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January 2005                                 Draft - Do Not Quote or 'Cite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January 2005                                              Draft - Do Not Quote or Cite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

                                  ^

                               *«§

                               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

-------