'**
\
Risk and Exposure Assessment to Support the
Review of the NC>2 Primary National Ambient
Air Quality Standard
-------
EPA document # EPA-452/R-08-008a
November 2008
Risk and Exposure Assessment to Support the
Review of the NC>2 Primary National Ambient
Air Quality Standard
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina
-------
Disclaimer
This document has been prepared by staff from the Ambient Standards Group, Office of Air
Quality Planning and Standards, U.S. Environmental Protection Agency. Any opinions,
findings, conclusions, or recommendations are those of the authors and do not necessarily reflect
the views of the EPA. For questions concerning this document, please contact Dr. Stephen
Graham (919-541-4344; graham.stephen@epa.gov), Mr. Harvey Richmond (919-541-5271;
richmond.harvey@epa.gov), or Dr. Scott Jenkins (919-541-1167; jenkins.scott@epa.gov).
-------
Table of Contents
List of Tables v
List of Figures ix
List of Acronyms/Abbreviation xiii
1. INTRODUCTION 1
1.1 OVERVIEW 1
1.2 HISTORY 5
1.2.1 History of the NO 2 NAAQS. 5
1.2.2 Health Evidence from Previous Review 6
1.2.3 Assessment from Previous Review 7
1.3 SCOPE OF THE RISK AND EXPOSURE ASSESSMENT FOR THE CURRENT REVIEW 7
2. SOURCES, AMBIENT LEVELS, AND EXPOSURES 9
2.1 SOURCES OF NO2 9
2.2 AMBIENT LEVELS OF NO2 9
2.2.1 Background on NO2 monitoring network 9
2.2.2 Trends in ambient concentrations ofNO2 13
2.2.3 Uncertainty Associated with the Ambient NO2 Monitoring Method 14
2.3 EXPOSURE TO NO2 15
2.3.1 Overview 15
2.3.2 Uncertainty Associated with Ambient Levels as a Surrogate for Exposure 16
3. AT RISK POPULATIONS 17
3.1 OVERVIEW 17
3.2 SUSCEPTIBILITY: PRE-EXISTING DISEASE 18
3.3 SUSCEPTIBILITY: AGE 19
3.4 SUSCEPTIBILITY: GENETICS 19
3.5 SUSCEPTIBILITY: GENDER 20
3.6 VULNERABILITY: PROXIMITY (ON OR NEAR) TO ROAD WAYS 20
3.7 VULNERABILITY: SOCIOECONOMIC STATUS 21
3.8 CONCLUSIONS 21
4. HEALTH EFFECTS 23
4.1 INTRODUCTION 23
4.2ADVERSE RESPIRATORY EFFECTS FOLLOWING SHORT-TERM EXPOSURES 26
4.2.1 Overview 26
4.2.2 Respiratory Emergency Department Visits andHospitalizations 27
4.2.3 Respiratory Symptoms 28
4.2.4 Lung Host Defenses and Immunity 31
4.2.5 Airway Response 32
4.2.6 Airway Inflammation 35
4.2.7 Lung Function 36
4.2.8 Conclusions and Coherence of Evidence for Short-Term Respiratory Effects 36
4.3 OTHER ADVERSE EFFECTS FOLLOWING SHORT-TERM EXPOSURES 37
4.4 AD VERSE EFFECTS FOLLOWING LONG-TERM EXPOSURES 38
-------
4.4.1 Respiratory Morbidity 38
4.4.2 Mortality 40
4.4.3 Other Long-Term Effects 40
4.5 RELEVANCE OF SPECIFIC HEALTH EFFECTS TO THE NO2 RISK CHARACTERIZATION 41
4.5.1 Overview 41
4.5.2 Epidemiology 42
4.5.3 Controlled Human Exposure Studies 44
5. IDENTIFICATION OF POTENTIAL ALTERNATIVE STANDARDS FOR ANALYSIS 46
5.1 INTRODUCTION 46
5.2 INDICATOR 46
5.3 AVERAGING TIME 46
5.4 FORM 48
5.5 LEVEL 49
6. OVERVIEW OF APPROACHES TO ASSESSING EXPOSURES AND RISKS 55
6.1 INTRODUCTION 55
6.2 POTENTIAL HEALTH BENCHMARK LEVELS 57
6.3 SIMULATING THE CURRENT AND ALTERNATIVE STANDARDS 59
6.3.1 Adjustment of Ambient Air Quality 59
6.3.2 Adjustment of Potential Health Effect Benchmark Levels 64
1. AMBIENT AIR QUALITY ASSESSMENT AND HEALTH RISK CHARACTERIZATION 67
7.1 OVERVIEW 67
7.2 APPROACH 69
7.2.1 Air Quality Data Screen 70
7.2.2 Selection of Locations for Air Quality Analysis 71
7.2.3 Site Characteristics of Ambient NO2 Monitors 73
7.2.4 Estimation ofOn-Road Concentrations using Ambient Concentrations 80
7.2.5 Air Quality Concentration Metrics 83
7.3 AIR QUALITY AND HEALTH RISK CHARACTERIZATION RESULTS 85
7.3.1 Ambient Air Quality (As Is) 85
7.3.2 On-Road Concentrations Derived From Ambient Air Quality (As Is) 97
7.3.3 Ambient Air Quality Adjusted to Just Meet the Current and Alternative Standards 102
7.3.4 On-Road Concentrations Derived From Ambient Air Quality Adjusted to Just Meet the Current and
Alternative Standards 114
7.4 UNCERTAINTY ANALYSIS 122
7.4.1 Air Quality Database 124
7.4.2 Measurement Technique for Ambient NO2 125
7.4.3 Temporal Representation 125
7.4.4 Spatial Representation 727
7.4.5 Air Quality Adjustment Procedure 130
7.4.6 On-Road Concentration Simulation 133
7.4.7 Health Benchmark. 143
7.5 KEY OBSERVATIONS 144
8. EXPOSURE ASSESSMENT AND HEALTH RISK CHARACTERIZATION 146
8.1 OVERVIEW 146
8.2 OVERVIEW OF HUM AN EXPOSURE MODELING USING APEX 149
8.3 CHARACTERIZATION OF STUDY AREA 151
8.3.1 Study Area Selection 151
8.3.2 Study Area Description 152
8.3.3 Time Period of Analysis 152
8.3.4 Populations Analyzed 152
8.4 CHARACTERIZATION OF AMBIENT AIR QUALITY USING AERMOD 154
8.4.1 Overview 154
-------
8.4.2 General Model Inputs 154
8.4.2.1 Meteorological Inputs 154
8.4.2.2 Surface Characteristics and Land Use Analysis 155
8.4.2.4 Other AERMOD Input Specifications 155
8.4.3 Major Link On-Road Emission Estimates 156
8.4.3.1 Emission Sources and Locations 156
8.4.3.2 Emission Source Strength 158
8.4.3.3 Other Emission Parameters 160
8.4.4 Minor Link On-road Emission Estimates 161
8.4.5 Adjustment of On-road Mobile Source Strengths to 2002 NEI Vehicle Emissions 163
8.4.5 Stationary Sources Emissions Preparation 164
8.4.6 Airport Emissions Preparation 165
8.4.7 Receptor Locations 167
8.4.8 Modeled Air Quality Evaluation 168
8.4.8.1 Comparison of Hourly Cumulative Density Functions 168
8.4.8.2 Comparison of annual average diurnal concentration profiles 170
8.4.8.3 Comparison of estimated on-roadNO2 concentrations 173
8.4.8. Using unadjusted AERMOD predicted NO2 concentrations 177
1.5 SIMULATED POPULATION 177
;.6 CONSTRUCTION OF LONGITUDINAL ACTIVITY SEQUENCES 179
1.7 CALCULATING MICROENVIRONMENTAL CONCENTRATIONS 179
8.7.1 Microenvironments Modeled. 180
8.7.2 Microenvironment Descriptions 181
8.7.2.1 Microenvironment 1: Indoor-Residence 181
8.7.2.2 Microenvironments 2-7: All other indoor microenvironments 184
8.7.2.3 Microenvironments 8 and 9: Outdoor Microenvironments 184
8.7.2.4 Microenvironment 10: Outdoors-General 185
8.7.2.5 Microenvironments 11 and 12: In Vehicle- Cars and Trucks, and Mass Transit 185
;.8 EXPOSURE MEASURES AND HEALTH RISK CHARACTERIZATION 185
8.8.1 Adjustment for Just Meeting the Current and Alternative Standards 187
1.9 EXPOSURE MODELING AND HEALTH RISK CHARACTERIZATION RESULTS 189
8.9.1 Overview 189
8.9.2 Annual Average Exposure Concentrations (as is) 190
8.9.3 Daily Average Exposures (as is) 192
8.9.4 One-Hour Exposures 197
8.9.4.1 Overview 197
8.9.4.2 Estimated Number of 1 -hour Exposures Above Selected Levels (as is) 197
8.9.4.3 Estimated Number of 1 -hour Exposures Above Selected Levels (current standard) 207
8.9.4.4 Estimated Number of 1 -hour Exposures Above Selected Levels (alternative standards) 209
1.10 KEY OBSERVATIONS 213
i.ll REPRESENTATIVENESS OF EXPOSURE RESULTS 214
8.11.1 Introduction 214
8.11.2 Description of Data Compiled and Summarized 214
8.11.2.1 HAPEM6 Near-Road Population Data Base 215
8.11.2.2 American Housing Survey (AHS) Data 218
8.11.2.3 Federal Highway Administration (FHWA) Data 220
8.11.4 Discussion 221
1.12 UNCERTAINTY ANALYSIS 223
8.12.1 Dispersion Modeling Uncertainties 224
8.12.1.1 AERMOD Algorithms 224
8.12.1.2 Meteorological Inputs 226
8.12.1.3 Mobile Source Characterization 228
8.12.1.4 On-Road Emissions Estimates 230
8.12.1.5 O3 Monitoring Data for OLM and PVMRM Options 234
8.12.1.6 Use of Unadjusted AERMOD NO2 Concentrations 235
8.12.2 Exposure Modeling Uncertainties 237
8.12.2.1 Population Data Base 237
8.12.2.2 Commuting Data Base 238
8.12.2.2 Human Time-Location-Activity Pattern Data 239
111
-------
8.12.2.3 Longitudinal Profile 240
8.12.2.4 Meteorological Data 242
8.12.2.5 Air Exchange Rates (AER) 243
8.12.2.6 Air Conditioning Prevalence 246
8.12.2.7 Indoor Source Estimation 248
9. CHARACTERIZATION OF HEALTH RISKS USING DATA FROM EPIDEMIOLOGICAL
STUDIES 252
9.1 INTRODUCTION 252
9.2 GENERAL APPROACH 253
9.3 AIR QUALITY INFORMATION 258
9.4 CONCENTRATION-RESPONSE FUNCTIONS 259
9.5 BASELINE HEALTH EFFECTS INCIDENCE DATA 261
9.6 ADDRESSING UNCERTAINTY AND VARIABILITY 262
9.7 RISK ESTIMATES FOR EMERGENCY DEPARTMENT VISITS 269
10. EVIDENCE- AND EXPOSURE/RISK-BASED CONSIDERATIONS RELATED TO THE PRIMARY
NO2NAAQS 275
10.1 INTRODUCTION 275
10.2 GENERAL APPROACH 276
10.3 ADEQUACY OF THE CURRENT ANNUAL STANDARD 278
10.3.1 Evidence-based considerations 279
10.3.2 Exposure- and risk-based considerations 282
10.3.2.1 Key uncertainties 284
10.3.2.2 Assessment results 287
10.3.3 Conclusions regarding the adequacy of the current standard 290
10.4 POTENTIAL ALTERNATIVE STANDARDS 291
10.4.1 Indicator 291
10.4.2 Averaging Time 291
10.4.3 Form 296
10.4.4 Level 299
10.4.4.1 Evidence-based considerations 299
10.4.4.2 Exposure- and risk-based considerations 304
10.4.4.3 Conclusions regarding level 309
11. REFERENCES 310
APPENDICES
APPENDIX A - Supplement to the NO2 Air Quality Characterization
APPENDIX B - Supplement to the NO2 Exposure Assessment
APPENDIX C - Nitrogen Dioxide Health Risk Assessment for Atlanta, GA
IV
-------
List of Tables
Number Page
Table 2-1. NOx Network Distribution of Monitor Objectives 12
Table 2-2. NOx Network Distribution across Measurement Scales 13
Table 3-1. Overview of Susceptibility and Vulnerability Factors 18
Table 4-1. Weight of Evidence for Causal Determination 24
Table 4-2. Causality judgments made in the ISA for endpoints associated with short-term NO2
exposures 25
Table 4-3. Causality judgments made in the ISA for endpoints associated with long-term NO2
exposures 26
Table 4-4. Proposed Mechanisms Whereby NO2 and Respiratory Virus Infections May
Exacerbate Upper and Lower Airway Symptoms 32
Table 4-5. Fraction of nitrogen dioxide-exposed asthmatics with increased nonspecific 34
airway hyperresponsiveness 34
Table 7-1. Counts of complete and incomplete site-years of NO2 ambient monitoring data 71
Table 7-2. Locations selected for NO2 Air Quality Characterization, associated abbreviations,
and values of selection criteria 73
Table 7-3. Percent of ambient NO2 monitors with selected monitoring objectives, using all valid
site-years of historical air quality (1995-2000) 75
Table 7-4. Percent of ambient NO2 monitors with selected monitoring objectives, using all valid
site-years of recent air quality (2001-2006) 75
Table 7-5. Percent of ambient NO2 monitors with selected measurement scales, using all valid
site-years of historical air quality (1995-2000) 76
Table 7-6. Percent of ambient NO2 monitors with selected measurement scales, using all valid
site-years of recent air quality (2001-2006) 77
Table 7-7. Percent of ambient NO2 monitors with selected land use, using all valid site-years of
historical air quality (1995-2000) 77
Table 7-8. Percent of ambient NO2 monitors with selected land use, using all valid site-years of
recent air quality (2001-2006) 78
Table 7-9. Distance of ambient monitors to the nearest major sources in selected locations 79
Table 7-10. Derived Cv/Cb ratios (m) for two season groups used for adjusting ambient NO2
concentrations to simulate on-road NO2 concentrations 82
Table 7-11. Monitoring site-years and annual average NO2 concentrations, using recent air
quality data (as is) and monitors sited >100 m of a major road 86
Table 7-12. Monitoring site-years and annual average NO2 concentrations, using recent air
quality data (as is) and monitors sited >20 m and <100 m of a major road 87
Table 7-13. Monitoring site-years and annual average NO2 concentrations, using recent air
quality data (as is) and monitors sited <20 m of a major road 88
Table 7-14. Number of daily maximum exceedances of short-term (1-hour) potential health
effect benchmarks in a year, using 2001-2003 recent NO2 air quality (as is) and
monitors sited > 100 m of a major road 91
Table 7-15. Number of daily maximum exceedances of short-term (1-hour) potential health
effect benchmarks in a year, using 2004-2006 recent NO2 air quality (as is) and
monitors sited > 100 m of a major road 92
v
-------
Table 7-16. Number of daily maximum exceedances of short-term (1-hour) potential health
effect benchmarks in a year, using 2001-2003 recent NC>2 air quality (as is) and
monitors sited >20 m and <100 m of a major road 93
Table 7-17. Number of daily maximum exceedances of short-term (1-hour) potential health
effect benchmarks in a year, using 2004-2006 recent NC>2 air quality (as is) and
monitors sited >20 m and <100 m of a major road 94
Table 7-18. Number of daily maximum exceedances of short-term (1-hour) potential health
effect benchmarks in a year, using 2001-2003 recent NC>2 air quality (as is) and
monitors sited <20 m from a major road 95
Table 7-19. Number of daily maximum exceedances of short-term (1-hour) potential health
effect benchmarks in a year, using 2004-2006 recent NC>2 air quality (as is) and
monitors sited <20 m from a major road 96
Table 7-20. Estimated annual average NC>2 concentrations on-roads, using recent air quality data
(as is) and an on-road adjustment factor 99
Table 7-21. Estimated number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks on-roads in a year, using 2001-2003 recent NO2 air quality
(as is) and an on-road adjustment factor 100
Table 7-22. Estimated number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks on-roads in a year, using 2004-2006 recent NC>2 air quality
(as is) and an on-road adjustment factor 101
Table 7-23. Estimated annual mean NO2 concentration and the number of daily maximum
exceedances of short-term (1-hour) potential health effect benchmarks in a year, using
2001-2003 air quality adjusted to just meeting a 1-hour 100 ppb 98th percentile
alternative standard and monitors sited > 100 m of a major road 109
Table 7-24. Estimated annual mean NO2 concentration and the number of daily maximum
exceedances of short-term (1-hour) potential health effect benchmarks in a year, using
2001-2003 air quality adjusted to just meeting a 1-hour 100 ppb 98th percentile
alternative standard and monitors sited >20 m and <100 m from a major road 110
Table 7-25. Estimated annual mean NO2 concentration and the number of daily maximum
exceedances of short-term (1-hour) potential health effect benchmarks in a year, using
2001-2003 air quality adjusted to just meeting a 1-hour 100 ppb 98th percentile
alternative standard and monitors sited <20 m from a major road Ill
Table 7-26. Estimated mean number of daily maximum exceedances of 100 ppb 1-hour NC>2
concentrations in a year, using 2001-2003 air quality as is and that adjusted to just
meeting the current and alternative standards (98th percentile) for monitors sited >100
m, >20 m and <100 m, and <20 m of a major road 112
Table 7-27. Estimated mean number of daily maximum exceedances of 150 ppb 1-hour NC>2
concentrations in a year, using 2001-2003 air quality as is and air quality adjusted to
just meeting the current and alternative standards (98th percentile) for monitors sited
>100 m, >20 m and <100 m, and <20 m of a major road 113
Table 7-28. Estimated annual mean NC>2 concentration and the number of daily maximum
exceedances of short-term (1-hour) potential health effect benchmarks on-roads in a
year, using recent air quality (2001-2003) adjusted to just meeting a 1-hour 100 ppb
98th percentile alternative standard and an on-road adjustment factor 119
Table 7-29. Estimated mean number of daily maximum exceedances of 100 ppb 1-hour NC>2
concentrations on-roads in a year, using air quality as is and air quality adjusted to just
VI
-------
meeting the current and alternative standards (98th percentile) and an on-road
adjustment factor 120
Table 7-30. Estimated mean number of daily maximum exceedances of 150 ppb 1-hour NO2
concentrations on-roads in a year, using air quality as is and air quality adjusted to just
meeting the current and alternative standards (98th percentile) and an on-road
adjustment factor 121
Table 7-31. Summary of qualitative uncertainty analysis for the air quality and health risk
characterization 123
Table 7-32. Percent difference in 1-hour NC>2 concentrations for three modeled receptors in
Atlanta at different vertical heights, using AERMOD predicted 2002 air quality 129
Table 7-33. Number of 1-hour NC>2 concentrations above 100 ppb for three modeled receptors in
Atlanta at different vertical heights, using AERMOD predicted 2002 air quality 130
Table 7-34. Comparison of empirical distribution of on-road adjustment factors used in on-road
concentration estimation with a fitted lognormal distribution 140
Table 7-35. Absolute difference in the estimated number of exceedances of potential health
effect benchmarks on-roads using either a fitted lognormal distribution or empirical
distribution of the on-road adjustment factors and 2004-2006 air quality as is and air
quality adjusted to just meet the current annual standard 142
Table 8-1. Statistical summary of average annual daily traffic (AADT) volumes (one direction)
for Atlanta AERMOD simulations 157
Table 8-2. Average heavy duty vehicle (HDV) fraction for Atlanta AERMOD simulations. .. 158
Table 8-3. Average calculated speed by link type in Atlanta modeling domain 160
Table 8-4. On-road area source sizes 160
Table 8-5. On-road emissions from major and minor links in Atlanta, 2002 162
Table 8-6. On-road vehicle emission strengths by county for Atlanta modeling domain: modeled
vsNEI2002 163
Table 8-7. Summary statistics of estimated on-road hourly NO2 concentrations (ppb) and the
numbers of hourly concentrations above 100, 150, and 200 ppb in a year using both
the AERMOD and the on-road ambient monitor simulation approaches in Atlanta. 176
Table 8-8. Asthma prevalence rates by age and gender used for Atlanta 178
Table 8-9. List of microenvironments modeled and calculation methods used 181
Table 8-10. Geometric means (GM) and standard deviations (GSD) for air exchange rates by
A/C type and temperature range used for Atlanta exposure assessment 182
Table 8-11. Data used to estimate removal rate constant for indoor microenvironments 183
Table 8-12. Probability of gas stove cooking by hour of the day 183
Table 8-13. Adjusted potential health effect benchmark levels used by APEX to simulate just
meeting the current standard and various alternative standards considered 187
Table 8-14. Percent of population within selected distances of a major road in several locations.
217
Table 8-15. Residential A/C prevalence and roadway distance statistics for housing units in
several locations (AHS, 2008) 219
Table 8-16. Population and roadway statistics for several locations (FHWA, 2002) 220
Table 8-17. Summary of qualitative uncertainty analysis for the exposure assessment 223
Table 8-18. National vehicle miles traveled by roadway category and vehicle type 232
Table 8-19. Observed peak hour truck percentages on Interstate 75 (1-75) using 2002 traffic
count data 232
vn
-------
Table 8-20. Comparison of exposure results using a 0.55 versus 0.97 A/C prevelance for 2002
air quality without indoor sources 248
Table 8-21. Comparison of exposure results using a uniform versus lognormal NO2 indoor
decay distribution for 2002 air quality without indoor sources 249
Table 9-1. Characterization of Key Uncertainties in the Emergency Department Visits Health
Risk Assessment for the Atlanta Region 264
Table 9-2. Estimated Percent of Total Annual Incidence of Respiratory ED Visits Associated
with "As Is" NO2 Concentrations and NO2 Concentrations that Just Meet Alternative
Standards in Atlanta, GA, Based on Adjusting 2005 NO2 Concentrations.* 270
Table 9-3. Estimated Percent of Total Annual Incidence of Respiratory ED Visits Associated
with "As Is" NO2 Concentrations and NO2 Concentrations that Just Meet Alternative
Standards in Atlanta, GA, Based on Adjusting 2006 NO2 Concentrations.* 271
Table 9-4. Estimated Percent of Total Annual Incidence of Respiratory ED Visits Associated
with "As Is" NO2 Concentrations and NO2 Concentrations that Just Meet Alternative
Standards in Atlanta, GA, Based on Adjusting 2007 NO2 Concentrations.* 272
Table 10-1. Ratios of short-term to annual average NO2 concentrations 293
Table 10-2. Ratios of 1-h daily maximum NO2 concentrations to 24-h average concentrations
(ppm) 295
Table 10-3. Mean annual NO2 concentrations for 2004-2006 given just meeting alternative 1-h
standards (98th percentile) 296
Table 10-4. NO2 concentrations (ppm) corresponding to 2nd-9th daily maximum and 98th/99th
percentile forms (2004-2006) 298
Table 10-5. Mean number of days per year (averaged over the 2004-2006 time period) estimated
to have ambient (central site monitor) 1-h daily maximum NO2 concentrations > 0.10
ppm assuming 98th and 99th percentile forms of a 0.20 ppm standard 299
Table 10-6. Percent of counties that may be above the level of the standard, given different levels
(based on years 2004-2006) 305
vin
-------
List of Figures
Number Page
Figure 1-1. Overview of the analyses described in this document and their interconnections 4
Figure 5-1. NO2 effect estimates (95% CI) for ED visits/HA and associated 1-h daily maximum
NO2 levels (98th and 99th percentile values in boxes) 52
Figure 5-2. NO2 effect estimates for respiratory symptoms and associated 1-h daily maximum
NO2 levels (98th and 99th percentile values in boxes) 53
Figure 6-1. Comparison of measured daily maximum NO2 concentration percentiles in Atlanta
for three high concentration years (1985, 1986, 1988) versus three low concentration
years (2005-2007) at one ambient monitor 61
Figure 6-2. Distributions of hourly NO2 concentrations at twelve ambient monitors in the
Boston CMSA, as is (top) and air quality adjusted to just meet the current standard
(bottom), Year 1995 62
Figure 6-3. Comparison of adjusted ambient monitoring concentrations (CS) or adjusted
benchmark level (dashed line) to simulate just meeting the current annual average
standard in Atlanta for year 2001 65
Figure 6-4. Comparison of the upper percentiles for where ambient monitoring NO2
concentrations (CS) and the benchmark level (dashed line) were adjusted to simulate
just meeting the current annual standard in Atlanta for year 2001. The hourly NO2
concentration distributions are provided in Figure 6-3 66
Figure 7-1. Illustration of three roadway distance categories used to characterize ambient
monitors in the Air Quality Characterization 80
Figure 7-2. Estimated mean number of daily maximum exceedances of short-term (1-hour)
potential health effect benchmarks in a year, using recent NO2 air quality (2001-2003)
adjusted to just meeting the current annual standard (0.053 ppm). Left graph:
monitors >100m from a major road; Middle graph: monitors >20 m and <100 m from
a major road; Right graph: monitors <20 m from a major road 105
Figure 7-3. Estimated mean number of daily maximum exceedances of short-term (1-hour)
potential health effect benchmarks in a year, using recent NO2 air quality (2004-2006)
adjusted to just meeting the current annual standard (0.053 ppm). Left graph:
monitors >100m from a major road; Middle graph: monitors >20 m and <100 m from
a major road; Right graph: monitors <20 m from a major road) 106
Figure 7-4. Estimated number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks (100 ppb, top; 200 ppb, bottom) in Chicago in a year, using
recent NO2 air quality data (2001-2006) adjusted to just meeting alternative 1-hour
standard levels (98th percentile, left; and 99th percentile, right) and monitors sited >100
m, > 20m and < 100 m, < 20 m of major roads 107
Figure 7-5. Estimated mean number of daily maximum exceedances of 200 ppb in four locations
(Phoenix, Los Angeles, Washington DC, and St. Louis) in a year, using recent NO2 air
quality data (2001-2006) adjusted to just meeting alternative 1-hour 98th percentile
standard levels and monitors sited >100 m, > 20 m and < 100 m, < 20 m of major
roads 108
Figure 7-6. Estimated mean number of daily maximum exceedances of short-term (1-hour)
potential health effect benchmarks on-roads in a year, using recent NO2 air quality
IX
-------
adjusted to just meeting the current annual standard (0.053 ppm) and an on-road
adjustment factor. Left graph: 2001-2003 air quality; Right graph: 2004-2006 air
quality 116
Figure 7-7. Estimated number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks (100 ppb, top; 200 ppb, bottom) on-roads in Chicago in a
year, using recent NC>2 air quality (2001-2006) adjusted to just meeting alternative 1-
hour standard levels (98th percentile, left; and 99th percentile) and an on-road
adjustment factor 117
Figure 7-8. Estimated mean number of daily maximum exceedances of 200 ppb on-roads in four
locations (Phoenix, Los Angeles, Washington DC, and St. Louis) in a year, using
recent NO2 air quality (2001-2006) adjusted to just meeting alternative 1-hour 98th
percentile standard levels and an on-road adjustment factor 118
Figre 7-9. Distribution of 1-hour NC>2 concentrations for three modeled receptors in Atlanta at
different vertical heights, using AERMOD predicted 2002 air quality 129
Figure 7-10. Comparison of measured daily maximum NO2 concentration percentiles in
Philadelphia for one high concentration years (1984) versus a low concentration years
(2007) at four ambient monitors 133
Figure 7-11. Comparison of the distribution of estimated C/Q, ratios or m for the not summer
category with fitted distributions 141
Figure 7-12. Comparison of the distribution of estimated Cv/Cb ratios or m for the summer
category with fitted distributions 141
Figure 8-1. General flow used for NC>2 exposure assessment 148
Figure 8-2. Four county modeling domain used for Atlanta exposure assessment 153
Figure 8-3. The 478 U.S. Census tracts representing area sources for on-road mobile emissions
that do not occur on major roadway links 162
Figure 8-4. Location of major roadway links and major stationary emission sources in Atlanta
modeling domain 166
Figure 8-5. Location of modeled receptors in Atlanta modeling domain 168
Figure 8-6. Comparison of measured ambient monitor NC>2 concentration distribution with the
modeled monitor receptor and receptors within 4 km of the monitors at three locations
in Atlanta for Year 2002 171
Figure 8-7. Comparison of measured ambient monitor NC>2 concentration diurnal profile with
the modeled monitor receptor and receptors within 4 km of the monitors at three
locations in Atlanta for Year 2002 172
Figure 8-8. Comparison of on-road/non-road ratios developed from AERMOD concentration
estimates for year 2002 and those derived from data reported in published NC>2
measurement studies 175
Figure 8-9. Comparison of annual average AERMOD predicted NO2 concentrations (on-road
and non-road receptors) and APEX modeled NO2 exposures (with and without
modeled indoor sources) in Atlanta modeling domain for year 2002 191
Figure 8-10. Comparison of estimated annual average NO2 exposures for Years 2001-2003 in
Atlanta modeling domain without modeled indoor sources 192
Figure 8-11. Distribution of measured daily average personal NO2 exposures for individuals in
Atlanta, stratified by two seasons (fall or spring) and cooking fuel (gas or electric).
Minimum (min), median (p50), and maximum (max) were obtained from each
-------
individual's multi-day exposure measurements. The figure generated here was based
on personal exposure measurements obtained from Suh (2008) 195
Figure 8-12. Distribution of estimated daily average NC>2 exposures for individuals in Atlanta,
stratified by two seasons (fall or spring) and with and without indoor sources, for Year
2002 APEX simulation. Lower bound (2.5th percentile, p2.5), median (p50), and
upper bound (97.5th percentile, p97.5) were calculated from each simulated persons
365 days of exposure. A random sample of 5% of persons (about 2,500 individuals) is
presented in each figure to limit the density of the graphs 196
Figure 8-13. Estimated number of all simulated asthmatics in the Atlanta model domain with at
least one NC>2 exposure at or above the potential health effect benchmark levels, using
modeled 2001 -2003 air quality (as is), without indoor sources 199
Figure 8-14. Estimated number of simulated asthmatic children in the Atlanta model domain
with at least one NC>2 exposure at or above the potential health effect benchmark
levels, using modeled 2001-2003 air quality (as is), without modeled indoor sources.
199
Figure 8-15. Estimated number of all simulated asthmatics in the Atlanta model domain with at
least one NC>2 exposure at or above potential health effect benchmark levels, using
modeled 2002 air quality (as is), both with and without modeled indoor sources 201
Figure 8-16. Estimated number asthmatic person-days in the Atlanta model domain with an NC>2
exposure at or above potential health effect benchmark levels, using modeled 2002 air
quality (as is), both with and without modeled indoor sources 201
Figure 8-17. Fraction of time all simulated persons in the Atlanta model domain spend in the
twelve microenvironments that corresponds with exceedances of the potential NC>2
health effect benchmark levels, a) > 100 ppb, b) > 200 ppb, and c) > 300 ppb, year
2002 air quality (as is) without indoor sources 204
Figure 8-18. Fraction of time all simulated persons in the Atlanta model domain spend in the
twelve microenvironments that corresponds with exceedances of the potential NC>2
health effect benchmark levels, a) > 100 ppb, b) > 200 ppb, and c) > 300 ppb, year
2002 air quality (as is) with indoor sources 205
Figure 8-19. Estimated percent of all asthmatics in the Atlanta modeling domain with repeated
NO2 exposures above potential health effect benchmark levels, using modeled 2002
air quality (as is), without indoor sources 206
Figure 8-20. Estimated percent of all asthmatics in the Atlanta modeling domain with repeated
NC>2 exposures above potential health effect benchmark levels, using modeled 2002
air quality (as is), with indoor sources 207
Figure 8-21. Estimated number of all asthmatics in the Atlanta modeling domain with at least
one NC>2 exposure at or above the potential health effect benchmark level, using
modeled 2002 air quality just meeting the current standard (cur std), with and without
modeled indoor sources 208
Figure 8-22. Estimated percent of asthmatics in the Atlanta modeling domain with repeated NC>2
exposures above health effect benchmark levels, using modeled 2002 air quality just
meeting the current standard, without modeled indoor sources 209
Figure 8-23. Estimated percent of asthmatics in the Atlanta modeling domain with NC>2
exposures at or above potential health effect benchmark levels, using modeled 2002
air quality adjusted to just meeting potential alternative standards, without indoor
sources 211
XI
-------
Figure 8-24. Estimated percent of asthmatics in the Atlanta modeling domain with multiple
exposures at or above potential health effect benchmark levels, using modeled 2002
air quality adjusted to just meeting a 50 ppb level 99th percentile form alternative
standard, without indoor sources 211
Figure 8-25. Estimated percent of asthmatics in the Atlanta modeling domain with multiple NCh
exposures at or above potential health effect benchmark levels, using modeled 2002
air quality adjusted to just meeting a 50 ppb level 99th percentile form alternative
standard, with indoor sources 212
Figure 8-26. Estimated percent of asthmatics in the Atlanta modeling domain with multiple NCh
exposures at or above potential health effect benchmark levels, using modeled 2002
air quality adjusted to just meeting a 100 ppb level 99th percentile form alternative
standard, without indoor sources 212
Figure 8-27. Comparison of estimated population and total roadway miles in 18 locations (data
from FHWA (2002) provided in Table 8-16) 221
Figure 8-28. Comparison of high ranked AERMOD 1-hour NC>2 concentrations (ug/m3) from
mobile sources across four NO2 monitoring locations based on 1ST vs. ATL
meteorological inputs for 2002 228
Figure 8-29. Comparison of the average ratios of predicted/observed concentrations of NO2
across four ambient monitors based on weekday vs. weekend only 234
Figure 8-30. Example comparison of estimated geometric mean and geometric standard
deviations of AER (h-1) for homes with air conditioning in several cities 244
Figure 8-31. Example of boot strap simulation results used in evaluating random sampling
variation of AER (h-1) distributions (RTF data) 245
Figure 8-32. Example of boot strap simulation results used in evaluating random sampling
variation of AER (h-1) distributions (outside CA) 246
Figure 9-1. Major components of nitrogen dioxide health risk assessment for emergency
department visits 256
xn
-------
List of Acronyms/Abbreviations
AADT Annual average daily traffic
A/C Air conditioning
AER Air exchange rate
AERMOD American Meteorological Society (AMS)/EPA Regulatory Model
AHS American Housing Survey
APEX EPA's Air Pollutants Exposure model, version 4
ANOVA One-way analysis of variance
AQS EPA's Air Quality System
AS Asthma symptoms
BAL bronchoalveolar lavage
BRFSS Behavioral Risk Factor Surveillance System
C Cough
CAA Clean Air Act
CAMD EPA's Clean Air Markets Division
CAMP Childhood Asthma Management Program
CASAC Clean Air Scientific Advisory Committee
CDC Centers for Disease Control
CHAD EPA's Consolidated Human Activity Database
CHF Congestive Heart Failure
Clev/Cinn Cleveland and Cincinnati, Ohio
CMSA Consolidated metropolitan statistical area
CO Carbon monoxide
COPD Chonic Obstructive Pulmonary Disease
COV Coefficient of Variation
C-R Concentration-Response
CTPP Census Transportation Planning Package
DVRPC Delaware Valley Regional Planning Council
ECP Eosinophil cationic protein
EDR Emergency department visits for respiratory disease
EDA Emergency department visits for asthma
ED AC Emergency department visits for asthma - children
FHWA Federal Highway Administration
HAAC Hospital admissions for asthma - children
ER Emergency room
EPA United States Environmental Protection Agency
EOC Exposure of Concern
GM Geometric mean
GSD Geometric standard deviation
GST Glutathione S-transferase (e.g., GSTM1, GSTP1, GSTT1)
h Hour
HNO3 Nitric acid
HONO Nitrous acid
ID Identification
ISA Integrated Science Assessment
xin
-------
ISH Integrated Surface Hourly Database
km Kilometer
L95 Lower limit of the 95th confidence interval
LA Los Angeles, California
m Meter
max Maximum
ME Microenvironment
med Median
MI Myocardial Infarction
min Minimum
MS Morning symptoms
MSA Metropolitan statistical area
NAAQS National Ambient Air Quality Standards
NAICS North American Industrial Classification System
NCEA National Center for Environmental Assessment
NEI National Emissions Inventory
NEM NAAQS Exposure Model
NCDC National Climatic Data Center
NHAPS National Human Activity Pattern Study
NHIS National Health Interview Survey
NC>2 Nitrogen dioxide
NOX Oxides of nitrogen
MV Nitrate ion
NWS National Weather Service
NYC New York City
NYDOH New York Department of Health
Os Ozone
OAQPS Office of Air Quality Planning and Standards
OR Odds ratio
ORD Office of Research and Development
ORIS Office of Regulatory Information Systems identification code
POC Parameter occurrence code
ppb Parts per billion
PEN Penetration factor
PM Particulate matter
PMN Polymorphonuclear
ppm Parts per million
PRB Policy-Relevant Background
PROX Proximity factor
PVMRM Plume Volume Molar Ratio Method
RECS Residential Energy Consumption Survey
RIU Rescue inhaler use
RR Relative risk
SAS Statistical Analysis Software
SB Shortness of breath
SEP Social-economic position
xiv
-------
SIC Standard Industrial Code
SD Standard deviation
se Standard error
TDM Travel Demand Modeling
tpy Tons per year
TRIM EPA's Total Risk Integrated Methodology
U95 Upper limit of the 95th confidence interval
US DOT United States Department of Transportation
US EPA United States Environmental Protection Agency
USGS United States Geological Survey
VMT Vehicle miles traveled
W Wheeze
xv
-------
1. INTRODUCTION
1.1 OVERVIEW
The U.S. Environmental Protection Agency (EPA) is conducting a review of the national
ambient air quality standards (NAAQS) for nitrogen dioxide (NC^). Sections 108 and 109 of the
Clean Air Act (The Act) govern the establishment and periodic review of the air quality criteria
and the NAAQS. These standards are established for pollutants that may reasonably be
anticipated to endanger public health or welfare, and whose presence in the ambient air results
from numerous or diverse mobile or stationary sources. The NAAQS are based on air quality
criteria, which reflect the latest scientific knowledge useful in indicating the kind and extent of
identifiable effects on public health or welfare that may be expected from the presence of the
pollutant in ambient air. The EPA Administrator promulgates and periodically reviews primary
(health-based) and secondary (welfare-based) NAAQS for such pollutants. Based on periodic
reviews of the air quality criteria and standards, the Administrator makes revisions in the criteria
and standards and promulgates any new standards as may be appropriate. The Act also requires
that an independent scientific review committee advise the Administrator as part of this NAAQS
review process, a function now performed by the Clean Air Scientific Advisory Committee
(CASAC).
The Agency has recently made a number of changes to the process for reviewing the
NAAQS (described at http://www.epa.gov/ttn/naaqs/). In making these changes, the Agency
consulted with CASAC. This new process, which is being applied to the current review of the
NC>2 NAAQS, contains four major components. Each of these components, as they relate to the
review of the NC>2 primary NAAQS, is described below.
The first of these components is an integrated review plan. This plan presents the
schedule for the review, the process for conducting the review, and the key policy-relevant
science issues that will guide the review. The integrated review plan for this review of the NC>2
primary NAAQS is presented in the Integrated Review Plan for the Primary National Ambient
Air Quality Standard for Nitrogen Dioxide (EPA, 2007a). The policy-relevant questions
identified in this document to guide the review are:
-------
• Has new information altered the scientific support for the occurrence of health effects
following short- and/or long-term exposure to levels of nitrogen oxides (NOX) found in
the ambient air?
• What do recent studies focused on the near-roadway environment tell us about health
effects of NOX?
• At what levels of NOX exposure do health effects of concern occur?
• Has new information altered conclusions from previous reviews regarding the plausibility
of adverse health effects caused by NOX exposure?
• To what extent have important uncertainties identified in the 1996 review been reduced
and/or have new uncertainties emerged?
• What are the air quality relationships between short-term and long-term exposures
toNOx?
Additional questions will become relevant if the evidence suggests that revision of the current
standard might be appropriate. These questions are:
• Is there evidence for the occurrence of adverse health effects at levels of NOX lower than
those observed previously? If so, at what levels and what are the important uncertainties
associated with that evidence?
• Do exposure estimates suggest that exposures of concern for NOx-induced health effects
will occur with current ambient levels of NC>2 or with levels that just meet current, or
potential alternative, standards? If so, are these exposures of sufficient magnitude such
that the health effects might reasonably be judged to be important from a public health
perspective? What are the important uncertainties associated with these exposure
estimates?
• Do the evidence, the air quality assessment, and the risk/exposure assessment provide
support for considering different standard indicators or averaging times?
• What range of levels is supported by the evidence, the air quality assessment, and the
risk/exposure assessments? What are the uncertainties and limitations in the evidence
and the assessments?
-------
• What is the range of forms supported by the evidence, the air quality assessment, and the
exposure/risk assessments? What are the uncertainties and limitations in the evidence
and the assessments?
The second component of the review process is a science assessment. A concise
synthesis of the most policy-relevant science has been compiled into the Integrated Science
Assessment (ISA). The ISA is supported by a series of annexes that contain more detailed
information about the scientific literature. The ISA to support this review of the NC>2 primary
NAAQS is presented in the Integrated Science Assessment for Oxides of Nitrogen - Health
Criteria, henceforth referred to as the ISA (EPA, 2008a).
The third component of the review process is a risk and exposure assessment, which is
described in this document. The purpose of this document is to communicate EPA's assessment
of exposures and risks associated with ambient NCh. In this assessment, we have developed
estimates of human exposures and risks associated with current ambient levels of NO2, with
levels that just meet the current standard, and with levels that just meet potential alternative
standards. Figure 1-1 (below) presents a schematic overview of the analyses described in this
document and how those analyses fit together. Each of the steps highlighted in Figure 1-1 is
described in more detail in subsequent sections.
-------
Evaluate health evidence in ISA
Identification of
potential health
benchmark levels
'3 .0
•H 15
i
p
*~
Air quality
characterization in
locations across U.S.
1
I
Qualitativ
controlle
Inform city
selection _
1 J
Compare NO2 levels with
potential health benchmark
studies
L
Qualitative
characterization of U.S.
epidemiology studies
Exposure
characterization in
select city/cities
1 1
0} o
Identify potential
alternative
standards
Epidemiology-based
quantitative risk
assessment
levels
Compare modeled NO2
exposures to potential
health benchmark levels
Output: Number of times per
year that NO2 concentrations
exceed potential health
benchmark levels
Output: Persons exposed and
total occurrences of exposures
to N02 levels that exceed
potential health benchmark
levels
Output: Number of
occurrences and percent of
I incidence for specified
health effect
Risk Characterization Based on
Air Quality
Risk Characterization Based on
Exposure
Risk Characterization Based on
Epidemiology
Risk-based
considerations to
Inform standard setting
Figure 1-1. Overview of the analyses described in this document and their interconnections
The results of the risk and exposure assessment is considered alongside the health evidence, as
evaluated in the final ISA, to inform the policy assessment and rulemaking process, as discussed
below in chapter 10. The draft plan for conducting the risk and exposure assessment to support
the NC>2 primary NAAQS is presented in the Nitrogen Dioxide Health Assessment Plan: Scope
and Methods for Exposure and Risk Assessment, henceforth referred to as the Health Assessment
Plan (EPA, 2007b). The first draft of the risk and exposure assessment is presented in Risk and
Exposure Assessment to Support the Review of the NO2 Primary National Ambient Air Quality
Standard: First Draft (EPA, 2008b). The second draft is presented in Risk and Exposure
-------
Assessment to Support the Review of the NO2 Primary National Ambient Air Quality Standard:
Second Draft (EPA, 2008c).
The fourth component of the process is the policy assessment and rulemaking. The
Agency's views on policy options will be published in the Federal Register as an advance notice
of proposed rulemaking (ANPR). This policy assessment will address the adequacy of the
current standard and of potential alternative standards, which will be defined in terms of
indicator, averaging time, form,1 and level. To accomplish this, the policy assessment will
consider the results of the final risk and exposure assessment as well as the scientific evidence
(including evidence from the epidemiologic, controlled human exposure, and animal
toxicological literatures) evaluated in the ISA, drawing from the discussion in chapter 10.
Taking into consideration CASAC advice and recommendations, as well as public comment on
the ANPR, the Agency will publish a proposed rule, to be followed by a public comment period.
Taking into account comments received on the proposed rule, the Agency will issue a final rule
to complete the rulemaking process.
1.2 HISTORY
1.2.1 History of the NO2 NAAQS
On April 30, 1971, EPA promulgated identical primary and secondary NAAQS for NC>2
under section 109 of the Act. The standards were set at 0.053 parts per million (ppm), annual
average (36 FR 8186). In 1982, EPA published Air Quality Criteria for Oxides of Nitrogen
(EPA, 1982), which updated the scientific criteria upon which the initial NC>2 standards were
based. On February 23, 1984, EPA proposed to retain these standards (49 FR 6866). After
taking into account public comments, EPA published the final decision to retain these standards
on June 19, 1985 (50 FR 25532).
On July 22, 1987, EPA announced that it was undertaking plans to revise the 1982 air
quality criteria (52 FR 27580). In November 1991, EPA released an updated draft air quality
criteria document for CASAC and public review and comment (56 FR 59285). The draft
document provided a comprehensive assessment of the available scientific and technical
information on health and welfare effects associated with NO2 and other oxides of nitrogen. The
1 The "form" of a standard defines the air quality statistic that is to be compared to the level of the standard
in determining whether an area attains the standard.
-------
CAS AC reviewed the draft document at a meeting held on July 1, 1993 and concluded in a
closure letter to the Administrator that the document "provides a scientifically balanced and
defensible summary of current knowledge of the effects of this pollutant and provides an
adequate basis for EPA to make a decision as to the appropriate NAAQS for NO2" (Wolff,
1993). The Air Quality Criteria Document for the Oxides of Nitrogen was then finalized (EPA,
1993).
The EPA also prepared a Staff Paper that summarized an air quality assessment for NO2
conducted by the Agency (McCurdy, 1994), summarized and integrated the key studies and
scientific evidence contained in the revised air quality criteria document, and identified the
critical elements to be considered in the review of the NO2 NAAQS. The CASAC reviewed two
drafts of the Staff Paper and concluded in a closure letter to the Administrator (Wolff, 1995) that
the document provided a "scientifically adequate basis for regulatory decisions on nitrogen
dioxide." In September of 1995, EPA finalized the Staff Paper entitled, "Review of the National
Ambient Air Quality Standards for Nitrogen Dioxide: Assessment of Scientific and Technical
Information" (EPA, 1995).
In October 1995, the Administrator announced her proposed decision not to revise either
the primary or secondary NAAQS for NO2 (60 FR 52874; October 11, 1995). A year later, the
Administrator made a final determination not to revise the NAAQS for NO2 after careful
evaluation of the comments received on the proposal (61 FR 52852, October 8, 1996). The level
for both the existing primary and secondary NAAQS for NO2 is 0.053 parts per million (ppm)
(100 micrograms per cubic meter of air [|j,g/m3]), annual arithmetic average, calculated as the
arithmetic mean of the l-hourNO2 concentrations.
1.2.2 Health Evidence from Previous Review
The prior Air Quality Criteria Document (AQCD) for Oxides of Nitrogen (EPA, 1993)
concluded that there were two key health effects of greatest concern at ambient or near-ambient
levels of NO2, increased airway responsiveness in asthmatic individuals after short-term
exposures and increased occurrence of respiratory illness in children with longer-term exposures.
Evidence also was found for increased risk of emphysema, but this was of major concern only
with exposures to levels of NO2 much higher than then-current ambient levels. The evidence
regarding airway responsiveness was drawn largely from controlled human exposure studies.
-------
The evidence for respiratory illness was drawn from epidemiologic studies that reported
associations between respiratory symptoms and indoor exposures to NC>2 in people living in
homes with gas stoves. The biological plausibility of the epidemiologic results was supported by
toxicological studies that detected changes in lung host defenses following NO2 exposure.
Subpopulations considered potentially more susceptible to the effects of NO2 included
individuals with preexisting respiratory disease, children, and the elderly.
1.2.3 Assessment from Previous Review
In the previous review of the NC>2 NAAQS, risks were characterized by comparing
ambient monitoring data, which were used as a surrogate for exposure, with potential health
benchmark levels identified from controlled human exposure studies. At the time of the review,
a few studies indicated the possibility for adverse health effects due to short-term (e.g., 1-hour)
exposures between 0.20 ppm and 0.30 ppm NC>2. Therefore, the focus of the assessment was on
the potential for short-term (i.e., 1-hour) exposures to NC>2 levels above potential health
benchmarks in this range. The assessment used monitoring data from the years 1988-1992 and
screened for sites with one or more hourly exceedances of potential short-term health effect
benchmarks. Predictive models were then constructed to relate the frequency of hourly
concentrations above short-term health effect benchmarks to a range of annual average
concentrations, including the current standard. Based on the results of this analysis, both
CAS AC (Wolff, 1995) and the Administrator (60 FR 52874) concluded that the minimal
occurrence of short-term peak concentrations at or above a potential health effect benchmark of
0.20 ppm (1-h average) indicated that the existing annual standard would provide adequate
health protection against short-term exposures. This conclusion was a key element in the
decision in the 1996 review to retain the existing annual standard.
1.3 SCOPE OF THE RISK AND EXPOSURE ASSESSMENT FOR THE
CURRENT REVIEW
NOX, for purposes of this document, include multiple gaseous (e.g., NC>2, NO, HONO)
and particulate (e.g., nitrate) species. As discussed in the integrated review plan (2007a), the
current review of the NC>2 NAAQS will focus on the gaseous species of NOX and will not
consider health effects directly associated with particulate species of NOX. Of the gaseous
-------
species, EPA has historically determined it appropriate to specify the indicator of the standard in
terms of NC>2 because the majority of the information regarding health effects and exposures is
for NO2. In the current review, staff notes that no alternative to NO2 has been advanced as being
a more appropriate surrogate for ambient gaseous NOX. Controlled human exposure studies and
animal toxicology studies provide specific evidence for health effects following exposure to
NC>2. Epidemiologic studies also typically report levels of NC>2, as opposed to other gaseous
NOX, though the degree to which monitored NC>2 reflects actual NC>2 levels, as opposed to NC>2
plus other gaseous NOX, can vary (e.g.,. see section 2.2.3 of this document). Therefore, NC>2 will
be used as the indicator for the gaseous NOX in the risk and exposure assessments described in
this document.
-------
2. SOURCES, AMBIENT LEVELS, AND EXPOSURES
2.1 SOURCES OF NO2
Ambient levels of NO2 are the product of both direct NO2 emissions and emissions of
other NOX (e.g, NO), which can then be converted to NO2 (for a more detailed discussion see the
ISA, section 2.2). Nationally, anthropogenic sources account for approximately 87% of total
NOX emissions. Mobile sources (both on-road and off-road) account for about 60% of total
anthropogenic emissions of NOX, while stationary sources (e.g., electrical utilities and industry)
account for the remainder (annex table 2.6-1). Highway vehicles represent the major mobile source
component. In the United States, approximately half the mobile source emissions are contributed by
diesel engines and half are emitted by gasoline-fueled vehicles and other sources (annex section
2.6.2 and Table 2.6-1). Apart from these anthropogenic sources, there are also natural sources of
NOX including microbial activity in soils, lightning, and wildfires (ISA, section 2.2.1 and annex
section 2.6.2).
2.2 AMBIENT LEVELS OF NO2
2.2.1 Background on NOi monitoring network
From the inception of the NC>2 monitoring network in the late 1970's through the present
day, the number of monitoring sites has remained relatively stable (Watkins, 2008). As of
October 2008, there were 409 NOX monitors within the United States actively reporting NO2 data
into EPA's Air Quality System (AQS). The NO2 network was originally established for
implementation of the NO2 NAAQS promulgated in 1971. The first requirements for NO2
monitoring to implement the 1971 NO2 NAAQS were established in May of 1979. At that time,
two NO2 national ambient monitoring stations (NAMS) were required in areas of the country
with populations greater than 1,000,000. 40 CFR Part 58, Appendix D, section 3.5. The
regulations noted that within urban areas, two permanent monitors are sufficient, and with
respect to those two monitors provided:
The first station (category (a), middle scale or neighborhood scale) would be to
measure the photochemical production of NO2 and would best be located in that
part of the urban area where the emission density of NOX is the highest. The
second station (category (b) urban scale), would be to measure the NO2 produced
-------
from the reaction of NO with 63 and should be downwind of the area peak NOX
emission areas.
40 CFRPart 58, Appendix D, section 3.5.
In October 2006, EPA revised the monitoring requirements for NO2 in light of the fact
that there are no NO2 non-attainment areas under the current standards. The 2006 rule
eliminated the minimum requirements for the number of NO2 monitoring sites. 40 CFR Part 58,
Appendix D, section 4.3. However, the rule requires continued operation of existing State and
local monitoring stations (SLAMS) until discontinuation is approved by the EPA Regional
Administrator. The revised rule further requires that where SLAMS NO2 monitoring is ongoing,
"at least one NO2 site in the area must be located to measure the maximum concentration of
N02."
As noted above, the size of the NO2 network has remained fairly stable through time,
even though no minimum monitoring sites were required under the 2006 rule. The maintenance
of the NO2 monitoring network has been driven by several factors, including the need to support
ozone (63) modeling and forecasting, the need to track PM precursors, and a general desire on
the part of states to continue to understand trends in ambient NC>2.
To characterize the current NC>2 network, staff has reviewed the NC>2 network meta-data.
The data reviewed are those available from AQS in October 2008, for monitors reporting data in
2008. The meta-data fields are typically created by state and local agencies when a monitor site
is opened, moved, or re-characterized. While these files are useful for characterizing specific
monitors, there is some uncertainty surrounding this meta-data given that there is no routine or
enforced process for updating or correcting meta-data fields. With this uncertainty in mind, staff
has compiled information on the monitoring objectives and measurement scales for monitors in
the NO2 network.
The monitor objective meta-data field describes the purpose of the monitor. For example
the purpose of a particular monitor could be to characterize health effects, photochemical
activity, transport, and/or welfare effects. As of October 2008, there were 489 records of NO2
monitor objective values (some monitors have multiple monitor objectives). Table 2-1 lists the
distribution of monitoring objectives across the network. There are 12 categories of monitor
objectives for NO2 monitors within AQS. The "other" category is for sites likely addressing a
10
-------
state or local need outside of the routine objectives, and the "unknown" category represents
missing meta-data. The remaining categories stem directly from categorizations of site types
within CFR. In 40 CFR Part 58 Appendix D, there are six examples of NC>2 site types:
1. Sites located to determine the highest concentration expected to occur in the area
covered by the network (Highest Concentration).
2. Sites located to measure typical concentrations in areas of high population
(Population Exposure).
3. Sites located to determine the impact of significant sources or source categories
on air quality (Source Oriented).
4. Sites located to determine general background concentration levels (General
Background).
5. Sites located to determine the extent of regional pollutant transport among
populated areas; and in support of secondary standards (Regional Transport).
6. Sites located to measure air pollution impacts on visibility, vegetation damage, or
other welfare-based impacts (Welfare Related Impacts).
The remaining four categories available are a result of updating the AQS database. In the more
recent upgrade to AQS, the data handlers inserted the available site types for Photochemical
Assessment Monitoring Stations (PAMS) network. These PAMS site types are spelled out in 40
CFR Part 58 Appendix D:
1. Type 1 sites are established to characterize upwind background and transported
Os and its precursor concentrations entering the area and will identify those areas
which are subjected to transport (Upwind Background).
2. Type 2 sites are established to monitor the magnitude and type of precursor
emissions in the area where maximum precursor emissions are expected to impact
and are suited for the monitoring of urban air toxic pollutants (Max. Precursor
Impact).
3. Type 3 sites are intended to monitor maximum O3 concentrations occurring
downwind from the area of maximum precursor emissions (Max. Os
Concentration).
11
-------
4. Type 4 sites are established to characterize the downwind transported 63 and its
precursor concentrations exiting the area and will identify those areas which are
potentially contributing to overwhelming transport in other areas (Extreme
Downwind).
Table 2-1. NOx Network Distribution of Monitor Objectives
NOx Monitor
Objective
Population Exposure
Highest Concentration
General Background
Max. Precursor Impact
(PAMS Type 2 Site)
Source Oriented
Upwind Background
(PAMS Type 1 Site)
Regional Transport
Other
Max. Os Concentration
(PAMS Type 3 Site)
Extreme Downwind
(PAMS Type 4 Site)
Welfare Related Impacts
Unknown
Totals:
Number of Monitor
Objective Records
177
58
51
21
19
18
12
9
8
3
1
112
489
Percent Distribution
36.20
11.86
10.43
4.29
3.89
3.68
2.45
1.84
1.64
0.61
0.20
22.90
100%
The spatial measurement scales are laid out in 40 CFR Part 58, Appendix D, Section 1
"Monitoring Objectives and Spatial Scales." This part of the regulation spells out what data
from a monitor can represent in terms of air volumes associated with area dimensions:
Microscale - 0 to 100 meters
Middle Scale - 100 to 500 meters
Neighborhood Scale - 500 meters to 4 kilometers
Urban Scale - 4 to 50 kilometers
Regional Scale - 50 kilometers up to 1000km
There are meta-data records for the NO2 network to indicate what the measurement scale of a
particular monitor represents. There are 386 NO2 monitor records in AQS with available
measurement scale information. Table 2-2 shows the measurement scale distribution across all
NO2 sites form the available data in AQS of monitors reporting data in 2008.
12
-------
Table 2-2. NOx Network Distribution across Measurement Scales.
Measurement Scale
Microscale
Middle Scale
Neighborhood
Urban Scale
Regional Scale
Totals:
Number of Measurement
Scale Records
3
23
212
119
29
386
Percent Distribution
0.78
5.96
54.92
30.83
7.51
100%
In summary, upon review of the known 409 monitors reporting data to AQS in 2008, and
the distribution of the available data from the categories of monitor objective and measurement
scale, we see the NO2 network is primarily targeting public health and photochemical process
monitoring objectives. We note that nearly half of the monitor objective records are directly
targeting public health through the population exposure (36.2%) and highest concentration
(1 1.8%) categories alone. The other categories serve to inform public health concerns, but also
address photochemistry issues where NOX serves as a precursor to ozone. Further, it appears that
approximately 10% of NO2 monitors are in place to serve the PAMS network. In reality, a large
majority of sites likely could serve both public health and photochemistry related objectives due
to their proximity to urban areas. The exceptions would likely be categories such as upwind
background, extreme downwind, regional transport, and possibly maximum 63 concentration.
These four categories only represent approximately 7% of the NO2 network, and have a higher
likelihood of being rural and likely regional in scale.
2.2.2 Trends in ambient concentrations
As noted above, NO2 is monitored largely in urban areas and, therefore, data from the
NO2 monitoring network is generally more representative of urban areas than rural areas.
According to monitoring data, nationwide levels of ambient NO2 (annual average) decreased
41% between 1980 and 2006 (ISA, Figure 2.4-15). Between 2003 and 2005, national mean
concentrations of NO2 were about 15 ppb for averaging periods ranging from a day to a year.
The average daily maximum hourly NO2 concentrations were approximately 30 ppb. These
values are about twice as high as the 24-h averages. The highest maximum hourly concentrations
(-200 ppb) between 2003 and 2005 are more than a factor often higher than the mean hourly or
13
-------
24-h concentrations (ISA, Figure 2.4-13). The highest levels of NO2 in the United States can be
found in and around Los Angeles, in the Midwest, and in the Northeast. Policy-relevant
background concentrations, which are those concentrations that would occur in the United States
in the absence of anthopogenic emissions in continental North America (defined here as the
United States, Canada, and Mexico), are estimated to range from only 0.1 ppb to 0.3 ppb (ISA,
section 2.4.6).
Ambient levels of NO2 exhibit both seasonal and diurnal variation. In southern cities,
such as Atlanta, higher concentrations are found during winter, consistent with the lowest mixing
layer heights being found during that time of the year. Lower concentrations are found during
summer, consistent with higher mixing layer heights and increased rates of photochemical
oxidation of NO2. For cities in the Midwest and Northeast, such as Chicago and New York City,
higher levels tend to be found from late winter to early spring with lower levels occurring from
summer though the fall. In Los Angeles the highest levels tend to occur from autumn though
early winter and the lowest levels from spring though early summer. Mean and peak
concentrations in winter can be up to a factor of two larger than in the summer at sites in Los
Angeles. In terms of daily variability, NO2 levels typically peak during the morning rush hours.
Monitor siting plays a key role in evaluating diurnal variability as monitors located further away
from traffic will show cycles that are less pronounced over the course of a day than monitors
located closer to traffic.
2.2.3 Uncertainty Associated with the Ambient NOi Monitoring Method
The method for estimating ambient NO2 levels (i.e., subtraction of NO from a measure of
total NOX) is subject to interference by NOX oxidation products (e.g., PAN, HNOs) (ISA, section
2.3). Limited evidence suggests that these compounds result in an overestimate of NO2 levels by
roughly 20 to 25% at typical ambient levels. Smaller relative errors are estimated to occur in
measurements taken near strong NOX sources since most of the mass emitted as NO or NO2
would not yet have been further oxidized. Relatively larger errors appear in locations more
distant from strong local NOX sources. Additionally, many NO2 monitors are elevated above
ground level in the cores of large cities. Because most sources of NO2 are near ground level (i.e.,
combustion emissions from traffic), there is a gradient of NO2 with higher levels near ground
level and lower levels being detected at the elevated monitor. One comparison has found an
14
-------
average of a 2.5-fold higher NO2 concentration measured at 4 meters above the ground compared
to 15 meters above the ground. The ISA notes that levels are likely even higher at elevations
below 4 meters (ISA, section 2.5.3.3). Another source of uncertainty in exposure estimates can
result from monitor location. NC>2 monitors are sited for compliance with air quality standards
rather than for capturing small-scale variability in NC>2 concentrations near sources such as
roadway traffic. Significant gradients in NC>2 concentrations near roadways have been observed
in several studies, and NC>2 concentrations have been found to be correlated with distance from
roadway and traffic volume (ISA, section 2.5.3.2).
2.3 EXPOSURE TO NO2
2.3.1 Overview
Human exposure to an airborne pollutant can be characterized by contact between a
person and the pollutant at a specific concentration for a specified period of time (ISA, section
2.5.1). The integrated exposure of a person to a given pollutant is the time-weighted average of
the exposures over all time intervals for all microenvironments in which the individual spends
time. Microenvironments in which people are exposed to air pollutants such as NC>2 typically
include residential indoor environments and other indoor locations, near-traffic outdoor
environments and other outdoor locations, and in vehicles (ISA, Figure 2.5-1).
There is a large amount of variability in the time that individuals spend in different
microenvironments, but on average people spend the majority of their time (about 87%) indoors.
Most of this time is spent at home with less time spent in an office/workplace or other indoor
locations (ISA, Figure 2.5-1). On average in the U.S., people spend about 8% of their time
outdoors and 6% of their time in vehicles. Significant variability surrounds each of these broad
estimates, particularly when considering influential personal attributes such as age or gender;
when accounting for daily, weekly, or seasonal factors influencing personal behavior; or when
characterizing individual variability in time spent in various locations (McCurdy and Graham,
2003; Graham and McCurdy, 2004). Typically, the time spent outdoors or in vehicles could vary
by 100% or more depending on which of these influential factors are considered. Exposure
misclassification can result when the time spent in different microenvironments is not taken into
consideration and may obscure the true relationship between ambient air pollutant exposures and
health outcomes.
15
-------
2.3.2 Uncertainty Associated with Ambient Levels as a Surrogate for Exposure
Many epidemiologic studies rely on measures of ambient NC>2 concentrations as
surrogates for personal exposure to ambient NC>2. Results have been mixed regarding the
appropriateness of using ambient levels of NC>2 as a surrogate for personal exposures to ambient
NC>2. Studies examining the association between ambient NC>2 and personal exposure to NC>2
have generated mixed results due to 1) the prevalence of indoor sources of NO2; 2) the spatial
heterogeneity of NC>2 in study areas; 3) the seasonal and geographic variability in the infiltration
of ambient NC>2; 4) differences in the time spent in different microenvironments; and 5)
differences in study design. As a result, some researchers have concluded that ambient NO2 may
be a reasonable proxy for personal exposure, while others have noted that caution must be
exercised (ISA, section 2.5.9). However, the possible consequences of this exposure error do not
bias conclusions in a positive direction (see chapter 4 of this document) since it generally tends
to reduce, rather than increase, effect estimates (ISA, section 5.2.2).
16
-------
3. AT RISK POPULATIONS
3.1 OVERVIEW
Specific groups within the general population are at increased risk for suffering adverse
effects from NC>2 exposure. This could occur because they are affected by lower levels of NC>2
than the general population (susceptibility), because they experience a larger health impact than
the general population to a given level of exposure (susceptibility), and/or because they are
exposed to higher levels of NC>2 than the general population (vulnerability). The term
susceptibility generally encompasses innate (e.g., genetic or developmental) and/or acquired
(e.g., age or disease) factors that make individuals more likely to experience effects with
exposure to pollutants. Given the likely heterogeneity of individual responses to air pollution,
the severity of health effects experienced by a susceptible subgroup may be much greater than
that experienced by the population at large. Factors that may influence susceptibility to the
effects of air pollution include age (e.g., infants, children, elderly); gender; race/ethnicity;
genetic factors; and pre-existing disease/condition (e.g., obesity, diabetes, respiratory disease,
asthma, chonic obstructive pulmonary disease (COPD), cardiovascular disease, airway
hyperresponsiveness, respiratory infection, adverse birth outcome) (ISA, sections 4.3.1, 4.3.5,
and 5.3.2.8). In addition, some population groups are vulnerable to pollution-related effects
because their air pollution exposures are higher than those of the general population. Factors that
may influence vulnerability to the effects of air pollution include socioeconomic status,
education level, air conditioning use, proximity to roadways, geographic location, level of
physical activity, and work environment (e.g., indoor versus outdoor) (ISA, section 4.3.5). The
ISA discusses factors that can confer susceptibility and/or vulnerability to air pollution with most
of the discussion devoted to factors for which NO2-specific evidence exists (ISA, section 4.3).
These factors are presented in table 3-1 below (from section 4.3.5 of the ISA) and are discussed
in subsequent sections of this chapter (see ISA, chapter 4 for more detail).
17
-------
Table 3-1. Overview of Susceptibility and Vulnerability Factors
Susceptibility Factors Vulnerability Factors
, „ , • Socioeconomic status
• Age. Gender
, , , . ,, , . , , , . , • Education level
• Adverse birth outcomes: e.g., pretemi birth, low birth
weight, growth restriction, birth defects • Air conditioning Use
• Race/etliuicity • Proximity to Roadways
1 Genetic factors • Geographic Location (West
• Pre-existing disease, e.g.. diabetes " l':'
,-., .. " Level of Exercise
1 Obesity
r> • * i- *i ,-./-.ivr» * Work Eiivirouinent (e.g..
• Respiratory diseases, e.g.. asthma. COPD , . , ~ '-
*• •' *- outdoor workers)
1 Cardiovascular diseases
3.2 SUSCEPTIBILITY: PRE-EXISTING DISEASE
A number of health conditions have been found to put individuals at greater risk for
adverse events following exposure to air pollution. In general, these include asthma, COPD,
respiratory infection, conduction disorders, congestive heart failure (CHF), diabetes, past
myocardial infarction (MI), obesity, coronary artery disease, low birth weight/prematurity, and
hypertension (ISA, sections 4.3.1, 4.3.5, and 5.3.2.9). In addition to these conditions,
epidemiologic evidence indicates that individuals with bronchial or airway hyperresponsiveness,
as determined by methacholine provocation, may be at increased risk for experiencing
respiratory symptoms (ISA, section 4.3.1). In considering NC>2 specifically, the ISA evaluated
studies on asthmatics, individuals with cardiopulmonary disease, and diabetics (ISA, sections
4.3.1.1 and 4.3.1.2). These groups are discussed in more detail below.
Epidemiologic and controlled human exposure studies, supported by animal toxicology
studies, have provided evidence for associations between NC>2 exposure and respiratory effects in
asthmatics (ISA, section 4.3.1.1). The ISA found evidence from epidemiologic studies for an
association between ambient NC>2 and children's hospital admissions, emergency department
(ED) visits, and calls to doctors for asthma. NO2 levels were associated with aggravation of
asthma effects that include symptoms, medication use, and lung function. Time-series studies
also demonstrated a relationship in children between hospital admissions or ED visits for asthma
and ambient NO2 levels, even after adjusting for co-pollutants such as particulate matter (PM)
and carbon monoxide (CO) (ISA, section 4.3.1.1). Important evidence was also available from
18
-------
epidemiologic studies of indoor NC>2 exposures. Recent studies have shown associations with
asthma attacks and severity of virus-induced asthma (ISA, section 4.3.1.1). In addition, in
controlled human exposure studies, airway hyperresponsiveness in asthmatics occurred
following exposure to lower NC>2 concentrations than the concentrations that caused effects on
other endpoints (ISA, sections 5.3.2.1-5.3.2.6).
Compared to asthma, less evidence is available to support cardiovascular disease as a
mediator of susceptibility to NC>2. However, recent epidemiologic studies report that individuals
with preexisting conditions (e.g., including diabetes, CHF, prior MI) may be at increased risk for
adverse cardiac health events associated with ambient NC>2 concentrations (ISA, section 4.3.1.2).
The small number of controlled human exposure and animal toxicological studies that have
evaluated cardiovascular endpoints provide only limited supporting evidence for susceptibility to
NC>2 in persons with cardiovascular disease (ISA, section 4.3.1.2).
3.3 SUSCEPTIBILITY: AGE
The ISA identifies infants, children (i.e., <18 years of age), and older adults (i.e., >65
years of age) as groups that are potentially more susceptible than the general population to the
health effects associated with ambient NO2 concentrations (ISA, section 4.3.2). The ISA found
evidence that associations of NC>2 with respiratory ED visits and hospitalizations were stronger
among children and older adults, though not all studies had comparable findings on this issue
(ISA, section 4.3.2). In addition, long-term exposure studies suggest effects in children that
include impaired lung function growth, increased respiratory symptoms and infections, and onset
of asthma (ISA, section 3.4 and 4.3.2). In some studies, associations between NC>2 and
hospitalizations or ED visits for CVD have been observed in elderly populations. Among studies
that observed positive associations between NC>2 and mortality, a comparison indicated that, in
general, the elderly population was more susceptible than the non-elderly population to NO2
effects (ISA, section 4.3.2).
3.4 SUSCEPTIBILITY: GENETICS
As noted in the ISA (section 4.3.4), genetic factors related to health outcomes and
ambient pollutant exposures merit consideration. Several criteria must be satisfied in selecting
and establishing useful links between polymorphisms in candidate genes and adverse respiratory
19
-------
effects. First, the candidate gene must be significantly involved in the pathogenesis of the
adverse effect of interest. Second, polymorphisms in the gene must produce a functional change
in either the protein product or in the level of expression of the protein. Third, in epidemiologic
studies, the issue of confounding by other environmental exposures must be carefully considered
(ISA, section 4.3.4).
Investigation of genetic susceptibility to NC>2 effects has focused on the glutathione S-
tranferase (GST) gene. Several GST genes have common, functionally-important alleles that
affect host defense in the lung (ISA, section 4.3.4). GST genes are inducible by electrophilic
species (e.g., reactive oxygen species) and individuals with genotypes that result in enzymes with
reduced or absent peroxidase activity are likely to have reduced defenses against oxidative insult.
This could potentially result in increased susceptibility to inhaled oxidants and radicals.
However, data on genetic susceptibility to NC>2 are only beginning to emerge and, while it
remains plausible that there are genetic factors that can influence health responses to NC>2, the
few available studies do not provide specific support for genetic susceptibility to NC>2 exposure
(ISA, section 4.3.4).
3.5 SUSCEPTIBILITY: GENDER
As reported in the ISA, a limited number of NC>2 studies have stratified results by gender.
The results of these studies were mixed, and the ISA does not draw conclusions regarding the
potential for gender to confer susceptibility to the effects of NC>2 (ISA, section 4.3.3).
3.6 VULNERABILITY: PROXIMITY (ON OR NEAR) TO ROADWAYS
The ISA includes discussion of vulnerable populations that experience increased NO2
exposures on or near roadways (ISA, section 4.3.6). Large gradients in NOx concentrations near
roadways lead to increased exposures for individuals residing, working, or attending school in
the vicinity of roadways. Many studies find that indoor, personal, and outdoor NC>2 levels are
strongly associated with proximity to traffic or to traffic density (ISA, section 4.3.6). Due to
high air exchange rates, NC>2 levels inside a vehicle could rapidly approach levels outside the
vehicle during commuting (ISA, section 4.3.6). Mean in-vehicle NC>2 levels are between 2 and 3
times ambient levels measured at fixed sites nearby (ISA, section 4.3.6). Therefore, individuals
with occupations that require them to be in traffic or close to traffic (e.g., bus and taxi drivers,
20
-------
highway patrol officers, toll collectors) and individuals with long commutes could be exposed to
relatively high levels of NO2 compared to ambient levels. Due to the high peak exposures while
driving, total personal exposure could be underestimated if exposures while commuting are not
considered.
3.7 VULNERABILITY: SOCIOECONOMIC STATUS
The ISA discusses evidence that socioeconomic status (SES) modifies the effects of air
pollution (section 4.3.6). Many recent studies examined modification by SES indicators on the
association between mortality and PM or other indices such as traffic density, distance to
roadway, or a general air pollution index (ISA, section 4.3.6). SES modification of NC>2
associations has been examined in fewer studies. For example, in a study conducted in Seoul,
South Korea, community-level SES indicators modified the association of air pollution with ED
visits for asthma. Of the five criteria air pollutants evaluated, NC>2 showed the strongest
association in lower SES districts compared to high SES districts (Kim et al., 2007). In addition,
Clougherty et al. (2007) evaluated exposure to violence (a potential surrogate for SES) as a
modifier of the effect of traffic-related air pollutants, including NO2, on childhood asthma. The
authors reported an elevated risk of asthma with an increase in NO2 exposure solely among
children with above-median exposure to violence in their neighborhoods (ISA, section 4.3.6).
Although these recent studies have evaluated the impact of SES on vulnerability to NC>2, they are
too few in number to draw definitive conclusions (ISA, section 5.3.2.8).
3.8 CONCLUSIONS
The population potentially affected by NO2 is large. A considerable fraction of the
population resides, works, or attends school near major roadways, and these individuals are
likely to have increased exposure to NC>2 (ISA, section 4.4). Based on data from the American
Housing Survey, approximately 36 million individuals live within 300 feet (-90 meters) of a
four-lane highway, railroad, or airport (ISA, section 4.4). Furthermore, in California, 2.3% of
schools with a total enrollment of more than 150,000 students were located within -500 feet of
high-traffic roads, with a higher proportion of non-white and economically disadvantaged
students attending those schools (ISA, section 4.4). Of this population, asthmatics and members
of other susceptible groups discussed above will have even greater risks of health effects related
21
-------
to NO2. In the United States, approximately 10% of adults and 13% of children have been
diagnosed with asthma, and 6% of adults have been diagnosed with COPD (ISA, section 4.4).
The prevalence and severity of asthma is higher among certain ethnic or racial groups such as
Puerto Ricans, American Indians, Alaskan Natives, and African Americans (ISA, section 4.4).
Furthermore, a higher prevalence of asthma among persons of lower SES and an excess burden
of asthma hospitalizations and mortality in minority and inner-city communities have been
observed (ISA, section 4.4). In addition, population groups based on age also comprise
substantial segments of the population that may be potentially at risk for NCVrelated health
impacts. Based on U.S. census data from 2000, about 72.3 million (26%) of the U.S. population
are under 18 years of age, 18.3 million (7.4%) are under 5 years of age, and 35 million (12%) are
65 years of age or older. Hence, large proportions of the U.S. population are in age groups that
are likely to have increased susceptibility and vulnerability for health effects from ambient NC>2
exposure. The considerable size of the population groups at risk indicates that exposure to
ambient NC>2 could have a significant impact on public health in the United States.
22
-------
4. HEALTH EFFECTS
4.1 INTRODUCTION
The ISA, along with its associated annexes, provides a comprehensive review and
assessment of the scientific evidence related to the health effects associated with NC>2 exposures.
For these health effects, the ISA characterizes judgments about causality with a hierarchy (for
discussion see ISA, section 1.3) that contains the following five levels.
• Sufficient to infer a causal relationship
• Sufficient to infer a likely causal relationship (i.e., more likely than not)
• Suggestive but not sufficient to infer a causal relationship
• Inadequate to infer the presence or absence of a causal relationship
• Suggestive of no causal relationship
Judgments about causality are informed by a series of criteria that are based on those set forth by
Sir Austin Bradford Hill in 1965 (ISA, table 1.3-1). These criteria include strength of the
observed association, availability of experimental evidence, consistency of the observed
association, biological plausibility, coherence of the evidence, temporal relationship of the
observed association, and the presence of an exposure-response relationship. A summary of
each of the five levels of the hierarchy is provided in table 1.3-2 of the ISA, which has been
included below (table 4-1).
23
-------
Table 4-1. Weight of Evidence for Causal Determination
Sufficient to infer a
causal relationship
Evidence is sufficient to conclude that there is a causal relationship between
relevant pollutant exposure and the outcome. Causality is supported when an
association has been observed between the pollutant and the outcome in
studies in which chance, bias, and confounding could be ruled out with
reasonable confidence. That is. human clinical studies provide the strongest
evidence for causality. Causality is also supported by findings from
epidemiologic "natural experiments" or observational studies supported by
other lines of evidence. Generally, determination is based on multiple studies
from more than one research group.
Sufficient to infer a
likely causal
relationship (i.e.,
more likely than not).
Evidence is sufficient to conclude that there is a likely causal association
between relevant pollutant exposures and the outcome. That is. an association
has been observed between the pollutant and the outcome in studies in which
chance, bias and confounding are minimized, but uncertainties remain. For
example, observational studies show associations but confounding and other
issues are difficult to address and/or other lines of evidence (human clinical,
animal, or mechanism of action information) are limited or inconsistent.
Generally, determination is based on multiple studies from more than one
research group.
Suggestive, but not
sufficient to infer a
causal relationship
Evidence is suggestive of an association between relevant pollutant exposures
and the outcome, but is weakened because chance, bias and confounding
cannot be ruled out. For example, at least one high-quality study shows an
association, while the results of other studies are inconsistent.
Inadequate to infer the
presence or absence of
a causal relationship
The available studies are inadequate to infer the presence or absence of a
causal relationship. That is. studies are of insufficient quality, consistency or
statistical power to permit a conclusion regarding the presence or absence of
an association between relevant pollutant exposure arid the outcome. For
example, studies which fail to control for confounding or which have
inadequate exposure assessment, fall into this category.
Suggestive of no causal
relationship
The available studies are suggestive of no causal relationship. That is. several
adequate studies, examining relationships between relevant population
exposures and outcomes, and considering sensitive subpopulations. are
mutually consistent in not showing an association between exposure and the
outcome at any level of exposures. In addition, the possibility of a small
elevation in risk at the levels of exposure studied can never be excluded.
The judgments of the ISA, along with the rationale supporting those judgments, are summarized
in tables 4-2 and 4-3 below (ISA, table 5.3-1) and are presented in more detail in subsequent
sections of this chapter.
24
-------
Table 4-2. Causality judgments made in the ISA for endpoints associated with short-term NO2
exposures
HEALTH OUTCOME
CONCLUSION FROM
PREVIOUS NAAQS REVIEW
CONCLUSION FROM 2008 ISA
SHORT-TERM EXPOSURE TO N02
Respiratory Morbidity
Lung Host Defense
Airway Inflammation
Airway Hyperresponsiveness
Respiratory Symptoms
Lung Function
ED Visits / Hospital
Admissions
Cardiovascular Morbidity
Cardiovascular Effects
ED Visits / Hospital
Admissions
Mortality
AH Cause and
Cardiopulmonary Mortality
No Overall Conclusion
Human clinical studies suggest NO2
effects; Animal toxicological studies
indicate that alveolar macrophages
and humoral and cell-mediated
immune systems are affected and
show that exposure can impair the
respiratory host defense system
resulting in susceptibility to infection.
No Studies
An increase in responsiveness to
bronchoconstrictors was found in
asthmatics and healthy individuals
exposed to NO; at rest.
Children living in homes with gas
stoves are at increased risk for
developing respiratory diseases and
illnesses compared to children living
in homes without gas stoves.
Lung function changes in asthmatics
reported at low (0.2 to 0,5 ppm), but
not higher (up to 4 ppm), NO2
concentrations. No convincing
evidence of lung function
decrements in healthy individuals
below 1 ,0 ppm.
No Studies
No Studies
No Studies
No Studies
No Studies
No Studies
"sufficient to infer a likely causal relationship"
Impaired host-defense systems and increased risk of susceptibility to both
viral and bacterial infections after NO; exposures have been observed in
epidemiologic, human clinical, and animal toxicologicai studies.
Human clinical studies report effects of NO; (1-2 ppm) on airway inflammation
in healthy humans. Animal toxicological studies and limited available
epidemiologic studies on children support these findings.
Human dinical studies of allergen and nonspecific bronchial challenges in
asthmatics observed increased airway hyperresponsiveness at near ambient
concentrations (0.1-0.3 ppm). Increased responsiveness to nonspecific
challenges was also observed in animals at higher NO, levels (e.g., 0.5 ppm).
Epidemiologic studies provide consistent evidence of an association of
respiratory effects with indoor and personal NO2 exposures in children.
Multicity studies provide further support for associations between ambient
NO; concentrations (means of 7-70 ppb) and respiratory symptoms in
asthmatic children.
The association between ambient NOj concentrations and lung function in
epidemiologic studies were generally inconsistent. Recent clinical evidence
generally confirms prior findings.
Positive and generally robust associations observed between ambient NO2
levels (means of 3-50 ppb) and increased ED visits and hospital admissions
for respiratory causes, especially asthma.
"inadequate to infer the presence or absence of a causal relationship"
Evidence from epidemiologic studies of heart rate variability, repolarization
changes, and cardiac rhythm disorders among heart patients with ischemic
cardiac disease are inconsistent.
Generally positive associations between ambient NO;, concentrations and
hospital admissions or ED visits for cardiovascular disease; however, the
effects were not robust to adjustment for copollutants.
"suggestive but not sufficient to infer a causal relationship"
Positive and generally robust associations between ambient NO;
concentrations and risk of nonaccidental and Cardiopulmonary mortality.
25
-------
Table 4-3. Causality judgments made in the ISA for endpoints associated with long-term NO2
exposures
HEALTH OUTCOME
CONCLUSION FROM
PREVIOUS NAAQS REVIEW
CONCLUSION FROM 2008 ISA
LONG-TERM EXPOSURE TO N02
Respiratory Morbidity
Respiratory Effects
Other Morbidity
Cancer
Cardiovascular Effects
Birth Outcomes
Mortality
All Cause and
Cardiopulmonary Mortality
No Overall Conclusion
NO2 can cause emphysema
{meeting the human definition
criteria) in animals at high
concentrations of NO2.
No Studies
No Studies
No Studies
No Studies
No Studies
No Studies
"suggestive but not sufficient to infer a causal relationship"
Epidemiologic studies observed decrements in lung function growth
associated with long-term exposure to NO2.
"inadequate to infer the presence or absence of a causal relationship"
Limited epidemiologic studies observed an association between long-term
NO2 exposure and cancer; animal toxicologicai studies have not provided
clear evidence that NO2 acts as a carcinogen.
Very limited epidemiologic and toxicologicai evidence does not suggest that
long-term exposure to NO2 has cardiovascular effects.
The epidemiologic evidence for an association between long-term exposure to
NO2 and birth outcomes is generally inconsistent, with limited support from
animal toxicologicai studies.
"inadequate to infer the presence or absence of a causal relationship"
The results of epidemiologic studies examining the association between long-
term exposure to NO;: and mortality were generally inconsistent-
4.2 ADVERSE RESPIRATORY EFFECTS FOLLOWING SHORT-TERM
EXPOSURES
4.2.1 Overview
The ISA concludes that, taken together, recent studies provide scientific evidence that is
sufficient to infer a likely causal relationship between short-term NC>2 exposure and adverse
effects on the respiratory system (ISA, section 5.3.2.1). This finding is supported by the large
body of recent epidemiologic evidence as well as findings from human and animal experimental
studies. These epidemiologic and experimental studies encompass a number of endpoints
including ED visits and hospitalizations, respiratory symptoms, airway hyperresponsiveness,
airway inflammation, and lung function. Effect estimates from epidemiologic studies conducted
in the United States and Canada generally indicate a 2-20%2 increase in risks for ED visits and
hospital admissions and higher risks for respiratory symptoms (ISA, section 5.4). The findings
2 Effect estimates in the ISA were standardized to a 30 ppb increase in NO2 for studies that evaluated 1-h daily
maximum NO2 concentrations and to a 20 ppb increase for studies that evaluated 24-h average concentrations.
26
-------
relevant to these endpoints, which provide the rationale to support the judgment of a likely causal
relationship, are described in more detail below.
4.2.2 Respiratory Emergency Department Visits and Hospitalizations
Epidemiologic evidence exists for positive associations of short-term ambient NO2
concentrations below the current NAAQS with increased numbers of ED visits and hospital
admissions for respiratory causes, especially asthma (ISA, section 5.3.2.1). Total respiratory
causes for ED visits and hospitalizations typically include asthma, bronchitis and emphysema
(collectively referred to as COPD), pneumonia, upper and lower respiratory infections, and other
minor categories. Temporal associations between ED visits or hospital admissions for respiratory
diseases and ambient levels of NO2 have been the subject of over 50 peer-reviewed research
publications since the review of the NO2 NAAQS that was completed in 1996. These studies
have examined morbidity in different age groups and have often utilized multi-pollutant models
to evaluate potential confounding effects of co-pollutants. Associations are particularly
consistent among children (< 14 years) and older adults (> 65 years) when all respiratory
outcomes are analyzed together (ISA, figures 3.1-8 and 3.1-9) and among children and subjects
of all ages for asthma admissions (ISA, figures 3.1-12 and 3.1-13). When examined with co-
pollutant models, associations of NO2 with respiratory ED visits and hospital admissions were
generally robust and independent of the effects of co-pollutants (ISA, figures 3.1-10 and 3.1-11).
The plausibility and coherence of these effects are supported by experimental (i.e., toxicologic
and controlled human exposure) studies that evaluate host defense and immune system changes,
airway inflammation, and airway responsiveness (see subsequent sections of this document and
ISA, section 5.3.2.1).
Of the ED visit and hospital admission studies reviewed in the ISA, 6 key studies were
conducted in the United States (ISA, table 5.4-1). Of these 6 studies, 4 evaluated associations
with NO2 using multi-pollutant models (Peel et al., 2005 and Tolbert et al., 2007 in Atlanta; New
York Department of Health (NYDOH), 2006 and Ito et al., 2007 in New York City) while 2
studies used only single pollutant models (Linn et al., 2000; Jaffe et al., 2003). In the study by
Peel and colleagues, investigators evaluated ED visits among all ages in Atlanta, GA during the
period of 1993 to 2000. Using single pollutant models, the authors reported a 2.4% (95% CI:
0.9, 4.1) increase in respiratory ED visits associated with a 30-ppb increase in 1-h max NO2
27
-------
concentrations. For asthma visits, a 4.1% (95% CI: 0.8%, 7.6%) increase was estimated in
individuals 2 to 18 years of age. Tolbert and colleagues reanalyzed these data with 4 additional
years of information and found essentially similar results in single pollutant models (2.0%
increase, 95% CI: 0.5, 3.3). This same study found that the associations were positive, but not
statistically-significant, in multi-pollutant models that included PMio or Os. In the study
conducted by the New York Department of Health, investigators evaluated asthma ED visits in
Bronx and Manhattan, New York over the period of January, 1999 to November, 2000. In
Bronx, the authors estimated a 6% (95% CI: 1%-10%) increase in visits per 20 ppb increase in
24-h average concentrations of NO2 and a 7% increase in visits per 30 ppb increase in daily 1-h
maximum concentrations. These effects were not statistically-significant in 2-pollutant models
that included PM2.5 or SO2. In Manhattan, the authors found non-significant decreases (3% for
24-h and a 2% for daily 1-h maximum) in ED visits associated with increasing NC>2. In the study
by Ito and colleagues, investigators evaluated ED visits for asthma in New York City during the
years 1999 to 2002. The authors estimated a 12% (95% CI: 7%, 15%) increase in risk per 20
ppb increase in 24-h ambient NC>2. Risk estimates were robust and remained statistically
significant in multi-pollutant models that included PM2.5, 63, CO, and 862. With regard to the
studies that evaluated only single pollutant models, Linn et al. (2000) detected a statistically-
significant increase in hospital admissions and Jaffee et al. (2003) detected a positive, but
statistically-nonsignificant, increase in ED visits associated with 24-h NC>2 concentrations.
4.2.3 Respiratory Symptoms
Evidence for associations between NC>2 and respiratory symptoms is derived primarily
from the epidemiologic literature, although the experimental evidence for airway inflammation
and immune system effects (described in the ISA, section 3.1 and summarized in subsequent
sections of this document) does provide some plausibility and coherence for the epidemiologic
results (ISA, section 5.3.2.1). Consistent evidence has been observed for an association of
respiratory effects with indoor and personal NO2 exposures in children (ISA, sections 3.1.5.1 and
5.3.2.1) and with ambient levels of NC>2 as measured by community monitors (ISA, sections
3.1.4.2 and 5.3.2.1, see Figure 3.1-6). In the results of multi-pollutant models, NC>2 associations
in multicity studies are generally robust to adjustment for co-pollutants including Os, CO, and
28
-------
(ISA, sections 3.1.4.3, 5.3.2.1 and Figure 3.1-7). Specific studies of respiratory symptoms
are discussed in more detail below.
Studies of Ambient NO 2
Epidemiologic studies using community ambient monitors have found associations
between ambient NO2 concentrations and respiratory symptoms (ISA, sections 3.1.4.2 and
5.3.2.1, Figure 3.1-6) in cities where NC>2 concentrations were within the range of 24-h average
concentrations observed in recent years. Several studies have been published since the 1996
review of the NO2 NAAQS including single-city studies (e.g., Ostro et al., 2001; Delfmo et al.,
2002) and multi-city studies in urban areas covering the continental United States and southern
Ontario (Schwartz et al., 1994; Mortimer et al., 2002; Schildcrout et al., 2006). The multi-city
studies are discussed in more detail below.
Schwartz el at (1994) studied 1,844 schoolchildren, followed for 1 year, as part of the Six
Cities Study that included the cities of Watertown, MA, Baltimore, MD, Kingston-Harriman,
TN, Steubenville, OH, Topeka, KS, and Portage, WI. Respiratory symptoms were recorded
daily. The authors reported a significant association between 4-day mean NO2 levels and
incidence of cough among all children in single-pollutant models, with an odds ratio (OR) of
1.61 (95% CI: 1.08, 2.43) standardized to a 20-ppb increase in NO2. The incidence of cough
increased up to approximately mean NO2 levels (-13 ppb) (p = 0.01), after which no further
increase was observed. The significant association between cough and 4-day mean NO2 level
remained unchanged in models that included O3 but lost statistical significance in two-pollutant
models that included PMio (OR = 1.37 [95% CI: 0.88, 2.13]) or SO2 (OR = 1.42 [95% CI: 0.90,
2.28]).
Mortimer et al. (2002) studied the risk of asthma symptoms among 864 asthmatic
children in New York City, NY, Baltimore, MD, Washington, DC, Cleveland, OH, Detroit, MI,
St Louis, MO, and Chicago, IL. Subjects were followed daily for four 2-week periods over the
course of nine months with morning and evening asthma symptoms and peak flow recorded.
The greatest effect was observed for morning symptoms using a 6-day moving average, with a
reported OR of 1.48 (95% CI: 1.02, 2.16) per 20 ppb increase in NO2. Although the magnitudes
of effect estimates were generally robust in multi-pollutant models that included Os (OR for 20-
ppb increase inNO2 = 1.40 [95% CI: 0.93, 2.09]), O3 and SO2 (ORforNO2 = 1.31 [95% CI:
29
-------
0.87, 2.09]), or O3, SO2, and PMio (OR for NO2 = 1.45 [95% CI: 0.63, 3.34]), they were not
statistically-significant.
Schildcrout et al. (2006) investigated the association between ambient NO2 and
respiratory symptoms and rescue inhaler use as part of the Childhood Asthma Management
Program (CAMP) study. The study reported on 990 asthmatic children living within 50 miles of
an NO2 monitor in Boston, MA, Baltimore, MD, Toronto, ON, St. Louis, MO, Denver, CO,
Albuquerque, NM, or San Diego, CA. Symptoms and use of rescue medication were recorded
daily, resulting in each subject having an average of approximately two months of data. The
authors reported the strongest association between NO2 and increased risk of cough for a 2-day
lag, with an OR of 1.09 (95% CI: 1.03, 1.15) for each 20-ppb increase inNO2 occurring 2 days
before measurement. Multi-pollutant models that included CO, PMi0, or SO2 produced similar
results (ISA, Figure 3.1-5, panel A). Additionally, increased NO2 exposure was associated with
increased use of rescue medication, with the strongest association for a 2-day lag, both for
single- and multi-pollutant models (e.g., for an increase of 20-ppb NO2 in the single-pollutant
model, the RR for increased inhaler usage was 1.05 (95% CI: 1.01, 1.09).
Studies of Indoor M?2
Evidence supporting increased respiratory morbidity following NO2 exposures is also
found in studies of indoor NO2 (ISA, section 3.1.4.1). For example, in a randomized
intervention study in Australia (Pilotto et al., 2004), students attending schools that switched out
unvented gas heaters, a major source of indoor NO2, experienced a decrease in both levels of
NO2 and in respiratory symptoms (e.g., difficulty breathing, chest tightness, and asthma attacks)
compared to students in schools that did not switch out unvented gas heaters (ISA, section
3.1.4.1). An earlier indoor study by Pilotto and colleagues (1997) also found that students in
classrooms with higher levels of NO2 had higher rates of respiratory symptoms (e.g., sore throat,
cold) and absenteeism than students in classrooms with lower levels of NO2. This study detected
a significant concentration-response relationship, strengthening the argument that NO2 is
causally related to respiratory morbidity. A number of other indoor studies conducted in homes
have also detected significant associations between indoor NO2 and respiratory symptoms (ISA,
section 3.1.4.1).
30
-------
4.2.4 Lung Host Defenses and Immunity
Impaired host-defense systems and increased risk of susceptibility to both viral and
bacterial infections after NC>2 exposures have been observed in epidemiologic, controlled human
exposure, and animal toxicological studies (ISA, section 3.1.1 and 5.3.2.1). A recent
epidemiologic study (Chauhan et al., 2003) provides evidence that increased personal exposure
to NC>2 worsened virus-associated symptoms and decreased lung function in children with
asthma. The limited evidence from controlled human exposure studies indicates that NC>2 may
increase susceptibility to injury by subsequent viral challenge at exposures of as low as 0.6 ppm
for 3 hours in healthy adults (Frampton et al., 2002). Toxicological studies have shown that lung
host defenses, including mucociliary clearance and immune cell function, are sensitive to NC>2
exposure, with effects observed at concentrations of less than 1 ppm (ISA, section 3.1.7). When
taken together, epidemiologic and experimental studies linking NC>2 exposure with viral illnesses
provide coherent and consistent evidence that NC>2 exposure can result in lung host defense or
immune system effects (ISA, sections 3.1.7 and 5.3.2.1). This group of outcomes also provides
some plausibility for other respiratory system effects. For example, effects on ciliary action
(clearance) or immune cell function (i.e. macrophage phagocytosis) could be the basis for the
effects observed in epidemiologic studies, including increased respiratory illness or respiratory
symptoms (ISA, section 5.3.2.1). Proposed mechanisms by which NC>2, in conjunction with viral
infections, may exacerbate airway symptoms are summarized in table 4-4 below (ISA, table 3.1-
1).
31
-------
Table 4-4. Proposed Mechanisms Whereby NO2 and Respiratory Virus Infections May Exacerbate
Upper and Lower Airway Symptoms
PROPOSED MECHANISMS
Upper Airway
Epithelium
1 Ciliary beat frequency
* Epithelial permeability
Lower Airway
Epithelium
Cytokines
Inflammatory cells
Inflammatory mediators
Allergens
(as in upper airway)
1 Epithelial-derived IL-8, GM-CSF, TNF-d
* Macrophage-derived IL-lb, IL-6, IL-8, TNF-a
* Mast cell tryptase
* Neutrophils
* Total lymphocytes
" NK lymphocytes
1 T-helper.'T-cytotoxic cell ratio
* Free radicals, proteases, TXA2, TXB2, LTB4
* Penetrance due to ciliostasis
I PD20-FEV,
* Antigen-specific IgE
* Epithelial permeability
Peripheral Blood
i B and NK lymphocytes
1 Total lymphocytes
4.2.5 Airway Response
In acute exacerbations of asthma, bronchial smooth muscle contraction occurs quickly to
narrow the airway in response to exposure to various stimuli including allergens or irritants.
Bronchoconstriction is the dominant physiological event leading to clinical symptoms and
interference with airflow (National Heart, Lung, and Blood Institute, 2007). Inhaled pollutants
such as NC>2 may enhance the inherent responsiveness of the airway to a challenge by allergens
and nonspecific agents (ISA, section 3.1.3). In the laboratory, airway responses can be measured
by assessing changes in pulmonary function (e.g., decline in FEVi) or changes in the
inflammatory response (e.g., using markers in bronchoalveolar lavage (BAL) fluid or induced
sputum) (ISA, section 3.1.3).
The ISA (section 5.3.2.1) draws two broad conclusions regarding the airway response
following NC>2 exposure. First, the ISA concludes that NC>2 exposure may enhance the
32
-------
sensitivity to allergen-induced decrements in lung function and increase the allergen-induced
airway inflammatory response at exposures as low as 0.26 ppm NO2 for 30 minutes (ISA, section
5.3.2.1 and Figure 3.1-2). Second, exposure to NO2 has been found to enhance the inherent
responsiveness of the airway to subsequent nonspecific challenges in controlled human exposure
studies (section 3.1.4.2). In general, small but significant increases in nonspecific airway
responsiveness were observed in the range of 0.2 to 0.3 ppm NO2 for 30-minute exposures and at
0.1 ppm NO2 for 60-minute exposures in asthmatics. These conclusions are consistent with
results from animal toxicological studies which have detected 1) increased immune-mediated
pulmonary inflammation in rats exposed to house dust mite allergen following exposure to 5
ppm NO2 for 3-h and 2) increased responsiveness to non-specific challenges following sub-
chronic (6-12 weeks) exposure to 1 to 4 ppm NO2 (ISA, section 5.3.2.1). Enhanced airway
responsiveness could have important clinical implications for asthmatics since transient increases
in airway responsiveness following NO2 exposure have the potential to increase symptoms and
worsen asthma control (ISA, section 5.4). In addition, the ISA cites the controlled human
exposure literature on the NO2 airway response as being supportive of the epidemiologic
evidence on respiratory morbidity (ISA, section 5.4). Because studies on airway responsiveness
have been used to identify potential health effect benchmark values and to inform the
identification of potential alternative standards for evaluation (see sections 4.5 and 5 of this
document), more detail is provided below on the specific studies that form the basis for the
conclusions in the ISA regarding this endpoint.
Folinsbee (1992) conducted a meta-analysis using individual level data from 19 clinical
NO2 exposure studies measuring airway responsiveness in asthmatics (ISA, section 3.1.3.2).
These studies included NO2 exposure levels between 0.1 ppm and 1.0 ppm and most of them
used nonspecific bronchoconstricting agents such as methacholine, carbachol, histamine, or cold
air. The largest effects were observed for subjects at rest. Among subjects exposed at rest, 76%
experienced increased airway responsiveness following exposure to NO2 levels between 0.2 and
0.3 ppm. Results from an update of this meta-analysis (results combined only from nonspecific
responsiveness studies) are presented in the ISA (Table 3.1-3) and in Table 4-5 below.
33
-------
Table 4-5. Fraction of nitrogen dioxide-exposed asthmatics with increased nonspecific
airway hyperresponsiveness
NO; ppm
0.1
0.1 -0.15
0,2 - 0,3
>0.3
0.1 -0.6
ALL EXPOSURES
0.86 (50)B
0.86 (87 f
G,5S(187)B
0.59(81)
0.60 (355)c
EXPOSURE WITH EXERCISE
—
0.59(17)
0,52(136}
0.49 (48)
0.52(201}
EXPOSURE AT REST
0.66 (50f
0.67 (70)e
0.75 (51 f
0.73 (33)s
0.71 (154)c
As noted in Table 4-5, when exposed at rest 66% of subjects experienced an increase in
airway responsiveness following exposure to 0.1 ppm NO2, 67% of subjects experienced an
increase in airway responsiveness following exposure to NO2 concentrations between 0.1 and
0.15 ppm (inclusively), 75% of subjects experienced an increase in airway responsiveness
following exposure to NO2 concentrations between 0.2 and 0.3 ppm (inclusively), and 73% of
subjects experienced an increase in airway responsiveness following exposure to NO2
concentrations above 0.3 ppm. Effects of NO2 exposure on the direction of airway
responsiveness are statistically-significant at all of these levels. Because this meta-analysis
evaluates only the direction of the change in airway responsiveness, it is not possible to discern
the magnitude of the change from these data. However, the results do suggest that short-term
exposures to NO2 at near-ambient levels (<0.3 ppm) can alter airway responsiveness in people
with mild asthma (ISA, section 3.1.3.2).
Several studies published since the 1996 review address the question of whether low-
level exposures to NO2 enhance the response to specific allergen challenge in mild asthmatics
(ISA, section 3.1.3.1). These recent studies suggest that NO2 may enhance the sensitivity to
allergen-induced decrements in lung function and increase the allergen-induced airway
inflammatory response. Strand et al. (1997) demonstrated that single 30-minute exposures to
0.26-ppm NCh increased the late phase response to allergen challenge 4 hours after exposure, as
measured by changes in lung function. In a separate study (Strand et al., 1998), 4 daily repeated
exposures to 0.26-ppm NO2 for 30 minutes increased both the early and late-phase responses to
allergen, as measured by changes in lung function. Barck et al. (2002) used the same exposure
and challenge protocol in the earlier Strand study (0.26 ppm for 30 min, with allergen challenge
Values are the fraction of asthmatics (out of the total number of individuals in parenthesis)
having an increase in airway responsiveness following NO2 versus air exposure. See table 3.1-3
in the ISA for more detail. B indicates p < 0.05 and c indicates p < 0.01.
34
-------
4 hours after exposure), and performed BAL 19 hours after the allergen challenge to determine
NO2 effects on the allergen-induced inflammatory response. Compared with air followed by
allergen, NO2 followed by allergen caused an increase in the BAL recovery of
polymorphonuclear (PMN) cells and eosinophil cationic protein (ECP) as well as a reduction in
total BAL fluid volume and cell viability. ECP is released by degranulating eosinophils, is toxic
to respiratory epithelial cells, and is thought to play a role in the pathogenesis of airway injury in
asthma. Subsequently, Barck et al. (2005) exposed 18 mild asthmatics to air or 0.26 ppm NO2
for 15 minutes on day 1, followed by two 15 minute exposures separated by 1 hour on day 2,
with allergen challenge after exposures on both days 1 and 2. Sputum was induced before
exposure on day 1 and after exposures (morning of day 3). Compared to air plus allergen, NO2
plus allergen resulted in increased levels of ECP in both sputum and blood and increased
myeloperoxidase levels in blood. All exposures in these studies (Barck et al., 2002, 2005; Strand
et al., 1997, 1998) used subjects at rest. They used an adequate number of subjects, included air
control exposures, randomized exposure order, and separated exposures by at least 2 weeks.
Together, they indicate the possibility for effects on allergen responsiveness in some asthmatics
following brief exposures to 0.26 ppm NO2. However, other recent studies have failed to find
effects using similar, but not identical, approaches (ISA, section 3.1.3.1). The differing findings
may relate in part to differences in timing of the allergen challenge, the use of multiple versus
single-dose allergen challenge, the use of BAL versus sputum induction, exercise versus rest
during exposure, and differences in subject susceptibility (ISA, section 3.1.3.1).
4.2.6 Airway Inflammation
Effects of NO2 on airway inflammation have been observed in controlled human
exposure and animal toxicological studies at higher than ambient levels (0.4-5 ppm). The few
available epidemiologic studies were suggestive of an association between ambient NO2
concentrations and inflammatory response in the airway in children, though the associations
were inconsistent in the adult populations examined (ISA, section 3.1.2 and 5.3.2.1). Controlled
human exposure studies provide evidence for increased airway inflammation at NO2
concentrations of <2.0 ppm. The onset of inflammatory responses in healthy subjects appears to
be between 100 and 200 ppm-minutes, i.e., 1 ppm for 2 to 3 hours (ISA, Figure 3.1-1). Increases
in biological markers of inflammation were not observed consistently in healthy animals at levels
35
-------
of less than 5 ppm; however, increased susceptibility to NC>2 concentrations of as low as 0.4 ppm
was observed when lung vitamin C was reduced (by diet) to levels that were <50% of normal.
These data provide some evidence for biological plausibility and one potential mechanism for
other respiratory effects, such as exacerbation of asthma symptoms and increased ED visits for
asthma (ISA, section 5.3.2.1).
4.2.7 Lung Function
Recent epidemiologic studies that examined the association between ambient NC>2
concentrations and lung function in children and adults generally produced inconsistent results
(ISA, sections 3.1.5.1 and 5.3.2.1). Controlled human exposure studies generally did not find
direct effects of NC>2 on lung function in healthy adults at levels as high as 4.0 ppm (ISA, section
5.3.2.1). For asthmatics, the direct effects of NC>2 on lung function also have been inconsistent at
exposure concentrations of less than 1 ppm NC>2.
4.2.8 Conclusions and Coherence of Evidence for Short-Term Respiratory Effects
As noted previously, the ISA concludes that the findings of epidemiologic, controlled
human exposure, and animal toxicological studies provide evidence that is sufficient to infer a
likely causal relationship for respiratory effects following short-term NC>2 exposure (ISA,
sections 3.1.7 and 5.3.2.1). The ISA (section 5.4) concludes that the strongest evidence for an
association between NC>2 exposure and adverse human health effects comes from epidemiologic
studies of respiratory symptoms, ED visits, and hospital admissions. These studies include panel
and field studies, studies that control for the effects of co-occurring pollutants, and studies
conducted in areas where the whole distribution of ambient 24-h average NC>2 concentrations
was below the current NAAQS level of 0.053 ppm (53 ppb) (annual average). The effect
estimates from the U.S. and Canadian studies generally indicate a 2-20% (see footnote 2 above)
increase in risks for ED visits and hospital admissions. Risks associated with respiratory
symptoms are generally higher (ISA, section 5.4).
Overall, the epidemiologic evidence for respiratory effects can be characterized as
consistent, in that associations are reported in studies conducted in numerous locations with a
variety of methodological approaches. Considering this large body of epidemiologic studies
alone, the findings are also coherent in the sense that the studies report associations with
respiratory health outcomes that are logically linked together. In addition, a number of these
36
-------
associations are statistically-significant, particularly the more precise effect estimates (ISA,
section 5.3.2.1). These epidemiologic studies are supported by evidence from toxicological and
controlled human exposure studies, particularly by controlled human exposure studies that
evaluate airway hyperresponsiveness in asthmatic individuals (ISA, section 5.4). Together, the
epidemiologic and experimental data sets form a plausible, consistent, and coherent description
of a relationship between NC>2 exposures and an array of adverse respiratory health effects that
range from the onset of respiratory symptoms to hospital admission.
However, as noted in the ISA (section 5.4), it is difficult to determine "the extent to
which NC>2 is independently associated with respiratory effects or if NC>2 is a marker for the
effects of another traffic-related pollutant or mix of pollutants." On-read vehicle exhaust
emissions are a nearly ubiquitous source of combustion pollutant mixtures that include NOX and
can be an important contributor to NC>2 levels in near-road locations. Although this complicates
the efforts to quantify specific NO2-related health effects, the evidence summarized in the ISA
indicates that NC>2 associations generally remain robust in multi-pollutant models and supports a
direct effect of short-term NC>2 exposure on respiratory morbidity at ambient concentrations
below the current NAAQS level. The robustness of epidemiologic findings to adjustment for co-
pollutants, coupled with data from animal and human experimental studies, support the
determination that the relationship between NC>2 and respiratory morbidity is likely causal, while
still recognizing the relationship between NC>2 and other traffic related pollutants and the
potential for confounding.
4.3 OTHER ADVERSE EFFECTS FOLLOWING SHORT-TERM
EXPOSURES
The ISA concludes that the epidemiologic evidence is suggestive but not sufficient to
infer a causal relationship between short-term exposure to NC>2 and all-cause and
cardiopulmonary-related mortality (ISA, section 5.3.2.3). Results from several large U.S. and
European multi-city studies and a meta-analysis study indicate positive associations between
ambient NO2 concentrations and the risk of all-cause (nonaccidental) mortality, with effect
estimates ranging from 0.5 to 3.6% excess risk in mortality per standardized increment (20 ppb
for 24-h averaging time, 30 ppb for 1-h averaging time) (ISA, section 3.3.1, Figure 3.3-2, section
5.3.2.3). In general, the NC>2 effect estimates were robust to adjustment for co-pollutants. Both
37
-------
cardiovascular and respiratory mortality have been associated with increased NC>2 concentrations
in epidemiologic studies (ISA, Figure 3.3-3); however, similar associations were observed for
other pollutants, including PM and SCh. The range of risk estimates for excess mortality is
generally smaller than that for other pollutants such as PM. In addition, while NC>2 exposure,
alone or in conjunction with other pollutants, may contribute to increased mortality, evaluation
of the specificity of this effect is difficult. Clinical studies showing hematologic effects and
animal toxicological studies showing biochemical, lung host defense, permeability, and
inflammation changes with short-term exposures to NC>2 provide limited evidence of plausible
pathways by which risks of mortality may be increased, but no coherent picture is evident at this
time (ISA, section 5.3.2.3).
The ISA concludes that the available evidence on cardiovascular health effects following
short-term exposure to NC>2 is inadequate to infer the presence or absence of a causal relationship
at this time (ISA, section 5.3.2.2). Evidence from epidemiologic studies of heart rate variability,
repolarization changes, and cardiac rhythm disorders among heart patients with ischemic cardiac
disease are inconsistent (ISA, section 5.3.2.2). In most studies, associations with PM were found
to be similar or stronger than associations with NC>2. Generally positive associations between
ambient NC>2 concentrations and hospital admissions or ED visits for cardiovascular disease have
been reported in single-pollutant models (ISA, section 5.3.2.2); however, most of these effect
estimate values were diminished in multi-pollutant models that also contained CO and PM
indices (ISA, section 5.3.2.2). Mechanistic evidence of a role for NC^in the development of
cardiovascular diseases from studies of biomarkers of inflammation, cell adhesion, coagulation,
and thrombosis is lacking (ISA, section 5.3.2.2). Furthermore, the effects of NC>2 on various
hematological parameters in animals are inconsistent and, thus, provide little biological
plausibility for effects of NC>2 on the cardiovascular system (ISA, section 5.3.2.2).
4.4 ADVERSE EFFECTS FOLLOWING LONG-TERM EXPOSURES
4.4.1 Respiratory Morbidity
The ISA concludes that overall, the epidemiologic and experimental evidence is
suggestive but not sufficient to infer a causal relationship between long-term NC>2 exposure and
respiratory morbidity (ISA, section 5.3.2.4). The available database evaluating the relationship
between respiratory illness in children and long-term exposures to NC>2 has increased since the
38
-------
1996 review of the NC>2 NAAQS. A number of epidemiologic studies have examined the effects
of long-term exposure to NC>2 and reported positive associations with decrements in lung
function and partially irreversible decrements in lung function growth (ISA, section 3.4.1, figures
3.4-1 and 3.4-2). Specifically, results from the California-based Children's Health Study, which
evaluated NO2 exposures in children over an 8-year period, demonstrated deficits in lung
function growth (Gauderman et al., 2004). This effect has also been observed in Mexico City,
Mexico (Rojas-Martinez et al., 2007a,b) and in Oslo, Norway (Oftedal et al., 2008), with
decrements ranging from 1 to 17.5 ml per 20- ppb increase in annual NC>2 concentration. Similar
associations have been found for PM, 63, and proximity to traffic (<500 m), though these studies
did not report the results of co-pollutant models. The high correlation among traffic-related
pollutants makes it difficult to accurately estimate independent effects in these long-term
exposure studies (ISA, section 5.3.2.4). With regard to asthma incidence and long-term NC>2,
two major cohort studies, the Children's Health Study (Gauderman et al., 2005) and a birth
cohort study in the Netherlands (Brauer et al., 2007), observed significant associations.
However, several other studies failed to find consistent associations between long-term NC>2
exposure and asthma outcomes (ISA, section 5.3.2.4). Similarly, epidemiologic studies
conducted in the United States and Europe have produced inconsistent results regarding an
association between long-term exposure to NC>2 and respiratory symptoms (ISA, sections 3.4.3
and 5.3.2.4). While some positive associations were noted, a large number of symptom
outcomes were examined and the results across specific outcomes were inconsistent (ISA,
section 5.3.2.4).
Animal toxicological studies may provide biological plausibility for the chronic effects of
NC>2 that have been observed in epidemiologic studies (ISA, sections 3.4.5 and 5.3.2.4). The
main biochemical targets of NC>2 exposure appear to be antioxidants, membrane polyunsaturated
fatty acids, and thiol groups. NC>2 effects include changes in oxidant/antioxidant homeostasis
and chemical alterations of lipids and proteins. Lipid peroxidation has been observed atNC>2
exposures as low as 0.04 ppm for 9 months and at exposures of 1.2 ppm for 1 week, suggesting
lower effect thresholds with longer durations of exposure. Other studies showed decreases in
formation of key arachidonic acid metabolites in AMs following NO2 exposures of 0.5 ppm.
NO2 has been shown to increase collagen synthesis rates at concentrations as low as 0.5 ppm.
This could indicate increased total lung collagen, which is associated with pulmonary fibrosis, or
39
-------
increased collagen turnover, which is associated with remodeling of lung connective tissue.
Morphological effects following chronic NO2 exposures have been identified in animal studies
that link to these increases in collagen synthesis and may provide plausibility for the deficits in
lung function growth described in epidemiologic studies (ISA, section 3.4.5).
4.4.2 Mortality
The ISA concludes that the epidemiologic evidence is inadequate to infer the presence or
absence of a causal relationship between long-term exposure to NO2 and mortality (ISA, section
5.3.2.6). In the United States and European cohort studies examining the relationship between
long-term exposure to NO2 and mortality, results have been inconsistent (ISA, section 5.3.2.6).
Further, when associations were suggested, they were not specific to NO2 but also implicated PM
and other traffic indicators. The relatively high correlations reported between NO2 and PM
indices make it difficult to interpret these observed associations at this time (ISA, section
5.3.2.6).
4.4.3 Other Long-Term Effects
The ISA concludes that the available epidemiologic and toxicological evidence is
inadequate to infer the presence or absence of a causal relationship for carcinogenic,
cardiovascular, and reproductive and developmental effects related to long-term NCh exposure
(ISA, section 5.3.2.5). Epidemiologic studies conducted in Europe have shown an association
between long-term NO2 exposure and increased incidence of cancer (ISA, section 5.3.2.5).
However, the animal toxicological studies have provided no clear evidence that NO2 acts as a
carcinogen (ISA, section 5.3.2.5). The very limited epidemiologic and toxicological evidence
does not suggest that long-term exposure to NO2 has cardiovascular effects (ISA, section
5.3.2.5). The epidemiologic evidence is not consistent for associations between NO2 exposure
and fetal growth retardation; however, some evidence is accumulating for effects on preterm
delivery (ISA, section 5.3.2.5). Scant animal evidence supports a weak association between NO2
exposure and adverse birth outcomes and provides little mechanistic information or biological
plausibility for the epidemiologic findings.
40
-------
4.5 RELEVANCE OF SPECIFIC HEALTH EFFECTS TO THE NO2 RISK
CHARACTERIZATION
4.5.1 Overview
As described previously, the ISA characterizes judgments about causality with a hierarchy
(for discussion see ISA, section 1.3) that contains the following five levels.
• Sufficient to infer a causal relationship
• Sufficient to infer a likely causal relationship (i.e., more likely than not)
• Suggestive but not sufficient to infer a causal relationship
• Inadequate to infer the presence or absence of a causal relationship
• Suggestive of no causal relationship
In order to be judged sufficient to infer a causal relationship, an association must have been
observed between the pollutant and the outcome in studies where chance, bias, and confounding
can be ruled out with reasonable confidence. Human clinical studies provide the strongest
evidence for causality while other lines of evidence (e.g., epidemiologic studies) provide support
for this determination. An inference of a causal relationship is generally based on multiple
studies from more than one research group. In order to be judged sufficient to infer a likely
causal relationship, an association must have been observed between the pollutant and the
outcome in studies where chance, bias, and confounding are minimized even though
uncertainties remain. These uncertainties could be due to the difficulty associated with
addressing chance, bias, and confounding and/or due to the fact that other lines of evidence are
limited or inconsistent. An inference of a likely causal relationship is generally based on
multiple studies from more than one research group. In order to be judged suggestive, but not
sufficient to infer a causal relationship, existing evidence must suggest an association between
the pollutant and the outcome but that evidence is weakened because chance, bias, and
confounding cannot be ruled out (see table 4-1). For example, this determination might apply if
at least one high-quality study shows an association, but the results of other studies are
inconsistent (ISA, Table 1.3-2). For purposes of the quantitative characterization of NC>2 health
risks, staff has judged it appropriate to focus on endpoints for which the ISA concludes that the
available evidence is sufficient to infer either a causal or a likely causal relationship. This is
consistent with judgments that have been made in other recent NAAQS reviews (e.g., see EPA,
41
-------
2005) and it will help ensure that our risk characterization is based on endpoints for which a
causal relationship with NC>2 is judged to be more likely than not (see table 4-1 above and ISA,
table 1.3-2).
The only endpoint evidence for which the evidence is judged to be sufficient to infer
either a causal or a likely causal relationship is respiratory morbidity following short-term NC>2
exposure. Therefore, for purposes of characterizing health risks associated with NO2, we have
focused on respiratory morbidity endpoints that have been associated with short-term NC>2
exposures. Other endpoints (e.g., long-term effects) will be considered as part of the evidence-
based evaluation of potential alternative standards during the rulemaking stage of the NAAQS
review. In evaluating the appropriateness of specific endpoints for use in the NC>2 risk
characterization, we have considered both epidemiologic and controlled human exposure studies.
4.5.2 Epidemiology
The ISA characterizes the epidemiologic evidence for respiratory effects as consistent, in
that associations are reported in studies conducted in numerous locations and with a variety of
methodological approaches (ISA, section 5.3.2.1). The findings are also coherent in the sense
that the studies report associations with respiratory health outcomes that are logically linked
together (ISA, section 5.3.2.1). When the epidemiologic literature is considered as a whole,
there are generally positive associations between NC>2 and respiratory symptoms, hospitalization,
and ED visits. A number of these associations are statistically significant, particularly the more
precise effect estimates (ISA, section 5.3.2.1). However, the ISA (section 5.4) offers the
following caveat to consider when interpreting the epidemiologic results: "It is difficult to
determine from these new studies the extent to which NC>2 is independently associated with
respiratory effects or if NC>2 is a marker for the effects of another traffic-related pollutant or mix
of pollutants (see Section 5.2.2 for more details on exposure issues). A factor contributing to
uncertainty in estimating the MVrelated effect from epidemiologic studies is that NC>2 is a
component of a complex air pollution mixture from traffic related sources that include CO and
various forms of PM." These caveats should be considered when interpreting a quantitative NO2
risk estimate based on the epidemiology literature. Despite these uncertainties, the ISA (section
5.4) concludes that, "Although this complicates the efforts to disentangle specific NO2-related
health effects, the evidence summarized in this assessment indicates that NO2 associations
42
-------
generally remain robust in multi-pollutant models and supports a direct effect of short-term NC>2
exposure on respiratory morbidity at ambient concentrations below the current NAAQS. The
robustness of epidemiologic findings to adjustment for copollutants, coupled with data from
animal and human experimental studies, support a determination that the relationship between
NC>2 and respiratory morbidity is likely causal, while still recognizing the relationship between
NC>2 and other traffic related pollutants."
When evaluating epidemiologic studies as to their appropriateness for use as the basis for
a quantitative risk assessment, staff has considered several factors. First, we have judged that
studies conducted in the United States are preferable to those conducted outside the United States
given the potential for effect estimates to be impacted by factors such as the ambient pollutant
mix, the placement of monitors, activity patterns of the population, and characteristics of the
healthcare system. Second, we judged that studies of ambient NC>2 are preferable to those of
indoor NC>2. This does not suggest that indoor studies are uninformative in the review of an
ambient standard. In fact, indoor studies provide a large part of the evidence base used in the
ISA to reach conclusions regarding causality. However, studies of indoor NC>2 focus on
individuals exposed to NC>2 from indoor sources. These indoor sources can result in exposure
patterns, NC>2 levels, and co-pollutants that are different from those typically associated with
ambient NC>2. Because the purpose of a quantitative risk assessment based on the
epidemiological literature would be to inform decisions regarding an ambient NC>2 standard, the
preferred approach would be to consider studies of ambient NO2. Third, we judged it
appropriate to focus on studies of ED visits and hospital admissions. When compared to studies
of respiratory symptoms, the public health significance of ED visits and hospital admissions are
less ambiguous (e.g., because of the potential disconnect between health outcomes and
subjective symptom ratings). In addition, baseline incidence data are more readily available for
these endpoints. Finally, we judged it appropriate to focus on studies that evaluated NC>2 health
effect associations using both single- and multi-pollutant models. Taking these factors into
consideration, we have chosen to focus on the studies by Peel and colleagues (2005) and by
Tolbert and colleagues (2007) in Atlanta, Georgia. The epidemiology-based risk assessment is
described in more detail in subsequent sections of this document.
43
-------
4.5.3 Controlled Human Exposure Studies
Controlled human exposure studies have addressed the consequences of short-term (e.g.,
30-minutes to several hours) NC>2 exposures for a number of health endpoints including airway
responsiveness, host defense and immunity, inflammation, and lung function (ISA, section 3.1).
In identifying health endpoints from controlled human exposure studies on which to focus the
characterization of NC>2 health risks, staff judges it appropriate to focus on endpoints that occur
at or near ambient levels of NC>2 and endpoints that are of clinical significance. With regard to
the NC>2 levels at which different effects have been documented, the ISA concludes that 1) in
asthmatics NO2 may increase the allergen-induced airway inflammatory response at exposures as
low as 0.26-ppm for 30 min (ISA, Figure 3.1-2) and NC>2 exposures between 0.2 and 0.3 ppm for
30 minutes or 0.1 ppm for 60-minutes can result in small but significant increases in nonspecific
airway responsiveness (ISA, section 5.3.2.1); 2) limited evidence indicates thatNO2 may
increase susceptibility to injury by subsequent viral challenge following exposures of 0.6-1.5
ppm for 3 hours; 3) evidence exists for increased airway inflammation atNC>2 concentrations less
than 2.0 ppm; and 4) the direct effects of NC>2 on lung function in asthmatics have been
inconsistent at exposure concentrations below 1 ppm (ISA, section 5.3.2.1). The ISA notes that
epidemiologic studies have reported health effects associations in areas reporting maximum
ambient concentrations from 100 to 300 ppb (ISA, Tables 5.3-2 and 5.3-3). Therefore, of the
health effects caused by NC>2 in controlled human exposure studies, the only effect identified by
the ISA to occur at or near ambient levels is increased airway responsiveness in asthmatics.
Staff judges that airway responsiveness in the asthmatic population is an appropriate
focus for the risk characterization for several reasons. First, the ISA concludes that "persons
with preexisting pulmonary conditions are likely at greater risk from ambient NC>2 exposures
than the general public, with the most extensive evidence available for asthmatics as a potentially
susceptible group" (ISA, section 5.3.2.8). Second, when discussing the clinical significance of
NO2-related airway hyperresponsiveness in asthmatics, the ISA concludes that "transient
increases in airway responsiveness following NC>2 exposure have the potential to increase
symptoms and worsen asthma control" (ISA, sections 3.1.3 and 5.4). That this effect could have
public health implications is suggested by the large size of the asthmatic population in the United
States (see above and ISA, Table 4.4-1). Third, NC>2 effects on airway responsiveness in
asthmatics are part of the body of experimental evidence that provides plausibility and coherence
44
-------
for the effects observed on hospital admissions and ED visits in epidemiologic studies (ISA,
section 5.3.2.1). As a result of these considerations, although studies on other endpoints
evaluated in controlled human exposure studies provide qualitative support for the ability of NO2
to cause adverse effects on respiratory health, the focus for purpose of quantifying risks
associated with ambient NO2 is airway responsiveness in asthmatics (see below).
Because many of the studies of airway responsiveness evaluate only a single level of NO2
and because of methodological differences between the studies, staff has judged that the data are
not sufficient to derive an exposure-response relationship in the range of interest. Therefore, the
most appropriate approach to characterizing risks based on the controlled human exposure
evidence for airway responsiveness is to compare estimated NO2 air quality and exposure levels
with potential health effect benchmark levels. Estimates of hourly peak air quality
concentrations and personal exposures to ambient NO2 concentrations at and above specified
potential health effect benchmark levels provide some perspective on the potential health impacts
of NO2 exposure. Staff recognizes that there is high inter-individual variability in NO2-induced
effects on airway responsiveness such that only a subset of asthmatic individuals exposed at and
above a given benchmark level may actually be expected to experience an adverse effect.
Potential health benchmark levels and the approach to using these benchmarks to characterize
health risks are described in more detail in chapter 6.
45
-------
5. IDENTIFICATION OF POTENTIAL ALTERNATIVE
STANDARDS FOR ANALYSIS
5.1 INTRODUCTION
The primary goals of the NO2 risk and exposure assessment described in this document
are to estimate short-term exposures and potential human health risks associated with 1) recent
levels of ambient NO2; 2) NO2 levels associated with just meeting the current standard; and 3)
NO2 levels associated with just meeting potential alternative standards. This section identifies
potential alternative standards in terms of indicator, averaging time, form, and level and provides
the rationale that was used to select them.
5.2 INDICATOR
The NOX, for purposes of this document, include multiple gaseous (e.g., NO2, NO) and
particulate (e.g., nitrate) species. In considering the appropriateness of different indicators, we
note that the health effects associated with particulate species of NOX have been considered
within the context of the health effects of ambient particles in the Agency's review of the
NAAQS for PM. Thus, as discussed in the integrated review plan (2007a), the current review of
the NO2 NAAQS is focused on the gaseous species of NOX and will not consider health effects
directly associated with particulate species of NOX. Of the gaseous species, EPA has historically
determined it appropriate to specify the indicator of the standard in terms of NO2 because the
majority of the information regarding health effects and exposures is for NO2. The final ISA has
found that this continues to be the case and, therefore, staff believes that NO2 remains the most
appropriate indicator.
5.3 AVERAGING TIME
The current annual standard for NO2 was originally set in 1971 based on epidemiologic
studies that supported a link between adverse respiratory effects and long-term exposure to low-
levels of NO2. Although the quantitative basis for the annual averaging time was later called into
question (60 FR 52876), the annual standard was retained in the most recent review (60 FR
52876) for two key reasons. First, the evidence showing the most serious health effects
associated with long-term exposures (e.g., emphysematous-like alterations in the lung and
46
-------
increased susceptibility to infection) came from animal studies conducted at concentrations well
above those permitted in the ambient air by the annual standard. Second, an air quality
assessment conducted by EPA concluded that areas that meet the annual standard would be
unlikely to experience short-term peaks above levels that had been shown in controlled human
exposure studies to impact endpoints of potential concern (i.e., airway responsiveness).
The issue of averaging time will be reconsidered in the current review. As described
above, the ISA concludes that, when taken together, "recent studies provide scientific evidence
that NC>2 is associated with a range of respiratory effects and is sufficient to infer a likely causal
relationship between short-term NC>2 exposure and adverse effects on the respiratory system"
(ISA, section 5.3.2.1). This conclusion is based, in part, on the observation that a number of
epidemiologic studies have detected positive associations between short-term (e.g., 1-h, 24-h)
NC>2 concentrations and health effects. Many of these studies have been conducted in locations
where long-term ambient levels of NCh are well below the current annual standard. As a result,
staff has concluded that it is appropriate to consider alternative averaging times for their ability
to protect against health effects associated with short-term NC>2 levels and/or exposures.
In contrast to the conclusion in the ISA concerning respiratory morbidity associated with
short-term exposures to NC>2, the ISA concludes that the "evidence examining the effect of long-
term exposure to NC>2 on respiratory morbidity is suggestive but not sufficient to infer a causal
relationship" (ISA, section 5.3.2.4). In addition, the ISA concludes that the available evidence
for the effect of long-term exposure to NO2 on other health outcomes (i.e., mortality, cancer,
cardiovascular effects, reproductive and developmental effects) is "inadequate to infer the
presence or absence of a causal relationship" (ISA, sections 5.3.2.5 and 5.3.2.6). As a result,
staff has not considered alternative long-term standards in the current assessment.
In considering appropriate short-term averaging times, staff has considered evidence from
both experimental and epidemiologic studies. New evidence from controlled human exposure
studies generally evaluates exposures between 30 minutes and 3 hours while epidemiologic
studies have used different short-term averaging periods, most commonly 1-h and 24-h (ISA,
section 3.1). A few epidemiologic studies have considered both 1-h and 24-h averaging times,
allowing comparisons to be made. The ISA reports that such comparisons failed to reveal
differences between effect estimates based on a 1-h averaging time versus those based on a 24-h
averaging time (ISA, section 5.3.2.7). Therefore, the ISA concludes that it is not possible to
47
-------
discern whether effects observed in epidemiologic studies are attributable to average daily (or
multiday) concentrations (24-h avg) or high, peak exposures (1-h max) (ISA, section 5.3.2.7). In
addition, the ISA concludes that experimental studies in both animals and humans provide
evidence that NC>2 exposures from less than 1 hour up to 3 hours can result in respiratory effects
(section 5.3.2.7). Given that the epidemiologic evidence does not provide clear guidance in
choosing between 1-h and 24-h averaging times, and given that the experimental literature
provides support for the occurrence of effects following exposures of shorter duration than 24
hours (e.g., 1-h), staff has chosen to evaluate standards with a 1-h averaging time.
5.4 FORM
In evaluating alternative forms for the primary standard, staff recognizes that it is important
to have a form that 1) reflects the health risks posed by elevated NC>2 concentrations and 2)
achieves a balance between limiting the occurrence of peak concentrations and providing a stable
and robust regulatory target. Consistent with judgments made in recent reviews of the PM (71
FR 61144) and O3 (73 FR 16436) NAAQS, staff judges that a concentration-based form
averaged over 3 years for the NC>2 standard would better reflect health risks and would provide
greater stability than a form based on expected exceedances. A concentration-based form would
give proportionally greater weight to hours when NC>2 concentrations are well above the level of
the standard than to hours when concentrations are just above the standard, while an expected
exceedance form would give the same weight to hours that just exceed the standard as to hours
that greatly exceed the standard. Therefore, a concentration-based form averaged over 3 years
better reflects the health risks posed by elevated NC>2 concentrations and, in developing potential
alternative standards for consideration, we have focused on standards with this concentration-
based form. The most recent review of the PM NAAQS (completed in 2006) judged that using a
98th percentile form averaged over 3 years provides an appropriate balance between limiting the
occurrence of peak concentrations and providing a stable regulatory target (71 FR 61144). In
consideration of this balance, we have determined it appropriate in the current review to evaluate
both the 98th and 99th percentile NC>2 concentrations averaged over 3 years.4 We have judged
that these percentiles, when combined with the range of alternatives identified for the level of the
4 98th or 99th percentiles of the 1-h daily maximum NO2 concentrations would be calculated for each of 3
consecutive years. The 98th or 99th percentile concentrations for each of these 3 years would then be averaged
together.
48
-------
standard (see below), offer a sufficient range of options to balance the objective of providing a
stable regulatory target against the objective of limiting the occurrence of peak concentrations.
5.5 LEVEL
In developing an approach to formulating an appropriate range of NO2 levels for analysis,
staff has taken into account several considerations including the following. First, since the
review of the NO2 NAAQS that was completed in 1996, a large number of published
epidemiologic studies have evaluated associations between respiratory morbidity and short-term
levels of ambient NO2. In general, these studies report positive associations and a number of
these associations are statistically-significant. The ISA notes that many of these studies have
been conducted in locations where ambient levels of NO2 are well below the level of the current
NAAQS (ISA, section 5.3.2.1). Second, controlled human exposure studies have detected
effects of NO2 exposure on several health endpoints. Of these, only airway hyperresponsiveness
is associated with exposures to NO2 concentrations at or near ambient levels. In fact, the NO2
exposure levels associated with increased airway responsiveness overlap the maximum ambient
NO2 concentrations in some locations where associations with respiratory effects have been
detected. Third, limitations in both epidemiologic studies (e.g., confounding by co-pollutants)
and controlled human exposure studies (e.g., most sensitive populations likely not evaluated)
suggest that an appropriate approach to identifying levels for potential alternative standards is to
consider both types of studies.
In considering both types of studies, we note that NO2 concentrations represent different
metrics when reported in epidemiologic studies versus controlled human exposure studies.
Concentrations of NO2 reported in epidemiologic studies are typically based on ambient
monitoring data while NO2 levels reported in controlled human exposure studies represent the
concentration of NO2 in the breathing zone of the individual. Therefore, consideration of NO2
levels from controlled human exposure studies when identifying alternative levels for an ambient
standard introduces some uncertainty. For example, elevated NO2 monitors, particularly in inner
cities, likely underestimate personal exposures that occur at lower elevations closer to traffic
(ISA, section 5.2.2). In situations where personal exposure to ambient NO2 is higher than
ambient levels measured at a monitor, ambient standard levels based on controlled exposure
studies could be less health-protective than levels based on concentrations reported in
49
-------
epidemiologic studies at ambient monitors. However, in studies where it has been determined,
the ratio of personal exposure to NO2 of ambient origin to the ambient NC>2 concentration ranges
from approximately 0.3 to 0.6 (ISA, section 5.2.2). This suggests that in some cases personal
exposure concentrations for ambient NC>2 are lower than the levels of NO2 measured at ambient
monitors. In situations where personal exposure concentrations for ambient NC>2 are lower than
the levels measured at ambient monitors, an ambient standard level based on controlled exposure
studies could be more health-protective than a level based on concentrations reported in
epidemiologic studies at ambient monitors. Overall, because some individuals are likely exposed
to NC>2 levels higher than those measured at ambient monitors (ISA, section 5.2.2) while others
are likely exposed to NC>2 levels lower than those measured at ambient monitors (see ISA, tables
2.5-4 and 2.5-5), we have chosen to use NO2 concentrations associated with both epidemiologic
studies and controlled human exposure studies for purposes of selecting alternative levels for
analysis.
As a result of the above considerations, to determine the levels that should be evaluated
staff has relied on both key epidemiologic studies conducted in the United States that evaluate
associations between short-term levels of NC>2 and respiratory morbidity (symptoms, hospital
admissions, ED visits) and on controlled human exposure studies that evaluate airway
hyperresponsiveness following NO2 exposure. Figures 5-1 and 5-2 below show standardized
effect estimates5 and the 98th and 99th percentile concentrations of daily 1-h maximum NC>2 for
locations and time periods that correspond to key U.S. epidemiologic studies identified in the
ISA (see table 5.4-1 in ISA for a list of key studies; Thompson and Jenkins, 2008).
Of the key U.S. epidemiologic studies included in figures 5-1 and 5-2, the highest 1-h
NO2 concentrations were detected in the two studies conducted in Los Angeles (Linn et al.,
2000; Ostro et al., 2001). For these studies, the 98th and 99th percentile 1-h daily maximum
concentrations of NO2 overlap levels that the ISA concludes are associated with increased airway
responsiveness in controlled human exposure studies (ISA, section 5.3.2.1). Therefore, staff
judges that the combination of the epidemiologic studies by Linn et al. (2000) and Ostro et al.
(2001), as well as the meta-analysis (Folinsbee, 1992; ISA, table 3.1-3; table 4-2 of this
document) of controlled human exposure studies on airway responsiveness, provide an
5 The effect estimates presented in figures 5-1 and 5-2 are for those endpoints included in figure 5.3-1 and table 5.4-
1 of the IS A.
50
-------
appropriate basis for identifying the upper end of the range of standard levels to be considered.
Given that the ISA concludes that significant increases in airway responsiveness are associated
with short-term exposures to NO2 at 0.2 to 0.3 ppm and given that the epidemiologic studies by
Linn et al. (2000) and Ostro et al. (2001) are associated with 98th and 99th percentile 1-h daily
maximum NO2 levels that are just below (Linn et al., 2000) and just above (99th percentile level
for Ostro et al., 2001) 0.2 ppm (see figures 1 and 2 below), staff judges that an appropriate upper
end of the range of potential standard levels is a daily maximum 1-h NO2 concentration of 0.20
ppm.
51
-------
15
10
1 5
as
in
8 o
LLI
C
d> r-
a -5
Q-
-1O
-15
-.U
1 -hour effect estimates
^
EC
EDA
EDAC
EDR
r
1-h 98: 0.085
1-h 99: 0.093
V
Tolbert 2007
Atlanta
Legend:
EDR E9A
\
V_
I
1-h 99: 0,095
Y
Peel 2005
Atlanta
J
Bronx
.
1-h 98: 0.088
L
~~V
A
EDA^
1-h 98 0,086
1-h 99: 0.098
Manhattan
1-h 96:0.086
1-h 99: 0,093
_J
NYDOH 2006
NYC
EDR = Emergency department visits for respiratory disease
EDA = Emergency department visits for asthma
EDAC = Emergency department visits for asthma - children
HAA = Hospital admissions for asthma
V
Jaffe 2003
Clev/Cinn
4
DA
1-h 98: 0.094
1-h 99:0
112
V
Ito 2007
NYC
24-hour effect estimates
1-h 98: 0.178
1-h 99: 0.197
\
HAA
V
E[
)A
1-h 98: 0.088
Linn 2000
LA
EDA
Manhattan
1-h 98- n rwfi
1-h 99:
0.093
1
I
y
_J
NYDOH 2006
NYC
Figure 5-1. NO2 effect estimates6 (95% CI) for ED visits/HA and associated 1-h daily
maximum NOi levels (98th and 99th percentile values in boxes7)
6Effect estimates presented in figures 5-1 and 5-2 are from single pollutant models only. The studies by Tolbert et
al, (2007); Peel et al., (2005); NYDOH (2006); Ito et al., (2007); and Delfino et al. (2002) also evaluated multi-
pollutant models. NO2 effect estimates retained statistical-significance in the study by Ito, but not in the other
studies.
7 Authors of relevant U.S. and Canadian studies were contacted and air quality statistics from the study monitor that
recorded the highest NO2 levels were requested. In cases where authors provided 1-hour daily maximum air quality
statistics, this information is presented in figures 1 and 2 (studies by Tolbert, Peel, NYDOH, Delfino). In one case
(study by Ito) authors provided 24-hour air quality data, but identified a specific monitor in AQS. We used AQS to
reconstruct the 1-hour daily maximum air quality for that monitor during the time period of the study. In three cases
(studies by Jaffe, Linn, Ostro), we were not able to identify appropriate statistics from the information provided by
the authors and the authors did not provide monitor identification information. In these cases, we attempted to
reconstruct the air quality data set for the location and time of the study using EPA's Air Quality System (AQS).
Prior to identifying potential alternative standards, we did not receive air quality information from any of the
Canadian authors contacted and we were unable to reconstruct the air quality data sets for the Canadian studies.
Therefore, for purposes of identifying levels of potential alternative standards, our analysis was based on these key
U.S. studies. Note that the NO2 concentrations reported in table 1 of the study by Jaffe are labeled as 24-hour
concentrations, but the author indicated in a personal communication (Jaffe, 2008) that they actually represent 1-
hour daily maximum concentrations.
52
-------
1 -hour effect estimates
4-hour effect estimate
24-hour effect estimates
130 -
110 -
90 -
| 70-
o> 50 -
u
X
uj 30 .
•*•*
c
o
S 10 -
Q.
-10 -
-30 -
1-h98<": 0.050
1-h 99*: 0.053
i i
] 1
1-h 98'"; 0,180
1-h 99'": 0.210
*
j
w c
AS
AS
k
* -i
1 1
t
*
*
i i i i r i i
1 \ i 1 1 1 1 1
AS C AS
MS
V
Delfino2002
Alpine, CA
Ostro2001 Delfino 2003 Mortimer 2002 Schwartz 1994 Sohildcrout 2006 Linn 1996
LA and Pasadena East LA County 8 Cities 6 Cities 8 Cities LA
Legend:
AS = Asthma symptoms
W = Wheeze
C = Cough
MS = Morning symptoms
*We do not have 1 -h 98lh and 99lh percentile NO2 levels for several of the U.S. respiratory symptom studies identified
in table 5.4-1 of the ISA. Comparison of averages (see ISA, table 5.4-1) suggests that 24-h NO2 levels in the studies
by studies by Schildcrout and Schwartz are somewhat lower than the 24-hour levels reported in other U.S. studies,
24-h levels in the study by Linn are similar to 24-h levels reported in other U.S. studies, and 1-h maximum levels in
the study by Delfino are lower than 1-h maximum levels reported in other U.S. studies. Such comparisons have not
been made for the study by Mortimer because it is the only study that reports 4-hour NO2 levels.
Figure 5-2. NOi effect estimates for respiratory symptoms and associated 1-h daily
maximum NOi levels (98th and 99th percentile values in boxes)
In identifying additional standard levels that should be analyzed, staff has considered that
1) health effect associations in epidemiologic studies are observed in locations with 1-h daily
maximum levels of NC>2 below 0.2 ppm (i.e., 99th percentile levels in several studies are close to
0.1 ppm); 2) controlled human exposure studies that evaluate the ability of NC>2 to elicit airway
hyperresponsiveness have assessed mild asthmatics and more severely affected asthmatics could
experience increased airway responsiveness at lower levels of NC>2 than observed in these
53
-------
studies; and 3) a meta-analysis presented in the ISA (see Table 4-2) detects statistically-
significant effects on the direction of airway responsiveness following short-term NC>2 exposures
as low as 0.1 ppm. As a result of these considerations, staff judges that it would be appropriate
to consider additional standard levels that provide a margin of safety relative to 0.20 ppm.
Therefore, we will also consider daily maximum 1-h NC>2 standard levels of 0.10 ppm and 0.15
ppm.
In identifying the lower end of the range of standards that will be analyzed, staff has
considered the fact that the study by Delfmo et al., (2002) provides evidence for associations
between short-term ambient NC>2 concentrations and respiratory morbidity in a location where
the 98th and 99th percentile concentrations of the 1-h daily maximum levels of NC>2 were well
below 0.1 ppm (Delfmo et al., 2002). This study detects associations between 1-h and 8-h (only
8-h associations were statistically-significant) levels of NC>2 and asthma symptoms in a location
where the 98th and 99th percentile 1-h daily maximum NC>2 concentrations were 0.050 and 0.053
ppm, respectively. The 8-h effect estimate in this study remained positive, but became
statistically non-significant, in a two-pollutant model that also included PMi0. Staff judges that it
is appropriate to base the lower end of the range of alternative standard levels on this study by
Delfmo et al. (2002). Therefore, we will also consider a 1-h daily maximum standard level of
0.050 ppm.
54
-------
6. OVERVIEW OF APPROACHES TO ASSESSING EXPOSURES
AND RISKS
6.1 INTRODUCTION
The purpose of the assessments described in this document is to characterize exposures
and risks associated with recent ambient levels of NC>2, with levels associated with just meeting
the current NC>2 NAAQS, and with levels associated with just meeting potential alternative
standards (see chapter 5 of this document for discussion of potential alternative standards). To
characterize health risks, we have employed three approaches. With each approach, we have
characterized health risks associated with the air quality scenarios of interest (i.e., recent air
quality unadjusted, air quality adjusted to simulate just meeting the current standard, and air
quality adjusted to simulate just meeting potential alternative standards). In the first approach,
NC>2 air quality levels have been compared to potential health effect benchmark values derived
from the controlled human exposure literature (see section 6.2 below for discussion of
benchmark levels). In the second approach, modeled estimates of actual exposures have been
compared to potential health effect benchmarks. In the third approach, exposure-response
relationships from epidemiologic studies have been used to estimate health impacts. An
overview of the approaches to characterizing health risks is provided below and each approach is
described in more detail in subsequent sections of this document and the associated appendices.
In the first approach, we have compared NC>2 air quality with potential health effect
benchmark levels for NC>2. Scenario-driven air quality analyses have been performed using
ambient NC>2 concentrations for the years 1995 though 2006. With this approach, NC>2 air
quality serves as a surrogate for exposure. All U.S. monitoring sites where NC>2 data have been
collected are represented by this analysis and, as such, the results generated are considered a
broad characterization of national air quality and human exposures that might be associated with
these concentrations. An advantage of this approach is its relative simplicity; however, there is
uncertainty associated with the assumption that NO2 air quality can serve as an adequate
surrogate for exposure to ambient NC>2. Actual exposures might be influenced by factors not
considered by this approach, such as the spatial and temporal variability in human activities.
55
-------
In the second approach, we have used an inhalation exposure model to generate more
realistic estimates of personal exposures. Estimates of personal exposure have been compared to
potential NO2 health benchmark levels. For this exposure analysis, a probabilistic approach was
used to model individual exposures considering the time people spend in different
microenvironments and the variable NO2 concentrations that occur within these
microenvironments across time, space, and microenvironment type. This approach to assessing
exposures was more resource intensive than using ambient levels as a surrogate for exposure;
therefore, staff has included the analysis of only one specific location in the U.S. (Atlanta
MSA)8. Although the geographic scope of this analysis is restricted, the approach provides
realistic estimates of NO2 exposures, particularly those exposures associated with important
emission sources of NOX and NO2, and serves to complement the broad air quality
characterization.
For the characterization of risks in both the air quality analysis and the exposure
modeling analysis described above, staff has used a range of short-term potential health effect
benchmarks. The levels of potential benchmarks are based on NO2 exposure levels that have
been associated with increased airway responsiveness in asthmatics in controlled human
exposure studies (ISA, section 5.3.2.1; see above for discussion). Benchmark values of 100,
150, 200, 250, and 300 ppb have been compared to both NO2 air quality levels and to estimates
of NO2 exposure. When NO2 air quality is used as a surrogate for exposure, the output of the
analysis is an estimate of the number of times per year specific locations experience 1-h levels of
NO2 that exceed a particular benchmark. When personal exposures are simulated, the output of
the analysis is an estimate of the number of individuals at risk for experiencing daily maximum
1-h levels of NO2 of ambient origin that exceed a particular benchmark. An advantage of using
potential health effect benchmark levels to characterize health risks is that the effects observed in
controlled human exposure studies clearly result from NO2 exposure. This is in contrast to
health effects associated with NO2 in epidemiologic studies, which may also be associated with
pollutants that co-occur with NO2 in the ambient air. Thus, when using epidemiologic studies as
the basis for risk characterization, the unique contribution of NO2 to a particular health effect
In the document titled Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient
Air Quality Standard: First Draft, we have presented the results of an exposure analysis for Philadelphia. Based on
CASAC comments received on that exposure analysis, we have refined our approach and applied those refinements
to the Atlanta analysis presented in this document. The original Philadelphia analysis is presented in the appendix to
this document, but has not been modified since the first draft.
56
-------
may be difficult to quantify. A disadvantage of the potential benchmark approach is that the
magnitude of the NC>2 effect on airway responsiveness can vary considerably from individual to
individual and not all asthmatics would be expected to respond to the same levels of NO2
exposure. Therefore, the public health impacts of NO2-induced airway hyperresponsiveness are
difficult to quantify.
In the third approach, we have estimated respiratory ED visits as a function of ambient
levels of NC>2 measured at a fixed-site monitor representing ambient air quality for an urban area.
In this approach, concentration-response functions are derived from NC>2 epidemiologic studies
and are used to estimate the impact of ambient levels of NC>2, as measured at a fixed-site
monitor, on ED visits. By focusing on a different health endpoint from the first two approaches
described above, this epidemiology-based approach provides additional perspective on the
potential public health impacts of NC>2. Relative to the approaches that use controlled human
exposure studies, this approach to characterizing health risks has several advantages. For
example, the public health significance of the effect in question (i.e., ED visits) is less
ambiguous in terms of its impact on individuals than is an increase in the airway response
measured in a controlled human exposure study. In addition, the concentration-response
relationship reflects real-world levels of NC>2 and co-pollutants present in ambient air. However,
a disadvantage of this approach is the ambiguity and complexity associated with quantifying the
contribution of NC>2 to the reported health impacts relative to the contributions of co-occurring
pollutants.
6.2 POTENTIAL HEALTH BENCHMARK LEVELS
As noted above (section 4.5.3 and 6.1), staff has developed potential health benchmark
levels that are based on results from controlled human exposure studies of airway
responsiveness. The purpose of these potential health benchmark levels is to provide a basis for
comparing NC>2 concentrations shown to increase airway responsiveness in asthmatics (from
controlled human exposure studies) to estimates of NC>2 exposures derived from our air quality
and exposure analyses. Because the purpose of the benchmarks is to provide a way to link
estimates of NC>2 exposure with levels known to produce respiratory effects in individual
asthmatics, benchmark levels have not been developed from the epidemiologic literature.
57
-------
To identify potential health effect benchmarks, staff has relied on the ISA's evaluation of
the NO2 human exposures studies. Controlled human exposure studies involving allergen
challenge in asthmatics suggest that NO2 exposure may enhance the sensitivity to allergen-
induced decrements in lung function and increase the allergen-induced airway inflammatory
response at exposures as low as 0.26-ppm NO2for 30 min (ISA, Figure 3.1-2 and section
5.3.2.1). Exposure to NO2 also has been found to enhance the inherent responsiveness of the
airway to subsequent nonspecific challenges (ISA, section 5.3.2.1). In asthmatics, small but
significant increases in nonspecific airway responsiveness have been observed in the range of 0.2
to 0.3 ppm NO2 for 30 minute exposures and at 0.1 ppm NO2 for 1-h exposures (ISA, section
5.3.2.1). Therefore, for the risk characterization, staff judges that 1-h NO2 levels in this range
are appropriate to consider as potential health benchmarks for comparison to air quality levels
and exposure estimates. To characterize health risks with respect to this range, potential health
effect benchmark values of 0.10 ppm, 0.20 ppm, 0.25 ppm, and 0.30 ppm have been employed to
reflect the lower- middle- and upper-end of the range identified in the ISA as levels at which
controlled human exposure studies have provided evidence for the occurrence of NO2-induced
airway hyperresponsiveness.
In choosing this range, we recognize that uncertainties exist regarding the percentage of
asthmatics expected to experience an increase in responsiveness following NO2 exposure and in
the clinical implications of such an increase. A meta-analysis presented in the ISA (see Table 4-
2 above) suggests that between 66% and 75% of asthmatics may experience an increase in
airway responsiveness following short-term NO2 exposures in the range of 0.1 to 0.3 ppm.
However, this meta-analysis provides information only on the direction of the NO2 effect and not
on its magnitude. In addition, the NO2 controlled human exposure studies of airway
responsiveness have focused primarily on mild asthmatics. It is possible that more severely
affected asthmatics could experience a more severe response following NO2 exposures in this
range. It is also possible that they could experience a response at lower levels of NO2 than the
mild asthmatics included in these studies. However, even considering these uncertainties, staff
judges that the identified range of concentrations is sufficient to provide some perspective on the
potential public health impacts of NO2 exposures, especially when the results of the risk
characterization based on airway responsiveness are considered in conjunction with the risk
assessment based on the epidemiology literature.
58
-------
6.3 SIMULATING THE CURRENT AND ALTERNATIVE STANDARDS
A primary goal of these risk and exposure assessments is to evaluate the ability of the
current NO2 standard (0.053 ppm annual average) and potential alternative standards (0.05, 0.10,
0.15, and 0.20 ppm; see chapter 5 of this document) to protect public health. In order to evaluate
the ability of a specific standard to protect public health, ambient NC>2 concentrations need to be
adjusted such that they simulate levels of NC>2 that just meet that standard. Such adjustments
allow comparisons of the level of public health protection that could be associated with just
meeting the current and potential alternative standards. All areas of the United States currently
have ambient NC>2 levels below the current annual standard. Therefore, to simulate just meeting
the current annual standard, NC>2 air quality levels must be adjusted upward. Similarly, to
simulate a potential standard that is below current air quality levels, those current levels must be
adjusted downward. This process of adjusting air quality to simulate just meeting a specific
standard is described in more detail below.
6.3.1 Adjustment of Ambient Air Quality
Based on the level of U.S. policy-relevant background (PRB) and observed trends in
ambient monitoring, ambient NC>2 concentrations were proportionally adjusted at each location
using the maximum monitored concentration that occurred in each year. Policy-relevant
background is defined as the distribution of NC>2 concentrations that would be observed in the
U.S. in the absence of anthropogenic (man-made) emissions of NC>2 precursors in the U.S.,
Canada, and Mexico. Policy-relevant background for most of the continental U.S. is estimated to
be less than 300 parts per trillion (ppt) (ISA, Section 2.4.6). In the Northeastern U.S. where
present-day NO2 concentrations are highest, this amounts to a contribution of less than 1%
percent of the total observed ambient NC>2 concentration. This low contribution of PRB to NC>2
concentrations provides support for a proportional method to adjust air quality, i.e., an equal
adjustment of air quality values across the entire air quality distribution to just meet a target
value.
While annual average concentrations have declined significantly over the time period of
analysis, the variability in the concentrations, both the annual average and 1-hour concentrations,
have remained relatively constant. This trend is apparent when considering the air quality data
collectively (Appendix A, section 7) and when considering individual locations (Rizzo, 2008).
59
-------
As an example, Figure 6-1 compares the trends in daily maximum NC>2 1-hour
concentration percentiles at the one ambient monitor in Atlanta that was in operation as far back
as 1985 and is currently part of the monitoring network. Three recent years of data (2005-2007)
were selected to constitute a series of low concentration year data along with three historical
years of data (1985, 1986, and 1988) constituting a series of high concentration year data. As
shown in the figure, the relationships between the low and high concentration years at each of
the daily maximum concentration percentiles are mostly linear, with R2 values ranging from 0.88
to 0.99. Where deviation from linearity did occur, it occurred primarily at a single point, either
at the maximum daily maximum or the minimum daily maximum 1-hour concentration. This
indicates that the rate of decrease in ambient air quality concentrations at the mean value is
consistent with the rate of change at the lower and upper daily maximum 1-hour concentration
percentiles. This trend provides support for the use of a proportional approach to adjust current
ambient concentrations to represent air quality under both the current and alternative standard
scenarios.
Aflanta-Sandy Springs-Gainesville, GA-AL: 131210048
a
o
R"2: 0.93
* !-
0.02
I I I I T
0.06 0.1D
Hijfi Year, t§85
R*2: D.96
I I
D.02
a J
I I I I
0.06 i. 10
R*2:tL9§
l I I
0.02 cure
High Year 1965
0.1Q
HNjhYear 19B5
8
O
\
\
0.05 0.10 0.15
Hip Year !W6
D.D5 DJO 0.15
Hp Year 1966
O.B5
0.1D 0.15
19M
H
f
I
I
o -
0
i i i i
R*2:0.99
0.02
Q.05 0.10
QJ2
D.06
0.1Q
D.Q2
DJO
Up Year I98i
H\f\ Year 1961
HlgftYear 19BS
60
-------
Figure 6-1. Comparison of measured daily maximum NOi concentration percentiles in
Atlanta for three high concentration years (1985,1986,1988) versus three low
concentration years (2005-2007) at one ambient monitor.
To adjust concentrations that simulate just meeting the current annual average standard of
0.053 ppm, proportional adjustment factors F for each location (/') and year (/) were derived by
the following:
F^S/C^ equation (6-1)
where,
FJJ = NO2 concentration adjustment factor (unitless) in location /' given the annual
average standard and for each yeary
S = Current standard level (i.e., 53 ppb, annual average NO2 concentration)
Cmax,ij = The maximum annual average NO2 concentration at a monitor in each
location /' and for each yeary (ppb)
In these cases where staff simulated a proportional adjustment in ambient NO2
concentrations using equation (6-2), it was assumed that the current temporal and spatial
distribution of air concentrations (as characterized by the current air quality data) is maintained
and increased NOX emissions contribute to increased NO2 concentrations, with the highest
monitor (in terms of annual averages) being adjusted so that it just meets the current 0.053 ppm
annual average standard. Values for each air quality adjustment factor used for each location
evaluated in the air quality and risk characterization are given in Appendix A (section 7.2). For
each location and calendar year, all the hourly NO2 concentrations in a location were multiplied
by the same constant value F to make the highest annual mean equal to 53 ppb for that location
and year. For example, of twelve monitors measuring NO2 in Boston for year 1995 (Figure 6-2,
top), the maximum annual average concentration was 30.5 ppb, giving an adjustment factor ofF
= 53/30.5 = 1.74 for that year. All hourly concentrations measured at all monitoring sites in that
location are then multiplied by 1.74, resulting in an upward scaling of hourly NO2 concentrations
for that year. Therefore, one monitoring site in Boston for year 1995 would have an annual
average concentration of 0.053 ppm, while all other monitoring sites would have an annual
61
-------
average concentration below that value, although still proportionally scaled up by 1.74 (Figure 6-
2, bottom). Then, using the adjusted hourly concentrations to simulate just meeting the current
standard, the metrics of interest (e.g., annual mean NO2 concentration, the number of potential
health effect benchmark exceedances) were estimated for each site-year.
BOSTON CMSA AIR QUALITY (1995)
(As Is)
100 H
90
u
4
3
O
BOSTON CMSA AIR QUALITY (1995)
Adjusted to Just Meet the Current Standard (CS)
100 -F
25
50
75 100
1-hour NO^ (ppb)
125
150
175
Figure 6-2. Distributions of hourly NO2 concentrations at twelve ambient monitors in the
Boston CMSA, as is (top) and air quality adjusted to just meet the current standard
(bottom), Year 1995.
62
-------
Proportional adjustment factors were also derived considering the form, averaging time,
and levels of the potential alternative standards under consideration. Discussion regarding the
staff selection of each of these components is provided in chapter 5 of this document. The 98th
and 99th percentile 1-hour NO2 daily maximum concentrations averaged across three years of
monitoring were used in calculating the adjustment factors at each of four standard levels as
follows:
f 3
Fm = S,
equation (6-2)
where,
= NO2 concentration adjustment factor (unitless) in location /' given alternative
standard percentile form k and standard level / across a 3 -year period
Si = Standard level / (i.e., 50, 100, 150, 200 ppb 1-hour NO2 concentration (ppb))
Cjjk = Selected percentile k (i.e., 98th or 99th) 1-hour daily maximum NO2
concentration at a monitor in location /' (ppb) for each yeary
As described above for adjustments made in simulating just meeting the current standard,
it was assumed that the current temporal and spatial distribution of air concentrations (as
characterized by the current NO2 air quality data) is maintained and increased NOX emissions
contribute to increased NO2 concentrations, with the highest monitor (in terms of the 3 -year
average at the 98th or 99th percentile) being adjusted so that it just meets the level of the
particular 1-hour alternative standard. Since the alternative standard levels range from 50 ppb
through 200 ppb, both proportional upward and downward adjustments were made to the 1-hour
ambient NO2 concentrations. The values for each air quality adjustment factor used for each
location evaluated in the air quality and risk characterization are given in Appendix A, section
7.2. Due to the form of the alternative standards, the expected utility of such an analysis, and the
limited time available to conduct the analysis, only the more recent air quality data were used
(i.e., years 2001-2006) and separated into two 3-year periods, 2001-2003 and 2004-2006. The 1-
hour ambient NO2 concentrations were adjusted in a similar manner described above for just
meeting the current standard, however, due to the form of these standards, only one factor was
63
-------
derived for each 3-year period, rather than one factor for each calendar year as was done with
just meeting the current standard.
6.3.2 Adjustment of Potential Health Effect Benchmark Levels
Rather than proportionally modify the air quality concentrations used for input to the
exposure model, a proportional adjustment of the potential health effect benchmark level was
performed. This was done to reduce the processing time associated with the exposure modeling
simulations since there were several thousands of receptors modeled in the Atlanta exposure
assessment. In addition, because the adjustment procedure is proportional, the application of a
downward adjustment of the selected benchmark level is mathematically equivalent to a
proportional upward adjustment of the air quality concentrations. The same approach used in the
air quality adjustment described above was used in the exposure modeling to scale the
benchmark levels to simulate just meeting the current and potential alternative standards. For
example, an adjustment factor of 2.27 was determined for Atlanta for year 2001 to simulate
ambient concentrations just meeting the current standard, based on the maximum annual average
NC>2 concentration of 23.3 ppb observed at an ambient monitor for that year (see Appendix A,
section 7.2). Therefore, the 1-hour potential health effect benchmark levels of 100, 200, and 300
ppb were proportionally adjusted to 44, 88, and 132 ppb, respectively for year 2001.
A comparison of the two procedures is presented in Figure 6-3 where air quality is
adjusted to simulate just meeting the current annual standard and where the benchmark is
adjusted to simulate air quality that just meets the current standard with using the as is air
quality. This example uses the distribution of hourly NC>2 concentrations measured at the
maximum ambient monitor (ID 1312100481) within the Atlanta modeling domain for year 2001.
If we were interested in the number of exceedances of 200 ppb 1-hour under the current standard
scenario for example, this would be equivalent to counting the number of exceedances of 88 ppb
using the as is air quality.
For additional clarity, the same ambient air quality data are presented in Figure 6-4, only
with expansion of the highest percentiles on the graph to allow for the counting of the number of
exceedances. In using the air quality adjusted to just meet the current standard, i.e., the as is air
quality was adjusted upwards by a factor 2.27, there are twelve exceedances of 200 ppb 1-hour.
When considering the as is air quality without adjustment but with a downward adjustment of
64
-------
the benchmark by the same factor of 2.27, there are the same number of exceedances of 200 ppb
1-hour. This benchmark adjustment procedure was applied in Atlanta where exposure modeling
was performed to simulate just meeting the current and alternative standards. Additional details
regarding derivation of the adjusted benchmark levels used in the exposure modeling are
provided in chapter 8 of this document.
ATLANTA MSA AIR QUALITY (2001)
As Is and Adjusted to Just Meet the Current Standard (CS)
0 ATL1312100481
ATL1312100481_CS
Adjust AQ up
Adjust Benchmark Down
25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400
1-hour NO2 (ppb)
Figure 6-3. Comparison of adjusted ambient monitoring concentrations (CS) or adjusted
benchmark level (dashed line) to simulate just meeting the current annual average
standard in Atlanta for year 2001.
65
-------
100 --
99.9 --
*
\
, , [
+-
f
4
4
4
f
1
r
^^- — "
-— "*^
— ©— ATL1 31 21 00481
^»— ATL1312100481_CS
Adjust AQ up
— 'Adjust Benchmark Down
50 100 150 200 250
1-hour NO2 (ppb)
300
350
400
Figure 6-4. Comparison of the upper percentiles for where ambient monitoring
concentrations (CS) and the benchmark level (dashed line) were adjusted to simulate just
meeting the current annual standard in Atlanta for year 2001. The hourly
concentration distributions are provided in Figure 6-3.
66
-------
7. AMBIENT AIR QUALITY ASSESSMENT AND HEALTH RISK
CHARACTERIZATION
7.1 OVERVIEW
Ambient monitoring data for each of the years 1995 through 2006 were used in this
analysis to characterize NC>2 air quality across the U.S. This air quality data, as well as other
NC>2 concentrations derived from ambient levels, were used as a surrogate to estimate potential
human exposure. While an ambient monitor measures NC>2 concentrations at a stationary
location, the monitor may well represent the concentrations that persons residing nearby are
exposed to. The quality of the extrapolation of ambient monitor concentration to personal
exposure will be dependent upon the spatial distribution of important emission sources, the siting
of the ambient monitors, and consideration of places that persons visit. It is within this context
that the approach for characterizing the ambient NC>2 air quality was designed.
Based on the health effects information from the human clinical and epidemiological
studies, the averaging time of interest for the air quality characterization was 1-hour, with
concentration levels ranging from between 100 and 300 ppb. Since the current standard is based
on annual average levels of NC>2 while the most definitive health effects evidence is associated
with short-term exposures (i.e., 30-minute to 1-hour, or one to several day), the air quality
analysis required the development of a model that relates annual average and short-term levels of
NC>2. To characterize this relationship and to estimate the number of exceedances of the
potential health effect benchmarks in specific locations, an empirical model, employing the
annual average and hourly concentrations, was chosen to avoid some of the difficulties in
extrapolating outside the range of the observed air quality.
The available NC>2 air quality were first divided into two six-year groups; one contained
data from years 1995-2000, representing an historical data set; the other contained the
monitoring years 2001-2006, representing recent ambient monitoring. Each of these monitoring
year-groups were evaluated considering the NO2 concentrations as they were reported and
representing the conditions at that time (termed in this assessment "as is"). This served as the
first air quality scenario, with the results within each year-group separated by monitor distance
67
-------
from a major road.l The ambient monitor data were categorized in this manner to account for
the potential influence of vehicle emissions on concentrations measured at the monitors within
close proximity to roadways. There is potential for different concentration levels measured at
each of these locations (i.e., near-road versus away from road) and thus potentially different
exposure concentrations experienced by those persons spending time in these locations. A
second scenario used the as is ambient monitoring data obtained from monitors sited >100 m
from a major road and a simplified on-road simulation approach (described below in section
7.2.4) to estimate on-road NC>2 concentrations for each of the year-groups. This scenario was
developed by recognizing that vehicles are important emission sources of NOX and NC>2 and that
people spend time inside vehicles on roads.
Two additional scenarios followed in similar fashion to the as is air quality analysis,
however these scenarios considered the ambient NC>2 concentrations simulated to just meeting
the current standard of 0.053 ppm annual average and each of the alternative 1-hour standards of
50, 100, 150, and 200 ppb.2 Due to the form of the alternative standards considered here (98th
and 99th percentiles of the daily maximum concentrations averaged over 3 years), the recent
ambient monitoring data set was divided into two three-year groups, 2001-2003 and 2004-2006.
Thus, the air quality characterization results are separated into two broad analyses, one
using air quality as is and the other where air quality was adjusted to just meeting the current and
alternative standards. Within both of these analyses, an additional simulation was performed to
estimate NO2 concentrations on roads. The first scenario described above is the only scenario
that uses purely measurement data. Each of the other scenarios either uses a simulation
procedure to estimate on-road concentrations (scenario 2), concentrations that just meet a
particular standard level (scenario 3), or both (scenario 4).
Because many of the NO2 ambient monitoring sites used in this analysis are primarily
targeting public health monitoring objectives, the results are considered a broad characterization
of national air quality and potential human exposures that might be associated with these
scenario-driven concentrations. The output of this air quality characterization is an estimate the
1 As part of our earlier analysis reported in the 2nd draft REA, the historical data were separated into two-road
distance categories, <100 m and >100 m from a major road. The recent data were separated into both the two- and
three-road distance categories.
2 As part of our earlier analysis reported in the 1st draft REA, the historical data were evaluated using concentrations
as is and for air quality adjusted to just meet the current standard. Only the recent data (2001-2006) were evaluated
using air quality adjusted to just meet potential alternative standards and divided into two three-year groups.
68
-------
number of times per year specific locations experience daily maximum levels3 of NC>2 that may
cause adverse health effects in susceptible individuals. Each location that was evaluated
contained one to several monitors operating for a few to several years, generating a number of
site-years of data. The number of site-years in a location were used to generate a distribution of
two exposure and risk characterization metrics; the annual average concentrations and the
number of daily maximum exceedances that did (observed data) or could occur (simulated data)
in a year for that location. The mean and median values were reported to represent the central
tendency of each metric for the four scenarios in each air quality year-group. For example, the
mean annual average concentration for a location is the arithmetic average of all site-year annual
average concentrations in that location given the particular year-group. The minimum annual
average concentration served to represent a lower bound and could be a single or multiple site-
year(s) of data in the location given the particular year-group, dependent on the distribution of
annual average concentrations for each site-year. Since there were either multiple site-years of
monitoring or numerous simulations performed for each location using all available site-years of
data, results for the upper percentiles generally included the 95th, 98th, and 99th percentiles of the
distribution. As described for the minimum value, these upper percentile estimates could also
represent either a single or multiple site-year(s) of data at a location given the particular year-
group.
7.2 APPROACH
There were five broad steps to allow for the characterization of the air quality. The first
step involved collecting, compiling, and screening the ambient air quality data collected since the
prior review in 1995 to ensure consistency with the NC>2 NAAQS requirements. Then, criteria
based on the current standard and the potential health effect benchmark levels were used to
identify specific locations for analysis using descriptive statistical analysis of the screened data
set. All remaining monitoring data not identified by the selection criteria were grouped into one
of two non-specific categories.
3 The historical data (1995-2000) were only evaluated for total number of exceedances, not for daily maximum
exceedances as done for the recent air quality data (2001-2006). Because of the differences in benchmarks levels,
differences in the number of road-to-monitor categories used, and different exceedance levels, the historical data
analyses are presented in Appendix A section A-9.1 to A-9.3, along with the comparable metrics using the recent
(2001-2006) air quality.
69
-------
The locations (both the specific and non-specific) served as the geographic centers of the
analysis, where application of the empirical model was done to estimate concentrations and the
number of exceedances of potential health effect benchmark levels. Next, due to expected
variable influence of road emissions on ambient monitor concentrations, the monitors within
each of the named locations were categorized according to particular attributes, including land
use characteristics, location type, monitoring objective, monitoring height above ground, and
distance to major roadways. In addition to the use of the ambient concentrations (as is) and
ambient concentrations just meeting the current and alternative standard levels, on-road NO2
concentrations were estimated in each location to approximate the potential exposure and risk
metrics associated with these concentrations. And finally, air quality metrics of interest were
calculated using air quality data from each scenario.
7.2.1 Air Quality Data Screen
NC>2 air quality data and associated documentation from the years 1995 through 2006
were downloaded from EPA's Air Quality System (AQS) for this purpose (EPA, 2007c, d). A
site was defined by the state, county, site code, and parameter occurrence code (POC), which
gives a 10-digit monitor ID code. As required by the NO2 NAAQS, a valid year of monitoring
data is needed to calculate the annual average concentration. A valid year at a monitoring site
was comprised of 75% of valid days in a year, with at least 18 hourly measurements for a valid
day (thus at least 274 or 275 valid days depending on presence of a leap year and a minimum of
4,932 or 4,950 hours). This served as the screening criterion for ambient monitoring data used in
the air quality characterization.
Site-years of data are the total numbers of years the collective monitors in a location were
in operation. Of a total of 5,243 site-years of data in the entire NO2 1-hour concentration
database, 1,039 site-years did not meet the above completeness criterion and were excluded from
any further analyses. In addition, since shorter term average concentrations are of interest, the
remaining site-years of data were further screened for 75% completeness on hourly measures in a
year (i.e., containing a minimum of 6,570 or 6,588, depending on presence of a leap year).
Twenty-seven additional site-years were excluded, resulting in 4,177 complete site-years in the
analytical database. Table 7-1 provides a summary of the site-years included in the analysis,
70
-------
relative to those excluded, by location and by two site-year groups.4 The air quality data from
AQS were separated into these two groups, one representing historical data (1995-2000) and the
other representing recent ambient monitoring data (2001-2006) to account for anticipated long-
term temporal variability in NC>2 concentrations within each location. The selection of specific
locations was a companion analysis to this data screening, and is discussed in section 7.2.2.
Table 7-1. Counts of complete and incomplete site-years of NO2 ambient monitoring data.
Location
Atlanta
Boston
Chicago
Cleveland
Colorado Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Total
Com
1995-2000
24
58
47
11
26
26
12
14
6
16
193
24
93
46
22
6
56
69
1135
200
Number of
plete
2001-2006
29
47
36
11
ND
10
12
30
4
35
177
20
81
39
27
6
43
66
1177
243
4177
Site-Years1
Incorr
1995-2000
5
16
20
2
4
10
4
11
0
4
16
1
12
6
8
0
3
21
249
112
i plete
2001-2006
1
34
22
2
4
4
1
0
2
9
19
4
24
8
25
0
9
18
235
141
1066
Site-^
% Cor
1995-2000
83%
78%
70%
85%
87%
72%
75%
56%
100%
80%
92%
96%
89%
88%
73%
100%
95%
77%
82%
64%
fears
iplete
2001-2006
97%
58%
62%
85%
ND
71%
92%
100%
67%
80%
90%
83%
77%
83%
52%
100%
83%
79%
83%
63%
80%
Notes:
1 The average number of monitors operating per year within the six-year group can be estimated by
dividing the number of site-years by 6 (the total would be divided by 12). The actual number of
monitors operating in any specific year is variable. See Appendix A-4, Table A-1 as an example.
ND no available monitoring data.
7.2.2 Selection of Locations for Air Quality Analysis
Criteria were established for selecting monitoring sites with high annual means and/or
frequent exceedances of potential health effect benchmarks. Selected locations were those that
4 At the time data were downloaded from AQS, 14 of 18 named locations and the 2 grouped locations contained
enough data to be considered valid for year 2006.
71
-------
had a maximum annual mean NC>2 level at a particular monitor greater than or equal to 25.7 ppb,
which represents the 90th percentile across all locations and site-years, and/or had at least one
reported 1-hour NO2 level greater than or equal to 200 ppb, the lowest level of the potential
health effect benchmarks.5 A location in this context would include a geographic area that
encompasses more than a single air quality monitor (e.g., particular city, metropolitan statistical
area (MSA), or consolidated metropolitan statistical area or CMSA). First, all ambient monitors
were identified as either belonging to a CMSA, a MSA, or neither. Then, locations of interest
were identified through statistical analysis of the ambient NO2 air quality data for each site
within a location.
Fourteen locations met both selection criteria and an additional four met at least one of
the criteria (see Table 7-2).6 In addition to these 18 specific locations, the remaining sites were
grouped into two broad location groupings. The Other CMSA location contains all the other sites
that are in MS As or CMS As but are not in any of the 18 specified locations. The Other Not MSA
location contains all the sites that are not in an MSA or CMSA. The final database for analysis
included air quality data from a total of 204 monitors within the named locations, 331 monitors
in the Other CMSA group, and 92 monitors in the Other Not MSA group.
5 At the time the locations were selected for analysis, the ISA identified 200 ppb as a lower level where health
effects were observed in human clinical studies.
6 New Haven, CT, while meeting both criteria, did not have any recent exceedances of 200 ppb and contained one of
the lowest maximum concentration-to-mean ratios. Therefore this location was not separated out as a specific
location for analysis and was grouped within a non-specfic category.
72
-------
Table 7-2. Locations selected for NO2 Air Quality Characterization, associated abbreviations, and
values of selection criteria.
Location
Type1
MSA
CMSA
CMSA
CMSA
MSA
CMSA
CMSA
MSA
MSA
MSA
CMSA
CMSA
CMSA
CMSA
MSA*\
MSA
MSA
CMSA
MSA/CMSA
.
Code
0520
1122
1602
1692
1720
2082
2162
2320
3600
4120
4472
4992
5602
6162
6200
6520
7040
8872
.
.
Description
Atlanta, GA
Boston-Worcester-Lawrence, MA-
NH-ME-CT
Chicago-Gary-Kenosha, IL-IN-WI
Cleveland-Akron, OH
Colorado Springs, CO
Denver-Boulder-Greeley, CO
Detroit-Ann Arbor-Flint, Ml
El Paso, TX
Jacksonville, FL
Las Vegas, NV-AZ
Los Angeles-Riverside-Orange
County, CA
Miami-Fort Lauderdale, FL
New York-Northern New Jersey-
Long Island, NY-NJ-CT-PA
Philadelphia-Wilmington-Atlantic
City, PA-NJ-DE-MD
Phoenix-Mesa, AZ
Provo-Orem, UT
St. Louis, MO-IL
Washington-Baltimore, DC-MD-
VA-WV
Other MSA/CMSA
Other Not MSA
Abbreviation
Atlanta*
Boston*
Chicago
Cleveland*
Colorado
Springs*
Denver*
Detroit*
El Paso*
Jacksonville
Las Vegas*
Los Angeles*
Miami
New York*
Philadelphia*
Phoenix*
Provo
St. Louis*
Washington DC*
Other MSA
Other Not MSA
Maximum # of
Exceedances
of 200 ppb
1
1
0
1
69
2
12
2
2
11
5
3
3
3
37
0
8
2
10
2
Maximum
Annual
Mean (ppb)
26.6
31.1
33.6
28.1
34.8
36.8
25.9
35.1
15.9
27.1
50.6
16.8
42.2
34.0
40.5
28.9
27.2
27.2
31.9
19.7
Notes:
1 CMSA is consolidated metropolitan statistical area; MSA is metropolitan statistical area according to the
1 999 Office of Management and Budget definitions (January 28, 2002 revision).
* Indicates locations that satisfied both the annual average and exceedance criteria.
7.2.3 Site Characteristics of Ambient NOi Monitors
The siting of the ambient monitors is of particular importance, recognizing that the
purpose of the monitoring could have an influence on the measured NC>2 concentrations and
subsequent interpretation in the air quality characterization. Specific monitoring site
characteristics provided in AQS were obtained, including the monitoring objective, measurement
73
-------
scale, and predominant land-use. Additional features such as proximity to NOX emission
sources, including mobile and stationary sources, were estimated using each monitoring site and
emission source geographic coordinates. Each of these attributes is summarized here to provide
perspective on the representation the ambient NCh monitoring network within each location. A
more thorough discussion of the ambient monitoring network is provided in Chapter 2.
Individual monitor site characteristics are given in Appendix A-4.
The monitor objective meta-data field describes the nature of the network in terms of its
attempt to generally characterize health effects, photochemical activity, transport, or welfare
effects. In recognizing that there were variable numbers of ambient monitors in operation or
with valid data in a given year, and that the air quality characterization was performed for
particular year-groups of data, the monitoring objectives were weighted by the number of site-
years. In addition, the monitors can have more than one objective. To evaluate the
representation of the monitors used for the purposes of the REA, the four objective categories
were used in the following order (i.e., population exposure, high concentration,
general/background, unknown) to characterize the monitor with one objective. All other
objectives were grouped into an "Other" category. Tables 7-3 and 7-4 summarize the monitoring
objectives for each location using the historical and recent air quality data, respectively. Most
locations contained monitoring site-years of data that would target public health through the
population exposure and highest concentration categories. Where these categories were not the
predominant objective, the monitoring objective was mainly unknown.
74
-------
Table 7-3. Percent of ambient NO2 monitors with selected monitoring objectives, using all valid
site-years of historical air quality (1995-2000).
Location
Atlanta
Boston
Chicago
Cleveland
Colorado Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Population
Exposure
88
43
55
100
23
42
50
50
23
50
26
85
73
21
38
45
22
High
Concentration
21
32
19
42
14
13
9
50
40
13
27
11
19
14
7
General
/Background
11
27
29
13
3
2
4
17
Unknown
13
33
2
100
31
17
100
13
64
32
100
68
39
32
35
Other
3
7
13
2
2
4
5
19
Table 7-4. Percent of ambient NO2 monitors with selected monitoring objectives, using all valid
site-years of recent air quality (2001-2006).
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Population
Exposure
79
66
53
100
50
50
80
54
20
50
40
79
67
28
48
45
22
High
Concentration
15
47
50
50
20
14
10
50
23
10
22
21
17
16
5
General
/Background
14
2
1
9
25
Unknown
21
15
100
64
36
10
4
100
51
27
23
21
Other
4
17
5
7
8
6
27
75
-------
Similarly, the overall measurement scale of the monitors used for the air quality
characterization in each location was evaluated based on the valid site-years of air quality data.
The measurement scale represents the air volumes associated with the monitoring area
dimensions. While a monitor can have multiple objectives, each monitor has only one
measurement scale. Tables 7-5 and 7-6 summarize the measurement scales of the monitors in
each location using the historical and recent air quality data, respectively. Most locations
contained monitoring site-years of data with measurement scales of urban (4 to 50 km) or
neighborhood (500 m to 4 km). Where these categories were not the predominant objective, the
measurement scale was commonly not indicated.
The land use meta-data indicate the prevalent land use within 1A mile of the monitoring
site. Tables 7-7 and 7-8 summarize the land use surrounding the monitors in each location using
the historical and recent air quality data, respectively. Most locations contained monitoring site-
years of data from areas within residential and commercial areas, but were generally dominated
by residential land use. Two locations however were characterized with site-years of data
associated entirely with commercial land use (i.e., Jacksonville and Provo).
Table 7-5. Percent of ambient NO2 monitors with selected measurement scales, using all valid site-
years of historical air quality (1995-2000).
Location
Atlanta
Boston
Chicago
Cleveland
Colorado Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Regional
3
11
1
4
8
Urban
63
26
6
45
23
42
29
13
17
38
6
13
23
11
26
21
22
Neighborhood
38
28
55
55
65
42
50
75
14
63
53
74
55
21
42
42
26
Middle
26
14
6
14
2
Micro
12
7
1
None
31
2
100
12
17
14
100
13
55
34
13
9
100
68
30
30
45
76
-------
Table 7-6. Percent of ambient NO2 monitors with selected measurement scales, using all valid site-
years of recent air quality (2001-2006).
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Regional
4
3
4
19
Urban
59
23
14
27
50
50
14
18
50
6
13
26
14
27
22
20
Neighborhood
41
57
53
73
50
50
60
83
15
50
46
77
37
35
48
44
27
Middle
33
11
5
22
4
1
Micro
11
1
None
4
40
100
56
43
10
15
100
51
24
25
33
Table 7-7. Percent of ambient NO2 monitors with selected land use, using all valid site-years of
historical air quality (1995-2000).
Location
Atlanta
Boston
Chicago
Cleveland
Colorado
Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Residential
33
41
57
45
54
83
14
27
55
63
36
72
100
41
57
42
21
Commercial
25
45
55
15
19
17
86
100
33
32
25
36
13
100
32
36
33
20
Industrial
10
6
27
38
13
4
13
13
13
11
7
15
Mobile
26
20
5
4
7
Agricultural
42
3
11
4
42
3
13
2
16
7
14
35
Other
7
1
1
2
77
-------
Table 7-8. Percent of ambient NO2 monitors with selected land use, using all valid site-years of
recent air quality (2001-2006).
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Residential
38
28
53
45
100
40
40
58
75
43
64
93
42
64
42
23
Commercial
21
62
55
50
60
100
23
34
25
35
21
100
30
27
34
13
Industrial
11
14
17
3
8
15
14
8
14
Mobile
33
17
3
2
9
Agricultural
41
50
1
13
7
14
9
13
39
Other
3
1
1
1
Mobile and stationary sources (i.e., primarily power generating utilities using fossil fuels)
are the most significant contributors to nitrogen oxides (NOX) emissions in the U.S. (ISA, section
2.2.1). Therefore, the distances of each ambient monitor in the named locations to major roads
and stationary sources were calculated.7 The estimated distances of the monitors to major roads
ranged from a few meters to several hundred meters (Table 7-9). On average, most of the
ambient monitors are placed at a distance of 100 meters or greater from a major road, however in
locations with a large monitoring network such as Boston, Chicago, or New York, there were a
few monitors sited within close proximity (<20 meters) of a road. Most of the monitors were
sited at much greater distances to NOX stationary sources than major roads. In general, monitors
were located at least 1 km from NOX stationary sources, with over half of the monitors located at
distances > 5 km.
Because there is potential for roadway emissions to affect concentrations at monitors
sited close to major roads, each of the ambient monitors were further categorized based on the
monitor distance from major roads. Three categories were identified, the first containing those
7 Major road types were defined as: primary limited access or interstate, primary US and State highways, secondary
State and County, freeway ramp, or other ramps. Distances were estimated for stationary sources within 10 km
having emissions greater than 5 tons per year (tpy). See Appendix A-4 for details for the approach used.
78
-------
monitors sited at or within 20 meters, (<20 m), those between 20 meter and 100 meters (>20 m,
<100 m), and those located at least 100 meters from a major road (>100 m) (Figure 7-1).
Table 7-9. Distance of ambient monitors to the nearest major sources in selected locations.
Location
Atlanta
Boston
Chicago
Cleveland
Colorado Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
n1
4
21
12
4
6
7
3
7
1
10
43
4
26
10
7
1
13
16
Distance (m) of monitors to
nearest major road
min
134
7
2
2
79
18
339
33
144
1
1
15
6
45
7
353
5
14
med
505
70
93
134
180
65
393
128
181
89
55
119
167
141
97
83
max
809
337
738
187
386
748
415
718
914
570
103
508
630
433
421
338
Distance (m) of monitor to
nearest 5 ton per year (tpy) NOX
stationary source
min
656
142
411
956
782
910
321
119
708
3837
140
1323
103
231
833
1214
396
288
med
7327
5363
7277
7278
6340
5904
7549
6085
5720
7237
6165
7611
6467
5689
6355
8178
7120
6254
max
9847
9988
9994
9884
9933
9979
9997
9991
9558
9950
9991
9117
9983
9982
9890
9433
9990
9973
Notes:
1 n is the number of monitors operating in a particular location between 1995 and 2006. The min, med, and
max represent the minimum, median, and maximum percentiles of the distribution for the distance in meters
(m) of the monitor to the nearby sources. Monitors >1km from a major road and monitors having no
stationary sources within 10 km are not included in this distribution. Individual monitor distances and
stationary source emissions within 10 km is provided in Appendix A-4.
79
-------
>100m
+ ambient monitor location
major road
Figure 7-1. Illustration of three roadway distance categories used to characterize ambient
monitors in the Air Quality Characterization.
7.2.4 Estimation of On-Road Concentrations using Ambient Concentrations
Since mobile sources can account for a large part of personal exposures to ambient NO2
in some individuals, the potential impact of roadway levels of NO2 was evaluated. A strong
relationship has been reported between NO2 levels measured on roadways and NO2 measured at
increasing distance from the road. This relationship has been described previously (e.g., Cape et
al., 2004) using an exponential decay equation of the form:
C=Ch+Ce
equation (7-1)
where,
Cx
Cv
k
NO2 concentration at a given distance (x) from a roadway (ppb)
NO2 concentration (ppb) at a distance from a roadway, not directly influenced
by road or non-road source emissions
NO2 concentration contribution from vehicles on a roadway (ppb)
Combined formation/decay constant describing NO2 with perpendicular
distance from roadway (meters'1)
Distance from roadway (meters)
80
-------
Based on the findings of several researchers, much of the decline in NC>2 concentrations
with distance from the road has been shown to occur within the first few meters (approximately
90% within a 10 meter distance), returning to near ambient levels between 200 to 500 meters
(Rodes and Holland, 1981; Bell and Ashenden, 1997; Gilbert et al., 2003; Pleijel et al., 2004).
At a distance of 0 meter, referred to here as on-road, the equation reduces to the sum of the non-
source influenced NC>2 concentration and the concentration contribution expected from vehicle
emissions on the roadway using
Cr =Ca(l+ m) equation (7-2)
where,
Cr = 1-hour on-road NC>2 concentration (ppb)
Ca = 1-hour ambient monitoring NC>2 concentration (ppb) either as is or adjusted to
just meet the current or alternative standards
m = Ratio derived from estimates of Cv/Cb (from equation (7-1))
and assuming that Ca = C&.8
To estimate on-road NC>2 levels as a function of the level recorded at ambient monitors at
a distance from a roadway, empirical data from published scientific literature were used. A
literature review was conducted to identify published studies containing NC>2 concentrations on
roadways and at varying distances from roadways. Relevant data identified from this literature
review were used to estimate the ratio (m) of the on-road vehicle concentrations (Cv) to NC>2
concentration at a distance from a roadway (Cb) (equation 7-1), generating a distribution of
values for use in estimating on-road concentrations from the ambient monitor concentrations
(Table 7-10). The distribution of derived m ratios were evaluated for possible stratification using
potential influential factors reported in the collection of studies, the number of values of m
available for potential groupings, and how the data were to be applied to the ambient monitoring
8 Note that Ca may differ from Cb since Ca could include the influence of on-road as well as non-road sources.
However, it is expected that for most monitors sited at a distance of >100 from a major road, the influence of on-
road emissions would be neglibible so that Ca = Cb.
81
-------
data. In general, categorizing the data based on summer and not summer seasons were
determined appropriate, containing 21 and 19 samples, respectively. Then, on-road NC>2
concentrations are estimated by probabilistically applying the distribution of adjustment factors
(l+m) to the ambient monitor concentrations used in this study that were sited at distances > 100
m of a major road. See Appendix A, section 8 for a detailed explanation of the m ratio derivation
and the literature sources used.
Table 7-10. Derived Cv/Cb ratios (m) for two season groups used for adjusting ambient NO2
concentrations to simulate on-road NO2 concentrations.
Cv/Cjb ratios (m)
Summer
0.49
0.51
0.52
0.67
0.70
0.74
0.75
0.78
0.78
0.79
0.90
0.92
0.93
0.94
1.13
1.19
1.21
1.32
1.95
2.43
2.45
2.70
Not
Summer
0.22
0.25
0.36
0.36
0.42
0.47
0.58
0.59
0.64
0.75
0.78
0.79
0.79
0.82
0.86
1.08
1.14
1.50
1.54
Theoretically, NC>2 concentrations can increase at a distance from the road due to
chemical interaction of NOX with Os, the magnitude of which can be driven by certain
meteorological conditions (e.g., wind direction). As such, the maximum NC>2 concentration may
not occur on the road but at a distance from the road. However, there are two important
components of this estimation procedure that need consideration. First, the relationship
developed from the peer-reviewed NC>2 roadway and near-road measurement studies was used to
82
-------
estimate NC>2 concentrations that occur on the road and not used to estimate NC>2 concentrations
that could occur at a distance from the road. If there are peak concentrations at a given distance
from a roadway that occur frequently in a location, the ambient monitors located within 20 m or
100 m of a road would capture these concentrations, where such monitors are available in a
location. Second, since there is potential for monitors that are sited near roadways to be
influenced by vehicle emissions and equation (7-2) assumes the ambient concentration is
approximating NC>2 concentrations not directly influenced by the roadway, the ambient monitors
<100 m were not used for estimating the on-road NC>2 concentrations in this analysis. The
uncertainty regarding any additional issues or assumptions and the potential effect on exposure
estimates are discussed in section 7.4.
To estimate on-road NO2 concentrations, each monitoring site was randomly assigned
one on-road adjustment factor (\+rn) for summer months and one for non-summer months from
the derived empirical distribution. On-road adjustment factors were assigned randomly because
we expect the empirical relationship between Cv and Cb to vary from place to place and we do
not have sufficient information to match specific ratios with any of the locations simulated in this
assessment. Hourly NC>2 levels were estimated for each site-year of data in a location using
equation (7-2) and the randomly assigned on-road adjustment factors. The process was
simulated 100 times for each site-year of hourly data. For example, the Boston CMS A location
had 210 random selections from the on-road distributions applied independently to the total site-
years of data (105). Following 100 simulations, a total of 10,500 site-years of data were
generated using this procedure (along with 21,000 randomly assigned on-road values selected
from the appropriate empirical distribution).
7.2.5 Air Quality Concentration Metrics
For each of the four air quality characterization scenarios considered, two concentration
metrics were calculated, including the annual average NC>2 concentrations for each site-year and
the number of exceedances of the potential health effect benchmark levels. To characterize this
relationship and to estimate the number of exceedances of the potential health effect benchmarks
in specific locations, several possible models were explored (i.e., exponential regression, logistic
regression, a regression assuming a Poisson distribution, and an empirical model). An empirical
model, employing the annual average and hourly NC>2 concentrations, was chosen to avoid some
83
-------
of the difficulties in extrapolating outside the range of the measurement data. In addition, an
empirical model could be used for any averaging time of interest. A detailed discussion
justifying the selection of the empirical approach rather than using a regression approach is
provided in Appendix A, section 6.
Using Figure 7-1 as an illustrative example of a location, assume that for a recent air
quality monitoring year-group (2001-2003) there are an equal number of valid monitoring years
(i.e., 3 years in this example) at each monitoring site indicated in the figure. In total there would
be 27 site-years of data (9 monitors by 3 years) and, when separated into the major road
categories, there would be 6, 15, and 6 site-years of ambient concentration data at < 20 m, < 20
and < 100 m, > 100 m from a major road, respectively. In the first scenario, the air quality is
analyzed without adjustment, giving mean annual average concentrations based on averaging the
6, 15, and 6 annual average NC>2 concentrations within each respective monitor-to-road distance
category. Median annual average concentrations are also provided to represent the central
tendency and are the median concentration values of the 6, 15, and 6 site-years of data. Due to
the limited number of site-years of data for each of the road distance categories, the 98th and 99th
percentile values are essentially the same, i.e., the maximum site-year annual average
concentration. The same approach was used for the counts of exceedances in a year, reporting
each the central tendency values and the lower and upper percentiles for each location.
In the second scenario, air quality metrics for the estimated on-road NC>2 concentrations
were generated in a similar manner. The concentration distributions for the annual average
concentrations and the distributions for the number of exceedances of short-term potential health
effect benchmark levels were calculated for each location and year-group. In using the
hypothetical location described above (Figure 7-1) for years 2001-2003, 600 site-years of hourly
on-road concentrations were simulated (6 site-years by 100 simulations each). Mean (average of
the 600 site-year values) and median (the average of the 300th and 301st site-year values) are
reported to represent the central tendency of each air quality metric. Since there were multiple
site-years and numerous simulations performed at each location using all valid site-years of data,
results for the upper percentiles also included the 98th and 99th percentiles of the distribution.
The 98th and 99th percentiles were the 588th and the 594th highest site-year values, respectively,
of the 600 calculated and ranked values. Roadways with high vehicle densities are likely better
represented by on-road concentration estimates at these upper tails of the distribution.
84
-------
7.3 AIR QUALITY AND HEALTH RISK CHARACTERIZATION
RESULTS
7.3.1 Ambient Air Quality (As Is)
As described earlier, this first scenario analyzing the as is air quality is based purely on
the measurement data. The air quality data obtained from AQS were first separated into two six-
year groups, one representing historical data (1995-2000) and the other representing more recent
data (2001-2006). The two broad six-year groups of data were compared using each location's
distribution of annual average concentrations and the total number of 1-hour exceedances of the
potential health effect benchmark levels. Briefly, annual average concentrations were about 15%
higher for the historical data when compared with the more recent data using corresponding
locations. In general, concentrations were about 20-25% higher at monitors sited <100 m from a
major road considering either six-year group of data. While the exceedances of potential health
effect benchmark levels were limited to a few locations, in general, a greater number of
exceedances were observed using the historical data set and at locations sited <100 m from a
major road when compared with the recent air quality. Detailed descriptive statistics regarding
concentration distributions for all locations, monitoring sites, and all monitoring years are
provided in the Appendix A, section 5.
Detailed analyses were performed using the recent as is ambient monitoring data. As
described in section 6.2.1, to remain consistent with the planned analysis of alternative standards
(sections 7.3.3 and 7.3.4), the recent ambient monitoring data were separated into two three-year
groups, 2001-2003 and 2004-2006. A summary of the descriptive statistics for the annual
average ambient NC>2 concentrations at each selected location for years 2001-2003 and 2004-
2006 are provided in Tables 7-11, 7-12 and 7-13 for monitors sited >100 m, 20 m to <100 m, and
<20 m from a major road, respectively. None of the locations contained a measured exceedance
of the current annual average standard of 0.053 ppm at any monitor, although Los Angeles and
New York had at least one annual average concentration >40 ppb during 2001-2003. There were
a fewer number of locations with monitors sited <100 m of a major road, however in most of the
locations where comparative monitoring data were available, the annual average concentrations
were greater at the monitors within 100 m of a major road (in 33 of 42 possible location/year-
group combinations) when compared with monitors > 100 m of a major road. Annual average
85
-------
concentrations measured at monitors located <20 m from a major road were also frequently
greater than those measured at monitors sited <20 m and >100 m for most location/year
combinations (68%). Where concentrations were greater at the monitors <100 m from a major
road, the concentrations were on average about 40-65% higher when compared with the more
distant monitors in each corresponding location. A comparison of the three-year groups of data
within each monitor site-group indicates that the more recent monitoring concentrations (i.e.,
2004-2006) were frequently lower, on average by about 13-19%. These average trends in
concentration across year-group and monitor site group were generally observed across all
percentiles of the distribution.
Table 7-11. Monitoring site-years and annual average NO2 concentrations, using recent air quality
data (as is) and monitors sited >100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
2001-2003
Site-
Years1
14
6
9
3
2
6
12
2
16
51
6
26
14
5
3
9
18
612
127
Annual Mean NO2 (ppb)2
mean
12
10
22
18
24
21
15
14
10
22
9
20
20
27
24
17
18
13
7
min
4
5
17
17
21
19
10
14
2
5
7
11
15
22
22
14
9
1
1
med
15
11
20
17
24
20
16
14
7
24
9
18
18
29
24
17
21
13
6
p95
23
12
28
19
26
23
18
15
22
34
10
28
28
29
25
21
25
21
14
p98
23
12
28
19
26
23
18
15
22
36
10
31
28
29
25
21
25
22
15
p99
23
12
28
19
26
23
18
15
22
37
10
31
28
29
25
21
25
24
16
2004-2006
Site-
Years1
15
8
8
ND
3
6
12
2
11
54
4
22
12
9
3
4
17
565
116
Annual Mean NO2 (ppb)2
mean
11
9
19
20
17
14
14
9
18
8
19
17
24
24
15
15
11
7
min
3
7
16
18
14
8
13
1
5
7
10
14
21
21
12
7
1
1
med
14
9
18
20
17
15
14
6
18
8
20
16
24
22
14
16
11
6
p95
18
10
24
21
20
18
14
20
27
8
25
25
26
29
18
22
19
15
p98
18
10
24
21
20
18
14
20
31
8
27
25
26
29
18
22
21
16
p99
18
10
24
21
20
18
14
20
31
8
27
25
26
29
18
22
23
16
Notes:
1 The average number of monitors operating per year within the three-year group can be estimated by dividing
the number of site-years by 3.
2 Annual means for each monitor were first calculated based on all hourly values in a year. Then the mean of the
annual means was estimated as the sum of all the annual means in a particular location divided by the number of
site-years across the monitoring period. The min, med, p95, p98, p99 represent the minimum, median, 95th, 98th,
and 99th percentiles of the distribution for the annual means in the selected three-year group.
ND no available monitoring data.
86
-------
Table 7-12. Monitoring site-years and annual average NO2 concentrations, using recent air quality
data (as is) and monitors sited >20 m and <100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
2001-2003
Site-
Years1
ND
14
6
ND
ND
ND
3
ND
3
35
3
13
7
2
ND
11
10
ND
ND
Annual Mean NO2 (ppb)2
mean
17
31
21
6
24
14
31
24
23
14
20
min
9
28
20
3
4
13
21
19
22
9
14
med
19
31
21
6
24
14
30
24
23
12
22
p95
25
32
22
9
40
16
40
30
24
25
26
p98
25
32
22
9
41
16
40
30
24
25
26
p99
25
32
22
9
41
16
40
30
24
25
26
2004-2006
Site-
Years1
ND
11
6
2
ND
ND
3
ND
ND
22
2
11
6
3
ND
8
12
ND
ND
Annual Mean NO2 (ppb)2
mean
15
29
15
15
25
13
28
22
19
12
18
min
10
28
14
13
9
13
18
18
19
8
13
med
16
29
15
13
27
13
29
22
19
10
18
p95
19
31
17
18
33
14
36
26
20
22
24
p98
19
31
17
18
34
14
36
26
20
22
24
p99
19
31
17
18
34
14
36
26
20
22
24
Notes:
1 The average number of monitors operating per year within the three-year group can be estimated by dividing
the number of site-years by 3.
2 Annual means for each monitor were first calculated based on all hourly values in a year. Then the mean of the
annual means was estimated as the sum of all the annual means in a particular location divided by the number of
site-years across the monitoring period. The min, med, p95, p98, p99 represent the minimum, median, 95th, 98th,
and 99th percentiles of the distribution for the annual means.
ND no available monitoring data.
87
-------
Table 7-13. Monitoring site-years and annual average NO2 concentrations, using recent air quality
data (as is) and monitors sited <20 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
2001-2003
Site-
Years1
ND
5
4
3
2
ND
ND
ND
3
9
3
7
ND
3
ND
6
4
ND
ND
Annual Mean NO2 (ppb)2
mean
21
22
23
36
22
30
6
28
35
18
23
min
7
22
22
35
21
23
6
25
34
16
20
med
23
22
22
36
22
29
6
28
35
19
24
p95
30
24
24
37
23
37
7
30
37
20
26
p98
30
24
24
37
23
37
7
30
37
20
26
p99
30
24
24
37
23
37
7
30
37
20
26
2004-2006
Site-
Years1
ND
3
3
3
3
ND
ND
ND
2
6
2
2
ND
5
ND
5
5
ND
ND
Annual Mean NO2 (ppb)2
mean
24
19
21
28
19
27
6
28
23
16
19
min
23
18
18
27
19
20
6
27
11
15
14
med
23
20
22
28
19
29
6
28
31
16
18
p95
25
20
22
29
20
31
6
28
32
17
23
p98
25
20
22
29
20
31
6
28
32
17
23
p99
25
20
22
29
20
31
6
28
32
17
23
Notes:
1 The average number of monitors operating per year within the three-year group can be estimated by dividing
the number of site-years by 3.
2 Annual means for each monitor were first calculated based on all hourly values in a year. Then the mean of the
annual means was estimated as the sum of all the annual means in a particular location divided by the number of
site-years across the monitoring period. The min, med, p95, p98, p99 represent the minimum, median, 95th, 98th,
and 99th percentiles of the distribution for the annual means.
ND no available monitoring data.
The number of daily maximum exceedances of the potential health effect benchmark
levels (100, 150, 200, 250, and 300 ppb NO2 for 1-hour) is shown in Tables 7-14 through 7-19
using recent as is ambient monitoring data and considering the three road distance categories.
As a reminder, these exceedance data are based on whether the daily maximum concentration at
a monitor was above the benchmark level; a single monitor value of 10 would represent ten days
in the year where there was at least one 1-hour exceedance of the benchmark level. Since there
are multiple monitors and monitoring years in a location, a distribution of these values can be
generated. The mean represents the central tendency value for each location; it is the average
number of days per year that there were daily maximum 1-hour NO2 concentration exceedances
observed over the three year period in a location. The upper percentiles (98th and 99th) represent
-------
an upper-level estimate of the number of days in one year there were daily maximum 1-hour
concentration exceedances at a particular monitor (or possibly more than one) in ae location.
In general, the number of daily maximum 1-hour benchmark exceedances was low across
all locations and considering both three-year groups of the as is air quality and monitors located
>100 m from a major road (Tables 7-14 and 7-15). The average number of exceedances of the
100 ppb 1-hour NC>2 concentration across each location was typically zero to a few, with one
location (Provo) containing an average of fourteen exceedances. Considering that there are 365
days in a year, this many exceedances amounts to a small fraction of the year (at most 4%) with
an exceedance of the lowest potential health effect benchmark level at a few locations. For the
locations with greater than one annual average exceedance, the numbers were primarily driven
by a single site-year of data. For example, Detroit contained a largest number of exceedances of
200 ppb (a maximum of 4 days in the year) for as is air quality data from years 2001-2003
(Table 7-14). All of these exceedances occurred at one monitor (ID 2616300192) during one
year (2002). Provo contained that greatest number of daily maximum exceedances of both the
100 and 150 ppb benchmark level (Table 7-15), associated with measurements from one monitor
(ID 4904900021) in 2006. However, when considering the collective locations (both the named
locations and aggregated), daily maximum exceedances of 150 and above were rarely occurred.
When considering monitors sited <20 m and >100 m of a major road (Table 7-16 and 7-
17), only a few locations contained exceedances of the potential health effect benchmark levels,
driven mainly by observations from one or two monitors. For example, in Los Angeles a single
year (2001) for two monitors (IDs 060370030 and 060371103) were responsible for many of the
observed exceedances of 100 ppb in the 2001-2003 year-group (each had 31 exceedances in a
year). Each of these monitors are located about 50 m from a major road along with around 40
stationary sources located within 10 km of this monitor, over half of which contained estimated
emissions of less than 15 tons per year (tpy) (Appendix A, Table A-8). When considering the
higher benchmark levels, nearly all locations had no daily maximum l-hourNO2 concentrations
>150ppb.
At monitoring locations <20 m to major roads, about half of the locations contained a
non-zero average number of daily maximum exceedances of 100 ppb (Tables 7-18 and 7-19). Of
those, Denver, Los Angeles, and Phoenix contained the highest average number of exceedances
given either the 2001-2003 or 2004-2006 year-groups. On average, the percent of days in a year
89
-------
with a daily maximum exceedance at these locations ranged from about 1-2%. As far as the
higher benchmark levels, nearly all locations had no daily maximum l-hourNO2 concentrations
>150ppb.
90
-------
Table 7-14. Number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks in a year, using 2001-2003
recent NO2 air quality (as is) and monitors sited >100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100
mean
0
0
0
0
2
3
0
1
0
4
0
0
0
0
0
0
0
0
0
min
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
2
2
0
1
0
1
0
0
0
0
0
0
0
0
0
p98
3
0
2
1
2
7
1
1
1
17
0
3
1
0
0
1
1
1
5
ppb1
p99
3
0
2
1
2
7
1
1
1
18
0
3
1
0
0
1
1
4
5
Exceedances of 150
mean
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
1
0
0
0
0
5
1
1
0
1
0
0
1
0
0
0
0
0
1
ppb1
p99
1
0
0
0
0
5
1
1
0
8
0
0
1
0
0
0
0
0
1
Exceedances of 200
mean
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
4
0
1
0
0
0
0
1
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
4
0
1
0
4
0
0
1
0
0
0
0
0
1
Exceedances of 250
mean
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
4
0
1
0
0
0
0
1
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
4
0
1
0
0
0
0
1
0
0
0
0
0
0
Exceedances of 300
mean
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the
distribution for the number of daily maximum exceedances in any one year within the monitoring period.
August 2008 - Draft
91
-------
Table 7-15. Number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks in a year, using 2004-2006
recent NO2 air quality (as is) and monitors sited >100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100 |
mean
0
0
0
ND
1
0
0
2
0
0
0
0
0
0
14
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
p98
1
0
0
3
0
0
3
0
2
0
2
0
1
43
0
2
1
2
Dpb1
p99
1
0
0
3
0
0
3
0
2
0
2
0
1
43
0
2
1
3
Exceedances of 150
mean
0
0
0
0
0
0
1
0
0
0
0
0
0
7
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
2
0
0
0
0
0
0
20
0
0
0
0
ppb1
p99
0
0
0
0
0
0
2
0
0
0
0
0
0
20
0
0
0
2
Exceedances of 200 |
mean
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
Dpb1
p99
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
1
Exceedances of 250
mean
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
Exceedances of 300
mean
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the
distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
August 2008 - Draft
92
-------
Table 7-16. Number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks in a year, using 2001-2003
recent NO2 air quality (as is) and monitors sited >20 m and <100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100 |
mean
ND
0
2
ND
ND
ND
2
ND
0
6
0
1
0
0
ND
0
0
ND
ND
min
0
0
0
0
0
0
0
0
0
0
0
med
0
1
2
0
1
0
0
0
0
0
0
p98
0
6
3
0
31
0
8
1
0
0
0
Dpb1
p99
0
6
3
0
31
0
8
1
0
0
0
Exceedances of 150
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
2
0
0
0
0
0
0
ppb1
p99
0
0
0
0
2
0
0
0
0
0
0
Exceedances of 200 |
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
1
0
0
0
0
0
0
Dpb1
p99
0
0
0
0
1
0
0
0
0
0
0
Exceedances of 250
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
1
0
0
0
0
0
0
ppb1
p99
0
0
0
0
1
0
0
0
0
0
0
Exceedances of 300
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
Notes:
1 The mean number of daily maximum exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular
location divided by the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th
percentiles of the distribution for the number of daily maximum exceedances in any one year within the monitoring period.
sID no available monitoring data.
August 2008 - Draft
93
-------
Table 7-17. Number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks in a year, using 2004-2006
recent NO2 air quality (as is) and monitors sited >20 m and <100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100
mean
ND
0
1
1
ND
ND
0
ND
ND
1
1
1
1
0
ND
0
0
ND
ND
min
0
0
0
0
0
0
0
0
0
0
0
med
0
1
1
0
1
1
1
1
0
0
0
p98
1
5
1
0
9
2
3
1
0
1
1
ppb1
p99
1
5
1
0
9
2
3
1
0
1
1
Exceedances of 150
mean
0
0
1
0
0
1
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
1
0
0
1
0
0
0
0
0
p98
0
0
1
0
1
1
1
0
0
0
0
ppb1
p99
0
0
1
0
1
1
1
0
0
0
0
Exceedances of 200
mean
0
0
0
0
0
1
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
1
0
0
0
0
0
p98
0
0
0
0
0
1
0
0
0
0
0
ppb1
p99
0
0
0
0
0
1
0
0
0
0
0
Exceedances of 250
mean
0
0
0
0
0
1
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
1
0
0
0
0
0
p98
0
0
0
0
0
1
0
0
0
0
0
ppb1
p99
0
0
0
0
0
1
0
0
0
0
0
Exceedances of 300
mean
0
0
0
0
0
1
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
1
0
0
0
0
0
p98
0
0
0
0
0
1
0
0
0
0
0
ppb1
p99
0
0
0
0
0
1
0
0
0
0
0
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the
distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
August 2008 - Draft
94
-------
Table 7-18. Number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks in a year, using 2001-2003
recent NO2 air quality (as is) and monitors sited <20 m from a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100 |
mean
ND
0
0
0
7
ND
ND
ND
1
6
0
1
ND
2
ND
0
0
ND
ND
min
0
0
0
3
0
0
0
0
1
0
0
med
0
0
0
7
0
6
0
1
1
0
0
p98
1
0
0
10
2
9
0
1
4
1
1
Dpb1
p99
1
0
0
10
2
9
0
1
4
1
1
Exceedances of 150
mean
0
0
0
1
0
0
0
0
0
0
0
min
0
0
0
1
0
0
0
0
0
0
0
med
0
0
0
1
0
0
0
0
0
0
0
p98
0
0
0
1
0
0
0
0
0
0
0
ppb1
p99
0
0
0
1
0
0
0
0
0
0
0
Exceedances of 200 |
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
Dpb1
p99
0
0
0
0
0
0
0
0
0
0
0
Exceedances of 250
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
Exceedances of 300
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the
distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
August 2008 - Draft
95
-------
Table 7-19. Number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks in a year, using 2004-2006
recent NO2 air quality (as is) and monitors sited <20 m from a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other
MSA/CMSA
Other Not
MSA
Exceedances of 100 |
mean
ND
0
0
0
2
ND
ND
ND
0
3
0
1
ND
2
ND
0
0
ND
ND
min
0
0
0
1
0
0
0
0
0
0
0
med
0
0
0
1
0
2
0
1
2
0
0
p98
0
0
0
3
0
8
0
1
3
0
0
Dpb1
p99
0
0
0
3
0
8
0
1
3
0
0
Exceedances of 150
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
1
0
0
0
0
0
0
0
ppb1
p99
0
0
0
1
0
0
0
0
0
0
0
Exceedances of 200 |
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
Dpb1
p99
0
0
0
0
0
0
0
0
0
0
0
Exceedances of 250
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
Exceedances of 300
mean
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
ppb1
p99
0
0
0
0
0
0
0
0
0
0
0
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the
distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
August 2008 - Draft
96
-------
7.3.2 On-Road Concentrations Derived From Ambient Air Quality (As Is)
Descriptive statistics for estimated on-road NO2 concentrations for each recent three-year
group are presented in Table 7-20. These estimated on-road concentrations were generated using
the simulation procedure described above (section 7.2.3) and represent the second air quality
scenario. For the 17 named locations, the calculation only used monitors sited at a distance >100
m of a major road. The two aggregate locations (i.e., Other CMSA and Not MSA) did not have
estimated monitor distances to major roads therefore all monitoring data available were used to
estimate the distribution of on-road NO2 concentrations.
The simulated on-road annual average NO2 concentrations are, on average, 80% higher
than the respective ambient levels at distances >100 m from a road. This falls within the range
of on-road to distant monitor concentration ratios reported in the ISA (about 2-fold higher
concentrations on-roads) (ISA, section 2.5.4). Denver, Los Angeles, and Phoenix were predicted
to have the highest on-road annual average NO2 levels. This is a direct result of these locations
already containing some of the highest as-is NO2 concentrations prior to the on-road simulation
(see Table 7-11). Estimated on-road annual average concentrations were greater by about 10%
when using the 2001-2003 monitoring data compared with the 2004-2006 monitoring data.
The median of the simulated concentration estimates for Los Angeles were compared
with NO2 measurements provided by Westerdahl et al. (2005) for arterial roads and freeways in
the same general location during spring 2003. Although the averaging time is not exactly the
same, comparison of the medians is judged to be appropriate.9 The estimated median on-road
concentration for 2001-2003 is 41 ppb which falls within the range of 31 ppb to 55 ppb identified
by Westerdahl et al. (2005).
On average, most locations are predicted to have fewer than 3 daily maximum
exceedances of 1-hour NO2 concentrations per year given the 200 ppb potential health effect
benchmark, while the median frequency of exceedances in most locations is estimated to be 1 or
less per year (Tables 7-21 and 7-22). When considering the lowest 1-hour benchmark of 100
ppb, most locations (30 out of 38 location/year-groups) were estimated to have fewer than 50
daily maximum exceedances per year, on average. There are generally fewer predicted mean
9 Table 7-12 considers annual average of hourly measurements, while Westerdahl et al. (2005) reported time-
averaged concentrations of 2-4 hours. Over time, the mean of 2-4 hour averages will be similar to the mean of
hourly concentrations, with the main difference being in the variability (and hence the various percentiles of the
distribution outside the central tendency).
August 2008 - Draft 97
-------
exceedances of the potential health effect benchmark levels when considering the 2004-2006
recent air quality compared with the 2001-2003 air quality. Areas with a relatively high number
of estimated exceedances (e.g., Provo, as described in section 7.3.1) are likely influenced by the
presence of a small number of monitors and one or a few exceptional site-years where there were
unusually high concentrations at the upper percentiles of the concentration distribution. When
considering higher benchmark levels (i.e., 250 ppb), the mean number of estimated daily
maximum exceedances was none or one for most locations.
The upper percentiles (98th and 99th) for estimated number of daily maximum
exceedances of 100 ppb in most locations were under 200 per year, outside of Denver, Los
Angeles, Phoenix, and Provo. In general, there were fewer estimated exceedances using the
2001-2003 air quality compared with the 2004-2006 data. Most locations had >100 estimated
daily maximum 1-hour exceedances of 150 ppb per year at the 98th and 99th percentiles and as
expected, the frequency of benchmark exceedances at all locations was lower when considering
any of the higher benchmark levels (i.e., 200, 250, 300 ppb, 1-hour average) compared with the
lower benchmarks. For example, it is estimated that between about 10 and 30 daily maximum
exceedances of 200 ppb could occur in a year in most locations as an upper bound estimate. The
estimated upper percentile number of exceedances of 250 and 300 ppb 1-hour were generally
less than 10 in most locations.
August 2008 - Draft 98
-------
Table 7-20. Estimated annual average NO2 concentrations on-roads, using recent air quality data
(as is) and an on-road adjustment factor.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
2001-2003
Site-
Years1
1400
600
900
300
200
600
1200
200
1600
5100
600
2600
1400
500
300
900
1800
61200
12700
Annual Mean NO2 (ppb)2
mean
22
17
39
32
42
37
27
26
19
41
16
36
36
49
43
31
33
23
12
min
5
7
21
22
27
24
13
18
3
6
9
14
18
28
28
18
11
1
1
med
24
18
37
32
40
36
27
26
14
40
15
34
33
47
41
30
34
22
11
p95
41
27
60
42
61
51
40
34
44
69
23
58
56
70
58
43
54
41
27
p98
47
29
65
43
63
54
43
36
48
77
24
65
64
72
61
48
58
47
31
p99
53
30
68
45
64
57
44
37
51
82
25
73
66
77
64
50
63
50
33
2004-2006
Site-
Years1
1500
800
800
ND
300
600
1200
200
1100
5400
400
2200
1200
900
300
400
1700
56500
11600
Annual Mean NO2 (ppb)2
mean
20
16
35
36
31
25
24
16
33
14
35
31
43
43
27
28
20
12
min
4
9
20
23
18
10
17
2
6
9
12
18
26
26
16
9
1
1
med
22
15
33
36
30
25
23
11
32
13
35
30
42
41
26
28
20
10
p95
36
22
51
50
43
38
33
40
55
19
54
45
59
68
38
46
36
27
p98
40
24
57
51
45
42
36
44
60
19
58
51
64
70
41
51
41
31
p99
42
24
60
53
47
43
37
46
65
20
61
59
65
71
42
52
45
33
Notes:
1 The average number of monitors operating per year within the three-year group can be estimated by dividing
the number of site-years by 300.
2 Annual means for each monitor were first calculated based on all simulated hourly values in a year. Then the
mean of the annual means was estimated as the sum of all the annual means in a particular location divided by
the number of simulated site-years across the monitoring period. The min, med, p95, p98, p99 represent the
minimum, median, 95th, 98th, and 99th percentiles of the distribution for the annual means.
ND no available monitoring data.
August 2008 - Draft
99
-------
Table 7-21. Estimated number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks on-roads in a
year, using 2001-2003 recent NO2 air quality (as is) and an on-road adjustment factor.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100
mean
23
5
52
31
89
41
32
13
23
71
7
42
37
101
61
25
36
16
4
min
0
0
0
0
8
1
0
0
0
0
0
0
0
1
1
0
0
0
0
med
5
1
35
21
74
30
19
7
4
57
2
28
19
83
38
12
17
3
0
p98
130
36
180
83
242
130
136
55
171
231
50
177
149
280
248
128
169
105
43
ppb1
p99
169
40
191
102
259
141
145
56
194
251
64
201
172
315
289
139
205
129
62
Exceedances of 150
mean
4
0
9
5
17
9
4
1
4
17
1
7
6
16
9
3
6
2
1
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
2
1
5
5
1
1
0
6
0
1
1
2
0
0
0
0
0
p98
36
3
62
30
80
44
24
6
54
94
6
55
49
113
62
28
46
22
9
ppb1
p99
51
5
68
30
94
46
26
7
62
108
10
63
62
124
64
37
54
32
15
Exceedances of 200
mean
1
0
2
1
3
3
0
1
0
5
0
2
1
2
1
0
1
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
1
2
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
8
1
24
10
25
16
5
1
8
41
1
24
11
17
11
5
9
4
2
ppb1
p99
13
1
29
10
26
18
6
1
9
48
2
24
24
20
11
6
14
7
5
Exceedances of 250
mean
0
0
0
0
1
2
0
1
0
1
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
3
0
5
3
5
7
1
1
1
16
0
6
3
4
2
1
1
1
1
ppb1
p99
4
0
12
3
5
7
1
1
1
23
0
9
5
5
2
1
1
2
2
Exceedances of 300
mean
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
1
0
1
1
1
7
0
1
0
7
0
1
1
0
0
1
0
0
1
ppb1
p99
1
0
3
1
1
7
1
1
0
10
0
2
1
0
0
1
0
0
1
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of simulated site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles
of the distribution for the number of daily maximum exceedances in any one year within the monitoring period.
August 2008 - Draft
100
-------
Table 7-22. Estimated number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks on-roads in a
year, using 2004-2006 recent NO2 air quality (as is) and an on-road adjustment factor.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Exceedances of 100
mean
17
2
36
ND
63
20
24
11
15
38
6
35
22
77
51
15
21
10
4
min
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
2
0
20
49
9
12
5
0
25
1
23
10
53
44
5
7
1
0
p98
114
18
148
190
90
114
48
133
150
46
149
101
275
160
76
119
79
43
ppb1
p99
120
21
161
195
103
143
59
148
169
46
171
130
293
160
83
143
100
65
Exceedances of 150
mean
2
0
5
10
2
3
2
2
6
0
5
2
10
17
1
2
1
1
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
4
0
0
2
0
1
0
1
0
1
2
0
0
0
0
p98
26
1
41
47
20
20
5
54
43
6
40
20
57
68
20
20
12
7
ppb1
p99
27
1
53
52
24
23
6
64
55
6
45
22
71
70
24
22
18
13
Exceedances of 200
mean
0
0
1
2
0
0
1
0
1
0
1
0
1
12
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
6
0
7
11
3
4
4
6
11
1
13
3
7
44
2
4
2
2
ppb1
p99
7
0
18
14
6
4
4
8
14
1
15
3
8
44
5
6
3
5
Exceedances of 250
mean
0
0
0
0
0
0
1
0
0
0
0
0
0
7
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p98
1
0
1
3
0
1
3
0
2
0
3
0
1
43
0
1
0
2
ppb1
p99
1
0
2
3
0
1
3
1
3
0
4
0
1
43
0
2
1
2
Exceedances of 300
mean
0
0
0
0
0
0
1
0
0
0
0
0
0
3
0
0
0
0
min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
med
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
3
0
1
0
1
0
0
40
0
0
0
0
ppb1
p99
0
0
0
0
0
0
3
0
1
0
1
0
0
40
0
0
0
2
Notes:
1 The mean number of exceedances represents the sum of the daily maximum exceedances occurring at all monitors in a particular location divided by
the number of simulated site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles
of the distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
August 2008 - Draft
101
-------
7.3.3 Ambient Air Quality Adjusted to Just Meet the Current and Alternative
Standards
As described in section 6.2, each of the current and alternative standards were evaluated
using the more recent air quality data set (i.e., 2001-2006) serving as the third air quality
scenario. Analysis results are presented for a few selected locations, potential health effect
benchmarks, and alternative standard levels, since there were a total of 10 air quality analyses (8
alternative standards, the current standard, and as is air quality), for each year-group of data
(2001-2003 and 2004-2006), for each of the three monitor-to-road categories (<20 m; <20 and
<100m; and >100 m from a major road), and evaluated at five potential health effect benchmark
levels (100, 150, 200, 250, 300 ppb 1-hour). All of the results for each location are provided in
Appendix A, section 9, much of which is summarized here in a series of key figures.
Figures 7-2 and 7-3 illustrate the estimated mean number of daily maximum exceedances
of the three lowest benchmark concentrations (i.e., 100, 150, and 200 ppb) using air quality data
adjusted to just meeting the current annual average standard for year-groups 2001-2003 and
2004-2006, respectively. The number of estimated daily maximum exceedances of 100 ppb
generally ranges from ten to fifty, with pattern of exceedances based on monitor siting consistent
with that noted above for the as is data. In general, there were a greater number of daily
maximum exceedances at the near road monitors compared with those sited >100 m from a
major road, although a few monitors sited at >100 m from a major road contained more
estimated exceedances than the monitors sited within 20 m of a major road (e.g., Denver).
There were also differences in the estimates for each three-year group from what was expected.
For example, a few of the locations had an estimated daily maximum number of exceedances of
100 ppb that were slightly higher for the 2004-2006 year-group when compared with the 2001-
2003 year-group (e.g., Detroit in Figures 7-2 and 7-3). The estimated number of daily maximum
exceedances of 150 and 200 ppb were much lower than that of 100 ppb, the mean number of
exceedances was fewer than 20 for most location years and roadway-monitor groupings. Note
that fifty-one of the 81 possible year-group and monitoring site data combinations at the 19
locations did not have any exceedances of the 200 ppb level when using air quality adjusted to
just meeting the current standard.
August 2008 Draft 102
-------
Figure 7-4 presents the mean estimated number of daily maximum exceedances when
considering the air quality adjusted to just meeting the potential alternative standard levels, using
Chicago as an example to illustrate the patterns in the estimated exceedances for two forms of
the standard. These patterns presented for Chicago apply to the other locations, with a few
exceptions. As expected, the estimated number of daily maximum exceedances is lower for a
99th percentile form compared with each corresponding level using the 98th percentile form of
alternative standard. In general, the number of estimated daily maximum exceedances of the
potential health effect benchmark levels at monitoring sites < 100 m from a major road is greater
than the numbers estimated for monitors sited > 100 m from a major road. This is what one
would expect given the greater potential for vehicle emissions influencing ambient
concentrations at near road monitors. There were also a slightly greater number of estimated
daily maximum exceedances at the monitors sited < 20 m compared with those sited between 20-
100 m. As expected, the number of exceedances of the potential health effect benchmark levels
decreases with decreasing alternative standard level. Regardless of three-year group or
monitoring group, an alternative standard level of 100 ppb tended to reduce the number of
estimated exceedances of 100 ppb to either a few to none.
Figure 7-5 presents mean estimated number of daily maximum exceedances of the 200
ppb concentration level for a few additional locations, Phoenix, Los Angeles, Washington DC,
and St. Louis. Again, there are trends in these results that are consistent with that reported for
the Chicago results, with few exceptions. For example, in St. Louis the estimated number of
daily maximum exceedances at monitors located > 100 m from a major road were greater than
those estimated using the monitoring sites < 100 m from a major road. Also note that there were
variable results when comparing year-groups across the different locations within the monitor
site-group; sometimes the year 2001-2003 contained greater numbers of exceedances when
compared with 2004-2006 (e.g., St. Louis), and other times it did not (e.g. Los Angeles).
However, the alternative standard levels of either 100 or 150 ppb at either percentile consistently
reduced the mean number of daily maximum exceedances of 200 ppb to about zero.
Tables 7-23, 7-24, and 7-25 summarize the annual mean concentrations and estimated
number of exceedances in a year given 2001-2003 air quality adjusted that just meets the 1-hour
100 ppb 98th percentile standard at monitors sited > 100 m, >20 m and < 100 m, and <20 m from
August 2008 Draft 103
-------
a major road, respectively. The tables provide a more comprehensive comparison of the
numbers of daily maximum exceedances for the lowest potential health effect benchmarks (i.e.,
100, 150, and 200 ppb), as well as providing upper percentile estimates for each of the
parameters. The results for this particular year-group are provided to describe patterns within a
given standard level, results were similar for the 2004-2006 air quality data. The complete
results for all of the standard levels and year-groups of air quality, including the observed
number of daily maximum exceedances (as is air quality) are provided in Appendix A, section 9.
Most locations contained a mean of fewer than 5 daily maximum exceedances of the 100 ppb
concentration level, with upper percentile estimates ranging from the 5 to about 15. These
results are comparably less than those estimated using air quality adjusted to just meeting the
current standard (Figure 7-2). At potential health effect benchmark levels above 100 ppb, there
were few estimated exceedances, particularly at and above the 200 ppb level, considering both
the mean and the upper percentiles.
Tables 7-26 summarizes the observed and estimated mean numbers of exceedances of
100 ppb using the 2001-2003 as is air quality and air quality adjusted to just meet the current
standard and the potential alternative 98th percentile standards at each location. The number of
daily maximum exceedances for the as is air quality generally fell within the number of
exceedances estimated using alternative 1-hour 98th percentile standards of 50 ppb and 100 ppb
at each location. When the air quality was adjusted to just meeting the current annual average
standard, the estimated number of daily maximum exceedances was generally near that estimated
using the alternative 1-hour 98th percentile standard of 150 ppb at each location. In a similar
manner, Table 7-27 summarizes the observed and estimated mean numbers of exceedances of
150 ppb 1-hour at each location. The number of daily maximum exceedances using as is air
quality in each location was most similar to that estimated using the alternative 1-hour 98th
percentile standard of 50 ppb, while estimates using the air quality adjusted to just meeting the
current standard again approached the estimated numbers of exceedance using the alternative 1-
hour 98th percentile standard of 150 ppb at each location.
August 2008 Draft 104
-------
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Monitors Sited >=100 m from a Major Road Monitors Sited >20 m, <100 m from a Major Road Monitors Sited >=20 m from a Major Road
]
ij
paaasa
r*
aaaj
•
11
JJJIJJJIJJJIJJJIJJJIJJJI
^
j
1
U
I
D
H
ZZI
iii
iii
iii
iii
iii
(_!_!_!
n> 100 ppb
• > 150 ppb
• > 200 ppb
iii
iii
-
-
-
-
-
No data
l
No data
No data
No data
No data
1
•
^
3
No data
n
No data
No data
D> 100 ppb
• > 150 ppb
• > 200 ppb
-
_
-
_
.
-
No data
I i
I
i
1 i
i
i
•mini |
1
f
1
No data
i
No data
i
No data
i
i
i
1 i
i
i
No data
i
i
i
No data
i
iii i
i
No data
i
No data
n> 100 ppb
• > 150 ppb
• > 200 ppb
0 25 50 75 100 125 150 175 200 2250 25 50 75 100 125 150 175 200 2250 25 50 75 100 125 150 175 200 225
Estimated mean number of daily maximum exceedances of 1-hour NO2 concentrations of 100,150, 200 ppb
Figure 7-2. Estimated mean number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks
in a year, using recent NO2 air quality (2001-2003) adjusted to just meeting the current annual standard (0.053 ppm). Left
graph: monitors >100m from a major road; Middle graph: monitors >20 m and <100 m from a major road; Right graph:
monitors <20 m from a major road.
105
-------
Monitors Sited >=100 m from a Major Road Monitors Sited >20 m, <100 m from a Major Road Monitors Sited <=20 m from a Major Road
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
n
No data
D
'
1
•
_l
•
=1
J
— -1
_l
J
]
=]
1
1
1
3
zn
i
'
G
C
•
> 100
> 150
>200
ppb
ppb
ppb
No data
F=j
•
No data
No data
No data
No data
1
^M~ 1
_l
No data
1
1
No data
No data
D> 100 ppb
• > 150 ppb
• > 200 ppb
-
-
_
-
No data
I
1
No data
No data
No data
I
j
i 1—1
^^
No data
I
No data
I
J
No data
No data
D> 100 ppb
D> 150 ppb
• > 200 ppb
0 25 50 75 100 125 150 175 200 225 0 25 50 75 100 125 150 175 200 2250 25 50 75 100 125 150 175 200 225
Estimated mean number of daily maximum exceedances of 1-hour NO2 concentrations of 100,150, 200 ppb in a Year
Figure 7-3. Estimated mean number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks
in a year, using recent NO2 air quality (2004-2006) adjusted to just meeting the current annual standard (0.053 ppm). Left
graph: monitors >100m from a major road; Middle graph: monitors >20 m and <100 m from a major road; Right graph:
monitors <20 m from a major road).
106
-------
6 --
-2001-2003,<=20m
- 2001-2003,>20m<100m
-2001-2003,>=100m
-2004-2006,<=20m
-2004-2006,>20m<100m
-2004-2006,>=100m
-2001-2003,<=20m
-2001-2003,>20m<100m
-2001-2003,>=100m
-2004-2006,<=20m
-2004-2006,>20m<100m
-2004-2006,>=100m
-B-2001-2003,<=20m
-A-2001-2003,>20m <100m
-e-2001-2003,>=100m
-»-2004-2006,<=20m
-A-2004-2006,>20m <100m
-»-2004-2006,>=100m
-B- 2001 -2003, <=20m
-A-2001-2003,>20m <100m
-e- 2001 -2003, >=100m
-m- 2004-2006, <=20m
-A- 2004-2006, >20m <100m
-•-2004-2006 >=100m
50
100 150 200 50 100 150 200
98th Percentile Alternative 1-Hour NO2 Standard Level 99th Percentile Alternative 1-Hour NO2 Standard Level
Figure 7-4. Estimated number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks (100
ppb, top; 200 ppb, bottom) in Chicago in a year, using recent NOi air quality data (2001-2006) adjusted to just meeting
alternative 1-hour standard levels (98th percentile, left; and 99th percentile, right) and monitors sited >100 m, > 20 m and < 100
m, < 20 m of major roads.
107
-------
- Phoenix -
- Los Angeles -
01
u
6 --
5 --
r-
uj .E
II
~
4--
3 --
CD o
Q Q.
03 2
°!
'o
1 --
0 -\
-B-2001-2003,<=20m
-A-2001-2003,>20m <100m
-e-2001-2003,>=100m
-«-2004-2006,<=20m
-*-2004-2006,>20m <100m
-»-2004-2006,>=100m
-2001-2003,<=20m
-2001-2003,>20m<100m
-2001-2003,>=100m
-2004-2006,<=20m
-2004-2006,>20m<100m
-2004-2006,>=100m
- Washington DC -
- St. Louis -
2001-2003,<=20m
2001-2003,>20m<100m
2001-2003,>=100m
2004-2006, <=20m
2004-2006,>20m<100m
2004-2006, >= 100m
2001-2003,<=20m
2001-2003,>20m<100m
2001-2003,>=100m
2004-2006,<=20m
2004-2006,>20m<100m
2004-2006,>=100m
50 100 150 200
98th Percentile Alternative 1-Hour NO2 Standard Level
50 100 150 200
98th Percentile Alternative 1-Hour NO2 Standard Level
Figure 7-5. Estimated mean number of daily maximum exceedances of 200 ppb in four locations (Phoenix, Los Angeles,
Washington DC, and St. Louis) in a year, using recent NOi air quality data (2001-2006) adjusted to just meeting alternative 1-
hour 98th percentile standard levels and monitors sited >100 m, > 20 m and < 100 m, < 20 m of major roads.
108
-------
Table 7-23. Estimated annual mean NO2 concentration and the number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks in a year, using 2001-2003 air quality adjusted to just meeting a 1-hour 100 ppb 98th percentile alternative
standard and monitors sited > 100 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Site-
Years1
14
6
9
3
2
6
12
2
16
51
6
26
14
5
3
9
18
612
127
Annual Mean NO2 (ppb)2
Mean
15
13
25
25
24
23
20
26
15
18
16
22
27
31
37
27
26
13
7
Min
5
7
19
25
22
21
14
26
3
4
13
12
20
26
34
22
12
1
1
Med
19
15
23
25
24
23
21
26
11
19
16
20
25
33
38
26
29
13
7
p98
29
16
32
26
27
26
24
26
32
29
19
34
39
34
39
32
35
23
17
p99
29
16
32
26
27
26
24
26
32
29
19
34
39
34
39
32
35
25
18
Number of Daily Maximum Exceedances of 1 -Hour Level3
>100ppb
Mean
2
0
1
2
2
3
3
7
1
1
4
1
3
1
6
3
4
0
0
Min
0
0
0
1
1
1
0
3
0
0
0
0
0
0
4
0
0
0
0
Med
0
0
0
3
2
3
2
7
0
0
3
0
2
1
6
1
3
0
0
p98
15
1
4
3
2
7
8
10
7
5
14
5
15
2
9
15
11
1
6
p99
15
1
4
3
2
7
8
10
7
10
14
5
15
2
9
15
11
4
6
>150 ppb
Mean
0
0
0
0
0
2
0
1
0
0
0
0
0
0
0
0
0
0
0
Min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Med
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
1
0
1
0
0
7
1
1
0
0
2
0
1
0
0
1
0
0
1
p99
1
0
1
0
0
7
1
1
0
5
2
0
1
0
0
1
0
0
1
> 200 ppb
Mean
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
Min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Med
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
p98
1
0
0
0
0
4
1
1
0
0
0
0
1
0
0
0
0
0
1
p99
1
0
0
0
0
4
1
1
0
0
0
0
1
0
0
0
0
0
1
Notes:
1 The average number of monitors operating per year within the three-year group is estimated by dividing the number of site-years by 3.
2 Annual means for each monitor were first calculated based on all simulated hourly values in a year. Then the mean of the annual
means was estimated as the sum of all the annual means in a particular location divided by the number of simulated site-years across the
monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the distribution for the
annual average concentration in any one year within the monitoring period.
3 The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location
divided by the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and
99th percentiles of the distribution for the number of daily maximum exceedances in any one year within the monitoring perod.
109
-------
Table 7-24. Estimated annual mean NO2 concentration and the number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks in a year, using 2001-2003 air quality adjusted to just meeting a 1-hour 100 ppb 98th percentile
alternative standard and monitors sited >20 m and <100 m from a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Site-
Years1
ND
14
6
ND
ND
ND
3
ND
3
35
3
13
7
2
ND
11
10
ND
ND
Annual Mean NO2 (ppb)2
Mean
23
35
27
8
19
27
33
34
26
22
29
Min
12
33
26
4
3
24
23
26
25
13
20
Med
26
35
28
8
19
27
33
33
26
18
31
p98
35
37
28
12
32
29
44
42
27
38
36
p99
35
37
28
12
32
29
44
42
27
38
36
Number of Daily Maximum Exceedances of 1 -Hour Level3
> 100 ppb
Mean
3
7
6
1
1
8
3
5
0
2
2
Min
0
1
2
0
0
4
0
0
0
0
0
Med
2
4
7
0
0
6
1
4
0
0
2
p98
11
21
9
4
7
15
13
11
0
11
5
p99
11
21
9
4
7
15
13
11
0
11
5
> 150 ppb
Mean
0
1
0
0
0
0
0
0
0
0
0
Min
0
0
0
0
0
0
0
0
0
0
0
Med
0
0
0
0
0
0
0
0
0
0
0
p98
0
3
0
0
1
1
2
0
0
0
0
p99
0
3
0
0
1
1
2
0
0
0
0
> 200 ppb
Mean
0
0
0
0
0
0
0
0
0
0
0
Min
0
0
0
0
0
0
0
0
0
0
0
Med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
1
0
0
0
0
0
0
p99
0
0
0
0
1
0
0
0
0
0
0
Notes:
1 The average number of monitors operating per year within the three-year group is estimated by dividing the number of site-years by 3.
2 Annual means for each monitor were first calculated based on all simulated hourly values in a year. Then the mean of the annual
means was estimated as the sum of all the annual means in a particular location divided by the number of simulated site-years across the
monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the distribution for the
annual average concentration in any one year within the monitoring period.
3 The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location
divided by the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and
99th percentiles of the distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
110
-------
Table 7-25. Estimated annual mean NO2 concentration and the number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks in a year, using 2001-2003 air quality adjusted to just meeting a 1-hour 100 ppb 98th percentile
alternative standard and monitors sited <20 m from a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Site-
Years1
ND
5
4
3
2
ND
ND
ND
3
9
3
7
ND
3
ND
6
4
ND
ND
Annual Mean NO2 (ppb)2
Mean
29
26
32
37
32
23
12
30
41
28
33
Min
10
25
31
37
31
18
11
27
40
25
28
Med
32
26
32
37
32
23
12
30
40
29
34
p98
41
27
34
38
32
29
12
33
43
30
36
p99
41
27
34
38
32
29
12
33
43
30
36
Number of Daily Maximum Exceedances of 1 -Hour Level3
> 100 ppb
Mean
2
2
7
7
7
1
1
1
6
3
6
Min
0
0
3
3
1
0
0
0
4
1
3
Med
0
1
7
7
10
0
1
1
4
3
6
p98
6
6
10
10
11
3
1
2
11
6
9
p99
6
6
10
10
11
3
1
2
11
6
9
> 150 ppb
Mean
0
0
0
2
0
0
0
0
0
0
0
Min
0
0
0
1
0
0
0
0
0
0
0
Med
0
0
0
2
0
0
0
0
0
0
0
p98
1
0
0
3
0
0
0
0
0
1
1
p99
1
0
0
3
0
0
0
0
0
1
1
> 200 ppb
Mean
0
0
0
0
0
0
0
0
0
0
0
Min
0
0
0
0
0
0
0
0
0
0
0
Med
0
0
0
0
0
0
0
0
0
0
0
p98
0
0
0
0
0
0
0
0
0
0
0
p99
0
0
0
0
0
0
0
0
0
0
0
Notes:
1 The average number of monitors operating per year within the three-year group is estimated by dividing the number of site-years by 3.
2 Annual means for each monitor were first calculated based on all simulated hourly values in a year. Then the mean of the annual
means was estimated as the sum of all the annual means in a particular location divided by the number of simulated site-years across the
monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the distribution for the
annual average concentration in any one year within the monitoring period.
3 The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location
divided by the number of site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and
99th percentiles of the distribution for the number of daily maximum exceedances in any one year within the monitoring period.
ND no available monitoring data.
111
-------
Table 7-26. Estimated mean number of daily maximum exceedances of 100 ppb 1-hour NO2 concentrations in a year, using 2001-2003 air
quality as is and that adjusted to just meeting the current and alternative standards (98th percentile) for monitors sited >100
m, >20 m and <100 m, and <20 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Sites >100 m of a major road
As
Is
0
0
0
0
2
3
0
1
0
4
0
0
0
0
0
0
0
0
0
Cur
Std
80
3
23
70
99
99
115
152
69
21
106
8
29
32
162
65
61
16
37
Alternative 1 -hour 98tn
percentile standard
50
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
100
2
0
1
2
2
3
3
7
1
1
4
1
3
1
6
3
4
0
0
150
20
3
31
49
37
17
35
74
41
9
41
16
53
89
200
60
58
3
1
200
65
34
106
133
149
73
125
152
114
37
107
72
171
213
327
175
153
18
6
Sites >20 m and <100 m of a major road
As
Is
ND
0
2
ND
ND
ND
2
ND
0
6
0
1
0
0
ND
0
0
ND
ND
Cur
Std
38
74
170
24
31
152
26
44
4
41
57
Alternative 1 -hour 98tn
percentile standard
50
0
0
0
0
0
0
0
0
0
0
0
100
3
7
6
1
1
8
3
5
0
2
2
150
35
90
77
11
15
73
50
84
49
37
57
200
116
212
178
46
53
155
154
216
210
119
168
Sites <20 m of a major road
As
Is
ND
0
0
0
7
ND
ND
ND
1
6
0
1
ND
2
ND
0
0
ND
ND
Cur
Std
19
21
111
48
211
42
78
18
72
56
78
Alternative 1 -hour 98tn
percentile standard
50
0
0
0
0
0
0
0
0
0
0
0
100
2
2
7
7
7
1
1
1
6
3
6
150
29
28
79
68
141
21
28
38
160
46
73
200
104
119
187
224
280
72
76
143
293
175
209
Notes:
The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location divided
by the number of site-years across the monitoring period.
ND No available monitoring data.
112
-------
Table 7-27. Estimated mean number of daily maximum exceedances of 150 ppb 1-hour NO2 concentrations in a year, using 2001-2003 air
quality as is and air quality adjusted to just meeting the current and alternative standards (98th percentile) for monitors sited
>100 m, >20 m and <100 m, and <20 m of a major road.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington
DC
Other MSA
Other Not
MSA
Sites >100 m of a major road
As
Is
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
Cur
Std
18
0
1
5
21
14
13
44
6
2
23
0
1
0
4
5
4
1
6
Alternative 1 -hour 98tn
percentile standard
50
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
100
0
0
0
0
0
2
0
1
0
0
0
0
0
0
0
0
0
0
0
150
2
0
1
2
2
3
3
7
1
1
4
1
3
1
6
3
4
0
0
200
10
1
14
27
17
9
16
41
19
5
23
7
27
41
112
31
30
1
1
Sites >20 m and <100 m of a major road
As
Is
0
0
ND
ND
ND
0
ND
0
0
0
0
0
0
ND
0
0
ND
ND
Cur
Std
3
5
32
2
3
42
1
2
0
3
2
Alternative 1 -hour 98tn
percentile standard
50
0
0
0
0
0
0
0
0
0
0
0
100
0
1
0
0
0
0
0
0
0
0
0
150
3
7
6
1
1
8
3
5
0
2
2
200
19
53
40
5
8
42
25
42
11
21
28
Sites <20 m of a major road
As
Is
ND
0
0
0
1
ND
ND
ND
0
0
0
0
ND
0
ND
0
0
ND
ND
Cur
Std
1
1
14
5
22
4
17
1
2
3
7
Alternative 1 -hour 98tn
percentile standard
50
0
0
0
0
0
0
0
0
0
0
0
100
0
0
0
2
0
0
0
0
0
0
0
150
2
2
7
7
7
1
1
1
6
3
6
200
15
15
50
33
76
7
16
18
88
23
38
Notes:
The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location divided
by the number of site-years across the monitoring period.
ND no available monitoring data.
113
-------
7.3.4 On-Road Concentrations Derived From Ambient Air Quality Adjusted to Just
Meet the Current and Alternative Standards
Just as was done with the as is air quality data, on-road NC>2 concentrations were
estimated using the air quality adjusted to just meeting the current and alternative standard and
the approach described in section 7.2.3. The analysis was performed using the more recent air
quality separated into two three-year groups (2001-2003 and 2004-2006) based on the form of
the potential alternative standards (i.e., a 3-year average) and represents the fourth air quality
scenario. Results are presented in a manner consistent with section 7.3.3, whereby the number of
daily maximum exceedances of the potential benchmark levels was estimated. However, for the
sake of brevity only key figures and tables are provided here. The complete results for the
estimated on-road concentrations and numbers of benchmark exceedances are provided in
Appendix A, section 9.
Figures 7-6 illustrates the estimated mean number of daily maximum exceedances of the
100, 150, and 200 ppb levels on-roads, given air quality adjusted to just meeting the current
annual average standard for two three-year groups. Most locations contained an average of two-
hundred or more estimated daily maximum exceedances of 100 ppb, much greater than those
estimated using either the ambient monitors sited > 100 m, >20 and <100 m, an <20 m of a major
road (Figures 7-2 and 7-3). The estimated numbers of daily maximum exceedances of the 150
and 200 ppb levels were also higher on-roads. Most locations were estimated to contain at least
one-hundred exceedances of 150 ppb and between 50 and 100 exceedances of 200 ppb on-roads
when using air quality concentrations adjusted to just meeting the current standard.
The effect of the potential alternative standards on the estimated on-road NC>2
concentrations was also analyzed at each of the locations. Figure 7-7 illustrates each of the
standard levels (50, 100, 150, and 200 ppb 1-hour) and the two forms (98th and 99th percentiles)
evaluated, again using Chicago as an example to describe patterns in the number of exceedances.
The patterns observed in Figure 7-2 and described in section 7.3.3 for the ambient monitors are
similar to that observed here, albeit with greater numbers of exceedances estimated on-roads
compared with those estimated for monitors near-roads or sited at a distance from major roads.
Estimated numbers of daily maximum concentrations above 100 ppb are between 50 and one
114
-------
hundred considering a standard level of 100 ppb (either percentile), however daily maximum
exceedances of 200 ppb are estimated to be between one and four.
Similar numbers of exceedances on-roads were estimated at other locations using air
quality adjusted to just meeting the potential alternative standards. Figure 7-8 illustrates the
estimated number of exceedances of 200 ppb at four selected locations as an example, Phoenix,
Los Angeles, Washington DC, and St. Louis, using a 98th percentile form of a 1-hour standard.
The number of concentrations above 200 ppb is similar at each of the locations (including
Chicago), particularly when comparing the 100 ppb standard level, ranging from two to seven.
Table 7-28 presents a more comprehensive comparison at this particular standard level (98th
percentile at 100 ppb) using 2001-2003 adjusted air quality at each of the locations. For most
locations, the estimated mean number of daily maximum exceedances of 200 ppb on-roads was
seven or less, with upper percentiles estimated to number about 30 to 70 of exceedances. The
mean number of exceedances of 250 and 300 ppb were less, ranging from a few to tens of
occurrences in a year.
Tables 7-29 and 7-30 summarizes the observed and estimated mean numbers of
exceedances of 100 and 150 ppb on-roads, respectively, using all the recent air quality as is and
that adjusted to just meet the current standard and the potential alternative 98th percentile
standards at each location. Patterns for the estimated on-road concentrations using as is air
quality and air quality adjusted to just meet the current annual standard followed similar patterns
observed for the monitors sited >100 m, > 20 and <100 m, and <20 m of a major road (see
Tables 7-25 and 7-26, for the daily maximum exceedances of 100 and 150 ppb using 2001-2003
air quality). The estimated number of daily maximum exceedances on-roads using the as is air
quality was within the range of estimates provided by the alternative 1-hour 98th percentile
standards of 50 and 100 ppb, while the estimated on-road exceedances of 150 ppb was within the
range of estimated exceedances using the 150 and 200 ppb alternative standard levels.
115
-------
2001 - 2003
2004 - 2006
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
•-1 ;
i i i
— — i — -
i i
i i
• r1
^ '
• ' '
I |
— '
— — '
1 1
^_ r1
1 1
D> 100 ppb
D> 150 ppb
• >200 ppb
.
•-1 i
i i i
No data' i i
i i i
^ ! '
I |
I |
• — r1
^ — '
• ' ' '
i i
^ — H
B~ i i
i i
^= !
— — r1
D2 100 ppb
D> 150 ppb
• 2200 ppb
50
100 150
200
250 300 350 0
50
100
150 200 250 300
350
Estimated mean number of daily maximum exceedances of 1-hour NO2
concentrations of 100,150, 200 ppb on-roads in a Year
Figure 7-6. Estimated mean number of daily maximum exceedances of short-term (1-hour)
potential health effect benchmarks on-roads in a year, using recent NO2 air quality
adjusted to just meeting the current annual standard (0.053 ppm) and an on-road
adjustment factor. Left graph: 2001-2003 air quality; Right graph: 2004-2006 air quality.
116
-------
300
Era
01
250
•= 3
>. o
200
1*
E Q.
150
S 8
"O O
II
I s
w> u
o 100
50
100
200
100 150 200 50 100 150
98th Percentile Alternative Standard Level (ppb) 99th Percentile Alternative Standard Level (ppb)
Figure 7-7. Estimated number of daily maximum exceedances of short-term (1-hour) potential health effect benchmarks (100
ppb, top; 200 ppb, bottom) on-roads in Chicago in a year, using recent NOi air quality (2001-2006) adjusted to just meeting
alternative 1-hour standard levels (98th percentile, left; and 99th percentile) and an on-road adjustment factor.
117
-------
175
- Phoenix -
- Los Angeles -
50
200
100 150 200 50 100 150
98th Percentile Alternative Standard Level (ppb) 98th Percentile Alternative Standard Level (ppb)
Figure 7-8. Estimated mean number of daily maximum exceedances of 200 ppb on-roads in four locations (Phoenix, Los
Angeles, Washington DC, and St. Louis) in a year, using recent NOi air quality (2001-2006) adjusted to just meeting
alternative 1-hour 98th percentile standard levels and an on-road adjustment factor.
118
-------
Table 7-28. Estimated annual mean NO2 concentration and the number of daily maximum exceedances of short-term (1-hour) potential
health effect benchmarks on-roads in a year, using recent air quality (2001-2003) adjusted to just meeting a 1-hour 100 ppb
98th percentile alternative standard and an on-road adjustment factor.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Site-
Years1
1400
600
900
300
200
600
1200
200
1600
5100
600
2600
1400
500
300
900
1800
61200
12700
Annual Mean NO2 (ppb)2
Mean
27
24
45
46
44
41
36
47
27
32
29
39
50
56
67
48
47
23
13
Min
6
9
25
32
28
26
17
33
4
5
16
15
26
33
44
28
15
1
1
Med
29
24
43
45
42
40
35
47
20
32
29
37
46
55
65
47
48
23
12
p98
57
40
75
61
65
59
56
64
69
61
45
70
89
83
96
75
82
48
34
p99
65
41
78
64
67
63
58
67
73
65
46
79
92
88
101
76
88
51
37
Number of Daily Maximum Exceedances of 1 -Hour Level3
> 100 ppb
Mean
46
26
83
104
99
59
89
119
82
33
78
57
116
153
257
125
109
17
6
Min
0
0
2
19
8
4
4
26
0
0
2
0
1
2
60
2
0
0
0
Med
22
13
68
98
82
46
79
118
43
19
74
43
102
152
277
118
99
4
0
p98
190
125
238
209
252
170
231
217
298
142
181
212
284
319
353
274
287
110
64
p99
229
131
257
225
269
186
249
235
307
160
189
226
294
337
358
288
310
133
86
> 150 ppb
Mean
10
3
18
24
19
13
17
32
19
6
19
10
27
35
75
28
27
2
1
Min
0
0
0
0
0
1
0
2
0
0
0
0
0
0
1
0
0
0
0
Med
1
0
7
15
6
8
8
25
3
1
10
3
12
14
51
15
10
0
0
p98
70
22
92
73
86
56
80
97
149
46
92
67
118
182
273
144
139
23
12
p99
104
25
106
86
103
57
91
114
172
55
110
73
137
206
301
153
168
34
17
> 200 ppb
Mean
2
0
4
7
4
5
3
7
5
1
5
3
7
6
18
7
7
0
0
Min
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Med
0
0
0
2
1
3
0
4
0
0
1
0
1
0
4
1
0
0
0
p98
26
3
43
37
36
26
23
34
63
15
37
33
54
44
86
45
56
5
4
p99
38
6
49
38
37
28
24
34
71
20
48
34
68
48
106
51
63
8
8
Notes:
1 The average number of monitors operating per year within the three-year group is estimated by dividing the number of site-years by 300.
2 Annual means for each monitor were first calculated based on all simulated hourly values in a year. Then the mean of the annual
means was estimated as the sum of all the annual means in a particular location divided by the number of simulated site-years across the
monitoring period. The min, med, p98, and p99 represent the minimum, median, 98th, and 99th percentiles of the distribution for the
annual average concentration in any one year within the monitoring period.
3 The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location
divided by the number of simulated site-years across the monitoring period. The min, med, p98, and p99 represent the minimum, median,
98th, and 99th percentiles of the distribution for the number of daily maximum exceedances in any one year within the monitoring period.
119
-------
Table 7-29. Estimated mean number of daily maximum exceedances of 100 ppb 1-hour NO2 concentrations on-roads in a year, using air
quality as is and air quality adjusted to just meeting the current and alternative standards (98th percentile) and an on-road
adjustment factor.
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
2001 -2003 Air Quality
As is
23
5
52
31
89
41
32
13
23
71
7
42
37
101
61
25
36
16
4
Current
std
183
86
193
264
267
282
272
290
189
152
232
129
222
245
338
262
222
133
126
Alternative 1-hour 98tn percentile
standard
50
2
0
4
7
4
5
3
7
5
1
5
3
7
6
18
7
7
0
0
100
46
26
83
104
99
59
89
119
82
33
78
57
116
153
257
125
109
17
6
150
117
95
205
235
232
178
216
233
167
111
174
169
254
293
343
258
221
73
28
200
170
163
281
305
288
265
278
289
218
191
232
249
312
332
351
316
279
138
63
2004-2006 Air Quality
As is
17
2
36
ND
63
20
24
11
15
38
6
35
22
77
51
15
21
10
4
Current
std
193
95
189
257
287
281
295
177
160
202
147
232
284
306
233
207
143
124
Alternative 1-hour 98tn percentile
standard
50
3
0
2
9
13
5
9
6
2
3
3
5
4
13
4
5
0
0
100
58
19
59
148
165
108
127
83
54
56
75
112
146
63
107
96
20
6
150
133
84
176
263
273
229
241
161
155
128
192
237
299
209
226
200
80
28
200
181
153
259
296
313
281
293
210
227
182
264
295
338
298
287
260
143
60
Notes:
The mean number of exceedances represents the sum of daily maximum exceedances occurring at all monitors in a particular location divided by the
number of site-years across the monitoring period.
ND no available monitoring data
120
-------
Table 7-30. Estimated mean number of daily maximum exceedances of 150 ppb
quality as is and air quality adjusted to just meeting the current and
adjustment factor.
1-hour NO2 concentrations on-roads in a year, using air
alternative standards (98th percentile) and an on-road
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
2001 -2003 Air Quality
As is
4
0
9
5
17
9
4
1
4
17
1
7
6
16
9
3
6
2
1
Current
std
110
20
74
134
157
166
168
205
106
55
146
36
87
96
235
131
111
47
63
Alternative 1-hour 98tn percentile
standard
50
0
0
0
1
0
1
0
1
0
0
0
0
1
0
2
1
0
0
0
100
10
3
18
24
19
13
17
32
19
6
19
10
27
35
75
28
27
2
1
150
46
26
83
104
99
59
89
119
82
33
78
57
116
153
257
125
109
17
6
200
95
72
168
196
198
140
180
201
143
81
145
132
217
262
331
224
190
51
19
2004-2006 Air Quality
As is
2
0
5
ND
10
2
3
2
2
6
0
5
2
10
17
1
2
1
1
Current
std
126
25
69
134
189
198
216
99
58
130
45
110
124
192
121
102
59
62
Alternative 1-hour 98tn percentile
standard
50
0
0
0
1
1
0
2
1
0
0
0
0
0
4
0
0
0
0
100
13
1
10
38
50
23
35
22
10
13
14
24
28
20
23
22
2
1
150
58
19
59
148
165
108
127
83
54
56
75
112
146
63
107
96
20
6
200
111
59
138
239
249
197
211
138
122
106
157
204
264
160
194
171
57
19
121
-------
7.4 UNCERTAINTY ANALYSIS
Uncertainty refers to the lack of knowledge regarding both the actual values of model input
variables (parameter uncertainty) and the physical systems or relationships (model uncertainty -
e.g., the shape of the concentration-response functions). In any risk assessment, uncertainty is,
ideally, reduced to the maximum extent possible, but significant uncertainty often remains. It
can be reduced by improved measurement and improved model formulation. In addition, the
degree of uncertainty can be characterized, ranging from qualitative to quantitative assessments.
Uncertainty can be distinct from variability, which commonly refers to the heterogeneity in a
population or variable of interest that is inherent and cannot be reduced through further research.
The approach for evaluating uncertainty was adapted from guidelines outlining how to
conduct a qualitative uncertainty characterization (WHO, 2008). First, the key sources of the
assessment that contribute to uncertainty are identified, and the rationale for why they are
included is discussed. Second, a qualitative characterization follows for the types and
components of uncertainty, resulting in a summary describing, for each source of uncertainty, the
level and direction of influence the uncertainty may have on the air quality characterization
results.
The overall characterization of uncertainty is qualitatively evaluated by considering the
degree of severity of the uncertainty, implied by the relationship between the source of the
uncertainty and the output of the assessment. To the extent possible, an appraisal of the
knowledge base (e.g., the accuracy of the data used, acknowledgement of data gaps) and
evaluation of the decisions made (e.g., selection of particular model forms) is also included in
this uncertainty rating. The characterization is subjectively scaled by the assessors using a
designation of low, medium, and high. Briefly, a low level of uncertainty suggests large changes
within the source of uncertainty would have only a small effect on the results, there is
completeness and scientific consistency in the knowledge base, and decisions made regarding the
particular source of uncertainty would be widely accepted. A designation of medium implies that
a change within the source of uncertainty would likely have a proportional effect on the results,
there may be limited scientific backing, and limited selection of inputs or models to choose from.
A characterization of high implies that a small change in the source would have a large effect on
122
-------
results, there may be inconsistencies present in the scientific support, and assumptions made
would be considered unusual and restrictive by others.
The bias direction indicates how the source of uncertainty has been judged to influence
estimated concentrations, either the concentrations are likely over- or under-estimated. In the
instance where two or more types or components of uncertainty are present that potentially offset
the direction of influence, the bias has been judged as both. An unknown bias has been assigned
where there was no evidence reviewed to judge the direction of uncertainty bias associated with
the source. Table 7-31 provides a summary of the sources of uncertainty identified in the air
quality characterization, the level of uncertainty, and the overall judged bias of each. A
discussion regarding each of these sources of uncertainty and how conclusions were drawn is
given in the sections that follow.
Table 7-31. Summary of qualitative uncertainty analysis for the air quality and health risk
characterization.
Source
Air Quality Data
Ambient
Measurement
Temporal
Representation
Spatial
Representation
Air Quality
Adjustment
On-Road
Simulation
Health
Benchmarks
Type
Database quality
Interference
Scale
Missing data
Years evaluated
Emission source changes
Scale
Monitor objectives
Vertical siting of monitor
Monitor extrapolation < 4m
Proportional approach used
Spatial scale
Temporal scale
Scenario modeled
Spatial scale
Exponential model
Influential factors
Distribution form
Non US studies used
Averaging time
Susceptibility
Concentration/
Exceedance
Bias Direction
both
over
none
both
both
over
unknown
both
under
unknown
both
over
both
over
over
both
unknown
both
unknown
none
under
Characterization
of Uncertainty
Low
Low - Medium
Low
Low - Medium
Low
Low - Medium
Medium
Medium
Low - Medium
Low - Medium
Medium - High
Medium
Medium - High
Low - Medium
Low - Medium
Medium - High
Medium
Low - Medium
Low- Medium
Low - Medium
Medium
123
-------
7.4.1 Air Quality Database
One basic assumption is that the AQS NC>2 air quality data used are quality assured
already. Methods exist for ensuring the precision and accuracy of the ambient monitoring data
(e.g., EPA, 1983). Reported concentrations contain only valid measures, since values with
quality limitations are not entered to the system, removed following determination of being of
lower quality or flagged. There is likely no selective bias in retention of data that is not of
reasonable quality if the data are in error, it is assumed that selection of high concentration poor
quality data would be just as likely as low concentration data of poor quality. Given the numbers
of measurements used for this analysis, it is likely that even if a few low quality data are present
in the data set, they would not have any significant effect on the results presented here. There
are no alternative data sets available that are as comprehensive, and where monitoring data are
available that are not included in the AQS, it is expected that given the same methods and quality
assurances, would be complimentary to the data existing in the AQS. Therefore, the air quality
data and database used likely contributes minimally to the uncertainty level, there is low
uncertainty in the knowledge base, and the uncertainty in the subjectivity of choices is also
considered low.
Temporally, the data are hourly measurements and appropriately account for variability
in concentrations that are commonly observed for NC>2 and by definition are representative of an
entire year. In addition, having more than one monitor does account for some of the spatial
variability in a particular location. However, the degree of representativeness of the monitoring
data used in this analysis can be evaluated from several perspectives, one of which is how well
the temporal and spatial variability are represented. In particular, missing hourly measurements
at a monitor may introduce bias (if different periods within a year or different years have
different numbers of measured values) and increase the uncertainty. Furthermore, the spatial
representativeness will be poor if the monitoring network is not dense enough to resolve the
spatial variability (causing increased uncertainty) or if the monitors are not appropriately
distributed to reflect population exposure (causing a bias). Additional uncertainty regarding
temporal and spatial representation by the monitors is expanded below.
124
-------
7.4.2 Measurement Technique for Ambient
One source of uncertainty for NC>2 air quality data is due to interference with other
oxidized nitrogen compounds. Nitrogen dioxide is not directly measured. It is estimated by
subtracting measured NO from total nitrogen oxides. The ISA identifies several studies
conducted that have shown a constant positive interference when oxides of nitrogen other than
NO2 are present (ISA section 2.3). Most commonly the interference is from HNOs and has been
reported to contribute to up to 50% the calculated NO2. This has been shown to occur primarily
during the afternoon hours in the summer and would result in an overestimation of ambient NO2
concentrations. During winter, positive interference in the measurement of NO2 is estimated to
be less, generally at 10% or lower. At any one particular site, however, there is uncertainty in
how much the interference will be, and is dependent on the presence of the NOZ compounds
which are largely not measured. In addition, it is not known whether there is a concentration
dependence on the amount of interference. This is an important uncertainty when air quality
concentrations adjusted upwards to just meet the current standard. Therefore, the bias would be
a consistent overestimation of NO2 concentrations, the level of which may range from affecting
the concentrations minimally upwards to a moderate effect. While the science demonstrating the
interference is consistent, there remain uncertainties about the application of the level of
interference to individual monitoring sites.
7.4.3 Temporal Representation
Data are valid hourly measures and are of similar temporal scale as identified health
effect benchmark concentrations. There are frequent missing values within a given valid year
which contribute to the uncertainty as well as introducing a possible bias if some seasons, day
types (e.g., weekday/weekend), or time of the day (e.g., night or day) are not equally represented.
Since a 75 percent daily and hourly completeness rule was applied, some of these uncertainties
and biases were reduced in these analyses. Additional validity criteria could have included
completeness for monitoring based on quarters, rather than the entire year. This would screen
for air quality data potentially missing an entire season of monitoring. The use of validity
criteria that included quarterly completeness would likely exclude a few monitor site-years of
data considered in the current analysis. This would likely have a greater effect at locations with
125
-------
fewer site-years of air quality if large numbers of missing data exist in a quarter, although in
some locations, actual seasons may be something other than a quarterly classification.
Ambient monitoring data were not interpolated to substitute for any missing values. It is
assumed that missing values are not systematic, i.e., high concentrations would be absent as well
low concentrations in equal proportions. There are methods available that can account for time-
of-day, day-of-week, and seasonal variation in ambient monitoring concentrations. However, if
a method were selected it would have to not simply interpolate the data, but accurately estimate
the probability of peak concentrations as well. It was judged that if such a method was available
or one was developed to substitute data, it would likely add to a similar level of uncertainty as
not choosing to substitute the data. Again, this can be viewed as having a limited affect on
uncertainty because using the validity criteria should select for the most representative and
complete ambient monitoring data sets possible.
There may be bias and uncertainty in the air quality characterization results if the years
monitored vary significantly between locations. Although monitoring locations within a region
do change over time, the NC>2 network has been reasonably stable over years 1995-2006,
particularly at locations with larger monitoring networks. While it is possible for monitors to
move from high concentration areas to low concentration areas or perhaps in the other direction,
regulations exist that specifies the design and measurement requirements for these networks
(e.g., 40 CFR Part 58). Given this, it is expected that the level of uncertainty in the specific
monitors operating from year to year is low with a variable bias direction of over-estimation for
some years and under-estimation for others.
It should also be noted that use of the older data in some of the analyses here assumes
that the sources present at that time are the same as current sources, adding uncertainty to results
if this is not the case. Separating the data into two 6-year groups (historical and recent for the as
is evaluation) and two additional subsets of the recent air quality (2001-2003 and 2004-2006)
before analysis reduces the potential impact from changes in national- or location-specific source
influences and is judged to have a minimal bias in representing air quality concentrations for
those selected years. There is some variability expected from year-to-year, that is, there may be
differences in the air quality results if the year-groups included a different 3-year period, such as
2002-2004 or 2003-2005. Deciding to bound the total period rather than characterize all possible
126
-------
3-year combinations was judged appropriate, given the small differences in the observed results
over time and the resources available for the analysis.
7.4.4 Spatial Representation
Relative to the physical area, there are only a small number of monitors in each location.
Even considering sparse siting, the monitoring data are assumed to be spatially representative of
the locations analyzed here because the monitors are used in determining whether areas meet or
do not meet the NAAQS. This could include areas between the ambient monitors that may or
may not be influenced by similar local sources of NC>2. For these reasons the uncertainty and
bias due to the spatial network may be moderate, although the monitoring network design should
have addressed these issues within the available resources and other monitoring constraints. Bias
will depend on ambient monitoring objectives and scale and whether there is large variability in
monitoring surface, i.e., areas of differing terrain that are not be well represented by the
distribution of monitors. The direction of this bias is largely unknown due to the differences in
the true representativeness of the network and the particular terrain in each location. In addition,
the air quality characterization used all monitors meeting the 75 percent completeness criteria,
without taking into account the monitoring objectives or land use for individual monitors. Thus,
there will be some lack of spatial representation and uncertainty due to the inclusion/exclusion of
some monitors that are very near local sources (including mobile sources) potentially resulting in
either over- or under-estimations.
According to one study conducted in the South Bronx, NY in November and December
2001, negative vertical gradients can exist for monitor concentrations (ISA, section 2.5.3.3). On
average, measured concentrations can be 2.5 times higher at 4 meters than at a 15 meter vertical
monitor siting. Therefore, monitors positioned on building rooftops may underestimate NC>2
concentrations at lower vertical heights and possibly at the standard breathing height of 1.8 m.
In this REA, only 7 of the 17810 monitors in the named locations contained monitoring heights of
15 meters or greater, with 58% at 4 meters or less height, and 79% at 5 meters or less in height.
In the aggregate locations (i.e., Other MSA, Other Not MSA), a total of 4 monitors of 34011
contained monitoring heights of 15 meters or greater, with 50% at 4 meters or less in height, and
10 26 monitors in the named locations did not have height reported (therefore, 178 + 26 = 204 total number of
monitors).
11 84 monitors in the aggregate locations did not have height reported (therefore, 340 + 84 = 424 total number of
monitors).
127
-------
73% at 5 meters above ground. Not accounting for this potential vertical gradient in NC>2
concentrations may generate underestimates of exceedances for some sites, however the overall
impact of inferences made for the locations included in this assessment without considering
vertical monitor height is likely minimal since most monitors used for analysis are sited at less
than 4-5 meters above ground. In addition, the relationship at heights below 4 meters is largely
uncertain (e.g., a breathing height of 1.8 meters is commonly used) and therefore would add an
additional unknown bias to the estimated NC>2 concentrations above a benchmark when used as a
surrogate for human exposure.
Staff evaluated the potential bias in vertical siting of monitors by using the AERMOD
predictions at four receptor heights for each of the three ambient monitors located in the Atlanta
exposure modeling domain (monitors 130890002, 130893001, and 131210048), each located at
heights 0, 1.8, 5, and 15 meters above ground. An example of the predicted hourly NC>2
concentration distributions at each receptor height for one site (ID 131210048) is presented in
Figure 7-9, and is similar to that predicted at each of the other monitors. Consistent with the one
study reported in the ISA, the estimated concentrations at the 15 meter monitoring height were
the lowest, with progressively greater concentrations with decreasing receptor height. However,
the level of differences in concentration at each of the different receptor heights were lower
using the modeled concentrations in Atlanta than when compared with those reported for the
South Bronx using the measured ambient concentrations. On average, the NC>2 concentrations
estimated at a 0 m height were >0.2% than those at a height of 1.8 m, while the largest difference
in concentration (7.9%) occurred with comparison of the 0 m to the 15m height (Table 7-32).
When comparing concentrations for each of the receptor heights at the upper tails of the
distribution, occasionally a similar pattern was observed, i.e., small increases in the numbers of
exceedances of a selected benchmark level (Table 7-33). At one location however (ID
130893001), there were no differences in the number of exceedances of the selected benchmark
level.
These modeling results support the ISA cited measurement study in that there is an
inverse relationship between monitor vertical siting height and NC>2 concentration, only the
magnitude of the relationship differs. The lack of a similar magnitude could be the result of
several factors such differing influential features of the study area versus the modeled area (e.g.,
seasonal/meteorological factors, presence of nearby sources, terrain) or perhaps a limited
128
-------
sensitivity of the model to variable receptor height. Based on these two limited evaluations,
there can be no clear determination as to whether the monitor vertical siting effect is as large as
the single study estimated or as small as the dispersion model predicted. Further, since there are
limited measurement and model results available to inform a decision and that there were few
monitors used above 5 m in vertical height, staff did not adjust concentrations using vertical
siting characteristics.
Monitor 131210048
AERMOD Predicted N02 (ppb)
Om Height
1.8m Height
5 m Height
15 m Height
20 40 60 80 100
AERMOD Predicted N02 (ppb)
120
140
160
Figre 7-9. Distribution of 1-hour NOi concentrations for three modeled receptors in
Atlanta at different vertical heights, using AERMOD predicted 2002 air quality.
Table 7-32. Percent difference in 1-hour NO2 concentrations for three modeled receptors in
Atlanta at different vertical heights, using AERMOD predicted 2002 air quality.
Monitor ID
130890002
130893001
131210048
% Difference in 1-hour NO2
(Lower to higher vertical height)
Om to 1.8 m
0.2
0.1
0.1
1.8 m to 5 m
0.9
0.6
1.6
0 m to 15 m
6.7
4.6
7.9
129
-------
Table 7-33. Number of 1-hour NO2 concentrations above 100 ppb for three modeled receptors in
Atlanta at different vertical heights, using AERMOD predicted 2002 air quality.
Monitor ID
130890002
130893001
131210048
Number of 1 -hour NO2 > 100 ppb
(AERMOD Estimated NO2, Year 2002)
0 m
21
2
12
1.8m
21
2
12
5 m
21
2
11
15m
19
2
10
7.4.5 Air Quality Adjustment Procedure
There is uncertainty in the air quality adjustment procedures due to the uncertainty of the
true relationship between the adjusted concentrations that are simulating a hypothetical scenario
and the as is air quality. The adjustment factors used for the current and alternative standards
each assumed that all hourly concentrations will change proportionately at each ambient
monitoring site. Two principal uncertainties are discussed, namely uncertainty regarding the
proportional approach used and the universal application of the approach to all ambient monitors
within each location.
Different sources have different temporal emission profiles, so that equally applied
changes to the concentrations at the ambient monitors to simulate hypothetical changes in
emissions may not correspond well with all portions of the concentration distribution. When
adjusting concentrations upward to just meeting the current standard, the proportional adjustment
used an equivalent multiplicative factor derived from the annual mean concentration and equally
applied to all portions of the concentration distribution, i.e., the upper tails were treated the same
as the area of central tendency. This may not necessarily reflect changes in an overall emissions
profile that may result from, for example, an increase in the number of sources in a location. It is
possible that while the mean concentration measured at an ambient monitor may increase with an
increase in the source emissions affecting concentrations measured at the monitor, the tails of the
hourly concentration distribution might not have the same proportional increase. The increase
could be greater or it could be less than that observed at the mean, dependent largely on the type
of sources and inherent operating conditions. Adjusting the ambient concentrations upwards to
simulate the alternative standards also carries a similar level of uncertainty although the
multiplicative factors were derived from the upper percentiles of the 1-hour NC>2 concentrations,
rather than the mean, and then applied to the 1-hour NC>2 concentrations equally.
130
-------
In each of these instances of adjusting the concentrations upwards, one could argue that
there may be an associated over-estimation in the concentrations at the upper tails of the
distributions, possibly leading to over-estimation in the numbers of exceedances of benchmark
levels. An analysis was performed on monitors within six of the locations used in the air quality
characterization to investigate how distributions of hourly nitrogen dioxide concentrations have
changed over time (Rizzo, 2008). The analysis indicates that a proportional approach can be
appropriate in simulating higher concentrations at most monitoring sites, since historically, NC>2
concentrations have decreased linearly across the entire concentration distribution at each of the
monitoring sites and locations evaluated. In addition, when adjusting concentrations downward
(e.g., the alternative standard level of 50 ppb 1-hour, 99th percentile), the use of the proportional
multiplicative adjustment derived from the upper tails and applied to concentration distribution
may be less uncertain because NC>2 concentrations have been observed to decrease linearly over
time, and only assumes that the downward trend would continue similarly in the future with
added source controls.
At some of monitoring sites analyzed however, there were features not consistent with a
completely proportional relationship, including deviation from linearity primarily at the
maximum or minimum percentile concentrations, some indication of curvilinear relationships,
and the presence of either a positive or negative regression intercept (Rizzo, 2008). Where
multiple monitors were present in a location there tended to be a mixture of each of these
condition, including proportionality. Not all of the locations analyzed as part of the air quality
characterization were included in the evaluation. It was also assumed that the analysis conducted
at the six locations would reflect what would be observed at the other locations if evaluated for
trends in concentration over long periods of time. High concentration year to low concentration
year comparisons were also limited to 3-years to generate appropriate and a manageable number
of comparisons. It was assumed that if additional years of data were compared that similar
relationships would be developed. Further, there is uncertainty in adjusting concentrations
upwards or downwards considering assumptions regarding future source emission scenarios and
how these would relate to observed trends in current and historical air quality. The uncertainty
about future source emission scenarios is largely unknown.
Universal application of the proportional simulation approach for each of the locations
and within each location was done for consistency and was designed to preserve the inherent
131
-------
variability in the concentration distribution. There is uncertainty regarding emission changes
that would affect the concentrations at the design monitor containing the highest concentration
(annual mean, 98th or 99th percentile 1-hour) that may not necessarily affect lower concentration
sites proportionately. This could result in either over- or under-estimations in the number of
exceedances at lower concentration sites within a location where the current or alternative
standard scenarios were evaluated. When comparing the low concentration years and the high
concentration years at multiple ambient monitors within a location however, most monitors
contained similar linear relationships (e.g., comparative regression slopes and intercepts). For
example, Figure 7-10 shows the daily maximum NCh concentration percentiles for four ambient
monitors in Philadelphia, where each of 6 ambient monitors were in operation for years 1984 and
2007. The similarity in slope for each of the monitors indicates that an adjustment factor derived
from one ambient monitor can be applied to the other monitors in the monitoring network.
Furthermore, when calculating the number of exceedances of the potential health effect
benchmark levels, the greatest numbers of exceedances typically were noted at the monitoring
sites with the highest concentrations within the location (Appendix A, section 7), with little
contribution from the low concentration sites within a location. A few locations though were
noted that may have an exceptional number of estimated exceedances as a result of the air quality
adjustment approach, particularly those locations with few monitoring sites that contained very
low concentrations and/or atypical variability in hourly concentrations. These few locations
(e.g., Miami, Jacksonville, and Provo) may contain overestimations at the upper tails of the
concentration distribution, leading to bias in estimated number of exceedances at both the upper
percentiles and the mean when using the air quality simulated to just meet the current and most
of the alternative standards. It should also be noted that where deviations from proportionality
occur, the magnitude of the uncertainty in the results is likely related to the magnitude of the
extrapolation to the adjusted concentration level. This means that there is likely greater
uncertainty in the results for evaluating the current annual and the 200 ppb 98th percentile
alternative standards, than when considering the 50 ppb and 100 ppb 99th percentile alternative
standard.
Given the limited deviations in linearity and proportionality at each monitor site that may
result in both over- or under-estimations in concentrations following either an adjustment
upwards or downwards and the limited time and resources available to develop a new universal
132
-------
approach that addresses each of the observed deviations, staff judged the proportional approach
used to simulate just meeting the current and alternative standards as adequate and appropriate
for the scenario considered.
High Year: 1984 Low Year: 2007
0.00 0.05 0.10 0.15 0.20
o 0 15 -f
o
D_
>- 0.10 -
0.05 -
420910013
R"2: 0.84
340070003
R"2: 0.94
421010004
R"2: 0.91
420170012
R"2: 0.89
421010047
R"2: 0.96
420450002
RA2: 0.91
0.00 0.05 0.10 0.15 0.20
0.00 0.05
0.15 0.20
High Year Percent!le Cone, (ppm)
Figure 7-10. Comparison of measured daily maximum NO2 concentration percentiles in
Philadelphia for one high concentration years (1984) versus a low concentration years
(2007) at four ambient monitors.
7.4.6 On-Road Concentration Simulation
On-road and ambient monitoring NC>2 concentrations have been shown to be correlated
significantly on a temporal basis (e.g., Cape et al., 2004) and motor vehicles are a significant
emission source of NOX, providing support for estimating on-road concentrations using ambient
monitoring data. The relationship used in this analysis to estimate on-road NC>2 concentrations
was derived from data reported in measurement studies containing mostly long-term averaging
times, typically 7-14 days or greater in duration (e.g., Roorda-Knape, 1998; Pleijel et al., 2004;
Cape et al, 2004). One study was conducted over a one-hour time averaging period however the
results were reported for time-averaging of at least 1-day (Rodes and Holland, 1981). Use of
such data is considered appropriate in this analysis to estimate on-road hourly concentrations
from hourly ambient measures, assuming a direct relationship exists between the short-term
- 0.15
- 0.05
133
-------
peaks and time-averaged concentrations (e.g., hourly on-road NO2 concentrations are correlated
with 24-hour averages). While this should not impact the overall contribution relationship
between vehicles and ambient concentrations on roads, the relationship will likely differ for
shorter averaging times. However, the longer-term data used to develop the algorithm were
likely collected during variable conditions (e.g., shifting wind direction, variable diurnal rate of
transformation of NO to NO2) than would be observed across shorter time periods. Therefore,
distribution of the adjustment factors based on 7-14 day averaging-times may be biased at the
tails, that is, upper percentile values used may be biased low and the lower percentiles may be
biased high. This could result in either over- or under-estimations of 1-hour NO2 concentrations,
depending on the time of day. In addition, the ambient concentration level could not be
considered in the application of the on-road factor to the hourly concentration because the
relationship between the derived adjustment factor and the ambient concentration on that
temporal scale is unknown. If there is a concentration dependent relationship, this would bias
the estimated on-road concentrations with unknown direction. Application of the on-road
concentration estimation also assumes that concentration changes that occur on-roads and at the
monitor are simultaneous (i.e., within the hour time period of estimation). While this may not be
the case, because time-activity patterns of individuals are not considered in this air quality
benchmark characterization, there would be no bias in the number of estimated exceedances.
If assessing personal exposures to individuals near roadways or within vehicles that are
traveling on a road, it is likely that their exposure concentrations would also vary due to differing
roadway concentrations. There was limited data available in the development of the on-road
adjustment factor for differing road-types and most of the data used were reported for urban
areas (Appendix A, Table A-108). The factors developed should be appropriate for use in urban
areas, as was done in this REA, however representation of all road-types (e.g., freeways,
arterials, or local roads) in the urban areas modeled is largely unknown. On-road concentrations
were not adjusted in this analysis to account for in-vehicle penetration and decay. Therefore, in-
vehicle concentrations would be overestimated if using the estimated on-road concentrations as a
surrogate, given that reactive pollutants (e.g., PM2 5) tend to have a lower indoor/outdoor (I/O)
concentration ratio (Rodes et al., 1998). One study reported mean (I/O) ratios of NO2 for a few
roadways and driving conditions in Hong Kong (Chan and Chung, 2003). On highways and
134
-------
urban streets, the I/O is centered about 0.6 to 1.0, indicating removal of outdoor NO2 as it enters
inside a vehicle.
At locations where traffic counts are very low (e.g., on the order of hundreds/day) the on-
road contribution has been shown to be negligible (Bell and Ashenden, 1997; Cape et al., 2004),
therefore any monitors sited in rural areas near roads with minimal traffic volumes could result
in small overestimations of NC>2 concentrations when using equation (7-2) at these locations.
This is not of great concern because most of the monitors used in the on-road simulation were
sited within large CMSA/MSA, likely encompassing urban/suburban features of a location rather
than rural areas. Monitors sited within 100 m of the roadway in the named locations were not
used in the calculation of on-road concentrations due to the possibility of these monitors already
accounting for notable impact from vehicle emissions (e.g., Beckerman et al., 2008), thus
controlling for over-estimating the on-road concentrations. However, there is potential for
influence by non-road source emissions on the measured concentrations at the monitors used (>
100 m from a major road), contrary to an assumption that there is an absence of direct source
influence (only that mobile sources were controlled for by selecting these monitors). Therefore,
if using ambient monitors directly affected by emissions from non-road sources, the simulated
on-road concentrations may be over-estimated. For example, the estimated number of on-road
exceedances was greater for Jacksonville and Provo than at the other locations, even though all
on-road simulations used monitors sited >100 m from a major road. The predominant land use
however for both of these monitors was commercial, although monitoring objective and
measurement scale were unknown. In addition, the ambient monitors in the aggregate locations
(i.e., Other MSA, Other Not MSA) did not have distances to major roads calculated and may
include a number of site-years of data from monitors <100 m from a major road. The estimated
number of daily maximum on-road exceedances may be over-estimated for these aggregate
locations.
Another source of uncertainty in the spatial heterogeneity of NO2 concentrations regards
the presence of street canyons on roadways. These localized areas may be subject to highly
variable and higher mean concentrations within a short span of a road, often defined by the
presence of man-made structures, such as buildings, on both sides of the road. In one study, a
comparison of street canyon measured NOX concentrations with those measured at a reference
site (termed background) indicated that there is about a factor of 2.3 difference in the
135
-------
concentrations (Ghenu et. al, 2007). Vardoulakis et al. (2004) reported mean NO2 concentrations
at a major intersection can be a factor of about 2.1 times greater than on-road concentrations
measured at a few hundred meters distance within a street canyon.12 Because these factors are
within the range of adjustment factors used here in estimating the on-road concentration, i.e.,
ranging from a factor of 1.2 to 3.7 times the ambient concentrations, it is likely that some of the
estimated on-road concentrations in the air quality characterization are similar in magnitude to
those found in street canyons.
To represent the relationship between the on-road concentrations and the concentrations
measured at a distance, a simple exponential model was used. The selection of an exponential
model was based on independent peer-reviewed studies that reported this type of relationship
using NO2 measurements. There are uncertainties and possible biases with the selection of the
exponential model. For example, NOX is primarily emitted as NO (e.g., Heeb et al., 2008;
Shorter et al., 2005), with substantial secondary formation due predominantly to NO + Os -^
NO2 + O2. Numerous studies have demonstrated the Os reduction that occurs near major roads,
reflecting the transfer of odd oxygen to NO to form NO2, a process that can impact NO2
concentrations both on- and downwind of the road. Some studies report NO2 concentrations
increasing just downwind of roadways and that are inversely correlated with Os (e.g., Beckerman
et al., 2008), suggesting that peak concentration of NO2 may not always occur on the road, but at
a distance downwind. While an exponential model may fit well (or for portions of the data), the
peak may be occurring at a distance from the road rather than on the road. Model convergence
was one criteria used in selecting for useful parameter estimates. One of the principal reasons
for lack of convergence is that the measurement data did not fit the exponential form considered.
Therefore, if the study measurement data contained peak concentrations at a distance of the road
and other lower concentrations closer to the road and along the transect, it was likely that there
were no valid parameters estimated for that study data. It follows that for studies where the
nearest roadway distance was not at the edge of the road and the overall concentrations pattern is
to increase with decreasing distance from the road, the estimated on-road adjustment factors may
be biased high. This would occur when the concentration pattern followed the exponential
model well, peaking at the nearest road measurement, but in the absence of additional
measurements closer to the road, the model assumes a further increase in concentrations with
12 Ambient concentrations at a site not influenced by mobile sources were not reported by Vardoulakis et al. (2004).
136
-------
decreased distance to the road. Therefore, the uncertainty regarding where the peak
concentration occurs (on-road or at a distance from the road) in combination with the form of the
exponential model (the highest concentration occurs at zero distance from road) and the selection
of studies that fit the exponential model, may also add a moderate level of uncertainty in the
estimated on-road concentrations and the number of exceedances.
The manner in which the on-road adjustment factor distribution was constructed and
applied also introduces uncertainty to the results. Based on the few influential variables
available from the on-road studies, the number of values derived for the adjustment factors
(n=41), a comparison of the distributions that would result in considering the potentially
influential variables, and considering how the factors would be applied to the ambient
monitoring data, it was decided that a two season categorization was the most appropriate
characterization of the derived data (Appendix A, section 8.2). There is some bias in the
application of the season categorization due to the presence of on-road adjustment factors
derived from annual average data within the not summer distribution. Staff judged that on-road
adjustment factors that included four seasons (spring, fall, summer, and winter) would be more
reflective of expected conditions during non-summer months rather than during summer months.
Rather than excluding 27% of the derived ratios, staff decided to retain the ratios and include
them in the not summer category.
Using an un-weighted mixture of urban and rural on-road adjustment factors within each
season category assumes that the distribution of each appropriately reflects the balance of these
factors within each location. First, some of the studies used in developing the on-road
adjustment factors did not distinctly report whether the data were from urban, suburban, or rural
areas. There is a moderate level of uncertainty in the judgment by staff in characterizing the
reported study data as urban or rural, an uncertainty that was used to support the decision to not
characterize the on-road adjustment factor distribution based on this potentially influential factor.
Second, the values for the m ratios derived from studies either described as by the original
researchers or inferred by staff as rural areas (min 0.36, median 0.89, max 2.44) were
comparable to those described as by the original researchers or inferred by staff as urban areas
(min 0.21, median 0.74, max 2.70). And finally, given the schedule and resources available to
produce the REA, the uncertainty in the categorization by area type, and little difference between
137
-------
the potential distributions, staff decided not to stratify the on-road adjustment factors by area
type.
Each season category was represented by an empirical distribution, with each value from
the distribution having an equal probability of selection. While there may be other distribution
forms that could be alternatively selected, staff judged that use of a fitted distribution would not
improve the representation of the true population of m values compared with a empirical form,
and that there would likely be minimal effect on the estimated number of exceedances. Use of
an empirical distribution was done because neither season group of data could be assigned to a
particular distribution type (e.g., normal, lognormal, exponential, gamma; see Figures 7-11 and
7-12), although the summer data set was significantly different from a normal distribution
(p<0.01). While there is uncertainty associated with the use of the empirically-derived data in
representing the true population of m values, assuming a fitted distribution is not without its own
uncertainties. For example, using a lognormal distribution may underestimate the observed
frequency of certain values of m (Figures 7-11 and 7-12), and while allowing for values outside
of the empirical distribution, would still need to have realistic bounds placed on the minimum
and maximum values, further adding to uncertainty regarding the shape and form of the fitted
distribution. In addition, allowing for the selection of on-road adjustment factors outside of the
range of the empirical data using a fitted distribution would have a low frequency such that the
overall impact to the estimated on-road NC>2 concentrations would likely be limited (Table 7-34).
Furthermore, the m ratios were derived from measurement data, they are not actual
measurements but are measurement-based. Fitting a distribution from the modeled data may also
add to the existing model uncertainty. Each of these factors mentioned (the number of samples,
uncertainty in the limits and shape of the distribution, fitting distributions to modeled data) were
considered and it was decided by staff that the empirical distribution derived from the
measurement data would be most representative.
Staff did however investigate the affect on the number of exceedances when using an
alternative fitted distribution. Lognormal distributions were selected, with lower and upper
bounds of the on-road adjustment factor defined by the 0.5th and 99.5th percentiles (Table 7-34).
On-road adjustment factors were obtained by sampling from the fitted m distribution for each
season13, and then adding 1 (see equation 7-2). The on-road simulation was performed using this
13 {geometric mean, geometric standard deviation}: Summer {0.989, 1.65}, Not summer {0.643, 1.73}.
138
-------
fitted distribution in the same manner done using the empirical distribution (section 7.2.4) using
2004-2006 air quality. The average number of estimated exceedances at 100, 200, and 300 ppb
were compared with those generated using the empirical distribution by taking the difference of
the two exposure estimates. In general, the difference between the estimated number of
exceedances obtained using the two different distributions was small when considering
unadjusted air quality (Table 7-35). Most locations had between 0 to 3 additional daily
maximum on-road exceedances at the 100 ppb benchmark using the fitted lognormal
distribution, with no difference in daily maximum exceedances at the 200 ppb or 300 ppb level.
Differences between the two on-road simulations were similarly small considering air quality
adjusted to just meet the current annual standard, although differences were also present at the
higher benchmark levels. In addition, the differences in the two simulations were variable at
each location, that is, at some locations, the lognormal distribution generated a greater number of
exceedances (e.g., El Paso, New York), while at other locations a fewer number of exceedances
(e.g., Atlanta, Detroit) than when using the empirical distribution.
139
-------
Table 7-34. Comparison of empirical distribution of on-road adjustment factors used in on-road
concentration estimation with a fitted lognormal distribution.
Season Group
Not Summer
Summer
Percentile
0.5
1.0
2.3
2.5
5.0
94.4
95.0
97.5
99.0
99.5
0.5
1.0
2.5
5.0
7.8
95.0
97.5
97.8
99.0
99.5
Distribution Form
Empirical
1.22
1.22
1.22
1.22
1.22
2.50
2.54
2.54
2.54
2.54
1.49
1.49
1.49
1.51
1.51
3.45
3.70
3.70
3.70
3.70
Lognormal
1.16
1.18
1.22
1.22
1.26
2.54
2.59
2.88
3.30
3.64
1.27
1.31
1.37
1.44
1.49
3.25
3.63
3.70
4.16
4.58
Notes:
Bold font indicates percentile where empirical distribution
minimum and maximum intersect with a fitted lognormal
distribution.
On-road adjustment factors are 1+m, see section 7.2.4 of
REA.
140
-------
season2=Hot Summer
0.0
O.G
1 .2
1 .8
2.4
3.0
Curves: ~ Normal(Mu=0.734 Sigma=0.3769)
"" Lognormal(Theta = 0 Shape=.B5 Scale=-.4)
Ueibull(Theta=0 Shape=2.1 Scale=.B3)
Gamma(Theta=0 Shape=3.92 Scale=0.19)
Figure 7-11. Comparison of the distribution of estimated C/C/, ratios or m for the not
summer category with fitted distributions.
season 2 =Summer
0.0
O.G
1 .2
1 .8
2.4
3.0
Curves: ~ Normal(Mu=1.1271 Signa=0.6553 )
~~ Lognormal(Theta=0 Shape=0.5 Scale=0)
Ueibul HTheta=0 Shape=1.9 Scale=1.3)
Ganna(Theta=0 Shape=3.99 Sca1e=0.28)
Figure 7-12. Comparison of the distribution of estimated Cv/Cb ratios or m for the summer
category with fitted distributions.
141
-------
Table 7-35. Absolute difference in the estimated number of exceedances of potential health effect
benchmarks on-roads using either a fitted lognormal distribution or empirical
distribution of the on-road adjustment factors and 2004-2006 air quality as is and air
quality adjusted to just meet the current annual standard.
Location
Atlanta
Boston
Chicago
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Other MSA
Other Not MSA
Difference in Mean Number of Daily Maximum
Exceedances2
As Is
>100
ppb
0
0
-1
-1
-1
-2
0
0
0
-2
-3
-1
0
-9
1
0
0
0
>200
ppb
0
0
0
0
0
0
0
0
0
0
0
0
0
-1
0
0
0
0
>300
ppb
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Current Annual Standard
>100
ppb
0
0
0
0
1
-1
-1
0
0
-1
-1
0
0
-3
0
0
0
0
>200
ppb
2
2
-2
0
2
-1
-2
1
1
0
-3
-1
-3
-3
3
1
0
1
>300
ppb
1
0
0
-1
3
-1
-4
1
0
-1
-1
-1
2
-11
-1
-1
0
1
Notes:
1 Differences are obtained by subtracting on-road exposure results using fitted
lognormal distribution from the results obtained using an empirical distribution.
Another source of uncertainty is the extent to which the near-road study locations used to
derive the on-road simulation factors represent the locations in these analyses. The on-road and
near-road data were collected in a few locations, most of them outside of the United States. The
source mixes (i.e., the vehicle fleet) in study locations may not be representative of the U.S. fleet.
Without detailed information characterizing the emissions patterns for the on-road study areas,
there was no attempt to match the air quality characterization locations to specific on-road study
areas, which might improve the precision of the estimates. When considering the two U.S.
studies containing the required measurement data (Rodes et al. 1981; Singer et al.; 2004), three m
ratios were estimated (i.e., 0.93, 1.54, and 2.43) similar in range with the ratios estimated using
data obtained from non-U.S. studies. This evidence implies that the level of uncertainty in
142
-------
applying the non-US studies for the purposes of this analysis may not be large, although there is
limited data available to make this judgment.
7.4.7 Health Benchmark
The choice of potential health effect benchmarks, and the use of those benchmarks to
assess risks, can introduce uncertainty into the risk assessment. For example, the potential health
effect benchmarks used were based on studies where volunteers were exposed to NC>2 for
varying lengths of time. Typically, the NC>2 exposure durations were between 30 minutes and 2
hours. This introduces some uncertainty into the characterization of risk, which compared the
potential health effect benchmarks to estimates of exposure over a 1-hour time period.
Therefore, the use of a 1-hour averaging-time could either over- or under-estimate risks.
In addition, the human exposure studies evaluated airways responsiveness in mild
asthmatics. For ethical reasons, more severely affected asthmatics and asthmatic children were
not included in these studies. Severe asthmatics and/or asthmatic children may be more
susceptible than mildly asthmatic adults to the effects of NC>2 exposure. Therefore, the potential
health effect benchmarks based on these studies could underestimate risks in populations with
greater susceptibility.
143
-------
7.5 KEY OBSERVATIONS
Presented below are key observations resulting from the air quality characterization:
• NO2 concentrations and estimates of benchmark exceedances are typically higher
for monitors that are within 20 m of a major roadway than when monitors are
farther (i.e., between 20 m and 100 m or >100 m) from a major roadway.
• Estimated on-road annual average NO2 concentrations, based on simulated air
quality, are, on average, 80% higher than the respective ambient levels at
distances >100 m from a road. This falls within the range of on-road to distant
monitor concentration ratios reported in the ISA (about 2-fold higher
concentrations on-roads) (ISA, section 2.5.4).
• For unadjusted air quality, representing a recent year, many locations are
estimated to have, on average, 0 days per year where the 1-hour daily maximum
ambient NO2 concentrations are > 100 ppb. Only one location is estimated to
experience more than 10 such days, though results were from a monitor sited
within a predominantly commercial area. Most locations are estimated to have,
on average, 0 days per year with 1-hour daily maximum ambient NO2
concentrations > 200-300 ppb. No location is estimated to have more than 1 such
day per year, on average (Tables 7-14 to 7-19). The corresponding annual
average NO2 concentrations typically ranged from 10 to 30 ppb (Tables 7-11 to 7-
13). In contrast, most locations are estimated to have between 10 and 50 days per
year where 1-hour daily maximum NO2 concentrations are > 100 ppb based on
simulated on-road air quality. On average, most locations are estimated to have
fewer than 5 days per year where 1-hour daily maximum on-road NO2
concentrations are > 200 ppb (Table 7-21). The annual average of estimated on-
road NO2 concentrations typically ranged from 15 to 45 ppb (Table 7-20).
• When air quality is adjusted to simulate just meeting the current annual standard,
a hypothetical scenario requiring air quality to be adjusted upward, all locations
evaluated are estimated to have multiple days per year where 1-hour daily
maximum ambient NO2 concentrations are > 100 ppb. Most locations are
estimated to have, on average, 50 days or more per year with 1-hour daily
144
-------
maximum ambient NC>2 concentrations > 100 ppb, and six locations are estimated
to have 100 days or more per year with 1-hour daily maximum ambient NC>2
concentrations > 100 ppb. Fewer benchmark exceedances are estimated to occur
with higher benchmark levels. For example, only two locations are estimated to
have 10 or more days per year with 1-hour daily maximum ambient NC>2
concentrations that equal or exceed 200 ppb (Figures 7-2 and 7-3). Most
locations are estimated to have between 100 and 300 days per year with 1-hour
daily maximum on-road NC>2 concentrations > 100 ppb and between 25 and 100
days per year with 1-hour daily maximum on-road NC>2 concentrations > 200 ppb
(Figure 7-6). The corresponding annual average NC>2 concentrations were
typically between 30 and 50 ppb (Table 7-28).
In a number of locations, potential alternative standard levels of 0.05 and 0.10 ppm are estimated
to result in far fewer days per year than standard levels of 0.15 and 0.20 ppm with NC>2
concentrations > 100 ppb (Tables 7-26 and 7-27). When considering the potential alternative
standard levels of 0.05 and 0.10 ppm, corresponding annual average NC>2 concentrations were
typically between 10 and 30 ppb, similar to a range of concentrations using unadjusted air
quality. When considering the potential alternative standard levels of 0.15 and 0.20 ppm,
corresponding annual average NC>2 concentrations were typically between 25 and 55 ppb, similar
to the range of concentrations observed when using adjusted air quality that just meets the
current annual standard.
145
-------
8. EXPOSURE ASSESSMENT AND HEALTH RISK
CHARACTERIZATION
8.1 OVERVIEW
This section documents the methodology and data used in the inhalation exposure
assessment and associated health risk characterization for NC>2 conducted in support of the
current review of the NC>2 primary NAAQS. Two important components of the analysis include
estimating temporally and spatially variable ambient NC>2 concentrations and simulating human
contact with these pollutant concentrations. The approach was designed to better reflect
exposures that occur nearby or on a roadway, not necessarily reflected by the existing ambient
monitoring data.
Both air quality and exposure modeling approaches have been used to generate estimates
of 1-hour NC>2 exposures within Atlanta, Georgia based on a 3-year period (2001-2003).
AERMOD, an EPA recommended dispersion model, was used to estimate 1-hour ambient NC>2
concentrations using emissions estimates from stationary and on-road mobile sources. The Air
Pollutants Exposure (APEX) model, EPAs human exposure model, was then used to estimate
population exposures using the hourly census block level NC>2 concentrations estimated by
AERMOD.
Exposure and potential health risk were characterized considering recent air quality
conditions (as is), for air quality adjusted upward to just meet the current NC>2 standard (0.053
ppm annual average), and for just meeting several potential alternative standards (see chapter 5).
The estimated 1-hour exposures for each of these air quality scenarios were compared with the 1-
hour potential health effect benchmark levels identified in chapter 6. Specifically, the number of
times an individual experienced a daily maximum 1-hour exposure concentration in excess of
100 ppb through 300 ppb was recorded. The exposures for each individual were estimated over
an entire year therefore, multiple occurrences of exceedances were recorded, giving the number
of days per year with an exceedance of the potential health effect benchmark levels.
The approaches used for assessing exposures in Atlanta are described below. Detailed
input data and supporting discussion of the Atlanta case-study is provided in Appendix B-4, in
addition to containing the methodology and results for the first exposure modeling case-study
146
-------
conducted in Philadelphia County as part of the 1st draft REA (EPA, 2008b). The Philadelphia
County assessment is not included in this REA. There were a few major differences in the
approaches used that would not necessarily allow for a direct comparison of the estimated
exposures with those estimated for Atlanta, therefore the approach and the exposure results for
the Philadelphia County case-study are discussed entirely in Appendix B.
Briefly, the discussion that follows here includes:
• Description of the inhalation exposure model and associated input data used for
Atlanta
• Evaluation of estimated NC>2 air quality concentrations and exposures
• Assessment of the quality and limitations of the input data for supporting the goals of
the NC>2 NAAQS exposure and risk characterization.
The overall flow of the exposure modeling process performed for this NC>2 NAAQS
review is illustrated in Figure 8-1. Several models were used in addition to APEX and
AERMOD including emission factors, meteorological processing, and travel demand models, as
well as a number of data bases and literature sources to populate the model input parameters.
Each of these are described within this chapter, supplemented with additional details in
Appendix B.
147
-------
Upper Air
Meteorological
Data
(NOAA
Radiosonde
Database)
Surface
Meteorological
Data
(Integrated
Surface Hourfy
Database)
J
M
|| E™
MOBILE6.2
Emission Factor
Model
J_
AERMET/
AERSURFACE
Major Stationary
Source Emtsshans
(NH)
Major Roadway
Link Emissions
{AADT>15,000)
Minor Roadway
Link Emissions
(AADT<15,000)
yUport Emissions
(NEI)
LMeteorological;
and Surface Data
NO; Outdoor Total
Concentration
Estimates
Human Actvify
Patterns
(CHAD)
Census Block
Populations
(US Census)
Home-toWork
Corn muting
(US Census)
KI02 Outdoor
Total
Concentration
Estimates
Cooking
NO2 Indoor
Source
Concentration
Contributions
(GARB study)
•Air Exchange
Rates I
Penetration
Factors
(Published
Studies)
VPE
Model-to-
Monitor
Comparison
|NO2 Exposure!
Iconcentrationl
Estimates
Figure 8-1. General flow used for NOi exposure assessment.
148
-------
8.2 OVERVIEW OF HUMAN EXPOSURE MODELING USING APEX
The EPA has developed the Air Pollutants Exposure Model (APEX) model for estimating
human population exposure to criteria and air toxic pollutants. APEX serves as the human
inhalation exposure model within the Total Risk Integrated Methodology (TRIM) framework
(EPA 2006a; 2006b) and was recently used to estimate population exposures in 12 urban areas
for the O3 NAAQS review (EPA, 2007g; 2007h).
APEX is a probabilistic model designed to account for sources of variability that affect
people's exposures. APEX simulates the movement of individuals through time and space and
estimates their exposure to a given pollutant in indoor, outdoor, and in-vehicle
microenvironments. The model stochastically generates a sample of simulated individuals using
census-derived probability distributions for demographic characteristics. The population
demographics are drawn from the year 2000 Census at the tract, block-group, or block level, and
a national commuting database based on 2000 census data provides home-to-work commuting
flows. Any number of simulated individuals can be modeled, and collectively they approximate
a random sample of people residing in a particular study area.
Daily activity patterns for individuals in a study area, an input to APEX, are obtained
from detailed diaries that are compiled in the Consolidated Human Activity Database (CHAD)
(McCurdy et al., 2000; EPA, 2002). The diaries are used to construct a sequence of activity
events for simulated individuals consistent with their demographic characteristics, day type, and
season of the year, as defined by ambient temperature regimes (Graham and McCurdy, 2004).
The time-location-activity diaries input to APEX contain information regarding an individuals'
age, gender, race, employment status, occupation, day-of-week, daily maximum hourly average
temperature, the location, start time, duration, and type of each activity performed. Much of this
information is used to best match the activity diary with the generated personal profile, using
age, gender, employment status, day of week, and temperature as first-order characteristics. The
approach is designed to capture the important attributes contributing to an individuals' behavior,
and of likely importance in this assessment (i.e., time spent outdoors) (Graham and McCurdy,
2004). Furthermore, these diary selection criteria give credence to the use of the variable data
that comprise CHAD (e.g., data collected were from different seasons, different states of origin,
etc.).
149
-------
APEX has a flexible approach for modeling microenvironmental concentrations, where
the user can define the microenvironments to be modeled and their characteristics. Typical
indoor microenvironments include residences, schools, and offices. Outdoor microenvironments
include for example near roadways, at bus stops, and playgrounds. Inside cars, trucks, and mass
transit vehicles are microenvironments which are classified separately from indoors and
outdoors. APEX probabilistically calculates the concentration in the microenvironment
associated with each event in an individual's activity pattern and sums the event-specific
exposures within each hour to obtain a continuous series of hourly exposures spanning the time
period of interest. The estimated microenvironmental concentrations account for the
contribution from ambient (outdoor) pollutant concentration and influential factors such as the
penetration rate into indoor microenvironments, air exchange rates, decay/deposition rates,
proximity to important outdoor sources, and indoor source emissions. Each of these influential
factors are dependent on the microenvironment modeled, the available data to define each of the
parameters, and the estimation method selected by the user. And, because the modeled
individuals represent a random sample of the population of interest, the distribution of modeled
individual exposures can be extrapolated to the larger population within the modeling domain.
The exposure modeling simulations can be summarized by five steps, each of which is
detailed in the subsequent sections of this document. Briefly, the five steps are as follows.
1. Characterize the study area. APEX selects the census blocks within that study area
- and thus identifies the potentially exposed population - based on user-defined
criteria and availability of air quality and meteorological data for the area.
2. Generate simulated individuals. APEX stochastically generates a sample of
hypothetical individuals based on the demographic data for the study area and
estimates anthropometric and physiological parameters for the simulated individuals.
3 Construct a sequence of activity events APEX constructs an exposure event
sequence spanning the period of the simulation for each of the simulated individuals
using the time-location-activity pattern data.
4. Calculate hourly concentrations in microenvironments. APEX users define
microenvironments that people in the study area visit by assigning location codes in
the activity pattern to the user-specified microenvironments. The model then
calculates hourly pollutant concentrations in each of these microenvironments for the
150
-------
period of simulation, based on the user-provided microenvironment descriptions, the
hourly air quality data, and for some of the indoor microenvironments, indoor sources
of NO2. Microenvironmental concentrations are calculated for each of the simulated
individuals.
5. Estimate exposures. APEX estimates a concentration for each exposure event based
on the microenvironment occupied during the event. These values can be averaged
by clock hour to produce a sequence of hourly average exposures spanning the
specified exposure period. These hourly values may be further aggregated to produce
daily, monthly, and annual average exposure concentrations.
8.3 CHARACTERIZATION OF STUDY AREA
8.3.1 Study Area Selection
The selection of the location used for this exposure analysis was based on the location of
field and epidemiology studies, the availability of ambient monitoring and other input data, the
desire to represent a range of geographic areas, population demographics, general climatology,
and results of the ambient air quality characterization.
Atlanta, along with several other locations, was initially selected as a location of interest
through statistical analysis of the ambient NO2 air quality data (see section 7 and Appendix A).
Briefly, criteria were established for selecting ambient monitoring sites having high annual mean
concentrations and/or exceedances of potential health effect benchmark concentrations. The 90th
percentile served as the point of reference for the annual mean concentrations and, across all
complete site-years for 2001-2006, this value was 23.5 ppb. Seventeen locations had one or
more site-years with an annual average concentration at or above the 90th percentile, of which
Atlanta had one site-year (26.6 ppb annual average). A 1-hour potential health effect benchmark
level of 200 ppb was selected as the second criteria for location selection, and Atlanta had one
measured concentration above this level. An additional grouping of locations was done base on
geographic regions of the U.S., with Atlanta as one of the locations in the southeastern U.S.
EPA was also able to obtain measured daily NO2 exposures of several individuals residing in
Atlanta (Suh, 2008), potentially for use in evaluating the APEX estimated exposures. Therefore,
in considering each of these selection criteria, 1) the availability of health effects data associated
with ambient concentrations (Tolbert et al., 2007), 2) the availability of personal exposure
151
-------
measurements, 3) the analysis of the air quality data, and 4) geographic representation, Atlanta
was selected as the second case-study location.
8.3.2 Study Area Description
The greater Atlanta metropolitan area covers the 13 counties within a radius of
approximately 40 km about the Atlanta city center (33.65 °N 84.42 °W) in Fulton County. Due
to the complexity of the air quality and exposure modeling to be performed in this exposure
assessment, the study location (or modeling domain) was designated as the four counties directly
surrounding the city of Atlanta (i.e., Cobb, DeKalb, Fulton, and Gwinnet Counties) (see Figure
8-2). These four counties comprise the urban center of the Atlanta MSA, and contain a large
portion of the urbanized road systems in the area. This four county modeling domain contains
27,315 U.S. Census blocks with a combined population of 2,678,078 (2000 Census), comprising
approximately 65% of the Atlanta MSA population.
8.3.3 Time Period of Analysis
Calendar years 2001 through 2003 were simulated to envelop the most recent year of
emissions data available for the study location (i.e., 2002) and to include a total of 3 years of
meteorological data to achieve a degree of representativeness in the dispersion and exposure
model estimates. In considering the past 30-years of annual average temperature and
precipitation in Atlanta, the three years were variable. On a scale of high to low, in 2001 the
temperature was about average (ranked 17th) though dry (28th lowest precipitation level), 2002
had average precipitation (18th) although warmer than most years (12th), and 2003 was cooler
(25th) and wetter (11th) than many other years (NCDC, 2007)
8.3.4 Populations Analyzed
The exposure assessment included the total population residing in each modeled area and
considered susceptible and vulnerable populations as identified in the ISA. These include
population subgroups defined from either an exposure or health perspective. The population
subgroups identified by the ISA (EPA, 2008b) that were included and that can be modeled in the
exposure assessment include:
• Children (5-18 years in age)
• Asthmatic children (5-18 years in age)
152
-------
• All persons (all ages)
• All Asthmatics (all ages)
In addition to these population subgroups, individuals anticipated to be exposed more
frequently to NC>2 were assessed, including those commuting on roadways and persons residing
near major roadways.
Major-Facility NOx-Emitting Stacks (2002 NEI)
Ozone Monitors
Tier-3 NOxMonitors
Upper-Air Station
A KATLISH Surface Met Station
GA_Counties_ProjectUTM 1BN
9,103 Random Census Block Sample
18.212 Unused Census Blocks
Greene Ts/liaferro
Figure 8-2. Four county modeling domain used for Atlanta exposure assessment.
153
-------
8.4 CHARACTERIZATION OF AMBIENT AIR QUALITY USING
AERMOD
8.4.1 Overview
Air quality data used for input to APEX were generated using AERMOD, a steady-state,
Gaussian plume model (EPA, 2004). The following steps were performed to estimate air
concentrations using AERMOD.
1 Collect and analyze general input parameters. Meteorological data, processing
methodologies, and information on surface characteristics and land use were used
to determine pollutant dispersion characteristics, atmospheric stability, and
mixing heights.
2. Define sources and estimate emissions. The emission sources modeled included
major stationary emission sources, on-road emissions that occur on major and
minor roadways, and emissions from Atlanta Hartsfield International Airport.1
3. Define receptor locations. Three sets of receptors were identified for the
dispersion modeling, and included, ambient monitoring locations, census block
centroids, and links along major roadways.
4. Estimate concentrations at receptors. Hourly concentrations were estimated for
each year simulated (2001-2003) by combining the estimated concentration
contributions from each of the emission sources at each of the defined receptors.
A brief description of input data and approaches used for estimating source emissions are
described below. Additional details on the inputs and assumptions used in the dispersion
modeling are provided in Appendix B-4.
8.4.2 General Model Inputs
8.4.2.1 Meteorological Inputs
All meteorological data used for the AERMOD dispersion model simulations were
processed with the AERMET meteorological preprocessor, version 06341. Raw meteorological
1 Fugitive emissions from major point sources in the Atlanta area were not included as was done in the Philadelphia
County case study, since the NEI shows all emissions to be accounted by stack totals.
154
-------
data from the Southeast Aerosol Research and Characterization study (SEARCH) site in Atlanta
were used as the primary source of meteorology for the AERMOD runs for the years 2001
through 2003. Raw hourly surface meteorological data for the 2001 to 2003 period were
obtained from the Integrated Surface Hourly (ISH) Database,2 primarily for use in modeling the
emissions from the Atlanta Hartsfield International Airport (KATL). Upper air data in the
Forecast System Laboratory (FSL) format was downloaded from the FSL, (now Global Systems
Division) website, http://www.fsl.noaa.gov/. Details regarding the data preparation and
processing are given in Appendix B, Attachment 1.
8.4.2.2 Surface Characteristics and Land Use Analysis
In addition to the standard meteorological observations of wind, temperature, and cloud
cover, AERMET analyzes three principal variables to help determine atmospheric stability and
mixing heights: the Bowen ratio, surface albedo as a function of the solar angle, and surface
roughness. A draft version of AERSURFACE (08256) was used to estimate land-use patterns
and calculate these three variables as part of the AERMET processing, using the US Geological
Survey (USGS) National Land Cover Data 2001 archives.3 Details for the seasonal specification
definitions, land-use sectors, and data processing are given in Appendix B, Attachment 1.
8.4.2.4 Other AERMOD Input Specifications
All emission sources in the Atlanta modeling domain were characterized as urban, using
the 2000 census population of approximately 4.1 million people in the Atlanta MSA.4 The
AERMOD toxics enhancements were also employed to speed calculations from area sources.
NOX chemistry was applied to all sources to determine NC>2 concentrations. For the roadway and
airport emission sources the Ozone Limiting Method (OLM) (EPA, 2006c) was used, with
plumes considered grouped. For all point source simulations, the Plume Volume Molar Ratio
Method(PVMRM) was used to estimate the conversion of NOX to NC>2 (Hanrahan, 1999a,
1999b). The equilibrium value for the NO2:NOX ratio was taken as 75%, the national average
ambient background ratio.5 The initial NC>2 fraction of NOX is anticipated to be about 10% or
2 National Climatic Data Center (NCDC), http://wwwl.ncdc.noaa.gov/pub/data/techrpts/tr200101/tr2001-01.pdf
3 http://seamless.usgs.gov/
4 http://www.census.gov/Press-Release/www/2001/sumfilel.html
5 Appendix W to CFR 51, page 466. http://www.epa.gov/scramOO 1/guidance/guide/appw 03.pdf.
155
-------
less (Finlayson-Pitts and Pitts, 2000; Yao et al., 2005), therefore a conservative value of 10%
was selected from the upper range of this estimate and used for all sources.
Hourly surface Oj data for years 2001-2003 were obtained from five ambient monitors
operating as part of EPA's Air Quality System (AQS)6 and from one ambient monitor operating
as part of the South Eastern Aerosol Research and Characterization (SEARCH) study.7 Missing
data were substituted based on seasonal and time of day characteristics, and hourly values were
averaged across each of the Os monitors which were available for a particular hour. None of the
AQS monitors had data available for November, December, January, and February, for these
months only the SEARCH monitor data were used. The locations of these monitors are shown in
Figure 8-2.
8.4.3 Major Link On-Road Emission Estimates
Information on traffic data in the Atlanta area was obtained from the Atlanta Regional
Commission (ARC) - the regional planning and intergovernmental coordination agency for the
10-county metropolitan area - via their most recent, baseline travel demand modeling (TDM)
simulation for year 2005. Although considerable effort was expended to maintain consistency
between the ARC approach to analysis of TDM data and that employed in this analysis, complete
consistency was not possible due to the differing analysis objectives. The ARC creates county
emission inventories. This study created spatially and temporally resolved emission strengths for
dispersion modeling. Information about expected differences in traffic between the 2005 data
year and 2001-2003 modeled years was not provided by ARC, nor was information about
seasonal differences in MOBILE6.2 inputs. The approach used for estimating these major road
emissions is discussed further below.
8.4.3.1 Emission Sources and Locations
The TDM simulation's data file outputs include a description of the fixed information for
the highway network links and traffic descriptors for four time periods: morning, afternoon,
evening, and nighttime. Each period's data includes free-flow speed, total vehicle count, total
heavy duty truck count, total single occupancy vehicle count, and TDM-calculated congested
6 http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata. htm
7 Ambient data were obtained from the Jefferson Street ozone monitor, maintained by Atmospheric Research &
Analysis, Inc. Available at http://www.atmospheric-research.com/studies/SEARCH/index.html.
156
-------
speeds for the period. The description of the network consists of a series of nodes joining
individual model links (i.e., roadway segments) to which the traffic volumes are assigned, and
the characteristics of those links, such as endpoint location, number of lanes, link distance, and
TDM-defined link daily capacity.
First, all links with annual average daily traffic (AADT) values greater than 15,000
vehicles per day (one-way) were classified as major within the four counties (Cobb, DeKalb,
Fulton, and Gwinnett) and a part of a fifth county (Clayton), which contains a small portion of
the beltway in the MSA. Then, link locations from the TDM were modified through a GIS
analysis to represent the best known locations of the actual roadways, since there was not always
a direct correlation between the two (see Appendix B-4.1.1). There were no hourly scaling
factors provided for the ARC's TDM predictions, therefore the total period volume was spread
uniformly amongst all hours contributing to that period (morning: 6AM-10AM, midday: 10AM-
3PM, afternoon: 3PM-7PM, or nighttime: 7PM-6AM). A 5-hour rolling average, centered on the
present hour, was applied to the emission scaling factors to allow for a smoothing of the
distribution. The heavy-duty vehicle (FtDV) fraction for each hour of each period was obtained
by dividing the total period truck count by the total vehicle count, fixing the value as constant for
all hours of the period, but allowing it to vary between periods and across links, according to the
TDM parameterization. It should be noted that trucks, as defined in the TDM, include heavy and
medium duty vehicles as well as commercial vehicles. Because no information on seasonal
variation in vehicle activity was available, no seasonal variation was used in the simulations.
However, seasonal variations in emission factors from MOBILE6.2 were implemented - see
section 8.4.3.2. The AADT and truck fraction from the ARC TDM used in the AERMOD
simulations of major links are shown in Tables 8-1 and 8-2, respectively. Note that no rural,
local designated links meet the major link AADT criteria, and are thus omitted in Tables 8-1, 8-
2, and 8-3.
Table 8-1. Statistical summary of average annual daily traffic (AADT) volumes (one direction) for
Atlanta AERMOD simulations.
Statistic
Count
Minimum AADT
Road Type
Arterial
Freeway
Local
Arterial
Freeway
CBD1
229
109
41
15,015
15,049
Fringe
180
94
60
15,019
16,745
Rural
14
2
16,603
23,569
Suburban
1,299
616
168
15,002
15,111
Urban
1,221
616
250
15,017
15,025
157
-------
Statistic
Maximum AADT
Average AADT
Road Type
Local
Arterial
Freeway
Local
Arterial
Freeway
Local
CBD1
15,442
51,820
150,047
110,425
24,814
73,598
25,737
Fringe
15,052
49,853
109,204
98,420
21,732
56,741
26,536
Rural
23,433
24,028
19,016
23,799
Suburban
15,111
64,487
144,434
98,909
21,383
59,164
23,781
Urban
15,017
46,824
155,083
127,085
22,434
64,744
25,745
Notes:
1 Central business district
Table 8-2. Average heavy duty vehicle (HDV) fraction for Atlanta AERMOD simulations.
Functional Class
Arterial
Freeway
Local
Time Period1
Nighttime
Morning
Midday
Afternoon
Nighttime
Morning
Midday
Afternoon
Nighttime
Morning
Midday
Afternoon
Region Ty
CBD2
12%
14%
17%
10%
8%
9%
12%
7%
10%
12%
14%
9%
Fringe
18%
19%
28%
16%
20%
20%
27%
16%
24%
24%
33%
19%
Rural
15%
18%
27%
15%
24%
26%
33%
21%
pe
Suburban
12%
15%
20%
12%
19%
19%
26%
15%
18%
19%
25%
15%
Urban
13%
15%
20%
12%
14%
15%
21%
12%
15%
16%
21%
12%
Notes:
1 morning: 6AM-10AM, midday: 10AM-3PM, afternoon: 3PM-7PM, or nighttime: 7PM-6AM
2 Central business district
8.4.3.2 Emission Source Strength
On-road mobile emission factors were derived from the MOBILE6.2 emissions model
using ARC input files describing the 2002 vehicle registration distribution and corresponding to
the 2008 Os season. To maintain consistency with the recent ARC simulations and current
modeling parameters and maximize temporal resolution, the ARC's 63 season input files were
used as a basis for all MOBILE6.2 simulations, but were modified as follows. First, the 24-hour
158
-------
series of temperature and humidity values included in the ARC files were those derived as
average values over peak 63 days. To modify the focus from peak 63 to average summer days,
these values in the input files were modified by converting to average daily minimum and
maximum temperature and corresponding specific humidity, determined by the same
meteorological record used in the dispersion simulations. Also, winter and summer-specific
fuels for the Atlanta region were used for all years, which differ only until the phase-in of
Georgia Phase 2 gasoline in 2003, at which point winter and summer sulfur levels are identical.
Finally, anti-tampering and inspection/maintenance programs, which were not included in the
original ARC input files, were taken from MOBILE input files prepared by the State of Georgia
for a previous project.
The simulations were executed to calculate average running NOX emission factors in
grams per mile for a specific functional class (Freeway, Arterial, Local, or Ramp), speed, and
season. Iterative MOBILE6.2 simulations were conducted to create tables of average Atlanta
region emission factors resolved by speed (2.5 to 65 mph, in 1 mph increments from 3 to 65
mph), functional class, season, and year (2001, 2002, or 2003) for each of eight combined
MOBILE vehicle classes. The resulting tables were then consolidated into speed, functional
class, and seasonal values for combined light- and heavy-duty vehicles. To create seasonal-
hourly resolved emissions, spring and fall values were taken as the average of corresponding
summer and winter values. See Appendix B-4 for an example of the calculated emission factors
for Summer, 2001.
To determine the emission strengths for each link for each hour of the year, the Atlanta
regional average MOBILE6.2 speed-resolved emissions factor tables were merged with the TDM
link data, which had been processed to determine time-resolved speeds. The spatial-mean speed
of each link at each time was calculated following the methodology of the Highway Capacity
Manual.8 Table 8-3 shows the resulting average speed for each functional class within each
TDM region. The resulting emission factors were then coupled with the TDM-based activity
estimates to calculate emissions from each of the major roadway links.
Table 8-3. Average calculated speed by link type in Atlanta modeling domain.
| Average Speed (mph)
8 As defined in Chapter 9 of Recommended Procedure for Long-Range Transporation Planning and Sketch
Planning. NCHRP Report 387, National Academy Press, 1997. 151 pp., ISBN No: 0-309-060-58-3.
159
-------
Link Type
Arterial
Freeway
Local
CBD1
22
54
26
Fringe
37
62
40
Suburban
40
60
40
Urban
30
57
34
Rural
51
64
N/A
Notes:
1 Central Business District
8.4.3.3 Other Emission Parameters
Each roadway link is characterized as a rectangular area source with the width given by
the number of lanes and an assumed universal lane width of 12 ft (3.66 m). The length and
orientation of each link is determined as the distance and angle between end nodes from the
adjusted TDM locations. In cases where the distance is such that the aspect ratio is greater than
100:1, the links were disaggregated into sequential links, each with a ratio less than that
threshold. There were 737 links that exceeded this ratio and were converted to 1,776 segmented
sources. Thus, the total number of area sources included in the dispersion simulations is 5,570.
Note that there are some road segments whose length was zero after GIS adjustment of node
location. This is assumed to be compensated by adjacent links whose length will have been
expanded by a corresponding amount. Table 8-4 shows the distribution of on-road area source
sizes.
Table 8-4. On-road area source sizes.
Statistic
Minimum
Median
Mean
1-a Deviation
Maximum
Number of
Lanes
1
2
2.7
1.2
8
Segment
Width (m)
3.7
7.3
9.9
4.5
29.3
Segment
Length (m)
0.0
352.8
426.3
330.0
2218.1
Resulting daily emission estimates were temporally allocated to hour of the day and
season using MOBILE6.2 emission factors, coupled with calculated hourly speeds from the post-
processed TDM and allocated into SEASHR emission profiles for the AERMOD dispersion
model. That is, 96 emissions factors are attributed to each roadway link to describe the emission
strengths for 24 hours of each day of each of four seasons and written to the AERMOD input
control file.
For light duty vehicles (LDV) it was assumed that the initial vertical extent of the plume
is about 1.7 times the average vehicle height, or about 2.6 meters for an average vehicle height of
160
-------
about 1.53 meters (5 feet), to account for the effects of vehicle-induced turbulence among other
factors. The source release height is based on the midpoint of the initial vertical extent, or about
1.3 meters. The initial vertical dispersion coefficient (sigma-Z0) was based on the initial vertical
extent divided by 2.15, or 1.2 meters.
For the heavy duty vehicles (HDV) as with LDVs, the initial vertical extent of the plume
was assumed 1.7 times the average vehicle height, or about 6.8 meters for an average vehicle
height of about 4.0 meters. Similarly, source release heights were based on the midpoint of the
initial vertical extent, or 3.4 meters. The initial sigma-Zo also based on the initial vertical extent
divided by 2.15, was 3.2 meters.
For effective source parameters representing a mix of LDV and HDV for a particular
major roadway link, the source release height and initial sigma-Zo were then assigned using an
emissions-weighted average based on the vehicle mix for that roadway link.
The total NOX emissions on the major roadways links were estimated to be 88,438 tons
per year (tpy) or approximately 70% of the total on-road emissions.
8.4.4 Minor Link On-road Emission Estimates
On-road mobile emissions that do not occur on major roadway links were assigned to US
Census tracts and simulated as area sources represented by the tract polygons. There are 478
census tract area sources across the 4-county Atlanta modeling domain, and a small part of
Clayton County (Figure 8-3). Emission magnitudes and temporal profiles were derived with the
same procedure as for the major roadway links, however individual link values were not stored.
Instead, each link was assigned to its respective tract and the combined emission total across
161
-------
BartOW Cartersville
Ozone Monitors
Tier-JNOx Monitors
Cities, Pop. >15,000 (2000 Census)
• 15.000-30,000
• 30.001-50,000
• 50.001-75,000
41 Paulding
ioonno
if Upper-Air Station
A KAIL ISH Surface Met Station
Fayette
Peaihtree City
Putnam
Hancock
Figure 8-3. The 478 U.S. Census tracts representing area sources for on-road mobile
emissions that do not occur on major roadway links.
links for a specific season and hour were determined for each tract. The resulting total seasonal-
hourly emissions profile for each tract area source was then used in AERMOD. Tract-wide
emission release parameters for minor links in the dispersion modeling were determined as
emissions-weighted averages of light- and heavy-duty vehicle contributions to the tract total
values. Estimated NOX emissions on the minor roadway links within the five counties was 38,039
tpy (Table 8-5).
Table 8-5. On-road emissions from major and minor links in Atlanta, 2002.
County
FIPS1
13063
13067
13089
13121
13135
Name
Clayton
Cobb
DeKalb
Fulton
Gwinnett
Total On-Ro
(t
Minor Link2
1,693
8,329
7,134
12,047
8,835
ad Emissions
PV)
Major Link3
6,185
15,816
19,871
30,999
15,568
% Minor4
21%
34%
26%
28%
36%
Total 38,039 88,438 30%
Notes:
1 Federal Information Processing Codes for each county.
2 Minor links are those roads with < 15,000 AADT.
3 Major links are those roads > 15,000 AADT.
162
-------
4 % Minor is the percent of minor roads in each county.
8.4.5 Adjustment of On-road Mobile Source Strengths to 2002 NEI Vehicle
Emissions
As noted above, the TDM data received from ARC specified traffic count projections for
2005 instead of the 2001-2003 target years for this analysis. All other model inputs were
estimated for the target years, e.g., on-road mobile source emission factors, point source
emissions (see section 8.4.6 below), airport emissions (see section 8.4.7 below), and
meteorological data. Therefore, to maintain temporal consistency for all emissions used, the on-
road emission strengths using the 2005 TDM data were adjusted to match 2002 totals for the 4-
county modeling domain from the National Emissions Inventory (NEI).
Table 8-6 compares the on-road mobile source emissions estimated for 2002 as described
above (i.e., the 2005 traffic counts combined with 2002 emission factors) with the NEI estimates
for 2002. Note that the differences in these estimates may be the result of differences in other
factors in addition to the target year traffic counts, such as fleet mix and heavy-duty vehicle
fractions. Based on this comparison, an adjustment factor of 0.78 was uniformly applied to all
on-road mobile source emission strengths in the Atlanta modeling domain, for both major and
minor links.
Table 8-6. On-road vehicle emission strengths by county for Atlanta modeling domain: modeled vs
NEI 2002.
County
Cobb
DeKalb
Fulton
Gwinett
TOTAL
Modeled
major link NOX
emissions
(tpy)
15,816
19,871
30,999
15,568
82,254
Modeled
minor link
NOX
emissions
(tpy)
8,329
7,134
12,047
8,835
36,346
Total modeled
on-road
vehicle NOX
emissions
(tpy)
24,145
27,006
43,046
24,403
118,599
NEI on-road
vehicle NOX
emissions for
2002 (tpy)
18,754
21,715
33,886
18,080
92,434
Ratio of NEI-
2002-to-
modeled NOX
emissions
0.78
0.80
0.79
0.74
0.78
163
-------
8.4.5 Stationary Sources Emissions Preparation
Data for the parameterization of major point sources in Atlanta comes primarily from
three sources: the 2002 National Emissions Inventory (NEI; US EPA, 2007e), Clean Air Markets
Division (CAMD) Unit Level Emissions Database (US EPA, 2007f), and temporal emission
profile information contained in the EMS-HAP (version 3.0) emissions model.9 The NEI
database contains stack locations, emissions release parameters (i.e., height, diameter, exit
temperature, exit velocity), and annual emissions for NOx-emitting facilities. The CAMD
database contains information on hourly NOX emission rates for units in the US, where the units
are the boilers or equivalent, each of which can have multiple stacks.
First, major stationary sources were selected from the NEI where stacks within facilities
contain at least 100 tpy total NOX emissions and are located either within the 4-county modeling
domain or within 10 km of the modeling domain. Seven NOx-emitting facilities met these
criteria (Figure 8-4). Stacks within the facilities that were listed separately in the NEI were
combined for modeling purposes if they had identical stack physical parameters and were co-
located within about 10m. This resulted in 28 combined stacks (stack parameters are in
Appendix B-4) and accounts for 16% of the total number of NOX point sources and 51% of the
total NOX point source emissions in this buffered four county Atlanta area.
The CAMD database was then queried for facilities that matched the facilities identified
from the NEI database using the facility name, the Office of Regulatory Information Systems
(ORIS) identification code, and facility total emissions. Only one of the 7 major facilities
identified was found in the CAMD data base: the Georgia Power Company McDonough Steam-
Generating Plant. The CAMD hourly emissions profiles for two units in this facility were
summed together and then, after appropriate scaling, used to represent 2 major-facility combined
stacks.
For the remaining 26 major-facility combined stacks, hourly NOX emissions profiles were
created based on the hourly profile typical of that stack's Source Classification Codes (SCC), the
season, and the day of week. These SCC-based temporal profiles are year-independent, and
were developed for the EPA's EMS-HAP model,10 described in the EMS-HAP model Version 2
9 http://www.epa.gov/ttn/chief/emch/projection/emshap30.html
1 ° http://www.epa.gov/scramOO l/dispersion_related.htm#ems-hap
164
-------
User's Guide, Section D-7.u As with CAMD hourly emissions, these SCC-based emission
profiles are scaled such that the annual total emissions are equal to those of NEI2002.
8.4.6 Airport Emissions Preparation
The Atlanta-Hartsfield International Airport emissions were assigned to a polygon that
defined an area source for simulation. The perimeter dimensions of the Atlanta-Hartsfield
International Airport were determined by GIS analysis of aerial photograph. As with some point
source emissions, the annual NOX emission totals were extracted from the NEI and the temporal
profiles from the EPA's EMS-HAP model. These seasonal, SCC-based emissions were scaled
such that the annual total emissions are equal to those of NEI 2002: 5,761 tpy, with about 90%
coming from commercial aircraft (see Figure 8-2 for airport location, Appendix B-4 for depiction
of area source polygon).
The initial vertical extent of the plume for aircraft emissions was estimated as 10 m to
account for typical emission heights and initial dilution parameters. A source release height of 5
m was selected based on the midpoint of the initial vertical extent and the initial vertical
dispersion coefficient was estimated using the initial vertical extent divided by 2.15, or 4.6
meters. For cargo-handling equipment a release height of 3.15 m was assumed, which is the
average for cargo-handling equipment from a study by the California Air Resources Board
(CARB 2006). The initial vertical dispersion coefficient was estimated as the release height
divided by 2.15, or 1.47 m. For effective source parameters representing a mix of aircraft and
cargo-handling equipment, the source release height and initial sigma-Zo were estimated using
an emissions-weighted average with 92% of emissions contributed by aircraft. The aggregate
value for release height was 4.85 m with an intial sigma- Zo of 4.22 m.
http://www.epa.gov/scramOOl/userg/other/emshapv2ug.pdf
165
-------
Bartow
iersville
Cherokee
[
—Dawsorr
Forsyth
*
A
Hall
Gaine
NKennesaw
Paulding
+ NOx Monitors
%£ Major NOx-Emitting Facilities (2002 NEI)
-^f Upper-Air Station
A ISH Surface Met Stations
Cities, Pop. >15,000 (2000 Census)
» 15,000-30,000
• 30,001-50,000
• 50,001-75,000
• 75,001-100.000
f >100,000
Preliminary Road Links
- Null
^^^ Rural Interstate
Rural Principal Arterial
— Rural Minor Arterial
^^^ Rural Major Collector
- Rural Minor Collector
Rural Local
^^^ Urbanized Interstate
— Urban Freeway
^^^ Urbanized Principal Arterial
- Urbanized Minor Arterial
Urbanized Collector
- Urbanized Local
10
20
JC
Coweta
Newnan
40 Kilometers
.On
Figure 8-4. Location of major roadway links and major stationary emission sources in Atlanta modeling domain.
166
-------
8.4.7 Receptor Locations
Three sets of receptors were chosen to represent the locations of interest. The first set
was selected to represent the locations of the residential population of the modeling domain.
These receptors were the 27,315 US Census block centroids12. In an effort to make the Atlanta
case-study more time efficient, a statistical analysis was performed on the Philadelphia exposure
assessment results reported in the 1st draft REA (see Appendix B, section 3) to determine the
degree of uncertainty introduced by modeling a subset of the block receptors only. The findings
of that analysis indicated that the use of a random selection of 1/3 of the block centroids would
provide estimates of exposure to exceedances of 200 ppb that were within 4% (90% confidence
bounds = 0.3% - 10%) of the estimates obtained based on using all the block receptors. It was
judged that this uncertainty was minimal when compared to other uncertainties in the analysis,
and therefore, a random selection of 9,103 (1/3) of the block centroids was used for this analysis.
These 9,103 Census block receptors are shown along with the other 18,212 block centroids are
shown in Figure 8-5. For modeling efficiency, each receptor was assigned a height of 0.0 ft (0.0
m). The effect of this on the exposure estimates in comparison with a standard breathing height
of 1.8 m is negligible (see section 7.4.4). Concentrations estimated at these centroid receptors
were used by APEX, along with other factors to estimate an individual's microenvironmental
concentrations (see section 8.7).
The second set of receptors was chosen to represent the on-road microenvironment
(Figure 8-5). For this set, one receptor was placed at the center of each of the 5,570 sources.
Receptor concentration estimates were used by APEX, along with other factors, to estimate an
individual's on-road and in-vehicle microenvironmental exposures (see section 8.7). The
distribution of distances of the on-road and the block centroid receptors was estimated to
determine the distance relationship between the on-road emissions and population-based
receptors. Approximately 1% of the block centroids are within 50 m of the center of a major
roadway link and 26% within 400 m, with a geometric mean of the distribution between 750 m
and 800 m (a detailed distribution is provided in Appendix B-4).13 The population distribution
The block centroids used for this analysis are actually population-weighted locations reported in the ESRI data
base. They were derived from geocoded addresses within the block taken from the Acxiom Corporation InfoBase
household database (Skuta and Wombold 2008; ESRI 2008). These centroids differ from the "internal points"
reported by the US Census, which are often referred to as centroids because they are designed to represent the
approximate geographic center of the block.
13 The distance to the roadway edge is shorter by half the roadway width. As noted in Table 8-4 modeled roadway
widths ranged from 3.7 m to 29.3 m with an average of 9.9 m and a median of 7.3 m.
167
-------
within close proximity of the road is actually a different value than the block centroids. For the
population simulated in Atlanta, 17% of the population resides within 75 m, 25% were between
75 and 200 m, and 58% were > 200 m of a major road.
The third set of receptors included the locations of the available ambient NOX/NO2
monitors. These receptors were used in evaluating the dispersion model performance. When
considering the four Atlanta counties and period of analysis, there were three monitors within the
modeling domain with valid ambient NC>2 monitoring concentrations (Figure 8-5).
»***
-
Selected NOx Monitors
• 9,103 Random Census Block Sample
® Major-Facility NOx-Emitting Stacks (2002 NEI)
• Road Link Centroids
{c 10 20 40 Kilometers
K • ' ' —Sock
Figure 8-5. Location of modeled receptors in Atlanta modeling domain.
8.4.8 Modeled Air Quality Evaluation
8.4.8.1 Comparison of Hourly Cumulative Density Functions
The hourly NC>2 concentrations estimated from each of the four source categories were
combined at each receptor. These concentration predictions were then compared with measured
concentrations at ambient NC>2 monitors. Rather than compare concentrations just at the single
modeled receptor point to the ambient monitor concentrations, a distribution of concentrations
was developed for the predicted concentrations for all receptors within a 4 km distance of the
168
-------
monitors, not including receptors within 100 m of a major road. Further, instead of a comparison
of central tendency values alone for the number of receptors modeled (mean or median), the
complete modeled and measurement concentration distributions were used for comparison.
As an initial comparison of modeled versus measured air quality, all modeled receptors
within 4 km of each ambient monitor location, excluding those receptors on roadways or within
100 m of a major roadway, were used to generate a prediction envelope.14 This envelope was
constructed based on selected percentiles from the modeled concentration distribution at each
receptor for comparison to the ambient monitor concentration distribution. The 2.5th and 97.5th
percentiles from all monitor distribution percentiles15 were selected to create the lower and upper
bounds of the envelope, while the 50th percentile concentrations were combined to create a
distribution representing the central tendency (Figure 8-6). The distribution of the modeled
values estimated for the monitor receptor is also presented, along with the complete hourly
concentration distribution measured at each ambient monitor. A table providing the values used
to generate the figure is provided in Appendix B-4.
The hourly concentration distributions modeled at receptors within 4 km of each of the
ambient monitor locations provide a reasonable representation of the measured ambient NC>2
concentrations. The lower and upper bounds of the predicted concentration distributions
surround the measured ambient concentration distribution at many of the percentiles. The actual
modeled monitor receptor concentration distributions were generally above that of the
corresponding measured concentrations, resulting in overestimation at some of the upper
percentiles by about 20-50%. At monitor 131210048 however, the overestimate in
concentrations was generally less than 10 ppb, or within 10-20% of that measured. In fact the
maximum estimated concentration at this monitor was 137 ppb, just 1 ppb above that measured
(136 ppb).
When considering the lowest potential health effect benchmark levels, the modeled
monitor receptor had 19, 2, and 9 daily maximum predicted values above 100 ppb 1-hour,
compared with 0, 0, and 3 of the measured values at monitors 130890002, 130893001, and
14 500 m to 4 km is the area of representation of a neighborhood-scale monitor, according to EPA guidance.
15 As an example, suppose there are 1000 receptors surrounding a monitor, each receptor containing 8,760 hourly
values used to create a concentration distribution. Then say the 73rd percentile concentration prediction is to be
estimated for each receptor. The lower bound of the 73rd percentile of the modeled receptors would represented by
the 2.5th percentile of all the calculated 73rd percentile concentration predictions, i.e., the 25th highest 73rd percentile
concentration prediction across the 1000 73rd percentile values generated from all of the receptors. Note that, at any
given percentile along either of the envelope bounds as well as at the central tendency distribution (the receptor 50th
percentile), the concentration from a different receptor may be used.
169
-------
131210048, respectively. There were only two predicted daily maximum exceedances of 150
ppb 1-hour at one monitor (ID 130890002), although there were no measured exceedances at this
monitor. None of the modeled monitors had estimated daily maximum NC>2 concentrations
above 200 ppb 1-hour as observed with the measurement data for all three monitors. This
indicates that while overestimating the number of exceedances of the lowest potential health
effect benchmark level of 100 ppb at these three locations, there is not an overestimate of higher
benchmark levels, such as those > 200 ppb considering as is air quality.
8.4.8.2 Comparison of annual average diurnal concentration profiles
A second comparison between the modeled and monitored data was performed to
evaluate the diurnal variation in NC>2 concentrations. For AERMOD, receptor concentrations
during each hour-of-the-day were averaged (i.e., 365 values for hour 1, 365 values for hour 2,
and so on), to generate an annual average NC>2 concentration for each hour at each receptor.
Then a prediction envelope was constructed similar to that described above from modeled
receptors located within 4 km of each ambient monitor. These modeled diurnal distributions,
along with that of each ambient monitor hour-of-the-day annual average concentration are
illustrated in Figure 8-7. A table providing the values used to generate the figure is provided in
Appendix B-4.
When comparing the modeled predicted and ambient measured diurnal profiles, there was
agreement between the patterns and several hours of the day where the observed values fell
within the model prediction envelope. This occurred primarily during the late night (9PM-
12AM), early morning (1AM-3AM), and late morning through many of the midday hours (8AM-
4PM). However, NC>2 concentrations were overestimated at certain times of the day, generally
between the hours of 4-6AM and 5-8PM. The overestimation in concentrations is not entirely
unexpected given the results of the distribution of hourly concentrations illustrated in Figure 8-6.
170
-------
Monitor 130890002
100 -F
^—AMBIENT MONITOR
- - - AERMOD P2.5
AERMOD P97.5
AERMOD MONITOR
'J- -J—
Monitor 130893001
AMBIENT MONITOR
AERMOD P2.5
AERMOD P97.5
AERMOD MONITOR
Monitor 131210048
AMBIENT MONITOR
AERMOD P2.5
AERMOD P97.5
AERMOD MONITOR
40
60
160
180
200
80 100 120 140
NO2 Concentration (ppb)
Figure 8-6. Comparison of measured ambient monitor NOi concentration distribution
with the modeled monitor receptor and receptors within 4 km of the monitors at three
locations in Atlanta for Year 2002.
171
-------
Monitor 130890002
.Q
Q.
Q.
.Q
Q.
a.
c
o
I
0)
o
o
-------
8.4.8.3 Comparison of estimated on-road NO2 concentrations
The two independent approaches used to estimate on-road NO2 concentrations, one using
ambient monitor data combined with an on-road simulation factor (section 7) and the other using
the AERMOD dispersion model, were compared to one another. There are no on-road NO2
concentration measurements in Atlanta for the modeled data to be compared with, although it
should be noted that the data used to estimate the simulation factors and applied to the monitor
data were measurement based.
First a comparison can be made between the adjustment factors used for estimating on-
road concentrations in the air quality analysis and similar factors that can be generated using
AERMOD estimated concentrations for year 2002. As described above in section 7, an
empirical distribution of on-road simulation factors was derived from on-road and near-road NC>2
concentration measurements published in the extant literature. The derived empirical
distribution was separated into two components, one for application to summertime ambient
concentrations, and the second for all other seasons. The two empirical distributions are
presented in Figure 8-8, and represent the factors that are multiplied by the ambient monitor
concentration (i.e., at monitors > 100 m from a major road) and used to estimate the on-road
concentration in the air quality characterization. The measurement data from which these were
derived were mainly time-averaged over 7 to 14 days. The one-hour NC>2 concentrations
estimated at AERMOD receptors > 100 m from a major road were compared with the
concentrations estimated at their closest on-road receptor to generate a similar ratio (i.e., on-
road/non-road NO2 concentrations). In this case, 7-day averages were calculated using the
hourly AERMOD concentration predictions. These 7-day average AERMOD generated ratios
were also stratified into two seasonal categories, one containing the summer ratios (June, July,
and August) and the other for all other times of the year. The AERMOD on-road factor
distributions in semi-empirical form are also presented in Figure 8-8. The values for each of the
methods and season distributions are provided in Appendix B-4.
Both the modeled and measurement derived distributions have similar seasonal
relationships, that is the summer ratios are consistently greater than the non-summer ratios
throughout the entire distribution. There are small differences when comparing the two
approaches at the lowest distribution percentiles, with the AERMOD ratios consistently below
173
-------
that of the empirically derived factors. This is likely due to the differences in the population of
samples used to generate each type of distribution. The measurement study-derived distribution
used data from on-road concentration measurements and from monitoring sites located at a
distance from the road, sites that by design of the algorithm and the factor selection criteria are
likely not under the influence of non-road NC>2 emission sources. The measurement study
derived ratios never fall below one, there are no on-road concentrations less than any
corresponding non-road influenced concentrations. This is, by design, a function of the
algorithm used to derive the ratio (i.e., concentrations away from the road are always less). This
was reasonable and conservative assumption used in estimating the on-road concentrations for
the air quality characterization performed in section 7. The population of AERMOD receptors
used to generate the depicted distribution however includes concentrations estimated at non-road
receptors that are greater than corresponding on-road receptors. This is likely a more realistic
depiction of the actual relationship between on-road and non-road receptors. It is possible that
an NC>2 emission source at a distance from a road could contribute to local concentrations more
than a mobile source contributes to corresponding on-road concentrations.
There are some similarities that follow when comparing each of the AERMOD with the
measurement study derived distributions the lower to mid percentiles. Overlap of the two
different approaches occurs through about the 40th percentile and tracks closely through the 90th
percentile. The AERMOD predicted ratio distributions then extend beyond the range of values
offered by the measurement study derived ratios at about the 90th percentile. Given the greater
number of receptors modeled by AERMOD, the AERMOD approach may be better representing
the variability in NO2 concentrations than when using the on-road adjustment factor approach.
While not directly comparable to Atlanta, a few U.S studies report similar concentrations
for inside vehicles and near roads. For example, Riediker et al. (2003) measured NO2
concentrations inside North Carolina State patrol cars while on duty in Raleigh as well as at a
fixed site monitor located near local major roadways. Mean concentrations inside the vehicles
averaged over about 9 hours were 41.7 ppb (minimum 1.6, maximum 548.5 ppb), similar to the
mean and range of concentrations estimated by AERMOD and using the on-road adjustment
factor (Table 8-7). Reported roadside NO2 concentrations were also comparable in mean
concentration, (49.9 ppb), although they had a smaller range (minimum 13.0, maximum 212.1
ppb). This is likely a result of a much greater averaging time (i.e., approximately 4-days) used
174
-------
for these roadside measurements. CARB (2003) measured NO2 concentrations inside school
buses during several commutes on urban and suburban/rural bus routes in Los Angeles, CA.
Measurements were collected over an average commute time of approximately 85 minutes and
corresponded to average concentrations of about NO2 70 ppb along an urban route (minimum 34,
maximum 120 ppb). Rural/suburban areas averaged less, NO2 concentrations were about 45 ppb
(minimum 23, maximum 68 ppb).
0.1
Not Summer - AERMOD
Summer-AERMOD
+ Summer - Studies
Not Summer - Studies
1 10 100
On-Road/Non-Road NO2 Concentration Ratio
1000
Figure 8-8. Comparison of on-road/non-road ratios developed from AERMOD
concentration estimates for year 2002 and those derived from data reported in published
measurement studies.
A second comparison was conducted using the hourly on-road NC>2 concentrations
estimated by AERMOD for 3,259 on-road receptors in Atlanta for the years 2001-2003. The 24
hourly values modeled for each day at each receptor were rounded to the nearest 1 ppb. The
second set of estimated on-road NO2 concentrations was generated as part of the Air Quality
Characterization by applying randomly selected on-road adjustment factors to the ambient
monitor concentrations in the Atlanta MSA, using the same three ambient monitors which were
all located > 100 m from a major road. Table 8-7 compares the summary statistics of the hourly
175
-------
concentrations and the number of estimated exceedances of three potential health effect
benchmark levels (i.e., 100, 150, and 200 ppb) using the two different approaches to estimate on-
road concentrations. The AERMOD predicted and ambient monitor simulated concentration
distributions have very similar variances, although the AERMOD estimated concentrations are
about 40% greater at the mean and about 15 ppb higher at each of the percentiles (save the max).
The AERMOD on-road receptors also consistently had a greater number of exceedances of
potential health effect benchmark levels than that estimated using the on-road monitor
simulation. For example, the AERMOD receptors had an average of 241 exceedances of 100
ppb per site-year while the simulated on-road monitors had an average of 169 exceedances per
year, a difference of about 40%. This difference between the two approaches was prevalent
throughout each of the percentiles and when considering each of the 1-hour concentration levels.
The differences could be due to the greater number of receptors modeled by AERMOD
(n=3,259) compared with the on-road monitor simulation (n=800) and that the AERMOD
generated on-road receptors could include locations with greater influence by roadway emissions
that are not captured by the simplified approach conducted in the Air Quality Characterization.
Table 8-7. Summary statistics of estimated on-road hourly NO2 concentrations (ppb) and the
numbers of hourly concentrations above 100,150, and 200 ppb in a year using both the
AERMOD and the on-road ambient monitor simulation approaches in Atlanta.
Statistic
Mean
Std
Var
N1
pO
P5
p10
p15
p20
p25
p30
p35
p40
p45
p50
p55
p60
On-Road Hourly
N02 (ppb)
AERMOD
43
25
631
28,548,840
0
9
15
19
23
26
29
32
34
37
40
44
46
AQ
Monitors
31
25
646
6,622,300
0
3
6
9
11
13
15
17
20
22
25
28
31
Number of hours
>100 ppb
AERMOD
241
307
94,102
3,259
0
2
4
7
11
16
23
30
41
56
79
119
185
AQ
Monitors
169
227
51,427
800
0
2
8
11
16
23
33
42
55
67
87
111
132
Number of hours
>150 ppb
AERMOD
28
51
2,577
3,259
0
0
0
0
0
0
0
0
0
1
1
3
6
AQ
Monitors
20
43
1,856
800
0
0
0
0
0
0
0
1
1
3
3
5
8
Number of hours
>200 ppb
AERMOD
5
14
190
3,259
0
0
0
0
0
0
0
0
0
0
0
0
0
AQ
Monitors
3
7
54
800
0
0
0
0
0
0
0
0
0
0
0
0
1
176
-------
Statistic
p65
p70
p75
p80
p85
p90
p95
p100
On-Road Hourly
N02 (ppb)
AERMOD
49
52
56
60
66
74
88
556
AQ
Monitors
35
38
43
49
56
65
81
437
Number of hours
>100 ppb
AERMOD
253
317
399
485
584
706
879
1,929
AQ
Monitors
160
184
220
280
353
426
649
1,595
Number of hours
>150 ppb
AERMOD
12
22
37
50
71
96
132
542
AQ
Monitors
9
13
21
26
36
53
110
373
Number of hours
>200 ppb
AERMOD
1
2
4
7
11
18
31
181
AQ
Monitors
1
1
2
3
3
9
17
55
Notes:
1 For the on-road hourly NO2 concentration, N is the number of 1 -hour concentrations generated for each
simulation. In the exceedance columns, N represents the number of AERMOD receptors or monitor site-
years simulated.
8.4.8. Using unadjusted AERMOD predicted NO2 concentrations
The NC>2 concentrations estimated using AERMOD may be biased upwards based on a
comparison with measurement data from the three available ambient monitors. Given an
apparent systematic bias, one could argue for adjusting concentrations to improve the
comparison of the model predictions with the measurement data. However, data were not
adjusted based on these model-to-monitor comparisons for a few reasons, primarily regarding the
confidence in the dispersion modeling system, the spatial representation of the monitors
compared with receptors modeled, and the number of comparisons available. Details on the
reasoning are provided in section 8.12.1.
8.5 SIMULATED POPULATION
One of the important population subgroups for the exposure assessment is asthmatics.
Evaluating exposures for this population requires an estimation of both adult and children asthma
prevalence rates. The proportion of the population of children characterized as being asthmatic
was estimated by statistics on asthma prevalence rates recently used in the NAAQS review for
Os (US EPA, 2007g). See Appendix B, Attachment 4 for details in the derivation. Specifically,
the analysis generated age and gender specific asthma prevalence rates for children ages 0-17
using data provided in the National Health Interview Survey (NHIS) for 2003 (CDC, 2007).
Adult asthma prevalence rates for Atlanta were derived from the Behavioral Risk Factor
177
-------
Surveillance System (BRFSS) survey information for years 2004 - 2005 (Blackwell and Kanny,
2007). Table 8-8 provides a summary of the prevalence rates used in the exposure analysis by
age and gender. Additional information on the variability in these prevalence rates is given in
Appendix B-4.
Table 8-8. Asthma prevalence rates by age and gender used for Atlanta.
Region
(Study Area)
Atlanta
(South)
Age1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
>17
Asthma Prevalence2
Female
0.034
0.052
0.071
0.088
0.099
0.119
0.122
0.112
0.093
0.091
0.108
0.132
0.123
0.097
0.095
0.100
0.115
0.145
0.083
Male
0.041
0.070
0.102
0.129
0.144
0.165
0.164
0.133
0.138
0.168
0.178
0.162
0.145
0.143
0.153
0.151
0.140
0.122
0.050
Notes:
1 Ages 0-17 from the National Health Interview
Survey (NHIS) for 2003 (CDC, 2007), ages >17
from the Behavioral Risk Factor Surveillance
System (BRFSS) survey information (Blackwell
and Kanny, 2007)
2 Asthma prevalence is given as fraction of the
population. Multiply by 100 to obtain the percent.
The total population simulated within the Atlanta model domain was approximately 2.68
million persons, of which there was a total simulated population of about 212,000 asthmatics.
The model simulated approximately 500,000 children, of which there were about 64,000
asthmatics. For comparison, the Georgia Department of Human Resources reports the 2001
asthma prevalence for children in middle and high school as ranging from 9.6 to 13.8%, though
for Fulton County middle school estimates were higher (15.8%) (Blackwell et al, 2003).
178
-------
And when considering this population simulated in Atlanta and their proximity to roadways,
17% of the population resides within 75 m, 25% were between 75 and 200 m, and 58% were >
200 m of a major road.
8.6 CONSTRUCTION OF LONGITUDINAL ACTIVITY SEQUENCES
Exposure models use human activity pattern data to predict and estimate exposure to
pollutants. Different human activities, such as spending time outdoors, indoors, or driving, will
result in varying pollutant exposure concentrations. To accurately model individuals and their
exposure to pollutants, it is critical to understand their daily activities. EPA's Consolidated
Human Activity Database (CHAD) provides data for where people spend time and the activities
performed. Typical time-activity pattern data available for inhalation exposure modeling consist
of a sequence of location/activity combinations spanning 24-hours, with 1 to 3 diary-days for any
single study individual.
The exposure assessment performed here requires information on activity patterns over a
full year. Long-term multi-day activity patterns were estimated from single days by combining
the daily records using an algorithm that represents the day-to-day correlation of activities for
individuals. The algorithm first uses cluster analysis to divide the daily activity pattern records
into groups that are similar, and then select a single daily record from each group. This limited
number of daily patterns is then used to construct a long-term sequence for a simulated
individual, based on empirically-derived transition probabilities. This approach is intermediate
between an assumption of no day-to-day correlation (i.e., re-selection of diaries for each time
period) and perfect correlation (i.e., selection of a single daily record to represent all days).
Details regarding the algorithm and supporting evaluations are provided in Appendix B-4,
Attachments 2 and 3.
8.7 CALCULATING MICROENVIRONMENTAL CONCENTRATIONS
Probabilistic algorithms are used to estimate the pollutant concentration associated with
each exposure event. The estimated pollutant concentrations account for temporal and spatial
variability in ambient (outdoor) pollutant concentration and factors affecting indoor
microenvironment, such as a penetration, air exchange rate, and pollutant decay or deposition
rate. APEX calculates air concentrations in the various microenvironments visited by the
179
-------
simulated person by using the ambient air data estimated for the relevant blocks/receptors, the
user-specified algorithm, and input parameters specific to each microenvironment. The method
used by APEX to estimate the microenvironmental concentration depends on the
microenvironment, the data available for input to the algorithm, and the estimation method
selected by the user. The current version of APEX calculates hourly concentrations in all the
microenvironments at each hour of the simulation for each of the simulated individuals using one
of two methods: by mass balance or a transfer factors method. Details regarding the algorithms
used for estimating specific microenvironments and associated input data derivations are
provided in Appendix B.
Briefly, the mass balance method simulates an enclosed microenvironment as a well-
mixed volume in which the air concentration is spatially uniform at any specific time. The
concentration of an air pollutant in such a microenvironment is estimated using the following
processes:
• Inflow of air into the microenvironment
• Outflow of air from the microenvironment
• Removal of a pollutant from the microenvironment due to deposition, filtration, and
chemical degradation
• Emissions from sources of a pollutant inside the microenvironment.
A transfer factors approach is simpler than the mass balance model, however, most
parameters are derived from distributions rather than single values to account for observed
variability. It does not calculate concentration in a microenvironment from the concentration in
the previous hour as is done by the mass balance method, and the transfer factors approach
contains only two parameters. A proximity factor is used to account for proximity of the
microenvironment to sources or sinks of pollution, or other systematic differences between
concentrations just outside the microenvironment and the ambient concentrations (at the
measurements site or modeled receptor). The second parameter, a penetration factor, quantifies
the amount of outdoor pollutant penetrates into the microenvironment.
8.7.1 Microenvironments Modeled
In APEX, microenvironments represent the exposure locations for simulated individuals.
For exposures to be estimated accurately, it is important to have realistic microenvironments that
180
-------
match closely to the locations where actual people spend time on a daily basis. As discussed
above, the two methods available in APEX for calculating pollutant levels within
microenvironments were mass balance or a transfer factors approach. Table 8-9 lists the
microenvironments used in this study, the calculation method used, and the type of parameters
needed to calculate the microenvironment concentrations.
Table 8-9. List of microenvironments modeled and calculation methods used.
Microenvironment
Indoors - Residence
Indoors - Bars and restaurants
Indoors - Schools
Indoors - Day-care centers
Indoors - Office
Indoors- Shopping
Indoors - Other
Outdoors - Near road
Outdoors - Public garage - parking lot
Outdoors - Other
In-vehicle - Cars and Trucks
In-vehicle - Mass Transit (bus, subway,
train)
1 AER=air exchange rate, DE=decay-depos
PE=penetration factor
Calculation
Method
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Factors
Factors
Factors
Factors
Factors
Parameter Types
used1
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
PR
PR
None
PE and PR
PE and PR
tion rate, PR=proximity factor,
8.7.2 Microenvironment Descriptions
8.7.2.1 Microenvironment 1: Indoor-Residence
The Indoors-Residence microenvironment uses several variables that affect NC>2
exposure: whether or not air conditioning is present, the average outdoor temperature, the NC>2
removal rate, and an indoor concentration source.
Air conditioning (A/C) status of an individual's residential microenvironment was
simulated randomly using the probability that a residence has an air conditioner. For the Atlanta
modeling domain an air-conditioning prevalence of 97.0 % was used (American Housing Survey
or AHS, 2004).
Air exchange rate (AER) data for the indoor residential microenvironment were the same
used in APEX for the most recent Oj NAAQS review (EPA, 2007g; see Appendix B,
181
-------
Attachment 5). Briefly, AER data were reviewed, compiled, and evaluated from the extant
literature to generate location-specific AER distributions categorized by influential factors,
namely, location, temperature, and presence of A/C. The AER data obtained was limited in the
number of samples, particularly when considering these influential factors. When categorizing
by temperature, a range of temperatures was used to maintain a reasonable number of samples
within each category to allow for some variability within the category, while still allowing for
differences across categories. Several distribution forms were investigated (i.e., exponential,
log-normal, normal, and Weibull) and in general, lognormal distributions provided the best fit.
Fitted lognormal distributions were defined by a geometric mean (GM) and standard deviation
(GSD). Because no fitted distribution was available specifically for Atlanta, distributions were
selected from other locations thought to have similar characteristics, qualitatively considering
factors that might influence AERs including the age composition of housing stock, construction
methods, and other meteorological variables not explicitly treated in the analysis, such as
humidity and wind speed patterns. To avoid unusually extreme simulated AER values, bounds
of 0.1 and 10 were selected for minimum and maximum AER, respectively. Table 8-10
summarizes the distributions used by A/C prevalence and temperature categories. See Appendix
B, Attachment 2 for additional details.
Table 8-10. Geometric means (GM) and standard deviations (GSD) for air exchange rates by A/C
type and temperature range used for Atlanta exposure assessment.
A/C Type
No A/C1
Central or
Room A/C2
Temp
(°C)
<=10
10-20
>20
<=10
10-20
20-25
>25
N
61
87
44
157
320
196
145
GM
0.9258
0.7333
1.3782
0.9617
0.5624
0.3970
0.3803
GSD
2.0836
2.3299
2.2757
1.8094
1.9058
1.8887
1.7092
Notes:
1 Distribution derived from Research Triangle Park study. See Appendix
B, Attachment 5.
2 Distribution derived from non-California cities. See Appendix B,
Attachment 5.
The same NC>2 removal rate distribution was used for all indoor microenvironments that
use the mass balance method. This removal rate is based on data provided by Spicer et al. (1993)
and was approximated with a uniform distribution, U{1.02, 1.45 h"1} based on the six reported
values from a single house (Table 8-12).
182
-------
Table 8-11. Data used to estimate removal rate constant for indoor microenvironments.
Source
Introduction
injection
injection
radiant heater
convective heater
range
injection
House
Temperature
o
c
23
21
26
24
22
22
Relative
Humidity
(%)
52
54
54
55
54
55
Air
Exchange
Rate
(h-1)
0.15
0.16
0.15
0.12
-
0.15
Removal
Constant
k
(h-1)
1.28
1.02
1.15
1.04
1.45
1.13
Notes:
Data from Table 1 of Spicer et al. (1 993).
An indoor emission source term was included in the APEX simulations to estimate NCh
exposure to gas cooking (hereafter referred to as "indoor sources"). This was the only indoor
source considered in this assessment. Three types of data were used to generate the emission
factor for this indoor source: (1) the fraction of households in the Atlanta MSA that use gas for
cooking fuel, (2) the range of contributions to indoor NCh concentrations that occur from
cooking with gas, and (3) the diurnal pattern of cooking in households.
The fraction of households in Atlanta that use gas cooking fuel (i.e., 39%) was obtained
from AHS (2004). Data used for estimating the contribution to indoor NCh concentrations that
occur during cooking with gas fuel were derived from a study sponsored by the California Air
Resources Board (CARB, 2001). A uniform distribution of concentration contributions for input
to APEX was estimated as U{4, 188 ppb}. An analysis by Johnson et al (1999) of survey data
on gas stove usage collected by Koontz et al (1992) showed an average number of meals
prepared each day with a gas stove of 1.4. The diurnal allocation of these cooking events was
estimated using food preparation time obtained from CHAD diaries, stratified by hour of the day,
and normalized to the expected value of daily food preparation events of 1.4 (Table 8-12).
Table 8-12. Probability of gas stove cooking by hour of the day.
Hour of the
Day
0
1
2
3
4
Probability
of Cooking
(%)1
0
0
0
0
0
183
-------
Hour of the
Day
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Probability
of Cooking
(%)1
5
10
10
10
5
5
5
10
5
5
5
15
20
15
10
5
5
0
0
Notes:
1 Values rounded to the nearest 5%. Data sum to 1 45%
due to rounding convention and the scaling to represent
1 .4 cooking events/day.
8.7.2.2 Microenvironments 2-7: All other indoor microenvironments
The remaining five indoor microenvironments, which represent Bars and Restaurants,
Schools, Day Care Centers, Office, Shopping, and Other environments, were all modeled using
the same data and functions. An air exchange rate distribution (GM = 1.109, GSD = 3.015, Min
= 0.07, Max = 13.8) was based on an indoor air quality study (Persily et al, 2005). This is the
same distribution in APEX used for the most recent Oj NAAQS review (EPA, 2007g). See
Appendix B, Attachment 5 for details in the data used and derivation. The removal rate is the
same uniform distribution used in the Indoor-Residence microenvironment (section 8.7.2.1).
The Bars and Restaurants microenvironment included an estimated contribution from indoor
sources as was described for the Indoor-Residence, only there was an assumed 100% prevalence
rate for cooking with a gas appliance and it occurred at any hour of the day.
8.7.2.3 Microenvironments 8 and 9: Outdoor Microenvironments
Two outdoor microenvironments, the Near Road and Public Garage/Parking Lot, used the
transfer factors method to calculate pollutant exposure. Penetration factors are not applicable to
184
-------
outdoor environments (effectively, PEN=1). The distribution for proximity factors were
developed from the dispersion model estimated concentrations, using the relationship between
on-road to receptor estimated concentrations.
8.7.2.4Microenvironment 10: Outdoors-General
The general outdoor environment concentrations are represented by the AERMOD
predicted concentrations. Therefore, both the penetration factor and proximity factor for this
microenvironment were set to 1.
8.7.2.5 Microenvironments 11 and 12: In Vehicle- Cars and Trucks, and Mass Transit
Penetration factors were developed from data provided in Chan and Chung (2003). Since
major roads were the focus of this assessment, reported indoor/outdoor ratios for highway and
urban streets were used here. Mean values range from about 0.6 to just over 1.0, with higher
values associated with increased ventilation (i.e., window open). A uniform distribution U{0.6,
1.0} was selected for the penetration factor for Inside-Cars/Trucks due to the limited data
available to describe a more formal distribution and the lack of data available to reasonably
assign potentially influential characteristics such as use of vehicle ventilation systems for each
location. Mass transit systems, due to the frequent opening and closing of doors, was assigned a
uniform distribution U{0.8, 1.0} based on the reported mean values for fresh-air intake (0.796)
and open windows (1.032) on urban streets. Proximity factors were developed from the
dispersion model estimated concentrations, using the relationship between the on-road to
receptor estimated concentrations. The proximity distributions were stratified based using time-
of day and season bins and are provided in Appendix B-4.
8.8 EXPOSURE MEASURES AND HEALTH RISK CHARACTERIZATION
APEX calculates the time series of exposure concentrations that a simulated individual
experiences during the simulation period. APEX determines the exposure using hourly ambient
air concentrations, calculated concentrations in each microenvironment based on these ambient
air concentrations (and indoor sources if present), and the minutes spent in a sequence of
microenvironments visited according to the composite diary. The hourly exposure concentration
at any clock hour during the simulation period is determined using the following equation:
185
-------
N
/~< hourly-mean
1 (}}
/-i hourly-mean
^ME(j)
r _ 7=1 _
^ i — ~ equation (8-1)
where,
Ci = Hourly exposure concentration at clock hour /' of the simulation period
(ppm)
N = Number of events (i.e., microenvironments visited) in clock hour /' of
the simulation period.
CMEU™"" = Hourly mean concentration in microenvironmenty (ppm)
t(j) = Time spent in microenvironmenty (minutes)
T =60 minutes
From the hourly exposures, APEX calculates time series of 1-hour average exposure
concentrations that a simulated individual would experience during the simulation period.
APEX then statistically summarizes and tabulates the hourly (or daily, annual average)
exposures. In this analysis, the exposure indicator is 1-hr exposures above selected health effect
benchmark levels. From this, APEX can calculate two general types of exposure estimates:
counts of the estimated number of people exposed at or above a specified NC>2 concentration
level and the number of times per year that they are so exposed; the latter metric is in terms of
person-occurrences or person-days. The former highlights the number of individuals exposed at
least one or more times per modeling period to the potential health effect benchmark level of
interest. APEX can also report counts of individuals with multiple exposures. This person-
occurrences measure estimates the number of times per season that individuals are exposed to the
exposure indicator of interest and then accumulates these estimates for the entire population
residing in an area.
APEX tabulates and displays the two measures for exposures above levels ranging from
100 to 300 ppb by 50 ppb increments for 1-hour average exposures. These results are tabulated
for the population and subpopulations of interest.
186
-------
8.8.1 Adjustment for Just Meeting the Current and Alternative Standards
We used a different approach to simulate just meeting the current and alternative
standards than was used in the Air Quality Characterization (Appendix A). In this case, instead
of adjusting upward16 the air quality concentrations, to reduce computer processing time, we
adjusted the health effect benchmark levels by the same factors described for each specific
location and simulated year (Table 8-13). Since it is a proportional adjustment, the end effect of
adjusting concentrations upwards versus adjusting benchmark levels downward within the model
is the same. The same follows for where as is concentrations were in excess of an alternative
standard level (e.g., 50 ppb for the 98th percentile averaged over three years), only the associated
benchmarks are adjusted upwards (i.e., a higher threshold simulating lower exposures).
Table 8-13. Adjusted potential health effect benchmark levels used by APEX to simulate just
meeting the current standard and various alternative standards considered.
Model
Scenario
As- is
Current
Standard
Alternative
Standards
Averaging
Time
Annual
1 hour
Cone
(ppb)
53
50
100
150
200
Conditions
Year 2001
Year 2002
Year 2003
98th %ile
99th %ile
98th %ile
99th %ile
98th %ile
99th %ile
98th %ile
99th %ile
Potential health effect benchmark level (ppb)
100
100
44
37
31
163
177
82
89
54
59
41
44
150
150
66
55
46
nd
nd
nd
nd
nd
nd
nd
nd
200
200
88
73
62
327
355
163
177
109
118
82
89
250
250
110
91
77
nd
nd
nd
nd
nd
nd
nd
nd
300
300
132
110
93
490
532
245
266
163
177
123
133
Notes:
nd not done due to model constraints on number of levels possible in one model simulation.
When modeling indoor sources, the indoor concentration contributions needed to be
scaled by the similar proportions. This additional scaling was necessary so as not to affect the
impact of the estimated indoor concentrations while adjusting the benchmark levels. The
following presents the justification of why it can appropriate to use the same proportional factor
to adjust the indoor source concentration contribution.
16 To evaluate the current and most of the alternative standards proposed, ambient concentrations were lower than air
quality that would just meet the standards.
187
-------
Exposure concentrations an individual experiences are first defined as the sum of the
contribution from ambient concentrations and from indoor sources (if present) and this
concentration can be either above or below a selected concentration level of interest. This is
represented by the following equation:
^" exposure ^ambient ^indoor ^thresh
mid
where,
Cexposure = individual exposure concentration (ppm)
a = proportion of exposure concentration from ambient (unitless fraction)
= ambient concentration in the absence of indoor sources
= proportion of exposure concentration from indoor (unitless fraction,
equivalent to 1-a)
= direct indoor source concentration contribution in the absence of
ambient influence (ppm)
= an exposure concentration of interest (ppm)
It follows that if we are interested in adjusting the ambient concentrations upwards by
some proportional factor/(a unitless number), this can be described with the following:
- ambient indoor > ^ threshold
This is equivalent to
a^ ambient + ^(^ indoor ' J ) > (^ threshold ' J )
Therefore, if the potential health effect benchmark level and the indoor concentrations are
both proportionally scaled downward by the same adjustment factor, the contribution of both
sources of exposure (i.e., ambient and indoor) are maintained and the same number of estimated
exceedances would be obtained as if the ambient concentration were proportionally adjusted
upwards by factor/
188
-------
8.9 EXPOSURE MODELING AND HEALTH RISK
CHARACTERIZATION RESULTS
8.9.1 Overview
The results of the exposure and risk characterization are presented here for the four
county modeling domain in Atlanta. Several exposure scenarios were considered for the
exposure assessment including an analysis of three averaging times for NO2 concentrations
(annual, 24-hour, and 1-hour), an analysis of the contribution to NO2 exposures of both indoor
and outdoor sources, and an analysis of NO2 exposures assuming air quality that just meets the
current annual and several alternative 1-hour daily maximum standards. The year 2002 served as
the base year for all scenarios, while 2001 and 2003 were only evaluated for a limited number of
scenarios. Exposures were simulated for four population groups; all persons, all children (ages
5-17), all asthmatics, and asthmatic children (ages 5-17).
The exposure results that are summarized below focus on asthmatics. Key results are
presented in the next three subsections, with complete results for each of these two population
subgroups provided in Appendix B-4. In addition, due to limitations in the data summaries
output from the current version of APEX, certain exposure data could only be output for the
entire population modeled (i.e., all persons - includes asthmatics and healthy persons of all ages)
rather than the particular subpopulation. The summary exposure results for the entire population
(e.g., annual average exposure concentrations, time spent in microenvironments at or above a
potential health effect benchmark level) is assumed representative of the asthmatic population in
the modeling results because the asthmatic population does not have its microenvironmental
concentrations and activities estimated any differently from those of the total population. The
assumption of modeling asthmatics similarly to healthy individuals (i.e., using the same time-
location-activity profiles) is supported by the findings of van Gent et al. (2007), at least when
considering children 7-10 years in age. These researchers used three different activity-level
measurement techniques; an accelerometer recording 1-minute time intervals, a written diary
considering 15-minute time blocks, and a categorical scale of activity level. Based on analysis of
5-days of monitoring, van Gent et al. (2007) showed no difference in the activity data collection
methods used as well as no difference between asthmatic children and healthy children when
comparing their activity levels.
189
-------
8.9.2 Annual Average Exposure Concentrations (as is)
Figure 8-9 illustrates the annual average exposure concentrations for the total simulated
population (i.e., both asthmatics and healthy individuals of all ages), considering the modeled
year 2002 air quality (as is) and both with and without indoor sources. Also plotted on this
figure is the distribution of the annual average NC>2 concentrations predicted by AERMOD
separated into two broad receptor categories. As a point of reference, the measured annual
average concentration for the three ambient monitors in the Atlanta modeling domain ranged
from 15 ppb to 19 ppb in year 2002. About one-half of the AERMOD predicted annual average
NC>2 concentrations for the non-road receptors were below the range of the ambient monitoring
annual average concentrations, with most receptors predicted to be less than 30 ppb. About 5%
of these receptors had concentrations above this level. It should be noted that the non-road
receptors included here could have a number of block centroids located near a major road.
Consistent with what was observed in the air quality characterization data for on-road
concentration estimates, the AERMOD long-term average concentrations predicted at the
roadway links are about twice that of the estimated concentrations at non-road receptors.
The hourly NO2 concentrations output from AERMOD were input into the exposure
model, providing a wide range of estimated exposures calculated by APEX (Figure 8-9). All
persons were estimated to experience exposures below an annual average exposure of 53 ppb,
even when considering indoor source concentration contributions. The estimated annual average
exposures were below that of both the modeled receptors and the measured air quality. For
example, the median annual average exposure was about 6 ppb less than the modeled median
non-road receptor concentration when the exposure estimation included indoor sources, and
about 9 ppb less when annual average exposures were estimated without the indoor sources. In
the absence of indoor source contributions, personal exposure concentrations for most of the
simulated individuals are estimated to be about 40 to 70 percent that of the local ambient or
outdoor concentration. This estimate is consistent with studies reporting such a relationship
based on measurements of personal exposure and ambient concentrations that ranges from
around 0.3 to 0.6 (Table AX3.5-lb, ISA ANNEX).
In comparing the estimated exposures with and without indoor sources, indoor sources
were estimated to contribute between 1 and 4 ppb to the total annual average exposures. This
would correspond to indoor sources contributing approximately 1/3 of the annual average
190
-------
exposures for persons using gas cooking appliances. Again, while Figure 8-9 summarizes the
entire population, the data are representative of what would be observed for the population of
asthmatics or asthmatic children.
Year-to year-variation was evaluated by comparing the estimated annual average
exposure distributions for each year simulated. Each simulated year of data was very similar,
with estimated median annual average exposures at about 10 ppb and 95% of the simulated
individuals' annual average exposures within the interval from 5.9 to 15.8 ppb (Figure 8-10).
100 -f
AERMOD - Non Road Receptors
AERMOD - On-Road Receptors
APEX - Exposure - Indoor Sources
APEX - Exposure - No Indoor Sources
10
20 30 40 50 60
Annual Average NO2 Concentration (ppb)
70
80
Figure 8-9. Comparison of annual average AERMOD predicted NOi concentrations (on-
road and non-road receptors) and APEX modeled NOi exposures (with and without
modeled indoor sources) in Atlanta modeling domain for year 2002.
191
-------
100 f
x 2001 - No Indoor Sources
+ 2002 - No Indoor Sources
o 2003 - No Indoor Sources
10 15 20 25
Annual Average NO2 Exposure (ppb)
30
35
Figure 8-10. Comparison of estimated annual average NOi exposures for Years 2001-2003
in Atlanta modeling domain without modeled indoor sources.
8.9.3 Daily Average Exposures (as is)
As mentioned earlier, APEX is capable of providing exposure results across a variety of
averaging times, including 24-hour average exposures. This averaging time serves as a good
point of comparison with the personal exposures reported in the published literature. As
mentioned above regarding APEX default results, the daily mean exposures were estimated for
the total simulated population. In this simulation, each person has 365 daily mean personal
exposures, thus each individual experiences a daily average concentration distribution (i.e., each
person has a median daily average exposure, a 99th percentile daily average exposure, etc.).
These modeled exposures were compared with personal NC>2 measurement data obtained from
Suh (2008) for the participants of an Atlanta epidemiological study conducted by Wheeler et al.
(2006). The personal exposure measurements were collected across two seasons (fall and
spring)17 and considered cooking fuel (gas or electric cooking) as an influential variable for
personal exposures. A total of 30 individuals participated in the study, of whom 13 subjects had
personal exposure measurements for both seasons, with no persons using both cooking fuels.
17 Fall was designated here for sample collection dates reported in the months of September, October, and November
1999; Spring was designated where sample collection dates were reported in the months of April and May 2000.
192
-------
An average of 6 daily average personal exposure measurements was available for each individual
when stratified by season and cooking fuel (minimum number of days = 3, max = 7). Because
there were few personal exposure measurements, an exposure distribution was constructed for
each individual, simply using their minimum, median, and maximum daily mean exposures and
are summarized in Figure 8-11. In comparing the median personal daily mean exposures using
the two stratification variables, two patterns can be noted. First, the use of gas as a cooking fuel
increased daily median personal exposures by about 3 to 10 ppb in both seasons. Second,
seasonal differences were also present, with personal daily average exposures higher during the
spring by about 1 to 3 ppb when comparing the individual median values for the persons
employing gas or electric cooking. While these general patterns are noted, it should be added
that the maximum daily average exposures were highest in the spring and similar for both of the
cooking fuel categories.
Daily mean exposures estimated using APEX were also evaluated in a similar manner, by
stratifying the results based on the same seasons and whether or not indoor sources were
included in the model simulation. The specific period from 1999-2000 was not modeled by
APEX although this period was included in the personal exposure measurement study. The
APEX simulation results for year 2002 were selected for comparison with the exposure
measurements obtained from Suh (2008). A distribution of each person's estimated daily
exposure was constructed, using the median daily mean exposure to represent the central
tendency and a 95 % prediction interval to represent the lower and upper bounds of exposure
(i.e., the 2.5th and the 97.5th percentiles). This prediction interval was chosen rather than using
the minimum and maximum as done with the personal measurements because APEX estimated
61 and 91 days of exposure for each individual in the spring and fall months, respectively. The
APEX results would likely capture more variability in exposures given the greater number of
days in comparison with the personal exposure measures that contained at most 7 days of data
per season.
The daily mean exposures estimated using APEX, stratified by season and by inclusion of
indoor sources, are presented in Figure 8-12. The distributions of median daily mean exposures
are comparable to one another, although the fall season was about 1 ppb higher than the spring
exposures. The range of estimated daily mean exposures, given by the 95% prediction interval,
was also similar across the season categories. In comparing the simulations where indoor
193
-------
sources were modeled to the simulations conducted without indoor source contributions, the
estimated exposures were between 1 to 4 ppb greater for the indoor source simulations. It should
be noted that the indoor source exposure distributions include exposures for all of the simulated
individuals, some of which do not have gas cooking occurring at home. The APEX simulated
daily mean exposures are similar to the measured personal exposures (Figure 8-11) when
considering the values and range of the median concentrations as well as the values and range of
the bounding percentiles given by the 95% prediction intervals.18
18 While a direct comparison of APEX estimated maximum daily exposure concentrations with the maximum
observed daily personal exposure concentrations is considered questionable given the large discrepancy in sample
sizes, it should be noted that approximately 99.1% of APEX simulated persons had their estimated maximum daily
exposure concentrations within the maximum observed daily personal exposure measurement of 78.2 ppb without
gas cooking. Approximately 97.5% of APEX simulated persons had their estimated maximum daily exposure
concentrations within the maximum observed daily personal exposure measurement of 85.4 ppb with gas cooking.
194
-------
FALL 1999
O
O
O
O
O
oi
o
LU
_i
LLI
O
O
O
O
100 -F
90
SPRING 2000
1 10
Daily Mean NO2 Exposure (ppb)
1003.1
1 10
Daily Mean NO2 Exposure (ppb)
100
Figure 8-11. Distribution of measured daily average personal NOi exposures for individuals in Atlanta, stratified by two
seasons (fall or spring) and cooking fuel (gas or electric). Minimum (min), median (p50), and maximum (max) were obtained
from each individual's multi-day exposure measurements. The figure generated here was based on personal exposure
measurements obtained from Suh (2008).
195
-------
FALL
0)
0
O
O
0)
o
o
(O
1_
o
o
100 -F
90
80
SPRING
0)
£
a)
Q.
g
1
3
O
a)
Q.
g
O
100 -F
1
10
100 0.1
Daily Mean NO2 Exposure (ppb)
1 10
Daily Mean NO2 Exposure (ppb)
100
Figure 8-12. Distribution of estimated daily average NOi exposures for individuals in Atlanta, stratified by two seasons (fall or
spring) and with and without indoor sources, for Year 2002 APEX simulation. Lower bound (2.5th percentile, p2.5), median
(p50), and upper bound (97.5th percentile, p97.5) were calculated from each simulated persons 365 days of exposure. A
random sample of 5% of persons (about 2,500 individuals) is presented in each figure to limit the density of the graphs.
196
-------
8.9.4 One-Hour Exposures
8.9.4.1 Overview
Because the focus of the exposure and risk characterization is on short-term 1-hour daily
maximum exposures, analyses were performed using the APEX estimated 1-hour exposure
concentrations. The number of exposures above the selected potential health effect benchmark
levels (i.e., 100, 150, 200, 250, and 300 ppb, 1-hour average) were estimated. An exceedance
was recorded when the maximum exposure concentration estimated for the individual was above
the selected benchmark level in a day. Estimates of repeated exposures are also recorded, that is
where 1-hour exposure concentrations were above a selected benchmark level in a day added
together across multiple days (therefore, the maximum number of multiple exceedances per
individual is 365). Persons of interest in this exposure analysis are those with particular
susceptibility to NC>2 exposure, namely individuals with asthma. The potential health effect
benchmark levels used are appropriate for characterizing the potential risk of adverse health
effects for asthmatics. The majority of the results presented in this section are for the entire (i.e.,
all ages) simulated asthmatic population because the pattern of exposure results for asthmatic
children were very similar. However, the exposure analysis was also performed for the total
population to assess numbers of persons exposed to these levels and to provide additional
information relevant to the asthmatic population (such as time spent in particular
microenvironments). The 1-hour exposure results are presented separately for three scenarios,
(1) considering the exposures associated with as is air quality, (2) simulating exposures with air
quality adjusted upwards to represent just meeting the current annual average standard, and (3)
simulating exposures associated with air quality adjusted to represent just meeting alternative 1-
hour daily maximum standards. In addition, the presence (or not) of indoor sources was also
considered within each of these three scenarios.
8.9.4.2 Estimated Number of 1-hour Exposures Above Selected Levels (as is)
The results presented in this section were generated from the modeled air quality as input
to APEX without any adjustment to the air concentrations or the potential health effect
benchmark levels. Figure 8-13 summarizes the estimated number of asthmatics exposed at each
of the potential health effect benchmark levels using the modeled air quality for each year,
197
-------
without any contribution from indoor sources. As observed with the annual average exposure
concentrations, there is great similarity in the estimated numbers of exceedances for each of the
three years modeled. Year-to-year variability in the number of asthmatics exposed as indicated
by a coefficient of variation (COV=mean/standard deviation) was at most 3.3%, calculated for
the 300 ppb benchmark level. All persons (i.e., just over 212,000) were estimated to be exposed
at least one time to a 1-hour daily maximum concentration of 100 ppb in a year. The number of
asthmatics exposed to greater concentrations (e.g., 200 or 300 ppb) drops only slightly and is
estimated to be somewhere between 117,000 - 196,000 depending on the 1-hour concentration
level and year of air quality simulated. Similar patterns across the benchmark levels were
observed for simulated asthmatic children, albeit with lower total numbers of asthmatic children
with exposures at or above the potential health effect benchmark levels.
The results for all asthmatics and asthmatic children were similar in terms of the
proportion of the population exposed and the year-to year variability in numbers of exceedances.
For example, nearly 61,000 asthmatic children were estimated to be exposed one time to a 1-
hour daily maximum NC>2 concentration of at least 200 ppb for year 2002, comprising about 95%
of that subpopulation (Figure 8-14). The number of children with at least one exceedance of 300
ppb was less, estimated to be about 41,000 using the 2002 air quality, or about 64% of all
asthmatic children. As a comparison, the percent of all asthmatics experiencing exposures at or
above 200 and 300 ppb was 92% and 59%, respectively. The year-to-year variability in the
number of asthmatic children exposed at or above the selected benchmark levels was also small,
although slightly higher than that estimated for all asthmatics. The highest COV for asthmatic
children using the 3-year exposure estimates was also observed for exceedances of the 300 ppb
benchmark (COV = 4.9%).
198
-------
2.5E+5-
»-
.32
E c ^-Ob+o
.c —
•% -5
2 5
o ~" 1 C;F+C;
Estimated Number
Exposed to Selectee
3 01
3 0 0
m m
3 i 01
,-
|^— — .
"~~T— — •
*T
I
^
-^l
^
^f
— L,
x'
/
-^
-V
r
--
_£
•—
--
-
- -----
/
150
200
2003 AQ (as is) - No Indoor
2002 AQ (as is) - No Indoor
2001 AQ (as is) - No Indoor
250
300
Potential Health Effect Benchmark Level (ppb)
Figure 8-13. Estimated number of all simulated asthmatics in the Atlanta model domain
with at least one NOi exposure at or above the potential health effect benchmark levels,
using modeled 2001-2003 air quality (as is), without indoor sources.
o ai
^ S
o 01
^ OT
O
1 &
CO
|s
fflS
o
100
150
200
2003 AQ (as is) - No Indoor
2002 AQ (as is) - No Indoor
2001 AQ (as is) - No Indoor
250
300
Potential Health Effect Benchmark Level (ppb)
Figure 8-14. Estimated number of simulated asthmatic children in the Atlanta model
domain with at least one NOi exposure at or above the potential health effect benchmark
levels, using modeled 2001-2003 air quality (as is), without modeled indoor sources.
199
-------
Additional exposure estimates were generated using the modeled 2002 air quality (as is).
Those estimates include an evaluation of the contribution of indoor sources. APEX allows for
the same persons to be simulated (i.e., demographics of the population were conserved), as well
as using the same individual time-location-activity profiles generated for each person. Figure 8-
15 illustrates the estimated number of asthmatics experiencing exposures above the potential
health effect benchmarks, both with indoor sources and without indoor sources included in the
model runs. The number of asthmatics at or above the selected benchmark levels at least one
time in a year is very similar when including indoor source concentration contributions (i.e., gas
cooking) compared to the number of persons whose exposure estimates did not include indoor
sources. The reduction in numbers of asthmatics exposed at least once at or above any potential
health effect benchmark level ranged from 0 to around 5,000 when indoor source contributions
were excluded.
The number of person-days of exposure at or above a given potential benchmark level
gives a different perspective on the contribution of indoor sources. Figure 8-16 illustrates the
total number of days where the particular concentration level was exceeded, representing the sum
of all multiple exposures (in contrast to focusing on persons as was done for example in Figure
8-13) for the simulated population in a given year. Since most individuals were exposed at least
one time at many of the 1-hour levels, it was difficult to discern the effect that indoor sources
had on the estimated exposures. Now it can be seen that the indoor source contribution increases
not just the number of persons exposed, but more importantly how many times they would be
exposed per year above the selected benchmark level. It appears that on average, there is an
increase in the number of person-days by about a factor of 2.1 and 1.8 for the 100 and 150 ppb 1-
hour concentration levels, respectively, while the higher benchmark levels are largely unaffected
by the presence of indoor sources.
An evaluation of the time spent in the 12 microenvironments was performed to estimate
where simulated individuals are exposed to concentrations above the potential health effect
benchmark levels. Currently, the output generated by APEX is limited to compiling the
microenvironmental time for the total population (includes both asthmatic individuals and
healthy persons) and the summaries provide the total time spent above the selected potential
health effect benchmark levels. As mentioned above, the data still provide a reasonable
approximation for each of the population subgroups (e.g., asthmatics or asthmatic children)
200
-------
because their microenvironmental concentrations and activities are not estimated any differently
from those of the total population simulated by APEX.
«- i; ra
in ra
&.E
2.5E+5
2.0E+5
•n w
2 ~
in 01
I 3 1.0E+5
E ^3
» ai 5.0E+4
< OT
O.OE+0
100
150
200
250
~1——______
300
Potential Health Effect Benchmark Level (ppb)
2002 AQ (as is) with indoor
2002 AQ (as is) no indoor
Figure 8-15. Estimated number of all simulated asthmatics in the Atlanta model domain
with at least one NOi exposure at or above potential health effect benchmark levels, using
modeled 2002 air quality (as is), both with and without modeled indoor sources.
£ c
E °
= £
•c <" ra
S^ £ c
p 5 -o ~
.E E a)
100
150
200
2002 AQ (as is) with indoor
2002 AQ (as is) no indoor
250
300
Potential Health Effect Benchmark Level (ppb)
Figure 8-16. Estimated number asthmatic person-days in the Atlanta model domain with
an NO2 exposure at or above potential health effect benchmark levels, using modeled 2002
air quality (as is), both with and without modeled indoor sources.
201
-------
As an example, Figure 8-17 (a, b, c) summarizes the percent of total time spent in each
microenvironment for simulation year 2002 that was associated with estimated exposure
concentrations at or above 100, 200, and 300 ppb (results for years 2001 and 2003 were similar).
These estimated exposures summarized in this figure did not include the contribution from
indoor sources. Time spent in the indoor microenvironments contributed little to the occurrence
of estimated exposures at or above the selected benchmark levels. Most indoor
microenvironments contributed < 1% of exposures to 1-h concentrations above 100 ppb and
none of them contributed at all to exceedances of the 200 and 300 ppb benchmark levels. Most
of the time associated with the high short-term exposures was associated with the transportation
microenvironments (In-Vehicle or In-Public Transport) or outdoors (Outdoors-Near Road,
Outdoors-Parking Lot, Outdoors-Other). The time spent outdoors near roadways exhibited an
increase in contribution of exceedances of potential health benchmark levels, increasing from
around 25 to 29% of time associated with concentrations of 100 and 300 ppb, respectively. The
in-vehicle microenvironment showed a corresponding decrease, estimated as contributing to 65%
of the time associated with 100 ppb exceedances, while contributing to 58% of 1-hour daily
maximum exposures at or above 300 ppb. While more persons are likely to spend more time
inside a vehicle than outdoors near roads, there is attenuation of the estimated on-road
concentration that penetrates the in-vehicle microenvironment, leading to lowered
concentrations. The result of this is that exposures above 300 ppb occur less frequently in-
vehicles when compared with the outdoor near-road microenvironment that involves no
attenuation of concentrations.
The microenvironments where the exposure exceedances occur were also identified for
the estimated exposures that included indoor source contributions (Figure 8-18). While the
transportation-associated microenvironments remained important for exposures above each of
the selected levels, the time spent in the indoor microenvironments was also important for
exceedances of hourly levels of 100 ppb, contributing to approximately 26% (inside a home) and
33% (inside bar/restaurant) of the time persons were exposed (Figure 8-18a). This is a result of
the indoor source contribution to each individual's exposure concentrations and is consistent
with what was observed regarding the effect of indoor sources on the total person-days of
exposure. However, the importance of the indoor microenvironments decreases with the
increasing benchmark levels. Exposures at or above 200-300 ppb occur rarely in the indoor
202
-------
microenvironments, even when considering the indoor source contributions. The exposures at
the higher benchmarks are associated mainly with the transportation microenvironments,
increasing from about 39% of the time exposures occurred at the lowest potential health effect
benchmark level (100 ppb) to comprising 100% of the time exposures occurred at the highest
benchmark level (300 ppb, Figure 8-18c).
In the above analysis of persons exposed, the results show the number or percent of those
with at least one day on which the 1-hour exposure was at or above the selected potential health
effect benchmark level. Given that the benchmark is for a relatively short averaging time (i.e., 1-
hour) it may be possible that individuals are exposed to concentrations at or above the potential
health effect benchmark levels on several days in a given year. Since APEX simulates the
longitudinal diary profile for each individual, the number of days with a 1-hour daily maximum
exposure above a selected level is retained for each person. Figure 8-19 presents such an
analysis for the year 2002, where the estimated exposures did not include indoor source NC>2
contributions. Nearly all simulated asthmatics (98.7%) experienced up to six exposures at or
above 100 ppb, with nearly 78% experiencing at least six exposures at or above the 150 ppb
level. Multiple exposures at or above the higher potential health effect benchmark levels were
less frequent, with around 58, 28, and 12 percent of asthmatics exposed annually to four or more
1-hour NC>2 concentrations greater than or equal to 200, 250, and 300 ppb, respectively.
203
-------
a) > 100 ppb
In-Public Trans
In-Vehicle
b) > 200 ppb
In-Public Trans^
c) > 300 ppb
In-Public Trans
In-Other pOther
-In-Residence
\
\ -In-Bar& Restaurant
-In-School
-In-Day Care
\
Out-Near Road
Out-Parking Lot
^—Out-Other
-Other
-In-Residence
-In-Bar & Restaurant
-In-School
-In-Day Care
-In-Office
-In-Shopping
Out-Near Road
—Out-Parking Lot
Out-Other
In-Other
\
pOther
-In-Residence
-In-Bar & Restaurant
-In-School
-In-Day Care
-In-Office
lopping
Out-Near Road
—Out-Parking Lot
— Out-Other
Figure 8-17. Fraction of time all simulated persons in the Atlanta model domain spend in
the twelve microenvironments that corresponds with exceedances of the potential NOi
health effect benchmark levels, a) > 100 ppb, b) > 200 ppb, and c) > 300 ppb, year 2002 air
quality (as is) without indoor sources.
204
-------
a) > 100 ppb
In-Public Trans
In-Vehicle^
Out-Other^
Out-Parking Lot—
Out-Near Road
In-Day Care
In-Shopping—^/
In-Office—^\
/—In-Residence
b) > 200 ppb
In-Residem
Other
\_ In-Bar&
Restaurant
In-School |n.Bar& Restaurant
A—In-Other
In-Public Trans \
//X In-Shopping
-—Out-Near Road
— Out-Parking Lot
c) > 300 ppb
In-School ,—\n-Bar & Restaurant
/^In-Other
//^In-Office
\l // In-Shopping
In-Public Trans^ \ »-ln-Day Care
—Out-Near Road
—Out-Parking Lot
Figure 8-18. Fraction of time all simulated persons in the Atlanta model domain spend in
the twelve microenvironments that corresponds with exceedances of the potential NOi
health effect benchmark levels, a) > 100 ppb, b) > 200 ppb, and c) > 300 ppb, year 2002 air
quality (as is) with indoor sources.
205
-------
The contribution of indoor sources to the occurrence of repeated exposure exceedances
was also evaluated. Figure 8-20 illustrates that nearly all asthmatics (about 93%) would be
exposed at least six times to either the 1-hour daily maximum 100 ppb or 150 ppb concentration
level in a year when considering exposure to ambient NC>2 combined with indoor source
emissions. This is approximately 15% more persons than was estimated for the simulations
without indoor source contributions. The percent of asthmatics experiencing multiple exposures
above the 200, 250 and 300 ppb was only about 1-4% greater than that observed for asthmatics
without indoor sources. This is consistent with the person-day results that indicate the indoor
source emissions contribute primarily to numbers of exposures experienced at or above the 100
or 150 ppb benchmark levels.
s «
i=
Q- T;
x >
uj £
in -J
o -a
•a S
«» J1^
<2 w i-
2s s
0 (0 >•
•S «>
01 a,
51
nr ^
100
Potential Health Effect Benchmark Level
(PPb)
Estimated Number
of Repeated
Exposures in a
Year
Figure 8-19. Estimated percent of all asthmatics in the Atlanta modeling domain with
repeated NO2 exposures above potential health effect benchmark levels, using modeled
2002 air quality (as is), without indoor sources.
206
-------
100
Estimated Number
of Repeated
Exposures in a
Year
150
Potential Health Effect Benchmark Level
(PPb)
Figure 8-20. Estimated percent of all asthmatics in the Atlanta modeling domain with
repeated NOi exposures above potential health effect benchmark levels, using modeled
2002 air quality (as is), with indoor sources.
8.9.4.3 Estimated Number of 1-hour Exposures Above Selected Levels (current
standard)
To simulate just meeting the current annual average NC>2 standard, the potential health
effect benchmark levels were adjusted in the exposure model, rather than adjusting all of the
hourly concentrations for each receptor and year simulated (see section 8.8). Similar to what
was performed for the as is air quality, estimates of short-term exposures (i.e., 1-hour daily
maximum) were generated for the total population and population subgroups of interest (i.e.,
asthmatics and asthmatic children).
When considering the estimated exposures associated with air quality simulated to just
meet the current annual average NC>2 standard, the number of persons experiencing
concentrations at or above the potential health effect benchmarks is increased in comparison with
as is air quality. Figure 8-21 illustrates the percent of asthmatics estimated to experience at least
one exposure at or above the selected potential health effect benchmark concentrations, with air
quality adjusted to just meet the current standard. The exposure results for both including and
excluding indoor source contributions are presented. While it was estimated that about 92, 76,
207
-------
and 59% of asthmatics would be exposed to 200, 250, and 300 ppb (1-hour average) at least once
in a year for the as is air quality, it was estimated that nearly all asthmatics would experience at
least one exposure above any of the potential health effect benchmark levels in a year when air
quality is adjusted to just meet the current standard. Exposure estimates where indoor sources
were included were not greatly different than the results without indoor source contributions,
with nearly all asthmatics estimated to have at least one exposure at or above even the highest
potential health effect benchmark level.
For air quality simulated to just meet the current standard, repeat exposures at the
selected potential health effect benchmarks are more frequent than that estimated for the
modeled as is air quality. Figure 8-22 illustrates this using the simulated asthmatic population
for year 2002 data as an example. Nearly all asthmatics (>97%) were estimated to be exposed at
or above any one of the selected levels for at least six times in a year. Results for asthmatics
when exposures were estimated considering the contribution from indoor sources were similar,
only slightly higher (data not shown).
_ £ ra
2.5E+5-T
2.0E+5-1
.* 7i 1.5E+5-
ai
o _i
"g
1.0E+5
« ai 5.0E+4
< OT
O.OE+0
100
150
200
2002 AQ (cur std) with indoor
2002 AQ (cur std) no indoor
250
300
Potential Health Effect Benchmark Level (ppb)
Figure 8-21. Estimated number of all asthmatics in the Atlanta modeling domain with at
least one NOi exposure at or above the potential health effect benchmark level, using
modeled 2002 air quality just meeting the current standard (cur std), with and without
modeled indoor sources.
208
-------
100
Estimated Number
of Repeated
Exposures in a
Year
150
200
250
300
Potential Health Effect Benchmark Level
(PPb)
Figure 8-22. Estimated percent of asthmatics in the Atlanta modeling domain with
repeated NOi exposures above health effect benchmark levels, using modeled 2002 air
quality just meeting the current standard, without modeled indoor sources.
8.9.4.4 Estimated Number of 1-hour Exposures Above Selected Levels (alternative
standards)
To simulate just meeting the alternative NC>2 standards, the potential health effect
benchmark level was adjusted in the exposure model, rather than adjusting all of the hourly
concentrations for each receptor and year simulated (see section 8-8). Similar to exposure
analyses performed with the as is air quality, estimates of short-term exposures (i.e., 1-hour daily
maximum) were generated for the total population and population subgroups of interest (i.e.,
asthmatics and asthmatic children). Due to limitations on the number of concentration levels
allowed in an APEX simulation, only the potential health effect benchmark levels of 100, 200,
300 ppb were evaluated for the alternative 1-hour daily maximum standards.
In considering exposures estimated to occur associated with air quality simulated to just
meet the alternative NO2 1-hour daily maximum standards, the number of persons experiencing
concentrations at or above the potential health effect benchmarks varied, depending on the form
and level of the standard. Figure 8-23 illustrates the different forms (a 98th percentile or p98;
99th percentile or p99) at various 1-hour concentration levels of the standard. The number of
209
-------
persons exposed at least once at each of the 98th and 99th percentiles alternative standards and
considering a potential benchmark level of 100 ppb is similar to that observed for the as is air
quality and for air quality adjusted to just meet the current standard. That is, most persons are
exposed at least once to 100 ppb in a year, regardless of the standard form and level chosen. It is
not until the level of the 1-hour daily maximum standard approaches 50 ppb for either of the
percentile forms that the number of persons exposed to the higher benchmark levels is
substantially reduced. For example, while nearly all asthmatics are exposed to 100 ppb at least
once in a year as was observed in the above analyses, the percent of asthmatics exposed to at
least one 1-hour concentration at or above the 200 or 300 ppb is reduced to about 49% and 14%
of the subpopulation, respectively, when considering the 50 ppb, 98th percentile standard.
The estimated number of repeated NO2 exposures above selected levels can be sharply
reduced for potential alternative standards at the lower end of the range of alternative standards
considered. As an example, Figure 8-24 illustrates the number of multiple exposures above the
potential health effect benchmark levels using a 50 ppb, 99th percentile alternative standard. This
is the first instance where multiple exposures of the 100 ppb benchmark are estimated to be
reduced, with about 57% of asthmatics estimated to have greater than six in a year. A greater
reduction in the number of multiple exposures is observed when considering the 200 ppb
benchmark level. For example, only 5% of asthmatics are estimated to be exposed four or more
times, compared with 58% using the 2002 air quality as is.
The effect of indoor source contributions to the exposures was also evaluated for the
same level and form of alternative standard (50 ppb, 99th percentile). Figure 8-25 illustrates what
has been consistently shown in the above analyses, the indoor sources primarily affect the
numbers of persons and the number of times a person is exposed at or above 100 or 150 ppb,
with limited contribution to the higher potential health effect benchmark levels.
In addition, for comparison with the results presented in Figure 8-24, the percent of
asthmatics exposed to the selected health effect benchmark levels considering the 100 ppb, 99th
percentile alternative standard is presented in Figure 8-26. A greater proportion of asthmatics
have multiple exposures at all of the 1-hour benchmarks, nearly all of which were estimated to
have at least six exposures at or above a 1-hour concentration of 100 ppb.
210
-------
IS
1/1
o
Q.
2.5E+5
2.0E+5
•S -3; 1.5E+5
V)
•D O
*J
IS
to
UJ
5.0E+4
O.OE+0
100 200
200 300
Potential Health Effect
Benchmark Level (ppb)
2002 AQ (p98-150) no indoor
2002 AQ (p99-150) no indoor
2002 AQ (p98-100) no indoor
2002 AQ (p99-100) no indoor
2002 AQ (p98-50) no indoor
2002 AQ (p99-50) no indoor
Figure 8-23. Estimated percent of asthmatics in the Atlanta modeling domain with
exposures at or above potential health effect benchmark levels, using modeled 2002 air
quality adjusted to just meeting potential alternative standards, without indoor sources
^
1 =
xg
UJ a>
U) -I
II
c t3
*- 2
0 «
*;
-------
^
II
tl
"* o 5
ot;£
11
o c
0) 0
li
•x 2
as
Estimated Number
of Repeated
Exposures in a
Year
100
200
300
Potential Health Effect
Benchmark Level (ppb)
Figure 8-25. Estimated percent of asthmatics in the Atlanta modeling domain with
multiple NOi exposures at or above potential health effect benchmark levels, using
modeled 2002 air quality adjusted to just meeting a 50 ppb level 99th percentile form
alternative standard, with indoor sources.
100%
UJ £
v> -I
.Si 'C
I 8
§•-
li
2§
3 Estimated Number
4 of Repeated
Exposures in a
Year
100
200
Potential Health Effect
Benchmark Level (ppb)
Figure 8-26. Estimated percent of asthmatics in the Atlanta modeling domain with
multiple NOi exposures at or above potential health effect benchmark levels, using
modeled 2002 air quality adjusted to just meeting a 100 ppb level 99th percentile form
alternative standard, without indoor sources.
212
-------
8.10 KEY OBSERVATIONS
Presented below are key observations resulting from the exposure assessment:
• Modeled concentrations are reasonable given comparisons to available data
o AERMOD predicted upper percentile NO2 concentrations are about 10 to
50% (or 10 to 30 ppb) higher than ambient measurements at three fixed-
site monitors
o APEX modeled daily mean NO2 exposures in Atlanta (medians 3-24 ppb)
are comparable to personal exposure measurements in Atlanta (median 3-
14 ppb)
o APEX modeled annual average NO2 exposures, expressed as a percent of
the ambient NO2 concentration (40 to 70%), are consistent with findings
reported in the ISA (30 to 60%)
• Estimated exposures above 1-hour potential health effect benchmark levels using
APEX were due largely to roadway-related exposures (>99%). Of this,
approximately 70% were from in-vehicle exposures, with the remainder
associated with outdoor near-road exposures
• When included, indoor sources contribute to the occurrence of NC>2 exposures at
or above 100 ppb (61%), but little to the occurrence of higher exposures (i.e.,
above 200, 300 ppb)
• The estimated effect of air quality on benchmark exceedances differs by
benchmark level considered:
o 100 ppb
• For all air quality scenarios considered, more than 90% of
asthmatics in Atlanta are estimated to be exposed at least one time
per year
• Of the standard levels evaluated, 50 ppb was the only level
estimated to reduce repeat exposures to NO2 concentrations above
100 ppb compared to recent air quality levels
o 200 ppb
• Of all the air quality scenarios considered, only alternative
standards set at 50 ppb are estimated to reduce the percent of
213
-------
asthmatics exposed at least one time (by approximately 40 to 50%)
relative to recent air quality levels
o 300 ppb
• Of all the air quality scenarios considered, only alternative
standards set at 50 ppb or 100 ppb are estimated to reduce the
percent of asthmatics exposed at least one time (by approximately
80% and 15% respectively) relative to recent air quality levels
• When simulating air quality that just meets the current annual standard, virtually
all asthmatics in Atlanta are estimated to experience six or more daily maximum
1-hour exposures per year to NC>2 concentrations above the highest benchmark
level evaluated (300 ppb)
• Using a 98th versus a 99th percentile form made a difference of approximately 5 to
10% in the number of daily maximum benchmark exceedances, with the 99th
percentile form generating fewer exceedances.
8.11 REPRESENTATIVENESS OF EXPOSURE RESULTS
8.11.1 Introduction
Due to time and resource constraints the exposure assessment evaluating the current and
alternative standards was only applied to the Atlanta urban area. A natural question is how
representative are the estimates from this assessment of exposures in Atlanta to other urban areas
in the United States. To address this question, additional data were compiled and analyzed to
provide perspective on how representative the Atlanta exposure modeling results might be for
other urban areas. Because most estimated exceedances were associated with the near-road or
in-vehicle microenvironments, the analysis and discussion is centered on a variety of population
and road statistics to allow for comparison of Atlanta with several other urban locations.
8.11.2 Description of Data Compiled and Summarized
Three sources were used for comparing Atlanta with several urban locations: 1) the
Human Air Pollutants Exposure Model near-road population data base, 2) American Housing
Survey (AHS) data, and 3) statistics from the Federal Highway Administration (FHWA).
214
-------
8.11.2.1 HAPEM6 Near-Road Population Data Base
The first type of data considered employs a data base developed for the Human Air
Pollutants Exposure Model (HAPEM6) and is available as part of the HAPEM6 installation
package.19 The data base provides estimates of the fraction of people in each of 6 age groups in
each US Census tract living near major roadways.20 The distance-from-roadway bins are: 0 m to
75 m, 75 m to 200 m, and > 200 m. Details regarding the data base development are provided in
Appendix B, Attachment 6 and 7.
Briefly, the development of the HAPEM6 near-road population data base was for use in
estimating the enhancement near major roadways of air toxic pollutant concentrations from on-
road motor vehicle emissions relative to concentrations at other outdoor locations. First, several
measurement studies of near roadway concentration gradients were reviewed (Kwon, 2005;
Meng et al., 2004; Riediker et al., 2003; Rodes et al., 1998; Weisel et al., 2004; Zhu et al., 2002).
None of these studies provided sufficient data to estimate the required concentration ratios.
Measurements in the Riediker et al. (2003) and Rodes et al. (1998) studies were made at
distances shorter than those of interest for this study. Available ratios for Zhu et al. (2002) were
only for downwind distances, and did not represent ratios under more general meteorological
conditions. The ability of regression models applied to the Relationships of Indoor, Outdoor,
and Personal Air (RIOPA) study data (Kwon et al., 2005; Meng et al., 2004; and Weisel et al.,
2004) to predict the near roadway concentrations was generally poor, likely due to the problem
that the near roadway concentrations are also impacted by other emissions sources that cannot be
easily adjusted for.
CALPUFF air dispersion model predictions from the Portland (OR) Air Toxics
Assessment were analyzed (PATA; Cohen et al., 2005) using summary statistics and regression
modeling in order to obtain distributions of concentration ratios. A distribution was developed
based on the ratio of concentrations within a distance D} meters of a major roadway, to
concentrations at locations greater than a second distance, D2 meters from a major roadway. A
second distribution was developed using the ratio of concentrations at DI - D2 meters of a major
roadway to concentrations at locations greater than D2 meters from a major roadway.
19
http://www.epa.gov/ttn/fera/human hapenxhtml.
20 Ages 0-1, 2-4, 5-15, 16-17, 18-64, and >65.
215
-------
To determine the best choices for the distance-to-roadway bins, regression models were
developed to estimate two distance terms D} and D2 for each season and for the annual mean.
Generally, the two-distance term regression models applied to the PATA CALPUFF data
showed that the optimal road distances, based on R2 and Akaike Information Criterion (AIC)
statistics, were DI = 75 m and D2 = 300 m. The regression models favored the higher distances
for D2 because the CALPUFF estimates for PATA continue to decrease significantly with
distance from the road. However, the R2 and AIC values were not much different between the
models with D2 = 200 or 300 m. Because a few other studies, including Zhu et al. (2002), have
shown typical zones of influence for roadways no further than 200 m, distances of DI = 75 m and
D2 = 200 m were selected for development of the population data base.
Then an analysis was conducted using the US Census block level populations stratified
by age. The block level data were then aggregated up to the tract level for populations stratified
by age for use in HAPEM6. The data bases used were:
• The Environmental Sciences Research Center (ESRI) StreetMap US roadway
geographic database (which includes NavTech, GDT and TeleAtlas rectified
street data)
• A geographic database of US Census block boundaries, extracted using the
PCensus 2000 Census data extraction tool for Census file SF1
• A geographic database for US Census block boundaries in Puerto Rico and the
US Virgin Islands obtained from Proximity
Because populations are not generally evenly distributed within blocks, it was assumed
that the block populations all reside within 150 m of at least one road within the block of
designation "local" or greater as defined by the Census Feature Class Codes (CFCC). Thus, the
first step was to create a 150 m buffer around all roadways within the block. This buffer served
as a "clipped" block boundary defining the portion of the block containing residential
populations. The block population was assumed to be uniformly distributed within the "clipped"
block boundary. Next, a 75 m buffer and a 200 m buffer were created around all major roadways
within the block. These buffers were overlaid on the "clipped" block boundary, and the fraction
of the "clipped" block area that fell within each buffer was calculated. This area fraction was
assumed to equal the population fraction that fell within each buffer, and the fractions were
applied to each population subgroup. The 75 m buffer and the 200 m buffer were also overlaid
216
-------
on the undipped block boundary to determine the fraction of the total block area that fell within
each of the buffers. The block level fractions for area and populations were then aggregated up
to the tract level.
For each tract in the CMSA or MSA, the tract-level, age-stratified distance-to-roadway
population fractions were combined with 2000 US Census tract-level, age-stratified residential
populations to calculate the CMSA- or MSA-level distance-to-roadway population fractions, as
follows.
CMSA fraction = ^tracts (tract-fraction*tract-population)/ ^tracts (tract-population)
Results of the estimated population within the selected roadway distances are provided in
Table 8-14 for each of the 18 named locations identified in the air quality characterization.
Based on the HAPEM6 data base, Atlanta has the lowest percent of its population living within
each of the near-roadway distance bins. Absolute differences for the closest road distance bin
ranged from 4 to 16 percentage points lower for Atlanta compared with all other locations.
Atlanta also was estimated to have the greatest percent of the population living at distances > 200
m from a major road. Consideration of just this attribute would suggest that on a population-
weighted basis, the number of daily maximum exposures of NC>2 concentrations above
benchmark levels may be lower for Atlanta than many of the other urban areas listed in Table 8-
14. However, note that most of the exceedances were associated with in-vehicle exposures.
Table 8-14. Percent of population within selected distances of a major road in several locations.
Location
Atlanta
Boston
Chicago
Cleveland
Colorado Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Percent of Population at Distance from a Major Road1
<75m
16%
27%
27%
20%
21%
23%
22%
24%
21%
22%
26%
26%
32%
>75 m, <200 m
23%
38%
38%
36%
32%
34%
33%
35%
29%
33%
38%
37%
37%
>200m
62%
35%
35%
44%
47%
43%
45%
41%
49%
45%
37%
37%
30%
217
-------
Location
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Percent of Population at Distance from a Major Road1
<75m
25%
22%
19%
20%
25%
>75 m, <200 m
34%
34%
31%
31%
27%
>200m
42%
44%
51%
49%
48%
Notes:
1 Major roadways were all roadways except those classified as local by the US
Census Bureau.
8.11.2.2 American Housing Survey (AHS) Data
The American Housing Survey (AHS), conducted by the Bureau of the Census for the
Department of Housing and Urban Development (HUD), collects data on the nation's housing
(AHS, 2008). Relevant housing characteristic data, including housing units within 300 m of a
major highway, railroad, or airport and residential prevalence of air conditioning are summarized
for 14 of the named locations21 evaluated in the air quality characterization, using the available
metropolitan areas surveyed by the AHS (Table 8-15). Because survey years differ for each
location and some locations contained more than one survey, the most recent data or data closest
to 2002 were selected (the base year for the exposure modeling). Consistent with the pattern
noted for the population living near roadways, Atlanta also has the lowest percent of housing
units (9.7%) within 300 m of a 4 or more lane highway, railroad, or airport (AHS, 2008).
Denver, Phoenix, and St. Louis were only slightly higher though, estimated to have 10.1, 11.2,
and 11.4% of housing units, respectively < 300 m from these same locations. Again,
consideration of this attribute alone, suggests daily maximum exposures toNO2 concentrations
above selected benchmarks would tend to be lower for Atlanta than many of the other urban
areas listed in Table 8-15. As noted previously, the estimated exceedances were dominated by
exposure occurring in-vehicles.
The AHS also provides data on A/C prevalence rate for several urban areas (Table 8-
15). The A/C prevalence can vary greatly across urban areas, based largely on climate
differences. The air conditioning prevalence can influence the air exchange rate in a residence,
potentially affecting the infiltration rate of outdoor air concentrations into the indoors residential
microenvironment. Atlanta was estimated to have one of the highest air conditioning prevalence
21 There are no AHS data for Colorado Springs, El Paso, Jacksonville, and Las Vegas.
218
-------
rates (97.2%), though similar rates could be found in Miami, Phoenix, St. Louis, and Washington
D.C. A few of the urban areas listed have much lower A/C prevalence rates, including Los
Angeles with 57.4% and Boston with 63.1%. For locations having a low A/C prevalence, it is
expected that the number of indoor residential exposures to daily maximum NC>2 concentrations
above selected benchmarks would be greater compared to those estimated in Atlanta. However,
results of this exposure assessment indicate the indoor residential microenvironment does not
contribute to exceedances of the selected benchmarks (see Figure 8-17) even when considering
alternative A/C prevalence (section 8.12.2.6).
Table 8-15. Residential A/C prevalence and roadway distance statistics for housing units in several
locations (AHS, 2008).
Location
Atlanta
Boston
Chicago
Cleveland
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
AHS
Survey
Year
2004
1998
2003
2004
2004
2003
ND
ND
ND
2003
2002
2003
2003
2002
ND
2004
1998
Housing units
< 300 m from a
mobile source1
(%)
9.7
14.3
18.8
19.0
10.1
17.7
ND
ND
ND
14.2
15.8
14.8
14.3
11.2
ND
11.4
17.6
A/C
Use2
(%)
97.2
63.1
89.6
75.8
66.9
82.4
ND
ND
ND
57.4
98.1
83.3
91.4
94.4
ND
96.7
96.0
Notes:
ND No data available
1 Represents the percent of total year-round housing units
located within 300 meters of a 4 or more lane highway,
railroad, or airport (AHS, 2008).
2 Represents the percent of total year-round housing units
having central or room unit air conditioners (AHS, 2008).
219
-------
8.11.2.3 Federal Highway Administration (FHWA) Data
The Federal Highway Administration (FHWA) produces annual highway statistics
reports based on highway data submitted by individual States. Because 2002 served as the base
year for the exposure assessment, these data were compiled for each of the named locations
evaluated in the air quality characterization and are presented in Table 8-16. We note that
Atlanta contains the highest per capita daily vehicle miles traveled and also the highest miles of
roadway per 1,000 persons, both attributes which would tend to result in greater population
exposure to peak NC>2 concentrations associated with roadway exposures. However, as shown in
Figure 8-27, Atlanta is roughly in the middle of the distribution with respect to estimated
population and total roadway miles when compared with the other locations examined.
Table 8-16. Population and roadway statistics for several locations (FHWA, 2002).
Location
Atlanta
Boston
Chicago
Cleveland
Colorado Springs
Denver
Detroit
El Paso
Jacksonville
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
Provo
St. Louis
Washington DC
Total
roadway
miles
13,438
13,809
23,832
6,975
1,887
7,261
13,755
2,225
4,769
3,206
26,329
15,436
37,854
15,743
10,684
1,384
9,123
10,561
Estimated
Population
(thousands)
2,873
3,854
7,702
1,785
439
1,989
3,835
657
906
1,456
12,365
5,021
17,307
4,813
2,949
345
2,067
3,807
Miles of
roadway
per 1,000
persons
4.7
3.6
3.1
3.9
4.3
3.7
3.6
3.4
5.3
2.2
2.1
3.1
2.2
3.3
3.6
4.0
4.4
2.8
Total
DVMT1
per
capita
35.3
20.9
21.5
20.6
20.4
22.9
25.1
17.3
31.0
18.1
23.7
23.9
15.9
19.4
21.2
21.2
29.2
22.7
220
-------
1 O,UUU
16,000
"5/T
c 14,000
ra
(A
3
0 12,000
[3,
0 10,000
1
§. 8,000
O
* 6,000
•4-1
ra
J 4,000
(A
LiJ
2,000
n
o
d* °
& ^
o o
o o 4
2
concentrations above selected benchmark levels than other locations included in Tables 8-14 and
8-15, in the absence of other influential attributes. However, as noted above, Atlanta contains
221
-------
the highest per capita daily vehicle miles traveled and also the highest miles of roadway per
1,000 persons, both attributes which would tend to result in greater population exposure to NC>2
concentrations above selected benchmark levels associated with increased roadway exposures.
While the miles of roads per person is generally accounted for by the current exposure modeling,
the DVMT specific to Atlanta is not well represented and is likely more similar to a national
DVMT.
Furthermore, Atlanta is roughly in the middle of the distribution with respect to estimated
population and total roadway miles when compared with the other locations examined. Given
that there are attributes that go in both directions with respect to the influence of on-road and
other mobile sources on the representativeness of Atlanta relative to the other 17 urban areas
examined in the air quality analysis, staff judges that the Atlanta exposure estimates are likely
representative of other moderate to large urban areas included in this comparison. Staff does
recognize that the Atlanta exposure results are likely lower on a population-weighted basis
compared to the largest urban areas such as Los Angeles, New York, and Chicago given the
greater proximity of the population to mobile sources in these large urban areas. For example,
64, 69, and 65%, respectively, of Los Angeles, New York, and Chicago's populations live within
200 m of a major road compared to only 39% for Atlanta. Similarly, there is a much higher
percentage of housing units within 300 m of a mobile source (i.e., 14.2, 14.8, and 18.8%,
respectively, for Los Angeles, New York, and Chicago), compared to only 9.7% for Atlanta.
As discussed above, with respect to residential A/C prevalence, we expect that urban
areas with lower A/C prevalence would tend to result in higher exposures to NC>2 of ambient
origin, all other factors being equal. We note that in comparing A/C prevalence rates for each
location, Boston, Cleveland, Denver, and Los Angeles had between 20 to 40 percentage points
less prevalence of A/C in residences than that observed for Atlanta. Thus, based on this
consideration, the estimated percent of population exposed and person days with exposures
above the selected health effect benchmarks for Atlanta may be somewhat lower than would be
expected for other urban areas within the parts of the U.S. (e.g., urban areas in the midwest,
northeast, southern California, and northwest) with lower prevalence of residential A/C.
222
-------
8.12 UNCERTAINTY ANALYSIS
The methods and the model used in this exposure assessment conform to the most
contemporary modeling methodologies available. APEX is a powerful and flexible model that
allows for the realistic estimation of air pollutant exposure to individuals. Since it is based on
human time-location-activity diaries and accounts for the most important variables known to
affect exposure (where people are located and what they are doing), it has the ability to
effectively approximate actual exposure conditions. In addition, the input data selected were the
best available data to generate the exposure results. However, there are constraints and
uncertainties with the modeling approaches and the input data that limit the realism and accuracy
of the modeling results.
Uncertainties and assumptions associated with NC>2 specific model inputs, their
utilization, and application are discussed in the following sections. Analyses for certain
components of APEX performed previously in other NAAQS reviews (see EPA, 2007g;
Langstaff, 2007) that are relevant to the NC>2 NAAQS review are only summarized below. This
includes a sensitivity analyses performed on the CHAD data base using 63 exposures and an
analysis of the air exchange rate data.
Following the same general approach described in section 7.8 and adapted from WHO
(2008), a qualitative analysis of the components contributing to uncertainty in the exposure
results was performed. This includes an identification of the important uncertainties, an
indication of the potential bias direction, and a scaling of the uncertainty using low, medium, and
high categories. Even though uncertainties in AERMOD concentrations predictions are an
APEX input uncertainty, they are addressed separately here for clarity.
Table 8-17. Summary of qualitative uncertainty analysis for the ex
Source
AERMOD
Inputs and
Algorithms
APEX
Inputs and
Type
AERMOD formulations for
mobile sources
On-road emissions
O3 monitoring data
Use of unadjusted NO2
concentrations
Meteorological data
Population data base
Commuting data base
Concentration/
Exceedance
Bias Direction
unknown
over
over
unknown
unknown
both
both
posure assessment.
Characterization
of Uncertainty
Low
Low- Medium
Low
Low- Medium
Low- Medium
Low
Low - Medium
223
-------
Algorithms
CHAD data base
Meteorological data
Air exchange rates
A/C prevalence
Indoor sources not modeled
Indoor decay distribution
Indoor concentration
distribution
Longitudinal profile
under
both
unknown
none
under
under
under
both
Low - High
Low
Medium
Low
Medium
Low - Medium
Medium - High
Low
8.12.1 Dispersion Modeling Uncertainties
Air quality data used in the exposure modeling was determined through use of EPA's
recommended regulatory air dispersion model, AERMOD (version 07026 (EPA, 2004)), with
meteorological data discussed above and emissions data based on the EPA's National Emissions
Inventory for 2002 (EPA, 2007e) and the CAMD Emissions Database (EPA, 2007f) for
stationary sources and mobile sources determined from local travel demand modeling and EPA's
MOBILE6.2 emission factor model. All of these are high quality data sources. Parameterization
of meteorology and emissions in the model were made in as accurate a manner as possible to
ensure best representation of air quality for exposure modeling. Thus, the resulting air quality
values are likely free of systematic errors to the best approximation available through application
of modeled data.
An analysis of uncertainty associated with application of a model is generally broken
down into two main categories of uncertainty: 1) model algorithms, and 2) model inputs. While
it is convenient to discuss uncertainties in this context, it is also important to recognize that there
is some interdependence between the two in the sense that an increase in the complexity of
model algorithms may entail an increase in the potential uncertainty associated with model
inputs.
8.12.1.1 AERMOD Algorithms
The AERMOD model was promulgated by in 2006 as a "refined" dispersion model for
near-field applications (with plume transport distances nominally up to 50 kilometers), based on
a demonstration that the model produces largely unbiased estimates of ambient concentrations
across a range of source characteristics, as well as a wide range of meteorological conditions and
topographic settings (Perry, etal., 2005; EPA, 2003). While a majority of the 17 field study
224
-------
databases used in evaluating the performance of AERMOD are associated with elevated plumes
from stationary sources (typically power plants), a number of evaluations included low-level
releases more typical of the dominant emissions category in this assessment. Moreover, the
range of dispersion conditions represented by these evaluation studies provides some confidence
that the fundamental dispersion formulations within the model will provide robust performance
in other settings.
AERMOD is a steady-state, straight-line plume model, which implies limitations on the
model's ability to simulate certain aspects of plume dispersion. For example, AERMOD treats
each hour of simulation as independent, with no memory of plume impacts from one hour to the
next. As a result, AERMOD may not adequately treat dispersion under conditions of
atmospheric stagnation or recirculation when emissions may build up within a region over
several hours. This could lead to a bias toward under-prediction by AERMOD during such
periods. On the other hand, AERMOD assumes that each plume may impact the entire domain
for each hour, regardless of whether the actual transport time for a particular source-receptor
combination exceeds an hour. While these assumptions imply some degree of physically
unrealistic behavior when considering the impacts of an individual plume simulation, their
importance in terms of overall uncertainty will vary depending upon the application. The degree
of uncertainty attributable to these basic model assumptions is likely to be more significant for
individual plume simulations than for a cumulative analysis based on a large inventory. This
question deserves further investigation to better define the limits and capabilities of a modeling
system such as AERMOD for large scale exposure assessments such as this. The evidence
provided by the model-to-monitor comparisons presented in Chapter 8.4.8 is encouraging as to
the viability of the approach in this application when adequate meteorological and other inputs
are available. However, each modeling domain and inventory will present its own challenges
and will require a separate assessment based on the specifics of the application.
Beyond the basic dispersion algorithms in AERMOD, another component of model
formulation uncertainty for this application is the use of the Ozone Limiting Method (OLM) and
Plume Volume Molar Ratio Method (PVMRM) algorithms for simulating the conversion of NO
to NO2. Model performance evaluations for these NOX chemistry algorithms are more limited
than for non-reactive plume dispersion. However, an assessment of the potential for bias
225
-------
associated with the PVMRM algorithm showed generally good performance based on the
available evaluation data (MACTEC, 2005).
While the AERMOD model algorithms are not considered to be a significant source of
uncertainty for this assessment, the representativeness of modeled concentrations for any
application are strongly dependent on the quality and representativeness of the model inputs.
The main categories of model inputs that may contribute to bias and uncertainty are emission
estimates and meteorological data. Use of the OLM and PVMRM chemistry algorithms also
introduces the potential uncertainty associated with the dependence on ambient 63 concentration
inputs since both algorithms treat the reaction of NO with 63 to produce NC>2 as the primary
mechanism for conversion of NOX emissions. These issues are addressed in the following
sections.
8.12.1.2 Meteorological Inputs
Details regarding the representativeness of the meteorological data inputs for AERMOD
are addressed separately in Attachment 1 to Appendix B-4 of the REA. One of the main issues
associated with the representativeness for this application is the sensitivity of the AERMOD
model to the surface roughness of the meteorological tower site used to process the
meteorological data for use in AERMOD relative to the surface roughness across the full domain
of sources. This issue has been shown to be more significant for low-level sources, including
mobile sources, due to the importance of mechanical shear-stress induced turbulence on
dispersion for such sources. In particular, concerns were raised in preliminary modeling due the
typically low surface roughness associated with the meteorological tower located at ATL airport
compared with the much higher roughness environment of most low-level emission sources.
Based on this concern, alternative meteorological data from the Jefferson Street (1ST) Southeast
Aerosol Research and Characterization study (SEARCH) site were determined to be more
representative of the majority of NOX emissions across the modeling domain, as discussed in
Attachment 1 to Appendix B-4. The meteorological data obtained for the 1ST site were used to
model all emissions with the exception of the Atlanta airport. Meteorological data for ATL were
used to model the airport emissions.
To assess the uncertainty associated with the sensitivity of AERMOD to surface
roughness effects, a comparison was made between modeled NO2 concentrations from mobile
226
-------
source emissions based on the 1ST meteorological data versus the ATL meteorological data.
Surface roughness lengths were generally about an order of magnitude higher at the 1ST site
relative to the ATL site. Results of this comparison are presented in Figure 8-28, which shows a
plot of high ranked 1-hour concentrations, unpaired in time and space, calculated at four Atlanta
NC>2 monitoring locations for 2002. This figure shows relatively good agreement in modeled
concentrations based on the two sets of meteorological inputs, at least for the peak of the
concentration distribution at these four receptor locations. This suggests that the sensitivity of
AERMOD model results to variations in surface roughness may be less significant than
commonly believed, provided that meteorological data inputs are processed with surface
characteristics appropriate for the meteorological site. The overall peak concentration should
about a factor of 2 bias between the two sets of meteorological inputs, with the concentrations
based on the higher roughness 1ST data showing higher impacts. While these results are
encouraging in relation to the assessment of uncertainty for the REA, they may not be indicative
of the degree of sensitivity to surface roughness for individual sources in other modeling
contexts.
227
-------
10000
1000
100
1000
Figure 8-28. Comparison of high ranked AERMOD 1-hour NOi concentrations (ug/ni )
from mobile sources across four NOi monitoring locations based on JST vs. ATL
meteorological inputs for 2002.
8.12.1.3 Mobile Source Characterization
Given the predominance of mobile source emissions in this assessment, source
characterization of non-stationary mobile sources is another important category to consider for
the uncertainty analysis. Determining the most appropriate source characteristics for modeling
emissions from mobile sources presents several technical challenges and involves a number of
uncertainties. Unlike typical stationary emission sources simulated by AERMOD, for which
source characteristics such as release height and effluent parameters can be clearly defined and
measured, emissions from mobile sources represent an aggregate of emissions from non-
stationary sources of various sizes, shapes, and speeds. Since mobile source emissions (other
than aircraft) are emitted near the ground, the plumes can be significantly influenced by the
turbulent wake associated with the emitting vehicle, as well as turbulence generated by nearby
vehicles and other roughness elements such as sound barriers, median barriers, trees, buildings,
etc. Some of these effects may vary depending upon the orientation of the roadway to the wind
direction since the vehicle wake effect is most pronounced for wind directions parallel to the
228
-------
road, where the wakes of multiple vehicles can combine to enhance dispersion of the roadway
emissions.
Representative source characteristics for mobile emissions may also depend on the
pollutant of interest, and whether the pollutant is primarily associated with direct vehicle
exhaust, as in the case of NC>2, or includes mechanical sources, such as re-entrained road dust,
tire wear and brake wear for PM. Emissions associated with vehicle exhaust may experience
some buoyancy due to the exhaust temperature exceeding the ambient temperature. Such
buoyancy effects might be negligible for vehicles moving at highway speeds, where
mechanically-induced turbulence would likely dominate the initial plume characteristics, but
could be a significant factor for slow moving vehicles on a cold day during rush hour. In
addition to the influence of roughness elements, characteristics of the roadway itself may be a
factor. For example, thermally-induced turbulence generated by direct sunlight on dark asphalt
could enhance the initial plume spread compared to the same conditions for a more reflective
concrete road surface. The best approach for determining source characteristics for mobile
emissions may also depend on the scope and nature of the application. Characterizing mobile
sources for a large-scale urban study, such as the Atlanta NC>2 modeling, may necessitate a
different approach than characterizing mobile sources for a particular highway project within a
more localized modeling domain. A detailed treatment accounting for influences of specific
local features may be possible for the latter. However, for larger scale applications such as this,
a simplified approach with the goal of characterizing emissions based on a reasonable estimate of
the aggregate effect of these various factors is necessitated by practical limitations.
The factor of 1.7 times the vehicle height used to account for vehicle-induced turbulence
is cited in Gilles et al. (2005) based on some field measurements for an unpaved road. The factor
of 1.7 is somewhat less than the typical formula for the turbulent wake downwind of a building,
which is 2.5 times the building height. This difference seems reasonable based on the more
aerodynamic shape of vehicles as compared to buildings. While some differences may be
expected between paved and unpaved roads, these differences are probably minor compared to
other uncertainties.
However, the value used to estimate vehicle-induced turbulence could be conservative in
terms of modeled concentrations since it may not account for the other influences mentioned
above, especially the influence of roadway orientation relative to the wind direction. These
229
-------
influences may include considerable variability both temporally and spatially, and this variability
is difficult to quantify within the current model formulations. However, given the size of the
domain, the number of sources included in the modeled inventory, and the length of period
modeled, it is likely that some of the uncertainties associated with source characterization will
tend to cancel out in the cumulative impact assessment. This expectation is supported for off-
road exposures by the generally good agreement in model-to-monitor comparisons presented in
Chapter 8.4.8, but is more difficult to assess in relation to on-road exposures.
The matching of TDM links, which are straight segments by their nature, to roads, which
may be curvilinear, was done as objectively as possible. It is possible that once links were
segmented to avoid aspect ratio issues in the modeling, these segments could deviate from the
actual road layout more than they did prior to segmenting. However, the GIS analysis and
matching was done before the segmenting and no revision was done after. Thus staff cannot
quantify any additional spatial mismatch that came from segmenting. Spatial mismatch from
segmenting could affect offroad concentrations relative to a different methodology where
predicted concentrations from segments were mated to road locations rather than full TDM links
due to both orientation of the areas and location relative to receptors. However, these segments
were not considered individual links, but rather partial values to avoid numerical modeling issues
and were maintained as consistent with the original data source from which they were drawn
(i.e., the TDM), which did not have data at the resolution of the individual segments. On-road
concentrations using an alternative methodology, where segments rather than links were mated
to actual road locations, would differ due to orientation of the area relative to wind direction.
However, as the onroad receptors are centered within the source, any differences are likely to be
small.
8.12.1.4 On-Road Emissions Estimates
It should be noted that free flow speed represents all traffic, that is, cars and trucks on
each link are assigned the same speed. This adds to uncertainty due to the prominence of trucks
in the NOX inventories, and may contribute positive bias to diurnal patterns in AERMOD
predictions.
The truck fractions used in the traffic modeling may also contribute to bias and
uncertainty in the exposure results, although adjustments made based on the NEI on-road
230
-------
emissions may correct for the potential bias. A comparison with National estimates of vehicle
miles traveled for year 2002 (FHWA, 2002) and a study conducted in Georgia (FHWA, 2007)
indicate that the average truck fractions produced by the Atlanta Regional TDM are potentially
higher, possibly contributing to overestimations in NOX emissions. Table 8-18 summarizes
National estimates of the vehicle mile traveled (VMT) by two highway categories and vehicle
types. Depending on roadway category, rural VMT consisted of about 7 to 20% trucks and
urban road miles ranged from 4 to 8% trucks. Comparable results can be found in data reported
by FHWA (2007) for a roadway in Atlanta, where urban traffic counts were estimated to be
comprised of 3 to 8% trucks and rural traffic counts ranged from 10 to 20% trucks (Table 8-19).
The rural roads were estimated to have from 15 to 33% of VMT as trucks based on the
Atlanta TDM modeling, while urban roads ranged from 12 to 20% (Table 8-2). The discrepancy
is likely due to an inconsistency between the definitions of trucks between the present and
FHWA documents. The truck fraction used in the exposure assessment is based directly on the
total truck volume included in the TDM, defined as:
Total Trucks = Heavy (class 8-13) + Medium (class 4-7) + Commercial (any vehicle
used for commercial purposes)
By definition this truck fraction contains a larger fraction of vehicles than those from the
DEIS. To maintain internal consistency in our approach, we used the Atlanta-provided
MOBILE6 input files and fleet characterizations and computed composite emissions for light and
heavy duty vehicles, where the heavy duty fraction is a VMT-weighted composite of all medium
and heavy duty vehicles from MOBILE classes 2B to class 8B. It is possible that bias could
result from the addition of commercial vehicles in the TDM, but this was not resolvable with the
data provided by ARC. However, even with a mismatched HDV fraction and emission factor,
for example by coupling MDV+HDV VMT to an HDV8B emission factor, this would have been
corrected by the scaling of the emissions down to the 2002 NEI on-road emissions to account for
differences between 2005 and 2001-03 activity (see Table 8-6). In addition, the reported truck
fractions in Table 8-18 are likely associated with peak counts for light duty vehicles during work
commutes. Therefore these may be biased low, even when comparing with morning and
afternoon HDV fractions in Table 8-2 which are based on broader time periods.
231
-------
Table 8-18. National vehicle miles traveled by roadway category and vehicle type.
Area
Rural
Urban1
Roadway
Interstate
Arterial
Other
Interstate
Other
Millions of Vehicle-Miles Traveled
Buses
941
1,104
1,901
802
2,101
Single Unit
Trucks
8,745
14,606
14,963
9,106
28,467
Combination
Trucks
45,633
27,818
14,090
23,887
27,215
All
Vehicles
279,962
433,805
414,393
408,618
1,318,978
Trucks +
Buses
20
10
7
8
4
Notes:
1 Urban consists of travel on all roads and streets in urban places with 5,000 or greater
population.
Data from table vm1 .xls of FHWA (2002).
Table 8-19. Observed peak hour truck percentages on Interstate 75 (1-75) using 2002 traffic count
data.
Roadway1
I-75 N of I-
285
I-75 N of
Wade
Green
Road
Peak
Hour2
AM
PM
AM
PM
Direction
Northbound
Southbound
Northbound
Southbound
Northbound
Southbound
Northbound
Southbound
FHWA Vehicle Classes3
6-7
(MDT)
1
0
1
1
1
1
1
1
8-13
(HOT)
7
3
5
6
19
11
10
14
Total
Trucks4
8
3
6
6
20
12
10
14
Notes:
1 I-75 N of I-285 is just east of Fulton county, likely represents an urban area. The
roadway I-75 N of Wade Green Road in northern Cobb county likely represents a
rural area.
2 Peak hours for AM were between 7:30-9:15, PM peak hours were 4:45-5:14.
3 MDT and HOT are medium and heavy duty trucks, respectively.
4 Totals were summed before rounding, therefore MDT +HDT may not always
sum to total trucks
Data are from Table 2-4 of FHWA (2007).
Another important source of uncertainty is the diurnal pattern of on-road vehicle activity
and emissions. Vehicle activity data was provided by the Atlanta Regional Council (ARC)
showing average daily traffic for 4 daily time periods on each roadway link (see section 8.4.3.1).
However, the air dispersion model requires hourly allocations of the data. Assigning equal
vehicle activity levels to each hour in an ARC-specified time period would have created an
unrealistic "step function" time series. In order to create a more realistic diurnal pattern while
minimizing alteration of the provided data, a 5-hour running averages of vehicle activity were
used to smooth the "step function". There is uncertainty about how well the smoothed diurnal
232
-------
pattern represents the unknown actual hourly diurnal pattern. This uncertainty is important,
because typical meteorological conditions at various times of day have differential effects on the
relationship between emissions and concentrations. For example, at dawn and dusk typical low
mixing heights and low wind speeds enhance the concentration impacts of emissions, so that any
emission overestimates at those times are likely to result in disproportionate concentration
overestimates. The model-to-monitor comparisons in Figure 8-7 show disproportionate
overestimates of concentrations during those hours of the day, suggesting that corresponding
vehicle activity levels may be overestimated.
Another temporal component of the emission profile for mobile sources worth noting is
the variability by day-of-week. The temporal profile of mobile source emissions used for the
REA accounts for variability by season and hour-of-day, but does not distinguish between
weekday and weekend emission profiles. An analysis of model-to-monitor comparisons for
weekday vs. weekend periods was conducted to assess this potential source of uncertainty.
Figure 8-29 compares the average ratios of predicted to observed NC>2 concentrations at selected
percentiles across the concentration distribution based on the four NC>2 monitors (including the
SEARCH 1ST monitor). While this analysis showed that model performance was significantly
better for weekdays than weekends, the affect of this simplification on the overall concentration
distributions was not significant. The absence of the morning rush hour peak in NC>2
concentrations evident in the ambient concentrations for the weekend comparison suggests that
this factor may have contributed somewhat to the conservative bias in modeled concentrations
for the morning rush hour period, but was not the major factor.
233
-------
- Pred/Obs Weekday
- Pred/Obs Weekend
20
30
40 50 60
Percentile
70
80
90
100
Figure 8-29. Comparison of the average ratios of predicted/observed concentrations of
NOi across four ambient monitors based on weekday vs. weekend only.
Another potential source of uncertainty in relation to on-road emission estimates is the
assumption of 0.1 as the "in-stack" ratio of NC>2 to NOX emissions. While this ratio is considered
to be generally representative of many sources of NOX emissions and is considered a default
value for applications of the OLM option, modeled concentrations of NC>2 will be sensitive to
this user-specified value, especially for modeled impacts close to low-level sources. As a result,
the uncertainty associated with this issue is likely to be greater for on-road concentration
estimates than for off-road locations. However, the overall uncertainty associated with this
modeling input is expected to be low.
8.12.1.5 O3 Monitoring Data for OLM and PVMRM Options
Monitored hourly Os concentrations were used as input to the OLM and PVMRM
atmospheric chemistry modules in AERMOD for conversion of NO to NO2. Since AERMOD is
currently limited to the use of a single hourly Oj concentration for all sources within the modeled
inventory, this may introduce some uncertainty in the modeled concentrations if there is
significant spatial variability of Os concentrations across the domain. One suggested source of
234
-------
this uncertainty and potential bias is the influence of an 63 deficit over major roadways due to
the high concentration of NOX emissions depleting 63, especially during rush hour. Lacking
more complete spatial coverage of O3 concentrations, it is difficult to quantify this potential
cause of bias and uncertainty. If such spatial variability of Os concentrations were present, it
would likely contribute to more conservative modeled concentrations during periods of higher
than average NOX emissions and lower than average Os concentrations, which would limit
conversion of NO to NC>2. This may be one factor contributing to the apparent conservative bias
that appears at all NO2 monitors during the morning rush hour period, in addition to the
uncertainties regarding temporal profiles of mobile source emissions. The simple linear
treatment of NOX chemistry in the AERMOD model options does not explicitly account for the
O3 depleting effect of the NO-to-NO2 conversion.
Another issue that has been raised as a potential source of uncertainty in relation to the
NOX chemistry for this application is that the AERMOD modeling was conducted separately for
mobile sources, point sources, and the ATL airport source. The OLM option was used for the
mobile source and airport source emissions, while the PVMRM option was used for the point
sources. As a result, the ozone-limiting effects simulated by these chemistry options may have
been underestimated and may lead to a conservative bias in modeled concentrations. While this
may have been an issue with preliminary modeling that separated the on-road and off-road
components of mobile source emissions into separate AERMOD runs, the final application of
AERMOD combined the on-road and off-road emissions into a single run. Given the relative
contributions of these three categories of emissions, the potential impact of Os depletion due to
point source and airport emissions on total modeled concentrations, which are dominated by the
mobile source category, is likely to be insignificant.
8.12.1.6 Use of Unadjusted AERMOD NO2 Concentrations
While the number and range of field study data bases used in AERMOD's evaluation is
large relative to evaluation of other models, these field studies are predominantly associated with
elevated buoyant releases from power plant stacks. In most of these cases emission rates are
known with a high degree of accuracy, often based on continuous emission monitors (CEMs) for
operational facilities or based on controlled emissions in the case of tracer releases. Model
performance for mobile source emissions across an urban area is not well documented, and
235
-------
evaluating model performance for such applications is complicated by a number of challenges
and uncertainties. These include uncertainties regarding the temporal and spatial distribution of
emissions, physical characteristics of the emission sources (such as release height and initial
dilution parameters), and the influence of spatial variability of surface characteristics on
dispersion of low-level plumes, among other factors. In addition, application of the AERMOD
model to simulate NOX emissions across an urban domain in support of an exposure assessment
imposes different requirements than a more typical application of AERMOD to support a
specific regulatory permit application. Regulatory applications of AERMOD are generally
motivated by a need to estimate the peak concentrations across a domain of interest without
regard to any specific spatial or temporal pattern of impacts. In contrast, the use of AERMOD to
support an exposure assessment implies that modeled concentrations will be coupled with
population and other information in a way that implies some spatial and temporal pairing of
impacts, and also places more emphasis on the significance of the full concentration distribution
predicted by the model.
Despite these additional concerns and caveats regarding the use of AERMOD in this
REA, the evaluation of modeled air quality presented in Chapter 8.4.8 shows overall good
agreement between AERMOD modeled NO2 concentrations and available ambient monitored
NO2 concentrations. The model evaluation results are consistent across the available ambient
monitors and across all seasons. Bias in the predicted concentration distribution at each of the
monitor locations is generally within the range expected of refined dispersion models across the
cumulative concentration distribution. While some systematic positive bias is evident in the
diurnal profiles associated with morning rush hour peak in mobile source emissions, the degree
of bias in those cases is within the factor 2 commonly used to indicate relatively unbiased model
performance. In considering that this upward bias occurs mainly in the early morning hours, it is
possible that there may not be a large proportion of the simulated population exposed at these
times of the day. Therefore the upward bias may not have a large effect on the exposure results
presented. The actual affect on the exposure results from this bias remains uncertain, because
the time-of-day simulated individuals were exposed was not generated by APEX.
Given the uncertainties associated with determining emission profiles and source
characteristics for an urban-wide exposure assessment, previous assessments have often included
adjustments to modeled concentrations based on available ambient monitored concentrations as a
236
-------
means of reducing overall uncertainty in the analysis. Two main factors have argued against
such an approach for this REA. First, advances in the science of dispersion modeling, including
the promulgation of AERMOD model and more refined techniques for treating the temporal and
spatial variation of emissions, provide a basis for more confidence in the representativeness of
modeled concentrations for the exposure assessment. Second, the limited number of available
ambient NO2 monitors, the predominance of mobile source emissions of NOX within the
inventory, and the significant horizontal gradients of concentration associated with low-level
emissions from major roadways raise significant concerns regarding the spatial
representativeness of the ambient NC>2 concentrations for purposes of the exposure assessment.
Given that the air quality concentration estimates are estimated to be conservative, and that the
values at the upper tails of the hourly distribution are not unusual in comparison with the other
portions of the concentration distribution, it was determined that adjustment of the modeled air
quality based on the three monitors was not necessary. Any effort to adjust concentration
estimates based on monitored values would present a range of options and issues regarding how
the modeled concentrations would be adjusted both temporally and spatially in relation to the
observations. Based on these considerations, such an approach could actually increase the
uncertainty of the REA in ways that are difficult to characterize or justify.
8.12.2 Exposure Modeling Uncertainties
8.12.2.1 Population Data Base
The population and commuting data are drawn from U.S. Census data from the year
2000. This is a high quality data source for nationwide population data in the U.S. however, the
data do have limitations. The Census used random sampling techniques instead of attempting to
reach all households in the U.S., as it has in the past. While the sampling techniques are well
established and trusted, they may introduce uncertainty to exposure results. The Census has a
quality section (http://www.census.gov/quality/) that discusses these and other issues with
Census data. It is likely the bias within this data would not affect the results in any particular
direction, and given the use of the sampled demographics to represent the simulated population,
it is expected that the uncertainty in the exposure results from this source is low.
237
-------
8.12.2.2 Commuting Data Base
Commuting pattern data were derived from the 2000 U.S. Census. The commuting data
address only home-to-work travel. A few simplifying assumptions needed to be made to allow
for practical use of this data base to reflect a simulated individual's commute. First, there were a
few commuter identifications that necessitated a restriction of their movement from a home
block to a work block. This is not to suggest that they never travelled on roads, only that their
home and work blocks were the same. This includes the population not employed outside the
home, individuals indicated as commuting within their home block, and individuals that
commute over 120 km a day. This could lead to either over- or under-estimations in exposure if
they were in fact to visit a block with either higher or lower NC>2 concentrations. Given that the
number of individuals who meet these conditions is likely a small fraction of the total population
and that the bias is likely in either direction, the overall uncertainly is considered low.
Second, although several of the APEX microenvironments account for time spent in
travel, the travel is assumed to always occur in basically a composite of the home and work
block. No other provision is made for the possibility of passing through other census blocks
during travel. This could also contribute to bias in either direction, dependent on the number of
blocks the simulated individual would actually traverse and the spatial variability of the
concentration across different blocks. This could potentially affect a large portion of the
population, since we expect that at the block level, many persons would have a commute transect
that included more than two blocks, although the actual number of persons and the number of
blocks per commute and the spatial variability across blocks has not been quantified. In
addition, the commuting route (i.e., which roads individuals are traveling on during the
commute) is not accounted for. This may bias the exposure results in either direction, with some
individual under-estimated and others over-estimated.
Furthermore, the estimation of block-to-block commuter flows relied on the assumption
that the frequency of commuting to a workplace block within a tract is proportional to the
amount of commercial and industrial land in the block. This assumption may introduce a bias in
overestimating exposures if 1) the blocks with greater commercial/industrial land density also
have greater concentrations when compared with lower density commercial/industrial density
blocks, and 2) most persons commute to lower commercial/industrial density blocks. It should
also be noted that recent surveys, notably the National Household Transportation Survey
238
-------
(NHTS), have found that most trips taken and most VMT accrued by households are non-work
trips, particularly social/recreational and shopping-related travel (Hu and Reuscher, 2004). This
constitutes an unquantified source of uncertainty that is not be addressed by the Census
commuter dataset.
8.12.2.2 Human Time-Location-Activity Pattern Data
The CHAD time-location activity diaries used are the most comprehensive source of such
data and realistically represent where individuals are located and what they are doing. The
diaries are sequential records of each persons activities performed and microenvironments
visited. There are however, uncertainties the exposure results as a result of the CHAD diaries
used for simulating individuals may introduce uncertainty to the exposure results. First, much of
the data used to generate the daily diaries were collected in surveys conducted over 20 years ago.
While the trends in people's daily activities may not have changed much over the years, it is
certainly possible that some differences do exist. For example, it is estimated that between 1983
and 2001, the average miles traveled by people in the U.S. increased by 55%, corresponding to a
2.4% annual increase in miles traveled per person (Hu and Reuscher, 2004). Therefore, it is
possible that the overall commute times in the CHAD diaries used to estimate in-vehicle
exposures are biased low, resulting in under-estimation of exposures.
Second, the CHAD data are taken from numerous surveys that were performed for
different purposes. Some of these surveys collected only a single diary-day while others went on
for several days. Some of the studies were designed to not be representative of the U.S.
population, although a large portion of the data is from National surveys. In addition, study
collection periods occur at different times of the year, possibly resulting in seasonal differences.
This could add uncertainty to the results if there are characteristics of the survey population that
are distinct from the simulated population.
The CHAD diaries that are selected from APEX to represent the Atlanta population are
not all from Atlanta, the state of Georgia, or from the Southeast, albeit some of the diaries may
be. As stated above, most of the diaries are from National surveys, therefore there are diaries
from locations other than Atlanta that are used to simulate the Atlanta population. A few of the
limitations associated with the use of diaries from different locations or seasons are corrected by
the approaches used in the exposure modeling. For example, diaries used are weighted by
239
-------
population demographics (i.e., age and gender) for a particular location and temperature is used
as a classification variable to account for its affect on human activities.
A sensitivity analysis was performed to evaluate the affect of using different CHAD
studies has on APEX results for the recent 63 NAAQS review (see Langstaff (2007) and EPA
(2007g)). Briefly, Os exposure results were generated using APEX with all of the CHAD diaries
and compared with results generated from running APEX using only the CHAD diaries from the
National Human Activity Pattern Study (NHAPS), a nationally representative study in CHAD.
There was agreement between the APEX exposure results for the 12 metropolitan areas
evaluated (one of which was Atlanta), whether all of CHAD or only the NHAPS component of
CHAD is used. The absolute difference in percent of persons above a particular concentration
level ranged from -1% to about 4%, indicating that the exposure model results are not being
overly influenced by any single study in CHAD. It is likely that similar results would be
obtained here for NC>2 exposures, although it remains uncertain due to different averaging times
(1-hour vs. 8-hour average).
This is not to suggest that the uncertainty is low in using the CHAD data to represent the
Atlanta area, but that similar results would be obtained in using the diaries available, so long as
the population was appropriately stratified and certain characteristics influencing exposure were
considered. One particular influential factor that is not modeled by APEX is the commute
time/distances for the Atlanta population. The Atlanta population is spread over a larger area
than most other locations and as a result, individual spend more time driving (Table 8-16). Not
taking this added drive time into account when using the CHAD diaries could lead to under-
estimation of exposures for the Atlanta population. Given the difference in Atlanta DVMT in
comparison with other locations, it is possible that this underestimation is large. However, in
considering this lack of accounting for the greater Atlanta commute times that exist and that an
important driver for exposures above selected levels was from the in-vehicle microenvironment,
the Atlanta exposure results may to some degree be representative of other locations in the U.S.
with more nationally representative commute times.
8.12.2.3 Longitudinal Profile
APEX creates seasonal or annual sequences of daily activities for a simulated individual
by sampling human activity data from more than one subject. Each simulated person essentially
240
-------
becomes a composite of several actual people in the underlying activity data. Certain aspects of
the personal profiles are held constant, though in reality they change as an individual ages. This
is only important for simulations with long timeframes, particularly when simulating young
children (e.g., over a year or more).
The cluster algorithm used in constructing longitudinal profiles was evaluated against a
sequence of available multiday diaries sets collected as part of the Harvard Southern California
Chronic Ozone Exposure Study (Xue et al. 2005, Geyh et al. 2000). Briefly, the activity pattern
records were characterized according to time spent in each of 5 aggregate microenvironments:
indoors-home, indoors-school, indoors-other, outdoors, and in-transit. The predicted value for
each stratum was compared to the value for the corresponding stratum in the actual diary data
using a mean normalized bias statistic. See Appendix B, Attachment 2 and 3 for details. The
evaluation indicated the cluster algorithm can replicate the observed sequential diary data, with
some exceptions. The predicted time-in-microenvironment averages matched well with the
observed values. For combinations of microenvironment/age/gender/season, the normalized bias
ranges from -35% to +41%. Sixty percent of the predicted averages have bias between -9% and
+9%, and the mean bias across any microenvironment ranges from -9% to +4%. Although, on
occasion there were large differences in replicating variance across persons and within-person
variance subsets, about two-thirds of the predictions for each case were within 30% of the
observed time spent in each microenvironment.
The longitudinal approach used in the exposure assessment was an intermediate between
random selection of diaries (a new diary used for every day for each person in the year) and
perfect correlation (same diary used for every day for each person in the year). The cluster
algorithm used here was previously compared with two other algorithms, one that used random
sampling and the other employing both diversity (D) and autocorrelation (^4) statistics (see EPA,
2007g for details on this algorithm). The number of persons with at least one or more exposures
to a given 63 concentration was about 30% less when using the cluster algorithm than when
using random sampling, while the number of multiple exposures for those persons exposed was
greater using the cluster algorithm (by about 50%). The algorithm employing the D and A
statistics exhibited similar patterns, although were lower in magnitude when compared with
random sampling (about 5% fewer persons with one or more exposures, about 15% greater
multiple exposures). These exposure results using the cluster algorithm in APEX appeared to be
241
-------
the result of a greater correlation of diaries selected in comparison with the other two algorithms.
This outcome conforms to an expectation of correlation between the daily activities of
individuals. While the evaluation was performed using 8-hour O3 as the exposure output it is
expected that similar results would be obtained for 1-hour NC>2 exposures. That is, the
characteristics of the diaries that contribute greatly to any pollutant exposure above a given
threshold (e.g., time spent outdoors, vehicle driving time) are likely a strong component in
developing each longitudinal profile. Given these results and that the REA in not necessarily
focused on health effects resulting from multi-day exposures, the particular longitudinal
approach used likely contributes minimally to uncertainty. See Appendix B, Attachments 2 and
3 for further details in the cluster algorithm and the evaluations performed.
8.12.2.4 Meteorological Data
Meteorological data are taken directly from monitoring stations in the assessment areas.
It is assumed that most of the data used are error free and have undergone required quality
assurance review. One strength of these data is that it is relatively easy to see significant errors if
they appear in the data. Because general climactic conditions are known for the simulated area,
it would have been apparent upon review if there were outliers in the dataset, and at this time
none were identified. If there were a bias in the data, it would be expected to be limited in
extend and randomly occurring, therefore contributing to both under and over-estimations
equally to a marginal degree. To reduce the number of calms and missing winds in the 1-hour
MET data, archived one-minute winds for the ASOS station at ATL were used to calculate
hourly average wind speed and directions. This approach reduces the number of estimated zero
concentrations that would be output by AERMOD if not supplemented by the additional wind
data, thus preventing a downward bias in the predicted 1-hour concentrations.
However, there are limitations in the use of these data. APEX only uses one temperature
value per day in selecting an appropriate CHAD diary and indoor microenvironment air
exchange rate. Because the model does not represent hour-to-hour variations in meteorological
conditions throughout the day, there may be uncertainty in some of the exposure estimates for
indoor microenvironments (see the next section).
242
-------
8.12.2.5 Air Exchange Rates (AER)
The residential air exchange rate (AER) distributions used to estimate indoor exposures
contribute to uncertainty in the exposure results. Three components of the AER analyzed
previously by EPA (2007g) include 1) the extrapolation of air exchange rate distributions
between-CMSAs, 2) analysis of within-CMSA uncertainty due to sampling variation, and 3) the
uncertainty associated with estimating daily AER distributions from AER measurements with
different averaging times. The results of those previous investigations are briefly summarized
here. See Appendix B, Attachments 8 and 9 for details.
Extrapolation of AER among locations
Air exchange rate (AER) distributions were assigned in the APEX model, as detailed in
the indoors-residential microenvironment. Since specific AER data for Atlanta were not
available, data from another location were used to represent AERs in Atlanta based on having
potentially similar influential characteristics. Such factors include age composition of housing
stock, construction methods used, and other meteorological variables not explicitly treated in the
analysis, such as humidity and wind speed patterns. AER data from measurements in Research
Triangle Park, NC (RTF) were selected to represent the distribution of AERs in Atlanta (see
Appendix B, Attachment 5).
In order to assess the uncertainty associated with this extrapolation, between-location
uncertainty was evaluated by examining the variation of the geometric means and standard
deviations across several cities and studies. The evaluation showed a relatively wide variation
across different cities in their AER geometric means and standard deviations, stratified by air-
conditioning status, and temperature range. For example, Figure 8-30 Illustrates the GM and
GSD of AERs estimated for several cities in the U.S. where A/C was present and within the
temperature range of 20-25 °C. The wide range in GM and GSD pairs implies that the modeling
results may be very different if the matching of modeled location to study location was changed.
For example, the NC>2 exposure estimates may be sensitive to use of an alternative distribution,
say those in New York City, compared with results generated using the RTF AER distributions.
It is likely though that the true distribution is more similar to the selected distribution from RTF
than New York City or some other location given the population of available AER data. It is
unclear as to the direction of bias given the limited number of data available for comparison. It
243
-------
should be noted that Houston, the only other "southern" location, generally coincided with the
RTF AERs distribution in the 20-25 °C and other temperature ranges for homes with A/C (see
Appendix B, Attachment 8).
Geometric mean and standard deviation of air exchange rate
For different cities and studies
Air Conditioner Type: Central or Room A/C
Temperature Range: 20-25 Degrees Celsius
4.0—
3.5—
O 3.0—
p
5 ? n—
tD ^-u
O
A c
G
D
B E
I
0.5
I
1.5
Geometric Mean
2.5
I
3.0
AAAHouston B B BLosAngeles C C CLosAngeles-Avol D D DLosAngeles-RIOPA
EEELosAnge1es-Wi1son]984 F F FNewYorkCity GGGNewYorkCity-RIOPA HHHNewYorkCity-TEACH
1 1 iRedBluff J J JResearchTrianglePark
Figure 8-30. Example comparison of estimated geometric mean and geometric standard
deviations of AER (h-1) for homes with air conditioning in several cities.
Within location uncertainty
There is also variation in AERs within studies for the same location (e.g., Research
Triangle Park, NC), but this is much smaller than the observed variation across different
CMSAs. This finding tends to support the approach of combining different studies for a CMSA,
where data were available. The within-city uncertainty was assessed by using a bootstrap
distribution to estimate the effects of sampling variation on the fitted geometric means and
standard deviations for the RTF data used to represent the Atlanta AERs. These bootstrap
distributions assess the uncertainty due to random sampling variation. They do not address other
244
-------
uncertainties such as the lack of representativeness of the available study data or the variation in
the lengths of the AER monitoring periods. Because only the GM and GSD were used, the
bootstrap analyses does not account for uncertainties about the true distributional shape, which
may not necessarily be lognormal.
One-thousand bootstrap samples were randomly generated for each AER subset (of size
N), producing a set of 1,000 geometric mean (GM) and geometric standard deviation (GSD)
pairs. The analysis of the RTF data used to represent Atlanta indicated that the GSD uncertainty
for a given AER temperature group tended to have a range within ±0.25 fitted GSD (hr"1), with
smaller intervals surrounding the GM (i.e, about ± 0.10 fitted GM (hr"1) (Figure 8-31). Broader
ranges were generated from the bootstrap simulation for AER distributions used for Atlanta
homes without A/C (Figure 8-32), although both still within ±0.5 of the fitted GM and GSD
values. See Appendix B, Attachment 8 for further details.
Geometric mean and standard deviation of air exchange rate
Bootstrapped distributions for different cities
City: Research Triangle Park
Air Conditioner Type: Central or Room A/C
Temperature Range: 20-25 Degrees Celsius
4.0—
3.5 —
I 3.0—
"3
V.
O O «
1
g 2.0—
C
\
0.5
I
1.0
I
1.5
Geometric Mean
2.0
I
2.5
3.0
•Bootstrapped Data +++Onginal Data
Figure 8-31. Example of boot strap simulation results used in evaluating random sampling
variation of AER (h-1) distributions (RTF data).
245
-------
Geometric mean and standard deviation of air exchange rate
Bootstrapped distributions for different cities
City: Outside California
Air Conditioner Type: No ,VC
Temperature Range: >20 Degrees Celsius
4.0 —
3.5 —
P 3.0-
OT
.
-------
Atlanta AHS, 2004). EPA (2007g) details the specification of uncertainty estimates in the form
of confidence intervals for the air conditioner prevalence rate, and compares these with
prevalence rates and confidence intervals developed from the Residential Energy Consumption
Survey (RECS) of 2001 for several aggregate geographic subdivision (e.g., states, multi-state
Census divisions and regions) (EIA, 2001).
Briefly, Air conditioning prevalence rates were 97% for Atlanta, with reported standard
errors of 1.2% (AHS, 2004). Estimated 95% confidence intervals were also small and span
approximately 4.6% (AHS, 2003). The RECS prevalence estimates for Atlanta and confidence
intervals compared well with a value of 95.0% and a 95% confidence interval spanning 5.8%.
This suggests that there is limited bias in the A/C prevalence estimates used and that the
uncertainty in the estimated value is likely low.
A sensitivity analysis was performed to evaluate changes in the estimated exposures
when using a lower A/C prevalence. Changing the A/C prevalence form the actual Atlanta value
used allows for a greater percentage of homes to use estimated AER distributions for homes
without A/C. An A/C prevalence of 0.55 (or 55%) was input to a new simulation using 2002 air
quality without indoor sources, based on the lower bound of observed A/C prevalence rates in
Table 8-15). Table 8-21 indicates that there is no change in the percent of asthmatics exposed at
or above each of the potential health effect benchmark levels, whether considering a single
exposure or up to six exposures in a year. There are however, several thousand additional person
days, or additional days for persons that are already exposed to daily maximum concentrations at
or above the benchmarks when considering the simulation conducted using the lower prevalence
rate. Only a few additional persons were exposed to benchmarks >200 ppb that did not have
such exposures in the simulation using the higher A/C prevalence rate. These results suggest that
the indoor-residential microenvironment contributes much less to exposures above any of the
benchmarks when compared with the estimated contribution from on-road and near-road
microenvironments. Most individuals (>99%) were already estimated to experience at least one
exposure at concentrations at or above 100 to 150 ppb through the roadway related exposures,
one of the main reasons why there are no additional persons exposed at these benchmark levels
when considering the lower A/C prevalence. Even though there is a large fraction of the
population not exposed to benchmark levels >200 ppb (18-41%) using the higher prevalence
rate, the outdoor ambient concentrations rarely would exceed these concentrations. Thus only a
247
-------
few additional persons experience these higher benchmark level concentrations with the lower
A/C prevalence.
Table 8-20. Comparison of exposure results using a 0.55 versus 0.97 A/C prevelance for 2002 air
quality without indoor sources.
Simulation
no-indoor -
0.97 AC
prevalence
no indoor-
0.55 AC
prevalence
Absolute
Difference
(0.97
ACprev-
0.55
ACprev)
Benchmark
Level (ppb)
100
150
200
250
300
100
150
200
250
300
100
150
200
250
300
Percent of All Asthmatics with Indicated
Number of Multiple Daily Maximum 1-hour
Exposures At or Above Benchmark Level
1
100%
99%
92%
76%
59%
100%
99%
92%
76%
59%
0
0
0
0
0
2
100%
96%
80%
56%
32%
100%
96%
81%
56%
33%
0
0
0
0
0
3
100%
93%
69%
40%
19%
100%
93%
69%
40%
20%
0
0
0
0
0
4
99%
88%
58%
28%
12%
99%
88%
58%
28%
12%
0
0
0
0
0
5
99%
83%
48%
20%
8%
99%
83%
48%
20%
8%
0
0
0
0
0
6
99%
78%
40%
15%
6%
99%
78%
40%
15%
6%
0
0
0
0
0
Person
Days
Above
Benchmark
7393854
2839603
1271622
618725
323273
7442239
2849885
1274646
620237
323878
-48385
-10282
-3024
-1512
-605
Number of
Persons
with at
Least One
Exposure
212426
209855
195766
161863
124531
212426
209855
195875
161918
124637
0
0
-109
-54
-106
8.12.2.7 Indoor Source Estimation
Other indoor NO2 emission sources, such as emissions from gas pilot lights, gas heating,
unvented gas fire places, gas water heaters, or gas clothes drying were not included in this
analysis due to lack of adequate data readily usable for characterization, modeling complexities
regarding the assignment of particular sources to the simulated population (e.g, correlations of
sources), a limited time to conduct analyses of potential data distributions including the analysis
of their uncertainties, and limited resources allocated for inclusion in the review. Exclusion of
these sources would bias all indoor concentrations low when these sources are present, however,
it is largely uncertain how much it would affect any estimates of the benchmark level
exceedances, and the number of persons affected in Atlanta in the absence of source emission
and prevalence data.
There may be uncertainty added to the exposure results when considering the form (i.e.,
uniform) and limits (limited by the bounds of the measurement data) of the distribution used to
248
-------
represent indoor decay. The measurement data used to develop the distribution were obtained
from a single study, conducted in a single home, under limited temperature and humidity
conditions (see Table 8-11). The experimental design included four different methods for
introducing NC>2 into the home, including direct injection and home appliances. Different
homes, with varying construction materials and furnishings may have different NC>2 decay rates,
and as a result of the uniform distribution selected, the number of exposures may be either over-
or under-estimated depending on the extent of how far the true population of NC>2 values extend
outside the range of the uniform distribution used (i.e., 1.02-1.45).
A sensitivity run was performed using an alternative fitted distribution, assuming a
lognormal form with GM and GSD of 1.17 and 1.14, respectively, and lower and upper bounds
defined by 50% of observed minimum (0.51) and 100% of observed maximum (2.9). The
exposure results generated using the lognormal decay distribution were compared with the
simulations performed using the uniform distribution for 2002 air quality and without indoor
sources (Table 8-21). There was no difference in the percent of asthmatics estimated to
experience one through six daily maximum exposures in a year above any of the benchmarks.
There were however a few additional person days above each of the benchmark levels except for
<300 ppb and 54 additional persons exposed at or above 200 and 250 ppb when using the
lognormal distribution. This suggests that the simulated exposures above the selected
benchmarks are not sensitive changes in the indoor decay rate NC>2. Whether the same outcome
would occur with additional alternative distributions of different forms and bounds or that the
indoor microenvironment is not sensitive to indoor decay based on the algorithm used by APEX
remains largely unknown.
Table 8-21. Comparison of exposure results using a uniform versus lognormal NO2 indoor decay
distribution for 2002 air quality without indoor sources.
Simulation
no-indoor
uniform
decay
no indoor-
Benchmark
Level (ppb)
100
150
200
250
300
100
Percent of All Asthmatics with Indicated
Number of Multiple Daily Maximum 1-hour
Exposures at or Above Benchmark Level
1
100%
99%
92%
76%
59%
100%
2
100%
96%
80%
56%
32%
100%
3
1 00%
93%
69%
40%
19%
1 00%
4
99%
88%
58%
28%
12%
99%
5
99%
83%
48%
20%
8%
99%
6
99%
78%
40%
15%
6%
99%
Person
Days
Above
Benchmark
7393854
2839603
1271622
618725
323273
7402926
Number of
Persons
with at
Least One
Exposure
212426
209855
195766
161863
124531
212426
249
-------
lognormal
decay
absolute
difference
(uniform-
lognormal)
150
200
250
300
100
150
200
250
300
99%
92%
76%
59%
0
0
0
0
0
96%
80%
56%
32%
0
0
0
0
0
93%
69%
40%
19%
0
0
0
0
0
88%
58%
28%
12%
0
0
0
0
0
83%
48%
20%
8%
0
0
0
0
0
78%
40%
15%
6%
0
0
0
0
0
2842022
1272227
619027
323273
-9072
-2419
-605
-302
0
209855
195821
161918
124531
0
0
-54
-54
0
The data used to estimate the average number of daily food preparation events is older
than the time period assessed (1992 versus 2001-2003) and may therefore be not reflect current
conditions to some degree, possibly leading to either under- or over-estimates of exposure to
concentration exceedances. For example, if the population of Atlanta in 2003 that uses gas
stoves to prepare food at home does so less frequently than reported the 1992 survey population,
then the number of such exposures may be over-estimated. The variability associated with the
mean usage of 1.4 that was used in the model is also under-represented in that there are likely
some individuals that cook more or less than this value. Furthermore, the estimate is not specific
for the Atlanta population. The uncertainty regarding each of these issues and how they may
affect the exposure results is largely unknown.
As noted in the microenvironmental description, it was assumed that the probability that a
food preparation event included stove use was the same no matter what hour of the day the food
preparation event occurred. If such probabilities differ, then the diurnal allocation of cooking
events may differ from the actual pattern. To the extent that the gas stove usage patterns may
correlate with ambient concentration patterns, the number of exposures to exceedances of
threshold concentrations of concern may be under- or over-estimated. For example, if gas stove
usage and ambient concentrations are positively correlated (e.g., if cooking tends to occur during
evening rush hour) and the diurnal allocation assumed here results in a lower correlation (e.g., if
the diurnal allocation understates the probability of gas stove usage at times of high ambient
concentrations) then the number of such exposures may be under-estimated. As another example,
if the diurnal pattern allocation assumed here understates the probability of gas stove usage at
times when simulated subjects are assumed to be at home, then the number of such exposures
may be under-estimated.
250
-------
There is also uncertainty regarding the distribution used to estimate the indoor source
concentration contribution. Concentrations were obtained from CARB (2001) from a variety of
described cooking conditions (e.g., with or without ventilation, different pans) and foods cooked
(e.g., bacon, french fries, broiled fish, lasagna) in a single test home. While an alternative
distribution form may be fitted to such data, there already exist large uncertainties regarding the
representation of these measured concentrations obtained under the limited experimental
conditions to the population of all possible cooking conditions, the foods cooked, and the proper
weighting of such cooking events for the simulated population in Atlanta. In the absence of such
knowledge, it is likely that a fitted distribution would be biased high. In considering these
uncertainties, staff elected to use a uniform distribution, bounded by the upper and lower range
of the experimental data. Use of this uniform distribution does exclude concentrations outside
the upper value, suggesting that concentrations in excess of the upper bound are an impossibility.
This is unlikely and may add to the uncertainty in the estimated exposures when cooking with
gas, and implies a bias in underestimating indoor source contribution to indoor concentrations.
Although it appears that the study was primarily designed to estimate upper percentile PM
concentrations, it is possible that the uniform distribution selected for NC>2 does capture the
upper range of concentrations very well due to the presence of study-designed "worst-case"
cooking scenarios.
The durations of the CARB (2001) cooking tests ranged from 21 minutes to 3 hours with
an average of about 70 minutes. For implementation in APEX it was assumed that each cooking
event lasts exactly an hour. That is, the randomly selected net concentration contribution was
added to hourly average indoor concentration for the hour it was selected to occur. Because the
mass balance algorithm leads to carryover from one hour to the next, some of the indoor cooking
impact will influence subsequent hours. However, the affect of the cooking event may be
overstated or understated for cooking events longer or shorter than 1 hour.
251
-------
9. CHARACTERIZATION OF HEALTH RISKS USING DATA
FROM EPIDEMIOLOGICAL STUDIES
9.1 INTRODUCTION
As mentioned above in chapter 6, in response to advice received from the CASAC NO2
Panel on the 1st draft REA, we have conducted a focused quantitative risk assessment in which
estimates of respiratory ED visits as a function of ambient levels of NO2 have been developed for
a single urban area (i.e., the Atlanta MSA). In this approach, concentration-response functions
derived from NO2 epidemiological studies are used in conjunction with 1) ambient air quality
data representing as is and alternative air quality scenarios and 2) baseline incidence data for
respiratory ED visits to estimate the impact of ambient levels of NO2 on ED visits associated
with these air quality scenarios.
The purpose of this chapter is to present the results for the current risk assessment which
is an illustrative case study that provides information on the magnitude and potential changes in
NO2-related public health impacts associated with recent air quality and alternative air quality
scenarios simulating attainment of the current and alternative NO2 standards. We note that
chapters 4 and 5 of this document provide additional qualitative assessment of the
epidemiological evidence most relevant to characterizing NO2-related health effects in the
United States; this includes both a discussion of respiratory-related ED visits as well as other
health endpoints. We also note that integration of the scientific evidence presented in the ISA
(EPA, 2008a) with the air quality, exposure, and risk characterization results presented in
chapters 7 through 9 of this document is presented in chapter 10. Chapter 10 also discusses
staffs assessment of how this information might be considered in evaluating the adequacy of the
current NO2 NAAQS and the need for potential alternative primary NO2 standards.
Previous reviews of the NO2 primary NAAQS, completed in 1985 and 1996, did not
include quantitative health risk assessments. Thus, the risk assessment described in this
document builds upon the methodology and lessons learned from the risk assessment work
conducted for the recently concluded PM and Os NAAQS reviews (Abt Associates, 2005; Abt
Associates, 2007). Many of the same methodological issues are present in conducting a risk
252
-------
assessment for each of these criteria air pollutants where epidemiological studies provided the
basis for the concentration-response relationships used in the quantitative risk assessment.
The NO2 health risk assessment described in this chapter estimates the incidence of
respiratory-related ED visits associated with short-term exposures to NO2 under recent ("as is")
air quality levels, upon just meeting the current NO2 standard of 0.053 ppm annual average, and
upon just meeting several potential alternative NO2 primary NAAQS in the Atlanta MSA.l As
discussed in more detail in chapter 6 above, staff has elected to evaluate daily maximum 1-h
standard levels of 0.05, 0.10, 0.15, and 0.20 ppm using both 98th and 99th percentile forms and
averaged over a thee-year period.2 The risk assessment is intended as a tool that, together with
other information on this health endpoint and other health effects evaluated in the final ISA and
discussed elsewhere in this document, can aid the Administrator in judging whether the current
primary standard protects public health with an adequate margin of safety, or whether revisions
to the standard are appropriate.
Section 9.2 describes the general approach used to conduct the risk assessment for ED
visits. Sections 9.3, 9.4, and 9.5 discuss in more detail the three types of inputs required to
conduct the assessment. Section 9.6 presents a discussion of uncertainties and variability and
section 9.7 presents a summary of results from the assessment and key observations.
9.2 GENERAL APPROACH
The general approach used for the NO2-related ED risk assessment is dictated by the fact
that it is based on concentration-response functions which have been estimated in
epidemiological studies evaluated in the final ISA. Since these studies estimate concentration-
response functions using ambient air quality data from fixed-site, population-oriented monitors,
the appropriate application of these functions in a risk assessment similarly requires the use of
ambient air quality data at fixed-site, population-oriented monitors. In order to estimate the
incidence of respiratory-related ED visits associated with recent air quality conditions in a set of
counties attributable to ambient NO2 exposures, as well as the change in incidence of this health
effect in that set of counties corresponding to a given simulated change in NO2 levels
1 The current NO2 standard refers to a two-year period and requires that the annual average NO2 level be less than or
equal to 0.053 ppm in each of the two years.
2 As an example, for the alternative standards using the 98th percentile form, the standard is met when the average of
the annual 98th percentile daily maximum 1-hour concentrations for a 3-year period is at or below the specified
standard level.
253
-------
representing just meeting the current or alternative 1-h daily maximum NC>2 standards, the
following thee elements are required:
• Air quality information including: (1) "as is" air quality data for NC>2 from
ambient monitors in the assessment location, and (2) "as is" concentrations adjusted
to reflect patterns of air quality estimated to occur under a simulation where the
area's air quality is adjusted to just meet the specified standard. (These air quality
inputs are discussed in more detail in section 6.2 of this document).
• Concentration-response functions which provide an estimate of the relationship
between the health endpoint of interest and ambient NC>2 concentrations.
• Baseline health effects incidence. The baseline incidence of the health effect in
the assessment location in the target year is the incidence corresponding to "as is"
NC>2 levels in that location in that year.
Figure 9-1 provides a broad schematic depicting the role of these components in the NC>2
risk assessment. Each of the key components (i.e., air quality information, estimated
concentration-response functions, and baseline incidence) is discussed below, highlighting those
points at which judgments have been made.
These inputs are combined to estimate health effect incidence changes associated with
specified changes in NC>2 levels. Although some epidemiological studies have estimated linear
or logistic concentration-response functions, by far the most common form, and the form
relevant for the epidemiological study used in the current risk assessment is the exponential (or
log-linear) form:
y = Beftc, (Equation 9-1)
where x is the ambient NO2 level, y is the incidence of the health endpoint of interest at NO2
level x, ft is the coefficient of ambient NC>2 concentration (describing the extent of change my
with a unit change in x), and B is the incidence at x=0, i.e., when there is no ambient NC>2. The
relationship between a specified ambient NC>2 level, XQ, for example, and the incidence of a given
health endpoint associated with that level (denoted asj/o) is then
y0 = Be^ . (Equation 9-2)
254
-------
If we let x0 denote the baseline (upper) NC>2 level, and x} denote the lower NC>2 level, and
denote the corresponding incidences of the health effect, we can derive the following
255
-------
Air Quality
Ambient Monitoring for
Selected Urban Areas
Recent ("As Is") Ambient
NO, Levels
Air Quality Adjustment
Procedures
Current and Alternative
Proposed Standards
Changes in
Distribution of
NO2 Air
Quality
Concentration-Response
Human Epidetniological
Studies
Concentration -
Response
Relationships
Estimates of City-specific
Baseline Health Effects
Incidence or Incidence
Rates and Population
Data
Health
Risk
Model
Risk Estimates:
* Recent Air
Quality
* Current
Standard
• Alternative
Standards
Figure 9-1. Major components of nitrogen dioxide health risk assessment for emergency
department visits.
256
-------
relationship between the change in x, Ax= (XQ- xj), and the corresponding change in;;, Ay, from
equation (9-1)3:
Ay = (J0 - Ji) = Jo[l - ^A*] • (Equation 9-3)
Alternatively, the difference in health effects incidence can be calculated indirectly using
relative risk. Relative risk (RR) is a measure commonly used by epidemiologists to characterize
the comparative health effects associated with a particular air quality comparison. The risk of
ED visits for respiratory illness at ambient NC>2 level XQ relative to the risk of ED visits for
respiratory illness at ambient NC>2 level xi, for example, may be characterized by the ratio of the
two rates: the rate of ED visits for respiratory illness among individuals when the ambient NC>2
level is x0 and the rate of ED visits for respiratory illness among (otherwise identical) individuals
when the ambient NC>2 level is xj. This is the RR for ED visits for respiratory illness associated
with the difference between the two ambient NC>2 levels, XQ and xj. Given a concentration-
response function of the form shown in equation (9-1) and a particular difference in ambient NC>2
levels, Ax, the RR associated with that difference in ambient NC>2, denoted as RR-Ax, is equal to
epAx rpjie Difference m health effects incidence, Ay, corresponding to a given difference in
ambient NC>2 levels, Ax, can then be calculated based on this RR-Ax as
AX = (y0 ~ *) = JVotl - (l/^Ax)] • (Equation 9-4)
Equations (9-3) and (9-4) are simply alternative ways of expressing the relationship
between a given difference in ambient NC>2 levels, Ax > 0, and the corresponding difference in
health effects incidence, Ay. These health impact equations are the key equations that combine
air quality information, concentration-response function information, and baseline health effects
incidence information to estimate health risks related to changes in ambient NC>2 concentrations.
If Ax < 0 - i.e., if Ax = (xi - XQ) - then the relationship between Ax and Ay can be shown to be
Ay = (yl - y0) = y0 [epta - 1]. If Ax < 0, Ay will similarly be negative. However, the magnitude of Ay will be the
same whether Ax > 0 or Ax < 0 - i.e., the absolute value of Ay does not depend on which equation is used.
257
-------
9.3 AIR QUALITY INFORMATION
As illustrated in Figure 9-1, and noted earlier, air quality information required to conduct
the NC>2 risk assessment includes (1) recent air quality data for NC>2 from a suitable monitor for
the assessment location and (2) air quality adjustment procedures to modify the recent data to
simulate air quality data just meeting the current annual and potential alternative 1-h daily
maximum standards. The approach used to adjust air quality data to simulate meeting specified
standards is discussed above in section 6.2.
In the first part of the risk assessment, we estimate the incidence of the health effect
associated with "as is" levels of NC>2 (or equivalently, the change in health effect incidence, Ay,
associated with a change in NC>2 concentrations from "as is" levels of NC>2 to 0 ppb). In the
second part, we estimate the incidence of the health effect associated with NC>2 concentrations
simulated to just meet a specified standard (i.e., the current NC>2 standard of 0.053 ppm annual
average as well as each of potential alternative 1-h daily maximum standards).
To estimate the incidence of a health effect associated with "as is" NC>2 levels in a
location, we need a time series of hourly "as is" NC>2 concentrations for that location. We have
used monitor data from the Georgia Tech monitor (monitor id =131210048), the monitor that
was used in Tolbert et al. (2007), the epidemiological study from which we obtained the
concentration-response functions (see section 9.4 below). Complete hourly data were available
on over 93 percent of the days - 348 days in 2005, 345 days in 2006, and 340 days in 2007.
Missing NC>2 concentrations were filled in, as described in section 3.5 of Appendix C.
Because Tolbert et al. (2007) estimated a relationship between daily respiratory-related
ED visits and the 3-day moving average (i.e., NC>2 levels on the same day, the previous day, and
the day before that) of the daily 1-h maximum NC>2 concentrations, we calculated the 3-day
moving average of the daily 1-h maximum NO2 concentrations at the monitor to provide the air
quality input to the risk assessment.
The calculations for the second part of the risk assessment, in which we estimated risks
associated with NC>2 levels simulated to just meet the current annual standard and potential
alternative 1-h daily maximum standards were done analogously, using the monitor-specific
series of adjusted daily maximum hourly concentrations rather than the monitor-specific series of
"as is" daily maximum hourly concentrations.
258
-------
9.4 CONCENTRATION-RESPONSE FUNCTIONS
As indicated in Figure 9-1, another key component in the risk assessment model is the set
of concentration-response functions which provide estimates of the relationship between the
health endpoint of interest and ambient NO2 concentrations. As discussed above, the health
endpoint of interest for this focused quantitative risk assessment is respiratory-related ED visits.
As discussed in sections 4.2.2 and 4.5.2 several community epidemiological studies have been
conducted in the U.S. that examined the relationship between NO2 and other air pollutants and
increased ED visits either for all respiratory causes or for asthma-related visits. Figure 5-1 in
this document summarizes the single pollutant model effect estimates from these studies. As
discussed in section 4.5.2, staff has considered several factors in selecting the urban area and
epidemiological studies upon which the current risk assessment is based. First, we have judged
that studies conducted in the United States are preferable to those conducted outside the United
States given the potential for effect estimates to be impacted by factors such as the ambient
pollutant mix, the placement of monitors, activity patterns of the population, and characteristics
of the healthcare system. Second, we judged that studies of ambient NO2 are preferable to those
of indoor NO2 given that studies of indoor NO2 focus on exposures in locations with indoor
sources of NO2. These indoor sources can result in exposure patterns, NO2 levels, and co-
pollutants that are different from those typically associated with ambient NO2. Third, we judged
it appropriate to focus on studies of ED visits. When compared to studies of respiratory
symptoms, the public health significance of ED visits is less ambiguous for the individuals
affected. In addition, baseline incidence data are more readily available for these endpoints.
Finally, we judged it appropriate to focus on studies that evaluated NO2 health effect associations
using both single- and multi-pollutant models. Taking these factors into consideration, we have
chosen to focus on the studies by Tolbert and colleagues (2007) in Atlanta, Georgia that address
ED visits for respiratory causes as a case study to illustrate the magnitude and changes in
estimated NO2-related risks for this endpoint for various air quality scenarios.
Tolbert et al. (2007) estimated concentration-response functions using both single
pollutant models (i.e., where NO2 was the only pollutant entered into the health effects model)
and multi-pollutant models (i.e., where one or two co-pollutants (PMio, 63, CO) were entered
into the health effects model). To the extent that any of the co-pollutants present in the ambient
air may have contributed to the health effects attributed to NO2 in single pollutant models, risks
259
-------
attributed to NO2 might be overestimated where concentration-response functions are based on
single pollutant models. However, if co-pollutants are highly correlated with NO2, their
inclusion in an NO2 health effects model can lead to misleading conclusions in identifying a
specific causal pollutant. When collinearity exists, inclusion of multiple pollutants in models
often produces unstable and statistically insignificant effect estimates for both NO2 and the co-
pollutants. Given that single and multi-pollutant models each have both potential advantages and
disadvantages, with neither type clearly preferable over the other in all cases, we report risk
estimates based on both single- and multi-pollutant models in the NO2 risk assessment.
All of the models in Tolbert et al. (2007) used a 3-day moving average of pollution levels
(i.e., the average of 0-, 1-, and 2-day lags), so the issue of which of several different lag
structures to select does not arise. The issue of how well a given lag structure captures the actual
relationship between the pollutant and the health effect, however, is still relevant. Models in
which the pollutant-related incidence on a given day depends only on same-day or previous-day
pollutant concentration (or some variant of those, such as a two- or thee-day average
concentration) necessarily assume that the longer pattern of pollutant levels preceding the
pollutant concentration on a given day does not affect incidence of the health effect on that day.
To the extent that a pollutant-related health effect on a given day is affected by pollutant
concentrations over a longer period of time, then these models would be mis-specified, and this
mis-specification would affect the predictions of daily incidence based on the model. The extent
to which short-term NO2 exposure studies may not capture the possible impact of long-term
exposures to NO2 is unknown. A number of epidemiologic studies have examined the effects of
long-term exposure to NO2 and observed associations with decrements in lung function and
partially irreversible decrements in lung function growth. The final ISA (EPA, 2008a)
concludes, however, that "overall, the epidemiological evidence was suggestive but not sufficient
to infer a causal relationship between long-term NO2 exposure and respiratory morbidity" (ISA,
section 3.4). Currently, there is insufficient information to adequately adjust for the potential
impact of longer-term exposure on respiratory ED visits associated with NO2 exposures, if any,
and this uncertainty should be kept in mind as one considers the results from the short-term
exposure NO2 risk assessment.
260
-------
9.5 BASELINE HEALTH EFFECTS INCIDENCE DATA
As illustrated in Equation 9-1, the most common health risk model based on air pollution
epidemiological studies expresses the reduction in health risk (Ay) associated with a given
reduction in NC>2 concentrations (Ax) as a percentage of the baseline incidence (y). To
accurately assess the impact of changes in NC>2 air quality on health risk in a given urban area,
information on the baseline incidence of health effects in that location is therefore needed. For
this assessment, baseline incidence is the incidence under recent ("as is") air quality conditions.
We obtained annual estimates of the baseline incidence of respiratory ED visits in
Atlanta, GA via personal communication with the authors of the study conducted in the Atlanta
area (Tolbert, 2007). Tolbert et al. (2007) notes that there are 42 hospitals with emergency
departments in the 20-county Atlanta MSA. Of these, 41 were able to provide incidence data for
at least part of the study period (1993 - 2004). For purposes of the NC>2 risk assessment, we
need incidences for the years of the risk assessment (2005 - 2007). Assuming that baseline
incidence of respiratory ED visits does not change appreciably in the span of a few years, we
have used the incidence of respiratory ED visits for the most recent year (i.e., 2004) in the
Tolbert et al. study, which was 121,818 respiratory ED visits.4 Because this baseline incidence
estimate is based on 36 hospitals, rather than the total 42 hospitals in Atlanta, this will be an
underestimate of baseline incidence. This is a source of downward bias in our estimates of NO2-
related risk. While there is some year-to-year variability in respiratory-related baseline incidence
(e.g., there were roughly 130,000 and 140,000 respiratory-related ED visits in 2002 and 2003,
respectively, in the Atlanta area), the estimate used for the risk assessment based on 2004 was
within 10% of the average for the most recent three year period available.
Average daily baseline incidences, necessary for short-term daily concentration-response
functions, were calculated by dividing the annual incidence by the number of days in the year for
which the baseline incidences were obtained. To the extent that NC>2 affects health, however,
actual incidence rates would be expected to be somewhat higher than average on days with high
NC>2 concentrations; using an average daily incidence would therefore result in underestimating
the changes in incidence on such days. Similarly, actual incidence rates would be expected to be
4 The specific definition of "respiratory-related" emergency department visits used in Tolbert et al. (2007) included
visits with the following respiratory illnesses as the primary diagnosis (specified by ICD-9 diagnostic codes):
asthma (493, 786.07, and 786.09), COPD (491, 492, and 496), upper respiratory illness (460 - 465, 460.0, and 477),
pneumonia (480 - 486), and bronchiolitis (466.1, 466.11, and 466.19).
261
-------
somewhat lower than average on days with low NC>2 concentrations; using an average daily
incidence would, therefore, result in overestimating the changes in incidence on low NC>2 days.
Both effects would be expected to be small, however, and should largely cancel one another out.
9.6 ADDRESSING UNCERTAINTY AND VARIABILITY
An important issue associated with any population health risk assessment is the
characterization of uncertainties and variability. Uncertainty refers to the lack of knowledge
regarding both the actual values of model input variables (parameter uncertainty) and the
physical systems or relationships (model uncertainty - e.g., the shape of the concentration-
response functions). In any risk assessment, uncertainty is, ideally, reduced to the maximum
extent possible, but significant uncertainty often remains. It can be reduced by improved
measurement and improved model formulation. In addition, the degree of uncertainty can be
characterized, sometimes quantitatively. For example, for the NC>2 risk assessment the statistical
uncertainty surrounding the estimated NC>2 coefficients in the concentration-response functions is
reflected in the confidence intervals provided for the risk estimates presented in this chapter and
in Appendix C. Additional uncertainties are discussed briefly below and in more detail in
Appendix C.
Variability refers to the heterogeneity in a population or variable of interest that is
inherent and cannot be reduced through further research. The current risk assessment for Atlanta
is based on locations-specific inputs (i.e., air quality data, baseline incidence data, and
concentration-response functions are for the Atlanta MSA). Variability in air quality data is
considered to some extent by the inclusion of thee years of data. Temporal variability is more
difficult to address, because the risk assessment focuses on some unspecified time in the future
when a given standard is just being met. To minimize the degree to which values of inputs to the
analysis may be different from the values of those inputs at that unspecified time: we have used
recent input data - for example, air quality data for the period 2005-2007 and baseline incidence
data for 2004. However, future changes in these inputs have not been predicted (e.g., future
population levels or changes in baseline incidence).
A number of important sources of uncertainty have been addressed qualitatively. Using a
similar approach to that described in section 7.8 for the air quality characterization and in section
8.12 for the exposure assessment in this document, staff have evaluated uncertainty in the
262
-------
respiratory-related ED visits risk assessment using an approach adapted from the recent
guidelines for qualitative uncertainty characterization (WHO, 2008). Uncertainties have been
qualitatively evaluated with respect to the level of uncertainty and the direction of bias. The level
of uncertainty was evaluated by considering the degree of severity of the uncertainly, implied by
the relationship between the source of the uncertainty and the output of the assessment. We have
used a designation of low, medium, and high as described in WHO (2008).
The bias direction indicates how the source of uncertainty was judged to influence
estimated ED visits associated with NO2 concentrations, either the estimated number or percent
of ED visits are likely "over-" or "under-estimated". In the instance where two or more types or
components of uncertainty result in offsetting direction of influence, the bias was judged as
"both". An "unknown" bias was assigned where there was no evidence reviewed to judge the
uncertainty associated with the source. Table 9-1 provides a summary of the sources of
uncertainty identified in the health risk assessment, the level of uncertainty, and the overall
judged bias of each. A brief summary discussion regarding those sources of uncertainty not
already examined in chapter 7 is included in the comments section of Table 9-1 and is elaborated
on in the bulleted points below.
• Causality. There is uncertainty about whether the association between NO2 and ED
visits actually reflects a causal relationship. Our judgment, drawing on the
conclusions in the ISA and as discussed in more detail in chapter 4, is that there is, at
a minimum, a likely causal relationship with either short-term NO2 itself or with NO2
serving as an indicator for itself and other components of ambient air associated with
combustion processes.
• Empirically estimated concentration-response relationships. In estimating the
concentration-response relationships, there are uncertainties: (1) surrounding
estimates of NO2 coefficients in concentration-response functions used in the
assessment, (2) concerning the specification of the concentration-response model
(including the shape of the relationships) and whether or not a population threshold or
non-linear relationship exists within the range of concentrations examined in the
studies, and (3) concerning the possible role of co-pollutants. The uncertainty
resulting from the statistical uncertainty associated with the estimated NO2 coefficient
in the concentration-response function has been characterized by confidence intervals
reflecting sample size. These confidence intervals do not reflect the uncertainties
related to the concentration-response functions, such as whether or not the model
263
-------
Table 9-1. Characterization of Key Uncertainties in the Emergency Department Visits Health Risk Assessment for the Atlanta Region.
Uncertainty
Level of
Uncertainty
Direction of Bias
Comments
Causality
low
Upward, if
causality
assumption isn't
true.
Statistical association does not prove causation. However, the risk
assessment considers only a health endpoint for which the overall weight
of the evidence supports the assumption that NO2 is likely causally
related based on the totality of the health effects evidence. If the
assumption of a causal relationship is incorrect, then a positive estimated
coefficient in the concentration-response function would be upward
biased, since it is greater than zero.
Empirically
estimated
concentration-
response
relations
medium
No obvious bias,
if concentration-
response model is
correctly
specified.
Otherwise,
unclear.
Because concentration-response functions are empirically estimated,
there is uncertainty surrounding these estimates. If the model is correctly
specified, there is no bias in the coefficient estimates. If the model is
mis-specified, there can be bias. Omitted confounding variables, for
example, could cause upward bias in the estimated NO2 coefficients if
the omitted variables are positively correlated with both NO2 and the
health effect. However, including potential confounding variables that
are highly correlated with one another can lead to unstable estimators.
Because both single- and multi-pollutant models were available, both
were used.
Functional form
of concentration-
response relation
medium
Unclear
Statistical significance of coefficients in an estimated concentration-
response function does not necessarily mean that the mathematical form
of the function is the best model of the true concentration-response
relation. If the "true" functional relationship between NO2 and a health
effect is different from the one specified, there can be bias in the
resulting estimates of effect. The direction of the bias will depend on
how the specified model differs from reality. For example, if the
specified concentration-response function is log-linear down to 0 ug/m3,
but there is actually a threshold in the true relationship, then the effect
will be overstated by the model corresponding to levels of NO2 below the
threshold.
264
-------
Uncertainty
Level of
Uncertainty
Direction of Bias
Comments
Lag structure of
concentration-
response relation
low
Downward, if
important lags are
omitted (e.g., if
concentration-
response function
includes a single
lag, while "truth"
is a distributed
lag).
Unclear, if
concentration-
response function
includes a single
lag, but it's the
wrong lag.
The actual lag structure for short-term NO2 exposures is uncertain.
Omitted lags could cause an underestimation in the predicted incidence
associated with a given reduction in NO2 concentrations. The level of
uncertainty (in the sense of the impact of the uncertainty) may depend on
the situation. For example, suppose the health effect is actually affected
largely by same-day NO2 concentrations but the model (incorrectly)
includes only a 1-day lag. In this case, the impact on the outcome of the
analysis may be minimal if, as is likely, there is a high degree of
autocorrelation in NO2 concentrations from day to day (so that
yesterday's NO2 level would act as a good proxy for today's NO2 level).
If, on the other hand, there is a distributed lag - e.g., if risk of the health
effect on day t depends on NO2 concentrations for the entire week
leading up to day t - and the model includes only a single lag, then the
understatement of effect could be substantial.
Transferability
of concentration-
response
relations
low
No obvious bias.
Concentration-response functions may not provide an adequate
representation of the concentration-response relationship in times and
places other than those in which they were estimated. For example,
populations in the assessment location/time period may have more or
fewer members of sensitive subgroups than the location/time period in
which functions were derived, which would introduce additional
uncertainty related to the use of a given concentration-response function
in the analysis. This problem was minimized in the NO2 risk assessment,
however, because it relies on concentration-response functions estimated
in a recent study conducted in the assessment location.
265
-------
Uncertainty
Level of
Uncertainty
Direction of Bias
Comments
Extrapolation of
concentration-
response
relations beyond
the range of
observed NO2
data
low
Unclear.
A concentration-response relationship estimated by an epidemiological
study may not be valid at concentrations outside the range of
concentrations observed during the study. This problem should be
minimal in the NO2 risk assessment, however, because the NO2
concentrations observed in the study from which C-R functions were
obtained covered a wide range - from 1 ppb to 181 ppb.
Adequacy of
ambient NO2
monitors as
surrogate for
population
exposure
low
No obvious bias.
Possible differences in how the spatial variation in ambient NO2 levels
across an urban area are characterized in the original epidemiological
study compared to the more recent ambient NO2 data used to characterize
current air quality would contribute to uncertainty in the health risk
estimates. The NO2 risk assessment uses the same monitor used in the
epidemiological study from which the C-R functions were obtained,
which should minimize this source of uncertainty.
Adjustment of
air quality
distributions to
simulate just
meeting current
and alternative
NO2 standards.
medium to
high
No obvious bias.
The pattern and extent of daily reductions in NO2 concentrations that
would result if the current NO2 standard or alternative NO2 standards
were just met is not known. There remains significant uncertainty about
the shape of the air quality distribution of hourly levels upon just meeting
an NO2 standard, especially for alternative standards that are at levels
higher than recent NO2 air quality levels.
Baseline health
effects data
Low-
medium
Downward bias.
Data on baseline incidence may be uncertain for a variety of reasons.
For example, location- and age-group-specific baseline rates may not be
available in all cases. Baseline incidence may change over time for
reasons unrelated to NO2. This source of uncertainty is relatively minor
in the NO2 risk assessment, however, because a baseline incidence
estimate has been obtained from the study authors for the assessment
area. There is a known downward bias to this estimate, however,
because it is based on an incomplete set of hospitals providing ED data
(36 out of 42) in the Atlanta MSA.
266
-------
used in the epidemiological study is the correct model form. With respect to
uncertainties about model form and whether or not a population threshold exists, the
available epidemiological studies neither support nor refute the existence of
thresholds at the population level. Concerning the possible role of co-pollutants in
the Tolbert et al. (2007) study, NO2 was only moderately correlated with the other
pollutants considered (i.e., PMio, O^) that produced the concentration-response
functions that have been used in the risk assessment, although it was fairly highly
correlated (r = 0.7) with CO. When a study, such as Tolbert et al. (2007) is conducted
in a single location, the problem of possible confounding is particularly difficult.
Single-pollutant models, which omit co-pollutants, may produce overestimates of the
NC>2 effect, if some of the effects are really due to one or more of the other pollutants.
On the other hand, effect estimates based on a multi-pollutant model can be uncertain
and even result in statistically insignificant estimates where there is a true
relationship, if the co-pollutants included in the model are highly correlated with
NO2. As a result of these considerations, we report risk estimates based on both the
single- and multi-pollutant models from Tolbert et al. (2007). It should be noted that
use of a concentration-response relationship based on an epidemiological study
conducted in the same location for this risk assessment reduces some potential
uncertainties since it does not involve extrapolation of the relationship across
different geographic areas with different population characteristics, land uses, source
mixtures and other factors.
• Adequacy of ambient NO? monitors as surrogate for population exposure. The
Tolbert et al. (2007) study used ambient concentrations at fixed-site monitors to
represent ambient exposure and for several reasons this may or may not provide a
good representation of ambient NO2 exposure for the population. The final ISA
identifies the following thee components to exposure measurement error: (1) the use
of average population rather than individual exposure data; (2) the difference between
average personal ambient exposure and ambient concentrations at central monitoring
sites; and (3) the difference between true and measured ambient concentrations (final
ISA, section 1.3.2, p. 1-5). While a concentration-response function may understate
the effect of personal exposure to NO2 on the incidence of a health effect, it will give
an unbiased estimate of the effect of ambient concentrations on the incidence of the
health effect, if the ambient concentrations at monitoring stations provide an unbiased
estimate of the ambient concentrations to which the population is exposed. If NO2 is
the causal agent, the understatement of the impact of personal exposures is not a
concern, since NO2 NAAQS are expressed in terms of ambient, not personal
exposure, levels. However, if NO2 is not the causal agent, and the effects are due to
confounding copollutants or other factors, then reducing ambient NO2 levels might
not result in the estimated reductions in the health effects.
• Adjustment of air quality distributions to simulate just meeting the current annual
standard and alternative 98th and 99th percentile daily maximum 1-h standards. The
current annual standard and many of the alternative 1-h standards analyzed in the
current risk assessment requires an upward adjustment of recent ambient NO2 levels.
In adjusting air quality to simulate just meeting these standards, we have assumed that
the overall shape of the distribution of 1-h and 24-h concentrations would not change.
While we believe this is a reasonable assumption in the absence of evidence
267
-------
supporting a change in the distribution, we recognize this as an important additional
uncertainty, especially for those scenarios where considerable upward adjustment is
required to simulate just meeting some of the standards.
Baseline incidence. There are uncertainties related to the baseline incidence
including: (1) the extent to which baseline incidence varies between the year used in
the assessment (i.e., 2004) and some unspecified future year when air quality is
adjusted to simulate just meeting the current and alternative standards; (2) the extent
to which baseline incidence is underestimated because only 36 of the 42 emergency
departments provided baseline incidence for the study in 2004; (3) the use of annual
incidence data to develop daily baseline incidence; and (4) the extent to which
Atlanta area residents visited emergency departments outside of the Atlanta MSA.
As noted previously, the use of the available baseline incidence for 2004 results in
some underestimation of the risk for the Atlanta MSA since data were only available
from 36 of the 42 emergency departments for that year (i.e., about 14% of emergency
departments were not included). Concerning the use of annual baseline incidence to
estimate daily incidence, to the extent that NO2 affects health, actual incidence would
be expected to be somewhat higher than average on days with high NO2
concentrations and using an average daily incidence would result in underestimating
the changes in incidence on such days. Similarly, actual incidence would be expected
to be somewhat lower on days with low NO2 concentrations and using an average
daily incidence would result in overestimating the changes in incidence on such days.
Both of these effects would be expected to be small and should largely cancel each
other out. With respect to the last uncertainty, we consider this to be a relatively
minor uncertainty since most ED visits are likely to be made to the closest emergency
department available, which, for residents of the Atlanta MSA are likely to be within
that MSA. The baseline incidence data has not been adjusted for any future changes
268
-------
such as aging of the population over time or possible changes in ED visits due to
increased in-migration of younger individuals.
9.7 RISK ESTIMATES FOR EMERGENCY DEPARTMENT VISITS
In this section, we present risk estimates associated with several air quality scenarios,
including three recent years of air quality as represented by 2005, 2006, and 2007 monitoring
data. In addition, risk estimates are presented for a hypothetical scenario, where air quality from
2006 and 2007 is adjusted upward to simulate just meeting the current annual NC>2 standard, and
for scenarios where the three year period (2005-2007) is adjusted (either up or down) to simulate
just meeting potential alternative 98th and 99th percentile daily maximum 1-h standards. As
discussed previously in chapter 5, potential alternative 1-h standards with levels set at 0.05, 0.10,
0.15, and 0.20 have been included in the risk assessment.
Throughout this section and Appendix C the uncertainty surrounding risk estimates
resulting from the statistical uncertainty of the NCh coefficients in the concentration-response
functions used is characterized by ninety-five percent confidence intervals around estimates of
incidence, incidence per 100,000 population, and percent of total incidence that is NO2-related.
In some cases, the lower bound of a confidence interval falls below zero. This does not imply
that additional exposure to NC>2 has a beneficial effect but only that the estimated coefficient in
the concentration-response function was not statistically significantly different from zero. Lack
of statistical significance could reflect insufficient statistical power to detect a relationship that
exists or could reflect that no relationship exists.
Tables 9-2, 9-3, and 9-4 present the risk estimates for NCVrelated ED visits associated
with recent air quality (2005, 2006, and 2007, respectively). Table 9-2 for 2005 also includes
risk estimates for just meeting several alternative 1-h daily maximum standards based on
adjusting 2005-2007 air quality data to simulate just meeting these alternative standards.
Similarly, Tables 9-3 (based on 2006) and 9-4 (based on 2007) include risk estimates associated
with just meeting these same alternative 1-h standards, as well as risk estimates associated with a
simulation where air quality is adjusted upward to represent just meeting the current 0.053 ppm
annual NC>2 standard. Since attainment of the current annual standard is based on the most
recent two year period, risk estimates for the annual standard are only included in the tables
based on 2006 and 2007 air quality.
269
-------
Table 9-2. Estimated Percent of Total Annual Incidence of Respiratory ED Visits Associated with "As Is" NO2 Concentrations and NO2
Concentrations that Just Meet Alternative Standards in Atlanta, GA, Based on Adjusting 2005 NO2 Concentrations.*
Other
Pollutants in
Model
none
CO
03
PM10
PM10, O3
Percent of Total Incidence of Respiratory Emergency Department Visits Associated with "As is" NO2 Concentrations and NO2 Concentrations that
Just Meet Alternative Standards**
"as is"
3%
(1.6% -4. 3%)
2.5%
(0.8% - 4.2%)
1.5%
(-0.1% -3.1%)
1.1%
(-0.6% -2.7%)
0.6%
(-1.1% -2. 3%)
Alternative 98th percentile 1-hr daily maximum standards
(ppm)
0.05***
2.1%
(1.1% -3.1%)
1.8%
(0.6% - 3%)
1.1%
(0% - 2.2%)
0.8%
(-0.4% -1.9%)
0.4%
(-0.8% -1.7%)
0.1
4.2%
(2.2% -6.1%)
3.6%
(1.2% -5.9%)
2.1%
(-0.1% -4. 3%)
1.5%
(-0.9% -3. 8%)
0.9%
(-1.6% -3. 3%)
0.15
6.2%
(3.3% - 8.9%)
5.3%
(1.8% -8.7%)
3.2%
(-0.1% -6. 3%)
2.2%
(-1.3% -5.6%)
1.3%
(-2.5% -4. 9%)
0.2
8.1%
(4.4% -11. 7%)
7%
(2.4% -11. 3%)
4.2%
(-0.2% - 8.4%)
3%
(-1.7% -7.4%)
1.7%
(-3.3% -6.4%)
Alternative 99th percentile 1-hr daily maximum standards
(ppm)
0.05
2%
(1%-2.9%)
1.7%
(0.6% - 2.8%)
1%
(0% - 2%)
0.7%
(-0.4% -1.8%)
0.4%
(-0.8% -1.5%)
0.1
3.9%
(2.1% -5. 7%)
3.3%
(1.1% -5. 5%)
2%
(-0.1% -4%)
1.4%
(-0.8% - 3.5%)
0.8%
(-1.5% -3%)
0.15
5.8%
(3.1% -8. 3%)
4.9%
(1.7% -8.1%)
2.9%
(-0.1% -5.9%)
2.1%
(-1.2% -5.2%)
1.2%
(-2. 3% -4. 5%)
0.2
7.6%
(4.1% -10. 9%)
6.5%
(2.2% -10.6%)
3.9%
(-0.2% -7.8%)
2.8%
(-1.6% -6.9%)
1 .6%
(-3.1% -6%)
*Estimated incidences of respiratory emergency department visits are based on the concentration-response functions estimated in Tolbert et al. (2007) [results
corresponding to Figure 2 in Tolbert et al. (2007) were obtained via personal communication with P. Tolbert]. All models use a 3-day moving average of the
daily 1-hr, maximum NO2 concentration and apply to all ages.
**lncidence was quantified down to 0 ppb. Percents are rounded to the nearest tenth.
***Alternative 1-hr daily maximum standards are characterized by a concentration of m ppm and an nth percentile, requiring that the average of the 3 annual
nth percentile 1-hr daily maxima over a 3-year period be at or below m ppm.
Note: Numbers in parentheses are 95% confidence intervals based on statistical uncertainty surrounding the NO2 coefficient.
270
-------
Table 9-3. Estimated Percent of Total Annual Incidence of Respiratory ED Visits Associated with "As Is" NO2 Concentrations and NO2
Concentrations that Just Meet Alternative Standards in Atlanta, GA, Based on Adjusting 2006 NO2 Concentrations.*
Other
Pollutants in
Model
none
CO
03
PM10
PM10, 03
Percent of Total Incidence of Respiratory Emergency Department Visits Associated with "As is" NO2 Concentrations and NO2 Concentrations that Just Meet the
Current and Alternative Standards**
"as is"
3.1%
(1 .6% - 4.5%)
2.6%
(0.9% - 4.4%)
1 .6%
(-0.1% -3.2%)
1.1%
(-0.6% - 2.8%)
0.6%
(-1 .2% - 2.4%)
current annual
standard
9%
(4.9% -12.9%)
7.7%
(2.6% -12.5%)
4.6%
(-0.2% - 9.2%)
3.3%
(-1 .9% - 8.2%)
1 .9%
(-3.6% -7.1%)
Alternative 98th percentile 1-hr daily maximum standards
(ppm)
0.05***
2.2%
(1 .2% - 3.2%)
1 .9%
(0.6% -3.1%)
1.1%
(-0.1% -2.3%)
0.8%
(-0.4% - 2%)
0.4%
(-0.8% - 1 .7%)
0.1
4.3%
(2.3% - 6.3%)
3.7%
(1.2% -6.1%)
2.2%
(-0.1% -4.5%)
1 .6%
(-0.9% - 3.9%)
0.9%
(-1 .7% - 3.4%)
0.15
6.4%
(3.5% - 9.3%)
5.5%
(1 .8% - 9%)
3.3%
(-0.2% - 6.6%)
2.3%
(-1 .3% - 5.8%)
1 .3%
(-2.5% -5.1%)
0.2
8.5%
(4.6% -12.2%)
7.3%
(2.5% - 1 1 .8%)
4.4%
(-0.2% - 8.7%)
3.1%
(-1 .8% - 7.7%)
1 .8%
(-3.4% - 6.7%)
Alternative 99th percentile 1-hr daily maximum standards
(ppm)
0.05
2%
(1.1% -3%)
1 .7%
(0.6% - 2.9%)
1%
(0%-2.1%)
0.7%
(-0.4% - 1 .8%)
0.4%
(-0.8% - 1 .6%)
0.1
4%
(2.2% - 5.9%)
3.4%
(1 .2% - 5.7%)
2.1%
(-0.1% -4.1%)
1 .4%
(-0.8% - 3.7%)
0.8%
(-1 .6% - 3.2%)
0.15
6%
(3.2% - 8.7%)
5.1%
(1 .7% - 8.4%)
3.1%
(-0.1% -6.2%)
2.2%
(-1 .2% - 5.4%)
1 .2%
(-2.4% - 4.7%)
0.2
7.9%
(4.3% - 1 1 .4%)
6.8%
(2.3% -11%)
4.1%
(-0.2% -8.1%)
2.9%
(-1 .7% - 7.2%)
1 .6%
(-3.2% - 6.2%)
*Estimated incidences of respiratory emergency department visits are based on the concentration-response functions estimated in Tolbert et al. (2007) [results corresponding
to Figure 2 in Tolbert et al. (2007) were obtained via personal communication with P. Tolbert]. All models use a 3-day moving average of the daily 1-hr, maximum NO2
concentration and apply to all ages.
"Incidence was quantified down to 0 ppb. Percents are rounded to the nearest tenth.
"""Alternative 1-hr daily maximum standards are characterized by a concentration of m ppm and an nth percentile, requiring that the average of the 3 annual nth percentile 1-
hr daily maxima over a 3-year period be at or below m ppm.
Note: Numbers in parentheses are 95% confidence intervals based on statistical uncertainty surrounding the NO2 coefficient.
271
-------
Table 9-4. Estimated Percent of Total Annual Incidence of Respiratory ED Visits Associated with "As Is" NO2 Concentrations and NO2
Concentrations that Just Meet Alternative Standards in Atlanta, GA, Based on Adjusting 2007 NO2 Concentrations.*
Other
Pollutants in
Model
none
CO
03
PM10
PM10, 03
Percent of Total Incidence of Respiratory Emergency Department Visits Associated with "As is" NO2 Concentrations and NO2 Concentrations that Just Meet the
Current and Alternative Standards**
"as is"
2.8%
(1 .5% - 4%)
2.4%
(0.8% - 3.9%)
1 .4%
(-0.1% -2.8%)
1%
(-0.6% - 2.5%)
0.6%
(-1.1% -2.2%)
current annual
standard
8.1%
(4.4% - 1 1 .6%)
6.9%
(2.3% - 1 1 .3%)
4.1%
(-0.2% - 8.3%)
2.9%
(-1 .7% - 7.3%)
1 .7%
(-3.2% - 6.4%)
Alternative 98th percentile 1-hr daily maximum standards
(ppm)
0.05***
2%
(1%-2.9%)
1 .7%
(0.6% - 2.8%)
1%
(0% - 2%)
0.7%
(-0.4% - 1 .8%)
0.4%
(-0.8% - 1 .5%)
0.1
3.9%
(2.1% -5.7%)
3.3%
(1.1% -5.5%)
2%
(-0.1% -4%)
1 .4%
(-0.8% - 3.5%)
0.8%
(-1 .5% - 3%)
0.15
5.8%
(3.1% -8.4%)
4.9%
(1.7% -8.1%)
2.9%
(-0.1% -5.9%)
2.1%
(-1 .2% - 5.2%)
1 .2%
(-2.3% - 4.5%)
0.2
7.6%
(4.1% -11%)
6.5%
(2.2% -10.6%)
3.9%
(-0.2% - 7.8%)
2.8%
(-1 .6% - 6.9%)
1 .6%
(-3% - 6%)
Alternative 99th percentile 1-hr daily maximum standards
(ppm)
0.05
1 .8%
(1%-2.7%)
1 .6%
(0.5% - 2.6%)
0.9%
(0% - 1 .9%)
0.6%
(-0.4% - 1 .7%)
0.4%
(-0.7% - 1 .4%)
0.1
3.6%
(1 .9% - 5.3%)
3.1%
(1%-5.1%)
1 .8%
(-0.1% -3.7%)
1 .3%
(-0.7% - 3.3%)
0.7%
(-1 .4% - 2.8%)
0.15
5.4%
(2.9% - 7.8%)
4.6%
(1 .5% - 7.5%)
2.7%
(-0.1% -5.5%)
1 .9%
(-1.1% -4.9%)
1.1%
(-2.1% -4.2%)
0.2
7.1%
(3.8% -10.2%)
6.1%
(2% - 9.9%)
3.6%
(-0.2% - 7.3%)
2.6%
(-1 .5% - 6.4%)
1 .5%
(-2.8% - 5.6%)
*Estimated incidences of respiratory emergency department visits are based on the concentration-response functions estimated in Tolbert et al. (2007) [results corresponding
to Figure 2 in Tolbert et al. (2007) were obtained via personal communication with P. Tolbert]. All models use a 3-day moving average of the daily 1-hr, maximum NO2
concentration and apply to all ages.
"Incidence was quantified down to 0 ppb. Percents are rounded to the nearest tenth.
""""Alternative 1-hr daily maximum standards are characterized by a concentration of m ppm and an nth percentile, requiring that the average of the 3 annual nth percentile 1-
hr daily maxima over a 3-year period be at or below m ppm.
Note: Numbers in parentheses are 95% confidence intervals based on statistical uncertainty surrounding the NO2 coefficient.
272
-------
In Table 9-2, and similarly in Tables 9-3 and 9-4, the first row of percent of total annual
incidence estimates is based on a single pollutant model (i.e., NO2 only) and results in the largest
estimates for NO2-related respiratory ED visits. The next three rows present risk estimates based
on two pollutant models (i.e., NO2 + CO, NO2 + 63, NO2 + PMio). The last row presents risk
estimates based on a three pollutant model (i.e., NO2 + PMio + Os). As noted above in this
chapter, effect estimates based on a multi-pollutant model can be uncertain and even result in
statistically insignificant estimates where there is a true relationship, if the co-pollutants included
in the model are highly correlated with NO2. In the case of this study in Atlanta, NO2 was
moderately correlated with PMio and 63 concentrations. The negative lower bounds of the
confidence intervals for many of the risk estimates based on multi-pollutant models may in part
be due to correlation in these pollutant concentrations and staff do not view these estimates as
suggesting any health beneficial health effect of increasing NO2 exposure levels.
Tables 4-1, 4-2, and 4-3 in Appendix C present these same results expressed in terms of
incidence of respiratory-related ED visits in the Atlanta MSA based on recent air quality and just
meeting altertative standards based on 2005, 2006, and 2007 air quality data. Finally, Tables 4-
4, 4-5, and 4-6 in Appendix C present these same risk estimates expressed in terms of incidence
per 100,000 general population in the Atlanta MSA based on recent air quality and simulating
just meeting alternative standards based on the same three years of air quality data.
Key Observations
Presented below are key observations resulting from the respiratory-related ED visits risk
assessment:
• Respiratory-related ED visits estimated to result from exposures to NO2 were
estimated for a single urban area (i.e., Atlanta) for several recent years of air
quality (2005-2007) and for air quality adjusted to simulate just meeting the
current annual NO2 standard and several alternative 1-hour daily maximum NO2
standards. While we would expect some differences in estimated NO2-related ED
respiratory visits across different locations due to differences in populations, land
use patterns, access to medical facilities, co-pollutants and other factors affecting
exposure and the concentration-response relationships, we believe that the risk
estimates do provide a useful perspective on the likely overall magnitude and
pattern of ED visits associated with various NO2 air quality scenarios in urban
areas within the U.S.
• The largest risk estimates were associated with single-pollutant NO2
concentration-response functions based on the effect estimates reported in Tolbert
et al. (2007). Risk estimates based on various co-pollutant models with Os, CO,
273
-------
and PMio resulted in significant reduction in the risk estimates, often by a factor
of two or greater and resulted in much wider confidence intervals.
• The only standards that resulted in a reduction in risk estimates from the baseline
of recent air quality for the three year period examined were the 98th and 99th
percentile 1-hour daily maximum standards set at the level of 0.05 ppm.
• The impact of changing the level of the alternative 1-hour daily maximum
standards is substantially greater than the impact of changing from a 98th to a
99th percentile standard. For example, changing from a 98th percentile 1-hour
daily maximum standard set at 0.10 ppm to one set at 0.05 ppm reduces the
estimated incidence of respiratory-related ED visits in Atlanta by about 49 percent
in 2007 (from 4700 to 2400); however, changing from a 98th percentile 1-hour
daily maximum standard based on 0.05 ppm to a 99th percentile 1-hour daily
maximum standard based on 0.05 reduces the incidence in 2007 by only about 8
percent (from 2400 to 2200).
• The overall pattern of risk estimates is similar across the three years examined.
For the three years examined, there was not significant year-to-year variability in
the risk estimates.
• Important uncertainties and limitations associated with the risk assessment which
were discussed above in section 9.6 and which should be kept in mind as one
considers the quantitative risk estimates include:
- uncertainty about the extent to which the associations between NC>2 and ED
visits for respiratory causes actually reflect causal relationships;
- statistical uncertainty due to sampling error which is characterized in the
assessment;
- uncertainties associated with the air quality adjustment procedure that was
used to simulate just meeting the current annual and several alternative 1-h
daily maximum standards;
-uncertainties associated with the estimated baseline incidence for ED
respiratory visits;
- uncertainties related to how changes in population, activity patterns, air
quality, and other factors over time might impact the risk estimates;
- there is uncertainty about the extent to which the risk estimates presented
for the Atlanta urban area are representative of other urban locations in the
U.S.
274
-------
10. EVIDENCE-AND EXPOSURE/RISK-BASED
CONSIDERATIONS RELATED TO THE PRIMARY NO2
NAAQS
10.1 INTRODUCTION
This chapter considers the scientific evidence in the ISA (EPA, 2008a) and the
exposure and risk characterization results presented in this document as they relate to the
adequacy of the current NC>2 primary NAAQS and potential alternative primary NC>2
standards. The available scientific evidence includes epidemiologic, controlled human
exposure, and animal toxicological studies. The NC>2 exposure and risk characterizations
described in chapters 6-9 of this document include estimates of exposures and health
risks associated with recent NC>2 concentrations and with NC>2 concentrations adjusted to
simulate scenarios just meeting the current and potential alternative standards. In
considering the scientific evidence and the exposure- and risk-based information, we
have also considered relevant uncertainties. Section 10.2 of this chapter presents our
general approach to considering the adequacy of the current standard and potential
alternative standards. Section 10.3 focuses on evidence- and exposure-/risk-based
considerations related to the adequacy of the current standard, and section 10.4 focuses
on such considerations related to potential alternative standards (in terms of the indicator,
averaging time, form, and level).
These considerations are intended to inform the Agency's policy assessment of a
range of options with regard to the NO2 NAAQS. We note that the final decision on
retaining or revising the current NC>2 primary standard, taking into account the Agency's
policy assessment, is largely a public health policy judgment. A final decision will draw
upon scientific information and analyses about health effects, population exposure and
risks, and policy judgments about the appropriate response to the range of uncertainties
that are inherent in the scientific evidence and analyses. Our approach to informing these
judgments, discussed more fully below, is based on a recognition that the available health
effects evidence reflects a continuum consisting of ambient levels at which scientists
generally agree that health effects are likely to occur through lower levels at which the
likelihood and magnitude of the response become increasingly uncertain. This approach
275
-------
is consistent with the requirements of the NAAQS provisions of the Act and with how
EPA and the courts have historically interpreted the Act. These provisions require the
Administrator to establish primary standards that, in the Administrator's judgment, are
requisite to protect public health with an adequate margin of safety. In so doing, the
Administrator seeks to establish standards that are neither more nor less stringent than
necessary for this purpose. The Act does not require that primary standards be set at a
zero-risk level but rather at a level that avoids unacceptable risks to public health,
including the health of sensitive groups.
10.2 GENERAL APPROACH
This section describes the general approach that staff is taking to inform decisions
regarding the need to retain or revise the current NC>2 NAAQS. The current standard,
which is an annual average of 0.053 ppm, was retained by the Administrator in the most
recent review in 1996 (61 FR 52854 (October 8, 1996)). The decision in that review to
retain the annual standard was based on consideration of available scientific evidence for
health effects associated with NO2 and on air quality information. With regard to these
considerations, the Administrator noted that "a 0.053 ppm annual standard would keep
annual NC>2 concentrations considerably below the long-term levels for which serious
chronic effects have been observed in animals" and that "Retaining the existing standard
would also provide protection against short-term peak NO2 concentrations at the levels
associated with mild changes in pulmonary function and airway responsiveness observed
in controlled human studies" (60 FR 52874, 52880 (Oct. 11, 1995)). As a result, the
Administrator concluded that "the existing annual primary standard appears to be both
adequate and necessary to protect human health against both long- and short-term NO2
exposures" and that "retaining the existing annual standard is consistent with the
scientific data assessed in the Criteria Document (U.S. EPA, 1993) and the Staff Paper
(U.S. EPA, 1995a) and with the advice and recommendations of CASAC" (61 FR 52852
at 52854).
To inform the range of options that the Agency will consider in this review of the
current primary NO2 standard, the general approach we have adopted builds upon the
approaches used in reviews of other criteria pollutants, including the most recent reviews
276
-------
of the Pb, O3, and PM NAAQS (EPA, 2008c; EPA, 2007h; EPA, 2005). As in these
other reviews, we consider the implications of placing more or less weight or emphasis
on different aspects of the scientific evidence and the exposure/risk-based information,
recognizing that the weight to be given to various elements of the evidence and
exposure/risk information is part of the public health policy judgments that the
Administrator will make in reaching decisions on the standard.
A series of general questions frames our approach to considering the scientific
evidence and exposure/risk-based information. First, our consideration of the scientific
evidence and exposure/risk-based information with regard to the adequacy of the current
standard is framed by the following questions:
• To what extent does evidence and exposure/risk-based information that has
become available since the last review reinforce or call into question evidence for
NO2-associated effects that were identified in the last review?
• To what extent has evidence for different health effects and/or sensitive
populations become available since the last review?
• To what extent have uncertainties identified in the last review been reduced
and/or have new uncertainties emerged?
• To what extent does evidence and exposure/risk-based information that has
become available since the last review reinforce or call into question any of the
basic elements of the current standard?
To the extent that the available evidence and exposure/risk-based information suggests it
may be appropriate to consider revision of the current standard, we consider that evidence
and information with regard to its support for consideration of a standard that is either
more or less protective than the current standard. This evaluation is framed by the
following questions:
• Is there evidence that associations, especially causal or likely causal associations,
extend to ambient NO2 concentrations as low as, or lower than, the concentrations
that have previously been associated with health effects? If so, what are the
important uncertainties associated with that evidence?
• Are exposures above benchmark levels and/or health risks estimated to occur in
areas that meet the current standard? If so, are the estimated exposures and health
277
-------
risks important from a public health perspective? What are the important
uncertainties associated with the estimated risks?
To the extent that there is support for consideration of a revised standard, we then
consider the specific elements of the standard (indicator for gaseous NOX, averaging time,
form, and level) within the context of the currently available information. In so doing,
we address the following questions:
• Does the evidence provide support for considering a different indicator for
gaseous NOX?
• Does the evidence provide support for considering different averaging times?
• What ranges of levels and forms of alternative standards are supported by the
evidence, and what are the associated uncertainties and limitations?
• To what extent do specific averaging times, levels, and forms of alternative
standards reduce the estimated exposures above benchmark levels and risks
attributable to NC>2, and what are the uncertainties associated with the estimated
exposure and risk reductions?
The following discussion addresses the questions outlined above and presents
staffs conclusions regarding the scientific evidence and the exposure-/risk-based
information specifically as they relate to the current and potential alternative standards.
This discussion is intended to inform the Agency's consideration of policy options that
will be presented in an Advanced Notice of Proposed Rulemaking (ANPR), together with
the scientific support for such options, and which will be further considered in the
Agency's proposed and final rule-making notices. Section 10.3 considers the adequacy
of the current standard while section 10.4 considers potential alternative standards in
terms of indicator, averaging time, form, and level. Each of these sections considers key
conclusions as well as the uncertainties associated with the evidence and/or exposure/risk
analyses.
10.3 ADEQUACY OF THE CURRENT ANNUAL STANDARD
In the last review of the NO2 NAAQS, the AQCD for NOX concluded that there
were two key health effects of greatest concern at ambient or near-ambient concentrations
of NC>2 (ISA, section 5.3.1). The first was increased airway hyperresponsiveness in
278
-------
asthmatic individuals after short-term exposures. The second was increased respiratory
illness among children associated with longer-term exposures to NC>2. Evidence also was
found for increased risk of emphysema, but this appeared to be of major concern only
with exposures to levels of NC>2 that were much higher than current ambient levels (ISA,
section 5.3.1). Controlled human exposure and animal toxicological studies provided
qualitative evidence for airway hyperresponsiveness and lung function changes while
epidemiologic studies provided evidence for increased respiratory symptoms with
increased indoor NC>2 exposures. Animal toxicological findings of lung host defense
system changes with NC>2 exposure provided a biologically-plausible basis for the
epidemiologic results. Subpopulations considered potentially more susceptible to the
effects of NO2 exposure included persons with preexisting respiratory disease, children,
and the elderly. The epidemiologic evidence for respiratory health effects was limited,
and no studies had considered effects such as hospital admissions, ED visits, or mortality
(ISA, section 5.3.1).
10.3.1 Evidence-based considerations
Evidence published since the last review generally has confirmed and extended
the conclusions articulated in the 1993 AQCD (ISA, section 5.3.2). The epidemiologic
evidence has grown substantially with the addition of field and panel studies, intervention
studies, time-series studies of effects such as hospital admissions, and a substantial
number of studies evaluating mortality risk associated with short-term NC>2 exposures.
As noted above, no epidemiologic studies were available in 1993 that assessed
relationships between NC>2 and outcomes such as hospital admissions, ED visits, or
mortality. In contrast, dozens of epidemiologic studies on such outcomes are now
included in this evaluation (ISA, chapter 3). While not as marked as the growth in the
epidemiologic literature, a number of recent toxicological and human clinical studies also
provide insights into relationships between NO2 exposure and health effects.
In considering this evidence, we note that different scientific methodologies
provide different types of information. For example, controlled human exposure studies
provide information on health effects that are specifically associated with exposure to
NC>2 in the absence of the co-pollutants that are commonly found in ambient air.
279
-------
However, these studies do not provide information directly related to the public health
implications of real-world NC>2 air quality. Epidemiologic studies provide information on
the public health implications of real-world NO2 concentrations; however, interpretation
of specific NO2-related effects in these studies is complicated by a number of factors,
including the presence of co-pollutants in the ambient air.
As an initial consideration with regard to the adequacy of the current standard,
staff notes that the evidence relating long-term (weeks to years) NC>2 exposures to
adverse health effects is judged to be either "suggestive but not sufficient to infer a causal
relationship" (respiratory morbidity) or "inadequate to infer the presence or absence of a
causal relationship" (mortality, cancer, cardiovascular effects,
reproductive/developmental effects) (ISA, sections 5.3.2.4-5.3.2.6). In contrast, the
evidence relating short-term (minutes to hours) NC>2 exposures to respiratory morbidity is
judged to be "sufficient to infer a likely causal relationship" (ISA, section 5.3.2.1). This
judgment is supported primarily by a large body of recent epidemiologic evidence that
evaluates associations of short-term NC>2 concentrations with respiratory symptoms, ED
visits, and hospital admissions. It suggests that, at a minimum, consideration of the
adequacy of the current annual standard should take into account the extent to which that
standard provides protection against respiratory effects associated with short-term NC>2
exposures. Such an emphasis on health endpoints for which evidence has been judged
sufficient to infer a likely causal relationship would be consistent with other recent
NAAQS reviews (e.g., EPA, 2005; EPA, 2007h; EPA, 2007i) and would ensure that
decisions are based on endpoints for which a causal relationship with NC>2 is judged to be
"more likely than not" (ISA, Table 1.3-2).
Because there was concern in the 1996 review of the NC>2 NAAQS about the
potential for respiratory effects associated with short-term exposure to NC>2
concentrations around 0.2 ppm, the extent to which the then-current standard (which
remains the current standard for purposes of this review) could be expected to afford
protection from NO2 concentrations at this level was considered. In that review, the issue
was examined with an air quality analysis that evaluated 1-h NC>2 concentrations. The
conclusion from that analysis was that locations meeting the current standard are unlikely
280
-------
to experience 1-h concentrations exceeding levels (e.g., 0.2 ppm) that have been
associated with respiratory effects in controlled human exposure studies (EPA, 1995).
In the current review, a larger number of epidemiologic studies are available. In
considering these epidemiologic studies, we note that annual average NC>2 concentrations
were below the level of the current annual NC>2 NAAQS in many of the locations where
positive associations with respiratory morbidity endpoints have been detected (ISA,
section 5.4). With regard to these studies, we note that the ISA characterizes the
evidence for respiratory effects as consistent and coherent. The evidence is consistent in
that associations are reported in studies conducted in numerous locations and with a
variety of methodological approaches (ISA, section 5.3.2.1). It is coherent in the sense
that the studies report associations with respiratory health outcomes that are logically
linked together (ISA, section 5.3.2.1). When the epidemiologic literature is considered as
a whole, there are generally positive associations between NC>2 and respiratory
symptoms, hospital admissions, and ED visits. A number of these associations are
statistically significant, particularly the more precise effect estimates (ISA, section
5.3.2.1).
Interpretation of these NC>2 epidemiologic studies is complicated by the fact that
on-road vehicle exhaust emissions are a nearly ubiquitous source of combustion pollutant
mixtures that include NC>2. In recognition of this complication, the ISA notes that it is
difficult to determine the extent to which NO2 is independently associated with
respiratory effects versus being a marker for the effects of another traffic-related pollutant
or mix of pollutants (see section 5.4). This uncertainty calls into question the extent to
which effect estimates from epidemiologic studies reflect the independent contributions
of NC>2 to the adverse respiratory outcomes assessed in these studies.
In order to provide some perspective on this uncertainty, the ISA has evaluated
epidemiologic studies that employed multi-pollutant models, epidemiologic studies of
indoor NC>2 exposure, and experimental studies. Specifically, the ISA notes that a
number of NO2 epidemiologic studies have attempted to disentangle the effects of NO2
from those of co-occurring pollutants by employing multi-pollutant models. When
evaluated as a whole, NC>2 effect estimates in these models generally remained robust
when co-pollutants were included. Therefore, despite uncertainties associated with
281
-------
separating the effects of NC>2 from those of co-occurring pollutants, the ISA (section 5.4,
p. 5-16) concludes that "the evidence summarized in this assessment indicates thatNO2
associations generally remain robust in multi-pollutant models and supports a direct
effect of short-term NC>2 exposure on respiratory morbidity at ambient concentrations
below the current NAAQS." With regard to indoor studies, the ISA notes that these
studies can test hypotheses related to NC>2 specifically (ISA, section 3.1.4.1). Although
confounding by indoor combustion sources is a concern, indoor studies are not
confounded by the same mix of co-pollutants present in the ambient air or by the
contribution of NC>2 to the formation of secondary particles or 63 (ISA, section 3.1.4.1).
The ISA notes that the findings of indoor NC>2 studies are consistent with those of studies
using ambient concentrations from central site monitors (also see chapter 4 of this
document) and concludes that indoor studies provide evidence of coherence for
respiratory effects (ISA, section 3.1.4.1). With regard to experimental studies, we note
that they have the advantage of providing information on health effects that are
specifically associated with exposure to NC>2 in the absence of co-pollutants. The ISA
concludes that the NC>2 epidemiologic literature is supported by 1) evidence from
controlled human exposure studies of airway hyperresponsiveness in asthmatics, 2)
controlled human exposure and animal toxicological studies of impaired host-defense
systems and increased risk of susceptibility to viral and bacterial infection, and 3)
controlled human exposure and animal toxicological studies of airway inflammation
(ISA, section 5.3.2.1 and 5.4). When taken together, the results of epidemiologic and
experimental studies form a plausible and coherent data set that supports a relationship
between NC>2 exposures and respiratory endpoints, including symptoms and ED visits
(ISA, section 5.4), at ambient concentrations that are present in areas that meet the
current NO2 NAAQS.
10.3.2 Exposure- and risk-based considerations
In addition to the evidence-based considerations described above, staff has
considered the extent to which exposure- and risk-based information can inform
decisions regarding the adequacy of the current annual NC>2 standard, taking into account
key uncertainties associated with the estimated exposures and risks. For this review,
282
-------
exposures have been addressed in two ways. In the first, NO2 air quality in 18 locations
around the country has been used as a surrogate for exposure. In the second, exposures
have been estimated for all asthmatics and for asthmatic children considering time spent
in different microenvironments in one urban area, Atlanta, GA,. For both of these
analyses, health risks have been characterized by comparing estimates of air quality or
exposure to potential health benchmark levels (see chapters 4 and 6). The benchmarks
are based on controlled human exposure studies involving known NC>2 exposure levels
and measured airway hyperresponsiveness in asthmatics. The outputs of these analyses
are estimates of the occurrence of exposures greater than or equal to benchmark levels,
which provide some perspective on the MVrelated health risks that could exist. In
another approach to characterizing NO2-related health risks, we have estimated the
occurrences of NO2-related respiratory ED visits in Atlanta. This quantitative risk
assessment is based on NO2 concentration-response relationships identified in an
epidemiologic study of air pollution-related ED visits in Atlanta. We have selected these
endpoints because they are considered adverse to the health of individuals and because
the data necessary for the assessment are available.
In making judgments as to whether MVrelated effects should be regarded as
adverse to the health of individuals, staff has relied upon the guidelines published by the
American Thoracic Society (ATS) (2000) and conclusions from the ISA. Of the
morbidity endpoints used to characterize risks, ED visits are clearly indicative of effects
that are adverse to the health of the individual. The ATS notes that detectable effects of
air pollution on clinical measures, including ED visits, should be considered adverse. In
addition, regarding airway responsiveness, we recognize the following:
• NO2-related airway hyperresponsiveness has the potential to increase asthma
symptoms and worsen asthma control (ISA, sections 5.3.2.1 and 5.4).
• The majority of asthmatics may experience MVrelated airway
hyperresponsiveness following short-term M>2 exposures between 0.1 ppm and
0.3 ppm (ISA, table 3.1-3).
• Over 20 million people in the U.S. have asthma (ISA, table 4.4-1).
Despite uncertainty as to the magnitude of NO2-related airway hyperresponsiveness in
any single individual (see below) and despite the fact that not all asthmatics are expected
283
-------
to respond to NC>2 concentrations between 0.1 and 0.3 ppm, these considerations suggest
that NCVrelated airway hyperresponsiveness is an adverse effect when viewed from the
perspective of the asthmatic population as a whole.
10.3.2.1 Key uncertainties
The way in which exposure and risk results will inform ultimate decisions
regarding the NC>2 standard will depend upon the weight placed on each of the analyses
when uncertainties associated with those analyses are taken into consideration. The
uncertainties associated with each of the analyses (air quality, Atlanta exposure, and
Atlanta risk) are summarized below and are described in more detail in chapters 7-9 of
this document. Although we are discussing these uncertainties within the context of the
adequacy of the current standard, they apply equally to consideration of alternative
standards.
Air Quality Analyses
A number of key uncertainties should be considered when interpreting these
results with regard to decisions on the standard. These uncertainties are discussed briefly
below and in more detail in chapter 7.
• In order to simulate just meeting the current annual standard and many of the
alternative 1-h standards analyzed, an upward adjustment of recent ambient NC>2
concentrations was required. We note that this adjustment does not reflect a
judgment that levels of NC>2 are likely to increase under the current standard or
any of the potential alternative standards under consideration. Rather, these
adjustments reflect the fact that the current standard, as well as some of the
alternatives under consideration, could allow for such increases in ambient NO2
concentrations. In adjusting air quality to simulate just meeting these standards,
we have assumed that the overall shape of the distribution of NC>2 concentrations
would not change. While we believe this is a reasonable assumption in the
absence of evidence supporting a different distribution and we note that available
analyses support this approach (Rizzo, 2008), we recognize this as an important
uncertainty. It may be an especially important uncertainty for those scenarios
where considerable upward adjustment is required to simulate just meeting one or
more of the standards.
• In order to estimate NC>2 concentrations on roadways, empirically-derived
relationships between ambient concentrations measured at fixed-site monitors and
on-road concentrations were used. We have judged this to be an appropriate
approach to estimating on-road NC>2 concentrations given that these
284
-------
concentrations have been shown to be correlated with concentrations measured at
fixed-site monitors (Cape et al., 2004). However, the data used to develop the
relationships were likely collected under different conditions (e.g., with regard to
meteorology, rate of transformation of NO to NO2). We do not know the extent
to which it is appropriate to assume that these conditions are representative of the
times and places included in our analyses. Therefore, there is uncertainty in the
degree to which the relationships used to estimate on-road NO2 concentrations
reflect the actual relationship in the locations and over the time periods of interest.
• The potential health benchmark levels introduce sources of uncertainty including
the following:
o The meta-analysis that formed a large part of the basis for potential health
benchmark levels included primarily mild asthmatics. For ethical reasons,
more severely affected asthmatics were not included in the studies that formed
the basis for the meta-analysis. Severe asthmatics may be more susceptible
than mildly asthmatic individuals to the effects of NO2 exposure (ISA, section
3.1.3.2). Therefore, the potential health effect benchmarks based on these
studies could underestimate risks in populations with greater susceptibility.
Although approaches to classifying asthma severity differ, some estimates
indicate that over half of asthmatics could be classified as moderate/severe
(Fuhlbrigge et al., 2002; Stout et al., 2006).
o This meta-analysis provides information on the direction of the NO2-induced
airway response, but not on the magnitude of the response. Therefore,
although the ISA does conclude that increased airway responsiveness
associated with NO2 exposure could increase symptoms and worsen asthma
control (ISA, section 5.4), the full public health implications of benchmark
exceedances are uncertain.
Atlanta Exposure Assessment
For our Atlanta exposure assessment, we have considered the occurrence of NO2
exposures, in asthmatics, that exceed potential health benchmark levels. As with the air
quality analyses, these exposures are considered for each of the air quality scenarios
evaluated. A number of key uncertainties should be considered when interpreting these
results with regard to decisions on the standard. Some of these uncertainties, including
the approach used to adjust air quality to simulate just meeting different standards and
uncertainties associated with benchmark levels, are shared with the air quality analyses.
Additional uncertainties associated with the Atlanta exposure assessment are discussed
briefly below. A more extensive discussion of uncertainties is provided in chapter 8.
285
-------
• A number of uncertainties are associated with exposure modeling, many of them
with the activity data used in APEX.
• When compared to ambient measurement data, predicted upper percentile NO2
concentrations from AERMOD may be 10-50% higher. Because these AERMOD
outputs are used as inputs for our exposure modeling, this suggests the possibility
that we are over-predicting upper percentile NO2 exposures. Other approaches
used to evaluate our exposure results (i.e., comparison to personal exposure
monitoring results and comparison of exposure-to-ambient concentration ratios
with those identified in the ISA) suggest that exposure estimates are reasonable.
However, we cannot rule out the possibility that we are over-predicting
benchmark exceedances with our Atlanta exposure analysis.
• The exposure assessment is limited to Atlanta and the extent to which these
results are representative of other locations in the U.S. is uncertain. As noted in
section 8.11 above, staff has judged that the Atlanta exposure estimates are likely
representative of other moderate to large urban areas. However, staff also
recognizes that, given the greater proximity of the population to mobile sources in
large urban areas such as Los Angeles, New York, and Chicago (see Tables 8-14
and 8-15), the Atlanta exposure estimates likely underestimate the fraction of
asthmatics in these cities that is exposed to NO2 concentrations greater than or
equal to potential health benchmark levels.
Atlanta Risk Assessment
For our risk assessment, we have considered the prevalence of NCVrelated
respiratory ED visits in Atlanta. As with the air quality and Atlanta exposure analyses,
ED visits are considered for each of the air quality scenarios evaluated. A number of key
uncertainties should be considered when interpreting these results with regard to
decisions on the standard. Some of these, including the approach used to adjust air
quality to simulate just meeting different standards and the appropriateness of
generalizing results from Atlanta, are uncertainties shared with the air quality and/or
Atlanta exposure analyses. Additional uncertainties associated with the Atlanta risk
assessment are discussed briefly below. A more extensive discussion of uncertainties is
provided in chapter 9.
• There is uncertainty about whether the association between NO2 and ED visits
actually reflects a causal relationship. Our judgment, that there exists at least a
likely causal relationship with either short-term NO2 itself or with NO2 serving as
an indicator for itself and other components of ambient air, draws on the
conclusions in the ISA and is discussed in more detail in chapter 4.
286
-------
• The statistical uncertainty associated with the estimated NC>2 coefficient in the
concentration-response function has been characterized by confidence intervals
reflecting sample size. However, these confidence intervals do not reflect all of
the uncertainties related to the concentration-response functions, such as whether
or not the model used in the epidemiologic study is the correct model form.
• Concerning the possible role of co-pollutants in the Tolbert et al. (2007) study,
single-pollutant models may produce overestimates of the NC>2 effects if some of
those effects are really due to one or more of the other pollutants. On the other
hand, effect estimates based on multi-pollutant models can be uncertain, and can
even result in statistically non-significant estimates where a true relationship
exists, if the co-pollutants included in the model are highly correlated with NC>2.
As a result of these considerations, we report risk estimates based on both the
single- and multi-pollutant models from Tolbert et al. (2007).
10.3.2.2 Assessment results
As noted previously, the current annual NC>2 standard was retained in 1996 based
largely on an evaluation of short-term NC>2 air quality. In that review, an air quality
analysis demonstrated that locations meeting the current annual standard were unlikely to
experience short-term ambient NC>2 concentrations at central site monitors that have been
associated with respiratory effects (i.e., airway hyperresponsiveness) in controlled human
exposure studies (i.e., around 0.2 ppm). Therefore, the current annual standard was
considered requisite to protect the public health against potential effects associated with
short-term (as well as long-term) exposures. We note that a similar analysis of air quality
in the current review produced similar results. That is, 1-h NC>2 concentrations greater
than or equal to 0.20 ppm are unlikely to occur in locations around the U.S., all of which
meet the current annual standard based on recent ambient air quality as measured at
central site monitors (i.e., see tables 7-14 to 7-19).
However, in the current review, in addition to evaluating the adequacy of the
current standard with ambient air quality as measured at central site monitors, we
consider the results of additional analyses that provide perspective on potential NO2-
associated health risks. For example, in our exposure analyses, we have evaluated NC>2
concentrations on roadways which are, on average, 80% higher than concentrations
measured at central site monitors (section 7.3.2). Staff notes that high concentrations of
NC>2 on or near roadways could impact asthmatics living or walking nearby (e.g., as
287
-------
would be common in an urban environment) or commuting in cars. In addition, we have
adjusted NC>2 concentrations to simulate NC>2 air quality that could occur upon just
meeting the current and potential alternative standards. As noted above (section
10.3.2.1), these adjustments provide information on potential health risks that could be
allowed to occur under different standard options. For our exposure analyses, we have
compared NC>2 concentrations to potential health benchmark levels from 100 ppb to 300
ppb, a range that extends beyond that considered in the 1996 review. We have also
evaluated NCVrelated ED visits in the current review. Epidemiological studies that form
the basis for this analysis were not available in the 1996 review. When taken together,
these analyses provide additional information, not available in the 1996 review, on which
to base a decision regarding the adequacy of the current annual standard (and potential
alternative standards) to protect the public health. The uncertainties associated with these
analyses (see 10.3.2.1 and chapter 7) should be carefully considered when interpreting
the results of these assessments.
Air Quality and Exposure Results
The results of our air quality and exposure assessments provide some perspective
on the public health impacts of effects that we cannot currently evaluate in a quantitative
risk assessment. As noted previously, we have addressed potential exposures with two
approaches. In the first, we have estimated air quality exceedances of health benchmark
levels in 18 locations across the U.S. In the second, we have estimated exposure
exceedances for asthmatics in Atlanta, GA. Results of these analyses, as they relate to
the adequacy of the current standard, are discussed below.
When considering the air quality-based results, where air quality serves as a
surrogate for exposure, as they relate to the adequacy of the current standard, we note the
number of benchmark exceedances estimated to occur given air quality that just meets
that standard. As noted above (section 10.3.2.1), this adjustment does not reflect a
judgment that levels of NC>2 are likely to increase under the current standard. Rather, it
reflects the fact that ambient NO2 concentrations could increase under the current
standard. In situations where annual NC>2 concentrations are adjusted upward to simulate
just meeting the current standard, 1-h NC>2 concentrations measured at fixed-site monitors
in locations across the U.S. could exceed concentrations that have been associated with
288
-------
increased airway responsiveness. Most locations are estimated to experience at least 50
days per year with 1-h ambient NC>2 concentrations at fixed-site monitors greater than or
equal to 100 ppb (Figures 7-2 and 7-3) under this hypothetical scenario. Far fewer
ambient exceedances are predicted for the higher benchmark levels. For example, only 5
areas are estimated to experience any days with 1-h ambient NC>2 concentrations at
central site monitors greater than or equal to 300 ppb, and none of those locations are
estimated to experience more than 2 such days per year, on average (Appendix A).
However, as noted above on-road NC>2 concentrations are estimated to be 80%
higher (on average) than concentrations at fixed-site monitors. In the majority of
locations, roadway exceedances of the 100 ppb benchmark level could occur on most
days of the year when air quality is adjusted upward to simulate just meeting the current
standard (Figure 7-6). Even for higher benchmark levels, most locations are estimated to
have exceedances on roadways. All locations evaluated except one (Boston) are
estimated to experience on-road NC>2 concentrations greater than or equal to 300 ppb
(Appendix A). Four of these locations are estimated to experience an average of greater
than 20 days per year with on-road NC>2 concentrations greater than or equal to 300 ppb
(Appendix A).
When considering the Atlanta exposure results as they relate to the adequacy of
the current standard, we note the number of benchmark exceedances estimated to occur
given air quality that is adjusted upward to simulate just meeting the current standard. If
NC>2 concentrations were such that the Atlanta area just meets the current standard, nearly
all asthmatics in Atlanta (>97%) would be estimated to experience six or more days per
year with 1-h NC>2 exposure concentrations greater than or equal to our highest
benchmark level (0.3 ppm) (Figure 8-22).
Risk Results
When considering the Atlanta risk assessment results as they relate to the
adequacy of the current standard, there was a range of central estimates since a two year
period (2006-2007) was included in the assessment. We note that the central estimates of
incidence of NO2-related respiratory ED visits in Atlanta ranged from about 8-9% of total
respiratory-related ED visits per year based on single pollutant models (or 9,800-10,900
NO2-related incidences) when air quality is adjusted upward to simulate a situation where
289
-------
Atlanta just meets the current standard. Central estimates of incidence of NCVrelated
respiratory ED visits ranged from 2.9-7.7% of total respiratory-related ED visits per year
based on two-pollutant models (or 3,600-9,400 NO2-related incidences) In addition,
inclusion of Os and/or PMio in multi-pollutant models results in the inclusion of an
estimate of zero NO2-related respiratory ED visits within the 95% confidence intervals.
10.3.3 Conclusions regarding the adequacy of the current standard
As noted above, several lines of evidence are relevant to consider when making a
decision regarding the adequacy of the current standard to protect the public health.
These include causality judgments made in the ISA regarding the level of support for
effects associated with short-term and long-term exposures, the epidemiologic evidence
described in the ISA (and summarized in chapter 4 of this document), the conclusions in
the ISA regarding the robustness of this evidence, and the support provided for
epidemiologic findings by experimental studies. To the extent that these considerations
are emphasized, the adequacy of the current standard to protect the public health would
clearly be called into question. Such a conclusion would provide support for
consideration of an NC>2 standard that would provide increased health protection for
sensitive groups, including asthmatics and individuals who spend time on or near major
roadways (see chapter 3), against health effects ranging from increased asthma symptoms
to respiratory-related ED visits and hospital admissions associated with short-term
exposures, as well as potential effects associated with long-term exposures.
In examining the exposure- and risk-based information with regard to the
adequacy of the current annual NC>2 standard to protect the public health, we note that the
results described above (and in more detail in chapters 7-9) indicate risks associated with
air quality adjusted upward to simulate just meeting the current standard that can
reasonably be judged important from a public health perspective. Therefore, exposure-
and risk-based considerations reinforce the scientific evidence in supporting the
conclusion that consideration should be given to revising the current standard so as to
provide increased public health protection, especially for sensitive groups, from
related adverse health effects associated with short-term, and potential long-term,
exposures.
290
-------
10.4 POTENTIAL ALTERNATIVE STANDARDS
10.4.1 Indicator
In the last review, EPA focused on NC>2 as the most appropriate indicator for
ambient NOX. In this review, while the presence of gaseous NOX species other than NC>2
has been recognized (e.g., see section 1.3 of this document), no alternative to NC>2 has
been advanced as being a more appropriate surrogate for ambient gaseous NOX.
Controlled human exposure studies and animal toxicology studies provide specific
evidence for health effects following exposure to NC>2. Epidemiologic studies also
typically report levels of NC>2, as opposed to other gaseous NOX, though the degree to
which monitored NC>2 reflects actual NC>2 levels, as opposed to NC>2 plus other gaseous
NOX, can vary (e.g.,. see section 2.2.3 of this document). Because emissions that lead to
the formation of NC>2 generally also lead to the formation of other NOX oxidation
products, measures leading to reductions in population exposures to NC>2 can generally be
expected to lead to reductions in population exposures to other gaseous NOX. Therefore,
meeting an NC>2 standard that protects the public health can also be expected to provide
some degree of protection against potential health effects that may be independently
associated with other gaseous NOX even though such effects are not discernable from
currently available studies indexed by NC>2 alone. Given these key points, staff judges
that the available evidence supports the retention of NC>2 as the indicator in the current
review.
10.4.2 Averaging Time
The current annual averaging time for the NC>2 NAAQS was originally set in
1971, based on epidemiologic studies that supported a link between adverse respiratory
effects and long-term exposure to low-levels of NC>2. As noted in section 10.3.2.2, that
annual standard was retained in subsequent reviews in part because an air quality
assessment conducted by EPA concluded that areas that meet the annual standard would
be unlikely to experience short-term ambient peaks above levels that had been shown in
291
-------
controlled human exposure studies to impact endpoints of potential concern (see section
10.3.2.2). Based on currently available evidence, the issue of averaging time is being
considered in the current review, as discussed below. In order to inform judgments on
averaging time, staff has considered causality judgments from the ISA, results from
experimental and epidemiologic studies, and NC>2 air quality correlations. These
considerations are described in more detail below.
To inform general decisions regarding averaging time (e.g., short-term versus
long-term), we note the causality judgments made in the ISA regarding different health
endpoints. As described in chapter 4 of this document, the evidence relating short-term
(minutes to hours) NC>2 exposures to respiratory morbidity is judged in the ISA to be
"sufficient to infer a likely causal relationship" (ISA, section 5.3.2.1) while the evidence
relating long-term (weeks to years) NC>2 exposures to adverse health effects is judged to
be either "suggestive but not sufficient to infer a causal relationship" (respiratory
morbidity) or "inadequate to infer the presence or absence of a causal relationship"
(mortality, cancer, cardiovascular effects, reproductive/developmental effects) (ISA,
sections 5.3.2.4-5.3.2.6). These judgments most directly support an averaging time that
focuses protection on short-term exposures to NC>2.
As has been done in past reviews, it is instructive to evaluate the potential for a
standard based on annual average NC>2 concentrations, as is the current standard, to
provide protection against short-term NO2 exposures. To this end, Table 10-1 reports the
ratios of short term to annual average NC>2 concentrations. Ratios of 1-h daily maximum
concentrations (98th and 99th percentile) to annual average concentrations range from 2.5
to 8.7 while ratios of 24-h average concentrations to annual average concentrations range
from 1.6 to 3.8 (see Thompson, 2008a for more details). The variability in these ratios
across locations, particularly those for 1-h to annual average concentrations, suggests that
a standard based on annual average NC>2 concentrations would not likely be an effective
or efficient approach to focus protection on short-term NC>2 exposures.
292
-------
Table 10-1. Ratios of short-term to annual average NO2 concentrations
Location
Atlanta
Boston
Chicago
Cleveland
Denver
El Paso
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
St. Louis
Washington DC
1-h Daily
Max(99*h):Annual
4.36
2.73
3.03
3.35
4.22
4.09
8.65
3.06
7.41
2.90
3.62
2.95
3.82
3.70
1-h Daily
Max(98*h):Annual
4.00
2.50
2.86
3.03
3.78
3.65
8.21
2.70
7.03
2.60
3.34
2.70
3.74
3.02
24-h
Avg(99*h):Annual
2.37
1.81
1.68
1.99
2.44
2.05
2.84
1.97
3.76
1.88
2.40
1.81
1.99
2.10
24-h
Avg(98*h):Annual
2.13
1.66
1.62
1.78
2.25
1.91
2.65
1.79
3.42
1.75
2.07
1.69
1.84
1.88
For example, in an area with a relatively high ratio (e.g., 8), the current annual
standard (0.053 ppm) would be expected to allow 1-h daily maximum NC>2
concentrations of about 0.4 ppm. In contrast, in an area with a relatively low ratio (e.g.,
3), the current standard would be expected to allow 1-h daily maximum NC>2
concentrations of about 0.15 ppm. Thus, for purposes of protecting against the range of
1-h NC>2 exposures considered in this review (i.e., 0.1 to 0.3 ppm), a standard based on
annual average concentrations would likely require more control than necessary in some
areas and less control than necessary in others, depending on the standard level selected.
In considering the level of support available for specific short-term averaging
times, we take note of evidence from both experimental and epidemiologic studies.
Controlled human exposure studies and animal toxicological studies provide evidence
that NO2 exposures with exposure durations from less than 1-h up to 3-h can result in
respiratory effects such as increased airway responsiveness and inflammation (ISA,
section 5.3.2.7). Specifically, the ISA concludes thatNO2 exposures of 0.1 ppm for 1-h
(or 0.2-0.3 ppm for 30-min) can result in small but significant increases in nonspecific
airway responsiveness (ISA, section 5.3.2.1). In contrast, the epidemiologic literature
does not provide clear support for one short-term averaging time versus another (ISA,
section 5.3.2.7). A number of epidemiologic studies detect positive associations between
respiratory morbidity and 1-h (daily maximum) and/or 24-h NO2 concentrations. A few
293
-------
epidemiologic studies have considered both 1-h and 24-h averaging times, allowing
comparisons to be made. The ISA reports that such comparisons in studies that evaluate
asthma ED visits fail to reveal differences between effect estimates based on a 1-h
averaging time and those based on a 24-h averaging time (ISA, section 5.3.2.7).
Therefore, the ISA concludes that it is not possible, from the available epidemiologic
evidence, to discern whether effects observed are attributable to average daily (or multi-
day) concentrations (24-h average) or high, peak exposures (1-h maximum) (ISA, section
5.3.2.7).
Given the above conclusions, the experimental evidence provides support for an
averaging time of shorter duration than 24 hours (e.g., 1-h) while the epidemiologic
evidence provides support for both 1-h and 24-h averaging times. At a minimum, this
suggests that a primary concern with regard to averaging time is the level of protection
provided against 1-h daily maximum NC>2 concentrations. However, it is also worthwhile
to consider the ability of averaging times under consideration to protect against 24-h
average NC>2 concentrations. To this end, Table 10-2 presents correlations between 1-h
daily maximum NC>2 concentrations and 24-h average NC>2 concentrations (98th and 99th
percentile) across 14 locations (see Thompson, 2008a for more detail). Typical ratios
range from a 1.5 to 2.0, though one ratio (Las Vegas) is 3.1. These ratios are far less
variable than those discussed above for annual average concentrations, suggesting that a
standard based on 1-h daily maximum NO2 concentrations could also be effective at
providing adequate protection against 24-h NC>2 concentrations.
294
-------
Table 10-2. Ratios of 1-h daily maximum NO2 concentrations to 24-h average
concentrations (ppm)
Location
Atlanta
Boston
Chicago
Cleveland
Denver
El Paso
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
St. Louis
Washington DC
1-h
99th
0.078
0.064
0.093
0.072
0.086
0.075
0.039
0.095
0.059
0.093
0.060
0.093
0.064
0.079
24 h
99th
0.042
0.043
0.052
0.043
0.050
0.038
0.013
0.061
0.030
0.060
0.040
0.057
0.034
0.045
Ratio
1.84
1.50
1.80
1.68
1.73
1.99
3.04
1.56
1.97
1.55
1.51
1.63
1.91
1.76
1-h
98th
0.071
0.059
0.088
0.065
0.077
0.067
0.037
0.083
0.056
0.083
0.056
0.085
0.063
0.065
24-h
98th
0.038
0.039
0.050
0.038
0.046
0.035
0.012
0.055
0.027
0.056
0.035
0.053
0.031
0.040
Ratio
1.88
1.50
1.77
1.70
1.68
1.91
3.10
1.50
2.06
1.48
1.61
1.60
2.03
1.61
As an additional matter, we note that a short-term standard (i.e., 1-h daily
maximum) within the lower part of the range of standards considered in this risk and
exposure assessment document could have the effect of maintaining annual average NO2
concentrations below the level of the current standard (0.053 ppm). For example, in all
locations evaluated, a 1-h standard with a level of 0.05 ppm is estimated to result in
annual average NO2 concentrations less than or equal to approximately 0.02 ppm. A 1-h
standard with a level of 0.10 ppm is estimated to result in annual average NO2
concentrations less than or equal to approximately 0.04 ppm. However, a 1-h standard
with a level of 0.15 ppm could result in annual average NO2 concentrations up to
approximately 0.06 ppm and a 1-h standard with a level of 0.20 ppm could result in
annual average NO2 concentrations up to approximately 0.07 ppm (Table 10-3).
295
-------
Table 10-3. Mean annual NO2 concentrations for 2004-2006 given just meeting alternative
1-h standards (98th percentile)
Location
Atlanta
Boston
Boston
Boston
Chicago
Chicago
Chicago
Cleveland
Cleveland
Denver
Denver
Detroit
El Paso
El Paso
Jacksonville
Las Vegas
Las Vegas
Los Angeles
Los Angeles
Los Angeles
Miami
Miami
Miami
New York
New York
New York
Philadelphia
Philadelphia
Phoenix
Phoenix
Phoenix
Provo
St Louis
St Louis
St Louis
Washington
Washington
Washington
Monitor Distance from Major
Roadway
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
<=20m
> 20 and < 100 m
<=20m
>= 100 m
>= 100 m
> 20 and < 100 m
>= 100 m
>= 100 m
<=20m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
> 20 and < 100 m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
<=20m
> 20 and < 100 m
>= 100 m
Annual NO2 Concentrations (ppm) for
Different Standard Levels
0.05
0.01
0.02
0.01
0.01
0.01
0.02
0.01
0.02
0.01
0.02
0.01
0.02
0.01
0.01
0.01
0.02
0.01
0.02
0.01
0.01
0.01
0.01
0.01
0.02
0.02
0.01
0.02
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.10
0.02
0.03
0.02
0.01
0.02
0.03
0.02
0.03
0.02
0.04
0.03
0.03
0.02
0.02
0.03
0.03
0.01
0.03
0.03
0.02
0.01
0.02
0.01
0.03
0.03
0.02
0.03
0.03
0.03
0.02
0.03
0.03
0.03
0.02
0.02
0.03
0.03
0.02
0.15
0.02
0.05
0.03
0.02
0.03
0.05
0.03
0.05
0.04
0.06
0.04
0.05
0.03
0.03
0.04
0.05
0.02
0.05
0.04
0.03
0.02
0.03
0.02
0.05
0.05
0.04
0.05
0.04
0.04
0.03
0.04
0.04
0.04
0.03
0.04
0.04
0.04
0.03
0.20
0.03
0.07
0.04
0.02
0.04
0.07
0.04
0.06
0.05
0.07
0.05
0.06
0.04
0.04
0.05
0.06
0.03
0.06
0.05
0.04
0.02
0.05
0.03
0.07
0.07
0.05
0.07
0.05
0.06
0.05
0.06
0.05
0.05
0.04
0.05
0.06
0.06
0.05
10.4.3 Form
When evaluating alternative forms in conjunction with specific levels, staff
considers the adequacy of the public health protection provided by the combination of
level and form to be the foremost consideration. In addition, we recognize that it is
296
-------
important to have a form that is reasonably stable and relatively insulated from the
impacts of extreme meteorological events. A standard set with a high degree of
instability could have the effect of reducing public health protection because shifting in
and out of attainment due to meteorological conditions could disrupt an area's ongoing
implementation plans and associated control programs.
Therefore, as noted in chapter 5 and consistent with recent reviews of the Os and
PM NAAQS, we have focused in the current review on concentration-based forms
averaged over 3 years. As noted in the review of the Os NAAQS (EPA, 2007h),
concentration-based forms better reflect pollutant-associated health risks than forms
based on expected exceedances because concentration-based forms give proportionally
greater weight to periods of time when pollutant concentrations are well above the level
of the standard than to times when the concentrations are just above the standard, while
an expected exceedance form would give the same weight to periods of time with
concentrations that just exceed the standard as to times when concentrations greatly
exceed the standard. Concentration-based forms also provide greater regulatory stability
than a form based on allowing only a single expected exceedance.
In considering specific concentration-based forms on which to focus the current
review, we note the need to minimize the number of days per year that an area could
exceed the standard level and still attain the standard. Given this, we have focused on
98th and 99th percentile forms averaged over 3 years. With regard to these alternative
forms, staff notes that a 99th percentile form for a 1-h daily maximum standard would
correspond to the 4th highest daily maximum concentration in a year while a 98th
percentile form would correspond approximately to the 7th to 8th highest daily maximum
concentration in a year (Table 10-4; see Thompson, 2008 for methods). As noted in
chapter 5, staff has judged that these forms would provide an appropriate balance
between limiting peak NC>2 concentrations and providing a stable regulatory target. This
is consistent with judgments made in the 2006 review of the PM NAAQS (EPA, 2005).
297
-------
Table 10-4. NO2 concentrations (ppm) corresponding to 2" -9 daily maximum and 98 799
percentile forms (2004-2006)
->th
2004-2006
Location
Atlanta
Boston
Chicago
Cleveland
Denver
El Paso
Las Vegas
Los Angeles
Miami
New York
Philadelphia
Phoenix
St. Louis
Washington DC
NO2 Daily Maximums
2nd
0.083
0.069
0.103
0.075
0.094
3rd
0.079
0.067
0.094
0.074
0.089
0.085 0.080
0.042 I 0.040
0.110
0.065
0.112
0.065
0.107
0.066
0.102
0.095
0.060
0.099
0.062
0.097
0.065
0.088
4th
0.078
0.064
0.093
0.072
0.086
0.075
0.039
0.095
0.059
0.093
0.060
0.093
0.064
0.079
5th
0.074
0.063
0.090
0.070
0.082
0.072
0.039
0.089
0.058
0.090
0.059
0.090
0.064
0.075
6th
0.073
0.062
0.090
0.069
0.079
0.071
0.038
0.088
0.057
0.086
0.058
0.089
0.063
0.072
7th
0.072
0.060
0.088
0.066
0.077
0.068
0.038
0.084
0.056
0.084
0.057
0.086
0.063
0.066
8th
0.070
0.059
0.088
0.065
0.073
0.067
0.037
0.083
0.054
0.082
0.056
0.084
0.063
0.065
9th
0.070
0.059
0.088
0.064
0.072
0.066
0.037
0.081
0.053
0.081
0.054
0.083
0.063
0.063
Percentiles
99th
0.078
0.064
0.093
0.072
0.086
0.075
0.039
0.095
0.059
0.093
0.060
0.093
0.064
0.079
98th
0.071
0.059
0.088
0.065
0.077
0.067
0.037
0.083
0.056
0.083
0.056
0.085
0.063
0.065
When considering the extent to which exposure and risk analyses inform
judgments on standard form, staff notes that a 99th percentile form could be appreciably
more protective than a 98th percentile form in some locations, as judged by the results of
our air quality analyses. For example, in Boston, Philadelphia, and Washington, D.C. a
99th percentile standard of 0.20 ppm is estimated to decrease benchmark exceedances,
relative to a 98th percentile form, by approximately 50-70% (Table 10-5). However, in
St. Louis, Detroit, and Las Vegas a 99th percentile form could decrease benchmark
exceedances by only approximately 10% (Table 10-5). For most locations, the difference
is estimated to be between approximately 10 and 50% (Table 10-5). With regard to the
Atlanta exposure assessment, we note that adoption of a 98th percentile form versus a 99th
percentile form would have virtually no effect on exposure concentrations at or above 0.1
ppm. However, choice of form could make a difference of approximately 5-10% on the
number of exposure concentrations at or above the 0.2 and 0.3 ppm benchmark levels
(Figure 8-22). With regard to the Atlanta risk assessment, a 99th percentile form is
estimated to be associated with approximately 6% to 8% fewer NO2-related ED visits
than a 98th percentile form, across the levels of the potential 1-h standards examined
(Tables 9-2 to 9-4).
298
-------
Table 10-5. Mean number of days per year (averaged over the 2004-2006 time period)
estimated to have ambient (central site monitor) 1-h daily maximum NO2
concentrations > 0.10 ppm assuming 98th and 99th percentile forms of a 0.20
ppm standard
Exceedances
Location Name 98th 99th
Atlanta 81 66
Boston 29 9
Chicago 84 62
Denver 211 159
Detroit 209 194
ElPaso 159 116
Jacksonville 178 53
Las Vegas 108 97
Los Angeles 70 50
Miami 74 64
New York 108 63
Philadelphia 163 77
Phoenix 224 163
Provo 84 74
St. Louis 146 134
Washington 133 69
When considering these results as they relate to standard form, we note that a decision on
form must be made in conjunction with selection of a particular standard level. The primary
emphasis in such a decision will be on the level of public health protection provided by the
combination of form and level. With regard to a decision on form, we note that the geographic
heterogeneity of the impact of form, as indicated in the air quality analysis, suggests that caution
should be exercised when using the Atlanta results as a basis for selecting the most appropriate
form. To the extent that a decision regarding form emphasizes this geographic uncertainty, it
would likely place more weight on the air quality results.
10.4.4 Level
10.4.4.1 Evidence-based considerations
In considering alternative standard levels that would provide greater protection than
that afforded by the current standard against NO2-related adverse health effects, staff has
taken into account scientific evidence from both experimental and epidemiologic studies,
as well as the uncertainties and limitations in that evidence. In particular, we have
299
-------
considered the extent to which controlled human exposure studies provide evidence for a
lowest-observed-effects level and the extent to which epidemiologic studies provide
evidence for potential effect thresholds and/or for positive associations that extend down
to the lower levels of NC>2 concentrations observed in studies. We note that the scientific
evidence can provide insights into alternative standard levels only within the context of
specific averaging times and forms. Therefore, while this section considers the evidence
as it relates to alternative levels, such considerations assume particular averaging times
and forms. Additional discussion of averaging time can be found in sections 5.3 and 10.5
of this document. Additional discussion of form can be found in sections 5.4 and 10.6 of
this document.
When considering the scientific evidence as it relates to alternative levels, we note
that NC>2 concentrations represent different metrics when reported in experimental studies
versus epidemiologic studies. Concentrations of NC>2 reported in epidemiologic studies
are typically based on ambient monitoring data while NC>2 levels reported in controlled
human exposure studies represent the concentration of NC>2 in the breathing zone of the
individual. In some locations and at some points in time, individuals are likely exposed
to NC>2 concentrations that are higher than those measured at ambient monitors. The ISA
concludes that elevated NC>2 monitors (e.g., monitors sited on building roofs), particularly
in inner cities, likely underestimate concentrations and personal exposures occurring at
lower elevations, closer to motor vehicle emissions (ISA, section 5.2.2). In this situation,
average NC>2 concentrations measured at an elevated ambient monitor could be well
below concentrations associated with effects in controlled human exposure studies while
actual NC>2 exposures on and/or near roadways could be comparable to, or higher than,
these concentrations. We note that in this situation, where personal exposure
concentrations to ambient NC>2 are higher than ambient levels measured at a fixed-site
monitor, ambient standards based principally on controlled exposure studies could be less
health-protective than standards primarily based on concentrations reported in
epidemiologic studies at ambient monitors. However, the ISA also concludes that, in
exposure measurement field studies where personal exposures have been measured,
personal exposures to NC>2 of ambient origin were generally lower than ambient NC>2
concentrations based on fixed-site monitors (ISA, tables 2.5-4 and 2.5-5; section 5.2.2).
300
-------
In this type of situation, where personal exposure concentrations to ambient NC>2 are
lower than the levels measured at ambient monitors, we note that an ambient standard
based principally on effects levels observed in controlled exposure studies could be more
health-protective than a standard based primarily on concentrations reported in
epidemiologic studies at ambient monitors.
Controlled Human Exposure Studies
In considering the available controlled human exposure studies (see chapter 4 for
more detail), we note that these studies have addressed the consequences of short-term
(e.g., 30-minutes to several hours) NC>2 exposures for a number of health endpoints
including airway responsiveness, host defense and immunity, inflammation, and lung
function (ISA, section 3.1). In identifying health endpoints on which to focus for
purposes of informing decisions about potential alternative standard levels, staff judges it
appropriate to focus on those endpoints that occur at or near ambient levels of NC>2 and
endpoints that are of clinical significance. As described in more detail in section 4.5.3,
the only endpoint to meet these criteria is increased airway responsiveness in asthmatics.
The ISA concludes thatNO2 exposures between 0.2 and 0.3 ppm for 30 minutes or 0.1
ppm for 60-minutes can result in small but significant increases in nonspecific airway
responsiveness (ISA, section 5.3.2.1) and that "transient increases in airway
responsiveness following NC>2 exposure have the potential to increase symptoms and
worsen asthma control" (ISA, sections 3.1.3 and 5.4). This effect could have important
public health implications due to the large size of the asthmatic population in the United
States (ISA, Table 4.4-1). In addition, NC>2 effects on airway responsiveness in
asthmatics are part of the body of experimental evidence that provides plausibility and
coherence for the observed NO2-related increased in hospital admissions and ED visits in
epidemiologic studies (ISA, section 5.3.2.1). For all of these reasons, we have focused
on increased airway responsiveness in asthmatics when considering the controlled human
exposure literature in terms of its ability to inform decisions on alternative standard
levels.
With regard to controlled human exposure studies of airway responsiveness, we
note that a meta-analysis of individual level data from 19 studies (section 3.1.3.2 and
table 4-5 in chapter 4 of this document) indicates that 66% of resting asthmatics
301
-------
experienced increased airway responsiveness following exposure to 0.1 ppm NO2, 67%
experienced an increase following exposure to NO2 concentrations between 0.1 and 0.15
ppm (inclusively), 75% experienced an increase following exposure to NO2
concentrations between 0.2 and 0.3 ppm (inclusively), and 73% experienced an increase
following exposure to NO2 concentrations above 0.3 ppm. Effects of NO2 exposure on
the direction of airway responsiveness are statistically-significant at all of these levels.
As noted in section 10.3.2.1, one of the important uncertainties associated with these
results is that, because the meta-analysis evaluated only the direction of the change in
airway responsiveness, it is not possible to discern the magnitude of the change from
these data. This limitation makes it particularly difficult to quantify the public health
implications of these results.
Epidemiologic Studies
When evaluating the epidemiologic literature for its potential to inform decisions
on standard level, we note that the ISA concludes that NO2 epidemiologic studies provide
"little evidence of any effect threshold" (section 5.3.2.9, p. 5-15). In studies that have
evaluated concentration-response relationships, they appear linear within the observed
range of data (ISA, section 5.3.2.9). In the absence of an apparent threshold, we are
focusing on the range of levels that have been associated with key U.S. studies for
purposes of identifying the range of standard levels supported by the epidemiologic
literature. When identifying this range, we focus on the higher percentiles of NO2
concentrations measured at the ambient monitors used in the study (i.e., 98th and 99th
percentiles), as these percentiles are likely most relevant for the health effects observed in
epidemiologic studies.
Figures 5-1 and 5-2 (see chapter 5) show standardized effect estimates and the
98th and 99th percentile concentrations of daily 1-h maximum NO2 for locations and time
periods that correspond to key U.S. epidemiologic studies identified in the ISA (see table
5.4-1 in ISA for a list of key studies). These key studies are associated with a range of
98th/99th percentile 1-h daily maximum levels from 0.05 ppm to 0.21 ppm. In
considering this information, we note that, toward the lower end of the range of NO2
concentrations observed in epidemiologic studies, there is increasing uncertainty as to
302
-------
whether observed health effects remain plausibly related to exposures to ambient NC>2, as
opposed to the broader mix of air pollutants present in the ambient air.
When considering an appropriate lower end of the range of levels that are
supported by the evidence, staff has considered two primary factors. First, the study by
Delfino et al., (2002) provides evidence for associations between short-term ambient NC>2
concentrations and respiratory morbidity in a location where NC>2 concentrations were
well below levels in most other key U.S. epidemiologic studies. This study reports
positive associations between 1-h and 8-h (only 8-h associations were statistically-
significant) levels of NC>2 and asthma symptoms in a location where the 98th and 99th
percentile 1-h daily maximum NC>2 concentrations were 0.05 and 0.053 ppm,
respectively. Second, the controlled human exposure studies of airway responsiveness
that formed the basis for the meta-analysis detected a lowest-observable-effect level of
0.1 ppm NC>2. However, these studies did not evaluate severe asthmatics. Most of the
subjects included in these studies were mild asthmatics. More severely affected
asthmatics may be more susceptible than mild asthmatics to the effects of NC>2 exposure
(ISA, section 3.1.3.2). As a result, staff judges that it is appropriate to base the lower end
of the range of alternative standard levels on the epidemiologic study by Delfino et al.
(2002) and on providing increased protection relative to the lowest-observed-effects level
for airway hyperresponsiveness in asthmatics. Therefore, staff concludes that the lower
end of the range of potential alternative 1-h daily maximum standards that is reasonably
supported by the evidence is 0.05 ppm (50 ppb).
When considering an appropriate upper end of the range of 1-h daily maximum
standard levels that is supported by the evidence, we note the following:
• Positive and statistically-significant associations were observed in several key
U.S. epidemiologic studies associated with 1-h daily maximum levels of NC>2
close to 0.1 ppm (Peel et al., 2005; NYDOH, 2006; Ito et al., 2007; Tolbert et al.,
2007) (see Figure 5-1). In multi-pollutant models, effect estimates remained
statistically-significant in the study by Ito and positive, but non-significant, in the
other studies.
• Positive and statistically-significant NC>2 effect estimates were also observed in
the two key U.S. studies associated with the highest 1-h NC>2 concentrations (Linn
et al., 2000; Ostro et al., 2001). These studies were associated with 98th and 99th
percentile 1-h daily maximum NC>2 concentrations from 0.18 ppm to 0.21 ppm.
These studies did not evaluate multi-pollutant models. Therefore, they do not
303
-------
provide additional support for an independent association between NC>2 and
respiratory morbidity beyond that provided by the studies noted above.
• The meta-analysis of airway responsiveness presented in the ISA reports
statistically-significant effects on the direction of airway responsiveness following
short-term NC>2 exposures from 0.1 ppm to 0.3 ppm. The ISA does not draw
distinctions between levels within this range with regard to the likely magnitude
of the response or the percent of asthmatics expected to respond.
Given these observations, staff notes that the scientific evidence provides strong support
for a standard at or below 0.1 ppm (100 ppb). However, to the extent that a decision
regarding standard level emphasizes the general uncertainties associated with quantifying
the contributions of NO2 to respiratory effects in epidemiologic studies and uncertainties
regarding the public health significance of NO2-associated airway hyperresponsiveness
(particularly at 0.1 ppm), a level as high as 0.2 could be supported. Therefore, staff
concludes that the upper end of the range of 1-h daily maximum standard levels that is
reasonably supported by the evidence is 0.2 ppm (200 ppb).
10.4.4.2 Exposure- and risk-based considerations
Staffs consideration of exposure- and risk-based information as it relates to
alternative levels for the primary NC>2 NAAQS builds upon our conclusion, discussed
above in section 10.3, that the overall body of scientific evidence clearly calls into
question the adequacy of the current standard to protect the public health. Therefore, we
have judged it appropriate to consider a range of alternative levels that would improve
upon the level of protection provided by the current standard. As noted in chapter 5, this
range of levels (0.05-0.20 ppm) is based on results from controlled human exposure and
epidemiologic studies. When considering this range of levels, we note that, given recent
air quality, only a level of 0.05 ppm would be estimated to result in any counties in the
U.S. that are above the level of the standard (Table 10-6; Thompson, 2008).
304
-------
Table 10-6. Percent of counties that may be above the level of the standard, given different levels (based on years 2004-2006)
Alternative Standards
and Levels (ppm)
Number of counties
with monitors
(population in 1000s)
Percent of counties, total and by region, (and total population) not likely to meet standard and level
Total counties
(population in
millions)
138(93.9)
Northeast
31
Southeast
32
Industrial
Midwest
20
Upper
Midwest
12
Southwest
9
Northwest
20
Southern
CA
13
Outside
Regions**
1
3 year 99th percentile daily 1 hour max:
0.20
0.15
0.10
0.05
0
0
0
59 (79.6)
0
0
0
77
0
0
0
41
0
0
0
80
0
0
0
17
0
0
0
56
0
0
0
55
0
0
0
77
0
0
0
0
3 year 98th percentile daily 1 hour max:
0.20
0.15
0.10
0.05
0
0
0
46 (63.4)
0
0
0
55
0
0
0
34
0
0
0
60
0
0
0
17
0
0
0
44
0
0
0
45
0
0
0
69
0
0
0
0
305
-------
The results of the air quality analysis are presented in chapter 7 of this document. In that
chapter, we present estimates of the number of days per year with ambient (based on fixed-site
ambient monitors) and on-road (based on on-road adjustment) NC>2 concentrations at or above
our potential health benchmark levels for the years 2001-2006. These estimates are based on as-
is air quality and air quality that has been adjusted to simulate just meeting the current and
potential alternative standards. In considering the results presented chapter 7, we note the
following key points:
• Given unadjusted air quality for the years 2001-2006, it is estimated that 1-h NC>2
concentrations would not exceed 0.1 ppm more than 14 days per year, on average, at
fixed-site monitors in any location evaluated (Tables 7-13 through 7-18).
• Given air quality for the years 2001-2003 adjusted to simulate just meeting the
current standard, it is estimated that that an average of up to 211 days per year (most
locations between 20 and 100 days per year) could occur with ambient NC>2
concentrations at fixed-site monitors that exceed 0.1 ppm (Table 7-26).
• Given air quality for the years 2001-2003 adjusted to simulate just meeting a standard
level of 200 ppb (98th percentile), it is estimated that an average of up to 327 days per
year (most locations between 50 and 200 days per year) could occur with ambient
NC>2 concentrations at fixed-site monitors that exceed 0.1 ppm (Table 7-26).
• Given air quality for the years 2001-2003 adjusted to simulate just meeting a standard
level of 150 ppb (98th percentile), it is estimated that an average of up to 200 days per
year (most locations between 20 and 100 days per year) could occur with ambient
NC>2 concentrations at fixed-site monitors that exceed 0.1 ppm (Table 7-26).
• Given air quality for the years 2001-2003 adjusted to simulate just meeting a standard
level of 100 ppb (98th percentile), it is estimated that an average of up to 8 days per
year could occur with ambient NC>2 concentrations at fixed-site monitors that exceed
0.1 ppm (Table 7-26).
• Given air quality for the years 2001-2003 adjusted to simulate just meeting a standard
level of 50 ppb (98th percentile), it is estimated that an average of up to 1 day per year
could occur with 1-h ambient NC>2 concentrations at fixed-site monitors greater than
or equal to 0.1 ppm. All of the locations evaluated except for 2 would be expected to
experience 0 days per year with ambient NC>2 concentrations at fixed-site monitors
that exceed this benchmark (Table 7-26).
• Mean estimates of days per year with on-road exceedances are higher than estimates
of ambient exceedances at fixed-site monitors (Up to 18 days per year for a standard
of 50 ppb, 257 for a standard of 100 ppb, 343 for a standard of 150 ppb, and 351 for a
306
-------
standard of 200 ppb based on the years 2001 to 2006 and 98th percentile standards)
(Table 7-29).
The results of the Atlanta exposure assessment are presented in chapter 8 of this document.
In Figures 8-19 through 8-26, we present estimates of the percent of asthmatics in Atlanta
expected to experience NO2 exposure concentrations at or above our potential health benchmark
levels for the year 2002 (estimates were similar for the years 2001-2003 (Figure 8-5)), given
unadjusted air quality and air quality that has been adjusted to simulate just meeting the current
and potential alternative standards. In considering the results presented in those figures, we note
the following key points:
• Given unadjusted air quality, it is estimated that virtually all asthmatics in Atlanta
could experience 6 or more exposures to NO2 concentrations greater than or equal to
0.1 ppm. It is estimated that just under 60% of Atlanta asthmatics could experience at
least one exposure to NO2 concentrations greater than or equal to 0.3 ppm and that
fewer than 10% could experience 6 or more exposures to NO2 concentrations greater
than or equal to 0.3 ppm (Figure 8-19).
• Given air quality adjusted to simulate just meeting the current standard, it is estimated
that virtually all Atlanta asthmatics could experience 6 or more days per year with
NO2 exposure concentrations greater than or equal to 0.3 ppm (Figure 8-22).
• Given air quality adjusted to simulate just meeting a standard level of either 150 or
200 ppb (98th or 99th percentile standard), it is estimated that more than 95% of
Atlanta asthmatics could experience 1 or more days per year with NO2 exposure
concentrations greater than or equal to 0.3 ppm (Figure 8-23).
• Given air quality adjusted to simulate just meeting a standard level of 100 ppb (99th
percentile standard), it is estimated that approximately 70% of Atlanta asthmatics
could experience 1 or more days per year with NO2 concentrations of 0.3 ppm or
above. With a 99th percentile standard, approximately 10% of Atlanta asthmatics
could experience 6 or more days per year with NO2 concentrations of 0.3 ppm or
above (Figure 8-26).
• Given air quality adjusted to simulate just meeting a standard level of 50 ppb (99th
percentile standard), it is estimated that fewer than 10% of asthmatics could
experience 1 or more days per year with NO2 concentrations of 0.3 ppm or above.
With a 99th percentile standard, virtually none are estimated to be exposed 6 or more
times per year. However, this standard is estimated to result in over 90% of
asthmatics being exposed 6 or more times per year to NO2 concentrations of 0.1 ppm
or above (Figure 8-24).
307
-------
The results of the Atlanta risk assessment are presented in chapter 9 of this document. In
Tables 9-2 through 9-4, we present estimates, for the years 2005-2007, of the percent of total
annual respiratory ED visits in Atlanta associated with NO2. These results are also presented as
incidence of NO2-associated respiratory ED visits in Appendix C. In considering the results
presented in Tables 9-2 through 9-4, we note the following key points:
• Based on single-pollutant models, central estimates of annual NO2-related respiratory
ED visits associated with recent air quality for the years 2005-2007 range from 2.8 to
3.1% (or 3,400 to 3,800 NCVrelated incidences per year). Based on multi-pollutant
models, central estimates for this same time period range from 0.6 to 2.6% (or 700 to
3,200 NO2-related incidences per year).
• Central estimates of annual NCVrelated respiratory ED visits associated with air
quality adjusted upward to simulate just meeting the current annual standard (based
on 2006-2007) range from 8.1 to 9.0% (or 9,800 to 10,900 NO2-related incidences
per year) based on single-pollutant models and from 1.7 to 7.7% (or 3,100 to 9,400
NO2-related incidences per year) based on multi-pollutant models. .
• Central estimates of annual NCVrelated respiratory ED visits associated with air
quality adjusted upward to simulate just meeting a 200 ppb, 1-h daily maximum, 98th
percentile standard (based on 2005-2007) ranges from 7.6 to 8.5% based on single-
pollutant models and from 1.6 to 7.3% based on multi-pollutant models.
• Central estimates of annual NO2-related respiratory ED visits associated with air
quality adjusted upward to simulate just meeting a 150 ppb, 1-h daily maximum, 98th
percentile standard (based on 2005-2007) ranges from 5.8 to 6.4% based on single-
pollutant models and from 1.2 to 5.5% based on multi-pollutant models.
• Central estimates of annual NCVrelated respiratory ED visits associated with air
quality adjusted to simulate just meeting a 100 ppb, 1-h daily maximum, 98th
percentile standard (based on 2005-2007) ranges from 3.9 to 4.3% based on single-
pollutant models and from 0.8 to 3.7% based on multi-pollutant models.
• Central estimates of annual NO2-related respiratory ED visits associated with air
quality adjusted to simulate just meeting a 50 ppb, 1-h daily maximum, 98th percentile
standard (based on 2005-2007) ranges from 2.0 to 2.2% based on single-pollutant
models and from 0.4 to 1.9% based on multi-pollutant models.
• Central estimates of annual NCVrelated respiratory ED visits associated with air
quality adjusted to simulate 99th percentile 1-h daily maximum standards in the range
of 50 to 200 ppb are generally on the order of 10% lower than the estimates
summarized above for standards with a 98th percentile form.
308
-------
10.4.4.3 Conclusions regarding level
As noted in section 10.7.1, staff concludes that the scientific evidence reasonably
supports a range of standard levels from 50 ppb to 200 ppb, with strong support for a level at or
below 100 ppb. In also considering the exposure-based information, we note that standard levels
of 150 and 200 ppb are generally estimated to be associated with a similar or greater number of
benchmark exceedances than are associated with just meeting the current standard, with standard
levels of 100 and 50 ppb providing appreciable reductions in estimated benchmark exceedances.
In considering the risk-based information, we note that all of the standard levels evaluated are
estimated to be associated with fewer NO2-related ED visits, on average, than are associated with
just meeting the current standard, though the reduction associated with a standard level of 200
ppb is relatively small, with reductions notably increasing with standard levels going from 150
ppb down to 50 ppb. When the scientific evidence is considered in conjunction with exposure
and risk results, the strongest support is for standard levels between 50 and 100 ppb. This
represents a range of levels that is consistent with the scientific evidence and that would be
expected to provide improved public health protection relative to that provided by the current
annual standard.
309
-------
11. References
Abt Associates Inc. (2005). Participate Matter Health Risk Assessment for Selected Urban
Areas. Prepared for Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. June 2005. Available online at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_td.html.
Abt Associates Inc. (2007). Ozone Health Risk Assessment for Selected Urban Areas. Prepared
for Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, Research Triangle Park, NC., July 2007, Under Contract No. 68-D-03-002,
Work Assignment 3-39 and 4-56. Available online at:
http://www. epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html .
AHS. (2003). Source and Accuracy Statement for the 2003 AHS-N Data Chart. Available at:
http: //www .census. gov/hhes/www/housing/ahs/03 dtchrt/source .html
AHS. (2004). US Census Bureau's American Housing Survey for the for the Atlanta
Metropolitan Area: 2004. Available at: http://www.census.gov/prod/2005pubs/hl70-04-
21.pdf.
AHS. (2008). US Census Bureau's American Housing Survey, Metropolitan Data.
http://www.census.gov/hhes/www/housing/ahs/metropolitandata.html.
American Thoracic Society (2000). What Constitutes an Adverse Health Effect of Air Pollution?
American Journal of Respiratory and Critical Care Medicine. 161:665-673.
Barck, C, Sandstrom T, Lundahl J, Hallden G, Svartengren M, Strand V, Rak S, Bylin G.
(2002). Ambient level of NO2 augments the inflammatory response to inhaled allergen in
asthmatics. RespirMed. 96:907-917.
Barck, C, Lundahl J, Hallden G, Bylin G. (2005). Brief exposures to NO2 augment the allergic
inflammation in asthmatics. Environ Res. 97:58-66.
Beckerman, B, Jerrett M, Brook JR, Verma DK, Arain MA, Finkelstein MM. (2008).
Correlation of nitrogen dioxide with other traffic pollutants near a major expressway.
AtmosEnviron. 42:275-290.
Bell, S and Ashenden TW. (1997). Spatial and temporal variation in nitrogen dioxide pollution
adj acent to rural roads. Water Air Soil Pollut. 95:87-98.
110
-------
Blackwell A and Kanny D. (2007). Georgia Asthma Surveillance Report. Georgia Department
of Human Resources, Division of Public Health, Chronic Disease, Injury, and
Environmental Epidemiology Section, February 2007. Publication Number:
DPH07/049HW. Available at: http://health.state.ga.us/epi/cdiee/asthma.asp.
Blackwell AD, Wu M, Mertz KJ, Powell KE, Williams CP, Chowdhury P. (2003). The Burden of
Asthma in Georgia 2003. Georgia Department of Human Resources; Division of Public
Health; Chronic Disease, Injury, and Environmental Epidemiology Section, December
2003. Publication number DPH03/127HW. Available at:
http://health.state.ga.us/pdfs/epi/cdiee/burdenofasthma.03.pdf
Brauer, M, Hoek G, Smit HA, De Jongste JC, Gerritsen J, Postma DS, Kerkhof M, Brunekreef B.
(2007). Air pollution and development of asthma, allergy and infections in a birth cohort.
Eur. Respir. J. 29:879-5
Bylin, G, Lindvall T, Rehn T, Sundin B. (1985) Effects of short-term exposure to ambient
nitrogen dioxide concentrations on human bronchial reactivity and lung function. Eur. J.
Respir. Dis. 66:205-217.
Cape, JN, Tang YS, van Dijk N, Love L, Sutton MA, Palmer SCF. (2004). Concentrations of
ammonia and nitrogen dioxide at roadside verges, and their contribution to nitrogen
deposition. Environ Pollut. 132:469-478.
CARB. (2001). Indoor air quality: residential cooking exposures. Final report. California Air
Resources Board, Sacramento, California. Available at:
http://www.arb. ca. gov/research/indoor/cooking/cooking.htm.
CARB (2003). Characterizing the Range of Children's Pollutant Exposure during School Bus
Commutes. Prepared for the California Air Resources Board under Contract No. 00-322.
October 10, 2003. Available at: http://www.arb.ca.gov/research/apr/past/00-322.pdf.
CARB. (2006). Diesel Particulate Matter Exposure Assessment Study for the Ports of Los
Angeles and Long Beach, California Air Resources Board. Available at:
http://www.arb.ca.gov/regact/marine2005/portstudy0406.pdf
CDC. (2007). National Center for Health Statistics. National Health Interview Survey (NHIS)
Public Use Data Release (2003). Available at:
http://www.cdc.gov/nchs/about/major/nhis/quest data related 1997 forward.htm.
Chan, AT and Chung MW. (2003). Indoor-outdoor air quality relationships in vehicle: effect of
driving environment and ventilation modes. Atmos Environ. 37:3795-3808.
Ill
-------
Chauhan, AJ, Inskip HM, Linaker CH, Smith S, Schreiber J, Johnston SL, Holgate ST (2003).
Personal exposure to nitrogen dioxide (NO2) and the severity of virus-induced asthma in
children. Lancet 361:1939-1944.
Clougherty, JE, Levy JI, Kubzansky LD, Ryan PB, Suglia SF, Canner MJ, and Wright RJ.
(2007). Synergistic effects of traffic-related air pollution and exposure to violence on
urban asthma etiology. Environ. HealthPerspect. 115:1140-1146.
Cohen J, Cook R, Bailey C, Carr E. (2005). Relationship between motor vehicle emissions of
hazardous pollutants, roadway proximity, and ambient concentrations in Portland, OR.
Environmental Modeling and Software. 20:7-12.
Delfino, RJ, Gone H, Linn WS, Pellizzari ED, Hu Y. (2003). Asthma symptoms in Hispanic
children and daily ambient exposures to toxic and criteria air pollutants. Environ. Health
Perspect. 111:647-656.
Delfino, RJ, Zeiger RS, Seltzer JM, Street DH, McLaren CE. (2002). Association of asthma
symptoms with peak particulate air pollution and effect modification by anti-
inflammatory medication use. Environ. Health Perspect. 110:A607-A617.
Devalia, JL, Rusznak C, Herdman MJ, Trigg CJ, Tarraf H, Davies RJ. (1994) Effect of nitrogen
dioxide and sulphur dioxide on airway response of mild asthmatic patients to allergen
inhalation. Lancet 344:1668-1671.
Dietert, RR, Etzel RA, Chen D, Halonen M, Holladay SD, Jarabek AM, Landreth K, Peden DB,
Pinkerton K, Smialowicz RJ, Zoetis T. (2000). Workshop to identify critical window of
exposure for children's health: immune and respiratory systems work group summary.
Environ Health Perspect. Suppl. 108(3):483-490.
EIA. (2001). Residential Energy Consumption Survey of 2001. Energy Information
Administration. Available at:
http://www.eia.doe.gov/emeu/recs/recs2001 he/200 ltblhp.html.
EPA. (1982). Air Quality Criteria for Oxides of Nitrogen. National Center for Environmental
Assessment, Research Triangle Park, NC. EPA-600/8-82-026. Available at
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=44436.
EPA (1983). Guideline on the Meaning and Use of Precision and Accuracy Data Required by
40 CFRParts 58, Appendices A and B. June 1983. PB83-238949.
EPA. (1993). Air Quality Criteria Document for the Oxides of Nitrogen. National Center for
Environmental Assessment, Research Triangle Park, NC. EPA-600/8-91/049F.
Available at: http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=40179.
112
-------
EPA. (1995). Review of the National Ambient Air Quality Standards for Nitrogen Dioxide:
Assessment of Scientific and Technical Information. OAQPS Staff Paper. Office of Air
Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-95-005.
September 1995.
EPA. (2002). Consolidated Human Activities Database (CHAD) Users Guide. Database and
documentation available at: http://www.epa.gov/chadnet 1 /.
EPA. (2003). AERMOD: Latest Features and Evaluation Results. EPA-454/R-03-003. U.S.
Environmental Protection Agency, Research Triangle Park, NC. Available at:
http://www.epa.gov/scramOO l/7thconf/aermod/aermod_mep.pdf
EPA. (2004). AERMOD: Description of Model Formulation. Office of Air Quality Planning
and Standards, Research Triangle Park, NC. EPA-454/R-03-004. Available at:
http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf
EPA (2005). Review of the National Ambient Air Quality Standards for Particulate Matter:
Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Office
of Air Quality Planning and Standards, Research Triangle Park, NC. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/pmstaffpaper_20051221.pdf.
EPA (2005). National Ambient Air Monitoring Strategy. Draft. Available at:
http://www.epa.gov/ttn/amtic/files/ambient/monitorstrat/naamstrat2005.pdf
EPA. (2006a). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model
Documentation (TRIM.Expo / APEX, Version 4) Volume I: User's Guide. Office of Air
Quality Planning and Standards, Research Triangle Park, NC. June 2006. Available at:
http://www.epa.gov/ttn/fera/human apex.html.
EPA. (2006b). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model
Documentation (TRIM.Expo / APEX, Version 4) Volume II: Technical Support
Document. Office of Air Quality Planning and Standards, Research Triangle Park, NC.
Available at: http://www.epa.gov/ttn/fera/human_apex.html.
EPA. (2006c). User's Guide for the AMS/EPA Regulatory Model - AERMOD (EPA-454/B-03-
001, September 2004), Addendum. U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina 27711.
EPA. (2007a). Plan for Review of the Primary National Ambient Air Quality Standard for
Nitrogen Dioxide. Available at:
http://www.epa.gOv/ttn/naaqs/standards/nox/s nox cr pd.html.
-------
EPA. (2007b). Nitrogen Dioxide Health Assessment Plan: Scope and Methods for Exposure
and Risk Assessment. Office of Air Quality Planning and Standards, Research Triangle
Park, NC. Available at: http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_pd.html.
EPA. (2007c). US EPA Air Quality System (AQS). Download Detailed AQS Data. Available
at: http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm.
EPA(2007d). Field Guide to Air Quality Data (vl.0.0). February 28, 2007. Available at:
http://www.epa.gov/ttn/airs/aqsdatamart/documentation/index.htm.
EPA. (2007e). 2002 National Emissions Inventory Data & Documentation. Office of Air
Quality Planning and Standards, Research Triangle Park, NC. Available at:
http://www.epa.gov/ttn/chief/net/2002inventory. html.
EPA. (2007f). Clean Air Markets - Data and Maps. Office of Atmospheric Programs,
Washington, DC. Emissions Prepackaged Data Sets. Available at:
http://camddataandmaps.epa.gov/gdm/index.cfm?fuseaction=emissions. wizard.
EPA. (2007g). Ozone Population Exposure Analysis for Selected Urban Areas (July 2007).
Office of Air Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-
07-010. Available at: http://epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html.
EPA. (2007h). Review of the National Ambient Air Quality Standards for Ozone: Assessment
of Scientific and Technical Information. OAQPS Staff paper. Office of Air Quality
Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-007a. Available
at: http://epa.gOv/ttn/naaqs/standards/ozone/s o3 cr sp.html.
EPA (20071). Review of the National Ambient Air Quality Standards for Lead: Policy
Assessment of Scientific and Technical Information. OAQPS Staff paper. Office of Air
Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-013.
Available at: http://www.epa.gov/ttn/naaqs/standards/pb/data/20071101_pb_staff.pdf.
EPA. (2008a). ISA for Oxides of Nitrogen - Health Criteria. National Center for
Environmental Assessment, Research Triangle Park, NC. Available at:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm? del d=l 94645.
EPA (2008b). Risk and Exposure Assessment to Support the Review of the NO2 Primary
National Ambient Air Quality Standard: First Draft. Office of Air Quality Planning and
Standards, Research Triangle Park, NC.
114
-------
EPA (2008c). Risk and Exposure Assessment to Support the Review of the NC>2 Primary
National Ambient Air Quality Standard: Second Draft. Office of Air Quality Planning
and Standards, Research Triangle Park, NC.
ESRI. (2008). ESRI Demographic Update Methodology: 2008/2013 (page 8). Available at:
http://www.esri.com/library/whitepapers/pdfs/demographic-update-methodology-
2008.pdf
Finlayson-Pitts, BJ and Pitts JN. (2000). Chemistry of the Upper and Lower Atmosphere.
Academic Press, San Diego CA. Page 17.
Folinsbee, LJ. (1992). Does nitrogen dioxide exposure increase airways responsiveness?
ToxicollndHealth. 8:273-283.
Fuhlbrigge, AL, Adams RJ, Guilbert TW, Grant E, Lozano P, Janson SL, Martinez F, Weiss KB,
Weiss ST. (2002). The Burden of Asthma in the United States: Level and Distribution
Are Dependent on Interpretation of the National Asthma Education and Prevention
Program Guidelines. Am JRespir Crit Care Med. 166:144-1049.
FFIWA. (2002). Highway Statistics 2002, Section V: Roadway Extent, Characteristics and
Performance. United States Department of Transportation, Federal Highway
Administration. Available at: http://www.fhwa.dot.gov/policy/ohim/hs02/re.htm.
FHWA. (2007). Draft Environmental Impact Statement (DEIS) for the Northwest Corridor
Project. Produced by the Georgia Department of Transportation (GDOT) and Georgia
Regional Transportation Authority (GRTA), in collaboration with the Federal Highway
Administration and Federal Transit Administration. Available at:
http://www.nwhovbrt.com/pages/deis.htmtftrs.
Gauderman, W J, Avol E, Gilliland F, Vora H, Thomas D, Berhane K, McConnell R, Kuenzli N,
Lurmann F, Rappaport E, Margolis H, Bates D, Peters J. (2004). The effect of air
pollution on lung development from 10 to 18 years of age. N. Engl. J. Med. 351:1057-
1067.
Gauderman, WJ, Avol E, Lurmann F, Kuenzli N, Gilliland F, Peters J, McConnell R. (2005).
Childhood asthma and exposure to traffic and nitrogen dioxide. Epidemiology 16:737-
743.
Ghenu, A, Rosant J-M, Sini J-F. (2007). Dispersion of pollutants and estimation of emissions in
a street canyon in Rouen, France. Environmental Modeling & Software. 23:314-321.
Gilbert, NL, Woodhouse S, Stieb DM, Brook JR. (2003). Ambient nitrogen dioxide and
distance from a major highway. Sci Total Environ. 312:43-46.
115
-------
Grilles, JA, Etyemezian V, Kuhns H, Nikolic D, Gillette DA. (2005). Effect of vehicle
characteristics on unpaved road dust emissions. Atmos Environ. 39:2341-2347.
Graham, SE and McCurdy T. (2004). Developing meaningful cohorts for human exposure
model s. J Expos Anal Environ Epidemiol. 14( 1 )23 -43.
Hanrahan, PL. (1999a). The plume volume molar ratio method for determining NO2/NOX ratios
in modeling. Part I: Methodology. J Air & Waste Manage Assoc. 49:1324-1331.
Hanrahan, PL. (1999b). The plume volume molar ratio method for determining NO2/NOX ratios
in modeling. Part II: Evaluation Studies. J Air & Waste Manage Assoc. 49:1332-1338.
Heeb, NV, Saxer CJ, Forss AM, Bruhlmann S. (2008). Trends of NO-. NO2-. and NHr
emissions from gasoline-fueled Euro-3- to Euro-4-passenger cars. Atmos Environ.
42:2543-2554
Hu PS and Reuscher TR. (2004). Summary of travel trends. 2001 National Household Travel
Survey. US Department of Transportation, Federal Highway Administration. Available
at: http://nhts.ornl.gov/2001/pub/STT.pdf
Ito, K, Thurston, GD, Silverman, RA. (2007). Characterization of PM2.5, gaseous pollutants,
and meteorological interactions in the context of time-series health effects models. J. of
Expos. Science and Environ. Epidemiology. 17:S45-S60.
Jaffe, DH, Singer ME, Rimm AA. (2003). Air pollution and emergency department visits for
asthma among Ohio Medicaid recipients, 1991-1996. Environ. Res. 91:21-28.
Jaffe, DH. (2008). Personal Communication to Scott Jenkins. See NCh Docket.
Jenkins, HS, Devalia JL, Mister RL, Bevan AM, Rusznak C, Davies RJ. (1999). The effect of
exposure to ozone and nitrogen dioxide on the airway response of atopic asthmatics to
inhaled allergen: dose- and time-dependent effects. Am. J. Respir. Crit. CareMed.
160:33-39.
Johnson, T, Mihlan G, LaPointe J, Fletcher K, Capel J. (1999). Estimation Of Carbon
Monoxide Exposures and Associated Carboxyhemoglobin Levels In Denver Residents
Using pNEM/CO (version 2.0). Prepared by ICF Kaiser Consulting Group for the U.S.
Environmental Protection Agency under Contract No. 68-D6-0064, March 1999.
116
-------
Torres, R, Magnussen H. (1990). Airways response of asthmatics after a 30 min exposure, at
resting ventilation, to 0.25 ppm NCh or 0.5 ppm SCh. Eur. Respir. J. 3:132-137.
Torres, R, Magnussen H. (1991). Effect of 0.25 ppm nitrogen dioxide on the airway response to
methacholine in asymptomatic asthmatic patients. Lung 169:77-85.
Torres, R, Nowak D, Grimminger F, Seeger W, Oldigs M, Magnussen H. (1995). The effect of 1
ppm nitrogen dioxide on bronchoalveolar lavage cells and inflammatory mediators in
normal and asthmatic subjects. Eur Respir J. 8:416-424.
Kim, S-Y, O'Neill M, Lee T-T, Cho Y, Kim T, and Kim H. (2007). Air pollution, socioeconomic
position, and emergency hospital visits for asthma in Seoul, Korea. Int. Arch. Occup.
Environ. Health 80:701-710.
Koontz, MD, Mehegan LL, and Nagda NL. (1992). Distribution and Use of Cooking Appliances
That Can Affect Indoor Air Quality, Report No. GRI-93/0013. Gas Research Institute,
Chicago, IL.
Kwon T. (2005). Development of a RIOPA database and evaluation of the effect of proximity
on the potential residential exposure to VOCS from ambient sources. PhD. dissertation.
Graduate School, New Brunswick, Rutgers, the State University of New Tersey and the
University of Medicine and Dentistry of New Tersey.
Langstaff, TE. (2007). OAQPS Staff Memorandum to Ozone NAAQS Review Docket (OAR-
2005-0172). Subject: Analysis of Uncertainty in Ozone Population Exposure Modeling.
[January 31,2007]. Available at:
http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html.
Linn, WS, Shamoo DA, Anderson KR, Peng RC, Avol EL, Hackney JD, Gong, H Tr. (1996).
Short-term air pollution exposures and responses in Los Angeles area schoolchildren. J.
Exposure Anal. Environ. Epidemiol. 6:449-472.
Linn, WS, Szlachcic Y, Gong H, Kinney PL, Berhane KT. (2000). Air pollution and daily
hospital admissions in metropolitan Los Angeles. Environ. Health Perspect. 108:427-
434.
MACTEC (2005). Evaluation of Bias in AERMOD-PVMRM, Final Report. Alaska DEC
Contract No. 18-9010-12. MACTEC Federal Programs, Inc., Research Triangle Park,
NC.
McCurdy, TR. (1994). Analysis of high 1 hour NO2 values and associated annual averages
using 1988-1992 data. Office of Air Quality Planning and Standards, Research Triangle
Park, NC. Available in Docket A-93-06.
117
-------
McCurdy, T, Glen G, Smith L, Lakkadi Y. (2000). The National Exposure Research
Laboratory' s Consolidated Human Activity Database. JExp Anal Environ Epidemiol.
10: 566-578.
McCurdy T and Graham SE. (2003). Using human activity data in exposure models: analysis of
discriminating factors. J Expos Anal Environ Epidemiol. 13:294-317.
Meng QY, Turpin BJ, Korn L, Weisel CP, Morandi M, Colome S, Zhang JJ, Stock T, Spektor D,
Winer A, Zhang L, Lee JH, Giovanetti R, Cui W, Kwon J, Alimokhtari S, Shendell D,
Jones J, Farrar C, Maberti S. (2004). Influence of ambient (outdoor) sources on
residential indoor and personal PM2.5 concentrations: Analyses of RIOPA data. JExp
Anal Environ Epidemiol. 15:17-28.
Mohsenin, V. (1987) Airway responses to nitrogen dioxide in asthmatic subjects. J. Toxicol.
Environ. Health 22:371-380.
Mortimer, KM, Neas LM, Dockery DW, Redline S, Tager IB. (2002). The effect of air
pollution on inner-city children with asthma. Eur Respir J. 19:699-705.
National Cooperative Highway Research Program (NCHRP) (1997). Recommended Procedure
for Long-Range Transportation Planning and Sketch Planning, NCHRP Report 387,
National Academy Press. 151 pp., ISBN No: 0-309-060-58-3.
National Heart, Lung, and Blood Institute (NHLBI). (2007) Expert panel report 3: guidelines for
the diagnosis and management of asthma. Bethesda, MD: U.S. Department of Health and
Human Services, National Institutes of Health, National Asthma Education and
Prevention Program; NUT Publication No. 07-4051.
NCDC. (2007). 2007 Local Climatological Data Annual Summary with Comparative Data.
Atlanta, Georgia (Katl). National Climate Data Center. ISSN 0198-1560.
New York Department of Health. (2006). A study of ambient air contaminants and asthma in
New York City, Final Report Part B: Air contaminants and emergency department visits
for asthma in the Bronx and Manhattan. Prepared for: The U.S. Department of Health
and Human Services, Agency for Toxic Substance and Disease Registry.
Oftedal, B, Brunekreef B, Nystad W, Madsen C, Walker SE, Nafstad P. (2008). Residential outdoor
air pollution and lung function in schoolchildren. Epidemiology 19:129-137.
118
-------
Ostro, B, Lipsett M, Mann J, Braxton-Owens H, White M. (2001). Air pollution and
exacerbation of asthma in African-American children in Los Angeles. Epidemiology
12:200-208.
Peel, JL, Tolbert PE, Klein M, Metzger KB, Flanders WD, Knox T, Mulholland JA, Ryan PB,
Frumkin H. (2005). Ambient air pollution and respiratory emergency department visits.
Epidemiology. 16:164-174.
Perry SG, Cimorelli AJ, Paine RJ, Erode RW, Weil JC, Venkatram A, Wilson RB, Lee RF,
Peters WD. (2005). AERMOD: A Dispersion Model for Industrial Source Applications.
Part II: Model Performance against 17 Field Study Databases. J. Appl. Meteor. 44:694-
708.
Persily, A, Gorfain J, Brunner G. (2005). Ventilation design and performance in U.S. office
buildings. ASHRAE Journal. April 2005, 30-35.
Pilotto, LS, Douglas RM, Attewell RG, Wilson SR (1997) Respiratory effects associated with
indoor nitrogen dioxide exposure in children. Int. J. Epidemiol. 26: 788-796.
Pilotto, LS, Nitschke M, Smith BJ, Pisaniello D, Ruffm RE, McElroy HJ, Martin J, Killer JE.
(2004). Randomized controlled trial of unflued gas heater replacement on respiratory
health of asthmatic schoolchildren. IntJEpidemiol. 33:208-214.
Pleijel, H, Karlsson GP, Gerdin EB. (2004). On the logarithmic relationship between NC>2
concentration and the distance from a highroad. Sci Total Environ. 332:261-264.
Riediker M, Williams R, Devlin R, Griggs T, Bromberg P. (2003). Exposure to
participate matter, volatile organic compounds, and other air pollutants inside
patrol cars. Environ Sci Technol. 37(10):2084-2093.
Rizzo (2008). Investigation of how distributions of hourly nitrogen dioxide concentrations have
changed over time in six cities. Nitrogen Dioxide NAAQS Review Docket (EPA-HQ-
OAR-2006-0922). Available at
http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_rea.html
Rodes, C, Sheldon L, Whitaker D, Clayton A, Fitzgerald K, Flanagan J, DiGenova F, Hering S,
Frazier C. (1998). Measuring Concentrations of Selected Air Pollutants Inside
California Vehicles. California Environmental Protection Agency, Air Resources Board.
Final Report, December 1998.
119
-------
Rodes, CE and Holland DM. (1981). Variations of NO, NO2 and O3 concentrations downwind
of a Los Angeles freeway. Atmos Environ. 15:243-250.
Roger, LJ, Horstman DH, McDonnell W, Kehrl H, Ives PJ, Seal E, Chapman R, Massaro E.
(1990). Pulmonary function, airway responsiveness, and respiratory symptoms in
asthmatics following exercise inNCh. Toxicol. Ind. Health 6:155-171.
Romieu, I, Ramirez-Aguilar M., Sienra-Monge JJ, Moreno-Macias, H, Del Rio-Navarro BE,
David G, Marzec J, Hernandez-Avila M, London S. (2006). GSTM1 and GSTP1 and
respiratory health in asthmatic children exposed to ozone. Eur. Respir. J. 28:953-959.
Rojas-Martinez, R, Perez-Padilla R, Olaiz-Fernandez G, Mendoza-Alvarado L, Moreno-Macias
H, Fortoul T, McDonnell W, Loomis D, Romieu I. (2007a) Lung function growth in
children with long-term exposure to air pollutants in Mexico City. Am. J. Respir. Crit.
CareMed. 176:377-384.
Rojas-Martinez, R, Perez-Padilla R, Olaiz-Fernandez G, Mendoza-Alvarado L, Moreno-Macias
H, Fortoul T, McDonnell W, Loomis D, Romieu I. (2007b) Lung function growth in
children with long-term exposure to air pollutants in Mexico City. Online data
supplement. Available: http://ajrccm.atsjournals.Org/cgi/data/176/4/377/DCl/l [3
October, 2007].
Roorda-Knape, MC, Janssen NAH, De Hartog JJ, Van Vliet PHN, Harssema H, Brunekreef B.
(1998). Air Pollution from traffic in city districts near major roadways. Atmos Environ.
32(11)1921-1930.
Rubenstein, I, Bigby BG, Reiss TF, Boushey HA Jr. (1990) Short-term exposure to 0.3 ppm
nitrogen dioxide does not potentiate airway responsiveness to sulfur dioxide in asthmatic
subjects. Am. Rev. Respir. Dis. 141:381-385.
Schildcrout, JS, Sheppard L, Lumley T, Slaughter JC, Koenig JQ, Shapiro GG. (2006).
Ambient air pollution and asthma exacerbations in children: an eight-city analysis. Am J
Epidemiol. 164:505-517.
Schwartz, J, Dockery DW, Neas LM, Wypij D, Ware JH, Spengler JD, Koutrakis P, Speizer FE,
Ferris BG, Jr. (1994). Acute effects of summer air pollution on respiratory symptom
reporting in children. Am J Respir Crit Care Med. 150:1234-1242.
320
-------
Shorter, JH, Herndon S, Zahniser MS, Nelson DD, Wormhoudt J, Demerjian KL, Kolb CE.
(2005). Real-Time Measurements of Nitrogen Oxide Emissions from In-Use New York
City Transit Buses Using a Chase Vehicle. Environ Sci Technol. 39:7991-8000.
Skuta and Wombold. (2008). The Secret Life of Polygons: Understanding the relative accuracy
of user-defined areas. Available at:
http://www.esri.com/news/arcuser/1008/polygons.html.
Spicer, CW, Kenny DV, Ward GF, Billick IH. (1993). Transformations, lifetimes, and sources
of NO2, HONO, and HNO3 in indoor environments. JAWMA. 43(11):1479-1485.
Stout, JW, Visness CM, Enright P, Lamm, C, Shapiro G, Gan, VN, Adams K, Mitchell HE.
(2006). Classification of asthma severity in children. Arch Pediatr Adolesc Med.
160:844-850.
Strand, V, Salomonsson P, Lundahl J, Bylin G. (1996). Immediate and delayed effects of
nitrogen dioxide exposure at an ambient level on bronchial responsiveness to histamine in
subj ects with asthma. Eur Respir J. 9:73 3 -740.
Strand, V, Rak S, Svartengren M, Bylin G. (1997). Nitrogen dioxide exposure enhances
asthmatic reaction to inhaled allergen in subjects with asthma. Am J Respir Crit Care
Med. 155:881-887.
Strand, V, Svartengren M, Rak S, Barck C, Bylin G. (1998). Repeated exposure to an ambient
level of NC>2 enhances asthmatic response to nonsymptomatic allergen dose. Eur Respir
J. 12:6-12.
Suh, H. (2008). Personal Communication, August 27, 2008, regarding personal exposure data
for Atlanta.
Thompson, R. (2008). Nitrogen Dioxide (NC^) Descriptive Statistics Tables. Memo to the NC>2
NAAQS docket. Available at
http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_rea.html
Thompson, R. and Jenkins SM (2008). Air Quality Statistics Associated with Nitrogen Dioxide
Epidemiologic Studies. Memo to the NC>2 NAAQS docket. Available at
http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_rea.html
Tolbert, P. (2008a). Personal communication (email) to H. Richmond, U.S. EPA - "Atlanta
Emergency Department Visit and Air Quality Data used in Tolbert et al. (2007)," May
30.
321
-------
Tolbert, P. (2008b). Personal communication (email) to H. Richmond, U.S. EPA - "Response
to Harvey Richmond regarding CASAC comments on SOPHIA ED study incidence data
for the NC>2 risk and exposure assessment," November 20.
Tolbert. PE. Klein M. Peel JL. Sarnat SE. Sarnat JA. (2007). Multipollutant modeling issues in
a study of ambient air quality and emergency department visits in Atlanta. J Expos Sci
Environ Epidemiol. 17S2:S29-35.
Tunnicliffe, WS, Burge PS, Ayres JG. (1994) Effect of domestic concentrations of nitrogen
dioxide on airway responses to inhaled allergen in asthmatic patients. Lancet 344:1733-
1736.
US DOT. (2007). Part 3-The Journey To Work files. Bureau of Transportation Statistics
(BTS). Available at: http://transtats.bts.gov/.
van Gent, R, van der Ent K, van Essen-Zandvliet LEM, Rovers MM, Kimpen ILL, de Meer G,
Klijn PHC. (2007). No difference in physical activity in (un)diagnosed asthma and
healthy controls. PediatricPulmonology. 42:1018-1023.
Vardoulakis, S, Gonzalez-Flesca N, Fisher BEA, Pericleous K. (2004). Spatial variability of air
pollution in the vicinity of a permanent monitoring station in central Paris. Atmos.
Environ. 39:2725-2736.
Watkins, N. and Thompson, R. (2008). NOX Network Review and Background. Memo to the
NO2 NAAQS docket.
Weisel CP, Zhang JJ, Turpin BJ, Morandi MT, Colome S, Stock TH, Spektor DM, Korn L,
Winer A, Alimokhtari S, Kwon J, Mohan K, Harrington R, Giovanetti R, Cui W, Afshar
M, Maberti S, Shendell D. (2004). Relationship of Indoor, Outdoor and Personal Air
(RIOPA) study; study design, methods and quality assurance / control results. JExp Anal
Environ Epidemiol. 15:123-137.
Westerdahl, D, Fruin S, Sax T, Fine PM, Sioutas C. (2005). Mobile platform measurements of
ultrafme particles and associated pollutant concentrations on freeways and residential
streets in Los Angeles. Atmos Environ. 39:3597-3610.
Wheeler, A, Zanobetti A, Gold DR, Schwartz J, Stone P, Suh HH. (2006). The relationship
between ambient air pollution and heart rate variability differs for individuals with heart
and pulmonary disease. Environ Health Perspect. 114(4):560-6.
322
-------
WHO (2008). Harmonization Project Document No. 6. Part 1: Guidance document on
characterizing and communicating uncertainty in exposure assessment. Available at:
http://www.who.int/ipcs/methods/harmonization/areas/exposure/en/.
Witten, A, Solomon C, Abbritti E, Arjomandi M, Zhai W, Kleinman M, Balmes J. (2005).
Effects of nitrogen dioxide on allergic airway responses in subjects with asthma. J.
Occup. Environ. Med. 47:1250-1259.
Wolff, GT. (1993). Letter to EPA Administrator Carol Browner: "CASAC Closure on the Air
Quality Criteria Document for Oxides of Nitrogen." EPA-SAB-CASAC-LTR-93-015,
September 30.
Wolff, GT. (1995). Letter to EPA Administrator Carol Browner: "CAS AC Review of the Staff
Paper for the Review of the National Ambient Air Quality Standards for Nitrogen
Dioxide: Assessment of Scientific and Technical Information." EPA-SAB-CASAC-
LTR-95-004, August 22.
Yao, X, Lau NT, Chan CK, Fang M. (2005). The use of tunnel concentration profile data to
determine the ratio of NO2/NOX directly emitted from vehicles. Atmos Chem Phys
Discuss. 5:12723-12740. Available at: http://www.atmos-chem-phys-
discuss.net/5/12723/2005/acpd-5-12723-2005.pdf.
Zhu Y, Hinds WC, Kim S, Shen S, Sioutas C. (2002). Study of ultrafine particles near a major
highway with heavy-duty diesel traffic. Atmos Environ. 36:4323-4335.
323
-------
United States Office of Air Quality Planning and Standards EPA-452/R-08-008a
Environmental Protection Health and Environmental Impacts Division November 2008
Agency Research Triangle Park, NC
324
------- |