-452/D-03-001
August 2003
First Draft
Review of the National Ambient Air Quality
Standards for Particulate Matter:
Policy Assessment of Scientific
and Technical Information
OAQPS Staff Paper - First Draft
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
August 2003 Draft - Do Not Quote or Cite
-------
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 release. This draft
OAQPS Staff Paper contains the preliminary 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.
August 2003 Draft - Do Not Quote or Cite
-------
Table of Contents
List of Tables vi
List of Figures viii
Abbreviations and Acronyms xiii
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 APPROACH 1-8
REFERENCES 1-10
2. AIR QUALITY CHARACTERIZATION 2-1
2.1 INTRODUCTION 2-1
2.2 CHARACTERIZATION OF AMBIENT PARTICULATE MATTER 2-1
2.2.1 Particle Size Distributions 2-2
2.2.1.1 Modes 2-2
2.2.1.2 Sampler Cut Points 2-6
2.2.2 Sources and Formation Processes 2-8
2.2.3 Chemical Composition 2-9
2.2.4 Fate and Transport 2-10
2.3 SOURCE EMISSIONS 2-12
2.3.1 Primary PM Emissions 2-12
2.3.2 Secondary PM Precursor Emissions 2-14
2.4 AMBIENT PM MEASUREMENT METHODS 2-15
2.4.1 Particle Mass Measurement Methods 2-15
2.4.2 Size-Differentiated Particle Number Concentration Measurement Methods
2-17
2.4.3 Chemical Composition Measurement Methods 2-17
2.4.4 Measurement Issues 2-18
2.5 PM CONCENTRATIONS, TRENDS, AND SPATIAL PATTERNS IN THE U.S.
2-19
2.5.1 PM25 2-19
2.5.2 PM10 2-29
2.5.3 PM10.2.5 2-29
2.5.4 Ultrafme Particles 2-37
2.5.5 Components of PM 2-38
August 2003 i Draft - Do Not Quote or Cite
-------
2.5.6 Relationships Among PM25, PM10, and PM10_25 2-41
2.6 TEMPORAL PATTERNS IN PM CONCENTRATIONS 2-42
2.6.1 PM2 5 and PM10.2 5 Patterns 2-42
2.6.2 Ultrafme Patterns 2-48
2.7 PM BACKGROUND LEVELS 2-51
2.8 RELATIONSHIP BETWEEN HUMAN EXPOSURE TO AMBIENT PM AND
CENTRAL MONITOR MEASUREMENTS OF PM 2-54
2.8.1 Definitions 2-55
2.8.2 Centrally Monitored PM Concentration as a Surrogate for Particle
Exposure 2-56
2.9 OPTICAL AND RADIATIVE PROPERTIES OF PARTICLES 2-61
2.9.1 PM Properties Affecting Visibility 2-61
2.9.2 PM Properties Affecting Climate 2-62
2.9.2 PM Properties Affecting Transmission of Ultraviolet Radiation .... 2-64
REFERENCES 2-65
3. CHARACTERIZATION OF PM-RELATED HEALTH EFFECTS 3-1
3.1 INTRODUCTION 3-1
3.2 MECHANISMS 3-3
3.3 NATURE OF EFFECTS 3-9
3.3.1 Premature Mortality 3-12
3.3.1.1 Mortality and Short-term PM Exposure 3-13
3.3.1.1.1 Multi-city Studies of Total Daily Mortality 3-16
3.3.1.1.2 Other Studies of Total Daily Mortality 3-23
3.3.1.1.3 Cause-specific Daily Mortality 3-29
3.3.1.2 Mortality and Long-term PM Exposure 3-31
3.3.1.3 Infant Mortality 3-40
3.3.1.4 Mortality Displacement and Life-Shortening 3-41
3.3.2 Indices of Morbidity 3-43
3.3.2.1 Hospital Admissions and Emergency Room Visits 3-44
3.3.2.2 Effects on the Respiratory System 3-53
3.3.2.2.1 Acute Respiratory Effects 3-53
3.3.2.2.2 Chronic Respiratory Effects 3-58
3.3.2.3 Effects on the Cardiovascular System 3-62
3.3.2.4 Developmental Effects 3-65
3.4 SENSITIVE GROUPS FOR PM-RELATED HEALTH EFFECTS 3-65
3.4.1 Preexisting disease as a risk factor 3-65
3.4.2Age-related risk factors 3-67
3.5 EVALUATION OF PM-RELATED HEALTH EFFECTS EVIDENCE .... 3-68
3.5.1 Role of Gaseous Co-pollutants 3-69
3.5.2 Issues Related to Epidemiological Modeling 3-74
3.5.2.1 Model Specification 3-75
3.5.2.2 Lag Periods 3-78
3.5.2.3 Measurement Error 3-80
August 2003 ii Draft - Do Not Quote or Cite
-------
3.5.3 Consideration of Evidence on PM Size and Composition 3-82
3.5.3.1 Ultrafme particles 3-82
3.5.3.2 PM Components and Source-Related Particles 3-83
3.5.4 Consistency and Coherence of Health Effects Evidence 3-89
3.5.4.1 Consistency 3-89
3.5.4.2 Coherence 3-98
REFERENCES 3-102
4. CHARACTERIZATION OF HEALTH RISKS 4-1
4.1 INTRODUCTION 4-1
4.1.1 Goals for Updated PM Risk Assessment 4-3
4.1.2 Summary of Risk Assessment Conducted During Prior PM NAAQS
Review 4-3
4.2 GENERAL SCOPE OF PM RISK ASSESSMENT 4-5
4.2.1 Overview of Components of the Risk Model 4-9
4.2.2 Criteria for Selection of Health Endpoints and Urban Study Areas .. 4-10
4.2.2.1 Selection of Health Endpoint Categories 4-12
4.2.2.2 Selection of Study Areas 4-13
4.2.3 Air Quality Considerations 4-17
4.2.3.1 Estimating PM Background Levels 4-24
4.2.3.2 Simulating PM Levels That Just Meet Specified Standards . . 4-24
4.2.4 Approach to Estimating PM-Related Health Effects Incidence 4-27
4.2.5 Baseline Health Effects Incidence Rates and Population Estimates .. 4-29
4.2.6 Concentration-Response Functions Used in Risk Assessment 4-30
4.2.6.1 Hypothetical Thresholds 4-31
4.2.6.2 Single and Multi-Pollutant Models 4-33
4.2.6.3 Single, Multiple, and Distributed Lag Functions 4-34
4.2.6.4 Long-term Exposure Mortality PM2 5 Concentration-Response
Functions 4-35
4.2.7 Characterizing Uncertainty and Variability 4-36
4.3 PM2 5, PM10.2 5, and PM10 RISK ESTIMATES FOR CURRENT ("AS IS") AIR
QUALITY 4-40
4.3.1 Summary of Risk Estimates 4-40
4.3.2 Sensitivity Analyses 4-58
4.3.3 Key Observations 4-64
4.4 RISK ESTIMATES ASSOCIATED WITH JUST MEETING THE CURRENT
PM25 STANDARDS 4-67
4.4.1 Summary of Risk Estimates 4-67
4.4.2 Sensitivity Analyses 4-71
4.4.3 Key Observations 4-73
4.5 RISK ESTIMATES ASSOCIATED WITH JUST MEETING ALTERNATIVE
PM25 AND PM10.25 STANDARDS 4-73
REFERENCES 4-74
August 2003 iii Draft - Do Not Quote or Cite
-------
5. CHARACTERIZATION OF PM-RELATED WELFARE EFFECTS 5-1
5.1 INTRODUCTION 5-1
5.2 EFFECTS ON VISIBILITY 5-2
5.2.1 Overview of Visibility Impairment 5-3
5.2.2 Effects of PM on Visibility 5-4
5.2.2.1 Measures of Visibility Impairment 5-4
5.2.2.2 Rayleigh Scattering and Natural Background Conditions 5-9
5.2.2.3 Contribution of PMto Visibility Conditions 5-10
5.2.3 Visibility Conditions in Class I and Non-Urban Areas 5-11
5.2.3.1 IMPROVE Visibility Monitoring Network 5-11
5.2.3.2 Current Conditions and Trends Based on IMPROVE Data . . 5-12
5.2.4 Urban Visibility Conditions 5-13
5.2.4.1 ASOS Airport Visibility Monitoring Network 5-14
5.2.4.2 ASOS Data: Urban Visibility and Correlation to PM25 Mass
' 5-15
5.2.5 Significance of Visibility to Public Welfare 5-18
5.2.5.1 Value of Improving Visual Air Quality 5-18
5.2.5.2 Visibility Goals and Programs 5-21
5.2.6 Evaluating Public Perceptions of Visibility Impairment 5-27
5.2.6.1 Photographic Representations of Visibility Impairment 5-27
5.2.6.2 Future Assessment of Public Opinions on Air Pollution-Related
Visibility Impairment 5-28
5.2.7 Summary 5-29
5.3 EFFECTS ON MATERIALS 5-31
5.3.1 Materials Damage Effects 5-32
5.3.2 Soiling Effects 5-34
5.3.3 Summary 5-35
5.4 EFFECTS ON VEGETATION AND ECOSYSTEMS 5-35
5.4.1 PM Characterization 5-36
5.4.2 PM Deposition 5-37
5.4.3 Direct Effects on Vegetation 5-39
5.4.4 Ecosystem Effects 5-43
5.4.5 Urban Ecosystems 5-60
5.4.6 Rural PM Air Quality Networks 5-62
5.4.7 Summary 5-62
5.5 EFFECTS ON CLIMATE CHANGE AND SOLAR RADIATION 5-63
5.5.1 Climate Change and Potential Human Health and Environmental Impacts
5-64
5.5.2 Alterations in Solar UV-B Radiation and Potential Human Health and
Environmental Impacts 5-65
5.5.3 Summary 5-67
REFERENCES 5-69
August 2003 iv Draft - Do Not Quote or Cite
-------
6. STAFF CONCLUSIONS AND RECOMMENDATIONS ON PM NAAQS 6-1
6.1 INTRODUCTION 6-1
6.2 ADEQUACY OF CURRENT PRIMARY PM NAAQS 6-2
6.2.1 Adequacy of Current PM25 Standards 6-3
6.2.2 Adequacy of Current PM10 Standards 6-8
6.3 ALTERNATIVE PRIMARY STANDARDS FOR FINE-FRACTION
PARTICLES 6-9
6.3.1 Alternative Indicators for Fine-Fraction Particle Standards 6-9
6.3.2 Alternative Averaging Times for PM25 Standards 6-13
6.3.3 Alternative Annual PM25 Standards 6-15
6.3.4 Alternative 24-Hour PM25 Standards 6-22
6.4 ALTERNATIVE PRIMARY STANDARDS FOR COARSE-FRACTION
PARTICLES 6-26
6.4.1 Alternative Indicators for Coarse-Fraction Particle Standards 6-27
6.4.2 Alternative Averaging Times for PM10_2 5 Standards 6-28
6.4.3 Alternative Annual PM10_2 5 Standards 6-30
6.4.4 Alternative 24-Hour PM10_2 5 Standards 6-33
6.5 SUMMARY OF STAFF RECOMMENDATIONS ON PRIMARY PM NAAQS
6-39
6.6 ALTERNATIVE SECONDARY STANDARDS FORPM 6-41
6.6.1 Alternative Standards for Visibility Protection 6-41
6.6.2 Alternative Standards for Other Welfare Effects 6-46
6.6.3 Summary of Staff Recommendations on Secondary PM NAAQS ...6-50
6.7 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
RECOMMENDATIONS RELATED TO STANDARD SETTING 6-52
REFERENCES 6-57
APPENDIX A:
Estimated Total Cardiovascular and Respiratory Mortality and Morbidity Effects
per Increments in 24-h Concentrations of PM10, PM25 and PM10_25 from U.S. and
Canadian Studies A-l
August 2003 v Draft - Do Not Quote or Cite
-------
List of Tables
Number Page
2-1 Particle Size Fraction Terminology Used in Staff Paper 2-3
2-2 Comparison of Ambient Particles: Fine Mode (Nuclei Mode plus Accumulation
Mode) and Coarse Mode 2-11
2-3 Annual National Emissions Estimates of Primary PM and PM Precursors
(thousand short tons), 1999 2-13
2-4 Summary of PM25 FRM Data Analysis in 27 Metro Areas, 1999-2001 2-28
2-5 Summary of Estimated PM10_25 analysis in 17 Metro Areas, 1999-2001 2-36
2-6 Estimated Range of Annual Average PM Regional Background Levels 2-53
3-1 Summary of Potential Mechanisms Based on Emerging Toxicological Evidence ... 3-7
3-2 Results of U.S. and Canadian Multi-city Studies on Associations between Short-
term PM Exposure and Mortality 3-17
3-3 Effect Estimates for Mortality per Increments in Long-term Mean Levels of Fine-
and Coarse-fraction Particle Indicators from U.S. and Canadian Studies 3-34
3-4 Effect Estimates for Morbidity per Increments in Long-term Mean Levels of Fine-
and Coarse-fraction Particle Indicators from U.S. and Canadian Studies 3-60
4-1 Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the
PM25 Risk Assessment 4-18
4-2 Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in
the PM2 5 Risk Assessment 4-19
4-3 Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the
PM10_2 5 Risk Assessment 4-20
4-4 Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in
the PM10_2 5 Risk Assessment 4-20
August 2003 vi Draft - Do Not Quote or Cite
-------
4-5 Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the
PM10 Risk Assessment 4-21
4-6 Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in
the PM10 Risk Assessment 4-23
4-7 Summary of PM Ambient Air Quality Data for Risk Assessment Study Areas 4-25
4-8 Summary of Estimated PM Background Concentrations Used in Risk Assessment. . 4-26
4-9 Sensitivity Analyses and Quantitative Comparisons 4-41
4-10 Sensitivity Analysis: Estimated Annual Health Risks of Short-Term and Long-
Term Exposure Mortality Associated with "As Is" PM2 5 Concentrations Using
Alternative Cutpoint Models, Detroit, MI, 2000 4-63
4-11 Sensitivity Analysis: The Effect of Assumptions About Historical Air Quality on
Estimates of Long-Term Exposure Mortality Associated with "As Is" PM25
Concentrations, Detroit, MI, 2000 4-65
4-12 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 4-70
4-13 Comparison of Annual Estimates of Short- and Long-Term Exposure Mortality
Reductions Associated with Just Meeting the Current PM2 5 Standards 4-72
6-1 Predicted Percentage of Counties with Monitors Not Likely to Meet Alternative
PM2 5 Standards 6-26
6-2 Predicted Percentage of Counties with Monitors Not Likely to Meet Alternative
PM10_2 5 and Current PM10 Standards 6-37
August 2003 vii Draft - Do Not Quote or Cite
-------
List of Figures
Number Page
2-1 Distribution of coarse, accumulation, and nucleation-mode particles by three
characteristics 2-5
2-2 An idealized distribution of ambient particulate matter PM showing fine and
coarse particles by size-selective samplers 2-7
2-3 Regions used in Staff Paper PM data analysis summaries 2-20
2-4 Distribution of annual mean PM2 5 and estimated annual mean PM10_2 5
concentrations by region, 2000-2002 2-21
2-5 Distribution of 98th percentile 24-hour average PM2 5 and estimated PM10_2 5
concentrations by region (|ig/m3), 2000-2002 2-22
2-6 County-level maximum annual mean PM2 5 concentrations, 2000-2002 2-23
2-7 County-level maximum 98th percentile 24-hour average PM2 5 concentration,
2000-2002 ' 2-24
2-8 Average measured annual mean PM2 5 concentration trend to IMPROVE sites,
1992-2001 2-26
2-9 County-level maximum annual mean PM10 concentrations, 2000-2002 2-30
2-10 County-level maximum 98th percentile 24-hour average PM10 concentrations,
2000-2002 2-31
2-11 Estimated county-level maximum annual mean PM10_25 concentrations, 2000-
2002 2-33
2-12 Estimated county-level maximum 98th percentile 24-hour average PM10_25
concentrations,, 2000-2002 2-34
2-13 Average measured annual average PM10_25 concentration trend at IMPROVE
sites, 1992-2001 2-35
2-14 Annual average composition of PM25 by region 2-39
2-15 Average annual average trend in fine particle ammonium sulfate, total carbon, and
crustal material at IMPROVE sites, 1992-2001 2-40
August 2003 viii Draft - Do Not Quote or Cite
-------
2-16 Distribution of ratios of annual mean PM25 and PM10 by region, 2000-2002 2-42
2-17 Regional average correlations of 24-hour average PM by size fraction 2-43
2-18 Urban 24-hour average PM2 5 concentration distributions by region and month,
2000-2002 ' 2-44
2-19 Urban 24-hour average estimated PM10_25 concentration distributions by region
and month, 2000 - 2002 2-45
2-20 Distribution of annual mean vs. 98th percentile 24-hour average PM25
concentrations, 2000 - 2002 2-47
2-21 Distribution of estimated annual mean vs. 98th percentile 24-hour average PM10_2 5
concentrations, 2000-2002 2-48
2-22 Hourly average PM_2 5 concentrations at a Cleveland, OH monitoring site, 2000-
2002 ' 2-49
2-23 Hourly average PM10_2 5 concentrations at a Cleveland, OH monitoring site, 2000-
2002 '. 2-50
3-1 Estimated excess risks for PM mortality (1 day lag) for the 88 largest U.S. cities
as shown in the revised NMMAPS analysis 3-20
3-2 An EPA-derived plot showing relationship of PM10 total mortality effects
estimates and 95% confidence intervals for all cities in the Samet et al. (2000a;
Domenici, 2003) NMMAPS 90-cities analyses in relation to study size 3-21
3-4 Effect estimates for PM10 and mortality from total, respiratory and cardiovascular
causes from U.S. and Canadian cities 3-26
3-5 Effect estimates for PM2 5 and mortality from total, respiratory and cardiovascular
causes from U.S. and Canadian cities 3-27
3-6 Effect estimates for PM10_2 5 and mortality from total, respiratory and
cardiovascular causes from U.S. and Canadian cities 3-28
3-7 Effect estimates for PM10 and hospital admissions, emergency room visits or
physicians office visits for various respiratory and cardiovascular diseases from
U.S. and Canadian studies 3-47
August 2003 ix Draft - Do Not Quote or Cite
-------
3-8 Effect estimates for PM2 5 and hospital admissions or emergency room visits for
various respiratory and cardiovascular diseases from U.S. and Canadian studies. . . 3-48
3-9 Effect estimates for PM10_2 5 and hospital admissions or emergency room visits for
various respiratory and cardiovascular diseases from U.S. and Canadian studies. . . . 3-49
3-10 Marginal posterior distributions for effect of PM10 on total mortality at lag 1 with
and without control for other pollutants, for the 90 cities 3-72
3-1 la Estimated excess mortality and morbidity risks per 25 |ig/m3 PM25 from U.S. and
Canadian studies 3-92
3-1 Ib Estimated excess mortality and morbidity risks per 25 |ig/m3 PM10_25 from U.S.
and Canadian studies 3-93
3-12 Associations between PM25 and total mortality from U.S. studies, plotted against
gaseous pollutant concentrations from the same locations 3-97
4-1 Major Components of Parti culate Matter Health Risk Assessment 4-11
4-2 Legend for Health Endpoints in Figures 4-42
4-3 Estimated Annual Percent of Total (Non-Accidental) Mortality Associated with
Short-Term Exposure to PM2 5 (and 95 Percent Confidence Interval): Single-
Pollutant, Single-City Models 4-43
4-4 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM2 5 (and 95 Percent Confidence Interval): Results Based on Single-
City versus Multi-City Models 4-44
4-5 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM2 5 (and 95 Percent Confidence Interval): Results Based on Single-
Pollutant versus Multi-Pollutant Models 4-45
4-6 Estimated Annual Percent of Mortality Associated with Short-Term Exposure to
PM25 (and 95 Percent Confidence Interval): Effect of Different Lag Models 4-46
4-7 Estimated Annual Percent of Mortality Associated with Long-Term Exposure to
PM2 5 (and 95 Percent Confidence Interval): Single-Pollutant Models 4-47
4-8 Estimated Annual Percent of Mortality Associated with Long-Term Exposure to
PM2 5 (and 95 Percent Confidence Interval): Single-Pollutant and Multi-Pollutant
Models 4-48
August 2003 x Draft - Do Not Quote or Cite
-------
4-9 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM10_2 5 (and 95 Percent Confidence Interval) 4-50
4-10 Estimated Annual Percent of Total (Non-Accidental) Mortality Associated with
Short-Term Exposure to PM10 (and 95 Percent Confidence Interval): Single-
Pollutant, Single-City Models 4-52
4-11 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM10 (and 95 Percent Confidence Interval): Results Based on Single-
City versus Multi-City Models: Mortality 4-53
4-12 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM10 (and 95 Percent Confidence Interval): Results Based on Single-
Pollutant versus Multi-Pollutant Models 4-54
4-13 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM10 (and 95 Percent Confidence Interval): Results Based on Single-
City versus Multi-City Models: Morbidity 4-56
4-14 Estimated Annual Percent of Mortality Associated with Short-Term Exposure to
PM10 (and 95 Percent Confidence Interval): Effect of Different Lag Models 4-57
4-15a Distribution of 24-Hour PM25 Concentrations in St. Louis (2002 Air Quality
Data) 4-68
4-15b Estimated Non-Accidental Mortality in St. Louis Associated with PM25
Concentrations (2002 Air Quality Data) 4-68
4-16a Distribution of 24-Hour PM10_2 5 Concentrations in Phoenix (1997 Air Quality
Data) 4-69
4-16b Estimated Cardiovascular Mortality in Phoenix Associated with PM10_2 5
Concentrations (1997 Air Quality Data) 4-69
5-1 Relationship between Light Extinction, Deciviews, and Visual Range 5-8
5-2 Relationship between Quarterly Distributions of Estimated Correlation
Coefficients and Longitude of ASOS Sites 5-16
5-3 Relationship between Predicted and Measured PM2 5 Concentrations in
Washington, DC 5-17
August 2003 xi Draft - Do Not Quote or Cite
-------
5-4 Relationship between Predicted and Measured PM2 5 Concentrations in
Milwaukee, WI 5-17
Figures 5-5 through 5-34 are available on the EPA website
http:/www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html
5-5 through 5-12 Images of visual air quality in Denver, CO under a range of
visibility conditions
5-13 through 5-20 Images of visual air quality in Phoenix, AZ under a range of
visibility conditions
5-21 through 5-28 Images of visual air quality in Washington, DC under a range of
visibility conditions
5-29 through 5-34 Images of visual air quality in Chicago, IL under a range of
visibility conditions
5-35 Illustration of the nitrogen cascade showing the movement of the human-
produced reactive nitrogen (Nr) as it cycles through the various environmental
reservoirs in the atmosphere, terrestrial and aquatic ecosystems 5-47
August 2003 xii Draft - Do Not Quote or Cite
-------
Abbreviations and Acronyms
AC
ACS
AHSMOG
AIRS
APHEA
AQCD
AQS
ASOS
BC
BS
CAA
CAMM
CAP
CASAC
CASTNet
CB
CD
CDC
CDPHE
CFR
C:N
CO
COH
COPD
C-R
CSS
cv
EC
Automated colorimetry
American Cancer Society
Adventist Health and Smoke Study
Aerometric Information Retrieval System
Air Pollution and Health, a European Approach
Air Quality Criteria Document
Air Quality System
Automated Surface Observing System
Black carbon
British or black smoke
Clean Air Act
Continuous Ambient Mass Monitor
Concentrated ambient particles
Clean Air Scientific Advisory Committee
Clean Air Status and Trends Network
Base cation
Criteria Document
Centers for Disease Control
Colorado Department of Public Health and Environment
Code of Federal Regulations
Carbon-to-nitrogen ratio
Carbon monoxide
Coefficient of haze
Chronic obstructive pulmonary disease
Concentration-response
Coastal sage scrub community
Contingent valuation
Elemental carbon
August 2003
Xlll
Draft - Do Not Quote or Cite
-------
ECG Electrocardiogram
EEA Essential Ecological Attribute
EPA Environmental Protection Agency
EPEC
Framework Ecological Processes and Effects Subcommittee document, Framework for
Assessing and Reporting on Ecological Condition
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
hosp. adm. Hospital admissions
1C Ion chromatography
IFS Integrated Forest Study
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
NMMAPS National Mortality and Morbidity Air Pollution Study
N2 Nonreactive, molecular nitrogen
NO2 Nitrogen dioxide
non-accid
mort Non-accidental mortality
August 2003
xiv
Draft - Do Not Quote or Cite
-------
Nr Reactive nitrogen
NSMPS Nano-scanning mobility particle sizer
NuCM Nutrient cycling model
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
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
PM10_2 5 Particles less than or equal to 10 jim in diameter and greater than 2.5 jim in
diameter
Particles less than or equal to 2.5 jim in diameter
10 Particles less than or equal to 10 jim in diameter
POPs Persistent organic pollutants
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
TMO Thermal manganese oxidation method
TOR Thermal/optical reflectance method
PM
PM
August 2003
XV
Draft - Do Not Quote or Cite
-------
TOT Thermal/optical transmission method
TSP Total suspended particulates
|ig micrograms
l-ig/m3 micrograms per cubic meter
URS Upper respiratory symptoms
U.S. United States
UV Ultraviolet
UV-B Ultraviolet-B
Vd Deposition velocity
VOCs Volatile organic compounds
XRF X-ray fluorescence
August 2003 xvi Draft - Do Not Quote or Cite
-------
1 1. INTRODUCTION
2
3 1.1 PURPOSE
4 This draft Staff Paper, prepared by the U.S. Environmental Protection Agency's (EPA)
5 Office of Air Quality Planning and Standards (OAQPS), identifies the key policy-relevant
6 scientific information contained in the draft document, Air Quality Criteria for Particulate
1 Matter - Fourth External Review Draft (EPA, 2003; henceforth referred to as the draft Criteria
8 Document (CD) and cited as CD), prepared by EPA's National Center for Environmental
9 Assessment (NCEA). Building upon an earlier preliminary draft Staff Paper (EPA, 2001), this
10 document includes results from initial staff analyses (e.g., analyses of air quality and visibility
11 data, and human health risk analyses), identifies additional analyses to be incorporated into a
12 subsequent draft of this document, and presents preliminary staff conclusions and
13 recommendations.
14 The final version of this Staff Paper will evaluate the policy implications of the key
15 studies and scientific information contained in the final EPA document, Air Quality Criteria for
16 Particulate Matter (targeted for completion by the end of 2003), and identify the critical
17 elements that EPA staff believes should be considered in the current review of the national
18 ambient air quality standards (NAAQS) for particulate matter (PM). This policy assessment is
19 intended to help "bridge the gap" between the scientific review contained in the CD and the
20 judgments required of the EPA Administrator in determining whether it is appropriate to revise
21 the NAAQS for PM. Emphasis will be placed on identifying those conclusions and
22 uncertainties in the available scientific literature that the staff believes should be considered in
23 selecting PM indicators, averaging times, forms1, and levels for the primary (health-based) and
24 secondary (welfare-based) standards, which must be considered collectively in evaluating the
25 health and welfare protection afforded by PM standards. The final Staff Paper will present
26 factors relevant to the evaluation of current primary and secondary PM NAAQS, as well as staff
27 conclusions and recommendations of options for the Administrator to consider.
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.
August 2003 1 -1 Draft - Do Not Quote or Cite
-------
1 While this document should be of use to all parties interested in the PM NAAQS review,
2 it is written for those decision makers, scientists, and staff who have some familiarity with the
3 technical discussions contained in the CD.
4
5 1.2 BACKGROUND
6 1.2.1 Legislative Requirements
7 Two sections of the Clean Air Act (Act) govern the establishment and revision of the
8 NAAQS. Section 108 (42 U.S.C. 7408) directs the Administrator to identify "air pollutants" that
9 "in his judgment, may reasonably be anticipated to endanger public health and welfare" and
10 whose "presence ... in the ambient air results from numerous or diverse mobile or stationary
11 sources" and, if listed, to issue air quality criteria for them. These air quality criteria are
12 intended to "accurately reflect the latest scientific knowledge useful in indicating the kind and
13 extent of identifiable effects on public health or welfare which may be expected from the
14 presence of [a] pollutant in ambient air ... ."
15 Section 109 (42 U.S.C. 7409) directs the Administrator to propose and promulgate
16 "primary" and "secondary" NAAQS for pollutants identified under section 108. Section
17 109(b)(l) defines a primary standard as one "the attainment and maintenance of which in the
18 judgment of the Administrator, based on such criteria and allowing an adequate margin of safety,
19 are requisite to protect the public health."2 A secondary standard, as defined in Section
20 109(b)(2), must "specify a level of air quality the attainment and maintenance of which, in the
21 judgment of the Administrator, based on such criteria, is requisite to protect the public welfare
22 from any known or anticipated adverse effects associated with the presence of [the] pollutant in
23 the ambient air."3
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."
August 2003 1 -2 Draft - Do Not Quote or Cite
-------
1 In setting standards that are "requisite" to protect public health and welfare, as provided
2 in section 109(b), EPA's task is to establish standards that are neither more nor less stringent
3 than necessary for these purposes. In so doing, EPA may not consider the costs of implementing
4 the standards. See generally Whitman v. American Trucking Associations, 531 U.S. 457, 464,
5 475-76 (2001).
6 The requirement that primary standards include an adequate margin of safety was
7 intended to address uncertainties associated with inconclusive scientific and technical
8 information available at the time of standard setting. It was also intended to provide a
9 reasonable degree of protection against hazards that research has not yet identified. Lead
10 Industries Association v. EPA, 647 F.2d 1130, 1154 (D.C. Cir 1980), cert, denied. 101 S. Ct. 621
11 (1980); American Petroleum Institute v. Costle, 665 F.2d 1176, 1177 (D.C. Cir. 1981), cert.
12 denied. 102 S.Ct. 1737 (1982). Both kinds of uncertainties are components of the risk associated
13 with pollution at levels below those at which human health effects can be said to occur with
14 reasonable scientific certainty. Thus, in selecting primary standards that include an adequate
15 margin of safety, the Administrator is seeking not only to prevent pollution levels that have been
16 demonstrated to be harmful but also to prevent lower pollutant levels that may pose an
17 unacceptable risk of harm, even if the risk is not precisely identified as to nature or degree.
18 In selecting a margin of safety, the EPA considers such factors as the nature and severity
19 of the health effects involved, the size of the sensitive population(s) at risk, and the kind and
20 degree of the uncertainties that must be addressed. The selection of any particular approach to
21 providing an adequate margin of safety is a policy choice left specifically to the Administrator's
22 judgment. Lead Industries Association v. EPA, supra. 647 F.2d at 1161-62.
23 Section 109(d)(l) of the Act requires that "not later than December 31, 1980, and at 5-
24 year intervals thereafter, the Administrator shall complete a thorough review of the criteria
25 published under section 108 and the national ambient air quality standards . . . and shall make
26 such revisions in such criteria and standards and promulgate such new standards as may be
27 appropriate . . . ." Section 109(d)(2) requires that an independent scientific review committee
28 "shall complete a review of the criteria . . . and the national primary and secondary ambient air
29 quality standards . . . and shall recommend to the Administrator any new . . . standards and
August 2003 1 -3 Draft - Do Not Quote or Cite
-------
1 revisions of existing criteria and standards as may be appropriate . . . ." Since the early 1980's,
2 this independent review function has been performed by the Clean Air Scientific Advisory
3 Committee (CASAC), which is administered by EPA's Science Advisory Board Staff Office.
4
5 1.2.2 History of PM NAAQS Reviews
6 EPA first established national ambient air quality standards for PM in 1971, based on the
7 original criteria document (DHEW, 1969). Particulate matter is the generic term for a broad
8 class of chemically and physically diverse substances that exist as discrete particles (liquid
9 droplets or solids) over a wide range of sizes. Particles originate from a variety of anthropogenic
10 stationary and mobile sources as well as natural sources. Particles may be emitted directly or
11 formed in the atmosphere by transformations of gaseous emissions such as sulfur oxides,
12 nitrogen oxides, and volatile organic compounds. The chemical and physical properties of PM
13 vary greatly with time, region, meteorology, and source category, thus complicating the
14 assessment of health and welfare effects.
15 The reference method specified for determining attainment of the original standards was
16 the high-volume sampler, which collects PM up to a nominal size of 25 to 45 micrometers (|im)
17 (referred to as total suspended particles or TSP). The primary standards (measured by the
18 indicator TSP) were 260 |ig/m3, 24-hour average, not to be exceeded more than once per year,
19 and 75 |ig/m3, annual geometric mean. The secondary standard was 150 |ig/m3, 24-hour average,
20 not to be exceeded more than once per year.
21 In October 1979 (44 FR 56731), EPA announced the first periodic review of the criteria
22 and NAAQS for PM, and significant revisions to the original standards were promulgated in
23 1987 (52 FR 24854, July 1, 1987). In that decision, EPA changed the indicator for particles from
24 TSP to PM10, the latter referring to particles with a mean aerodynamic diameter4 less than or
25 equal to 10 jim. EPA also revised the level and form of the primary standards by: (1) replacing
26 the 24-hour TSP standard with a 24-hour PM10 standard of 150 |ig/m3 with no more than one
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.
August 2003 1 -4 Draft - Do Not Quote or Cite
-------
1 expected exceedance per year; and (2) replacing the annual TSP standard with a PM10 standard
2 of 50 |ig/m3, annual arithmetic mean. The secondary standard was revised by replacing it with
3 24-hour and annual standards identical in all respects to the primary standards. The revisions
4 also included a new reference method for the measurement of PM10 in the ambient air and rules
5 for determining attainment of the new standards. On judicial review, the revised standards were
6 upheld in all respects. Natural Resources Defense Council v. Administrator, 902 F. 2d 962 (D.C.
7 Cir. 1990). cert, denied. Ill S. Ct. 952 (1991).
8 In December 1994, EPA presented its plan for the second periodic review of the criteria
9 and NAAQS for PM to the CASAC, and promulgated significant revisions to the NAAQS in
10 1997 (62 FR 38652, July 18, 1997). In that decision, EPA revised the PM NAAQS in several
11 respects. While it was determined that the PM NAAQS should continue to focus on particles
12 less than or equal to 10 jim in diameter, it was also determined that the fine and coarse fractions
13 of PM10 should be considered separately. New standards were added, using PM25, referring to
14 particles with a mean aerodynamic diameter less than or equal to 2.5 jim, as the indicator for fine
15 particles, with PM10 standards retained for the purpose of regulating coarse-fraction particles.
16 EPA established two new PM25 standards: an annual standard of 15 |ig/m3, based on the 3-year
17 average of annual arithmetic mean PM25 concentrations from single or multiple community -
18 oriented monitors; and a 24-hour standard of 65 |ig/m3, based on the 3-year average of the 98th
19 percentile of 24-hour PM25 concentrations at each population-oriented monitor within an area.
20 A new reference method for the measurement of PM25 in the ambient air was also established, as
21 were rules for determining attainment of the new standards. To continue to address coarse-
22 fraction particles, the annual PM10 standard was retained, while the 24-hour PM10 standard was
23 revised to be based on the 99th percentile of 24-hour PM10 concentrations at each monitor in an
24 area. EPA revised the secondary standards by making them identical in all respects to the
25 primary standards.
26
27 1.2.3 Litigation Related to 1997 PM Standards
28 Following promulgation of the revised PM NAAQS, petitions for review were filed by a
29 large number of parties, addressing a broad range of issues. In May 1998, a three-judge panel of
August 2003 1-5 Draft - Do Not Quote or Cite
-------
1 the U.S. Court of Appeals for the District of Columbia Circuit issued an initial decision that
2 upheld EPA's decision to establish fine particle standards, holding that "the growing empirical
3 evidence demonstrating a relationship between fine particle pollution and adverse health effects
4 amply justifies establishment of new fine particle standards." American Trucking Associations v.
5 EPA , 175 F. 3d 1027, 1055-56 (D.C. Cir. 1999) (rehearing granted in part and denied in part,
6 195 F. 3d 4 (D.C. Cir. 1999), affirmed in part and reversed in part, Whitman v. American
1 Trucking Associations., 531 U.S. 457 (2001). The Panel also found "ample support" for EPA's
8 decision to regulate coarse particle pollution, but vacated the 1997 revised PM10 standards,
9 concluding in part that PM10 is a "poorly matched indicator for coarse particulate pollution"
10 because it includes fine particles. Id. at 1053-55. As a result of this aspect of the court's ruling,
11 which EPA did not appeal, the 1987 PM10 standards remain in effect. In the current review, EPA
12 is addressing coarse-fraction particles in the current review by considering standards based on an
13 indicator of PM10_25, referring to particles with a mean aerodynamic diameter greater than 2.5 jim
14 but less than or equal to 10 jim.
15 More generally, the Panel held (with one dissenting opinion) that EPA's approach to
16 establishing the level of the standards in 1997, both for PM and for ozone NAAQS promulgated
17 on the same day, effected "an unconstitutional delegation of legislative authority." Id. at 1034-
18 40. Although the Panel stated that "the factors EPA uses in determining the degree of public
19 health concern associated with different levels of ozone and PM are reasonable," it remanded the
20 NAAQS to EPA, stating that when EPA considers these factors for potential non-threshold
21 pollutants "what EPA lacks is any determinate criterion for drawing lines" to determine where
22 the standards should be set. Consistent with EPA's long-standing interpretation, the Panel also
23 reaffirmed prior rulings holding that in setting NAAQS EPA is "not permitted to consider the
24 cost of implementing those standards." Id. at 1040-41.
25 Both sides filed cross appeals on these issues to the United States Supreme Court, and
26 the Court granted certiorari. In February 2001, the Supreme Court issued a unanimous decision
27 upholding EPA's position on both issues. Whitman v. American Trucking Associations, 531 U.S.
28 457, 464, 475-76. On the constitutional issue, the Court held that the statutory requirement that
29 NAAQS be "requisite" to protect public health with an adequate margin of safety sufficiently
August 2003 1 -6 Draft - Do Not Quote or Cite
-------
1 guided EPA's discretion, affirming EPA's approach of setting standards that are neither more
2 nor less stringent than necessary. The Supreme Court remanded the case to the Court of
3 Appeals for resolution of any remaining issues that had not been addressed in that court's earlier
4 rulings. Id. at 475-76. In March 2002, the Court of Appeals rejected all remaining challenges to
5 the standards, holding under the traditional standard of review that EPA's PM2 5 standards were
6 reasonably supported by the administrative record and were not "arbitrary and capricious."
7 American Trucking Associations v. EPA, 283 F. 3d 355, 369-72 (D.C. Cir. 2002). Thus, the
8 1997 PM25 standards are in effect.
9
10 1.2.4 Current PM NAAQS Review
11 In October 1997, EPA published its plans for the current periodic review of the PM
12 NAAQS (62 FR 55201, October 23, 1997), including the 1997 PM25 standards and the 1987
13 PM10 standards. As part of the process of preparing the PM CD, in April 1999, NCEA hosted a
14 peer review workshop on drafts of key chapters of the CD. The first external review draft CD
15 was reviewed by CAS AC and the public at a meeting held in December 1999. Based on CAS AC
16 and public comment, NCEA revised the draft CD and released a second external review draft
17 (March 2001) for review by CAS AC and the public at a meeting held in July 2001. A
18 preliminary draft Staff Paper (EPA, 2001) was released in June 2001 for public comment and for
19 consultation with CAS AC at the same public meeting. A third external review draft CD was
20 released in May 2002 for review at a meeting held in July 2002.
21 Shortly after EPA released the third external review draft CD, the Health Effects Institute
22 (HEI)5 announced that researchers at Johns Hopkins University had discovered problems with
23 applications of statistical software used in a number of important epidemiological studies that
24 had been discussed in that draft CD. In response to this significant issue, EPA took steps in
25 consultation with CASAC to encourage researchers to reanalyze affected studies and to submit
26 them expeditiously for peer review by a special expert panel convened at EPA's request by HEI.
27 EPA subsequently incorporated the results of this reanalysis and peer-review process into a
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.
August 2003 1 -7 Draft - Do Not Quote or Cite
-------
1 fourth external review draft CD, which was released in June 2003 for review by CASAC and the
2 public at an August 2003 meeting.
3 This draft Staff Paper is being provided to CASAC and the public for review at a meeting
4 planned for November 2003. Following that meeting, EPA will complete the staff analyses and
5 take CASAC and public comments into account in preparing a second draft Staff Paper, to be
6 based on the final CD, and will make that draft document available for further review and
7 comment by CASAC and the public.
8 The schedule for completion of this review is now governed by a consent decree
9 resolving a lawsuit filed in March 2003 by a group of plaintiffs representing national
10 environmental organizations. The lawsuit alleged that EPA had failed to perform its mandatory
11 duty, under section 109(d)(l) [42 U.S.C. 7409(d)(l)], of completing the current review within
12 the period provided by statute. American Lung Association v. Whitman (No. 1:03CV00778,
13 D.D.C. 2003). Entered by the court after an opportunity for public comment, the consent decree
14 provides that EPA will issue a final PM CD no later than December 19, 2003, and that EPA will
15 sign for publication notices of proposed and final rulemaking concerning its review of the PM
16 NAAQS no later than March 31, 2005 and December 20, 2005, respectively. These dates are
17 premised on the expectation that a series of interim milestones will be met, including the release
18 of a second draft PM Staff Paper by April 30, 2004, followed by CASAC and public review of
19 that draft by July 31, 2004, with completion of a final PM Staff Paper by September 30, 2004.
20
21 1.3 APPROACH
22 The final Staff Paper will take into account the scientific evidence reviewed in the final
23 CD, and will include: 1) the results of comparative air quality analyses, the final human health
24 risk assessment, and analyses examining visibility impairment; 2) the staffs overall evaluation
25 of the adequacy of the current primary and secondary NAAQS; and 3) staff conclusions and
26 recommendations as to whether any revisions are appropriate to address public health and
27 welfare effects associated with fine- and coarse-fraction particles. For these purposes, the staff
28 will assess and integrate new scientific and technical findings with information gained in
August 2003 1-8 Draft - Do Not Quote or Cite
-------
1 previous reviews in the context of those critical elements that the staff believes should be
2 considered.
3 In conducting the various technical analyses presented in this draft Staff Paper, the staff
4 has focused separately on fine- and coarse-fraction particles, building upon the conclusions
5 reached in the last review, and taking into account the new information that has become
6 available. More specifically, air quality analyses to characterize spatial and temporal air quality
7 patterns have been conducted in terms of PM25 and PM10_25 as the indicators for fine- and coarse-
8 fraction particles, respectively. Similarly, the current draft human health risk assessment focuses
9 on analyzing various health effects associated with exposure to ambient PM2 5 and PM10_2 5, as
10 well as analyses for PM10 that provide insight into the health risks associated with exposure to
11 PM2 5 and/or PM10.2 5.
12 Following this introductory chapter, this draft Staff Paper is organized into five other
13 chapters. Chapter 2 focuses on air quality characterizations (e.g., information on atmospheric
14 chemistry, sources of PM, measurement methods, and spatial and temporal patterns in ambient
15 PM concentrations based on extensive newly available air quality monitoring data) as well as
16 information on the relationships between ambient air quality and human exposure. Chapter 3
17 presents key information on PM-associated health effects, relying primarily on the review of
18 recent epidemiological and toxicological studies in the draft CD and integrating the new
19 information with findings from previous criteria and NAAQS reviews. The results of an interim
20 draft human health risk assessment are presented in Chapter 4, together with information on
21 plans to conduct additional analyses of risk reductions associated with just attaining specified
22 alternative PM25 and PM10_25 standards. Information on welfare effects of ambient PM is
23 presented in Chapter 5, together with analyses of data on visibility impairment. Chapter 6
24 presents preliminary staff conclusions and recommendations on the current PM NAAQS and
25 possible alternative PM2 5 and PM10_2 5 primary and secondary standards for consideration by the
26 Administrator. The preliminary staff recommendations on possible alternative standards,
27 together with comments received from CAS AC and the public on this draft Staff Paper, will help
28 to form the basis for the additional risk analyses to be conducted for inclusion in the second draft
29 Staff Paper.
August 2003 1 -9 Draft - Do Not Quote or Cite
-------
REFERENCES
3 Environmental Protection Agency. (2001) Review of the National Ambient Air Quality Standards for Paniculate
4 Matter: Policy Assessment of Scientific and Technical Information - Preliminary Draft OAQPS Staff
5 Paper. June.
6
7 Environmental Protection Agency. (2003) Air Quality Criteria for Paniculate - Fourth External Review Draft.
8 Research Triangle Park, NC: Office of Research and Development; report no. EPA/600/P-99/002. June.
9
10 U.S. Department of Health, Education and Welfare. (1969) Air Quality Criteria for Paniculate Matter. U.S.
11 Government Printing Office, Washington DC, AP-49.
August 2003 1-10 Draft - Do Not Quote or Cite
-------
1 2. AIR QUALITY CHARACTERIZATION
2
3 2.1 INTRODUCTION
4 This chapter defines the various classes of PM and then briefly discusses the physical and
5 chemical properties of PM in the atmosphere, sources of PM, PM measurement methods, and
6 recent PM concentrations and trends. This information is useful for interpreting the available
7 health and welfare effects information, and for making recommendations on appropriate
8 indicators for PM. Section 2.2 presents information on the basic physical and chemical
9 properties of classes of PM, and is not substantially different from information contained in the
10 1996 Criteria Document (EPA, 1996a) and Staff Paper (EPA, 1996b). Section 2.3 discusses
11 sources of ambient PM and provides a summary of national estimates of source emissions.
12 Section 2.4 presents information on the methods used to measure ambient PM and some
13 important considerations in the design of these methods. Section 2.5 presents data on PM
14 concentrations, trends, and spatial patterns in the United States. Section 2.6 provides
15 information on the temporal variability of PM. Much of the information in Sections 2.5 and 2.6
16 is derived from analyses of data collected by the nationwide networks of PM25 and PM10
17 monitors. Section 2.7 defines and discusses background levels of ambient PM. Section 2.8
18 addresses the relationships between ambient PM levels and human exposure to PM. Finally,
19 Section 2.9 summarizes relevant information on the optical and radiative effects of ambient
20 particles.
21
22 2.2 CHARACTERIZATION OF AMBIENT PARTICULATE MATTER
23 PM represents a broad class of chemically and physically diverse substances that exist as
24 discrete particles in the condensed (liquid or solid) phase. Particles can be described by size,
25 formation mechanism, origin, chemical composition, and atmospheric behavior. Fine particles
26 and coarse particles, which are defined in Section 2.2.1.1, are distinct entities with fundamentally
27 different sources and formation processes, chemical composition, atmospheric residence times
28 and behaviors, and transport distances. The 1996 Criteria Document concluded that these
29 differences justified consideration of fine and coarse particles as separate pollutants (EPA 1996a,
August 2003 2-1 Draft - Do Not Quote or Cite
-------
1 p. 13-3), and this conclusion is reiterated in the draft CD (CD, p. 2-36 and 9-4). The
2 fundamental differences between fine and coarse particles are also important considerations in
3 assessing the available health effects and exposure information.
4
5 2.2.1 Particle Size Distributions
6 Particle properties and their associated health and welfare effects differ by size. The
7 diameters of atmospheric particles span 5 orders of magnitude, ranging from 0.001 micrometers
8 to 100 micrometers (iim).1 The size and associated composition of particles determine their
9 behavior in the respiratory system, including how far the particles are able to penetrate, where
10 they deposit, and how effective the body's clearance mechanisms are in removing them.
11 Furthermore, particle size is one of the most important parameters in determining the residence
12 time of particles in ambient air, a key consideration in assessing exposure. Particle size is also a
13 major determinant of visibility impairment, a welfare effect linked to ambient particles. Particle
14 surface area, number, chemical composition, and water solubility all vary with particle size, and
15 are also a variable of the formation processes and emissions sources.
16 Common conventions for classifying particles by size include: (1) modes, based on
17 observed particle size distributions and formation mechanisms; and (2) "cut points," based on the
18 inlet characteristics of a specific PM sampling device. The terminology used in this draft Staff
19 Paper for describing these classifications is summarized in Table 2-1 and discussed in detail in
20 the following subsections.
21 2.2.1.1 Modes
22 Based on extensive examinations of particle size distributions in several U.S. locations in
23 the 1970's, Whitby (1978) found that particles display a consistent multi-modal distribution over
24 several physical metrics, such as mass or volume (CD, p. 2-7). These modes are apparent in
25 Figure 2-1, which shows average ambient distributions of particle number, surface area, and
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).
August 2003 2-2 Draft - Do Not Quote or Cite
-------
Table 2-1. Particle Size Fraction Terminology Used in Staff Paper
Term
Description
Size Distribution Modes
Coarse Particles ("Coarse Mode")
Fine Particles
Accumulation-Mode Particles
Ultrafine Particles
Aitkin-Mode Particles
Nucleation-Mode Particles
The distribution of particles larger than the intermodal minimum in
volume or mass distributions; this minimum generally occurs between 1
and 3 um. These particles usually represent the greatest amount of mass.
The distribution of particles 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, Aitkin, 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. Particles in this mode usually represent the most surface area.
A subset of fine particles with diameters below about 0.1 um,
encompassing the Aitkin and nucleation modes. Particles in this size
range are usually the most numerous.
A subset of ultrafine particles with diameters between 0.01 and 0.1 um.
Freshly formed particles with diameters below 0.01 um.
Sampling Measurements
Total Suspended Particles (TSP)
PM,,
PM2 5 ("fine fraction")
PM10_2 5 ("coarse fraction")
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 the inhalable part of coarse particles, and is an
indicator for particles that penetrate to the tracheo-bronchial and the gas-
exchange regions of the lung.
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. The collected particles
include most of the fine particles. A small portion of coarse-mode
particles may be included depending on the sharpness of the sampler
efficiency curve and the tail of the coarse-mode may extend into the fine
fraction.
Particles measured directly using a dichotomous sampler or subtraction
of particles measured by a PM25 sampler from those measured by a PM10
sampler. This measurement is an indicator for the inhalable fraction of
coarse-mode particles.
August 2003
1-"-
Draft - Do Not Quote or Cite
-------
1 volume by particle size.2 Panel (a) illustrates that by far, the largest number of ambient particles
2 in a typical distribution are very small, below 0.1 jim in diameter, while panel (c) indicates most
3 of the particle volume, and therefore most of the mass, is found in particles with diameters larger
4 than 0.1 jim.3 Most of the surface area (panel b) is between 0.1 and 1.0 jim. The surface area
5 distribution in panel (b) peaks around 0.2 jim. Distributions may vary across locations,
6 conditions, and time due to differences in sources, atmospheric conditions, topography, and the
7 age of the aerosol.
8 As illustrated in panel (c) of Figure 2-1, volume distributions typically measured in
9 ambient air in the U.S. are found to be bimodal, with overlapping tails, and an intermodal
10 minimum between 1 and 3 jim (CD, p. 2-7). The distribution of particles that are mostly larger
11 than this minimum are called "coarse particles," and the distribution of particles that are mostly
12 smaller than the minimum are called "fine particles." Fine particles can be subcategorized into
13 smaller modes: "nucleation mode," "Aitkin mode," and "accumulation mode." Together,
14 nucleation-mode and Aitkin-mode particles make up "ultrafine particles."4 The nucleation mode
15 is apparent as the largest peak in the number distribution in panel (a), and is also visible in the
16 surface area distribution in panel (b). Nucleation-mode and Aitkin-mode particles have
17 relatively low mass and grow rapidly into accumulation-mode particles, so they are not
18 commonly observed as a separate mode in volume or mass distributions. The accumulation
19 mode is apparent as the leftmost peak in the volume distribution in panel (c) and the largest peak
20 in the surface area distribution in panel (b).
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 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 by density.
4 Whitby (1978) did not identify a separate ultrafine particle mode between 0.01 and 0.1 um, and therefore
the Aitkin mode is not illustrated in Figure 2-1.
August 2003 2-4 Draft - Do Not Quote or Cite
-------
CD
.Q
E
x
IE
o
Q
en
15-
10 -
5-
Nn = 7.7x 10
DGNn= 0.013
(a)
agn=1.7
Na= 1.3x 10
DGNa = 0.069
= 2.03
Nc=4.2
DGNC=0.97
agc=2.15
0 600
O3 CM
8 ^ 400
ro "-1
t en
zs .9
W ^ 200 -
(b)
Sa = 535
DGSa = 0.19
DGSn = 0.023
Sc=41
DCS =3.1
30 H
| £20H
Vn = 0.33
DGVn = 0.031
0.001
0.01
0.1
1.0
i
10
100
Figure 2-1. Distribution of coarse [subscript c], accumulation [subscript a], and
nucleation [subscript n] -mode particles by three characteristics: Panel
(a) number [N], Panel (b) surface area [S], and Panel (c) volume [V] for
the grand average continental size distribution. Dp = particle diameter
(um, log scale); DGN = geometric mean diameter by number; DGS =
geometric mean diameter by surface area; DGV = geometric mean
diameter by volume.
Source: Whitby (1978); CD, page 2-8.
August 2003
2-5
Draft - Do Not Quote or Cite
-------
1 2.2.1.2 Sampler Cut Points
2 Another set of particle size classifications is derived from the characteristics of ambient
3 particle samplers. Particle samplers typically use size-selective inlets that are defined by their 50
4 percent cut point, which is the particle aerodynamic diameter at which exactly 50 percent of
5 particles of a specified diameter are captured by the inlet. The usual notation for these
6 definitions is "PM,.", where x refers to measurements with a 50% cut point of x jim aerodynamic
7 diameter. Because of the overlap in the distributions of fine- and coarse-mode ambient particles,
8 and the fact that samplers do not have perfect cut points, no single sampler can precisely separate
9 them. The objective of size-selective sampling is usually to measure particle size fractions with
10 some special relationship to human health impacts, visibility impairment, or emissions sources.
11 The EPA has historically defined indicators of PM for NAAQS using cut points of
12 interest. Figure 2-2 presents an idealized distribution of ambient PM showing the fractions
13 collected by size-selective samplers. Prior to 1987, the indicator for the PM NAAQS was total
14 suspended particulate matter (TSP), and was defined by the design of the High Volume Sampler
15 (Hi Vol).5 As illustrated in Figure 2-2, TSP typically includes particles with diameters less than
16 about 40 jim, but could include even larger particles under certain conditions. When EPA
17 established new PM standards in 1987, the selection of PM10 as an indicator was intended to
18 focus regulatory concern on particles small enough to pass deeply into the human respiratory
19 tract. In 1997, EPA established new standards for fine particles measured as PM25. The dashed
20 lines in Figure 2-2 illustrate the distribution of particles captured by the PM10 Federal Reference
21 Method (FRM) sampler6, including all fine and some coarse-mode particles, and the distribution
22 captured by the PM2 5 FRM sampler7, including the potential capture of a small subset of coarse-
23 mode particles.
5 40 CFR Part 50, Appendix B, Reference Method for the Determination of Suspended Paniculate Matter in
the Atmosphere (High-Volume Method).
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.
August 2003 2-6 Draft - Do Not Quote or Cite
-------
Accumulation Mode
0,1
0.2
0.5 1 2 5 10
Particle Diameter, Dp(|jm)
Total Suspended Particles (TSP) —
20
PM
10
PM
2.5
PM
10-2.5
100
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 2-16.
4
5
6
The EPA is now considering establishing standards for another PM indicator identified in
Table 2-1 as PM10_25, which represents the inhalable fraction of coarse-mode particles. Section
2.4 discusses measurement methods for this indicator. The reader should note that the terms
"fine fraction" and "coarse fraction" are used interchangeably with PM2 5 and PM10_2 5,
respectively, to refer to specific portions of the fine and coarse modes collected by size selective
samplers.
August 2003
2-7
Draft - Do Not Quote or Cite
-------
1 2.2.2 Sources and Formation Processes
2 In most locations, a variety of activities contribute to ambient PM concentrations. Fine
3 and coarse particles generally have distinct sources and formation mechanisms, although there is
4 some overlap (CD, p. 3-60). Coarse particles are generally primary particles, meaning they are
5 emitted from their source directly as particles. Most coarse particles result from mechanical
6 disruption of large particles by crushing or grinding, from evaporation of sprays, or from dust
7 resuspension. Specific sources include construction and demolition activities, mining and
8 mineral processing, sea spray, wind-blown dust, and resuspension of settled biological material
9 from soil surfaces and roads. The amount of energy required to break down primary particles
10 into smaller particles normally limits coarse particle sizes to greater than 1.0 |im diameter (EPA
11 1996a, p. 13-7). Some combustion-generated particles, such as fly ash, are also found in the
12 coarse mode.
13 By contrast, fine particles are produced in high energy processes such as combustion or
14 by atmospheric photochemistry. Common directly emitted fine particles include unburned
15 carbon particles from combustion, and nucleation-mode particles emitted as combustion-related
16 vapors that condense within seconds of being exhausted to ambient air. Fossil-fuel combustion
17 sources include motor vehicles, power generation facilities, industrial facilities, residential wood
18 burning, agricultural burning, and forest fires.
19 The formation and growth of fine particles are influenced by three processes: (1)
20 nucleation (i.e., gas molecules coming together to form a new particle); (2) condensation of
21 gases onto existing particles; and (3) coagulation of particles, the weak bonding of two or more
22 particles into one larger particle. Gas phase material condenses preferentially on smaller
23 particles since they have the greatest surface area, and the rate constant for coagulation of two
24 particles decreases as the particle size increases. Thus, ultrafme particles grow into the
25 accumulation mode, but accumulation-mode particles do not normally grow into the coarse mode
26 (CD, p. 2-20).
27 Secondary formation processes can result in either new particles or the addition of PM to
28 pre-existing particles. Examples of secondary particle formation include: (1) the conversion of
29 sulfur dioxide (SO2) to sulfuric acid (H2SO4) droplets that further react with ammonia (NH3) to
30 form various sulfate particles (e.g., ammonium sulfate (NH4)2SO4 or ammonium bisulfate
August 2003 2-8 Draft - Do Not Quote or Cite
-------
1 NH4HSO4); (2) the conversion of nitrogen dioxide (NO2) to nitric acid (HNO3) droplets which
2 react further with ammonia to form ammonium nitrate (NH4NO3) particles; and (3) reactions
3 involving volatile organic compounds (VOC) yielding organic compounds with low ambient
4 temperature (saturation) vapor pressures that nucleate or condense on existing particles to form
5 secondary organic aerosol particles (CD, p. 3-62 to 3-72).
6
7 2.2.3 Chemical Composition
8 Based on studies conducted in most parts of the U.S., the draft CD reports that sulfate,
9 ammonium, and hydrogen ions; black carbon, secondary organic compounds, and primary
10 organic species from cooking and combustion; and certain metals, primarily from combustion
11 processes, are found predominately in fine particles. Crustal-related materials such as calcium,
12 aluminum, silicon, magnesium, and iron; and primary organic materials such as pollen, spores,
13 and plant and animal debris are found predominantly in coarse particles (CD, p. 2-29).
14 Some components, such as potassium and nitrate, may be found in both fine and coarse
15 particles. Potassium in coarse particles comes almost entirely from soil. Potassium in fine
16 particles comes mainly from emissions of burning wood or cooking meat. Nitrate in fine
17 particles comes mainly from the reaction of gas-phase nitric acid with gas-phase ammonia to
18 form ammonium nitrate particles. Nitrate in coarse particles comes primarily from the reaction
19 of gas-phase nitric acid with pre-existing coarse particles (CD, p. 2-29).
20 Many ambient particles also contain water (i.e., particle-bound water) as a result of an
21 equilibrium between water vapor and hygroscopic particles (CD, p. 2-31). Particle-bound water
22 influences the size of particles and in turn their aerodynamic and light scattering properties
23 (discussed in section 2.9). The amount of particle-bound water will vary with the composition of
24 particles. Sulfates, nitrates, and secondary organic compounds are more hygroscopic than BC,
25 primary OC, and crustal material.
August 2003 2-9 Draft - Do Not Quote or Cite
-------
1 2.2.4 Fate and Transport
2 Fine- and coarse-mode particles typically exhibit different behavior in the atmosphere.
3 These differences may affect several exposure-related considerations including the
4 representativeness of central-site monitored values and the penetration of particles formed
5 outdoors into indoor spaces. The ambient residence time of atmospheric particles varies with
6 size. Ultrafine particles grow rapidly into the accumulation mode, and have a very short life, on
7 the order of minutes to hours. Ultrafine particles are also small enough to be removed through
8 diffusion to falling rain drops. Accumulation-mode particles remain suspended longer, due to
9 collisions with air molecules, and have relatively low surface deposition rates. They can be
10 transported thousands of kilometers and remain in the atmosphere for days to weeks.
11 Accumulation-mode particles serve as condensation nuclei for cloud droplet formation and are
12 eventually removed from the atmosphere as falling rain drops. Accumulation-mode particles
13 that are not involved in cloud processes are removed from the atmosphere by falling rain, or
14 eventually by gravitational settling and impaction on surfaces.
15 By contrast, coarse-mode particles can settle rapidly from the atmosphere with lifetimes
16 ranging from minutes to hours depending on their size, atmospheric conditions, and their
17 altitude. Larger coarse-mode particles are not readily transported across urban or broader areas,
18 because they are generally too large to follow air streams, and they tend to be easily removed by
19 gravitational settling and by impaction on surfaces. Smaller-sized coarse-mode particles can
20 have longer lifetimes and longer travel distances, especially in extreme circumstances, such as
21 intercontinental dust storms (CD, p. 2-35). Coarse-mode particles also are readily removed by
22 falling rain drops (CD, p. 2-35).
23 The characteristics of ultrafme, accumulation-mode, and coarse-mode particles that were
24 discussed in the preceding sections are summarized in Table 2-2.
August 2003 2-10 Draft - Do Not Quote or Cite
-------
Table 2-2. Comparison of Fine and Coarse Particles
Fine
Coarse
Ultrafme
Accumulation
Aerodynamic
Diameter
um
0.1 to < 3.0 um
> 1.0 um
Formed from:
Formed by:
Combustion, high temperature
processes and atmospheric reactions
Nucleation
Condensation
Coagulation
Condensation
Coagulation
Reaction of gases in or on
particles
Evaporation of fog and
cloud droplets in which
gases have dissolved
and reacted
Break-up of large solids/droplets
Mechanical disruption (crushing,
grinding, abrasion of surfaces)
Evaporation of sprays
Suspension of dusts
Reactions of gases in or on particles
Composed of:
Solubility:
Sources:
Sulfate ions, Black
carbon, Metal
compounds, Organic
compounds with very
low saturation vapor
pressure at ambient
temperatures, particle-
bound water.
Probably less soluble
than accumulation
mode
Combustion of coal, oil,
gasoline, diesel
fuel, wood
Atmospheric
transformation of
SO2 and some
organic compounds
High temperature
industrial processes,
smelters, steel
mills, etc.
Volcanic activity
Wildfires
Sulfate (SO;2), Nitrate (NOj
), Ammonium (NHJ),
and Hydrogen (H+) ions
Black carbon,
Large variety of organic
compounds
Metal compounds of Pb, Cd,
V, Ni, Cu, Zn, Mn, Fe,
K, etc.)
Particle-bound water
Largely soluble,
hygroscopic and
deliquescent
Combustion of coal, oil,
gasoline, diesel fuel,
wood
Atmospheric transformation
products of NOX, SO2,
and organic
compounds including
biogenic organic
species (e.g., terpenes)
High temperature industrial
processes, smelters,
steel mills, etc.
Volcanic activity
Wildfires
Suspended soil or street dust
Fly ash from uncontrolled combustion of
coal, oil, wood
Nitrates/chlorides associated with basic
metal oxides
Oxides of crustal elements (Si, Al, Ti, Fe,
Mg)
CaCO3, NaCl, sea salt
Pollen, mold, fungal spores
Plant and animal fragments
Tire, brake pad, and road wear debris
Largely insoluble and non-hygroscopic
except for sea salt
Resuspension of industrial dust and soil
tracked onto roads and streets
Suspension from disturbed soil (e.g.,
farming, mining, unpaved roads)
Construction and demolition
Uncontrolled coal and oil combustion
Ocean spray
Biological sources
Wind-blown dust
Atmospheric
half-life:
Removal
Processes:
Minutes to hours
Grows into accumulation
mode
Diffusion to falling rain
drops
Travel distance: <1 to 10s of km
Days to weeks
Forms cloud droplets and
rains put
Dry deposition
100s to 1000s of km
Minutes to hours
Days for particles suspended by dust storms
Dry deposition by fallout
Scavenging by falling rain drops
<1 to 10s of km
(100s to 1000s in dust storms)
Source: Adapted from Wilson and Suh (1997); CD, p. 2-37.
August 2003
2-11
Draft - Do Not Quote or Cite
-------
1 2.3 SOURCE EMISSIONS
2 The distribution and amount of emissions of pollutants that contribute to ambient PM can
3 provide insights into observed ambient levels. The links between source emissions and ambient
4 concentrations of PM can include complex, non-linear atmospheric processes, including gaseous
5 chemical reactions and pollution transport.
6 Source emissions can be measured using monitoring equipment or estimated using
7 emission inventory methods. For most source types, emissions inventory methods are the most
8 practical. The EPA routinely publishes national estimates of annual source emissions of
9 pollutants that contribute to ambient PM concentrations. In general, national emissions estimates
10 are uncertain, and there have been few field studies to test emission inventories against
11 observations. The draft CD concludes that uncertainties in national emissions estimates could be
12 as low as ±10 percent for the best characterized source categories (e.g., SO2 from power plants
13 measured by continuous instruments), while fugitive dust sources should be regarded as order-
14 of-magnitude (CD, p. 3-98). The EPA is working to reduce these uncertainties through advances
15 in the understanding of the fate and transport characteristics of fugitive dust emissions released
16 at ground level.
17 Table 2-3 provides a summary of recent annual estimates of national emissions of
18 primary PM and PM precursors. While reviewing the following discussion on emissions
19 estimates, the reader should keep in mind that national estimates, while instructive, can obscure
20 important distinctions in the relative contributions of different sources across smaller geographic
21 regions, including important differences between urban and rural areas.
22
23 2.3.1 Primary PM Emissions
24 The majority of directly emitted anthropogenic PM is estimated to be coarse particles.
25 Though highly uncertain, recent national estimates of PM10_2 5 emissions shown in Table 2-3 are
26 about 2.5 times higher than estimates of PM25 emissions - 16.9 million short tons compared to
27 6.8 million short tons. A large portion of primary PM emissions are attributed to a variety of
28 small area-wide sources, which are often more difficult to characterize and are more uncertain
29 than larger point source emissions.
August 2003 2-12 Draft - Do Not Quote or Cite
-------
Table 2-3. Annual National Emissions Estimates of Primary PM and PM Precursors (thousand short tons), 1999
Source
Fuel Combustion
Industrial Processes
On-Road Vehicles
Nonroad Engines and Vehicles
Agriculture
Fire
Livestock
Miscellaneous
Fugitive Dust
Biogenic/Natural
TOTAL
NOX
10,026
942
8,590
5,515
0
-
0
320
0
2,315
27,708
SO2
16,091
1,465
363
936
0
-
0
12
0
-
18,867
NH3
48
289
260
10
769
-
3,552
-
0
35
4,964
voca
904
7,996
5,297
3,232
8
699
0
9
0
48,500
66,645
Primary
PM25
766
913
229
411
860
853
89
22
2,631
n/e
6,773
Primary
PM10.2.5
263
350
66
47
3,438
133
501
2
12,105
n/e
16,906
n/e = not estimated. Direct PM emissions due to wind erosion on natural surfaces are not estimated because this source is highly sporadic, resulting in emissions
too highly uncertain to be included.
a VOC emissions have been estimated for compounds known to contribute to ozone formation; the organic compounds that are most important to secondary
organic aerosols are a subset of this category, and may include other semi-volatile compounds.
Source: EPA (2001), Tables A-4 through A-10; Guenther et al. (2000).
August 2003
2-13
Draft - Do Not Quote or Cite
-------
1 National estimates of primary PM10_25 are dominated by fugitive dust and agricultural
2 sources. Fugitive dust sources include paved and unpaved road dust, dust from construction and
3 agricultural activities, and natural sources like geogenic wind erosion (not estimated in Table 2-
4 3). Fugitive dust is also a significant source of primary PM2 5. Unlike PM10_25, where fugitive
5 dust emissions comprise about 75 percent of total emissions, fugitive dust emissions of PM25 is
6 only about one-third of total emissions. Recent research has found that about 75 percent of these
7 emissions are within 2 meters of the ground when measured. A significant portion of these
8 coarse-mode particles are removed or deposited within a few kilometers of their release point
9 due to turbulence associated with surface topography, and the presence of vegetation or
10 structures (DRI, 2000). This is consistent with the generally small amount of crustal material
11 found in ambient PM2 5 samples in most locations. As shown in Table 2-3, direct emissions from
12 fuel combustion, industrial processes, fires, and motor vehicles contribute more to primary PM2 5
13 than to primary PM10_2 5.
14
15 2.3.2 Secondary PM Precursor Emissions
16 Major precursors of secondarily formed fine particles include SO2, nitrogen oxides
17 (NOX), which encompasses NO and NO2, and certain organic compounds. Table 2-3 shows the
18 estimated contribution of various sources to nationwide emissions of SO2 NOX, VOC, and NH3.
19 Fuel combustion in the power generation and industrial sectors dominates nationwide estimates
20 of SO2 emissions and contributes significantly to NOX emissions. However, emissions from
21 motor vehicles comprise the greatest portion of nationwide NOX emissions. Motor vehicle
22 emissions also make up a substantial portion of nationwide VOC emissions, with additional
23 contributions from the use of various solvents in industrial processes and commercial products.
24 The vast majority of nationwide NH3 emissions are estimated to come from livestock operations
25 and fertilizer application, but in urban areas there is a significant contribution from light-duty
26 cars and trucks, as well as certain industrial processes.
27 The relationship between changes in precursor emissions and resulting changes in
28 ambient PM25 can be nonlinear. Thus, it is difficult to project the impact on ambient PM2 5
29 arising from expected changes in PM precursor emissions without air quality simulation models
30 that incorporate treatment of complex chemical transformation processes and meteorology.
August 2003 2-14 Draft - Do Not Quote or Cite
-------
1 Generally SO2 emissions reductions lead to reductions in sulfate aerosol, and NOX emissions
2 reductions lead to reductions in nitrate aerosol. However, the direction and extent of changes
3 will vary by location and season, depending on fluctuations in NH3 emissions and changes in
4 prevailing meteorology and photochemistry.
5
6 2.4 AMBIENT PM MEASUREMENT METHODS
7 The methods used to measure ambient PM are important to understanding population
8 exposure to PM, evaluating health risks, and developing and evaluating the effectiveness of risk
9 management strategies. Because PM is not a homogeneous pollutant, measuring and
10 characterizing particles suspended in the atmosphere is a significant challenge.8 Ambient
11 measurements include particle mass, composition, and particle number. Most instruments
12 collect PM by drawing a controlled volume of ambient air through a size-selective inlet, usually
13 defined by the inlet's 50 percent cut point. Measurable indicators of fine particles include PM25,
14 PMj 0, British or black smoke (BS), coefficient of haze (COH), and PM10 (in areas dominated by
15 fine particles). Measurable indicators of coarse-mode particles include PM10_25, PM15.25, and
16 PM10 (in areas dominated by coarse-mode particles).
17
18 2.4.1 Particle Mass Measurement Methods
19 Ambient PM mass can be measured directly, by gravimetric methods, or indirectly, using
20 methods that rely on the physical properties of particles. The most common direct measurement
21 methods include filter-based methods where ambient aerosols are collected for a specified period
22 of time (e.g., 24 hours) on filters that are weighed to determine mass by difference before and
23 after collection. Examples include the FRM monitors for PM2 5 and PM10. Dichotomous
24 samplers contain a separator that splits the air stream from a PM10 inlet into two streams so that
25 both fine- and coarse-fraction particles can be collected on separate filters.
Refer to EPA 1996a, Chapter 4 and draft CD Chapter 2 for more comprehensive assessments of particle
measurement methods. A recent summary of PM measurement methods is also given in Fehsenfeld 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.htmlX
August 2003 2-15 Draft - Do Not Quote or Cite
-------
1 Another widely used method is the Tapered Element Oscillating Microbalance
2 (TEOM®) sensor, consisting of a replaceable filter mounted on the narrow end of a hollow
3 tapered quartz tube. The air flow passes through the filter, and the aerosol mass collected on the
4 filter causes the characteristic oscillation frequency of the tapered tube to change in direct
5 relation to particle mass. This approach allows mass measurements to be recorded on a near-
6 continuous basis (i.e., every few minutes).
7 Other methods that produce near-continuous PM mass measurements include the beta
8 attenuation sampler and the Continuous Ambient Mass Monitor (CAMM). A beta attenuation
9 (or beta gauge) sampler determines the mass of particles deposited on a filter by measuring the
10 absorption of electrons generated by a radioactive isotope, where the absorption is closely
11 related to the mass of the particles. The CAMM measures the pressure drop increase that occurs
12 in relation to particle loading on a membrane filter. Both methods (beta-attenuation and
13 CAMM) require calibration against standard mass measurements as neither measures PM mass
14 directly by gravimetric analysis.
15 PM has also been characterized in the U.S. and elsewhere by indirect optical methods
16 that rely on the light scattering or absorbing properties of either suspended PM or PM collected
17 on a filter.9 These include BS, COH, and estimates derived from visibility measurements. In
18 locations where they are calibrated to standard mass units, these indirect measurements can be
19 useful surrogates for particle mass. The BS method typically involves impacting samples from a
20 4.5 jim inlet onto white filter paper where blackness of the stain is measured by light absorption.
21 Smoke particles composed of black carbon, including black carbon (BC), typically make the
22 largest contribution to stain darkness. COH is determined using a light transmittance method.
23 This involves impacting samples from a 5.0 jim inlet onto filter tape where the opacity of the
24 resulting stain is determined. This technique is somewhat more responsive to non-carbon
25 particles than the BS method. Nephelometers measure the light scattered by ambient aerosols in
26 order to calculate light extinction. This method results in measurements that can correlate well
27 with the mass of fine particles below 2 jim diameter. Since the mix of ambient particles varies
9 See Section 2.9 of this chapter for a discussion of the optical properties of PM.
August 2003 2-16 Draft - Do Not Quote or Cite
-------
1 widely by location and time of year, the correlation between BS, COH, and nephelometer
2 measurements and PM mass is highly site- and time-specific.
O
4 2.4.2 Size-Differentiated Particle Number Concentration Measurement Methods
5 Recently there has been increasing interest in examining the relationship between the
6 particle number concentration by size and health effects. Several instruments are needed to
7 provide size distribution measurements (number and size) over the 5 orders of magnitude of
8 particle diameters of interest. A nano-scanning mobility particle sizer (NSMPS) counts particles
9 in the 0.003 to 0.15 jim range. A standard scanning mobility particle sizer (SMPS) counts
10 particles in the 0.01 to 1 |im range, and a laser particle counter (LPC) counts particles in the 0.1
11 to 2 |im range. An aerodynamic particle sizer measures particles in the 0.7 to 10 jim range.
12 These techniques, while widely used in aerosol research, have not yet been widely used in health
13 effects studies.
14
15 2.4.3 Chemical Composition Measurement Methods
16 There are a variety of methods used to identify and describe the characteristic
17 components of ambient PM.10 X-ray fluorescence (XRF) is a commonly used laboratory
18 technique for analyzing the elemental composition of primary particles deposited on filters. Wet
19 chemical analysis methods, such as ion chromatography (1C) and automated colorimetry (AC)
20 are used to measure ions such as nitrate (NO3"), sulfate (SO4=), chloride (Cl"), ammonium (NH+),
21 sodium (Na+), organic cations (such as acetate), and phosphate (PO43").
22 There are several methods for separating organic carbon (OC) and black carbon (BC) in
23 ambient filter samples. Thermal/optical reflectance (TOR) and thermal manganese oxidation
24 (TMO) have been commonly applied in aerosol studies in the United States. Still another
25 method is the thermal/optical transmission (TOT) method, which uses a different temperature
26 profile than TOR. The two methods yield comparable estimates of total carbon, but give a
27 different split between OC and BC. Monitoring methods capable of separately measuring
10 The reader is referred to Chapter 2, section 2.2, of the draft CD for a more thorough discussion of
sampling and analytical techniques for measuring PM. Methods used in EPA's National PM Speciation Trends
Network and other special monitoring programs are summarized in Solomon et al. (2001).
August 2003 2-17 Draft - Do Not Quote or Cite
-------
1 sulfate, nitrate, and carbon particles on a near-continuous basis have been developed recently,
2 but have not yet been widely tested or used.
O
4 2.4.4 Measurement Issues
5 There is no perfect sampler under all conditions, so there are uncertainties between the
6 mass and composition collected and measured by a sampler and the mass and composition of
7 material that exists as suspended PM in ambient air (Fehsenfeld et al., 2003). To date few
8 standard reference materials exist to estimate the accuracy of measured PM mass and chemical
9 composition relative to what is found in air. At best, uncertainty is estimated based on
10 collocated precision and comparability or equivalency to other similar methods, which
11 themselves have unknown uncertainty, or to the FRM, which is defined for regulatory purposes
12 but is not a standard in the classical sense. There are a number of measurement-related issues
13 that can result in positive or negative measurement artifacts which could affect the associations
14 epidemiological researchers find between ambient particles and health effects.
15 The semivolatile components of PM can create both positive and negative measurement
16 artifacts. Negative artifacts arise from evaporation of the semivolative components of PM
17 during or after collection, which is caused by changes in temperature, relative humidity, or
18 aerosol composition, or due to the pressure drop as collected air moves across the filter. Nitrate
19 losses due to evaporation may represent as much as 10-20% of total PM2 5 mass, as shown in
20 southern California studies (CD, p. 2-97). Positive artifacts arise when gas-phase compounds
21 absorb onto or react with filter media or already collected PM, or when particle-bound water is
22 not removed. The chemical interaction of gases being collected with particles already on the
23 filter can also result in negative artifacts. These interactions depend on the compounds
24 contained in collected particles and in the gas phase, and also depend on both location and time.
25 Particle-bound water can represent a significant fraction of PM mass under conditions
26 where relative humidity is more than 60% (CD, p. 2-46). The amount of particle-bound water
27 will vary with the composition of particles, as discussed in section 2.2.3. The use of heated
28 inlets to remove particle-bound water (e.g. TEOM at 50° C) can result in loss of semi-volatile
29 compounds unless corrective techniques are applied (CD, p. 2-101).
August 2003 2-18 Draft - Do Not Quote or Cite
-------
1 Particle bounce from the filter can result in negative artifacts. This may be more
2 prevalent under lower relative humidity conditions. Impactor coatings can be used to limit
3 particle bounce, but can interfere with mass and chemical composition measurements..
4 In areas with significant amounts of dust, high wind conditions resulting in blowing dust
5 can interfere with accurate separation of fine- and coarse-fraction particles. In these unique
6 conditions a significant amount of coarse-fraction material can be found in the inter-modal
7 region between 1 and 3 jim. The addition of a PMl 0 measurement in these circumstances can
8 provide greater insights into the magnitude of this problem.
9
10 2.5 PM CONCENTRATIONS, TRENDS, AND SPATIAL PATTERNS IN THE U.S.
11 This section provides analysis of the latest available PM air quality data, including PM
12 levels, composition, and spatial patterns. The EPA and the States have been using a national
13 network to measure and collect PM10 concentrations since 1987, and PM2 5 concentrations since
14 1999. Summaries through the end of 2002, based on data publically available from EPA's Air
15 Quality System (AQS) as of April 2003, are presented here. PM25 data from the network for
16 Interagency Monitoring of Protected Visual Environments (IMPROVE) are also presented.
17 Many data summaries are presented by region, as shown in Figure 2-3. These regions were first
18 defined in the 1996 CD, and later revised by researchers in the NMMAPS study, and have
19 proven useful for understanding potential differences in the characteristics of PM in different
20 parts of the U.S..
21
22 2.5.1 PM25
23 Following the establishment of new standards for PM2 5 in 1997, the EPA led a national
24 effort to deploy and operate over 1000 PM25 monitors. These monitors use the FRM which,
25 when its procedures are followed, assures that PM data are collected using standard equipment,
August 2003 2-19 Draft - Do Not Quote or Cite
-------
4
5
6
7
Figure 2-3. Regions used in Staff Paper PM data analysis summaries.
operating procedures, and data handling techniques.11 Most of these FRM monitors began
operation in 1999. The EPA has analyzed the available data collected by this network from
2000-2002. Data from the monitors were screened for completeness with the purpose of
avoiding seasonal bias. To be included in these analyses a monitor needed at least a full year of
data, defined as either 4, 8, or 12 consecutive quarters with eleven or more observations per
quarter. A total of 1142 FRM monitors in the United States met these criteria12.
The annual PM2 5 mean concentrations range from about 2 to 29 |ig/m3, with a median of
about 13 |ig/m3. The 98th percentiles of the distribution of 24-hour average concentrations range
from about 8 to 94 |ig/m3, with a median of about 33 |ig/m3. Figures 2-4 and 2-5 depict the
12
See 40 CFR Parts 50 and 58 for monitoring program requirements.
1129 of the 1142 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.
August 2003 2-20 Draft - Do Not Quote or Cite
-------
4
5
65:
55:
E
"DJ
CD
^
£=
25:
-5
TO
0)
-e
o
en
TO
O
T3
5
"CD
1
I
§
o
I
-e
o
CO
O
PM25 PM10.2.6 PM25 PM10.2.6 PM25 PM10.2.6 PM25 PM10.2.6 PM25 PM10.25 PM25 PM10.2.6 PM25 PM10.2.6
N= 200 68 290 86 260 106 94 39 60 31 177 109 48 36
Figure 2-4. Distribution of annual mean VM2S and estimated annual mean PM10_25
concentrations by region, 2000-2002. Box depicts interquartile range and
median; whiskers depict 5th and 95th percentiles; asterisks depict minimum and
maximum. Number below indicates the number of sites in each region.
Source: Schmidt et al. (2003)
regional distribution of site-specific annual mean and 98th percentile 24-hour average PM25 (and
PMio-2.5, discussed in section 2.5.3) concentrations, respectively, by geographic region
(excluding Alaska, Hawaii, Puerto Rico, and the Virgin Islands). Figures 2-6 and 2-7 are national
maps that depict county-level annual mean PM2 5 concentrations and 98th percentile 24-hour
average
August 2003
2-21
Draft - Do Not Quote or Cite
-------
O)
^.
_CD
§
CO
CT)
200
190
180"
170"
160
ISO
HO
130-
120"
110
100 "
90
70:
GO
50'
40'
30
20 '
10
0"
CO
CO
CD
O
Z
*
^
y
J
CO
CO
CD
O
W
*
*
^
*
-J
^
-H- <
CO
CD
•a
Jco.
•a
*
Jk_
-T
-L
*
J
H
s
"co
CD
•a
-CD—
Q.
Q.
>
H
?
H
^
^
' ^L~
"co
CD
(—
t—i
O
W
nl
a.
\
-+-
00
CD
C
O
Z
*
Jl
si
CO
E
p
t >k
CO
O
s_
CD
JZ
O
W
V JL
N= 200
68 290
86
260 106
94
39
60
31
177
109 48
36
Figure 2-5. Distribution of 98th percentile 24-hour average PM2 5 and estimated PM
10_
25 concentrations by region, 2000-2002. Box depicts interquartile range and
median; whiskers depict 5th and 95th percentiles; asterisks depict minimum
and maximum. Number below indicates the number of sites in each region.
Source: Schmidt et al. (2003)
August 2003
2-22
Draft - Do Not Quote or Cite
-------
PM25 Concentration (|jg/m3)
693 counties
18
Figure 2-6. County-level maximum annual mean PM2 5 concentrations, 2000-2002.
Source: Schmidt et al. (2003)
August 2003
2-23
Draft - Do Not Quote or Cite
-------
PM2 5 Concentration (|jg/m3)
693 counties
< or = 30
40 < or = 65
30 < or = 40
>65
Figure 2-7. County-level maximum 98th percentile 24-hour average PM2 5 concentration, 2000-2002.
Source: Schmidt et al. (2003)
August 2003
2-24
Draft - Do Not Quote or Cite
-------
1 concentrations, respectively, from the FRM network.13 The monitor with the highest
2 concentration in each monitored county is used to represent the value in that county. The map
3 and box plots show that many locations in the eastern U.S. and in California had annual mean
4 PM25 concentrations above 15 |ig/m3. Annual mean PM25 concentrations were above 18 |ig/m3
5 in several urban areas throughout the eastern U.S., including Atlanta, Birmingham, Chicago,
6 Cincinnati, Cleveland, Detroit, Indianapolis, Knoxville, Louisville, Pittsburgh, and St. Louis.
7 Los Angeles and the central valley of California also were above 18 |ig/m3. Sites in the upper
8 midwest, southwest, and northwest regions had generally low annual mean PM2 5 concentrations,
9 most below 12 |ig/m3. The 98th percentile 24-hour average PM2 5 concentrations above 65 |ig/m3
10 appear only in California and Montana. Values in the 40 to 65 |ig/m3 range were more common
11 in the eastern U.S. and on the west coast, mostly in or near urban areas, but relatively rare in the
12 upper midwest and southwest regions. In these regions the 98th percentile 24-hour average PM25
13 concentrations were more typically below 40 |ig/m3, with many below 25 |ig/m3.
14 The IMPROVE monitoring network, which consists of sites located primarily in national
15 parks and wilderness areas throughout the U.S., provides data for long-term PM25 trends for
16 generally rural areas.14 Figure 2-8 shows the composite long-term trend at 9 eastern sites, 23
17 western sites, and one urban site in Washington, D.C. At the rural eastern sites, measured PM25
18 decreased about 16 percent from 1992 to 2001. At the rural western sites PM25 decreased about
19 10 percent from 1992 to 2001. At the Washington, D.C. site the annual average PM25
20 concentration in 2001 was about 30 percent lower than the 10-year peak in 1994.
13 Readers should be cautioned not to draw conclusions regarding the potential attainment status of any area
from these data summaries. 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. Not all of the PM federal reference method (FRM) monitors that contributed data to the
summaries presented here recorded valid PM measurements for all four calendar quarters for each of the 3 years.
14 Since IMPROVE monitoring instruments and protocols are not identical to FRM monitors, the data are
not directly comparable to the FRM monitor data.
August 2003 2-25 Draft - Do Not Quote or Cite
-------
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
92
93
94
95
96
97
99
00
01
Year
- Washington D.C. Site
• 9 Eastern Sites
-23 Western Sites
Figure 2-8. Average measured annual mean PM2 5 concentration trend at IMPROVE
sties, 1992-2001. Included sites must have 8 of 10 valid years of data; missing years are
interpolated. Measured mass represents measurement from the filter.
Source: Schmidt et al. (2003)
1 The relative spatial homogeneity of the ambient air across a specified area can be
2 assessed by examining the values at multiple sites of several indicators, including: (1) site pair
3 correlations, (2) differences in long-term (e.g., annual) average concentrations, and (3)
4 differences in short-term (e.g., daily) average concentrations. An analysis of these indicators for
5 site pairs in 27 urban areas using PM25 FRM monitoring data from 1999-2001 is included in the
6 draft CD.
7 An analysis of monitor pairs from each of the 27 urban areas indicates that multiple sites
8 in these areas were highly correlated throughout the period. More than 86 percent (426 out of
9 491) of the between-site correlation coefficients in all 27 areas were greater than or equal to
10 0.80, and more than 53 percent (268 out of 491) of the correlations were greater than or equal to
August 2003 2-26 Draft - Do Not Quote or Cite
-------
1 0.90. Further, every area had at least one monitor pair with a correlation coefficient greater than
2 or equal to 0.85 (CD, Appendix 3A).
3 A summary of the analyses of long-term and short-term concentration differences is
4 shown in Table 2-4. The difference in annual mean PM2 5 concentrations between monitor pairs
5 in the 27 cities ranged from less than 1 |ig/m3 in Baton Rouge to about 8 |ig/m3 in Pittsburgh.
6 Large differences in annual mean concentrations across a metropolitan area may be due to
7 differences in emissions sources, meteorological, or topography. Small differences may be due
8 only to measurement imprecision (CD, p. 3-46). In urban areas, the site pair with the maximum
9 and minimum annual mean concentration was highly correlated (r(maxmin) >0.70); the most notable
10 exception was the site pair in Gary, IN (r(maxmin)=0.56).
11 The analysis in the draft CD also examined differences in short-term 24-hour average
12 concentrations between the urban site pairs. Small differences throughout the distribution would
13 indicate relatively homogeneous concentration levels between the sites. Table 2-4 presents a
14 summary of the 90th percentile of the distribution (P90) of daily site pair differences in each urban
15 area. The site pair with the largest difference (max pair) and the smallest difference (min pair)
16 are shown. The P90 values for the 491 monitor pairs in the 27 urban areas ranged from about 2 to
17 21 |ig/m3. Often the site pair with the maximum P90 value in each city was also the pair with the
18 largest annual mean difference. The site pair with the highest P90 values in each city were
19 generally highly correlated (rmax>0.70), and in some cases were more highly correlated than the
20 sites with the largest annual mean differences.
August 2003 2-27 Draft - Do Not Quote or Cite
-------
Table 2-4. Summary of PM25 FRM Data Analysis in 27 Metropolitan Areas, 1999-2001.
City
Pittsburgh, PA
Salt Lake City, UT
Detroit, MI
Cleveland, OH
St. Louis, MO
Portland, OR
Chicago, IL
Seattle, WA
Birmingham, AL
Los Angeles, CA
Gary, IN
Washington, DC
Kansas City, MO
Riverside, CA
Dallas, TX
Boise, ID
Atlanta, GA
Grand Rapids, MI
San Diego, CA
Tampa, FL
Steubenville, OH
Philadelphia, PA
Louisville, KY
Milwaukee, WI
Norfolk, VA
Columbia, SC
Baton Rouge, LA
N
Sites
11
6
10
8
11
4
11
4*
5
6
4
5*
6
5
7
4
6*
4
4
4
5
7
4
8
5
3
o
J
Annual Mean (jig/m3)
Max Site
22.0
13.6
19.9
20.2
20.2
9.1
20.6
11.9
21.6
23.7
17.6
16.7
13.8
30.0
13.7
10.3
21.2
14.0
17.0
12.7
18.9
16.0
17.4
14.4
13.7
15.7
14.5
Min Site
13.8
8.8
13.5
14.0
13.9
6.3
14.5
8.9
16.6
18.3
14.0
13.8
11.4
25.0
11.5
8.7
18.3
12.1
14.6
11.1
16.5
14.1
15.7
13.1
12.6
14.7
14.1
% Diff
37%
35%
32%
31%
31%
31%
30%
25%
23%
23%
20%
17%
17%
17%
16%
16%
14%
14%
14%
13%
13%
12%
10%
9%
8%
6%
3%
* fmax.miiri
0.69
0.86
0.89
0.84
0.69
0.79
0.91
0.91
0.80
0.76
0.56
0.84
0.87
0.93
0.89
0.79
0.81
0.93
0.73
0.87
0.86
0.85
0.86
0.89
0.96
0.93
0.97
P90 Gig/m3)
Max Pair
21.0
11.4
13.8
14.3
15.2
6.5
11.3
8.5
15.2
18.2
11.3
7.7
6.5
17.8
6.3
8.8
10.8
6.1
11.0
5.0
10.0
7.5
6.0
5.3
5.0
3.3
2.9
Min Pair
4.2
4.4
5.0
3.3
2.8
4.1
3.5
3.6
6.6
6.2
4.2
3.5
1.9
3.6
1.9
3.8
5.3
3.1
6.3
3.1
6.2
3.3
3.8
2.8
2.6
2.8
2.5
r
"max
0.69
0.86
0.84
0.84
0.69
0.79
0.92
0.75
0.80
0.66
0.59
0.84
0.90
0.81
0.89
0.79
0.75
0.93
0.73
0.71
0.79
0.84
0.90
0.89
0.91
0.93
0.93
* Does not include 1 additional site >100 km from the others in the urban area.
P90 = 90th percentile of the distribution of differences in 24-hour averages between two sites in the same urban area.
r(max,mn) = correlation between intra-urban sites with the largest difference in annual mean concentrations.
r(max) = correlation between intra-urban sites with the largest difference in P90 values.
Source: CD, Appendix 3A
August 2003
2-28
Draft - Do Not Quote or Cite
-------
1 2.5.2 PM10
2 For the purpose of comparison to PM2 5 and PM10_2 5 concentrations, PM10 data from
3 2000-2002 are presented in Figures 2-9 and 2-10. Figure 2-9 shows the PM10 annual mean
4 concentrations, and Figure 2-10 shows the 98th percentile 24-hour average concentrations.15 As
5 in the earlier PM2 5 maps, the monitor with the highest value in each monitored county is used to
6 represent the value in each county. Most areas of the country had concentrations below the level
7 of the annual mean PM10 standard of 50 |ig/m3. Exceptions include several locations in the
8 southwest and in central California. Most areas of the country also had concentrations below the
9 level of the 24-hour standard of 150 |ig/m3, with exceptions mostly in the western U.S.16
10 In the National Air Quality and Emissions Trends Report (EPA, 200Ib), EPA examined
11 national and regional PM10 trends for the 10-year period from 1992 to 2001. The EPA found a
12 national average decline in concentrations of approximately 14 percent over the 10 year period,
13 with regional average declines across the eastern U.S. ranging from 7 to 28 percent and declines
14 across the western U.S. ranging from 10 to 31 percent.
15
16 2.5.3 PM1025
17 PM10_2 5 is a measure of the coarse-mode fraction of PM10 being considered in this review.
18 It can be measured by a dichotomous sampler, or by using a difference method with collocated
19 PM10 and PM25 monitors employing the same sampling protocol. A nationwide network of
20 samplers with the specific intent to consistently and accurately measure PM10_2 5 does not
21 currently exist. The EPA is currently evaluating a variety of monitoring platforms to establish a
22 FRM for PM10_2 5, which would be used in the future to design a national network of monitors to
23 measure coarse-fraction particles. Until such a network is established, estimates of PM10_2 5 can
24 be generated for a limited number of locations using a difference method on same-day data
15 These figures do not depict officially designated PM10 nonattainment areas. As of June 23, 2003, there
were a total of 64 areas classified as moderate or serious nonattainment areas, mostly in the western U.S. See
designated nonattainment areas atwww.epa.gov/oar/oaqps/greenbk/pnc.html.
16 The form of the 1987 PM10 standard is based on the annual 2nd highest daily concentration rather than the
98th percentile concentration shown in Figure 2-12. The annual 98th percentile concentration is presented here for
consistency with the depictions of PM2 5 and PM10_2 5 concentrations
August 2003 2-29 Draft - Do Not Quote or Cite
-------
PM10 Concentration (|jg/m3)
574 counties
< or = 20
30 < or = 50
20 < or = 30
>50
Figure 2-9. County-level maximum annual mean PM10 concentrations, 2000-2002.
Source: Schmidt et al. (2003)
August 2003
2-30
Draft - Do Not Quote or Cite
-------
PM10 Concentration (|jg/m3)
574 counties
< or = 50
75 150
Figure 2-10. County-level maximum 98th percentile 24-hour average PM10 concentrations, 2000-2002.
Source: Schmidt et al. (2003)
August 2003
2-31
Draft - Do Not Quote or Cite
-------
1 collected from co-located PM10 and PM2 5 FRM monitors. Since the protocol for each monitor is
2 not usually identical, the consistency of these PM10_2 5 measurements are relatively uncertain, and
3 are referred to as "estimates" in this draft Staff Paper.
4 The 98th percentiles of the distribution of estimated 24-hour average PM10_2 5
5 concentrations range from about 5 to 315 |ig/m3, with a median of about 29 |ig/m3. The box
6 plots in Figures 2-4 and 2-5 (introduced above in section 2.5.1 on PM2 5) depict the regional
7 distribution of site-specific estimated annual mean and 98th percentile 24-hour average PM10_2 5
8 concentrations, respectively, by geographic region (excluding Alaska, Hawaii, Puerto Rico, and
9 the Virgin Islands). Figures 2-11 and 2-12 are national maps that depict estimated county-level
10 annual mean PM10_25 concentrations, and 98th percentile 24-hour average concentrations,
11 respectively. To construct the maps, the site with the highest concentration in each monitored
12 county is used to represent the value in that county. The annual mean PM10_25 concentrations are
13 generally estimated to be below 45 |ig/m3, with one maximum value as high as 66 |ig/m3 (see
14 Figure 2-4), and with a median of about 10 |ig/m3. Compared to annual mean PM25
15 concentrations, annual mean PM10_25 estimates are more variable, with more distinct regional
16 differences. As shown in Figure 2-4, eastern U.S. estimated annual mean PM10_25 levels tend to
17 be lower than annual mean PM25 levels, and in the western U.S. estimated PM10_25 levels tend to
18 be higher than PM2 5 levels. The highest estimated annual mean PM10_2 5 concentrations appear in
19 the southwest region and southern California. The estimated 98th percentile 24-hour average
20 PM10_2 5 concentrations are generally highest in the southwest, southern California, and upper
21 midwest, where a few sites have estimated concentrations well above 100 |ig/m3 (see Figure 2-
22 5).
23 The IMPROVE monitoring network provides data for long-term PM10_2 5 trends for
24 generally rural areas. Figure 2-13 presents the composite long-term trend at 9 eastern sites, 23
25 western sites, and one urban site in Washington D.C. At the rural eastern sites, measured PM10.
26 2 5 in 2001 was about 34 percent lower than the 10-year peak in 1994. At the rural western sites,
27 measured PM10_2 5 was about 19 percent lower in 2001 than the 10-year peak in 1994. At the
28 Washington, D.C. site the annual average PM10_25 concentration in 2001 was about 30 percent
29 lower than the 10-year peak in 1994, but nearly 1 |ig/m3 higher than the 1998 low point.
August 2003 2-32 Draft - Do Not Quote or Cite
-------
PM10_2 5 Concentration (|jg/m3)
351 counties
< or = 10
15 < or = 25
10 < or = 15
>25
Figure 2-11. Estimated county-level maximum annual mean PM10_25 concentrations, 2000-2002.
1 Source: Schmidt et al. (2003)
August 2003
2-33
Draft - Do Not Quote or Cite
-------
PM10_25 Concentration (|jg/m3)
351 counties
75
Figure 2-12. Estimated county-level maximum 98th percentile 24-hour average PM10_2 5 concentrations,
2000-2002.
Source: Schmidt et al. (2003)
August 2003
2-34
Draft - Do Not Quote or Cite
-------
00
01
-Washington D.C. Site -B-9 Eastern Sites
•23 Western Sites
Figure 2-13. Average measured annual average PM10_2 5 concentration trend at
IMPROVE sties, 1992-2001. Included sites must have 8 of 10 valid years of data; missing
years are interpolated. Measured mass represents the measurement from the filter.
Source: Schmidt et al. (2003)
1 The draft CD contains an analysis of 1999-2001 PM10_25 estimates in 17 urban areas that
2 is useful for assessing the spatial homogeneity of PM10_25 across the urban areas (CD, Appendix
3 3 A). This analysis is similar to the 27-city analysis for PM25 discussed in section 2.5.1 and
4 summarized earlier in Table 2-4. However, since there were fewer site pairings, and fewer urban
5 areas covered, the PM10_2 5 results are not as robust as the PM2 5 results. The PM10_2 5 analysis is
6 summarized in Table 2-5. The analysis reveals generally lower correlations for PM10_2 5
7 compared to the PM2 5 correlations in the same city. Of the 65 monitor pairs analyzed, only 4
8 had correlation coefficients greater than or equal to 0.80, in contrast to more than 86% (426 of
9 491) of the pairs for PM2 5.
10
August 2003
2-35
Draft - Do Not Quote or Cite
-------
Table 2-5. Summary of Estimated PMin.2.s Analysis in 17 Metropolitan Areas, 1999-2001.
City
Cleveland, OH
Detroit, MI
Salt Lake City, UT
St. Louis, MO
Riverside, CA
Dallas, TX
San Diego, CA
Baton Rouge, LA
Los Angeles, CA*
Steubenville, OH
Gary, IN
Columbia, SC
Chicago, IL
Louisville, KY
Portland, OR
Milwaukee, WI
Tampa, FL
N
Sites
6
3
o
J
3
4
4
4
2
4
4
o
3
2
o
5
2
2
2
2
Annual Mean (jig/m3)
Max Site
26.4
19.4
27.5
22.5
46.2
19.1
19.4
19.1
24.1
14.3
5.1
9.6
16.1
9.1
6.7
9.1
11.3
Min Site
7.2
7.3
14.8
12.1
25.5
11.2
11.6
12.8
16.1
10.2
3.9
7.4
12.8
7.6
5.7
7.9
10.1
% Diff
73%
62%
46%
46%
45%
41%
40%
33%
33%
29%
24%
23%
20%
16%
15%
13%
11%
* fmax.miiri
0.41
0.39
0.72
0.70
0.32
0.66
0.65
0.40
0.58
0.54
0.79
0.70
0.53
0.65
0.69
0.65
0.81
P90 Gig/m3)
Max Pair
40.0
34.9
28.7
27.2
42.6
16.5
14.7
22.4
17.3
18.5
8.0
8.0
24.6
5.5
5.1
9.2
5.3
Min Pair
10.6
15.7
9.8
13.0
13.3
4.5
8.3
22.4
15.5
10.9
6.3
8.0
11.1
5.5
5.1
9.2
5.3
r
"max
0.41
0.39
0.72
0.70
0.36
0.66
0.63
0.40
0.58
0.48
0.60
0.70
0.53
0.65
0.69
0.65
0.81
* Does not include 1 additional site >100 km from the others in the urban area.
P90 = 90th percentile of the distribution of differences in 24-hour averages between two sites in the same urban area.
r(max,min) = correlation between intra-urban sites with the largest difference in annual mean concentrations.
r(max) = correlation between intra-urban sites with the largest difference in P90 values.
Source: CD, Appendix 3A
4
5
The difference in estimated annual mean PM10_2 5 between site pairs in the 17 cities also
covered a greater range than was seen for PM25, with differences up to about 21 |ig/m3 in
Riverside, CA. Similarly, the P90 values (described in section 2.5.1) for the 65 site pairs ranged
from about 5 |ig/m3 to about 43 |ig/m3, which is wider than the range of about 2 |ig/m3 to 21
|ig/m3 observed for PM2 5.
August 2003
2-36
Draft - Do Not Quote or Cite
-------
1 These indicators provide evidence that PM10_2 5 is more heterogenous than PM2 5 in some
2 locations (e.g., Cleveland, Detroit, Steubenville), and may be similar in other locations (e.g.,
3 Portland, Tampa, St. Louis). Any conclusions should be tempered by the inherent uncertainty in
4 the PM10_2 5 estimation method (discussed at the beginning of this section), and the relatively
5 small sample size for PM10_2 5 relative to PM2 5.
6
7 2.5.4 Ultrafine Particles
8 There are no nationwide monitoring networks for ultrafine particles (i.e., those with
9 diameters < 0.1 jim), and only a few recently published studies of ultrafine particle counts in the
10 U.S. At an urban site in Atlanta, GA, particles in three size classes were measured on a
11 continuous basis between August 1998 and August 1999 (Woo et al., 2001). The classes
12 included ultrafine particles in two size ranges, 0.003 to 0.01 |im and 0.01 to 0.1 jim, and a subset
13 of accumulation-mode particles in the range of 0.1 to 2 jim. In Atlanta, the vast majority (89
14 percent) of the number of particles were in the ultrafine mode (smaller than 0.1 jim), but 83
15 percent of the particle volume was in the subset of accumulation-mode particles. The
16 researchers found that for particles with diameters up to 2 jim, there was little evidence of any
17 correlation between number concentration and either volume or surface area. Kim et al. (2002)
18 confirmed similarly poor correlations between PM25 mass and number of ultrafine particles for
19 sites in Los Angeles and nearby Riverside, CA. This suggests that PM2 5 cannot be used as a
20 surrogate for ultrafine mass or number, so ultrafine particles need to be measured independently.
21 Studies of near-roadway particle number and size distributions have shown sharp
22 gradients in ultrafine concentrations around Los Angeles roadways (Zhu et al. 2002a,b).
23 Ultrafine PM concentrations were found to decrease exponentially with distance from the
24 roadway source, and was equal to the upwind "background" location at 300 m downwind.
August 2003 2-37 Draft - Do Not Quote or Cite
-------
1 2.5.5 Components of PM
2 Atmospheric PM is comprised of many different chemical components that vary by
3 location, time of day, and time of year. Further, as discussed in section 2.3, fine- and coarse-
4 fraction particles have fundamentally different sources and composition. Recent data from the
5 rural IMPROVE network and from the urban speciation network provide indications of regional
6 composition differences for fine-fraction particles. Figure 2-14 shows recent annual average
7 PM25 mass apportionment among chemical components at several sites in 9 different regions.17
8 In general, fine-fraction particles in the eastern U.S. regions are dominated by ammonium
9 sulfate, and carbon compounds. In the western U.S. regions fine-fraction particles have a greater
10 proportion of carbon compounds. With the exception of rural locations in the desert west region,
11 crustal material is a very small portion of fine-fraction particles. The ammonium nitrate
12 component is more prevalent in urban aerosols than in rural aerosols, especially in the California
13 region, and may be an indication of population-driven NOX sources, such as transportation
14 activity and urban combustion sources.
15 Trends in rural area and Washington D.C. concentrations of fine-fraction particle
16 components (with the exception of ammonium nitrate) based on data from the IMPROVE
17 network are shown in Figure 2-15. Annual average ammonium sulfate has declined at the urban
18 Washington D.C. site and the rural eastern sites. Both total carbon and crustal material at the
19 Washington D.C. site dropped significantly in 1995, but have not shown significant
20 improvements since then. All other components in all areas have shown small changes over the
21 10-year period.
17 Figure 2-14 identifies ammonium as a separate component of PM25 mass, however, it is associated with
either sulfate or nitrate (as ammonium sulfate or ammonium nitrate) roughly in proportion to the amount of sulfate
and nitrate indicated.
August 2003 2-3 8 Draft - Do Not Quote or Cite
-------
|Sulfate @ Estimated Ammonium • Nitrate Q Total Carbon nCrustal
Figure 2-14. Annual average composition of PM25 by region. Rural data from IMPROVE
network, urban data from EPA Speciation Trends Network, September 2001 - August 2002.
(components are displayed in the same order as the legend from bottom to top in the stacked
bars)
Source: Schmidt et al. (2003)
1
2
August 2003
2-39
Draft - Do Not Quote or Cite
-------
E
Total Carbon Trend
93 94
Figure 2-15. Average annual average trend in fine particle ammonium sulfate, total
carbon, and crustal material at IMPROVE sites, 1992-2001. Included sites
must have 8 of 10 valid years of data; missing years are interpolated. Ammonium
nitrate trend is not shown because the methodology changed in 1996.
Source: Schmidt et al. (2003)
August 2003
2-40
Draft - Do Not Quote or Cite
-------
1 2.5.6 Relationships Among PM2 5, PM10, and PM10_2 5
2 In this section, information on the relationships among PM indicators in different regions
3 is presented based on data from the nationwide PM FRM monitoring networks. Figure 2-16
4 shows the distribution of ratios of annual mean PM25 to PM10 at sites in different geographic
5 regions for 2000-2002. The ratios are highest in the eastern U.S. regions with median ratios
6 between 0.6 and 0.7, and lowest in the Southwest region, with a median ratio less than 0.4.
7 These data are generally consistent with earlier findings reported in the 1996 CD from a more
8 limited set of sites. Ratios greater than one are an artifact of the imprecision in the independent
9 PM10 and PM2 5 measurement methods.
10 Correlations among pollutant indicators can provide insights into how well one indicator
11 can represent the variability in another indicator. Many earlier epidemiological studies use PM10
12 as an indicator of fine particles. Figure 2-17 shows the results of a nationwide analysis of
13 correlations among PM size fractions using 24-hour average data from the FRM monitoring
14 networks for 2000-2002. PM2 5 and PM10 measured on the same days at collocated monitors are
15 fairly well correlated, on average, in the eastern regions, and not as well correlated, on average,
16 in the upper midwest and southwest regions. PM10 is fairly well correlated with estimated PM10.
17 25 in most regions, with the highest average correlation in the upper midwest and southwest
18 regions. PM10 is more highly correlated, on average, with PM25 than with estimated PM10_25 in
19 the northeast and industrial midwest regions. Their correlations are similar in the southeast, and
20 PM10 is more highly correlated, on average, with PM10_2 5 in the northwest and southern
21 California regions. These data suggest that PM10 might be a suitable indicator for either fme-
22 fraction or coarse-fraction particles, depending upon the locational factors. However, in all
23 locations estimated PM10_2 5 and PM2 5 are very poorly correlated, which should be expected due
24 to their differences in origin, composition, and atmospheric behavior.
August 2003 2-41 Draft - Do Not Quote or Cite
-------
1.50-
i.oo---
1
0.50---
o.oo-l
I
Northeast
N = 68
Southeast
86
Industrial
Midwest
106
Upper
Midwest
39
Southwest
31
Northwest
109
Southern
California
36
Figure 2-16. Distribution of ratios of annual mean PM2 5 to PM10 by region, 2000-2002.
Box depicts interquartile range and median; whiskers depict 5th and 95th
percentiles; asterisks depict minimum and maximum. Number below indicates
the number of sites in each region.
Source: Schmidt et al. (2003)
1 2.6 TEMPORAL PATTERNS IN PM CONCENTRATIONS
2 2.6.1 PM25 and PM10_25 Patterns
3 Data from the PM FRM network from 2000-2002 show distinct seasonal variations in
4 PM25 and estimated PM10_25 concentrations. Figure 2-18 shows the monthly distribution of 24-
5 hour average urban PM2 5 concentrations in different geographic regions. The months with peak
6 urban PM2 5 concentrations vary by region. The urban areas in the northeast, industrial midwest,
August 2003
2-42
Draft - Do Not Quote or Cite
-------
0 9 -
0 8 -
0 7
¥
oi n R
.- U.D
1
0 05
c
.2
13 04
0) w ^
o
0 n-
0 2
A
A
n
n
* A A
D
A D
A
D
*
*
* * *
ro ro cu CD cu cu
.c .c -! -! ! !
| o ^ i g o
"w CL
Z3 ZJ
c
»PM2.5vs. estimated PM10-2.5 DPMIOvs. PM2.5 APM10 vs. estimated PM10-2.5
Southern California
Figure 2-17. Regional average correlations of 24-hour average PM by size fraction.
Source: Schmidt et al. (2003)
1 and upper midwest regions all exhibit peaks in both the winter and summer months. In the
2 northeast and industrial midwest regions the summer peak is slightly more pronounced than the
3 winter peak, and in the upper midwest region the winter peak is slightly more pronounced than
4 the summer peak. In the southeast a single peak period in the summer is evident. In western
5 regions peaks occur in the late fall and winter months.
6 Figure 2-19 shows the distributions of estimated 24-hour average urban PM10_2 5
7 concentrations in different geographic regions. In all regions except the southwest, the lowest
8 concentrations occur in the winter months. Elevated levels are apparent in the easternmost
9 regions in either April or May. In the upper midwest, northwest, and southern California regions
10 the highest levels occur in the late summer to early fall. The southwest region exhibits the
August 2003
2-43
Draft - Do Not Quote or Cite
-------
Figure 2-18. Urban 24-hour average PM2 5
concentration distributions by region and
Northeast
range and median; line connects monthly 35_
means. 30_
25-
Source: Schmidt et al. (2003) 20_
15-
10-
5-
o-
40-
35-
30-
25-
20-
15-
10-
5-
o-
40-
35-
30-
25-
20-
15-
10-
5-
o-
40-
35-
30-
25-
20-
15-
10-
5-
o-
Upper Midwest
35-
30-
25-
074 207. 222B 20-
2=53 2033 2]BO 2"3 215B 15-
- ~ ^"&^flnnfl-fl-&B 10"
o-
4=0= 403= 5 B ™°
I 1 43=7 I I 4B12
4969 5 67 4992
LJnR'^ 0 D n = D"^
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Industrial Midwest
7695 79B9 794
7B99
| 1 7730 i 1 ^ — ~"---_^ | 1
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Northwest
40-
35-
30-
45=, 25 -
4045 4756
43=0 I 1 20-
_^X 4053 4405 ^ _/ — _
o-
Southeast
B726 B5BO
ftftftB^^^
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Southwest
40-
35-
30-
25-
20-
«« n 15^
o-
Southern California
505 15Q7
1579
\ 40B ]555 ]4B] /"^-
\ 1 1 450 ™° ,432 505 450 /
D -ftffl ' W D D
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
August 2003
2-44
Draft - Do Not Quote or Cite
-------
Figure 2-19. Urban 24-hour average
estimated PM10_2 5 concentration distributions
Northeast
oy region anu iiioiiui, zuuu-zuuz. BOX ciepicis 45n
interquartile range and median; line connects 40
monthly means.
25-
20-
Source: Schmidt et al. (2003) 15
10-
5-
o-
-5 -
45-
40-
35-
30-
25-
20-
15-
10-
5-
o-
-5 -
45-
40-
35-
30-
25-
20-
15-
10-
5-
o-
-5 -
025 BBB
H^nn nBHEJ^R^i
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Upper Midwest Industrial Midwest
40-
35-
30-
552 539 25 -
50, 555 513 547
I 1 I II 1 \ ™ 2°~
533 5BB L — -^ ~~--~^^_ 530
§4§n0°° D nHfl :
o-
-5 -
Btiiiwfiii
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec ^ Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Northwest
45-
40-
35-
30-
657 559
,074 7,5 ,0,4 1B2B I" I |— I 770 2Q .
[^r""r\~ D D — — U ~"~M-O 10"
n D U n 5"
o-
-5 -
Southeast
,32 ,,B,
r;VS~H^nH~^""^fl~H^^^H^n
|-j n ti anuHn^y
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Southwest
Southern California
40-
35-
30-
25-
20-
15-
10-
5-
o-
-5 -
40-
527 M1 533 499 520 ™ 35-
~~
40,
—
^
1
—
/
/~~-
\
~^
500
5,5 ! !
533
\
~^
-—
— "
1—1 i— ' l_l
—
__
^
—
547
| — , 30-
25-
20-
15-
10-
5-
o-
-5 -
333
333
—
—
| 1
-TT
U
370
~
_
373
3B5 | 1
392 3BB 394
^L
—
— '
—
t—
' — '
—
!
^
—
300
\n 35,
V i — i
x
—
—
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
August 2003
2-45
Draft - Do Not Quote or Cite
-------
1 greatest range of variability throughout the year. Elevated levels are apparent in the spring,
2 consistent with winds that contribute to blowing dust.
3 The relationship between the annual mean at a site and the shorter-term 24-hour average
4 peaks is useful for examining the relationships between short- and long-term air quality
5 standards. The box plots in Figures 2-20 and 2-21 show the relationships for PM25 and
6 estimated PM10_2 5, respectively, between annual mean PM concentrations and peak daily
7 concentrations as represented by the 98th percentile of the distribution of daily average
8 concentrations at FRM sites across the U.S. There is considerable variability in peak daily
9 values for sites with similar annual means. For annual mean PM25 values between 10 and 15
10 Hg/m3 the interquartile range spans about 5 to 6 |ig/m3 for each 1 |ig/m3 interval. The range
11 between the 5th and 95th percentile values for each interval varies substantially. Estimated PM10.
12 2.5 generally exhibits greater variability in 98th percentile values for sites with similar annual
13 means than seen for PM2 5 The maximum estimated PM10_2 5 values are quite high relative to the
14 rest of the distribution starting with the annual mean interval over 16 |ig/m3.
15 Monitors that provide near-continuous measurements can provide insights into short-term
16 (e.g., hourly average) patterns in PM, which could be important to understanding associations
17 between elevated PM levels and adverse health effects. Examples of average hourly profiles for
18 PM2 5 and PM10_2 5 from 2000-2002 are shown in Figures 2-22 and 2-23 for a monitoring site in
19 the Cleveland, OH metropolitan area. The PM2 5 profile in Figure 2-22 indicates elevated hourly
20 average levels occurred most often between the hours of 6:00 am and 9:00 am, corresponding to
21 the typical morning rush of automobile traffic, with the highest levels in the summer and fall.
22 The 95th percentile concentrations during peak hours can be as high as four times the median
23 level for the same hour. Slightly lower peak levels occur between the hours of 10:00 pm and
24 midnight, corresponding to the onset of the nighttime inversion. This profile of hourly average
25 PM2 5 levels is typical of many, but not all, urban areas. The PM10_2 5 profile in Figure 2-23
26 indicates a pattern of hourly average levels similar to the pattern for PM25.
August 2003 2-46 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
100
90--
= 60
OJ
u
k_
OJ
£b
CD
50
40
10
Annual mean
(ug/m3)
56 6-8 8-9 9-10 10-11 11-12 12-13 13-14 14-15 15-16 16-17 >1 7
N= 49 90 69 86 112 138 113 128 126 111 67 73
th
Figure 2-20. Distribution of annual mean vs. 98 percentile 24-hour average PM2
-2.5
concentrations, 2000-2002. Box depicts interquartile range and median;
whiskers depict 5th and 95th percentiles; asterisks depict maximum and minimum.
N = number of sites in each interval.
The hourly ranges shown in Figure 2-22 suggest hour-to-hour changes in PM2 5
concentrations encompass several |ig/m3, however, extreme values for hour-to-hour variations
can be much larger (Fitz-Simons et al., 2000). An analysis of the distribution of increases in
hour-to-hour concentrations at multiple sites across the U.S. for 1999 finds site-level median
increases of about 1 |ig/m3 to 30 |ig/m3, and 95th percentile increases ranging from about 4 |ig/m3
to 16 |ig/m3.
August 2003
2-47
Draft - Do Not Quote or Cite
-------
140
130-
110
1
g 90-
6 80
0 ™
"c
0 Crt
O bO
30
= 88 73 81 59 58 39 47 32 11
Figure 2-21. Distribution of estimated annual mean vs. 98th percentile 24-hour average
PM10_25 concentrations, 2000-2002. Box depicts interquartile range and
median; whiskers depict 5th and 95th percentiles; asterisks depict minimum and
maximum. N = number of sites in each interval.
Source: Schmidt et al. (2003)
4
5
6
2.6.2 Ultrafine Patterns
Diurnal or seasonal patterns for ultrafine particles have been extensively studied in
relatively few areas of the U.S. At an urban site in Atlanta, GA, ultrafine particle number
concentrations were found to be higher on weekdays than on weekends (Woo et al., 2001).
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
August 2003
2-48
Draft - Do Not Quote or Cite
-------
60-
50-
40-
30-
20-
10-
o-L
60-
50-
40-
o
o>
o
c
o
O
O.
20-
10-
Oi
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
^~^~ Spring (Mar-May)
~~ ~~ •" Summer (Jun-Aug)
Fall (Sep-Nov)
— Winter(Dec-Feb)
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 2-22. Hourly average VM25 concentrations at a Cleveland, OH monitoring
site, 2000-2002. Upper panel shows the distribution of concentrations (box
plot of interquartile range, mean, median, 5th and 95th percentiles); lower
panel shows seasonal averages by hour.
Source: Schmidt et al. (2003)
August 2003
2-49
Draft - Do Not Quote or Cite
-------
110-
100-
90-
80-
70-
60-
50-
40-
30-
20-
10-
o-
V I
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
110-
100-
90-
1 70;
c
.9. 60-
1
I 50-
o
<3 40-
S
J 30-
D.
20-
10-
o-
Spring (Mar-May)
Summer (Jun-Aug)
Fall (Sep-Nov)
Winter(Dec-Feb)
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 2-23. Hourly average PM10_2 5 concentrations at a Cleveland, OH monitoring
site, 2000-2002. Upper panel shows the distribution of concentrations
(box plot of interquartile range, mean, median, 5th and 95th percentiles);
lower panel shows seasonal averages by hour.
Source: Schmidt et al. (2003)
August 2003
2-50
Draft - Do Not Quote or Cite
-------
1 traffic is heaviest. Smaller particles in the range of 0.004 to 0.01 jim were elevated during the
2 peak traffic period, most notably in cooler temperatures, below 50°F. Kim et al. (2002) report
3 similar results for Riverside, California. At the Atlanta site, episodes of relatively high ultrafme
4 particle levels were observed during corresponding periods of peak SO2 levels. In El Paso, TX,
5 ultrafme particle concentrations were also shown to vary in patterns similar to traffic volumes
6 (Noble et al., 2003).
7
8 2.7 PM BACKGROUND LEVELS
9 For the purposes of this document, background PM is defined as the distribution of PM
10 concentrations that would be observed in the U.S. in the absence of anthropogenic, or man-made,
11 emissions of primary PM and precursor emissions of VOC, NOX, SO2, and NH3 in North
12 America. Thus, background includes PM from natural sources and transport of PM from both
13 natural and man-made sources outside of North America. Estimating background PM
14 concentrations is important for the health risk analyses presented in Chapter 4and the assessment
15 of ecosystem and visibility effects in Chapter 5. The draft CD states that "recent but limited
16 information about background PM concentrations have not provided sufficient evidence to
17 warrant any changes in estimates of the annual average background concentrations given in the
18 1996 PM AQCD" (CD, p. 3-104). However, the draft CD does discuss the increasing
19 recognition and understanding of the long-range transport of PM from outside the U.S.
20 Background levels of PM vary by geographic location and season, and have a natural
21 component and an anthropogenic component. The natural background arises from: (1) physical
22 processes of the atmosphere that entrain coarse-mode particles (e.g., windblown crustal material,
23 sea salt spray); (2) volcanic eruptions (e.g., sulfates); (3) natural combustion such as wildfires
24 (e.g., elemental and organic carbon, and inorganic and organic PM precursors); and (4) biogenic
25 sources such as vegetation, microorganisms, and wildlife (e.g., organic PM, inorganic and
26 organic PM precursors). The exact magnitude of the natural portion of PM for a given
27 geographic location cannot be precisely determined because it is difficult to distinguish local
28 sources of PM from the long-range transport of anthropogenic particles and precursors.
29 PM can be transported long distances from natural or quasi-natural events occurring
30 outside the continental U.S. (CD, p. 3-68). The occurrence and location of these long-range
August 2003 2-51 Draft - Do Not Quote or Cite
-------
1 transport events are highly variable and their impacts on the U.S. are equally variable. The
2 contributions to background from sources outside of North America can be significant on an
3 episodic, but probably not on an annual basis (CD, p. 3-89). Several studies have focused on
4 identifying the origin, sources, and impacts of recent transnational transport events from North
5 America and extra-continental sources.
6
7 • The transport of PM from biomass burning in Central America and southern Mexico in
8 1998 has been shown to contribute to elevated PM levels in southern Texas and
9 throughout the entire central and southeastern United States (CD, p. 3-71).
10
11 • Wildfires in the boreal forests of northwestern Canada may impact large portions of the
12 eastern United States. The CD estimates that a July 1995 Canadian wildfire episode
13 resulted in excess PM2 5 concentrations ranging from 5 |ig/m3 in the Southeast, to nearly
14 100 |ig/m3 in the northern Plains States (CD, p. 3-73).
15
16 • Windblown dust from dust storms in the North African Sahara desert has been observed
17 in satellite images as plumes crossing the Atlantic Ocean and reaching the southeast coast
18 of the U.S., primarily Florida; North African dust has also been tracked as far as Illinois
19 and Maine. These events have been estimated to contribute 6 to 11 |ig/m3 to 24-hour
20 average PM25 levels in affected areas during the events (CD, p. 3-69).
21
22 • Dust transport from the deserts of Asia (e.g., Gobi, Taklimakan) across the Pacific Ocean
23 to the northwestern U.S. also occurs. Husar et al. (2000) report that the average PM10
24 level at over 150 reporting stations throughout the northwestern U.S. was 65 |ig/m3
25 during an episode in the last week in April 1998, compared to an average of about 20
26 Hg/m3 during the rest of April and May (CD, p. 3-69).
27
28 Section 3.3.3 of the draft CD provides broad estimates of annual average background PM
29 levels, as shown in Table 2-6. The lower bounds of the ranges are based on compilations of
30 natural versus anthropogenic emissions levels, ambient measurements in remote areas, and
31 regression studies using anthropogenic and/or natural tracers (the results from 52 publications
32 were used in this effort) (NAPAP, 1991). The upper bounds are derived from the multi-year
33 annual averages of the "clean" remote monitoring sites in the IMPROVE network (Malm et al.,
34 1994). Since the IMPROVE data reflect the effects of anthropogenic emissions from within
35 North America as well as background, they may overestimate the actual background
36 concentrations. There is a distinct geographic difference in background levels, with lower levels
August 2003 2-52 Draft - Do Not Quote or Cite
-------
Western U.S. (ug/m3)
PM10
PM25
PM10.2.5
4
1
<1
-8
-4
-7
Eastern U.S. (ug/m3)
5- 11
2-5
<1 - 9
1 in the western U.S. and higher levels in the eastern U.S. The eastern U.S. is estimated to have
2 more natural organic fine-mode particles and more water associated with hygroscopic fine-mode
3 particles than the western U.S. due to generally higher humidity levels. For PM10_2 5, the draft
4 CD (p. 3-82) presents a very rough point estimate of 3 |ig/m3 for an annual average in both the
5 eastern and western U.S.
6
7 Table 2-6. Estimated Range of Annual Average PM Regional Background Levels
8
9
10
11
12 Source: draft CD, p. 3-82
13
14 Over shorter periods of time (e.g., days or weeks), the range of expected background
15 concentrations is much broader. Specific natural events such as wildfires, volcanic eruptions,
16 and dust storms, both of North American and international origin, can lead to very high levels of
17 PM comparable to, or greater than, those driven by man-made emissions in polluted urban
18 atmospheres. Because such excursions are essentially uncontrollable, EPA has in place a
19 "natural events" policy that removes consideration of them from attainment decisions18.
20 Disregarding such large and unique events, an estimate of the range of "typical"
21 background on a daily basis can be obtained from reviewing multi-year data at remote locations.
22 In the last review, EPA staff analyzed PM2 5 concentrations from selected IMPROVE sites and
23 found 24-hour peak to annual mean ratios in the range of two to four. Applying these ratios to
24 estimated annual background levels suggested that the highest 24-hour PM2 5 background
25 concentrations over the course of a year could be in the range of 15 to 20 |ig/m3 (EPA, 1996b, p.
1 8
Under this policy (Nichols, 1996), 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 PM events are specified: volcanic or seismic activity,
wildland fires, and high wind dust events.
August 2003 2-53 Draft - Do Not Quote or Cite
-------
1 IV-14). Based on current air quality information, EPA staff has conducted an analysis of five
2 recent years (1997-2001) of data at IMPROVE sites across the U.S. (Langstaff, 2003a). Sites
3 with five-year means below the upper end of the ranges for annual mean background in Table 2-
4 6 in the eastern and western U.S. were selected for analysis. There are three such sites in the
5 East and 31 in the West19. Staff focused on the 99th percentile concentrations at each of these
6 sites, so as to avoid including excursions that likely reflect exceptional natural events. The 99th
7 percentile concentrations at the eastern sites ranged from 13 to 17 |ig/m3, and from 6 to 20 |ig/m3
8 at the western sites. Even though these monitors are probably among the least likely to be
9 impacted by North American anthropogenic pollution, there will be some impact, and therefore
10 they will still tend to overestimate background concentrations. Considering these factors, the
11 distributions of daily PM25 concentrations at these sites provide an indication of the ranges for
12 the daily variability of PM25 background concentrations, and the 99th percentiles of these
13 distributions are a reasonable estimate of the highest daily background concentrations. Staff
14 notes that these recent findings are generally consistent with those from the last review,
15 continuing to suggest a range of about 15 to 20 |ig/m3 as the upper end of the distribution of
16 daily PM2 5 background concentrations in the U.S.
17
18 2.8 RELATIONSHIP BETWEEN HUMAN EXPOSURE TO AMBIENT PM AND
19 CENTRAL MONITOR MEASUREMENTS OF PM
20 The statutory focus of the primary NAAQS for PM is to protect public health from the
21 adverse effects associated with the presence of PM in the ambient air - that is, the focus is on
22 particles in the atmosphere that are emitted by sources to the outdoors or formed by chemical
23 reactions in the atmosphere. We refer to the concentrations of PM in the ambient air as ambient
24 PM. An understanding of human exposure to ambient PM helps inform the evaluation of
25 underlying assumptions and interpretation of results of epidemiologic studies that characterize
26 relationships between monitored ambient PM concentrations and observed health effects
27 (discussed in Chapter 3).
19 One of the western sites (Lasson Volcanic National Park) was removed from the analysis since it had
anomalously high concentrations.
August 2003 2-54 Draft - Do Not Quote or Cite
-------
1 An important exposure-related issue for this review is the characterization of the
2 relationships between ambient PM concentrations measured at one or more centrally located
3 monitors and personal exposure to ambient PM, as characterized by particle size, composition, or
4 other factors. Information on the type and strength of these relationships, discussed below, is
5 relevant to the evaluation and interpretation of associations found in epidemiologic studies that
6 use measurements of PM concentrations at centrally located monitors as a surrogate for exposure
7 to ambient PM.20 The focus here is on particle size distinctions; the draft CD (CD, Section 5.4)
8 also discusses exposure relationships related to compositional differences.
9
10 2.8.1 Definitions
11 Exposure to a contaminant is defined as contact at a boundary between a human and the
12 environment (e.g., the breathing zone) at a specific contaminant concentration for a specific
13 interval of time; it is measured in units of concentration(s) multiplied by time (or time interval)
14 (National Research Council, 1991). An individual's total personal exposure to PM results from
15 breathing air containing PM in different types of environments (e.g., outdoors near home,
16 outdoors away from home, indoors at home, indoors at office or school, commuting, restaurants,
17 malls, other public places) These environments may have different concentrations of PM with
18 particles originating from a wide variety of sources.
19 Ambient PM is comprised of particles emitted by anthropogenic and natural sources to
20 the outdoors and particles formed in the atmosphere from emissions of gaseous precursors. This
21 includes emissions not only from outdoor sources such as smokestacks and automobiles, but also
22 from sources located indoors, such as fireplaces, wood stoves, and some industrial processes,
23 that are vented to the outdoors. Exposure to ambient PM can occur both outdoors and indoors to
24 the extent that ambient PM penetrates into indoor environments - we use the term PM of
25 ambient origin to refer to both outdoor and indoor concentrations of ambient PM. We use the
26 term nonambient PMio refer to concentrations of PM that are due to other indoor sources of
27 particles such as smoking, cooking, other sources of combustion, cleaning, resuspension of
20 Consideration of exposure measurement error and the effects of exposure misclassification on the
interpretation of the epidemiologic studies are addressed in Chapter 3.
August 2003 2-55 Draft - Do Not Quote or Cite
-------
1 particles, mechanical processes, and chemical interactions occurring indoors. In characterizing
2 human exposure to PM concentrations relevant to setting standards for ambient air quality, the
3 draft CD conceptually separates an individual's total personal exposure to PM into exposure to
4 PM of ambient origin and exposure to all other sources of PM (i.e., nonambient exposure}.
5 Outdoor concentrations of PM are affected by emissions, meteorology, atmospheric
6 chemistry, and removal processes. Indoor concentrations of PM are affected by several factors,
7 including outdoor concentrations, processes that result in infiltration of ambient PM into
8 buildings, indoor sources of PM, aerosol dynamics and indoor chemistry, and removal
9 mechanisms such as particle deposition, ventilation, and air-conditioning and air cleaning
10 devices (CD, p. 5-119). Concentrations of PM inside vehicles are subject to essentially the same
11 factors as concentrations of PM inside buildings. Personal exposure to PM also includes a
12 component which results specifically from the activities of an individual that typically generate
13 particles affecting only the individual or a small localized area surrounding the person, such as
14 walking on a carpet.
15 Epidemiologic studies use measurements from central monitors to represent the variety of
16 concentrations in an area. We use the term central site to mean the site of a PM monitor
17 centrally located with respect to the area being studied. In some cases epidemiologic studies
18 combine the measurements from more than one monitor to obtain a broader representation of
19 area-wide PM concentrations than a single monitor provides.
20
21 2.8.2 Centrally Monitored PM Concentration as a Surrogate for Particle Exposure
22 The 1996 Criteria Document (EPA, 1996a) presented a thorough review of PM exposure-
23 related studies up to that time. The previous Staff Paper (EPA, 1996b) drew upon the studies,
24 analyses, and conclusions presented in the 1996 Criteria Document and discussed two
25 interconnected PM exposure issues: (1) the ability of central fixed-site PM monitors to represent
26 population exposure to ambient PM and (2) how differences between fine and coarse particles
27 affect population exposures. Distinctions between PM size classes and components were found
28 to be important considerations in addressing the representativeness of central monitors. For
29 example, fine particles have a longer residence time and generally exhibit less variability in the
30 atmosphere than the coarse fraction. As discussed in the 1996 Staff Paper (EPA, 1996b), the
August 2003 2-56 Draft - Do Not Quote or Cite
-------
1 1996 Criteria Document concluded that measurements of daily variations of PM have a plausible
2 linkage to daily variations of human exposures to PM of ambient origin for the populations
3 represented by the ambient monitoring stations, and that this linkage is stronger for fine particles
4 than for PM10 or the coarse fraction of PM10. The 1996 Criteria Document further concluded that
5 central monitoring can be a useful, if imprecise, index for representing the average exposure of
6 people in a community to PM of ambient origin (EPA, 1996b, p. IV-15, 16).
7 Exposure studies published since 1996 and reanalyses of studies that appeared in the
8 1996 Criteria Document are reviewed in the draft CD, and provide additional support for the
9 findings made in the 1996 Criteria Document and 1996 Staff Paper.
10 An individual's total personal exposure to PM generally differs from the ambient
11 concentration measured at the central site monitor because: (1) spatial differences in ambient PM
12 concentrations exist across a city or region; (2) generally only a fraction of the ambient PM is
13 present in indoor or in-vehicle environments, whereas individuals generally spend a large
14 percentage of time indoors; and (3) a variety of indoor sources of PM contribute to total personal
15 exposure. Thus, the amount of time spent outdoors, indoors, and in vehicles and the types of
16 activities engaged in (e.g., smoking, cooking, vacuuming) also will heavily influence personal
17 exposure to PM.
18 With regard to the first factor that influences the relationship between total personal
19 exposure and concentrations measured at central sites, the spatial variability of PM plays a large
20 role. As discussed in Section 2.5, for many areas PM25 concentrations are fairly uniform
21 spatially, with higher concentrations near roadways and other direct sources of PM25. Analyses
22 of PM2 5 data for 27 urban areas indicate that differences in annual mean concentrations between
23 monitoring sites in an urban area range from less than 1 |ig/m3 to as much as 8 |ig/m3. However,
24 the correlations of daily PM2 5 between sites are typically high. Daily mean PM2 5 concentrations
25 exhibit much higher spatial variability than annual means, even when the daily concentrations at
26 sites are highly correlated. Although the spatial variability of PM25 varies for different urban
27 areas, overall, some degree of uniformity results from the widespread formation and long
28 lifetime of the high regional background of secondary PM2 5. In summarizing the key findings
29 related to spatial variability in PM25 concentrations, the draft CD states (p. 3-103):
August 2003 2-57 Draft - Do Not Quote or Cite
-------
1 Although PM2 5 concentrations within an MSA can be highly correlated between
2 sites, there can still be significant differences in their concentrations. The degree
3 of spatial uniformity in PM2 5 concentrations and the strength of site to site
4 correlations in urban areas varies across the country. These factors should be
5 considered in using data obtained by the PM2 5 FRM network to approximate
6 community-scale human exposures, and caution should be exercised in
7 extrapolating conclusions as to spatial uniformity or correlations obtained in one
8 urban area to another. Limited information also suggests that the spatial
9 variability in urban source contributions is likely to be larger than for regional
10 source contributions to PM2 5 and for PM2 5, itself.
11
12 Relative to fine particles, coarse particles are likely to be more variable across urban scales.
13 Daily mean PM10_2 5 concentrations tend to be more variable and have lower inter-site
14 correlations than PM2 5, possibly due to their shorter atmospheric lifetime and the more sporadic
15 nature of PM10_25 sources (CD, Section 3.2.5).
16 The second factor influencing the relationship between ambient PM concentrations
17 measured at central sites and total personal exposure to PM is the extent to which ambient PM
18 penetrates indoors and remains suspended in the air. If the flow of ambient PM into the home
19 from the outdoors is very restricted, the relationship between ambient PM concentrations
20 measured at a central site and total exposure to PM will be weaker than in a situation where
21 ambient PM flows more readily into the home and is a greater part of the overall indoor PM
22 concentrations. This is heavily dependent on the air exchange rate, and also on penetration
23 efficiency and deposition or removal rate, both of which vary with particle aerodynamic size.
24 Air exchange rates (the rates at which the indoor air in a building is replaced by outdoor air) are
25 influenced by building structure, the use of air conditioning and heating, opening and closing of
26 doors and windows, and meteorological factors (e.g., difference in temperature between indoors
27 and outdoors). Based on physical mass-balance considerations, usually the higher the air
28 exchange rate the greater the fraction of PM of ambient origin found in the indoor and in-vehicle
29 environments. Rates of infiltration of outdoor PM into homes through cracks and crevices are
30 higher for PM25 than for PM10, PM10_25, or ultrafme particles (CD, p. 5-120). Since PM10_25
31 infiltrates indoors less readily than PM2 5 and settles out more rapidly than PM2 5, a greater
32 proportion of PM2 5 of ambient origin is found indoors than PM10_2 5, relative to their outdoor
August 2003 2-58 Draft - Do Not Quote or Cite
-------
1 concentrations. Thus, particle size also influences the amounts of PM of ambient origin found
2 indoors.
3 The third factor influencing the relationship between ambient concentrations measured at
4 central sites and total personal exposure is the contribution of indoor sources to total personal
5 exposure. On average, individuals spend nearly 90 percent of their time indoors. The
6 contribution of indoor sources to indoor concentrations of PM is significant, and can be quite
7 variable on different days and between individuals. Indoor sources such as combustion devices
8 (e.g., stoves and kerosene heaters) generate predominantly fine particles; cooking produces both
9 fine and coarse particles; and resuspension (e.g., dusting, vacuuming, and walking on rugs)
10 generates predominantly coarse particles (CD, p. 5-79).
11 These three factors related to total personal exposure can give rise to measurement error
12 (CD, p. 5-113 to 5-118), which makes the quantification of relationships between concentrations
13 measured at central site monitors and health effects more difficult due to reduction in statistical
14 power. However, as discussed in the draft CD and below in Chapter 3, exposure measurement
15 errors are not expected to influence the interpretation of findings from either the chronic or time-
16 series epidemiological studies that have used ambient concentration data (CD, p. 5-117).
17 In looking more specifically at the relationship between personal exposure to PM of
18 ambient origin and concentrations measured at central site monitors, an analysis of data from the
19 PTEAM study21 provides important findings, as discussed in the draft CD (p. 5-60 to 5-63 and 5-
20 122 to 5-123). The PTEAM study demonstrated that central site ambient PM concentrations are
21 well correlated with personal exposure to PM of ambient origin, while such concentrations are
22 only weakly correlated with total personal exposure. This study also found that estimated
23 exposure to nonambient PM is effectively independent of PM concentrations at central site
24 monitors, and that nonambient exposures are highly variable due to differences in indoor sources
25 across the study homes.
21 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 PM, 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.
August 2003 2-59 Draft - Do Not Quote or Cite
-------
1 In other studies in which subjects experienced small indoor source contributions to their
2 personal exposures, such that total exposure is mostly from PM of ambient origin, high
3 correlations are generally found between total personal exposure and ambient PM measured at a
4 central site (CD, p. 5-51). For example, measurements of ambient sulfate, which is mostly in
5 fine PM, have been found to be highly correlated with total personal exposure to sulfate (CD, p.
6 5-121). Since in these studies there were minimal indoor sources of sulfate, the relationship
7 between ambient concentrations and total personal exposure to sulfate was not weakened by
8 variability in indoor-generated sulfates, explaining the higher correlations.
9 The draft CD discusses the finding by some researchers that some epidemiologic studies
10 yield statistically significant associations between ambient concentrations measured at a central
11 site and health effects even though there is a very small correlation between ambient
12 concentrations measured at a central site and total personal exposures (CD, section 5.3.2.4). The
13 explanation of this finding is that, as discussed above, total personal exposure includes both
14 ambient and nonambient generated components, and while the nonambient portion of personal
15 exposure is not generally correlated with ambient concentrations, the exposure to ambient
16 concentrations is correlated with ambient concentrations. Thus, it is not surprising that health
17 effects might correlate with central site PM concentrations, because exposure to ambient PM
18 correlates with these concentrations, and the lack of correlation of total exposure with central site
19 PM concentrations does not alter that relationship. By their statistical design, time-series
20 epidemiologic studies of this type only address the ambient component of exposure.
21 There is not a clear consensus among exposure analysts at this time as to how well
22 community monitor measurements of ambient air PM concentrations represent a surrogate for
23 personal exposure PM of ambient origin. It is recognized that existing PM exposure
24 measurement errors or uncertainties most likely will reduce the statistical power of PM health
25 effects analyses, thus making it difficult to detect a true underlying association between the
26 exposure metric and the health outcome. However, the use of ambient PM concentrations as a
27 surrogate for exposures is not expected to change the principal conclusions from PM
28 epidemiological studies that use community average health and pollution data. Based on these
29 considerations and on the review of the available exposure-related studies, the draft CD
August 2003 2-60 Draft - Do Not Quote or Cite
-------
1 concludes that for time-series epidemiology, ambient PM concentration as measured at central
2 site monitors is a useful surrogate for exposure to PM of ambient origin (CD, p. 9-117).
O
4 2.9 OPTICAL AND RADIATIVE PROPERTIES OF PARTICLES
5 By scattering and absorbing electromagnetic radiation, ambient particles can impair
6 visibility, affect climate processes, and affect the amount of ultraviolet radiation that reaches the
7 earth. The sun emits electromagnetic radiation across the full spectrum, including at ultraviolet
8 (0.028 to 0.4 |im) and visible (0.4 to 0.7 jim) wavelengths, while the earth emits at thermal
9 infrared (0.7 to 3.2 jim) wavelengths. The effects of ambient particles on the transmission of
10 these segments of the electromagnetic spectrum depend on the radiative properties of the
11 particles, which in turn are dependent on the size and shape of the particles, their composition,
12 the distribution of components within individual particles, and their vertical and horizontal
13 distribution in the lower atmosphere.
14
15 2.9.1 PM Properties Affecting Visibility
16 Visibility is affected by scattering and absorption of light in visible wavelengths by
17 particles and gases in the atmosphere. As discussed in section 4.3 of the draft CD, the efficiency
18 of particles in causing visibility impairment depends on particle size, shape, and composition.
19 Accumulation mode particles, within the fine mode, are generally most effective in impairing
20 visibility. Accumulation-mode particle components principally responsible for visibility
21 impairment are sulfates, nitrates, organic matter, black carbon, and soil dust. All such particles
22 scatter light to some degree, but only black carbon plays a significant role in light absorption.
23 Since black carbon, which is a product of incomplete combustion from activities such as the
24 burning of wood or diesel fuel, is a relatively small component of PM in most areas, impairment
25 is generally dominated by scattering rather than absorption.
26 Because humidity causes hygroscopic particles to grow in size, humidity plays a
27 significant role in particle-related impairment. The amount of increase in particle size with
28 increasing relative humidity depends on particle composition. Humidity-related particle growth
29 is a more important factor in the eastern U.S., where annual average relative humidity levels are
August 2003 2-61 Draft - Do Not Quote or Cite
-------
1 70 to 80 percent compared to 50 to 60 percent in the western U.S. Due to relative humidity
2 differences, aerosols of a given mass, size distribution, and composition would likely cause
3 greater visibility impairment in an eastern versus a western location.
4
5 2.9.2 PM Properties Affecting Climate
6 The effects of PM on the transfer of radiation in the visible and infrared spectral regions
7 also play a role in global and regional climate. As discussed in section 4.5.1 of the draft CD,
8 particles can have both direct and indirect effects on climatic processes. The direct effects are
9 the result of the same processes responsible for visibility degradation, namely radiative
10 scattering and absorption. However, while visibility impairment is caused by particle scattering
11 in all directions, climate effects result mainly from scattering light away from the earth and into
12 space. This reflection of solar radiation back to space decreases the transmission of visible
13 radiation to the surface and results in a decrease in the heating rate of the surface and the lower
14 atmosphere. At the same time, absorption of either incoming solar radiation or outgoing
15 terrestrial radiation by particles, primarily black carbon, results in an increase in the heating rate
16 of the lower atmosphere.
17 The extent to which ambient particles scatter and absorb radiation is highly dependent on
18 their composition and optical properties and on the wavelength of the radiation. For example,
19 sulfate and nitrate particles effectively scatter solar radiation, and they weakly absorb infrared,
20 but not visible, radiation. The effects of mineral dust particles are complex; depending on
21 particle size and degree of reflectivity, mineral aerosol can reflect or absorb radiation. Dark
22 minerals absorb across the solar and infrared radiation spectra leading to warming of the
23 atmosphere. Light-colored mineral particles in the appropriate size range can scatter visible
24 radiation, reducing radiation received at the earth's surface. Organic carbon particles mainly
25 reflect radiation, whereas black carbon particles strongly absorb radiation; however, the optical
26 properties of carbonaceous particles are modified if they become coated with water or sulfuric
27 acid. Upon being deposited onto surfaces, particles can also either absorb or reflect radiation
28 depending in part on the relative reflectivity of the particles and the surfaces on which they are
29 deposited.
August 2003 2-62 Draft - Do Not Quote or Cite
-------
1 In addition to these direct radiative effects, particles can also have a number of indirect
2 effects on climate related to their physical properties. For example, sulfate particles can serve as
3 condensation nuclei which alter the size distribution of cloud droplets by producing more
4 droplets with smaller sizes. Because the total surface area of the cloud droplets is increased, the
5 amount of solar radiation that clouds reflect back to space is increased. Also, smaller cloud
6 droplets have a lower probability of precipitating, causing them to have longer atmospheric
7 lifetimes. An important consequence of this effect on cloud properties is the suppression of rain
8 and potentially major disruption of hydrological cycles downwind of pollution sources, leading
9 to a potentially significant alteration of climate in the affected regions (CD, p. 4-218).
10 The overall radiative and physical effects of particles, both direct and indirect, are not the
11 simple sum of effects caused by individual classes of particles because of interactions between
12 particles and other atmospheric gases (CD, p. 4-219). As discussed in Section 4.5.1.2 of the
13 draft CD, the effects of sulfate particles have been the most widely considered, with globally
14 averaged radiative effects of sulfate particles generally estimated to have partially offset the
15 warming effects caused by increases in greenhouse gases. On the other hand, global-scale
16 modeling of mineral dust particles suggests that even the sign as well as the magnitude of effects
17 depends on the vertical distribution and effective particle radius.
18 The draft CD makes clear that atmospheric particles play an important role in climatic
19 processes, but that their role at this time remains poorly quantified. In general, on a global scale,
20 the direct effect of radiative scattering by atmospheric particles is to likely exert an overall net
21 effect of cooling the atmosphere, while particle absorption may lead to warming. The net impact
22 of indirect effects on temperature and rainfall patterns remains difficult to generalize. However,
23 deviations from global mean values can be very large even on a regional scale, with any
24 estimation of more localized effects introducing even greater complexity (CD, p. 217). The draft
25 CD concludes that any effort to model the impacts of local alterations in particle concentrations
26 on projected global climate change or consequent local and regional weather patterns would be
27 subject to considerable uncertainty (CD, p. 4-238).
28
August 2003 2-63 Draft - Do Not Quote or Cite
-------
1 2.9.2 PM Properties Affecting Transmission of Ultraviolet Radiation
2 As discussed in section 4.5.2 of the draft CD, the transmission of solar radiation in the
3 ultraviolet (UV) range through the earth's atmosphere is affected by ozone, clouds and particles.
4 Of particular interest is the effect of particles on radiation in the ultraviolet-B (UV-B) range
5 (generally from 0.280 to 0.320 jim), which has been associated with various biological effects.
6 Relative to ozone, the effects of ambient particles on the transmission of UV-B radiation are
7 more complex. The draft CD notes that even the sign of the effect can reverse as the
8 composition of the particle mix in an air mass changes from scattering to absorbing types (e.g.,
9 from sulfate to elemental carbon), and that there is an interaction in the radiative effects of
10 scattering particles and absorbing molecules, such as ozone, in the lower atmosphere (CD, p. 4-
11 223).
12 The effects of particles in the lower atmosphere on the transmission of solar UV-B
13 radiation have been examined both by field measurements and by radiative transfer model
14 calculations. The draft CD cites several studies that reinforce the idea that particles can play an
15 important role in modulating the attenuation of solar UV-B radiation, although none included
16 measurements of ambient PM concentrations, so that direct relationships between PM levels and
17 UV-B radiation transmission could not be determined (CD, pp. 4-223 to 4-225). The available
18 studies, conducted in diverse locations around the world, demonstrate that relationships between
19 particles and solar UV-B radiation transmission can vary considerably over location, conditions,
20 and time. While ambient particles are generally expected to decrease the flux of solar UV-B
21 radiation reaching the surface, any comprehensive assessment of the radiative effects of particles
22 would be location-specific and complicated by the role of particles in photochemical activity in
23 the lower atmosphere. Whether the photochemical production of ozone is enhanced, remains the
24 same, or reduced by the presence of ambient particles will be location-specific and dependent on
25 particle composition. Also complicating any assessment of solar UV-B radiation penetration to
26 specific areas of the earth's surface are the influences of clouds, which in turn are affected by the
27 presence of ambient particles (CD p. 4-226).
August 2003 2-64 Draft - Do Not Quote or Cite
-------
1 REFERENCES
2
3 Clayton, C. A.; Perritt, R. L.; Pellizzari, E. D.; Thomas, K. W.; Whitmore, R. W.; Wallace, L. A.; Ozkaynak, H.;
4 Spengler, J. D. (1993). Particle total exposure assessment methodology (PTEAM) study: distributions of
5 aerosol and elemental concentrations in personal, indoor, and outdoor air samples in a southern California
6 community. J. Exposure Anal. Environ. Epidemiol. 3: 227-250.
7
8 Desert Research Institute (2000). Watson, John G. and Judith C. Chow, "Reconciling Urban Fugitive Dust
9 Emissions Inventory and Ambient Source Contribution Estimates: Summary of Current Knowledge and
10 Needed Research," Document No. 6110.4F, Reno, NV, May, 2000. Available:
11 www.epa.gov/ttn/chief/efdocs/fugitivedust.pdf.
12
13 Environmental Protection Agency (1996a). Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
14 Office of Research and Development; report no. EPA/600/P-95/001aF-cF.3v
15
16 Environmental Protection Agency (1996b). Review of the National Ambient Air Quality Standards for Paniculate
17 Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research
18 Triangle Park, NC: Office of Air Quality Planning and Standards; report no. EPA-452/R-96-013.
19
20 Environmental Protection Agency (200Ib). National Air Quality and Emissions Trends Report, 1999. Research
21 Triangle Park, NC: Office of Air Quality Planning and Standards; report no. EPA/454/R-01-004. March,
22 2001. Available: www.epa.gov/oar/aqtrnd99.
23
24 Environmental Protection Agency (2002). Latest findings on national air quality: 2001 status and trends. Research
25 Triangle Park, NC: U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards;
26 report no. EPA-454/K-02-001.
27
28 Fehsenfeld, F.; D. Hastie; C. Chow; and P.A. Solomon. "Gas and Particle Measurements, Chapter 5." In NARSTO
29 Paniculate Matter Science Assessment. McMurray, P., Shepherd, M., and Vickery, J. eds. NARSTO, U.S.
30 Environmental Protection Agency, Research Triangle Park, NC. 2003.
31
32 Fitz-Simons, T.; Mathias, S.; Rizzo, M. (2000). U.S. EPA Memorandum to File. Subject: Analyses of 1999 PM
33 Data for the PMNAAQS Review. November 17, 2000. Available:
34 www.epa.gov/oar/oaqps/pm25/docs.html
35
36 Fitz-Simons, T.; Mathias, S. (2002). U.S. EPA Memorandum to File. Subject: Analyses of 1999-2001 PM Data for
37 the PM NAAQS Review. July 21, 2002. Available: www.epa.gov/oar/oaqps/pm25/docs.html
38
39 Guenther, A.; Geron, C.; Pierce, T.; Lamb, B.; Harley, P.; Fall, R. (2000) Natural emissions of non-methane volatile
40 organic compounds, carbon monoxide, and oxides of nitrogen from North America. Atmos. Environ. 34:
41 2205-2230.
42
43 Husar, R. B.; Schichtel, B. A.; Falke, S. R.; Li, F.; Wilson, W. E.; Pinto, J.; Malm, W. C.; Fox, D. G.; Feldman,
44 G. C.; McClain, C.; Kuring, N.; Holben, B. N.; Vermote, E. F.; Herman, J. R.; Elvidge, C. D. (2000). The
45 impact of the 1998 Central American smoke on the atmospheric environment of eastern North America.
46 J. Geophys. Res.: submitted.
47
48 Kim, S.; Shen S.; Sioutas, C.; Zhu, Y.; Hinds, W.C. (2002). Size distribution and diurnal and seasonal trends of
49 ultrafine particles in source and receptor sites of the Los Angeles basin. J. Air & Waste Manage. Assoc.
50 52:297-307.
51
52 Langstaff, J. (2003a) OAQPS Staff Memorandum to PMNAAQS Review Docket (OAR-2001-0017). Subject:
53 PM2 5 concentrations at IMPROVE sites with low annual mean concentrations. August 29, 2003.
August 2003 2-65 Draft - Do Not Quote or Cite
-------
2 Malm, W.C.; Sisler, J.F.; Huffman, D.; Eldred, R.; Cahill, T.A. (1994). Spatial and seasonal trends in particle
3 concentration and optical extinction in the United States. J. Geophys. Res. 29: 1347-1370.
4
5 Mage, D.; Wilson, W.; Hasselblad, V.; Grant, L. (1999). Assessment of human exposure to ambient paniculate
6 matter. J. Air Waste Manage. Assoc. 49: 174-185.
7
8 National Acid Precipitation Assessment Program (NAPAP), (1991). Office of the Director, Acid Deposition: State
9 of Science and Technology. Report 24, Visibility: Existing and Historical Conditions - Causes and Effects.
10 Washington, D.C.
11
12 National Research Council (1991). Human exposure assessment for airborne pollutants: advances and opportunities.
13 Washington, DC: National Academy of Sciences.
14
15 Nichols, M. D. (1996) Memorandum to EPA Air Division Directors regarding Areas Affected by PM-10 Natural
16 Events. May 30, 1996.
17
18 Noble, C.A.; S. Mukerjee; M. Gonzales; C.E. Rodes; P.A. Lawless; s. Natarajan; E.A. Myers; G.A. Norris; L. Smith;
19 H. Ozkaynak; L.M. Neas. (2003). Continuous measurement of fine and ultrafine paniculate matter, criteria
20 pollutants and meteorological conditions in urban El Paso, Texas. Atmos. Environ. 37: 827-840.
21
22 Ozkaynak, H.; Xue, J.; Spengler, J.; Wallace, L.; Pellizzari, E.; Jenkins, P. (1996a). Personal exposure to airborne
23 particles and metals: results from the particle TEAM study in Riverside, California. J. Exp. Anal. Environ.
24 Epidemiol. 6: 57-78.
25
26 Ozkaynak, H.; Xue, J.; Weker, R.; Bulter, D.; Koutrakis, P.; Spengler, J. (1996b). The particle TEAM (PTEAM)
27 study: analysis of the data: final report, volume III. Research Triangle Park, NC: U.S. Environmental
28 Protection Agency, Atmospheric Research and Exposure Assessment Laboratory; report no. EPA/600/R-
29 95/098. Available from: NTIS, Springfield, VA; PB97-102495.
30
31 Schmidt, M.; D. Mintz; V. Rao; L. McCluney (2003). U.S. EPA Memorandum to File. Subject: Analyses of 2000-
32 2002 PM Data for the PM NAAQS Review. August 29, 2003. Available:
33 www.epa.gov/oar/oaqps/pm25/docs.html.
34
35 Solomon, P.A.; M.P. Tolocka; G. Norris; and M. Landis. "Chemical Analysis Methods for Atmospheric Aerosol
36 Components." In Aerosol Measurement: Principles, Techniques, and Application, Second Edition, Eds. P.
37 Barron and K. Willeke. John Wiley & Sons, Inc., New York, NY. 2001.
38
39 Trijonis, J. (1982). Existing and natural background levels of visibility and fine particles in the rural East. Atmos.
40 Environ. 16:2431-2445.
41
42 Whitby, K. T. (1978). The physical characteristics of sulfur aerosols. Atmos. Environ. 12: 135-159.
43
44 Wilson, W. E.; Suh, H.H. (1997) Fine particles and coarse particles: concentration relationships relevant to
45 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
46
47 Wilson, W. E.; Mage, D. T.; Grant, L. D. (2000) Estimating separately personal exposure to ambient and
48 nonambient paniculate matter for epidemiology and risk assessment: why and how. J. Air Waste Manage.
49 Assoc. 50: 1167-1183.
50
51 Woo, K.S.; Chen, D.R.; Pui, D.Y.H.; McMurry, P.H. (2001). Measurement of Atlanta Aerosol Size Distributions:
52 Observation of Ultrafine Particle Events. Aerosol Science and Technology, 34:75-87.
53
August 2003 2-66 Draft - Do Not Quote or Cite
-------
1 Wotawa, G.; Trainer, M. (2000). The influence of Canadian forest fires on pollutant concentrations in the United
2 States. Science 288: 324-328.
August 2003 2-67 Draft - Do Not Quote or Cite
-------
-------
3. CHARACTERIZATION OF PM-RELATED HEALTH EFFECTS
1 3.1 INTRODUCTION
2 This chapter summarizes key information relevant to assessment of the known and
3 potential health effects associated with exposure to ambient PM, alone and in combination with
4 other pollutants that are routinely present in ambient air. A comprehensive discussion of this
5 information, focusing on the new scientific studies published since the last review, can be found
6 in Chapters 6 through 9 of the draft CD.
7 The presentation here organizes the key health effects information into those elements
8 essential for the evaluation of current and alternative standards for PM. Drawing primarily upon
9 the epidemiology, lexicological, dosimetry, and exposure-related studies in the draft CD, this
10 Chapter summarizes: (1) information and hypotheses regarding mechanisms by which particles
11 that penetrate to and deposit in various regions of the respiratory tract may exert effects; (2) the
12 nature of effects that have been associated with ambient PM, with a focus on fine- and coarse-
13 fraction PM; (3) the identification of sensitive populations that appear to be at greater risk of
14 effects of exposure to ambient PM; and (4) issues related to interpretation and evaluation of the
15 health effects evidence, including discussion of the role of co-pollutants, issues related to
16 epidemiological modeling, evidence for effects of various PM components, and the consistency
17 and coherence of the evidence. Preliminary staff conclusions and recommendations related to
18 primary standards for PM are incorporated into Chapter 6 of this draft Staff Paper.
19 In the last review, a variety of health effects had been associated with ambient PM at
20 concentrations extending from those elevated levels found in the historic London episodes down
21 to levels below the 1987 PM10 standards. Of special importance from the last review were the
22 conclusions that (1) ambient particles smaller than 10 |im that penetrate into the thoracic region
23 of the respiratory tract remain of greatest concern to health, (2) the fine and coarse fractions of
24 PM10 should be considered separately for the purposes of setting ambient air quality standards,
25 and (3) the consistency and coherence of the health effects evidence greatly adds to the strength
26 and plausibility of the observed PM associations. Important uncertainties remained, however,
27 such as issues related to interpreting the role of gaseous co-pollutants in PM associations with
August 2003 3-1 Draft - Do Not Cite or Quote
-------
1 health effects, and the lack of demonstrated biological mechanisms that could explain observed
2 effects.
3 A large number of new studies containing further evidence of serious health effects have
4 been published since the last review, with important new information coming from
5 epidemiologic, lexicological, controlled human exposure, and dosimetry studies. Examples of
6 important new epidemiologic studies include:
7 • New multi-city studies that use uniform methodologies to investigate the effects of PM
8 on health with data from multiple locations with varying climate and air pollution mixes,
9 contributing to increased understanding of the role of various potential confounders,
10 including gaseous co-pollutants, on observed PM associations. These studies can provide
11 more precise estimates of the magnitude of a PM effect than most smaller-scale
12 individual city studies.
13
14 • More studies of various health endpoints evaluating independent associations between
15 effects and fine- and coarse-fraction particles, as well as ultrafme particles or specific
16 components (e.g., sulfates, metals).
17
18 • Numerous new studies of cardiovascular endpoints, with particular emphasis on
19 assessment of cardiovascular risk factors as well as symptoms.
20
21 • New analyses and approaches to addressing issues related to potential confounding by
22 gaseous co-pollutants, possible thresholds for effects, and measurement error and
23 exposure misclassification.
24 • Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures
25 including PM components using factor analysis or source apportionment methods to link
26 effects with different PM source types (e.g., combustion, crustal1 sources).
27
28 • Several new "intervention studies" providing evidence for improvements in respiratory or
29 cardiovascular health with reductions in ambient concentrations of particles and gaseous
30 co-pollutants.
31
32 Important new lexicological, controlled human exposure, and dosimetry studies include, for
33 example:
34 • Animal and controlled human exposure studies using concentrated ambient particles
35 (CAPs), new indicators of response (e.g., heart rate variability), and animal models
36 representing sensitive subpopulations, that are relevant to the plausibility of the
1 "Crastal" is used here to describe particles of geologic origin, which can be found in both fine- and
coarse-fraction PM.
August 2003 3-2 Draft - Do Not Cite or Quote
-------
1 epidemiological evidence and provide insights into potential mechanisms for PM-related
2 effects.
3
4 • Dosimetry studies using new modeling methods and controlled exposures that provide
5 increased understanding of the dosimetry of different particle size classes and in
6 members of potentially sensitive subpopulations, such as people with chronic respiratory
7 disease.
8
9 • Studies relating population exposure to PM and other pollutants measured at centrally
10 located monitors to estimates of exposure to ambient pollutants at the individual level
11 have added new insights to our understanding of community health studies.
12
13 3.2 MECHANISMS
14 This section briefly summarizes available information concerning the penetration and
15 deposition of particles in the respiratory tract and outlines potential physiological and
16 pathological responses to PM, drawing from information presented in previous PM criteria and
17 standard reviews and in Chapters 6-9 of the draft CD.
18 The 1996 Staff Paper concluded that the available toxicological and clinical information
19 yields no demonstrated biological mechanism(s) that can explain the associations between
20 ambient PM exposure and mortality and morbidity reported in community epidemiologic studies
21 (EPA, 1996b, p. V-2). While that conclusion still holds true, substantial progress has been made
22 in identifying and understanding a number of potential pathways that were the subject of
23 speculation in the last review. The major purposes of the discussion presented here are to note
24 the available information of greatest relevance in identifying those fractions of PM that are most
25 likely to be of concern to health, to examine possible links between ambient particles and
26 reported respiratory and cardiovascular effects in humans, to identify factors that may contribute
27 to susceptibility in sensitive populations, and to focus attention on the advances in mechanistic
28 research that are providing evidence in support of a biological basis for causal links between
29 ambient PM exposures and reported health effects.
30 As discussed in the 1996 Staff Paper, an evaluation of the ways by which inhaled particles
31 might ultimately affect human health must take account of patterns of deposition and clearance
32 in the respiratory tract. The draft CD observes that the probability of any biological effect of PM
33 depends on particle deposition and retention, as well as underlying dose-response relationships
August 2003 3-3 Draft - Do Not Cite or Quote
-------
1 (CD, p. 6-42). The major elements of these considerations have been developed in previous
2 reviews and are summarized briefly here. The human respiratory tract can be divided into three
3 main regions: (1) extra-thoracic, (2) tracheobronchial, and (3) alveolar (CD, Figure 6-1). The
4 regions differ markedly in structure, function, size, mechanisms of deposition and removal, and
5 sensitivity or reactivity to deposited particles; overall, the concerns related to ambient particles
6 are greater for the two lower regions (EPA, 1996b).
7 The junction of conducting and respiratory airways appears to be a key anatomic focus;
8 many inhaled particles of critical size are deposited in the respiratory bronchioles that lie just
9 distal to this junction. Recent studies have indicated that ultrafine, as well as larger particles
10 show enhanced deposition of particles at airway bifurcations (CD, p. 6-20). The draft CD
11 summarizes simulations of deposition of ambient particle distributions that indicate fine- and
12 coarse-fraction particles are deposited in both the tracheobronchial and alveolar regions. Peak
13 deposition for ultrafine and coarse-fraction particles is in the upper part of the respiratory tract,
14 while peak deposition for accumulation-mode particles occurs in the alveolar regions (CD, p. 6-
15 23, Figure 6-8). The draft CD notes that the number dose (particles/cm2/day) of fine particles to
16 the lung is orders of magnitude higher than that for coarse-fraction particles; for the smallest
17 particles, ultrafine particles, there are even larger doses in terms of particle number.
18 The potential effects of deposited particles are influenced by the speed and nature of
19 removal. Clearance and translocation mechanisms include dissolution of particles or particle
20 constituents, mucociliary transport, or endocytosis by cells such as macrophages; the relative
21 importance of the different mechanisms varies with each of the three regions (CD, Table 6-2,
22 Figure 6-11). The presence of respiratory disease is a factor that can alter clearance of particles
23 (CD, p. 6-57). Recent studies have suggested that ultrafine particles may move directly from the
24 lungs into the systemic circulation, providing a pathway by which ambient PM exposure could
25 rapidly affect extrapulmonary organs (CD, p. 6-53).
26 Information from the last review, as well as important new studies discussed in the draft
27 CD, add to evidence from the earlier 1987 review, showing how breathing patterns and
28 respiratory disease status can affect regional particle deposition patterns. The 1996 CD showed
29 that as mouth-breathing or workload increases so does deposition in the bronchial and alveolar
30 regions. For those individuals considered to be mouth breathers, deposition increases for coarse-
August 2003 3-4 Draft - Do Not Cite or Quote
-------
1 fraction particles in the tracheobronchial region (EPA, 1996a, pp. 166-168). New evidence
2 indicates that people with chronic lung disease can have increased total lung deposition, and can
3 also show increases in local deposition ("hot spots") due to uneven airflow in diseased lungs
4 (CD, p. 6-103). In such a case, the respiratory condition can enhance sensitivity to inhaled
5 particles by increasing the delivered dose to sensitive regions. Such dosimetry studies are of
6 obvious relevance to identifying sensitive populations, which is discussed in section 3.4.
7 As discussed in the 1996 Staff Paper, evidence from epidemiologic studies of
8 occupational and historical community exposures and laboratory studies of animal and human
9 responses to simulated ambient particle components suggested that at exposures well above the
10 then-current PM standards, particles may produce physiological and ultimately pathological
11 effects by a variety of mechanisms. Previous criteria and standards reviews included an
12 integrated extensive examination of available literature on the potential mechanisms,
13 consequences, and observed responses to particle deposition organized according to major
14 regions of the respiratory tract (EPA, 1982b, 1996a,b). Evidence from dosimetry studies has
15 indicated that particles that deposit in the thoracic region (tracheobronchial and alveolar
16 regions), i.e., particles smaller than 10 |im diameter, were of greatest concern for standard setting
17 (EPA, 1996b, p. V-3, Figure V-l). Although more recent information has expanded our
18 understanding of these issues, no basis has emerged to change that fundamental conclusion.
19 In the last two reviews, staff identified a number of hypothesized mechanisms and
20 supporting observations by which common components of ambient particles that deposit in the
21 thoracic region, alone or in combination with pollutant gases, might produce health effects (EPA,
22 1982b, Table 5-2; 1996b, Table V-2). While there has been little doubt in the scientific
23 community that the historical London air pollution episodes had profound effects on daily
24 mortality and morbidity, no combination of the mechanisms and observations advanced in the
25 past reviews has been sufficiently tested or generally accepted as explaining the historical
26 community results. Moreover, the hypothesized mechanisms cited in those previous reviews
27 were based on insights developed from laboratory and occupational/community epidemiologic
28 studies that involved concentrations that were substantially higher than those observed in
29 contemporary U.S. atmospheres, and in many cases using laboratory-generated particles that
30 may be of limited relevance to community exposures (EPA, 1996b, p V-4).
August 2003 3-5 Draft - Do Not Cite or Quote
-------
1 Fully defining the mechanisms of action for PM would involve description of the
2 pathogenesis or origin and development of any related diseases or processes resulting in
3 premature mortality. While the substantial recent progress presented in Chapters 6, 7 and 9 of
4 the draft CD and summarized below has provided important insights that contribute to the
5 plausibility of community study results, this more ambitious goal of fully understanding
6 fundamental mechanisms has not yet been reached. Some of the more important findings
7 presented therein, including those related to the cardiovascular system, may be more accurately
8 described as intermediate responses potentially caused by PM exposure rather than complete
9 mechanisms. It appears unlikely that the complex mixes of particles that are present in
10 community air pollution would act alone through any single pathway of response. Accordingly,
11 it is plausible that several responses might occur in concert to produce reported health endpoints.
12 By way of illustration, Mauderly et al. (1998) discussed particle components or
13 characteristics hypothesized to contribute to PM health, producing an illustrative list of 11
14 components or characteristics of interest for which some evidence existed. The list included: 1)
15 PM mass concentration, 2) PM particle size/surface area, 3) ultrafme PM, 4) metals, 5) acids, 6)
16 organic compounds, 7) biogenic particles, 8) sulfate and nitrate salts, 9) peroxides, 10) soot, and
17 11) co-factors, including effects modification or confounding by co-occurring gases and
18 meteorology. The authors stress that this list is neither definitive nor exhaustive, and note that
19 "it is generally accepted as most likely that multiple toxic species act by several mechanistic
20 pathways to cause the range of health effects that have been observed" (Mauderly et al., 1998).
21 In assessing the more recent animal, controlled human, and epidemiologic information,
22 the draft CD developed a summary of current thinking on pathophysiological mechanisms for the
23 effects of low concentrations of PM (CD, pp. 7-120 to 7-127). The potential mechanisms
24 discussed in the draft CD, organized by effect category, are summarized in Table 3-1. The list of
25 potential mechanisms in Table 3-1 was developed mainly in reference to effects from short-term
26 rather than long-term exposure to PM. Repeated occurrences of some short-term insults, such as
27 inflammation, might contribute to long-term effects, but wholly different mechanisms might also
28 be important in the development of chronic responses. The draft CD highlights emerging
29 evidence for these effects and potential mechanisms, as summarized below. The relative support
30 for these potential mechanisms/intermediate effects and their relevance to real world inhalation
August 2003 3-6 Draft - Do Not Cite or Quote
-------
1 of ambient particles varies significantly. Moreover, the draft CD highlights the variability of
2 results that exist among different approaches, investigators, animal models, and even day-to-day
3 within studies. "Nevertheless, . . . much progress has been made since the 1996 PM AQCD in
4 evaluating pathophysiological mechanisms involved in PM-associated cardiovascular and
5 respiratory health effects" (CD, p. 7-120).
7
8
9
10
11
Table 3-1. Summary of Potential Mechanisms Based on Emerging Toxicological Evidence
(CD, pp. 7-120 to 7-127)
Effect
Direct Pulmonary Effects
Systemic Effects Secondary
to Lung Injury
Direct Effects on the Heart
Potential Mechanisms
Lung injury and inflammation
Increased susceptibility to respiratory infections
Increased airway reactivity and exacerbation of asthma
Lung injury leading to impairment of heart function by
lowering blood oxygen levels and increasing the work of
breathing
Lung inflamation and cytokine production leading to adverse
systemic hemodynamic effects (e.g., arrhythmia)
Lung inflammation leading to increased risk of heart attacks
and strokes due to increased blood coagulability
PM/lung interactions potentially affecting hematopoiesis (e.g.,
blood cell formation)
Uptake of particles and/or distribution of soluble components
from the lungs into the systemic circulation
Effects on the autonomic control of the heart and
cardiovascular system
12
13
14
15
16
17
18
19
20
Direct Pulmonary Effects. Substantial new toxicologic information summarized in the
draft CD provides evidence for the occurrence of direct pulmonary effects; while many of the
newer studies use high doses (in mg or hundreds of jig), some have used relatively low doses
that are close to ambient concentrations. Evidence that supports hypotheses on direct pulmonary
effects includes toxicological and controlled human exposure studies using both sources of
August 2003
5-7
Draft - Do Not Cite or Quote
-------
1 ambient particles and combustion-related particles. Toxicologic studies using intratracheal
2 instillation of ambient particles from various locations (e.g., St. Louis, MO; Washington DC;
3 Dusseldorf, Germany; Ottawa, Canada; Provo and Utah Valley, Utah; Edinburgh, Scotland) have
4 "shown clearly that PM can cause lung inflammation and injury" (CD, p. 7-16). Several studies
5 using filter extracts from Utah Valley ambient samples collected before, during and after the
6 shut-down of a major particle-emitting facility have reported effects such as increases in oxidant
7 generation, release of cytokines such as IL-8, and evidence of pulmonary injury such as
8 increased levels of lactose dehydrogenase (CD, p 7-16 to 7-18). Administration of residual oil
9 fly ash (ROFA, an example of a combustion source particle type) has been shown to produce
10 severe inflammation, with effects including recruitment of neutrophils, eosinophils and
11 monocytes into the airway (CD, p. 7-28). Several new toxicologic or controlled human exposure
12 studies using exposure to CAPs have shown evidence of inflammatory responses that appear to
13 be dependent on particle composition (CD, p. 7-21). In vitro studies, summarized in draft CD
14 Table 7-9, also provide support for the observed inflammatory effects of ambient PM and
15 constituent substances (CD, p. 7-68). Evidence from a limited number of toxicologic studies
16 suggests that exposure to particles can increase susceptibility to respiratory infections (CD, p. 7-
17 122). Moreover, a number of epidemiologic studies have linked daily changes in ambient PM
18 concentrations with increased risk of hospitalization for pneumonia or respiratory infections
19 (CD, Table 8-19). Some toxicologic evidence also indicates that PM can aggravate asthmatic
20 symptoms or increase airway reactivity, especially in studies of the effects of diesel exhaust
21 particles (CD, p. 7-122), while a number of epidemiologic studies have reported increased risk of
22 exacerbation of asthma with ambient PM exposure (CD, p. 8-174).
23 Systemic Effects Secondary to Lung Injury. Secondary systemic effects may also result
24 from lung injury or inflammation. While more limited than for direct pulmonary effects, some
25 new evidence from toxicologic studies suggests that injury or inflammation in the respiratory
26 system can lead to changes in heart rhythm, reduced oxygenation of the blood, changes in blood
27 cell counts, or changes in the blood that can increase the risk of blood clot formation, a risk
28 factor for heart attacks or strokes (CD, pp. 7-123 to 7-125).
29 Direct Effects on the Heart. Animal studies have also provided initial evidence that
30 particles can have direct cardiovascular effects. As shown in draft CD Figure 9-15, there are
August 2003 3-8 Draft - Do Not Cite or Quote
-------
1 several pathways by which particle deposition in the respiratory system could lead to
2 cardiovascular effects, such as PM-induced pulmonary reflexes resulting in changes in the
3 autonomic nervous system that then could affect heart rhythm (CD, p. 9-62). Also, inhaled PM
4 could affect the heart or other organs if particles or particle constituents are released into the
5 circulatory system from the lungs; some new evidence indicates that the smaller ultrafine
6 particles can move directly from the lungs into the systemic circulation (CD, p. 6-53). Recent
7 epidemiologic studies have begun to include more sensitive measures of cardiovascular
8 responses, linking ambient PM exposure with increased levels of fibrinogen or C-reactive
9 protein in the blood, and with changes in heart function such as decreased heart rate variability
10 or heart rhythm (CD, p. 9-103). The effects observed in epidemiologic studies could result either
11 from direct effects of PM on the cardiovascular system, or secondarily through changes in the
12 pulmonary system.
13 Summary. Dosimetric information shows that both fine- and coarse-fraction particles
14 (smaller than 10 |im) can penetrate and deposit in the tracheobronchial and alveolar regions of
15 the lung. Particles also may carry other harmful substances with them to these regions, with the
16 smaller particles having the greatest surface area available for such transport (CD, p. 9-50). New
17 studies have identified a variety of responses to constituents of ambient PM that may contribute
18 to the health effects reported in epidemiologic studies. The draft CD observes "A primary
19 causative attribute may not exist but rather many attributes may contribute to a complex
20 mechanism driven by the nature of a given PM and its contributing sources." (CD, p. 9-61).
21 Even though specific mechanisms or pathways for effects have not yet been demonstrated, the
22 increasing laboratory evidence of pathways by which particles could affect the respiratory and
23 cardiovascular systems adds to the plausibility of the conclusion that particles, alone or in
24 combination with other pollutants, are playing a causal role in the effects observed in
25 epidemiologic studies.
26
27 3.3 NATURE OF EFFECTS
28 The 1996 Staff Paper identified the following key health effects categories associated
29 with PM exposure (EPA, 1996b, pp V-8 and V-9):
30 • Increased mortality
August 2003 3-9 Draft - Do Not Cite or Quote
-------
1 • Indices of morbidity associated with respiratory and cardiovascular disease
2 • Hospital admissions and emergency room visits
3 • School absences
4 • Work loss days
5 • Restricted activity days
6 • Effects on lung function and symptoms
7 • Morphological changes
8 • Altered host defense mechanisms
9 Additional evidence is now available to identify the following new indices of morbidity:
10 • Physicians'office or clinic visits
11 • Cardiovascular health indicators, such as heart rate variability or C-reactive
12 protein levels
13 • Developmental effects, such as low birth weight, and infant mortality
14 In considering the nature of effects, it is important to note some key characteristics and
15 limitations of the kinds of studies — epidemiologic, toxicologic and controlled human exposure
16 studies - used to identify them.
17 The general strengths and weaknesses of epidemiologic studies were discussed in detail
18 in the 1996 CD (Chapter 12) and are briefly reviewed in section 8.1.3 of the draft CD.
19 Epidemiologic studies can identify associations between actual community-level air pollution
20 containing PM and population-level health effects, and can provide evidence useful in making
21 inferences with regard to the causality of such relationships, although they cannot alone be used
22 to demonstrate mechanisms of action. Epidemiologic studies can also provide information that
23 can help to identify sensitive populations particularly at risk for effects (summarized below in
24 section 3.4).
25 A central issue in the analysis of epidemiological evidence considered throughout the
26 discussion of effects in this section (and further in section 3.5) is the role of co-pollutants as
27 potential confounders or effect modifiers in associations between health effects and PM. In
28 addition, co-pollutants may act as indicators for fine particles derived from specific combustion
29 sources; for example, the Criteria Document for CO concluded that ambient CO may be a
30 surrogate for air pollution from combustion sources (EPA, 2000a). Confounding occurs when a
August 2003 3-10 Draft - Do Not Cite or Quote
-------
1 health effect that is caused by one risk factor is attributed to another variable that is correlated
2 with the causal risk factor; epidemiological analyses attempt to adjust or control for potential
3 confounders. A gaseous copollutant (e.g., O3, CO, SO2 and NO2) meets the criteria for potential
4 confounding in PM-health associations if: (1) it is a potential risk factor for the health effect
5 under study; (2) it is correlated with PM; and (3) it does not act as an intermediate step in the
6 pathway between PM exposure and the health effect under study (CD, p. 8-8). Effect modifiers
7 include variables that may influence the health response to the pollutant exposure (e.g., co-
8 pollutants, individual susceptibility, smoking or age); epidemiological analyses do not attempt to
9 control for effect modifiers, but rather to identify and assess the level of effect modification (CD,
10 p. 8-10).
11 The potential for confounding by co-pollutants is an important question, and throughout
12 the sections that follow, results of epidemiologic studies that consider both PM alone and PM
13 with co-pollutants will be discussed. Testing for confounding using multi-pollutant models has
14 been the most commonly used method to assess potential confounding. As discussed in the draft
15 CD section 8.4.3.3, there are statistical issues to be considered with multi-pollutant models, such
16 as possibly creating mis-fitting models by forcing all pollutants to fit the same lag structure, by
17 adding correlated but non-causal variables, or by omitting important variables. However, where
18 pollutant concentrations are not highly correlated, effect estimates from multi-pollutant models
19 could be more sound than those from single pollutant models (CD, p. 9-69). In addition, it is
20 possible that pollutants may act together, or that the effects of a single pollutant may be mediated
21 by other components of an ambient pollution mix. As stated in the draft CD, "Specific
22 attribution of effects to any single pollutant may therefore be overly simplistic" (CD, p. 9-69).
23 Several other key issues and considerations related to evaluating epidemiologic studies
24 are also discussed below in section 3.5. For example, issues related to epidemiologic modeling
25 are discussed, including statistical model specification and methods used to adjust for other
26 variables, including weather variables, the lag period between PM exposure and associated
27 effects, and the influence of measurement error on the results of epidemiologic studies. Other
28 considerations related to the evaluation of PM-related health effects evidence discussed in
29 section 3.5 includes the role of various components within the fine and coarse fractions of PM in
30 the different health endpoints that have been associated with PM and the consistency and
August 2003 3-11 Draft - Do Not Cite or Quote
-------
1 coherence of the health effects evidence.
2 Animal toxicologic, controlled human exposure, and dosimetry studies can provide
3 important support to epidemiologic studies and can help elucidate biological mechanisms that
4 explain observed effects (discussed above in section 3.2). Such studies can also provide
5 important information on risk factors for individual or population susceptibility to effects and on
6 characteristics of particles (e.g., constituents and subclasses) that may play key roles in the
7 production of health effects. However, as discussed in more detail in Chapter 8 of the draft CD,
8 the doses used in animal studies are generally much higher than community-level concentrations,
9 and important differences in dosimetry can exist across species. While animal models are
10 intended to mimic aspects of human disease, the results of animal studies may not directly reflect
11 human health responses. Further, controlled human exposure studies can only address the least
12 severe health endpoints, for obvious ethical reasons, and are limited in their ability to examine
13 health effects in the most sensitive populations. The need remains to link effects observed in
14 such studies under simulated exposure conditions (e.g., with regard to chemical composition,
15 particle size, and concentration) to those that would likely occur in real-world environments.
16 Recognizing the different strengths and limitations of these various kinds of studies, key
17 evidence illustrating these major PM effects categories is outlined below, with an emphasis on
18 the most recent information. Mortality effects are discussed in section 3.3.1, with discussion of
19 indices of morbidity in section 3.3.2, organized into four general categories: increased hospital
20 admissions and emergency room visits for both cardiovascular and respiratory diseases, effects
21 on the respiratory system, effects on the cardiovascular system, and developmental effects.
22
23 3.3.1 Premature Mortality
24 This section discusses (1) mortality associations with short-term PM exposure, with
25 emphasis on results from newly available multi-city analyses; (2) mortality associations with
26 long-term PM exposure; and (3) issues related to interpreting the results of mortality studies,
27 including mortality displacement and life shortening.
August 2003 3-12 Draft - Do Not Cite or Quote
-------
1 3.3.1.1 Mortality and Short-term PM Exposure
2 Historical reports of dramatic pollution episodes have provided clear evidence of
3 mortality associated with high levels of PM and other pollutants, as summarized in the 1996 CD
4 (EPA, 1996a, pp. 12-28 to 12-31) and Staff Paper (EPA, 1996b, p. V-l 1). More recently,
5 associations between increased daily mortality and PM have been reported at much lower PM
6 concentrations in a large number of areas with differing climates, PM composition, and levels of
7 gaseous co-pollutants. The 1996 CD summarized about 35 time-series mortality studies using
8 various PM indicators; the majority of these studies reported positive, statistically significant2
9 associations for PM10, as well as for PM25 and other indicators of fine-fraction particles (e.g.,
10 sulfates and H+). Significant associations were reported for total mortality3 for PM10 and
11 indicators of fine-fraction particles (EPA, 1996b, Tables V-3, V-l 1, V-l2) and cause-specific
12 mortality (i.e., respiratory- and cardiovascular-related mortality) in the general population and in
13 the elderly for PM10 (EPA, 1996b, Table V-4). In the 1996 CD, one daily mortality study
14 addressed coarse-fraction particles (PM10_25), reporting no statistically significant association
15 across the six cities included in the study, although a significant association was reported in one
16 of the six cities (EPA, 1996b, Table V-l 4).
17 In the previous PM NAAQS review, much consideration was given to the effects of PM
18 and co-pollutants, acting alone and in combination, in the associations with adverse health
19 effects reported in epidemiologic studies. The 1996 CD evaluated the findings of studies that
20 used single- and multiple-pollutant models to assess the potential for copollutant confounding
21 and effects modification. In some studies, PM effect estimate sizes were relatively unchanged
22 when gaseous pollutants were included in the models, and where the estimate was reduced, it
23 typically remained statistically significant (EPA, 1996a, p. 13-57). Much attention was focused
24 on a series of analyses and reanalyses using data from one U.S. city, Philadelphia, the most
25 comprehensive of which was a study funded by the Health Effects Institute (HEI). This study
Unless otherwise noted, statistically significant results are reported at a 95% confidence level.
3In these discussions, "total" mortality represents mortality from all causes excluding accidents and
suicides, as the term is typically used in epidemiologic studies on mortality and air pollution.
August 2003 3-13 Draft - Do Not Cite or Quote
-------
1 reported associations between mortality and TSP and other pollutants, concluding that it was
2 difficult to distinguish the effects of TSP from one or more gaseous co-pollutants for this single
3 location due in part to the fact that the co-pollutants were generally correlated with TSP. Indeed,
4 the limitations of even the most comprehensive single-city analyses precluded definitive
5 conclusions concerning the role of PM. For this reason, both the 1996 CD and Staff Paper
6 examined the consistency and coherence of effects across studies of individual cities having
7 different pollutant mixtures, climate, and other factors. Based on the consistent positive
8 associations found in such multiple studies, the draft CD concluded that PM effects were not
9 sensitive to other pollutants and the "findings regarding the PM effects are valid" (EPA, 1996a, p
10 13-57, EPA, 1996b, p V-56).
11 Taking into account these findings, the summary report from the HEI review panel
12 recommended that future research into the role of co-pollutants should improve upon the
13 examination of multiple single-city studies by different investigators by conducting multi-city
14 studies, using consistent analytical approaches across cities. The HEI review panel observed that
15 "[Consistent and repeated observations in locales with different air pollution profiles can provide
16 the most convincing epidemiologic evidence to support generalizing the findings from these
17 models" (HEI, 1997, p. 38). In fact, the HEI-sponsored study of PM10 and mortality in 90 U.S.
18 cities was designed to use such a multi-city approach (Samet et al., 2000, p. 1).
19 Since the last review, more than 80 new time-series studies of the relationship between
20 PM and mortality have been published (CD, p. 8-23), including several multi-city studies that are
21 responsive to the recommendations from the last review. In considering the body of evidence
22 on associations between PM and mortality in this standards review, the multi-city studies are of
23 particular relevance. The multi-city studies combine data from a number of cities that may vary
24 in climate, air pollutant sources or concentrations, and other potential risk factors. The
25 advantages of multi-city analyses include: (1) evaluation of associations in larger data sets can
26 provide more precise effect estimates than pooling results from separate studies; (2) consistency
27 in data handling and model specification can eliminate variation due to study design; (3) effect
28 modification or confounding by co-pollutants can be evaluated by combining data from areas
29 with differing air pollutant combinations; (4) regional or geographical variation in effects can be
30 evaluated; and (5) "publication bias" or exclusion of reporting of negative or nonsignificant
August 2003 3-14 Draft - Do Not Cite or Quote
-------
1 findings can be avoided (CD, p. 8-30). While considering results from the full body of studies,
2 staff will place emphasis on the results of multi-city studies in the following discussions.
3 Regarding the full body of multi-city and single-city studies, the draft CD notes that with
4 only a few exceptions, these newly reported associations are generally positive, many are
5 statistically significant (using both single- and multi-pollutant models), and the reported effect
6 estimates are generally consistent with the range of estimates from the last review (CD, p. 8-29).
7 Consistent with Table 9-10 in the draft CD (p. 9-91), the Table in Appendix A summarizes daily
8 mortality effect estimates for increments of PM10, PM2 5, and PM10_25 from all available multi-city
9 and single-city U.S. and Canadian studies4 as a consolidated reference for the following
10 discussion of associations between daily PM and increased total and cause-specific mortality. A
11 number of the new time-series epidemiologic studies have used generalized additive models
12 (GAM) in their analyses, and issues have been found with the convergence criteria and the
13 method for determining standard errors when using GAM, as discussed more fully in section
14 3.5.2.1. In Appendix A and the figures presented in this Chapter, results will be presented from
15 studies that have been reanalyzed to address issues related to GAM use, or where GAM were not
16 used in analysis; results from the full body of studies have been included in the assessment in the
17 CD and are described in the appendices to Chapter 8 of the draft CD.
4 Findings of U.S. and Canadian studies are more directly applicable for the review of the PM NAAQS,
though all study results are considered in the overall review of new scientific information. For consideration of
consistency in effects across studies, the results presented in figures that follow, and the effect estimates summarized
in Appendix A are from single-pollutant models, though results of multi-pollutant models are discussed in the text.
August 2003 3-15 Draft - Do Not Cite or Quote
-------
1 3.3.1.1.1 Multi-city Studies of Total Daily Mortality
2 The results from the U.S. and Canadian multi-city studies are summarized in Table 3-2.
3 In the previous review, a single multi-city study evaluated associations between daily mortality
4 and PM, including fine- and coarse-fraction particle measurements in six U.S. cities (Schwartz et
5 al., 1996, reanalyzed in Schwartz, 2003a). Reanalyses using the Six City data have reported
6 results consistent with the findings reported by Schwartz and colleagues (Klemm et al., 2000,
7 reanalyzed in Klemm et al., 2003). As shown in Table 3-2, significant increases in total
8 mortality of 2 to over 3% were reported per 25 |ig/m3 of PM25 and 50 |ig/m3 of PM10 (using more
9 stringent GAM and GLM)5 while PM10_2 5 was not significantly associated with mortality. The
10 role of gaseous co-pollutants was not directly addressed in these analyses.
11 Several recent multi-city analyses provide valuable new insights on associations between
12 PM and mortality, including more direct evaluation of the role of co-pollutants in PM-mortality
13 associations through the use of multi-pollutant modeling.
14 The National Morbidity, Mortality and Air Pollution Study (NMMAPS) included
15 analyses of PM10 effects on mortality in 90 U.S. cities, with additional, more detailed, analyses
16 being conducted in a subset of the 20 largest U.S. cities (discussed below in sections on cause-
17 specific mortality and morbidity) (Samet et al., 2000a,b,c; reanalyzed in Dominici et al., 2003).
18 A uniform methodology was used to evaluate the relationship between mortality and PM10 for
19 the different cities, and the results were synthesized to provide a combined estimate of effects
20 across the cities. These analyses are "marked by extremely sophisticated approaches addressing
21 issues of measurement error biases, copollutant evaluations, regional spatial correlation, and
22 synthesis of results from multiple cities by hierarchical Bayesian meta-regressions and meta-
23 analyses" (CD, p. 8-30). Based on the results shown in Figure 3-1, the overall risk estimate for
24 all cities6 is a statistically significant increase of 1.4% (using more stringent GAM) or 1.1%
25 (using GLM) in total mortality per 50 |ig/m3 increase in PM10 lagged one day (Samet et al.,
26 2000a,b, reanalyzed in
5GAM and GLM refer to generalized additive models and generalized linear models, respectively, and are
further discussed in section 3.5.2.1 on model specification.
6Note that Figure 3-1 includes results for 88 cities in the continental U.S.; Anchorage, AK and Honolulu,
HI are not included.
August 2003 3-16 Draft - Do Not Cite or Quote
-------
Table 3-2. Results of U.S. and Canadian Multi-city Studies on Associations
between Short-term PM Exposure and Mortality
Study
Six U.S. Cities
Schwartz et al.,
1996, re analysis in
Schwartz, 2003a
Six U.S. Cities
(reanalysis)
Klemm et al.,
2003
90 U.S. Cities
Sametetal.,
2000a,b;
reanalysis in
Domenici, 2003
10 U.S. Cities
Schwartz et al.,
2000, reanalysis
Schwartz 2003b
8 Canadian Cities
Burnett et al.,
2000, reanalysis
Burnett and
GnlHhem 9003
% Increase in Total Mortality per:
50 ug/m3 PM15/10
(*)
3.5(2.0,5.1)**
2.0(0.3,3.8)***
1.4(0.9, 1.9)**
1.1(0.5, 1.7)***
3.3(2.6,4.1)**
2.8 (2.0, 3.6)***
3.2(1.1,5.5)**
2.7 (-0.1, 5. 5)***
25 ug/m3 PM2 5 25 ug/m3 PM10.2 5
3.5(2.5,4.5)** (*)
3.3(2.2,4.3)***
3.0(2.0,4.0)** 0.8 (-0.6, 2.1)**
2.0 (0.9, 3.2)*** 0.3 (-1.2, 1.8)***
2.8(1.2,4.4)** 1.9 (-0.1, 3.9)**
2.1(0.1,4.2)*** 1.8 (-0.6, 4.4)***
Range of City
PM Mean Levels
(ug/m3)
PM1017.8-45.6
PM25 11.2-29.6
PM10_25 6.6-16.1
PM15/10 medians 14.4-30.3
PM2 5 medians 9. 0-23.1
PM10_2.5 medians 5.0-13.0
PM10 15.3-52.0
PM10 27. 1-40.6
PM10 20.4-3 1.0
PM2 5 9.5-17.7
PM10.2.5 8.9-16.8
Studies in italics were available for review in the 1996 PM CD.
* reanalysis results not available
** results from GAM using more stringent convergence criteria
*** results from GLM using natural splines
1
2 Domenici, 2003). Further, PM10 was also positively associated with mortality at 0-day and 2-day
3 lags (CD Figure 8-5, p. 8-34).
4 One key objective of the NMMAPS analyses was to characterize the effects of PM10 and
5 each of the gaseous co-pollutants, alone and in combination. An important result of this
6 assessment is the finding from the 90-city analysis of the "weak influence of gaseous co-
7 pollutants on the PM10 effect size estimates" (CD, Figure 8-6, p. 8-35). Less consistent
8 associations were found for the gaseous pollutants than for PM10, though in the overall analyses,
9 weak associations were found with SO2, NO2 and CO, and there was a significant association
10 with ozone in the summer months (CD, p. 8-35). The authors concluded that this "suggests that
August 2003
3-17
Draft - Do Not Cite or Quote
-------
1 the effect of PM10 is robust to the inclusion of other pollutants." (Samet et al., 2000b, p. 19); this
2 conclusion was affirmed in reanalysis (Domenici, 2003, p. 18). Potential confounding of PM
3 effects by gaseous pollutants is discussed further in section 3.5.1.
4 As seen in Figure 3-1, effect estimates for many individual cities exhibit wide confidence
5 ranges; the estimates vary and some are negative, though not statistically significant, suggesting
6 potentially more heterogeneity in effect estimates across cities than had been seen with single-
7 city studies. However, the NMMAPS investigators "caution against attempts to interpret
8 estimates for any specific city" (Samet et al., 2000b, p. 43). For many single-city studies, study
9 locations were selected based on the population size and availability of PM measurements. In
10 contrast, NMMAPS, while advantageous in including data from many different locations with
11 different climates and pollutant mixes, included many locations for which the sample size (i.e.,
12 population size and PM10 data) was inherently smaller for a given study period. This can result
13 in reduced precision for estimated effect estimates for the individual cities. As observed in the
14 draft CD:
15 ... increasing scientific weight should be accorded to [studies] in proportion to the
16 precision of their estimate of a health effect. Small studies and studies with an
17 inadequate exposure gradient generally produce less precise estimates than large studies
18 with an adequate exposure gradient. Therefore, the range of exposures (e.g., as indicated
19 by the IQR), the size of the study as indexed by the total number of observations (e.g.,
20 days) and total number of events (i.e., total deaths), and the inverse variance for the
21 principal effect estimates are all important indices useful in determining the likely
22 precision of health effect estimates and in according relative scientific weight to the
23 findings of a given study (CD, p. 8-29).
24 This is illustrated in Figure 3-2, where the individual city effect estimates are plotted
25 against the natural log of mortality-days (a product of each city's daily mortality rate and the
26 number of days for which PM data were available) as an indicator of the precision of the analysis
27 of each individual city. The use of the mortality-day product combines two of these indicators of
28 precision, and it can be seen that as the mortality-day product increases, the effect estimates are
29 generally more precise with narrower confidence intervals. In developing the combined 90-city
30 PM10-mortality effect estimate, the results for individual cities were weighted based on the
31 variance of the city effect estimates, which is another indicator for estimate precision. Thus,
32 staff has placed primary emphasis on the results of the overall multi-city effect estimates.
August 2003 3-18 Draft - Do Not Cite or Quote
-------
1 Also shown in Figure 3-1 are the results based on a grouping of these cities into the same
2 seven U.S. regions depicted in Figure 2-3. Samet et al. (2000a,b, reanalyzed in Domenici, 2003)
3 report that some variability in effects can be seen across cities and between regions. Combined
4 effect estimates for each of the seven U.S. regions varied somewhat, with generally higher
5 effects reported in the Northeast States and in Southern California. Data on some county-
6 specific variables (e.g., mean household income, percent of people not graduating from high
7 school, percent of people using public transportation) were included in analyses to investigate
8 regional differences, but the investigators did not identify any factors that resulted in significant
9 differences in effect across regions (CD, p. 8-35).
10 Schwartz (2000, reanalyzed in Schwartz, 2003b) conducted a series of multi-city
11 analyses using data from 10 U.S. cities where every-day PM monitoring data were available (in
12 many areas, PM is monitored on a l-in-3 or l-in-6 day basis). Using inverse variance weighting
13 methods to combine results across cities, a statistically significant association was reported
14 between PM10 and total mortality, with an effect estimate of an increase of 3.3% per 50 |ig/m3
15 PM10 (using GAM) or 2.8% per 50 |ig/m3 PM10 (using GLM), and effect estimate sizes were the
16 same in summer and winter (CD, p. 8-37).
17 In a combined analysis of data for the 8 largest Canadian cities, Burnett et al. (2000,
18 reanalyzed in Burnett and Goldberg, 2003) reported that mortality was significantly associated
19 with both PM2 5 and PM10, but not PM10_25. Overall effect estimates for increased total mortality
20 of appproximately 2-3% (using more stringent GAM and GLM) were reported per 25 |ig/m3 and
21 50 |ig/m3 increases in PM2 5 and PM10, respectively. Additional analyses were conducted using
22 PM25 components, including sulfates and a number of metals, and these results are discussed
23 further in section 3.5.3. The Canadian 8-city study also showed that the associations between
24 mortality and PM2 5 and PM10 generally remained significant in a number of analyses when
25 gaseous co-pollutants and 0- and 1-day lags were included in the models, although in a few
26 instances the effect estimates were reduced and lost statistical significance. The authors
27 conclude that mortality is associated with both PM and gaseous pollutants (Burnett et al., 2000);
28 copollutant effects and multi-pollutant models were not included in the reanalysis of this study's
29 results.
August 2003 3-19 Draft - Do Not Cite or Quote
-------
% Increase in Mortality per 10 Unit Increase in
-4-202
Me
Overall
Overall
Overall
Wichita
Overall
Madison
San Jose
El Paso —
_____
'-.an LJ'Cgo
Minneapolis/ St Paul ©
L
Chicago-
Cleveland 6-
^ ^
Le
Overall
Norfolk
Arlington
Overall
Little Rock
i -^r^r,, , /-~--.h ,
JUI _iLj' L-ILji
n
i— |-|/,-f1ni'illi"1
._ JU.jUll /IIIU
/•\
^
Jackson
j/ , ,j|| ^ _
uno/wine u
~
Overall
Overall
1 1
-e Seattle
0 f
^.acramenro Northwest
u idcuind
n Colorado Sonnas
<-/
-e — Spokane
••-
^ nt-, .-,.-,.-,;..,
* — ' »4
„ J OOUinWCut
p „ pL - .-
uorpuj unrioti
*-
9— Los Angeles
n ^'int-i Ann ' Annhpim ^nuthprn
j -^diuj And ; Andiieuii \>oucn6rn
i? n 1 f IJ
e,er->,
U r-,al 1-jClvj ^-.ILy (.UCr/p JaCI^On, MldnCj Uppct
— Des Moines Midwest
0 0 Topeka
•-
-e — Detroit
e — Pittsburgh
° Cincinnati Industrial
|— >• 1 Af
i on v y3/n_
o Evan svi lie
-m-
e New York
n Mr-i'f-irl-1
i-,
Northeast
' ' ^* luyncy
-T o, -
u ^yracuoe
0 Richmond
Q Kingston
— | —
-e Houston
^.ridiiuu
/-, D- • u
o tfirmmgnam Southea-t
i jrjjn •> mo
"-' Mn\v rirlnnn-
-, r^.~^_^ n^.
L/iucn jDoro
n Shreveport
•
_ o Huntsville
-»-
Figure 3-1. Estimated excess risks for PM mortality (1 day lag) for the 88 largest U.S.
cities as shown in the revised NMMAPS analysis (CD Figure 8-3)
Source: Dominici et al. (2002; 2003).
August 2003
3-20
Draft - Do Not Cite or Quote
-------
1
2
11 -
9 -
7-
e
S~ 5
0_
"E
O)
o
en
a:
I
o
3-
0)
Q. 0-
-1 -
-7 -
-9-
-11
4
^~Q i ± *
7 8 9 10 11
Natural Log of Mortality - Days
i
12
13
Figure 3-2. An EPA-derived plot showing relationship of PM10 total mortality effects
estimates and 95% confidence intervals for all cities in the Samet et al.
(2000a; Domenici, 2003) NMMAPS 90-cities analyses in relation to study size
(i.e., the natural logarithm or numbers of deaths times days of PM
observations). The dashed line depicts the overall nationwide effect estimate
(grand mean) of approximately 0.28% per 10 ug/m3 PM10 for models with no
co-pollutants (from draft CD Figure 8-21).
August 2003
3-21
Draft - Do Not Cite or Quote
-------
1 In addition, a European multi-city study, Air Pollution and Health: A European Project
2 (APHEA), has resulted in a series of analyses that are summarized in the draft CD (pp. 8-34 to 8-
3 37). This study included data from 12 European countries in the initial analyses, and reported
4 effect estimates for western cities of an approximately 2% increase in mortality per 50 |ig/m3
5 PM10 (CD, p. 8-44). Further analyses were done in a second series of studies (APHEA2) that
6 included data from 29 European cities, where an overall effect estimate of a 3.0% increase in risk
7 of mortality per 50 |ig/m3 PM10 was reported; reanalysis produced essentially identical results
8 (CD, p. 8-47). Using this larger set of cities, little evidence for potential confounding by gaseous
9 pollutants was reported, except for NO2, which the APHEA investigators considered an indicator
10 of high influence from traffic emissions (CD, p. 8-47). Although the APHEA studies used
11 consistent analytical methodologies, the PM measurement methods varied between cities,
12 including TSP, BS, PM13, and PM10, thus making quantitative comparisons with U.S. and
13 Canadian findings more difficult; however, these PM-mortality effect estimates are generally
14 consistent with the U.S. and Canadian results summarized in Table 3-2.
15 Analysis of data from the initial 12 APHEA countries was done to investigate factors that
16 may influence the heterogeneity found between western and eastern European cities. In
17 reanalysis of these data to address GAM-related modeling questions, the contrast between
18 western and eastern cities was less clear in the models using more stringent convergence criteria
19 or alternative models, suggesting that the apparent heterogeneity could be sensitive to model
20 specification (CD, p. 8-46). In further investigation of potential heterogeneity across the 29
21 APHEA cities, the city-specific variables that were found to influence the PM10-mortality
22 associations included climate (larger effects with warmer climates), health status of the
23 population (smaller effects with higher baseline mortality rates) and pollution mix (larger effects
24 with increasing average NO2 concentration). The investigators suggested that the relationship
25 found with climate may be due to more open windows thus greater exposure to ambient air in
26 warmer areas, and the relationship with NO2 concentration indicates the importance of traffic as
27 a PM source in the association with mortality (CD, p. 8-47).
28 In summary, the draft CD concludes that the findings of the Harvard Six-Cities study that
29 were available during the previous review have been confirmed by new analyses, and powerful
30 new multi-city analyses have provided important new evidence showing associations between
August 2003 3-22 Draft - Do Not Cite or Quote
-------
1 daily mortality and changes in PM10 and PM2 5 alone and in combination with gaseous co-
2 pollutants routinely present in the ambient air. Effect estimates from the multi-city studies range
3 from approximately 1.0 to 3.5% per 50 |ig/m3 increase in PM10 and from 2 to over 3% per 25
4 Hg/m3 increase in PM2 5 (CD, p. 8-50). The draft CD notes that the combined daily mortality
5 estimates from these multi-city studies are all consistent with the range of PM10 effect estimates
6 reported in the last review (CD, p. 8-50), with the 90-city NMMAPS estimate toward the lower
7 end of the range. Further, similarly sized effect estimates are reported between total mortality
8 and PM10 and PM2 5 but no significant associations are reported with PM10_2 5.
9 3.3.1.1.2 Other Studies of Total Daily Mortality
10 Numerous studies have been conducted in single cities or locations in the U.S. or Canada
11 (summary of results in Appendix A), as well as locations in Europe, Mexico City, South
12 America, Asia or Australia (summary of results in Table 8A-1 of the draft CD). As was
13 observed based on the more limited studies available in the last review, the associations reported
14 in the recent studies on PM10 and mortality are largely positive, and frequently statistically
15 significant. Similarly, a number of new studies also provide evidence of positive associations
16 between mortality and both PM25 and/or PM10_25, though the associations with PM10_25 less often
17 reached statistical significance. While some of the studies conducted in Europe, Mexico or
18 South America use gravimetric PM measurements (e.g., PM10, PM25, PM10_25), many of the non-
19 North American studies use PM indicators such as TSP, BS or COH, and the Australian studies
20 used nephelometric measures of PM. As summarized in Appendix 8A-1 of the draft CD, these
21 studies also show largely positive, significant associations between PM and mortality. While
22 effect estimates for different PM indicators may not be quantitatively comparable, the results
23 from all of these studies taken together show qualitative consistency in generally finding
24 significant associations between changes in PM and daily mortality.
25 From the full body of new studies reviewed and summarized in the draft CD, staff have
26 focused on the results of U.S. and Canadian studies for PM10, PM25, and PM10_25. The results of
27 single-city studies are combined with those from the multi-city studies discussed above, and
28 shown in Figures 3-4, 3-5 and 3-6, respectively. Using the same approach taken in the draft CD
29 in presenting the NMMAPS results (Figure 3-2 above), the effect estimates are plotted against
30 the natural log of the mortality-days product, again serving as an indicator of study precision. As
August 2003 3-23 Draft - Do Not Cite or Quote
-------
1 was seen in Figure 3-2, there is the expected tendency of studies to have more precise effect
2 estimates as the mortality-days product increases. Effect estimates from single-pollutant models
3 are presented in these figures for purposes of comparing results across studies; results of multi-
4 pollutant models are briefly discussed below and in section 3.5.1, and discussed more fully in
5 Appendix 8A-1 of the draft CD. Each figure includes effect estimates for both new studies and
6 studies included in the 1996 CD and, for comparison purposes, the range of statistically
7 significant effect estimates from the previous review is indicated in the right-hand margin. Also,
8 for the purpose of comparing effect estimates across the studies, where reanalysis was done, the
9 results from GAM analyses using more stringent criteria are presented. It should be recognized
10 that the standard errors may be larger in models using GLM, but GAM analyses may be more
11 appropriate for model fitting and effect estimate determination (CD, p. 8-196); staff notes that
12 there was little apparent difference seen when comparing figures using GAM stringent or GLM
13 results. Effect estimates for total, cardiovascular and respiratory mortality are included to give
14 an overview of the entire body of mortality studies, though cause-specific findings will be
15 discussed further in the next section. Staff notes that, as a subset of the full body of evidence
16 summarized in the draft CD, the range of effect estimates seen in these figures may differ from
17 the effect estimate ranges presented in the draft CD.
18 A number of new single-city analyses have included multi-pollutant modeling for
19 evaluating effects of PM and co-pollutants. As was found in the previous review, some of these
20 analyses report that PM effect sizes are little affected by the inclusion of copollutant gases in the
21 models, while others report potential confounding by one or more co-pollutants. In U.S. studies
22 conducted in Coachella Valley and Santa Clara County, California and Detroit, Michigan,
23 investigators concluded that generally positive associations (both significant and non-significant)
24 between PM and mortality were relatively unchanged in multi-pollutant models (Ostro et al.,
25 1999, 2000, reanalyzed in Ostro et al., 2003; Fairley, 1999, reanalyzed in Fairley, 2003;
26 Lippmann et al., 2000, reanalyzed in Ito, 2003). On the other hand, Moolgavkar (2000a,
27 reanalyzed in Moolgavkar, 2003) reported that the inclusion of gaseous co-pollutants resulted in
28 large reductions in PM effect estimates in Cook, Los Angeles, and Maricopa Counties, with
29 general loss of statistical significance.
August 2003 3-24 Draft - Do Not Cite or Quote
-------
1 The CD finds "nearly all of the newly reported analyses with a few exceptions continue
2 to show statistically significant associations between short-term (24 h) PM exposures indexed by
3 a variety of ambient PM measurements and increases in daily mortality in numerous U.S. and
4 Canadian cities, as well as elsewhere around the world" (CD, p. 8-29). For studies conducted in
5 the U.S. and Canada, Figure 3-4 indicates that nearly all of the associations between PM10 and
6 total mortality are positive and many reach statistical significance, especially among the more
7 precise effect estimates. Considering the results from all studies, the draft CD reports that effect
8 estimates for associations between PM10 and total mortality range from 1.5 to 8.5% per 50 |ig/m3
9 PM10.
10 Studies including both PM25 and PM10_25 are summarized in Table 8-2 of the draft CD (p.
11 8-52) and the effect estimates are presented in Figure 8-7 (CD, p. 8-55); studies from the U.S. as
12 well as other countries are included. The draft CD observes, "All of the studies found positive
13 associations between both the fine and coarse PM indices and increased mortality risk" (CD, p.
14 8-54). As can be seen in Figure 8-7 of the draft CD, several studies report stronger associations
15 with PM25 than for PM10_2 5, but some new studies suggest effect estimates for PM10_25 that are
16 similar or slightly larger in size than those for PM2 5. The draft CD also observes that there are a
17 number of considerations to take into account in interpreting the results of studies on different
18 PM size fractions, including differences in measurement error for the two size fractions,
19 potential variation from city to city in constituents of each size fraction, and the limited sample
20 sizes for some of the studies (CD, p. 8-57). Focusing on the studies with greater precision, the
21 draft CD reports that effect estimates for total mortality with both PM2 5 and PM10_2 5 generally
22 range from about 2 to 6% per 25 |ig/m3 of PM2 5 or PM10.25 (CD, p. 8-57).
23 In addition to those studies that included both PM2 5 and PM10_2 5 there are a number of
24 new studies that evaluated associations with PM25 but not PM10_25. Figure 3-5 includes effect
25 estimates for PM25 and mortality from U.S. and Canadian studies, including multi-city studies.
26 All effect estimates for PM2 5 and total mortality are positive and, especially for those more
27 precise effect estimates, many are statistically significant. The findings of new studies are
August 2003 3-25 Draft - Do Not Cite or Quote
-------
§ -
•Su. .
ro co
CO 0
<| "
^ .
t
£ o -
^ CN
2«> -
2 *~
0.
i- 0 .
!«•
l£>
ro
CD "? -
15
.io
to V
LLJ
Z'\
I
1
1
5
35 '
3
5
(
(
4
35
i
i
i
1
-a
T
i
,
1
37 1§
5 9fif
i zij i
T '
1 i
, t
1
I
I
I
7
1
i
1
i
15
1
ft
i
i
i
•
4
5
. .
-K 24
26
i
•
19
'
<
i
i
i
i
t
\
\
i
•) |
OOl
i
'i
i
i
»
/
1
i
i
>i
i
,'
J
Legend:
• = total mortality
A = cardiovascular, circulatory, or cardiorespiratory mortality
O = respiratory mortality
29
J 28
?i
i
i
i
1 1
16
t • 27
9 •" m-™ 2F,
r
i
3 ,. J I T» •
: !? «]fif Ji iif i fi
Q
O
CD
i
E
o
V)
0
"ro
E
K
0
0
(0
Q^
I
6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5
Natural Log of Mortality-Days
11.0
11.5
12.0
12.5
13.0
13.5
14.0
Figure 3-4. Effect estimates for PM10 and mortality from total, respiratory and cardiovascular causes from U.S. and Canadian
cities in relation to the mortality-days product (the product of study days and the number of deaths per day ~ an indicator of
study precision). Study locations are identified below; multi-city studies denoted by ^. Results of GAM stringent reanalyses;
studies not originally using GAM denoted by • (data in Appendix A).
1. * Burnett and Goldberg, 2003, 8 Canadian
cities
2. Chock et al., 2000, Pittsburgh •
3. * Domenici, 2003, 90 U.S. cities
4. Fairley, 2003, Santa Clara
5. Ito, 2003, Detroit
6. Ito and Thurston, 1996, Chicago •
7. Kinney et al., 1995, LA •
8. * Klemm and Mason, 2003, Six Cities
overall
9. Klemm and Mason, 2003, Boston
10. K loin in and Mason, 2003, Kinston/Harriman
11. Klemm and Mason, 2003, Portage
12. Klemm and Mason, 2003, Steubenville
13. Klemm and Mason, 2003, St. Louis
14. Klemm and Mason, 2003, Topeka
15. Klemm and Mason, 2000, Atlanta •
16. Lipfert et al., 2000a, Philadelphia •
17. Mar et al., 2003, Phoenix
18. Moolgavkar. 2003, LA County
19. Moolgavkar, 2003, Cook County
20. Ostro et al., 2000, Coachella Valley
21. Pope et al., 1992, Utah Valley •
22. * Schwartz, 2003b, 10 Cities overall
23. Schwartz, 2003b, Birmingham
24. Schwartz, 2003b, Canton
25. Schwartz, 2003b, Chicago
26. Schwartz, 2003b, Colorado Springs
27. Schwartz, 2003b, Detroit
28. Schwartz, 2003b, Minneapolis
29. Schwartz, 2003b, New Haven
30. Schwartz, 2003b, Pittsburgh
31. Schwartz, 2003b, Seattle
32. Schwartz, 2003b, Spokane
33. Schwartz, 1993, Birmingham •
34. Styer et al., 1995, Chicago «
35. Tsai et al., 2000, Camden •
36. Tsai et al., 2000, Elizabeth •
37. Tsai et al., 2000, Newark •
August 2003
5-26
Draft - Do Not Cite or Quote
-------
(0 ^
o
'2%
o
o
S
1±
o
1 O
c\i
s
Q_ V)
a
0°
03 ,„
(0
E
to
LJJ
LU
i i
18
<
(
4
c
'19
t
<
1 <
i
18
i
i
I
i
f 19
i
i
i
i
?
{
t
<
. <
7
3
320
'i-
i
i
»
3
i
1fi
i
i
i
i i
i
i
i
4
A /-
, 1.° '
" t I
il i
>i
i
i
1
?
i
i
1 i
i '
t 1
i '
I •
II
'I
i
Legend:
• = total mortality
A = cardiovascular, circulatory, or cardiorespiratory mortality
O = respiratory mortality
2 11
8 It 1 17
I | If! v! i
1 I
n
o
§
O>
P
i mates fror
-------
ro
'88
3
o
n
o
CL
i_ O
O)
ro
lo
E
m
3
t
i
i
4 <3
T1 j
O>
E
o
M—
"ro
"«)
LLJ
7.0
7.5
8.0
8.5 9.0 9.5
Natural Log of Mortality-Days
10.0
10.5
11.0
11.5
Figure 3-6. Effect estimates for PM10_2 5 and mortality from total, respiratory and cardiovascular causes from U.S. and
Canadian cities in relation to the mortality-days product (the product of study days and the number of deaths per day — an
indicator of study precision). Study locations are identified below; multi-city studies denoted by ^. Results of GAM stringent
reanalyses; studies not originally using GAM denoted by • (data in Appendix A)
1. 'A' Burnett and Goldberg, 2003, 8 Canadian cities
2. Chock et al., 2000, Pittsburgh •
3. Fairley, 2003, Santa Clara
4. Ito, 2003, Detroit
5. Klemm and Mason, 2000, Atlanta •
6. Lipfert et al., 2000a, Philadelphia •
7. Mar et al., 2003, Phoenix
8. Ostro et al., 2003, Coachella Valley
9. Schwartz, 2003a, Boston
10. Schwartz, 2003a, Kingston/ Harriman
11. Schwartz, 2003a, Portage
12. Schwartz, 2003a, St. Louis
13. Schwartz, 2003a, Steubenville
14. Schwartz, 2003a, Topeka
August 2003
5-28
Draft - Do Not Cite or Quote
-------
1 consistent with associations found between PM2 5 and total mortality in the previous PM NAAQS
2 review (CD, p. 8-76).
3 As shown in Figure 3-6, effect estimates between PM10_2 5 and total mortality are also
4 generally positive, though less likely to reach statistical significance. In the previous PM
5 NAAQS review, very few studies tested associations with PM10_2 5, with only the multi-city study
6 using Six Cities data presenting results for PM10_2 5 with mortality. In that study, a statistically
7 significant association between total mortality and PM10_2 5 was reported for only one of the
8 cities. Some new U.S. and Canadian studies have now reported positive associations with PM10.
9 25, and some show statistically significant results (Mar et al., 2003; Ostro et al., 2003; results for
10 total mortality not reanalyzed) but most do not. The CD concludes that, while the evidence is
11 not as consistent, recent studies also suggest that PM10_2 5 may also be associated with daily
12 mortality (CD, p. 8-77).
13 3.3.1.1.3 Cause-specific Daily Mortality
14 In the 1996 Staff Paper, several studies also reported associations between PM10 and
15 respiratory and cardiovascular mortality (EPA, 1996b, p. V-13). The associations reported with
16 mortality from respiratory or cardiovascular diseases were generally consistent with the results
17 for total mortality, and the draft CD concluded that this lent support to the biological plausibility
18 of the PM associations (EPA, 1996a, p. 12-69).
19 Appendix A and Figures 3-4 through 3-6 include effect estimates for mortality from
20 respiratory and cardiovascular causes, as well as for all nonaccidental causes, from multi-city
21 and single-city studies. For example, the recent U.S. multi-city study, NMMAPS, included a
22 comparison of findings for total and cardiorespiratory mortality for the 20 largest U.S. cities.
23 The effect estimate for deaths from cardiorespiratory causes was somewhat larger (1.6% increase
24 per 50 |ig/m3 increase in PM10) than that for deaths from all causes (1.1% increase per 50 |ig/m3
25 increase in PM10) (GLM results, Samet et al., 2000c, reanalyzed in Domenici, 2003). In general,
26 for respiratory and cardiovascular mortality, the U.S. and Canadian studies tend to show
27 somewhat larger effect estimates than for total mortality associations with PM10 (or PM15) and
28 PM2 5 (e.g., Gwynn et al., 2000, not reanalyzed; Ostro et al., 1999, reanalyzed in Ostro et al.,
29 2003; Pope et al., 1999a, not reanalyzed; Fairley, 1999, reanalyzed in Fairley, 2003; Lippmann et
30 al., 2000, reanalyzed in Ito, 2003; Mar et al., 2000, reanalyzed in Mar et al., 2003; Goldberg et
August 2003 3-29 Draft - Do Not Cite or Quote
-------
1 al., 2000, reanalyzed in Goldberg and Burnett, 2003; Tsai et al., 2000) (Figures 3-4 and 3-5). As
2 was observed in the 1996 CD, some of the effect estimates for respiratory mortality are larger in
3 magnitude but less precise, with large confidence intervals, which is likely because respiratory-
4 related deaths comprise a small proportion of daily mortality rates. As was found with total
5 mortality, few significant associations were reported with PM10_2 5 for cause-specific mortality;
6 however, in those few studies, the effect estimates for cardiovascular mortality tended to be
7 greater than those for total mortality (Mar et al., 2000; Ostro et al., 2000) (both reanalyzed only
8 the cardiovascular mortality results (Mar et al., 2003; Ostro et al., 2003)) (Figure 3-6).
9 A number of investigators have also tested for associations between PM and "other" or
10 non-cardiorespiratory deaths, and generally no such associations are reported. However, in a
11 few studies, positive, though not always statistically significant, associations were reported for
12 the "other" mortality category. Regarding these findings, the draft CD concludes, "at least some
13 of these 'other' associations may also be due to seasonal cycles that include relationships to
14 peaks in influenza epidemics that may imply respiratory complications as a 'contributing' cause
15 to the 'other' deaths. Or, the 'other' category may include sufficient numbers of deaths due to
16 diabetes or other diseases which may also involve cardiovascular complications as contributing
17 causes" (CD, p. 8-75).
18 In addition to the evidence from epidemiologic studies, new, though limited, information
19 is available from toxicologic studies that offers insight into PM-related mortality. In some of the
20 toxicologic studies using an animal model of cardiopulmonary disease, some animals died after
21 exposure to ROFA, a combustion source related particle sample, though none of these studies
22 was designed to assess lethality (CD, p. 7-45). Indicators of inflammation or cardiac arrhythmia
23 were also measured in these studies (CD, Table 7-7a). While the applicability of effects found
24 for this animal model to human health may be questioned, and the doses were relatively high, the
25 findings lend plausibility to the associations with cardiorespiratory mortality reported in
26 epidemiologic studies. Since the studies were designed to assess effects on cardiovascular or
27 respiratory systems, the toxicological evidence for PM-related effects is more fully discussed in
28 the sections on respiratory and cardiovascular morbidity effects.
29 In summary, considering studies conducted in the U.S. and other countries, the new
30 studies continue to report risks for mortality from cardiovascular and respiratory diseases with
August 2003 3-30 Draft - Do Not Cite or Quote
-------
1 increasing PM, and the findings suggest that associations reported for total mortality are
2 indicative of associations with deaths from cardiorespiratory-related causes. For all studies, as
3 noted above, the draft CD observed that effect estimates for total mortality ranged generally from
4 about 2 to 6% per 25 |ig/m3 of PM2 5 or PM10_25. Again considering all studies, the draft CD
5 reports that cause-specific effect estimates generally range from about 2 to 20% per 25 |ig/m3 of
6 PM2 5 for cardiovascular or cardiorespiratory mortality and from about 2 to 14% per 25 |ig/m3 of
7 PM2 5 for respiratory mortality (CD, p. 9-85). For PM10_2 5, cause-specific effect estimates from
8 all studies generally range from about 3 to 7% per 25 |ig/m3 of PM10_25 for cardiovascular or
9 cardiorespiratory mortality and from about 3 to 6% per 25 |ig/m3 of PM10_2 5 for respiratory
10 mortality (CD, p. 9-87)
11
12 3.3.1.2 Mortality and Long-term PM Exposure
13 The 1996 CD summarized the findings of a number of cross-sectional studies that had
14 been conducted over several previous decades. The cross-sectional studies had generally
15 identified associations between residence in communities with higher pollution levels and
16 increased mortality in adult populations, but concern had been raised about the lack of
17 information on potentially important covariates and methodological limitations (EPA, 1996a, p.
18 12-159).
19 In contrast with cross-sectional studies, prospective cohort studies include subject-
20 specific information on potential confounders (e.g., smoking history, occupation, health history)
21 and have been considered to provide more reliable results (EPA, 1996a, p. 13-33). In the
22 previous review, results were presented for three recent prospective cohort studies of adult
23 populations (i.e., the Six Cities, American Cancer Society (ACS), and California Seventh Day
24 Adventist (AHSMOG) studies). The Six Cities study (Dockery et al., 1993) evaluated effects of
25 many PM size classes, and significant associations were reported with PM15, PM2 5, sulfates and
26 non-sulfate fine particles, but not with TSP or coarse particles (TSP-PM15 or PM15-PM2 5) (EPA,
27 1996a, Table 12-18). The ACS study (Pope et al., 1995) reported significant associations for
28 PM2 5 and sulfates (a fine particle surrogate). Both the Six Cities and ACS studies reported
29 associations with mortality from all causes and cardiorespiratory causes, with larger effect
30 estimates for cardiorespiratory causes. The AHSMOG study did not find an association between
August 2003 3-31 Draft - Do Not Cite or Quote
-------
1 TSP and mortality (Abbey et al., 1991). The 1996 CD concluded that the chronic exposure
2 studies, taken together, suggested associations between increases in mortality and long-term
3 exposure to PM (EPA, 1996a, p. 13-34).
4 The new studies of adult mortality that are available for the current review include a
5 comprehensive reanalysis of data from the Six Cities and ACS studies (Krewski et al., 2000),
6 new analyses using updated data from the AHSMOG study (Abbey et al., 1999) and the ACS
7 study (Pope et al., 2002), and a new analysis using data from a cohort of veterans (Lipfert et al.,
8 2000b). Findings from the original Six Cities, ACS, and AHSMOG investigations together with
9 those from new studies and reanalyses are summarized in Table 3-3. In addition, a new cross-
10 sectional study of adult populations has been published (Lipfert and Morris, 2002) and several
11 recent studies have suggested relationships between long-term PM exposure (e.g., months) and
12 mortality in infants, as discussed later in this section.
13 The reanalysis of the Six Cities and ACS studies included two major components, a
14 replication and validation study, and a sensitivity analysis, where alternative risk models and
15 analytic approaches were used to test the robustness of the original analyses. In the first phase,
16 Krewski and colleagues (2000) reported the data from the two studies to be of generally high
17 quality, and were able to replicate the original results, confirming the original investigators'
18 findings of associations with both total and cardiorespiratory mortality (CD, p. 8-83).
19 In the second phase, the sensitivity analyses generally reported that the use of alternative
20 models, including variables that had not been used in the original analyses (e.g., physical
21 activity, lung function, marital status), did not materially alter the original findings. Krewski and
22 colleagues (2000) also obtained data on additional city-level variables that were not available in
23 the original data sets (e.g., population change, measures of income, maximum temperature,
24 number of hospital beds, water hardness) and included these data in the models. The
25 associations between fine particles and mortality were generally unchanged in these new
26 analyses, with the exception of population change, which did somewhat reduce the size of the
27 associations with fine particles or sulfates.
28 Further sensitivity analyses were conducted using data for potentially susceptible
29 subgroups, and the results did not show differences in the PM-mortality associations among most
30 subgroups, including those defined by gender, smoking status, exposure to occupational dusts
August 2003 3-32 Draft - Do Not Cite or Quote
-------
1 and fumes, and marital status. However, the effects of fine particles appeared to be larger in the
2 subgroup of persons without a high school education than in one of persons with more
3 education; Krewski and colleagues (2000) postulated that this relationship could be due to some
4 unidentified socioeconomic effect modifier.
5 The original Six Cities and ACS analyses had not included assessment of potential
6 confounding by co-pollutants. The Six Cities study had included assessment of associations
7 between mortality and the gaseous pollutants individually (Dockery et al., 1993), and significant
8 or borderline significant associations were reported with SO2 and NO2, but it was observed that
9 these pollutants were strongly correlated with PM (EPA, 1996a, p. 12-168). For the reanalyses,
10 Krewski and colleagues obtained additional data on gaseous pollutant concentrations and
11 evaluated the effects of these pollutants both alone and with PM in multi-pollutant models.
12 Significant associations were reported between mortality and sulfur dioxide, and in multiple
13 pollutant models, the sulfur dioxide associations often appeared stronger than those for fine
14 particles and sulfates. The authors suggest that sulfur dioxide may be acting as a marker for
15 other mortality-associated pollutants, and conclude "[nonetheless, both fine particles and sulfate
16 continued to demonstrate a positive association with mortality even after adjustment for the
17 effects of sulfur dioxide in our spatial regression analyses." (Krewski et al., 2000, p. 233, 234)
18 Several methods were used to address variation from city to city, or spatial correlation
19 among cities, using the larger sulfate data set (approximately 150 cities, as compared with
20 approximately 50 cities for which there were PM2 5 data) The resulting sulfate associations were
21 sometimes smaller and sometimes larger than the original effect estimates. The authors
22 concluded that this "suggests that uncontrolled spatial autocorrelation accounts for 24% to 64%
23 of the observed relation. Nonetheless, all our models continued to show an association between
24 elevated risks of mortality and exposure to airborne sulfate." (Krewski et al., 2000, p. 228).
August 2003 3-33 Draft - Do Not Cite or Quote
-------
Table 3.3 Effect Estimates for Mortality per Increments in Long-term Mean Levels of
Fine- and Coarse-fraction Particle Indicators from U.S. and Canadian Studies (from draft
CD Table 9-11)
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)
SOJ (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)
PM25 (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)
18** (9, 34)
11** (4, 24)
NR(18, 47)
MR (11, 30)
59(34, 101)
20 (10, 38)
7.1(9,42)
21 (10, 30)***
14 (5, 20)***
18 (±4)
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
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)
****
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)
NR(18, 47)
MR (11, 30)
NR(18, 47)
MR (11, 30)
18** (9, 34)
59(34, 101)
20 (10, 38)
7.1(9,42)
August 2003
3-34
Draft - Do Not Cite or Quote
-------
Study
ACS Study Extended
AnalysesE
Southern CaliforniaF
Southern California11
Increased Lung Cancer
Six CityA
Six City ReanalysisD
ACS Study0
ACS Study ReanalysisD
ACS Study Extended
AnalysesE
Southern CaliforniaF
Southern California11
Indicator (Increment)
PM25 (10 ug/m3) (1979-83)
PM25 (10 ug/m3) (1999-00)
PM25 (10 ug/m3) (average)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
PM10.25 (10 ug/m3)
Mortality in Adults
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)
PM2 5 (10 ug/m3) (1979-83)
PM25 (10 ug/m3) (1999-00)
PM25 (10 ug/m3) (average)
PM10 (20 ug/m3)
PM25 (10 ug/m3)
Relative Risk (95% CI)
1.06(1.02, 1.10)
1.08(1.02, 1.14)
1.09(1.03, 1.16)
1.01 (0.92, 1.10)
1.23 (0.97, 1.55) (males)
1.20 (0.87, 1.64) (males)
****
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)
Study
Concentrations *
(us/m3)
21 (10, 30)***
14 (5, 20)***
18 (±4)
51(0,84)
32 (17, 45)
27 (4, 44)
NR(18, 47)
MR (11, 30)
NR(18, 47)
NR (11,30)
18** (9, 34)
59(34, 101)
20 (10, 38)
7.1(9,42)
21 (10, 30)***
14 (5, 20)***
18 (±4)
51(0,84)
32 (17, 45)
27 (4, 44)
* Presented as mean (min, max), or mean (±SD).
** Median
*** Minimum and maximum values estimated from Figure 1 (Pope et al., 2002)
**** Results only for smoking category subgroups.
References:
ADockeryetal. (1993)
B EPA (1996a)
c Pope etal. (1995)
D Krewski et al. (2000)
E Pope et al. (2002)
F Abbey etal. (1999)
G Lipfert et al. (2000b)
H McDonnell et al. (2000)
1 In summary, the draft CD concludes that the reanalyses generally confirm the original
2 investigators' findings of associations between mortality and long-term exposure to fine particles
3 with the Six Cities and ACS cohort data. As shown in draft CD Table 8-5, the mortality relative
4 risk estimates reported in the replication analysis were nearly identical to those reported in the
5 original studies (CD, p. 8-84). In the sensitivity analyses, Krewski et al. (2000) reported risk
6 estimates that were "remarkably robust to alternative risk models" (p. 25). While recognizing
August 2003
3-35
Draft - Do Not Cite or Quote
-------
1 that increased mortality may be attributable to more than one component of ambient air
2 pollution, the reanalysis confirmed the association between mortality and fine particle and
3 sulfate exposures (CD, p. 8-87).
4 A new study included extended analyses using the ACS cohort data with additional
5 follow-up health data and air quality data from the new fine particle monitoring network for
6 1999-2000 (Pope et al., 2002). As shown in Table 8-7 of the draft CD, this new study reports
7 significant associations between long-term exposure to fine particles (using the original air
8 quality data for 1979-1982) and premature mortality from all causes, cardiopulmonary diseases,
9 and lung cancer (p. 8-90). Thus, the relationship persisted with the longer follow-up period for
10 health data, and the effect estimates, while somewhat smaller than those in the original analysis,
11 were consistent with those from the original analysis. Significant associations were also found
12 when using the more recent air quality data, and average concentrations for all air quality data.
13 The authors noted that the PM2 5 concentrations from 1999-2000 were highly correlated with
14 PM25 concentrations from 1979-1983, indicating that "the rank ordering of cities by relative
15 pollution levels remained nearly the same" (Pope et al., 2002, p. 1136). As was seen in the
16 reanalysis study (Krewski et al., 2000), significant associations were reported for fine particles
17 and sulfates, but not coarse fraction particles or TSP.
18 This extended analysis included the use of data on gaseous pollutant concentrations,
19 more recent data on fine particle concentrations, and evaluated further the influence of other
20 covariates (e.g., dietary intake data, occupational exposure) and model specification in the PM-
21 mortality relationship (e.g., new methods for spatial smoothing and random effects models in the
22 Cox proportional hazards model) (Pope et al., 2002). The investigators reported that the
23 associations found with sulfate and fine particle concentrations were robust to the inclusion of
24 many covariates for socioeconomic factors or personal health variables (e.g., dietary factors,
25 alcohol consumption, body mass index). Their findings in this regard were consistent with those
26 of Krewski et al. (2000). Pope et al. (2002) also found no significant spatial autocorrelation after
27 controlling for fine particles. While Krewski et al. (2000) found that effect estimates change
28 with different methods of control for spatial variability, Pope and colleagues tested new methods
29 for modeling spatial variability, and reported that, after adjusting for spatial correlation, the
August 2003 3-36 Draft - Do Not Cite or Quote
-------
1 estimated PM2 5 effects were significant and persistent for cardiopulmonary and lung cancer
2 mortality, and were nearly significant for total mortality (CD, p. 8-96).
3 Other new analyses using updated data from the AHSMOG cohort included more recent
4 air quality data for PM10 and estimated PM2 5 concentrations from visibility data, along with new
5 health information from continued follow-up of the Seventh Day Adventist cohort (Abbey et al.,
6 1999; McDonnell et al., 2000). In contrast with the original study in which no statistically
7 significant results were reported with TSP, a significant association was reported between total
8 mortality and PM10 (number of days exceeding 100 |ig/m3) for males (CD, p. 8-98). However,
9 no significant associations were reported for other PM10 indices (e.g., 20 |ig/m3 increase), for
10 deaths from contributing respiratory causes, or for females. Additional analyses were conducted
11 using only data from males and estimated PM2 5 and PM10_2 5 concentrations; larger effect
12 estimates were reported for mortality with PM25 than with PM10_2 5, but again, the estimates were
13 generally not statistically significant (CD, Table 8-11).
14 A new long-term exposure study uses data from a cohort of up to 70,000 military men
15 assembled by the Veterans Administration in the 1970s (Lipfert et al., 2000b). The cohort was
16 male, middle-aged (mean age 51 years), included a higher proportion of African-Americans and
17 current or former smokers than the general population, and the participants were selected on the
18 basis of having mild or moderate hypertension (CD, p. 8-101). As such, this cohort represents an
19 "at-risk" population group that is markedly distinct from the general U.S. population.
20 The authors report that, when methods consistent with those in the other prospective
21 cohort studies were used, similar results were found, stating "[t]he single-mortality-period
22 responses without ecological variables are qualitatively similar to what has been reported before
23 (SO4= > PM25 > PM15)" (CD, p. 8-103). In analyses using shorter PM exposure and mortality
24 time periods, the investigators report inconsistent and largely nonsignificant associations
25 between PM exposure (including, depending on availability, TSP, PM10, PM2 5, PM15 and PM15.
26 2 5) and mortality. Ecological variables included education, income, population race proportions,
27 altitude, poverty rate, and number of heating-degree days. The final model used by the authors
28 included 233 variables, of which 162 were interaction terms of systolic blood pressure, diastolic
29 blood pressure, and body mass index variables with age. One hypothesis offered by the authors
August 2003 3-37 Draft - Do Not Cite or Quote
-------
1 is that the "relatively high fraction of mortality within this cohort may have depleted it of
2 susceptible individuals in the later periods of follow-up" (CD, p. 8-103).
3 Three of the prospective cohort studies described above (Six Cities, ACS, AHSMOG)
4 also examined long-term exposure to PM in relation to lung cancer mortality. In the results
5 available during the previous review, none of the these studies reported a significant association
6 between long-term exposure to fine particles and lung cancer mortality, although the ACS study
7 reported a significant association between lung cancer mortality and sulfates (EPA, 1996b, p. V-
8 17). The reanalysis study confirmed these findings for the Six Cities and ACS studies (Krewski
9 et al., 2000).
10 One new effect reported in the extended analysis of the ACS study were statistically
11 significant associations between fine particle and sulfate concentrations and lung cancer
12 mortality. Pope et al. (2002) report an 8% increased risk of lung cancer mortality per 10 |ig/m3
13 change in PM2 5, and this effect estimate is little changed and remains significant with adjustment
14 for covariates, random effects modeling and spatial smoothing methods (CD, Figure 8-10). In
15 new analyses using updated data from the AHSMOG cohort, significant associations were
16 reported between long-term PM10 exposure and lung cancer mortality for males, but not females
17 and some associations were also reported with other gaseous pollutants (Beeson et al., 1998). In
18 contrast, in the original analyses (Pope et al., 1995; Abbey et al., 1991) no significant
19 associations were reported with lung cancer mortality. The findings were based on a small
20 number of lung cancer deaths in the cohort, and the effect estimates were quite variable (CD, p.
21 8-98). Further analysis using data for males and estimated PM2 5 and PM10_2 5 concentrations
22 reported no statistically significant associations with lung cancer mortality for either PM2 5 or
23 PM10.25 (McDonnell et al., 2000; CD, p. 8-101).
24 The epidemiologic findings of associations between fine particles and lung cancer
25 mortality are supported by the results of a few recent toxicologic studies that have examined the
26 mutagenic potential of ambient particles and "have shown some degree of evidence that appears
27 to support the biologic plausibility of the long-term lung cancer effects," though it is noted that
28 the studies generally lack adequate information on dose (CD, p. 7-69). In addition, the Health
29 Assessment Document for diesel engine exhaust concludes that diesel engine exhaust, one
August 2003 3-38 Draft - Do Not Cite or Quote
-------
1 source of PM emissions, is a likely human carcinogen (EPA, 2002). On the results of the new
2 epidemiologic studies, the draft CD concluded "[o]verall, these new cohort studies confirm and
3 strengthen the published older ecological and case-control evidence indicating that living in an
4 area that has experienced higher PM exposures can cause a significant increase in RR of lung
5 cancer incidence and associated mortality" (CD, p. 8-228).
6 The draft CD relies heavily on the results of the ACS and Six Cities study, the reanalysis
7 of these studies, and the extended ACS study in evaluating the results of the prospective cohort
8 studies. The draft CD concludes that the "lack of consistent findings in the AHSMOG study and
9 negative results of the VA study, do not negate the findings of the Six Cities and ACS studies"
10 (CD, p. 8-105). The Six Cities and ACS studies (and reanalyses and new analyses) included
11 measured PM data (in contrast with AHSMOG PM estimates based on TSP or visibility
12 measurements), have study populations more similar to the general population than the VA study
13 cohort, and have been validated through an exhaustive reanalysis (CD, p. 8-108 and 8-109). The
14 draft CD concludes that the recent publications "collectively appear to confirm earlier cross-
15 sectional study indications that the fine mass component of PM10 (and usually especially its
16 sulfate constituent) are more strongly correlated with mortality than is the coarse PM10_25
17 component" (CD, p. 8-109). The effect estimates for deaths from all causes fall in a range of 4
18 to 14% increased risk per 10 |ig/m3 increase in PM25, with associations for cardiopulmonary
19 mortality (6 to 19% increased risk per 10 |ig/m3 increase in PM25) and lung cancer (1 to 39%
20 increased risk per 10 |ig/m3 increase in PM25) being generally larger than those for total
21 mortality (Table 3-3).
22 In addition to the prospective cohort studies, cross-sectional studies have provided
23 information on the relationship between long-term PM exposure and adverse health effects,
24 though emphasis has been placed on the findings of cohort studies due to the availability of
25 individual health data on study participants. One new U.S. cross-sectional study has been
26 conducted that uses pollution and mortality data from five time periods (Lipfert and Morris,
27 2002). Significant associations were reported for all pollutants considered, though the
28 associations varied by age group and pollutant. The authors emphasized their findings of
29 temporal changes in the pollution-mortality association; there was a trend toward reduced effects
August 2003 3-39 Draft - Do Not Cite or Quote
-------
1 when using more recent data sets, though the associations found using PM2 5 data from the new
2 fine particle monitoring network were somewhat larger than those using older data.
3 One additional new analysis of the Six Cities data has evaluated mortality risk with
4 different estimates of long-term PM25 exposure. The original study (Dockery et al., 1993)
5 averaged PM concentrations over a period of years to represent long-term PM exposure
6 estimates, while the new analysis includes PM averages over a range of time periods, such as 2
7 or 3-5 years preceding the individual's death (Villeneuve et al., 2002). The authors reported that
8 effect estimates for mortality were lower with time-dependent PM2 5 exposure indicators (e.g., 2
9 years before individual's death), than with the longer-term average concentrations. They
10 postulate that this is likely due to the "influence of city-specific variations in mortality rates and
11 decreasing levels of air pollution that occurred during follow-up" (CD, p. 8-88). This might be
12 expected, if the most polluted cities had the greatest decline in pollutant levels as controls were
13 applied (CD, p. 8-85). The authors observe that the fixed average concentration window may be
14 more representative of cumulative exposures and thus a more important predictor of mortality
15 than a shorter time period just preceding death (Villeneuva et al., 2002, p. 574).
16 Summary. Positive, statistically significant associations between mortality from total or
17 cardiorespiratory causes and long-term exposure to particles, especially to fine particles, were
18 reported in the original Six Cities and ACS studies. These results were confirmed in extensive
19 reanalyses and new analyses using updated health and air quality data for the ACS study. In
20 considering these results, as well as the other inconsistent and negative results discussed above,
21 the draft CD concludes that long-term PM2 5 exposure durations are likely associated with serious
22 human health effects. (CD, p. 8-277).
23 3.3.1.3 Infant Mortality
24 In addition to findings of the prospective cohort studies for adult populations, several
25 new studies have also reported associations between mortality in infants and long-term pollutant
26 concentrations. In these studies, the exposure periods were one month or more during gestation
27 or around the time of birth. Each of the studies reviewed in the draft CD (section 8.2.3.4)
28 reported significant associations between infant mortality and PM exposure. One study, using
29 data from 80 U.S. cities, reported significant associations between PM10 concentrations during
August 2003 3-40 Draft - Do Not Cite or Quote
-------
1 the first 2 months of the infant's life and mortality from respiratory causes and sudden infant
2 death syndrome (Woodruff et al., 1997). Lipfert et al. (2000c) conducted a similar analysis using
3 annual average PM10 to represent the infants' pollution exposure levels (instead of the 2-month
4 postnatal average), and reported associations similar to those in Woodruff et al. (1997), but also
5 reported that geographic gradients in pollution levels and infant mortality rates make interpreting
6 these results difficult (CD, p. 8-114). In addition, Chay and Greenstone (2001a,b) linked
7 reductions in TSP concentrations with reduced infant mortality in the U.S. (CD, p. 115). Studies
8 conducted in the Czech Republic and Mexico City also reported associations between PM
9 exposure and infant mortality. The draft CD concludes that the findings from this group of
10 studies are "suggestive of a causal relationship between PM exposure and infant mortality" (CD,
11 p. 8-115).
12 3.3.1.4 Mortality Displacement and Life-Shortening
13 The 1996 CD and Staff Paper discussed the issue of mortality displacement, or whether
14 the associations reported between mortality and short-term exposure represent deaths among the
15 weakest individuals who might have died within days even without PM exposure (sometimes
16 referred to as "harvesting"). Limited data were available, and it was concluded that there may be
17 evidence of mortality displacement occurring in some portion of the population, but that further
18 research was needed to more fully address this question (EPA, 1996b, p. V-19).
19 In addition, the draft CD assessed potential life-shortening that may occur with long-
20 term exposure to PM, and concluded that increased mortality results from both short-term and
21 long-term ambient PM exposure, and that the amount of life shortening could potentially be on
22 the order of years (EPA, 1996a, p. 13-45).
23 More recently, the extent to which mortality displacement may be occurring was
24 investigated using two new types of analyses. One type of study separated time-series data into
25 three components — seasonal and longer fluctuations, intermediate fluctuations, and short-term
26 fluctuations — and varied the cutoff between the intermediate and short-term cycles to test for
27 the presence of harvesting. While there was evidence from the analysis done in Boston that
28 mortality from chronic obstructive pulmonary disease (COPD) may be displaced by a only few
29 months, effect sizes for deaths from pneumonia, heart attacks, and all causes were reported to
August 2003 3-41 Draft - Do Not Cite or Quote
-------
1 increase as longer time scales were included, thus offering no evidence for harvesting effects
2 (Schwartz, 2000c, reanalyzed in Schwartz, 2003). Using data from Milan, Italy, positive
3 associations were reported between TSP and mortality up to 13 days, with no effect reported in
4 the next few days, then positive coefficients from 20 days to 45 days (maximum time scale used
5 in study), possibly providing evidence for an initial "rebound" due to depletion of the
6 susceptible population, but with an overall increase in effect size when considering mortality
7 over the longer time scale (Zanobetti et al., 2000). Using first simulation analyses, then analyses
8 using data from Philadelphia, effects of harvesting were assessed at 3 days, 30 days, and 300
9 days (Zeger et al., 1999), and larger effect sizes were reported for the longer frequency ranges.
10 The results of this group of studies "suggest that the extent of harvesting, if any, is not a matter
11 of only a few days" (CD, p. 8-273).
12 In contrast, two recent studies analyzed data from Chicago and Philadelphia (Smith et al.,
13 1999; Murray and Nelson, 2000, respectively) using methods that included estimation of the size
14 of the "frail" population, and the effect of air quality changes on the life expectancy of those in
15 that population. Findings from both studies "suggest that the frail population is very small and
16 its lifetime short, such that PM or any external stress cannot have more than a few days of life-
17 shortening impacts" (CD, p. 8-273). The draft CD notes that this may be an inherent limitation
18 of the model used in these studies, since it was assumed that mortality was occurring only in the
19 fairly small frail population and PM exposure can also be associated with the development of
20 underlying health problems, thus increasing the size of the frail population (CD, p. 9-138).
21 Consistency is apparent in these studies when comparing results within similar models, but not
22 across the different types of models. Overall, in considering the results of this new body of
23 research, the draft CD concludes that "there appears to be no strong evidence to suggest that PM
24 is shortening life by a few days" (CD, p. 8-273).
25 The extent of life-shortening that may be associated with long-term PM exposure has
26 been investigated in a recent analysis using effect estimates from existing studies and life-Table
27 analysis methods (Brunekreef, 1997). Chronic exposure to PM, with an exposure difference of
28 10 |ig/m3, was associated with a reduction in 1.31 years in the population's life expectancy at age
29 25. Taking into account the evidence from a few new studies showing associations between
August 2003 3-42 Draft - Do Not Cite or Quote
-------
1 infant mortality and PM exposure, the draft CD finds that these data suggest that potential loss of
2 population life expectancy associated with long-term PM exposure may be even greater than
3 Brunekreef s (1997) estimate (CD, p. 8-274).
4
5 3.3.2 Indices of Morbidity
6 As noted in the 1996 PM Staff Paper, given the statistically significant positive
7 associations between community PM concentrations and mortality, it is reasonable to anticipate
8 that comparable epidemiologic studies should find increased morbidity with elevated levels of
9 PM (EPA, 1996b, p. V-21). This was indeed the case in the past review, where positive
10 associations were reported between PM and morbidity effects ranging from the more severe
11 (e.g., hospitalization for respiratory or cardiovascular diseases) to moderate exacerbation of
12 respiratory conditions or decreases in lung function. Staff noted the logical relationships
13 between the cause-specific mortality and hospital admissions results, as well as those across the
14 range of morbidity effects and sensitive populations.
15 A number of more recent epidemiologic studies also find increased hospital admissions
16 or emergency room visits, as well as changes in lung function and respiratory symptoms with
17 PM exposure. Other new epidemiologic studies have expanded the range of morbidity indices
18 associated with PM, including physicians' office or clinic visits for respiratory disease, and
19 cardiovascular health indicators such as heart rate or heart rate variability. In the previous
20 review, several epidemiologic studies also reported increased numbers of school absences, lost
21 work days or restricted activity days with increased PM (EPA, 1996b, p. V-22); little new
22 evidence is provided for the latter indices in the draft CD.
23 Recent toxicological and controlled human exposure studies have provided new evidence
24 on morbidity effects that suggests linkages with effects reported in epidemiologic studies.
25 Effects related to some new endpoints measured in the recent epidemiologic studies, such as
26 heart rate variability, were first reported in animal toxicologic studies. Some toxicologic studies
27 have used ambient PM samples from areas in which epidemiologic studies were conducted (e.g.
28 Ohio, 1999a,b). In addition, many laboratory studies have measured cellular or physiological
29 changes, such as changes in numbers of immune cell types, levels of cytokines, or measures of
August 2003 3-43 Draft - Do Not Cite or Quote
-------
1 pulmonary or cardiovascular function following exposure to CAPs or instilled ambient particles.
2 The more subtle biological responses measured in such studies may provide supporting evidence
3 for morbidity associations reported without being considered themselves as separate indices of
4 morbidity.
5 3.3.2.1 Hospital Admissions and Emergency Room Visits
6 Hospitalization and emergency room visits are measures of more severe respiratory or
7 cardiovascular morbidity, and associations with these health outcomes have been evaluated in
8 numerous studies. The 1996 Staff Paper observed that epidemiologic studies demonstrated
9 associations between hospital admissions and emergency room visits for respiratory and cardiac
10 causes and PM10 exposure (EPA, 1996b, p. V-21). Most studies evaluated relationships with
11 admissions/visits for respiratory diseases, including asthma, COPD and pneumonia, and nearly
12 all associations were statistically significant. The 1996 CD included some limited information
13 on the relationship between PM10 and cardiovascular admissions, but while few in number, the
14 studies did report significant associations for these effects. Where multi-pollutant models were
15 evaluated, associations reported with PM10 were not substantially changed with the inclusion of
16 gaseous co-pollutants in the models. The 1996 CD included results from only one study where
17 PM2 5 and PM10_2 5 data were available, and positive associations with total respiratory
18 admissions/visits were reported for both; statistically significant associations were reported with
19 fine particles or fine particle components, and these associations were larger and less influenced
20 by copollutant confounding than those with PM10_25 (Thurston et al., 1994). As noted in the
21 1996 Staff Paper, the associations reported with hospital admissions and emergency room visits
22 were coherent with the findings of significant associations with mortality, especially mortality
23 from cardiovascular and respiratory causes.
24 Numerous recent studies have continued to report significant associations between PM
25 and hospital admissions or emergency room visits for respiratory or cardiovascular diseases.
26 The new studies have included multi-city analyses, numerous assessments using cardiovascular
27 admissions/visits, and evaluation of the effects of fine- and coarse-fraction particles. The results
28 for all new cardiovascular and respiratory admissions/visits studies, including those using
29 nongravimetric PM measurements and studies from non-North American locations, are
August 2003 3-44 Draft - Do Not Cite or Quote
-------
1 summarized in the draft CD in Tables 8-17, 8-19, 8-20, 8-21, and 8-22 (with further descriptive
2 information about the studies in Tables 8B-1 and 8B-2), and the effect estimates for PM10, PM2 5
3 or PM10_25 from U.S. and Canadian studies are summarized in Appendix A. In general, the
4 recent studies provide evidence that support previously identified associations between PM and
5 admissions/visits for respiratory and cardiovascular diseases (CD pp. 8-136 and 8-167).
6 As for mortality, staff have focused on the findings from U.S. and Canadian studies for
7 associations with PM10, PM25 and PM10_25, and the results are presented in Figures 3-7, 3-8 and
8 3-9, respectively. In these figures, effect estimates are presented by general respiratory or
9 cardiovascular effects categories, separated into more specific subcategories in cases where
10 available (e.g., COPD, asthma). Effect estimates for PM10 presented in Figure 3-7 include
11 findings from multi-city studies, as well as results from studies available for review in the 1996
12 CD, with the range of statistically significant effect estimates from the 1996 CD indicated at the
13 right-hand margin; for PM2 5 or PM10_25, the effect estimates from the only study on respiratory
14 admissions/visits available in the 1996 CD are indicated in the right-hand margins in Figures 3-7
15 and 3-8.
16 The recent U.S. multi-city study, NMMAPS, reported statistically significant associations
17 between PM10 and hospital admissions in the elderly for cardiovascular diseases, pneumonia or
18 COPD in 14 cities (Samet et al., 2000b; reanalyzed in Zanobetti and Schwartz, 2003), with
19 somewhat larger effect estimates when a distributed lag approach was used (Zanobetti et al.,
20 2000; reanalyzed in Zanobetti and Schwartz, 2003). Increases of 5% in hospital admissions for
21 cardiovascular disease and 8% and 6% in hospital admissions for COPD or pneumonia
22 (respectively) per 50 |ig/m3 increase in PM10 were reported (CD Tables 8-16 and 8-17) (in GLM
23 models using penalized splines). In addition, the authors used a new approach for evaluating
24 potential confounding by testing for associations between the PM effect estimate and the PM-
25 gaseous pollutant relationships in each location (as was done in multi-city mortality analyses
26 described in section 3.3.1.1.1). No evidence was found for trends between the PM10-copollutant
27 coefficients (for O3 or SO2) and PM10-respiratory admissions associations, or between the PM10-
28 copollutant coefficients (for CO, O3 or SO2) and PM10-cardiovascular admissions associations,
August 2003 3-45 Draft - Do Not Cite or Quote
-------
1 indicating that confounding by co-pollutants is unlikely (Samet et al., 2000b; reanalyzed in
2 Zanobetti and Schwartz, 2003).
3 These data were also used to explore potential factors that might result in variation in
4 effect estimates across cities. The authors reported that PM10-admissions effect estimates were
5 smaller in cities where air conditioning use was greater, which is an indicator of heterogeneity of
6 exposure rather than heterogeneity in biological effects. In addition, there was evidence of
7 larger effect size with greater percentage of PM10 emissions from highway vehicles in an area,
8 suggesting the potential importance of PM from motor vehicles in these health associations
9 (Janssen et al., 2001; reanalyzed in Zanobetti and Schwartz, 2003)..
10 A multi-city study analysis for 8 U.S. counties also reported statistically significant
11 associations between PM10 and hospital admissions for cardiovascular diseases among the
12 elderly. An increase of 5% in admissions was associated with a 50 |ig/m3 increase in PM10, with
13 no evidence of confounding with ambient CO (Schwartz, 1999); this study used GAM in the
14 original analyses and results of reanalyses are not available.
15 In an analysis of data from 8 European cities from the APHEA study, associations
16 between PM10 and admissions/visits for all respiratory diseases, asthma or COPD were all
17 positive, though not always statistically significant. The overall effect estimates were 4.5% per
18 50 |ig/m3 increase in PM10 for hospital admissions for all respiratory diseases, 5.1% per 50 |ig/m3
19 increase in PM10 for COPD admissions and 6.1% per 50 |ig/m3 increase in PM10 for asthma
20 hospital admissions (Atkinson et al., 2001). Le Tertre and colleagues (2002; reanalyzed in Le
21 Tertre et al., 2003) reported a 2.5% increase in risk of cardiac hospital admissions per 50 |ig/m3
22 increase in PM10. Several previous APHEA publications (summarized in Table 8B-2 of the draft
23 CD) have reported associations between hospital admissions for respiratory disease or asthma
24 and PM measurements from a variety of methods (e.g., suspended particles, black smoke).
25 These findings are generally consistent with those from U.S. multi-city studies.
August 2003 3-46 Draft - Do Not Cite or Quote
-------
in
is
E
T3
CL
,0
O
ra
to
LU
"5
£
LU
1
t
14
f- I <
111
?
<
<
3
i
<
I
1 J
i
17
(
)
I
)
2
1
19 4
!
0
: 13
: 9 '
8T '
T
:
respi
i
4
(
20 ^«
Y y
' VI1
1
1
i
i
•atory
1Q
i
18
]
5
!
1
[
[
r 1
1
1
!
1
11
1 !
, '
]
I O ° = all respiratory
2 1
i
"6'
i
i
I ;
i
!
1
r
-,
' 90 4 '
Mil J.
t
x =COPD
v = pneumonia
* = asthma
A = all cardiovascular
^ = ischemic heart disease
* = congestive heart disease/heart failure
^ =dysrhythmia
0 =stroke/c
srebrovascular
^ = peripheral circulation
7 4
12 ,,,
I til Ii
M
4 <
° 7 1
31
D! «.
3
3
S 3
3
7
f 4 1
5 .
5 1
%
<
J
Ji
•
cardiovascular
C
C
c
4.
"
C
1
1
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3-7. Effect estimates for PM10 and hospital admissions, emergency room visits (denoted 0) or physicians office visits
(denoted o) for various respiratory and cardiovascular diseases from U.S. and Canadian studies. Study locations are
identified below; multi-city studies denoted by "^. Results of GAM stringent reanalyses; studies not using GAM originally
denoted • (data in Appendix A)
1. Burnett et al., 1997, Toronto •
2. Choudbury et al., 1997, Anchorage o •
3. Delflno et al., 1997, Montreal O •
4. Ito, 2003, Detroit
5. Linn et al., 2000, LA •
6. Lipsett et al., 1997, Santa Clara 0 •
7. Metzger et al., 2003, Atlanta 0 •
8. Moolgavkar, 2003, Cook County •
9. Moolgavkar, 2003, LA County •
10. Morris and Naumova, 1998, Chicago
11. Nauenberg and Basu, 1999, LA •
12. Schwartz and Morris, 1995, Detroit •
13. Schwartz et al., 1996, Cleveland «
14. Schwartz, 1994, Detroit •
15. Sheppard, 2003, Seattle
16. Stieb et al., 2000, St. John 0 •
17. Thurston et al., 1994 Toronto •
18. Tolbert et al., 2000b, Atlanta 0 •
19. Tolbert et al., 2000a, Atlanta 0 •
20. * Zanobetti and Schwartz, 2003,14 U.S. cities
August 2003
5-47
Draft - Do Not Cite or Quote
-------
(/)
° 8 -
^3 CD
TO
'o
I
"w
1 ° "
I
T3
^
IT)
S 0
2 (M
CL
£
O
£
in
O) o
T3
ra
1
t5 ™
Hi
° = all respiratory A = all cardiovascular
x = COPD % = ischemic heart disease
v = pneumonia * = congestive heart disease/heart failure
^ = asthma ^ = dysrhythmia
2
I
<
1
9 I
I 1
10
o 9
3 Q
c
(
i I i
1
1
(
6
i X
;
A
\
4
<
, :
:
> i
8
10
7
i"
i
1
respiratory
E
[
' 1
1
5 I !
< t * •
6 •• ..
! I
i
i
v = stroke/cerebrovascular
^ = peripheral circulation
5
4
i 4
4 T '
!..
.. i
i
5
f
5 4
T T 4
< >
ss s! <. i[
* *
cardiovascular
Q
O
CD
O)
O)
E
o
^
CD
ro
E
LU
1
2
3
4
5
6
7
Figure 3-8. Effect estimates for PM2 5 and hospital admissions or emergency room visits (denoted 0) for various respiratory
and cardiovascular diseases from U.S. and Canadian studies. Study locations are identified below. Results of GAM stringent
reanalyses; studies not using GAM originally denoted • (data in Appendix A)
1. Burnett et al., 1997, Toronto •
2. Delfino et al., 1997, Montreal 0
3. Delfino et al., 1998, Montreal 0
4. Ito, 2003, Detroit
5. Metzger et al., 2003, Atlanta 0
6. Moolgavkar, 2003, LA
7. Sheppard, 2003, Seattle
8. Stieb et al., 2000, St. John 0 •
9. Thurston et al., 1994, Toronto •
10. Tolbert et al., 2000a, Atlanta 0
August 2003
5-48
Draft - Do Not Cite or Quote
-------
in
c
o
£3
ra
'o
o
CO
c
g
'
E
-a
CL
^O
O
ra
-------
1 Results for U.S. and Canadian PM10 studies are shown in Figure 3-7, where it can be seen
2 that nearly all associations between PM10 and admissions/visits for respiratory or cardiovascular
3 causes are positive and many are statistically significant, especially those with greater precision.
4 These results are generally consistent with findings of studies available during the previous
5 review.
6 Less evidence was available on fine particles during the previous review, but the studies
7 that included fine particles or fine particle indicators consistently reported statistically significant
8 associations with hospital admissions for respiratory diseases (EPA, 1996b, Table V-12, p. V-
9 60b). Recent studies have also reported significant associations between PM2 5 and
10 admissions/visits for respiratory diseases (Figure 3-8). In Figure 3-8, it can be seen that
11 associations between PM2 5 and hospitalization or emergency room visits for the general category
12 of respiratory diseases are all positive and statistically significant, while the results for individual
13 disease categories (COPD, pneumonia, and asthma) are less consistent, perhaps due to smaller
14 sample sizes for the specific categories. Associations with the general category of
15 cardiovascular diseases are also all positive and statistically significant or nearly so, but again
16 the results for specific diseases (ischemic heart disease, dysrhythmia, congestive heart disease or
17 heart failure, and stroke) are positive but often not statistically significant.
18 In the previous PM NAAQS review, only one hospital admissions study had included
19 coarse-fraction particles, and it did not report significant associations with hospitalization for
20 respiratory disease (Thurston et al., 1994). However, several new studies show significant
21 associations between PM10_2 5 and admissions/visits for both respiratory and cardiovascular
22 diseases (Figure 3-9). As shown in Figure 3-9, the associations reported between PM10_25 are
23 generally but not always positive, and less often statistically significant than the associations
24 reported with PM25. In the previous review, staff recognized that information about the effects
25 of coarse-fraction particles can also come from studies linking health effects with PM10 in areas
26 where coarse-fraction particles are predominant, and evidence available at that time suggested
27 that aggravation of asthma and respiratory infections and symptoms were associated with daily
28 or episodic increases in PM10 dominated by coarse-fraction particles (62 FR 38677). Recent
29 studies conducted in coarse fraction-dominated areas such as Las Vegas, Tucson, Utah Valley,
30 Phoenix, Southern California, and Anchorage have also reported associations between PM10 and
August 2003 3-50 Draft - Do Not Cite or Quote
-------
1 adverse health effects. The effects included in these studies include associations between PM10
2 and medical visits for asthma in Anchorage, AK (Choudhury et al., 1997) and hospital
3 admissions for COPD in Las Vegas (Chen et al., 2000) and for cardiovascular disease in Tucson
4 (Schwartz, 1997) (the last two studies were analyzed using GAM and were not reanalyzed).
5 However, in Maricopa County, AZ (Phoenix), Moolgavkar (2000 a,b,c; 2003) did not report
6 significant associations between PM10 and hospital admissions for COPD or cardiovascular
7 diseases. Overall, the recent studies continue to show associations between PM10 and respiratory
8 morbidity in areas where coarse-fraction particles predominate, and there is evidence suggestive
9 of cardiovascular morbidity as well.
10 Figures 3-7 through 3-9 present effect estimates from single-pollutant models. As was
11 found for mortality studies, the multi-city analyses of hospital admissions have not found
12 evidence of significant confounding by copollutant gases. In single-city studies, a number of
13 investigators evaluated the effects of gaseous co-pollutants independently and in multi-pollutant
14 models with PM. As discussed in further detail in section 3.5.1, some gaseous pollutants have
15 been reported to have independent effects on the respiratory system and might be expected to act
16 as confounders in PM-admissions/visits associations. For example, several studies have
17 indicated that O3 is associated with increased admission/visits for respiratory diseases, such as
18 asthma, as has been summarized in the Air Quality Criteria for Ozone (EPA, 1996c). A number
19 of the studies included in Tables 8-19 and 8-22 of the draft CD report significant associations
20 with O3. In some of these studies, PM effect estimates were reduced in two-pollutant models
21 with O3 (e.g., Tolbert et al., 2000b; Delfino et al., 1998), but in others, PM associations were
22 generally reported to be robust to inclusion of O3 in the models (e.g., Lippmann et al., 2000,
23 reanalyzed in Ito, 2003; Gwynn et al., 2000, not reanalyzed; Burnett et al., 1997) and less
24 evidence was found for potential confounding by other gaseous pollutants (results discussed in
25 more detail in Table 8B-2 of the draft CD). In considering studies of cardiovascular
26 admissions/visits, the draft CD focused on CO as a copollutant of interest, due to the known
27 effects of CO on the cardiovascular system (EPA, 1999). The draft CD finds that "[t]he above
28 analyses of daily PM10 and CO in U.S. cities, overall, suggest that elevated concentrations of
29 both PM10 and CO may enhance risk of cardiovascular (CVD)-related morbidity leading to acute
30 hospitalizations" (CD, p. 8-130). In studies of cardiovascular and chronic respiratory disease
August 2003 3-51 Draft - Do Not Cite or Quote
-------
1 admissions/visits, Moolgavkar (2000b,c, reanalyzed in Moolgavkar, 2003) reports that
2 associations with PM10 and PM2 5 were dramatically reduced with the inclusion of either CO or
3 NO2 (differs by location and health endpoint) in the models. For cardiovascular
4 admissions/visits (but equally true for respiratory diseases) the draft CD concludes: "In some
5 studies, PM clearly carries an independent association after controlling for gaseous co-pollutants.
6 In others, the 'PM effects' are markedly reduced once co-pollutants are added to the model; but
7 this may in part be due to both PM and co-pollutants and/or gaseous pollutant (e.g., CO) having
8 independent effects on cardiovascular function" (CD, p. 8-147).
9 Summary. PM is associated with admissions/visits for respiratory and cardiovascular
10 diseases and specific disease categories, including ischemic heart disease, asthma, COPD, and
11 pneumonia; the findings are generally consistent with those reported in the 1996 CD. Positive,
12 often statistically significant associations for both respiratory and cardiovascular
13 admissions/visits are seen with PM10 and PM2 5, and positive, but less frequently statistically
14 significant, associations are found with PM10_25. As is also the case for mortality studies, U.S.
15 multi-city studies (Samet et al., 2000a,b, reanalyzed in Domenici, 2003; Schwartz, 1999, not
16 reanalyzed) likely provide the most precise estimates for relationships of U.S. ambient PM10
17 exposure to increased risk for hospitalization, and in these studies increases of 5% in hospital
18 admissions for cardiovascular disease and 8% and 6% in hospital admissions for COPD or
19 pneumonia (respectively) per 50 |ig/m3 increase in PM10 were reported (CD Table 8-18, Table 8-
20 19). As stated previously, in single-city studies, effect estimates for cardiovascular admissions
21 generally range from about 1% to 10% per 25 |ig/m3 PM25 or PM10.25 (CD, p. 8-231), and effect
22 estimates for respiratory admissions generally range from about 5% to 25% per 25 |ig/m3 PM2 5
23 or PM10_2 5 Taken together, the findings of new studies and those reviewed in the 1996 CD offer
24 reasonably consistent evidence for associations between ambient PM concentrations, across PM
25 indicators, and admissions/visits to the hospital or emergency room for respiratory or
26 cardiovascular diseases.
August 2003 3-52 Draft - Do Not Cite or Quote
-------
1 3.3.2.2 Effects on the Respiratory System
2 Evidence available in the previous review suggested associations between PM exposure
3 and effects on the respiratory system such as changes in lung function, increases in respiratory
4 symptoms or disease, as well as related morbidity indices such as school absences, lost work
5 days and restricted activity days (EPA, 1996b, pp. V-21 and V-22). From epidemiologic or
6 controlled human exposure studies of short-term PM exposure, it was reported that sensitive
7 individuals (especially those with asthma or pre-existing respiratory symptoms) may have
8 increased or aggravated symptoms, with or without reduced lung function (EPA, 1996b, p. V-
9 23). Long-term (months to years) exposure to PM was linked with decreased lung function and
10 increased incidence of respiratory diseases such as bronchitis (EPA, 1996b, p. V-26). The
11 results of studies using long-term and short-term PM exposure data were reported to be
12 consistent with one another. In addition, toxicologic studies using particles such as acid
13 aerosols, ROFA or PM components, generally at high concentrations, and autopsy studies of
14 humans and animals reported evidence of pulmonary effects, including morphological damage
15 (e.g., changes in cellular structure of the airways), and changes in resistance to infection.
16 Recently published studies summarized in the draft CD have included epidemiologic
17 studies of lung function or symptoms and new morbidity indicators such as physician visits, for
18 both short-term and long-term exposures to PM. Also, recent toxicological or controlled human
19 exposure studies have used exposures to ambient PM (e.g., inhalation exposures to CAPs or
20 intratracheal instillation of ambient PM samples) to assess responses such as lung injury or
21 inflammation. These studies provide additional new evidence linking PM with respiratory
22 effects. Among the many new epidemiologic studies are several assessing relationships between
23 PM and additional health endpoints, including physicians' office visits. A number have
24 evaluated effects on lung function or respiratory symptoms, while few new studies have assessed
25 effects such as school absences or work loss days, which are indirect measures that may be
26 linked with respiratory illness.
27 3.3.2.2.1 Acute Respiratory Effects
28 Epidemiologic Studies. Among the new epidemiologic studies are several using medical
29 visits for respiratory illness as a measure of health effects, including Choudhury et al (1997), in
30 Anchorage AK (shown in Figure 3-7) and a number of studies conducted outside the U.S. or
August 2003 3-53 Draft - Do Not Cite or Quote
-------
1 Canada (CD, p. 8-164). These studies have evaluated effects of pollutant exposure on visits to
2 physicians' offices or doctors' visits to patients. Visits for asthma were significantly increased
3 with PM exposure in children and people of all ages, and significant associations were found
4 with visits for lower respiratory diseases in children (CD, p. 8-164).
5 The draft CD notes that these studies "provide new insight into the fact that there is a
6 broader scope of severe morbidity associated with PM air pollution exposure than previously
7 documented" (CD, p. 8-165). These studies find associations that range widely up to 60%
8 increases in medical visits with a 50 |ig/m3 change in PM10 (CD Table 8-22). The results of
9 these studies offer a link between the more severe endpoints, such as increased mortality and
10 hospital admissions or emergency room visits for respiratory diseases, and moderate effects such
11 as respiratory symptoms and decreased lung function. These new studies also indicate the
12 potentially more widespread public health impact of less severe respiratory health endpoints
13 (CD, p. 8-168).
14 New epidemiologic studies on PM-related effects on respiratory symptoms or lung
15 function are summarized in draft CD Tables 8-23 through 8-33; the studies are grouped by health
16 status of the study subjects (asthmatic or nonasthmatic), type of effects (e.g., symptoms, lung
17 function changes) and PM exposure (short- and long-term). Only a few recent North American
18 publications are available; the results for U.S. and Canadian studies using gravimetric PM data
19 are included in Appendix A. Most U.S. and Canadian studies used gravimetric PM data,
20 generally PM10 and sometimes PM2 5 and PM10_2 5, and most were studies using children.
21 For asthmatic subjects, associations were reported between PM10 or PM2 5 decreases in
22 lung function measures (e.g., decreased peak expiratory flow rate); some but not all of the
23 associations reached statistical significance. In addition, positive associations were reported
24 between PM10 or PM25 and one or more of a range of respiratory symptoms (e.g., cough, wheeze,
25 shortness of breath) but the findings were less consistent than those for lung function (CD, p. 8-
26 170). In studies of nonasthmatic subjects, inconsistent results were reported for changes in lung
27 function, while studies generally reported increases in respiratory symptoms that were often not
28 statistically significant. Generally similar results were found for both PM10 or PM2 5 (CD, p. 8-
29 180).
August 2003 3-54 Draft - Do Not Cite or Quote
-------
1 Few studies have included both PM2 5 and PM10_2 5 data. In those studies, the findings
2 suggest roles for both fine- and coarse-fraction PM in reduced lung function and increased
3 respiratory symptoms (CD, p. 9-111). For example, using data from the Six Cities study, lower
4 respiratory symptoms were found to be significantly increased for children with PM2 5 but not
5 with PM10_2 5, while the reverse was true for cough (Schwartz and Neas, 2000). When both PM2 5
6 and PM10_2 5 were included in models, the effect estimates were reduced for each, but PM2 5
7 retained significance in the association with lower respiratory symptoms and PM10_2 5 retained
8 significance in the association with cough. In the last review, several studies reported significant
9 associations between symptoms or lung function changes with PM10 and fine particles or fine
10 particle surrogates, but no data were available for coarse-fraction particles (EPA 1996b, Table
11 V-12). The new studies continue to show effects of short-term exposure to PM10 and PM2 5 and
12 offer additional evidence for associations between PM10_2 5 and respiratory morbidity.
13 The newly available epidemiologic studies include some that used ambient PM
14 concentrations averaged over 1- or 8-hours, in addition to 24-hour averages. Two panel studies
15 of symptoms in asthmatic subjects are summarized in the CD, and one reported larger effect
16 estimates for 1- or 8-hour concentrations than for 24-hour PM10 concentrations (Delfmo et al.,
17 1998) while the other reported larger effect estimates for 24-hour PM10 concentrations (Ostro et
18 al., 2001) (CD, p. 8-179).
19 Considering also results from studies conducted outside the U.S. and Canada, the draft
20 CD finds evidence supporting increases in respiratory symptoms associated with short-term
21 exposures to PM for both asthmatic and nonasthmatic subjects, though many associations did not
22 reach statistical significance. Again, considering the full body of literature, short-term PM
23 exposure was associated with decreases in lung function (e.g., peak expiratory flow rate) in
24 studies of asthmatics but little evidence was reported for associations between lung function and
25 short-term PM exposure in nonasthmatic subjects.
26 Laboratory Studies. Key toxicologic or controlled human exposure studies summarized
27 in the draft CD include: (1) exposures of human volunteers in a clinical setting to CAPs; (2)
28 animal studies with exposure to ambient PM by inhalation of CAPs or intratracheal installation
29 of ambient PM samples; and (3) in vitro exposures to ambient particles using cells from the
30 respiratory system (e.g., bronchial epithelial cells, macrophages). The principal effects studied
August 2003 3-55 Draft - Do Not Cite or Quote
-------
1 have been inflammatory responses and other indicators of lung injury. As has often been the
2 case, many of the new studies have used high doses (e.g., doses in the hundreds of jig to mg), but
3 some new CAPs studies have reported effects with doses that are closer to the range of ambient
4 PM concentrations.
5 Numerous studies have reported inflammatory changes with exposure to particles,
6 including changes in levels or increased release of cytokines, or chemicals released as part of the
7 inflammatory process (e.g., interleukins such as IL-8) (CD Tables 7-1, 7-2 and 7-3). A variety of
8 particles have been used in these studies, with many involving exposure to ROFA or other
9 combustion-related particles, but some using inhalation of CAPs or instillation of ambient
10 particles.
11 Although less evidence is available from studies using ambient particle exposures, Costa
12 and Dreher (1997) summarized evidence from studies showing increased inflammatory cell
13 counts with exposure to ambient particles collected in U.S., Canadian, and German cities, and
14 Brain et al. (1998) showed that similar levels of acute inflammatory injury were caused by urban
15 air particles and Kuwaiti oil fire particles (on an equal mass basis). One new controlled human
16 exposure study also reported evidence of inflammatory changes in the lung with exposure to
17 CAPs (Ohio et al., 2000). In addition, Ohio and Devlin (2001) exposed human volunteers by
18 intratracheal instillation to ambient PM that had been collected in Utah Valley before, during,
19 and after the time period when a major local PM source, a steel mill had been shut down. A
20 greater inflammatory response was found with exposure to particles collected while the plant
21 was operating than during the period when the steel mill was closed. Similar results were
22 reported in animal toxicologic studies using PM from Utah Valley filters (Dye et al., 2001;
23 Molinelli et al., 2002). The draft CD concludes that these results indicate that the pulmonary
24 effects observed after experimental exposure of humans to the Utah Valley PM are in good
25 accord with epidemiologic reports of adverse respiratory health effects in Utah Valley residents
26 (CD, p. 7-18). The types of effects reported included increases in neutrophils, eosinophils or
27 alveolar macrophages (either total number or percent) and expression of cytokines such as
28 interleukin-6 (CD, Tables 7-2 and 7-3).
29 A number of animal studies have shown that exposure to diesel exhaust particles can
30 increase the production or release of inflammatory cells, such as eosinophils, and may
August 2003 3-56 Draft - Do Not Cite or Quote
-------
1 exacerbate the allergic response. Controlled exposures of humans to diesel exhaust particles also
2 have resulted in increases in inflammatory cells indicative of enhanced response to allergens.
3 Together, the human and animal studies provide evidence that particle exposures can produce
4 inflammatory changes in the respiratory system (CD, p. 7-120).
5 Animal studies also have reported evidence of general lung injury, including increased
6 protein levels in lung fluids with exposure to ambient particles (CD Table 7-2) or combustion-
7 related particles such as ROFA (CD, Table 7-3). One general cause of lung cell injury may be
8 the production of reactive oxidant species that can damage the epithelial cells in the lung; these
9 chemicals can be produced as part of an inflammatory response to particle exposure. In in vitro
10 experiments, ambient PM exposures were reported to have effects that included increased release
11 of inflammatory chemicals, evidence of oxidant stress on the cells, and evidence of general
12 cellular toxicity (e.g., release of proteins) (CD Table 7-10). As discussed previously, some
13 toxicological studies have used ambient particles collected in Utah Valley over a time period that
14 included closure of a major PM source, and these studies have reported evidence of general lung
15 injury as well as inflammation (CD, pp. 7-17 and 7-18). Several in vitro studies have also
16 reported evidence of increased oxidative stress in lung cell cultures exposed to particles collected
17 in Utah Valley; notably, the particle doses used in these studies were only 2-3-fold greater than
18 generally estimated doses for humans breathing ambient air (Ohio et al., 1999a,b; Soukup et al.,
19 2000). In two of these studies, the transition metal content of the particles appeared to be more
20 closely linked to reported effects than the quantity of particles (Ohio et al., 1999a,b). Soukup
21 and colleagues (2000) also tested the effects of particles collected in Utah Valley, and found
22 evidence of oxidant activity with particles collected at times when a major industrial PM source
23 was in operation, but not when the industrial source was shut down.
24 The draft CD concludes that studies using "intratracheal instillation of ambient PM
25 obtained from specific ambient sources have shown clearly that PM can cause lung inflammation
26 and injury" (CD, p. 7-16). Findings of inflammation and lung injury are generally consistent
27 with epidemiologic results showing increases in respiratory symptoms or exacerbation of
28 respiratory diseases. The 1996 CD observed that impairment of pulmonary host defense
29 mechanisms by acidic particles was consistent with observations of increased prevalence of
30 bronchitis in communities with higher levels of acidic PM (EPA 1996a, p. 13-75).
August 2003 3-57 Draft - Do Not Cite or Quote
-------
1 Summary. The recent epidemiologic findings are consistent with those of the previous
2 review in showing associations with both respiratory symptom incidence and decreased lung
3 function. For asthmatic subjects, PM10 or PM2 5 were associated with small decreases in lung
4 function and increases in respiratory symptoms, though the associations were not always
5 statistically significant. The findings from studies of physicians' office visits for respiratory
6 diseases offer new evidence of acute respiratory effects with exposure to ambient PM that is
7 coherent with evidence of increased respiratory symptoms and admissions/visits to the hospital
8 or emergency room for respiratory disease. While urging caution in interpreting the findings of
9 the toxicologic studies where higher doses were used, the draft CD concludes that "[t]he fact that
10 instillation of ambient PM collected from different geographical areas and from a variety of
11 emission sources consistently caused pulmonary inflammation and injury tends to corroborate
12 epidemiologic studies that report increased respiratory morbidity and mortality associated with
13 PM in many different geographical areas and climates." (CD, p. 7-19).
14 3.3.2.2.2 Chronic Respiratory Effects
15 In the 1996 CD, only a few epidemiologic studies had assessed associations between
16 long-term PM exposure and lung function changes or respiratory symptoms. Among U.S. and
17 Canadian studies, studies using data from the Six Cities and 24-Cities cohorts suggested
18 associations between fine particles and find particle indicators (e.g., acid aerosols) and chronic
19 bronchitis and decreased lung function in children (CD, p. 6-205). For adults, the 1996 CD
20 summarized the results of several cross-sectional studies as well as one cohort study
21 (AHSMOG), and found evidence for increased incidence of respiratory diseases, especially
22 bronchitis, with long-term PM exposure (EPA, 1996a, p. 12-197). Overall, significant
23 associations were observed between decreased lung function or increased incidence of bronchitis
24 in children with fine particles or fine particle surrogates, with less evidence for associations with
25 PM10.25, PM10, PM15 or TSP (EPA, 1996b, Table V-13).
26 Several new epidemiologic analyses have been conducted on long-term pollutant
27 exposure effects on respiratory symptoms or lung function in the U.S.; numerous European,
28 Asian, and Australian studies have also been published. Little new evidence is available from
29 toxicologic or controlled human exposure studies regarding long-term effects of PM exposure.
30 The new U.S. epidemiologic studies are based on data from two main cohort studies, a study of
August 2003 3-58 Draft - Do Not Cite or Quote
-------
1 schoolchildren in 12 Southern California Communities and an adult cohort of Seventh Day
2 Adventists (AHSMOG). Results for the new studies, together with the results of the Six Cities
3 and 24-Cities studies, are presented in Table 3-4.
4 As seen in Table 3-4, initial analysis of data from the 12 Southern California
5 Communities children's cohort show significant associations between long-term exposure to
6 PM10 and incidence of bronchitis or increased production of phlegm among the subgroup of
7 children with asthma, though no significant associations were found for the subgroups of
8 children without asthma (Peters et al., 1999a; McConnell et al., 2000). After a four-year follow-
9 up period for 4th grade children, decreases in lung function growth were associated with
10 increasing exposure to PM10, PM25, and PM10_2 5, with generally similar findings for both fme-
11 and coarse-fraction particles. Two-pollutant models were tested in this study, and the effect
12 estimates for the various PM indices, NO2 and acid vapor were generally reduced in size. The
13 authors observe that motor vehicle emissions are a major source of ambient particles, NO2 and
14 inorganic acids and thus they were unable to identify the independent effects of each pollutant
15 (Gauderman et al., 2000, p. 1388). Analysis of data from a second group of children recruited as
16 4th-graders showed decreases in lung function growth with long-term exposure to PM10 and
17 PM2 5, but the results were generally not statistically significant (Gauderman et al., 2002). Avol
18 et al. (2001) also reported increased lung function growth among cohort participants who moved
19 from areas with higher to lower PM10 concentrations during the study period, compared with
20 participants who remained in high pollution areas. These findings are generally consistent with
21 those from short-term exposure studies where respiratory morbidity is associated with PM10
22 PM2 5 and PM10.2 5.
23 Further analyses have been done for the AHSMOG adult cohort, and significant
24 decreases in lung function were reported only for the subgroup of males with a family history of
25 lung disease, but not for other subgroups of the cohort. In two-pollutant models, the coefficients
26 for PM10 and sulfates were found to remain unchanged or increase in size, while O3 and SO2
27 were reduced and lost statistical significance.
28 A number of long-term studies of respiratory effects have also been conducted in non-
29 North American countries, and many report significant associations between indicators of long-
30 term PM exposure and either decreases in lung function or increased respiratory disease
August 2003 3-59 Draft - Do Not Cite or Quote
-------
1
2
3
4
5
6
7
prevalence (summarized in Table 8-B8 of the draft CD). The results of U.S. and Canadian
studies, presented in Table 3-4, generally show increased symptoms and decreased lung function
with increases in PM exposure. These new findings, along with those from non-North American
studies, are generally consistent with those of the previous review as well as with findings of
associations between short-term PM exposure and increased respiratory symptoms or decreased
lung function.
Table 3.4 Effect Estimates for Morbidity per Increments in Long-term Mean Levels of
Fine- and Coarse-fraction Particle Indicators from U.S. and Canadian Studies (from draft
CD Table 9-11)
Type of Health Effect Indicator (Increment)
Study
Odds Ratio (95% CI)
Range of City
PM Levels **
Means (ug/m3)
Increased Bronchitis in Children
Six CityA PM15/10 (20 ug/m3)
PM25 (10 ug/m3)
24 CityB SO; (15 ug/m3)
PM21(10ug/m3)
PM10 (20 ug/m3)
AHSMOG0 SOJ (15 ug/m3)
12 Southern California PM10 (20 ug/m3)
communities13 (1986-1990 data)
(all children)
12 Southern California PM10 (20 ug/m3)
communitiesE 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)
communities13 (1986-1990 data)
(all children)
12 Southern California PM10 (20 ug/m3)
communitiesE 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
AHSMOGF PM10 (20 ug/m3)
1.19(0.84, 1.68)
NR
Decreased Lung Function in Children
Six City
A
PM15/10 (50 ug/m3)
NS Changes*
NR (20, 59)
August 2003
3-60
Draft - Do Not Cite or Quote
-------
Type of Health Effect
Study
24 City0
12 Southern California
communities11
(all children)
12 Southern California
communities11
(all children)
12 Southern California
communities1
(4th grade cohort)
12 Southern California
communities1
(4th grade cohort)
12 Southern California
communities1
(second 4th grade cohort)
12 Southern California
communities1
(second 4th grade cohort)
12 Southern California
communities1
(second 4th grade cohort)
12 Southern California
communitiesK
12 Southern California
communitiesK
12 Southern California
communities14
Lung Function Changes
AHSMOGL (% predicted
FEVb females)
AHSMOGL (% predicted
FEVb males)
Indicator (Increment)
SOI (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(10ug/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)
PM10 (20 ug/m3)
in Adults
PM10 (cutoff of
54.2 days/year
>100 ug/m3)
PM10 (cutoff of
54.2 days/year
>100 ug/m3)
Odds Ratio (95% CI)
-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
-70 (-120, -20) PEFR growth
+0.9 % (-0.8, 2.5) FEVj
+0.3 % (-2.2, 2.8) FEVj
Range of City
PM Levels **
Means (ug/m3)
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)***
NR(10, 35)***
NR
NR(15, 70)***
NR(10, 35)***
NR
NR(10, 80)****
NR(5, 30)****
NR(10, 80)****
NR(5, 30)****
NR(10, 80)****
NR(5, 30)****
NR (15.0, 66.2)
NR (15.0, 66.2)
NR (15.0, 66.2)
52.7 (21.3, 80.6)
54.1(20.0,80.6)
August 2003
3-61
Draft - Do Not Cite or Quote
-------
Type of Health Effect
Study
Indicator (Increment)
Odds Ratio (95% CI)
Range of City
PM Levels **
Means (ug/m3)
AHSMOGL (% predicted PM10 (cutoff of
FEVb males whose parents 54.2 days/year
had asthma, bronchitis, >100 ug/m3)
emphysema)
AHSMOGL (% predicted
FEV,, males)
; (1.6 ug/m3)
-7.2 % (-11.5,-2.7) FEVj
-1.5% (-2.9, -0.1) FEVJ
54.1(20.0,80.6)
7.3 (2.0, 10.1)
* NS Changes = No significant changes (no quantitative results reported).
** Presented as overall study mean (min, max), unless indicated as mean (±SD); NR=not reported.
*** Estimated from Figure 1, Gauderman et al. (2000)
****Estimated from figures available in online data supplement to Gauderman et al. (2002)
References:
A Dockery et al. (1989)
B Dockery et al. (1996)
c Abbey etal. (1995a,b,c)
D Peters etal. (1999a)
EMcConnelletal. (1999)
FBerglundetal. (1999)
G Raizenne et al. (1996)
H Peters etal. (1999b)
1 Gauderman et al. (2000)
J Gauderman et al. (2002)
KAvoletal. (2001)
L Abbey etal. (1998)
1 3.3.2.3 Effects on the Cardiovascular System
2 In the last review, evidence was available from a few epidemiologic studies indicating
3 that PM was associated with increased mortality and hospital admissions for cardiovascular
4 diseases. These findings inspired further research so that a significantly expanded body of
5 evidence is available in this review from epidemiologic, toxicologic, and controlled human
6 exposure studies. As described in the previous section, many new epidemiologic studies have
7 continued to show relationships between mortality and hospital admissions or emergency
8 department visits for cardiovascular diseases. Numerous new controlled human exposure,
9 toxicological and epidemiologic studies have provided evidence on cardiovascular health
10 endpoints, such as changes in cardiac function and changes in blood components or
11 characteristics, that is building new understanding of how exposure to ambient particles could
12 affect the cardiovascular system.
13 Measures of changes in cardiac function included in these new studies include
14 arrhythmia, alterations in electrocardiogram (ECG) patterns, heart rate or heart rate variability
15 changes, and incidence of myocardial infarction. One recent epidemiologic study reported an
August 2003
3-62
Draft - Do Not Cite or Quote
-------
1 association between the occurrence of nonfatal myocardial infarction (or a heart attack) and
2 ambient fine particle concentrations averaged over the 2 hours preceding the event (Peters et al.,
3 2001). In addition, Peters et al. (2000) used data on discharge frequency from implanted cardiac
4 defibrillators (discharges occur when the patient is experiencing cardiac arrythmia) and reported
5 generally positive associations with several PM indicators, including statistically significant
6 associations with PM2 5. As summarized in draft CD Table 7-7, animal studies have also shown
7 evidence of arrythmia or alterations in ECG patterns that included increases in the S-T peak of
8 the heart beat pattern; this type of change indicates effects on repolarization of the heart which is
9 linked with increased risk of sudden death (CD, p. 9-63).
10 Heart rate variability changes have been reported in a number of epidemiologic studies.
11 In several studies, tests of cardiac function (e.g., heart rate, heart rate variability) were done
12 repeatedly for panels of elderly people over a period of several weeks. Low heart rate variability
13 has been implicated as a predictor of increased cardiovascular morbidity and mortality (CD, p. 9-
14 62). Most, but not all, new studies reported decreases in several measures of heart rate
15 variability with increased PM10 and PM25 (CD, p. 8-143). While many epidemiology studies
16 have used 24-hour average PM measurements, two new studies have linked effects with PM
17 averaged over 4-hours. Gold et al., 2000 (reanalyzed in Gold et al., 2003) reported significant
18 decreases in heart rate variability with 4-hour average PM2 5 but not with PM10_2 5; Magari et al.
19 (2001) also reported decreased heart rate variability in 4-hour PM25 concentrations. The
20 findings on changes in heart rate are less consistent than those for heart rate variability, with
21 some studies reporting significant increases in heart rate with PM exposure, but one study
22 reporting an association with decreased heart rate (CD, p. 8-143).
23 Decreased heart rate was also reported in an animal study using intratracheal installation
24 of urban PM (but not with Mt. St. Helens volcanic ash) (Watkinson et al., 2000). In a study
25 using rats and hamsters, no effects were reported in hamsters, but increased heart rate and blood
26 cell differential counts were reported in rats (Gordon et al., 2000). Changes in heart rate or heart
27 rate variability generally indicate effects on the autonomic nervous system, and the draft CD
28 concludes "There is now ample evidence that inhaled particles can affect the heart through the
29 [autonomic nervous system]" (CD, p. 9-61).
August 2003 3-63 Draft - Do Not Cite or Quote
-------
1 Some studies have reported increases in blood components or characteristics. In
2 epidemiologic studies, exposure to particles has resulted in increased levels of C-reactive
3 protein, which is considered an important indicator of inflammation, tissue damage and
4 infection, and related to increased risk of cardiac events (CD, p. 9-65). Fibrinogen is a blood
5 clotting factor released in inflammatory processes; it has been reported to be a risk factor for
6 ischemic heart disease and cerebrovascular disease, and it contributes to blood plasma viscosity
7 (Gardner et al., 2000). In humans exposed to PM, fibrinogen levels were increased in some, but
8 not all, studies, and plasma viscosity was increased with PM exposure in one European study
9 (CD, pp. 8-143 to 8-145). One new controlled human exposure study reported mild increases in
10 neutrophils in bronchoalveolar lavage samples and increased blood fibrinogen levels after 2-hour
11 exposure to concentrated ambient PM2 5 (Ohio et al., 2000). Finally, changes in blood cell
12 counts, including platelets or red blood cells, was reported in some epidemiologic studies (CD, p.
13 8-145). A number of toxicologic studies have also reported such hemolytic effects as changes in
14 blood factors such as hemoglobin levels or platelet counts (CD, Table 7-7).
15 Summary. The draft CD concludes that". . .animal studies, to date, have provided
16 evidence indicating that high concentrations of inhaled or instilled particles can have systemic,
17 especially cardiovascular, effects." (CD, p. 7-52). A more detailed discussion of the potential
18 mechanisms or pathways by which PM could be linked with effects on the cardiovascular system
19 is included in draft CD section 9.7.3, where the draft CD concludes "[tjaken as a whole, these
20 studies are difficult to interpret but clearly indicate that PM can affect the circulatory system.
21 However, a complete understanding of the pathways by which very small concentrations of
22 inhaled ambient PM can produce vascular changes that can contribute to increased
23 mortality/morbidity remains to be more fully elucidated." (CD, p. 9-66).
24 Regarding the epidemiologic studies, the draft CD concludes: "[t]he above findings add
25 support for some intriguing hypotheses regarding possible mechanisms by which PM exposure
26 may be linked with adverse cardiac outcomes. They are interesting in implicating both increased
27 blood viscosity and C-reactive protein, a biological marker of inflammatory responses thought to
28 be predictive of increased risk for serious cardiac events" (CD, p. 8-145). The results of recent
29 toxicologic studies, epidemiologic panel studies, and controlled human exposure studies
30 generally provide coherence with the findings from community health studies in finding
August 2003 3-64 Draft - Do Not Cite or Quote
-------
1 associations with increased heart rate, decreased heart rate variability, increases in inflammatory
2 substances such as C-reactive protein, and in plasma viscosity or blood fibrinogen levels (CD, p.
3 9-127).
4 3.3.2.4 Developmental Effects
5 Some new evidence is available that is suggestive of adverse effects of air pollution on
6 prenatal development. Several recent studies have shown significant associations between PM10
7 concentration averaged over a month or a trimester of gestation and risk of IUGR (intrauterine
8 growth reduction) and low birth weight (CD, pp. 8-112, 8-113). However, when gaseous co-
9 pollutants were included in the studies, it was sometimes difficult to separate the effects of the
10 pollutants. The CD concludes that these effects are emerging as potentially more important than
11 was appreciated in the 1996 CD but the evidence is still limited regarding these effects (CD, pp.
12 9-136 and 9-137).
13
14 3.4 SENSITIVE GROUPS FOR PM-RELATED HEALTH EFFECTS
15 In general, subpopulations that have been identified in previous PM NAAQS reviews as
16 being potentially more sensitive to the adverse health effects of PM have included individuals
17 with pre-existing respiratory and cardiovascular disease, the elderly and children (EPA 1996b,
18 pp. V-33 to V-36). The draft CD (sections 9.9.2 and 9.9.3) summarizes the results of new
19 studies as they relate to two major at-risk categories: (1) persons with preexisting disease; and
20 (2) age-related population groups. The new studies continue to support consideration of people
21 with preexisting diseases, the elderly and children as potentially sensitive to PM; new evidence
22 also suggests that infants are a potentially sensitive group, including evidence of development
23 effects during the prenatal period (CD, p. 9-136). Some studies have suggested that other
24 factors, such as socioeconomic status, race, or gender may play a role in susceptibility to PM-
25 related effects, but only limited and inconsistent evidence on such groups is available.
26
27 3.4.1 Preexisting disease as a risk factor
28 Numerous epidemiologic studies have identified individuals with cardiorespiratory
29 diseases (e.g., asthma, COPD) as being at greater risk for adverse effects with PM exposure (CD,
30 p. 9-73). New information from studies of cardiovascular health measures such as plasma
August 2003 3-65 Draft - Do Not Cite or Quote
-------
1 viscosity or changes in heart rate or heart rate variability provides additional support for
2 consideration of persons with cardiovascular disease as being susceptible to the PM-related
3 effects (CD, p. 9-76).
4 Several new epidemiologic studies have evaluated associations between PM and
5 mortality or hospital admissions for subpopulations with existing heart or lung disease groups,
6 also including diabetes. Associations for PM10 with hospital admissions for cardiovascular
7 diseases were larger in the subgroup of people with diabetes (Zanobetti and Schwartz, 2001). In
8 addition, larger effect estimates for hospital admissions for cardiovascular diseases were reported
9 in people with acute respiratory infections, and for hospital admissions for pneumonia in people
10 with asthma, and for hospital admissions for COPD in people with heart failure (Zanobetti et al.,
11 2000). Staff notes that both analyses used GAM and have not been reanalyzed; though in
12 reanalyses of other studies using the same data, there was little effect on the results (Zanobetti
13 and Schwartz, 2003). One new study also indicated in original analyses that risk of mortality for
14 nonaccidental causes was increased in subpopulations with diabetes, heart or lung diseases, but
15 when reanalyzed, the effect estimates generally lost statistical significance (Goldberg et al.,
16 2000; Goldberg and Burnett, 2003).
17 Asthma has been of particular public interest as a respiratory condition that may lead to
18 sensitivity to air pollution effects. Included in Appendix A are numerous epidemiologic studies
19 reporting increased medical visits for asthma with exposure to PM10, PM25 or PM10_2 5, and most
20 studies reported significant associations. In considering asthmatics as a susceptible
21 subpopulation, the results for studies evaluating changes in lung function and respiratory
22 symptoms were evaluated separately for asthmatic and nonasthmatic subjects. In the studies of
23 lung function or symptoms, asthmatic subjects generally had greater reduction in pulmonary
24 function with PM exposure, but both asthmatic and non-asthmatic subjects had similar responses
25 in respiratory symptom studies. A number of toxicologic studies have evaluated the effects of
26 particles or surrogate particles on allergic diseases, including allergic asthma, and the draft CD
27 finds that "[tjhese studies provide biological plausibility for the exacerbation of allergic asthma
28 associated with episodic exposure to PM" (CD, p. 7-58).
29 New dosimetry studies have shown that, among people with small airways disease such
30 as COPD, airflow may be unevenly distributed due to airway obstruction, and there can be
August 2003 3-66 Draft - Do Not Cite or Quote
-------
1 closure of small airways, resulting in deeper penetration of particles in the better ventilated
2 regions, or increased local deposition of particles. In addition, ventilation rate and rate of air
3 flow are often increased with airway obstruction. The findings of these studies suggest that total
4 lung deposition generally is increased with obstructed airways, regardless of deposition
5 distribution between the tracheobronchial or alveolar regions (CD, pp. 6-34).
6 A number of animal models of susceptible populations have been used in toxicologic
7 studies examining PM. These models include: monocrotaline treatment of rats as a model of
8 cardiorespiratory disease; SO2-induced chronic bronchitis in rats; ovalbumin sensitization in
9 rodents as a model of airway hyperresponsiveness; and genetically predisposed animals such as
10 the spontaneously hypertensive rat. The advantages and disadvantages of these animal models
11 are discussed more fully in section 7.4 of the draft CD. In general, these studies reported greater
12 effects (including lung function changes, evidence of lung injury, or immunological changes) in
13 the animal models of susceptibility. While recognizing that further research is needed, the draft
14 CD concludes that these studies "do provide evidence of enhanced susceptibility to inhaled PM
15 in 'compromised' hosts" (CD, p. 7-50). In addition, a number of new studies, though using
16 animal models and high particle doses, have suggested that genetic susceptibility may play a role
17 in differential responses to inhaled particles across a population (CD, p. 7-52).
18 3.4.2 Age-related risk factors
19 In the previous review, numerous studies indicated that the elderly and children are more
20 susceptible to PM-related health effects (EPA, 1996a, p. 12-364). Similarly, in reviewing the
21 recent studies of PM-related medical visits or admissions/visits for respiratory diseases, the draft
22 CD finds that the groups identified as being most strongly affected by PM are older adults,
23 children and infants (CD, p. 9-135).
24 Evidence from dosimetry studies has not identified elderly adults to be at increased risk
25 due to difference in lung deposition, clearance or retention of inhaled particles associated with
26 aging, per se (CD, p. 6-103), and limited evidence is available from toxicologic studies on older
27 animal models. However, numerous epidemiologic studies have reported stronger and more
28 consistent associations between mortality or morbidity effects and PM in older age groups (CD,
29 pp. 8-146, 8-166). The newer studies of cardiac function have also suggested that elderly persons
30 with preexisting cardiopulmonary disease are susceptible to changes in cardiac function (CD, p.
August 2003 3-67 Draft - Do Not Cite or Quote
-------
1 8-146), indicating a connection between the elderly and preexisting disease status in
2 susceptibility to PM-related effects.
3 Children have been identified as being particularly susceptible to the effects of PM on the
4 respiratory system (CD, p. 8-152). As summarized above, associations have been found between
5 PM and acute respiratory symptoms or changes in lung function, hospital admissions or
6 emergency room visits for respiratory disease, as well as chronic effects such as decreased lung
7 function growth. The draft CD highlights findings of a number of new studies that raise the
8 possibility that deposition may be greater in children than adults, though the studies "still do not
9 provide unequivocal evidence for significant differences in deposition between children and
10 adults" (CD, p. 6-31). However, children have generally higher activity levels with
11 accompanying higher ventilation rates that might contribute to increased particle deposition.
12 Children also have smaller lungs and higher minute volumes relative to lung size, which would
13 likely result in greater doses of particles per lung surface area compared to adults; they are more
14 likely to spend time outdoors, increasing exposure to ambient particles; and they have the
15 highest rates of respiratory illnesses related to infection (CD, p. 9-135).
16 Several new epidemiologic studies have reported significant associations between PM
17 exposure and intrauterine growth reduction or low birth weight, known to be infant health risk
18 factors, as well as excess infant mortality, suggesting that the prenatal development period may
19 be a time of potential susceptibility to air pollutants (CD, p. 9-106). It is known that the early
20 post-natal period of lung development is a time of high susceptibility for lung damage by
21 pollutants, which is concordant with findings of reduced lung function growth in older children
22 in the Southern California children's cohort (CD, p. 9-136).
23
24 3.5 EVALUATION OF PM-RELATED HEALTH EFFECTS EVIDENCE
25 The preceding sections summarized evidence from new health studies and integrated
26 them with findings from previous PM NAAQS reviews. As in those previous reviews, much of
27 the currently available health evidence is derived from epidemiologic studies, though critical
28 new insights are offered in the results of toxicologic and controlled human exposure studies.
29 The 1996 CD and Staff Paper discussed at some length issues related to the interpretation and
30 evaluation of epidemiological evidence. While recognizing that additional research was needed
August 2003 3-68 Draft - Do Not Cite or Quote
-------
1 on some issues, the 1996 CD concluded that "the epidemiologic findings cannot be wholly
2 attributed to inappropriate or incorrect statistical methods, misspecification of concentration-
3 effect models, biases in study design or implementation, measurement errors in health endpoint,
4 pollution exposure, weather, or other variables, nor confounding of PM effects with effects of
5 other factors" (EPA, 1996a, p. 13-92).
6 In this section, the new findings relevant to the interpretation of epidemiological
7 information will be discussed. Throughout the preceding discussions on the nature of health
8 effects associated with PM, and the consistency and coherence of the health evidence,
9 consideration of potential confounding by co-pollutants has been discussed. Here, additional
10 considerations regarding potential confounding by gaseous co-pollutants will be discussed in
11 section 3.5.1. Several additional issues related to analytical modeling in epidemiological studies
12 is discussed in section 3.5.2: (1) the influence of model specification on epidemiologic findings
13 and the related issue of appropriate control for potential confounding by weather and related
14 time-varying factors; (2) the lag period between exposure and occurrence of health effects; and
15 (3) the influence of exposure error or exposure misclassification on reported PM-health
16 associations. In addition, new information is available on potential health effects of PM
17 components or source-related PM, as summarized in section 3.5.3. Finally, evidence regarding
18 the consistency and coherence of PM-related effects is summarized in section 3.5.4.
19
20 3.5.1 Role of Gaseous Co-pollutants
21 The extent to which PM-related effects are confounded or modified by other pollutants
22 was discussed at some length in the 1996 Staff Paper (EPA, 1996b, pp. V-45 to V-54) as one of
23 the important uncertainties considered in making recommendations concerning the PM NAAQS
24 (EPA, 1996b, p. VII-13). As stated previously, another pollutant meets the criteria for potential
25 confounding in PM-health associations if: (1) it is a potential risk factor for the health effect
26 under study; (2) it is correlated with PM; and (3) it does not act as an intermediate step in the
27 pathway between PM exposure and the health effect under study (CD, p. 8-8). In addition, there
28 may be effects associated with co-exposure to multiple pollutants. The draft CD states, "it is
29 important to continue to recognize that health effects associated statistically with any single
30 pollutant may actually be mediated by multiple components of the complex ambient mix" (CD,
August 2003 3-69 Draft - Do Not Cite or Quote
-------
1 p. 9-69). Recent animal toxicologic studies have tested effects of exposure to PM or PM
2 surrogates (e.g., urban PM, carbon particles, acid aerosols) in combination with O3. In a series
3 of recent studies, co-exposure to O3 and urban particles resulted in greater effects than those
4 reported with exposure to O3 alone, while mixed results were reported from studies using
5 combinations of acid aerosols and O3 (CD Table 7-12).
6 In epidemiologic studies, there are a number of methods for assessing potential
7 confounding by co-pollutants, including multi-pollutant modeling in multiple or single locations
8 and assessing the relationship between PM-mortality associations and correlations between PM
9 and copollutant concentrations in multi-city studies. All methods have issues to be considered in
10 interpreting their findings (CD, p. 8-13 to 8-14). It is important to recognize that there are issues
11 relating to potential copollutant confounding that multi-pollutant models may not be able to
12 address. Inclusion of pollutants in a multi-pollutant model that are highly correlated with one
13 another can lead to misleading conclusions in identifying a specific causal pollutant.
14 Collinearity between pollutants may occur if the gaseous pollutants and PM come from the same
15 sources, or if PM constituents are derived from gaseous pollutants (e.g., sulfates from SO2) (CD,
16 p. 8-205). This situation certainly occurs. For example, sources of PM constituents include
17 combustion of various fuels, gasoline or diesel engine exhaust, and some industrial processes
18 (CD, Table 9-1); these sources also emit gaseous pollutants. When collinearity exists, multi-
19 pollutant models would be expected to produce unstable and statistically insignificant effect
20 estimates for both PM and the co-pollutants. Methods that use data from multiple cities in
21 hierarchical or two-stage analyses have been used in some more recent studies, but these also
22 may be influenced by model misspecification or heterogeneity across the cities (CD, p. 8-14).
23 As discussed in the section 3.3, multi-pollutant models have been commonly used to
24 assess potential confounding among pollutants. This approach was used in the NMMAPS
25 mortality analyses for 20 and 90 U.S. cities, in which the authors added first ozone, then ozone
26 and another co-pollutant (e.g., CO, NO2 or SO2) to the models (CD, p. 8-35). As shown in
27 Figure 3-10, the relationship between PM10 and mortality was little changed in models including
28 control for ozone and other gaseous pollutants. The authors concluded that the PM10-mortality
29 relationship was not confounded by co-pollutant concentrations across 90 U.S. cities (Samet et
30 al., 2000a,b, reanalyzed Domenici 2003). Among the recent single-city studies, a number of
August 2003 3-70 Draft - Do Not Cite or Quote
-------
1 research groups have reported little influence of co-pollutants on PM associations in such models
2 (e.g., Lippmann et al., 2000, reanalyzed Ito, 2003; Fairley 1999, reanalyzed Fairley 2003; Ostro
3 et al., 2000, reanalyzed Ostro et al., 2003), although some have reported substantial reduction in
4 associations with PM when gaseous pollutants are added to the model (e.g., Chock et al., 2000;
5 Moolgavkar et al., 2000a,b,c; reanalyzed Moolgavkar 2003).
6 An example of the second type of evaluation of multi-pollutant confounding was used in
7 the NMMAPS morbidity analyses in 14 U.S. cities. The authors used an alternative approach,
8 testing for relationships between the coefficients for the PM10-admissions with PM10-co-pollutant
9 correlations for each city. No such relationships were found between the PM10 effect estimates
10 and PM10-co-pollutant correlations, leading the authors to conclude that the associations with
11 PM10 were not confounded by the gaseous co-pollutants (CD, pp. 8-125, 8-151).
12 Based on the results of multi-pollutant modeling methods, the draft CD concludes that
13 while there are still difficulties in separating effects of pollutants, "it now appears unlikely that
14 such confounding accounts for the vast array of effects attributed to ambient PM based on the
15 rapidly expanding PM epidemiology database" (CD, p. 9-112). This conclusion is supported not
16 only by statistical approaches, but also by: (1) the biological plausibility of effects of the
17 potentially confounding pollutants; and (2) consideration of exposure relationships.
August 2003 3-71 Draft - Do Not Cite or Quote
-------
-0.25 0.0 0.25 0.5 0.75
% Change in Mortality per 10 jjg/m3 Increase in PM
10
2
3
4
5
Figure 3-10. Marginal posterior distributions for effect of PM10 on total mortality at lag 1
with and without control for other pollutants, for the 90 cities. The numbers in
the upper right legend are the posterior probabilities that the overall effects
are greater than 0. (CD Figure 8-6)
Source: Dominici et al. (2003).
Biological plausibility. With regard to biological plausibility, the draft CD observes that
"It is generally accepted that O3, NO2 and SO2 are associated with diminished pulmonary
function and increased respiratory symptoms as well as more serious consequences, and CO
exposure has been associated with cardiovascular effects" (CD, p. 8-206). For example, if
concentrations of CO and PM are correlated, CO is a potential confounder for PM-related effects
on the cardiovascular system, while it is an implausible confounder for effects on the respiratory
August 2003
3-72
Draft - Do Not Cite or Quote
-------
1 system. Reflecting the evidence summarized in the recent Criteria Document for CO, the draft
2 CD finds that "[a]mong the gaseous criteria pollutants, CO has emerged as the most consistently
3 associated with cardiovascular (CVD) hospitalizations. The CO effects are generally robust in
4 the multi-pollutant model, sometimes as much so as PM effects. However, the typically low
5 levels of ambient CO concentrations in most such studies and minimal expected impacts on
6 carboxyhemoglobin levels and consequent associated hypoxic effects thought to underlie CO
7 CVD effects complicate interpretation of the CO findings and argue for the possibility that CO
8 may be serving as a general surrogate for combustion products (e.g., PM) in the ambient
9 pollution mix." (CD, p. 9-37).
10 Similarly, ozone would plausibly be a potential confounder for respiratory effects, but not
11 cardiovascular effects. In the discussion of studies on respiratory hospital admissions and
12 respiratory symptoms or lung function in asthmatics (CD sections 8.3.2 and 8.3.3), however,
13 where PM and ozone are considered together, they were generally found to have independent
14 effects (CD, pp. 8-174, 8-185).
15 SO2 exposure can affect the airways, especially for people with asthma, resulting in
16 breathing difficulties that can be accompanied by symptoms such as wheezing, chest tightness,
17 or shortness of breath. Short-term exposures (e.g., less than 3 hours) to low levels of NO2 may
18 lead to changes in airway responsiveness and lung function in individuals with preexisting
19 respiratory illnesses and increases in respiratory illnesses in children. Long-term exposures to
20 NO2 may lead to increased susceptibility to respiratory infection and may cause irreversible
21 alterations in lung structure. Thus, both SO2 and NO2 are plausibly linked with effects on the
22 respiratory system, but it is less plausible to consider SO2 and NO2 as confounders in
23 associations with effects such as cardiovascular mortality.
24 Exposure relationships. In addition to considering correlations between ambient
25 concentrations of PM and the gaseous pollutants, several new exposure studies have assessed
26 correlations between ambient pollutant concentrations and individuals' personal exposure to
27 ambient-origin pollutants. Some recent exposure studies have collected personal and ambient
28 monitoring data, collected at a single central site, for PM2 5 and gaseous pollutants (e.g., O3, SO2
29 and NO2), and assessed the degree of day-to-day correlation between the different measures of
30 personal and ambient concentrations. The investigators reported that the personal and ambient
August 2003 3-73 Draft - Do Not Cite or Quote
-------
1 PM2 5 measurements were correlated, as were personal exposure to PM2 5 and ambient
2 concentrations of the gaseous pollutants. However, the personal and ambient concentrations of
3 each of the gaseous pollutants were not well correlated. The draft CD concludes that ambient
4 concentrations of NO2, SO2, and O3, and likely CO, are unlikely to confound reported
5 associations of ambient PM with health effects, but can possibly serve as surrogates for ambient
6 PM exposure or more likely exposure to source categories with which the gases are correlated,
7 i.e., NO2 and CO with motor vehicle-associated PM and SO2 and O3 with regional sulfate (CD, p.
8 9-37).
9 Summary. Several lines of evidence provide information on potential confounding of
10 PM-health relationships by co-pollutants. Among studies using multi-pollutant modeling
11 methods, most but not all report no substantial change in PM effect estimates with control for
12 gaseous co-pollutants. Powerful new evidence is available from multi-city studies, particularly
13 the NMMAPS, where the relationship between PM10 and mortality was relatively unchanged
14 with control for gaseous co-pollutants in 90 U.S. cities. Considering biological plausibility,
15 clearly other pollutants such as the gaseous criteria pollutants have been associated with certain
16 health effects; indeed EPA has established NAAQS for the other criteria pollutants based in part
17 on such associations. Where PM and the other pollutants are correlated, it can be difficult to
18 distinguish effects of the various pollutants in multi-pollutant models. However, a number of
19 research groups have found the effects of PM and gases to be independent of one another, for
20 example, in the evidence from asthma symptom studies with PM and ozone. In addition, new
21 evidence on exposure considerations suggests that it is less likely that a relationship found
22 between a health endpoint and ambient PM concentrations is actually representing a relationship
23 with another pollutant. Taking into consideration the findings of single- and multi-city studies
24 and other evaluations of potential confounding by gaseous co-pollutants described in preceding
25 sections, the evidence generally indicates that PM, alone or in combination with other pollutants,
26 has independent effects on morbidity and mortality (CD, p. 8-279).
27
28 3.5.2 Issues Related to Epidemiological Modeling
29 The 1996 CD included extensive discussions of methodological issues for epidemiologic
30 studies, including questions about model specification or selection, measurement error in
August 2003 3-74 Draft - Do Not Cite or Quote
-------
1 pollutant measurements and exposure misclassification. As summarized in the 1996 Staff Paper,
2 PM-health effects associations reported in epidemiologic studies were not likely an artifact of
3 model specification, since analyses or reanalyses of data using different modeling strategies
4 reported similar results (EPA 1996b, p. V-39). Little information was available for the 1996 CD
5 to allow comparison of differing lag periods for different health effects. Also, few studies
6 quantitatively evaluated the potential influence of measurement or exposure error in interpreting
7 epidemiologic study findings or compared exposure time windows for PM-related health effects.
8 Recent studies have provided some new information, as discussed below. Broader
9 questions about model selection resurfaced recently in the process of addressing questions about
10 time-series epidemiologic models using GAM. These new questions have been evaluated in a
11 series of reanalyses of some time-series studies, and are discussed in section 3.5.2.1. Some
12 recent studies have evaluated appropriate lag periods between the concentration exposure period
13 and the health outcome being measured, and the results of these studies are discussed in section
14 3.5.2.2. Finally, section 3.5.2.3 summarizes new information from studies that have assessed the
15 influence of measurement error on epidemiologic analyses.
16 An additional issue, the form of the concentration-response function, is described in more
17 detail in Chapter 4 (section 4.2.6.1), with the discussion of the risk analyses. To briefly
18 summarize that discussion, a number of short-term exposure studies have evaluated
19 concentration-response functions for mortality, and while one study suggested a potential
20 threshold level for PM2 5, overall, the studies do not provide evidence for a discernible threshold
21 level (see p. 4-32). For long-term exposure to PM, the shape of the concentration-response
22 function was also evaluated using data from the ACS cohort, where mean PM2 5 concentrations
23 ranged from about 10 to 34 |ig/m3, and the concentration-response relationships for all-cause and
24 cardiopulmonary mortality could be reasonably approximated by a linear model (CD, p. 8-89).
25 Staff concludes that, while it is reasonable to expect that for individuals or groups of individuals
26 there may be biological thresholds for different effects, discernible population threshold levels
27 have not been detected in available epidemiologic studies.
28 3.5.2.1 Model Specification
29 In 2002, questions were raised about the default convergence criteria and standard error
30 calculations made using generalized additive models (GAM), which has been commonly used in
August 2003 3-75 Draft - Do Not Cite or Quote
-------
1 recent time-series epidemiologic studies. As discussed more completely in section 8.4.2 of the
2 draft CD, a number of time-series studies were reanalyzed using alternative methods, typically
3 GAM with more stringent convergence criteria and an alternative model such as generalized
4 linear models (GLM) with natural smoothing splines, and the results of the reanalyses have been
5 compiled and reviewed in a recent HEI publication (HEI, 2003 a). In most, but not all, of the
6 reanalyzed studies, it was found that risk estimates were reduced and confidence intervals
7 increased with the use of GAM with more stringent convergence criteria or GLM analyses (CD,
8 p. 8-193). PM10 mortality study results are presented in Figure 8-17 of the draft CD, where it can
9 be seen that the reanalyses generally did not substantially change the findings of the original
10 analyses, and the changes in risk estimates with alternative analysis methods were much smaller
11 than the variation in effects across studies. The HEI review committee concluded:
12 • While the number of studies showing an association of PM with
13 mortality was slightly smaller, the PM association persisted in the
14 majority of studies.
15 • In some of the large number of studies in which the PM
16 association persisted, the estimates of PM effect were substantially
17 smaller.
18 • In the few studies in which investigators performed further
19 sensitivity analyses, some showed marked sensitivity of the PM
20 effect estimate to the degree of smoothing and/or the specification
21 of weather. (HEI, 2003b, p. 269)
22 Thus, these reanalyses indicated that weather continues to be a potential confounder of
23 concern. As summarized in the draft CD, there remains no altogether satisfactory way to select
24 the "best" model to address time and weather variables. If the model does not adequately
25 address daily changes in weather variables, then some effects of temperature on health would be
26 falsely ascribed to the pollution variable. Conversely, if the analyst overcontrols for weather,
27 such that the temperature-health relationship is more "wiggly" than the true dose-response
28 function, then the result will be a much less efficient estimate of the pollutant effect. This would
29 result in incorrectly ascribing some of the true pollution effect to the temperature variable, which
30 would make it difficult to detect a real but small pollution effect. Thus, the HEI review panel
31 recommended "further exploration of these studies to a wider range of alternative degrees of
32 smoothing and to alternative specifications of weather variables in time-series models." (CD, p.
33 8-202).
August 2003 3-76 Draft - Do Not Cite or Quote
-------
1 Among the new studies included in Appendix 8 A of the draft CD are some that use case-
2 crossover methods. The case-crossover study design has only recently been applied in studies of
3 the health effects of air pollutants. This type of study uses the health event (e.g., hospital
4 admission for heart disease) as the case period, and selects a control period from some specific
5 time before or after the event, and assesses whether there are differences in risk factors (air
6 pollutants and other factors) between the periods. Where both case-crossover and time-series
7 methods have been used to analyze the same or similar data, the results have been similar. For
8 example, Neas et al. (1999) found associations between TSP and daily mortality in Philadelphia
9 similar to those found in previous time-series studies summarized in the 1996 CD (EPA 1996a,
10 pp. 12-283 to 12-302), while Lee et al. (1999) and Lee and Schwartz (1999) used time-series and
11 case-crossover methods with TSP and daily mortality data from Seoul, Korea, and neither
12 reported a statistically significant association.
13 Methods used in assessing effects associated with long-term exposure to pollutants were
14 also reviewed as a part of the reanalysis of the long-term mortality studies (Krewski et al., 2000).
15 The authors applied an array of different models and variables to determine whether the original
16 results would remain robust to different analytic assumptions and model specifications. The
17 draft CD concludes "[n]one of these alternative models produced results that materially altered
18 the original findings" (CD, p. 8-84).
19 In summary, questions about the analytical methods used in epidemiologic studies,
20 especially the time-series studies, have been raised in this review as well as previous reviews.
21 The recently discovered issues with GAM analyses have raised new questions about addressing
22 time-varying variables such as weather adequately, but the overall conclusion has remained that
23 the original findings are generally supported in reanalyses. In addition, the major reanalysis of
24 the prospective cohort studies reported that the findings were not sensitive to alternative
25 modeling strategies, but additional questions remain about adequately addressing spatial
26 correlation in the data (Krewski et al., 2000).
27 With the substantial amount of evidence now available on PM-related health effects, EPA
28 has not focused on evaluating the methods used in each individual study, but rather on the full
29 body of evidence, including epidemiologic studies that use a variety of methods and data from
August 2003 3-77 Draft - Do Not Cite or Quote
-------
1 multiple or single locations, as well as information from other scientific disciplines. The
2 approach is summarized in the draft CD:
3 Rather, a three-pronged approach is likely to yield useful evaluative information: (1) an
4 overall characterization of evident general commonalities (and/or notable marked
5 differences) among findings from across the body of studies dealing with particular PM
6 exposure indices and types of health outcomes, looking for convergence of evidence
7 regarding types of effects and effect-sizes attributable to ambient PM indices across
8 various methodologically acceptable analyses; (2) thorough, critical assessment of newly
9 published multi-city analyses of PM effects, assuming that greater scientific weight is
10 generally ascribable to their results than those of smaller-sized studies (often of
11 individual cities) yielding presumably less precise effect size estimates; and (3)
12 evaluation of coherence of the findings among different types of effects and across
13 various geographic locations, as well as with other types of pertinent biological
14 information (e.g., exposure, dosimetry, toxicity, etc.). (CD, p. 8-192)
15
16 3.5.2.2 Lag Periods
17 One important model selection question in studies of short-term PM exposures is the
18 selection of a lag period between air pollutant concentration changes and health outcome
19 measures. Commonly used lags are 0 day (effects occurring on the same day as the pollution
20 measurement), 1 to several days, or average pollution measures over several days preceding the
21 health outcome. Many epidemiologic studies on the health effects of acute PM exposure have
22 tested several lag periods, or time delays between the pollution measurement and the occurrence
23 of the health outcome being measured, and the results for the most statistically significant lag
24 period are reported in the publication. As stated in the draft CD, "[w]hile this practice may bias
25 the chance of finding a significant association, without a firm biological reason to establish a
26 fixed pre-determined lag, it appears reasonable" (CD, p. 8-234). In some more recent studies,
27 goodness-of-fit criteria have been used to help select the best modeling approach. Some multi-
28 city studies, such as the NMMAPS, have used a fixed lag period for all sites. In addition, an
29 alternative approach, the distributed lag, has been introduced in several new studies. With this
30 approach, the effect of pollution on health is assessed as the effect of a weighted average
31 pollution variable, recognizing that effects of air pollution can occur on several subsequent days.
32 In the NMMAPS analysis of PM10 associations with total mortality, lag periods of 0, 1
33 and 2 days were used across all cities. The authors reported associations with all three lags, with
34 the largest association being reported for a 1-day lag period. As stated in the draft CD, "since
August 2003 3-78 Draft - Do Not Cite or Quote
-------
1 the cardiovascular, respiratory or other causes of acute mortality usually associated with PM are
2 not at all specific, there is little a priori reason to believe that they must have the same relation to
3 current or previous PM exposures at different sites" (CD, p. 8-235). For morbidity studies, the
4 findings are similar. The draft CD reports that time series studies of hospital admissions or
5 emergency room visits for cardiovascular diseases suggest that the strongest effects are reported
6 at lag 0, with some effects seen at lag 1 but little beyond a one-day lag (CD, p. 8-147). Recent
7 studies of hospital admissions for respiratory diseases, summarized in Tables 8-17, 8-19 and 8-
8 20 of the draft CD, report results for varying lag periods for different categories of respiratory
9 disease, ranging from 0 to 3 days. In the NMMAPS evaluation of PM10 associations with
10 hospital admissions among the elderly, the distributed lag approach was reported to generally
11 result in stronger associations (Samet et al., 2000b). In an analysis using data on cause-specific
12 mortality from 10 U.S. cities, distributed lag estimates were substantially larger than estimates
13 for 1- or 2-day average lag periods (Schwartz, 2003b).
14 As noted in the draft CD, misspecification of the lag structure can result in modeling
15 biases (CD, p. 8-237). One concern is where PM measurements are available only every other or
16 every sixth day, which makes evaluation of appropriate lag periods difficult, since it isn't
17 possible to test distributed lag or moving average models adequately. Also, as the draft CD
18 states "[o]ne would then expect to see different best-fitting lags for different cause-specific
19 mortality or hospital admissions" (CD, p. 8-238). Based on these potential concerns, and the
20 results of some studies indicating larger effect sizes in distributed lag models, the draft CD
21 concludes that single-day or two-day average lag periods may be underestimating PM effects
22 (CD, p. 8-238).
23 In summarizing questions about lag periods for the short-term exposure studies, the draft
24 CD states:
25 It may be possible that different PM size components or particles with different
26 composition or sources produce effects by different mechanisms manifested at different
27 lags, or that different preexisting conditions may lead to different delays between
28 exposure and effect. Thus, although maximum effect sizes for PM effects have often
29 been reported for 0-1 day lags, evidence is also beginning to suggest that more
30 consideration should be given to lags of several days. Also, if it is considered that all
31 health effects occurring at different lag days are all real effects, so that the risks for each
32 lag day should be additive, then higher overall risks may exist that are higher than
August 2003 3-79 Draft - Do Not Cite or Quote
-------
1 implied by maximum estimates for any particular single or two-day lags. In that case,
2 multi-day averages or distributed lag models should be used. (CD, p. 8-280).
3 3.5.2.3 Measurement Error
4 In this and previous reviews of the PM NAAQS, much of the health evidence for PM-
5 related effects comes from epidemiologic studies where ambient PM measurements are used to
6 represent community PM exposures. One key issue is the use of PM concentrations measured at
7 central locations to represent the community's exposure to ambient PM. As discussed in section
8 2.8, daily changes in individuals' personal exposure to ambient PM is well correlated with daily
9 changes in ambient PM measured at central monitors. While particles can be generated from
10 indoor sources, particles from nonambient sources are not correlated temporally with ambient
11 particle concentrations. Thus, the draft CD concludes that ambient PM concentrations measured
12 at central site monitors are a useful surrogate for exposure to ambient PM (CD, p. 9-117, p. 8-
13 252).
14 Another key issue in interpreting epidemiologic study findings is related to error in the
15 measurements of the pollutants. Analyses available for the 1996 Staff Paper indicated that
16 random measurement error in pollutant concentration data is not likely to bias the findings of
17 epidemiologic analyses using these data. However, a remaining question was the existence of
18 differential measurement error, where one pollutant was measured with more error than another,
19 and the effect this might have in comparing epidemiologic findings for the two pollutants (EPA,
20 1996b, p. V-42).
21 The draft CD summarizes the findings of several new analyses that show the potential
22 influence of differential measurement error on epidemiological analysis results, though the
23 conditions required for the error to substantially influence the epidemiological findings are
24 severe and unlikely to exist in current studies. In simulation analyses of a "causal" pollutant and
25 a "confounder" with differing degrees of measurement error and collinearity between the
26 pollutants it was found that, in some circumstances, a causal variable measured with error may
27 be overlooked and its significance transferred to a surrogate. However, for "transfer of apparent
28 causality" from the causal pollutant to the confounder to occur, there must be high levels of both
29 measurement error in the causal variable and collinearity between the two variables (CD, p. 8-
30 249).
August 2003 3-80 Draft - Do Not Cite or Quote
-------
1 An additional analysis applied measurement error models to data from the Harvard Six
2 Cities study, specifically testing relationships between mortality and either fine or coarse fraction
3 particles. The authors identified several variables that could influence bias in effect estimates for
4 fine- or coarse-fraction particles: the true correlation of fine- and coarse-fraction particles,
5 measurement errors for both, and the underlying true ratio of the toxicity of fine- and coarse-
6 fraction particles. The existence of measurement error and collinearity between pollutants could
7 result in underestimation of the effects of the less well-measured pollutant. However, the
8 authors conclude "it is inadequate to state that differences in measurement error among fine and
9 coarse particles will lead to false negative findings for coarse particles. If the underlying true
10 ratio of the fine and coarse particle toxicities is large (i.e., greater than 3:1), fine particle
11 exposure must be measured significantly more precisely in order not to underestimate the ratio
12 of fine particle toxicity versus coarse particle toxicity" (Carrothers and Evans, 2000, p. 72).
13 Thus, while the potential remains for differential error in pollutant measurements to influence the
14 results of epidemiologic studies, it is unlikely that the levels of measurement error and
15 correlation between pollutants reported in existing studies would result in transfer of apparent
16 causality from one pollutant to another.
17 The influence of exposure misclassification on the results of epidemiological analyses
18 has been further investigated in one major new analysis that was conducted as a part of
19 NMMAPS (Zeger et al., 2000). Using data collected in previous exposure studies, the authors
20 developed a relationship between personal exposure to ambient particles and ambient PM10
21 concentrations. The authors reported that the association between PM10 and mortality using
22 ambient PM10 concentrations underestimated the association between personal ambient PM10
23 exposure and mortality.
24 Sheppard and Damian (2000) did a simulation study using asthma hospital admissions
25 and PM data from Seattle. As was found for mortality, effect estimates and their standard errors
26 did increase with increasing magnitude of measurement error variance. However, "The change
27 is small and does not influence the relative risk estimate" (Sheppard and Damian, 2000, p. 686).
28 In reviewing these new studies, along with analyses that were available in previous
29 reviews, the draft CD concludes "the studies that examined joint effects of correlation and error
30 suggest that PM effects are likely underestimated, and that spurious PM effects (i.e., qualitative
August 2003 3-81 Draft - Do Not Cite or Quote
-------
1 bias such as change in the sign of the coefficient) due to transferring of effects from other
2 covariates require extreme conditions and are, therefore, unlikely." (CD, p. 8-255)
3
4 3.5.3 Consideration of Evidence on PM Size and Composition
5 Much of the focus of the preceding discussions on the nature of PM-related effects has
6 been epidemiologic studies that use gravimetric PM measurements, with an emphasis on PM10,
7 PM2 5 and PM10_2 5. In the 1996 Staff Paper, evidence from health studies for the two main PM
8 size categories, fine- and coarse-fraction particles, was extensively discussed (EPA, 1996b, p. V-
9 58 to V-76). The body of evidence on fine- and coarse-fraction particles has been greatly
10 expanded, and the results of recent studies have been discussed throughout the discussions of
11 mortality and morbidity effects in section 3.3.
12 Beyond studies of fine- and coarse-fraction particles, the newly available evidence also
13 includes studies of ultrafine particles, various components in the fine and coarse fractions, and
14 source-related particles. Section 5.3.5.1 will focus on the new information regarding ultrafine
15 particles. Building upon information available in the last review, section 3.5.3.2 discusses
16 reported health associations with different PM components and with various source-related
17 particles.
18 3.5.3.1 Ultrafine particles
19 As described in Chapter 2, ultrafine particles generally include particles smaller than 0.1
20 |im in diameter. Ultrafine particles are a portion of fine PM; they predominate in the number of
21 particles, but comprise only a very small portion of fine PM mass. It has been suggested, based
22 on toxicological evidence, that ultrafine particles may be relatively more toxic than larger
23 particles (e.g., accumulation mode particles). It has also been proposed that particle surfaces, or
24 the chemical composition of particle surfaces, may be related to PM toxicity, and ultrafine
25 particles have relatively large surface area for a given mass (CD, p. 7-84). It is expected that
26 ultrafine particles "because of their small size, are not effectively phagocytized by alveolar
27 macrophages and can easily penetrate the airway epithelium, gaining access to the interstitium"
28 (CD, p. 7-85). In addition, some recent studies indicate that ultrafine particles can be rapidly
29 cleared from the lungs into the systemic circulation and reach other organ systems (CD, p. 6-53).
August 2003 3-82 Draft - Do Not Cite or Quote
-------
1 The toxicologic studies available to date addressing potential effects of ultrafine particles
2 have used PM surrogates or model particles, such as ultrafine carbon or TiO2 particles. Several
3 new studies are reviewed in the draft CD with somewhat mixed findings on whether greater
4 effects are reported with ultrafine particles than with fine particles. In studies using metal oxide
5 dusts, the health response was increased with increasing total surface area, suggesting that
6 particle surface chemistry is an important component of biological responses (CD, p. 7-86).
7 However, the draft CD concludes that there is insufficient toxicological evidence to conclude
8 that ambient ultrafine particle concentrations are more strongly linked to health effects than mass
9 concentrations of fine particles (CD, p. 7-118).
10 A limited number of epidemiologic studies, all conducted in European nations, have
11 evaluated health associations with ultrafine particles. One study reported associations between
12 total mortality and both fine particle mass and ultrafine particle number count data, with effects
13 of about the same magnitude reported for each PM size fraction. The authors concluded that
14 both fine and ultrafine particles showed independent effects on mortality at ambient
15 concentrations (Wichmann et al., 2000; reanalyzed in Stolzel et al., 2003). Three studies, using
16 panels of asthmatic children or adults, have reported associations between ultrafine particles and
17 increased symptoms or decreased pulmonary function. All reported associations with both
18 ultrafine particle number concentrations and mass concentrations of BS, PM2 5 or PM10,
19 generally reporting significant associations with both ultrafine particle counts and PM mass
20 (Peters et al., 1997; Pekkanen et al., 1997; Tiittanen et al., 1999).
21 Finally, some new evidence from human exposure studies has indicated that infiltration
22 rates for ultrafine particles into buildings are lower than those for fine (accumulation mode)
23 particles (CD, p. 9-32). This would suggest that community exposure to PM is greater for
24 ambient fine particles than ambient ultrafine particles, and makes it unlikely that health
25 associations found with ambient PM2 5 mass are truly reflecting underlying associations with
26 ultrafine PM.
27 3.5.3.2 PM Components and Source-Related Particles
28 In the 1996 CD, evidence from toxicological studies on the effects of acid aerosols,
29 metals, ultrafine particles, diesel emission particles, silica, and bioaerosols was available. In
30 addition, a substantial body of epidemiologic studies had evaluated relationships between
August 2003 3-83 Draft - Do Not Cite or Quote
-------
1 mortality and morbidity and ambient sulfate or acid aerosol concentrations. The 1996 CD
2 concluded that the epidemiologic studies suggest that strongly acidic PM, including sulfates as
3 an indicator of acid aerosols, was associated with both acute and chronic health effects (EPA,
4 1996a, p. 12-253).
5 Recent studies have evaluated the effects of not only numerous PM components (e.g.,
6 sulfates, nitrates, acids, metals, elemental carbon, biological components), but also PM from
7 different sources (e.g., motor vehicle or industrial emissions, crustal material). Many of the
8 components or sources listed above are more likely to be linked with fine-fraction than with
9 coarse-fraction particles, such as sulfates, nitrates, acid aerosols, metals, and motor vehicle
10 emissions. In addition, the CAPs used in toxicology studies has been comprised of concentrated
11 fine-fraction or accumulation mode particles. Biological components (e.g., fragments of pollen
12 grains, mold spores) and crustal material are more likely linked with coarse-fraction particles.
13 Thus, more new information regarding PM components is relevant to fine-fraction than to
14 coarse-fraction particles. In the following discussion, evidence on components that are likely to
15 be found in fine-fraction particles is summarized first, following by components or sources
16 reflecting coarse-fraction particles.
17 Among epidemiologic studies that examined the effects of specific components of PM,
18 most commonly used were sulfates and acids, COH, and elemental carbon or organic carbon (as
19 indicators of motor vehicle emissions). Several new epidemiologic studies, and one toxicologic
20 study, have used factor analysis or source apportionment methods in an attempt to distinguish
21 effects of PM from different source types. A larger body of evidence on effects of specific PM
22 components is available from toxicological studies.
23 As was reported in the previous review, numerous epidemiologic studies have indicated
24 that both mortality and morbidity effects are associated with ambient exposures to sulfates and
25 acid aerosols (H+). Similarly, associations reported in recent studies between ambient sulfates
26 and mortality are positive and most are statistically significant (CD, figure 8-8). The draft CD
27 concludes that, in these studies, the relative significance of sulfate and H+ varied from city to
28 city, and the associations were stronger in cities where the sulfate and H+ levels were relatively
29 high (CD, p. 8-66). Significant associations were also reported using sulfates as the PM
30 indicator in studies of long-term PM exposure and mortality (CD, p. 8-107).
August 2003 3-84 Draft - Do Not Cite or Quote
-------
1 One new study with exposures to CAPs in dogs reported an association between the
2 sulfur factor of the particles with changes in red blood cell count and hemoglobin levels (Clarke
3 et al., 2000). However, in toxicologic and controlled human exposure studies using exposure to
4 acid aerosols (CD, Table 7-5), limited evidence for effects are reported. The draft CD concludes
5 that the new studies are consistent with the findings from the previous review, where effects
6 were reported in toxicological or controlled human exposure studies only when levels were very
7 high, although "acidic components should not be entirely dismissed as possible mediators of
8 ambient PM health effects, since so little is known about potential cardiovascular impacts or
9 impacts in compromised subjects" (CD, p. 9-57). One difference between the epidemiologic and
10 toxicological studies is that the epidemiologic studies were measuring sulfates or acidity of the
11 ambient aerosol, while toxicological studies were using exposures to acid aerosols alone. The
12 draft CD notes that while the independent toxicities of sulfates and nitrates are found to be low
13 in toxicologic studies, they may influence the toxicity or bioavailability of other PM
14 components, such as the metals (CD, p. 9-69).
15 Elemental carbon and organic carbon concentrations were used in studies conducted in
16 Atlanta (Klemm and Mason, 2000; Metzger et al., 2003) and Phoenix (Mar et al., 2000,
17 reanalyzed Mar et al., 2003). Both were significant predictors of mortality in the Phoenix study,
18 and of emergency department visits for cardiovascular diseases in Atlanta (Metzger et al., 2003),
19 although no PM indicators were reported to be significantly associated with mortality in Atlanta,
20 possibly due to the small sample size in this preliminary analysis (Klemm and Mason, 2000).
21 The draft CD observes that the correlations between COH, elemental carbon and organic carbon
22 and other mobile source related pollutants (fine PM, NO2, CO) were high and are likely
23 indicators of particles from motor vehicle emissions, concluding that associations found with
24 these components suggest that "PM components from motor vehicle sources are likely associated
25 with mortality" (CD, p. 8-70).
26 Recent studies have provided substantial new evidence on effects of particulate metals
27 and diesel exhaust particles. The substantial body of evidence on both cancer and noncancer
28 effects of diesel exhaust emissions (including diesel exhaust particles) is summarized in the
29 Health Assessment Document for Diesel Engine Exhaust (EPA, 2002). These effects include
30 respiratory symptoms and exacerbation of the allergic response to inhaled antigens, potentially
August 2003 3-85 Draft - Do Not Cite or Quote
-------
1 contributing to the incidence and severity of allergic rhinitis and asthma (CD, p. 9-60). In
2 addition, long-term exposure to diesel exhaust is likely to pose a lung cancer hazard to humans
3 (EPA, 2002).
4 Metals, especially water soluble metals, have been reported to cause cell injury and
5 inflammatory changes in toxicologic studies; it is noted that many studies have used relatively
6 high doses (CD, p.7-116). For example, Costa and Dreher (1997) summarized results from a
7 number of toxicologic studies that have reported that effects were more closely linked to the
8 metal content of particles than particle mass. The transition metals, such as iron, vanadium or
9 nickel, have been most commonly associated with effects in toxicologic studies, and autopsy
10 studies have shown that exposure to these metals is widespread as indicated by their presence in
11 the lung tissues of urban residents. The draft CD concludes that combustion particles with high
12 metal content (e.g., ROFA) have been found to cause adverse effects, although it has not yet
13 been established that the same effects are found with the generally lower metal content of
14 ambient particles (CD, p. 9-58). Limited evidence on metals is also available from
15 epidemiologic studies, although one new study reported associations between mortality and
16 particulate iron, nickel and zinc in 8 Canadian Cities (Burnett et al., 2000; these results not
17 reanalyzed). As discussed above, a series of studies used ambient particles or particle extract
18 from Utah Valley PM, where a major source had been closed for a period of time, and reported
19 that effects were greater with particles collected during the period when the steel mill was
20 operating, and that the effects were also correlated with the metal content of the particles (CD, p.
21 9-63).
22 In one new toxicologic study, dogs were exposed to CAPs and numerous indicators of
23 lung injury or inflammation (e.g., white blood cell counts, protein in lung lavage fluid) and
24 cardiovascular health (e.g., platelet and red blood cell counts, hemoglobin or fibrinogen levels)
25 were measured (Clarke et al., 2000). The authors conducted factor analysis and identified four
26 PM factors: aluminum/silicon, sulfur, vanadium/nickel, and bromine/lead, each of which was
27 found to be associated with different effects. The authors conclude that specific components of
28 particles may be responsible for effects, but do not distinguish PM sources that would be linked
29 to each of the PM factors or components.
August 2003 3-86 Draft - Do Not Cite or Quote
-------
1 Four new epidemiologic studies and one toxicologic study have used factor analysis or
2 source apportionment methods to investigate health associations with PM (PM2 5 and PM10 or
3 PM15) from different sources (Laden et al., 2000, reanalyzed in Schwartz, 2003; Mar et al., 2000,
4 reanalyzed in Mar et al., 2003; Tsai et al., 2000; Ozkaynak et al., 1996; Clarke et al., 2000).
5 These studies used elements or other PM components as indicators of the emissions sources; for
6 example, Laden et al. (2000) use silicon as an indicator for fine particles of crustal or geologic
7 origin (CD, Table 8-4). In addition to testing associations between PM mass and mortality, the
8 four studies evaluated relationships with the PM source factors. The four epidemiologic studies
9 are fairly consistent in finding associations for mortality with indicators of PM (both PM10/15 and
10 PM2 5) from combustion sources, but not from geologic sources. The draft CD concludes that the
11 results of the epidemiologic studies generally indicate that a "number of source types were
12 associated with mortality, including motor vehicle emissions, coal combustion, oil burning and
13 vegetative burning" (CD, p. 8-70).
14 Bioaerosols are more typically found in coarse-fraction than fine-fraction particles. The
15 1996 CD concluded that effects of bioaerosols (e.g., endotoxin) were reported in toxicological or
16 controlled human exposure studies only when levels were very high. The recent toxicological
17 and controlled human exposure studies on the effects of ambient bioaerosols, primarily
18 endotoxins, are summarized in draft CD Table 7-6. These studies of workers exposed in
19 agricultural settings showed respiratory changes, such as reduced lung function or increased
20 airway responsiveness, with increasing dust or endotoxin exposure levels. These occupational
21 study findings were supported by evidence for inflammatory responses in animal or controlled
22 human exposure studies. However, the endotoxin levels measured in these studies were far
23 greater than levels generally reported in ambient air. The draft CD concludes: "[although these
24 exposures are massive compared to endotoxin levels in ambient PM in U.S. cities, these studies
25 serve to illustrate the effects of endotoxin and associated bioaerosol material in healthy
26 nonsensitized individuals" (CD, p. 7-39). In addition, a number of epidemiologic studies have
27 reported associations of mold spore concentrations on lung function or asthma symptom severity
28 (Delfino et al., 1996, 1997; Neas et al., 1996). In evaluating the results of new epidemiologic
29 studies on the association between mortality and coarse-fraction particles, the draft CD suggests
30 that the findings of associations in some areas "hint at possible contributions of biogenic
August 2003 3-87 Draft - Do Not Cite or Quote
-------
1 materials (e.g., molds, endotoxins, etc.) to the observed coarse particle effects in at least some
2 locations" but sufficient evidence is not yet available to support or refute this hypothesis (CD,
3 pp. 9-85).
4 The effects of PM of crustal or geologic origin were also investigated in two
5 epidemiologic studies that used meteorological data in conjunction with air quality data to
6 identify days where wind-blown crustal particles predominate. Both studies reported no
7 evidence of associations between mortality and wind-blown crustal particles (Schwartz et al.,
8 1999; Pope et al., 1999). The draft CD finds that the results of these studies, taken together,
9 suggest that particles of crustal origin (whether in the fine or coarse fraction of PM) are not
10 likely associated with acute mortality (CD, pp. 8-59). However, the draft CD observes that
11 "crustal" particles may carry biological components (e.g., endotoxin), toxic trace elements, or
12 components of previously deposited fine PM, such as metals from smelters, steel mills or other
13 industrial operations (CD, p. 8-281). In addition, the existing studies have assessed only
14 mortality as a health endpoint and there are numerous morbidity indices of potential concern.
15 These recent studies provide some new evidence for health effects associations with
16 many different PM components such as sulfates, acids and metals. It is also possible that
17 interaction between some PM components may be an important factor in some health effects
18 associations; for example, sulfates may increase the bioavailability of metals, and diesel particles
19 may increase allergic effects of bioaerosols. From the results of toxicologic studies, the draft CD
20 concludes "It appears that many biological responses are produced by PM whether it is
21 composed of a single component or a complex mixture" (CD, p. 7-79). For mortality, the factor
22 analysis studies appear to implicate ambient PM from combustion-related sources in associations
23 with total mortality, but not particles of crustal or geologic origin (CD, p. 9-89). The draft CD
24 concludes:
25 A primary causative attribute may not exist but rather many attributes may contribute to a
26 complex mechanism driven by the nature of a given PM and its contributing sources.
27 The multiple interactions that may occur in eliciting a response in a host may make the
28 identification of any single causal component difficult and may account for the fact that
29 mass as the most basic metric shows the relationships to health outcomes that it does. (p.
30 9-61)
31
August 2003 3-88 Draft - Do Not Cite or Quote
-------
1 3.5.4 Consistency and Coherence of Health Effects Evidence
2 The 1996 Staff Paper pointed out the inherent limitations in trying to determine the role
3 of PM by examining even the most thorough studies of individual cities that show associations
4 between ambient PM and various health effects. Accordingly, the staff presented a more
5 comprehensive synthesis that considered the consistency and coherence of the available evidence
6 in evaluating the likelihood of PM being causally associated with the observed effects (EPA,
7 1996b, V-54 to 58). Substantially more evidence of associations between ambient PM and
8 health effects is now available, including multi-city studies that address some of the limitations
9 of single-city studies, and these studies provide further support for the consistency and coherence
10 of the body of evidence on PM-related effects.
11 As discussed in the last review, consistency of an association is evidenced by repeated
12 observations by different investigators, in different places, circumstances and time; and by the
13 consistency of the association with other known facts (EPA, 1996a, Chapter 13). Beyond
14 considering the consistency of associations for individual health endpoints, coherence refers to
15 the logical or systematic interrelationship between different health indices that would be
16 expected to be seen across studies of different endpoints or from different disciplines. The
17 consistency and coherence of the expanded body of evidence now available is discussed and
18 evaluated below.
19 3.5.4.1 Consistency
20 The 1996 Criteria Document summarized over 80 community epidemiologic studies
21 evaluating associations between short-term PM levels and mortality and morbidity endpoints in a
22 number of locations throughout the world, using a variety of statistical techniques, of which over
23 60 studies found consistent, positive, significant associations (EPA, 1996a, Tables 12-2 and 12-8
24 to 12-13). The 1996 Staff Paper displayed the relative risk estimates for mortality and morbidity
25 effects associated with PM10 from the U.S. and Canadian studies, concluding that despite the
26 variations in study locations and approaches, the estimates for each health endpoint were
27 relatively consistent among the studies; although, as would be expected, some variation was seen
28 (EPA, 1996b, V-55 and Figure V-2).
29 Since the last review, more than 80 new studies have been published on mortality alone
August 2003 3-89 Draft - Do Not Cite or Quote
-------
1 and a large number of new morbidity studies. The draft CD notes that the effect estimates from
2 the new studies in the U.S. and throughout the world are generally consistent with those
3 observed in the last review, not only from PM10 multi- and single-city studies, but also from the
4 significantly expanded body of studies using PM2 5 (CD, p. 8-276). A number of the new studies
5 have included PM10_2 5 and suggest that coarse-fraction particles may also be linked with serious
6 health effects (CD, pp. 8-276, 8-278).
7 In the 1996 Staff Paper, staff also recognized a number of factors related to addressing
8 why effect estimates might be expected to vary from one location to another. These factors
9 include, for example, variation in ambient PM levels, differing socioeconomic or demographic
10 characteristics, variation in composition of PM across areas, and differing health status (i.e.,
11 potential susceptibility) between cities (EPA, 1996b, p. V-55). As discussed above in section
12 3.3, staff has focused on the results of recent multi-city studies in addressing questions about
13 variation across cities that are difficult to answer in single-city studies. Several multi-city
14 studies have explored potential differences in effect estimates among cities, and have provided
15 some evidence for factors that might explain variation in effect estimates. In addition to the
16 factors recognized in the previous PM NAAQS review, the CD observes that, especially in the
17 NMMAPS analyses for 90 U.S. cities, inclusion of cities with smaller sample sizes (with respect
18 to mortality rate and air quality data availability) may result in greater apparent heterogeneity
19 simply due to more statistical imprecision of results in the smaller cities (CD, p. 8-246). In
20 addition, the CD recognizes that factors such as air conditioner use, as an indicator for reduced
21 exposure to outdoor air pollutants, can potentially explain some variability across areas; this
22 would reflect differences in population exposures, rather than variations in biological response.
23 While it is reasonable to expect some variation in effects across areas, staff observes that
24 in combining the results for PM2 5 from short-term exposure studies conducted across the U.S.
25 and Canada for a range of health endpoints from mortality to varying indices of morbidity
26 (Figure 3-1 la), the effect estimates for each health endpoint are relatively consistent among the
27 studies, which is similar to the pattern observed for PM10 studies in the last review (EPA, 1996b,
28 Figure V-2). Effect estimates for associations with PM2 5 are positive for all health endpoints,
29 and many are statistically significant (Figure 3-1 la). For mortality, effect estimates for total
30 mortality fall within the same general range, especially when focusing on the results of studies
August 2003 3-90 Draft - Do Not Cite or Quote
-------
1 with greater precision, and effect estimates for cardiovascular and respiratory diseases are
2 somewhat more variable than those for total mortality but are still generally consistent across the
3 different locations. For morbidity, the effect estimates for hospital admissions or medical visits
4 for both cardiovascular and respiratory diseases vary within a larger range, but that might be
5 expected since results for specific causes (e.g., dysrhythmia, pneumonia) are grouped with
6 results for all cardiovascular or all respiratory diseases in this presentation.
7 For PM10.2 5 fewer studies are available, and staff observes more evidence for consistency
8 across morbidity effects than for mortality in the results of these studies (Figure 3-1 Ib). The
9 effect estimates for PM10_2 5 with cause-specific mortality are generally positive, with several
10 being statistically significant or nearly so, but only one of the effect estimates for PM10_2 5 and
11 total mortality is clearly positive and statistically significant. For hospital admissions/medical
12 visits for respiratory and cardiovascular diseases, the associations with PM10_2 5 are generally, but
13 not always, positive, and the effect estimates vary somewhat in size, especially those from the
14 preliminary analyses in Atlanta (Tolbert et al., 2000b). In Figures 3-1 la and 3-1 Ib, staff
15 observes that the patterns of results for PM2 5 and PM10_2 5 are generally similar, though the
16 evidence for consistency is greater for PM2 5 than for PM10_2 5, especially for mortality.
August 2003 3-91 Draft - Do Not Cite or Quote
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
in
c
o
•*=
ro
'o
o
0.
•e
o
in
O)
T3
c
ro
0
to
LU
S-
1
lfii.-l??1.8llfiJffllSSl? ii.-1.2i-
0
• 1
6
\\f
L ]
t
5
on 29 T
T 19 I I
I? Ifll
LI IT
s
26
11 11 T2
jj|[j|lil{|!
' i
total cardiovascular respiratory cardiovascular
mortality mortality
mortality admissions
8
2
7 2g 2 25
I
2
8 7
13
j
4
8
I
LI
respiratory
admissions
5 27 1
1
1
III '
4
14
4
2
2
4
4
respiratory
symptoms
Figure 3-lla. Estimated excess mortality and morbidity risks per 25 ug/m3 PM25from U.S. and Canadian studies (below).
Results of GAM stringent reanalyses; studies not originally using GAM denoted by •. Multi-city studies denoted by ^".
1 .* Burnett and Goldberg, 2003, 8
Canadian cities
2. Burnett etal, 1997, Toronto •
3. Chock et al, 2000, Pittsburgh •
4. Delfino etal., 1997, Montreal •
5. Delfino etal., 1998, Montreal •
6. Fairley, 2003,, Santa Clara
7. Ito, 2003, Detroit
8. Klemm and Mason, 2000, Atlanta
7. Lipfert et al., 2000a, Philadelphia •
10. Mar etal., 2003, Phoenix
11. Metzger et al., 2003 •
12. Moolgavkar, 2003, Los Angeles Co.
13. Neas etal., 1995, Uniontown •
14. Neas et al., 1996, State College •
15. Ostro et al., 2003, Coachella Valley
16. Ostro et al., 1995, So. California •
17. * Schwartz 2003a, 6 cities overall
18. Schwartz 2003a, Boston
19. Schwartz 2003a, Kingston/Harriman
20. Schwartz 2003a, Portage
21. Schwartz 2003a, Steubenville
22. Schwartz 2003a, St. Louis
23. Schwartz 2003a, Topeka
24. * Schwartz and Neas, 2000, 6 cities
•
25. Sheppard, 2003, Seattle
26. Stieb et al., 2000, St. John •
27. Thurston et al., 1994, Toronto •
28. Tolbert et al., 2000a, Atlanta •
29. Tsai et al., 2000, Newark •
30. Tsai et al., 2000, Elizabeth •
31. Tsai et al., 2000, Camden •
August 2003
5-92
Draft - Do Not Cite or Quote
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
in
c
o
'ro
'o
in
S
i
0.
£
o
in
O)
T3
c
ro
ro
LU
..
o
o -
^~
o
IT) -
i
1
4
4 5 1
4
5 T T 14 7 T 8
1 15 11 12 13 j 3 3 T lie } J J f
•SIIJJ.IIJ.11 •* *
.
v° •
2
2
f
1 5 9 9 9 2
jinm j IK
[ J J l1!!1]
0
19
I 20 T ,
I|
total cardiovascular respiratory cardiovascular respiratory
f
1
7
respira
7
ory
mortality mortality mortality admissions admissions 1 symptoms
Figure 3-llb. Estimated excess mortality and morbidity risks per 25 ug/m3 PM10_25from U.S. and Canadian studies (below).
Results of GAM stringent reanalyses; studies not originally using GAM denoted by •. Multi-city studies denoted by ^".
1 .* Burnett and Goldberg, 2003, 8
Canadian cities
2. Burnett etal, 1997, Toronto •
3. Chock et al, 2000, Pittsburgh •
4. Fairley, 2003,, Santa Clara
5. Ito, 2003, Detroit
6. Klemm and Mason, 2000, Atlanta
7. Lipfert et al., 2000a, Philadelphia •
8. Mar et al., 2003, Phoenix
9. Metzger etal., 2003 •
10. Ostro et al., 2003, Coachella Valley
11. Schwartz 2003a, Boston
12. Schwartz 2003a, Kingston/Harriman
13. Schwartz 2003a, Portage
14. Schwartz 2003a, Steubenville
15. Schwartz 2003a, St. Louis
16. Schwartz 2003a, Topeka
17. * Schwartz andNeas, 2000, 6 cities
•
18. Sheppard, 2003, Seattle
19. Thurston et al., 1994, Toronto •
20. Tolbert et al., 2000a, Atlanta •
August 2003
5-93
Draft - Do Not Cite or Quote
-------
1 Another consideration for the consistency of the reported PM effects is the sensitivity of
2 PM effect estimates to the differing levels of co-pollutants present in various study locations.
3 Such an evaluation supplements the multi-city and single city analyses discussed in earlier
4 sections. In the last review, the Staff Paper examined PM10 effect estimates, to consider whether
5 the reported PM effects can be interpreted appropriately as being likely independent effects
6 attributable to PM, or whether the evidence suggests that the reported PM effects likely result
7 from the influence of other pollutants present in the ambient air in the study locations, either
8 through confounding or effect modification. As discussed in the 1996 Staff Paper, if PM is
9 acting independently, then a consistent association should be observed in a variety of locations
10 of differing levels of co-pollutants. Effect estimates for PM10-mortality associations were plotted
11 against concentrations of gaseous pollutants in the study area, and there was no evidence that
12 associations reported between PM10 and mortality were correlated with copollutant
13 concentrations. (EPA, 1996b, Figure V-3a,b). Similarly, Figure 3-12 shows the reported effect
14 estimates for PM25 and mortality (from single-pollutant models) from U.S. and Canadian studies
15 relative to the levels of O3, NO2, SO2, and CO present in the study locations. As was seen in the
16 last review for PM10, the magnitude and statistical significance of the associations reported
17 between PM2 5 and mortality in these studies show no trends with the levels of any of the four
18 gaseous co-pollutants. While not definitive, these consistent patterns indicate that it is more
19 likely that there is an independent effect of PM25, as seen for PM10 in the 1996 Staff Paper, that
20 is not appreciably modified by differing levels of the gaseous pollutants.
21 In addition, several new "intervention studies" have linked improvements in health with
22 reduction in concentrations of PM and other pollutants. Staff considers these studies to reflect
23 consistency in effect, in that findings of health improvement with reduced pollution is consistent
24 with epidemiological evidence of such associations. Numerous epidemiologic studies in the
25 Utah Valley area reported associations between daily PM10 concentrations and a range of health
26 effects, from respiratory symptoms to mortality, and the investigators observed that respiratory
27 hospital admissions and frequency of school absences decreased during a period when a major
28 PM source (a steel mill) was closed (Pope et al., 1992; 1989). The draft CD also summarizes
29 results from studies in Atlanta, Hong Kong, and East Germany, where policy or regulatory
30 changes resulted in reduced air pollutant concentrations, with subsequent observations of
August 2003 3-94 Draft - Do Not Cite or Quote
-------
1 improvements in health, though it was difficult to distinguish effects of reductions in the
2 individual pollutants. The draft CD concludes: "[tjaken together, these epidemiologic
3 intervention studies tend to support the conclusion that reductions in ambient air pollution
4 (especially PM) exposures resulted in decreased respiratory and cardiovascular health effects"
5 (CD, p. 8-218)
6 Summary. New evidence from multi-city analyses has pointed to several new factors to
7 recognize as potentially resulting in varying effect estimates for different locales. As in the
8 previous review, staff recognizes that there are a number of variables that might be expected to
9 result in different health risk estimates across communities. Overall, however, in evaluating the
10 evidence from the many epidemiology studies, staff observes that there are fairly consistent
11 associations between PM2 5 and a range of effects, from respiratory symptoms to mortality from
12 cardiopulmonary diseases. For PM10_2 5, there is consistency across findings for morbidity,
13 including respiratory symptoms and hospitalization for cardiopulmonary diseases; while there
14 are some significant associations with mortality for heart or lung diseases, there is less evidence
15 and less consistency in associations with total mortality.
August 2003 3-95 Draft - Do Not Cite or Quote
-------
1
2
0 0
J CO
4
5 °
s?
6 LO o.
OS f\j "
7 ?
ro
8 JB
03
9P
^3
10 2^"
11 |
12 ffi
13
14
6
<
2
i.
A
8
16
1
1
. ,
5
f -r 1- H
• • T '' T
I5 0.08
10
14I 1 '
0.10
17
: t
i
o
CO"
0
CN'
0.
CD-
^
I
4
i
M2 3
1 '-
< i
!
-L 1
0.12 0.14 0.16 0.05 0.
16 Ozone (ppm), 4th highest
17
18 g-
19
20 <2
s?
21 88-
22 ?
ro
23 £
ro
24 |o.
26 |
27
28
29
£
8
i
1
|
1
4
5
7
-L .
[ ?
1
i
?
B
8
,
i
.
14
I
-
16
T 17
T " ?I? ,
I f " ^
1 1
--
10 0.15 0.20
8-hour daily maximum NO2 (ppm), 2nd highest 1-hour daily maximum
C
17
!
I
CD
CO'
0
CN'
O.
8.
4
i
s :
-L <
<
i
iP2 '
,!'9
t
i
?
' 1
1
L D
2
4
17
; i\f | j
J_
0.02 0.04 0.06 0.08
SO2 (ppm), 2nd highest 24-hour maximum
0.10
4 6 8 10
CO (ppm), 2nd highest 8-hour daily maximum
August 2003
5-96
Draft - Do Not Cite or Quote
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
0
CO"
O
in o
CD
LU
B
c
3
4
1
-
f '
L J
'f 1
E
16
10
[ "f
17
1
1J 13
•
4
• 12
5 1
JI3-
I:
t
s
,
3
3
2
1
i
)
F
3
17
10
' I
I
18
1 f
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 3-12. 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 Aerometric Information Retrieval System (AIRS)
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; (E)
annual mean SO2 concentration; (F) annual mean NO2 concentration. Study locations are identified below (data in Appendix
A)
1. Chock et al., 2000, Pittsburgh
2. Fairley, 2003, Santa Clara
3. Ito, 2003, Detroit
4. Klemm and Mason, 2000, Atlanta
5. Lipfert et al., 2000a
6. Mar et al., 2003, Phoenix
7. Moolgavkar, 2003
8. Ostro et al., 2003, Coachella Valley
9. Ostro et al., 1995, So. California
10. Schwartz, 2003a, Boston
11. Schwartz, 2003a, Knoxville
12. Schwartz, 2003a, Portage
13. Schwartz, 2003a, St. Louis
14. Schwartz, 2003a, SteubenvUle
15. Schwartz, 2003a, Topeka
16. Tsai et al., 2000, Camden NJ
17. Tsai et al., 2000, Elizabeth NJ
18. Tsai et al., 2000, Newark NJ
August 2003
5-97
Draft - Do Not Cite or Quote
-------
1 3.5.4.2 Coherence
2 The newly available epidemiologic and toxicological evidence substantially reinforces
3 and adds to the coherence in the kinds of health effects associated with PM exposure noted in the
4 last review (EPA, 1996b, V-56). The 1996 Criteria Document provided a qualitative review of
5 the coherence of the health effects associated with both short- and long-term exposure to PM
6 (EPA, 1996a, Tables 13-6 and 13-7). In that review, it was noted that PM is related to a number
7 of logically linked effects of both the respiratory and cardiovascular systems. Respiratory
8 system effects included premature mortality and increased hospital and emergency room
9 admissions for respiratory-related causes, as well as increased respiratory disease and symptoms
10 and decreased lung function. Cardiovascular system effects included premature mortality and
11 limited evidence for increased hospital and emergency room admissions for cardiovascular-
12 related causes. In addition to this observed qualitative coherence, quantitative coherence was
13 also observed in that the increases in respiratory- and cardiovascular-related hospital admissions
14 were more frequently occurring than the increases in mortality for the same causes, based on
15 reported relative risk estimates and baseline population incidence statistics (EPA, 1996a, Table
16 13-8).
17 Qualitative coherence is further supported by the new PM-related effects associations
18 that have now been reported, including increased physicians' visits for respiratory causes and
19 various new cardiovascular-related endpoints. These new findings serve to fill in the spectrum
20 of observed effects from physiological changes that are linked to more serious health outcomes
21 through premature mortality.
22 For respiratory endpoints, there is evidence of coherence in effects for both PM2 5 and
23 PM10_2 5. As seen in Figure 3-1 la for PM2 5, associations with respiratory mortality are positive,
24 but with wide confidence intervals, generally not reaching statistical significance. Associations
25 between PM2 5 and hospital admissions or medical visits for respiratory diseases are all positive,
26 and especially for the more precise findings, a number of effect estimates are statistically
27 significant. Effect estimates for PM2 5 with respiratory symptoms are also all positive, with many
28 also statistically significant. For PM10_2 5, Figure 3-1 Ib shows positive associations with
29 respiratory mortality, hospital admissions or medical visits for respiratory disease and respiratory
30 symptoms, except for the preliminary results from Tolbert et al. (2000b). Some associations,
August 2003 3-98 Draft - Do Not Cite or Quote
-------
1 particularly for respiratory admissions/visits, reach statistical significance, but the evidence is
2 less consistent than that for PM2 5.
3 For cardiovascular effects, the new epidemiologic and toxicologic evidence on discussed
4 in section 3.3.2.3 has added important new evidence for coherence in effects on the
5 cardiovascular system for PM2 5 or PM10. These studies include evidence on more subtle changes
6 in cardiovascular health, such as levels of fibrinogen or C-reactive protein, or heart rate
7 variability. The draft CD observes that the findings of increased levels of fibrinogen or plasma
8 viscosity indicate a potential link between ambient PM, especially PM2 5, exposure and the
9 occurrence of ischemic events, and the increases seen in blood factors such as C-reactive protein
10 provide evidence for inflammatory changes that can be linked with more serious cardiac effects
11 (CD, p. 9-128). For PM2 5, the results on new cardiovascular health endpoints add to evidence of
12 coherence for cardiovascular effects. Associations between PM25 and cardiovascular mortality
13 are all positive, and a number are statistically significant, as are associations with hospital
14 admissions or medical visits for cardiovascular diseases (Figure 3-1 la).
15 Less evidence is available for PM10_2 5 for these new cardiovascular endpoints. Few
16 studies of the more subtle cardiovascular health endpoints have included PM10_2 5, and one study
17 that did reported associations with PM25 but not PM10_25 (Peters et al., 2001). Associations
18 between PM10_25 cardiovascular mortality are positive and some reach statistical significance,
19 and associations with cardiovascular admissions/visits are also generally positive, but the results
20 are less consistent than those for fine-fraction particles (Figure 3-1 Ib). Thus, while there is
21 evidence that suggests associations between coarse-fraction particles and cardiovascular health
22 effects, the evidence is limited and results are not as consistent as those for fine-fraction
23 particles.
24 The new evidence contributes to the quantitative coherence observed in the last review.
25 In the 1996 Staff Paper, results from studies linking short-term PM10 exposure with an array of
26 health endpoints were plotted together, and showed coherence in the findings across the various
27 endpoints. This pattern continues when adding results from the newer studies of short-term PM10
28 exposures. In NMMAPS, 1.1% and 1.6% increases in total and cardiorespiratory mortality,
29 respectively, were reported for a 50 |ig/m3 increase in daily PM10 (in GLM reanalyses) while
30 increases in hospital admissions of 5% (for cardiovascular causes, with a range across other
August 2003 3-99 Draft - Do Not Cite or Quote
-------
1 studies of approximately 3% to 6%) and 8% (for COPD or pneumonia, with a range across other
2 studies of approximately 5% to 25% for respiratory-related causes) were similarly reported. In
3 addition, several new studies have reported associations with visits to physicians' offices for
4 respiratory disease, ranging up to a 75% increase for a 50 |ig/m3 increase in PM10. In the new
5 studies on lung function changes or respiratory symptoms incidence, increases in risk of
6 respiratory-related symptoms range up to over 50% per 50 |ig/m3 increase in daily PM10.
7 Updated baseline incidence rates for respiratory and heart diseases reported in the draft CD
8 (Table 9-13), considered together with these illustrative ranges of effect estimates (and with the
9 ranges shown above in Figures 3-4 through 3-9), continue to show that the quantitative
10 coherence across all PM-related endpoints, especially for PM10 is consistent with expectations
11 (CD, p. 9-128). Staff observes that these ranges are primarily drawn from single-pollutant model
12 results, and the ranges might differ somewhat if including results of multi-pollutant studies,
13 though the coherence across effect categories would not be expected to change.
14 Data now also permit an examination of quantitative coherence for effects associated
15 with PM25 and PM10_25 (Figures 3-1 la and 3-1 Ib). As summarized in the CD, effect estimates
16 for a 25 |ig/m3 change in PM25 with mortality range from 2-3% for total mortality from multi-
17 city studies, 2-6% for total mortality in single-city studies, and 2-20% for mortality from
18 cardiovascular or respiratory diseases. From single-city studies, effect estimates generally range
19 from about 1% to 10% per 25 |ig/m3 PM2 5 for cardiovascular admissions and for respiratory
20 admissions, from about 5% to 25% per 25 |ig/m3 PM2 5. Larger effect estimates are reported with
21 respiratory symptoms, ranging up to 150% increase in symptoms per 25 |ig/m3 PM25. As
22 described earlier in section 3.3, effect estimates with PM10_25 with differing health endpoints are
23 of about the same magnitude as those for PM2 5, though, as has been stated earlier, the findings
24 are less often statistically significant for PM10_2 5 than for PM2 5 and less consistent for mortality.
25 The coherence of PM-related effects is further strengthened by studies demonstrating
26 associations with a range of effects in the same population, as illustrated by studies in a number
27 of locations (EPA, 1996b, V-57). As discussed above, studies in Utah Valley have shown a
28 number of closely related health outcomes associated with PM exposures, including decreased
29 lung function, increased respiratory symptoms, increased medication use in asthmatics, and
30 increased elementary school absences (frequently due to upper respiratory illness) (EPA, 1996b,
August 2003 3-100 Draft - Do Not Cite or Quote
-------
1 V-57). Several recent toxicologic and clinical human exposure studies have used particulate
2 matter collected during the same time periods, and in all studies, more health effects were found
3 with exposure to PM collected when the steel mill was operating than with PM collected while
4 the mill was closed (CD, p. 8-215). In addition, the 1996 Staff Paper observed that a series of
5 epidemiologic studies conducted in cities such as Detroit, Birmingham or Utah Valley reported
6 associations with a range of respiratory and cardiovascular effects (EPA, 1996b, p. V-57). These
7 studies generally reported associations with PM10, and a number of recent studies have expanded
8 upon this evidence with findings of associations with a series of morbidity and mortality effects
9 in individual cities, such as Minneapolis or Pittsburgh (see Chapter 4 discussion of city selection
10 for risk assessment). For PM2 5 (or PMj), there have been findings of significant associations
11 with multiple respiratory health endpoints, including hospital admissions and emergency room
12 visits for asthma, and increased asthma symptoms. In Boston, associations have been reported
13 with mortality, respiratory symptoms and several cardiovascular health indicators, ranging from
14 increased risk of myocardial infarction to decreases in heart rate variability.
15 Summary. These observations strongly underscore coherence in the array of health
16 effects for which associations with exposure to ambient PM have now been reported, from subtle
17 changes in cardiac and lung function to increased aggravation of heart and lung diseases to
18 increases in utilization of health care facilities to increased mortality from cardiorespiratory
19 diseases. PM2 5 has been associated with effects on both the respiratory and cardiovascular
20 systems, ranging from the more subtle effects, such as changes in lung function or levels of C-
21 reactive protein, to serious effect such as mortality. For PM10_2 5, there is evidence for coherence
22 in effects on the respiratory system, with less consistent evidence for effects on the
23 cardiovascular system. The expanded evidence for coherence in effects, along with previously
24 described observations of marked consistency in the results of recent studies and those available
25 in the last review, provide increased support for a causal link between PM and effects on the
26 cardiovascular and respiratory systems (CD, p. 9-67).
27
August 2003 3-101 Draft - Do Not Cite or Quote
-------
1 REFERENCES
2
3 Abbey, D. E.; Mills, P. K.; Petersen, F. F.; Beeson, W. L. (1991) Long-term ambient concentrations of total
4 suspended particulates and oxidants as related to incidence of chronic disease in California Seventh-Day
5 Adventists. Environ. Health Perspect. 94: 43-50.
6
7 Abbey, D. E.; Burchette, R. I; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
8 paniculate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med. 158:
9 289-298.
10
11 Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, L.; Yang, J. X. (1999) Long-
12 term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit.
13 Care Med. 159:373-382.
14
15 Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
16 Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2001) Acute effects of paniculate air pollution
17 on respiratory admissions: results from APHEA 2 project. Am. J. Respir. Crit. Care Med. 164: 1860-1866.
18
19 Atkinson, R. W.; Anderson, H. R.; Sunyer, J.; Ayres, J.; Baccini, M.; Vonk, J. M.; Boumghar, A.; Forastiere, F.;
20 Forsberg, B.; Touloumi, G.; Schwartz, J.; Katsouyanni, K. (2003) Acute effects of paniculate air pollution
21 on respiratory admissions. In: Revised analyses of time-series studies of air pollution and health. Special
22 report. Boston, MA: Health Effects Institute; pp. 81-84. Available: http://www.healtheffects.org/news.htm
23 [16 May, 2003].
24
25 Avol, E. L.; Gauderman, W. J.; Tan, S. M.; London, S. J.; Peters, J. M. (2001) Respiratory effects of relocating to
26 areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
27
28 Beeson, W. L.; Abbey, D. E.; Knutsen, S. F. (1998) Long-term concentrations of ambient air pollutants and incident
29 lung cancer in California adults: results from the AHSMOG study. Environ. Health Perspect. 106: 813-823.
30
31 Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms
32 and pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
33
34 Braga, A. L. F.; Zanobetti, A.; Schwartz, J. (2001) The lag structure between paniculate air pollution and respiratory
35 and cardiovascular deaths in ten U.S. cities. J. Occup Environ. Med. 43:927-933.
36
37 Brain, J. D.; Long, N. C.; Wolfthal, S. F.; Dumyahn, T.; Dockery, D. W. (1998) Pulmonary toxicity in hamsters of
38 smoke particles from Kuwaiti oil fires. Environ. Health Perspect. 106: 141-146.
39
40 Brunekreef. B. (1997) Air pollution and life expectancy: is there a relation? Occup. Environ. Med. 54: 781-784.
41
42 Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of paniculate size and chemistry in the
43 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
44 Environ. Health Perspect. 105:614-620.
45
46 Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
47 (2000) Association between paniculate- and gas-phase components of urban air pollution and daily
48 mortality in eight Canadian cities. Inhalation Toxicol. 12(suppl. 4): 15-39.
49
50 Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian
51 cities. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
52 Health Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
53
August 2003 3-102 Draft - Do Not Cite or Quote
-------
1 Carrothers, T. I; Evans, J. S. (2000) Assessing the impact of differential measurement error on estimates of fine
2 particle mortality. J. Air Waste Manage. Assoc. 50:65-74.
3
4 Chay, K. Y.; Greenstone, M. (200 Ib) The impact of air pollution on infant mortality: evidence from geographic
5 variation in pollution shocks induced by a recession. Berkeley, CA: University of California, Berkeley,
6 Department of Economics. Available: http://elsa.berkeley.edu/users/kenchay/ [30 May, 2003].
7
8 Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air paniculate pollution and hospital admissions for
9 chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12: 281-298.
10
11 Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
12 pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
13
14 Choudhury, A. H.; Gordian, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PM10 air
15 pollution. Arch. Environ. Health 52: 113-117.
16
17 Clarke, R. W.; Catalano, P.; Coull, B.; Koutrakis, P.; Krishna Murthy, G. G.; Rice, T.; Godleski, J. J. (2000) Age-
18 related responses in rats to concentrated urban air particles (CAPs). Inhalation Toxicol. 12:(Suppl 1): 73-
19 84.
20
21 Costa, D. L.; Dreher, K. L. (1997) Bioavailable transition metals in paniculate matter mediate cardiopulmonary
22 injury in healthy and compromised animal models. Environ. Health Perspect. Suppl. 105(5): 1053-1060.
23
24 Delfino, R. J.; Coate, B. D.; Zeiger, R. S.; Seltzer, J. M.; Street, D. H.; Koutrakis, P. (1996) Daily asthma severity in
25 relation to personal ozone exposure and outdoor fungal spores. Am. J. Respir. Crit. Care Med. 154:
26 633-641.
27
28 Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M. R. (1997) Effects of air pollution
29 emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Crit. Care Med.
30 155: 568-576.
31
32 Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. G. (1998) Symptoms in pediatric asthmatic and air pollution:
33 differences in effects by symptom severity, anti-inflammatory medication use and paniculate averaging
34 time. Environ. Health Perspect. 106:751-761.
35
36 Dockery, D. W.; Spengler, J. D.; Neas, L. M.; Speizer, F. E.; Ferris, B. G., Jr.; Ware, J. H.; Brunekreef, B. (1989)
37 An epidemiologic study of respiratory health status and indicators of indoor air pollution from combustion
3 8 sources. In: Harper, J. P., ed. Combustion processes and the quality of the indoor environment: transactions
39 of an international specialty conference; September 1988; Niagara Falls, NY. Pittsburgh, PA: Air & Waste
40 Management Association; pp. 262-271. (A&WMA transactions series: TR-15).
41
42 Dockery, D. W.; Schwartz, J.; Spengler, J. D. (1992) Air pollution and daily mortality: associations with particulates
43 and acid aerosols. Environ. Res. 59: 362-373.
44
45 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.
46 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
47 329: 1753-1759.
48
49 Dockery, D. W.; Cunningham, J.; Damokosh, A. L; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
50 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
51 symptoms. Environ. Health Perspect. 104: 500-505.
52
53 Dominici, F.; Daniels, M.; McDermott, A.; Zeger, S. L.; Samet, J. M. (2003) Shape of the exposure-response
54 relation and mortality displacement in the NMMAPS database. In: Revised analyses of time-series studies
August 2003 3-103 Draft - Do Not Cite or Quote
-------
1 of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 91-96. Available:
2 http://www.healtheffects.org/news.htm [16 May, 2003].
3
4 Dye, J. A.; Lehmann, J. R.; McGee, J. K.; Winsett, D. W.; Ledbetter, A. D.; Everitt, J. I.; Ohio, A. I; Costa, D. L.
5 (2001) Acute pulmonary toxicity of paniculate matter filter extracts in rats: coherence with epidemiological
6 studies in Utah Valley residents. Environ. Health Perspect. 109(suppl. 3): 395-403.
7
8 EPA. (1987) National ambient air quality for paniculate matter; final rule. 62 FR 38651. July 18, 1997
9
10 EPA. (1996a) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC: National Center for
11 Environmental Assessment-RTF Office; report no. EPA/600/P-95/00laF-cF. 3v
12
13 EPA. (1996b) Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment of
14 Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office of
15 Air Quality Planning and Standards; report no. EPA-452YR-96-013.
16
17 EPA. (1996c) Air Quality Criteria for Ozone and Related Photochemical Oxidants. Research Triangle Park, NC:
18 National Center for Environmental Assessment-RTF Office; report no. EPA/600/P-93/004aF-cF. 3v
19
20 EPA. (2000a) Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: National Center for
21 Environmental Assessment-RTF Office; report no. EPA/600/P-99/00IF.
22
23 EPA. (2002) Health assessment document for diesel exhaust emissions. Washington, DC: Office of Research and
24 Development, National Center for Environmental Assessment; report no. EPA/600/8-90/057F
25
26 EPA. (2003) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC: National Center for
27 Environmental Assessment-RTF Office; report no. EPA/600/P-99/002aD.
28
29 Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
30 Perspect. 107:637-641.
31
32 Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
3 3 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp.
34 97-106. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
35
36 Friedman, M. S.; Powell, K. E.; Hutwagner, L.; Graham, L. M.; Teague, W. G. (2001) Impact of changes in
37 transportation and commuting behaviors during the 1996 summer Olympic games in Atlanta on air quality
38 and childhood asthma. JAMA J. Am. Med. Assoc. 285: 897-905.
39
40 Gauderman, W. J.; McConnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
41 Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and
42 lung function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
43
44 Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
45 Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
46 lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
47 Med. 166: 76-84.
48
49 Ohio, A. J.; Stoneheurner, J.; McGee, J. K.; Kinsey, J. S. (1999a) Sulfate content correlates with iron concentration
50 in ambient air pollution particles. Inhalation Toxicol. 11:293-307.
51
52 Ohio, A. J.; Stoneheurner, J.; Dailey, L. A.; Carter, J. D. (1999b) Metals associated with both the water-soluble and
53 insoluble fractions of an ambient air pollution particle catalyze an oxidative stress. Inhalation Toxicol.
54 11:37-49.
August 2003 3-104 Draft - Do Not Cite or Quote
-------
1 Ohio, A. I; Kim, C.; Devlin, R. B. (2000) Concentrated ambient air particles induce mild pulmonary inflammation
2 in healthy human volunteers. Am. J. Respir. Crit. Care Med. 162:981-988.
3
4 Ohio, A. J.; Devlin, R. B. (2001) Inflammatory lung injury after bronchial instillation of air pollution particles. Am J.
5 Respir. Crit. Care Med. 164:704-708.
6
7 Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, L.; Allen, G.; Verrier, M; Cherry, R.;
8 Verrier,R. (2000) Ambient pollution and heart rate variability. Circulation 101:1267-1273.
9
10 Gold, D. R.; Schwartz, J.; Litonjua, A.; Verrier, R.; Zanobetti, A. (2003) Ambient pollution and reduced heart rate
11 variability. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
12 MA: Health Effects Institute; pp. 107-112. Available: http://www.healtheffects.org/news.htm [16 May,
13 2003].
14
15 Gordon, T.; Nadziejko, C.; Chen, L. C.; Schlesinger, R. (2000) Effects of concentrated ambient particles in rats and
16 hamsters: an exploratory study. Cambridge, MA: Health Effects Institute; research report no. 93
17
18 Gwynn, R. C.; Burnett, R. T.; Thurston, G. D. (2000) A time-series analysis of acidic paniculate matter and daily
19 mortality and morbidity in the Buffalo, New York, region. Environ. Health Perspect. 108: 125-133.
20
21 Health Effects Institute. (2003a) Revised analyses of the National Morbidity, Mortality, and Air Pollution Study
22 (NMMAPS), part II. In: Revised analyses of time-series studies of air pollution and health. Special report.
23 Boston, MA: Health Effects Institute; pp. 9-72. Available: http://www.healtheffects.org/news.htm [16 May,
24 2003].
25
26 Health Effects Institute. (2003b) Commentary on revised analyses of selected studies. In: Revised analyses of
27 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
28 pp. 255-290. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
29
30 Ito, K.; Thurston, G. D. (1996) Daily PM10/mortality associations: an investigation of at-risk subpopulations. J.
31 Exposure Anal. Environ. Epidemiol.6:79-95.
32
33 Ito, K. (2003) Associations of paniculate matter components with daily mortality and morbidity in Detroit,
3 4 Michigan. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
35 MA: Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/news.htm [16 May,
36 2003].
37
38 Janssen, H. A. J.; Schwartz, J.; Zanobetti, A.; Suh, H. H. (2002) Air conditioning and source-specific particles as
3 9 modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ. Health Perspect.
40 110:43-49.
41
42 Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM-10 associations in Los Angeles.
43 Inhalation Toxicol. 7:59-69.
44
45 Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
46 and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50: 1433-1439.
47
48 Klemm, R. J.; Mason, R. M., Jr.; Heilig, C. M.; Neas, L. M.; Dockery, D. W. (2000) Is daily mortality associated
49 specifically with fine particles? Data reconstruction and replication of analyses. J. Air Waste Manage.
50 Assoc. 50:1215-1222.
51
52 Klemm, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
53 of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
54 pp. 165-172. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
August 2003 3-105 Draft - Do Not Cite or Quote
-------
1 Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J,; Jerrett, M.; Abrahamowicz, M.; White,
2 W. H. (2000) Reanalysisof the Harvard Six Cities Study and the American Cancer Society Study of
3 paniculate air pollution and mortality. A special report of the Institute's particle epidemiology reanalysis
4 project. Cambridge, MA: Health Effects Institute.
5
6 Laden, F.; Neas, L. M.; Dockery, D. W.; Schwartz, J. (2000) Association of fine paniculate matter from different
7 sources with daily mortality in six U.S. cities. Environ. Health Perspect. 108:941-947.
8
9 Lee, J.-T.; Schwartz, J. (1999) Reanalysis of the effects of air pollution on daily mortality in Seoul, Korea:
10 a case-crossover design. Environ. Health Perspect. 107: 633-636.
11
12 Lee, J.-T.; Shin, D.; Chung, Y. (1999) Air pollution and daily mortality in Seoul and Ulsan, Korea. Environ. Health
13 Perspect. 107: 149-154.
14
15 Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
16 Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2002) Short term effects of paniculate
17 air pollution on cardiovascular diseases in eight European cities. J. Epidemiol. Community Health
18 56:773-779.
19
20 Le Tertre, A.; Medina, S.; Samoli, E.; Forsberg, B.; Michelozzi, P.; Boumghar, A.; Vonk, J. M.; Bellini, A.;
21 Atkinson, R.; Ayres, J. G.; Sunyer, J.; Schwartz, J.; Katsouyanni, K. (2003) Short-term effects of paniculate
22 air pollution on cardiovascular diseases in eight European cities. In: Revised analyses of time-series studies
23 of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 173-176. Available:
24 http://www.healtheffects.org/news.htm [16 May, 2003].
25
26 Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
27 admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
28
29 Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and size-
30 classified paniculate matter. J. Air Waste Manage. Assoc. 50:1501-1513.
31
32 Lipfert, J. W.; Perry, H. M., Jr.; Miller, J. P.; Baty, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) the Washington
33 University-EPRI veteran's cohort mortality study: preliminary results. Inhalation Toxicol. 12(Suppl. 4):41-
34 73.
35
36 Lipfert, F. W.; Zhang, J.; Wyzga, R. E. (2000c) Infant mortality and air pollution: a comprehensive analysis of U.S.
37 data for 1990. J. Air Waste Manage. Assoc. 50: 1350-1366.
38
39 Lipfert, F. W.; Morris, S. C. (2002) Temporal and spacial relationships between age-specific mortality and ambient
40 air quality in the United States: preliminary results for counties, 1960-97. Occup. Environ. Med. 59:
41 156-174.
42
43 Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
44 mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report 95.
45
46 Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
47 California. Environ. Health Perspect. 105: 216-222.
48
49 Magari, S. R.; Hauser, R.; Schwartz, J.; Williams, P. L.; Smith, T. J.; Christiani, D. C. (2001) Association of heart
50 rate variability with occupational and environmental exposure to paniculate air pollution. Circulation
51 104:986-991.
52
53 Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
54 Phoenix, 1995-1997. Environ. Health Perspect. 108:347-353.
August 2003 3-106 Draft - Do Not Cite or Quote
-------
1 Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003) Air pollution and cardiovascular
2 mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
3 Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
4 http://www.healtheffects.org/news.htm [16 May, 2003].
5
6 Mauderly, J.; Neas, L.; Schlesinger, R. (1998) PM monitoring needs related to health effects. In: Atmospheric
7 observations: helping build the scientific basis for decisions related to airborne paniculate matter; Report of
8 the PM measurements research workshop, July 22-23, 1998. Available from "PM Measurements Report",
9 Health Effects Institute, 955 Massachusetts Ave., Cambridge MA 02139.
10
11 McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gauderman, W. J.; Margolis, H. G.;
12 Lurmann, F.; Thomas, D. C.; Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
13 California children with asthma. Environ. Health Perspect. 107:757-760.
14
15 McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of
16 mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers. J.
17 Exposure Anal. Environ. Epidemiol. 10:427-436.
18
19 Metzger, K. B.; Tolbert, P. E.; Klein, M.; Peel, J. L.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
20 Frumkin, H. (2003) Ambient air pollution and cardiovascular emergency department visits in Atlanta,
21 Georgia, 1993-2000. Epidemiology: in press.
22
23 Molinelli, A. R.; Madden, M. C.; McGee, J. K.; Stonebuerner, J. G.; Ohio, A. J. (2002) Effect of metal removal on
24 the toxicity of airborne particulate matter from the Utah Valley. Inhalation Toxicol. 14: 1069-1086.
25
26 Moolgavkar, S. H. (2000a) Air pollution and mortality in three U.S. counties. Environ. Health Perspect. 108:777-
27 784.
28
29 Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
30 metropolitan areas. J. Air Waste Manage. Assoc. 50:271-280.
31
32 Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
33 metropolitan areas of the United States. Inhalation Toxicol. 12(Suppl. 4):75-90.
34
3 5 Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
36 In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
37 Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
38
39 Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
40 evidence of an increased effect at low temperatures. Environ. Health Perspect. 106: 649-653.
41
42 Murray, C. J.; Nelson, C. R. (2000) State-space modeling of the relationship between air quality and mortality. J. Air
43 Waste Manage. Assoc. 50: 1075-1080.
44
45 Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations
46 and exposure to air pollution. Public Health Rep. 114: 135-148.
47
48 Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air
49 pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
50 141: 111-122.
51
52 Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
53 determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807.
54
August 2003 3-107 Draft - Do Not Cite or Quote
-------
1 Neas, L. M; Dockery, D. W.; Koutrakis, P.; Speizer, F. E. (1999) Fine particles and peak flow in children: acidity
2 versus mass. Epidemiology 10:550-553.
3
4 Neas, L. M.; Schwartz, I; Dockery, D. (1999) A case-crossover analysis of air pollution and mortality in
5 Philadelphia. Environ. Health Perspect. 107: 629-631.
6
7 Ostro, B. (1995) Fine paniculate air pollution and mortality in two Southern California counties. Environ. Res.
8 70:98-104.
9
10 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
11 Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10:412-419.
12
13 Ostro, B.; Lipsett, M.; Mann, J.; Braxtron-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
14 African-American children in Los Angeles. Epidemiology 12: 200-208.
15
16 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coachella Valley,
17 California. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
18 MA: Health Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/news.htm [16 May,
19 2003].
20
21 Ozkaynak, H.; Xue, J.; Zhou, H.; Raizenne, M. (1996) Associations between daily mortality and motor vehicle
22 pollution in Toronto, Canada. Boston, MA: Harvard University School of Public Health, Department of
23 Environmental Health; March 25.
24
25 Pekkanen, J.; Timonen, K.L.; Ruuskanen, J.; Reponen, A.; Mirme, A. (1997) Effects of ultrafine and fine particles in
26 urban air on peak expiratory flow among children with asthmatic symptoms. Environ. Res. 74:24-33.
27
28 Peters, A.; Wichmann, H. E.; Tuch, T.; Heinrich, J.; Heyder, J. (1997) Respiratory effects are associated with the
29 number of ultrafine particles. Am. J. Respir. Crit. Care Med. 155: 1376-1383.
30
31 Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
32 Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities
3 3 with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit.
34 Care Med. 159: 760-767.
35
36 Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
37 Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
3 8 differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
39 159:768-775.
40
41 Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.;
42 Monahan, K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology
43 11: 11-17.
44
45 Peters, A.; Dockery, D. W.; Muller, J. E.; Mittleman, M. A. (2001) Increased paniculate air pollution and the
46 triggering of myocardial infarction. Circulation 103:2810-2815.
47
48 Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah Valley. Arch.
49 Environ. Health 47:211-217.
50
51 Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999) Paniculate air pollution and daily mortality on Utah's Wasatch
52 Front. Environ. Health Perspect. 107:567-573.
53
August 2003 3-108 Draft - Do Not Cite or Quote
-------
1 Pope, C. A., Ill; Thun, M. I; Namboodiri, M. M; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
2 (1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
3 Respir. Crit. Care Med. 151:669-674.
4
5 Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
6 cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. J. Am. Med. Assoc.
7 287:1132-1141.
8
9 Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer, F.
10 E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
11 Perspect. 104: 506-514.
12
13 Samet, J. M.; Zeger, S. L.; Domenici, F.; Curriero, F.; Coursac, L; Dockery, D.W.; Schwartz, J.; Zanobetti, A.
14 (2000a) The national morbidity, mortality, and air pollution study. Part I: methods and methodological
15 issues. Cambridge, MA: Health Effects Institute: research report no. 94.
16
17 Samet, J. M.; Zeger, S. L.; Domenici, F.; Curriero, F.; Coursac, L; Dockery, D.W.; Schwartz, J.; Zanobetti, A.
18 (2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
19 pollution in the United States. Cambridge, MA: Health Effects Institute: research report no. 94.
20
21 Samet, J. M.; Domenici, F.; Curriero, F.; Coursac, L; Zeger, S. L. (2000c) Fine paniculate air pollution and
22 mortality in20 U.S. cities, 1987-1994. N. Engl. J. Med. 343:1742-9.
23
24 Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137:1136-1147.
25
26 Schwartz, J. (1994) Air pollution and hospital admissions for the elderly in Detroit, Michigan. Am J. Respir. Crit
27 Care Med. 150:589-598.
28
29 Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
30 8:371-377.
31
32 Schwartz, J. (1999) Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology
33 10: 17-22.
34
3 5 Schwartz, J. (2000) Harvesting and long term exposure effects in the relation between air pollution and mortality.
36 Am. J. Epidemiol. 151:440-448.
37
3 8 Schwartz, J. (2003a) Daily deaths associated with air pollution in six US cities and short-term mortality displacement
39 in Boston. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
40 MA: Health Effects Institute; pp. 219-226. Available: http://www.healtheffects.org/news.htm [16 May,
41 2003].
42
43 Schwartz, J. (2003b) Airborne particles and daily deaths in 10 US cities. In: Revised analyses of time-series studies
44 of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 211-218. Available:
45 http://www.healtheffects.org/news.htm [16 May, 2003].
46
47 Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit,
48 Michigan. Am. J. Epidemiol 142:23-35.
49
50 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J.
51 Air Waste Manage. Assoc. 46:927-939.
52
53 Schwartz, J.; Spix, C.; Touloumi, G.; Bacharova, L.; Barumamdzadeh, T.; le Tertre, A.; Piekarksi, T.; Ponce de
54 Leon, A.; Ponka, A.; Rossi, G.; Saez, M.; Schouten, J. P. (1996b) Methodological issues in studies of air
August 2003 3-109 Draft - Do Not Cite or Quote
-------
1 pollution and daily counts of deaths or hospital admissions. In: St Leger, S., ed. The APHEA project. Short
2 term effects of air pollution on health: a European approach using epidemiological time series data.
3 J. Epidemiol. Community Health 50(suppl. 1): S3-S11.
4
5 Schwartz, J.; Norris, G.; Larson, T.; Sheppard, L.; Clairborne, C.; Koenig, J. (1999) Episodes of high coarse
6 particles concentrations are not associated with increased mortality. Environ. Health Perspect. 107:339-
7 342.
8
9 Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
10 respiratory health effects in schoolchildren. Epidemiology 11:6-10.
11
12 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
13 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
14
15 Sheppard, L.; Damian, D. (2000) Estimating short-term PM effects accounting for surrogate exposure measurements
16 from ambient monitors. Environmetrics 11: 675-687.
17
18 Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
19 1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
20 MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/news.htm [16 May,
21 2003].
22
23 Smith, R. L.; Davis, J. M.; Speckman, P. (1999) Assessing the human health risk of atmospheric particles. In:
24 Environmental Statistics: Analysing Data for Environmental Policy: proceedings of a Novartis Foundation
25 Symposium; May 1998; London, England. New York, NY: Wiley. [Novartis Foundation Symposium 220,
26 pp. 59-79].
27
28 Soukup, J. M.; Ohio, A. J.; Becker, S. (2000) Soluble components of Utah Valley paniculate pollution alter alveolar
29 macrophage function in vivo and in vitro. Inhalation Toxicol. 12:401-414.
30
31 Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
32 Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in
33 Saint John, Canada. J. Exposure Anal. Environ. Epidemiol.: 10: 461-477.
34
3 5 Stolzel, M.; Peters, A.; Wichmann, H.-E. (2003) Daily mortality and fine and ultrafine particles in Erfurt, Germany.
36 In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
37 Effects Institute; pp. 231-240. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
38
39 Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne paniculate matter on daily
40 death counts. Environ Health Perspect. 103:490-497.
41
42 Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
43 summertime haze air pollution in Toronto, Ontario: Consideration of the role of acid aerosols. Environ. Res.
44 65:271-290.
45
46 Tiitanen, P.; Timonen, K.L.; Ruuskanen, J.; Mirme, A.; Pekkanen, J. (1999) Fine particulate air pollution,
47 resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13:266-273.
48
49 Tolbert, P. G.; Klein, M.; Metzger, K. B.; Peel, J.; Flanders, W. D.; Todd, K.; Mulholland, J. A.; Ryan, P. B.;
50 Frumkin, H. (2000a) Interim results of the study of particulates and health in Atlanta (SOPHIA). J.
51 Exposure Anal. Environ. Epidemiol. 10:446-460.
52
August 2003 3-110 Draft - Do Not Cite or Quote
-------
1 Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. I; Carlin, B. P.; Klein, M;
2 Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frumkin, H.; Ryan, P. B.; White, M. C. (2000b) Air quality and
3 pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151: 798-810.
4
5 Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
6 chemical composition of airborne paniculate matter. Inhalation Toxicol. 12(suppl.): 121-135.
7
8 Villeneuve, P. J.; Goldberg, M. S.; Krewski, D.; Burnett, R. T.; Chen, Y. (2002) Fine paniculate air pollution and
9 all-cause mortality within the Harvard six-cities study: variations in risk by period of exposure. Ann.
10 Epidemiol. 12: 568-576.
11
12 Watkinson, W. P.; Campen, M. J.; Dreher, K. L.; Su, W.-Y.; Kodavanti, U. P.; Highfill, J. W.; Costa, D. L. (2000)
13 Thermoregulatory effects following exposure to paniculate matter in healthy and cardiopulmonary-
14 compromised rats. J. Therm. Biol. 25:131-137.
15
16 Wichmann, H.-E.; Spix, C.; Tuch, T.; Wolke, G.; Peters, A.; Heinrich, J.; Kreyling, W.G.; Heyder, J. (2000) Daily
17 mortality and fine and ultrafine particles in Erfurt, Germany. Part I: Role of particle number and particle
18 mass. Cambridge, MA: Health Effects Institute: research report no. 98.
19
20 Woodruff, T. J.; Grille, J.; Schoendorf, K. C. (1997) The relationship between selected causes of postneonatal infant
21 mortality and paniculate air pollution in the United States. Environ. Health Perspect. 105:608-612.
22
23 Zanobetti, A.; Schwartz, J. (2000) Race, gender, and social status as modifiers of the effects of PM10 on mortality. J.
24 Occup. Environ. Med. 42:469-474.
25
26 Zanobetti, A.; Schwartz, J. (2001) Are diabetics more susceptible to the health effects of airborne particles? Am. J.
27 Respir. Crit. Care Med. 164: 831-833.
28
29 Zanobetti, A.; Schwartz, J. (2003) Airborne particles and hospital admissions for heart and lung disease. In: Revised
30 analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
31 Institute; pp. 241-248. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
32
33 Zanobetti, A.; Schwartz, J.; Dockery, D. W. (2000a) Airborne particles are a risk factor for hospital admissions for
34 heart and lung disease. Environ. Health Perspect. 108: 1071-1077.
35
36 Zanobetti, A.; Wand, M. P.; Schwartz, J.; Ryan, L. M. (2000b) Generalized additive distributed lag models:
37 quantifying mortality displacement. Biostatistics 1:279-292.
38
39 Zeger, S. L.; Dominici, F.; Samet, J. (1999) Harvesting-resistant estimates of air pollution effects on mortality.
40 Epidemiology 10:171-175.
41
42 Zeger, S. L.; Thomas, D.; Dominici, F.; Samet, J. M.; Schwartz, J.; Dockery, D.; Cohen, A. (2000) Exposure
43 measurement error in time-series studies of air pollution: concepts and consequences. Environ. Health
44 Perspect. 108: 419-426.
45
August 2003 3-111 Draft - Do Not Cite or Quote
-------
-------
4. CHARACTERIZATION OF HEALTH RISKS
1 4.1 INTRODUCTION
2 This chapter describes and presents the initial results from an updated PM health risk
3 assessment that is being conducted for EPA's current review of the PM NAAQS. This updated
4 risk assessment builds upon the methodology used in the more limited PM risk assessment
5 (summarized below) that was conducted as part of EPA's prior PM NAAQS review. This
6 updated assessment focuses on (1) the risks of mortality, morbidity, and symptoms associated
7 with recent ambient PM25 PM10_25, and PM10 levels, (2) the risk reductions associated with just
8 meeting the current suite of PM2 5 NAAQS, and (3) the risk reductions associated with just
9 meeting various alternative PM2 5 standards and a range of PM10_2 5 standards, consistent with
10 ranges of standards recommended by staff for consideration. The third component of the risk
11 assessment, focusing on alternative PM2 5 and PM10_2 5 standards, will be included in the next
12 draft of this Staff Paper, based on the preliminary staff recommendations on possible alternative
13 standards contained in Chapter 6 of this draft Staff Paper and taking into account comments
14 received from CASAC and the public on those preliminary recommendations.
15 As discussed in Chapter 2, the fact that the sources and composition of PM2 5 and PM10_2 5
16 are largely distinct, along with the new health effects evidence discussed in Chapter 3, supports
17 the recommendation in the 1996 Staff Paper that fine- and coarse-fraction particles should be
18 considered as separate pollutants. At that time, a number of health studies indicated differences
19 in health effects between fine- and coarse-fraction particles, and suggested that serious health
20 effects, such as premature mortality, were more closely associated with fine-fraction particles.
21 The new studies, summarized in Chapter 8 of the draft CD, continue to show associations
22 between serious health effects, including premature mortality, and fine-fraction PM, but they
23 also offer new evidence indicating associations between coarse-fraction PM and health effects.
24 In June 2001, O AQPS released a draft document, PM NAAQS Risk Analysis Scoping
25 Plan, (EPA, 200Ic) describing EPA's overall plan for conducting the PM health risk assessment
26 for the current review. The CASAC PM Panel provided feedback on this draft plan in a
27 consultation held July 24, 2001, and the Agency also received comments from the general
August 2003 4-1 Draft - Do Not Quote or Cite
-------
1 public. In January 2002, EPA released a draft document, Proposed Methodology for Paniculate
2 Matter Risk Analyses for Selected Urban Areas, (Abt Associates, 2002a) for public and CAS AC
3 review. This draft document described EPA's plans to conduct a risk assessment for PM2 5-
4 related risks for several health endpoints, including mortality, hospital admissions, and
5 respiratory symptoms, and PM10_2 5-related risks for hospital admissions and respiratory
6 symptoms. The CASAC PM Panel discussed this draft document in a February 27, 2002
7 teleconference and provided its comments in a May 23, 2002 Advisory letter to EPA's
8 Administrator (Hopke, 2002). OAQPS also received several comments from the public. In its
9 May 23, 2002 Advisory, the CASAC PM Panel "concluded that the general methodology as
10 described in the report is appropriate. ... Thus, the general framework of the approach is the
11 sensible approach to this risk analysis" (Hopke, 2002). Among its comments, the CASAC Panel
12 suggested extending the risk assessment to include PM10 as an indicator in the risk assessment,
13 since many health studies used PM10 as an indicator and PM10 air quality data are generally
14 available (Hopke, 2002). Risks associated with PM10 ambient levels are likely to reflect the
15 contribution of PM25, PM10_25, or some combination of both depending on the relative
16 composition of PM in various urban areas. The CASAC Panel also offered a number of
17 comments related to the scope and details of the proposed risk assessment.
18 In response to a request from CASAC to provide additional details about the proposed
19 scope of the PM10_25 and PM10 components of the planned risk assessment, in April 2003 EPA
20 released a draft memorandum (Abt, 2003a) to the CASAC and the public addressing this topic.
21 On May 1, 2003, the CASAC PM Panel held a consultation with EPA to provide advice on staff
22 plans for conducting the PM10_25 and PM10 components of the health risk assessment. OAQPS
23 has carefully considered these comments in preparing the risk assessments described in this
24 Chapter. These comments also are addressed further in a draft technical report (Abt Associates,
25 2003b) which provides additional details about the PM risk assessment.1
26
JWe hereafter refer to the "PM risk assessment" unless reference to a specific PM indicator (e.g., PM2 5) is
required. The current PM risk assessment considers each of three PM indicators - PM2 5, PM10_2 5, and PM10 - in
turn.
August 2003 4-2 Draft - Do Not Quote or Cite
-------
1 4.1.1 Goals for Updated PM Risk Assessment
2 The goals of the updated PM risk assessment are (1) to develop a better understanding of
3 the influence of various inputs and assumptions on the risk estimates and (2) to gain insights into
4 the nature of the risks associated with exposures to ambient PM. In addition, the risk assessment
5 provides a rough sense of the potential magnitude of PM-associated mortality and morbidity
6 associated with current PM2 5, PM10_2 5, and PM10 levels and with attaining the current suite of
7 PM2 5 NAAQS (as well as alternative PM2 5 and PM10_25 standards identified as part of this
8 review) in specific urban areas. The staff recognizes that due to the many sources of uncertainty
9 inherent in conducting the PM risk assessment, the resulting PM risk estimates should not be
10 viewed as precise measures of the health impacts now occurring or anticipated to occur in the
11 future in any given location or nationally. Further, the staff recognizes that the role of the risk
12 assessment in this standards review will necessarily be limited by the significant uncertainties
13 associated with this assessment, discussed in section 4.2.7 below.
14
15 4.1.2 Summary of Risk Assessment Conducted During Prior PM NAAQS Review
16 For the prior review, EPA conducted a health risk assessment that estimated population
17 risk for two defined urban study areas: Philadelphia and Los Angeles counties. The PM health
18 risk model combined information about daily PM air quality for these two study areas with
19 estimated concentration-response (C-R) functions derived from epidemiological studies and
20 baseline health incidence data for specific health endpoints to derive estimates of the annual
21 incidence of specific health effects occurring under "as is" air quality.2 Since site-specific
22 relative risks were not available for all endpoints in both locations (and in the absence of more
23 information concerning which individual studies might best characterize the health risk in a
24 given location), a form of meta analysis (referred to as a "pooled analysis") was conducted
25 which combined the results of the studies that met specified criteria. The assessment also
26 examined the reduction in estimated incidence that would result upon just attaining the existing
27 PM10 standards and several sets of alternative PM2 5 standards. In addition, the assessment
2"As is" PM concentrations are defined here as a recent year of air quality.
August 2003 4-3 Draft - Do Not Quote or Cite
-------
1 included sensitivity analyses and integrated uncertainty analyses to better understand the
2 influence of various inputs and assumptions on the risk estimates. The methodological approach
3 followed in conducting the prior risk assessment is described in Chapter 6 of the 1996 Staff
4 Paper (EPA, 1996b) and in several technical reports (Abt Associates, 1996; Abt Associates,
5 1997a,b) and publications (Post et al., 2000; Deck et al., 2001).
6 Summarized below are the key observations resulting from the prior risk assessment
7 which were most pertinent to the decision on the PM NAAQS, as well as several important
8 caveats and limitations associated with that assessment:
9 • EPA placed greater weight on the overall qualitative conclusions derived from the health
10 effect studies - that PM air pollution is likely causing or contributing to significant
11 adverse effects at levels below those permitted by the existing PM10 standards - than on
12 the specific C-R functions and quantitative risk estimates derived from them. The
13 quantitative risk estimates included significant uncertainty and, therefore, were not
14 viewed as demonstrated health impacts. Nevertheless, EPA did state that it believed the
15 assessment presented reasonable estimates as to the possible extent of risk for these
16 effects given the available information (62 FR at 38656).
17
18 • Consideration of key uncertainties and alternative assumptions resulted in fairly wide
19 ranges in estimates of the incidence of PM-related mortality and morbidity effects and
20 risk reductions associated with attainment of alternative standards in both locations in the
21 risk assessment. Significantly, the combined results for these two cities alone found that
22 the risk remaining after attaining the current PM10 standards was on the order of hundreds
23 of premature deaths each year, hundreds to thousands of respiratory-related hospital
24 admissions, and tens of thousands of additional respiratory-related symptoms in children
25 (62 FR at 38656).
26
27 • Based on the results from the sensitivity analyses of key uncertainties and the integrated
28 uncertainty analyses, the single most important factor influencing the uncertainty
29 associated with the risk estimates was whether or not a threshold concentration exists
30 below which PM-associated health risks are not likely to occur (62 FR at 38656).
31
32 • Over the course of a year, the few peak 24-hour PM2 5 concentrations appeared to
33 contribute a relatively small amount to the total health risk posed by the entire air quality
34 distribution as compared to the aggregated risks associated with the low to mid-range
35 PM25 concentrations (62 FR at 38656).
36
August 2003 4-4 Draft - Do Not Quote or Cite
-------
1 • There was greater uncertainty about both the existence and the magnitude of estimated
2 excess mortality and other effects associated with PM2 5 exposures as one considered
3 lower concentrations that approach background levels (62 FR at 38656).
4
5 • Based on the results from the sensitivity analyses of key uncertainties and/or the
6 integrated uncertainty analyses, the following uncertainties had a much more modest
7 impact on the risk estimates: the use of C-R functions from multi-pollutant, rather than
8 single-pollutant models; the choice of approach to adjusting the slope of the C-R
9 functions in analyzing alternative cutpoints (i.e., hypothesized thresholds); the value
10 chosen to represent average annual background PM concentrations; and the choice of
11 approach to adjusting air quality distributions for simulating attainment of alternative
12 PM25 standards (EPA, 1996b).
13
14 4.2 GENERAL SCOPE OF PM RISK ASSESSMENT
15 As discussed in Chapter 3 above, the draft CD concludes that "the overall weight of
16 evidence, based on information concerning PM exposure, dosimetry, toxicology, and
17 epidemiology, supports the conclusion that PM, especially fine PM, is the primary contributor to
18 a variety of adverse health effects associated with air pollution. However, difficult technical
19 issues still remain in further separating the effects of fine and coarse particles and in delineating
20 respective contributions of PM acting alone or in combination with gaseous co-pollutants in
21 increasing risks of health effects anticipated to occur in response to exposures to contemporary
22 particle-containing ambient air mixes in the U.S." (CD, p. E-44) For coarse-fraction particles the
23 strongest evidence is found relating PM10_2 5 ambient concentrations and increased respiratory
24 hospital admissions and respiratory symptoms. While not as strong as the case for PM2 5 and
25 premature mortality, there is sufficient evidence to suggest that PM10_2 5 also is likely to be linked
26 with premature mortality. The draft CD finds that the expanded body of evidence substantiates
27 associations between PM2 5 and mortality, with some studies suggesting associations between
28 PM10_2 5 and mortality as well (CD, p. 8-57). Although mortality effects for PM10_2 5 were not
29 included in OAQPS's proposed Risk Analysis Scoping Plan (EPA, 2001c), such effects have
30 been added consistent with advice provided by CASAC PM Panel members in the May 1, 2003
31 consultation with staff.
32 The updated risk assessment being conducted for the current NAAQS review is premised
33 on the assumption that each of the ambient PM indicators (i.e., PM2 5, PM10_2 5, and PM10) are
August 2003 4-5 Draft - Do Not Quote or Cite
-------
1 causally related to the mortality, morbidity, and symptomatic effects (alone and/or in
2 combination with other pollutants) observed in the epidemiological studies. Staff concludes that
3 this assumption is well supported by the evaluation contained in the draft CD and is consistent
4 with the advice provided by the CAS AC PM Panel. However, staff recognizes that there are
5 varying degrees of uncertainty associated with whether or not there is a causal relationship for
6 each of the three PM indicators and the various health endpoints and that the degree of
7 uncertainty is directly related to differences in the relative weight of evidence.
8 This PM risk assessment focuses on selected health endpoints such as increased excess
9 daily mortality and mortality associated with long-term exposure, increased hospital admissions
10 for respiratory and cardiopulmonary causes, and increased respiratory symptoms for children. A
11 consequence of limiting the assessment to selected health endpoints is that the risk estimates
12 likely understate the type and extent of potential health impacts of ambient PM exposures.
13 Although the risk assessment does not address all health effects for which there is some evidence
14 of association with exposure to PM, all such effects are identified and considered previously in
15 Chapters.
16 Like the prior risk assessment done as part of the last review (EPA, 1996b), this current
17 updated risk assessment uses C-R functions from epidemiological studies based on ambient PM
18 concentrations measured at fixed-site, population-oriented, ambient monitors. As discussed
19 earlier in Chapter 2 (section 2.8) and Chapter 3 (section 3.5.3.3), measurements of daily
20 variations of ambient PM2 5 concentrations, as used in the time-series studies that provide the C-
21 R relationships for this assessment, have a plausible linkage to the daily variations of exposure
22 from ambient sources for the populations represented by ambient monitoring stations. The draft
23 CD concludes that ambient PM concentrations (i.e., those measured at ambient monitoring
24 stations) are a useful surrogate for exposure to ambient PM (CD, p. 9-117, p. 8-252). A more
25 detailed discussion of the possible impact of exposure misclassification on the estimated C-R
26 relationships derived from the community epidemiological studies is presented above in Chapter
27 3 (see section 3.5.3.3).
28 While quantitative estimates of personal or population exposure do not enter into
29 derivations of the PM risk estimates for this review, an understanding of the nature of the
August 2003 4-6 Draft - Do Not Quote or Cite
-------
1 relationships between ambient PM and its various components and human exposure underlies the
2 conceptual basis for the risk assessment. Unlike recent reviews for O3 and CO, where exposure
3 analyses played an important role, a quantitative exposure analysis will not be conducted as part
4 of this review since the currently available epidemiology health effects evidence relates ambient
5 PM concentrations, not exposures, to health effects. As discussed in Chapter 4 of the draft CD,
6 EPA and the exposure analysis community are working to improve exposure models designed
7 specifically to address PM. Both EPA and the broader scientific community also are in the
8 process of collecting new information in PM exposure measurement field studies that will
9 improve the scientific basis for exposure analyses that may be considered in future reviews.
10 While the NAAQS are intended to provide protection from exposure to ambient PM,
11 EPA recognizes that exposures to PM from other sources (i.e., non-ambient PM) also have the
12 potential to affect health. The EPA's Office of Radiation and Indoor Air and other Federal
13 Agencies, such as the Consumer Product Safety Commission (CPSC) and the Occupational
14 Safety and Health Administration (OSHA), address potential health effects related to indoor,
15 occupational, environmental tobacco smoke, and other non-ambient sources of PM exposure. As
16 with the prior PM risk assessment, contributions to health risk from non-ambient sources are
17 beyond the scope of the proposed risk assessment for this NAAQS review.
18 This current PM health risk assessment is similar in many respects to the one conducted
19 for the last PM NAAQS review. Both the prior and the current PM risk assessment:
20 • estimate risks for the urban centers of example cities, rather than attempt a nationwide
21 assessment.
22
23 • analyze risks for a recent 12-month period of air quality (labeled "as is") and for a
24 scenario in which air quality just meets the current set of standards.
25
26 • analyze risks for scenarios in which air quality is simulated to just meet potential
27 alternative standards that are recommended by staff for consideration.
28
August 2003 4-7 Draft - Do Not Quote or Cite
-------
1 • estimate risks only for concentrations exceeding estimated background levels or the
2 lowest measured level (LML) observed in the study, if it is higher than the estimated
3 background level in the assessment location.3
4
5 • present qualitative and quantitative considerations of uncertainty, including sensitivity
6 analyses of key individual uncertainties.
7
8 Both the prior and the current PM risk assessment focus on health endpoints for which C-
9 R functions have been estimated in epidemiological studies. Since these studies estimate C-R
10 functions using air quality data from fixed-site, population-oriented monitors, the appropriate
11 application of these functions in a PM risk assessment similarly requires the use of air quality
12 data from fixed-site, population-oriented, ambient monitors. This is identical to the approach
13 taken in the last PM NAAQS review.
14 The current risk assessment includes risk estimates for 9 urban areas for PM2 5, 5 urban
15 areas for PM10_2 5, and 12 urban areas for PM10 These areas have been chosen based on
16 availability of PM C-R relationships and adequate PM air quality data. The selection of these
17 areas also reflects a desire to include areas from the various regions of the United States to the
18 extent possible in order to reflect regional differences in the composition of PM and other factors
19 (e.g., different levels of co-pollutants, air-conditioning use).
20 Finally, estimates of risks above background PM concentrations are provided for short-
21 term exposure health endpoints because they are judged to be more relevant to policy decisions
22 about the NAAQS than estimates that include risks potentially attributable to uncontrollable
23 background PM concentrations. For long-term exposure mortality associated with PM2 5, the
24 LMLs for the PM2 5 epidemiology studies are 10 and 11 i-ig/m3, which are higher than the range
25 of estimated PM2 5 background levels in either the East or West. Estimating risks outside the
26 range of the original epidemiology studies that were the source of the C-R functions would
27 introduce significant additional uncertainties into the risk assessment. Therefore, the risks
3 A C-R relationship estimated by an epidemiological study may not be representative of the relationship
that exists outside the range of concentrations observed during the study. To partially address this problem, risk was
not calculated for PM levels below the lowest measured level (LML) in the study, if reported. The LML's for each
study that provided a C-R relationship for the current PM risk assessment, where reported, are provided in Appendix
CofAbt(2003b).
August 2003 4-8 Draft - Do Not Quote or Cite
-------
1 associated with long-term exposure were only estimated in excess of the LML. Since we do not
2 estimate risks below the LML, the overall long-term exposure mortality risks may be
3 underestimated if annual average PM2 5 concentrations below the LMLs contribute to long-term
4 exposure mortality.
5 The following sections provide an overview of the components of the risk model,
6 describe the selection of urban areas and health endpoints included in the PM risk assessment,
7 discuss each of the major components of the risk model, address characterization of uncertainty
8 and variability associated with the risk estimates, and summarize the currently available PM risk
9 estimates. A separate draft technical report (Abt Associates, 2003b) also is available which
10 provides a more detailed discussion of the risk assessment methodology and risk estimates.
11
12 4.2.1 Overview of Components of the Risk Model
13 In order to estimate the incidence of a particular health effect associated with "as is"
14 conditions in a specific county or set of counties attributable to ambient PM25 (or PM10_25 or
15 PM10) exposures in excess of background and the change in incidence of the health effect in that
16 county or set of counties corresponding to a given change in PM2 5 (or PM10_2 5) levels resulting
17 from just meeting a specified set of PM25 (or PM10_25) standards, the following three elements are
18 required:
19 • air quality information including: (1) "as is" air quality data for PM25, PM10_25, and PM10
20 from population-oriented monitors for the selected location, (2) estimates of background
21 PM2 5, PM10_2 5, and PM10 concentrations appropriate for that location, and (3) a method
22 for adjusting the "as is" data to reflect patterns of air quality estimated to occur when the
23 area just meets a given set of PM2 5 (or PM10_2 5) standards.
24
25 • relative-risk based C-R functions (preferably derived in the assessment location) which
26 provide an estimate of the relationship between the health endpoints of interest and
27 ambient PM concentrations.
28
29 • annual or seasonal baseline health effects incidence rates and population data which are
30 needed to provide an estimate of the annual or seasonal baseline incidence of health
31 effects corresponding to "as is" PM levels.
32
August 2003 4-9 Draft - Do Not Quote or Cite
-------
1 Figure 4-1 provides a broad schematic depicting the role of these components in the risk
2 assessment. Those points where EPA proposes to conduct analyses of alternative assumptions,
3 procedures, or data are indicated by a circle with Sx in it. A summary description of the type of
4 sensitivity analyses performed is included later in section 4.2.7 (See Table 4-9). Each of the
5 three key components (i.e., air quality information, estimated PM-related C-R functions, and
6 baseline incidence) are discussed below, highlighting those points at which judgments have been
7 made.
8
9 4.2.2 Criteria for Selection of Health Endpoints and Urban Study Areas
10 Only two urban counties were included in the risk assessment conducted for the prior PM
11 NAAQS review due to the very limited number of urban areas that had sufficient recent PM2 5
12 ambient air quality monitoring data and because of the limited number of epidemiological
13 studies that directly measured PM25. As discussed in more detail in Chapter 3, since the last
14 review, a significant number of epidemiological studies have been published examining a variety
15 of health effects associated with ambient PM25, PM10_2 5, and PM10 in various urban areas
16 throughout the U.S. and Canada, as well as Europe and other parts of the world. Tables 9-8 and
17 9-10 in the draft CD summarize the available U.S. and Canadian short-term exposure studies that
18 provide effect estimates for PM (i.e., PM2 5, PM10_2 5, and PM10) for mortality and morbidity
19 endpoints,
20 respectively. Table 9-11 summarizes the available U.S. and Canadian studies that provide effect
21 estimates for mortality and other health endpoints for long-term exposure to PM25 and other PM
22 indicators. While a significant number of new epidemiological studies also have been published
23 since the last review, and are evaluated in the draft CD, the PM risk assessment relies only on
24 U.S. studies given the possible differences in population and characteristics of PM and co-
25 pollutants between the U.S. and these other locations. The approach and criteria that EPA has
26 used to select the health endpoints and urban areas to include in the risk assessment for the three
27 PM indicators are described below.
August 2003 4-10 Draft - Do Not Quote or Cite
-------
Ambient Population-
oriented Monitoring
and Estimated
Background Levels for
Selected Cities
Air Quality Adjustment
Procedures
Alternative Proposed
Standards
Human Epidemiological
Studies (various health
endpoints)
Estimates of City-specific
Baseline Health Effects
Incidence Rates (various
health endpoints) and
Populations
As is" Analysis
Changes in
Distribution
ofPMAir
Quality
Concentration -
Response
Relationships
Health
Risk
Model
Risk Estimates:
• "As is"
• "Alternative
Scenarios"
= kth Sensitivity Analysis (See Table 4-9): Analysis of effects of alternative assumptions, procedures or data
occurs at these points.
Figure 4-1. Major Components of Particulate Matter Health Risk Assessment
August 2003
4-11
Draft - Do Not Quote or Cite
-------
1 4.2.2.1 Selection of Health Endpoint Categories
2 OAQPS staff carefully reviewed the health effects evidence evaluated in the draft CD in
3 order to identify potential health effect categories to include in the current PM risk assessment.
4 Given the large number of endpoints and studies addressing PM-related effects, staff
5 recommended for inclusion in the PM risk assessment only the more severe and better
6 understood (in terms of health consequences) health endpoint categories for which the weight of
7 the evidence supports the existence of a relationship between PM and the health effects category.
8
9 For the health effect categories included in the current PM risk assessment, the risk
10 assessment assumes that a causal relationship exists. On this topic the draft CD states,
11 ... although associations of PM with harmful effects continue to be observed consistently
12 across most new studies, the newer findings do not fully resolve issues concerning
13 relative contributions to the observed epidemiological associations of (a) PM acting
14 along, (b) PM acting in combination with gaseous co-pollutants, (c) the gaseous
15 pollutants per se, and (d) the overall ambient mix. ... However, with considerable new
16 experimental evidence now in hand, it is possible to hypothesize various ways in which
17 ambient exposure to PM acting alone or in combination with other co-pollutants can
18 plausibly be involved in the complex chain of biological events leading to harmful health
19 effects in the human population. This newer experimental evidence, coupled with new
20 exposure analyses results, adds much support for interpreting the epidemiological
21 findings discussed here as likely being indicative of causal relationships between
22 exposures to ambient PM (or specific size or chemical components) and consequent
23 associated increased mortality and morbidity effects.(CD, p. 9-70)
24
25 For the three PM indicators (PM2 5, PM10, PM10_2 5), staff considered only those endpoint
26 categories which provided C-R relationships based on U.S. and Canadian studies that used PM
27 concentrations obtained by one of the following approaches: (1) directly measuring the fine
28 fraction of PM using PM2 5 or PM2 b (2) estimating the fine fraction using nepholometry data,
29 (3) directly measuring PM10, and (4) estimating PM10_2 5 concentrations based on co-located PM10
30 and PM2 5 monitors or based on measurements using dichots.
31 The staff selected health effect endpoints for each of the three PM indicators based on
32 whether the weight of the evidence available in the draft CD, taking into account advice from
33 CASAC, supports the assumption that the health effects are causally related to the PM indicators.
34 In cases where all of the available studies failed to find a statistically significant relationship, the
August 2003 4-12 Draft - Do Not Quote or Cite
-------
1 effect endpoint was excluded. In situations where there is a mixture of statistically significant
2 and non-significant findings for a given health effect endpoint and PM
3 indicator (e.g., hospital admissions for COPD patients and PM25), staff also considered evidence
4 from available PM10 studies in making a judgment on whether effects are likely related to PM25.
5 Based on a review of the evidence evaluated in the draft CD and the criteria discussed
6 above, staff included the following broad categories of health endpoints associated with short-
7 term exposures in the risk assessment for all three PM indicators:
8 • non-accidental total, cardiovascular, and respiratory mortality;
9 • hospital admissions for cardiovascular and respiratory causes;
10 • respiratory symptoms not requiring hospitalization
11 In addition, total, cardiopulmonary, and lung cancer mortality due to long-term exposure are
12 included in the PM25 risk assessment. Other effects reported to be associated with PM,
13 including, but not limited to, decreased lung function, changes in heart rate variability, and
14 increased emergency room visits are addressed in Chapter 3, but are not included in the
15 quantitative risk assessment.
16 4.2.2.2 Selection of Study Areas
17 A primary goal of the current PM risk assessment has been to identify and include urban
18 areas in the U.S. for which epidemiological studies are available that estimate C-R relationships
19 for those locations. This goal is in large part motivated by the evidence from the NMMAPS
20 (Samet et al., 2001) that suggests there is geographic variability in C-R relationships across
21 different urban areas in the U.S. The selection of urban areas to include in the PM risk
22 assessment was based on the following criteria:
23 • An area had sufficient air quality data for a recent year (1999 or later). Sufficient
24 PM2.5 or PM10 air quality data is defined as having at least one PM monitor at
25 which there are at least 11 observations per quarter for a one year period.4
26 Sufficient air quality data for PM10_2 5 is defined as a one year period with at least
4For 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.
August 2003 4-13 Draft - Do Not Quote or Cite
-------
1 11 daily values per quarter based on data from co-located PM2 5 and PM10
2 monitors.5'6
3
4 • An area is the same as or close to the location where at least one C-R function for
5 one of the recommended health endpoints has been estimated by a study that
6 satisfies the study selection criteria (see below).
7
8 • An area is one in which studies exist that had relatively greater precision, as
9 indicated by a relatively greater number of effect-days observations.
10
11 • For the hospital admission effects category relatively recent area-specific baseline
12 incidence data, specific to International Classification of Disease (ICD) codes, are
13 available.
14
15 For the PM2 5 risk assessment, staff focused on selecting urban areas based primarily on a
16 location's having non-accidental total and cause-specific mortality PM25 C-R functions since this
17 was the largest data base in terms of number of studies in different locations. Staff then
18 supplemented this by consideration of other morbidity endpoints (e.g,, hospital admissions).
19 Based on a review of studies listed in Tables 9-8 and 9-10 of the draft CD (see also Appendix A
20 of this SP), a candidate pool of 17 urban locations was initially suggested based on short-term
21 exposure mortality studies (16 of the candidate locations); Seattle was added based on a hospital
22 admissions study.7
23 Staff next considered an indicator of study precision for the short-term exposure
24 mortality studies identified in the first step. As discussed above in Chapter 3 (section 3.3.1.1.1)
25 and in Chapter 8 of the draft CD, the natural logarithm of the mortality-days (a product of each
26 city' s daily mortality rate and the number of days for which PM data were available) can be used
27 as an indicator of the degree of precision of effect estimates; studies with larger values for this
5The criterion of at least 11 observations per quarter is based on EPA guidance on measuring attainment of
the daily and annual PM standards outlined in Appendix N of the July 18, 1997 Federal Register notice.
6To be consistent with the epidemiological studies which generally used only population-oriented monitors,
any monitors where the monitoring objective was listed as "highest concentration monitor" were excluded from
consideration. The few monitors excluded were in industrial or commercial areas and are intended to characterize
local conditions near major point sources.
7The Tolbert et al. (2000) study in Atlanta was excluded from consideration because it presented only
preliminary results, and because the draft CD urged caution in interpreting these preliminary results.
August 2003 4-14 Draft - Do Not Quote or Cite
-------
1 indicator should be accorded relatively greater study weight. As discussed previously in
2 Chapter 3, the draft CD notes that based on an analysis of the 90-cities NMMAPS data, there are
3 "generally narrower confidence intervals for more homogeneously positive effects estimates as
4 study size increases beyond about In (mortality-days) = 9.0" (CD, Figure 8-21, p. 8-244). Based
5 on this observation, staff considered only those urban areas in which studies with relatively
6 greater precision were conducted, specifically including studies that have a natural log of
7 mortality-days greater than or equal to 9.0 for total non-accidental mortality.8 As a result of
8 applying this criterion, six urban areas were excluded as potential study areas (Camden, NJ;
9 Coachella Valley, CA; Elizabeth, NJ; Newark, NJ; Steubenville, OH; and Topeka, KS).
10 Finally, staff considered which of the potential study locations identified from steps 1 and
11 2 above also had sufficient PM2 5 ambient monitoring data. A location was considered to have
12 sufficiently complete air quality data if it had at least one monitor at which there were at least 11
13 observations per quarter and at least 122 observations per year (i.e., equivalent to 1 in 3 day
14 monitoring). This final criterion excluded two of the remaining potential study areas (Knoxville,
15 TN and Portage, WI), leaving eight cities in which epidemiological studies reported C-R
16 relationships for PM25 and mortality and which had sufficient air quality data in a recent year.
17 Eight of the nine urban areas, excluding Seattle, are included in the PM25 risk assessment
18 based on short-term exposure mortality. Since the studies reporting C-R functions for PM2 5-
19 related mortality associated with long-term exposure were conducted for multiple cities across
20 the U.S., the issue of matching risk assessment locations with city-specific studies did not arise.
21 Therefore, the PM2 5 risk assessment for long-term exposure mortality was conducted for these
22 same eight urban areas.
23 Most of the morbidity and respiratory symptom studies reporting PM2 5-related effects
24 were conducted in the same set of locations as the short-term exposure mortality studies. In
25 considering these other health endpoints, staff applied similar criteria (i.e., studies providing
26 effects estimates with relatively greater precision and availability of recent and adequate PM25
Most of the epidemiological 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 C-R functions from such studies,
as long as the natural log of total mortality-days was greater than or equal to 9.0.
August 2003 4-15 Draft - Do Not Quote or Cite
-------
1 ambient air quality data). In addition, for the hospital admissions effect category, staff limited
2 selection to those urban areas where the necessary baseline incidence data could be obtained.
3 Based on applying the above criteria and considerations, the health endpoints and urban
4 locations selected for the PM2 5 risk assessment are summarized in Tables 4-1 and 4-2, for
5 mortality and morbidity endpoints, respectively. These tables also list the specific studies that
6 provided the estimated C-R functions used in the PM2 5 risk assessment. More detailed
7 information on the studies selected can be found in Exhibit C of the PM risk assessment
8 technical support document (Abt Associates, 2003b).
9 The selection of urban areas to include for the PM10_2 5 risk assessment was based on
10 examining the pool of epidemiological studies reporting associations for PM10_25 in any of the
11 urban areas already selected for the PM2 5 risk assessment. As summarized in Tables 4-3 and
12 4-4, the PM10_25 risk assessment is more limited because of the more limited air quality data
13 (requiring co-located PM2 5 and PM10 monitors or availability of dichot data) as well as the
14 smaller number of studies. As noted previously, consistent with advice from the CASAC PM
15 Panel, short-term exposure mortality was included among the health effects in the PM10_25 risk
16 assessment. The areas included for short-term exposure mortality in the PM10_25 risk assessment
17 are Detroit, Philadelphia, Phoenix, and St. Louis.9 In addition, increased hospital admissions
18 associated with PM10_2 5 were estimated for Detroit and Seattle, and increased respiratory
19 symptoms were estimated for St. Louis.
20 In addition to the criteria listed above, selection of areas to include in the PM10 risk
21 assessment was guided by the desire to include urban areas that would further inform
22 comparisons both across the PM indicators (i.e., PM25, PM10_25) and across health effect
23 categories. Staff also wanted to include, at a minimum those urban areas already selected for the
24 PM2 5 risk assessment, for which city-specific C-R functions for short-term exposure mortality
25 are available from the NMMAPS and/or other studies. Tables 4-5 and 4-6 summarize the urban
26 locations selected for the PM10 risk assessment for the mortality and morbidity health endpoints,
27 respectively. Similar to the prior tables, the studies providing the C-R relationships also are
28 identified. Additional details about the epidemiological studies and the C-R functions used in
9For Phoenix, there were not sufficient data within AIRS to calculate PM10_2 5 concentrations in any of the
years from 1999 through 2002. Instead, we used data from 1997, based on co-located TEOM monitors. Because of
this, the PM10_2 5 risk assessment in Phoenix is treated as a separate special case in the PM risk assessment.
August 2003 4-16 Draft - Do Not Quote or Cite
-------
1 the risk assessment from these studies are provided in Appendix A of Chapter 3 of this draft
2 Staff Paper and in Appendix C of the risk assessment technical support document (Abt, 2003b),
3 respectively.
4
5 4.2.3 Air Quality Considerations
6 As mentioned earlier, air quality information required to conduct the PM risk assessment
7 includes: (1) "as is" air quality data for PM2 5, PM10_2 5, and PM10 from population-oriented
8 monitors for each selected location, (2) estimates of background PM25, PM10_2 5, and PM10
9 concentrations appropriate for each location, and (3) a method for adjusting the "as is" data to
10 reflect patterns of air quality estimated to occur when an area just meets a given set of PM25 (or
11 PM10_2 5) standards. OAQPS retrieved ambient air quality data for PM2 5 and PM10 for the
12 potential study areas for the years 1999 through 2002 from EPA's Aerometric Information
13 Retrieval System (AIRS). Urban areas were only included in the risk assessment if there was at
14 least one monitor with 11 or more observations per quarter. Staff calculated PM10_2 5
15 concentrations from co-located PM2 5 and PM10 monitors that met the minimum observation
16 cutoff. The most recent year of PM data was used for each study area and PM indicator subject
17 to meeting this requirement. Table 4-7 provides a summary of the PM25, PM10_25, and PM10
18 ambient air quality data for the urban study areas included in the risk assessment. Additional
19 tables providing more detailed information on PM ambient concentrations for these locations,
20 including the number of observations available on a quarterly and annual basis for each monitor,
21 can be found in Appendix A of the technical support document (Abt, 2003b).
August 2003 4-17 Draft - Do Not Quote or Cite
-------
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
St. Louis, MO
Mortality Associated with Short-Term Exposure
Total (non-accidental)
Schwartz etal. (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) *
Mar et al. (2000)E
Fairley (1999)F
Klemm etal. (2000)B
- ischemic heart
disease *
Circulatory
Lippmann et al.
(2000)°
Respiratory
Klemm et al. (2000)B -
COPD *,
pneumonia *
Lippmann et al. (2000)°
Moolgavkar (2000a)D -
COPD
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)-6cities
Krewski et al. (2000)-ACS
Pope et al. (2002)-ACS extended
* Includes a multi-city or multi-county C-R 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 provides total and cardiopulmonary mortality and
Pope et al. (2002)-ACS extended provides total and lung cancer mortality
August 2003
4-18
Draft - Do Not Quote or Cite
-------
1
2
3
4
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
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
A Reanalyzed in Ito (2003)
B Reanalyzed in Moolgavkar (2003)
c Reanalyzed in Sheppard (2003)
August 2003
4-19
Draft - Do Not Quote or Cite
-------
1
2
3
Table 4-3. Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the PM10_2 5 Risk Assessment
Urban Location
Detroit, MI
Philadelphia, PA
Phoenix, AZ
St. Louis, MO
Total (non-accidental or all
cause)
Lippmann et al. (2000)A
Lipfert et al. (2000)
Schwartz etal. (1996)°
Cardiovascular
Lipfert etal. (2000)*
Mar et al. (2000)B
Circulatory
Lippmann et al.
(2000)A
Respiratory
Lippmann et al. (2000)A
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
* Includes multi-county, multi-city, regional, or national C-R function
A Reanalyzed in Ito (2003)
B Reanalyzed in Mar et al. (2003)
c Reanalyzed in Schwartz (2003b)
Table 4-4. 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, regional, or national C-R function
A Reanalyzed in Ito (2003)
B Reanalyzed in Sheppard (2003)
August 2003
4-20
Draft - Do Not Quote or Cite
-------
Table 4-5. Mortality Health Endpoints, Urban Locations, and Studies Selected for Use in the PM10 Risk Assessment
Urban Location
Boston, MA
Chicago, IL
Detroit, MI
Los Angeles, CA
Minneapolis, MN
Philadelphia, PA
Phoenix, AZ
Pittsburgh, PA
Provo, UT
Total (non-accidental or
all cause)
Klemm et al. (2000)A *
Samet et al. (2000)B *
Schwartz (2000b)c*
Samet et al. (2000)B' *
Styeretal. (1995)
Ito and Thurston (1996)
Moolgavkar (2000a)D
Schwartz (2000a)c*
Schwartz (2000b)c*
Schwartz (2000a)c*
Samet et al. (2000)B' *
Lippmann et al. (2000)E
Samet et al. (2000)B *
Kinneyetal. (1995)
Moolgavkar (2000a)D
Schwartz (2000b)c *
Schwartz (2000a)c *
Samet et al. (2000)B' *
Samet et al. (2000)B *
Lipfert et al. (2000)
Samet et al. (2000)B *
Schwartz (2000b)c *
Schwartz (2000a)c *
Samet et al. (2000)B- *
Chock et al. (2000)
Samet et al. (2000)B *
Cardiovascular
Moolgavkar (2000a)D
Bragaetal. (2001)°*
Bragaetal. (2001)°*
Moolgavkar (2000a)D
Bragaetal. (2001)°*
Lipfert et al. (2000) *
Maretal. (2000)F
Bragaetal. (2001)°*
Cardiorespiratory
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Circulatory
Ito and Thurston
(1996)
Lippmann et al.
(2000)E
Respiratory
Ito and Thurston (1996)
Moolgavkar (2000a)D - COPD
Braga et al. (2001)° - COPD *,
pneum. *
Lippmann et al. (2000)E
Braga et al. (2001)° - COPD *
pneum. *
Moolgavkar (2000a)D - COPD
Braga et al. (2001)° - COPD *
pneum. *
Braga et al. (2001)° - COPD *
pneum. *
August 2003
4-21
Draft - Do Not Quote or Cite
-------
Urban Location
San Jose, CA
Seattle, WA
St. Louis, MO
Total (non-accidental or
all cause)
Samet et al. (2000)B *
Fairley (1999)°
Schwartz (2000b)c *
Schwartz (2000a)c *
Samet et al. (2000)B' *
Klemm et al. (2000)A *
Samet et al. (2000)B *
Cardiovascular
Fairley (1999)°
Bragaetal. (2001)°*
Cardiorespiratory
Samet et al. (2000)B *
Samet et al. (2000)B *
Samet et al. (2000)B *
Circulatory
Respiratory
Fairley (1999)°
Braga et al. (2001)° - COPD *
pneum. *
4
5
6
7
8
9
10
11
*Includes multi-county, multi-city, regional, or national C-R function
A Reanalyzed in Klemm and Mason (2003)
B Reanalyzed in Dominici et al. (2003)
c Reanalyzed in Schwartz (2003a)
D Reanalyzed in Moolgavkar (2003)
E Reanalyzed in Ito (2003)
F Reanalyzed in Mar et al. (2003)
G Reanalyzed in Fairley (2003)
August 2003
4-22
Draft - Do Not Quote or Cite
-------
1
2
3
4
Table 4-6. Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in the PM10 Risk Assessment
Urban Location
Boston, MA
Chicago, IL
Detroit, MI
Los Angeles,
CA
Minneapolis,
MN
Pittsburgh, PA
Provo, UT
Seattle, WA
St. Louis, MO
Cardiovascular Hospital Admissions
Samet et al. (2000)A *
Moolgavkar (2000b)B
Morris and Naumova (1998) - congestive
heart failure
Samet et al. (2000)A *
Schwartz and Morris (1995) - congestive
heart failure, ischemic heart disease,
dysrhythmias
Lippmann et al. (2000)° - congestive heart
disease, ischemic heart disease,
dysrhythmias
Moolgavkar (2000b)B
Linn et al. (2000)
Samet et al. (2000)A *
Samet et al. (2000)A *
Samet et al. (2000)A *
Samet et al. (2000)A *
Respiratory Hospital Admissions
Samet et al. (2000)A - pneumonia, COPD*
Moolgavkar (2000c)B - COPD
Samet et al. (2000)A - pneumonia, COPD*
Lippmann et al. (2000)° - Pneumonia, COPD
Schwartz (1994a) - pneumonia, COPD
Moolgavkar (2000c)B - COPD
Linn et al. (2000) - respiratory, COPD, asthma
Nauenberg and Basu (1999) - asthma
Samet et al. (2000)A - pneumonia, COPD*
Schwartz (1994b) - pneumonia, COPD
Samet et al. (2000)A - pneumonia, COPD*
Samet et al. (2000)A - pneumonia, COPD*
Samet et al. (2000)A - pneumonia, COPD*
Sheppard et al. (1999)D - asthma
Respiratory Symptoms
Schwartz et al. (1994) - lower
respiratory symptoms (LRS), upper
respiratory symptoms (URS)
Pope et al. (1991) - LRS, URS
Yu et al. (2000) - asthma
Schwartz et al. (1994) - LRS, URS
9
10
11
12
13
14
15
16
17
18
19
*Includes multi-city, regional, or national C-R function
A Reanalyzed in Zanobetti and Schwartz (2003)
B Reanalyzed in Moolgavkar (2003)
c Reanalyzed in Ito (2003)
D Reanalyzed in Sheppard (2003)
August 2003
4-23
Draft - Do Not Quote or Cite
-------
1 4.2.3.1 Estimating PM Background Levels
2 Background PM concentrations used in the PM risk assessment are defined above in
3 Chapter 2 as the PM concentrations that would be observed in the U.S. in the absence of
4 anthropogenic emissions of PM and its precursors in North America. For the base case risk
5 estimates, the midpoint of the appropriate ranges of annual average estimates for PM2 5 and PM10
6 background presented in Section 2.7 were used (i.e., eastern values were used for eastern study
7 locations and western values were used for western study locations). For PM10_2 5 the rough point
8 estimate of 3 |ig/m3 for both eastern and western portions of the U.S. was used (CD, p.3-82). In
9 addition, sensitivity analyses were also conducted in which the lower- and upper-end of the
10 estimated ranges of background for each PM indicator were used to estimate risk. A sensitivity
11 analysis also was performed examining the potential impact of using a distribution of estimated
12 daily background PM2 5 concentrations on the risk estimates. Table 4-8 summarizes the range
13 and midpoint background levels used in the base case and sensitivity analyses.
14 4.2.3.2 Simulating PM Levels That Just Meet Specified Standards
15 To estimate the health risks associated with just meeting the current PM25 standards and
16 alternative PM25 and PM10_2 5 standards it is necessary to estimate the distribution(s) of PM
17 concentrations that would occur under each specified standard (or sets of standards). While
18 compliance with the standards is based on a 3-year average, because of the limited number of
19 years for which PM2 5 and PM10_2 5 data are available, only a single year of data is used in the risk
20 assessment. When assessing the risks associated with long-term exposures, using
21 epidemiological studies that use an annual average concentration, the annual mean is simply set
22 equal to the standard level. In contrast, when assessing the risks associated with short-term
23 exposures using epidemiological studies which consider daily average concentrations, the
24 distribution of 24-hour values that would occur upon just attaining a given 24-hour PM standard
25 has to be simulated.
August 2003 4-24 Draft - Do Not Quote or Cite
-------
Table 4-7. Summary of PM Ambient Air Quality Data for Risk Assessment Study Areasi
Area
Boston, MAa
Chicago, ILb
Detroit, MP
Los Angeles County, CAd
Minneapolis, MNe
Philadelphia County, PAf
Phoenix, AZg
Pittsburgh, PAh
Provo, UT
San Jose, CA
Seattle, WAk
St. Louis1
Population
(millions)
2.8
5.4
2.1
9.5
1.6
1.5
3.1
1.3
0.5
1.7
1.7
2.5
PM,;*
Annual Avg.
11.4
15.8
20.4
14.5
10.4
15.2
12.4
9.3
14.7
24-hr Avg, 98th%
33.0
36.7
47.2
40.9
28.9
41.4
57.5
24.8
34.2
PM,,,,;*
Annual Avg.
6.5
9.2
33.3
5.7
3.5
24-hr , 98th%
23.5
22.1
70.6
11.4
11.4
PM,;*
Annual Avg.
24.3
25.7
20.2
36.6
27.5
25.4
47.9
20.5
28.6
24.6
18.0
22.8
24-hr, 98th%
50
55
49
55
61
72
83.9
58
72
77
44
69
'Based on air quality data for the year 2002, unless otherwise noted in footnotes below; all concentrations are in \ig/m3.
"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; PM10 data are for 1999.
blncludes Cook County.
Includes Wayne County.
Includes Los Angeles County.
Includes Hennepin and Ramsey Counties.
Includes Philadelphia County.
includes Maricopa County; PM25 air quality data are for 2001 and PM10_2 5 air quality data are based on TEOM data from 1997.
Includes Allegheny County
'Includes Utah County
includes Santa Clara County; PM2 5 air quality data are for 2001 and PM10 air quality data are for 1999.
"includes King County; PM10_2 5 air quality data are for 2000.
'includes St. Louis, Franklin, Jefferson, St. Charles Counties in MO, Clinton, Madison, Monroe, and St. Claire Counties inIL and St. Louis City.
August 2003
4-25
Draft - Do Not Quote or Cite
-------
2
3
Table 4-8. Summary of Estimated PM Background Concentrations Used in
Risk Assessment*
Eastern
Locations**
Western
Locations***
PM7,
Range
2-5
1 -4
Midpoint
3.5
2.5
PM,n.,.,
Range
<1 -9
<1 -7
Point
Estimate
3
3
PM,n
Range
5- 11
4-8
Midpoint
8
6
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
*A11 concentrations are specified in ng/m3
"Eastern locations: Boston, Chicago, Detroit, Minneapolis-St. Paul, Philadelphia, Pittsburgh, St. Louis
"'Western locations: Los Angeles, Phoenix, Provo, San Jose, Seattle
There are many possible ways to create an alternative distribution of daily concentrations
that just meets a specified set of PM standards. Both the assessment conducted during the last
NAAQS review (see Abt Associates, 1996, section 8.2) and a new analysis of historical air
quality (see Abt Associates, 2003b, Appendix B) have found that PM2 5 levels in excess of
estimated background concentrations have in general historically decreased in a proportional
manner (i.e., concentrations at different points in the distribution of 24-hour PM25 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 PM reductions that would result from just meeting a set of standards is to use
proportional rollback (i.e., to decrease PM levels on all days by the same percentage) for all
concentrations exceeding the background level. A sensitivity analysis also has been conducted
to examine the impact on the 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%).
In the PM25 risk assessment, air quality just meeting the current PM25 standards is
simulated by reducing the PM2 5 concentrations at the composite monitor by the same percentage
August 2003
4-26
Draft - Do Not Quote or Cite
-------
1 on all days. The percentage reduction is determined by comparing the maximum of the monitor-
2 specific annual averages (or the maximum of the monitor-specific ninety-eighth percentile daily
3 values) adjusted for background with the level of the annual (or daily) standard adjusted for
4 background.10 Because pollution abatement methods are applied largely to anthropogenic
5 sources of PM25, rollbacks were applied only to PM25 concentrations above estimated
6 background levels. The percent reduction is determined by the controlling standard, which for
7 the current suite of PM25 standards is the existing annual standard for the four urban study areas
8 that do not currently meet the annual average standard. The same basic approach will be applied
9 to simulate just meeting alternative PM25 and PM10_25 standards in the assessments that will be
10 included in the next draft of the Staff Paper.
11
12 4.2.4 Approach to Estimating PM-Related Health Effects Incidence
13 The C-R relationships used in the PM risk assessment are empirically estimated relations
14 between average ambient PM concentrations and the health endpoints of interest reported by
15 epidemiological studies for specific urban areas. Most epidemiological studies estimating
16 relationships between PM and health effects used a method referred to as "Poisson regression" to
17 estimate exponential (or log-linear) C-R functions.11 In this model,
y = B e^ , (Equation 4-1)
18 where x is the ambient PM level, y is the incidence of the health endpoint of interest at PM level
19 x, p is the coefficient of ambient y0 concentration, and B is the incidence at x=0, i.e., when there
20 is no ambient PM2 5. The difference in health effects incidence, Ay=y0 - y, from y0 to the baseline
21 incidence rate, y, that corresponds to a given difference in ambient PM25 levels, Ax = x0 - x, is
22 then
23 Ay = y[e^ - 1] (Equation 4-2)
10Since an area is allowed, if it meets certain requirements, to determine whether it meets the current annual
average standard based on the spatial average of its population-oriented monitors, in section 4.4 the percent rollbacks
that would have resulted from using this alternative approach in each study area also are presented.
nFor some studies on respiratory hospital admissions used in the risk assessment a linear C-R function was
estimated.
August 2003 4-27 Draft - Do Not Quote or Cite
-------
or, alternatively,
Ay = y(RR^ - 1) (Equation 4-3)
2 where RRAx is the relative risk associated with the change in ambient PM2 5 levels, Ax. Equations
3 4-2 and 4-3 are simply alternative ways of expressing the relation between a given difference in
4 ambient PM2 5 levels and the corresponding difference in health effects. These equations are the
5 key equations that combine air quality information, C-R information, and baseline health effects
6 incidence information to estimate ambient PM2 5 health risk.
7 The daily time-series epidemiological studies, discussed above in Chapter 3, used models
8 estimating C-R functions in which the PM-related incidence on a given day depends only on
9 same-day PM concentration or previous-day PM concentration (or some variant of those, such as
10 two-day average concentrations). Such models necessarily assume that the longer pattern of PM
1 1 levels preceding the PM concentration on a given day does not affect mortality on that day. To
12 the extent that PM-related mortality on a given day is affected by PM concentrations over a
13 longer period of time, then these models would be mis-specified, and this mis-specification
14 would affect the predictions of daily incidence based on the model. The extent to which longer-
15 term (i.e., weekly, monthly, seasonal, or annual) PM2 5 exposures affect the relationship observed
16 in the daily time-series studies is unknown. However, there is some evidence, based on analyses
17 of PM10 data, that mortality on a given day is influenced by prior PM exposures up to more than
18 a month before the date of death (Schwartz, 2000a, reanalyzed in Schwartz, 2003b). Currently,
19 there is insufficient information to adjust for the impact of longer-term exposure on mortality
20 associated with PM2 5 exposures and this is an important uncertainty that should be kept in mind
21 as one considers the results from the short-term exposure PM25 risk assessment. However, for
22 the PM10 risk assessment, distributed lag model results will be used where available.
23 The estimated PM2 5-related mortality associated with long-term exposure studies is
24 likely to include mortality related to short-term exposures as well as mortality related to longer-
25 term exposures. As discussed above, estimates of daily mortality based on the time-series
26 studies also are likely to be affected by prior exposures. Therefore, the estimated annual
August 2003 4-28 Draft - Do Not Quote or Cite
-------
1 incidences of mortality calculated based on the short- and long-term exposure studies are not
2 likely to be completely independent and should not simply be added together.
3 The statistical uncertainty surrounding the estimated PM2 5, PM10_2 5, and PM10
4 coefficients in the reported C-R functions is reflected in the intervals provided for the risk
5 estimates in sections 4.3 to 4.5. In addition, sensitivity analyses examine how the risk estimates
6 would vary if, instead of the reported C-R relationship, different hypothetical threshold models
7 were applied instead. Another sensitivity analysis addresses how the PM2 5 risk estimates would
8 change if a distributed lag model could be applied instead of the single lag models reported in
9 the literature for short-term exposure mortality. Finally, a third sensitivity analysis addresses the
10 possible impact of different assumptions about the role of historical air quality concentrations in
11 contributing to the reported effects associated with long-term exposure. The results of these
12 sensitivity analyses are summarized in section 4.3 .
13
14 4.2.5 Baseline Health Effects Incidence Rates and Population Estimates
15 The most common health risk model expresses the reduction in health risk (Ay)
16 associated with a given reduction in PM concentrations (Ax) as a percentage of the baseline
17 incidence (y). To accurately assess the impact of PM air quality on health risk in the selected
18 urban study locations, information on the baseline incidence of health effects (i.e., the incidence
19 under "as is" air quality conditions) and population size in each location is therefore needed.
20 Where possible, county-specific incidence or incidence rates have been used. County-specific
21 mortality incidences were available for the year 1998 from CDC Wonder, an interface for public
22 health data dissemination provided by the Centers for Disease Control.12 The baseline mortality
23 rates for each risk assessment location are provided in Exhibits 5.4, 5.5, and 5.6 of the technical
24 support document (Abt, 2003b). Population sizes, for both total population and various age
25 ranges used in the PM risk assessment were obtained for the year 2000 from the 2000 U.S.
26 Census data13 and are summarized in Exhibits 5.1, 5.2, and 5.3 of the technical support document
27 (Abt, 2003b).
12See http://wonder.cdc.gov/.
13See http://factfinder.census.gov/.
August 2003 4-29 Draft - Do Not Quote or Cite
-------
1 For some of the morbidity endpoints, however, county-specific incidence rates were
2 difficult to obtain. County-specific rates for cardiovascular and respiratory hospital discharges,
3 and various subcategories (e.g., pneumonia, asthma), have been obtained, where possible, from
4 state, local, and regional health departments and hospital planning commissions for each of the
5 study areas where there were C-R relationships available for hospital admissions associated with
6 PM concentrations.14 Baseline hospitalization rates used in each PM risk assessment location are
7 summarized in Exhibits 5.7, 5.8, and 5.9 of the technical support document (Abt, 2003b). For
8 respiratory symptoms in children, the only available estimates of baseline incidence rates were
9 from the studies that estimated the C-R relationships for those endpoints. However, because the
10 risk assessment locations for these endpoints were selected partly on the basis of where studies
11 were carried out, baseline incidence rates reported in these studies should be appropriate for the
12 risk assessment locations to which they were applied.
13
14 4.2.6 Concentration-Response Functions Used in Risk Assessment
15 A key component in the risk model is the set of C-R functions which provide estimates of
16 the relationship between each health endpoint of interest and ambient PM concentrations. As
17 discussed above, the health endpoints that have been included in the PM2 5, PM10_2 5, and PM10
18 risk assessment for short-term exposure include mortality, hospital admissions, and respiratory
19 symptoms not requiring hospitalization. For PM2 5 long-term exposure mortality is also
20 estimated. Health endpoints were included in the risk assessment if the overall weight of the
21 evidence from the collective body of studies supported the conclusion that there was likely to be
22 a causal relationship between PM and the specific health endpoint. Once it had been determined
23 that a health endpoint was to be included in the assessment, inclusion of a study on that health
24 endpoint was not based on the existence of a statistically significant result. Both single-
25 pollutant and, where available, multi-pollutant, C-R functions were used from the studies listed
26 in Tables 9-8, 9-10, and 9-11 of the draft CD (see also Appendix A of this SP).
14The 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.
By using the annual discharge rate, it is assumed 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.
August 2003 4-30 Draft - Do Not Quote or Cite
-------
1 As discussed in the draft CD (section 8.4.2) and Chapter 3 (section 3.5.3.1), questions
2 were raised in 2002 about the default convergence criteria (which impact the mean estimate) and
3 standard error calculations (which result in understated standard errors) used in many of the
4 short-term PM epidemiological studies that used generalized additive models (GAMs) in S-Plus.
5 To address these concerns, many of the study authors performed reanalyses of certain of the
6 studies using alternative statistical estimation approaches (e.g., GLM with different degrees of
7 freedom and different types of splines), in addition to using GAMs with a more stringent
8 convergence criterion. To avoid producing a prohibitively large set of results, the PM risk
9 assessment included C-R functions using only GAM with the more stringent convergence
10 criterion (denoted "GAM (stringent)") for all urban locations, except Chicago and Los Angeles.
11 While the GAM (stringent) functions are likely to provide better central tendency estimates, they
12 do not address the issue of understated standard errors of the coefficient estimates. Thus, the
13 confidence intervals included in the risk assessment involving use of the GAM (stringent) C-R
14 functions are somewhat understated. As indicated in the draft CD, "the extent of downward bias
15 in standard error reported in these data (a few percent to -15%) also appears not to be very
16 substantial, especially when compared to the range of standard errors across studies due to
17 differences in population size and number of days available" (CD, p.8-197). The reanalyzed
18 GLM C-R functions reflect a corrected approach to calculation of the standard errors of the
19 coefficient estimates, but are less likely to provide the best central tendency effect estimate. For
20 Chicago and Los Angeles, a wider array of C-R models was included to illustrate the impact of
21 alternative model specifications on the risk estimates.
22 More detailed information about the C-R relationships used in the PM risk assessment is
23 provided in Appendix C (Exhibits C. 1 through C.27 of the technical report (Abt, 2003b) for
24 each of the study areas. This information includes population characteristics (e.g., age and
25 disease status), form of the model (e.g., log-linear, logistic), whether other pollutants were
26 included in the model, lags used, observed minimum and maximum ambient PM concentrations,
27 and PM coefficients along with lower and upper 5th and 95th confidence intervals.
28 4.2.6.1 Hypothetical Thresholds
29 In assessing or interpreting public health risk associated with exposure to PM, the form
30 of the C-R function is a critical component. The 1996 Criteria Document (EPA, 1996a)
August 2003 4-31 Draft - Do Not Quote or Cite
-------
1 evaluated evidence from epidemiological studies regarding both functional form and whether a
2 threshold for effects could be identified; this evaluation raised some key questions, but there was
3 not sufficient evidence to draw conclusions (EPA, 1996a, section 13.6.5).
4 Among the new epidemiological studies are several that use different modeling methods
5 to investigate hypothetical threshold levels and C-R forms. As summarized in the draft CD, the
6 results of multi-city studies suggest there is no strong evidence of the existence of a threshold for
7 associations between PM and acute mortality; some single-city studies suggest a hint of a
8 threshold, but not in a statistically significant manner (CD, p. 8-241).
9 Using data from 20 U.S. cities to analyze the PM10 and short-term exposure mortality
10 relationship, roughly linear associations were found for total and cardiorespiratory mortality,
11 consistent with the lack of a threshold (CD, p. 8-239; Daniels et al., 2000). Cakmak et al. (1999)
12 tested different methods for detecting the presence of a threshold for the PM-mortality
13 relationship, using Toronto pollution and mortality data. The study authors concluded that "if a
14 threshold exists, it is highly likely that standard statistical analysis can detect it" (CD, p. 8-240).
15 Some evidence for thresholds in the relationship between PM25, but not PM10_25, and mortality
16 was found using data from Phoenix. Smith et al. (2000) found evidence suggesting a potential
17 threshold level of 20-25 |ig/m3 daily concentrations for short-term exposure mortality
18 associations with PM2 5 but no evidence of a threshold in the relationship between PM10_2 5 and
19 mortality. The draft CD (CD, p. 8-241) observes that the data set used in this analysis is small
20 but the findings warrant further analysis. Overall, considering the results of these new studies,
21 the draft CD concludes that "more data may need to be examined with alternative approaches ...
22 meanwhile, the use of a linear PM effect model appears to be appropriate" (CD, p. 8-241), which
23 is consistent with the conclusions of the previous Criteria Document (EPA, 1996a).
24 For long-term exposure to PM, the shape of the C-R function also was evaluated using
25 data from the ACS cohort, where mean PM25 concentrations ranged from abut 10 to 34 |ig/m3.
26 As shown in Figure 8-9 of the draft CD, the C-R relationships for all-cause and cardiopulmonary
27 mortality can be reasonably approximated by a linear model. However, for lung cancer, the
28 relationship appears to have a steeper linear slope at lower concentrations, with a flatter linear
29 slope at PM2 5 concentrations that exceed about 13 |ig/m3 (CD, p.8-89). It also is apparent in this
30 figure that the confidence intervals around the estimated C-R functions expand at the low and
August 2003 4-32 Draft - Do Not Quote or Cite
-------
1 high ends of the ranges of data, indicating greater uncertainty in the shape of the C-R
2 relationship across the lowest and highest concentration ranges.
3 Staff recognizes that it is reasonable to expect that, for individuals or groups of
4 individuals, there may be biological thresholds for different effects. While epidemiological
5 analyses have not identified a threshold level in the range of air quality concentrations in these
6 studies, it is possible that a threshold level exists within these ranges but cannot be detected due
7 to variability in susceptibility across a population. It is also possible that such a level may exist
8 at concentrations lower than the air quality levels observed in current studies. Staff have
9 examined the potential impact on risk estimates of possible threshold levels through sensitivity
10 analyses using various hypothetical threshold levels.
11 The PM2 5 base case risk assessment summarized in section 4.3 and 4.4 below does not
12 include a threshold. Rather risk estimates reflect the potential contribution of PM2 5 down to
13 either an estimated background level or the LML in the study, whichever is higher. For a
14 number of studies, including all of the long-term exposure mortality studies, the LML is
15 significantly above the estimated background concentrations and, therefore, there is no
16 contribution to the risk estimates from PM2 5 concentrations below the LML in these cases. As
17 noted earlier, sensitivity analyses have been conducted that do include hypothetical alternative
18 thresholds or cutpoints, where risks only are estimated due to PM25 concentrations exceeding the
19 assumed threshold concentrations. Three hypothetical thresholds (10, 15, and 20 |ig/m3) were
20 used for short-term exposure mortality and three different hypothetical thresholds (10, 12.5, and
21 15 jig/m3) were used for long-term exposure mortality.15 Results of these sensitivity analyses
22 are discussed below in section 4.3.
23 4.2.6.2 Single and Multi-Pollutant Models
24 For several of the epidemiological studies from which C-R relationships for the PM risk
25 assessment were obtained, C-R functions are reported both for the case where only PM levels
26 were entered into the health effects model (i.e., single-pollutant models) and where PM and one
27 or more other measured gaseous co-pollutants (i.e., ozone, nitrogen dioxide, sulfur dioxide,
15The groups of hypothetical thresholds (or cutpoints) used for both short- and long-term exposure mortality
are similar to those used in the prior PM2 5 risk assessment, except that the highest cutpoints (i.e., 30 ng/m3 for short-
term and 18 ng/m3 for long-term) were dropped since there is now more substantial evidence supporting the
existence of effects below these levels.
August 2003 4-33 Draft - Do Not Quote or Cite
-------
1 carbon monoxide) were entered into the health effects model (i.e., multi-pollutant models). To
2 the extent that any of the co-pollutants present in the ambient air may have contributed to the
3 health effects attributed to PM in single-pollutant models, risks attributed to PM might be
4 overestimated where C-R functions are based on single-pollutant models. However, as shown in
5 Figure 3-12 (pp. 3-75 - 3-76) for PM25 and in the last review (EPA, 1996b, p. V-55) for PM10,
6 the magnitude and statistical significance of the associations reported between PM25 (and PM10)
7 and mortality due to short-term exposure show no trends with the levels of any of the four
8 gaseous co-pollutants examined. As stated earlier, "While not definitive, these consistent
9 patterns indicate that it is more likely that there is an independent effect of PM25, as well as
10 PM10, that is not confounded or appreciably modified by the gaseous co-pollutants." (SP, p. 3-
11 74)
12 For some of the gaseous co-pollutants, such as carbon monoxide, nitrogen dioxide, and
13 sulfur dioxide, which tend to be highly correlated with ambient PM2 5 concentrations in some
14 cities, it is difficult to sort out whether these pollutants are exerting any independent effect from
15 that attributed to PM25. As discussed in the draft CD, inclusion of pollutants that are highly
16 correlated with one another can lead to misleading conclusions in identifying a specific causal
17 pollutant. When collinearity exists, multi-pollutant models would be expected to produce
18 unstable and statistically insignificant effects estimates for both PM and the co-pollutants (CD,
19 p. 9- 73). Given that single and multi-pollutant models each have both potential advantages and
20 disadvantages, with neither type clearly preferable over the other, risk estimates based on both
21 single and multi-pollutant models are reported where both are available.
22 4.2.6.3 Single, Multiple, and Distributed Lag Functions
23 The question of lags and the problems of correctly specifying the lag structure in a model
24 are discussed extensively in the draft CD (section 8.4.5). The draft CD points out that, in most
25 PM times-series studies, after the basic model is fit (before considering PM), several different
26 PM lags are typically fit in separate single-lag models and the most significant lag is chosen.
27 The draft CD notes that "while this practice may bias the chance of finding a significant
28 association, without a firm biological reason to establish a fixed pre-determined lag, it appears
29 reasonable" (p. 8-234). There is recent evidence (Schwartz., 2000a, reanalyzed in Schwartz,
30 2003b), that the relation between PM and health effects may best be described by a distributed
August 2003 4-34 Draft - Do Not Quote or Cite
-------
1 lag (i.e., the incidence of the health effect on day n is influenced by PM concentrations on day n,
2 day n-1, day n-2 and so on). If this is the case, a model that includes only a single lag (e.g., a 0-
3 day lag or a 1-day lag) is likely to understate the total impact of PM. The draft CD makes this
4 point, noting that "if one chooses the most significant single lag day only, and if more than one
5 lag day shows positive (significant or otherwise) associations with mortality, then reporting a RR
6 [relative risk] for only one lag would also underestimate the pollution effects" (p. 8-236).
7 Because of this, a distributed lag model is considered preferable to a single lag model.
8 However, distributed lag models have been used in only a few cases and only for PM10 (e.g.,
9 Schwartz, 2000b, reanalyzed in Schwartz, 2003b). Where distributed lag models were available,
10 they were included in the PM risk assessment. When a study reports several single lag models,
11 unless the study authors identify a "best lag", the following lag models were included in the risk
12 assessment based on the assessment in the draft CD:
13 • both 0- and 1-day lag models for mortality (both total and cause specific),
14
15 • both 0- and 1-day lag models for cardiovascular and respiratory hospital
16 admissions, and
17
18 • 0-, 1-, and 2-day lag models (if all three were available) for COPD hospital
19 admissions.
20
21 In addition, for two urban locations (Chicago and Los Angeles), the risk assessment included all
22 single lag models used in a study to illustrate the effect of lag structure on the risk estimates. A
23 sensitivity analysis was also conducted to examine the potential impact of using a distributed lag
24 approach for short-term exposure mortality associated with PM2 5 based on the distributed lag
25 analysis of PM10 and mortality (Schwartz, 2000b, reanalyzed in Schwartz, 2003b).
26 4.2.6.4 Long-term Exposure Mortality PM2 5 Concentration-Response Functions
27 There are far fewer long-term exposure studies than short-term exposure studies in the
28 peer-reviewed literature. The available long-term exposure mortality C-R functions are all based
29 on cohort studies, in which a cohort of individuals is followed over time. Two cohorts that have
30 been studied are particularly relevant for the purposes of this risk assessment. One cohort, in six
31 U.S. cities, was originally followed in a study referred to as the Six Cities study. The other
32 cohort, of participants enrolled by American Cancer Society (ACS) volunteers, was composed of
33 a much larger sample of individuals from many more cities. It was originally followed in a study
August 2003 4-3 5 Draft - Do Not Quote or Cite
-------
1 referred to as the ACS study. There have recently been reanalyses of both the Six Cities study
2 and the ACS study by Krewski et al. (2000), referred to here as Krewski et al. (2000) - Six Cities
3 reanalysis, and Krewski et al. (2000) - ACS reanalysis. Both of these reanalyses are used in the
4 PM2 5 risk assessment. In addition, Pope et al. (2002) extended the follow-up period for the ACS
5 cohort to sixteen years and published findings on the relation of long-term exposure to PM2 5 and
6 all-cause mortality as well as cardiopulmonary and lung cancer mortality (referred to here as
7 Pope et al. (2002) - ACS extended. Three different indicators of PM25 exposure were considered
8 in this extended ACS study. The first indicator, 1979-1983, was the period considered in the
9 original ACS study as well as in the Krewski reanalysis. The second indicator was 1999-2000,
10 and the third was an average of the 1979-1983 and 1999-2000 PM2 5 ambient concentrations.
11 The C-R functions based on PM2 5 from 1979-1983 are included in the PM2 5 risk assessment,
12 because this period is the one used by the study authors in their main presentation of results.
13 Table 3-3 summarizes the total mortality findings from the long-term exposure mortality
14 studies discussed above, as well as two other PM cohort studies (i.e., the AHSMOG and
15 Veterans' Cohort studies) that are not included in the PM25 risk assessment. In addition, Table
16 8-11 from the draft CD presents results for total, cardiovascular and lung cancer mortality. As
17 discussed in Chapter 3 (see section 3.3.1.2), the draft CD concludes that the "lack of consistent
18 findings in the AHSMOG study and negative results of the VA study, do not negate the findings
19 of the Six Cities and ACS studies." (CD, p. 8-105) The Six Cities and ACS studies were based
20 on measured PM data (in contrast with AHSMOG PM estimates based on TSP or visibility
21 measurements), have study populations more similar to the general population than the VA study
22 cohort, and have been validated through an exhaustive reanalysis (CD, p. 8-105). Lacking these
23 advantages, the AHSMOG and VA studies were excluded from the risk assessment. Overall, the
24 draft CD concludes that "there is substantial evidence for a positive association between long-
25 term PM exposure to PM (especially fine particles) and mortality" (CD, p. 8-105).
26
27 4.2.7 Characterizing Uncertainty and Variability
28 An important issue associated with any population health risk assessment is the
29 characterization of uncertainty and variability. Uncertainty refers to the lack of knowledge
30 regarding the actual values of model input variables (parameter uncertainty) and of physical
August 2003 4-36 Draft - Do Not Quote or Cite
-------
1 systems or relationships (model uncertainty - e.g., the shapes of C-R functions). In any risk
2 assessment uncertainty is, ideally, reduced to the maximum extent possible. It can be reduced by
3 improved measurement and improved model formulation.
4 Variability refers to the heterogeneity in a population or parameter. For example, there
5 may be variability among C-R functions describing the relation between PM2 5 and mortality
6 across urban areas. This variability may be due to differences in population (e.g., age
7 distribution), population activities that affect exposure to PM (e.g., use of air conditioning),
8 levels of PM and/or co-pollutants, and/or other factors that vary across urban areas.
9 The current risk assessment incorporates some of the variability in key inputs to the
10 assessment by using location-specific inputs (e.g., location-specific C-R functions, baseline
11 incidence rates, and air quality data). Although spatial variability in these key inputs across all
12 U.S. locations has not been fully characterized, variability across the selected locations is
13 imbedded in the assessment by using, to the extent possible, inputs specific to each urban area.
14 Temporal variability is more difficult to address, because the risk reduction portions of the risk
15 assessment (i.e., estimated risk reduction associated with just meeting specified standards) focus
16 on some unspecified time in the future when specified PM standards are just met. To minimize
17 the degree to which values of inputs to the assessment may be different from the values of those
18 inputs at that unspecified time, we have used the most current inputs available (i.e., year 2002 air
19 quality data for most locations and the most recent available mortality baseline incidence rates
20 (from 1998)). However, we have not tried to predict future changes in inputs (e.g., future
21 population levels or possible changes in baseline incidence rates).
22 The following are among the major sources of uncertainty in the PM health risk
23 assessment:
24 • There is uncertainty about whether each of the estimated associations between the three
25 PM indicators (PM2 5, PM10_2 5, and PM10) and the various health endpoints included in
26 this risk assesment actually reflect a causal relationship. There are varying degrees of
27 uncertainty associated with the various PM indicators and health endpoints related to
28 differences in the weight of evidence supporting judgments about whether an observed
29 association truly reflects a causal relationship.
30
31 • There are a number of uncertainties related to estimating the C-R functions, including the
32 following:
33
August 2003 4-37 Draft - Do Not Quote or Cite
-------
1 • There is uncertainty surrounding estimates of PM coefficients in C-R functions
2 used in the assessment.
3
4 • There is uncertainty about the specification of the model (including the shape of
5 the C-R relationship), particularly whether or not there are thresholds below
6 which no response occurs.
7
8 • There is uncertainty related to the transferability of PM C-R functions from study
9 locations to the locations selected for the risk assessment and from study
10 location.16 A C-R function in a study location may not provide an accurate
11 representation of the C-R relationship in the assessment location(s) because of (1)
12 variations in PM composition across cities, (2) the possible role of associated co-
13 pollutants in influencing PM risk, (3) variations in the relation of total ambient
14 exposure (both outdoor and ambient contributions to indoor exposure) to ambient
15 monitoring in different locations (e.g, due to differences in air conditioning use in
16 different regions of the U.S.), (4) differences in population characteristics (e.g.,
17 the proportions of members of sensitive subpopulations) and population behavior
18 patterns across locations.
19
20 • There is uncertainty related to the extent to which PM C-R functions derived from
21 studies in a given location and time when PM concentrations were higher provide
22 accurate representations of the C-R relationships for the same location with lower
23 annual average concentrations and fewer high daily average concentrations.
24
25 • There are a number of uncertainties related to use of baseline incidence rates and
26 population data in the risk assessment, including the following:
27
28 • There is uncertainty related to the extent to which baseline incidence rates, age
29 distribution, and other demographic variables that impact the risk estimates vary
30 for the year(s) when the actual epidemiology studies were conducted, the recent
31 year of air quality used in the assessment, and some unspecified future year when
32 air quality is adjusted to just meet the current or alternative standards.
33
34 • There is uncertainty related to the use of annual incidence rate data to develop
35 daily health effects incidence data.
36
37 • For the respiratory symptoms endpoint, baseline health effects incidence data
38 were only available as a total estimate for all six urban areas combined. This
39 introduces additional uncertainty because the risk assessment applied the overall
40 incidence rate from the six cities to individual cities (i.e., Boston and St. Louis).
41
16 The risk assessment locations were selected partly on the basis of where C-R functions were estimated,
specifically to avoid this important source of uncertainty. Therefore, this is a source of uncertainty in the risk
assessment only when C-R functions from multi-city studies are applied to a risk assessment location.
August 2003 4-3 8 Draft - Do Not Quote or Cite
-------
1 • There are uncertainties related to the air quality adjustment procedure used to simulate
2 just meeting the current PM2 5 standards and alternative PM2 5 and PM10_2 5 standards and
3 about estimated background concentrations for each location.
4
5 • There are uncertainties associated with use of baseline health effects incidence
6 information that is not specific to the assessment locations.17
7
8 The uncertainties from some of these sources - in particular, the statistical uncertainty
9 surrounding estimates of the PM coefficients in C-R functions — were characterized
10 quantitatively. It was possible, for example, to calculate confidence intervals around risk
11 estimates based on the uncertainty associated with the estimates of PM coefficients used in the
12 risk assessment. These confidence intervals express the range within which the risks are likely
13 to fall if the sampling error uncertainty surrounding PM coefficient estimates were the only
14 uncertainty in the assessment.18 There are, of course, several other significant uncertainties in
15 the risk assessment, as noted above. If there were sufficient information to characterize these
16 sources of uncertainty quantitatively, they could be included in a Monte Carlo analysis to
17 produce confidence intervals that more accurately reflect all sources of uncertainty.
18 We handled other uncertainties in the risk assessment in several ways:
19 • Limitations and assumptions in estimating risks and risk reductions were clearly stated
20 and explained.
21
22 • Sensitivity analyses were conducted to illustrate the effects of changing key default
23 assumptions on the results of the assessment, and quantitative comparisons were
24 presented to inform other analytic choices.19 See Table 4-9 for a summary of the
25 sensitivity analyses and quantitative comparisons conducted for the PM risk assessment.
26
27
17 Location-specific baseline incidence rates were obtainable for most health endpoints. The only health
endpoints for which this is not the case are cough and lower respiratory symptoms, for which baseline incidence
rates were reported in the study (Schwartz and Neas, 2000). These rates were based on six cities combined. Boston
and St. Louis, the two assessment locations where these endpoints are evaluated, were two of the six cities.
18However, as discussed earlier in section 4.2.6, for the short-term C-R functions based on reanalyzed
GAM (stringent) models the confidence intervals are somewhat understated.
19" Sensitivity analyses" refers to assessing the effects of uncertainty on some of the final risk estimates;
"quantitative comparisons" refer to numerical comparisons (e.g. comparisons of monitor values) that are not carried
that far.
August 2003 4-39 Draft - Do Not Quote or Cite
-------
1 4.3 PM2 5, PM10 2 5, and PM10 RISK ESTIMATES FOR CURRENT ("AS IS") AIR
2 QUALITY
3
4 4.3.1 Summary of Risk Estimates
5 The risk estimates associated with "as is" PM2 5, PM10_2 5, and PM10 concentrations in
6 excess of background levels are summarized in a series of figures in this section. The risk
7 estimates are expressed in terms of percent of total incidence for each health endpoint in these
8 figures. The legend of symbols used for the different health endpoints is presented in Figure 4-2.
9 For each series of estimates, a mean effect estimate is provided along with 95% confidence
10 intervals.20 Additional detailed tables which present the estimated incidence (both as the number
11 of effects and as a percentage of total incidence) for each risk assessment location are included in
12 the technical support document (see Appendix D). Risk estimates in a given assessment location
13 are presented only for those health endpoints for which there is at least one acceptable C-R
14 function reported for that location. Therefore, the set of health effects shown in the figures
15 varies for the different locations.
16 Figures 4-3 through 4-8 summarize the PM25 risk estimates across the various assessment
17 locations. Figure 4-3 shows risk estimates for six urban areas for total non-accidental mortality
18 associated with short-term exposure to PM25 using single-pollutant, single-city models. Most of
19 the mean effect estimates are in the range 0.5 to 1.5% of total non-accidental mortality, with one
20 location, San Jose, showing the highest risk (about 3%) with 0-day lag and the same area also
21 having the lowest risk (0%) with 1-day lag. Figure 4-4 shows risk estimates for two locations
22 (Boston and St. Louis) for various short-term exposure mortality endpoints using single-city
23 versus multi-city models. Generally, the mean risk estimates for the single- and multi-city
24 models are roughly comparable, with lower risk estimates seen in Boston for the multi-city
25 models compared to the single-city models and the reverse being observed in St. Louis. As
26 expected, the 95% confidence intervals are somewhat smaller for most of the multi-city models
27 compared to the respective single-city model given the greater sample size.
20Risk estimates less than zero were truncated at zero since we have no reason or evidence to assume that
increasing ambient PM concentrations results in improved health. In cases where the lower confidence bound was <
0, the interval shown does not represent a 95% confidence interval around the mean estimate. In these cases, the
upper bound is the 97.5 percentile and the lower confidence bound has been truncated at zero.
August 2003 4-40 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
Table 4-9. Sensitivity Analyses and Quantitative Comparisons
Analysis
Number
(Figure 4-1)
1
2
3
4
5
6
7
PM
Indicator
PM25
PM25,
PM10,
PM10.2.5
PM25
PM25
PM25,
PM10,
PM10.2.5
PM25
PM25,
PM10,
PM10.2.5
Component of
the Risk
assessment
Air Quality
Air Quality
Air Quality
Air Quality
Concentration-
Response
Concentration-
Response
Concentration-
Response
Sensitivity Analysis or Comparison
A sensitivity analysis of the effect of assuming a
constant background PM level versus a distribution of
daily background levels
A sensitivity analysis of the effect of assuming
different (constant) background PM levels
A sensitivity analysis 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
A comparison of the effect of just meeting the current
annual PM25 Standard of 15 i-ig/m3 using the maximum
versus the average of monitor-specific averages
A sensitivity analysis using an approach to estimate the
possible impact of using a distributed lag C-R function
A sensitivity analysis 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
A sensitivity analysis assuming alternative
hypothetical threshold concentration levels for the
occurrence of PM25-related response at concentrations
above those for background
10
11
12
13
14
15
16
Source: Abt Associates (2003b).
August 2003
4-41
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
o
X
o
v
All-cause Mortality
Asthma Hospital Admissions
Asthma Symptoms
Cardiovascular Hospital Admissions
Cardiovascular Mortality
Circulatory Mortality
Congestive Heart Failure Hospital Admissions
COPD Hospital Admissions
COPD Mortality
Cough
Dysrhythmias Hospital Admissions
Ischemic Heart Disease Hospital Admissions
Ischemic Heart Disease Mortality
Lower Respiratory Symptoms
Non-accidental Mortality
Pneumonia Hospital Admissions
Pneumonia Mortality
Respiratory Mortality
Figure 4-2. Legend for Health Endpoints in Figures
August 2003
4-42
Draft - Do Not Quote or Cite
-------
1
2
o
3 #
£
4 !„ .
5 |
'o
6 |« -
O
7 £
"c
8 1- •
D^
9 B
10 J - •
11 ^
o
12
1
1
1
1
1
1
i
•
2
1
1
3 *
1
1
1
C
£
i
'
5
i
i
i
i
i
5
i
3
1
1
I I
I I
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Boston
Detroit
Los Angeles
Pittsburgh
San Jose
St. Louis
Figure 4-3. Estimated Annual Percent of Total (Non-Accidental) Mortality Associated with Short-Term Exposure to PM2 5
(and 95 Percent Confidence Interval): Single-Pollutant, Single-City Models*
Source: Abt (2003b)
1. Schwartz etal. (1996)*
2. Lippmannetal. (2000)**
3. Moolgavkar (2000a)*** - 0-day lag
4. Moolgavkar (2000a)*** - 1-day lag
5. Chock et al. (2000) (age<75)
6. Chock et al. (2000) (age>75)
7. Fairley (1999)**** - 0-day lag
8. Fairley (1999)**** - 1-day lag
*Reanalyzed in Schwartz (2003 a)
"Reanalyzed in Ito (2003)
* "Reanalyzed in Moolgavkar (2003)
****Reanalyzed in Fairley (2003)
*In cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-43
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
O (0
eS
in
O)
c
o
c
|
)
1
I
1
1
£
7
3
Boston
St. Louis
Figure 4-4. Estimated Annual Percent of Health Effects Associated with Short-Term Exposure to PM2 5 (and 95 Percent
Confidence Interval): Results Based on Single-City versus Multi-City Models*
(Single-city models are always on the left, followed by the corresponding multi-city models)
Source: Abt (2003b)
1. Schwartz et al. (1996)* - single-city; non-accid. mort. 7. Klemm et al. (2000)** - single-city; pneumonia mort.
2. Schwartz et al. (1996)* - 6 cities; non-accid. mort. 8. Klemm et al. (2000)** - 6 cities; pneumonia mort.
3. Klemm et al. (2000)** - single-city; COPD mort. *Reanalyzed in Schwartz (2003a)
4. Klemm et al. (2000)** - 6 cities; COPD mort. ** Reanalyzed in Klemm and Mason (2003)
5. Klemm et al. (2000)** - single-city; ischemic heart disease mort.
6. Klemm et al. (2000)** - 6 cities; ischemic heart disease mort.
JIn cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-44
Draft - Do Not Quote or Cite
-------
1
2
3 » -
4 o
^ LO
6 ? m •
o -D
7 ^
0
O c
9 |* '
10 |
11 i*
12 °"
13 §
14 1 M
15 I
(0
16 |
17 ^
18 |
19 E
20
o
21
10
i
13
i
4
£
1
1
1
2
3
i i
i
,
10
7 1
9
i
6 i
I J
;
9
1
8
X
i
i
i
i
1
11 i
I
|
|
1
2
i i
i
1
i
i
4
i
16
i
I
1
|
I .
19
1
,718
i
99
^^ Los Angeles Pittsburgh San Jose
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
Figure 4-5. Estimated Annual Percent of Health Effects Associated with Short-Term Exposure to PM2 5 (and 95 Percent
Confidence Interval): Results Based on Single-Pollutant versus Multi-Pollutant Models*
(Single-pollutant models are always on the left, followed by the corresponding multi-pollutant models)
Source: Abt (2003b)
1. Moolgavkar (2000a)* - non-accid. mort.
2. Moolgavkar (2000a)* - non-accid. mort.; + CO
3. Moolgavkar (2000a)* - cardiovasc. mort.
4. Moolgavkar (2000a)* - cardiovasc. mort.; + CO
5. Moolgavkar (2000b)* - cardiovasc. hosp. adm.
6. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; + CO
7. Moolgavkar (2000c)* - COPD hosp. adm.
8. Moolgavkar (2000c)* - COPD hosp. adm.; + NO2
10. Chock et al. (2000) (age<75) - non-accid. mort.;
+ C0,03, S02,PM10-2.5
11. Chock et al. (2000) (age>75) - non-accid. mort.
12. Chock et al. (2000) (age>75) - non-accid. mort.;
+ CO, O3,SO2,PM10-2.5
13. Fairley (1999)** -non-accid. mort.;
14. Fairley (1999)** -non-accid. mort.; + COH
15. Fairley (1999)** - non-accid. mort.; + NO3
16. Fairley (1999)** -non-accid. mort.; with SO4
17. Fairley (1999)** - non-accid. mort.; with NO2
18. Fairley (1999)** -non-accid. mort.; + CO
19. Fairley (1999)** - non-accid. mort.; + 8 hr O3
*Reanalyzed in Moolgavkar (2003)
"Reanalyzed in Fairley (2003)
9. Chock et al. (2000) (age<75) - non-accid. mort.
JIn cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-45
Draft - Do Not Quote or Cite
-------
1
2
~ CO -
3 0
4 ss
in
0)
J T3
C
fi ™
° 'CB CD -
7 |
8 ^
o c
9 a
10 t. .
11 1
O
12 |
13 1
03
14 ! " •
CO
15 ^
16 1
LJJ
17
18
1
i
i
|
19
20
21
2
i i
i "
1 1 1
<
3
ii
4 5
:• T I s J
Los Angeles
7 i
. <
.
5 c
'
1
<
11
^
10
3
i
,
<
^
*
Phoenix
22 Figure 4-6. Estimated Annual Percent of Mortality Associated with Short-Term Exposure to PM2 5 (and 95 Percent
23 Confidence Interval): Effect of Different Lag Models*
24 Source: Abt (2003b)
25
26 1 . Moolgavkar (2000a)* - non-accid. mort. - 0-day lag 7. Moolgavkar (2000a)* - cardiovascular mort. - 0-day lag
27 2. Moolgavkar (2000a)* - non-accid. mort. - 1-day lag 8. Moolgavkar (2000a)* - cardiovascular mort. - 1-day lag
28 3. Moolgavkar (2000a)* - non-accid. mort. - 2-day lag 9. Moolgavkar (2000a)* - COPD mort
29 4. Moolgavkar (2000a)* - non-accid. mort. - 3-day lag 10. Moolgavkar (2000a)* -
- 0-day lag
COPD mort. -1 -day lag
30 5 . Moolgavkar (2000a)* - non-accid. mort. - 4-day lag 11. Mar et al. (2000)* * - cardiovascular mort. - 0-day lag
31 6. Moolgavkar (2000a)* - non-accid. mort. - 5-day lag 12. Mar etal. (2000)** -cardiovascular mort. - 1 -day lag
32
33
*Reanalyzed in Moolgavkar (2003)
"Reanalyzed in Mar et al. (2003)
34 *In cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
3 5 97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-46
Draft - Do Not Quote or Cite
-------
1
2
3
4 |
5 °> (D -
T3
A c
r> ro
w
-------
1
3
4 o
LO
O "D
7 ra
8 ?» -
" c
9 1
10 ^
1 1 -
11 0 tD
12 £
13 -
14 g
15 -
16 1
17 •«
18 £" '
19 i
20
21 ° -
2*4
ills
P"I
5
3
9
2 4
9 9
1 "
-------
1 Figure 4-5 compares risk estimates for three locations for various short-term exposure to
2 PM2 5 health endpoints examining single-pollutant versus multi-pollutant models. In some cases
3 there is relatively little difference in the risk estimates between the single-pollutant and multi-
4 pollutant models (e.g., Pittsburgh), while in other cases there are larger differences depending on
5 the pollutant(s) added to the model.
6 Figure 4-6 illustrates the impact of using different lag models for short-term exposure
7 mortality associated with PM2 5 in two locations (Los Angeles and Phoenix). Generally, the risk
8 estimates are greatest for the 0- and 1-day lag models. This is consistent with the earlier
9 discussion of lags and short-term exposure mortality in the draft CD and Chapter 3.
10 Figures 4-7 and 4-8 show risk estimates for mortality related to long-term exposure to
11 PM2 5 based on single- and multi-pollutant models, respectively. The mean risk estimates for the
12 single-pollutant models, which include C-R functions based on the Six Cities and ACS studies
13 (Krewski et al., 2000), as well the updated ACS study (Pope et al., 2002), range from less than
14 0.5% in Boston and Phoenix to as high as roughly 4 to 5% of total mortality in Los Angeles and
15 St. Louis, with most estimates falling in the 1.5 to 2.5% range. As shown in Figure 4-9, the risk
16 estimates based on multi-pollutant models, involving addition of a single co-pollutant in the ACS
17 study, show generally greater risk associated with PM25 when CO, NO2, or O3 were added to the
18 models and lower risk associated with PM25 when SO2 was added.21
19 Figure 4-9 shows risk estimates for short-term exposure mortality, hospital admission,
20 and respiratory symptoms associated with PM10_2 5 for five urban areas (Detroit, Philadelphia,
21 Phoenix, Seattle, and St. Louis). For Detroit risk estimates are provided for non-accidental
22 mortality, and two types of cause specific mortality, as well as several categories of
23 cardiovascular and respiratory-related hospital admissions. For Philadelphia and Phoenix only
24 cardiovascular mortality risk estimates were able to be developed and only asthma hospital
25 admissions for Seattle. There is significant variability in the effect estimates, expressed as
26 percent of total incidence, for mortality associated with PM10_25, with three (Detroit,
27 Philadelphia,
21 The addition of a second pollutant reduced the number of cities available for estimating the C-R 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 38
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.
August 2003 4-49 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
(N
O "~
£
in
O)
-o o
C T-
ro
o
c
0
| « -
&
'o <°
1
0
0^
F
to
LU
1 ™
LU
O
1
! ;
>
3
i!
| J
3
1 £
1
J
1
1
1
{
1 1
1
3
i
[
£
4
<
3
1
4
<
«
0
1
1
11
{
I
1
12
-3
1
1
4
4
4
4
Detroit
Philadelphia Phoenix
Seattle
St. Louis
Figure 4-9. Estimated Annual Percent of Health Effects Associated with Short-Term Exposure to PM10_2 5 (and 95 Percent
Confidence Interval)*
Source: Abt (2003b)
1. Lippmann et al. (2000)* - non-accid. mort.
2. Lippmann et al. (2000)* - circulatory mort.
3. Lippmann et al. (2000)* - respiratory mort.
4. Lippmann et al. (2000)* - pneumonia hosp. adms.
5. Lippmann et al. (2000)* - COPD hosp. adms.
6. Lippmann et al. (2000)* - ischemic heart disease hosp. adms.
7. Lippmann et al. (2000)* - dysrhythmia hosp. adms.
8. Lippmann et al. (2000)* - congestive heart failure hosp. adms.
9. Lipfert et al. (2000) - 7-county model; cardiovasc. mort.
JIn cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
10. Mar et al. (2000)** - cardiovasc. mort.
11. Sheppard et al. (1999)* * * - asthma hosp. adms.
12. Schwartz et al. (1996)**** - non-accid. mort.
13. Schwartz and Neas (2000) - 6-city model; lower resp. symptoms
14. Schwartz and Neas (2000) - 6-city model; cough
*Reanalyzed in Ito (2003)
"Reanalyzed in Mar et al. (2003)
""Reanalyzed in Schwartz (2003b)
August 2003
4-5Dmft - Do Not Quote or Cite
-------
1 and Phoenix) of the four urban areas showing mean effect estimates ranging from about 0.3 to
2 over 6% In the fourth area, St. Louis, no mean effect on mortality is reported for PM10_2 5.
3 However, for St. Louis there were mean effect estimates ranging from roughly 3 to 4% of total
4 incidence for lower respiratory symptoms and cough.
5 Figures 4-10 through 4-14 summarize the PM10 risk estimates across the various
6 assessment locations. Figure 4-10 shows risk estimates for total (non-accidental) mortality
7 associated with short-term exposure to PM10 in 11 urban areas across the U.S. for a recent year of
8 air quality. Most, but not all, of the mean effect estimates fall within a range of about 0.5 to
9 2.5% of total incidence. Only the risk estimates for Minneapolis, which are based on distributed
10 lag models that are expected to yield higher estimates, result in higher mean effect estimates for
11 short-term exposure mortality, showing a range of roughly 3.5 to 4%. The 12 urban areas
12 included in the PM10 risk assessment vary considerably in terms of the relative contribution of
13 PM25 and PM10_25 to total PM10 mass. Several cities are dominated by PM25 (e.g., Detroit,
14 Pittsburgh, St. Louis), several have roughly equal PM25 and PM10_25 (Los Angeles, San Jose, and
15 Seattle) and at least one is dominated by PM10_25 (Phoenix). There is no apparent pattern in the
16 PM10 mortality risk estimates related to the relative ratio of fine- and coarse-fraction PM across
17 the 12 urban areas.
18 Single- versus multi-city model short-term exposure mortality and morbidity risk
19 estimates are displayed in Figures 4-11 and 4-13, respectively. Single-city models are always on
20 the left, followed by corresponding multi-city models. Generally the mean effect estimates are
21 similar between the single- and multi-city models for short-term mortality, although the 95%
22 confidence intervals are generally smaller for the multi-city models due to larger sample size.
August 2003 4-51 Draft - Do Not Quote or Cite
-------
1
2
o
3 #
J in
O)
4 1
o ^
5 1
o
c
6 i „
|2
7 |
0
o
§ 0 (N -
9 1
•+=
10 fi - -
•5
11 £
LU
12
o
13
1
i
i
i
2
i
i
i
i
i
i
i
i
1 o
3
J
t
1 c
i
1
1
f 6 7
if
!
,
|
14
1^ Boston
15
16
17
2
9
•
i
£
i
3 i
1.
5
Chicago
Detroit
Q
i
3
•
i
i
i
i
i
(.
1
,
i -
i
i
7
\
1
i
(
(
3
•
i
3
3
i
2 1
9 '
3
li,
3
11
•
i
i
Los Angeles Minneapolis Phoenix Pittsburgh
18 Figure 4-10. Estimated Annual Percent of Total
Philadelphia
i
1
i
12
i
i
3 '
i
i
13
i
i
T
I ,
i
i
<
i
>
3
5
1
i
i
i
i
Prove San Jose Seatle St. Louis
(Non-Accidental) Mortality Associated with Short-Term Exposure to PM10
19 (and 95 Percent Confidence Interval): Single-Pollutant, Single-City Models*
20 Source: Abt (2003b)
21 1. Klemm et al. (2000)*
22 2. Schwartz (2000b)« —
distr.
lag
9. Kinney et al.
10. Chock etal
23 3. Samet etal. (2000)***
24 4. Sty er etal. (1995
25 5. Ito and Thurston( 1996)
11. Chock etal
(1995)
(2000) - age<75 *Reanalyzed in Klemm and Mason (2003)
(2000) - age>75 "Reanalyzed
in Schwartz (2003a)
12. Fairley (1999)****** -0-day lag ***Reanalyzed in Dominici etal. (2002)
13. Fairley (1999)****** _ 1-day lag ""Reanalyzed in Moolgavkar (2002)
26 6. Moolgavkar(2000a)****- 0-day lag
27 7. Moolgavkar(2000a)****- 1-day lag
28 8. Lippmann et al.
*****Reanalyzed in Ito (2003)
"""Reanalyzed in Fairley (2003)
(2000)*****
29 JIn cases where the lower bound of the confidence interval was less than zero, the
interval does not represent a 95% confidence interval. In these cases, the upper bound is the
3 0 97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-52
Draft - Do Not Quote or Cite
-------
1
2
3
4 ge "> '
5 °>
•^ T3
6 ™
7 I" -
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
1
1
1
1 ^
1
1
1
1
1
3
14
1 '
T '
i
', ]
' 6
i
7
Pf
I
[
]
[
i i
]
[
I
3
'1
]
i
i
i
i
0
1
1
i
i
j
i i
1\
1 i
li
21
i i
i i
i
1
i
3
i
41
i
i
5
1
i
i
1
i
6
i
i
i
i
1
t
7
i
i
18
i •
9
i
2
4
0
2
=
2
i
1
j
22
J
3
[
2
!
'.
2
3
4 ,
3
5
'.
2
i
!
i
2
1
6
E
i
E
i
72
i i
i i
i
8
2
1
|
1
3
i
9
i
i
i
, '
0
3
1
i
3
i
1
i
j
2
I
I
oo
4
Chicago
Detroit
Los Angeles
Minneapolis Pittsburgh
Seatle
Figure 4-12. Estimated Annual Percent of Health Effects Associated with Short-Term Exposure to PM10 (and 95 Percent
Confidence Interval): Results Based on Single-Pollutant versus Multi-Pollutant Models1
(Single-pollutant models are always on the left, followed by the corresponding multi-pollutant models)
Source: Abt (2003b)
1. Ito and Thurston (1996) - non-accid. mort.
2. Ito and Thurston (1996) - non-accid. mort.; +O3
3. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; 0-day lag
4. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; 0-day lag +CO
5. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; 1-day lag
6. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; 1-day lag +CO
7. Morris and Naumova (1998) - cong. heart failure hosp. adm.
8. Morris and Naumova (1998) - cong. heart failure hosp. adm.;
+CO, SO2, NO2, O3
9. Schwartz and Morris (1995)- cong. heart failure hosp. adm.
10. Schwartz and Morris (1995)- cong. heart failure hosp. adm.;+CO
11. Schwartz and Morris (1995) - ischemic heart disease hosp. adm.
12. Schwartz and Morris (1995) - ischemic heart disease hosp. adm.; +CO
13. Schwartz and Morris (1995) - ischemic heart disease hosp. adm.; +SO2
14. Kinney etal. (1995) - non-accid. mort.
15. Kinney etal. (1995) -non-accid. mort.;+O3
16. Kinney etal. (1995) -non-accid. mort.;+CO
August 2003
4-54
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
29. Chock et al. (2000), age<75 - non-accid. mort.
30. Chock et al. (2000), age<75 - non-accid. mort.; + NO2
31. Chock et al. (2000), age>75 - non-accid. mort.
32. Chock et al. (2000), age>75 - non-accid. mort.; + CO
33. Yu et al. (2000) - asthma symptoms**
34. Yu et al. (2000) - asthma symptoms; + CO, SO2
*Reanalyzed in Moolgavkar (2003)
17. Moolgavkar (2000a)* - non-accid. mort.; 2-day lag*
18. Moolgavkar (2000a)* - non-accid. mort.; 2-day lag; + CO*
19. Moolgavkar (2000a)* - cardiovasc. mort.; 2-day lag*
20. Moolgavkar (2000a)* - cardiovasc. mort.; 2-day lag; + CO*
21. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; 2-day lag*
22. Moolgavkar (2000b)* - cardiovasc. hosp. adm.; 2-day lag; + CO*
23. Moolgavkar (2000c)* - COPD. hosp. adm.; 0-day lag
24. Moolgavkar (2000c)* - COPD. hosp. adm.; 0-day lag; + NO2
25. Moolgavkar (2000c)* - COPD. hosp. adm.; 1-day lag
26. Moolgavkar (2000c)* - COPD. hosp. adm.; 1-day lag; + NO2
27. Schwartz (1994b) - pneumonia hosp. adm.
28. Schwartz (1994b) - pneumonia hosp. adm.; + O3
*In cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-55
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
(N
O "~
ro ^~
"a?
o
c
£
!
!
j
3
1
f 1
i 2
X | I
I x
i
3 '
i (
(
1
(
C
3
| 3
i
1 3
^ £
s S
( i
i
1
i !
J /
1
2
x
i
3
/
I
I
1
£
3
i '
i
;
5
1
1
!2
¥ '
X
i
3
i
1
1
1
C
i
3
3
\
6
3 E
,e
i!
Chicago
Detroit
Minneapolis
Pittsburgh
Provo
Seattle
Figure 4-13. Estimated Annual Percent of Health Effects Associated with Short-Term Exposure to PM10 (and 95 Percent
Confidence Interval): Results Based on Single-City versus Multi-City Models: Morbidity*
(Single-city models are always on the left, followed by the corresponding multi-city models) Source: Abt (2003b)
1. Samet et al. (2000)* - single-city model; cardiovasc. hosp. adms.
2. Samet et al. (2000)* - 14-city model; cardiovasc. hosp. adms.
3. Samet et al. (2000)* - single-city model; pneumonia hosp. adms.
4. Samet et al. (2000)* - 14-city model; pneumonia hosp. adms.
5. Samet et al. (2000)* - single-city model; COPD hosp. adms.
6. Samet et al. (2000)* - 14-city model; COPD hosp. adms.
*Reanalyzed in Zanobetti and Schwartz (2003)
*In cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-56
Draft - Do Not Quote or Cite
-------
I
2
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
0 CM
En
-a
£=
(0 0
1 "
0)
-a
'o
_E co
1
I—
'o
-i_. CO
£=
B
(D
^
CD f
ro
'V>
LU
1N •
HI
8 1
o
12345 | * 2 10
1
1
<
1
2 •
1
3
<
4
1
1
5 .
S
18
. _ ^
H
8
I1
'
|
1 2
III
S £
•
5
4 T
ills'
} 1
. ,
1
0
<
1
1
2
'
8
T
i
1 — r
«
4
<
y
2
0
8
2
1
1
1
1
I
22
i ¥
8
j'
Chicago
Detroit
Los Angeles
Minneapolis Phoenix Pittsburgh San Jose Seattle
Figure 4-14. Estimated Annual Percent of Mortality Associated with Short-Term Exposure to PM10 (and 95 Percent
Confidence Interval): Effect of Different Lag Models
Source: Abt (2003b)
Moolgavkar (2000a)* - single-city model; non-accid. mort.; 0-day lag
Moolgavkar (2000a)* - single-city model; non-accid. mort.; 1-day lag
Moolgavkar (2000a)* - single-city model; non-accid. mort.; 2-day lag
Moolgavkar (2000a)* - single-city model; non-accid. mort.; 3-day lag
Moolgavkar (2000a)* - single-city model; non-accid. mort.; 4-day lag
Moolgavkar (2000a)* - single-city model; non-accid. mort.; 5-day lag
Schwartz (2000b)** - 10-city model; all cause mort.; mean of 0- and 1-day lags
Schwartz (2000b)** - 10-city model; all cause mort.; distr. lag
Moolgavkar (2000a)* - single-city model; cardiovasc. mort.; 0-day lag
10. Moolgavkar (2000a)* - single-city model; cardiovasc. mort.; 1-day lag
11. Moolgavkar (2000a)* - single-city model; COPD mort.; 0-day lag
13. Braga et al. (2001)** - 10-city model; COPD mort.; mean of 0- and 1-day lags
14. Braga et al. (2001)** - 10-city model; COPD mort.; distr. lag
15. Braga et al. (2001)** - 10-city model; pneumonia mort.; mean of 0- and 1-day lags
16. Braga et al. (2001)** - 10-city model; pneumonia mort.; distr. lag
17. Braga et al. (2001)** - 10-city model; cardiovasc. mort.; mean of 0- and 1-day lags
18. Braga et al. (2001)** - 10-city model; cardiovasc. mort.; distr. lag
19. Maretal. (2000)***-cardiovasc. mort.; 0-day lag
20. Mar et al. (2000)*** - cardiovasc. mort.; 1-day lag
21. Fairley (1999)**** -non-accid. mort.; 0-day lag
22. Fairley (1999)**** -non-accid. mort.; 1-day lag
*Reanalyzed in Moolgavkar (2002); "Reanalyzed in Schwartz (2003a);
***Reanalyzed in Mar et al. (2003); ""Reanalyzed in Fairley (2003)
12. Moolgavkar (2000a)* - single-city model; COPD mort.; 1-day lag
*In cases where the lower bound of the confidence interval was less than zero, the interval does not represent a 95% confidence interval. In these cases, the upper bound is the
97.5th percentile and the lower bound has been truncated at zero.
August 2003
4-57
Draft - Do Not Quote or Cite
-------
1 In contrast to the mortality figure, there were somewhat larger differences between single- and
2 multi-city models for hospital admissions in four of the six risk assessment locations. Where there
3 were appreciable differences between single- and multi-city models for a given location, there was
4 no consistent pattern in terms of risk estimates being either smaller or larger based on the type of
5 model.
6 Figure 4-12 compares single- and multi-pollutant model risk estimates for PM10 for short-
7 term exposure mortality, hospital admissions, and respiratory symptom health endpoints. In most
8 cases, addition of a single pollutant (e.g., CO, O3, or NO2) made relatively little difference in the
9 PM10 risk estimates. The exception to this pattern is the set of risk estimates based on Moolgavkar
10 (2000c) (reanalyzed in Moolgavkar, 2003), which show much larger declines in PM10 risk
11 estimates in Los Angeles when CO or NO2 are added to the model.
12 Risk estimates associated with various single day lag models, as well as a distributed lag
13 model, associated with short-term exposure mortality for PM10 are shown in Figure 4-14. Risk
14 estimates associated with the distributed lag model are always larger than the corresponding
15 single-day lag based on the same study. In the two cities where 0- through 5-day lags were
16 examined, the 0- and 1-day lag were associated with the largest risk estimates in Chicago and the
17 3-day lag was associated with the largest risk estimate in Los Angeles.
18
19 4.3.2 Sensitivity Analyses
20 Several sensitivity analyses were carried out to provide some perspective on the impact of
21 various assumptions and uncertainties on the health risk estimates (see Table 4-9 for summary of
22 different types of sensitivity analyses). These sensitivity analyses were conducted in each of the
23 study areas and the complete results are in the technical support document (Abt, 2003b). The PM2 5
24 risk results for one study area (Detroit), are shown here for some of the sensitivity analyses for
25 illustrative purposes. Detroit has been selected because it provides an opportunity to examine both
26 mortality and morbidity risk estimates and includes both single and multi-pollutant C-R functions.
27 For purposes of informing decisions about the PM NAAQS, we are interested in PM-related
28 risks due to concentrations over background levels, where background includes PM from natural
29 sources and transport of PM from sources outside of North America. One set of sensitivity
30 analyses examined the impact of using the lower and upper end of the range of estimated
August 2003 4-58 Draft - Do Not Quote or Cite
-------
1 background concentrations cited in the draft CD. For Detroit, the use of alternative estimated PM2 5
2 background levels had no impact on the short-term exposure mortality or hospital admission risk
3 estimates because the LML for PM25 in Ito (2003) [reanalysis of Lippmann et al. (2000)] was 6
4 M-g/m3, which is larger than the range of background levels considered in the sensitivity analysis
5 (i.e., 2 to 5 |ig/m3). In the other eight PM25 locations, using the upper and lower end of the range of
6 estimated background generally had a small to modest impact, on the order of+/- 10-20% change
7 in short-term exposure health endpoint risk estimates compared to use of the midpoint of the
8 estimated range of background levels in the base case estimates. Alternative estimated PM25
9 background levels had no impact on long-term exposure mortality in Detroit, or any of the other
10 PM25 locations, because the LML in the long-term studies was 10 or 11 i-ig/m3, which is larger than
11 the range of estimated PM2 5 background levels.
12 A sensitivity analysis also has been conducted that focuses on the impact of using a varying
13 estimated PM25 background concentration instead of the fixed level used in each study area in the
14 base case assessment. While there are no empirical data that provide the distribution of
15 background PM25 concentrations, there is information available from a rural monitoring network
16 (IMPROVE) that provides the best available information on the potential variation in background
17 concentrations. Based on an examination of distributions of ambient PM2 5 24-hour data at
18 IMPROVE sites (Langstaff, 2003b), which are mostly located in rural environments near Class I
19 areas, two lognormal distributions were estimated, one for the East (which was applied to Boston,
20 Philadelphia, St. Louis, and Detroit) and one for the West (which was applied to San Jose, Los
21 Angeles, Phoenix, and Seattle). It is important to recognize that all IMPROVE sites measure some
22 PM2 5 from anthropogenic sources, and this typically will inflate the standard deviation over what it
23 would be if we were able to measure background concentrations as defined for purposes of this
24 assessment. Because of this, we selected the smallest standard deviation among all IMPROVE
25 sites in the East and in the West. The resulting standard deviation for the East is 3.8 and for the
26 West is 1.5. These are not substantially smaller than the next two smallest standard deviations (3.8
27 and 4.0 in the East, and 1.6 and 1.6 in the West). The mean value for the Eastern distribution is 3.5
28 M-g/m3 and for the Western distribution is 2.5 |ig/m3 which are based on the midpoint of the
29 estimated PM25 background concentrations in the draft CD (p. 3-104). It should be noted that the
30 estimated distributions for background may not fully reflect peak 24-h average natural background
August 2003 4-59 Draft - Do Not Quote or Cite
-------
1 concentrations which can be substantially higher than the annual or seasonal average natural
2 background concentrations within areas affected by wildfires and dust storms and long range
3 transport from outside North America (CD, p. 3-103). While the current PM2 5 risk assessment,
4 therefore, does not capture these unusual events, it should be noted that there are regulatory
5 provisions to exclude such events for purposes of judging whether an area is meeting the current
6 NAAQS (as noted above in section 2.7).
7 Background levels are likely to be positively correlated with "as is" levels in a given
8 location. To reflect this likelihood, we assigned a background value to each day that is the same
9 percentile in the distribution of daily background levels as the corresponding "as is" level is in its
10 distribution. Using this approach, the differences in risk estimates of using a constant background
11 versus a varying daily background are so small that they do not show up after rounding.
12 For PM10_2 5, the sensitivity analysis examining the effects of using the lower and upper end
13 of the range of estimated background levels showed no difference in any of the risk estimates for
14 short-term exposure mortality or hospital admissions in Detroit between the base case (which used
15 a value of 3 |ig/m3 for background) and the lower end where background was estimated to be 0
16 M-g/rn3. Again there was no effect due to different background levels because the LML for the
17 study providing the C-R relationship (i.e., 5 jig/m3) was above the estimated background levels. At
18 the upper end, where background was estimated to be 9 i-ig/m3, the short-term exposure mortality
19 and hospital admission risk estimates were reduced by about 60-70% (see Exhibit 7.14 in Abt
20 (2003b). The effect of different background concentrations for the other PM10_2 5 locations can be
21 found in Exhibits D.84 and D.86 through D.89 in Abt (2003b). In addition, the effect of different
22 background concentrations on PM10 risk estimates for the 12 PM10 locations can be found in
23 Appendix D (Abt, 2003b).
24 One of the most significant uncertainties continues to be the issue of hypothetical thresholds
25 below which there may be no PM2 5 health effects. As discussed above in section 4.2.6.1, there is
26 very limited evidence addressing whether or not thresholds exist for PM2 5, with most analyses
27 failing to find evidence that thresholds exist. As a sensitivity analysis, three hypothetical
28 thresholds or cutpoints (10, 15, and 20 |ig/m3) are used to examine the potential impact on risk
29 estimates for short-term exposure mortality and three different hypothetical thresholds or cutpoints
30 (10, 12.5, and 15) are used to examine the potential impact on risk estimates for long-term
August 2003 4-60 Draft - Do Not Quote or Cite
-------
1 exposure mortality. In conjunction with defining such cutpoints for these sensitivity analyses, the
2 slopes of the C-R functions have been increased to reflect the effect of hypothetical thresholds at
3 the selected levels. A simple slope adjustment method has been used that assumes the slope for the
4 upward-sloping portion of a hockeystick would be approximately a weighted average of the two
5 slopes of a hockeystick - namely, zero and the slope of the upward-sloping portion of the
6 hockeystick (see the technical support document (Abt, 2003b) for additional details). If the data
7 used in a study do not extend down below the cutpoint or extend only slightly below the cutpoint,
8 then the extent of the downward bias of the reported PM coefficient will be minimal or non-
9 existent. This is the case, for example, when the cutpoint is 10 |ig/m3 or 12.5 |ig/m3 for long-term
10 exposure mortality, given that the LMLs in the long-term exposure mortality studies were 10 |ig/m3
11 and 11 i-ig/m3. The slope adjustment methods used in these sensitivity analyses are intended only to
12 be illustrative of the possible impact on the risk estimates of alternative hypothetical thresholds. A
13 more thorough evaluation of the effect of possible thresholds would require re-analysis of the
14 original health and air quality data.
15 Table 4-10 shows the results of these sensitivity analyses examining the impact of
16 hypothetical thresholds for short- and long-term exposure mortality risk estimates for the "as is"
17 PM25 levels in Detroit. From Table 4-10 it is apparent that the short-term exposure risk estimates
18 are particularly sensitive to the application of hypothetical thresholds. For short-term exposure
19 mortality, a hypothetical threshold of 10 |ig/m3 reduces the percent of total non-accidental mortality
20 incidence associated with PM2 5 by about 30% and the highest hypothetical threshold (20 |ig/m3)
21 reduces it by about 70%. For long-term exposure mortality, a hypothetical threshold of 10 |ig/m3
22 has no impact on the risk estimates, because the LML for the epidemiological studies providing the
23 C-R relationships in Detroit is 10 |ig/m3. A hypothetical threshold of 12.5 |ig/m3 reduces the
24 percent of total mortality by about 37 to 43% depending on the study and model (i.e., single vs.
25 multi-pollutant) used. A hypothetical threshold of 15 |ig/m3 reduces the base case long-term
26 exposure mortality incidence associated with PM2 5 by about 81 to 86%.
27 Another sensitivity analysis illustrates how different the risk estimates would be if the C-R
28 functions used for short-term exposure mortality had used distributed lag models instead of single
29 lag models. Schwartz (2000a) has shown in a study of short-term exposure mortality in 10 cities
30 using PM10 as the indicator that a distributed lag model predicted the same relative risk that a single
August 2003 4-61 Draft - Do Not Quote or Cite
-------
1 lag model would have predicted if the coefficient was approximately two times what it was
2 estimated to be. To simulate the possible impact of using a distributed lag model, the PM25
3 coefficients were multiplied by two. As would be expected, the risk estimates are almost doubled
4 using the distributed lag approximation (see Abt, 2003b; Appendix D).
5 The influence of using different periods of exposure on the risks estimated in long-term
6 exposure mortality studies also has been examined in a sensitivity analysis. Two alternatives were
7 examined: assuming the relevant PM2 5 ambient concentrations were respectively 50% higher than
8 and twice as high as the PM2 5 ambient concentrations used in the original epidemiological study.
9 The impact of these varying assumptions about the role of historical air quality on estimates of
10 long-term exposure mortality associated with "as is" PM2 5 concentrations is shown for Detroit in
11 Table 4-11. Assuming that PM2 5 concentrations were 50% higher than and twice as high as that in
12 the original studies reduces long-term exposure mortality risk estimates by about one-third and
13 one-half, respectively.
August 2003 4-62 Draft - Do Not Quote or Cite
-------
Table 4-10. Sensitivity Analysis: Estimated Annual Health risks of Short-Term and Long-Term Exposure Mortality Associated
with "As Is" PM2 5 Concentrations Using alternative Hypothetical Threshold Models, Detroit, MI, 2000*
Health
Effects **
Short-
Term
Exposure
Mortality
Long-Term
Exposure
Mortality
Study
Single Pollutant Models:
Ito (2003) [reanalysis of
Lippmann et al. (2000)]
Single Pollutants Models:
Krewski et al. (2000) - ACS
Pope et al. (2002) - ACS
extended
Multi-Pollutant Models
Krewski et al. (2000) - ACS
Krewski et al. (2000) - ACS
Krewski et al. (2000) - ACS
Krewski et al. (2000) - ACS
Type
Non-
accidental
All cause
All cause
All cause
All cause
All cause
All cause
Ages
all
30+
30+
30+
30+
30+
30+
Lag
3 day
Other
Pollutants
in
Model
CO
NO2
O3
S02
Percent of total Incidence Associated with PM25 Above Hypothetical Threshold ***
BASE CASE:
Background
= 3.5 |-igm3
0.7%
(0.0% - 2.2%)
BASE CASE:
Background
= 3.5 |-igm3
2.7%
(1.4% -4.1%)
2.3%
(0.5% -4.1%)
3.9%
(2.2% -5. 5%)
4.6%
(2.5% - 6.6%)
3.9%
(2.2% -5. 5%)
0.7%
(0.0%) - 2.9%)
Hypothetical
Threshold
= 10 |-igm3
0.5%
(0.0% -1.5%)
Hypothetical
Threshold
= 10 |-igm3
2.7%
(1.4% -4.1%)
2.3%
(0.5% -4.1%)
3.9%
(2.2% -5. 5%)
4.6%
(2.5% - 6.6%)
3.9%
(2.2% -5. 5%)
0.7%
(0.0% - 2.9%)
Hypothetical
Threshold
= 15 |-igm3
0.3%
(0.0% - 0.9%)
Hypothetical
Threshold
= 12.5 |-igm3
1.7%
(0.9% - 2.6%)
1.5%
(0.3% - 2.7%)
2.4%
(1.4% -3. 5%)
2.9%
(1.5% -4.2%)
2.4%
(1.4% -3.5%)
0.4%
(0.0% -1.8%)
Hypothetical
Threshold
= 20 |-igm3
0.2%
(0.0% -0.5%)
Hypothetical
Threshold
= 15 |-igm3
0.5%
(0.2% - 0.7%)
0.4%
(0.1% -0.8%)
0.7%
(0.4% -1.0%)
0.8%
(0.4% -1.2%)
0.7%
(0.4% -1.0%)
0.1%
(0.0% -0.5%)
* This sensitivity analysis was performed only for those studies which reported highest measured levels in the study. See text for an explanation of the slope adjustment method.
** Health effects are associated with short-term exposure to PM25 unless otherwise specified.
*** Health effects incidence was quantified across the range of PM concentrations observed in each study, when possible, but not below background level. Average background
PM25 is taken to be 3.5 |_ig/m3 in the East and 2.5 |_ig/m3 in the West. Estimates less than zero were truncated at zero. Incidences are rounded to the nearest 10; percentages are
rounded to the nearest tenth.
Note: Numbers in parentheses are 95% confidence intervals based on statistical uncertainty surrounding the PM2 5 coefficient.
August 2003
4-63
Draft - Do Not Quote or Cite
-------
1 4.3.3 Key Observations
2 Sections 4.3.1 and 4.3.2 have summarized the PM health risk estimates and sensitivity
3 analyses associated with "as is" PM air quality levels. Summarized below are key observations
4 resulting from this part of the risk assessment:
5 • A fairly wide range of risk estimates are observed for PM-related morbidity and mortality
6 health effects across the urban areas analyzed associated with "as is" air quality for the
7 three PM indicators (PM2 5, PM10_2 5, and PM10).
8
9 • Most of the mean effect estimates for PM2 5 for the base case analysis are in the range 0.5 to
10 1.5% for short-term exposure total non-accidental mortality. Generally, the mean risk
11 estimates for the single- and multi-city models are roughly comparable in most of the urban
12 areas analyzed. The impact of adding additional co-pollutants to the models was variable;
13 sometimes there was relatively little difference, while in other cases there were larger
14 differences.
15
16 • For long-term exposure mortality associated with PM25, the mean effect estimates range
17 from about 0.2% to as high as 4 to 5% with most estimates falling in the 1.5 to 2% range for
18 single-pollutant models. Addition of a single co-pollutant resulted in higher risk estimates
19 when CO, NO2, or O3 were added to the models for the ACS study and lower risk estimates
20 when SO2 was added.
21
22 • There is significant variability in the mortality effect estimates associated with PM10_2 5, with
23 three of the four urban areas showing mean effect estimates ranging from about 0.3 to over
24 6% of total incidence and a fourth area reporting no mean effect on mortality, but about 4 to
25 nearly 6% of respiratory symptoms in this same area associated with PM10_2 5
26
27 • There is no apparent pattern in the PM10 mortality risk estimates related to the relative ratio
28 of fine- and coarse-fraction PM across the 12 urban areas analyzed. Most, but not all, of
29 the mean risk estimates for total (non-accidental) mortality associated with short-term
30 exposure to PM10 in these 12 urban areas associated with a recent year of air quality range
31 from about 0.5 to 2.5% of total incidence.
32
33
August 2003 4-64 Draft - Do Not Quote or Cite
-------
Table 4-11. Sensitivity Analysis: The Effect of Assumptions About Historical Air Quality on Estimates of Long-Term Exposure
Mortality Associated with "As Is" PM2 5 Concentrations, Detroit, MI, 2000
Health Effects
Long-Term
Exposure Mortality
Study
Single Pollutant Models
Krewski et al. (2000) -
ACS
Krewski et al. (2000) -
ACS
Pope et al. (2002)
Pope et al. (2002)
Pope et al. (2002)
Multi-Pollutants Models
Krewski et al. (2000) -
ACS
Krewski et al. (2000) -
ACS
Krewski et al. (2000) -
ACS
Krewski et al. (2000) -
ACS
Type
All cause
Cardiopulmonary
All cause
Cardiopulmonary
Lung Cancer
All cause
All cause
All cause
All cause
Ages
30+
30+
30+
30+
30+
30+
30+
30+
30+
Other Pollutants in
Model
CO
N02
03
SO2
Percent of Total Incidence
Base
Case:Assuming
AQ as Reported
2.7%
(1.4% -4.1%)
5.3%
(3. 5% -7.4%)
2.3%
(0.5% -4.1%)
3.3%
(0.9% -5. 7%)
4.5%
(0.6% - 8.2%)
3.9%
(2.2% -5. 5%)
4.6%
(2.5% -4.4%)
3.9%
(2.2% -5. 5%)
0.7%
(0.0% - 2.9%)
Asuming relevant
AQ 50% higher
1.8%
(0.9% - 2.7%)
3.6%
(2. 3% -5.0%)
1.6%
(0.3% - 2.8%)
2.2%
(0.6% -3. 8%)
3.0%
(0.4% -5. 5%)
2.6%
(1.5% -3.7%)
3.1%
(1.6% -4.4%)
2.6%
(1.5% -3.7%)
0.5%
(0.0% -1.9%)
Assuming relevant
AQ twice as high
1.3%
(0.7% -2.1%)
2.7%
(1.8% -3. 8%)
1.2%
(0.2% -2.1%)
1.7%
(0.4% -2. 9%)
2.3%
(0.3% - 4.2%)
2.0%
(1.1% -2. 8%)
2.3%
(1.2% -3.3%)
2.0%
(1.1% -2. 8%)
0.4%
(0.0% -1.4%)
* Health effects incidence was quantified across the range of PM concentration observed in each study, when posible, but not below background level. Average background PM2 5 is
taken to be 3.5 ug/m3 in the East and 2.5 ug/m3 in the West. Estimates less than zero were truncated at zero. Incidences are rounded to the nearest 10; percents are rounded to the
nearest tenth.
Note: Numbers in parentheses are 95% confidence intervals based on statistical uncertainty surrounding the PM2.5 coefficient.
August 2003
4-65
Draft - Do Not Quote or Cite
-------
1 The wide variability in risk estimates associated with a recent year of air quality for the
2 three different PM indicators is to be expected given the wide range of PM levels across the urban
3 areas analyzed and the variation observed in the C-R relationships obtained from the orginal
4 epidemiology studies. Among other factors, this variability may reflect differences in populations,
5 exposure considerations (e.g., degree of air conditioning use), differences in co-pollutants and/or
6 other stressors, differences in study design, and differences related to exposure and monitor
7 measurement error.
8 There are considerably fewer PM10_2 5 risk estimates available for the five urban areas
9 examined due to the much smaller number of epidemiology studies examining health effects
10 associated with coarse fraction particles. The confidence intervals for the two locations with the
11 highest effect estimates (i.e., Phoenix and Philadelphia), were fairly wide.
12
13 • Based on the results from the sensitivity analyses, the single most important factor
14 influencing the risk estimates is whether or not a hypothetical threshold exists below which
15 PM-related health effects do not occur.
16
17 • Based on the results from the sensitivity analyses, the following uncertainties have a
18 moderate impact on the risk estimates in some or all of the cities: choice of an alternative
19 estimated background level, use of a distributed lag model, and alternative assumptions
20 about the relevant air quality for long-term exposure mortality. Use of a distribution of
21 daily background concentrations had very little impact on the risk estimates.
22
23 During the previous review of the PM NAAQS, EPA provided an illustrative example
24 based on the PM health risk assessment that showed the distribution of mortality risk associated
25 with short-term exposure over a 1-year period. EPA concluded that peak 24-hour PM2 5
26 concentrations appeared "to contribute a relatively small amount to the total health risk posed by an
27 entire air quality distribution as compared to the risks associated with low to mid-range
28 concentrations" (61 FR at 65652, December 13, 1996). Figure 4-15 (a,b) for PM2 5 in St. Louis and
29 Figure 4-16 (a,b) for PM10_2 5 in Phoenix provide examples of annual distributions of 24-hour PM
30 concentrations and the corresponding distribution of estimated mortality incidence based on two
31 short-term exposure epidemiology studies included in the current PM risk assessment.22
22The St. Louis PM2 5 example uses the C-R function for non-accidental mortality from Schwartz et al.
(1996), reanalyzed in Schwartz (2003). The Phoenix PM10_2 5 example uses the C-R function for cardiovascular
mortality from Mar et al. (2000), reanalyzed in Mar et al. (2003).
August 2003 4-66 Draft - Do Not Quote or Cite
-------
2 • Consistent with the observation made in the previous PM NAAQS review, for both PM2 5
3 and PM10_2 5 the peak 24-hour concentrations contribute a relatively small amount to the
4 total health risk associated with short-term exposures on an annual basis.
5
6
7 4.4 RISK ESTIMATES ASSOCIATED WITH JUST MEETING THE CURRENT PM2 5
8 STANDARDS
9
10 4.4.1 Summary of Risk Estimates
11 The second part of the PM2 5 risk assessment estimates the risk reductions that would result
12 if the current annual PM25 standard of 15 |ig/m3 and the current daily PM25 standard of 65 |ig/m3
13 were just met in the assessment locations. This part of the risk assessment only considers those
14 locations that do not meet the current standards based on 2002 data (i.e., Detroit, Philadelphia,
15 Pittsburgh, Los Angeles, and St. Louis). Again, this section occasionally uses Detroit for
16 illustrative purposes, however, the remaining risk estimates can be found in the technical support
17 document (Abt Associates, 2003b). As noted previously, the 15 |ig/m3 annual average standard is
18 the controlling standard in all five study areas, consequently, just meeting this standard also results
19 in each of these areas meeting the 65 i-ig/m3, 24-hour standard.
20 The percent rollback necessary to just meet the annual standards depends on whether the
21 maximum or the spatial average of the monitor-specific annual averages is used. For the current
22 risk assessment, the approach used to simulate just meeting the current annual average standard
23 used the maximum of the monitor-specific annual averages as shown in Table 4-12. Since an area
24 could potentially use the spatial average of the population-oriented monitors to determine whether
25 or not it met the annual average standard, Table 4-12 also presents the percent rollbacks that would
26 have resulted from using this alternative approach in each urban study area.
August 2003 4-67 Draft - Do Not Quote or Cite
-------
10
15
25 30
24-Hr PM -2.5
Concentration (ug/m3)
35
40
45
50
55
Figure 4-15a. Distribution of 24-Hour PM25 Concentrations in St. Louis (2002 Air Quality
Data)
«—Mean Estimate of# of Deaths
• 2.5th% Estimate of# of Deaths
A 97.5th% Estimate of# of Deaths
20 25 30 35 40
24-Hr PM-2.5 Concentration (ug/m3)
Figure 4-15b. Estimated Non-Accidental Mortality in St. Louis Associated with PM2 5
Concentrations (2002 Air Quality Data)
August 2003
4-68
Draft - Do Not Quote or Cite
-------
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160
24-Hr PM10-2.5
Concentration (ug/m3)
Figure 4-16a. Distribution of 24-Hour PM10_25 Concentrations in Phoenix (1997 Air Quality Data)
130
120 -
-*— Mean Estimate of # of Deaths
• 2.5th% Estimate of # of Deaths
A 97.5th% Estimate of # of Deaths
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160
24-Hr PM10-2.5 Concentration (ug/m3)
Figure 4-16b. Estimated Cardiovascular Mortality in Phoenix Associated with PM10_2 5
Concentrations (1997 Air Quality Data)
August 2003
4-69
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
Table 4-12. Air Quality Adjustments Required to Just Meet the Current Annual PM25
Standard of 15 fig/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 Annual PM2 5
Standard
Using Maximum of Monitor-
Specific Annual Averages
29.5%
41.9%
9.5%
8.7%
3.4%
Using Average of Monitor-
Specific Annual Averages
10.9%
30.2%
0.0%
0.0%
0.0%
10
11
12
13
14
15
16
Source: Abt (2003b)
August 2003
4-70
Draft - Do Not Quote or Cite
-------
1 To gain some perspective on how the risk reductions compare across the four study areas,
2 Table 4-13 shows the reduction in short- and long-term mortality incidence associated with going
3 from "as is" PM25 air quality (based on 2002 air quality data) to just meeting the current annual
4 PM2 5 standard. For short-term exposure mortality, single-pollutant, non-accidental mortality
5 estimates are selected since they are available across the five study areas. For long-term exposure
6 mortality, the ACS single-pollutant model reanalysis estimates for total (all cause) mortality are
7 selected for comparison. Risk reductions for total mortality related to PM2 5 long-term exposure
8 associated with the Six City reanalysis are roughly two to three times larger than the ACS
9 reanalysis estimates and risk reductions associated with the ACS extended study (Pope et al., 2002)
10 are slightly smaller than the ACS reanalysis estimates shown in Table 4-14. In Table 4-13 risk
11 reductions are expressed both as a percentage reduction in the PM2 5-associated mortality and as a
12 percentage of the total mortality due to PM25 and other causes. As expected the largest reduction
13 (250 deaths representing a 45.9% reduction in PM2 5-related incidence and a 0.4% reduction in total
14 mortality ) for short-term exposure mortality is observed for Los Angeles which has the highest
15 PM25 ambient levels and largest population. The smallest reduction for short-term exposure
16 mortality is observed for St. Louis (10 deaths representing a 3.4% reduction in PM25-related
17 incidence and a 0.0% reduction (taking into account rounding) in total mortality).
18
19 4.4.2 Sensitivity Analyses
20 The base case risk assessment uses a proportional rollback approach to adjust air quality
21 distributions to simulate the pattern that would occur in an area improving its air quality so that it
22 just meets the current annual average PM2 5 standard. The support for this approach is briefly
23 discussed in section 4.2.3 and in more detail in Appendix B of the technical support document
24 (Abt, 2003b). While the available data suggest that this is a reasonable approach, other patterns of
25 change are possible. In a sensitivity analysis an alternative air quality adjustment approach is used
26 which reduces the top 10 percent of the distribution of PM25 concentrations by 1.6 times as much
27 as the lower 90 percent of concentrations. The result of this alternative hypothetical adjustment
28 which reduces the highest days more than the rest of the distribution showed only a
29
August 2003 4-71 Draft - Do Not Quote or Cite
-------
Table 4-13. Comparison of Annual Estimates of Short- and Long-Term Exposure Mortality Reductions Associated with Just
Meeting the Current PM25 Standards*
Health Effect and Model
non-accidental mortality, single pollutant,
s.t. exposure, 3 day lag
non-accidental mortality, single pollutant,
s.t exposure, 1 day lag
non-accidental mortality, single pollutant,
s.t. exposure, 0 day lag***
non-accidental mortality (age > 75 ),
single pollutant, s.t. exposure, 0 day lag
non-accidental mortality, single pollutant,
s.t. exposure, lag 0/lday
total mortality (age > 30)**
long-term exposure
total mortality (age > 30)**
long-term exposure
total mortality (age > 30)**
long-term exposure
total mortality (age > 30)**
long-term exposure
total mortality (age > 30)**
long-term exposure
Urban Study
Area
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
Detroit
Los Angeles
Philadelphia
Pittsburgh
St. Louis
"As Is"
Incidence
130
(0 - 400)
550
(0 - 1060)
340
60
(0 - 250)
260
(100 - 420)
500
(260 - 780)
2730
(1420-4140)
360
(190 - 560)
350
(180 - 540)
500
(260 -770 )
Reduction in
Incidence
50
(0 - 140)
250
(0 - 500)
30
10
(0 - 20)
10
(0 -10 )
310
(160 - 480)
1980
(1030-3010)
80
(40 - 130)
70
(40-110)
40
(20 - 60)
Percent of PM25-Related
Incidence Reduced
36.2%
(n/a -36.5%)
45.9%
(45.7 - 46%)
9.5%
8.8%
(n/a - 8.9%)
3.4%
(3.4-3.4%)
62.6%
(62.5 - 62.8%)
72.5%
(72.2 - 72.7%)
23.3%
(23.2-23.4%)
19.8%
(19.8 - 19.9%)
8.1%
(8.1-8.1%)
Reduction in Incidence
Expressed as Percent of
Total Incidence
0.3%
(0.0 - 0.8%)
0.4%
(0.0 - 0.9%)
0.2%
0.1%
(0.0 - 0.3%)
0%
(0-0.1%)
1.7%
(0.9 - 2.5%)
3.4%
(1.8-5.2%)
0.5%
(0.3 -0.7%)
0.5%
(0.2 - 0.7%)
0.2%
(0.1-0.3%)
*Risk reductions are relative to the "as is" (year 2002) air quality base case risk estimates.
**These risk reductions are based on the Krewski et al. (2000) ACS reanalysis, single-pollutant model.
n/a = Incidence reduction as a percent of PM-related incidence was calculated by dividing the incidence reduction achieved by rolling back PM2 5 above background
to just meet the current standards by the incidence associated with "as is" PM2 5 above background. In those cases in which the incidence associated with "as is"
PM2 5 above background was estimated to be zero, this percent could therefore not be calculated.
***The Lipfert et al. (2000) study does not provide the statistical uncertainties surrounding the PM2 5 coefficients and, therefore, the 95% confidence intervals could
not be calculated.
August 2003
4-72
Draft - Do Not Quote or Cite
-------
1 small difference (less than 1%) in the percent change in PM-associated incidence (see Exhibit 7.8
2 and Appendix D in Abt, 2003b).
3
4 4.4.3 Key Observations
5 Sections 4.4.1 and 4.4.2 have summarized the PM health risk estimates and sensitivity
6 analyses associated with just meeting the current PM25 standards. Summarized below are key
7 observations resulting from this part of the risk assessment:
8 • There is a wide range of reductions in PM25-related incidence across the five urban areas
9 analyzed which is largely due to the varying amount of reduction in ambient PM25
10 concentrations required in these urban areas to just meet the current PM2 5 standard. For
11 example, using single-pollutant models the percent of PM25-related incidence reduced for
12 short-term, non-accidental mortality ranges from about 45% in Los Angeles to about 3% in
13 St. Louis. Similarly, using the ACS study the percent of PM25-related incidence reduced
14 for long-term exposure mortality ranges from roughly 70% in Los Angeles to about 8% in
15 St. Louis.
16
17 • The risk estimates associated with just meeting the current PM2 5 standards incorporate
18 several additional sources of uncertainty, including: (1) uncertainty in the pattern of air
19 quality concentration reductions that would be observed across the distribution of PM
20 concentrations in areas attaining the standards ("rollback uncertainty") and (2) uncertainty
21 concerning the degree to which current PM risk coefficients may reflect contributions from
22 other pollutants, or the particular contribution of certain constituents of PM2 5, and whether
23 such constituents would be reduced in similar proportion to the reduction in PM2 5 as a
24 whole. At least one alternative approach to rolling back the distribution of daily PM2 5
25 concentrations, in which the upper end of the distribution of concentrations was reduced by
26 a greater amount than the rest of the distribution, had little impact on the risk estimates.
27
28 4.5 RISK ESTIMATES ASSOCIATED WITH JUST MEETING ALTERNATIVE PM2 5
29 AND PM10 2 5 STANDARDS
30
31 In the next draft of the PM Staff Paper, staff plans to include risk estimates associated with
32 just meeting alternative PM2 5 and alternative PM10_2 5 standards. The standards analyzed will be
33 based on the range of standards recommended for consideration in Chapter 6 of this draft Staff
34 Paper, taking into account CASAC and public comments on the proposed ranges received as part of
35 the review of this Staff Paper.
August 2003 4-73 Draft - Do Not Quote or Cite
-------
1 REFERENCES
2 Most Chapter 4 references are available at the end of Chapter 3. References not listed at the end
3 of Chapter 3 are listed here.
4
5 Abt Associates Inc. July 3, 1996 (Revised November 1996). "A Paniculate Matter Risk Assessment for Philadelphia
6 and Los Angeles." Prepared for the Office of Air Quality Planning and Standards, U. S. Environmental
7 Protection Agency, Contract No. 68-W4-0029. Available electronically on the web at:
8 http://www.epa.gOv/ttn/naaqs/standards/pm/s pm pr td.html
9
10 Abt Associates Inc. 1997a. Revisionof Mortality Incidence Estimates Based on Pope etal. (1995) in the Abt
11 Paniculate Matter Risk Assessment Report. Memorandum from Ellen Post and John Voyzey, Abt Associates
12 Inc. to John Bachmann, Allyson Siwik, Michele McKeever, and Harvey Richmond, U.S. EPA/OAQPS. June
13 5, 1997.
14
15 Abt Associates Inc. 1997b. Revision of Mortality Incidence Estimates Based on Pope et al. (1995) in the December
16 1996 Supplement to the Abt Paniculate Matter Risk Assessment Report. Memorandum from Ellen Post, Abt
17 Associates, Inc. to John Bachmann, Allyson Siwik, Michele McKeever, and Harvey Richmond, U.S.
18 EPA/OAQPS. June 6.
19
20 Abt Associates Inc. 2002. Proposed Methodology for Paniculate Matter Risk Analyses for Selected Urban Areas:
21 Draft Report. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental Protection
22 Agency, Contract No. 68-D-03-002. Available electronically on the web at:
23 http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html
24
25 Abt Associates Inc. 2003a. Preliminary Recommended Methodology for PM10 and PM10.25 Risk Analyses in Light of
26 Reanalyzed Study Results." Draft Memorandum from Ellen Post, Abt Associates, Inc. to Harvey Richmond,
27 U.S. EPA/OAQPS. Aprils. Available electronically on the web at:
28 http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html
29
30
31 Abt Associates Inc. 2003b. Paniculate Matter Health Risk Assessment for Selected Urban Areas: Draft Report.
32 Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency,
33 Contract No. 68-D-03-002. Available electronically on the web at:
34 http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html
35
36 Deck, L. B.; Post, E.S.; Smith, E.; Wiener, M.; Cunningham, K.; Richmond, H. (2001). Estimates of the health risk
3 7 reductions associated with attainment of alternative paniculate matter standards in two U.S. cities. Risk Anal.
38 21(5): 821-835.
39
40 EPA, 200 Ic. Paniculate Matter NAAQS Risk Analysis Scoping Plan. Draft. Office of Air Quality Planning and
41 Standards, U.S. Environmental Protection Agency. Available electronically on the web at:
42 http://www.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html
43
44 Hopke, 2002. Letter from Dr. Phil Hopke, Chair, Clean Air Scientific Advisory Committee (CASAC) to Honorable
45 Christine Todd Whitman, Administrator, U.S. EPA. Final advisory review report by the CASAC Particulate
46 Matter Review Panel on the proposed paniculate matter risk assessment. May 23. Available electronically on
47 the web at: http://www.epa.gov/sab/pdf/casacadv02002.pdf
48
49 Langstaff, J. (2003). OAQPS Staff Memorandum to PM NAAQS Review Docket (OAR-2001-0017). Subject:
50 Variability of PM25 Background Concentrations. August 15, 2003.
51
52 Post, E.; Deck, L.; Larntz, K.; Hoaglin. D. (2001). An application of an empirical Bayes estimation technique to the
53 estimation of mortality related to short-term exposure to paniculate matter. Risk Anal. 21(5): 837-842.
August 2003 4-74 Draft - Do Not Quote or Cite
-------
1 5. CHARACTERIZATION OF PM-RELATED WELFARE EFFECTS
2
3 5.1 INTRODUCTION
4 This chapter summarizes key information relevant to assessing the welfare effects
5 associated with ambient PM, alone and in combination with other pollutants commonly present
6 in the ambient air, drawing upon the most relevant information contained in the draft CD and
7 other significant reports referenced therein. The chapter is organized into a discussion of the
8 effects on public welfare to be considered in this review of the secondary standards for PM.
9 Specifically, this chapter addresses PM-related effects on visibility (section 5.2), materials
10 (section 5.3), vegetation and ecosystems (section 5.4), and climate change and solar radiation
11 (section 5.5). For each category of PM-related effects, this chapter presents a brief summary of
12 the relevant scientific information and a staff assessment of whether the available information is
13 sufficient to be considered as the basis for secondary standards distinct from primary standards
14 for PM. In addition, in assessing information on PM-related effects on climate change and solar
15 radiation, consideration is given to potential indirect impacts on human health and the
16 environment that may be a consequence of radiative and climatic changes attributable to changes
17 in ambient PM. Preliminary staff conclusions and recommendations related to secondary
18 standards for PM are presented in Chapter 6.
19 It is important to note that the discussion of PM-related effects on visibility, vegetation
20 and ecosystems, and climate change and solar radiation in Chapter 4 of the draft CD builds upon
21 and includes by reference extensive information from several other significant reviews of these
22 areas. Most notably, these reports include the Recommendations of the Grand Canyon Visibility
23 Transport Commission (1996), the National Research Council's Protecting Visibility in National
24 Parks and Wilderness Areas (1993), reports of the National Acid Precipitation Assessment
25 Program (1991, 1998), previous EPA Criteria Documents, including Air Quality Criteria for
26 Particulate Matter and Sulfur Oxides (EPA, 1982) and Air Quality Criteria for Oxides of
27 Nitrogen (EPA, 1993), recent reports of the National Academy of Sciences (NAS, 2001) and the
28 Intergovernmental Panel on Climate Change (IPCC, 1998, 2001a,b), and numerous other U.S.
29 and international assessments of stratospheric ozone depletion and global climate change carried
August 2003 5-1 Draft - Do Not Quote or Cite
-------
1 out under U.S. Federal interagency programs (e.g., the U.S. Global Climate Change Research
2 Program), the World Meteorological Organization (WMO), and the United Nations Environment
3 Programme (UNEP).
4
5 5.2 EFFECTS ON VISIBILITY
6 Visibility impairment is caused by the scattering and absorption of light by particles and
7 gases in the atmosphere. It is the most noticeable effect of fine particles present in the
8 atmosphere. Air pollution degrades the visual appearance and perceived color of distant objects
9 to an observer and reduces the range at which they can be distinguished from the background.
10 The effect of ambient particles on visibility is dependant upon particle size and composition,
11 atmospheric illumination, the optical properties of the atmosphere, and the optical properties of
12 the target being viewed.
13 This section discusses the role of ambient PM in the impairment of visibility, building
14 upon the information presented in the last Staff Paper (EPA, 1996b) and drawing upon the most
15 relevant information contained in the draft CD and significant reports on the science of visibility
16 referenced therein. In particular, this section includes new information on the following topics:
17
18 • Data analyses for a number of cities in which 1999 daily PM25 measurements are
19 correlated with visibility data from Automated Surface Observation System (ASOS)
20 installations.
21
22 • An overview of existing and planned visibility programs, goals, and methods for the
23 evaluation of visibility impairment as a basis for standard setting, in the U.S. and abroad,
24 illustrating the significant value placed on efforts to improve visibility outside of national
25 parks and wilderness areas.
26
27 The presentation here organizes the available information on visibility impairment into
28 elements related to the evaluation of current and alternative standards for PM. Beyond providing
29 an overview of visibility impairment, this section summarizes: (1) the effects of PM on visibility
30 (building upon information presented above in Section 2.9); (2) conditions in Class I and non-
31 urban areas, as well as in urban areas; (3) information on the significance of visibility to public
32 welfare; and (4) approaches to evaluating public perceptions of visibility impairment and
33 judgments about the acceptability of varying degrees of impairment.
August 2003 5-2 Draft - Do Not Quote or Cite
-------
1 5.2.1 Overview of Visibility Impairment
2 Visibility can be defined as the degree to which the atmosphere is transparent to visible
3 light (NRC, 1993; CD, 4-153). Visibility effects are manifested in two principal ways: (1) as
4 local impairment (e.g., localized hazes and plumes); and (2) as regional haze. This distinction is
5 significant both to the ways in which visibility goals may be set and to how air quality
6 management strategies may be devised.
7 Local-scale visibility degradation is commonly in the form of either a plume resulting
8 from the emissions of a specific source or small group of sources, or in the form of a localized
9 haze, such as an urban "brown cloud." Impairment caused by a specific source or small group of
10 sources has been generally termed as "reasonably attributable" impairment. Plumes are
11 comprised of smoke, dust, or colored gas that obscure the sky or horizon relatively near sources.
12 Sources of locally visible plumes, such as the plume from an industrial facility or a burning field,
13 are often easy to identify. There have been a limited number of cases in which Federal land
14 managers have certified the existence of visibility impairment in a Class I area (i.e., 156 national
15 parks, wilderness areas, and international parks identified for visibility protection in section
16 162(a) of the Clean Air Act) that is considered "reasonably attributable" to a particular source.1
17 Urban visibility impairment often results from the combined effect of stationary, mobile,
18 and area source emissions. Complex local meteorological conditions may contribute to such
19 impairment as well. Localized or layered haze often results from emissions from many sources
20 located across an urban or metropolitan area. This type of impairment may be seen as a band or
21 layer of discoloration appearing well above the terrain. A common manifestation of this type of
22 visibility impairment is the "brown cloud" situation experienced in some cities particularly in the
23 winter months, when cooler temperatures limit vertical mixing of the atmosphere. The long-
24 range transport of emissions from sources outside the urban area may also contribute to urban
25 haze levels. A number of studies have been conducted in the past in cities like Denver, Dallas,
26 and Seattle to characterize urban visibility problems.
JTwo of the most notable cases leading to emissions 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.
August 2003 5-3 Draft - Do Not Quote or Cite
-------
1 The second type of impairment, regional haze, results from pollutant emissions from a
2 multitude of sources located across a broad geographic region. It impairs visibility in every
3 direction over a large area, in some cases over multi-state regions. Regional haze masks objects
4 on the horizon and reduces the contrast of nearby objects. The formation, extent, and intensity
5 of regional haze is a function of meteorological and chemical processes, which sometimes cause
6 fine particle loadings to remain suspended in the atmosphere for several days and to be
7 transported hundreds of kilometers from their sources (NRC, 1993). It is this second type of
8 visibility degradation that is principally responsible for impairment in national parks and
9 wilderness areas across the country (NRC, 1993). Visibility in urban areas at times may be
10 dominated by local sources, but often may be significantly affected by long-range transport of
11 haze due to the multi-day residence times of fine particles in the atmosphere. Fine particles
12 transported from urban and industrialized areas in turn may be significant contributors to
13 regional-scale impairment in Class I and other rural areas.
14
15 5.2.2 Effects of PM on Visibility
16 The efficiency with which a unit mass of particles causes visibility impairment depends
17 on a number of factors, including particle size, composition, and humidity. These basic concepts
18 are discussed above in section 2.9.1. Building on this information, this section discusses
19 common measures of visibility impairment, estimated natural visibility conditions, and other
20 important factors in the relationship between PM and visibility impairment.
21 5.2.2.1 Measures of Visibility Impairment
22 Several atmospheric optical indices and approaches can be used for characterizing
23 visibility impairment. As summarized below and discussed in more detail in the CD, there are
24 several indicators that could be used in regulating air quality for visibility protection, including:
25 (1) human observation of visual range; (2) light extinction (and related parameters of visual
26 range and deciview); (3) light scattering by particles; and (4) fine particle mass concentration
27 (CD, page 4-166).
28 Human Observation. For many decades, the National Weather Service has recorded
29 hourly visibility at major airports based on human observations of distant targets. This approach
August 2003 5-4 Draft - Do Not Quote or Cite
-------
1 has provided a historical record of visibility across the U.S. and has allowed a general
2 interpretation of regional visibility trends. Airport visibility monitoring has been automated in
3 recent years, however, through deployment of the Automated Surface Observing System (ASOS)
4 at more than 900 airports across the country (discussed below in section 5.2.4). While human
5 observations have been very effective for the purposes of air safety, these data are not as well
6 correlated to air quality levels as data obtained from automated monitoring methods.
7 Light Extinction and Related Measures. The light extinction coefficient has been
8 widely used in the U.S. for many years as a metric to describe the effect of pollutant
9 concentrations on visibility. It can be defined as the fraction of light lost or redirected per unit
10 distance through interactions with gases and suspended particles in the atmosphere. The light
11 extinction coefficient represents the summation of light scattering and light absorption due to
12 particles and gases in the atmosphere. Both anthropogenic and non-anthropogenic sources
13 contribute to light extinction. The light extinction coefficient (oext) is represented by the
14 following equation (CD, 4-155):
15
16 oex, = osg + oag + osp + oap (5-1)
17
18 where osg = light scattering by gases (also known as Rayleigh scattering)
19 oag = light absorption by gases
20 osp = light scattering by particles
21 oap = light absorption by particles.
22 Light extinction is commonly expressed in terms of inverse kilometers (km"1) or inverse
23 megameters (Mm"1), where increasing values indicate increasing impairment.
24 Total light extinction can be measured directly by a transmissometer or it can be
25 calculated from ambient pollutant concentrations. Transmissometers measure the light
26 transmitted through the atmosphere over a distance of 1 to 15 kilometers. The light transmitted
27 between the light source (transmitter) and the light-monitoring component (receiver) is
28 converted to the path-averaged light extinction coefficient. Transmissometers operate
29 continuously, and data is often reported in terms of hourly averages.
August 2003 5-5 Draft - Do Not Quote or Cite
-------
1 Direct relationships exist between measured ambient pollutant concentrations and their
2 contributions to the extinction coefficient. The contribution of each aerosol constituent to total
3 light extinction is derived by multiplying the aerosol concentration by the extinction efficiency
4 for that aerosol constituent. Extinction efficiencies vary by type of aerosol constituent and have
5 been obtained for typical atmospheric aerosols by a combination of empirical approaches and
6 theoretical calculations. For certain aerosol constituents, extinction efficiencies increase
7 significantly with increases in relative humidity.
8 The IMPROVE visibility monitoring program has developed an algorithm for calculating
9 total light extinction as the sum of aerosol light extinction for each of the five major fine particle
10 components and coarse fraction mass, plus 10 Mm"1 for Rayleigh extinction. The mass for each
11 component is multiplied by its dry extinction efficiency, and in the case of sulfate and nitrate, by
12 a relative humidity adjustment factor to account for their hygroscopic behavior (CD, p. 4-170).
13 The relative humidity adjustment factor increases significantly with higher humidity, ranging
14 from about 2 at 70%, to 4 at 90%, and 10 at 95% relative humidity (CD, p. 4-171, Figure 4-38).
15 In addition to the optical effects of atmospheric constituents as characterized by the
16 extinction coefficient, lighting conditions and scene characteristics play an important role in
17 determining how well we see objects at a distance. Some of the conditions that influence
18 visibility include whether a scene is viewed towards the sun or away from the sun, whether the
19 scene is shaded, and the color and reflectance of the scene (NAPAP, 1991). For example, a
20 mountain peak in bright sun can be seen from a much greater distance when covered with snow
21 because lighter colored terrain will reflect more light than darker colored terrain.
22 One's ability to clearly see an object is degraded both by the reduction of image forming
23 light (transmitted radiance) from the object caused by scattering and absorption, and by the
24 addition of non-image forming light that is scattered into the viewer's sight path. This non-image
25 forming light is called path radiance (EPA, 1996a, p. 8-23). A common example of this effect is
26 the inability to see stars in the daytime due to the brightness of the sky caused by Rayleigh
27 scattering. At night, in the absence of sunlight, the path radiance is small so the stars are readily
28 seen. More light is generally scattered in the forward direction than back towards the light
August 2003 5-6 Draft - Do Not Quote or Cite
-------
1 source, thus causing a haze to appear bright when looking at scenes that are towards the direction
2 of the sun and dark when looking away from the sun.
3 Though these non-air quality related influences on visibility can sometimes be
4 significant, they cannot be accounted for in any practical sense in formulation of national or
5 regional measures to minimize haze. Lighting conditions change continuously as the sun moves
6 across the sky and as cloud conditions vary. Non-air quality influences on visibility also change
7 when a viewer of a scene simply turns his head. Regardless of the lighting and scene conditions,
8 however, sufficient changes in ambient concentrations of PM will lead to changes in visibility
9 (and the extinction coefficient). The extinction coefficient integrates the effects of aerosols on
10 visibility, yet is not dependent on scene-specific characteristics. It measures the changes in
11 visibility linked to emissions of gases and particles that are subject to some form of human
12 control and potential regulation, and therefore can be useful in comparing the potential impact of
13 various air quality management strategies on visibility (NAPAP, 1991).
14 By apportioning the extinction coefficient to different aerosol constituents, one can
15 estimate changes in visibility due to changes in constituent concentrations (Pitchford and Malm,
16 1994). The National Research Council's 1993 report Protecting Visibility in National Parks and
17 Wilderness Areas states that "[P]rogress toward the visibility goal should be measured in terms
18 of the extinction coefficient, and extinction measurements should be routine and systematic."
19 Thus, it is reasonable to use the change in the light extinction coefficient, determined in multiple
20 ways, as the primary indicator of changes in visibility for regulatory purposes.
21 Visual range is a measure of visibility that is inversely related to the extinction
22 coefficient. Visual range can be defined as the maximum distance at which one can identify a
23 large black object against the horizon sky. The colors and fine detail of many objects will be lost
24 at a distance much less than the visual range, however. Visual range has been widely used in air
25 transportation and military operations in addition to its use in characterizing air quality.
26 Conversion from the extinction coefficient to visual range can be made with the following
27 equation (NAPAP, 1991):
28
29 VisualRange(km) = 3.91/oext(km-1) (5-2)
30
August 2003 5-7 Draft - Do Not Quote or Cite
-------
1 Another important visibility metric is the deciview, a unitless metric which describes
2 changes in uniform atmospheric extinction that can be perceived by a human observer. It is
3 designed to be linear with respect to perceived visual changes over its entire range in a way that
4 is analogous to the decibel scale for sound (Pitchford and Malm, 1994). Neither visual range nor
5 the extinction coefficient has this property. For example, a 5 km change in visual range or 0.01
6 km"1 change in extinction coefficient can result in a change that is either imperceptible or very
7 apparent depending on baseline visibility conditions. Deciview allows one to more effectively
8 express perceptible changes in visibility, regardless of baseline conditions. A one deciview
9 change is a small but perceptible scenic change under many conditions, approximately equal to a
10 10% change in the extinction coefficient (Pitchford and Malm, 1994). The deciview metric is
11 also useful in defining goals for perceptible changes in visibility conditions under regulatory
12 programs. Deciview can be calculated from the light extinction coefficient (oext) by the equation:
13
14 Haziness (dv)= 10 ln(oext/10 Mm'1) (5-3)
15
16 Figure 5-1 graphically illustrates the relationships among light extinction, visual range, and
17 deciview.
18
1
0
1
i i i i i mi
7 11 14 16 19 23
i i i i i mi
i
30
I
I
34
I
I
37
I
I
39
I
i mi
42 46
i mi
Extinction (Mm"1) 10 20 30 ^ 50 70 10° 20° 30° 40° 50° 70° 100°
Deciviews (dv)
Visual Range (km) 400 2oo 130 100 so eo 40 20 13 10 s 64
Figure 5-1. Relationship Between Light Extinction, Deciviews, and Visual Range.
Source: Malm, 1999.
19
20 Light Scattering Coefficient. Across the U.S., light scattering is typically a much larger
21 contributor to total light extinction than light absorption. Of the main categories of particles,
22 only elemental carbon is a key contributor to light absorption, and it commonly represents only
23 5-10% of total light extinction (Malm et al., 2000). Light scattering data taken by a
24 nephelometer can be correlated fairly well with total light extinction measurements using certain
August 2003 5-8 Draft - Do Not Quote or Cite
-------
1 assumptions for light absorption. Nephelometers measure the scattering of light by particles
2 contained in a small volume of air, and thus provide a point measurement of scattering.
3 Fine Particle Mass Concentration. Fine particle (e.g., PM2 5) mass concentrations can
4 be used as a general surrogate for visibility impairment. However, as described in many reviews
5 of the science of visibility, the different constituents of PM2 5 have variable effects on visibility
6 impairment. For example, crustal material in general accounts for less light scattering per unit
7 mass than other constituents, and sulfates and nitrates contribute greater amounts of light
8 scattering as relative humidity levels exceed 70%. Thus, while higher PM2 5 mass concentrations
9 generally indicate higher levels of visibility impairment, it is not as precise a metric as the light
10 extinction coefficient. By using historic averages or regional estimates of the component-
11 specific percentage of total mass, however, one can develop reasonable estimates of light
12 extinction from PM mass concentrations (see section 5.2.3.1 for further discussion).
13 5.2.2.2 Rayleigh Scattering and Natural Background Conditions
14 Rayleigh scattering represents the degree of natural light scattering found in a particle-
15 free atmosphere, caused by the gas molecules that make up "blue sky" (e.g., N2, 02). The
16 magnitude of Rayleigh scattering depends on the wavelength or color of the light being scattered
17 as well as on the density of gas in the atmosphere and varies by site elevation, generally from 9
18 to 11 Mm"1 for green light at about 550 r|m (CD, p. 4-158). A standard value of 10 Mm"1 is often
19 used to simplify comparisons of light extinction values across a number of sites with varying
20 elevations (Malm, 2000; CD, p. 4-158). The concept of Rayleigh scattering can be used to
21 establish a theoretical maximum horizontal visual range in the earth's atmosphere. At sea level,
22 this maximum visual range is approximately 330 kilometers. Since certain meteorological
23 conditions can lead to visibility conditions that are close to "Rayleigh," it is analogous to a
24 baseline or boundary condition against which other extinction components can be compared.
25 Light extinction caused by PM from natural sources can vary significantly from day to
26 day and location to location due to natural events such as wildfire, dust storms, and volcanic
27 eruptions. It is useful to consider estimates of natural background concentrations of PM on an
28 annual average basis, however, when evaluating the relative contributions of anthropogenic
29 (man-made) and non-anthropogenic sources to total light extinction.
August 2003 5-9 Draft - Do Not Quote or Cite
-------
1 As discussed in Chapter 2, background PM is defined for the purpose of this document as
2 the distribution of PM concentrations that would be observed in the U.S. in the absence of
3 anthropogenic emissions of primary PM and precursor emissions of VOC, NOX, SO2, and NH3 in
4 North America. Table 2-4 describes the range for annual average regional background PM2 5
5 mass in the eastern U.S. as 2 to 5 |ig/m3, and in the western U.S. as 1 to 4 |ig/m3. For PM10, the
6 estimated annual average background concentrations range from 5 to 11 |ig/m3 in the eastern
7 U.S., and 4 to 8 |ig/m3 in the western U.S.
8 The NAPAP report provides estimates of extinction contributions from Rayleigh
9 scattering plus background levels of fine and coarse particles. In the absence of anthropogenic
10 emissions of visibility-impairing particles, these estimates are 26 + 7 Mm"1 in the East, and 17 +
11 2.5 Mm"1 in the West. These equate to a naturally-occurring visual range in the East of 150 + 45
12 km, and 230 + 40 km in the West. Excluding light extinction due to Rayleigh scatter, annual
13 average background levels of fine and coarse particles are estimated to account for 14 Mm"1 in
14 the East and about 6 Mm"1 in the West. The primary non-anthropogenic substances responsible
15 for natural levels of visibility impairment are naturally-occurring organics, suspended dust
16 (including coarse particles), and water associated with hygroscopic particles. At the ranges of
17 fine particle concentrations associated with natural background conditions noted above, small
18 changes in fine particle mass have a large effect on total light extinction. Thus, higher levels of
19 background fine particles and associated average humidity levels in the East result in a fairly
20 significant difference between naturally occurring visual range in the rural East as compared to
21 the rural West.
22 5.2.2.3 Contribution of PM to Visibility Conditions
23 On an annual average basis, the concentrations of background fine particles are generally
24 small when compared with concentrations of fine particles from anthropogenic sources (NRC,
25 1993). The same relationship holds true when one compares annual average light extinction due
26 to background fine particles with light extinction due to background plus anthropogenic sources.
27 Table VIII-4 in the 1996 Staff Paper makes this comparison for several locations across the
28 country by using background estimates from Table VIII-2 and light extinction values derived
29 from monitored data from the IMPROVE network. These data indicate that anthropogenic
August 2003 5-10 Draft - Do Not Quote or Cite
-------
1 emissions make a significant contribution to average light extinction in most parts of the country,
2 as compared to the contribution from background fine particle levels. Man-made contributions
3 account for about one-third of the average extinction coefficient in the West and more than 80%
4 in the rural East (NAPAP, 1991).
5 It is important to note that even in those areas with relatively low concentrations of
6 anthropogenic fine particles, such as the Colorado plateau, small increases in anthropogenic fine
7 particle concentrations can lead to significant decreases in visual range. This is one reason why
8 Class I areas have been given special consideration under the Clean Air Act. As discussed in the
9 1996 Staff Paper, visibility in an area with lower concentrations of air pollutants (such as many
10 western Class I areas) will be more sensitive to a given increase in fine particle concentration
11 than visibility in a more polluted atmosphere (EPA, 1996b, p. VIII-10, Figure VIII-9).
12 Conversely, to achieve a given amount of visibility improvement, a larger reduction in fine
13 particle concentration is required in areas with higher existing concentrations, such as the East,
14 than would be required in areas with lower concentrations. This relationship between changes in
15 fine particle concentrations and changes in visibility (in deciviews) also illustrates the relative
16 importance of the overall extinction efficiency of the pollutant mix at particular locations. At a
17 given ambient concentration, areas having higher average extinction efficiencies due to the mix
18 of pollutants would have higher levels of impairment. In the East, the combination of higher
19 humidity levels and a greater percentage of sulfate as compared to the West causes the average
20 extinction efficiency for fine particles to be almost twice that for sites on the Colorado Plateau.
21
22 5.2.3 Visibility Conditions in Class I and Non-Urban Areas
23 5.2.3.1 IMPROVE Visibility Monitoring Network
24 In conjunction with the National Park Service, other Federal land managers, and State
25 organizations, EPA has supported monitoring in national parks and wilderness areas since 1988.
26 The network was originally established at 20 sites, but it has now been expanded to 110 sites that
27 represent all but one (Bering Sea) of the 156 mandatory Federal Class I areas across the country.
28 This long-term visibility monitoring network is known as IMPROVE (Interagency Monitoring of
29 PROtected Visual Environments). The following discussion briefly describes the IMPROVE
August 2003 5-11 Draft - Do Not Quote or Cite
-------
1 protocol and provides the rationale supporting use of the light extinction coefficient, derived
2 from either direct optical measurements and/or measurements of aerosol constituents, for
3 purposes of implementing air quality management programs to improve visibility.
4 IMPROVE provides direct measurement of fine particles and precursors that contribute
5 to visibility impairment. The IMPROVE network employs aerosol measurements at all sites, and
6 optical and scene measurements at some of the sites. Aerosol measurements are taken for PM10
7 and PM2 5 mass, and for key constituents of PM2 5, such as sulfate, nitrate, organic and elemental
8 carbon, soil dust, and several other elements. Measurements for specific aerosol constituents are
9 used to calculate "reconstructed" aerosol light extinction by multiplying the mass for each
10 constituent by its empirically-derived scattering and/or absorption efficiency. Knowledge of the
11 main constituents of a site's light extinction "budget" is critical for source apportionment and
12 control strategy development. Optical measurements are used to directly measure light
13 extinction or its components. Such measurements are taken principally with either a
14 transmissometer, which measures total light extinction, or a nephelometer, which measures
15 particle scattering (the largest human-caused component of total extinction). Scene
16 characteristics are typically recorded 3 times daily with 35 millimeter photography and are used
17 to determine the quality of visibility conditions (such as effects on color and contrast) associated
18 with specific levels of light extinction as measured under both direct and aerosol-related
19 methods. Directly measured light extinction is used under the IMPROVE protocol to cross-
20 check that the aerosol-derived extinction levels are reasonable in establishing current visibility
21 conditions. Aerosol-derived extinction is used to document spatial and temporal trends and to
22 determine how proposed changes in atmospheric constituents would affect future visibility
23 conditions.
24 5.2.3.2 Current Conditions and Trends Based on IMPROVE Data
25 Annual average visibility conditions (i.e., total light extinction due to anthropogenic and
26 non-anthropogenic sources) vary regionally across the U.S. The rural East generally has higher
27 levels of impairment than remote sites in the West, with the exception of urban-influenced sites
28 such as San Gorgonio Wilderness (CA) and Point Reyes National Seashore (CA), which have
29 annual average levels comparable to certain sites in the Northeast. Regional differences are
August 2003 5-12 Draft - Do Not Quote or Cite
-------
1 illustrated by Figures 4-39a and 4-39b in the CD, which show that visibility levels on the 20%
2 haziest days in the west (15 deciviews) are about equal to levels on the 20% best days in the east
3 (CD, p. 4-181).
4 Higher visibility impairment levels in the East are due to generally higher concentrations
5 of anthropogenic fine particles, particularly sulfates, and higher average relative humidity levels.
6 In fact, sulfates account for 60-86% of the haziness in eastern sites (CD, 4-236). Aerosol light
7 extinction due to sulfate on the 20% haziest days is significantly larger in eastern Class I areas as
8 compared to western areas (CD, p. 4-183; Figures 4-40a and 4-40b). With the exception of
9 remote sites in the northwestern U.S., visibility is typically worse in the summer months. This is
10 particularly true in the Appalachian region, where average extinction in the summer exceeds the
11 annual average by 40% (Sisler et al., 1996).
12 Regional trends in Class I area visibility are updated annually in the EPA's National Air
13 Quality and Emissions Trends Report (EPA, 2001). Eastern trends for the 20% haziest days
14 from 1992-1999 showed a 1.5 deciview improvement, or about 16%. However, visibility in the
15 east remains significantly impaired, with an average visual range of 14 miles (22 km) on the 20%
16 haziest days. In western Class I areas, aggregate trends showed little change during 1990-1999
17 for the 20% haziest days, and modest improvements on the 20% mid-range and clearest days.
18 Average visual range on the 20% haziest days in western Class I areas is 80 km.
19
20 5.2.4 Urban Visibility Conditions
21 For many years, urban visibility has been characterized using data describing airport
22 visibility conditions. Until the mid-1990's, airport visibility was typically reported on an hourly
23 basis by human observers. An extensive database of these assessments has been maintained and
24 analyzed to characterize visibility trends from the late-1940's to mid-1990's (Schichtel et al.,
25 2000).
26 As noted earlier, visibility impairment has been studied in several major cities in the past
27 decades because of concerns about fine particles and their potentially significant impacts (e.g.,
28 health-related and aesthetic) on the residents of large metropolitan areas (e.g,. Middleton, 1993).
29 Urban areas generally have higher loadings of fine PM and visibility impairment levels than
August 2003 5-13 Draft - Do Not Quote or Cite
-------
1 monitored Class I areas. Urban area annual mean and 98th percentile 24-hour average PM2 5
2 levels for 1999-2001 are presented in Chapter 2. These levels are generally higher than those
3 found in the IMPROVE database for rural Class I areas. Urban areas have higher concentrations
4 of organic carbon, elemental carbon, and particulate nitrate than rural areas due to a higher
5 density of fuel combustion and diesel emissions.
6 5.2.4.1 ASOS Airport Visibility Monitoring Network
7 In 1992, the National Weather Service (NWS), Federal Aviation Administration (FAA),
8 and Department of Defense began deployment of the Automated Surface Observing System
9 (ASOS). ASOS is now the largest instrument-based visibility monitoring network in the U.S.
10 (CD, p. 4-175). The ASOS visibility monitoring instrument is a forward scatter meter that has
11 been found to correlate well with light extinction measurements from the Optec transmissometer
12 (NWS, 1998). It is designed to provide consistent, real-time visibility and meteorological
13 measurements to assist with air traffic control operations. A total of 569 FAA-sponsored and
14 313 NWS-sponsored automated observing systems are installed at airports throughout the
15 country. ASOS visibility data are typically reported for aviation use in small increments up to a
16 maximum of 10 miles visibility. While this truncated data is not useful for characterizing actual
17 visibility levels, the raw, non-truncated data from the 1-minute light extinction and
18 meteorological readings are now archived and available for analysis.
19
August 2003 5-14 Draft - Do Not Quote or Cite
-------
1 5.2.4.2 ASOS Data: Urban Visibility and Correlation to PM2 5 Mass
2 EPA obtained archived 1999 ASOS data for 63 cities across the country. The purpose of
3 analyzing the ASOS data was to improve characterizations of current visibility conditions in
4 non-Class I areas, particularly in urban areas, and to evaluate the relationship between ASOS
5 visibility sensor data and PM2 5 mass concentrations at ASOS sites.
6 EPA staff evaluated the 1-minute ASOS data and developed correlations between ASOS
7 measurements and PM2 5 concentrations, as well as relationships between quarterly correlation
8 coefficient values and longitude of ASOS sites, monthly ASOS visibility patterns, and diurnal
9 patterns by season, hour of day, and day of week (Szykman and Damberg, 2003). Figure 5-2
10 shows quarterly distributions of estimated correlation coefficients in 49 cities between ASOS
11 extinction values and PM2 5 concentrations at nearby Federal Reference Method sites that were
12 operational in 1999 as a function of the longitude of ASOS sites. Correlations exceeded 0.70 in
13 at least one quarter for 40 of the 49 cities. Correlations were more consistently high in the 3rd
14 quarter, with correlations exceeding 0.70 in 29 (or 59%) of the cities.
15 Figures 5-3 and 5-4 provide two examples of relatively well-characterized relationships
16 between predicted PM25 concentrations (based on ASOS extinction values) and measured PM2 5
17 concentrations in Washington, DC and Milwaukee, with correlation coefficients of 0.84 and
18 0.87, respectively. At least in some areas of the country, as seen in these example cities, the use
19 of ASOS data to understand the linkage between PM25 concentrations and urban visibility
20 shows a significant amount of promise. EPA intends to conduct additional, more detailed
21 analyses comparing ASOS measurements to PM25 concentrations from the recently deployed
22 urban speciation network, which may provide more consistently high correlations.
August 2003 5-15 Draft - Do Not Quote or Cite
-------
2
3
4
5
7
8
9
0)
s
o
1
fl
Quarter 1 (January-March 1999)
1 2 ™.,™™™™™™™™™™™™™..^^^
1 _
0.6 -
0 -
-0 4 -
* *» «•
• » .»*?«.
•
» * * <
» • *
-130 -120 -110 -100 -90 -80 -70
ASOS Longitude
1.2 i
o 04
U 02
c Q
i_ 06
Quarter 2 (April -June 1999)
1 •••-..
1 • • * • ' • ^_B •
• ••. • • »
' •' •
•
_ •
•
• Qtr2
-130 -120 -110 -100 -90 -80 -70
ASOS Longitude
10
11
12
Correlation Coefficient
Quarter 3 (July - September 1999)
i*4^4**
***** A ^A **A**
, *. * 4* ^ '
A*
AQtr3]
-130 -120 -110 -100 -90 -80 -70
ASOS Longitude
Correlation Coefficient
1.2 -,
1 _
0 -
-n a -
Quarter 4 (October - December 1999)
(» 0° 0 0 %
o °o°°o «£ %
° ° „ „ o o ° oo0 0
g 0
8 <
w w
o Qtr4J
-130 -120 -110 -100 -90 -80 -70
ASOS Longitude
Figure 5-2. Relationship Between Quarterly Distributions of Estimated Correlation
Coefficients and Longitude of ASOS Sites.
August 2003
5-16
Draft - Do Not Quote or Cite
-------
1
2
3
4
Site Pair: ASOS=DCA , PM=510591004
K 21 31
limnd PI !uj/m3l
- -M O 0 O 2 DDD3 i £ (1 4 5
Figure 5-3. Relationship between Predicted and
Measured PM2 5 Concentrations in Washington, DC.
Diagonal line indicates the regression line.
Source: Szykman and Damberg, 2003.
Sits Pair: ASOS = MKE , PM=550790099
10 30
HiiEund PI In
quailer +^-M o Q o 2 n n n 3 i A £
Figure 5-4. Relationship between Predicted and
Measured PM2 5 Concentrations in Milwaukee, WI.
Diagonal line indicates the regression line.
Source: Szykman and Damberg, 2003.
August 2003
5-17
Draft - Do Not Quote or Cite
-------
1 5.2.5 Significance of Visibility to Public Welfare
2 Visibility is an air quality-related value having direct significance to people's enjoyment
3 of daily activities in all parts of the country. Survey research on public awareness of visual air
4 quality using direct questioning typically reveals that 80% or more of the respondents are aware
5 of poor visual air quality (Cohen et al., 1986). The importance of visual air quality to public
6 welfare across the country has been demonstrated by a number of studies designed to quantify
7 the benefits (or willingness to pay) associated with potential improvements in visibility. More
8 recently, the importance of visual air quality to the policymakers and the general public alike has
9 also been demonstrated by a number of regional, state, and local efforts to address visibility
10 impairment in urban and non-urban areas.
11 5.2.5.1 Value of Improving Visual Air Quality
12 Individuals value good visibility for the well-being it provides them directly, both in the
13 places where they live and work, and in the places where they enjoy recreational opportunities.
14 Millions of Americans appreciate the scenic vistas in national parks and wilderness areas each
15 year. Visitors consistently rate "clean, clear air" as one of the most important features desired in
16 visiting these areas (Department of Interior, 1998). A 1998 survey of 590 representative
17 households by researchers at Colorado State University found that 88% of the respondents
18 believed that "preserving America's most significant places for future generations" is very
19 important, and 87% of the respondents supported efforts to clean up air pollution that impacts
20 national parks (Hass and Wakefield, 1998).
21 Economists have performed many studies in an attempt to quantify the economic benefits
22 associated with improvements in current visibility conditions both in national parks and in urban
23 areas. Economists distinguish between use values and non-use values. Use values are those
24 aspects of environmental quality that directly affect an individual's welfare. These include the
25 aesthetic benefits of better visibility, improved road and air safety, and enhanced recreation in
26 activities like hunting and hiking.
27 Non-use values are those for which an individual is willing to pay for reasons that do not
28 relate to the direct use or enjoyment of any environmental benefit. The component of non-use
29 value that is related to the use of the resource by others in the future is referred to as the bequest
August 2003 5-18 Draft - Do Not Quote or Cite
-------
1 value. This value is typically thought of as altruistic in nature. Another potential component of
2 non-use value is the value that is related to preservation of the resource for its own sake, even if
3 there is no human use of the resource. This component of non-use value is sometimes referred to
4 as existence value or preservation value. Non-use values are not traded, directly or indirectly, in
5 markets. For this reason, the estimation of non-use values has proved to be significantly more
6 difficult than the estimation of use values. Non-use values may be related to the desire that a
7 clean environment be available for the use of others now and in the future, or they may be related
8 to the desire to know that the resource is being preserved for its own sake, regardless of human
9 use. Non-use values may be a more important component of value for recreational areas,
10 particularly national parks and monuments.
11 It is well recognized in the U.S. and abroad that there is an important relationship
12 between good air quality and economic benefits due to tourism. A 1998 Department of Interior
13 study found that travel-related expenditures by national park visitors alone average $14.5 billion
14 annually (1996 dollars) and support 210,000 jobs (Peacock, 1998). A similar estimate of
15 economic benefits resulting from visitation to national forests and other public lands could
16 increase this estimate significantly.
17 McNeill and Roberge (2000) studied the impact of poor visibility episodes on tourism
18 revenues in Greater Vancouver and the Lower Fraser Valley in British Columbia as part of the
19 Georgia Basin Ecosystem Initiative of Environment Canada. Through this analysis a model was
20 developed that predicts future tourist revenue losses that would result from a single extreme
21 visibility episode. They found that such an episode would result in a $7.45 million loss in the
22 Greater Vancouver area and $1.32 million loss in the Fraser Valley.
23 The results of several valuation studies addressing both urban and rural visibility are
24 presented in the 2003 draft Criteria Document (CD, p 4-187 to 4-191), 1996 Criteria Document
25 (EPA, 1996a, p. 8-83, Table 8-5; p. 8-85, Table 8-6) and in the 1996 Staff Paper (EPA, 1996b,
26 Table VIII-1) (Chestnut et al., 1994). Past studies by Schulze (1983) and Chestnut and Rowe
27 (1990b) have estimated the preservation values associated with improving the visibility in
28 national parks in the Southwest to be in the range of approximately $2-6 billion annually. An
29 analysis of the residential visibility benefits in the eastern U.S. due to reduced sulfur dioxide
August 2003 5-19 Draft - Do Not Quote or Cite
-------
1 emissions under the acid rain program suggests an annual value of $2.3 billion (in 1994 dollars)
2 in the year 2010 (Chestnut and Dennis, 1997). The authors suggest that these results could be as
3 much as $1-2 billion more because the above estimate does not include any value placed on
4 eastern air quality improvements by households in the western U.S.
5 Estimating benefits for visibility can be difficult because visibility is not directly or
6 indirectly valued in markets. The studies cited above are based on a valuation method known as
7 contingent valuation (CV). Concerns have been identified about the reliability of value estimates
8 from contingent valuation studies because research has shown that bias can be introduced easily
9 into these studies if they are not carefully conducted. Accurately estimating willingness-to-pay
10 for avoided health and welfare losses depends on the reliability and validity of the data collected.
11 However, there is an extensive scientific literature and body of practice on both the theory and
12 technique of contingent valuation. EPA believes that well-designed and well-executed CV
13 studies are useful for estimating the benefits of environmental effects such as improved visibility
14 (EPA, 2000).
15 Society also values visibility because of the significant role it plays in transportation
16 safety. Serious episodes of visibility impairment can increase the risk of unsafe air
17 transportation, particularly in urban areas with high air traffic levels (EPA, 1982b). In some
18 cases, extreme haze episodes have led to flight delays or the shutdown of major airports,
19 resulting in economic impacts on air carriers, related businesses, and air travelers. For example,
20 on May 15, 1998 in St. Louis, Missouri, it was reported that a haze episode attributed to
21 wildfires in central America resulted in a reduction in landing rates and significant flight delays
22 at Lambert International Airport. The 24-hour PM2 5 levels reached 68 |ig/m3 during that
23 episode. In addition, the NTSB has concluded in accident reports that high levels of pollution
24 and haze, such as those experienced during the July 1999 air pollution episode in the
25 northeastern U.S., have played a role in air transportation accidents and loss of life (NTSB,
26 2000). During this episode, 24-hour levels of PM2 5 ranged from 35-52 |ig/m3 in the New
27 England states.
August 2003 5-20 Draft - Do Not Quote or Cite
-------
1 5.2.5.2 Visibility Goals and Programs
2 The value placed on protecting visual air quality is further demonstrated by the existence
3 of a number of programs, goals, standards, and planning efforts that have been established in the
4 U.S. and abroad to address visibility concerns in urban and non-urban areas. These regulatory
5 and planning activities are of particular interest here as they are illustrative of the significant
6 value that the public places on improving visibility, and because they have developed methods
7 for evaluating public perceptions and judgments about the acceptability of varying degrees of
8 visibility impairment. As discussed below, such methods could be applied to develop additional
9 information that would help to inform EPA's review of the secondary PM NAAQS. Specific
10 discussion is provided below on the Act's special emphasis on protection of visibility in certain
11 areas, the Denver visibility standard, the Phoenix Brown Cloud Summit, and visibility protection
12 efforts in the Canadian province of British Columbia.
13 In addition to the above visibility protection efforts, staff notes that the State of
14 California (California Code of Regulations) and the Lake Tahoe Regional Planning Agency
15 (Molenar, 2000) have also established visibility standards in the U.S. Internationally, the
16 Australian state of Victoria has established a visibility objective (State Government of Victoria,
17 2000a and 2000b), and a visibility guideline is under consideration in New Zealand (New
18 Zealand National Institute of Water & Atmospheric Research, 2000a and 2000b; New Zealand
19 Ministry of Environment, 2000).
20 Sections 169A and 169B of the CAA. In section 302(h) of the Act, visibility impairment
21 is identified as an effect on public welfare to be protected against by secondary NAAQS, among
22 other basic programs under the Act. Additional protection against visibility impairment in
23 special areas is provided for in sections 169A and 169B of the Act. Section 169A, added by the
24 1977 CAA Amendments, established a national visibility goal to "remedy existing impairment
25 and prevent future impairment" in 156 national parks and wilderness areas (Class I areas). The
26 Amendments also called for EPA to issue regulations requiring States to develop long-term
27 strategies to make "reasonable progress" toward the national goal. EPA issued initial regulations
28 in 1980 focusing on visibility problems that could be linked to a single source or small group of
August 2003 5-21 Draft - Do Not Quote or Cite
-------
1 sources. Action was deferred on regional haze until monitoring, modeling, and source
2 apportionment methods could be improved.
3 The 1990 CAA Amendments placed additional emphasis on regional haze issues through
4 the addition of section 169B. In accordance with this section, EPA established the Grand
5 Canyon Visibility Transport Commission (GCVTC) in 1991 to address adverse visibility impacts
6 on 16 Class I national parks and wilderness areas on the Colorado Plateau. The GCVTC was
7 comprised of the Governors of nine western states and leaders from a number of Tribal nations.
8 The GCVTC issued its recommendations to EPA in 1996, triggering a requirement in section
9 169B for EPA issuance of regional haze regulations.
10 EPA accordingly promulgated a final regional haze rule in 1999. Under the regional
11 haze program, States are required to establish goals for improving visibility on the 20% most
12 impaired days in each Class I area, and for allowing no degradation on the 20% least impaired
13 days. Each state must also adopt emission reduction strategies which, in combination with the
14 strategies of contributing States, assure that Class I area visibility improvement goals are met.
15 The first State implementation plans are to be adopted in the 2003-2008 time period, with the
16 first implementation period extending until 2018. Five multistate planning organizations are
17 evaluating the sources of PM2 5 contributing to Class I area visibility impairment to lay the
18 technical foundation for developing strategies coordinated among many States in order to make
19 reasonable progress in Class I areas across the country.
20 Denver Visibility Program and Standard-Setting Methodology. The State of Colorado
21 adopted a visibility standard for the city of Denver in 1990. The Denver standard is violated
22 when the four-hour average light extinction level exceeds 76 Mm"1 (equivalent to approximately
23 50 km visual range and 20 deciviews) during the hours between 8 a.m. and 4 p.m.
24 Transmissometer readings taken when relative humidity is greater than 70% are excluded. The
25 staff note that in setting this standard, the State chose to use a fairly short averaging period (4
26 hours) in recognition of the fact that visibility conditions can change significantly over the
27 course of the day. Of particular interest is the process by which the Denver visibility standard
28 was developed, which relied on citizen judgments of acceptable and unacceptable levels of
29 visual air quality (Ely et al., 1991).
August 2003 5-22 Draft - Do Not Quote or Cite
-------
1 Representatives from Colorado Department of Public Health and Environment (CDPHE)
2 conducted a series of meetings with 17 civic and community groups in which a total of 214
3 individuals were asked to rate slides having varying levels of visual air quality for a well-known
4 vista in Denver. The CDPHE representatives asked the participants to base their judgments on
5 three factors: 1) the standard was for an urban area, not a pristine national park area where the
6 standards might be more strict; 2) standard violations should be at visual air quality levels
7 considered to be unreasonable, objectionable, and unacceptable visually; and 3) judgments of
8 standards violations should be based on visual air quality only, not on health effects.
9 The participants were shown slides in 3 stages. First, they were shown seven warm-up
10 slides describing the range of conditions to be presented. Second, they rated 25 randomly-
11 ordered slides based on a scale of 1 (poor) to 7 (excellent), with 5 duplicates included. Third,
12 they were asked to judge whether the slide would violate what they would consider to be an
13 appropriate urban visibility standard (i.e. whether the level of impairment was "acceptable" or
14 "unacceptable").
15 The Denver visibility standard setting process produced the following findings:
16
17 • Individuals' judgments of a slide's visual air quality and whether the slide violated a
18 visibility standard are highly correlated (Pearson correlation coefficient greater than
19 80%) with the group average.
20
21 • When participants judged duplicate slides, group averages of the first and second ratings
22 were highly correlated.
23
24 • Group averages of visual air quality ratings and "standard violations" were highly
25 correlated. The strong relationship of standard violation judgments with the visual air
26 quality ratings is cited as the best evidence available from this study for the validity of
27 standard violation judgments (Ely et al., 1991).
28
29 The CDPHE researchers sorted the ratings for each slide by increasing order of light
30 extinction and calculated the percent of participants that judged each slide to violate the
31 standard. The Denver visibility standard was then established based on a 50% acceptability
32 criterion. Under this approach, the standard was identified as the light extinction level that
33 divides the slides into two groups: those found to be acceptable and those found to be
34 unacceptable by a majority of study participants. The CDPHE researchers found this level to be
August 2003 5-23 Draft - Do Not Quote or Cite
-------
1 reasonable because for the slides at this level and above, a majority of the study participants
2 judged the light extinction levels to be unacceptable. In fact, when researchers evaluated all
3 citizen judgments made on all slides at this level and above as a single group, more than 85% of
4 the participants found visibility impairment at and above the level of the selected standard to be
5 unacceptable.
6 Images are available on an EPA website
7 (http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html, labeled as Figures 5-5 to 5-12)
8 that illustrate visual air quality in Denver under a range of visibility conditions (generally
9 corresponding to 10th, 20th, 30th, 40th, 50th, 60th 80th, and 90th percentile values). These images
10 were generated using WinHaze, a state-of-the-art image modeling program developed by Air
11 Resource Specialists, Inc. (ARS), discussed below in section 5.2.6.1.
12 Phoenix, Arizona - Governor's Brown Cloud Summit. On March 15, 2000, Governor
13 Jane Dee Hull established an advisory group process called the Brown Cloud Summit, directing
14 the Summit "to establish options for a visibility standard or other method to track progress in
15 improving visibility in the Phoenix area." The Summit process was formed because of citizen
16 and government concerns about the poor visibility conditions that had been prevalent in the
17 Phoenix metropolitan area over the past several years, particularly in the winter months. The
18 Summit issued its final report in January 2001 (Arizona Department of Environmental Quality,
19 2001). A key recommendation of the Summit was that citizen-defined visibility goals should be
20 established, based on a citizen survey process, similar to the Denver approach. The final report
21 also recommended a number of short- and long-term emissions reduction strategies to help
22 improve visibility and public health.
23 In 2002, the Arizona Department of Environmental Quality acted on the recommendation
24 of the Summit and formed the Visibility Index Oversight Committee. The Committee's goal was
25 to coordinate the involvement of Phoenix-area residents in the development of a visibility index.
26 The Phoenix committee patterned its survey process after the process developed by Denver
27 officials to develop a visibility standard there.
28 The committee contracted with a consulting firm to design a survey instrument and
29 administer the survey. A final report of the survey methods and results is now available (BBC
August 2003 5-24 Draft - Do Not Quote or Cite
-------
1 Research & Consulting, prepared for Arizona Department of Environmental Quality, 2003). The
2 survey included 385 participants in 27 separate sessions. A sample size of 385 was carefully
3 chosen so that responses would be representative of the area's population. Participants were
4 carefully recruited to form a sample group that was demographically representative of the larger
5 Phoenix population. Three sessions were held in Spanish.
6 Participants were shown a series of 25 images of the same vista of downtown Phoenix,
7 with South Mountain in the background at a distance of about 40 km. Photographic slides of the
8 images were developed using the WinHaze program. The visibility impairment levels ranged
9 from 15 to 35 deciviews. Participants first rated the randomly-shown slides on a scale of 1
10 (unacceptable) to 7 (excellent). Next, the participants rated slides, again shown in random order,
11 as acceptable or unacceptable. This phase of the survey produced the following findings:
12 • At least 90 percent of all participants found visible air quality acceptable between 15
13 deciviews (87 km visual range) and 20 deciviews (53 km);
14
15 • At 24 deciviews (36 km), nearly half of all participants thought the visible air quality was
16 unacceptable; and
17
18 • By 26 deciviews (29 km), almost three-quarters of participants said it was unacceptable,
19 with nearly all participants considering levels of 31 deciviews (18 km) and higher to be
20 unacceptable.
21
22 The information developed in this survey informed the development of future
23 recommendations by the Visibility Index Oversight Committee for a visibility index for the
24 Phoenix Metropolitan Area. Based on the Committee's recommendations, the Arizona
25 Department of Environmental Quality created the Blue Sky Index. Similar to the approach
26 followed for the Denver standard, the Blue Sky Index focuses on a relatively short averaging
27 time (6-hours). The index establishes a target for tracking the number of days each year (250
28 days in 2001, 260 in 2002, 275 in 2003) during which the 6-hour visual range is 40 km or more,
29 as measured by a transmissometer. The Phoenix survey demonstrates that the rating
30 methodology developed for gathering citizen input for establishing the Denver visibility standard
31 can be reliably transferred to another city while relying on updated imaging technology to
32 simulate a range of visibility impairment levels.
August 2003 5-25 Draft - Do Not Quote or Cite
-------
1 Images are available at the same EPA website
2 (http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html, labeled as Figures 5-13 to 5-20)
3 that illustrate visual air quality in Phoenix under a range of visibility conditions. These images
4 were also generated using the WinHaze program. The modeled light extinction levels are based
5 on correlations derived from 1994-1997 ambient air quality monitoring data.
6 Province of British Columbia, Canada. In 1993, the REVEAL (Regional Visibility
7 Experimental Assessment in the Lower Fraser Valley) field study was undertaken to characterize
8 summertime visibility and ambient aerosol loadings in southwestern British Columbia. In 1994,
9 researchers at the University of British Columbia conducted a pilot study on the perception of
10 acceptable visibility conditions in the area, using photographs and optical measurements taken
11 during the summer of 1993 (Pryor, 1996). The study was based on the methodology used in
12 setting the Denver visibility standard (Ely et al., 1991).
13 Participants in the study were shown slides of two suburban locations in British
14 Columbia: Chilliwack and Abbotsford. After using the same general protocol, Pry or found that
15 responses from this pilot study would indicate a standard in terms of visual range of
16 approximately 40 km for Chilliwack and 60 km for Abbotsford. Pryor (1996) discusses some
17 possible reasons for the variation in standard visibility judgments between the two locations.
18 Factors discussed include the relative complexity of the scenes, different levels of development
19 at each location, potential local source influence on site-specific nephelometer data, and
20 potential bias of the sample population since only students participated. The author observed
21 that the applicable 1995 health objective for PM10, 50 |ig/m3 for a 24 hour period, would roughly
22 correspond to a visual range of 26 km and therefore would not be protective of a visibility
23 standard set in the range of 40-60 km, as suggested by the pilot study findings. The author
24 expressed the view that the pilot study reinforced the conclusion that the methodology originally
25 developed for the Denver standard-setting process is a sound and effective one for obtaining
26 public participation in a standard-setting process, and that it could be adapted for such use in
27 another geographic location with only minor modifications (Pryor, 1996).
28
August 2003 5-26 Draft - Do Not Quote or Cite
-------
1 5.2.6 Evaluating Public Perceptions of Visibility Impairment
2 New methods and tools are now available to communicate and evaluate public
3 perceptions of varying visual effects associated with alternative levels of visibility impairment
4 relative to varying pollution levels and environmental conditions. Survey methods are being
5 refined, applied, and evaluated in various studies, such as those described above in section
6 5.2.5.2 for Denver, Phoenix, and British Columbia. These methods are intended to assess public
7 perceptions as to the acceptability of varying levels of visual air quality, considered in these
8 studies to be an appropriate basis for developing goals and standards for visibility protection.
9 Staff finds that the survey methods used in the cases discussed above produced reasonably
10 consistent results from location to location, each with a majority of participants finding visual
11 ranges within about 40 to 60 km to be acceptable. As discussed below, new tools for the
12 development of photographic representation of visibility impairment, such as the WinHaze
13 software developed by ARS, are also now available that facilitate the use of such survey
14 methods.
15 5.2.6.1 Photographic Representations of Visibility Impairment
16 In the past, the principal method for recording and describing visual air quality has been
17 through 35 millimeter photographs. Under the IMPROVE program, EPA, federal land
18 management agencies, and Air Resource Specialists, Inc. (ARS) have developed an extensive
19 archive of visual air quality photos for national parks and wilderness areas. In comparison, we
20 have only a limited archive of photos of urban areas.
21 The draft CD discusses some of the methods that are now available to represent different
22 levels of visual air quality (CD, p. 4-174). In particular, Molenar (1994) describes a
23 sophisticated visual air quality simulation technique, incorporated into the WinHaze program
24 developed by ARS, which combined various modeling systems under development for the past
25 20 years. The technique relies on first obtaining an original base image slide of the scene of
26 interest. The slide should be of a cloudless sky under the cleanest air quality conditions possible.
27 The light extinction represented by the scene should be derived from aerosol and optical data
28 associated with the day the image was taken, or it should be estimated from contrast
29 measurements of features in the image. The image is then digitized to assign an optical density
August 2003 5-27 Draft - Do Not Quote or Cite
-------
1 to each pixel. At this point, the radiance level for each pixel is estimated. Using a detailed
2 topographic map, technicians identify the specific location from which the photo was taken, and
3 they determine the distances to various landmarks and objects in the scene. With this
4 information, a specific distance and elevation is assigned to each pixel.
5 Using the digital imaging information, the system then computes the physical and optical
6 properties of an assumed aerosol mix. These properties are input into a radiative transfer model
7 in order to simulate the optical properties of varying pollutant concentrations on the scene.
8 WinHaze, an image modeling program for personal computers that employs simplified
9 algorithms based on the sophisticated modeling technique, is now available (Air Resource
10 Specialists, 2003).
11 An alternative technique would be to obtain actual photographs of the site of interest at
12 different ambient pollution levels. However, long-term photo archives of this type exist for only
13 a few cities. In addition, studies have shown that observers will perceive an image with a cloud-
14 filled sky as having a higher degree of visibility impairment than one without clouds, even
15 though the PM concentration on both days is the same. The simulation technique has the
16 advantage that it can be done for any location as long as one has a very clear base photo. In
17 addition, the lack of clouds and consistent sun angle in all images in effect standardizes the
18 perception of the images and enables researchers to avoid potentially biased responses due to
19 these factors.
20 5.2.6.2 Future Assessment of Public Opinions on Air Pollution-Related Visibility
21 Impairment
22 In the preliminary draft Staff Paper (EPA, 2001), information was presented on a small
23 pilot project conducted by EPA that applied the survey and photographic techniques discussed
24 above (Abt Associates, 2001). In conjunction with this pilot project, images that illustrate visual
25 air quality in Washington, DC under a range of visibility conditions were prepared and are
26 available at http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html (labeled as Figures 5-
27 21 to 5-28). Two example images from the set of images found on the website are included
28 below, showing that the U.S. Capitol building, located less than 5 miles from where the base
29 photo was taken, becomes almost entirely obscured at a PM25 concentration of 65 |ig/m3.
August 2003 5-28 Draft - Do Not Quote or Cite
-------
Figure 5-26. Washington, DC - 30 ug/m3
PM25 (16 km visual range)
Figure 5-28. Washington, DC - 65 ug/m3
PM2 5 (8 km visual range)
1
2 Information on the pilot project was presented in the preliminary draft Staff Paper to
3 elicit CAS AC and public comment on the use of this type of approach to help inform EPA's
4 review of the secondary PM NAAQS, and, more specifically, to elicit comments on various
5 aspects of the survey methodology used in the pilot project. The project was premised on the
6 view that public perceptions of and judgments about the acceptability of visibility impairment in
7 urban areas are relevant factors in assessing what constitutes an adverse level of visibility
8 impairment in the context of this NAAQS review.
9 EPA received general support for the use of this type of approach, and also received
10 advice from members of CAS AC as to how the survey methodology could be improved. At that
11 time, EPA staff expressed the intention of refining the approach based on that advice, and
12 preparing a revised methodology document for additional review by CASAC and the public prior
13 to conducting a more extensive survey that could appropriately inform this review. Although
14 resource constraints have since prevented this work from being conducted, EPA hopes to pursue
15 it in the future so the results of a more extensive survey can be used to help inform the next
16 periodic review of the PM secondary standards.
17
18 5.2.7 Summary
19 The draft CD and other reports referenced in section 5.1 provide a significant body of
20 information documenting the effects of PM and its components on atmospheric visibility. The
August 2003
5-29
Draft - Do Not Quote or Cite
-------
1 draft CD provides updated information on studies evaluating particle scattering efficiencies and
2 the relative humidity effects primarily on sulfate and nitrate particle size. Visibility trends in
3 national parks and wilderness areas are presented using data from the IMPROVE visibility
4 monitoring network. While recent emphasis has been placed on characterizing rural visibility
5 conditions for the purposes of implementing the regional haze rule, urban visibility conditions
6 continue to be poorly characterized. In an effort to improve characterization of urban visibility,
7 staff conducted an initial analysis of newly available high-resolution (1-minute) data from the
8 ASOS monitoring network and comparison to nearby PM2 5 concentrations, concluding that such
9 analysis appears to be a promising approach. Staff anticipates that additional analyses
10 comparing visibility data from ASOS to data from continuous PM monitors, as well as
11 information from the PM2 5 speciation monitoring network and from expanded use of continuous
12 visibility monitoring in urban areas, will allow for better characterization of current conditions
13 and temporal variations in urban visibility conditions across the country in the future.
14 Information is presented above on the efforts of various urban areas to develop local
15 visibility standards based on a survey methodology that elicits citizen judgments about
16 acceptable and unacceptable levels of visibility as portrayed through photographic slides. In the
17 preliminary Staff Paper, staff discussed a pilot focus group project conducted by EPA in 2000 to
18 evaluate the methodology used in establishing the Denver visibility standard in combination with
19 the WinHaze photographic modeling technique. Similar approaches were used as part of studies
20 to address visibility problems in Phoenix and the Frasier Valley of British Columbia. The
21 common survey approach used in the Denver, Phoenix, and British Columbia studies yielded
22 reasonably consistent results, with each study indicating that a majority of citizens find value in
23 protecting local visibility to within a visual range of about 40 to 60 km. Staff finds this approach
24 to evaluating citizen judgments about acceptable visibility to be promising, and hopes to
25 continue to pursue this approach in the future.
26 Staff believes that currently available information provides a basis for considering
27 whether any revisions to the PM2 5 secondary standards may be appropriate for protecting against
28 PM-related visibility effects. Staff recognizes that until additional monitoring, data analyses,
29 and assessments of citizen judgments of acceptable urban visibility as discussed above can be
August 2003 5-30 Draft - Do Not Quote or Cite
-------
1 conducted, limitations in the available information significantly constrain development of a
2 quantitative basis for a secondary standard(s) based solely on such information. However, staff
3 believes it remains appropriate in this review to consider whether any revisions that may be
4 made to the primary PM2 5 standards are also appropriate for revising the PM2 5 secondary
5 standards to protect against PM-related effects on visibility. Staff also believes that the
6 consideration of any revisions to the secondary PM2 5 standards can be appropriately informed by
7 comparing photographic representations of the degree of visibility impairment allowed under the
8 current 24-hour standard with the degree of visibility impairment that is possible under any
9 alternate 24-hour standard. Photographic representations of visibility conditions associated with
10 a range of PM25 concentrations are presented in Figures 5-5 through 5-28 for Denver, Phoenix,
11 and Washington, DC. In addition, actual photographs of Chicago illustrating visibility
12 conditions associated with a range of PM25 concentrations are presented in Figures 5-29 through
13 5-34. All of these images are available on an EPA website at
14 http ://www. epa. gov/ttn/naaqs/standards/pm/s_pm_cr_sp.html.
15
16
17 5.3 EFFECTS ON MATERIALS
18 The effects of the deposition of atmospheric pollution, including ambient PM, on
19 materials are related to both physical damage and aesthetic qualities. The deposition of PM
20 (especially sulfates and nitrates) can physically affect materials, adding to the effects of natural
21 weathering processes, by potentially promoting or accelerating the corrosion of metals, by
22 degrading paints, and by deteriorating building materials such as concrete and limestone.
23 Particles contribute to these physical effects because of their electrolytic, hygroscopic and acidic
24 properties, and their ability to sorb corrosive gases (principally SO2). As noted in the last
25 review, only chemically active fine-mode or hygroscopic coarse-mode particles contribute to
26 these physical effects (EPA 1996b, p. VIII-16).
27 In addition, the deposition of ambient PM can reduce the aesthetic appeal of buildings
28 and culturally important articles through soiling. Particles consisting primarily of carbonaceous
29 compounds cause soiling of commonly used building materials and culturally important items
August 2003 5-31 Draft - Do Not Quote or Cite
-------
1 such as statues and works of art (CD, p. 4-191). Soiling is the deposition of particles on surfaces
2 by impingement, and the accumulation of particles on the surface of an exposed material results
3 in degradation of its appearance. Soiling can be remedied by cleaning or washing, and
4 depending on the soiled material, repainting (EPA, 1996b, p. VIII-19).
5 Building upon the information presented in the last Staff Paper (EPA, 1996b), and
6 including the limited new information presented in Chapter 4 of the draft CD, the following
7 sections summarize the physical damage and aesthetic soiling effects of PM on materials
8 including metals, paint finishes, and stone and concrete.
9
10 5.3.1 Materials Damage Effects
11 Physical damage such as corrosion, degradation, and deterioration occurs in metals, paint
12 finishes, and building materials such as stone and concrete, respectively. Metals are affected by
13 natural weathering processes even in the absence of atmospheric pollutants. Atmospheric
14 pollutants, most notably SO2 and particulate sulfates, can have an additive effect, by promoting
15 and accelerating the corrosion of metals. The rate of metal corrosion depends on a number of
16 factors, including the deposition rate and nature of the pollutants; the influence of the protective
17 corrosion film that forms on metals, slowing corrosion; the amount of moisture present;
18 variability in electrochemical reactions; the presence and concentration of other surface
19 electrolytes; and the orientation of the metal surface. Historically, studies have shown that the
20 rate of metal corrosion decreases in the absence of moisture, since surface moisture facilitates
21 the deposition of pollutants and promotes corrosive electrochemical reactions on metals (CD, pp.
22 4-192 to 4-194).
23 The draft CD (p. 4-195, Table 4-18) summarizes the results of a number of studies
24 investigating the roles of particles and SO2 on the corrosion of metals. The draft CD concludes
25 that the role of particles in the corrosion of metals is not clear (CD, p. 4-193). While several
26 studies suggest that particles can promote the corrosion of metals, others have not demonstrated
27 a correlation between particle exposure and metal corrosion. Although the corrosive effects of
28 SO2 exposure in particular have received much study, there remains insufficient evidence to
August 2003 5-32 Draft - Do Not Quote or Cite
-------
1 relate corrosive effects to specific participate sulfate levels or to establish a quantitative
2 relationship between ambient particulate sulfate and corrosion.
3 Similar to metals, paints also undergo natural weathering processes, mainly from
4 exposure to environmental factors such as sunlight, moisture, fungi, and varying temperatures.
5 Beyond these natural processes, atmospheric pollutants can affect the durability of paint finishes
6 by promoting discoloration, chalking, loss of gloss, erosion, blistering, and peeling. Historical
7 evidence indicates that particles can damage painted surfaces by serving as carriers of more
8 corrosive pollutants, most notably SO2, or by serving as concentration sites for other pollutants.
9 If sufficient damage to the paint occurs, pollutants may penetrate to the underlying surface. A
10 number of studies available in the last review showed some correlation between PM exposure
11 and damage to automobile finishes. In particular, Wolff et al. (1990) concluded that damage to
12 automobile finishes resulted from calcium sulfate forming on painted surfaces by the reaction of
13 calcium from dust particles with sulfuric acid contained in rain or dew. In addition, paint films
14 permeable to water are also susceptible to penetration by acid-forming aerosols (EPA 1996b, p.
15 VIII-18). The erosion rate of oil-based house paint has reportedly been enhanced by exposure to
16 SO2 and humidity; several studies have suggested that this effect is caused by the reaction of SO2
17 with extender pigments such as calcium carbonate and zinc oxide, although Miller et al. (1992)
18 suggest that calcium carbonate acts to protect paint substrates (CD, p. 4-194).
19 With respect to damage to building stone, numerous studies discussed in the draft CD
20 (pp. 4-197 to 4-202; Table 4-19) suggest that air pollutants, including sulfur-containing
21 pollutants and wet or dry deposition of atmospheric particles and dry deposition of gypsum
22 particles, can enhance natural weathering processes. Exposure-related damage to building stone
23 results from the formation of salts in the stone that are subsequently washed away by rain,
24 leaving the surface more susceptible to the effects of air pollutants. Dry deposition of sulfur-
25 containing pollutants and carbonaceous particles promotes the formation of gypsum on the
26 stone's surface. Gypsum is a black crusty material that occupies a larger volume than the
27 original stone, causing the stone's surface to become cracked and pitted, leaving rough surfaces
28 that serve as sites for further deposition of airborne particles (CD, p. 4-197).
August 2003 5-33 Draft - Do Not Quote or Cite
-------
1 The rate of stone deterioration is determined by the pollutant mix and concentration, the
2 stone's permeability and moisture content, and the pollutant deposition velocity. Dry deposition
3 of SO2 between rain events has been reported to be a major causative factor in pollutant-related
4 erosion of calcareous stones (e.g., limestone, marble, and carbonated cement). While it is clear
5 from the available information that gaseous air pollutants, in particular SO2, will promote the
6 decay of some types of stones under specific conditions, carbonaceous particles (non-carbonate
7 carbon) and particles containing metal oxides may help to promote the decay process (CD, p. 4-
8 302).
9
10 5.3.2 Soiling Effects
11 Soiling affects the aesthetic appeal of painted surfaces. In addition to natural factors,
12 exposure to PM may give painted surfaces a dirty appearance, although few studies are available
13 that evaluate the soiling effects of particles (CD, p. 4-221). Early studies demonstrated an
14 association between particle exposure and increased frequency of cleaning painted surfaces.
15 More recently, Haynie and Lemmons (1990) conducted a study to determine how various
16 environmental factors contribute to the rate of soiling on white painted surfaces. They reported
17 that coarse-mode particles initially contribute more to soiling of horizontal and vertical surfaces
18 than do fine-mode particles, but are more easily removed by rain, leaving stains on the painted
19 surface. The authors concluded that the accumulation of fine-mode particles, rather than coarse-
20 mode particles, more likely promotes the need for cleaning of the painted surfaces (EPA 1996b,
21 p. VIII-21-22; CD, pp. 4-202 to 4-204). Haynie and Lemmons (1990) and Creighton et al.
22 (1990) reported that horizontal surfaces soiled faster than vertical surfaces and that large
23 particles were primarily responsible for the soiling of horizontal surfaces not exposed to rainfall.
24 Additionally, a study was conducted to determine the potential soiling of artwork in five
25 Southern California museums (Ligocki, et al., 1993). Findings were that a significant fraction of
26 fine elemental carbon and soil dust particles in the ambient air penetrates to the indoor
27 environment and may constitute a soiling hazard to displayed artwork (EPA 1996b, p. VIII-22).
28 As for stone structures, the presence of gypsum is related to soiling of the stone surface
29 by providing sites for particles of dirt to concentrate. Lorusso et al. (1997) attributed the need
August 2003 5-34 Draft - Do Not Quote or Cite
-------
1 for frequent cleaning and restoration of historic monuments in Rome to exposure to total
2 suspended particles (TSP). Further, Davidson et al. (2000) evaluated the effects of air pollution
3 exposure on a limestone structure on the University of Pittsburgh campus using estimated
4 average TSP levels in the 1930s and 1940s and actual values for the years 1957 to 1997.
5 Monitored levels of SO2 were also available for the years 1980 to 1998. Based on the available
6 data on pollutant levels and photographs, the authors concluded that soiling began while the
7 structure was under construction. With decreasing levels of pollution, the soiled areas have been
8 slowly washed away, the process taking several decades, leaving a white, eroded surface (CD,
9 pp. 4-203 to 4-204).
10
11 5.3.3 Summary
12 Damage to building materials results from natural weathering processes that are
13 enhanced by exposure to airborne pollution, most notably sulfur-containing pollutants. Ambient
14 PM has been associated with contributing to pollution-related damage to materials, and can
15 cause significant detrimental effects by soiling painted surfaces and other building materials.
16 Available data indicate that particle-related soiling can result in increased cleaning frequency
17 and repainting, and may reduce the useful life of the soiled materials. However, to date, no
18 quantitative relationships between particle characteristics (e.g., concentrations, particle size, and
19 chemical composition) and the frequency of cleaning or repainting have been established.
20
21 5.4 EFFECTS ON VEGETATION AND ECOSYSTEMS
22 Environmental impacts of ambient PM are considered here in relation to effects on
23 vegetation and other components of the environment, such as soils, water, and wildlife, that
24 make up ecosystems. The following sections discuss the components of PM most relevant for
25 vegetation and ecosystems (5.4.1); the mechanisms by which vegetation or ecosystems come in
26 contact with PM through the processes of deposition (5.4.2); the nature and extent of direct
27 effects of PM to vegetation (5.4.3); the nature and extent of direct and indirect effects of PM on
28 ecosystems (5.4.4); urban ecosystems (5.4.5); PM speciation data and monitoring networks
August 2003 5-35 Draft - Do Not Quote or Cite
-------
1 (5.4.6); and an overall summary of information (5.4.7). These discussions are based on
2 information contained and referenced in Chapter 4 of the draft CD.
3 This review also introduces the Framework for Assessing and Reporting on Ecological
4 Condition recommended by the Ecological Processes and Effects Committee (EPEC) of EPA's
5 Science Advisory Board (hereafter EPEC Framework; SAB, 2002) and described in subsections
6 4.2.1 and 4.2.3 of the draft CD. Such a framework can provide a useful approach for organizing
7 discussions of stressor effects on ecosystem components at successive levels of complexity,
8 facilitating elucidation of the linkages between effects at different hierarchical scales and putting
9 them in the broader context of ecological integrity. The vegetation and ecosystem effects of PM
10 identified below encompass the full range, scales, and properties of biological organization listed
11 under the Framework's Essential Ecological Attribute (EEA) "Biotic Condition" which includes
12 effects at the organism, species/population, and community/ecosystem levels (CD, pp. 4-2 to 4-
13 4; 4-55 to 4-60).
14
15 5.4.1 PM Characterization
16 As previously discussed, PM is not a single pollutant, but a heterogeneous mixture of
17 particles differing in size, origin, and chemical composition. Fine PM, which is typically more
18 diverse than coarse PM, is predominantly secondary in nature, having condensed from the vapor
19 phase or been formed by chemical reaction from gaseous precursors in the atmosphere. Sulfur
20 and nitrogen oxides are often oxidized to their respective acids and neutralized with ammonium
21 cations as particulate salts. Fine PM may also contain VOCs, volatilized metals, and products of
22 incomplete combustion. Coarse PM, by contrast, is predominantly primary in nature, having
23 been emitted from area or point sources, and may contain iron, silica, aluminum, and base
24 cations from soil, fragments of plants and insects, pollen, fungal spores, bacteria, and viruses, as
25 well as automobile related debris (e.g., brake linings) (CD, pp. 4-6, 4-7).
26 Though the chemical constitution of individual particles can be strongly correlated with
27 size, the relationship between particle size and particle composition can also be quite complex,
28 making it difficult in most cases to use particle size as a surrogate for chemistry. For example,
29 nitrogen exhibits a strongly bimodal size distribution, with the peak above l|im attributed to
August 2003 5-36 Draft - Do Not Quote or Cite
-------
1 HNO3 adsorption onto coarse alkaline particles, and the peak below 1 jim attributed to gas phase
2 condensation of ammonia with either sulfuric or nitric acid. This heterogeneity of PM exists not
3 only within individual particles or samples from individual sites, but to an even greater extent,
4 between samples from different sites. At this point in time, this heterogeneity of PM across
5 space and time has not been adequately characterized . Since vegetation and other ecosystem
6 components are affected more by particulate chemistry than size fraction, exposure to a given
7 mass concentration of airborne PM may lead to widely differing plant or ecosystem responses,
8 depending on the particular mix of deposited particles. At this time, there has been relatively
9 little research aimed at defining the effects of unspeciated PM mass on plants or ecosystems,
10 though effects of some specific chemical components of PM have been described (CD, pp. 4-6 to
11 4-8).
12
13 5.4.2 PM Deposition
14 For PM to affect plants and ecosystems, it must first be removed from the atmosphere
15 and deposited onto an ecological surface (e.g. plants, water, soils) through the processes of
16 deposition. There are three major routes of deposition: (1) wet deposition in which particles are
17 deposited in rain and snow; (2) occult deposition in which particles are deposited in fog, cloud-
18 water and mists; and (3) dry deposition in which dry particles are deposited directly onto
19 surfaces. Available evidence suggests that all modes of deposition must be considered in
20 determining potential impacts to vegetation and ecosystems because each mode may dominate
21 over specific intervals of time or space. Wet deposition is generally more effective for removing
22 fine-mode PM from the atmosphere, whereas dry deposition is more effective for coarse-mode
23 particles (CD, p. 4-9).
24 Several current national monitoring networks (discussed in more detail below in section
25 5.4.6) have measured both total wet and total dry deposition of a number of ambient air
26 pollutants for the past several decades. However, though much is known about the total
27 deposition of certain chemical species of airborne pollutants, the percentage of that total
28 deposition attributed to particles is often not known or specifically measured.
August 2003 5-37 Draft - Do Not Quote or Cite
-------
1 All forms of particle deposition are affected by many factors (CD, p. 4-10, Table 4-2).
2 For example, dry deposition is influenced by particle size, shape, chemical composition, macro-
3 and micro-surface characteristics (e.g., canopy aerodynamics/roughness, leaf properties such as
4 stickiness/wettability, shape, orientation), and atmospheric characteristics (e.g., wind speed,
5 temperature, humidity) and stability. Wet deposition is influenced by some of the same factors
6 as well as the timing, intensity and duration of the precipitation event, the temperature and
7 humidity conditions following the event, and the amount of dry PM previously deposited on the
8 surface that becomes soluble, potentially increasing foliar uptake, or washed off, transferring it
9 to the soil or other media. The deposition of PM in fog droplets in occult deposition is directly
10 proportional to wind speed, droplet size, concentration, and fog density (liquid water content per
11 volume air). Occult deposition is particularly effective for delivery of dissolved and suspended
12 materials to vegetation because: (1) concentrations of ions are often many-fold higher in clouds
13 or fog than in precipitation or ambient air (e.g., acidic cloud water, which is typically 5-20 times
14 more acid than rainwater, can increase pollutant deposition and exposure to vegetation and soils
15 at high elevation sites by more than 50% of wet and dry deposition levels); (2) PM is delivered
16 in a hydrated and bioavailable form to foliar surfaces and remains hydrated due to conditions of
17 high relative humidity and low radiation; and (3) the mechanisms of sedimentation and
18 impaction for submicron particles that would normally be low in ambient air are increased.
19 High-elevation forests receive larger particulate deposition loadings than equivalent low-
20 elevation sites, due to a number of orographic (mountain related) effects. These orographic
21 effects include higher wind speeds that enhance the rate of aerosol impaction, enhanced rainfall
22 intensity and composition, and increased duration of occult deposition. Additionally, the needle-
23 shaped leaves of the coniferous species often found growing in these high elevation sites,
24 enhance impaction and retention of PM delivered by all three deposition modes (CD, pp. 4-12 to
25 4-44).
26 Given all the factors enumerated above and in the CD, direct measurement of PM
27 deposition to the environment can be extremely difficult, and available tabulations of PM
28 deposition velocity (Vd) remain highly variable and suspect (CD, p. 4-9). Due to these
29 complexities, numerous models have been developed to calculate Vd. However, modeling
August 2003 5-38 Draft - Do Not Quote or Cite
-------
1 particle deposition to vegetation is at a relatively early stage of development, and it is not
2 possible to identify a best or most generally applicable modeling approach at this time (CD, p. 4-
3 20). Further, a review by Wesely and Hicks (2000) concluded that a comprehensive
4 understanding of particulate deposition remains a distant goal. It is clear that substantially
5 improved techniques for monitoring and predicting particulate deposition will be required to
6 characterize with some degree of certainly quantitative relationships between ambient PM
7 concentrations, rates and quantities of PM deposition, and associated biological effects on plants
8 and ecosystems (CD, pp. 4-39, 4-54).
9
10 5.4.3 Direct Effects on Vegetation
11 Plant effects can result from either the physical or the chemical properties of PM, or both,
12 and may be caused directly by particle deposition onto the affected vegetation or indirectly
13 through deposition to other media such as soils or water. Particulate matter that deposits directly
14 from the atmosphere onto above-ground plant surfaces may: (1) reside on the leaf, twig, or bark
15 surface for an extended period; (2) be taken up through the leaf surface; or (3) be removed from
16 the plant via resuspension to the atmosphere, washing off by rainfall, or litter-fall with
17 subsequent transfer to the soil (CD, p. 4-62). The following discussion focuses on those particles
18 that are intercepted by and remain on the leaves.
19 Physical effects of PM occur mainly in areas where deposition rates for particles in the
20 coarse mode are high, such as near roadways, agricultural areas and industrial sites. High
21 deposition rates have in some cases led to crust formation on plant leaves. Physical effects that
22 have been observed in vegetation in such areas include reduced photosynthesis and subsequent
23 reductions in carbohydrate formation, and root and plant growth; blockage of the stomata
24 preventing adequate gas exchange; changes in leaf temperature (e.g., heat stress); destruction of
25 leaf tissue (e.g., chlorosis, necrosis, and/or abscission); and premature leaf-fall (CD, p. 4-63).
26 In most areas, however, where deposition rates are not high enough for significant
27 physical effects from PM to occur, the chemical composition of PM becomes the key phytotoxic
28 factor leading to plant injury. Often, it is the chemical composition or class of PM in the fine
29 mode that produces phytotoxic effects when deposited onto plant surfaces. Most information
August 2003 5-39 Draft - Do Not Quote or Cite
-------
1 currently available on plant effects focuses on nitrate particle deposition, in particular, and more
2 generally on acidic deposition, primarily from nitrogen- and sulfur- containing particles and
3 gaseous pollutants. To a lesser degree, the effects of trace metals and organics are also
4 considered. However, studies of the direct effects of chemical additions to foliage through
5 particle deposition have found little or no effects of PM on foliar processes unless exposure
6 levels were significantly higher than typically would be experienced in the ambient environment.
7 Further, only a few studies have been completed on the direct effects of fine-mode particles on
8 vegetation, and the conclusion reached in the 1982 PM/SOx Air Quality Criteria Document (U.S.
9 EPA, 1982), that sufficient data were not available for adequate quantification of dose-response
10 functions, continues to be true today (CD, p. 4-68).
11 Effects of Nitrogen Deposition. Nitrogen has long been recognized as the nutrient most
12 important for plant growth. For instance, approximately 75% of the nitrogen in a plant leaf is
13 used during the process of photosynthesis, and to a large extent, it governs the utilization of
14 phosphorus, potassium, and other nutrients (CD, p. 4-95). Particle deposition of nitrate, together
15 with other nitrogen-containing gaseous and precipitation-derived sources, represents a
16 substantial fraction of total nitrogen reaching vegetation. However, much of this nitrogen is
17 contributed by gaseous nitric acid vapor, and a considerable amount of the particulate nitrate is
18 taken up indirectly through the soil. Though plants usually absorb nitrogen (as NH4+ or NO3")
19 through their roots, it is known that foliar uptake of nitrate can also occur. However, the
20 mechanism of foliar uptake is not well established, plants vary in their ability to absorb
21 ammonium and nitrate, and it is not currently possible to distinguish sources of chemicals
22 deposited as gases or particles using foliar extraction. Since it has proven difficult to quantify
23 the percentage of nitrogen uptake by leaves that is contributed by ambient particles, direct foliar
24 effects of nitrogen-containing particles have not been documented. (CD, pp. 4-68, 4-71).
25 Effects of Sulfur Deposition. Similar to nitrogen, sulfur is an essential plant nutrient
26 that can deposit on vegetation in the form of sulfate particles, or be taken up by plants in gaseous
27 form. Greater than 90% of anthropogenic sulfur emissions are as sulfur dioxide (SO2), with most
28 of the remaining emissions in the form of sulfate. However, sulfur dioxide is rapidly
29 transformed in the atmosphere to sulfate, which is approximately 30-fold less phytotoxic than
August 2003 5-40 Draft - Do Not Quote or Cite
-------
1 SO2. Low dosages of sulfur can serve as a fertilizer, particularly for plants growing in sulfur-
2 deficient soils. There are only a few field demonstrations of foliar sulfate uptake, however, and
3 the relative importance of foliar leachate and prior dry-deposited sulfate particles remains
4 difficult to quantify. Though current levels of sulfate deposition reportedly exceed the capacity
5 of most vegetative canopies to immobilize the sulfur, sulfate additions in excess of needs do not
6 typically lead to plant injury. Additional studies are needed, however, on the effects of sulfate
7 particles on physiological characteristics of plants following chronic exposures (CD, pp. 4-71, 4-
8 72).
9 Effects of Acidic Deposition. Though dry deposition of nitrate and sulfate particles does
10 not appear to induce foliar injury at current ambient exposures, when found in acidic
11 precipitation, such particles do have the potential to cause direct foliar injury. This is especially
12 true when the acidic precipitation is in the form of fog and clouds, which may contain solute
13 concentrations many times those found in rain. In experiments on seedling and sapling trees,
14 both coniferous and deciduous species showed significant effects on leaf surface structures after
15 exposure to simulated acid rain or acid mist at pH 3.5, while some species have shown subtle
16 effects at pH 4 and above. Epicuticular waxes, which function to prevent water loss from plant
17 leaves, can be destroyed by acid rain in a few weeks, which suggests links between acidic
18 precipitation and aging. Due to their longevity and evergreen foliage, the function of
19 epicuticular wax is more crucial in conifers. For example, red spruce seedlings, which have been
20 extensively studied, appear to be more sensitive to acid precipitation (mist and fog) when
21 compared with other species (CD, pp. 4-72, 4-73). In addition to accelerated weathering of leaf
22 cuticular surfaces, other direct responses of forest trees to acidic precipitation include increased
23 permeability of leaf surfaces to toxic materials, water, and disease agents; increased leaching of
24 nutrients from foliage; and altered reproductive processes (CD, p. 4-86). All of these effects
25 serve to weaken trees so that they are more susceptible to other stresses (e.g., extreme weather,
26 pests, pathogens).
27 Trace elements. Of the 90 elements that make up the inorganic fraction of the soil, 80
28 exist in concentrations of less than 0.1% and are known as "trace elements". Trace elements
29 with a density greater than 6 g/cm3 are referred to as "heavy metals." Although some trace
August 2003 5-41 Draft - Do Not Quote or Cite
-------
1 metals are essential for vegetative growth or animal health, in large quantities, they are all toxic.
2 Most trace metals found in the atmosphere are produced by industrial combustion processes and
3 exist predominantly as metal chloride particles, which tend to be volatile, or as metal oxides,
4 which tend to be nonvolatile. Generally, only the heavy metals cadmium, chromium, nickel and
5 mercury are released from stacks (CD, pp. 4-74, 4-75).
6 Investigations of trace elements present along roadsides and in industrial and urban
7 environments have indicated that impressive burdens of particulate heavy metal can accumulate
8 on vegetative surfaces. Once on the surface, these metals can potentially impact either the
9 metabolism of above-ground plant tissues or the activity of populations of organisms resident on
10 and in the leaf surface (e.g., bacteria, fungi and arthropods). A trace metal must be brought into
11 solution before it can enter into the leaves or bark of vascular plants. Since the solubility of most
12 trace metals is low, foliar uptake and direct heavy metal toxicity is limited. In those instances
13 when trace metals are absorbed, they are frequently bound in leaf tissue and are lost when the
14 leaf later drops off. Only a few metals have been documented to cause direct phytotoxicity in
15 field conditions, with copper, zinc and nickel toxicities observed most frequently. It is unlikely,
16 therefore, that deposition of trace metals to vegetative surfaces at ambient levels is causing wide-
17 spread acute plant toxicity. Only a few studies have been conducted on the effects of trace
18 metals on leaf surface organisms, specifically fungi, and these used soluble compounds
19 containing heavy metals. Results indicated variations in fungal sensitivity and tolerance, with
20 iron, aluminum, nickel, zinc, manganese and lead exhibiting the broadest spectrum of growth
21 suppression. Though trace metals probably occur naturally on leaf surfaces as low-solubility
22 compounds, given sufficient solubility and dose, changes in microbial community structure on
23 leaf surfaces are possible. Because the fungi and bacteria living on and in the surfaces of leaves
24 play an important role in the microbial succession that prepares leaves for decay and litter
25 decomposition, such impacts could affect the larger ecosystem. Trace metal toxicity of lichens
26 has also been demonstrated in a few cases (CD, pp. 4-75, to 4-77).
27 On the other hand, the effects of chronic low-level metal deposition on perennial plant
28 species may be more significant than the acute effects referred to above. When trees are exposed
29 to sub-lethal concentrations of heavy metals, levels of intracellular metal-binding peptides,
August 2003 5-42 Draft - Do Not Quote or Cite
-------
1 phytochelatins, increase. In studies designed to test the relationship between heavy metals and
2 the decline of forest tree species in certain areas in the U.S., the data showed a systematic and
3 significant increase in phytochelatin concentrations associated with the extent of tree injury.
4 Though there has been no direct evidence of a physiological association between tree injury and
5 exposure to metals, metals have been implicated because their deposition pattern has been
6 correlated with the decline of certain tree species (CD, pp. 4-76, 4-77).
7 Organics. Many different chemical compounds can fall under the generic classification
8 of "organics." These compounds may also be referred to as toxic substances, pesticides,
9 hazardous air pollutants (HAPs), air toxics, semivolatile organic compounds (SOCs), and
10 persistent organic pollutants (POPs). While these substances are not criteria pollutants, they are
11 discussed here because many of them partition between gas and particle phases and are removed
12 from the atmosphere by both wet and dry deposition. As particles they can become airborne, be
13 distributed over wide areas, and impact remote ecosystems. Some notable organics include such
14 compounds as DDT, polychlorinated biphenyls (PCBs), and polynuclear aromatic hydrocarbons
15 (PAHs). These substances may enter plants via the roots, be deposited as particles onto the waxy
16 cuticle of leaves, or be taken up through the stomata. Which pathway is followed is a function of
17 the chemical and physical properties of the pollutant, environmental conditions, and the plant
18 species. However, the direct uptake of organic contaminants through the cuticle or in the vapor
19 phase through the stomates is poorly characterized for most trace organics. Additionally, the
20 toxicity of organic contaminants to plants and soil microorganisms is not well studied (CD, pp.
21 4-77 to 4-79).
22
23 5.4.4 Ecosystem Effects
24 As discussed in the draft CD, human existence on this planet depends on the life-support
25 services ecosystems provide. Both ecosystem structure (biotic condition) and functions
26 (ecological processes) play an essential role in providing ecosystem goods (products) and
27 services. Ecosystem processes maintain clean water, clean air, a vegetated Earth, and a balance
28 of organisms: the functions that enable humans to survive. The benefits they impart include
29 absorption and breakdown of pollutants, cycling of nutrients, binding of soil, degradation of
August 2003 5-43 Draft - Do Not Quote or Cite
-------
1 organic waste, maintenance of a balance of gases in the air, regulation of radiation balance and
2 climate, and the fixation of solar energy (CD, pp. 4-57, 4-58; Table 4-12).
3 Ecosystem-level responses related to PM occur when the effects of PM deposition on the
4 biological and physical components of ecosystems become sufficiently widespread as to impact
5 essential ecological attributes such as cycling of nutrients and materials or biodiversity. Such
6 responses can be a result of physical effects caused by high levels of PM dust being deposited
7 directly onto vegetative surfaces over a large portion of a plant community, or more importantly,
8 from chemical effects resulting from constituents of PM deposited directly onto vegetative
9 surfaces or acting indirectly through deposition into soil and/or aquatic environments.
10 At an experimental site near limestone quarries and processing plants in southwestern
11 Virginia, where PM dust accumulation occurred for at least 30 years, long-term changes in the
12 structure and composition of the various strata within plant communities (e.g., herbs, seedlings,
13 saplings, trees) were observed. Specifically, red maple was more abundant in all strata when
14 compared with the control site where it was present only as a seedling. The growth of tulip
15 poplar, dogwood, hop-hornbeam, black haw and red bud appeared to be favored by the dust,
16 while the growth of conifers and other acid tolerant species such as rhododendron, was limited.
17 It can be assumed that changes in soil alkalinity also occurred at the site due to the heavy
18 deposition of limestone dust, but in the absence of soil analyses, no conclusion was reached as to
19 the role that chemical changes to the soils may have played in these plant community changes.
20 This site exemplifies how the direct physical effects of chronic PM dust accumulation can
21 impact ecosystems by favoring the growth of some species and limiting others (CD, pp. 4-83 to
22 4-85).
23 The most significant PM-related ecosystem-level effects are the result of long-term
24 cumulative deposition of a given chemical species (e.g., nitrate) or mix (e.g., acidic deposition)
25 that exceeds the natural buffering or storage capacity of the ecosystem and/or affects the nutrient
26 status of the ecosystem, either by direct foliar uptake or by directly or indirectly changing soil
27 chemistry, populations of bacteria involved in nutrient cycling, and/or populations of fungi
28 involved in plant nutrient uptake (CD, pp. 4-90, 4-91). To understand these effects, long-term,
August 2003 5-44 Draft - Do Not Quote or Cite
-------
1 detailed ecosystem or site-specific data usually are required. The following discussion is
2 organized according to the speciated effects of PM.
3 Nitrogen Deposition. In the natural environment, nitrogen may be divided into two
4 groups: nonreactive, molecular (N2) and reactive (Nr). Though N2 is the most abundant element
5 in the atmosphere, it only becomes available to support the growth of plants and microorganisms
6 after it is converted into a reactive form. Reactive nitrogen includes inorganic reduced forms
7 (e.g., ammonia [NH3] and ammonium [NH4+]), inorganic oxidized forms (e.g., nitrogen oxides
8 [NOJ, nitric acid [HNO3], nitrous oxide [N2O], and nitrate [NO3"]), and organic compounds (e.g.,
9 urea, amine, proteins, and nucleic acids) (CD, pp. 4-95, 4-96).
10 Evidence shows that Nr is now accumulating in the environment on all spatial scales -
11 local, regional and global. The increase of global Nr is the result of three main causes: (1)
12 widespread cultivation of legumes, rice and other crops that promote the conversion of N2 to
13 organic nitrogen through biological nitrogen fixation; (2) combustion of fossil fuels, which
14 converts both atmospheric N2 and fossil nitrogen to reactive NOX; and (3) the Haber-Bosch
15 process, which converts N2 to Nr to sustain food production and some industrial activities. The
16 deposition of nitrogen in the U.S. from human activity doubled between 1961 and 1997 due
17 mainly to the use of inorganic nitrogen fertilizers and the emissions of NOX from fossil fuel
18 emissions with the largest increase occurring in the 1960s and 1970s (CD, pp. 4-96 to 4-98).
19 Despite the many beneficial effects of Nr on the health and welfare of humans (e.g., the
20 synthetic fertilizers used in cultivation and the cultivation-induced bacterial nitrogen fertilization
21 (BNF) sustain a large portion of the world's population), increased Nr in the global system also
22 contributes to many contemporary environmental problems such as:
23
24 • production of tropospheric ozone and aerosols and associated environmental (and human
25 health) problems;
26
27 • productivity increases in forests and grasslands, followed by decreases wherever
28 atmospheric Nr deposition increases significantly and critical thresholds are exceeded; Nr
29 additions probably also decrease biodiversity in many natural habitats;
30
31 • acidification and loss of biodiversity in lakes and streams in many regions of the world,
32 in association with sulfur deposition;
33
August 2003 5-45 Draft - Do Not Quote or Cite
-------
1 • eutrophication, hypoxia, loss of biodiversity, and habitat degradation in coastal
2 ecosystems, which is now considered the biggest pollution problem in coastal waters; and
3
4 • global climate change and stratospheric ozone depletion, both of which impact the health
5 of ecosystems (and humans) (CD, pp. 4-96 to 4-98).
6
7 Unfortunately, large uncertainties regarding the rates of Nr accumulation in the various
8 reservoirs (e.g., atmosphere, soil, water) limits our ability to determine the temporal and spatial
9 distribution of environmental effects. These uncertainties are of great significance because Nr
10 can alter a wide array of biogeochemical processes through exchanges among environmental
11 reservoirs. As depicted in Figure 5-35, this sequence of nitrogen transfers, transformations, and
12 environmental effects is referred to as the nitrogen cascade (CD, p. 4-96).
13 Nitrogen has long been recognized as the nutrient that is most often limiting (and
14 therefore most important) for plant growth. In soils low in nitrogen, atmospherically deposited
15 nitrogen can act as a fertilizer, though other secondary nutrients may then become limiting (e.g.,
16 phosphorus, calcium or magnesium). However, because not all plants are equally capable of
17 utilizing extra nitrogen, some plants will gain a competitive advantage and will replace those that
18 are better adapted to living in lower nitrogen environments, such as those growing in infertile
19 soil, shaded understories, deserts, or tundra. Plants adapted to these low resource environments
20 have been observed to have a slow growth rate, low photosynthetic rate, low capacity for
21 nutrient uptake (e.g., they tend to respond less than other plant species even when provided with
22 an optimal supply and balance of resources), and low soil microbial activity. The effect of
23 additions of nitrates on plant community succession patterns and biodiversity, has been studied
24 in several long-term nitrogen fertilization studies in both New England and Europe. These
25 studies suggest that some forests receiving chronic inputs of nitrogen may decline in productivity
26 and experience greater mortality. Specifically, the findings at Mount Ascutney, Vermont,
27 suggest that declining coniferous forest stands with slow nitrogen cycling may be replaced by
28 deciduous fast-growing forest species that cycle nitrogen rapidly (CD, pp. 4-95 to 4-105).
29 When additions of nitrogen above soil background levels exceed the capacity of plants
30 and soil microorganisms to utilize and retain it, a condition known as "nitrogen saturation"
August 2003 5-46 Draft - Do Not Quote or Cite
-------
1
2
3
NOX
Energy
Production
1
1
Ozi
Eff<
J
AAAAAAAAAAA
A
Atmosphere
Food
Production
A
A
A
NHX *
People
(Food; Fiber)
Human Activities
The Nitrogen
Cascade
Indicates denitrification potential
4
5
6 Figure 5-35 Illustration of the nitrogen cascade showing the movement of the human-
7 produced reactive nitrogen (Nr) as it cycles through the various
8 environmental reservoirs in the atmosphere, terrestrial and aquatic
9 ecosystems (Figure 4-15, CD p. 4-97).
August 2003
5-47
Draft - Do Not Quote or Cite
-------
1 results. Specific ecosystem effects associated with nitrogen saturation include: (1) changes in
2 plant uptake and allocation (i.e., a permanent increase in foliar nitrogen and reduced foliar
3 phosphorus and lignin due to the lower availability of carbon, phosphorus, and water); (2)
4 increased litter production; (3) increased ammonification (the release of ammonia) and trace gas
5 emissions; (4) decreased root biomass; (5) reduced soil fertility (the results of increased cation
6 leaching); (6) increased nitrification (conversion of ammonia to nitrate during decay of litter and
7 soil organic matter); and (7) nitrate leaching resulting in increased nitrate and aluminum
8 concentrations in streams and decreased water quality. Additionally, studies suggest that during
9 nitrogen saturation, soil microbial communities change from predominantly fungal (mycorrhizal)
10 communities to those dominated by bacteria (Aber et al., 1998; CD, pp. 4-98, 4-99).
11 Some U.S. forests are now showing severe symptoms of nitrogen saturation, including
12 the northern hardwoods and mixed conifer forests in the Adirondack and Catskill Mountains of
13 New York; the red spruce forests at Whitetop Mountain, Virginia, and Great Smoky Mountains
14 National Park, North Carolina; mixed hardwood watersheds at Fernow Experimental Forest in
15 West Virginia; American beech forests in Great Smoky Mountains National Park, Tennessee;
16 mixed conifer forests and chaparral watersheds in southern California and the southwestern
17 Sierra Nevada in Central California; the alpine tundra/subalpine conifer forests of the Colorado
18 Front Range; and red alder forests in the Cascade Mountains in Washington. All these systems
19 have been exposed to highly elevated nitrogen deposition, and nitrogen saturated watersheds
20 have been reported in the above-mentioned areas. Annual nitrogen additions through deposition
21 in the southwestern Sierra Nevada are similar in magnitude to nitrogen storage in vegetation
22 growth increments of western forests, suggesting that current nitrogen deposition rates may be
23 near the assimilation capacity of the overstory vegetation. Ongoing urban expansion will
24 increase the potential for nitrogen saturation of forests from urban sources (e.g., Salt Lake City,
25 Seattle, Tucson, Denver, central and southern California) unless there are improved emission
26 controls (Fenn et al., 1998).
27 Not all forest ecosystems react in the same manner to nitrogen deposition. The Harvard
28 Forest in Massachusetts was the site for a nitrogen amendment study over 8 years (1988-96).
29 During the study, nitrate leaching was observed in the pine stand after the first year in the high
August 2003 5-48 Draft - Do Not Quote or Cite
-------
1 nitrogen plots. Further increases in nitrate leaching were observed in the pine stand in 1995 and
2 1996, while the hardwood stand showed no significant increases in nitrate leaching until 1996.
3 This sharp contrast in the responses of pine and hardwood stands indicates that the mosaic of
4 community types across the landscape must be considered when determining regional scale
5 response to nitrogen deposition (Magill, et al., 2002). Johnson et al. (1991) reported that
6 measurements showing the leaching of nitrates and aluminum (Al+3) from high elevation forests
7 in the Great Smoky Mountains indicate that these forests have reached saturation (CD, p. 4-101).
8 Plant succession patterns and biodiversity are affected significantly by chronic nitrogen
9 additions in some North American ecosystems. In one example, experimental studies of nitrogen
10 deposition conducted over a 12-year period on Minnesota grasslands observed that plots
11 originally dominated by native warm-season grasses shifted to low-diversity mixtures dominated
12 by cool-season grasses at all but the lowest rates of nitrogen addition. Grasslands with high
13 nitrogen retention and carbon storage rates were the most vulnerable to loss of species and major
14 shifts in nitrogen cycling. The shift to low-diversity mixtures was associated with the decrease
15 in biomass carbon-to-nitrogen (C:N) ratios, increased nitrogen mineralization, increased soil
16 nitrate, high nitrogen losses, and low carbon storage (Weldin and Tilman, 1996). Naeem et al.
17 (1994) experimentally demonstrated under controlled environmental conditions that the loss of
18 biodiversity, genetic resources, productivity, ecosystem buffering against ecological
19 perturbation, and loss of aesthetic and commercially valuable resources also may alter or impair
20 ecosystems services (CD, p. 4-105).
21 The carbon-to-nitrogen (C:N) ratio of the forest floor can also be changed by nitrogen
22 deposition over time. This change appears to occur when the ecosystem becomes nitrogen
23 saturated. Long-term changes in C:N status have been documented in Central Europe and
24 indicate that nitrogen deposition has changed the forest floor. In Europe, low C:N ratios
25 coincide with high deposition regions. Therefore, to predict the rate of changes in nitrate
26 leaching, it is necessary to be able to predict the rate of changes in the forest floor C:N ratio.
27 Decreased foliar and soil nitrogen and soil C:N ratios, as well as changes in nitrogen
28 mineralization rates, have been observed when comparing responses to nitrogen deposition in
29 forest stands east and west of the Continental Divide in the Colorado Front Range.
August 2003 5-49 Draft - Do Not Quote or Cite
-------
1 Understanding the variability in forest ecosystem response to nitrogen input is essential in
2 assessing pollution risks (CD, p. 4-107).
3 The mutualistic relationship between plant roots, fungi, and microbes is critical for the
4 growth of the organisms involved, because the plant root zone (rhizosphere) is an important
5 region of nutrient dynamics. The plant roots provide shelter and carbon for the symbiont,
6 whereas the symbiont provides access to limiting nutrients such as nitrogen and phosphorus for
7 the plant. Thus, as indicated above, changes in soil nitrogen that influence below-ground
8 mycorrhizal fungal diversity also impact above-ground plant biodiversity, ecosystem variability,
9 and productivity (CD, pp. 4-107, 4-108).
10 These types of effects have been observed in the field; for example, iln studies of the
11 coastal sage scrub (CSS) community in southern California, which is composed of the drought-
12 deciduous shrub $ Artemisia californica, Encelia farinosa, andEriogonumfasciculatum. The
13 CSS in California has been declining in land area and in shrub density over the past 60 years and
14 is being replaced in many areas by Mediterranean annual grasses. At one time, nitrogen
15 deposition was considered as a possible cause, since the deposition rate is high in the Los
16 Angeles Air Basin (Bytnerowicz and Fenn, 1996). However, if increased nitrogen availability
17 were the only variable between the invasive annuals and native shrubs, neither shrubs nor
18 grasses would have a particular advantage. Later studies found that nitrogen enrichment of the
19 soils also induced a shift in the arbuscular mycorrhizal community composition. Larger-spored
20 fungal species (Scutellospora and Gigaspora), due to a failure to sporulate, decreased in number
21 with a concomitant proliferation of small-spored species of Glomus aggregatum, G. leptotichum,
22 and G. geosporum, indicating a strong selective pressure for the smaller spored species of fungi
23 (Edgerton-Warburton and Allen, 2000). These results demonstrate that nitrogen enrichment of
24 the soil significantly alters the arbuscular mycorrhizal species composition and richness, and
25 markedly decreases the overall diversity of the arbuscular mycorrhizal community. The decline
26 in coastal sage scrub species can, therefore, directly be linked to the decline of the arbuscular
27 mycorrhizal community (Edgerton-Warburton and Allen, 2000) (CD, pp. 4-108, 4-109).
28 Excessive nitrogen inputs to terrestrial ecosystems also affect aquatic ecosystems. The
29 primary pathways of nitrogen loss from terrestrial ecosystems are hydrological transport beyond
August 2003 5-50 Draft - Do Not Quote or Cite
-------
1 the rooting zone into groundwater or stream water, or surface flows of organic nitrogen as nitrate
2 and nitrogen loss associated with soil erosion. Nitrogen saturation of a high elevation watershed
3 in the southern Appalachian Mountains was observed to affect stream water chemistry. The
4 Great Smoky Mountains National Park in Tennessee and North Carolina receives high total
5 atmospheric deposition of sulfur and nitrogen. A major portion of the atmospheric loading is
6 from dry and cloud deposition. Nitrogen saturation of the watershed resulted in extremely high
7 exports of nitrate and promoted both chronic and episodic stream acidification in streams
8 draining undisturbed watersheds. Significant export of base cations was also observed.
9 Nitrification of the watershed soils resulted in elevations of soil solution aluminum
10 concentrations to levels known to inhibit calcium uptake in red spruce (CD, p. 4-110).
11 There has been a 3- to 8-fold increase in nitrogen flux from 10 watersheds in the
12 northeastern U. S. since the early 1900s. These increases are associated with nitrogen oxide
13 emissions from combustion which have increased 5-fold. Riverine nitrogen fluxes have been
14 correlated with atmospheric deposition onto their landscapes and also with nitrogen oxides
15 emissions into their airsheds. Data from 10 benchmark watersheds with good historical records
16 indicate that about 36-80% of the riverine total nitrogen export, with an average of 64%, was
17 derived directly or indirectly from nitrogen oxide emissions (CD, pp. 4-110, 4-111).
18 Direct atmospheric nitrogen deposition and increased nitrogen inputs via runoff into
19 streams, rivers, lakes, and oceans can noticeably affect aquatic ecosystems as well. Estuaries are
20 among the most intensely fertilized ecosystems on Earth, receiving far greater nutrient inputs
21 than other systems. In the Northeast, nitrogen is the element most responsible for eutrophication
22 in coastal waters of the region. An illustrative example is recently reported research (Paerl et al.,
23 2001) characterizing the effects of nitrogen deposition on the Pamlico Sound, NC, estuarine
24 complex, which serves as a key fisheries nursery supporting an estimated 80% of commercial
25 and recreational fmfish and shellfish catches in the southeastern U.S. Atlantic coastal region.
26 Such direct atmospheric nitrogen deposition onto waterways feeding into the Pamlico Sound or
27 onto the Sound itself and indirect nitrogen inputs via runoff from upstream watersheds contribute
28 to conditions of severe water oxygen depletion; formation of algae blooms in portions of the
29 Pamlico Sound estuarine complex; altered fish distributions, catches, and physiological states;
August 2003 5-51 Draft - Do Not Quote or Cite
-------
1 and increases in the incidence of disease. Under extreme conditions (e.g., hurricanes) the effects
2 of nitrogen runoff (in combination with excess loadings of metals or other nutrients) can be
3 massive, such as the creation of the widespread "dead-zone" affecting large areas of the estuary
4 (CD, pp. 4-109, 4-110).
5 The impact of increasing nitrogen inputs on the nitrogen cycle and forests, wetlands, and
6 aquatic ecosystems is discussed in detail elsewhere (EPA, 1993, 1997; Garner, 1994; World
7 Health Organization, 1997). Understanding the variability in forest ecosystem response to
8 nitrogen input is essential in assessing pollution-related impacts (CD, p. 4-98).
9 As noted above, sulfur is another essential plant nutrient, the most important source of
10 which for plants is sulfate taken up by the roots, even though plants can also utilize atmospheric
11 SO2. Atmospheric deposition of sulfate onto the soils, therefore, is an important component of
12 the sulfur cycle. The biochemical relationship between sulfur and nitrogen in plant proteins
13 indicates that neither element can be assessed adequately without reference to the other.
14 Nitrogen uptake in forests may be loosely regulated by sulfur availability, but sulfate additions in
15 excess of needs do not necessarily lead to injury (CD, pp. 4-112, 4-113).
16 Acidic Deposition. Acidic deposition over the past quarter of a century has emerged as a
17 critical environmental stress that affects forested landscapes and aquatic ecosystems in North
18 America, Europe, and Asia. Acidic deposition can originate from transboundary air pollution
19 and affect large geographic areas. It is composed of ions, gases, particles derived from gaseous
20 emissions of sulfur dioxide (SO2), nitrogen oxides (NOX), ammonia (NH3), and particulate
21 emissions of acidifying and neutralizing compounds, and is highly variable across space and
22 time. It links air pollution to diverse terrestrial and aquatic ecosystems (CD, p. 4-114).
23 Acidic deposition has been firmly implicated as a causal factor in the northeastern high-
24 elevation decline of red spruce (DeHayes et al., 1999). The frequency of freezing injury of red
25 spruce has increased over the past 40 years, a period that coincides with increase emissions of
26 sulfur and nitrogen oxides and increased acidic deposition. DeHayes et al. (1999) indicate that
27 there is a significant positive association between cold tolerance and foliar calcium in trees
28 exhibiting deficiency in foliar calcium, and further state that their studies raise the strong
29 possibility that acid rain alteration of foliar calcium is not unique to red spruce but has been
August 2003 5-52 Draft - Do Not Quote or Cite
-------
1 demonstrated in many other northern temperate forest tree species including yellow birch (Betula
2 alleghaniensis\ white spruce (Picea glaucus), red maple (Acer rubrum) eastern white pine
3 (Pinus strobus), and sugar maple (Acer saccharuni) (CD, pp. 4-120, 4-121).
4 Although forest ecosystems other than the high-elevation spruce-fir forests are not
5 currently manifesting symptoms of injury directly attributable to acid deposition, less sensitive
6 forests throughout the U.S. are experiencing gradual losses of base cation nutrients, which in
7 many cases will reduce the quality of forest nutrition over the long term (National Science and
8 Technology Council, 1998). In some cases, such effects may not even take decades to occur
9 because these forests have already been receiving sulfur and nitrogen deposition for many years.
10 Soil acidification and its effects result from the deposition of nitrate, sulfate and the associated
11 hydrogen ion. The introduction of the hydrogen ion (H+) will directly impact the fluxes of base
12 cations such as Ca, K, and Mg via cation exchange or weathering processes and influence the
13 availability of other elements (e.g., aluminum and mercury). Soil leaching is often of major
14 importance in cation cycles, and many forest ecosystems show a net loss of base cations (CD, pp.
15 4-115,4-117).
16 In aluminum-rich soils, acid deposition, by lowering the pH, can increase Al
17 concentrations in soil water through dissolution and ion-exchange processes. There is abundant
18 evidence that Al is toxic to plants, and it is believed that the toxic effect of Al on forest trees
19 could be due to its interference with Ca uptake. Once it enters forest tree roots, Al accumulates
20 in root tissue. Because Ca plays a major role in cell membrane integrity and cell wall structure,
21 reductions in Ca uptake suppress cambial growth, reduce the rate of wood formation, decrease
22 the amount of functional sapwood and live crown, and predispose trees to disease and injury
23 from stress agents when the functional sapwood becomes less than 25% of cross sectional stem
24 area. There are large variations in Al sensitivity among ecotypes, between and within species
25 due to differences in nutritional demands and physiological status, which are related to age and
26 climate, which change over time (CD, pp. 4-119, 4-120 and 4-126).
27 The Integrated Forest Study (IPS) (Johnson and Lindberg, 1992) has characterized the
28 complexity and variability of ecosystem responses to atmospheric inputs and provided the most
29 extensive data set available on the effects of atmospheric deposition, including particle
August 2003 5-53 Draft - Do Not Quote or Cite
-------
1 deposition, on the cycling of elements in forest ecosystems. For example, at the IFS sites on
2 average, particulate deposition contributes 47% to total Ca deposition (range: 4 to 88%), 49% of
3 total K deposition (range: 7 to 77%), 41% to total Mg deposition (range: 20 to 88%), 36% to
4 total sodium deposition (range: 11 to 63%), and 43% to total base cation deposition (range: 16 to
5 62%). Of the total particulate deposition, the vast majority (> 90%) is > 2 jim. Given the wide
6 ranges of particulate deposition for each base cation across the IPS sites, however, the unique
7 characteristics of various sites need to be better understood before assumptions are made about
8 the role particulate pollution plays in ecosystem impacts, though it is clear that particulate
9 deposition contributes considerably to the total impact of base cations to most of the IFS sites
10 (CD, pp. 4-127, 4-128).
11 These inputs of base cations have considerable significance, not only for base cation
12 status, but also for the potential of incoming precipitation to acidify or alkalize the soils in the
13 IFS ecosystems. The actual rates, directions, and magnitudes of changes that may occur in soils
14 (if any), however, will depend on rates of inputs from weathering and vegetation outputs, as well
15 as deposition and leaching. In other words, these net losses or gains of base cations must be
16 placed in the context of the existing soil pool size of exchangeable base cations. For example,
17 some sites identified as sensitive have large stores of weatherable minerals, while other soils,
18 with smaller stores of weatherable minerals but larger exchangeable cation reserves, are
19 considered less sensitive. In addition, atmospheric deposition may have significantly affected
20 the nutrient status of some IFS sites through the mobilization of Al. However, the connection
21 between Al mobilization and forest response is still not clear and warrants further study. The
22 IFS project further concluded that acidic deposition is having a significant, often overwhelming
23 effect on both nutrient cycling and cation leaching from the soils in most of the forest
24 ecosystems studied, though the nature of the effects varies from one location to another. It
25 appears that particle deposition has a greater effect on base cation inputs to soils than on base
26 cation losses associated with inputs of sulfur, nitrogen, and FT (CD, pp. 4-132 to 4-137).
27 The data collected in the IFS give some idea of the importance of particulate deposition
28 in these forest ecosystems, but they cannot account for the numerous potential feedbacks
29 between vegetation and soils, nor for the dynamics through time that can influence the ultimate
August 2003 5-54 Draft - Do Not Quote or Cite
-------
1 response. One way to examine some of these interactions and dynamics is to use simulation
2 modeling. The nutrient cycling model (NuCM) has been developed specifically for this purpose
3 and has been used to explore the effects of atmospheric deposition, fertilization, and harvesting
4 on some of the IPS sites (Johnson et al., 1993). The NuCM model is a stand-level model that
5 incorporates all major nutrient cycling processes (uptake, translocation, leaching, weathering,
6 organic matter decay, and accumulation) (CD, p. 4-137). Johnson et al. (1999) used the NuCM
7 model to simulate the effects of reduced S, N, and base cation (CB) deposition on nutrient pools,
8 fluxes, soil, and soil solution chemistry in two contrasting southern Appalachian forest
9 ecosystems: the extremely acidic red spruce sites and the less acidic Coweeta hardwood site
10 from the IPS project. In summary, the authors found that in an extremely acidic system, CB
11 deposition can have a major effect on CB leaching through time and S and N deposition had a
12 major effect on Al leaching. At the less acidic Coweeta site, CB deposition had a minor effect on
13 soils and soil solutions; whereas S and N deposition had delayed but major effects on CB
14 leaching (CD, pp. 4-137, 4-142).
15 Seasonal and episodic acidification of surface waters have been observed in the eastern
16 U. S., Canada and Europe. In the Northeast, the Shenandoah National Park in Virginia, and the
17 Great Smoky Mountains, episodic acidification has been associated with the nitrate ion. The
18 acidification of aquatic ecosystems and effects on aquatic biota are discussed in more detail in
19 the EPA documentor Quality Criteria for Nitrogen Oxides (EPA, 1993) (CD, p. 4-121).
20 Further, an extensive discussion of the various effects of acidic deposition is presented in the
21 U.S. National Acid Precipitation Assessment Program (NAPAP) Biennial Reports to Congress
22 (National Science and Technology Council, 1998).
23 Given the significant reductions in sulfur emissions that have occurred in the U. S. and
24 Europe in recent decades, the findings of Driscoll et al. (1989, 2001) and Hedin et al. (1994) are
25 especially relevant. Driscoll et al. (1989, 2001) noted a decline in both SO4"2 and base cations in
26 both atmospheric deposition and stream water over the past two decades at Hubbard Brook
27 Watershed, NH. However, the reductions in SO2 emissions in Europe and North America in
28 recent years have not been accompanied by equivalent declines in net acidity related to sulfate in
29 precipitation, and may have, to varying degrees, been offset by steep declines in atmospheric
August 2003 5-55 Draft - Do Not Quote or Cite
-------
1 base cation concentrations over the past 10 to 20 years (Hedin et al., 1994). Analysis of the data
2 from the IPS supports the authors' contention that atmospheric base cation inputs may seriously
3 affect ecosystem processes (CD, p. 4-122).
4 Critical Loads. The critical load has been defined as a "quantitative estimate of an
5 exposure to one or more pollutants below which significant harmful effects on specified sensitive
6 elements of the environment do not occur according to present knowledge" (Lokke et al., 1996).
7 The critical loads framework originated in Europe where the concept has generally been
8 accepted as the basis for abatement strategies to reduce or prevent injury to the functioning and
9 vitality of forest ecosystems caused by long-range transboundary chronic acidic deposition. The
10 concept is useful for estimating the amounts of pollutants that sensitive ecosystems can absorb
11 on a sustained basis without experiencing measurable degradation. The estimation of ecosystem
12 critical loads requires an understanding of how an ecosystem will respond to different loading
13 rates in the long term and is a direct function of the level of sensitivity of the ecosystem to the
14 pollutant and its capability to ameliorate change. In Europe, the elements used in the critical
15 load concept are a biological indicator, a chemical criterion, and a critical value (CD, p. 4-124).
16 In the U.S. in 1989, a program was designed to develop and evaluate a framework for
17 setting critical loads of nitrogen and sulfur. The flexible six-step approach includes:
18 (1) selection of ecosystem components, indicators, and characterization of the resource;
19 (2) definition of the functional subregions; (3) characterization of deposition within each of the
20 subregions; (4) definition of an assessment endpoint; (5) selection and application of models;
21 and (6) mapping projected ecosystem responses. This approach allows for consideration of
22 variability in ecosystem characteristics and data availability (Strickland et al., 1993; CD, p. 4-
23 124).
24 In the first step, ecological endpoints or indicators that are measurable characteristics
25 related to the structure, composition, or functioning of ecological systems (i.e., indicators of
26 condition) must be selected. A number of different indicators for monitoring ecosystem status
27 have been proposed. Biogeochemicals (e.g., foliar nitrogen; nutrient ratios (N:P, N:cation);
28 foliar nitrate; foliar 615 N; arginine concentration; soil C:N ratio; NO3" in soil extracts or in soil
29 solution; and flux rates of nitrogenous trace gasses from soil) have been discussed in Fenn et al.,
August 2003 5-56 Draft - Do Not Quote or Cite
-------
1 1998. Alternatively, a widely recognized indicator of nitrogen saturation in all ecosystem types,
2 including California forests and chaparral, is increased and prolonged NO3" loss below the main
3 rooting zone and in stream water. Seasonal patterns of stream water nitrate concentrations are
4 especially good indicators of watershed nitrogen status. Biological indicators that have been
5 suggested for use in the critical load calculation include mycorrhizal fungi (Lokke et al. 1996)
6 and fine roots, since they are an extremely dynamic component of below-ground ecosystems and
7 can respond rapidly to stress. The physiology of carbon allocation has also been suggested as an
8 indicator of anthropogenic stress (Andersen and Rygiewicz, 1991). It should be kept in mind,
9 however, that the response of a biological indicator is an integration of a number of different
10 stresses. Furthermore, there may be organisms more sensitive to the pollutant(s) than the species
11 selected (Lokke et al., 1996; National Science and Technology Council, 1998) (CD, pp. 4-124 to
12 126).
13 Assessment endpoints, on the other hand, selected in step (4) above, represent a formal
14 expression of the environmental value that is to be protected. If the assessment endpoint is to be
15 used as a regulatory limit, it should be socially relevant and the time scales of response
16 appropriate. With regard to the latter, for example, surface water acidification associated with
17 nitrate leaching should respond to decreases in nitrogen loading in a short period of time.
18 However, changes in growth responses of vegetation resulting from soil nutrient imbalances may
19 require years or decades to detect (CD, pp. 4-124, 4-125).
20 The critical loads concept has significant potential for identifying the level of protection
21 needed to achieve ecosystem sustainability for any specific, well-defined ecosystem over the
22 long-term. However, it is a very data-intensive approach, and, at the present time, there is
23 insufficient data for the vast majority of U.S. ecosystems to allow for practical application of this
24 approach as a basis for developing national standards to protect sensitive U.S. ecosystems from
25 adverse effects related to PM deposition.
26 Trace Elements. Some trace elements deposited directly onto vegetative surfaces can be
27 toxic to the populations of fungi and other microorganisms living on the leaves. Since these
28 organisms play an important role in leaf decomposition after litterfall, changes in these
29 communities can affect the rate of litter decomposition and subsequently nutrient availability for
August 2003 5-57 Draft - Do Not Quote or Cite
-------
1 vegetation (CD, p. 4-77). Alternatively, trace elements can be absorbed and bound in the leaf
2 tissue, which has also been shown to have a depressing effect on the rates of litter
3 decomposition. Heavy metals deposited from the atmosphere to forests accumulate either in the
4 top, richly organic layer of the forest floor or in the soil layers immediately beneath it, areas
5 where the activity of plant roots and soil organisms is greatest. Only toxicity from copper,
6 nickel, and zinc
7 has been frequently documented, though toxicity from cadmium, cobalt and lead compounds has
8 been seen under unusual situations. Since these metals can all be toxic to roots and soil
9 organisms, they change the litter decomposition processes that influence the availability of
10 essential soil nutrients, ultimately interfering with ecosystem nutrient cycling. Therefore, any
11 heavy metal effects on structure and function of an ecosystem are likely to occur through the soil
12 and litter. A number of toxic effects of metals on soil microbes have also been documented. For
13 example, cadmium was observed to decrease and prolong logarithmic rates of microbial
14 increase, to reduce microbial respiration and fungal spore formation and germination, to inhibit
15 bacterial transformation, and to induce abnormal morphologies. Additionally, it has been
16 suggested that the effect of heavy metals on mycorrhizae fungi can vary from host to host, and
17 in some cases, symbiotic associations of mycorrhizal fungi with certain plants provide some
18 additional degree of tolerance to metals (CD, pp. 4-145 to 4-147).
19 There is some evidence that invertebrates inhabiting soil litter do accumulate metals.
20 Earthworms from roadsides were shown to contain elevated concentrations of cadmium, nickel,
21 lead, and zinc, though interference with earthworm activity was not cited. A study of the
22 accumulation of these metals in earthworms suggested that cadmium and zinc were concentrated,
23 but not lead. Thus, organisms that feed on earthworms from soils with elevated concentrations
24 of certain metals (e.g. zinc) for extended periods would be expected to accumulate these metals
25 to toxic levels. Increased concentrations of heavy metals have been found in a variety of small
26 mammals living in areas with heavy metal concentrations in the soils. Biological accumulation
27 of metals through the plant-herbivore and litter-detritivore chains also can occur. When soils are
28 acidic, earthworm abundance decreases and bioaccumulation of metals from the soil may
29 increase exponentially with decreasing pH (CD, pp. 4-146, 4-147).
August 2003 5-58 Draft - Do Not Quote or Cite
-------
1 Organics. At the ecosystem level, some organic chemicals are of concern because they
2 may reach toxic levels in both animal and human food chains. Of particular ecological and
3 public concern are the polychlorinated hydrocarbons, such as the dioxins (e.g., polychlorinated
4 dibenzo-p-dioxins and dibenzofurans [PCDD/F]). As discussed above, wet and dry particle
5 deposition are the most important pathways for the accumulation of these more highly
6 chlorinated congeners in vegetation. Though not studied extensively in plants, biodegradation
7 probably does not occur readily since these compounds are found primarily in the lipophilic
8 cuticle and are very resistant to microbial degradation. Therefore, the grass-cattle-milk/beef
9 pathway can be a critical one for humans since exposure often comes from ingestion of animal
10 fat from fish, meat and dairy products. Alternatively, feed contaminated with soil containing the
11 pollutant can be another source of exposure for beef and dairy cattle as well as chickens.
12 Likewise in natural ecosystems, these chemicals tend to bioaccumulate up the food chain. For
13 example, PCDD/Fs have been found in arctic seals and polar bears (CD, pp. 4-87 to 4-89).
14 Section 112 of the CAA provides the legislative basis for U.S. hazardous air pollutant
15 (HAP) programs. In response to mounting evidence that air pollution contributes to water
16 pollution, Congress included Section 112m (Atmospheric Deposition to Great Lakes and Coastal
17 Waters) in the 1990 CAA Amendments that direct the EPA to establish a research program on
18 atmospheric deposition of HAPS to the "Great Waters." Actions taken by EPA and others to
19 evaluate and control sources of Great Waters pollutants of concern appear to have positively
20 affected trends in pollutant concentrations measured in air, sediment, and biota. Details
21 concerning these effects may be found in "Deposition of Air Pollutants to the Great Waters,"
22 Third Report to Congress (EPA, 2000). The Third Report (EPA-453/R-00-005, June 2000), like
23 the First and Second Reports to Congress, focuses on 15 pollutants of concern, including
24 pesticides, metal compounds, chlorinated organic compounds, and nitrogen compounds. The
25 new scientific information in the Third Report supports and builds on three broad conclusions
26 presented in the previous two EPA Reports to Congress:
27 (1) Atmospheric deposition from human activities can be a significant contributor of toxic
28 chemicals and nitrogen compounds to the Great Waters. The relative importance of atmospheric
29 loading for a particular chemical in a water body depends on many factors (e.g., characteristics
August 2003 5-59 Draft - Do Not Quote or Cite
-------
1 of the water body, properties of the chemical, and the kind and amount of atmospheric deposition
2 versus water discharges).
3 (2) A plausible link exists between emissions of toxic pollutants of concern into the air above
4 the Great Waters; the deposition of these pollutants (and their transformation products); and the
5 concentrations of these pollutants found in the water, sediments, and biota, especially fish and
6 shellfish. For mercury, fate and transport modeling and exposure assessments predict that the
7 anthropogenic contribution to the total amount of methylmercury in fish is, in part, the result of
8 anthropogenic mercury releases from industrial and combustion sources increasing mercury
9 body burdens (i.e., concentrations) in fish. Also, the consumption offish is the dominant
10 pathway of exposure to methylmercury for fish-consuming humans and wildlife. However, what
11 is known about each stage of this process varies with each pollutant (for instance, the chemical
12 species of the emissions and its transformation in the atmosphere).
13 (3) Airborne emissions from local as well as distant sources, from both within and outside
14 the United States, contribute pollutant loadings to waters through atmospheric deposition.
15 Determining the relative roles of particular sources — local, regional, national, and possibly
16 global, as well as anthropogenic, natural, and re-emission of pollutants — contributing to
17 specific water bodies is complex, requiring careful monitoring, atmospheric modeling, and other
18 analytical techniques (CD, pp. 4-89, 4-90).
19
20 5.4.5 Urban Ecosystems
21 Though humans now clearly influence virtually all ecosystems on the planet, nowhere
22 has human activity been more intense than in cities, suburbs, and surrounding areas. Although
23 the study of ecological phenomena in urban environments is not new, the concept of a city as an
24 ecosystem is relatively new for the field of ecology. Urban ecology implicitly recognizes the
25 role humans play in developing unique ecosystems. This new, integrative ecology, which now
26 explicitly incorporates human decisions, culture, institutions, and economic systems, along with
27 primary production, species richness, biogeophysical budgets, ecosystem patterns and processes,
28 and exchanges of materials and influence between cities and surrounding landscapes, is what is
August 2003 5-60 Draft - Do Not Quote or Cite
-------
1 needed if urban ecosystems, and ecosystems in general, are to be understood (CD, pp. 4-149, 4-150).
2 Though there is a large body of knowledge on concentrations and chemical reactions of
3 air pollutants in cities, there has been little work done on the rates of atmospheric deposition, and
4 especially particle deposition and effects, within urban ecosystems. Existing research indicates
5 that cities are often sources of nitrogen oxides, sulfur oxides, and dust, among many other
6 pollutants. Since some of these pollutants are major plant nutrients, they may be affecting
7 nutrient cycles in plant dominated areas in and around cities. Gases and particles in urban air
8 can increase atmospheric deposition within and downwind of cities. Lovett et al. (2000)
9 observed that concentrations and fluxes of NO3", NH4+, Ca+2, Mg+2, SO4"2, and Cl" in oak forest
10 throughfall all declined significantly with distance from New York City, while hydrogen ion
11 concentration and flux increased significantly with distance from the city. McDonnell et al.
12 (1997) found that levels of heavy metals in the foliar litter in urban forest soils were higher than
13 those in rural areas, with urban levels approaching or exceeding levels reported to affect soil
14 invertebrates, macrofungi, and soil microbial processes. The urban forests exhibited reduced
15 fungal biomass and arthropod densities when compared to rural stands. In addition to heavy
16 metal stress, urban forests also experience other anthropogenic stress such as poor air quality and
17 low water availability caused by hydrophobic soils. The potential affect of these conditions on
18 ecosystem processes of decomposition and nitrogen cycling appear to be ameliorated by two
19 other anthropogenic factors: increased average temperatures caused by the heat island effect and
20 the introduction and successful colonization of earthworms in the urban forests (McDonnell et
21 al., 1997). Another study (Pouyat and McDonnell, 1991) that looked at heavy metal
22 accumulations in forest soils in southeastern New York found variations in amounts of Zn, Cu,
23 Ni, and Cd that corresponded closely with the urban-rural land use gradient, with pollutants
24 highest near the urban end of the gradient (CD, pp. 4-150 to 4-152).
25 Though the role of particulate deposition in urban areas has not been explicitly studied, it
26 appears likely to be a significant contributor to the ecosystem effects that have been observed.
August 2003 5-61 Draft - Do Not Quote or Cite
-------
1 5.4.6 Rural PM Air Quality Networks
2 Atmospheric concentrations of dry particles began to be routinely measured in 1986, with
3 the establishment of EPA's National Dry Deposition Network (NDDN). After new monitoring
4 requirements were added in the 1990 Clean Air Act Amendments, EPA, in cooperation with the
5 National Oceanic and Atmospheric Association, created the Clean Air Status and Trends
6 Network (CASTNet) from the NDDN. CASTNet comprises 85 sites and is considered the
7 nation's primary source for atmospheric data to estimate concentrations for ground-level ozone
8 and the chemical species that make up the dry deposition component of total acid deposition
9 (e.g., sulfate, nitrate, ammonium, sulfur dioxide, and nitric acid), as well as the associated
10 meteorology and site characteristics data that are needed to model dry deposition velocities (CD,
11 pg. 4-21: www.epa.gov/castnet/overview/html).
12 To provide data on wet deposition levels in the U.S., the National Atmospheric
13 Deposition Program (NADP) was initiated in the late 1970's as a cooperative program between
14 federal, state, and other public and private groups. By the mid-1980's, it had grown to nearly
15 200 sites and it stands today as the longest running national atmospheric deposition monitoring
16 network (www.epa.gov/castnet/overview/html).
17 In addition to these deposition monitoring networks, other networks collect data on
18 ambient aerosol concentrations and chemical composition. Such networks include the
19 IMPROVE network, discussed above in sections 2.5 and 5.2.3, and the newly implemented PM25
20 chemical Speciation Trends Network (STN) that consists of 54 core National Ambient
21 Monitoring Stations and approximately 250 State and Local Air Monitoring Stations (CD, p. 2-
22 82).
23 In the future, data from these networks, combined with data from the deposition
24 monitoring networks may allow better understanding of the components of PM that are most
25 strongly influencing PM-related ecological effects.
26
27 5.4.7 Summary
28 The draft CD presents evidence of effects on vegetation and ecosystems from ambient
29 PM, both in the U.S. and Europe, including in particular effects related to the cascade of Nr in
August 2003 5-62 Draft - Do Not Quote or Cite
-------
1 terrestrial and aquatic ecosystems and those related to nitrate and acidic deposition. Among the
2 observed effects of chronic nitrate deposition and its accumulation in ecosystems are nitrogen
3 saturation, altering of ecosystem structure (i.e., biodiversity), and the functioning of ecosystem
4 processes (e.g., nitrogen cycling). Though nitrogen deposited to ecosystems has not in general
5 been speciated and measured, there is abundant evidence that nitrogen emissions from fossil-fuel
6 burning deposited as nitrate over time has, as cited in the CD and summarized above, affected
7 the functioning of ecosystems. At this time, however, there remains a paucity of widespread,
8 long-term speciated rural PM data (needed for understanding ecosystem level effects); a lack of
9 studies relating unspeciated ambient PM mass/size fraction exposures to vegetation or ecosystem
10 effects; major complexities and regional and meteorological variability associated with the
11 atmospheric deposition processes;and a lack of detailed location-specific environmental data that
12 would be needed to determine whether a given amount of PM deposition occurring in a given
13 location represents a beneficial or an adverse effect. As a result, currently available information
14 does not yet provide a basis for quantitative characterization of the complex relationships
15 between observed PM-related adverse effects on vegetation and ecosystems in various locations
16 across the U.S. and levels of PM in the ambient air. Thus, while evidence of PM-related effects
17 clearly exists, there is insufficient information available at this time to serve as a basis for a
18 secondary national air quality standard for PM, specifically selected to protect against adverse
19 effects on vegetation and ecosystems.
20
21 5.5 EFFECTS ON CLIMATE CHANGE AND SOLAR RADIATION
22 Atmospheric particles alter the amount of solar radiation transmitted through the earth's
23 atmosphere by both scattering and absorbing radiation. More specifically, most components of
24 ambient PM (especially sulfates) scatter and reflect incoming solar radiation back into space,
25 thus offsetting the "greenhouse effect" to some degree by having a cooling effect on climate. In
26 contrast, some components of ambient PM (especially black carbon) also absorb incoming solar
27 radiation and are believed to contribute to some degree to atmospheric warming. Lesser impacts
28 of atmospheric particles are associated with their role in altering the amount of ultraviolet solar
29 radiation (especially UV-B) penetrating through the earth's atmosphere to ground level, where it
August 2003 5-63 Draft - Do Not Quote or Cite
-------
1 can exert a variety of effects on human health, plant and animal biota, and other environmental
2 components (CD, p. 205). The extensive research and assessment efforts into global climate
3 change and stratospheric ozone depletion provide evidence that atmospheric particles play
4 important roles in these two types of atmospheric processes, not only on a global scale, but also
5 on regional and local scales as well.
6 Information on the role of atmospheric particles in these atmospheric processes is briefly
7 summarized above in Chapter 2 (section 2.9). The effects on human health and the environment
8 associated with these atmospheric processes are briefly summarized below, based on the
9 information in section 4.5 of the draft CD and referenced reports. These effects are discussed
10 below in conjunction with consideration of the potential indirect impacts on human health and
11 the environment that may be a consequence of climatic and radiative changes attributable to
12 local and regional changes in ambient PM.
13
14 5.5.1 Climate Change and Potential Human Health and Environmental Impacts
15 Potential effects of global warming and climate change on both the environment and
16 human health in the U.S. are discussed in section 4.5.1.1 in the draft CD and related references.
17 The draft CD (p. 4-209) notes that while current climate models are successful in simulating
18 present annual mean climate and the seasonal cycle on continental scales, they are lass
19 successful at regional scales. Findings from various referenced assessments illustrate well the
20 considerable uncertainties and difficulties in projecting likely climate change impacts on
21 regional or local scales. For example, uncertainties in calculating the direct radiative effects of
22 atmospheric particles arise from a lack of knowledge of their vertical and horizontal variability,
23 their size distribution, chemical composition, and the distribution of components within
24 individual particles. Any complete assessment of the radiative effects of PM would require
25 computationally intensive calculations that incorporate the spatial and temporal behavior of
26 particles of varying composition that have been emitted from, or formed by precursors emitted
27 from, different sources. In addition, the draft CD (p. 219) notes that calculations of indirect
28 physical effects of particles on climate (e.g., related to alteration of cloud properties and
29 disruption of hydrological cycles, as discussed above in section 2.9) are subject to much larger
August 2003 5-64 Draft - Do Not Quote or Cite
-------
1 uncertainties than those related to the direct radiative effects of particles. The draft CD
2 concludes that at the present time impacts on human health and the environment due to aerosol
3 effects on the climate system can not be calculated with confidence, and notes that the
4 uncertainties associated with such aero sol-related effects will likely remain much larger than
5 those associated with greenhouse gases (CD, p. 4-219). Nevertheless, the draft CD concludes
6 that substantial qualitative information available from observational and modeling studies
7 indicates that different types of atmospheric aerosols (i.e., different components of PM) have
8 both warming and cooling effects on climate, both globally and regionally. Studies also suggest
9 that global and regional climate changes could potentially have both positive and negative
10 effects on human health, human welfare, and the environment.
11
12 5.5.2 Alterations in Solar UV-B Radiation and Potential Human Health and
13 Environmental Impacts
14 This section briefly summarizes information in section 4.5.2 of the draft CD on the health
15 and environmental effects associated with UV-B radiation exposure, and considers the potential
16 impacts that may result from changes in UV-B radiation penetration to the earth's surface
17 attributable to changes in ambient PM. The main types of effects associated with exposure to
18 UV-B radiation include direct effects on human health and agricultural and ecological systems,
19 indirect effects on human health and ecosystems, and effects on materials. The study of these
20 effects has been driven by international concern over potentially serious increases in the amount
21 of solar UV-B radiation reaching the earth's surface due to the depletion of the stratospheric
22 ozone layer by the release of various man-made ozone-depleting substances. Extensive
23 qualitative and quantitative characterizations of these global effects attributable to projections of
24 stratospheric ozone depletion have been periodically assessed in studies carried out under WMO
25 and UNEP auspices, with the most recent projections being published in UNEP (1998, 2000) and
26 WMO (1999).
27 Direct human health effects of UV-B radiation exposure include: skin damage (sunburn)
28 leading to more rapid aging and increased incidence of skin cancer; effects on the eyes, including
29 retinal damage and increased cataract formation possibly leading to blindness; and suppression
30 of some immune system components, contributing to skin cancer induction and possibly
August 2003 5-65 Draft - Do Not Quote or Cite
-------
1 increasing susceptibility to certain infectious diseases. Direct environmental effects include
2 damage to terrestrial plants, leading to possible reduced yields of some major food crops and
3 commercially important tress, as well as to biodiversity shifts in natural terrestrial ecosystems;
4 and adverse effects on aquatic life, including reductions in important components of marine food
5 chains as well as other aquatic ecosystem shifts. Indirect health and environmental effects are
6 primarily those mediated through increased tropospheric ozone formation and consequent
7 ground-level ozone-related health and environmental impacts. Effects on materials include
8 accelerated polymer weathering and other effects on man-made materials and cultural artifacts.
9 In addition, there are emerging complex issues regarding interactions and feedbacks between
10 climate change and changes in terrestrial and marine biogeochemical cycles due to increased
11 UV-B radiation penetration. (CD, p. 4-221, 4-222).
12 In contrast to these types of negative impacts associated with increased UV-B penetration
13 to the Earth's surface, the draft CD (p. 4-222, 4-223) summarizes research results that are
14 suggestive of possible beneficial effects of increased UV-B radiation penetration. For example,
15 a number of studies have focused on the protective effects of UV-B radiation with regard to non-
16 skin cancer incidence, which proved suggestive evidence that UV-B radiation, acting through the
17 production of vitamin D, may be a risk-reduction factor for mortality due to several types of
18 cancer, including cancer of the breast, colon, ovary, and prostate, as well as non-Hodgkin
19 lymphoma.
20 The various assessments of these types of effects that have been conducted consistently
21 note that the modeled projections quantitatively relating changes in UV-B radiation (attributable
22 to stratospheric ozone depletion) to changes in health and environmental effects are subject to
23 considerable uncertainty, with the role of atmospheric particles being one of numerous
24 complicating factors. Taking into account the complex interactions between ambient particles
25 and UV-B radiation transmission through the lower atmosphere, the draft CD concludes that any
26 effort to quantify projected indirect effects of variations in atmospheric PM on human health or
27 the environment due to particle impacts on transmission of solar UV-B radiation would require
28 location-specific evaluations that take into account the composition, concentration, and internal
29 structure of the particles; temporal variations in atmospheric mixing heights and depths of layers
August 2003 5-66 Draft - Do Not Quote or Cite
-------
1 containing the particles; and the abundance of ozone and other absorbers within the planetary
2 boundary layer and the free troposphere (CD, page 4-227).
3 At present, models are not available to take such complex factors into account, nor is
4 sufficient data available to characterize input variables that would be necessary for any such
5 modeling. The draft CD concludes, however, that the outcome of such modeling efforts would
6 likely vary from location to location, even as to the direction of changes in the levels of
7 exposures to UV-B radiation, due to location-specific changes in ambient PM concentrations
8 and/or composition (CD, p. 4-227). Beyond considering just average levels of exposures to UV-
9 B radiation in general, the draft CD notes that ambient PM can affect the directional
10 characteristics of UV-B radiation scattering at ground-level, and thus its biological effectiveness.
11 Also, ambient PM can affect not only biologically damaging UV-B radiation, but can also reduce
12 the ground-level ratio of photorepairing UV-A radiation to damaging UV-B radiation. Further,
13 the draft CD notes that ambient PM deposition is a major source of PAH in certain water bodies,
14 which can enhance the adverse effects of solar UV-B radiation on aquatic organisms, such that
15 the net effect of ambient PM in some locations may be to increase UV-B radiation-related
16 biological damage to certain aquatic and terrestrial organisms. (CD, p. 4-227).
17
18 5.5.3 Summary
19 A number of assessments of the factors affecting global warming and climate change as
20 well as those affecting the penetration of solar UV-B radiation to the earth's surface clearly
21 recognize ambient PM as playing various roles in these processes. These assessments, however,
22 have focused on global- and regional-scale impacts, allowing for generalized assumptions to take
23 the place of specific, but unavailable, information on local-scale atmospheric parameters and
24 characteristics of the distribution of particles present in the ambient air. As such, the available
25 information provides no basis for estimating how localized changes in the temporal, spatial, and
26 composition patterns of ambient PM, likely to occur as a result of expected future emissions of
27 particles and their precursor gases across the U.S., would affect local, regional, or global changes
28 in climate or UV-B radiation penetration - even the direction of such effects on a local scale
29 remains uncertain. Moreover, similar concentrations of different particle components can
August 2003 5-67 Draft - Do Not Quote or Cite
-------
1 produce opposite net effects. It follows, therefore, that there is insufficient information available
2 to project the extent to which, or even whether, such location-specific changes in ambient PM
3 would indirectly affect human health or the environment secondary to potential changes in
4 climate and UV-B radiation.
5 Based on currently available information, the potential indirect effects of ambient PM on
6 public health and welfare, secondary to potential PM-related changes in climate and UV-B
7 radiation, can play no quantitative role in considering whether any revisions of the primary or
8 secondary PM NAAQS are appropriate at this time. Even qualitatively, the available
9 information is very limited in the extent to which it can help inform an assessment of the overall
10 weight of evidence in an assessment of the net health and environmental effects of PM in the
11 ambient air, considering both its direct effects (e.g., inhalation-related health effects) and indirect
12 effects mediated by other routes of exposure and environmental factors (e.g., dermal exposure to
13 UV-B radiation).
August 2003 5-68 Draft - Do Not Quote or Cite
-------
1 REFERENCES
2
3 Section 5.2
4
5 Abt Associates, Inc. (2001) Assessing Public Opinions on Visibility Impairment Due to Air Pollution: Summary
6 Report. Prepared for EPA Office of Air Quality Planning and Standards; funded under EPA Contract No.
7 68-D-98-001. Bethesda, Maryland. January 2001.
8
9 Air Resource Specialists, Inc. (2003) WinHaze Air Quality Modeler, version 2.9.0. Available from
10 http://www.air-resource.com/whatsnew.htm
11
12 Arizona Department of Environmental Quality. (2001) Governor's Brown Cloud Summit: Final Report. January 16,
13 2001. http://www.adeq.state.az.us/environ/air/browncloud/ffinal
14
15 California Code of Regulations. Title 17, Section 70200, Table of Standards.
16
17 Chestnut, L. G.; Rowe, R. D. (1990) Preservation of values for visibility in the national parks. Washington, DC:
18 U.S. Environmental Protection Agency.
19
20 Chestnut, L.G.; Dennis, R. L.; Latimer, D. A. (1994) Economic benefits of improvements invisibility: acid rain
21 provisions of the 1990 clean air act amendments. Proceedings of Aerosols and Atmospheric Optics:
22 Radiative Balance and Visual Air Quality. Air & Waste Management Association International Specialty
23 Conference, pp. 791-802.
24
25 Chestnut, L. G.; Dennis, R. L. (1997) Economic benefits of improvements invisibility: acid rain. Provisions of the
26 1990 clean air act amendments. J. Air Waste Manage. Assoc. 47:395-402.
27
28 Cohen, S.; Evans, G.W.; Stokols, D.; Krantz, D.S. (1986) Behavior, Health, and Environmental Stress. Plenum
29 Press. New York, NY.
30
31 Departmentof Interior. (1998) Air Quality in the National Parks. Natural Resources Report 98-1. National Park
32 Service, Air Quality Division. Denver, Colorado.
33
34 Ely, D.W.; Leary, J.T.; Stewart, T.R.; Ross, D.M. (1991) The Establishment of the Denver Visibility Standard. For
3 5 presentation at the 84th Annual Meeting & Exhibition of the Air and Waste Management Association, June
36 16-21, 1991.
37
3 8 Environmental Protection Agency. (1979) Protecting Visibility: An EPA Report to Congress. Research Triangle
3 9 Park, NC: Office of Air Quality Planning and Standards. Report no. EPA-45-/5-79-008.
40
41 Environmental Protection Agency. (1982) Review of the National Ambient Air Quality Standards for Paniculate
42 Matter, Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park,
43 N.C.: Office of Air Quality Planning and Standards, Strategies and Air Standards Division. Report no.
44 EPA-450/5-82-001.
45
46 Environmental Protection Agency. (1996a) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
47 National Center for Environmental Assessment-RTF Office; report no. EPA/600/P-95/001aF-cF. 3v.
48
49 Environmental Protection Agency. (1996b) Review of the National Ambient Air Quality Standards for Paniculate
50 Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research
51 Triangle Park, NC 27711: Office of Air Quality Planning and Standards; report no. EPA-452VR-96-013.
52
August 2003 5-69 Draft - Do Not Quote or Cite
-------
1 Environmental Protection Agency. (1999) Regional Haze Regulations. 40 CFR Part 51.300-309. 64 Federal
2 Register 35713.
3
4 Environmental Protection Agency. (2000) Guidelines for Preparing Economic Analyses. Washington, DC: Office of
5 the Administrator. EPA240-R-00-003.
6
7 Environmental Protection Agency. (200 la) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
8 Office of Research and Development; report no. EPA/600/P-99/002. March.
9
10 Environmental Protection Agency. (200Ib) National Air Quality and Emissions Trends Report, 1999. Research
11 Triangle Park, NC: Office of Air Quality Planning and Standards. Report no. EPA/454/R-01-004. March.
12
13 Environmental Protection Agency. (2003) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
14 National Center for Environmental Assessment-RTF Office; report no. EPA/600/P-99/002aD.
15
16 Grand Canyon Visibility Transport Commission (1996) Report of the Grand Canyon Visibility Transport
17 Commission to the United States Environmental Protection Agency.
18
19 Hass, G. E.; Wakefield, T. J. (1998) National Parks and the American Public: A National Public Opinion Survey of
20 the National Park System. Colorado State University, Department of Natural Resource Recreation and
21 Tourism, College of Natural Resources, Fort Collins, CO. Report prepared for the National Parks and
22 Conservation Association. June 1998.
23
24 Hill, B.; Harper,W.; Halstead, J.; Stevens, T.H.; Porras, I.; Kimball, K.D. (2000) "Visitor Perceptions and Valuation
25 of Visibility in the Great Gulf Wilderness, New Hampshire" in Cole, etal. Proceedings: Wilderness
26 Science in a Time of Change. Proc., RMRS-P-000. Ogden, VT: U.S.D. A. Forest Service, Rocky
27 Mountain Research Station.
28
29 Malm, W.C.; Sisler, J. F.; Pitchford, M.; Scruggs, M.; Ames, R.; Copeland, S.; Gebhart, K.; Day, D. (2000) Spatial
30 and Seasonal Patterns and Temporal Variability of Haze and Its Constituents in the United States: Report
31 III. Colorado State University, Cooperative Institute for Research in the Atmosphere. Fort Collins, CO.
32
33 Malm, W.C. (1999) Introduction to Visibility. Cooperative Institute for Research in the Atmosphere (CIRA),
34 Colorado State University. Fort Collins, CO.
35
36 Malm, W. C.; Kelley, K.; Molenar, J.; Daniel, T. (1981) Human Perception of Visual Air Quality (Uniform Haze).
37 Atmospheric Environment. Volume 15, Issue 10/11. 1875-1890.
38
39 Malm, W.C.; Sisler, J. F.; Pitchford, M.; Scruggs, M.; Ames, R.; Copeland, S.; Gebhart, K.; Day, D. (2000) Spatial
40 and Seasonal Patterns and Temporal Variability of Haze and Its Constituents in the United States: Report
41 III. Colorado State University, Cooperative Institute for Research in the Atmosphere. Fort Collins, CO.
42
43 McNeill, R. and Roberge, A. (2000) The Impact of Visual Air Quality on Tourism Revenues in Greater Vancouver
44 and the Lower Fraser Valley. Environment Canada, Georgia Basin Ecosystem Initiative. GBEI report no.
45 EC/GB-00-028.
46
47 Middleton, P. (1993) Brown Cloud II: The Denver Air Quality Modeling Study, Final Summary Report. Metro
48 Denver Brown Cloud Study, Inc. Denver, CO.
49
50 Molenar, J.V.; Malm, W.C.; Johnson, C.E.(1994) Visual Air Quality Simulation Techniques. Atmospheric
51 Environment. Volume 28, Issue 5, 1055-1063.
52
August 2003 5-70 Draft - Do Not Quote or Cite
-------
1 Molenar, John V. (2000) Visibility Science and Trends in the Lake Tahoe Basin: 1989-1998. Report by Air
2 Resource Specialists, Inc., to Tahoe Regional Planning Agency. February 15, 2000.
3
4 National Acid Precipitation Assessment Program. (1991) Acid Deposition: State of Science and Technology. Report
5 24. Visibility: Existing and Historical Conditions-Causes and Effects. Washington, DC.
6
7 National Research Council. (1993) Protecting Visibility in National Parks and Wilderness Areas. National
8 Academy of Sciences Committee on Haze in National Parks and Wilderness Areas. National Academy
9 Press: Washington, DC.
10
11 National Transportation Safety Board (NTSB). (2000) NTSB Report NYC99MA178, July 6, 2000. Report on July
12 16, 1999 fatal accident at Vineyard Haven, MA.
13
14 National Weather Service. (1998) Automated Surface Observing System (ASOS) User's Guide. ASOS Program
15 Office. Silver Spring, MD.
16
17 New Zealand Ministry for the Environment. (2000) Proposals for Revised and New
18 Ambient Air Quality Guidelines: Discussion Document. Air Quality Report No. 16. December.
19
20 New Zealand National Institute of Water & Atmospheric Research (NIWAR). (2000a) Visibility in New Zealand:
21 Amenity Value, Monitoring, Management and Potential Indicators. Air Quality Technical Report 17.
22 Prepared for New Zealand Ministry for the Environment. Draft report.
23
24 New Zealand National Institute of Water & Atmospheric Research (NIWAR). (2000b) Visibility in New Zealand:
25 National Risk Assessment. Air Quality Technical Report 18. Prepared for New Zealand Ministry for the
26 Environment. Draft report.
27
28 Peacock, B.; Killingsworth, C.; Simon, B. (1998) State and National Economic Impacts Associated with Travel
29 Related Expenditures by Recreational Visitors to Lands Managed by the U.S. Department of Interior. U.S.
30 Department of the Interior. January.
31
32 Pitchford, M.; Malm, W. (1994) Development and Applications of a Standard Visual Index. Atmospheric
33 Environment. Vol. 28, no. 5, pp. 1049-1054.
34
35 Pryor, S.C. (1996) Assessing Public Perception of Visibility for Standard Setting Exercises. Atmospheric
36 Environment, vol. 30, no. 15, p. 2705-2716.
37
38 Schichtel, B.A., Husar, R.B., Husar, J. B., Falke, S. R., and Wilson, W.E. (2001) "Haze Trends of the United
39 States, 1980-1995," Atmospheric Environment (inpublication).
40
41 Schulze, W. D.; Brookshire, D. S.; Walther, E. G.; MacFarland, K. K.; Thayer, M. A.; Whitworth, R. L.; Ben-Davis,
42 S.; Malm, W.; Molenar, Jr. (1983) The Economic Benefits of Preserving Visibility in the National
43 Parklands of the Southwest. Nat. Resour. J. 23: 149-173.
44
45 Sisler, J.; Malm, W.; Molenar, J.; Gebhardt, K. (1996) Spatial and Seasonal Patterns and Long Term Variability of
46 the Chemical Composition of Haze in the U.S.: An Analysis of Data from the IMPROVE Network. Fort
47 Collins, CO: Cooperative Institute for Research in the Atmosphere, Colorado State University.
48
49 State Government of Victoria, Australia. (2000a) Draft Variation to State Environment Protection Policy (Air
50 Quality Management) and State Environment Protection Policy (Ambient Air Quality) and Draft Policy
51 Impact Assessment. Environment Protection Authority. Publication 728. Southbank, Victoria.
52
August 2003 5-71 Draft - Do Not Quote or Cite
-------
1 State Government of Victoria, Australia. (2000b) Year in Review. Environment Protection Authority. Southbank,
2 Victoria.
3
4 Stewart, T. R.; Middleton, P.; Ely, D. (1983) Urban Visual Air Quality Judgements: Reliability and Validity.
5 Journal of Environmental Psycho logy. Volume 3, 129.
6
7 Szykman, J.; Damberg, R. (2003) Analysis of Automated Surface Observation System (ASOS) Visibility Sensor
8 Data to Evaluate Correlation With Nearby PM2 5 Monitors. Memorandum to the docket for the PM Staff
9 Paper.
10
11 Section 5.3
12
13 Creighton, P. J.; Lioy, P. J.; Haynie, F. H.; Lemmons, T. J.; Miller, J. L.; Gerhart, J. (1990) Soiling by atmospheric
14 aerosols in an urban industrial area. J. Air Waste Manage. Assoc. 40: 1285-1289.
15
16 Davidson, C. L; Tang, W.; Finger, S.; Etyemezian, V.; Striegel, M. F.; Sherwood, S. I. (2000) Soiling patterns on a
17 tall limestone building: changes over 60 years. Environ. Sci. Technol. 34: 560-565.
18
19 Environmental Protection Agency. (1996b) Review of the National Ambient Air Quality Standards for Paniculate
20 Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research
21 Triangle Park, NC 27711: Office of Air Quality Planning and Standards; report no. EPA-452VR-96-013.
22
23 Environmental Protection Agency. (2001) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC:
24 Office of Research and Development; report no. EPA/600/P-99/002. March.
25
26 Haynie, F.H.; Lemmons, T. J. (1990) Particulate matter soiling of exterior paints at a rural site. Aerosol Sci. Technol.
27 13:356-367.
28
29 Ligocki, M. P.: Salmon, L. G.; Fall, T.; Jones, M. C.; Nazaroff, W. W.; Cass, G. R. (1993) Characteristics of
30 airborne particles inside southern California museums. Atmos. Environ. Part A 27: 697-711.
31
32 Lorusso, S.; Marabelli, M.; Troili, M. (1997) Air pollution and the deterioration of historic monuments. J. Environ.
33 Pathol. Toxicol. Oncol. 16: 171-173.
34
35 Miller, W. C.; Fornes, R. E.; Gilbert, R. D.; Speer, A.; Spence, J. (1992) Removal of CaCO3 extender in residential
36 coatings by atmospheric acidic deposition. In: Measurement of toxic and related air pollutants: proceedings
37 of the 1992 U. S. EPA/A&WMA international symposium. Pittsburgh, PA: Air & Waste Management
3 8 Association; pp. 129-134. (A&WMA publication VTP-25)
39
40 Wolff, G. T.; Collins, D. C.; Rodgers, W. R.; Verma, M. H.; Wong, C. A. (1990) Spotting of automotive finishes
41 from the interactions between dry deposition of crustal material and wet deposition of sulfate. J. Air Waste
42 Manage. Assoc. 40: 1638-1648.
43
44 Section 5.4
45
46 Aber, J. D.; Nadelhoffer, K. J.; Steudler, P.; Melillo, J. M. (1989) Nitrogen saturation in northern forest ecosystems:
47 excess nitrogen from fossil fuel combustion may stress the biosphere. Bioscience 39: 378-386.
48
49 Aber, J.; McDowell, W.; Nadelhoffer, K.; Magill, A.; Berntson, G.; Kamakea, M.; McNulty, S.; Currie, W.;
50 Rustad, L.; Fernandez, I. (1998) Nitrogen saturation in temperate forest ecosystems. BioScience 48:
51 921-934.
52
August 2003 5-72 Draft - Do Not Quote or Cite
-------
1 Andersen, C. P.; Rygiewicz, P. T. (1991) Stress interactions and mycorrhizal plant response: understanding carbon
2 allocation priorities. Environ. Pollut. 73: 217-244.
3
4 Bytnerowicz, A.; Fenn, M. E. (1996) Nitrogen deposition in California forests: a review. Environ. Pollut.
5 92: 127-146.
6
7 DeHayes, D. H.; Schaberg, P. G.; Hawley, G. I; Strimbeck, G. R. (1999) Acid rain impacts on calcium nutrition and
8 forest health. Bioscience 49: 789-800.
9
10 Driscoll, C. T.; Wyskowski, B. I; DeStaffan, P.; Newton, R. M. (1989) Chemistry and transfer of aluminum in a
11 forested watershed in the Adirondack region of New York, USA. In: Lewis, T. E., ed. Environmental
12 chemistry and toxicology of aluminum. Chelsea, MI: Lewis Publishers, Inc.; pp. 83-105.
13
14 Driscoll, C. T.; Lawrence, G. B.; Bulger, A. I; Butler, T. I; Cronan, C. S.; Eagar, C.; Lambert, K. F.; Likens, G. E.;
15 Stoddard, J. L.; Weathers, K. C. (2001) Acidic deposition in the northeastern United States: sources and
16 inputs, ecosystem effects, and management strategies. BioScience 51: 180-198.
17
18 Edgerton-Warburton, L. M.; Allen, E. B. (2000) Shifts in arbuscular mycorrhizal communities along an
19 anthropogenic gradient nitrogen deposition gradient. Ecol. Appl. 10: 484-496.
20
21 Environmental Protection Agency. (1982) Air quality criteria for paniculate matter and sulfur oxides. Research
22 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
23 Assessment Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA;
24 PB84-156777.
25
26 Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park, NC:
27 Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
28 nos. EPA/600/8-9!/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525,
29 andPB95-124517.
30
31 Environmental Protection Agency. (1997) Nitrogen oxides: impacts on public health and the environment.
32 Washington, DC: Office of Air and Radiation; August. Available:
33 www.epa.gov/ttncaaal/tl/reports/noxrept.pdf [1999, November 24].
34
3 5 Environmental Protection Agency. (2000) Deposition of air pollutants to the great waters. Third report to Congress.
36 [Executive Summary]. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office of Air
37 Quality Planning and Standards; report no. EPA-453/R-00-005.
38
39 Environmental Protection Agency. (2001) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC:
40 Office of Research and Development; report no. EPA/600/P-99/002. March.
41
42 Fenn, M. E.; Poth, M. A.; Aber, J. D.; Baron, J. S.; Bormann, B. T.; Johnson, D. W.; Lemly, A. D.; McNulty, S. G.;
43 Ryan, D. F.; Stottlemyer, R. (1998) Nitrogen excess in North American ecosystems: predisposing factors,
44 ecosystem responses, and management strategies. Ecol. Appl. 8: 706-733.
45
46 Garner, J. H. B. (1994) Nitrogen oxides, plant metabolism, and forest ecosystem response. In: Alscher, R. G.;
47 Wellburn, A. R., eds. Plant responses to the gaseous environment: molecular, metabolic and physiological
48 aspects, [3rd international symposium on air pollutants and plant metabolism]; June 1992; Blacksburg, VA.
49 London, United Kingdom: Chapman & Hall; pp. 301-314.
50
51 Hedin, L. O.; Granat, L.; Likens, G. E.; Buishand, T. A.; Galloway, J. N.; Butler, T. J.; Rodhe, H. (1994)
52 Steep declines in atmospheric base cations in regions of Europe and North America. Nature (London)
53 367: 351-354.
August 2003 5-73 Draft - Do Not Quote or Cite
-------
1 Jaworski, N. A.; Howarth, R. W.; Hetling, L. J. (1997) Atmospheric deposition of nitrogen oxides onto the landscape
2 contributes to coastal eutrophication in the northeast United States. Environ. Sci. Technol. 31: 1995-2004.
3
4 Johnson, D. W.; Van Miegroet, H.; Lindberg, S. E.; Todd, D. E.; Harrison, R. B. (1991) Nutrient cycling in red
5 spruce forests of the Great Smoky Mountains. Can. J. For. Res. 21: 769-787.
6
7 Johnson, D. W.; Lindberg, S. E., eds. (1992) Atmospheric deposition and forest nutrient cycling: a synthesis of the
8 integrated forest study. New York, NY: Springer-Verlag, Inc. (Billings, W. D.; Golley, F.; Lange, O. L.;
9 Olson, J. S.; Remmert, H., eds. Ecological studies: analysis and synthesis: v. 91).
10
11 Johnson, D. W.; Swank, W. T.; Vose, J. M. (1993) Simulated effects of atmospheric sulfur deposition on nutrient
12 cycling in a mixed deciduous forest. Biogeochemistry 23: 169-196.
13
14 Johnson, D.W.; Susfalk, R.B.; Brewer, P.,F.; Swank, W.T. (1999) Simulated effects of reduced sulfur, nitrogen, and
15 base cation deposition on soils and solutions in southern Appalachian forests. J. Environ. Qua!
16 28: 1336-1346.
17
18 Lekke, H.; Bak, J.; Falkengren-Grerup, U.; Finlay, R. D.; Ilvesniemi, H.; Nygaard, P. H.; Starr, M. (1996)
19 Critical loads of acidic deposition for forest soils: is the current approach adequate. Ambio 25: 510-516.
20
21 Lovett, G. M.; Traynor, M. M.; Pouyat, R. V.; Carreiro, M. M.; Zhu, W.-X.; Baxter, J. W. (2000) Atmospheric
22 deposition to oak forest along an urban-rural gradient. Environ. Sci. Technol. 34: 4294-4300.
23
24 Magill, A. H.; Aber, J. D.; Berntson, G. M.; McDowell, W. H.; Nadelhoffer, K. J.; Melillo, J. M.; Steudler, P. (2000)
25 Long-term nitrogen additions and nitrogen saturation in two temperate forests. Ecosystems 3: 238-253.
26
27 McDonnell, M. J.; Pickett, S. T. A.; Groffman, P.; Bohlen, P.; Pouyat, R. V.; Zipperer, W. C.; Parmelee, R. W.;
28 Carreiro, M. M.; Medley, K. (1997) Ecosystem processes along an urban-to-rural gradient. Urban Ecosyst.
29 1:21-36.
30
31 Naeem, S.; Thompson, L. J.; Lawler, S. P.; Lawton, J. H.; Woodfin, R. M. (1994) Declining biodiversity can alter
32 the performance of ecosystems. Nature 368: 734-737.
33
34 National Science and Technology Council. (1998) National acid precipitation assessment program biennial report to
35 Congress: an integrated assessment; executive summary. Silver Spring, MD: U.S. Department of
36 Commerce, National Oceanic and Atmospheric Administration. Available:
37 www.nnic.noaa.gov/CENR/NAPAP/NAPAP_96.htm [24 November 1999].
38
39 Paerl, H. W.; Bales, J. D.; Ausley, L. W.; Buzzelli, C. P.; Crowder, L. B.; Eby, L. A.; Go, M.; Peierls, B. L.;
40 Richardson, T. L.; Ramus, J. S. (2001) Ecosystem impacts of three sequential hurricanes (Dennis, Floyd,
41 and Irene) on the United States' largest lagoonal estuary, Pamlico Sound, NC. Proc. Nat. Acad. Sci. U. S. A.
42 98:5655-5611.
43
44 Pouyat, R. V.; McDonnell, M. J. (1991) Heavy metal accumulations in forest soils along an urban-rural gradient in
45 southeastern New York, USA. Water Air Soil Pollut. 57/58: 797-808.
46
47 Science Advisory Board. (2002) A framework for assessing and reporting on ecological condition: an SAB report.
48 Young, T. F.; Sanzone, S., eds. Washington, DC: U.S. Environmental Protection Agency, Ecological
49 Processes and Effects Committee; report no. EPA-SAB-EPEC-02-009. Available at:
50 http://www.epa.gov/sciencel/index.html (27 January 2003).
51
August 2003 5-74 Draft - Do Not Quote or Cite
-------
1 Strickland, T. C.; Holdren, G. R., Jr.; Ringold, P. L.; Bernard, D.; Smythe, K.; Fallen, W. (1993) A national critical
2 loads framework for atmospheric deposition effects assessment: I. method summary. Environ. Manage.
3 17:329-334.
4
5 Wall, D. H.; Moore, J. C. (1999) Interactions underground: soil biodiversity, mutualism, and ecosystem processes.
6 Bioscience 49: 109-117.
7
8 Wedin, D. A.; Tilman, D. (1996) Influence of nitrogen loading and species composition on the carbon balance of
9 grasslands. Science 274: 1720-1723.
10
11 Wesely, M. L.; Hicks, B. B. (2000) A review of the current status of knowledge on dry deposition. Atmos. Environ.
12 34:2261-2282.
13
14 World Health Organization. (1997) Nitrogen oxides. 2nd ed. Geneva, Switzerland: World Health Organization.
15 (Environmental health criteria 188).
16
17 Section 5.5
18
19 Environmental Protection Agency. (2003) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC:
20 Office of Research and Development; report no. EPA/600/P-99/002. June.
21
22 Intergovernmental Panel on Climate Change (IPCC). (1998) The regional impacts of climate change: an assessment
23 of vulnerability. Cambridge, United Kingdom: Cambridge University Press.
24
25 Intergovernmental Panel on Climate Change (IPCC). (200la) Climate change 2001: the scientific basis.
26 Contribution of working group I to the third assessment report of the Intergovernmental Panel on Climate
27 Change. Cambridge, United Kingdom: Cambridge University Press.
28
29 Intergovernmental Panel on Climate Change (IPCC). (200 Ib) Climate change 2001: impacts, adaptation, and
3 0 vulnerability. Contribution of working group II to the third assessment report of the Intergovernmental
31 Panel on Climate Change. Cambridge, United Kingdom: Cambridge University Press.
32
33 National Academy of Sciences (NAS). (2001) Committee on the Science of Climate Change, National Research
34 Council. Climate Change Science: An Analysis of Some Key Questions, National Academy Press,
35 Washington, DC.
36
37 United Nations Environment Programme (UNEP). (1998) Environmental effects of ozone depletion: 1998
38 assessment. J. Photochem. Photobiol. B 46: 1-4.
39
40 United Nations Environment Programme (UNEP). (2000) Environmental effects of ozone depletion: interim
41 summary. Available at: http://www.gcrio.org/ozone/unep2000summarv.html (9 April 2002).
42
43 World Meteorological Organization. (1999) Scientific assessment of ozone depletion: 1998. Geneva, Switzerland:
44 World Meteorological Organization, Global Ozone and Monitoring Project; report no. 44.
August 2003 5-75 Draft - Do Not Quote or Cite
-------
-------
1 6. STAFF CONCLUSIONS AND RECOMMENDATIONS ON PM NAAQS
2 6.1 INTRODUCTION
3 This chapter presents preliminary staff conclusions and recommendations for the
4 Administrator to consider in deciding whether the existing PM standards should be revised and,
5 if so, what revised or new standards are appropriate.1 Drawing from the synthesis of information
6 and analyses contained in both the draft CD and in Chapters 2 through 5 herein, this chapter
7 begins with staff consideration of the overall adequacy of the current primary standards for PM.
8 Subsequent sections address each of the major components needed to define specific standards:
9 pollutant indicator (for both fine- and coarse-fraction particles), averaging time, form and level.
10 Staff conclusions and recommendations for ranges of alternative primary standards are then
11 presented, based on considering how both the major components of an individual standard and a
12 suite of standards operate together to protect public health with an adequate margin of safety. A
13 discussion of secondary PM standards follows, with staff conclusions and recommendations on
14 alternative secondary standards for both fine- and coarse-fraction particles. This chapter
15 concludes with a summary of key uncertainties associated with establishing PM standards and
16 related staff research recommendations.
17 In recommending a range of primary standard options for the Administrator to consider,
18 the staff notes that the final decision is largely a public health policy judgment. A final decision
19 must draw upon scientific information and analyses about health effects and risks, as well as
20 judgments about how to deal with the range of uncertainties that are inherent in the scientific
21 evidence and analyses. The staffs approach to informing these judgments is based on a
22 recognition that the available health effects evidence generally reflects a continuum consisting of
23 levels at which scientists generally agree that health effects are likely through lower levels at
24 which the likelihood and magnitude of the response become increasingly uncertain. This
25 approach is consistent with the requirements of the NAAQS provisions of the Act and with how
26 EPA and the courts have historically interpreted the Act. These provisions require the
1 As noted in Chapter 1, staff conclusions and recommendations presented in the next draft Staff Paper will
be further informed by consideration of the information and analyses in the final CD, additional staff analyses and
the results of the completed human health risk assessment, and CASAC and public comments received on this draft.
August 2003 6-1 Draft - Do Not Quote or Cite
-------
1 Administrator to establish primary standards that are requisite to protect public health with an
2 adequate margin of safety. In so doing, the Administrator seeks to establish standards that are
3 neither more nor less stringent than necessary for this purpose. The provisions do not require
4 that primary standards be set at a zero-risk level, but rather at a level that avoids unacceptable
5 risks.
6 In recommending secondary standard options, the staff similarly notes that the final
7 decision is largely a public policy judgment that draws upon scientific information and analyses
8 about effects on public welfare, as well as judgments about how to deal with the range of
9 uncertainties that are inherent in the scientific evidence and analyses. In making
10 recommendations, the staff also takes into account the extent to which the recommended ranges
11 of primary standards would provide the requisite protection against adverse PM-related effects
12 on public welfare.
13 Further, the staff notes that especially where considerable uncertainty exists with regard
14 to appropriate policy choices based on the scientific information and analyses, it is appropriate to
15 consider the risk management implications for public health and welfare of alternative
16 approaches that represent scientifically sound options. Thus, staff has considered risk
17 management implications together with the scientific evidence in assessing whether alternative
18 approaches to establishing PM standards would provide both the level of protection that is
19 requisite and an effective and efficient basis for pollution control strategies that will result in the
20 attainment and maintenance of that level of public health and welfare protection.
21 6.2 ADEQUACY OF CURRENT PRIMARY PM NAAQS
22 In 1997, EPA concluded that revision of the PM NAAQS was appropriate, given the
23 body of scientific evidence indicating that health effects were associated with PM at ambient
24 concentrations below the levels of the then effective PM10 standards. EPA further concluded that
25 the most effective and efficient approach to revising the PM NAAQS to achieve the requisite
26 level of protection would be to establish separate standards for the fine and coarse fractions of
27 PM10 (62 FR at 38,665-66, July 18, 1997). Thus, as summarized in Chapter 1, EPA added PM25
28 standards to address effects associated with fine-fraction particles, and retained (with minor
August 2003 6-2 Draft - Do Not Quote or Cite
-------
1 revision) PM10 standards, in conjunction with the PM2 5 standards, to address effects associated
2 with coarse-fraction particles.
3 As an initial step in considering the adequacy of the current suite of PM standards, staff
4 first considered whether information now available in this review continues to support the
5 previous conclusion that fine- and coarse-fraction particles should be regulated separately. As
6 discussed in Chapter 2, fine- and coarse-fraction particles are distinct entities with fundamentally
7 different sources and formation processes, chemical composition, atmospheric residence times,
8 behaviors in the atmosphere, and patterns of human exposure. Also, as discussed in Chapter 3,
9 studies have reported associations of increased mortality and morbidity with both fine- and
10 coarse-fraction particles, and the two fractions of PM appear to have independent effects.
11 Further, as was observed in the previous review (EPA, 1996b), the strategies needed for control
12 of fine- and coarse-fraction particles are different. The recent air quality, exposure, and health
13 effects information and analyses presented and evaluated in the draft CD provide further support
14 for and increase staffs confidence in the conclusion that fine- and coarse-fraction particles
15 should be treated as separate pollutants. As the draft CD concludes:
16 Fine and thoracic coarse PM, indexed respectively by PM25 and PM10_25, should
17 be considered as separate subclasses of PM. Considerations of emissions sources,
18 atmospheric chemistry, physical behavior, exposure relationships, respiratory
19 deposition, toxicologic findings, and epidemiologic observations argue for
20 monitoring fine and thoracic coarse particles separately. (CD, p. E-42)
21 Thus, staff concludes that it is again appropriate to recommend retaining separate standards for
22 fine- and coarse-fraction particles. Based on this recommendation, the following sections focus
23 separately on the adequacy of the current PM25 standards for controlling fine-fraction particles
24 and the adequacy of the current PM10 standards for controlling coarse-fraction particles.
25 6.2.1 Adequacy of Current PM2 5 Standards
26 Staffs assessment of the adequacy of the current PM2 5 standards draws upon the
27 extensive PM epidemiologic data base, as well as information from toxicologic, dosimetric, and
28 exposure-related research, that has become newly available since the last review, as assessed in
29 the draft CD and Chapter 3. Numerous new epidemiologic studies have shown statistically
30 significant associations between short- and long-term ambient PM exposures (based on
August 2003 6-3 Draft - Do Not Quote or Cite
-------
1 community air quality measures) with a variety of human health endpoints, including total and
2 cardiorespiratory mortality; hospital admissions, emergency department visits, and other medical
3 visits for various respiratory or cardiovascular diseases; respiratory illness and symptoms; and
4 lung function changes.
5 The draft CD notes that there is considerable coherence across the newly available
6 epidemiologic study findings (CD, p. 9-127), and that the "general internal consistency of the
7 epidemiologic database and available findings demonstrate well that notable human health
8 effects are associated with exposures to ambient PM at concentrations currently found in many
9 geographic locations across the United States" (CD, p. 9-66). These epidemiologic findings are
10 further enhanced by new findings from toxicologic and dosimetric research, which have
11 provided supportive evidence that certain particle attributes can be causally related to the
12 observed health effects and insights into potential mechanisms by which PM may be affecting
13 human health (CD, p. 9-61). The draft CD concludes that newly available epidemiologic studies
14 are important in extending earlier results to many more cities (CD, p. 9-68), and that the relative
15 risk estimates from the new studies generally comport well with previously reported effects
16 estimates (CD, p. 9-75). Further, the draft CD concludes that there is now a large and reasonably
17 convincing body of evidence that confirms earlier associations between short- and long-term
18 ambient PM10 exposures and mortality/morbidity effects, as well as a growing body of studies
19 that confirm such associations with ambient PM25 exposures and suggest that PM25 is a probable
20 contributing cause of observed PM-associated health effects (CD, p. E-23). In addition, the draft
21 CD highlights new findings that implicate fine-fraction particles as likely contributing to
22 exacerbation of asthma and to increased lung cancer risk (CD, p. E-25, 26).
23 In evaluating this evidence, staff notes that there are many more epidemiologic studies
24 now available in the current review that provide evidence of associations between PM2 5 and
25 serious health effects in areas with air quality at and above the level of the current annual PM25
26 standard. In addition, as shown in Chapter 3, Appendix A, and discussed below in section 6.3.3,
27 there are also a few newly available short-term exposure studies that provide evidence of
28 statistically significant associations with PM2 5 in areas with air quality below the annual
29 standard level of 15 |ig/m3. For example, three such studies (Burnett et al., 2000 and Burnett and
30 Goldberg, 2003; Mar et al., 1999, 2003; Fairley, 1999, 2003) report associations between short-
August 2003 6-4 Draft - Do Not Quote or Cite
-------
1 term PM2 5 exposure and total and cardiovascular mortality in areas in which long-term average
2 PM25 concentrations ranged between 13 and 14 |ig/m3. Other short-term exposure studies with
3 even lower long-term average PM2 5 concentrations reported statistically significant associations
4 with morbidity effects, although the weight to be appropriately accorded these studies is far more
5 uncertain. New evidence is also available from long-term exposure studies suggesting that
6 associations extend below the level of the current standard. For example, limited results from
7 the extended ACS long-term exposure study (Pope et al., 2002), are suggestive of an association
8 between mortality and long-term PM2 5 exposure across cities in which the average aggregate
9 mean concentration for the most recent period evaluated in the study was approximately 14
10 Hg/m3. The entire group of epidemiologic PM25 studies provides direct and strong support for
11 PM2 5 standards that provide at least the level of protection afforded by the current standards, and
12 also clearly calls into question the adequacy of that level of public health protection.
13 Beyond the epidemiologic studies using PM25 as an indicator of fine-fraction particles,
14 there is also a large body of newly available evidence from studies that used PM10, as well as
15 other indicators or components of fine-fraction particles (e.g., sulfates, combustion-related
16 components), that provides additional support for the conclusions reached in the last review as to
17 the likely causal role of ambient PM, and the likely importance of fine-fraction particles, in
18 contributing to observed health effects. Such studies notably include new multi-city studies,
19 intervention studies (that relate reductions in ambient PM to observed improvements in
20 respiratory or cardiovascular health), and source-oriented studies (e.g., suggesting associations
21 with combustion- and vehicle-related sources of fine-fraction particles). Further, the draft CD
22 concludes that new epidemiologic studies of ambient PM associations with infant mortality
23 and/or developmental effects, if further substantiated, suggest life shortening that could well be
24 significantly larger than previously estimated, and that new associations with asthma-related
25 increased physicians visits and symptoms suggest likely much larger public health impacts due
26 to ambient PM than just those indexed by the mortality/morbidity effects considered in the last
27 review (CD, p. E-28). Taken together, this group of studies also provides strong support for
28 standards for fine-fraction particles that are at least as protective as the current PM2 5 standards.
29 These studies also raise questions as to whether those standards should be revised to provide
30 increased protection to reflect the generally stronger and broader body of evidence now available
August 2003 6-5 Draft - Do Not Quote or Cite
-------
1 relative to the evidence that served as the basis for establishing the current PM2 5 standards in the
2 last review.
3 In considering the adequacy of the current standards, staff recognizes, however, that there
4 are important limitations and uncertainties associated with this expanded body of evidence, as
5 discussed at length in the draft CD and in Chapter 3, that need to be carefully considered in
6 determining the weight to be placed on the newly available studies in the current review. The
7 draft CD notes that several key issues need to be considered when interpreting this body of
8 evidence. For example, while PM-effects associations continue to be observed across most new
9 studies, the newer findings do not fully resolve the extent to which the associations are properly
10 attributed to PM acting alone or in combination with other gaseous co-pollutants, or to the
11 gaseous co-pollutants themselves. The draft CD notes that available statistical methods for
12 assessing potential confounding by gaseous co-pollutants may not yet be fully adequate,
13 although the various approaches that have now been used to evaluate this issue tend to
14 substantiate that reported PM effects are at least partly due to PM acting alone or in the presence
15 of other co-varying gaseous pollutants (CD, p. E-26).
16 Another issue of particular importance that has recently been highlighted is the
17 sensitivity of various statistical models to the approach used to address potential confounding by
18 weather- and time-related variables in time-series epidemiological studies. This issue resurfaced
19 in the course of reanalyses of a number of the newer studies that were being conducted to
20 address a more narrow issue related to problems associated with the use of commonly used
21 statistical software (CD, 8-201; Chapter 3, section 3.5.3.1 above). These reanalyses suggest that
22 weather continues to be a potential confounder of concern, and that there remains no altogether
23 satisfactory way to select an appropriate model to address such potential confounding. The HEI
24 Special Panel, in reviewing these reanalyses, concluded that this awareness introduces a degree
25 of uncertainty in the time-series epidemiological studies that had not been widely appreciated
26 previously (CD, p. 8-204).
27 In looking beyond PM mass indicators, a number of newly available studies highlight the
28 issue of the extent to which observed health effects may be associated with various specific
29 chemical components within the mix of fine particles. The potential for various fine particle
30 components to have differing relative toxicities with regard to the various health endpoints being
August 2003 6-6 Draft - Do Not Quote or Cite
-------
1 considered adds complexity to the interpretation of the study results. The draft CD (CD, p. E-
2 28) recognizes that more research is needed to address uncertainties about the extent to which
3 various components may be relatively more or less toxic than others, or than undifferentiated
4 PM25 mass across the range of health endpoints studied.
5 These and other issues that introduce uncertainties into the interpretation of the available
6 body of evidence are reflected in the draft CD's recognition that "many challenges still exist
7 with regard to delineating the magnitudes and variabilities of risk estimates for ambient PM, the
8 ability to attribute observed health effects to specific PM constituents, the time intervals over
9 which PM health effects are manifested, the extent to which findings in one location can be
10 generalized to other locations, and the nature and magnitude of the overall public health risk
11 imposed by ambient PM exposure." (CD, p. 9-66).
12 Staff well recognizes that as the body of available evidence has expanded, it has both
13 added greatly to our knowledge of PM-related effects, as well as to the complexity and
14 uncertainties inherent in efforts to interpret the evidence in a policy-relevant context as a basis
15 for setting appropriate standards. In considering both the available evidence and the inherent
16 uncertainties in the evidence highlighted above, staff concludes that the available evidence
17 sufficiently calls into question the adequacy of the current suite of PM25 standards that
18 consideration should be given to revising the current PM25 standards to provide increased public
19 health protection. Staff also concludes that there remain sufficiently important uncertainties in
20 the available body of evidence such that consideration could also be given to retaining the
21 current PM2 5 standards at this time. Staff conclusions and recommendations on possible
22 alternative primary standards for fine-fraction particles are discussed below in section 6.3.
23 6.2.2 Adequacy of Current PM10 Standards
24 EPA's decision in 1997 to continue to use PM10 as the indicator for standards to control
25 coarse-fraction particles, rather than a more selective indicator (e.g., PM10_2 5), was based in part
26 on the recognition that the only studies of clear quantitative relevance to health effects most
27 likely associated with coarse-fraction particles used undifferentiated PM10 as the indicator of
28 coarse-fraction particles (in areas where the coarse fraction was the predominant component of
29 PM10) and on the very limited ambient air quality data then available on PM10_2 5, in contrast to
August 2003 6-7 Draft - Do Not Quote or Cite
-------
1 the extensive monitoring network already in place for PM10 (62 FR at 38,668). However, as
2 discussed in Chapter 1, in subsequent litigation regarding the 1997 PM NAAQS revisions, the
3 court rejected as arbitrary EPA's reasons for adopting the PM10 indicator. In so doing, the court
4 held in part that PM10 is a "poorly matched indicator" for coarse-fraction particles in the context
5 of a rule that also includes PM25 standards because PM10 includes PM25. 175 F. 3d. at 1054.
6 Although the court found "ample support" (id.) for EPA's decision to regulate coarse-fraction
7 particles, the court nonetheless vacated the 1997 revised PM10 standards for the control of
8 coarse-fraction particles.
9 In considering the adequacy of the current PM10 standards to control coarse-fraction
10 particles, in conjunction with separate standards for PM25, staff now concludes, consistent with
11 the court's opinion, that PM10 is not an appropriate indicator for such standards. In reaching this
12 conclusion, staff has taken into account information now available in this review, including the
13 body of newly available evidence on health effects associated with coarse-fraction particles from
14 studies that directly used PM10_25 as the PM indicator. In addition, staff notes that there is now
15 much more information available to characterize air quality in terms of estimated PM10_252 than
16 was available in the last review. Thus, the reasons used in support of the 1997 decision to retain
17 PM10 as the indicator for coarse-fraction particles no longer hold. As a consequence of
18 concluding that PM10 is no longer an appropriate indicator for coarse-fraction particles, staff also
19 concludes that the current standards defined in terms of a PM10 indicator are not adequate or
20 appropriate for the purpose of providing effective and efficient protection from health effects
21 associated with coarse-fraction particles.
22 Taking into account the newly available information, staff also concludes that it is
23 reasonable and appropriate in this review to consider replacing the current PM10 standards with
24 standards based on an indicator of coarse-fraction particles that does not also include the fine
25 fraction (e.g., PM10_25). Staff conclusions and recommendations on possible alternative
As noted above 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.
August 2003 6-8 Draft - Do Not Quote or Cite
-------
1 indicators, averaging times, forms and levels for coarse-fraction primary standards are discussed
2 below in section 6.4.
3 6.3 ALTERNATIVE PRIMARY STANDARDS FOR FINE-FRACTION PARTICLES
4 6.3.1 Alternative Indicators for Fine-Fraction Particle Standards
5 In 1997, EPA determined that it was more appropriate to control fine-fraction particles
6 (referred to below as fine particles) as a group, as opposed to singling out particular components
7 or classes of fine particles; thus, an undifferentiated mass-based indicator (PM25) was selected
8 rather than an indicator based on PM composition. A number of animal toxicologic and
9 controlled human exposure studies had reported health effects associations with high
10 concentrations of numerous fine particle components (e.g., sulfates, nitrates, transition metals,
11 organic compounds), although such associations were not consistently observed. In addition,
12 community health studies had found significant associations between fine particles or PM10 and
13 health effects in areas with significant mass contribution of differing components or sources of
14 fine particles, including sulfates, wood smoke, nitrates, secondary organic compounds and acid
15 sulfate aerosols. It was also not possible to rule out any one PM component as contributing to
16 the fine particle effects found in epidemiology studies. Thus, it was determined that total mass
17 of fine particles was the most appropriate indicator for fine particle standards. (62 FR at
18 38,667).
19 In considering whether the information available in this review warrants selection of a
20 different indicator for fine particles, staff notes that since the last review many new studies have
21 continued to show associations between various health effects and short- and long-term exposure
22 to fine particles as indexed by PM2 5. In addition, an extensive PM2 5 monitoring network has
23 been deployed and operated in conjunction with the development of strategies and programs to
24 implement the 1997 PM25 standards. This network has provided substantial new air quality
25 information, in terms of PM2 5, that has been and is being used in ongoing PM research and has
26 provided substantial information on PM2 5 air quality that informs this review. Further, staff
27 continues to recognize the importance of an indicator that not only captures all the fine PM
28 components of potential concern (i.e., an effective indicator), but also places greater emphasis
August 2003 6-9 Draft - Do Not Quote or Cite
-------
1 for control on those constituents or fractions that are most likely to result in the largest risk
2 reduction (i.e., an efficient indicator).
3 While many new studies continue to link adverse health effects with short- and long-term
4 exposure to PM2 5, different types of studies also have continued to suggest links between
5 adverse health effects and a range of fine particle components and characteristics. Animal
6 toxicologic and controlled human exposure studies have provided evidence linking a variety of
7 fine particle components or particle types (e.g., sulfates or acid aerosols, metals, organic
8 constituents, bioaerosols, diesel particles) with health effects, although often at high
9 concentrations. In addition, some toxicologic studies have raised questions about the relative
10 toxicity of size-differentiated subsets of fine particles, such as ultrafme particles (at or below 0.1
11 |im in diameter), focusing in particular on the chemical composition of such particles or their
12 increased surface area. These findings are discussed in detail in Chapter 7 of the draft CD and
13 summarized above in Section 3.5.2.2.
14 Some epidemiologic studies also have implicated various PM components (e.g., sulfates,
15 nitrates, carbon, organic compounds, and metals) as being associated with adverse effects. For
16 example, the draft CD discusses associations reported in a number of studies between sulfates
17 and mortality, including both short-term (CD, Figure 8-8) and long-term (CD, p. 8-110) exposure
18 studies. In addition, some recent studies have suggested that the ultrafme subset of fine particles
19 may also be associated with adverse effects. For example, some European studies have
20 suggested independent effects of both ultrafme and fine particles. Further, several important
21 new studies have used factor analysis methods to test for associations between mortality and PM
22 from different PM sources. These studies report associations generally with particles from
23 combustion sources, including motor vehicles and burning of coal, wood or vegetation, but not
24 with particles of crustal origin.
25 Thus, toxicologic and epidemiologic studies summarized above and discussed in the draft
26 CD have provided evidence for effects associated with various fine particle components, size-
27 differentiated subsets of fine particles, and sources of fine particles. The draft CD concludes
28 "the new evidence suggests that exposure to particles from several different source categories,
29 and of different composition and size, may have independent associations with health outcomes"
August 2003 6-10 Draft - Do Not Quote or Cite
-------
1 (CD, p. 9-110). Conversely, the draft CD provides no basis to conclude that any individual fine
2 particle component cannot be associated with adverse health effects.
3 Given the range of health effects with which PM has been associated, and the range of
4 potential mechanisms that may explain how PM may be causally related to such effects, the
5 available evidence provides no basis for expecting that any one component or source of fine
6 particles would be solely responsible for all PM-related effects. As summarized in the draft CD:
7 Toxicological studies have provided considerable supportive evidence that certain
8 physicochemical particle attributes can provide elements of 'causality' to observed health
9 effects of ambient PM. A primary causative attribute may not exist but rather many
10 attributes may contribute to a complex mechanism driven by the nature of a given PM
11 and its contributing sources. The multiple interactions that may occur in eliciting a
12 response in a host may make the identification of any single causal component difficult
13 and may account for the fact that mass as the most basic metric shows the relationships to
14 health outcomes that it does. (CD, p. 9-61).
15 However, it is likely that, for a given health response, some PM components are more
16 closely linked with that response than others. As a consequence, for some specific effects, such
17 as changes in specific immune cell numbers or release of cytokines, there may be stronger
18 correlation with individual PM components than with particle mass. For example, in some
19 toxicologic studies of cardiovascular effects, PM exposures of equal mass did not produce the
20 same effects, indicating that PM composition was important (CD, p. 7-40). Staff recognizes that
21 different PM constituents may have differing biological responses, and that this is an important
22 source of uncertainty in interpreting epidemiologic evidence.
23 In addition to potential differences in biological responses for different mixes of PM
24 components, there may be differences across communities that can influence the PM-health
25 relationship. As summarized in the CD:
26 ... magnitudes and significance levels of observed air pollution-related effects
27 estimates would be expected to vary somewhat from place to place, if the
28 observed epidemiologic associations denote actual effects, because (a) not only
29 would the complex mixture of PM vary from place to place, but also (b) affected
30 populations may differ in characteristics that could affect susceptibility to air
31 pollution health effects. Such characteristics include sociodemographic factors
32 underlying health status, indoor-outdoor activities, diet, medical care access,
33 exposure to risk factors other than ambient air pollution (such as extreme weather
34 conditions), and variations in factors (e.g., air-conditioning) affecting human
35 exposures to ambient-generated PM. (CD, p. 9-87).
August 2003 6-11 Draft - Do Not Quote or Cite
-------
1 Taking into account the above considerations, staff concludes that it remains appropriate
2 to use undifferentiated particle mass as the basis for the indicator of fine particle standards. Staff
3 further concludes that there is no adequate basis for supplementing mass-based fine particle
4 standards with standards for any specific fine particle component or subset of fine particles.
5 Staff also recognizes, however, that potential differences in PM, arising from different mixes of
6 PM sources across the country, are a source of uncertainty in interpreting the epidemiologic
7 evidence as a basis for establishing national fine particle standards.
8 In selecting a specific mass-based indicator for fine particles, it is also necessary to
9 specify a specific size fraction, in terms of a sampler cut point, to differentiate the fine and
10 coarse fractions. As recognized in the last review and summarized in Chapter 2, the particle
11 diameter distinguishing fine- and coarse-fraction particles falls between 1 and 3 jim. As relative
12 humidity increases, fine particles grow into this intermodal size range; as relative humidity
13 decreases, more coarse-mode particles may be suspended in this size range (CD, p. 2-119). In its
14 1997 revisions to the PM NAAQS, EPA recognized that because of the potential overlap
15 between fine and coarse particle mass in this intermodal range, the choice of any specific
16 sampling cut point within this range is largely a policy judgment. In making this judgment, EPA
17 selected 2.5 jam as the cut point for the fine particle indicator, based on considerations of
18 consistency with the epidemiologic studies, the limited potential for intrusion of coarse-fraction
19 particles into the fine fraction, and availability of monitoring technology (62 FR 38668).
20 Further, EPA noted that PM2 5 encompasses many specific components of potential concern in
21 the fine fraction, including most sulfates, acids, transition metals, organics, and ultrafine
22 particles.
23 While additional information is now available on effects associated with various
24 components of fine particles, staff sees no basis in the new information for reaching a different
25 judgment as to the most appropriate indicator of fine particles. Thus, staff concludes that 2.5 jam
26 remains an appropriate cut point for including the larger accumulation-mode particles of the fine
27 fraction while limiting intrusion of coarse fraction particles, and recommends that PM2 5 be
28 retained as an effective and efficient indicator for fine-fraction particles.
August 2003 6-12 Draft - Do Not Quote or Cite
-------
1 6.3.2 Alternative Averaging Times for PM2 5 Standards
2 In the last review, EPA established PM2 5 standards with annual and 24-hour averaging
3 times (62 FR at 38,668-70). This decision was based in part on evidence of health effects related
4 to both short-term (from less than 1 day to up to several days) and long-term (from a year to
5 several years) measures of PM. EPA recognized that a 24-hour averaging time was consistent
6 with the majority of community epidemiologic studies reporting daily associations. Further,
7 EPA noted that a 24-hour standard could also effectively protect against episodes lasting several
8 days, as well as providing some degree of protection from potential effects associated with
9 shorter duration exposures. EPA also recognized that an annual standard would provide effective
10 protection against both annual and multi-year, cumulative exposures that had been associated
11 with an array of health effects, and that a much longer averaging time would complicate and
12 unnecessarily delay control strategies and attainment decisions. The possibility of seasonal
13 effects also was considered, although the very limited available evidence and the seasonal
14 variability of sources of fine particle emissions across the country did not provide a satisfactory
15 basis for establishing a seasonal averaging time.
16 In deciding to set both annual and 24-hour standards, EPA concluded that the most
17 effective and efficient approach was to treat the annual standard as the generally controlling
18 standard for lowering the entire distribution of PM25 concentrations, with the 24-hour standard
19 providing protection against the occurrence of peak 24-hour concentrations, particularly those
20 that present localized or seasonal exposures of concern in areas where the highest 24-hour-to-
21 annual mean PM25 ratios are appreciably above the national average. This conclusion was
22 supported by results of the PM risk assessment from the last review that indicated that peak 24-
23 hour PM2 5 concentrations contributed a relatively small amount to total health risk, such that
24 much if not most of the aggregated annual risk associated with short-term exposures results from
25 the large number of days during which the 24-hour average concentrations are in the low- to
26 mid-range. Further, no evidence suggested that risks associated with long-term exposures are
27 likely to be disproportionately driven by peak 24-hour concentrations. Thus, a generally
28 controlling annual standard was judged to reduce risks associated with both short- and long-term
29 exposures effectively and with more certainty than a 24-hour standard. Further, an annual
30 standard is more stable over time, likely resulting in the development of more consistent risk
August 2003 6-13 Draft - Do Not Quote or Cite
-------
1 reduction strategies, since an area's attainment status would be less likely to change due solely to
2 year-to-year variations in meteorological conditions that affect the atmospheric formation of fine
3 particles.
4 Information available in this review generally is consistent with and supportive of the
5 conclusions reached in the last review that provided the basis for setting both annual and 24-hour
6 PM2 5 standards, and for establishing the annual standard as the generally controlling standard.
7 In considering the new information, staff makes the following observations:
8 • The conclusion that peak 24-hour PM2 5 concentrations contribute a relatively small
9 amount to the total health risk associated with short-term exposures on an annual basis,
10 such that much if not most of the aggregated annual risk results from the large number of
11 days during which the 24-hour average concentrations are in the low- to mid-range, is
12 supported by EPA's updated risk assessment, as discussed in section 4.3.3. Support for
13 this conclusion is also found in studies in which health effect associations remain when
14 high-concentration days are removed from the analysis (Schwartz et al., 1996; Ostro et
15 al., 1999, 2000).
16 • It continues to be the case, as discussed in section 4.2.6.1, that available short-term
17 exposure studies do not provide evidence of population thresholds, but rather reflect
18 relationships between health effects and ambient PM across a distribution of PM
19 concentrations. Thus, as in the last review, staff recognizes that attempting to identify a
20 "lowest observed effect level" in terms of a 24-hour concentration to serve as a basis for
21 establishing a generally controlling 24-hour PM25 standard would not be an appropriate
22 approach in this review.
23 • Some recent studies have used a distributed lag over several days preceding the health
24 event, as discussed in chapter 8 of the draft CD and summarized in section 3.5.2.2.
25 While such studies continue to suggest consideration of a multiple day averaging time,
26 staff notes that limiting 24-hour concentrations of fine particles will also protect against
27 effects found to be associated with PM averaged over several days in health studies.
28 Consistent with the conclusion reached in the last review, staff again concludes that the
29 complexity of a multiple-day averaging time would not provide more effective protection
30 than a 24-hour average.
31 • There is a growing body of studies that provide additional evidence of effects associated
32 with exposure periods shorter than 24-hours (e.g., one to several hours), as noted in
33 section 3.3. While staff concludes that this information remains too limited to serve as a
34 basis for establishing a shorter-than-24-hour fine particle standard at this time, staff
35 recognizes this as an important area of research that could provide a basis for the
36 consideration of a shorter-term standard in the future.
August 2003 6-14 Draft - Do Not Quote or Cite
-------
1 • While some newer studies have investigated seasonal effects, as noted in section 3.3,
2 staff concludes that currently available evidence of such effects is still too limited to
3 serve as a basis for considering seasonal standards.
4 Based on the above considerations, staff concludes that the currently available
5 information supports keeping and provides no basis for changing the averaging times of the
6 current PM25 standards. Staff recommends that both annual and 24-hour alternative PM2 5
7 standards be considered together. Further, staff recommends that the annual standard be
8 established so as to be the generally controlling standard, with the 24-hour standard continuing
9 to provide protection against unusually high peak short-term PM2 5 concentrations.
10 6.3.3 Alternative Annual PM2 5 Standards
11 In the last review, EPA concluded that the level of a generally controlling annual PM2 5
12 standard should be selected so as to limit annual PM2 5 concentrations to somewhat below those
13 where the body of epidemiologic evidence is most consistent and coherent (62 FR at 38,675-77).
14 This approach recognized both the strengths and the limitations of the full range of scientific and
15 technical information on the health effects of PM, as well as the associated uncertainties. In so
16 doing, EPA placed the greatest weight on those U.S. and Canadian epidemiologic studies
17 reporting statistically significant health effects associations with direct measures of fine particle
18 mass. Further, EPA placed the greatest emphasis on the short-term exposure studies, while also
19 considering whether the long-term exposure studies suggested the need for a lower level. This
20 approach recognized that while effects may occur over the full range of concentrations observed
21 in the studies, the strongest evidence for short-term PM2 5 effects occurs at concentrations near
22 the long-term average across the study period. Based on this approach, EPA set the annual PM2 5
23 standard at a level of 15 |ig/m3, noting that this standard would provide substantial protection
24 against short-term as well as long-term exposures to fine particles. In so doing, EPA recognized
25 that while the possibility of effects at lower levels cannot be excluded, the evidence for that
26 possibility is highly uncertain and the likelihood of significant health risk, if any, becomes
27 smaller as concentrations approach the lower end of the range of air quality observed in the key
28 epidemiologic studies and/or background levels. In addition, EPA recognized that such a
29 standard could not be expected to offer an adequate margin of safety against the effects of all
August 2003 6-15 Draft - Do Not Quote or Cite
-------
1 potential short-term peak exposures in areas with strong local or seasonal sources of fine
2 particles. Thus, this annual standard was set in conjunction with a 24-hour standard to
3 supplement health protection from unusually high peak levels, as discussed below in section
4 6.3.4.
5 Since the last review, many new studies relating ambient PM2 5 concentrations to health
6 effects have been published; many of these studies have made progress in addressing some of the
7 key uncertainties identified in the last review; and two or more years of ambient air quality data
8 have been collected from the PM2 5 monitoring network. Notwithstanding this extensive body of
9 new information, staff notes that important uncertainties remain that need to be taken into
10 account in weighing the new information as a basis for considering alternative PM2 5 annual
11 standards. In considering the newly available information and the associated uncertainties, staff
12 believes that the approach taken in the last review to setting a generally controlling annual
13 standard remains appropriate as a framework for evaluating the currently available information.
14 Staff sees nothing fundamentally different in the currently available information that would
15 invalidate the previous approach or indicate that a different approach would be more appropriate.
16 In first considering the results of currently available studies of short-term exposure to
17 fine particles, and the associated fine particle levels, staff continues to focus on U.S. and
18 Canadian studies for drawing quantitative conclusions for standard-setting purposes. In so
19 doing, staff believes it is important to take into account the following factors:
20 • The extent to which reported health effects associations with fine particles are
21 statistically significant, not only in single-pollutant models, but also in multi-pollutant
22 models or other types of analyses that assess the likelihood that the association is robust
23 to the inclusion of potentially confounding co-pollutants.
24 • The extent to which results modeled using GAM functions were re-analyzed to address
25 statistical problems that may have been present in initially reported results.
26 • The relative reliability of the air quality data used in the study, including consideration of
27 the type of monitoring used and the extent to which daily air quality was measured or
28 estimated based on less-than-every-day measurements.
29 • The relative precision of the study results, with greater weight being placed on studies
30 with greater statistical power to produce more precise effects estimates.
August 2003 6-16 Draft - Do Not Quote or Cite
-------
1 • The extent to which the weight of evidence supports an association between fine particles
2 and the health endpoint assessed in a given study.
3 Effects estimates for mortality and morbidity effects from single-pollutant PM2 5 short-
4 term exposure studies are presented in the table in Appendix A of Chapter 3, along with reported
5 mean PM2 5 concentrations. Studies included in this table are those in which GAM was not used
6 and those for which GAM results have been reanalyzed. Studies using GAM that were not
7 reanalyzed are not included in this table, but are discussed in the draft CD (Chapter 8) and in
8 section 3.3 above. Similarly, information on the extent to which co-pollutants were assessed in
9 these studies, and the extent to which study results were robust to the inclusion of co-pollutants,
10 is discussed in the draft CD and in section 3.3. Information on the type of monitoring used and
11 the nature of the air quality data used in the study is generally summarized in the draft CD and in
12 section 3.3, and discussed more specifically in the individual studies. The relative precision of
13 mortality and morbidity PM25 studies is generally shown in Figures 3-5 and 3-8, respectively.
14 Based on this information, and taking the factors identified above into consideration, staff
15 makes the following observations:
16 • A number of epidemiologic studies reporting statistically significant associations with
17 mortality and morbidity effects have been conducted in areas in which the long-term (i.e.,
18 annual or multi-year) mean 24-hour PM25 concentrations ranged from over 40 |ig/m3
19 down to approximately 8.5 |ig/m3.
20 Many of these studies reported associations that remained significant upon reanalysis,
21 were generally robust to the inclusion of co-pollutants in multi-pollutant models, had adequate
22 air quality data for PM2 5 mass, and resulted in relatively precise effects estimates of health
23 effects for which the weight of the entire body of evidence supports the likelihood of an
24 association for the health endpoint with fine particles. Other studies not included in the table in
25 Appendix A of Chapter 3 also showed statistically significant associations based on using a
26 GAM function, but because reanalyses have not been done, it is unclear whether the associations
27 would remain significant with the use of more appropriate models. Taken together, these studies
28 greatly extend the evidence of PM2 5 associations beyond that available in the last review, in
29 which 15.7 |ig/m3 was the lowest long-term mean 24-hour PM2 5 concentration for which a
30 statistically significant association was reported. Many of the newly available studies shown in
August 2003 6-17 Draft - Do Not Quote or Cite
-------
1 Appendix A in which the mean concentrations were above 15 |ig/m3 lend additional support for
2 an annual standard at least as protective as the current standard. Further, studies showing
3 statistically significant associations that have been conducted in areas in which the long-term
4 mean 24-hour PM25 concentrations were lower than 15 |ig/m3 support consideration of a lower
5 standard level, as discussed below.
6 • Studies reporting statistically significant associations in areas where the long-term mean
7 24-hour PM2 5 concentrations ranged from approximately 14 |ig/m3 down to 8.5 |ig/m3
8 provide evidence of PM2 5-related total and cardiovascular mortality and emergency
9 department visits related to asthma and cardiovascular illness at levels below the current
10 annual standard.
11 In considering this group of studies, staff has focused first on those studies that included
12 adequate gravimetric PM2 5 mass measurements, and where mortality associations with fine
13 particles were generally robust in models including gaseous copollutants. Studies conducted in
14 Phoenix (Mar et al., 1999), Santa Clara County, CA (Fairley, 1999), and in eight Canadian cities
15 (Burnett et al., 2000) reported significant relationships between PM25 and mortality where mean
16 PM2 5 concentrations ranged between 13 and 14 |ig/m3. These studies were reanalyzed to address
17 questions about the use of GAM, and the study results from Phoenix and Santa Clara County
18 were little changed in alternative models (Mar et al., 2003; Fairley, 2003), although Burnett and
19 Goldberg (2003) reported that their results were sensitive to using different temporal smoothing
20 methods. Beyond these mortality studies, two other studies provided evidence of statistically
21 significant associations with emergency department visits. One such GAM study (Norris et al.,
22 1999) reported a highly significant association with asthma visits where the mean PM25
23 concentration was approximately 12 |ig/m3, primarily based on nephelometry data. However,
24 this study has not been reanalyzed, such that it is not clear if the association would remain
25 significant with a more appropriate model, nor did the study assess the potential for confounding
26 by co-pollutants. The other such study (Stieb et al., 2000) did not use GAM and reported a
27 significant association with cardiovascular visits in a single pollutant model where the mean
28 PM2 5 concentration was approximately 8.5 |ig/m3. However, the PM2 5 association was not
29 robust to the inclusion of co-pollutants, such that the authors concluded that the results imply no
30 effects of PM independent of co-pollutants (CD, 8B-15).
August 2003 6-18 Draft - Do Not Quote or Cite
-------
1 Based on consideration of the available PM25 short-term exposure studies, staff believes
2 that it would be appropriate to consider a range of PM25 levels for an annual standard that
3 extends down from 15 |ig/m3 to as low as 12 |ig/m3. Staff judges that consideration of a level as
4 low as 12 |ig/m3 would be precautionary and would place much weight on the overall strength
5 and coherence of the evidence, while giving less weight to the uncertainties that add to the
6 difficulties inherent in attempting to interpret the available evidence in a policy-relevant context.
7 Staff notes that while one study provides some suggestion that PM25 associations may possibly
8 extend to quite low levels beyond this range, it is not possible to distinguish the potential
9 contribution of fine particles from that of the co-pollutants in this study. At the lower end of this
10 range, consideration of air quality data measurement methods, potential confounding by co-
11 pollutants, and uncertainties in appropriate model selection adds considerable uncertainty to the
12 use of study results as a basis for quantitative standard-setting decisions.
13 Beyond the short-term exposure PM2 5 studies, staff also has considered the results of
14 currently available U.S. and Canadian studies of long-term exposure to fine particles.
15 Information on these studies, and the associated mean fine particle levels, is discussed in the
16 draft CD (Chapter 8) and section 3.3 above, especially in Tables 3-3 and 3-4 for mortality and
17 morbidity effects, respectively. Based on this information, staff makes the following
18 observations:
19 • Since the last review, reanalyses and extensions of previously available long-term
20 exposure studies have confirmed and strengthened the evidence of associations between
21 fine particles and mortality and morbidity effects.
22 • Newly available long-term mortality and morbidity studies lend support for considering
23 an annual PM2 5 standard at a level below the current standard, consistent with the
24 evidence from the short-term exposure studies.
25 Two of the long-term mortality studies that were available in the previous review used
26 data from the Six Cities and the ACS cohorts, and had aggregate mean PM25 concentrations of
27 18 |ig/m3 (ranging from approximately 11 to 30 |ig/m3 across cities) and 21 |ig/m3 (ranging from
28 approximately 9 to 34 |ig/m3 across cities), respectively. Reanalyses of data from these cohorts
29 continued to report significant associations with PM2 5, using essentially the same air quality
30 distributions, and new analyses using the ACS cohort include PM25 concentrations from 1999-
August 2003 6-19 Draft - Do Not Quote or Cite
-------
1 2000 that range from about 6 to 21 |ig/m3 (from figure 1, Pope et al., 2002) with a mean of
2 approximately 14 |ig/m3 (Pope et al., 2002). Significant associations, though with somewhat
3 smaller effect estimates, were reported with PM2 5 in the new analysis (Pope et al., 2002).
4 New reports of associations between morbidity effects and long-term PM2 5 exposure
5 were available from the Southern California children's health cohort, where the means of 2-week
6 average PM25 concentrations, measured in 1994 using an acid/aerosol sampler designed for this
7 study, ranged from approximately 7 to 32 |ig/m3, with an across-city average of approximately
8 15 |ig/m3 (Peters et al., 1999). In figures depicting relationships between lung function growth
9 and average PM concentration, there is no evidence of a threshold in the data (Gauderman et al.,
10 2000, 2002). As summarized in Table 3.4, the series of reports from this cohort study generally
11 show decreases in lung function and lung function growth and increased risk of developing
12 respiratory illness, though the findings do not always reach statistical significance.
13 Taking into account the above observations from both the short- and long-term exposure
14 studies, staff concludes that the findings of new epidemiologic studies are consistent with the
15 evidence that supported the 1997 decision to set a generally controlling annual PM25 standard
16 level of 15 |ig/m3, and that these findings clearly provide a basis for consideration of annual
17 PM2 5 standard levels lower than the current standard. In making a preliminary recommendation
18 on a range of annual standard levels for consideration, staff believes that the range should extend
19 down to a level of 12 |ig/m3. However, staff recognizes that information on the estimated
20 reduction in health risks associated with just attaining alternative standards in this range is not
21 yet available, and staff will consider such information in presenting conclusions and
22 recommendations on this standard in the next draft of this Staff Paper.
23 In addition to considering the level of an annual standard, consideration must also be
24 given to the form of the standard. In 1997 EPA established the form of the annual PM2 5
25 standard as an annual arithmetic mean, averaged over 3 years, from single or multiple
26 community-oriented monitors. This form was intended to represent a relatively stable measure
27 of air quality and to characterize area-wide PM25 concentrations. The arithmetic mean serves to
28 represent the broad distribution of daily air quality values, and a 3-year average provides a more
29 stable risk reduction target than a single-year annual average. The annual PM2 5 standard level is
30 to be compared to measurements made at either a single representative community-oriented
August 2003 6-20 Draft - Do Not Quote or Cite
-------
1 monitoring site, or the spatial average of measurements from multiple community-oriented
2 monitoring sites (62 FR at 38,672). This approach was judged to be consistent with the
3 epidemiologic studies on which the PM2 5 standard was primarily based, in which air quality data
4 were generally averaged from across several monitors in an area or taken from a single monitor
5 that was selected to represent community-wide exposures, not localized "hot spots."
6 In this review, in conjunction with recommending that consideration be given to
7 alternative standard levels, staff is also reconsidering the appropriateness of allowing for spatial
8 averaging across monitors in an area. There now exists much more PM2 5 air quality data than
9 were available in the last review, and consideration of the spatial variability across urban areas
10 that is revealed by this new database (see draft CD, section 3.2.5, and section 2.5 above) raises
11 questions as to whether an annual standard that is based on averaging across monitors in an area
12 will result in sufficient uniformity in public health protection in implementing such a standard in
13 areas across the country. Staff intends to further assess this issue and will consider any such
14 additional analysis in the next draft of this Staff Paper.
15 6.3.4 Alternative 24-Hour PM2 5 Standards
16 In the last review, EPA considered a number of factors in establishing the level of the 24-
17 hour PM25 standard, including the significant protection afforded against short-term exposures
18 by the annual PM25 standard and the role of the 24-hour standard in providing supplemental
19 protection against peak exposures not addressed by the annual standard (62 FR at 38,676-77).
20 Air quality and effects information from the studies that helped to inform the selection of the
21 annual standard also was considered, as were uncertainties in the risks associated with infrequent
22 and isolated peak exposures in areas that meet the annual standard. Having decided to set the
23 annual PM25 standard at a level of 15 |ig/m3, EPA set the 24-hour standard at a level of 65
24 Hg/m3, which was the upper end of the range under consideration. In so doing, EPA placed
25 much weight on the uncertainties associated with understanding the extent to which infrequent
26 peak 24-hour exposures that could occur in areas that attain the annual standard contribute to the
27 total risk associated with daily exposures over the course of a year.
28 In setting the level of the 24-hour standard, EPA first selected the form of the standard to
29 be the 98th percentile of 24-hour concentrations at each population-oriented monitor within an
August 2003 6-21 Draft - Do Not Quote or Cite
-------
1 area, averaged over three years (62 FR at 38671-74). EPA selected such a concentration-based
2 form because of its advantages over the previously used expected-exceedance form3. A
3 concentration-based form is more reflective of the health risk posed by elevated PM2 5
4 concentrations because it gives proportionally greater weight to days when concentrations are
5 well above the level of the standard than to days when the concentrations are just above the
6 standard. Further, a concentration-based form better compensates for missing data and less-than-
7 every-day monitoring; and, when averaged over 3 years, it has greater stability and, thus,
8 facilitates the development of more stable implementation programs. After considering a range
9 of concentration percentiles from the 95th to the 99th, EPA selected the 98th percentile as an
10 appropriate balance between adequately limiting the occurrence of peak concentrations and
11 providing increased stability and robustness. Further, by basing the form of the standard on
12 concentrations measured at population-oriented monitoring sites (as specified in 40 CFR part
13 58), EPA intended to provide protection for people residing in or near localized areas of elevated
14 concentrations.
15 As discussed above in section 6.3.3, since the last review, many new studies relating
16 ambient PM2 5 concentrations to health effects provide evidence of associations at air quality
17 levels below those for which statistically significant associations were observed in the last
18 review. In addition, the substantial PM25 air quality database that is now available allows for
19 better characterization of the relationship between annual concentrations and associated
20 distributions of 24-hour concentrations in a large number of areas across the country.
21 Nevertheless, significant uncertainty remains in understanding the risks associated with peak
22 exposures in areas that meet alternative annual standards.
23 In considering the newly available information, the associated uncertainties, and the
24 inherent difficulties in using this information as a basis for quantitative standard setting, staff has
25 identified two different approaches to setting a 24-hour PM2 5 standard to provide an appropriate
26 degree of supplemental protection relative to the range of staff-recommended alternative annual
27 standards for consideration in this review. One approach would focus on the upper end of the
3 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.
August 2003 6-22 Draft - Do Not Quote or Cite
-------
1 distribution of daily PM2 5 concentrations from short-term exposure studies reporting statistically
2 significant health effects associations where the annual mean concentrations were within the
3 staff-recommended range of annual standard levels. A second approach would focus on an
4 assessment of recent PM25 air quality data that presents the distributions of 98th percentile 24-
5 hour average PM2 5 concentrations as a function of annual mean PM2 5 concentrations (shown in
6 Chapter 2, Figure 2-20). The information presented in this figure provides some indication of
7 the extent to which specific 24-hour standard levels would likely result in some degree of
8 supplemental protection, beyond that provided by an annual standard alone, against peak
9 exposures at sites that would likely attain specific annual standard levels. Staff believes that
10 either approach, or both approaches considered together, can provide an appropriate basis for
11 recommending a range of alternative 24-hour PM2 5 standards for consideration in this review. In
12 considering these approaches as a basis for identifying alternative levels for a 24-hour PM2 5
13 standard, staff has continued to focus on standards with the same 98th percentile form as the
14 current standard. While staff notes that alternative forms could also be considered in this review,
15 staff sees nothing in the new information that calls into question the appropriateness of the
16 current form or the basis for the selection of the form in the last review. Thus, staff recommends
17 that consideration be given to retaining the 98th percentile form for a 24-hour PM25 standard.
18 Using the first approach to consider alternative levels, staff focused on those short-term
19 exposure studies discussed above in section 6.3.3 that provided the basis for the range of
20 recommended alternative annual standards. The upper end of the PM25 concentration ranges
21 from these studies as well as other studies reporting health effects associations where annual
22 mean PM25 concentrations were at or below around 18 |ig/m3 were obtained from the published
23 study or from the study authors and are documented in Ross (2003). In particular, staff has
24 focused on the upper end of the concentration ranges for studies reporting statistically significant
25 effects where annual mean PM25 concentrations ranged from around 15 |ig/m3 down to
26 approximately 12 |ig/m3. Looking first at the range of 98th percentile concentrations in these
27 studies, the concentrations generally ranged from approximately 32 to 42 |ig/m3, with one study
28 having a 98th percentile concentration near 60 |ig/m3. Looking more broadly at the 95th to 99th
29 percentile concentrations in these studies, the concentrations generally ranged from somewhat
30 below 30 |ig/m3 to approximately 45 |ig/m3, with one study having a 99th percentile
August 2003 6-23 Draft - Do Not Quote or Cite
-------
1 concentration near 70 |ig/m3. Taken together, this information suggests consideration of 24-hour
2 PM2 5 standard levels below the level of the current standard, ranging from approximately 45
3 Hg/rn3 down to about 30 |ig/m3.
4 Using the second approach, staff looked at the information in Figure 2-20, based on an
5 assessment of air quality data from 2000 to 2003 at over 1000 monitoring sites, to determine
6 what 24-hour standard levels would likely result in some degree of supplemental protection
7 relative to the protection afforded by an annual standard alone, for annual standards within the
8 range of 15 to 12 |ig/m3. As an initial matter, staff observes that the current standard level of 65
9 Hg/rn3 provides essentially no supplemental protection (with only one maximum 98th percentile
10 concentration occurring above that level) in areas where the 3-year average annual means ranged
11 from 16 |ig/m3 down to the lowest levels observed. Staff notes that the maximum values from
12 the 98th percentile concentration distributions in areas with annual means ranging from 15 to 12
13 ng/m3 are generally around 60 |ig/m3, whereas the top 5th percentile of these distributions are
14 generally within the range of approximately 50 down to somewhat above 40 |ig/m3.
15 To assist in understanding the public health implications of various combinations of
16 alternative annual and 24-hour standards, staff assessed (based on the same air quality database)
17 the percentage of counties that would not likely attain various PM2 5 annual standards alone in
18 comparison to the percentage of counties that would not likely attain alternative combinations of
19 annual and 24-hour PM2 5 standards. This assessment is intended to provide some rough
20 indication of the breadth of supplemental protection potentially afforded by various
21 combinations of alternative standards. The results of such an assessment, based on air quality
22 data from 693 counties (including 1152 monitoring sites), are shown in Table 6-1. From this
23 table it can be seen that for an annual standard set at 15 |ig/m3, 24-hour standard levels ranging
24 from 50 to 40 |ig/m3 would add approximately 2 to 6 % to the percentage of counties nationwide
25 that would not likely attain both standards relative to the number of counties that would not
26 likely attain the annual standard alone. For an annual standard set at 13 |ig/m3, 24-hour standard
27 levels ranging from 50 to 35 |ig/m3 would add approximately 1 to 7 % to the percentage of
28 counties not likely to attain both standards; and for an annual standard set at 12 |ig/m3, 24-hour
29 standard levels in this range would add approximately 0 to 5 % to the percentage of counties.
August 2003 6-24 Draft - Do Not Quote or Cite
-------
1 Staff believes that both approaches provide useful information to frame the range of
2 alternative 24-hour standards that are appropriate for consideration in this review. By combining
3 the ranges suggested by both approaches, staff makes a preliminary recommendation that
4 consideration be given in this review to setting a 24-hour PM2 5 standard, with a 98th percentile
5 form, within a range of levels below the level of current standard, extending from approximately
6 50 |ig/m3 down to 30 |ig/m3. As in the last review, staff believes that selecting a 24-hour PM25
7 standard from within this range should take into account the degree of protection likely afforded
8 by the selected annual PM25 standard. However, as noted above in section 6.3.3, staff will
9 consider information not yet available from the human health risk assessment in presenting
10 conclusions and recommendations on this standard in the next draft of this Staff Paper.
August 2003 6-25 Draft - Do Not Quote or Cite
-------
1 TABLE 6-1 PREDICTED PERCENTAGE OF COUNTIES WITH MONITORS NOT
2 LIKELY TO MEET ALTERNATIVE PM2 5 STANDARDS
Alternative
Standards and
Levels (ug/m3)
No. of Counties
with Monitors
\nnual:
15
14
13
12
Percent of counties not likely to meet alternative PM25 standards *
Total
693
19
31
42
53
Region 1:
Northeast
111
20
35
48
67
Region 2:
Southeast
197
19
33
49
64
Region 3:
Industrial
Midwest
145
41
63
79
87
Region 4:
Upper
Midwest
62
0
0
6
13
Region 5:
Southwest
32
6
6
6
6
Region 6:
Northwest
114
4
7
10
17
Region 7:
Southern
CA
15
60
67
67
67
Outside
Regions*
17
0
0
6
6
Combined
\nnual/24-Hour:
15/65
15/50
15/40
15/35
15/30
15/25
19
21
25
34
59
80
20
20
26
48
84
97
19
19
19
20
53
87
41
41
41
54
89
99
0
0
0
2
11
50
6
6
6
9
19
44
4
14
28
41
51
64
60
67
80
80
80
80
0
0
6
6
6
18
13/50
13/40
13/35
13/30
13/25
43
46
49
63
80
48
50
54
84
97
49
49
50
60
87
79
79
80
94
99
6
6
8
16
50
6
6
9
19
44
16
29
41
51
64
73
80
80
80
80
6
6
6
6
18
12/50
12/40
12/35
12/30
12/25
53
56
58
67
81
67
68
69
86
97
64
64
65
72
88
87
87
87
95
99
13
13
15
21
50
6
6
9
19
44
18
31
41
51
64
73
80
80
80
80
6
6
6
6
18
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
* Based on 4, 8, or 12 consecutive quarters of data with at least 11 samples per quarter from the 2000 to 2002
database. As such, these estimates are not based on the same amount of 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.
August 2003
6-26
Draft - Do Not Quote or Cite
-------
1 6.4 ALTERNATIVE PRIMARY STANDARDS FOR COARSE-FRACTION
2 PARTICLES
3 6.4.1 Alternative Indicators for Coarse-Fraction Particle Standards
4 In 1997, in conjunction with new standards for fine particles, EPA decided to use PM10 as
5 an indicator to protect against the effects most likely associated with coarse-fraction particles
6 (referred to below as coarse particles) (62 FR at 38,668). As discussed above in section 6.2.2, in
7 light of currently available information, the reasons for that decision no longer hold, and staff
8 now concludes that PM10 is not an appropriate indicator for such standards.
9 The recent scientific evidence summarized in the draft CD does support, nonetheless,
10 EPA's previous decision generally to use indicators based on PM mass (e.g., PM25, PM10) for the
11 PM NAAQS, as was discussed above for fine particles. In addition, recent evidence from
12 dosimetric studies supports retaining 10 jim as the appropriate cut point for particles capable of
13 penetrating to the thoracic regions of the lung. Thus, in conjunction with PM25 standards, staff
14 concludes that an appropriate mass-based indicator for coarse particles would be PM10_2 5.
15 There is limited evidence to support consideration of other indicators for coarse particles,
16 such as individual components within the coarse fraction. In general, less is known about the
17 composition of coarse particles than fine particles. Even less evidence is available from health
18 studies that would allow identification of specific components or groups of components of coarse
19 particles that may be more closely linked with specific health outcomes. Several studies have
20 focused on the crustal, or geological component of particles, and the results suggest that crustal
21 material, either in the fine or coarse fraction, is not associated with mortality (CD, p. 8-59).
22 Studies conducted in Spokane, WA, and Utah Valley, UT reported that mortality was not
23 associated with high PM10 concentrations resulting from dust storms, although there were
24 statistically significant associations between PM10 and mortality in both (Schwartz et al., 1999
25 and Pope et al., 1999; studies not reanalyzed). However, particles of crustal origin may be
26 linked with morbidity effects under some conditions, or resuspended coarse particles may serve
27 as carriers for other more toxic components, such as metals from previously deposited fine
28 particles or pesticides applied to agricultural lands. The draft CD concludes, however, that more
29 research is needed to identify conditions under which one or another class of coarse particles
30 may cause little or no adverse health effects, as well as conditions under which such particles
August 2003 6-27 Draft - Do Not Quote or Cite
-------
1 may cause notable effects (CD, p. 8-280). Thus, as for fine particles, there is no evidence that
2 would lead toward the selection of one or more PM components as being primarily responsible
3 for effects associated with coarse particles, nor is there any component that can at this time be
4 eliminated from consideration.
5 Taking into account the above considerations, staff concludes that an undifferentiated
6 mass-based indicator continues to be the most appropriate indicator for coarse particle standards.
7 Staff recommends that such an indicator would appropriately retain 10 jim as the upper cut point
8 for the coarse fraction, since larger-sized particles are less likely to penetrate into the thoracic
9 regions, and that the lower cut point of 2.5 jam be used so as to most clearly differentiate
10 between coarse (PM10_25) and fine (PM25) particles. Staff notes that consideration of a PM10_25
11 indicator is supported by EPA's ongoing efforts to establish a FRM for PM10_2 5 and to design a
12 national network of such monitors (see section 2.5.3).
13 6.4.2 Alternative Averaging Times for PM10_2 5 Standards
14 In the last review, EPA retained both annual and 24-hour standards to provide protection
15 against the known and potential effects of short- and long-term exposures to coarse-fraction
16 particles (62 FR at 38,677-79). This decision was based in part on qualitative considerations
17 related to the expectation that deposition of coarse-fraction particles in the respiratory system
18 could aggravate effects in individuals with asthma. In addition, quantitative support came from
19 limited epidemiologic evidence suggesting that aggravation of asthma and respiratory infection
20 and symptoms may be associated with daily or episodic increases in PM10, where dominated by
21 coarse-fraction particles including fugitive dust. Further, potential build-up of insoluble coarse-
22 fraction particles in the lung after long-term exposures to high levels also was considered.
23 Information available in this review on coarse-fraction particles, while still limited,
24 represents a significant expansion of the evidence available in the last review. A number of
25 epidemiologic studies are now available that report statistically significant associations between
26 short-term exposure to PM10_2 5 and both morbidity effects and mortality. In addition, one recent
27 study linked reduced lung function growth with long-term exposure to both PM10_2 5 and PM2 5.
28 Staff believes that the newly available evidence continues to support the decision made in the
29 last review to maintain both annual and 24-hour standards for control of coarse-fraction particles,
August 2003 6-28 Draft - Do Not Quote or Cite
-------
1 while recognizing that the quantitative evidence related to long-term exposures remains very
2 limited.
3 In setting such standards, staff believes it is appropriate to use the same approach as
4 discussed above in section 6.3.2 for PM2 5 standards; that is, staff recommends that consideration
5 be given to setting a generally controlling annual PM10_2 5 standard in conjunction with a 24-hour
6 standard to provide supplemental protection against unusually high peak short-term
7 concentrations. This recommendation is supported by the results of EPA's updated risk
8 assessment, showing that peak 24-hour PM10_2 5 concentrations contribute a relatively small
9 amount to the total health risk associated with short-term exposures on an annual basis, such that
10 much if not most of the aggregated annual risk results from the large number of days during
11 which the 24-hour average concentrations are in the low- to mid-range, as discussed in section
12 4.3.3. In addition, staff recognizes that available short-term exposure studies do not provide
13 evidence of population thresholds, but rather reflect relationships between health effects and
14 ambient PM10_25 across a distribution of concentrations. Thus, as for PM25, staff concludes that
15 attempting to identify a "lowest observed effect level" in terms of a 24-hour concentration to
16 serve as a basis for establishing a generally controlling 24-hour PM10_2 5 standards would not be
17 an appropriate approach in this review.
18 On the other hand, staff recognizes that the available evidence is far less certain with
19 regard to potential long-term effects than for short-term effects. Thus, staff believes that
20 consideration also could be given to establishing only a 24-hour PM10_2 5 standard, possibly using
21 an alternative approach as discussed below in section 6.4.4..
22 In considering the available information, staff concludes that there is no basis for
23 considering any averaging times for PM10_2 5 standards other than annual and 24-hour averages.
24 While it is possible that peak concentrations of shorter duration may be associated with health
25 effects, there is little published evidence to support a PM10_2 5 standard with an averaging time of
26 less than 24 hours. In the following sections, staff provides recommendations regarding
27 alternative annual PM10_2 5 standards, followed by recommendations on alternative short-term
28 PM10_2 5 standards, either in combination with an annual standard or alone.
August 2003 6-29 Draft - Do Not Quote or Cite
-------
1 6.4.3 Alternative Annual PM10_2 5 Standards
2 In the last review, EPA's decision to retain the annual PM10 standard was based on the
3 same approach as was used in 1987 for establishing the standard at a level of 50 |ig/m3. That
4 approach considered the very limited quantitative evidence of associations linking higher long-
5 term concentrations of various indicators of PM with health effects, together with qualitative
6 evidence of particle deposition in the lung, to conclude that it is possible that cumulative
7 deposition of coarse-fraction particles could be of concern. Other supporting evidence included
8 qualitative information about long-term build-up of silica-containing materials and evidence
9 related to biological aerosols, which supported the need to limit coarse-fraction particles but
10 were not appropriately addressed by traditional air pollution control programs. Thus, this annual
11 standard was set to address effects potentially associated with long-term exposures, in
12 conjunction with a 24-hour standard set to address effects associated with short-term exposures.
13 In this review, staff is taking into consideration the still limited, but significantly
14 expanded evidence now available that directly links primarily short-term PM10_2 5 concentrations
15 with morbidity effects (e.g., respiratory effects and hospital admissions) and mortality. Staff
16 believes it is more appropriate to consider this evidence as a basis for setting a generally
17 controlling annual PM10_25 standard in conjunction with a supplemental 24-hour standard, as
18 discussed above in section 6.4.2, rather than to use the approach from the last review of looking
19 only at long-term exposure evidence as a basis for an annual standard. In so doing, staff is
20 generally taking into account the same factors as those considered for PM2 5, as discussed above
21 in section 6.3.3. Also as discussed above, mortality and morbidity effects estimates and reported
22 mean PM10_2 5 concentrations are presented in the table in Appendix A of Chapter 3, or are
23 discussed in the draft CD and in section 3.3 above (for the non-reanalyzed GAM studies).
24 Discussion about the extent to which reported associations are robust to the inclusion of co-
25 pollutants and the nature of the air quality monitoring used in the studies is generally
26 summarized in the draft CD and in section 3.3 above. The relative precision of the mortality and
27 morbidity PM10_2 5 studies is depicted above in Figures 3-6 and 3-9, respectively.
28 Based on this information from short-term exposure studies, and taking the factors
29 outlined in section 6.3.3 into consideration, staff makes the following observation:
August 2003 6-30 Draft - Do Not Quote or Cite
-------
1 • A number of short-term exposure studies reporting statistically significant associations
2 with mortality and morbidity effects have been conducted in areas in which the annual
3 mean 24-hour PM10_2 5 concentrations ranged from over 33 |ig/m3 down to approximately
4 11 |ig/m3.
5 Several of these studies reported significant associations that remained significant upon
6 reanalysis; and some were generally robust to the inclusion of co-pollutants in multi-pollutant
7 models, had adequate air quality data of PM10_25 mass, and resulted in relatively precise effects
8 estimates of health endpoints for which the weight of the entire body of evidence supports the
9 likelihood of an association with coarse-fraction particles. Other studies not included in the table
10 in Appendix A also showed statistically significant associations based on using a GAM function;
11 but, because reanalyses have not been done, it is unclear whether the associations would remain
12 significant with the use of more appropriate models. Taken together, these studies significantly
13 extend the evidence of PM10_2 5 associations beyond that available in the last review.
14 In considering this group of studies, staff has focused first on those studies where
15 significant mortality associations with PM10_25 were generally robust in models including
16 gaseous copollutants. Studies conducted in Phoenix (Mar et al., 1999; Mar et al., 2003) and
17 Coachella Valley, CA (Ostro et al., 2000; Ostro et al., 2003) reported relatively precise
18 statistically significant associations between PM10_25 and total or cardiovascular mortality where
19 mean PM10_25 concentrations ranged from approximately 33.2 to 30.5 |ig/m3, respectively. In
20 addition, a significant association was reported with total mortality in Steubenville (Schwartz et
21 al., 1996; Schwartz, 2003), where the annual mean PM10_25 concentration was approximately
22 16.1 |ig/m3, although this particular study did not assess the potential for confounding by co-
23 pollutants.
24 A few studies also reported significant associations with morbidity effects, including
25 respiratory and cardiovascular hospital admissions and respiratory symptoms. One such
26 reanalyzed study in Detroit (Lippmann et al., 2000; Ito, 2003) reported relatively precise
27 significant associations with hospital admissions for ischemic heart disease and pneumonia that
28 were generally robust to inclusion of co-pollutants, where the annual mean PM10_2 5 concentration
29 was approximately 13.3 |ig/m3. Another study in Seattle (Sheppard et al., 1999; Sheppard et al.,
30 2003) reported a relatively precise significant association with asthma hospital admissions where
August 2003 6-31 Draft - Do Not Quote or Cite
-------
1 the annual mean PM10_2 5 concentration was approximately 16.2 |ig/m3, although the results were
2 not robust to inclusion of co-pollutants, and were based on air quality mass measurements with
3 nephelometry data used to fill in missing days. A few other studies reporting significant
4 morbidity associations were either not reanalyzed or lasted less than a full year, such that annual
5 mean PM10_2 5 concentrations were not available.
6 Based on consideration of the available PM10_25 short-term exposure studies, staff
7 believes that it would be appropriate to consider a range of PM10_25 levels for an annual standard
8 that extends down from approximately 30 |ig/m3 to as low as 13 |ig/m3. Staff judges that
9 consideration of a level as low as 13 |ig/m3 would be precautionary and would place much
10 weight on the strength of the more limited and less consistent body of evidence now available on
11 coarse particles as compared to the available evidence on fine particles (discussed above in
12 section 6.3.3). At the lower end of this range, consideration of air quality data measurement
13 methods, potential confounding by co-pollutants, and uncertainties in appropriate model
14 selection adds considerable uncertainty to the use of study results as a basis for quantitative
15 standard-setting decisions. Staff notes that these uncertainties are generally greater with regard
16 to evidence from coarse particle studies than from fine particle studies.
17 Beyond the short-term exposure PM10_25 studies, staff has also considered the results of
18 currently available long-term exposure studies, and makes the following observation:
19 • Since the last review, only one long-term exposure study has become available that has
20 assessed associations between PM10_2 5 and morbidity effects, showing limited evidence of
21 an association with reduced lung function growth in children.
22 This study (Gauderman et al., 2000) reported a significant association between long-term PM10_2 5
23 concentrations (as well as PM10 and PM2 5) and reduced lung function growth in children across
24 12 southern California communities, although the study results were mixed and limited and the
25 authors found it difficult to separate the effects of the different PM indicators. Based on the
26 graphical presentation of PM10_25 concentrations in the study (from which a mean PM10_25
27 concentration of roughly 20 |ig/m3 can be inferred), staff concludes that this study provides
28 suggestive evidence of possible long-term PM10_2 5 respiratory effects within the range of annual
29 mean PM10_2 5 concentrations recommended for consideration based on the short-term exposure
30 study evidence (13 to 30 |ig/m3), suggesting that an annual standard at the upper end of that
August 2003 6-32 Draft - Do Not Quote or Cite
-------
1 range may not be protective against possible long-term effects. Staff notes, however, that other
2 long-term exposure studies do not provide evidence of PM10_25 mortality or morbidity effects,
3 suggesting that it may be appropriate to place little weight on such effects in considering a
4 standard within this range.
5 Taking into account the above observations from both the short- and long-term exposure
6 studies, staff concludes that the findings of new epidemiologic studies support consideration of a
7 range from approximately 30 |ig/m3 down to 13 |ig/m3 in setting a generally controlling annual
8 PM10_2 5 standard. As a caution in judging the weight to be placed on the studies discussed
9 above, staff notes that the overall body of evidence for PM10_25 health effects associations is
10 more limited and less consistent than the evidence for PM25 associations, especially with regard
11 to mortality effects. In making a preliminary recommendation on a range of annual PM10_2 5
12 standard levels for consideration, staff recognizes that information on the estimated reduction in
13 health risks associated with just attaining alternative standards in this range is not yet available,
14 and staff will consider such information in presenting conclusions and recommendations in the
15 next draft Staff Paper.
16 In addition to considering the level of an annual standard, consideration must also be
17 given to the form of the standard. Staff sees no basis at this time for recommending any revision
18 to the form of the current annual PM10 standard, which is based on an annual arithmetic mean,
19 averaged over 3 years, measured at each monitor within an area. However, as noted above in
20 section 6.3.3, staff is reconsidering the appropriateness of allowing for spatial averaging across
21 monitors as part of the form of an annual PM25 standard. In further assessing this issue, staff
22 also intends to consider the issue of spatial averaging with regard to the form of a PM10_2 5 annual
23 standard.
24 6.4.4 Alternative 24-Hour PM10_2 5 Standards
25 In the last review, EPA's decision to retain the level of the 24-hour PM10 standard of 150
26 Hg/m3 (with revision to the form of the standard) was based on two community studies of
27 exposure to fugitive dust that showed health effects only in areas experiencing large exceedances
28 of that standard, as well as on qualitative information regarding the potential for health effects
29 related to short-term exposure to coarse-fraction particles. Because of the very limited nature of
August 2003 6-33 Draft - Do Not Quote or Cite
-------
1 this evidence, staff concluded that while it supported retention of a standard to control coarse-
2 fraction particles, it provided no basis for considering a more protective standard. However,
3 because of concerns about the expected-exceedance-based form of the 1987 PM10 standard,
4 primarily related to the stability of the attainment status of an area over time and complex data
5 handling conventions needed in conjunction with less-than-every-day sampling, EPA adopted a
6 concentration-based form for the 24-hour standard, as was done for the 24-hour PM2 5 standard,
7 as discussed above in section 6.3.4. In making this change, EPA selected a 99th percentile form,
8 in contrast to the 98th percentile form adopted for the 24-hour PM2 5 standard, so as not to allow
9 any relaxation in the level of protection that had been afforded by the previous 1-expected-
10 exceedance form.
11 As discussed above in section 6.4.3, since the last review, new evidence has become
12 available that directly links primarily short-term PM10_2 5 concentrations with morbidity effects
13 and mortality. In considering this evidence as a basis for setting a 24-hour PM10_25 standard,
14 staff has focused primarily on alternative 24-hour standards that would supplement the
15 protection afforded by the range of recommended annual standards, consistent with the approach
16 used in considering alternative 24-hour PM25 standards.
17 In considering alternative levels for a supplemental 24-hour PM10_2 5 standard, staff has
18 used the same two approaches that were used to consider alternative levels for the 24-hour PM2 5
19 standard, discussed above in section 6.3.4. In considering these approaches as a basis for
20 identifying alternative levels for a 24-hour PM10_25 standard, staff has focused on standards with
21 a 98th percentile form, consistent with the form of the 24-hour PM2 5 standard. While alternative
22 forms could also be considered in this review, staff believes there is good reason to make the
23 forms consistent across standards, given that the reasoning behind setting different forms in the
24 last review is no longer relevant.
25 The first approach focuses on the upper end of the distribution of daily PM10_2 5
26 concentrations from short-term exposure studies reporting statistically significant health effects
27 associations where the annual mean concentrations were within the staff-recommended range of
28 annual standard levels. In using this approach, staff has focused on those short-term exposure
29 studies discussed above in section 6.4.3 that provided the basis for the range of recommended
30 alternative annual standards. The upper ends of the PM10_25 concentration ranges from these and
August 2003 6-34 Draft - Do Not Quote or Cite
-------
1 other studies were obtained from the published study or from the study authors and are
2 documented in Ross (2003). In particular, staff has focused on the upper end of the
3 concentration ranges for studies reporting statistically significant effects where annual mean
4 PM2 5 concentrations ranged from approximately 13 |ig/m3 to over 30 |ig/m3. Looking first at the
5 range of 98th percentile concentrations in these studies, the concentrations generally ranged from
6 over 30 |ig/m3 to approximately 70 |ig/m3, with one study having a 98th percentile concentration
7 over 100 |ig/m3. Looking more broadly at the 95th to 99th percentile concentrations in these
8 studies, the concentrations generally ranged from somewhat below 30 |ig/m3 to approximately 75
9 Hg/m3, with one study having a 99th percentile concentration over 130 |ig/m3. Taken together,
10 this information suggests consideration of 24-hour PM10_25 standard levels ranging from
11 approximately 30 to 75 |ig/m3 to supplement the protection afforded by the range of
12 recommended annual standards.
13 The second approach focuses on an assessment of recent PM10_25 air quality data that
14 presents the distributions of 98th percentile 24-hour average PM10_25 concentrations as a function
15 of annual mean PM10_25 concentrations (shown in Chapter 2, Figure 2-21). The information
16 presented in this figure provides some indication of the extent to which specific 24-hour standard
17 levels would likely result in some degree of supplemental protection, beyond that provided by an
18 annual standard alone, against peak exposures at sites that would likely attain specific annual
19 standard levels. In using this approach, staff has looked at the information in Figure 2-21, based
20 on an assessment of air quality data from 2000 to 2002 at almost 500 monitoring sites, to
21 determine what 24-hour standard levels would likely result in some degree of supplemental
22 protection relative to the protection afforded by an annual standard alone, for annual standards
23 within the range of 13 to 30 |ig/m3. As an initial matter, staff observes that alternative 24-hour
24 standard levels based on this approach are highly dependent upon the level of the annual
25 standard being considered. For example, for an annual standard at the lower end of the range,
26 the top 5th percentile values from the distributions are generally with the range of approximately
27 40 |ig/m3 up to about 60 |ig/m3. For an annual standard around the middle of the range, the top
28 5th percentile values are generally with the range of approximately 50 |ig/m3 up to about 70
29 ng/m3. At the upper end of the range of alternative annual standard levels, the top 5th percentile
30 values are generally around 70 |ig/m3 or above.
August 2003 6-3 5 Draft - Do Not Quote or Cite
-------
1 To assist in understanding the public health implications of various combinations of
2 alternative annual and 24-hour standards, staff assessed the PM10_2 5 air quality database in the
3 same manner as was done for the PM2 5 database, as discussed above in section 6.3.4. The
4 results of this assessment, based on air quality data from 351 counties (including 488 monitoring
5 sites), is shown in Table 6-2. (To provide a rough comparison with the current PM10 standards,
6 similar information is also provided in this table for the current combined PM10 standards.)
7 From this table it can be seen that for an annual PM10_25 standard set at 13 |ig/m3, 24-hour
8 standard levels ranging from 60 to 40 |ig/m3 would add approximately 1 to 5 % to the percentage
9 of counties that would not likely attain both standards relative to the number of counties that
10 would not likely attain the annual standard alone. Similar increments are seen for an annual
11 PM10_2 5 standard set at 16 |ig/m3, with 24-hour standard levels ranging from 70 to 50 |ig/m3. For
12 an annual PM10_2 5 standard set at 30 |ig/m3, 24-hour standard levels at and above 70 |ig/m3 would
13 add approximately 4% or less to the percentage of counties that would not likely attain the
14 annual standard alone.
15 Staff believes that both approaches provide useful information to frame the range of
16 alternative 24-hour standards that are appropriate for consideration in this review. By combining
17 the ranges suggested by both approaches, staff makes a preliminary recommendation that
18 consideration be given in this review to setting a 24-hour PM10_2 5 standard, with a 98th percentile
19 form, within a range of levels from 30 |ig/m3 to 75 |ig/m3. Staff recognizes that selecting a 24-
20 hour PM10_2 5 standard from within this range should take into account the degree of protection
21 afforded by the selected annual PM10_2 5 standard.
August 2003 6-36 Draft - Do Not Quote or Cite
-------
1 TABLE 6-2 PREDICTED PERCENTAGE OF COUNTIES WITH MONITORS NOT LIKELY
2 TO MEET ALTERNATIVE PM,n 2, and CURRENT PM,n STANDARDS
Alternative
Standards and
Levels (ug/m3)
No. of Counties with
PMmjS Monitors
3M10.25: Annual
30
20
16
13
Percent of counties not likely to meet alternative PM10_25 standards *
Total
351
3
9
19
31
Region 1 :
Northeast
53
0
0
9
13
Region 2:
Southeast
70
0
1
7
14
Region 3:
Industrial
Midwest
73
0
4
7
14
Region 4:
Upper
Midwest
29
3
14
31
52
Region 5:
Southwest
19
21
37
42
53
Region 6:
Northwest
83
0
4
22
46
Region 7:
Southern
CA
14
21
64
64
73
Outside
Regions*
*
1C
10
40
70
80
>M10.25: 24-Hour
75
70
60
50
40
35
4
6
11
18
29
36
0
0
2
13
23
25
1
1
7
7
11
20
1
4
7
8
15
21
10
14
14
31
55
66
21
26
37
37
47
53
2
5
7
17
35
46
21
21
50
57
64
71
10
20
50
80
90
90
DM10_25: Combined
\nnual/24-Hour:
30/75
30/70
30/60
5
7
11
0
0
2
1
1
7
1
4
7
10
14
14
21
26
37
2
5
7
29
29
50
20
20
50
20/70
20/60
20/50
11
13
19
0
2
13
3
7
7
4
7
8
17
17
31
37
37
37
8
10
18
60
60
60
40
50
80
16/70
16/60
16/50
20
21
24
9
9
15
9
11
11
7
8
10
34
34
38
42
42
42
23
24
29
64
64
64
70
70
90
13/60
13/50
13/40
No. of Counties with
PMW Monitors
DM10: Combined
\nnual/24-Hour
50/150
32
33
36
574
3
13
19
25
82
0
16
16
16
118
0
14
15
18
117
1
55
55
69
50
6
53
53
53
30
17
46
47
47
141
2
79
79
79
16
50
80
90
90
2C
0
August 2003
6-37
Draft - Do Not Quote or Cite
-------
1 * Based on 4, 8, or 12 consecutive quarters of data with at least 11 samples per quarter from the 2000 to 2002
2 database. As such, these estimates are not based on the same amount of air quality data that would be used to
3 determine whether an area would attain a given standard or set of standards. These estimates can only approximate
4 the number of counties that are likely not to attain the given standards and should be interpreted with caution.
5 ** "Outside Regions" includes Alaska, Hawaii, Puerto Rico, and the Virgin Islands.
6 Staff has also considered alternative 24-hour PM10_25 standards that could serve as the
7 sole means of protecting against known and potential effects associated with short- and long-
8 term exposures to PM10_25. In so doing, staff recognizes that since currently available short-term
9 exposure studies provide no evidence of population thresholds, it would clearly not be
10 appropriate to attempt to identify a "lowest observed effect level" in terms of a 24-hour
11 concentration to serve as a basis for establishing a 24-hour PM10_2 5 standard as the sole standard
12 for coarse-fraction particles. Rather, staff recommends consideration of the same range of 24-
13 hour PM10_2 5 levels identified above, on the basis that such standards could potentially provide a
14 roughly equivalent degree of protection on a national scale as would the generally controlling
15 annual standards combined with the supplemental 24-hour standards, although clearly the
16 distribution of counties not likely to attain would be somewhat different across the various
17 regions presented in Table 6-2. In light of the very limited information on potential long-term
18 effects, staff believes that consideration of a sole 24-hour PM10_25 standard may be appropriate,
19 although staff recognizes that using the currently available evidence to select such a standard
20 may well be more difficult than it would be to select from a range of alternative annual standards
21 in conjunction with a supplemental 24-hour standard.
22 In making the above preliminary recommendations on a range of 24-hour standards,
23 either in combination with an annual standard or alone, staff recognizes that information on the
24 estimated reduction in health risks associated with just attaining alternative standards in this
25 range is not yet available. Staff will consider such information in presenting conclusions and
26 recommendations in the draft of this Staff Paper.
August 2003 6-3 8 Draft - Do Not Quote or Cite
-------
1 6.5 SUMMARY OF STAFF RECOMMENDATIONS ON PRIMARY PM NAAQS
2 The major staff recommendations and supporting conclusions from sections 6.2 through
3 6.4 are briefly summarized below:
4 1. Consideration should be given to revising the current PM2 5 primary standards to provide
5 increased public health protection from fine particles based primarily on newly available
6 evidence of mortality and morbidity health effects in areas where the annual mean
7 concentrations are below the level of the current annual PM25 standard.
8 2. The current PM10 primary standards should be revised and replaced with standard(s) that
9 are defined in terms of an indicator that includes the coarse fraction without including the
10 fine fraction. Such standards should be based primarily on evidence now available
11 showing health effects associations with PM10_2 5.
12 3. The suite of primary PM NAAQS should continue to focus on ambient particles capable
13 of penetrating to the thoracic region, with separate standards for the fine and coarse
14 fractions of PM10. Staff concludes that the available information continues to support the
15 use of PM25 as the indicator for fine particles and supports the use of PM10_25 as the
16 indicator for coarse particles.
17 4. Staff recommends that PM25 primary standards should continue to be based on both
18 annual and 24-hour averaging times.
19 a. Staff recommends that the Administrator consider selecting the level of the annual
20 PM25 primary standard from within a range of 15 |ig/m3 to approximately 12
21 Hg/m3. This recommended range is based primarily on quantitative results from
22 both short- and long-term exposure studies in the U.S. and Canada, as well as
23 taking into account the strength of the entire body of evidence showing health
24 effects associated with exposure to PM2 5 and the uncertainties and limitations
25 inherent in evidence and in the methodological approaches available for
26 interpreting the evidence. This recommended range is based on an approach that
27 continues to view the annual standard as the generally controlling standard
28 intended to reduce the entire distribution of PM25 concentrations so as to provide
29 public health protection by reducing risks associated with both short- and long-
30 term exposures.
August 2003 6-39 Draft - Do Not Quote or Cite
-------
1 b. Staff recommends that the Administrator consider revising the 24-hour PM2 5
2 primary standard, selecting the level from within a range of approximately 50
3 Hg/m3 to 30 |ig/m3. This recommended range is based on a combination of
4 approaches that focus on the upper end of the PM25 concentration distributions
5 both in key U.S. and Canadian epidemiologic studies and from the network of
6 current PM2 5 monitoring sites. These approaches are consistent with the
7 continuing view that the 24-hour standard should be set so as to provide
8 supplemental protection against peak exposures not addressed by the annual
9 standard. Staff recommends that consideration be given to retaining the current
10 98th percentile form of the standard.
11 5. Staff recommends that new PM10_25 primary standards be established for both annual and
12 24-hour averaging times, with possible consideration given to establishing just a 24-hour
13 standard.
14 a. Staff recommends that the Administrator consider selecting the level of the annual
15 PM10_2.5 primary standard from within a range of approximately 30 |ig/m3 to
16 approximately 13 |ig/m3. This recommended range is based primarily on
17 quantitative results from both short- and long-term exposure studies in the U.S.
18 and Canada, as well as taking into account the strength of the entire body of
19 evidence showing health effects associated with exposure to PM10_2 5 and the
20 uncertainties and limitations inherent in evidence and in the methodological
21 approaches available for interpreting the evidence. This recommended range is
22 based on the same approach used for the annual PM2 5 standard, in that it views
23 the annual standard as the generally controlling standard intended to reduce the
24 entire distribution of PM10_25 concentrations so as to provide public health
25 protection by reducing risks associated with short-term and possibly long-term
26 exposures.
27 b. Staff recommends that the Administrator consider selecting the level of a 24-hour
28 PM10_25 primary standard from within a range of approximately 75 |ig/m3 to 30
29 Hg/m3. Consistent with the approach used for the 24-hour PM2 5 standard, this
30 recommended range is based on a combination of approaches that focus on the
August 2003 6-40 Draft - Do Not Quote or Cite
-------
1 upper end of the PM10_25 concentration distributions both in key U.S. and
2 Canadian epidemiologic studies and from the network of current PM10 and PM2 5
3 monitoring sites. These approaches are consistent with the continuing view that
4 the 24-hour standard should be set so as to provide supplemental protection
5 against peak exposures not addressed by the annual standard. Alternatively, if
6 consideration is given to establishing a 24-hour standard as the sole standard for
7 PM10_25, staff recommends that consideration be given to selecting such a standard
8 from within the same range. In either case, staff recommends that consideration
9 be given to setting a 24-hour PM10_2 5 standard in terms of a 98th percentile form,
10 consistent with the form of the 24-hour PM2 5 standard.
11 6. In recommending these primary standards and ranges of levels, staff is mindful that the
12 Act requires standards to be set that are requisite to protect public health with an
13 adequate margin of safety, such that the standards are to be neither more nor less
14 stringent than necessary. Thus, the Act does not require that NAAQS be set at zero-risk
15 levels, but rather at levels that avoid unacceptable risks to public health.
16 6.6 ALTERNATIVE SECONDARY STANDARDS FOR PM
17 6.6.1 Alternative Standards for Visibility Protection
18 In 1997, EPA decided to address the effects of PM on visibility by setting secondary
19 standards identical to the suite of PM25 primary standards, in conjunction with the establishment
20 of a regional haze program under sections 169A and 169B of the Act (62 FR at 38,679-83). In
21 reaching this decision, EPA first concluded that PM, especially in the fine fraction, produces
22 adverse effects on visibility in various locations across the country, including multi-state regions,
23 urban areas, and remote Class I Federal areas (e.g., national parks and wilderness areas). EPA
24 also concluded that addressing visibility impairment solely through setting more stringent
25 national secondary standards would not be an appropriate means to protect the public welfare
26 from adverse impacts of PM on visibility in all parts of the country. As a consequence, EPA
27 determined that an approach that combined national secondary standards with a regional haze
28 program was the most appropriate and effective way to address visibility impairment.
August 2003 6-41 Draft - Do Not Quote or Cite
-------
1 In reaching these conclusions, EPA recognized that the selection of an appropriate level
2 for a national secondary standard to address visibility protection was complicated by regional
3 differences in visibility impairment due to several factors, including background and current
4 levels of PM, the composition of PM, and average relative humidity. As a result of these
5 regional differences, EPA noted that a national standard intended to maintain or improve
6 visibility conditions in many parts of the West would have to be set at or below natural
7 background levels in the East; conversely, a national standard that would improve visibility in
8 the East would permit further degradation in the West. Beyond such problems associated with
9 regional variability, EPA also determined that there was not sufficient information available to
10 establish a standard level that would represent a threshold above which visibility conditions
11 would always be adverse and below which conditions would always be acceptable.
12 These considerations led EPA to assess whether the protection afforded by the
13 combination of the selected primary PM25 standards and a regional haze program would provide
14 appropriate protection against the effects of PM on visibility. Based on such an assessment,
15 EPA determined that attainment of the primary PM25 standards through the implementation of
16 regional control strategies would be expected to result in visibility improvements in the East at
17 both urban and regional scales, but little or no change in the West, except in and near certain
18 urban areas. Further, EPA determined that a regional haze program that would make significant
19 progress toward the national visibility goal in Class I areas would also be expected to improve
20 visibility in many urban and non-Class I areas as well. EPA also noted, however, that the
21 combined effect of the PM NAAQS and regional haze programs may not address all situations in
22 which people living in certain urban areas may place a particularly high value on unique scenic
23 resources in or near these areas. EPA concluded that such situations were more appropriately
24 and effectively addressed by local visibility standards, such as those established by the city of
25 Denver, than by national standards and control programs.
26 As called for in the last review, EPA promulgated a regional haze program in 1999. That
27 program requires States to establish goals for improving visibility in Class I areas and to adopt
28 control strategies to achieve these goals. More specifically, States are required to establish goals
29 for improving visibility on the 20% most impaired days in each Class I area, and for allowing no
30 degradation on the 20% least impaired days. Since strategies to meet these goals are to reflect a
August 2003 6-42 Draft - Do Not Quote or Cite
-------
1 coordinated approach among States, multistate planning organizations have been formed and are
2 now developing strategies, to be adopted in the 2003 to 2008 time period, that will make
3 reasonable progress in meeting these goals. Thus, although the regional haze program is now
4 moving forward, there is as yet little additional information available as to the visibility
5 improvements likely to result from this program in urban areas and other non-Class I areas.
6 In considering the information available in this review, as discussed in section 5.2, staff
7 notes that while new research has led to improved understanding of the optical properties of
8 particles and the effects of relative humidity on those properties, it has not changed the
9 fundamental characterization of the role of PM in visibility impairment from the last review.
10 Further, staff notes that regional factors continue to complicate any effort to identify an
11 appropriate level for a national secondary standard to address visibility protection. Thus, staff
12 continues to conclude that PM, especially in the fine fraction, produces adverse effects on
13 visibility in various locations across the country, and that addressing visibility impairment solely
14 through setting more stringent national secondary standards would not be appropriate.
15 Information now available from visibility and fine particle monitoring networks allows
16 for an updated characterization of visibility trends and current levels in Class I areas and for
17 future analyses to better characterize visibility impairment in urban areas. In considering the
18 most recent monitoring information, as discussed in section 5.2, staff makes the following
19 observations:
20 • In Class I areas, visibility levels on the 20% haziest days in the West are about equal to
21 levels on the 20% best days in the East. Despite improvement through the 1990's,
22 visibility in the rural East remains significantly impaired, with an average visual range of
23 approximately 20 km on the 20% haziest days (compared to the naturally occurring
24 visual range of about 150 + 45 km). In the rural West, the average visual range showed
25 little change over this period, with an average visual range of approximately 100 km on
26 the 20% haziest days (compared to the naturally occurring visual range of about 230 + 40
27 km).
28 • Urban areas generally have higher visibility impairment than Class I areas. Although
29 urban visibility remains poorly characterized at this time, automated visibility
30 measurements are now available for analysis in the form of high-resolution data from the
31 ASOS network located at major airports to better characterize visibility in urban areas.
32 Initial analysis of ASOS data, as discussed in section 5.2.4.2, shows promise in
33 developing correlations between urban visibility data and PM2 5 concentrations, although
34 more analysis is needed.
August 2003 6-43 Draft - Do Not Quote or Cite
-------
1 New information is also now available from local regulatory efforts in the U.S. and
2 abroad to establish standards and programs to address specific visibility concerns in particular
3 urban and non-urban areas, as discussed in section 5.2.5. These efforts have produced new
4 methods and tools to communicate and evaluate public perceptions about varying visual effects
5 associated with alternative levels of visibility impairment relative to varying pollution levels and
6 environmental conditions. Methods involving the use of focus groups to elicit citizen judgments
7 about the acceptability of varying levels of visual air quality in an urban area have been
8 developed by the State of Colorado, and used to develop a visibility standard for Denver. These
9 methods have now been adapted and applied in other areas, including Phoenix, AZ, and the
10 province of British Columbia, Canada, producing reasonably consistent results in terms of the
11 visual ranges found to be generally acceptable by the participants in the various studies. These
12 applications and results lend support to staffs conclusion that this approach holds promise for
13 developing information in the future that will help inform judgments about acceptable urban
14 visibility levels. Although resource constraints have prevented additional development of this
15 approach since staff conducted the initial pilot study that was presented in the preliminary draft
16 Staff Paper (EPA, 2001), EPA hopes to pursue it in the future so the results of a more extensive
17 study can be used to help inform the next periodic review of the PM secondary standards.
18 Until such methods have been sufficiently developed and more broadly applied, staff
19 believes it is appropriate to make use of photographic representations of visibility impairment to
20 help inform judgments about the acceptability of varying levels of visual air quality in urban
21 areas. As discussed in section 5.2, photographic representations of varying levels of visual air
22 quality have been developed for several urban areas, in conjunction with efforts to develop local
23 programs and with EPA's initial pilot study, and are available on EPA's website
24 (http:/www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_sp.htmn as an attachment to this
25 document. In considering these images for Washington, D.C., Chicago, Denver, and Phoenix,
26 staff makes the following observations:
27 • At concentrations at or near the level of the current 24-hour PM2 5 standard, scenic views
28 (e-g-, mountains, historic monuments) as depicted in these images around and within the
29 urban areas are significantly obscured from view.
August 2003 6-44 Draft - Do Not Quote or Cite
-------
1 • Appreciable improvement in the visual clarity of the scenic views depicted in these
2 images occurs at concentrations toward the lower end of the staff-recommended range of
3 consideration for the 24-hour primary PM2 5 standards.
4 In making these observations, staff recognizes the limitations in using such visual
5 representations as a basis for selecting a level for a visibility-based PM2 5 secondary standard.
6 Nonetheless, staff believes that these observations, together with the information discussed
7 above and considered in the last review, support consideration of revising the current PM25
8 secondary standards to be consistent with any revisions made to the PM2 5 primary standards that
9 would afford greater visibility protection, especially toward the lower end of the staff-
10 recommended ranges for the primary standards. As in the last review, staff believes that such
11 standards should be considered in conjunction with the regional haze program now being
12 implemented as an effective approach to providing appropriate protection against PM-related
13 visibility impairment across the country. Staff further believes that the development of local
14 programs continues to be an effective and appropriate approach to provide additional protection
15 for unique scenic resources in and around certain urban areas that are particularly highly valued
16 by people living in those areas. In addition, staff expects that the results of any future local
17 surveys of citizens' judgments as to the acceptability of varying visual air quality levels will
18 provide useful information for consideration in subsequent PM NAAQS reviews.
19 6.6.2 Alternative Standards for Other Welfare Effects
20 EPA's decision in 1997 to revise the suite of PM secondary standards took into account
21 not only visibility protection, but also the other PM-related welfare effects considered in the last
22 review (i.e., materials damage and soiling). Based on this broader consideration, EPA
23 established secondary standards for PM identical to the suite of primary standards, including
24 both PM2 5 and PM10 standards, to provide appropriate protection against the welfare effects
25 associated with fine and coarse particle pollution (62 FR at 38,683). This decision was based on
26 considering both visibility effects associated with fine particles, as discussed above in section
27 6.6.1, and materials damage and soiling effects associated with both fine and coarse particles.
28 With regard to effects on materials, EPA concluded that both fine and coarse particles can
August 2003 6-45 Draft - Do Not Quote or Cite
-------
1 contribute to materials damage and soiling effects. However, EPA also concluded that the
2 available data did not provide a sufficient basis for establishing a distinct secondary standard
3 based on materials damage or soiling alone. These considerations led EPA to address whether
4 the reductions in fine and coarse particles likely to result from the suite of PM primary standards
5 would provide appropriate protection against the effects of PM on materials. Taking into
6 account the available information and the limitations in that information, EPA judged that setting
7 secondary standards identical to the suite of PM25 and PM10 primary standards would provide
8 increased protection against the effects of fine particles and retain an appropriate degree of
9 control on coarse particles.
10 In this review, the draft CD and this draft Staff Paper have been broadened in scope to
11 address not only PM-related effects on materials, but also effects on ecosystems and vegetation,
12 as discussed above in section 5.4. In addition, staff has included in this draft Staff Paper
13 consideration of the role of ambient PM in atmospheric process associated with climate change
14 and the transmission of solar radiation, as discussed above in section 5.5. In considering the
15 currently available evidence on each of these types of PM-related welfare effects, staff notes that
16 there is much information linking ambient PM to potentially adverse effects on materials and
17 ecosystems and vegetation, and on characterizing the role of atmospheric particles in climatic
18 and radiative processes. However, based on the limitations in the evidence, especially with
19 regard to the lack of evidence linking various effects to specific levels of ambient PM, staff
20 concludes that the available evidence does not provide a sufficient basis for establishing distinct
21 secondary standards based on any of these effects alone. These considerations lead staff to
22 address whether the reductions in fine and coarse particles likely to result from the current
23 secondary standards or the range of recommended revised PM25 and PM10_25 primary standards
24 would provide appropriate protection against these types of PM-related welfare effects.
25 With regard to PM-related effects on materials, staff notes that the available evidence
26 continues to support the following observations:
27 • Materials damage and soiling that occur through natural weathering processes are
28 enhanced by exposure to atmospheric pollutants, most notably SO2 and particulate
29 sulfates.
August 2003 6-46 Draft - Do Not Quote or Cite
-------
1 • While ambient particles play a role in the corrosion of metals and in the weathering of
2 paints and building materials, no quantitative relationships between ambient particle
3 concentrations and rates of damage have been established.
4 • Similarly, while soiling associated with fine and coarse particles can result in increased
5 cleaning frequency and repainting of surfaces, no quantitative relationships between
6 particle characteristics (e.g., concentrations, particle size, and chemical composition) and
7 the frequency of cleaning or repainting have been established.
8 Staff believes that these observations and the underlying available evidence continue to support
9 consideration of retaining an appropriate degree of control on both fine and coarse particles.
10 Lacking any specific quantitative basis for establishing distinct standards to protect against PM-
11 related adverse effects on materials, staff recommends consideration be given to (1) retaining the
12 current PM2 5 secondary standards or to revising those standards to be consistent with any
13 revisions made to the current PM2 5 primary standards, and (2) retaining secondary standards for
14 coarse particles, using a PM10_2 5 indicator consistent with the primary standards, at a level that
15 either retains the degree of protection afforded by the current PM10 standards or that is consistent
16 with any new PM10_2 5 primary standards.
17 With regard to PM-related effects on ecosystems and vegetation, staff notes that the draft
18 CD presents evidence of such effects, particularly related to nitrate and acidic deposition, and
19 concludes that current PM levels in the U.S. "have the potential to alter ecosystem structure and
20 function in ways that may reduce their ability to meet societal needs" (CD, p. 4-153). Much of
21 the associated uncertainty surrounding the characterization of the relationships between ambient
22 PM levels and ecosystem or vegetation responses is related to the extreme complexity and
23 variability that exist in predicting particle deposition rates, which are affected by particle size
24 and composition, associated atmospheric conditions, and the properties of the surfaces being
25 impacted. Though several national deposition monitoring networks have been successfully
26 measuring wet and dry deposition for several decades, they often do not distinguish the form
27 (e.g., particle, wet, and dry gaseous) in which a given chemical species is deposited, so that it is
28 difficult to know what percentage of total deposition is attributable to ambient PM. Further, data
29 from monitoring sites generally do not address all the variables affecting deposition that come
30 into play in a natural system.
August 2003 6-47 Draft - Do Not Quote or Cite
-------
1 In addition to these uncertainties, many of the documented PM-related ecosystem-level
2 effects only became evident after long-term, chronic exposures to specific chemical
3 constituent(s) of PM eventually exceeded the natural buffering or assimilative capacity of the
4 system. In most cases, PM deposition is not the only source of the chemical species to the
5 affected system and the percentage of the deposition due to ambient PM is often not known.
6 Because ecosystems have different sensitivities and capacities to buffer or assimilate pollutants,
7 it is difficult to predict the rate of deposition that would be likely to lead to the observed adverse
8 effects within any particular ecosystem. Equally difficult is the prediction of recovery rates for
9 already affected areas if deposition of various chemical species were to be reduced.
10 Despite these uncertainties, a number of significant and adverse environmental effects
11 that either have already occurred or are currently occurring have been linked to chronic
12 deposition of chemical constituents found in ambient PM. Staff notes, for example, that the
13 following effects have been linked with chronic additions of nitrate and its accumulation in
14 ecosystems:
15 • Productivity increases in forests and grasslands, followed by decreases in productivity
16 and possible decreases in biodiversity in many natural habitats wherever atmospheric
17 reactive nitrogen deposition increases significantly and critical thresholds are exceeded;
18 • Acidification and loss of biodiversity in lakes and streams in many regions, especially in
19 conjunction with sulfate deposition; and
20 • Eutrophication, hypoxia, loss of biodiversity, and habitat degradation in coastal
21 ecosystems.
22 Staff notes that effects of acidic deposition have been extensively documented, as
23 discussed in the draft CD and other reports referenced therein. For example, effects on some
24 species of forest trees linked to acidic deposition include increased permeability of leaf surfaces
25 to toxic materials, water, and disease agents; increased leaching of nutrients from foliage; and
26 altered reproductive processes; all of which serve to weaken trees so that they are more
27 susceptible to other stresses (e.g., extreme weather, pests, pathogens). In particular, acidic
28 deposition has been implicated as a causal factor in the northeastern high-elevation decline of red
29 spruce. Although U.S. forest ecosystems other than the high-elevation spruce-fir forests are not
30 currently manifesting symptoms of injury directly attributable to acid deposition, less sensitive
August 2003 6-48 Draft - Do Not Quote or Cite
-------
1 forests throughout the U.S. are experiencing gradual losses of base cation nutrients, which in
2 many cases will reduce the quality of forest nutrition over the long term.
3 Further, staff notes that chronic exposures to some metal constituents of ambient particles
4 may be linked to adverse effects on vegetation. For example, although there has been no direct
5 evidence of a physiological association between tree injury and exposure to metals, metals have
6 been implicated because their deposition pattern has been correlated with the decline of certain
7 tree species.
8 Taking into account the available evidence linking chemical constituents of both fine and
9 coarse PM to these types of known and potential adverse effects on ecosystems and vegetation,
10 staff believes that further reductions in ambient PM would likely contribute to long-term
11 recovery and prevent further degradation of sensitive ecosystems and vegetation. Staff
12 recognizes, however, that the available evidence does not provide any quantitative basis for
13 establishing distinct national standards for ambient PM. Further, staff recognizes that due to
14 site-specific sensitivities to various components of ambient PM, differing buffering and
15 assimilative capacities, and local and regional differences in the percentage of total deposition
16 that is likely attributable to ambient PM, national standards alone would not be an appropriate
17 means to protect against adverse impacts of ambient PM on ecosystems and vegetation in all
18 parts of the country. Nonetheless, staff believes that reductions in fine and coarse particles likely
19 to result from the current suite of secondary standards or the range of recommended revised
20 PM2 5 and new PM10_2 5 primary standards would contribute to increased protection against PM-
21 related effects on ecosystems and vegetation. Staff recommends that the potential for increased
22 protection of ecosystems and vegetation be taken into account in considering whether to revise
23 the current PM secondary standards consistent with any revisions that may be made to the PM
24 primary standards. Further, staff believes that any such increased protection should be
25 considered in conjunction with protection afforded by other programs intended to address
26 various aspects of air pollution effects on ecosystems and vegetation, such as the Acid
27 Deposition Program and other regional approaches to reducing pollutants linked to nitrate or
28 acidic deposition.
29 With regard to the role of ambient PM in climate change processes and in altering the
30 penetration of solar UV-B radiation to the earth's surface, staff notes that information available
August 2003 6-49 Draft - Do Not Quote or Cite
-------
1 in this review derives primarily from broad-scale research and assessments related to the study of
2 global climate change and stratospheric ozone depletion. As such, this information is generally
3 focused on global- and regional-scale processes and impacts and provides essentially no basis for
4 characterizing how differing levels of ambient PM in areas across the U.S. would affect local,
5 regional, or global climatic changes or alter the penetration of UV-B radiation to the earth's
6 surface. As noted in section 4.5.3, even the direction of such effects on a local scale remains
7 uncertain. Moreover, similar concentrations of different particle components can produce
8 opposite net radiative effects. Thus, staff concludes that there is insufficient information
9 available to help inform consideration of whether any revisions of the current PM secondary
10 standards are appropriate at this time based on ambient PM's role in atmospheric processes
11 related to climate or the transmission of solar radiation.
12 6.6.3 Summary of Staff Recommendations on Secondary PM NAAQS
13 The major staff recommendations and supporting conclusions from section 6.5 are briefly
14 summarized below:
15 1. Anthropogenic particles especially in the fine fraction impair visibility in various
16 locations across the country, although the selection of an appropriate level for a national
17 secondary standard to address visibility protection is complicated by regional differences
18 in several factors, including background and current levels of PM, the composition of
19 PM, and average relative humidity. In addition, both fine and coarse particles contribute
20 to materials damage and soiling and to adverse effects on ecosystems and vegetation
21 linked to PM deposition.
22 2. Staff recommends that consideration be given to revising the current PM25 secondary
23 standards to be consistent with any revisions made to the PM2 5 primary standards that
24 could be expected to afford greater visibility protection.
25 a. Such standards, especially if set toward the lower end of the staff-recommended
26 ranges, would be expected to result in appreciable improvements in visual air
27 quality, especially in urban areas, beyond the protection likely to be afforded by
28 the regional haze program now being implemented.
August 2003 6-50 Draft - Do Not Quote or Cite
-------
1 b. Such standards would also be expected to provide increased protection against the
2 effects of fine particles related to materials damage and soiling and potential
3 adverse impacts on ecosystems and vegetation linked to fine-particle deposition.
4 3. Staff recommends that consideration be given to replacing the current PM10 secondary
5 standards with new PM10_2 5 secondary standards that either retain the level of protection
6 afforded by the current PM10 standards or are identical to any new PM10_2 5 primary
7 standards so as to continue control of coarse particles that contribute to materials damage
8 and soiling and potential adverse impacts on ecosystems and vegetation.
9 4. In making these recommendations, staff has taken into account both the available
10 evidence linking fine and coarse particles with effects on visibility, materials, ecosystems
11 and vegetation, as well as the limitations in the available evidence. In so doing, staff
12 recognizes that the available information does not provide a sufficient basis for the
13 development of distinct national secondary standards to protect against such effects
14 beyond the protection likely to be afforded by the suite of PM primary standards.
15 6.7 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
16 RECOMMENDATIONS RELATED TO STANDARD SETTING
17 Staff believes it is important to continue to highlight the unusually large uncertainties
18 associated with establishing standards for PM relative to other single component pollutants for
19 which NAAQS have been set. Key uncertainties and staff research recommendations on both
20 health-related and welfare-related topics are outlined below. In some cases, research in these
21 areas can go beyond aiding in standard setting to aiding in the development of more efficient and
22 effective control strategies. Staff notes, however, that a full set of research recommendations to
23 meet standards implementation and strategy development needs is beyond the scope of this
24 discussion.
25 The 1996 PM Staff Paper included a discussion of uncertainties and research
26 recommendations (EPA, 1996b, pp. VII-41-44) that addressed the following issues related to
27 understanding health effects associated with exposure to PM:
28 • lack of demonstrated biological mechanisms for PM-related effects,
29 • potential influence of measurement error and exposure error,
August 2003 6-51 Draft - Do Not Quote or Cite
-------
1 • potential confounding by copollutants,
2 • evaluation of the effects of components or characteristics of particles,
3 • the shape of concentration-response relationships,
4 • methodological uncertainties in epidemiological analyses,
5 • the extent of life span shortening,
6 • characterization of annual and daily background concentrations, and
7 • understanding of the effects of coarse fraction particles.
8 As has been discussed in depth in the draft CD, especially in Chapters 5 through 8, an
9 extensive body of new studies related to understanding health effects associated with exposure to
10 PM is now available that provides important information on many of the topics listed above. For
11 example, regarding the lack of demonstrated biological mechanisms, new evidence from
12 toxicologic and controlled human exposure studies has provided information on an array of
13 potential mechanisms for effects on the cardiac and respiratory systems, as discussed in Chapters
14 7 and 9 of the draft CD. Still, the draft CD emphasizes that much remains to be learned to fully
15 understand the pathways or mechanisms by which PM is linked with different health endpoints.
16 For each of the issues listed above, new evidence has become available that helps to reduce
17 uncertainties, although uncertainty has been reduced in some areas more than others. Staff has
18 identified the following key uncertainties and research questions that have been highlighted in
19 this review of the health-based primary standards
20 1. The body of evidence on effects of coarse fraction particles has been greatly expanded,
21 but the uncertainties regarding coarse fraction particles are still greater than those for fine
22 particles. As discussed in Chapter 2, the spatial variability of coarse fraction particles is
23 generally greater than that for fine particles, which will increase uncertainty in the
24 associations between health effects and coarse-fraction particles measured at central site
25 monitors. Additional exposure research is needed to understand the influence of
26 measurement error and exposure error on coarse fraction particle epidemiology results.
27 In addition, little is known about coarse particle composition, and less about the health
28 effects associated with individual components or sources of coarse fraction particles, but
August 2003 6-52 Draft - Do Not Quote or Cite
-------
1 it is possible that there are components of coarse fraction particles (e.g., crustal material)
2 that are less likely to have adverse effects, at least at lower concentrations, than other
3 components.
4 2. Identification of specific components, properties, and sources of fine particles that are
5 linked with health effects remains an important research need. Available evidence
6 provides no basis for expecting that any one component would be solely responsible for
7 PM-related effects, but it is likely that some components are more closely linked with
8 specific effects than others. Continued source characterization, exposure,
9 epidemiological, and toxicological research is needed to help identify components,
10 characteristics, or sources of particles that may be more closely linked with various
11 specific effects to aid in our understanding of causal agents and in the development of
12 efficient and effective control strategies for reducing health risks. Conducting human
13 exposure research in parallel with such health studies will help reduce the uncertainty
14 associated with interpreting health studies and provide a stronger basis for drawing
15 conclusions regarding observed effects.
16 3. An important aspect in characterizing risk and making decisions regarding air quality
17 standard levels is the shape of concentration-response functions for PM, including
18 identification of potential threshold levels. Recent studies continue to show no strong
19 evidence for a threshold level in the relationship between PM and mortality, though some
20 studies have suggested potential levels (CD, p. 8-241).
21 4. The relationship between PM and other air pollutants in causing health effects remains an
22 important question in reducing public health risk from air pollution. Numerous new
23 analyses have indicated that associations found between PM and adverse health effects
24 are not simply reflecting actual associations with some other pollutant. However, effects
25 have been found with the gaseous copollutants, and it is possible that pollutants may
26 interact or modify effects of one another. Further understanding of the sources,
27 exposures, and effects of PM and other air pollutants can assist in the design of effective
28 strategies for public health protection.
29 5. Methodological issues in epidemiology studies were discussed at length in the previous
30 review, and it appeared at the time that the epidemiology study results were not greatly
August 2003 6-53 Draft - Do Not Quote or Cite
-------
1 affected by selection of differing statistical approaches or methods of controlling for
2 other variables, such as weather. However, investigation of recently discovered
3 questions on the use of generalized additive models in time-series epidemiology studies
4 has again raised model specification issues. While reanalyses of studies using different
5 modeling approaches generally did not generally result in substantial differences in
6 model results, some studies showed marked sensitivity of the PM effect estimate to
7 different methods of adjusting for weather variables. There remains a need for further
8 study on the selection of appropriate modeling strategies and appropriate methods to
9 control for time-varying factors, such as temperature, (see Section 3.5.3.1; CD section
10 8.4.2)
11 6. Selection of appropriate averaging times for PM air quality standards is important for
12 public health protection, and available information suggests that effects may be linked to
13 exposures of very short duration (e.g., one or more hours). Data on effects linked with
14 such peak exposures, such as those related to wildfires, agricultural burning, or other
15 episodic events, would be an important aid to public health response and communication
16 programs. Investigation into the PM exposure time periods that are linked with effects
17 will provide valuable information both for the standard-setting process and for risk
18 communication and management efforts.
19 7. There remain significant uncertainties in the characterization of annual and daily
20 background concentrations for fine particles, and especially for coarse particles. Further
21 analyses of air quality monitoring and modeling that improved these background
22 characterizations would help reduce uncertainties in estimating health risks relevant for
23 standard setting (i.e., those risks associated with exposure to PM in excess of background
24 levels) and would aid in the development and implementation of associated control
25 programs.
26 With regard to welfare-related effects, discussed in Chapter 4 of the draft CD, staff has
27 identified the following key uncertainties and research questions that have been highlighted in
28 this review of the welfare-based secondary standards:
August 2003 6-54 Draft - Do Not Quote or Cite
-------
1 1. Urban visibility remains poorly characterized. Important information that would improve
2 urban visibility characterization could be obtained from analyses of data from a number
3 of monitoring networks and programs including: ASOS airport visibility data; PM2 5
4 speciation trends network data; State and local continuous visibility monitoring (e.g.,
5 transmissometer or nephelometer data); and State and local continuous PM2 5 monitoring.
6 High-resolution data and analyses that focus on shorter than 24-hour averaging times
7 would likely be particularly informative.
8 2. Refinement and broader application of survey methods designed to elicit citizens'
9 judgments about the acceptability of varying levels of local visibility impairment could
10 help inform future decisions on national secondary standards. Such research could
11 appropriately build upon the methodology developed by the State of Colorado and used
12 as a basis for setting a visibility standard for the city of Denver, which has been adapted
13 and applied in other areas in the U.S. and abroad.
14 3. There remain significant uncertainties associated with the characterization and prediction
15 of particle deposition rates to natural surfaces in general, and most importantly, with
16 respect to particulate nitrate. Reduction in these uncertainties will be key to developing
17 the capability of quantitatively linking ambient PM concentrations with environmental
18 exposures and response. In order to better understand the nature of the role that PM
19 plays in cumulative long-term environmental impacts, more research needs to be
20 conducted on the percentage of total deposition contributed by PM and where necessary,
21 better tools and monitoring methods should be developed.
22 4. The immense variability in sensitivity to PM deposition across U.S. ecosystems has not
23 yet been adequately characterized, specifically the factors controlling ecosystem
24 sensitivity to and recovery from chronic nitrogen and acid inputs. Data should be
25 collected on a long term basis on a greater variety of ecosystems in conjunction with the
26 development of improved predictive models. Such research could help in future
27 consideration within the U.S. of the "critical loads" concept, which is generally accepted
28 in Europe as the basis for abatement strategies to reduce or prevent injury to the
29 functioning and vitality of forest ecosystems caused by long-range transboundary chronic
30 acidic deposition.
August 2003 6-55 Draft - Do Not Quote or Cite
-------
1 REFERENCES
2 Burnett, R. T.; Brook, I; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
3 (2000) Association between paniculate- and gas-phase components of urban air pollution and daily
4 mortality in eight Canadian cities. Inhalation Toxicol. 12(suppl. 4): 15-39.
5 Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated paniculate mass and daily mortality in eight Canadian
6 cities. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
7 Health Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
8 Chock, D. P.; Winkler, S.; Chen, C. (2000) A study of the association between daily mortality and ambient air
9 pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50: 1481-1500.
10 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.
11 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
12 329: 1753-1759.
13 Dockery, D. W.; Cunningham, J.; Damokosh, A. I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
14 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
15 symptoms. Environ. Health Perspect. 104: 500-505.
16 EPA. (1996a) Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC: National Center for
17 Environmental Assessment-RTF Office; report no. EPA/600/P-95/00laF-cF. 3v
18 EPA. (1996b) Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment of
19 Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office of
20 Air Quality Planning and Standards; report no. EPA-452YR-96-013.
21 EPA. (2003) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC: National Center for
22 Environmental Assessment-RTF Office; report no. EPA/600/P-99/002aD.
23 Fairley, D. (1999) Daily mortality and air pollution in Santa Clara County, California: 1989-1996. Environ. Health
24 Perspect. 107:637-641.
25 Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
26 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp.
27 97-106. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
28 Gauderman, W. J.; McConnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
29 Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and
30 lung function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
31 Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
32 Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
33 lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
34 Med. 166: 76-84.
3 5 Ito, K. (2003) Associations of paniculate matter components with daily mortality and morbidity in Detroit,
3 6 Michigan. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
37 MA: Health Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/news.htm [16 May,
38 2003].
August 2003 6-56 Draft - Do Not Quote or Cite
-------
1 Lippmann, M; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of paniculate matter components with daily
2 mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report 95.
3 Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
4 Phoenix, 1995-1997. Environ. Health Perspect. 108:347-353.
5 Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.; Koenig, J. Q. (2003) Air pollution and cardiovascular
6 mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
7 Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
8 http://www.healtheffects.org/news.htm [16 May, 2003].
9 Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
10 In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
11 Effects Institute; pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
12 Norris, G.; Young-Pong, S. N.; Koenig, J. Q.; Larson, T. V.; Sheppard, L.; Stout, J. W. (1999) An association
13 between fine particles and asthma emergency department visits for children in Seattle. Environ. Health
14 Perspect. 107: 489-493.
15 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
16 Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10:412-419.
17 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coachella Valley,
18 California. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
19 MA: Health Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/news.htm [16 May,
20 2003].
21
22 Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
23 Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999a) A study of twelve southern California communities
24 with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit.
25 Care Med. 159: 760-767.
26 Peters, J. M.; Avol, E.; Gauderman, W. J.; Linn, W. S.; Navidi, W.; London, S. J.; Margolis, H.; Rappaport, E.;
27 Vora, H.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities with
28 differing levels and types of air pollution. II. Effects on pulmonary function. Am. J. Respir. Crit. Care Med.
29 159:768-775.
30 Pope, C. A., Ill; Hill, R. W.; Villegas, G. M. (1999) Paniculate air pollution and daily mortality on Utah's Wasatch
31 Front. Environ. Health Perspect. 107:567-573.
32 Pope, C. A., Ill; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
33 (1995) Paniculate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
34 Respir. Crit. Care Med. 151:669-674.
35 Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
36 cardiopulmonary mortality, and long-term exposure to fine paniculate air pollution. J. Am. Med. Assoc.
37 287:1132-1141.
3 8 Ross, M. (2003). OAQPS Staff Memorandum to PM NAAQS Review Docket (OAR-2001-0017). Subject:
3 9 Statistical information on air quality data from epidemiologic studies. August 29, 2003.
40
41 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J.
42 Air Waste Manage. Assoc. 46:927-939.
August 2003 6-57 Draft - Do Not Quote or Cite
-------
1 Schwartz, I; Norris, G.; Larson, T.; Sheppard, L.; Clairborne, C.; Koenig, J. (1999) Episodes of high coarse
2 particles concentrations are not associated with increased mortality. Environ. Health Perspect. 107:339-
3 342.
4
5 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution on
6 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
7 Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
8 1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
9 MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/news.htm [16 May,
10 2003].
11 Stieb, D. M.; Beveridge, R. C.; Brook, J. R.; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.
12 Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in
13 Saint John, Canada. J. Exposure Anal. Environ. Epidemiol.: 10: 461-477.
August 2003 6-58 Draft - Do Not Quote or Cite
-------
Appendix A. Estimated Total Cardiovascular and Respiratory Mortality and Morbidity Effects per Increments in 24-h
Concentrations of PM10, PM25 and PM10_25 from U.S. and Canadian Studies
Original study*
Study Location
Reanalysis study
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
Schwartz etal., 1996
Steubenville, OH
Schwartz, 2003a
Analysis % increase (95% CI)
Comments per
50 ug/m3 PM10
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
GAM Strict
GLMNS
GLMBS
GLMPS
% increase (95% CI)
per
25 ug/m3 PM2 5
—
—
—
—
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)
2.4 (-0.4, 5.3)
1.7 (-1.34.8)
1.5 (-1.5, 4.6)
1.8 (-1.2, 4.9)
% increase (95% CI)
per
25 ug/m3 PM10.25
—
—
—
—
(GLMPS)
0.7 (-1.9, 3.4)
(GLMPS)
1.7 (-2.7, 6.3)
(GLMPS)
0.3 (-2.1, 2.7)
(GLMPS)
5.2(0.0, 10.7)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
PM10 38 (max 128)
PM1058(15, 177)
PM1047 (11, 297)
PM10 48 (21, 80)
PM1024.5 (SD12.8)
PM2515.7 (SD9.2)
PM10.258.8(SD7.0)
PM10 32.0 (SD 14.5)
PM25 20.8 (SD 9.6)
PM10_25 11.2 (SD 7.4)
PM10 30.6 (SD 16.2)
PM2 5 18.7 (SD 10.5)
PM10_2511.9(SD8.5)
PM10 45.6 (SD 32.3)
PM25 29.6 (SD 21.9)
PMW.2516.1 (SD 13.0)
August 2003
A-l
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
Schwartz etal, 1996
Portage, WI
Schwartz, 2003a
Schwartz etal., 1996
Topeka, KS
Schwartz, 2003a
Schwartz etal., 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
Clyde et al., 2000
Phoenix, AZ
Analysis
Comments
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
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
2.6 (-1.2, 6.6) (GLMPS)
0.8 (-3. 3, 5.1) 0.7 (-4.0, 5.6)
1.5 (-2.7, 5.8)
1.1 (-3.1, 5.4)
1.6 (-5.3, 9.0) (GLMPS)
2. 7 (-5. 0, 1 0. 9) -3. 0 (-8. 1,2.3)
1.3 (-6.2, 9.3)
1.4 (-6.3, 9.6)
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)
6(>0, 11)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
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)
PM10^5 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)
NR
PM10 mean 45.4
August 2003
A-2
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
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 et al., 2003
Klemm et al., 2000
Six City reanalysis -
Steubenville
Klemm et al., 2003
Klemm et al., 2000
Six City reanalysis - Topeka
Klemm et al., 2003
Klemm et al., 2000
Six City reanalysis - Knoxville
Klemm et al., 2003
Klemm et al., 2000
Six City reanalysis - Boston
Klemm et al., 2003
Klemm et al., 2000
Six City reanalysis - Madison
Klemm et al., 2003
Analysis
Comments
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
GAM Strict
GLMNS
% increase (95% CI)
per
50 ug/m3 PM10
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)
1.0 (-4.6, 7.0)
-1.5 (-7.7, 5.1)
% 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)
1.5 (-2.7, 5.9)
-1.2 (-5.7, 3.5)
% increase (95% CI)
per
25 ug/m3 PM10.25
4.5 (-7.6, 18.1)
3.3 (-5.3, 12.6)
—
___
1.4 (-11.3, 15.9)
0.0 (-2.2, 2.3)
-0.5 (-3.0, 2.0)
4.6 (-0.7, 10.1)
4.0 (-1.6, 10.0)
-3.7 (-9.2, 2.1)
-4.7 (-10.8, 1.8)
3.5 (-1.0, 8.2)
3.0 (-1.9, 8.2)
1.3 (-1.1, 3.7)
1.8 (-1.0, 4.6)
0.0 (-4.8, 5.0)
-1.0 (-6.2, 4.5)
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
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)
PM25 18.7 (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)
PM25 15.7 (SD 9.2)
PM10.25 8.8 (SD 7.0)
PM10 17.8(SD11.7)
PM25 11.2(SD7.8)
PM10.256.6(SD6.8)
August 2003
A-3
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
Klemm et al., 2000
Six City reanalysis - overall
Klemm et al., 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
Analysis
Comments
GAM Strict
GLMNS
GLMPS
GAM not used
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
% increase (95% CI)
per
50 ug/m3 PM10
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
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)
—
% increase (95% CI)
per
25 ug/m3 PM10.25
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 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)
PM25 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)
Moolgavkar, 2003
Ostro, 1995
San Bernadino and Riverside
Counties, CA
Schwartz, 2000c
Boston, MA
Schwartz, 2003a
GAM not used
GLMNS
0 (-1.4, 1.4)
5.8 (4.5, 73) (15-day)
9.7(8.2, 11.2) (60-day)
PM2532.5(9.3, 190.1)
(estimated from
visibility)
PM25 15.6 (±9.2)
August 2003
A-4
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
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
Schwartz, 2000
Spokane, WA
Schwartz, 2003b
Analysis % increase (95% CI) % increase (95% CI)
Comments per per
50 ug/m3 PM10 25 ug/m3 PM25
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)
StrictGAM 5.62 (-0.31, 11.91)
(dist. lag)
% increase (95% CI) PM10, PM2 5 and PM10.2 5
per Mean (Range) Levels
25 ug/m3 PM10.25 Reported"
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
PM10
August 2003
A-5
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
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)
GAM not used
GAM not used
GAM not used
% increase (95% CI) % increase (95% CI) % increase (95% CI) PM10, PM2 5 and PM10.2.5
per per per Mean (Range) Levels
50 ug/m3 PM10 25 ug/m3 PM25 25 ug/m3 PM10.25 Reported"
8.58 (-3.94, 22.73)
5.65 (4.62, 6.70) 4.34 (2.82, 5.89)
11.07 (0.70, 22.51) 5.65 (0.11, 11.51)
-4.88 (-17.88, 10.19) 1.77 (-5.44, 9.53)
PM10
PM15 55 (SD 6.5)
PM2542.1(SD22.0)
PM15 47.0 (SD 20.9)
PM25 39.9 (SD 18.0)
PM15 47.5 (SD 18.8)
PM2537.1(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
Total Cardiovascular Mortality
/to and Thurston, 1996
Chicago, IL
Pope et al., 1992
Utah Valley, UT
GLMNS
GAM not used
GAM not used
GAM not used
GAM not used
GAM not used
1.6 (0.8, 2.4)
7.79 (3.65, 12.10) 5.13 (3.09, 7.21)
15.03 (4.29, 26.87) 6.18 (0.61, 12.06)
3.05 (-11.04, 19.36) 2.28 (-4.97, 10.07)
1.49 (-0.72, 3.74)
9.36(1.91,17.36)
PM10 mean range
15.3-52.0
PM15 55 (SD 6.5)
PM2542.1(SD22.0)
PM15 47.0 (SD 20.9)
PM25 39.9 (SD 18.0)
PM15 47.5 (SD 18.8)
PM,,37.1(SD 19.8)
PM10 38 (max 128)
PM10 47 (11,297)
August 2003
A-6
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
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
Ostro et al., 2000
Coachella Valley, CA
Ostro et al., 2003
Ostro, 1995
San Bernadino and Riverside
Counties, CA
Analysis
Comments
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not used
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM Strict
GLMNS
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
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)
5.5(1.6,9.5)
5.1(1.2,9.1)
___
% 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)
...
9.8 (-5.7, 27.9)
10.2 (-5.3, 28.3)
0.69 (-0.34, 1.74)
% increase (95% CI)
per
25 ug/m3 PM10.25
5.0 (-13.3, 27.3)
—
5.4 (-0.4, 11.2)
6.7 (-1.0, 15.0)
6.0 (-1.6, 14.3)
6.4(1.3, 11.7)
6.2 (0.8, 12.0)
...
2.9 (0.7, 5.2)
2.7(0.4,5.1)
___
PM10, PM2 5 and PM10.2.5
Mean (Range) Levels
Reported"
PM10 34 (6, 165)
PM2513(2, 105)
PM10.2511(0,45)
PM25 17.6 (4.6, 71.7)
PM10 32.20 (7.0, 95.0)
PM25 17.28 (-0.6, 72.6)
PM10.25 6.80 (-20.0, 28.3)
PM1031(12, 105)
PM25 18(6,86)
PM10.2513(4,50)
mean (10%, 90%)
PM1046.5(5, 213)
PM2513.0(0, 42)
PM10.2.5 33.5 (5, 187)
PM10 median 44 (7, 166)
PM2 5 median 22 (4, 86)
PM10 median 35 (3, 365)
PM10 47.4 (3, 417)
PM25 16.8(5,48)
PM10.25 17.9 (0, 149)
PM2532.5(9.3, 190.1)
(estimated from
visibility)
August 2003
A-7
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
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 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
6.77(1.97,
19.78(3.51,
10.7 (-3.7,
10.8 (-3.4,
7.5 (-10.5,
7.9 (-10.2,
11.79)
38.61)
27.2) 11.7 (-9.8, 38.3)
27.1) 13.5 (-3.6, 33.7)
29.2) 2.3 (-10.4, 16.7)
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,
6.2 (-3.4,
12.6) 1.0 (-5. 1,7.4)
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"
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)
PM10.25 13 (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)
August 2003
A-8
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
Analysis
Comments
% increase (95% CI)
per
50 ug/m3 PM10
% 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"
CARDIOVASCULAR MORBIDITY
Total Cardiovascular Hospital
Samet et al., 2000
14 U.S. Cities (>65 years)
Zanobetti and Schwartz (2003b)
Linn et al., 2000
Los Angeles, CA (>29 years)
Metzger et al., 2003
Atlanta, GA
Moolgavkar, 2000b
Cook Co., IL (all ages)
Moolgavkar, 2003
Moolgavkar, 2000b
Los Angeles, CA (all ages)
Moolgavkar, 2003
Tolbert et al., (2000a)
Atlanta, GA 1993-1998
Admissions:
strict GAM
strict GAM
(dist lag)
GLMNS
GLMPS
GAM not used
GAM not used
strict GAMloodf
GLMNSloodf
GAM30df
GAMloodf
GLMNSloodf
GAM not used
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)
4.58 (-0.7, 10.2)
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)
-8.2(p=0.002)
PM10 means 24.4-45.3
PM10 45.5 (5, 132)
8.5 (2.6, 14.7) 3.0 (-3.7, 10.3) PM10 27.9 (0.5, 90.1)
PM25 19.2(1.8,54.6)
PM10.25 9.7 (0.5, 34.2)
PM10 median 35 (3, 365)
3.95 (2.2, 5.7) — PM10 median 44 (7, 166)
2.9(1.2,4.6) PM2 5 median 22 (4, 86)
3.15(1.1,5.2)
PM1030.1(SD 12.4)
(Period 1)
Tolbert et al., 2000a
Atlanta, GA (all ages)
Stieb et al., 2000
St. John, CAN (all ages)
GAM not used
GAM not used
5.1 (-7.9, 19.9)
39.2 (5.0, 84.4)
6.1 (-3.1, 16.2)
15.11(0.25,32.8)
17.6 (-4.6, 45.0)
PM1029.1(SD 12.0)
PM25 19.4 (SD 9.35)
PM10.25 9.39 (SD 4.52)
summer 93
PM10 14.0 (max 70.3)
PM25 8.5 (max 53.2)
August 2003
A-9
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
Burnett et al., 1997
Toronto, CAN (all ages)
Ischemic Heart Disease Hospital
Schwartz and Morris, 1995
Detroit (> 65 years)
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito 2003
Metzger et al., 2003
Atlanta, GA
Analysis
Comments
GAM not used
Admissions:
GAM not used
Strict GAM
GLMNS
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
12.07(1.43,23.81)
5.0(1.9,8.3)
8.0 (-0.3, 17.1)
6.2 (-2.0, 15.0)
5.6 (-3.9, 16.0)
% increase (95% CI)
per
25 ug/m3 PM2 5
7.18 (-0.61, 15.6)
—
3.65 (-2.05, 9.7)
3.0 (-2.7, 9.0)
5.8 (-4.1, 16.9)
% increase (95% CI)
per
25 ug/m3 PM10.25
20.46 (8.24, 34.06)
—
10.2 (2.4, 18.6)
8.1(0.4, 16.4)
-4.5 (-13.0, 11.6)
PM10, PM2 5 and PM10.2.5
Mean (Range) Levels
Reported"
PM10 28.4 (4, 102)
PM2 5 16.8 (1,66)
PM10.251 1.6 (1,56)
PM10 48 (22, 82)
mean (10%, 90%)
PM1031(maxl05)
PM25 18(6,86)
PM10.25 13 (4, 50)
PM10 27.9 (0.5, 90.1)
PM25 19.2(1.8,54.6)
PM10.259.7(0.5, 34.2)
Dysrhythmias Hospital Admissions:
Lippmann et al., 2000
Detroit, MI (>65 years)
Strict GAM
GLMNS
2.8 (-10.9-18.7)
2.0 (-11.7-17.7)
3.2 (-6.6-14.0)
2.6 (-7.1-13.3)
0.1 (-12.4-14.4)
0.0 (-12.5-14.3)
PM1031(maxl05)
PM25 18 (6, 86)
Ito (2003)
Metzger et al., 2003
Atlanta, GA
Tolbert et al., 2000a
Atlanta, GA (all ages)
GAM not used
GAM not used
4.1 (-5.9, 14.9)
13.4 (-14.1, 49.0)
3.8 (-5.8, 14.4)
6.1 (-12.6, 28.9)
5.3 (-6.3, 18.5)
53.2(2.1, 129.8)
PM10.25 13 (4, 50)
PM10 27.9 (0.5, 90.1)
PM25 19.2 (1.8, 54.6)
PM10.25 9.7 (0.5, 34.2)
PM2519.4(SD9.35)
PM10.25 9.39 (SD 4.52)
August 2003
A-10
Draft - Do Not Quote or Cite
-------
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"
Heart Failure/Congestive Heart Disease Hospital Admissions:
Schwartz and Morris, 1995 GAM not used 2.8 (0.7, 5.0)
Detroit (> 65 years)
Linn et al., 2000 GAM not used 2.02 (-0.94, 5.06)
Los Angeles, CA (>29 years)
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito, 2003
Metzger et al., 2003
Atlanta, GA
Morris and Naumova, 1998
Chicago, IL (>65 years)
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 GAM not used
Los Angeles, CA (>29 years)
9.2 (-0.3-19.6)
8.4 (-1.0-18.7)
GAM not used -3.9 (-14.9, 8.4)
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)
14.3(1.7,28.6)
4.4 (-4.0-13.5)
4.9 (-3.55-14.1)
5.1 (-8.7, 21.0)
PM1048(22,82)
mean (10%, 90%)
PM10 45.5 (5, 132)
PM1031(maxl05)
PM25 18(6,86)
PM10.25 13 (4, 50)
PM10 27.9 (0.5, 90.1)
PM25 19.2(1.8,54.6)
PM10.25 9.7 (0.5, 34.2)
PM1041(6, 117)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
PM10 45.5 (5, 132)
August 2003
A-ll
Draft - Do Not Quote or Cite
-------
Original study* Analysis % increase (95% CI)
Study Location Comments per
Reanalysis study 50 ug/m3 PM10
Lippmann et al., 2000 Strict GAM 5.00 (-5.27, 16.38)
Detroit, MI (>65 years) GLM NS 4.41 (-5.81, 15.74)
Ito, 2003
Peripheral Circulation Hospital Admissions:
Metzger et al., 2003 GAM not used 10.4 (-0.9, 23.0)
Atlanta, GA
% 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"
1.94 (-5.16, 9.57) 5.00 (-4.59, 15.56) PM10 31 (max 105)
0.97 (-6.06, 8.52) 5.63 (-4.02, 16.25) PM25 18 (6, 86)
PM10.25 13 (4, 50)
13.0(2.1,25.0) 5.6 (-7.0, 8.6) PM10 27.9 (0.5, 90.1)
PM25 19.2 (1.8, 54.6)
PMin.,, 9.7 (0.5, 34.2)
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 GAM not used 2.89 (1.09, 4.72)
Los Angeles, CA (>29 years)
Schwartz etal., 1996 GAMnotused 5.8(0.5,11.4)
Cleveland, OH (>65 years)
Lumley and Heagerty, 1999 GAMnotused
King County, WA (all ages)
Burnett et al., 1997 GAMnotused 10.93 (4.53, 17.72)
15.00 (1.97, 28.03) 22.25 (-9.53, 54.03) PM10 29.5-38.8 (max
96.0)
PM25 15.8-22.3 (max
66.0)
PMW_25 12.7-16.5 (max
33.0)
PM10 45.5 (5, 132)
PM10 43
5.91(1.10,10.97) — PMjNR
8.61 (3.39, 14.08) 12.71 (5.33, 20.74) PM10 28.1 (4, 102)
Toronto, CAN (all ages)
Delfinoetal., 1997
Montreal, CAN (>64 years)
PM2 5 16.8 (1,66)
PM10.25 11.6 (1,56)
GAM not used 36.62 (10.02, 63.21)
23.88 (4.94, 42.83)
summer 93
PM1021.7(max51)
PM25 12.2 (max 31)
August 2003
A-12
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
Delfinoetal., 1998
Montreal, CAN (>64 years)
Stieb et al., 2000
St. John, CAN (all ages)
Pneumonia Hospital Admissions:
Schwartz, 1995
Detroit (> 65 years)
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
Analysis
Comments
GAM not used
GAM not used
GAM not used
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
% increase (95% CI)
per
50 ug/m3 PM10
—
8.8(1.8, 16.4)
5.9(1.9, 10.0)
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)
% 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"
13. 17 (-0.22, 26.57) — PM25 18.6 (SD 9.3)
5.69(0.61,11.03) — summer 93
PM10 14.0 (max 70.3)
PM2 5 8.5 (max 53. 2)
PM1048(22,82)
mean (10%, 90%)
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)
Los Angeles, CA (>29 years)
Tolbert et al., 2000a
Atlanta, GA (all ages)
GAM not used
-3.5 (33.0, -29.9)
12.44 (-7.89, 37.24) -23.03 (-50.69, 20.15)
PM10 29.1(SD12.0)
PM25 19.4 (SD 9.35)
PM10.259.39(SD4.52)
August 2003
A-13
Draft - Do Not Quote or Cite
-------
Original study*
Study Location
Reanalysis study
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito (2003)
Moolgavkar, 2000c
Cook Co., IL (allages)
Moolgavkar 2003
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., 2000b
Atlanta, GA (<17 years)
Tolbert et al., 2000a
Atlanta, GA (all ages)
Analysis
Comments
Strict GAM
GLMNS
Strict GAM:
100 df
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
GAM not used
% increase (95% CI)
per
50 ug/m3 PM10
6.5 (-7.8, 23.0)
4.6 (-9.4, 20.8)
3.24 (.03, 6.24)
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)
18.8 (-8.7, 54.4)
% 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"
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)
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)
2.3 (-14.8, 22.7) 21.1 (-18.2, 79.3) PM10 29.1 (SD 12.0)
PM25 19.4 (SD 9.35)
PM10.25 9.39 (SD 4.52)
August 2003
A-14
Draft - Do Not Quote or Cite
-------
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
Sheppardetal., 1999
Seattle, WA (<65 years)
Sheppardetal., 2003
Strict GAM
GLMNS
10.9 (2.8, 19.6)
8.1(0.1, 16.7)
8.7 (3.2, 14.4)
6.5(1.1,12.0)
5.5 (0, 14.0)
5.5 (-2.7, 11.1)
PM1031.5(9055)
PM25 16.7(9032)
PMin.7, 16.2 (90 29)
PMin.7, 16.2 (90 29)
Odds Ratio (95% CI) for Odds Ratio (95% CI) for PM10.2.5 Mean (Range)
25 ug/m3 % increase in 25 ug/m3 % increase in Levels Reported**
PIV/T PIV/T
Respiratory Symptoms
Odds Ratio (95% CI) for Odds Ratio (95% CI) for Odds Ratio (95%
50 ug/m3 % increase in 25 ug/m3 % increase in 25 ug/m3 % incn
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
GAM not used
1.09(0.57, 2.10)
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 5 median 18.0 (max
86)
PM10 median 30.0 (max
117)
PM2 s median 18.0 (max
86)
PM25 24.5 (max 88.1)
PM1022(0.5,73)
PM1044(11, 195)
PM1044 (11, 195)
August 2003
A-15
Draft - Do Not Quote or Cite
-------
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) PM10, PM2 5 and PM10.2.5
per Mean (Range) Levels
25 ug/m3 PM10_2.5 Reported"
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)
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)
GAM not used
GAM not used
GAM not used
NR
NR
NR
GAM not used 1.05 (0.64, 1.73)
GAMnotused 1.51(1.04,2.17)
GAM not used
GAM not used
GAMnotused 1.40(1.14,1.73)
1.48(1.17, 1.88) (1-d)
1.59 (0.93, 2.70) (1-d)
1.61 (1.21, 2.17) (0-d)
1.28 (0.98, 1.67)
1.61(1.20,2.16)
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)
PM10 55.87 (19.63,
101.42)
1.77 (1.23, 2.54) PM25 (same as Six Cities)
PM10.25NR
1.51 (0.66, 3.43) PM25 (same as Six Cities)
PM10.25NR
PM10 median 22.1(0.2,
159.0) (north site)
August 2003
A-16
Draft - Do Not Quote or Cite
-------
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)
per
25 ug/m3 PM10.25
PM10, PM2 5 and PM10.2 5
Mean (Range) Levels
Reported"
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)
Vedaletal., 1998
Port Alberni, CAN
(children, dyspnea)
Vedaletal., 1998
Port Alberni, CAN
(children, any symptom)
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)
GAM not used 1.05 (0.74, 1.49)
GAMnotused 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)
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)
August 2003
A-17
Draft - Do Not Quote or Cite
-------
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
August 2003
A-18
Draft - Do Not Quote or Cite
-------
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.
August 2003
A-19
Draft - Do Not Quote or Cite
-------
-------
TECHNICAL REPORT DATA
(Please read Instructions on reverse before completing)
1. REPORT NO.
EPA-452/D-03-001
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Review of the National Ambient Air Quality Standards for
Particulate Matter: Policy Assessment of Scientific and Technical
Information. OAQPS Staff Paper, First Draft
5. REPORT DATE
August 2003
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Ross, M.., Mathias, S., Richmond, H., Langstaff, J., Sandiford, V.,
Damberg, R., Martin, K.
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Health and Ecosystems Effect Group, Mail Code C539-01
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Health and Ecosystems Effect Group, Mail Code C539-01
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
14. SPONSORING AGENCY CODE
EPA/200/04
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The purpose of the Staff Paper is to evaluate the policy implications of the key scientific and technical
information contained in a related EPA document, Air Quality Criteria for Particulate Matter (AQCD), and
identify critical elements that EPA staff believe should be considered in reviewing the PM National ambient
air quality standards (NAAQS). The Staff Paper is intended to "bridge the gap" between the scientific
review contained in the AQCD and the public health and welfare policy judgments required of the
Administrator in reviewing the NAAQS. A fourth external review draft of the AQCD has recently been made
available for public review and comment (66 FR 18929, June 30, 2003), and for review by CASAC at a
public meeting on August 25-26, 2003. Building upon an earlier preliminary draft Staff Paper (EPA, 2001),
and taking into account the information in the fourth external review draft of the AQCD, this first draft Staff
Paper includes results from initial staff analyses (e.g., analyses of air quality and visibility data, and human
health risk assessment), plans for additional analyses to be incorporated into a subsequent draft of this
document, and preliminary staff conclusions or recommendations on potential revision or retention of the PM
NAAQS.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Particulate Matter
PM
Air Pollution
Health Effects
Welfare Effects
Mortality
Morbidity
Risk Assessment
Air Quality Standards
18. DISTRIBUTION STATEMENT
Release Unlimited
19. SECURITY CLASS (Report)
Unclassified
21. NO. OF PAGES
20. SECURITY CLASS (Page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
-------
-------
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-452/D-03-001
Environmental Protection Air Quality Strategies and Standards Division August 2003
Agency Research Triangle Park, NC
------- |