t
Review of the National Ambient Air Quality
Standards for Particulate Matter:
Policy Assessment of Scientific
and Technical Information
OAQPS Staff Paper - Second Draft
January 2005
-------
s
-------
EPA-452/D-05-001
January 2005
Second Draft
Review of the National Ambient Air Quality
Standards for Particulate Matter:
Policy Assessment of Scientific
and Technical Information
OAQPS Staff Paper - Second Draft
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
January 2005 Draft - Do Not Quote or 'Cite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 provisional 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.
January 2005 Draft - Do Not Quote or Cite
-------
Table of Contents
List of Tables .............................. ............................. : . . . viii
-\
List of Figures . . .' [[[ xi
Abbreviations and Acronyms .................................................. xvi
1 INTRODUCTION,. \ ............................... • ...... .. ........... 1-1
1.1. PURPOSE ""•:;;.,.. ............. .............. . . .' ................... 1-1
. 1.2 BACKGROL%D\ .................. .' ............ ' ................ 1-2
1.2.1 Legislative Requirements .......... . ........ ................ 1-2
1.2.2 History of tJM NAAQS Reviews .............................. 1-4
1.2.3 -Litigation Related to 1997 PM Standards ....................... 1-5
1.2.4 Current PM NAAQS Review ............ .v .................... 1-7
1.3 GENERAL APPROACH AND ORGANIZATION OF DOCUMENT ...... 1-8
REFERENCES [[[ 1-11
2 CHARACTERIZATION OF AMBIENT PM ............................... 2-1
2.1 INTRODUCTION ....................... .' ....................... 2-1
2.2 PROPERTIES OF AMBIENT PM .............. .............. .' ..... 2-1
2.2.1 Particle Size Distributions ................ . .................. 2-2
2.2.1.1 Modes ........................ : . .' ................. .2-2
2.2.1.2 Sampler Cut Points ........ '. ......... '. ............... 2-4
2.2.2 Sources and Formation Processes ............................. 2-8
2.2.3 Chemical Composition ........ . ................ : ............ 2-9
2.2.4 Fate and Transport ........................................ 2-10
2.2.5 Optical Properties of Particles ............................... 2-11
2.2.6 Radiative Properties of Particles ...... . .................. ____ 2-12
2.3 AMBIENT PM MEASUREMENT METHODS ....................... 2-14
2.3.1 Particle Mass Measurement Methods ......................... 2-14
2.3.2 Particle Indirect Optical Methods ............................ 2-16
2.3.3 Size-Differentiated Particle Number Concentration Measurement
Methods ................... . ........................... . 2-17
2.3.4 Chemical Composition Measurement Methods . ................. 2-17
2.3.5 Measurement Issues ........ . .............................. 2-18
2.4 PM CONCENTRATIONS, TRENDS, AND SPATIAL PATTERNS ...... 2-20
2.4.1 PM25 ................. ; ........... .' ................. .... 2-20
2.4.2 PM10 [[[ 2-32
2.4.3 PM}0.25 ..................... . ............................ 2-32
2.4.4 Ultr'afme Particles .................. . ..................... 2-39
2.4.5 Components of PM ..... ..... ' ............................. 2-41
-------
2.5 PM TEMPORAL PATTERNS : 2-46
2.5.1 PM25 and PMi0.25 Patterns 2-46
2.5.2 Ultrafine Patterns 2-60
2.6 PM BACKGROUND LEVELS 2-60
2.7 RELATIONSHIP BETWEEN AMBIENT PM MEASUREMENTS AND
HUMAN EXPOSURE 2-65
2.7.1 Definitions 2-66
2.7.2 Centrally Monitored PM Concentration as a Surrogate for Particle
Exposure 2-67
2.8 RELATIONSHIP BETWEEN AMBIENT PM AND VISIBILITY 2-72
2.8.1 Particle Mass and Light Extinction 2-72
2.8.2 Other Measures of Visibility 2-75
2.8.3 Visibility at PM Background Conditions 2-76
REFERENCES 2-79
PM-RELATED HEALTH EFFECTS
AND PRIMARY PM NAAQS
3. POLICY-RELEVANT ASSESSMENT OF HEALTH EFFECTS EVIDENCE 3-1
3.1 INTRODUCTION '. .. 3-1
3.2 MECHANISMS ' 3-5
3.3 NATURE OF EFFECTS ' 3-12
3.3.1 Premature Mortality 3-13
3.3.1.1 Mortality and Short-term PM Exposure 3-13
3.3.1.2 Mortality and Long-term PM Exposure 3-20
3.3.2 Morbidity 3-23
3.3.2.1 Hospitalization and Medical Visits 3-23
3.3.2.2 Effects on the Respiratory System from Short-term Exposures
3-26
3.3.2.3 Effects on the Respiratory System from Long-term Exposures
3-28
3.3.2.4 Effects on the Cardiovascular System 3-29
3.3.3 Developmental effects 3-30
3.3.4 Summary 3-31
3.4 INTEGRATIVE ASSESSMENT OF HEALTH EVIDENCE 3-31
3.4.1 Strength of Associations 3-32
3.4.2 Robustness of Associations 3-33
3.4.3 Consistency 3-34
3.4.4 Coherence and Plausibility 3-36
3.4.5 Summary 3-38
3.5 PM-RELATED IMPACTS ON PUBLIC HEALTH 3-39
3.5.1 Potentially Susceptible and Vulnerable Subpopulations 3-39
3.5.2 Potential Public Health Impact 3-41
January 2005
11
Draft - Do Not Quote or Cite
-------
I
3.6 ISSUES RELATED TO QUANTITATIVE ASSESSMENT OF
EPIDEMIOLOGIC EVIDENCE 3-42
.3:6,1 Air Quality Data in Epidemiologic Studies 3-43
3.6.2 Exposure Error " 3-46
3.6.3 Alternative Model Specifications ' 3-48
3.6.4 Co-pollutant Confounding and Effect Modification 3-50
3.6.5 Temporality in Concentration-Response Relationships 3-56
" . 3.6.5.1 PM short-term exposure time periods 3-56
3.6.5.2 Lag Structure in Short-term Exposure Studies 3-57
3.6.5.3 Seasonal Differences in Time-Series Epidemiologic Results . 3-60
'" 3.6.5.4 Exposure Time Periods in Long-term Exposure Studies 3-61
3.6.6 Concentration-Response Relationships and Potential Population
Thresholds 3-63
3.7 SUMMARY AND CONCLUSIONS • 3-66
REFERENCES ' 3-69
4 CHARACTERIZATION OF HEALTH RISKS 4-1
4.1 INTRODUCTION' T '... 4-1
4.1.1 Summary of Risk Assessment Conducted During Prior PM N AAQS
Review : 4-3
4.1.2 Development of Updated Assessment 4-5
4.2 GENERAL SCOPE OF PM RISK ASSESSMENT 4-6
4.2.1 Overview of Components of the Risk Model 4-10
4.2.2 Criteria for Selection of Health Endpoints and Urban Study Areas .. 4-11
4.2.2.1 Selection of Health Endpoint Categories 4-13
-4.2.2.2 Selection of Study Areas 4-14
4.2.3 Air Quality Considerations 4-17
4.2.3.1 Estimating PM Background Levels 4-20
4.2.3.2 Simulating PM Levels That Just Meet Specified Standards .. 4-20
4.2.4 . Approach to Estimating PM-Related Health Effects Incidence 4-24
4.2.5 Baseline Health Effects Incidence Rates and Population Estimates .. 4-27
4.2.6 Concentration-Response Functions Used in Risk Assessment 4-28
4.2.6.1 Hypothetical Thresholds • 4-34
4.2.6.2 Single and Multi-Pollutant Models !... 4-34
4.2.6.3 Single, Multiple, and Distributed Lag Functions 4-35
4.2.6.4 Long-term Exposure Mortality PM2 5 Concentration-Response
. Functions 4-37
4.2.7 Characterizing Uncertainly and Variability 4-37
4.3 PM25 and PM10.2.5 RISK ESTIMATES FOR CURRENT ("AS IS") AIR
QUALITY ' 4-41
4.3.1 Base Case Risk Estimates : 4-41
4.3.2 Sensitivity Analyses ; i 4-50
4.3.2.1 Alternative Background Levels 4-50
4.3.2.2 Hypothetical Thresholds 4-53
January 2005 iii Draft - Do Not Quote or Cite
-------
4.3.2.3 Alternative Concentration-Response Models 4-54
4.3.3 Key Observations 4-56
4.4 RISK ESTIMATES ASSOCIATED WITH JUST MEETING THE CURRENT
PM2 5 STANDARDS 4-59
4.4.1 Base Case Risk Estimates 4-59
4.4.2 Sensitivity Analyses 4-65
4.4.3 Key Observations 4-66
4.5 RISK ESTIMATES ASSOCIATED WITH JUST MEETING ALTERNATIVE
PM2.5 AND PM10.15 STANDARDS 4-67
4.5.1 Base Case Risk Estimates for Alternative PM25 Standards 4-67
4.5.2 Base Case Estimates for Alternative PM10.25 Standards 4-74
4.5.3 Sensitivity Analyses for Alternative PM25 and PM10.2.5 Standards ... 4-75
4.5.3.1 Hypothetical Thresholds 4-75
4.5.3.2 Spatial Averaging Versus Maximum Community Monitor... 4-79
4.5.4 Key Observations 4-81
REFERENCES 4-84
5 STAFF CONCLUSIONS AND RECOMMENDATIONS ON PRIMARY
PM NAAQS 5-1
5.1 INTRODUCTION 5-1
5.2 APPROACH 5-2
5.3 FINE PARTICLE STANDARDS 5-5
5.3.1 Adequacy of Current PM25 Standards 5-5
5.3.2 Indicators 5-17
5.3.3 Averaging Times 5-21
5.3.4 Alternative PM25 Standards to Address Health Effects Related to
Long-term Exposure 5-24_
5.3.4.1 Evidence-based Considerations 5-25
5.3.4.2 Risk-based Considerations 5-27
5.3.4.3 Summary 5-32
5.3.5 Alternative PM25 Standards to Address Health Effects Related to
Short-term Exposure 5-33
5.3.5.1 Evidence-based Considerations 5-33
5.3.5.2 Risk-based Considerations 5-42
5.3.5.3 Summary 5-47
5.3.6 Alternative Forms for Annual and 24-Hour PM25 Standards 5-48
5.3.6.1 Form of Annual Standard 5-48
5.3.6.2 Form of 24-Hour Standard 5-51
5.3.7 Summary of Staff Recommendations on Primary PM23 NAAQS .... 5-54
5.4 THORACIC COARSE PARTICLE STANDARDS ! 5-57
5.4.1 Adequacy of Current PM10 Standards 5-57
5.4.2 Indicators ' 5-62
5.4.3 Averaging Times 5-64
January 2005
IV
Draft - Do Not Quote or Cite
-------
5.4.4 Alternative PM10.25 Standards to Address Health Effects Related to
' Short-term Exposure 5-65
5.4.5 Summary of Staff Recommendations on Primary PM10.25 NAAQS .. 5-73
5.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
RECOMMENDATIONS RELATED TO SETTING PRIMARY PM
STANDARDS 5-75
REFERENCES , 5-79
PM-RELATED WELFARE EFFECTS
AND SECONDARY PM NAAQS
6 POLICY-RELEVANT ASSESSMENT OF PM-RELATED WELFARE
EFFECTS '. 6-1
6.1 INTRODUCTION , 6-1
6.2 EFFECTS ON VISIBILITY 6-2
6.2.1 Overview of Visibility Impairment '. 6-2
6.2.2 Visibility Trends and Current Conditions in Class I and Non-Urban
Areas : 6-4
6.2.3 Visibility Conditions in Urban Areas 6-5
6.2.3.1 ASOS Airport Visibility Monitoring Network 6-6
6.2.3.2 Correlation between Urban Visibility and PM25 Mass 6-7
6.2.4 Economic and Societal Value of Improving Visual Air Quality 6-15
6.2.5 Programs and Goals for Improving Visual Air Quality 6-18
6.2.5.1 Regional Protection 6-18
6.2.5.2 Local, State, and International Goals and Programs 6-19
6.2.6 Approaches to Evaluating Public Perceptions and Attitudes 6-21
6.2.6.1 Photographic Representations of Visual Air Quality 6-22
6.2.6.2 Survey Methods 6-23
6.2.7 Summary 6-26
6.3 EFFECTS ON VEGETATION AND ECOSYSTEMS . 6-27
6.3.1 Major Ecosystem Stressors in PM 6-29
6.3.2 Direct Vegetation Effects of PM Stressor Deposition 6-30
6.3.3 Ecosystem Effects of PM Stressor Deposition 6-31
6.3.3.1 Environmental Effects of Reactive Nitrogen Deposition .... 6-32
6.3.3.2 Environmental Effects of PM-Related Acidic Deposition ... 6-42
6.3.4 Characteristics and Location of Sensitive Ecosystems in the U.S. ... 6-51
6.3.5 Ecosystem Exposures to PM Stressor Deposition 6-53
6.3.6 'Critical Loads 6-55
6.3.7 Summary and Conclusions 6-59
6.4 EFFECTS ON MATERIALS 6-61
6.4.1 Materials Damage Effects 6-62
6.4.2 Soiling Effects 6-64
6.4.3 Summary and Conclusions 6-65
January 2005 v Draft - Do Not Quote or Cite
-------
6.5 EFFECTS ON CLIMATE CHANGE AND SOLAR RADIATION 6-65
6.5.1 Climate Change and Potential Human Health and Environmental
Impacts 6-66
6.5.2 Alterations in Solar UV-B Radiation and Potential Human Health
and Environmental Impacts 6-68
6.5.3 Summary and Conclusions 6-71
REFERENCES 6-73
7 STAFF CONCLUSIONS AND RECOMMENDATIONS ON SECONDARY
PM NAAQS 7-1
7.1 INTRODUCTION 7-1
7.2 APPROACH 7-2
7.3 STANDARDS TO ADDRESS VISIBILITY IMPAIRMENT 7-3
7.3.1 Adequacy of Current PM25 Standards 7-5
7.3.2 Indicators 7-7
7.3.3 Averaging Times : 7-8
7.3.4 Alternative PM25 Standards to Address Visibility Impairment 7-9
7.3.5 Alternative Forms of a Short-term PM25 Standard 7-13
7.3.5 Summary of Staff Recommendations 7-16
7.4 STANDARDS TO ADDRESS OTHER PM-RELATED WELFARE
EFFECTS 7-17
7.4.1 Vegetation and Ecosystems 7-18
7.4.2 Materials Damage and Soiling 7-21
7.4.3 Climate Change and Solar Radiation 7-21
7.4.4 Summary of Staff Recommendations 7-22
7.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
RECOMMENDATIONS RELATED TO STANDARD SETTING 7-23
REFERENCES 7-25
APPENDIX 2A:
Source Emissions 2A-1
APPENDIX 3A:
Mortality and Morbidity Effect Estimates and PM Concentrations from U.S. and
Canadian Studies for Short-term Exposures to PM10, PM25, and PM10.25 3A-1
APPENDIX 3B:
Mortality and Morbidity Effect Estimates and PM Concentrations from U.S. and
Canadian Studies for Long-Term Exposures to PM,0, PM25,, and PM10.2S 3B-1
January 2005
VI
Draft - Do Not Quote or Cite
-------
APPENDIX 4A:
Study-Specific Information on Short- and Long-term Exposure Studies in Cities
included in PM2 5 Assessment and on Short-term Exposure Studies in Cities
included in PM10.25 Assessment 4A-i
APPENDIX 4B:
Sensitivity Analyses: Estimated PM-Related Incidence Associated with Short-
and Long-term Exposure to PM2 5 and Short-term Exposure to PMia.25 • • •: 4B-i
ATTACHMENT 6A:
Images of Visual Air Quality in Selected Urban Areas in the U.S. available at:
.-..'..'. hrtp://wwv.epa.gov/ttnAiaaQs/standards/prn/s pm cr sp.html
January 2005 " vii Draft - Do Not Quote or Cite
-------
List of Tables
Number
2-1 Particle Size Fraction Terminology Used in Staff Paper 2-3
2-2 Comparison of Ambient Particles, Fine Particles (Ultrafine plus
Accumulation-Mode) and Coarse Particles . . 2-12
2-3 Summary of PM25 FRM Data Analysis in 27 Metropolitan Areas, 1999-2001' 2-31
2-4 Summary of Estimated PM10_25 Analysis in 17 Metropolitan Areas, 1999-2001 2-40
2-5 Estimated Ranges of Annual Average PM Regional Background Levels 2-61
2-6 IMPROVE Sites Selected for Estimates of Regional Background 2-64
2-7 Estimates of Long-Term Means, Daily Standard Deviations and 99th Percentiles
of PM25 Background Concentrations (ug/m3) 2-64
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 PM25 Risk Assessment ;' 4-19
4-3 Morbidity Health Endpoints, Urban Locations, and Studies Selected for Use in
the PM10.25 Risk Assessment 4-19
4-4 Summary of PM Ambient Air Quality Data for Risk Assessment Study Areas
4-21
4-5 Relevant Population Sizes for PM Risk Assessment Locations 4-29
4-6 Baseline Mortality Rates for 2001 for PM2 5 Risk Assessment Locations 4-30
4-7 Baseline Hosphalization Rates for PM Risk Assessment Locations 4-32
4-8 Sensitivity Analyses 4-42
January 2005
Vlll
Draft - Do Not Quote or Cite
-------
4-9 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 ..- 4-55
4-10 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 4-60
4-11 Comparison of Annual Estimates of Short- and Long-Term Exposure Mortality
Reductions Associated with Just Meeting the Current PM15 Standards 4-63
4-12 Alternative Sets of PM2S Standards Considered in the PM25 Risk Assessment 4-69
4-13 Estimated Design Values for Annual and 98th and 99th Percentile Daily PM2 s
Standards Based on 2001-2003 Air Quality Data 4-69
4-14 Alternative PM1(WS Standards Considered in the PM10_25 Risk Assessment 4-76
4-15 Estimated Design Values for 98th and 99th Percentile Daily PM10.2 5 Standards
Based on 2001-2003 Air Quality Data 4-76
5-1 Estimated PM2S-related Annual Incidence of Total Mortality when Current PM25
Standards are Met (Base Case and Assumed Alternative Hypothetical Thresholds) . 5-14
5-2 Estimated Percent Reduction in PM2 5-related Long-term Mortality Risk (ACS
Extended Study) for Alternative Standards Relative to Current Standards (Base
Case and Assumed Alternative Hypothetical Thresholds) 5-28
5-3 a Predicted percent of counties with monitors (and percent of population in counties
with monitors) not likely to meet alternative annual and 24-hour (98th percentile
form) PM2 5 standards ..: 5-38
5-3b Predicted percent of counties with monitors (and percent of population in counties
with monitors) not likely to meet alternative annual and 24-hour (99th percentile
form) PM2 5 standards 5-40
5-4 Estimated Percent Reduction in PM2 5-attributable Short-term Risk
(mortality/morbidity) for Alternative Standards Relative to Meeting Current
Standards (Base Case and Assumed Alternative Hypothetical Thresholds 5-43
5-5 Estimated PM10.2J-related Annual Incidence of Hospital Admissions and Cough in
Children with 2003 Air Quality in Areas that Meet the Current PM10 Standards
(Base Case and Assumed Alternative Hypothetical Thresholds) 5-61
January 2005 ' ix Draft - Do Not Quote or Cite
-------
5-6a Predicted percent of counties with monitors (and percent of population in counties
with monitors) not likely to meet alternative 24-hour (98th percentile form) PM]0.
25 standards or current PM10 standards 5-70
5-6b Predicted percent of counties with monitors (and percentage of population in
counties with monitors) not likely to meet alternative 24-hour (99th percentile
form) PM10.25 standards or current PM10 standards 5-71
January 2005
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 PM showing fine and coarse particles and
the fractions collected by size-selective samplers '._ 2-7
2-3 Regions used in PM Staff Paper data analyses summaries 2-21
2-4 Distribution of annual mean PM2 5 and estimated annual mean PM10.2 5
concentrations by region, 2001-2003 2-23
2-5 Distribution of 98th percentile 24-hour average PM2S and estimated PM,0.25 '
concentrations by region, 2001-2003 2-24
2-6 County-level maximum annual mean PM25 concentrations, 2001-2003 2-25
2-7 County-level maximum 98th percentile 24-hour average PM2 5 concentrations,
2001-2003 ......:. . 2-26
2-8 Regional trends in annual average PM2 5 concentrations in the EPA FRM network,
1999-2003 2-28
2-9 Average annual average trend in fine particle mass, ammonium sulfate,
ammonium nitrate, total carbon, and crustal material at IMPROVE sites, 1993-
2003 '....' .' ' 2-29
2-10 County-level maximum annual mean PM10 concentrations, 2001-2003 2-33
2-11 County-level maximum 98th percentile 24-hour average PM10 concentrations,
2001-2003 . . ..- 2-34
2-12 Estimated county-level maximum annual mean PM,0.2S concentrations, 2001-
2003 :... ....... 2-36
2-13 Estimated county-level maximum 98th percentile 24-hour average PM10.2.5
concentrations, 2001-2003 2-37
2-14 Average measured annual average PMi0.2 s concentration trend at IMPROVE
sites, 1993-2003 '. ". 2-38
January 2005 xi Draft - Do Not Quote or Cite
-------
2-15 Annual average composition of PM25 by region, 2003 2-42
2-16 Average PM]0.25 PM25, and PM01 (ultrafme) chemical composition at an EPA
'supersite' monitor in Los Angeles, CA, 10/2001 to 9/2002 2-44
2-17 Distribution of ratios of PM25 to PM10 by region, 2001-2003 ..' 2-45
2-18 Regional average correlations of 24-hour average PM by size fraction 2-47
2-19 Urban 24-hour average PM2 5 concentration distributions by region and month,
2001-2003 2-48
2-20 Urban 24-hour average estimated PM10^2 5 concentration distributions by region
and month, 2001-2003 2-49
2-21 Seasonal average composition of urban PM25 by region, 2003 2-51
2-22 Seasonal average composition of rural PM25 by region, 2003 2-52
2-23 Distribution of annual mean vs. 98th percentile 24-hour average PM2 5
concentrations, 2001-2003 2-53
2-24 Distribution of estimated annual mean vs. 98th percentile 24-hour average PM10.2 5
concentrations, 2001-2003 2-54
2-25 Hourly average PM25 and PM,0.25 concentrations at a Greensboro, NC monitoring
site, 2001-2003 2-56
2-26 Seasonal hourly average PM2s and PM10_25 concentrations at a Greensboro, NC
monitoring site, 2001-2003 2-57
2-27 Hourly average PM_2 5 and PM10.2 5 concentrations at an El Paso, TX monitoring
site, 2001-2003 2-58
2-28 Hourly PM2 5 and PM10.2 5 concentrations at an El Paso, TX monitoring site, April
26, 2002-April 27, 2002 2-59
2-29 Relationship between light extinction, deciviews, and visual range 2-76
3-1 Excess risk estimates for total nonaccidental, cardiovascular, and respiratory
mortality with short-term exposure to PM in single-pollutant models for U.S. and
Canadian studies '. 3-18
January 2005
XII
Draft - Do Not Quote or Cite
-------
3-2 Excess risk estimates for hospital admissions and emergency department visits for
cardiovascular and respiratory diseases in single-pollutant models from U.S. and
Canadian studies ; 3-24
3-3 Associations between PM25 and total mortality from U.S. studies, plotted against
gaseous pollutant concentrations from the same locations .3-54
i
3-4 Natural logarithm of relative risk for total and cause-specific mortality with long-
term exposure to PM2 5.'. 3-65
4-1 Major Components of Paniculate Matter Health Risk Assessment 4-12
4-2 Estimated Annual Percent of Total (Non-Accidental) Mortality Associated with
Short-Term Exposure to PM25 (and 95 Percent Confidence Interval): Single-
Pollutant, Single-City Models ' 4-43
4-3 Estimated Annual Percent of Health Effects Associated with' Short-Term
Exposure to PM25 (and 95 Percent Confidence Interval): Results Based on Single-
City versus Multi-City Models '.C: 4-44
4-4 Estimated Annual Percent of Health Effects Associated with Short-Term
Exposure to PM25 (and 95 Percent Confidence Interval): Results Based on Single-
Pollutant versus Multi-Pollutant Models 4-45
4-5 Estimated Annual Percent of Mortality Associated with Long-Term Exposure to
PM25 (and 95 Percent Confidence Interval): Single-Pollutant Models 4-46
4-6 Estimated Annual Percent of Mortality Associated with Long-Term Exposure to
PM2 5 (and 95 Percent Confidence Interval): Single-Pollutant and Multi-Pollutant
Models (Based on Krewski et al. (2000) - ACS Study) 4-47
4-7 Estimated Annual Percent of Health Effects Associated with Short-Term
" . Exposure to PM10.25 (and 95 Percent Confidence Interval) 4-49
4-8a Distribution of 24-Hour PM25 Concentrations in Detroit (2003 Air Quality Data) .. 4-58
4-8b Estimated Non-Accidental Mortality in Detroit Associated with PM25
Concentrations (2003 Air Quality Data) (Based on Itp, 2003 4-58
4-9 .Estimated Annual Percentage Reduction of Health Risks Associated with Rolling
Back PM2S Concentrations to Just Meet the Current Standards (and 95 Percent
Confidence Intervals): Non-Accidental Mortality Associated with Short-Term
Exposure to PM2 5 4-61
January 2005 xiii Draft - Do Not Quote or Cite
-------
4-10
4-11
4-12
4-13
4-14
4-15
Estimated Annual Percentage Reduction of Health Risks Associated with Rolling
Back PM25 Concentrations to Just Meet the Current Annual Standards (and 95
Percent Confidence Interval): Mortality Associated with Long-Term Exposure to
PM,<
4-62
Estimated Annual Reduction in Short-Term Exposure Mortality Associated with
Rolling Back PM25 Concentrations that Just Meet the Current Annual Standard of
15 ug/m3 and the Current Daily Standard of 65 ug/m3 to PM25 Concentrations
that Just Meet Alternative Suites of PM25 Annual and Daily 98th Percentile
Standards: Detroit, MI, 2003 4-70
Estimated Annual Reduction in Short-Term Exposure Mortality Associated with
Rolling Back PM2 5 Concentrations that Just Meet the Current Annual Standard of
15 ug/m3 and the Current Daily Standard of 65 ug/m3 to PM25 Concentrations
that Just Meet Alternative Suites of PM2.5 Annual and Daily 99th Percentile
Standards: Detroit, MI, 2003 4-71
Estimated Annual Reduction in Long-Term Exposure Mortality Associated with
Rolling Back PM2 s Concentrations that Just Meet the Current Annual Standard of
15 ug/m3 and the Current Daily Standard of 65 ug/m3 to PM23 Concentrations
that Just Meet Alternative Suites of PM25 Annual and Daily 98th Percentile
Standards: Detroit, MI, 2003 4-72
Estimated Annual Reduction in Long-Term Exposure Mortality Associated with
Rolling Back PM2 5 Concentrations that Just Meet the Current Annual Standard of
15 ug/m3 and the Current Daily Standard of 65 ug/m3 to PM25 Concentrations
that Just Meet Alternative Suites of PM25 Annual and Daily 99th Percentile
Standards: Detroit, MI, 2003 '.....' 4-73
Estimated Annual Reduction of Hospital Admissions for Ischemic Heart Disease
Associated with Rolling Back "As Is" PM,0.2 5 Concentrations to PM10_2 5
Concentrations that Just Meet Alternative PM10.25 Daily Standards: Detroit, MI,
2003 4-77
6-1 PM25 concentration differences between eastern and western areas and between
rural and urban areas for 2003 6-8
6-2 Distribution of hourly and 24-hour average relative humidity at eastern and
western U.S. National Weather Service sites, 2003 6-10
6-3 Relationship between reconstructed light extinction (RE) and 24-hour average
PM25, 2003 6-11
January 2005
xiv
Draft - Do Not Quote or Cite
-------
6-4 Model slope for relationship between reconstructed light extinction (RE) and
hourly PM25 (increase in RE due to incremental increase in PM25), 2003 6-13
6-5 Relationship between reconstructed light extinction (RE) and 12 p.m. - 4 p.m.
average PM25, 2003 -. 6-14
6-6 Illustration of the nitrogen cascade 6-34
7-1 Distributions of PM2 s concentrations for 12 p.m. - 4 p.m. corresponding to visual
ranges of 25 km (panel a), 30 km (panel b), and 35 km (panel c) - by region 7-11
7-2 Estimated exceedances (%) of various PM25 levels for 12 p.m. - 4 p.m. based on
daily county maximum, 2001-2003 • ' 7-15
January 2005
xv
Draft - Do Not Quote or Cite
-------
Abbreviations and Acronyms
AC Automated colorimetiy
ACS American Cancer Society
AHSMOG Adventist Health and Smoke Study
AIRS Aerometric Information Retrieval System
ANC Acid neutralizing capacity
APHEA Air Pollution and Health, a European Approach
AQCD Air Quality Criteria Document
AQS . Air Quality System
ARS Air Resource Specialists, Inc.
ASOS Automated Surface Observing System
BC Black carbon
BS British or black smoke
CAA Clean Air Act
CAMM Continuous Ambient Mass Monitor
CAP Concentrated ambient particles
CASAC Clean Air Scientific Advisory Committee
CASTNet Clean Air Status and Trends Network
CB Base cation
CD Criteria Document
CDC Centers for Disease Control
CDPHE Colorado Department of Public Health and Environment
CFR Code of Federal Regulations
CL Critical loads
C:N Carbon-to-nitrogen ratio
CO Carbon monoxide
COH Coefficient of haze
COPD Chronic obstructive pulmonary disease
CPSC Consumer Product Safety Commission
C-R Concentration-response
CSS Coastal sage scrub community
CV Contingent valuation
EC Elemental carbon
ECG Electrocardiogram
ED Emergency department
EEA Essential Ecological Attribute
January 2005
xvi
Draft - Do Not Quote or Cite
-------
EMAP Environmental Monitoring and Assessment Program
EPA Environmental Protection Agency
EPEC Ecological Processes and Effects Committee
ERP . Episodic Response Project
FDMS Filter Dynamics Measurement System
FLM Federal Land Manager
FRM Federal reference method •
GAM Generalized additive models
GCVTC ' Grand Canyon Visibility Transport Commission
GLM Generalized linear models
HAPs Hazardous air pollutants
HEI Health Effects Institute
HF Heart failure
hosp. adm. Hospital admissions
1C Ion chromatography
IPS Integrated Forest Study'
IHD Ischemic heart disease
IMPROVE mteragency Monitoring of Protected Visual Environments
LML Lowest measured level
LPC Laser particle counter
LRS Lower respiratory symptoms
mort. Mortality • . .- .
NAAQS National ambient air quality standards
NADP , National Atmospheric Deposition Program
NAPAP National Acid Precipitation Assessment Program
NCEA • National Center for Environmental Assessment
NDDN National Dry Deposition Network
NEG/ECP New England Governors/Eastern Canadian Premiers
NMMAPS National Mortality and Morbidity Air(Pollution Study
N2 Nonreactive, molecular nitrogen
NO2 Nitrogen dioxide
non-accid
mort Non-accidental mortality
Nr Reactive nitrogen
NSMPS Nano-scanning mobility particle sizer
NuCM Nutrient cycling model
NWS National Weather Service
January 2005
xvii
Draft - Do Not Quote or Cite
-------
O3 Ozone
OAQPS Office of Air Quality Planning and Standards
OAR Office of Air and Radiation
OC Organic carbon
ORD Office of Research and Development
OSHA Occupational Safety and Health Administration
PAHs Polynuclear aromatic hydrocarbons
pneum. Pneumonia
PTEAM EPA's Particle Total Exposure Assessment Methodology
PCBs Polychlorinated biphenyls
PCDD/F Polychlorinated dibenzo-p-dioxins/dibenzofurans
PM Particulate matter
PMio-2.5 Particles less than or equal to 10 um in diameter and greater than 2.5 urn in
diameter
PM25 Particles less than or equal to 2.5 [am in diameter
PMW Particles less than or equal to 10 um in diameter
PnET-BGC A forest net productivity model (PnET) linked to a soil model (BGC)
POPs Persistent organic pollutants
PRB Policy relevant background
REVEAL Regional Visibility Experimental Assessment in the Lower Fraser Valley
RR Relative risk
SAB Science Advisory Board
* SMPS Standard scanning mobility particle sizer
SO2 Sulfur dioxide
SO4 Sulfate
SOCs Semivolatile organic compounds
STN PM2 5 Chemical Speciation Trends Network
SP Staff Paper
TEOM Tapered Element Oscillating Microbalance sensor
TIME/LTM Temporally Integrated Monitoring of Ecosystems/Long-Term Monitoring Proj ect
TL Target load
TMO Thermal manganese oxidation method
. TOR Thermal/optical reflectance method
TOT Thermal/optical transmission method
TSD Technical support document
TSP Total suspended particulates
|ig micrograms
January 2005
xvm
Draft - Do Not Quote or Cite
-------
jig/m3 micrograms per cubic meter
UNEP United Nations Environmental Program
URS Upper respiratory symptoms
U.S. United States
UV Ultraviolet
UV-B Ultraviolet-B
Vd Deposition velocity •
VOCs Volatile organic compounds
WMO World Meteorological Organization
XRF X-ray fluorescence
January 2005 xix Draft - Do Not Quote or Cite
-------
1 1. INTRODUCTION
2 1.1 PURPOSE
3 This draft Staff Paper, prepared by staff in the U.S. Environmental Protection Agency's
4 (EPA) Office of Air Quality Planning and Standards (OAQPS), evaluates the policy implications
5 of the key studies and scientific information contained in the document, Air Quality Criteria for
6 Paniculate Matter (EPA, 2004; henceforth referred to as the Criteria Document (CD) and cited
7 as CD), prepared by EPA's National Center for Environmental Assessment (NCEA). This
8 document, which builds upon an earlier first draft Staff Paper (EPA, 2003), also presents and
9 interprets results from updated and expanded staff analyses (e.g., air quality analyses, human
10 health risk assessments, and visibility analyses) that staff believes should be considered in EPA's
11 current review of the national ambient air quality standards (NAAQS) for particulate matter
12 (PM). This draft Staff Paper presents provisional staff conclusions and recommendations as to
13 potential revisions of the primary (health-based) and secondary (welfare-based) PM NAAQS,
14 based on consideration of the available scientific information and analyses and related
15 limitations and uncertainties. The final version of this document will be informed by comments
16 received through an independent scientific review and public comments on this draft document.
17 The policy assessment presented in this document is intended to help "bridge the gap"
18 between the scientific review contained in the CD and the judgments required of the EPA
19 Administrator in determining whether, and if so, how, it is appropriate to revise the NAAQS for
20 PM. This assessment focuses on the basic elements of PM air quality standards: indicators,
21 averaging times, forms1, and levels. These elements, which serve to define each standard within
22 the suite of PM NAAQS, must be considered collectively in evaluating the health and welfare
23 protection afforded by the standards.
24 While this document should be of use to all parties interested in the PM NAAQS review,
25 it is written for those decision makers, scientists, and staff who have some familiarity with the
26 technical discussions contained in the CD.
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.
January 2005
1-1
Draft - Do Not Quote or Cite
-------
1 1.2 BACKGROUND
2 1.2.1 Legislative Requirements
3 Two sections of the Clean Air Act (Act) govern the establishment and revision of the
4 NAAQS. Section 108 (42 U.S.C. 7408) directs the Administrator to identify "air pollutants" that
5 "in his judgment, may reasonably be anticipated to endanger public health and welfare" and
6 whose "presence... in the ambient air results from numerous or diverse mobile or stationary
7 sources" and, if listed, to issue air quality criteria for them. These air quality criteria are
8 intended to "accurately reflect the latest scientific knowledge useful in indicating the kind and
9 extent of identifiable effects on public health or welfare which may be expected from the
10 . presence of [a] pollutant in ambient air...."
11 Section 109 (42 U.S.C. 7409) directs the Administrator to propose and promulgate
12 "primary" and "secondary" NAAQS for pollutants identified under section 108. Section
13 109(b)(l) defines a primary standard as one "the attainment and maintenance of which in the
14 judgment of the Administrator, based on such criteria and allowing an adequate margin of safety,
15 are requisite to protect the public health."2 A secondary standard, as defined in Section
16 109(b)(2), must "specify a level of air quality the attainment and maintenance of which, in the
17 judgment of the Administrator,' based on such criteria, is requisite to protect the public welfare
18 from any known or anticipated adverse effects associated with the presence of [the] pollutant in
19 the,ambient air."3
20 In setting standards that are "requisite" to protect public health and welfare, as provided
21 in section 109(b), EPA's task is to establish standards that are neither more nor less stringent
22 than necessary for these purposes. In so doing, EPA may not consider the costs of implementing
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, 9l"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."
January 2005 1 -2 Draft - Do Not Quote or Cite
-------
1 the standards. See generally Whitman v. American Trucking Associations, 531 U.S. 457,464,
2 475-76 (2001).
3 The requirement that primary standards include an adequate margin of safety was
4 intended to address uncertainties associated with inconclusive scientific and technical
5 information available at the time of standard setting. It was also intended to provide a
6 reasonable degree of protection against hazards that research has not yet identified. Lead
1 Industries Association v. EPA, 647 F.2d 1130,1154 (D.C. Cir 1980), cert, denied. 101 S. Ct. 621
8 (1980); American Petroleum Institute v. Costle, 665 F.2d 1176,1186 (D.C. Cir. 1981), cert.
9 denied. 102 S.Ct 1737 (1982). Both kinds of uncertainties are components of the risk associated
10 with pollution at levels below those at which human health effects can be said to occur with
11 reasonable scientific certainly. Thus, in selecting primary standards that include an adequate
12 margin of safety, the Administrator is seeking not only to prevent pollution levels that have been
13 demonstrated to be harmful but also to prevent lower pollutant levels that may pose an
14 unacceptable risk of harm, even if the risk is not precisely identified as to nature or degree.
15 In selecting a margin of safety, the EPA considers such factors as the nature and severity
16 of the health effects involved, the size of the sensitive population(s) at risk, and the kind and
17 degree of the uncertainties that must be addressed. The selection of any particular approach to
18 providing an adequate margin of safety is a policy choice left specifically to the Administrator's
19 judgment. Lead Industries Association v. EPA, supra. 647 F.2d at 1161-62.
20 Section 109(d)(l) of the Act requires that "not later than December 31,1980, and at 5-
21 year intervals thereafter, the Administrator shall complete a thorough review of the criteria
22 published under section 108 and the national ambient air quality standards ... and shall make
23 such revisions in such criteria and standards and promulgate such new standards as may be
24 appropriate...." Section 109(d)(2) requires that an independent scientific review committee
25 "shall complete a review of the criteria... and the national primary and secondary ambient air
26 quality standards... and shall recommend to the Administrator any new... standards and
27 revisions of existing criteria and standards as may be appropriate...." Since the early 1980's,
28 this independent review function has been performed by the Clean Air Scientific Advisory
29 Committee (CASAC), a standing committee of EPA's Science Advisory Board.
January 2005
1-3
Draft - Do Not Quote or Cite
-------
1 1.2.2 History of PMNAAQS Reviews
2 Paniculate matter is the generic term for a broad class of chemically and physically
3 diverse substances that exist as discrete particles (liquid droplets or solids) over a* wide range of:
4 sizes. Particles originate from a variety of anthropogenic stationary and mobile sources as well
5 as natural sources. Particles may be emitted directly or formed in the atmosphere by
6 transformations of gaseous emissions such as sulfur oxides, nitrogen oxides, and volatile organic
7 compounds. The chemical and physical properties of PM vary greatly with time, region,
8 meteorology, and source category, thus complicating the assessment of health and welfare
9 effects.
10 EPA first established national ambient air quality standards for PM in 1971, based on the
11 original criteria document (DHEW, 1969). The reference method specified for determining
12 attainment of the original standards was the high-volume sampler, which collects.PM up to a
13 nominal size of 25 to 45 micrometers (urn) (referred to as total suspended particles or TSP). The
14 primary standards (measured by the indicator TSP) were 260 ug/m3, 24-hour average, not to be
15 exceeded more than once per year, and 75 ug/m3, annual geometric mean. The secondary
16 standard was 150 ug/m3,24-hour average,'not to be exceeded more than once per year.
17 In October 1979 (44 FR 5 6731), EPA announced the first periodic review of the criteria
18 and NAAQS for PM, and significant revisions to the original standards were promulgated in
19 1987 (52 FR 24854, July 1,1987). In that decision, EPA changed the indicator for particles from
20 TSP to PM10, the latter including particles with a mean aerodynamic diameter4 less than or equal
21 to 10 um, which delineates that subset of inhalable particles small enough to penetrate to the
22 thoracic region (including the fracheobronchial and alveolar regions) of the respiratory tract
23 (referred to as thoracic particles). EPA also revised the level and form of the primary standards
24 by: (1) replacing the 24-hour TSP standard with a 24-hour PM10 standard of 150 ug/m3 with no
25 more than one expected exceedance per year; and (2) replacing the annual TSP standard with a
26 PM10 standard of 50 ug/m3, annual arithmetic mean. The secondary standard was revised by
4 The more precise term is 50 percent cut point or 50 percent diameter (D5<>). 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.
January 2005 1-4 " Draft - Do Not'Quote or Cite
-------
1 replacing it with 24-hour and annual standards identical in all respects to the primary standards.
2 The revisions also included a new reference method for the measurement of PM,0 in the ambient
3 air and rules for determining attainment of the new standards. On judicial review, the revised
4 standards were upheld in all respects. Natural Resources Defense Council v. Administrator, 902
5 F. 2d 962 (D.C. Cir. 1990), cert, denied. 111 S. Ct 952 (1991).
6 In April 1994, EPA announced its plans for the second periodic review of the criteria and
7 NAAQS for PM, and promulgated significant revisions to the NAAQS in 1997 (62 FR 38652,
8 July 18, .1997). In that decision, EPA revised the PM NAAQS in several respects. While it was
9 determined that the PM NAAQS should continue to focus on particles less than or equal to 10
10 urn in diameter, it was also determined that the fine and coarse fractions of PM 10 should be
11 considered separately. New standards were added, using PM2 5, referring to particles with a
12 nominal mean aerodynamic diameter less than or equal to 2.5 um, as the indicator for fine
13 particles, with PM10 standards retained for the purpose of regulating the coarse fraction of PM10
14 (referred to as thoracic coarse.particles or coarse-fraction particles; generally including particles
15 with a nominal mean aerodynamic diameter greater than 2.5 um and less than or equal to 10 um,
16 or PMi0.2.5). EPA established two new PM25 standards: an annual standard of 15 ug/m3, based
17 on the 3-year average of annual arithmetic mean PM2 5 concentrations from single or multiple
18 community-oriented monitors; and a 24-hour standard of 65 ug/m3, based on the 3-year average
19 of the 98th percentile of 24-hour PM25 concentrations at each population-oriented monitor within
20 an area. A new reference method for the measurement of PM2 5 in the ambient air was also
21 established, as were rules for determining attainment of the new standards. To continue to
22 address thoracic coarse particles, the annual PM]0 standard was retained, while the 24-hour PM10
23 standard was revised to be based on the 99th percentile of 24-hour PMJO concentrations at each
24 monitor in an area. EPA revised the secondary standards by making them identical in all
25 respects to the primary standards.
26 1.2.3 Litigation Related to the 1997 PM Standards
27 Following promulgation of the revised PM NAAQS, petitions for review were filed by a
28 large number of parties, addressing a broad range of issues. In May 1998, a three-judge panel of
January 2005
1-5
Draft - Do Not Quote or Cite
-------
Q
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
7 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 PM10 standards, concluding in
9 part that PM10 is a "poorly matched indicator for coarse paniculate pollution" because it includes
10 fine particles. Id. at 1053-55. Pursuant to the court's decision, EPA deleted 40 CFR section
11 50.6 (d), the regulatory provision controlling the transition froni the pre-existing 1987 PM,0
12 standards to the 1997 PM10 standards (65 FR' 80776, December 22, 2000). The pre-existing 1987
13 PM,0 standards remained in place. Ii at 80777. In the current review, EPA is addressing
14 thoracic coarse particles in part by considering standards based on an indicator of PM,0.25.
JPN 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 oh 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 the constitutional and cost issues. Whitman v. American'
28 Trucking Associations, 531 U. S. 457,464, 475-76. On the constitutional issue, the Court held
29 that the statutory requirement that NAAQS.be "requisite" to protect public health with an
©
January 2005 ' 1-6 Draft - Do Not Quote or Cite
-------
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
elements of the PM standards under review. While the CD focuses on new scientific
information available since the last criteria review, it appropriately integrates that information
with scientific criteria from previous reviews. The quantitative analyses presented herein (and
described in more detail in a number of technical support documents) are based on the most
recently available air quality information, so as to provide current characterizations of PM air
quality patterns, estimated human health risks related to exposure to ambient PM, and PM-
related visibility impairment .
Partly as a consequence, of EPA's decision in the last review to consider fine particles and
thoracic coarse particles separately, much new information is now available on PM air quality
and human health effects directly in terms of PM2 5 and, to a much more limited degree, PM10.25.
This information adds to the body of evidence on PM10 that has continued to grow since the
introduction of that indicator in the first PM NAAQS review. Since the purpose of this review is
to evaluate the adequacy of the current standards that separately address fine and thoracic,coarse
particles, staff has focused this policy assessment and associated quantitative analyses primarily
on the evidence related directly to PM25 and PM10.2 5. In so doing, staff has considered PM10-
related evidence primarily to help inform our understanding of key issues and to help interpret
and provide context for the more limited PM2 5 and PM10.2 5 evidence.
Following this introductory chapter, this draft Staff Paper is organized into three main
parts: the characterization of ambient PM; PM-related health effects and primary PM NAAQS;
and PM-related welfare effects and secondary PM NAAQS. The characterization of ambient PM
is presented in Chapter 2, which focuses on properties of ambient PM, measurement methods,
spatial and temporal patterns in ambient PM concentrations, PM background levels, and ambient
PM relationships with human exposure and with visibility impairment. Thus, Chapter, 2 provides
information relevant to both the health and welfare assessments in the other two main parts of ,
this document.
Chapters 3 through 5 comprise the second main part of this draft Staff Paper dealing with
human health and primary standards. Chapter 3 presents a policy-relevant assessment of PM
health effects evidence, including an overview of the evidence, key human health-related
conclusions from the CD, and an examination of issues related to the quantitative assessment of
o
January 2005
1-9
Draft - Do Not Quote or Cite
-------
1 the epidemiologic health evidence. Chapter 4 presents a quantitative assessment of PM-related
2 health risks, including risk estimates for current air quality levels as well as those associated with
3 just meeting the current NAAQS and various alternative standards that might be considered in
4 this review. Chapter 5 presents the staff review of the current primary standards for fine and
5 thoracic coarse particles. This chapter begins with a discussion of the broader approach used by
6 staff in this review of the primary PM NAAQS than in the last review, generally reflecting both
7 evidence-based and quantitative risk-based considerations. This review includes consideration
8 of the adequacy of the current standards, conclusions as to alternative indicators, averaging
9 - times, levels and forms, and provisional recommendations on ranges of alternative primary
10 standards for consideration by the Administrator.
11 Chapters 6 and 7 comprise the third main part of this draft Staff Paper dealing with
12 welfare effects and secondary standards. Chapter 6 presents a policy-relevant assessment of PM
13 welfare effects evidence, including evidence related to visibility impairment as well as to effects
14 on vegetation and ecosystems, climate change processes, and man-made materials. This
15 chapter's emphasis is on visibility impairment, reflecting the availability of a significant amount
16 of policy-relevant information and staff analyses which serve as the basis for staff consideration
17 of a secondary standard specifically for visibility protection. Chapter 7 presents the staff review
18 of the current secondary standards, beginning with a discussion of the approach used by staff in
19 this review of the secondary PM NAAQS. This review includes consideration of the adequacy
s 20 of the current standards, conclusions as to alternative indicators, averaging times, levels and
21 forms, and provisional recommendations on ranges of alternative secondary standards for
22 consideration by the Administrator.
23 The staff conclusions and recommendations presented herein are provisional; final staff
24 conclusions and recommendations, to be presented in the final version of this document, will be
25 informed by comments received from CAS AC and the public in their reviews of this draft
26 document.
January 2005 1-10 Draft - Do Not Quote or Cite
-------
REFERENCES
2 Environmental Protection Agency. (2001) Review of the National Ambient Air Quality Standards for Particulate
3 Matter: Policy Assessment of Scientific and Technical Information - Preliminary Draft OAQPS Staff
4 Paper. June.
5 Environmental Protection Agency. (2003) Review of the National Ambient Air Quality Standards for Particulate
6 Matter: Policy Assessment of Scientific and Technical Information - First Draft OAQPS Staff Paper.
7 • August.
8 Environmental Protection Agency. (2004) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC:
9 Office of Research and Development; report no. EPA/600/P-99/002a,bF. October.
10 U.S. Department of Health, Education and Welfare (DEHW). (1969) Air Quality Criteria for Particulate Matter.
11 U.S. Government Printing Office, Washington DC, AP-49.
January 2005
1-11
Draft - Do Not Quote or Cite
-------
1 2. CHARACTERIZATION OF AMBIENT PM
2
3 2.1 INTRODUCTION
4 This chapter generally characterizes various classes of ambient PM in terms of physical
5 and chemical properties, measurement methods, recent concentrations'and trends, and ,
(
6 relationships with human exposure and visibility impairment. This information is useful for
7 interpreting the available health and welfare effects information, and for making
8 recommendations on appropriate indicators for primary and secondary PM standards. The
9 information presented in this chapter was drawn from the CD and additional analyses of data
10 from various PM monitoring networks. ' •
11 Section 2.2 presents information on the basic physical and chemical properties of classes
12 of PM. Section 2.3 presents information on the methods used to measure ambient PM and some
13 important considerations in the design of these methods. Section 2.4-presents data on PM
14 concentrations, trends, and spatial patterns in the U.S. Section 2.5 provides information on the
15 temporal variability of PM. Much of the information in Sections 2.4 and 2.5 is derived from
16 analyses of data collected by the nationwide networks of PM2 5 and PM10 monitors through 2003.
17 Section 2.6 defines and discusses background levels of ambient PM. Section 2.7 addresses the
18 relationships between ambient PM levels and human exposure to PM. Section 2.8 addresses the
19 relationship between ambient PM2 5 levels and visibility impairment. An appendix to this
20 chapter (Appendix 2-A) discusses sources of ambient PM and provides a summary of national
21 estimates of source emissions.
22 • ' " -
23 2.2 PROPERTIES OF AMBIENT PM
24 PM represents a broad class of chemically and physically diverse substances that exist as
25 discrete particles in the condensed (liquid or solid) phase. Particles can be characterized by size,
26 formation mechanism, origin, chemical composition, and atmospheric behavior. This section
27 generally focuses on size since classes of particles have historically been characterized largely in
28 that manner. Fine particles and coarse particles, which are defined in Section 2.2.1.1, are
29 - relatively distinct entities with fundamentally different sources and formation processes,
30 chemical composition, atmospheric residence times and behaviors, transport distances, and
January 2005 2-1 Draft - Do Not Quote or Cite
-------
1 optical and radiative properties. The CD concludes that these differences justify consideration of
2 fine and coarse particles as separate subclasses of PM pollution (CD, pp. 2-111 and 9-21).
3
4 2.2.1 Particle Size Distributions
5 Particle properties and their associated health and welfare effects differ by size. The
6 diameters of atmospheric particles span 5 orders of magnitude, ranging from 0.001 micrometers
7 to 100 micrometers (um).1 The size and associated composition of particles determine their
8 behavior in the respiratory system, including how far the particles are able to penetrate, where
9 they deposit, and how effective the body's clearance mechanisms are in removing them.
10 Furthermore, particle size is one of the most important parameters in determining the residence
11 time and spatial distribution of particles in ambient air, key considerations in assessing exposure.
12 Particle size is also a major determinant of visibility impairment, a welfare effect linked to
13 ambient particles. Particle surface area, number, chemical composition, and water solubility all
14 vary with particle size, and are also influenced by the formation processes and emissions
15 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 specific PM sampling devices. The terminology used in this Staff Paper
19 for describing these classifications is summarized in Table 2-1 and discussed in the following
20 subsections.
21
22 2.2.1.1 Modes
23 Based on extensive examinations of particle size distributions in several U.S. locations in
24 the 1970's, Whitby (1978) found that particles display a consistent multi-modal distribution over
25 several physical metrics, such as mass or volume (CD, p. 2-7). These modes are apparent in
26 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).
January 2005 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
Thoracic Coarse Particles
Fine Particles
*
Accumulation-Mode Particles
Ultrafine Particles
Aitkin-Mode Particles
Nucleation-Mode Particles
The distribution of particles that are mostly larger than the intermodal
minimum in volume or mass distributions; also referred to as coarse-mode
particles. This intermodal minimum generally occurs between 1 and 3 urn.
A subset of coarse particles that includes particles that can be inhaled and
penetrate to the thoracic region (i.e., the tracheobronchial and the
gas-exchange regions) of the lung. This subset includes the smaller coarse
particles, ranging in size up to those with a nominal aerodynamic diameter less
than or equal to 10 microns.
The distribution of particles that are mostly smaller than the intermodal.
minimum in volume or mass distributions; this minimum generally occurs
between 1 and 3 um. This includes particles in the nucleation, Aitkin, and
accumulation modes.
A subset of fine particles with diameters above about 0.1 fun. Ultrafine
particles grow by coagulation or condensation and "accumulate" in this size
range.
A subset of fine particles with diameters below about 0.1 ^m, encompassing
the Aitkin and nucleation modes.
A subset of Ultrafine particles with diameters between about 0.01 and 0.1 um.
Freshly formed particles with diameters below about 0.01 um.
Sampling Measurements'
Total Suspended Particles (TSP)
PMU
PM2.5
PM
10-2.5
Particles measured by a high volume sampler as described in 40 CFR Part 50,
Appendix B. This sampler has a cut point of aerodynamic diameters that
varies between 25 and 40 um depending on wind speed and direction.
Particles measured by a sampler that contains a size fractionator (classifier)
designed with an effective cut point (50% collection efficiency) of 10 um
aerodynamic diameter. This measurement includes the fine particles and a
subset of coarse particles, and is an indicator for particles that can be inhaled
and penetrate to the thoracic region of the lung; also referrred to as thoracic
particles.
Particles measured by a sampler that contains a size fractionator (classifier)
designed with an effective cut point (50% collection efficiency) of 2.5 um
aerodynamic diameter. This measurement, which generally includes all fine
particles', is an indiator for fine particles; also referred to as fine-fraction
particles. A small portion of coarse particles may be included depending on
the sharpness of the sampler efficiency curve.
Particles measured directly using a dichotomous sampler or by subtraction of
particles measured by a PMj 5 sampler from those measured by a PM10
sampler. This measurement is an indicator for the coarse fraction of thoracic
particles; also referred to as thoracic coarse particles or coarse-fraction
particles.
January 2005
2-3
Draft - Do Not Quote or Cite
-------
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
volume by particle size.2 Panel (a) illustrates that by far, the largest number of ambient particles
in a typical distribution are very small, below 0.1 \im in.diameter, while panel (c) indicates most
of the particle volume, and therefore most of the mass, is found in particles with diameters larger
than 0.1 um.3 Most of the surface area (panel b) is between 0.1 and 1.0 \im. The surface area
distribution in panel (b) peaks around 0.2 jam. Distributions may vary across locations,
conditions, arid time due to differences in sources, atmospheric conditions, topography, and the
age of the aerosol.
As illustrated in panel (c) of Figure 2-1, volume distributions typically measured in
ambient air in the U.S. are found to be bimodal, with overlapping tails, and an intermodal
minimum between 1 and 3 um (CD, p. 2-25). The distribution of particles that are mostly larger
than this minimum make up the coarse mode and are called "coarse particles," and the
distribution of particles that are mostly smaller than the minimum are called "fine particles."
Fine particles can be subcategorized into smaller modes: "nucleation mode," "Aitkin mode,"
and "accumulation mode." Together, nucleation-mode and Aitkin-mode particles make up
"ultrafine particles.'"1 Ultrafine particles are apparent as the largest peak in the number
distribution in panel (a), and are also visible in the surface area distribution in panel (b).
Nucleation-mode and Aitkin-mode particles have relatively low mass and grow rapidly into
accumulation-mode particles, so they are not commonly observed as a separate mode in volume
or mass distributions. The accumulation mode is apparent as the leftmost peak in the volume
distribution in panel (c) and the largest peak in the surface area distribution in panel (b).
2.2.1.2 Sampler Cut Points
Another set of particle size classifications is derived from the characteristics of ambient
particle samplers. Particle samplers typically use size-selective inlets that are defined by their 50
percent cut point, which is the particle aerodynamic diameter at which 50 percent of particles of
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.
. Mass is proportional to volume times density.
4 Whitby (1978) did not identify multiple ultrafine particle modes between 0.01 and 0.1 um, and therefore
separate nucleation and Aitkin modes are not illustrated in Figure 2-1. See CD Figure 2-6 for a depiction of all
particle modes.
January 2005
2-4
Draft - Do Not Quote or Cite
-------
Figure 2-1. Distribution of coarse (c), accumulation (a), and nuclei (n) mode particles by
three characteristics: (a) number, N; (b) surface area, S; and (c) volume, V
for the grand average continental size distribution. DGV = geometric mean
diameter by volume; DCS = geometric mean diameter by surface area;
DGN = geometric mean diameter by number; Dp - particle diameter.
Source: Whitby (1978); CD, p. 2-8.
January 2005
2-5
Draft - Do Not Quote or Cite
-------
1 a specified diameter are captured by the inlet, and their penetration efficiency as a function of
2 particle size. The usual notation for these classifications is "PM/', where x refers to
3 measurements with a 50 percent cut point of x (am aerodynamic diameter. Because of the
4 overlap in the size distributions of fine and coarse-mode ambient particles, and the fact that inlets
5 do not have perfectly sharp cut points, no single sampler can completely separate them. Given a
6 specific size cut, the smaller the particles the greater the percentage of particles that are captured.
7 The objective of size-selective sampling is usually to measure particle size fractions that provide
8 a relationship to human health impacts, visibility impairment, or emissions sources.
9 The EPA has historically defined indicators of PM for NAAQS using cut points of
10 interest. Figure 2-2 presents an idealized distribution of ambient PM showing the fractions
11 collected by size-selective samplers. Prior to 1987, the indicator for the PM NAAQS was total
12 suspended particulate matter (TSP), and was defined by the design of the High Volume Sampler
13 (Hi Vol).5 As illustrated in Figure 2-2, TSP typically includes particles with diameters less than
14 about 40 urn, but could include even larger particles under certain conditions. When EPA
15 established new PM standards in 1987, the selection of PM10 as an indicator was intended to
16 focus regulatory attention on particles small enough to be inhaled and to penetrate into the
17 thoracic region of the human respiratory tract. In 1997, EPA established standards for fine
18 particles measured as PM2 5 (i.e., the fine fraction of PM10). The dashed lines in Figure 2-2
19 illustrate the distribution of particles captured by the PM10 Federal Reference Method (FRM)
20 sampler6, including all fine and some coarse particles, and the distribution captured by the PM2 5
21 FRM sampler7, including generally all fine particles and potentially capturing a small subset of
22 coarse particles. .
23 The EPA is now considering establishing standards for another PM indicator identified in
24 Table 2-1 as PM10_25, which represents the subset of coarse particles small enough to be inhaled
25 and to penetrate into the thoracic region of the respiratory tract (i.e., the coarse fraction of PM10,
26 or thoracic coarse particles). Section 2.3 discusses measurement methods for this indicator.
27
. 5 40 CFR Part 50, Appendix B, Reference Method for the Determination of Suspended Particulate Matter in
the Atmosphere (High-Volume Method).
6 40 CFR Part 50, Appendix J, Reference Method for the Determination of Particulate Matter as PM10 in the
Atmosphere.
7 40 CFR Part 50, Appendix L, Reference Method for the Determination of Fine Particulate Matter as PM25
in the Atmosphere.
January 2005 2-6 Draft - Do Not Quote or Cite
-------
;S
(€ui/ciuri)da
i
O)
a
00
a
I
a
o
•I
T3
t
I
-------
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 um diameter (EPA
11 1996a, p. 13-7). Some combustion-generated particles, such as fly ash, are also found as coarse
12 particles.
13 By contrast, a significant amount of fine particles are produced through combustion
14 processes and atmospheric chemistry reactions. Common directly emitted fine particles include
15 unburned carbon particles from combustion, and nucleation-mode particles emitted as
16 combustion-related vapors that condense within seconds of being exhausted to ambient air.
17 Fossil-fuel combustion sources include motor vehicles and off-highway equipment, power
18 generation facilities, industrial facilities, residential wood burning, agricultural burning, and
19 forest fires.
20 The formation and growth of fine particles are influenced by several processes
21 including: (1) nucleation (i.e., gas molecules coming together to form a new particle); (2)
22 condensation of gases onto existing particles; (3) coagulation of particles, the weak bonding of
23 two or more particles into one larger particle; (4) hygroscopic uptake of water; and (5) gas phase
24 reactions which form secondary PM. Gas phase material condenses preferentially on smaller
25 particles since they have the greatest surface area, and the rate constant for coagulation of two
26 particles decreases as the particle size increases. Thus, ultrafine particles grow into the
27 , accumulation mode, but accumulation-mode particles do not normally grow into coarse particles
28 (CD, p. 2-29).
29 Secondary formation processes can result in either new particles or the addition of PM to
30 pre-existing particles. Examples of secondary particle formation include: (1) the conversion of
31 sulfur dioxide (SO2) to sulfuric acid (H2SO4) droplets that further react with ammonia (NH3) to
32 form various sulfate particles (e.g., ammonium sulfate (NH4)2SO4 or ammonium bisulfate
January 2005
2-8
Draft - Do Not Quote or Cite
-------
1 NH4HSO4); (2) the conversion of nitrogen dioxide (NO2) to nitric acid (HNO3) vapor that reacts
2 further with ammonia to form ammonium nitrate (NH4NO3) particles; and (3) reactions involving
3 volatile organic compounds (VOC) yielding organic compounds with low ambient temperature
4 (saturation) vapor pressures that nucleate or condense on existing particles to form secondary
5 organic aerosol particles (CD, p. 3-65 to 3-71). In most of the ambient monitoring data displays
6 shown later in this chapter, the first two types of secondary PM are generally labeled phirally as
7 'sulfates' and 'nitrates' (respectively), which implies that the ammonium content is
8 encompassed. The third type of secondary PM may be lumped with the directly emitted
9 elemental carbon particles and labeled 'total carbonaceous mass/ or the two types of
10 carbonaceous PM may be reported separately as elemental carbon (EC) and organic carbon
11 (OC). •
12 . ' . ' "•
13 2.2.3 Chemical Composition
* * • *
14 Based on studies conducted in most parts of the U.S., the CD reports that a number of
15 chemical components of ambierit PM are found predominately in fine particles including:
16 sulfate, ammonium, and hydrogen ions; elemental carbon8, secondary organic compounds, and
17 primary organic species from cooking and combustion; and certain metals, primarily from •
18 combustion processes. Chemical components found predominately in coarse particles include:
19 ' crustal-related materials such as calcium, aluminum, silicon, magnesium, and iron; and primary
20 organic materials such as pollen, spores, arid plant and animal debris (CD, p. 2-38),
21 Some components, such as nitrate and potassium, may be found in both fine and coarse
22 particles. Nitrate in fine particles comes mainly from the reaction ofgas-phase nitric acid with
23 gas-phase ammonia to form ammonium nitrate particles. Nitrate in coarse particles comes •
24 primarily from the reaction ofgas-phase nitric acid with preexisting coarse particles (CD, p. 2-
25 38). Potassium in coarse particles comes primarily from soil, with additional contributions from
26 sea salt in coastal areas. Potassium in fine particles, generally not a significant contributor to
27 overall mass, comes mainly from emissions of burning wood, with infrequent but large
28 contributions from fireworks, as well as significant proportions from the tail of the distribution
29 of coarse soil particles (i.e., < 2.5 um in diameter) in areas with high soil concentrations.
Also called light absorbing carbon and black carbon.
January 2005 ' "2-9 Draft - Do Not Quote or Cite
-------
1 Many ambient particles also contain water (i.e., particle-bound water) as a result of an
2 equilibrium between water vapor and hygroscopic PM (CD, p. 2-40). Particle-bound water
3 influences the size of particles and in turn their aerodynamic and light scattering properties
4 (discussed in section 2.2.5). Particle-bound water can also act as a carrier to convey dissolved
5 gases or reactive species into the lungs which, in turn, may cause heath consequences. (CD, p.
6 , 2-112). The amount of particle-bound water in ambient particulate matter will vary with the
7 particle composition and the ambient relative humidity. Sulfates, nitrates, and some secondary
8 organic compounds are much more hygroscopic than elemental carbon (BC), primary organic
9 carbon (OC), and crustal material.
10
11 2.2.4 Fate and Transport
12 Fine and coarse particles typically exhibit different behaviors in the atmosphere. These
13 differences may affect several exposure-related considerations, including the representativeness
14 of central-site monitored values and the penetration of particles formed outdoors into indoor
15 spaces. The ambient residence time of atmospheric particles varies with size. Ultrafine particles
16 have a very short life, on the order of minutes to hours, since they grow rapidly into the
17 accumulation mode. However, their chemical content persists in the accumulation mode.
18 Ultrafine particles are also small enough to be removed through diffusion to falling rain drops.
19 Accumulation-mode particles remain suspended longer, due to collisions with air molecules, and
20 have relatively low surface deposition rates. They can be transported thousands of kilometers
21 and remain in the atmosphere for days to weeks. Accumulation-mode particles serve as
22 condensation nuclei for cloud droplet formation and are eventually removed from the
23 atmosphere in falling rain drops. Accumulation-mode particles that are not involved in cloud
24 processes are eventually removed from the atmosphere by gravitational settling and impaction on
25 surfaces.
26 By contrast, coarse particles can settle rapidly from the atmosphere with lifetimes
27 ranging from minutes to days depending on their size, atmospheric conditions, and their altitude.
28 Larger coarse particles are not readily transported across urban or broader areas, because they
29 are generally too large to follow air streams, and they tend to be easily removed by gravitational
30 settling and by impaction on surfaces. Smaller coarse particles extending into the tail of the
31 distribution can have longer lifetimes and travel longer distances, especially in extreme
January 2005
2-10
Draft - Do Not Quote or Cite
-------
1 circumstances, such as intercontinental dust storms (CD, p. 2-49). Coarse particles also are
2 readily removed by falling rain drops (CD, p. 2-50).
3 The characteristics of ultrafine, accumulation-mode, and coarse-mode particles that were
4 discussed in the preceding sections are summarized in Table 2-2.
5
6 2.2.5 Optical Properties of Particles
7 Particles and gases in the atmosphere scatter and absorb light and, thus, affect visibility.
8 As discussed in section 4.3 of the CD, the efficiency of particles in causing visibility impairment
9 depends on particle size, shape, and composition. Accumulation-mode particles are more
10 efficient per unit mass than coarse particles in causing visibility impairment. The accumulation-
11 mode particle components principally responsible for visibility impairment are sulfates, nitrates,
12 organic matter, and elemental carbon. Soil dust in the fine tail of the coarse particle distribution
13 can also impair visibility. All of these particles scatter light to some degree, but, of these, only
14 elemental carbon (also called light absorbing carbon) plays a significant role in light absorption.
15 Since elemental carbon, which is a product of incomplete combustion from activities such as the
16 burning of wood or diesel fuel, is a relatively small component of PM in most areas, visibility
17 impairment is generally dominated by light scattering rather than by light absorption.
18 Because humidity causes hygroscopic particles to grow in size, humidity plays a
. i •
19 significant role in particle-related visibility impairment. The amount of increase in particle size
20 with increasing relative humidity depends on particle composition. Humidity-related particle
21 growth is a more important factor in the eastern U.S., where annual average relative humidity
22 levels are 70 to 80 percent compared to 50 to 60 percent in the western U.S. Due to relative
23 humidity differences, aerosols of a given mass, dry particle size distribution, and composition
24 would likely cause greater visibility impairment in an eastern versus a western location. The
25 relationship between ambient PM and visibility impairment is discussed below in Section 2.8.
26 X . '
27 2.2.6 Radiative Properties of Particles
28 Ambient particles scatter and absorb electromagnetic radiation across the full spectrum,
29 including ultraviolet, visible, and thermal infrared wavelengths, affecting climate processes and
30 the amount of ultraviolet radiation that reaches the earth. As discussed in section 4.5 of the CD,
31 the effects of ambient particles on the transmission of these segments of the electromagnetic
32 spectrum depend on the radiative properties of the particles, which in turn are dependent on the
January 2005 2-11 Draft - Do Not Quote or Cite
-------
TABLE 2-2. COMPARISON OF AMBIENT PARTICLES,
FINE PARTICLES (Ultiafine plus Accumulation-Mode) AND COARSE PARTICLES
Fine
Ultrafine
Accumulation
Coarse
Formation
Processes:
Penned by:
Composed
of:
Solubility:
Sources:
Combustion, high-temperature
processes, and atmospheric reactions
Nucleation
Condensation
Coagulation
Sulfate
Elemental carbon
Metal compounds
Organic compounds
with very low
saturation vapor
pressure at ambient
temperature
Probably less soluble
than accumulation
mode
Combustion
Atmospheric
transformation of
SO2 and some
organic compounds
High temperature
processes
Atmospheric Minutes to hours
half-life:
Condensation
Coagulation
Reactions of gases in or
on particles
Evaporation of fog and cloud
droplets in which gases have
dissolved and reacted
Sulfate, nitrate, ammonium,
and hydrogen ions
Elemental carbon
Large variety of organic
compounds
Metals: compounds of Pb, Cd,
V, Ni, Cu, Zn, Mn, Fe, 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 processes,
smelters, steel mills, etc.
Days to weeks
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
Suspended soil or street dust
Fly ash from uncontrolled combustion
of coal, oil, and wood
Nitrates/chlorides/sulfates from
HNO3/HC1/SO2 reactions with
coarse particles
Oxides of crustal elements
(Si,Al,Ti,Fe)
CaCOj, CaSO4, NaCl, sea salt
Pollen, mold, fungal spores
Plant and animal fragments •
Tire, brake pad, and road wear debris
Largely insoluble and nonhygroscopic
Resuspension of industrial dust and
soil tracked onto roads and streets
Suspension from disturbed soil (e.g.,
fanning, mining, unpaved roads)
Construction and demolition
Uncontrolled coal and oil combustion
Ocean spray
Biological sources
Minutes to hours
Removal
Processes:
Travel
distance:
Grows into
accumulation mode
Diffuses to raindrops
<1 to 10s of km
Forms cloud droplets and
rains out
Dry deposition
100s to 1000s of km
Dry deposition by fallout
Scavenging by falling rain drops
< 1 to 10s of km (small size tail,
100s to 1000s in dust storms)
Source: Adapted from Wilson and Suh (1997); CD, p. 2-52.
January 2005
2-12
Draft - Do Not Quote or Cite
-------
1 size and shape of the particles, their composition, the distribution of components within
2 individual particles, and their vertical and horizontal distribution in the lower atmosphere.
3 The effects of PM on the transfer of radiation in the visible and infrared spectral regions
4 play a role in global and regional climate. Direct effects of particles on climatic processes are the
5 result of the same processes responsible for visibility degradation, namely radiative scattering and
6 absorption. However, while visibility impairment is caused by particle scattering in all directions,
7 climate effects result mainly from scattering light away from the earth and into space. This
8 reflection of solar radiation back to space decreases the transmission of visible radiation to the
9 surface and results in a decrease in the heating rate of the surface and the lower atmosphere. At
10 the same time, absorption of either incoming solar radiation or outgoing terrestrial radiation by
11 particles, primarily elemental carbon, results in an increase in the heating rate of the lower
12 atmosphere.
13 The extent to which ambient particles scatter and absorb radiation is highly dependent on
14 their composition and optical properties and on the wavelength of the radiation. For example,
15 sulfate and nitrate particles effectively scatter solar radiation, and they weakly absorb infrared,
16 but not visible, radiation. The effects of mineral dust particles are complex; depending on particle
17 size and degree of reflectivity, mineral aerosol can reflect or absorb radiation. Dark minerals
18 absorb across the solar and infrared radiation spectra leading to warming of the atmosphere.
19 Light-colored mineral particles in the appropriate size range can scatter visible radiation, reducing
20 radiation received at the earth's surface. Organic carbon particles mainly reflect radiation,
21 whereas elemental carbon particles strongly absorb radiation; however, the optical properties of
22 carbonaceous particles are modified if they become coated with water or sulfuric acid. Upon
23 being deposited onto surfaces, particles can also either absorb or reflect radiation depending in
24 part on the relative reflectivity of the particles and the surfaces on which they are deposited.
25 The transmission of solar radiation in the ultraviolet (UV) range through the earth's
26 atmosphere is affected by ozone and clouds as well as by particles. The effect of particles on
27 radiation in the ultraviolet-B (UV-B) range, which has been associated with various biological
28 effects, is of particular interest. Relative to ozone, the effects of ambient particles on the
29 transmission of UV-B radiation are more complex. The CD notes that even the sign of the effect
30 can reverse as the composition of the particle mix in an air mass changes from scattering to
31 absorbing types (e.g., from sulfate to elemental carbon), and that there is an interaction in the
January 2005 2-13 Draft - Do Not Quote or Cite
-------
1 radiative effects of scattering particles and absorbing molecules, such as ozone, in the lower
2 atmosphere.
3
4 2.3 AMBIENT PM MEASUREMENT METHODS
5 The methods used to measure ambient PM are important to understanding population
6 exposure to PM, evaluating health and welfare risks, and developing and evaluating the
7 effectiveness of risk management strategies. Because PM is not a homogeneous pollutant,
8 measuring and characterizing particles suspended in the atmosphere is a significant challenge.9
9 Ambient measurements include particle mass, composition, and particle number. Most
10 instruments collect PM by drawing a controlled volume of ambient air through a size-selective
11 inlet, usually defined by the inlet's 50 percent cut point. Measurable indicators of fine particles
12 include PM25, PM, 0, British or black smoke (BS), coefficient of haze (CoH), and PM10 (in areas
13 dominated by fine particles). Measurable indicators of coarse-mode particles include PM,0.2 5,
14 PM15.2J, and PM10 (in areas dominated by coarse-mode particles).
15
16 2.3.1 Particle Mass Measurement Methods
17 Ambient PM mass can be measured directly, by gravimetric methods, or indirectly, using
18 methods that rely on the physical properties of particles. Methods can also be segregated as either
19 discrete or continuous according to whether samples require laboratory analysis or the data are
20 available in real-time. Discrete methods provide time integrated data points (typically over a 24-
21 hour period) that allow for post-sampling gravimetric analyses in the laboratory. These methods
22 are typically directly linked to the historical data sets that have been used in health studies that
23 provide the underlying basis for having a NAAQS. Continuous methods can provide time
24 resolution on the order of minutes and automated operation up to several weeks, facilitating the
25 cost-effective collection of greater amounts of data compared with discrete methods.
26 The most common direct measurement methods include filter-based methods where
27 ambient aerosols are collected for a specified period of time (e.g., 24 hours) on filters that are
28 weighed before and after collection to determine mass by difference. Examples include the FRM
29 monitors for PM2 5 and PM10. Dichotomous samplers contain a separator that splits the air stream
9 Refer to 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.htaii').
January 2005 2-14 Draft - Do Not Quote or Cite
-------
1 from a PM10 inlet into two streams so that both fine- and coarse-fraction particles can be collected
2 . on separate filters. These gravimetric methods require weighing the filters after they are
3 subjected to specific equilibrium conditions (e.g., 22° C, 35 percent RH).
4 Discrete, gravimetric methodologies have been refined over the past 20 years as PM
5 monitoring networks have evolved from sampling based on the high volume TSP and PM1)?
6 method to the PM2 5 FRM. The inclusion of such measures as size-selective inlets and separators,
7 highly specific filter media performance criteria, active flow control to account for ambient
8 changes in temperature and pressure, and highly prescriptive filter weighing criteria have reduced
9 levels of measurement uncertainty, compared with earlier methods.
10 National quality assurance data analyzed .by EPA between 1999-2001 indicate that the
11 PM2 5 FRM has been a robust indicator of ambient levels by meeting the data quality objectives
12 (DQO) established at the beginning of the monitoring program. Three-year average estimates
13 from reporting organizations aggregated on a national basis for collocated sampler precision (7.2
14 percent), flow rate accuracy (0.18 percent), and method bias (-2.06 percent, from the Performance
15 Evaluation Program)10 are well within their respective goals of ±10 percent^ +4 percent, and +10
16 percent.
17 There are a number of continuous PM measurement techniques. A commonly used -
18 method is the Tapered Element Oscillating Microbalance (TEOM®) sensor, consisting of a
19 replaceable filter mounted on the narrow end of a hollow tapered quartz tube. The air flow passes
20 through the filter, and the aerosol mass collected on the filter causes the characteristic oscillation
21 frequency of the tapered tube to change in direct relation to particle mass. This approach allows
22 . mass measurements to be recorded on a near-continuous basis (i.e., every few minutes).
23 The next generation of the TEOM® is the Filter Dynamics Measurement System
24 (FDMS® monitor). This method is based upon the differential TEOM that is described in the CD
25 (CD, p. 2-78). The FDMS method employs an equilibration system integrated with a TEOM®
26 having alternating measurements of ambient air and filtered air. This self-referencing approach
27 allows the method to determine the amount of volatile PM that is evaporating from the TEOM
28 sensor for 6 of every 12 minutes of operation. An hourly measurement of the total aerosol mass
10 The Performance Evaluation Program (PEP) is designed to determine total bias for the PM2 5 sample
collection and laboratory analysis processes. Federally referenced audit samplers are coEocated adjacent to a
monitoring site's routine sampler and run for a 24-hour period. The concentrations are then determined
independently by EPA laboratories and compared in order to assess bias. The performance evaluations are
conducted four times per year (once per quarter) at one-fourth (25 percent) of the sampling sites in a reporting
organization.
January 2005' 2-15 Draft - Do Not Quote or Cite
-------
1 concentration, including non-volatile and volatile PM, is calculated and reported every 6 minutes.
2
3 Other methods that produce near-continuous PM mass measurements include the beta
4 attenuation sampler and the Continuous Ambient Mass Monitor (CAMM). A beta attenuation (or
5 beta gauge) sampler determines the mass of particles deposited on a filter by measuring the
6 absorption of electrons generated by a radioactive isotope, where the absorption is closely related
7 to the mass of the particles. The CAMM measures the pressure drop increase that occurs in
8 relation to particle loading on a membrane filter. Both methods (beta-attenuation and CAMM)
9 require calibration against standard mass measurements as neither measures PM mass directly by
10 gravimetric analysis.
11
12 2.3.2 Particle Indirect Optical Methods
13 PM has also been characterized in the U.S. and elsewhere by indirect optical methods that
14 rely on the light scattering or absorbing properties of either suspended PM or PM collected on a
15 filter.11 These include BS, CoH, and estimates derived from visibility measurements. In locations
16 where they are calibrated to standard mass units, these indirect measurements can be useful
17 surrogates for particle mass. The BS method typically involves collecting samples from a 4.5 um
18 inlet onto white filter paper where blackness of the stain is measured by light absorption. Smoke
19 particles composed primarily of elemental carbon (EC), including black carbon (BC), typically
20 make the largest contribution to stain darkness. CoH is determined using a light transmittance
21 method. This involves collecting samples from a 5.0 ^m inlet onto filter tape where the opacity
22 of the resulting stain is determined. This technique is somewhat more responsive to non-carbon
23 particles than the BS method. Nephelometers measure the light scattered by ambient aerosols in
24 order to calculate light extinction. This method results in measurements that can correlate well
25 with the mass of fine particles below 2 urn diameter. Since the mix of ambient particles varies
26 widely by location and time of year, the correlation between BS, COH, and nephelometer
27 measurements and PM mass is highly site- and time-specific. The optical methods described
28 here, as well as the particle counters described below, are based on the measurement of properties
11 See Section 2.2.5 of this chapter for a discussion of the optical properties of PM.
January 2005
2-16
Draft - Do Not Quote or Cite
-------
1 such as light scattering and electric mobility, which are inherently different than previous
2 methods described based on aerodynamic diameter.
3
4 2.3.3 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 arid 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 um 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 to
11 2 um range. An aerodynamic particle sizer measures particles in the 0.7 to 10 um range. These
12 techniques, while widely used in aerosol research, have not yet been widely used in health effects
13 studies.
14 '••'••' •
15 2.3.4 Chemical Composition Measurement Methods
16 There are a variety of methods used to identify and describe the characteristic components
17 of ambient PM.i2 X-ray fluorescence (XRF) is a commonly used laboratory technique for
18 analyzing the elemental composition of primary particles deposited on filters. Wet chemical
19 analysis methods, such as ion chromatography (1C) and automated coloiimetry (AC) are used to
20 measure ions such as nitrate (NO3"), sulfate (S04"), chloride (Cl~), ammonium (NH+), sodium
21 (Na+), organic cations (such as acetate), and phosphate (P043~).
22 There are several methods for separating organic carbon (OC) and elemental carbon (EC)
23 or black carbon (BC) in ambient filter samples. Thermal optical reflectance (TOR), thermal
24 manganese oxidation (TMO), and thermal optical transmittance (TOT) have been commonly
25 applied in aerosol studies in the United States. The thermal optical transmission (TOT) method,
26 used in the EPA speciation program, uses a different temperature profile than TOR, which is used
27 in the Interagency Monitoring of Protected Visual Environments (IMPROVE) visibility
12 The reader is referred to Chapter 2, section 2.2, of the CD for a more thorough discussion of sampling
and analytical techniques for measuring PM. Methods used in EPA's National PM Speciation Trends Network and
other special monitoring programs are summarized in Solomon et al. (2001).
January 2005 2-17 Draft - Do Not Quote or Cite
-------
1 monitoring program. The two methods yield comparable estimates of total carbon, but give a
2 different split between OC and EC.
3 Commercial instruments are now available to measure carbon (OC, EC, TC); nitrate; and
4 sulfate on a near-continuous basis. These instruments provide time-resolved measurements from
5 a few minutes to a few hours. The semi-continuous methods involved a variety of techniques that
6 include thermal reduction; wet impaction and flash vaporization; and thermal oxidation with
7 non-dispersive infrared (NDIR) detection. They have been field tested and compared through the
8 EPA's Environmental Technology Verification (ETV) program and the Supersites program and
9 proven to be good candidates for additional testing (EPA, 2004a). Data are now becoming
10 available from regional planning and multi-state organizations and the EPA to understand the
11 comparison with filter-based methods and the inherent limitations of these technologies.
12 The. U.S. EPA is coordinating a pilot study of semi-continuous speciation monitors at five
13 Speciation Trends Network (STN) sites. The pilot study began in 2002. The goals of the pilot
14 study are to assess the operational characteristics and performance of continuous carbon, nitrate,
15 and sulfate monitors for routine application at STN sites; work with the pilot participants and the
16 vendors to improve the measurement technologies used; and evaluate the use of an automated
17 data collection and processing system for real time display and reporting. After the pilot
18 monitoring and data evaluation phase, proven semi-continuous monitors will become the
19 framework for a long-term network of up to 12 STN sites equipped with semi-continuous sulfate,
20 nitrate, and carbon monitors.
21
22 2.3.5 Measurement Issues
23 There is no perfect PM sampler under all conditions; so there are uncertainties between ,
24 the mass and composition collected and measured by a sampler and the mass and composition of
25 material that exists as suspended PM in ambient air (Fehsenfeld et al., 2003). To date, few
26 standard reference materials exist to estimate the accuracy of measured PM mass and chemical
27 composition relative to what is found in air. At best, uncertainty is estimated based on collocated
28 precision and comparability or equivalency to other similar methods, which themselves have
29 unknown uncertainty, or to the FRM, which is defined for regulatory purposes but is not a
30 standard in the classical sense. There are a number of measurement-related issues that can result
January 2005 2-18 Draft - Do Not Quote or Cite
-------
1 in positive or negative measurement artifacts which could affect the associations epidemiological
2 researchers find between ambient particles and health effects.
3 The semivolatile components of PM can create both positive and negative measurement
4 artifacts. Negative artifacts arise from evaporation of the semivolatile components of PM during
5 or after collection, which is caused by. changes in temperature, relative humidity, or aerosol
6 composition, or due to the pressure drop as collected air moves across the filter. Nitrate losses
7 due to evaporation may represent as much as 10-20 percent of total PM^mass, as shown in
8 southern California studies (CD, p. 2-68). Positive artifacts arise when gas-phase compounds
-' 9 absorb onto or react with filter media or already collected PM, or when particle-bound water is
10 not removed. The chemical interaction of gases being collected with particles already on the
11 filter and conversion of PM components to gas-phase chemicals can also result in negative
12 artifacts. These interactions depend on the compounds contained in collected particles and in the
13 gas phase, and also depend on both location and time.
14 Particle-bound water can represent a significant fraction of ambient PM mass under
15 conditions where relative humidity is more than 60 percent (CD; p. 2-63, p. 2-109). It can also
16 represent a substantial fraction of gravimetric mass at normal equilibrium conditions (i.e., 22° C,
17 35 percent RH) when the aerosol has high sulfate content. * The amount of particle-bound water
18 will vary with the composition of particles, as discussed'in section 2.2.3. The use of heated inlets
19 to remove particle-bound water (e.g. TEOM at'50° C) can result in loss of semi-volatile
20 compounds unless corrective techniques are applied, although the newer generation TEOM's
21 incorporates less reliance on heat for water management (CD, p. 2-100, Table 2-7).
22 Particle bounce from the impaction plate can result in negative artifacts. This may be
23 more prevalent under lower relative humidity conditions: Impactor coatings can be used to limit
24 particle bounce, but can interfere with mass and chemical composition measurements.
25 In areas with significant amounts of dust, high wind conditions resulting in blowing dust
26 can interfere with accurate separation of fine- and coarse-fraction particles. In these unique
27 conditions a significant amount of coarse-fraction material can be found in the inter-modal region
28 between 1 and 3 um, thus overstating the mass of fine-fraction particles. The addition of a PM, 0
29 measurement in these circumstances can provide greater insights into the magnitude of this
30 problem (CD, p. 9-12).
31
January 2005 2-19 Draft - Do Not Quote or Cite
-------
1 2.4 PM CONCENTRATIONS, TRENDS, AND SPATIAL PATTERNS
2 This section provides analysis of the latest available PM air quality data, including PM
3 levels, composition, and spatial patterns. The EPA and the States have been using a national
4 network to measure and collect PM10 concentrations since 1987, and PM25 concentrations since
5 1999. Summaries through the end of 2003,,based on data publically available from EPA's Air
6 Quality System (AQS) as of August 2004, are presented here. PM25 data from the IMPROVE
7 network are also presented. Many data summaries are presented by region, as shown in Figure 2-
8 3. These regions are the same as those defined in the CD and have proven useful for
9 understanding potential differences in the characteristics of PM in different parts of the U.S.
10 As is the case with all surface-based ambient monitoring data, these data can be considered
11 representative of exposures in typical breathing zones in the lowest 15 meters of the atmosphere.
12 .
13 2.4.1 PM25
14 Following the establishment of new standards for PM2 5 in 1997, the EPA led a national
15 effort to deploy and operate over 1000 PM25 monitors. Over 90 percent of the monitors are
16 located in urban areas. These monitors use the PM2S FRM which, when its procedures are
17 followed, assures that PM data are collected using standard equipment, operating procedures, and
18 filter handling techniques.13 Most of these FRM monitors began operation in 1999. TheEPAhas
19 analyzed the available data collected by this network from 2001-2003. Data from the monitors
20 were screened for completeness with the purpose of avoiding seasonal bias. To be included in
21 these analyses, amonitoring site needed all 12 quarters (2001-2003), each with 11 or more
22 observations.. A total of 827 FRM sites in the U. S. met these criteria."
23 The 3-year average annual PM25 mean concentrations range from about 4 to 28 ng/m3,
24 with a median of about 13 ug/m3. The 3-year average annual 98th percentiles of the 24-hour
25 average concentrations range from about 9 to 76 ug/m3, with a median of about 32 ug/m3.
26 Figures 2-4 and 2-5 depict the regional distribution of site-specific 3-year average annual mean
27 and 3-year average 98th percentile 24-hour average PM25 (and PM]0_25, discussed in section 2.4.3)
See 40 CFR Parts 50 and 58 for monitoring program requirements.
810 of the 827 monitors are located in the contiguous continental U.S. covered by the regions shown in
Figure 2-3. The remainder are located in Alaska, Hawaii, and U.S. territories.
January 2005 2-20 Draft - Do Not Quote or Cite
-------
=
s
0
VI
at
I
ft
a
93
03
PN
56
O
*a.
I
I
a
-------
1 concentrations, respectively, by geographic region (excluding Alaska, Hawaii, Puerto Rico, and
2 the Virgin Islands). In general, with the exception of southern California, PM25 annual average
3 mass is greater in the eastern regions than in the western regions, whereas PM10.2.5 annual average
4 mass is greater in the western regions. Figures 2-6 and 2-7 are national maps that depict county-
5 level 3-year average annual mean and 3-year average annual 98fe percentile 24-hour average
6 PM2 5 concentrations, respectively, from the FRM network.15 The site with the highest
7 concentration in each monitored county is used to represent the value in that county. The map
8 and box plots show that many locations in the eastern U.S. and in California had annual mean
9 PM25 concentrations above 15 ug/m3. Mean PM25 concentrations were above 18 ug/m3 in several
10 urban areas throughout the eastern U.S., including Chicago, Cleveland, Detroit, Indianapolis,
11 Pittsburgh, and St. Louis. Los Angeles and the central valley of California also were above 18
12 ug/m3. Sites in the upper midwest, southwest, and northwest regions had generally low 3-year
13 average annual mean PM25 concentrations, most below 12 \ig/m3. Three-year average annual 98th
14 percentile 24-hour average PM25 concentrations above 65 ug/m3 appear only in California.
15 Values in the 40 to 65 ug/m3 range were more common in the eastern U.S. and on the west coast,
16 mostly, in or near urban areas, but relatively rare in the upper midwest and southwest regions. In
17 these regions, the 3-year average 98th percentile PM25 concentrations were more typically below
18 40 ug/m3, with many below 25 ug/m3.
19 The PM maps shown in this chapter encompass all valid data, including days that were
20 flagged for episodic events, either natural or anthropogenic. Examples of such events include
21 biomass burning, meteorological inversions, dust storms, and volcanic and seismic activity. PM
22 concentrations can increase dramatically with these 'natural' or 'exceptional' events. Although
23 these events are rare (e.g., affecting less than 1 percent of reported PM25 concentrations between
24 ' 2001 and 2003), they can affect people's short-term PM exposure, briefly pushing daily PM
25 levels into the unhealthy ranges of the Air Quality Index (AQI). An analyses of 2001-2003 PM25
26 data found that over 9 percent of the days above (site-based) 98th percentile 24-hour
Readers are 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. Although 11 samples per quarter, as required in these analyses, is sufficient to show nonattainment,
additional data capture (at least 75 percent per quarter) is required to show attainment of the standards. Not all of the
PM federal reference method (FRM) sites that contributed data to the summaries presented here recorded 75 percent
data capture for all four calendar quarters for each of the 3 years.
January 2005 2-22 Draft - Do Not Quote or Cite
-------
v>
8
o
BS
g
a -5
s s
d 2
iti
o s Ja
*^o ^* *^
9
f
Q
uea|/\i
a c^
_H O 'C
•C .M »
.a MU
o 2 I
^
5
s
<->W-)
-------
•;::::<>:; :;;:O:M:::p;:::::
-------
•n
Xoo
r~><
II
oo
X^
o
o
t/5
I
S3
a
OJ
'**"'' to
c .2
.2 i
I I
8C C4
c S
o
O
in
2N
0.
a
a
«8
U
o
o
(§)
t -
z t
U ' "
^o
NO' ,'P
en
«e
-------
o
Tf
V
X«r>
Vs0
A
O
m x!
I
Ml
to
c .2
•5 c
CO ^
i- O
c «
8CJ
s
c •"
o
O
in
8
o
-------
1 concentrations were flagged for events. The events, in fact, were found to cause the 98th
2 percentiles to inflate by up to 18 ug/m3, with an average increase of 0.8 ug/m3. Natural and
3 exceptional events, however, rarely have a significant effect on annual or longer averages of PM.
4 In the afore-mentioned analyses of 2001-2003 PM2.5 data, the average effect of natural and
5 exceptional events on 3-year annual means was less than 0.1 ug/m3 (Schmidt, et al., 2005).
6 Episodic event-flagged data are often excluded from trends-type analyses and are addressed for
7 the purpose of determining compliance with the NAAQS by EPA's national and exceptional
8 events policies, as described below in section 2.6.
9 PM2s short-term trends were recently evaluated by EPA in The Particle Pollution Report
10 (EPA, 2004, p. 14). In the EPA FRM network, PM2 5 annual average concentrations decreased
11 10 percent nationally from 1999.to 2003. The Northeast, where moderate concentrations are
12 found, was the only region that did not show a decline between these years; annual concentrations
13 in that region rose about 1 percent over the 5-year period. Except in the Northeast, PM2 5
14 generally decreased the most in the regions with the highest concentrations - the Southeast (20
15 percent), southern California (16 percent), and the Industrial Midwest (9 percent) from 1999 to
16 2003. The remaining regions with lower concentrations (the Upper Midwest, 1he Southwest, and
17 the Northwest) posted modest declines in PM2.5; see Figure 2-8 (EPA, 2004, p. 15).
18 The IMPROVE monitoring network, which consists of sites located primarily in national
19 parks and wilderness areas throughout the U.S., provides data for long-term PM25 trends for
20 generally rural areas.16 Figure 2-9 shows the composite long-term trend at 8 eastern sites, 17
21 western sites, and one urban site in Washington, D.C. The 4 westmost U.S. subregions
22 (Northwest, southern California, Upper Midwest, and Southwest) are considered the 'west' and
23 the 3 eastern ones (Northeast, Southeast, and Industrial Midwest) are termed the 'east' At the
24 rural eastern sites, measured PM25 mass decreased about 23 percent from 1993 to 2003. At the
25 rural western sites PM25 mass decreased about 21 percent from 1993 to 2003. At the
26 Washington, D.C., site the annual average PM2 5 concentration in 2003 was about 31 percent
27 lower than the value in 1993.
28 The relative spatial homogeneity of the ambient air across a specified area can be assessed
29 by examining the values at multiple sites using several indicators, including: (1) site pair
16IMPROVE monitoring instruments and protocols (defined at httpr/Msta.cira.colostate.edu/improve/) are
not identical to FRM monitors.
January 2005 2-27 Draft - Do Not Quote or Cite
-------
o
o
u
in
t*
CL
(II
O»
2
I
§
c
«**,
4*3V
'''/4*4",'Z/^», , -> t'tf'"1
%&$&%&*$&*'&#
?3$%./4 ;*«^r<#.^
^4/4^4^'
^ /^ Vjl7'.^ ' *3*i^SSs&i
3t
I
^
lyft-
vo
f~
•*
ll
c c
2 ,2
S g*
o
o
-------
12
10
25
20
15
10
WESTERN UNITED STATES
EASTERN UNITED STATES
T3 B ET
WASHINGTON, DC
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
-»- PMz.5 Mass -B- Amm. Sulfate "*~ TCM -*" Crustal "**• Amm. Nitrate
Figure 2-9. Average annual average trend in PM2§ mass, ammonium
sulfate, ammonium nitrate, total carbonaceous mass, and
crustal material at IMPROVE sites, 1993-2003.
Source: Schmidt et al. (2005)
January 2005
2-29
Draft - Do Not Quote or Cite
-------
1 correlations, (2) differences in long-term (e.g., annual) average concentrations, and (3)
2 differences in short-term (e.g., daily) average concentrations. An analysis of these indicators for
3 site pairs in 27 Metropolitan Statistical Areas (MSAs) using PM2 5 FRM monitoring data from
4 1999-2001 is included in the CD (CD, Appendix 3 A).
5 An analysis of site pairs from each of the 27 urban areas indicates that multiple sites in
6 these areas were highly correlated throughout the period. More than 86 percent (426 out of 491)
7 of the between-site correlation coefficients in all 27 areas were greater than or equal to 0.80, and
8 more than 53 percent (268 out of 491) of the correlations were greater than or equal to 0.90.
9 Further, every area had at least one monitor pair with a correlation coefficient greater than or
10 equal to 0.85 (CD, Appendix 3 A). A larger, more recent (2001 -2003) PM2 5 FRM database was
11 similarly analyzed; the median between-site correlation for more than 2,000 site pairs across the
12 nation was about 0.9 (Schmidt, et al., 2005).
13 A summary of the analyses of long-term and short-term concentration differences for the
14 27 urban areas is shown in Table 2-3. The difference in annual mean PM25 concentrations
15 between monitor pairs in the 27 cities ranged from less than 1 ug/m3 in Baton Rouge to about 8
16 ng/m3 in Pittsburgh. Large differences in annual mean concentrations across a metropolitan area
17 may be due to differences in emissions sources, meteorology, or topography. Small differences
18 may be due only to measurement imprecision (CD, p. 3-46). In urban areas, the site pair with the
19 maximum and minimum annual mean concentration was highly correlated (r(max>a,in) > 0.70); the
20 most notable exception was the site pair in Gary, IN (r(niaXiniin)=0.56).
21 The analysis in the CD also examined differences in 24-hour average concentrations
22 between the urban site pairs. Small differences throughout the distribution would indicate
23 relatively homogeneous concentration levels between the sites. Table 2-3 presents a summary of
24 the 90th percentile of the distribution (P90) of daily site pair differences in each urban area. The
25 site pairs with the largest difference (max pair) and the smallest difference (min pair) are shown.
26 The P90 values for the 491 monitor pairs in the 27 urban areas ranged from about 2 to 21 ug/rn3.
27 Often the site pair with the maximum P90 value in each city was also the pair with the largest
28 annual mean difference. The site pair with the highest P90 values in each city was generally
29 highly correlated (rfflaxi 0.70), and in some cases was more highly correlated than the sites with the
30 largest annual mean differences.
31
January 2005 2-30 Draft - Do Not Quote or Cite
-------
Table 2-3. Summary
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
i -.
Norfolk, V A
Columbia,' SC
Baton Rouge, LA
N
Sites
.11
6
10
' 8
ii
' 4"
11
.. 4 *
5
6
4
5*
6
5
7
4
6*
4"
4
4
•- 5-r
7
4.
8
5
3
3
•>•• ' 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
MinSite
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%
' 's%
''6%
3%
IVn,,Ymln> '
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
P,0 (ng/m3)
Max Pair
. 21.0
11.4
13.8
143
15.2
'6.5
11.3 ..
8.5
15.2
18.2
11.3'
7.7
6.5
17.8
6.3
. 8.8 v
10.8
6.1
• ii.o
'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
. «•„„
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 = 90* percentile of the distribution of differences in 24-hour averages between two sites in the same urban area.
r(nuxjnin) = correlation between intra-urban sites with the largest difference iri annual mean concentrations.
r{m!K) = correlation between intra-urban sites with the largest difference in P90 values.
Source: CD, Appendix 3A
January 2005
2-31
Draft - Do Not Quote or Cite
-------
-------
1 Since the protocol for each monitor is not usually identical, the consistency of these PM10.2 5
2 measurements is relatively uncertain, and they are referred to as "estimates" in this Staff Paper.19
3 The 98th percentile 24-hour average PM10.25 concentrations range from about 5 to 208
4 ug/m3, with a median of about 28 ug/m3. The box plots in Figures 2-4 and 2-5 (introduced in
5 section 2.4.1) depict the regional distribution of site-specific estimated annual mean and 98th
6 percentile 24-hour average PM10_2 5 concentrations, respectively, by geographic region (excluding
7 Alaska, Hawaii, Puerto Rico, and the Virgin Islands). Figures 2-12 and 2-13 are national maps
8 that depict estimated county-level annual mean PMi0,2 5 concentrations and 98th percentile 24-hour
9 average concentrations, respectively. To construct the maps; the site with the highest
10 concentration in each monitored county is used to represent the value in that county. The annual
11 mean PM10.2J concentrations are generally estimated to be below 40 ug/m3, with one maximum
12 value as high as 64 ug/m3 (see Figure 2-4), and with a median of about 10-11 ug/m3. Compared
13 to annual mean PM2 5 concentrations, annual mean PM10.2 5 estimates are more variable, with more
14 distinct regional differences. As shown in Figure 2-4, eastern U.S. estimated annual mean PM10.
15 25 levels tend to be lower than annual mean PM25 levels, and in the western U.S. estimated PM,0.
16 25 levels tend to be higher than PM2 5 levels. The highest estimated annual mean PM10.2 5
17 concentrations appear in the southwest region and southern California. The estimated 98th
18 percentile 24-hour average PM10.2 5 concentrations are generally highest in the.southwest,
19 southern California, and upper midwest, where a few sites have estimated concentrations well
20 above 100 ug/m3 (see Figure 2-5). As noted before, these maps include days mat were flagged
21 for natural or exceptional episodic events. Episodic events can affect PM10.25 98th percentiles
22 even more than for PM25. An evaluation of 2001-2003 PM10.25 data found that events caused 98th
23 percentiles to be elevated by an average of 2.5 ug/m3 (Schmidt, et al., 2005).
24 The IMPROVE monitoring network provides long-term PM]0.2 s trends for generally rural
25 areas. Figure 2-14 presents the composite long-term trend at 7 eastern sites, 17 western sites, and
I
19Note that the urban PM10.25 estimates derived in this review, labeled '2001-2003', actually represent
either the entire 12-quarter period or the most recent consecutive 4- or 8-quarter period (from that 3-year period)
with 11 or more samples each. This technique was used to maximize the number of usable sites (arid not introduce
seasonal bias). Of the 489 total sites, 230 had 12 complete quarters, 122 sites had 8 quarters, and 137 had 4. Similar
to Plv^ 5 and PM10 processing, 'annual' means and 'annual' 98th percentiles were first constructed from 4-quarter
periods, albeit for PM10.2 5, not all necessarily from the same calender year. The 4-quarter statistics were then
averaged together for the 8- and 12-quarter sites. Hence there is some temporal variability intrinsic in 2001-2003
estimates. The 1-, 2-, or 3-year averages of the 'annual' statistics are subsequently referred to simply as 'annual
means' or' 98th percentiles'.
January 2005 2-35 Draft - Do Not Quote or Cite
-------
\
o
o
I
I
•O i O
~- A Q
V
-------
t
o
s
a
4)
>n u
* 8
11 °
* i
X»n 3
t^ ^
V &<
A «
m M
O
%
o
O)
5
^t
o
Q
1
-------
\
\
\
\
o
o
0 I
w 1
i
o>
o> t
o> '
CO
O>
CO
I
r*« t
o>
o> ®
T"
in
o> S
0> E
T- ^
II)
a
10 !"
O)
O)
o>
o>
ro
o>
0>
O>
CO
(O
ELU/6rl
I
C8
*O
J
O
«
I
g
I
W>
h
es
•a
§
|
ft fi
« o
HB
S J
i
a
00
cl
fl
8
*s
v—'
•a
•4-f
U
-*-)
•a
5
DO
<3
b
I
-------
1 one urban site in Washington, D.C. At the rural eastern sites, measured PM10.2 5 in 2003 was
2 about 33 percent lower then the corresponding value in 1993. At the rural western sites, *
3 measured PM10_25 was about 17 percent higher in 2003 than the corresponding value in 1993. At
4 the Washington, D.C., site, the annual average PM10.2 5 concentration in 2003 was about 25
5 percent lower than the 10-year peak in 1994, but nearly 2 ug/m3 higher than the 1998 low point.
6 The CD contains an analysis of 1999-2001 PM10.25 estimates in 17 MS As that is useful for
7 assessing the spatial homogeneity of PM10_2 5 across the urban areas (CD, Appendix 3 A). This
8 analysis is similar to the 27-city analysis for PM25 discussed in section 2.4.1 and summarized
9 earlier in Table 2-4. However, since there were fewer site pairings, fewer urban areas covered,
i-
10 and because of higher uncertainty in daily concentration estimates, the PM10.2 5 results are not as
11 robust as the PM25 results. The PM10.25 analysis is summarized in Table 2-4. The analysis •
12 reveals generally lower correlations for PM,o.2 5 compared to the PM2S correlations in the same
13 city. Of the 65 monitor pairs analyzed, only 4 had correlation coefficients greater than or equal to
14 0.80, in contrast to more than 86 percent (426 of 491) of the pairs for PM2 s.
15 The difference in estimated annual mean PM10.2 5 between site pairs in the 17 cities also
16 covered a greater range than was seen for PM2 5, with differences up to about 21 ug/m3 in
17 Riverside, CA. Similarly, the ?«, values (described in section 2.4.1) for the 65 site pairs ranged
18 from about 5 ug/m3 to about 43 ug/m3, which is wider than the range of about 2 |ig/m3 to 21
19 ug/m3 observed for*PM2 5.
20 These analyses indicate that PM10.25 is more heterogeneous than PM25 in many locations
21 (e.g., Cleveland, Detroit, Steubenville) and may be similar in other locations (e.g., Portland,
22 Tampa, St. Louis). Any conclusions should be tempered by the inherent uncertainty in the PM10.
23 25 estimation method (discussed at the beginning of this section), and the relatively small sample
24 size for PM10.2 5 relative to PM2 5. ' . '
25 ..
26 2.4.4 Ultrafine Particles
27 . There are no nationwide monitoring networks for ultrafine particles (i.e., those with
28 diameters < 0.1 um), and only a few recently published studies of ultrafine particle counts in the
29 U.S. At an urban site in Atlanta, GA, particles in three size classes were measured on a
30 continuous basis between August 1998 and August 1999 (CD, p. 2B-21). The classes included
31 ultrafine particles in two size ranges, 0.003 to 0.01 um and 0.01 to 0.1 um, and a subset of
January.200S 2-39 Draft - Do Not Quote or Cite
-------
Table 2-4. Summary of Estimated PMin.2^ Analysis in 17 Metropolitan Areas, 1999-2001.
City
Cleveland, OH
Detroit, M
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
3
3
4
4
4
2
4
4
3
2
3
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
Mia 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%
ft mar mini
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
PTOOig/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»«
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 = 90* pereentile of the distribution of differences in 24-hour averages between two sites in the same urban area.
r(msn-min) = correlation between ultra-urban sites with the largest difference in annual mean concentrations.
r
-------
1 accumulation-mode particles in the range of 0.1 to 2 um. in Atlanta, the vast majority (89
2 percent) of the number of particles were in the ultrafine mode (smaller than 0.1 um), but 83
3 percent of the particle volume was in the subset of accumulation-mode particles. The researchers
4 . found that for particles with diameters up to 2 um, there was little evidence of any correlation
5 between number concentration and either volume or surface area. Similarly poor correlations
6' between PM2 5 mass and number of ultrafine particles were confirmed for sites in Los Angeles
7 and nearby Riverside, CA (Kim et al, 2002). This suggests that PM25 cannot be used as a
8 surrogate for ultrafine mass or number, so ultrafine particles need to be measured independently.
9 Studies of near-roadway particle number and size distributions have shown sharp
10 gradients in ultrafine concentrations around Los Angeles roadways (CD, p. 2-35 to 2-36).
11 Ultrafine PM concentrations were found to decrease exponentially with distance from the
12 roadway source, and were equal to the upwind "background" location at 300 m downwind.
13
14 2.4.5 Components of PM
15 Atmospheric PM is comprised of many different chemical components that vary by
16 location, time of day, and time of year. Further, as discussed in section 2.2, fine and coarse
17 particles have fundamentally different sources and composition. • Recent data from the rural
18 IMPROVE network and from the EPA urban speciation network provide indications of regional
19 composition differences for fine particles. Although both programs provide detailed estimates of
20 specific PM chemical components (individual metals, ions, etc.), only gross-level speciation
21 breakouts are shown here. Figure 2-15 shows urban and rural 2003 annual average PM25 mass
22 apportionment among chemical components averaged over several sites within each of the U.S.
23 regions. In general:
24 • PM2 5 mass is higher in urban areas than in rural areas.
25
26 • PM2 5 in the eastern U.S. regions is dominated by ammonium sulfate and carbon.
27
28 • PM2 5 in the western U.S. regions has a greater proportion of carbon.
29
30 • Ammonium nitrate is more prevalent in urban aerosols than in rural aerosols, especially in
31 the midwest regions and in southern California.
32 '••'.'
33 Though most of the speciation data available are from PM2 s, there is a limited amount of
34 data available on speciation profiles for other size fractions as well. One such data source is the
January 2005 2-41 Draft - Do Not Quote or Cite
-------
C
o
to
8
8
A
<0
€
o
(0
o
(O
re
s
in
•a
o.
3
o
V)
t:
o
i
"a
O
<5
i O
Figure 2-15. Annual average composition of PM25 by region, 2003. Rural data
(top panel) from IMPROVE network, urban data (bottom panel) from
EPA Speciation Network. Components (from top to bottom) are
crustal material, total carbonaceous mass (TCM), ammonium
nitrate, and ammonium sulfate.
Source: Schmidt etal. (2005)
January 2005
2-42
Draft - Do Not Quote or Cite
-------
1 PM Supersites program, established by EPA. This monitoring program addresses a number of
2 scientific issues associated with PM.20 At a Supersite location in the Los Angeles metropolitan
3 area, speciation data have been collected for fine, coarse, and ultrafme. Speciated data from this
4 source-influenced site are shown in Figure 2-16. These data show that fine, coarse, and ultrafine
5 PM have different compositions (in the Los Angeles area). For these PM size fractions, there are
6 differences in the relative amounts of nitrates, sulfates, crustal (metals and trace elements), and
7 carboa Carbon, shown here as organic (OC) and elemental carbon (EC), makes up a large
8 fraction of ultrafine and fine PM; crustal material dominates the coarse traction.
9 Trends in rural area and urban Washington, D.C., concentrations of fine particle
10 components based on data from the IMPROVE network from 1993 to 2003 are shown in Figure
11 2-9 (introduced above in section 2.4.1 on PM2 5). The top two panels of this figure aggregate
12 rural IMPROVE sites in the eastern and western U.S. The bottom panel shows the urban
13 IMPROVE data for Washington, D.C., for the same time period. Levels of rural annual average
14 PM2 5 mass are significantly higher in the east than in the west Annual levels of ammonium
15 sulfates have decreased the most (and contributed the most to the reductions in PM25 mass) both
16 in eastern and western rural areas. At the Washington, D.C., IMPROVE site, mass has decreased
17 31 percent from 1993-2003. Total carbon (34 percent reduction) and ammonium sulfates (down
18 29 percent) are the biggest contributors to the mass reduction over the past 10 years. In addition,
19 at the Washington, B.C., site, bom total carbon and sulfates dropped significantly in 1995, but
20 have not shown significant improvements since then. All other components in all areas have
21 shown small changes over the 10-year period.
22
23 2.4.6 Relationships Among PM2 s, PM10, and PM10.2 5
24 In this section, information on the relationships among PM indicators in different regions
25 is presented based on data from the nationwide PM FRM monitoring networks.21 Figure 2-17
26 shows the distribution of ratios of annual mean PM25 to PM10 at sites in different geographic
27 regions for 2001-2003. The ratios are highest in the eastern U.S. regions with median ratios of
20 More information can be found at http://www.epa.gov/ttn/amtic/supersites.html.
21 In this section's analyses, information was gleaned from the 489 site (4-, 8-, 12-quarter) PM10.2.j database
for all 3 sizes in order to get seasonally unbiased estimates of their statistical relationships (i.e., to ensure a minimum
number of data pairs each quarter for 4-, 8-, or 12 quarters).
January 2005 ' 2-43 Draft - Do Not Quote or Cite
-------
PMio-2.5 ('coarse')
rO.7%
12.9%
5.8%
m Amm. Nitrate
H Amm. Sulfate
• Crustal
H3 oc
• EC
59.6%
PM2 5 ('fine')
4.4%
30.0%
13.0%
22.5%
Amm. Nitrate
H Amm. Sulfate
• Crustal
m oc
• EC
PM0! ('ultrafine')
13.0%
8.9%
I Amm. Nitrate
I Amm. Sulfate
I Crustal
I OC
I EC
71.2%
Figure 2-16. Average PM10_2 5 PM2 5, and PM0 j (ultrafine) chemical
composition at an EPA 'supersite' monitor in Los Angeles,
CA, 10/2001 to 9/2002. Components shown in clockwise order
(starting with ammonium nitrate) as listed in legend from top to
bottom.
2-44
Draft - Do Not Quote or Cite
-------
O)
a
-------
1 about 0.6 to 0.65, and lowest in the Southwest region, with a median ratio near 0.3. These data
2 are generally consistent with earlier findings reported in the 1996 CD from a more limited set of
3 sites. Ratios greater than one are an artifact of the uncertainty in the independent PM10 and PM2.5
4 measurement methods.
5 Correlations among pollutant indicators can provide insights into how well one indicator
6 can represent the variability in another indicator. Figure 2-18 shows the results of a nationwide
7 analysis of correlations among PM size fractions using 24-hour average data from the FRM
8 monitoring networks for 2001-2003. PM25 and PMj0 measured on the same days at collocated -
9 monitors are fairly well correlated, on average, in the eastern regions, and not as well correlated,
10 on average, in the upper midwest and southwest regions. PM10 is fairly well correlated with
11 estimated PMI0.25 in most regions, with the highest average correlation in the upper midwest and
12 southwest regions.. PM10 is more highly correlated, on average, with PM25 than with estimated
13 PM10.2 j in the northeast and industrial midwest regions. Their correlations are similar in the
14 southeast, and PM10 is more highly correlated, on average, with PM10.25 in the northwest and
15 southern California regions? These data suggest that PM10 might be a suitable indicator for either
16 fine or coarse particles, depending upon location-specific factors. However, in all locations
17 estimated PM10.2 5 and PM2 5 are very poorly correlated, which should be expected due to their
18 differences in origin, composition, and atmospheric behavior.
19 "
20 2.5 PM TEMPORAL PATTERNS
21 2.5.1 PM2.S and PM10.2 s Patterns
22 Data from the PM FRM network from 2001-2003 generally show distinct seasonal
23 variations in PM25 and estimated PM^5 concentrations. Although distinct, the seasonal
24 fluctuations are generally not as sharp as those seen for ozone concentrations. Figure 2-19 shows
25 the monthly distribution of 24-hour average urban PM2 5 concentrations in different geographic
26 regions. The months with peak urban PM2 5 concentrations vary byVegion. The urban areas in
27 the northeast, industrial midwest, and upper midwest regions all exhibit peaks in both the winter
28 and summer months. In the northeast and industrial midwest regions, the summer peak is slightly
29 more pronounced than the winter peak, and in the upper midwest region the winter peak is
30 slightly more pronounced than the summer peak. In the southeast, a single peak period in the
31' summer is evident. In western regions, peaks occur in the late fall and winter months. ^fe
January 2005 2-46 Draft - Do Not Quote or Cite
-------
p; ;
o
«
I|:|I:II:IO:I
Hi!
liii^iiiii
;;«:
s:
o
in
iiBii
iroilis
I
I
nc;
JiMi;
liEii
CS
•a
o
en
"•»
-------
1 Figure 2-20 shows the distributions of estimated 24-hour average urban PM,0.2 5
2 concentrations by U.S. geographic region. The lowest concentrations generally occur in the
3 winter months. Elevated levels are apparent in the easternmost regions in April. In the upper
4 midwest, northwest, and southern California regions, the highest levels occur in the mid- to late-
5 summer to mid-fall. The southwest region exhibits the greatest range of variability throughout
6 the year. Elevated levels are apparent in the spring, consistent with winds that contribute to
7 windblown dust. In the southwest and southern California, highly elevated levels in the fall,
8 especially October, were caused by forest fires in the vicinity of the monitoring sites.
9 The chemical components of fine particles also exhibit seasonal patterns. Figures 2-21
10 and 2-22 show seasonal 2003 urban and rural patterns for each of the U.S. regions. Seasonal
11 patterns are shown by calendar quarter. In general:
12 • PM25 values in the east are typically higher in the third calendar quarter (July-September)
13 when sulfates are more readily formed from SO2 emissions from power plants
14 predominantly located there and sulfate formation is supported by increased
15 photochemical activity.
16 . . •
17 • Urban PM2 5 values tend to be higher in the first (January-February) and fourth (October-
18 December) calendar quarters in many areas of the western U.S., in part because more
19 carbon is produced when woodstoves and fireplaces are used and paniculate nitrates are
20 more readily formed in cooler weather. In addition, the effective surface layer mixing
21 depth often is restricted due to inversion events, as well as limited by reduced radiative
22 heating.
23
24 • Urban concentrations of PMZ5 are seen to be generally higher than rural concentrations in
25 all four quarters, though in the west the difference seems to be greatest in the cooler
26 months. ' ' •
27 . , . .
28 The relationship between the annual mean at a site and the shorter-term 24-hour'average
29 peaks is useful for examining the relationships between short- and long-term air quality standards.
30 The box plots in Figures 2-23 and 2-24 show the relationships for PM25 and estimated PM10.25,
31 respectively, between annual mean PM concentrations and peak daily concentrations as
32 represented by the 98th percentile of the distribution of daily average concentrations at FRM sites
33 across the U.S. Although mere is a clear monotonic relationship between 98th percentiles and
34 annual means, there is considerable variability in peak daily values for sites with similar annual
35 means. For annual mean PM25 values between 10 and 15 ug/m3, the interquartile range of 98th
36 percentiles spans about 5 to 6 ug/m3 for each 1 ug/m3 interval.. The range between the 5th and 95th
January 2005 2-50 Draft - Do Not Quote or Cite
-------
iAmra
Atnnt,
Sulfate!
Figure 2-21. Seasonal average composition of urban PM25 by region, 2003.
Data from EPA Speciation Network. Components (from top to bottom)
are crustal material, total carbonaceous mass (TCM), ammonium nitrate,
and ammonium sulfate.
Source: Schmidt et al. (2005)
January 2005 ' 2-51 Draft - Do Not Quote or Cite
-------
Irustal
E Amm. I
ffltratei
;Amm. 1
Figure 2-22. Seasonal average composition of rural PM2 5 by region, 2003.
Data from IMPROVE Network. Components (from top to bottom)
are crustal material, total carbonaceous mass (TCM), ammonium nitrate,
and ammonium sulfate.
Source: Schmidt et al. (2005)
January 2005
2-52
Draft - Do Not Quote or Cite
-------
^
* n
i -a
g
"?
(S
(eLU/6i1) a
-------
fc
::::::: -if:: :::::::::(;: ::::::::::*::;:;;:;::;: J:::::::::::{::::::::::: i:::::::::::::.: ::::::::
(eiu/6Ti)
«
•
C8
^M
V
>
C3
t,
§
43
4
2
u
•a
2
£
* n
S
^
O
£
T3
_
c/i
+^
'55
<*H
O
fe
"c
(O
es
c
ON
*n
8
r^
*03
is
£>
5 S
-------
1 percentile values for each interval varies substantially. Estimated PMi0.25 generally exhibits
2 greater variability in 98th percentile values for sites with similar annual means than seen for PM2 5
3 The maximum estimated PM10.2 5 values are quite high relative to the rest of the distribution for
4 annual mean intervals above 20 ng/m3.
5 Monitors that provide near-continuous measurements can provide insights into short-term
6 (e.g., hourly average) patterns in PM, which could be important to understanding associations
7 between elevated PM levels and adverse health and welfare effects. Examples of average hourly
8 profiles for PM2 5 and PM10_2 5 from 2001 -2003 are shown in Figures 2-25 and 2-26 for a
9 monitoring site in the Greensboro, NC, metropolitan area. As with most eastern urban sites, the
10 PM2 5 concentrations are significantly higher than those for PM10.2 5. Profiles, for both PM2 5 and
11 PM10.2 5, in Figure 2-25 indicate that elevated hourly average levels occurred most often between
12 the hours of 6:00 am and 9:00 am, corresponding to the typical morning rush of automobile
13 traffic. An evening peak starting about 5:00pm is also evident for both size indicators! The 95th
14 percentile concentrations during peak hours can be as high as three to four times the median level
15 for the same hour. As indicated in Figure 2-26, the lowest seasonal levels for both size fractions
16 occur in the winter. For PM2 5} the summer concentrations are considerably higher than the other
17 season. These profiles of hourly average PM25 and PM10.25 levels are typical of many, but not
18 all, eastern U.S. urban areas.
19 Figure 2-27 shows hourly average PM2 5 and PM10.2 5 concentrations for a monitoring site
20 in the El Paso, TX metropolitan area from 2001-2003. Like many western U.S. sites for all hours
21 of the day, the PM10_25 concentrations are higher than the PM25 levels. However, this particular
22 site is atypical of most urban ones, even in the west. Note the increased variability in the hourly
23 concentrations compared to the Greensboro site; the 95th percentile concentrations for some hours
24 are more than ten times the median levels. Note also that hourly means are significantly higher
25 than the medians, and in some cases, the 75th percentiles.. Episodic events are causing these
/
26 significant excursions from the typical day. Figure 2-28 highlights one of several such episodic
27 events that affected this site. On April 26, 2002, there was a dust storm that caused the PM25 and
28 PM10.2 5 concentrations to be extremely elevated. The dust particles from the storm had a greater
29 impact on the PM10.2 5 concentrations than the PM^. (Note that the PM10.2 5 scale is about 3 times
30 as large as the PM2.5 scale.) Hourly PM10.2S levels approaching 3000 ug/m3 were recorded this
31 day.
January 2005 2-55 Draft - Do Not Quote or Cite
-------
4ff
30
20
Iff
tn
C
o
I
o
O
0'
40
30
20
10
PM2.5
.0 1 23456 789 10 11 12 13 14 15 16 17 16 19 20 21 22 23
PM
10-2.5
0 1 2 34 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour
Figure 2-25. Hourly average PM2 5 and PM10 2 5 concentrations at a Greensboro,
NC monitoring site, 2001-2003. Upper panel shows the distribution of
PMj 5 concentrations and the lower panel shows the distribution of PM10.2 5
concentrations. (Box plots of interquartile ranges, means, medians, 5th and
95th percentiles.)
Source: Schmidt et al. (2005)
January 2005
2-56
Draft - Do Not Quote or Cite
-------
30
20
10
o
1
I
O 3° .
o
20
10
taring (Mar-May)
rummer (Jun-Aug)
=all (Sep-Noy)
''inter (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
PM
10-2.5
0 1 23 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour
Figure 2-26. Seasonal hourly average PM2 s and PM10_2 5 concentrations at a
Greensboro, NC monitoring site, 2001-2003. Upper panel shows the
PM2 3 concentrations and the lower panel, shows the PM10_2 5
concentrations.
Source: Schmidt etal. (2005)
January 2005 2-57 Draft - Do Not Quote or Cite
-------
CO
I
c
o
I
§
c
o
O
220
2tO
200
10 -
00 -
BO-
00
70
60 -
50 -
40 •
30 -
20
10 '
PML,
.5
30 31 33 33
PM
10-2.5
Hour
Figure 2-27. Hourly average PM2 s and PM10.2 5 concentrations at an El Paso.
monitoring site, 2001-2003. Upper panel shows the distribution of PIV^ 5
concentrations and the lower panel shows the distribution of PM10_2 5
concentrations- (Box plots of interquartile ranges, means, medians, 5th
and 95th percentiles.)
Source: Schmidt et al. (2005)
January 2005 2-58 Draft - Do Not Quote or Cite •
-------
I
O
I
O
012345J7S81I1II1I
!j 5 j jjjj33?un11i 11n 131 n11«i
April 26,2002 April 27. 2002
PM
10-2.5
Q12345B7B911 1 111 Il)1222.2222222d33333333»444444
-------
1 The hourly ranges shown in Figures 2-25 and 2-27 suggest that hour-to-hour changes in
2 PM2.5 concentrations encompass several ug/m3; however, extreme values for hour-to-hour
3 variations can be much larger. An analysis of the distribution of increases in hour-to-hour
4 concentrations at multiple sites across the U.S. for 2001-2003 found site-level median hourly
5 increases ranging up to 6 ug/m3 (maximum), with an average median increase of about 1.8 ug/m3.
6
7 2.5.2 Ultrafme Patterns
8 Diurnal or seasonal patterns for ultrafine particles have been studied in relatively few
9 areas of the U.S. A study done at the most extensively studied urban location in the U.S., Atlanta,
10 GA, is discussed in the CD (p.3-32). In this study, (CD, p. 3-32 to 3-33) ultrafine particle number
11 concentrations were found to be higher in the winter than in the summer. Concentrations of
^
12 particles in the range of 0.01 to 0.1 um were higher at night than during the daytime, and tended
13 to reach their highest values during the morning period when motor vehicle traffic is heaviest.
14 Smaller particles in the range of 0.004 to 0.01 um were elevated during the peak traffic period,
15 most notably in cooler temperatures, below 50°F..
16
17 2.6 PM BACKGROUND LEVELS
18 For the purposes of this document, policy-relevant background (PRB) (referred to as
19 "background" in the rest of this section) PM is defined as the distribution of PM concentrations
20 that would be observed in the U.S. in the absence of anthropogenic (man-made) emissions of
21 primary PM and precursor emissions (e.g., VOC, NOXJ SO2, and NH3) in the U.S., Canada, and
22 Mexico. The reason for defining background in this manner is that for purposes of determining
23 the adequacy of current standards and the need, if any, to revise the standards, EPA is focused on
24 the effects and risks associated with pollutant levels that can be controlled by U.S. regulations or
25 through international agreements with border countries. Thus, as defined here, background
26 includes PM from natural sources in the U. S. and transport of PM from both natural and
27 man-made sources outside of the U.S. and its neighboring countries.
28 Section 3.3.3 of the CD discusses annual average background PM levels, and states that
29 " [estimates of annually averaged PRB concentrations or their range have not changed from the
30 1996 PM AQCD" (CD, p. 3-105). These ranges for PM25 and PM10 are reproduced in Table 2-5.
31 The lower bounds of these ranges are based on estimates of "natural" background midrange
January 2005 2-60 Draft - Do Not Quote or Cite
-------
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
concentrations. The upper bounds are derived from the multi-year annual averages of the remote
monitoring sites in the IMPROVE network (EPA, 1996a, p. 6-44). The ranges for PM,0.25 are
derived from the PM2.5 ranges and the PM,0 ranges by subtraction (CD, p. 3-83). Since the
IMPROVE data unavoidably reflect some contributions from the effects of anthropogenic •
emissions from within the U.S., Canada, and Mexico, as well as background, they likely
overestimate the U.S. background concentrations as defined here.
There is a distinct geographic difference in background levels, with lower levels in the
western U.S. and higher levels in the eastern U.S. The eastern U.S. is estimated to have more
natural organic fine particles and more water associated with hygroscopic fine particles than the
western U.S. due to generally higher humidity levels.
Table 2-5. Estimated Ranges of Annual Average PM Regional Background Levels
'
PM10
PM25
PM10.2.5
Western U.S. (ug/m3)
4-8
1-4
0-7
Eastern U.S. (ue/m3)
5-11
2-5
0-9
Background levels of PM vary by geographic location and season, and have a natural
component and an anthropogenic component. The natural background arises from: (1) physical
processes of the atmosphere that entrain coarse particles (e.g., windblown crustal material, sea
salt spray); (2) volcanic eruptions (e.g., sulfates); (3) natural combustion such as wildfires (e.g.,
elemental and organic carbon, and inorganic and organic PM precursors); and (4) biogenic
sources such as vegetation, microorganisms, and wildlife (e.g., organic PM, inorganic and organic
PM precursors). The exact magnitude of the natural portion of background PM for a given
geographic location cannot be precisely determined because it is difficult to distinguish local
sources of PM from the long-range transport of anthropogenic particles and precursors.
PM can be transported long distances from natural or quasi-natural events occurring
outside the continental U.S. (CD, p. 3-82). The occurrence and location of these long-range
transport events are highly variable and their impacts on the U.S. are equally variable. The
contributions to background from sources outside of the U.S., Canada, and Mexico can be
significant on an episodic, but probably not on an annual basis (CD, p. 3-91). Several studies
January 2005
2-61
Draft - Do Not Quote or Cite
-------
1 have focused on identifying the origin, sources, and impacts of recent transnational transport
2 events from Canada, Mexico, and extra-continental sources.
3
4 • The transport of PM from biomass burning in Central America and southern Mexico in
5 1998 has been shown to contribute to elevated PM levels in southern Texas and
6 throughout the entire central and southeastern United States (CD, p. 3^86).
7
8 • Wildfires in the boreal forests of northwestern Canada may impact large portions of the
9 eastern United States. The CD estimates that a July 1995 Canadian wildfire episode
10 resulted in excess PM2 5 concentrations ranging from 5 ug/m3 in the Southeast, to nearly
11 100 ug/m3 in the northern Plains States (CD, p. 3-87).
12
13 . • Windblown dust from dust storms in the North African Sahara desert has been observed in
14 satellite images as plumes crossing the Atlantic Ocean and reaching the southeast coast of
15 the U.S., primarily Florida; North African dust has also been tracked as far as Illinois and
16 Maine. These events have been estimated to contribute 6 to 11 ug/m3 to 24-hour average
17 PM25 levels in affected areas during the events (CD, p. 3-84).
18
19 • Dust transport from the deserts of Asia (e.g., Gobi, Taklimakan) across the Pacific Ocean
20 to the northwestern U.S. also occurs. Husar et al. (2001) report that the average PM10
21 level at over 150 reporting stations throughout the northwestern U.S. was 65 ug/m3 during
22 an episode in the last week in April 1998, compared to an average of about 20 ng/m3
23 during the rest of April and May (CD, p. 3-84).
24
25 Background concentrations of PM2.5, PM]0_25, and PM10 may be conceptually viewed as
26 comprised of baseline and episodic components. The baseline component is the contribution
27 from natural sources within the U.S., Canada, and Mexico and from transport of natural and
28 anthropogenic sources outside of the U.S., Canada, and Mexico that is reasonably well
29 characterized by a consistent pattern of daily values each year, although they may vary by region
30 and season.
31 In addition to this baseline contribution to background concentrations, a second
32 component consists of more rare episodic high-concentration events over shorter periods of time
33 (e.g., days or weeks) both within the U.S., Canada, and Mexico (e.g., volcanic eruptions, large
34 forest fires) and from outside of the U.S., Canada, and Mexico (e.g., transport related to dust
35 storms from deserts in North Africa and Asia). Over shorter periods of time (e.g., days or weeks),
36 the range of background concentrations is much broader than the annual averages. Specific
37 natural events such as wildfires, volcanic eruptions, and dust storms, both of U.S. and
38 international origin, can lead to very high levels of PM comparable to, or greater than, those
January 2005 2-62 Draft - Do Not Quote or Cite
-------
1 driven by man-made emissions in polluted urban atmospheres. Because such excursions can be
2 essentially uncontrollable, EPA has in place policies that can remove consideration of them,
3 where appropriate, from attainment decisions.22
4 Disregarding such large and unique events, an estimate of the range of "typical"
5 background on a daily basis can be obtained from reviewing multi-year data at remote locations.
6 EPA staff have conducted an analysis of daily PM2 5 measurements from 1990 to 2002 at
7 IMPROVE sites across the U.S., focused on the non-sulfate components of PM25 (Langstaff,
8 2005). Ambient sulfate concentrations are almost entirely due to anthropogenic sources (with the
9 exception of sulfates from volcanic eruptions), so while non-sulfate PM25 is partly of
10 anthropogenic origin, it captures almost all of the background.
11 Based on regional differences in geography and land.use, the U. S. is divided into a
12 number of regions for estimating regional background levels. The "Eastern U.S." region extends
13 west to include Minnesota, Iowa, Missouri, Arkansas, and Louisiana. The "Central West" region
14 comprises states west of the Eastern U.S. region and east of Washington, Oregon, and California.
15 Washington, Oregon, and northern California make up the "North West Coast" and southern
16 California (south of about 40 degrees-latitude) makes up the "South West Coast" regions.23
17 To arrive at estimates of background we use the averaged measured non-sulfate PM2 5
18 values at IMPROVE sites in these regions. The Eastern U.S. region is heavily impacted by
19 anthropogenic emissions and we selected sites in northern states, which we judge to be affected to
20 a lesser extent by anthropogenic pollution, to derive estimates of background concentrations,
21 using all IMPROVE sites in the selected states. In all of the other regions we include all of the
22 There are two policies which allow PM data to be flagged for special consideration due to natural events:
the Exceptional Events Guideline (EPA, 1986) and the PM10 Natural Events Policy (Nichols, 1996). Under these .
policies, EPA will exercise its discretion not to designate areas as nonattainment and/or to discount data in
circumstances where an area would attain but for exceedances that result from uncontrollable natural events. Three
categories of natural PM10 events are specified in the natural events policy: volcanic or seismic activity, wildland
fires, and high wind dust events. The exceptional events policy covers natural and other events not expected to recur
at a given location and applies to all criteria pollutants. Categories of events covered in the exceptional events
guidance include, but are not limited to, high winds, volcanic eruptions, forest fires, and high pollen counts. EPA is
drafting further guidance concerning how to handle data affected by natural events related to the PM standards.
23 The 'Eastern' region roughly equates to the combined Southeast, Northeast, Industrial Midwest, and
eastern portion (MN, 1A, & MO) of the Upper Midwest regions as defined previously in this chapter (Figure 2-4).
The 'Central West' region roughly corresponds to the western portion of the Upper Midwest region and the eastern .
two thirds (ID, MT, CO, UT, NV) of the Northwest region. The 'North West Coast' approximates the remaining
one third (northern CA, OR, and WA) of the Northwest region. The' South West Coast' area is similar to the
southern California region.
•
January 2005 ' 2-63 " Draft - Do Not Quote or Cite
-------
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
IMPROVE sites. Table 2-6 describes the IMPROVE sites selected to represent these different
regions of the U.S. We recognize that these estimates will likely be biased high, as they include
an anthropogenic component, some sites more than others.
The 99th percentile concentrations at each of these sites were calculated to assess high
values measured at these sites, while avoiding excursions that potentially reflect exceptional
natural events. Standard deviations were also calculated for characterization of the daily variation
of background concentrations. Table 2-7 presents the results of this analysis as means and ranges
of individual site statistics within each of the background regions.
Table 2-6. IMPROVE sites selected for estimates of regional background
Region
Eastern U.S.
Central West
North West Coast
South West Coast
Alaska
Hawaii
IMPROVE Sites
All sites in Maine, New Hampshire, Vermont, Minnesota, and Michigan
All sites in this region (sites in ID, MT, WY, ND, SD, CO, UT, NV, AZ)
All sites in this region (all Washington and Oregon sites, and the northern
California sites REDW and LAVO)
All sites in this region (all California sites except the northern sites REDW
and LAVO)
All sites in Alaska
All sites in Hawaii
Table 2-7. Estimates of long-term means, daily standard deviations and 99th percentiles of
PM2f background concentrations (|ig/m3)
Region
Eastern U.S.
Central West
North West Coast
South West Coast
Alaska
Hawaii
# Sites
7
37
8
8
1
3
Means
3.0 (2.5-3.6)
2.5 (1.6-4.6)
3.4 (2.2-6.6)
5.2 (2.6-8.6)
1.2
1.1 (0:7-1.8)
St Devs
2.5 (2.1-2.8)
1.9(1.3-3.7)
2.8(2.1-4.2)
3.7(1.8-6.8)
1.5
0.9(0.8-1.0)
99th %iles
13(11-15)
10 (6-17)
14(10-21)
20 (9-33)
9
4(4-5)
Notes:
1) Some of these estimates likely contain a significant North American anthropogenic component.
2) The "Means" column has the mean of the long-term averages of the sites representing the region followed by the
minimum and maximum of the long-term averages of these sites in parentheses. Similarly for the "St Devs" column,
which presents standard deviations of the daily concentrations about the annual means, and the "99* %iles" column,
which presents the 99* percentiles of the daily concentrations over the 23-year period.
January 2005
2-64
Draft - Do Not Quote or Cite
-------
1 Considering these factors, the distributions of daily PM2 5 concentrations at these sites
2 provide an indication of the ranges for the daily variability of PM2 5 background concentrations,
3 and the 99th percentiles'of these distributions are an estimate of the highest daily background
4 concentrations. Staff notes that these recent findings are generally consistent with those from the
5 last review, which suggested a range of about 15 to 20 jig/in3 as the upper end of the distribution
6 of daily PM25 background concentrations in the U.S. (EPA, 1996b).
7 - . ' ,
8 2.7 RELATIONSHIP BETWEEN AMBIENT PM MEASUREMENTS AND HUMAN
9 EXPOSURE
10
11 The statutory focus of the primary NAAQS for PM is protection of public health from the
12 . adverse effects associated with'the presence of PM in the ambient air - that is, the focus is on
13 particles in the outdoor atmosphere that are either emitted directly by sources or formed in the '
14 atmosphere from precursor emissions. We refer to the concentrations of PM in the ambient air as
15 ambient PM. An understanding of human exposure to ambient PM helps inform the evaluation of
16 underlying assumptions and interpretation of results of epidemiologic studies that characterize
17 relationships between monitored ambient PM concentrations and observed health effects
18 (discussed in Chapter 3). '
19 An important exposure-related issue for this review is the characterization of the
20 relationships between ambient PM concentrations measured at one or more centrally located
21 monitors and personal exposure to ambient PM, as characterized by particle size, composition,
22 source origin, and other factors. Information on the type and sfrength of these relationships,
23 discussed below, is relevant to the evaluation and interpretation of associations found in
24 epidemiologic studies that use measurements of PM concentrations at centrally located monitors
25 as a surrogate for exposure to ambient PM.24 The focus here is on particle size distinctions; the
26 CD (CD, Section 5.4) also discusses exposure relationships related to compositional differences.
27 ' '
24 Consideration of exposure measurement error and the effects of exposure misclassification on the
interpretation of the epidemiologic studies are addressed in Chapter 3.
January 2005 2-65 Draft - Do Not Quote or Cite
-------
1 2.7.1 Definitions
2 Exposure to a contaminant is defined as contact at a boundary between a human and the
3 environment (e.g., the breathing zone) at a specific contaminant concentration for a specific
4 interval of time; it is measured in units of concentrations) multiplied by time (or time interval)
5 (National Research Council, 1991). An individual's total personal exposure to PM results from
6 breathing air containing PM in different types of environments (e.g., outdoors near home,
7 outdoors away from home, indoors at home, indoors at office or school, commuting, restaurants,
8 malls, other public places). These environments may have different concentrations of PM with
9 particles originating from a wide variety of sources.
10 Ambient PM is comprised of particles emitted by anthropogenic and natural sources and
11 particles formed in the atmosphere from emissions of gaseous precursors. This includes
12 emissions not only from outdoor sources such as smokestacks, industrial sources, and
13 automobiles, but also from sources located indoors with emissions vented outdoors, such as
14 fireplaces, wood stoves, and some cooking appliances. Exposure to ambient PM can occur both
15 outdoors and indoors to the extent that ambient PM penetrates into indoor environments - we use
16 the term PM of ambient origin to refer to both outdoor and indoor concentrations of ambient PM.
17 We use the term nonambient PM to refer to concentrations of PM that are only due to indoor
18 sources of particles that are not vented outdoors such as smoking, cooking, other non-vented
19 sources of combustion, cleaning, mechanical processes, and chemical interactions producing
20 particles. In characterizing human exposure to PM concentrations relevant to setting standards
21 for ambient air quality, the CD conceptually separates an individual's total personal exposure to
22 PM into exposure to PM of ambient origin and exposure to all other sources of PM (i.e.,
23 nonambient PM exposure).
24 Outdoor concentrations of PM are affected by emissions, meteorology, topography,
25 atmospheric chemistry, and removal processes. Indoor concentrations of PM are affected by
26 several factors, including outdoor concentrations, processes that result in infiltration of ambient
27 PM into buildings, indoor sources of PM, aerosol dynamics and indoor chemistry, resuspension of
28 particles, and removal mechanisms such as particle deposition, ventilation, and air-conditioning
29 and air cleaning devices (CD, p. 5-122). Concentrations of PM inside vehicles are subject to
30 essentially the same factors as concentrations of PM inside buildings. Personal exposure to PM
31 also includes a component which results specifically from the activities of an individual that
January 2005 2-66 Draft - Do Not Quote or Cite
-------
0
1 typically generate particles affecting only the individual of a small localized area surrounding the
2 person, such as walking on a carpet, referred to as the personal cloud.
3 Epidemiologic studies generally use measurements from central monitors to represent the
• •
4 ambient concentrations in an urban or rural area. We use the term central'site to mean the site of
5 a PM monitor centrally located with respect to the area being studied. In many cases,
6 epidemiologic studies combine'the measurements from more than one monitor to obtain a broader
7 representation of area-wide PM concentrations than a single monitor provides.
8 •.'•„. ' . ;' •
9 2.7.2 Centrally Monitored PM Concentration as a Surrogate for Particle Exposure
10 The 1996 Criteria Document (EPA, 1996a) presented a thorough review of PM exposure-
11 related studies up to that time. The 1996 Staff Paper (EPA, 1996b) drew upon the studies;
12 analyses, and conclusions presented in the 1996 Criteria Document and discussed two '
* »
13 interconnected PM exposure issues: (1) the ability of central fixed-site PM monitors to represent
14 population exposure to ambient PM and (2) how differences between fine and coarse particles
15 affect population exposures. Distinctions between PM size classes and components were found to
1
16 be important considerations in addressing the representativeness of central monitors. For ' '-
17 example, fine particles have a longer'residence time and generally exhibit less variability in the
18 atmosphere than coarse fraction particles. As discussed in the 1996 Staff Paper, the 1996' Criteria
19 Document concluded that measurements of daily variations, of PM have a plausible linkage to
20 daily variations of human exposures to PM of ambient origin for the populations represented by
23 monitoring can be a useful, if imprecise, index for representing the average exposure of people in
24 a community to PM "of ambient origin (EPA, 1996b, p. IV-15,16).
25 Exposure studies published since 1996 and reanalyses of studies that appeared in the 1996
26 Criteria Document are reviewed in the current CD, and provide additional support for these
27 findings. The CD discusses two classes of fine particles: ultrafine and accumulation-mode
28 particles (see Chapter 2). Ultrafine, accumulation-mode, arid coarse particles have different
29 chemical and physical properties which affect personal exposures in different ways (CD, Table 9-
30 2, p. 9-17)1 ' '
Q
^^
-January 2005 2-67 Draft - Do Not Quote or Cite
-------
©
1 PM]0.2 5 ahd ultrafine particles penetrate indoors less readily than PM2 5 and deposit to surfaces
2 more rapidly than PM2 5, a greater proportion of PM25 of ambient origin is found indoors than
3 PM10.2 5 and ultrafine particles, relative to their outdoor concentrations. Thus, the particle size
4 distribution influences the amounts of PM of ambient origin found indoors.
5 Since people typically spend a large part of their time indoors at home, the air exchange "
• i ' • r * * .
6 rate of the home has a large impact on exposures to ambient pollution. Homes with low air
7 exchange rates are more protected from outdoor sources, and vice-versa. Homes in regions with
8 moderate climate tend to be better ventilated and have higher air exchange rates than areas which
9 have very cold or very hot climates. Thus, climate plays an important role in regional population
10 exposure to ambient pollution.
11 " The third factor influencing the relationship between ambient concentrations measured at
1 2 central sites and total personal exposure is the contribution of indoor sources to total personal
1 " ..(;,.
13 exposure. On average, individuals spend nearly 90 percent of their time indoors. The
T K '-, , .' • •
14 contribution of indoor sources to indoor concentrations of PM is significant, and can be quite
1 5 variable on different days and between individuals. Indoor sources such as combustion devices
16 (e.g., stoves and kerosene heaters) generate predominantly fine particles; cooking produces both
1 1 fine and coarse particles; and resuspension (e.g., dusting, vacuuming, and walking on rugs)
1 8 generates predominantly coarse particles (CD, p. 5-82). This factor, however, does not influence
' " • ' ' • . , 4 '
1 9 exposure to PM of ambient origin.
...;.. -, i .
20 These three factors related to total personal exposure can give rise to measurement error in
21 estimating exposures to fine and coarse PM (CD, Section 5.5.3), thus making the quantification of
i ' ' f •••',')
22 relationships between concentrations measured at central site monitors and health effects more
23 difficult due to reduction in statistical power. Moreover, exposure measurement errors can also
24 affect the magnitude and the precision of the health effects estimates. However, as discussed in
25 the CD and below in Chapter 3, exposure measurement errors under most ordinary circumstances
26 are not expected to influence the overall interpretation of findings from either the long-term
27 exposure of time-series epidemiologic studies that have used ambient concentration data (CD, p.
28 5-121). •'
29 The CD discusses the finding by some researchers that some epidemiologic studies yield
30 statistically significant associations between ambient concentrations measured at a central site and
3 1 health effects even though there is a very small correlation between ambient concentrations
January 2005 . ,. 2-70 Draft - Do Not Quote or Cite
-------
1 measured at a central site and total personal exposures. The explanation of this finding is that
2 total personal exposure includes both ambient and nonambient generated components, and while
3 the nonambient portion of personal exposure is not generally correlated with ambient
4 concentrations, the exposure to concentrations of ambient origin is correlated with ambient
5 concentrations. Thus, it is not surprising that health effects might correlate with central site PM
6 concentrations, because exposure to PM of ambient origin correlates with these concentrations,
7 and the lack of correlation of total exposure with central site PM concentrations does not
8 statistically 'alter that relationship. By their statistical design, time-series epidemiologic studies of
9 this type only address the ambient component of exposure, since the impact of day-to-day
10 fluctuations in ambient PM on acute health effects is examined.
11 In looking more specifically at the relationship between personal exposure to PM of
12 ambient origin and concentrations measured at central site monitors, an analysis of data from the
13 PTEAM study25 provides important findings, as discussed in the CD (p. 5-63 to 5-66 and 5-125 to
14 5-126). The PTEAM study demonstrated that central site ambient PM10 concentrations are well
15 correlated with personal exposure to PM10 of ambient origin, while such concentrations are only
16 weakly correlated with total personal exposure. This study also found that estimated exposure to
17 nonambient PM10 is effectively independent of PM10 concentrations at central site monitors, and
18 . that nonambient"exposures are highly variable due to differences in indoor sources across the
19 study homes.
20 When indoor sources only have minor contributions to personal exposures, total exposure
21 is mostly from PM of ambient origin. In these cases high correlations are generally found
22 between total personal exposure and ambient PM measured at a central site (CD, p. 5-54). For
23 example, measurements of ambient sulfate, which is mostly in the fine fraction, have been found
24 to be highly correlated with total personal exposure to sulfate (CD, p. 5-124). Since.in these
25 studies there were minimal indoor sources of sulfate, the relationship between ambient
26 concentrations and total personal exposure to sulfate was not weakened by possible presence of
27 small indoor-generated sulfates in some environments.
25 EPA's Particle Total Exposure Assessment Methodology (PTEAM) field study (Clayton et al., 1993;
Ozkaynak et al., 1996a;b) is a large-scale probability sample based field study. The study measured indoor, outdoor,
and personal PM10, the air exchange rate for each home, and time spent in various indoor residential and outdoor
environments for 147 subjects/households, 12-hr time periods in Riverside, California.
January 2005 2-71 ' Draft - Do Not Quote or Cite
-------
1 It is recognized that existing PM exposure measurement errors or uncertainties most likely
2 will reduce the statistical power.of PM health effects analyses, thus making it more difficult to
3 detect a true underlying association between the exposure metric and the health outcome of
4 interest. However, the use of ambient PM. concentrations as a surrogate for personal ambient
5 exposures is not expected to change the principal conclusions from PM epidemiological studies
6 mat use community average health and pollution data (CD, p. 5-121). Based on these
7 considerations and on the review of the available exposure-related studies, the CD concludes that
8 for epidemiologic studies, ambient PM2 5 concentration as measured at central site monitors is a
9 useful surrogate for exposure to PM2 5 of ambient origin. However, for coarse and ultrafine PM,
10 such ambient concentrations are not likely to be as good a surrogate for personal ambient
11 exposure. While nonambient PM may also be responsible for health effects, since the ambient
12 and nonambient components of personal exposure are independent, the health effects due to
13 nonambient PM exposures generally will not bias the risk estimated for ambient PM exposures
14 (CD, p. 9-17).
15
16 2.8 RELATIONSHIP BETWEEN AMBIENT PM AND VISIBILITY
17 . The effect of ambient particles on visibility is dependent upon particle size and
18 composition, atmospheric illumination, the optical properties of the atmosphere, and the optical
19 properties of the target being viewed. The optical properties of particles, discussed in section
20 2.2.5, can be well characterized in terms of a light extinction coefficient. For a given distribution
21 of particle sizes and compositions, the light extinction coefficient is strictly proportional to the
22 particle mass concentration. Light extinction is a measure of visibility impairment, and, as such,
23 provides a linkage between ambient PM and visibility, as discussed below in section 2.8.1. Other
24 measures directly related to the light extinction coefficient are also used to characterize visibility
25 impairment, including visual range and deciviews, as discussed below in section 2.8.2. Light
26 extinction associated with background levels of PM is also discussed below in section 2.8.3.
27
28 2.8.1 Particle Mass and Light Extinction
29 Fine particle mass concentrations can be used as a general surrogate for visibility
30 impairment. However, as described in many reviews of the science of visibility, the different
31 constituents of PM2 5 have variable effects on visibility impairment. For example, sulfates and
January 2005 "' 2-72 Draft - Do Not Quote or Cite
-------
1 nitrates contribute substantially more to light scattering per unit mass than other constituents,
2 especially as relative humidity levels exceed 70 percent. Thus, while higher PM2 5 mass
3 concentrations generally indicate higher levels of visibility impairment, it is not as precise a
4 metric as the light extinction coefficient. By using historic averages, regional estimates, or actual
5 day-specific measurements of the component-specific percentage of total mass, however, one can
6 • develop reasonable estimates of light extinction from PM mass concentrations (see section 6.2.2
7 for further discussion).
8 The light extinction coefficient has been widely used in the U.S. for many years as a
9 metric to describe the effect of concentrations of particles and gases on visibility. It can be
10 defined as the fraction of light lost or redirected per unit distance through interactions with gases
11 and suspended particles in the atmosphere. The light extinction coefficient represents the
12 summation of light scattering and light absorption due to particles and gases in the atmosphere.
13 Both anthropogenic and non-anthropogenic sources contribute to light extinction. The light
14 extinction coefficient (£>«,) is represented by the following equation (CD, 4-155):
15
16 6ext = 6ap + bag + isg+6sp (5-1)
17
18 "where £>ap = light absorption by particles
19 &gg = light absorption by gases
20 bss = light scattering by gases (also known as Rayleigh scattering)
21 65p = light scattering 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 converted
28 to the path-veraged light extinction coefficient Transmissometers operate continuously, and data
29 are often reported in terms of hourly averages.
3 0 Direct relationships exist between measured ambient pollutant concentrations and their
f 31 contributions to the extinction coefficient. The contribution of each aerosol constituent to total
January 2005 . 2-73 Draft - Do Not Quote or Cite
-------
1 light extinction is derived by multiplying the aerosol concentration by the extinction efficiency
2 for that aerosol constituent. Extinction efficiencies vary by type of aerosol constituent and have
3 been obtained for typical atmospheric aerosols by a combination of empirical approaches and
4 theoretical calculations. For certain aerosol constituents, extinction efficiencies increase
5 significantly with increases in relative humidity.
6 The IMPROVE visibility monitoring program has developed an algorithm for calculating
7 total light extinction as the sum of aerosol light extinction for each of the five major fine particle
8 components and for the coarse fraction mass, plus 10 Mm"1 for light extinction due to Rayleigh
9 scattering, discussed below. This algorithm is represented by the following equation (CD, 4-
10 169):
11
12 bta= (3)f(RH) [SULFATE]
13 + (3)f(RH) [NITRATE]
14 - +(1.4) [ORGANIC CARBON]
15 +(W) [LIGHTABSORBING CARBON] (5-2)
16 +(1)[SOIL]
17 +(0.6) [COARSE PM]
18 +10 (for Rayleigh scattering by gases)
19
20 The mass for each component is multiplied by its dry extinction efficiency and, in the case
21 of sulfate and nitrate, by a relative humidity adjustment factor, f(RH), to account for their
22 hygroscopic behavior (CD, p. 4-169). The relative humidity adjustment factor increases
23 significantly with higher humidity, ranging from about 2 at 70 percent, to 4 at 90 percent, and
24 over 7 at 95 percent relative humidity (CD, p. 4-170, Figure 4-38).
25 Rayleigh scattering represents the degree of natural light scattering found in a particle-free
26 atmosphere, caused by the gas molecules that make up "blue sky" (e.g., N2, O2). The magnitude
27 of Rayleigh scattering depends on the wavelength or color of the light being scattered, as well as
28 on the density of gas in the atmosphere, and varies by site elevation, generally from 9 to 11 Mm"1
29 for green light at about 550 nm (CD, p. 4-156 to 4-157). A standard value of 10 Mm"1 is often
30 used to simplify comparisons of light extinction values across a number of sites with varying
31 elevations (Malm, 2000; CD, p. 4-157). The concept of Rayleigh scattering can be used to
-------
1 establish a theoretical maximum horizontal visual range in the earth's atmosphere. At sea level,
2 mis maximum visual range is approximately 330 kilometers. Since certain meteorological
3 conditions can lead to visibility conditions that are close to "Rayleigh," it is analogous to a
4 baseline or boundary condition against which other extinction components can be compared.
5 The light extinction coefficient integrates the effects of aerosols on visibility, yet is not
6 dependent on scene-specific characteristics. It measures the changes in visibility linked to
7 emissions of gases and particles. By apportioning the light extinction coefficient to different
8 aerosol constituents, one can estimate changes in visibility due to changes in constituent
9 concentrations (Pitchford and Malm, 1994).
10
11 2.8.2 Other Measures of Visibility
12 Visual range is a measure of visibility that is inversely related to the extinction coefficient.
13 Visual range can be defined as the maximum distance at which one can identify a large black
14 object against the horizon sky. The colors and fine detail of many objects will be lost at a
15 distance much less than the visual range, however. Visual range has been widely used in air
16 transportation and military operations in addition to its use in characterizing air quality.
17 Conversion from the extinction coefficient to visual range can be made with the following
18 equation (NAPAP, 1991):
19
20 Visual Range (km) = 3912/Z^/Mm-1) (5-3)
21
22 , Another important visibility metric is the deciview, a unitless metric which describes
23 changes in uniform atmospheric extinction that can be perceived by a human observer. It is
24 designed to be linear with respect to perceived visual changes over its entire range in a way that is
25 analogous to the decibel scale for sound (Pitchford and Malm, 1994). Neither visual range nor
26 the extinction coefficient has this property. For example, a 5 km change in visual range or 0.01
27 km"1 change in extinction coefficient can result in a change that is either imperceptible or very
28 apparent depending on baseline visibility conditions. Deciview allows one to more effectively
29 express perceptible changes in visibility, regardless of baseline conditions. A one.deciview
30 change is a small but perceptible scenic change under many conditions, approximately equal to a
31 10 percent change in the extinction coefficient (Pitchford and Malm, 1994). Deciview can be
32 calculated from the light extinction coefficient (&«,) by the equation:
January 2005 2-75 Draft - Do Not Quote or Cite
-------
1
2 Haziness (dv) = 10
3
4 Figure 2-29 graphically illustrates the relationships among light extinction, visual range, and
5 deciview.
6
Extinction (Mm"1) 10 20 30 40 so 70 100 200 300 «o soo 7001000
7 """"
Deciviews (dv)
8
I
0
!
I
7
I
I
11
I
I
14
1
1
16
I
I INI
19 23
I INI
I
30
I
I
34
I
I
37
I
1
39
1
i mi
42 46
i nil
Visual Range (km) 400 2oo iso too so so AO 20 13 10 s s 4
9
I o Figure 2-29. Relationship between light extinction, deciviews, and
visual range.
11
Source: Malm (1999)
12
13 2.8.3 Visibility at PM Background Conditions
14 Light extinction caused by PM from natural sources can vary significantly from day to day
15 and location to location due to natural events such as wildfire, dust storms, and volcanic
16 eruptions. It is useful to consider estimates of natural background concentrations of PM on an
17 annual average basis, however, when evaluating the relative contributions of anthropogenic (man-
18 made) and non-anthropogenic sources to total light extinction. Background PM is defined and
19 discussed in detail in section 2.6, and Table 2-5 provides the annual average regional background
20 PM25 mass ranges for the eastern and western U.S.
21 The National Acid Precipitation Assessment Program report (NAPAP, 1991) provides
22 estimates of extinction contributions from background levels of fine and coarse particles, plus
23 Rayleigh scattering. In the absence of anthropogenic emissions of visibility-impairing particles,
24 these estimates are 26 ±7 Mm'1 in the East, and 17 + 2.5 Mm"1 in the West These equate to a
25 naturally-occurring visual range in the East of 150 + 45 km, and 230 + 40 km in the West.
26 Excluding light extinction due to Rayleigh scattering, annual average background levels of fine
27 and coarse particles are estimated to account for approximately 14 Mm"1 in the East and about 6
28 Mm"1 in the West. The primary non-anthropogenic substances responsible for natural levels of
29 visibility impairment are naturally-occurring organic?, suspended dust (including coarse
30 particles), and water associated with hygroscopic particles. At the ranges of fine particle
31 concentrations associated with background conditions, discussed above in section 2.6, small
January 2005 2-76 Draft - Do Not Quote or Cite
-------
1 changes in fine particle mass have a large effect on total light extinction. Thus, higher levels of
2 background fine particles and associated average humidity levels in the East result in a fairly
3 significant difference between naturally occurring visual range in the rural East as compared to
4 the rural West. This issue is discussed further in Chapter 6, section 6.2.
t
5 Fine particles originate from both natural and anthropogenic, or man-made, sources.
6 Background concentrations of fine particles are those originating from natural sources. On an
7 annual average basis, concentrations of background fine particles are generally small when
8 compared with concentrations of fine particles from anthropogenic sources (NRC, 1993). The
9 same relationship holds true when one compares annual average light extinction due to
10 background fine particles with light extinction due to background plus anthropogenic sources.
11 Table VIII-4 in the 1996 Staff Paper makes this comparison for several locations across the
12 country by using background estimates from Table VIII-2 and light extinction values derived
13 from monitored data from the IMPROVE network. These data indicate that anthropogenic
14 emissions make a significant contribution to average light extinction in most parts of the country,
15 as compared to the contribution from background fine particle levels. Anthropogenic
16 contributions account for about one-third of the average extinction coefficient in the rural West
17 and more than 80 percent in the rural East (NAPAP, 1991).
18 It is important to note that, even in areas with relatively low concentrations of
19 anthropogenic fine particles, such as the Colorado plateau, small increases in anthropogenic fine
20 particle concentrations can lead to significant decreases in visual range. As discussed in the CD,
21 visibility in an area with lower concentrations of air pollutants (such as many western Class I
22 areas) will be more sensitive to a given increase in fine particle concentration than visibility in a
23 more polluted atmosphere. Conversely, to achieve a given amount of visibility improvement, a
24 larger reduction in fine particle concentration is required in areas with higher existing
25 concentrations, such as the East, than would be required in areas with lower concentrations. This
26 relationship between changes in fine particle concentrations and changes in visibility (in
27 deciviews) also illustrates the relative importance of the overall extinction efficiency of the
28 pollutant mix at particular locations. At a given ambient concentration, areas having higher
29 average extinction efficiencies, due to the mix of pollutants, would have higher levels of
30 impairment. In the East, the combination of higher humidity levels and a greater percentage of
January 2005 2-77 Draft - Do Not Quote or Cite
-------
1 sulfate as compared to the West causes the average extinction efficiency for fine particles to be
2 almost twice that for sites on the Colorado Plateau.
3
January 2005 2-78 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 Emissions
9 Inventory and Ambient Source Contribution Estimates: Summary of Current Knowledge and Needed
10 Research," Document No. 6110.4F, Reno, NV, May, 2000. Available:
11 www.ena. eov/ttn/chief/efdoc s/fugitivedust. pdf.
12
13 Environmental Protection Agency (1986). Guideline on the Identification and Use of Air Quality Data Affected by
14 Exceptional Events. EPA-450/4-86-007.
15
16 Environmental Protection Agency (1996a). Air Quality Criteria for Particulate Matter. Research Triangle Park, NC:
17 . Office of Research and Development;, report no. EPA/600/P-95/001aF-cF.3v
18 . .
19 Environmental Protection Agency (1996b). Review of the National Ambient Air Quality Standards for Particulate
20 Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle
21 Park, NC: Office of Air Quality Planning and Standards; report no. EPA-452/R-96-013.
22
23 Environmental Protection Agency (2004a). EPA's Environmental Technology Verification Program. Research
24 Triangle Park, NC: Office of Research and Development; report no. EPA/600F-04/064.
25
26 Environmental Protection Agency (2004b). The Particle Pollution Report. Current Understanding of Air Quality and
27 Emissions through 2003. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office of Air
28 Quality Planning and Standards; report rib. EPA-454-R-04-002.
29
30 Fehsenfeld, F.; D. Hastie; C. Chow, and PA. Solomon. "Gas and Particle Measurements, Chapter 5." In NARSTO
31 Particulate Matter Science Assessment. McMurray, P., Shepherd, M., and Vickery, J. eds. NARSTO, U.S.
3 2 Environmental Protection Agency, Research Triangle Park, NC. 2003.
33
34 Fitz-Simons,T.;Mathias, S.;Rizzo,M. (2000). U.S. EPA Memorandum to File. Subject: Analyses of 1999PM Data
35 for the PMNAAQS Review. November 17, 2000. Available: www.epa.gov/oar/oaaps/pm25/docs.html
36
37 Husar, R. B."; Tratt,D. M.; Schichtel, B. A; Falke, S. R.; Li, F.; Jaffe, D.; Gasso, S.; Gill, T.; Laulainen, N. S.; Lu, F.;
38 Reheis, M. C.; Chun, Y.; Westphal, D.; Holben, B. N.; Gueymard, C.; McKendry, I.; Kuring,N.;Feldman,
39 G. C.; McClain, C.; Frouin, R. I; Merrill, J.; DuBois, D.; Vignola, F.; Murayama, T.; Nickovic, S.; Wilson,
40 W. E.; Sassen, K.; Sugimoto, N.; Malm, W. C. (2001) Asian dust events of April 1998. J. Geophys. Res.
41 (Atmos.) 106:18,317-18,330.
42
43 Langstaff, J. E. (2005). OAQPS Staff Memorandum to PMNAAQS Review Docket (OAR-2001-0017). Subject:
44 Estimation of Policy-Relevant Background Concentrations of Particulate Matter. [January 27,2005].
45
46 Malm, W.C. (2000) Spatial and seasonal patterns and temporal variability of haze and its constituents in the United
47 States. Report III. Colorado State University, Cooperative Institute for Research in the Atmosphere. Fort
48 Collins, CO. Available: http://vista.cira.colostate.edu/improve/Publications/Reports/2000/2000.htm [22
49 March, 2002].
50
51 Malm, W.C.; Day, D.E.; Kreidenweis, S.M. (2000). Light scattering characteristics of aerosols as a function of
52 relative humidity: a comparison of measured scattering and aerosol concentrations using the theoretical
53 models. J. Air Waste Manage. Assoc. 50: 686-709
54
5 5 National Acid Precipitation Assessment Program (NAPAP), (1991). Office of the Director, Acid Deposition: State of
56 Science and Technology. Report 24, Visibility: Existing and Historical Conditions - Causes and Effects.
57 Washington, D.C.
January 2005 2-79 - Draft - Do Not Quote or Cite
-------
1
2 National Research Council (1991). Human exposure assessment for airborne pollutants: advances and opportunities.
3 Washington, DC: National Academy of Sciences.
4
5 National Weather Service. (1998) Automated Surface Observing System (ASOS) User's Guide. ASOS Program
6 Office. Silver Spring, MD.
7
8 Nichols, M. D. (1996) Memorandum to EPA Air Division Directors regarding Areas Affected by PM-10 Natural
9 Events. May 30,1996.
10
11 Noble, C.A.; S. Mukerjee; M. Gonzales; C.E. Rodes; P.A. Lawless; s. Natarajan; E.A. Myers; G.A. Norris; L. Smith;
12 H. Ozkaynak; L.M. Neas. (2003). Continuous measurement of fine and ultrafine particulate matter, criteria
13 pollutants and meteorological conditions in urban El Paso, Texas. Atmos. Environ. 37: 827-840.
14
15 Ozkaynak, H.; Xue, J.; Spengler, J.; Wallace, L.; Pellizzari, E.; Jenkins, P. (1996a). Personal exposure to airborne
16 particles and metals: results from the particle TEAM study in Riverside, California. J. Exp. Anal. Environ.
17 Epidemiol. 6: 57-78.
18
19 Ozkaynak, H.; Xue, J.; Weker, R.; Bulter, D.; Koutrakis, P.; Spengler, J. (1996b). The particle TEAM (PTEAM)
20 study: analysis of the data: final report, volume III. Research Triangle Park, NC: U.S. Environmental
21 Protection Agency, Atmospheric Research and Exposure Assessment Laboratory; report no. EPA/600/R-
22 95/098. Available from: NTIS, Springfield, VA; PB97-102495.
23
24 Pitchford, M.; Malm, W. (1994) Development and Applications of a Standard Visual Index. Atmospheric
25 Environment. Vol. 28, no. 5, pp. 1049-1054.
26
27 Schmidt, M.; D. Miritz; V. Rao; L. McCluney (2005). U.S. EPA Memorandum to File. Subject: Analyses of 2001-
28 2003PM Data for the PMNAAQS Review. January 31, 2005. Available:
29 www.epa.gov/oar/oaqps/pm25/docs.html.
30
31 Solomon, P.A.; M.P. Tolocka; G. Norris; and M. Landis (2001). "Chemical Analysis Methods for Atmospheric
32 Aerosol Components." In Aerosol Measurement: Principles, Techniques, and Application, Second Edition,
33 Eds. P. Barron and K. Willeke. John Wiley & Sons, Inc., New York, NY.
34
35 Watson, J.G., Robinson, N.F., Lewis, C.W., Coulter, C.T., Chow, J.C., Fujita, E.M., Lowenthal, D.H., Conner, T,L.,
36 Henry, R.C., and Willis, R.D. (1997). Chemical mass balance receptor model version 8 (CMB) user's
37 manual. Prepared for U.S. Environmental Protection Agency, Research Triangle Park, NC, Desert Research
38 Institute, Reno, NV.
39
40 Whitby, K. T. (1978). The physical characteristics of sulfur aerosols. Atmos. Environ. 12:135-159.
41
42 Wilson, W. E.; Sub, H.H. (1997) Fine particles and coarse particles: concentration relationships relevant to
43 epidemiologic studies. J. Air Waste Manage. Assoc. 47: 1238-1249.
44
January 2005 . . 2-80 Draft - Do Not Quote or Cite
-------
3. POLICY-RELEVANT ASSESSMENT OF HEALTH EFFECTS EVIDENCE
1 3.1 INTRODUCTION
2 This chapter assesses key policy-relevant information on the known and potential health"
3 effects associated with exposure to ambient PM, alone and in combination with other pollutants
4 that are routinely present in ambient air. More specifically, this assessment focuses on health
s
5 effects associated with exposures to ambient fine particles and to thoracic coarse particles,
6' consistent with EPA's decision in the last review to establish new standards for fine particles
7 • separate from those intended to address effects related to thoracic coarse particles. The
8 presentation here first summarizes the qualitative assessment of health evidence contained in the
9 CD, as a basis for development of staff conclusions and recommendations related to primary
10 standards for PM, as discussed in Chapter 5. Secondly, this assessment addresses key issues
11 relevant to quantitative assessment of the epidemiologic health evidence available in this review
12 so as to provide a foundation for quantitative health risk assessment, as discussed in Chapter 4.
13 Iri the last review of the PM NAAQS, a variety of health effects had been associated with
14 ambient PM at concentrations extending from those elevated levels found in the historic London
15 episodes down tb'levels below the 1987 PM10 standards. The epidemiologic evidence for PM-
16 related effects was found to be strong, suggesting a "likely causal role" of ambient PM in
17 contributing to a range of health effects (62 FR 38657). Of special importance in the last review
18 were the conclusions that (1) ambient particles smaller than 10 jim that penetrate into the
19 thoracic region of the respiratory tract remained of greatest concern to health, (2) the fine and
20 coarse fractions of PM10 should be considered separately for the purposes of setting ambient air
21 quality standards, and (3) the consistency and coherence of the health effects evidence greatly
22 . added to the strength and plausibility of the observed PM.associations. Important uncertainties
23 remained, however, such as issues related to interpreting the role of gaseous co-pollutants in PM
24 associations with health effects, and-the lack of demonstrated biological mechanisms that could
25 explain observed effects. "
26 EPA's conclusion in the last review that fine and thoracic coarsVparticles should be
27 considered as separate pollutants was based on differences in physical and chemical properties,
28 sources, atmospheric formation and transport, relationships with human exposure, and evidence
January 2005 , 3-1 Draft - Do Not Cite or Quote
-------
1 of health effects (62 FR 38667). In this review, the CD has evaluated the newly available
2 evidence related to the physics and chemistry of participate matter, exposure relationships, and
3 particle dosimetry. The CD notes that the chemical and physical distinctions between fine and
4 coarse particles recognized in the last review remain generally unchanged; recent studies
5 continue to show that fine and coarse particles generally have different sources and composition
6 and different formation processes (see Table 2-2 herein). Recent exposure research finds that
7 accumulation-mode fine particles can infiltrate into buildings more readily than can thoracic
8 coarse particles, and that ambient concentrations of PM,0.2 5 are less well correlated and less
9 uniform across a community than ambient concentrations of PM2 5 (CD, p. 9-21). The CD also
10 concludes that the new evidence from.dosimetry studies continues to reinforce some distinctions
11 between fine and coarse particles, and submodes within fine particles, with regard to deposition
)
12 patterns in the respiratory tract (CD, p. 9-21). While there is significant overlap between particle
13 size classes, thoracic coarse particles have somewhat greater deposition fractions in the upper,
14 regions of the respiratory tract, while fine particles generally (though not the larger
15 accumulation-mode particles) are more likely to be deposited in the alveolar region than are
16 thoracic coarse particles (CD, p. 9-21). Based on these considerations, the CD concludes that it
17 remains appropriate to consider fine and thoracic coarse particles as separate subclasses of PM
18 (CD, p. 9-22).
19 The assessment of health evidence in this chapter therefore focuses on health effects
20 associated with fine and thoracic coarse particles. This assessment is based on the CD's
21 evaluation and conclusions on the body of evidence from health studies, summarized in Chapters
22 6 through 9 of the CD, with particular emphasis on the integrative synthesis presented in Chapter
23 9. That integrative synthesis focuses on integrating newly available scientific information with
24 that available from the last review, as well as integrating information from various disciplines, so
25 as to address a set of issues central to EPA's assessment of scientific information upon which
26 this review of the PM NAAQS is to be based. It is intended to provide a coherent framework for
27 assessment of human health effects posed by ambient PM in the U.S., and to facilitate
28 consideration of the key. policy-related issues to be addressed in this Staff Paper, including
29 recommendations as to appropriate indicators, averaging times, levels, and forms for PM
30 NAAQS. As described in section 9.1 of the CD, the integrative synthesis focuses not only on
January 2005 3-2 Draft - Do Not Cite or Quote
-------
1 what has been learned since the last review, but also highlights important uncertainties that
2 remain and the value of continuing PM research efforts in a number of areas.
3 As summarized in Chapters 6 through 9 of the CD, a large number of new studies
4 containing further evidence of serious health effects have been published since the last review,
5 with important new information coming from epidemiologic, toxicologic, controlled human
6 exposure, and dosimetry studies. As was true in the last review, evidence from epidemiologic
7 studies plays a key role in the CD's evaluation of the scientific evidence. As discussed further in
8 section 3.3, some highlights of the new evidence include:
9 • New multi-city studies that use uniform methodologies to investigate the-effects of PM
10 on health with data from multiple locations with varying climate and air pollution mixes,
11 contributing to increased understanding of the role of various potential confounders,
12 including gaseous co-pollutants, on observed PM associations. These studies provide
13 more precise estimates of the magnitude of a PM effect than most smaller-scale
14 individual city studies.
15 '
16 • More studies of various health endpoints evaluating independent associations between
17 effects and fine and thoracic coarse particles, as well as dtrafine particles or specific
18 components (e.g., sulfates, metals).,
19 .
20 • Numerous new studies of cardiovascular endpoints, with particular emphasis on
21 assessment of cardiovascular risk factors or physiological changes.
22
23 • Studies relating population exposure to PM and other pollutants measured at centrally
24 located monitors to estimates of exposure to ambient pollutants at the individual level
25 have lead to a better understanding of the relationship between ambient PM levels and
26 /• personal exposures to PM of ambient origin.
27
28 • New analyses and approaches to addressing issues related to potential confounding by
> 29 gaseous co-pollutants, possible thresholds for effects, and measurement error and
30 exposure misclassification. .
31
32 • Preliminary attempts to evaluate the effects of air pollutant combinations or mixtures
33 including PM components using factor analysis or source apportionment methods to link
34 effects with different PM source types (e.g., combustion, crustal1 sources).
35
1 "Crustal" is used here to describe particles of geologic origin, which can be found in both fine- and
coarse-fraction PM. ' -
January 2005 3 -3 Draft - Do Not Cite or Quote
-------
1 • Several new "intervention studies" providing evidence for improvements in respiratory or
2 cardiovascular health with reductions in ambient concentrations of particles and gaseous
3 co-pollutants.
4
5 In addition, the body of evidence on PM-related effects has greatly expanded with
6 • findings from studies that help inform mechanism of action, including important new dosimetry,
7 toxicologic and controlled human exposure studies.
8
9 • Animal and controlled human exposure studies using concentrated ambient particles
10 (CAPs), new indicators of response (e.g., C-reactive protein levels, heart rate variability),
11 and animal models representing sensitive subpopulations, that are.relevant to the
12 > plausibility of the epidemiologic evidence and provide insights into potential mechanisms
13 for PM-related effects.
14
15 • Dosimetry studies using new modeling methods that provide increased understanding of
16 the dosimetry of different particle size classes and in members of potentially sensitive
17 subpopulations, such as people with chronic respiratory disease.
18
19 In presenting that evidence and conclusions based on it, this chapter first summarizes
20 information from the CD's evaluation of health evidence from the different disciplines. Sections
21 3.2 and 3.3 provide overviews of the CD's findings on the evidence of potential mechanisms for
22 PM-related effects and on the nature of effects associated with PM exposures, respectively.
23 Drawing from the integration of evidence in Chapter 9 of the CD, the chapter summarizes the
24 CD's integrative findings and conclusions regarding causality in section 3.4, with a particular
25 focus on results for fine and thoracic coarse particles. Section 3.5 also draws from the CD's
26 integrative synthesis to characterize potential at-risk subpopulations and potential public health
27 impacts of exposure to ambient PM. Finally, section 3.6 addresses several key issues relevant to
28 the staffs interpretation and quantitative assessment of the health evidence, including: (1)
29 considerations related to air quality measurements and data used in the health studies; (2)
30 exposure error in fine and thoracic coarse particle studies; (3) specification of models used in
31 epidemiologic studies; (4) approaches to evaluating the role of co-pollutants and potential
32 confounding in PM-effects associations; (5) questions of temporality in associations between air
33 quality and health effects, including lag periods used in short-term exposure studies and the
34 selection of time periods used to represent exposures in long-term exposures studies; and (6)
35 questions related to the form of concentration-response relationships and potential threshold
January 2005 3-4 Draft - Do Not Cite or Quote
-------
1 levels. In this final section, staff builds upon the CD's detailed evaluation and integration of the
2 scientific evidence on these issues to reach conclusions regarding the use of the health study
3 results in quantitative evaluation and risk assessments that inform staff recommendations on
4 potential revisions to the primary PM NAAQS presented in Chapter 5.
5
6 3.2 MECHANISMS
7 This section provides an overview of evidence presented in the CD on potential
8 ^ mechanisms by which exposure to PM may'result in effects, drawing from Chapters 6 and 7 of
9 the CD. Evidence from dosimetry studies has played a key role in previous PM NAAQS •
10 reviews, especially in the decision to revise the indicator from TSP to PM10 to focus on thoracic
11 particles (52 FR 24634, July 1,1987). In contrast, in previous reviews of the PM NAAQS there
12 has been little available evidence on potential biological mechanisms by which deposited
13 particles could affect the lungs or heart.
14 An evaluation of the ways by which inhaled particles might ultimately affect human
15 health must take account of patterns of deposition and clearance in the respiratory tract.
16 Particles must be deposited and retained in the respiratory tract for biological effects to occur
17 (CD, p. 6-1). Briefly, the human respiratory tract can be divided into three main regions: (1)
18 extra-thoracic, (2) tracheobronchial, and (3) alveolar (CD, Figure 6-1). The regions differ
19 markedly in structure, function, size, mechanisms of deposition and removal, and sensitivity or
20 reactivity to deposited particles; overall, the concerns related to ambient particles are greater for
21 the two lower regions.
22 Fine particles, including accumulation mode and ultrafme prticles, and thoracic coarse
23 particles can all penetrate into and be deposited in the alveolar and tracheobronchial regions of
24 the respiratory tract, though there are differences among these size fractions. The CD finds that
25 deposition patterns are generally similar for ultrafme and coarse particles, with a large fraction of
26 particles being deposited in the extrathoracic region. Removal of particles by the extrathoracic
27 region is less efficient for accumulation-mode fine particles, and thus penetration is increased to
28 the tracheobronchial and alveolar regions (CD, 6-105). The CD concludes that fractional
29 deposition into the alveolar region of the respiratory system for healthy individuals is greatest for
30 particles iri the size ranges of approximately 2.5 to 5 urn and 0.02 to 0.03 [am, and fractional
January 2005 3-5 Draft - Do Not Cite or Quote
-------
1 deposition to the tracheobronchial region is greatest for particles in the size range of
2 approximately 4 to 6 um (CD, p. 6-109). The junction of conducting and respiratory airways
3 appears to be a key anatomic focus; many inhaled particles of critical size are deposited in the
4 respiratory bronchioles that lie just distal to this junction. Recent studies have indicated that
5 ultrafme and thoracic coarse particles show enhanced deposition of particles at airway
6 bifurcations (CD, p. 6-20).
7 Breathing patterns and respiratory disease status can affect regional particle deposition
8 patterns. New evidence indicates that people with chronic lung disease can have increased total
4
9 lung deposition, and can also show increases in local deposition ("hot spots") due to uneven
10 airflow in diseased lungs (CD, p. 6-34). In such cases, the respiratory condition can enhance
11 sensitivity to inhaled particles by increasing the delivered dose to sensitive regions. Such *
12 dosimetry studies are of obvious relevance to identifying sensitive populations (see section 3.5).
13 The potential effects of deposited particles are influenced by the speed and nature of
s*
14 removal. The predominant clearance and translocation mechanisms vary across the three regions
15 of the respiratory system. For example, dissolution or absorption of particles or particle
16 constituents and endocytosis by cells such as macrophages are two primary mechanisms
17 operating in the alveolar region. These mechanisms also apply in the tracheobronchial region,
18 where two key additional mechanisms for particle clearance or translocation are mucociliary
19 transport and coughing (CD, 6-44, Table 6-2). Soluble components of particles may also move
20 into the circulatory system and thus throughout the body. Recent studies have also suggested
21 that ultrafine particles may be able to move directly from the lungs into the systemic circulation,
22 providing a pathway by which ambient PM exposure could rapidly affect extrapulmonary organs
23 (CD, p. 6-55).
24 In summary, new evidence from dosimetry studies has advanced our understanding of the
25 complex and different patterns of particle deposition and clearance in the respiratory tract
26 exhibited by fine particles in the accumulation mode, ultrafine particles, and thoracic coarse
27 particles. The evidence shows that all size fractions of thoracic particles can enter the *
28 tracheobronchial or alveolar regions of the respiratory system and potentially cause effects.
29 A major research need identified in the last review was the need to understand the
30 potential biological mechanisms by which deposited particles could result in the varying effects
January 2005 3-6 Draft - Do Not Cite or Quote
-------
1 observed in epidemiological studies with PM exposure. New evidence from toxicologic and
2 controlled human exposure studies has helped to identify and provide support for a number of
3. potential pathways by which particles could have biological effects, as discussed in Chapter 7 of
4 the CD. Fully defining the mechanisms of action for PM would involve description of the
5 pathogenesis or origin and development of any related diseases or processes resulting in
6 premature mortality. While the evidence summarized in the CD has provided important insights
7 that contribute to the plausibility of effects observed in community health studies, this more
8 ambitious goal of fully understanding fundamental mechanisms has not yet been reached. Some
9 of the more important findings presented therein, including those related to the cardiovascular
10 system, may be more accurately described as intermediate responses potentially caused by PM
11 exposure rather than complete mechanisms. It appears unlikely that the complex mixes of
12 particles that are present in ambient air would act alone through any single pathway of response.
13 Accordingly, it is plausible that several health responses might occur in concert to produce
14 reported health endpoints.
15 By way of illustration, Mauderly et al. (1998) discussed particle components or
16 characteristics hypothesized to contribute to PM health, producing an illustrative list of 11
17 components or characteristics of interest for which some evidence existed. The list included: 1)
18 PM mass concentration, 2) PM particle size/surface aVea, 3) ultrafine PM, 4) metals, 5) acids, 6)
19 organic compounds, 7) biogenic particles, 8) sulfate and nitrate salts, 9) peroxides, 10) soot, and
20 11) co-factors, including effects modification or confounding by co-occurring gases and
21 meteorology. The authors stress that this list is neither definitive nor exhaustive, and note that
22 "it is generally accepted as most likely that multiple toxic species act by several mechanistic
23 pathways to cause the range of health effects that have been observed" (Mauderly et al., 1998).
24 In assessing the more recent animal, controlled human, and epidemiologic information,
25 the CD developed a summary of current thinking on pathophysiological mechanisms for the
26 effects related to PM exposure. Section 7.10.1 of the CD discusses a series of potential
27 mechanisms or general pathways for effects on the heart and lung, and the CD's conclusions on
28 the evidence supporting different types of effects is briefly summarized below. The relative
29 support for these potential mechanisms/intermediate effects and their relevance to real world
30 inhalation of ambient particles varies significantly. Moreover, the CD highlights the variability
January 2005 . 3-7 - Draft - Do Not Cite or Quote
-------
1 of results that exist among different approaches, investigators, animal models, and even day-to-
2 day within studies. Nonetheless, the CD states that "Findings since 1996 have provided
3 evidence supporting many hypotheses regarding induction of PM effects; and this body of
4 evidence has grown substantially." (CD, p. 7-205). For the most part, the evidence from
5 toxicologic and controlled human exposure studies discussed below reflects the effects of fine
6 particles or fine particle constituents.
7 Direct Pulmonary Effects. Potential pathways for direct pulmonary effects include:
8 lung injury and inflammation; increased airway reactivity and asthma exacerbation; and
9 increased susceptibility to respiratory infections. The CD finds "particularly compelling"
10 evidence that PM exposure causes lung injury and inflammation. Evidence that supports
11 hypotheses on direct pulmonary effects includes toxicologic and controlled human exposure
12 studies using both sources of ambient particles and combustion-related particles. Toxicologic
13 studies using intratracheal instillation of ambient particles from various locations (e.g., St. Louis,
14 MO; Washington DC; Dusseldorf, Germany; Ottawa, Canada; Provo and Utah Valley, Utah;
15 Edinburgh, Scotland) have shown that ambient particles can cause lung inflammation and injury
16 (CD, p. 7-48). Several studies using filter extracts from Utah Valley ambient samples collected
17 .; • before, during and after the shut-down of a maj or particle-emitting facility have reported effects
18 such as increases in oxidant generation, release of cytokines such as IL-8, and evidence of
19 pulmonary injury such as increased levels of lactose dehydrogenase (CD, p 7-46,7-47).
20 Administration of residual oil fly ash (ROFA, an example of a combustion source particle type)
21 has been shown to produce acute lung injury and severe inflammation, with effects including
22 recruitment of neutrophils, eosinpphils and monocytes into the airway (CD, p. 7-60). New
23 toxicologic or controlled human exposure studies using exposure to CAPs have reported some
24 evidence of inflammatory responses in animals, as well as increased susceptibility to infections,
25 though the results of this group of studies are more equivocal (CD, p. 7-85). In vitro studies,
26 summarized in section 7.4.2 of the CD, also report evidence of lung injury, inflammation, or
27 altered host defenses with exposure to ambient particles or particle constituents. Some
28 toxicologic evidence also indicates that PM can aggravate asthmatic symptoms or increase
29 airway reactivity, especially in studies of the effects of diesel exhaust particles (CD, section
30 7.3.5). Finally, some new evidence suggests that particles can initiate neurogenic responses in
January 2005. 3-8 Draft - Do Not Cite or Quote
-------
• 1 the respiratory system, for example, several studies have indicated that some particles can
2 activate sensory nerve receptors in the airways, leading to inflammatory responses such as "
3 cytokine release (CD, section 7.4.4.4) .
4 Systemic Effects Secondary to Lung Injury. Adding to the list of direct pulmonary
5 effects, these pathways include: impairment of lung function leading to heart injury; pulmonary
6 inflammation and cytokine production leading to systemic hemodynamic effects; lung
7 inflammation leading to increased blood coagulability; and lung inflammation leading to
8 hematopoiesis effects. While more.limited than for direct pulmonary effects, some new evidence
9 from toxicologic studies suggests that injury or inflammation in the respiratory system can lead
10 to changes in heart rhythm, reduced oxygenation of the blood, changes in blood cell counts, or
11 changes in the blood that can increase the risk of blood clot formation, a risk factor for heart,
12 attacks or strokes (CD, pp..7-209 to 7-212). '
13 Effects on the Heart. In addition, potential pathways for effects on the heart include:
14 effects on the heart from uptake of particles or particle constituents in the blood; and effects on
15 the autonomic control of the heart and circulatory system. In the last review,- little or no evidence
16 was available on potential cardiovascular effects from toxicologic studies. More recent studies
17 have provided some initial evidence that particles can have direct cardiovascular effects. As
18 shown in Figure 7-1 of the CD, there are several pathways by which particle deposition in the
19 respiratory system could lead to cardiovascular effects, such as PM-induced pulmonary reflexes
20 resulting in changes in the autonomic nervous system that then could affect heart rhythm (CD, p.
21 -7-8). Also, inhaled PM could affect the heart or other organs if particles or particle constituents
22 are released into the circulatory system from the lungs; some new evidence indicates that the
23 smaller ultrafine particles can move directly from the lungs into the systemic circulation (CD, p.
24 6-55). The CD concludes that the data remain limited but provide some new insights into
25 mechanisms by which particles, primarily fine particles, could affect the cardiovascular system
26 (CD, 7-35,7-212). • *
27 The above list of potential mechanisms was developed mainly in reference to effects
28 from short-term rather than long-term exposure to PM. Repeated occurrences of some short-
29 term insults, such as inflammation, might contribute to long-term effects, but wholly different
30 mechanisms might also be important in the development of chronic responses. Some
January2005 ' 3-9 . Draft - Do Not Cite or Quote
-------
1 mechanistic evidence is available, however, for potential carcinogenic or genotoxic effects of
2 particles. Section 7.10.1 of the CD also includes a discussion of the evidence for mutagenic or
3 genotoxic effects of particles or particle constituents, concluding that "both ambient PM and
4 combustion products of coal, wood, diesel, and gasoline are mutagenic/genotoxic." (CD, p.7-
5 215).
6 While new evidence is available from studies exposing animals or humans to ambient
7 fine particles, many toxicologic and controlled human exposure studies have used exposures to
8 fine particle constituents or emission-related particles, such as fly ash or diesel exhaust particles.
9 The evidence related to particle types or components is summarized in section 7.10.2 of the CD.
10 Overall, the findings indicate that different health responses are linked with different particle
11 characteristics, and that both individual components and the complex particle mixtures appear to
12 be responsible for many biological responses (CD, p. 7-206).
13 Particles may also help carry other airborne substances into the respiratory tract, as
14 summarized in section 7.9 of the CD. Particles can take up moisture and grow in the humid
15 atmosphere of the respiratory tract, thus potentially altering the deposition and clearance patterns
16 of the particles. Water-soluble gases can be carried into the lung on particles, and delivery of
17 reactive gases such as SO2 and formaldehyde to the lower respiratory regions can be increased
18 when carried on particles since these gases would otherwise be more likely trapped in the upper
19 airways. Particles can also carry reactive oxygen species, such as hydrogen peroxide, and other
20 toxic compounds such as polynuclear aromatic hydrocarbons or allergens, into the lower
21 respiratory regions (CD, pp, 7-203, 7-204).
22 Beyond the dosimetric evidence summarized above, few studies have assessed potential
23 . biological mechanisms for effects seen with PM10.25, for either acute or chronic exposures (CD,
24 p. 9-55). However, the CD includes results from a few new toxicologic studies that assess the
25 effects of thoracic coarse particles. Section 7.4.2 of the CD includes discussion of two studies
26 that report inflammatory responses in cells exposed to ambient thoracic coarse particles collected
27 in Chapel Hill, NC, that appeared to be linked to the endotoxin content of the particles (CD, pp.
28 7-83, 7-102). A study in Japan also reported effects on immune cells with exposure to
29 resuspended coarse particles (CD, p. 7-135). Another research group exposed blood cells to
January 2005 3-10 Draft - Do Not Cite or Quote
-------
1 ambient fine and thoracic coarse particles, and reported greater effects with fine particles (CD,- p.
2 7-102). • " ' '•''.'-•'
3 Many of the newer studies use high doses (in mg or hundreds of ug), though some have
4 used doses that are close to ambient concentrations. A key consideration for evaluating the •
5 results of animal toxicologic studies is the relation between effects reported with high dose
6 exposures to animals to effects that would be expected in human populations with ambient
7 exposures. The CD presents an illustrative set of analyses evaluating the doses and responses
8 reported in human and animal studies in Appendix 7 A of the CD. In the analyses, dosimetry
9 models were used to predict doses of deposited and retained particles in various regions of the
10 respiratory system for humans and rats. In this series of analyses, the dose ratios for humans to
11 rats were quite variable across dose metrics and respiratory system regions. For example, using
12 data'from combustion particle (residual oil fly ash) exposures, the equivalent exposure ratios for
13 rats to humans in Table 7 A-8a of the CD range from about 0.1 to 16 (CD, p. 7 A-34). Using
14 particle number and surface area-based dose metrics resulted in a broader range of equivalent •
x*v 15 exposure ratios, for example, ranging from 0.008 to 1,300 for particle surface area1 (CD,-p. 7A- •
16 36). The CD also evaluated relative dose levels using data from two sets of studies in which
17 toxicologic and controlled human exposure studies used the same type of ambient particles (Utah
18 Valley dust and concentrated ambient particles). Tables 7A-1 la through 7A-1 Ic in the CD show
19 estimations for both deposited or retained doses in the alveolar and tracheobronchial regions for
20 three scenarios. In each case the differences between humans and rats is not overly large; for '
21 example, deposited doses were roughly two-'to four-fold higher for rats than for humans in
22 analyses from'inhalation exposure studies using concentrated ambient particles (CD, pp. 7Ai52,
23 7A-53). Recognizing the limitations of this small set of illustrative analyses, the CD concludes
24 that larger doses in rats may be dosimetrically equivalent to lower doses in humans, given the
25 faster particle clearance rates in rats (CD, p. 7A-62). However, the CD also observed that the
26 prediction of dose levels depends on a number of factors, and estimated equivalent exposure
27 ratios for rats and humans vary substantially (CD, 7-163). •••••'. :
28 In summary, while investigation of potential mechanisms for the effects'of particles
29 remains an important research question, new mechanistic studies provide evidence to support a
30 number of hypothesized mechanisms of action. In evaluating this new body of evidence; the CD
January 2005 • 3-11 Draft - Do Not Cite or Quote
-------
fo^--^---.:^^
u
1 more than.80 new time-series studies of the relationship between short-term exposure to PM and
2 mortality have been published, including several multi-city studies that are responsive to the •
3 recommendations from the last review (CD, p. 8-23).
4 In the last review, much consideration was given to assessing the relative roles of PM and
5 co-pollutants, acting alone and in combination, in producing the associations with adverse health
6 effects in epidemiolpgic studies. Much attention was focused on a series of analyses and
7 reanalyses using data from one U.S. city, Philadelphia, which reported associations between
8 mortality-and TSP and gaseous co-pollutants. However, it was difficult to distinguish the effects
9 of TSP from one or more gaseous co-pollutants for this single location due in part to the fact that
10 the co-pollutants were generally correlated with TSP (Samet et al., 1997; EPA, 1996a, p. 13-56).
11 Indeed, the limitations of even the most comprehensive single-city analyses precluded definitive
12 conclusions concerning the role of PM. The results of reanalyses of these data were reviewed
13 by an expert panel, the Health Effects Institute review panel, who observed that "[cjonsistent and
14 repeated observations in locales with different air pollution profiles can provide.the most
15 convincing epidemiolpgic evidence to support generalizing the findings from these models"
16 (HEI, 1997, p. 38). The summary report from this panel recommended that future research into ] x^X
17 the role of co-pollutants should improve upon the examination of multiple single-city studies by,
18 different investigators and by conducting multi-city studies, using consistent analytical
19 approaches across cities.. Consistent with these views, the 1996 CD and Staff Paper examined
20 the consistency and coherence of reported effects across studies of individual cities having
21 different pollutant mixtures, climate, and other factors. , . _
22 In this review, the CD has emphasized the results of the multi-city studies as being of
23 particular relevance. The multi-city studies combine datafrom 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
January 2005 3-14 Draft -DoNot Cite or.,Quote \
-------
1 evaluated; and (5) "publication bias" or exclusion of reporting of negative or nonsignificant
2 findings can be avoided (CD, p. 8-30).
3 The National Morbidity, Mortality and Air Pollution Study (NMMAPS) is the largest
4 available multi-city analysis, and included analyses of PM]0 effects on mortality in 90 U.S. cities
5 (Samet et al., 2000a,b; Dominici et al., 2003a). Additional, more detailed, analyses were
6 conducted in a subset of the 20 largest U.S. cities (Samet et al., 2000b). The NMMAPS study
7 was, in fact, designed to use a multi-city approach such as that recommended above (Samet et
8 al., 2000c, p; 1). A uniform methodology was used to evaluate the relationship between
9 mortality and PM10 for the different cities, and the results were synthesized to provide a
10 combined estimate of effects across the cities. The authors reported associations between total
11 and cardiorespiratory mortality and PM]0 that were robust to different modeling approaches and
12 to adjustment for gaseous co-pollutants. For total mortality, the overall risk estimate for all cities
13 is a statistically significant increase of 1.4% (using more stringent GAM) or 1.1% (using GLM)
14 per 50 [ig/m3 PM10, lagged one day (Dominici et al., 2003a; CD, p. 8-33).. Key components to
15 the NMMAPS analyses include assessment of the potential heterogeneity in effects and effects of
16 co-pollutants!, as discussed below in sections 3.4.3 and 3.6.4, respectively.
17 Another major multi-city study used data from 10 U.S. cities where every-day PM]0
18 monitoring data were available (in many areas, monitoring is done on a l-in-3 or l-in-6 day'
19 basis) (Schwartz, 2003b). The authors reported a statistically significant association between
20 PM10 and total mortality, with an effect estimate of an increase of 3.4% per 50 ng/m3 PM10 (in
21 reanalyzed GAM results) or 2.8% per 50 ug/m3 PM10 (using GLM) (Schwartz, 2003b; CD, p. 8-
22 38). The CD observes that the effect estimates from this study are larger than those reported in
23 NMMAPS, and suggests that the availability of more frequent monitoring data may partly
24 account for the differences (CD, p. 8-39).
25 . In the previous review, results for one key multi-city study were available, in which
26 associations were assessed between daily mortality and PM, using fine and thoracic coarse
27 particle measurements from six U.S. cities (the "Six Cities" study) (Schwartz, et al., 1996). The
28 authors reported significant associations for total mortality with PM2? and PMi0, but not with
29 PM10.2 5. Reanalyses of Six Cities data have reported results consistent with the findings of the
30 original study, with statistically significant increases in total mortality ranging from 2% to over
January 2005 3-15 Draft - Do Not Cite or Quote
-------
1 3% reported for results from more stringent GAM or GLM analyses using either PM2S (per 25
2 ug/m3 increment) or PM10 (per 50 ug/m3 increment), whereas PM!0.25 was not significantly
3 associated with mortality (Schwartz, 2003a; Klemm and Mason, 2003; CD, p. 8-40).
4 Using data for the eight largest Canadian cities, mortality was associated with PM25,
5 PM10, and PM10.2 5 and the effect estimates were of similar magnitude for each.PM indicator
6 (Burnett et al., 2000; Burnett and Goldberg, 2003). Using either more stringent GAM or GLM,
7 the authors reported increases ranging from 2% to 3% in total mortality for each PM indicator.
8 The association between mortality and PM2 5 generally remained statistically significant in a
9 number of analyses when gaseous co-pollutants and 0- and 1 -day lags were included in the
10 models, although in a few instances the effect estimates were reduced and lost statistical
11 significance. Associations with PM10, and PM10_2 5 did not reach statistical significance, though
12 the effect estimates were similar in magnitude to those for PM25. While the associations
13 reported with PM10.2 5 were somewhat increased in magnitude in reanalyses, they did not reach
14 statistical significance. The CD concludes that it is difficult to compare the relative significance
15 of associations with PM25 and PM10.25, but for this study, "overall, they do not appear to be
16 markedly different" (Burnett and Goldberg, 2003; CD, p. 8-42).
17 The CD also highlights results of analyses from a major European multi-city study, the
18 Air Pollution and Health: A European Approach (APHEA) study, that evaluated associations
19 between mortality and various PM measures (CD, section 8.2.2.3.3). In the analyses that
20 included data from 29 European cities, overall effect estimates of 2 to 3% increased risk of
21 mortality per 50 ug/m3 PM10 were reported; reanalysis produced essentially identical results to
22 those of the initial studies (Katsouyanni et al., 2003; CD, p. 8-47).
23 Numerous studies have been conducted in single cities or locations in the U.S. or Canada,
24 as well as locations in Europe, Mexico City, South America, Asia and Australia (Table 8A in the
25 CD). As was observed based on the more limited studies available in the last review, the
26 associations reported in the recent studies on short-term exposure to PM10 and mortality are
27 largely positive, and frequently statistically significant Overall, the CD concludes that multi-
28 city studies in the U.S., Canada, and Europe reported statistically significant associations with
29 effect estimates ranging from ~1.0 to 3.5% increased risk of total mortality per 50 ug/m3 PMi0,
30 and from 2 to over 3% increased risk in mortality per 25 ug/m3 PM25 (CD, p. 8-50). Combining
January 2005 3-16 Draft - Do Not Cite or Quote
-------
1 total mortality effect estimates from many individual-city studies with those from the multi-city
2 studies, the CD finds that they generally fall in the range of-1.0 to 8,0% per 50 ug/m3 PM10
3 (CD, p. 8-337).
4 Effect estimates from U.S. and Canadian multi-city and single-city studies are presented
5 in Figure 3-1 for associations between PM10, PM25 and PM10.25 and mortality.5 Figure 3-1 shows
6 that, for PM2 5, almost all effect estimates are positive and a number are statistically significant,
7 particularly when focusing on the results of studies with greater precision. As summarized in the
8 CD, effect estimates for total mortality from the multi-city studies range from ~1 to 3.5% per 25
9 ug/m3 PM25, and from approximately 2 to 6% per 25 ug/m3 PM25 from the relatively more
10 precise single-city studies (CD, p. 9-28). Figure 3-1 also shows effect estimates for PM10.25 that
11 are generally positive and similar in magnitude to those for PM2 5 and PM10 but for total
12 mortality, none reach statistical significance. Staff notes that on a unit mass basis, the effect
13 estimates for both PM2 5 and PM10.2 s are generally larger than those for PM10, which is consistent
14 with PM25 and PM]0.25 having independent effects (CD, p.-9-25).
15 In general, effect estimates are somewhat larger for respiratory and cardiovascular
16 mortality than for total mortality. In the NMMAPS analyses using data from the 20 largest U.S.
17 cities, the effect estimates for deaths from cardiorespiratory causes were somewhat larger than
18 those for deaths from all causes (1.6% versus 1.1% increased risk per 50 ng/m3 PMi0, using
19 GLM) (Domenici, 2003; CD, p. 8-78). In Figure 3-1, for all three PM indicators, it can be seen
20 that not only is the effect estimate size generally larger for cardiovascular mortality, but the
21 effect estimates are also more likely to reach statistical significance. This is particularly true for
22 PMj0.25, where two of the five effect estimates for cardiovascular mortality shown are positive
23 and statistically significant (Mar et al., 2003; Ostro et al., 2003). For respiratory mortality,
24 effect estimates are often larger than those for either total or cardiovascular mortality, but they
25 are often less precise, which would be expected since respiratory deaths comprise a small
5 The effect estimates in Figure 3-1 (for mortality effects) and in Figure 3-2 (for morbidity effects;
discussed below in section 3.3.2) have been plotted in order of decreasing study power, using as an indicator the
natural log of the product of the number of study days and number of health events per day.
January 2005 3-17 Draft - Do Not Cite or Quote
-------
2 S
*-£
I .a -g
2 "S B
•a ** «
M 8^
41 S-*
e^* £
•=.2 I g.
•IP!
lili
a S =
I
o
A «.
.2 M
•O « o
£ M*;- 2
2 .S i $
-••3 * *
33 g S
a u iS °
u c « y
"o •• s "5
|.ff|g
CQ *g B« g.
MS a
I
i
1
u
00
£> S & -8 .a § gt
gul o,§ a>r5 ss
»^^ giU gTL-'SCJ-'^Jj
.1
fe
-------
f
1 proportion of-total deaths. The CD concludes that effect estimates fall in the range of 3 to 7%
2 per 25 \ig/m3 PM25 for cardiovascular or cardiorespiratory mortality, and 2 to 7% per 25 ng/m3
3 PM25 for respiratory mortality in U.S. and Canadian cities. The magnitude of the effect
4 estimates for PM10_2 5 are similar to those for PM2 5, generally falling in the range of 3 to 8% for
5 cardiovascular mortality and 3 to 16% for respiratory mortality per 25 ng/m3 PMj0.25 (CD, p. 8-
6 306).
7 While some of the studies conducted in Europe, Mexico or South America use
8 gravimetric PM measurements (e.g., PM10, PM25, PM10.25), many of the non-North American
9 studies use PM indicators such as TSP, BS or COH, and the Australian studies used
10 nephelometric measures of PM. While effect estimates for different PM indicators may not be
11 quantitatively comparable, the CD observes that "many of the newly reported analyses continue
12 to show statistically significant associations between short-term (24-hr) PM exposures indexed
13 by a variety of ambient PM measurements and increases in daily mortality in numerous U.S. and
14 Canadian cities, as well as elsewhere around the world" (CD, p. 8-24). These effect estimates are
15 generally within (but toward the lower end of) the range of PM10 estimates previously reported in
16 the 1996 PM AQCD. ' .
17 As discussed in section 8.2.2.5 of the CD, associations have been reported between
18 mortality and short-term exposure to a number of PM components, especially fine particle
19 components. Recent studies have evaluated the effects of air pollutant combinations or mixtures
20 including PM components using factor analysis or source apportionment methods to link effects
21 with different PM source types (for example, combustion and crustal sources). These studies
22 have suggested that fine particles of some source types, especially combustion sources, may
j
23 contribute more to associations with mortality than other particles, such as those from crustal
24 material in fine particles (CD, p. 8-85).
25 The evidence from time-series studies is also buttressed by findings of several
26 "intervention studies" that have assessed improvement in health in areas where policy, economic
27 or regulatory changes resulted in reduced air pollutant concentrations (section 8.2.3.4 in'the CD).
28 Studies conducted in Dublin and Hong Kong reported reduced mortality risk following
29 regulations that banned the use of bituminous coal and reduced sulfur in fuel oil, respectively,
30 though it was difficult to distinguish effects of reductions in the individual pollutants.
January 2005 3-19 Draft = Do Not Cite or Quote
-------
1 Overall, the CD finds that the expanded body of evidence provides "especially strong"
2 evidence for associations between short-term exposure to thoracic particles and mortality (CD, p.
3 8-335). From the full body of multi-city and single-city studies, the CD observes that "many of
4 the newly reported analyses continue to show statistically significant associations between short-
5 term (24 h) PM exposures indexed by a variety of ambient PM measurements and increases in
6 daily mortality in numerous U.S. and Canadian cities, as well as elsewhere around the world"
7 (CD, p. 8-24).
8 3.3.1.2 Mortality and Long-term PM Exposure
9 In the 1996 PM CD, results were presented for three recent prospective cohort studies of
10 adult populations (i.e., the Six Cities, American Cancer Society (ACS), and California Seventh
11 Day Adventist (AHSMOG) studies). The 1996 CD concluded that the chronic exposure studies,
12 taken together, suggested associations between increases in mortality and long-term exposure to
13 PM (EPA, 1996a, p. 13-34). New studies discussed in the CD (section 8.2.3) include a
14 comprehensive reanalysis of data from the Six Cities and ACS studies, new analyses using
15 updated data from the AHSMOG and ACS studies, and a new analysis using data from a cohort
16 of veterans. Effect estimates from all four of these studies are provided in Appendix 3B.
17 The reanalysis of the Six Cities and ACS studies included two major components, a
18 replication and validation study, and a sensitivity analysis, where alternative risk models and
19 analytic approaches were used to test the robustness of the original analyses. The reanalysis
20 investigators replicated the original results, confirming the original investigators' findings of
21 associations with both total and cardiorespiratory mortality (Krewski et al., 2000; CD, p. 8-95).
22 In single-pollutant models, none of the gaseous co-pollutants was significantly associated with
23 mortality except SO2. The reanalyses included multi-pollutant models with the gaseous
24 pollutants, and the associations between mortality and both fine particles and sulfates were
25 unchanged in these models, except for those including SO2. SO2 is a precursor for fine particle
26 sulfates, making it difficult to distinguish effects of S02 and sulfates or fine particles (CD, p. 9-
27 37). While recognizing that increased mortality may be attributable to more than one component
28 of ambient air pollution, the reanalysis confirmed the association between mortality and fine
29 particle and sulfate exposures (Krewski et al., 2000; CD, p. 8-95).
January 2005 3-20 Draft = Do Not Cite or Quote
-------
1 The extended analyses for the ACS cohort study included follow-up health data and air
2 quality data from the new fine particle monitoring network for 1999-2000, and reported
3 significant associations between long-term exposure to fine particles (using various averaging
4 periods"for air quality concentrations) and premature mortality from all causes, cardiopulmonary
5 diseases, and lung cancer (Pope et al., 2002; CD p. 8-102). This extended analysis included the
6 use of data on gaseous pollutant concentrations, more recent data on fine particle concentrations,
7 and evaluated further the influence of other covariates (e.g., dietary intake data, occupational
8 exposure) and model specification for the PM-mortality relationship (e.g., new methods for
9 spatial smoothing and random effects models in the Cox proportional hazards model) (CD, p. 8-
10 97). The investigators reported that the associations found with sulfate and fine particle
11 concentrations were robust to the inclusion of many covariates for socioeconomic factors or
12 personal health variables (e.g., dietary factors, alcohol consumption, body mass index); however,
13 as was found in the reanalysis of the original ACS study, education level was found to be an
14 effect modifier, in that larger and more statistically significant effect estimates were reported in
15 the group with the lowest education level (Pope et al., 2002; CD, p. 8-104). In both the
16 reanalyses and extended analyses of the ACS cohort study, long-term exposure to PM10.25 was
17 not significantly associated with mortality (Krewski et al., 2000; Pope et al., 2002).
18 Other new analyses using updated data from the AHSMOG cohort included more recent
19 air quality data for PM10 and estimated PM25 concentrations from visibility data, along with new
20 health information from continued follow-up of the Seventh Day Adventist cohort (Abbey et al.,
21 1999; McDonnell et al., 2000). In contrast to the original study in which no statistically
22 significant results were reported with TSP, a significant association was reported between total
23 mortality and PM,0 for males, but not for females (CD, pp. 3-41, 3-42). Additional analyses
24 were conducted using only data from males and estimated PM2 5 and PM10.2 5 concentrations;
25 larger effect estimates Were reported for mortality with PM2 5 than with PM10.2 5, but the estimates
26 were generally not statistically significant (McDonnell et al., 2000; CD, p. 8-117). In the VA
27 cohort study, analyses were done using subsets of PM exposure and mortality time periods, and
28 the investigators report inconsistent and largely nonsignificant associations'between PM
29 exposure (includirig, depending on availability, TSP, PM,0, PM2 5, PM15 and PM]5.2 5) and
30 mortality (Lipfert et al., 2000b). . •
January 2005 3-21 Draft = Do Not Cite or Quote
-------
1 ' Based on an evaluation of all the available long-term exposure studies, the CD places
2 greatest weight on the results of the Six Cities and ACS studies. In so doing, the CD notes that
3 the Six Cities and ACS, studies (including reanalyses and extended analyses) included measured
4 PM data (in contrast with AHSMOG PM estimates based on TSP or visibility measurements),
5 have study populations more similar to the general population than the VA study cohort, and
6 have been validated through an exhaustive reanalysis (CD, pp. 8-116; 9-33).
7 One new effect reported in the extended analysis of the ACS study was a statistically
8 significant association between fine particle and sulfate concentrations and'lung cancer.
9 mortality, with a 13% increased risk of lung cancer mortality per 10 ug/m3 PM2 5, using air
10 quality data averaged across all available years (CD, p. 8-99). This effect estimate is little
11 changed and remains significant with adj ustment for covariates, random effects modeling and
12 spatial smoothing methods (CD, Figure 8-8). Also, in new analyses using updated data from the
13 AHSMOG cohort, significant associations were reported between long-term PM 10 exposure and
14 lung cancer mortality for males, but not females (CD, p. 8-317). *
15 The epidemiologic findings of associations between fine particles and lung cancer
16 mortality are supported by the results of recent toxicologic studies that have examined the
17 mutagenic potential of ambient particles. These toxicologic studies have provided evidence of
18 mutagenicity or genotoxicity with exposure to combustion-related particles or to ambient
19 particles collected in Los Angeles, Germany and the Netherlands (CD, p. 9-76). In addition, the
20 Health Assessment Document for diesel engine exhaust concludes that diesel engine exhaust,
21 one source of PM emissions, is a likely human carcinogen (EPA, 2002). On the results of the
22 new epidemiologic studies, the CD concluded "[o]verall, these new cohort studies confirm and
23 strengthen the published older ecological and case-control evidence indicating that living in an .
24 area that has experienced higher PM exposures can cause a significant increase in RR of lung
25 cancer incidence and associated mortality" (CD, p. 8-318). A number of toxicologic studies,
26 summarized in section 7.10.1 of the CD, report evidence of genotoxicity or mutagenicity with
27 particles. The CD also finds that the evidence indicates that fine particles may be more
28 mutagenic than thoracic coarse particles (CD, p. 7-214), which is consistent with the evidence
29 from epidemiologic studies. Considered with the results of toxicologic studies, the CD finds that
January 2005 3-22 Draft = Do Not Cite or Quote
-------
1 this new evidence supports the plausibility of a relationship between fine particles and lung
2 cancer mortality (CD, p. 9-78).
3 Thus, emphasizing the results from the Six Cities and ACS cohorts, the CD finds that
4 there are significant associations for mortality with long-term exposure to PM2 5. The effect
5 estimates for deaths from all causes fall in a range of 6 to 13% increased risk per 10 ng/m3 PM25,
6 while effect estimates for deaths from cardiopulmonary'causes fall in a range of 6 to 19% per 10
7 ug/m3 PM25. For lung cancer mortality, the effect estimate was a 13% increase per 10 ng/m3
8 PM15 in the results of the extended analysis from the ACS cohort (Pope etl al., 2002; CD, Table
9 8-12). In addition, based on evidence from reanalyses and extended analyses using ACS cohort
10 data, the CD concludes that the long-term exposure studies provide evidence that long-term
11 exposure to thoracic coarse particles is not associated with mortality (CD, p. 8-307).
12 '.
13 3.3.2 Morbidity
14 The epidemiologic evidence also includes associations between various indicators of PM
»
15 and a wide range of endpoints reflecting both respiratory- and cardiovascular-related morbidity
16 effects. The following sections summarize the CD's findings on PM-related morbidity effects,
17 beginning with hospital admissions and medical visits for respiratory and cardiovascular
18 diseases. Subsequent sections provide overviews of the CD's evaluation of evidence for effects
19 on the respiratory and cardiovascular systems. Effect estimates for associations between short-
20 term exposure to PM2 5 or PM10.2 s with hospitalization and medical visits from U.S. and
21 Canadian studies are presented below in Figure 3-2. Appendix 3A includes effect estimates for
22 associations with hospitalization and medical visits, as well as those for respiratory symptoms
•
23 and lung function and physiological cardiovascular effects, with short-term exposures to PM]0,
24 PM2,5 or PM10.i5 from U.S. and Canadian studies. The results for all new cardiovascular and
25 respiratory admissions/visits studies, including those using nongravimetric PM measurements
26 and studies from non-North American locations, are summarized in the CD in section 8.3, and a
27 more complete discussion of all studies is available in Appendix 8B of the CD.
28 3.3.2.1 Hospitalization and Medical Visits
29 Numerous recent studies have continued to report significant associations between short-
30 term exposures to PM and hospital admissions or emergency department visits for respiratory or
January 2005 • 3-23 Draft = Do Not Cite or Quote
-------
Q
3
O)
I
m
-------
1 cardiovascular diseases. The new studies have included multi-city analyses, numerous
2 assessments using cardiovascular admissions/visits,' and evaluation of the effects of fine and
3 thoracic coarse particles, . ' " ' ' - '
4 ' One new'multi-city study, the NMMAPS, included analyses of associations with hospital
5 admissions among the elderly,.and reported statistically significant associations between PMIO
6 and hbspital'admissions in the elderly for cardiovascular diseases, pneumonia arid chronic" " !
7 obstructive pulmonary disease (COPD) in 14 cities (Samet et al., 2000; Schwartz et al., 2003).
* ' ' .
8 Increases of 5% in hospital admissions for cardiovascular disease arid 8% and 6% in hospital
9 admissions for COPD or pneumonia; respectively, per 50 ug/m3 PM10 were reported. In the ' '
10 NMMAPS multi-city analyses on hospitalization for respiratory and cardiovascular diseases,
11 effect estimates with PM10 were not correlated with city-specific correlations between PM,0 and
12 co-pollutant levels, which the authors conclude indicates a lack of confounding by co-pollutants
13 (CD, p. 8-146, 8-175).
14 . Numerous single-city studies have also'been published that report associations between
15 short-term PM exposure and hospitalization or rriedical visits for respiratory diseases. The effect
16 estimates from these studies generally fall in a range of 5 to 20% increased risk per 50 ng/m?' '
17 PM10, with somewhat higher estimates for~asthma visits (CD, p. 8-193). The findings from
18 studies of medical visits for respiratory diseases offer new evidence of acute respiratory effects '
19 with exposure to ambient PM (the studies generally used PM10) that provides new insight into
20 the scope of respiratory morbidity (CD, p. 9-180). ' ' ' ' '«
21 Figure 3 -2 shows associations between PM2 5 and hospitalization or emergency; room'
22 visits for the general category of respiratory diseases that are all positive and statistically
23 significant, while the results for individual disease categories (COPD, pneumonia, and asthma)
24 are less consistent, perhaps due to smaller sample sizes for the specific categories. Associations
25 with the general category of cardiovascular diseases are also all positive and statistically •
26 significant or nearly so, but again the results for specific diseases (ischemic heart disease, -
27 dysrhythmia, congestive heart disease or heart failure, and stroke) are positive but often not
28 statistically significant. Similarly, associations between hospital admissions for respiratory and'
29 cardiovascular diseases and PM10.2 5 are generally positive and, as evident in Figure3-2, the more
30 precise estimates are statistically significant. Overall, the CD finds that excess risks for
January 2005 3-25 Draft-Do Not Cite or Quote
-------
0
1 , The CD finds that the recent epidemiologic findings are consistent with those of the
2 previous review in showing associations with both respiratory symptom incidence and decreased
3 lung function (CD, p. 9-70). PM10 and 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, and a few new studies reported associations between PM10.2 5 and
6 respiratory morbidity. The findings from studies of physicians' office visits for respiratory
7 diseases offer new evidence of acute respiratory effects with exposure to ambient PM that is
8 coherent with evidence of increased respiratory symptoms and admissions/visits to the hospital
9 or emergency.room for respiratory disease. .While urging caution in interpreting the findings of
10 the toxicologic studies where higher doses were used, the CD concludes that "[t]he fact that
11 instillation of ambient PM collected from different geographical areas has been shown to cause
12 pulmonary inflammation and injury tends to support epidemiologic studies that report increased
13 PM-associated respiratory effects living in some of the same geographical areas" (CD, p, 7-48).
14 3.3.2.3 Effects on the Respiratory System from Long-term Exposures
15 .In the last review, several studies had reported that long-term PM exposure was linked
16 with increased respiratory disease and decreased lung function. One study, using data from 24 ,
17 U.S. and Canadian cities.("24 Cities" study), reported associations with these effects and long-
18 term exposure to fine particles or acidic particles, but not with-PM]0 exposure (Dockery et al.,
19 1996; Raizenne et al., 1996). The 1996 Staff Paper included further evaluation of the evidence
20 that indicated no relationship between lung function decrements and long-term exposure to
21 thoracic coarse particles (EPA, 1996b, p. V-67a). , . . ..
22 Several new epidemiologic-analyses have been conducted on long-term pollutant
23 exposure effects on respiratory symptoms or lung function in the U.S.; numerous European,
24 Asian, and Australian studies have also been published. In the U.S., studies have been based on
25 data from two cohorts, a cohort of schoolchildren in 12 Southern California Communities and an
• , ' • •. .. ...... • • >
26 adult cohortof Seventh Day Adventists (AHSMOG). Results for the new studies, together with
27 the findings available in the last review, are presented in Appendix 3B.
28 In general, these studies have indicated that long-term exposure to PM, for both PM10 or
29 PM2 5, is associated ,with reduced lung function growth and increased risk of developing chronic
30 respiratory illness (CD, p. 8-215). In section 8.3.3.2.2, the CD describes results from the
January 2005 , .3-28 Draft - Do Not Cite or Quote L.J
-------
t
1 Southern California cohort, where significant decreases in lung function growth were associated
2 with increasing exposure to PM10, PM2.5 and PMi0.2 5 in one analysis (Gauderman et al., 2000),
3 while in a second group of children recruited in this cohort there were decreases in lung function
4 growth with long-term exposure to PM]0 and PM2 5 (PMi0.2 s data were not included in this study)
5 but the results were generally not statistically significant (Gauderman et al., 2002). In an
6 analysis of cohort participants who moved during the course of the study, those who moved to
7 areas with lower PM concentrations (using PM10 as the indicator) showed increased lung
8 function growth, whereas lung function growth decreased in the group of children who moved to
9 areas with high pollution levels (Avol et al., 2001; CD, p. 8-213). A number of long-term
. 10 studies of respiratory effects also have been conducted in non-North American countries, and
• 11 many report significant associations between indicators of long-term PM exposure and either
12 decreases in lung function or increased respiratory disease prevalence (Table 8-B8 of the CD).
13 Considered together, the CD finds that the long-term exposure studies on respiratory
14 morbidity reported positive and statistically associations between fine particles or fine particle
15 components and lung function decrements or chronic respiratory diseases, such as chronic
16 bronchitis (CD pp. 8-313, 8-314). The CD observes that little evidence is available on potential
17 effects of long-term to exposure to PM]0.2 5 (CD pp. 8-313, 8-314); one analysis from a Southern
18 California cohort suggests a link between decreased lung function and long-term PM10.2 5
19 exposure, but an earlier report from the 24 Cities study finds no such associations.
20 33.2.4 Effects on the Cardiovascular System
21 In contrast with the limited information available in the previous review, the CD observes
22 that new toxicologic and epidemiologic studies provide much more evidence of effects on the
23 cardiovascular system with short-term exposures to PM (CD, p. 9-67). These new findings help
24 to shed light on biological mechanisms that underlie associations between short-term PM
25 exposure and cardiovascular mortality and hospitalization that have been reported previously.
26 The CD also observes that, while epidemiologic studies have shown associations between long-
27 term exposure to particles, especially fine particles, and cardiovascular mortality, only limited
28, evidence is available on potential cardiopulmonary morbidity responses to long-term PM
29 exposure, or mechanisms underlying such responses (CD, p. 9-69).
t
January 2005 • ' 3-29 Draft - Do Not Cite or Quote
-------
1 Epidemiologic studies have reported associations between short-term exposures to
2 ambient PM (often using PM10) and measures of changes in cardiac function such as arrhythmia,
3 alterations in electrocardiogram (ECG) patterns, heart rate or heart rate variability changes, and
4 incidence of myocardial infarction (CD, p. 8-166). Recent studies have also reported increases
5 in blood components or characteristics such as increased levels of C-reactive protein and
6 fibrinogen (CD, p. 8-169). Several of these studies report significant associations between short-
7 term PM2 5 exposures and cardiovascular health indicators. Only one of the new set of studies
8 included PM,0_2 5, in which significant associations were reported between onset of myocardial
9 infarction and short-term PM25 exposures but not with PM10_25 exposures (Peters et al., 2001).
10 As noted in section 3.2, a number of toxicologic and controlled human exposure studies
11 have reported some similar cardiovascular responses with exposure to different types of
12 ' particles. In section 9.2.3.2.1, the CD summarizes evidence from both epidemiologic and
13 toxicologic studies on subtle changes in cardiovascular health. These changes include increased
14 blood fibrinogen and fibrin formation, certain ECG parameters (e.g., heart rate variability or
15 HRV), and vascular inflammation. The CD notes that vascular inflammation induces release of
16 C-reactive proteins and cytokines that may cause further inflammatory responses which, on a
17 chronic basis, could lead to atherosclerosis.
18 ' Where a series of studies have been conducted in the same location, these studies can
19 provide evidence for coherence of effects, linking results from different study types for exposure
20 to PM in the same airshed. As discussed in the CD, in Boston, epidemiologic associations were
21 reported between PM2.5 and incidence of myocardial infarction, increases in recorded discharges
22 from implanted cardiovertex defibrillators, and decreases in HRV measures. Toxicologic.studies
23 in Boston, using PM2 5 CAPs exposures in dogs, also suggested changes in cardiac rhythm with
24 PM25 mass and changes in blood parameters with certain PM2 5 components (CD, p. 9-68, 9-69).
25
26 3.3.3 Developmental effects
27 Some new evidence is available that is suggestive of adverse effects of exposure to PM
28 and gaseous co-pollutants on.prenatal development, including both mortality and morbidity
29 effects. Several recent studies have shown significant associations between PM10 concentration
30 averaged over a month or a trimester of gestation and risk of intrauterine growth reduction
January 2005 3-30 Draft - Do Not Cite or Quote
t
-------
1 (IUGR) and low birth weight. In addition, several new studies have suggested that infant
2 mortality may be associated with exposure to PM and gaseous co-pollutants during gestation:
3 The CD concludes that these effects are emerging as potentially more important than was"
4 appreciated in the 1996 CD, but the evidence is still preliminary regarding these effects (CD,
5 pp. 8-347).
6 ••••'•
7 3.3.4 Summary
8 In summary., the CD finds that the many new available studies build upon what was
9 previously known, reporting associations between PM exposure, using various PM indicators,
10 with a broad range of cardiovascular and respiratory health endpoints (CD, p. 9-23). The new
11 studies support findings from the last review on associations between PM and cardiorespiratory
12 mortality, hospitalization and emergency department visits for respiratory diseases, respiratory
-»u
13 symptoms and decreased lung function. Recent studies also broaden the range of health effects
14 associated with exposure to PM. Evidence for respiratory effects is expanded with studies
«15 showing associations with visits to physicians or clinics for respiratory illnesses. New evidence
16 is available to link PM exposure, especially fine particles, with effects on the cardiovascular
17 system, including changes in physiological indicators or biomafkers for cardiovascular health.
18
19 3.4 INTEGRATIVE ASSESSMENT OF HEALTH EVIDENCE
20 In Chapter 9, the CD assesses the new health evidence, integrating findings from
21 epidemiologic studies with experimental (e.g., dosimetric and toxicologic) studies, to make
22 judgments about the extent to which causal inferences can be made about observed associations
23 between health endpoints and various indicators or constituents of ambient PM, acting alone
24 and/or in combination with other pollutants. In evaluating the evidence from epidemiologic
25 studies in section 9.2.2, the CD focuses on well-recognized criteria, including (1) the strength of
26 reported associations; (2) the robustness of reported associations to the use of alternative model
27 specifications, potential confounding by co-pollutants, and exposure misclassification related to
28 measurement error; (3) the consistency of findings in multiple studies of adequate power, and in
29 different persons, places, circumstances and times; (4) temporality between exposure and
• 30 observed effects; (5) the nature of concentration-response relationships; and (6) information
January 2005 3-31 Draft - Do Not Cite or Quote
-------
1 from so-called natural experiments or intervention studies (CD, p. 9-23). Integrating more
2 broadly across epidemiologic and experimental evidence in section 9.2.3, the CD focuses on the
3 coherence and plausibility of observed PM-related health effects to reach judgments about
4 causality.
5 The following discussion summarizes the conclusions and judgments from the CD's
6 integrative assessment, focusing first on the strength, robustness, and consistency of the
7 epidemiologic evidence, and ending with a focus on the CD's assessment of coherence and
8 biological plausibility of PM-related health effects. Other related issues, including temporality
9 of effects and the form of PM concentration-response relationships, are discussed below in
10 section 3.6, with a focus on how these issues affect the use of epidemiologic results in the
11 quantitative risk assessments discussed in Chapter 4.
12 .
13 3.4.1 Strength of Associations
14 Considering the magnitude, statistical significance, and the degree of precision of the
15 effect estimates derived from epidemiologic analyses, the CD finds that the results from recent
16 studies expand and support epidemiologic evidence that was found to be "fairly strong" in the
17 last review (EPA, 1996a, p. 13-92). From the short-term exposure studies, the CD concludes
18 that the "epidemiologies! evidence is strong" for associations between PM2 5 and PMi0 and total
19 or cardiovascular mortality (CD, p. 9-32). Associations between PM10.25 and mortality are
20 similar in magnitude, but less precise, than those for PM2 5 or PM10; the CD finds this evidence
21 "not as strong" but suggestive of associations with mortality (CD, p. 9-32). For both PM2 5 and
22 PM10.2 5 there is a series of positive associations with hospitalization and emergency department
23 visits for cardiovascular or respiratory diseases; many are statistically significant, but the
24 associations with PM10_2 5 are somewhat less precise than those for PM2 5 (CD, p. 9-29). Studies
25 of respiratory symptoms or lung function changes show associations with both fine and thoracic
26 coarse particles (CD, p. 8-343), while the studies of more subtle cardiovascular health outcomes
27 have shown associations with fine, but not thoracic coarse particles. Taken together, the CD
28 concludes that there is strong epidemiologjcal evidence linking short-term exposures to fine
29 particles with a range of cardiorespiratory morbidity and mortality effects. The more limited
January 2005 3-32 Draft - Do Not Cite or Quote
t
-------
1 evidence on effects of PM,0.2.5 is suggestive of both mortality and morbidity effects, with greater
2 strength in the evidence for morbidity, especially respiratoiy morbidity.
3 For long-term exposures, the evidence supports associations between PM2 5 and mortality
4 for cardiovascular and respiratory diseases and lung cancer, as well as the development of
5 chronic respiratory illness and decreased lung function (CD, p. 9-34). For PM,0.2 5, available
6 studies provide evidence of the absence of associations with mortality. Since long-term
7 exposure morbidity studies have generally not included PM10.2 5 data, no conclusions can be
8 drawn regarding long-term exposure to PM10.2 5 and morbidity effects (CD, p. 9-34).
9
10 3.4.2 Robustness of Associations
11 In section 9.2.2.2, the CD evaluates the robustness of epidemiologic associations in part
12 by considering the effect of differences in statistical model specification, potential confounding
13 by co-pollutants' and exposure error on'PM-health associations. The 1996 CD included an
14 assessment of evidence then available on these issues, and concluded that the effects observed in
• 15 epidemiologic studies "cannot be wholly attributed to" issues such as confounding by co-
16 pollutants, differing model specifications, or measurement error (EPA, 1996a, p. 13-92). These
17 issues have been further evaluated in many new studies available in this review.
18 As discussed below in section 3.6.3, the CD assesses the findings of studies that
19 evaluated alternative modeling strategies, with a particular focus on the recent set of analyses to
20 address issues related to the use of GAM in time-series epidemiologic studies. The reanalyses
21 included the use of alternative statistical 'models and methods of control for time-varying effects,
22 such as weather or season. In the results of these reanalyses, some studies showed little change
23 in effect estimates, while others reported reduced effect estimate size, though the CD observes
24 that the reductions were often not substantial (CD, p. 9-35). Overall, the CD concludes that
25 associations between short-term exposure to PM and various health outcomes are generally
26 robust to the use of alternative modeling strategies, though further evaluation of alternative
27 modeling strategies is warranted (CD, p. 9-35). In addition, the reanalysis and extended analyses
28 of data from prospective cohort studies have shown that reported associations between mortality
/ •
29 and long-term exposure to fine particles are robust to alternative modeling strategies, as
• 30 discussed below in section 3.6.3.
January 2005 3-33 Draft - Do Not Cite or Quote
-------
1 . The CD also included extensive evaluation of the sensitivity of PM-health responses to
2 confounding by gaseous co-pollutants, as discussed in detail in section 8.4,3 of the CD, and more
3 briefly below in section 3.5.6. In the new multi-city studies, as well as many of the single-city
4 studies, health outcome associations with short-term exposures to PMi0 PM2 5 and PM10_2 5 are
5 little changed in multi-pollutant models including one or more of the gaseous co-pollutants (CD,
6 p. 8-253). However, in some single-city analyses, PM-health outcome associations were
7 attenuated in multi-pollutant models; the CD observes that collinearity between co-pollutants can
8 make interpretation of multi-pollutant models difficult (CD, p. 8-253). Overall, the CD
9 concludes that these studies indicate that effect estimates for associations between mortality and
10 morbidity and various PM indices are robust to confounding by co-pollutants (CD, p. 9-37).
11 Finally, as discussed in section 3.6.2, anumber of recent studies have evaluated the
12 influence of exposure error on PM-health associations. Exposure error includes both
13 consideration of measurement error, and the degree to which measurements from an individual
14- monitor reflect exposures to the surrounding community. Several studies have shown that fairly
15 extreme conditions (e.g., very high correlation between pollutants and no measurement error in
16 the "false" pollutant) are needed for complete "transfer, of causality" of effects from one
17 pollutant to another (CD, p. 9-38). In comparing fine and thoracic coarse particles, the CD
18 observes that exposure error is likely to be more important for associations with PM10.2 5 than
19 with PM25, since there is generally greater error in PM]0.25 measurements, PM,0.25
20 concentrations are less evenly distributed across a community, and less likely to penetrate into
21 buildings (CD, p. 9-38). Therefore, while the CD concludes that associations reported with
22 PM10 PM2 5 and PM10.2 5 are generally robust, the CD recognizes that factors related to exposure
23 error may result in reduced precision for epidemiologic associations with PM1(W5 (CD, p. 9-46).
24
25 3.4.3 Consistency
26 The 1996 CD reported associations between short-term PM exposure and mortality or
27 morbidity from studies conducted in locations across the U.S. as well as in other countries, and
28 concluded that the epidemiologic data base had "general internal consistency" (EPA, 1996a, p.
29 13-30). This epidemiologic data base has been greatly expanded with numerous studies
30 conducted in single locations, as well as several key multi-city studies. As described above, the
January 2005 3-34 Draft - Do Not Cite or Quote
t
-------
•
1 CD finds that the epidemiologic studies generally report positive and often statistically
2 significant associations with various cardiorespiratory health outcomes. The larger body of
3 evidence also has shown more variability in effect estimate size for a given health outcome than
4 was apparent in the last review.
5 New multi-city studies have allowed evaluation of consistency in effect estimates across
6 geographic locations, vising uniform statistical modeling approaches. In the NMMAPS results,
7 effect estimates for many individual cities exhibited wide confidence ranges, with varied etTect
8 estimate sizes, that suggested potentially more heterogeneity in effect estimates across cities than
9 had been seen with single-city studies in the last review. However, the authors observed that
•\
1 0 there was no statistically significant heterogeneity across the effect estimates in the NMMAPS
1 1 analyses (Samet et al., 2000; Dominici et al., 2003a). The Canadian multi-city study also
12 reported some limited evidence suggesting heterogeneity in responses for PMZ5 and PM^s in
13 the reanalysis to address GAM questions; whereas there been no evidence of heterogeneity in
14 initial study findings (Burnett and Goldberg, 2003; CD, p. 9-39). Finally, in the European multi-
1 5 city, there were differences seen between effect estimates from eastern and western European
16 cities in initial analyses, but these differences were less clear with reanalysis to address GAM
17 issues (Katsouyanni et al., 2003). Overall, the new multi-city study results suggest that effect
1 8 estimates differ from one location to another, but the extent of heterogeneity is not clear.
1 9 The CD discusses a number of factors that would be likely to cause variation in PM-
20 health outcomes in different populations and geographic areas in section 9.2.2.3. The CD
21 recognizes that differences might well be expected in effects across locations, and discusses
22 investigation of a number of factors that appeared to be associated with variation in effect
23 estimates, including indicators of exposure to traffic-related pollution and climate-related
24 increases in exposure to ambient pollution (CD, p. 9-39). Other factors might also be expected
25 to cause variation in observed effects between locations, including population characteristics that
26 affect susceptibility or exposure differences, distribution of PM sources, or geographic features
27 that would affect the spatial distribution of PM (CD, p. 9-41). In addition, the CD observes that
28 NMMAPS, while advantageous in including data from many different locations with different
29 climates and pollutant mixes, included many locations for which the sample size (i.e., population
30 size and PM10 data) was inherently smaller for a given study period (CD, p. 9-40). The Canadian
January 2005 3-35 Draft - Do Not Cite or Quote
-------
1 8-city study, as well, used PM'data from a monitoring network that operated primarily on a 1-in-
2 6 day collection schedule, although the data were available for a long time period. In general,
3 use of data collected on every sixth day results in reduced statistical power, resulting in less
4 precision for estimated effect estimates for the individual cities and increased potential
5 variability in results (CD, p. 9-40).
6 Overall, the CD finds that "[fjocusing on the studies with the most precision, it can be
7 concluded that there is much consistency in epidemiologic evidence regarding associations
8 between short-term and long-term exposures to fine particles and cardiopulmonary mortality and
9 morbidity." (CD, p. 9-47). The CD also concludes that for short-term exposure to thoracic
10 coarse particles, there is some consistency in effect estimates for hospitalization for
11 cardiovascular and respiratory causes, though fewer studies are available on which to make such
12 an assessment (CD, p. 9-47).
13
14 3.4.4 Coherence and Plausibility
15 Section 9.2.3 of the CD integrates and evaluates evidence from the different health
16 disciplines to draw conclusions regarding the coherence of effects observed in the cardiovascular
17 and respiratory systems, as well as evidence for biological plausibility of these effects. The CD
18 finds that progress has been made in substantiating and expanding epidemiologic findings on ,
19 cardiovascular- and respiratory-related effects of PM, and in obtaining evidence bearing on the
20 biological plausibility of observed effects and potential mechanisms of action for particles (CD,
21 p. 9-49).
22 As was concluded in the previous review, in considering evidence from epidemiologic
23 studies using PM10 and other PM indicators the CD finds coherence for effects on the
24 cardiovascular and respiratory systems. Figures 8-24 through 8-28 of the CD show effect
25 estimates for associations between short-term exposures to PM10 and a range of cardiovascular
26 and respiratory health endpoints from within the same geographic location. This evidence from
27 epidemiologic studies in one location provides some broad support for coherence of effects
28 related to PM. In addition, the new series of toxicologic and controlled human exposure studies
29 using ambient particles (primarily PM10) collected in Utah Valley show inflammatory effects mat
30 are consistent with evidence of respiratory effects from the epidemiologic studies (CD, p. 9-71).
January 2005 -3-36 Draft - Do Not Cite or Quote
t
t
-------
1 Considering'epidemiologic evidence for PM2 5, the CD finds that epidemiologic studies
2 report associations with a broad range of effects on the cardiovascular and respiratory systems,
3 primarily from short-term exposure studies, but also supported by associations reported for long-
4 term fine particle exposure with cardiovascular mortality (CD, pp. 9-67). As described briefly in
5 section 3.2 above, and in more depth in Chapter 7 of the CD, the findings of newtoxicologic and
6 controlled human exposure studies, while still limited, support a number of potential biological
7 mechanisms or pathways for PM-related effects, and this evidence is largely from studies of fine
8 particles or fine particle components. The experimental and epidemiologic evidence together
9 support the biological plausibility of observed effects on the cardiovascular system (CD, p. 9-
10 70).. In addition, the CD highlights evidence from a series of epidemiologic and toxicologic
11 studies using ambient PM2 5 exposures in Boston that provide evidence of coherence in effects on
12 the cardiovascular system (CD, pp. 9-68, 9-69). The CD observes: "While many research
13 - questions remain, the convergence of evidence related to cardiac health from epidemiologic and
14 toxicologic studies indicates both coherence and plausibility in this body of evidence." (CD, p.
• 15 9-78). In the last review, evidence was available suggesting coherence of effects on the
16 respiratory system, and the CD finds that new epidemiologic and toxicologic studies expand
17 upon that knowledge, particularly for PM2 5 (CD, p. 9-74). In locations where epidemiologic
18 studies have been conducted, toxicologic or controlled human exposure studies using exposures
19 to concentrated ambient particles have shown effects related to lung inflammation, though
20 minimal effects on lung function have been reported (CD, p. 9-72).
21 As was true in the last review, there is some coherence in epidemiologic evidence linking
22 long-term exposure to fine particles with mortality and effects on the respiratory system.
23 However, toxicologic studies that are currently available have generally not studied effects of
24 long-term or chronic exposures to air pollution, so for the most part, no conclusions can be
25 drawn regarding biological plausibility of observed effects with long-term PM2 5 exposures (CD,
26 ' p. 9-69). However, for lung cancer, the CD summarizes evidence that supports coherence and
27 plausibility in the associations reported between long-term exposures to fine particles and lung
28 cancer mortality. Toxicologic evidence of mutagenicity or genotoxicity of particles lends
29 coherence and plausibility to evidence from epidemiologic studies linking long-term exposure to
• 30 fine particles with lung cancer mortality (CD, p. 9-76).
January 2005 3-37 Draft - Do Not Cite or Quote
-------
1 ' Less information is available to allow conclusions to be drawn about coherence or
2 plausibility for associations with PM10.2 5. Based on the epidemiologic evidence discussed
3 previously, the CD concludes that the results are suggestive of associations between short-term
4 exposure to PM10.2 5 and morbidity effects, including data on hospitalization for respiratory
5 diseases as well as increased respiratory symptoms (CD, p. 9-90). Only limited evidence is
6 available from toxicologic studies of PM10.2 5, as noted in section 3.2, though the available
7 evidence does provide some coherence for effects on the respiratory system As discussed
8 above, fractional deposition to the tracheobronchial region is greatest for thoracic coarse .
9 particles in the size range of 4 to 6 nm (CD, p. 6-109). This would be consistent with
10 epidemiological evidence linking PM10.25 with respiratory morbidity, such as increased
11 respiratory symptoms or risk of hospitalization for asthma. In addition, as observed in the CD,
12 reduced precision in PM10_2 5 effect estimates may be heavily influenced by the increased error in
13 PM10_25 measurements and exposure error related to greater spatial variability and reduced
14 penetration indoors, thus larger standard errors would be expected for associations with PM10.2S
15 than for either PM10 or PM25 (CD, p. 9-91).
16
17 3.4.5 Summary
18 The new evidence from epidemiologic studies builds upon the conclusions of the last
19 review regarding the strength, robustness.and consistency of the evidence. While uncertainties
20 remain and the new studies raise some new questions, the CD concludes:
21 In conclusion, the epidemiological evidence continues to support likely causal
22 associations between PM2 5 and PM10 and both mortality and morbidity from
23 cardiovascular and respiratory diseases, based on an assessment of strength, robustness,
24 and consistency in results. For PM10.25, less evidence is available, but the studies using
25 short-term exposures have reported results that are of the same magnitude as those for
26 PM10 and PM25, though less often statistically significant and thus having less strength,
27 and the associations are generally robust to alternative modeling strategies or
28 consideration of potential confounding by co-pollutants. (CD, p. 9-48).
29 Much more evidence is now available related to the coherence and plausibility of effects
30 than in the last review. For short-term exposures, the CD finds that the integration of evidence
31 from epidemiologic and toxicologic studies indicates both coherence and plausibility of effects
32 on the cardiovascular and respiratory systems, particularly for fine particles (CD, p. 9-78). Also,
January 2005 3-38 Draft - Do Not Cite or Quote
I
-------
1 there is evidence supporting coherence and plausibility for the observed associations between
2 long-term exposures to fine particles and lung cancer mortality (CD, p. 9-79). The smaller body
3 of evidence on thoracic coarse particles, especially 'the limited evidence from toxicologjc studies,
4 provides only limited evidence of coherence for effects of thoracic coarse particles.
5 Epidemiologic and dosimetric evidence, along with limited support from toxicologic studies,
6 support associations between PM10.25 and the respiratory system, with less evidence available on
7 cardiovascular effects.
8 Finally, the evaluation of these criteria leads the CD to draw conclusions regarding
9 causality of effects seen with fine or with thoracic coarse particles. Overall, the CD concludes
10 that the available evidence supports the general conclusion that PM25 or fine particle components
11 are "likely causally related to cardiovascular and respiratory mortality and morbidity" (CD, p. 9-
12 79). For PM10.2 5, the more limited body of evidence is suggestive of causality between short-
13 term (but not long-term) exposures and various mortality and morbidity effects, with stronger
14 evidence for associations with morbidity (CD, p. 9-79, 9-80).
15 - •'.•'"••"'
16 3.5 PM-RELATED IMPACTS ON PUBLIC HEALTH
17 The following discussion draws from sections 9.2.4 and 9.2.5 of the CD to characterize
18 subpopulations potentially at risk for PM-related effects and potential public health impacts
19 associated with exposure to ambient PM. In particular, the potential magnitude of at-risk
20 population groups is discussed, along with other key considerations related to impacts on public
21 health, such as the concept of "mortality displacement" or "harvesting."
22
23 3.5.1 Potentially Susceptible and Vulnerable Subpopulations
24 The CD summarizes information on potentially susceptible or vulnerable groups in
25 section 9.2.4, As described there, the term susceptibility refers to innate (e.g., genetic or
26 developmental) or acquired (e.g., personal risk factors, age) factors that make individuals more
i
27 likely to experience effects with exposure to pollutants. A number of population subgroups
28 have been identified as potentially susceptible to health effects as a result of PM exposure,
29 including people with existing heart and lung diseases, including possibly diabetes, older adults
30 and children. In addition, new attention has been paid to the concept of some population groups
January 2005 3-39 Draft - Do Not Cite or Quote
-------
1 having increased vulnerability to pollution-related effects due to factors including socioeconomic
2 status (e.g., reduced access to health care or low socioeconomic status) or particularly elevated
3 exposure levels, such as residence near sources such as roadways (CD, p. 9-81). Most available
4 studies have used PM10 or other measures of thoracic particles, with little specific evidence on
5 potential susceptibility to effects of PM25 or PM]0_25.
6 . A good deal of evidence indicates that people with existing heart or lung diseases are
7 more susceptible to PM-related effects. In addition, new studies have suggested that people with
8 diabetes, who are at risk for cardiovascular disease, may have increased susceptibility to PM
9 exposures. This body of evidence includes findings from epidemiologic studies that associations
10 with mortality or morbidity are greater in those with preexisting conditions, as well as evidence
11 from toxicologic studies using animal models of cardiopulmonary disease (CD, section 9.2.4.1).
12 Two age groups, older adults and the very young, are also potentially at greater risk for
13 PM-related effects. Epidemiologic studies have generally not shown striking differences
14 between adult age groups. However, some epidemiologic studies have suggested that serious
15 health effects, such as premature mortality, are greater among older populations (CD, p. 8-328).
16 In addition, preexisting respiratory or cardiovascular conditions are more prevalent in older
17 adults than younger age groups; thus there is some overlap between potentially susceptible
18 groups of older adults and people with heart or lung diseases.
19 Epidemiologic evidence has reported associations with emergency hospital admissions
20 for respiratory illness and asthma-related symptoms in children (CD, p. 8-328). The CD also
21 observes that several factors may make children more susceptible to PM-related effects,
22 including the greater ventilation per kilogram body weight in children and the fact that children
23 are more likely to be active outdoors and thus have greater exposures (CD, p. 9-84). In addition,
24 the CD describes a limited body of new evidence from epidemiologic studies for potential PM-
25 related health effects in infants, but concludes that the available new results are too mixed to
26 allow any clear conclusions to be drawn,(CD, p. 8-335).
27 The CD also discusses other potentially susceptible groups for which less evidence is
28 available. Gender is a potential factor, and there are suggested differences in epidemiologic
29 study results, but the findings are not always consistent (CD, section 9.2.4.4). There is some new
January 2005 3-40 Draft - Do Not Cite or Quote
-------
1 suggestive evidence on genetic susceptibility to air pollution, but no conclusions can be drawn at
2 this time (CD section 9.2.4.3).
3 Finally, there is some new evidence from epidemiologic studies that people from lower
4 socioeconomic strata, or who have greater exposure to sources such as roadways, may be more
5 vulnerable to PM exposure. Such population groups would be considered to be more vulnerable
6 to potential effects on the basis of socioeonomic status or exposure conditions, as distinguished
7 from susceptibility due to biological or individual health characteristics (CD, section 9.2.4.5).
8 In summary, there are several population groups that may be more susceptible or
9 vulnerable to PM-related effects. These groups include those with preexisting heart and lung
10 diseases, older adults and children. The available evidence does not generally allow distinctions
11 to be drawn between PM2 5 and PM10.2 5.
12
13 3.5.2 Potential Public Health Impact
14 As summarized above, there are several populations groups that may be susceptible or
15 vulnerable to effects from exposure to PM. The CD provides estimates of the size of population
16 subgroups, such as young children or older adults, and people with prevalent heart or lung
17 diseases (CD, section 9.2.5.1) that are the subpopulations considered to be likely susceptible to
18 the effects of PM exposure. As shown in Table 9-4 of the CD, approximately 22 million people,
19 or 11 % of the U. S. population, have received a diagnosis of heart disease, about 20% of the
20 population have hypertension and about 9% of adults and 11% of children in the U.S. have been
21 diagnosed with asthma. In addition, about 26% of the U.S. population are under 18 years of age,
22 and about 12% are 65 years of age or older (CD, p. 9-89). The CD concludes that combining
23 fairly small risk estimates and small changes in PM concentration with large groups of the U.S.
24 population would result in large public health impacts (CD, p. 9-93).
25 These health statistics'also generally illustrate increasing frequency of less serious health
26 outcomes that would be expected in a "pyramid of effects." In general, many PM-health studies
27 have used the more severe outcome measures for which data are readily available, such as
28 mortality or hospitalization. Incidence or frequency would be expected to increase in the
29 population for less severe effects along the spectrum of severity, for example, from
January 2005 3-41 Draft - Do Not Cite or Quote
-------
1 cardiovascular mortality to the subtle measures of cardiovascular health, such as changes in heart
2 rhythm or increased levels of C-reactive protein. • •
3 One issue that is important for interpreting the public health implications of the
4 associations reported between mortality and short-term exposure to PM is whether mortality is
5 occurring only in very frail individuals (sometimes referred to as "harvesting"), resulting in loss
6 of just a few days of life expectancy. A number of new analyses are discussed in the CD
7 (section 8.4.10.1) that assess the likelihood of such "harvesting" occurring in the short-term
8 exposure studies. Overall, the CD concludes from the time-series studies that there appears to be
9 no strong evidence to suggest that short-term exposure to PM is only shortening life by a few
.10 days (CD, p. 8-329).
11 In addition to evidence from short-term exposure studies discussed above, one new report
12 used the mortality risk estimates from the ACS prospective cohort study to estimate potential
13 loss of life expectancy from PM-related mortality in a population. The authors estimated that the
14 loss of population life expectancy associated with long-term exposure to PM25 was substantial,
15 on the order of a year or so (CD, p. 9-94). Taken together, these results suggest that exposure to
16 ambient PM can have substantial public health impacts (CD, p. 9-93).
17
18 3.6 ISSUES RELATED TO QUANTITATIVE ASSESSMENT OF EPIDEMIOLOGIC
19 EVIDENCE
20 The 1996 CD included extensive discussions of methodological issues for epidemiologic
21 studies, including questions about model specification or selection, co-pollutant confounding,
22 measurement error in pollutant measurements, and exposure misclassification. Based on
23 information available in the last review, the 1996 PM CD concluded that PM-health effects
24 associations reported in epidemiologic studies were not likely an artifact of model specification,
25 since analyses or reanalyses of data using different modeling strategies reported similar results
t
26 (EPA 1996a, p. 13-92). Little information was available at that time to allow for evaluation of
27 these and other related methodological issues.
28 A large number of studies now available in this review have provided new insights on
29 these and other issues as evaluated in Chapters 8 and 9 of the CD. The following discussion
30 builds upon the CD's evaluation of key methodological issues related to epidemiologic studies as
January 2005 3 -42 Draft - Do Not Cite or Quote
-------
1 a basis for staff judgments specifically regarding the use of epidemiologic evidence in
2 quantitative assessments, as discussed in Chapters 4 and 5.
3 This section addresses a number of key methodological issues. Section 3.6.1 discusses
4 air quality data reported in epidemiologic studies, which is one key component of quantitative
5 risk assessment. Section 3.6.2 discusses the issue of exposure error associated with the use of
6 ambient air concentrations as indicators of population exposures in epidemiologic studies.
7 Section 3.6.3 addresses statistical modeling and model specifications used in epidemiologic
8 studies. Section 3.6.4 addresses potential confounding by co-pollutants, to draw staff
9 conclusions about the use of specific study results in quantitative assessments. Finally, two of
10 the criteria discussed in the CD's integrative assessment of the health evidence - temporality and
11 the nature of concentration-response functions - are discussed. Section 3.6.5 includes discussion
12 of several topics in temporal relations between PM exposure and health outcomes. In section
13 3.6.6, the form of concentration-response relationships in both short-term and long-term
14 exposure studies is discussed, as is evidence related to the potential existence of population
15 threshold levels for effects.
16
17 3.6.1 Air Quality Data in Epidemiologic Studies
18 In general, epidemiologic studies use ambient measurements to represent population
19 exposures to PM of ambient origin. This section discusses some considerations with regard to
20 the ambient PM measurements: (1) whether the type of monitoring method influences the
21 epidemiologic study findings; (2) how measurement error might affect estimates of effects for
22 PM2 5 and PM]0.2 s and (3) how the frequency of PM measurement collection can influence the
23 power and certainty of study results. Questions related to the influence of exposure error on
24 epidemiologic study results are discussed in the following section.
25 Many studies have used PM25 and PM10_25 measurements from dichotomous samplers or
26 Harvard impactors, but PM2 5 and PM]0 measurements from co-located TEOMs or BAMs also
27 have been used, along with other methods (see Chapter 2 for more detailed descriptions of
28 monitors). In reviewing results from studies using various monitoring methods for PM2.5 and-
29 PMi0.25, staff finds that mere appear-to be no systematic differences in the effect estimates
30 related to the use of differing monitoring methods.
January 2005 3-43 Draft - Do Not Cite or Quote
-------
1 For these various monitoring methods, however, another factor to consider is the degree
2 to which uncertainty in the air quality measurements may affect epidemiologic associations with
3 PM^ 5 or PM2 5 The CD summarizes the findings of several new analyses that show the
4 potential influence of differential measurement error on epidemiologic analysis results, for either
5 PM with gaseous pollutants, or PM10.25 and PM25 as separate pollutants (section 8.4.5). Several
6 studies used simulation analyses of a "causal" pollutant and a "confounder" with differing
7 degrees of measurement error and collinearity between the pollutants. These studies found that,
8 in some circumstances, a causal variable measured with error may be overlooked and its
9 significance transferred to a surrogate. However, for "transfer of apparent causality" from the
10 causal pollutant to the confounder to occur, there must be high levels of both measurement error
11 in the causal variable and collinearity between the two variables (CD, p. 8-282, 8-283). The
12 conditions required for the error to substantially influence the epidemiologic findings are severe
13 and unlikely to exist in current studies. Thus, while the potential remains for differential error in
14 pollutant measurements to influence the results of epidemiologic studies, it is unlikely that the
15 levels of measurement error and correlation between pollutants reported in existing studies
16 would result in transfer of apparent causality from one pollutant to another (CD, p. 9-38).
17 One analysis applied measurement error models to data from the Harvard Six Cities
18 study, specifically testing relationships between mortality and either fine or thoracic coarse
19 particles (Carrothers and Evans, 2000). The authors identified several variables that could result
20 in biased effect estimates for fine- or coarse-fraction particles: the true correlation of fine- and
21 coarse-fraction particles, measurement errors for both, and the underlying true ratio of the
22 toxicity of fine- and coarse-fraction particles. The existence of measurement error and
23 collinearity between pollutants could result in underestimation of the effects of the less well-
24 measured pollutant. However, the authors conclude "it is inadequate to state that differences in
25 measurement error among fine and coarse particles will lead to false negative findings for coarse
26 particles. If the underlying true ratio of the fine and coarse particle toxicities is large (i.e.,
27 greater than 3:1), fine particle exposure must be measured significantly more precisely in order
28 not to underestimate the ratio of fine particle toxicity versus coarse particle toxicity" (Carrothers
29 and Evans, 2000, p. 72; CD, p. 8-286). These analyses, using data from a study in which
30 significant associations were reported for mortality with PM2 5, but not with PM10.2 5, indicate that
January 2005 3-44 Draft - Do Not Cite or Quote
-------
1 it is unlikely that measurement error in one PM measurement will result in "false negative"
2 results for coarse particles'or "false positive" results for fine particles (CD, p. 8-286). Thus, for
3 either PM2 5 or PM10.2^5 measurement error is not likely to be falsely attributing effects fronYohe:
4 pollutant to another pollutant in the existing epidemiologic studies. ' '
5 ' However, it must'be recognized that measurement error is a larger issue for PM,0.2 5 than
6 for fine particles, especially when"PM10.25'concentrations are determined by subtraction of PM10
7 and PM25 measurements (see section 2.4.3). It is likely that measurement error would increase
8 the uncertainty of an epidemiologic association. With increased error in PM10_2 5 monitoring
9 methods, any reported epidemiologic associations would be less likely to" reach statistical
10 significance (CD/p. 5-126). Thus, a set of positive but generally not statistically significant
11 associations between PMJ0.2 5 and a health outcome could be reflecting a true association that is •'
12 measured with error: Decreases in study precision would also occur even'if gravimetric PMJ0.25
13 were"perfectly measured, but the sources and relative composition of the coarse particles were
14 highly variable.' In evaluating the implications of the epidemiologic studies showing effects of
15 PM10.2 5, therefore, staff places more emphasis on the pattern of results in a series of studies than
16 on the statistical significance of any single effect estimate.
17 Finally, frequency of data collection can also affect the results reported from ' •
18 epidemiologic analyses. ,The CD discusses the use of less-than-everyday monitoring data as a '
19 source of uncertainly for time-series analyses (CD, p. 8-296). Many such studies were
20 conducted in areas where PM was monitored on a daily basis; in fact, the availability of every-
21 day monitoring is cited as a basis for study location in a number of reports. This is particularly
22 true for panel studies on respiratory or cardiovascular symptoms, all of which use daily PM
23 monitoring data,; though generally for shorter time periods.
*• r' ' '
24 • However, staff observes that a small number of the recent studies have been based on-less
25 frequently collected data. Data'collection frequency is one component of statistical power for
t •*• . f,
26 time-series studies; and missing data would result in increased uncertainty in study results. In
27 addition, for either PM2 5 or PM10.2 5, one would expect that a substantial proportion of missing
28 data may complicate time-series analyses (CD, p. 9-41). As illustrated in the CD, effect
29 estimates for PM10 and mortality varied in size and statistical significance in a series of analyses
30 of data collected on a 1 -in-6 day schedule (CD, p. 8-297). The CD presents results from a study
January 2005 3-45 Draft - Do Not Cite or Quote
-------
1 had not been used in the original analyses (e.g., physical activity, lung function, marital status),
2 did not alter the original findings. Data were also obtained for additional city-level variables that
3 were not available in the original data sets (e.g., population change, measures of income,
4 maximum temperature, number of hospital beds, water hardness) and reanalysis investigators
5 included these data in the models. The associations between fine particles and mortality were
6 > generally unchanged in these new analyses, with the exception of population change, which did
7 somewhat reduce the size of the associations with fine particles or sulfates (CD, p. 8-92). ,
8 In summary, the sensitivity of epidemiologic study results to model specification has L
9 been investigated for both short-term and long-term exposure studies. In both cases, the
10 reanalyses generally support the findings of the original studies, while raising questions for
11 further research. For short-term exposure studies, staff concludes that it.is appropriate to use the
12 results of the reanalyzed time-series epidemiologic studies or the results of studies that did not .
13 use GAM in the original analyses. In addition, staff observes that the use of more appropriate ,
14 convergence criteria in GAM has generally addressed questions about the magnitude of the
15 effect estimate. To obtain correct standard errors for the estimates, additional reanalyses used
16 GLM and parametric smoothing approaches that generally produced larger standard errors. For
17 quantitative risk assessment, staff concludes that models using more stringent GAM criteria
18 likely provide the most representative effect estimate sizes, while in illustrating the significance
19 of associations (e.g., as presented in Figures 3-1 and 3-2) staff has chosen to use results from
20 GLM-based analyses when available. For long-term exposure studies, staff concludes that
21 results from the reanalyses or extended analyses, in particular the extended analysis of the ACS
22 study, are most appropriate for use in quantitative assessment.
23 ..
24 3.6.4 Co-pollutant Confounding and Effect Modification
25 . Confounding occurs when a health effect that is caused by one risk factor is attributed to
26 another variable that is correlated with the causal risk factor; epidemiologic analyses attempt to
27 adjust or control for potential confounders. A gaseous copollutant (e.g., O3, CO, SO2 and NOj)
28 meets the criteria for potential confounding in PM-health associations if: (1) it is a potential risk
29 factor for the health effect under study; (2) it is correlated with PM; and (3) it does not act as an
30 intermediate step in the pathway between PM exposure and the health effect under study (CD, p.
January 2005 . 3-50 Draft - Do Not Cite or Quote
-------
n
1 it is unlikely that measurement error in one PM measurement will result in "false negative"
•
2 results for coarse particles 'or "false positive" results for fine particles (CD,' p. 8-286). Thus, for
3 either PM15 or PMi0.2'5 measurement error is not likely to be falsely attributing effects from" one
4 pollutant to another pollutant in the existing epidemiologic studies.
5 However, it must be recognized that measurement error is a larger issue for PM]0.2 5 than
6 for fine particles, especially when PM10_2J" concentrations are determined by subtraction of PM,0
7 and PM2 5 measurements (see section 2. 4. 3). It is likely that measurement error would increase
8 Hie uncertainty of an epidemiologic association. With increased error in PM10.2 5 monitoring
9 methods, any reported epidemiologic associations would be less likely to" reach statistical
10 significance (CD,'p. 5-126). Thus, a set of positive but generally not statistically significant
1 1 associations between PM10.2 5 and a health outcome could be reflecting a true association that is •'
1 2 measured with error; Decreases in study precision would also occur even if gravimetric PM10.2 5
13 were'perfectly measured, but the sources and relative composition of the coarse particles were
14 highly variable. In evaluating the implications of the epidemiologic studies showing effects of
15 PM10.25, therefore, staff places more emphasis on the pattern of results in a series of studies than
16 on the statistical significance of any single effect estimate. '
1 7 Finally, frequency of data collection can also affect the results reported from
18 epidemiologic analyses. ,The CD discusses the use of less-than-everyday monitoring data as a
19 source of uncertainty for time-series analyses (CD, p. 8-296). Many such studies were
20 conducted iri areas where PM was' monitored on a daily basis; in fact, the availability of every-
21 day monitoring is cited as a Basis for study location in a number of reports. This is particularly
22 true for panel studies on respiratory or cardiovascular symptoms, all of which use daily PM
i ,
23 monitoring data, though generally for shorter time periods.
24 • However, staff observes that a small number of the recent studies have been based on less
25 frequently collected data. Data1 collection frequency is one component of statistical power for
1 'T • I .
26 time-series studies; and missing data would result in increased uncertainty in study results. Iri
27 addition, for either PM2 5 or PM10.2 5, one would expect that a substantial proportion of missing
28 data may complicate time-series analyses (CD, p. 9-41). As illustrated in the CD, effect
29 estimates for PM10 and mortality varied in size and statistical significance in a series of analyses
30 of data collected on a l-in-6 day schedule (CD, p. 8-297). The CD presents results from a study
January 2005 3-45 Draft - Do Not Cite or Quote
-------
u
1 correlations. In such situations, while the epiderhiologic associations may be illustrating true
2 time-series relationships between PM and a health outcome, it is more difficult to draw
3 inferences about the population exposure levels at which those effects are seen. From studies in
4 which significant associations are reported with PM,0,2 5, the distribution of ambient monitoring
5 data available for the study may reflect levels that are higher or lower,than those experienced by
6 neighborhoods in other parts of the community. •
7 • • '
8 3.6.3 Alternative Model Specifications
9 As observed earlier, statistical modeling issues for epidemiologic studies were discussed
10 in great detail in the 1996 PM CD (EPA, 1996a, sections 12.6.2 and 12.6.3). This evaluation
11 lead to the conclusion that PM-related effects observed in epidemiologic studies were unlikely to ,
12 be seriously biased by inadequate statistical modeling or confounded by weather (CD, p. 8-22).
13 Statistical modeling issues have re-emerged in this review, however, and much attention has
14 been given to further investigations of approaches to model specification for epidemiological
15 analyses. The following discussions draw from the CD's evaluation of model specification . , \
16 issues for bom short-term and long-term exposure studies.
17 Time-series epidemiologic studies
18 In 2002, questions were raised about the default convergence criteria and standard error
19 calculations made using GAM, which have been commonly used in recent time-series
20 epidemiologic studies. As discussed more completely in the CD (section 8.4.2), a number of
21 time-series studies were reanalyzed using alternative methods, typically GAM with more
22 stringent convergence criteria and alternative models such as GLM with natural smoothing
23 splines. The results of the reanalyses have been compiled and reviewed in an HEI publication
24 (HEI, 2003a). Reanalyzed PM10 mortality study results are presented in Figure 8-15 in the CD,
25 where it can be seen that the reanalyses generally did not substantially change the findings of the
26 original analyses, and the changes in effect estimates with alternative analysis methods were
27 much smaller than the variation in effects across studies. Taking into account the conclusions of
28 the HEI review, the CD finds that mortality effect estimates were often, but not always, reduced
29 with the use of GAM with more stringent convergence criteria; however, the extent of these
30 changes was not substantial in most cases (CD, p. 8-232). Further, for morbidity studies, the
January 2005 3-48 Draft - Do Not Cite or Quote
-------
1 CD finds that the impact of the reanalyses was relatively small and the basic conclusions
2 regarding the significance of PM-related hospital admissions remained unchanged when more
3 stringent GAM criteria were used (CD; p. 8-235).
4 These reanalyses also investigated alternative model specifications to control for '
5 potential weather effects and temporal trends. As shown in Figures 8-20 and 8-21 in the CD, the
6 magnitude of the effect estimate for PM can decrease with increasing control for weather and
7 temporal trend, though it generally stabilizes at some point. The CD observes that there is no
8 clear consensus at this time as to what constitutes appropriate control for such variables, while
9 recognizing that no single approach is likely to be most appropriate in all cases (CD, p. 8-340).
10 If the model does not adequately address daily changes in weather variables, then some effects of
11 temperature on health would be falsely ascribed to the pollution variable. Conversely, if the
12 model overcontrols for weather," such that the temperature-health relationship is more "wiggly" '
13 man the true dose-response function, then the result will be a much less efficient estimate of the
14 pollutant effect (CD, p. 8-236). This would result in incorrectly ascribing some of the true
15 pollution effect to the temperature variable, which would make it difficult to detect a real but
16 small pollution effect. The CD concludes that the available studies appear to demonstrate that
17 there are PM-related effects independent of weather influences, but that further evaluation is
18 needed on how to best characterize possible combined effects of air pollution and weather (CD,
19 p. 8-340). ' " ' "
20 Prospective cohort epldemiologic studies
21 Data from the ACS and Six Cities prospective cohort studies were used in a major
22 reanaly sis study that evaluated a number of issues that had.been raised for the long-term
23 exposure studies. These issues included whether the results were sensitive to alternative
24 modeling strategies. The reanaly sis included two major components, a replication and validation
25 study,'and a sensitivity analysis, where alternative risk models and analytic approaches were
26 used to test the robustness of the original analyses. In the first phase, the data from the two
27 studies were found to be of generally high quality, and the original results were replicated,
28 confirming the original investigators'"findings of associations with both total and
29 cardiorespiratory mortality (Krewski et al., 2000; CD, p. 8-91). In the second phase, the
30 sensitivity analyses generally reported that the use of alternative models, including variables that
January 2005 ' 3-49 • Draft-Do Not'Cite or Quote
-------
1 had not been used in the original analyses (e.g., physical activity, lung function, marital status),
2 did not alter the original findings. Data were also obtained for additional city-level variables that
3 were not available in the original data sets (e.g., population change, measures of income,
4 maximum temperature, number of hospital beds, water hardness) and reanalysis investigators
5 included these data in the models. The associations between fine particles and mortality were
6 ( generally unchanged in these new analyses, with the exception of population change, which did
7 somewhat reduce the size of the associations with fine particles or sulfates (CD, p. 8-92).
8 In summary, the sensitivity of epidemiologic study results to model specification has
9 been investigated for both short-term and long-term exposure studies. In both cases, the
10 reanalyses generally support the findings of the original studies, while raising questions for
11 further research. For short-term exposure studies, staff concludes that it is appropriate to use the
12 results of the reanalyzed time-series epidemiologic studies or the results of studies that did not
13 use GAM in the original analyses. In addition, staff observes that the use of more appropriate
14 convergence criteria in GAM has generally addressed questions about the magnitude of the
15 effect estimate. To obtain correct standard errors for the estimates, additional reanalyses used
16 GLM and parametric smoothing approaches that generally produced larger standard errors. For
17 quantitative risk assessment, staff concludes that models using more stringent GAM criteria
18 likely provide the most representative effect estimate sizes, while in illustrating the significance
19 of associations (e.g., as presented in Figures 3-1 and 3-2) staff has chosen to use results from
20 GLM-based analyses when available. For long-term exposure studies, staff concludes that
21 results from the reanalyses or extended analyses, in particular the extended analysis of the ACS
22 study, are most appropriate for use in quantitative assessment.
23
24 3.6.4 Co-pollutant Confounding and Effect Modification
25 Confounding occurs when a health effect that is caused by one risk factor is attributed to
26 another variable that is correlated with the causal risk factor; epidemiologic analyses attempt to
27 adjust or control for potential confounders. A gaseous copollutant (e.g., O3, CO, SO2 and NO2)
28 meets the criteria for potential confounding in PM-health associations if: (1) it is a potential risk
29 factor for the health effect under study; (2) it is correlated with PM; and (3) it does not act as an
30 intermediate step in the pathway between PM exposure and the health effect under study (CD, p.
January 2005 ' 3-50 Draft - Do Not Cite or Quote
-------
1 8-10). Effect modifiers include variables that may influence the health response to the pollutant
2 exposure (e.g., co-pollutants, individual susceptibility, smoking or age). Both are important
3 considerations for evaluating effects in a mixture of pollutants, but for confounding, the
4 emphasis is on controlling or adjusting for potential confounders in estimating the effects of one
5 pollutant, while the emphasis for effect modification is on identifing and assessing the level of
6 effect modification (CD, p. 8-12).
7 Co-pollutant Confounding
8 Potential confounding by gaseous copollutants has been most commonly assessed by
9 using multi-pollutant models. As discussed in the CD (section 8.4.3.2), there are statistical
10 issues to be considered with multi-pollutant models, such as possibly creating mis-fitting models
11 by forcing all pollutants to fit the same lag structure, by adding correlated but non-causal
12 variables, or by omitting important variables. There are issues relating to potential copollutant
13 confounding that multi-pollutant models may not be able to address. Inclusion of pollutants in a
14 multi-pollutant model that are highly correlated'with one another can lead to misleading
15 conclusions in identifying a specific causal pollutant. Collinearity between pollutants may occur
16 if the gaseous pollutants and PM come from the same sources, or if PM constituents are derived
17 from gaseous pollutants (e.g., sulfates from S02) (CD, p. 8-12). This situation certainly occurs.
18 For example, sources of fine particle constituents include combustion of various fuels, gasoline
19 or diesel engine exhaust, and some industrial processes (CD, Table 9-1); these sources also emit
20 gaseous pollutants. When collinearity exists, multi-pollutant models would be expected to
21 produce unstable and statistically insignificant effect estimates for both PM and the co-
22 pollutants. .
23 In the NMMAPS multi-city analyses, one key objective was to characterize the effects of
24 PM10 and the gaseous co-pollutants, alone and in combination. Multi-pollutant modeling was
25 used in the NMMAPS mortality analyses for 20 and 90 U.S. cities, in which the authors added
26 first O3, then O3 and another co-pollutant (e.g., CO, NO2 or SO2) to the models (CD, p. 8-35).
27 The relationship between PM]0 and mortality was little changed in models including control for
28 O3 and other gaseous pollutants (CD, Figure 8-4, p. 8-35). The authors concluded that the PM10-
29 mortality relationship was not confounded by co-pollutant concentrations across 90 U.S. cities
30 (Samet et al., 2000a,b; Domenici, 2003).
January 2005 3-51 Draft - Do Not Cite or Quote
-------
1 Single- and multi-pollutant model results for a range of health outcomes with PM10, PM25
2 and PM10_25 from multi- and single-city studies are presented in Figures 8-16 through 8-19 of the
3 CD. For the most part, the addition of gaseous co-pollutants had little influence on PM
4 associations, although substantial reduction in associations with PM could be seen in some cases
5 when gaseous pollutants are added to the model.
6 In the long-term exposure studies, multi-pollutant models have been tested in some
7 analyses. The reanalysis of data from the ACS cohort indicated that associations between
8 mortality and PM2 5 or sulfates were reduced in size in co-pollutant models including SO2 but not
9 with the other gaseous pollutants. Since SO2 is a precursor for fine particle sulfates, it is
10 naturally difficult to distinguish effects from the precursor SO2 and fine particles, as discussed
11 above (CD, p. 9-37). -
12 In addition to statistical approaches for assessing potential confounding, the CD also
13 discusses information available on the biological plausibility of effects of the potentially
14 confounding pollutants and consideration of exposure relationships. Information about the
15 biological plausibility of effects can inform conclusions about which pollutant from a mixture of
16 correlated pollutants is more likely responsible for the observed associations. For example, in
17 evaluating results of the ACS study analyses described above, the authors concluded that an
18 association between SO2 and mortality was less plausible than the association between PM25 and
19 mortality (CD, p. 8-15). Further research is needed on biological mechanisms underlying air
20 pollution-related effects to support future assessments.
21 Some recent exposure studies have collected personal and ambient monitoring data,
22 collected at a single central site, for PM25 and gaseous pollutants (e.g., O3, SO2 and NO2)5 and
23 assessed the degree of day-to-day correlation between the different measures of personal and
24 ambient concentrations. The investigators reported that the personal and ambient PM2 5
25 measurements were correlated, as were personal exposure to PM2 5 and ambient concentrations
26 of the gaseous pollutants. However, the personal and ambient concentrations of each of the
27 gaseous pollutants were not well correlated. These findings suggest that associations reported
28 with ambient PM2 5 are truly reflecting associations with fine particles and that fine particles are
29 unlikely to be simply acting as surrogates for other gaseous pollutants (Samat et al., 2000,2001;
30 CD, p. 5-90).
January 2005 3-52 Draft - Do Not Cite or Quote
-------
1 Effect Modification
2 Some new studies have also assessed the potential for effect modification by the gaseous
3 pollutants. In the NMMAPS morbidity analyses for!4 U.S. cities, the authors tested for
4 relationships between the coefficients for the PM]0-admissions with PM10-co-pollutant
5 correlations for each city. No such relationships were found between the. PM,0 effect estimates
6 for cardiovascular .or respiratory hospitalization and PMi0-co-pollutant correlations (CD, pp. 8-
7 146, 8-175), These results indicate that associations reported in this study for PM10 are not
8 dependent on the correlation between PM10 and the gaseous copollutants.
9 An alternative way to evaluate the effect of co-pollutants on associations reported with
10 PM15 is illustrated in Figure 3-3. As discussed in the 1996 Staff Paper, if PM is acting
11 independently, then a consistent association should be observed in a variety of locations of
12 differing levels of co-pollutants. Effect estimates for PM10-mortality associations were plotted
13 against concentrations of gaseous pollutants in the study area, and there was no evidence that
14 associations reported between PM10 and mortality were correlated with copollutant
15 concentrations. (EPA, 1996b, Figure V-3a,b). Similarly, Figure 3-3 shows the reported effect
16 estimates for PM25 and mortality (from single-pollutant models) from U.S. and Canadian studies
17 relative to the levels of O3, NO2, SO2, and CO present in the study locations. As was seen in the
18 last review for PM,0, the magnitude and statistical significance of the associations reported
19 between PM25 and mortality in these studies show no trends with the levels of any of the four
20 gaseous co-pollutants. While not definitive, these consistent patterns indicate that it is more .
21 likely that there is ah independent effect of PM25 that is not appreciably modified by differing
22 levels of the gaseous pollutants.
23 In summary, the available evidence does not indicate that exposure to the gaseous
24 pollutants is an effect modifier for PM-related health outcomes. With regard to confounding
25 effects between pollutants, where PM and the other pollutants are correlated, it can be difficult to
26 distinguish effects of the various pollutants in multi-pollutant models. However, a number of
27 research groups have found the effects of PM and gases to be independent of one another, as
28 illustrated in Figures 8-16 through 8-19 of the CD. In addition, new evidence on exposure
29 considerations suggests that it is less likely that a relationship found between a health endpoint
30 and ambient PM concentrations is actually representing a relationship with another pollutant
January 2005 3-53 Draft - Do Not Cite or Quote
-------
ffi
• [*TJ n
«~ i W n
io| — $~H
1 1 II
oe oz 01 o
• -I i ' 1
y • I
• i -^ M
OS 05 (H 0
0.08 0. tO 0.12 0,14 0.1 6 O.OS 0.1 0 0.1 5 0.20
Ozone (ppm), 4th highest &-bour daily maximum NO2 (ppm), 2nd highest 1-hour daily maximum
„ V J?
0 " .
i • .Til
^ tJr
-l-tH
I
10
ollutant concentration
time period: (A) mean
2nd highest 24-hour S(
' • s* ^ «« ••
} ^1i?
1 till
H f t> Q1 g g
~" * S e S s
''8 a S S e
i 1 2- 1
i ,^ "3 i3 s
* oz n o a a 5 55
IBB) O^ JS
u
a o.oa 0.10
and total mortality fron
ta obtained from the Aii
,B) mean of 2nd highest 1
-hour CO concentration
L .^ 1
0.02 0.04 0.0
S02 (ppm), 2nd highest 24-hoiir maxhium
ssociations between PM2 s
e locations. Air quality dat
lour ozone concentration;
i; (D) mean of 2nd highest 8
^ S ' S
A « s s
oc oz 01 o JJ *^ M S
10 %SQ PW a»«UJ!»sa »»«3 & § * *•
E 1 "^ 8
13. Schwartz, 2003 a, St. Louis, MO
14. Schwartz, 2003a, Steubenville, OH
15. Schwartz, 2003a, Topefca, KS
16. Tsai et al., 2000, Camden NJ
17. Tsai et al., 2000, Elizabeth NJ
18. Tsai et al., 2000, Newark NJ
1. Chock et al, 2000, Pittsburgh, PA 7. Moolgavkar, 2003, Los Angeles, CA
2. Fairley, 2003, Santa Clara County, CA 8. Ostro et al., 2003, Coachella Valley, CA
3. Ito, 2003, Detroit, Ml 9. Ostro et al., 1995, Southern California
4. Klemm and Mason, 2000, Atlanta, GA 10. Schwartz, 2003a, Boston, MA
5. Lipfert et al,, 2000a, Philadelphia, PA 11. Schwartz, 2003a, Knoxville, TN
6. Mar et al, 2003, Phoenix, AZ 12. Schwartz, 2003a, Portage, WI
o
0
a
i
Wi
I
-------
Or-
COr-
oe oz
10 %96
If)
o
o
OJ
o
i
in
r o
I
°P!
o O
o «
s
- o
IJ-i
s!^
s
el
ea
tifi
W> fa B
«a ^
s /«— •••
II!
"S^o
I a 'S
T; to u
* >> o
8 w >»
111
K) , S
i—J ,J3 O
«« -^- -™-
ilo
ill
2 oa
* 8
£ «4
i 8 •g •% -5
11 a "S s
i ee H-H H
; «i « t-^ »
Sir!
S S3
-
&
«5:-ift
linn
rf a
JJ a
S-s-9
!,»*
•5? £ S * * *
It S-B-6-e
«-rf
§£
Ifi-lfil
S3Stfrf|
fStf I i fa-
«slis
ar« * - s
-r§«-§
ock
mm an
fert et a
etal,
1. Cho
2. Fairley,
3. Ito, 2
4. Kle
S. Lipf
6. Mar
a
I
-------
1 Finally, it is possible that pollutants may act together, or that the effects of a single pollutant may
2 be mediated by other components of an ambient pollution mix. For example, recent animal
3 toxicologic studies have tested effects of exposure to PM (e.g., urban PM, carbon particles, acid
4 aerosols) in combination with O3 and suggeted that co-exposure to O3 and urban particles resulted
5 in greater effects than those reported with exposure to O3 alone, while mixed results were
6 reported from studies using combinations of acid aerosols and 03 (CD, Table 7-13). Taking into
7 consideration the findings of single- and multi-city studies and other evaluations of potential
8 confounding by gaseous co-pollutants described in preceding sections, the CD concludes that
9 while research questions remain, in general, "associations for various PM indices with mortality
10 or morbidity are robust to confounding by co-pollutants." (CD, p. 9-37). This indicates that effect
11 estimates from single-pollutant models can be used to represent the magnitude of a concentration-
12 response relationship, though there will remain uncertainty with regard to potential contributions
13 from other pollutants. For quantitative assessment, staff concludes that single-pollutant model
14 results provide reasonable indicators of the magnitude of PM-related effects for the purpose of
15 comparing risk estimates with different alternative standard scenarios, with additional sensitivity
16 analyses to include multi-pollutant model results.
17
18 3.6.5 Temporality in Concentration-Response Relationships
19 3.6.5.1 PM short-term exposure time periods
20 While most time-series epidemiologic studies use 24-hour average PM measurements,
21 several new studies have used ambient PM concentrations averaged over shorter time intervals,
22 such as 1- or 4-hour averages. Many such studies have evaluated associations with
23 cardiovascular health biomarkers or physiological changes. Section 8.3.1.3.4 of the CD describes
24 several epidemiologic studies that report statistically significant associations between 2- to 4-hour
25 PM10 or PM2 5 concentrations and cardiovascular health endpoints, including myocardial
26 infarction incidence and heart rate variability (CD, pp. 8-162 to 8-165). One study reported effect
27 estimates for myocardial infarction incidence with PM2 5 averaged over 2- and 24 hours that are
28 quite similar in magnitude, and both are statistically significant (Peters et al., 2001; CD, p. 8-165).
29 For respiratory health outcomes, two panel studies of symptoms in asthmatic subjects are
30 summarized in the CD (section 8.3.3.1.1). One study in a small Southern California community,
January 2005 3-56 Draft - Do Not Cite or Quote
-------
1 reported larger effect estimates for 1- or 8-hour concentrations than for 24-hour PMi0
2 concentrations (Delfino et al.; 1998), while the other, in Los Angeles, reported larger effect
3 estimates for 24-hour PMj0 concentrations (Ostro et al., 2001; CD, p. 8-206). However, several
4 studies of hospital admissions or medical visits for respiratory diseases reported the strongest
5 associations with several-day average PM concentrations (CD, p. 8-279).
6 Evidence of health effects associations with different exposure time periods can inform
7 staff conclusions and recommendations regarding potential NAAQS averaging times. Staff
8 observes that the very limited information available in the CD suggests that cardiovascular effects
9 may be associated with acute exposure time periods on the order of an hour or so.
10 3.6.5.2 Lag Structure in Short-term Exposure Studies
11 In the short-term exposure epidemiologic studies, many investigators have tested
12 associations for a range of lag periods between the health outcome and PM concentration (see
13 CD, sections 8.4.4 and 9.2.2.4). As discussed in the CD, it is important to consider the pattern of
14 results that is seen across the series of lag periods. If there is an apparent pattern of results across
15 the different lags, then selecting the single-day lag with the largest effect from a series of positive
16 associations is likely to underestimate the overall effect size, since single-day lag effect estimates
17 do not fully capture the risk that may be distributed over adjacent or other days (CD, p. 8-270). •
18 Where effects are found for a series of lag periods, a distributed lag model will more accurately
19 characterize the effect estimate size. However, if there is no apparent pattern or reported effects
20 vary across lag days, the use of any single result may be inappropriate for quantitative assessment
21 -(CD, p. 9-42). .' : ' •
22 For selecting effect estimates from studies for use in quantitative risk assessment, or for
23 evaluation of potential revisions to the standards, staff considered patterns of results for PM2 5 or
24 PM10_25 across lag periods from U.S. and Canadian studies.' Numerous investigators have
25 reported quantitative results only for the strongest associations, after testing associations over a
26 range of lags and finding a reasonably consistent pattern across lags. An example of such an
27 evaluation is provided in an analysis using hospitalization for asthma (Sheppard et al.,1999;
28 2003). This study tested lags to 3-days and beyond, and reported consistent patterns across lags
29 for associations between asthma hospitalization and PM,0, PM2 5 or PM]0.2 5. Results for the
30 strongest associations are presented, with the authors observing 'When considering single (vs.
January 2005 3-57 Draft - Do Not Cite or Quote
-------
1 distributed) lag estimates, it is important to put the estimate in the context of the pattern of lags
2 nearby and to recognize that effect estimates contain information from adjacent days owing to
3 serial correlation of the pollutant series. The pollutant effects given for asthma are larger than
4 and consistent with estimates obtained for adjacent lags. In contrast, adjacent lags to the same-
5 day PM and SO2 effects on appendicitis change much more abruptly, and the overall pattern is ,
6 unstable." (Sheppard et al., 1999, p. 27)
7 Most of the studies included in Appendix 3 A either selected lag periods a priori, or
8 , evaluated results for a range of lag periods, reporting effect estimates for one lag period based on
9 this evaluation. An example of results that do not follow a consistent pattern across lags can be
10 found in a study in Coachella Valley (Ostro et al., 2000; 2003). In this study, results for a series
11 of lags show fairly consistent patterns for associations between PM10 and PM10_25 and
12 cardiovascular mortality, but not with total or respiratory mortality, nor for associations between
13 PM2.5 and total and cardiovascular mortality. Based on the greater uncertainty on the effect
14 estimate size for the PM2 5-mortality association from this study, staff concludes that it would not
15 be appropriate to use the results for quantitative assessments.6 In addition, a series of studies in
16 Cook County, IL and Los Angeles County, CA, include effect estimates for 0- to 5-day lag
17 periods and, in general, the results follow a pattern. However, the pattern of results for COPD '
18 mortality with PM25 was quite inconsistent (Moolgavkar, 2000a,b,c; Moolgavkar, 2003, p. 191).7
19 Based on the considerations described above, the results for COPD mortality from this study were
20 not used in the risk assessment discussed in Chapter 4. >
21 The CD concludes that it is likely that the most appropriate lag period for a study will
22 vary, depending on the health outcome and the specific pollutant under study. Some general
23 observations can be made about lag periods for different health outcomes. For total and
24 cardiovascular mortality, it appears that the greatest effect size is generally reported for the 0-day
air quality measurements available for PMj 5 and PM10.2 5 may also contribute to the more uncertain
findings for PM2.s in this study. For PM10.j 3j a 10-year series of concentrations was modeled from a 2 Vi year series
of ambient measurements at co-located beta attenuation monitors, while predictive models for PM^j concentrations
were not reported to be adequate, so only the 2 '/4 year series of measurements were used in PM2 3 analyses.
That only l-in-6 day PM measurements were available in Los Angeles County is likely to be an important
factor contributing to less consistent findings there.
January 2005 3-58 Draft - Do Not Cite or Quote
-------
1 lag and 1-day lag, generally tapering off for longer lag periods (CD, p. 8-279). This is true also
2 for hospitalization for cardiovascular diseases. For cardiovascular effects such as myocardial
3 infarction or HRV change, there appears to be a pattern of larger effects with shorter lag periods,
4 such as 1- to 4-hours. For respiratory symptoms, many studies report effects over a series of lags,
5 with larger effect estimates for moving average or distributed lag models. Similarly, for asthma
6 hospitalization, there appear to be larger effects over longer average time periods, out to 5- to 7-
7 day average lags.
8 A number of recent studies that have investigated associations with distributed lags
9 provide effect estimates for health responses that persist over a period of time (days to weeks)
10 after the exposure period. The available studies have generally used PM,0 or other PM indicators,
11 but not PM2 5 or PM10.2 5. Effect estimates from distributed lag models are often, but not always,
12 larger in size mat those for single-day lag periods (CD, p. 8-281). For example, in multi-city
13 analyses of data from 10 U.S. cities, the effect estimates for total mortality from distributed lag
14 models are about twice those from 0-1 day average lag models (Schwartz, 2003b). In the 14-city
15 NMMAPS analysis of hospitalization in the elderly, the combined city effect estimate for COPD
16 hospitalization is larger (about doubled) in results of distributed lag models than in 0-1 day
17 average lag models, while the CVD hospitalization effect estimate is only increased by a small
18 amount, and the effect estimate for pneumonia hospitalization is somewhat smaller in distributed
19 lag models, compared with a 0-1 day average lag (Schwartz, et al., 2003).
20 In summary, the CD concludes that distributed lag results would likely provide more
21 accurate effect estimates for quantitative assessment than an effect estimate for a single lag period
22 (CD, p. 9-42). However, at this time, studies using PM2 5 and PM10.25 have not included
23 distributed lag models Most U.S. and Canadian studies have reported consistent patterns in
24 results for different lags; for these studies, an effect estimate for a single-day lag period is likely
25 to underestimate the effect. In quantitative assessments for PM2 5 and PM10.2 5, since results are
26 not available for distributed lag models, staff conclude that it is appropriate to use single-day lag
27 period results, recognizing that this is likely to underestimate the effect. For quantitative
28 assessment, staff concludes that it is appropriate to use results from lag period analyses consistent
29 with those reported in the CD, focusing on shorter lag periods for cardiovascular effects and lag'
30 periods of several days for respiratory effects, depending on availability of results. For the few
January 2005 3-59 Draft - Do Not Cite or. Quote
-------
1 studies that show inconsistent patterns, the use of single-day lag results are not appropriate for
2 quantitative assessment.
3 3.6.5.3 Seasonal Differences in Time-Series Epidemiologic Results
4 As discussed in section 3.5.3, time-series epidemiologic studies generally use some
5 temporal or seasonal terms in the models to control for seasonal changes in health outcomes. In
6 addition, a few epidemiologic studies have also evaluated PM-health associations across seasons
7 by doing analyses on data subdivided into different seasons, thus evaluating differences in effects
8 across the season, rather than trying to control for seasonal influences. The CD observes that
9 there can be seasonal differences in correlations between PM and other pollutants, or in PM levels
10 across seasons (CD, p. 8-57).
11 The CD presents results for seasonal analyses for individual studies in Chapter 8 and the
12 Appendices to Chapter 8. In 10 U.S. cities, the relationship between PM10 and mortality was the
13 same in analyses for data divided into summer and winter seasons (Schwartz, et al., 2000). In
14 Pittsburgh, relationships between PM]0.2S and PM25 and mortality were "unstable" when statified
15 by season, and there was evidence of differing multi-collinearily between seasons (Chock et al.,
16 2000). In Coachella Valley, associations between mortality and several PM indicators were
17 stronger in the winter season (October-May) than in the summer season (Ostro et al., 2000).
18 However, an earlier analysis in two Southern California counties reported significant associations
19 between estimated PM25 and mortality in the summer (April-September) quarter only (Ostro et
20 al., 1995). Seasonal analyses were done for the mortality-PM25 relationship in San Jose, and
21 there were no significant differences between the four seasons (Fairley, 2003). In Phoenix, the
22 association between PM10_2 s and mortality was reported to be highest in spring and summer, when
23 PM10.Z5 concentrations were lowest (Mar et al., 2003). Associations between PM10 and
24 hospitalization for cardiovascular diseases in Los Angeles were greater in the winter and fall
25 . seasons than in spring or summer (Linn et al., 2000). Asthma hospitalization was significantly
26 associated wiHi PM10 for both "wet" and "dry" seasons in Los Angeles, but the association was
27 larger in magnitude during the wet season (January-March) (Nauenberg and Basu, 1999). In
28 Seattle, associations between PM10, PM2 5 and PM10.2 5 and asthma hospitalization were positive in
29 all seasons, but higher in spring and fall (Sheppard et al., 2003).
January 2005 3-60 Draft - Do Not Cite or Quote
-------
1 Staff observes that these few studies show no apparent pattern in results across seasons.
2 "The largest of these studies showed no seasonal differences in the results combining data from 10
3 U.S. cities (Schwartz et al., 2000). Most of the studies listed above show generally positive
4 results across all seasons'tested, with some reporting larger effect estimates in one or more
5 season(s), but the differences were not statistically significant. Staff concludes that the available
6 evidence does not support quantitative assessment of seasonal differences in relationships
7 between PM and health outcomes.
8 3.6.5.4 Exposure Time Periods in Long-term Exposure Studies
9 Studies of effects related to long-term PM exposures have generally used air'quality levels
10 averaged over months of years as exposure indicators. It is important to recognize that these
11 studies do not test specifically for latency in an exposure-effect relationship. Instead, the average
12 PM levels are used to represent long-term exposure to ambient PM, and the exposure
13 comparisons are basically cross-sectional in nature (CD, p. 9-42). As discussed in the CD, it is
14 not easy to differentiate the role of historic exposures from more recent exposures, leading to
15 potential exposure measurement error (CD, p. 5-118). 'This potential misclassification of
16 exposure is increased if average PM concentrations change over time differentially between areas.
17 Several new studies have used different air quality periods for estimating long-term
18 exposure and tested associations with mortality for the different exposure periods. In the
19 extended analysis of the ACS study, Pope et al. (2002) reported associations between mortality
20 and'PM2.5 using the original air quality data (1979-1983), data from the new fine particle
21 monitoring network (1999-2000), and the average PM2S concentrations from both time periods.
22 The authors reported that the two data sets were well correlated, indicating that the ordering of the
,'
23 cities from low to high pollution levels had changed little. When using average PM2 s levels from
24 all years, the associations for total, cardiopulmonary and lung cancer were slightly larger in size,
25 though not significantly so, than for either individual air quality data set.
26 - A new analysis of the Six Cities data has evaluated mortality risk with different estimates
27 of long-term PM25 exposure. The original study (Dockery et al., 1993) averaged PM
28 concentrations over a period of years (1979 to 1986) to represent long-term PM exposure
29 estimates, while the new analysis includes PM2.5 data from more recent years and evaluates
30 associations with PM25 averaged over a range of time periods, such as 2 or 3-5 years preceding
January 2005
3-61
Draft - Do Not Cite or Quote
-------
1 the individual's death (Villeneuve et al, 2002). The authors reported that effect estimates for
2 mortality were lower with time-dependent PM15 exposure indicators (e.g., 2 years before
3 individual's death), than with the longer-term average concentrations. They postulate that this is
4 likely due to the "influence of city-specific variations in mortality rates and decreasing levels of
5 air pollution that occurred during follow-up" (CD, p. 8-97). This might be expected, if the most
6 polluted cities had the greatest decline in pollutant levels as controls were applied (CD, p. 8-93).
7 The authors observe that the fixed average concentration window may be more representative of
8 cumulative exposures, and thus a more important predictor of mortality, than a shorter time period
9 just preceding death (Villeneuve et al., 2002, p. 574).
10 Using essentially the same air quality data set as that used in the original ACS analyses,
11 Lipfert et al. (2000b) investigated associations between mortality and PM (using several PM
12 indicators) over numerous averaging periods. When using methods similar to those of the other
13 prospective cohort studies, the authors report finding similar associations between fine particles
14 and mortality (CD, p. 8-115). However, in analyses using mortality and PM data in different time
15 segments, the results were varied, with some statistically significant negative associations
16 reported. The authors report that the strongest positive, associations were found with air quality
17 data from the earliest time periods, as well as the average across all data.
18 All three analyses indicate that averaging PM concentrations over a longer time period
19 results in stronger associations; as the Six Cities study authors observe, the longer series of data is
20 likely a better indicator of cumulative exposure. In these studies, spatial variation in the PM
21 concentrations is the key exposure indicator, and one key question is the extent to which
22 concentrations change over time, particularly whether there are differential changes across cities.
23 As observed above, the order of cities from high to low pollution levels changed little across time
24 periods in the cities used in the ACS analyses. Where lower effect estimates are reported with
25 data collected in more recent years, the CD observes: "This is likely indicative of the
26 effectiveness of control measures in reducing source emissions importantly contributing to the
27 toxicity of ambient particles in cities where PM levels were substantially decreased over time"
28 (CD, p. 9-43). The CD concludes that further study is warranted on the importance of different
29 time windows for exposure indicators in studies of effects of chronic PM exposure.
January 2005 3-62 Draft - Do Not Cite or Quote
-------
1 For use in quantitative assessments, staff concludes that it appropriate to use results from
2 analyses that are based on averaging PM levels over longer time periods, since the recent studies
3 indicate that this provides a better indicator of long-term PM exposure. Thus, as described in
4 Chapter 4, the results from the extended ACS analyses using average PM25 concentrations from
5 both the original and more recent time periods are used in the PM risk assessment. Staff notes
6 that this is consistent with the advice to EPA from the Health Effects Subcommittee (HES) of the
7 SAB's Clean Air Act Compliance Council (SAB, 2004), in their review of methods used for
#'
8 EPA's health benefits assessments. The HES recommended using the results of ACS cohort
9 analyses that used air quality data averaged over the full study time period, indicating that this
10 represented the best period to us e in order to reduce measurement error.
11 . •
12 3.6.6 Concentration-Response Relationships and Potential Population Thresholds
13 In assessing or interpreting public health risk associated with exposure to PM, the form of
14 the concentration-response function is a critical component. The CD recognized that it is
15 reasonable to expect that, for individuals or groups of individuals with similar innate
16 characteristics and health status, there may be biological thresholds for different effects.
17 Individual thresholds would presumably vary substantially from person to person due to
18 individual differences in genetic-level susceptibility and pre-existing disease conditions (and
19 could even vary from one time to another for a given person). Thus, it would be difficult to detect
20 a distinct threshold at the population level, especially if the most sensitive members of a
21 population are unusually sensitive even down to very low concentrations. The person-to-person
22 difference in the relationship between personal exposure to PM of ambient origin and the
23 concentration observed at a monitor may also add to the variability in observed exposure-
24 response relationships, further obscuring potential population thresholds (CD, p. 9-43, 9-44).
25 The 1996 CD evaluated evidence from epidemiologic studies regarding both functional
26 form and whether a threshold for effects could be identified. Based on the few available studies,
27 the 1996 CD concluded that linear model results "appear adequate for assessments of PM10 and
28 PM25 effects" (EPA, 1996a, p. 13-91). Among the new epidemiologic studies of short-term PM
29 exposure are several that use different modeling methods to investigate potential threshold levels
30 and concentration-response forms.
January 2005
3-63
Draft - Do Not Cite or Quote
-------
1 Several time-series studies have evaluated potential threshold levels for associations
2 between mortality and short-term PM exposures. In plots of concentration-response curves from
3 multi-city analyses, using the NMMAPS data, it is difficult to discern any evident threshold for
4 relationships between PM10 and total or cardiorespiratory mortality. The authors also present
i
5 posterior probabilities for the existence of thresholds at different levels of PM 10 showing that if
6 there is a threshold in the relationships between PM10 and total or cardiorespiratory mortality, the
7 likelihood of the threshold being above about 25 |ig/m3 is essentially zero (Dominici et al., 2003b;
8 CD, pp. 8-320, 8-321). One single-city analysis used various statistical methods to test for
9 thresholds in simulated data sets that were created with assumed threshold levels ranging from
10 12.8 to 34.4 jig/m3 for the relationship between PM10 and mortality. The authors of this analysis
11 concluded that it was highly likely that standard statistical methods could detect a threshold level,
12 if one existed (Cakmak et al., 1999; CD, p. 8-319).
13 One single-city study used PM25 and PM10.2i measurements in Phoenix and reported that
14( there was no indication of a threshold in the association between PM10.2.5 and mortality, but that
15 there was suggestive evidence of a threshold for the mortality association with short-term
16 exposure to PM2 5 up to levels of about 20-25 u.g/m3 (Smith et al., 2000; CD, 8-322). In addition,
17 single-city analyses in Birmingham and Chicago suggested that the concentration-response
18 functions for PM10 and mortality changed to show increasing effects at levels of 80 to 100 jig/m3
19 PM10, but "not to an extent that statistically significant distinctions were demonstrated" (CD, p. 8-
20 322).
21 For long-term exposure to PM and mortality, the shape of the concentration-response
22 function was evaluated using data from the ACS cohort. The concentration-response
23 relationships for associations between PM2 s and all-cause, cardiopulmonary and lung cancer
24 mortality are shown in Figure 3-4. The authors reported that the associations for all-cause,
25 cardiovascular and lung cancer mortality "were not significantly different from linear
26 associations" (Pope, et al., 2002). It is apparent in this figure that the confidence intervals around
27 each of the estimated concentration-response functions expand significantly as one looks below
28 around 12-13 u.g/m3, indicating greater uncertainty in the shape of the concentration-
Jantiary 2005, 3-64 Draft - Do Not Cite or Quote
-------
Figure 3-4. Natural logarithm of relative risk for total and cause-specific mortality per
10 ug/m3 PM2 5 (approximately the excess relative risk as a fraction), with
smoothed concentration-response functions. Based on Pope et al. (2002) mean
curve (solid line) with pointwise 95% confidence intervals (dashed lines).
(Source: CD, Figure 8-7).
January 2005
3-65
Draft - Do Not Cite or Quote
-------
1 response relationship at concentration ranges below this level. In addition, for lung cancer, the
2 relationship appears to have a steeper linear slope at lower concentrations, with a flatter linear
3 slope at PM25 concentrations that exceed about 13 jig/m3 (CD, p. 8-98).
4 In summary, while staff recognizes that individual thresholds may likely exist for specific
5 health responses, existing studies do not support or refute the existence of population thresholds
6 for PM-mortality relationships, for either long-term or short-term PM exposures within the range
7 of air quality observed in the studies (CD, p. 9-44). While epidemiologic analyses have not
8 identified population thresholds in the range of air quality concentrations in the studies, it is
9 possible that such thresholds exist within or below these ranges but cannot be detected due to
10 variability in susceptibility across a population. Even in those few studies with suggestive
11 evidence of population thresholds, the potential thresholds are at fairly low concentrations (CD, p.
12 9-45). Based on the above considerations, staff concludes that it is appropriate to focus on linear
13 or log-linear concentration-response models reported in the studies for quantitative risk
14 assessment Recognizing that population thresholds may well exist below the lowest air quality
15 levels observed in the studies, staff concludes it is not appropriate to extrapolate below these
16 levels. Further, to address the possibility that population thresholds may exist at fairly low levels
17 within the range of air quality observed in the studies, staff concludes that it is appropriate to
18 consider alternative hypothetical threshold levels in the context of sensitivity analyses within the
19 PM risk assessment.
20
21 3.7 SUMMARY AND CONCLUSIONS
22 Based on the available evidence and the evaluation of that evidence in the CD,
23 summarized briefly above, staff concludes that the body of evidence supports an inference of
24 causality for associations between PM25 and a broad range of health effects. Short-term exposure
25 to PM2 5 is likely causally associated with mortality from cardiopulmonary diseases,
26 hospitalization and emergency department visits for cardiopulmonary diseases, increased
27 respiratory symptoms, decreased lung function, and physiological changes or biomarkers for
28 cardiac changes. Long-term exposure to PM2 5 is likely causally associated with mortality from
29 cardiopulmonary diseases and lung cancer, and effects on the respiratory system such as
30 decreased lung function or the development of chronic respiratory disease. Staff concludes that
January 2005 3 -66 Draft - Do Not Ci te or Quote
-------
1 there is less strength,"but suggestive evidence of causality for short-term exposure to PM10.2 5 and
2 indicators of morbidity, including hospitalization for cardiopulmonary diseases, increased
3 respiratory symptoms and decreased lung function. Staff concludes that it is appropriate to
4 consider including the health outcomes listed above in quantitative assessments for PM25 and
5 • PMio.^. Further, staff notes that more equivocal evidence is available for other PM-health
6 responses, such as associations between short-term exposure to PM10.25 and mortality, and
7 between PM and effects on infants. Staff believes that less certain evidence, while not
8 appropriate for quantitative assessment, can inform more general assessments of the evidence.
9 Several issues that are relevant to the interpretation of health evidence for quantitative
10 assessment of PM-related effects are discussed above. Measurement error and exposure error are
11 issues that are distinctly more important for interpretation of results for PM10.2.5 than PM25. For
12 PM10.2j, there is greater uncertainty in the relationship between ambient PM measured at central
13 monitors and individuals' exposure to ambient PM, based on both variability in PM10.25
14 concentrations across an area and decreased ability for coarse particles to penetrate into buildings.
15 This uncertainty is likely to increase the confidence intervals around effect estimates. In
16 interpreting results of associations with PM10.2 5, staff places greater emphasis on evaluating
17 results from the pattern of findings in multiple studies than on statistical significance of any
18 individual result.
19 In the evaluation of different epidemiologic model specifications, as described above,
20 some effect estimates differ upon reanalysis to address issues associated with the use of the
21 default GAM procedures, but many are little affected. Recognizing that there is no single
22 "correct" analytical approach, staff concludes that it is appropriate for quantitative assessment to
23 use results from short-term exposure studies that were reanalyzed with more stringent GAM
24 criteria or with other approaches such as GLM, or that did not use GAM in the original analysis.
25 Regarding potential confounding by co-pollutants, the CD concludes that the evidence
26 supports the existence of independent effects of PM, while recognizing the difficulties in
27 distinguishing effects from mixtures of correlated pollutants. Staff concludes that single-pollutant
28 model effect estimates can be used as reasonable indicators of the magnitudes of effect sizes, with
29 sensitivity analyses to evaluate the influence of adjustment for co-pollutants.
January 2005
3-67
Draft - Do Not Cite or Quote
-------
1 The CD concludes that distributed lag periods may provide the most representative
2 quantitative estimates of effect for some health outcomes, such as mortality. Recognizing that
3 distributed lags have not been used in the available studies of PM2 5 and PM10.2 5, staff concludes
4 that a reasonable approach to selection of effect estimates for use in quantitative assessment is to
5 evaluate the pattern of lag results available from studies. If the data show a reasonable pattern of
6 results, then selecting a single lag period is appropriate, recognizing that this result is likely to
7 underestimate effects. Conversely, if the pattern of results across lag periods is unstable, staff
8 concludes that it is inappropriate to use such results for quantitative assessment
9 For the long-term exposure studies, recent studies indicate that long-term PM exposure is
10 likely to be better estimated from air quality data averaged over longer time periods (e.g., multiple
11 years of data). Staff concludes that effect estimates based on PM data averaged over longer times
12 periods are more representative of population health responses for use in risk assessment.
13 Specifically, for the results from the extended analysis of the ACS study, staff concludes that it is
14 most appropriate to use the concentration-response functions from the models using averaged air
15 quality data over the full study time period for quantitative assessment.
16 Finally, evaluation of the health effects data summarized in the CD provides no evidence
17 to support selecting any particular population threshold for PM25 or PM1(K2 5, recognizing that it is
18 reasonable to expect that, for individuals, there may be thresholds for specific health responses.
19 Staff observes that uncertainty in the concentration-response function increases at the low end of
20 the range of concentrations. Even in those studies where the existence of population thresholds is
21 suggested, they are at fairly low concentrations. For the PM risk assessment, staff concludes that
22 it is appropriate to focus on linear or log-linear concentration-response models reported in the
23 studies, while considering alternative hypothetical threshold levels in the context of sensitivity
24 analyses. Staff also concludes it is not appropriate to extrapolate below the lowest PM
25 concentrations reported in the studies.
January 2005 3-68 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
33
34
35
36
37
38
39
40
41
42-
43
44
45
46
47
48
49
50
51
52
53
REFERENCES
• X"™"N
Abbey, D. E.; Lebowitz, M. D.; Mills, P. K.; Petersen, F. F.; Beeson, W. L.; Burchette, R. J. (1995a) Long-term
ambient concentrations of participates and oxidants and development of chronic disease in a cohort of
nonsmoking California residents. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the colloquium on
particulate air pollution and human mortality and morbidity; January 1994; Irvine, CA. Inhalation Toxicol.
.7:19-34.
Abbey, D. E.; Ostro, B. E.; Fraser, G.; Vancuren, T.; Burchette, R. J. (1995b) Estimating fine particulates less that' ,
2.5 microns in aerodynamic diameter (PM, 5) from airport visibility data in California. J. Exposure Anal.
Environ. Epidemiol. 5: 161-180.
Abbey, D. E.; Burchette, R.' J.; Knutsen, S. F.; McDonnell, W. F.; Lebowitz, M. D.; Enright, P. L. (1998) Long-term
particulate and other air pollutants and lung function in nonsmokers. Am. J. Respir. Crit. Care Med. 158:
289-298.
Abbey, D. E.; Nishino, N.; McDonnell, W. F.; Burchette, R. J.; Knutsen, S. F.; Beeson, L.; Yang, J. X. (1999) Long-
term inhalable particles and other air pollutants related to mortality in nonsmokers. Am. J. Respir. Crit. Care
Med. 159:373-382.
Avol, E. L.; Gauderman, W. J.; Tan, S. M.; London, S. J.; Peters, J. M. (2001) Respiratory effects of relocating to
areas of differing air pollution levels. Am. J. Respir. Crit. Care Med. 164: 2067-2072.
Berglund, D. J.; Abbey, D. E.; Lebowitz, M. D.; Knutsen, S. F.; McDonnell, W. F. (1999) Respiratory symptoms and
pulmonary function in an elderly nonsmoking population. Chest 115: 49-59.
Burnett, R. T.; Cakmak, S.; Brook, J. R.; Krewski, D. (1997) The role of particulate size and chemistry in the ^
association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
Environ. Health Perspect. 105:614-620.
Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips," O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
(2000) Association between particulate- and gas-phase components of urban air pollution and daily mortality
in eight Canadian cities. Inhalation Toxicol. 12(suppl. 4): 15-39.
Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated particulate mass and daily mortality in eight Canadian cities.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003];
Cakmak, S.; Burnett, R. T.; Krewski, D. (1999) Methods for detecting and estimating population threshold
concentrations for air pollution-related mortality with exposure measurement error. Risk Anal. 19:487-496.
*^
Carrothers, T. J.; Evans, J. S. (2000) Assessing the impact of differential measurement error on estimates of fine
particle mortality. J. Air Waste Manage. Assoc. 50:65-74.
Chen, L.; Yang, W.; Jennison, B. L.; Omaye, S. T. (2000) Air particulate pollution and hospital admissions for
chronic obstructive pulmonary disease in Reno, Nevada. Inhalation Toxicol. 12:281-298
Chock, D. P.; Winkler, S.; Chen; C. (2000) A study of the association between daily mortality and ambient air
pollutant concentrations in Pittsburgh, Pennsylvania. J. Air Waste Manage. Assoc. 50:1481-1500.
Choudhury, A. H.; Gordiari, M. E.; Morris, S. S. (1997) Associations between respiratory illness and PMlo air
pollution. Arch. Environ. Health 52:113-117.
January 2005
3-69
Draft - Do Not Cite or Quote
-------
1 Clyde, M. A.; Guttorp, P.; Sullivan, E. (2000) Effects of ambient fine and coarse particles on mortality in Phoenix,
2 Arizona. Seattle, WA: University of Washington, National Research Center for Statistics and the
3 Environment; NRCSE technical report series, NRCSE-TRS no. 040. Available:
4 http://www.nrcse.wshmgton.edu/pdf/trs40jm.pdf [18 October, 2004].
5 '
6 Delfino, R. J.; Murphy-Moulton, A. M.; Burnett, R. T.; Brook, J. R.; Becklake, M, R. (1997) Effects of air pollution
7 on emergency room visits for respiratory illnesses in Montreal, Quebec. Am. J. Respir. Grit. Care Med.
8 155:568-576.
9
10 Delfino, R. J.; Zeiger, R. S.; Seltzer, J. M.; Street, D. G. (1998) Symptoms in pediatric asthmatic and air pollution:
11 differences in effects by symptom severity, anti-inflammatory medication use and particulate averaging time.
12 Environ. Health Perspect 106:751-761.
13
14 Dockery, D. W.; Pope, C. A, III; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G., Jr.; Speizer, F. E.
15 (1993) An association between air pollution and mortality in six U.S. cities. N. Engl. J. Med.
16 329:1753-1759.
17
18 Dockery, D. W.; Speizer, F. E.; Stram, D. O.; Ware, J. H.; Spengler, J. D.; Ferris, B. G., Jr. (1989) Effects of
19 inhalable particles onrespiratory health of children. Am. Rev. Respir. Dis. 139: 587-594.
20
21 Dockery, D. W.; Cunningham, J.; Damokosh, A I.; Neas, L. M.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.;
22 Raizenne, M.; Speizer, F. E. (1996) Health effects of acid aerosols on North American children: respiratory
23 symptoms. Environ;Health Perspect. 104: 500-505.
24
25 Dominici, F.; McDermott, A; Daniels, M.; Zeger, S. L.; Samet, J. M. (2003a) Mortality among residents of 90 cities.
26 • In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
27 Effects Institute; pp. 9-24. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
28
29 Dominici, F.; Daniels, M.; McDermott, A; Zeger, S. L.; Samet, J. M. (2003b) Shape of the exposure-response
30 relation and mortality displacement in the NMMAPS database. In: Revised analyses of time-series studies of
31 air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 91 -96. Available:
32 http://wvvW.healtheffects.org/Pubs/TimeSeries.pdf [12 May, 2004].
33
34 EPA (1996a) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC: National Center for
35 Environmental Assessment-RTF Office; report no. EPA/600/P-95/001aF-cF. 3v
36
3 7 EPA. (1996b) Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment of
3 8 . Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office of Air
39 Quality Planning and Standards; report no. EPA-452\R-96-013.
40
41 EPA. (2002) Health assessment document for diesel engine exhaust. Washington, DC: Office of Research and
42 Development, National Center for Environmental Assessment; report no. EPA/600/8-90/057F. Available:
43 http://cfpub.epa.gov/ncea/[22 May, 2003].
44
45 Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1989-1996. In: Revised analyses of
46 time-series studies of air pollution and health. Special report Boston, MA: Health Effects Institute; pp. .
47 97-106. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
48
49 Gamble, J. L. (1998) Effects of ambient air pollution on daily mortality: a time series analysis of Dallas, Texas,
50 1990-1994. Presented at: 91st annual meeting and exhibition of the Air & Waste Management Association;
51 June; San Diego, CA. Pittsburgh, PA: Air & Waste Management Association; paper no. 98
52
January 2005 3-70 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
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
Gauderman, W. J.; McConnell, R.; Gilliland, F.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
Rappapoit, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and lung
function growth in southern California children. Am. J.Respir. Crit. CareMed. 162:1383-1390.
Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
Med. 166: 76-84.
Gold, D. R.; Litonjua, A.; Schwartz, J.; Lovett, E.; Larson, A.; Nearing, L.; Allen, G'.; Verrier, M; Cherry, R.;
Verrier, R. (2000) Ambient pollution and heart rate variability. Circulation 101:1267-1273.
Goldberg, M. S.; Bailor, J. C., Ill; Burnett, R. T.; Brook, J. R.; Tamblyn, R.; Bonvalot, Y.; Ernst, P.; Flegel, K. M.;
Singh, R. K.; Valois, M.-F. (2000) Identifying subgroups of the general population that may be susceptible
to short-term increases in particulate air pollution: a time-series studi in Montreal, Quebec. Cambridge, MA:
Health Effects Institute; research report 97.
HEI. (1997) Particulate air pollution and daily mortality: analyses of the effects of weather and multiple air pollutants
(The phase I.B. report of the Particle Epidemiology Evaluation Project). Health Effects Institute. Cambridge,
MA. • ....
HEI. (2003c) Commentary on revised analyses of selected studies. In: Revised analyses of time-series studies of air
pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 255-290. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Ito, K.; Kinney, P.; Thurston, G. D. (1995) Variations in PM-10 concentrations within two metropolitan areas and
their implications for health effects analyses. In: Phalen, R. F.; Bates, D. V., eds. Proceedings of the
colloquium on particulate air pollution and human mortality and morbidity, part II; January 1994; Irvine, CA.
Inhalation Toxicol. 7: 735-745.
Ito, K.; Thurston, G. D. (1996) Daily PM10/mortality associations: an investigation of at-risk subpopulations. J.
Exposure Anal. Environ. Epidemiol.6:79-95.
Ito, K. (2003) Associations of particulate matter components with daily mortality and morbidity in Detroit, Michigan.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 143-156. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdfil2 May, 2004].
Jacobs, J.; Kreutzer, R.; Smith, D, (1997) Rice burning and asthma hospitalizations, Butte County, California,
1983-1992. Environ. Health Perspect. 105: 980-985.
Katsouyanni, K.; Touloumi, G.; Samoli, E.; Petasakis, Y.; Analitis, A.; Le Tertre,' A.; Rossi, G.; Zmirou, D.;
Ballester, F.; Boumghar, A.; Anderson, H. R.; Wojryniak, B.; Paldy, A.; Braunstein, R.; Pekkanen, J.;
. Schindler, C.; Schwartz, J. (2003) Sensitivity analysis of various models of short-term effects of ambient
particles on total mortality in 29 cities in APHEA2. In: Revised analyses of time-series studies of air
pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 157-164. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Kinney, P. L.; Ito, K.; Thurston, G. D. (1995) A sensitivity analysis of mortality/PM-10 associations in Los Angeles.
Inhalation Toxicol. 7:59-69.
Klemm, R. J.; Mason, R. M., Jr. (2000) Aerosol research and inhalation epidemiological study (ARIES): air quality
and daily mortality statistical modeling—interim results. J. Air. Waste Manage. Assoc. 50:1433-1439.
t January 2005
3-71
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
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Klenun, R. J.; Mason, R. (2003) Replication of reanalysis of Harvard Six-City mortality study. In: Revised analyses
of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute;
pp. 165-172. Available: http://www.heaItheffects.org/Pubs/TimeSeries.pdfI12 May, 2004].
Krewski, D.; Burnett, R. T.; Goldberg, M. S.; Hoover, K.; Siemiatycki, J,; Jerrett, M.; Abrahamowicz, M.; White, W.
H. (2000) Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of particulate
air pollution and mortality. A special report of the Institute's particle epidemiology reanalysis project.
Cambridge, MA: Health Effects Institute.
Linn, W. S.; Szlachcic, Y.; Gong, H., Jr.; Kinney, P. L.; Berhane, K. T. (2000) Air pollution and daily hospital
admissions in metropolitan Los Angeles. Environ. Health Perspect. 108: 427-434.
Lipfert, F. W.; Morris, S. C.; Wyzga, R. E. (2000a) Daily mortality in the Philadelphia metropolitan area and size-
classified particulate matter. J. Air Waste Manage. Assoc. 50:1501-1513.
Lipfert, J. W.; Perry, H. M., Jr.; Miller, J. P.; Bary, J. D.; Wyzga, R. E.; Carmody, S. E. (2000b) the Washington
University-EPPJ veteran's cohort mortality study: preliminary results. Inhalation Toxicol. 12(Suppl. 4):41-
73.
Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of particulate matter components with daily
mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report 95.
'Lipsett, M.; Hurley, S.; Ostro, B. (1997) Air pollution and emergency room visits for asthma in Santa Clara County,
California. Environ. Health Perspect. 105: 216-222.
Mar, T. F.; Nonis, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
Phoenix, 1995-1997. Environ. Health Perspect 108:347-353.
Mar, T. F.; Norris, G. A.; Larson, T. V.; Wilson, W. E.fKoenig, J. Q. (2003) Air pollution and cardiovascular
mortality in Phoenix, 1995-1997. In: Revised analyses of time-series studies of air pollution and health.
Special report. Boston, MA: Health Effects Institute; pp. 177-182. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdftl8 October, 2004].
Mauderly, J.; Neas, L.; Schlesinger, R. (1998) PM monitoring needs related to health effects. In: Atmospheric
observations: helping build the scientific basis for decisions related to airborne particulate matter; Report of
the PM measurements research workshop, July 22-23,1998. Available from "PM Measurements Report",
Health Effects Institute, 955 Massachusetts Ave., Cambridge MA 02139.
McConnell, R.; Berhane, K.; Gilliland, F.; London, S. J.; Vora, H.; Avol, E.; Gaudermah, W. J.; Margolis, H. G.;
Luimann, F.; Thomas, D. C., Peters, J. M. (1999) Air pollution and bronchitic symptoms in southern
California children with asthma. Environ. Health Perspect. 107:757-760.
McDonnell, W. F.; Nishino-Ishikawa, N.; Petersen, F. F.; Chen, L. H.; Abbey, D. E. (2000) Relationships of mortality
with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers. J. Exposure
Anal. Environ. Epidemiol. 10:427-436.
Moolgavkar, S. H. (2000a) Air pollution and mortality in three U.S. counties. Environ. Health Perspect. 108:777-784.
Moolgavkar, S. H. (2000b) Air pollution and hospital admissions for diseases of the circulatory system in three U.S.
metropolitan areas. J. Air Waste Manage. Assoc. 50:271-280.
Moolgavkar, S. H. (2000c) Air pollution and hospital admissions for chronic obstructive pulmonary disease in three
metropolitan areas of tfie United States. Inhalation Toxicol. 12(Suppl. 4):75-90.
January 2005
3-72
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
33
34
35
36
37
'38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Moolgavkar, S. H. (2003) Air pollution and daily deaths and hospital admissions in Los Angeles and Cook counties.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute;pp. 183-198. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
Morris, R. D.; Naumova, E. N. (1998) Carbon monoxide and hospital admissions for congestive heart failure:
evidence of an increased effect at low temperatures. Environ. Health Perspect. 106:649-653.
Morris, R. D.; Naumova, E. N.; Munasinghe, R. L. (1995) Ambient air pollution and hospitalization for congestive
heart failure among elderly people in seven large US cities. Am. J. Public Health 85:1361-1365.
Naeher, L. P.; Holfo'rd, T. R.; Beckett, W. S.; Belanger, K.; Triche, E. W.; Bracken, M. B,; Leaderer, B. P. (1999)
Health women's PEF variations with ambient summer concentrations of PM,0, PM25, SO42", H*, and Oj. Am.
J.Respir.Crit Care Med. 160: 117-125. '
Nauenberg, E.; Basu, K. (1999) Effect of insurance coverage on the relationship between asthma hospitalizations and
exposure to air pollution. Public Health Rep. 114:135-148.
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Tollerud, D. J.; Speizer, F. E. (1995) The association of ambient air '
pollution with twice daily peak expiratory flow rate measurements in children. Am. J. Epidemiol.
141:111-122.
Neas, L. M.; Dockery, D. W.; Burge, H.; Koutrakis, P.; Speizer, F. E. (1996) Fungus spores, air pollutants, and other
-. determinants of peak expiratory flow rate in children. Am. J. Epidemiol. 143: 797-807. ,
Neas, L. M.; Dockery, D. W.; Koutrakis, P.; Speizer, F. E. (1999) Fine particles and peak flow in children: acidity
versus mass. Epidemiology 10:550-553.
Ostro, B. (1995) Fine particulate air pollution and mortality in two Southern California counties. Environ. Res.
70:98-104.
Ostro, B. D.; Lipsett, M. J.; Wiener, M. B.; Seiner, J. C. (1991) Asthmatic responses to airborne acid aerosols. Am. J.
Public Health 81: 694-702.
Ostro, B. D.; Lipsett, M. J.; Mann, J. K.; Braxton-Owens, H.; White, M. C. (1995) Air pollution and asthma
exacerbations among African-American children in Los Angeles. Inhalation Toxicol. 7:711-722.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coachella
Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10:412-419.
Ostro, B.; Lipsett, M.; Mann, J.;Braxtron-Owens, H.; White, M. (2001) Air pollution and exacerbation of asthma in
African-American children in Los Angeles. Epidemiology 12: 200-208.
Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coachella Valley, California.
In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health
Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/Pubs/rimeSeries.pdf [18 October,
2004].
Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Gauderman, W. J.; Lurmanh, F.; Linn, W. S.; Margolis, H.;
. Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999b) A study of twelve southern California communities,
with differing levels and types of air pollution. L Prevalence of respiratory morbidity. Am. J. Respir. Grit
Care Med. 159:760-767.
Peters, A.; Liu, E.; Verrier, R. L.; Schwartz, J.; Gold, D. R.; Mittleman, M.; Baliff, J.; Oh, J. A.; Allen, G.; Monahan,
K.; Dockery, D. W. (2000a) Air pollution and incidence of cardiac arrhythmia. Epidemiology 11:11-17.
January 2005
3-73
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
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Peters, A.; Dockeiy, D. W.; Muller, J. E.; Mittleman, M, A. (2001) Increased particulate air pollution and the
triggering of myocardial infarction. Circulation 103:2810-2815.
Pope, C. A., Ill; Schwartz, J.; Ransom, M. R. (1992) Daily mortality and PM10 pollution in Utah Valley. Arch.
Environ. Health 47:211-217.
Pope, C. A., HI; Thun, M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E.; Heath, C. W., Jr.
(1995) Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am. J.
Respir. Crit. Care Med. 151: 669-674.
Pope, C. A, III; Dockery, D. W.; Kanner, R. E.; Villegas, G. M.; Schwartz, J. (1999b) Oxygen saturation, pulse rate
and particulate pollution: a daily time-series panel study. Am. J. Respir. Crit. Care Med. 159: 365-372.
Pope, C. A, III; Verrier, R. L.; Lovett, E. G.; Larson, A. C.; Raizenne, M. E.; Kanner, R. E.; Schwartz, J.; Villegas,
G. M., Gold, D. R.; Dockery, D. W.(1999c) Heart rate variability associated with particulate air pollution.
Am. Heart J.I 38:890-899.
Pope, C, A, III; Thun,_M. J.; Namboodiri, M. M.; Dockery, D. W.; Evans, J. S.; Speizer, F. E., Heath, C. W., Jr.
(1995) Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am'. J.
Respir. Crit. Care Med. 151:669-674.
Pope, C. A, III; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
cardiopulmonary mortality, and long-term exposure to fine particulate air pollution, J. Am. Med. Assoc.
287:1132-1141.
Raizenne, M.; Neas, L. M.; Damokosh, A. I.; Dockery, D. W.; Spengler, J. D.; Koutrakis, P.; Ware, J. H.; Speizer, F.
E. (1996) Health effects of acid aerosols on North American children: pulmonary function. Environ. Health
Perspect. 104: 506-514.
SAB. (2004) Advisory for plans on health effects analysis in the analytical plan for EPA's second prospective
analysis - benefits and costs of the clean air act, 1990-2000. Advisory by the Health Effects Subcommittee
of the Advisory Council for Clean Air Compliance Analysis. EPA SAB Council - ADV-04-002. March,
2004. Available: liltp:/Avww.epa.gov/sciencel/pdiycouncii adv_04002.pdf.
Samet, J. M.; Zeger, S. L.; Kelsall, J. E.; Xu, J.; Kalkstein, L. S. (1996) Weather, air pollution and mortality in
Philadelphia, 1973-1980, report to the Health. Effects Institute on phase IB, Particle Epidemiology
Evaluation Project. Cambridge, MA: Health Effects Institute; review draft.
Samet, J. M.; Zeger, S. L.; Domenici, F.; Curriero, F.; Coursac, I.; Dockery, D.W.; Schwartz, J.; Zanobetti, A.
(2000a) The national morbidity, mortality, and air pollution study. Part I: methods and methodological
issues. Cambridge, MA: Health Effects Institute: research report no. 94.
Samet, J. M.; Zeger, S. L.; Domenici, F.; Curriero, F.; Coursac, I.; Dockery, D.W.; Schwartz, J.; Zanobetti, A.
(2000b) The national morbidity, mortality, and air pollution study. Part II: morbidity, mortality, and air
pollution in the United States. Cambridge, MA: Health Effects Institute: research report no. 94.
Samet, J. M.; Domenici, F.; Curriero, F.; Coursac, I.; Zeger, S. L. (2000c) Fine particulate air pollution and mortality
in20 U.S. cities, 1987-1994. N. Engl. J. Med. 343:1742-9.
Samat, J. A.; Koutrakis, P.; Sun, H. H. (2000) Assessing the relationship between personal particulate and gaseous
exposures of senior citizens living in Baltimore, MD. J. Air Waste Manage. Assoc. 50:1184-1198.
Sarnat, J. A; Schwartz, J.; Catalano, P. J.; Suh. H. H. (2001) Gaseous pollutants in particulate matter epidemiology:
confounders or surrogates? Environ. Health Perspect. 109:1053-1061.
January 2005
3-74
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
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
Schwartz, J. (1993) Air pollution and daily mortality in Birmingham, Alabama. Am. J. Epidemiol. 137:1136-1147.
Schwartz, J. (1997) Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology
8:371-377.
Schwartz, J. (2000a) Assessing confounding, effect modification, and thresholds in the association between ambient
particles and daily deaths. Environ. Health Perspect. 108:563-568.
Schwartz, J.; Dockery, D. W.; Neas, L. M.; Wypij, D.; Ware, J. H.; Spengler, J. D.; Koutrakis;P.; Speizer, F. E.;
Ferris, B. G.( Jr. (1994) Acute effects of summer air pollution on respiratory symptom reporting in children.
Am. J. Respir. Grit. Care Med. 150:1234-1242.
Schwartz, J.; Morris, R. (1995) Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan.
Am. J. Epidemiol 142:23-35.
Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J. Air
Waste Manage. Assoc. 46:927-939.
Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
respiratory health effects in schoolchildren. Epidemiology 11:6-10.
Schwartz, J. (2001) Air pollution and blood markers of cardiovascular risk. Environ. Health Perspect.
Suppl. 109(3): 405-409.
Schwartz, J.; Zanobetti, A.; Bateson, T. (2003) Morbidity and mortality among elderly residents of cities with daily
PM measurements. In: Revised analyses of time-series studies of air pollution and health. Special report'
Boston, MA: Health Effects Institute; pp. 25-58. Available:
h%://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
Schwartz, J. (2000a) Assessing confounding, effect modification, and thresholds in the association between ambient
particles and daily deaths. Environ Health Perspect. 108: 563-568.
Schwartz, J. (2003a) Daily deaths associated with air pollution in six US cities and short-term mortality displacement
in Boston. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 219-226. Available: htrp://www,healtheffects.org/FTibs/TlmeSeries.pdf
[18 October, 2004].
Schwartz, J. (2003b) Airborne particles and daily deaths in 10 US cities. In: Revised analyses of time-series studies of
air pollution and health. Special report. Boston, MA: Health Effects Institute; pp. 211-218. Available:
http://www.healtheffects.org/Pubs/TimeSeries.pdftl8 October, 2004].
Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects of ambient air pollution onnonelderly
asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
1987-1994. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18
October, 2004].
Smith, R. L.; Spitzner, D.; Kim, Y.; Fuentes, M. (2000) Threshold dependence of mortality effects for fine and coarse
particles in Phoenix, Arizona. J. Air Waste Manage. Assoc. 50: 1367-1379.
January 2005
3-75
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
33
34
35
Stieb, D. M.; Beveridge, R. C.; Brook, J. R,; Smith-Doiron, M.; Burnett, R. T.; Dales, R. E.; Beaulieu, S.; Judek, S.;
Mamedov, A. (2000) Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint
John, Canada. J. Exposure Anal. Environ. Epidemiol.: 10: 461-477.
Styer, P.; McMillan, N.; Gao, F.; Davis, J.; Sacks, J. (1995) Effect of outdoor airborne particulate matter on daily
death counts. Environ Health Perspect. 103:490-497.
Thurston, G. D.; Ito, K.; Hayes, C. G.; Bates, D. V.; Lippmann, M. (1994) Respiratory hospital admissions and
summertime haze air pollution in Toronto, Ontario: Consideration of the role of acid aerosols. Environ. Res.
65:271-290.
Tiittanen, P.; Timonen, K. L.; Ruuskanen, J.; Mime, A.; Pekkanen, J. (1999) Fine particulate air pollution,
resuspended road dust and respiratory health among symptomatic children. Eur. Respir. J. 13: 266-273.
Tolbert, P. E.; Mulholland, J. A.; Macintosh, D. L.; Xu, F.; Daniels, D.; Devine, O. J.; Carlin, B. P.; Klein, M.;
Dorley, J.; Butler, A. J.; Nordenberg, D. F.; Frurnkin, H.; Ryan, P. B.; White, M. C. (2000) Air quality and
pediatric emergency room visits for asthma in Atlanta, Georgia. Am. J. Epidemiol. 151:798-810.
Tsai, F. C.; Apte, M. G.; Daisey, J. M. (2000) An exploratory analysis of the relationship between mortality and the
chemical composition of airborne particulate matter. Inhalation Toxicol. 12(suppl.): 121-135.
Vedal, S.; Petkau, J.; White, R.; Blair, J. (1998) Acute effects of ambient inhalable particles in asthmatic and
nonasthmatic children. Am. J. Respir. Crit. Care Med. 157: 1034-1043.
Villeneuve, P. J.; Goldberg, M. S.; Krewski, D.; Burnett, R. T.; Chen, Y. (2002) Fine particulate air pollution and
all-cause mortality within the Harvard six-cities study: variations in risk by period of exposure. Ann.
Epidemiol. 12: 568-576.
Yang, W.; Jennison, B. L.; Omaye, S. T. (1997) Air pollution and asthma emergency room visits in Reno, Nevada.
Inhalation Toxicol. 9: 15-29.
Zanobetti, A.; Schwartz, J. (2003a) Airborne particles and hospital admissions for heart and lung disease. In: Revised
analyses of time-series studies of air pollution and health. Special report. Boston, MA: Health Effects
Institute; pp. 241-248. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004].
January 2005
3-76
Draft - Do Not Cite or Quote
-------
o
4. CHARACTERIZATION OF HEALTH RISKS
1 4.1 INTRODUCTION ,
2 This chapter describes and presents the 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 1997 PM NAAQS review. This
* 6 updated assessment includes estimates of (1) the risks of mortality, morbidity, and symptoms
7 associated with recent ambient PM2 5 PM,0_2 5 and PM10 levels, (2) the risk reductions associated
8 with just meeting the current suite of PM2 5 NAAQS, and (3) the risk reductions associated with
9 just 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 and discussed in Chapter 5 of mis
11 draft Staff Paper. The risk assessment discussed in this Chapter is more fully described and
12 presented in the draft technical support document, Paniculate Matter Health Risk Assessment for
13 Selected Urban Areas: Draft Report (Abt Associates, 2005; henceforth referred to as the
14 Technical Support Document (TSD) and cited as TSD).
15 An important issue associated with any population health risk assessment is the
16 characterization of uncertainty and variability. Uncertainty refers to the lack of knowledge
17 regarding both the actual values of model input variables (parameter uncertainty) and the
18 physical systems or relationships (model uncertainty - e.g., the shapes of concentration-response
19 (C-R) functions). In any risk assessment uncertainty is, ideally, reduced to the maximum extent
20 possible, but significant uncertainty often remains. It can be reduced by improved measurement
21 and improved model formulation. In addition, the degree of uncertainty can be characterized,
22 sometimes quantitatively. For example, the statistical uncertainty surrounding the estimated
23 PM25 and PM10.2 5 coefficients in the reported C-R functions is reflected in the confidence '
24 intervals provided for the risk estimates in this chapter and in the TSD. Additional uncertainties
25 are addressed quantitatively through sensitivity analyses and/or qualitatively and are discussed in
26 more detail in section 4.2.7.
January 2005
4-1
Draft - Do Not Quote or Cite
-------
1 As noted above, the updated risk assessment presents qualitative and quantitative
2 considerations of uncertainty, including sensitivity analyses of key individual uncertainties.
3 Given the existing data gaps in the scientific evidence and associated uncertainties, a more
4 comprehensive integrated assessment of uncertainties, while desirable, would require use of
5 techniques involving elicitation of probabilistic judgments from health scientists. While the
6 Agency is currently developing these approaches, such comprehensive assessments of
7 uncertainty are not available for the current risk assessment for this PM NAAQS review.
8 Variability refers to the heterogeneity in a population or parameter. For example, there
9 may be variability among C-R functions describing the relation between PM2 5 and mortality
10 across urban areas. This variability may be due to differences in population (e.g., age
11 distribution), population activities that affect exposure to PM (e.g., use of air conditioning),
12 levels and composition of PM and/or co-pollutants, and/or other factors that vary across urban
13 areas.
14 The current risk assessment incorporates some of the variability in key inputs to the
15 assessment by using location-specific inputs (e.g., location-specific C-R functions, baseline
16 incidence rates, and air quality data). Although spatial variability in these key inputs across all
17 U.S. locations has not been fully characterized, variability across the selected locations is
18 imbedded in the assessment by using, to the extent possible, inputs specific to each urban area.
19 Temporal variability is more difficult to address, because the risk reduction portions of the risk
20 assessment (i.e., estimated risk reduction associated with just meeting specified standards) focus
21 on some unspecified time in the future when specified PM standards are just met. To minimize
22 the degree to which values of inputs to the assessment may be different from the values of those
23 inputs at that unspecified time, we have used the most current inputs available (i.e., year 2003 air
24 quality data for most locations and the most recent available mortality baseline incidence rates
25 (from 2001)). However, we have not tried to predict future changes in inputs (e.g., future
26 population levels or possible changes in baseline incidence rates).
27 The goals of the updated PM risk assessment are: (1) to provide estimates of the potential
28 magnitude of PM-associated mortality and morbidity associated with current PM25, and PM10.25
29 levels and with attaining the current suite of PM2 5 NAAQS (as well as the additional estimated
January 2005
4-2
Draft - Do Not Quote or Cite
-------
1 reductions in effects associated with alternative PM2 5 and PM10.2 5 standards identified as part of
2 this review) in specific urban areas,1 (2) to develop a better understanding of the influence of
3 various inputs and assumptions oh the risk estimates (e.g., choice of policy-relevant background
4 (PRB) levels, consideration of potential hypothetical thresholds), and (3) to gain insights into the
5 nature of the risks associated with exposures to ambient PM (e.g., patterns of risk reduction
6 associated with meeting alternative annual and daily standards). The staff recognizes that due to
7 the many sources of uncertainty inherent in conducting the PM risk assessment, the resulting PM
8 risk estimates should not be viewed as precise measures of the health impacts now occurring or
9 anticipated to occur in the future in any given location or nationally. Further, the staff
10 recognizes that the role of the risk assessment in this standards review must take into account the
11 significant uncertainties associated with this assessment, discussed in section 4.2.7 below.
12
13 4.1.1 Summary of Risk Assessment Conducted During Prior PM NAAQS Review
14 For the last review cycle, EPA conducted a health risk assessment that estimated
15 population risk for two defined urban study areas: Philadelphia'and Los Angeles counties. The
16 PM health risk model combined information about daily PM air quality for these two study areas
17 with estimated concentration-response (C-R) functions derived from epidemiological studies and
18 baseline health incidence data for specific health endpoints to derive estimates of the annual
19 incidence of specific health effects occurring under "as is" air quality.2 Since site-specific
20 relative risks were not available for all endpoints in both locations (and in the absence of more
21 information concerning which individual studies might best characterize the health risk in a
22 given location), a form of meta analysis (referred to as a "pooled analysis") was conducted
23 which combined the results of the studies that met specified criteria. The assessment also
24 examined the reduction in estimated incidence that would result upon just attaining the existing
25 PM10 standards and several sets of alternative PM2 5 standards. In addition, the assessment
26 included sensitivity analyses and integrated uncertainty analyses to better understand the,
'Risk estimates associated with ciirrent PM10 levels also have been included in an appendix to the TSD for
those urban areas where PM2 5 risks have been estimated to provide, additional context.
2«
:As is" PM concentrations are defined here as a recent year of air quality.
January 2005
4-3
Draft - Do Not Quote or Cite
-------
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
influence of various inputs and assumptions on the risk estimates. The methodological approach
followed in conducting the prior risk assessment is described in Chapter 6 of the 1996 Staff
Paper (EPA, 1996b) and in several technical reports (Abt Associates, 1996; Abt Associates,
1997a,b) and publications (Post et al., 2000; Deck et al., 2001).
Summarized below are the key observations resulting from the prior risk assessment
which were most pertinent to the 1997 decision on the PM NAAQS, as well as several important
caveats and limitations associated with that assessment:
• EPA placed greater weight on the overall qualitative conclusions derived from the health
effect studies - that ambient PM is likely causing or contributing to significant adverse
effects at levels below those permitted by the existing PM10 standards - than on the
specific C-R functions and quantitative risk estimates derived from them. The
quantitative risk estimates included significant uncertainty and, therefore, were not
viewed as demonstrated health impacts. Nevertheless, EPA did state that it believed the
assessment presented reasonable estimates as to the possible extent of risk for these
effects given the available information (62 FR at 38656).
• Consideration of key uncertainties and alternative assumptions resulted in fairly wide
ranges in estimates of the incidence of PM-related mortality and morbidity effects and
risk reductions associated with attainment of alternative standards in both locations in the
risk assessment. Significantly, the combined results for these two cities alone found that
the risk remaining after attaining the current PM10 standards was on the order of hundreds
of premature deaths each year, hundreds to thousands of respiratory-related hospital
admissions, and tens of thousands of additional respiratory-related symptoms in children
(62 FR at 38656).
• Based on the results from the sensitivity analyses of key uncertainties and the integrated
uncertainty analyses, the single most important factor influencing the uncertainty
associated with the risk estimates was whether or not a threshold concentration exists
below which PM-associated health risks are not likely to occur (62 FR at 38656).
• Over the course of a year, the few peak 24-hour PM2 5 concentrations appeared to
contribute a relatively small amount to the total health risk posed by the entire air quality
distribution as compared to the aggregated risks associated with the low to mid-range
PM25 concentrations (62 FR at 38656).
• There was greater uncertainty about both the existence and the magnitude of estimated
excess mortality and other effects associated with PM2 5 exposures as one considered
lower concentrations that approach background levels (62 FR at 38656).
January 2005
4-4
Draft - Do Not Quote or Cite
-------
i •: - - . •
2 • Based on the results from the sensitivity analyses of key uncertainties and/or the
3 integrated uncertainty analyses, the following uncertainties had a much more modest ..
4 impact on the risk estimates: the use of C-R functions from multi-pollutant, rather than
5 single-pollutant models; the choice of approach to adjusting the slope of the C-R
6 functions in analyzing alternative cutpoints (i.e., hypothesized thresholds); the value
7 chosen to represent average annual background PM concentrations; and the choice of
8 approach to adjusting air quality distributions for simulating attainment of alternative
9 PM25 standards (EPA, 1996b).
10
11 4.1.2 Development of Updated Assessment
12 The scope and methodology for the updated PM risk assessment have been developed
13 over the last three years. In June 2001, OAQPS released a draft document, PMNAAQS Risk
14 Analysis Scoping Plan, (EPA, 2001 c) describing EPA's overall plan for conducting the PM-
•i
15 health risk assessment for the current review. The CAS AC PM Panel provided feedback on this
16 draft plan in a consultation held July 24,2001, and the Agency also received comments from the
17 general public. In January 2002, EPA released a draft document, Proposed Methodology for
v»> ' • "
18 Paniculate Matter Risk Analyses for Selected Urban Areas, (Abt Associates, 2002) for public
" l '
19 and CASAC review. This draft document described EPA's plans to conduct a risk assessment
20 for PM2 5-related risks for several health endpoints, including mortality, hospital admissions, and
21 respiratory symptoms, and PM,0.2 s-related risks for hospital admissions and respiratory
22 symptoms. The CASAC PM Panel discussed this draft document in a February 27,2002
23 teleconference and provided its comments in a May 23, 2002 Advisory letter to EPA's
24 Administrator (Hopke, 2002). OAQPS also received several comments from the public. In its
25 May 23, 2002.Advisory, the CASAC PM Panel "concluded that the general methodology as
26 described in the report is appropriate.... Thus, the general framework of the approach is the
27 sensible approach to this risk analysis" (Hopke, 2002). Among its comments, the CASAC Panel
28 • suggested extending the risk assessment to include PM10 (Hopke, 2002).
29 In response to a request from CASAC to provide additional details about the proposed •
30 scope of the PMi0.25 and PM10 components of the planned risk assessment, in April 2003 EPA
31 released a draft memorandum (Abt, 2003a) to the CASAC and the public addressing this topic.
32 On May 1,2003, the CASAC PM Panel held a consultation with EPA to provide advice on staff
January 2005
4-5
Draft - Do Not-Quote or Cite
-------
1 plans for conducting the PM,0.2.5 and PM10 components of the health risk assessment. In August
2 2003 OAQPS released a draft technical report describing its draft PM risk assessment (Abt
3 Associates, 2003b) 'in conj unction with the 1sl draft Staff Paper.3 The CAS AC provided its
4 comments on the draft PM risk assessment in its letter to the Administrator (Hopke, 2004). The
5 revised draft risk assessment discussed in this Chapter and in the TSD (Abt Associates, 2005)
6 has taken into consideration the CAS AC and public comments received on the 2003 drafts and
7 the evaluation of the health effects literature contained in the final CD.
8
9 4.2 GENERAL SCOPE OF PM RISK ASSESSMENT
10 As discussed in Chapter 3 above, the CD concludes (p. 9-79) that "a growing body of
11 evidence both from epidemiological and toxicological studies also supports the general
12 conclusion that PM2S (or one or more PM25 components), acting alone and/or in combination
13 with gaseous co-pollutants are likely causally related to cardiovascular and respiratory mortality
14 and morbidity." With respect to PM10.25, the CD (p.9-80) finds that there is "a much more
15 limited body of evidence ... suggestive of associations between short-term (but not long-term )
16 exposures to ambient coarse-fraction thoracic particles... and various mortality and morbidity
17 effects observed at times in some locations." The CD further concludes that there is somewhat
18 stronger evidence for coarse-fraction particle associations with morbidity (especially respiratory)
19 endpoints than for mortality. As discussed in greater detail in Chapter 3, the evidence relating
20 PM10.25 concentrations and premature mortality is equivocal and, therefore, the quantitative risk
21 assessment presented here and included in the TSD (Abt Associates, 2005) only includes
22 morbidity health endpoints for PM]0.2 5. The PM10.2 5 risk assessment is more limited than the
23 PM25 assessment because of the more limited air quality data as well as the smaller number of
24 studies for which there is sufficient evidence to use in this assessment.
25 The updated risk assessment being conducted for the current NAAQS review is premised
26 on the assumption that elevated ambient PM2 5 concentrations are causally related to the
27 mortality, morbidity, and symptomatic effects (alone and/or in combination with other
We hereafter refer to the "PM risk assessment" unless reference to a specific PM indicator (e.g., PMj 5) is
required. The current PM risk assessment primarily focuses on two PM indicators — PM2 5 and PMIM 5.
January 2005
4-6
Draft - Do Not Quote or Cite
-------
1 pollutants) observed in the epidemiological studies. Similarly, the risk assessment for PM10,25 is
2 premised on the assumption that elevated ambient PM10,25 concentrations are causally related to
3 morbidity and symptomatic effects observed in the epidemiological studies. Staff concludes that
4 these assumptions are well supported by the evaluation contained in the CD and is'consistent
5 with the advice provided by the CAS AC PM Panel. However, staff recognizes that there are
6 varying degrees of uncertainty associated with whether or not there is a causal relationship for
7 each of the PM indicators and the specific health endpoints (e.g., cardiovascular hospital
8 admissions, COPD hospital admissions) and that the degree of uncertainty is directly related to
9 • differences in the relative weight of evidence.
10 This PM25 risk assessment focuses on selected health endpoints such as increased excess
11 daily mortality and mortality associated with long-term exposure, and increased hospital
12 admissions for respiratory and cardiopulmonary causes and increased respiratory symptoms for
13 children associated with short-term exposure. The PM10.2 5 risk assessment includes increased
14 hospital admissions for respiratory and cardiopulmonary causes and increased respiratory
15 symptoms for children associated with short-term" exposure. A consequence of limiting the
16 assessment to these selected health endpoints is that the risk estimates likely understate the type
17 and extent of potential health impacts of ambient PM exposures. Although the risk assessment
18 does not address all health effects for which there is some evidence of association with exposure
19 to PM, the broad range of effects are identified and considered previously in Chapter 3.
20 Like the prior risk assessment done as part of the last review (EPA, 1996b), this current
21 updated risk assessment uses C-R functions from epidemiological studies based on ambient PM
22 concentrations measured at fixed-site, community-oriented, ambient monitors.' As discussed
23 earlier in Chapter 2 (section 2.7) and Chapter 3 (section 3.6.2), measurements of daily variations"
24 of ambient PM concentrations, as used in the time-series studies that provide the C-R
25 relationships for this assessment, have a plausible linkage to the daily variations of exposure to
26 ambient PM2 5 for the populations represented by ambient monitoring stations. The CD
27 concludes that "at this time, the use of ambient PM concentrations as a surrogate for exposures is
28 not expected to change the principal conclusions from PM epidemiological studies that use
29 community average health and pollution data" (CD, p. 5-121). A more detailed discussion of the
January 2005
4-7
Draft - Do Not Quote or Cite
-------
1 possible impact of exposure misclassification on the estimated C-R relationships derived from
2 the community epidemiological studies is presented above in Chapter 3 (see section 3.6.2).
3 While quantitative estimates of personal or population exposure do not enter into
4 derivations of the PM risk estimates for this review, an understanding of the nature of the
5 relationships between ambient PM and its various components and human exposure underlies the
6 conceptual basis for the risk assessment. Unlike recent reviews for 03 and CO, where exposure
7 analyses played an important role, a quantitative exposure analysis will not be conducted as part
8 of this review since the currently available epidemiology health effects evidence relates ambient
9 PM concentrations, not exposures, to health effects. As discussed in Chapter 5 of the CD, EPA
10 and the exposure analysis community are working to improve exposure models designed
11 specifically to address PM, Both EPA and the broader scientific community also are in the
12 process of collecting new information in PM exposure measurement field studies that will
13 improve the scientific basis for exposure analyses that may be considered in future reviews.
14 While the NAAQS are intended to provide protection from exposure to ambient PM,
15 EPA recognizes that exposures to PM from other sources (i.e., non-ambient PM) also have the
16 potential to affect health. The EPA's Office of Radiation and Indoor Air and other Federal
17 Agencies, such as the Consumer Product Safety Commission (CPSC) and the Occupational
18 Safety and Health Administration (OSHA), address potential health effects related to indoor,
19 occupational, environmental tobacco smoke, and other non-ambient sources of PM exposure. As
20 with the prior PM risk assessment, contributions to health risk from non-ambient sources are
21 beyond the scope of the risk assessment for this NAAQS review.
22 This current PM health risk assessment is similar in many respects to the one conducted
23 for the last PM NAAQS review. Both the prior and the current PM risk assessment:
24 • estimate risks for the urban centers of example cities, rather than attempt a nationwide
25 assessment;
26
27 • analyze risks for a recent 12-month period of air quality (labeled "as is") and for
28 scenarios in which air quality just meets the current set of standards;
29
30 • analyze additional reductions in risks for scenarios in which air quality is simulated to
31 just meet potential alternative standards that are recommended by staff for consideration;
January 2005
4-8
Draft - Do Not Quote or Cite
-------
2 • estimate risks only for concentrations exceeding estimated background levels or the
3 lowest measured level (LML) observed in the study, if it is higher than the estimated
4 background level in the assessment location; and
s ' ; •••
6 • present qualitative and quantitative considerations of uncertainty, including sensitivity
7 analyses of key individual uncertainties.
8 .
9 Both the prior and the current PM risk assessment focus on health endpoints for which C-
10 R functions have been estimated in epidemiological studies. Since these studies estimate C-R
11 functions using air quality data from fixed-site, community-oriented monitors, the appropriate
12 application of these functions in a PM risk assessment similarly requires the use of air quality
13 data from fixed-site, community-oriented, ambient monitors. This is identical to the approach
14 taken in the last PM N AAQS review.
15 The current risk assessment includes risk estimates for 9 urban areas for PM2 5 and 3
16 urban areas for PM10.2 5 In addition, to provide some additional context, PM10risk estimates are
17 provided in Appendix 1 of Abt Associates (2005), for the same urban areas and short-term
18 exposure health endpoints for which PM2 s and PM,0.2 s risk estimates are available. As
19 discussed in section 4.2.2. these areas have been chosen based on availability of PM C-R
20 relationships, adequate PM air quality data, and baseline incidence data. The selection of these
21 areas also reflects a desire to include areas from the various regions of the United States to the
22 extent possible in order to reflect regional differences in the composition of PM and other factors
23 (e.g., different levels of co-pollutants, air-conditioning use).
*•* - '
24 A C-R relationship estimated ,by an epidemiological study may not be representative of
25 the relationship mat exists outside the range of concentrations observed during the study. To
26 partially address this problem, risk was not calculated for PM levels below the LML in the study,
27 if reported. The LML's for each study that provided a C-R relationship for the current PM risk
28 assessment, where available, are provided in Appendix 4A.
29 For long-term exposure mortality associated with PM25, the LMLs for the relevant PM25
30 epidemiology studies are 7.5,10, and 11 jig/m3, for the ACS-extended, ACS, and Six Cities
31 studies, respectively. These LMLs are higher than the range of estimated PM25 background
32 levels in either the Eastern or Western U.S.. Estimating risks outside the range of the original
January 2005 4-9 Draft-Do Not Quote or Cite
-------
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
epidemiology studies that were the source of the C-R functions would introduce significant
additional uncertainties into the risk assessment Therefore, the risks associated with long-term
exposure were only estimated in excess of the LML. Since we do not estimate risks below the
LML, the overall long-term exposure mortality risks would be underestimated to the extent that
annual average PM25 concentrations below the LMLs contribute to long-term exposure
mortality. Where the LML for the epidemiology study that served as the basis for the C-R
relationship was either below the estimated background PM concentration for an area or was not
available, risks were only estimated above background PM concentrations. The rationale for this
choice is that risks associated with concentrations above background are judged to be more
relevant to policy decisions about the NAAQS than estimates that include risks potentially
attributable to uncontrollable background PM concentrations.
The following sections provide an overview of the components of the risk model,
describe the selection of urban areas and health endpoints included in the PM risk assessment,
discuss each of the major components of the risk model, address characterization of uncertainty
and variability associated with the risk estimates, and present key results from the assessment A
separate TSD (Abt Associates, 2005) is available which provides a more detailed discussion of
i
the risk assessment methodology and additional risk estimates.
4.2.1 Overview of Components of the Risk Model
In order to estimate the incidence of a particular health effect associated with "as is"
conditions in a specific county or set of counties attributable to ambient PM2 5 (or PM10_2 5)
exposures in excess of background and the change in incidence of the health effect in that county
or set of counties corresponding to a given change in PM2 5 (or PM10.2.5) levels resulting from just
meeting a specified set of PM25 (or PMi0.25) standards, the following three elements are required:
air quality information including: (1) "as is" air quality data for PM2.5 and PM10_2 5 from
ambient monitors for the selected location, (2) estimates of background PM2.5 and PM10.25
concentrations appropriate for that location, and (3) a method for adjusting the "as is"
data to reflect patterns of air quality estimated to occur when the area just meets a given
set of PM2 5 (or PM,o.25) standards;
January 2005.
4-10
Draft - Do Not Quote or Cite
-------
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
• relative-risk based C-R functions (preferably derived in the assessment location) which
provide an estimate of the relationship between the health endpoints of interest and
ambient PM concentrations; and
• annual or seasonal baseline health effects incidence rates and population data which are
needed to provide an estimate of the annual or seasonal baseline incidence of health
effects in an area before any changes in PM air quality.
•
Figure 4-1 provides abroad schematic depicting the role of these components in the risk
assessment. Those points where EPA has conducted analyses of alternative assumptions,
procedures, or data are indicated by a diamond with Sx in it. A summary description of the type
of sensitivity analyses performed is included later in section 4.2.7 (See Table 4-8). Each of the
three key components (i.e., air quality information, estimated PM-related C-R functions, and
baseline incidence) are discussed below, highlighting those points at which judgments have been
made.
+ '
4.2.2 Criteria for Selection of Health Endpoints and Urban Study Areas
Only two urban counties were included in the risk assessment conducted for the prior PM
NAAQS review due to the very limited number of urban areas that had sufficient recent PM25
ambient air quality monitoring data and because of the limited number of epidemiological
studies that directly measured PM2 5. As discussed in more detail in Chapter 3, since the last
review, a significant number of epidemiological studies have been published examining a variety
of health effects associated with ambient PM15, PM10.2 5, and PM10 in various urban areas
throughout the U.S. and Canada, as well as Europe and other parts of the world. While a
significant number of new epidemiological studies have been published since the last review, •
and are evaluated in the CD, the PM risk assessment relies only on U.S. arid Canadian studies to
limit introducing uncertainty associated with the possible differences in population and
characteristics of PM and co-pollutants between the U.S. and Canada and these other locations.
The approach and criteria that EPA has used to select the health endpoints and urban areas to.
include in the risk assessment for the PM indicators are described below.
January 2005
4-11
Draft - Do Not Quote or Cite
-------
w SB 2 >«
0) 3 ^ ."S
fill
fe
1
I
•
O T—I
-------
1 4.2.2.1 Selection of Health Endpoint Categories
2 As discussed in Chapter 3, OAQPS staff carefully reviewed the health effects evidence
3 evaluated in the CD in order to identify potential health effect categories to include in the current
4 PM risk assessment. Given the large number of endpoints and studies addressing PM-related
5 effects, staff recommended for inclusion in the PM risk assessment only the more severe and
6 better understood (in terms of health consequences) health endpoint categories for which the
7 overall weight of the evidence from the collective body of studies supports the conclusion that
8 there is likely to be a causal relationship between PM and the health effects category and for
9 which baseline incidence data were available. In addition, for the three PM indicators (PM2 5,
10 PM10, PM10_2 5), staff considered only those endpoint categories which provided C-R relationships
11 based on U.S. and Canadian studies that used PM concentrations obtained by one of the
12 following approaches: (1) directly measuring fine particles using PM2.5 or PM2j, (2) estimating
13 the concentration of fine particles vising nepholometry data, and (3) estimating PM10.2 5
14 concentrations based on co-located PMlo and PM2 5 monitors or based on measurements using
15 dichotomous samplers.
16 Based on a review of the evidence evaluated in the CD and discussed in Chapter 3, as
17 well as the criteria discussed above, staff included the following broad categories of health
18 endpoints in the risk assessment for PM2 5:
19 related to short-term exposure:
20 • total (non-accidental), cardiovascular, and respiratory mortality;
21 • hospital admissions for cardiovascular and respiratory causes;
22 • respiratory symptoms not requiring hospitalization
23 . related to long-term exposure:
24 • total, cardiopulmonary, and lung cancer mortality.
25 •
26 Other effects reported to be associated with PM2 s, including, but not limited to, decreased lung
27 function, changes in heart'rate variability, and increased emergency room visits are addressed in
28 Chapter 3, but are not included in the quantitative risk assessment.
January 2005
4-13
Draft - Do Not Quote or Cite
-------
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
Based on a review of the evidence evaluated in the CD and discussed in Chapter 3, as
well as the criteria discussed above, staff included the following categories of health endpoints
associated with short-term exposures in the risk assessment for PM10.2 5:
hospital admissions for cardiovascular and respiratory causes;
• respiratory symptoms.
As discussed in Chapter 3 (section 3.4), more equivocal evidence is available for other health
responses, such as associations between short-term exposure to PM10_2 5 and mortality. Staff
believe that these health endpoints, which are based on less certain evidence, are not appropriate
for inclusion in the quantitative risk assessment. Staff have considered these endpoints in more
qualitative assessments of the evidence presented in Chapter 3.
4.2.2.2 Selection of Study Areas
A primary goal of the current PM risk assessment has been to identify and include urban
areas in the U.S. for which epidemiological studies are available that estimate'C-R relationships
for those locations. This goal is in large part motivated by the evaluation contained in the CD
and staff assessment in Chapter 3 that suggests there may be geographic variability in C-R
relationships across different urban areas in the U.S. The selection of urban areas to include in
the PM risk assessment was based on the following criteria:
An area had sufficient air quality data for a recent year (1999 or later). Sufficient PM2.5
data is defined as having at least one PM monitor at which there are at least 11
observations per quarter for a one year period.4 Sufficient air quality data for PMJCU2 $ is
defined as a one year period with at least 11 daily values per quarter based on data from
co-located PM25 and PMIO monitors. The criterion of at least 11 observations per quarter
is based on EPA guidance on measuring attainment of the daily and annual PM standards
and is contained in Appendix N of the July 18,1997 Federal Register notice.
An area is the same as or close to the location where at least one C-R function, for one of
the recommended health endpoints, has been estimated by a study that satisfies me study
selection criteria (see below).
Tor 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.
January 2005 4-14 Draft - Do Not Quote or Cite
-------
1 • An area is one in which studies exist that had relatively greater precision, as indicated by
2 a relatively greater number of effect-days observations.
3 ,
4 • -Where an area was considered based on PM-related hospital admission effects; an area
5 had relatively recent area-specific baseline incidence data.
6
7 For the PM2 5 risk assessment, staff focused on selecting urban areas based primarily on a
8 location's having non-accidental total and cause-specific mortality PM25 C-R functions since this
9 was the largest data base in terms of number of studies in different locations. Staff then
10 supplemented this by consideration of other morbidity endpoints (i.e., hospital admissions). •
11 Based on a review of studies listed in Tables 8 A and 8B of the CD (see also Appendices 3-A and
12 3-B of this SP), a candidate pool of 17 urban locations was initially suggested based on short-
13 term exposure mortality studies (16 of the candidate locations); Seattle was added based on a
14 hospital admissions study.5 -
15 Staff next considered an indicator of study precision for the urban areas associated with
16 the short-term exposure mortality studies identified in the first step. As discussed above in
17 Chapter 3 (section 3.3.1.1) and in Chapter 8 of the CD '(pp. 8-324 - 8-325), the natural logarithm
18 of the mortality-days (a product of each city' s daily mortality rate and the number of days for'
19 which PM data were available) can be used as a rough indicator of the degree of precision of
20 effect estimates; studies with larger values for this indicator should be accorded relatively greater
21 study weight. While there is no bright line for selecting any particular cutoff, it was the staffs
22 judgment to consider only those urban areas in which studies with relatively greater precision
23 were conducted, specifically including studies that have a natural log of mortality-days greater
24 than or equal to 9.0 for total non-accidental mortality^ As a result of applying this criterion, six
25 urban areas Were excluded as potential study areas (Camden, NJ; Coachella Valley, CA;
26 Elizabeth, NJ; Newark, NJ; Steubenville, OH; and Topeka, KS).
5The Tolbert et al. (2000) study in Atlanta was excluded from consideration because the CD urged caution
in interpreting these preliminary results given the incomplete and variable nature of the databases analyzed.
^ost 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.
January 2005 ' 4-15 Draft - Do Not Quote or Cite
-------
1 Finally, staff considered which of the potential study locations identified from steps 1 and
2 t 2 above also had sufficient PM2 5 ambient monitoring data. A location was considered to have
3 sufficiently complete air quality data if it had at least one monitor at which there were at least 11
4 observations per quarter and at least 122 observations per year (i.e., equivalent to 1 in 3 day
5 monitoring). This final criterion excluded two of the remaining potential study areas (Knoxville,
6 TN and Portage, WI), leaving nine urban areas (i.e., Boston, Detroit, Los Angeles, Philadelphia,
7 Phoenix, Pittsburgh, San Jose, Seattle, and St. Louis) in which epidemiological studies reported
8 C-R relationships for PM2 5 and mortality or hospital admissions and which had sufficient air
9 quality data in a recent year.
10 The PM2 5 risk assessment for long-term exposure mortality was conducted for nine urban
11 areas. Eight of the nine urban areas, excluding Seattle, were already included in the PM2 5 risk
12 assessment based on short-term exposure mortality and are listed above. Since the C-R
13 functions for PM25-related mortality associated with long-term exposure used in the risk
14 assessment are based on differences in long-term PM averages across multiple cities in the U.S.,
15 the issue of matching risk assessment locations with city-specific studies did not arise.
16 Therefore, long-term exposure mortality risk estimates also were developed for Seattle.
17 Most of the short-term morbidity and respiratory symptom studies reporting PM23-related
18 effects were conducted in the same set of locations as the short-term exposure mortality studies.
19 In considering these other health endpoints, staff applied similar criteria (i.e., studies providing
20 effects estimates with relatively greater precision and availability of recent and adequate PM2 5
21 ambient air quality data). In addition, for the hospital admissions effect category, assessment
22 was limited to those urban areas where the necessary baseline incidence data could be obtained.
23 Based on applying the above criteria and considerations, the health endpoints and urban
24 locations selected for the PM2 5 risk assessment are summarized in Tables 4-1 and 4-2, for
25 mortality and morbidity endpoints, respectively. These tables also list the specific studies that
26 provided the estimated C-R functions used in the PM2 5 risk assessment More detailed
27 information on the studies selected can be found in Appendices 3A, 3B, and 4A of this draft
28 Staff Paper and Appendix C of the TSD (Abt Associates, 2005).
29 The selection of urban areas to include for the PM^^ risk assessment was based on
30 examining the pool of epidemiological studies reporting associations for PM10_25 with the
January 2005 4-16 ' Draft - Do Not Quote or Cite
-------
1 morbidity endpoints (hospital admissions and respiratory symptoms) in any of the urban areas
2 . already selected for the PM2 5 risk assessment. As summarized in Table 4-3 and noted earlier,
3 the PM10_25 risk assessment is more limited because of the more limited air quality, data
4 (requiring co-located PM2 5 and PM10 monitors or availability of dichot data) as well as the
5 smaller number of health endpoints and studies. Based on the available data, EPA has included
6 in the PM,0_2 j risk assessment the following health endpoints and locations: increased hospital
7 admissions in Detroit and Seattle, and increased respiratory symptoms in St. Louis. Additional
8 details about the epidemiological studies and the.C-R functions used in the risk assessment based
9 on these studies are provided in Appendices 3A, 3B, and 4A of this draft Staff Paper and
10 Appendix C of the TSD (Abt Associates, 2005).
11 With respect to the PM10.2 5 risk assessment, staff notes that the locations used in this part
12 of the risk assessment are not representative of urban locations in the U.S. that experience the
13 most significant elevated 24-hour PM10.2 5 ambient concentrations. Thus, observations about risk
14 reductions associated with alternative standards in the three urban areas (i.e., Detroit, Seattle,
15 and St. Louis) may not be very relevant to the areas expected to have the greatest health risks
16 associated with peak daily ambient PM10.2 5 concentrations.
17
18 4.2.3 Air Quality Considerations
19 As mentioned earlier, air quality information required to conduct the PM risk assessment
20 includes: (1) "as is" air quality data for PM25 and PM10.25 from suitable monitors for each
21 selected location, (2) estimates of background PM25 and PM10.25 concentrations appropriate for
22 each location, and (3) a method for adj usting the "as is" data to reflect patterns of air quality
23 estimated to occur when an area just meets a given set of PM25 (or PM10.25) standards. OAQPS
24 retrieved ambient air quality data for PM2 5 and PM10 for the potential study areas for the years
, 25 1999 through 2003 from EPA's Air Quality System (AQS). As noted earlier, consistent with
26 EPA guidance, urban areas were only included in the risk assessment if there was at least one
27 monitor with I'l or more observations per quarter. Staff calculated PM10.2 5 concentrations from
28 co-located PM^5 and PM10 monitors that met the minimum observation cutoff. Generally, the
29 most recent year of PM data was used for each study area and PM indicator subject to meeting
30 this requirement..
January 2005 4-17 Draft - Do Not Quote or Cite
-------
.ssessmen
•8
CA
s
=
•O
a
rtality H
w
1
H
!
a°
* 1
•a 3
I |
If
2 9
•< h
1
I
an Locatio
•a
t>
I
(u
o
s
§
2
•a
al. (2000>ACS
;2002>ACS ext
8
ert et a
"8
!
O) t<
O S
< 02
8^
a|
-so.
P-4
•a
ts
£
o
8
•a
ts
1
s
999;
iirl
citi
CS
i '-a
o
fe
a
i
r-l (N
oo
O t-
-------
s
e
1
5
PM
v
«s
•e
€
I
w
W5
•g
.§
u
|
1
ar
I
5
A
O
£
8
£
00
s ^
_j
3
u
O
of
03
•i .s .s .a
£
£
1
S
e
o.
•e
i
t*ia-1
3
5
•s
I
fe
o
1
so
\ o i— i
-------
1 Consistent with the approach used in the last PM risk assessment, a composite monitor
2 data set was created for each assessment location based on averaging the 24-hour values from all
3 monitors eligible for comparison with the standards for each day with any monitoring data. The
4 resulting composite monitor data set provides a single series of daily concentrations for the urban
5 area which serves as the surrogate index of exposure for the urban area. Table 4-4 provides a
6 summary of the PM2 5 and PM10.2.5 ambient air quality data for the urban study areas based on the
7 composite monitor values used in the risk assessment. Additional tables providing more detailed
8 information on PM ambient concentrations for these locations, including the number of
9 observations available on a quarterly and annual basis for each monitor, can be found in
10 Appendix A of the TSD (Abt Associates, 2005).
11 4.2.3.1 Estimating PM Background Levels
12 Background PM concentrations used in the PM risk assessment are defined above in
13 Chapter 2 as the PM concentrations that would be observed in the U.S. in the absence of
14 anthropogenic emissions of PM and its precursors in the U.S., Canada, and Mexico. For the base
15 case risk estimates, the midpoint of the appropriate ranges of annual average estimates for PM2 5
16 background presented in section 2.6 were used (i.e., eastern values were used for eastern study
17 locations and western values were used for western study locations). For PM10.2 5 the
18 approximate mid-point of the annual average estimates for PM10.2 5 background presented in
19 section 2.6 were used. In sensitivity analyses, we examine the impact of assuming 1) a constant
20 background set at the lower and upper ends of the range of estimated background levels for the
21' eastern and western United States, depending on the assessment location and 2) a variable daily
22 PM2 5 background, using distributions whose means are equal to the values used in the base case
23 analysis and whose distributions are based on an analysis of PM25 data from relatively remote
24 sites with the sulfate component removed (see Langstaff (2005)).
25 4.2.3.2 Simulating PM Levels That Just Meet Specified Standards
26 To estimate the health risks associated with j ust meeting the current PM2 5 standards and
27 alternative PM2 5 and PM10_2 5 standards it is necessary to estimate the distribution(s) of PM
28 concentrations that would occur under each specified standard (or sets of standards). Since
29 compliance with the standards is based on a 3-year average, air quality data from 2001 to 2003
30 have been used to determine the amount of reduction in PM2 5 concentrations required to meet
January 2005 4-20 Draft - Do Not Quote or Cite
-------
*
£•
1
«*•*
i
v>
J
\n
b
«g
S
OS
Q
g^
5
u
3
•+ri
1
M
i
o
1
3
Jg
«•*
••I
H
*
V
§
*&
Ml
5 „
^J
0,
£
J3
1
OS
jg
5
--3
• 3 °"
d | 5
al ^
O -M
•5 § oo
s I 1
g u o"
t! ^
fe ^"* C3*
| | '-3
w "efl *S
^j £ 2
|1 |
*g rs G
l! i"
* S !-<" S
11 1 's
II - £ 1
° M «i § U
1 c 1 i i
t"" '" CO
_ a ^
o '5? ^ o '53 £2
u S-l ">• ^ ^
*^l tjjiStt |
iiiffi|ilsfrl
XM rf § -S "-H^iSS0-
illlilfll^li
rills II Ifll'3
2 ^S^JKcuS^wl^M
sScacocac'jcQGacQcnMM
O2UUUUUUUVUU
•g ni-s-gi-sllsl
|lssggas"iss1
•F: s-£-l63ft3*-cfe:ljP6Hi-1.!?
0
1
a
-------
1 the current or alternative suites of standards. The amount of control has then been applied to a
2 single year of data (i.e., 2003, unless otherwise specified) to estimate risks for a single year.
3 Estimated design values (see Table 4-13 later in this Chapter) based on the highest community-
4 oriented monitor within each study area are used to determine the percent adjustment necessary
5 to just meet annual, 98th percentile daily, and 99th percentile daily standards.
6 Under the current annual PM2.5 standard urban areas may, under certain circumstances,
7 use the average of the annual averages of several monitors within an urban area to determine
8 compliance, commonly referred to as the "spatial averaging approach." Therefore, a sensitivity
9 analysis has been conducted for 3 urban areas to allow comparison of the estimated incidence
10 and percent reduction in incidence associated with using either the highest monitor or the spatial
11 average for determining the percent adjustment necessary to just meet the current and alternative
12 annual standards.
13 The percent adjustment to simulate just meeting alternative standards is applied to the
14 composite monitor for the urban area. The composite monitor is used because it is the best
15 surrogate indicator of exposure that matches the type of exposure measure used in the original
16 epidemiological studies.
17 When assessing the risks associated with long-term exposures, which use C-R functions
18 from epidemiological studies that are specified in terms of long-term average concentrations, the
19 annual mean is simply set equal to the standard level. In contrast, when assessing the risks
20 associated with short-term exposures, which use C-R functions from epidemiological studies that
21 consider the sequence of daily average concentrations, the distribution of 24-hour values that
22 would occur upon just attaining a given 24-hour and/or annual PM'Standard has to be simulated.
23 • There are many possible ways to create an alternative distribution of daily concentrations
24 that just meets a specified set of PM standards. Both the assessment conducted during the last
25 NAAQS review (see Abt Associates, 1996, section 8.2) and a more recent analysis of historical
26 air quality data (see Abt Associates, 2005, Appendix B) have found that PM2 5 levels in excess of
27 estimated background concentrations in general have historically decreased in a roughly
28 proportional manner (i.e., concentrations at different points in the distribution of 24-hour PM2 5
29 values in excess of an estimated background concentration have decreased by approximately the
30 same percentage). This suggests that, in the absence of detailed air quality modeling, a
January 2005 . 4-22 Draft - Do Not Quote or Cite
-------
1 reasonable method to simulate PM25 reductions that would result from just meeting a set of
2 standards is to use a proportional adjustment (i.e., to decrease non-background PM levels on all
3 days by the same percentage) for all concentrations exceeding the background level.7 A
4 sensitivity analysis also has been conducted to examine the impact on the PM2.5 risk estimates of
5 an alternative air quality adjustment procedure (e.g., a method that reduces the top 10% of daily
6 PM2.5 concentrations more than the lower 90%).
7 Because the PM10.25historical air quality data are substantially more sparse, there was •
8 insufficient data to carry out the lype of evaluation of historical data that was done for PM2 5to
9 see whether the shape of the distribution of daily values has changed over time^ In the absence
10 of a clearly preferable alternative, the same proportional rollback approach used for PM25 has
(
11 been used for the PM10.25 assessment. This increases the uncertainty about the PM,o.25 risk
•
12 estimates associated with meeting alternative PM10.25 standards. ' '
13 . In assessing health risks associated with PM2 5 and PM10.2 5, air quality just meeting the
14 current or alternative PM25 standards and alternative PM10.2S standards is simulated by reducing
15 the PM2 5 or PM10.2 5 concentrations at the composite monitor by the same percentage on all days.
16 The percentage reduction is determined by comparing the maximum of the monitor-specific
17 annual averages (or the maximum of the monitor-specific ninety-eighth or ninety-ninth
18 percentile daily values, depending on the form of the standard) with the level of the annual (or
19 daily)
20 standard.8" Because pollution abatement methods are applied largely to anthropogenic sources of
21 PM25 or PM10.2.5, rollbacks were applied only to PM25 or PM,0.25 concentrations above estimated
22 background levels. Where sets of standards are considered, as is the case for PM2 5 where both
23 an annual and a daily standard are specified, the percent reduction is determined by the
24 "controlling standard." The "controlling standard" is defined as the standard which would
25 require the greatest reduction in PM levels to just meet the standard. For the current suite of
The portion of the distribution below the estimated background concentration is not rolled back, since air
quality strategies adopted to meet the standards will not reduce the background contribution to PM concentrations.
'since 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 community-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.
January 2005 4-23 Draft - Do Not Quote or Cite
-------
1 PM25 standards, the existing annual standard of 15 jig/m3 is the controlling standard for the five
2 urban study areas (i.e., Detroit, Los Angeles, Philadelphia, Pittsburgh, and St. Louis) that do not
3 meet the current standards. In four of these five urban areas suites of annual standards within the
4 range of 12 to 15 (ig/m3 combined with the current daily standard of 65 jig/m3, using a 98th
5 percentile form, requires the same reduction as when these annual standards are combined with a
6 daily standard of 40 jig/m3, using the same daily form. Therefore, the risk assessment only
7 included the 14 p,g/m3 annual standard combined with the current daily standard for the one
8 location (i.e., Philadelphia) and annual standard scenario where there was a difference in the
9 reduction required between daily standards of 40 and 65 [ig/m3.
10 .
11 4.2.4 Approach to Estimating PM-Related Health Effects Incidence
12 The C-R relationships used in the PM risk assessment are empirically estimated relations
13 between average ambient PM concentrations and the health endpoints of interest reported by
14 epidemiological studies for specific urban areas. Most epidemiological studies estimating
15 relationships between PM and health effects used a method referred to as "Poisson regression" to
16 estimate exponential (or log-linear) C-R functions.9 In this model,
y = B C** (Equation 4-1)
17 where y is the incidence of the health endpoint of interest associated with ambient PM level x, f5
18 is the coefficient of ambient PM concentration, and B is the incidence of the health endpoint at x
19 = 0, i.e.,.when there is no ambient PM2S (or PM10.25) The difference in health effects incidence,
20 Ay = y0 - y, from y0 to the baseline incidence rate, y, that corresponds to a given difference in
21 ambient PM2 5 (or PM10.2S) levels, Ax= x0-x, is then
22
Ay = y[e&* - 1] (Equation 4-2)
r or some studies on respiratory hospital admissions used in the risk assessment a linear C-R function was
estimated.
January 2005 4-24 Draft - Do Not Quote or Cite
-------
1 , or, alternatively,
Ay = y(RRte - 1) (Equation 4-3)
'*
2 where RRto is the relative risk associated with the change in ambient PM2 5 (or PM10.2.5) levels,
3 Ax. Equations 4-2 and 4-3 are simply alternative ways of expressing the relation between a
4 given difference in ambient PM25 (or PM10.25) levels and the corresponding difference in health
5 effects. These equations are the key equations that combine air quality information, C-R
6 information, and baseline health effects incidence information to estimate ambient PM2 5 and
7 jPMjo-as health risk.
8 • For the first part of the risk assessment, characterizing'risks associated with "as is"
9 ambient PM concentrations, Ax is the difference between the as is ambient PM concentration (on
10 each day for the short-term exposure (i.e, daily or 24-hour) endpoints or the annual average for
11 the long-term exposure (i.e., annual average or longer) endpoints and either the estimated PRB
12 concentration or the LML in the epidemiology study providing the p, whichever is greater. For
13 the second part of the risk assessment, characterizing the reduction in health effects incidence
14 associated with alternative PM standards, Ax is the difference between Ihe ambient PM
15 concentration when the current PM standards are just met (on each day for the short-term
16 exposure endpoints or the annual average for the long-term exposure endpoints) and the ambient
17 PM concentration associated with just meeting the specified alternative standards.10
18 For short-term exposure health endpoints, the risk assessment first calculated the daily
19 changes in incidence. Since most areas had at least some days for which no ambient PM
20 concentration data were available, the estimated annual incidence was summed up for each
21 quarter of the year and adjusted by using the ratio of the total number of days in each quarter to
22 the number of days in the quarter for which air quality data was available.u This simple
«
23 adjustment assumes that missing air quality data occur randomly within a quarter and that the
10For those areas already meeting the current PM^ standards, Ax is'the difference between the as is ambient
PM concentration and the ambient PM concentration associated with just meeting the specified standards.
"Adjustment was done on a quarterly basis to reduce possible bias that would be introduced where missing
data are not uniformly distributed throughout the year.
January 2005 • 4-25 Draft - Do Not Quote or Cite
-------
1 distribution of PM concentrations on the days with missing data is essentially the same as the
2 distribution on days for which there are PM data. The quarterly incidence estimates were then
3 summed to derive an annual estimate.
4 The daily time-series epidemiological studies used models estimating C-R functions in
5 which the PM-related incidence on a given day depends only on some specified lagged PM
6 concentration measure (e.g., 0-day lag,-1-day lag, 2-day lag, average of 0- and 1-day lag). As
7 discussed in Chapter 3 (section 3.6.5.2), such models necessarily assume that the longer pattern
8 of PM levels preceding the PM concentration on a given day does not affect mortality on that
9 day. To the extent that PM-related mortality on a given day is affected by PM concentrations
10 over a longer period of time, then these models would be mis-specified, and this mis-
11 specification would affect the predictions of daily incidence based on the model. The extent to
12 which longer-term (i.e., weekly, monthly, seasonal, or annual) PM2 5 exposures affect the
13 relationship observed in the daily time-series studies is unknown. However, there is some
14 evidence, based on analyses of PM10 data, that mortality on a given day is influenced by prior
15 PM exposures up to more than a month before the date of death (Schwartz, 2000a, reanalyzed in
16 Schwartz, 2003b). As indicated in section 3.6.5.2, our use of single day lag models which ignore
17 longer-term influences may result in the risk being underestimated. Currently, there is
18 insufficient information to adjust for the impact of longer-term exposure (on the order of weeks
19 or months) on mortality associated with short-term PM25 exposures and this is an important
20 uncertainty that should be kept in mind as one considers the results from the short-term exposure
21 PM25 risk assessment.
22 The estimated PM2 5-related mortality associated with long-term exposure studies is
23 likely to include mortality related to short-term exposures as well as mortality related to longer-
24 term exposures. As just discussed, estimates of daily mortality based on the time-series studies
25 also are likely to be affected by prior exposures. Therefore, the estimated annual incidences of
26 mortality calculated based on the short- and long-term exposure studies are not likely to be
27 completely independent and should not be added together.
28 The statistical uncertainty surrounding the estimated PM25 and PM10.2s coefficients in the
29 reported C-R functions is reflected in the confidence intervals provided for the risk estimates in
30 sections 4.3 to 4.5. In addition, sensitivity analyses examine how the short- and long-term PM2 5
January 2005 4-26 Draft - Do Not Quote or Cite
-------
1 exposure mortality risk estimates would vary if, instead of the reported C-R relationships,
2 different hypothetical threshold models were applied instead. Another sensitivity analysis
3 addresses how the PM2 5 risk estimates would change if a distributed lag model were applied
4 instead of the single lag models reported in the literature for short-term exposure mortality. "A
5 third sensitivity analysis addresses the possible impact of different assumptions about the role of
6 historical air quality "concentrations in contributing to the reported effects associated with long-
7 term exposure. Finally, PM2 s risk estimates based on alternative model specifications, including
8 the impact of different lags, statistical models (i.e., GAM vs. GLM), and degrees of freedom
.9 allowed (i.e., 30 vs. 100) are shown for short-term exposure mortality and morbidity endpoints in
10 Los Angeles are included in the TSD (Abt Associates, 2005). The results of these sensitivity
11 analyses are discussed in section 4.3 .
12 •' -
13 4.2.5 Baseline Health Effects Incidence Rates and Population Estimates
14 The most common health risk model expresses the reduction in health risk (Ay)
15 associated with a given reduction in PM concentrations (Ax) as a percentage of the baseline
16 incidence (y). To accurately assess the impact of PM air quality on health risk in the selected
17 urban study locations, information on the baseline incidence of health effects (i.e., the incidence
18 under "as is" air quality conditions) and population size in each location is therefore needed.
19 Population sizes, for both total population and various age ranges used in the PM risk assessment
20 were obtained for the year 2000 from the 2000 U.S. Census data12 and are summarized in Table
21 4-5. Where possible, county-specific incidence or incidence rates have been used. County-
22 specific mortality incidences were available for the year 2001 from CDC Wonder (CDC, 2001),
23 an interface for public health data dissemination provided by the Centers for Disease Control
24 (CDC). The baseline mortality rates for each risk assessment location are provided in Table 4-6.
25 County-specific rates for cardiovascular and respiratory hospital discharges, and various
26 subcategories (e.g., pneumonia, asthma), have been obtained, where possible, from state, local,
27 and regional health departments and hospital planning commissions for each of the risk
See http://factfinder.census.gov/.
January 2005 . 4-27 Draft - Do Not Quote or Cite
-------
1 assessment locations.13 Baseline hospitalization rates used in each PM25 and PM10.25 risk
2 assessment location are summarized in Table 4-7. For respiratory symptoms in children, the
3 only available estimates of baseline incidence rates were from the studies that estimated the C-R
4 relationships for those endpoints. However, because the risk assessment locations for these
5 endpoints were selected partly on the basis of where studies were carried out, baseline incidence
6 rates reported in these studies should be appropriate for the risk assessment locations to which
7 they were applied.
8 . ,
9 4.2.6 Concentration-Response Functions Used in Risk Assessment
10 A key component in the risk model is the set of C-R functions which provide estimates of
11 the relationship between each health endpoint of interest and ambient PM concentrations. As
12 discussed above, the health endpoints that have been included in the PM25 risk assessment for
13 short-term exposure include mortality, hospital admissions, and respiratory symptoms not
14 requiring hospitalization and long-term exposure mortality is also estimated. The health
15 endpoints that have been included in the PM]0_2 5 risk assessment for short-term exposure include
16 hospital admissions and respiratory symptoms not requiring hospitalization. These health
17 endpoints were included in the risk assessment because the overall weight of the evidence from
18 the collective body of studies supported the conclusion that there was likely to be a causal
19 relationship between PM and these specific health endpoints. Once it had been determined that a
20 health endpoint was to be included in the assessment, inclusion of a study on that health endpoint
21 to estimate the magnitude of the response was not based on the existence of a statistically
22 significant result. Both single-pollutant and, where available, multi-pollutant, C-R functions
23 were used from the studies listed in Tables 8A and 8B of the CD (see also Appendices 3 A and
24 3B of this SP).
The 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 cany over to the beginning of the current year.
January 2005 4-28 Draft - Do Not Quote or Cite
-------
a
e
•e
i
o
1
eu
o
3
/M
in
*
%)
60
8
o
oo"
^
IS
0.
£
sP
o^
00
tN
00
?
4S
£
8
on
6
'
.«43g
;I31|8
1
1
^.sf
0
fe
a
t
'3
^- (S
-fr «n
-------
t
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Table 4-7. Baseline Hospitalization Rates for PM Risk Assessment Locations*
Health Effect
Detroit1
Los Angeles3
Hospital Admissions (per 100,000 general population/year)
Pneumonia admissions (65 and over): ICD codes 480-486
COPD and asthma admissions (all ages): ICD codes 490-496
COPD and asthma admissions (65 and over): ICD codes 490-
496
Asthma (<65): ICD code 493
Cardiovascular admissions (65 and over): ICD codes: 390-
429
Ischemic heart disease (65 and over): ICD codes 410-414
Dysrhythmias (65 and over): ICD code 427
Congestive heart failure (65 and over): ICD code 428
. 250
—
- 192
~
—
487
161
341
—
318
—
—
i 728
—
— .
, __
Seattle3
-
—
—
—
92
—
—
4—
^_
"Hospitalization rates are presented only for the locations in which the C-R functions were estimated. For each
location, the number of discharges was divided by the location's population from the 2000 U.S. Census estimates to
obtain rates. All incidence rates are rounded to the nearest unit.
1. Wayne County. Year 2000 hospitalization data were obtained from the Michigan Health and Hospital
Association. '
2. Los Angeles County; Year 1999 hospitalization data were obtained from California's Office of Statewide Health
Planning and Development - Health Care Information Resource Center.
3. King County. Year 2000 hospitalization data were obtained from the State of Washington Department of Health,
Center for Health Statistics, Office of Hospital and Patient Data Systems.
January 2005
4-32
Draft - Do Not Quote or Cite
-------
1 As discussed in the CD (section 8.4.2) and Chapter 3 (section 3.6.3), questions were
2 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 time-series studies employing generalized additive models (GAMs) in a
5 commonly used statistical software package. To address these concerns, many of the study
6 authors performed reanalyses of certain of the studies using alternative statistical estimation
7 approaches (e.g., GLM with different degrees of freedom and different types of splines), in
8 addition to using GAMs with a more stringent convergence criterion. To avoid producing a
9 prohibitively large set of results, and based on the earlier staff conclusion in Chapter 3 (section
10 3.6.3) that models using more stringent GAM criteria provide the most representative effect
11 estimate sizes, the PM risk assessment included C-R functions using only GAM with the more
12 stringent convergence criterion (denoted "GAM (stringent)") for all urban locations, except Los
13 Angeles.14 It should be noted that the GAM stringent C-R functions do not address the issue of
14 understated standard errors of the coefficient estimates. Thus, the confidence intervals included
15 in the risk assessment involving use of the GAM (stringent) C-R functions are somewhat
16 understated. As indicated in the CD, "the extent of downward bias in standard error reported in
17 these data (a few percent to -15%) also appears not to be very substantial, especially when
18 compared to the range of standard errors across studies due to differences in population size and
19 number of days available" (CD, p. 9-35).
20 More detailed information about the C-R relationships used in the PM risk assessment is
21 provided in Appendix 4A of this draft Staff Paper. This information includes population
22 characteristics (e.g., age and disease status), form of the model (e.g., log-linear, logistic),
23 whether other pollutants were included in the model, lags used, observed minimum and
24 maximum ambient PM concentrations, and PM coefficients along with lower and upper 5th and
25 95th confidence intervals.
26
5 risk estimates for various combinations of statistical estimation approaches (GAM and GLM with
varying degrees of freedom) have been included for Los Angeles as a sensitivity analysis to illustrate the impact of
•
alternative model specification choices.
January 2005 4-33 Draft - Do Not Quote or Cite
-------
1 4.2.6.1 Hypothetical Thresholds
2 In assessing or interpreting public health risk associated with exposure to PM, the form
3 of the C-R function is a critical component. The health effects evidence examining whether or
4 not a population threshold might exist for short- and long-term exposure health outcomes for
5 PM25 and short-term exposure health outcomes for PM10_25 is discussed in section 3.6.6 of this
6 SP and section 8.4.7 of the CD.
7 The PM2 5 base case risk assessments presented in sections 4.3 and 4.4 below do not
8 include a threshold, based on the conclusions in the CD that "there is no strong evidence of a
9 clear threshold for PM mortality effects" and that the use of linear PM effect models appears to
10 be appropriate (CD, p.8-345) . The base case risk estimates reflect the potential contribution of
11 PM2 5 down to either an estimated background level or the LML in the study, whichever is
12 higher. For a number of studies, including all of the long-term exposure mortality studies, the
13 LML is significantly above the estimated background concentrations and, therefore, there is no
14 contribution to the risk estimates from PM25 concentrations below the LML in these cases.
15 As discussed in section 3.6.6, while the CD concludes that there is no strong evidence of
16 a clear threshold for PM mortality effects, it also notes "nor is there clear evidence against iff
17 possible thresholds for PM-related effects" (p. 8-322). The CD also states that "some single-city
18 studies do provide some suggestive hints forpossible thresholds, but not in a statistically clear
19 manner" (p. 8-322). Therefore, as noted earlier, sensitivity analyses have been conducted that do
20 include hypothetical alternative thresholds, where risks only are estimated due to PM2 5 or
21 PM10.25 concentrations exceeding the assumed threshold concentrations. Based on the staff
22 evaluation contained in section 3.6.6, three hypothetical thresholds (10,15, and 20 jig/m3) were
23 included in sensitivity analyses for short-term exposure mortality for PM25 and short-term
24 exposure morbidity for PM]0.25 and two hypothetical thresholds (10 and 12 [ig/m3) were included
25 in sensitivity analyses for long-term exposure mortality associated with PM25 Results of these
26 sensitivity analyses are discussed below in section 4.3.
27 4.2.6.2 Single and Multi-Pollutant Models
28 For several of the epidemiological studies from which C-R relationships for the PM risk
29 assessment were obtained, C-R functions are reported both for the case where only PM levels
30 were entered into the health effects model (i.e., single-pollutant models) and where PM and one
January 2005 4-34 Draft - Do Not Quote or Cite
-------
1 or more other measured gaseous co-pollutants (i.e., ozone, nitrogen dioxide, sulfur dioxide,
2 carbon monoxide) were entered into the health effects model (i.e., multi-pollutant models). To
3 the extent mat any of the co-pollutants present in the ambient air may have contributed to the
4 health effects attributed to PM in single-pollutant models, risks attributed to PM might be
5 overestimated where C-R functions are based on single-pollutant models. However, as discussed
6 in section 3.6.4 the statistical significance of the associations reported between PM2S (and PM10)
7 and mortality due to short-term exposure show no trends with the levels of any of four gaseous
8 co-pollutants examined. While not definitive, these consistent patterns indicate that it is more
9 likely that there is an independent effect of PM2 s that is not appreciably modified by the gaseous
10 co-pollutants.
11 For some of the gaseous co-pollutants, such as CO, NO2, and SO2, which tend to be
12 highly correlated with ambient PM2.5 concentrations in some cities (and, in the case ofNOx and
13 SOX, are PM precursors as well), it is difficult to sort out whether these pollutants are exerting
14 any independent effect from that attributed to PM2 5. As discussed in section 3.6.4, inclusion of
15 pollutants that are highly correlated with one another can lead to misleading conclusions in
16 identifying a specific causal pollutant. When such collinearity exists, multi-pollutant models
17 would be expected to produce unstable and statistically insignificant effects estimates for both
18 PM and the co-pollutants (CD, p.8-241). Given that single and multi-pollutant models each' have
19 both potential advantages and disadvantages, with neither type clearly preferable over the other
20 in all cases, risk estimates based on both single and multi-pollutant models have been developed.
21 4.2.6.3 Single, Multiple, and Distributed Lag Functions
22 The question of lags and the problems of correctly specifying the lag structure in a model
23 are discussed extensively in the CD (section 8.4.4) and in section 3.6.5 of this SP. As noted in
24 those discussions, it is important to consider the pattern of-results that is seen across the series of
25 lag periods. When there is an observed pattern showing effects across different lags, use of the
26 single-day lag with the largest effect, while reasonable, is likely to underestimate the overall
27 effect size (since the largest single-lag day results do not fully capture the risk also distributed
28 over adjacent or other days)(CD, p.8-270).
29 As discussed in the CD, a number of the PM2 5 short-term exposure mortality studies
30 reported stronger associations with shorter lags, with a pattern of results showing larger
January 2005 4-35 Draft - Do Not Quote'o'r Cite
-------
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
associations at the 0- and 1-day lag period that taper off with successive lag days for the varying
PM indicators. Several studies included in the PM25 risk assessment only included 0- and 1-day
lags in presenting results. Therefore, when a study reports several single day lag models, unless
the study authors identify a "best lag", both the 0- and 1-day lag models for mortality (both total
and cause-specific) were chosen for inclusion in the PM2 5 risk assessment. In one study
conducted in Los Angeles (Moolgavkar, 2003), there was no consistent pattern observed across
the various lags examined for COPD mortality. Therefore, EPA did not include this particular
endpoint in the PM2 5 risk assessment for Los Angeles.
For hospital admissions, unless the study authors specified an optimal lag, both 0- and 1 -
day lag models were included for cardiovascular admissions since the CD indicates that recent
evidence from time series studies strongly suggests maximal effects at 0-day lag with some
carryover to 1-day lag and little evidence for effects beyond 1-day for this health endpoint (CD,
p.8-279). Since many of the studies addressing COPD hospital admissions report effects at
somewhat longer lags, 0-, 1-, and 2-day lag models (if all three were available) were included in
the risk assessment for this health endpoint, unless the authors-selected a different "best lag."
As discussed in section 3.6.5.2, there is recent evidence (Schwartz, 2000b, reanalyzed in
Schwartz, 2003 b), that the relation between PM and health effects may best be described by a
distributed lag (i.e., the incidence of the health effect on day n is influenced by PM
concentrations on day n, day n-1, day n-2 and so on). As noted above, if this is the case, a model
that includes only a single lag (e.g., a 0-day lag or a 1-day lag) is likely to understate the total
impact of PM. Because of this, a distributed lag model may be preferable to a single lag model.
However, distributed lag models have been used in only a few cases and only for PM10. When a
study reports several single lag models, unless the study authors identify a "best lag," the
following lag models were included in the risk assessment based on the assessment in CD and in
section 3.6.5.2:
• both 0- and 1 -day lag models for mortality (both total and cause specific),
• both 0- and 1-day lag models for cardiovascular and respiratory hospital admissions, and
• 0-, l-5 and 2-day lag models (if all three were available) for COPD hospital admissions.
January 2005 4-36 Draft - Do Not Quote or Cite
-------
1 A sensitivity analysis was also conducted to examine the potential impact of using a distributed
2 lag approach for short-term exposure mortality associated with PM2 5 based on the distributed lag
3 analysis of PM10 and mortality (Schwartz, 2000b, reanalyzed in Schwartz, 2003b). This
4 sensitivity analysis was included to provide a very rough sense of the possible underestimation
5 of risk due to use of single-day lags models.
6 4.2.6.4 Long-term Exposure Mortality PM2 s Concentration-Response Functions
7 The available long-term exposure mortality C-R functions are all based on cohort studies,
8 in'which a cohort of individuals is followed over time. As discussed in section 3.3.1.2, based on
9 the evaluation contained in the CD and the staffs assessment of the complete data base
10 addressing mortality associated with long-term exposure to PM2 s, staff have concluded that two
11 cohorts that have been studied are particularly relevant for the PM2 s risk assessment These
12 include the Six Cities study cohort (referred to here as Krewski et al (2000) - Six Cities) and the
13 American Cancer Society (ACS) cohort (referred to as Krewski et al. (2000) - ACS) containing
14 a larger sample of individuals from many more cities. In addition, Pope et al. (2002) extended
15 the follow-up period for the ACS cohort to sixteen years and published findings on the relation
16 of long-term exposure to PM25 and all-cause mortality as well as cardiopulmonary and lung
17 cancer mortality (referred to here as Pope et al. (2002) - ACS extended). EPA's use of these
18 particular cohort studies to estimate health risks associated with long-term exposure to PMZ5 is
19 consistent with the views expressed in the NAS (2002) report, "Estimating the Public Health
20 Benefits of Proposed Air Pollution Regulations," and the SAB Clean Air Act Compliance
21 Council review of the proposed methodology to estimate the health benefits associated with the
22 Clean Air Act (SAB, 2004). Risk estimates have been developed using C-R functions from the
23 Six Cities, ACS, and ACS-extended studies. As explained in section 3.6.5.4, three different
24 indicators of long-term PM2.5 exposure were considered in this extended ACS study and staff
25 have selected the C-R function associated with an average of the 1979-1983 and 1999-2000
26 PM2 s ambient concentrations to use in the current risk assessment
27
28 4.2.7 Characterizing Uncertainty and Variability
29 This section discusses the approaches used in the current PM risk assessment to address,
30 and characterize, where feasible, uncertainties and variability. Although the weight of the
January 2005 4-37 Draft - Do Not Quote or Cite
-------
1 evidence is sufficient to support the conclusions in the CD that a variety of health endpoints are
2 likely causally related to short- and long-terra ambient exposures to PM2 5 and short-term
3, ambient exposures to PM]0_2 5, significant uncertainties remain affecting the quantitative
4 assessment of health risks associated with varying exposure levels. The following briefly
5 summarizes the major sources of these uncertainties and variability and how they are dealt with
6 in the risk assessment:
7
8 • • Causality. There is uncertainty about whether each of the estimated associations between
9 the two PM indicators (PM2 5 and PM]0_2 5) and the various health endpoints included in
10 this risk assessment actually reflect a causal relationship. There are varying degrees of
11 uncertainty associated with the various PM indicators and health endpoints related to
12 differences in the weight of evidence supporting judgments about whether an observed
13 association truly reflects a causal relationship. For example, there is much greater
14 uncertainty associated with the morbidity effects associated with PM]0.25 exposures
15 compared to PM2 5 due to the much smaller health effects data base. Chapter 3 presents a
16 more detailed discussion of the staffs qualitative assessment of the varying weight of
17 evidence associated with the effects included in the risk assessment.
18
19 • Empirically estimated C-R relationships. In estimating the C-R relationships, there are
20 uncertainties: (1) surrounding estimates of PM coefficients in C-R functions used in the
21 assessment, (2) concerning the specification of the C-R model (including the shape of the
22 C-R relationship) and whether or not a population threshold exists within the range of
23 concentrations examined in the studies, and (3) related to the extent to which PM C-R
24 functions derived from studies in a given location and time when PM concentrations were
25 higher provide accurate repres entations of the C-R relationships for the same location
26 with lower annual and daily PM concentrations. For the few instances where multi-city
27 PM C-R functions are included in the risk assessment (e.g., use of the Six-Cities study
28 function for respiratory symptoms associated with short-term exposures to PM2 s applied
29 in Boston and St. Louis), there also is uncertainty related to the transferability of PM C-R
30 functions from multiple locations to the specific location selected for the risk
31 assessment.15 Statistical uncertainty, based on the standard errors reported in the
32 epidemiology studies is incorporated in the risk assessment and is discussed below.
33 Sensitivity analyses of potential alternative hypothetical thresholds also have been
34 included in the risk assessment.
1 A C-R function derived from a multi-cities study may not provide an accurate representation of the C-R
relationship in a specific assessment location because of (1) variations in PM composition across cities, (2) the
possible role of associated co-pollutants in influencing PM risk, (3) variations in the relation of total ambient
exposure (both outdoor exposure and ambient contributions to indoor exposure) to ambient monitoring in different
locations (e.g, due to differences in air conditioning use in different regions of the U.S.), (4) differences in
population characteristics (e.g., the proportions of members of sensitive subpopulations) and population behavior
patterns across locations.
January 2005 4-38 Draft - Do Not Quote or Cite
t
-------
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
44
45
• Lag structure. There is some evidence from a few PM,0 studies that the impact of any
single day of exposure may be to cause effects across a number of subsequent days (i.e., a
distributed lag), however most epidemiology studies have only analyzed single day lags.
The use of single day lag C-R functions could result in a downward bias in the estimated
incidence associated with a given reduction in PM concentrations. However, there are no
available PM2 5 or PMi0.2 5 studies that included distributed lag models. As discussed
below, a limited sensitivity analysis has been conducted to illustrate the potential impact
• on PM25 mortality risk estimates associated with short-term exposures. .
* 'Extrapolation of C-R relationship beyond the range of observed PM data. There is
significant uncertainty about the shape of the C-R relationship beyond the range of the
PM data observed in the epidemiology studies. Risk estimates have not been calculated
for PM levels below the lowest measured level (LML) in a study, if it was available.
Where the LML was not available, risk was estimated only down to an estimated
background level. This approach minimizes the uncertainty for risk estimates associated
with concentrations within the range of the studies.
• Adequacy of ambient PM monitors as surrogate for population exposure. The extent to
which there are differences in the relationship between spatial variation in ambient PM2 5
or PM10.2 5 concentrations and ambient exposures in the original epidemiology studies •
compared to more recent ambient PM25 or PM10.25 data introduces additional uncertainly
in the risk estimates. This is expected to be more of a concern for PM10_2 5 where greater
spatial variability in ambient monitoring data within urban areas and over time has been
observed.
• Adjustment of air quality distributions to simulate just meeting alternative standards.
. The shape of the daily distribution of PM25 and PM10.25 ambient concentrations that
would result upon meeting alternative PM standards is unknown. Based on an analysis
of historical data, staff believes it is a reasonable assumption that PM25 concentrations
would be reduced by roughly the same percentage. However, there is much greater
uncertainty associated with the use of this same approach for meeting PM10.2S standards
given the lack of sufficient data to evaluate the reasonableness of this assumption.
• Background concentrations. Since risks have only been estimated in excess of
background, where the LML is either not available or is lower than the estimated
background, uncertainty about background concentrations contributes to uncertainty
about the risk estimates. As discussed below, sensitivity analyses examining the impact
of alternative constant and varying daily background levels on the risk estimates have
been conducted.
• Baseline incidence rates and population data. There are uncertainties related to; (1) the
extent to which baseline incidence rates, age distribution, and other demographic
variables that impact the risk estimates vary for the year(s) when the actual epidemiology
studies were conducted, the recent year of air quality used in the assessment, and some
unspecified future year when air quality is adjusted to simulate just meeting the current or
January 2005 4-39 Draft - Do Not Quote or Cite
-------
1 alternative standards; (2) the use of annual incidence rate data to develop daily health
2 effects incidence data; and (3) related to the use of an overall combined incidence rate for
3 six cities for the respiratory symptoms endpoint which is applied to individual cities (i.e.,
4 Boston and St. Louis). Variability in baseline incidence and population data is taken into
5 account by use of city-specific data in most cases.
6
7 The uncertainties from some of these sources — in particular, the statistical uncertainty
8 surrounding estimates of the PM coefficients in C-R functions — are characterized quantitatively
9 in the PM risk assessment. It is possible, for example, to calculate confidence intervals around
10 risk estimates based on the uncertainty associated with the estimates of PM coefficients used in
11 the risk assessment. These confidence intervals express the range within which the risks are
12 likely to fall if the sampling error uncertainty surrounding PM coefficient estimates were the
13 only uncertainly in the assessment.16 In situations where the point estimate for a C-R function is
14 positive, but the lower confidence limit estimate is less than 1.0, the lower confidence limit of
15 the risk estimate is a negative value. Based on the overall body of evidence on the relationships
16 between PM and health effects, the staff believes that these negative estimates should not be
17 interpreted as implying that increasing PM levels will result in reduced risks, but rather that the
18 negative risk estimates are simply a result of statistical uncertainty in the reported C-R
19 relationships in the epidemiological studies.
20 Steps also have been taken to minimize some of the uncertainties noted above. For
21 example, the current PM risk assessment includes only health endpoints for which the CD
22 evaluation and staff assessment (see Chapter 3) find that the overall weight of the evidence
23 supports the conclusion that PM25 is likely causally related, or for PM10.25 is suggestive of a
24 causal relationship. Also, for most of the health endpoints and locations included in the risk
25 assessment, this assessment uses the C-R functions derived from epidemiological studies carried
26 out in those same locations. This serves to minimize the uncertainties, such as differences in
27 composition and differences in factors affecting human exposure associated with applying C-R
28 functions developed in one location to a different location.
29 As noted above, a variety of sensitivity analyses, summarized in Table 4-8, have been
30 included in the risk assessment to address some of the major uncertainties. The results of these
Tiowever, as discussed earlier in section 4.2.6, for the short-termC-R functions based on reanalyzed GAM
(stringent) models the confidence intervals are somewhat understated.
January 2005 4-40 Draft - Do Not Quote or Cite
-------
1 sensitivity analyses are summarized in sections 4.3.2 (for as is risk estimates), 4.4.2 (for just
2 meeting the current PM25standards), and 4.5.3 (for meeting alternative PM2S and PM10.2.5
3 standards).
4
5 4.3 PM2 s and PMlo.2.5 RISK ESTIMATES FOR CURRENT ("AS IS") AIR QUALITY
6 43.1 Base Case Risk Estimates
7 . The base case risk estimates associated with "as is" PM2 5 and PM10.2 5 concentrations in
8 excess of background levels are presented in a series of figures in this section. The risk
9 estimates are expressed in terms of percent of total incidence for each health endpoint in these
10 figures. For each series of estimates, a point estimate is provided along with 95% confidence
11 intervals. As noted above, in some cases,' where the lower confidence limit of the C-R function
12 is less than 1.0, the resulting lower confidence limit of the risk estimate is a negative value. The
13 staffs interpretation of these negative values is that while they.indicate statistical uncertainty
.14 about the C-R relationships, they do not at all suggest that risk reductions would be associated
15 with an increase in PM levels. Additional detailed tables which present the estimated incidence
16 (both as the number of effects and as a percentage of total incidence) for each risk assessment
17 location are included in the TSD (Abt Associates, 2005). Risk estimates in a given assessment
18 location are presented only for those health endpoints for which there is at least one acceptable
19 C-R function reported for that location. Therefore, the set of health effects shown in the figures .
20 varies for the different locations.
21 Figures 4-2 through 4-6 present the PM25 risk estimates across the various assessment
22 locations associated with "as is" concentrations in excess of either background or the LML in the
23 study providing the C-R function, whichever is greater. Figure 4-2 compares risk estimates for
24 total non-accidental mortality incidence associated with short-term (i.e.j 24-hour) exposure to
25 PM2 5 using single-pollutant, single-city models. The point estimates are in the range from about
26 0.5 to 2.5% of total non-accidental mortality incidence. Figure 4-3 compares the estimated
27 percent of total incidence for non-accidental and cause-specific mortality associated with short-
28 term exposure to PM2 5 based on single city versus multi-city models. Generally, the estimated
29 incidence for the single- and multi-city models are roughly comparable, with somewhat lower
30 risk estimates seen in Boston for the multi-city models compared to the single-city models and
January 2005 4-41 . Draft - Do Not Quote or Cite
-------
Table 4-8. Sensitivity Analyses
Analysis
Number
(Figure 4-1)
1
2
3
4
5
6
7
PM
Indicator
PM2>
PM10.2.5
PM2.5
PM2.5
PM2.5
PM2.5
PM2.5
PM2.5,
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
different (constant) background PM levels
A sensitivity analysis of ihe effect of assuming a
constant background PM level versus a distribution of
daily background 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 sensitivity analysis of the effect of just meeting the
current and alternative annual PM2 5 standards 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
Sensitivity analyses assuming alternative hypothetical
threshold concentration levels for the occurrence of
PM2 j- and PMi0.23-related response at concentrations
above those for background or the LML for as is air
quality, and for just meeting the current and alternative
PM, , standards.
2
3
4
5
6
8
9
10
11
12
13
14
Source: Abt Associates (2005).
January 2005
4-42
Draft - Do Not Quote or Cite
-------
€
: ; : i J,
(6e Asp ffi
*1:
-1
Koeooz
ZMBMipS;
00
oo
-------
iiftjjejJoweseiaBJp;;
;:Xiije|ibwaSBSSjp;:;;
i;.sas»:fl;;;T*T.;
g
OS <-'
1 i
~'i
I
13
-Q
O
JS X
w a
*£ «
F S
||
If
<» SJD
1 I
K«
VM ^
O MM
+- 3
e »
s 5
P
"*• u
"S c
•»- o>
II
w u
P?^!»?3:i;;iiiiilii;iii| «
^
u
o
I
oo ON
vo
oo
o ^
c>» cs
-------
!JK*M»
:
I
CO O
ft* £
V) <
O
i
a
II
01 .»
o e
E ft.
£2
'i"
•^ *•
1
o
Sfl
<
1
A
i
£
I
"B
ft.
i
e
iZ
e
e
•e
u
3
M
»« J
3
o «
*- 9
tf va
®
8
o
-------
t-» oo
-------
G
fc
•a*
o
i
WJ
•9
IS
II
A
£
•3
e
J
si
on T3
H 9
e ss
S(£
*S ^
I
e e
« -r
H
c £
li
I
I
S5
U K)
I
.a
j
J"
•8
3,
i> oo
. o -^
VO t~ OO O\ O i—i
,—c -1 ^M -4 (N| (S
-------
1 the reverse being observed in St. Louis. As expected, the 95% confidence intervals are
2 somewhat smaller for the multi-city models compared to the respective single-city models which
3 is due to the greater sample size in the multi-city models.
4 Figure 4-4 compares risk estimates based on single-pollutant versus multi-pollutant C-R
5 models provided in the epidemiological studies for PM2 5 short-term exposure health endpoints.
6 In two cases there is relatively little difference in the risk estimates between the single-pollutant
7 and multi-pollutant models (i.e., Pittsburgh and San Jose), while in the third case (Los Angeles)
8 there are larger differences when either CO or N02 are added to the model along with PM.
9 Figures 4-5 and 4-6 show risk estimates for mortality related to long-term (i.e., annual average)
10 exposure to PM2 5 based on single- and multi-pollutant models, respectively. The point estimates
11 for the single-pollutant models, based on the ACS-extended study (Pope et al., 2002), range from
12 0.5% in Seattle to as high as 6.6% of total mortality in Los Angeles, with most point estimates
13 falling in the 2 to 5% range. The point estimates based on the original ACS study (Krewski et '
14 al., 2000) are lower in Phoenix, Seattle, and San Jose (ranging from 0 to 0.5%) because the "as
15 is" annual averages at the composite monitors in these locations were not much higher than the
16 LML in the ACS study (i.e., 10 (ig/m3) and risk estimates only were calculated down to the
17 LML. As shown in Figure 4-6, the risk estimates based on multi-pollutant models, involving
18 addition of different single co-pollutants in the ACS study, show generally greater risk
19 associated with PM2 5 when CO, NO2, or 03 were added to the models and lower risk associated
20 with PM2 s when SO2 was added.17
21 Figure 4-7 shows risk estimates for hospital admissions and respiratory symptoms
22 associated with short-term exposure to PM10.25 for three urban areas (Detroit, Seattle, and St
23 Louis). For Detroit risk estimates are provided for several categories of cardiovascular and
24 respiratory-related hospital admissions and show point estimates ranging from about 2 to 7% of
25 cause-specific admissions being associated with as is short-term exposures to PM]0.25. The point
26 estimate for asthma hospital admissions associated with PM]p_2 5 exposures for Seattle, an area
17 The addition of a second pollutant reduced the number of cities available for estimating the C-R function
from 50 for PM25 alone to 44 with addition of CO, to 3 3 with addition of NO2, to 45 with addition of Oj 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.
January 2005 4-48 Draft - Do Not Quote or Cite
-------
0
mMmmMoti. pis
igjuownou
ijtilciQ&cjili
I
1
O)
I
a
I
3
V3
e
o.
*
W
Q
•0
V)
I
w
1
OS
in
c .*• •
-------
1 with lower PM10.25 ambient concentrations, is about 1%. Point estimates for lower respiratory
2 symptoms and cough in St. Louis are about 12 and 15%, respectively.
3
4 4.3.2 Sensitivity Analyses
5 Several sensitivity analyses were carried out to provide some perspective on the impact
6 of various assumptions and uncertainties on the health risk estimates (see Table 4-8 above for a
7 summary of different types of sensitivity analyses). Most of these sensitivity analyses were
8 conducted in each of the study areas and the complete results are in the TSD (Abt Associates,
9 2005). The PM25 risk results for one study area (Detroit), are shown here for some of the
10 sensitivity analyses for illustrative purposes. Detroit has been selected because it provides an
11 opportunity to examine bom mortality and morbidity risk estimates and includes both single and
12 multi-pollutant C-R functions. In some cases, sensitivity analyses were conducted only in one
13 location due to data constraints (e.g., only Los Angeles for alternative C-R model specifications
14 since it was the only study that presented results for a wide range of alternative model
15 specifications). '
16 4.3.2.1 Alternative Background Levels
17 For purposes of informing decisions about the PM N AAQS, we are interested in PM-
18 related risks due to concentrations over background levels, where background includes PM from
19 natural sources and transport of PM from sources outside of the U.S., Canada, and Mexico
20 (discussed in section 2.6). One set of sensitivity analyses examined the impact of using the
21 lower and upper end of the range of estimated background concentrations provided in section
22 2.6. For Detroit, the use of alternative estimated PM2 5 background levels had only a relatively
23 small impact on the short-term exposure mortality or hospital admission risk estimates because
24 the LML for PM2 5 in Ito (2003) [reanalysis of Lippmann et al. (2000)] was 4 iig/m3, which is
25 lower than the upper range of background levels considered in the sensitivity analysis (i.e., 2 to 5
26 ng/m3). There was no difference in the base case where background was assumed to be 3.5
27 jig/m3 versus setting background at the lower end of the range (2.0 iig/m3). With the background
28 set at 5 jig/m3, the short-term exposure risk estimates were about 10% smaller than the base case.
January 2005 4-50 Draft, do not cite or quote
-------
1 In tiie other eight PM2 5 locations, using the upper and lower end of the range of estimated
2 background generally had a small to modest impact, on the order of roughly +/-10-20% change
3 in short-term exposure health endpoint risk estimates compared to use of the midpoint of the
4 estimated range of background levels in the base case estimates. Alternative estimated PM2 5
5 background levels had no impact on long-term exposure mortality in Detroit, or any of the other
6 PM2 5 locations, because the LMLs in the long-term studies were 7.5,10 or 11 ug/m3, which all
7 are larger than tiie range of estimated PM2 5 background levels.
8 A sensitivity analysis also was conducted that focused on the impact of using a varying
9 estimated PM2 5 background concentration instead of the fixed level used in each study area in
10 the base case assessment. Staff developed a Monte Carlo simulation approach to generate a year
11 long series of daily PM25 background concentrations for specific urban areas based on using
12 available distributional information for the observed and background concentrations to estimate
13 ' their joint distribution, which yields the distribution of the background concentrations
14 conditioned on the level of the observed concentrations (see Langstaff, 2004 for additional
15 details describing the approach). This approach involved assigning a background value to an
16 observed concentration by randomly selecting a value from the conditional distribution
17 corresponding to the observed value. The analysis was done both without any correlation
18 assumed and with a 0.4 correlation between background and observed concentrations. To
19 * implement this approach, the mean of the background distribution was assumed to^be the mid-
20 point estimate of PM2 5 background discussed in section 2.6. Estimates of the variation in
21 background concentrations for different regions of the United States were obtained by an
22 analysis of daily data from IMPROVE sites with the sulfate component removed (Langstaff,
23 2005). It is important to recognize that all IMPROVE sites measure some PM2 5 from
24 anthropogenic sources, and that removing sulfate from the PM2 5 component considered does not
25 completely remove all anthropogenic contributions to the observed concentrations.
26 The sensitivity analysis examining varying daily background was carried out in Detroit
27 and St. Louis using as is air quality levels for short-term exposure non-accidental mortality
28 associated with PM2 5. As shown in exhibit 7.8 (Abt Associates, 2005), the difference between
29 the risk estimates based on a constant versus a varying daily background were extremely small in
January 2005 4-51 Draft - Do Not Quote or Cite
-------
1 Detroit (i.e., 0.8 percent of total incidence with varying daily background vs. 0.9 percent with
2 assumed constant background). The difference was even smaller in St. Louis in both the no
3 correlation and 0.4 correlation cases, with essentially no difference in risk estimates between the
4 constant and varying daily background cases (Abt Associates, 2005).
5 It should be noted that the estimated distributions for background may not fully reflect
6 peak 24-h average natural background concentrations which can be substantially higher than the
7 annual or seasonal average background concentrations within areas affected by wildfires and
8 dust storms and long range transport from outside the United States, Canada, and Mexico (see
9 section 2.6). While the current PM25 base case risk estimates, therefore, do not capture these
10 unusual events, it should be noted that there are regulatory provisions to exclude such events for
11 purposes of judging whether an area is meeting the current NAAQS (as noted above in section
12 2.6). The PM25 risk assessment also included a sensitivity analysis which used 2002 air quality
13 data for Boston to examine the impact of an extreme example (i.e., the Quebec fire episode in
14 July 2002) of this type of natural episodic event on short- and long-term exposure mortality (see
15 Exhibits 7.9 and 7.10 in Abt Associates, 2005). This sensitivity analysis showed that there was
16 hardly any difference (i.e., differences ranged from 0 to 0.1 % of total incidence) in estimated
17 short-term exposure mortality associated with PM2 5 when one included or excluded this fairly
18 extreme, but
19 short-term episode.18 This same sensitivity analysis showed a difference of about 0.2% in total
20 long-term exposure mortality incidence associated with PM25 with and without inclusion of the
21 Quebec fire episode days.
22 For PM10_2 5, the sensitivity analysis examining the effects of using the lower and upper
23 end of the range of estimated background levels showed about a 16% increase in the risk
24 estimates for various respiratory and cardiovascular-related short-term exposure hospital
25 admissions in Detroit between the base case (which used a value of 4.5 u.g/m3 for background)
\
26 and the lower end where background was estimated to be 1 ng/m3. At the upper end, where
27 background was estimated to be 9 jig/m3, the short-term exposure hospital admission risk
. This extreme episode included 2 days with PMj 5 levels above 30 ^g /m3 and 1 day above
January 2005 4-52 Draft - Do Not Quote or Cite "•<
-------
1 estimates were reduced by about 19% (see Exhibit 7.12 in the TSD (Abt Associates (2005)).
2 The effect of different background concentrations for the other two PM10.2 5 locations can be
3 found in Exhibits D. 84 and D. 86 through D. 89 in me TSD.
4 4.3.2.2 Hypothetical Thresholds
5 • One of the most significant uncertainties continues to be the issue of hypothetical
6 thresholds below which there may be no PM2 5 or PM10.2 5 health effects. As discussed above in
7 sections 3.6.6 and 4.2.6.1, there is very limited evidence addressing whether or not thresholds
8 exist for PM2 5, with most analyses failing to find evidence that population thresholds exist
9 within the range of concentrations examined. As a sensitivity analysis, three hypothetical
10 thresholds^ or cutpoints (10,15, and 20 |ig/m3) are used to examine the potential impact on risk
11 estimates for short-term exposure mortality and two different hypothetical thresholds or •
12 cutpoints (10 and 12 |ig/m3) are used to examine the potential impact on risk estimates for long-
13 term exposure mortality. In conjunction with defining such cutpoints for these sensitivity
14 analyses, the slopes of the C-R functions have been increased to reflect the effect of hypothetical
15 thresholds at the selected levels. A simple slope adjustment method has been used that assumes
16 the slope for the upward-sloping portion of a hockey stick would be approximately a weighted
17 average of the two slopes of a hockey stick - namely, zero and the slope of the upward-sloping
18 portion of the hockey stick (see the TSD (Abt Associates, 2005) for additional details). If the
19 data used in a study do not extend down below the cutpoint or extend only slightly below the
20 cutpoint, then the extent of the downward bias of the reported PM coefficient will be minimal or
21 non-existent. This is the case, for example, when the cutpoint is 10 jig/m3 or 12 jig/m3 for long-
22 term exposure mortality, given that the LMLs in the long-term exposure mortality studies were
23 7.5,10, and 11 p-g/m3. Staff believes that the slope adjustment method used in this risk
24 assessment is a reasonable approach to illustrate the potential impact of using a non-linear
25 approach. A more definitive evaluation of the effect of hypothetical thresholds and use of
26 alternative non-linear approaches would require re-analysis of the original health and air quality
27 data, which is beyond the scope of this risk assessment. • "
28 The results of these sensitivity analyses examining the impact of hypothetical thresholds
29 for short-term exposure mortality risk estimates for the "as is" PM2 5 levels in Detroit show mat
January 2005 4-53 Draft - Do Not Quote or Cite
-------
1 the short- and long-term exposure risk estimates are particularly sensitive to the application of
2 hypothetical thresholds. A hypothetical threshold of 10 [ig/m3 reduces the percent of total non-
3 accidental mortality incidence associated with short-term exposure to PM25 from 0.9% to 0.5%,
4 a 44% decrease in the risk estimate and the highest hypothetical threshold of 20 jig/m3 reduces it
5 to 0.2%, nearly an 80% reduction from the base case. To illustrate the impact on long-term
6 exposure mortality, using the risk estimates based on the ACS-extended all cause mortality
7 results, a hypothetical threshold of 10 ng/m3 reduces the risk estimate from 4.7% of total
8 incidence to 3.7%, a reduction of about 20% from the base case estimate and a hypothetical
9 threshold of 12 \ig/m3 reduces the risk estimate to 2.7%, a reduction of over 40% from the base
10 case estimate.
11 . 4.3.2.3 Alternative Concentration-Response Models
12 Another sensitivity analysis illustrates how different the risk estimates would be if the C-
13 R functions used for short-term exposure mortality had used distributed lag models instead of
14 single lag models. Schwartz (2000a) has shown in a study of short-term exposure mortality in 10
15 cities using PM10 as the indicator that a distributed lag model predicted the same relative risk that
16 a single lag model would have predicted if the coefficient was approximately two times what it
17 was estimated to be. To simulate the possible impact of using a distributed lag model, the PM25
18 coefficients were multiplied by two. As would be expected, the risk estimates are almost
19 doubled using the distributed lag approximation (see Abt Associates, 2005; Appendix D).
20 The influence of using different periods of exposure on the risks estimated in long-term
21 . exposure mortality studies also has been examined in a sensitivity analysis. Two alternatives
22 were examined: assuming the relevant PM25 ambient concentrations were respectively 50%
23 higher than and twice as high as the PM25 ambient concentrations used in the original
24 epidemiological study. The impact of these varying assumptions about the role of historical air
25 quality on estimates of long-term exposure mortality associated with "as is" PM25 concentrations
26 is shown for Detroit in Table 4-9. Assuming that PM2 5 concentrations were 50% higher than
27 and twice as high as that in the original studies reduces long-term exposure mortality risk
28 estimates by about one-third and one-half, respectively.
29
January 2005 4-54 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
til
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Various respiratory and cardiovascular cause-specific hospital admission point estimates
associated with short-term exposure to PM10.2.5 range from 1 to 7%, depending on
location and type of admission. Point estimates for lower respiratory symptoms and
cough were about 12 and 15% of total incidence for as is levels in a single urban area (St.
Louis)
The wide variability in risk estimates associated with a recent year of air quality for the
two different PM indicators is to be expected given the wide range of PM levels across the urban
areas analyzed and the variation observed in the C-R relationships obtained from the original
epidemiology studies. Among other factors, this variability may reflect differences in
populations, exposure considerations (e.g., degree of air conditioning use), differences in co-
pollutants and/or other stressors, differences in study design, and differences related to exposure
and monitor measurement error.
Based on the results from the sensitivity analyses, the following key observations are
made: ' •..-.-.
• The single most important factor influencing the risk estimates is whether or not a
hypothetical threshold exceeding the estimated background level or LML in the studies
exists. . " " '
• The following uncertainties have a moderate impact on the risk estimates in some or all
of the cities: choice of an alternative estimated constant background level, use of a
distributed lag model, and alternative assumptions about the relevant air quality for long-
term exposure mortality. Use of a distribution of daily background concentrations had,
very little impact on the risk estimates.
During the previous review of the PM NAAQS, EPA provided an illustrative example
based on the PM health risk assessment that showed the distribution of mortality risk associated
with short-term exposure over a 1-year period. EPA concluded that peak 24-hour PM25:
concentrations appeared "to contribute a relatively small amount to the total health risk posed by
an entire air quality distribution as compared to the risks associated with low to mid-range
concentrations" (61 FR at 65652, December 13,1996). Figures 4-8 (a,b) provide an example of
the annual distribution of 24-hour PM2S concentrations in Detroit and the corresponding
January2005
4-57
Draft - Do Not Quote or Cite
-------
§
•p| *)
w •«
s *S .-
•o c o-
rt w
0.
S
R 8
tw C
P
f >»
in
•s
"§
1
!
s
o
t r
i ••
O «
00
00
••a
1
4SS
.
••a
s
od
.O -H
•06 '
*
"
cn
en
£
» '
1 1 ^
||f
SB en &
o
cs o
cS
JN
«
5
Al
13 8
o
3 ^
a S
O CJ
8 <
I"3
|*
I
•^ u
•c|
red
1
.
I
S *
-------
1 • Various respiratory and cardiovascular cause-specific hospital admission point estimates
2 associated with short-term exposure to PM,0.2 5 range from 1 to 7%, depending on
3 location and type of admission. Point estimates for lower respiratory symptoms and
4 cough were about 12 and 15% of total incidence for as is levels in a single urban area (St.
5 Louis)
6
7 ^ The wide variability in risk estimates associated with a recent year of air quality for the
8 two different PM indicators is to be expected given the wide range of PM levels across the urban
9 areas analyzed and the variation observed in the C-R relationships obtained from the original
10 epidemiology studies. Among other factors, this variability may reflect differences in
11 populations, exposure considerations (e.g., degree of air conditioning use), differences in co-
12 pollutants and/or other stressors, differences in study design, and differences related to exposure
13 and monitor measurement error.
14 Based on the results from the sensitivity analyses, the following key observations are
15 made:
16 • The single most important factor influencing the risk estimates is whether or not a
17 hypothetical threshold exceeding the estimated background level or LML in the studies
18 exists.
19
20 • The following uncertainties have a moderate impact on the risk estimates in some or all
21 of the cities: choice of an alternative estimated constant background level, use of a
22 distributed lag model, and alternative assumptions about the relevant air quality for long-
23 term exposure mortality. Use of a distribution of daily background concentrations had
24 very little impact on the risk estimates.
25
26 During the previous review of the PM NAAQS, EPA provided an illustrative example
27 based on the PM health risk assessment that showed the distribution of mortality risk associated
28 wilh short-term exposure over a 1 -year period. EPA concluded that peak 24-hour PM25
29 concentrations appeared "to contribute a relatively small amount to the total health risk posed by
30 an entire air quality distribution as compared to the risks associated with low to mid-range
31 concentrations" (61 FR at 65652, December 13,1996). Figures 4-8 (a,b) provide an example of
32 the annual distribution of 24-hour PM2 5 concentrations in Detroit and the corresponding
33
January 2005 4-57 Draft ~ Do Not Quote or Cite
-------
too
80-
80-
70
|.j
I
30
20-
10-
0
15 20
25 30
24-Hr PM-PM2.5
Concentration (ug/m3)
35 40
1 Figure 4-8a. Distribution of 24-Hour PMZ 5 Concentrations in Detroit (2003 Air Quality
2 Data).
3 Fi
4
as-
83-
7S-
73-
«»•
83-
58-
3 53-
§ 4S-
2 43-
!f»:
*l-
ji IB
| 13-
-2!
-7-
-12
-17-
-22-
-27-
... .. ...
A
4
A
1 5 10 15 20 25 30 35
• •
•
• • •
24-Hr PM-2.5 Concentration (ug/m3(
-*- Mean Estimate of # of Deaths
• 2.5th% Estimate of « of Deaths
A 97.5tti% Estimate of # of Deaths
* A
A
A
• |f jj ft e:>
gure 4-8b. Estimated Non-Accidental Mortality in Detroit Associated with PM2 5
Concentrations (2003 Air Quality Data) (Based on Ito, 2003).
January 2005
4-58
Drq/T - Do Not Quote or Cite
-------
1 distribution of estimated mortality incidence (for PM2 5) based on the short-term exposure
2 epidemiology study included in the current PM risk assessment.19 Consistent with the
3 observation made in the previous PM NAAQS review, the highest peak 24-hour PM2 5
4 concentrations contribute a relatively small amount to the total health risk associated with short-
5 term exposures on an annual basis based on typical distributions observed in urban areas.
6 •
7 4.4 RISK ESTIMATES ASSOCIATED \VITH JUST MEETING THE CURRENT
8 PM2 s STANDARDS
9 4.4.1 Base Case Risk Estimates
10 The second part of the PM2S risk assessment estimates the risk reductions that would
11 result if the current annual PM2 s standard of 15 (ig/m3 and the current daily PM25 standard of 65
12 jig/m3 were just met in.the assessment locations. This part of the risk assessment only considers
13 those locations that do not meet the current standards based on 2001-2003 air quality data (i.e.,
14 Detroit, Philadelphia, Pittsburgh, Los Angeles, and St. Louis). As noted previously, the 15
15 |ig/m3 annual average standard is the controlling standard in all five study areas, consequently,
16 just meeting this standard also results in each of these areas meeting the 65 ng/ni3, 24-hour
17 standard.
18 ' The percent rollback necessary to just meet the annual standards depends on whether the
19 maximum or the spatial average of the monitor-specific annual averages is used. For the risk
20 assessment described in the TS D and discussed here, the approach used to simulate just meeting
21 the current annual average standard for the base case risk estimates used the maximum of the
22 monitor-specific annual averages as.shown in Table 4-10. Since an area could potentially use
23 the spatial average of the community-oriented monitors to determine whether or not it met the
24 annual average standard, Table 4-10 also presents the percent rollbacks and annual average
25 design values that would have resulted from using this alternative approach in each urban study
26 area which does not meet the current annual standard and which meets the minimum criteria for
19The Detroit PM25 example uses the C-R function for non-accidental mortality from Lippmann et al.
(2000), reanalyzed in Ito (2003).
January 2005 4-59 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
Table 4*10. Air Quality Adjustments Required to Just Meet the Current Annual PMZ 5
Standard of 15 lig/m3 Using the Maximum vs. the Average of Monitor-
Specific Averages
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Assessment
Location
Detroit
Los Angeles*
Philadelphia
Pittsburgh
St. Louis
Percent Rollback Necessary to
Just Meet the Current Annual
PM1S Standard
Using
Maximum of
Monitor-
Specific
Annual
Averages
28.1%
59.2%
10.9%
35.0%
17.9%
Using Average
of Monitor-
Specific Annual
Averages
11.5%
—
-0.9%
22.8%
13.5%
Design Value Based on 2001-2009
Data,
Annual
Based on
Maximum
Monitor
19.5
23.6
16.4
21.2
17.5
Annual Based on
Average of
Monitor-Specific
Annual Averages
16.5
— •
14.9
18:4
16.8
*Los Angeles does not meet the minimum requirements for use of spatial averaging.
Source: Abt Associates (2005)
use of spatial averaging. A sensitivity analysis examining the impact of using design values
based on spatial averaging is discussed in section 4.5.3.2.
Drawing on the detailed risk estimates contained in Exhibit 8.1 and Appendix E of the
TSD (Abt Associates, 2005), Figure 4-9 displays the estimated percent reductions in total
incidence for non-accidental mortality associated with short-term exposure to PM25
concentrations when air quality goes from as is concentrations to just meeting the current annual
and daily PM25 suite of standards in four of the risk assessment study area that do not meet the
current standards.20 The point estimates generally are in the range of 0.3 - 0.5 percent reduction
20C
Short-term exposure non-accidental mortality estimates were not included for Philadelphia because the C-
R function did not include confidence limits for this endpoint.
January 2005
4-60
Draft - Do Not Quote or Cite
-------
ia>
«,::
OZ?W
ftl!
P0«
m
too
JBitef
wc*»
;;•»;;:
!J!!J;!JJ!J;!J;r;;;r;
SiqwSpg: ;i;
M>?:?::J':;;; :;
P"w: :..:.::.:..:, .
•«:;•!••••••••••••••
ap
«-
e
uo
tjunpH
V
j?
««•
il!*!!:;!;:!!!!!!!!!
1
-J
•!
:;ni:;i:;::!::;i;:!;;st;-tb«H»:=;M;;(;;!:;
:k PMj s Concentrations
idental Mortality
;tb«:Ar>g«ile!9:::::::::::::::::::::::::::"::::::::::::::::::-::::::;Plltsburiah::: I::::::::::::::::::::::;:::::;:::
Reduction of Health Risks Associated with Rolling Ba<
dards (and 95 Percent Confidence Intervals): Non-Ace
xposure to PM2.S.
!:i:!:;;;!::i;ri:;;:::;;:;;:O«ti*Hi:;;Ei;;i:ii:ii:ii:!i:i;:!i:!i:!i;;;;;;;;i
Figure 4-9. Estimated Annual Percentage
to Just Meet the Current Stan
Associated with Short-Term E
Source: Abt Associates, 2005
S
i
I
.
-------
*"T":TTT"""":""*:"::*::**':?:T*":""*!;":";:"*:;*?;; ;:"~:1~7?:^*1""*^:":^*"
i if ill
: • •;: ^'
it!
*;;;; ;;;;
-------
'i
1
A '
on
§
I
-------
•sss*i
O H "S t* 2
•_B « oa C «j
« S N a. S
a S*g
o
&
cn -O
n
es
£
5 "*
DS
en
Effect
ealt
0 •
I/I t— 1
-a
1
I
Al
3>g
it
13 t9D
0
o ^
3
55
a-
w ^.
11
p
1«
** ^ 4C
I"8!?
c ^*
•a | g
S S®
* Is
** S: 9*
4 &-S
fN (S
(N O
cK
*2\
o
Ci
"*•
00
30)
$*
fa
£ M
^ «
C
§5
"3 S°
SI
,2
^B
S.-S
D O
OQ G
'« x
§3
•o u
fl) •,
M O
13
TH flj
•c|
S d
u o
S g
"S *
!1
OT 13
"C CU
M
•c
.•^sr
•3 *
I!
r*i c
8.8
£>!
O °3
•S M
5 ~
11
a §
•II
o
>,
?l
8
a
t
t
-------
1 in total incidence, which represent from about 11 'to 45% reductions in the PM-related incidence.
2 Figure 4-10 shows the estimated percent reductions in total incidence for mortality associated
3 with long-term exposure to PM2 5 concentrations for this same air quality change in all five of
4 the risk assessment study areas that do not meet the current standards. The point estimates are in
5 the range 0.5 to nearly 4.0 percent reduction in total incidence, which represents from about 18
6 to 59% reductions in the PM-related incidence. Table 4-11 shows the estimated short- and long-
7 exposure mortality incidence to facilitate a comparison both within and across the five study
8 areas. For short-term exposure mortality, single-pollutant, non-accidental mortality estimates are
9 selected since they are available for four of the study areas, and cardiovascular mortality is
10 shown for the fifth area, Philadelphia. For long-term exposure mortality, the ACS-extended
11 estimates for total (all cause) mortality are selected for comparison. In Table 4-11 risk
12 ' reductions are expressed both as a percentage reduction in the PM25-associated mortality and as
13 a percentage of the total mortality due to PM2.5 and other causes. As expected, the reductions in
14 both short- and long-term exposure mortality.associated with PM2S are ranked in the same order
15 as the percent rollback required to bring as is concentrations down to just attaining the current
16 standards, with Los Angeles having the biggest percentage reduction in risk and Philadelphia the
17 least. Also, both the risk remaining upon just meeting the current PM2 5 standards and the size of
18 the reduction in risk in moving from as is concentrations to just meeting the current standards are
19 larger associated with long-term exposure mortality estimates.
20 .
21 4.4.2 Sensitivity Analyses
22 The base case risk assessment used a proportional rollback approach to adjust air quality
23 distributions to simulate the pattern that would occur in an area improving its air quality so that it
24 just meets the current annual average PM2 5 standard. The support for this approach is briefly
25 discussed in section 4.2.3 and in more detail in Appendix B of the TSD(Abt Associates, 2005). •
26 While the available data suggest that this is a reasonable approach, other patterns of change are
27 possible. In a sensitivity analysis an alternative air quality adjustment approach was used which
28 reduced the top 10 percent of the distribution of PM2S concentrations by 1.6 times as much as the
January 2005 4-65 Draft - Do Not Quote or Cite
-------
2
3
4
5
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
lower 90 percent of concentrations. The result of this alternative hypothetical adjustment which
reduces the highest days more than the rest of the distribution showed only a small difference
(less than 1%) in the percent change in PM-associated incidence (see Exhibit 8.2 and Appendix
E, exhibits E5-E8 in Abt Associates, 2005).
4.4.3 Key Observations
Sections 4.4.1 and 4.4.2 have presented the PM health risk estimates and sensitivity
analyses associated with just meeting the current PM25 standards. Presented below are key
observations resulting from this part of the risk assessment:
There is a wide range of reductions in PM25-related incidence across the five urban areas
analyzed which is largely due to the varying amount of reduction in ambient PM25
concentrations required in these urban areas to just meet the current PM2 5 standard. For
example, using single-pollutant models the percent of PM25-related incidence reduced for
short-term, non-accidental mortality ranges from about 45% in Los Angeles to about 18%
in St. Louis. Similarly, using the ACS-extended study the percent of PM25-related
incidence reduced for long-term exposure mortality ranges from roughly 60% in Los
Angeles to about 18% in Philadelphia.
The risk estimates associated with just meeting the current PM25 standards incorporate
several additional sources of uncertainty, including: (1) uncertainty in the pattern of air
quality concentration reductions that would be observed across the distribution of PM
concentrations in areas attaining the>standards ("rollback uncertainty") and (2)
uncertainty concerning the degree to which current PM risk coefficients may reflect
contributions from other pollutants, or the particular contribution of certain constituents
of PM2 5, and whether such constituents would be reduced in similar proportion to the
reduction in PM2 5 as a whole.
At least one alternative approach to rolling back the distribution of daily PM2 5
concentrations, in which the upper end of the distribution of concentrations was reduced
by a greater amount than the rest of the distribution, had little impact on the risk
estimates.
January 2005
4-66
Draft - Do Not Quote or Cite
-------
1 4.5 RISK ESTIMATES ASSOCIATED WITH JUST MEETING ALTERNATIVE
2 PMj s AND PM1M3 STANDARDS
3 4.5.1 Base Case Risk Estimates for Alternative PM2 5 Standards
4 The third part of the PM2.5 risk assessment estimates the risk reductions associated with
5 just meeting alternative suites of annual and-daily PM25 standards. For the five urban areas that
6 exceeded the current PM2 5 suite of standards (i.e., Detroit, Los Angeles, Philadelphia,
7 Pittsburgh, and St. Louis), the estimated risk reductions were those associated with a further
8 reduction in PM25 concentrations from just meeting the current standards to just meeting various
9 suites of alternative PM2 5 standards. For the four urban areas that met the current PM2 5
10 standards based on 2001-2003 levels (i,e., Boston, Phoenix, San Jose, and Seattle), Ihe estimated
11 risk reductions were those associated with a reduction from as is air.quality levels to just meeting
12 various suites of alternative PM2 5 standards.
13 • The selection of the suites of alternative annual and daily standards included in the risk
14 assessment was based on the preliminary staff recommendations described in Chapter 6 of the
15 draft 2003 Staff Paper (EPA, 2003) and consideration of public and CAS AC comments. Annual
16 standards of 15,14,13, and 12 |xg/m3 were each combined with 98th percentile daily standards of
17 40, 35,30, and 25 ng/m3, and 99th percentile daily standards at the same levels.21 In addition, an
18 annual standard of 15 jig/m3 was combined with a ninety -ninth percentile daily standard of 65
19 jig/m3. The combinations of annual and daily alternative standards used in the PM25 risk
20 assessment are summarized in Table 4-12. The same proportional adjustment approach used to
. +
21 simulate air quality just meeting the current standards, described previously in section 4.2.3.2
22 and in section 2.3 of Abt Associates (2005), was used to simulate air quality just meeting the
23 various alternative suites of standards. Table 4-13 provides the design values for the annual and
21In four of the five urban areas that do not meet the current suite of PlV^.s standards, annual standards
within the range of 12 to 15 pg/m3 combined with the current daily standard of 65 ng/m3, using a 98th percentile
form, require the same reduction as when these annual standards are combined with a daily standard of 40 ng/m1,
using the same daily form. Therefore, the risk assessment only included the 14 pg/m5 annual standard combined
with the current daily standard for the one location (i.e., Philadelphia) and annual standard scenario where there was
a difference in the reduction required between daily standards of 40 and 65
J
January 2005
4-67
Draft - Do Not Quote or Cite
-------
1 98* and 99th percentile daily standards for all of the PM2 5 risk assessment study areas based on
2 air quality data from 2001-2003 for the base case risk estimates.
3 The estimated risk reduction in total non-accidental mortality, presented both in terms of
4 percent reduction in total incidence and in number of cases avoided, associated with short-term
5 PM2s exposures for alternative annual standards combined with ninety-eighth and ninety-ninth
6 percentile daily standards, respectively, are given in Figures 4-11 and 4-12 for Detroit.
7 Similarly, the estimated risk reduction in total mortality associated with long-term PM25
8 exposures for these same alternative standards are given in Figures 4-13 and 4-14 for Detroit.
9 Similar figures for the other risk assessment locations and additional risk estimates for cause-
10 specific mortality, hospital admissions, and respiratory symptoms (depending on location)
11 associated with alternative standards are presented in Chapter 8 and Appendix F of Abt
12 Associates (2005). As with the estimated risk reductions presented earlier for just meeting the
13 current PM2S standards, when the percent reduction is expressed in terms of the estimated
14 reduction in PM-related incidence rather than total incidence, the changes are much larger. The
15 complete set of risk estimates is presented in Exhibits 8.5a through 8.5h for Detroit and the
16 exhibits in Appendix F for the other 4 locations in the TSD (Abt Associates, 2005).
17 Some interesting patterns can be observed in the estimated risk reductions displayed in
18 Figures 4-11 through 4-14. For example, in Figures 4-11 and 4-13 one observes there are no
19 estimated reductions in risk in going fromjust meeting the current 15 |ig/m3 annual standard/65
20 |ig/m3 98th percentile daily standard to either a 40 or 35 jig/m3 98th percentile daily standard with
21 the same 15 |ig/m3 annual standard. The reason for this is that the 28.1% reduction, required
22 based on the 3-year estimated design value, when applied to the 2003 PM2 5 distribution for the
23 composite monitor to meet the current 15 [ig/m3 annual standard, brings down the 98th percentile
24 daily value to below 35 ng/m3. Thus, there is no additional reduction in air quality or risk when
25 either a 40 or 35 ^g/m3 98th percentile daily standard is considered in combination with a 15
26 |ig/m3 annual standard. Meeting lower daily 98th percentile standards of 30 or 25 jig/m3 when
27 combined with the current annual standard do require additional air quality reductions and, thus,
28 result in additional estimated risk reductions compared to just meeting the current suite of
January 2005
4-68
Draft - Do Not Quote or Cite
-------
t
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
Table 4-12. Alternative Sets of PM,, Standards Considered in the PM,« Risk
•is •
Assessment"1
Annual
Standard
15
14
13
12
98th Percentile Daily Standard
65
X**
40
X
X
X
X
35
X
X
X
X
, 30
X
X
X
X
25
X
X
X
X
99th Percentile Daily Standard
65
X
40
X
X
X
X
35
X
X
X
X
• 30
X
X
X
X
25
X
X
X
X
*A11 standards are in
**Only in Philadelphia.
Table 4-13. Estimated Design Values for Annual and 98th and 99th Percentile Daily PM2.5
Standards Based on 2001-2003 Air Quality Data*
Location
Boston
Detroit
Los Angeles
Philadelphia
Phoenix
Pittsburgh
St. Louis
San Jose
Seattle
Annual
14.4
19.5
23.6
16.4 . .
11.5
21.2
17.5
14.6
11.1
98"1 Percentile Daily
44
44
62
51
35
63
42
47
41
99th Percentile Daily
60
48
96
89
41
70
46
53
48
*The calculation of design values is explained in Schmidt (2005). All design values are in |ag/m3.
January 2005
4-69
Draft - Do Not Quote or Cite
-------
V)
VA
fe
I
I
i ;;;;;As«8.
::::if.:
;;!!l7;
•a.
I
III
III
ill
w
'5F5
4
£
-------
0
t
^wx;:
%xsii$i\l
::::::*£"::::::::«
;<*.;
S
v/
w
I
oo
o "^
oo o\ o i— i
<— — (S CS
en
tN
-------
V)
\o
•s
fe
; *;=
...
:".:: a :.;;.:::
B S fi
= = -
£<3 §
fill
>•§ s
•8
I j 2
£ J 1
ill
|ls
a« «
.s r
s n
IIP
s.«» *3 -.
I5 §8
^ eg S ^
VJ
. -d I
C"3 AI *^ O
. M JT u o
5 O ^ Q aj «
|| | If!
J I 1 I 1 |
r« ® "gD S (S §
W W S 00 » co
s
a
<$
rp
•4
t
4
£
b
3
s
-------
t
•
p^
; :: Reduction in Incroence (Num&sr of c.
m^^^-^^^^
:;;;;;;::;:::::: i^|iflp-0f ; w*^yw*wwwww.
' -:--:••• .frjft^ ^ • i;-yy ^;.- VQ-J^A'I!
iiiiiMi
a
ft
*
*
»F
«0
»
**
kfM
ira
":«:
•4::^-
;;:::*»
:::::::&:
:::; :F*
V»B»rt«y>.
JS0
..„•„
J&
:"Aj!
**
••v
ipr
;-!*
PT ^
Sfr
l> P
»pa
cie
- • -i
:"
jsy'stei ;;:
!:: ::;::::
::; ; ::.»•;
::: : ::.«T
J::::: :::;::
|^;;|i
iiiii; ;;:in
Rolling Back PM2S
Current Daily Standard o
and Daily 99th Percentile
ng'Term Exposure Mortality Associated with 1
Current Annual Standard of 15 ug/m3 and the
t Just Meet Alternative Suites of PM2 5 Annual
d
. Estimated Annual Reduction in Lo
Concentrations that Just Meet the *
ug/m3 to PMj s Concentrations that
Standards: Detroit, MI, 2003.*
*Based on Pope et al. (2002) - ACS extende
Source : Abt Associates (2005)
Q
fe
8
S
'CS
-------
1 standards. The maximum incremental risk reduction from the current standards, with respect to
2 both short- and long-term exposure PM2 5-associated mortality, is estimated to occur for meeting
3 the daily 98th and 99th percentile daily standards set at 25 [ig/m3. For daily standards set at this
4 level the estimated risk reduction does not depend on the level of the annual standard within the
5 range of standards considered. Within four of the five study areas, just meeting 98th or 99th
6 percentile daily standards set at 30 p.g/m3 results in the same short- or long-term exposure
7 mortality risk reductions no matter which annual standards (from 12 to 15 ng/m3) they are paired
8 with. Similar, although not identical, patterns are observed in the other four risk assessment
9 locations that do not meet the current PM25 standards (see Figures Fl through F14 in the TSD
10 (Abt Associates (2005)).
11
12 4.5.2 Base Case Estimates for Alternative PMto_2S Standards
13 The second part of the PM]0.2 5 risk assessment estimates the risk reductions associated
14 with just meeting alternative daily PM10_2 5 standards for the three locations examined earlier
15 (Detroit, St. Louis, and Seattle). Estimated reductions in risk were developed for going from as is
16 levels (based on 2003 air quality) to just meeting alternative PM10.2 5 standards. Staff selected
17 the alternative daily standards to be included in the risk assessment based on the preliminary staff
18 recommendations described in Chapter 6 of the draft 2003 Staff Paper (EPA, 2003) and
19 consideration of public and CASAC comments. Table 4-14 summarizes the sets of 98th and 99th
20 percentile daily standards that were included in the PM10_2 5 risk assessment. The estimated design
21 values which were used to determine the air quality adjustment to be used in simulating just
22 meeting alternative PM10_25 standards are shown in Table 4-15.
23 The estimated annual reduction in hospital admissions for ischemic heart disease,
24 presented both in terms of percent reduction in total incidence and in number of cases avoided,
25 associated with short-term PM,0.2 5 exposures for alternative 98th and 99th percentile daily
26 standards, respectively, are given in Figure 4-15 for Detroit. Daily PM10_25 standards set at 80
27 (for 98th percentile form) and 100 or 80 (for 99th percentile form) result in no reduction in risk in
28 Detroit. The reason why no estimated risk reductions are observed with these alternative
29 standards is that the percent reduction of PM10.2 5 concentrations at the composite monitor to just
30 meet a standard is determined by comparing the alternative standard level with the design value
January 2005 4-74 Draft - Do Not Quote or Cite
S
-------
••
t
1 for that location based on 2001-2003 air quality data. In Detroit, the design value for the 98th
2 percentile daily PM10.25 standards is 70 jig/m3 whereas the 98th percentile daily value in 2003 is
3 105.9 ug/m3. Because the design value is lower than 80 ng/m3, the highest 98th percentile daily
4 PM10_2 5 standard, zero risk reductions were estimated to result from this standard, even though the
5 98th percentile daily value at the composite monitor in 2003,105.9 jig/m3, is well above the
6 standard level. Similarly, the design value for the 99th percentile daily PM10.2 5 standards is 77
7 jig/rn3 for Detroit, whereas the 99th percentile daily value at the composite monitor in Detroit in
8 2003 is substantially greater than 100 (ig/m3, the highest 99th percentile daily PM10.2 s standard. So
9 zero risk reductions were similarly estimated to result from both a 100 and 80 ng/m3 standards. In
10 general, estimated risk reductions increase and the confidence intervals around the estimates
11 widen as lower daily standards are considered.
12 As expected, the maximum reduction in risk is achieved with the 98* percentile 25 ^ig/m3
13 standard and 99th percentile 30 jig/m3 standard. The point estimate is that about a 4% reduction in
14 hospital admissions for ischemic heart disease, equating to roughly 450 fewer cases, would result
15 from meeting either of these daily standards. Similar patterns in risk reduction are observed for
16 the other hospital admission endpoints in Detroit which are included in Chapter 9 of Abt
17 Associates (2005). Additional risk estimates for hospital admissions for asthma in Seattle and
t
18 cough and lower respiratory symptoms in St. Louis can be found in Appendix G of Abt
19 Associates (2005). Based on the point estimates, there are no risk reductions associated with just
20 meeting daily 98th percentile PM10.2 5 standards of 80 |ig/m3 in Detroit, and 80,65, and 50 ng/m3 in
21 St. Louis or Seattle. Similarly, there are no risk reductions associated with just meeting daily 99th
22 percentile PM10.25 standards of 100 or 80 [ig/m3 in Detroit, and 100, 80, or 60 [ig/m3 in St. Louis..
23 or Seattle.
24
25 4.5.3 Sensitivity Analyses for Alternative PM1S and PM10.j 5 Standards
26 4.5.3.1 Hypothetical Thresholds
27 An important observation from the sensitivity analyses on estimated health risks
28 associated with "as is" PM2 5 concentrations was that the impact of hypothetical thresholds was
29 the greatest on the estimated risks.. In order to gain insight into the impact of this important
30
January 2005 • . 4-75 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
$
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Table 4-14. Alternative PM,n,. Standards Considered in the PM.n •, < Risk Assessment*
M8-2.5
Daily
Standards Based on the 98th Percentile
Value
80
65
50
30
25
Daily Standards Based on the 99th Percentile
Value
100
80
60
35
30
*A11 standards are in |ig/m3.
Table 4-15. Estimated Design Values for 98th and 99th Percentile Daily PM10,2S Standards
Based on 2001-2003 Air Quality Data9"
Location
Detroit
St. Louis
Seattle
. 98th Percentile Daily
70
33
31
99th Percentile Daily
77
47
39
"The calculation of design values is explained in Schmidt (2005). All design values are in
January 2005
4-76
Draft - Do Not Quote or Cite
f
-------
Reduction i
imoiiP
«
A|BP
i i
II
II
"o to
ft >,
•£ 1
•a
e
£
09
on £
5«
a
cn
e
"5
fi
1
e
S
e
• m
73 |
*S.§
S*
ss
0 B
if
I
in
O
8
•*—/
CO
I
3
S §^«
« y g
= ^1:
5 2 i
< *?« '
« R rf §
S « g-g
fe
a
t
s
R
I
ts tn
vo
oo
O ^^
en
vo
-------
1 uncertainty on the risk estimates, an additional set of sensitivity analyses was developed to
2 examine the impact of different hypothetical threshold assumptions on estimated risks associated
3 with just meeting the current and alternative PM25 standards and alternative PM10_25 standards.
4 For those locations and cases where either the current PM25 standards or any of the alternative
5 suites of standards were already met under as is air quality, the estimated risks associated with
6 "as is" PM2 5 (or PM]0_2 5) concentrations in excess of either background or the LML for the health
7 endpoint, whichever is greater, were calculated.
8 For PM25 this sensitivity analysis included estimates of risk for all cause mortality,
9 cardiopulmonary mortality, and lung cancer mortality associated with long-term exposure to
10 PM15 based on Pope et al. (2002) - ACS extended. Since the patterns observed were identical,
11 only the all cause mortality results are presented in Appendix 4B (See Abt Associates, 2005 for
12 the cause-specific mortality estimates). In addition, this sensitivity analysis also included non-
13 accidental mortality (or cause-specific if there was no suitable function for non-accidental
14 mortality available) associated with short-term exposure to PM25. As in the earlier sensitivity
15 analysis for as is air quality, hypothetical thresholds of 10,15, and 20 [ig/m3 were considered for
16 health endpoints associated with short-term exposures, and hypothetical thresholds of 10 and 12
17 M-§/m3 were considered for the mortality endpoints associated with long-term exposure.
18 The sensitivity analysis results for all-cause mortality associated with long-term exposure
19 and mortality associated with short-term exposure for Detroit, Los Angeles, Philadelphia,
20 Pittsburgh, and St. Louis are shown in Appendix 4B to this Chapter (Tables 4B-1 through 4B-10)
21 The results for cardiopulmonary and lung cancer mortality associated with long-term exposure to
22 PM25 based on Pope et al. (2002) - ACS extended are shown in Appendix H of Abt Associates
23 (2005). Not surprisingly, estimated PM-related incidences varied substantially with both
24 hypothetical threshold assumptions and alternative standards. In Detroit, for example, the
25 estimated number of cases of non-accidental mortality associated with short-term exposure to
26 PM2 5 when the current standards are just met decreases from 115, under the assumption of no
27 threshold, to 54, 26, and 12 under hypothetical threshold assumptions of 10,15, and 20 jig/m3,
28 respectively. Because meeting increasingly lower level standards removes estimated cases at the
29 higher concentrations and considering higher hypothetical thresholds increasingly removes
30 estimated cases at concentrations between background (or the LML) and the threshold, one would
31 expect to see an increase in the percent reduction associated with just meeting alternative
January 2005 4-78 Draft - Do Not Quote or Cite
-------
1 standards for higher hypothetical thresholds. This is exactly what is found. For example, as seen •
2 in Table 4B-1, going from just meeting the current standards (15 [ig/rri3 annual and 65 |ig/m3 daily •
3 98th percentile value)'to just meeting the lowest set of standards considered (12 jig/m3 annual and
4 25 ng/m3 'daily 99th percentile value) results in a reduction in short-term exposure mortality
5 incidence of (115 -75)7115 = 34.8 percent under the assumption of no threshold, but under the
6 assumption of a threshold of 10 |ig/m3 it results in a reduction of (54 - 22)/54 = 59 percent." Under
7 hypothetical short-term exposure thresholds of 15 and 20 jig/m3, the reductions are 73 percent and
8 83 percent, respectively. 'As shown in Table 4B-2 for all-cause mortality associated with long-
9 term exposure in Detroit, the reduction in mortality incidence is even more dramatic when'
r . ,
10 alternative hypothetical thresholds are considered. Going from just meeting the current standards
11 to j ust meeting the lowest set of standards considered (12 ng/m3 annual and 25 jig/m3 daily 99th
12 percentile value) results in a reduction in long:term exposure mortality incidence of (522-
13 207X522= 60% under the assumption of no threshold, but under the assumptions of a long-term
14 exposure threshold of 10 ng/m3 it results in a reduction of (282 - 0)/282 =100 percent. With a
15 hypothetical long-term exposure threshold of 12 jig/m3 estimated incidence is reduced to 41 upon
16 just meeting the current suite of standards and a 100% reduction is achieved upon meeting either
17 a. 15 jig/m3 annual standard with a 30 [ig/m3 daily 98th percentile standard or a 1'4 (ig/m3 annual
18 with a 40 (ig/m3 daily 98th percentile value. The same general patterns can be seen in all
19 locations and for all health endpoints considered:
20 The sensitivity analysis results examining alternative PM^ 5 standards with hypothetical
21 thresholds associated with short-term exposure morbidity endpoints for Detroit, Seattle, and St.
22 Louis also are shown in Appendix B to this Chapter (Tables 4B-11 through 4B-13). The health
23 endpoints included hospital admissions for ischemic heart disease in Detroit; hospital admissions
24 for asthma (age < 65) in Seattle; and days of cough among children in St. Louis, all associated
25 with short-term exposures to PM10.2 5 exposures Hypothetical short-term exposure thresholds 'of
26 10,15, and 20 jig/m3 were considered.
27 4.5.3.2 Spatial Averaging Versus Maximum Community Monitor
28 As -discussed previously in section 4.2.3.2, under the current annual PM2 5 standard urban
29 areas may, under certain circumstances,' use the average of the annual averages of several
30 monitors within an urban area to determine compliance with the annual standard, commonly
31 referred to as the "spatial averaging approach." Four of the five urban areas included in the PM25
January 2005 ' 4-79 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
J7
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
• . =• In four of the five risk assessment locations that do not meet the current PM2 s standards,
daily standards of 40 )ig/m3, 98th percentile or 65 (ig/m3, 99th percentile when combined
with the current 15 [ig/m3 annual standard provide no additional risk reduction in terms of
short-term exposure mortality. • . t
• In all five of the risk assessment locations that do not meet the current PM2.5 standards,
the maximum risk reduction with respect to both short- and long-term PM2 5-associated
mortality is estimated to occur upon meeting the 98th and 99th percentile daily standards
set at 25 jig/m3. For these standards the estimated risk reduction does not depend on the
level of the annual standard within the range of standards examined.
• For four of the five risk assessment locations the estimated risk reduction within each
area associated with meeting either a 98th or 99th percentile daily PM25 standard set at 30
|ig/m3 is the same no matter which annual standard is included within the range of
standards examined. -
• For the PM10.2 5 risk estimates, the maximum reduction in risk is achieved with the 98th
percentile 25 u.g/m3 standard or 99th percentile 30 jig/m3 standard. The point estimate is '
that about a 4% reduction in hospital admissions for ischemic heart disease, equating to .
roughly 450 fewer cases, would result from meeting either of these daily standards. The
confidence intervals get significantly larger as lower PM,0.2 5 standards are considered.
Similar patterns in risk reduction are observed for the other hospital admission endpoints
in Detroit
4
• Based on the point estimates, there are no risk reductions associated with just meeting
daily 98th percentile PM10.2 5 standards of 80 (ig/m3 in Detroit, and 80,65, and 50 [ig/m3 in
St. Louis or Seattle. Similarly, there are no risk reductions associated with just meeting
daily 99th percentile PM10.2 5 standards of 100 or 80 [ig/m3 in Detroit, and 100, 80, or 60
|ig/m3 in St. Louis or Seattle.
Section 4.5.3 presented the results of the following two sets of sensitivity analyses: (1)
considering the impact on risk estimates associated with just meeting the current and alternative
PM2 j standards and alternative PM10.2 5 standards when hypothetical threshold models are included
and (2) considering the impact on risk estimates associated with just meeting the current and
alternative PM25 standards when the average of the annual averages of several monitors within an '
urban area are used to determine compliance with the annual standard, commonly referred to as
the "spatial averaging approach." Presented below are key observations resulting from this part of
the risk assessment: • •
• For short-term exposure mortality associated with PM2 5 there is a significant decrease in •
the incidence avoided as one considers higher hypothetical thresholds. There also is a
January 2005
4-82
Draft - Do Not Quote or Cite
-------
1 significant increase observed in the percent reduction in PM-associated incidence upon
2 ' just meeting alternative standards with higher hypothetical thresholds. The reduction in
3 incidence and increase in percent reduction in PM-associated incidence is even more
4 dramatic for long-term exposure mortality as higher alternative hypothetical thresholds
5 are considered.
6
7 • For short-term exposure morbidity associated with PM10_25, there is a significant decrease
8 in the incidence avoided as one considers higher hypothetical thresholds.
9 '
10 • There is an increase in estimated short-term exposure mortality incidence associated with
11 PM15 when a spatial averaging approach is used to determine compliance with the current
12 annual standard or alternative suites of standards where the daily standard is not the
13 controlling standard.
t
January 2005 4-83 Draft - Do Not Quote or Cite
-------
I REFERENCES
2
3 Most Chapter 4 references are available at the end of Chapter 3. References not listed at the end
4 of Chapter 3 are listed here.
6 Abt Associates Inc. (1996). "A Particulate Matter Risk Assessment for Philadelphia and Los Angeles." Bethesda,
7 MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental Protection
8 Agency, Contract No. 68-W4-0029. July 3 (revised November). Available:
9 http:/.'Vww.epa.gov/ita/naags/standards/pm/s_pm pr td.html.
10
11 Abt Associates Inc. (1997a). Abt Associates Memorandum to U.S. EPA. Subject: Revision of Mortality Incidence
12 Estimates Based on Pope et al. (1995) in the Abt Particulate Matter Risk Assessment Report. June 5,1997.
13
14 Abt Associates Inc. (1997b). Abt Associates Memorandum to U.S. EPA. Subject: Revision of Mortality Incidence
15 Estimates Based on Pope et al. (1995) in the December 1996 Supplement to the Abt Particulate Matter Risk
16 Assessment Report. June 6,1997.
17
18 Abt Associates Inc. (2002). Proposed Methodology for Particulate Matter Risk Analyses for Selected Urban Areas:
19 Draft Report. Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S.
20 Environmental Protection Agency, Contract No. 68-D-03-002. Available:
21 http://wvvw.epa.gOv/ttn/naaqs/standards/pm/s pm cr td.html.
22
23 Abt Associates Inc. (2003a). Abt Associates Memorandum to U.S. EPA. Subject: Preliminary Recommended
24 Methodology for PMi0 and PM10.2.s Risk Analyses in Light of Reanalyzed Study Results. April 8,2003.
25 Available: http://w^w.epa.aov/ttn/iiaaqs/sta!idards/piii/s pm cr td.html.
26
27 Abt Associates Inc. (2003b). Particulate Matter Health Risk Assessment for Selected Urban Areas: Draft Report.
28 Bethesda, MD: Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
29 Protection Agency, Contract No. 68-D-03-002. Available:
30 http://mvw.epa,eov/ttrj/naaqs/standards/CTn/s pm cr td.html.
31
32 Abt Associates Inc. (2005). Particulate Matter Health Risk Assessment for Selected Urban Areas. Draft Report.
33 Bethesda, MD. Prepared for the Office of Air Quality Planning and Standards, U.S. Environmental
34 Protection Agency, Contract No. 68-D-03-002. Available:
35 hUo://www epa.aov/ttn/naaQ3/stendards/pra/s_pm cr ld.html.
36
37 Center for Disease Control (2001). CDC Wonder. Available: hitp://wonder.cdc.aov/.
38
39 Deck, L. B.; Post, E.S.; Smith, E.; Wiener, M.; Cunningham, K.; Richmond, H. (2001). Estimates of the health risk
40 reductions associated with attainment of alternative particulate matter standards in two U.S. cities. Risk
41 Anal. 21(5): 821-835.
42
43 Environmental Protection Agency (2001). Particulate Matter NAAQS Risk Analysis Scoping Plan, Draft. Research
44 Triangle Park, NC: Office of Air Quality Planning and Standards. Available:
45 httr>://www.eKa.eov/ttn,''!iaaqs/standai'ds/pfii/sj3m cr tdJitml.
46
47 Hopke, P. (2002). Letter from Dr. Phil Hopke, Chair, Clean Air Scientific Advisory Committee (CASAC) to
48 Honorable Christine Todd Whitman, Administrator, U.S. EPA. Final advisory review report by the CASAC
49 Particulate Matter Review Panel on the proposed particulate matter risk assessment May 23,2002.
50 Available: http:/AA^vTv.epa.gpy/sab/pdt7casacady02002.pdi.
51
January 2005 4-84 Draft - Do Not Quote or Cite
t
-------
1 Langstaff, J. (2004). OAQPS Staff Memorandum to PMNAAQS Review Docket (OAR-2001-0017). Subject A
2 Methodology for Incorporating Short-term Variable Background Concentrations in Risk Assessments.
3 December 17,2004. Available: http://www.epa.gOv/ttii/riaaqs/standards/Pin/s otn.cr td.html.
*4
5 Langstaff, J. (2005). OAQPS Staff Memorandum to PMNAAQS Review Docket (OAR-2001-0017). Subject:
6 Estimation of Policy-Relevant Background Concentrations of Particulate Matter. January 27,2005.
7 Available: htlp://w\vw.er>a gov/tfo'naacis/standards/pm/s pm crtd.html.
8
9 National Academy of Sciences (2002). Estimating the Public Health Benefits of Proposed Air Pollution Regulations.
10 Washington, D.C.: The National Academy Press. Available:
11 httn://www.rnD.edu/books/0309086094/hurd/. !
12 . .
13 Post, E.; Deck, L.; Laratz, K.; Hoaglin. D. (2001). An application of an empirical Bayes estimation technique to the
14 estimation of mortality related to short-term exposure to particulate matter. Risk Anal. 21(5): 837-842.
15
16 Schmidt, M.: Mintz, D.; Rao, V.; McCluney, L. (2005). U.S. EPA Memorandum to File. Subject: Draft Analyses of
17 2001-2003 PM Data for the PMNAAQS Review. January 31,2005. Available:
18 btto:/Avww.ena.eov/oar/oaaiiS/mn25/docs.httul.
19
20 Science Advisory Board (2004). Advisory on Plans for Health Effects Analysis in the Analytical Plan for EPA's
21 Second Prospective Analysis - Benefits and Costs of the Clean Air Act, 1990-2000. Advisory by the Health
22 Effects Subcommittee of the Advisory Council for Clean Air Compliance Analysis. EPA SAB Council -
23 ADV-04-002. March. Available: http://ww\^gpa.gov/sciencel/{xif/coujicil adv 04QQ2.pdf.
24
t
January 2005 4-85 Draft - Do Not Quote or Cite
-------
1 5. STAFF CONCLUSIONS AND RECOMMENDATIONS ON PRIMARY PM NAAQS
2 5.1 INTRODUCTION
3 This chapter presents staff conclusions and recommendations for the Administrator to
4 consider in deciding whether the existing primary PM standards should be revised and, if so,
5 what revised standards are appropriate.1 The existing suite of primary PM standards includes
6 annual and 24-hour PM2 s standards, to protect public health from exposure to fine particles, and
7 annual and 24-hour PM!0 standards, to protect public health from exposure to thoracic coarse
8 particles. Each of these standards is defined in terms of four basic elements: indicator,
9 averaging time, level and form. Staff conclusions and recommendations on these standards are
10 based on the assessment and integrative synthesis of information presented in the CD and on
11 staff analyses and evaluations presented in Chapters 2 through 4 herein.
*
12 N In recommending a range of primary standard options for the Administrator to consider,
13 staff notes mat the final decision is largely a public health policy judgment. A final decision
14 must draw upon scientific information and analyses about health effects and risks, as well as
/
15 judgments about how to deal with the range of uncertainties that are inherent in the scientific
16 evidence and analyses. The staffs approach to informing these judgments, discussed more fully
17 below, is based on a recognition that the available health effects evidence generally reflects a
18 continuum consisting of ambient levels at which scientists generally agree that health effects are
19 likely to occur through lower levels at which the likelihood and magnitude of the response
20 become increasingly uncertain. This approach is consistent with the requirements of the
21 NAAQS provisions of the Act and with how EPA and the courts have historically interpreted the
22 Act. These provisions require the Administrator to establish primary standards that are requisite
23 to protect public health with an adequate margin of safety. In so doing, the Administrator seeks
24 to establish standards that are neither more nor less stringent than necessary for this purpose.
25 The provisions do not require that primary standards be set at a zero-risk level, but rather at a
26 level that avoids unacceptable risks to public health.
1 As noted in Chapter 1, staff conclusions and recommendations presented herein are provisional; final staff
conclusions and recommendations, to be included in the final version of this document, will be informed by
comments received from CASAC and the public in their reviews of this draft document
January 2005 5-1 • Draft - Do Not Quote or Cite
-------
1 " 5.2 APPROACH
2 As an initial matter, PM25 standards for fine particles and PM]0 standards for thoracic
3 coarse particles are addressed separately, consistent with the decision made by EPA in the last
4 review and with the conclusion in the CD that fine and thoracic coarse particles should continue-
5' to be considered as separate subclasses of PM pollution. As discussed in Chapter 3, section •
6 3.2.3, this conclusion is based in part on long-established information on the differences in
7 sources, properties; and atmospheric behavior between fine and coarse particles, and is
8 reinforced by new information that advances our understanding of differences in human
9 exposure relationships and dosimetric patterns, and the apparent independence of health effects
10 that have been associated with these two subclasses of PM pollution in epidemiologic studies.
11 "In general, in evaluating whether the current primary standards are adequate or whether
12 revisions are appropriate, and in developing recommendations on the elements of possible
13 alternative standards for consideration, staffs approach in this review builds upon and broadens
14 the general approach used by EPA in the last review. In setting PM25 standards in 1997, the
15 Agency mainly used an evidence-based approach that placed primary emphasis on epidemiologic
16 evidence from short-term exposure studies of fine particles, judged to be the strongest evidence
17 at that time, in reaching decisions to set a generally controlling annual PM25 standard and a 24-
18 hour PM2 5 standard to provide supplemental protection. The risk assessment conducted in the
19 last review provided qualitative insights, but was judged to be too limited to serve as a
20 quantitative basis for decisions on the standards. In this review, the more extensive and stronger
21 body of evidence now available on health effects related to both short- and long-term exposure
22 to PM25, together with the availability of much more extensive PM2 5 air quality data, have .
23 facilitated a more comprehensive risk assessment for PM25. As a result, staff has used'a broader
24 approach in this review of the PM2 5 standards that takes into account both evidence-based and
25 quantitative risk-based considerations, placing greater emphasis on evidence from long-term
26 exposure studies and quantitative risk assessment results for fine particles than was done in the
27 last review. Staff has applied this approach to a more limited degree in reviewing the PM10
28 standards, reflecting the far more limited nature of the health effects evidence and air quality
29 data available for thoracic coarse particles.
January 2005
5-2
Draft - Do Not Quote or Cite
-------
1 Staff has taken into account evidence-based considerations primarily by assessing the
2 epidemiologic evidence of associations with health endpoints that the CD has judged to be likely
3 causal based on an integrative synthesis of the entire body of evidence. Less weight is given to
4 evidence of associations that are judged to be only suggestive of possible causal relationships,
5 taking this information into account as part of margin of safety considerations. In so doing, staff
6 has placed greater weight on U.S. and Canadian studies reporting statistically significant
7 associations, providing relatively more precise effects estimates, using relatively more reliable
8 air quality data, and reporting associations that are generally robust to alternative model
9 specifications and the inclusion of potentially confounding co-pollutants. By considering the
10 ambient particle levels present during specific studies, staff has reached conclusions as to the
11 degree to which alternative standards could be expected to protect against the observed health
12 effects, while being mindful of the inherent limitations and uncertainties in such evidence.
13 Staff has also taken into account quantitative risk-based considerations, drawn from.the
14 results of the risk assessment conducted in several example urban areas (discussed in Chapter 4).
15 More specifically, staff has considered estimates of the magnitude of PM-related risks associated
16 with current air quality levels, as well as the risk reductions likely to be associated with attaining
17 the current or alternative standards. In so doing, staff recognizes the considerable uncertainties
18 inherent in such risk estimates, and has taken such uncertainties into account by considering the
19 sensitivity of the risk estimates to alternative assumptions likely to have substantial impact on
20 the estimates.
21 More specifically, in this review a series of questions frames staffs approach to reaching
22 conclusions and recommendations, based on the available evidence and information, as to
23 whether consideration should be given to retaining or revising the current primary PM standards.
24 Staffs review of the adequacy of the current standards begins by considering whether the
25 currently available body of evidence assessed in the CD suggests that revision of any of the basic
26 elements of the standards would be appropriate. This evaluation of the adequacy of the current
27 standards involves addressing questions such as the following:
28 • To what extent does newly available information reinforce or call into question evidence
29 of associations with effects identified in the last review?
January 2005
5-3
Draft - Do Not Quote or Cite
-------
1 • To what extent does newly available information reinforce or call into question any of the
2 basic elements of the current standards?
3 • To what extent have important uncertainties identified in the last review been reduced
4 and have new uncertainties emerged?
5 To the extent that the evidence suggests that revision of the current standards would be
6 appropriate, staff then considers whether the currently available body of evidence supports
7 consideration of standards that are either more or less protective by addressing the following •
8 questions:
9 • Is there evidence that associations, especially likely causal associations, extend to air
10 quality levels that are as low as or lower than had previously been observed, and what are
11 • the important uncertainties associated with that evidence?
12 • Are health risks estimated to occur in areas that meet the current standards; are they
13 important from a public health perspective; and what are the important uncertainties
14 associated with the estimated risks?
15 To the extent that there is support for consideration of revised standards, staff then identifies
16 ranges of standards (in terms of indicators, averaging times, levels and forms) that would reflect
17 a range of alternative public health policy judgments, based on the currently available evidence,
18 as to the degree of protection that is requisite to .protect public health with an adequate margin of
19 safety. In so doing, staff addresses the following questions:
20 • Does the evidence provide support for considering different PM indicators?
21 • Does the evidence provide support for considering different averaging times?
22 • What range of levels and forms of alternative standards is supported by the evidence, and
23 what are the uncertainties and limitations in that evidence?
24 • To what extent do specific levels and forms of alternative standards reduce the estimated
25 risks attributable to PM, and what are the uncertainties in the estimated risk reductions?
26s Based on the evidence, estimated risk reductions, and related uncertainties, staff makes'
27 recommendations as to ranges of alternative standards for the Administrator's consideration in
28 reaching decisions as to whether to retain or revise the primary PM NAAQS. •'
January 2005 5-4 Draft - Do Not Quote or Cite
-------
1 Standards for fine particles are addressed in section 5.3 below, beginning with staffs
2 consideration of the adequacy of the current primary PM25 standards. Subsequent subsections
3 address each of the major elements that define specific PM standards: pollutant indicator,
4 averaging time, level and form. Staff has evaluated separately the protection that a suite of PM25
5 standards would likely provide against effects associated with long-term exposures (section
6 5.3.4) and^those associated with short-term exposures (section 5.3.5). These separate evaluations
7 provide the basis for integrated recommendations on alternative suites of standards that protect
8 against effects associated with both long- and short-term exposures, based on considering how a
9 suite of standards operate together to protect public health. In a similar manner, standards for
10 thoracic coarse particles are addressed in section 5.4 below. This chapter concludes with a
11 summary of key uncertainties associated with establishing primary PM standards and related
12 staff research recommendations in section 5.5.
13 5.3 FINE PARTICLE STANDARDS
14 5.3.1 Adequacy of Current PMZ s Standards
15 In considering the adequacy of the current PM25 standards, staff has first considered the
16 extent to which newly available information reinforces or calls into question evidence of
17 associations with effects identified in the last review, as well as considering the extent to which
18 important uncertainties have been reduced or have resurfaced as being more important than
19 previously understood. In looking across the extensive epidemiologic evidence available in this
20 review, the CD addresses these questions by concluding that "the available findings demonstrate
21 well that human health outcomes are associated with ambient PM" (CD, p. 9-24) and, more
22 specifically, that there is now "strong epidemiological evidence" for PM2 5 linking short-term
23 exposures with cardiovascular and respiratory mortality and morbidity, and long-term exposures
24 with cardiovascular and lung cancer mortality and respiratory morbidity (CD, p. 9-46). This
25 latter conclusion reflects greater strength in the epidemiologic evidence specifically linking
26 PM2 5 and various health endpoints than was observed in the last review, when the CD concluded
27 that the epidemiologic evidence for PM-related effects was "fairly strong," noting that the
January 2005
5-5
Draft - Do Not Quote or Cite
-------
t
1 studies "nonetheless provide ample reason to be concerned" about health effects attributable to
2 PM at levels below the then-current PM NAAQS (EPA, 1996, p. 13-92).
3 As discussed in Chapter 3 (section 3.5) and the CD (section 9.2.2), the CD concludes that
4 the extensive body of epidemiologic evidence now available continues to support likely causal
5 associations between PM25 and the above health outcomes based on an assessment of strength,
6 robustness, and consistency in results. The CD finds "substantial strength" in the evidence of
7 PM23 associations, especially for total and cardiovascular mortality (CD, p. 9-28). The CD
8 recognizes that while the relative risk estimates are generally small in magnitude, a number of
9 new studies provide more precise estimates that are generally positive and often statistically
10 significant.' Overall, the CD firids the new evidence substantiates that the associations are
11 generally robust to confounding by co-pollutants, noting that much progress has been made in
12 sorting out contributions to observed health effects of PM and its components relative to other
13 co-pollutants. On the other hand, the CD notes that effect estimates are generally more sensitive
14 than previously recognized to different modeling strategies to adjust for temporal trends and
15 weather variables. While some studies showed little sensitivity, different modeling strategies
16 . altered conclusions in other studies.
17 Although greater variability in effects estimates across study locations is seen in the
18 much larger set of studies now available, in particular in the new multi-city studies, the CD finds
19 i much consistency in the epidemiologic evidence particularly in studies with the most precision.
20 There also are persuasive reasons why variation in associations in different locations could be
21 expected. Further, the CD concludes that new source apportionment studies and "found
22 experiments," showing improvements in community health resulting from reductions in PM and
23 other air pollutants, lend additional support to the results of other studies that focused
24 specifically on'PM2 5.
25 Looking more broadly to integrate epidemiologic evidence with that from exposure-
26 related, dosimetric and toxicologic studies, the CD (section 9.2.3) considered the coherence of
27 the evidence and-the extent to which the new evidence provides insights into mechanisms by
28 which PM, especially fine particles, may be affecting human health. Progress made in gaining
29 insights into mechanisms lends support to the biological plausibility of results observed in
January 2005 5 -6 Draft - Do Not Quote or Cite
-------
1 epidemiologic studies. For cardiovascular effects, the CD finds that the convergence of
2 important new epidemiologic and toxicologic evidence builds support for the plausibility of
3 associations especially between fine particles and physiological endpoints indicative of increased
4 risk of cardiovascular disease and changes in cardiac rhythm. This finding is supported by new
5 cardiovascular effects research focused on fine particles that has notably advanced our
6 understanding of potential mechanisms by which PM exposure, especially in susceptible
7 individuals, could result in changes in cardiac function or blood characteristics that are risk
8 factors for cardiovascular disease. For respiratory effects, the CD finds that toxicologic studies
9 have provided evidence that supports plausible biological pathways for fine particles, including
10 inflammatory responses, increased airway responsiveness, or altered responses to infectious
11 , agents. Further, the CD finds coherence across a broad range of cardiovascular and respiratory
12 , health outcomes from epidemiologic and toxicologic studies done in the same location,
13 particularly noting, for example, the series of studies conducted in or evaluating ambient PM
14 from Boston and the Utah Valley. The CD also finds that toxicologic evidence examining
15 combustion-related particles supports the plausibility of the observed relationship between fine
16 particles and lung cancer mortality. With regard to PM-related infant mortality and
17 developmental effects, the CD finds this to be an emerging area of concern, but notes that current
18 information is still very limited in support of the plausibility of potential ambient PM
19 relationships.
20 Based on the above considerations and findings from the CD, staff concludes that the
21 newly available information generally reinforces the associations between PM2 5 and mortality
22 and morbidity effects observed in the last review. Staff recognizes that important uncertainties
23 and research questions remain, notably including questions regarding modeling strategies to
24 adjust for temporal trends and weather variables in time-series epidemiologic studies.
25 Nonetheless, staff notes mat progress has been made in reducing some key uncertainties since
26 the last review, including important progress in advancing our understanding of potential
27 mechanisms by which ambient PM2 5, alone and in combination with other pollutants, is causally
28 linked with cardiovascular, respiratory, and lung cancer associations observed in epidemiologic
January 2005 5-7 . Draft - Do Not Quote or Cite
t
-------
1 studies. Thus, staff finds clear support in the available evidence, as assessed in the CD, for fine
2 particle standards that are at least as protective as the current PM2 5 standards.
3 Having reached this initial conclusion, staff also has addressed the question of whether
4 the available evidence supports consideration of standards that are more protective than the
5 current PM25 standards. In so doing, staff has considered first whether there is evidence that
6 health effects associations with short- and long-term exposures to fine particles extend to lower
7 air quality levels than had previously been observed, or to levels below the current standards. In
8 addressing this question, staff first notes that the available evidence does not either support or
9 refute the existence of thresholds for the effects of PM on mortality across the range of
10 concentrations in the studies, as discussed in Chapter 3 (section 3.4.6) and the CD (section
11 9.2.2.5). More specifically, while there are likely threshold levels for individuals and specific
12 health responses, existing studies show little evidence for thresholds for PM-mortality
13 relationships in populations, for either long-term or short-term PM exposures (CD, p. 9-44).
14 Further, the CD notes that in the multi-city and most single-city studies, statistical tests
15 comparing linear and various nonlinear or threshold models have not shown statistically
16 significant distinctions between them (CD, p. 9-44). Even in those few studies with suggestive
17 evidence for thresholds, the potential thresholds are at fairly low concentrations (CD, p. 9-45).
18 While acknowledging that for some health endpoints, such as total nonaccidental mortality, it is
19 likely to be extremely difficult to detect thresholds, the CD concludes that "epidemiblogic
20 studies suggest no evidence for clear thresholds in PM-mortality relationships within the range
21 of ambient PM concentrations observed in these studies." (CD, p. 9-48).
22 In considering the available epidemiologic evidence (summarized in Chapter 3, section
23 3.3 and Appendices 3 A and 3B), staff has focused on specific epidemiologic studies that show
24 statistically significant associations between PMZ5 and health effects for which the CD judges
25 associations with PM25 to be likely causal. Many more U.S. and Canadian studies are now
26 available in the current review that provide evidence of associations between PM?5 and serious
27 health effects in areas with air quality at and above the level of the current annual PM25 standard
28 (15 ug/m3), which was set to provide protection against health effects related to both short- and
29 long-term exposures to fine particles. Notably, a few of the newly available short-term exposure
January 2005 5-8 . Draft - Do Not Quote or Cite
-------
1 mortality studies provide evidence of statistically significant associations with PM25 in areas
2 with long-term average air quality below that ambient level (summarized in Appendix 3A). In
3 considering this group of studies, staff has focused on those studies that include adequate
4 gravimetric PM2 5.mass measurements, and where the associations are generally robust to
5 alternative model specification and to the inclusion of potentially confounding co-pollutants.
6 Three such studies conducted in Phoenix (Mar et al., 1999,2003), Santa Clara County, CA
7 (Fairley, 1999,2003) and eight Canadian cities (Burnett et al., 2000 and Burnett and Goldberg,
8 2003) report statistically significant associations between short-term PM2 5 exposure and total
9 and cardiovascular mortality in areas in which long-term average PM2 5 concentrations ranged
10 between 13 and 14 ug/m3. These studies were reanalyzed to address questions about the use of
11 GAM with default convergence criteria, and the study results from Phoenix and Santa Clara
12 County were little changed in alternative models (Mar et al., 2003; Fairley, 2003), although
13 Burnett and Goldberg (2003) reported that their results were sensitive to using different temporal
14 smoothing methods.
15 Beyond .these mortality studies, other studies provide evidence of statistically significant
16 associations with morbidity. Three studies of emergency department visits were conducted in
17 areas where the mean PM25 concentrations were approximately 12 ng/m3 or below, although
18 these studies either had not been reanalyzed to address the default convergence criteria problem
19 with GAM, did not assess the potential for confounding by co-pollutants or were not robust to
20 the inclusion of co-pollutants, or were done only during a single season. Another new study
21 reported statistically significant associations with incidence of myocardial infarction where the
22 mean PM2.5 concentration was just above 12 ug/m3; however, the CD urges caution in
23 interpreting the results of the new body of evidence related to such cardiovascular effects (CD, p.
24 8-166). Thus, these studies provide no clear evidence of statistically significant associations
25 with PM2 5 at such low concentrations.
26 New evidence is also available from U.S. and Canadian studies of long-term exposure to
27 fine particles (summarized in Appendix 3B). In evaluating this evidence (CD, section 9.2.3), the
28 CD notes that new studies have built upon studies available in the last review and these studies
29 have confirmed and strengthened the evidence of associations for both mortality and respiratory
January 2005 5 -9 Draft - Do Not Quote or Cite
-------
1 morbidity. For mortality, the CD places greatest weight on the reanalyses and extensions of the
2 Six Cities and the ACS studies, finding that these studies provide "strong evidence" for
3 associations with fine particles (CD, p. 9-34), notwithstanding the lack of consistent results in
4 other long-term exposure studies. For morbidity, the CD finds that new studies of a cohort of
5 children in Southern California have built upon earlier limited evidence to provide "fairly
6 strong" evidence that long-term exposure to fine particles is associated with development of
7 chronic respiratory disease and reduced lung function growth (CD, p. 9-34). ;
8 As discussed in the CD and in Chapter 3 above, mortality studies of the Six Cities and
9 ACS cohorts available in the last review had aggregate long-term mean PM2 5 concentrations of
10 18 ug/m3 (ranging from approximately 11 to 30 ug/m3 across cities) and 21 ug/m3 (ranging from
11 approximately 9 to 34 ug/m3 across cities), respectively. Reanalyses of data from these cohorts
12 continued to report significant associations with PM25, using essentially the same air quality
13 distributions. The extended analyses using the ACS cohort also continued to report statistically
14 significant associations with PM2 5 with the inclusion of more recent PM2_s air quality data, with
15 an average range across the old and new time periods from about 7.5 to 30 ug/m3 (from figure 1,
16 Pope et al., 2002) with a long-term mean of approximately 17.7 ug/m3 (Pope et al., 2002). As
17 with the earlier cohort studies, no evidence of a threshold was observed in the relationships with
18 total, cardiovascular, and lung cancer mortality reported in this extended study. In the morbidity
19 studies of the Southern California children's cohort, the means of 2-week average PM25
20 concentrations ranged from approximately 7 to 32 ug/m3, with an across-city average of
21 approximately 15 ug/m3 (Peters et al., 1999). Staff notes that in figures depicting relationships
22 between lung function growth and average PM concentration, there is no evidence of a threshold
23 in this study (Gauderman et al., 2000,2002).
24 Beyond the epidemiologic studies using PM2 5 as an indicator of fine particles, a large
25 body of newly available evidence from studies that used PM10, as well as other indicators or
26 components of fine particles (e.g.,' sulfates, combustion-related components), provides additional
27 support for the conclusions reached in the last review as to the likely causal role of ambient PM,
28 and the likely importance of fine particles in contributing to observed health effects. Such
29 studies notably include new multi-city studies, intervention studies (that relate reductions in
January 2005 5-10 Draft - Do Not Quote or Cite
-------
1 ambient PM to observed improvements in respiratory or cardiovascular health), and source-
2 oriented studies (e.g., suggesting associations with combustion- and vehicle-related sources of
3 fine particles). Further, the CD concludes that new epidemiologic studies of ambient PM
4 associations with potential PM-related infant mortality and/or developmental effects are very
5 limited, although if further substantiated by future research, would significantly increase
6 estimates of the extent of life shortening due to PM-related premature mortality (CD, p, 9-94).
7 The CD also notes that new epidemiologic studies of asthma-related increased physicians visits
8 and symptoms, as well as new studies of cardiac-related risk factors, suggest likely much larger
9 public health impacts due to ambient fine particles than just those indexed by the
10 mortality/morbidity effects considered in the last review (CD, p. 9-94).
11 Staff recognizes, however, mat important limitations and uncertainties associated with
12 this expanded body of evidence for PM2 5 and other indicators or components of fine particles, as
13 discussed in Chapter 3 herein and section 9.2.2 of the CD, need to be carefully considered in
14 determining the weight to be placed on the studies available in this review. For example, the CD
15 notes that while PM-effects associations continue to be observed across most new studies, the
16 newer findings do not fully resolve the extent to which the associations are properly attributed to
17 PM acting alone or in combination with other gaseous co-pollutants, or to the gaseous co-
18 pollutants themselves. The CD notes that available statistical methods for assessing potential
19 confounding by gaseous co-pollutants may not yet be fully adequate, although the various
20 approaches that have now been used to evaluate this issue tend to substantiate that associations
21 for various PM indicators with mortality and morbidity are robust to confounding by co-
22 pollutants (CD, p. 9-37).
23 Another issue of particular importance is the sensitivity of various statistical models to
24 the approach used to address potential confounding by weather- and time-related variables in
25 time-series epidemiological studies. As discussed in section 3.5.3 herein and in section 9.2.2 of
26 the CD, this issue resurfaced in the course of reanalyses of a number of the newer studies that
27 were being conducted to address a more narrow issue related to problems associated with the use
28 of commonly used statistical software. These reanalyses suggest that weather continues to be a
29 potential confounder of concern and highlight that no one model is likely to be most appropriate
January 2005 .. 5-11 Draft - Do Not Quote or Cite
-------
1 in all cases. The HEI Review Panel, in reviewing these reanalyses, concluded that this
2 awareness introduces a degree of uncertainty in evaluating the findings from time-series
3 epidemiological studies that had heretofore not been widely appreciated.
4 ' In looking beyond PM mass indicators, a number of newly available studies highlight the
5 issue of the extent to which observed health effects may be associated with various specific
6 chemical components within the mix of fine particles. The potential for various fine particle
7 components to have differing relative toxicities with regard to the various health endpoints being
8 considered adds complexity to the interpretation of the study results. The CD recognizes that
9 more research is needed to address uncertainties about the extent to which various components
10 may be relatively more or less toxic than'others, or than undifferentiated PM2 5 mass across the
11 range of health endpoints studied.
12 While the limitations and uncertainties in the available evidence suggest caution in
13 interpreting the epidemiologic studies at the lower levels'of air quality observed in the studies,
\
14 staff concludes that the evidence now available provides strong support for considering fine
15 particle standards that would provide increased protection from that afforded by. the current
* " . .
16 PM2 5 standards. More protective standards would reflect the generally stronger and broader .
17 body of evidence of associations with mortality and morbidity now available in this review, at
18 lower levels of air quality and at levels below the current standards, and with more
19 understanding of possible underlying mechanisms.
20 In addition to this evidence-based evaluation, staff has also considered the extent to
21 which health risks estimated to occur upon attainment of the current PM25 standards may be
22 judged to be important from a public health perspective, taking into account key uncertainties
23 associated with the estimated risks. Based on the risk assessment presented in Chapter 4, staff
24 considered as a base case the estimated risks attributable to PM2 5 concentrations above
25 background levels, or above the lowest measured levels in a given study if that was higher than
26 background, so as to avoid extrapolating risk estimates beyond the range of air quality upon
27 which the concentration-response functions were based. In the case of estimated risk associated
28 with long-term exposure, based on the extended ACS study, risk was estimated down to an
29 annual level of 7.5 ug/m3, the lowest measured level in that study; for estimated risk associated
January 2005 5-12 Draft - Do Not Quote or Cite
-------
1 with short-term exposure, risk was estimated down to daily levels ranging from 2.5 to 4 ug/m3,
2 based on estimated background or the lowest measured level in a particular study.
3 In the absence of evidence for clear thresholds in any of the studies used in this risk
4 assessment, the base case estimates in this analysis reflect the linear or near-linear concentration-
5 response functions reported in the studies. To reflect the uncertainty as to whether thresholds
6 may exist within the range of air quality observed in the studies, but may not be discernable with
7 currently applied statistical methods, staff also has considered estimates of risk based on
8 concentration-response functions modified to incorporate various assumed hypothetical
9 threshold levels, as discussed in Chapter 4. Based on the sensitivity analyses conducted as part
10 of the risk assessment, the uncertainly associated with alternative hypothetical thresholds had by
11 • far the greatest impact on estimated risks. Other uncertainties have a more moderate and often
12 variable impact on the risk estimates in some or all of the cities, including the use of single-
13 versus multi-pollutant models, single- versus multi-city models, use of a distributed lag model,
14 alternative assumptions about the relevant air quality for long-term exposure mortality, and
15 alternative constant or varying background levels. v
16 - Table 5-1 summarizes the estimated PM25-related annual incidence of total mortality
17 associated with long- and short-term exposure for the base case and for alternative hypothetical
18 thresholds in the nine example urban areas included in the risk assessment. In looking
19 particularly at the annual incidence of PM2 5-related mortality estimated to occur upon attainment
20 of the current PM25 standards in the five study areas that do not meet the current standards based
21 on 2001 -2003 air quality data, staff notes that there is a fairly wide range of estimated incidence
22 across the areas. Such variation would be expected considering, for example, differences in total
23 population, demographics, exposure considerations (e.g., degree of air conditioning use),
24 presence of co-pollutants and other environmental stressors, and exposure measurement error
25 across urban areas; as well as differences in concentration-response relationships across studies
26 that might be due in part to variation in these factors across locations. Staff also recognizes that
27 mere are uncertainties associated with the procedure used to simulate air quality that would just
28 attain the current standards and in the degree to which various components of the fine particle
29 mix would likely be reduced in similar proportion to the simulated reduction in PM25 as a whole.
January 2005 5-13 Draft - Do Not Quote or Cite
-------
X
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Table 5-1 Estimated PM2 s-related Annual Incidence of Total Mortality when Current
PM2 5 Standards are Met (Base Case and Assumed Alternative Hypothetical
Thresholds)*
Short4erm Exposure:
Annual Non-Accidental Mortality
(except as noted)
Base case
Estimate,
95% Cl
Assumed Hypothetical Short-term
Exposure Thresholds
10(jg/m3
15pg/m*
20 jjg/m3
Long-term Exposure:
Annual All-Cause Mortality
Base case
Estimate,
95% Cl
Assumed Hypothetical
Long-term Exposure
Thresholds
10 ug/m3
12 |jg/m3
Risks associated with just meeting current PMts standards
Detroit
Los Angeles
Philadelphia
{short-term: cardiovascular
mortality)
Pittsburgh .
(short-term: over age 74)
St. Louis
115
-116 to 338
248
-31 to 51 9
367
175 to 560
50
-108 to 200
191
70 to 311
54
-55 to 159
115
"-14 to 240
189
90 to 288
. 22
-48 to 87
75
28 to 122
26
-27 to 77
58
-7 to 121
106 -
51 to 162
10
-23 to 41
29
11 to46
12
-12 to 35
29
-4 to 61
57
27 to 87
5
-11 to 18
9
3 to 14
522
181 to 910
1,507 .
531 to 2,587
536
185 to 943
403
141 to 699
596
206 to 1,047
282
98 to 494
823
290 to 1415
' 338
116 to 597
215
75 to 373
311
107 to 548
41
14 to 72
138
48 to 237
137
47 to 244
25
9 to 43
23
8 to 40
Risks associated with "as is "air quality (in areas that moot current PM2S standards)
Boston
Phoenix
(short-term: cardiovascular
mortality over age 64)
San.Jose
Seattle"
390
265 to 514
323
97 to 536
218
45 to 387
-
173
118 to 228
115
35 to 190
80
17 to 141
82
56 to 109
67
21 to 109
44
9 to 77
41
28 to 53
43
13 to 69
28
6 to 50
594
204 to 1053
349
119 to 620
172
59 to 306
50
17 to 89
309
106 to 551
76
26 to 136
58
20 to 104
0
OtoO
20
7to36
0
OtoO
0
OtoO •
0
OtoO
* These estimates of annual incidence of PM2.5-related mortality are based on using the maximum monitor in an area
to calculate the percent rollback needed to just attain the current PM^s annual standard, and applying that percent
rollback to the composite monitor in the area, as described in Chapter 4, section 4.2.3. Estimates of annual mortality
incidence based on using a spatially averaged concentration to calculate the percent rollback needed to just attain the
current standard, where this is allowed, would be higher than the estimates shown here.
** No short-term exposure concentration-response function is available for mortality in Seattle.
26 Staff observes that base case point estimates of annual incidence of total PMZ5-related
27 mortality associated with just meeting the current PM25 standards in the five areas shown range
January 2005
5-14
Draft - Do Not Quote or Cite
-------
1 from approximately 400 to 600 in four areas (or from roughly 25 to 35 deaths per 100,000
2 population in these areas) to over 1500 annual deaths in Los Angeles (i.e., roughly 16 deaths per
3 100,000 population) associated with long-term exposure. These estimated incidences associated
4 with long-term exposure represent 2.6 to 3.2 percent of total mortality incidence due to all
5 causes. Expressing the risk estimates in terms of percentage of total incidence takes into
6 account city-to-city differences in population size and baseline mortality incidence rate. In some
7 areas, the 95% confidence ranges associated with the estimates of total annual mortality
8 incidence related to short-term exposure (but not long-term exposure) extend to below zero,
9 reflecting appreciably more uncertainty in estimates based on positive but not statistically
10 significant associations. In the other four areas mat meet the current standards based on recent
11 air quality data, base case point estimates of annual incidence of total PM2 5-related mortality
12 associated with long-term exposure range from a lower end of about 50 deaths in Seattle (which
13 represents a rate of about 3 per 100,000 population) to an upper end of almost 600 deaths in
14 Boston (a rate of 21 per 100,000 population). It is much more difficult to make comparisons
15 among the urban areas with regard to short-term exposure mortality incidence or incidence rates
16 because of the different population groups and mortality types examined in the epidemiology
17 studies'for the different locations. There also is greater variability in the estimates for mortality
18 associated with short-term exposure due to the use of different city-specific concentration-
19 response relationships.
20 In looking beyond the base case estimates, staff also considered the extent to which the
21 assumption of the presence of hypothetical thresholds in the concentration-response relationships
22 would influence the risk estimates. As expected, risk estimates are substantially smaller when
23 hypothetical threshold concentration-response functions are considered. Point estimates of
24 annual incidence of total PM25-related mortality associated with long-term exposure are roughly
25 50% of base case estimates when a hypothetical threshold of 10 ug/m3 is assumed, whereas when
26 a hypothetical threshold of 12 ug/m3 is assumed, point estimates are roughly 5 to 20% of base
27 case estimates in nonattainment areas (and even smaller in attainment areas). A similar pattern is
28 seen when considering the impact of alternative hypothetical thresholds in the range of 10 to 20
29 ng/m3 on risks associated with short-term exposure.
January 2005 5-15 , Draft-DoNot Quote or Cite
-------
1 In considering these estimates of PM2 5-related incidence of annual total mortality upon
2 . meeting the current standards in a number of example urban areas, together with the
3 uncertainties in these estimates, staff concludes that they are indicative of risks that can
4 reasonably be judged to be important from a public health perspective and provide support for
5 consideration of standards mat would provide increased protection from that afforded by the
6 current PM2 5 standards. In the absence of evidence of clear thresholds, staff believes that it is
7 appropriate to give most weight to the base case risk estimates. These estimates indicate the
8 likelihood of thousands of premature deaths per year in urban areas across the U.S. Beyond the
9 estimated incidences of mortality discussed above, staff also recognizes that similarly substantial
10 numbers of incidences of hospital admissions, emergency room visits, aggravation of asthma and
11 other respiratory symptoms, and increased cardiac-related risk are also likely in many urban
12 areas, based on risk assessment results presented in Chapter 4 and on the discussion related to •
13 the pyramid of effects drawn from section 9.2.5 of the CD. Staff also believes that it is important
14 to recognize how highly dependent these risk estimates are on the shape of the underlying
15 concentration-response functions. In so doing, staff nonetheless notes that in considering even
16 the largest assumed hypothetical thresholds, estimated mortality risks are not completely
17 eliminated when current PM2S standards are met in a number of example urban areas, including
18 all such areas that do not meet the standards based on recent air quality.
19 Staff well recognizes that as the body of available evidence has expanded, it has added
20 greatly both to our knowledge of PM-related effects, as well as to the complexity inherent in
21 interpreting the evidence in a policy-relevant context as a basis for setting appropriate standards.
22 In considering available evidence, risk estimates, and related limitations and uncertainties, staff
23 concludes that the available information clearly calls into question the adequacy of the current
24 suite of PM2 5 standards', and provides strong support for giving consideration to revising the
25 current PM2 5 standards to provide increased public health protection. Staff conclusions and
26 recommendations for indicators, averaging times, and levels and forms of alternative, more
27 protective primary standards for fine particles are discussed in the following sections.
January 2005' 5-16 Draft - Do Not Quote or Cite
-------
t
1 5.3.2 Indicators
2 " In 1997, EPA established PM25 as the indicator for fine particles. In reaching this
3 decision, the Agency first considered whether the indicator should be based on the mass of a
4 size-differentiated sample of fine particles or on one or more components within the mix of fine
5 particles. Secondly, in establishing a size-based indicator, a size cut point needed to be selected
6 that would appropriately distinguish fine particles from particles in the coarse mode.
7 In addressing the first question in the last review, EPA determined that it was more
8 appropriate to control fine particles as a group, as opposed to singling out any particular
9 component or class of fine particles based on the following considerations. Community health
10 studies had found significant associations between various indicators of fine particles (including
11 PM25 or PM10 in areas dominated by fine particles) and health effects in areas with significant
12 mass contributions of differing components or sources of fine particles, including sulfates, wood
13 smoke, nitrates, secondary organic compounds and acid sulfate aerosols. In addition, a number
14 of animal toxicologic and controlled human exposure studies had reported health effects
15 associations with high concentrations of numerous fine particle components (e.g., sulfates,
16 nitrates, transition metals, organic compounds), although such associations were not consistently
17 observed. It also was not possible to rule out any component within the mix of fine particles as
18 not contributing to the fine particle effects found in epidemiologic studies. Thus, it was
19 determined that total mass of fine particles was the most appropriate indicator for fine particle
20 standards rather than an indicator based on PM composition (62 FR 38667, July 18,1997).
21 Having selected a size-based indicator for fine particles, the Agency then based its
22 selection of a specific cut point on a number of considerations. In focusing on a cut point within
23 the size range of 1 to 3 \im (i.e., the intermodal range between fine and coarse mode particles),
24 EPA recognized that the choice of any specific sampling cut point within this range was largely a
25 policy judgment. In making this judgment, the Agency noted that the available epidemiologic
26 studies of fine particles were based largely on PM25; only very limited use of PM] monitors had
27 been made. While it was recognized that using PMj as an indicator of fine particles would
28 exclude the tail of the coarse mode in some locations, in other locations it would miss a portion
29 of the fine PM, especially under high humidity conditions, which would result in falsely low fine
January 2005
5-17
Draft - Do Not Quote or Cite
-------
1 PM measurements on days with some of the highest fine PM concentrations. The selection of a
2 2.5 um cut point reflected the regulatory importance that was placed on defining an indicator for
3 fine particle standards that would more completely capture fine particles under all conditions
4 likely to be encountered across the U.S., especially when fine particle concentrations are likely
5 to be high, while recognizing that some small coarse particles would also be captured by PM2 5
6 monitoring.2 Thus, EPA's selection of 2.5 um as the cut point for the fine particle indicator was
7 based on considerations of consistency with the epidemiologic studies, the regulatory importance
8 of more completely capturing fine particles under all conditions, and the limited potential for
9 intrusion of coarse particles in some areas; it also took into account the general availability of
10 monitoring technology (62 FR 3 8668).
11 In this current review, staff observes that the same considerations apply for selection of
12 an appropriate indicator for fine particles. As an initial matter, staff notes that the available
13 epidemiologic studies linking mortality and morbidity effects with short- and long-term
14 exposures to fine particles continue to be largely indexed by PM2 5. Some epidemiologic studies
15 also have continued to implicate various PM components (e.g., sulfates, nitrates, carbon, organic
16 compounds, and metals) as being associated with adverse effects; effects have been reported
17 with a broad range of PM components, as summarized in Table 9-13 of the CD (p. 9-31).
18 Animal toxicologic and controlled human exposure studies, evaluated in Chapter 7 of the CD,
19 have continued to link a variety of PM components or particle types (e.g., sulfates or acid
20 aerosols, metals, organic constituents, bioaerosols, diesei particles) with health effects, though
21 often at high concentrations (CD section 7.10.2). In addition, some recent studies have
22 suggested that the ultrafine subset of fine particles may also be associated with adverse effects
23 (CD, pp. 8-66,8-199). . '
24 . Staff recognizes that, for a given health response, some PM components are likely to be
25 more closely linked with that response than others (CD, p. 9-30). That different PM constituents
26 may have differing biological responses is an important source of uncertainty in interpreting
27 epidemiologic evidence. For specific effects there may be stronger correlation with individual
2 In reaching this decision, EPA indicated that it might be appropriate to address undue intrusion of coarse
mode particles resulting in violations of PM2 5 standards in the context of policies established to implement such
standards (62 FR 38668).
January 2005 5-18 Draft - Do Not Quote or Cite
-------
1 PM components than with particle mass. For example, in some toxicologic studies of
2 cardiovascular effects, such as changes in heart rate, electrocardiogram measures, or increases in
3 arrhythmia, PM exposures of equal mass did not produce the same effects, indicating that PM
4 composition was important (CD, p. 7-30). In addition, section 9.2.3.1.3 of the CD indicates that
5 particles, or particle-bound water, can act as carriers to deliver other toxic agents into the
6 respiratory tract, highlighting the fact that exposure to particles may elicit effects that are linked
7 with a mixture of components more than with any individual PM component
8 Thus, epidemiologic and toxicologic studies summarized above and discussed in the CD
9 have provided evidence for effects associated with various fine particle components or size-
10 differentiated subsets of fine particles. The CD concludes: "These studies suggest that many
11 different chemical components of fine particles and a variety of different types of source
12 categories are all associated with, and probably contribute to, mortality, either independently or
13 in combinations" (CD, p. 9-31). Conversely, the CD provides no basis to conclude that any
14 individual fine particle component cannot be associated with adverse health effects. There is no
15 evidence that would lead toward the selection of one or more PM components as being primarily
16 responsible for effects associated with fine particles, nor is there any component that can be
17 eliminated from consideration. Staff continues to recognize the importance of an indicator that
18 not only captures all of the most harmful components of fine PM (i.e., an effective indicator), but
19 also places greater emphasis for control on those constituents or fractions, including most
20 sulfates, acids, transition metals, organics, and ultrafine particles, that are most likely to result in
21 the largest risk reduction (i.e., an efficient indicator). Taking into account the above
22 considerations, staff concludes that it remains appropriate to control fine particles as a group;
23 i.e., that total mass of fine particles is the most appropriate indicator for fine particle standards.
24 With regard to an appropriate cut point for a size-based indicator of total fine particle
25 mass, the CD most generally concludes that advances in our understanding of the characteristics
26 of fine particles continue to support the use of particle size as an appropriate basis for
27 distinguishing between these subclasses, and that a nominal cut point of 2.5 um remains
28 appropriate (CD, p. 9-22). This conclusion follows from a recognition that within the intermodal
29 range of 1 to. 3 jim there is no unambiguous definition of an appropriate cut point for the
January 2005 5-19 Draft - Do Not Quote or Cite
-------
1 separation of the overlapping fine and coarse particle modes (CD, p. 9-8). Within this range,
2 staff considered cut points of both 1 jirri and 2.5 jim. Consideration of these two cut points took
3 into account that there is generally very little mass in this intermodal range, although in some
4 circumstances (e.g., windy, dusty areas) the coarse mode can extend down to and below 1 um,
5 whereas in other circumstances (e.g., high humidity conditions, usually associated with very high
6 fine particle concentrations) the fine mode can extend up to and above 2.5 um The same
7 considerations that led to the selection of a 2.5 \im cut point in the last review - that the
8 epidemiologic evidence was largely based on PM2 5 and that it was more important from a
9 regulatory perspective to more completely capture fine particles under all conditions likely to be
10 encountered across the U.S. (especially when fine particle concentrations are likely to behigh)
11 than to avoid some coarse-mode intrusion into the fine fraction in some areas - also lead to the
12 same conclusion in this review. In addition, section 9.2.1.2.3. of the CD discusses the potential
13 health significance of particles as carriers of water, oxidative compounds, and other components
14 into the respiratory system. This consideration adds to the importance of ensuring that larger
15 accumulation-mode particles are included in the fine particle size cut. Therefore, as observed
i\ 16 previously in section 3.1.2, the scientific evidence leads the CD to conclude that 2.5 uni remains
17 an appropriate upper cut point for a fine particle mass indicator.
18 . Thus, consistent with the CD's conclusion that 2.5 um remains an appropriate cut point
i
19 for including the larger accumulation-mode fine particles while limiting intrusion of coarse
20 particles, staff recommends that PM2 5 be retained as the indicator for fine particles. Staff further
21 concludes that currently available studies do not provide a sufficient basis for supplementing
22 mass-based fine particle standards with standards for any specific fine particle component or
23 subset of fine particles, or for eliminating any individual component or subset of components
24 from fine particle mass standards.
* i
25 Further, staff notes that since the' last review an extensive PM2 5 monitoring network has
26 been deployed and operated in cooperative efforts with State, local and Tribal agencies and with
27 instrument manufacturers. At the same time, EPA has been working on the development of
28 strategies and programs to implement the 1997 PM2.5 standards, based on the Federal Reference
29 Monitor (FRM).for PM2 5. The new monitoring network has provided substantial new air quality
January 2005 5-20 Draft - Do Not Quote or Cite
-------
1 potentially be used to provide information to the public based on episodic very short-term peak
2 fine particle levels that may be of public health concern.
3 In the last review, having decided to set both annual and 24-hour PM2 5 standards, EPA
4 also made judgments as to the most effective and efficient approach to establishing a suite of
5 standards that, taken together, would appropriately protect against effects associated with both
6 long- and short-term exposures. At that time, EPA selected an approach that was based on
7 treating the annual standard as the generally controlling standard for lowering the entire
8 distribution of PM2 5 concentrations, with the 24-hour standard providing additional protection
9 against the occurrence of peak 24-hour concentrations. The 24-hour standard was intended to
10 address in particular those peaks that result in localized or seasonal exposures of concern in areas
11 where the highest 24-hour-to-annual mean PM25 ratios are appreciably above the national
12 average. This approach was supported by results of the PM risk assessment from the last review
13 which indicated that peak 24-hour PM2 5 concentrations contribute a relatively small amount to
14 total health risk, such that much if not most of the aggregated annual risk associated with short-
15 term exposures results from the large number of days during which the 24-hour average •
16 ' concentrations are in the low- to mid-range. Further, no evidence suggested that risks associated
17 with long-term exposures are likely to be disproportionately driven by peak 24-hour
18 concentrations. Thus, a generally controlling annual standard was judged to reduce risks
19 associated with both short- and long-term exposures effectively and with more certainty than a
20 24-hour standard. Further, an annual standard was seen to be more stable over time, likely
21 resulting in the development of more consistent risk reduction strategies, since an area's
22 attainment status would be less likely to change due solely to year-to-year variations in
23 meteorological conditions that affect the atmospheric formation of fine particles.
24 In this review, staff recognizes that some key considerations that led to establishing a
25 generally controlling annual standard in the last review are still valid. In particular, staff
26 observes that:
27 • EPA's updated risk assessment supports the conclusion that peak 24-hour PM2 5
28 concentrations contribute a relatively small amount to the total health risk associated with
29 short-term exposures on an annual basis, such that much if not most of the aggregated
30 annual risk results from the large number of days during which the 24-hour average
31 concentrations are in the low- to mid-range, as discussed in Chapter 4 (section 4.3.3).
January 2005 , 5-23 Draft - Do Not Quote or Cite
-------
1 Support for this conclusion is also found in studies in which health effect associations
2 remain when high-concentration days are removed from the analysis (Schwartz et al,
3 " 1996; Ostroetal., 1999,2000).
4 • It continues to be the case, as discussed in section 4.2.6.1, that available short-term
5 exposure studies do not provide evidence of clear population thresholds, but rather reflect
6 relationships between health effects and ambient PM across a wide distribution of PM
7 concentrations. Thus, as in the last review, staff recognizes that these studies do not
8 provide a basis for identifying a lowest-observed-effect level that would clearly translate
9 into a 24-hour standard that would protect against all effects related to short-term
10 exposures.
11 Nonetheless, staff believes that the greatly expanded body of epidemiologic evidence and
12 air quality data provide the basis for considering alternative approaches to establishing a suite of
13 PM25 standards. Thus, staff has not focused a priori on an annual standard as the generally
14 controlling standard for protection against effects associated with both long- and short-term
15 exposures. Rather, staff has broadened its view to consider botii evidence-based and risk-based
16 approaches to evaluating the protection that a suite of PM2 5 standards can provide against effects
17 associated with long-term exposures arid against short-term exposures. These evaluations,
18 discussed in the next two sections, provide the basis for integrated recommendations on ranges
19 of alternative suites of standards that, when considered together, protect against effects
20 associated with both long- and short-term exposures.
21 5.3.4 Alternative PM2.5 Standards to Address Health Effects Related to Long-term
22 Exposure
23 In considering alternative PMj 5 standards that would provide protection against health
24 effects related to long-term exposures, staff has taken into account both evidence-based and risk-
25 based considerations. As discussed below in this section, staff has first evaluated the available
26 . evidence from long-term exposure studies, as well as the uncertainties and limitations in that
27 evidence, to assess the degree to which alternative annual PM2 5 standards can be expected to
28 provide protection against effects related to long-term exposures. Secondly, staff has considered
29 the quantitative risk estimates for long-term exposure effects, discussed in Chapter 4, to assess
30 the extent to which alternative annual and/or 24-hour standards can be expected to reduce the
January 2005
5-24
Draft - Do Not Quote or Cite
-------
1 estimated risks attributable to long-term-exposure to PM25. Staff conclusions as to ranges of
2 alternative annual and/or 24-hour standards that would provide protection against health effects
3 related to long-term exposures are summarized at the end of this section. The integrated staff
4 recommendations presented in section 5.3.7 are based in part on the conclusions from this
5 section and in part on staff conclusions from the next section, in which alternative PM2 5
6 standards to address health effects related to short-term exposures are assessed.
7 5.3.4.1 Evidence-based Considerations
8 In taking into account evidence-based considerations, staff has focused on long-term
9 exposure studies of fine particles in the U.S. As discussed above, staff notes that the reanalyses
10 and extensions of earlier studies have confirmed and strengthened the evidence of long-term
11 associations for both mortality and morbidity effects. The assessment in the CD of these
12 mortality studies, taking into account study design, the strength of the study (in terms of
13 statistical significance and precision of result), and the consistency and robustness of results,
14 concluded that it was appropriate to give the greatest weight to the reanalyses of the Six Cities
15 study and the ACS study, and in particular to the results of the extended ACS study (CD, p.
16 9-33). The assessment in the CD of the relevant morbidity studies noted in particular the results
17 of the new studies of the children's cohort in Southern California as providing evidence of
18 respiratory morbidity with long-term PM exposures.
19 Staff believes it is appropriate to consider a level for an annual PM2} standard that is
20 somewhat below the averages of the long-term concentrations across the cities in each of these
21 studies, recognizing that the evidence of an association in any such study is strongest at and
22 around the long-term average where the data in the study are most concentrated. For example,
23 the interquartile range of long-term average concentrations within a study, or a range within one
24 standard deviation around the study mean, might be used to characterize the range over which
25 the evidence of association is strongest. Staff also believes it is appropriate to consider the long-
26 term average concentration at the point where the confidence interval becomes notably wider,
27 suggestive of a concentration below which the association becomes appreciably more uncertain
28 and the possibility that an effects threshold may exist becomes more likely. Staff further notes
29 that in considering a level for a standard that is to provide protection with an adequate margin of
January 2005
5-25
Draft ~ Do Not Quote or Cite
-------
1 safety, it is appropriate to take into account evidence of effects' for which'the reported ^ !-
2 associations provide only suggestive evidence of a potentially causal association. " -"•'
3 In looking first at the long:term exposure mortality studies, staff notes that the long-term
4 mean PM2 5 concentration in the Six Cities study was 18 ug/m3, within an overall range of 11 to
5 . 30 |ig/m3. In'the studies using the ACS cohort, the long-term mean PM2 5 concentration across
6 the cities was 21 u.g/m3 in the initial study and in the reanalysis of that study, within an overall
7 range of 9 to 34 ug/m3. In the extended ACS study, the mean for the more recent time period
8 used in the analysis (from 1999 to 2000) was 14 u.g/m3; in looking at the association based on the
9 air quality averaged over both time periods (which was the basis for the concentration-response
10 functions from this study used in the risk assessment), the long-term mean PM25 concentration
11 was 17.7 ug/m3, with a standard deviation of ± 4, ranging down to 7.5 fig/m3. The CD notes that
12 the confidence intervals around the relative risk functions in this extended study, as in the initial
»
13 ACS study, start to become appreciably wider below approximately 12 to 13 u.g/m3. In
14 considering the Southern California children's cohort study showing evidence of decreased lung'
15 function growth, staff notes that the long-term mean PM2 5 concentration was 15 ug/m3, ranging
16 from 7 to 32 ng/m3 across the cities. This is approximately equal to the long-term mean PM2 j
17 concentration in the earlier 24 City study, showing effects on children's lung function; in which '
18 the long-term mean concentration was 14.5 ug/m3, ranging from 9 to 17 ug/m3 across the cities:
19 In considering this evidence, staff concludes that these studies provide a basis for
20 considering an annual PM25 standard somewhat below 15 ug/m3,' down to about -12 ug/m3. A
21 standard of 14'u.g/m3 would reflect some consideration of the more recent long-term exposure
22 studies that show associations over a somewhat lower range of air quality than had been
23 observed in the studies available in the last review. A standard of 13 jig/m3 would be consistent
24 with a judgment that appreciable weight should be accorded these long-term exposure studies,
25 particularly taking into account the most recent extended ACS mortality study' 'and the Southern
26 California children's cohort morbidity study. A standard level of 13-u.g/m3 would be well below
27 the long-term mean in the Six Cities mortality study 'and approximately 'one standard deviation
28 below the extended ACS mortality study mean, while being somewhat closer to the long-term
29 means in the morbidity studies discussed above. A standard of 12 ug/m3 would be consistent
January 2005 5-26 Draft - Do Not Quote or Cite
-------
r
I
w
o
€:
|!
3 •
10
*,
o
S3
-8
*3
1-^
^ •<
li
3
S
9-,
R
8
8 /
s
5-
3
S
8
S
S
S
* I
fe
S
a
g
S
S
fe
ft
a
*
9
i
O)
I
OS
oo Os p i— i
>r> vo
o\ o
-------
1 The alternative annual PM2 5 standards considered here include a range of levels from 15
2 to 12 ug/m3, and simulating attainment of the standards is based on a percent rollback calculated
3 using the highest monitor in an area, as noted in Table 5-1 and discussed in Chapter 4, section
4 4.2.3, The alternative 24-hour PM2 5 standards considered here include a range of levels from 65
5 to 25 ug/m3 in conjunction with two different forms, including the 98th percentile form of the
6 current 24-hour PM2 5 standard and an alternative 99th percentile form. Further discussion of
7 alternative forms of the annual and 24-hour standards is presented below in section 5.3.6.
8 In looking at the base case estimates, staff has first considered the estimated reductions
9 associated with lower levels of the annual PM2 5 standard, without changing the 24-hour
10 standard. From Table 5-2, staff observes that alternative annual standard levels of 14,13, and 12
11 ug/m3 result in generally consistent estimated risk reductions from long-term exposure to PM2 5
12 of roughly 20, 30, and 50 percent, respectively, across all five'example cities. Thus, for the base
13 case assessment in which mortality risks are estimated down to the lowest measured level in the
14 extended ACS study, estimated reductions in mortality associated with long-term exposure to
15 PM2S are no greater than 50 percent in any of the five example cities with changes in the annual
16 standard down to a level of 12 ug/m3.
17 Staff also examined the effect on mortality reduction if the 24-hour standard were to
18 change. Staff first notes that the estimated reductions in long-term mortality risk associated with
19 changes to the 24-hour standard are much more variable across cities than with changes in just
20 the annual standard. Further, no combination of standards within the ranges that staff has
21 considered result in the elimination of all estimated long-term mortality risk in all example cities.
22 This assessment indicates that estimated reductions in long-term mortality risk of approximately
23 50 percent or greater in the five example cities generally result from 24-hour standards set at 30
24 to 25 ug/m3, based on either the 98th or 99th percentile form of such a standard, depending on the
25 city. ' ''
26 Staff further considered the effects of various combinations of the annual and 24-hour
27 standard. Staff notes in particular that the base case estimates of long-term mortality risk
28 reduction associated with a 24-hour standard set at 25 ug/m3 provides the same degree of risk
29 reduction regardless of the level of the annual standard within the range of 15 to 12 ug/m3; a 24-
January2005 5-30 Draft - Do Not Quote or Cite
-------
1 hour standard set at 30 |ig/m3 provides the same degree of risk reduction in most but not all
2 cases. That is, in the range of 30 to 25 ug/m3, the 24-hour standard would be the generally
3 controlling standard in most cases relative to an annual standard in the range of 15 to 12 ug/m3;
4 and, in those cases, lowering the annual standard to as low as 12 ug/m3 would result in no
5 additional estimated reductions in long-term mortality risks.
6 Beyond this base case assessment, staff also has considered the extent to which the
7 assumption of the presence of hypothetical thresholds in the concentration-response relationships
8 would influence the estimated risk reductions. As noted above (section 5.3.1), the estimated
9 incidence of PM25-related mortality associated with long-term exposure when the current
10 standards are met are appreciably smaller, although still present, under these assumed
11 hypothetical thresholds, hi considering an assumed threshold of 10 ug/m3, staff observes that
12 lowering the annual standard to alternative levels of 14,13, and 12 ng/m3 (without changing the
13 24-hour standard) results in estimated risk reductions of roughly 30 to 40 percent, 50 to 70
14 percent, and 80 to 100 percent, respectively, across the five example cities. In considering
15 changes to the annual and/or 24-hour PM2 5 standards in this case, staff first notes that mortality
16 risk associated with long-term exposure is estimated to be reduced by 100 percent in all five
17 cities with a 24-hour standard set at 25 ug/m3, in combination with the current annual standard.
18 For a 24-hour standard set at 35 ug/m3, with a 99th percentile form, estimated risk reductions
19 remained at 100 percent in three of the cities, but were only 40 and 6 percent in the other two
20 cities. Under this assumed threshold of 10 ug/m3, similar to the base case, there is little if any
21 additional reduction obtained in lowering the annual standard below 15 ug/m3 in conjunction
22 with 24-hour standards in this range. Thus, in this case, as in the base case, changes in the 24-
23 hour standard, while retaining the current annual standard, can result in larger but much more
24 variable estimated reductions in risks associated with long-term exposures across the five cities.
25 Further, in considering an assumed hypothetical threshold of 12 ug/m3, staff observes
26 that lowering the annual standard to a level of 14 ug/m3 (without changing the 24-hour standard)
27 results in estimated risk reductions of 100 percent in all five cities. In considering changes to the
28 24-hour PM2 5 standard alone in this case, staff notes that long-term mortality risk is estimated to
January 2005 5-31 Draft - Do Not Quote or Cite
-------
1 be reduced by 100 percent in all five cities with a 24-hour standard set at 30 ug/m3, 98*
2 percentile form.
3 5.3.4.3 Summary s
4 In summary, in considering the epidemiologic evidence, estimates of risk reductions
5 associated with alternative annual and/or 24-hour standards, and the related limitations and
6 uncertainties, staff concludes that there is clear support for considering revisions to the suite of
7 current PM2 5 standards to provide additional protection against health effects associated with
8 long-term exposures.. In looking specifically at the evidence of associations between long-term
9 exposure to PM25 and serious health effects, including total, cardiovascular, and lung cancer
10 mortality, as well as respiratory-related effects on children, staff concludes that it is appropriate
11 to consider an annual PM25 standard in the range of 15 down to 12 ug/m3. In considering the
12 results of the quantitative risk assessment, in the absence of evidence of clear thresholds, staff
13 believes that it is appropriate to give significant weight to base case risk estimates, while also
14 considering the implications of potential thresholds within the range of the air quality data from
15 the relevant studies. In so doing, staff finds further support for considering an annual PM2 5
16 standard in the range of 14 to 12 ug/m3. Alternatively, staff also finds support for a revised 24-
17 hour standard, in conjunction with retaining the current annual standard, in the range of 35 to 25
18 ug/m3, with an emphasis on a 99th percentile form especially with a standard level in the middle
19 or upper end of this range. Staff notes that a 24-hour standard at a level of 40 ug/m3 is estimated
20 to provide no additional protection against the serious health effects associated with long-term
21 PM25 exposures in two or three of the five example cities (for a 99th or 98th percentile form,
22 respectively) relative to that afforded by the current annual PM25 standard, regardless of the
23 weight that is given to the potential for a threshold within the range considered by staff. Staff
24 believes that a suite of PM2 5 standards selected from the alternatives identified above could
25 provide an appropriate degree of protection against the mortality and morbidity effects
26 associated with long-term exposure to PM25 in studies in urban areas across the U.S..
January 2005
5-32
Draft - Do Not Quote or Cite
-------
1 53.5 Alternative PMW Standards to Address Health Effects Related to Short-term
1 Exposure
3 In considering alternative PM2 5 standards that would provide protection against health
4 effects related to short-term exposures, staff has similarly taken into account both evidence-
5 based and risk-based considerations. As discussed below in this section, staff has first evaluated
6 the available evidence from short-term exposure studies, as well as the uncertainties and
7 limitations in that evidence, to assess the degree to which alternative 24-hour and/or annual
8 PM2 5 standards can be expected to provide protection against effects related to short-term
9 exposures. Secondly, staff has considered the quantitative risk estimates for short-term exposure
10 effects, discussed in,Chapter 4, to assess the extent to which alternative annual and/or 24-hour
11 standards can be expected to reduce the estimated risks attributable to short-term exposure to
12 PM2 5. Staff conclusions as to ranges of alternative annual and/or 24-hour standards that would
13 provide protection against health effects related to short-term exposures are summarized at .the
14 end of this section. As noted above, the integrated staff recommendations presented in section
15 5.3.7 are based in part on the conclusions from this section and in part on staff conclusions from
16 the previous section, in which alternative PM2 5 standards to address health effects related to
17 long-term exposures are assessed.
18 53.5.1 Evidence-based Considerations
19 In taking into account evidence-based considerations, staff has evaluated the available
20 evidence from short-term exposure studies, as well as the uncertainties and limitations in that
21 evidence. In so doing, staff has focused on U.S. and Canadian short-term exposure studies of
22 fine particles (Appendix 3A). We took into account reanalyses that addressed GAM-related
23 statistical issues and considered the extent to which the studies report statistically significant and
24 relatively precise relative risk estimates; the reported associations are robust to co-pollutant
25 confounding and alternative modeling approaches; and the studies used relatively reliable air
26 quality data. In particular, staff has focused on those specific studies, identified above in section
27 5.3.1, that provide evidence of associations in areas that would have met the current annual and
28 24-hour PMZ5 standards during the time of the study. Staff believes that this body of evidence
January 2005
5-33
Draft - Do Not Quote or Cite
-------
1 can serve as a Basis for 24-hour and/or annual PM2 j standards that would provide increased
1 protection against effects related to short-term exposures; ' ; ' • • •
3 As an initial matter, staff recognizes, as discussed above, that these short-term exposure
4 studies provide' no evidence of clear thresholds, or lowest-observed-effects levels, in terms of 24-
5 hour average concentrations. Staff notes that of the two PM25 studies that explored potential
6 thresholds, one study in Phoenix provided some suggestive evidence of a threshold possibly as
7 high as 20 to 25 ug/m3, whereas the other study provided evidence suggesting that if a threshold
8 existed, it would likely be appreciably below 25 ug/m3. While'there is no evidence for clear
9 thresholds within the range of air quality observed hi the epidemiologic studies, for some health
10 endpoints (such as total nonaccidental mortality) it is likely to be extremely difficult to detect
11 threshold levels (CD, p.9-45). As a consequence, this body of evidence is difficult to'translate
12 directly into a specific 24-hour standard that would independently protect against all effects
13 associated with short-term exposures. Staff notes that the distributions of daily PM2S
14 concentrations in these studies often extend down to or below background levels, such that
15 consideration of the likely range of background concentrations across the U.S., as discussed in
16 Chapter 2, section 2,6, becomes important in identifying a lower bound of a range of 24-hour
17 standards appropriate for consideration. ' .
18 Being mindful of the difficulties posed by issues relating to threshold and background
19 levels, staff has first considered this short-term exposure epidemiologic evidence as a basis for
20 alternative 24-hour PM25 standards. In so doing, staff has focused on the upper end of the
21 distributions of daily PM25 concentrations, particularly in terms of the 98th and 99th percentile
22 values, reflecting the form of the current 24-hour standard and an alternative form considered in
23 the risk assessment, respectively. In looking at the specific studies identified in section 5.3.1 that
24 report statistically significant association in areas that would have met the current PM2 5
25 standards, including studies in Phoenix (Mar et al., 1999, 2003), Santa Clara County, CA
26 (Fairley, 1999/2003) and eight Canadian cities (Burnett et al., 2000 and Burnett and Goldberg,
27 2003), staff notes that the 98th percentile values range from approximately 32 to 39 ug/m3 in
28 Phoenix and the eight Canadian cities, up to 59 ug/m3 in Santa Clara Country; 99th percentile
29 values range from 34 to 45 ug/m3 in Phoenix and the eight Canadian cities, up to 69 'ug/m3 in •
January 2005 " 5-34 Draft - Do Not Quote or die
-------
r
1 evidence discussed above, giving little weight to the remaining uncertainties in the broader body
2 of short-term exposure evidence, including the possibility of a threshold within the range of air
3 quality in the studies and the recognition that results may be sensitive to selection of models
4 bey ond the range of models examined in these particular studies.
5 Consistent with the conclusions reached in the last review (62 FR 3 8674-7), however,
6 staff continues to believe that an annual standard cannot be expected to offer an adequate margin
7 of safety against the effects of all short-term exposures, especially in areas with unusually high
8 peak-to-mean ratios of PM25 levels, possibly associated with strong local or seasonal sources, or
9 for potential PM2 5-related effects that may be associated with shorter-than-daily exposure
10 periods (noted above in section 5.3.3). As a result, if an alternative annual standard were
11 adopted to provide primary protection against effects associated with short-term exposures, staff
12 believes it is appropriate also to consider an alternative 24-hour PM2 5 standard to provide such
13 supplemental protection. Such a supplemental 24-hour standard could reasonably be based on
14 air quality information (from 2001 to 2003) in Chapter 2, Figure 2-23, that shows the distribution
15 of 98th percentile values as a function of annual means values in urban areas across the U.S.
16 Based on this information, staff concludes that a supplemental standard in the range of
17 approximately 40 to 35 ng/m3 would limit peak concentrations in areas with relatively high
18 peak-to-mean ratios (i.e., generally in the upper quartile to the upper 5th percentile, respectively)
19 and with annual mean concentrations in the range of 12 to 15 ug/m3.
20 To assist in understanding the public health implications of various combinations of
21 alternative annual and 24-hour standards, staff assessed (based on the same air quality database)
22 the percentage of counties, and the population in those counties, that would not likely attain
23 various PM25 annual standards alone in comparison to the percentage of counties that would not
24 likely attain alternative combinations of annual and 24-hour PM2 5 standards. This assessment is
25 intended to provide some rough indication of the breadth of supplemental protection potentially
' 26 afforded by various combinations of alternative standards. The results of such an assessment,
27 based on air quality data from 562 counties, are shown in Tables 5-3(a) and (b).
January 2005 5-37 Draft - Do Not Quote or Cite
,
-------
i
^>
s.
S
1
E
€
'I
s
1
— w
s
II
I!
ja e
•I
1
1
S
s
o
i
.
S
S
8?
a
3
ffl
s
• O
O)
I
a
I
oo
^^ o^O
CNCNcNtStNcNCS
-------
!
standard anc
meet stated
not likely to
c
o
'•a
i
o
Q.
Q.
*•»
1
fe
t
S"
•a
e
cs
5
__
o
o
^
W
1
•n
c
(0
0)
1
1
£
CO
i
I
it
**
°cs
I5
CO
1
Z
Southwest
&!
= =
Industrial
Midwest
1
c/>
1
"c"
o
1
al
°f
3
8
^— .
t
a>
3.
m
*~
v>
OO
CM
g
S
V-
S
S
«
l!
6 S
-c o
-|€
« .=
8 r«
5& C
ii
P
o
fe
r-.
m
^-
R
§
£
in
SO.
o
S
CO
0
r~.
CO
o
in
^
f5
§
t-w
CO
in.
O
o
in
CO
o
t—
CO
CM
o
^g-
S
£*-*
°
f*.
E
•r-
in
•q-
CO
O
r~-
CD
CD
O
^J-
S
°
£
OO
in
CM
§
CO
r»-
i»-
co
8
o
•*_••
^-
F:
5
CO
CM
CO
m
«
C*3
CO
§
in
o>
OO
S
9
f°
£
r~
o
CO
CO
CO
o
OO
in
CO
CO
^r
3
i»-
«
£
OO
in
CM
co
o
&
CM
in
CM
3
co
p
CO
CO
S
in
CO
CM
o
fe
CM
in
CM
S
CO
0
1
S
o
in
CM
o
N-
CD
«JT
O
CM
S
CO
o
S
S
in
•*
CM
O
P—
CO
S
o
CM
S
CO
p
OO
CO
in
§
CM
r*~
&
%
o
•*
S
CO
S
e
S
o
CO
CM
CO
S
£
C*J
*r
3
r—
OO
CO
a
o>
in
CM
CM
1
£•
1*3
& fc
.Sa *
I 1
U
33
§|
•8 1
1 1 '
•s >.
« "a
3J O
M g
m °
Ii «
ra j>
•5 ^
J ,+5 .
"
1 S^ a
& £; TS **
S '^3 1
3, i -S • 8
|i a a
3 ^3 co o
IM 5
^ i 1 ^
r- TO .£ 1^-1
73 i-c 00 ^
!|| I
i|; i
Hi i
S "° >>
8 s-s §
<^ T3 HS 2
fll ^^ *.
SW 4J Jf?*
,3 H *J
s>« 1
§2*3 -1
Tfl D M -g
p^ cd s M.
» -3 S S
o
I
i— • (S m
so ox o i— '
en
>n vo
-------
i
o
B
/^^
i
o
U in
.S'S
-I I
« i
3ts
_ >n
Is
(U
&5
19 =
S.
It
.
p
•0 C
§1
S §
"I
ll
11
a. S
3
JS
H
ja
«
J.
•o
C
CO
in
*
•o
S
«
"S
I
E
S
o
1
i
fc
s>
•o
§
1
I
c
3
o
8
_«
•S
fc
C
3
s
§
**
"*
ii
of
C
u
f S
w
1
1
o
5
w
ll
°-s
Industrial
Midwest
I
o
w
1
3>
1
z
*£*
o
"So
3
-S ft
"~ z
C
J
E
in
in
CO
CN
.
§
co
CO
s
S"
S"
in
"g C
i i
p
* C
s- t*
^ .1
13 g.
^5
•
.
f
3
c
o
o
CO
"*
o
o
OJ
CM
Is-
CT>
O
CO
^r
in
o
to
in
in
o
in
T—
CO
CM
5
in
T~
O
cS
r-.
in
-
CO
r-
o
••a-
-
in
0
co
o
cS
CN
in
CN
.
CO
o
»
in
CO
o
CO
in
ro
o
r—
^
S
§
5
in
CO
in
CO
o
CO
s
«
»
&
s
£
00
CO
CO
8
in
CO
CO
e
s
.
.
g
S
en
CO
in
CO
S
in
o
fe
CO
in
o
U)
CM
CO
CM
5
in
m
CD
•*
O
55
in
o
o
kO
CM
CO
CM
5
CM
O
in
>^-
O
co
CM
O
o
CM
in
T""
CM
S
sg
CO
CM
in
•«
CO
*
o
o
c—
in
CM
CO
co
in
in
CO
o
•q-
CO
O
CO
in
en
o
CO
EM
CM
r-
^
Is-
S
••a-
CO
CO
s
CO
CO
%
&
s
CO
en
CM
CO
CO
CO
S
•*f
CO
r—
CO
s
CO
.
§1
CO
CO
s
CO
eg
in
CN
•*
0
a
i
•—< cs
-------
.
i
a
U)
CQ
?
CD
m
"S
13
"55
1
E
£
"5
"S
o
-.=
2
S
•*•*
I
1
e
o
§?
•o
a
i!
1
|
1
<£
§
•p
8
1
i
TB ^
**
c
^ 5
o
co
ts
1
•e
8
§
CO
«3 1
a- *
Q.-5
11
11
1
n
o
CO
1
z
1
ll
"8
€
8
-;?•
OJ
••w*
10
!P
00
iN
en
1
fO
S3
&
r>-
S
to
1?
1 §
•'g ^
* c
•S ^>
^ o
1!
£--
o
CO
en
in
fs
°
5?
S
o
in
CO
CO
O
to
in
?
CO
0
£
r—
5
,—,
in
CO
O
CO
CM
o
CD
°
^
CO
CO
r-
CD
CO
o
r-.
°
tn
ur>
S.
^
e^
••J-
CO
CO
o
oo
in
en
in
CO
§
in
S
S
in
CO
CO
CO
o
oo
in
oo
CO
in
en
S
8
s
f=
,—,
CO
CO
CO
f—
oo
CO
CO
en
en
co
O
S
08
in
CM
CO
O
CO
CM
in
CM
en
CO
CO
g
1
S
in
CD
CM
O
E5
CD
?
CM
OO
CO
0
£
8
C-J
in
CM
O
fe
CM
CM
O
CM
cn
oo
CO
r—
£
<0
in
CM
r—
f2
co
co
o
CM
en
OO
S
K
E.
05
O
CM
CO
S
in
en
CM
cS
CO
§
b
S
in
-------
1 For example, from Table 5-3 (a) it can be seen that for an annual standard set at 15 ug/m3,
2 24-hour standard levels ranging from 40 to 35 ug/m3, with a 98* percentile form, would add
3 approximately 3 to 13 percent to the percentage of counties nationwide that would not likely
4 attain both standards relative to the number of counties that would not likely attain the annual
5 standard alone; with a 99th percentile form, as seen in Table 5-3(b), these percentages increase to
6 13 to 30 percent. For an annual standard set at 12 ug/m3, 24-hour standard levels in this range
7 would add approximately 1 to 4 percent, or 5 to 9 percent, to the percentage of counties for
8 standards with a 98th or 99th percentile form, respectively. As seen in Tables 5-3(a) and (b), the
9 percentage of the population that would be afforded greater public health protection from these
10 alternative standards would increase somewhat more than would the percentage of counties not
11 likely to attain the standards.
12 5.3.5.2 Risk-based Considerations . .."
13 Beyond looking directly at the relevant epidemiologic evidence, staff has also considered
V14 the extent to which specific levels and forms of alternative 24-hour and annual PM25-standards
15 are likely to reduce the estimated risks attributable to short-term exposure to PM25, and the
16 uncertainties in the estimated risk reductions. As discussed above (section 5.3.1), staff has based
17 this evaluation on the risk assessment results presented in Chapter 4, in which short-term ,
18 exposure risks were estimated down to background or the lowest measured level (LML) in a
•19 particular study, whichever is higher. Staff also has considered the sensitivity of these results to
20 the uncertainty related to potential thresholds by using concentration-response functions
21 modified to incorporate assumed hypothetical threshold levels.
<~
22 Table 5-4 summarizes estimated percentage reductions in mortality attributable to short-
23 term exposure to PM2S in going from meeting the current PM25 standards to meeting alternative
24 annual and 24-hour PM2 5 standards in the five example cities that do not meet the current
25 standards based on 2001-2003 air quality data. Base case estimated percentage risk reductions
26 are given in the table, along with reductions associated with assumed alternative hypothetical
27 thresholds. The percentage reductions presented in Table 5-4 represent approximate reductions
28 relative to the total'estimated short-term mortality incidence presented above in Table 5-1.
January 2005 5-42 Draft - Do Not Quote or Cite
-------
2
0>
"3
_l
z
"I
in
*3
b-
n
f
EL
1
<
§
.±2
rS
•
2J
0
•
8
m
^
o
v"
1
.a
3
in
o
1
s
m
o
0)
I
«N
m
o
8
£
8
in
o
IMSsumea inresnoia ujgiim)
(base = backqround or LML1
o
v»
8
§
§
^~
>o
g
a
§
§
o
§
s
CO
s
s '
§
8
Cg
O
8
s
**^
Incidence Associated with
Meetina Current Standards
«
\-
\
\
(-
'\
'»
;>
S
}
,» "
\
^
,}
• »
.1.. -
r
jl
T"
;%
TO*
"H ^
i&
•p> '
$
0
o
o
0
o
CM
o
•*!•
co
o
^-
8
a
TT
O
O
o
o
o
o
o
§
^^
CO
o
-
s
8
CD
o
o
o
in
CO
h-
co
OJ
in
u>
CO
co
§
g
s
CM
m
CO
oo
co
r-
CJ
CD
Eo
&
\s
CO
M-
«
CO
CO
«
CM
r—
o
o
CO
s
S3
CO
CO
CM
CO
§
8
fc
5
s
3
e
en
XT
co
CO
s
s
CO
•«J-
R
S
OO
xr
?l
cQ
o
o
o
o
o
o
o
o
3
to
•sT
8
en
o
o
o
o
o
o
o
.£
•-P
1
0>
a.
%
f
rt
1
LO
CO
O
o
o
o
o
*r
§
S
CO
oo
CO
CM
m
CO
r-.
K
CO
CO
CO
o
o
o
C3
•*
5
3
CM
o
§
8
s
a
o
o
fe
CO
£
g
CO
CO
s
o>
•*
CO
^
CM
in
CO
R
s
9
a
o
CO
§
CO
CO
oo
CO
s
3
s
s
£
OJ
s
o
in
s
51
CO
CO
R
§
s
c
8
§
§
CM
CO
o
vo
§
1
r—
en
R
B
fe
SI
S
S
CO
r~-
%
8
?
T-,
<*> :
§2
«>
§
^
to :
«& '
9
f?
« i
^
* ..
8
fe"" ",:
» ;
13
S'
1^ •
V-
» ,
1
i^
•6
la
*
^
CO
en
CO
o
•^-
CM
CD
CO
§
CO
CO
CM
T
s
Si
t~-
O
8
CO
CM
O
o
rr
5
*—
CO
CM
CO
CO
CO
R
S
h-.
CO
S
in
m
CO
•<»-
S
S
CO
CM
O
S
S
S3
s
CM
CO
§
S
r--
r^
5
£
§
co
f-.
CT)
Tf
CO
CO
1".
CD
CO
9
R
CM
CO
•*
CO
in
CM
5
CO
CO
en
•«r
a
o
oo
s
s
s
§
s
s
in
CO
fe
T™"
o>
oS
en
to
o
in
3
CO
CO
8
§
s
-------
s»
Li
1
£
in
S
2'
o _
!!
E H
381
^1
S
s
00
8
o
W)
Associated with
tandards
Inci
M
"
w
O
•»
B
to
CO
-------
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
The same alternative standards are considered here as were considered above in section
5.2.4. That is, the alternative annual PM2 5 standards considered here include a range of levels
from 15 to 12 ug/m3, and simulating meeting these standards is based on a percent rollback
calculated using the highest monitor in an area, as noted in Table 5-1 and discussed in Chapter 4,
section 4.2.3. The alternative 24-hour PM25 standards considered here again include a range of
levels from 65 to 25 ug/m3 in conjunction with two different forms, including the 98th percentile
form of the current 24-hour PM25 standard and an alternative 99* percentile form. Further
discussion of these alternative forms for annual and 24-hour standards is presented below in
section 5.3.6.
In looking at the base case estimates, staff first considered the estimated reductions
associated with lower levels of the annual PM25 standard, without changing the 24-hour
standard. .From Table 5-4, staff observes that lowering the annual standard to alternative levels
of 14,13, and 12 ug/m3 results in small but generally consistent estimated risk reductions of
roughly 10 to 15 percent, 15 to 20 percent, and 25 to 30 percent, respectively, across all five
example cities, Thus, for the base case assessment in which mortality risks are estimated down
to background or the lowest measured level in the relevant study, estimated reductions in
mortality associated with short-term exposure to PM2 5 are no greater than 30 percent in any of
the five example cities with changes in the annual PMZ5 down to a level of 12 ug/m3.
In considering changes to the 24-hour and/or annual PM2 5 standards for base case
estimates, staff first notes that the estimated reductions in short-term mortality risk associated
with changes to the 24-hour standard are generally larger and much more variable across cities
than with changes in just the annual standard. Further, no combination of standards within the
ranges that staff has considered results in the elimination of all estimated mortality risk
associated with short-term exposure in all example cities. More specifically, a 24-hour standard
of 25 ug/m3 results in base case estimates of reductions in short-term mortality ranging from
approximately 30 to 50 percent (98th percentile form) and 35 to 70 percent (99th percentile form)
across the five cities in conjunction with any annual standard in the range of 15 to 12 ug/m3. A
24-hour standard of 30 ug/m3 results in base case estimates of reductions in short-term mortality
ranging from approximately 25 to 35 percent (98th percentile form) and 25 to 65 percent (99th
January 2005
5-45
Draft - Do Not Quote or Cite
-------
1 percentile form) across the five cities in conjunction with an annual standard of 12 ug/m3; the
2 lower end, but not the upper end, of these ranges decreases somewhat in conjunction with annual
3 standards from 13 to 15 ug/m3. As in the assessment of risk related to long-term exposures
4 discussed in section 5.3.4.2, this assessment indicates that 24-hour standards of 30 to 25 ug/m3
5 become generally controlling standards in most cases within this range of annual standards.
6 Beyond this base case assessment, staff also has considered the extent to which the
7 assumption of the presence of hypothetical thresholds in the concentration-response relationships
8 would influence the estimated risk reductions. As noted above (section 5.3.1), the estimated
9 incidence of PM2 5-related mortality associated with short-term exposure when the current
10 standards are met are appreciably smaller under these assumed hypothetical thresholds. In
11 considering an assumed threshold of 10 ug/m3, staff observes that lowering the annual standard
12 to alternative levels of 14,13, and 12 ug/m3 (without changing the 24-hour standard) results in
\.
13 estimated risk reductions of roughly 15 to 25 percent, 30 to 35 percent, and 45 to 55 percent,
14 respectively, across all five example cities. In considering changes to the 24-hour and/or annual
15 PM25 standards in this case, staff notes that a 24-hour standard of 25 ug/m3 results in estimates
16 of reductions in short-term mortality ranging from approximately 45 to 80 percent (98*
17 percentile form) and 60 to 95 percent (99th percentile form) across the five cities in conjunction
18 with any annual standard in the range of 15 to 12 ug/m3. A 24-hour standard of 30 ug/m3 results
19 in estimates of reductions in short-term mortality ranging from approximately 45 to 60 percent
20 (98th percentile form) and 50 to 95 percent (99th percentile form) across the five cities in
21 conjunction with an annual standard of 12 ug/m3; as with the base case, the lower end, but not
22 the upper end, of these ranges decreases appreciably in conjunction with annual standards from
23 13 to 15 ug/m3. Thus, in this case, as in the base case, changes in the 24-hour standard, while
24 retaining the current annual standard, can result in generally larger but much more variable
25 estimated reductions in risks associated with short-term exposures across the five cities than with
26 changes in just the annual standard.
27 Further, in considering assumed hypothetical thresholds of 15 or 20 ug/m3, staff observes
28 that lowering the annual standard to alternative levels of 14,13, and 12 ug/m3 (without changing
29 the 24-hour standard) results in estimated risk reductions of roughly 20 to 45 percent, 40 to 65
January 2005 5-46 Draft - Do Not Quote or Cite
-------
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
percent, and 60 to 90 percent, respectively, across all five example cities. In considering
changes to the 24-hour and/or annual PM25 standards in this case, staff notes that a 24-hour
standard of 25 ug/m3 results in estimates of reductions in short-term mortality ranging from
approximately 60 to 100 percent (98th percentile form) and 70 to 100 percent (99th percentile
form) across the five cities in conjunction with any annual standard in the range of 15 to 12
ug/m3. A 24-hour standard of 30 ug/m3 results in estimates of reductions in short-term mortality
ranging from approximately 60 to 90 percent (98th percentile form) and 60 to 100 percent (99th
percentile form) across the five cities in conjunction with an annual standard of 12 ug/m3;
similarly, the lower end, but not the upper end, of these ranges decreases appreciably in
conjunction with annual standards from 13 to 15 ug/m3. Thus, in this case as .well, changes in
«
the 24-hour standard, while retaining the current annual standard, can result in generally larger
but much more variable estimated reductions in risks associated with short-term exposures
across the five cities than with changes in just the annual standard.
53.5.3 Summary
In summary, in considering the relevant epidemiologic evidence, estimates of risk
reductions associated with alternative annual and/or 24-hour standards, and the related
limitations and uncertainties, staff concludes that there is clear support for considering revisions
to the suite of current PM25 standards to provide additional protection against health effects
associated with short-term exposures. In looking specifically at the evidence of associations
between short-term exposure to PM2 5 and serious health effects, with a particular focus on
mortality associations, staff concludes that it is appropriate to consider a revised 24-hour
standard in the range of 30 to 25 ug/m3 in conjunction with retaining the current annual standard
level of 15 ug/m3. Alternatively, staff also believes the evidence supports consideration of a
revised annual standard, in the range of 13 to 12 ug/m3., in conjunction with a revised 24-hour
standard, to provide supplemental protection, in the range of 40 to 35 ug/m3. In considering the
results of the quantitative risk assessment, in the absence of evidence of clear thresholds, staff
believes mat it is appropriate to give significant weight to base case risk estimates, while also
considering the implications of potential thresholds within the range of the air quality data from
the relevant studies. In so doing, staff also finds support for considering a revised 24-hour
January 2005
5-47
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
g
9
10
standard, in conjunction with retaining an annual standard level of 15 ug/m3, in the range of 30
to 25 ug/m3. Staff notes that a 24-hour standard at a level of 35 ug/m3 is estimated to provide
less than 30 percent reduction in mortality incidence in two or three of the five example cities
(for a 99th or 98* percentile form, respectively), in either the base case or under an assumed
hypothetical threshold of 10 ug/m3, relative to that afforded by the current annual PM25 standard
alone. Further, staff finds little support based on the risk assessment for addressing short-term
exposure effects solely with a revised annual standard in a range down to 12 ug/m3. Staff
believes that a suite of PM25 standards selected from the alternatives identified above could
provide an appropriate degree of protection against the mortality and morbidity effects
associated with short-term exposure to PM25 in studies in urban areas across the U.S..
11 5.3.6 Alternative Forms for Annual and 24-Hour PM2.5 Standards
12 53.6.1 Form of Annual Standard
13 ^ In 1997 EPA established the form of the annual PM2S standard as an annual arithmetic
14 mean, averaged over 3 years, from single or multiple community-oriented monitors. This form
15 was intended to represent a relatively stable measure of air quality and to characterize area-wide
16 PM25 concentrations. The arithmetic mean serves to represent the broad distribution of daily air
17 . quality values, and a 3-year average provides a more stable risk reduction target than a single-
18 year annual average. The annual PM2 s standard level is to be compared to measurements made
19 at the community-oriented monitoring site recording the highest level, or, if specific constraints
20 are met, measurements from multiple community-oriented monitoring sites may be averaged (62
21 FR at 38,672). The constraints on allowing the use of spatially averaged measurements were
22 intended to limit averaging across poorly correlated or widely disparate air quality values. This
23 approach was judged to be consistent with the epidemiologic studies on which the PM2 5 standard
24 was primarily based, in which air quality data were generally averaged across multiple monitors
25 in an area or were taken from a single monitor that was selected to represent community-wide
26 exposures, not localized "hot spots."
27 In this review, in conjunction with recommending that consideration be given to
28 , alternative annual standard levels, staff is also reconsidering the appropriateness of continuing to
January 2005
5-48
Draft - Do Not Quote or Cite
-------
\
1 allow spatial averaging across monitors as part of the form of an annual standard. There now
2 exist much more PM2 5 air quality data than were available in the last review. Consideration of
3 the spatial variability across urban areas that is revealed by this new database (see Chapter 2,
4 section 2.4 above, and the CD Chapter 3, section 3.2.5) raises questions as to whether an annual
5 standard that allows for spatial averaging, within currently specified or alternative constraints,
6 would provide appropriate public health protection. In conducting analyses to assess these
7 questions, as discussed below, staff has taken into account both aggregate population risk across
8 an entire urban area and the potential for disproportionate impacts on potentially vulnerable
9 subpopulations within an area
10 The effect of allowing the us e of spatial averaging on aggregate population risk was
11 considered as part of the sensitivity analyses included in the health risk assessment discussed in
12 Chapter 4. In particular, a sensitivity analysis was done in several example urban areas (Detroit,
13 Pittsburgh, and St. Louis) that compared estimated mortality risks (associated with both long-
14 and short-term exposures) based on air quality values from the highest community-oriented
15 monitor in an area with estimated risks based on air quality values averaged across all such
16 monitors within the constraints allowed by the current standard. As discussed in Chapter 4,
17 section 4.2.3, the monitored air quality values were used to determine the design value for the
18 annual standard in each area, as applied to a "composite" monitor to reflect area-wide exposures.
19 Changing the basis of the annual standard design value from the concentration at the highest
20 monitor to the average concentration across all monitors reduces the air quality adjustment
21 needed to just meet the current or alternative annual standards. As expected, the estimated risks
22 remaining upon attainment of the current annual standard are greater when spatial averaging is
23 used than when the highest monitor is used. Based on the results of this analysis in the three
24 example cities, estimated mortality incidence associated with long-term exposure based on the
25 use of spatial averaging is about 10 to over 40% higher than estimated incidence based on the
26 use of the highest monitor. For estimated mortality incidence associated with short-term
27 exposure, the use of spatial averaging results in risk estimates that range from about 5 to 25%
28 higher. In considering estimated risks remaining upon attainment of alternative suites of annual
29 and 24-hour PM25 standards, spatial averaging only has an impact in those cases when the
January 2005
5-49
Draft - Do Not Quote or Cite
-------
1 annual standard is the "controlling" standard. For such cases in the three example cities, the
2 estimated mortality incidence associated with long-term exposure in most cases ranges from
3 i about 10 to 60% higher when spatial averaging is used, and estimated mortality incidence
4 associated with shortrterm exposure in most cases ranges from about 5 to 25% higher.
5 'In considering the potential for disproportionate impacts on potentially vulnerable
6 subpopulations, staff has assessed whether any such groups are more likely to live in census
7 tracts in which the monitors recording the highest air quality values in an area are located. Data
j f
^ 8 were obtained for demographic parameters measured at the census tract level, including
9 education level, income level, and percent minority. These data from the census tract in which
10 the highest air quality values were monitored were compared to area-wide average values
11 ' (Schmidt etal., 2005). Recognizing the limitations of such cross-sectional analyses, staff-
12 observes that the results suggest that the highest concentrations in an area tend to be measured at
13 monitors located in areas where the surrounding population is more likely to have lower
14 , education and income levels, and higher percentage minority levels. Staff notes that some
~v
15 epidemiologic study results, most notably the associations between mortality and long-term
16 PM2 5 exposure in the ACS cohort, have shown larger effect estimates in the cohort subgroup
17 with lower education levels (CD, p. 8-103). As discussed in Chapter 3, section 3.4, people with
18 lower socioeconomic status (e.g., lower education and income levels), or who have greater
19 exposure to sources such as roadways, may have increased vulnerability to the effects of PM
20 exposure. Combining evidence from health studies suggesting that people with lower
21 socioeconomic status may be considered a population more vulnerable to PM-related effects
22 with indications'from air quality analyses showing that higher PM25 concentrations are measured
23 in local communities with lower socioeconomic status, staff finds that this is additional evidence
24 which supports a change.from spatial averaging across PM25 monitors to provide appropriate
25 protection from public health risks associated with exposure to ambient PM2 5.
26 In considering whether alternative constraints on the use of spatial averaging may be
27 , appropriate to consider, staff has analyzed existing data on the correlations and differences
28 '' between monitor pairs in metropolitan areas (Schmidt etal., 2005). For all pairs of PM25
29 • monitors, the median correlation coefficient based on annual air quality data is approximately
January 2005 '
5-50
Draft - Do Not Quote or Cite
-------
1 0.9; i.e., substantially higher than the current criterion for correlation of at least 0.6, which was
2 met by nearly all monitor pairs. Similarly, the current criterion that differences in mean air
3 quality values between monitors not exceed 20% was met for most monitor pairs, while the
4 annual median and mean differences for all monitor pairs are 5% and 8%, respectively. This
5 analysis also showed that in some areas with highly seasonal air quality patterns (e.g., due to
6 seasonal woodsmoke emissions), substantially lower seasonal correlations and larger seasonal
7 differences can occur relative .to those observed on an annual basis. The spatial averaging
8 requirements established in 1997 were intended to represent a relatively stable measure of air
9 quality and to characterize area-wide PM2 5 concentrations, while also precluding averaging
10 across monitors that would leave a portion of a metropolitan area with substantially greater
11 exposures than other areas (62 FR 38672). Based on the PM2 5 air quality data now available,
12 staff believes.that the existing constraints on spatial averaging may not be adequate to achieve
13 this result
14 In considering the results of the analyses discussed above, staff concludes that it is
15 appropriate to consider eliminating the provision that allows for spatial averaging from the form
16 of an annual PM15 standard. Further, staff concludes that if consideration is given to retaining an
17 allowance for spatial averaging, more restrictive criteria should be considered. Staff believes
18 that it would be appropriate to consider alternative criteria such as a correlation coefficient of at
19 least 0.9, determined on a seasonal basis, with differences between monitor values not to exceed
20 about 10%.
21 5.3.6.2 Form of 24-Hour Standard
22 In 1997 EPA established the form of the 24-hour PM2 5 standard as the 98th percentile of
23 .24-hour concentrations at each population-oriented monitor within an area, averaged over three
24 years (62 FR at 38671-74). EPA selected such a concentration-based form because of its
25 advantages over the previously used expected-exceedance form.4 A concentration-based form is
26 more reflective of the health risk posed by elevated PM2 5 concentrations because it gives
27 proportionally greater weight to days when concentrations are well above the level of the
4 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.
January 2005
5-51
Draft - Do Not Quote or Cite
-------
1 standard than to days when the concentrations are just above tiie standard. Further, a
2 concentration-based form better compensates for missing data and less-than-every-day
3 monitoring; and, when averaged over 3 years, it has greater stability and, thus, facilitates the
4 development of more stable implementation programs. After considering a range of
5 concentration percentiles from the 95th to the 99th, EPA selected the 98th percentile as an
6 appropriate balance between adequately limiting the occurrence of peak concentrations and
7 providing increased stability and robustness. Further, by basing the form of the standard on
8 concentrations measured at population-oriented'monitoring sites (as specified in 40 CFR part
9 58), EPA intended to provide protection for people residing in or near localized areas of elevated
10 concentrations.
11 : In this review, in conjunction with recommending that consideration be given to
12 alternative 24-hour standard levels, staff is also considering the appropriateness of
13 recommending that the current 98* percentile form, averaged over 3 years, be retained or
14 revised. As an initial matter, staff believes that it is appropriate to retain a concentration-based
15 form that is defined in terms of a specific percentile of the distribution of 24-hour PM2 5
16 concentrations at each population-oriented monitor within an area, averaged over 3 years, Staff
17 bases this recommendation on the same reasons that were the basis for EPA's selection of this
18 type of form in the last review. As to the specific percentile value to be considered, staff has
19 narrowed the focus of this review to the 98th and 99th percentile forms. This focus is based on the
20 observation that the current 98th percentile form already allows the level of the standard to be '
21 exceeded seven days per year, on average (with every-day monitoring), while potentially
22 allowing many more exceedance days in the worst year within the 3-year averaging period
23 (Schmidt et al., 2005). As a result, in areas that just attain the standards, EPA's communication
24 to the public through the Air Quality Index will on one hand indicate that the general level of air
25 quality is satisfactory (since the standards are being met), but on the other hand it may identify
26 many days throughout the year as being unhealthy, particularly for sensitive groups. Thus, staff
27 does not believe it would be appropriate to consider specifying the form in terms of an even
28 lower percentile value.
f
January 2005
5-52
Draft - Do Not Quote or Cite
-------
1 In considering differences between 98th and 99th percentile forms, staff believes it is
2 appropriate to take into consideration the relative risk reduction afforded by these alternative
3 forms at the same standard level. Based on the risk assessment results discussed in Chapter 4,
4 and the risk reductions associated with alternative levels and forms discussed above in sections
5 5.3.4 and 5.3.5, staff notes that the 99th percentile can, in some instances, result in appreciably
6 greater risk reductions in particular areas than that associated with a standard at the same level
7 but with a 98th percentile form. More specifically, staff considered the differences in risk
8 reductions associated with attaining alternative standards with 98th and 99th percentile forms in
9 five example urban areas (Detroit, Los Angeles, Philadelphia, Pittsburgh, and St. Louis). In
10 looking at estimated risk reductions associated with meeting a 24-hour standard of 30 |ig/m3, for
11 - example, estimated risk reductions for mortality associated with long-term exposures were
12 higher with the use of a 99th percentile form in some areas by approximately 15%, ranging up to
13 over 50% higher in Los Angeles. For estimated risk reductions for mortality associated with
14 short-term exposures, the use of a 99th percentile form resulted in estimated reductions that were
15 higher by less than 10% to over 30% across the five urban areas.
16 Staff also analyzed the available air quality data from 2001 to 2003 to compare the 98th
17 and 99th percentile forms in terms of the numbers of days that would be expected to exceed the
18 level of the standard (on average over 3 years and in the worst year within a 3-year averaging
19 period) and by how much the standard would typically be exceeded on such days (Schmidt et al.,
20 2005). In so doing, as noted above, staff observes that the current 98th percentile form allows the
21 level of the standard to be exceeded seven days per year, on average (with every-day
22 monitoring), and finds that this form allows up to about 20 days in the worst year within the 3-
23 year averaging period. A 99th percentile form would allow the level of the standard to be
24 exceeded three days per year, on average (with every-day monitoring), while allowing up to
25 about 13 days in the worst year within the 3-year averaging period. Further, staff observes that
26 for either form, daily peak concentrations in the upper 1 to 2% of the annual air quality
27 distributions are within 5 |ig/m3 of the 98* or 99th percentile value somewhat more than half the
January 2005
5-53
Draft - Do Not Quote or Cite
-------
1 time and are almost always within 10 to 15 jig/m3 above the 98th or 99th percentile values, with '
2 very few excursions above this range.5
3 Based on these considerations, staff recommends either retaining the 98th percentile form
4 or revising it to be based on the 99th percentile air quality value. In selecting between these
5 alternative forms, staff believes primary consideration should be given to the estimated level of
6 risk reduction mat is associated with standards based on either form. Staff also believes it is
7 appropriate to take into account whether the 24-hour standard is set to supplement protection
8 afforded by an annual standard, or is intended to be the primary basis for providing protection
9 against effects associated with short-term exposures. In choosing between forms of alternative
10 standards that provide generally equivalent levels of public health protection, staff believes it is
11 appropriate to consider the implications from a public health communication perspective of the
12 extent to which alternative forms allow different numbers of days in a year to be above the level
13 of the standard in areas that attain Hie standard. In particular, staff notes that the use of a 99th
14 percentile form would result in a more consistent public health message to the general public in
15 the context of the wide-spread use of the Air Quality Index.
16 5.3.7 Summary of Staff Recommendations on Primary PMZ s NAAQS
17 Staff recommendations for the Administrator's consideration in making decisions on the
18 primary PM25 standards, together with supporting conclusions from sections 5.3.1 through 5.3.6,
19 are briefly summarized below.- Staff recognizes that selecting from among alternative standards
20 will necessarily reflect consideration of the qualitative and quantitative uncertainties inherent in
21 the relevant evidence and in the assumptions that underlie the quantitative risk assessment. In
22 recommending these alternative suites of primary standards and ranges of levels for
23 consideration, staff is mindful that the Act requires standards to be set that are requisite to
5 This analysis also looked at the number of days in which the reported air quality values were "flagged" as
being heavily influenced by natural events (including forest fires, dust storms) or exceptional events, for which the
Agency's natural and exceptional events policies would likely apply. While flagged days generally account for less
than 1% of all reported 24-hour average PMj 5 concentrations, they account for about 40% of the highest 100 days
across the country. In looking at the reported values that are above the 99* or 98th percentiles of the distribution of
values, approximately 3 to 6% of the highest 2% of days (above the 98th percentile) were flagged, and approximately
5 to 10% of the highest 1% of days (above the 99th percentile) were flagged.
January 2005 5-54 Draft - Do Not Quote or Cite
-------
1 protect public health with an adequate margin of safety, such that the standards are to be neither
2 more nor less stringent than necessary. Thus, the Act does not require that NAAQS be set at
3, zero-risk levels, but rather at levels that avoid unacceptable risks to public health.
4 ( 1) Consideration should be given to revising the current PM2 5 primary standards to provide
5 increased public health protection from the effects of both long- and short-term exposures
6 to fine particles in the ambient air. This recommendation is based in general on the
7 evaluation in the CD of the newly available epidemiologic, toxicologic, dosimetric, and
8 exposure-related evidence, and more specifically on the evidence of mortality and
9 morbidity effects in areas where the current standards were met, together with judgments
10 as to the public health significance of the estimated incidence of effects upon just
11 attaining the current standards.
12 (2) The indicator for fine particle standards should continue to be PM25. This
13 recommendation is based on the conclusion that the available evidence does not provide
14 a sufficient basis for replacing or supplementing a mass-based fine particle indicator with
15 an indicator for any specific fine particle component or subset of fine particles, nor does
16 it provide a basis for excluding any components; on the evaluation in the CD of air
17 quality within the interrnodal particle size range of 1 to 3 urn; and on the policy judgment
18 made in the last review to place regulatory importance on defining an indicator that
19 would more completely capture fine particles under all conditions likely to be
20 encountered across the U.S., while recognizing that some limited intrusion of small
21 coarse particles will occur in some circumstances. Consideration should be given to
22 modifying the FRM for PM2 5 based on instrumentation and operational improvements
23 that have been made since the PM25 monitoring network was deployed in 1999, and to
24 the adoption of FEMs for appropriate continuous measurement methods.
25 ( 3) Averaging times for PM2 5 standards should continue to include annual and 24-hour
26 averages to protect against health effects associated with short-term (hours to days) and
January 2005 5-55 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
long-term (seasons to years) exposure periods; consideration of other averaging times,
especially on the order of one or more hours, was limited by a lack of adequate
information at this time.
( a) Consideration should be given to revising the form of the annual standard to one
based on Hie highest community-oriented monitor in an area or, alternatively, to
one with more constrained requirements for the use of spatial averaging across
community-oriented monitors.
( b) Consideration should be given to revising the form of the 24-hour standard to a
99th percentile form or, alternatively, to retaining the 98th percentile form, based in
part on considering the level of risk reduction likely to result from a standard
using either form.
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
t
( 4) Consideration should be given'to alternative suites of PM25 standards to provide
protection against effects associated with both long- and short-term exposures, taking
into account both evidence-based and risk-based considerations. Integrated
recommendations on ranges'of alternative suites of standards that, when considered
together, protect against effects associated with both long- and short-term exposures
include: ' •
( a) Staff recommends consideration of an annual PM2 5 standard at the current level
of 15 ug/m3 together with a revised 24-hour PM2 5 standard in the range of 35 to
25 ug/m3. Staff judges that such a suite of standards, particularly in conjunction
with a 99th percentile form for a 24-hour standard set at the middle to upper end of
this range, could provide an appropriate degree of protection against serious
mortality and morbidity effects associated with long- and short-term exposures to
fine particles.
( b) Alternatively, staff also recommends consideration of a revised annual PM2 5
standard, within the range of 14 to 12 ng/m3, together with a revised 24-hour
PM2 5 standard to provide supplemental protection against episodic localized or
seasonal peaks, in the range of 40 to 35 ug/m3. Staff judges that such a suite of
January 2005
5-56
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
standards, particularly with an annual standard set at the middle or low end of this
range, could provide an appropriate degree of protection against serious mortality
and morbidity effects associated with long- and short-term exposures to fine
particles.
5.4 THORACIC COARSE PARTICLE STANDARDS
5.4.1 Adequacy of Current PM10 Standards
In considering the adequacy of the current PM10 standards to control thoracic coarse
particles, in conjunction with separate standards for PM25, staff has first considered the
appropriateness of using PM,0 as an indicator for thoracic coarse particles. In 1997, in
conjunction with establishing new PM2 5 standards, EPA determined that the new function of
PM,0 standards was to protect against potential effects associated with thoracic coarse particles
in the size range of 2.5 to 10 urn (62 FR 38,677). Although staff had given some consideration
to a more narrowly defined indicator that did not include fine particles (e.g., PM10.2 5), EPA
decided to continue to use PM10 as the indicator for standards to control thoracic coarse particles.
This decision was based in part on the recognition that the only studies of clear quantitative
relevance to health effects most likely associated with thoracic coarse particles used PM,0 in
areas where the coarse fraction was the predominant component of PM10, namely two fugitive
dust studies in areas that substantially exceeded the PM10 standards (62 FR 38,679). Also, the
decision reflected the fact that there was only very limited ambient air quality data then available
specifically on thoracic coarse particles, in contrast to the extensive monitoring network already
in place for PM10. In essence, EPA concluded at that time that it was appropriate to continue to
control thoracic coarse particles, which, like fine particles, are capable of penetrating to the
thoracic region of the respiratory tract, but that the only information available upon which to
base such standards was indexed in terms of PM10.6
As discussed in Chapter 1, however, in subsequent litigation regarding the 1997 PM NAAQS revisions,
the court held in part that PM10 is a "poorly matched indicator" for thoracic coarse particles in the context of a rule
that also includes PM25 standards because PMto includes PM25. 175 F. 3d. at 1054. Although the court found
"ample support" (idj for EPA's decision to regulate thoracic coarse particles, the court nonetheless vacated the 1997
revised PM10 standards for the control of thoracic coarse particles.
January 2005
5-57
Draft - Do Not Quote or Cite
f
-------
1 taking into account key uncertainties associated1 with the estimated risks. Estimates of risks
2 attributable to short-term exposure to PM10.2.5 are presented in Chapter 4 for Detroit, Seattle and
3 St. Louis, the urban areas in which the studies discussed above were conducted. These estimated
4 risks are attributable to PM10.25 concentrations above background levels, or above the lowest
5 measured levels in a given study if higher than background, so as to avoid extrapolating risk
6 estimates beyond the range of air quality upon which the concentration-response functions were
7 based.
8 In the absence of evidence for clear thresholds in any of the studies used in the risk
9 assessment, the base case estimates in the analysis reflect the linear or near-linear concentration-
10 response functions reported in the studies. To reflect the uncertainty as to whether thresholds
11 may exist within the range of air quality observed in the studies, but may not be discemable with
12 currently applied statistical methods, staff has also considered estimates of risk based on
r
13 concentration-response functions modified to incorporate various assumed hypothetical •
14 ^threshold levels. Based on the sensitivity analyses conducted as part of the risk assessment, the
15 'uncertainty associated with alternative hypothetical thresholds had by far the greatest impact on
16 estimated risks.
17 Table 5-5 summarizes the estimated PM10.25-related annual incidence of hospital
18 admissions and respiratory symptoms (cough) in children associated with short-term exposure
19 for the base case and for alternative hypothetical thresholds in the three example urban areas
20 included in the risk assessment. Staff observes that the base case estimates of cardiac-related
21 hospital admissions in Detroit are an order of magnitude greater than asthma-related admissions
22 in Seattle. Such large differences are in part attributable to the large differences in PM,0.2S,air
23 quality levels in these two areas, in which the 2003 annual average PM10.25 concentration in
24 Detroit (21.7 (ig/m3) is much higher than in Seattle (111.4; jig/m3).- Further, staff notes that the
25 2003 annual average PM10.2S concentration in St. Louis (12:0 iig/m3) is similarly far below that
26 in Detroit. In looking beyond the base case estimates, staff observes that, as expected, the risk
27 estimates are substantially smaller when concentration-response functions adjusted to reflect
28 hypothetical thresholds are considered.' At the largest assumed hypothetical threshold, estimates
r\
January 2005
5-60
Draft - Do Not Quote or Cite
-------
1 effects of the crustal contribution in thoracic coarse particles. The CD notes that particles of.
2 crustal origin may be linked with morbidity effects, or may serve as carriers for other more toxic
3 components, such as metals or organic compounds (CD, p. 9-63). The CD discusses some
*>
4 coarse particle components (e.g., metals, biogenic constituents) or sources contributing to coarse
5 particles (e.g., wood burning) that may be linked with health effects, but little evidence is
6 available on any of the components or sources within the coarse fraction at present (CD, p. 9-32).
7 Thus, as for fine particles, there is no evidence that would lead toward the selection of one or
8 more PM components as being primarily responsible for effects associated with coarse particles,
9 nor is there any component that can be eliminated from consideration. . ...
10 Taking into account the above considerations, staff concludes that a mass-based indicator
11 continues to be the most appropriate indicator for any thoracic coarse particle standards. Staff
12 recommends that such an indicator retain 10 urn as the upper cut point, and that the lower cut
13 point of 2.5 um be used so as to most clearly differentiate between thoracic coarse (PM10_2 5) and
14 fine (PM2 5) particles. In considering the evidence that suggests that high PM concentrations
15 linked with dust storm events may be of less concern, staff notes that EPA has historically used
16 natural events policies to address such issues in the implementation of PM standards. , , -.. ~"
17 In conjunction with considering PM]0.2 5 as an indicator for standards to address thoracic . ^^^
18 coarse particles, EPA is evaluating various ambient monitoring methods. This evaluation is
19 being performed through field studies of commercially ready and prototype methods to
20 characterize the measurement of thoracic coarse particles.8 The PM10_25 methods evaluation has
21 resulted in characterizing the performance of multiple PM10.2 5 measurement technologies under a
22 variety of aerosol and meteorological conditions. This characterization has demonstrated that
23 the majority of commercially available methods for the measurement of PM10.25 have good
24 .precision and are well correlated with filter-based gravimetric methods such as the difference
25 method that has primarily been used to date (i.e., operation of collocated PM10 and PM25 low
26 volume FRMs and calculating PM10.25 by difference). EPA is working with the instrument
27 manufacturers to address design issues that should reduce biases that have been observed among
28 methods, in preparation for another field study examining the performance of the methods.
This work is being done in consultation with the CASAC AAMM Subcommittee.
January 2005 5r63 Draft - Do Not Quote or Cite \ /
-------
1 taking into account key uncertainties associated with the estimated risks. Estimates of risks
2 attributable to short-term exposure to PM10.Z5 are presented in Chapter 4 for Detroit, Seattle and
3 St. Louis, the urban areas in which the studies discussed above were conducted. These estimated
4 risks are attributable to PM10.2.5 concentrations above background levels, or above the lowest
5 measured levels in a given study if higher than background, so as to avoid extrapolating risk
6 estimates beyond the range of air quality upon which the concentration-response functions were
7 based.
"J
8 In the absence of evidence for clear thresholds in any of the studies used in 1he risk
9 assessment, the base case estimates in the analysis reflect the linear or near-linear concentration-
\ 10 response functions reported in the studies. To reflect the uncertainty as to whether thresholds
11 may exist within the range of air quality observed in the studies, but may not be discernable with
12 currently applied statistical methods, staff has also considered estimates of risk based on
13 concentration-response functions modified to incorporate various assumed hypothetical
14 threshold levels. Based on the sensitivity analyses conducted as part of the risk assessment, the
15 uncertainty associated with alternative hypothetical thresholds had by far the greatest impact on
16 estimated risks.
17 Table 5-5 summarizes the estimated PM10_2 5-related annual incidence of hospital
18 admissions and respiratory symptoms (cough) in children associated with short-term exposure
19 for the base case and for alternative hypothetical thresholds in the three example urban areas
20 included in the risk assessment. Staff observes that the base case estimates of cardiac-related
21 hospital admissions in Detroit are an order of magnitude greater than asthma-related admissions
22 in Seattle. Such large differences are in part attributable to the large differences in PM10.25,air
23 quality levels in these two areas, in which the 2003 annual average PM10.2 s concentration in
24 Detroit (21.7 |ig/m3) is much higher than in Seattle (11'.4. n.g/m3).- Further, staff notes that the
25 2003 annual average PM10.25 concentration in St. Louis (12.0 jig/m3) is similarly far below that
26 in Detroit. In looking beyond the base case estimates, staff observes that, as expected, the risk
27 estimates are substantially smaller when concentration-response functions adjusted to reflect
28 hypothetical thresholds are considered. At the largest assumed hypothetical threshold, estimates
January 2005
5-60
Draft- Do Not Quote or Cite
-------
1 in Detroit are 50 percent smaller than base case estimates, whereas in St. Louis estimates are 90
2 percent smaller.
3
4
5
6
Table 5-5 Estimated PM10 2 rrelated Annual Incidence of Hospital Admissions and
Cough in Children with 2003 Air Quality in Areas that Meet the Current
PM10 Standards (Base Case and Assumed Alternative Hypothetical
Thresholds)
Detroit: hospital admissions for
ischemic heart disease
Seattle: hospital admissions for
asthma (age <65)
St. Louis: days of cough in
children
Short-term Exposure
Base case
Estimate,
95% Cl
654
169 to 1083
27
Oto65
27,000
11, 000 to 40,900
Assumed Hypothetical Thresholds
10 pg/m3
505
131 to 836
11
Oto26
11,500
4,700 to 17,400
15 (jg/m3
386
100 to 636
4
OtolO
5,400
2,200 to 8,000
20 Mg/m3
294
77 to 483
1
Oto3
2,600
1,100 to 3,700
7
8
9
10
11
12
13 Beyond the specific health endpoints presented in Table 5-5, for which sensitivity
14 analyses have been done, staff notes that hundreds of additional hospital admissions for other
15 cardiac- and respiratory-related diseases are also estimated in Detroit, based on risk assessment
16 results presented in Chapter 4, as are thousands of additional days in which children are likely to
17 experience other symptoms of the lower respiratory tract in St. Louis. In considering these
18 limited estimates, even when hypothetical thresholds are assumed, staff concludes that they are
19 indicative of risks that can reasonably be judged to be important from a public health
20 perspective, especially in areas in which PM,0.25 concentrations approach those observed in
21 Detroit.
22 In considering the evidence and risk estimates for thoracic coarse particles discussed
23 above, and the related limitations and uncertainties, staff concludes that this information is
24 sufficient to support consideration of revised standards for thoracic coarse particles to afford
January 2005
5-61
Draft - Do Not Quote or Cite
f
-------
1 protection from effects related to short-term exposure to current ambient levels of PM10.2.5 in
2. some urban areas. Staff conclusions and recommendations on an indicator and associated
3 monitoring methods, averaging times, and alternative levels and forms for thoracic coarse
4 particle standards mat would afford an appropriate degree of protection from'such effects are
5 discussed in the following sections.
6 5.4.2 Indicators ' ' ,
7 Section 5.4.1 above discusses EPA's decision in 1997 to continue to use PM10 as the-
8 indicator for standards intended to protect against the effects most likely associated with thoracic
9 coarse particles. In considering the adequacy of such standards, staff has taken into account
10 information now available on health effects and air quality in which thoracic coarse particles are
11 indexed by PM10.2 5, concluding that such information should form the basis for consideration of
.12 standards for thoracic coarse particles using an indicator that does not include the fine fraction of
13 PM10.
14 The CD concludes that the recent scientific information supports EPA's previous
15 decision to use an indicator based on PM mass, as discussed above in section 5.3.2 for fine
16 particles. In addition, currently available information from dosimetric studies supports retaining
17 10 urn as the appropriate cut point for particles capable of penetrating to the thoracic regions of
18 the lung. In conjunction with PM2 5 standards, an appropriate mass-based indicator for thoracic
19 coarse particles thus would be PM10.2 5. As noted above, this is the indicator that has been used
20 to index thoracic coarse particles in newly available epidemiologic studies and in
21 ' characterizations of air quality.
22 There is limited evidence to support consideration of other indicators for thoracic coarse
23 particles, such as individual components within mis PM fraction. In general, less is known about
24 the composition of thoracic coarse particles than fine particles. Even less evidence is available
25 from health studies that would allow identification of specific components or groups of
26 components of coarse particles that may be more closely linked with specific health outcomes.
27 While several studies have suggested that the crustral or geological component of fine particles
28 is not significantly associated with mortality (CD, p. 8-66), no studies have focused on potential
t
January 2005
5-62
Draft - Do Not Quote or Cite
-------
1 effects of the crustal contribution in thoracic coarse particles, The CD notes that particles of
2 crustal origin may be linked with morbidity effects, or may serve as carriers for other more toxic
3 components, such as metals or organic compounds (CD, p. 9-63). The CD discusses some
£.
4 coarse particle components (e.g., metals, biogenic constituents) or sources contributing to coarse
5 particles (e.g., wood burning) that may be linked with health effects, but little evidence is
6 available on any of the components or sources within the coarse fraction at present (CD, p.9-32).
7 Thus, as for fine particles, there is no evidence that would lead toward the selection of one or
8 more PM components as being primarily responsible for effects associated with coarse particles,
9 nor is there any component that can be eliminated from consideration.
10 Taking into account the above considerations, staff concludes that a mass-based indicator
11 continues to be the most appropriate indicator for any thoracic coarse particle standards. Staff
12 recommends that such an indicator retain 10 um as the upper cut point, and that the lower cut
13 point of 2.5 urn be used so as to most clearly differentiate between thoracic coarse (PM10.25) and
14 fine (PM2 5) particles. In considering the evidence that suggests that high PM concentrations
15 linked with dust storm events may be of less concern, staff notes that EPA has historically used
16 natural events policies to address such issues in the implementation of PM standards.
17 In conjunction with considering PM10.2 5 as an indicator for standards to address thoracic
18 coarse particles, EPA is evaluating various ambient monitoring methods. This evaluation is
19 being performed through field studies of commercially ready and prototype methods to
20 characterize the measurement of thoracic coarse particles.8 The PM10.25 methods evaluation has
21 resulted in characterizing the performance of multiple PM,0.2 5 measurement technologies under a
22 variety of aerosol and meteorological conditions. This characterization has demonstrated that
23 the majority of commercially available methods for the measurement of PM10.25 have good
24 precision and are well correlated with filter-based gravimetric methods such as the difference
25 method that has primarily been used to date (i.e., operation of collocated PM10 and PM25 low
26 volume FRMs and calculating PM10_25 by difference). EPA is working with the instrument
27 manufacturers to address design issues that should reduce biases that have been observed among
28 methods, in preparation for another field study examining the performance of the methods.
This work is being done in consultation with the CASAC AAMM Subcommittee,
January 2005 5-63 Draft - Do Not Quote or Cite
t
-------
1
2
3
4
5
6
7
8
9
10
11
EPA has begun the process of examining data quality objectives for potential PM10_2 5
standards. On the basis of preliminary analyses, it is apparent that greater sampling frequency
will be important due to the high variability of PM10.2 s in Ihe atmosphere; this would be
particularly important for a short-term PM10.2 5' standard. Due to the resource intensive nature of
filter sampling on a daily basis, staff believes that it will be critical to include continuous
monitoring in any network deployment strategy for a possible PM10.2 s standard. In addition to
providing high temporal resolution to PMj0.2 5 data, continuous monitors would also support
public reporting of PM10_2 5 episodes and inclusion of PM10_2 5 in an air quality forecasting
program. As noted above and elsewhere in this document, PM10.25 is more highly variable in the
atmosphere than PM2 5, such that spatial robustness will be a particularly important consideration
in monitoring network design.
12 5.4.3 Averaging Times
13 In the last review, EPA retained both annual and 24-hour standards to provide protection
14 against the known and potential effects of short- and long-term exposures to thoracic coarse
15 particles (62 FR at'38,677-79). This decision was based in part on qualitative considerations
16 related to the expectation that deposition of thoracic coarse particles in the respiratory system
17 could aggravate effects in individuals with asthma. In addition, quantitative support came from
18 limited epidemiologic evidence suggesting that aggravation of asthma and respiratory infection
19 and symptoms may be associated with daily or episodic increases in PM10, where dominated by
20 thoracic coarse particles including fugitive dust. Further, potential build-up of insoluble thoracic
21 coarse particles in the lung after long-term exposures to high levels was also considered
22 plausible.
23 Information available in this review on thoracic coarse particles, while still limited,
24 represents a significant expansion of the evidence available in the last review. As discussed '
25 above in section 5.4.1, a number of epidemiologic studies are now available that report
26 statistically significant associations between short-term (24-hour) exposure to PMi0.2 s and
27 morbidity effects, which the CD concludes are suggestive of causal associations, and mortality,
28 which the CD concludes provide less support for possible causal associations. With regard to
January 2005
5-64
Draft - Do Not Quote or Cite
-------
1 long-term exposure studies, while one recent study reported a link between reduced lung
2 function growth and long-term exposure to PM,0.2 5 and PM2.5? the CD concludes that the
3 . evidence is not sufficient to be suggestive of a causal association. Staff notes that no evidence is
4 available to suggest associations between PM10_2 5 and very short exposure periods of one or
5 more hours.
6 Based on these considerations, staff concludes that the newly available evidence provides
7 support for considering a 24-hour standard for control of thoracic coarse particles, based
8 primarily on evidence suggestive of associations between short-term exposure and morbidity
9 effects, reflecting as well the potential for associations with mortality. Noting the absence of
10 evidence judged to be even suggestive of an association with long-term exposures, staff
11 concludes that there is little support for an annual standard, although staff recognizes that it may
12 be appropriate to consider an annual standard to provide a margin of safety against possible
13 effects related to long-term exposure to thoracic coarse particles that future research may reveal.
14 Staff observes, however, that a 24-hour standard that would reduce 24-hour exposures would
15 also likely reduce long-term average exposures, thus providing some margin of safety against the
16 possibility of health effects associations with long-term exposures.
17 5.4.4 Alternative PM10.2.S Standards to Address Health Effects Related to Short-term
18 Exposure
19 In the last review, EPA's decision to retain the level of the 24-hour PMIO standard of 150
20 ug/mf (with revision of the form of the standard) was based on two community studies of
21 exposure to fugitive dust that showed health effects only in areas experiencing large exceedances
22 of that standard, as well as on qualitative information regarding the potential for health effects
23 related to short-term exposure to thoracic coarse particles. Because of the very limited nature of
24 this evidence, staff concluded mat while it supported retention of a standard to control thoracic
25 coarse particles, it provided no basis for considering a more protective standard. However,
26 because of concerns about the expected-exceedance-based form of the 1987 PM!0 standard,
27 primarily related to the stability of the attainment status of an area overtime and complex data
28 handling conventions needed in conjunction with less-than-every-day sampling, EPA adopted a
January 2005
5-65
Draft - Do Not Quote or Cite
-------
1 concentration-based form for the 24-hour standard, as was done for the 24-hour PM2 5 standard,
2 as discussed above in section 5.3.6. In making this change, EPA selected a 99th percentile form,9
3 in contrast to the 98th percentile form adopted for the 24-hour PM2 5 standard, so as not to allow
4 any relaxation in the level of protection that had been afforded by the previous 1-expected-
5 exceedance form. •
6 Since the last review, as discussed above in section 5.4.1, new evidence specific to
7 PM10.2 5 has become available mat suggests associations between short-term PM10.2 5
8 concentrations and morbidity effects and, to a lesser degree,- mortality. In considering this
9 evidence as a basis for setting a 24-hour PM10_2 5 standard, staff has focused on U.S. and
10 Canadian short-term exposure studies of thoracic coarse particles (Appendix 3 A). In so doing,
11 staff has taken into account reanalyses that addressed GAM-related statistical issues and has
12 considered the extent to which the studies report statistically significant and relatively precise
13 relative risk estimates; the reported associations are robust to co-pollutant confounding and
14 alternative modeling approaches; and the studies used relatively reliable air quality data. In
15 particular, staff has focused first on those specific morbidity studies that provide evidence of
16 associations in areas that would have met the current PM10 standards during the time of the
17 study.
18 As an initial matter, staff recognizes, as discussed in Chapter 3 (section 3.6.6), that these
19 short-term exposure studies provide no evidence of clear thresholds, or lowest-observed-effects
20 levels, in terms of 24-hour average concentrations. Staff notes that in the one study that explored
21 a potential PM10.2 5 threshold, conducted in Phoenix, no evidence of a threshold was observed for
22 PM10.2 5, even though that study provided some suggestion of a potential threshold for PM2 5. The
23 CD concludes that while' there is no evidence of a clear threshold within the range of air quality
24 observed in the studies, for some health endpoints (such as total nonaccidental mortality) it is
25 likely to be extremely difficult to detect threshold levels (CD, p. 9-45). As a consequence, mis
26 body of evidence is difficult to translate directly into a specific 24-hour standard that would
27 protect against all effects associated with short-term exposures. Staff notes that the distributions
9 As noted above, the court vacated the 1997 24-hour PMlo standard that had been revised to incorporate a
1 99th percentile form.
January 2005
5-66
Draft - Do Not Quote or Cite
-------
1 of daily PM10.25 concentrations in these studies often extend down to or below background
2 levels, such that the likely range of background concentrations across the U.S., as discussed in
3 Chapter 2, section 2.6, could be a relevant consideration in this policy evaluation. Staff
4 recognizes, however, that there is insufficient data to estimate daily distributions of background
5 PMi0.25 levels (as was done for background PM25 levels, as discussed in Chapter 2, section 2.6).
6 Being mindful of the difficulties posed by uncertainties related to potential thresholds and
7 insufficient data to characterize daily distributions of PM10.2 5 background concentrations, as well
8 as the limited nature of the available evidence, staff has considered the short-term exposure
9 epidemiologic evidence as a basis for alternative 24-hour PM10.2 5 standards. In so doing, staff
10 has focused on the upper end of the distributions of daily PM10.2.5 concentrations, particularly in
11 terms of the 98th and 99th percentile values, consistent with the forms considered in section 5.3.6
12 above for PM25. In looking at the specific morbidity studies identified in section 5.4.1 that
13 report statistically significant associations with respiratory- and cardiac-related hospital
14 admissions in areas that had ambient air quality levels that would have met the current PM10
15 standards at the time of the study, including studies in Toronto (Burnett et al., 1997), Seattle
16 (Sheppard et al.,1999, 2003), and Detroit (Lippmann et al., 2000; Ito, 2003), staff notes that the
17 reported 98th percentile values range from approximately 30 to 36 ug/m3 in all three areas, and
18 the 99th percentile values range from 36 to 40 ug/m3 (Ross and Langstaff, 2005).
19 In looking more closely at these studies, staff recognizes that the uncertainty related to
20 exposure measurement error is potentially quite large in epidemiologic studies linking effects to
21 PM10.25 air quality measures. For example, in looking specifically at the Detroit study, staff
22 notes that the PM,0.2 5 air quality values were based on air quality monitors located in Windsor,
23 Canada The study authors determined that the air quality values from these monitors were
24 generally well correlated with air quality values monitored in Detroit, where the hospital
25 admissions data were gathered, and, thus concluded that these monitors were appropriate for use
26 in exploring the association between air quality and hospital admissions in Detroit. Staff has
27 observed, however, mat the PM10.2 5 levels reported in this study are significantly lower than the
28 PMto.25 levels measured at some of the Detroit monitors in 2003 - an annual mean level of 13.3
29 ug/m3 is reported in the study, based on 1992 to 1994 data, as compared to an average annual
January 2005
5-67
Draft - Do Not Quote or Cite
-------
1 mean level of 21.7 ug/m3 measured at two urban-center monitors in 2003 (which is used as the
2 basis for the risk assessment presented in Chapter'4). This observation prompted staff to further
3 explore the comparison between PM10.2 5 levels monitored at Detroit arid Windsor sites.' This
4 exploration has shown that in'fecent years, based on available Windsor and Detroit data from
5 1999 to 2003, the Windsor monitors used in this study typically have recorded PM1(W5 levels that
6 are generally less man half the levels recorded at urban-center Detroit monitors, though the
7 concentrations measured in Windsor are more similar to concentrations reported for suburban
8 areas well outside the city (Ross and Langstaff, 2005). These observations lead staff to conclude
9 that the statistically significant, generally robust association with hospital admissions in Detroit
10 reflects population exposures that may be appreciably higher than what would be estimated
11 using data from the Windsor monitors. Taking these observations into account, staff nonetheless
12 believes that these studies in general, and the Detroit study in particular, do provide evidence of
13 associations between short-term exposures to PM10.25 and hospital admissions. Staff does
14 conclude, however, that the association observed in the Detroit study, which' staff judges to be
15 the strongest of these studies, likely reflects exposure levels potentially much higher in the
16 central city area than those reported in that study. Based on this information, staff believes that
17 alternative 24-hour PM]0_2 5 standards appropriate for consideration in this review need not ;
18 necessarily extend to levels down to or below the ranges reported in these studies in order to'
19 provide protection from the morbidity effects related to short-term exposures to PM,0.2 5.
20 Staff has also looked at the evidence from U.S. and Canadian studies thai report
21 statistically significant and generally robust associations with mortality and short-term exposures
22 to PM10.25. As discussed in section 9.2.3 of the CD, the evidence associating mortality with
23 short-term exposures to PM10_25 is too uncertain to infer a likely causal relationship, although it
24 is suggestive of a possible causal relationship. Staff identified two such studies, conducted in
25 Phoenix (Mar et al, 2000,2003) and Coachella Valley, CA (Ostro et al., 2000,2003), that report
26 98th percentile PM10_25 values of approximately 70 and 107 ug/m3, and 99th percentile values of
27 75 and 134 ng/m3, respectively. Staff notes that these studies were conducted in areas with air
28 quality levels that would not have met the current PM10 standards. A staff analysis of PM10 and
29 estimated PM10.25 concentrations from the AQS database for 2001 to 2003 suggests that 98th
January 2005
5-68
Draft - Do Not Quote or Cite
-------
s
S «
If
ss
Jg
-1
"
I "
c •?
8 s
g g
fli PH
1!
s
S O
§3
•3"
12
•s
l.-I
o «
eo
•5
•g
CD
*&•
\M
8
R
Jj
€ ^
1l
« J
!l
a
8
a
s
s
S
s
^
s
s
ffi
s
s
88
•s &.
and
ilia
u
a
o
O)
I
I
S
cn
-------
1 Beyond looking direcHy at the relevant epidemiologic evidence, staff has also considered
2 the extent to which the PM,0.2 5 risk assessment results discussed in Chapter 4 can help inform
3 consideration of alternative 24-hour PM10_2.5 standards. While one of the goals of the PMi0.2 s
4 risk assessment was to provide estimates of the risk reductions associated with just meeting
5 alternative PM10.2 5 standards, staff has concluded that the nature and magnitude of the
6 uncertainties and concerns associated with this portion of the risk assessment weigh against use
7 of these risk estimates as a basis for recommending specific standard levels. These uncertainties
8 and concerns include, but are not limited to the following:
9 • as discussed above, concerns that the current PM10.2 5 levels measured at ambient
10 monitoring sites in recent years may be quite different from the levels used to
11 characterize exposure in the original epidemiologic studies based on monitoring sites in
12 different location, thus possibly over- or underestimating population risk levels; "
13 • greater uncertainty about the reasonableness of the use of proportional rollback to
14 simulate attainment of alternative PM10.2.s daily standards in any urban area due to the
15 limited availability of PM10.2 5 air quality data over time;
16 • concerns that the locations used in the risk assessment are not representative of urban
17 areas in the U.S. that experience the most significant 24-hour peak PM10.2.5
18 concentrations, and thus, observations about relative risk reductions associated with
19 alternative standards may not be relevant to the areas expected to have the greatest health
20 'risks associated with elevated ambient PM10L25 levels; and
21 • concerns about the much smaller health effects database that supplies the C-R
22 relationships used in the risk assessment, compared to that available for PM2 5, which
23 limits our ability to evaluate the robustness of the risk estimates for the same health
24 endpoints across different locations.
25 In summary, in considering the relevant epidemiologic evidence and the related
26 limitations and uncertainties, staff concludes that there is support for considering a 24-hour
27 PM10.2 5 standard to replace the current PM10 standards to provide protection against health
28 effects associated with short-term exposures to thoracic coarse particles. In looking primarily at
29 the evidence of associations between short-term exposure to PM10.2.5 and mortality, staff
30 concludes that it is appropriate to consider a 24-hour standard in the range of 65 to 75 ug/m3,
31 with a 98th percentile form, or in the range of 75 to 85 ug/m3, with a 99th percentile form. A
January 2005 5-72 Draft ~ Do Not Quote or Cite
-------
1 standard set within either of these ranges could be expected to provide a margin of safety to
2 protect against the potential, but uncertain, mortality effects of PM,0.2 5, while continuing to
3 provide protection against the effects of PM10.2 5 associated with high levels of PM10 that were the
4 basis for the decision made by EPA in 1997 to retain the levels of the PMj0 standards. In
5 addition, staff observes that several epidemiologic studies have reported associations with
6 morbidity effects in areas with lower PM10.2 5 that could support consideration of standard levels
7 as low as approximately 30 ug/m3,98th percehtile, or 35 ug/m3, 99th percentile.
8 Staff recognizes, however, that the epidemiologic evidence on morbidity and mortality
9 effects related to PM10.25 exposure is very limited at this time. A key area of uncertainty in this
10 evidence is the potentially quite large uncertainty related to exposure measurement error for
11 PM10.2.5, as compared with fine particles. PM10.2 5 concentrations can vary substantially across a
12 metropolitan area and thoracic coarse particles are less able to penetrate into buildings than fine
13 particles; thus, the ambient concentrations reported in epidemiologic studies may not well
14 represent area-wide population exposure levels. Other key uncertainties include the lack of
15 information on the composition of thoracic coarse particles and the effects of thoracic coarse
16 particles from various sources, and the lack of evidence on potential mechanisms for effects of
17 thoracic coarse particles. Staff believes that taking these uncertainties into account leads to
18 consideration of standard levels toward the upper end of the ranges identified above.
19 5.4.5 Summary of Staff Recommendations on Primary PMto.2 s NAAQS
20 Staff recommendations for the Administrator's consideration in making decisions on
21 standards for thoracic coarse particles, together with supporting conclusions from sections 5.4.1
22 through 5.4.4, are briefly summarized below. In making these recommendations, staff is mindful
23 that the Act requires standards to be set that are requisite to protect public health with an
24 adequate margin of safety, such that the standards are to be neither more nor less stringent than
25 necessary. Thus, the Act does not require that NAAQS be set at zero-risk levels, but rather at
26 levels that avoid unacceptable risks to public health.
January 2005
5-73
Draft - Do Not Quote or Cite
-------
1 (1) The current primary PM10 standards should be revised in part by replacing the PM10
2 indicator with an indicator of thoracic coarse particles that does not include fine particles.
3 Any such revised standards should be based on available health effects evidence and air
4 quality data generally indexed by PMJO_2 5, to provide public health protection more
5 specifically directed toward effects related to exposure to thoracic coarse particles in the
6 ambient air.
7 (2) The indicator for a thoracic coarse particle standard should be PM ]0_2 5, consistent with
8 the recommendation made in section 5.3.7 to retain PM25 as the indicator for fine particle
9 standards.
10 ( a) As noted above, this recommendation is based primarily on the evaluation in the
1 1 CD of air quality within the intermodal particle size range of 1 to 3 um and the
12 . policy judgment made in the last review to place regulatory importance on
»
13 - . defining an indicator that would more completely capture fine particles under all
14 , conditions likely to be encountered across the U.S., while.recognizing that some
15 limited intrusion of small coarse particles will occur in some circumstances.
16 ( b) In support of this recommendation, work should continue on the development of
17 an FRM for PM10.2 5 based on the ongoing fiel'd program to evaluate various types
18 of monitors, and consideration should be given to the adoption of FEMs for
19 appropriate continuous measurement methods.
20 ( 3) A 24-hour averaging time should be retained for a PM^s standard to protect against
21 health effects associated with short-term exposure periods, with consideration given to
22 the use of either a 98th or 99th percentile form. Consideration could also be given to
23 retaining an annual averaging time, in considering the appropriate margin of safety
24 against possible health effects that might be associated with long-term exposure periods.
25 ( 4) Consideration should be given to setting a 24-hour PM10.2 5 standard about as protective
26 as the current daily PM10 standard, with a level in the range of approximately 65 to 75
January 2005 5-74 Draft - Do Not Quote or Cite
-------
1 Hg/m3, 98* percentile, or approximately 75 to 85 ng/rn3, 99th percentile. Staff also
2 believes there is some support for consideration of a PM10.2 5 standard level down to
3 approximately 30 ug/m3, 98th percentile, or 35 ng/m3,99th percentile, while recognizing
4 that a standard set at such a relatively low level would place a great deal of weight on
5 very limited and uncertain epidemiologic associations. Consideration of PM10.25
6 standards within Ihe ranges recommended above, and design considerations for an
7 associated PM10.2 5 monitoring network, should take into account the especially large
8 variability seen in currently available information on ambient concentrations and
9 composition of PM10.25.
10 5.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
11 RECOMMENDATIONS RELATED TO SETTING PRIMARY PM STANDARDS
12 Staff believes it is important to continue to highlight the unusually large uncertainties
13 associated with establishing standards for PM relative to other single component pollutants for
14 which NAAQS have been set. Key uncertainties and staff research recommendations on health-
15 related topics are outlined below. In some cases, research in these areas can go beyond aiding in
16 standard setting to aiding in the development of more efficient and effective control strategies.
17 Staff notes, however, that a full set of research recommendations to meet standards
18 implementation and strategy development needs is beyond the scope of this discussion.
19 The 1996 PM Staff Paper included a discussion of uncertainties and research
20 recommendations (EPA, 1996b, pp. VII-41-44) that addressed the following issues related to
21 understanding health effects associated with exposure to PM:
22 • lack of demonstrated biological mechanisms for PM-related effects, •
23 • potential influence of measurement error and exposure error,
24 • potential confounding by copollutants,
25 • evaluation of the effects of components or characteristics of particles,
26 • the shape of concentration-response relationships,
27 • methodological uncertainties in epidemiological analyses,
28 • the extent of life span shortening,
January 2005 % 5-75 Draft - Do Not Quote or Cite
-------
1
2
characterization of annual'and daily background concentrations, and
understanding of the effects of coarse fraction particles.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
As has been discussed in depth in the CD, especially in Chapters 5 through 8, an
extensive body of new studies related to understanding health effects associated with exposure to
PM is now available that provides important information on many of the topics listed above. For
example, regarding the lack of demonstrated biological mechanisms, new evidence from
toxicologic and controlled human exposure studies has provided information on an array of
potential mechanisms for effects on the cardiac and respiratory systems, as discussed in Chapters
7 and 9 of the CD. Still, the CD emphasizes that much remains to be learned to fully understand
the pathways or mechanisms by which PM is linked with different health endpoints. For each of
> *
the issues listed above, new evidence has become available that helps to reduce uncertainties,
although uncertainty has been reduced in some areas more than others. Staff has identified the
following key uncertainties and research questions that have been highlighted in this review of
the health-based primary standards
(1)
The body of evidence on effects of thoracic coarse particles has been expanded, but the
uncertainties regarding thoracic coarse particles are still much greater than those for fine
particles. As discussed in Chapter 2, the spatial variability of thoracic coarse particles is
generally greater than that for fine particles, which will increase uncertainty in the
associations between health effects and thoracic coarse particles measured at central site
monitors. Additional exposure research is needed to understand the influence of
measurement error and exposure error on thoracic coarse particle epidemiology results.
In addition, little is known about coarse particle composition, and less about the health
effects associated with individual components or sources of thoracic coarse particles, but
it is possible mat there are components of thoracic coarse particles (e.g., crustal material)
that are less likely to have adverse effects, at least at lower concentrations, than other
components.
January 2005
5-76
Draft - Do Not Quote or Cite
-------
1 (2) Identification of specific components, properties, and sources of fine particles that are
2 linked with health effects remains an important research need. Available evidence
3 provides no basis for expecting that any one component would be solely responsible for
4 PM2 5-related effects, but it is likely that some components are more closely linked with
5 specific effects than others. Continued source characterization, exposure,
6 epidemiological, and toxicological research is needed to help identify components,
7 characteristics, or sources of particles that may be more closely linked with various
8 specific effects to aid in our understanding of causal agents and in the development of
9 efficient and effective control strategies for reducing health risks. Conducting human
10 . exposure research in parallel with such health studies will help reduce the uncertainty
11 associated with interpreting health studies and provide a stronger basis for drawing
12 conclusions regarding observed effects.
13 (3) An important aspect in characterizing risk and making decisions regarding air quality
14 standard levels is the shape of concentration-response functions for PM, including
15 identification of potential threshold levels. Recent studies continue to show no evidence
16 for a clear threshold level in relationships between various PM indicators and mortality,
17 within the range of concentrations observed in the studies, though some studies have
18 suggested potential levels.
19 (4) The relationship between PM and other air pollutants in causing health effects remains an
20 important question in reducing public health risk from air pollution. Numerous new
21 analyses have indicated that associations found between PM and adverse health effects
22 are not simply reflecting actual associations with some other pollutant. However, effects
23 have been found with the gaseous co-pollutants, and it is possible that pollutants may
24 interact or modify effects of one another. Further understanding of the sources,
25 exposures, and effects of PM and other air pollutants can assist in the design of effective
26 strategies for public health protection.
27 (5) Methodological issues in epidemiology studies were discussed at length in the previous
28 review, and it appeared at the time that the epidemiology study results were not greatly
29 affected by selection of differing statistical approaches or methods of controlling for
January 2005
5-77
Draft - Do Not Quote or Cite
-------
1 other variables, such as weather. However, investigation of recently discovered
2 questions on the use of generalized additive models in time-series epidemiology studies
3 has again raised model specification issues. While reanalyses of studies using different
4 modeling approaches generally did not result in substantial differences in model results,
5 some studies showed marked sensitivity of the PM effect estimate to different methods of
6 adjusting for weather variables. There remains a need for further study on the selection
7 of appropriate modeling strategies and appropriate methods to control for time-varying
8 factors, such as temperature.
9 (6) Selection of appropriate averaging times for PM air quality standards is important for
10 public health protection, and available information suggests that some effects, including
11 cardiac-related risk factors, may be linked to exposures of very short duration (e.g., one
12 or more hours). Data on effects linked with such peak exposures, such as those related to
13 wildfires, agricultural burning, or other episodic events, would be an important aid to
14 public health response and communication programs. Investigation into the PM exposure
15 time periods that are linked with effects will provide valuable information both for the
16 standard-setting process and for risk communication and management efforts.
17 (7) There remain significant uncertainties in the characterization of annual and daily
18 background concentrations for fine particles and especially for thoracic coarse particles.
19 Further analyses of air quality monitoring and modeling that improved these background
20 characterizations would help reduce uncertainties in estimating health risks relevant for
21 . standard setting (i.e., those risks associated with exposure to PM in excess of background
22 levels) and would aid in the development and implementation of associated control
23 programs.
January 2005 5-78 Draft - Do Not Quote or Cite
-------
REFERENCES
2 Burnett, R. T,; Cakmak, S.; Brook, J. R; Krewski, D. (1 997) The role of particulate size and chemistry in the
3 association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases.
4 Environ. Health Perspect. 105:614-620.
5 Burnett, R. T.; Brook, J.; Dann, T.; Delocla, C.; Philips, O.; Cakmak, S.; Vincent, R.; Goldberg, M. S.; Krewski, D.
6 (2000) Association between particulate- and gas-phase components of urban air pollution and daily
7 mortality in eight Canadian cities. Inhalation Toxicol. 1 2(suppl. 4): 15-39.
8 Burnett, R. T.; Goldberg, M. S. (2003) Size-fractionated particulate mass and daily mortality in eight Canadian
9 cities. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston, MA:
10 . Health Effects Institute; pp. 85-90. Available: http://www.healtheffects.org/news.htm [16 May, 2003].
1 1 EPA. (1 996) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC: National Center for
1 2 Environmental Assessment-RTP Office; report no. EPA/600/P-95/001 aF-cF. 3v.
1 3 EPA. (2004) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC : National Center for
1 4 Environmental Assessment-RTP Office; report no. EPA/600/P-99/002aD.
1 5 Fairley, D. (1 999) Daily mortality and air pollution in Santa Clara County, California: 1 989-1 996. Environ. Health
16 Perspect. 107:637-641.
1 7 Fairley, D. (2003) Mortality and air pollution for Santa Clara County, California, 1 989-1 996. In: Revised analyses of
1 8 time-series studies of air pollution and health. Special report. Boston, MA: Health Effects Institute; pp.
19 97-106. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf [18 October, 2004]. '
20 Gauderman, W. J.; McConnell, R; Gilliland, P.; London, S.; Thomas, D.; Avol, E.; Vora, H.; Berhane, K.;
21 Rappaport, E. B.; Lurmann, F.; Margolis, H. G.; Peters, J. (2000) Association between air pollution and
22 lung function growth in southern California children. Am. J. Respir. Crit. Care Med. 162: 1383-1390.
23 Gauderman, W. J.; Gilliland, G. F.; Vora, H.; Avol, E.; Stram, D.; McConnell, R.; Thomas, D.; Lurmann, F.;
24 Margolis, H. G.; Rappaport, E. B.; Berhane, K.; Peters, J. M. (2002) Association between air pollution and
25 lung function growth in southern California children: results from a second cohort. Am. J. Respir. Crit. Care
26 Med. 166: 76-84.
27 Ito, K. (2003) Associations of particulate matter components with daily mortality and morbidity in Detroit,
28 Michigan. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
29 MA: Health Effects Institute; pp. 143-1 56. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
30 [12 May, 2004].
3 1 Langstaff, J. (2004). Estimation of Policy- Relevant Background Concentrations of Particulate Matter. Memorandum
32 to PM NAAQS review docket OAR-2001-0017. January 27, 2005.
33 Lippmann, M.; Ito, K.; Nadas, A.; Burnett, R. T. (2000) Association of particulate matter components with daily
34 mortality and morbidity in urban populations. Cambridge, MA: Health Effects Institute; research report 95.
35 Mar, T. F.; Norris, G. A.; Koenig, J. Q.; Larson, T. V. (2000) Associations between air pollution and mortality in
36 Phoenix, 1995-1997. Environ. Health Perspect. 108:347-353.
January 2005 5-79 Draft - Do Not Quote or Cite
-------
1 Mat, T. F.; Morris, 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:/Avww.healtheffects.org/Pubs/rimeSeries.pdf [18 October, 2004].
5 Ostro, B, D.; Hurley, S.; Lipsett, M. J. (1999) Air pollution and daily mortality in the Coacheila Valley, California:
6 a study of PM10 dominated by coarse particles. Environ. Res. 81: 231-238.
7 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2000) Coarse and fine particles and daily mortality in the Coacheila
8 Valley, CA: a follow-up study. J. Exposure Anal. Environ. Epidemiol. 10:412-419.
• <
9 Ostro, B. D.; Broadwin, R.; Lipsett, M. J. (2003) Coarse particles and daily mortality in Coacheila Valley,
10 California. In: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
11 MA: Health Effects Institute; pp. 199-204. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
12 [18 October, 2004].
13 Peters, J. M.; Avol, E.; Navidi, W.; London, S. J.; Qauderman, W. J.; Lurmann, F.; Linn, W. S.; Margolis, H.;
14 ' Rappaport, E.; Gong, H., Jr.; Thomas, D. C. (1999) A study of twelve southern California communities
15 with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am. J. Respir. Crit.
16 Care Med. 159:760-767.
17 Pope, C. A., Ill; Burnett, R. T.; Thun, M. J.; Calle, E. E.; Krewski, D.; Ito, K.; Thurston, G. D. (2002) Lung cancer,
18 cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. J. Am. Med. Assoc.
19 287:1132-1141.
20 Ross, M.; Langstaff, J. (2005) Updated statistical information on air quality data from epidemiologic studies.
21 Memorandum to PMNAAQS review docket OAR-2001-0017. January 31, 2005.
22 Schmidt et al., (2005) Draft analysis of PM ambient air quality data for the PM NAAQS review. Memorandum to
23 PMNAAQS review docket OAR-2001-0017. January 31, 2005.
24 Schwartz, J.; Dockery, D. W.; Neas, L. M. (1996a) Is daily mortality associated specifically with fine particles? J.
25 Air Waste Manage. Assoc. 46:927-939.
26 Schwartz, J.; Neas, L. M. (2000) Fine particles are more strongly associated than coarse particles with acute
27 respiratory health effects in schoolchildren. Epidemiology 11:6-10.
28 Sheppard, L.; Levy, D.; Norris, G.; Larson, T. V.; Koenig, J. Q. (1999) Effects'of ambient air pollution on
29 nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology 10: 23-30.
30 Sheppard, L. (2003) Ambient air pollution and nonelderly asthma hospital admissions in Seattle, Washington,
31 1987-1994. Iri: Revised analyses of time-series studies of air pollution and health. Special report. Boston,
32 MA: Health Effects Institute; pp. 227-230. Available: http://www.healtheffects.org/Pubs/TimeSeries.pdf
33 [18 October, 2004].
January 2005 ' 5-80 Draft - Do Not Quote or Cite
-------
1 6. POLICY-RELEVANT ASSESSMENT OF PM-RELATED WELFARE EFFECTS
2 .
3 6.1 INTRODUCTION
4 This chapter assesses key policy-relevant information on the known and potential effects
5 on public welfare associated with ambient PM, alone and in combination with other pollutants
6 commonly present in the ambient air, drawing upon the most relevant information contained in
7 the CD and other significant reports referenced therein. The welfare effects to be considered in
8 this review of the secondary PM NAAQS include effects on visibility (section 6.2), vegetation
9 and ecosystems (section 6.3), materials (section 6.4), and climate change processes1 (section
10 6.5). For each category of effects, this chapter presents a summary of the relevant scientific
11 information and a staff assessment of whether the available information is sufficient to be •
12 considered as the basis for secondary standards distinct from primary standards for PM. Staff
13 conclusions and recommendations related to secondary standards for PM are presented in
14 Chapter 7.
15 It is important to note that discussion of PM-related effects on visibility, vegetation and
16 ecosystems, and climate change processes in Chapters 4 and 9 of the CD builds upon and
17 includes by reference extensive information from several other significant scientific reviews of
18 these topics. Most notably, these reports include the Recommendations of the Grand Canyon
19 Visibility Transport Commission (1996), the National Research Council's Protecting Visibility
20 in National Parks and Wilderness Areas (1993), reports of the National Acid Precipitation
21 Assessment Program (1991,1998), previous EPA Criteria Documents, including^//- Quality
22 Criteria for Particulate Matter and Sulfur Oxides (EPA, 1982) and Air Quality Criteria for
23 Oxides of Nitrogen (EPA, 1993), recent reports of the National Academy of Sciences (NAS,
24 2001) and the Intergovernmental Panel on Climate Change (IPCC, 1998, 2001a,b), and
25 numerous other U.S. and international assessments of stratospheric ozone depletion and global
26 climate change carried out under U.S. Federal interagency programs (e.g., the U.S. Global
27 Climate Change Research Program), the World Meteorological Organization (WMO), and the
28 United Nations Environment Programme (UNEP).
In assessing information on PM-related effects on climate change processes, consideration is given to
potential indirect impacts on human health and the environment that may be a consequence of changes in climate
and solar radiation attributable to changes in ambient PM.
January 2005 6-1 Draft - Do Not Quote or Cite
-------
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 .
6.2 EFFECTS ON VISIBILITY
Visibility can be defined as the degree to which the atmosphere is transparent to visible
light (NRCi, 1993; CD, 4-153). Visibility impairment is the most noticeable effect of fine
particles present in the atmosphere. Particle pollution degrades the visual appearance and
perceived color of distant objects to an observer and reduces the range at which they can be
distinguished from the background.
This section discusses the role of ambient PM in the impairment of visibility, drawing
upon the most relevant information contained in the CD (Chapters 4 and 9), as well as significant
reports on the science of visibility referenced therein, and building upon information presented in
section 2.8 of this document. In particular, this section includes new information on the
following topics:
• Summary findings of analyses of hourly PM25 measurements and reconstructed light
extinction coefficients for urban areas, for 2003, that demonstrate a significant
correlation between PM25 and light extinction across the U.S. during daylight hours.
• An overview of visibility programs, goals, and methods for the evaluation of visibility
impairment as a basis for standard setting, in the U.S. and abroad, illustrating Ihe
significant value placed on visual air quality, as demonstrated by efforts to improve \
visibility in national parks and wilderness areas, as well as in urban areas.
This section summarizes available information as follows: (1) information on the general
types of visibility impairment; (2) trends and conditions in Class I and non-urban areas; (3)
visibility conditions in urban areas; (4) studies of the economic value of improving visual air
quality; (5) current policy approaches to addressing visibility impairment; and (6) approaches to
evaluating public perceptions of visibility impairment and judgments about the acceptability of
varying degrees of visibility impairment.
6.2.1 Overview of Visibility Impairment
Visibility effects are manifested in two principal ways: as local impairment (e.g.,
localized hazes and plumes) and as regional haze. This distinction is significant because this
difference impacts both how visibility goals may be set and how air quality management
strategies may be devised.
January 2005
6-2
Draft - Do Not Quote or Cite
-------
1 Local-scale visibility degradation commonly occurs either in the form of a plume
2 resulting from the emissions of a specific source or small group of sources, or in the form of a
3 localized haze, such as an urban "brown cloud." Visibility impairment caused by a specific
4 source or small group of sources has been generally termed "reasonably attributable"
5 impairment. Plumes are comprised of smoke, dust, or colored gas that obscure the sky or
6 horizon relatively near sources. Sources of locally visible plumes, such as the plume from an
7 industrial facility or a burning field, are often easy to identify. There have been a limited number
8 of cases in which Federal land managers have certified the existence of visibility impairment in a
9 Class I area (i.e., 156 national parks, wilderness areas, and international parks identified for
10 visibility protection in section 162(a) of the Act) as being "reasonably attributable" to a
11 particular source.2
12 The second type of impairment, regional haze, results from pollutant emissions from a
13 multitude of source's located across a broad geographic region. Regional haze impairs visibility
14 in every direction over a large area, in some cases over multi-state regions. It also masks objects
15 on the horizon and reduces the contrast of nearby objects. The formation, extent, and intensity
16 of regional haze is a function of meteorological and chemical processes, which sometimes cause
17 fine particle loadings to remain suspended in the atmosphere for several days and to be
18 transported hundreds of kilometers from their sources (NRC, 1993). It is this second type of
19 visibility degradation, regional haze, that is principally responsible for impairment in national
20 parks and wilderness areas across the country (NRC, 1993).
21 While visibility impairment in urban areas at times may be dominated by local sources, it
22 often may be significantly affected by long-range transport of haze due to the multi-day
23 residence times of fine particles in the atmosphere. Fine particles transported from urban and
24 industrialized areas, in turn, may be significant contributors to regional-scale impairment in
25 Class I and other rural areas.
. 2Two 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.
January 2005
6-3
Draft - Do Not Quote or Cite
-------
1 6.2.2 Visibility Trends and Current Conditions in Class I and Non-Urban Areas
2 In conjunction with the National Park Service, other Federal land managers, and State
3 organizations, EPA has supported visibility monitoring in national parks and wilderness areas
4 since 1988. The monitoring network was originally established at 20 sites, but it has now been
5 expanded to 110 sites that represent all but one (Bering Sea) of the 156 mandatory Federal Class
6 I areas across the country. This long-term visibility monitoring network is known as IMPROVE
7 (Interagency Monitoring of PROtected Visual Environments).
8 IMPROVE provides direct measurement of fine particles and precursors that contribute
9 to visibility impairment The IMPROVE network employs aerosol measurements at all sites, and
10 optical and scene measurements at some of the sites. Aerosol measurements are taken for PM10
11 and PM2 5 mass, and for key constituents of PM2 *5, such as sulfate, nitrate, organic and elemental
12 carbon, soil dust, and several other elements. Measurements for specific aerosol constituents are
13 used to calculate "reconstructed" aerosol light extinction by multiplying the mass for each
14 constituent by its empirically-derived scattering and/or absorption efficiency, with adjustment
15 for the relative humidity. Knowledge of the main constituents of a site's light extinction
16 "budget" is critical for source apportionment and control strategy development Optical
17 measurements are used to directly measure light extinction or its components. Such
18 measurements are taken principally with either a transmissometer, which measures total light
19 extinction, or a nephelometer, which measures particle scattering (the largest human-caused
20 component of total extinction). Scene characteristics are typically recorded 3 times daily with 35
21 millimeter photography and are used to determine the quality of visibility conditions (such as
22 effects on color and contrast) associated with specific levels of light extinction as measured
23 under both direct and aerosol-related methods. Directly measured light extinction is used under
24 the IMPROVE protocol to cross-check that the aerosol-derived light extinction levels are
25 reasonable in establishing current visibility conditions. Aerosol-derived light extinction is used
26 to document spatial and temporal trends and to determine how proposed changes in atmospheric
27 constituents would affect future visibility conditions.
28 Annual average visibility conditions (reflecting light extinction due to both
29 anthropogenic and non-anthropogenic sources) vary regionally across the U.S. The rural East
January 2005
6-4
Draft - Do Not Quote or Cite
-------
1 generally has higher levels of impairment than remote sites in the West, with the exception of
2 urban-influenced sites such as San Gorgonio Wilderness (CA) and Point Reyes National
3 Seashore (CA), which have annual average levels comparable to certain sites in the Northeast.
4 Regional differences are illustrated by Figures 4-3 9a and 4-39b in the CD, which show that, for
5 Class I areas, visibility levels on the 20% haziest days in the West are about equal to levels on
6 the 20% best days in the East (CD, p 4-179).
7 Higher visibility impairment levels in the East are due to generally higher concentrations
8 of anthropogenic fine particles, particularly sulfates, and higher average relative humidity levels.
9 In fact, sulfates account for 60-86% of the haziness in eastern sites (CD, 4-236). Aerosol light
10 extinction due to sulfate on the 20% haziest days is significantly larger in eastern Class I areas as
11 compared to western areas (CD, p. 4-182; Figures 4-40a and 4-40b). With the exception of
12 remote sites in the northwestern U.S., visibility is typically worse in the summer months. This is
13 particularly true in the Appalachian region, where average light extinction in the summer
14 exceeds the annual average by 40% (Sisler et al., 1996).
15 Regional trends in Class I area visibility are updated and presented in the EPA's National
16 Air Quality and Emissions Trends Report (EPA, 2001). Eastern trends for the 20% haziest days
17 from 1992-1999 showed a 1.5 deciview improvement, or about a 16% improvement. However,
'18 visibility in the East remains significantly impaired, with an average visual range of
19 approximately 20 km on the 20% haziest days. In western Class I areas, aggregate trends
20 showed little change during 1990-1999 for the 20% haziest days, and modest improvements on
21 the 20% mid-range and clearest days. Average visual range on the 20% haziest days in western
22 Class I areas is approximately 100 km.
23 .
24 6.2.3 Visibility Conditions in Urban Areas
25 Urban visibility impairment often results from the combined effect of stationary, mobile,
26 and area source emissions. Complex local meteorological conditions may contribute to such
27 impairment as well. Localized or layered haze often results from emissions from many sources
28 located across an urban or metropolitan area. A common manifestation of this type of visibility
29 impairment is the "brown cloud" situation experienced in some cities particularly in the winter
January 2005 6-5 Draft - Do Not Quote or Cite
-------
t
1 ' months, when cooler temperatures limit vertical mixing of the atmosphere. The long-range
2 transport of emissions from sources outside the urban area may also contribute to urban haze
3 levels.
4 Visibility impairment has been studied in several major cities in the past decades because
5 of concerns about fine particles and their potentially significant impacts (e.g., health-related and
6 aesthetic) on the residents of large metropolitan areas (e.g., Middletori, 1993). Urban areas•
7 generally have higher loadings of PM2 5 and, thus, higher visibility impairment than monitored
8 Class I areas. As discussed in Chapter 2, sections 2.4 and 2.5, annual meanlevels of 24-hour
9 average PM2 s levels are generally higher in urban areas than those found in the IMPROVE
10 database for rural Class I areas. Urban areas have higher concentrations of organic carbon,
11 elemental carbon, and paniculate nitrate than rural areas due to a higher density of fuel
12 combustion and diesel emissions.
13 6.2.3.1 ASOS Airport Visibility Monitoring Network
14 For many years, urban visibility has been characterized using data describing airport
15 visibility conditions. Until the mid-1990's, airport visibility was typically reported on an hourly
16 basis by human observers. An extensive database of these assessments has been maintained and
17 analyzed to characterize visibility trends from the late-1940's to mid-1990's (Schichtel et al.,
18 2001).
19 In 1992, the National Weather Service (NWS), Federal Aviation Administration (FAA),
20 and Department of Defense began deployment of the Automated Surface Observing System
21 (ASOS). ASOS is now the largest instrument-based visibility monitoring network in the U.S. •
22 (CD, p. 4-174). The ASOS visibility monitoring instrument is a forward scatter meter that has
23 been found to correlate well with light extinction measurements from the Optec transmissometer
24 (NWS, 1998). It is designed to provide consistent, real-time visibility and meteorological
25 measurements to assist with air traffic control operations. A total of 569 FAA-spbnsored and
26 313 NWS-sponsored automated observing systems are installed at airports throughout the
27 country. ASOS visibility data are typically reported for aviation use in small increments up to a
28 maximum of 10 miles visibility. While these truncated data are not ideal for characterizing
29 actual visibility levels, the raw, non-truncated data from the 1-minute light extinction and
January 2005 6-6 Draft - Do Not Quote or Cite
-------
1 meteorological readings are now archived and available for analysis for a subset of the ASOS
2 sites.3
3 6.2.3.2 Correlation between Urban Visibility and PM2S Mass
4 In an effort to better characterize urban visibility, staff has analyzed the extensive new
5 data now available on PM2 5 primarily in urban areas. This rapidly expanding national database,
6 including FRM measurements of PM25 mass, continuous measurements of hourly PM2.5 mass,
7 and PM25 chemical speciation measurements, has now provided the opportunity to conduct such
8 an analysis. In this analysis, described below and documented in detail in Schmidt et al. (2005),
9 staff has sought to explore the factors that have historically complicated efforts to address
10 visibility impairment nationally, including regional differences related to levels of primarily fine
11 particles and relative humidity. Taking these factors into account, staff has compared
12 correlations between visibility, in terms of reconstructed light extinction (using the IMPROVE
13 methodology discussed in Chapter 2, section 2.8), with hourly PM25 concentrations in urban
14 areas across the U.S. and in eastern and western regions.
15 As an initial matter, staff has explored the factors contributing to the substantial
16 East/West differences that have been characterized primarily for Class I areas across the country,
17 as discussed above in section 6.2.2. In considering fine particle levels, staff notes that East/West
18 differences are substantially smaller in urban areas than in rural areas. As shown in Figure 6-1,
19 24-hour average PM2 5 concentrations in urban areas in the East and West are much more similar
t
20 than in rural areas. A significantly lower East/West ratio is observed in urban areas, based on
21 data from either the FRM or the EPA Speciation Network, than in rural areas, based on data from
22 the IMPROVE network.
23 In considering relative humidity levels, staff notes that, while the average daily relative
24 humidity levels are generally higher in eastern than western areas, in both regions relative
25 humidity levels are appreciably lower during daylight as compared to night time hours. These
3 A preliminary analysis of the archived data for 63 cities across the U.S. was presented in the first draft
Staff Paper (August 2003), but further analysis has not been conducted. While the preliminary analysis
demonstrated relatively well-characterized correlations between predicted PM, j concentrations (based on ASOS
extinction values) and measured PM25 concentrations in some urban areas, such correlations were not consistently
observed in urban areas across the country.
January 2005 6-7 Draft - Do Not Quote or Cite
f
-------
a
t
00
vi
m
I
•a
•s
^
&
I
-------
1 differences can be seen in Figure 6-2, based on data from National Weather Service (NWS) sites.
2 As discussed in Chapter 2, section 2.8, the reconstructed light extinction coefficient, for a given
3 mass and concentration, increases sharply as relative humidity rises. Thus, visibility impacts
4 related to East/West differences in average relative humidity are minimized during daylight
5 hours, when relative humidity is generally lower.
6 Taking these factors into account, staff has considered both 24-hour and shorter-term
7 daylight hour averaging periods in evaluating correlations between PM2 5 concentrations in urban,
8 areas and visibility, in terms of reconstructed light extinction (RE), in eastern and western areas,
9 as well as nationwide. Figure 6-3 shows clear and similarly strong correlations between RE and
10 24-hour average PM2 s in eastern, western, and all urban areas. Figure 6-3 is based on data from
11 161 urban continuous PM25 mass monitoring sites across the country with co-located or nearby
12 24-hour PM2 5 speciation data. RE values were calculated based on a constructed hourly
13 speciated PMZ5 data set, hourly relative humidity data (either co-located or from nearby NWS
14 sites), and a coarse PM data set (estimated either by difference method from the continuous
15 PM25 and co-located continuous PMi0 instruments, or based on regional ratios of PM fractions)
1.6 (Schmidt et al., 2005). In calculating RE, the relative humidity was capped at 95%, reflecting
17 the lack of accuracy in higher relative humidity values and their highly disproportionate impact
18 on RE.
19 For these analyses, staff has considered both 10 years of relative humidily data,
20 converted to 10-year average hourly f(RH)4 values (Figure 6-3, panel a), as well as actual hourly
21 relative humidity data for 2003, converted to f(RH) values (Figure 6-3, panel b). Staff
22 recognizes that 10-year average hourly f(RH) data are more reflective of long-term humidity
23 patterns, and may provide a more appropriate basis for relating ambient PM2 5 levels to visibility
24 impairment in the context of consideration of a potential secondary standard to protect against
25 PM-related visibility impairment. On the other hand, since there can be significant day-to-day
26 variance in relative humidity that is not reflected in long-term average f(RH) data, actual hourly
*f(RH) is the relative humidily adjustment factor; it increases significantly with higher humidity. See
section 2.8.1 and Chapter 4 of the CD (CD, pp. 4-149 to 4-170) for further information.
January 2005
6-9
Draft - Do Not Quote or Cite
-------
CM
I—I
H:
Ht
i i
^
Hi
Ht
H{
I i"
HH±t
^
I
>h
>h
^
HP
ir.|... I,,.,!.,,,-!.
u
•* «w
S IS
CM
tM
CM
si
£ *
M
CM
3
89 4>
s
>
-------
Figure 6-3. Relationship between reconstructed light extinction (RE) and 24- hour
average PM2 s, 2003. RE in top panel (a) computed with 10-year average
f(RH)\ RE in bottom panel (b) computed using actual f(RH).
Source: Schmidt et al. (2005)
January 2005
6-11
Draft - Do Not Quote or Cite
-------
1 f(RH) data were also included in the analyses, to reflect the potential ranges of high and low
2 relative humidity levels likely to occur over the course of a year.
3 In considering shorter-term daylight hour averaging periods, staff also evaluated the
4 slope and strength of the correlations between RE and PM25 concentrations on an hourly basis
5 (Schmidt et al., 2005). Figure 6-4 shows plots of the average slope of the correlation between
6 hourly RE and corresponding PM2 5 concentrations (i.e., the increase in RE due to Ihe
7 incremental increase in PM2 s) by region, in eastern and western areas, and nationwide. The
8 slopes are all lower during daytime hours when the disproportionate effects of relative humidity
9 on the light extinction coefficients for fine particle sulfates and nitrates are diminished. Thus,
10 during daylight hours, the slope more closely represents the influence of PM25 mass on visibility
11 than the influence of relative humidity. In addition, Figure 6-4 shows that the slopes (and hence,
12 the relationships between RE and PM2 5) are more comparable across regions during daylight
13 hours. In considering the strength of these correlations, staff notes that the correlations between
14 RE and PM25, as indicated by the model R2 values, are strong for individual daylight hours,
15 similar to that for the 24-hour average (Schmidt et al., 2005).' On a national basis, daytime (9
16 a.m. to 6 p.m.) hourly model R2 values are all above 0.6 for the RE's calculated with actual f(RH)
11 values and above 0.8 for the RE's calculated with 10-year average f(RH) values (Schmidt et al.,
18 2005).
19 On the basis of lower slopes and more inter-region comparability, staff selected anumber
20 of daylight time periods to consider in evaluating additional correlations between PM2 5
21 concentrations and RE in eastern and western regions, as well as nationwide. Evaluated time'
22 periods included 7 a.m. to 7 p.m.; 9 a.m. to 5 p.m.; 10 a.m. to 6 p.m.; 10 a.m. to 4 p.m.; 12 p.m.
23 to 4 p.m.; and 8 a.m. to 4 p.m. With a focus on minimizing slope, minimizing regional and
24 East/West slope differences, maximizing R2 values, and considering other related factors, staff
25 selected the 12 p.m. to 4 p.m. time period for further analyses (Schmidt et al., 2005).
26 Using the same data as were used for Figure 6-3, Figure 6-5 shows examples of the
27 correlations between RE and PM25 concentrations averaged over a 4-hour time period, for 10-
28 year average hourly f(RH) data (panel a) and for actual hourly f(RH) data in 2003 (panel b). As
29 seen in this figure, the correlations between RE and PM2 5 concentrations during daylight hours
January 2005
6-12
Draft - Do Not Quote or Cite
-------
i U>
CO
o> CO
If
6
o
tj **
2 w
O (0
CO UJ
II
III
CO Z
«t *
di
I *
.2 o>
51
to ^
I s
1^
in (0
<8
-------
Figure 6-5. Relationship between reconstructed light extinction (RE) and 12 p.m. -
4 p.m. average PM2 5,2003. RE in top panel (a) computed with 10-year
avsragsffRH); RE in bottom panel (b) computed using actual f(RH).
Source: Schmidt et al. (2005)
January 2005
6-15
Draft - Do Not Quote or Cite
-------
1 in urban areas are comparably strong (similar R2 values), yet more reflective of PM2 5 mass rather
2 than relative humidity effects (i.e., lower slopes), in comparison to the correlations based on a
3 24-hour averaging time. Further, these correlations in urban areas are generally similar in the
4 East and West, in sharp contrast to the East/West differences observed in rural areas.
5
6 6.2.4 Economic and Societal Value of Improving Visual Air Quality
7 Visibility is an air quality-related value having direct significance to people's enjoyment
8 of daily activities in all parts of the country. Survey research on public awareness of visual air
9 quality using direct questioning typically reveals that 80% or more of the respondents are aware
10 of poor visual air quality (Cohen et al., 1986). The importance of visual air quality to public
11 welfare across the country has been demonstrated by a number of studies designed to quantify
12 the benefits (or willingness to pay) associated with potential improvements in visibility
13 (Chestnut and Rowe, 1991).
14 Individuals value good visibility for the sense of well-being it provides them directly,
15 both in the places where they live and work, and in the places where they enjoy recreational
16 opportunities. Millions of Americans appreciate the scenic vistas in national parks and
17 wilderness areas each year. Visitors consistently rate "clean, clear air" as one of the most
18 important features desired in visiting these areas (Department of Interior, 1998). A 1998 survey
19 of 590 representative households by researchers at Colorado State University found that 88% of
20 the respondents believed that "preserving America's most significant places for future
21 generations" is very important, and 87% of the respondents supported efforts to clean up air
22 pollution that impacts national parks (Hass and Wakefteld, 1998).
23 . Economists have performed many studies in an attempt to quantify the economic benefits
24 associated with improvements in current visibility conditions both in national parks and in urban
25 areas. These economic benefits are often described by economists as either use values or non-
26 use values. Use values are those aspects of environmental quality that directly affect an
27 individual's welfare. These include improved aesthetics during daily activities (e.g., driving or
28 walking, looking out windows, daily recreations), for special activities (e.g., visiting parks and
January 2005
6-15
Draft - Do Not Quote or Cite
-------
1 scenic vistasj hiking, hunting), and for viewing scenic photography. Aesthetic benefits of better
2 visibility also include improved road and air safety.
3 Non-use values are those for which an individual is willing to pay for reasons that do not
4 relate to the direct use or enjoyment of any environmental benefit The component of non-use
5 value that is related to the use of the resource by others in the future is referred to as the bequest
6 value. This value is typically thought of as altruistic in nature. Another potential component of
7 non-use value is the value that is related to preservation of the resource for its own sake, even if
8 there is no human use of the resource. This component of non-use value is sometimes referred to
9 as existence value or preservation value. Non-use values are not traded, directly or indirectly, in
10 markets. For this reason, the estimation of non-use values has proved to be significantly more
11 difficult than the estimation of use values. Non-use values may be related to the desire that a
12 clean environment be available for the use of others now and in the future, or they may be related
13 to the desire to know that the resource is being preserved for its own sake, regardless of human
14 , use. Non-use values may be a more important component of value for recreational areas,
15 particularly national parks and monuments, and for wilderness areas.
16 In addition, staff notes that the concept of option value is a key component of the
17 measured values. The option value represents the value that is tied to preserving improved
18 visibility in the event of a visit, even though a visit is not certain. This component is considered
19 by some as a use value and by others as a non-use value.
20 Tourism in the U.S. is a significant contributor to the economy. A1998 Department of
21 Interior study found that travel-related expenditures by national park visitors alone average $.14.5
22 billion annually (1996 dollars) and support 210,000 jobs (Peacock et al., i998). A similar
23 estimate of economic benefits resulting from visitation to national forests and other public lands
24 could increase this estimate significantly.
25 It is well recognized in the U.S. and abroad that there is an important relationship
26 between good air quality and economic benefits due to tourism. McNeill and Roberge (2000)
27 studied the impact of poor visibility episodes on tourism revenues in Greater Vancouver and the
28 Lower Fraser Valley in British Columbia as part of the Georgia Basin Ecosystem Initiative of
29 Environment Canada Through this analysis, a model was developed that predicts future tourist
January 2005
6-16
Draft - Do Not Quote or Cite
-------
1 revenue losses that would result from a single extreme visibility episode. They found that such
2 an episode would result in a $7.45 million loss in the Greater Vancouver area and $1.32 million
3 loss in the Fraser Valley.
4 The results of several valuation studies addressing both urban and rural visibility are
5 presented in the CD (CD, pp. 4-187 to 4-190), the 1996 Criteria Document (EPA, 1996a, p. 8-83,
6 Table 8-5; p. 8-85, Table 8-6) and in Chestnut and Rowe (1991) and Chestnut et al. (1994). Past
7 studies by Schulze et al. (1983) and Chestnut and Rowe (1990) have estimated the preservation
l
8 values associated with improving the visibility in national parks in the Southwest to be in the
9 range of approximately $2-6 billion annually. An analysis of the residential visibility benefits in
10 the eastern U.S. due to reduced sulfur dioxide emissions under the acid rain program suggests an
11 annual value of $2.3 billion (in 1994 dollars) in the year 2010 (Chestnut and Dennis, 1997). The
12 authors suggest that these results could be as much as $1-2 billion more because the above
13 estimate does not include any value placed on eastern air quality improvements by households in
14 the western U.S.
15 Estimating benefits for improvements in visibility can be difficult because visibility is not
16 directly or indirectly valued in markets. Many of the studies cited above are based on a
17 valuation method known as contingent valuation (CV). Concerns have been identified about the
18 reliability of value estimates from contingent valuation studies because research has shown that
19 bias can be introduced easily into these studies if they are not carefully conducted. Accurately
20 estimating willingness-to-pay for avoided health and welfare losses depends on the reliability
21 and validity of the data collected. However, there is an extensive scientific literature and body of
22 practice on both the theory and technique of contingent valuation. EPA believes that well-
23 designed and well-executed CV studies are useful for estimating the benefits of environmental
24 effects such as improved visibility (EPA, 2000).
25 Some of the studies cited above used an alternative valuation method known as hedonic
26 pricing. Hedonic pricing is a technique used to measure components of property value (e.g.,
27 proximity to schools). It relies on the measurement of differentials in property values under
28 various environmental quality conditions, including air pollution and environmental amenities,
29 such as aesthetic views. This method works by analyzing the way that market prices change
January 2005
6-17
Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
II
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
with changes in environmental quality or amenity. EPA believes that well-designed and well-
executed hedonic valuation studies, in combination with public perception surveys, are useful for
estimating the benefits of environmental effects such as improved visibility.
Society also values visibility because of the significant role it plays in transportation
safety. Serious episodes of visibility impairment can increase the risk of unsafe air
transportation, particularly in urban areas with high air traffic levels (EPA, 1982). In some
cases, extreme haze episodes have led to flight delays or the shutdown of major airports,
resulting in economic impacts on air carriers, related businesses, and air travelers. For example,
on May 15,1998 in St. Louis, Missouri, it was reported that a haze episode attributed to
i.
wildfires in central America resulted in a reduction in landing rates and significant flight delays
at Lambert International Airport. The 24-hour PM25 levels reached 68 ng/m3 during that
episode. In addition, the National Transportation Safety Board (NTSB) has concluded in
accident reports that high levels of pollution and haze, such as those experienced during the July
1999 air pollution episode in the northeastern U. S., have played a role in air transportation
accidents and loss of life (NTSB, 2000). During this episode, 24-hour levels of PM2 5 ranged
from 35-52 ug/m3 in the New England states.
6.2.5 Programs and Goals for Improving Visual Air Quality
Specific discussion is provided below on regional visibility programs in the U.S.; as well
as local visibility programs established by States, localities, and other countries in an effort to
protect visual air quality.
6.2.5.1 Regional Protection
Due to differences in visibility impairment levels (due to differences in chemical
composition of haze and in relative humidity levels) between the'East and West, EPA, land
managers, and States have taken a regional approach, rather man a national approach, to
protecting visibility in non-urban areas in the U.S.. Protection against visibility impairment in
special areas is provided for in sections 169A/169B, and 165 of the Act, in addition to that
provided by the secondary NAAQS. Section 169A, added by the 1977 CAA Amendments,
established a national visibility goal to "remedy existing impairment and prevent future
January 2005
6-18
Draft - Do Not Quote or Cite
-------
1 impairment" in 156 national parks and wilderness areas (Class I areas). The Amendments also
2 called for EPA to issue regulations requiring States to develop long-term strategies to make
3 "reasonable progress" toward the national goal. EPA issued initial regulations in 1980 focusing
4 on visibility problems that could be linked to a single source or small group of sources. Action
5 was deferred on regional haze until monitoring, modeling, and source apportionment methods
6 could be improved.
7 The 1990 CAA Amendments placed additional emphasis on regional haze issues through
8 the addition of section 169B. In accordance with this section, EPA established the Grand
9 Canyon Visibility Transport Commission (GCVTC) in 1991 to address adverse visibility impacts
10 on 16 Class I national parks and wilderness areas on the Colorado Plateau. The GCVTC was
11 comprised of the Governors of nine western states and leaders from a number of Tribal nations.
12 The GCVTC issued its recommendations to EPA in 1996, triggering a requirement in section
13 169B for EPA issuance of regional haze regulations.
14 EPA accordingly promulgated a final regional haze rule in 1999 (EPA, 1999; 65 FR
15 35713). Under the regional haze program, States are required to establish goals for improving
16 visibility on the 20% most impaired days in each Class I area, and for allowing no degradation
17 on the 20% least impaired days. Each state must also adopt emission reduction strategies which,
18 in combination with the strategies of contributing States, assure that Class I area visibility
19 improvement goals are met. The first State implementation plans are to be adopted in the 2003-
20 2008 time period, with the first implementation period extending until 2018. Five multistate
21 planning organizations are evaluating the sources of PM2 5 contributing to Class I area visibility
22 impairment to lay the technical foundation for developing strategies, coordinated among many
23 States, in order to make reasonable progress in Class I areas across the country.
24 6.2.5.2 Local, State, and International Goals and Programs
25 The value placed on protecting visual air quality is further demonstrated by the existence
26 of a number of programs, goals, standards, and planning efforts that have been established in the
27 U.S. and abroad to address visibility concerns in urban and non-urban areas. These regulatory
28 and planning activities are of particular interest because they are illustrative of the significant
29 value that the public places on improving visibility, and because they have made use of
January 2005
6-19
Draft - Do Not Quote or Cite
-------
f'.
1 developed methods for evaluating public perception's and judgments about the acceptability of
2 varying degrees of visibility impairment.
3 Several state and local governments have developed programs to improve visual air
4 quality in specific urban areas, including Denver, CO; Phoenix, AZ; and, Lake Tahoe, C A. At
5 least two States have established statewide standards to protect visibility. In addition, visibility
6 protection efforts have been undertaken in other countries, including Australia, New Zealand,
7 and Canada. Examples of these efforts are highlighted below.
8 In 1990, the State of Colorado adopted a visibility standard for the city of Denver. The
9 Denver standard is a short-term standard that establishes a limit of a four-hour average light
10 extinction level of 76 Mm"1 (equivalent to a visual range of approximately 50 km) during the
11 hours between 8 a.m. arid 4 p.m. .(Ely et al., 1991). In 2003, the Arizona Department of
12 Environmental Quality created the Phoenix Region Visibility Index, which focuses on an
13 averaging time of 4 hours during actual daylight hours. This visiblity index establishes visual air
14 quality categories (i.e., excellent to very poor) and establishes the goals of moving days in the
15 poor/very poor categories up to the fair category, and moving days in the fair category up to the
16 good/excellent categories (Arizona Department of Environmental Quality, 2003). This approach
17 results in a focus on improving visibility to a visual range of approximately 48-36 km. In 1989,
18 the state of California revised the visibility standard for the Lake Tahoe Air Basin and
19 established an 8-hour visibility' standard equal to a visual range of 30 miles (approximately 48
20 km) (California Code of Regulations).
21 California and Vermont each have standards to protect visibility, though they are based
22 on different measures. Since 1959, the state of California has had an air quality standard for
23 particle pollution where the "adverse" level was defined as the "level at which there will be...
24 reduction in visibility or similar effects." California's general statewide visibility standard is a .
25' visual range of 10 miles (approximately 16 km) (California Code'of Regulations). In 1985,
26 Vermont established a state visibility standard that is expressed as a summer seasonal sulfate
27 concentration of 2 ug/m3, that equates to a visual range of approximately 50 km. This standard
28 was established to represent "reasonable progress toward attaining the congressional visibility
t
January 2005
6-20
Draft - Do Not Quote or Cite
-------
1 goal for the Class 1 Lye Brook National Wilderness Area, and applies to this Class 1 area and to
2 all other areas of the state with elevations greater than 2500 ft.
3 Outside of the U.S., efforts have also been made to protect visibility. The Australian
4 state of Victoriahas established a visibility objective (State Government of Victoria, 2000a and
5 2000b), and a visibility guideline is under consideration in New Zealand (New Zealand National
6 Institute of Water & Atmospheric Research, 2000a and 2000b; New Zealand Ministry of
7 Environment, 2000). A survey was undertaken for the Lower Fraser Valley in British Columbia,
8 with responses from this pilot study being supportive of a standard in terms of a visual range of
9 approximately 40 km for the suburban township of Chilliwack and 60 km for the suburban
10 township of Abbotsford, although no visibility standard has been adopted for the Lower Fraser
11 Valley at this time.
12
13 6.2.6 Approaches to Evaluating Public Perceptions and Attitudes
14 New methods and tools have been developed to communicate and evaluate public
15 perceptions of varying visual effects associated with alternative levels of visibility impairment
16 relative to varying pollution levels and environmental conditions. New survey methods have
17 been applied and evaluated in various studies, such as those for Denver, Phoenix, and the Lower
18 Fraser Valley in British Columbia, and these studies are described below in more detail. These
19 methods are intended to assess public perceptions as to the acceptability of varying levels of
20 visual air quality, considered in these studies to be an appropriate basis for developing goals and
21 standards for visibility protection. For the Denver and British Columbia studies, actual slides
22 taken in the areas of interest, and matched with transmissometer and nephelometer readings,
23 respectively, were used to assess public perceptions about visual air quality. For the Phoenix
24 study, WinHaze, a newly available image modeling program, discussed below, was used for
25 simulating images. Staff finds that, even with variations in each study's approaches, the survey
26 methods used for the Denver, Phoenix, and British Columbia studies produced reasonably
27 consistent results from location to location, each with a majority of participants finding visual
28 ranges within about 40 to 60 km to be acceptable.
January 2005
6-21
Draft - Do Not Quote or Cite
t
-------
1 6.2.6.1 Photographic Representations of Visual Air Quality
2 In the past, the principal method for recording and describing visual air quality has been
3 through 35 millimeter photographs. Under the IMPROVE program, EPA, federal land
4 management agencies, and Air Resource Specialists, Inc. (ARS) have developed an extensive
5 archive of visual air quality photos for national parks and wilderness areas. In comparison, we
6 have only a limited archive of photos of urban areas.
7 The CD discusses some of the methods that are now available to represent different
8 levels of visual air quality (CD, p. 4-174). In particular, Molenar et al. (1994) describes a
9 sophisticated visual air quality simulation technique, incorporated into the WinHaze program
10 developed'by ARS, which combined various modeling systems under development for Ihe past
11 20 years. The technique relies on first obtaining an original base image slide of the scene of
12 interest. The slide should be of a cloudless sky under the cleanest air quality conditions possible.
13 The light extinction represented by the scene should be derived from aerosol and optical data
14 associated with the day the image was taken, or it should be estimated from contrast
15 measurements of features in the image. The image is then digitized to assign an optical density
16 to each pixel. At this point, the radiance level for each pixel is estimated. Using a detailed
17 topographic map, technicians identify the specific location from which the photo was taken, and
18 they determine the distances to various landmarks and objects in the scene. With this
19 information, a specific distance and elevation is assigned to each pixel.
20 Using the digital imaging information, the system then computes the physical and optical
21 properties of an assumed aerosol mix. These properties are input into a radiative transfer model
22 in order to simulate the optical properties of varying pollutant concentrations on the scene.
23 WinHaze, an image modeling program for personal computers that employs simplified
24 algorithms based on the sophisticated modeling technique, is now available (Air Resource
25 Specialists, 2003).
26 The simulation technique has the advantage of being readily applicable to any location
27 as long as a very clear base photo is available for that location. In addition, the lack of clouds
28 and consistent sun angle in all images, in effect, standardizes the perception of the images and
29 enables researchers to avoid potentially biased responses due to these factors. An alternative
January 2005 6-22 Draft - Do Not Quote or Cite
-------
1 technique to using simulated images is to obtain actual photographs of the site of interest at
2 different ambient pollution levels. However, long-term photo archives of this type exist for only
3 a few cities. In addition, studies have shown that observers will perceive an image with a cloud-
4 filled sky as having a higher degree of visibility impairment than one without clouds, even
5 though the PM concentration on both days is the same.
6 As part of a pilot study5 in Washington, D.C., both survey and photographic techniques
7 were applied (Abt Associates, 2001). In conjunction with this pilot project, images that illustrate
8 , visual air quality in Washington, DC under a range of visibility conditions were prepared and are
9 available at http://www,epa.goy/ttii/naaqs/sta^^ (labeled as Attachment
10 6-A: Images of Visual Air Quality in Selected Urban Areas in the U.S.). Included as part of
11 Attachment 6-A, this website also contains actual photographs of Chicago illustrating visibility
12 conditions associated with a range of PM23 concentrations, as well as simulated images for
13 Denver and Phoenix, as discussed below.
14 6.2.6.2 Survey Methods
15 Denver, Colorado: Visibility Standard
16 The process by which the Denver visibility standard was developed relied on citizen
17 judgments of acceptable and unacceptable levels of visual air quality (Ely et al., 1991).
18 Representatives from Colorado Department of Public Health and Environment (CDPHE)
19 conducted a series of meetings with 17 civic and community groups in which atotal of 214
20 individuals were asked to rate slides having varying levels of visual air quality for a well-known
21 vista in Denver. The CDPHE representatives asked the participants to base their judgments on
22 three factors: 1) the standard was for an urban area, not a pristine national park area where the
23 standards might be more strict; 2) standard violations should be at visual air quality levels
24 considered to be unreasonable, objectionable, and unacceptable visually; and 3) judgments of
25 standards violations should be based on visual air quality only, not on health effects.
26 The participants were shown slides in 3 stages. First, they were shown seven warm-up
27 slides describing the range of conditions to be presented. Second, they rated 25 randomly-
5 A small pilot study for Washington, D.C. was conducted by EPA and was briefly discussed in the
preliminary draft staff paper (2001).
January 2005
6-23
Draft - Do Not Quote or Cite
f
-------
* 1 ordered slides based on a scale of 1 (poor) to 7 (excellent), with 5 duplicates included. Third,
2 they were asked to judge whether the slide would violate what they would consider to be an
3 appropriate urban visibility standard (i.e., whether the level of impairment was "acceptable" or
4 "unacceptable"). . . .
5 The Denver visibility standard setting process produced the following findings:
6 ' . ' , * * ••'- ' '; ' ' '
7 • Individuals' judgments of a slide's visual air quality and whether the slide violated a •
8 visibility standard are highly correlated (Pearson correlation coefficient greater than
9 " 80%) with the group average. ' •''•
10 - •. '...-..
11 • When participants judged duplicate slides, group averages of the first and second ratings
12 were highly correlated.
13 . . ' •• . . . /,'..:
14 • Group averages of visual air quality ratings and "standard violations" were highly
15 correlated.' The strong relationship of standard violation judgments with the visual air ;
16 quality ratings is cited as the best evidence available from this study for the validity of
17 standard violation judgments (Ely et al., 1991).
18 ' -. ' •
19 The CDPHE researchers sorted the ratings for each slide by increasing order of light
20 extinction and calculated the percent of participants that judged each slide to violate the
21 standard. The Denver visibility standard1 was then established based on a 50% acceptability
22 criterion. Under this approach, the standard was identified as the light extinction level that
23 divides the slides into two groups: those found to be acceptable and those found to be
24 unacceptable by a majority of study participants. The CDPHE researchers found this level to be
25 reasonable because, for the slides at this level and above, a majority of the study participants
26 judged the light extinction levels to be unacceptable. In fact, when researchers evaluated all
27 citizen judgments made on all slides at this level and above as a single group, more than 85% of
28 the participants found visibility impairment at and above the level of the selected standard to be
29 unacceptable.'
30 Though images used in the Denver study were actual photographs, more recently,
31 WinHaze has been used to generate images that illustrate visual air quality in Denver under a
32 range of visibility conditions (generally corresponding to 10th, 20*, 30*, 40th, 50*, 60th 80*, and
33 90* percentile values), and these images are available in Attachment 6-A at
34 http://www.epa.gov/ttn/naaqs/staiidards/pni/s_pm_cr_sp.htiTil.
January 2005 6-24 Draft - Do Not Quote or Cite
-------
1 Data analyses using extensive new monitoring data now available on.PM2 3 primarily in
2 urban areas show a consistently high correlation between hourly PM2.S data and RE coefficients ,
3 for urban areas across regions of the U.S. during daylight hours. These correlations in urban
4 areas are generally similar in the East and West, in sharp contrast to the East/West differences
5 observed in rural areas.
6 The importance of visual air quality to public welfare across the country has been
7 demonstrated by a number of studies designed to quantify the benefits (or willingness to pay).
8 associated with potential improvements in visibility. The value placed on protecting visual air
9 quality is further demonstrated by the existence of a number of programs, goals, standards, and
10 planning efforts that have been established in the U.S. and abroad to address visibility concerns
11 in urban and non-urban areas.
12 In some urban areas, poor visibility has led to more localized efforts to better
13 characterize, as well as improve, urban visibility conditions. The public perception survey -
14 approach used in the Denver, Phoenix, and British Columbia studies yielded reasonably
15 consistent results, with each study indicating that a maj ority of citizens find value in protecting
16 local visibility to within a visual range of about 40 to 60 km. In the cases of Denver and
17 Phoenix, these studies provided the basis for the establishment of their visibility standards and
18 goals. ..,.-. ,
19 Staff believes ,that the findings of the new data analyses, in combination with recognized
20 benefits to public welfare of improved visual air quality and an established approach for
21 • determining acceptable visual range, provide a basis for considering revisions to the secondary
22 PM25 standards to protect against PM-related visibility effects in urban areas.
23 ....,--,
24 6.3 EFFECTS ON VEGETATION AND ECOSYSTEMS
25 Information and conclusions regarding what is currently known about the impacts of
26 ambient PM on ecosystems and individual components of ecosystems such as vegetation, soils,
27 water, and wildlife are discussed in Chapters 4 and 9 of the CD. This section seeks to build upon
28 and focus this body of science using EPA's ecological risk paradigm in a manner that highlights
29 the usefulness and policy relevance of the scientific information. In so doing, staff has drawn
January 2005
6-27
Draft - Do Not Quote or Cite
t
-------
1 from EPA's Guidelines for Ecological Risk Assessment (Guidelines) (EPA, 1998), which
2 expanded upon the earlier document, Framework for Ecological Risk Assessment (EPA, 1992),
3 with the goal of improving the quality of ecological risk assessments and increasing the
\ '
4 consistency of assessments across the Agency.
5 According to the Guidelines document, the three main phases of ecological risk
6 assessment are problem formulation, analysis, and risk characterization. In problem formulation,
7 the purpose for the assessment is articulated, the problem is defined, assessment endpoints are
8 selected, a conceptual model is prepared and an analysis plan is developed: Initial work in
9 problem formulation includes the integration of available information on sources, stressors,
10 effects, and ecosystem and receptor characteristics.
11 In the analysis phase data are evaluated to determine how exposure to stressors is likely
12 to occur (exposure profile) and the relationship between stressor levels and ecological effects
13 (stressor-response profile). These products provide the basis for the risk characterization phase.
14 During the third phase, risk characterization, the exposure and stressor-response profiles
15 are integrated through the risk estimation process. Risk characterization includes a summary of
16 assumptions, scientific uncertainties, and strengths and limitations of the analyses. The final
17 product is a risk description in which the results of the integration are presented, including an
18 interpretation of ecological adversity and description of uncertainty and lines of evidence.
19 Keeping these goals and guidelines in mind, this section organizes information into the
20 following seven subsections: major ecosystem stressors in PM (6.3.1); direct vegetation effects
21 of PM stressor deposition (6.3.2); ecosystem effects of PM stressor deposition (6.3.3);
22 characteristics and location of sensitive ecosystems within the U.S. (6.3.4); ecosystem exposures
23 to PM deposition (6.3.5); consideration of critical loads as an approach for effects
24 characterization and/or as a management tool (6.3.6); and summary and conclusions (6.3.7).
25 This review will also consider and reference where applicable the extent to which PM
26 affects the essential ecological attributes (EEAs) outlined in the Framework for Assessing and
21 Reporting on Ecological Condition, recommended by the Ecological Processes and Effects
28 Committee (EPEC) of EPA's Science Advisory Board (hereafter EPEC Framework; SAB,
29 2002), ad described in subsections 4.2.1 and 4.2.3 of the CD.
January 2005 * 6-28 Draft - Do Not Quote or Cite
-------
1 6.3.1 Major Ecosystem Stressors in PM
2 As previously discussed, PM is not a single pollutant, but a heterogeneous mixture of
3 particles differing in size, origin, and chemical composition. This heterogeneity of PM exists not
4 only within individual particles or samples from individual sites, but to an even greater extent,
5 between samples from different sites. Since vegetation and other ecosystem components are
6 affected more by particulate chemistry than size fraction, exposure to a given mass concentration
7 of airborne PM may lead to widely differing plant or ecosystem responses, depending on the
8 particular mix of deposited particles. Though the chemical constitution of individual particles
9 can be strongly correlated with size, the relationship between particle size and particle
10 composition can also be quite complex, making it difficult in most cases to use particle size as a
11 surrogate for chemistry. Because PM size classes do not necessarily have specific differential
12 relevance for vegetation or ecosystem effects (Whitby, 1978; EPA, 1996a), it is the opinion of
13 the staff that an ecologically relevant indicator for PM would be based on one or multiple
14 chemical stressors found in ambient PM. At this time it remains to be studied as to what extent
15 NAAQS standards focused on a given size fraction would result in reductions of the ecologically
16 relevant constituents of PM for any given area.
17 A number of different chemical species found within ambient PM and their effects on
18 vegetation and ecosystems were discussed in chapter 4 of the PM CD. In particular, the CD
19 focused on nitrates and sulfates, concluding that these PM constituents are considered to be the
20 stressors of greatest environmental significance (CD, p. 9-114): Other components of PM, such
21 as dust, trace metals, and organics, which can also be toxic to plants and other organisms at high
22 levels, were also discussed. However, because such high levels occur only near a few limited
23 point sources and/or on a very local scale, they do not appear significant at the national level.
24 Therefore, the remainder of this section will narrow its focus to consideration of the impacts of
25 particulate nitrates and sulfates, both separately and in combination with acidifying compounds,
26 on sensitive ecosystem components and essential ecological attributes, which in turn, impact
27 overall ecosystem structure and function.
28
January 2005
6-29
Draft - Do Not Quote or Cite
f
-------
1 6.3.2 Direct Vegetation Effects of PMStressor Deposition
2 Nitrogen is a critical limiting nutrient for plant growth. The process of photosynthesis
3 uses approximately 75% of the nitrogen in a plant leaf, and,.thus, to a large extent, governs the
4 utilization of other nutrients such as phosphorus, potassium (CD, p. 4-95). Plants usually absorb
5 nitrogen (as NH4+ of N03") through their roots. However, particle deposition of nitrate, together •
6 with other nitrogen-containing gaseous and precipitation-derived sources, can represent a
• 7 substantial fraction of total nitrogen reaching vegetation. In nitrogen-limited ecosystems, this
8 influx of N can act as a fertilizer. Though it is known that foliar uptake of nitrate can occur, the
9 mechanism of foliar uptake is not well established, and it is not currently possible to distinguish
f
10 sources of chemicals deposited as gases or particles using foliar extraction. Since it has proven
11 difficult to quantify the percentage of nitrogen uptake by leaves that is contributed by ambient
12 particles, direct foliar effects of nitrogen-containing particles have not been documented. (CD,
13 pp. 4-69, 4-70).
14 Similar.to nitrogen, sulfur is an essential plant nutrient that can deposit on vegetation in
15 the form of sulfate particles, or be taken up by plants in gaseous form." Greater than 90% of
16 anthropogenic sulfur emissions are as sulfur dioxide (SO2)S with most of the remaining emissions
17 in the form of sulfate. However, sulfur dioxide is rapidly transformed in the atmosphere to
18 sulfate, which is approximately 30-fold less phytotoxic than SO2. Low dosages of sulfur can
19 also serve as a fertilizer, particularly for plants growing in sulfur-deficient soils. There are only
20 a few field demonstrations of foliar sulfate uptake, however, and the relative importance of foliar
21 leachate and prior dry-deposited sulfate particles remains difficult to quantify. Though current
22 levels of sulfate deposition reportedly exceed the capacity of most vegetative canopies to
23 immobilize the sulfur, sulfate additions in excess of needs do not typically lead to plant injury
24 (CD, .pp. 4-71,4-72). , .-.-.'
25 Staff therefore conclude that at current ambient levels, risks to vegetation from short term
26 exposures to dry deposited particulate nitrate or sulfate are low. Additional studies are needed,
27 however, on the effects of sulfate particles on physiological characteristics of plants following
28 chronic exposures (CD, p. 4-72).
January 2005
6-30
Draft - Do Not Quote or Cite
-------
1 Though dry deposition of nitrate and sulfate particles does not appear to induce foliar
2 injury at current ambient exposures, when found in acidic precipitation, such particles do have
3 the potential to cause direct foliar injury. This is especially true when the acidic precipitation is
4 in the form of fog and clouds, which may contain solute concentrations many times those found
5 in rain. In experiments on seedling and sapling trees, both coniferous and deciduous species
6 showed significant effects on leaf surface structures after exposure to simulated acid rain or acid
7 mist at pH 3.5, .while some species have shown subtle effects at pH 4 and above. Epicuticular
8 waxes, which function to prevent water loss from plant leaves, can be destroyed by acid rain in a
9 few weeks, which suggests links between acidic precipitation and aging. Due to their longevity
10 and evergreen foliage, the function of epicuticular wax is more crucial in conifers. For example,
11 red spruce seedlings, which have been extensively studied, appear to be more sensitive to acid
12 precipitation (mist and fog) when compared with other species (CD, pp. 4-72, 4-73). In addition
13 to accelerated weathering of leaf cuticular surfaces, other direct responses of forest trees to
^14 acidic precipitation include increased permeability of leaf surfaces to toxic materials, water, and
15 disease agents; increased leaching of nutrients from foliage; and altered reproductive processes
16 (CD, p. 4-86). All of these effects serve to weaken trees so that they are more susceptible to
17 other stresses (e.g., extreme weather, pests, pathogens).
18 Acid precipitation with levels of acidity associated with the foliar effects described above
19 are currently found in some locations in the U.S.. For example, in the eastern U.S., the mean
20 ' precipitation pH ranges from 4.3 (Pennsylvania and New York) to 4.8 (Maine)(EPA, 2003). It
21 can be assumed that occult (mist or fog) deposition impacting high elevations more frequently,
22 would contain even higher concentrations of acidity. Thus, staff conclude that the risks of foliar
23 injury occurring from acid deposition is high. The contribution of particulate sulfates and
24 nitrates to the total acidity found in the acid deposition impacting eastern vegetation is not clear.
25
26 6.3.3 Ecosystem Effects of PMStressor Deposition
27 Ecosystem-level responses related to PM occur when the effects of PM deposition on the
28 biological and physical components of ecosystems become sufficiently widespread as to impact
29 essential ecological attributes such as nutrient cycling and/or shifts in biodiversity. The most
January 2005
6-31
Draft - Do Not Quote or Cite
-------
1 significant PM-related ecosystem-level effects result from long-term cumulative deposition of a
2 given chemical species (e.g., nitrate) or mix (e.g., acidic deposition) that exceeds the natural
3 buffering or storage capacity of the ecosystem and/or affects the nutrient status of the ecosystem,
4 usually by indirectly changing soil chemistry, populations of bacteria involved in nutrient
5 cycling, and/or populations of fungi involved in plant nutrient uptake (CD, pp. 4-90,4-91). To
6 understand these effects, long-term, detailed ecosystem or site-specific data usually are required.
7 The availability of this type of long-term data is limited. The following discussion is organized
, 8 according to the speciated effects of PM on ecosystems.
9 63.3.1 Environmental Effects of Reactive Nitrogen (Nr) Deposition
10 In the environment, nitrogen may be divided into two types: nonreactive, molecular
11 nitrogen (N2) and reactive nitrogen (Nr). Molecular nitrogen is the most abundant element in the
12 atmosphere. However, it only becomes available to support the growth of plants and
13 microorganisms after it is converted into a reactive form. In nature, Nr creation is accomplished
14 by certain organisms that have developed the capability of converting N2 to biologically active
15 reduced forms (Galloway and Cowling, 2002; Homung and Langan, 1999; EPA, 1993). By the
16 mid-1960's, however, Nr creation through natural terrestrial processes'had been overtaken by Nr
17 creation as a result of human processes (CD, p. 4-95). The deposition of nitrogen in the U. S.
18 from human activity doubled between 1961 and 1997, with the largest increase occurring in the
19 1960s and 1970s (CD, p. 4-98), Reactive nitrogen is now accumulating in the environment on
20 all spatial scales - local, regional and global. The three main sources of anthropogenic Nr are:.
21 (1) the Haber-Bosch process, which converts N2 to Nr to sustain food production and some
22 industrial activities; (2) widespread cultivation of legumes, rice and other crops that promote the
23 conversion of N2 to organic nitrogen through biological nitrogen fixation; and (3) combustion of
24 fossil fuels, which converts both atmospheric N2 and fossil nitrogen to reactive NOX (CD, pp. 4-
25 95,4-96; Galloway and Cowling, 2002; Galloway et al., 2003). Currently available forms of -
26 reactive nitrogen include inorganic reduced forms (e.g., ammonia [NH3] and ammonium [NH4+]),
27 inorganic oxidized forms (e.g., nitrogen oxides [NOJ, nitric acid [HNO3], nitrous oxide [N2O],
28 and nitrate [NO3'])S and organic compounds (e.g., urea, amine, proteins, and nucleic acids (CD,
29 p. 4-95).
January 2005
6-32
Draft - Do Not Quote or Cite
-------
1 Emissions of nitrogen oxides from fuel burning increased exponentially from!940 until
2 the 1970s, leveled off after the passage of the I970 amendments to the Clean Air Act, and
3 stabilized at approximately 7 Tg NOX /yr in the late 1990s. Contemporary emissions of NO* in
4 the U.S. from fossil fuel burning are nearly two-thirds the rate of Nr released from the use of
5 inorganic fertilizers and comprise 30% of the global emissions of NOX from fossil fuel
6 combustion. Despite decreases in emissions from fossil fuel burning industries, emissions from
7 automobiles have increased approximately 10% since 1970 due to greater total miles driven
8 (Howarth et at, 2002). Some NOX emissions are transformed into a portion of ambient air PM
9 (paniculate nitrate) and deposited onto sensitive ecosystems.
10 The term "nitrogen cascade" refers to the sequential transfers and transformations of Nr
11 molecules as they move from one environmental system or reservoir (atmosphere, biosphere,
12 hydrosphere) to another, and the multiple linkages that develop among the different ecological
13 components, as shown in Figure 6-6. Because of these linkages, adding anthropogenic Nr alters
14 a wide range of biogeochernical processes and exchanges as the Nr moves among the different
15 environmental reservoirs, with the consequences accumulating through time (Galloway and
16 Cowling, 2002; Galloway et at., 2003). These changes in the nitrogen cycle are contributing to
17 both beneficial and detrimental effects to the health and welfare of humans and ecosystems
18 (Rabalais, 2002; van Egmond et al., 2002; Galloway, 1998).
19 Large uncertainties, still exist, however, concerning the rates of Nr accumulation in Hie
20 various environmental reservoirs which limit our ability to determine the temporal and spatial
21 distribution of environmental effects for a given input of Nr. These uncertainties are of great
22 significance because of the sequential nature of Nr effects on environmental processes. Reactive
23 nitrogen does not cascade at the same rate through all environmental systems. The only way to
24 eliminate Nr accumulation and stop the cascade is to convert Nr back to nonreactive N2
25 (Galloway et al., 2003).
January 2005
'6-33
Draft - Do Not Quote or Cite
f
-------
I
2
3
4
5
6
7
8
9
10
11
12
13
14
t
NOx
Energy
Production
5
s
3
*
^
Oz<
EITc
j
AAAAAAAAAAA
Atmosphere
Food
Production
NHy
People
(Food; Fiber)
Human Activities
The Nitrogen
Cascade
- Indicates dtnltrlflcitlon potential
Figure 6-6 Illustration of the nitrogen cascade showing the movement of human-
produced reactive nitrogen (Nr) as it cycles through the various
environmental reservoirs in the atmosphere and in terrestrial and aquatic
ecosystems (Galloway et al., 2003; Figure 4-15, CD p. 4-97).
Some of the more significant detrimental effects resulting from chronic increased inputs
of atmospheric Nr (e.g., participate nitrates) include: (1) decreased.productivity, increased
mortality, and/or shifts in terrestrial plant community composition, often leading to decreased
biodiversity in many natural habitats wherever atmospheric Nr deposition increases significantly
and critical thresholds are exceeded (Aber et al., 1995); (2) leaching of excess nitrate and
associated base cations from terrestrial soils into streams, lakes and rivers and mobilization of
soil aluminum; (3) eutrophication, hypoxia, loss of biodiversity, and habitat degradation in
coastal ecosystems, now considered a major pollution problem in coastal waters (Rabalais,
January 2005
6-34
Draft - Do Not Quote or Cite
-------
1 2002); (4) acidification and loss of aquatic flora and fauna biodiversity in lakes and streams in
2 many regions of the world when associated with sulfur deposition (Vitousek et al., 1997); and
3 (5) alteration of ecosystem processes such as nutrient and energy cycles through changes in the
4 functioning and species composition of beneficial soil organisms (Galloway and Cowling 2002).
5 Additional, indirect detrimental effects of excess Nr on societal values include: (1)
6 increases in fine PM resulting in regional hazes that decrease visibility at scenic rural and urban
7 vistas and airports (discussed above in section 6.2); (2) depletion of stratospheric ozone by N2O
8 emissions which can in turn affect ecosystems and human health; (3) global climate change
9 induced by emissions of N2O (Galloway et al., 2003); (4) formation of O3 and ozone-induced
10 injury to crops, forests, and natural ecosystems and the resulting predisposition to attack by
11 pathogens and insects, as well as human health related impacts (EPA, 1996); (5) decrease in
12 quantity or quality of available critical habitat for threatened and endangered species (Fenn et al.,
13 2003); and (6) alteration of fire cycles in a variety of ecoystem types (Fenn et al., 2003).
14 A number of the more significant effects of chronic, long-term deposition of Nr on
15 terrestrial and aquatic ecosystems will be discussed below, specifically those effects which seem
16 to pose the greatest long-term risks to species or ecosystem health and sustainability or that
17 threaten ecosystem flows of goods and services important to human welfare.
18 Nitrogen Saturation of Terrestrial Ecosystems
19 Long-term, chronic additions of Nr (including nitrate deposition from ambient PM) to
20 terrestrial ecosystems is resulting in numerous ecosystems shifting to a detrimental ecological
21 condition known as "nitrogen saturation." Nitrogen saturation does not occur at a specific point
22 in time, but is a set of gradually developing critical changes in ecosystem processes which
23 represent the integrated response of a system to increased nitrogen availability over time (Aber,
24 1992). It occurs when nitrogen inputs exceed the capacity of plants and soil microorganisms to
25 utilize and retain the nitrogen (Aber et al., 1989, 1998; Garner, 1994; EPA, 1993). Under
26 conditions of nitrogen saturation, some other resource generally replaces nitrogen in limiting
27 biotic functions. The appearance of nitrate in soil solution (leaching) is an early symptom of
28 excess Nr accumulation.
29 Not all vegetation, organisms, or ecosystems react in the same manner to increased Nr
30 availability from atmospheric deposition. This is due in part to the variation both within and
January 2005 6-35 Draft - Do Not Quote or Cite
-------
1 across species in their inherent capacity to utilize additional Mr and the suite of other factors that
2 influence the range of community or ecosystem types possible at any given location. Such
3 factors can include the mineral composition of the underlying bedrock, the existing soil nutrient
4 pools, the local climatic.conditions including weather extremes such as drought, High/low
5 temperatures, topography, elevations, natural/land use history, and fire regimes.
6 In U.S. ecosystems, the nutrient whose supply most often sets the limit of possible
7 primary productivity at a given site is biologically available nitrogen. However, in any given
8 ecosystem, not all plants are equally capable of utilizing extra nitrogen. Those plants that are
9 predisposed to capitalize on any increases in Nr availability gain an advantage over those that are
10 not as responsive to added nutrients. Over time, this shift in the competitive advantage may lead
11 to shifts in overall plant community composition. Whether or not this shift is considered adverse
12 would depend on the management context within which that ecosystem falls and the ripple
13 effects of this shift on other ecosystem components, essential ecological attributes (EEAs), and
14 ecosystems. '
15 The effect of additions of nitrates on plant community succession patterns and
^ 16 biodiversity has been studied in several long-term nitrogen fertilization studies in both the U.S.
(
17 and Europe. These studies suggest that some forests receiving chronic inputs of nitrogen may
18 decline in productivity and experience greater mortality (Fenn et al. 1998). For example,
19 fertilization and nitrogen gradient experiments at Mount Ascutney, VT suggest that nitrogen
20 saturation may lead to the replacement of slow-growing, slow nitrogen-cycling spruce-fir forest
21 stands by fast-growing deciduous forests that cycle nitrogen rapidly (Fenn et al. 1998).
22 Similarly, experimental studies of the effects of Nr deposition overa 12-year period on
23 Minnesota grasslands dominated by native warm-season grasses observed the shift to low-
24 diversity mixtures dominated by cool-season grasses at all but the lowest rates of Nr addition
25 (Wedin and Tilman, 1996). The shift to low-diversity mixtures was associated with the decrease
26 in biomass carbon to N (C:N) ratios, increased Nr mineralization, increased soil nitrate, high
27 nitrogen losses, and low carbon storage. Grasslands with high nitrogen retention arid carbon
28 storage rates were the most vulnerable to loss of species and major shifts in nitrogen cycling.
29 (Wedin and Tilman, 1996).
t
January 2005 6-36 Draff - Do Not Quote or Cite
-------
1 The carbon-to-nitrogen (C:N) ratio of the forest floor can be changed by nitrogen
2 deposition over time. In Europe, low C:N ratios coincide with high deposition regions. A strong
3 decrease in forest floor root biomass has also been observed with increased nitrogen availability,
4 and appears to occur when the ecosystem becomes nitrogen saturated. If root growth and
5 mycorrhizal formation are impaired by excessive nitrogen deposition, the stability of the forest
6 floor vegetation may be affected. The forest floor C:N ratio has been used as a rough indicator
7 of ecosystem nitrogen status in mature coniferous forests and the risk of nitrate leaching. Nitrate
8 leaching has been significantly correlated with forest floor nitrate status, but not with nitrate
9 deposition. Therefore, to predict the rate of changes in nitrate leaching, it is necessary to be able
10 to predict the rate of changes in the forest floor C:N ratio. Understanding the variability in forest
11 ecosystem response to nitrogen input is essential in assessing pollution risks (Gundersen et al,
12 1998; CD, pp. 4-106,4-107).
13 In the U.S., forests that are now showing severe symptoms of nitrogen saturation include:
14 the northern hardwoods and mixed conifer forests in the Adirondack and Catskill Mountains of
15 New York; the red spruce forests at Whitetop Mountain, Virginia, and Great Smoky Mountains
16 National Park, North Carolina; mixed hardwood watersheds at Femow Experimental Forest in
17 West Virginia; American beech forests in Great Smoky Mountains National Park, Tennessee;
18 mixed conifer forests and chaparral watersheds in southern California and the southwestern
19 Sierra Nevada in Central California; .the alpine tundra/subalpine conifer forests of the Colorado
20 Front Range; and red alder forests in the Cascade Mountains in Washington. All these systems
21 have been exposed to elevated nitrogen deposition, and nitrogen saturated watersheds have been
22 reported in the above-mentioned areas. Annual nitrogen additions through deposition in the
23 southwestern Sierra Nevada are similar in magnitude to nitrogen storage in vegetation growth
24 increments of western forests, suggesting that current nitrogen deposition rates may be near the
25 assimilation capacity of the overstory vegetation. Ongoing urban expansion will increase the
26 potential for nitrogen saturation of forests from urban sources (e.g., Salt Lake City, Seattle,
27 Tucson, Denver, central and southern California) unless there are improved emission controls
28 (Fenn et al., 1998).
January 2005
6-37
Draft - Do Not Quote or Cite
t
-------
'1 The composition and structure of the plant community within an ecosystem in large part
2 determines the food supply and habitat types available for use by other organisms. In terrestrial
3 systems, plants serve as the integrators between above-ground and below-ground environments
4 and are influenced by and influence conditions in each. It is because of these linkages that
5 chronic excess Nr additions can lead to complex, dramatic, and severe ecosystem level/wide
6 changes/responses. Changes in soil Nr influence below ground communities as well. A
7 mutualistic relationship exists in the rhizosphere (plant root zone) between plant roots, fungi, and
8 microbes. Because the rhizosphere is an important region of nutrient dynamics, its function is
9 critical for the growth of the organisms involved. The plant roots provide shelter and carbon for
10 the symbionts, whereas the symbionts provide access to limiting nutrients such as nitrogen and
11 phosphorus for the plant. Bacteria make N, S, Ca, P, Mg, and K available for plant use while
12 fungi in association with plant roots form mycorrhizae that are essential in the uptake by plants
13 of mineral nutrients, such as N and P (Section 4.3.3; Wall and Moore, 1999; Rovira and Davy,
14 1974). Mycorrhizal fungal diversity is associated with above-ground plant biodiversity,
15 ecosystem variability, and productivity (Wall and Moore, 1999). Studies suggest that during
16 nitrogen saturation, soil microbial communities change from being predominately fungal, and
17 dominated by mycdrrhizae, to being dominated by bacteria (Aber et al., 1998; CD, pp. 4-107,4-
18 108), dramatically affecting both above- and belowiground ecosytems. These types of effects
19 have been observed in the field. For example, the coastal sage scrub (CSS) community in
20 California has been declining in land area and in drought deciduous shrub density over the past
21 60 years, and is being replaced in many areas by Mediterranean annual grasses. At the same
22 time, larger-spored below-ground fungal species (Scutellospora and Gigaspord), due to a failure
23 to sporulate, decreased in number with a concomitant proliferation of small-spored species of
24 Glomus aggregatum, G leptotichum, and G. geosporum, indicating a strong selective pressure
25 for the smaller spored species of fungi (Edgerton-Warburton and Allen, 2000). These results
26 demonstrate that nitrogen enrichment of the soil significantly alters the arbuscular mycorrhizal
27 species composition and richness, and markedly decreases the overall diversity of the arbuscular
28 mycorrhizal community. The decline in the coastal sage scrub species can be directly linked to
f
January 2005 6-3 8 Draft - Do Not Quote or Cite
-------
1 the decline of the arbuscular mycorrhizal community (Edgerton-Warburton and Allen, 2000;
2 Allen et al., 1998; Padgett et al., 1999)(CD, pp. 4-108,4-109). -
3 Impacts on threatened and endangered species. In some rare and unique U.S.
4 ecosystems, the chronic additions of atmospherically-derived nitrogen have already had some
5 dire and perhaps irreversible consequences. For example, California has many species that occur
6 in shrub, forb, and grasslands affected by N deposition, with up to 200 sensitive plant species in
7 southern California CSS alone (Skinner and Pavlik, 1994). Some 25 plant species are already
8 extinct in California, most of them annual and perennial forbs that occurred in sites now
9 experiencing conversion to annual grassland. As CSS converts more extensively to annual
10 grassland dominated by invasive species, loss of additional rare species may be inevitable.
11 Though invasive species are often identified as the main threat to rare species, it is more likely
12 that invasive species combine with other factors, such as excess N deposition, to promote
13 increased productivity of invasive species and resulting species shifts.
14 Not surprisingly, as sensitive vegetation is lost, wildlife that depend on these plants are
15 adversely affected. Included among these wildlife species are several threatened or endangered
16 species listed by the U.S. Fish and Wildlife Service, such as the desert tortoise and checkerspot
17 butterfly. A native to San Francisco Bay area, the bay checkerspot butterfly (Euphydryas editha
18 bayensis), has been declining steadily over the past decade, with local extirpations in some
19 reserves. This decline has been associated with the invasion of exotic grasses replacing the
20 native forbs on which the butterfly depends. In particular, the larval stage is dependent on
21 primarily one host plant, Plantago erecta, which is increasingly being out-competed by exotic
22 grasses.
23 Similarly, the desert tortoise has declined due to a number of co-occurring stresses,
24 including grazing, habitat destruction, drought, disease, and a declining food base. In the desert
25 shrub inter-spaces, sites where native forbs once flourished, invasive grasses now dominate,
26 reducing the nutritional quality of foods available to the tortoise (Fenn et al., 2003; Nagy et al.,
27 1998). Nitrogen deposition contributes to the productivity and density of N-fertilized grasses at
28 the expense of native forbs (Brooks, 2003). "Thus, protection of endangered species will
29 require increased exotic grass control, but local land management strategies to protect these
January 2005 • 6-39 Draft - Do Not Quote or Cite
f
-------
1 endangered species may not succeed unless they are accompanied by policy changes at the
2 regional or national level that reduce air pollution" (Fenn et al., 2003).
3 Community composition of epiphytic lichens is readily altered by small increases in
4 nitrogen deposition, an effect that seems to be widespread in the West (Fenn et al., 2003). Most
5 epiphytic lichens meet their nutritional requirements from atmospheric deposition and can store
6 N in excess of their nutritional needs (van Herk, 1999). In the San Bernardino Mountains, up to
7 50% of the lichen species that occurred in the region in the early 1900s have disappeared, with a
8 disproportionate number of the locally extinct species being (epiphytic) cyanolichens (Fenn et
9 al., 2003; Nash and Sigal, 1999). The Pacific Northwest, in contrast, still has widespread
10 populations of pollution-sensitive lichens (Fenn et al., 2003). However, in urban areas, intensive
11 agricultural zones and downwind of major urban and industrial centers, there is a sparsity of
12 sensitive lichen species and high levels of N concentrations have been measured in lichen tissue
13 (Fenn et al., 2003). Replacement of sensitive lichens by nitrophilous species has'undesirable
14 ecological consequences. In late-successional, naturally N-limited forests of the Coast Range
15 and western Cascades, epiphytic cyanolichens make important contributions to mineral cycling
16 and soil fertility (Pike 1978, Sollins et al., 1980, Antoine, 2001), and together with other large,
17 pollution-sensitive macrolichens, are an integral part of the food web for large and small
18 mammals, insects and birds (McCune and Geiser, 1997).
19 Alteration of native fire cycles. Several lines of evidence suggest that N deposition may
20 be contributing to greater fuel loads and thus altering the fire cycle in a variety of ecosystem
21 types, although further study is needed (Fenn et al., 2003). Invasive grasses promote a rapid fire
22 cycle in many locations (D'Antonio and Vitousek, 1992). The increased productivity of
23 flammable understory grasses increases the spread of fire and has been hypothesized as one
24 mechanism for the recent conversion of CSS to grassland (Minnich and Dezzani, 1998).
25 Thus, through its effect on habitat suitability, genetic diversity, community dynamics and
26 composition, nutrient status, energy and nutrient cycling, and frequency and intensity of natural
27 disturbance regimes (fire), excess Nr deposition is having profound and adverse impact on the
28 essential ecological attributes associated with terrestrial ecosystems. Strong correlation between
29 the stressor and adverse environmental response exists in many locations, and N-addition studies
January 2005 . 6-40 Draft - Do Not Quote or Cite
-------
1 have confirmed this relationship between stressor and response. Loss of species and genetic
2 diversity are clearly adverse ecological effects and adverse to the public welfare. Research
3 efforts should be made to elucidate what role paniculate deposition is playing in contributing to
4 these effects so as to facilitate the mitigation of such effects.
5 Effects of Nitrogen Addition on Aquatic Habitats
6 Aquatic ecosystems (streams, rivers, lakes, estuaries or oceans) receive increased
7 nitrogen inputs either from direct atmospheric deposition (including nitrogen-containing
8 particles), surface runoff, or leaching from nitrogen saturated soils into ground or surface waters.
9 The primary pathways of Nr loss from forest ecosystems are hydrological transport beyond the
10 rooting zone into groundwater or stream water, or surface flows of organic nitrogen as nitrate
11 and Nr loss associated with soil erosion (Fenn et al., 1998). In the east, high nitrate
12 concentrations have been observed in streams draining nitrogen saturated watersheds in the
13 southern Appalachian Mountains (Fenn et al., 1998). The Great Smoky Mountains National
14 Park in Tennessee and North Carolina receives elevated levels of total atmospheric deposition of
15 sulfur and nitrogen. A major portion of the atmospheric loading is from dry and cloud
16 deposition. Nitrogen saturation of the watershed resulted in extremely high exports of nitrate
17 and promoted both chronic and episodic stream acidification in streams draining undisturbed
18 watersheds. Significant export of base cations was also observed (CD, pp. 4-110,4-111; see also
19 section 6.3.3.2 on acidification from PM deposition).
20 In the west, the Los Angeles Air Basin exhibited the highest stream water NCy
21 concentrations in wilderness areas of North America (Bytnerowicz and Fenn, 1996; Fenn et al.,
22 1998). Chronic N deposition in southern California, in the southwestern Sierra Nevada, and in
23 the Colorado Front Range leads to increased net N mineralization and nitrification rates in soil
24 and to elevated NO3" concentrations in lakes and streams. These symptoms occur in low- and
25 mid-elevation, high-deposition areas (>15 kg N/ha/yr) and in high elevation sites with relatively
26 low N deposition (4 to 8 kg N/ha/yr) but little capacity to assimilate and retain added N.
27 Estuaries are among the most intensely fertilized systems on Earth (Fenn et al., 1998).
28 They receive far greater nutrient inputs than other systems. In the Northeast, for example,
29 nitrogen is the element most responsible for eutrophication in coastal waters of the region. Since
30 the early 1900s, there has been a 3r to 8-fold increase in nitrogen flux fromlO watersheds in the
January 2005 - 6-41 Draft - Do Not Quote or Cite
f
-------
northeast These increases are associated with nitrogen oxide emissions from combustion which
2 have increased 5-fold. Riverine nitrogen fluxes have been correlated with atmospheric
3 deposition onto their landscapes and also with nitrogen oxides emissions into their airsheds.
4 Data from 10 benchmark watersheds with good historical records indicate that about 36-80% of
5 the riverine total nitrogen export, averaging approximately 64%, was derived directly or
6 indirectly from nitrogen oxide emissions (CD, pp. 4-109,4-110).
7 The Pamlico Sound, NC estuarine complex, which serves as a key fisheries nursery
8 supporting an estimated 80% of commercial and recreational finfish and shellfish catches in the
9 southeastern U.S. Atlantic coastal region, has also been the subject of recent research (Paerl et
10 al., 2001) to characterize the effects of nitrogen deposition on the estuary. Direct atmospheric
11 nitrogen deposition onto waterways feeding into the Pamlico Sound or onto the Sound itself and
12 indirect nitrogen inputs via runoff from upstream watersheds contribute to conditions of severe
13 water oxygen depletion; formation of algae blooms in portions of the Pamlico Sound estuarine
14 complex; altered fish distributions, catches, and physiological states; and increases in the
15 incidence of disease. Especially under extreme rainfall events (e.g., hurricanes), massive -
16 influxes of nitrogen (in combination with excess loadings of metals or other nutrients) into
17 watersheds and sounds can lead to dramati c decreases of oxygen in water and the creation of
18 widespread "dead zones" and/or increases in algae blooms that can cause extensive fish kills and
• ' * ' -
19 damage to commercial fish and sea food harvesting (Paerl et al., 2001; CD, pp. 4-109,4-110).
20 6.3.3.2 Environmental Effects of PM-Related Acidic Deposition.
21 <' Acidic deposition has emerged over the past quarter century as a critical environmental
22 stress that affects diverse terrestrial and aquatic ecosystems in North America, Europe, and Asia
23 (Driscoll et al., 2001). In the eastern U.S. for example, the current acidity in precipitation is at
24 least twice as high as in pre-industrial times, with mean precipitation pH ranges from 4.3
25 (Pennsylvania and New York) to 4.8 (Maine) (EPA, 2003). Acidic deposition is highly variable
26 across space and time, can originate from transboundary air pollution, can travel hundreds of
27 miles before being deposited, thereby affecting large geographic areas. It is composed of ions,
28 gases, and particles derived from the precursor gaseous emissions of SO2, NOX, NH3 and
January 2005 6-42 Draft - Do Not Quote or Cite
-------
1 particulate emissions of other acidifying compounds. Acid deposition disturbs forest and aquatic
2 ecosystems by giving rise to harmful chemical conditions (Dricoll et al., 2001).
3 Terrestrial Effects
4 Acidic deposition has changed the chemical composition of soils by depleting the content
5 of available plant nutrient cations (e.g., Ca2+, Mg2+, K+) by increasing the mobility of Al, and by
6 increasing the S and N content (Driscoll et al., 2001). Soil leaching is often of major
7 importance in cation cycles, and many forest ecosystems show a net loss of base cations (CD, pp.
8 4-118). In acid sensitive soils, mineral weathering (the primary source of base cations in most
9 watersheds) is insufficient to keep pace with leaching rates accelerated by acid deposition
10 (Driscoll et al.3 2001).
11 In the absence of acid deposition, cation leaching in northeastern forest soils is driven
12 largely by naturally occurring organic acids derived from the decomposition of organic matter.
13 Organic acids tend to mobilize Al through formation of organic-Al complexes, most of which are
14 deposited lower in the soil profile through adsorption to mineral surfaces. This process, termed
15 podzolization, results in surface waters with low concentrations of Al. Such concentrations are
16 primarily in a nontoxic, organic form (Driscoll et al., 1998). Acid deposition, however, has
17 altered podzolization by solubilizing Al with mobile inorganic anions, facilitating the transport .
18 of inorganic Al into surface waters. In forest soils with base saturation values less than 20%,
19 acidic deposition leads to increased Al mobilization and a shift in chemical speciation of Al from
20 organic to inorganic forms that are toxic to terrestrial and aquatic biota.
21 The toxic effect of Al on forest vegetation is attributed to its interference with plant
22 uptake of essential nutrients, such as Ca and Mg. Because Ca plays a major role in cell
23 membrane integrity and cell wall structure, reductions in Ca uptake suppress cambial growth,
24 reduce the rate of wood formation, decrease the amount of functional sapwood and live crown,
25 and predispose trees to disease and injury from stress agents when the functional sapwood
26 becomes less than 25% of cross sectional stem area (Smith, 1990a). There are large variations in
27 Al sensitivity among ecotypes, between and within species, due to differences in nutritional
28 demands and physiological status, that are related to age and climate, which change over time
29 (CD, pp. 4-126).
January 2005 ' 6-43 Draft - Do Not Quote or Cite
f
-------
N1 Acidic deposition has been firmly implicated as a causal factor in the northeastern high-
2 elevation decline of red spruce (DeHayes et al., 1999). Red spruce is common in Maine, where
3 it is an important commercial species. It is also common at high elevations in mountainous
4 regions throughout the Northeast, where it is valued for recreation and aesthetics, as well as for
5 providing a habitat for unique and endangered species. Dieback has been most severe at high
6 elevations in the Adirondack and Green Mountains, where more than 50% of the canopy trees
7 died during the 1970s and 1980s. In the White Mountains, about 25% of the canopy spruce died
8 during that same period (Craig and Friedland 1991). Dieback of red spruce trees has also been
9 observed in mixed hardwood-conifer stands at relatively low elevations in the western
10 Adirondack Mountains, areas that receive high inputs of acidic deposition (Shortie et al., 1997).
11 Results of controlled exposure studies show that acidic mist or cloud water reduces the cold
12 tolerance of current-year red spruce needles by 3-10 degrees C (DeHayes et al., 1999). This
13 increased susceptibility to freezing occurs due to the loss of membrane-associated Ca2+ from
14 needles through leaching caused by the hydrogen ion. The increased frequency of winter injury
15 in the Adirondack and Green Mountains since 1955 coincides with increased exposure of red
16 spruce canopies to highly acidic cloud water (Johnson et al., 1984). Recent episodes of winter
17 injury have been observed throughout much of the range of red spruce' in the Northeast
18 (DeHayes et al., 1999). DeHayes etal. (1999) indicate that there is a significant positive
19 association between cold tolerance and foliar calcium in trees exhibiting deficiency in foliar
20 calcium, and further state that their studies raise the strong possibility that acid rain alteration of
21 foliar calcium is not unique to red spruce but has been demonstrated in many other northern
22 temperate forest tree species including yellow birch (Betula alleghaniensis), white spruce (Picea
23 r glaucus), red maple (Acer rubrum) eastern white pine (Pinus strobus), and sugar maple (Acer
24 saccharum) (CD, p. 4-120).
25 Although less well established, there is also a strong possibility mat low Ca to Al ratios
26 in soils may also be impacting northeastern red spruce. Cronan and Grigal (1995) concluded that
27 a Ca:Al ratio of less than 1.0 in soil water indicated a greater than 50% probability of impaired
28 growth in red spruce. They cite examples of studies from the northeast where soil solutions in
29 the field were found to exhibit Ca: Al ionic ratios less than 1.0.
January 2005 .
6-44
Draft - Do Not Quote or Cite
-------
1 Acidic deposition may also be contributing to episodic dieback of sugar maple in the
2 Northeast through depletion of nutrient cations from marginal soils. Horsley et al. (1999) found
3 that dieback at 19 sites in northwestern and north-central Pennsylvania and south-western New
4 York was correlated with combined stress from defoliation and deficiencies of Mg and Ca.
5 Dieback occurred predominately on ridgetops and on upper slopes, where soil base availability
6 was much lower than at mid and low slopes of the landscape (Bailey et al., 1999). Because
7 multiple factors such as soil mineralogy and landscape position affect soil base status, the extent
8 to which sugar maple dieback can be attributed to acidic deposition is not clear.
9 Less sensitive forests throughout the U.S. are experiencing gradual losses of base cation
10 nutrients, which in many cases will reduce the quality of forest nutrition over the long term
11 (National Science and Technology Council, 1998). In some cases, such effects may not even
12 take decades to occur because these forests have already been receiving S and N deposition for
13 many years.
14 In contrast to contributing to the adverse impacts of acid deposition, particles can also
15 provide a beneficial supply of base cations to sites with very low rates of supply from mineral
16 sources. In these areas, atmospheric inputs of bass cations can help ameliorate the acidifying
17 effects of acid particles. The Integrated Forest Study (IFS) (Johnson and Lindberg, 1992) has
18 characterized the complexity and variability of ecosystem responses to atmospheric inputs and
19 provided the most extensive data set available on the effects of atmospheric deposition, including
20 particle deposition, on the cycling of elements in forest ecosystems. This study showed that in
21 the IFS ecosystems, inputs of base cations have considerable significance, not only for base
22 • cation status, but also for the potential of incoming precipitation to acidify or alkalize the soils.
23 The actual rates, directions, and magnitudes of changes that may occur in soils (if any), however,
24 will depend on rates of inputs from weathering and vegetation outputs, as well as deposition and
25 leaching. In other words, these net losses or gains of base cations must be placed in the context
26 of the existing soil pool size of exchangeable base cations (CD, p. 4-132). Given the wide
27 ranges of paniculate deposition for each base cation across the IFS sites, however, the unique
28 characteristics of various sites need to be better understood before assumptions are made about
29 the role particulate pollution plays in ecosystem impacts (CD, pp. 4-127,4-128).
January 2005
6-45
Draft - Do Not Quote or Cite
-------
\m
1 In a follow up study, Johnson et al. (1999) used the nutrient cycling moded, NuCM, to
2 simulate the effects of reduced S, N, and base cation (CB) deposition on nutrient pools, fluxes,
3 soil, and soil solution chemistry in two contrasting southern Appalachian forest ecosystems. The
4 authors found that in an extremely acidic system, CB deposition can have a major effect on CB
5 leaching through time and S and N deposition had a major effect on Al leaching. At the less
6 acidic Coweeta site, CB deposition had only a minor effect on soils and soil solutions; whereas S
7 and N deposition had delayed but major effects on CB leaching (CD, pp. 4-136,4-137).
8 Aquatic Effects
9 Inputs of acidic deposition to regions with base-poor soils have resulted in the
10 acidification of soil waters, shallow ground waters, streams, and lakes in anumber of locations
11_ within the U.S. In addition, perched seepage lakes, which derive water largely from direct
12 precipitation inputs, are highly sensitive to acidic deposition (Charles, 1991). These processes
13 usually result in lower pH and, for drainage lakes, higher concentrations of inorganic monomeric
14 Al. Such changes in chemical conditions are toxic to fish and other aquatic animals. (Driscoll et
15 al., 2001).
16 A recent report, Response of Surface Water Chemistry to the Clean Air Act of 1990
17 (EPA, 2003), analyzes data from 1990 through 2000 obtained from EPA's Long Term
18 .Monitoring (LTM) and Temporally Integrated Monitoring of Ecosystems (TIME) projects, part
19 of EMAP (Environmental Monitoring and Assessment Program). The report assesses recent
20 changes in surface water chemistry in response to changes in deposition, in the northern and
21 eastern U.S., specifically in the acid sensitive regions defined as New England (Maine, New
22 Hampshire, Vermont and Massachusetts), the Adirondack Mountains of New York, the Northern
23 Appalachian Plateau (New York, Pennsylvania and West Virginia), the Ridge and Blue Ridge
24 Provinces of Virginia, and the Upper Midwest (Wisconsin and Michigan). Acidic waters are
25 defined as having acid neutralizing capacity '(ANC) less than zero (i.e., no acid buffering
26 capacity in the water), corresponding to a pH of about 5.2. Increases in surface water ANC
27 values and/or pH would indicate improved buffering capacity and signal the beginning of
28 recovery (EPA, 2003).
29 Using National Atmospheric Deposition Program (NADP) data, trends in sulfate and N
30 (nitrate + ammonium) deposition were analyzed, along with CB deposition, sulfate and nitrate
January 2005 6-46 Draft - Do Not Quote or Cite
-------
1 concentrations in surface waters, ANC and pH levels. Over this timeframe, sulfate deposition
2 declined significantly across all regions, while N declined slightly in the Northeast and increased
3 slightly in the Upper Midwest. Base cation deposition showed no significant changes in the East
4 and increased slightly in the Upper Midwest. Concurrently, all regions except the Ridge/Blue
5 Ridge province in the mid-Atlantic showed significant declines in sulfate concentrations in
6 surface waters, while nitrate concentrations decreased in two regions with the highest ambient
7 nitrate concentrations (Adirondacks, Northern Appalachian Plateau) but were relatively
8 unchanged in regions with low concentrations.
9 Given the declines in S and N deposition measured for these areas, one would expect to
10 find increasing values of ANC, pH or both in response. ANC values did increase in the
11 Adirondacks, Northern Appalachian Plateau and Upper Midwest, despite a decline in base
12 cations (Ca and Mg) in each region. The loss of base cations limited the extent of ANC and pH
13 increase. Toxic Al concentrations also declined slightly in the Adirondacks. In New England
14 and Ridge/Blue Ridge, however, regional surface water ANC did not change significantly (EPA,
15 2003).
16 Modest increases in ANC have reduced the number of acidic lakes and stream segments
17 in some regions. There are an estimated 150 Adirondack lakes with ANC less than 0, or 8.1% of
18 the population, compared to 13% (240 lakes) in the early 1990s. In the Upper Midwest, an
19 estimated 80 of 250 lakes that were acidic in mid-1980s are no longer acidic. TIME surveys of
20 streams in the Northern Appalachian Plateau region estimated that 8.5% (3,600 kilometers) of
21 streams remain acidic at the present time, compared to 12% (5,014 kilometers) of streams that
22 were acidic in 1993-94. In these three regions taken together, approximately one-fourth to one-
23 third of formerly acidic surface waters are no longer acidic, although still with very low ANC.
24 The report finds little evidence of regional change in the acidity status of New England or the
25 Ridge/Blue Ridge regions and infers that the numbers of acidic waters remain relatively
26 unchanged. Despite a general decline in base cations and a possible increase in natural organic
27 acidity, there is no evidence that the number of acidic waters have increased in any region (EPA,
28 2003).
January 2005 6-47 Draft - Do Not Quote or Cite
-------
1 Acidification has marked effects on the trophic structure of surface waters. Decreases in
2 pH and increases in Al concentrations contribute to declines in species richness and in the
3 abundance of zooplankton, macroinvertebrates, and fish (Schindler et al.,1985; Keller and Gunn
4 1995). Numerous studies have shown that fish species richness (the number offish species in a
5 water body) is positively correlated with pH and ANC values (Rago and Wiener, 1986 Kretser et
6 al.j 1989). Decreases in pH result in decreases in species richness by eliminating acid-sensitive
7 species (Schindler et al. 1985). Of the 53 fish species recorded by the Adirondack Lakes Survey
8 Corporation, about half (26 species) are absent from lakes with pH below 6.0. Those 26 species
9 include important recreational fishes, such as Atlantic salmon, tiger trout, redbreast sunfish,
10 bluegill, tiger musky, walleye, alewife, and kokanee (Kretser et al. 1989), plus ecologically
11 important minnows that serve as forage for sport fishes.
«
12 A clear link exists between acidic water, which results from atmospheric deposition of
13 strong acids, and fish mortality. The Episodic Response Project (ERP) study showed that
14 streams with moderate to severe acid episodes had significantly higher fish mortality during
15 bioassays than nonacidic streams (Van Sickle et al., 1996). The concentration of inorganic
16 monomeric Al was the chemical variable most strongly related to mortality in the four test
17 species (brook trout, mottled sculpin, slimy sculpin, and blacknose dace). The latter three
18 species are acid sensitive. In general, trout abundance was lower in ERP streams with median
19 episode pH less than 5.0 and inorganic monomeric Al concentrations greater than 3.7-7.4 mmol
20 L"1. Acid sensitive species were absent from streams with median episode pH less than 5.2 and
21 with a concentration of inorganic monomeric Al greater than 3.7 mmol L'1..
22 Given the significant reductions in sulfur emissions that have occurred in the U. S. and
23 Europe in recent decades, the findings of Driscoll et al. (1989,2001 j and Hedin et al. (1994) are
24 especially relevant. Driscoll et al. (1989, 2001) noted a decline in both SO4"2 and base cations in
25 both atmospheric deposition and stream water over the past two decades at Hubbard Brook
26 Watershed, NH. However, the reductions in S02 emissions in Europe and North America in
27 recent years have not been accompanied by equivalent declines in net acidity related to sulfate in
28 precipitation, and may have, to varying degrees, been offset by steep declines in atmospheric
29 base cation concentrations over the past 10 to 20 years (Hedin et al., 1994).
January 2005 6-48 Draft - Do Not Quote or Cite
-------
1 Driscoll et al. (2001) envision a recovery process that will involve two phases. Initially,
2 a decrease in acidic deposition following emissions controls will facilitate a phase of chemical
3 recovery in forest and aquatic ecosystems. Recovery time for this phase will vary widely across
4 ecosystems and will be a function of the following:
5 • the magnitude of decreases in atmospheric deposition
6 • the local depletion of exchangeable soil pools of base cations
7 • the local rate of mineral weathering and atmospheric inputs of base cations
8 • the extent to which soil pools of S and N are released as SO42" or as NO3" to drainage
9 waters and the rate of such releases (Galloway et al. 1983),
10
11 In most cases, it seems likely that chemical recovery will require decades, even with additional
12 controls on emissions. The addition of base cations, e.g., through liming, could enhance
13 chemical recovery at some sites.
14 The second phase in ecosystem recovery is biological recovery, which can occur only if
15 chemical recovery is sufficient to allow survival and reproduction of plants and animals. The
16 time required for biological recovery is uncertain. For terrestrial ecosystems, it is likely to be at
17 least decades after soil chemistry is restored because of the long life of tree species and the
18 complex interactions of soil, roots, microbes, and soil biota. For aquatic systems, research
19 suggests that stream macroinvertebrate populations may recover relatively rapidly
20 (approximately 3 years), whereas lake populations of zooplankton are likely to recover more
21 slowly (approximately 10 years) (Gunn and Mills 1998). Some fish populations may recover in
22 5 to 10 years after the recovery of zooplankton populations. Stocking could accelerate fish
23 population recovery (Driscoll et al., 2001)
24 Projections made using an acidification model (PnET-BGC) indicate that full
25 implementation of the 1990 CAAA will not afford substantial chemical recovery at Hubbard
26 Brook EF and at many similar acid-sensitive locations (Driscoll et al., 2001) . Model
27 calculations indicate that the magnitude and rate of recovery from acidic deposition in the
28 northeastern U.S. are directly proportional to the magnitude of emissions reductions. Model
29 evaluations of policy proposals calling for additional reductions in utility SO2 and NOX
January 2005
6-49
Draft - Do Not Quote or Cite
-------
t
1' emissions, year round emissions controls, and early implementation indicate greater success in
2 facilitating the recovery of sensitive ecosystems (Driscoll et al., 2001).
3 Indirect Vegetation and Ecosystem Effects from Atmospheric PM
4 In addition to the direct and indirect effects of deposited PM, ambient atmospheric PM
5 can effect radiation and climate conditions that influence overall plant/ecosystem productivity.
6 The degree to which these effects occur in any given location will depend on the chemical and
7 physical composition and concentration of the ambient PM. Because plants are adapted to the
8 overall light and temperature environments in which they grow, any PM-related changes to these
9 conditions potentially alter the overall competititive success these plants will have in that
10 ecosystem.
11 With respect to radiation, the characteristics and net receipts of solar and terrestrial
12 radiation determine rates of both photosynthesis and the heat-driven process of water cycling.
13 Atniospheric turbidity (the degree of scattering occurring in the atmosphere due to particulate
14 loading) influences the light environment of vegetative canopy in two ways: through conversion
15 of direct to diffuse radiation and by scattering or reflecting incoming radiation back out into
16 space. Diffuse radiation increases canopy photosynthetic productivity by distributing radiation
17 more uniformly throughout the canopy so that it also reaches the lower leaves and improves the
18 canopy radiation use efficiency (RUE). Acting in the opposite direction, non-absorbing,
19 scattering aerosols present in PM reduce the overall amount of radiation reaching vegetative
20 surfaces, by scattering or reflecting it back into space. It appears that global albedo has been
21 increasing due to an increasing abundance of atmospheric particles. Using World
22 Meteorological Organization (WMO) data, Stanhill and Cohen (2001) have estimated that
23 average solar radiation receipts have declined globally by an average of 20 W m-2 since 195 8.
24 The net effect of atmospheric particles on plant productivity is not clear, however, as the
25 enrichment in photosynthetically active radiation (PAR) present in diffuse radiation may offset a
26 portion of the effect of decreased solar radiation receipts in some instances (CD, pp.' 4-92,4-93).
1 t
27 Plant processes also are sensitive to temperature. Some atmospheric particles (most
28 notably black carbon) absorb short-wavelength solar radiation, leading to atmospheric heating
29 and reducing total radiation received at the surface. Canopy temperature and transpirational
30 water use by vegetation are particularly sensitive to long-wave, infrared radiation. Atmospheric
January 2005 6-50 Draft - Do Not Quote or Cite
-------
^ \
1 heating by particles can potentially reduce photosynthetic water uptake efficiency and vertical
2 temperature gradients, potentially reducing the intensity of atmospheric turbulent mixing,
3 Stanhill and Cohen (2001) suggested that plant productivity is more affected by changes in
4 evapotranspiration induced by changes in the amount of solar radiation plants receive than by
5 changes in the amount of PAR plants receive (CD, p. 4-93).
6
• 7 63.4 Characteristics and Location of Sensitive Ecosystems in the U.S.
8 Ecosystems sensitive to anthropogenically derived nitrogen and/or acid deposition tend to
9 have similar characteristics. Some of these ecosystems and characteristics have already been
10 mentioned in earlier sections but are repeated here to provide a more comprehensive list that can
11 help ecological risk assessors/managers identify areas of known or potential concern. For
12 example, lower nitrogen and/or resource environments, such as those with infertile soils, shaded
13 understories, deserts, or tundras, are populated with organisms specifically adapted to survive
14 under those conditions. Plants adapted to these conditions have been observed to have similar
15 characteristics, including inherently slower growth rates, lower photosynthetic rates, and lower
16 capacity for nutrient uptake, and grow in soils with lower soil microbial activity. When N
17 becomes more readily available, such plants will be replaced by nitrophilic plants which are
18 better able to use increased amounts of Nr (Fenn et al., 1998).
19 Additionally, in some instances, there seem to be important regional distinctions in
20 exposure patterns, environmental stressors, and ecosystem characteristics between the eastern
21 A and western U.S.. A seminal report describing these distinctive characteristics for the western
22 U.S. (11 contiguous states located entirely west of the 100th meridian) is Fenn et al., 2003.
23 In the western U.S., vast areas receive low levels of atmospheric deposition, interspersed
24 with hotspots of elevated N deposition downwind of large, expanding metropolitan centers or
25 large agricultural operations, hi other words, spatial patterns of urbanization largely define the
26 areas where air pollution impacts are most severe. The range of air pollution levels for western
27 wildlands is extreme, spanning from near-background to the highest exposures in all of North
28 America, with the possible exception of forests downwind of Mexico City. Over the same
29 geographic expanse, climatic conditions and ecosystem types vary widely. Some regions receive
January 2005 6-51 Draft - Do Not Quote or Cite
-------
7,
1
2
3
4
5
6
7
8
9
10
II
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
more than 1000 millimeters of precipitation, namely the Pacific coastal areas, the Sierra Nevada,
x
the Colorado Rockies, and northern Idaho, while other regions are arid or semiarid, with more
than 300 clear days per year (Riebsame et al., 1997). In these latter regions, the contribution of
atmospheric dry deposition is likely to be most important These characteristics which are
unique to the West require special consideration, and often make application of models and
ecological effects thresholds developed for other regions inappropriate.
In summary, sensitive or potentially sensitive ecosystems in the west include those that:
• are located downwind of large urban source areas; regions with a mix of emissions
sources that may include urban, mobile, agricultural, and industrial sources; and/or sites
near large point.sources of N.
• contain inherently N sensitive ecosystem components, such as lichens, diatoms, or poorly
buffered watersheds which produce high streamwater NO3- levels: These sensitive
components can be affected by N deposition rates as low as 3-8 kg/ha/yr.
• occur on top of siliclastic/crystalline bedrock with little potential for buffering acidity.
• are naturally nitrogen limited. For example, the approximately 16,000 high elevation
western mountain lakes are generally oligotrophic and especially sensitive to the effects
of atmospheric deposition.
A seminal report describing key characteristics of sensitive ecosystems for the eastern
and in particular the northeastern U.S. is Driscoll et al. (2001). In the northeastern United States,
atmospheric deposition is largely a regional problem. Because S and N most often occur
together in the eastern atmosphere and deposit to the environment as acidic deposition, acidic
deposition is seen as a critical environmental stress.
Several critical chemical thresholds appear to coincide with the onset of deleterious
effects to biotic resources resulting from acid deposition. Thus, ecosystems sensitive to
additional acid inputs include those with the following characteristics:
a molar Ca:Al ratio of soil water that is less than 1.0; ' •
• soil percentage base cation saturation less than 20%;
• surface water pH less than 6.0; *
• ANC less than 50 meqL-1; and
• concentrations of inorganic monomeric Al greater than 2 mmol L-l.
Knowledge of such indicators is necessary for restoring ecosystem structure and function.
January 2005
6-52
Draft - Do Not Quote or Cite
-------
1 6.3.5 Ecosystem Exposures to PM Stressor Deposition
2 In order for any specific chemical stressor present in ambient PM to impact ecosystems,
3 it must first be removed from the atmosphere through deposition. Deposition can occur in three
4 modes: wet (rain/frozen precipitation), dry, or occult (fog, mist or cloud). At the national scale,
5 all modes of deposition must be considered in determining potential impacts to vegetation and
6 ecosystems because each mode may dominate over specific intervals of time or space. (CD, p.
7 4-8 to 4-10). For example, in large parts of the western U.S. which are arid or semiarid, dry
8 deposition may be the source of most deposited PM (Fenn, et al., 2003). However, in coastal
9 areas or high elevation forests, wet or occult deposition may predominate. Where the latter is the
10 case, deposition levels may greatly exceed PM levels measured in the ambient air. Occult
11 deposition is particularly effective for delivery of dissolved and suspended materials to
12 vegetation because: (1) concentrations of ions are often many-fold higher in clouds or fog than in
13 precipitation or ambient air (e.g., acidic cloud water, which is typically 5-20 times more acid
14 than rainwater, can increase pollutant deposition and exposure to vegetation and soils at high
15 elevation sites by more than 50% of wet and dry deposition levels); (2) PM is delivered in a
16 hy drated and bioavailable form to foliar surfaces and remains hy drated due to conditions of high
17 relative humidity and low radiation; and (3) the mechanisms of sedimentation and impaction for
18 submicron particles that would normally be low in ambient air are increased. High-elevation
19 forests can be especially at risk from depositional impacts because they receive larger paniculate
20 deposition loadings than equivalent low-elevation sites, due to a number of orographic
21 (mountain related) effects. These orographic effects include higher wind speeds that enhance the
22 rate of aerosol impaction, enhanced rainfall intensity and composition, and increased duration of
23 occult deposition. Additionally, the needle-shaped leaves of the coniferous species often found
24 growing in these high elevation sites, enhance impaction and retention of PM delivered by all
25 three deposition modes (CD, pp. 4-29,4-44).
26 In order to establish exposure-response profiles useful in ecological risk assessments, two
27 types of monitoring networks need to be in place. First, a deposition network is needed that can
28 track changes in deposition rates of PM stressors (nitrates/sulfates) occurring in sensitive or
29 symptomatic areas/ecosystems. Secondly, a network or system of networks that measure the
January 2005
6-53
Draft - Do Not Quote or Cite
-------
1 response of key ecological indicators sensitive to changes in atmospheric deposition of PM
2 stressors is also needed.
3 ' Currently in the U.S., national deposition monitoring networks routinely measure total
4 wet or dry deposition of certain compounds. Atmospheric concentrations of dry particles began
5 to be routinely measured in 1986, with the establishment of EPA's National Dry Deposition
6 Network (NDDN). After new monitoring requirements were added in the 1990 CAAA, EPA, in
7 cooperation with the National Oceanic and Atmospheric Association, created the Clean Air
8 Status and Trends Network (CASTNet) from the NDDN. C ASTNet comprises 85 sites and is
9 considered the nation's primary source for atmospheric data to estimate concentrations for
10 ground-level ozone and the chemical species that make up the'dry deposition component of total
11 acid deposition (e.g., sulfate, nitrate, ammonium, sulfur dioxide, and nitric acid), as well as the
12 associated meteorology and site characteristics data that are needed to model dry deposition
13 velocities (CD. pg. 4-21: ftittp://.wvvw.e^p..a.g<)Tv/castneu')..
14 . To provide data on wet deposition levels in the U.S., the National Atmospheric
15 Deposition Program (NADP) was initiated in the late 1970's as a cooperative program between
16 federal, state, and other public and private groups. By the mid-1980's, it had grown to nearly
17 200 sites, and it stands today as the longest running national atmospheric deposition monitoring
18 network (http://nadp.sws.. liiuc. eduA.
19 In addition to these deposition monitoring networks, other networks collect data on
20 ambient aerosol concentrations and chemical composition. Such networks include the
21 IMPROVE network, discussed above in section 2.5, and the newly implemented PM25 chemical
22 Speciation Trends Network (STN) that consists of 54 core National Ambient Monitoring
23 Stations and approximately 250 State and Local Air Monitoring Stations.
24 Data from these deposition networks demonstrate that N and S compounds are being
25 deposited onto soils and aquatic ecosystems in sufficient amounts to impact ecosystems at local,
26 regional and national scales. Though the percentages of N and S containing compounds in PM
27 vary spatially and temporally, nitrates and sulfates make up a substantial portion of the chemical
28 composition of PM. In the future, speciated data from these networks may allow better
January 2005 6-54 Draft - Do Not Quote or Cite
-------
Y
1 understanding of the specific components of total deposition that are most strongly influencing
2 PM-related ecological effects.
3 Unfortunately, at this time there is only limited long-term ecosystem response monitoring
4 taking place at the national level. Two exceptions are the Hubbard Brook Experimental Forest
5 research site, that provides the longest continuous record of precipitation and stream chemistry in
6 the U.S. (Likens and Bormann, 1995) and EPA's LTM and TIME projects which monitor
7 changes in surface water chemistry in the acid sensitive regions of the northern and eastern U.S..
8 Because the complexities of ecosystem response make predictions of the magnitude and timing
9 of chemical and biotic recovery uncertain, it is strongly recommended that this lype of long-term
10 surface water chemistry monitoring network be continued, and that a biological monitoring
11 program be added. Data from these long-term monitoring sites will be invaluable for the
12 evaluation of the response of forested watersheds and surface waters to a host of research and
13 regulatory issues related to acidic deposition, including soil and surface water recovery, controls
14 on N retention, mechanisms of base cation depletion, forest health, sinks for S in watersheds,
15 changes in dissolved organic carbon and speciation of Al, and various factors related to climate
16 change (EPA, 2003).
17
18 63.6 Critical Loads
19 The critical load (CL) has been defined as a "quantitative estimate of an exposure to one
20 or more pollutants below which significant harmful effects on specified sensitive elements of the
21 environment do not occur according to present knowledge" (Lokke et al., 1996). The critical
22 load framework originated in Europe where the concept has generally been accepted as the basis
23 for abatement strategies to reduce or prevent injury to the functioning and vitality of forest
24 ecosystems caused by long-range transboundary chronic acidic deposition. The concept is
25 useful for estimating the amounts of pollutants that sensitive ecosystems can absorb on a
26 sustained basis without experiencing measurable degradation. The estimation of ecosystem
27 critical loads requires an understanding of how an ecosystem will respond to different loading
28 rates in die long term and is a direct function of the level of sensitivity of the ecosystem to the
29 pollutant and its capability to ameliorate pollutant stress.
January 2005 6-55 Draft - Do Not Quote or Cite
-------
1 Key to the establishment of a critical load is the selection of appropriate ecological
2 endpoints or indicators that are measurable characteristics related to the structure, composition,'
3 or functioning of ecological systems (i.e., indicators of condition). In Europe, the elements used
4 in the critical load concept are a biological indicator, a chemical criterion, and a critical value
5 (CD, p. 4-124). A number of different indicators for monitoring ecosystem status have been
6 proposed. Indicators of ecosystems at risk of N saturation could include: foliar nitrogen, nutrient
7 ratios (N:P, N:cation); foliar nitrate; foliar 515 N; arginine concentration; soil C:N ratio; NO3" in
8 soil extracts or increased and prolonged NO3~ loss below the main rooting zone and in stream
9 water or in soil solution; and flux rates of nitrogenous trace gases from soil (Fenn et al., 1998),
10 Seasonal patterns of stream water nitrate concentrations are especially good indicators of
11 watershed N status. Biological indicators that have been suggested for use in the critical load
12 calculation in forest ecosystems include mycorrhizal fungi (Lokke et al., 1996) and fine roots,
13 since they are an extremely dynamic component of below-ground ecosystems and can respond
14 rapidly to stress. The physiology of carbon allocation has also been suggested as an indicator of
15 anthropogenic stress (Andersen and Rygiewicz, 1991). Lichen community composition, in
16 terrestrial ecosystems or lichen N tissue levels are also fairly responsive to changes in N
17 deposition over time (Fenn et al., 2003). In aquatic systems, diatom species composition can be
18 a good indicator of changes in water chemistry (Fenn et al., 2003). It should be kept in mind,
19 however, that the response of a biological indicator is an integration of a number of different
20 stresses. Furthermore, there may be organisms more sensitive to the pollutant(s) than the species
21 selected (Lokke et al., 1996; National Science and Technology Council, 1998) (CD, pp. 4-124 to
22 126).
23 Within North America, a number of different groups have recently begun to use or
24 develop critical loads. As discussed below, these groups include the U.S. Federal Land
25 Managers (FLMs), such as the National Park Service and the Forest Service, a binational group
26 known as New England Governors/Eastern Canadian Premiers (NEG/ECP), and several
27 Canadian Provinces.
28 Federal Land Managers have hosted a number of meetings over the last few years to
29 discuss how the CL concept might be used in helping them fulfill their mandate of providing
January 2005
6-56
Draft - Do Not Quote or Cite
-------
1 protection for the lands they manage. In trying to develop a consistent approach to using CL; a
i
2 number of issues and considerations have been identified. First, the distinction between critical
3 loads (which are based on modeled or measured dose-response data) and target loads (which can
4 be based on political, economic, spatial or temporal considerations in addition to scientific
5 information) needs to be recognized. When using the critical or target load (TL) approach, one
6 must indicate the spatial (or geographic) scope, the temporal scope (timeframe to ecological or
7 ecosystem recovery), and a description of the sensitive receptors (or resource) to be protected,
8 the sensitive receptor indicators (physical, chemical biological, or social characteristics of the
9 receptor that can be measured), and the harmful effect on the receptor that is of concern.
10 Additionally, one would need to specify what is the "desired condition" that the critical or target
11 load is meant to achieve. For any given location, there may be a range or suite of possible
12 critical or target loads based on different sensitive receptors and/or receptor indicators found at
13 that site. Alternatively, one could focus on the most sensitive receptor and select a single CL or
14 TL for that receptor. Several aspects of the CL approach make it attractive for use by the FLMs.
15 Specifically, it can provide a quantitative, objective and consistent approach for evaluating
16 resource impacts. In an effort to progress the CL approach, the Forest Service is testing the
17 applicability of the European protocol to several U.S. case study sites.
18 Under the auspices of the NEG/ECP, and other binational efforts, Canadian and U.S.
19 scientists are involved in joint forest mapping projects. A Forest Mapping Work Group has been
20 tasked with conducting a regional assessment of the sensitivity of northeastern North American
21 forests to current and projected sulfur and nitrogen emissions levels, identifying specific forested
22 areas most sensitive to continued deposition and estimating deposition rates required to maintain
23 forest health and productivity. They have completed the development of methods, models and
24 mapping techniques, and identification of data requirements. Some of these data requirements
25 include: pollution loading to forest landscapes; the interaction of pollutants with forest canopies;
26 plant nutrient requirements; and the ability of soils to buffer acid inputs and replenish nutrients
27 lost due to acidification.
28 In addition to the CL measure, they have also defined a "deposition index" as the
29 difference between the CL and current deposition levels. Positive values of the index reflect the
January 2005
6-57
Draft - Do Not Quote or Cite
-------
1 capacity of a forest ecosystem to tolerate additional acidic deposition. Negative index values
2 correspond to the reduction in S and N deposition required to eliminate or deter the development
3 of future nutrient limitations. This allows an assessor to identify areas where the deposition
4 problems are most severe, and which sites might be under the CL level currently but not far from
5 reaching or exceeding that level should deposition levels increase. Currently maps exist for
6 Vermont and Newfoundland, though the goal is to develop maps that will cover Quebec and the
7 Atlantic provinces of Canada, along with the remaining New England states. These maps show
8 that 31% of Vermont forests and 23% of Newfoundland forests are sensitive (e.g., current levels
9 of S and N deposition are causing cation depletion).
10 Though these current activities hold promise for using the CLs approach in
11 environmental assessments and in informing management decisions, widespread use of CL's in
12 the U.S. is not yet possible. Critical loads is a very data-intensive approach, and, at the present
13' time, there is a paucity of ecosystem- level data for most sites. However, for a limited number of
14 areas which already have a long-term record of ecosystem monitoring, (e.g., Rocky Mountain
i
15 National Park in Colorado and me Lye Brook Wilderness in Vermont), FLMs may be able to
16 develop site-specific CLs. Further, in areas already exceeding the CL, it may be difficult to
17 determine what the management goals are/should be for each mapped area (e.g., what is the
18 "desired condition" or level of protection) without historic baseline data. More specifically, with
19 respect to PM deposition, there are insufficient data for the vast maj ority of U. S. ecosystems that
20 differentiate the PM contribution to total N or S depostion to allow for practical application'bf
21 this approach as a basis for developing national standards to protect sensitive U.S. ecosystems
22 from adverse effects related to PM deposition. Though atmospheric sources of Nr and acidifying
23 compounds, including ambient PM, are clearly contributing to the overall excess pollutant load
24 or burden entering ecosystems annually, insufficient data are available at this time to quantify
25 the contribution of ambient PM to total Nr or acidic deposition as its role varies both temporally
26 and spatially along with a number of other factors. Thus, it is not clear whether a CL could be
27 developed just for the portion of the total N or S input that is contributed by PM.
28
January 2005
6-58
Draft - Do Not Quote or Cite
-------
\
1 6.3.7 Summary and Conclusions
2 The above discussions identify a group of ecosystems known to be sensitive to excess N
3 and S inputs and a list of characteristics that can be used to predict or locate other potentially
4 sensitive ecosystems within the U.S. Further, exposures of these sensitive ecosystems to
5 atmospherically derived pollutants (e.g., N and S) have been measured and documented, in some
6 cases for decades. Clear linkages between reduced atmospheric concentrations of these
7 pollutants and reduced deposition rates have been made. The mechanisms of environmental and
8 ecosystem responses to these inputs are increasingly understood, though very complex.
9 Fertilization and acidification studies have verified observed ecosystem responses to these
10 pollutants in the field. Ecosystem-level effects associated with excess N and S inputs are
11 profound, but in most cases potentially reversible. New assessment and management tools, such
12 as critical and target loads, are being developed to better characterize the relationship between
13 deposition loads and ecosystem response. The success of these tools will depend on the
14 availability of sufficient ecosystem response data, which is currently limited to a few long-term
15 monitoring networks/sites (e.g., TIME/LTM). The current risk to sensitive* ecosystems and
16 especially sensitive species like the checkerspot butterfly, desert tortoise, epiphytic lichens,
17 native shrub and forb species, and aquatic diatom communities is high. The loss of species and
18 whole ecosystem types is adverse and should receive increased protection.
19 A number of ecosystem-level conditions (e.g., nitrogen saturation, terrestrial and aquatic
20 acidification, coastal eutrophication) have been associated with chronic, long-term exposure of
21 ecosystems to elevated inputs of compounds containing Nr, sulfur and/or associated hydrogen
22 ions. These ecosystem level changes profoundly impact almost all of the EEAs identified in the
.23 EPEC Framework (SAB, 2002) and described in sections 4.2.1 and 4.2,3 of the CD. These
24 impacted EEAs include Landscape Condition, Biotic Condition, Chemical and Physical
25 Characteristics, Ecological Processes, and Natural Disturbance Regimes. Given that humans, as
26 well as other organisms, are dependent on the services ecosystems provide, ecosystem changes
27 of mis magnitude are of concern and can lead to adverse impacts on human health and welfare.
28 Based on the information included in the above discussions and Chapters 4 and 9 of the
29 CD, staff has reached the following conclusions:
January 2005
6-59
Draft - Do Not Quote or Cite
-------
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
• An ecologically-relevant indicator for PM would be based on one or multiple chemical
stressors found in ambient PM (e.g. N or S containing compounds).
• Ecosystem effects can be associated with long-term high or even low levels of excess
inputs. Thus, there is no bright line or threshold for effects, but rattier a "syndrome" of
complex changes over time. Additionally, ecosystem recovery can occur but may take
decades, and may require controls beyond those already established.
• Excess N or acid deposition acts in conjunction with other co-occurring stresses (e.g.,
invasive species, reduced grazing pressure) that jointly determine ecological outcomes.
Therefore, these pollution-related stresses should not be considered in isolation.
Additionally, all forms of airborne nitrogen and acidic compounds need to be considered
and managed in harmony.
• Monitoring networks may be sufficient to measure air concentrations or deposition but
are not generally sufficient to monitor ecosystem response. For example, in the West;
more environmental monitoring is needed downwind of large urban areas.
Unfortunately, our ability to relate ambient concentrations of PM to ecosystem response
is hampered by a number of significant data gaps and uncertainties. First, U.S. monitoring
networks have only recently begun to measure speciated PM. Historically, measurements were
focused only on a particular size fraction such as PM10 and, more recently, PM15. An exception
to this is the IMPROVE network, which collects speciated measurements. Additionally, except
for the IMPROVE and some CASTNet sites, much of the PM monitoring effort has focused on
urban or near urban exposures, rather than on those in sensitive ecosystems. Thus, the lack of a
long-term, historic database of annual speciated PM deposition rates precludes establishing
relationships between PM deposition (exposure) and ecosystem response at this time. As a
result, while evidence of PM-related effects clearly exists, there is insufficient information
available at this time to serve as a basis for a secondary national air quality standard for PM,
specifically selected to protect against adverse effects on vegetation and ecosystems. .
A second source of uncertainty lies in predicting deposition velocities based on ambient
concentrations of PM. There are a multitude of factors that influence the amounts of PM that get
deposited from the air onto sensitive receptors, including the mode of deposition (wet, dry, and
occult), wind speed, surface roughness or stickiness, elevation, particle characteristics (e.g., size,
January 2005
6-60
Draft - Efo Not Quote or Cite
-------
1 shape, chemical composition), and relative humidity. Therefore, modeled deposition rates, used
2 in the absence of monitored data, can be highly uncertain.
3 Third, each ecosystem has developed within a context framed by the topography,
4 underlying bedrock,'soils, climate, meteorology, hydrologic regime, natural and land use history,
5 species associations that co-occur at that location (e.g., soil organisms, plants), and successional
6 stage, making it unique from all others. Because of this variety, and insufficient baseline data on
7 each of these features for most ecosystems, it is currently not possible to extrapolate with much
8 confidence any effect from one ecosystem to another, or to predict an appropriate "critical load"
9 for the vast majority of U.S. ecosystems.
10 As additional PM speciated air quality and deposition monitoring data become available,
11 there is much room for fruitful research into the areas of uncertainty identified above. At this
12 time, however, staff concludes that there is insufficient information available to recommend for
13 consideration an ecologically defined secondary standard that is specifically targeted for
14 protection of vegetation and ecosystems against the adverse effects potentially associated with
t
15 the levels of PM-related stressors of nitrate and sulfate found in the ambient air.
16
17 6.4 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
January 2005
6-61
Draft - Do Not Quote or Cite
I
-------
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 (section 4.4) of the CD, the
7 following sections summarize the physical damage and aesthetic soiling effects of PM on
8 materials including metals, paint finishes, and stone and concrete.
9
10 6.4.1 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 S02 and participate 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 4-193).
23 The CD (p. 4-194, 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 CD concludes that
25 the role of particles in the corrosion of metals is not clear (CD, p. 4-193). While several studies
26 suggest that particles can promote the corrosion of metals, others have not demonstrated a
27 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
t
January 2005 6-62 Draft - Do Not Quote or Cite
-------
1 relate corrosive effects to specific participate sulfate levels or to establish a quantitative
2 relationship between ambient paniculate 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 VHI-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 S02
17 with extender pigments such as calcium carbonate and zinc oxide, although Miller et al. (1992)
18 suggest mat calcium carbonate acts to protect paint substrates (CD, p. 4-196).
19 With respect to damage to building stone, numerous studies discussed in the CD (pp.
20 4-196 to 4-202; Table 4-19) suggest that air pollutants, including sulfur-containing pollutants
21 and wet or dry deposition of atmospheric particles and dry deposition of gypsum particles, can
22 enhance natural weathering processes. Exposure-related damage to building stone results from
23 the formation of salts in the stone that are subsequently washed away by rain, leaving the surface
24 more susceptible to the effects of air pollutants. Dry deposition of sulfur-containing pollutants
25 and carbonaceous particles promotes the formation of gypsum on the stone's surface. Gypsum is
26 a black crusty material that occupies a larger volume than the original stone, causing the stone's
27 surface to become cracked and pitted, leaving rough surfaces that serve as sites for further
28 deposition of airborne particles (CD, p. 4-200).
January 2005
6-63
Draft - Do Not Quote or Cite
I
-------
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.
8 4-201 ,4-202).
9
10 6.4.2 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. Early studies demonstrated an
13 association between particle exposure and increased frequency of cleaning painted surfaces. •
14 More recently, Haynie and Lemmons (1990) conducted a study to determine how various
15 environmental factors contribute to the rate of soiling on white painted surfaces. They reported'
16 that coarse-mode particles initially contribute more to soiling of horizontal and vertical surfaces
17 man do fine-mode particles, but are more easily removed by rain, leaving stains on the painted •
18 surface. The authors concluded that the accumulation of fine-mode particles, rather than coarse-
19 mode particles, more likely promotes the need for cleaning of the painted surfaces (EPA 1996b,
20 p. VIII-21-22; CD, pp. 4-202 to 4-204). Haynie and Lemmons (1990) and Creighton et al.
21 (1990) reported that horizontal surfaces soiled faster than vertical surfaces and that large
22 particles were primarily responsible for the soiling of horizontal surfaces not exposed to rainfall.
23 Additionally, a study was conducted to determine the potential soiling of artwork in five
24 Southern California museums (Ligocki, et al., 1993). Findings were that a significant fraction of
25 fine elemental carbon and soil dust particles in the ambient air penetrates to the indoor
26 environment and may constitute a soiling hazard to displayed artwork (EPA 1996b, p. VIII-22).
27 As for stone structures, the presence of gypsum is related to soiling of the stone surface
28 by providing sites for particles of dirt to concentrate. Lorusso et al. (1997) attributed the need
S
29 for frequent cleaning and restoration of historic monuments in Rome to exposure to total
t
January 2005-
6-64
Draft - Do Not Quote or Cite
-------
1 suspended particles (TSP). Further, Davidson etal. (2000) evaluated the effects of air pollution
2 exposure on a limestone structure on the University of Pittsburgh campus using estimated
3 average TSP levels in the 1930s and 1940s and actual values for the years 1957 to 1997.
4 Monitored levels of SO2 were also available for the years 1980 to 1998. Based on the available
5 data on pollutant levels and photographs, the authors concluded that soiling began while the
6 structure was under construction. With decreasing levels of pollution, the soiled areas have been
7 slowly washed away, the process taking several decades, leaving a white, eroded surface (CD,
8 pp. 4-203).
9
10 6.4.3 Summary and Conclusions
11 Damage to building materials results from natural weathering processes that are
12 enhanced by exposure to airborne pollution, most notably sulfur-containing pollutants. Ambient
13 PM has been associated with contributing to pollution-related damage to materials, and can
14 cause significant detrimental effects by soiling painted surfaces and other building materials.
15 Available data indicate that particle-related soiling can result in increased cleaning frequency
16 and repainting, and may reduce the useful life of the soiled materials. However, to date, no
17 quantitative relationships between particle characteristics (e.g., concentrations, particle size, and
18 chemical composition) and the frequency of cleaning or repainting have been established. Thus,
19 staff-concludes that PM effects on materials can play no quantitative role in considering whether
20 any revisions of the secondary PM NAAQS are appropriate at this time.
21
22 6.5 EFFECTS ON CLIMATE CHANGE AND SOLAR RADIATION
23 Atmospheric particles alter the amount of electromagnetic radiation transmitted through
24 the earth's atmosphere by both scattering and absorbing radiation. As discussed above in .
6
25 Chapter 2 (section 2.2.6), most components of ambient PM (especially sulfates) scatter and
26 reflect incoming solar radiation back into space, thus offsetting the "greenhouse effect" to some
27 degree by having a cooling effect on climate. In contrast, some components of ambient PM
28 (especially black carbon) absorb incoming solar radiation or outgoing terrestrial radiation, and
29 are believed to contribute to some degree to atmospheric wanning. Lesser impacts of
January 2005
6-65
Draft - Do Not Quote or Cite
I
-------
1 atmospheric particles are associated with their role in altering the amount of ultraviolet solar
2 radiation (especially UV-B) penetrating through the earth's atmosphere to ground level, where it
3 can exert a variety of effects on human health, plant and animal biota, and other environmental
4 components (CD, p. 205). The extensive research and assessment efforts into global climate
5 change and stratospheric ozone depletion provide evidence that atmospheric particles play
6 important roles in these two types of atmospheric processes, not only on a global scale, but also
7 on regional and local scales as well.
8 Information on the role of atmospheric particles in these atmospheric processes and the
9 effects on human health and the environment associated with these atmospheric processes is
10 briefly summarized below, based on the information in section 4.5 of the CD and referenced"
11 reports. These effects are discussed below in conjunction with consideration of the potential
12 indirect impacts on human health and the environment that may be a consequence of climatic
13 and radiative changes attributable to local and regional changes in ambient PM.
14
15 6.5.1 Climate Change and Potential Human Health and Environmental Impacts
16 As discussed in section 4.5.1 of the CD, particles can have both direct and indirect effects
17 on climatic processes.' The direct effects are the result of the same processes responsible for
18 visibility degradation, namely radiative scattering and absorption. However, while visibility
19 impairment is caused by particle scattering in all directions, climate effects result mainly from
20 scattering light away from the earth and into space. This reflection of solar radiation back to
21 space decreases the transmission of visible radiation to the surface and results in a decrease in
22 the heating rate of the surface and the lower atmosphere. At the same time, absorption of either
23 incoming solar radiation or outgoing terrestrial radiation by particles, primarily black carbon,
24 results in an increase in the heating rate of the lower atmosphere.
25 In addition to these direct radiative effects, particles can also have a number of indirect
26 effects on climate related to their physical properties. For example, sulfate particles can-serve as
27 condensation nuclei which alter the size distribution of cloud droplets by producing more
28 droplets with smaller sizes. Because the total surface area of the cloud droplets is increased, the
29 amount of solar radiation that clouds reflect back to space is increased. Also, smaller cloud
I
January 2005
6-66
Draft - Do Not Quote or Cite
-------
1 droplets have a lower probability of precipitating, causing them to have longer atmospheric
2 lifetimes. An important consequence of this effect on cloud properties is the suppression of rain
3 and potentially major disruption of hydrological cycles downwind of pollution sources, leading
4 to a potentially significant alteration of climate in the affected regions (CD, p. 4-218).
5 The overall radiative and physical effects of particles, both direct and indirect, are not the
6 simple sum of effects caused by individual classes of particles because of interactions between
7 particles and other atmospheric gases. As discussed in Section 4.5.1.2 of the CD, the effects of
8 sulfate particles have been the most widely considered, with globally averaged radiative effects
9 of sulfate particles generally estimated to have partially offset the warming effects caused by
10 increases in greenhouse gases. On the other hand, global-scale modeling of mineral dust
11 particles suggests that even the sign as well as the magnitude of effects depends on the vertical
12 distribution and effective particle radius.
13 The CD makes clear that atmospheric particles play an important role in climatic
14 processes, but that their role at this time remains poorly quantified. In general, on a global scale,
15 the direct effect of radiative scattering by atmospheric particles is to likely exert an overall net
16 effect of cooling the atmosphere, while particle absorption may lead to wanning. The net impact
17 of indirect effects on temperature and rainfall patterns remains difficult to generalize. However,
18 deviations from global mean values can be very large even on a regional scale, with any
19 estimation of more localized effects introducing even greater complexity (CD, p. 216). The CD
20 concludes that any effort to model the impacts of local alterations in particle concentrations on
21 projected global climate change or consequent local and regional weather patterns would be
22 subject to considerable uncertainty (CD, p. 4-240).
23 More specifically, the CD notes that while current climate models are successful in
24 simulating present annual mean climate and the seasonal cycle on continental scales, they are
25 lass successful at regional scales (CD, p. 4-207). Findings from various referenced assessments
26 illustrate well the considerable uncertainties and difficulties in projecting likely climate change
27 impacts on regional or local scales. For example, uncertainties in calculating the direct radiative
28 effects of atmospheric particles arise from a lack of knowledge of their vertical and horizontal
29 variability, their size distribution, chemical composition, and the distribution of components
January 2005 6-67 Draft - Do Not Quote or Cite
f
-------
1 within individual particles. Any complete assessment of the radiative effects of PM would
2 require computationally intensive calculations that incorporate the spatial and temporal behavior
3 of particles of varying composition that have been emitted from, or formed by precursors emitted
4 from, different sources. In addition, calculations of indirect physical effects of particles on
5 climate (e.g., related to alteration of cloud properties and disruption of hydrological cycles) are
6 subj ect to much larger uncertainties than those related to the direct radiative effects of particles
7 (CD, p. 4-219). The CD concludes that at present impacts on human health and the environment
8 due to aerosol effects on the climate system can not be calculated with confidence, and notes that
9 the uncertainties associated with such aerosol-related effects will likely remain much larger than
10 those associated with greenhouse gases (CD, p. 4-219). Nevertheless, the CD concludes that
11 substantial qualitative information available from observational and modeling studies indicates
12 that different types of atmospheric aerosols (i.e., different components of PM) have both
13 warming and cooling effects on climate, both globally and regionally. Studies also suggest that
14 global and regional climate changes could potentially have both positive and negative effects on
15 human health, human welfare, and the environment.
16 '
17 6.5.2 Alterations in Solar UV-B Radiation and Potential Human Health and
18 - Environmental Impacts
19 As discussed in section 4.5.2 of the CD, the effects of particles in the lower atmosphere.
20 " on the transmission of solar UV-B radiation have been examined both by field measurements
21 and by radiative transfer model calculations. Several studies cited in the CD reinforce the idea
22 that particles can play an important role in modulating the attenuation of solar UV-B radiation,
23 although none included measurements of ambient PM concentrations, so that direct relationships
24 between PM levels and UV-B radiation transmission could not be determined. The available
25 studies, "conducted in diverse locations around the world, demonstrate that relationships between
26 particles and solar UV-B radiation transmission can vary considerably over location, conditions,
27 and time. While ambient particles are generally expected to decrease.the flux of solar UV-B
28 radiation reaching the surface, any comprehensive assessment of the radiative effects of particles
29 would be location-specific and complicated by the role of particles in photochemical activity in
30 . the lower atmosphere. Whether the photochemical production of ozone is enhanced, remains the
'January 2005 6-68 Draft - Do Not Quote or Cite
-------
\
1 same, or reduced by the presence of ambient particles will be location-specific and dependent on
2 particle composition. Also complicating any assessment of solar UV-B radiation penetration to
3 specific areas of the earth's surface are the influences of clouds, which in turn are affected by the
4 presence of ambient particles.
5 The main types of effects associated with exposure to UV-B radiation include direct
6 effects on human health and agricultural and ecological systems, indirect effects on human
7 health and ecosystems, and effects on materials (CD, p. 4-221). The study of these effects has
8 been driven by international concern over potentially serious increases in the amount of solar
9 UV-B radiation reaching the earth's surface due to the depletion of the stratospheric ozone layer
10 by the release of various man-made ozone-depleting substances. Extensive qualitative and
11 quantitative characterizations of these global effects attributable to proj ections of stratospheric
12 ozone depletion have been periodically assessed in studies carried out under WMO and UNEP
13 auspices, with the most recent projections being published in UNEP (1998,2000) and WMO
14 (1999).
15 Direct human health effects of UV-B radiation exposure include: skin damage (sunburn)
16 leading to more rapid aging and increased incidence of skin cancer; effects on the eyes, including
17 retinal damage and increased cataract formation possibly leading to blindness; and suppression
18 of some immune system components, contributing to skin cancer induction and possibly
19 increasing susceptibility to certain infectious diseases. Direct environmental effects include
20 damage to terrestrial plants, leading to possible reduced yields of some major food crops and
21 commercially important tress, as well as to biodiversity shifts in natural terrestrial ecosystems;
22 and adverse effects on aquatic life, including reductions in important components of marine food
23 chains as well as other aquatic ecosystem shifts. Indirect health and environmental effects are
24 primarily those mediated through increased tropospheric ozone formation and consequent
25 ground-level ozone-related health and environmental impacts. Effects on materials include
26 accelerated polymer weathering and other effects on man-made materials and cultural artifacts.
\
27 In addition, there are emerging complex issues regarding interactions and feedbacks between
28 climate change and changes in terrestrial and marine biogeochemical cycles due to increased
29 UV-B radiation penetration. (CD, p. 4-221, 4-222).
January 2005
6-69
Draft - Do Not Quote or Cite
-------
1 In contrast to these types of negative impacts associated with increased UV-B penetration
2 to the Earth's surface, the CD (p. 4-222,4-223) summarizes research results that are suggestive
3 of possible beneficial effects of increased UV-B radiation penetration. For example, a number of
4 studies have focused on the protective effects of UV-B radiation with regard to non-skin cancer
5 incidence, which proved suggestive evidence that UV-B radiation, acting through the production
6 of vitamin D, may be a risk-reduction factor for mortality due to several types of cancer,
7 including cancer of the breast, colon, ovary, and prostate, as well as non-Hodgkin lymphoma,
8 Hie various assessments of these types of effects that have been conducted consistently
9 note that the modeled projections quantitatively relating changes in UV-B radiation (attributable
10 to stratospheric ozone depletion) to changes in health and environmental effects are subject to
11 considerable uncertainty, with the role of atmospheric particles being one of numerous
12 complicating factors. Taking into account the complex interactions between ambient particles
13 and UV-B radiation transmission through the lower atmosphere, the CD concludes that any
14 effort to quantify projected indirect effects of variations in atmospheric PM on human health or
15 the environment due to particle impacts on transmission of solar UV-B radiation would require
16 location-specific evaluations that take into account the composition, concentration, and internal
17 structure of the particles; temporal variations in atmospheric mixing heights and depths of layers
18 containing the particles; and the abundance of ozone and other absorbers within the planetary
19 boundary layer and the free troposphere (CD, 4-226).
20 At present, models are not available to take such complex factors into account, nor is
21 sufficient data available to characterize input variables that would be necessary for any such
22 modeling. The CD concludes, however, that the outcome of such modeling efforts would likely
23 vary from location to location, even as to the direction of changes in the levels of exposures to
24 UV-B radiation, due to location-specific changes in ambient PM concentrations and/or
25 composition (CD, p. 4-227).' Beyond considering just average levels of exposures to UV-B
26 radiation in general, the CD notes that ambient PM can affect the directional characteristics of
27 UV-B radiation scattering at ground-level, and thus its biological effectiveness. Also, ambient
28 PM can affect not only biologically damaging UV-B radiation, but can also reduce the ground-
29 level ratio of photorepairing UV-A radiation to damaging UV-B radiation. Further, the CD notes
January 2005
6-70
Draft - Do Not Quote or Cite
-------
1 that ambient PM deposition is a major source of PAH in certain water bodies, which can enhance
2 the adverse effects of solar UV-B radiation on aquatic organisms, such that the net effect of
3 ambient PM in some locations may be to increase UV-B radiation-related biological damage to
4 certain aquatic and terrestrial organisms. (CD, p, 4-227).
5 /-
6 6.5.3 Summary and Conclusions
7 A number of assessments of the factors affecting global warming and climate change as
8 well as those affecting the penetration of solar UV-B radiation to the earth's surface clearly
9 recognize ambient PM as playing various roles in these processes. These assessments, however,
10 have focused on global- and regional-scale impacts, allowing for generalized assumptions to take
11 the place of specific, but unavailable, information on local-scale atmospheric parameters and
12 characteristics of the distribution of particles present in the ambient air. As such, the available
13 information provides no basis for estimating how localized changes in the temporal, spatial, and
14 composition patterns of ambient PM, likely to occur as a result of expected future emissions of
15 particles and their precursor gases across the U.S., would affect local, regional, or global changes
16 in climate or UV-B radiation penetration - even the direction of such effects on a local scale
17 remains uncertain. Moreover, similar concentrations of different particle components can
18 produce opposite net effects. It follows, therefore, that there is insufficient information available
19 to project the extent to which, or even whether, such location-specific changes in ambient PM
20 would indirectly affect human health or the environment secondary to potential changes in
21 climate and UV-B radiation.
22 - Based on currently available information, staff concludes that the potential indirect
23 effects of ambient PM on public health and welfare, secondary to potential PM-related changes
24 in climate and UV-B radiation, can play no quantitative role in considering whether any
25 revisions of the primary or secondary PMNAAQS are appropriate at this time. Even
26 qualitatively, the available information is very limited in the extent to which it can help inform
27 an assessment of the overall weight of evidence in an assessment of the net health and
28 environmental effects of PM in the ambient air, considering both its direct effects (e.g.,
January 2005
6-1 \
Draft - Do Not Quote or Cite
-------
1 inhalation-related health effects) and indirect effects mediated by other routes of exposure and
2 environmental factors (e.g., dermal exposure to UV-B radiation).
January 2005 6-72 Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
REFERENCES
Section 6.2 - Visibility Impairment
Abt Associates, Inc. (2001) Assessing Public Opinions on Visibility Impairment Due to Air Pollution: Summary
Report. Prepared for EPA Office of Air Quality Planning and Standards; funded under EPA Contract No.
68-D-98-001. Bethesda, Maryland. January 2001.
Air Resource Specialists, Inc. (2003) WinHaze Air Quality Modeler, version 2.9.0. Available from
http://www.air-resource.com/whatsnew.htm
Arizona Department of Environmental Quality. (2003) Visibility Index Oversight Committee Final Report:
Recommendation for a Phoenix Area Visibility Index. March 5,2003.
htl.t>://www.pfaoenixvis.nel/R)F/vis 031403fi.ual.ndf.
BBC Research & Consulting. (2002) Phoenix Area Visibility Survey. Draft Report. October 4, 2002.
htta://www.bbcresearch,com/lit)rarvAfisibili)v draft report.pdf
California Code of Regulations. Title 17, Section 70200, Table of Standards.
Chestnut, L. G.; Rowe, R. D. (1990) Preservation values for visibility in the national parks. Washington, DC: U.S.
Environmental Protection Agency.
Chestnut, L. G.; Rowe, R. D. (1991) Economic valuation of changes invisibility: A state of the science assessment.
Sector B5 Report 27. In Acidic Depositions: State of Science and Technology Volume IV Control
Technologies, Future Emissions and Effects Valuation. P.M. Irving (ed.). The U.S. National Acid
Precipitation Assessment Program. GPO, Washington, D.C.
Chestnut, L.G.; Dennis, R. L.; Latimer, D. A. (1994) Economic benefits of improvements invisibility: acid rain
. provisions of the 1990 clean air act amendments. Proceedings of Aerosols and Atmospheric Optics:
Radiative Balance and Visual Air Quality. Air & Waste Management Association International Specialty
Conference, pp. 791-802.
Chestnut, L. G.; Dennis, R. L. (1997) Economic benefits of improvements invisibility: acid rain. Provisions of the
1990 clean air act amendments. J. Air Waste Manage. Assoc, 47:395-402.
Cohen, S.; Evans, G.W.; Stokols, D.; Krantz, D.S. (1986) Behavior, Health, and Environmental Stress. Plenum
Press. New York, NY,
Department of Interior. (1998) Air Quality in the National Parks. Natural Resources Report 98-1. National Park
Service, Air Quality Division. Denver, Colorado.
Ely, D.W.; Leary, J.T.; Stewart, T.R.; Ross, D.M. (1991) The Establishment of the Denver Visibility Standard. For
presentation at the 84th Annual Meeting & Exhibition of the Air and Waste Management Association, June
16-21,1991.
Environmental Protection Agency. (1979) Protecting Visibility: An EPA Report to Congress. Research Triangle
Park, NC: Office of Air Quality Planning and Standards. Report no. EPA-45-/5-79-008.
January 2005
6-73
Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
52
53
Environmental Protection Agency. (1982) Review of the National Ambient Air Quality Standards for Particulate
Matter, Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park,
N.C.: Office of Air Quality Planning and Standards, Strategies and Air Standards Division. Report no.
EPA-450/5-82-001.
Environmental Protection Agency. (1993) Air Quality Criteria for Oxides of Nitrogen. Research Triangle Park, NC:
Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office. Report
no. EPA-600/8-91/049F.
Environmental Protection Agency. (1996a) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC:
National Center for Environmental Assessment-RTF Office; report no. EPA/600/P- 95/00 laF-cF. 3v.
Environmental Protection Agency. (1 996b) Review of the National Ambient Air Quality Standards for Particulate
Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research
Triangle Park, NC 27711: Office of Air Quality Planning and Standards; report no. EPA-452XR-96-013.
Environmental Protection Agency. (1999) Regional Haze Regulations. 40 CFR Part 51. 300-309. 64 Federal
3571 3.
Environmental Protection Agency. (2000) Guidelines for Preparing Economic Analyses. Washington, DC: Office of
the Administrator. EPA 240-R-00-003.
Environmental Protection Agency. (2001) National Air Quality and Emissions Trends Report, 1 999. Research
Triangle Park,NC: Office of Air 'Quality Planning and Standards. Report no. EPA/454/R-01-004.' March.
Grand Canyon Visibility Transport Commission (1996) Report of the Grand Canyon Visibility Transport
Commission to the United States Environmental Protection Agency.
Hass, G. E.; Wakefield, T. J. (1 998) National Parks and the American Public: A National Public Opinion Survey of
the National Park System. Colorado State University, Department of Natural Resource Recreation and
Tourism, College of Natural Resources, Fort Collins, CO. Report prepared for the National Parks and
Conservation Association. June 1998.
McNe ill, R. and Roberge, A. (2000) The Impact of Visual Air Quality on Tourism Revenues in Greater Vancouver
and the Lower Fraser Valley. Environment Canada, Georgia Basin Ecosystem Initiative. GBEI report no. .
' EC/GB-00-028.
Middleton, P. (1993) Brown Cloud II: The Denver Air Quality Modeling Study, Final Summary Report. Metro
Denver Brown Cloud Study, Inc. Denver, CO.
Molenar, J.V.; Malm, W.C.; Johnson, C.E. (1994) Visual Air Quality Simulation Techniques. Atmospheric
Environment. Volume 28, Issue 5, 1055-1063.
Molenar, J.V. (2000) Visibility Science and Trends in the Lake Tahoe Basin: 1989-1 998. Report by Air Resource
Specialists, Inc., to Tahoe Regional Planning Agency. February 15, 2000. '
National Acid Precipitation Assessment Program (NAPAP) (1991) Acid Deposition: State of Science and
Technology. Report 24. Visibility: Existing and Historical Conditions - Causes and Effects. Washington,
DC.
National Acid Precipitation Assessment Program (NAPAP). (1998) Biennial Report to' Congress: an
Integrated Assessment.
http://dwb.unl.edu/Teacher/NSF/C14/C14Links/www.nnic.noaa.gov/CENR/NAPAP/NAPAP_96.htm
January 2005
6-74
Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
National Research Council. (1993) Protecting Visibility in National Parks and Wilderness Areas. National
Academy of Sciences Committee on Haze in National Parks and Wilderness Areas. National Academy
Press: Washington, DC.
National Transportation Safety Board (NTSB). (2000) NTSB Report NYC99MAI78, July 6,2000. Report on July
16, 1999 fatal accidental Vineyard Haven, MA.
National Weather Service. (1998) Automated Surface Observing System (ASOS) User's Guide. ASOS Program
Office. Silver Spring, MD.
New Zealand Ministry for the Environment. (2000) Proposals for Revised and New
Ambient Air Quality Guidelines: Discussion Document. Air Quality Report No. 16. December.
New Zealand National Institute of Water & Atmospheric Research (NIWAR). (2000a) Visibility in New Zealand:
Amenity Value, Monitoring, Management and Potential Indicators. Air Quality Technical Report 17.
Prepared for New Zealand Ministry for the Environment. Draft report.
New Zealand National Institute of Water & Atmospheric Research (NIWAR). (2000b) Visibility in New Zealand:
National Risk Assessment. Air Quality Technical Report 18. Prepared for New Zealand Ministry for the
Environment. Draft report.
Peacock, B.; Killingsworth, C.; Simon, B. (1998) State and National Economic Impacts Associated with Travel
Related Expenditures by Recreational Visitors to Lands Managed by the U.S. Department of Interior. U.S.
Department of the Interior. January.
Pryor, S.C. (1996) Assessing Public Perception of Visibility for Standard Setting Exercises. Atmospheric
Environment, vol. 30, no. 15, pp. 2705-2716.
Schichtel, B.A., Husar, R.B., Falke, S.R., and Wilson, W.E. (2001) "Haze Trends over the United States,
1980-1995," Atmospheric Environment, vol. 35, no. 30, pp. 5205-5210.
Schmidt, S.M., Mintz, D., Rao, T., and McCluney, L. (2005) Draft analysis of PM ambient air quality data for the
PM NAAQS review. Memorandum to PM NAAQS review docket OAR-2001 -0017. January 31,2005.
Schulze, W. D.; Brookshire, D. S.; Walther, E. G.; MacFarland, K. K.; Thayer, M. A.; Whitworth, R. L.; Ben-Davis,
S.; Malm, W.; Molenar, Jr. (1983) The Economic Benefits of Preserving Visibility in the National
Parklands of the Southwest. Nat. Resour. J. 23: 149-173.
Sisler, J., Malm, W.; Molenar, J.; Gebhardt, K. (1996) Spatial and Seasonal Patterns and Long Term Variability of
the Chemical Composition of Haze in the U.S.: An Analysis of Data from the IMPROVE Network. Fort
Collins, CO: Cooperative Institute for Research in the Atmosphere, Colorado State University.
State Government of Victoria, Australia. (2000a) Draft Variation to State Environment Protection Policy (Air
Quality Management) and State Environment Protection Policy (Ambient Air Quality) and Draft Policy
Impact Assessment. Environment Protection Authority. Publication 728. Southbank, Victoria.
State Government of Victoria, Australia. (2000b) Year in Review. Environment Protection Authority. Southbank,
Victoria.
January 2005
6-75
'Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
52
53
Section 6.3 - Vegetation and Ecosystems
Aber, J. D.; Nadelhoffer, K. J.; Steudler, P.; Melillo.J. M. (1989) Nitrogen saturation in northern forest ecosystems:
excess nitrogen from fossil fuel combustion may stress the biosphere. Bioscience 39: 378-386.
Aber, J. D.; Magill, A; McNulty, S. G.; Boone, R. D.; Nadelhoffer, K. J.; Downs, M.; Hallett, R. (1995) Forest
biogeochemistry and primary production altered by nitrogen saturation. Water Air Soil Pollut. 85:
1665-1670.
Aber, J.; McDowell, W.; Nadelhoffer, K.; Magill, A.; Bemtson, G.; Kamakea, M.; McNulty, S.; Currie, W.;
Rustad, L.; Fernandez, I. (1998) Nitrogen saturation in temperate forest ecosystems. BioScience 48:
921-934.
Allen, E. B.; Padgett, P. E.; Bytnerowicz, A.; Minich, R. (1998) Nitrogen deposition effects on coastal sage
vegetation of southern California. USDA Forest Service Gen. Tech. Rep. PSW-GTR-166, pp. 131-139.
Andersen, C. P.; Rygiewicz, P. T. (1991) Stress interactions and mycorrhizal plant response: understanding carbon
allocation priorities. Environ. Pollut. 73: 217-244.
Antoine, M.E. (2001) Ecophysiology of the cyanolichen.Lo&wia oregana. Master's thesis. Oregon State University,
Corvallis.
Bailey, S.W., Horsley, S.B., Long, R.P., Hallet, R.A. (1999) Influence of geologic and pedologic factors on health of
sugar maple on the Allegheny Plateau, U.S. In Horsley, S.B. and Long, R.P., eds. Sugar Maple Ecology
and Health: Proceedings of an International Symposium. Radnor, PA: U.S. Department of Agriculture,
Forest Service. General Technical Report NE-261. PP. 63-65.
Brooks, ML. (2003) Effects of increased soil nitrogen on the dominance of alien annual plants in the Mojave Desert.
Journal of Applied Ecology. 40:344-353.
Bytnerowicz, A.; Fenn, M. E. (1996) Nitrogen deposition in California forests: a review. Environ. Pollut.
92: 127-146.
Charles, D.F., ed. (1991) Acidic Deposition and Aquatic Ecosystems. Regional Case Studies. New York: Springer-
Verlag.
Craig, B.W. and Friedland, A.J. (1991) Spatial patterns in forest composition and standing dead red spruce in
montane forests of the Adirondacks and northern Appalachians. Environmental Monitoring and
Assessment. 18:129-140.
Cronan, C. S.; Grigal, D. F. (1995) Use of calcium/aluminum ratios as indicators of stress in forest ecosystems.
J. Environ. Qual. 24: 209-226.
D'Antonio, C.M. and Vitousek, P.M. (1992) Biological invasions by exotic grasses: The grass-fire cycle and global
change. Annual Review of Ecology and Systematics. 23: 63-87.
DeHayes, D. H.; Schaberg, P. G.; Hawley, G. J.; Strimbeck, G. R. (1999) Acid rain impacts on calcium nutrition and
forest health. Bioscience 49: 789-800.
Driscoll, C. T.; Wyskowski, B. J.; DeStaffan, P.; Newton, R. M. (1989) Chemistry and transfer of aluminum in a
forested watershed in the Adirondack region of New York, USA. In: Lewis, T. E., ed. Environmental /
chemistry and toxicology of aluminum. Chelsea, MI: Lewis Publishers, Inc.; pp.'83-105.
January 2005
6-76
Draft - Do Not Quote or Cite
-------
1 Driscoll, C.T., Likens, G.E., Church, MR. (1998) Recovery of surface waters in the northeastern U.S. from
2 decreases in atmospheric deposition of sulfur. Water, Airand Soil Pollution. 105:319-329.
3
4 Driscoll, C. T.; Lawrence, G. B.; Bulger, A. J.; Butler, T. J.; Cronan, C. S.; Eagar, C.; Lambert, K. F.; Likens, G. E.,
5 Stoddard, J. L.; Weathers, K. C. (2001) Acidic deposition in the northeastern United States: sources and
. 6 inputs, ecosystem effects, and management strategies. BioScience 51:180-198.
7
8 Edgerton-Warburton, L. M.; Allen, E. B. (2000) Shifts in arbuscular mycorrhizal communities along an
9 anthropogenic gradient nitrogen deposition gradient. Ecol. Appl. 10: 484-496.
10
11 Environmental Protection Agency. (1982) Air quality criteria for particulate matter and sulfur oxides. Research
12 Triangle Park, NC: Office of Health and Environmental Assessment, Environmental Criteria and
13 Assessment Office; EPA report no. EPA-600/8-82-029aF-cF. 3v. Available from: NTIS, Springfield, VA;
14 PB84-156777.
15
16 Environmental Protection Agency. (1992) Framework for Ecological Risk Assessment Washington, D.C.: Risk
17 Assessment Forum, U.S. Environmental Protection Agency. EPA/630/R-92/001.
18
19 Environmental Protection Agency. (1993) Air quality criteria for oxides of nitrogen. Research Triangle Park, NC:
20 Office of Health and Environmental Assessment, Environmental Criteria and Assessment Office; report
21 nos. EPA/600/8-91/049aF-cF. 3v. Available from: NTIS, Springfield, VA; PB95-124533, PB95-124525,
22 and PB95-124517.
23
24 Environmental Protection Agency. (1996) Air quality criteria for particulate matter. Research Triangle Park, NC:
25 National Center for Environmental Assessment-RTP Office; report nos. EPA/600/P-95/001 aF-cF. 3v.
26
27 Environmental Protection Agency. (1997) Nitrogen oxides: impacts on public health and the environment.
28 Washington, DC: Office of Air and Radiation; August. Available:
29 www.epa.gov/ttncaaal/tl/reports/noxrept.pdf [1999, November 24].
30
31 Environmental Protection Agency. (1998) Guidelines for Ecological Risk Assessment Washington, D.C: Risk
32 Assessment Forum, U.S. Environmental Protection Agency. EPA/630/R-95/002F.
33
34 Environmental Protection Agency. (2000) Deposition of air pollutants to the great waters. Third report to Congress.
35 [Executive Summary]. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office of Air
36 Quality Planning and Standards; report no. EPA-453/R-00-005.
37
38 Environmental Protection Agency. (2001) Air Quality Criteria for Particulate Matter. Research Triangle Park, NC:
39 Office of Research and Development; report no. EPA/600/P-99/002.'March.
40
41 Environmental Protection Agency. (2002) A Framework for Assessing and Reporting on Ecological Condition: An
42 SAB Report. Washington, D.C.: Ecological Processes and Effects Committee, Science Advisory Board,
43 U.S. Environmental Protection Agency. EPA-SAB-EPEC-02-009.
44
45 Environmental Protection Agency. (2003) Response Of Surface Water Chemistry to the Clean Air Act Amendments
46 of 1990. National Health and Environmental Effects Research Laboratory, Office of Research and
47 Development, U.S. Environmental Protection Agency. Research Triangle Park, NC. EPA 620/R-03/001.
48
49 Fenn, M. E.; Poth, M. A.; Aber, J. D.; Baron, J. S.; Bormann, B. T.; Johnson, D. W.; Lemly, A. D.; McNulty, S. G.;
50 Ryan, D. F.; Stottlemyer, R. (1998) Nitrogen excess in North American ecosystems: predisposing factors,
51 ecosystem responses, and management strategies. Ecol. Appl. 8: 706-733.
52 •)
January 2005 6-77 Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
Fenn, M. E.; Baron, J. S.; Allen, E. B.; Rueth, H. M.; Nydick, K. R.; Geiser, L.; Bowman, W. D.; Sickman, J. O.;
Meixner, T.; Johnson, D. W.; Neitlich, P. (2003) Ecological effects of nitrogen deposition in the western
United States. BioScience 53: 404-420.
Galloway, J.N, Norton, S.N., Church, MR. (1983) Freshwater acidification from atmospheric deposition of sulfuric
acid: A conceptual model. Environmental Science and Technology. 17:541 A-545A
Galloway, J. N. (1998) The global nitrogen cycle: changes and consequences. Environ. Pollut 102(suppl. 1): 15-24.
Galloway, J. N.; Cowling, E. B. (2002) Reactive nitrogen and the world: 200 years of change. Ambio 31: 64-71.
Galloway, J. R; Aber, J. D.; Erisman, J. W.; Seitzinger, S. P.; Howarth, R. W.; Cowling, E. B.; Cosby, B. J. (2003)
The nitrogen cascade. BioScience 53: 341-356.
Gamer, J. H. B. (1994) Nitrogen oxides, plant metabolism, and forest ecosystem response. In: Alscher, R. G.;
Wellburn, A. R., eds. Plant responses to the gaseous environment: molecular, metabolic and physiological
aspects, [3rd international symposium on air pollutants and plant metabolism]; June 1992; Blacksburg, VA.
London, United Kingdom: Chapman & Hall; pp. 301-314.
Gunn, J.M. and Mills, K.H. (1998) The potential for restoration of acid-damaged lake trout lakes. Restoration
Ecology. 6:390-397.
Gundersen, P.; Callesen, I.; De Vries, W. (1998) Nitrate leaching in forest ecosystems is related to forest floor C/N
ratios. Environ. Pollut 102(suppl. 1): 403-407.
Hedin, L. O.; Granat, L.; Likens, G. E.; Buishand, T. A.; Galloway, J. N.; Butler, T. J.; Rodhe, H. (1994)
Steep declines in atmospheric base cations in regions of Europe and North America. Nature (London)
367:351-354.
Hornung, M; Langan, S. J. (1999) Nitrogen deposition, sources, impacts and responses in natural and semi-natural
ecosystems. In: Bangan, S. J., ed. Impact of nitrogen deposition on natural ecosystems and semi-natural
ecosystems. Dordrect, Netherlands: Kluwer Academic Publishers; pp. 1-14. [Environmental Pollution,
no.3].
Horsley, S.B., Long, R.P., Bailey, S.W., Hallet, R.A., Hall, T.J. (1999) Factors contributing to sugar maple decline
along topographic gradients on the glaciated and unglaciated Allegheny Plateau. In Horsley, S.B. and
fLong, R.P., eds. Sugar Maple Ecology and Health: Proceedings of an International Symposium. Radnor,
'PA: U.S. Department of Agriculture, Forest Service. General Technical Report NE-261. PP. 60-62.
Howarth, R. W.; Boyer, E. W.; Pabich, W. J.; Galloway, J. N. (2002) Nitrogen use in the United States from
1961-2000 and potential future trends. Ambio 31: 88-96.
Jaworski, N. A.; Howarth, R. W.; Hetling, L. J. (1997) Atmospheric deposition of nitrogen oxides onto the landscape
contributes to coastal eutrophication in the northeast United States. Environ. Sci. Technol. 31:1995-2004.
Johnson, A.H., Friedland, A.J., Dushoff, J.G. (1984) Recent and historic red spruce mortality: Evidence of climatic
influence. Water, Air and Soil Pollutioa 30:319-330.
Johnson, D. W.; Van Miegroet, H.; Lindberg, S. E.; Todd, D. E.; Harrison, R. B. (1991) Nutrient cycling in red
spruce forests of the Great Smoky Mountains. Can. J. For. Res, 21:769-787.
January 2005 '
6-78
Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
52
53
Johnson, D. W.; Lindberg, S. E., eds. (1992) Atmospheric deposition and forest nutrient cycling: a synthesis of the
integrated forest study. New York, NY: Springer-Verlag, Inc. (Billings, W. D.; Golley, F.; Lange, O. L.;
Olson, J. S.; Remmert, H., eds. Ecological studies: analysis and synthesis: v. 91).
Johnson, D. W.; Swank, W. T.; Vose, J. M. (1993) Simulated effects of atmospheric sulfur deposition on nutrient
cycling in a mixed deciduous forest. Biogeochemistry 23: 169-196.
Johnson, D.W.; Susfalk, R.B.; Brewer, P.,R; Swank, W.T. (1999) Simulated effects of reduced sulfur, nitrogen, and
base cation deposition on soils and solutions in southern Appalachian forests. J. Environ. Qual 28:1336-1346.
-Keller, W. and Gunn, J.M. (1995) Lake water quality improvements and recovering aquatic communities. In Gunn,
J.M. ed. Restoration and Recovery of an Industrial Region: Progress in Restoring the Smelter-damaged
Landscape Near Sudbury, Canada. New York: Springer-Verlag, PP. 67-80.
Kretser, W., Gallagher, J., Nicolette, J. (1989) Adirondack Lake Study. 1984-1987. An Evaluation of Fish
Communities and Water Chemistry. Ray Brook, New York: Adirondacks Lakes Survey Corporation.
Likens, G.E. and Bormann, F.H. (1995) Biogeochemistry of a Forested Ecosystem. 2nd ed., New York: Springer-
Verlag.
L0kke, H.; Bak, J.; Falkengren-Grerup, U.; Finlay, R. D.; Ilvesniemi, H.; Nygaard, P. H.; Starr, M. (1996)
Critical loads of acidic deposition for forest soils: is the current approach adequate. Ambio 25: 510-516.
(
Lovett, G. M; Traynor, M. M; Pouyat, R. V.; Carreiro, M. M.; Zhu, W.'-X.; Baxter, J. W. (2000) Atmospheric
deposition to oak forest along an urban-rural gradient. Environ. Sci. Technol. 34: 4294-4300.
Magill, A. H.; Aber, J. D.; Bemtson, G. M.; McDowell, W. H.; Nadelhoffer, K. J.; Melillo, J. M.'; Steudler, P. (2000)
Long-term nitrogen additions and nitrogen saturation in two temperate forests. Ecosystems 3: 238-253.
McCune, B. and Geiser, L. (1997) Macrolichens of the Pacific Northwest. Corvallis: Oregon State University Press.
McDonnell, M. J.; Pickett, S. T. A.; Groffman, P.; Bohlen, P.; Pouyat, R. V.; Zipperer, W. C.; Parmelee, R. W.;
Carreiro, M. M.; Medley, K. (1997) Ecosystem processes along an urban-to-rural gradient. Urban Ecosyst.
1:21-36.
Minnich, R. A. and Dezzani, R.J. (1998) Historical decline of coastal sage scrub in the Riverside-Ferris Plain,
California. Western Birds. 29:366-391.
*
Naeem, S.; Thompson, L. J.; Lawler, S. P.; Lawton, J. H.; Woodfin, R. M. (1994) Declining biodiversity can alter
the performance of ecosystems. Nature 368: 734-737.
i
Nagy, K.A., Henen, B.T., Vyas, D.B. (1998) Nutritional quality of native and introduced food plants of wild desert
tortoises. Journal of Herpetology 32: 260-267.
Nash, T.H. and Sigal, L.L. (1999) Ephiphytic lichens in the San Bernardino mountains in relation to oxidant
gradients. In Miller, P.R. McBride, J.R., eds. Oxidant Air Pollution Impacts on the Montane Forests of
Southern California: A Case Study of the San Bernardino Mountains. Ecological Studies 134. New York:
Springer-Verlag. PP. 223-234.
National Science and Technology Council. (1998) National acid precipitation assessment program biennial report to
Congress: an integrated assessment; executive summary. Silver Spring, MD: U.S. Department of
Commerce, National Oceanic and Atmospheric Administration. Available:
www.nnic.noaa.gov/CENR/NAPAP/NAPAP_96.htm [24 November 1999].
January 2005
6-79
Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
Padgett, P. E.; Allen, E. B.; Bytnerowicz, A.; Minich, R. A. (1999) Changes in soil inorganic nitrogen as related to
atmospheric nitrogenous pollutants in southern California. Atmos. Environ. 33:769-781.
Paerl, H. W.; Bales, J. D.; Ausley, L. W.; Buzzelli, C. P.; Crowder, L. B.; Eby, L. A.; Go, M.; Peierls, B. L.;
Richardson, T. L.; Ramus, J. S. (2001) Ecosystem impacts of three sequential hurricanes (Dennis, Floyd,
and Irene) on the United States' largest lagoonal estuary, Pamlico Sound, NC. Proc. Nat. Acad. Sci. U. S. A.
98:5655-5611.
Pike, L.H. (1978) The importance of epiphytic lichens in mineral cycling. Bryologist 81: 247-257.
Pouyat, R. V.; McDonnell, M. J. (1991) Heavy metal accumulations in forest soils along an urban-rural gradient in
southeastern New York, USA. Water Air Soil Pollut. 57/58:797-808.
Rabalais,N.N. (2002) Nitrogen in aquatic ecosystems. Ambio 31:102-112.
Rago, PJ. and Wiener, J.G. (1986) Does pH affect fish species richness when lake area is considered? Transactions
of the American Fisheries Society. llb:438-447.
Riebsame, W.E., Robb, J.J., Gosnell, H., Theobald, D., Breding, P., Hanson, C., Rokoske, K. (1997) Atlas of the
New West: Portrait of a Changing Region. New York: W. W. Norton.
Rovira, A. D.; Davey, C. B. (1974) Biology of the rhizosphere. In: Carson, E. W., ed. The plant root and its
environment: proceedings of an institute; My, 1971; Blacksburg, VA. Charlottesville, VA: University
Press of Virginia; pp. 153-204.
Schindler, D.W., Mills, K.H., Malley, D.F., Findlay, S., Schearer, J.A., Davies, I.J., Turner, M.A., Lindsey, G.A.,
Cruikshank, D.R. (1985) Long-term ecosystem stress: Effects of years of experimental acidification.
Canadian Journal of Fisheries and Aqautic Science. 37:342-354. '
Science Advisory Board. (2002) A framework for assessing and reporting on ecological condition: an SAB report.
Young, T. F.; Sanzone, S., eds. Washington, DC: U.S. Environmental Protection Agency, Ecological
Processes and Effects Committee; report no. EPA-SAB-EPEC-02-009. Available at:
http://www.epa.gov/sciencel/rndex.html (27 January 2003).
Shortle, W. C.; Smith, K. T.; Minocha, R; Lawrence, G. B.; David, M B. (1997) Acidic deposition, cation
mobilization, and biochemical indicators of stress in healthy red spruce. J. Environ. Qual. 26: 871-876.
Skinner, M.W.; Pavlik, B.M. (1994) CNPS Inventory of Rare and Endangered Vascular Plants of California.
Sacramento: California Native Plant Society.
Smith, W. H. (1990) Forests as sinks for air contaminants: soil compartment. In: Air pollution and forests:
interactions between air contaminants and forest ecosystems. 2nd ed.' New York, -NY: Springer-Verlag;
pp. 113-146. (Springer series on environmental management).
Sollins, P., Grier, C.C., McCorison, F.M., Cromack, K., Jr.Fogel, R., Fredriksen, R.L. (1980) The internal element
cycles of an old-growth Douglas-fir ecosystem in western Oregon. Ecological Monographs 50:261-285.
Stanhill, G.; Cohen, S. (2001) Global dimming: a review of the evidence for a widespread and significant reduction
in global radiation with discussion of its probable causes and possible agricultural consequences. Agric.
For. Meteorol. 107: 255-278.
January 2005
6-80
Draft - Do Not Quote or Cite
-------
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
44
45
46
47
48
49
50
51
52
Strickland, T. C.; Holdren, G. R., Jr.; Ringold, P. L.; Bernard, D.; Smythe, K.; Fallen, W. (1993) A national critical
loads framework for atmospheric deposition effects assessment: I. method summary..Environ. Manage.
17: 329-334.
Van Egmond, K.; Bresser, T.; Bouwman, L. (2002) The European nitrogen case. Ambio 31: 72-78.
van Herk, C.M. (1999) Mapping of ammonia pollution with epiphytic lichens in the Netherlands. Lichenologist
31:9-20.
Van Sickle, J., Baker, J.P., Simonin, H.A., Baldigo, B.P., Kretser, W.A., Sharpe, W.F. (1996) Episodic acidification
of small streamsin the northeastern U.S.: Fish mortality in field bioassays. Ecological Applications. 6:408-
421.
Vitousek, P. M.; Mooney, H. A.; Lubchenco, J.; Melillo, J. M. (1997) Human domination of Earth's ecosystems.
Science (Washington, DC) 277:494-499.
Wall, D. H.; Moore, J. C. (1999) Interactions underground: soil biodiversity, mutualism, and ecosystem processes.
Bioscience 49: 109-117.
Wedin, D. A.; Tilman, D. (1996) Influence of nitrogen loading and species composition on the carbon balance of
grasslands. Science 274: 1720-1723.
Wesely, M. L.; Hicks, B. B. (2000) A review of the current status of knowledge on dry deposition. Atmos. Environ.
34: 2261-2282.
Whitby, K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12:135-159.
World Health Organization. (1997) Nitrogen oxides. 2nd ed. Geneva, Switzerland: World Health Organization.
(Environmental health criteria 188).
Section 6.4 - Man-made Materials
Creighton, P. J.; Lioy, P. J.; Haynie, F. H.; Lemmons, T. J.; Miller, J. L.; Gerhart, J. (1990) Soiling by atmospheric
aerosols in an urban industrial area, J. Air Waste Manage. Assoc. 40:1285-1289.
Davidson, C. I.; Tang, W.; Finger, S.; Etyemezian, V.; Striegel, M. F.; Sherwood, S. I. (2000) Soiling patterns on a
tall limestone building: changes over 60 years. Environ. Sci. Technol. 34: 560-565.
Environmental Protection Agency. (1996b) Review of the National Ambient Air Quality Standards for Particulate
Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Research
Triangle Park, NC 27711: Office of Air Quality Planning and Standards; report no. EPA-452\R-96-013.
Haynie, F.H.; Lemmons, T. J. (1990) Particulate matter soiling of exterior paints at a rural site. Aerosol Sci. Technol.
13:356-367.
Ligocki, M. P.: Salmon, L. G.; Fall, T.; Jones, M. C.; Nazaroff, W. W.; Cass, G. R. (1993) Characteristics of
airborne particles inside southern California museums. Atmos. Environ. Part A 27: 697-711.
Lorusso, S.; Marabelli, M.; Troili, M. (1997) Air pollution and the deterioration of historic monuments. J. Environ.
Pathol. Toxicol. Oncol. 16:171-173.
January 2005
6-81
Draft - Do Not Quote or Cite
-------
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Miller, W. C,; Fomes, R. E.; Gilbert, R. D.; Speer, A.; Spence, J. (1992) Removal of CaCOj extender in residential
coatings by atmospheric acidic deposition. In: Measurement of toxic and related air pollutants: proceedings
of the 1992 U. S. EPA/A&WMA international symposium. Pittsburgh, PA: Air & Waste Management
Association; pp. 129-134. (A&WMA publication VIP-25)
Wolff, G. T.; Collins, D. C.; Rodgers, W. R.; Verma, M. H.; Wong, C. A. (1990) Spotting of automotive finishes
from the interactions between dry deposition of crustal material and wet deposition of sulfate. J. Air Waste
Manage. Assoc. 40:1638-1648.
1
2
3
4
5
6
7
8
9
10
11 Section 6.5 - Climate Change and Solar Radiation
Intergovernmental Panel on Climate Change (IPCC). (1998) The regional impacts of climate change: an assessment
of vulnerability. Cambridge, United Kingdom: Cambridge University Press.
Intergovernmental Panel on Climate Change (IPCC). (2001 a) Climate change 2001: the scientific basis.
Contribution of working group I to the third assessment report of the Intergovernmental Panel on Climate
Change. Cambridge, United Kingdom: Cambridge University Press.
Intergovernmental Panel on Climate Change (IPCC). (2001 b) Climate change 2001: impacts, adaptation, and
vulnerability. Contribution of working group II to the third assessment report of the Intergovernmental
Panel on Climate Change. Cambridge, United Kingdom: Cambridge University Press.
r
National Academy of Sciences (NAS). (2001) Committee on the Science of Climate Change, National Research
Council. Climate Change Science: An Analysis of Some Key Questions, National Academy .Press,
Washington, DC.
United Nations Environment Programme (UNEP). (1998) Environmental effects of ozone depletion: 1998
assessment. J. Photochem. Photobiol. B 46:1-4.
United Nations Environment Programme (UNEP). (2000) Environmental effects of ozone depletion: interim
summary. Available at: bttp://www.gcrio.or^oyoiie/unep20(X)5ummanLhtml (9 April 2002).
World Meteorological Organization. (1999) Scientific assessment of ozone depletion: 1998. Geneva, Switzerland:
World Meteorological Organization, Global Ozone and Monitoring Project; report no. 44.
January 2005
6-82
Draft - Do Not Quote or Cite
-------
1 7. STAFF CONCLUSIONS AND RECOMMENDATIONS ON
2 SECONDARY PMNAAQS
3 7.1 INTRODUCTION
4 This chapter presents staff conclusions and recommendations for the Administrator to
5 consider in deciding whether the existing secondary PM standards should be revised and, if so,
6 what revised standards are appropriate.1 The existing suite of secondary PM standards, which is
7 identical to the suite of primary PM standards, includes annual and 24-hour PM25 standards and
8 annual and 24-hour PM,0 standards to address visibility impairment associated with fine particles
9 and materials damage and soiling related to both fine and coarse particles. Each of these
10 standards is defined in terms of four basic elements: indicator, averaging time, level and form.
11 Staff conclusions and recommendations on these standards are based on the assessment and
12 integrative synthesis of information related to welfare effects presented in the CD and on staff
13 analyses and evaluations presented in Chapters 2 and 6 herein.
14 In recommending a range of secondary standard options for the Administrator to
15 consider, staff notes that the final decision is largely apublic policy judgment. A final decision
16 must draw upon scientific evidence and analyses about effects on public welfare, as well as'
17 judgments about how to deal with the range of uncertainties that are inherent in the relevant
18 information. The NAAQS provisions of the Act require the Administrator to establish secondary
19 standards that are requisite to protect public welfare2 from any known or anticipated adverse
20 effects associated with the presence of the pollutant in the ambient air. In so doing, the
21 Administrator seeks to establish standards that are neither more nor less stringent than necessary
22 . for this purpose. The provisions do not require that secondary standards be set to eliminate all
As noted in Chapter 1, staff conclusions and recommendations presented herein are provisional; final staff
conclusions and recommendations, to be included in the final version of this document, will be informed by
comments received from CASAC and the public in their reviews of this draft document.
2 As noted in Chapter 1, welfare effects as defined in section 302(h) [42 U.S.C. 7602(h)] include, but are
not limited to, "effects on soils, water, crops, vegetation, man-made materials, animals, wildlife, weather, visibility
and climate, damage to and deterioration of property, and hazards to transportation, as well as effects on economic
values and on personal comfort and well-being."
January 2005
7-1
Draft - Do Not Quote or Cite
-------
1 welfare effects, but rather at a level requisite to protect public welfare from those effects that are
2 judged to be adverse.
3 7.2 APPROACH
4 Similar to the approach discussed in Chapter 5, section 5.2, for the review of the primary
5 NAAQS, staffs approach here can be framed by a series of questions that may be applicable for
6 , each category of PM-related welfare effects identified in the CD as being associated with the
7 presence of the pollutant in the ambient air. Staffs review of the adequacy of the current PM
8 standards for each effects category involves addressing questions such as:
9 • To what extent does the available information demonstrate or suggest that PM-related
10 effects are occurring at current ambient conditions or at levels that would meet the
11 . current standards?
12 • To what extent does the available information inform j udgments as to whether any
13 observed or anticipated effects are adverse to public welfare?
w
14 • To what extent are the current secondary standards likely to be effective in achieving
15 protection against any identified adverse effects?
16 To the extent that the available information suggests that revision of the current secondary
17 standards would be appropriate for an effects category, staff then identifies ranges of standards
18 (in terms of indicators, averaging times, levels, and forms) that would reflect a range of
19 alternative policy judgments as to the degree of protection that is requisite to protect public
20 welfare from known or anticipated adverse effects. In so doing, staff addresses questions such
21 as:
22
23
24
25
26
27
28
Does the available information provide support for considering different PM indicators?
Does the available information provide support for considering different.averaging
times?
What range of levels and forms of alternative standards is supported by the information,
and what are the uncertainties and limitations in that information?
To what extent would specific levels and forms of alternative standards reduce adverse
impacts attributable to PM, and what are the uncertainties in the estimated reductions?
January 2005
7-2
Draft - Do Not Quote or Cite
-------
1 Based on the available information, estimated reductions in adverse impacts, and related
2 uncertainties, staff makes recommendations as to ranges of alternative standards for the
3 Administrator's consideration in reaching decisions as to whether to retain or revise the
4 secondary PM NAAQS.
5 In presenting this approach, staff well recognizes that for some welfare effects the
6 currently available information falls short of what is considered sufficient to serve as a basis for
7 a distinct standard defined specifically in terms of the relationship between ambient PM and that
8 effect In the case of visibility impairment, however, the available information may well provide
9 a basis for a distinctly defined standard. In either case, staff believes it is appropriate to consider
10 the extent to which the current or recommended primary standards may afford protection against
11 the identified welfare effects.
12 Staff first considers information related to the effects of ambient PM, especially fine
13 particles, on visibility impairment in section 7.3, and makes recommendations that consideration
14 be given to a revised PM25 standard. Other PM-related welfare effects, including effects on
15 vegetation and ecosystems, materials, and global climate change processes, are addressed in
16 section 7.4. This chapter concludes with a summary of key uncertainties associated with
17 establishing secondary PM standards and related staff research recommendations in section 7.5.
18 7.3 STANDARDS TO ADDRESS VISIBILITY IMPAIRMENT
19 In 1997, EPA decided to address the effects of PM on visibility by setting secondary
20 standards identical to the suite of PM2 5 primary standards, in conjunction with the future
21 establishment of a regional haze program under sections 169A and 169B of the Act (62 FR at
22 38,679-83). In reaching this decision, EPA first concluded that PM, especially fine particles,
23 produces adverse effects on visibility in various locations across the country, including multi-
24 state regions, urban areas, and remote Class I Federal areas (e.g., national parks and wilderness
25 areas). EPA also concluded that addressing visibility impairment solely through setting more
26 stringent national secondary standards would not be an appropriate means to protect the public
27 welfare from adverse impacts of PM on visibility in all parts of the country. As a consequence,
January 2005
7-3
Draft - Do Not Quote or Cite
-------
1 EPA determined mat an approach that combined national secondary standards with a regional
2 haze program was the most appropriate and effective way to address visibility impairment.
3. In reaching these conclusions in 1997, EPA recognized, based on observations from
4 available monitoring data, primarily from rural sites in the IMPROVE monitoring network, that
5 the selection of an appropriate level for a national secondary standard to address visibility
6 protection was complicated by regional differences in visibility impairment These differences
7 were due to several factors, including background and current levels of PM, the composition of
8 PM, and average relative humidity. As a result of these regional differences, EPA noted that a
9 national standard intended to maintain or improve visibility conditions in many parts of the West
10 would have to be set at or below natural background levels in the East; conversely, a national
11 standard that would improve visibility in the East would permit further degradation in the West
12 Beyond such problems associated with regional variability, EPA also determined that there was
13 not sufficient information available to establish a standard level to protect against visibility
14 .conditions generally considered to be adverse in1 all areas.
15 These considerations led EPA to assess whether the protection afforded by the
16 combination of the selected primary PM2 5 standards and a regional haze program would provide
17 appropriate protection against the effects of PM on visibility. Based on such an assessment,
18 EPA determined that attainment of the primary PM2 5 standards through the implementation of
19 regional control strategies would be expected to result in visibility improvements in the East at
20 both urban and regional scales, but little or no change in the West, except in and near certain
21 urban areas. Further, EPA determined that a regional haze program that would make significant
22 progress toward the national visibility goal in Class I areas would also be expected to improve
23 visibility in many urban and non-Class I areas as well. EPA also noted, however, that the
24 combined effect of the PM NAAQS and regional haze programs may not address all situations in
25 which people living in certain urban areas may place a particularly high value on unique scenic
26 resources in or near these areas. EPA concluded that such situations were more appropriately
27 and effectively addressed by local visibility standards, such as those established by the city of
28 Denver, than by national standards and control programs.
29 As anticipated in the last review, EPA promulgated a regional haze program in 1999.
30 That program requires States to establish goals for improving visibility in Class I areas and to
January 2005 7-4 Draft - Do Not Quote or Cite
-------
1 adopt control strategies to achieve these goals. More specifically, States are required to establish
2 goals for improving visibility on the 20% most impaired days in each Class I area, and for
3 allowing no degradation on the 20% least impaired days. Since strategies to meet these goals are
4 to reflect a coordinated approach among States, multistate regional planning organizations have
5 been formed and are now developing strategies, to be adopted over the next few years, that will
6 make reasonable progress in meeting these goals.
7 7.3.1 Adequacy of Current PMZ 5 Standards
8 In considering the information now available in this review, as discussed in Chapters 2
9 and 6 (section 6.2), staff notes that, while new research has led to improved understanding of the
10 optical properties of particles and the effects of relative humidity on those properties, it has not
11 changed the fundamental characterization of the role of PM, especially fine particles, in visibility
12 impairment from the last review. However, extensive new information now available from
i
13 visibility and fine particle monitoring networks has allowed for updated characterizations of
14 visibility trends and current levels in urban areas, as well as Class I areas. These new data are a
15 critical component of the analysis presented in section 6.2.3 that better characterizes visibility
16 impairment in urban areas.
17 Based on this information, staff has first considered the extent to which available
18 information shows PM-related impairment of visibility at current ambient conditions in areas
19 across the U. S. Taking into account the most recent monitoring information and analyses, staff
20 makes the following observations:
21 • In Class I areas, visibility levels on the 20% haziest days in the West are about equal to
22 levels on the 20% best days in the East. Despite improvement through the 1990's,
23 visibility in the rural East remains significantly impaired, with an average visual range of
24 approximately 20 km on the 20% haziest days (compared to the naturally occurring
25 visual range of about 150 + 45 km). In the rural West, the average visual range showed
26 little change over this period, with an average visual range of approximately 100 km on
27 the 20% haziest days (compared to the naturally occurring visual range of about 230 ± 40
28 km).
29 • In urban areas, visibility levels show far less difference between eastern and western
30 regions. For example, based on reconstructed light extinction values calculated from 24-
31 hour average PM2 5 concentrations, the average visual ranges on the 20% haziest days in
32 eastern and western urban areas are approximately 21 km and 28 km, respectively. Even
January 2005
7-5
Draft - Do Not Quote or Cite
Q
-------
1 more similarity is seen in considering 4-hour (12:00 to 4:00 pm) average PM25
2 concentrations, for which the average visual ranges on the 20% haziest days in eastern
3 and western urban areas are approximately 26 km and 3 Okm, respectively. (Schmidt et
4 al.,2005)
5 Based on this information, and on the recognition that efforts are now underway to
6 address all human-caused visibility impairment in Class I areas through the regional haze
7 program implemented under sections 169A and 169B of the Act, as discussed above, staff has
8 focused in this review on visibility impairment primarily in urban areas, hi so doing, staff has
9 considered whether information now available can inform judgments as to the extent to which
10 existing levels of visibility impairment in urban areas can be considered adverse to public
11 welfare. In so doing, staff has looked at studies in the U.S. and abroad that have provided the
12 basis for the establishment of standards and programs to address specific visibility concerns in
13 v local areas, as discussed in section 6.2.5. These studies have produced new methods and tools to
14 communicate and evaluate public perceptions about varying visual effects associated with
15 alternative levels of visibility impairment relative to varying particle pollution levels and
16 environmental conditions. As discussed in section 6.2.6, methods involving the use of surveys to
17 elicit citizen judgments about the acceptability of varying levels of visual air quality in an urban
18 area have been developed by the State of Colorado, and used to develop a visibility standard for
19 Denver. These methods have now been adapted and applied in other areas, including Phoenix,
20 AZ, and the province of British Columbia, Canada, producing reasonably consistent results in
21 terms of the visual ranges found to be generally acceptable by the participants in the various
22 studies, which ranged from approximately 40 to 60 km in visual range.
23 Beyond the information available from such programs, staff believes it is appropriate to
24 make use directly of photographic representations of visibility impairment to help inform
25 judgments about the acceptability of varying levels of visual air quality in urban areas. As
26 discussed in section 6.2.6, photographic representations of varying levels of visual air quality
27 have been developed for several urban areas and are available on EPA's website
28 Outp://vvvsw.epa.gov/ttnJ'fnaaqs/statidards/pm/sjjm_crjsp.htinl') as an attachment to this •
29 document. In considering these images for Washington, D.C., Chicago, and Phoenix (for which
30 PMZ5 concentrations are reported), staff makes the following observations:
January 2005 7-6 Draft - Do Not Quote or Cite
-------
1 • At concentrations at or near the level of the current 24-hour PM25 standard, scenic views
2 (e-g-> mountains, historic monuments), as depicted in these images around and within the
3 urban areas, are significantly obscured from view.
4 • Appreciable improvement in the visual clarity of the scenic views depicted in these
5 images .occurs at PM2.5 concentrations below 35 to 40 ug/m3, or at visual ranges generally
6 above 20 km for the urban areas considered.
7 While being mindful of the limitations in using visual representations from a small
8 number of areas as a basis for considering national visibility-based secondary standards, staff
9 nonetheless concludes that the observations discussed above support consideration of revising
10 the current PM2 5 secondary standards to enhance visual air quality, particularly with a focus on
11 urban areas. Thus, in the sections that follow, staff evaluates information related to indicator,
12 averaging time, level and form to identify a range of alternative PM standards for consideration
13 that would protect visual air quality, primarily in urban areas.
14 7.3.2 Indicators
15 . As discussed in Chapter 2, section 2.8, fine particles contribute to visibility impairment
16 directly in proportion to their concentration in the ambient air. Hygroscopic components of fine
17 particles, in particular sulfates and nitrates, contribute disproportionately to visibility impairment
18 under high humidity conditions, when such components can reach particle diameters up to and
19 even above 2.5 um. Particles in the coarse mode generally contribute only marginally to
20 visibility impairment in urban areas. Thus, fine particles, as indexed by PM2i, are an appropriate
21 indicator of PM pollution to consider for the purpose of standards intended to address visibility
22 impairment.
23 In analyzing how well PM2 s concentrations correlate with visibility in urban locations
24 across the U.S., as discussed above in section 6.2.3 and in more detail in Schmidt et al. (2005),
25 staff concludes that the observed correlations are strong enough to support the use of PM2 5 as the
26 indicator for such standards. More specifically, clear correlations exist between 24-hour average
27 PM2 j concentrations and reconstructed light extinction (RE), which is directly related to visual
28 range, and these correlations are similar in eastern and western regions. These correlations are
29 less influenced by relative humidity and more consistent across regions when PM2 5
January 2005
7-7
Draft - Do Not Quote or Cite
-------
1 concentrations are averaged over shorter, daylight time periods (e.g., 4 to 8 hours). Thus, staff
2 concludes that it is appropriate to use PM2 5 as an indicator for standards to address visibility
3 impairment in urban areas, especially when the indicator is defined for a relatively short period
4 of daylight hours.
5 7.3.3 Averaging Times
6 In considering appropriate averaging times for a standard to address visibility
7 impairment, staff has considered averaging times that range from 24 to 4 hours, as discussed in
8 section 6.2.3. Within this range, as noted above, correlations between PM25 concentrations and
9 RE are generally less influenced by relative humidity and more consistent across regions as the
10 averaging time gets shorter. Based on the regional and national average statistics considered in
11 this analysis, staff observes that in the 4-hour time period between 12:00 and 4:00 p.m., the slope
12 of the correlation between PM25 concentrations and hourly RE is lowest and most consistent
13 across regions than for any other 4-hour or longer time period within a day (Chapter 6, Figure .
14 6-4). Staff also recognizes that these advantages remain in looking at a somewhat wider time
15 period, from approximately 10:00am to 6:00 pm. Staff concludes that an averaging time from 4
16 to 8 hours, generally within the time period from 10:00 am to 6:00 pm, should be considered for
17 a standard to address visibility impairment.
18 In reaching this conclusion, staff recognizes that the national PM2 5 FRM monitoring
19- network provides 24-hour average concentrations, such that implementing a standard with a less-
20 than-«24-hour averaging time would necessitate the use of continuous monitors that can provide
21 hourly time resolution. Given that the data used in the analysis discussed above are from
22 commercially available PM2} continuous monitors, such monitors clearly could provide the
23 hourly data that would be needed for comparison with a potential visibility standard with a less-
24 than-24-hour averaging time. Decisions as to which PM2.5 continuous monitors are providing
,\
25 data of sufficient quality to be used in a visibility standard would follow protocols for approval
26 of reference and equivalent methods that can provide data in at least hourly intervals.
27 Development of the criteria for approval of these reference or equivalent methods for support of
28 a visibility standard would be based upon a data quality objective process,that considers
January 2005
7-8
Draft - Do Not Quote or Cite
-------
1 uncertainties associated with the measurement system and the level of the standard under
2 consideration.
3 73.4 Alternative VM1S Standards to Address Visibility Impairment
4 In considering alternative short-term (4- to 8-hour) PM2 5 standards that would provide
5 requisite protection against PM-related impairment of visibility primarily in urban areas, staff
6 has taken into account the results of public perception and attitude surveys in the U.S. and
7 Canada, State and local visibility standards within the U.S., and visual inspection of
8 photographic representations of several urban areas across the U. S. Staff believes that these
9 sources provide a basis for bounding a range of levels appropriate for consideration in setting a
10 national visibility standard primarily for urban areas.
11 As discussed above in section 6.2, public perception and attitude surveys conducted in
12 Denver, CO and Phoenix, AZ resulted in judgments reflecting the acceptability of a visual range
13 of approximately 50 and 40 km, respectively. A similar survey approach in the Fraser Valley in
14 British Columbia, Canada reflected the acceptability of a visual range of 40 to 60 km. Visibility
15 standards established for the Lake Tahoe area in California and for areas within Vermont are
16 both targeted at a visual range of approximately 50 km. Staff notes that, in contrast to this
17 convergence of standards and goals around a visual range from 40 to 60 km, California's long-
18 standing general state-wide visibility standard is a visual range of approximately 16 km. Staff
19 believes that consideration should be given to national visibility standards for urban areas across
20 the U.S. that are somewhat less stringent than local standards and goals set to protect scenic
21 resources in and around certain urban areas that are particularly highly valued by people living in
22 those areas, suggesting an upper end of the range of consideration below 40 km.
23 Staff has also inspected the photographic representations of varying levels of visual air
24 quality that have been developed for Washington, D.C., Chicago, Phoenix, and Denver
25 (available on EPA's website, http:/viww.epa.gov/tta/naaqs/st^ as an
26 attachment to this document). Staff observes that scenic views (e.g., historic monuments, lake
27 front and mountain vistas) depicted in these images (around and within the three urban areas for
28 which PM2 5 concentrations are reported) are significantly obscured from view at PM2 5
29 concentrations of 35 to 40 ^g/m3 in Chicago, Washington, D.C., and Phoenix, corresponding to
January 2005 7-9 Draft - Do Not Quote or Cite
-------
1 reported visual ranges in Washington, D.C. and Phoenix of 12 to 20 km, respectively. Staff also
2 observes that visual air quality appears to be good in these areas at PM2 s concentrations
3 generally below 20 ug/m3, corresponding to reported visual ranges in Washington, D.C. and
4 Phoenix above approximately 25 to 35 km, respectively. In looking at the images in Denver,
5 staff observes that visual air quality appears to be generally good, specifically in terms of the .
6 ability to view nearby mountain ranges, at a visual range above 52 km. These observations are
7 interpreted by staff as suggesting consideration of a national visibility standard in the range of 30
8 to 20 jAg/m3, The upper end of this range is below the levels at which scenic views are
9 significantly obscured, and the lower end is around the level at which visual air quality generally
10 appeared to be good in these areas. Staff recognizes that the above observations about visual air
11 quality in urban areas inherently take into account the nature and location of scenic views that
12 are notable within and around a given urban area, which has implications for the appropriate
13 design of an associated monitoring network.
i
14 Building upon the analysis discussed above in section 6.2.3, staff has characterized the
15 distributions of PM2 5 concentrations, 4-hour averages in the 12:00 to 4:00 pm time frame, by
16 region, that correspond to various visual range target levels; The results are shown in Figure 7-1,
17 panels (a) through (c), for visual range levels of 25, 30, and 35 km, respectively. This figure
18 shows notable consistency across regions in the median concentrations that correspond to the
19 target visual range level, with what more variation in regional mean values as well as notable
20 variation within each region. In focusing on the median values, staff observes that 4-hour
21 average PM2.5 concentrations of approximately 30, 25, and 20 |ig/m3 correspond to the target
22 visual range levels of 25, 30, and 35 km, respectively. Thus, a standard set within the range of
23 30 to 20 ug/m3 can be expected to correspond generally to median visual range levels of
24 approximately 25 to 35 km in urban areas across the U.S.. Staff notes, however, that a standard
25 set at any specific PM2 5 concentration will necessarily result in visual ranges that vary somewhat
26 in urban areas across the country, reflecting in part the less-than-perfect correlation between
27 PM2 5 concentrations and reconstructed light extinction. Staff also notes that the range of PM2 5
28 concentrations from 30 to 20 ug/m3, suggested by staffs analysis and observations of
29 photographic representations, is generally consistent with national target visual range levels
January 2005 7-10 Draft - Do Not Quote or Cite
-------
::•:::::-:•:::::•:•::: Visual range ~ 25kltt I!!-:-:li!!•!-!! ii-lii ii-l !!ii.::1'ii--.iitiii-iii'ii-iii'iii'iii
:;;;t;;.:;it Visual ranee = 3Skm
Northeast Southeast Industrial Uppef Southwes, Northwest Southern
Midwest Midwest California
Figure 7-1. Distributions of PM25 concentrations for 12 p.m. - 4 p.m.
corresponding to visual ranges of 25km (panel a), 30km (panel b),
and 35km (panel c) — by region. Box depicts interquartile range and
median; whiskers depict 5th and 95th percentiles; star denotes mean.
Source: Schmidt et al. (2005)
January 2005 7-11 Draft - Do Not Quote or Cite
-------
I
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
below 40 km, the level suggested by the public perception surveys and the.local visibility
standards and goals discussed above.
In considering a standard down to 20 ng/m3, staff has again looked to information on i
PM2S background concentrations, as was done in considering primary PM25 standard levels in
Chapter 5, section 5.3.5. In both instances, staff recognizes that an appropriate standard level
intended to provide protection from man-made pollution should be clearly above background '
levels. In considering background levels in conjunction with a primary standard, staff focused
on the 99th percentile of the distribution of estimated background levels; consistent with
consideration of a 98th or 99th percentile form for a primary standard, concluding in that case that
25 ug/m3 was an appropriate lower end to the range of 24-hour primary PM25 standards for
consideration. For reasons discussed below, staff believes that a lower percentile form would be
appropriate to consider for a visibility standard, and thus has looked to a lower percentile in the
distribution of estimated background levels as a basis for comparison with the lower end of the
range of short-term secondary PM2 s standards for consideration. As discussed in Chapter 2,
section 2.6, staff notes that, while long-term average daily PM2 s background levels are quite low
(ranging from 1 to 5 ug/m3 across the U.S.), the estimated 90th percentile values in distributions
of daily background levels are appreciably higher, but generally well below 15 ng/m3, with
levels below 10 ug/m3 in most areas, and these levels may include some undetermined
contribution from anthropogenic emissions (Langstaff, 2005). In addition, staff again notes that
even higher daily background levels result from episodic occurrences of extreme natural events
(e.g., wildfires, global dust storms), but levels related to such events are generally excluded from
consideration under EPA's natural events policy, as noted in section 2.6. Taking these
considerations into account, staff believes that 20 ug/m3 is an appropriate lower end to the range
of short-term PM2 5 standards for visibility protection for consideration in this review.
As in the last review, staff believes that a national visibility standard should be
considered in conjunction with the regional haze program as a means of achieving appropriate
levels of protection against PM-related visibility impairment in urban, non-urban, and Class I
areas across the country. Staff recognizes that programs implemented to meet a national
standard focused primarily on urban areas can be expected to improve visual air. quality in
surrounding non-urban areas as well, as would programs now being developed to address the
>
January 2005 7-12 Draft - Do Not Quote or Cite
-------
1 requirements of the regional haze rule established for protection of visual air quality in Class I
2 areas. Staff further believes that the development of local programs continues to be an effective
3 and appropriate approach to provide additional protection for unique scenic resources in and
4 around certain urban areas that are particularly highly valued by people living in those areas.
5 Based on these considerations, and taking into account the observations and analysis discussed
6 above, staff concludes that consideration should be given to a short-term (4- to 8-hour daylight
7 average) secondary PM25 standard in the range of 30 to 20 ug/m3 for protection of visual air
8 quality primarily in urban areas.
9 7.3.5 Alternative Forms of a Short-term PM2 s Standard
10 In considering an appropriate form for a short-term PM2 5 standard for visibility, staff has
11 taken into account the same general factors that were taken into account in considering an
12 appropriate form for the primary PM25 standard, as discussed above in Chapter 5, section 5.3.6.
13 In that case, as in the last review, staff has concluded that a concentration-based form should be
14 considered because of its advantages over the previously used expected-exceedance form3. One
15 such advantage is that a concentration-based form is more reflective of the impacts posed by
16 elevated PM2 5 concentrations because it gives proportionally greater weight to days when
17 concentrations are well above the level of the standard than to days when the concentrations are
18 just above the standard. Staff notes that the same advantage would apply for a visibility standard
19 as to a health-based standard, in that it would give proportionally greater weight to days when
20 PM-related visibility impairment is substantially higher than to days just above the standard.
21 Further, staff recognizes that a concentration-based form better compensates for missing data and
22 less-than-every-day monitoring; and, when averaged over 3 years, it has greater stability and,
23 thus, facilitates the development of more stable implementation programs. Taking these factors
24 into account, staff concludes that consideration should be given to a percentile-based form for a
25 visibility standard.
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.
January 2005
7-13
Draft - Do Not Quote or Cite
-------
1 To identify a range of concentration percentiles that would be appropriate for
2 consideration, staff first concludes that constraints on the number of days in which a standard
3 can be exceeded should be appreciably tighter for a standard intended to protect against serious
4 health effects than would be appropriate for a standard intended to protect against visibility
5 impairment, as noted above. Thus, staff believes that the upper end of the range of consideration
6 should be below the 98th to 99* percentiles being considered for a 24-hour primary PM25
7 standard. Staff has also considered that the regional haze program targets the 20% most
8 impaired days for improvements in visual air quality in Class I areas. If a similar target of the
9 20% most impaired days were judged to be appropriate for protecting visual air quality in urban
10 areas, a percentile well above the 80* percentile would be appropriate to increase the likelihood
11 that days in this range would be improved by control strategies intended to attain the standard. A
12 focus on improving the 20% most impaired days suggests to staff that the 90th percentile, which
13 represents the middle of the distribution of the 20% worst days, would be an appropriate form.
14 To assist in understanding the implications of alternative percentile forms in combination
15 with alternative levels of a standard, staff assessed the.percentage of days estimated to exceed
16 various PM25 concentrations in counties across the U.S., as shown in Figure 7-2. This analysis is
17 based on 2001 to 2003 air quality data, using the 4-hour average concentration from 12:00 to
18 4:00 pm at the maximum monitor in each county. This assessment is intended to provide some
19 rough indication of the breadth of additional protection potentially afforded by alternative
20 percentile forms for a given standard level: Staff notes that a 90th percentile form, averaged over
21 3 years, that allows 10% of the days to be above the level of the standard provides additional
22 protection of visual air quality in far fewer areas at a standard level of 30 ug/m3 than at a level of
23 20 ug/m3.
24 Based on the factors discussed above, staff concludes that a percentile-based form should
25 be considered, based on a percentile at or somewhat above the 90th percentile. Staff believes that
26 a form selected from within this range could provide an appropriate balance between adequately
27 limiting the occurrence of peak concentrations and providing for a relatively stable standard.
January 2005 7-14 Draft - Do Not Quote or Cite
-------
_ o
e
41
en
,2
*
ft
u
fe
I
•d
o
CO
[3A3| Bujpsaoxa sAep 10
I
-------
1 7.3.5 Summary of Staff Recommendations
2 Staff recommendations for the Administrator's consideration in making decisions on the
3 secondary PM2 5 standards to address PM-related visibility impairment, together with supporting
4 conclusions from sections 7.3.1 through 7.3.4, are briefly summarized below. Staff recognizes
5 that selecting from among alternative standards will necessarily reflect consideration of the
6 qualitative and quantitative uncertainties inherent in the relevant information. In making the
7 following recommendations, staff is mindful that the Act requires secondary standards to be set
8 . lhat are requisite to protect public welfare from those effects that are judged to be adverse, such
v
9 that the standards are neither more nor less stringent than necessary. • The provisions do not
10 require that secondary standards be set to eliminate all welfare effects.
11 (1) Consideration should be given to revising the current suite of secondary PM25 standards
12 to provide increased and more targeted protection primarily in urban areas from visibility
13 impairment related to fine particles.
14 (2) The indicator for a fine particle visibility standard should be PM25, reflecting the strong
15 correlation between short-term average PM25 in urban areas across the U.S. and light
16 extinction, which is a direct measure of visibility impairment.
17 (3) Consideration should be given to a short-term averaging time for a PM2 5 standard, within
18 the range of 4 to 8 hours, within a daylight time period between approximately 10:00 am
19 to 6:00 pm. To facilitate implementation of such a standard, consideration should be
20 given to the adoption of FEMs for appropriate continuous methods for the measurement
21 of short-term average PM2 5 concentrations.
22 (4) Consideration should.be given to alternative PM25 standards to provide protection against
23 visibility impairment primarily in urban areas. This recommendation reflects the
24 recognition that programs implemented to meet such a standard can be expected to
25 improve visual air quality in non-urban areas as well, just as programs now being
26 developed to address the requirements of the regional haze rule, for protection of visual
January 2005 7-16 Draft - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
air quality in Class I areas, can also be expected to improve visual air quality in some
urban areas. Recommendations on ranges of alternative levels and forms for such a
standard include:
( a) Staff recommends consideration of a 4- to 8-hour PM2 5 standard within the range
of 30 to 20 ng/m3. Staff judges that a standard within this range could provide an
appropriate degree of protection against visibility impairment, generally resulting
in a visual range of approximately 25 to 35 km, primarily in urban areas, as well
as improved visual air quality in surrounding non-urban areas.
(b) Staff also recommends consideration of a percentile-based form for such a
standard, focusing on a range at or somewhat above the 90* percentile of the
annual distribution of daily short-term PM2 s concentrations, averaged over 3
years.
13 7.4 STANDARDS TO ADDRESS OTHER PM-RELATED WELFARE EFFECTS
14 EPA's decision in 1997 to revise the suite of secondary PM standards took into account
15 not only visibility protection, but also materials damage and soiling, the other PM-related
16 welfare effect considered in the last review. Based on this broader consideration, EPA
17 established secondary standards for PM identical to the suite of primary standards, including
18 both PM2 5 and PM10 standards, to provide appropriate protection against the welfare effects
19 associated with fine and coarse particle pollution (62 FR at 38,683). This decision was based on
20 considering both visibility effects associated with fine particles, as discussed above in section
21 7.3, and materials damage and soiling effects associated with both fine and coarse particles.
22 With regard to effects on materials, EPA concluded that both fine and coarse particles can
23 contribute to materials damage and soiling effects. However, EPA also concluded that the
24 available data did not provide a sufficient basis for establishing a distinct secondary standard
25 based on materials damage or soiling alone. These considerations led EPA to consider whether
26 the reductions in fine and coarse particles likely to result from the suite of primary PM standards
27 would provide appropriate protection against the effects of PM on materials. Taking into
28 account the available information and the limitations in that information, EPA judged that setting
29 secondary standards identical to the suite of PM25 and PM10 primary standards would provide
January 2005 7-17 Draft - Do Not Quote or Cite
-------
1 increased protection against the effects of fine particles and retain an appropriate degree of
2 control on coarse particles.
3 In this review, in addition to addressing visibility impairment, the CD has broadened its
4 scope to include effects on ecosystems and vegetation, discussed in Chapter 6, section 6.3, and
5 also addresses PM-related effects on materials, discussed in section 6.4, and the role of ambient
6 PM in atmospheric processes associated with climate change and the transmission of solar
7 radiation, discussed in section 6.5. In considering the currently available evidence on each of
8 these types of PM-related welfare effects, staff notes that there is much information linking
9 ambient PM to potentially adverse effects'on materials and ecosystems and vegetation, and on
10 characterizing the role of atmospheric particles in climatic and radiative processes. However, on
11 the basis of the evaluation of the information discussed in Chapter 6, which highlighted the
12 substantial limitations in the evidence, especially with regard to the lack of evidence linking
13 various effects to specific levels of ambient PM, staff concludes that the available evidence does
14 not provide a sufficient basis for establishing distinct secondary standards based on any of these
15 ' effects alone. These considerations lead staff to address in the following sections'whether the
16 reductions in fine and coarse particles likely to result from the current secondary standards, or
17 the range of recommended revisions to the primary standards and the secondary PM25 standard
18 to address visibility impairment, would provide appropriate protection against these other PM-
19 related welfare effects.
20 7.4.1 Vegetation and Ecosystems
21 With regard to PM-related effects on ecosystems and vegetation, staff notes that the CD
22 presents evidence of such effects, particularly related to nitrate and acidic deposition, and
23 concludes that current PM levels in the U.S. "have the potential to alter ecosystem structure and
24 function in ways that may reduce their ability to meet societal needs" (CD, p, 4-153). Much of
25 the associated uncertainty surrounding the characterization of the relationships between ambient
26 PM levels and ecosystem or vegetation responses is related to the extreme complexity and
27 variability that exist in predicting particle deposition rates, which are affected by particle size
28 and composition, associated atmospheric conditions, and the properties of the surfaces being
29 impacted. Though several national deposition monitoring networks have been successfully
January 2005 ' 7-18 Draft - Do Not Quote or Cite
-------
1 measuring wet and dry deposition for several decades, they often do not distinguish the form
2 (e.g., particle, wet, and dry gaseous) in which a given chemical species is deposited, so that it is
3 difficult to know what percentage of total deposition is attributable to ambient PM. Further, data
4 from monitoring sites generally do not address all the variables affecting deposition that come
5 into play in a natural system.
6 In addition to these uncertainties, many of the documented PM-related ecosystem-level
7 effects only became evident after long-term, chronic exposures to specific chemical
8 constituents) of PM eventually exceeded the natural buffering or assimilative capacity of the
9 system. In most cases, PM .deposition is not the only source of the chemical species to the
10 affected system and the percentage of the deposition due to ambient PM is often not known.
11 Because ecosystems have different sensitivities and capacities to buffer or assimilate pollutants,
12 it is difficult to predict the rate of deposition that would be likely to lead to the observed adverse
13 effects within any particular ecosystem. Equally difficult is the prediction of recovery rates for
14 already affected areas if deposition of various chemical species were to be reduced.
15 Despite these uncertainties, a number of significant and adverse environmental effects
16 that either have already occurred or are currently occurring have been linked to chronic
17 deposition of chemical constituents found in ambient PM. Staff notes, for example, mat the
18 following effects have been linked with chronic additions of nitrate and its accumulation in
19 ecosystems:
20 • Productivity increases in forests and grasslands, followed by decreases in productivity
21 and possible decreases in biodiversity in many natural habitats wherever atmospheric
22 reactive nitrogen deposition increases significantly and critical thresholds are exceeded;
23 • Acidification and loss of biodiversity in lakes and streams in many regions, especially in
24 conjunction with sulfate deposition; and
25 • Eutrophication, hypoxia, loss of biodiversity, and habitat degradation in coastal
26 ecosystems.
27 Staff notes that effects of acidic deposition have been extensively documented, as
28 discussed in the CD and other reports referenced therein. For example, effects on some species
29 of forest trees linked to acidic deposition include increased permeability of leaf surfaces to toxic
30 materials, water, and disease agents; increased leaching of nutrients from foliage; and altered
January 2005
7-19
Draft - Do Not Quote or. Cite
-------
1 reproductive processes; all of which serve to weaken trees so that they are more susceptible to
2 other stresses (e.g., extreme weather, pests, pathogens). In particular, acidic deposition has been
3 implicated as a causal factor in the northeastern high-elevation decline of red spruce. Although
4 U. S. forest ecosystems other than the high-elevation spruce-fir forests are not currently
5 manifesting symptoms of injury directly attributable to acid deposition, less sensitive forests
6 throughout the U.S. are experiencing gradual losses of base cation nutrients, which in many
7 cases will reduce the quality of forest nutrition over the long term.
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 to the prevention of further degradation of sensitive ecosystems and vegetation.
12 Staff 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 may 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 revisions to
20 the primary standards would contribute to increased protection against PM-related effects on
21 ecosystems and vegetation. Staff recommends that the potential for increased protection of
22 ecosystems and vegetation be taken into account in considering whether to revise the current
23 secondary PM standards. Further, staff believes that any such increased protection should be
24 considered in conjunction with protection afforded by other programs intended to address
25 various aspects of air pollution effects on ecosystems and vegetation, such as the Acid
26 Deposition.Program and other regional approaches to reducing pollutants linked to nitrate or
27 acidic deposition.
January 2005 7-20 Draft - Do Not Quote or Cite
-------
1 7.4.2 Materials Damage and Soiling
2 With regard to PM-related effects on materials, staff notes that the available evidence
3 continues to support the following observations:
4 • Materials damage and soiling that occur through natural weathering processes are
5 enhanced by exposure to atmospheric pollutants, most notably SO2 and particulate
6 sulfates.
7 * While ambient particles play a role in the corrosion of metals and in the weathering of
8 paints and building materials, no quantitative relationships between ambient particle
9 concentrations and rates of damage have been established.
10 • Similarly, while soiling associated with fine and coarse particles can result in increased
11 cleaning frequency and repainting of surfaces, no quantitative relationships between
12 particle characteristics (e.g., concentrations, particle size, and chemical composition) and
13 the frequency of cleaning or repainting have been established.
14 Staff believes that these observations and the underlying avail able evidence continue to support
15 consideration of retaining an appropriate degree of control on both fine and coarse particles.
16 Lacking any specific quantitative basis for establishing distinct standards to protect against PM-
17 related adverse effects on materials, staff recommends consideration be given to (1) retaining the
18 current secondary PM2 5 standards or revising those standards to be consistent with any revisions
19 made to the primary PM2S standards or to the secondary PM25 standards to address visibility
20 impairment, and (2) retaining secondary standards for coarse particles, using a PM,0.2 5 indicator
21 consistent with the primary standards, at a level that either retains the degree of protection
22 afforded by the current PM10 standards or that is consistent with any new PMi0.2 5 primary
23 standards.
24 7.4.3 Climate Change and Solar Radiation
25 With regard to the role of ambient PM in climate change processes and in altering the
26 penetration of solar UV-B radiation to the earth's surface, staff notes that information available
27 in this review derives primarily from broad-scale research and assessments related to the study of
28 global climate change and stratospheric ozone depletion. As such, this information is generally
29 focused on global- and regional-scale processes and impacts and provides essentially no basis for
30 characterizing how differing levels of ambient PM in areas across the U.S. would affect local,
January 2005 . 7-21 Draft - Do Not Quote or Cite
-------
1 regional, or global climatic changes or alter the penetration of UV-B radiation to the earth's
2 surface. As noted in section 6.5, even the direction of such effects on a local scale remains
3 uncertain. Moreover, similar concentrations of different particle components can produce •
4 opposite net radiative effects. Thus, staff concludes that there is insufficient information
5 available to help inform consideration of whether any revisions of the current secondary PM
6 standards are appropriate at this time based on ambient PM's role in atmospheric processes
7 related to climate or the transmission of solar radiation.
8 7.4.4 Summary of Staff Recommendations
9 Taking into account the conclusions presented in sections 7.4.1 through 7.4.3 above, staff
10 makes the following recommendations with regard to PM-related effects on vegetation and
j
11 ecosystems and materials damage and soiling:
•».
12 (1) Consideration should be given to retaining secondary standards for fine and coarse
13 particles that at a minimum retain the level of protection afforded by the current PM2 5
14 and PM10 standards so as to continue control of ambient particles, especially long-term
15 deposition of particles, especially particulate nitrates and sulfates, that contribute to
16 adverse impacts on vegetation and ecosystems and materials damage and soiling.
17 (2) For consistency with the primary standards, secondary standards for fine and coarse
18 particles should be indexed by PM25 and PMi0_25. While staff recognizes that PM-related
19 impacts on vegetation and ecosystems in particular are associated with chemical
20 components in either size fraction rather than with particle size per se, staff also
21 recognizes that sufficient information is not available at this time to recommend
22 consideration of an ecologically based indicator in terms of a specific chemical
23 ' component of PM.
24 In making these recommendations, staff has taken into account both the available
25 evidence linking fine and coarse particles with effects on vegetation and ecosystems and material
26 damage and soiling, as well as the limitations in the available evidence. In so doing, staff
January 2005 7-22 Draft - Do Not Quote or Cite
-------
1 recognizes that the available information does not provide a sufficient basis for the development
2 of distinct national secondary standards to protect against such effects beyond the protection
3 likely to be afforded by the suite of primary PM standards.
4 7.5 SUMMARY OF KEY UNCERTAINTIES AND RESEARCH
5 RECOMMENDATIONS RELATED TO STANDARD SETTING
6 Staff believes it is important to continue to highlight the unusually large uncertainties
7 associated with establishing standards for PM relative to other single component pollutants for
8 which NAAQS have been set. Key uncertainties and staff research recommendations welfare-
9 related topics are outlined below. In some cases, research in these areas can go beyond aiding in
10 standard setting to aiding in the development of more efficient and effective control strategies.
11 Staff notes, however, that a full set of research recommendations to meet standards
12 implementation and strategy development needs is beyond the scope of this discussion.
13 With regard to welfare-related effects, discussed in Chapter 4 of the CD and Chapter 6
14 herein, staff has identified the following key uncertainties and research questions that have been
15 highlighted in this review of the welfare-based secondary standards:
16 (1) Refinement and broader application of survey methods designed to elicit citizens'
17 judgments about the acceptability of varying levels of local visibility impairment could
18 help inform future reviews of a visibility-based secondary standard. Such research could
19 appropriately build upon the methodology developed by the State of Colorado and used
20 as a basis for setting a visibility standard for the city of Denver, which has been adapted
21 and applied in other areas in the U.S. and abroad.
22 (2) There remain significant uncertainties associated with the characterization and prediction
23 of particle deposition rates to natural surfaces in general, and most importantly, with
24 respect to nitrogen deposition in particular. Reduction in these uncertainties will be key
25 to developing the capability of quantitatively linking ambient PM concentrations with
26 environmental exposures and response. In order to better understand the nature of the •
27 role that PM plays in cumulative long-term environmental impacts, more research needs
January 2005 7-23 Draft - Do Not Quote or Cite
-------
3
4
5
6
7
8
9
10
11
to be conducted on the percentage of total deposition contributed by PM and where
necessary, better tools and monitoring methods should be developed.
(3) The immense variability in sensitivity to PM deposition across U.S. ecosystems has not
yet been adequately characterized, specifically the factors controlling.ecosystem
sensitivity to and recovery from chronic nitrogen and acid inputs. Data should be
collected on a long-term basis on a greater variety of ecosystems in conjunction with the
development of improved predictive models. Such research could help in future
consideration within the U.S. of the "critical loads" concept, which is generally accepted
s
in Europe as the basis for abatement strategies to reduce or prevent injury to the
functioning and vitality of forest ecosystems caused by long-range transboundary chronic
acidic deposition.4
t
4 This recommendation is consistent with the views of the National Research Council (NRC) contained in
its recent review of air quality management in the U.S. (NRC, 2004). This report recognizes that for some resources
at risk from air pollutants, including soils, groundwaters, surface waters, and coastal ecosystems, a deposition-based
standard could be appropriate, and identifies "critical loads"as one potential approach for establishing such a
deposition-based standard.
January 2005
7-24
Draft - Do Not Quote or Cite
-------
1 REFERENCES
2 Langstaff, John E. (2004). Estimation of Policy-Relevant Background Concentrations of Particulate Matter.
3 Memorandum to PM NAAQS review docket OAR-2001 -0017. January 27,2005.
4 National Research Council (NRC) (2004). Air Quality Management in the United States. Committee on Air Quality
5 ' Management in the U.S., National Research Council of the National Academy of Science. The National
6 Academies Press, Washington, D.C. ISBN 0-309-08932-8.
7 Schmidt et al. (2005) Draft analysis of PM ambient air quality data for the PM NAAQS review. Memorandum to
8 PMNAAQS review docket OAR-2001-0017. January 31,2005.
January 2005 7-25 Draft - Do Not Quote or Cite
-------
t
1 APPENDIX 2A. Source Emissions
- 2 '
3 The distribution and amount of emissions of pollutants that contribute to ambient PM can
4 provide insights into observed ambient levels. The links between source emissions and ambient
5 concentrations of PM can include complex, non-linear atmospheric processes, including gaseous
6 chemical reactions and pollution transport.
7 Source emissions can be measured using monitoring equipment or estimated using
8 emission inventory methods. For most source types, emissions inventory methods are the most
9 practical. The EPA routinely publishes national estimates of annual source emissions of
10 pollutants that contribute to ambient PM concentrations. In general, national emissions estimates
11 are uncertain, and there have been few field studies to test emission inventories against
12 observations. The draft CD concludes that uncertainties in national emissions estimates could be
13 as low as ±10 percent for the best characterized source categories (e.g., SO2 from power plants
14 measured by continuous instruments), while fugitive dust sources should be regarded as order-
15 of-magnitude (CD, p. 3-98). The EPA is working to reduce these uncertainties through advances
f 16 in the understanding of the fate and transport characteristics of fugitive dust emissions released
17 at ground level. Episodic emissions from dust storms and forest fires are difficult to quantify and
18 to allocate accurately in space and time, and discerning between natural and anthropogenic
19 "causality" for these source categories is especially challenging.
20 Table 2A-1 provides a summary of recent annual estimates of national emissions of
21 primary PM and PM precursors. While reviewing the following discussion on emissions
22 estimates, the reader should keep in mind that national estimates, while instructive, can obscure
23 important distinctions in the relative contributions of different sources across smaller geographic
24 regions, including important differences between urban and rural areas.
25
26 Primary PM Emissions
27 The majority of directly emitted anthropogenic PM is estimated to be coarse particles.
28 Though highly uncertain, recent national estimates of PMi0.25 emissions (total of all sources)
29 shown in Table 2A-1 are about 2.5 times higher than estimates of PM25 emissions -16.3 million
30 short tons compared to 6.6 million short tons. A large portion of primary PM emissions are
January 2005 2A-1 Draft - Do Not Quote or Cite
-------
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
attributed to a variety of small area-wide sources, which are often more difficult to characterize
and are more uncertain than larger point source emissions.
National estimates of primary PM10_25 are dominated by fugitive dust and agricultural
sources. Fugitive dust sources include paved and unpaved road dust, dust from construction and
agricultural activities, and natural sources like geogenic wind erosion (not estimated in Table
2A-1). Fugitive dust is also a significant source of primary PM25. Unlike PMI0.25, where
fugitive dust emissions comprise about 75 percent of total emissions, fugitive dust emissions of
PM2S is only about one-third of total emissions. Recent research has found that about 75 percent
of these emissions are within 2 meters of the ground when measured. A significant portion of
these coarse-mode particles are removed or deposited within a few kilometers of their release
point due to turbulence associated with surface topography, and the presence of vegetation or
structures (DRI, 2000). This is consistent with the generally small amount of crustal material
found in ambient PM25 samples in most locations. As shown in Table 2A-1, direct emissions
from fuel combustion, industrial processes, fires, and motor vehicles contribute more to primary
PM2.5 than to primary PM10.2 5. Recent improvements to methodologies for estimating
emissions, reflected in the values in Table A-l, include:
• Wildfires and prescribed burning - use of state-specific fuel loading factors and improved
emission factors
• Residential wood combustion (woodstoves & fireplaces) - recalculation of emissions
using updated wood consumption data
• Condensible PM emissions - added these emissions; were not previously included
• Animal husbandry - updated NH3 emissions for this category based on recent work by
EPA's Emission Standards Division/OAQPS
• Mobile source emissions - updated estimates using the latest MOBILE and NONROAD
models
Secondary PM Precursor Emissions
Major precursors of secondarily formed fine particles include SO2, nitrogen oxides
(NOJ, which encompasses NO and NO2, and certain organic compounds. Table 2A-1 shows the
estimated contribution of various sources to nationwide emissions of SO2 NOX, VOC, and NH3.
January 2005 2A-2 Draft - Do Not Quote or Cite
t
-------
1 Fuel combustion in the power generation and industrial sectors dominates nationwide estimates
2 of SO2 emissions and contributes significantly to NOX emissions. However, emissions from
3 motor vehicles comprise the greatest portion of nationwide NOX emissions. Motor vehicle
4 emissions also make up a substantial portion of nationwide VOC emissions, with additional
5 contributions from the use of various solvents in industrial processes and commercial products.
6 The vast majority of nationwide NH3 emissions are estimated to come from livestock operations
7 • and fertilizer application, but in urban areas there is a significant contribution from light-duty
8 cars and trucks, as well as certain industrial processes.
9 The relationship between changes in precursor emissions 'and resulting changes in
10 ambient PM25 can be nonlinear. Thus, it is difficult to project the impact on ambient PM25
11 arising from expected changes in PM precursor emissions without air quality simulation models
12 that incorporate treatment of complex chemical transformation processes and meteorology.
13 Generally SO2 emissions reductions lead to reductions in sulfate aerosol, and NOX emissions
14 reductions lead to reductions in nitrate aerosol. However, the direction and extent of changes
15 will vary by location and season, depending on fluctuations in NH3 emissions and changes in
16 prevailing meteorology and photochemistry.
t
January 2005 2A-3 Draft - Do Not Quote or Cite
\
-------
p
2
w *
e «
a. s
•5 S3
A *»
I =
•g s
•- "C
2 o
"I w
il
i.i
S T3
w
Oi
vt •
3^2
fk
£3 Ofl t^
*^ S o
^ a a,
%**
6 a)
P
O jj
'u' S3 ^
^ ts
O V,
in 5
o\ (X
^^ 1H m_.
i °"^
G
g ^
^ ^"u
M f^ ^
0 ^
• S V!
.s •§
f i
<. §
^ O
Original study*
Study Location
Reanalysis study
•.
-
e
5
f
wmmmwm
'
s ^ ^ If s |
5 si xtL S-"" ^""* i
\ * * 1 3|
5 a, a, °" gs i
i i i i -
*-:
*s"'
! ! ! ! **"»"
! ! ! •
iii' ?f)
«->
^ "^ § "^
r^ ^^ ^ *-s , 1
^- sL ^L ^
s s s s
*tt tS "tt "tS
§ 3 3 § S
•*M i»j *« '*•* ?*
§o o o «
C K S !>3
!5 »5 -5 •§ *§
^ ^ ^ N ^
O C3 C) O O
O\ \j)
t |3afe,1 *
|tj ^-| ^f | j "§
1 | 11 "3r§ Jf «
•a S ^^ ^ ^ ^ s fc
a la ii il i
^ ^ 0, to <%
va
tC
T "l" ^ °0" oO \o~ OC 'O*
Ol
o
5*
t
vo
S
1^
**»
t
-------
u
g
S
32.3)
2L9)
^ Ci
aa&
vo \o 2
iX O\ —-
^ ** ^
0 »> cy
aaV
Q
7.8-45.6
. 2-29. 6
6. 1
1
means 1
s 6.6-
PM1
PU2
PMt
*f ^ _;
vi 2 ^ O 3
a 2 is if I
9 9
°° i
°! vi
P, P.
vq p
TT oo
dc5 5 3
O C3 C) O
•0
o o
'5 00
w Z
0 0
00
T ^-
rf -H"
O ^
£ !"
5
o o"
SO vi
(S -^
^i r-
•" cs
o
0
o5 &:
"S.^
"s^i
j 1
I i i
5-S -S
Scartz et al., 1
Portage, WI
Schwartz, 2003a
Schwartz et al., 1
Topeka, KS
Schwartz, 2003a
3a
2
t aL,
Over
Schwartz
6 Cities,
Schwart
«
o
•
en
1
-------
I-r
—r w
?|
a §
8*
O
S
u
u, <->
o
m
C3\
I
* It
>>-9 R
1
O 03
'
u-i \O O
P?
t-H
t~-
^ m t> -3
S3 i %
oo
ocT oC
oo m
K oo
o
O
ii
r-
(S
S I
o o
c\ ^ vq i": —i vo t
oj 2 o oo -^ in <
"*! 3 s 2 3 a
s" 2 2^^" 2 '
c-
CO
n
.a s
•30 T5
tS -S § '
CS "
(S P"
o "1
s-
S
r-
od
w "5 S
z a w
o o
r~* ^o
vo r-
•-< fS
.-. « o
q1 m t-^ c-
o •7 *7 "7
O OO O OO O OO OO
i
>-,
| o g
a s
s.-g-
s o
fi X!
_; o
S|
vb Ix
CO
*' ^"n "^w *•».*:
2j• o
5 M S
O
3
O)
-------
PM10.
Levels
"
PM10, PM2 j
Mean (Ran
Repo
u
vi
8-1
a.
o
*
>n
a!
0,
*
u
on
y
y Lo
lysis
S
R
-r^" q
i 80
p ^ V?
;«r«? 2
oo" ff
o'q^
is
o o
P
P
5SJ 5S3 §
§E g§2-
DH
*> O
00 >"
o o
q °!
o o
3
°\
f>
m
•n
I—I fl pvj t-"1
vi o1*
I
OH
o
!|I|i'|\glS
;S^f> >U> ^JJH
!J?« o3*jOS ^HnQ8>
D^M OOT500 ~^-a
o> 8 *
.$^$ a -g j» "T- a i
."3 § "3 3o"o £§S
I -Si j ° I j M a
I
I
a
I
en
-------
o
1
v>
*
^
60
"->
cs
*
in
O-l
-a -g
M O
a
"0
cs
I
r-
(N
6
o
I
s
0)
O
I
en
o\
CM
1
O
a
G
I
a
•3
ITS
00
s
o
r-;
-------
#«
&
2:
o
^ u 0
U S n
S S-
I
y
S rf
C>
vi
Vlf
» J •§
"3 >..2?
R
ts
vq
«n
O
t>
£
CO
i
a
00
00
v o ^ o
-------
Repo
t, 0*
B «
&
tS
O
— t- »n
u
Os
PM1047
a,
vo" m
IT) O,
06 od
3 o
* 'S
o o
O
ey, 1999
nta Clara Co
Fair
vo
S
vo
vo
•fi 60
=> 7; 00 __
S P " g
=>:^-Q ^.
•<. o rvi cs
00 r—
sag s « ao ^2
s
a
o
o
00
O O
vq o
fS rrj
O O
- §*
S 2*£
f-; O
fl 00
"* (N
VD
'oO
0 ~
oo a!
C-2-
r- >n
Os ON
<*>
O O
rr ~
^"«?
«> t-
ts ^
SO vo
T5
IS
3S
o o
c
o\,o o . o Q"SO
SS1O IN»J(S (S.JC-* O^S
3o -«J- »'-1-rs»>r»
M<;<^ 53"«3S S "S -oa'r
JSa),g jg o S ^3=5-3
|,3 fe S° "
etal
1 1 u 1 -
^ « l> «>-S J> « si
ooo ooo ^o
£J3 .3^,0 -go
22220
G
I
a
i
>0
-------
t
1
a
.
oo
O ^"i
i-; d
en
o
CM"-*
CO P1^
T.
TT
o
SS
O O
O
1 IB I3'
•n
I
-------
<£• *
es
33
11
£*
Ct
5P
OfJ
22?
8
CN
00
y—V 00
f: G^ ci
•n ^ in <*>
$S~$ %
o\ °; HI —i
M M « ,x
O
•^
00
tn
rf
S
cs
ON
oq
S
•° *«•
§48
. CN of es
,« JJ «
111
S < S
3> w ^P
J3J
o
T3
~O
•a d-
-5 •§
* ^
4> *J
sw
8P
I"
S
2 00
cK 9
„
-------
^
«i ^
_
1 ^ I
„ 31 €
«« § ^
S ^(2
S §'
g"l
G ~
x^ ^
0
vo;^
00 2
J? -
£ 2*
fa
Av
. '
1
H
^
0
0
^_^
«*1
—
P
vU
r-i
P
^
J^
C5
Z
J
O
Dysrhythmias Hospital Admi
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito (2003)
v
Q
g
en
1
1
a.
1
CA
«
U
M
5
r
Heart Failure/Congestive He
"^\ vS*
00 o
?f t'
v*" ^
§ rr
5 g
•§ 1
a. g
i
j
«s
oq
"S
i
o
Schwartz and Morris, 1995
Detroit f> 65 years)
rs
2
&
«n
29 yeare
^^
in /*^v £^~ ' f^
2 £g «i ^
g vo; S ~- '
ooo 2 ^
~ 1 3 """
"> 3 2 °
oJ" ± ^
rx^fc
/— v ^
*n •
m 2
•*•' H '
vL-^
t 0\
TT
y— v / — \
O 00
12 2
•5T f| !
q oq
ob vo
c?P g
o^ ao v«
^•H ^-1 • ^
25 S?
^~i/ ^
S 5 ^
13
2 |
o § §
w O O
Lippmann et al., 2000
Detroit, MI (>65 years)
Ito, 2003
Morris and Naumova, 1 998
Chicago, IL (>65 years)
B
_o
S
a
'
3
Myocardial Infarction Hospi
fN|
2
in"
m
>n
o
2*
fc
j
i
/^^
vd
S
e
S
*"
"8
3
1
o
Linn etal., 2000
Los Angeles, CA (>29 years
X-N /~\ X-N
(N n in v-i
O O O
2" S 2
fr, E OH
i i 1
i i i
£T ST Q
O P-. c\
•*fr ts o\
ro PO ^f
CT» ^^ ^n
T *? ^
O r*^ .
• to
i— i O • >tj
1 "8 • "8
3 3 3
•s -s -s
i; S (3 C
.So go o
a -g
~ ,s
i . - 1 ^ ^
CS MX 'IT^ •« W
-S b <^ £ 2 b
"S. S "• * S
o ^ a >> 'i ^
B 3 g. oj .2 g^
.2 & I t 1 ^
1 .-3 * < ^ ^
£'o° ,2 oo •« o1^
! a.| I S4 fM
u "3S 5 "3 ?• Da "3 S
a -« < }• •«< »-«<<
1. J Mil ia
w 3 u a J oS j
1
a
<^^
i^
S
•^j
O
5
^
a
i
^
k.
Q
o
-------
1
u
^
£
'-' t, "
u u
|
PM
25
%
sr*
OS n in
OS OS
— ^ O
I"- I-"
«N 00
22
S o
b
9
I
I
(H
U
t.
o
22
OH OH
O
0
o
o
— i
«
•^ v£> ^T v:
111 I
o
•a 2
» ^
s (§
PM1035.2(S
C-
ro
ss
on
v
1 8
00
0
1
O
1
o
«" B
a o
s •
J ffl
I
0
I
a
CO
O
o
o
•a">
«a
^1
I*
-------
sa
>n
T
1—4
ON"
\o
oo
00
ts
t
O
3
1
1
o
S
8
rn
cs
8
S
\d
fi
O
o
>%
QA ^7
2^
•a'--
TtS s
tc o
gS
.a
r-
<»?
T—*
^S
00
OO
1
"S
g
*S
.2
1
•<
1
E.
M
£
i
a*
.2 5^,
g tf-
a
1 i
£ I
a
1
o*
G
I
a
i
I
cs
3
S
-------
||
f*H "1 «
g 'ST'S
•0 S a
:£t &
S *;;&
& a
P
0 ,
£ i
^ H ^
a ^"s
nj -5
Q Esfl
1 a
0
^ M
i/i 2
0^ Oi
^ S^g
g "So
g
*
? a
1, i
1*1
g a.
1 *
» f
fi-
^ 0
Original study*
Study Location
ReanaJysis study
«jj
*»
S
i
o
s
p
'
I
oo o vo'cn'
"2 ""2
oC oT <*f in"
vi TT SM
V™> \s_X \^>
00 tr> <* m
00 00* ^ \O
5 5
O O ^Z fc
•*<-* •«-• <^F< ^r<
_65y
Zanobetti and Schwart
?ffs
I"H QQ
l-ej
s2al
*5?" n S
&, *
,_,
o »
vS
00 00
< M
OZ
S O
s
.5
•t?1 S
Lippmann et al., 2000
Detroit, MI (>65 year
Ito 2003
COPD Hospital Admi
1^1 /* "N /*^\ -^^
&=> ^ s 8j> sj !d in ,«« 1"H "o
^ Cs 2 ^" fn .: 5J
S* s 2 2^ -
S fc 0-
^v?
5J IN
II ! *o" **•
' ' 1 TT 7
^oo'
<» 2
ON1 O>"
H o
i i i ££
vo ON
1 I
O fO
-i
u ^T ^ S
|89 ofg.s
8s §13 |s2t
5^- ^s i -I, 1^^
£•* ««'§ 31 i-fs
Il«^o " " ll^
•§ Q 1 ^ s .! »3 .§! <5 ^j"
1
^v
I/I
-------
3.9
«.
1
H
1st
ji
0*
&
0,
u
£
v>
c\
|
1
O
5n
¥
C9
1
*
§
* e^
W>
v^X
!
t
ji
*~^ TJ
u
ftS
ft
g<2
<3
i
«*.£
(X
°"^
^
g
u g
01 "5b
i^
^
PL,
l\
a.
8
1
o
O
Original study*
Study Location
Reanalysis study
S§ '
r~ 4 ^
•If £j
g §
•3 «§
= *>
'ss
£ CL,
,-
o^ P S"
(^ CJ "1
r- «ri •"
8 R S
si • e 3,
S5 fc ov
VO OO (o
•* ^ cs
S? -^ '-*
<"">.?! Ov
S oo 06
-r «A ri
S "^ S
5 e d
^ C4 0
* W) p
< < 1»
O O ^ 2 ^
S o -P o jr o
O3 m 03 •— U «— i
Moolgavkar, 2000c
Los Angeles, C A (all ages)
Moolgavkar 2003
VI f*^ (S
VO VO ^
in „ -n -
t— r ^ vf
>/l ^^ tfl
rf +*•*-*
g 8 • S-
5 S S
O O C3
Asthma Hospital Admissions:
Choudbury et al, 1 997
Anchorage, AK
Medical Visits (all ages)
Jacobs et al, 1997
Butte County, CA (all ages)
Linn etal, 2000
Los Angeles, CA (>29 years)
§
oC
cs
— i
o
ex.
i
i
a\
vi
t-"
r-i
v^/
*J
5
O
fl
cs
c^-
o
Ov"
^^
s
Q\ O\ \
-*\
TV
S
3
Ts
1>
S
a o* ""i ^ vd
PK
^
1
I
Lipsett etal, 1997
Santa Clara Co, CA (all ages)
w,
00
1
1
1
Vs!
**!
fn"
o
1
5
O
Nauenberg and Basu, 1999
Los Angeles, CA (all ages)
OO
o
(X,
i
1
p
CN
H"
~
; not used
>
O
Tolbert etal, 2000
^H \D f
O "1
. PH &,
<^5 ^^
• (
SM"
vi ^
v\ v\
. o
s q.
cio
r- in
00 VO
<3~
O\ '
*-* ^2
«f ^"
2 «>
< 00
^2
i^
Atlanta, GA (<17 years)
Sheppardetal, 1999
Seattle, WA (<65 years)
-H
e*
£
Sheppard et al, 2003
O)
fe
I
8
i
<
.PI
b
-------
»- &
5 „
°<
o
£
o
.S
•i
•S
«s W*
§§
-~
K t ^
flo
ast
-------
tr,
PM10, PM25 and PM
Mean (Range) Lev
Reported**
? a
5n 3
^ H S
1 a t
N» CS
£••
o 3
a i!
^"^ JM **L-
S3 Hi "^
§ S
£5
'S tS
*
0 _ =
£h 2
\_^ . —
g "Sb
<3 "*"
* *
•§" 5
1°
Original study*
Study Location
Reanalysis study
P£> /•— \ ^"S
00 |> 00
cs vS r4 wS
oo oo oo oo
O "1 O ""!
^^ ^S^ wr* i_y
i i
: :
? ' ?
i? ^
—< £5 •' *O§H?
•3r "E . ^ ® , * « -^
p§2 #x2 pxiS.**
| j a « w a -g a a |
O -S »5 ^ w5 xS S1
rf
PM,a median 22. 1(0
159.0) (north site)
|
i
R
2
i-i
o
»— *
not used
2j
o
Vedal et al., 1998
Port Alberni, CAN
(children, cough)
O)
fe
I
-------
S
1
o
S I
¥ fc^
g
s
^ "53
8
O
etal.,1998
C
•8
§
1
o
W °^
I, J
^ a
'0 «
nose
8
o
etal.,1998
Alberni, C
(S
•-Tn
TO £*
I?
a 55
vq
s
8-
VO
O
998
i, C
^1
tJ 5
^o 1$
P
ts
s"
o o
l-
•dl"?
"S % S
o
•rfJ^
is < S
ill
r> ,o.
s
I
O
8
Ov
•3*
-s
I
>
f
C
•<] g
Jl
»:
I
*1
B§
a
J!
I
OJ
r*
i—i
<:
-------
S
nJ !
Q
£
8.
^ •< K
Ml
.3'
i
i
tudies in italics
mean (minimu
a
oo
I
b
-------
APPENDIX 3B. Mortality and Morbidity Effect Estimates and PM Concentrations
from U.S. and Canadian Studies for Long-Term Exposures to PM10, PM2 s, and PM10 2 s
Study
Indicator (Increment)
Relative Risk (95% CI)
Study
Concentrations
(us/m3)
Increased Total Mortality in Adults
Six City*
Six City8
ACS Studyc
(151 U.S. SMSA)
Six City ReanalysisD
ACS Study ReanalysisD
ACS Study Extended
Analyses5
Southern California1*
(
Southern California8
Veterans Cohort0
PM15/10(20ng/m3)
PMj.5 (10 fig/m3) .
SC-: (15 Mg/m3)
PMIMJ (10 Mg/m3)
PMj ., (10 Mg/m3)
SOJ (15 ug/m3)
PM15/10(20Mg/m3)
PMj., (10 Mg/m3)
PM15/1()(20ug/m3)(dichot)
FMu (10 Mg/m3)
PM^OOug/m3)
PM25 (10 ug/m3) (1979-83)
PMj5 (10 Mg/m3) (1999-00)
PMj5 (10 Mg/m3) (average)
PM10(20ug/m3)
PM10 (30 days/year>100 ug/m3)
PMIO (20 ug/m3)
PM10 (30 days/year>100 Mg/m3)
PH., (10 Mg/m3)
PM1M.5(10Mg/m3)
PM2.5 (10 Mg/m3) (1979-81)
1.18(1.06,1.32)
1.13(1.04,1.23)
1.54(1.15,2.07)
1.43(0.83,2.48)
1.07(1.04,1.10)
1.11(1.06,1.16)
1.19(1.06,1.34)
1.14(1.05,1.23)
1.04(1.01,1.07)
1.07(1.04, 1.10)
1.00(0.99,1.02)
1.04(1.01,1.08)
1.06(1.02,1.10)
1.06(1.02,1.11)
1.09(0.99, 1.21) (males)
1.08 (1.01, 1.16) (males)
0.95(0.87, 1.03) (females)
0.96(0.90, 1.02) (females)
1.09(0.98, 1.21) (males)
1.05(0.92, 1.21) (males)
0.90 (0.85, 0.95) (males)
NR(18,47)
MR (11, 30)
NR(5, 13)
18U(9, 34)
llu(4,24)
NR(18,47)
NR(11,30)
59(34,101)
20 (10, 38)
7.1(9,42)
21 (9, 34)
14(5,20)
18(7.5,30)
51 (0, 84)
51 (0, 84)
32 (17, 45)
27 (4, 44) '
24 (6, 42)
Increased Cardiopulmonary Mortality in Adults
Six City*
Six City ReanalysisD
ACS Studyc
ACS Study Reanalysis0
ACS Study Extended
Analyses8
PM15/10(20Mg/m3)
PMj5 (10 Mg/m3)
PM15/10(20(xg/m3)
PM2.5 (10 Mg/m3)
PMu (10 Mg/m3)
PM15/lc(20Mg/m3)(dichot)
PM2.5(10Mg/m3)
PM15.2.5(10Mg/m3)
PMJS (10 Mg/m3) (1979-83)
PM2 5 (10 Mg/m3) (1999-00)
PMj5 (10 (ig/m3) (average)
V
1.18(1.06,1.32)
1.20(1.03,1.41)
1.19(1.07,1.33)
1.12(1.07,1.17)
1.07(1.03,1.12)
1.12(1.07,1.17)
1.00(0.98,1.03)
1.06(1.02,1.10)
1.08(1.02,1.14)
1.09(1.03,1.16)
MR (18, 47)
NR(11,30)
NR(18,47)
NR(11,30)
18U(9, 34)
59 (34, 101)
20 (10, 38)
7.1(9,42)
21 (9, 34)
14(5,20)
18(7.5,30)
January 2005
3B-1
Draft - Do Not Cite or Quote
-------
Study
Southern CalifomiaF
Southern California15
Increased Lung Cancer
Six City*
Six City ReanalysisD
ACS Study0
ACS Study Reanalysis"
ACS Study Extended
Analyses6
Southern California'
Southern California11
Increased Bronchitis in
Six City1
24 CityJ
AHSMOG*
12 Southern California
communities1"
(all children)
12 Southern California
communities"
fchildren with asthma)
Indicator (Increment)
PM,o (20 ug/m3)
PMjiOOMg/m3)
PM1M.5(10Mg/m3)
Mortality in Adults
PMMO(20Mg/m3)
PM2.5(10Mg/m3)
PM1J/10(20Mg/m3)
PMj.5 (10 Mg/m3)
PM2.j (10 Mg/m3)
PHs/jo (20 ug/m3) (dichot)
PMj3 (10 Mg/m3)
PM15.2.5 (10 Mg/m3)
PM2 5 (10 Mg/m3) (1979-83)
PMj 5 (10 Mg/m3) (1999-00)
PM35 (10 Mg/m3) (average)
PM10 (20 Mg/m3)
PM-j, (10 Mg/m3)
-
Children
PM15/10(20Mg/m3)
PM2.5(10Mg/m3)
SOI (15 Mg/m3)
PMj.OOMg/m3)
PM10(20Mg/m3)
SOT (15 Mg/m3)
PM10(20Mg/m3)
(1986-1 990 data)
PM,o (20 Mg/m3)
PMj.5 (10 Mg/m3)
Relative Risk (95% CI)
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)
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),
Study
Concentrations
(Jis/m3)
51 (0, 84)
32 (17, 45)
27 (4, 44)
NR(18,47)
NR(11,30)
NR(18,47)
, NR(11,30)
18" (9, 34)
59(34,101)
20 (10, 38)
7.1(9,42)
21 (9, 34)
14(5,20)
18(7.5,30)
51 (0, 84)
32 (17, 45)
27(4,44)
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
communitiesL
(all children)
12 Southern California
communities**
(children with asthma)
PM)0 (20 Mg/m3)
(1986- 1990 data)
PM10 (20 Mg/m3)
PM, .5 (10 Mg/m3)
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
January 2005
3B-2
Draft - Do Not Cite or Quote
-------
Study
Indicator (Increment)
Relative Risk (95% CI)
Study
Concentrations
Cue/in3)
Increased Airway Obstruction in Adults
AHSMOGK
PM10(20ng/m3)
1.19(0.84,1.68)
NR
Decreased Lung Function in Children
Six City1
24 City3
12 Southern California
communities1"
(all children)
12 Southern California
communities11
(all children)
12 Southern California
communities'5
(4th grade cohort)
12 Southern California
communities'5
(4* grade cohort)
12 Southern California
communitiesR
(second 4* grade
cohort)
12 Southern California
communities11
(second 4* grade
cohort)
12 Southern California
communities11
(second 4* grade
cohort)
PM15/10(50ng/m3)
S0:(15ng/m3)
PMj L (10 ng/m3)
PM10 (20 ng/m3)
PM10(20ng/m3)
(1986-90 data)
PM10 (20 ng/m3)
(1986-1 990 data)
PM10(20ng/m3)
PM^aOug/m3)
PM1M.5(10ng/m3)
PM10(20ug/m3)
PMjsClOug/m3)
PMuwj (10 Mg/m3) -
PM10(20ug/m3)
PM^OOug/m3)
PM10(20ng/m3)
PM^s (10 ng/m3)
PM10(20|ig/m3)
PMjj (10 ng/m3)
NS Changes
-6.56% (-9.64, -3.43) FVC
-2.15% (-3.34, -0.95) FVC
-2.80% (-4.97, -0.59) FVC
-19.9 (-37.8, -2.6) FVC
-25.6 (-47.1, -5.1) MMEF
-0.23 (-0.44, -0.01) FVC %
growth
-0.1 8 (-0.36, 0.0) FVC %
growth
-0.22 (-0.47, 0.02) FVC %
growth
-0.51 (-0.94, -0.08) MMEF %
growth
•i - 0.4 (-0.75, -0.04) MMEF %
growth
-0.54 (-1 .0, -0.06) MMEF %
growth
/
-0.1 2 (-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
NR(20,59)
4.7 (0.7, 7.4)
14.5(5.8,20.7)
23.8(15.4,32,7)
NR (28.0, 84.9)
NR (28.0, 84.9)
NR(15,70)X
NR(10, 35)x
NR
*
NR(15,70)X
NR(10,35)X
NR
NR(10,80)Y
NR(5,30)Y
NR(10,80)Y
NR(5,30)Y
NR(10,80)Y
NR(5,30)Y
January 2005
3B-3
Draft - Do Not Cite or Quote
-------
Study
12 Southern California
communities8
12 Southern California
communities*
12 Southern California
communities8
Indicator (Increment)
PM10 (20 ug/m3)
PM10(20ug/m3)
PM10 (20 Mg/m3)
Relative Risk (95% CI)
-3.6 (-18, 11) FVC growth
-33 (-64, -2.2) MMEF growth
-70 (-120, -20) PEFR growth
Study
Concentrations
(ue/m3)
NR (15.0, 66.2)
NR (15.0, 66.2)
NR (15.0, 66.2)
Lung Function Changes in Adults
AHSMOGT(%
predicted FEV,,
females)
AHSMOGT
(% predicted FEV,,
males)
AHSMOGT
(% predicted FEV,,
males whose parents
had asthma, bronchitis,
emphysema)
AHSMOG7
(% predicted FEV^
males)
PM,0 (cutoff of 54.2 days/year
>100 ug/m3)
PM,0 (cutoff of 54.2 days/year
>100ug/m3)
PM,0 (cutoff of 54.2 days/year
>1 00 ug/m3)
SO; (1.6 ug/m3)
+0.9% (-0.8, 2.5) FEV,
+0.3 % (-2.2, 2.8) FEV,
-7.2% (-11.5, -2.7) FEV,
-1.5%(-2.9,-0.1)FEV,
52.7(21.3,80.6)
54.1(20.0,80.6)
54.1(20.0,80.6)
7.3 (2.0, 10.1)
References:
ADockeryetal.(1993)
BEPA(1996a)
c Pope etal. (1995)
DKrewskietal. (2000)
E Pope etal. (2002)
F Abbey etal. (1999)
°Lipfertetal. (2000b)
* McDonnell et al. (2000)
'Dockery etal. (1989)
'Dockery etal. (1996)
K Abbey etal. (1995a,b,c)
L Peters etal. (1999a)
"McConnelletal. (1999)
NBerglundetal. (1999)
°Raizenne etal. (1996)
pPeters etal. (1999)
Q Gauderman et al. (2000)
R Gauderman et al. (2002)
sAvoletal. (2001)
T Abbey etal. (1998)
Note: Study concentrations are presented as mean (min, max), or mean (±SD); NS Changes = No significant changes
(no quantitative results reported); NR=not reported.
"Median
v Results only for smoking category subgroups.
x Estimated from Figure 1, Gauderman et al. (2000) ,
Y Estimated from figures available in online data supplement to Gauderman et al. (2002)
January 2005
3B-4
Draft - Do Not Cite or Quote
-------
APPENDIX 4A
Study-Specific Information on Short- and Long-term Exposure
Studies in Cities included in PM2 5 Assessment and on Short-term
Exposure Studies in Cities included in PM10.2.5 Assessment
January 2005
4A-i
Draft - Do Not Quote or Cite
-------
I
!
§r
^1
Ii
1
8
p
o
8
o
s
I
!
i o
! D>
'. <*>.
=
S.
o
I
9
IS
I
I
> CO
i -5
°
If
P •*'
8
5
i?
s
s-
•o
o
is
£
-------
II
l
«>. it
CM o
o"
-CO
E
3 O
••9
f
Ill
•fe
jfl
£ S
•o
i
Dl
ISA
fliiill
11
,
<
I
P
£
I
.e
sr
s
o'
9
N.
I
o
c
» 1
ft:
o
,
o
Q_
09
<1J
IT
•5:=:
I
2
>,
ra
•?
s-
T3
i
Q_
sr- C
CO c
5? (D
H £
•^ Q.
_£$
g
d
nj
>s
ra
73
CO
ID
00
6i< £
o n w
S
a
a.
o
o
•5=:
f ^3
'« §
>\ j\|
"
9
8
d
>,
f
&
-D
CM
i I
II
!
ra '*-'
c —
ro CD
loJ
•"*• C
Cv c
il
O
TT
-
l'
a
O
N
-s
8
6
,
CD
•o
CN
<:
ra
o
ca •-"
c
(0
E
10 "—
C _:
ra ra
i
«•!
-------
I
I
,
a.
n
i
0.
ill
I
is
o '-
•3 ff
§> ^
s= I?
<
C3 r.
8
d
s
§
9
S
O
•J
t
i
is
o *-
I S
61 £
£ &
s" s
"^ o
IPS:
S
s-
T3
O
1
S
S
d
II
"^ O
I i
8
d
ii *-
"I?
t
5 o
-------
a
Is
3 g
i
o.
f
° s
f
1'
S
1
o
0.-.
o
o
s-
•o
o
P.
8
o
3
s
s
o
g8
S
S.
j_ O
II-
tr
9
8
1
8
il
11=
8
9
S
9
S
§
9
s-
8
(3 .
i
o
•|
-o
t
o
o
S Q
< s
o ^
S
S
if!
o
b
8
6
9
5
o
?
W
f
I
8
t
18
5-
S
1
o o
I?
iti
00
"^ a
s-s
1
a
-------
I
2
:
I,
S 6
t
3
S
§
o n
s
S
s
o
t
8.
ill
s
6
s-
5
s
s
o
s
d
t
a*
is"?
s
8
o
t
!$
ll
if
t
58
-S
d
s-
XI
o
8
f!
s
6
fi
II
o
b
s-
•o
o
S
g
m o
^j*- •*—
^ o
9
K
18
o ^l
ii
if-
8
o
I
I
8
o
I
t
,
18
s
§
I
>.
I
O
I*
II
-------
I
s-
•o .
_
tj
i
o
0.
i
I
I
55
I
*
;8
d
Ife
L;
I-B
8
a
IT
*
•5
a !2
* 2.—
tff
CN
d
IS
9
R
§
f
8
8
9
8
b
•*-" "p"
•H) Q)
= £
ii
at
s
i
S
III
-------
<
a.
_2
a.
a.
cu
x
UJ
E
I
*j
o
V)
1
1
i
ig
5 u
II
c ,s
«SE
UI
ra
I
1
(0
s
S
TJ
3
(0
Q.
0)
X
UJ
o
V)
£
o
•a
I
in*
I
n
1}
i
0,
a;
X •
ui
ui
•S
o
8
9
o>
3
(0
TO
O
is i
5 2.
0)
ID
O
10
c
10
8
3
o
6
6> o
i-8
O O
01
c c
ra a
c c
a) ffl
•O -n
88
11
22
"8 "5
1
w
I
o
r
sis
8"
in
iS
o
§
R
E
fc
8
9"
1
ra
•D
O
M
-------
O
II
o
od
p
d
o>
ra
S
0)
C
o
I
S
I
82
o ~—'
CM >.
— 4,
a-s
1.2
n
in
8
9
D>
n
s-
•o
S
C
o
ra ^
u <£,
= s|
§"0 2
o
S
n
82
o ^—
n ^-
U. 0
2
w
q
d
0)
C
o
il
V
a:
ra
re
(2
ley
(O
g
8
d
8
9
8
o>
ra
*
•o
o
S
CM
O
C
o
linear,
M
ngent
ill
i
ra
O
in
'9!
.a-
a
c
(D
ey (2003)
irley(19£
3
8
d
8
I
p
d
D)
ra
ra
•o
o
in
o
L-- £
IS"
Jo
§
10
'55
m
2S
« o>
8^
CM >.
|i
£ u.
.J-s
to
S
8
d
p
d
m
o
in
o
8
in
"35
irley (2003)
Failley (19
o
p
d
p
d
O)
>>
-S-
o
in
o
(O
O
8
1
o>
•D
8
m
(0
'55
"ra
c
m
8?
s —•
CJ-ST
tl
II
U
S
u
111
£
in
csi
o
c
ls£l
5
;>.
tfi O)
111
Q.
-------
5
n.
a.
3 I
i'
!!.
O c =
s
O
i
8
d
I
I
(M
II
01
6
• s c
6.1 £
°<3J2
g
05
V
1
i
ill
o
3
d
f
ill
8
1
o
?• >— 1
£ CO ^
^ 8.-
•2
7
CM
1.1
Q
Q.
8
q
6
g
d
a-
•o
o
en
d
.1
I!
a '»
£• ™ ^™
_ s ff
3
d
a
I
TJ
O
c
'•j
O J2
I
a.
11
A M
E "5 ;
II!
o
q
6
s-
73
O
S •£
c «
.£ ra
Q
Q.
8
5T «
315
|Il
III
6
I
O
O:
8
n-5
||
Is
s-s
= •§
ll§
HI
6
s
I
I
o
I
a
I
i
n
Q
5
o>
q
d
I
I
s-
•o
i
u
I
f
1
•5.
s
6
§
9
I
r
•o
C'
o
i
i (D
3
^
o
o
s
q
o
•v
1
T
*-
tv
A
1
%
J
§
?
(0
(O
I
-------
(A
£
•o
10
cv
I
IA
O
Q.
fi
O)
O
.Q
n
in
S
•f
Q!"
t^-
^~
^
o
^J-
CD
C
E
I
o
E
re
0
t
g"
O T3
CM £
"— -o
__.* ^
(0
00
CO
S
d
CD
C-.
§
d
annual
mean
1
00
CO
o
T~
O
O
(0
(U
c
1
^>
5
CO
_
0*
$
re
u
5
o"
o
§.
"re
K
'x.
1 w
£ o
* <
s
T-
o
d
JQ
S
o
d
CM
00
o
o
d
H ^
2 co
c «
co E
JO
c
CO
CO
0
1
^_
re
CD
c
1)
G
A
CO
^
re
(U
(0
^
8
<
4
o"
CM,
"re
0)
Jx:
S w
*U
o
o
d
O)
CO
CO
S
d
CO
§
d
1 c
5 re
c £
re E
re
"c
00
co
o
T—
' CO
0
k_
(0
CO
C
^
O
******"
o
CO
—
re
CD
o>
3
re
u
_ ?
g
o
£L
"re
V
!x
is
* <
en
o
d
en
0
CM
S
°
^
CN
T-
S
d
1 c
2 co
El
$
CO
CO
o
CM
O
CO
CO
CD
c
•1
_o
o
CO
^
OJ
§
CO
o
5
o*
o
8
"re
oi
^
5 w
2 o
* <
fe
I
Itvl
o
o
5
-------
10
s
a.
£
c
o
IS
o
E
u
a>
a.
(0
<0
s
II
3 CO
s
ii
UJ
1
CD
g
O
o
o
CO
CO
o
g
9
CO
CO
t
S
E.
Q>
CO CO
§ O
S.g
CO?
Ill
IO
to
o
o
o
g
to
CN
CO
o
o
O)
t
co
s
log-linear,
GAM
(stringent
m
to
o
0.
CO
-*-l
V
CO C
III
CO
o
o
in
t*-
i
o
o
d
in
en
oo
o
o
•a
CN
O
IO
u co
•= CO
t 40
co =g
"5 ti
U) CO
—
i
I
o
o.
in
V
c
o
c
ll
.s
"S co
So
x-s (TJ
si,
S;
o i
in
o
o
CM
CO
§
9'
CN
O
o
fr.
2
8?
T3
O
O
in
co
c
o
c
~ S
A
CD
CO
CN
> £
11
c ro
O CO
0 £
(A
'S^
"co to
s-g
o s
§
Ol
a
-------
:I
m
o &
E o
£ o
8-
S-:
Ul
u>
I
co"
n
o>
(O
_c
n
.S
•«rf
to
E
§
0
i
|
CO
u
E
o
0)
a.
to
i
S
o
o
o
§
o
o
O)
I
00
03
0)
c
log-li
GAM
(stri
co
O)
Asthm
i
Q.
__^ CD
*"*"(§
i^
ISS
Q. TO -—
«" TO "CB
£ (D j_,
OT ^-a>
Sh
£.•5
p?0
§,£?
1?
S.«
SITI
2 CM
to
|
w
co"
0. B
0. 3
O. O
3 m
o °
-J CO
S
u
X
111
Ol
|
£ i
0 §
o
li
u
T3
i
55
p
o
S
in
S
p
9
in
oo
eo
CD
q
o
S
o
CN
Is
£?
S
00
8
CO
p
9
CO
§
O)
g
s-
t
CM
in
CN
CO
"o>
I
tl
1
S?
N CO
CD t
OT CN
in
CM
(0
'Si
1
T3
(D 'o
£J CD
CO CM
I
Crt
0
I
13
I
I
'&
CN
-------
APPENDIX 4B
Sensitivity Analyses: Estimated PM-Related Incidence Associated
with Short- and Long-term Exposure to PM2 5 and Short-term
Exposure to PM,0_2 5
January 2005 4B-i Draft - Do Not Quote or Cite
-------
f:l
& O
IP
|Ii
s
»
t,
I'
£
1
13
•a,?
S
il?
iss;
;? i
R
;*i
§
§ §
o o
S «i
ll
II
I-
1!
as
II
tl
il «
lit
Ul M
1*
;?8
"-IS
!
s
-S
Standards
it
i
o
a
-------
8
i
I
g
it
ai.
•6 S
I!
.f
!1! I
II I
-.3 &
I;
««
tt <
f!!
-------
QJ
Q
fe
aj (0
•S «•
.s m
S^
S 5
<•£
10
C 10
S
ui -o
.2 c
o S;
>S
A 2
>
5 u>
gg
JlZ
«^s
I
., 10
la
^Sfi
(0
£
(0
U
ii
x °
Is
n w
E e
a) is
*^i o
t
S
2
a
£
5
1
re
ce Associ
§
2
u
c
1
jj
a
a.
00
le value ~~j
1
(U
Q.
1
o"
in
CN
^4
~—
sf
"~"
o"
to ^
00
3
3
ja>
1
&
i
o
in"
f*3
in
O
O '
o
v~ '
co.
m
9 rl
in
le value
§
1
O)
o-
co
10
0
O i
o
o"
O '
o.
So
°T
NS
le value
1
1
in"
CM
in
S
^4
• — •
CM ^
~~"
o"
in rl
00
ile value
•E
1
i
CD
in"
CD
in
£
^4
— *
0>
""""
o"
a?
in ,-
00
ile value
1
1
i
CO
o"
in
o"
CO
in"
CD ^
^ CM
o
~~"
DO
10 f2
0)
75
.S>
I
"c
S
a5
a.
1
in"
co
^r
Of
te
m
I
I
-------
0
b
1
2
Q.
H
s
l»
<
s
e
«
^«
•B
1
1
CL
to
^n
I
fj
1
(0
1
I
<
v
73
n
il
1~°
l«
o
|
•o
0
in
e!
*g
» 2
91 TT
€ "
o
Q.
X
4mt
S £
UJ % ^"1
,
S
"Bi
3
S
D
8
&
s
0)
o"
CO
•<>•
o
O i
o
o1
O i
o.
sr
^ in
fc°?
NR
le value
•E
?
o>
a.
CO
in"
CM
*a-
o
0 i
0
s
£ '.
CO
CO
in
*~
•rt- CO
CO CO
CM
S
to
0)
1
8
8.
of •
O)
o"
•^
CO
0
O i
0
M-
CD
^ *~
CO I
CO
CO
in"
a-?
"8
o>
3
§
0
•*j-
CO
0
0 i
o
S
& ',
co
CO
in
S5?
°S
le value
1
o>
a.
CO
ui"
CO
CO
o
O i
0
•*r
CD
£ '•
CO
CO
in"
TJ- CD
CO co
CM
0)
15
.a>
-s
P
CO
a.
£ •
o"
CO
CO
0
O '
0
o"
0 i
0
sr
fi»
"B
le value
•E
8
0)
Q.
.C
(S>
o>
in"
CM
CO
o
0 i
o
—
O i
0
sr
-1?
"si
e value
<=
1
8.
1
o" •
T
CM '
o
1
t
m
I
i
-------
$
U)
^
a
£
i
1
ce Asso
i
"5
c
1
i
a
a.
Standards
>
c
(Q
£
<
•o
0
Jm
vt
1 "
1- •§
13 A
O -.
v TT
£ "
Q.
X
s
o
£
II)
fi «
it
I "
gj
X
« c
g —
Q (U M
7^ S ^ ,E
111 9 D>
il~2
i" a "
CO 0)
sr
•~
3,
'ra
a
t
1
15
a
c
e
O
O i
2.
O i
2
^
O)
^- ^-
™$
•3
ra
o
•B
1
S.
1
in
CO
CM
o
0 i
0
0 i
o_
ifm^
en
T~ ^r
"i
0>
3
§
.2
a)
Q.
(-j-
00
CM
0
O i
o
0 I
.•—•.
00
O -*J-
^s
0)
"re
_0)
5
Q.
i
0)
LO'
CM
CM
O
0 i
O_
O t
^_^
O)
"*" "^
Ra
(U
75
J!J
"E
I
cl
0)
o"
CM
0
O i
Q
O 1
^ ^
o>
T™ ^
^S
O)
>
.2
^
8
d>
a.
O5
m"
CO
CM
0
O i
o.
T—
O i
0
^— ^
O)
T— ^J-
^g
o"
CO
CM
O
0 i
0
o"
0 •
/-•-.
00
t; fO
04 CM
ID
•i
1
1
.c
cr>
in"
CM
CM
•D
£
"c
i
'S
f
S.
o
•lofthesli
S
a.
™
£
jj
8!
«
1
i
x^
"5
tg
I
•g
o
u
i
M
1
j
•D O
I
o
1
in
03
-------
a
01
13
If
I- m
1 O
£
c
il
ra S!
•£
.£>:
m
o
S
*
2||
I ||l
|2 uj x 3
*l
•a a
£
•a
tfi»
Study
,f
8
8
s
r.
in
S0?
8
-------
I-
a
til
I
s
HI
I
to ^
!
e 'i
£
li
« 8
1!
-------
I
Jra
01
V II
s?
S
-------
O
fe
«J
t
|
*8
«
g
S
s
1
D.
Ill
i
£
i
I
!H
£ N
£ ^r
in
£ "i 9!
a. e
£H IT
ji'I
O
i
>
s
f
1
1
s
O i
S
O i
a
s
O i
a
intile value
A
a
si
i
m
*
o
o >
a
s
O i
a
s
O i
a
intile value
P
a
£
§
S
,
s
o >
jjT
Pm *"
N !(!
s.
§
n J:
5 <
'i
ntile value
«
S
i
§
fO
S
O i
a
.
pi. 1
N 1ft
ffi
n ^
O 1
*• «o
ntile value
8
S
8
in
n
S
o •
P,
p
18 T
£i
i
ntile value
8
8
i
n
<•>
S
o •
P.
S
O i
a
is
U>
S
o <
P.
S
O i
a
0
o •
e
mtile value
P
a
£
s
s
D
s
O i
P.
S
o •
a
o
S;
5
I
8
a
i
o
CM
S
o >
P,
S
O '
a
0
"§
ntile value
8
a
i
Ul
n
a
o
O i
3
intile value
?
a
i
o
CM
O
O •
O
O i
e
o
O i
e
mtile value
P
a
£
s
in
CM
fr£
¥ f:
-------
II
i=O
II
>
I E
'S «) «.,
.
o s .a
"3 -o -g
ilt
i!
.3
8
a
i*i
•s .
a
s
s
,ST
©
ss
Ei
§t
Ei
«s
&'
CJ
a
s
rRT
Ma
:,S
2 . !
'fiR
o
5
CVI ,
5
1
I
.2
8.
8
ff
1
S
a
fi
S
1
a>
I
1
S
§
£
s.
s
tf
-------
§
s
0
1
00 t*l <» (N
I
ical
a
£
SS
Ci
S
d
a
3?
! TI
to
Ci
S
"
°T
S
ci
Is
JS
II
s
!S«
s
I
£
I
5
I
i
s"
u
I
s
?
tf
xf
I!
•I
I E
i!
II
1
-------
§0)
III c
£. °3»
»2
§1
_l «
1 U
•a «i
S ia
•5 *
ro oj
v
«£
<£
£tf
i s
il
11
SI
!3*w
ill
g W 0
iS S s
^S 2
«! 5
p*
lit
** S«
fs'gg
S
[O (o
J
ss
1
n
>
i
a.
01
S
in
s
O i
a
5?
"S.
s
^
5
J
1
|
0)
§
in
S
O i
S
U)
N
ST
«
• ?
"I
1
i
Q.
n
!!?
in
S
O i
S
S
o >
8
S *?
^s
0)
g—
0>
&
£
§
3
u>
s
O i
s
s
o •
o
§
ST
§
2~~
i •
|
S
a
m
0
0 >
a
ftf
"•l
UJ •
5
1
s
a
c
I
01
iS
in
S
O •
S
s-
0 i
a
5
ST
§
!
i
Q.
S
§
M)
8
o •
a
s
o •
a
o
o •
9
•
o>
^
^
ffi
0>
s
in
S
O '
a
s
O i
a
s
O i
a
!
i
m
a>
S
in
S
o <
a
s
O i
S
S
0 i
a
!
i
a.
a>
S
u>
E
_
|
S
s
*
s
O '
a
s
O '
a
f
ST
O)
a
a>
J
a.
f
00
O)
in
«
^
S
o <
a
s
O I
e
5
ST
§
D
1
^
a.
i
s>
S
•»
-
-------
Of
*1
£
•o
JS
8
S
«
u
di
I
|
1
n
n
»
>
1
S3 <3
•s*l
£ M a
Ti
Hypothetic
Threshold
o10 M9'm3
ii
BASE CASE: Lov
Measured Level
Sbdy
=7.5 ug/ml
%
2?
I
|
i
1
S
O i
a
s
a >
8
S
0 <
a
1
M
C
1
S
B
T
&
O •
a.
s
0 >
a
s
o •
a
1
n
>
c
a
&
%
*
s-
o •
a
s
0 i
a
s
o •
S
|
1
8.
&
U)
C4
V
S
0 <
a
6*
^s
»&
fi|
S
5
_£
^S
<1>
&
i
§
CO
s
O '
a
8
PT
S
o> «
!
1
s
s
CO
s
o >
a
s
O i
S
5
a
8"
0 i
8
o
ST
o
a
J
or
>
s
c
k
&
s
n
S
0 i
a
o
o •
e
s
o <
a
J
5
c
Q.
S
s
CO
s
O i
a
0
O '
S
S
o •
o
d>
*
o.
§!
S
CO
S"
o •
a
c?
O i
a
s
0 •
3
1
C
1
£
in
CM
CO
S
o •
a
to
£T
CO
d
s
8*
CMg
0)
>
J
Q.
£
§
§
CM
S
(3 •
a
s
i-. »-
(O i
S
to
a^
•£
O
Q.
£
§
S
CM
S
O i
a
s
O 1
o
§
ST
§
!
m
09
A
£
S
f*
S
o <
o
0
O i
a
s
o •
a
J
CD
>
1
e
1
&
8
CM
O
0 •
O
O
O i
a
0
O i
8
0)
^
5
5
c
0.
£
tn
(N
CM
0> S 'Q
5 0.-
f*i
S a£
s*«
*f S
Is*
* *s
s i
n. *
B S
**§
li
€
8
|
*fl
-------
1
o
1
5*
Si
2
CO
1
CN
i?
Ml .£
£3
*1
CO «
SO
•n »
* 2
IS
8£
»*
< =
t!
i =
II
Sin
E
^5
in
K.
0.
I
.I
Si
S
2
CN
1
I
!i
I!
as v> g
= sl
OJ ^ 1™
wQ-P
. a *
I*
£ffl
§
CM
2
£~
S1
2
W
•i
I
I
I
1
8.
>
I
2
I
S
s
S.
£
|
S
$
6
S
f
£
I
I
§
To
-------
II
Sgtf
l!
-8
551
Si
a
1
S
1
Is
Ill
il*
a
Ill
f is
o
fc
o
I
a
-------
. fe
a
1
£
<5
•g
•
i!
o
g
• "
.2 m
8 «
ll
II
t
JC
s
I I (
s
s
o
i i
o
o
I 1
o
8
!T«
5
9
o
I I
S
,?!
S'
i
j
ii
a
i
S
a
ST
ST
Cl
> ^ «
l§li
w p i
£e!
5 2>
5
I
J
Ji
o
2
g
«
x = ±;
iu < n.
-------
II
O
a
£
S'- C
I U
s
8
T «
5
o
i C
o
i'
So
s
o
I i C
a
o
) i C
a
if!
;s •»
R
1?«
'o '
»>
f i
111
* S-ri
£
I
S3 8
ST '
°1 |
si*!
C9| ji
fill
-------
W X >. *J
h- 111 X (0
ed
2
II
- e
II
si
o> T
CO
8
S
m
0.
«
•f
in
in
8?
ID •
CM
1
0)
¥
R
S"
3.
in
<»
f
S
5"
in
m
d
O)
'S
in
$
m
O '
CO
R?
00?
in
GO
00
i i
Ci
to
I
d
in
R1?
in
in
,-?
CM
S
1C 7
J
S.
-------
8
II
CO
CM i
CO
O i
CN
CN
CM-
CM"
CM
O
|
a
&>
c4
c
w
II
O) ,
CO
to
in
s-
CO
ro
v
en
CO '
co
sa
c.
G.
"
G.
co-
in >
G
ffl T
Gi
CO i
a.
a?
t.
a?
00?
•—
s
8
'£
CM
^
HZ
§
jj
I
o>
Q.
3
S
I
a.
i
i
4
CO
R
CO
IR
8
(N
CM
CM
(N
I
-------
a
f!
o ^
j g
11
Si
11
£2
a
iy
.B
w
l!i
lil
w ? P
« So 3
j a a. J
a x >> «J
I- ui X M
ii
! £
11
£3.
0.7
s
*i
R?
s
55
3
a i
S
t 00
'is
) CO
.te.
§
s
s
-------
a
*:
Ss
Is
£3
t
87
S
§
*
,_ r-
s.
ST
S
in
a> 'V
G
S
s"?
s
.S
tf
.0
p .
i«S i
Hj 1
I 12 3
!!! !
»'!.!
s I
»«!•
!!!
it
if
' i
f1
S^
•I
111
-------
i
£
.2 «„
,>
w ^
"« S
«<
If
Q (A
•E 3
II
O V
13
oi O
ll
£ Q>
w £
!--
|i
< 3
ra
i E
1
^ w
JJ i. I
K>£ j« 7.
£?!
*• o o — -
=20
thetical
shold
Ill
I|s
o S o
O.£ S
£t- II
&'
=5 S.
*-.$£
w
o I ^
111 5 «
a
in
ID
S
ID
o> o>
in oi
8
S
s
8*?
cS
8?
s.
I
i-
'ra
•o
co
2
i
3 daily
S
8
ID
SJ
n
1
S.
I
i-
n
T3
co
percentile
N
o> °>
8? o>
pe
§
0?
10 a)
a>
a.
i
.£•
'
-------
_
|
w
JC
4>^
i
i
.2 ..
o d)
8"i
»
u 2
< E
fi?S
co 3-
*^ ^
v o>
S. c
<5
ro TJ
l§
** ni
« Si
S^
S«
W (M
c J2
•S?
o 2
.2 £
E c
35 ,
15 S 2
fi *- TS
O. in >
83 >
~> =5
x g S
7n «ff Bf
IS ^L i£
3 10 ?
il §
< «J s
•o ? &•
ill — .^
15 i
c W "»
-3 g 9
rt .£ 5
UJ IS «
*" W 3
• 1. 2
§>l^ -
« ^ S S
5 S o s
< £ S $
>, > _ 5
f ^ ^ •*
'> «q "o a
4s CM £ g
S?805
0) ^ t O «
Table4B-12. S
Exposure to Pl\
Hypothetical Tf
Seattle, WA, 20
(As Is = 11.4 ug/m:
I
&>
2
i
•n
1
8
1
•g
c
1
1
E
S n
T
o
S
CM N
O
5?
N. <°
CM i
O
g
J i
I !
f
5>
«
<
8 j
". <
N
!;
? J
•t
CM
C
^
ID ^
g
JR
CM N
O
5?
&• .
TO
D 1
n (
1
gu
(j
r ^
CM
> e
in *7
S
co
CM
a)
i
s
5
i s
n (
X
S"
i
O (
1 •
rj t
a u
r -5i
CM
> C
£
0)
1
B
S.
§ ;
n i
>.
IB
D T
•5 C
1 -
n '
r c
^~
1 C
c
•* T
g
1)
1
:
S I
>,
i 7
3 1
O C
1
R f
> «-
o •
1 C
in
CM i
9
O
ai
3
5
2 ;
E
u
2
0)
a.
i !
^
n i
? '
I ;
5 i
^
CM i
C
S
in "7
o
CO
CM "
^ i
o
r ^
CM
C
if T
o
O3
CM CM
*" O
in
Kf
o
ij
2
>
u
1 •
s
8.
1 J
J. ^
1 '
T ^
= C
1 -
— <
2»
CM1?
O
n
3
5
z*
j
:
1;
o t
1
3 c
i> T
o
oo
CM CM
O
in
CM?
0
Ol
S
>
a>
1 r
5
01
x
c j
1 i
X
s r
n
a t
o f
1
R S
CSI '
o
s
=
n
>
III
1 ;
C J
i I
X
= r
in
a i
T t
I 1
3 £
r c
O '
, S
^
CM '
0
<0 7
o
a?
o
-n 5
1
Gf
2 ^1
o n
1 *
1 *?
j t
i «
3
i
3
e
S
1
c
e
&
S.
|
.a w
J 1
JQ .3=
•g S &
J f c 8 "
I S| N
I i i 2
i ** z
tils S
fill *
"K tj « §
« | 5 S
* 3 " i
•s s o j
(- O *3 .(5"
1 g I 1
s If 1
co (S u a
a S e 1
5 § 2 f
e 8 « u
w 5 w *l
S u fl> .12
•2 w g J
U N s 8
i el %
$ s - ~8
(0 Q. r 2
— •sis
sf 2 5 -°
S S» J9
s f I I
1 if 1
i s o s
Ol
3
1
(U
•j
u
^,
|
>>
S
3
1
*
fijs I
Sni ^
S '1 I
1 f*J
fill
1 Ii
§ s! $
"In the base case, incidei
models, incidence was qu;
Note: Numbers in parenth
January 2005
-------
>*i S
w <
&•§
fl
I
£
v
§1
si
8
So
85
85
85
So
s-
i!
"8
si
§1
85
a
SI
85
8*7
CJ
•I
i
5 .
a
1
-------
(0
x
10 -8 «»
111
*1 J
ill
Hi
UJ O M
§1°
£ 'fl 6>
•I c 2
|||
S '5 E
£sa
-
!S 2 o)
^1 8
L; * °
D>
2s? 2
^
o
<
»
•
ii
o >
i^1
1|
i!
4 R
Hypothetical
Threshold
=15 ug/mS
Threshold
=10 ug/mS
BASE CASE:
Dwest Measured
Level in Study
=4
:i|
g a
Thres
=15 |j
M P
I"
s S
1 «*
c &
a CB
•i
^
*Ǥ
s
NO!
in <7
2.
§
a
<*>
CM 1
•4
8AS
r a?
S T ^
95
ff.
O) i O
"
1*
CO
"
X*?*
0 ' (0
TS
1
I
m
"
•S
•--•
00
"
S
S
P
B
O i r-
n
«M
o
f
S
in
*«*!
O ' ID
-go.
8"
Q
ID
a
t:
i
-------
II
II,
11
&
™ m S
f*
. *
,
'88
£*
I . O
n J. IJ
:M
ID . P
2.8
II
'88
*«
°E
r!JS?
9*
s -5
'aft
i *»
1 s^
111
li |i
MIS
i ii
» -• o i
•II*
Ms!
s * a
ll|
•3 *» « g «
ililU
tlnn
-------
si
«c
i!
II
r
i
11
II
lc
i
iif'
5 3 '
£ w •
ill
:•>-• T
I c
ui —' •
< 3
..«*
CM • 0
"i- e>
a"!
FT
§:
N. V- '
S T<
l~- ~ (
sr
i o>;
i i i
'5'
?*5>
s-
1 5 j
* - •
»
i ^ "
' T- CO
10 r
e ,
8
§"7«
»*
a-
Si
,8*
™
§"*
I ^ c»
's.Si
S"^
'a?
-------
K Ot
||
Ii
< g
|?
s!
if
ing Annual
of Monitor-
K
S
fe
o
1
1
li
s j?
IT
i* w v*
lli
|lf
1
IE
1
£
3
I
8
S
ii -,
,iit
ul f
t
3
i
<.
T-8?
°S§
B-^
fill
« i ^j
8,
sr
5?
en • ^
ffi-
;s§
°°
P.8
f*
S'
is
• 0
op
°°
a
c.
og
1
2
6
*
|
! !i
! SJ
* st-
!!
s?
III
ill
'I
?
S-S
ll
\\
h
si
l
f I
H
-------
l
— 3
I
00
I
-------
|