*«< PRC*4'
Risk and Exposure Assessment to Support the
Review of the NO2 Primary National Ambient
Air Quality Standard: Draft Technical Support
Document (TSD)
-------
EPA-452/P-08-002
April 2008
Risk and Exposure Assessment to Support the
Review of the NO2 Primary National Ambient
Air Quality Standard: Draft Technical Support
Document (TSD)
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, and was written with
support from technical documents from ICF International (through Contract No. EP-D-06-115).
Any opinions, findings, conclusions, or recommendations are those of the authors and do not
necessarily reflect the views of the EPA or ICF International. This is the first draft document
submitted to support the NC>2 Risk and Exposure Assessment for review and comment from the
Clean Air Scientific Advisory Committee (CASAC) and the general public. Any questions
concerning this draft document should be addressed to Dr. Stephen E. Graham, U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards, C504-06,
Research Triangle Park, North Carolina 27711 (email: graham.stephen@epa.gov).
-------
Table of Contents
List of Tables v
List of Figures vii
1 Introduction 1
2 Air Quality Characterization 2
2.1 Air Quality Data Screen 2
2.1.1 Introduction 2
2.1.2 Approach 2
2.1.3 Results 3
2.2 Selection of Locations 5
2.2.1 Introduction 5
2.2.2 Approach 5
2.2.3 Results 5
2.3 Ambient Monitor Characterization 7
2.3.1 Introduction 7
2.3.2 Approach 7
2.3.3 Results 8
2.4 Spatial and Temporal Air Quality Analyses 10
2.4.1 Introduction 10
2.4.2 Approach 10
2.4.3 Spatial Results 11
2.4.4 Temporal Results 16
2.5 Air Quality Simulation 22
2.5.1 Introduction 22
2.5.2 Approach 23
2.6 Method for Estimating On-Road Concentrations 27
2.6.1 Introduction 27
2.6.2 Derivation of On-Road Factors 28
2.6.3 Application of On-Road Factors 30
2.6.4 Interpretation of Estimated On-Road Concentrations 31
2.7 Estimation of Potential Health Effect Benchmark Exceedances 33
2.7.1 Introduction 33
2.7.2 Approach 33
2.7.3 Results 34
2.7.3.1 Air Quality Monitoring Data As Is 34
2.7.3.2 Simulated Air Quality Data to Just Meet The Current Standard 39
2.7.3.3 Simulated On-Road Concentrations, Air Quality Data As Is 43
2.7.3.4 Simulated On-Road Concentrations, Simulated Air Quality Data To Just Meet
The Current Standard 48
2.8 Variability and Uncertainty 52
2.8.1 Air Quality Data 52
2.8.2 Measurement Technique for Ambient NO2 52
2.8.3 Temporal Representation 53
2.8.4 Spatial Representation 53
2.8.5 Air Quality Adjustment Procedure 53
-------
2.8.6 On-Road Concentration Simulation 54
2.8.7 Health Benchmark 56
2.9 References 57
Exposure Assessment and Health Risk Characterization 59
3.1 Introduction 59
3.1.1 Selection of Study Areas 59
3.1.2 Exposure Periods 60
3.1.3 Populations Analyzed 60
3.2 Dispersion Modeling 60
3.2.1 Meteorological Inputs 61
3.2.1.1 Data Selection 61
3.2.2 Surface Characteristics and Land Use Analysis 66
3.2.3 Meteorological Analysis 67
3.2.4 On-Road Emissions Preparation 69
3.2.4.1 Philadelphia County Data Sources 69
3.2.5 Stationary Sources Emissions Preparation 76
3.2.5.1 Philadelphia Data Sources 76
3.2.5.2 Data Source Alignment 76
3.2.6 Fugitive and Airport Emissions Preparation 82
3.2.6.1 Philadelphia County 82
3.2.7 Receptor Locations 86
3.2.7.1 Philadelphia County 86
3.2.8 Other Modeling Specifications 88
3.2.9 Air Quality Concentration Estimation 89
3.3 Human Exposure Modeling using APEX 90
3.3.1 History 90
3.3.2 Model Overview 91
3.3.3 Study Area Descriptions 92
3.3.3.1 Air Quality Data 93
3.3.3.2 Meteorological Data 93
3.3.4 Simulated Individuals 94
3.3.4.1 Population Demographics 94
3.3.4.2 Commuting 95
3.3.4.3 Profile Functions 97
3.3.4.4 Asthma Prevalence Rates 97
3.3.5 Activity Pattern Sequences 99
3.3.5.1 Personal Information file 100
3.3.5.2 Diary Events file 100
3.3.5.3 Construction of Longitudinal Activity Sequences 101
3.3.6 Calculating Microenvironmental Concentrations 103
3.3.6.1 Mass Balance Model 103
3.3.6.2 Factors Model 104
3.3.6.3 Microenvironments Modeled 105
3.3.6.4 Microenvironment Descriptions 105
3.3.6.5 Microenvironment Mapping 112
3.3.7 Exposure Calculations 114
in
-------
3.3.8 Exposure Model Output 115
3.4 Exposure Modeling Results 117
3.4.1 Overview 117
3.4.2 Annual Average Exposure Concentrations (as is) 117
3.4.3 One-Hour Exposures (as is) 120
3.4.3.1 Maximum Estimated Exposure Concentrations 121
3.4.3.2 Number of Estimated Exposures above Selected Levels 121
3.4.3.3 Number of Repeated Exposures Above Selected Levels 127
3.4.4 One-Hour Exposures Associated with Just Meeting the Current Standard 129
3.4.4.1 Number of Estimated Exposures above Selected Levels 129
3.4.4.2 Number of Repeated Exposures Above Selected Levels 130
3.5 Variability and Uncertainty 131
3.5.1 Introduction 131
3.5.2 Input Data Evaluation 132
3.5.2.1 Meteorological Data 132
3.5.2.2 Air Quality Data 132
3.5.2.3 Population and Commuting Data 133
3.5.2.4 Activity Pattern Data 133
3.5.2.5 Air Exchange Rates 134
3.5.2.6 Air Conditioning Prevalence 135
3.6 References 137
IV
-------
List of Tables
Table 1. Example of monitors IDs and years of operation using the Boston CMS A 3
Table 2. Counts of complete site-years of NO2 monitoring data 4
Table 3. Locations selected for NO2 Air Quality Characterization, associated abbreviations, and
values of selection criteria 6
Table 4. Distribution of the distance (m) of ambient monitors to the nearest major road in
selected locations 8
Table 5. Distribution of the distance (m) of ambient monitors to stationary sources with NOX
emissions >5 tons per year (tpy) and within a 10 kilometers (km)1 radius 9
Table 6. Distribution of NOX emissions from stationary sources within 10 kilometers (km) of
monitoring site, where emissions were >5 tons per year1 9
Table 7. Statistical test results for spatial comparisons of all location parameter distributions.. 14
Table 8. Statistical test results for spatial comparisons of within location parameter distributions.
15
Table 9. Statistical test results for temporal comparisons of all location parameter distributions.
19
Table 10. Maximum annual average NO2 concentrations and air quality adjustment factors (F) to
just meet the current standard, historic monitoring data 25
Table 11. Maximum annual average NO2 concentrations and air quality adjustment factors (F) to
just meet the current standard, recent monitoring data 26
Table 12. Reviewed studies containing NO2 concentrations at a distance from roadways 28
Table 13. Monitoring site-years and annual average NO2 concentrations for two monitoring
periods, historic and recent air quality data (as is) 36
Table 14. Number of exceedances of short-term (1-hour) potential health effect benchmark
levels in a year, 1995-2000 historic NO2 air quality (as is) 37
Table 15. Number of exceedances of short-term (1-hour) potential health effect benchmark
levels in a year, 2001-2006 recent NO2 air quality (as is) 38
Table 16. Estimated annual average NO2 concentrations for two monitoring periods, historic and
recent air quality data adjusted to just meet the current standard (0.053 ppm annual
average) 40
Table 17. Estimated number of exceedances of short-term (1-hour) potential health effect
benchmark levels in a year, 1995-2000 NO2 air quality adjusted to just meet the
current standard (0.053 ppm annual average) 41
Table 18. Estimated number of exceedances of short-term (1-hour) potential health effect
benchmark levels in a year, 2001-2006 NO2 air quality adjusted to just meet the
current standard (0.053 ppm annual average) 42
Table 19. Estimated annual average on-road concentrations for two monitoring periods, historic
and recent ambient air quality (as is) 45
Table 20. Estimated number of exceedances of short-term (1-hour) potential health effect
benchmark levels in a year on-roads, 1995-2000 historic NO2 air quality (as is) 46
Table 21. Estimated number of exceedances of short-term (1-hour) potential health effect
benchmark levels in a year on-roads, 2001-2006 historic NO2 air quality (as is) 47
Table 22. Estimated annual average on-road concentrations for two monitoring periods, air
quality data adjusted to just meet the current standard (0.053 ppm annual average).. 49
-------
Table 23. Estimated number of exceedances of short-term (1-hour) potential health effect
benchmark levels in a year on-roads, 1995-2000 historic NC>2 air quality adjusted to
just meet the current standard (0.053 ppm annual average) 50
Table 24. Estimated number of exceedances of short-term (1-hour) potential health effect
benchmark levels in a year on-roads, 2001-2006 recent NC>2 air quality adjusted to just
meet the current standard (0.053 ppm annual average) 51
Table 25. Summary of qualitative uncertainty analysis for the air quality characterization 56
Table 26. Number of AERMET raw hourly surface meteorology observations and percent
acceptance rate, 2001-2003.a 62
Table 27. Number of calms reported by AERMET by year and location 62
Table 28. Number and AERMET acceptance rate of upper-air observations 2001-2003 64
Table 29. Seasonal specifications by study location 66
Table 30. Monthly precipitation compared to NCDC 30-year climatic normal, 2001-2003 68
Table 31. Hourly scaling factors (in percents) applied to Philadelphia County AADT volumes.
71
Table 32. Seasonal scaling factors applied to Philadelphia County AADT volumes 72
Table 33. Signals per mile, by link type, applied to Philadelphia County AADT volumes 72
Table 34. Statistical summary of AADT volumes (one direction) for Philadelphia County
AERMOD simulations 72
Table 35. Average calculated speed by link type 75
Table 36. On-road area source sizes 75
Table 37. Combined stacks parameters for stationary NOX emission sources in Philadelphia
County 78
Table 38. Matched stacks between the CAMD and NEI database 80
Table 39. Emission parameters for the three Philadelphia County fugitive NOX area emission
sources 83
Table 40. Philadelphia International airport (PHL)NOX emissions 85
Table 41. Philadelphia CMSA NOX monitors 86
Table 42. Comparison of ambient monitoring and AERMOD predicted NC>2 concentrations. ... 89
Table 43. The meteorological stations used for each study area 93
Table 44. Examples of profile variables in APEX 94
Table 45. Asthma prevalence rates by age and gender for 4 regions 98
Table 46. Summary of activity pattern studies used in CHAD 101
Table 47. Mass balance model parameters 103
Table 48. Factors model parameters 104
Table 49. List of microenvironments and calculation methods used 105
Table 50. Air conditioning prevalence estimates with 95% confidence intervals 106
Table 51. Geometric means (GM) and standard deviations (GSD) for air exchange rates by city,
A/C type, and temperature range 107
Table 52. Probability of gas stove cooking by hour of the day 110
Table 53. Mapping of CHAD activity locations to APEX microenvironments 112
Table 54. Adjustment factors and potential health effect benchmark levels used by APEX to
simulate just meeting the current standard 115
Table 55. Example of APEX output files 116
VI
-------
List of Figures
Figure 1. Spatial distributions of annual mean NC>2 ambient monitoring concentrations for
selected CMSA locations, years 1995-2006 12
Figure 2. Spatial distributions of annual mean NC>2 ambient monitoring concentrations for
selected MSA and grouped locations, years 1995-2006 12
Figure 3. Spatial distributions of hourly NC>2 ambient monitoring concentrations for selected
CMSA locations, years 1995-2006 13
Figure 4. Spatial distributions of hourly NC>2 ambient concentration for selected CMSA
locations, years 1995-2006 14
Figure 5. Spatial distribution of annual average NC>2 concentrations among 10 monitoring sites
in Philadelphia CMSA, years 1995-2006 15
Figure 6. Temporal distributions of annual mean NC>2 concentrations for all monitors, years
1995-2006 17
Figure 7. Temporal distributions of annual mean NC>2 concentrations for the Philadelphia
CMSA, years 1995-2006 18
Figure 8. Temporal distribution of hourly NC>2 concentrations in the Los Angeles CMSA, years
1995-2006 19
Figure 9. Temporal distributions of hourly NC>2 concentrations in the Jacksonville MSA, years
1995-2006, one monitor 20
Figure 10. Temporal distribution of annual average NC>2 concentrations in the Not MSA group
location, years 1995-2006 21
Figure 11. Trends in hourly and annual average NC>2 ambient monitoring concentrations and
their associated coefficients of variation (COV) for all monitors, years 1995-2006... 23
Figure 12. Distribution of on-road factors (Cv/Cb or m) for two season groups 30
Figure 13. Example of Light- and heavy-duty vehicle NOX emissions grams/mile (g/mi) for
arterial and freeway functional classes, 2001 74
Figure 14. Differences in facility-wide annual NOx emission totals between NEI and CAMD
data bases for Philadelphia County 2002 82
Figure 15. Locations of the four ancillary area sources. Also shown are centroid receptor
locations 84
Figure 16. Centroid locations within fixed distances to major point and mobile sources 87
Figure 17. Frequency distribution of distance between each Census receptor and its nearest road-
centered receptor 88
Figure 18. Example of a profile function file for A/C prevalence 97
Figure 19. Example input file from APEX for Indoors-residence microenvironment 108
Figure 20. Example input file from APEX for all Indoors microenvironments, other than
Indoors-residence Ill
Figure 21. Example input file from APEX for outdoor near road microenvironment Ill
Figure 22 . Distribution of AERMOD predicted annual average NC>2 concentrations at each of
the 16,857 receptors in Philadelphia County for years 2001-2003 118
Figure 23. Estimated annual average total NO2 exposure concentrations for all simulated persons
in Philadelphia County, using modeled 2001-2003 air quality (as is), with modeled
indoor sources 119
VII
-------
Figure 24. Comparison of AERMOD predicted and ambient monitoring annual average NO2
concentrations (as is) and APEX exposure concentrations (with and without modeled
indoor sources) in Philadelphia County for year 2002 120
Figure 25. Estimated maximum NO2 exposure concentration for all simulated persons in
Philadelphia County, using modeled 2001-2003 air quality (as is), with and without
modeled indoor sources. Values above the 99th percentile are not shown 121
Figure 26. Estimated number of all simulated asthmatics in Philadelphia County with at least
one NO2 exposure at or above the potential health effect benchmark levels, using
modeled 2001-2003 air quality (as is), with modeled indoor sources 122
Figure 27. Estimated number of simulated asthmatic children in Philadelphia County with at
least one NO2 exposure at or above the potential health effect benchmark levels, using
modeled 2001-2003 air quality (as is), with modeled indoor sources 122
Figure 28. Comparison of the estimated number of all simulated asthmatics in Philadelphia
County with at least one NO2 exposure at or above potential health effect benchmark
levels, using modeled 2002 air quality (as is) , with and without modeled indoor
sources 123
Figure 29. Fraction of time all simulated persons in Philadelphia County spend in the twelve
microenvironments associated with the potential NO2 health effect benchmark levels,
a) > 200 ppb, b) > 250 ppb, and c) > 300 ppb, year 2002 simulation with indoor
sources 125
Figure 30. Fraction of time all simulated persons in Philadelphia County spend in the twelve
microenvironments associated with the potential NO2 health effect benchmark levels,
a) > 200 ppb, b) > 250 ppb, and c) > 300 ppb, year 2002 simulation without indoor
sources 126
Figure 31. Estimated percent of all asthmatics in Philadelphia County with repeated NO2
exposures above potential health effect benchmark levels, using 2003 modeled air
quality (as is), with modeled indoor sources 128
Figure 32. Estimated percent of all asthmatics in Philadelphia County with repeated NO2
exposures above potential health effect benchmark levels, using modeled 2002 air
quality (as is), with and without indoor sources 128
Figure 33. Estimated percent of all asthmatics in Philadelphia with at least one exposure at or
above the potential health effect benchmark level, using modeled 2001-2003 air
quality just meeting the current standard, with modeled indoor sources 129
Figure 34. Estimated number of all asthmatics in Philadelphia with at least one exposure at or
above the potential health effect benchmark level, using modeled 2002 air quality just
meeting the current standard, with and without modeled indoor sources 130
Figure 35. Estimated percent of asthmatics in Philadelphia County with repeated exposures
above health effect benchmark levels, using modeled 2002 air quality just meeting the
current standard, with and without modeled indoor sources 131
Figure 36. Geometric mean and standard deviation of air exchange rate bootstrapped for Los
Angeles residences with A/C, temperature range from 20-25 degrees centigrade (from
US EPA, 2007d) 135
List of Appendices
Appendix A. Ambient Monitor Characterization
Appendix B. Temporal NO2 Air Quality Characterization
VIM
-------
Appendix C. Spatial NC>2 Air Quality Characterization
Appendix D. Technical Memorandum on Regression Modeling
Appendix E. Technical Memorandum on Land Use and Surface Analysis
Appendix F. Technical Memorandum on Longitudinal Diary Construction Approach
Appendix G. Exposure and Risk Results for All Asthmatics and Asthmatic Children
IX
-------
List of Acronyms/Abbreviations
AADT
A/C
AER
AERMOD
AHS
APEX
ANOVA
AQS
BRFSS
CAMD
CASAC
CDC
CHAD
CMSA
CO
COV
CTPP
DVRPC
EPA
EOC
GM
GSD
hr
ID
ISA
ISH
km
L95
m
ME
max
med
min
MSA
NAAQS
NAICS
NCEA
NEI
NEM
NCDC
NHAPS
NHIS
NO2
NOX
NWS
Annual average daily traffic
Air conditioning
Air exchange rate
American Meteorological Society (AMS)/EPA Regulatory Model
American Housing Survey
EPA's Air Pollutants Exposure model, version 4
One-way analysis of variance
EPA's Air Quality System
Behavioral Risk Factor Surveillance System
EPA's Clean Air Markets Division
Clean Air Scientific Advisory Committee
Centers for Disease Control
EPA's Consolidated Human Activity Database
Consolidated metropolitan statistical area
Carbon monoxide
Coefficient of Variation
Census Transportation Planning Package
Delaware Valley Regional Planning Council
United States Environmental Protection Agency
Exposure of Concern
Geometric mean
Geometric standard deviation
Hour
Identification
Integrated Science Assessment
Integrated Surface Hourly Database
Kilometer
Lower limit of the 95th confidence interval
Meter
Microenvironment
Maximum
Median
Minimum
Metropolitan statistical area
National Ambient Air Quality Standards
North American Industrial Classification System
National Center for Environmental Assessment
National Emissions Inventory
NAAQS Exposure Model
National Climatic Data Center
National Human Activity Pattern Study
National Health Interview Survey
Nitrogen dioxide
Oxides of nitrogen
National Weather Service
-------
Os Ozone
OAQPS Office of Air Quality Planning and Standards
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
ppm Parts per million
PRB Policy-Relevant Background
PROX Proximity factor
PVMRM Plume Volume Molar Ratio Method
RECS Residential Energy Consumption Survey
SAS Statistical Analysis Software
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
XI
-------
i 1 Introduction
2 The U.S. Environmental Protection Agency (EPA) is presently conducting a review of the
3 national ambient air quality standards (NAAQS) for nitrogen dioxide (NO2). Sections 108 and
4 109 of the Clean Air Act (The Act) govern the establishment and periodic review of the
5 NAAQS. These standards are established for pollutants that may reasonably be anticipated to
6 endanger public health and welfare, and whose presence in the ambient air results from
7 numerous or diverse mobile or stationary sources. The NAAQS are to be based on air quality
8 criteria, which are to accurately reflect the latest scientific knowledge useful in indicating the
9 kind and extent of identifiable effects on public health or welfare that may be expected from the
10 presence of the pollutant in ambient air. The EPA Administrator is to promulgate and
11 periodically review, at five-year intervals, primary (health-based) and secondary (welfare-based)
12 NAAQS for such pollutants. Based on periodic reviews of the air quality criteria and standards,
13 the Administrator is to make revisions in the criteria and standards and promulgate any new
14 standards as may be appropriate. The Act also requires that an independent scientific review
15 committee advise the Administrator as part of this NAAQS review process, a function now
16 performed by the Clean Air Scientific Advisory Committee (CASAC).
17
18 This report document, in detail, the methodology and input data used in the risk and exposure
19 assessment for NC>2 conducted in support of the current review of the NC>2 NAAQS.
20 Specifically, this report includes the following:
21
22 • Description of the areas assessed and populations considered
23 • Summary of the air quality modeling methodology and associated input data
24 • Description of the inhalation exposure model and associated input data
25 • Evaluation of estimated NC>2 exposures
26 • Assessment of the quality and limitations of the input data for supporting the goals of
27 the NC>2 NAAQS exposure analysis.
28
-------
i 2 Air Quality Characterization
2
3 2.1 Air Quality Data Screen
4 2.1.1 Introduction
5 The current NC>2 standard of 53 ppb annual arithmetic average was set in 1971 and has been
6 retained since by subsequent reviews (i.e., 1985, 1995). Minor revisions to the standard made in
7 1985 included an explicit rounding convention, stated annual averages would be determined on a
8 calendar year basis, and indicated an explicit 75% completeness requirement for monitoring (60
9 FR 52874). Each of these components of the standard were considered in characterizing the air
10 quality monitoring data, beginning first with the selection of valid data.
11 2.1.2 Approach
12 NC>2 air quality data from years 1995 through 2006 and associated documentation were
13 downloaded from EPA's Air Quality System (US EPA, 2007a; 2007b). As of the date of the
14 analyses performed, hourly measurements for year 2006 were only available for January 1
15 through October 31, 2006. A site was defined by the state, county, site code, and parameter
16 occurrence code (POC), which gives a 10-digit monitor ID code. The POC identifies collocated
17 measurements at the same monitoring location, so that each measuring instrument is treated as a
18 different site. Typically there was only one POC at a given monitoring location.
19
20 As required by the NO2 NAAQS, a valid year of monitoring data is needed to calculate the
21 annual average concentration. A valid year at a monitoring site is comprised of 75% of valid
22 days in a year, with at least 18 hourly measurements for a valid day (thus at least 274 or 275
23 valid days depending on presence of a leap year, a minimum of 4,932 or 4,950 hours). This
24 served as a screening criterion for data to be used for analysis.
25
26 Site-years of data are the total numbers of years the collective monitors in a location were in
27 operation. For example, from years 1995-2006, the Boston CMS A had 27 total monitors in
28 operation, some of which did not contain sufficient numbers of monitoring values, while others
29 contained upwards of 11 years (Table 1). Thus in summing the number of operating years, this
30 particular location contained a total of 105 site-years of data across the monitoring period.
31
32 In all of the subsequent analyses, where hourly values were missing they were treated as such.
33 Reported values of zero (0) concentration were also retained as is. For certain illustrations,
34 values of zero were substituted with 0.5 ppb, derived from one-half the lowest recorded 1-hour
35 concentration (1 ppb).
36
37
38
39
40
41
42
-------
1 Table 1. Example of monitors IDs and years of operation using the Boston CMSA.
Monitor ID
2303130021
2500510021
2500510051
2500900051
2500920061
2500940041
2500950051
2502100091
2502130031
2502500021
250250021 1
2502500351
2502500361
2502500401
2502500411
2502500421
2502510031
2502700201
2502700231
3301100161
3301100191
3301100201
3301110111
3301500091
3301500131
3301500141
3301500151
Complete
Incomplete
Year of monitoring (1995-2006)
95
i
c
c
c
c
c
c
c
c
c
c
c
c
12
1
96
c
i
c
c
c
c
c
c
c
c
c
10
1
97
c
c
c
c
c
c
c
c
c
c
c
11
0
98
c
c
c
c
c
c
c
c
c
c
c
i
11
1
99
i
i
i
i
c
c
c
i
c
c
i
i
c
c
7
7
00
c
i
i
i
c
c
c
i
i
c
c
i
c
7
6
01
c
i
c
c
c
c
c
c
c
c
i
i
i
c
i
10
5
02
c
i
i
c
i
i
c
c
c
i
c
c
c
c
c
10
5
03
c
c
i
i
i
c
i
c
i
c
i
i
i
5
8
04
i
c
i
i
i
c
c
i
c
c
c
i
c
7
6
05
i
c
i
c
i
c
c
i
c
c
c
i
c
8
5
06
c
i
c
i
c
i
i
c
c
c
i
c
7
5
Totals
Complete
7
0
2
0
10
5
2
1
0
11
8
1
1
11
1
6
5
8
3
4
1
5
0
5
4
3
1
105
Incomplete
4
1
4
1
2
7
1
0
5
1
0
0
0
1
7
1
0
1
0
1
2
1
3
2
2
1
2
50
Notes:
c = met criteria for valid year of monitoring data.
i = did not met criteria for valid year of monitoring data.
3 2.1.3 Results
4 Of a total of 5,243 site-years of data in the entire NC>2 1-hour concentration database, 1,039
5 site-years did not meet the above criterion and were excluded from any further analyses. In
6 addition, since shorter term average concentrations are of interest, the remaining site-years of
7 data were further screened for 75% completeness on hourly measures in a year (i.e., containing a
8 minimum of 6,570 or 6,588, depending on presence of a leap year). Twenty-seven additional
9 site-years were excluded, resulting in 4,177 complete site-years in the analytical database. Table
10 2 provides a summary of the site-years included in the analysis, relative to those excluded, by
11 location and by two site-year groupings.1 Location selection is defined in the Section 1.2.
14 of 18 named locations and the 2 grouped locations contained enough data to be considered valid for year 2006.
-------
1
2
Table 2. Counts of complete site-years of NO2 monitoring data.
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Total
Com
1995-2000
58
47
11
26
12
193
24
93
46
69
24
26
14
6
16
22
6
56
1135
200
Number of
Dlete
2001-2006
47
36
11
10
12
177
20
81
39
66
29
0
30
4
35
27
6
43
1177
243
4177
Site-Years
Incorr
1995-2000
16
20
2
10
4
16
1
12
6
21
5
4
11
0
4
8
0
3
249
112
iplete
2001-2006
34
22
2
4
1
19
4
24
8
18
1
4
0
2
9
25
0
9
235
141
1066
% Cor
1995-2000
78%
70%
85%
72%
75%
92%
96%
89%
88%
77%
83%
87%
56%
100%
80%
73%
100%
95%
82%
64%
iplete
2001-2006
58%
62%
85%
71%
92%
90%
83%
77%
83%
79%
97%
0%
100%
67%
80%
52%
100%
83%
83%
63%
80%
-------
i 2.2 Selection of Locations
2 2.2.1 Introduction
3 The next step in this analysis was to identify similarities and differences in air quality among
4 locations for the purpose of either aggregating or segregating data using a combination of
5 descriptive statistics and health based criteria. Location in this context would include a
6 geographic area that encompasses more than a single air quality monitor (e.g., particular city,
7 consolidated metropolitan statistical area or CMSA).
8 2.2.2 Approach
9 Criteria were established for selecting sites with high annual means and/or frequent
10 exceedances of potential health effect benchmarks. Selected locations were those that had a
11 maximum annual mean NO2 level at a particular monitor greater than or equal to 25.7 ppb, which
12 represents the 90th percentile across all locations and site-years, and/or had at least one reported
13 1-hour NC>2 level greater than or equal to 200 ppb, the lowest level of the potential health effect
14 benchmarks. A location in this context would include a geographic area that encompasses more
15 than a single air quality monitor (e.g., particular city, metropolitan statistical area (MSA), or
16 consolidated metropolitan statistical area or CMSA). First, all monitors were identified as either
17 belonging to a CMSA, a MSA, or neither. Then, locations of interest were identified through
18 statistical analysis of the ambient NC>2 air quality data for each site within a location.
19 2.2.3 Results
20 Fifteen locations met both selection criteria, that is, having at least one site-year annual mean
21 above 25.72 ppb and at least one exceedance of 200 ppb. Upon further analysis of the more
22 recent ambient data (2001-2006), four additional locations were observed to have met at least
23 one of the criteria (either high annual mean and/or at least one exceedance of 200 ppb). New
24 Haven, CT, while meeting the earlier criteria, did not have any recent exceedances of 200 ppb
25 and contained one of the lowest maximum concentration-to-mean ratios, therefore was not
26 separated out as a specific location. Thus, 14 locations were retained from the initial selection
27 and 4 locations selected from a second screening to provide additional geographical
28 representation. In addition to these 18 specific locations, the remaining sites were grouped into
29 two broad location groupings. The Other CMSA location contains all the other sites that are in
30 MSAs or CMSAs but are not in any of the 18 specified locations. The Not MSA location
31 contains all the sites that are not in an MSA or CMSA. The selected locations are summarized in
32 Table3.
33
34 The final database for analysis included air quality data from a total of 205 monitors within
35 the named locations, 331 monitors in the Other CMSA group, and 92 monitors in the Not MSA
36 group. Again, the monitors that were retained contained the criteria for estimating a valid annual
37 average concentration described above.
-------
Table 3. Locations selected for NO2 Air Quality Characterization, associated abbreviations, and values of selection criteria.
Location
Type1 Code Description Abbreviation
CMSA*
CMSA
CMSA*
CMSA*
CMSA*
CMSA*
CMSA
CMSA*
CMSA*
CMSA*
MSA*
MSA*
MSA*
MSA
MSA*
MSA*
MSA
MSA*
MSA/CMSA
-
1122
1602
1692
2082
2162
4472
4992
5602
6162
8872
0520
1720
2320
3600
4120
6200
6520
7040
-
-
Boston-Worcester-Lawrence, MA-NH-ME-CT
Chicago-Gary-Kenosha, IL-IN-WI
Cleveland-Akron, OH
Denver-Boulder-Greeley, CO
Detroit-Ann Arbor-Flint, Ml
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
Washington-Baltimore, DC-MD-VA-WV
Atlanta, GA
Colorado Springs, CO
El Paso, TX
Jacksonville, FL
Las Vegas, NV-AZ
Phoenix-Mesa, AZ
Provo-Orem, UT
St, Louis, MO-IL
Other MSA/CMSA
Other Not MSA
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington DC
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Maximum # of
Exceedances
of 200 ppb
1
0
1
2
12
5
3
3
3
2
1
69
2
2
11
37
0
8
10
2
Maximum
Annual Mean
(ppb)
31.1
33.6
28.1
36.8
25.9
50.6
16.8
42.2
34.00
27.2
26.6
34.8
35.1
15.9
27.1
40.5
28.9
27.2
31.9
19.7
1 CMSA is consolidated metropolitan statistical area; MSA is metropolitan statistical area according to the 1999 Office of Management and
Budget definitions (January 28, 2002 revision).
* Indicates locations that satisfied both the annual average and exceedance criteria.
-------
i 2.3 Ambient Monitor Characterization
2 2.3.1 Introduction
3 Siting of monitors is of particular importance, recognizing that proximity of local sources
4 could influence on measured NC>2 concentrations. As part of the risk and exposure scope and
5 methods document (US EPA, 2007c), both mobile and stationary sources (in particular power
6 generating utilities using fossil fuels) were indicated as significant contributors to nitrogen
7 oxides (NOX) emissions in the U.S. Analyses were performed to determine the distance of all
8 location-specific monitors to these source categories. In addition, emissions of NOX from
9 stationary sources within close proximity of the location-specific monitoring sites were
10 estimated.
11 2.3.2 Approach
12 Major road distances to each monitor were calculated using GIS. Distances of monitoring
13 sites to stationary sources and those source's emissions were estimated using data within the
14 2002 National Emissions Inventory (NEI; US EPA, 2007d). The NEI database reports emissions
15 of NOX in tons per year (tpy) for 131,657 unique emission sources at various points of release.
16 The release locations were all taken from the latitude longitude values within the NEI. First, all
17 NOX emissions were summed for identical latitude and longitude entries while retaining source
18 codes for the emissions (e.g., Standard Industrial Code (SIC), or North American Industrial
19 Classification System (NAICS)). Therefore, any facility containing similar emission processes
20 were summed at the stack location, resulting in 40,855 observations. These data were then
21 screened for sources with emissions greater than 5 tpy, yielding 18,798 unique NOX emission
22 sources. Locations of these stationary source emissions were compared with ambient monitoring
23 locations using the following formula:
24
25 d = arccos(sin(toj)x sin(to2) + cos(latl)x cos(lat2)x cos(lon2 -lon^))xr
26
27 where
28
29 d = distance (kilometers)
30 lati = latitude of a monitor (radians)
31 Iat2 = latitude of source emission (radians)
32 lonj = longitude of monitor (radians)
33 Iori2 = longitude of source emission (radians)
34 r = approximate radius of the earth (or 6,371 km)
35
36 Location data for monitors and sources provided in the AQS and NEI data bases were given
37 in units of degrees therefore, these were first converted to radians by dividing by 180/Ti. For
38 each monitor, source emissions with estimated distances within 10 km were retained.
-------
1 2.3.3 Results
2
3
4
5
6
7
8
9
10
11
The distribution of the nearest distance of the ambient monitors to major roads for each of the
named locations is summarized in Table 4.2 Physical attributes of individual monitors (e.g.,
latitude/longitude, probe height) including the distance of the nearest major road is provided in
the Appendix A. On average, most monitors are placed at a distance of 50 meters or greater
from a major road, however in locations with a large monitoring network such as Boston,
Chicago, or New York CMSA, there may be one or two monitors placed within close proximity
(<10 meters) of a road.
Table 4. Distribution of the distance (m) of ambient monitors to the nearest major road 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
n
4
21
12
4
6
7
3
7
1
10
43
4
26
10
7
1
13
16
Distance (m) of monitor to nearest major road
mean std min 2.5 50 97.5 max
488
101
158
114
196
166
382
282
144
244
155
57
145
247
190
353
126
129
283
93
212
90
103
260
39
266
286
150
45
130
199
177
123
104
134
7
2
2
79
18
339
33
1
1
15
6
45
7
5
14
134
7
2
2
79
18
339
33
1
2
15
6
45
7
5
14
505
70
93
134
180
65
393
128
181
89
55
119
167
141
97
83
809
337
738
187
386
748
415
718
914
522
103
508
630
433
421
338
809
337
738
187
386
748
415
718
914
570
103
508
630
433
421
338
1 n is the number of monitors operating in a particular location between 1995 and 2006. The min, 2.5,
med, 97.5, and max represent the minimum, 2.5th, median, 97.5th, and maximum percentiles of the
distribution for the distance in meters (m) to the nearest major road. Monitors > 1 km from road are not
included.
12
13
14
15
16
17
18
19
20
Table 5 contains a summary of the distance of stationary source emissions to monitors within
each named location. There were a number of sources emitting >5 tpy of NOX and located within
a 10 km radius for many of the monitors. On average though, most monitors are placed at
greater distances from stationary source emissions than roads with most sources at a distance of
greater than 5 km. Most of the stationary source emissions of NOX within a 10 km radius of
monitors were less than 50 tpy (Table 6). Details regarding individual monitors are provided in
Appendix A.
2 Distances between monitors and major roads were first determined using a Tele-Atlas roads database in a GIS
application. For road-monitor pairs that showed particularly close distances, the values were fine-tuned using
GoogleEarth® to estimate the distance to road edge.
8
-------
1
2
Table 5. Distribution of the distance (m) of ambient monitors to stationary sources with NOX emissions >5
tons per year (tpy) and within a 10 kilometers (km)1 radius.
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
n
9
595
394
19
66
140
87
126
20
18
523
11
736
382
59
7
253
160
Distance of monitor to NOX emission source (m)
mean std min 2.5 50 97.5 max
6522
5333
6586
7092
6109
5655
6889
5694
5125
6700
6003
6184
6101
5837
6298
6558
6799
6173
3164
2603
2657
2439
2632
2593
2254
3185
2962
2184
2435
3151
2555
2474
2279
3664
2337
2425
656
142
411
956
782
910
321
119
708
3837
140
1323
103
231
833
1214
396
288
656
761
770
956
1034
1029
1963
1384
708
3837
1483
1323
1383
1299
1312
1214
1989
704
7327
5363
7277
7278
6340
5904
7549
6085
5720
7237
6165
7611
6467
5689
6355
8178
7120
6254
9847
9733
9834
9884
9847
9862
9974
9945
9558
9950
9801
9117
9818
9754
9803
9433
9863
9777
9847
9988
9994
9884
9933
9979
9997
9991
9558
9950
9991
9117
9983
9982
9890
9433
9990
9973
1 n is the number of sources emitting >5 tons per year (tpy) of NOX within a 1 0 kilometer (km) radius of a
monitor in a particular location. The min, 2.5, med, 97.5, and max represent the minimum, 2.5th, median,
97.5th, and maximum percentiles of the distribution for the distance in meters (m) to the source emission.
4
5
Table 6. Distribution of NOX emissions from stationary sources within 10 kilometers (km) of monitoring
site, where emissions were >5 tons per year1.
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
n
9
595
394
19
66
140
87
126
20
18
523
11
736
382
59
7
253
160
Emissions (tpy) of NOX from sources within 10 km of monitor
mean std min 2.5 50 97.5 max
709
128
204
702
387
252
251
117
201
483
70
24
284
154
85
60
167
320
1621
344
919
612
1091
1286
637
286
407
636
310
16
1024
408
234
38
1032
1254
22
5
5
126
5
5
5
5
5
18
5
8
5
5
5
7
5
6
22
5
5
126
5
5
6
5
5
18
5
8
6
5
5
7
5
6
35
10
10
284
19
15
24
31
31
84
12
22
31
29
14
83
16
34
4895
1155
2204
1476
4205
5404
2398
912
1642
1665
577
51
3676
1304
1049
102
848
6009
4895
3794
8985
1476
4205
9483
3762
1679
1642
1665
4256
51
9022
4968
1049
102
14231
10756
1 n is the number of sources emitting >5 tons per year of NOX within a 1 0 kilometer radius of a monitor in
a particular location. The min, 2.5, med, 97.5, and max represent the minimum, 2.5th, median, 97.5th, and
maximum percentiles of the distribution for the source emission in tons per year (tpy).
-------
i 2.4 Spatial and Temporal Air Quality Analyses
2 2.4.1 Introduction
3 An analysis of the air quality was performed to determine spatial and temporal trends,
4 considering locations, monitoring sites within locations, and time-averaging of ambient NO2
5 concentrations collected from 1995 through 2006. The purpose is to present relevant information
6 on the air quality as it relates to both the current form of the standard (annual average
7 concentration) and the exposure concentration and duration associated with adverse health
8 effects (1-hour).
9 2.4.2 Approach
10 To evaluate variability in NO2 concentrations, temporal and spatial distributions of summary
11 statistics were computed in addition to use of statistical tests to compare distributions between
12 years and/or monitors and/or locations. For a given location, the variability within that location
13 is defined by the distribution of the annual summary statistics across years and monitors and by
14 the distribution of the hourly concentrations across hours and monitors. The summary statistics
15 were compiled into tables and used to construct figures for visual comparison and for statistical
16 analysis.
17
18 Boxplots were constructed to display the distribution across sites and years (or hours for the
19 hourly concentrations) for a single location. The box extends from the 25th to the 75th
20 percentile, with the median shown as the line inside the box. The whiskers extend from the box
21 to the 5th and 95th percentiles. The extreme values in the upper and lower tails beyond the 5th
22 and 95th percentiles are not shown to allow for similar scaling along the y-axis for the plotted
23 independent variables. The mean is plotted as a dot; typically it would appear inside the box,
24 however it will fall outside the box if the distribution is highly skewed. All concentrations are
25 shown in parts per billion (ppb).
26
27 Q-Q plots also display the distribution in the calculated air quality metrics across sites and
28 years (or hours for the hourly concentrations) for a single location. The Q-Q plot is used to
29 compare the observed cumulative distribution to a standard statistical distribution. In this case
30 the observed distributions are compared with a log-normal distribution, so that the vertical scale
31 is logarithmic. The horizontal scale is the quantile of a standard normal distribution, so that if
32 there are N observed values, then the kth highest value is plotted against the quantile probit(p),
33 where probit is the inverse of the standard normal distribution function, and/? is the plotting
34 point. The plotting points were chosen asp = (k-3/8)/(N+l/4) for the annual statistics andp =
35 k/(N+l) for the hourly concentrations. If the distribution were exactly log-normal, then the curve
36 would be a straight line. The median value is the y-value when the normal quantile equals zero.
37 The slope of the line is related to the standard deviation of the logarithms, so that the higher the
38 slope, the higher the coefficient of variation (standard deviation divided by the mean for the raw
39 data, before taking logarithms).
40
41 In addition to the tabular and graphical comparisons of the summary statistics, the
42 distributions of each variable were compared using various statistical tests. An F-Statistic
43 comparison compares the mean values between locations using a one way analysis of variance
10
-------
1 (ANOVA). This test assumed that for each location, the site-year or site-hour variables are
2 normally distributed, with a mean that may vary with the location and a constant variance (i.e.,
3 the same for each location). Statistical significance was assigned for p-values less than or equal
4 to 0.05. The Kruskal-Wallis Statistics are non-parametric tests that are extensions of the more
5 familiar Wilcoxon tests to two or more groups. The analysis is valid if the difference between
6 the variable and the location median has the same distribution for each location. If so, this
7 procedure tests whether the location medians are equal. The test is also consistent under weaker
8 assumptions against more general alternatives. The Mood Statistic comparisons are non-
9 parametric tests that compare the scale statistics for two or more groups. The scale statistic
10 measures variation about the central value, which is a non-parametric generalization of the
11 standard deviation. This test assumes that all the groups have the same median. Specifically,
12 suppose there is a total of N values, summing across all the locations to be compared. These N
13 values are ranked from 1 to N, and the jth highest value is given a score of (j - (N+l)/2}2. The
14 Mood statistic uses a one-way ANOVA statistic to compare the mean scores for each location.
15 Thus the Mood statistic compares the variability between the different locations assuming that
16 the medians are equal.
17 2.4.3 Spatial Results
18 A summary of the important spatial trends in NO2 concentrations is reported in this section.
19 Detailed air quality results (i.e., by year and within-location) are presented in Appendices B and
20 C, each containing both tabular and graphic summaries of the spatial and temporal concentration
21 distributions.
22
23 A broad view of the spatial differences in NO2 monitoring concentrations across locations is
24 presented in Figures 1 and 2. In general there is variability in NO2 concentrations between the
25 20 locations. For example, in Los Angeles, the mean of annual means is approximately 24.3 ppb
26 over the period of analysis, while considering the Not MSA grouping, the mean annual mean
27 was about 7.0 ppb. Phoenix contained the highest mean annual mean of 27.3 ppb. Variability in
28 the annual average concentrations was also present within locations, the magnitude of which
29 varied by location. On average, the coefficient of variation in the annual mean concentrations
30 was about 35%, however locations such as Jacksonville or Provo had COVs as low as 6% while
31 locations such as Las Vegas and Not MSA contained COVs above 60%. Reasons for differing
32 variability arise from the size of the monitoring network in a location, level of the annual mean
33 concentration, underlying influence of temporal variability within particular locations, among
34 others.
35
11
-------
5
6
7
8
9
10
Annual Mean
•15-
40-
35-
25-
20-
tioston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington
^ Location
2 Figure *\. Spatial distributions of annual mean NO2 ambient monitoring concentrations for selected
3 CMSA locations, years 1995-2006.
4
Annual Mean
40-
Atlanta Colorado I'.! Pisjio Jiieksonvillc La.s Vega^ Phoenix Provo Si. Louts Other MSA Othr Non-MSA
Location
Figure 2. Spatial distributions of annual mean NO2 ambient monitoring concentrations for selected MSA
and grouped locations, years 1995-2006.
Spatial differences in hourly concentrations were of course consistent with that observed for
the annual mean concentrations, and as expected there were differences in the COVs across
12
-------
1
2
3
4
5
6
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
locations, ranging from about 60 to 120%. However, in comparing the 90 percent intervals
(from the 5th to the 95th percentiles) of hourly concentrations across locations, the ranges are
somewhat similar (for example see Figure 3 for the CMSA locations). This means that the
intervals for the annual mean differ more than that of the hourly concentrations between
locations likely due to the influence of high 1-hour NC>2 concentrations for certain locations.
Hourly Cones
60-
55-
50-
45-
40-
35-
30-
Boston Chicago Cleveland Denver Detroit I,.os An^clcs \1iami New York Philadelphia Washington
Location
Figure 3. Spatial distributions3 of hourly NO2 ambient monitoring concentrations for selected CMSA
locations, years 1995-2006.
This presence of extreme NC>2 concentrations is best illustrated in Figure 4 using a Q-Q plot
that captures the full concentration distribution for each CMSA location. The Q-Q plots are
generally curved rather than straight, such that the distributions do not appear to be log-normal.
However, the annual mean and hourly concentration curves do tend to be approximately straight
and parallel for values above the median (normal quantile = 0) through the 3rd quantile,
suggesting that these upper tails of the distributions are approximately log-normal with
approximately the same coefficients of variation. Beyond the 3rd quantile though, each
distribution similarly and distinctly curves upwards, indicating a number of uncharacteristic NC>2
concentrations at each location when compared with the rest of their respective concentration
distributions.
3 The boxplots for hourly concentrations were created using a different procedure than for the annual
statistics, because of the large number of hourly values and the inability of the graphing procedure to
allow frequency weights. Therefore, the appropriate weighted percentiles and means were calculated
and plotted as shown, but the vertical lines composing the sides of the box were omitted.
13
-------
I [nurly Ortnc
1000.0 H
1
2
3
4
5
6
7
8
9
10
11
100.C
lor
o.i-
location —Boston
-3 -2
—Chicago —Cleveland
"Denver
\omnal Quantilc
Detroit Los Angeles Miami
345
—New York —Philadelphia —Washington
Zero values^ ere r
Figure 4. Spatial distributions of hourly NO2 ambient concentration for selected CMSA locations, years
1995-2006.
Distributions of each variable (annual means and hourly concentrations) were compared
between the different locations using statistical tests. The results in Table 7 show statistically
significant differences between locations for both variables and all three summary statistics
(means, medians, and scales). This supports the previous observation that the distributions for the
different locations are dissimilar.
Table 7. Statistical test results for spatial comparisons of all location parameter distributions.
Concentration
Parameter
Annual Mean
Hourly
Means Comparison
F Statistic p-value
148
330272
<0.0001
<0.0001
Central Values Comparison
Kruskal-Wallis p-value
1519
5414056
<0.0001
<0.0001
Scales Comparison
Mood p-value
729
1354075
<0.0001
<0.0001
12
13
14
15
16
17
18
19
20
21
22
23
The spatial distributions of NC>2 concentrations within locations were also evaluated. As an
example, Figure 5 illustrates the distribution of the annual mean NC>2 concentration at 10
monitoring sites within Philadelphia. The mean annual means vary from a minimum of 14.8 ppb
(site 1000310071) to a maximum of 30.5 ppb (site 4210100471). The range of within-site
variability can be attributed to the number of monitoring years available coupled with the
observed trends in temporal variability across the monitoring period (discussed below in Section
2.4.4).
Distributions of each variable (annual means and hourly NC>2 concentrations) within
locations (i.e., site distributions) were compared using statistical tests. The results in Table 8
indicate statistically significant differences within locations for both variables and the central
14
-------
1 tendency statistics (means and medians), while scales were statistically significant for 38 out of
2 40 possible tests. This supports the previous observation that the distributions for the different
3 locations are dissimilar.
5
6
7
Annual Mean
.14-
28-
27-
26-
25-
24-
23-
1001)310031 1000310071 1000,320041 3400700032 4201700121 4204500021 4209IOOI.il 4210)00043 4210100292 421010047]
Figure 5. Spatial distribution of annual average NO2 concentrations among 10 monitoring sites in
Philadelphia CMSA, years 1995-2006.
9
10
Table 8. Statistical test results for spatial comparisons of within location parameter distributions.
Concentration
Parameter
Annual Mean
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington DC
Atlanta
Colorado Springs
El Paso
Las Vegas
Means Comparison
F
Statistic p-value
47.3
123
12.1
85.3
13.2
49.0
111
106
48.9
48.6
119
8.7
36.0
137
0.001
O.001
0.001
0.001
O.001
0.001
O.001
0.001
0.001
O.001
0.001
O.001
0.001
0.001
Central Values
Comparison
Kruskal-
Wallis p-value
96.5
76.7
15.4
32.0
13.1
325
36.2
163
68.8
104
45.2
18.8
31.6
45.4
0.001
O.001
0.002
0.001
0.001
0.001
O.001
0.001
0.001
O.001
0.001
0.009
0.001
0.001
Scales Comparison
Mood p-value
79.9
68.5
7.5
23.0
7.8
240
29.9
151
33.0
71.2
28.6
8.7
35.3
35.2
0.001
O.001
0.058
0.001
0.020
0.001
O.001
0.001
0.001
O.001
0.001
0.273
0.001
0.001
15
-------
Concentration
Parameter
Hourly
Location
Phoenix
St. Louis
Other CMSA
Not MSA
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Las Vegas
Phoenix
St. Louis
Other CMSA
Not MSA
Means Comparison
F
Statistic p-value
20.4
51.5
82.5
76.9
17884
11611
4191
25130
4125
27288
10669
20052
13759
14262
35917
5541
10503
22567
5626
14807
19557
17630
0.001
0.001
O.001
0.001
O.001
0.001
0.001
O.001
0.001
O.001
O.001
0.001
O.001
0.001
O.001
O.001
0.001
O.001
0.001
O.001
O.001
0.001
Central Values
Comparison
Kruskal-
Wallis p-value
32.2
82.1
2152
424
312994
142034
14102
104800
10442
1050310
68580
404234
112129
223040
137022
48252
57694
136455
35645
178180
6306431
1580139
0.001
0.001
O.001
0.001
O.001
0.001
0.001
O.001
0.001
O.001
O.001
0.001
O.001
0.001
O.001
O.001
0.001
O.001
0.001
O.001
O.001
0.001
Scales Comparison
Mood p-value
23.6
69.0
1934
372
59896
37224
1985
2864
424
269190
43090
91104
4903
30974
17330
3921
18334
28972
6747
47842
2164452
491390
0.001
0.001
O.001
0.001
O.001
0.001
0.001
O.001
0.001
O.001
O.001
0.001
O.001
0.001
O.001
O.001
0.001
O.001
0.001
O.001
O.001
0.001
4
5
6
7
8
9
10
11
12
13
14
1
2 2.4.4 Temporal Results
A broad view of the trend of NC>2 monitoring concentrations over time is presented in Figure
6. The annual mean concentrations were calculated for each monitor site within each year to
create a distribution of annual mean concentrations for each year. The distribution of annual
mean concentrations generally decreases with each increasing year. On average, mean annual
mean NC>2 concentrations consistently decrease from a high of 17.5 ppb in 1995 to the most
recent mean of 12.3 ppb. Also notable is the consistent pattern in the decreasing concentrations
across each years distribution, the shape of each curve is similar indicating that while
concentrations have declined, the variability within each year is similar from year to year. The
variability within a given year is representing spatial differences in annual average
concentrations across the 20 locations.
16
-------
Annual Mean
100.0-
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
10.0-
-3
-2
-I
0123
Normal Quantilc
year - I
-------
Q-Q Plot of Annual Mean
2
3
4
5
6
7
8
9
10
11
12
13
14
15
loc_type=CMSA loc_name=Philadelphia-Wilmington-Atlantic City, PA-NJ-DE-MD CMSA
Annual Mean
100-
MB ~ 1995- 1996- 1997- 1998
Normal Qiiantili.1
1999 2000 - 2001 - 2002 - 2003 - 2004
2005 2006
Figure 7. Temporal distributions of annual mean NO2 concentrations for the Philadelphia CMSA, years
1995-2006.
In general, temporal trends within a location considering the hourly concentration data were
consistent with the above, particularly where the monitoring network was comprised of several
monitors. For example, Figure 8 illustrates the temporal distribution for hourly NC>2
concentration in the Los Angeles CMSA, comprising between 26 and 36 monitors in operation
per year. NC>2 concentrations are decreased with increasing calendar year of monitoring with the
distribution of hourly concentrations lowest in the more recent years of monitoring. The pattern
of variability in NC>2 concentration within a year at this location is also similar when comparing
across years based on similarities in the shape of each years respective curve.
18
-------
Q-Q Plot of Hourly Concentrations
loc_type=CMSA loc_ricime=Los Angeles-Riverside-Orange County, CA CMSA
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Hourly Cone
1000.0-
-3-2-1012315
Ncrmal Quanlilc
year - 1W5 - ] 9% - 1997 - 1998 1999 2000 - 2001 - 2002 - 2003 - 2004 - 2005 - 2006
Zcrovflhics ucrc rcpliKcd In II S
Figure 8. Temporal distribution of hourly NO2 concentrations in the Los Angeles CMSA, years 1995-
2006.
These temporal trends were confirmed by statistical comparison tests. The means and
medians of the annual means and hourly concentrations compared across the different years were
statistically significant (Table 9). The Mood test shows that for the annual means, the scales
were also significantly different. Note, however, that the Mood test derivation assumes that the
medians of the annual means are the same for each year, whereas the plots and the Kruskall-
Wallis test result implies that the medians are not the same. As noted before, Figure 8 indicates
that the Q-Q curves for different years have similar slopes but different intercepts, which implies
that the annual means for different years have different mean values but similar coefficients of
variation. In fact the coefficients of variation of the annual means are nearly identical for
different years, ranging from 52 % to 55 %.
Table 9. Statistical test results for temporal comparisons of all location parameter distributions.
Concentration
Parameter
Annual Mean
Hourly
Means Comparison
F Statistic p-value
15.0
47432
<0.0001
<0.0001
Central Values Comparison
Kruskal-Wallis p-value
146
494826
<0.0001
<0.0001
Scales Comparison
Mood p-value
32.5
24238
0.0006
<0.0001
17
18
19
20
21
22
23
There were some exceptions to this temporal trend, particularly when considering the
distribution of hourly concentrations and where a given location had only few monitors per year.
Using Jacksonville as an example, Figure 9 illustrates the same temporal trend in NC>2
concentrations as was observed above for much of the distribution, however distinctions are
noted at the upper tails of the distribution for two years of data, 2002 and 2004. For
Jacksonville, each years' hourly concentration distribution was based on only a single monitor.
19
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Where few monitors exist in a given location, atypical variability in one or a few monitors from
year to year can greatly influence the distribution of short-term concentrations, particularly at the
upper percentiles.
The same follows for assignment of statistical significance to temporal trends within
locations. While annual average concentrations are observed to have declined over time within a
location, the number of sites were typically few thus limiting the power of the statistical tests.
Only Los Angeles, El Paso, Phoenix, and Other CMSA were significant (p<0.05) for the central
tendency tests, while only Los Angeles and Other CMSA were significant (p<0.05) for scale
(data not shown). All hourly concentrations comparison tests for years within each location were
significant for all three test statistics (mean, median, scale).
Hourly Cone
1000.0-
1000
10.0
-4-3-2-101234
Nonnal Quantilc
year - 1995- 1996- 1997- 1998 1999 2000" 2002- 2003-2004- 2005
Zero values nere replaced bj 0 i>
Figure 9. Temporal distributions of hourly NO2 concentrations in the Jacksonville MSA, years 1995-
2006, one monitor.
One final temporal trend worthy of mention is that associated with the Not MSA grouped
location. There is very little difference in annual average concentrations across the 1995-2006
monitoring period. While percentage-wise the reduction in concentration is about 25%, on a
concentration basis, this amounts to a maximum of about 2 ppb reduced over the 11 year period
(Figure 10). In considering, the past 5 years, there was even less of a reduction in annual
average concentration with about a 0.5 ppb difference between 2001 and 2006. This could
indicate that many of these monitoring sites are receiving relatively less impact from local
sources of NO2 (e.g., emissions from major roads and stationary sources) compared with the
other locations. Therefore, the areas that these monitors represent may also be less likely to see
significant benefit by changes in source emissions and/or NO2 standard levels compared with the
named CMSA/MSA locations.
20
-------
Annual Mean
20-
19-
18-
17-
16-
15-
14-
13-
12-
II-
10-
9-
8-
7-
6-
5-
4-
3-
Year
2 Figure 10. Temporal distribution of annual average NO2 concentrations in the Not MSA group location,
3 years 1995-2006.
21
-------
i 2.5 Air Quality Simulation
2 2.5.1 Introduction
3 Every location across the U.S. meets the current NC>2 annual standard (US EPA, 2007e).
4 Even considering air quality data as far back as 1995, no location/monitoring site exceeded the
5 current standard. Therefore, simulation of air quality data was required to evaluate just meeting
6 the current standard or standards that are more stringent.
7
8 In developing a simulation approach to adjust air quality to meet a particular standard level,
9 policy-relevant background (PRB) levels in the U.S. were first considered. Policy-relevant
10 background is defined as the distribution of NO2 concentrations that would be observed in the
11 U.S. in the absence of anthropogenic (man-made) emissions of NC>2 precursors in the U.S.,
12 Canada, and Mexico. Estimates of PRB have been reported in the draft ISA (Section 1.5.5) and
13 the Annex (AX2.9), and for most of the continental U.S. the PRB is estimated to be less than 300
14 parts per trillion (ppt). In the Northeastern U.S. where present-day NC>2 concentrations are
15 highest, this amounts to a contribution of about 1% percent of the total observed ambient NC>2
16 concentration (AX2.9). This low contribution of PRB to NC>2 concentrations provides support
17 for a proportional method to adjust air quality, i.e., an equal adjustment of air quality values
18 across the entire air quality distribution to just meet a target value.
19
20 Next, the variability in NC>2 concentrations was evaluated to determine whether a
21 proportional approach would be reasonable if applied broadly across all years of data. Since the
22 adjustment factor to meet the current standard would likely increase with increasing year, it was
23 of interest to determine the trend in both the hourly concentrations and variability by year.
24 Figure 11 presents a summary of the annual average and hourly mean concentrations, as well as
25 the coefficient of variation (COV, standard deviation as a percent of the mean) for each
26 respective mean. Sample size for the annual average concentrations was about 350 per year,
27 while hourly concentrations numbered about 3 million per year.
28
29 As expected, there was no observed difference in the mean concentrations when comparing
30 each concentration metric within a year. The mean of the annual averages of all monitors is
31 nearly identical to the mean of the hourly concentrations. However, statistically significant
32 decreases in concentration are evident from year-to-year (p<0.0001), with concentrations
33 decreasing by about 30% across the monitoring period. Contrary to this, there is no apparent
34 trend in the COV for the annual average concentrations across the 12 years of data, generally
35 centered about 53%. The COV of the hourly concentrations is larger than the annual COV as
36 expected, however it increases with increasing year. The hourly COV ranges from a low of 84%
37 in 1998 to a high of 92% in 2006, amounting to a relative percent difference of only 10% across
38 the entire monitoring period. A non-parametric Mann-Whitney U-test indicates that there is a
39 significant difference in the COVs when comparing each year-group (p=0.004). This may result
40 in a small upward bias in the number of estimated exceedances of short-term (1-hour) potential
41 health benchmark levels if using a proportional roll-up on the more recent monitoring data
42 relative to that estimated by rolling up the historic data to just meet the current standard. While
43 the trend of increasing COV is apparent across the entire monitoring period, based on the limited
44 difference in COV from year-to-year for both the annual and hourly concentration data within
22
-------
1 each year-group (each is <4%), it is concluded that a proportional method could be broadly
2 applied.
20
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
19
18
I17
O
15 16
+J
O 15
O
O
o 14
13
12
11 -
10
.o
-o'
95
90
85
80
-•—Annual Mean
-•—Hourly Mean
-D--COV-Annual
-o--COV-Hourly
B- 0 ,
-Q-..
•-O-
100
50
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
O
03
75 >
O
+ 70 .S
1
65 O
60
55
Figure 11. Trends in hourly and annual average NO2 ambient monitoring concentrations and their
associated coefficients of variation (COV) for all monitors, years 1995-2006.
2.5.2 Approach
For the air quality characterization, data were first separated into two groups, an historic set
of monitoring data (1995-2000) and one containing the most recent air quality (2001-2006).
This grouping would further reduce any potential influential monitoring data affecting the
variability in hourly concentrations that may exist in one year to the next within a location.
Typically, ambient concentrations are not adjusted higher to simulate just meeting alternative
standards, therefore older historical data may be of use in better representing scenarios that are at
or near the current NC>2 standard. To date, the following air quality scenarios have been
considered:
• "as is" representing the historical and recent ambient monitoring hourly concentration
data as reported by US EPA's Air Quality System (AQS).
• "simulated' concentrations to just meet the current NC>2 NAAQS (53 ppb annual
average).
23
-------
1 Based on the form of the standard and observed trends in ambient monitoring, such as the
2 retention of similar hourly and annual COVs over time while annual average concentrations
3 significantly decrease over the same time period, NC>2 concentrations were proportionally
4 modified at each location using the maximum annual average concentration that occurred in each
5 year. To just meet the current standard adjustment factors F for each location (/') and year (/')
6 were derived by the following
7
9
10 where,
11
12 FJJ = Adjustment factor (unitless)
13 Cmax,ij = Maximum annual average NC>2 concentration at a monitor in a location /' (ppb)
14
15 Values for each air quality adjustment factor used for each location are given in Tables 10
16 and 1 1 . It should be noted that a different monitor could have been used for each year to
17 estimate F, the selection dependent only on whether the monitor contained the highest annual
18 concentration for that year in the particular location. For each location and calendar year, all the
19 hourly concentrations were multiplied by the same constant value F to make the highest annual
20 mean equal to 53 ppb for that location and year. For example, for Boston in 1995, the maximum
21 annual mean was 30.5 ppb, giving an adjustment factor ofF= 53/30.5 = 1.74. All hourly
22 concentrations in Boston in 1995 were multiplied by 1.74. Then, using the adjusted hourly
23 concentrations, the distributions of the annual means and annual number of exceedances are
24 computed in the same manner as the as-is scenario.4
4 Because of the large database, we did not implement this procedure exactly as stated. For the annual means we
computed and applied the adjustment factors directly to each annual mean. For the hourly concentrations we used
the frequency distributions of the rounded hourly values, so that, in effect, we applied the adjustment factors to the
hourly values after rounding them to the nearest integer. This has a negligible impact on the calculated number of
exceedances.
24
-------
Table 10. Maximum annual average NO2 concentrations and air quality adjustment factors (F) to just
meet the current standard, historic monitoring data.
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington DC
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Metric
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
1995
30.5
1.74
32.2
1.64
27.3
1.94
34.8
1.52
21.6
2.45
46.2
1.15
14.7
3.60
41.7
1.27
31.8
1.67
26.2
2.02
18.8
2.81
23.2
2.28
23.3
2.27
15.8
3.36
27.1
1.96
32.6
1.63
22.6
2.35
26.2
2.02
31.9
1.66
19.1
2.78
1996
31.0
1.71
32.0
1.66
25.9
2.04
33.1
1.60
21.5
2.47
42.3
1.25
16.0
3.30
42.2
1.26
33.9
1.56
26.9
1.97
26.6
1.99
23.6
2.24
35.1
1.51
14.9
3.55
26.7
1.99
31.6
1.68
24.3
2.18
24.8
2.14
30.3
1.75
14.5
3.66
1997
30.4
1.74
33.6
1.58
28.1
1.89
33.9
1.56
25.9
2.05
43.2
1.23
16.6
3.19
41.1
1.29
32.4
1.63
25.9
2.05
25.2
2.10
19.8
2.68
33.6
1.58
14.4
3.69
32.0
1.66
23.3
2.27
24.8
2.14
29.4
1.80
19.7
2.69
1998
30.7
1.73
32.2
1.64
27.3
1.94
35.3
1.50
22.9
2.31
43.4
1.22
15.2
3.49
41.9
1.26
34.0
1.56
27.2
1.95
24.1
2.20
20.5
2.59
30.7
1.72
15.0
3.52
25.3
2.09
35.0
1.52
23.9
2.22
25.8
2.05
31.0
1.71
18.8
2.82
1999
29.7
1.79
31.5
1.68
24.5
2.16
19.4
2.73
18.0
2.94
50.6
1.05
16.8
3.15
41.5
1.28
31.7
1.67
25.4
2.09
23.8
2.22
19.3
2.75
27.7
1.91
15.9
3.34
26.6
1.99
40.5
1.31
24.1
2.20
27.2
1.95
29.3
1.81
19.7
2.69
2000
29.0
1.83
32.0
1.66
23.1
2.30
14.9
3.55
23.9
2.22
43.9
1.21
15.7
3.37
40.6
1.31
27.9
1.90
23.5
2.26
22.9
2.31
34.8
1.52
24.3
2.18
15.4
3.45
25.1
2.12
36.3
1.46
23.6
2.25
26.3
2.02
26.5
2.00
18.7
2.83
25
-------
Table 11. Maximum annual average NO2 concentrations and air quality adjustment factors (F) to just
meet the current standard, recent monitoring data.
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington DC
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Metric
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
Max Annual Mean
F
2001
29.7
1.79
31.9
1.66
23.6
2.25
36.8
1.44
23.2
2.29
41.2
1.29
15.8
3.35
40.3
1.32
29.9
1.77
24.3
2.18
23.3
2.27
21.7
2.45
22.5
2.35
37.1
1.43
24.1
2.20
24.7
2.15
26.5
2.00
16.5
3.21
2002
25.3
2.10
32.4
1.63
22.3
2.38
35.4
1.50
21.4
2.47
40.2
1.32
14.3
3.71
39.7
1.33
29.5
1.80
24.8
2.14
19.4
2.73
21.4
2.48
14.6
3.62
22.3
2.38
34.7
1.53
24.8
2.14
22.9
2.32
27.4
1.93
16.4
3.23
2003
22.5
2.36
30.9
1.72
21.7
2.45
21.4
2.47
22.0
2.41
35.3
1.50
12.9
4.12
32.0
1.65
24.7
2.15
26.0
2.04
16.4
3.23
19.9
2.66
14.3
3.70
21.4
2.48
34.3
1.54
21.8
2.43
20.3
2.60
26.4
2.01
15.5
3.42
2004
25.0
2.12
29.3
1.81
22.2
2.38
27.2
1.95
18.9
2.80
33.7
1.57
13.0
4.08
30.5
1.74
25.6
2.07
24.0
2.20
17.0
3.12
18.0
2.94
13.7
3.88
19.7
2.69
31.4
1.69
22.3
2.37
22.3
2.37
25.3
2.09
15.8
3.36
2005
23.4
2.26
29.6
1.79
21.5
2.46
27.6
1.92
19.6
2.71
30.9
1.72
13.5
3.92
36.5
1.45
26.3
2.02
24.1
2.20
17.4
3.05
17.3
3.06
13.3
3.97
19.9
2.67
31.5
1.68
20.5
2.58
16.8
3.15
24.0
2.21
17.1
3.11
2006
22.5
2.35
30.6
1.73
18.2
2.91
29.1
1.82
15.9
3.34
29.7
1.78
34.2
1.55
17.8
2.98
19.6
2.70
17.9
2.96
18.0
2.94
30.6
1.73
28.9
1.83
15.0
3.52
18.5
2.87
15.6
3.39
26
-------
i 2.6 Method for Estimating On-Road Concentrations
2 2.6.1 Introduction
3 As an additional step in the air quality characterization, the potential impact of motor
4 vehicles on the surrogate exposure metrics was evaluated. Several studies have shown that
5 concentrations of NO2 are at elevated levels when compared to ambient concentrations measured
6 at a distance from the roadway (e.g., Rodes and Holland, 1981; Gilbert et al., 2003; Cape et al.,
7 2004; Pleijel et al., 2004; Singer et al., 2004). On average, concentrations on or near a roadway
8 are from 1.5 to 2 times greater than ambient concentrations (US EPA, 2007f), but on occasion, as
9 high as 7 times greater (Bell and Ashenden, 1997; Bignal et al., 2007). A strong relationship
10 between measured on-road NO2 concentrations and those with increasing distance from the road
11 has been reported under a variety of conditions (e.g., variable traffic counts, different seasons,
12 wind direction) and can be described (e.g., Cape et al., 2004) with an exponential decay equation
13 of the form
14
15 Cl=Cfe + Cve-fa eq(l)
16 where,
17
18 Cx = NO2 concentration at a given distance (x) from a roadway (ppb)
19 Cb = NO2 concentration (ppb) at a distance from a roadway, not directly influenced
20 by road or non-road source emissions
21 Cv = NO2 concentration contribution from vehicles on a roadway (ppb)
22 k = Rate constant describing NO2 combined formation/decay with perpendicular
23 distance from roadway (meters'1)
24 x = Distance from roadway (meters)
25
26 As a function of reported concentration measures and the derived relationship, much of the
27 decline in NO2 concentrations with distance from the road has been shown to occur within the
28 first few meters (approximately 90% within 10 meter distance), returning to near ambient levels
29 between 200 to 500 meters (Rodes and Holland, 1981; Bell and Ashenden, 1997; Gilbert et al.,
30 2003; Pleijel et al., 2004). At a distance of 0 meters, referred to here as on-road, the equation
31 reduces to the sum of the non-source influenced NO2 concentration and the concentration
32 contribution expected from vehicle emissions on the roadway using
33
34 Cr=Ca(l+m) eq(2)
35 where,
36
37 Cr = 1-hour on-road NO2 concentration (ppb)
38 Ca = 1-hour ambient monitoring NO2 concentration (ppb) either as is or modified to
39 just meet the current standard
40 m = Modification factor derived from estimates of Cv/Cb (from eq (1))
41
27
-------
1
2
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
and assuming that Ca =
2.6.2 Derivation of On-Road Factors
A literature review was conducted to identify published studies containing NC>2
concentrations both on-roads and with various distances from roadways. Principal criteria for
inclusion in this analysis were that either tabular, graphical, or equations were provided in the
paper that related distances from roadways and associated NO2 concentrations. Eleven papers
were identified using these criteria, spanning several countries, various time periods, roadway
locations, seasons, and wind direction (Table 12). The final data set contained 501 data points,
encompassing multiple NO2 measurements from a total of 56 individual roads.
Table 12. Reviewed studies containing NO2
First Author
Bell
Signal
Cape
Gilbert
Maruo
Monn
Nitta
Pleijel
Rodes
Roorda-Knape
Singer
Year
1987
2004
2002
2001
2001
1995
1982
1994
1978
1995
2001
Country/State
Wales
England
Scotland
Quebec
Japan
Switzerland
Japan
Sweden
California
Holland
California
concentrations at a distance
Season Type
Summer, winter Rural
Summer, fall Urban
Annual Urban
Summer Urban
Summer Urban
Summer, Winter Urban
Not reported Urban
Summer Rural
Summer Urban
Summer Urban
Spring through fall Urban
from roadways.
Wind Direction
Up, down
Combined
Combined
Down, up, combined
Combined
Combined
Combined
Combined
Down
Combined
Up, Down
Although there were, on occasion, several roads within a particular study, data for factors
thought to influence on-road concentrations were very limited or were not distinct for all studies.
The relationship noted in eq (1) was solved using the data collected from the above reviewed
literature and employing the SAS procedure proc nlin, generally as follows,
proc nlin data=no2 maxiter=1000 noprint NOITPRINT;
parms Cb=0 to 80 by 1
Cv= 0 to 80 by 1
k= 0 to 1 by .025;
model Cr=Cb + Cv*exp (-k*x);
by author road season wind;
output out=outdata parms=Cb Cv k;
run;
The procedure was run for all individual roads identified within each study location. Results
of this analysis were screened for data that yielded no unique solutions (lack of model
convergence) or irrational parameters. Criteria for censoring data included the following, as well
as the number of individual roads censored:
• Model did not converge (n=5)
5 Note that Ca differs from Cb since Ca may include the influence of on-road as well as non-road sources. However,
it is expected that for most monitors the influence of on-road emissions is minimal so that Ca = Cfr
28
-------
1 • k<0 (n=l)
2 • k>\ (n=2)
3 • Both k=0 and Cv =0 (n=l)
4 • Extremely large Cv (>8,000 ppb; n=2)
5 • Cb<0 (n=l)
6
7 Data were evaluated for trends using available influential factors and considering the number
8 of samples available for potential groupings. In general, the measurements reported in the
9 summer and resultant parameter estimates were observed as distinct from the measures and
10 parameter estimates from other seasons. The data were then grouped accordingly into two
11 seasonal groups, summer and not summer, containing 23 and 21 samples, respectively. These
12 two groups were also censored for any unusual parameter estimates. Resulting criteria for
13 censoring the grouped data included the following:
14
15 • Extreme value of & compared with others in group (n=l)
16 • Extreme values of estimated m due to combined low estimated Cb relative to high
17 estimated Cv (n=2)
18
19 Two approaches were considered for estimating m from the Cv and Cb pairs in each season.
20 The first approach was to regress Cb on Cv (either with or without an intercept) and use the fitted
21 slope to estimate m. Ignoring meteorological effects, Equation 1 implies that Cv results solely
22 from on-road emission sources and that Cb results solely from non-road emission sources. Since
23 these two source types are likely to have quite different diurnal profiles, we expect the hourly Cv
24 and Cb values to be approximately independent.6 Regressing Cb against Cv would imply that
25 there is some correlation between the values, which would be inconsistent with the conceptual
26 model underlying Equation 1. Further, if Cb were regressed against Cv using an intercept, the
27 physical meaning of the intercept would be unclear.
28
29 An empirical method was selected for the approach to estimate m based on the two seasonal
30 sets of ratios of Cv/Cb. The resulting distribution for each group is presented in Figure 12.
31 Neither group could be assigned to a particular distribution (e.g., normal, lognormal, exponential,
32 gamma). Means from the two seasons were tested for significant difference using a Student's t
33 (p=0.026), while the season distributions were compared using a Kolmogorov-Smirnov test (p=
34 0.196). It was decided to retain the groups as separate to allow for some apportioning of
35 variability resulting from an apparent seasonal influence, even though the statistical test results
36 were mixed.
37
38
6 Although the fact that Cv and Cb are subject to the same meteorology introduces some correlation, because
meteorology tends to vary on a longer time scale than hourly, it is likely to have less influence than the emissions on
the correlation between hourly concentrations.
29
-------
1
2
100%
90%
80%
70%
_o
'^
g 60%
2 site-seasons.
30
-------
1
2 A particular summer on-road factor has a 1/22 chance of selection, while a specific not
3 summer value has a 1/19 probability of selection, based on respective sample sizes. This random
4 assignment was repeated for all site-years of data. Hourly NO2 concentrations were estimated
5 for each site-year of data in a location using eq (2) and the randomly assigned on-road factors.
6 Finally, the process was simulated 100 times for each site-year of hourly data. For example, the
7 Boston CMS A location had 210 random selections from the on-road distributions applied
8 independently to the total site-years of data (105). Following 100 simulations, a total of 10,500
9 site-years of data were generated using this procedure (along with 21,000 randomly assigned on-
10 road values selected from the appropriate empirical distribution).
11
12 Simulated on-road NO2 concentrations were used to generate concentration distributions for
13 the annual average concentrations and distributions for the number of exceedances of short-term
14 potential health effect benchmark levels. Means and median values are reported to represent the
15 central tendency of each parameter estimate. Since there were multiple simulations performed at
16 each location using all available site-years of data, results for the upper percentiles were
17 expanded to the 95th, 98th and 99th percentiles of the distribution, rather than estimate a 95%
18 interval as was done above for the non-road scenarios. It is more appropriate to apply the
19 parameter estimates outside the central tendencies to particular sites, areas within locations, or
20 for certain conditions. Minimum values for the annual mean and annual number of exceedances
21 were also estimated. One approach would have been to use the minimum values across the 100
22 simulations. However, that approach may not give the lowest possible value, because it is
23 unlikely that in 100 simulations for a site-year there is a simulation where both seasonal
24 adjustment factors are chosen to be the lowest values of 1 + m. To obtain the lowest value, two
25 simulations were conducted for each site-year. The Summer seasonal adjustment factor was set
26 to the lowest possible value (1.49) and the Not-Summer seasonal adjustment factor was the
27 lowest possible value (1.22). The annual means and exceedances for those two separate
28 simulations were used to compute the minimum values for each distribution.
29
30 In addition and as part of the air quality characterization, the approach described in Section
31 2.7 below was used to estimate the number of short-term concentrations above selected levels
32 that might occur on roadways using the estimated hourly Cr values, associated with air quality as
33 is. For evaluating just meeting the current annual standard the approach described in Section 2.5
34 was used before estimating on-road concentrations.
35 2.6.4 Interpretation of Estimated On-Road Concentrations
36 The simulated on-road concentrations, simply put, are estimates. The algorithm is not
37 designed to estimate concentrations on a particular roadway, all roads, or to estimate on-road
38 exposures in a location. The algorithm assumes that the monitor is measuring the concentrations
39 that would be observed at a distance (>200m or so) of a particular road (could be any road type).
40 It then follows that the monitors within a location are linked proportionally to the distribution of
41 roads (and types) in a location. This is likely not the case, particularly in locations with few
42 monitoring sites, therefore available monitors will likely be either over- or under-representative
43 of some roadway types.
44
31
-------
1 The simulation is designed to estimate the potential concentrations associated with potential
2 on-road exposures, developing central tendencies and bounds to be interpreted qualitatively with
3 the expected emissions that would occur on-roads within a location. That is, the higher-traveled
4 roadways would be better represented by on-road concentration estimates at the upper tails of the
5 distribution, while other roads with less traffic density would be better represented at the lower
6 tails of the distribution. Additional consideration should be given to where few monitor sites
7 were available in a location, or even where monitor sites are more densely distributed within a
8 particular area of a location, before interpreting estimated concentrations.
32
-------
i 2.7 Estimation of Potential Health Effect Benchmark Exceedances
2 2.7.1 Introduction
3 A principal goal of the exposure assessment was to develop a model that estimates the
4 frequency of high short-term exposures, considering just meeting the current standard and any
5 alternative standards under consideration. Since the current standard is on an annual average
6 basis, the relationship between that NC>2 concentration and short-term NC>2 concentrations
7 needed to be evaluated. As part of the prior review, McCurdy (1994) used a non-linear
8 regression (i.e., exponential) to describe the relationship between annual average concentrations
9 and occurrence of short-term peak concentrations at two locations (i.e., one for Los Angeles and
10 one for all other locations combined). At the time of the McCurdy (1994) analysis, there were at
11 least a few monitors with reported annual average concentrations at or above the current standard
12 for the Los Angeles analysis. The non-linear model was applied to estimate number of
13 exceedances given selected annual average concentrations, and reasonably estimated the average
14 number of exceedances at selected annual average concentration levels.
15
16 The same type of regression model was explored as a first step in this analysis as well as
17 evaluating the feasibility of other models (i.e., a logistic regression, and another assuming a
18 possion distribution) using air quality monitoring data from 1995-2006 (see Appendix D). Each
19 of these models were developed for each location and applied to estimate the number of potential
20 health effect benchmark exceedances at various annual average concentration levels. Following
21 the construction of the models, a few issues with the approach became evident. Because of the
22 limited number of exceedances above 200 ppb in most locations, the best models could only be
23 developed at concentrations lower than this level. Second, some of the locations yielded
24 inadequate models (e.g., non-convergence) that led to a regrouping of the original 20 locations
25 identified above. Third, the predictability of the developed relationships using the varied
26 regression approaches was questionable. Consistently, predictions above the observed maximum
27 annual average concentrations were orders of magnitude higher than the maximum observed
28 number of exceedances. The same occurred with the McCurdy (1994) analysis, though to a
29 lesser degree, since there were at least a few site-years with concentrations at the annual
30 standard. However, upper bound estimations in that 1994 analysis needed to be stunted once
31 predictions were made for concentrations outside of the range of the measured data. Confidence
32 intervals in each both the McCurdy (1994) analyses and those generated as part of this analysis
33 were extremely large.
34
35 It is due to these issues surrounding the applicability of these statistical models to the current
36 ambient monitoring data given the reduced annual average concentrations that a new approach
37 was developed and applied. An empirical model was employed to avoid the difficulties in
38 extrapolating outside the range of the data, combined with the concentration roll-up procedure
39 described in Section 2.5 to estimate short-term concentration exceedances that might occur at
40 concentrations just meeting the current standard.
41 2.7.2 Approach
42 An empirical approach was selected to estimate exceedances at each location. A total of four
43 air quality scenarios were evaluated using the empirical model for each of two distinct ambient
33
-------
1 monitoring periods, resulting in a total of eight separate analyses. The available NO2 air quality
2 were divided into two groups; one contained data from years 1995-2000, representing an
3 historical data set; the other contained the monitoring years 2001-2006, representing recent
4 ambient monitoring. Each of these monitoring year-groups were evaluated considering the NO2
5 concentrations as they were reported and representing the conditions at that time (termed in this
6 assessment "as is"). This served as the first air quality scenario. The second scenario considered
7 the ambient NO2 concentrations simulated to just meeting the current standard of 0.053 ppm
8 annual average. The 3rd and 4th scenarios followed in similar fashion, however these scenarios
9 used the ambient monitoring data to estimate NO2 concentrations that might occur on roadways
10 to generate on-road concentrations for as is air quality and for ambient concentrations just
11 meeting the current standard. Again, each of these four scenarios was evaluated using both the
12 historical and recent data air quality data sets.
13
14 Since all of the NO2 ambient monitoring sites are represented by this analysis, the generated
15 results are considered a broad characterization of national air quality and human exposures that
16 might be associated with these concentrations. The output of this air quality characterization was
17 used to estimate the number of times per year specific locations experience levels of NO2 that
18 could cause adverse health effects in susceptible individuals. Each location that was evaluated
19 contained one to several monitors operating for a few to several years, generating a number of
20 site-years of data. The number of site-years in a location were used to generate a distribution of
21 two exposure and risk characterization metrics; the annual average concentrations and the
22 numbers of exceedances that did (observed data) or could occur (simulated data) in a year for
23 that location. The mean and median values were reported to represent the central tendency of
24 each metric for the four scenarios in each air quality year-group, while the minimum value
25 served to represent the lower bound. Since there were either multiple site-years or numerous
26 simulations performed at each location using all available site-years of data, results for the upper
27 percentiles included the 95th, 98th and 99th percentiles of the distribution.
28
29 2.7.3 Results
30 2.7.3.1 Air Quality Monitoring Data As Is
31 As mentioned previously, air quality data were separated into two groups, one representing
32 historic data (1995-2000) and the other more recent data (2001-2006). Detailed statistics
33 regarding concentration distributions for particular locations and specific monitoring years are
34 provided in Appendices B and C. All of the results in Tables 13-15 are based on air quality data
35 as reported by the AQ S.
36
37 Table 13 provides descriptive statistics for ambient NO2 concentrations and the site-years
38 available for each location and air quality grouping. For example, in Boston, there were 58
39 complete site-years during 1995-2000, for which the annual mean concentrations ranged from
40 about 5 ppb to 31 ppb with a mean annual mean of about 18 ppb. Los Angeles, New York,
41 Phoenix and Denver (recent data only), had higher annual average concentrations at the mean
42 and upper percentiles, considering both the recent and historic air quality data, compared with
43 other locations. Annual average NO2 concentrations have decreased on average by 14% at most
44 of the locations when comparing the historic to the recent year groups, although the mean annual
34
-------
1 average concentrations increased by about 67% at the Denver location using the more recent
2 data.
O
4 The number of short-term concentration exceedances follows in Tables 14 and 15 given
5 potential health effect benchmark levels of 200, 250, and 300 ppb. For example, the numbers of
6 exceedances of 200 ppb ranged from 0 to 1 with a mean estimated number of exceedances of 0
7 for Boston (Table 14). During the years 2001-2006, annual average ambient NC>2 concentrations
8 ranged from 14 to 23 ppb in Detroit considering 122 site-years of data (Table 14). On average
9 there was one exceedance of 200 ppb in Detroit across the total time period, however was as high
10 as 12 given a particular year and site (Table 15).
11
12 In general, the number of exceedances of the selected benchmark levels was low when
13 considering either air quality year-group and at any location. The average number of
14 exceedances of the lowest potential health effect benchmark level across each location was
15 primarily one or less, with very few locations deviating from this estimate. Where locations had
16 a larger mean estimate, it was largely driven be a single site-year of data that contained a number
17 of concentration exceedances. For example, the Colorado Springs mean estimate of exceedances
18 was 3 for the entire area (a total of 8 monitors in operation at some time over the 6 year period of
19 1995-2000), however there was one-site year that contained 69 concentrations above 200 ppb
20 (Table 14). That particular monitor (ID 0804160181) does not appear to have any unusual
21 attributes; the closest major road is beyond a distance of 160 meters, the closest stationary source
22 emitting > 5 tpy over 4 km away, and most sources within 10 km are emitting on average 430
23 tpy (Appendix A). However, one particular source is noted as driving the estimated mean
24 emissions upwards. A power generating utility (NAICS code 221112) located at a 7.2 km
25 distance contained an emission estimate of 4205 tpy, while 9 of the 11 sources located within 10
26 km are under 100 tpy (data not provided). It is not known at this time whether this particular
27 facility is influencing the observed concentration exceedances at this specific monitoring site.
28
29 The same can be stated for the Phoenix location across the same time period, whereas a
30 single year from one monitor (ID 0401330031) was responsible for all observed exceedances of
31 200 ppb (Table 14). While located closer to the roadway (at 78 m) than the Colorado Springs
32 monitor, 9 of 10 stationary sources located within 10 km of this monitor emitted less than 60 tpy
33 (one was at 272 tpy), none of which were located within 5 km. It is not known if observed
34 exceedances of 200 ppb at this monitor are a result of proximity of major roads or stationary
35 sources. Detroit contained the largest number of exceedances of 200 ppb (a maximum of 12)
36 when considering the air quality data from years 2001-2006 (Table 15). Again, all of those
37 exceedances occurred at one monitor (ID 2616300192) during one year (2002). Twelve sources
38 of NOX emissions are located within 2.6 to 5 km of this monitor, contributing between 6 and 27
39 tpy. The number of exceedances of higher potential health effect benchmark levels (i.e., 250 and
40 300 ppb) were of course less than those observed for 200 ppb, most of which were zero, with
41 maximum numbers isolated the same aforementioned cities.
35
-------
Table 13. Monitoring site-years and annual average NO2 concentrations for two monitoring periods, historic and recent air quality data (as is).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs2
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
1995-2000
Site-
Years
58
47
11
26
12
193
24
93
46
69
24
26
14
6
16
22
6
56
1135
200
Annual Mean (ppb) 1
mean min med p95 p98 p99
18
24
23
16
19
26
10
27
23
20
14
16
23
15
14
30
24
18
14
8
5
9
17
6
12
4
6
11
15
9
5
7
14
14
3
24
23
5
1
0
19
24
23
9
19
26
9
27
21
22
15
17
23
15
8
30
24
19
14
7
31
32
28
35
26
45
17
41
33
26
25
24
35
16
27
36
24
26
24
16
31
34
28
35
26
46
17
42
34
27
27
35
35
16
27
40
24
26
26
19
31
34
28
35
26
46
17
42
34
27
27
35
35
16
27
40
24
27
28
19
2001-2006
Site-
Years
47
36
11
10
12
177
20
81
39
66
29
-
30
4
35
27
6
43
1177
243
Annual Mean (ppb) 1
mean min med p95 p98 p99
15
24
19
26
19
22
9
23
20
18
12
-
15
14
11
25
24
15
12
7
5
16
14
18
14
4
6
10
14
7
3
-
8
13
1
11
21
8
1
1
13
23
19
27
19
22
8
24
19
19
14
-
16
14
9
24
23
15
12
6
25
32
24
37
23
36
15
36
29
25
19
-
21
15
22
35
29
22
20
14
30
32
24
37
23
37
16
40
30
26
23
-
22
15
23
37
29
25
22
16
30
32
24
37
23
40
16
40
30
26
23
-
22
15
23
37
29
25
24
16
1 The mean is the sum of the annual means for each monitor 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, 95 , 98th, and 99th percentiles of the distribution for the annual mean.
2 Colorado Springs monitoring data were collected as part of short-term study completed in September 2001 , therefore there are no 2001-2006 data.
36
-------
Table 14. Number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 1995-2000 historic NO2 air quality (as
is).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour 1
mean min med p95 p98 p99
0
0
0
0
0
0
0
0
0
0
0
3
0
0
1
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
3
1
0
0
0
0
0
3
2
0
11
0
0
0
0
0
0
0
1
2
3
2
1
1
3
1
1
69
2
0
11
37
0
0
0
0
1
0
1
2
3
4
1
3
3
2
1
69
2
0
11
37
0
8
0
1
Exceedances of 250 ppb 1-hour 1
mean min med p95 p98 p99
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
3
0
0
0
0
0
0
0
1
1
1
0
0
0
0
1
0
23
0
0
3
3
0
0
0
0
0
0
1
1
1
2
0
0
0
1
0
23
0
0
3
3
0
4
0
0
Exceedances of 300 ppb 1 -hour 1
mean min med p95 p98 p99
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
3
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
4
0
0
3
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
4
0
0
3
0
0
0
0
0
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
37
-------
Table 15. Number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 2001-2006 recent NO2 air quality (as
is).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour
mean min med p95 p98 p99
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
12
0
2
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
12
0
3
0
1
0
0
0
2
0
0
0
0
0
0
0
0
0
0
12
1
3
0
1
0
0
0
2
0
0
0
0
0
1
Exceedances of 250 ppb 1-hour
mean min med p95 p98 p99
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8
0
2
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
8
0
3
0
1
0
0
0
1
0
0
0
0
0
0
0
0
0
0
8
1
3
0
1
0
0
0
1
0
0
0
0
0
1
Exceedances of 300 ppb 1-hour
mean min med p95 p98 p99
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
0
3
0
0
0
0
0
0
0
0
0
0
0
0
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
38
-------
1 2.7.3.2 Simulated Air Quality Data to Just Meet The Current Standard
2 Descriptive statistics for the ambient NO2 concentrations simulated to just meet the current
3 standard is presented in Table 16. Note both year groups (1995-2000 and 2001-2006) contain
4 maximum concentrations of 53 ppb for the annual average concentration, a direct consequence of
5 the concentration roll-up procedure. On average, the mean simulated annual average NO2
6 concentrations are about 2.1-2.4 higher than the respective year group as is concentrations,
7 however the actual range in this factor could be as low as 1.2 or as high as 3.8 depending on year
8 group and particular location selected. This is a function of the location and year specific factors
9 used to simulate just meeting the current standard (Tables 10 and 11) that are also of similar
10 range.
11
12 As expected, the number of estimated short-term concentration exceedances is greater when
13 considering the current standard and considering all potential health effect benchmark levels
14 (Tables 17 and 18). For example, depending on location and year-group, the number of
15 exceedances of 200 ppb on average could be a low as 0 or as high as 88 considering air quality
16 simulated to meet the current standard, although about 75% the location-year groups were
17 estimated as containing a mean of less than 10 exceedances in a year. Median estimates of
18 exceedances of 200 ppb were typically well below that of the mean, 85% were either 1 or less,
19 indicating an upward bias in the mean influenced by the number of exceedances at the upper
20 ends of the distribution.
21
22 The same three cities noted in the as is evaluation above contained the highest mean and
23 maximum number of exceedances when considering the simulated historic air quality data (i.e.,
24 Colorado Springs, Detroit, Phoenix). The same reasoning applies here, primarily the influence
25 of concentration exceedances at a single monitor. Miami and Jacksonville are also indicated as
26 having a relatively higher estimate of mean number of exceedances than the other locations,
27 however this is driven mainly by the small network size (n=l for Jacksonville, n=5 for Miami).
28 Having a limited number of monitors in a given location could bias the mean estimate in either
29 direction (high or low), most notable here where there were an unusual number of peak
30 concentrations in a given year. In addition, Miami contained some of the lowest annual average
31 concentrations (Table 13), yielding the highest air quality simulation factors across all years of
32 data (Tables 10-11). That coupled with a high COV (-130%) for hourly concentrations at a two
33 of the monitors in Miami (IDs 1201180021, 1208600271) clearly played a significant role in the
34 higher estimated number of exceedances. Denver also contained a high COV (-110%) for the
35 earlier air quality period (1995-2000), likely associated with the higher estimate of maximum
36 exceedances at this location (141) following the concentration roll-up compared with only 2
37 observed exceedances when considering the air quality as is. Both the mean and maximum
38 estimate of exceedances for Provo (ID 4904900021) during 2001-2006 were also likely
39 influenced by the small network size (n=l) in this location and one particular year (2006) that
40 contained a number of concentrations above 150 ppb prior to the concentration roll-up.
41
39
-------
Table 16. Estimated annual average NO2 concentrations for two monitoring periods, historic and recent air quality data adjusted to just meet the
current standard (0.053 ppm annual average).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs2
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
1995-2000
Site-
Years
58
47
11
26
12
193
24
93
46
69
24
26
14
6
16
22
6
56
1135
200
Annual Mean (ppb) 1
mean min med p95 p98 p99
32
39
47
29
45
31
34
35
39
42
32
38
43
53
29
45
53
37
26
22
10
15
37
10
26
4
19
14
25
20
11
14
30
53
7
36
53
11
1
1
33
40
53
29
51
32
31
35
35
45
31
45
40
53
17
44
53
39
26
20
53
53
53
53
53
52
53
53
53
53
53
53
53
53
53
53
53
53
43
51
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
48
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
53
50
53
2001-2006
Site-
Years
47
36
11
10
12
177
20
81
39
66
29
-
30
4
35
27
6
43
1177
243
Annual Mean (ppb) 1
mean min med p95 p98 p99
32
41
48
47
49
33
35
35
41
40
34
-
42
53
28
40
53
38
25
22
11
27
41
33
42
5
19
15
26
19
9
-
24
53
4
19
53
19
1
3
28
39
53
53
50
33
32
35
40
44
40
-
43
53
21
40
53
38
26
20
53
53
53
53
53
53
53
53
53
53
53
-
53
53
53
53
53
53
43
46
53
53
53
53
53
53
53
53
53
53
53
-
53
53
53
53
53
53
48
53
53
53
53
53
53
53
53
53
53
53
53
-
53
53
53
53
53
53
51
53
1 The mean is the sum of the annual means for each monitor 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, 95 , 98th, and 99th percentiles of the distribution for the annual mean.
2 Colorado Springs monitoring data were collected as part of short-term study completed in September 2001 , therefore there are no 2001-2006 data.
40
-------
Table 17. Estimated number of exceedances of short-term (1-hour)
adjusted to just meet the current standard (0.053 ppm annual averac
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour
mean min med p95 p98 p99
0
0
3
8
13
1
10
0
0
1
4
30
4
12
3
12
1
0
0
4
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
13
0
8
0
0
0
0
0
1
15
0
0
0
0
0
0
1
0
24
19
25
5
27
1
1
4
19
180
14
20
28
57
5
1
1
18
1
1
24
141
25
8
34
2
12
9
21
241
14
20
28
198
5
1
3
53
2
1
24
141
25
9
34
3
12
17
21
241
14
20
28
198
5
15
6
87
potential health effect benchmark levels in a year, 1995-2000 NO2 air quality
e).
Exceedances of 250 ppb 1-hour
mean min med p95 p98 p99
0
0
1
2
4
0
2
0
0
0
0
15
1
2
1
4
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
10
5
15
0
6
0
0
1
2
123
6
7
13
4
0
0
0
4
1
0
10
28
15
2
15
1
9
3
3
135
6
7
13
92
0
0
1
15
1
0
10
28
15
2
15
3
9
3
3
135
6
7
13
92
0
14
1
42
Exceedances of 300 ppb 1-hour
mean min med p95 p98 p99
0
0
0
1
2
0
1
0
0
0
0
8
0
0
1
1
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
4
10
0
2
0
0
1
1
72
2
1
11
0
0
0
0
1
0
0
3
9
10
0
8
0
5
2
1
83
2
1
11
31
0
0
0
8
1
0
3
9
10
2
8
1
5
2
1
83
2
1
11
31
0
13
1
21
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
41
-------
Table 18. Estimated number of exceedances of short-term (1-hour)
adjusted to just meet the current standard (0.053 ppm annual averac
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour
mean min med p95 p98 p99
0
1
1
2
8
0
17
0
1
0
8
7
31
1
0
88
0
0
3
0
0
0
0
0
0
0
0
0
0
0
0
7
0
0
0
0
0
0
0
0
1
1
1
0
11
0
0
0
0
6
22
0
0
0
0
0
0
1
2
4
7
45
1
66
1
2
2
48
24
72
3
0
526
2
1
17
5
15
4
7
45
5
69
2
25
5
56
27
72
12
1
526
5
3
44
5
15
4
7
45
6
69
5
25
6
56
27
72
12
1
526
5
5
57
potential health effect benchmark levels in a year, 2001-2006 NO2 air quality
e).
Exceedances of 250 ppb 1-hour
mean min med p95 p98 p99
0
0
0
0
4
0
3
0
0
0
1
1
15
0
0
34
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7
0
0
0
0
0
0
0
0
1
2
34
0
18
0
1
1
9
3
46
0
0
205
1
0
4
1
0
1
2
34
0
23
1
7
1
10
6
46
2
0
205
1
1
14
1
0
1
2
34
1
23
1
7
2
10
6
46
2
0
205
1
2
20
Exceedances of 300 ppb 1-hour
mean min med p95 p98 p99
0
0
0
0
3
0
1
0
0
0
0
0
7
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
1
28
0
11
0
0
0
2
0
25
0
0
1
0
0
2
0
0
1
1
28
0
19
0
1
1
5
1
25
0
0
1
1
0
8
0
0
1
1
28
1
19
0
1
1
5
1
25
0
0
1
1
1
9
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
42
-------
1 2.7.3.3 Simulated On-Road Concentrations, Air Quality Data As Is
2 Descriptive statistics for estimated on-road NC>2 concentrations are presented in Table 19.
3 These on-road concentrations were generated by using the simulation procedure described in
4 Section 2.6 as applied to air quality data as is. On average, the simulated on-road annual average
5 concentrations are approximately a factor of 1.8 higher compared with their respective ambient
6 concentrations (see Table 13). This factor is consistent with the range of 1.5 to 2 reported in the
7 ISA (US EPA, 2007f) for studies that compared on-road to ambient NC>2 concentrations. Los
8 Angeles, New York, Phoenix and Denver (recent data only for this location), were locations
9 estimated to contain higher on-road concentrations at the mean and upper percentiles considering
10 both the recent and historic air quality data compared with other locations. This is a direct result
11 of these locations already containing the highest as is concentrations prior to the on-road
12 simulation.
13
14 As a point of reference, the median of the simulated concentration estimates for Los Angeles
15 were compared with NC>2 measurements provided by Westerdahl et al. (2005) for arterial roads
16 and freeways in the same general location during spring 2003. Although the averaging time is
17 not the same,7 comparison of the medians could be considered appropriate. Median on-road
18 concentrations from Westerdahl et al. (2005) ranged from 31 to 55 ppb and compare well with
19 the median of 40 ppb estimated here for years 2001-2006.
20
21 When considering the number of exceedances of 200 ppb estimated to occur on-road, most
22 locations, on average, would have had less than 10 in a year. As observed with the ambient NC>2
23 concentrations, the median frequency of exceedances in most locations were estimated to be
24 typically 1 or less per year, considering both the historic and recent air quality data (Tables 20
25 and 21). However, the number of exceedances at each location were consistently less when
26 considering the recent air quality compared with the historic air quality. There were a few
27 exceptions to these generalities, such as the high number of estimated on-road exceedances of
28 200 ppb for the Colorado Springs and Provo locations. Again, these were the result of these
29 locations having few monitoring sites and a number influential NC>2 concentrations at the upper
30 percentiles of the distribution in one or a few site-years. When considering the two largest
31 groups (all of the other CMS A/MS A and Not CMS A), it is estimated that, on average, about 1 or
32 less exceedances per year of 200 ppb could occur. The 95 percent interval indicates as many as
33 14 exceedances at a particular site within that large grouping for a given year considering the
34 historic data, while only as many as 4 when considering the more recent data.
35
36 There were similarities in the estimated distributions for Chicago, Los Angeles, and New
37 York. Each of these locations are large CMS A, contain several monitoring sites, and have an
38 abundance of roads and associated vehicles.8 Based on the calculations here, each of these
39 locations was estimated to have on average, about 10 exceedances of 200 ppb per year on-roads.
40 Assuming that the on-road exceedances distribution is proportionally representing the
41 distribution of roadways within each location, about one-half of the roads in these areas would
7 Table 13 here considers the median of the annual average while Westerdahl et al. (2005) reported median
concentrations averaged over 2 to 4 hours. In general, there are no differences for the mean annual averages versus
the mean hourly averages (see Appendix B), the main difference in these two metrics is in the variability (and hence
the various percentiles of the distribution outside the central tendency).
8 Of the named locations, Chicago, Los Angeles, and New York contain the highest daily vehicle miles traveled
(Federal Highway Administration (FHWA, 2005)).
43
-------
1 not have any concentrations in excess of 200 ppb. This is because the median value for
2 exceedances of 200 ppb in most locations is zero. However, Tables 20 and 21 indicate that there
3 is also a possibility of tens to just over a hundred exceedances in a year on certain roads/sites.
4
44
-------
Table 19. Estimated annual average on-road concentrations for two monitoring periods, historic and recent ambient air quality (as is).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
1995-2000
Site-
Years
5800
4700
1100
2600
1200
19300
2400
9300
4600
6900
2400
2600
1400
600
1600
2200
600
5600
113500
20000
Annual Mean (ppb)1
mean min med p95 p98 p99
33
44
42
29
35
48
19
50
43
37
26
30
42
28
26
54
43
33
26
14
7
11
22
8
15
5
7
14
19
12
6
9
17
18
4
30
29
7
1
0
33
44
41
19
34
47
17
49
40
38
25
30
40
27
16
52
42
33
25
12
59
68
61
67
52
87
33
81
68
56
49
51
67
37
56
76
58
51
47
31
67
75
65
78
57
97
38
91
76
61
57
64
75
39
62
83
62
58
53
35
71
79
67
81
59
104
39
96
80
64
60
73
82
41
63
88
64
61
57
39
2001-2006
Site-
Years
4700
3600
1100
1000
1200
17700
2000
8100
3900
6600
2900
-
3000
400
3500
2700
600
4300
117700
24300
Annual Mean (ppb)1
mean min med p95 p98 p99
27
43
36
48
34
41
17
43
37
33
21
-
27
25
20
45
43
27
21
12
7
20
18
23
18
5
7
12
18
9
4
-
10
17
2
14
26
10
1
1
25
42
35
46
34
40
15
41
34
33
23
-
27
25
15
43
41
27
21
11
51
66
51
74
47
71
30
70
57
52
40
-
42
34
45
70
61
44
39
27
57
72
54
83
52
80
33
79
63
57
43
-
45
36
50
79
69
49
45
31
60
76
58
87
54
85
36
85
68
61
47
-
48
37
53
84
70
52
48
33
1 The mean is the sum of the annual means for each monitor in a particular location divided by the number of site-years across the monitoring period.
The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the annual mean.
45
-------
Table 20. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a yearon-roads, 1995-2000
historic NO2 air quality (as is).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour
mean min med p95 p98 p99
3
12
10
7
10
45
0
20
5
4
4
20
7
0
6
36
2
2
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
4
0
1
0
0
0
0
2
0
0
3
0
0
0
0
14
79
74
41
48
236
4
109
31
23
31
170
33
1
37
256
9
14
6
2
37
142
108
94
72
417
6
230
60
43
57
264
58
2
66
319
33
25
18
7
54
183
129
102
86
550
8
384
84
58
87
320
76
4
97
390
34
35
32
14
Exceedances of 250 ppb 1-hour
mean min med p95 p98 p99
1
2
2
2
4
13
0
5
1
0
1
11
2
0
1
14
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
15
12
9
21
71
1
28
4
3
3
106
9
0
11
107
0
1
1
1
10
31
30
17
34
146
4
65
11
7
11
181
19
1
15
200
1
8
3
2
15
53
49
33
35
211
6
129
15
11
21
216
30
1
19
280
4
12
6
4
Exceedances of 300 ppb 1-hour
mean min med p95 p98 p99
0
0
1
1
2
4
0
1
0
0
0
6
1
0
1
7
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
1
4
14
21
0
5
1
1
1
47
5
0
6
26
0
0
0
0
1
6
10
6
21
48
3
14
4
2
1
119
7
0
11
103
0
4
1
1
3
10
17
7
26
78
4
31
7
2
2
159
11
0
11
181
0
10
2
2
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
46
-------
Table 21. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a yearon-roads, 2001-2006
historic NO2 air quality (as is).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour
mean min med p95 p98 p99
1
10
3
8
5
11
0
9
1
1
1
1
3
1
3
70
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
2
50
21
39
29
70
3
48
6
6
8
6
15
6
21
547
2
1
1
8
142
36
69
44
131
7
90
14
14
16
9
23
15
44
662
7
5
4
17
188
42
82
45
183
13
143
29
21
25
15
24
23
61
662
14
10
8
Exceedances of 250 ppb 1-hour
mean min med p95 p98 p99
0
2
1
2
2
2
0
2
0
0
0
0
2
0
0
33
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
11
4
8
16
13
2
8
1
0
1
1
8
0
2
234
0
0
0
1
29
7
15
22
29
5
19
1
1
3
1
15
1
5
606
1
1
2
4
44
9
20
28
48
5
25
2
2
6
2
15
3
7
612
2
1
3
Exceedances of 300 ppb 1-hour
mean min med p95 p98 p99
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
13
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
13
2
2
1
0
0
0
0
5
0
0
3
0
0
0
0
6
3
7
14
7
4
3
1
0
1
0
8
0
0
423
0
0
1
0
8
3
7
21
13
5
6
1
0
2
0
8
0
0
435
1
0
2
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
47
-------
1 2.7.3.4 Simulated On-Road Concentrations, Simulated Air Quality Data To Just Meet The Current
2 Standard
3 Descriptive statistics for estimated on-road NC>2 concentrations with just meeting the current
4 standard are presented in Table 22. These on-road concentrations were generated by using the
5 simulation procedure described in Section 2.6 and applied to simulated air quality data to just
6 meet the current standard using the approach described in Section 2.5. On average, the simulated
7 on-road annual average concentrations are also about 1.8 time higher than the ambient
8 concentrations rolled-up to just meet the current standard (see Table 16), similar to what was
9 observed for this relationship considering the air quality (as is).
10
11 The mean number of estimated exceedances of 200 ppb ranges from tens to several hundreds
12 (Tables 23 and 24), sharply increased from the previous on-road estimates using the air quality
13 (as is). Some of the highest exceedance estimates occurred in the locations described previously
14 as being influenced by a few concentrations at the upper percentiles of their distributions in a
15 small number of years and/or monitoring sites (e.g., Miami, Colorado Springs, Provo).
16 Compared to the means, median estimated exceedances of 200 ppb are lower, on average by
17 about 60%, indicating the presence of highly influential data at the upper percentiles of the
18 distribution at each location. This is evident when considering the 95th - 99th percentiles,
19 where several hundred to around two thousand exceedances of 200 ppb were estimated.
20 However, the estimated number of exceedances is lower for locations containing more site-years
21 of data than for the locations with the fewest site-years. This trend is consistent with those
22 described earlier, whereas estimates of exceedances in the simulated data for the large urban
23 areas are stabilized by greater sample size (both the number of monitors and 1-hour values). The
24 median number of exceedances of 200 ppb at the locations containing a larger monitoring
25 network (i.e. at least 40 site-years per year-group) was estimated to be between 10 and 100 per
26 year. Upper bounds for the locations with the greatest number of monitoring sites approach
27 around 1,000 estimated on-road exceedances per year upon just meeting the current standard.
28
29 It should be noted that the estimated on-road concentrations and number exceedances for
30 many of the locations were higher for the 2001-2006 rolled-up data when compared with the
31 1995-2000 rolled-up data. To obtain generally comparable results across the two time periods,
32 the assumption for the concentration roll-up was that a similar level of variability be maintained
33 from year-to-year (or year-group to year-group). As described in section 2.5 of the draft TSD, a
34 slight increase in hourly COV occurred from 1995-2006 (-10% for all locations). The effect
35 may have finally emerged in this combined simulation by generating a greater number of
36 concentrations above the potential health effect benchmarks that may have previously been just
37 below the threshold in the earlier on-road simulations considering the as is ambient
38 concentrations.
39
40
48
-------
Table 22. Estimated annual average on-road concentrations for two monitoring periods, air quality data adjusted to just meet the current standard
(0.053 ppm annual average).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
1995-2000
Site-
Years
5800
4700
1100
2600
1200
19300
2400
9300
4600
6900
2400
2600
1400
600
1600
2200
600
5600
113500
20000
Annual Mean (ppb)1
mean min med p95 p98 p99
58
72
86
53
81
56
62
64
71
77
57
69
77
96
53
82
96
68
46
39
13
18
47
12
33
6
24
18
31
26
14
18
38
67
8
46
67
14
1
1
57
72
84
49
83
55
56
62
67
77
55
73
74
95
34
78
95
68
46
35
103
112
123
112
124
102
111
104
111
116
111
118
122
128
113
115
129
106
84
90
117
123
128
124
129
114
124
117
123
124
126
127
129
131
125
127
139
118
95
104
125
130
136
129
133
122
128
123
128
130
129
131
138
144
130
129
144
124
103
115
2001-2006
Site-
Years
4700
3600
1100
1000
1200
17700
2000
8100
3900
6600
2900
-
3000
400
3500
2700
600
4300
117700
24300
Annual Mean (ppb)1
mean min med p95 p98 p99
58
74
88
85
90
61
63
63
74
73
61
-
75
96
50
72
95
69
46
39
14
35
53
42
54
7
25
18
33
23
12
-
30
67
5
24
67
25
1
3
53
72
86
85
87
60
57
61
71
74
66
-
74
94
36
71
93
67
45
35
105
113
123
124
123
105
112
103
111
114
111
-
112
129
112
110
128
106
84
89
120
124
130
130
129
116
126
119
125
124
126
-
124
139
124
125
131
118
95
101
126
130
146
141
134
123
129
125
128
128
129
-
128
145
129
127
138
126
102
109
1 The mean is the sum of the annual means for each monitor in a particular location divided by the number of site-years across the monitoring period.
The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the annual mean.
49
-------
Table 23. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a yearon-roads, 1995-2000
historic NO2 air quality adjusted to just meet the current standard (0.053 ppm annual average).
Location
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
Colorado Springs
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour1
mean min med p95 p98 p99
78
172
321
214
405
100
363
77
114
219
251
304
178
610
238
250
443
148
52
95
0
0
1
0
2
0
1
0
0
0
0
0
0
40
0
0
1
0
0
0
13
61
195
23
284
18
260
11
27
101
42
77
82
549
26
105
230
48
6
7
411
727
1045
1261
1227
489
1045
412
570
852
1094
1320
692
1426
1107
953
1643
620
268
549
677
1001
1221
1921
1439
791
1334
693
797
1070
1472
1756
951
1515
1674
1326
1871
871
444
928
790
1170
1439
2215
1589
927
1427
930
942
1185
1640
1879
1105
1801
1882
1435
2058
966
592
1203
Exceedances of 250 ppb 1-hour1
mean min med p95 p98 p99
23
59
124
97
175
33
162
23
32
73
106
120
57
263
89
83
135
46
15
39
0
0
0
0
2
0
0
0
0
0
0
0
0
2
0
0
0
0
0
0
1
7
38
5
97
2
93
1
4
18
7
11
24
195
5
17
32
6
0
1
131
303
566
511
576
173
579
127
181
351
535
565
215
773
574
379
543
259
84
221
257
512
663
1142
776
318
737
258
308
457
843
769
347
839
688
466
697
356
156
438
334
643
761
1574
872
432
791
420
364
525
947
930
447
1002
860
563
817
432
231
635
Exceedances of 300 ppb 1-hour1
mean min med p95 p98 p99
8
22
51
45
80
12
72
8
9
27
45
60
21
114
36
33
43
16
5
17
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
1
40
0
32
0
0
2
1
1
8
66
1
3
2
0
0
0
43
137
304
228
317
62
316
40
52
158
277
294
78
407
280
181
208
99
25
91
106
230
380
582
424
127
396
91
104
220
435
371
162
443
369
296
303
163
57
198
131
322
392
908
482
184
430
171
138
270
514
416
200
470
422
364
339
200
90
318
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
50
-------
Table 24. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a yearon-roads, 2001-2006 recent
NO2 air quality adjusted to just meet the current standard (0.053 ppm annual average).
OLocation
Boston
Chicago
Cleveland
Denver
Detroit
Los Angeles
Miami
New York
Philadelphia
Washington
Atlanta
El Paso
Jacksonville
Las Vegas
Phoenix
Provo
St. Louis
Other CMSA
Not MSA
Exceedances of 200 ppb 1-hour1
mean min med p95 p98 p99
87
176
387
277
440
106
406
84
174
208
335
389
607
278
149
516
182
64
101
0
0
14
0
17
0
3
0
0
0
0
4
56
0
0
1
0
0
0
12
61
268
113
309
23
306
14
60
83
135
257
542
43
19
345
69
6
7
458
805
1117
964
1214
533
1173
458
726
874
1293
1251
1385
1319
758
1664
762
333
569
753
1022
1322
1233
1444
788
1345
709
973
1171
1647
1604
1642
1929
1172
1966
1100
569
874
990
1139
1735
1560
1628
893
1416
872
1184
1310
1755
1737
1743
2196
1352
2115
1216
740
1095
Exceedances of 250 ppb 1-hour1
mean min med p95 p98 p99
23
59
149
87
166
31
193
25
51
63
143
144
273
101
33
228
59
19
39
0
0
0
0
0
0
0
0
0
0
0
0
5
0
0
0
0
0
0
1
7
65
22
90
2
113
1
7
10
21
66
202
6
1
72
8
0
1
137
335
573
337
513
186
669
149
239
327
687
530
789
680
203
729
302
105
232
263
560
676
430
689
290
855
295
383
426
973
858
924
828
303
818
468
207
419
330
620
846
557
744
363
923
413
521
558
1093
971
1027
1045
370
847
576
300
569
Exceedances of 300 ppb 1-hour1
mean min med p95 p98 p99
7
23
62
28
67
10
88
8
16
21
61
54
125
42
7
134
20
6
16
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
15
5
25
0
35
0
1
1
4
20
74
0
0
5
1
0
0
38
128
326
125
265
59
367
49
77
127
339
221
436
354
48
643
127
31
95
93
295
407
203
322
115
542
110
153
181
510
350
490
502
70
693
211
72
184
132
354
428
283
385
150
588
177
227
224
656
441
557
565
95
694
260
120
264
1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site-
years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for
the number of exceedances in any one year within the monitoring period.
51
-------
i 2.8 Variability and Uncertainty
2 This uncertainty analysis first identifies the sources of the assessment that do or do not
3 contribute to uncertainty, and provide a rationale for why this is the case. A qualitative
4 evaluation follows for the types and components of uncertainty, resulting in a matrix describing,
5 for each source of uncertainty, both the direction and magnitude of influence has on exposure
6 estimates. The bias direction indicates how the source of uncertainty is judged to influence
7 estimated concentrations, either the concentrations are likely "over-" or "under-estimated". In
8 the instance where two types or components of uncertainty result in offsetting direction of
9 influence, the uncertainty was judged as "both". The magnitude indicates an estimated size of
10 influence the uncertainty has on estimated concentrations. "Minimal" uncertainty was noted
11 where quantitative evidence indicates the influence is either conditional and/or limited to few
12 components in type. A characterization of "moderate" was assigned where multiple components
13 of uncertainty existed within a given type and act in similar direction, however the presence of
14 all at once may be dependent on certain conditions. "Major" uncertainty was used where
15 multiple components of uncertainty exist within a given type, the components have few limiting
16 conditions, and the components consistently act in similar bias direction. "Unknown" was
17 assigned where there was no evidence reviewed to judge the uncertainty associated with the
18 source. Table 25 provides a summary of the sources of uncertainty identified in the air quality
19 characterization and the judged bias and magnitude of each.
20 2.8.1 Air Quality Data
21 One basic assumption is that the AQS NC>2 air quality data used are quality assured already.
22 Reported concentrations contain only valid measures, since values with quality limitations are
23 either removed or flagged. There is likely no selective bias in retention of data that is not of
24 reasonable quality, it is assumed that selection of high concentration poor quality data would be
25 just as likely as low concentration data of poor quality. Given the numbers of measurements
26 used for this analysis, it is likely that even if a few low quality data are present in the data set,
27 they would not have any significant effect on the results presented here. Therefore, the air
28 quality data and database used likely contributes minimally to uncertainty. Temporally, the data
29 are hourly measurements and appropriately account for variability in concentrations that are
30 commonly observed for NO2 and by definition are representative of an entire year. In addition,
31 having more than one monitor does account for some of the spatial variability in a particular
32 location. However, the degree of representativeness of the monitoring data used in this analysis
33 can be evaluated from several perspectives, one of which is how well the temporal and spatial
34 variability are represented. In particular, missing hourly measurements at a monitor may
35 introduce bias (if different periods within a year or different years have different numbers of
36 measured values) and increase the uncertainty. Furthermore, the spatial representativeness will
37 be poor if the monitoring network is not dense enough to resolve the spatial variability (causing
38 increased uncertainty) or if the monitors are not evenly distributed (causing a bias). Additional
39 uncertainty regarding temporal and spatial representation by the monitors is expanded below.
40 2.8.2 Measurement Technique for Ambient NO2
41 One source of uncertainty for NO2 air quality data is due to interference with other oxidized
42 nitrogen compounds. The ISA points out positive interference, commonly from HNOs, of up to
43 50%, particularly during the afternoon hours, resulting in overestimation of concentrations.
52
-------
1 Also, negative vertical gradients exist for monitors (2.5 times higher at 4 meter vs. 15 meter
2 vertical siting (draft ISA, section 2.5.3.3), thus monitors positioned on rooftops may
3 underestimate exposures. Only 7 of the 1119 monitors in the named locations contained
4 monitoring heights of 15 meters or greater, with nearly 60% at 4 meters or less height, and 80%
5 at 5 meters or less in height. Not accounting for this potential vertical gradient in NC>2
6 concentrations may generate underestimates of exceedances for some site-years, however the
7 overall impact of inferences made for the locations included in this assessment is likely minimal
8 since most monitors sited at less than 4-5 meters in vertical height.
9 2.8.3 Temporal Representation
10 Data are valid hourly measures and are of similar temporal scale as the potential health effect
11 benchmark concentrations. There are frequent missing values within a given valid year which
12 contribute to the uncertainty as well as introducing a possible bias if some seasons, day types
13 (e.g., weekday/weekend), or time of the day (e.g., night or day) are not equally represented.
14 Since a 75 percent daily and hourly completeness rule was applied, some of these uncertainties
15 and biases were reduced in these analyses. Data were not interpolated in the analysis. Similarly,
16 there may be bias and uncertainly if the years monitored vary significantly between locations.
17 Although monitoring locations within a region do change over time, the NC>2 network has been
18 reasonably stable over the 1995-2006 period, particularly at locations with larger monitoring
19 networks, so the impact to uncertainty is expected to be minimal regarding both bias direction
20 and magnitude. It should also be noted that use of the older data in some of the analyses here
21 carries the assumption that the sources present at that time are the same as current sources,
22 adding uncertainty to results if this is not the case. Separating the data into two 5 year groups
23 (historic and recent) before analysis reduces the potential impact from changes in national- or
24 location-specific source influences and is judged to have a minimal magnitude.
25 2.8.4 Spatial Representation
26 Relative to the physical area, there are only a small number of monitors in each location.
27 Since most locations have sparse siting, the monitoring data are assumed to be spatially
28 representative of the locations analyzed here. This includes areas between the ambient monitors
29 that may or may not be influenced by similar local sources of NC>2. For these reasons the
30 uncertainty and bias due to the spatial network may be moderate, although the monitoring
31 network design should have addressed these issues within the available resources and other
32 monitoring constraints. This air quality characterization used all monitors meeting the 75
33 percent completeness criteria, without taking into account the monitoring objectives or land use
34 for the monitors. Thus, there will be some lack of spatial representation and likely moderate
35 uncertainty due to the inclusion/exclusion of some monitors that are very near local sources
36 (including mobile sources).
37 2.8.5 Air Quality Adjustment Procedure
38 The primary uncertainty of the empirical method used to estimate exceedances under the
39 current-standard scenario is due to the uncertainty of the true relationship between the annual
40 mean concentrations and the number of exceedances. The empirical method assumes that if the
41 annual means change then all the hourly concentrations will change proportionately. However,
' 28 monitors did not have height reported (therefore, 177 + 28 = 205 total number of monitors in named locations)
53
-------
1 different sources have different temporal emission profiles, so that applied changes to the annual
2 mean concentrations at monitors may not correspond well to all parts of the concentration
3 distribution equally. Similarly, emissions changes that affect the concentrations at the site with
4 the highest annual mean concentration will not necessarily impact lower concentration sites
5 proportionately. This could result in overestimations in the number of exceedances at lower
6 concentration sites within a location, however it is likely to be minimal given that the highest
7 concentrations typically were measured at the monitoring sites with the highest annual average
8 concentrations within the location (draft TSD, Appendix C). This minimal bias would apply to
9 areas that contain several monitors, such as Boston, New York, or Los Angeles. Universal
10 application of the proportional simulation approach at each of the locations was done for
11 consistency and was designed to preserve the inherent variability in the concentration profile. A
12 few locations were noted that may have an exceptional number of exceedances as a result of the
13 air quality adjustment approach, particularly those locations with few monitoring sites that
14 contained very low annual average concentrations and/or atypical variability in hourly
15 concentrations. These locations (e.g., Miami, Jacksonville, Provo) could contain moderate
16 overestimations at the upper tails of the concentration distribution, leading to bias in number of
17 estimated exceedances at both the upper percentiles and the mean for the scenarios using the air
18 quality simulated to just meet the current standard.
19 2.8.6 On-Road Concentration Simulation
20 On-road and ambient monitoring NC>2 concentrations have been shown to be correlated
21 significantly on a temporal basis (e.g., Cape et al., 2004) and motor vehicles are a significant
22 emission source of NOX, providing support for estimating on-road concentrations using ambient
23 monitoring data. The relationship used in this analysis to estimate on-road NO2 concentrations
24 was derived from data collected in measurement studies containing mostly long-term averaging
25 times, typically 14-days or greater in duration (e.g., Roorda-Knape, 1998; Pleijel et al., 2004;
26 Cape et al, 2004), although one study was conducted over a one-hour time averaging period
27 (Rodes and Holland, 1981). This is considered appropriate in this analysis to estimate on-road
28 hourly concentrations from hourly ambient measures, assuming a direct relationship exists
29 between the short-term peaks to time-averaged concentrations (e.g., hourly on-road NC>2
30 concentrations are correlated with 24-hour averages). While this should not impact the overall
31 contribution relationship between vehicles and ambient concentrations on roads, the decay
32 constant k will differ for shorter averaging times. The on-road concentration estimation also
33 assumes that concentration changes that occur on-road and at the monitor are simultaneous (i.e.,
34 within the hour time period of estimation). Since time-activity patterns of individuals are not
35 considered in this analysis, there is no bias in the number of estimated exceedances. The long-
36 term data used to develop the model were likely collected over variable meteorological
37 conditions (e.g., shifting wind direction) and other influential attributes (e.g., rate of
38 transformation of NO to NC>2 during the daytime versus nighttime hours) than would be observed
39 across shorter time periods. This could result in either over- or under-estimations of
40 concentrations, depending on the time of day. The variability in NC>2 concentration within an
41 hour is also not considered in this analysis, that is, the on-road concentration at a given site will
42 likely vary during the 1-hour time period. If considering personal exposures to individuals
43 within vehicles that are traveling on a road, it is likely that their exposure concentrations would
44 also vary due to differing roadway concentrations. This could also result in either over- or
54
-------
1 under-estimations of concentrations, depending on the duration of travel and type of road
2 traveled on.
O
4 On-road concentrations were not modified in this analysis to account for in-vehicle
5 penetration and decay. This indicates that in-vehicle concentrations would be overestimated if
6 using the on-road concentrations as a surrogate, given that reactive pollutants (e.g., PM2 5) tend
7 to have a lower indoor/outdoor (I/O) concentration ratio (Rodes et al., 1998). Chan and Chung
8 (2003) report mean (I/O) ratios of NO2 for a few roadways and driving conditions in Hong Kong.
9 On highways and urban streets, the value is centered about 0.6 to 1.0, indicating decay of NO2 as
10 it enters the vehicle.
11
12 At locations where traffic counts are very low (e.g., on the order of hundreds/day) the on-
13 road contribution has been shown to be negligible (Bell and Ashenden, 1997; Cape et al., 2004),
14 therefore any rural areas just meeting the standard with minimal traffic volumes would likely
15 have resulted in small overestimations of NO2 concentrations using eq (2). For any monitor that
16 is sited in close proximity of the roadway (14 monitors were sited at <10 m from a major road),
17 on-road concentrations may have been overestimated using eq (2), since the assumption is that
18 the ambient concentration is equivalent to the non-source impacted concentration. In some
19 locations (i.e., Boston, Chicago, Denver, Los Angeles, Miami, St. Louis, and Washington DC),
20 at least half of the monitors used in this analysis are sited < 100 m from a major road (see Table
21 5, section 2.3.3), a distance noted by some researchers a possibly receiving notable impact from
22 vehicle emissions (e.g., Beckerman et al., 2008). In addition, NOX is primarily emitted as NO
23 (e.g., Heeb et al., 2008; Shorter et al., 2005), with substantial secondary formation due
24 predominantly to NO + O3 -> NO2 + O2. Numerous studies have demonstrated the O3 reduction
25 that occurs near major roads, reflecting the transfer of odd oxygen to NO to form NO2, a process
26 that can impact NO2 concentrations both on- and downwind of the road. Some studies report
27 NO2 concentrations increasing just downwind of roadways and that are inversely correlated with
28 Os (e.g., Beckerman et al., 2008), suggesting that peak concentration of NO2 may not always
29 occur on the road, but at a distance downwind. Uncertainty regarding where the peak
30 concentration occurs (on-road or at a distance from the road) in combination with the form of the
31 exponential model used to estimate the on-road concentrations (the highest concentration occurs
32 at zero distance from road) could also lead to overestimation. However, the interpretation of the
33 estimate is what may be most uncertain, that is whether the exceedances are occurring on the
34 road or nearby.
35
36 Another source of uncertainty is the extent to which the near-road study locations represent
37 the locations studied in these analyses. The on-road and near-road data were collected in a few
38 locations, most of them outside of the United States. The source mixes (i.e., the vehicle fleet) in
39 study locations may not be representative of the U.S. fleet. Without detailed information
40 characterizing the emissions patterns for the on-road study areas, there was no attempt to match
41 the air quality characterization locations to specific on-road study areas, which might have
42 improved the precision of the estimates. However, since concentration ratios were selected
43 randomly from all the near-road studies and applied to each monitor individually, and since we
44 estimated overall minimum and upper bounds using multiple simulations, the analysis provides a
45 reasonable lower and upper bound estimate of the uncertainty.
55
-------
1 2.8.7 Health Benchmark
2
3
4
5
6
7
8
9
10
11
12
13
14
15
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. Use of a 1-hour
averaging time could 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.
Table 25. Summary of qualitative uncertainty analysis for the air quality characterization.
Source
Air Quality Data
Ambient Measurement
Temporal Representation
Spatial Representation
Air Quality Adjustment
On-Road Simulation
Health Benchmarks
Type
Database quality
Interference
Vertical siting
Scale
Missing data
Years monitored
Source changes
Scale
Monitor objectives
Temporal scale
Spatial scale
Temporal scale
Decay
Spatial scale
Model used
Non US studies used
Averaging time
Susceptibility
Bias Direction
both
over
under
none
both
both
over
both
both
over
over
both
over
over
over
unknown
unknown
under
Magnitude
minimal
moderate
minimal
none
minimal
minimal
minimal
moderate
moderate
moderate
moderate
minimal
minimal
moderate
minimal
unknown
moderate
moderate
Notes:
Bias Direction: indicates the direction the source of uncertainty is judged to influence either the
concentration or risk estimates.
Magnitude: indicates the estimated size of influence.
minimal - influence is either conditional and/or limited to few components in type
moderate - multiple components of uncertainty existed within a given type and act in similar
direction, however the presence of all at once may be dependent on certain conditions.
major - multiple components of uncertainty exist within a given type, the components have few
limiting conditions, and the components consistently act in similar bias direction.
16
56
-------
2 2.9 References
3 Bell S and Ashenden TW. (1997). Spatial and temporal variation in nitrogen dioxide pollution
4 adj acent to rural roads. Water Air Soil Pollut. 95:87-98.
5 Beckerman B, Jerrett M, Brook JR, Verma DK, Arain MA, Finkelstein MM. (2008).
6 Correlation of nitrogen dioxide with other traffic pollutants near a major expressway. Atmos
1 Environ. 42:275-290.
8 Signal KL, Ashmore MR, Headley AD, Stewart K, Weigert K. (2007). Ecological impacts of
9 air pollution from road transport on local vegetation. Applied Geochemistry. 22:1265-1271.
10 Cape JN, Tang YS, van Dijk N, Love L, Sutton MA, Palmer SCF. (2004). Concentrations of
11 ammonia and nitrogen dioxide at roadside verges, and their contribution to nitrogen
12 deposition. Environ Pollut. 132:469-478.
13 Chan AT and Chung MW. (2003). Indoor-outdoor air quality relationships in vehicle: effect of
14 driving environment and ventilation modes. Atmos Environ. 37:3795-3808.
15 FHWA. (2005). Highway Statistics 2005, Urbanized Areas - 2005, Miles and Daily Vehicle-
16 Miles of Travel (Table HM-71). Available at:
17 http://www.fhwa.dot.gov/policy/ohim/hs05/htm/hm71 .htm.
18 Gilbert NL, Woodhouse S, Stieb DM, Brook JR. (2003). Ambient nitrogen dioxide and distance
19 from a major highway. Sci TotalEnviron. 312:43-46.
20 Heeb NV, Saxer CJ, Forss A-M, Bruhlmann S. (2008). Trends of NO-, NO2-, and NH3-
21 emissions from gasoline-fueled Euro-3- to Euro-4-passenger cars. Atmos Environ.
22 42(10):2543-2554.
23 Maruo YY, Ogawa S, Ichino T, Murao N, Uchiyama M. (2003). Measurement of local
24 variations in atmospheric nitrogen dioxide levels in Sapporo, Japan using a new method with
25 high spatial and high temporal resolution. Atmos Environ. 37:1065-1074.
26 McCurdy TR. (1994). Analysis of high 1 hour NO2 values and associated annual averages
27 using 1988-1992 data. Report to the Office of Air Quality Planning and Standards, Durham
28 NC.
29 Monn Ch, Carabias V, Junker M, Waeber R, Karrer M, Wanner FIU. (1997). Small-scale spatial
30 variability of particulate matter <10 |j,m (PMio) and nitrogen dioxide. Atmos Environ.
31 31(15)2243-2247.
32 Nitta H, Sato T, Nakai S, Maeda K, Aoki S, Ono M. (1993). Respiratory health associated with
33 exposure to automobile exhaust. I. Results of cross-section studies in 1979, 1982, 1983.
34 Arch Environ Health. 48(l):53-58.
35 Pleijel H, Karlsson GP, Gerdin EB. (2004). On the logarithmic relationship between NO2
36 concentration and the distance from a highroad. Sci Total Environ. 332:261-264.
37 Rodes C, Sheldon L, Whitaker D, Clayton A, Fitzgerald K, Flanagan J, DiGenova F, Hering S,
38 Frazier C. (1998). Measuring Concentrations of Selected Air Pollutants Inside California
39 Vehicles. California Environmental Protection Agency, Air Resources Board. Final Report,
40 December 1998.
41 Rodes CE and Holland DM. (1981). Variations of NO, NO2 and Os concentrations downwind
42 of a Los Angeles freeway. Atmos Environ. 15:243-250.
43 Roorda-Knape MC, Janssen NAH, De Hartog JJ, Van Vliet PHN, Harssema H, Brunekreef B.
44 (1998). Air Pollution from traffic in city districts near major roadways. Atmos Environ.
45 32(11)1921-1930.
57
-------
1 Shorter JH, Herndon S, Zahniser MS, Nelson DD, Wormhoudt J, Demerjian KL, Kolb CE.
2 (2005). Real-Time measurements of nitrogen oxide emissions from in-use New York City
3 transit buses using a chase vehicle. Environ Sci Technol. 39:7991-8000.
4 Singer BC, Hodgson AT, Hotchi T, Kim JJ (2004). Passive measurement of nitrogen oxides to
5 assess traffic-related pollutant exposure for the East Bay Children's Respiratory Health
6 Study. AtmosEnviron. 38:393-403.
7 US EPA. (2007a). US EPA Air Quality System (AQS). Download Detailed AQS Data.
8 Available at: http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm.
9 US EPA. (2007b). Field Guide to Air Quality Data (vl.0.0). February 28, 2007. Available at:
10 http://www.epa.gov/ttn/airs/aqsdatamart/documentation/index.htm.
11 US EPA. (2007c). Nitrogen Dioxide Health Assessment Plan: Scope and Methods for Exposure
12 and Risk Assessment. September 2007. Office of Air Quality Planning and Standards.
13 Available at: http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_pd.html.
14 US EPA. (2007d). ALLNEICAP Annual 11302007 file posted at:
15 http://www.epa.gov/ttn/chief/net/2002inventory.htmlffinventorydata.
16 US EPA. (2007e). Air Trends. Nitrogen Dioxide, http://www.epa.gov/airtrends/nitrogen.html.
17 Westerdahl D, Fruin S, Sax T, Fine PM, Sioutas C. (2005). Mobile platform measurements of
18 ultrafine particles and associated pollutant concentrations on freeways and residential streets
19 in Los Angeles. Atmos Environ. 39:3597-3610.
20 US EPA. (2007f). Integrated Science Assessment for Oxides of Nitrogen-Health Criteria
21 (First External Review Draft) and Annexes (August 2007). Research Triangle Park, NC:
22 National Center for Environmental Assessment. Available at:
23 http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l 81712.
24
25
26
58
-------
i 3 Exposure Assessment and Health Risk Characterization
2 3.1 Introduction
3 This section documents the methodology and input data used in the inhalation exposure
4 assessment for NC>2 conducted in support of the current review of the NC>2 primary NAAQS.
5 Two important components of the analysis include the approach for estimating temporally and
6 spatially variable NC>2 concentrations and simulating contact of humans with these pollutant
7 concentrations. Both air quality and exposure modeling approaches have been used here to
8 generate estimates of 1-hour NC>2 exposures within selected urban areas of the U.S. Details on
9 the approaches used are provided below and include the following:
10
11 • Description of the areas assessed and populations considered
12 • Summary of the air quality modeling methodology and associated input data
13 • Description of the inhalation exposure model and associated input data
14 • Evaluation of estimated NC>2 exposures using modeling methodology
15 • Assessment of the quality and limitations of the input data for supporting the goals of the
16 NC>2 NAAQS exposure analysis.
17
18 The selected modeling approach was both time and labor intensive. To date, only the
19 exposure and risk results for the Philadelphia case-study are complete and are presented in this
20 draft document. Location-specific input data for Philadelphia and the other selected case-study
21 areas are presented where collected (mainly meteorological data) to provide information on the
22 relative variability of the input data to be used.
23 3.1.1 Selection of Study Areas
24 The selection of areas to include in the exposure analysis takes into consideration the location
25 of field and epidemiology studies, the availability of ambient monitoring and other input data,
26 the desire to represent a range of geographic areas, population demographics, general
27 climatology, and results of the ambient air quality characterization.
28 Locations of interest were initially identified through a similar statistical analysis of the
29 ambient NC>2 air quality data described above for each site within a location. Criteria were
30 established for selecting sites with high annual means and/or high numbers of exceedances of
31 potential health effect benchmark concentrations. The analysis considered all data combined, as
32 well as the more recent air quality data (2001-2006) separately.
33
34 The 90th percentile served as the point of reference for the annual means, and across all
35 complete site-years for 2001-2006, this value was 23.5 ppb. Seventeen locations contained one
36 or more site-years with an annual average concentration at or above the 90th percentile. When
37 combined with the number of 1-hour NC>2 concentrations at or above 200 ppb, only two locations
38 fit these criteria, Philadelphia and Los Angeles. Considering the short-term criterion alone,
39 Detroit contained the greatest number of exceedances of 200 ppb (numbering 12 for years 2001-
40 2006). Two additional locations were selected by considering geographic/climatologic
41 representation and also their historic ambient concentrations. Atlanta (1 exceedance of 200 ppb
42 and a maximum annual average concentration of 26.6 ppb for years 1995-2006) and Phoenix
59
-------
1 (maximum annual mean concentration of 37.1 ppb for 2001-2006 and 37 exceedances of 200
2 ppb for years 1995-2006) were selected to represent the southern and western region of the US
3 from the pool of remaining locations with either exceedances of the 90th percentile annual mean
4 concentration or 200 ppb 1-hour.
5
6 To summarize, the following 5 urban areas were selected for a detailed exposure analysis:
7 • Philadelphia, PA
8 • Atlanta, GA
9 • Detroit, MI
10 • Los Angeles, CA
11 • Phoenix, AZ
12
13 3.1.2 Exposure Periods
14 The exposure periods modeled were 2001 through 2003 to envelop the most recent year of
15 travel demand modeling (TDM) data available for the respective study locations (i.e., 2002) and
16 to include a 3 years of meteorological data to achieve a degree of stability in the dispersion and
17 exposure model estimates.
18 3.1.3 Populations Analyzed
19 A detailed consideration of the population residing in each modeled area was included where
20 the exposure modeling was performed. The assessment includes the general population (All
21 Persons) residing in each modeled area and considered susceptible and vulnerable populations as
22 identified in the ISA. These include population subgroups defined from either an exposure or
23 health perspective. The population subgroups identified by the ISA (US EPA, 2007a) that were
24 included and that can be modeled in the exposure assessment include:
25
26 • Children (ages 5-18)
27 • Asthmatic children (ages 5-18)
28 • All persons (all ages)
29 • All Asthmatics (all ages)
30
31 In addition to these population subgroups, individuals anticipated to be exposed more
32 frequently to NC>2 were considered, including those commuting on roadways and persons
33 residing near major roadways. To date, this document provides a summary of the subpopulations
34 of interest (all asthmatics and asthmatic children), supplemented with additional exposure and
35 risk results for the total population where appropriate.
36 3.2 Dispersion Modeling
37 Air quality data used for input to APEX were generated using AERMOD, a steady-state,
38 Gaussian plume model (EPA, 2004). For each identified case-study location, the following steps
39 were performed
40 1. Collect and analyze general input parameters. Meteorological data, processing
41 methodologies used to derive input meteorological fields (e.g., temperature, wind
60
-------
1 speed, precipitation), and information on surface characteristics and land use are
2 needed to help determine pollutant dispersion characteristics, atmospheric
3 stability and mixing heights.
4 2. Estimate emissions. The emission sources modeled included, major stationary
5 emission sources, on-road emissions that occur on major roadways, and fugitive
6 emissions.
7 3. Define receptor locations. Three sets of receptors were identified for the
8 dispersion modeling, including ambient monitoring locations, census block
9 centroids, and links along major roadways.
10 4. Estimate concentrations at receptors. Hourly concentrations were estimated for
11 each year of the simulation (years 2001 through 2003) by combining
12 concentration contributions from each of the emission sources and accounting for
13 sources not modeled.
14
15 The AERMOD model predictions were then used as input to the APEX model to estimate
16 population exposure concentrations for Philadelphia County. Hourly NO2 concentrations were
17 estimated for each of 3 years (2001-2003) at each of the defined receptor locations (census
18 blocks and roadway links) using hourly NOX emission estimates and dispersion modeling.
19 Relevant input data collected for Philadelphia as well some of the data collected as part of the
20 other selected case-study locations to be evaluated in the second draft risk and exposure
21 assessment are presented below.
22 3.2.1 Meteorological Inputs
23 All meteorological data used for the AERMOD dispersion model simulations were processed
24 with the AERMET meteorological preprocessor, version 06341. This section describes the input
25 data and processing methodologies used to derive input meteorological fields for each of the five
26 regions of interest.
27
28 3.2.1.1 Data Selection
29 Raw surface meteorological data for the 2001 to 2003 period were obtained from the
30 Integrated Surface Hourly (ISH) Database,10 maintained by the National Climatic Data Center
31 (NCDC). The ISH data used for this study consists of typical hourly surface parameters
32 (including air and dew point temperature, atmospheric pressure, wind speed and direction,
33 precipitation amount, and cloud cover) from hourly Automated Surface Observing System
34 (ASOS) stations. No on-site observations were used.
35
36 Surface meteorological stations for this analysis were those at the major airports of each of
37 the five cities in the study:
38
39 • Atlanta: Atlanta Hartsfield International (KATL)
40 • Detroit: Detroit Metropolitan (KDTW)
41 • Los Angeles: Los Angeles International (KLAX)
42 • Philadelphia: Philadelphia International (KPHL)
43 • Phoenix: Phoenix Sky Harbor International (KPHX).
1 http://wwwl .ncdc.noaa.gov/pub/data/techrpts/tr20010 l/tr2001-01 .pdf
61
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
The selection of surface meteorological stations for each city minimized the distance from
the station to city center, minimized missing data, and maximized land-use representativeness of
the station site compared to the city center.
The total number of surface observations per station, and the percentage of those
observations accepted by AERMET (i.e., those observations that were both not missing and
within the expected ranges of values), are shown by Table 26.
Note that instances of calm winds are not rejected by the AERMET processor, but are later
treated as calms in the dispersion analysis. There were 2,538 hours in Atlanta with calm winds
(10% of the total hourly records), 1,924 in Detroit (7%), 3,190 in Los Angeles (12%), 1,772 in
Philadelphia (7%), and 3,559 in Phoenix (14%) (see Table 27).
Table 26. Number of AERMET raw hourly surface meteorology observations and percent acceptance
rate, 2001-2003.3
Surface
Variable
Precipitation
Station Pressure
Cloud Height
Sky Cover
Horizontal
Visibility
Temperature
Dew Point
Temperature
Relative
Humidity
Wind Direction
Wind Speed
Atlanta
(KATL)
N=26281
% Accepted
100
99
99
97
99
99
99
99
94
99
Detroit
(KDTW)
N=26271
% Accepted
100
99
99
97
99
99
99
99
97
99
Los Angeles
(KLAX)
N=26276
% Accepted
100
99
99
97
99
100
100
100
92
100
Philadelphia
(KPHL)
N=26268
% Accepted
100
99
99
95
99
99*
99
99
97
99
Phoenix
(KPHX)
N=26279
% Accepted
100
99
99
97
100
85*
99
99
91
99
Notes:
3 Percentages are rounded down to the nearest integer. All data obtained from the NCDC ISH
database.
* The majority of unaccepted records are due to values being out of range.
<95% of observations were accepted.
17
18
Table 27. Number of calms reported by AERMET by year and location.
2001
2002
2003
Total
Atlanta
917
856
765
2538
Detroit
547
619
758
1924
Los Angeles
1051
1019
1120
3190
Philadelphia
610
470
692
1772
Phoenix
1152
1233
1174
3559
19
62
-------
1 Mandatory and significant levels of upper-air data were obtained from the NOAA
2 Radiosonde Database.11 Upper air observations show less spatial variation than do surface
3 observations; thus they are both representative of larger areas and measured with less spatial
4 frequency than are surface observations. The selection of upper-air station locations for each
5 city minimized both the proximity of the station to city center and the amount of missing data in
6 the records. The selected stations are:
7
8 • Atlanta: Peachtree City (KFFC)
9 • Detroit: Detroit/Pontiac (KDTX)
10 • Los Angeles: Miramar Naval Air Station near San Diego (KNKX)
11 • Philadelphia: Washington Dulles Airport (KIAD)
12 • Phoenix: Tucson (KTWC).
13 •
14 The total number of upper-air observations per station per height interval, and the percentage
15 of those observations accepted by AERMET, are shown in Table 28.
11 http://raob.fsl.noaa.gov/
63
-------
Table 28. Number and AERMET acceptance rate of upper-air observations 2001-2003.
Height
Level
Surface
0-500m
500-
1000m
1000-
1500m
1500-
2000m
2000-
2500m
Variable
Pressure
Height
Temperature
DewPoint Temperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPoint Temperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
Atlanta (KFFC)
n
2124
2124
2124
2124
2124
2124
3418
3418
3418
3418
3418
3418
4133
4133
4133
4133
4133
4133
4336
4336
4336
4336
4336
4336
3203
3203
3203
3203
3203
3203
3171
3171
3171
3171
3171
% Accepted
100
100
100
100
99
89*
100
100
100
99
29
29
100
100
100
99*
62
62
100
100
100
96*
72
72
100
100
100
95*
50
50
100
100
100
94*
52
Detroit (KDTX)
n
2125
2125
2125
2125
2125
2125
4577
4577
4577
4577
4577
4577
3059
3059
3059
3059
3059
3059
4739
4739
4739
4739
4739
4739
3351
3351
3351
3351
3351
3351
3078
3078
3078
3078
3078
% Accepted
100
100
100
100
100
98*
100
100
100
99*
64
64
100
100
100
98*
50
50
100
100
100
96*
67
67
100
100
100
95*
46
46
100
100
100
92*
50
Los Angeles (KNKX)
n
2166
2166
2166
2166
2166
2166
5775
5775
5775
5775
5775
5775
6058
6058
6058
6058
6058
6058
4473
4473
4473
4473
4473
4473
2478
2478
2478
2478
2478
2478
2229
2229
2229
2229
2229
% Accepted
100
100
100
100
99
99
100
100
100
99
47
47
100
100
100
99
62
62
100
100
100
98*
71
71
100
100
100
96*
50
50
100
100
100
94*
51
Philadelphia (KIAD)
n
2152
2152
2152
2152
2152
2152
4320
4320
4320
4320
4320
4320
3702
3702
3702
3702
3702
3702
4204
4204
4204
4204
4204
4204
3354
3354
3354
3354
3354
3354
3246
3246
3246
3246
3246
% Accepted
100
100
100
100
100
85*
100
100
100
99
63
62
100
100
100
99*
73
73
100
100
100
97*
71
71
100
100
100
95*
50
50
100
100
100
93*
50
Phoenix (KTWC)
n
2143
2143
2143
2143
2143
2143
3611
3611
3611
3611
3611
3611
2797
2797
2797
2797
2797
2797
1473
1473
1473
1473
1473
1473
1889
1889
1889
1889
1889
1889
3453
3453
3453
3453
3453
% Accepted
99*
99*
87*
99*
100
100
100
100
97*
100
63
62
100
100
100
99*
88
88
100
100
100
99*
54
54
100
100
100
95*
54
54
100
100
100
94*
82
64
-------
Height
Level
2500-
3000m
3000-
3500m
3500-
4000m
>4000
m
Variable
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Pressure
Height
Temperature
DewPointTemperature
WindDirection
WindSpeed
Atlanta (KFFC)
n
3171
4318
4318
4318
4318
4318
4318
2840
2840
2840
2840
2840
2840
2964
2964
2964
2964
2964
2964
7895
7895
7895
7895
7895
7895
% Accepted
52
100
100
100
94*
74
74
100
100
100
92*
49
49
100
100
100
90*
49
49
87*
73*
100
81 *
53
53
Detroit (KDTX)
n
3078
4257
4257
4257
4257
4257
4257
2932
2932
2932
2932
2932
2932
2775
2775
2775
2775
2775
2775
7279
7279
7279
7279
7279
7279
% Accepted
50
100
100
100
90*
71
71
100
100
99
88*
48
48
100
100
99
84*
49
49
77*
70*
98*
74*
59
59
Los Angeles (KNKX)
n
2229
2769
2769
2769
2769
2769
2769
2754
2754
2754
2754
2754
2754
2014
2014
2014
2014
2014
2014
6136
6136
6136
6136
6136
6136
% Accepted
51
100
100
100
90*
73
73
100
100
100
91 *
69
69
100
100
100
86*
53
53
82*
64*
99*
76*
59
59
Philadelphia (KIAD)
n
3246
3736
3736
3736
3736
3736
3736
3614
3614
3614
3614
3614
3614
2830
2830
2830
2830
2830
2830
7619
7619
7619
7619
7619
7619
% Accepted
50
100
100
100
90*
64
64
100
100
100
90*
65
65
100
100
100
87*
50
50
88*
71 *
99*
79*
55
55
Phoenix (KTWC)
n
3453
2213
2213
2213
2213
2213
2213
2344
2344
2344
2344
2344
2344
2423
2423
2423
2423
2423
2423
7483
7483
7483
7483
7483
7483
% Accepted
82
100
100
100
90*
55
55
100
100
100
88*
54
54
100
100
100
85*
55
55
58*
71 *
99*
69*
65
65
Notes:
a Percentages are rounded down to the nearest integer. All data obtained from the NCDC ISH database.
* The majority of unaccepted records are due to values being out of range
Shading:
<95 of observations were accepted.
<75 of observations were accepted.
<50 of observations were accepted.
65
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
3.2.2 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 ratio12, surface albedo13 as a function of the solar angle, and surface
roughness14.
The January 2008 version of AERSURFACE was used to estimate land-use patterns and
calculate the Bowen ratio, surface albedo, and surface roughness as part of the AERMET
processing. AERSURFACE uses the US Geological Survey (USGS) National Land Cover Data
1992 archives (NLCD92)15. Three to four land-use sectors were manually identified around the
surface meteorological stations using this land-use data. These land-use sectors are used to
identify the Bowen ratio and surface albedo, which are assumed to represent an area around the
station of radius 10 km, and to calculate surface roughness by wind direction.
A monthly temporal resolution was used for the Bowen ratio, albedo, and surface roughness
for all five meteorological sites. Because the five sites were located at airports, a lower surface
roughness was calculated for the 'Commercial/Industrial/Transportation' land-use type to reflect
the dominance of transportation land cover rather than commercial buildings. Los Angeles and
Phoenix are arid regions, which increases the calculated albedo and Bowen ratio values and
decreases the surface roughness values assigned to the 'Shrubland' and 'Bare Rock/Sand/Clay'
land-use types to reflect a more desert-like area. Philadelphia and Detroit each have at least one
winter month of continuous snow cover, which tends to increase albedo, decrease Bowen ratio,
and decrease surface roughness for most land-use types during the winter months compared to
snow-free areas.
Seasons were assigned for each site based on 1971-2000 NCDC 30-year climatic normals
and on input from the respective state climatologists. Table 29 provides the seasonal definitions
for each city.
Table 29. Seasonal specifications by study location.
Location
Atlanta
Detroit
Winter
(continuous
snow)
Dec, Jan, Feb,
Mar
Winter
(no snow)
Dec, Jan, Feb
Spring
Mar, Apr, May
Apr, May
Summer
Jun, Jul, Aug
Jun, Jul, Aug
Fall
Sep, Oct, Nov
Sep, Oct, Nov
For any moist surface, the Bowen Ratio is the ratio of heat energy used for sensible heating (conduction and
convection) to the heat energy used for latent heating (evaporation of water or sublimation of snow). The Bowen
ratio ranges from about 0.1 for the ocean surface to more than 2.0 for deserts. Bowen ratio values tend to decrease
with increasing surface moisture for most land-use types.
13 The ratio of the amount of electromagnetic radiation reflected by the earth's surface to the amount incident upon
it. Value varies with surface composition. For example, snow and ice vary from 80% to 85% and bare ground from
10% to 20%.
14 The presence of buildings, trees, and other irregular land topography that is associated with its efficiency as a
momentum sink for turbulent air flow, due to the generation of drag forces and increased vertical wind shear.
15 http://seamless.usgs.gov/
66
-------
Los Angeles
Philadelphia
Phoenix
Dec, Jan, Feb
Apr, May, Jun
Mar, Apr, May
Apr, May, Jun
Jul, Aug, Sep
Jun, Jul, Aug
Jul, Aug, Sep
Oct, Nov, Dec,
Jan, Feb, Mar
Sep, Oct, Nov
Oct, Nov, Dec,
Jan, Feb, Mar
Season definitions provided by the AERSURFACE manual as follows:
Winter (continuous snow): Winter with continuous snow on ground
Winter (no snow): Late autumn after frost and harvest, or winter with no snow
Spring: Transitional spring with partial green coverage or short annuals
Summer: Midsummer with lush vegetation
Fall: Autumn with unharvested cropland
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2 Further discussion of the land use and surface analysis, as well as a discussion of the
3 difference in results from employing the new AERSURFACE tool is given in Appendix E.
4 3.2.3 Meteorological Analysis
The AERMET application location and elevation were taken as the center of each modeled
city, estimated using Google Earth version 4.2.0198.2451 (beta). They are as follows:
• Atlanta: 33.755 °N, 84.391 °W, 306 m
• Detroit: 42.332 °N, 83.048 °W, 181 m
• Los Angeles: 34.053 °N, 118.245 °W, 91 m
• Philadelphia: 39.952 °N, 75.164 °W, 12 m
• Phoenix: 33.448 °N, 112.076 °W, 330 m
For each site in this study, the 2001-2003 AERSURFACE processing was run three times -
once assuming the entire period was drier than normal, once assuming the entire period was
wetter than normal, and once assuming the entire period was of average precipitation
accumulation. These precipitation assumptions influence the Bowen ratio, discussed above.
To create meteorological input records that best represent the given city for each of the three
years, the resulting surface output files for each site were then pieced together on a month-by-
month basis, with selection based on the relative amount of precipitation in each month. Any
month where the actual precipitation amount received was at least twice the 1971-2000 NCDC
30-year climatic normal monthly precipitation amount was considered wetter than normal, while
any month that received less than half the normal amount of precipitation amount was considered
drier than normal; all other months were considered to have average surface moisture conditions.
Table 30 indicates the surface moisture condition for each month-location combination for
this study. The final meteorological record includes wet conditions for the Bowen ratio for the
month-location combinations shown in green and dry conditions for those in orange. All other
region-month combinations used an average Bowen ratio.
67
-------
Table 30. Monthly precipitation compared to NCDC 30-year climatic normal, 2001-2003.
City
Atlanta
Detroit
Los Angeles
Philadelphia
Phoenix
City
Atlanta
Detroit
Los Angeles
Philadelphia
Phoenix
City
Atlanta
Detroit
Los Angeles
Philadelphia
Phoenix
2001
Jan
55.9%
27.4%
157.7%
74.8%
214.4%
Feb
77.6%
125.6%
237.0%
103.6%
1 1 1 .5%
Mar
169.3%
29.8%
52.7%
144.2%
71 .7%
Apr
91.7%
89.2%
175.0%
43.9%
429.9%
May
77.1%
106.0%
4.9%
102.9%
14.8%
Jun
185.9%
61 .6%
0.0%
180.1%
13.1%
Jul
50.0%
39.4%
0.0%
29.9%
68.8%
Aug
28.3%
82.0%
0.0%
26.0%
48.6%
Sep
53.9%
108.2%
0.0%
67.1%
0.0%
Oct
25.8%
280.7%
10.9%
30.6%
3.0%
Nov
21 .4%
75.8%
119.9%
17.9%
28.0%
Dec
58.7%
66.0%
60.0%
64.6%
95.9%
2002
Jan
105.5%
132.7%
76.1%
69.9%
6.2%
Feb
54.7%
73.9%
17.7%
17.7%
0.0%
Mar
102.2%
66.6%
28.4%
96.4%
7.7%
Apr
50.5%
123.9%
27.5%
52.7%
25.2%
May
86.0%
104.2%
42.7%
89.2%
0.0%
Jun
77.2%
25.4%
4276.6%
93.9%
0.0%
Jul
50.8%
133.7%
15656.2%
51 .0%
119.7%
Aug
21 .6%
36.8%
1358.3%
59.0%
0.0%
Sep
157.5%
60.1%
829.8%
89.1%
66.7%
Oct
191.7%
50.7%
1730.1%
202.7%
46.8%
Nov
131.4%
83.5%
277.0%
94.2%
57.7%
Dec
137.3%
47.8%
216.6%
117.9%
18.8%
2003
Jan
40.2%
13.4%
0.0%
53.2%
68.8%
Feb
75.4%
34.1%
121.7%
165.0%
413.1%
Mar
132.1%
59.8%
69.4%
102.7%
48.2%
Apr
95.6%
64.9%
78.7%
62.0%
69.3%
May
252.5%
155.0%
397.0%
108.5%
0.0%
Jun
198.5%
59.7%
0.0%
246.2%
0.0%
Jul
100.2%
28.2%
0.0%
46.5%
61 .6%
Aug
95.2%
109.0%
0.0%
86.1%
45.2%
Sep
59.5%
124.7%
0.0%
120.8%
36.2%
Oct
47.9%
99.7%
1 1 1 .5%
162.8%
26.9%
Nov
102.3%
99.2%
71.1%
92.9%
90.6%
Dec
70.9%
86.9%
64.9%
158.6%
20.5%
Shading:
At least twice the normal precipitation level
Less than twice the normal precipitation level and greater than half the normal amount
Less than or equal to half the normal monthly precipitation amount
68
-------
1 3.2.4 On-Road Emissions Preparation
2 3.2.4.1 Philadelphia County Data Sources
3 Information on traffic data in the Philadelphia area was obtained from the Delaware Valley
4 Regional Planning Council (DVRPC16) via their most recent, baseline travel demand modeling
5 (TDM) simulation - that is, the most recent simulation calibrated to match observed traffic data.
6 DVRPC provided the following files.
7
8 • Shapefiles of TDM outputs for the 2002 baseline year for all links in their network.
9 • Input files for the MOBILE6.2 emissions model that characterize local inputs that differ
10 from national defaults, including fleet registration distribution information.
11 • Postprocessing codes they employ for analysis of TDM outputs into emission inventory
12 data, to ensure as much consistency as possible between the methodology used for this
13 study and that of DVRPC. These include DVRPC's versions of the local SVMT.DEF,
14 HVMT.DEF, and FVMT.DEF MOBILE6.2 input files describing the vehicle miles
15 traveled (VMT) by speed, hour, and facility, respectively, by county in the Delaware
16 Valley area.
17 • A lookup table used to translate average annual daily traffic (AADT) generated by the
18 TDM into hourly values.
19
20 Although considerable effort was expended to maintain consistency between the DVRPC
21 approach to analysis of TDM data and that employed in this analysis, including several personal
22 communications with agency staff on data interpretation, complete consistency was not possible
23 due to the differing analysis objectives. The DVRPC creates countywide emission inventories.
24 This study created spatially and temporally resolved emission strengths for dispersion modeling.
25
26 Emission Sources and Locations
27 The TDM simulation's shapefile outputs include annual average daily traffic (AADT)
28 volumes and a description of the loaded highway network. The description of the network
29 consists of a series of nodes joining individual model links (i.e., roadway segments) to which the
30 traffic volumes are assigned, and the characteristics of those links, such as endpoint location,
31 number of lanes, link distance, and TDM-defined link daily capacity.17
32
33 To reduce the scope of the analysis, the full set of links in the DVRPC network was first
34 filtered to include only those roadway types considered major (i.e., freeway, parkway, major
35 arterial, ramp), and that had AADT values greater than 15,000 vehicles per day (one direction).
36
37 However, the locations of links in the model do not necessarily agree well with the roads
38 they are attempting to represent. While the exact locations of the links may not be mandatory for
39 DVRPC's travel demand modeling, the impacts of on-road emissions on fixed receptors is
40 crucially linked to the distance between the roadways and receptors. Hence, it was necessary to
41 modify the link locations from the TDM to the best known locations of the actual roadways.
16 http://www.dvrpc.org/
17 The TDM capacity specifications are not the same as those defined by the Highway Capacity Manual (HCM).
Following consultation with DVRPC, the HCM definition of capacity was used in later calculations discussed
below.
69
-------
1 The correction of link locations was done based on the locations of the nodes that define the
2 end points of links with a GIS analysis, as follows.
O
4 A procedure was developed to relocate TDM nodes to more realistic locations. The
5 nodes in the TDM represent the endpoints of links in the transportation planning network and are
6 specified in model coordinates. The model coordinate system is a Transverse Mercator
7 projection of the TranPlan Coordinate System with a false easting of 31068.5, false northing of-
8 200000.0, central meridian: -75.00000000, origin latitude of 0.0, scale factor of 99.96, and in
9 units of miles. The procedure moved the node locations to the true road locations and translated
10 to dispersion model coordinates. The Pennsylvania Department of Transportation (PA DOT)
11 road network database18 was used as the specification of the true road locations. The nodes were
12 moved to coincide with the nearest major road of the corresponding roadway type using a built-
13 in function of ArcGIS. Once the nodes had been placed in the corrected locations, a line was
14 drawn connecting each node pair to represent a link of the adjusted planning network.
15
16 To determine hourly traffic on each link, the AADT volumes were converted to hourly
17 values by applying DVRPC's seasonal and hourly scaling factors. To determine hourly traffic
18 on each link, the AADT volumes were converted to hourly values by applying DVRPC's
19 seasonal and hourly scaling factors. The heavy-duty vehicle fraction - which is assumed by
20 DVRPC to be about 6% in all locations and times - was also applied.19 Another important
21 variable, the number of traffic signals occurring on a given link, was taken from the TDM link-
22 description information.
23
24 Several of these parameters are shown in the following set of tables.
25
26 • Table 31: hourly scaling factors
27 • Table 32: seasonal scaling factors
28 • Table 33: number of signals per roadway mile
29 • Table 34: statistical summaries of AADT volumes for links included in the study.
18 http://www.pasda.psu.edu/
19 As shown by Figure 13, NOX emissions from HDVs tend to be higher than their LDV counterparts by about a
factor of 10. However, the HDV fraction is less than 10% of the total VMT in most circumstances, mitigating their
influence on composite emission factors, although this mitigating effect is less pronounced at some times than
others. For example, nighttimes on freeways tend to show a smaller reduction in HDV volume than in total volume,
and thus an increased HDV fraction. This effect is not captured in most TDMs or emission postprocessors and -
both to maintain consistency with the local MPO's vehicle characterizations and emissions modeling and due to lack
of other relevant data - was also not included here. The net result of this is likely to be slightly underestimated
emissions from major freeways during late-night times.
70
-------
Table 31. Hourly scaling factors (in percents) applied to Philadelphia County AADT volumes.
Road
Type
Freeway
Arterial
Local
Ramp
Road
Type
Freeway
Arterial
Local
Ramp
Region
CBD
Fringe
Urban
Suburban
Rural
CBD
Fringe
Urban
Suburban
Rural
CBD
Fringe
Urban
Suburban
Rural
CBD
Fringe
Urban
Suburban
Rural
Region
CBD
Fringe
Urban
Suburban
Rural
CBD
Fringe
Urban
Suburban
Rural
CBD
Fringe
Urban
Suburban
Rural
CBD
Fringe
Urban
Suburban
Rural
0:00
1.23
1.23
1.23
0.96
0.71
1.43
1.53
1.13
0.70
0.60
1.11
1.00
1.19
0.53
0.55
1.23
1.23
1.23
0.96
0.71
12:00
4.97
4.97
4.97
5.05
4.92
5.27
5.52
5.42
5.75
5.55
6.26
6.31
5.25
5.78
5.20
4.97
4.97
4.97
5.05
4.92
1:00
0.86
0.86
0.86
0.64
0.48
0.96
0.97
0.68
0.40
0.36
0.71
0.55
0.74
0.29
0.32
0.86
0.86
0.86
0.64
0.48
13:00
5.77
5.77
5.77
5.19
5.01
5.57
5.40
5.54
5.71
5.50
6.74
5.64
5.40
5.57
5.11
5.77
5.77
5.77
5.19
5.01
2:00
0.74
0.74
0.74
0.54
0.38
0.61
0.62
0.52
0.32
0.34
0.45
0.37
0.53
0.21
0.25
0.74
0.74
0.74
0.54
0.38
14:00
6.40
6.40
6.40
5.90
5.75
5.95
6.08
6.16
6.12
6.00
6.88
6.64
6.44
6.01
5.89
6.40
6.40
6.40
5.90
5.75
3:00
0.84
0.84
0.84
0.61
0.48
0.50
0.47
0.45
0.33
0.41
0.37
0.21
0.43
0.20
0.30
0.84
0.84
0.84
0.61
0.48
15:00
6.60
6.60
6.60
6.80
7.12
6.63
6.88
7.04
7.05
7.11
6.78
7.32
7.35
7.11
7.41
6.60
6.60
6.60
6.80
7.12
4:00
1.23
1.23
1.23
0.90
0.95
0.58
0.54
0.63
0.55
0.77
0.41
0.39
0.54
0.37
0.57
1.23
1.23
1.23
0.90
0.95
16:00
7.02
7.02
7.02
7.58
7.88
7.39
7.36
7.39
7.66
7.82
7.64
7.85
7.80
8.20
8.53
7.02
7.02
7.02
7.58
7.88
5:00
2.50
2.50
2.50
2.16
2.54
1.17
1.10
1.68
1.71
2.29
0.97
0.98
1.32
1.25
1.89
2.50
2.50
2.50
2.16
2.54
17:00
6.76
6.76
6.76
7.67
8.18
7.81
8.08
7.42
7.98
7.98
8.10
9.52
7.85
8.98
8.93
6.76
6.76
6.76
7.67
8.18
6:00
4.87
4.87
4.87
5.39
6.05
2.89
2.99
4.26
4.51
5.47
2.39
1.98
3.37
3.94
5.26
4.87
4.87
4.87
5.39
6.05
18:00
6.27
6.27
6.27
6.51
6.27
6.36
6.24
6.08
6.42
6.26
6.57
6.25
6.41
6.83
6.75
6.27
6.27
6.27
6.51
6.27
7:00
6.52
6.52
6.52
7.33
7.77
5.50
5.77
6.68
7.04
7.37
4.82
5.31
6.54
7.51
7.93
6.52
6.52
6.52
7.33
7.77
19:00
4.20
4.20
4.20
4.27
4.31
4.78
4.98
4.74
4.81
4.48
4.96
5.50
5.02
5.02
4.82
4.20
4.20
4.20
4.27
4.31
8:00
6.47
6.47
6.47
6.85
6.79
6.87
6.53
6.86
6.84
6.62
6.72
5.91
6.86
7.50
6.84
6.47
6.47
6.47
6.85
6.79
20:00
3.52
3.52
3.52
3.34
3.45
4.05
4.21
3.77
3.83
3.50
3.96
5.29
4.04
3.83
3.64
3.52
3.52
3.52
3.34
3.45
9:00
5.75
5.75
5.75
5.52
5.22
5.87
5.60
5.47
5.37
5.36
6.50
5.78
5.09
5.24
4.94
5.75
5.75
5.75
5.52
5.22
21:00
3.06
3.06
3.06
2.97
2.97
3.74
3.82
3.31
3.13
2.80
3.02
2.87
3.46
2.90
2.70
3.06
3.06
3.06
2.97
2.97
10:00
4.99
4.99
4.99
4.90
4.64
5.37
5.14
5.09
4.95
5.09
4.60
5.14
4.65
4.66
4.57
4.99
4.99
4.99
4.90
4.64
22:00
2.50
2.50
2.50
2.32
2.10
3.18
3.13
2.61
2.15
1.88
2.88
2.46
2.79
1.82
1.73
2.50
2.50
2.50
2.32
2.10
11:00
5.02
5.02
5.02
4.94
4.78
5.17
4.86
5.17
5.36
5.35
4.93
5.19
4.95
5.22
4.89
5.02
5.02
5.02
4.94
4.78
23:00
1.92
1.92
1.92
1.66
1.27
2.36
2.19
1.93
1.34
1.11
2.25
1.56
2.01
1.05
0.99
1.92
1.92
1.92
1.66
1.27
2
3
71
-------
1 Table 32. Seasonal scaling factors applied to Philadelphia County AADT volumes.
2
3
Season
Winter
Spring
Summer
Autumn
Winter
Spring
Summer
Autumn
Winter
Spring
Summer
Autumn
Winter
Spring
Summer
Autumn
Road
Type
Freeway
Freeway
Freeway
Freeway
Arterial
Arterial
Arterial
Arterial
Local
Local
Local
Local
Ramp
Ramp
Ramp
Ramp
Factor
0.945
1.006
1.041
1.009
0.942
1.004
1.041
1.013
0.933
1.012
1.05
1.004
0.944
1.005
1.041
1.011
Table 33. Signals per mile, by link type, applied to Philadelphia County AADT volumes.
Functional Class
Freeway
Local
Major Arterial
Minor Arterial
Parkway
Ramp
Region Type
CBD
0
8
8
8
4
0
Fringe
0
6
6
6
2
0
Rural
0
1.5
1
1.3
0.5
0
Suburban
0
3
2
2
1
0
Urban
0
5
4
4
1.5
0
4
5
6
7
8
9
Table 34. Statistical summary of AADT volumes (one direction) for Philadelphia County AERMOD
simulations.
Statistic
Count
Minimum
AADT
Maximum
AADT
Average
AADT
Road Type
Arterial
Freeway
Ramp
Arterial
Freeway
Ramp
Arterial
Freeway
Ramp
Arterial
Freeway
Ramp
CBD
186
11
0
15088
15100
44986
39025
21063
25897
Fringe
58
10
4
15282
18259
16796
44020
56013
40538
21196
40168
24468
Suburban
210
107
3
15010
15102
15679
48401
68661
24743
20736
33979
18814
Urban
580
98
1
15003
15100
16337
44749
68661
16337
22368
31294
16337
72
-------
1 Emission Source Strength
2 On-road mobile emission factors were derived from the MOBILE6.2 emissions model as
3 follows. The DVRPC-provided external data files describing the vehicle miles traveled (VMT)
4 distribution by speed, functional class, and hour, as well as the registration distribution and Post-
's 1994 Light Duty Gasoline Implementation for Philadelphia County were all used in the model
6 runs without modification. To further maintain consistency with the recent DVRPC inventory
7 simulations and maximize temporal resolution, the DVRPC's seasonal particulate matter (PM)
8 MOBILE6 input control files were also used. These files include county-specific data describing
9 the vehicle emissions inspection and maintenance (I/M) programs, on-board diagnostics (OBD)
10 start dates, VMT mix, vehicle age distributions, default diesel fractions, and representative
11 minimum and maximum temperatures, humidity, and fuel parameters. The simulations are
12 designed to calculate average running NOX emission factors.20
13
14 These input files were modified for the current project to produce running NOx emissions in
15 grams per mile for a specific functional class (Freeway, Arterial, or Ramp) and speed. Iterative
16 MOBILE6.2 simulations were conducted to create tables of average Philadelphia County
17 emission factors resolved by speed (2.5 to 65 mph), functional class, season, and year (2001,
18 2002, or 2003) for each of the eight combined MOBILE vehicle classes (LDGV, LDGT12,
19 LDGT34, HDGV, LDDV, LDDT, HDDV, and MC)21. The resulting tables were then
20 consolidated into speed, functional class, and seasonal values for combined light- and heavy-duty
21 vehicles. Figure 13 shows an example of the calculated emission factors for Autumn, 2001.
22
23
20 Basing the present emissions model input files on MPO-provided PM, rather than NOX input files should not cause
confusion. MPO-provided PM files were used because they contain quarterly rather than annual or biannual
information. In all cases the output species were modified to produce gaseous emissions. Further, many of the
specified input parameters do not affect PM emissions, but were included by the local MPO to best represent local
conditions, which were preserved in the present calculations of NOX emissions. This usage is consistent with the
overall approach of preserving local information wherever possible.
21 HDDV - Heavy-Duty Diesel Vehicle, HDGV - Heavy-Duty Gasoline Vehicle, LDDT - Light-Duty Diesel Truck,
LDDV - Light-Duty Diesel Vehicle, LDGT12 - Light-Duty Gasoline Truck with gross vehicle weight rating < 6,000
Ibs and a loaded vehicle weight of < 5,750 Ibs, LDGT 34 - Light-Duty Gasoline Truck with gross vehicle weight
rating between 6,001 - 8,500 and a loaded vehicle weight of < 5,750 Ibs, LDGV - Light-Duty Gasoline Vehicle, MC
- Motorcycles.
73
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
E
3
in
c
o
'in
v>
E
LLI
X
O
o>
O)
20
18
16
14
12
10
8
6
Fall Arterial LDV
- - - - Fall Freeway LDV
Fall Arterial HDV
Fall Freeway HDV
10
20
50
60
70
30 40
Average Speed (mph)
Figure 13. Example of Light- and heavy-duty vehicle NOX emissions grams/mile (g/mi) for arterial and
freeway functional classes, 2001.
To determine the emission strengths for each link for each hour of the year, the Philadelphia
County 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 speed calculations
were made as follows.
The spatial-mean speed of each link at each time was calculated following the methodology
of the Highway Capacity Manual.22 Generally, the spatial-mean speed calculation is a function of
the time-resolved volume-to-capacity ratio, with capacity the limiting factor. In the case of
freeway calculations, this is determined by the HDV fraction, posted speed, and the general
hilliness of the terrain, which was assumed to be uniformly flat for this region. The case of
arterials without intersections is similar, but also considers urban effects. The case of arterials
with intersections further considers the number of signals and length of each link and
signalization parameters. It was assumed that all signals are identical, operating with a 120-
second cycle and a protected left turn phase. Each link's speed is calculated independently. For
example, a series of adjacent arterial links could show very different spatial-mean speeds if one
link contains one or more intersections. That is, no up- or down-stream impacts are considered
on individual link speeds. Speeds were assumed to be equal for light- and heavy-duty vehicles.
Table 35 shows the resulting average speed for each functional class within each TDM
region. Several values are shown as N/A, due to the focus only on major links as discussed
above.
22 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.
74
-------
1
2
Table 35. Average calculated speed by link type.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Ramp
Arterial
Freeway
Average Speed (mph)
CBD
N/A
34
51
Fringe
35
31
62
Suburban
35
44
66
Urban
35
32
62
Rural
N/A
N/A
N/A
The resulting emission factors were then coupled with the TDM-based activity estimates to
calculate emissions from each of the 1,268 major roadway links. However, many of the links
were two sides of the same roadway segment. To speed model execution time, those links that
could be combined into a single emission source were merged together. This was done only for
the 628 links (314 pairs) where opposing links were paired in space and exhibited similar activity
levels within 20% of each other.
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 27 links that exceeded this ratio and were converted to 55 segmented
sources. Thus, the total number of area sources included in the dispersion simulations is 982.
Table 36 shows the distribution of on-road area source sizes. 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 36. On-road area source sizes.
Minimum
Median
Average
1-a Deviation
Maximum
Segment
Width (m)
3.7
11.0
13.7
7.7
43.9
Lanes
1.0
3.0
3.8
2.1
12.0
Segment
Length (m)
0.0
220.6
300.2
259.5
1340.2
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
postprocessed 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.
The release height of each source was determined as the average of the light- and heavy-duty
vehicle fractions, with an assumed light- and heavy-duty emission release heights of 1.0 ft
75
-------
1 (0.3048 m) and 13.1 ft (4.0 m), respectively.23 Because AERMOD only accepts a single release
2 height for each source, the 24-hour average of the composite release heights is used in the
3 modeling. Since surface-based mobile emissions are anticipated to be terrain following, no
4 elevated or complex terrain was included in the modeling. That is, all sources are assumed to lie
5 in a flat plane.
6
7 3.2.5 Stationary Sources Emissions Preparation
8 3.2.5.1 Philadelphia Data Sources
9 Data for the parameterization of major point sources in Philadelphia comes primarily from
10 two sources: the 2002 National Emissions Inventory (NEI; US EPA, 2007b) and Clean Air
11 Markets Division (CAMD) Unit Level Emissions Database (US EPA, 2007c). These two
12 databases have complimentary information.
13
14 The NEI database contains stack locations, emissions release parameters (i.e., height,
15 diameter, exit temperature, exit velocity), and annual emissions for 707 NOx-emitting stacks
16 (206 of which are considered fugitive release points) in Philadelphia County. The CAMD
17 database, on the other hand, has information on hourly NOX emission rates for all the units in the
18 US, where the units are the boilers or equivalent, each of which can have multiple stacks. The
19 alignment of facilities between the two databases is not exact, however. Some facilities listed in
20 the NEI, are not included in the CAMD database. Of those facilities that do match, in many cases
21 there is no clear pairing between the individual stacks assigned within the databases.
22
23 3.2.5.2 Data Source Alignment
24 To align the information between the two databases and extract the useful portion of each for
25 dispersion modeling, the following methodology was used.
26
27 1. Attention was limited stacks within the NEI data base that (a) lie within Philadelphia
28 County and (b) were part of a facility with total emissions from all stacks exceeding
29 lOOtpyNOx.
30 2. Individual stacks that had identical stack physical parameters and were co-located
31 within about 10m were combined to be simulated as a single stack with their
32 emissions summed.
33 3. All fugitive releases were removed from the list, to be analyzed as a separate source
34 group.
35
36 The resulting 19 distinct, combined stacks from the NEI are shown in Table 37.
37
38 The CAMD database was then queried for facilities that matched the facilities identified from
39 the NEI database. Facility matching was done on the facility name, Office of Regulatory
40 Information Systems (ORIS) identification code (when provided) and facility total emissions to
41 ensure a best match between the facilities. Once facilities were paired, individual units and
23 4.0 m includes plume rise from truck exhaust stacks. See Diesel Paniculate Matter Exposure Assessment Study
for the Ports of Los Angeles and Long Beach. State of California Air Resources Board, Final Report, April 2006.
76
-------
1 stacks in the data bases were paired, based on annual emission totals. Table 38 shows the
2 matching scheme for the seven major facilities in Philadelphia County.
24
4 In Table 38, there are sometimes multiple CAMD units that pair with a single NEI combined
5 stack. In these cases the hourly emission rates from the matching CAMD units are summed for
6 each hour. For example, in the case of stack 859 for "Sunoco, Inc - Philadelphia" five CAMD
7 hourly records are summed into a single hourly record. Then each resulting hourly value is
8 scaled by a factor of 1032.8 / 938.9 = 1.10, so that the annual total matches the NEI annual total.
9
10 Similarly, there are sometimes multiple combined stacks that pair with single units. In this
1 1 case the CAMD values are disaggregated according to NEI-defined stack contributions. For
12 example, "Sunoco, Inc - Philadelphia" stack 855's profile is determined by taking the hourly
13 profile from CAMD unit number 52106-150101, and scaling each value by a factor of 26.2 tpy /
14 48.2 tpy total = 0.54. Then each resulting hourly value is scaled by a factor of 48.2/162. 1 = 0.3
15 so that the sum of the annual totals for the 4 stacks corresponding to unit number 52106-150101
16 matches the NEI total. For consistency, in each case the 2001 and 2003 hourly emission profiles
17 were determined using the same scaling factors, but applied to the respective CAMD emission
18 profile.
19
20 It is clear from Table 38 that most facilities agree well in total annual NOX emissions between
21 the two databases. However, in the case of the "Sunoco Chemicals (Former Allied Signal)"
22 facility, nearly half of the NEI emissions (without fugitives) do not appear in the CAMD
23 database. The reason for this is unknown and no information was readily available on the
24 relative accuracy of the two databases.
25
26 Figure 14 illustrates the discrepancy versus fraction of hours with positive emissions,
27 according to the CAMD data base. The figure suggests that the discrepancies are not primarily
28 the result of facilities with episodic emissions (i.e., "peak load" facilities). Although there is
29 good agreement on facility-wide emissions between the two data bases, there are larger
30 discrepancies between CAMD unit emissions and NEI stack emissions. This is to be expected
3 1 given the discrepancy in resolution between the two data bases.
32
24 Note that Jefferson Smurfit does not exist in the CAMD database. The matching here was based on facility types
as follows. Smurfit in PA was taken as a packaging/recycling facility, and the stack assumed to be a Cogen facility,
based on information in the NEEDS database (http://www.epa.gov/interstateairquality/pdfs/NEEDS-NODA.xls).
The best matched cogen plant in Philadelphia County in both the NEEDS and CAMD database is the Gray's Ferry
Cogen Partnership (ORIS 54785), which was a reasonable match for Smurfit's total emissions. It was assumed that
the hourly emission profile also matches well.
77
-------
Table 37. Combined stacks parameters for stationary NOX emission sources in Philadelphia County.
Stack
No
817
818
819
820
821
855
856
857
858
859
860
861
NEI
Site ID
NEIPA2218
NEIPA2218
NEI40720
NEI40720
NEI40720
NEI40723
NEI40723
NEI40723
NEI40723
NEI40723
NEI7330
NEI7330
Facility Name
EXELON
GENERATION CO -
DELAWARE
STATION
EXELON
GENERATION CO -
DELAWARE
STATION
JEFFERSON
SMURFIT
CORPORATION
(US)
JEFFERSON
SMURFIT
CORPORATION
(US)
JEFFERSON
SMURFIT
CORPORATION
(US)
Sunoco Inc. -
Philadelphia
Sunoco Inc. -
Philadelphia
Sunoco Inc. -
Philadelphia
Sunoco Inc. -
Philadelphia
Sunoco Inc. -
Philadelphia
SUNOCO
CHEMICALS
(FORMER ALLIED
SIGNAL)
SUNOCO
CHEMICALS
SIC
Code
4911
4911
2631
2631
2631
2911
2911
2911
2911
2911
2869
2869
NAICS
Code
221112
221112
32213
32213
32213
32411
32411
32411
32411
32411
325998
325998
ORIS
Facil
ity
Code
3160
3160
Stack
Emiss
(tpy)
4.82
287.8
0.148
113.8
114.46
26.2
1.3
1.4
19.3
1032.8
0.033
49.1
Stack X
(deg)
-75.1358
-75.1358
-75.2391
-75.2391
-75.2391
-75.2027
-75.2003
-75.203
-75.2027
-75.2124
-75.0715
-75.0715
Stack Y
(deg)
39.96769
39.96769
40.03329
40.03329
40.03329
39.92535
39.91379
39.92539
39.92535
39.90239
40.00649
40.00649
Stack
Ht(m)
49
64
16
53
53
24
24
25
25
61
5
41
Exit
Tern
P(K)
515
386
477
427
477
450
644
511
527
489
476
422
Stack
Diam
(m)
4.2
3.7
0.4
2.4
2.4
2.1
1.5
1.9
1.9
5.8
0.5
1.4
Exit
Vel
(m/s)
0
17
19
10
12
9
22
10
11
11
7
22
Facility
Emiss
Incl
Fugitive
(tpy)
297.8
297.8
228.4
228.4
228.4
3112.2
3112.2
3112.2
3112.2
3112.2
160.9
160.9
78
-------
Stack
No
862
863
864
865
866
867
868
NEI
Site ID
NEI7330
NEI7330
NEIPA10135
3
NEIPA10135
3
NEIPA10135
6
NEIPA10135
6
NEIPA2222
Facility Name
(FORMER ALLIED
SIGNAL)
SUNOCO
CHEMICALS
(FORMER ALLIED
SIGNAL)
SUNOCO
CHEMICALS
(FORMER ALLIED
SIGNAL)
TRIGEN-
SCHUYLKILL
TRIGEN-
SCHUYLKILL
GRAYS FERRY
COGENERATION
PARTNERS
GRAYS FERRY
COGENERATION
PARTNERS
TRIGEN- EDISON
SIC
Code
2869
2869
4961
4961
4911
4911
4961
NAICS
Code
325998
325998
22
22
22
22
62
ORIS
Facil
ity
Code
54785
54785
Stack
Emiss
(tpy)
34.6
77.2
128.6
61.5
143.2
90.3
130.5
Stack X
(deg)
-75.0715
-75.0715
-75.1873
-75.1873
-75.1873
-75.1873
-75.1569
Stack Y
(deg)
40.00649
40.00649
39.94239
39.94239
39.94239
39.94239
39.94604
Stack
Ht(m)
42
42
69
78
78
85
78
Exit
Tern
P(K)
422
422
450
450
396
443
589
Stack
Diam
(m)
1.6
1.6
4.9
7.3
5.5
3.2
3.7
Exit
Vel
(m/s)
17
22
6
2
20
21
9
Facility
Emiss
Incl
Fugitive
(tpy)
160.9
160.9
190.1
190.1
233.5
233.5
130.5
79
-------
Table 38. Matched stacks between the CAMD and NEI database.
NEI Facility
Name
Exelon
Generation Co
- Delaware
Station
Sunoco Inc. -
Philadelphia
Sunoco
Chemicals
(Former Allied
Signal)
Trigen -
Schuylkill
NEI
Comb.
Stack
Number
817
818
855
856
857
858
859
860
861
862
863
864
NEI
Comb.
Stack
Emiss
(tpy)
4.8
287.8
26.2
1.3
1.4
19.3
1032.8
0.0
49.1
34.6
77.2
128.6
NEI
Unit
Emiss
(tpy)
4.8
287.8
48.2
1032.8
160.9
128.6
NEI
Facility
Emiss
(tpy,
w/out
Fugitive)
292.6
1081.0
160.9
190.1
CAMD
Facility
Name
Delaware
Philadelphia
Refinery
Sunoco
Chemicals
Frankford
Plant
Trigen
Energy -
CAMD
Units *
3160-9
3160-71
3160-81
52106-
150101
52106-
150137
52106-
150110
52106-
150138
52106-
150139
52106-
150140
880007-52
50607-23
CAMD
Unit
Emiss
(tpy)*
1.542
123.8
164
162.1
194.2
162.1
194.2
194.2
194.2
84.5
163.1
CAMD
Comb.
Unit
Totals
(tpy)
1.542
287.8
162.1
938.9
84.5
163.1
CAMD
Facility
Totals
(tpy)
289.3
1101.0
84.5
178.7
Stack 5
(%,
relative
to
CAMD
value)
213%
0%
-70%
10%
90%
-21%
Stack
5
(tpy)
3.3
0.0
113.9
93.9
76.4
-34.5
Facility
5(%
relative
to
CAMD
value)
1%
-2%
90%
6%
Facility
5 (tpy)
3.3
-20.3
76.4
11.4
80
-------
NEI Facility
Name
Grays Ferry
Cogeneration
Partners
Trigen -
Edison
Jefferson
Smurfit
Corporation
(U S) ***
NEI
Comb.
Stack
Number
865
866
867
868
819
820
821
NEI
Comb.
Stack
Emiss
(tpy)
61.5
143.2
90.3
130.5
0.1
113.8
114.5
NEI
Unit
Emiss
(tpy)
61.5
143.2
90.3
130.5
228.4
NEI
Facility
Emiss
(tpy,
w/out
Fugitive)
233.5
130.5
228.4
CAMD
Facility
Name
Schuykill
Grays Ferry
Cogen
Partnership
Trigen
Energy
Corporation-
Edison St
CAMD
Units *
50607-24
50607-26
54785-2
54785-25
880006-1
880006-2
880006-3
880006-4
54785-2
54785-25
CAMD
Unit
Emiss
(tpy)*
2.9
12.7
143.2
90.3
19.8
17.3
36.1
37.8
143.2
90.3
CAMD
Comb.
Unit
Totals
(tpy)
15.6
143.2
90.3
111
233.5
CAMD
Facility
Totals
(tpy)
233.5
111.0
233.5
Stack 5
(%,
relative
to
CAMD
value)
293%
0%
0%
18%
-2%
Stack
5
(tpy)
45.9
0.0
0.0
19.4
-5.1
Facility
5(%
relative
to
CAMD
value)
0%
18%
-2%
Facility
5 (tpy)
0.0
19.4
-5.1
Notes:
* In the format "ORIS ID - UNIT ID"
** All CAMD values are for 2002
*** Jefferson Smurfit not in CAMD; will use Grays Ferry as surrogate
81
-------
1
2
3
4
5
6
7
8
9
10
1 1
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
2002 Annual NOx Emission
Mass [NEI-CAMD] Difference
Facility-Wide 2002 NOx Emission Frequency
versus
Difference in 2002 NOx Emission Mass Between the CAMD and NEI Databases
(ORIS ID labeled)
cn
Art _
c
0
n
t u
0
-C
w 20
ACi
en
880007
50607 88i°°6
• g54785(Gravs Herrv)
3160 • .
O4/ OO(Jefferson Sr
52
urfit)
106
10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
NOx Emission Frequency for 2002
(% hours in 2002 where NOx was emitted)
Figure 14. Differences in facility-wide annual NOx emission totals between NEI and CAMD data bases
for Philadelphia County 2002.
3.2.6 Fugitive and Airport Emissions Preparation
3.2.6.1 Philadelphia County
Fugitive emission releases in Philadelphia County, as totaled in the NEI database, were
modeled as area sources with the profile of these releases determined by the overall facility
profile of emissions. In addition, emissions associated with the Philadelphia International
Airport were estimated.
Fugitive Releases
Thirty five combined stacks were identified during the point source analysis (see previous
section) that were associated with facilities considered major emitters, but where the emissions
from the stacks are labeled Fugitive in the NEI. These stacks have zero stack diameter, zero
emission velocity, and exit temperature equal to average ambient conditions (295 K). Thus, we
determined it was not appropriate to include these in the point source group simulation.
These 35 stacks occur at only two facilities in the County: Exelon Generation Co - Delaware
Station (NEI Site ID: NEIPA2218) and Sunoco Inc. - Philadelphia (NEI Site ID: NEI40723).
Consequently, they were grouped by facility. The Sunoco emissions further fall into two distinct
categories based on release heights. Thus, to accommodate all these sources most efficiently, we
created three area source groups: one for Sunoco emissions at 3.0 m, one for Sunoco emissions
greater than 23.0 m, and one for Exelon. The "stacks" within the NEI and their parameters
comprising each of these sources are shown in Table 39 along with their groupings and the
resulting combined area source parameters.
82
-------
1 Table 39. Emission parameters for the three Philadelphia County fugitive NOX area emission sources.
Grp.
No.
1
2
3
NEI
Site ID
NEIPA
2218
NEI40
723
NEI40
723
Facility Name
EXELON
GENERATIO
NCO-
DELAWARE
STATION
Sunoco Inc. -
Philadelphia
Sunoco Inc. -
Philadelphia
NEI 2002
Emissions
(tpy)
0.1
5.1
5.2
65.3
350.9
12.7
355.7
31.1
6.2
182.4
1.1
7.5
1.0
2.0
49.4
106.3
188.5
87.8
36.1
9.7
61.2
13.6
17.0
17.2
12.2
12.6
23.7
19.2
10.0
1,680.4
79.5
13.1
15.3
2.5
10.2
19.0
211.2
350.8
Stack X
-75.13582
-75.12528
-75.21408
-75.21300
-75.20972
-75.20945
-75.20876
-75.20845
-75.20809
-75.20707
-75.20651
-75.20301
-75.20114
-75.20090
-75.20079
-75.20047
-75.20043
-75.20024
-75.20020
-75.19995
-75.19766
-75.19751
-75.19735
-75.19723
-75.19720
-75.19713
-75.19699
-75.19644
-75.21322
-75.20833
-75.20850
-75.20844
-75.20838
-75.20828
-75.20889
Stack Y
39.96769
39.96680
39.90811
39.90878
39.90467
39.90778
39.90185
39.90708
39.91580
39.90946
39.90988
39.91362
39.91273
39.91621
39.91615
39.91366
39.91377
39.91406
39.91410
39.91596
39.91696
39.91696
39.91590
39.91597
39.91698
39.91596
39.91599
39.91493
39.90899
39.90278
39.90246
39.90239
39.90231
39.90237
39.90279
Stack
Height
(m)
5
8
6.5
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3.0
23
26
27
27
27
27
30
26.7
Stacks
Used for
Emission
Profile 1
817+818
855+856+
857+858+
859
855+856+
857+858+
859
Scaled Emissions (tpy) 2
2001
4.8
1,873.
8
391.2
2002
5.2
1,681.
4
351.0
2003
6.4
2,202
.4
459.8
1 See Table 37 for stack definitions.
2 Scaled emissions are determined by summing the scaled, hourly values
from the CAMD database, as used in the dispersion modeling.
83
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In the case of the Sunoco emissions, the vertices of the area sources were determined by a
convex hull encapsulating all the points. In the case of Excelon, only two points are provided,
which is insufficient information to form a closed polygon. Instead, the boundary of the facility
was digitized into a 20-sided polygon. Figure 15 shows the locations of these polygons.
Emission profiles for the fugitive releases were determined from the CAMD hourly emission
database in a method similar to that for the point sources. We determined scaling factors based
on the ratio of the 2002 fugitive releases described by the NEI to the total, non-fugitive point
source releases from the same facility. All stacks within that facility were combined on an
hourly basis for each year and the fugitive to non-fugitive scaling factor applied, ensuring that
the same temporal emission profile was used for fugitives as for other releases from the facility,
since the origins of the emissions should be parallel. We created external hourly emissions files
for each of the three fugitive area sources with appropriate units (grams per second per square
meter).
unoco (ReleaseHght = 3m)
\
/*
\ ^ Sunoco (ReleaseHght = 23+ m)
KPHL Airport Baggage Handling Area
Figure 15. Locations of the four ancillary area sources. Also shown are centroid receptor locations.
Philadelphia International Airport
Another significant source of NOx emissions in Philadelphia County not captured in the
earlier simulations is from operation of the Philadelphia International Airport (PHL). PHL is the
only major commercial airport in the County and is the largest airport in the Delaware Valley.
84
-------
1
2
3
4
5
6
7
The majority of NOx emissions in the NEI25 database attributable to airports in Philadelphia
County are from non-road mobile sources, specifically ground support equipment. There is
another airport in the County: Northeast Philadelphia Airport. However, because it serves
general aviation, is generally much smaller in operations than PHL, and has little ground support
equipment activity - which is associated primarily with commercial aviation - all airport
emissions in the County were attributed to PHL. The PHL emissions were taken from the non-
road section of the 2002 NEI, and are shown by Table 40.
Table 40. Philadelphia International airport (PHL) NOX emissions
State and
County
Philadelphia,
PA
PHL Total
scc
2265008005
2267008005
2270008005
2275020000
2275050000
NOx
(tpy)
4.6
5.1
196.2
0.01
2.5
208.4
SCC Level 1
Description
Mobile
Sources
Mobile
Sources
Mobile
Sources
Mobile
Sources
Mobile
Sources
SCC Level 3
Description
Off-highway
Vehicle
Gasoline, 4-
Stroke
LPG
Off-highway
Vehicle
Diesel
Aircraft
Aircraft
SCC Level 6
Description
Airport
Ground
Support
Equipment
Airport
Ground
Support
Equipment
Airport
Ground
Support
Equipment
Commercial
Aircraft
General
Aviation
SCC Level 8
Description
Airport
Ground
Support
Equipment
Airport
Ground
Support
Equipment
Airport
Ground
Support
Equipment
Total: All
Types
Total
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
As with the fugitive sources discussed above, the airport emissions are best parameterized as
area sources. The boundary of the area source was taken as the region of operation of baggage
handling equipment, including the terminal building and the region surrounding the gates. This
region was digitized into an 18-sided polygon of size 1,326,000 m2, and included in the
AERMOD input control file.
The activity profile for PHL was taken to have seasonal and hourly variation (SEASHR),
based on values from the EMS-HAP model.26 These factors are disaggregated in the EMS-HAP
model database based on source classification codes (SCCs), which were linked to those from
the NEI database. The EMS-HAP values provide hourly activity factors by season, day type, and
hour; to compress to simple SEASHR modeling, the hourly values from the three individual day
types were averaged together. The total emissions for each SCC were then disaggregated into
seasonal and hourly components and the resulting components summed to create total PHL
emissions for each hour of the four annual seasons. These parameterized emissions were then
normalized to the total cargo handling operational area, to produce emission factors in units of
grams per second per square meter and included in the AERMOD input file. Figure 15 also
shows the location of the PHL area source.
25 http://www.epa.gov/ttn/chief/net/2002inventory.html
26 EPA 2004, User's Guide for the Emissions Modeling System for Hazardous Air Pollutants (EMS-HAP) Version
3.0, EPA-454/B-03-006.
85
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
3.2.7 Receptor Locations
3.2.7.1 Philadelphia County
Three sets of receptors were chosen to represent the locations of interest. First, all NOx
monitor locations, shown by Table 41, within the Philadelphia CMS A were included as receptor
locations. Although all receptors are assumed to be on a flat plane, they are placed at the
standard breathing height of 5.9 ft (1 .8 m).
Table 41. Philadelphia CMSA NOX monitors.
CMSA
Philadelphia-
Wilmington-
Atl antic City,
PA-NJ-DE-MD
Site ID
100031003
100031007
100032004
340070003
420170012
420450002
420910013
421010004
421010029
421010047
Latitude
39.7611
39.5511
39.7394
39.923
40.1072
39.8356
40.1122
40.0089
39.9572
39.9447
Longitude
-75.4919
-75.7308
-75.5581
-75.0976
-74.8822
-75.3725
-75.3092
-75.0978
-75.1731
-75.1661
The second receptor locations were selected to represent the locations of census block
centroids near major NOX sources. GIS analysis was used to determine all block centroids in
Philadelphia County that lie within a 0.25 mile (400 m) of the roadway segments and also all
block centroids that lie within 6.2 miles (10 km) of any major point source. 12,982 block
centroids were selected due to their proximity to major roadways; 16,298 centroids were selected
due to their proximity to major sources. The union of these sets produced 16,857 unique block
centroid receptor locations, each of which was assigned a height of 5.9 ft (1.8 m). The locations
of centroids that met either distance criteria - and were thus included in the modeling - is shown
by Figure 16.
86
-------
1
2
3
4
5
6
1
8
9
10
11
81
Combined Point Stack
Census Block Crntrelds within 400m of M ajor Roadway Edg« OR within 10km of S
Figure 16. Centroid locations within fixed distances to major point and mobile sources.
The third set of receptors was chosen to represent the on-road microenvironment. For this
set, one receptor was placed at the center of each of the 982 sources.
The distance relationship between the road segments and block centroids can be estimated by
looking at the distance between the road-centered and the block centroid receptors. Figure 17
shows the histogram of the shortest distance between each centroid receptor and its nearest
roadway-centered receptor.
87
-------
rrrrrrrrrrf rrrrrrrrn
IHlL|lL,HL>L,LhLLiLLL,L,LL.| LLLLLLLLI
~ 1 UU /O
- 90%
80%
- 70%
- 60%
- 50%
- 40%
- 30%
- 20%
- 10%
- no/.
d
u_
a
o
o
o
LO O
IT) O
CM O
CM -^ CD CO
CM CM CM CM CM
LO
CD
CM
O
IT) h-
I Distance (m)
2 Figure 17. Frequency distribution of distance between each Census receptor and its nearest road-
3 centered receptor.
4
5 The centroids selected were those within 10 km of any major point source or 400 m from any
6 receptor edge, so the distances to the nearest major road segment can be significantly greater
7 than 400 m. The mode of the distribution is about 150 m and the median distance to the closest
8 roadway segment center is about 450 m. However, these values represent the distances of the
9 block centroids to road centers instead of road edges, so that they overestimate the actual
10 distances to the zone most influenced by roadway by an average of 14 m and a range of 4 m to
11 44 m (see Table 36 above).
12 3.2.8 Other Modeling Specifications
13 Since each of the case-study locations were MS A/CMS As, all emission sources were
14 characterized as urban. The AERMOD toxics enhancements were also employed to speed
15 calculations from area sources. NOX chemistry was applied to all sources to determine NC>2
16 concentrations. For the each of the roadway, fugitive, and airport emission sources, the ozone
17 limiting method (OLM) was used, with plumes considered ungrouped. Because an initial NC>2
18 fraction of NOX is anticipated to be about 10% or less (Finlayson-Pitts and Pitts, 2000; Yao et al.,
19 2005), a conservative value of 10% for all sources was selected. For all point source simulations
20 the Plume Volume Molar Ratio Method (PVMRM) was used to estimate the conversion of NOX
21 to NC>2, with the following settings:
22 1. Hourly series of Os concentrations were taken from EPA's AQS database27. The
23 complete national hourly record of monitored Os concentrations were filtered for the
24 four monitors within Philadelphia County (stations 421010004, 421010014,
25 421010024, and 421010136). The hourly records of these stations were then
26 averaged together to provide an average Philadelphia County concentrations of Os for
27 each hour of 2001-2003.
27 http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm
88
-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
17
18
2. The equilibrium value for the NO2:NOX ratio was taken as 75%, the national average
ambient ratio.28
3. The initial NC>2 fraction of NOX is anticipated to be about 10% or less. A default
value of 10% was used for all stacks (Finlayson-Pitts and Pitts, 2000).
3.2.9 Air Quality Concentration Estimation
The hourly concentrations estimated from each of the three source categories were combined
at each receptor. Then a local concentration, reflecting the concentration contribution from
emission sources not included in the simulation, was added to the sum of the concentration
contributions from each of these sources at each receptor. The local concentration was estimated
from the difference between the model predictions at the local NC>2 monitors and the observed
values. It should be noted that this local concentration may also include any model error present
in estimating concentration at the local monitoring sites. Table 42 presents a summary of the
estimated local concentration added to the AERMOD hourly concentration data.
16 Table 42. Comparison of ambient monitoring and AERMOD predicted NO2 concentrations.
Year and
Monitor ID
Annual Average NO2 concentration (ppb)
Monitor
AERMOD
Inititial
Difference1
AERMOD
Final2
2001
4210100043
4210100292
4210100471
mean
26
28
30
7
22
20
18
6
10
11
19
33
32
2002
4210100043
4210100292
4210100471
mean
24
28
29
7
21
19
17
7
10
11
18
32
31
2003
4210100043
4210100292
4210100471*
mean
24
25
25
7
22
26
17
3
-1
6
13
28
32
1 the difference represents concentrations attributed to sources
not modeled by AERMOD and model error.
2 the mean difference between measured and modeled was
added uniformly at each receptor hourly concentration to
generate the AERMOD final concentrations.
* monitor did not meet completeness criteria used in the air
quality characterization.
1 Appendix W to CFR 51, page 466. http://www.epa.gov/scramOOl/guidance/guide/appw_03.pdf.
89
-------
i 3.3 Human Exposure Modeling using APEX
2 The Air Pollutants Exposure model (APEX) is a personal computer (PC)-based program
3 designed to estimate human exposure to criteria and air toxic pollutants at the local, urban, and
4 consolidated metropolitan levels. APEX, also known as TREVI.Expo, is the human inhalation
5 exposure module of EPA's Total Risk Integrated Methodology (TRIM) model framework (US
6 EPA, 1999), a modeling system with multimedia capabilities for assessing human health and
7 ecological risks from hazardous and criteria air pollutants. It is being developed to support
8 evaluations with a scientifically sound, flexible, and user-friendly methodology. Additional
9 information on the TRIM modeling system, as well as downloads of the APEX Model, user's
10 guide, and other supporting documentation, can be found on EPA's Technology Transfer
11 Network (TTN) at http://www.epa.gov/ttn/fera.
12 3.3.1 History
13 APEX was derived from the National Ambient Air Quality Standards (NAAQS) Exposure
14 Model (NEM) series of models, developed to estimate exposure to the criteria pollutants (e.g.,
15 carbon monoxide (CO), ozone 63). In 1979, EPA began by assembling a database of human
16 activity patterns that could be used to estimate exposures to indoor and outdoor pollutants
17 (Roddin et al., 1979). These data were then combined with measured outdoor concentrations in
18 NEM to estimate exposures to CO (Biller et al., 1981; Johnson and Paul, 1983). In 1988,
19 OAQPS began to incorporate probabilistic elements into the NEM methodology and use activity
20 pattern data based on various human activity diary studies to create an early version of
21 probabilistic NEM for 63 (i.e., pNEM/Os). In 1991, a probabilistic version of NEM was
22 extended to CO (pNEM/CO) that included a one-compartment mass-balance model to estimate
23 CO concentrations in indoor microenvironments. The application of this model to Denver,
24 Colorado has been documented in Johnson et al. (1992). Additional enhancements to pNEM/Os
25 in the early- to mid-1990's allowed for probabilistic exposure assessments in nine urban areas for
26 the general population, outdoor children, and outdoor workers (Johnson et al., 1996a; 1996b;
27 1996c). Between 1999 and 2001, updated versions of pNEM/CO (versions 2.0 and 2.1) were
28 developed that relied on activity diary data from EPA's Consolidated Human Activities Database
29 (CHAD) and enhanced algorithms for simulating gas stove usage, estimating alveolar ventilation
30 rate (a measure of human respiration), and modeling home-to-work commuting patterns.
31
32 The first version of APEX was essentially identical to pNEM/CO (version 2.0) except that it
33 was capable of running on a PC instead of a mainframe. The next version, APEX2, was
34 substantially different, particularly in the use of a personal profile approach (i.e., simulation of
35 individuals) rather than a cohort simulation (i.e., groups of similar persons). APEX3 introduced
36 a number of new features including automatic site selection from national databases, a series of
37 new output tables providing summary exposure and dose statistics, and a thoroughly reorganized
38 method of describing microenvironments and their parameters. Most of the spatial and temporal
39 constraints of pNEM and APEX1 were removed or relaxed by version 3.
40
41 The version of APEX used in this exposure assessment is APEX4, described in the APEX
42 User's Guide and the APEX Technical Support Document (US EPA, 2006a; 2006b), referred to
43 as the APEX User's Guide and TSD.
90
-------
1 3.3.2 Model Overview
2 APEX estimates human exposure to criteria and toxic air . . . .. ..
-> ,, , , iiU , , , ,., 4 , 4 ri A microenvironment is athree-
3 pollutants at the local, urban, or consolidated metropolitan .. . . .. . .
„ i i • * u *• • • * i u dimensional space in which human
4 area levels using a stochastic, microenvironmental approach. . . ... . . .
« TU j ^ A 1 1 + A + f 1 fu +u +- 1 contact with an environmental
5 The model randomly selects data tor a sample or hypothetical „ . .. . . , ,. ,
. . ,. ., , ,, , , , 4- , 4 , / • T . pollutant takes place and which can
6 individuals from an actual population database and simulates £ . . . „ u ± • _i
-7 uu .u .- i • j- -j i> . ^ u.- j be treated as a well-characterized,
7 each hypothetical individual s movements through time and . .. . . . ..
0 ;F ,, . ,. , ,, .-..,• . relatively homogeneous location
8 space (e.g., at home, in vehicles) to estimate their exposure to ... ..,,..
n u ;/ ADCV • i+ +- A^ with respect to pollutant
9 a pollutant. APEX simulates commuting, and thus exposures . .. t .,. ,..
1A ,,, iU , , , 4. r • j- -j i u concentrations for a specified time
10 that occur at home and work locations, for individuals who . . r
11 work in different areas than they live.
12
13 APEX can be conceptualized as a simulated field study that would involve selecting an actual
14 sample of specific individuals who live in (or work and live in) a geographic area and then
15 continuously monitoring their activities and subsequent inhalation exposure to a specific air
16 pollutant during a specific period of time.
17
18 The main differences between APEX and an actual field study are that in APEX:
19 • The sample of individuals is a virtual sample, not actual persons. However, the
20 population of individuals appropriately balanced according to various demographic
21 variables and census data using their relative frequencies, in order to obtain a
22 representative sample (to the extent possible) of the actual people in the study area
23 • The activity patterns of the sampled individuals (e.g., the specification of indoor and
24 other microenvironments visited and the time spent in each) are assumed by the model to
25 be comparable to individuals with similar demographic characteristics, according to
26 activity data such as diaries compiled in EPA's Consolidated Human Activity Database
27 (or CHAD; US EPA, 2002; McCurdy et al., 2000)
28 • The pollutant exposure concentrations are estimated by the model using a set of user-
29 input ambient outdoor concentrations (either modeled or measured) and information on
30 the behavior of the pollutant in various microenvironments;
31 • Variation in ambient air quality levels can be simulated by either adjusting air quality
32 concentrations to just meet alternative ambient standards, or by reducing source
33 emissions and obtaining resulting air quality modeling outputs that reflect these potential
34 emission reductions, and
35 • The model accounts for the most significant factors contributing to inhalation exposure -
36 the temporal and spatial distribution of people and pollutant concentrations throughout
37 the study area and among microenvironments - while also allowing the flexibility to
38 adjust some of these factors for alternative scenarios and sensitivity analyses.
39
40 APEX is designed to simulate human population exposure to criteria and air toxic pollutants
41 at local, urban, and regional scales. The user specifies the geographic area to be modeled and the
42 number of individuals to be simulated to represent this population. APEX then generates a
43 personal profile for each simulated person that specifies various parameter values required by the
44 model. The model next uses diary-derived time/activity data matched to each personal profile to
45 generate an exposure event sequence (also referred to as activity pattern or diary) for the
46 modeled individual that spans a specified time period, such as one year. Each event in the
91
-------
1 sequence specifies a start time, exposure duration, geographic location, microenvironment, and
2 activity performed. Probabilistic algorithms are used to estimate the pollutant concentration
3 associated with each exposure event. The estimated pollutant concentrations account for the
4 effects of ambient (outdoor) pollutant concentration, penetration factors, air exchange rates,
5 decay/deposition rates, and proximity to emission sources, depending on the microenvironment,
6 available data, and estimation method selected by the user. Because the modeled individuals
7 represent a random sample of the population of interest, the distribution of modeled individual
8 exposures can be extrapolated to the larger population. The model simulation can be broadly
9 described in five steps that follow:
10
11 1. Characterize the study area. APEX selects census tracts within a study area - and thus
12 identifies the potentially exposed population - based on user-defined criteria and
13 availability of air quality and meteorological data for the area.
14 2. Generate simulated individuals. APEX stochastically generates a sample of
15 hypothetical individuals based on the census data for the study area and human profile
16 distribution data (such as age-specific employment probabilities).
17 3. Construct a sequence of activity events. APEX constructs an exposure event sequence
18 spanning the period of the simulation for each of the simulated individuals and based on
19 the activity pattern data.
20 4. Calculate hourly concentrations in microenvironments. APEX users define
21 microenvironments that people in the study area would visit by assigning location codes
22 in the activity pattern to the user-specified microenvironments. The model then
23 calculates hourly concentrations of a pollutant in each of these microenvironments for the
24 period of simulation, based on the user-provided microenvironment descriptions and
25 hourly air quality data. Microenvironmental concentrations are calculated for each of the
26 simulated individuals.
27 5. Determine exposures. APEX estimates a concentration for each exposure event based
28 on the microenvironment occupied during the event. These values can be averaged by
29 clock hour to produce a sequence of hourly average exposures spanning the specified
30 exposure period. These hourly values may be further aggregated to produce daily,
31 monthly, and annual average exposure values.
32 3.3.3 Study Area Descriptions
33 The APEX study area has traditionally been on the scale of a city or slightly larger
34 metropolitan area, although it is now possible to model larger areas such as combined statistical
35 areas (CSAs). In this analysis the study area is defined by a single or few counties. The
36 demographic data used by the model to create personal profiles is provided at the census block
37 level. For each block the model requires demographic information representing the distribution
38 of age, gender, race, and work status within the study population. Each block has a location
39 specified by latitude and longitude for some representative point (e.g., geographic center). The
40 current release of APEX includes input files that already contain this demographic and location
41 data for all census tracts, block groups, and blocks in the 50 United States, based on the 2000
42 Census.
43
44 Philadelphia County is comprised of 17,315 blocks containing a population of 1,517,550
45 persons. For this analysis the population studied was limited those residents of Philadelphia
92
-------
1 County residing in census blocks that were either within 400 meters of a major roadway or
2 within 10 km of a major emission source (see section 3.2.7 for definition). This was done to
3 maintain balance between the representation of the study area/objectives and the computational
4 load regarding file size and processing time. There were 16,857 such blocks containing a
5 population of 1,475,651.
6
7 3.3.3.1 Air Quality Data
8 Air quality data input to the model were generated by air quality modeling. Principal
9 emission sources included both mobile and stationary sources as well as fugitive emissions. The
10 methodology was described previously in Section 3.2. Briefly, hourly NC>2 concentrations were
11 estimated for each of 3 years (2001-2003) at each of the defined receptor locations using hourly
12 NOX emission estimates and dispersion modeling.
13
14 In APEX, the ambient air quality data are assigned to geographic areas called districts. The
15 districts are used to assign pollutant concentrations to the blocks/tracts and microenvironments
16 being modeled. The ambient air quality data are provided by the user as hourly time series for
17 each district. As with blocks/tracts, each district has a representative location (latitude and
18 longitude). APEX calculates the distance from each block/tract to each district center, and
19 assigns the block/tract to the nearest district, provided the block/tract representative location
20 point (e.g., geographic center) is in the district. Each block/tract can be assigned to only one
21 district. In this assessment the district was synonymous with the receptor modeled in the
22 dispersion modeling (see Sections 3.2).
23
24 3.3.3.2 Meteorological Data
25 Ambient temperatures are input to APEX for different sites (locations). As with districts,
26 APEX calculates the distance from each block to each temperature site and assigns each block to
27 the nearest site. Hourly temperature data are from the National Climatic Data Center Surface
28 Airways Hourly TD-3280 dataset (NCDC Surface Weather Observations). Daily average and 1-
29 hour maxima are computed from these hourly data.
30
31 There are two files that are used to provide meteorological data to APEX. One file, the
32 meteorological station location file, contains the locations of meteorological data recordings
33 expressed in latitude and longitude coordinates. This file also contains start and end dates for the
34 data recording periods. The temperature data file contains the data from the locations in the
35 temperature zone location file. This file contains hourly temperature readings for the period
36 being modeled for the meteorological stations in and around the study area. Table 43 lists the
37 meteorological stations used for each modeled area.
38 Table 43. The meteorological stations used for each study area.
Location
Atlanta, GA
Detroit, Ml
Los Angeles, CA
Philadelphia County, PA
Phoenix
World Meteorological Organization
ID (call sign)
72219 (KATL)
72537 (KDTW)
72295 (KLAX)
72408 (KPHL)
72278 (KPHX)
93
-------
1 3.3.4 Simulated Individuals
2 APEX stochastically generates a user-specified number of simulated persons to represent the
3 population in the study area. Each simulated person is represented by a personal profile, a
4 summary of personal attributes that define the individual. APEX generates the simulated person
5 or profile by probabilistically selecting values for a set of profile variables (Table 44). The
6 profile variables could include:
7 • Demographic variables, generated based on the census data;
8 • Physical variables, generated based on sets of distribution data;
9 • Other daily varying variables, generated based on literature-derived distribution data that
10 change daily during the simulation period.
11 APEX first selects demographic and physical attributes for each specified individuals, and
12 then follows the individual over time and calculates his or her time series of exposure.
13 Table 44. Examples of profile variables in APEX.
Variable
Type
Demographic
Physical
Profile Variables
Age
Gender
Home block
Work tract
Employment status
Air conditioner
Gas Stove
Description
Age (years)
Male or Female
Block in which a simulated person lives
Tract in which a simulated person works
Indicates employment outside home
Indicates presence of air conditioning at home
Indicates presence of gas stove at home
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Due to the large size of the air quality input files, the modeled area was separated into three
sections. The number of simulated persons in each model run (3 sections per 3 years) was set to
50,000, yielding a total of 150,000 persons simulated for each year. The parameters controlling
the location and size of the simulated area were set to include the county(s) in the selected study
area. The settings that allow for replacement of CHAD data that are missing gender,
employment or age values were all set to preclude replacing missing data. The width of the age
window was set to 20 percent to increase the pool of diaries available for selection. The variable
that controls the use of additional ages outside the target age window was set to 0.1 to further
enhance variability in diary selection. See the APEX User's Guide for further explanation of
these parameters. The total population simulated for Philadelphia County was approximately
1.48 million persons, of which there a total simulated population of 163,000 asthmatics. The
model simulated approximately 281,000 children, of which there were about 48,000 asthmatics.
Due to random sampling, the actual number of specific subpopulations modeled varied slightly
by year.
3.3.4.1 Population Demographics
APEX takes population characteristics into account to develop accurate representations of
study area demographics. Specifically, population counts by area and employment probability
94
-------
1 estimates are used to develop representative profiles of hypothetical individuals for the
2 simulation.
O
4 APEX is flexible in the resolution of population data provided. As long as the data are
5 available, any resolution can be used (e.g., county, census tract, census block). For this
6 application of the model, census block level data were used. Block-level population counts come
7 from the 2000 Census of Population and Housing Summary File 1 (SF-1). This file contains the
8 100-percent data, which is the information compiled from the questions asked of all people and
9 about every housing unit.
10
11 As part of the population demographics inputs, it is important to integrate working patterns
12 into the assessment. In the 2000 U.S. Census, estimates of employment were developed by
13 census information (US Census Bureau, 2007). The employment statistics are broken down by
14 gender and age group, so that each gender/age group combination is given an employment
15 probability fraction (ranging from 0 to 1) within each census tract. The age groupings used are:
16 16-19, 20-21, 22-24, 25-29, 30-34, 35-44, 45-54, 55-59, 60-61, 62-64, 65-69, 70-74, and >75.
17 Children under 16 years of age were assumed to be not employed.
18
19 Since this analysis was conducted at the census block level, block level employment
20 probabilities were required. It was assumed that the employment probabilities for a census tract
21 apply uniformly to the constituent census blocks.
22
23 3.3.4.2 Commuting
24 In addition to using estimates of employment by tract, APEX also incorporates home-to-
25 work commuting data. Commuting data were originally derived from the 2000 Census and were
26 collected as part of the Census Transportation Planning Package (CTPP) (US DOT, 2007). The
27 data used contain counts of individuals commuting from home to work locations at a number of
28 geographic scales. These data were processed to calculate fractions for each tract-to-tract flow to
29 create the national commuting data distributed with APEX. This database contains commuting
30 data for each of the 50 states and Washington, D.C.
31 Commuting within the Home Tract
32 The APEX data set does not differentiate people that work at home from those that
33 commute within their home tract.
34 Commuting Distance Cutoff
35 A preliminary data analysis of the home-work counts showed that a graph of log(flows)
36 versus log(distance) had a near-constant slope out to a distance of around 120 kilometers.
37 Beyond that distance, the relationship also had a fairly constant slope but it was flatter, meaning
38 that flows were not as sensitive to distance. A simple interpretation of this result is that up to
39 120 km, the majority of the flow was due to persons traveling back and forth daily, and the
40 numbers of such persons decrease fairly rapidly with increasing distance. Beyond 120 km, the
41 majority of the flow is made up of persons who stay at the workplace for extended times, in
42 which case the separation distance is not as crucial in determining the flow.
43 To apply the home-work data to commuting patterns in APEX, a simple rule was chosen. It
44 was assumed that all persons in home-work flows up to 120 km are daily commuters, and no
95
-------
1 persons in more widely separated flows commute daily. This meant that the list of destinations
2 for each home tract was restricted to only those work tracts that are within 120 km of the home
3 tract. When the same cutoff was performed on the 1990 census data, it resulted in 4.75% of the
4 home-work pairs in the nationwide database being eliminated, representing 1.3% of the workers.
5 The assumption is that this 1.3% of workers do not commute from home to work on a daily
6 basis. It is expected that the cutoff reduced the 2000 data by similar amounts.
7 Eliminated Records
8 A number of tract-to-tract pairs were eliminated from the database for various reasons. A fair
9 number of tract-to-tract pairs represented workers who either worked outside of the U.S. (9,631
10 tract pairs with 107,595 workers) or worked in an unknown location (120,830 tract pairs with
11 8,940,163 workers). An additional 515 workers in the commuting database whose data were
12 missing from the original files, possibly due to privacy concerns or errors, were also deleted.
13 Commuting outside the study area
14 APEX allows for some flexibility in the treatment of persons in the modeled population who
15 commute to destinations outside the study area. By specifying "KeepLeavers = No" in the
16 simulation control parameters file, people who work inside the study area but live outside of it
17 are not modeled, nor are people who live in the study area but work outside of it. By specifying
18 "KeepLeavers = Yes," these commuters are modeled. This triggers the use of two additional
19 parameters, called LeaverMult and LeaverAdd. While a commuter is at work, if the workplace is
20 outside the study area, then the ambient concentration is assumed to be related to the average
21 concentration over all air districts at the same point in time, and is calculated as:
22 Ambient Concentrat ion = LeaverMult x avg(t) + LeaverAdd eq (3)
23 where:
24 Ambient Concentration = Calculated ambient air concentrations for locations outside
25 of the study area (ppm or ppm)
26 LeaverMult = Multiplicative factor for city-wide average concentration,
27 applied when working outside study area
28 avg(t) = Average ambient air concentration over all air districts in
29 study area, for time t (ppm or ppm)
30 LeaverAdd = Additive term applied when working outside study area
31 All microenvironmental concentrations for locations outside of the study area are determined
32 from this ambient concentration by the same function as applies inside the study area.
33 Block-level commuting
34 For census block simulations, APEX requires block-level commuting file. A special software
35 preprocesser was created to generate this files for APEX on the basis of the tract-level
36 commuting data and finely-resolved land use data. The software calculates commuting flows
37 between census blocks for the employed population according to the following equation.
38
3 9 Fl°W block =Fl°W tract X Fpop X Fland
40 where:
96
-------
2 Flow biock = flow of working population between a home block and a work block.
3 Flow tract = flow of working population between a home tract and a work tract.
4 Fpop = fraction of home tract's working population residing in the home block.
5 F imd = fraction of work tract's commercial/industrial land area in the work block
6 Thus, it is assumed that the frequency of commuting to a workplace block within a tract is
7 proportional to the amount of commercial and industrial land in the block.
8
9 3.3.4.3 Profile Functions
10 A Profile Functions file contains settings used to generate results for variables related to
11 simulated individuals. While certain settings for individuals are generated automatically by
12 APEX based on other input files, including demographic characteristics, others can be specified
13 using this file. For example, the file may contain settings for determining whether the profiled
14 individual's residence has an air conditioner, a gas stove, etc. As an example, the Profile
15 Functions file contains fractions indicating the prevalence of air conditioning in the cities
16 modeled in this assessment (Figure 18). APEX uses these fractions to stochastically generate air
17 conditioning status for each individual. The derivation of particular data used in specific
18 microenvironments is provided below.
19
AC_Home
! Has air conditioning at home
TABLE
INPUT 1 PROBABILITY 2 "A/C probabilities"
0.850.15
RESULT INTEGER 2 "Yes/No"
12
#
20
21 Figure 18. Example of a profile function file for A/C prevalence.
22
23 3.3.4.4 Asthma Prevalence Rates
24 One of the important population subgroups for the exposure assessment is asthmatic children.
25 Evaluation of the exposure of this group with APEX requires the estimation of children's asthma
26 prevalence rates. The proportion of the population of children characterized as being asthmatic
27 was estimated by statistics on asthma prevalence rates recently used in the NAAQS review for
28 Os (US EPA, 2007d; 2007e). Specifically, the analysis generated age and gender specific asthma
29 prevalence rates for children ages 0-17 using data provided in the National Health Interview
30 Survey (NHIS) for 2003 (CDC, 2007). These asthma rates were characterized by geographic
31 regions, namely Midwest, Northeast, South, and West. Adult asthma prevalence rates for
32 Philadelphia County were obtained from the Behavioral Risk Factor Surveillance System
33 (BRFSS) survey information (PA DOH, 2008). The average rates for adult males and females in
34 Philadelphia for 2001-2003 were 7% and 12%, respectively. These rates were assumed to apply
35 to all adults uniformly. Table 45 provides a summary of the prevalence rates used in the
36 exposure analysis by age and gender for each of the four regions as applied to the five study
37 areas in the exposure assessment.
38
97
-------
1 Table 45. Asthma prevalence rates by age and gender for 4 regions.
Region
(Study Area)
Midwest
(Detroit)
Northeast
(Philadelphia)
South
(Atlanta)
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18+
0
1
2
3
4
5
6
7
Females
Prevalence se L95 U95
0.070
0.071
0.073
0.075
0.081
0.095
0.092
0.090
0.086
0.110
0.162
0.196
0.212
0.170
0.140
0.133
0.140
0.165
0.068
0.072
0.075
0.077
0.082
0.116
0.161
0.185
0.171
0.145
0.135
0.141
0.166
0.174
0.151
0.146
0.146
0.157
0.070
0.034
0.052
0.071
0.088
0.099
0.119
0.122
0.112
0.036
0.020
0.018
0.019
0.022
0.026
0.029
0.026
0.022
0.027
0.035
0.039
0.040
0.034
0.026
0.023
0.022
0.040
0.066
0.038
0.022
0.020
0.023
0.030
0.037
0.041
0.040
0.035
0.031
0.031
0.034
0.034
0.029
0.028
0.031
0.054
0.013
0.012
0.014
0.017
0.019
0.022
0.023
0.022
0.021
0.037
0.042
0.042
0.044
0.051
0.045
0.047
0.048
0.063
0.098
0.123
0.137
0.107
0.092
0.091
0.098
0.093
0.007
0.021
0.038
0.042
0.043
0.063
0.092
0.108
0.096
0.080
0.078
0.084
0.102
0.109
0.095
0.093
0.088
0.068
0.040
0.015
0.031
0.046
0.056
0.064
0.079
0.079
0.072
0.203
0.130
0.124
0.132
0.144
0.171
0.178
0.166
0.149
0.186
0.255
0.298
0.313
0.258
0.209
0.192
0.198
0.275
0.442
0.221
0.145
0.138
0.151
0.205
0.266
0.298
0.284
0.246
0.223
0.227
0.259
0.266
0.232
0.221
0.232
0.322
0.140
0.077
0.085
0.109
0.134
0.150
0.175
0.182
0.170
Males
Prevalence se L95 U95
0.031
0.063
0.108
0.158
0.216
0.178
0.128
0.121
0.128
0.147
0.177
0.190
0.195
0.169
0.168
0.180
0.201
0.237
0.048
0.046
0.052
0.068
0.100
0.149
0.207
0.228
0.222
0.212
0.177
0.166
0.183
0.171
0.170
0.182
0.204
0.242
0.120
0.041
0.070
0.102
0.129
0.144
0.165
0.164
0.133
0.015
0.018
0.021
0.027
0.037
0.035
0.028
0.026
0.027
0.030
0.030
0.030
0.031
0.028
0.026
0.026
0.030
0.058
0.033
0.018
0.015
0.018
0.023
0.029
0.042
0.045
0.043
0.041
0.037
0.035
0.036
0.031
0.029
0.029
0.032
0.061
0.019
0.016
0.017
0.021
0.024
0.024
0.025
0.023
0.010
0.033
0.070
0.107
0.145
0.113
0.078
0.074
0.079
0.093
0.120
0.131
0.135
0.115
0.117
0.130
0.142
0.132
0.010
0.019
0.027
0.037
0.059
0.094
0.129
0.143
0.142
0.136
0.108
0.102
0.116
0.113
0.115
0.127
0.142
0.133
0.090
0.015
0.041
0.070
0.088
0.099
0.118
0.116
0.090
0.090
0.115
0.163
0.228
0.308
0.270
0.204
0.193
0.200
0.226
0.254
0.266
0.272
0.242
0.235
0.243
0.277
0.388
0.200
0.108
0.097
0.120
0.164
0.226
0.316
0.343
0.332
0.316
0.275
0.259
0.276
0.250
0.244
0.254
0.284
0.399
0.150
0.110
0.116
0.146
0.184
0.205
0.224
0.226
0.194
98
-------
Region
(Study Area)
West
(Los
Angeles)
(Phoenix)
Notes:
se - Standard
L95 - Lower lin
U95- Upper lir
Age
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Females
Prevalence se L95 U95
0.093
0.091
0.108
0.132
0.123
0.097
0.095
0.100
0.115
0.145
0.013
0.031
0.054
0.074
0.077
0.077
0.073
0.081
0.091
0.102
0.122
0.127
0.131
0.120
0.111
0.112
0.122
0.019
0.018
0.020
0.023
0.020
0.017
0.016
0.016
0.016
0.029
0.010
0.013
0.015
0.018
0.021
0.021
0.020
0.020
0.020
0.023
0.027
0.027
0.025
0.024
0.021
0.021
0.022
17 0.144 0.040
error
nit on 95th confidence interval
nit on 95th confidence interval
0.059
0.059
0.071
0.090
0.085
0.065
0.064
0.070
0.084
0.091
0.002
0.012
0.029
0.043
0.042
0.042
0.039
0.047
0.055
0.061
0.074
0.079
0.084
0.076
0.071
0.074
0.081
0.076
0.144
0.139
0.162
0.191
0.175
0.142
0.137
0.141
0.156
0.223
0.067
0.078
0.098
0.127
0.137
0.139
0.131
0.138
0.146
0.167
0.194
0.200
0.197
0.183
0.167
0.166
0.180
0.256
Males
Prevalence se L95 U95
0.138
0.168
0.178
0.162
0.145
0.143
0.153
0.151
0.140
0.122
0.031
0.046
0.063
0.078
0.091
0.113
0.121
0.127
0.132
0.151
0.164
0.170
0.175
0.162
0.165
0.170
0.179
0.192
0.023
0.025
0.025
0.022
0.020
0.019
0.019
0.017
0.018
0.026
0.025
0.019
0.014
0.019
0.025
0.029
0.029
0.028
0.027
0.028
0.026
0.026
0.027
0.028
0.026
0.025
0.025
0.043
0.095
0.121
0.130
0.119
0.106
0.105
0.116
0.116
0.105
0.075
0.004
0.017
0.036
0.044
0.048
0.060
0.068
0.075
0.080
0.096
0.112
0.117
0.120
0.107
0.112
0.120
0.127
0.111
0.197
0.230
0.240
0.218
0.195
0.192
0.200
0.194
0.185
0.193
0.186
0.116
0.106
0.136
0.168
0.201
0.207
0.208
0.208
0.229
0.233
0.240
0.248
0.237
0.236
0.236
0.246
0.312
1
2 3.3.5 Activity Pattern Sequences
3 Exposure models use human activity pattern data to predict and estimate exposure to
4 pollutants. Different human activities, such as spending time outdoors, indoors, or driving, will
5 have varying pollutant exposure concentrations. To accurately model individuals and their
6 exposure to pollutants, it is critical to understand their daily activities.
7
8 The Consolidated Human Activity Database (CHAD) provides data for where people spend
9 time and the activities performed. CHAD was designed to provide a basis for conducting multi-
10 route, multi-media exposure assessments (McCurdy et al., 2000). The data contained within
11 CHAD come from multiple activity pattern surveys with varied structures (Table 46), however
99
-------
1 the surveys have commonality in containing daily diaries of human activities and personal
2 attributes (e.g., age and gender).
O
4 There are four CHAD-related input files used in APEX. Two of these files can be
5 downloaded directly from the CHADNet (http://www.epa.gov/chadnetl), and adjusted to fit into
6 the APEX framework. These are the human activity diaries file and the personal data file, and
7 are discussed below. A third input file contains metabolic information for different activities
8 listed in the diary file, these are not used in this exposure analysis. The fourth input file maps
9 five-digit location codes used in the diary file to APEX microenvironments; this file is discussed
10 in the section describing microenvironmental calculations (Section 3.3.6).
11
12 3.3.5.1 Personal Information file
13 Personal attribute data are contained in the CHAD questionnaire file that is distributed with
14 APEX This file also has information for each day individuals have diaries. The different
15 variables in this file are:
16
17 • The study, person, and diary day identifiers
18 • Day of week
19 • Gender
20 • Employment status
21 • Age in years
22 • Maximum temperature in degrees Celsius for this diary day
23 • Mean temperature in degrees Celsius for this diary day
24 • Occupation code
25 • Time, in minutes, during this diary day for which no data are included in the database
26
27 3.3.5.2 Diary Events file
28 The human activity diary data are contained in the events file that is distributed with APEX.
29 This file contains the activities for the nearly 23,000 people with intervals ranging from one
30 minute to one hour. An individuals' diary varies in length from one to 15 days. This file
31 contains the following variables:
32
33 • The study, person, and diary day identifiers
34 • Start time of this activity
35 • Number of minutes for this activity
36 • Activity code (a record of what the individual was doing)
37 • Location code (a record of where the individual was)
38
39
40
100
-------
1 Table 46. Summary of activity pattern studies used in CHAD.
Study Name
Baltimore
California
Adolescents
and Adults
(GARB)
California
Children
(GARB)
Cincinnati
(EPRI)
Denver
(EPA)
Los Angeles:
Elementary
School
Children
Los Angeles:
High School
Adolescents
National:
NHAPS-Air
National:
NHAPS-
Water
Washington,
D.C. (EPA)
Location
A single
building in
Baltimore
California
California
Cincinnati
MSA
Denver
MSA
Los
Angeles
Los
Angeles
National
National
Wash. DC
MSA
Study
time
period
01/1997-
02/1997,
07/1 998-
08/1 998
1 0/1 987-
09/1 988
04/1 989-
02/1 990
03/1 985-
04/1985,
08/1 985
11/1982-
02/1 983
10/1989
09/1 990-
10/1990
09/1 992-
10/1994
09/1 992-
10/1994
11/1982-
02/1 983
Ages
72-93
12-17
18-94
0-11
0-86
18-70
10-12
13-17
0-93
0-93
18-98
Persons
26
181
1,552
1,200
888
432
17
19
4,326
4,332
639
Person
-days
292
181
1,552
1,200
2,587
791
51
42
4,326
4,332
639
Diary type
/study
design
Diary
Recall
/Random
Recall
/Random
Diary
/Random
Diary
/Random
Diary
Diary
Recall
/Random
Recall
/Random
Diary
/Random
Reference
Williams et al. (2000)
Robinson et al.
(1989);
Wiley etal. (1991 a)
Wiley etal. (1991b)
Johnson (1989)
Johnson (1984);
Akland etal. (1985)
Spier etal. (1992)
Spier etal. (1992)
Klepeis etal. (1996);
Tsang and Klepeis
(1996)
Klepeis etal. (1996);
Tsang and Klepeis
(1996)
Hartwell etal. (1984);
Akland etal. (1985)
2
3
4
5
6
1
8
9
10
11
12
13
14
15
16
17
18
3.3.5.3 Construction of Longitudinal Activity Sequences
Typical time-activity pattern data available for inhalation exposure modeling consist of a
sequence of location/activity combinations spanning a 24-hour duration, with 1 to 3 diary-days
for any single individual. Exposure modeling requires information on activity patterns over
longer periods of time, e.g., a full year. For example, even for pollutant health effects with short
averaging times (e.g., NC>2 1-hour average concentration) it may be desirable to know the
frequency of exceedances of a concentration over a long period of time (e.g., the annual number
of exceedances of a 1-hour average NC>2 concentration of 200 ppb for each simulated individual).
Long-term multi-day activity patterns can be estimated from single days by combining the
daily records in various ways, and the method used for combining them will influence the
variability of the long-term activity patterns across the simulated population. This in turn will
influence the ability of the model to accurately represent either long-term average high-end
exposures, or the number of individuals exposed multiple times to short-term high-end
concentrations.
101
-------
1 A common approach for constructing long-term activity patterns from short-term records is
2 to re-select a daily activity pattern from the pool of data for each day, with the implicit
3 assumption that there is no correlation between activities from day to day for the simulated
4 individual. This approach tends to result in long-term activity patterns that are very similar
5 across the simulated population. Thus, the resulting exposure estimates are likely to
6 underestimate the variability across the population, and therefore, underestimate the high-end
7 exposure concentrations or the frequency of exceedances.
8
9 A contrasting approach is to select a single activity pattern (or a single pattern for each
10 season and/or weekday-weekend) to represent a simulated individual's activities over the
11 duration of the exposure assessment. This approach has the implicit assumption that an
12 individual's day-to-day activities are perfectly correlated. This approach tends to result in long-
13 term activity patterns that are very different across the simulated population, and therefore may
14 over-estimate the variability across the population.
15 Cluster-Markov Algorithm
16 A new algorithm has been developed and incorporated into APEX to represent the day-to-
17 day correlation of activities for individuals. The algorithms first use cluster analysis to divide the
18 daily activity pattern records into groups that are similar, and then select a single daily record
19 from each group. This limited number of daily patterns is then used to construct a long-term
20 sequence for a simulated individual, based on empirically-derived transition probabilities. This
21 approach is intermediate between the assumption of no day-to-day correlation (i.e., re-selection
22 for each time period) and perfect correlation (i.e., selection of a single daily record to represent
23 all days).
24
25 The steps in the algorithm are as follows.
26 1. For each demographic group (age, gender, employment status), temperature range, and
27 day-of-week combination, the associated time-activity records are partitioned into 3
28 groups using cluster analysis. The clustering criterion is a vector of 5 values: the time
29 spent in each of 5 microenvironment categories (indoors - residence; indoors - other
30 building; outdoors - near road; outdoors - away from road; in vehicle).
31 2. For each simulated individual, a single time-activity record is randomly selected from
32 each cluster.
33 3. A Markov process determines the probability of a given time-activity pattern occurring
34 on a given day based on the time-activity pattern of the previous day and cluster-to-
35 cluster transition probabilities. The cluster-to-cluster transition probabilities are
36 estimated from the available multi-day time-activity records. If insufficient multi-day
37 time-activity records are available for a demographic group, season, day-of-week
38 combination, then the cluster-to-cluster transition probabilities are estimated from the
39 frequency of time-activity records in each cluster in the CHAD data base.
40
41 Details regarding the Cluster-Markov algorithm and supporting evaluations are provided in
42 Appendix F.
102
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
3.3.6 Calculating Microenvironmental Concentrations
Probabilistic algorithms are used to estimate the pollutant concentration associated with each
exposure event. The estimated pollutant concentrations account for the effects of ambient
(outdoor) pollutant concentration, penetration factor, air exchange rate, decay/deposition rate,
and proximity to microenvironments can use the transfer factors method while the others use the
mass balance emission sources, depending on the microenvironment, available data, and the
estimation method selected by the user.
APEX calculates air concentrations in the various microenvironments visited by the
simulated person by using the ambient air data for the relevant blocks, the user-specified
estimation method, and input parameters specific to each microenvironment. 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.
3.3.6.1 Mass Balance Model
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.
Table 47 lists the parameters required by the mass balance method to calculate
concentrations in a microenvironment. A proximity factor (fpr0ximity) is used to account for
differences in ambient concentrations between the geographic location represented by the
ambient air quality data (e.g., a regional fixed-site monitor or modeled concentration) and the
geographic location of the microenvironment (e.g., near a roadway). This factor could take a
value either greater than or less than 1. Emission source (ES) represents the emission rate for the
emission source and concentration source (CS) is the mean air concentration resulting from the
source. Rrem0va/ is defined as the removal rate of a pollutant from a microenvironment due to
deposition, filtration, and chemical reaction. The air exchange rate (^atr exchange) is expressed in
air changes per hour.
Table 47. Mass balance model parameters.
Variable
' proximity
CS
" removal
" air exchange
V
Definition
Proximity factor
Concentration source
Removal rate due to deposition,
filtration, and chemical reaction
Air exchange rate
Volume of microenvironment
Units
unitless
ppb
1/hr
1/hr
mj
Value Range
' proximity — "
CS>0
"removal — "
" air exchange — U
V>0
37
38
The mass balance equation for a pollutant in a microenvironment is described by:
103
-------
dt
-=ACM-AC0!ri-ACr,
eq(4)
2
O
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
29
30
31
32
removal
where:
dCME(t) = Change in concentration in a microenvironment at time t (ppb),
= Rate of change in microenvironmental concentration due to influx
of air (ppb/hour),
= Rate of change in microenvironmental concentration due to outflux
of air (ppb/hour),
= Rate of change in microenvironmental concentration due to
removal processes (ppb/hour), and
= Rate of change in microenvironmental concentration due to an
emission source inside the microenvironment (ppb/hour).
Within the time period of an hour each of the rates of change, AC!n, AC0!d, ACremova/, and
murce, is assumed to be constant. At each hour time step of the simulation period, APEX
estimates the hourly equilibrium, hourly ending, and hourly mean concentrations using a series
of equations that account for concentration changes expected to occur due to these physical
processes. Details regarding these equations are provided in the APEX User's Guide. APEX
reports hourly mean concentration as hourly concentration for a specific hour. The calculation
then continues to the next hour by using the end concentration for the previous hour as the initial
microenvironmental concentration. A description of the input parameters estimates used for
microenvironments using the mass balance approach is provided below.
3.3.6.2 Factors Model
The factors method is simpler than the mass balance method. It does not calculate
concentration in a microenvironment from the concentration in the previous hour and it has
fewer parameters. Table 48 lists the parameters required by the factors method to calculate
concentrations in a microenvironment without emissions sources.
28 Table 48. Factors model parameters.
Variable
' proximity
' penetration
Definition
Proximity factor
Penetration factor
Units
unitless
unitless
Value Range
' proximity — "
0 ^ f penetration - 1
The factors method uses the following equation to calculate hourly mean concentration in a
microenvironment from the user-provided hourly air quality data:
s~i hourlymean /~i r r
' ^v ambient J proximity J penetration
ME
eq(5)
33 where:
34
35
36
37
/~i hourfymeai
^ambient
Jproximity
Jpenetration
Hourly concentration in a microenvironment (ppb)
Hourly concentration in ambient environment (ppb)
Proximity factor (unitless)
Penetration factor (unitless)
104
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
The ambient NC>2 concentrations are from the air quality data input file. The proximity factor
is a unitless parameter that represents the proximity of the microenvironment to a monitoring
station. The penetration factor is a unitless parameter that represents the fraction of pollutant
entering a microenvironment from outside the microenvironment via air exchange. The
development of the specific proximity and penetration factors used in this analysis are discussed
below for each microenvironment using this approach.
3.3.6.3 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
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 are: 1) factors and 2) mass balance. A list of microenvironments used in this
study, the calculation method used, and the parameters used to calculate the microenvironment
concentrations can be found in Table 49.
Table 49. List of microenvironments and calculation methods used.
Microenvironment
No.
1
2
3
4
5
6
7
8
9
10
11
12
0
Name
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)
Not modeled
Calculation
Method
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Factors
Factors
Factors
Factors
Factors
Parameter
Types used 1
AERandDE
AERandDE
AER and DE
AERandDE
AERandDE
AERandDE
AERandDE
PR
PR
None
PE and PR
PE and PR
1 AER=air exchange rate, DE=decay-deposition rate, PR=proximity factor, PE=penetration
factor
Each of the microenvironments is designed to simulate an environment in which people
spend time during the day. CHAD locations are linked to the different microenvironments in the
Microenvironment Mapping File (see below). There are many more CHAD locations than
microenvironment locations (there are 113 CHAD codes versus 12 microenvironments in this
assessment), therefore most of the microenvironments have multiple CHAD locations mapped to
them.
3.3.6.4 Microenvironment Descriptions
Microenvironment 1: Indoor-Residence
105
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
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. The first two of these variables affect the air exchange
rate.
Since the selection of an air exchange rate distribution is conditioned on the presence or
absence of an air-conditioner, for each modeled area the air conditioning status of the residential
microenvironments is simulated randomly using the probability that a residence has an air
conditioner. For this study, location-specific air conditioning prevalence was taken from the
American Housing Survey of 2003 (AHS, 2003a; 2003b). Previous analyses (US EPA, 2007d)
detail the specification of uncertainty estimates in the form of confidence intervals for the air
conditioner prevalence using the following:
Standard Error (P) =
3850 P(l-P)
N '
Confidence Interval (P) = P +1 .96 x Standard Error (P)
where P is the estimated percentage and N is the estimated total number of housing units.
Table 50 contains the values for air conditioning prevalence used for each modeled location.
Table 50. Air conditioning prevalence estimates with 95% confidence intervals.
AHS
Survey
Atlanta
Detroit
Los Angeles
Philadelphia
Phoenix
Housing
Units
797,687
1,877,178
3,296,819
1,943,492
-
A/C
Prevalence
(%)
97.0
81.4
55.1
90.6
-
se
1.2
1.8
1.7
1.3
-
L95
94.7
78.0
51.7
88.1
-
U95
99.3
84.9
58.4
93.2
-
Notes:
se - Standard error
L95 - Lower limit on 95th confidence interval
U95 - Upper limit on 95th confidence interval
Air exchange rate data for the indoor residential microenvironment were obtained from US
EPA (2007d). Briefly, residential air exchange rate (AER) data were obtained from several
studies (Avol et al., 1998; Williams et al., 2003a, 2003b; Meng et al., 2004; Weisel et al., 2004;
Chillrud at al, 2004; Kinney et al., 2002; Sax et al., 2004; Wilson et al., 1986, 1996; Colome et
al., 1993, 1994; Murray and Burmaster, 1995). Influential characteristics (e.g., temperature, air
conditioning), where reported in the study, were also compiled for use in statistical analyses.
Descriptive statistics were generated for each location/variable type and evaluated using
statistical comparison testing (e.g., ANOVA). Based on the summary statistics and the statistical
comparisons, different AER distributions were fit for each combination of A/C type, city, and
temperature. In general, lognormal distributions provided the best fit, and are defined by a
geometric mean (GM) and standard deviation (GSD). To avoid unusually extreme simulated
AER values, bounds of 0.1 and 10 were selected for minimum and maximum AER, respectively.
106
-------
1
2
3
4
5
6
7
8
9
10
Fitted distributions were available for one of the cities modeled in this assessment, Los
Angeles. For the other 4 of the cities to be modeled, a distribution was selected from one of the
other locations thought to have similar characteristics to the city to be modeled, qualitatively
considering factors that might influence AERs. These factors include 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. The distributions used for these each of
the modeled locations are provide in Table 51.
Table 51. Geometric means (GM) and standard deviations (GSD) for air exchange rates by city, A/C
type, and temperature range.
Area
Modeled
Los Angeles
Philadelphia
and Detroit
Atlanta (No
A/C)
Atlanta (A/C)
Study City
Houston
Inland
California
Los Angeles
New York
City
Outside
California
Research
Triangle
Park, NC
A/C Type
Central or
Room A/C
No A/C
Central or
Room A/C
No A/C
Central or
Room A/C
No A/C
Central or
Room A/C
No A/C
Central or
Room A/C
No A/C
Central or
Room A/C
Temp
(°C)
<=20
20-25
25-30
>30
<=10
10-20
>20
<=25
>25
<=10
10-20
20-25
>25
<=20
20-25
25-30
>30
<=10
10-20
20-25
>25
<=10
10-25
>25
<=10
10-20
>20
<=10
10-20
20-25
25-30
>30
<=10
10-20
>20
<=10
10-20
20-25
>25
N
15
20
65
14
13
28
12
226
83
17
52
13
14
721
273
102
12
18
390
148
25
20
42
19
48
59
32
179
338
253
219
24
61
87
44
157
320
196
145
GM
0.4075
0.4675
0.4221
0.4989
0.6557
0.6254
0.9161
0.5033
0.8299
0.5256
0.6649
1.0536
0.8271
0.5894
1.1003
0.8128
0.2664
0.5427
0.7470
1.3718
0.9884
0.7108
1.1392
1 .2435
1.0165
0.7909
1 .6062
0.9185
0.5636
0.4676
0.4235
0.5667
0.9258
0.7333
1.3782
0.9617
0.5624
0.3970
0.3803
GSD
2.1135
1.9381
2.2579
1.7174
1 .6794
2.9162
2.4512
1.9210
2.3534
3.1920
2.1743
1.7110
2.2646
1.8948
2.3648
2.4151
2.7899
3.0872
2.0852
2.2828
1 .9666
2.0184
2.6773
2.1768
2.1382
2.0417
2.1189
1.8589
1.9396
2.2011
2.0373
1 .9447
2.0836
2.3299
2.2757
1.8094
1.9058
1.8887
1.7092
11
107
-------
1 For this analysis, the same NC>2 removal rate distribution was used for all microenvironments
2 that use the mass balance method. This removal rate is based on data provided by Spicer et al.
3 (1993). A total of 6 experiments, under variable source emission characteristics including
4 operation of gas stove, were conducted in an unoccupied test house. A distribution could not be
5 described with the limited data set, therefore a uniform distribution was approximated by the
6 bounds of the 6 values, a minimum of 1.02 and a maximum of 1.45 h"1.
7 An excerpt from the APEX input file describing the indoor residential microenvironment is
8 provided in Figure 19. The first section of the input file excerpt specifies the air exchange rate
9 distributions for the microenvironment. Average temperature and air conditioning presence,
10 which are city-specific, were coded into air exchange rate conditional variables, Cl and C2,
11 respectively. Average temperatures were separated into five categories (variable Cl, numbered
12 1-5): 50 ° F, 50-68 ° F, 68-77 ° F, 77-86 ° F, and 86 ° F and above. For variable C2, air
13 conditioning status can range from 7 to 2 (7 for having air conditioning, 2 for not having it). The
14 air exchange rate estimates generated previously in the form of lognormal distributions were
15 entered into the appropriate temperature and A/C category for each location for a total often
16 distributions (i.e., 5 temperature distributions by 2 air conditioning distributions). In the input
17 file example however, there are actually four AER distributions for homes with an air
18 conditioner and three for those without; the last few distributions for each air conditioning setting
19 were the same due to the available data to populate the field. The parameter estimates for the
20 removal factor (DE) is also shown following the AER data.
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51 Figure 19. Example input file from APEX for Indoors-residence microenvironment.
108
Micro number
Parameter Type
Condition # 1
Condition #2
ResampHours
ResampDays
ResampWork
= 1 ! Indoors -
= AER
= AvgTempCat
= AC Home
= NO
= YES
= YES
Block DType Season Area
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
Micro number
Pollutant = 1
Parameter Type
ResampHours
ResampDays
ResampWork
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
= 1
= DE
= NO
= NO
= YES
Block DType Season Area
1 1
1 1
C1
1
2
3
4
5
1
2
3
4
5
C2
1
1
1
1
1
2
2
2
2
2
C3
1
1
1
1
1
1
1
1
1
1
residence -
Shape
Lognormal
Lognormal
Lognormal
Lognormal
Lognormal
Lognormal
Lognormal
Lognormal
Lognormal
Lognormal
AIR EXCHANGE RATES
Par1
0.711
1.139
1.139
1.244
1.244
1.016
0.791
1.606
1.606
1.606
Par2 Par3 Par4 LTrunc UTrunc
2
2
2
2
2
2
2
2
2
2
018
677
677
177
177
138
042
119
119
119
0
0
0 .
0 .
0 .
0 .
0 .
0 .
0 .
0 .
0
0
0
0
0
0
0
0
0
0
.1 10
.1 10
.1 10
.1 10
.1 10
.1 10
.1 10
.1 10
.1 10
.1 10
! DECAY RATES
C1
1
C2
1
C3
1
Shape
Uniform
Par1 Par2 ParS Par4
1.02
1.45
LTrunc
1.02
UTrunc
1.45
-------
1 Indoor source contributions. A number of studies, as described in section 2.5.5 of the NOX
2 ISA, have noted the importance of gas cooking appliances as sources of NCh emissions. An
3 indoor emission source term was included in the APEX simulations to estimate exposure to
4 indoor sources of NO2. Three types of data were used to implement this factor:
5 • The fraction of households in the Philadelphia MSA that use gas for cooking fuel
6 • The range of contributions to indoor NO2 concentrations that occur from cooking
7 with gas
8 • The diurnal pattern of cooking in households.
9
10 The fraction of households in Philadelphia County that use gas cooking fuel (i.e., 55%) was
11 taken from the US Census Bureau's American Housing Survey for the Philadelphia Metropolitan
12 Area: 2003.
13
14 Data used for estimating the contribution to indoor NO2 concentrations that occur during
15 cooking with gas fuel were derived from a study sponsored by the California Air Resources
16 Board (CARB, 2001). For this study a test house was set up for continuous measurements of
17 NO2 indoors and outdoors, among several other parameters, and conducted under several
18 different cooking procedures and stove operating conditions. A uniform distribution of
19 concentration contributions for input to APEX was estimated as follows.
20
21 • The concurrent outdoor NO2 concentration measurement was subtracted from each
22 indoor concentration measurement, to yield net indoor concentrations
23 • Net indoor concentrations for duplicate cooking tests (same food cooked the same
24 way) were averaged for each indoor room, to yield average net indoor concentrations
25 • The minimum and maximum average net indoor concentrations for any test in any
26 room were used as the lower and upper bounds of a uniform distribution
27
28 This resulted in a minimum average net indoor concentration of 4 ppb and a maximum net
29 average indoor concentration of 188 ppb.
30
31 An analysis by Johnson et al (1999) of survey data on gas stove usage collected by Koontz et
32 al (1992) showed an average number of meals prepared each day with a gas stove of 1.4. The
33 diurnal allocation of these cooking events was estimated as follows.
34
35 • Food preparation time obtained from CHAD diaries was stratified by hour of the day,
36 and summed for each hour, and summed for total preparation time.
37 • The fraction of food preparation occurring in each hour of the day was calculated as
38 the total number of minutes for that hour divided by the overall total preparation time.
39 The result was a measure of the probability of food preparation taking place during
40 any hour, given one food preparation event per day.
41 • Each hourly fraction was multiplied by 1.4, to normalize the expected value of daily
42 food preparation events to 1.4.
43 The estimated probabilities of cooking by hour of the day are presented in Table 52. For
44 this analysis it was assumed that the probability that food preparation would include stove usage
45 was the same for each hour of the day, so that the diurnal allocation of food preparation events
46 would be the same as the diurnal allocation of gas stove usage. It was also assumed that each
109
-------
1
2
3
4
cooking event lasts for exactly 1 hour, implying that the average total daily gas stove usage is 1.4
hours.
Table 52. Probability of gas stove cooking by hour of the day.
Hour of Day
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Probability of
Cooking (%)1
0
0
0
0
0
5
10
10
10
5
5
5
10
5
5
5
15
20
15
10
5
5
0
0
Values rounded to the nearest 5%. Data
to 1 45% due to rounding and scaling to 1 .
cooking events/day.
sum
4
6 Microenvironments 2-7: All other indoor microenvironments
1 The remaining five indoor microenvironments, which represent Bars and Restaurants,
8 Schools, Day Care Centers, Office, Shopping, and Other environments, are all modeled using the
9 same data and functions (Figure 20). As with the Indoor-Residence microenvironment, these
10 microenvironments use both air exchange rates and removal rates to calculate exposures within
11 the microenvironment. The air exchange rate distribution (GM = 1.109, GSD = 3.015, Min =
12 0.07, Max = 13.8) was developed based on an indoor air quality study (Persily et al, 2005; see
13 US EPA, 2007d for details in derivation). The decay rate is the same as used in the Indoor-
14 Residence microenvironment discussed previously. The Bars and Restaurants microenvironment
15 included an estimated contribution from indoor sources as was described for the Indoor-
16 Residence, only there was an assumed 100% prevalence rate and the cooking with the gas
17 appliance occurred at any hour of the day.
110
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19 Figure 20. Example input file from APEX for all Indoors microenvironments, other than Indoors-
20 residence.
21 Microenvironments 8 and 9: Outdoor microenvironments
22 Two outdoor microenvironments, the Near Road and Public Garage/Parking Lot, used the
23 factors method to calculate pollutant exposure. Penetration factors are not applicable to outdoor
24 environments (effectively, PEN=1). Proximity factors were developed from the AERMOD
25 concentration predictions, i.e., the block-centroid-to-nearest-roadway concentration ratios. Based
26 on the resulting sets of ratio values, the ratio distributions were stratified by hour of the day into
27 3 groups as indicated by the "hours-block" specification in the example file in Figure 21. The
28 lower and upper bounds for sampling were specified as the 5th and 95th percentile values,
29 respectively, of each distribution.
Micro number = 2 ! Bars & restaurants - AIR EXCHANGE RATES
Parameter Type = AER
ResampHours = NO
ResampDays = YES
ResampWork = YES
Block DType Season Area C1
1 1 111
Micro number = 2
Pollutant = 1
Parameter Type = DE
ResampHours = NO
ResampDays = YES
ResampWork = YES
Block DType Season Area C1
11111
C2 C3
1 1
Shape Par1 Par2 Par3 Par4
LogNormal 1.109 3.015 0
LTrunc UTrunc
0.07 13.8
! DECAY RATES
C2 C3
1 1
Shape Par1 Par2 Par3 Par4
Uniform 1.02 1.45 .
LTrunc UTrunc
1.02 1.45
30
31
32
33
34
35
36
37
38
39
40
41
42
43 Figure 21. Example input file from APEX for outdoor near road microenvironment.
44 Microenvironment 10: Outdoors-General.
45 The general outdoor environment concentrations are well represented by the modeled
46 concentrations. Therefore, both the penetration factor and proximity factor for this
47 microenvironment were set to 1.
Micro number =8 ! Outdoor near road PROXIMITY FACTOR
Pollutant = 1
Parameter Type = PR
Hours - Block = 111111222222222222233311
ResampHours = YES
ResampDays = YES
ResampWork = YES
Block DType Season Area C1 C2 C3 Shape Par1 Par2 Par3 Par4 LTrunc UTrunc ResampOut
111 1111 LogNormal 1.251 1.478 0. . 0.86 2.92 Y
211 1111 LogNormal 1.555 1.739 0. . 0.83 4.50 Y
311 1111 LogNormal 1.397 1.716 0. . 0.73 4.17 Y
111
-------
1 Microenvironments 11 and 12: In Vehicle- Cars and Trucks, and Mass Transit
2 Penetration factors were developed from data provided in Chan and Chung (2003). Inside-
3 vehicle and outdoor NC>2 concentrations were measured with for three ventilation conditions, air-
4 recirculation, fresh air intake, and with windows opened. Since major roads were the focus of
5 this assessment, reported indoor/outdoor ratios for highway and urban streets were used here.
6 Mean values range from about 0.6 to just over 1.0, with higher values associated with increased
7 ventilation (i.e., window open). A uniform distribution was selected for the penetration factor
8 for Inside-Cars/Trucks (ranging from 0.6 to 1.0) due to the limited data available to describe a
9 more formal distribution and the lack of data available to reasonably assign potentially
10 influential characteristics such as use of vehicle ventilation systems for each location. Mass
11 transit systems, due to the frequent opening and closing of doors, was assigned a uniform
12 distribution ranging from 0.8 to 1.0 based on the reported mean values for fresh air intake and
13 open windows.
14
15 Proximity factors were developed as described above for Microenvironments 8 and 9
16
17 3.3.6.5 Microenvironment Mapping
18 The Microenvironment Mapping file matches the APEX Microenvironments to CHAD
19 Location codes. Table 53 gives the mapping used for the APEX simulations.
20 Table 53. Mapping of CHAD activity locations to APEX microenvironments.
CHAD LOG.
U
X
30000
30010
30020
30100
30120
30121
30122
30123
30124
30125
30126
30127
30128
30129
30130
30131
30132
30133
30134
30135
30136
30137
30138
30139
30200
30210
30211
30219
30220
Description
Uncertain of correct code
No data
Residence, general
Your residence
Other residence
Residence, indoor
Your residence, indoor
. . . , kitchen
. . . , living room or family room
. . . , dining room
. . . , bathroom
. . . , bedroom
. . . , study or office
. . . , basement
. . . , utility or laundry room
. . . , other indoor
Other residence, indoor
. . . , kitchen
. . . , living room or family room
. . . , dining room
. . . , bathroom
. . . , bedroom
. . . , study or office
. . . , basement
. . . , utility or laundry room
. . . , other indoor
Residence, outdoor
Your residence, outdoor
. . . , pool or spa
. . . , other outdoor
Other residence, outdoor
APEX
-1
-1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
10
10
10
10
10
micro
Unknown
Unknown
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Indoors -Residence
Outdoors -Other
Outdoors -Other
Outdoors -Other
Outdoors -Other
Outdoors -Other
112
-------
30221 ..., pool or spa = 10
30229 ..., other outdoor = 10
30300 Residential garage or carport = 7
30310 ..., indoor = 7
30320 ..., outdoor = 10
30330 Your garage or carport = 1
30331 ..., indoor = 1
30332 ..., outdoor = 10
30340 Other residential garage or carport = 1
30341 ..., indoor = 1
30342 ..., outdoor = 10
30400 Residence, none of the above = 1
31000 Travel, general = 11
31100 Motorized travel = 11
31110 Car = 11
31120 Truck = 11
31121 Truck (pickup or van) = 11
31122 Truck (not pickup or van) = 11
31130 Motorcycle or moped = 8
31140 Bus = 12
31150 Train or subway = 12
31160 Airplane = 0
31170 Boat = 10
31171 Boat, motorized = 10
31172 Boat, other = 10
31200 Non-motorized travel = 10
31210 Walk = 10
31220 Bicycle or inline skates/skateboard = 10
31230 In stroller or carried by adult = 10
31300 Waiting for travel = 10
31310 •••, bus or train stop = 8
31320 ..., indoors = 7
31900 Travel, other = 11
31910 ..., other vehicle = 11
32000 Non-residence indoor, general = 7
32100 Office building/ bank/ post office = 5
32200 Industrial/ factory/ warehouse = 5
32300 Grocery store/ convenience store = 6
32400 Shopping mall/ non-grocery store = 6
32500 Bar/ night club/ bowling alley = 2
32510 Bar or night club = 2
32520 Bowling alley = 2
32600 Repair shop = 7
32610 Auto repair shop/ gas station = 7
32620 Other repair shop = 7
32700 Indoor gym /health club = 7
32800 Childcare facility = 4
32810 ..., house = 1
32820 ..., commercial = 4
32900 Large public building = 7
32910 Auditorium/ arena/ concert hall = 7
32920 Library/ courtroom/ museum/ theater = 7
33100 Laundromat = 7
33200 Hospital/ medical care facility = 7
33300 Barber/ hair dresser/ beauty parlor = 7
33400 Indoors, moving among locations = 7
33500 School = 3
33600 Restaurant = 2
33700 Church = 7
33800 Hotel/ motel = 7
33900 Dry cleaners = 7
34100 Indoor parking garage = 7
34200 Laboratory 7
Outdoors-Other
Outdoors-Other
Indoors-Other
Indoors-Other
Outdoors-Other
Indoors-Residence
Indoors-Residence
Outdoors-Other
Indoors-Residence
Indoors-Residence
Outdoors-Other
Indoors-Residence
In Vehicle-Cars_and_Trucks
In Vehicle-Cars_and_Trucks
In Vehicle-Cars_and_Trucks
In Vehicle-Cars_and_Trucks
In Vehicle-Cars_and_Trucks
In Vehicle-Cars_and_Trucks
Outdoors-Near_Road
In Vehicle-Mass_Transit
In Vehicle-Mass_Transit
Zero_concentration
Outdoors-Other
Outdoors-Other
Outdoors-Other
Outdoors-Other
Outdoors-Other
Outdoors-Other
Outdoors-Other
Outdoors-Other
Outdoors-Near_Road
Indoors-Other
In Vehicle-Cars_and_Trucks
In Vehicle-Cars_and_Trucks
Indoors-Other
Indoors-Office
Indoors-Office
Indoors-Shopping
Indoors-Shopping
Indoors-Bars_and_Restaurants
Indoors-Bars_and_Restaurants
Indoors-Bars_and_Restaurants
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Day_Care_Centers
Indoors-Residence
Indoors-Day_Care_Centers
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Schools
Indoors-Bars_and_Restaurants
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
Indoors-Other
113
-------
34300 Indoor, none of the above = 7 Indoors-
35000 Non-residence outdoor, general = 10 Outdoors
35100 Sidewalk, street = 8 Outdoors
35110 Within 10 yards of street = 8 Outdoors
35200 Outdoor public parking lot /garage = 9 Outdoors
35210 •••, public garage = 9 Outdoors
35220 ..., parking lot = 9 Outdoors
35300 Service station/ gas station = 10 Outdoors
35400 Construction site = 10 Outdoors
35500 Amusement park = 10 Outdoors
35600 Playground = 10 Outdoors
35610 ••-, school grounds = 10 Outdoors
35620 ••-, public or park = 10 Outdoors
35700 Stadium or amphitheater = 10 Outdoors
35800 Park/ golf course = 10 Outdoors
35810 Park = 10 Outdoors
35820 Golf course = 10 Outdoors
35900 Pool/ river/ lake = 10 Outdoors
36100 Outdoor restaurant/ picnic = 10 Outdoors
36200 Farm = 10 Outdoors
36300 Outdoor, none of the above = 10 Outdoors
Other
-Other
-Near_Road
-Near_Road
-Public_Garage-Parking
-Public_Garage-Parking
-Public_Garage-Parking
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
-Other
3
4
5
6
7
8
9
10
11
3.3.7 Exposure Calculations
APEX calculates exposure as a 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:
•» hourly-mean
'ME(j)
c,=
_ 7=1
T
eq(6)
12
13
14
15
16
17
18
19
20
21
22
23
24
25
where:
Ct
N
T
Hourly exposure concentration at clock hour /' of the simulation period
Number of events (i.e., microenvironments visited) in clock hour / of
the simulation period.
Hourly mean concentration in microenvironment y (ppm)
Time spent in microenvironment y' (minutes)
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:
114
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
counts of the estimated number of people exposed to 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 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 200
to 300 ppb by 50 ppb increments for 1-hour average exposures. These results are tabulated for
the population and subpopulations of interest.
To simulate just meeting the current standard, dispersion modeled concentration were not
rolled-up as done in the air quality characterization. A proportional approach was used as done
in the Air Quality Characterization, but to reduce processing time, the health effect benchmark
levels were proportionally reduced by the similar factors described for each specific location and
simulated year. 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
difference in the exposure and risk modeling was that the modeled air quality concentrations
were used to generate the adjustment factors. Table 54 provides the adjustment factors used and
the adjusted potential health effect benchmark concentrations to simulate just meeting the current
standard. When modeling indoor sources, the indoor concentration contributions needed to be
scaled downward by the same proportions.
Table 54. Adjustment factors and potential health effect benchmark levels used by APEX to simulate just
meeting the current standard
Simulated
Year
(factor)
2001
(1.59)
2002
(1.63)
2003
(1.64)
Potential Health
Effect Benchmark
Level (ppb)
Actual
150
200
250
300
150
200
250
300
150
200
250
300
Adjusted
94
126
157
189
92
122
153
184
91
122
152
183
27 3.3.8 Exposure Model Output
28 All of the output files written by APEX are ASCII text files. Table 55 lists each of the output
29 data files written for these simulations and provides descriptions of their content. Additional
115
-------
1 output files that can produced by APEX are given in Table 5-1 of the APEX User's Guide, and
2 include hourly exposure, ventilation, and energy expenditures, and even detailed event-level
3 information, if desired. The names and locations, as well as the output table levels (e.g., output
4 percentiles, cut-points), for these output files are specified by the user in the simulation control
5 parameters file.
6 Table 55. Example of APEX output files.
Output File Type
Log
Profile Summary
Microenvironment
Summary
Sites
Output Tables
Description
The Log file contains the record of the APEX model simulation as it progresses.
If the simulation completes successfully, the log file indicates the input files and
parameter settings used for the simulation and reports on a number of different
factors. If the simulation ends prematurely, the log file contains error messages
describing the critical errors that caused the simulation to end.
The Profile Summary file provides a summary of each individual modeled in the
simulation.
The Microenvironment Summary file provides a summary of the time and
exposure by microenvironment for each individual modeled in the simulation.
The Sites file lists the tracts, districts, and zones in the study area, and identifies
the mapping between them.
The Output Tables file contains a series of tables summarizing the results of the
simulation. The percentiles and cut-off points used in these tables are defined
in the simulation control parameters file.
116
-------
i 3.4 Exposure Modeling Results
2 3.4.1 Overview
3 The results of the exposure and risk characterization are presented here for Philadelphia
4 County. Several scenarios were considered for the exposure assessment, including two
5 averaging time for NC>2 concentrations (annual and 1-hour), inclusion of indoor sources, and for
6 evaluating just meeting the current standard. To date, year 2002 served as the base year for all
7 scenarios, years 2001 and 2003 were only evaluated for a limited number of scenarios.
8 Exposures were simulated for four groups; children and all persons, and the asthmatic population
9 within each of these.
10
11 The exposure results summarized below focus on the population group where exposure
12 estimations are of greatest interest, namely asthmatic individuals. Complete results for each of
13 these two population subgroups is provided in Appendix G. However, due to certain limitations
14 in the data summaries output from APEX, some exposure data could only be output for the entire
15 population modeled (i.e., all persons - includes asthmatics and healthy persons of all ages). The
16 summary data for the entire population (e.g., annual average exposure concentrations, time spent
17 in microenvironments at or above a potential health effect benchmark level) can be
18 representative of the asthmatic population since the asthmatic population does not have its
19 microenvironmental concentrations and activities estimated any differently from those of the
20 total population.
21 3.4.2 Annual Average Exposure Concentrations (as is)
22 Since the current NC>2 standard is 0.053 ppm annual average, the predicted air quality
23 concentrations, the measured ambient monitoring concentrations, and the estimated exposures
24 were summarized by annual average concentration. The distribution for the AERMOD predicted
25 NO2 concentrations at each of the 16,857 receptors for years 2001 through 2003 are illustrated in
26 Figure 22. Variable concentrations were estimated by the dispersion model over the three year
27 period (2001-2003). The NO2 concentration distribution was similar for years 2001 and 2002,
28 with mean annual average concentrations of about 21 ppb and a COV of just over 30%. On
29 average, NO2 annual average concentrations were lowest during simulated year 2003 (mean
30 annual average concentration was about 16 ppb), largely a result of the comparably lower local
31 concentration added (Table 26). While the mean annual average concentrations were lower than
32 those estimated for 2001 and 2002, a greater number of annual average concentrations were
33 estimated above 53 ppb for year 2003. In addition, year 2003 also contained greater variability
34 in annual average concentrations as indicated by a COV of 53%.
117
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
200
"o. 180
c
01
u
c
o
O
tf
140 -
120 -
100 -
01
0)
£
I
CO
I 80
c
•5
Q
O
o:
UJ
40 -
20 -
-5
-3 -2-101
Normal Quantile
Figure 22 . Distribution of AERMOD predicted annual average NO2 concentrations at each of the 16,857
receptors in Philadelphia County for years 2001-2003
The hourly concentrations output from AERMOD were input into the exposure model,
providing a range of estimated exposures output by APEX. Figure 23 illustrates the annual
average exposure concentrations for the entire simulated population (both asthmatics and healthy
individual of all ages), for each of the years analyzed and where indoor sources were modeled.
While years 2001 and 2002 contained very similar population exposure concentration
distributions, the modeled year 2003 contained about 20% lower annual average concentrations.
The lower exposure concentrations for year 2003 are similar to what was observed for the
predicted air quality (Figure 22), however, all persons were estimated to contain exposures
below an annual average concentration of 53 ppb, even considering indoor source concentration
contributions. Again, while Figure 23 summarizes the entire population, the data are
representative of what would be observed for the population of asthmatics or asthmatic children.
118
-------
100
90 -
80 -
70 -
£
o
01
Q.
c
o
JS
3
Q.
O
Q.
60 -
2001 with indoor sources
2002 with indoor sources
2003 with indoor sources
15 20 25 30
Annual Average NO2 Exposure (ppb)
2 Figure 23. Estimated annual average total NO2 exposure concentrations for all simulated persons in
3 Philadelphia County, using modeled 2001-2003 air quality (as is), with modeled indoor sources.
4
5 The AERMOD predicted air quality and the estimated exposures for year 2002 were
6 compared using their respective annual average NC>2 concentrations (Figure 24). As a point of
7 reference, the annual average concentration for 2002 ambient monitors ranged from 24 ppb to 29
8 ppb. Many of the AERMOD predicted annual average concentrations were below that of the
9 lowest ambient monitoring concentration of 24 ppb, although a few of the receptors contained
10 concentrations above the highest measured annual average concentration . Estimated exposure
11 concentrations were below that of both the modeled and measured air quality. For example,
12 exposure concentrations were about 5 ppb less than the modeled air quality when the exposure
13 estimation included indoor sources, and about 10 ppb less for when exposures were estimated
14 without indoor sources. In comparing the estimated exposures with and without indoor sources,
15 indoor sources were estimated to contribute between 1 and 5 ppb to the total annual average
16 exposures.
119
-------
100
90 -
80 -
70 -
60 -
01
Q_
50 -
40 -
30 -
20 -
10 -
AERMOD Predicted 2002 air quality (as is)
APEX Exposure 2002 no indoor sources
APEX Exposure 2002 with indoor sources
10 15 20 25 30
Annual Average NO2 Cocentrations (ppb)
35
40
45
2 Figure 24. Comparison of AERMOD predicted and ambient monitoring annual average NO2
3 concentrations (as is) and APEX exposure concentrations (with and without modeled indoor sources) in
4 Philadelphia County for year 2002.
5 3.4.3 One-Hour Exposures (as is)
6 Since there is interest in short-term exposures, a few analyses were performed using the
7 APEX estimated exposure concentrations. As part of the standard analysis, APEX reports the
8 maximum exposure concentration for each simulated individual in the simulated population.
9 This can provide insight into the proportion of the population experiencing any NC>2 exposure
10 concentration level of interest. In addition, exposures are estimated for each of the selected
11 potential health effect benchmark levels (200, 250, and 300 ppb, 1-hour average). An
12 exceedance was recorded when the maximum exposure concentration observed for the individual
13 was above the selected level in a day (therefore, the maximum number of exceedances is 365 for
14 a single person). Estimates of repeated exposures are also recorded, that is where 1-hour
15 exposure concentrations were above a selected level in a day added together across multiple days
16 (therefore, the maximum number of multiple exceedances is also 365). Persons of interest in this
17 exposure analysis are those with particular susceptibility to NC>2 exposure, namely individuals
18 with asthma. The health effect benchmark levels are appropriate for estimating the potential risk
19 of adverse health effects for asthmatics. The majority of the results presented in this section are
20 for the simulated asthmatic population. However, the exposure analysis was performed for the
21 total population to assess numbers of persons exposed to these levels and to provide additional
22 information relevant to the asthmatic population (such as time spent in particular
23 microenvironments), although most of the results for the total population are reported in Chapter
24 3oftheTSD.
25
120
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
3.4.3.1 Maximum Estimated Exposure Concentrations
A greater variability was observed in maximum exposure concentrations for the 2003 year
simulation compared with years 2001 and 2002 (Figure 25). While annual average exposure
concentrations for the total population were the lowest of the 3-year simulation, year 2003
contained a greater number of individual maximum exposures at and above the lowest potential
health effect benchmark level. When indoor sources are not modeled however, over 90% of the
simulated persons do not have an occurrence of a 1-hour exposure above 200 ppb in a year.
3.4.3.2 Number of Estimated Exposures above Selected Levels
When considering the total asthmatic population simulated in Philadelphia County and using
current air quality of 2001-2003, nearly 50,000 persons were estimated to be exposed at least one
time to a one-hour concentration of 200 ppb in a year (Figure 26). These exposures include both
the NC>2 of ambient origin and that contributed by indoor sources. The number of asthmatics
exposed to greater concentrations (e.g., 250 or 300 ppb) drops dramatically and is estimated to be
somewhere between 1,000 - 15,000 depending on the 1-hour concentration level and the year of
air quality data used. Exposures simulated for year 2003 contained the greatest number of
asthmatics exposed in a year consistently for all potential health effect benchmark levels, while
year 2002 contained the lowest number of asthmatics. Similar trends across the benchmark
levels and the simulation years were observed for asthmatic children, albeit with lower numbers
of asthmatic children with exposures at or above the potential health effect benchmark levels.
100
c
HI
o
HI
Q.
c
o
3
Q.
O
Q.
90 -
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
«2001 with indoor sources
o 2002 with indoor sources
A 2003 with indoor sources
x 2002 no indoor sources
50 100 150 200 250 300 350
Maximum 1-hour Exposure (ppb) in a Year
400
450
500
Figure 25. Estimated maximum NO2 exposure concentration for all simulated persons in Philadelphia
County, using modeled 2001-2003 air quality (as is), with and without modeled indoor sources. Values
above the 99th percentile are not shown
121
-------
6.0E+4
1
2
3
4
5
6
1
O.OE+0
200
250
300
2003 AQ (as is) - with indoor souces
2002 AQ (as is) - with indoor sources
2001 AQ (as is) - with indoor sources
Simulated Year - Scenario
Potential Health Effect Benchmark Level (ppb)
Figure 26. Estimated number of all simulated asthmatics in Philadelphia County with at least one NO2
exposure at or above the potential health effect benchmark levels, using modeled 2001-2003 air quality
(as is), with modeled indoor sources.
1.4E+4
O.OE+0
200
300
Potential Health Effect Benchmark Level (ppb)
2003 AQ (as is) - with indoor souces
2002 AQ (as is) - with indoor sources
2001 AQ (as is) - with indoor sources
Simulated Year - Scenario
Figure 27. Estimated number of simulated asthmatic children in Philadelphia County with at least one
NO2 exposure at or above the potential health effect benchmark levels, using modeled 2001-2003 air
quality (as is), with modeled indoor sources.
122
-------
5.0E+4-
200
Potential Health Effect Benchmark Level (ppb)
300
2002 AQ (as \s) - with indoor sources
2002 AQ (as is) - no indoor sources
Simulated Year - Scenario
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Figure 28. Comparison of the estimated number of all simulated asthmatics in Philadelphia County with
at least one NO2 exposure at or above potential health effect benchmark levels, using modeled 2002 air
quality (as is), with and without modeled indoor sources.
For example, nearly 12,000 were estimated to be exposed to at least a one-hour NC>2
concentration of 200 ppb in a year (Figure 27). Additional exposure estimates were generated
using the modeled 2002 air quality (as is) and where the contribution from indoor sources was
not included in the exposure concentrations. 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 28 compares 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 concentrations is reduced by between 50-80%, depending on
benchmark level, when not including indoor source (i.e., gas cooking) concentration
contributions.
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 is summarized to 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) since their
microenvironmental concentrations and activities are not estimated any differently from those of
the total population by APEX.
123
-------
1 As an example, Figure 29 (a, b, c) summarizes the percent of total time spent in each
2 microenvironment for simulation year 2002 that was associated with estimated exposure
3 concentrations at or above 200, 250, and 300 ppb (results for years 2001 and 2003 were similar).
4 Estimated exposures included the contribution from one major category of indoor sources (i.e.,
5 gas cooking). The time spent in the indoor residence and bars/restaurants were the most
6 important for concentrations >200 ppb, contributing to approximately 75% of the time persons
7 were exposed (Figure 29a). This is likely a result of the indoor source concentration contribution
8 to each individual's exposure concentrations. The importance of the particular
9 microenvironment however changes with differing potential health effect benchmark levels.
10 This is evident when considering the in-vehicle and outdoor near-road microenvironments,
11 progressing from about 19% of the time exposures were at the lowest potential health effect
12 benchmark level (200 ppb) to a high of 64% of the time exposures were at the highest
13 benchmark level (300 ppb, Figure 29c).
14
15 The microenvironments where higher exposure concentrations occur were also evaluated for
16 the exposure estimates generated without indoor source contributions. Figure 30 illustrates that
17 the time spent in the indoor microenvironments contributes little to the estimated exposures
18 above the selected benchmark levels. The contribution of these microenvironments varied only
19 slightly with increasing benchmark concentration, ranging from about 2-5%. Most of the time
20 associated with high exposures was associated with the transportation microenvironments (In-
21 Vehicle or In-Public Transport) or outdoors (Out-Near Road, Out-Parking Lot, Out-Other). The
22 importance of time spent outdoors near roadways exhibited the greatest change in contribution
23 with increased health benchmark level, increasing from around 30 to 44% of time associated
24 with concentrations of 200 and 300 ppb, respectively
25
124
-------
1
2
3
4
In-Public Trans
\Other
Out-Other
Out-Parking Lot
Out-Near Road
In-Other
In-Shopping—7
In-Office-/
In-Day Care—'
In-School—'
In-Bar & Restaurant
In-Residence
a) > 200 ppb
In-Public Trans
k Other
In-Residence
Out-Other
Out-Parking Lot
In-Bar & Restaurant
b) > 250 ppb
In-Bar & Restaurant
In-Public Trans |n.Residence J Hn-School
Out-Other
Out-Near Road
Out-Parking Lot
c) > 300 ppb
Figure 29. Fraction of time all simulated persons in Philadelphia County spend in the twelve
microenvironments associated with the potential NO2 health effect benchmark levels, a) > 200 ppb, b) >
250 ppb, and c) > 300 ppb, year 2002 simulation with indoor sources.
125
-------
1
2
3
4
In-Bar& Restaurant
In-Residence-
Other
In-Public Trans—
In-Vehicle
Out-Near Road
Out-Parking Lot
Out-Other
a) > 200 ppb
In-Bar& Restaurant /—In-School
In-Vehicle
Out-Near Road
Out-Other
Out-Parking Lot
b) > 250 ppb
In-Bar& Restaurant,-In-School
In-Residence^k //~tn-°al9.are
In-Public
In-Vehicle
Out-Near Road
Out-Other
Out-Parking Lot
c) > 300 ppb
Figure 30. Fraction of time all simulated persons in Philadelphia County spend in the twelve
microenvironments associated with the potential NO2 health effect benchmark levels, a) > 200 ppb, b) >
250 ppb, and c) > 300 ppb, year 2002 simulation without indoor sources
126
-------
1 3.4.3.3 Number of Repeated Exposures Above Selected Levels
2 In the analysis of persons exposed, the results show the number or percent of those with at
3 least one exposure at or above the selected potential health effect benchmark level. Given that
4 the benchmark is for a small averaging time (i.e., one-hour) it may be possible that individuals
5 are exposed to concentrations at or above the potential health effect benchmark levels more than
6 once in a given year. Since APEX simulates the longitudinal diary profile for each individual,
7 the number of times above a selected level is retained for each person. Figure 31 presents such
8 an analysis for the year 2003, the year containing the greatest number of exposure concentrations
9 at or above the selected benchmarks. Estimated exposures include both those resulting from
10 exposures to NC>2 of ambient origin and those resulting from indoor source NC>2 contributions.
11 While a large fraction of individuals experience at least one exposure to 200 ppb or greater over
12 a 1-hour time period in a year (about 32 percent), only around 14 percent were estimated to
13 contain at least 2 exposures. Multiple exposures at or above the selected benchmarks greater
14 than or equal to 3 or more times per year are even less frequent, with around 5 percent or less of
15 asthmatics exposed to 1-hour concentrations greater than or equal to 200 ppb 3 or more times in
16 a year.
17
18 Exposure estimates for year 2002 are presented to provide an additional perspective,
19 including a lower bound of repeated exposures for this population subgroup and for exposure
20 estimates generated with and without modeled indoor sources (Figure 32). Most asthmatics
21 exposed to a 200 ppb concentration are exposed once per year and only around 11 percent would
22 experience 2 or more exposures at or above 200 ppb when including indoor source contributions.
23 The percent of asthmatics experiencing multiple exposures a and abovet 250 and 300 ppb is
24 much lower, typically less than 1 percent of all asthmatics are exposed at the higher potential
25 benchmark levels. Also provided in Figure 32 are the percent of asthmatics exposed to selected
26 levels in the absence of indoor sources. Again, without the indoor source contribution, there are
27 reduced occurrences of multiple exposures at all of the potential health effect benchmark levels
28 compared with when indoor sources were modeled.
29
30
127
-------
Estimated Number of
Repeated Exposures in a Year
1
2
3
4
5
Potential Health Effect Benchmark
Level (ppb)
Figure 31. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures
above potential health effect benchmark levels, using 2003 modeled air quality (as is), with modeled
indoor sources.
Estimated Number of
Repeated Exposures
6
7
8
9
10
Health Effect Benchmark Level (ppb)
Figure 32. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures
above potential health effect benchmark levels, using modeled 2002 air quality (as is), with and without
indoor sources.
128
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
3.4.4 One-Hour Exposures Associated with Just Meeting the Current Standard
To simulate just meeting the current NO2 standard, 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. Similar estimates of short-term exposures
(i.e., 1-hour) were generated for the total population and population subgroups of interest (i.e.,
asthmatics and asthmatic children).
3.4.4.1 Number of Estimated Exposures above Selected Levels
In considering exposures estimated to occur associated with air quality simulated to just
meet the current annual average NO2 standard, the number of persons experiencing
concentrations at or above the potential health effect benchmarks increased. To allow for
reasonable comparison, the number of persons affected considering each scenario is expressed as
the percent of the subpopulation of interest. Figure 33 illustrates the percent of asthmatics
estimated to experience at least one exposure at or above the selected potential health effect
benchmark concentrations, with just meeting the current standard and including indoor source
contributions. While it was estimated that about 30% percent of asthmatics would be exposed to
200 ppb (1-hour average) at least once in a year for as is air quality, it was estimated that around
80 percent of asthmatics would experience at least one concentration above the lowest potential
health effect benchmark level in a year representing just meeting the current standard. Again,
estimates for asthmatic children exhibited a similar trend, with between 75 to 80 percent exposed
to a concentration at or above the lowest potential health effect benchmark level at least once per
year for a year just meeting the current standard (data not shown). The percent of all asthmatics
experiencing the higher benchmark levels is reduced to between 31 and 45 percent for the 250
ppb, 1-hour benchmark, and between 10 and 24 percent for the 300 ppb, 1-hour benchmark level
associated with air quality representing just meeting the current annual average standard.
200
250
300
2003 AQ (std) - with indoor soucrces
2002 AQ (std) - with indoor soucrces
2001 AQ (std) - with indoor soucrces
Simulated Year - Scenario
Potential Health Effect Benchmark Level (ppb)
Figure 33. Estimated percent of all asthmatics in Philadelphia with at least one exposure at or above the
potential health effect benchmark level, using modeled 2001-2003 air quality just meeting the current
standard, with modeled indoor sources
129
-------
1 In evaluating the influence of indoor source contribution for the scenario just meeting the
2 current standard, the numbers of individuals exposed at selected levels are reduced without
3 indoor sources, ranging from about 26 percent lower for the 200 ppb level to around 11 percent
4 for the 300 ppb level when compared with exposure estimates that accounted for indoor sources
5 (Figure 34).
200
Potential Health Effect Benchmark Level (ppb)
2002 AQ (std) - with indoor sources
2002 AQ (std) - no indoor sources
Simulated Year - Scenario
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Figure 34. Estimated number of all asthmatics in Philadelphia with at least one exposure at or above the
potential health effect benchmark level, using modeled 2002 air quality just meeting the current standard,
with and without modeled indoor sources.
3.4.4.2 Number of Repeated Exposures Above Selected Levels
For air quality simulated to just meet the current standard, repeated exposures at the selected
potential health effect benchmarks are more frequent than that estimated for the modeled as is air
quality. Figure 35 illustrates this using the simulated asthmatic population for year 2002 data as
an example. Many asthmatics that are exposed at or above the selected levels are exposed more
than one time. Repeated exposures above the potential health effect benchmark levels are
reduced however, when not including the contribution from indoor sources. The percent of
asthmatics exposed drops with increasing benchmark level, with progressively fewer persons
experiencing multiple exposures for each benchmark level.
130
-------
90
Estimated Number of
Repeated Exposures
in a Year
Health Effect Benchmark Level (ppb)
2 Figure 35. Estimated percent of asthmatics in Philadelphia County with repeated exposures above
3 health effect benchmark levels, using modeled 2002 air quality just meeting the current standard, with
4 and without modeled indoor sources
5 3.5 Variability and Uncertainty
6 3.5.1 Introduction
7 The methods and the model used in this assessment conform to the most contemporary
8 modeling methodologies available. APEX is a powerful and flexible model that allows for the
9 realistic estimation of air pollutant exposure to individuals. Since it is based on human activity
10 diaries and accounts for the most important variables known to affect exposure, it has the ability
11 to effectively approximate actual conditions. In addition, the input data selected were the best
12 available data to generate the exposure results. However, there are constraints and uncertainties
13 with the modeling approach and the input data that limit the realism and accuracy of the model
14 results.
15
16 All models have limitations that require the use of assumptions. Limitations of APEX lie
17 primarily in the uncertainties associated with data distributions input to the model. Broad
18 uncertainties and assumptions associated with these model inputs, utilization, and application
19 include the following, with more detailed analysis summarized below and presented previously
20 (see US EPA, 2007d; Langstaff, 2007).
21
22 • The CHAD activity data used in APEX are compiled from a number of studies in
23 different areas, and for different seasons and years. Therefore, the combined data set
24 may not constitute a representative sample for a particular study scenario.
131
-------
1 • Commuting pattern data were derived from the 2000 U.S. Census. The commuting data
2 address only home-to-work travel. The population not employed outside the home is
3 assumed to always remain in the residential census tract. Furthermore, although several
4 of the APEX microenvironments account for time spent in travel, the travel is assumed to
5 always occur in basically a composite of the home and work block. No other provision is
6 made for the possibility of passing through other blocks during travel.
7 • APEX creates seasonal or annual sequences of daily activities for a simulated individual
8 by sampling human activity data from more than one subject. Each simulated person
9 essentially becomes a composite of several actual people in the underlying activity data.
10 • The model currently does not capture certain correlations among human activities that
11 can impact microenvironmental concentrations (for example, cigarette smoking leading
12 to an individual opening a window, which in turn affects the amount of outdoor air
13 penetrating the microenvironment).
14 • Certain aspects of the personal profiles are held constant, though in reality they change as
15 individuals age. This is only important for simulations with long timeframes, particularly
16 when simulating young children (e.g., over a year or more).
17 3.5.2 Input Data Evaluation
18 Modeling results are heavily dependent on the quality of the data that are input to the system.
19 As described above, several studies were reviewed, and data from these studies were used to
20 develop the parameters and factors that were used to build the microenvironments in this
21 assessment. A constraint on this effort is that there are a limited number of NC>2 exposure studies
22 to use for evaluation.
23
24 The input data used in this assessment were selected to best simulate actual conditions that
25 affect human exposure. Using well characterized data as inputs to the model lessens the degree
26 of uncertainty in exposure estimates. Still, the limitations and uncertainties of each of the data
27 streams affect the overall quality of the model output. These issues and how they specifically
28 affect each data stream are discussed in this section.
29
30 3.5.2.1 Meteorological Data
31 Meteorological data are taken directly from monitoring stations in the assessment areas. One
32 strength of these data is that it is relatively easy to see significant errors if they appear in the data.
33 Because general climactic conditions are known for each area simulation, it would have been
34 apparent upon review if there were outliers in the dataset. However, there are limitations in the
35 use of these data. Because APEX only uses one temperature value per day, the model does not
36 represent hour-to-hour variations in meteorological conditions throughout the day that may affect
37 both NC>2 formation and exposure estimates within microenvironments.
38 3.5.2.2 Air Quality Data
39 Air quality data used in the exposure modeling was determined through use of EPA's
40 recommended regulatory air dispersion model, AERMOD (version 07026 (US EPA, 2004)), with
41 meteorological data discussed above and emissions data based on the EPA's National Emissions
42 Inventory for 2002 (US EPA, 2007b) and the CAMD Emissions Database (US EPA, 2007c) for
43 stationary sources and mobile sources determined from local travel demand modeling and EPA's
44 MOBILE6.2 emission factor model. All of these are high quality data sources. Parameterization
132
-------
1 of meteorology and emissions in the model were made in as accurate a manner as possible to
2 ensure best representation of air quality for exposure modeling. Further, minor sources not
3 included in the dispersion modeling were captured and any remaining long-term errors in the
4 results corrected through use of local concentrations derived from monitor observations. Thus,
5 the resulting air quality values are free of systematic errors to the best approximation available
6 through application of modeled data.
7
8 3.5.2.3 Population and Commuting Data
9 The population and commuting data are drawn from U.S. Census data from the year 2000.
10 This is a high quality data source for nationwide population data in the U.S. However, the data
11 do have limitations. The Census used random sampling techniques instead of attempting to
12 reach all households in the U.S., as it has in the past. While the sampling techniques are well
13 established and trusted, they introduce some uncertainty to the system. The Census has a quality
14 section (http://www.census.gov/quality/) that discusses these and other issues with Census data.
15
16 In addition to these data quality issues, certain simplifying assumptions were made in order
17 to better match reality or to make the data match APEX input specifications. For example, the
18 APEX dataset does not differentiate people that work at home from those that commute within
19 their home tract, and individuals that commute over 120 km a day were assumed to not commute
20 daily. In addition to emphasizing some of the limitations of the input data, these assumptions
21 introduce uncertainty to the results.
22
23 Furthermore, the estimation of block-to-block commuter flows relied on the assumption that
24 the frequency of commuting to a workplace block within a tract is proportional to the amount of
25 commercial and industrial land in the block. This assumption introduces additional uncertainty.
26
27 3.5.2.4 Activity Pattern Data
28 It is probable that the CHAD data used in the system is the most subject to limitations and
29 uncertainty of all the data used in the system. Much of the data used to generate the daily diaries
30 are over 20 years old. Table 46 indicates the ages of the CHAD diaries used in this modeling
31 analysis. While the specifics of people's daily activities may not have changed much over the
32 years, it is certainly possible that some differences do exist. In addition, the CHAD data are
33 taken from numerous surveys that were performed for different purposes. Some of these surveys
34 collected only a single diary-day while others went on for several days. Some of the studies
35 were designed to not be representative of the U.S. population, although a large portion of the
36 data are from National surveys. Furthermore, study collection periods occur at different times of
37 the year, possibly resulting in seasonal differences. A few of these limitations are corrected by
38 the approaches used in the exposure modeling (e.g., weighting by US population demographics
39 for a particular location, adjusting for effects of temperature on human activities).
40
41 A sensitivity analysis was performed to evaluate the impact of the activity pattern database
42 on APEX model results for O3 (see Langstaff (2006) and US EPA (2007d)). Briefly, exposure
43 results were generated using APEX with all of the CHAD diaries and compared with results
44 generated from running APEX using only the CHAD diaries from the National Human Activity
45 Pattern Study (NHAPS), a nationally representative study in CHAD. There was very good
46 agreement between the APEX results for the 12 cities evaluated, whether all of CHAD or only
47 the NHAPS component of CHAD is used. The absolute difference in percent of persons above a
133
-------
1 particular concentration level ranged from -1% to about 4%, indicating that the exposure model
2 results are not being overly influenced by any single study in CHAD. It is likely that similar
3 results would be obtained here for NC>2 exposures, although remains uncertain due to different
4 averaging times (1-hour vs. 8-hour average).
5
6 3.5.2.5 Air Exchange Rates
7 There are several components of uncertainty in the residential air exchange rate distributions
8 used for this analysis. US EPA (2007d) details an analysis of uncertainty due to extrapolation of
9 air exchange rate distributions between-CMSAs and within-CMSA uncertainty due to sampling
10 variation. In addition, the uncertainty associated with estimating daily air exchange rate
11 distributions from air exchange rate measurements with varying averaging times is discussed.
12 The results of those investigations are briefly summarized here.
13 Extrapolation among cities
14 Location-specific distributions were assigned in the APEX model, as detailed in the indoors-
15 residential microenvironment. Since specific data for all of the locations targeted in this analysis
16 were not available, data from another location were used based on similar influential
17 characteristics. Such factors include age composition of housing stock, construction methods,
18 and other meteorological variables not explicitly treated in the analysis, such as humidity and
19 wind speed patterns. In order to assess the uncertainty associated with this extrapolation,
20 between-CSA uncertainty was evaluated by examining the variation of the geometric means and
21 standard deviations across cities and studies.
22
23 The analysis showed a relatively wide variation across different cities in the air exchange rate
24 geometric mean and standard deviation, stratified by air-conditioning status and temperature
25 range. This implies that the air exchange rate modeling results would be very different if the
26 matching of modeled locations to study locations was changed. For example, the NC>2 exposure
27 estimates may be sensitive to the assumption that the Philadelphia air exchange rate distributions
28 can be represented by the New York City air exchange rate data.
29 Within CSA uncertainty
30 There is also variation within studies for the same location (e.g., Los Angeles), but this is
31 much smaller than the variation across CMSAs. This finding tends to support the approach of
32 combining different studies for a CMSA. In addition, within-city uncertainty was assessed by
33 using a bootstrap distribution to estimate the effects of sampling variation on the fitted geometric
34 means and standard deviations for each CMSA. The bootstrap distributions assess the
35 uncertainty due to random sampling variation but do not address uncertainties due to the lack of
36 representativeness of the available study data or the variation in the lengths of the AER
37 monitoring periods.
38
39 1,000 bootstrap samples were randomly generated for each AER subset (of size N),
40 producing a set of 1,000 geometric mean and geometric standard deviation pairs. The analysis
41 indicated that the geometric standard deviation uncertainty for a given CSA/air-conditioning-
42 status/temperature-range combination tended to have a range of at most from fitted GSD-1.0 hf1
43 to fitted GSD+1.0 hf1', but the intervals based on larger AER sample sizes were frequently much
44 narrower. The ranges for the geometric means tended to be approximately from fitted GM-0.5
134
-------
1 hf1 to fitted GM+0.5 hf1, but in some cases were much smaller. Figure 36 illustrates such
2 results for Los Angeles as an example.
4.0—
3.5 —
P 3.0-1
"3
55
.a 2.5—
\
0.5
1.0
I
1.5
Geometric Mean
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
•Bootstrapped Data
"Original Data
Figure 36. Geometric mean and standard deviation of air exchange rate bootstrapped for Los Angeles
residences with A/C, temperature range from 20-25 degrees centigrade (from US EPA, 2007d).
Variation in measurement averaging times
Although the averaging periods for the air exchange rates in the study data varied from one
day to seven days, the analyses did not take the measurement duration into account and treated
the data as if they were a set of statistically independent daily averages. To investigate the
uncertainty of this assumption, correlations between consecutive 24-hour air exchange rates
measured at the same house were investigated using data from the Research Triangle Park Panel
Study (US EPA, 2007d). The results showed extremely strong correlations, providing support
for the simplified approach of treating multi-day averaging periods as if they were 24-hour
averages.
3.5.2.6 Air Conditioning Prevalence
Because the selection of an air exchange rate distribution is conditioned on the presence or
absence of an air-conditioner, for each modeled area, the air conditioning status of the residential
microenvironments was simulated randomly using the probability that a residence has an air
conditioner, i.e., the residential air conditioner prevalence rate. For this study we used location-
specific data from the American Housing Survey of 2003. US EPA (2007d) 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 Energy Information Administration's Residential Energy Consumption Survey (RECS)
of 2001 for more aggregate geographic subdivision (e.g., states, multi-state Census divisions and
regions).
Air conditioning prevalence rates for the 5 locations from the American Housing Survey
(Table 50) ranged from 55% for Los Angeles to 97% for Atlanta. Reported standard errors were
relatively small, ranging from less than 1.2% for Atlanta to 1.8% for Detroit. The corresponding
95% confidence intervals are also small and range from approximately 4.6% to 6.9%. The
135
-------
1 RECS prevalence estimates and confidence intervals compared with the similar locations using
2 AHS data were mixed. Good agreements between the AHS and RECS confidence intervals was
3 found for Atlanta and Detroit. Poor agreement with the AHS for either the Census Region or
4 Census Division estimates was shown for Los Angeles and Philadelphia, with estimates of those
5 owning A/C lower when considering the RECS data. However, since the AHS survey results are
6 city-specific and were based on a more recent survey, the AHS prevalence estimates were used
7 for the APEX modeling.
8
9 Furthermore, some residences use evaporative coolers, also known as "swamp coolers," for
10 cooling. The estimation of air exchange rate distributions from measurement data used here did
11 not take into account the presence or absence of an evaporative cooler. Based on statistical
12 comparison tests (i.e., F-test, Kruskal-Wallis, Mood) for where information was available to
13 generate AER distributions with and without swamp cooler ownership, it was determined that
14 presence or absence of such data did not alter the statistical air exchange model (US EPA,
15 2007d).
136
-------
i 3.6 References
2 AHS. (2003a). American Housing Survey for 2003. Available at:
3 http://www.census.gov/hhes/www/housing/ahs/ahs.html.
4 AHS. (2003b). Source and Accuracy Statement for the 2003 AHS-N Data Chart. Available at:
5 http://www.census.gov/hhes/www/housing/ahs/03dtchrt/source.html.
6 Akland GG, Hartwell TD, Johnson TR, Whitmore RW. (1985). Measuring human exposure to
7 carbon monoxide in Washington, D. C. and Denver, Colorado during the winter of 1982-83.
8 Environ Sci Technol. 19:911-918.
9 Avol EL, Navidi WC, Colome SD. (1998) Modeling ozone levels in and around southern
10 California homes. Environ Sci Technol. 32:463-468.
11 Biller WF, Feagans TB, Johnson TR, Duggan GM, Paul RA, McCurdy T, Thomas HC. (1981).
12 A general model for estimating exposure associated with alternative NAAQS. Paper No. 81-
13 18.4 in Proceedings of the 74th Annual Meeting of the Air Pollution Control Association,
14 Philadelphia, PA.
15 CARB. (2001). Indoor air quality: residential cooking exposures. Final report. California Air
16 Resources Board, Sacramento, California. Available at:
17 http://www.arb.ca.gov/research/indoor/cooking/cooking.htm.
18 CDC. (2007). National Center for Health Statistics. National Health Interview Survey (NHIS)
19 Public Use Data Release (2003). Available at:
20 http://www.cdc.gov/nchs/about/major/nhis/quest_data_related_1997_forward.htm.
21 Chan AT and Chung MW. (2003). Indoor-outdoor air quality relationships in vehicle: effect of
22 driving environment and ventilation modes. Atmos Environ. 37:3795-3808.
23 Chilrud SN, Epstein D, Ross JM, Sax SN, Pederson D, Spengler JD, Kinney PL. (2004).
24 Elevated airborne exposures of teenagers to manganese, chromium, and iron from steel dust
25 and New York City's subway system. Environ Sci Technol. 38:732-737.
26 Colome SD, Wilson AL, Tian Y. (1993). California Residential Indoor Air Quality Study,
27 Volume 1, Methodology and Descriptive Statistics. Prepared for the Gas Research Institute,
28 Pacific Gas & Electric Co., San Diego Gas & Electric Co., Southern California Gas Co.
29 Colome SD, Wilson AL, Tian Y. (1994). California Residential Indoor Air Quality Study,
30 Volume 2, Carbon Monoxide and Air Exchange Rate: An Univariate and Multivariate
31 Analysis. Chicago, IL. Prepared for the Gas Research Institute, Pacific Gas & Electric Co.,
32 San Diego Gas & Electric Co., Southern California Gas Co. GRI-93/0224.3
33 Finlayson-Pitts BJ and Pitts JN. (2000). Chemistry of the Upper and Lower Atmosphere.
34 Academic Press, San Diego CA. Page 17.
35 Hartwell TD, Clayton CA, Ritchie RM, Whitmore RW, Zelon HS, Jones SM, Whitehurst DA.
36 (1984). Study of Carbon Monoxide Exposure of Residents of Washington, DC and Denver,
37 Colorado. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office of
38 Research and Development, Environmental Monitoring Systems Laboratory. EPA-600/4-84-
39 031.
40 Johnson TR and Paul RA. (1983). The NAAQS Exposure Model (NEM) Applied to Carbon
41 Monoxide. EPA-450/5-83-003. Prepared for the U.S. Environmental Agency by PEDCo
42 Environmental Inc., Durham, N.C. under Contract No. 68-02-3390. U.S. Environmental
43 Protection Agency, Research Triangle Park, North Carolina.
137
-------
1 Johnson T. (1984). A Study of Personal Exposure to Carbon Monoxide in Denver, Colorado.
2 Research Triangle Park, NC: U.S. Environmental Protection Agency, Environmental
3 Monitoring Systems Laboratory. EPA-600/4-84-014.
4 Johnson T. (1989). Human Activity Patterns in Cincinnati, Ohio. Palo Alto, CA: Electric
5 Power Research Institute. EPRIEN-6204.
6 Johnson T, Capel J, Olaguer E, Wijnberg L. (1992). Estimation of Ozone Exposures
7 Experienced by Residents of ROMNET Domain Using a Probabilistic Version of NEM.
8 Prepared by IT Air Quality Services for the Office of Air Quality Planning and Standards,
9 U.S. Environmental Protection Agency, Research Triangle Park, North Carolina.
10 Johnson T, Capel J, McCoy M. (1996a). Estimation of Ozone Exposures Experienced by Urban
11 Residents Using a Probabilistic Version of NEM and 1990 Population Data. Prepared by IT
12 Air Quality Services for the Office of Air Quality Planning and Standards, U.S.
13 Environmental Protection Agency, Research Triangle Park, North Carolina, September.
14 Johnson T, Capel J, Mozier J, McCoy M. (1996b). Estimation of Ozone Exposures
15 Experienced by Outdoor Children in Nine Urban Areas Using a Probabilistic Version of
16 NEM. Prepared for the Air Quality Management Division under Contract No. 68-DO-30094,
17 April.
18 Johnson T, Capel J, McCoy M, Mozier J. (1996c). Estimation of Ozone Exposures Experienced
19 by Outdoor Workers in Nine Urban Areas Using a Probabilistic Version of NEM. Prepared
20 for the Air Quality Management Division under Contract No. 68-DO-30094, April.
21 Johnson T, Mihlan G, LaPointe J, Fletcher K. (1999). Estimation Of Carbon Monoxide
22 Exposures and Associated Carboxyhemoglobin Levels In Denver Residents Using
23 pNEM/CO (version 2.0). Prepared for the U.S. Environmental Protection Agency under
24 Contract No. 68-D6-0064, March 1999.
25 Kinney PL, Chillrud SN, Ramstrom S, Ross J, Spengler JD. (2002). Exposures to multiple air
26 toxics in New York City. Environ Health Perspect. 110:539-546.
27 Klepeis NE, Tsang AM, Behar JV. (1996). Analysis of the National Human Activity Pattern
28 Survey (NHAPS) Respondents from a Standpoint of Exposure Assessment. Washington, DC:
29 U.S. Environmental Protection Agency, Office of Research and Development. EPA/600/R-
30 96/074.
31 Koontz, M. D., L. L. Mehegan, and N. L. Nagda. 1992. Distribution and Use of Cooking
32 Appliances That Can Affect Indoor Air Quality, Report No. GRI-93/0013. Gas Research
33 Institute, Chicago.
34 LangstaffJE. (2007). OAQPS Staff Memorandum to Ozone NAAQS Review Docket (OAR-
35 2005-0172). Subject: Analysis of Uncertainty in Ozone Population Exposure Modeling.
36 [January 31,2007]. Available at:
37 http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html.
38 McCurdy T, Glen G, Smith L, Lakkadi Y. (2000). The National Exposure Research
39 Laboratory's Consolidated Human Activity Database, JExp Anal Environ Epidemiol. 10:
40 566-578.
41 Meng QY, Turpin BJ, Korn L, Weisel CP, Morandi M, Colome S, Zhang JJ, Stock T, Spektor D,
42 Winer A, Zhang L, Lee JH, Giovanetti R, Cui W, Kwon J, Alimokhtari S, Shendell D, Jones
43 J, Farrar C, Maberti S. (2004). Influence of ambient (outdoor) sources on residential indoor
44 and personal PM2.5 concentrations: Analyses of RIOPA data. J Expos Anal Environ
45 Epidemiol. 15:17-28.
138
-------
1 Murray DM and Burmaster DE. (1995). Residential air exchange rates in the United States:
2 empirical and estimated parametric distributions by season and climatic region. Risk
3 Analysis. 15(4):459-465.
4 PA DOH. (2008). Behavioral Risk Factor Surveillance System. Pennsylvania Department of
5 Health, Bureau of Health Statistics and Research. Available at:
6 http://www.dsfhealth.state.pa.us/health/cwp/view.asp?a=175&Q=242623.
7 Persily A and Gorfain J. (2004). Analysis of ventilation data from the U.S. Environmental
8 Protection Agency Building Assessment Survey and Evaluation (BASE) Study. National
9 Institute of Standards and Technology, NISTIR 7145, December 2004.
10 Persily A, Gorfain J, Brunner G. (2005). Ventilation design and performance in U.S. office
11 buildings. ASHRAE Journal. April 2005, 30-35.
12 Robinson JP, Wiley JA, Piazza T, Garrett K, Cirksena K. (1989). Activity Patterns of California
13 Residents and their Implications for Potential Exposure to Pollution. California Air
14 Resources Board, Sacramento, CA. CARB-A6-177-33.
15 Roddin MF, Ellis HT, Siddiqee WM. (1979). Background Data for Human Activity Patterns,
16 Vols. 1, 2. Draft Final Report. Prepared for Strategies and Air Standards Division, Office of
17 Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
18 Triangle Park, N.C.
19 Sax SN, Bennett DH, Chillrud SN, Kinney PL, Spengler JD. (2004). Differences in source
20 emission rates of volatile organic compounds in inner-city residences of New York City and
21 Los Angeles. JExpos Anal Environ Epidemiol. 14(S):95-109.
22 Spicer CW, Kenny DV, Ward GF, Billick IH (1993). Transformations, lifetimes, and sources of
23 NO2, HONO, and HNO3 in indoor environments. JAWMA. 43(11):1479-1485.
24 Spier CE, Little DE, Trim SC, Johnson TR, Linn WS, Hackney JD. (1992). Activity patterns in
25 elementary and high school students exposed to oxidant pollution. J Expo Anal Environ
26 Epidemiol. 2:277-293.
27 Tsang AM and Klepeis NE. (1996). Descriptive Statistics Tables from a Detailed Analysis of
28 the National Human Activity Pattern Survey (NHAPS) Data. U.S. Environmental Protection
29 Agency. EPA/600/R-96/148.
30 US Census Bureau. (2007). Employment Status: 2000- Supplemental Tables. Available at:
31 http://www.census.gov/population/www/cen2000/phc-t28.html.
32 US DOT. (2007). Part 3-The Journey To Work files. Bureau of Transportation Statistics
33 (BTS). Available at: http://transtats.bts.gov/.
34 US EPA. (1999). Total Risk Integrated Methodology. Website:
35 http://www.epa.gov/ttnatw01/urban/trim/trimpg.html.
36 US EPA. (2002). Consolidated Human Activities Database (CHAD) Users Guide. Database
37 and documentation available at: http://www.epa.gov/chadnetl/.
38 US EPA. (2004). AERMOD: Description of Model Formulation. Office of Air Quality
39 Planning and Standards. EPA-454/R-03-004. Available at:
40 http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf
41 US EPA. (2006a). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model
42 Documentation (TRIM.Expo / APEX, Version 4) Volume I: User's Guide. Office of Air
43 Quality Planning and Standards, Research Triangle Park, NC. June 2006. Available at:
44 http ://www. epa.gov/ttn/fera/human_apex.html.
45 US EPA. (2006b). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model
46 Documentation (TRIM.Expo / APEX, Version 4) Volume II: Technical Support Document.
139
-------
1 Office of Air Quality Planning and Standards, Research Triangle Park, NC. June 2006.
2 Available at: http://www.epa.gov/ttn/fera/human_apex.html.
3 US EPA. (2007a). Integrated Science Assessment for Oxides of Nitrogen-Health Criteria
4 (First External Review Draft) and Annexes (August 2007). Research Triangle Park, NC:
5 National Center for Environmental Assessment. Available at:
6 http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l 81712.
7 US EPA. (2007b). 2002 National Emissions Inventory Data & Documentation. Available at:
8 http://www.epa.gov/ttn/chief/net/2002inventory.html.
9 US EPA. (2007c). Clean Air Markets - Data and Maps. Emissions Prepackaged Data Sets.
10 Available at: http://camddataandmaps.epa.gov/gdm/index.cfm?fuseaction=emissions.wizard.
11 US EPA. (2007d). Ozone Population Exposure Analysis for Selected Urban Areas (July 2007).
12 Research Triangle Park, NC: Office of Air Quality Planning and Standards. EPA-452/R-07-
13 010. Available at: http://epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html.
14 US EPA. (2007e). Review of the National Ambient Air Quality Standards for ozone:
15 assessment of scientific and technical information. OAQPS Staff paper (July 2007). Research
16 Triangle Park, NC: Office of Air Quality Planning and Standards. EPA-452/R-07-007a.
17 Available at: http://epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_sp.html.
18 Weisel CP, Zhang JJ, Turpin BJ, Morandi MT, Colome S, Stock TH, Spektor DM, Korn L,
19 Winer A, Alimokhtari S, Kwon J, Mohan K, Harrington R, Giovanetti R, Cui W, Afshar M,
20 Maberti S, Shendell D. (2004). Relationship of Indoor, Outdoor and Personal Air (RIOPA)
21 study; study design, methods and quality assurance/control results. J Exp Anal Environ
22 Epidemiol. 15:123-137.
23 Wiley JA, Robinson JP, Piazza T, Garrett K, Cirksena K, Cheng Y-T, Martin G. (1991a).
24 Activity Patterns of California Residents: Final Report. California Air Resources Board,
25 Sacramento, CA. ARB/R93/487. Available from: NTIS, Springfield, VA., PB94-108719.
26 Wiley JA, Robinson JP, Cheng Y-T, Piazza T, Stork L, Pladsen K. (1991b). Study of Children's
27 Activity Patterns: Final Report. California Air Resources Board, Sacramento, CA. ARB-R-
28 93/489.
29 Williams R, Suggs J, Creason J, Rodes C, Lawless P, Kwok R, Zweidinger R, Sheldon L.
30 (2000). The 1998 Baltimore particulate matter epidemiology-exposure study: Part 2.
31 Personal exposure associated with an elderly population. J Expo Anal Environ Epidemiol.
32 10(6):533-543.
33 Williams R, Suggs J, Rea A, Leovic K, Vette A, Croghan C, Sheldon L, Rodes C, Thornburg J,
34 Ejire A, Herbst M, Sanders, Jr W. (2003a). The Research Triangle Park particulate matter
35 panel study: PM mass concentration relationships. AtmosEnviron. 37:5349-5363.
36 Williams R, Suggs J, Rea A, Sheldon L, Rodes C, Thornburg J. (2003b). The Research
37 Triangle Park particulate patter panel study: modeling ambient source contribution to
38 personal and residential PM mass concentrations. Atmos Environ. 37:5365-5378.
39 Wilson AL, Colome SD, Baker PE, Becker EW. (1986). Residential Indoor Air Quality
40 Characterization Study of Nitrogen Dioxide, Phase I, Final Report. Prepared for Southern
41 California Gas Company, Los Angeles.
42 Wilson AL, Colome SD, Tian Y, Baker PE, Becker EW, Behrens DW, Billick IH, Garrison CA.
43 (1996). California residential air exchange rates and residence volumes. J Expos Anal
44 Environ Epidemiol. 6(3):311-326.
45 Yao X, Lau NT, Chan CK, Fang M. (2005). The use of tunnel concentration profile data to
46 determine the ratio of NO2/NOX directly emitted from vehicles. Atmos Chem Phys Discuss.
140
-------
1 5:12723-12740. Available at: http://www.atmos-chem-phys-discuss.net/5/12723/2005/acpd-
2 5-12723-2005.pdf.
141
-------
United States Office of Air Quality Planning and Standards EPA-452/P-08-002
Environmental Protection Air Quality Strategies and Standards Division April 2008
Agency Research Triangle Park, NC
-------
Risk and Exposure Assessment to Support the
Review of the NO2 Primary National Ambient
Air Quality Standard: Draft Technical Support
Document (TSD)
Appendices
-------
EPA-452/P-08-002
April 2008
Risk and Exposure Assessment to Support the
Review of the NO2 Primary National Ambient
Air Quality Standard: Draft Technical Support
Document (TSD)
Appendices
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina
-------
Appendix A. Ambient Monitor Characterization
Appendix A contains details regarding physical attributes of each monitor used within the
named locations (i.e., 18 specific locations were defined; it does not include the broadly grouped
locations of "Other CMS A" or Not MSA). Each of these monitors met the criteria for containing
a valid number of reported concentrations and were used throughout the air quality
characterization. Data provided include monitor location and purpose, ground height and
elevation above sea level, and distance to the nearest major roadway (Table A-l). In addition,
the distances and emissions of stationary sources that emit > 5 tons NOX per year were calculated
for each monitor (Table A-2).
A-1
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
ID
130890002
130893001
131210048
132230003
132470001
230313002
250051005
250092006
250094004
250095005
250210009
250250002
250250021
250250035
250250036
250250040
250250041
250250042
250251003
250270020
250270023
330110016
330110019
330110020
330150009
330150013
330150014
Lat
33.69
33.85
33.78
33.93
33.59
43.08
42.06
42.47
42.79
42.76
42.32
42.35
42.38
42.33
42.33
42.35
42.32
42.33
42.40
42.27
42.27
42.99
43.00
43.00
43.08
43.00
43.08
Long
-84.29
-84.21
-84.40
-85.05
-84.07
-70.75
-71.15
-70.97
-70.81
-71.11
-71.13
-71.10
-71.03
-71.12
-71.12
-71.04
-70.97
-71.08
-71.03
-71.80
-71.79
-71.46
-71.47
-71.47
-70.76
-71.20
-70.75
Land Use
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
AGRICULTURAL
AGRICULTURAL
RESIDENTIAL
AGRICULTURAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
INDUSTRIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
Location Type1
SUBURBAN
RURAL
URBAN AND CENTER
CITY
RURAL
RURAL
SUBURBAN
RURAL
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
RURAL
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
RURAL
URBAN AND CENTER
CITY
Objective2
POPULATION EXPOSURE
OTHER
HIGHEST CONCENTRATION
GENERAL/BACKGROUND
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
MAX OZONE CONCENTRATION
POPULATION EXPOSURE
UNKNOWN
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
UNKNOWN
UNKNOWN
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
OTHER
POPULATION EXPOSURE
Monitor3
Ht Elev
(m) (m)
5
5
5
4
5
-
4
5
4
-
4
5
4
4
6
5
4
3
4
5
5
3
1
2
308
0
290
417
219
40
61
52
1
0
0
6
6
0
0
0
10
6
59
145
145
75
61
61
3
0
4
Roadway4
Dist
(m) Type
432
579
134
>1000
809
70
17
158
15
337
144
7
7
158
158
37
>1000
26
228
44
49
168
70
70
48
>1000
266
3
2
3
-
3
2
3
3
3
3
3
2
3
3
3
3
-
3
4
3
3
3
3
3
3
-
3
A-2
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Boston
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Cleveland
Cleveland
Cleveland
Cleveland
Colorado Springs
Colorado Springs
Colorado Springs
Colorado Springs
Colorado Springs
Colorado Springs
Colorado Springs
Colorado Springs
Denver
Denver
ID
330150015
170310037
170310063
170310064
170310075
170310076
170313101
170313103
170314002
170314201
170314201
170318003
171971011
180890022
180891016
390350043
390350060
390350066
390350070
080416001
080416004
080416005
080416006
080416009
080416011
080416013
080416018
080013001
080050003
Lat
43.08
41.98
41.88
41.79
41.96
41.75
41.97
41.97
41.86
42.14
42.14
41.63
41.22
41.61
41.60
41.46
41.49
41.46
41.46
38.63
38.92
38.76
38.92
38.64
38.85
38.81
38.81
39.84
39.66
Long
-70.76
-87.67
-87.63
-87.60
-87.66
-87.71
-87.88
-87.88
-87.75
-87.80
-87.80
-87.57
-88.19
-87.30
-87.33
-81.58
-81.68
-81.58
-81.59
-104.72
-104.81
-104.76
-105.00
-104.71
-104.83
-104.82
-104.75
-104.95
-105.00
Land Use
COMMERCIAL
RESIDENTIAL
MOBILE
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
MOBILE
MOBILE
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
AGRICULTURAL
INDUSTRIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
INDUSTRIAL
RESIDENTIAL
AGRICULTURAL
RESIDENTIAL
INDUSTRIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
AGRICULTURAL
COMMERCIAL
Location Type1
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
RURAL
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
RURAL
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
RURAL
RURAL
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
RURAL
SUBURBAN
Objective2
POPULATION EXPOSURE
UNKNOWN
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
GENERAL/BACKGROUND
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
POPULATION EXPOSURE
HIGHEST CONCENTRATION
Monitor3
Ht Elev
(m) (m)
4
3
15
15
4
3
4
4
8
8
4
5
5
14
4
4
5
4
4
4
4
4
4
3
3
3
4
4
3
183
181
180
180
186
197
195
184
198
198
179
181
183
183
287
206
287
278
1673
1931
1747
2313
1707
1832
1823
1795
1559
1654
Roadway4
Dist
(m) Type
38
17
68
346
136
2
20
20
118
239
239
2
>1000
738
187
187
2
187
81
>1000
150
79
199
>1000
198
386
163
748
138
3
3
3
3
3
3
2
2
3
2
2
3
-
1
3
2
4
2
3
-
1
3
2
-
3
4
2
3
2
A-3
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Denver
Denver
Denver
Denver
Denver
Detroit
Detroit
Detroit
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
Jacksonville
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Los Angeles
Los Angeles
Los Angeles
Los Angeles
ID
080310002
080590006
080590008
080590009
080590010
260990009
261630016
261630019
481410027
481410028
481410037
481410044
481410055
481410057
481410058
120310032
320030022
320030023
320030073
320030078
320030539
320030557
320030563
320030601
320031019
320032002
060370002
060370016
060370030
060370113
Lat
39.75
39.91
39.88
39.86
39.90
42.73
42.36
42.43
31.76
31.75
31.77
31.77
31.75
31.66
31.89
30.36
36.39
36.81
36.17
35.47
36.14
36.16
36.18
35.98
35.79
36.19
34.14
34.14
34.04
34.05
Long
-104.99
-105.19
-105.17
-105.20
-105.24
-82.79
-83.10
-83.00
-106.49
-106.40
-106.50
-106.46
-106.40
-106.30
-106.43
-81.64
-114.91
-114.06
-115.33
-114.92
-115.09
-115.11
-115.10
-114.84
-115.36
-115.12
-117.92
-117.85
-118.22
-118.46
Land Use
COMMERCIAL
INDUSTRIAL
INDUSTRIAL
INDUSTRIAL
AGRICULTURAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
INDUSTRIAL
RESIDENTIAL
RESIDENTIAL
DESERT
MOBILE
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
DESERT
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
MOBILE
Location Type1
URBAN AND CENTER
CITY
RURAL
RURAL
RURAL
RURAL
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
RURAL
RURAL
SUBURBAN
RURAL
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
RURAL
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
Objective2
HIGHEST CONCENTRATION
UNKNOWN
GENERAL/BACKGROUND
GENERAL/BACKGROUND
UNKNOWN
UNKNOWN
HIGHEST CONCENTRATION
POPULATION EXPOSURE
GENERAL/BACKGROUND
SOURCE ORIENTED
MAX OZONE CONCENTRATION
MAX PRECURSOR EMISSIONS
IMPACT
UPWIND BACKGROUND
GENERAL/BACKGROUND
POPULATION EXPOSURE
UNKNOWN
SOURCE ORIENTED
POPULATION EXPOSURE
POPULATION EXPOSURE
REGIONAL TRANSPORT
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
POPULATION EXPOSURE
GENERAL/BACKGROUND
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
Monitor3
Ht Elev
(m) (m)
-
4
4
4
-
4
4
5
5
4
5
5
5
5
3
3.5
4
3.5
4
3.5
3
4
4
4
3.5
2
6
5
5
1589
1774
1715
1848
1877
189
191
192
1140
1126
1143
1128
0
0
0
7
0
490
0
1094
533
567
570
0
950
0
183
275
65
91
Roadway4
Dist
(m) Type
18
65
31
99
63
415
393
339
33
718
128
38
127
450
478
144
122
303
515
25
11
1
254
52
914
240
329
300
50
190
3
3
3
3
2
3
5
3
4
3
3
3
3
3
3
1
2
3
2
3
3
3
3
3
3
3
3
3
3
3
A-4
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
ID
060370206
060371002
060371103
060371201
060371301
060371601
060371701
060372005
060374002
060375001
060375005
060376002
060376012
060379002
060379033
060590001
060590007
060591003
060595001
060650012
060655001
060658001
060659001
060710001
060710012
060710014
060710015
060710017
060710306
060711004
Lat
33.96
34.18
34.07
34.20
33.93
34.01
34.07
34.13
33.82
33.92
33.95
34.39
34.38
34.69
34.67
33.83
33.83
33.67
33.93
33.92
33.85
34.00
33.68
34.90
34.43
34.51
35.78
34.14
34.51
34.10
Long
-117.84
-118.32
-118.23
-118.53
-118.21
-118.06
-117.75
-118.13
-118.19
-118.37
-118.43
-118.53
-118.53
-118.13
-118.13
-117.94
-117.94
-117.93
-117.95
-116.86
-116.54
-117.42
-117.33
-117.02
-117.56
-117.33
-117.37
-116.06
-117.33
-117.63
Land Use
COMMERCIAL
COMMERCIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
INDUSTRIAL
MOBILE
RESIDENTIAL
RESIDENTIAL
Location Type1
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
RURAL
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
Objective2
UNKNOWN
UNKNOWN
HIGHEST CONCENTRATION
UNKNOWN
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UPWIND BACKGROUND
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UPWIND BACKGROUND
Monitor3
Ht Elev
(m) (m)
5
13
6
7
6
6
4
6
.
4
-
_
5
3
5
4
6
82
4
6
4
-
8
-
4
-
4
4
6
300
168
87
226
27
75
270
250
6
21
21
375
397
725
725
45
10
0
82
677
171
250
1440
690
4100
876
498
607
913
369
Roadway4
Dist
(m) Type
>1000
58
55
206
29
78
15
385
1
10
149
2
143
61
146
225
225
202
570
432
75
133
522
64
30
18
42
64
38
349
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
1
3
3
4
3
3
3
3
3
3
2
A-5
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Miami
Miami
Miami
Miami
Miami
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
ID
060712002
060711234
060714001
060719004
061110005
061110007
061111003
061111004
061112002
061112003
061113001
120110003
120110031
120118002
120860027
120864002
090010113
090019003
090090027
090091123
340030001
340030005
340130011
340130016
340131003
340170006
340210005
340230011
340273001
Lat
34.10
35.76
34.42
34.11
33.20
32.71
34.45
34.45
34.28
34.28
34.26
26.28
26.27
26.09
25.73
25.80
41.18
41.12
41.30
41.31
40.81
40.90
40.73
40.72
40.76
40.67
40.28
40.46
40.79
Long
-117.49
-117.40
-117.28
-117.27
-117.37
-117.15
-119.27
-119.23
-118.68
-119.31
-119.14
-80.28
-80.30
-80.11
-80.16
-80.21
-73.19
-73.34
-72.90
-72.92
-73.99
-74.03
-74.14
-74.15
-74.20
-74.13
-74.74
-74.43
-74.68
Land Use
INDUSTRIAL
DESERT
RESIDENTIAL
COMMERCIAL
UNKNOWN
COMMERCIAL
MOBILE
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
INDUSTRIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
FOREST
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
INDUSTRIAL
INDUSTRIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
AGRICULTURAL
AGRICULTURAL
Location Type1
CITY
SUBURBAN
RURAL
SUBURBAN
SUBURBAN
UNKNOWN
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
RURAL
RURAL
SUBURBAN
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
RURAL
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
RURAL
RURAL
Objective2
UNKNOWN
OTHER
UNKNOWN
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
HIGHEST CONCENTRATION
GENERAL/BACKGROUND
POPULATION EXPOSURE
HIGHEST CONCENTRATION
MAX PRECURSOR EMISSIONS
IMPACT
POPULATION EXPOSURE
POPULATION EXPOSURE
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
HIGHEST CONCENTRATION
POPULATION EXPOSURE
MAX OZONE CONCENTRATION
POPULATION EXPOSURE
UNKNOWN
Monitor3
Ht Elev
(m) (m)
5
1
-
5
1
5
-
4
4
2
4
6
4
4
16
4
4
5
3.67
9
4
3
4
5
4
5
4
4
5
381
545
1006
0
320
244
231
262
314
3
43
3
3
3
2
5
3
4
11
18
61
6
3
3
48.45
3
30
21
274
Roadway4
Dist
(m) Type
81
>1000
111
169
63
89
18
56
471
90
307
22
103
>1000
15
87
8
508
237
14
82
172
232
6
25
266
442
298
227
3
-
3
3
3
3
2
3
1
1
3
3
4
-
3
3
3
4
1
2
3
5
1
1
3
3
1
3
3
A-6
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Phoenix
Phoenix
Phoenix
Phoenix
ID
340390004
340390008
360050080
360050083
360050110
360470011
360590005
360610010
360610056
360810097
360810098
360810124
361030009
100031003
100031007
100032004
340070003
420170012
420450002
420910013
421010004
421010029
421010047
040130019
040133002
040133003
040133010
Lat
40.64
40.60
40.84
40.87
40.82
40.73
40.74
40.74
40.76
40.76
40.78
40.74
40.83
39.76
39.55
39.74
39.92
40.11
39.84
40.11
40.01
39.96
39.94
33.48
33.46
33.48
33.46
Long
-74.21
-74.44
-73.92
-73.88
-73.90
-73.95
-73.59
-73.99
-73.97
-73.76
-73.85
-73.82
-73.06
-75.49
-75.73
-75.56
-75.10
-74.88
-75.37
-75.31
-75.10
-75.17
-75.17
-112.14
-112.05
-111.92
-112.12
Land Use
INDUSTRIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
INDUSTRIAL
COMMERCIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
AGRICULTURAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
INDUSTRIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
Location Type1
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
SUBURBAN
RURAL
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
SUBURBAN
SUBURBAN
Objective2
HIGHEST CONCENTRATION
POPULATION EXPOSURE
HIGHEST CONCENTRATION
UNKNOWN
UNKNOWN
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
GENERAL/BACKGROUND
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
OTHER
UNKNOWN
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
HIGHEST CONCENTRATION
POPULATION EXPOSURE
POPULATION EXPOSURE
Monitor3
Ht Elev
(m) (m)
4
4
15
15
6
5
38
10
12
8
-
-
-
-
5
2
2
4
7
11
11
4.3
9
5.8
4.2
5.4
0
15
24
0
9
27
38
15
0
6
8
0
65
20
0
7.6
12
3
53
22
25
21
333
339
368
325
Roadway4
Dist
(m) Type
37
99
122
132
76
171
32
55
62
197
9
150
116
189
144
82
405
393
413
630
45
103
66
401
141
78
7
4
3
3
5
3
3
3
3
3
3
3
3
2
2
3
3
3
3
3
1
3
3
2
3
3
3
3
A-7
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Phoenix
Phoenix
Phoenix
Provo
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
St. Louis
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
ID
040134005
040134011
040139997
490490002
171630010
291830010
291831002
291890001
291890004
291890006
291893001
291895001
291897002
291897003
295100072
295100080
295100086
110010017
110010025
110010041
110010043
240053001
245100040
245100050
510130020
510590005
510590018
Lat
33.41
33.37
33.50
40.25
38.61
38.58
38.87
38.52
38.53
38.61
38.64
38.77
38.73
38.72
38.62
38.68
38.67
38.90
38.98
38.90
38.92
39.31
39.30
39.32
38.86
38.89
38.74
Long
-111.93
-112.62
-112.10
-111.66
-90.16
-90.84
-90.23
-90.34
-90.38
-90.50
-90.35
-90.29
-90.38
-90.37
-90.20
-90.25
-90.24
-77.05
-77.02
-76.95
-77.01
-76.47
-76.60
-76.58
-77.06
-77.47
-77.08
Land Use
RESIDENTIAL
AGRICULTURAL
RESIDENTIAL
COMMERCIAL
INDUSTRIAL
AGRICULTURAL
AGRICULTURAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
COMMERCIAL
RESIDENTIAL
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
COMMERCIAL
AGRICULTURAL
RESIDENTIAL
Location Type1
URBAN AND CENTER
CITY
RURAL
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
RURAL
RURAL
SUBURBAN
SUBURBAN
RURAL
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
SUBURBAN
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
URBAN AND CENTER
CITY
RURAL
SUBURBAN
Objective2
UNKNOWN
SOURCE ORIENTED
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
UNKNOWN
HIGHEST CONCENTRATION
UNKNOWN
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
HIGHEST CONCENTRATION
POPULATION EXPOSURE
UNKNOWN
HIGHEST CONCENTRATION
MAX PRECURSOR EMISSIONS
IMPACT
HIGHEST CONCENTRATION
POPULATION EXPOSURE
UNKNOWN
POPULATION EXPOSURE
UNKNOWN
Monitor3
Ht Elev
(m) (m)
4
4
4
4
3
4
4
4
4
4
2
4
4
14
4
4
10
11
4.6
4.2
4
7
4
4
352
258
346
1402
125
0
131
183
183
175
161
168
168
0
154
152
0
20
91
8
50
5
12
49
171
77
11
Roadway4
Dist
(m) Type
259
12
433
353
18
340
31
161
95
97
5
421
59
112
43
116
133
54
106
141
278
186
14
338
80
315
54
3
3
3
2
4
3
3
2
2
3
1
3
3
3
4
3
3
3
3
4
3
3
3
2
3
5
3
A-8
-------
Table A-1 . Attributes of location-specific ambient monitors used for air quality characterization and the distance to nearest major roadway.
Location
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
ID
510591004
510591005
510595001
511071005
511530009
515100009
Lat
38.87
38.84
38.93
39.02
38.86
38.81
Long
-77.14
-77.16
-77.20
-77.49
-77.64
-77.04
Land Use
COMMERCIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
RESIDENTIAL
Location Type1
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
SUBURBAN
URBAN AND CENTER
CITY
Objective2
UNKNOWN
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
POPULATION EXPOSURE
UNKNOWN
Monitor3
Ht Elev
(m) (m)
11
_
4
4
4
11
110
83.9
106
0
111
23
Roadway4
Dist
(m) Type
84
50
18
75
196
83
5
3
5
3
2
3
1 Land use indicates the prevalent land use within 1/4 mile of that site.
2 Objective Indicates the reason for measuring air quality by the monitor.
3 Monitor probe height (Ht) and site elevation (Elev) above sea level are given in meters (m).
4 Distances (Dist) to roadway are given in meters (m). Major road types are defined as: 1=primary limited access or interstate, 2=primary US and State highways, 3=Secondary State
and County, 4=freeway ramp, 5=other ramps.
A-9
-------
Table A-2. Distance of location-specific ambient monitors used in Tier 1 analyses to stationary sources emitting > 5 tons of NOX per year, within a 1 0 kilometer
distance of monitoring site.
Location
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
Chicago
ID
130890002
130893001
131210048
132230003
132470001
230313002
250051005
250092006
250094004
250095005
250210009
250250002
250250021
250250035
250250036
250250040
250250041
250250042
250251003
250270020
250270023
330110016
330110019
330110020
330150009
330150013
330150014
330150015
170310037
170310063
170310064
170310075
170310076
170313101
170313103
170314002
170314201
170314201
Distance (km) to Source emissions >5 tpy and within 10
km
n
1
3
5
0
0
5
3
12
0
10
57
62
55
62
62
56
25
65
49
28
28
0
0
0
5
1
5
5
17
57
33
31
46
30
30
63
7
7
mean
4.9
7.2
6.4
3.5
6.7
6.8
5.8
5.8
4.6
6.1
5.1
5.1
5.3
7.8
5.3
6.4
3.7
3.6
3.3
8.4
4.0
3.1
5.6
4.9
6.9
7.3
7.8
6.6
6.6
6.7
6.5
6.5
std
4.0
3.3
1.5
1.6
2.7
2.3
2.5
2.4
2.3
2.6
2.6
2.4
2.0
2.8
2.4
2.5
2.4
1.0
1.8
0.9
2.7
3.2
2.5
2.7
2.3
2.2
2.2
2.6
1.5
1.5
min
4.9
2.7
0.7
1.0
5.5
2.5
1.7
1.0
0.6
1.5
0.3
0.3
0.4
0.7
0.7
0.6
0.1
0.4
2.0
8.4
1.0
1.9
0.7
0.4
1.2
0.8
1.3
2.7
2.7
0.5
4.0
4.0
2.5
4.9
2.7
0.7
1.0
5.5
2.5
1.7
1.8
1.1
1.7
0.8
0.8
0.9
0.7
1.0
1.0
0.1
0.4
2.0
8.4
1.0
1.9
0.7
0.5
1.2
0.8
1.6
2.7
2.7
0.5
4.0
4.0
50
4.9
9.2
7.3
3.8
6.0
7.4
6.7
5.9
4.3
6.5
5.1
5.1
5.6
8.2
4.9
7.0
2.9
3.0
3.3
8.4
4.4
3.0
5.7
4.9
6.9
8.4
8.4
7.2
7.2
7.2
6.6
6.6
97.5
4.9
9.8
8.9
4.9
8.5
9.9
8.6
9.9
9.4
9.8
9.0
9.0
9.0
9.9
10.0
9.6
8.6
8.4
4.4
8.4
5.5
4.1
9.5
9.4
10.0
9.9
9.8
9.7
9.7
9.8
9.0
9.0
max
4.9
9.8
8.9
4.9
8.5
9.9
8.6
9.9
9.7
9.8
9.6
9.6
9.3
9.9
10.0
9.6
8.6
8.4
4.4
8.4
5.5
4.1
9.5
10.0
10.0
9.9
9.9
9.7
9.7
9.9
9.0
9.0
Emissions (tpy) of Sources within 10 km and >5 tpy
mean
34
34
1249
642
9
439
201
106
98
130
99
99
106
81
94
145
58
58
642
29
642
642
18
110
94
10
170
313
313
122
8
8
std
2
2106
769
4
1083
347
283
273
304
273
273
286
206
267
319
165
165
769
769
769
31
416
428
7
463
1638
1638
407
3
3
min
34
32
22
31
5
5
6
5
5
5
5
5
5
5
5
5
5
5
31
29
31
31
5
5
5
5
5
5
5
5
5
5
2.5
34
32
22
31
5
5
6
5
5
5
5
5
5
5
5
5
5
5
31
29
31
31
5
5
5
5
5
5
5
5
5
5
50
34
34
39
203
8
21
29
9
9
11
9
9
9
11
9
11
13
13
203
29
203
203
7
9
10
7
10
9
9
9
8
8
97.5
34
36
4895
1860
14
3794
923
1155
1155
1155
1155
1155
1155
957
1155
1155
868
868
1860
29
1860
1860
126
1677
2465
36
1677
8985
8985
1677
14
14
max
34
36
4895
1860
14
3794
923
1419
1419
1419
1419
1419
1419
957
1419
1419
868
868
1860
29
1860
1860
126
2465
2465
36
2204
8985
8985
2465
14
14
A-10
-------
Table A-2. Distance of location-specific ambient monitors used in Tier 1 analyses to stationary sources emitting > 5 tons of NOX per year, within a 1 0 kilometer
distance of monitoring site.
Location
Chicago
Chicago
Chicago
Chicago
Cleveland
Cleveland
Cleveland
Cleveland
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Detroit
Detroit
Detroit
El Paso
El Paso
El Paso
El Paso
El Paso
ID
170318003
171971011
180890022
180891016
390350043
390350060
390350066
390350070
080416001
080416004
080416005
080416006
080416009
080416011
080416013
080416018
080013001
080050003
080310002
080590006
080590008
080590009
080590010
260990009
261630016
261630019
481410027
481410028
481410037
481410044
481410055
Distance (km) to Source emissions >5 tpy and within 10
km
n
63
1
8
8
5
4
5
5
4
10
9
0
4
14
14
11
34
19
52
9
9
10
7
4
51
32
22
24
15
25
24
mean
7.3
4.0
5.1
4.7
8.1
4.1
8.0
7.6
5.1
5.9
7.5
5.2
5.0
6.3
6.9
5.3
6.7
5.3
5.9
6.2
6.5
5.5
4.9
7.4
6.3
8.1
2.2
8.7
5.9
2.8
std
2.0
3.8
2.4
1.9
2.4
1.9
1.8
4.4
2.2
2.1
4.3
2.3
2.9
1.7
1.8
3.7
2.5
2.1
2.0
3.2
3.1
3.2
2.1
2.2
1.6
1.9
2.6
1.2
1.8
min
1.7
4.0
0.8
2.1
5.2
1.0
5.2
5.5
0.8
3.5
3.3
1.0
2.0
2.1
4.3
1.6
1.0
0.9
2.7
3.7
2.5
1.1
0.3
1.3
2.6
1.5
0.9
0.1
4.4
1.6
2.5
2.3
4.0
0.8
2.1
5.2
1.0
5.2
5.5
0.8
3.5
3.3
1.0
2.0
2.1
4.3
1.6
1.0
0.9
2.7
3.7
2.5
1.1
0.3
2.0
2.6
1.5
0.9
0.1
4.4
1.6
50
8.0
4.0
4.1
4.1
8.3
4.4
8.3
7.3
5.1
5.6
8.1
5.3
5.8
6.9
7.1
4.7
9.1
5.8
6.3
6.1
7.0
5.6
5.7
7.9
6.5
8.6
1.6
9.4
5.6
2.2
97.5
9.6
4.0
9.4
7.6
9.9
6.4
9.8
9.7
9.1
9.8
9.5
9.3
9.6
9.9
9.6
9.5
10.0
9.7
8.6
10.0
9.9
9.2
7.7
9.8
10.0
9.3
9.3
10.0
9.5
9.6
max
9.7
4.0
9.4
7.6
9.9
6.4
9.8
9.7
9.1
9.8
9.5
9.3
9.6
9.9
9.6
9.5
10.0
9.8
8.6
10.0
9.9
9.2
7.7
9.9
10.0
9.3
9.3
10.0
9.5
9.6
Emissions (tpy) of Sources within 10 km and >5 tpy
mean
361
20
815
815
673
810
673
673
780
48
490
780
345
346
430
310
313
319
63
59
53
73
63
387
57
99
127
135
158
127
std
1201
1680
1680
664
681
664
664
1374
80
1393
1374
1113
1113
1254
1622
1233
1495
66
68
66
71
70
797
168
195
338
230
366
338
min
5
20
8
8
126
165
126
126
16
5
5
16
5
5
5
5
5
5
11
8
6
12
7
5
5
5
5
5
5
5
2.5
5
20
8
8
126
165
126
126
16
5
5
16
5
5
5
5
5
5
11
8
6
12
7
6
5
5
5
5
5
5
50
18
20
243
243
284
800
284
284
133
17
11
133
22
27
34
15
17
14
39
13
13
44
46
41
12
29
32
38
32
32
97.5
6216
20
4936
4936
1476
1476
1476
1476
2835
267
4205
2835
4205
4205
4205
9483
5404
5404
182
182
182
182
152
3087
837
912
1679
912
1679
1679
max
7141
20
4936
4936
1476
1476
1476
1476
2835
267
4205
2835
4205
4205
4205
9483
5404
9483
182
182
182
182
152
3762
837
912
1679
912
1679
1679
A-11
-------
Table A-2. Distance of location-specific ambient monitors used in Tier 1 analyses to stationary sources emitting > 5 tons of NOX per year, within a 1 0 kilometer
distance of monitoring site.
Location
El Paso
El Paso
Jacksonville
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Las Vegas
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
ID
481410057
481410058
120310032
320030022
320030023
320030073
320030078
320030539
320030557
320030563
320030601
320031019
320032002
060370002
060370016
060370030
060370113
060370206
060371002
060371103
060371201
060371301
060371601
060371701
060372005
060374002
060375001
060375005
060376002
060376012
060379002
060379033
060590001
060590007
060591003
060595001
060650012
060655001
Distance (km) to Source emissions >5 tpy and within 10
km
n
0
16
20
7
0
0
0
5
4
1
0
0
1
7
7
35
7
11
18
31
7
45
22
13
10
55
32
25
5
6
4
4
17
17
14
16
0
0
mean
8.8
5.1
4.6
6.9
9.1
7.6
9.9
3.1
7.5
5.5
4.3
5.6
5.7
6.5
5.1
6.8
6.5
6.1
5.2
6.4
5.1
4.6
5.6
6.2
7.8
6.3
6.4
6.4
6.1
7.9
std
0.4
3.0
0.9
1.2
1.2
1.1
1.8
2.3
3.1
2.2
2.6
2.7
1.2
2.1
2.3
3.0
3.5
2.3
2.4
2.4
1.8
2.5
1.0
0.8
2.4
2.4
2.2
1.6
min
8.4
0.7
3.8
4.7
7.3
7.6
9.9
1.6
4.5
2.1
1.3
2.3
0.1
1.8
3.3
1.2
2.3
1.1
0.2
1.7
0.3
1.4
3.6
3.0
6.8
5.3
2.8
2.8
2.1
3.4
2.5
8.4
0.7
3.8
4.7
7.3
7.6
9.9
1.6
4.5
2.1
1.3
2.3
0.1
1.8
3.3
2.5
2.3
1.1
0.2
2.2
0.3
1.4
3.6
3.0
6.8
5.3
2.8
2.8
2.1
3.4
50
8.6
5.7
3.9
7.2
9.7
7.6
9.9
2.9
8.5
5.2
3.2
5.8
6.0
7.2
5.5
7.1
7.2
7.0
5.5
6.2
4.8
4.6
5.8
6.8
7.7
6.4
7.2
7.2
6.0
8.2
97.5
9.5
9.6
5.6
7.9
9.7
7.6
9.9
4.5
8.9
9.8
9.8
9.2
9.9
10.0
6.5
9.7
9.7
9.7
10.0
9.9
9.6
9.9
7.8
9.7
9.2
7.1
9.4
9.4
9.3
9.5
max
9.5
9.6
5.6
7.9
9.7
7.6
9.9
4.5
8.9
9.8
9.8
9.2
9.9
10.0
6.5
10.0
9.7
9.7
10.0
9.9
9.6
9.9
7.8
9.7
9.2
7.1
9.4
9.4
9.3
9.5
Emissions (tpy) of Sources within 10 km and >5 tpy
mean
31
201
175
816
807
84
84
10
12
23
15
32
47
18
10
22
28
22
12
76
205
224
29
26
22
22
14
14
65
19
std
30
407
222
760
877
4
8
27
10
31
59
21
4
24
33
20
8
159
754
850
20
19
28
28
12
12
116
26
min
5
5
30
18
18
84
84
5
5
5
5
6
6
5
6
5
5
5
5
5
6
6
8
8
6
6
5
5
5
6
2.5
5
5
30
18
18
84
84
5
5
5
5
6
6
5
6
5
5
5
5
5
6
6
8
8
6
6
5
5
5
6
50
23
31
77
851
772
84
84
9
9
11
13
20
24
10
10
12
12
16
9
16
21
21
18
18
9
9
8
8
10
9
97.5
106
1642
650
1665
1665
84
84
16
29
115
36
109
215
86
15
86
115
70
30
744
4256
4256
54
54
64
64
46
46
434
109
max
106
1642
650
1665
1665
84
84
16
29
115
36
109
215
86
15
115
115
70
30
789
4256
4256
54
54
64
64
46
46
434
109
A-12
-------
Table A-2. Distance of location-specific ambient monitors used in Tier 1 analyses to stationary sources emitting > 5 tons of NOX per year, within a 1 0 kilometer
distance of monitoring site.
Location
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Miami
Miami
Miami
Miami
Miami
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
ID
060658001
060659001
060710001
060710012
060710014
060710015
060710017
060710306
060711004
060711234
060712002
060714001
060719004
061110005
061110007
061111003
061111004
061112002
061112003
061113001
120110003
120110031
120118002
120860027
120864002
090010113
090019003
090090027
090091123
340030001
340030005
340130011
340130016
340131003
340170006
340210005
340230011
340273001
Distance (km) to Source emissions >5 tpy and within 10
km
n
12
2
3
0
3
3
0
3
19
2
20
1
8
5
20
0
0
4
3
7
0
0
0
3
8
7
3
5
6
48
18
43
44
32
42
8
20
1
mean
7.4
4.6
6.9
6.0
4.4
6.1
7.3
1.6
5.7
6.5
5.8
6.9
4.7
6.6
5.5
5.1
4.1
7.0
4.4
6.3
2.7
3.3
6.5
6.8
5.4
5.5
6.4
6.9
5.4
6.1
8.5
std
2.2
5.9
1.9
2.6
4.6
2.6
1.7
0.4
2.2
2.5
2.5
2.2
1.0
1.3
2.3
4.2
2.6
3.1
2.0
1.0
2.8
2.2
2.9
2.9
2.8
2.0
2.5
1.7
2.8
min
3.6
0.4
5.3
3.5
1.7
3.6
4.3
1.3
2.0
6.5
1.5
3.1
1.7
5.2
4.1
1.9
1.6
1.3
1.4
4.0
1.3
1.2
2.9
0.1
0.7
0.1
2.1
1.1
3.2
1.0
8.5
2.5
3.6
0.4
5.3
3.5
1.7
3.6
4.3
1.3
2.0
6.5
1.5
3.1
1.7
5.2
4.1
1.9
1.6
1.3
1.4
4.0
1.3
1.2
2.9
0.1
0.8
1.0
2.1
1.6
3.2
1.0
8.5
50
7.4
4.6
6.5
5.9
1.8
5.7
7.4
1.6
5.8
6.5
5.7
7.7
4.2
6.8
5.6
5.9
1.8
7.8
3.4
7.4
2.7
2.4
6.3
7.4
5.8
6.3
6.8
7.7
5.5
7.0
8.5
97.5
9.8
8.7
9.0
8.6
9.7
8.9
9.8
1.9
9.6
6.5
9.0
9.6
9.3
7.5
6.7
7.4
8.9
9.1
8.8
7.5
3.9
8.9
9.8
10.0
9.4
9.4
9.3
9.5
7.3
9.5
8.5
max
9.8
8.7
9.0
8.6
9.7
8.9
9.8
1.9
9.6
6.5
9.0
9.6
9.3
7.5
6.7
7.4
8.9
9.1
8.8
7.5
3.9
8.9
9.9
10.0
9.5
9.6
9.3
9.5
7.3
9.5
8.5
Emissions (tpy) of Sources within 10 km and >5 tpy
mean
119
11
209
199
752
199
57
1122
44
577
171
68
25
63
18
35
31
22
538
127
280
234
468
53
273
267
77
369
115
95
20
std
358
9
321
327
1045
327
120
1168
65
438
118
20
113
4
51
19
15
711
179
484
447
1506
79
1372
1357
149
1420
244
175
min
5
5
10
6
12
6
5
296
5
577
5
8
5
5
14
5
14
8
48
12
14
7
6
6
5
5
5
5
8
6
20
2.5
5
5
10
6
12
6
5
296
5
577
5
8
5
5
14
5
14
8
48
12
14
7
7
6
5
5
5
6
8
6
20
50
10
11
38
15
296
15
18
1122
17
577
10
19
18
7
20
13
27
18
192
37
86
64
31
21
18
18
22
24
32
36
20
97.5
1254
17
579
577
1948
577
492
1948
250
577
1254
278
76
232
22
146
51
51
1689
333
1144
1144
4440
307
640
640
640
2213
718
792
20
max
1254
17
579
577
1948
577
492
1948
250
577
1254
278
76
232
22
146
51
51
1689
333
1144
1144
9022
307
9022
9022
640
9022
718
792
20
A-13
-------
Table A-2. Distance of location-specific ambient monitors used in Tier 1 analyses to stationary sources emitting > 5 tons of NOX per year, within a 1 0 kilometer
distance of monitoring site.
Location
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
New York
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Philadelphia
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Provo
St Louis
St Louis
St Louis
St Louis
St Louis
St Louis
St Louis
ID
340390004
340390008
360050080
360050083
360050110
360470011
360590005
360610010
360610056
360810097
360810098
360810124
361030009
100031003
100031007
100032004
340070003
420170012
420450002
420910013
421010004
421010029
421010047
040130019
040133002
040133003
040133010
040134005
040134011
040139997
490490002
171630010
291830010
291831002
291890001
291890004
291890006
291893001
Distance (km) to Source emissions >5 tpy and within 10
km
n
46
12
54
37
55
56
7
52
54
11
48
24
3
39
11
32
69
10
30
12
32
74
73
11
6
10
10
11
1
10
7
48
1
9
10
6
8
16
mean
6.3
7.2
6.4
6.0
5.9
5.9
6.3
5.9
5.4
6.3
7.1
7.0
3.8
5.5
9.2
4.8
7.7
4.1
4.8
5.1
5.9
5.7
5.2
6.8
4.1
6.7
5.0
5.8
6.4
8.5
6.6
7.0
1.7
7.5
7.7
8.9
7.0
7.3
std
2.4
2.1
2.3
2.8
2.2
2.7
3.4
2.5
2.6
2.1
2.3
2.6
3.2
2.5
0.6
1.9
2.3
2.3
2.6
2.5
2.5
2.1
2.1
2.2
2.3
1.4
0.9
2.9
1.2
3.7
2.8
2.1
1.3
1.5
1.7
2.0
min
0.7
3.2
1.8
1.6
2.1
0.7
1.9
0.3
0.3
2.9
1.6
2.1
2.0
1.6
8.0
0.7
1.8
1.2
0.2
1.4
1.0
1.1
0.6
4.2
1.3
4.1
3.5
0.8
6.4
5.6
1.2
1.3
1.7
4.3
6.2
6.9
4.2
3.4
2.5
0.9
3.2
1.8
1.6
2.6
1.5
1.9
1.4
1.4
2.9
2.8
2.1
2.0
1.6
8.0
0.7
2.0
1.2
0.2
1.4
1.0
1.8
0.8
4.2
1.3
4.1
3.5
0.8
6.4
5.6
1.2
1.9
1.7
4.3
6.2
6.9
4.2
3.4
50
6.6
8.0
6.4
6.3
5.7
5.7
8.1
6.1
5.5
6.9
7.8
8.0
2.0
6.2
9.3
4.7
8.5
4.2
5.4
4.3
5.6
5.6
4.8
6.7
4.1
6.6
4.9
7.0
6.4
8.7
8.2
8.0
1.7
7.7
7.4
9.8
7.9
7.6
97.5
9.6
10.0
9.9
9.9
9.6
9.7
9.8
9.6
9.9
9.5
9.8
10.0
7.6
9.7
9.8
8.4
10.0
9.4
9.5
8.8
9.9
9.7
9.6
9.8
6.9
9.0
6.6
9.4
6.4
9.9
9.4
9.8
1.7
9.9
9.8
10.0
8.7
9.6
max
9.7
10.0
9.9
9.9
9.9
10.0
9.8
9.8
10.0
9.5
9.8
10.0
7.6
9.7
9.8
8.4
10.0
9.4
9.5
8.8
9.9
9.7
9.7
9.8
6.9
9.0
6.6
9.4
6.4
9.9
9.4
9.9
1.7
9.9
9.8
10.0
8.7
9.6
Emissions (tpy) of Sources within 10 km and >5 tpy
mean
134
23
241
171
236
296
372
494
470
65
262
436
537
282
323
223
87
85
504
89
58
74
95
106
21
50
115
81
18
115
60
112
7821
1868
24
38
25
22
std
341
36
776
725
769
787
500
1453
1429
77
820
1136
759
481
494
403
196
96
1055
232
111
148
221
313
19
80
328
116
328
38
178
4704
20
37
34
43
min
5
5
6
6
6
7
7
5
7
13
6
8
40
5
6
5
5
11
5
5
5
5
5
5
5
9
5
6
18
5
7
5
7821
7
5
7
6
5
2.5
6
5
6
6
6
7
7
7
7
13
7
8
40
5
6
5
5
11
5
5
5
5
5
5
5
9
5
6
18
5
7
5
7821
7
5
7
6
5
50
21
10
29
21
29
42
223
50
50
26
31
26
161
62
63
45
24
57
73
12
20
19
19
10
15
24
10
38
18
10
83
17
7821
8
15
28
11
11
97.5
594
134
3676
4440
3676
3676
1451
4440
4440
246
3676
4440
1410
2058
1351
1312
477
275
4968
823
571
477
1033
1049
56
272
1049
350
18
1049
102
538
7821
14231
60
105
105
181
max
2213
134
4440
4440
4440
4440
1451
9022
9022
246
4440
4440
1410
2058
1351
1312
1478
275
4968
823
571
1033
1478
1049
56
272
1049
350
18
1049
102
848
7821
14231
60
105
105
181
A-14
-------
Table A-2. Distance of location-specific ambient monitors used in Tier 1 analyses to stationary sources emitting > 5 tons of NOX per year, within a 1 0 kilometer
distance of monitoring site.
Location
St Louis
St Louis
St Louis
St Louis
St Louis
St Louis
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
Washington DC
ID
291895001
291897002
291897003
295100072
295100080
295100086
110010017
110010025
110010041
110010043
240053001
245100040
245100050
510130020
510590005
510590018
510591004
510591005
510595001
511071005
511530009
515100009
Distance (km) to Source emissions >5 tpy and within 10
km
n
11
16
16
46
31
35
13
6
10
12
11
26
24
14
2
6
10
8
4
5
0
9
mean
7.5
5.7
6.2
6.3
6.9
6.7
5.4
6.4
6.1
5.0
7.5
5.0
6.2
6.2
4.9
8.4
7.4
6.3
6.5
7.1
7.0
std
1.7
1.8
2.0
2.5
2.2
2.3
2.4
1.0
2.4
3.2
2.1
2.5
2.1
2.6
4.8
0.4
1.6
2.0
2.8
2.3
2.4
min
4.3
2.0
2.5
0.7
0.4
1.7
2.9
4.8
0.6
0.3
2.6
0.3
2.4
1.5
1.4
8.0
3.7
4.6
3.2
4.5
1.1
2.5
4.3
2.0
2.5
2.0
0.4
1.7
2.9
4.8
0.6
0.3
2.6
0.3
2.4
1.5
1.4
8.0
3.7
4.6
3.2
4.5
1.1
50
7.7
5.4
6.0
6.5
7.3
6.6
4.5
6.5
6.1
4.6
7.9
4.9
6.0
5.4
4.9
8.4
7.8
5.5
6.8
6.5
7.9
97.5
9.7
9.7
9.6
9.9
10.0
9.9
9.7
7.6
9.8
9.8
9.7
9.5
10.0
9.8
8.3
9.2
9.3
9.4
9.2
9.6
8.8
max
9.7
9.7
9.6
9.9
10.0
9.9
9.7
7.6
9.8
9.8
9.7
9.5
10.0
9.8
8.3
9.2
9.3
9.4
9.2
9.6
8.8
Emissions (tpy) of Sources within 10 km and >5 tpy
mean
46
28
24
77
98
94
557
40
124
109
1034
122
129
558
13
1104
80
94
30
14
809
std
62
37
33
150
176
168
1643
35
137
129
3225
220
227
1579
7
2413
173
193
19
8
1959
min
5
5
5
5
5
5
11
11
11
11
6
6
6
11
8
9
14
14
17
8
14
2.5
5
5
5
5
5
5
11
11
11
11
6
6
6
11
8
9
14
14
17
8
14
50
15
15
15
16
17
17
34
26
66
46
45
56
56
46
13
13
19
19
22
12
156
97.5
181
143
143
508
848
848
6009
98
410
410
10756
1118
1118
6009
18
6009
571
571
58
27
6009
max
181
143
143
848
848
848
6009
98
410
410
10756
1118
1118
6009
18
6009
571
571
58
27
6009
A-15
-------
Appendix B. Temporal Air Quality Characterization
Appendix B contains the ambient air quality analysis results by year for each of the named locations. Boxplots
were constructed to display the annual average and hourly concentration distributions across years for a single
location. The box extends from the 25th to the 75th percentile, with the median shown as the line inside the box. The
whiskers extend from the box to the 5th and 95th percentiles. The extreme values in the upper and lower tails beyond
the 5th and 95th percentiles are not shown to allow for similar scaling along the y-axis for the plotted independent
variables. The mean is plotted as a dot; typically it would appear inside the box, however it will fall outside the box
if the distribution is highly skewed. All concentrations are shown in parts per billion (ppb). The boxplots for hourly
concentrations were created using a different procedure than for the annual statistics, because of the large number of
hourly values and the inability of the graphing procedure to allow frequency weights. Therefore, the appropriate
weighted percentiles and means were calculated and plotted as shown, but the vertical lines composing the sides of
the box are essentially omitted. Tables are provided that summarize the complete distribution, with percentiles
given in segments of 10.
B-1
-------
List of Figures
Figure B-l. Temporal distribution of annual average NO2 ambient concentrations, Boston CMSA, years 1995-2006 4
Figure B-2. Temporal distribution of hourly NO2 ambient concentrations, Boston CMSA, years 1995-2006 4
Figure B-3. Temporal distribution of annual average NO2 ambient concentrations, Chicago CMSA, years 1995-2006 6
Figure B-4. Temporal distribution of hourly NO2 ambient concentrations, Chicago CMSA, years 1995-2006 6
Figure B-5. Temporal distribution of annual average NO2 ambient concentrations, Cleveland CMSA, years 1995-2006 7
Figure B-6. Temporal distribution of hourly NO2 ambient concentrations, Cleveland CMSA, years 1995-2006 8
Figure B-7. Temporal distribution of annual average NO2 ambient concentrations, Denver CMSA, years 1995-2006 9
Figure B-8. Temporal distribution of hourly NO2 ambient concentrations, Denver CMSA, years 1995-2006 10
Figure B-9. Temporal distribution of annual average NO2 ambient concentrations, Detroit CMSA, years 1995-2006 12
Figure B-10. Temporal distribution of hourly NO2 ambient concentrations, Detroit CMSA, years 1995-2006 12
Figure B-ll. Temporal distribution of annual average NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006 14
Figure B-12. Temporal distribution of hourly NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006 14
Figure B-13. Temporal distribution of annual average NO2 ambient concentrations, Miami CMSA, years 1995-2006 16
Figure B-14. Temporal distribution of hourly NO2 ambient concentrations, Miami CMSA, years 1995-2006 16
Figure B-15. Temporal distribution of annual average NO2 ambient concentrations, New York CMSA, years 1995-2006 18
Figure B-16. Temporal distribution of hourly NO2 ambient concentrations, New York CMSA, years 1995-2006 18
Figure B-17. Temporal distribution of annual average NO2 ambient concentrations, Philadelphia CMSA, years 1995-2006 20
Figure B-18. Temporal distribution of hourly NO2 ambient concentrations, Philadelphia CMSA, years 1995-2006 20
Figure B-19. Temporal distribution of annual average NO2 ambient concentrations, Washington DC CMSA, years 1995-2006 22
Figure B-20. Temporal distribution of hourly NO2 ambient concentrations, Washington DC CMSA, years 1995-2006 22
Figure B-21. Temporal distribution of annual average NO2 ambient concentrations, Atlanta MSA, years 1995-2006 24
Figure B-22. Temporal distribution of hourly NO2 ambient concentrations, Atlanta MSA, years 1995-2006 24
Figure B-23. Temporal distribution of annual average NO2 ambient concentrations, Colorado Springs MSA, years 1995-2006 26
Figure B-24. Temporal distribution of hourly NO2 ambient concentrations, Colorado Springs MSA, years 1995-2006 26
Figure B-25. Temporal distribution of annual average NO2 ambient concentrations, El Paso MSA, years 1995-2006 28
Figure B-26. Temporal distribution of hourly NO2 ambient concentrations, El Paso MSA, years 1995-2006 28
Figure B-27. Temporal distribution of annual average NO2 ambient concentrations, Jacksonville MSA, years 1995-2006 30
Figure B-28. Temporal distribution of hourly NO2 ambient concentrations, Jacksonville MSA, years 1995-2006 30
Figure B-29. Temporal distribution of annual average NO2 ambient concentrations, Las Vegas MSA, years 1995-2006 32
Figure B-30. Temporal distribution of hourly NO2 ambient concentrations, Las Vegas MSA, years 1995-2006 32
Figure B-31. Temporal distribution of annual average NO2 ambient concentrations, Phoenix MSA, years 1995-2006 34
Figure B-32. Temporal distribution of hourly NO2 ambient concentrations, Phoenix MSA, years 1995-2006 34
Figure B-33. Temporal distribution of annual average NO2 ambient concentrations, Provo MSA, years 1995-2006 36
Figure B-34. Temporal distribution of hourly NO2 ambient concentrations, Provo MSA, years 1995-2006 36
Figure B-35. Temporal distribution of annual average NO2 ambient concentrations, St. Louis MSA, years 1995-2006 38
Figure B-36. Temporal distribution of hourly NO2 ambient concentrations, St. Louis MSA, years 1995-2006 38
Figure B-37. Temporal distribution of annual average NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006 40
Figure B-38. Temporal distribution of hourly NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006 40
Figure B-39. Temporal distribution of annual average NO2 ambient concentrations, Other Not MSA, years 1995-2006 42
Figure B-40. Temporal distribution of hourly NO2 ambient concentrations, Other Not MSA, years 1995-2006 42
B-2
-------
List of Tables
Table B-l. Temporal distribution of annual average NO2 ambient concentrations, Boston CMS A, years 1995-2006 5
Table B-2. Temporal distribution of hourly NO2 ambient concentrations, Boston CMSA, years 1995-2006 5
Table B-3. Temporal distribution of annual average NO2 ambient concentrations, Chicago CMSA, years 1995-2006 7
Table B-4. Temporal distribution of hourly NO2 ambient concentrations, Chicago CMSA, years 1995-2006 7
Table B-5. Temporal distribution of annual average NO2 ambient concentrations, Cleveland CMSA, years 1995-2006 8
Table B-6. Temporal distribution of hourly NO2 ambient concentrations, Cleveland CMSA, years 1995-2006 8
Table B-7. Temporal distribution of annual average NO2 ambient concentrations, Denver CMSA, years 1995-2006 11
Table B-8. Temporal distribution of hourly NO2 ambient concentrations, Denver CMSA, years 1995-2006 11
Table B-9. Temporal distribution of annual average NO2 ambient concentrations, Detroit CMSA, years 1995-2006 13
Table B-10. Temporal distribution of hourly NO2 ambient concentrations, Detroit CMSA, years 1995-2006 13
Table B-ll. Temporal distribution of annual average NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006 15
Table B-12. Temporal distribution of hourly NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006 15
Table B-13. Temporal distribution of annual average NO2 ambient concentrations, Miami CMSA, years 1995-2006 17
Table B-14. Temporal distribution of hourly NO2 ambient concentrations, Miami CMSA, years 1995-2006 17
Table B-15. Temporal distribution of annual average NO2 ambient concentrations, New York CMSA, years 1995-2006 19
Table B-16. Temporal distribution of hourly NO2 ambient concentrations, New York CMSA, years 1995-2006 19
Table B-17. Temporal distribution of annual average NO2 ambient concentrations, Philadelphia CMSA, years 1995-2006 21
Table B-18. Temporal distribution of hourly NO2 ambient concentrations, Philadelphia CMSA, years 1995-2006 21
Table B-19. Temporal distribution of annual average NO2 ambient concentrations, Washington DC CMSA, years 1995-2006 23
Table B-20. Temporal distribution of hourly NO2 ambient concentrations, WashingtonDC CMSA, years 1995-2006 23
Table B-21. Temporal distribution of annual average NO2 ambient concentrations, Atlanta MSA, years 1995-2006 25
Table B-22. Temporal distribution of hourly NO2 ambient concentrations, Atlanta MSA, years 1995-2006 25
Table B-23. Temporal distribution of annual average NO2 ambient concentrations, Colorado Springs MSA, years 1995-2006 27
Table B-24. Temporal distribution of hourly NO2 ambient concentrations, Colorado Springs MSA, years 1995-2006 27
Table B-25. Temporal distribution of annual average NO2 ambient concentrations, El Paso MSA, years 1995-2006 29
Table B-26. Temporal distribution of hourly NO2 ambient concentrations, El Paso MSA, years 1995-2006 29
Table B-27. Temporal distribution of annual average NO2 ambient concentrations, Jacksonville MSA, years 1995-2006 30
Table B-28. Temporal distribution of hourly NO2 ambient concentrations, Jacksonville MSA, years 1995-2006 31
Table B-29. Temporal distribution of annual average NO2 ambient concentrations, Las Vegas MSA, years 1995-2006 33
Table B-30. Temporal distribution of hourly NO2 ambient concentrations, Las Vegas MSA, years 1995-2006 33
Table B-31. Temporal distribution of annual average NO2 ambient concentrations, Phoenix MSA, years 1995-2006 35
Table B-32. Temporal distribution of hourly NO2 ambient concentrations, Phoenix MSA, years 1995-2006 35
Table B-33. Temporal distribution of annual average NO2 ambient concentrations, Provo MSA, years 1995-2006 37
Table B-34. Temporal distribution of hourly NO2 ambient concentrations, Provo MSA, years 1995-2006 37
Table B-35. Temporal distribution of annual average NO2 ambient concentrations, St. Louis MSA, years 1995-2006 39
Table B-36. Temporal distribution of hourly NO2 ambient concentrations, St. Louis MSA, years 1995-2006 39
Table B-37. Temporal distribution of annual average NO2 ambient concentrations, Other MS A/CMS A, years 1995-2006 41
Table B-38. Temporal distribution of hourly NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006 41
Table B-39. Temporal distribution of annual average NO2 ambient concentrations, Other Not MSA, years 1995-2006 43
Table B-40. Temporal distribution of hourly NO2 ambient concentrations, Other Not MSA, years 1995-2006 43
B-3
-------
Annual Mean
40-
Annual Mean
loc_type=CMSA loc_name=Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-1. Temporal distribution of annual average NO2 ambient concentrations, Boston CMSA, years 1995-
2006.
Hourly Cones
50-
Hourly Concentrations
loc_type=CMSA loc_name=Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
0—
I'»5 1006 K»7 l'»8 10«} 2000 2001 2002 200.1 2004 2005 2005
Veur
Figure B-2. Temporal distribution of hourly NO2 ambient concentrations, Boston CMSA, years 1995-2006.
B-4
-------
Table B-1. Temporal distribution of annual average NO2 ambient concentrations, Boston CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
12
10
11
11
7
7
10
10
5
7
8
7
Mean
20
19
17
17
19
17
16
16
15
15
14
13
SD
7
8
8
8
9
8
8
7
6
6
6
6
cov
34
42
44
48
45
49
50
43
42
41
39
42
Min
6
6
6
6
6
5
7
5
9
7
7
8
p10
12
8
9
8
6
5
7
7
9
7
7
8
p20
14
11
11
10
9
10
8
10
10
9
10
9
p30
16
14
13
12
20
11
10
12
11
12
10
10
p40
21
17
15
15
20
11
12
13
11
12
11
10
p50
22
19
16
15
21
18
16
15
12
16
13
10
p60
22
21
19
19
21
20
20
19
17
16
15
15
p70
23
24
22
23
21
20
22
22
21
16
18
15
p80
23
26
22
23
27
22
24
24
22
17
19
19
p90
27
29
27
28
30
29
28
25
22
25
23
23
Max
31
31
30
31
30
29
30
25
22
25
23
23
Table B-2. Temporal distribution of hourly NO2 ambient concentrations, Boston CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
99946
83541
90161
89710
54043
56196
82048
80472
41198
56831
66244
57681
Mean
20
19
17
17
19
16
16
16
15
15
14
13
SD
12
14
12
13
13
12
13
12
11
10
11
10
COV
62
72
72
75
70
76
77
75
75
71
75
74
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
5
3
3
3
3
2
2
2
3
3
3
3
p20
9
7
6
5
7
5
4
5
5
5
5
4
p30
12
10
9
8
10
7
7
7
7
7
7
6
p40
15
13
11
11
13
11
10
10
10
10
9
8
p50
18
16
15
15
17
14
14
14
13
12
12
11
p60
22
21
18
18
21
18
18
17
16
15
15
14
p70
26
25
23
23
25
22
22
21
19
19
18
17
p80
30
30
28
28
30
27
27
26
24
23
23
22
p90
36
38
35
35
37
34
34
32
31
29
29
28
Max
100
205
134
112
117
95
114
93
99
96
113
79
B-5
-------
Annual Mean
loc_type=CMSA loc_name=Chicago-Gary-Kenosha, IL-IN-WI CMSA
Annual Mean
40-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-3. Temporal distribution of annual average NO2 ambient concentrations, Chicago CMSA, years 1995-
2006.
Hourly Concentrations
loc_type=CMSA loc_name=Chicago-Gary-Kenosha, IL-IN-WI CMSA
Hourly Cones
60-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-4. Temporal distribution of hourly NO2 ambient concentrations, Chicago CMSA, years 1995-2006.
B-6
-------
Table B-3. Temporal distribution of annual average NO2 ambient concentrations, Chicago CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
7
7
6
9
9
9
7
7
5
6
6
5
Mean
28
24
25
23
23
22
25
24
26
23
23
23
SD
3
8
8
7
7
7
5
6
5
6
5
6
cov
12
32
34
32
29
30
21
24
19
25
23
27
Min
23
9
9
9
10
9
18
17
20
16
17
16
p10
23
9
9
9
10
9
18
17
20
16
17
16
p20
24
21
23
17
17
18
18
19
21
18
18
17
p30
26
23
23
19
19
20
24
22
22
18
18
18
p40
26
23
24
23
22
21
24
22
25
20
20
20
p50
27
24
25
24
24
22
25
22
27
22
22
22
p60
29
28
27
25
24
23
28
23
29
24
24
25
p70
29
28
31
26
27
27
28
23
30
29
28
28
p80
30
31
31
31
31
29
28
30
31
29
28
29
p90
32
32
34
32
32
32
32
32
31
29
30
31
Max
32
32
34
32
32
32
32
32
31
29
30
31
Table B-4. Temporal distribution of hourly NO2 ambient concentrations, Chicago CMSA, years 1995-2006
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
58998
59447
51443
76365
74985
75327
58268
58383
42406
49210
51043
42009
Mean
28
24
25
23
23
22
25
24
26
23
23
23
SD
14
14
15
14
14
14
13
14
14
13
13
13
COV
51
58
59
61
61
62
54
59
54
57
59
57
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
11
7
7
6
7
6
9
8
10
8
8
8
p20
15
11
11
10
10
10
13
12
14
11
11
11
p30
19
15
15
13
13
13
16
15
17
14
14
14
p40
22
18
19
17
17
17
20
18
21
18
17
17
p50
26
22
23
21
21
20
23
21
24
21
21
21
p60
29
26
27
25
25
24
27
25
28
25
24
25
p70
33
31
31
29
30
29
31
29
32
28
29
29
p80
38
36
36
34
35
34
36
34
37
33
34
34
p90
47
43
44
41
42
41
43
42
45
41
41
41
Max
113
127
113
112
113
108
114
149
122
101
106
137
Annual Mean
loc_type=CMSA loc_name=Cleveland-Akron, OH CMSA
Annual Mean
20-
Ift-
17-
16-
15-
14-
1005 1')% 1007 1008 1000 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-5. Temporal distribution of annual average NO2 ambient concentrations, Cleveland CMSA, years 1995-2006.
B-7
-------
Hourly Concentrations
loc_type=CMSAloc_name=Cleveland-Akron, OH CMSA
Hourly Cones
70-
0-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-6. Temporal distribution of hourly NO2 ambient concentrations, Cleveland CMSA, years 1995-2006.
Table B-5. Temporal distribution of annual average NO2 ambient concentrations, Cleveland CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
2
2
1
2
2
2
2
2
2
1
2
2
Mean
24
23
28
24
21
20
21
20
20
22
19
16
SD
5
4
5
5
4
4
4
3
3
3
cov
19
19
0
22
26
19
17
18
15
0
17
17
Min
21
20
28
20
17
18
19
17
17
22
17
14
p10
21
20
28
20
17
18
19
17
17
22
17
14
p20
21
20
28
20
17
18
19
17
17
22
17
14
p30
21
20
28
20
17
18
19
17
17
22
17
14
p40
21
20
28
20
17
18
19
17
17
22
17
14
p50
24
23
28
24
21
20
21
20
20
22
19
16
p60
27
26
28
27
25
23
24
22
22
22
22
18
p70
27
26
28
27
25
23
24
22
22
22
22
18
p80
27
26
28
27
25
23
24
22
22
22
22
18
p90
27
26
28
27
25
23
24
22
22
22
22
18
Max
27
26
28
27
25
23
24
22
22
22
22
18
Table B-6. Temporal distribution of hourly NO2 ambient concentrations, Cleveland CMSA,
/ears 1995-2006.
Year
1995
1996
1997
1998
1999
2000
n
16042
16593
8300
16680
16743
16399
Mean
24
23
28
24
21
20
SD
13
12
17
13
12
11
COV
53
52
59
53
58
55
Min
2
1
0
0
0
0
p10
10
9
12
9
7
8
p20
13
13
15
13
10
10
p30
16
15
18
16
13
13
p40
19
18
21
19
16
16
p50
22
21
24
22
19
19
p60
25
24
28
25
22
22
p70
29
28
32
29
26
25
p80
34
32
38
33
30
30
p90
41
39
49
40
37
36
Max
108
148
253
89
86
74
B-8
-------
2001
2002
2003
2004
2005
2006
16566
16464
16948
8484
16558
16853
21
20
20
22
19
16
12
11
11
11
12
10
56
56
57
51
60
64
0
1
0
0
0
0
8
8
7
10
7
5
10
10
10
13
9
8
13
12
13
15
12
10
16
15
15
18
14
12
19
18
18
20
17
14
22
21
20
23
20
16
26
24
24
26
23
20
30
28
28
30
28
24
37
35
35
37
35
30
103
88
90
83
85
175
Annual Mean
40-
Annual Mean
loc_type=CMSA loc_name=Denver-Boulder-Oreeley, CO CMSA
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-7. Temporal distribution of annual average NO2 ambient concentrations, Denver CMSA, years 1995-2006.
B-9
-------
Hourly Concentrations
loc_type=CMSA loc_name=Denver-Boulder-Greeley, CO CMSA
Hourly Cones
60-
1'WS !<>% ll»7 IWg IQW 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-8. Temporal distribution of hourly NO2 ambient concentrations, Denver CMSA, years 1995-2006.
B-10
-------
Table B-7. Temporal distribution of annual average NO2 ambient concentrations, Denver CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
3
6
6
5
3
3
2
1
1
2
2
2
Mean
28
14
15
16
12
12
31
35
21
24
24
24
SD
6
11
11
13
6
3
8
4
5
8
cov
23
77
74
77
52
26
25
0
0
17
21
33
Min
23
6
6
7
8
9
26
35
21
21
20
18
p10
23
6
6
7
8
9
26
35
21
21
20
18
p20
23
7
8
7
8
9
26
35
21
21
20
18
p30
23
7
8
7
8
9
26
35
21
21
20
18
p40
26
8
9
8
9
10
26
35
21
21
20
18
p50
26
8
9
9
9
10
31
35
21
24
24
24
p60
26
9
9
16
9
10
37
35
21
27
28
29
p70
35
22
23
23
19
15
37
35
21
27
28
29
p80
35
22
23
29
19
15
37
35
21
27
28
29
p90
35
33
34
35
19
15
37
35
21
27
28
29
Max
35
33
34
35
19
15
37
35
21
27
28
29
Table B-8. Temporal distribution of hourly NO2 ambient concentrations, Denver CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
23204
46816
45049
40258
23164
24649
15204
7688
6989
15878
15467
13775
Mean
28
14
15
17
12
12
31
35
21
24
24
24
SD
17
15
15
17
13
13
17
13
17
15
16
15
COV
62
108
106
100
108
108
55
36
78
60
65
65
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
6
1
1
1
0
0
8
20
3
4
3
3
p20
11
2
3
3
2
1
15
24
5
10
8
7
p30
16
4
4
5
4
3
21
28
8
16
14
13
p40
22
6
6
7
6
5
27
31
13
20
19
19
p50
27
8
8
10
8
8
32
34
18
24
24
24
p60
32
11
12
15
10
10
36
38
25
28
29
28
p70
36
16
17
22
14
14
41
41
31
32
33
33
p80
41
25
26
31
21
19
45
45
37
37
38
38
p90
48
37
39
42
33
30
52
51
44
43
44
44
Max
286
137
141
148
96
141
157
159
136
115
114
169
B-11
-------
Annual Mean
20-
19-
18-
17-
16-
15-
14-
Annual Mean
loc_type=CMSA loc_name=Detroit-Ann Arbor-Flint, MI CMSA
IW5 !')% IW IW8 I'XN 2000 2001 2002 200 j 200-t 2005 2006
Year
Figure B-9. Temporal distribution of annual average NO2 ambient concentrations, Detroit CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Detroit-Ann Arbor-Flrnt, MI CMSA
Hourly Cones
50-
2000 200 1 2002 2MB 20I» 2005 2006
Figure B-10. Temporal distribution of hourly NO2 ambient concentrations, Detroit CMSA, years 1995-2006.
B-12
-------
Table B-9. Temporal distribution of annual average NO2 ambient concentrations, Detroit CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
2
3
3
2
1
1
2
2
2
2
2
2
Mean
20
18
17
21
18
24
21
20
20
17
18
15
SD
2
5
8
3
3
1
2
3
2
1
cov
10
28
44
14
0
0
14
7
12
16
9
9
Min
19
12
13
19
18
24
19
19
19
15
17
14
p10
19
12
13
19
18
24
19
19
19
15
17
14
p20
19
12
13
19
18
24
19
19
19
15
17
14
p30
19
12
13
19
18
24
19
19
19
15
17
14
p40
19
20
13
19
18
24
19
19
19
15
17
14
p50
20
20
13
21
18
24
21
20
20
17
18
15
p60
22
20
13
23
18
24
23
21
22
19
20
16
p70
22
21
26
23
18
24
23
21
22
19
20
16
p80
22
21
26
23
18
24
23
21
22
19
20
16
p90
22
21
26
23
18
24
23
21
22
19
20
16
Max
22
21
26
23
18
24
23
21
22
19
20
16
Table B-10. Temporal distribution of hourly NO2 ambient concentrations, Detroit CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
16629
23600
24117
14863
7110
8590
15154
16623
16569
14779
15827
17273
Mean
20
18
17
21
18
24
21
20
20
17
19
15
SD
12
13
16
14
13
13
13
15
13
11
12
10
COV
58
74
94
68
73
56
61
73
62
66
63
64
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
8
4
2
5
4
8
7
7
7
5
6
4
p20
10
7
5
9
7
12
9
10
9
7
8
6
p30
12
9
7
12
9
15
12
12
12
9
10
8
p40
15
12
10
15
12
19
15
15
15
12
13
10
p50
18
15
13
18
15
22
19
18
18
14
16
13
p60
21
18
16
22
19
26
23
22
21
17
19
16
p70
25
22
21
27
24
30
27
25
25
21
23
19
p80
29
27
26
31
29
35
32
30
30
26
28
23
p90
35
35
36
39
36
42
38
36
36
33
35
29
Max
117
167
322
136
104
128
194
443
139
78
84
58
B-13
-------
Annual Mean
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA
Annual Mean
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-11. Temporal distribution of annual average NO2 ambient concentrations, Los Angeles CMSA, years 1995-
2006.
Hourly Concentrations
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA
Hourly Cones
70-
W5 19% IW7 I'WS 1 2000 2001 2002 2003 2004 2005 2006
Yctir
Figure B-12. Temporal distribution of hourly NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006.
B-14
-------
Table B-11. Temporal distribution of annual average NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
36
29
33
32
31
32
31
32
32
28
28
26
Mean
29
25
25
25
27
25
25
24
23
22
21
19
SD
13
12
12
11
12
11
11
9
9
7
7
7
cov
47
46
47
44
44
43
43
39
37
33
34
35
Min
5
4
4
4
5
4
4
5
5
5
5
5
p10
8
6
8
9
10
10
9
10
11
13
12
9
p20
18
15
14
16
18
16
17
16
15
15
14
13
p30
20
17
16
19
20
20
19
18
18
17
16
15
p40
23
21
20
21
23
22
24
22
21
20
19
17
p50
30
28
26
26
28
25
24
24
24
21
21
19
p60
37
31
29
33
32
28
27
25
26
24
22
20
p70
39
35
33
34
35
32
33
29
29
27
25
23
p80
45
38
34
36
39
36
36
33
31
30
27
25
p90
46
41
42
39
39
39
37
36
34
31
31
27
Max
46
42
43
43
51
44
41
40
35
34
31
30
Table B-12. Temporal distribution of hourly NO2 ambient concentrations, Los Angeles CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
290519
232203
263050
257541
253401
263311
251895
258452
259935
225075
227769
184205
Mean
29
26
25
25
27
25
25
24
23
22
21
19
SD
22
19
19
19
20
18
18
17
17
15
14
14
COV
78
74
75
74
73
72
71
71
72
70
69
74
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
6
5
4
5
5
5
5
5
4
4
4
3
p20
9
8
7
8
8
8
8
8
7
7
7
6
p30
14
12
11
12
13
12
12
11
11
11
11
9
p40
19
17
16
17
18
17
17
16
15
15
15
12
p50
25
22
21
22
24
23
23
21
20
20
19
16
p60
30
28
27
28
30
28
28
26
25
25
23
20
p70
37
34
33
34
37
34
33
32
31
29
28
25
p80
45
40
40
40
43
40
39
38
37
35
33
31
p90
57
50
50
50
54
50
48
46
45
42
40
38
Max
239
250
200
255
307
214
251
262
163
157
136
107
B-15
-------
Annual Mean
17-
Annual Mean
loc_type=CMSA loc_name=Miami-Fort Lauderdale, FL CMSA
H)<» 2000 200! 2002 200.1 2004
2005
Figure B-13. Temporal distribution of annual average NO2 ambient concentrations, Miami CMSA, years 1995-2006.
Hourly Cones
Hourly Concentrations
loc_type=CMSA loc_name=Miami-Fort Lauderdale, FL CMSA
100? 1008 1000 2000 2001
2004 2005
Figure B-14. Temporal distribution of hourly NO2 ambient concentrations, Miami CMSA, years 1995-2006.
B-16
-------
Table B-13. Temporal distribution of annual average NO2 ambient concentrations, Miami CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
n
4
4
4
4
4
4
4
4
4
4
4
Mean
10
10
10
10
11
10
10
9
9
9
9
SD
3
4
4
4
4
4
4
4
3
3
3
cov
31
43
43
42
42
37
42
39
29
36
38
Min
7
6
7
6
6
7
6
6
7
6
6
p10
7
6
7
6
6
7
6
6
7
6
6
p20
7
6
7
6
6
7
6
6
7
6
6
p30
9
8
9
9
10
9
9
7
8
8
7
p40
9
8
9
9
10
9
9
7
8
8
7
p50
10
9
9
9
10
9
9
8
9
8
8
p60
10
9
9
9
10
10
10
9
9
8
8
p70
10
9
9
9
10
10
10
9
9
8
8
p80
15
16
17
15
17
16
16
14
13
13
14
p90
15
16
17
15
17
16
16
14
13
13
14
Max
15
16
17
15
17
16
16
14
13
13
14
Table B-14. Temporal distribution of hourly NO2 ambient concentrations, Miami CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
n
32713
33086
32754
30849
32721
31833
33063
33755
31031
33625
32342
Mean
10
10
10
10
11
10
10
9
9
9
9
SD
10
10
10
10
11
10
10
9
9
10
10
COV
95
103
97
98
99
99
98
96
97
117
109
Min
0
0
0
0
0
0
0
0
0
0
0
p10
1
1
1
1
1
1
1
1
1
1
0
p20
2
2
2
2
2
2
2
2
2
2
1
p30
3
3
3
3
3
4
3
3
3
2
2
p40
5
4
5
5
5
5
5
4
4
4
4
p50
7
6
7
7
7
7
7
6
6
5
6
p60
10
9
10
10
10
10
10
9
8
7
8
p70
13
12
13
12
14
13
13
12
11
10
11
p80
18
17
18
16
18
17
17
16
15
14
15
p90
25
25
25
23
26
24
24
22
21
21
22
Max
75
96
94
69
128
203
86
80
85
417
94
B-17
-------
Annual Mean
loc_type=CMSA loc_name=New York-Northern New Jersey-Long Island, N Y-NJ-CT-PA CMS
Annual Mean
50-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-15. Temporal distribution of annual average NO2 ambient concentrations, New York CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=New York-Northern New Jersey-Long Island, NY-NJ-CT-PA CMS
Hourly Cones
60-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-16. Temporal distribution of hourly NO2 ambient concentrations, New York CMSA, years 1995-2006.
B-18
-------
Table B-15. Temporal distribution of annual average NO2 ambient concentrations, New York CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
16
15
16
14
16
16
14
17
15
14
16
5
Mean
28
28
27
27
27
26
25
25
23
21
23
25
SD
8
8
8
9
9
8
8
8
6
7
7
6
cov
28
29
30
34
31
32
32
31
28
31
31
23
Min
12
12
12
11
11
11
11
11
12
10
11
18
p10
16
17
17
15
17
16
17
16
14
13
13
18
p20
24
22
23
18
19
18
17
17
16
14
16
21
p30
25
26
24
22
24
19
21
20
18
17
18
23
p40
26
27
26
27
26
25
24
22
21
20
22
24
p50
29
27
27
28
27
26
26
25
25
21
22
25
p60
30
29
29
30
29
29
27
28
26
24
25
26
p70
31
32
31
33
33
30
27
28
27
24
27
26
p80
33
34
35
36
33
32
31
29
29
28
27
30
p90
39
41
40
40
41
38
38
38
30
30
32
34
Max
42
42
41
42
42
41
40
40
32
30
36
34
Table B-16. Temporal distribution of hourly NO2 ambient concentrations, New York CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
133504
122074
131144
116748
132646
134037
114478
141480
122724
115578
133856
42223
Mean
28
28
27
27
27
26
25
24
23
21
23
25
SD
16
16
15
16
16
15
15
15
14
13
14
13
COV
56
57
56
58
57
58
61
60
61
64
63
51
Min
0
0
0
0
0
0
0
0
0
0
1
0
p10
9
8
9
8
8
8
7
7
6
5
6
10
p20
14
13
13
13
13
12
10
11
10
8
9
13
p30
18
18
17
17
17
16
15
14
13
12
13
17
p40
22
22
22
22
22
20
19
18
16
15
16
20
p50
26
26
26
26
26
24
23
23
20
19
20
24
p60
31
31
30
31
30
28
28
27
25
23
24
28
p70
35
35
35
35
35
33
33
32
29
27
29
32
p80
40
40
40
40
40
38
38
37
35
32
35
37
p90
48
48
47
48
48
46
45
44
42
40
42
43
Max
162
162
181
240
148
118
142
129
138
156
119
92
B-19
-------
Annual Mean
loc_type=CMSA loc_name=Philadelphia-Wilmington-Atlantic City, PA-NJ-DE-MD CMSA
Annual Mean
.14-
18-
17-
16-
15-
IW5 !')% IW l<»8 I'XW 2000 2001 2002 200 j 2004 2005 2006
Year
Figure B-17. Temporal distribution of annual average NO2 ambient concentrations, Philadelphia CMSA, years 1995-
2006.
Hourly Concentrations
loc_type=CMSA loc_name=Philadelphia-Wilmmgton-Atlantic City, PA-NJ-DE-MD CMSA
Hourly Cones
50-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-18. Temporal distribution of hourly NO2 ambient concentrations, Philadelphia CMSA, years 1995-2006.
B-20
-------
Table B-17. Temporal distribution of annual average NO2 ambient concentrations, Philadel
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
8
8
8
8
8
6
7
8
6
7
7
4
Mean
23
25
24
24
23
21
23
21
20
20
19
16
SD
6
6
6
7
6
4
5
5
4
4
4
1
cov
27
24
25
30
28
20
24
26
19
22
19
9
Min
15
19
18
16
16
17
16
15
16
14
16
14
p10
15
19
18
16
16
17
16
15
16
14
16
14
p20
17
21
19
18
17
18
18
16
17
16
17
14
p30
20
21
20
19
18
18
19
18
17
18
17
15
p40
20
21
20
19
18
19
19
19
18
18
17
15
p50
22
22
21
21
20
20
21
20
19
19
18
15
p60
24
24
22
22
22
20
26
20
19
23
20
16
phia CMSA, years 1995-2006.
p70
28
29
28
29
27
26
26
24
24
23
20
16
p80
31
33
32
33
30
26
28
28
24
25
22
18
p90
32
34
32
34
32
28
30
29
25
26
26
18
Max
32
34
32
34
32
28
30
29
25
26
26
18
Table B-18. Temporal distribution of hourly NO2 ambient concentrations, Philadelphia CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
65415
67989
68291
66847
64813
51145
59227
66779
49256
58509
56459
32357
Mean
24
25
24
24
22
21
23
21
20
20
19
16
SD
14
14
14
14
13
13
13
12
12
12
12
11
COV
60
55
57
58
59
60
59
59
62
59
62
69
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
8
8
8
7
6
6
6
6
5
6
6
4
p20
10
11
11
11
10
10
10
10
8
9
9
6
p30
14
17
15
15
14
13
14
13
11
12
11
8
p40
19
20
19
19
17
16
17
16
15
15
14
10
p50
20
24
22
23
21
19
21
20
18
18
17
13
p60
26
30
26
27
25
23
25
23
22
22
21
16
p70
30
30
30
31
29
27
29
27
26
26
25
20
p80
35
40
35
36
33
32
34
32
30
30
29
25
p90
40
42
42
42
40
39
40
38
36
36
36
31
Max
140
100
247
97
109
97
96
268
105
101
120
95
B-21
-------
Annual Mean
loc_type=CMSA loc_name=Washington-Baltimore, DC-MD-VA-WV CMSA
Annual Mean
10-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-19. Temporal distribution of annual average NO2 ambient concentrations, Washington DC CMSA, years 1995-
2006.
Hourly Concentrations
loc_type=CMSA loc_name=Washington-Baltimore, DC-MD-VA-WV CMSA
Hourly Cones
SO-
WS !<)% IW7 IW8 IW> 2000 2001 2002 2003 2004 2005 2005
Year
Figure B-20. Temporal distribution of hourly NO2 ambient concentrations, Washington DC CMSA, years 1995-2006.
B-22
-------
Table B-19. Temporal distribution of annual average NO2 ambient concentrations, Washington DC CMSA, years 1995-
2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
12
11
11
11
12
12
11
10
11
12
12
10
Mean
21
22
20
22
20
18
19
19
20
18
17
15
SD
5
4
5
5
5
5
5
6
6
5
5
4
cov
25
20
27
23
25
27
28
31
28
27
28
30
Min
11
11
10
12
11
9
9
9
10
10
9
7
p10
11
20
11
15
12
10
11
10
12
10
10
7
p20
19
20
17
18
14
13
14
13
16
15
14
10
p30
19
21
19
20
18
17
19
16
18
15
15
14
p40
22
22
21
22
20
18
20
20
18
17
17
15
p50
23
22
22
23
21
20
22
23
23
19
18
15
p60
23
24
22
24
23
21
23
23
23
21
21
16
p70
25
24
24
25
24
23
23
24
23
21
21
17
p80
25
25
25
26
24
23
23
25
25
22
21
18
p90
26
26
26
26
25
23
24
25
26
23
22
19
Max
26
27
26
27
25
23
24
25
26
24
24
20
Table B-20. Temporal distribution of hourly NO2 ambient concentrations, Washington DC CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
98349
91551
87646
89335
100112
101494
91594
83969
93111
99370
96396
83691
Mean
21
22
20
22
20
18
19
19
20
18
17
15
SD
13
12
12
12
12
12
12
12
12
11
12
11
COV
59
57
62
57
61
64
62
64
61
63
68
73
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
7
7
6
8
6
5
6
6
6
5
5
4
p20
10
11
9
11
9
8
9
9
9
8
7
6
p30
13
14
12
14
12
11
11
11
12
10
10
7
p40
16
17
15
16
15
13
14
14
14
13
12
9
p50
19
20
18
20
18
16
17
17
17
16
15
12
p60
23
24
21
23
21
19
20
20
21
19
18
14
p70
27
28
25
27
25
23
24
24
25
23
22
18
p80
31
32
30
32
30
28
29
30
30
28
27
23
p90
38
39
37
38
37
35
36
37
37
34
34
30
Max
145
107
155
285
114
141
89
108
102
115
115
129
B-23
-------
Annual Mean
loc_type=MSA loc_name=Atlanta,GA
Annual Mean
30-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-21. Temporal distribution of annual average NO2 ambient concentrations, Atlanta MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Atlanta,GA
Hourly Cones
SO-
WS 1W> IW I'WS 1 2000 2001 2002 2003 2004 2005 2006
Yctir
Figure B-22. Temporal distribution of hourly NO2 ambient concentrations, Atlanta MSA, years 1995-2006.
B-24
-------
Table B-21. Temporal distribution of annual average NO2 ambient concentrations, Atlanta MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
3
5
4
3
4
5
5
5
4
5
5
5
Mean
13
14
15
13
14
14
14
12
11
11
11
11
SD
6
9
7
10
9
7
8
6
6
6
6
6
cov
46
61
47
80
61
53
56
51
56
51
51
57
Min
7
6
8
6
7
5
4
4
5
4
4
3
p10
7
6
8
6
7
5
4
4
5
4
4
3
p20
7
6
8
6
7
6
6
6
5
5
5
5
p30
7
6
15
6
7
8
8
7
6
6
6
6
p40
15
11
15
8
7
12
12
11
6
10
10
9
p50
15
16
15
8
13
17
17
15
11
15
14
13
p60
15
17
15
8
20
17
17
15
16
15
14
14
p70
19
18
15
24
20
18
17
16
16
15
14
15
p80
19
22
25
24
24
21
20
17
16
16
16
17
p90
19
27
25
24
24
23
23
19
16
17
17
18
Max
19
27
25
24
24
23
23
19
16
17
17
18
Table B-22. Temporal distribution of hourly NO2 ambient concentrations, Atlanta MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
25213
40576
31069
24142
31121
40584
42761
42076
32215
42124
42279
41052
Mean
13
15
15
12
14
14
14
12
11
11
11
11
SD
12
13
13
13
14
14
14
12
11
11
11
11
COV
89
89
86
105
99
97
98
95
101
98
96
98
Min
1
1
1
0
0
1
1
1
0
1
1
1
p10
3
3
3
1
2
1
1
1
1
1
1
2
p20
3
3
5
3
4
3
3
3
2
3
3
3
p30
5
5
7
4
5
5
5
5
3
4
4
4
p40
7
8
9
6
7
7
7
6
5
6
6
5
p50
10
11
12
8
9
10
9
9
7
8
8
7
p60
13
14
15
11
12
13
13
11
9
10
10
9
p70
16
18
18
14
17
17
17
15
13
14
13
13
p80
22
24
23
20
23
23
23
20
17
19
18
18
p90
30
34
33
30
35
33
33
29
26
28
27
27
Max
93
122
181
124
242
110
172
136
91
127
97
73
B-25
-------
Annual Mean
40-
Annual Mean
loc_type=MSA loc_name=Colorado Springs,CO
Figure B-23. Temporal distribution of annual average NO2 ambient concentrations, Colorado Springs MSA, years 1995-
2006.
Hourly Cones
so-
Hourly Concentrations
loc_type=MSA loc_name=Colorado Springs,CO
1007
I
-------
Table B-23. Temporal distribution of annual average NO2 ambient concentrations, Colorado Springs MSA, years 1995-
2006.
Year
1995
1996
1997
1998
1999
2000
n
7
3
4
4
4
4
Mean
16
16
16
16
15
19
SD
7
9
6
6
6
11
cov
42
53
36
37
37
58
Min
7
7
7
7
7
9
p10
7
7
7
7
7
9
p20
8
7
7
7
7
9
p30
12
7
17
17
16
16
p40
12
18
17
17
16
16
p50
18
18
18
17
17
16
p60
21
18
19
18
18
17
p70
21
24
19
18
18
17
p80
22
24
20
20
19
35
p90
23
24
20
20
19
35
Max
23
24
20
20
19
35
Table B-24. Temporal distribution of hourly NO2 ambient concentrations, Colorado Springs MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
n
58569
25387
33469
34509
34472
33956
Mean
16
16
16
16
15
19
SD
14
16
13
12
12
20
COV
91
101
80
76
82
106
Min
0
0
0
0
0
0
p10
2
2
3
3
3
3
p20
4
4
5
5
4
6
p30
6
6
6
7
6
8
p40
8
8
9
9
9
11
p50
11
11
12
12
12
15
p60
16
16
16
16
16
20
p70
22
21
21
22
21
24
p80
29
28
27
27
26
28
p90
36
35
35
34
32
34
Max
148
246
118
85
230
308
B-27
-------
Annual Mean
loc_type=MSA loc_name=El Paso,TX
Annual Mean
40-
T
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-25. Temporal distribution of annual average NO2 ambient concentrations, El Paso MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=EI Paso,TX
Hourly Cones
60-
W5 19% IW7 IW8 I'W) 2000 2001 2002 2003 2004 2005 2006
YctIF
Figure B-26. Temporal distribution of hourly NO2 ambient concentrations, El Paso MSA, years 1995-2006.
B-28
-------
Table B-25. Temporal distribution of annual average NO2 ambient concentrations, El Paso MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
1
1
3
2
3
4
5
5
5
5
5
5
Mean
23
35
26
25
22
18
16
17
16
14
14
14
SD
7
8
6
5
4
4
3
4
3
4
cov
0
0
27
33
25
26
26
23
21
25
21
26
Min
23
35
21
19
17
14
10
11
11
9
10
8
p10
23
35
21
19
17
14
10
11
11
9
10
8
p20
23
35
21
19
17
14
12
13
13
11
11
11
p30
23
35
21
19
17
16
14
16
15
13
13
13
p40
23
35
23
19
23
16
16
16
16
13
14
14
p50
23
35
23
25
23
16
17
16
16
13
15
15
p60
23
35
23
31
23
16
17
17
17
15
15
16
p70
23
35
34
31
28
16
18
18
18
17
16
16
p80
23
35
34
31
28
24
20
20
19
18
17
17
p90
23
35
34
31
28
24
22
21
20
18
17
18
Max
23
35
34
31
28
24
22
21
20
18
17
18
Table B-26. Temporal distribution of hourly NO2 ambient concentrations, El Paso MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
6960
6627
22888
15523
23447
30772
38020
41466
39968
41952
41496
37203
Mean
23
35
26
25
22
17
16
17
16
14
14
14
SD
13
15
15
15
13
13
12
13
13
12
12
12
COV
58
43
58
61
60
72
77
77
80
83
86
84
Min
3
2
0
0
0
0
0
0
0
0
0
0
p10
9
20
10
7
6
3
3
4
3
2
2
2
p20
12
23
13
12
10
5
5
5
5
4
4
4
p30
14
27
16
15
14
8
7
7
7
6
5
6
p40
17
29
20
19
17
12
10
10
9
8
7
8
p50
21
32
23
23
21
16
13
13
12
11
10
10
p60
25
36
28
27
25
20
16
17
16
14
14
14
p70
29
40
32
32
28
24
21
22
21
19
19
19
p80
34
46
38
37
33
28
27
28
27
25
24
25
p90
41
54
45
45
40
34
34
35
35
32
31
32
Max
113
219
174
166
108
125
102
153
106
97
87
99
B-29
-------
Annual Mean
loc_type—MSA loc_name—Jacksonville,FL
Annual Mean
16-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Year
Figure B-27. Temporal distribution of annual average NO2 ambient concentrations, Jacksonville MSA, years 1995-2006.
Hourly Concentrations
loc_type-MSAloc_name-Jacksonville,FL
Hourly Cones
40-
I995 1996 1997 1998 1999 2000 200] 2002 2003 2004 2005
Year
Figure B-28. Temporal distribution of hourly NO2 ambient concentrations, Jacksonville MSA, years 1995-2006.
Table B-27. Temporal distribution of annual average NO2 ambient concentrations, Jacksonville MSA, years 1995-2006.
I Year I n I Mean I SD I COV I Min I p10 I p20 I p30 I p40 I p50 I p60 I p70 I p80 I p90 I Max
B-30
-------
1995
1996
1997
1998
1999
2000
2002
2003
2004
2005
1
1
1
1
1
1
1
1
1
1
16
15
14
15
16
15
15
14
14
13
0
0
0
0
0
0
0
0
0
0
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
16
15
14
15
16
15
15
14
14
13
Table B-28. Temporal distribution of hourly NO2 ambient concentrations, Jacksonville MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2002
2003
2004
2005
n
7755
8148
8326
8211
7795
7661
7944
7041
7451
7890
Mean
16
15
14
15
16
15
15
14
14
13
SD
10
10
9
10
10
10
10
10
11
9
cov
60
64
65
65
61
67
66
71
83
67
Min
0
0
0
0
0
0
0
0
0
0
p10
6
5
5
5
5
5
5
4
4
4
p20
8
7
6
7
7
7
7
6
6
6
p30
9
9
8
9
9
9
9
8
7
8
p40
11
11
10
11
12
11
11
10
9
9
p50
14
13
12
13
14
13
13
12
11
11
p60
16
15
15
15
16
15
15
14
13
13
p70
19
18
17
18
20
18
17
17
16
16
p80
23
21
21
22
24
23
21
21
20
20
p90
29
28
27
28
30
30
27
28
26
26
Max
76
80
92
66
63
72
294
76
201
64
B-31
-------
Annual Mean
loc_type=MSA loc_name=Las Vegas,NV-AZ
Annual Mean
10-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Year
Figure B-29. Temporal distribution of annual average NO2 ambient concentrations, Las Vegas MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Las Vegas,NV-AZ
Hourly Cones
70-
50-
I005 10% 1907 1008 1000 2000 2001 2002 200.1 2004 2005
Veur
Figure B-30. Temporal distribution of hourly NO2 ambient concentrations, Las Vegas MSA, years 1995-2006.
B-32
-------
Table B-29. Temporal distribution of annual average NO2 ambient concentrations, Las Vegas MSA, years 1995-2006.
Year
1995
1996
1998
1999
2000
2001
2002
2003
2004
2005
n
1
1
3
5
6
6
9
7
7
6
Mean
27
27
12
14
12
11
11
12
11
10
SD
12
10
9
9
8
8
8
8
cov
0
0
95
71
81
84
68
66
73
76
Min
27
27
4
4
3
2
3
2
1
2
p10
27
27
4
4
3
2
3
2
1
2
p20
27
27
4
6
4
5
3
6
4
5
p30
27
27
4
8
4
5
7
8
5
5
p40
27
27
7
8
8
6
7
8
5
6
p50
27
27
7
8
8
6
9
9
9
8
p60
27
27
7
16
8
7
10
19
19
9
p70
27
27
25
24
22
22
19
19
19
19
p80
27
27
25
25
22
22
22
21
19
19
p90
27
27
25
27
25
23
22
21
20
20
Max
27
27
25
27
25
23
22
21
20
20
Table B-30. Temporal distribution of hourly NO2 ambient concentrations, Las Vegas MSA, years 1995-2006.
Year
1995
1996
1998
1999
2000
2001
2002
2003
2004
2005
n
7951
8723
25234
43110
46403
49734
74814
58398
57484
48911
Mean
27
27
12
14
12
11
11
12
11
10
SD
20
22
14
16
14
14
13
14
13
12
COV
74
81
118
110
119
128
117
119
120
123
Min
0
0
0
0
0
0
0
0
0
0
p10
0
0
0
0
0
0
0
0
0
0
p20
11
9
0
0
0
0
0
0
0
0
p30
15
12
0
5
0
0
0
0
0
0
p40
20
17
5
6
5
0
5
5
0
0
p50
25
24
8
8
7
6
7
7
6
6
p60
31
31
10
12
10
8
10
10
9
9
p70
37
38
14
18
15
13
14
15
14
12
p80
42
44
23
28
23
21
21
24
23
18
p90
50
54
35
39
34
33
32
35
33
30
Max
410
149
103
110
100
104
87
103
73
75
B-33
-------
Annual Mean
loc_type=MSA loc_name=Phoemx-Mesa,AZ
Annual Mean
1 00? 1 008 1 000 2000 200 ! 2002 2003 2004 2005 2006
Figure B-31. Temporal distribution of annual average NO2 ambient concentrations, Phoenix MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Phoenix-Mesa,AZ
Hourly Cones
70-
0-
l«05 1006 iOi)7 |
-------
Table B-31. Temporal distribution of annual average NO2 ambient concentrations, Phoenix MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
3
3
2
4
5
5
5
3
2
5
6
6
Mean
29
28
30
29
33
30
27
29
32
25
23
22
SD
3
3
3
5
5
4
6
6
4
4
7
7
cov
12
12
10
15
14
13
23
19
11
18
29
30
Min
26
25
28
24
28
26
22
24
29
19
12
11
p10
26
25
28
24
28
26
22
24
29
19
12
11
p20
26
25
28
24
30
27
22
24
29
21
20
19
p30
26
25
28
28
31
29
22
24
29
23
20
19
p40
29
29
28
28
31
29
24
29
29
23
24
21
p50
29
29
30
29
31
29
26
29
32
24
24
22
p60
29
29
32
30
32
30
27
29
34
24
24
24
p70
33
32
32
30
34
30
29
35
34
25
26
25
p80
33
32
32
35
37
33
33
35
34
28
26
25
p90
33
32
32
35
40
36
37
35
34
31
32
31
Max
33
32
32
35
40
36
37
35
34
31
32
31
Table B-32. Tern
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
23196
23598
14629
32078
40996
41686
40463
25028
14195
42176
50583
48791
poral distribution of hourly NO2 ambient concentrations, Phoenix MSA, years 1 995-2006.
Mean
29
28
30
29
33
30
27
29
32
25
23
22
SD
17
17
16
17
22
21
16
17
17
15
15
16
COV
59
59
55
58
66
71
59
59
55
62
66
73
Min
0
0
0
0
0
0
1
0
0
0
0
0
p10
8
8
8
8
9
8
7
7
8
6
5
4
p20
12
12
13
12
13
12
11
12
14
9
8
7
p30
17
17
18
17
18
17
15
17
20
13
12
10
p40
23
22
25
23
24
22
21
23
27
18
16
13
p50
28
27
30
28
30
27
26
28
32
23
20
18
p60
33
32
35
33
36
32
31
34
37
28
25
24
p70
37
37
39
38
42
38
36
39
42
33
31
30
p80
44
43
44
44
49
45
41
45
48
39
36
37
p90
53
51
52
52
60
54
49
53
55
45
44
46
Max
128
115
114
116
198
267
118
108
101
104
131
111
B-35
-------
Annual Mean
loc_type=MSA loc_name=Provo-Orem,UT
Annual Mean
20-
1995 1996 1997 1998 1999 2000
2002 2003
2005 2006
Figure B-33. Temporal distribution of annual average NO2 ambient concentrations, Provo MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Provo-Orem,UT
Hourly Cones
130-
1 <»8 i '.«<)
200.1 2004 2005 200(>
2000 2001 2002
Year
Figure B-34. Temporal distribution of hourly NO2 ambient concentrations, Provo MSA, years 1995-2006.
B-36
-------
Table B-33. Temporal distribution of annual average NO2 ambient concentrations, Provo MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
1
1
1
1
1
1
1
1
1
1
1
1
Mean
23
24
23
24
24
24
24
25
22
22
21
29
SD
COV
0
0
0
0
0
0
0
0
0
0
0
0
Min
23
24
23
24
24
24
24
25
22
22
21
29
p10
23
24
23
24
24
24
24
25
22
22
21
29
p20
23
24
23
24
24
24
24
25
22
22
21
29
p30
23
24
23
24
24
24
24
25
22
22
21
29
p40
23
24
23
24
24
24
24
25
22
22
21
29
p50
23
24
23
24
24
24
24
25
22
22
21
29
p60
23
24
23
24
24
24
24
25
22
22
21
29
p70
23
24
23
24
24
24
24
25
22
22
21
29
p80
23
24
23
24
24
24
24
25
22
22
21
29
p90
23
24
23
24
24
24
24
25
22
22
21
29
Max
23
24
23
24
24
24
24
25
22
22
21
29
Table B-34. Temporal distribution of hourly NO2 ambient concentrations, Provo MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
8002
8430
7034
8210
8563
8406
8501
8200
7730
8302
8502
6993
Mean
23
24
23
24
24
24
24
25
22
22
21
29
SD
13
15
13
13
13
13
14
14
13
15
13
34
COV
55
61
57
56
55
56
57
57
59
66
62
118
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
7
7
7
7
7
7
6
6
6
5
5
5
p20
10
10
10
10
11
10
10
10
8
8
8
7
p30
13
14
14
14
14
14
14
15
12
12
11
10
p40
17
18
18
18
19
18
19
20
16
16
15
13
p50
22
23
22
23
23
22
23
25
21
20
19
17
p60
26
28
26
28
28
27
28
30
26
25
23
22
p70
30
32
31
32
33
32
33
34
30
30
28
30
p80
34
37
35
37
37
37
38
38
34
35
33
38
p90
40
43
41
42
42
42
43
43
39
42
39
61
Max
67
97
81
78
77
74
72
80
72
90
64
164
B-37
-------
Annual Mean
10-
Annual Mean
loc_type=MSA loc_name=St, Loms,MO-IL
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-35. Temporal distribution of annual average NO2 ambient concentrations, St. Louis MSA, years 1995-2006.
Hourly Cones
so-
Hourly Concentrations
loc_type=MSA loc_name=St, Louis,MO-IL
I«.H)7
2000 2001 2002 2003 2004 2005 2005
Figure B-36. Temporal distribution of hourly NO2 ambient concentrations, St. Louis MSA, years 1995-2006.
B-38
-------
Table B-35. Temporal distribution of annual average NO2 ambient concentrations, St. Louis MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
10
10
10
8
9
9
8
9
9
9
6
2
Mean
18
17
17
19
19
18
17
16
15
14
13
12
SD
6
6
6
5
5
5
5
4
4
4
3
5
cov
35
33
32
25
24
29
28
26
26
31
24
40
Min
5
6
6
11
12
9
10
10
9
8
9
8
p10
8
8
8
11
12
9
10
10
9
8
9
8
p20
12
12
12
13
14
12
12
11
10
10
10
8
p30
15
16
16
18
18
16
17
14
14
12
10
8
p40
19
19
19
19
18
17
17
15
14
13
12
8
p50
19
19
19
19
20
18
18
16
16
13
13
12
p60
20
20
19
19
21
19
19
17
16
16
15
15
p70
22
20
19
20
21
21
20
19
18
17
15
15
p80
22
21
21
22
24
21
20
21
19
18
15
15
p90
24
23
23
26
27
26
25
23
20
22
17
15
Max
26
25
25
26
27
26
25
23
20
22
17
15
Table B-36. Tern
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
85072
86085
86314
68308
77611
77327
67871
76693
77543
75493
49948
16688
poral distribution of hourly NO2 ambient concentrations, St. Louis MSA, years 1995-2006.
Mean
18
17
17
19
19
18
17
16
15
14
13
12
SD
12
11
11
11
12
11
11
11
10
10
9
8
COV
68
65
67
58
61
64
64
65
67
69
70
70
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
4
4
4
6
6
5
5
5
4
4
4
3
p20
7
7
7
9
9
8
7
7
6
6
5
5
p30
10
10
10
12
12
10
10
9
8
8
7
6
p40
13
13
12
14
14
13
13
12
11
10
9
8
p50
16
16
15
17
17
16
15
14
13
12
11
10
p60
19
19
18
20
20
19
19
17
16
15
13
12
p70
23
22
22
23
24
22
22
21
19
18
16
15
p80
28
26
26
28
29
27
27
25
23
22
20
18
p90
34
32
33
33
36
34
33
31
29
28
26
23
Max
103
84
274
97
99
85
95
124
123
130
70
53
B-39
-------
Annual Mean
loc_type=MSA/CMSA loc_name=Other MSA/CMSA
Annual Mean
40-
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Year
Figure B-37. Temporal distribution of annual average NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA/CMSA loc_name=Other MSA/CMSA
Hourly Cones
40-
IW7 I'WS l'W> 2000 2001 2002 2003 2004 2005 2006
Figure B-38. Temporal distribution of hourly NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006.
B-40
-------
Table B-37. Temporal distribution of annual average NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
186
186
187
185
192
199
201
209
202
211
207
147
Mean
15
14
14
14
15
14
13
12
12
11
11
10
SD
6
6
6
6
6
6
6
6
5
5
5
4
cov
44
43
43
43
42
41
43
45
42
44
43
41
Min
1
1
2
1
1
1
1
1
1
1
1
1
p10
5
5
5
5
6
5
5
5
5
5
5
4
p20
8
9
9
10
9
8
7
7
7
7
7
6
p30
11
11
11
11
11
11
10
9
9
9
9
9
p40
13
13
12
13
14
12
12
11
11
10
10
9
p50
15
15
14
14
15
14
13
13
12
11
11
11
p60
17
16
16
16
16
16
15
14
14
13
12
12
p70
18
18
18
18
18
17
17
16
15
14
14
13
p80
21
20
19
20
20
18
18
17
17
16
16
14
p90
22
22
22
22
23
21
20
20
18
17
17
16
Max
32
30
29
31
29
26
27
27
26
25
24
18
Table B-38. Temporal distribution of hourly NO2 ambient concentrations, Other MSA/CMSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
186
1520743
1520290
1503051
1560074
1630060
1648640
1713558
1661992
1738133
1706730
1168444
Mean
15
14
14
14
15
14
13
13
12
11
11
10
SD
6
12
11
11
12
11
11
11
10
10
10
9
COV
44
81
82
80
83
81
84
85
84
87
87
87
Min
1
0
0
0
0
0
0
0
0
0
0
0
p10
5
2
2
2
3
2
2
2
2
2
2
2
p20
8
5
4
5
5
4
4
4
4
3
3
3
p30
11
7
6
7
7
6
6
5
5
5
5
5
p40
13
9
9
9
9
8
8
7
7
7
6
6
p50
15
12
11
11
11
11
10
9
9
8
8
8
p60
17
15
14
15
14
13
13
12
12
11
11
10
p70
18
18
18
18
18
17
16
15
15
14
14
13
p80
21
23
23
23
24
22
21
20
19
18
18
17
p90
22
31
30
31
32
29
29
28
26
25
25
23
Max
32
336
313
300
172
289
193
158
148
160
153
240
B-41
-------
Annual Mean
20-
1')-
18-
17-
16-
15-
1-1-
II-
10-
')-
X-
7-
6-
5-
4-
Annual Mean
loc_type=Not MSA loc_name=Other Not MSA
IW) 2000 2001 2002 2003 200-1
2006
Figure B-39. Temporal distribution of annual average NO2 ambient concentrations, Other Not MSA, years 1995-2006.
Hourly Cones
30-
Hourly Concentrations
loc_type=Not MSA loc_name=Other Not MSA
I««7 |i|t)g |9<)Q 2000 2001 2002 2003 2004 2005
2006
Figure B-40. Temporal distribution of hourly NO2 ambient concentrations, Other Not MSA, years 1995-2006.
B-42
-------
Table B-39. Temporal distribution of annual average NO2 ambient concentrations, Other Not MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
28
29
35
33
36
39
41
42
44
47
43
26
Mean
8
7
7
7
8
8
7
7
7
6
7
6
SD
5
5
5
5
5
4
4
4
4
4
4
5
cov
59
71
67
62
67
57
60
65
61
64
63
71
Min
1
0
0
1
0
2
1
1
1
2
1
1
p10
2
0
1
1
1
2
2
2
2
2
2
1
p20
4
2
3
3
3
3
3
2
3
2
2
2
p30
5
4
4
4
4
5
4
3
3
3
3
2
p40
7
5
5
5
5
6
5
4
4
4
5
3
p50
7
5
7
7
7
8
6
6
6
5
6
5
p60
8
7
9
7
8
8
8
8
8
7
8
8
p70
10
10
10
10
9
10
9
8
9
8
9
10
p80
13
13
12
12
12
11
10
10
11
11
11
11
p90
15
14
14
14
16
14
13
13
13
13
12
12
Max
19
14
20
19
20
19
17
16
15
16
17
16
Table B-40. Temporal distribution of hourly NO2 ambient concentrations, Other Not MSA, years 1995-2006.
Year
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
n
225810
234628
278906
264015
290382
316568
328407
340873
351652
375716
353229
207114
Mean
8
7
7
8
8
8
7
7
7
6
7
6
SD
9
8
8
8
9
8
7
7
7
7
8
7
COV
104
118
113
105
113
104
109
112
110
115
114
119
Min
0
0
0
0
0
0
0
0
0
0
0
0
p10
0
0
0
1
0
1
1
1
1
1
1
0
p20
2
1
1
2
2
2
1
1
2
1
1
1
p30
3
2
2
3
2
3
2
2
2
2
2
2
p40
4
3
3
4
3
4
3
3
3
3
3
2
p50
6
4
5
5
5
5
4
4
4
4
4
3
p60
7
6
6
7
6
7
6
5
5
5
6
5
p70
10
8
9
9
9
9
8
8
7
7
8
7
p80
13
11
12
12
12
12
11
11
10
10
11
10
p90
19
17
18
18
18
18
16
17
16
16
17
16
Max
217
164
207
181
286
192
139
267
201
285
262
101
B-43
-------
Appendix C. Spatial NO2 Air Quality Characterization
Appendix C contains the ambient air quality analysis results by monitoring site within each of the named
locations. Boxplots were constructed to display the annual average and hourly concentration distributions across
sites for a single location. The box extends from the 25th to the 75th percentile, with the median shown as the line
inside the box. The whiskers extend from the box to the 5th and 95th percentiles. The extreme values in the upper
and lower tails beyond the 5th and 95th percentiles are not shown to allow for similar scaling along the y-axis for the
plotted independent variables. The mean is plotted as a dot; typically it would appear inside the box, however it will
fall outside the box if the distribution is highly skewed. All concentrations are shown in parts per billion (ppb). The
boxplots for hourly concentrations were created using a different procedure than for the annual statistics, because of
the large number of hourly values and the inability of the graphing procedure to allow frequency weights.
Therefore, the appropriate weighted percentiles and means were calculated and plotted as shown, but the vertical
lines composing the sides of the box are essentially omitted. Tables are provided that summarize the complete
distribution, with percentiles apportioned in segments of 10.
C-1
-------
List of Figures
Figure C-1. Spatial distribution of annual average NO2 concentration, Boston CMSAset a, years 1995-2006 4
Figure C-2. Spatial distribution of hourly NO2 concentration, Boston CMSA set a, years 1995-2006 4
Figure C-3. Spatial distribution of annual average NO2 concentration, Boston CMSAset b, years 1995-2006 5
Figure C-4. Spatial distribution of hourly NO2 concentration, Boston CMSA set b, years 1995-2006 5
Figure C-5. Spatial distribution of annual average NO2 concentration, Chicago CMSA, years 1995-2006 7
Figure C-6. Spatial distribution of hourly NO2 concentration, Chicago CMSA, years 1995-2006 7
Figure C-7. Spatial distribution of annual average NO2 concentration, Cleveland CMSA, years 1995-2006 9
Figure C-8. Spatial distribution of hourly NO2 concentration, Cleveland CMSA, years 1995-2006 9
Figure C-9. Spatial distribution of annual average NO2 concentration, Denver CMSA, years 1995-2006 11
Figure C-10. Spatial distribution of hourly NO2 concentration, Denver CMSA, years 1995-2006 11
Figure C-11. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006 13
Figure C-12. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006 13
Figure C-13. Spatial distribution of annual average NO2 concentration, Los Angeles CMSAset a, years 1995-2006 15
Figure C-14. Spatial distribution of hourly NO2 concentration, Los Angeles CMSAset a, years 1995-2006 15
Figure C-15. Spatial distribution of annual average NO2 concentration, Los Angeles CMSAset b, years 1995-2006 16
Figure C-16. Spatial distribution of hourly NO2 concentration, Los Angeles CMSAset b, years 1995-2006 16
Figure C-17. Spatial distribution of annual average NO2 concentration, Los Angeles CMSAset c, years 1995-2006 17
Figure C-18. Spatial distribution of hourly NO2 concentration, Los Angeles CMSA set c, years 1995-2006 17
Figure C-19. Spatial distribution of annual average NO2 concentration, Miami CMSA, years 1995-2006 20
Figure C-20. Spatial distribution of hourly NO2 concentration, Miami CMSA, years 1995-2006 20
Figure C-21. Spatial distribution of annual average NO2 concentration, New York CMSA set a, years 1995-2006 22
Figure C-22. Spatial distribution of hourly NO2 concentration, New York CMSA set a, years 1995-2006 22
Figure C-23. Spatial distribution of annual average NO2 concentration, New York CMSAset b, years 1995-2006 23
Figure C-24. Spatial distribution of hourly NO2 concentration, New York CMSA set b, years 1995-2006 23
Figure C-25. Spatial distribution of annual average NO2 concentration, Philadelphia CMSA, years 1995-2006 25
Figure C-26. Spatial distribution of hourly NO2 concentration, Philadelphia CMSA, years 1995-2006 25
Figure C-27. Spatial distribution of annual average NO2 concentration, Washington DC CMSAset a, years 1995-2006..27
Figure C-28. Spatial distribution of hourly NO2 concentration, Washington DC CMSA set a, years 1995-2006 27
Figure C-29. Spatial distribution of annual average NO2 concentration, Washington DC CMSAset b, years 1995-2006..28
Figure C-30. Spatial distribution of hourly NO2 concentration, Washington DC CMSA set b, years 1995-2006 28
Figure C-31. Spatial distribution of annual average NO2 concentration, Atlanta MSA, years 1995-2006 30
Figure C-32. Spatial distribution of hourly NO2 concentration, Atlanta MSA, years 1995-2006 30
Figure C-33. Spatial distribution of annual average NO2 concentration, Colorado Springs MSA, years 1995-2006 32
Figure C-34. Spatial distribution of hourly NO2 concentration, Colorado Springs MSA, years 1995-2006 32
Figure C-35. Spatial distribution of annual average NO2 concentration, El Paso MSA, years 1995-2006 34
Figure C-36. Spatial distribution of hourly NO2 concentration, El Paso MSA, years 1995-2006 34
Figure C-37. Spatial distribution of annual average NO2 concentration, Jacksonville MSA, years 1995-2006 36
Figure C-38. Spatial distribution of hourly NO2 concentration, Jacksonville MSA, years 1995-2006 36
Figure C-39. Spatial distribution of annual average NO2 concentration, Las Vegas MSA, years 1995-2006 38
Figure C-40. Spatial distribution of hourly NO2 concentration, Las Vegas MSA, years 1995-2006 38
Figure C-41. Spatial distribution of annual average NO2 concentration, Phoenix MSA, years 1995-2006 40
Figure C-42. Spatial distribution of hourly NO2 concentration, Phoenix MSA, years 1995-2006 40
Figure C-43. Spatial distribution of annual average NO2 concentration, Provo MSA, years 1995-2006 42
Figure C-44. Spatial distribution of hourly NO2 concentration, Provo MSA, years 1995-2006 42
Figure C-45. Spatial distribution of annual average NO2 concentration, St. Louis MSA, years 1995-2006 44
Figure C-46. Spatial distribution of hourly NO2 concentration, St. Louis MSA, years 1995-2006 44
C-2
-------
List of Tables
Table C-1. Spatial distribution of annual average NO2 concentration, Boston CMSA, years 1995-2006 6
Table C-2. Spatial distribution of hourly NO2 concentration, Boston CMSA, years 1995-2006 6
Table C-3. Spatial distribution of annual average NO2 concentration, Chicago CMSA, years 1995-2006 8
Table C-4. Spatial distribution of hourly NO2 concentration, Chicago CMSA, years 1995-2006 8
Table C-5. Spatial distribution of annual average NO2 concentration, Cleveland CMSA, years 1995-2006 10
Table C-6. Spatial distribution of hourly NO2 concentration, Cleveland CMSA, years 1995-2006 10
Table C-7. Spatial distribution of annual average NO2 concentration, Denver CMSA, years 1995-2006 12
Table C-8. Spatial distribution of hourly NO2 concentration, Denver CMSA, years 1995-2006 12
Table C-9. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006 14
Table C-10. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006 14
Table C-11. Spatial distribution of annual average NO2 concentration, Los Angeles CMSA, years 1995-2006 18
Table C-12. Spatial distribution of hourly NO2 concentration, Los Angeles CMSA, years 1995-2006 19
Table C-13. Spatial distribution of annual average NO2 concentration, Miami CMSA, years 1995-2006 21
Table C-14. Spatial distribution of hourly NO2 concentration, Miami CMSA, years 1995-2006 21
Table C-15. Spatial distribution of annual average NO2 concentration, New York CMSA, years 1995-2006 24
Table C-16. Spatial distribution of hourly NO2 concentration, New York CMSA, years 1995-2006 24
Table C-17. Spatial distribution of annual average NO2 concentration, Philadelphia CMSA, years 1995-2006 26
Table C-18. Spatial distribution of hourly NO2 concentration, Philadelphia CMSA, years 1995-2006 26
Table C-19. Spatial distribution of annual average NO2 concentration, Washington DC CMSA, years 1995-2006 29
Table C-20. Spatial distribution of hourly NO2 concentration, Washington DC CMSA, years 1995-2006 29
Table C-21. Spatial distribution of annual average NO2 concentration, Atlanta MSA, years 1995-2006 31
Table C-22. Spatial distribution of hourly NO2 concentration, Atlanta MSA, years 1995-2006 31
Table C-23. Spatial distribution of annual average NO2 concentration, Colorado Springs MSA, years 1995-2006 33
Table C-24. Spatial distribution of hourly NO2 concentration, Colorado Springs MSA, years 1995-2006 33
Table C-25. Spatial distribution of annual average NO2 concentration, El Paso MSA, years 1995-2006 35
Table C-26. Spatial distribution of hourly NO2 concentration, El Paso MSA, years 1995-2006 35
Table C-27. Spatial distribution of annual average NO2 concentration, Jacksonville MSA, years 1995-2006 37
Table C-28. Spatial distribution of hourly NO2 concentration, Jacksonville MSA, years 1995-2006 37
Table C-29. Spatial distribution of annual average NO2 concentration, Las Vegas MSA, years 1995-2006 39
Table C-30. Spatial distribution of hourly NO2 concentration, Las Vegas MSA, years 1995-2006 39
Table C-31. Spatial distribution of annual average NO2 concentration, Phoenix MSA, years 1995-2006 41
Table C-32. Spatial distribution of hourly NO2 concentration, Phoenix MSA, years 1995-2006 41
Table C-33. Spatial distribution of annual average NO2 concentration, Provo MSA, years 1995-2006 43
Table C-34. Spatial distribution of hourly NO2 concentration, Provo MSA, years 1995-2006 43
Table C-35. Spatial distribution of annual average NO2 concentration, St. Louis MSA, years 1995-2006 45
Table C-36. Spatial distribution of hourly NO2 concentration, St. Louis MSA, years 1995-2006 45
C-3
-------
Annual Mean
loc_type=CMSA loc_name=Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA Set a
Annual Mean
40-
Figure C-1. Spatial distribution of annual average NO2 concentration, Boston CMSA set a, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA Set a
Hourly Cones
50-
1
3
0
3
1
3
0
0
2
1
•>
s
0
0
5
1
0
0
5
1
"> t
5 5
0 0
0 0
9 9
2 4
0 0
0 0
6 4
1 1
•> i "^ *>
5555
0000
0222
9155
5000
0000
0002
5921
1111
"V "> 1 ">
5555
0000
2222
5555
0000
0000
3344
5601
1111
•> *>
5 5
0 0
2 2
5 5
0 1
0 0
4 0
2 3
1 1
^
5
0
2
7
0
0
2
0
1
Monitor
Figure C-2. Spatial distribution of hourly NO2 concentration, Boston CMSA set a, years 1995-2006.
C-4
-------
Annual Mean
loc_type=CMSA loc_name=Boston-Worcester-La»rence, MA-NH-ME-CT CMSA Set b
Annual Mean
20-
2502700231 3M1! 100)61 330M001'>I 3301 100201 330I5000')I 3301500131 3301500141 3301500151
Monuor
Figure C-3. Spatial distribution of annual average NO2 concentration, Boston CMSA set b, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA Set b
Hourly Cones
40-
2502700211 3301100161 3301100191 3301100201 3301500091 3301500131 3301500141 3301500151
Momlor
Figure C-4. Spatial distribution of hourly NO2 concentration, Boston CMSA set b, years 1995-2006.
C-5
-------
Table C-1. Spatial distribution of annual average NO2 concentration, Boston CMSA, years 1995-2006.
Monitor ID
2303130021
2500510051
2500920061
2500940041
2500950051
2502100091
2502500021
2502500211
2502500351
2502500361
2502500401
2502500411
2502500421
2502510031
2502700201
2502700231
3301100161
3301100191
3301100201
3301500091
3301500131
3301500141
3301500151
n
7
2
10
5
2
1
11
8
1
1
11
1
6
5
8
3
4
1
5
5
4
3
1
Mean
10
8
12
6
9
22
28
25
23
22
21
12
21
23
19
15
16
11
12
12
6
8
13
SD
1
1
3
0
1
3
3
2
3
2
1
0
2
1
1
1
0
cov
9
9
22
7
8
11
12
10
16
7
6
3
15
8
12
10
6
Min
9
8
9
6
9
22
23
21
23
22
16
12
17
21
17
15
14
11
10
9
5
7
13
p10
9
8
9
6
9
22
23
21
23
22
18
12
17
21
17
15
14
11
10
9
5
7
13
p20
9
8
10
6
9
22
25
22
23
22
20
12
19
21
18
15
14
11
11
11
5
7
13
p30
9
8
10
6
9
22
25
23
23
22
21
12
19
22
19
15
15
11
11
12
5
7
13
p40
9
8
11
6
9
22
29
27
23
22
21
12
19
22
19
15
15
11
12
12
5
7
13
p50
10
8
11
6
9
22
30
27
23
22
21
12
21
23
19
15
15
11
12
12
6
7
13
p60
10
9
13
6
10
22
30
27
23
22
22
12
22
23
19
15
16
11
12
12
6
7
13
p70
10
9
15
6
10
22
30
27
23
22
22
12
24
23
20
16
16
11
12
12
6
8
13
p80
10
9
15
7
10
22
31
28
23
22
23
12
24
24
20
16
19
11
12
13
7
8
13
p90
11
9
16
7
10
22
31
28
23
22
23
12
25
25
21
16
19
11
13
13
7
8
13
Max
11
9
16
7
10
22
31
28
23
22
23
12
25
25
21
16
19
11
13
13
7
8
13
Table C-2. Spatial distribution of hourly NO2 concentration, Boston CMSA, years 1995-2006.
Monitor ID
2303130021
2500510051
2500920061
2500940041
2500950051
2502100091
2502500021
2502500211
2502500351
2502500361
2502500401
2502500411
2502500421
2502510031
2502700201
2502700231
3301100161
3301100191
3301100201
3301500091
3301500131
3301500141
3301500151
n
58123
16732
80761
41337
16228
8546
87534
63990
8539
8542
91196
8319
48078
40775
63836
24267
33436
8022
41325
40978
33536
25372
8599
Mean
10
8
12
6
9
22
28
25
23
22
21
12
21
23
19
15
16
11
12
12
6
8
13
SD
9
7
10
7
8
10
11
11
10
11
12
10
10
12
11
9
10
9
9
9
7
7
9
COV
94
81
80
108
91
46
40
45
47
49
59
89
48
54
59
58
64
81
75
77
118
94
75
Min
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
p10
1
2
3
0
2
9
14
13
10
9
7
2
9
9
6
5
6
2
3
2
0
1
3
p20
2
3
4
1
3
13
18
16
13
12
10
3
12
12
9
8
8
3
4
4
1
2
5
p30
4
4
6
2
4
15
21
18
16
15
13
5
15
14
11
10
9
5
6
6
2
3
6
p40
5
5
7
3
5
18
24
21
19
19
16
6
17
17
14
12
11
7
7
8
2
4
8
p50
7
6
9
4
6
21
27
24
21
21
19
8
20
20
17
14
13
9
9
10
3
5
10
p60
9
8
12
6
8
23
30
26
24
24
22
11
22
24
21
16
16
11
11
12
5
7
12
p70
12
10
15
7
11
27
33
30
27
28
26
15
25
28
24
19
19
14
14
15
7
9
16
p80
16
13
20
10
14
30
37
34
31
31
31
19
29
33
29
22
23
18
18
19
10
12
20
p90
23
18
27
16
22
35
43
40
37
36
38
27
35
40
35
27
29
24
25
25
15
17
27
Max
100
50
90
70
51
75
134
205
74
100
113
81
79
94
95
93
158
54
62
63
50
48
65
C-6
-------
Annual Mean
loc_type=CMSA loc_name=Chicago-Gary-Kenosha, IL-IN-W1 CMSA
Annual Mean
20
10
0
_^ -^
1
^ 5 b
»
"^^^ ^^
=
1 1 1 1 1 1 1 1 1 1 1 1 1 1
77777777777788
00000000000100
33333333333988
11111111111799
00000334448101
000001 10220000
36677000000121
734561321 13126
1111111
121112
Figure C-5. Spatial distribution of annual average NO2 concentration, Chicago CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Chicago-Gary-Kenosha, 1L-IN-WI CMSA
Hourly Cones
so
40
10
20
10
<
(
i
i —
<
, <
i
1
_*
1
i
<
i
1
)
1
> 1
1
I
I — 1
1
1
) —
>
77777777777788
00000000000
0 0
33333333333988
111111111799
00000334448
0 1
000001 10220000
36677000000
73456 32113
111111121
2 1
2 6
1 2
Figure C-6. Spatial distribution of hourly NO2 concentration, Chicago CMSA, years 1995-2006.
C-7
-------
Table C-3. Spatial distribution
Monitor ID n Mean
1703100371 1 29
1703100631 12 31
1703100641 6 23
1703100751 4 24
1703100761 5 20
1703131011 3 31
1703131031 9 29
1703140021 12 26
1703142011 4 17
1703142012 4 17
1703180031 8 23
1719710111 5 9
1808900221 8 19
1808910162 2 22
Table C-4. Spatial distribution
of annual average NO2 concentration, Chicago CMSA, years 1995-2006.
SD COV
1 4
2 8
0 2
2 9
1 3
1 5
2 8
1 4
1 4
1 4
1 6
1 4
2 7
of hourly NO2
Min
29
29
21
23
18
30
28
22
17
16
22
9
18
21
p10
29
30
21
23
18
30
28
23
17
16
22
9
18
21
concentration,
Monitor ID n Mean SD COV Min
1703100371 8630 29
1703100631 101935 31
1703100641 52139 23
1703100751 34028 24
1703100761 42946 20
1703131011 25141 31
1703131031 75061 29
1703140021 102779 26
1703142011 32625 17
1703142012 32552 17
1703180031 68952 23
1719710111 41227 9
1808900221 63295 19
1808910162 16574 22
13 44
15 46
13 57
12 52
12 59
13 41
13 44
13 51
11 64
10 62
12 53
6 69
12 66
12 56
0
0
0
0
0
3
0
0
0
0
0
0
0
3
p10
15
13
8
10
7
16
14
11
5
6
9
3
4
9
p20
29
31
22
23
19
30
28
24
17
16
22
9
18
21
p30 p40
29 29
31 31
22 23
24 24
20 20
30 31
28 29
24 26
17 17
16 16
23 23
9 9
18 19
21 21
p50
29
32
23
24
20
31
29
27
18
17
23
9
19
22
p60
29
32
24
24
21
31
30
27
18
17
24
9
19
23
p70
29
32
24
24
22
32
30
27
18
17
24
9
19
23
p80
29
32
24
25
22
32
31
27
18
17
24
10
20
23
p90
29
32
26
25
22
32
31
28
18
17
25
10
20
23
Max
29
34
26
25
22
32
31
28
18
17
25
10
20
23
Chicago CMSA, years 1995-2006.
p20
19
19
11
13
10
20
18
14
7
8
12
4
7
12
p30 p40
22 25
23 27
15 18
16 19
12 15
23 27
22 25
17 20
10 12
10 12
15 18
5 6
10 13
14 16
p50
28
30
21
22
18
30
28
24
15
14
21
8
17
19
p60
31
34
25
26
21
33
31
27
19
17
25
9
20
22
p70
35
38
29
29
25
37
35
31
22
20
29
11
25
26
p80
39
43
34
34
30
41
39
36
27
25
33
13
29
31
p90
47
51
41
41
37
48
47
44
33
31
40
18
36
39
Max
113
137
127
113
98
105
149
106
77
70
97
52
131
125
C-8
-------
Annual Mean
loc_type=CMSA loc_name=Cleveland-Akron, OH CMSA
Annual Mean
29-
27-
26-
25-
24-
23-
3>)03 500431
Figure C-7. Spatial distribution of annual average NO2 concentration, Cleveland CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Cleveland-Akron, OH CMSA
Hourly Cones
50-
Figure C-8. Spatial distribution of hourly NO2 concentration, Cleveland CMSA, years 1995-2006.
C-9
-------
Table C-5. Spatial distribution of annual average NO2 concentration, Cleveland CMSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
3903500431 2 20 14 20 20 20 20 20 20 21 21 21 21 21
3903500601 12 24 3 12 18 22 22 22 22 23 25 26 27 27 28
3903500661 6 18 16 17 17 17 17 17 18 18 19 19 20 20
3903500701 2 15 2 12 14 14 14 14 14 15 17 17 17 17 17
Table C-6. Spatial distribution of hourly NO2 concentration, Cleveland CMSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
3903500431 16215 20 11 54 1 8 11 13 16 18 21 24 28 35 92
3903500601 99696 24 13 53 0 10 13 16 19 22 25 28 33 40 253
3903500661 50100 18 11 60 0 7 9 11 13 15 18 22 26 33 103
3903500701 16619 15 11 70 0 5 7 9 10 13 15 18 23 30 175
C-10
-------
Annual Mean
loc_type=CMSA loc_name=Denver-Boulder-Greeley, CO CMSA
Annual Mean
40-
Figure C-9. Spatial distribution of annual average NO2 concentration, Denver CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Denver-Boulder-Greeley, CO CMSA
Hourly Cones
60-
40-
Figure C-10. Spatial distribution of hourly NO2 concentration, Denver CMSA, years 1995-2006.
C-11
-------
Table C-7. Spatial distribution of annual average NO2 concentration, Denver CMSA, years 1995-2006.
Monitor ID
0800130011
0800500031
0803100021
0805900061
0805900081
0805900091
0805900101
n Mean
11 21
1 26
9 33
3 7
4 9
3 9
5 7
Table C-8. Spatial distribution
Monitor ID
0800130011
0800500031
0803100021
0805900061
0805900081
0805900091
0805900101
SD
3
4
0
1
1
1
cov
14
11
6
7
8
19
Min
15
26
27
7
9
8
6
p10
18
26
27
7
9
8
6
of hourly NO2 concentration,
n Mean SD
83703 21
7790 26
68630 33
22077 7
32449 9
24368 9
39124 7
17
15
15
8
9
9
8
COV
82
57
46
109
97
100
106
Min
0
0
0
0
0
0
0
p10
2
8
15
1
0
1
1
p20
19
26
28
7
9
8
6
Denver
p20
4
12
20
1
2
2
2
p30
20
26
29
7
9
8
6
CMSA
p30
7
16
24
3
3
3
2
p40
21
26
33
7
9
9
7
, years
p40
11
20
28
4
5
5
4
p50 p60
21 22
26 26
34 35
7 7
9 9
9 9
7 7
1995-2006.
p50 p60
17 25
25 29
31 35
5 6
7 9
6 8
5 6
p70
23
26
35
8
9
9
8
p70
32
34
39
9
12
10
9
p80
23
26
35
8
10
9
9
p80
38
39
44
12
15
14
12
p90 Max
23 26
26 26
37 37
8 8
10 10
9 9
9 9
p90 Max
45 239
45 176
51 286
18 66
22 68
20 88
17 98
C-12
-------
Annual Mean
loc_type=CMSA loc_name=Detroit-Ann Arbor-Flint, MI CMSA
Annual Mean
26-
24-
2.1-
15-
14-
1.1-
Figure C-11. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSAloc_name=Detroit-Ann Arbor-Flint, MI CMSA
Hourly Cones
50-
40-
2616300161
Monitor
2616300192
Figure C-12. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006.
C-13
-------
Table C-9. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
2609900091 2 12 0 3 12 12 12 12 12 12 13 13 13 13 13
2616300161 11 21 3 13 16 19 20 20 21 22 22 23 23 24 26
2616300192 11 18 3 14 13 14 15 17 18 19 19 19 19 19 21
Table C-10. Spatial distribution of annual average NO2 concentration, Detroit CMSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
2609900091 16523 12 9 75 0 3 5 6 8 10 12 15 19 25 322
2616300161 86487 21 13 62 0 7 10 13 16 19 23 26 31 38 244
2616300192 88124 18 13 75 0 5 7 9 12 15 18 22 27 35 443
C-14
-------
Annual Mean
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA Set a
Annual Mean
50
40
20
— * — | 1
j
, . « * r-1— 1 *
— i— • « * — T— — r—
~*~ Ly-l — r— [ *
• — — — m —
•
• i
—
000000000000000
6 6 6 6 (J b () 6 6 (> 6 6 6 b b
000000000000000
333333333333333
777777777777777
000001111112455
000120123670000
013100000000000
2603623 5215
211112122221211
Figure C-13. Spatial distribution of annual average NO2 concentration, Los Angeles CMSA set a, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA Set a
Hourly Cones
8(1
70
60
50
40
.10
20
10
t
<
— 4
i
1
(
1
I
>
— _J
i __
4
1
t
L_ t
*
h
h
1
L- *
i
— i
— i —
i
_ —
0000 0 0 0 0 0 000000
6 6 > <) <) 6 6 t> t> <> 6 6 6 > <)
000000000000000
333333333333333
777777777777777
00000 111112455
000120123670000
013100000000000
2603623 11 5215
211112122221211
Figure C-14. Spatial distribution of hourly NO2 concentration, Los Angeles CMSA set a, years 1995-2006.
C-15
-------
Annual Mean
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA Set b
Annual Mean
40-
Figure C-15. Spatial distribution of annual average NO2 concentration, Los Angeles CMSA set b, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA Set b
Hourly Cones
00
50
40
30
20
10
f) 1
{
\ —
_J
I
1
> .
1
>
>
<
\
\
t
1 —
— 1
I
1
<
h
i ~
I
_ (
i
i
1
: ^ :
t
000000000000000
6 6 i b b b ft > > b b f> (> i b
000000000000000
333355556666777
777799995555 11
669900150589000
000000000000000
01030000100001 1
2223173 21 24
11155121221111
Figure C-16. Spatial distribution of hourly NO2 concentration, Los Angeles CMSA set b, years 1995-2006.
C-16
-------
Annual Mean
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA Set c
Annual Mean
30
20
10
0
I — « —
•""" -r-
1 T 1
CD e=g ~"~
S- . _ ~- -^
000000000000000
666666666666666
000000001 1 1 1 1 1 1
777777771111111
111 III 1 1 1 1 1 1 1
000112490011223
003020000000000
110030000000000
576442145734231
111211111111111
Figure C-17. Spatial distribution of annual average NO2 concentration, Los Angeles CMSA set c, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Los Angeles-Riverside-Orange County, CA CMSA Set c
Hourly Cones
60
SO
40
30
30
10
1
>_ _J
1
i
1
1
}
— 1
I —
<
_<
i
I
mi
i
i —
1
~~ _!
i —
<
i —
— T—
II
— 1 —
000000 0 0000 0 0 0 0
66 >
00000000
7 7 7 7 '
1 1
0001
777
1 1
> (j 6 6 6 > <)
i
1
1
24900
1 i i
1 1 1
1 1 1
223
003020000000000
o o :
0000000000
57644214;
2
7 :
1
4231
1 1
Monitor
Figure C-18. Spatial distribution of hourly NO2 concentration, Los Angeles CMSA set c, years 1995-2006.
C-17
-------
Table C-11. Spatial distribution of annual average NO2 concentration, Los Angeles CMSA, years 1995-2006.
Monitor ID
0603700022
0603700161
0603700301
0603701131
0603702061
0603710022
0603711031
0603712012
0603713012
0603716012
0603717012
0603720051
0603740022
0603750011
0603750051
0603760021
0603760121
0603790021
0603790331
0605900015
0605900075
0605910031
0605950012
0606500121
0606550012
0606580012
0606590011
0607100011
0607100121
0607100141
0607100151
0607100171
0607103061
0607110042
0607112341
0607120021
0607140011
0607190041
0611100051
0611100071
0611110031
0611110041
0611120021
0611120031
0611130011
n
12
12
1
12
1
11
11
12
12
10
12
12
11
9
2
2
5
6
5
5
4
12
11
9
12
12
12
12
2
5
2
3
7
11
9
12
3
12
7
9
1
7
12
9
12
Mean
33
28
38
24
45
38
36
26
37
37
39
32
30
27
13
27
20
16
15
33
22
18
30
19
16
23
17
23
7
21
7
6
21
36
5
34
17
31
4
16
10
8
18
10
13
SD
7
5
4
6
6
4
6
4
7
5
5
2
0
4
1
2
0
3
2
3
6
3
3
4
2
1
0
1
2
0
1
4
1
5
1
5
0
1
1
4
1
2
cov
22
17
18
16
16
17
16
11
17
15
16
9
1
14
6
12
3
8
10
16
19
14
22
16
11
6
5
6
28
4
5
12
12
13
4
16
8
9
7
20
8
16
Min
20
20
38
16
45
26
27
17
28
31
30
23
20
23
13
25
18
14
15
29
20
12
21
15
9
17
14
20
7
20
5
6
19
31
5
27
16
25
4
14
10
7
13
9
9
p10
25
22
38
17
45
29
27
20
30
33
31
24
24
23
13
25
18
14
15
29
20
13
25
15
12
19
14
21
7
20
5
6
19
31
5
27
16
26
4
14
10
7
14
9
10
p20
25
24
38
20
45
33
32
21
31
35
31
27
28
23
13
25
19
15
15
31
20
16
25
16
13
21
15
22
7
20
5
6
20
34
5
30
16
26
4
14
10
7
15
9
11
p30
29
26
38
23
45
35
33
24
31
35
35
29
29
24
13
25
19
15
15
32
21
17
25
17
15
22
15
22
7
21
5
6
21
34
5
31
16
26
4
15
10
8
15
9
11
p40
33
26
38
24
45
40
34
25
36
35
36
32
29
27
13
25
19
16
15
32
21
18
27
18
16
22
17
22
7
21
5
6
21
36
5
33
16
29
4
16
10
8
17
9
11
p50
33
27
38
25
45
41
37
26
38
36
40
33
30
28
13
27
19
16
15
33
22
19
28
20
16
23
17
23
7
21
7
6
22
36
5
36
16
31
4
16
10
8
19
10
13
p60
36
28
38
26
45
41
39
26
39
37
43
34
32
28
13
30
20
16
15
33
24
19
33
20
16
24
17
24
7
22
8
6
22
37
5
36
16
33
4
16
10
8
20
10
14
p70
36
29
38
28
45
41
39
28
41
38
43
35
33
29
13
30
20
18
15
33
24
19
34
20
17
25
18
24
7
23
8
7
22
38
6
36
18
34
4
17
10
8
20
11
14
p80
39
32
38
28
45
42
40
28
43
39
44
37
34
29
13
30
21
18
15
35
24
20
35
22
18
26
18
24
7
23
8
7
22
38
6
38
18
35
5
17
10
8
22
11
14
p90
41
33
38
28
45
45
43
31
43
42
46
37
34
30
13
30
21
19
16
37
24
20
35
23
20
29
19
25
7
23
8
7
22
39
6
38
18
38
5
19
10
8
22
11
15
Ma
46
37
38
28
45
45
45
32
46
45
51
37
37
30
13
30
21
19
16
37
24
23
39
23
21
30
20
25
7
23
8
7
22
46
6
42
18
40
5
19
10
8
24
11
16
C-18
-------
"able C-12. Spatial distribution of hourly NO2 concentration,
Monitor ID
0603700022
0603700161
0603700301
0603701131
0603702061
0603710022
0603711031
0603712012
0603713012
0603716012
0603717012
0603720051
0603740022
0603750011
0603750051
0603760021
0603760121
0603790021
0603790331
0605900015
0605900075
0605910031
0605950012
0606500121
0606550012
0606580012
0606590011
0607100011
0607100121
0607100141
0607100151
0607100171
0607103061
0607110042
0607112341
0607120021
0607140011
0607190041
0611100051
0611100071
0611110031
0611110041
0611120021
0611120031
0611130011
n
97734
97838
6817
97124
7604
88656
88425
96922
97352
81411
98551
98151
88730
74014
15047
16534
39399
46871
40341
40987
33847
97546
88510
69857
95624
95642
95010
94741
14753
39719
15531
23713
56831
88766
69325
95054
24587
97785
54034
73031
8240
56869
94238
70332
95263
Mean
33
28
38
24
45
38
36
26
37
37
39
32
30
27
13
27
20
16
15
33
22
18
30
19
16
23
17
23
7
21
7
6
21
36
5
34
17
31
4
16
10
8
18
10
13
SD
20
18
17
16
25
19
19
16
17
18
18
17
17
19
15
15
12
11
11
17
15
15
16
17
12
16
13
17
5
14
6
5
15
17
5
18
11
16
4
12
5
5
13
8
8
cov
59
63
44
67
56
49
52
64
45
48
47
54
58
72
114
57
61
69
73
53
70
85
54
91
73
67
75
76
69
67
82
84
70
48
103
54
68
51
89
74
52
66
70
85
65
Min
0
0
8
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p10
11
8
21
7
19
17
15
7
19
17
19
13
12
5
0
10
4
5
5
14
5
4
12
3
4
6
4
5
2
7
2
2
5
17
1
12
6
12
0
4
4
3
4
1
4
Los Angeles
p20
16
13
25
9
25
23
20
11
24
23
25
18
16
9
1
14
9
7
6
19
9
6
17
5
6
10
6
7
4
9
3
3
8
22
2
19
7
18
1
6
6
4
7
2
6
p30
21
17
28
12
30
28
25
15
28
27
29
22
19
12
2
18
12
9
7
22
10
7
20
7
8
13
8
9
4
11
3
3
11
26
2
24
9
22
3
8
7
5
9
4
7
CMSA
p40
26
21
32
16
34
32
30
19
31
31
33
26
23
17
4
21
16
11
9
26
14
9
24
10
10
17
10
12
5
14
4
4
13
30
3
28
11
26
3
10
8
6
12
6
9
, years
p50
31
25
35
21
39
36
34
23
35
34
36
30
27
23
6
25
19
13
11
30
20
12
27
13
12
21
13
18
6
17
5
5
17
34
3
32
13
30
4
12
9
6
16
8
11
1995-2006.
p60
35
29
38
26
45
41
38
28
39
38
40
34
31
30
10
28
22
17
14
34
23
16
31
18
15
25
17
25
7
22
6
6
22
38
4
37
16
33
5
16
10
7
19
10
13
p70
41
34
42
32
51
45
43
33
43
42
45
38
37
37
17
32
25
21
18
38
30
23
35
25
19
30
22
33
8
27
7
7
28
43
5
42
21
38
5
20
12
9
24
13
15
p80
47
40
48
37
60
52
49
38
48
48
52
44
43
43
26
37
30
26
25
44
36
31
41
34
25
35
27
40
10
33
10
9
34
49
7
48
27
43
6
26
14
11
29
17
18
p90
58
50
57
45
75
62
60
47
57
58
63
52
52
51
36
46
36
32
32
55
42
40
50
43
33
44
34
48
14
41
14
13
42
58
12
58
34
51
8
33
16
14
36
21
23
Max
223
196
160
201
208
262
239
163
250
225
184
225
208
178
91
159
120
140
103
175
127
183
192
307
82
150
127
196
57
113
60
73
100
199
62
170
86
162
81
123
61
66
124
93
127
C-19
-------
Annual Mean
17-
Annual Mean
loc_type=CMSA loc_name=Miami-Fort Lauderdale, FL CMSA
Figure C-19. Spatial distribution of annual average NO2 concentration, Miami CMSA, years 1995-2006.
Hourly Cones
40-
Hourly Concentrations
loc_type=CMSA loc_name=Miami-Fort Lauderdale, FL CMSA
0-
I20MQQJIL 1201180021
Monitor
Figure C-20. Spatial distribution of hourly NO2 concentration, Miami CMSA, years 1995-2006.
C-20
-------
Table C-13. Spatial distribution of annual average NO2 concentration, Miami CMSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
1201100031
1201100311
1201180021
1208600271
1208640022
39 178
89 1 12 7
11 9 1 11 7
11 6 076
11 15 19 13
8
7
8
6
13
Table C-14. Spatial distribution of hourly NO2 concentration,
Monitor ID
1201100031
1201100311
1201180021
1208600271
1208640022
n Mean SD COV Min
24440 9 7 81 0
63306 9 7 78 0
92241 9 11 128 0
87068 6 8 132 0
90717 15 10 67 0
p10
2
2
0
1
5
8
8
8
6
14
Miami
p20
3
3
1
1
7
8
9
8
6
14
CMSA
p30
4
5
1
2
9
9
9
9
6
15
, years
p40
5
6
2
2
11
9
9
9
6
15
9
9
9
6
16
9
9
9
7
16
9
10
10
7
16
9
10
10
7
17
9
10
10
7
17
1995-2006.
p50
7
7
3
3
13
p60
8
9
5
4
15
p70
10
11
11
5
18
p80
13
14
18
9
22
p90
18
18
26
17
28
Max
65
64
128
75
417
C-21
-------
Annual Mean
loc_type=CMSA loc_name=New York-Northern New Jersey-Long Island, N Y-NJ-CT-PA CMS Set a
Annual Mean
50-
Figure C-21. Spatial distribution of annual average NO2 concentration, New York CMSA set a, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=New York-Northern New Jersey-Long Island, NY-NJ-CT-PA CMS Set a
Hourly Cones
70-
Figure C-22. Spatial distribution of hourly NO2 concentration, New York CMSA set a, years 1995-2006.
C-22
-------
Annual Mean
loc_type=CMSA loc_name=New York-Northern New Jersey-Long Island, NY-NJ-CT-PA CMS Setb
Annual Mean
50-
Figure C-23. Spatial distribution of annual average NO2 concentration, New York CMSA set b, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=New York-Northern New Jersey-Long Island, NY-NJ-CT-PA CMS Setb
Hourly Cones
70-
50-
3600500801 3600500831 3600501101 3604700111 3605900052 3606100101 3606100561 3608100971 3608100981 3608101241 3610300092
Monitor
Figure C-24. Spatial distribution of hourly NO2 concentration, New York CMSA set b, years 1995-2006.
C-23
-------
Table C-15. Spatial distribution of annual average
Monitor ID n Mean
0900101131 3 23
0900190031 8 17
0900900271 2 21
0900911231 9 26
3400300011 3 28
3400300051 4 21
3401300111 5 32
3401300161 1 29
3401310031 11 28
3401700061 11 25
3402100051 11 16
3402300111 11 18
3402730011 11 11
3403900042 1 1 38
3403900081 3 28
3600500801 5 35
3600500831 12 28
3600501101 6 30
3604700111 1 33
3605900052 1 1 23
3606100101 4 36
3606100561 10 39
3608100971 3 26
3608100981 7 29
3608101241 5 25
3610300092 6 15
TableC-16. Spatial distribution
Monitor ID n Mean
0900101131 25148 23
0900190031 67123 17
0900900271 16002 21
0900911231 76418 26
3400300011 25620 28
3400300051 34090 21
3401300111 41642 32
3401300161 8368 29
3401310031 93578 28
3401700061 93886 25
3402100051 94591 16
3402300111 94366 18
340273001 1 92642 1 1
3403900042 92472 38
3403900081 23611 28
3600500801 41120 35
3600500831 95448 28
3600501101 46299 29
3604700111 8300 33
3605900052 89801 23
3606100101 30694 36
3606100561 81341 39
3608100971 24104 26
3608100981 56186 29
3608101241 39406 25
3610300092 48236 15
SD COV
1 4
2 11
1 5
1 4
1 2
1 5
1 3
2 7
2 6
1 4
1 6
1 6
4 12
2 6
1 4
2 9
2 6
2 10
1 1
2 6
0 1
1 4
2 7
2 14
of hourly NO2
SD COV
13 55
13 75
14 65
13 50
14 50
14 66
16 50
15 52
14 51
14 56
11 67
12 65
9 82
15 41
13 47
14 40
13 47
13 45
14 41
13 56
11 31
13 33
14 54
13 46
13 50
10 67
Min
23
14
20
24
28
20
31
29
24
22
15
16
10
30
27
33
24
26
33
18
35
34
26
27
23
13
NO2 concentration
p10
23
14
20
24
28
20
31
29
26
23
15
17
11
32
27
33
25
26
33
20
35
35
26
27
23
13
concentration,
Min
0
0
0
0
3
3
3
3
3
2
2
3
0
3
3
0
0
0
3
0
0
0
0
0
0
0
p10
9
4
6
11
11
5
12
11
11
9
4
5
3
19
11
19
13
14
17
8
23
24
10
13
11
5
p20
23
15
20
25
28
20
31
29
27
23
15
18
11
32
27
34
27
29
33
21
35
37
26
28
24
13
New
p20
12
6
8
14
15
8
17
15
15
12
7
8
3
25
16
23
17
18
21
11
27
28
13
17
14
7
p30
23
16
20
25
28
20
31
29
28
25
16
18
11
39
27
35
27
29
33
22
35
38
26
28
25
13
, New York CMSA, years 1995-2006.
p40
24
18
20
25
28
20
32
29
28
26
16
18
11
40
27
35
28
30
33
22
35
38
26
28
25
14
p50
24
18
21
25
28
20
32
29
29
26
16
18
11
40
27
35
29
30
33
24
36
39
26
29
25
16
p60
24
18
22
26
28
20
32
29
29
26
16
19
11
41
27
36
30
30
33
24
36
40
26
30
26
17
p70
24
18
22
26
29
20
33
29
29
26
17
19
12
41
30
36
30
30
33
24
36
40
26
30
27
17
p80
24
19
22
27
29
22
33
29
29
26
17
19
12
41
30
36
31
30
33
25
36
41
26
30
27
17
p90
24
21
22
27
29
22
33
29
29
27
17
19
12
42
30
36
31
32
33
25
36
42
26
30
28
17
Max
24
21
22
27
29
22
33
29
31
27
17
20
12
42
30
36
32
32
33
26
36
42
26
30
28
17
York CMSA, years 1995-2006.
p30
15
8
11
17
19
11
21
18
19
16
8
10
5
29
20
26
20
21
25
14
29
32
17
20
17
8
p40
18
10
14
20
23
14
26
22
23
19
11
13
7
33
24
30
23
24
28
18
32
35
20
24
20
10
p50
22
14
18
24
26
18
31
26
27
23
13
16
8
37
27
33
26
28
32
21
35
38
24
27
23
12
p60
25
18
22
27
31
22
35
31
31
27
16
19
10
41
30
37
30
31
35
25
37
41
28
31
27
15
p70
29
23
27
31
35
27
40
36
35
32
20
23
13
45
34
40
34
35
39
29
40
44
33
35
31
19
p80
34
29
33
36
40
33
45
41
40
37
25
28
17
50
38
45
39
40
43
34
44
48
38
40
36
24
p90
40
36
40
43
47
40
53
49
47
44
32
35
24
58
44
54
46
47
51
40
50
55
45
47
43
31
Max
109
103
101
240
119
124
148
103
150
147
79
99
95
225
122
181
136
119
155
162
118
162
95
114
144
86
C-24
-------
Annual Mean
loc_type=CMSA loc_name=Pliiladelpliia-Wilmington-Atlantic City, PA-NJ-DE-MD CMSA
Annual Mean
34-
33-
32-
31-
30-
20-
38-
37-
3f>-
25-
24-
2.1-
Figure C-25. Spatial distribution of annual average NO2 concentration, Philadelphia CMSA, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Pliiladelpliia-Wilmington-Atlantic City, PA-NJ-DE-MD CMSA
Hourly Cones
60-
50-
40-
Figure C-26. Spatial distribution of hourly NO2 concentration, Philadelphia CMSA, years 1995-2006.
C-25
-------
Table C-17. Spatial distribution of annual average NO2 concentration, Philadelphia CMSA, years 1995-2006.
Monitor ID n Mean
1000310031 5 18
1000310071 1 15
1000320041 4 18
3400700032 10 21
4201700121 12 18
4204500021 12 19
4209100131 11 17
4210100043 11 26
4210100292 10 29
4210100471 9 31
TableC-18. Spatial distribution
SD COV
1 6
1 4
1 7
2 11
2 8
2 13
3 10
3 11
3 10
of hourly NO2
Min
16
15
18
19
15
16
14
22
25
26
p10 p20
16 17
15 15
18 18
20 20
16 16
17 17
14 15
23 24
25 26
26 26
p30
17
15
18
20
16
18
16
24
28
29
concentration, Philadelphia
Monitor ID n Mean SD COV Min
1000310031 40363 18
1000310071 6611 15
1000320041 31615 18
3400700032 84603 22
4201700121 102584 18
4204500021 100344 19
4209100131 93572 17
4210100043 90975 26
4210100292 81218 29
4210100471 75202 31
12 69
9 62
12 63
13 59
12 67
12 64
12 69
13 49
13 43
12 40
0
1
0
3
0
0
0
0
0
0
p10 p20
4 7
6 7
5 8
7 10
5 7
5 8
4 6
10 14
15 19
16 20
p30
10
9
11
13
9
10
9
18
21
23
p40
18
15
18
21
17
18
16
26
28
30
CMSA,
p40
12
10
13
16
12
13
11
20
25
26
p50
18
15
18
22
18
19
16
26
29
32
years
p50
16
12
16
19
15
16
15
24
29
30
p60
18
15
19
22
18
19
17
27
31
32
p70
18
15
19
22
18
19
18
28
32
32
p80
19
15
19
23
20
20
19
28
33
34
p90
19
15
19
24
20
20
19
29
33
34
Max
19
15
19
24
21
21
21
29
33
34
1995-2006.
p60
19
15
20
23
19
20
18
28
30
31
p70
23
17
23
27
23
24
22
31
35
36
p80
28
21
28
32
28
29
27
37
40
40
p90
34
28
34
39
34
36
33
43
46
47
Max
247
69
115
114
106
268
99
190
120
140
C-26
-------
Annual Mean
loc_type=CMSA loc_name=Washington-Baltimore, DC-MD-VA-WV CMSA Seta
Annual Mean
30-
Figure C-27. Spatial distribution of annual average NO2 concentration, Washington DC CMSA set a, years 1995-2006.
Hourly Concentrations
loc_type=CMSA locjiame=Washington-Baltimore, DC-MD-VA-WV CMSA Seta
Hourly Cones
50-
Figure C-28. Spatial distribution of hourly NO2 concentration, Washington DC CMSA set a, years 1995-2006.
C-27
-------
Annual Mean
loc_type=CMSA loc_name=Washington-Ba1timore, DC-MD-VA-WV CMSA Setb
Annual Mean
25-
Figure C-29. Spatial distribution of annual average NO2 concentration, Washington DC CMSA set b, years 1995-2006.
Hourly Concentrations
loc_type=CMSA loc_name=Washington-Baltimore, DC-MD-VA-WV CMSA Setb
Hourly Cones
50—
Figure C-30. Spatial distribution of hourly NO2 concentration, Washington DC CMSA set b, years 1995-2006.
C-28
-------
Table C-19. Spatial distribution of annual average NO2 concentration, Washington DC CMSA,
Monitor ID n Mean
1100100172 1 25
1100100251 12 22
1100100411 12 23
1100100431 12 20
2400530012 8 18
2451000401 11 25
2451000501 1 21
5101300201 12 23
5105900051 11 10
5105900181 3 19
5105910043 6 22
5105910051 4 17
5105950011 10 20
5110710051 8 14
5115300091 12 11
5151000093 12 24
SD COV
2 11
3 12
2 12
2 11
2 7
2 10
1 12
1 3
2 7
1 9
3 15
1 6
2 18
2 8
Min
25
17
16
17
15
22
21
18
7
19
20
15
14
13
7
20
p10
25
19
21
18
15
23
21
21
9
19
20
15
16
13
9
23
Table C-20. Spatial distribution of hourly NO2 concentration,
Monitor ID n Mean
1100100172 8584 25
1100100251 102444 22
1100100411 103173 23
1100100431 102217 20
2400530012 63983 18
2451000401 89589 25
2451000501 7872 21
5101300201 97517 23
5105900051 89964 10
5105900181 22689 19
5105910043 50294 22
5105910051 34022 17
5105950011 79051 20
5110710051 65327 14
5115300091 101671 11
5151000093 98221 24
SD COV
11 45
12 55
12 53
13 64
12 65
11 44
12 60
13 56
7 73
11 60
11 52
11 63
12 61
9 65
7 68
12 48
Min
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
p10
12
9
9
6
5
12
6
8
3
6
10
6
6
5
3
11
p20 p30
25 25
20 21
21 23
18 18
15 17
23 23
21 21
21 22
9 10
19 19
21 21
15 17
17 19
13 14
9 10
23 23
Washington
p20 p30
15 18
11 14
12 15
9 12
7 10
15 18
9 12
11 14
4 5
9 11
12 14
8 9
9 12
7 8
5 6
14 17
p40
25
22
23
19
18
24
21
22
10
19
22
17
20
14
10
24
p50
25
23
24
19
18
25
21
23
10
19
22
17
21
14
11
24
p60
25
23
24
21
18
26
21
23
10
19
23
17
22
14
11
25
years
p70
25
23
25
22
19
26
21
24
11
20
23
17
22
14
11
26
1995-2006.
p80
25
24
25
23
20
26
21
25
11
20
23
18
22
15
12
26
p90
25
24
25
23
20
26
21
25
11
20
23
18
23
16
12
26
Max
25
25
26
24
20
27
21
26
12
20
23
18
24
16
15
27
DC CMSA, years 1995-2006.
p40
20
16
18
15
12
21
16
17
6
13
17
12
14
10
7
20
p50
23
19
21
18
15
23
19
20
8
16
20
14
18
11
9
23
p60
27
23
24
22
19
26
23
24
10
20
23
17
21
14
11
26
p70
30
27
28
26
23
29
27
28
12
24
27
21
25
17
13
29
p80
33
32
33
31
28
33
32
34
15
29
31
26
30
21
16
34
p90
39
39
39
38
34
39
38
41
20
36
38
32
36
28
21
40
Max
113
285
141
258
114
108
75
110
101
89
91
129
155
64
84
115
C-29
-------
Annual Mean
loc_type=MSA loc_name=Atlanta,GA
Annual Mean
30-
1312100481
1322300031
1324700011
Figure C-31. Spatial distribution of annual average NO2 concentration, Atlanta MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Atlanta,GA
Hourly Cones
40-
1324700011
Figure C-32. Spatial distribution of hourly NO2 concentration, Atlanta MSA, years 1995-2006.
C-30
-------
Table C-21. Spatial distribution of annual average NO2 concentration, Atlanta MSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
1308900021
1308930011
1312100481
1322300031
1324700011
10 16 2
9 15 2
12 21 4
10 5 1
12 7 1
Table C-22. Spatial distribution of
Monitor ID
1308900021
1308930011
1312100481
1322300031
1324700011
n Mean
83891 16
72029 15
98975 21
80168 5
100149 7
11
10
17
20
11
hourly NO2
14 14
13 13
16 17
3 4
6 6
concentration,
SD COV Min p10
12 77
11 73
15 73
5 108
6 81
0 3
1 4
0 5
0 1
0 2
15 15
14 15
17 18
4 4
6 6
Atlanta MSA,
p20 p30
5 8
6 8
8 11
1 2
3 3
15
15
19
4
6
years
p40
10
10
14
3
4
15
16
21
5
7
16
16
23
5
7
17
17
24
5
8
18
17
24
6
8
19
18
25
6
8
20
18
27
7
8
1995-2006.
p50
13
12
17
3
5
p60
16
15
21
4
6
p70
20
19
26
5
8
p80
25
24
33
7
10
p90
33
32
43
11
14
Max
139
95
181
70
242
C-31
-------
Annual Mean
loc_type=MSA loc_name=Colorado Springs,CO
Annual Mean
40-
Figure C-33. Spatial distribution of annual average NO2 concentration, Colorado Springs MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Colorado Spnngs,CO
Hourly Cones
60-
40-
Figure C-34. Spatial distribution of hourly NO2 concentration, Colorado Springs MSA, years 1995-2006.
C-32
-------
Table C-23. Spatial distribution of annual average NO2 concentration, Colorado Springs MSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
0804160011
0804160041
0804160051
0804160061
0804160091
0804160111
0804160131
0804160181
6 8
6 17
1 18
1 7
1 12
6 21
1 22
4 22
Table C-24. Spatial distribution
Monitor ID
0804160011
0804160041
0804160051
0804160061
0804160091
0804160111
0804160131
0804160181
n Mean
51373 8
51288 17
8345 18
7993 7
8282 12
50707 21
8637 22
33737 23
1 10
2 10
3 12
8 37
of hourly NO2
SD COV
7 94
11 66
13 74
7 99
10 89
16 77
14 62
21 94
7
16
18
7
12
17
22
18
111
16 16 16
18 18 18
111
12 12 12
17 19 19
22 22 22
18 18 18
7
17
18
7
12
20
22
18
7
17
18
7
12
20
22
19
7
17
18
7
12
20
22
19
8
18
18
7
12
23
22
19
8
18
18
7
12
23
22
35
9
21
18
7
12
24
22
35
9
21
18
7
12
24
22
35
concentration, Colorado Springs MSA, years 1995-2006.
Min
0
0
1
0
0
0
0
0
p10 p20 p30
1 2 3
469
357
1 2 3
234
5 7 10
5811
5 7 10
p40
4
12
10
4
6
14
15
14
p50
5
15
15
5
7
18
20
18
p60
7
20
21
6
10
23
26
23
p70
9
24
27
8
14
27
31
28
p80
12
28
32
11
20
31
36
33
p90
18
34
36
16
29
37
41
41
Max
59
115
143
49
56
246
87
308
C-33
-------
Annual Mean
loc_type=MSA loc_name=EI Paso.TX
Annual Mean
40-
4814100271 4814100281 4814100371 4814100441 4814100551 4814100571 4814100581
Figure C-35. Spatial distribution of annual average NO2 concentration, El Paso MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=EI Paso.TX
Hourly Cones
60-
48N 100271 -18I4ICWCSI 4814 100.17 I 4SI4I0044I 4814100551 4814100571 4814100581
Figure C-36. Spatial distribution of hourly NO2 concentration, El Paso MSA, years 1995-2006.
C-34
-------
Table C-25. Spatial distribution of annual average NO2 concentration, El Paso MSA, years 1995-2006.
Monitor ID
4814100271
4814100281
4814100371
4814100441
4814100551
4814100571
4814100581
n Mean
4 32
1 23
11 18
8 19
7 17
7 14
6 10
Table C-26. Spatial distribution
Monitor ID
4814100271
4814100281
4814100371
4814100441
4814100551
4814100571
4814100581
n Mean
29730 32
8045 23
87748 18
62362 19
53960 17
57229 14
47248 10
SD
3
2
4
1
1
1
cov
10
12
22
5
6
11
of hourly NO2
SD
14
14
13
13
13
11
11
COV
45
60
71
67
78
79
109
Min
28
23
15
13
16
13
8
p10 p20
28 28
23 23
16 17
13 13
16 16
13 13
8 9
p30
31
23
17
18
16
14
9
concentration, El Paso MSA
Min
1
5
0
0
0
0
0
p10 p20
16 20
10 12
5 7
5 8
3 5
3 4
1 2
p30
24
13
9
11
7
6
3
p40 p50
31 32
23 23
17 18
20 21
16 16
14 14
10 10
p60
34
23
18
21
16
15
10
p70
34
23
18
22
16
15
11
p80
35
23
19
23
17
15
11
p90
35
23
21
24
18
16
11
Max
35
23
23
24
18
16
11
, years 1995-2006.
p40 p50
27 30
15 18
12 14
14 17
10 13
8 10
4 5
p60
33
22
18
21
18
14
7
p70
37
27
23
25
23
19
11
p80
42
34
29
30
28
25
18
p90
49
42
36
36
35
31
27
Max
219
117
153
125
87
85
84
C-35
-------
Annual Mean
loc_type=MSA loc_name=Jacksonville,FL
Annual Mean
16—
Figure C-37. Spatial distribution of annual average NO2 concentration, Jacksonville MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Jacksonville,FL
Hourly Cones
40—
Figure C-38. Spatial distribution of hourly NO2 concentration, Jacksonville MSA, years 1995-2006.
C-36
-------
Table C-27. Spatial distribution of annual average NO2 concentration, Jacksonville MSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
1203100322 10 15 16 13 14 14 14 15 15 15 15 16 16 16
Table C-28. Spatial distribution of hourly NO2 concentration, Jacksonville MSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
1203100322 78222 15 10 67 0 5 7 9 10 12 15 18 22 28 294
C-37
-------
Annual Mean
loc_type=MSA loc_name=Las Vegas,NV-AZ
Annual Mean
30-
3200300221 3200300231 3200300731 3200300781 3200305391 3200305571 3200305631 3200306011 3200310191 3200320021
Figure C-39. Spatial distribution of annual average NO2 concentration, Las Vegas MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Las Vegas,NV-AZ
Hourly Cones
70-
50-
.1200300221 1200100211 3200.100731 3200300781 12001053*)! 1200105571 3200305631 1200106011 12001IOI<)1 3200320021
Figure C-40. Spatial distribution of hourly NO2 concentration, Las Vegas MSA, years 1995-2006.
C-38
-------
Table C-29. Spatial distribution of annual average NO2 concentration, Las Vegas MSA, years 1995-2006.
Monitor ID
3200300221
3200300231
3200300731
3200300781
3200305391
3200305571
3200305631
3200306011
3200310191
3200320021
n Mean SD COV Min
75 1 26 4
47 2 28 5
78 197
1 9 9
8 23 3 12 19
2 27 0 1 27
3 19 0 1 19
56 2 34 3
73 1 38 1
7 21 27 19
p10
4
5
7
9
19
27
19
3
1
19
Table C-30. Spatial distribution of hourly NO2 concentration,
Monitor ID
3200300221
3200300231
3200300731
3200300781
3200305391
3200305571
3200305631
3200306011
3200310191
3200320021
n Mean SD COV Min
58087 5 7 152 0
34550 7 8 105 0
56906 8 10 124 0
8672 9 10 115 0
64921 23 16 70 0
16674 27 21 78 0
25061 19 15 78 0
42417 6 8 124 0
57230 3 5 186 0
56244 21 16 73 0
p10
0
0
0
0
5
0
0
0
0
0
p20
4
5
7
9
20
27
19
4
2
20
p30
4
6
8
9
21
27
19
6
2
21
p40
4
6
8
9
22
27
19
6
2
21
p50
5
7
8
9
22
27
19
7
2
22
p60
5
9
8
9
23
27
19
7
3
22
p70
5
9
8
9
25
27
19
8
3
22
p80
6
10
9
9
25
27
19
8
3
22
p90
7
10
9
9
27
27
19
8
4
24
Max
7
10
9
9
27
27
19
8
4
24
Las Vegas MSA, years 1995-2006.
p20
0
0
0
0
7
10
5
0
0
6
p30
0
0
0
0
10
14
7
0
0
9
p40
0
5
0
5
14
19
11
0
0
13
p50
0
6
5
7
21
24
17
5
0
20
p60
5
8
8
8
28
31
23
7
0
27
p70
7
10
11
10
33
37
28
8
0
32
p80
10
13
15
14
38
43
33
12
6
36
p90
15
18
22
22
44
52
39
18
9
42
Max
91
52
104
87
103
410
87
51
71
110
C-39
-------
Annual Mean
loc_type=MSA loc_name=Phoernx-Mesa,AZ
Annual Mean
50-
Figure C-41. Spatial distribution of annual average NO2 concentration, Phoenix MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Phoenix-Mesa,AZ
Hourly Cones
70-
50-
Figure C-42. Spatial distribution of hourly NO2 concentration, Phoenix MSA, years 1995-2006.
C-40
-------
Table C-31. Spatial distribution of annual average NO2 concentration, Phoenix MSA, years 1995-2006.
Monitor ID
0401300191
0401330026
0401330031
0401330101
0401340051
0401340111
0401399971
n Mean
10 27
12 29
10 24
9 35
1 22
2 12
5 24
Table C-32. Spatial distribution
Monitor ID
0401300191
0401330026
0401330031
0401330101
0401340051
0401340111
0401399971
n Mean
81411 27
97376 29
80162 24
73070 35
7420 22
16459 12
41521 24
SD
3
3
4
3
1
3
cov
10
10
17
9
6
12
of hourly NO2
SD
17
17
19
18
13
8
15
COV
63
59
78
53
58
69
60
Min
24
25
19
31
22
11
21
p10
24
25
19
31
22
11
21
concentration,
Min
0
0
0
0
2
0
0
p10
6
8
6
9
7
2
7
p20 p30
24 25
26 29
20 21
31 32
22 22
11 11
22 23
p40
27
29
23
34
22
11
23
p50
28
29
24
35
22
12
24
p60
28
30
24
35
22
12
25
p70
29
32
25
36
22
12
26
p80
29
32
28
37
22
12
27
p90 Max
30 31
33 34
30 31
40 40
22 22
12 12
28 28
Phoenix MSA, years 1995-2006.
p20 p30
9 14
12 17
9 12
16 23
9 13
4 6
10 14
p40
20
23
16
30
17
8
19
p50
26
28
20
35
21
10
23
p60
32
33
25
40
25
13
27
p70
37
38
30
45
29
16
32
p80
42
44
35
50
33
18
37
p90 Max
50 148
53 151
45 267
58 164
39 99
22 53
45 131
C-41
-------
Annual Mean
loc_type=MSA loc_name=Provo-Orem,UT
Annual Mean
29—
27-
Figure C-43. Spatial distribution of annual average NO2 concentration, Provo MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=Provo-Orem,UT
Hourly Cones
50—
40-
Figure C-44. Spatial distribution of hourly NO2 concentration, Provo MSA, years 1995-2006.
C-42
-------
Table C-33. Spatial distribution of annual average NO2 concentration, Provo MSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
4904900021 12 24 2 9 21 22 22 23 23 24 24 24 24 25 29
Table C-34. Spatial distribution of hourly NO2 concentration, Provo MSA, years 1995-2006.
Monitor ID n Mean SD COV Min p10 p20 p30 p40 p50 p60 p70 p80 p90 Max
4904900021 96873 24 16 68 0 6 9 13 17 22 27 31 36 42 164
C-43
-------
Annual Mean
30-
Annual Mean
loc_type=MSA loc_name=St, Louis,MO-lL
Figure C-45. Spatial distribution of annual average NO2 concentration, St. Louis MSA, years 1995-2006.
Hourly Concentrations
loc_type=MSA loc_name=St, Louis,MO-lL
Hourly Cones
50-
Figure C-46. Spatial distribution of hourly NO2 concentration, St. Louis MSA, years 1995-2006.
C-44
-------
Table C-35. Spatial distribution of annual average NO2 concentration, St.
Monitor ID n Mean
1716300102 12 18
2918300101 3 6
2918310021 12 10
2918900012 3 19
2918900041 6 15
2918900062 11 12
2918930012 11 20
2918950011 10 17
2918970022 6 20
2918970031 4 15
2951000722 10 25
2951000801 5 19
2951000861 6 19
SD COV
2 12
0 7
1 13
0 2
2 15
1 12
2 11
2 13
1 6
2 14
2 9
1 5
2 11
Min
15
5
8
19
12
10
17
13
19
12
20
19
15
p10
15
5
8
19
12
10
17
14
19
12
21
19
15
Table C-36. Spatial distribution of hourly NO2 concentration,
Monitor ID n Mean
1716300102 101236 18
2918300101 25873 6
2918310021 99623 10
2918900012 25801 19
2918900041 51987 15
2918900062 93770 12
2918930012 95589 20
2918950011 86912 17
2918970022 51777 20
2918970031 32235 15
2951000722 85643 25
2951000801 42884 19
2951000861 51623 19
SD COV
9 52
6 98
8 81
11 58
10 68
9 79
11 52
11 62
11 54
10 66
11 46
11 59
12 62
Min
0
0
0
0
0
0
0
0
0
0
0
0
0
p10
8
1
2
7
4
3
8
6
8
4
11
7
6
p20
16
5
9
19
13
10
18
15
20
12
23
19
18
p30
16
5
9
19
13
11
19
16
20
16
24
19
18
Louis MSA, years 1995-2006.
p40
17
6
9
19
14
12
20
17
20
16
25
19
19
p50
18
6
10
19
14
12
21
18
20
16
25
19
19
p60
18
6
11
19
14
12
22
19
20
16
25
19
20
p70
19
6
11
19
16
13
22
19
21
16
26
20
21
p80
19
6
11
19
16
13
22
19
21
17
26
20
21
p90
20
6
11
19
18
13
22
19
22
17
27
21
21
Max
21
6
12
19
18
14
24
20
22
17
27
21
21
St. Louis MSA, years 1995-2006.
p20
10
2
3
9
6
4
11
8
11
7
15
10
9
p30
12
2
4
12
8
5
13
10
13
9
18
12
11
p40
14
3
6
14
10
7
16
12
16
11
20
15
14
p50
16
4
8
17
12
9
19
15
18
13
23
17
16
p60
19
5
10
20
15
12
22
18
21
16
26
20
19
p70
21
7
12
23
18
15
25
21
25
19
29
23
23
p80
25
9
16
28
22
19
29
26
29
24
33
28
28
p90
31
13
21
34
29
25
35
32
36
30
40
34
36
Max
123
51
73
89
80
79
101
124
103
64
130
274
87
C-45
-------
Appendix D. Technical Memorandum on Regression Modeling
This appendix provides a technical memorandum submitted to EPA by ICF International.
The memo is as submitted, with the exception of modified page numbering and addition of
borders around each table.
D-1
-------
ICF
INTERNATIONAL
MEMORANDUM
To: Stephen Graham, US EPA
From: Jonathan Cohen and Arlene Rosenbaum
Date: February 15, 2008
Re: Regression Modeling of NO2 Exceedances of 150 ppb versus Annual Mean
si MM4KV
This document describes our regression analyses of 1995 to 2006 NO2 hourly concentration
data. Regression was used to estimate the annual number of exceedances of 150 ppb from the
annual mean, in 20 locations (mostly large urban areas). Exposures to concentrations above
certain thresholds may be associated with adverse health effects. These models were applied in
an as-is scenario to estimate the annual exceedances at sites with annual means equal to the
1995-2006 current average for their location. These models were also applied in a current-
standard scenario to predict the annual exceedances at sites with annual means equal to the
current NO2 standard of 53 ppb. The current-standard scenario is an extrapolation to higher
annual means than currently observed; the maximum annual mean across all complete site-
years was 51 ppb, in Los Angeles.
We found these results unsatisfactory, both because the regression models did not show a
strong relationship between the annual means and the exceedances, and because the
predicted numbers of exceedances for the current-standard scenario were in many cases
extremely high and quite uncertain. For this reason we decided not to apply the regression
modeling to the other concentration levels of interest (200, 250, and 300 ppb) but instead
decided to develop empirical exceedance estimates, as described elsewhere.
All of the 1995 to 2006 NO2 hourly concentration data from AQS were compiled and annual
summary statistics for each site-year combination were computed. Of particular interest is the
long-term air quality measured by the annual mean and the short-term air quality measured by
the annual numbers of hourly exceedances of selected levels 150, 200, 250 and 300 ppb.
Exposures to concentrations above these thresholds may be associated with adverse health
effects. To make the results temporally representative, we restricted the analyses to the 20
percent of site-years that were 75 % complete, as defined by having data for 75 % of the hours
in a year and having data for at least 75 % of the hours in a day (i.e., 18 hours or more) on at
least 75 % of the days in a year. We also spatially grouped the data into 18 urban areas with
high annual means and high exceedances; these locations were all CMSAs or MSAs either with
at least one site-year annual mean above 25.72 ppb (the 90th percentile) or with at least one
exceedance of 200 ppb, as follows.
• Boston
• Cleveland
D-2
-------
ICF International Regression Modeling of NO2 Exceedances
• Denver
• Detroit
• Los Angeles
• New York
• Philadelphia
• Washington DC
• Atlanta
• Colorado Springs
• El Paso
• Las Vegas
• Phoenix
• St. Louis
• Chicago
• Miami
• Jacksonville
• Provo.
The remaining site-years were analyzed as two additional location groups: "Other MSA/CMSA"
site-years in an MSA or CMSA, and "Other Not MSA" site-years not in an MSA. Thus we have a
total of 20 "locations."
ov v a ")•-.. s
The regression modeling of the 1995-2006 NO2 data continues the analyses by McCurdy
(1994)1 of the 1988-1992 data. A regression model is used to estimate the mean number of
exceedances from the annual mean. McCurdy (1994) assumed normally distributed
exceedances and an exponential link function to estimate exceedances of 150, 200, 250, and
300 ppb based on the 1988-1992 data. In this section we present the results of the regression
analyses for exceedances of 150 ppb using eight alternative models based on the 1995-2006
data. Throughout this discussion, "exceedances" will refer to annual numbers of hourly
exceedances of 150 ppb, unless otherwise stated.
Of the eight models, the two selected regression models were the Poisson exponential model
and the normal linear model, stratified by location. The Poisson exponential model is of the
form:
• Number of exceedances has a Poisson distribution.
• Mean exceedances = exp(a + b x annual mean).
• The intercept a, and slope b, depend on the location.
The normal linear model is of the form:
• Number of exceedances has a normal distribution with standard deviation s.
• Mean exceedances = a + b x annual mean.
• The intercept a, slope b, and s all depend on the location.
1 McCurdy TR (1994). Analysis of high 1 hour NO2 values and associated annual averages using 1988-
1992 data. Report to the Office of Air Quality Planning and Standards, Durham NC.
D-3
-------
ICF International
Regression Modeling of NO2 Exceedances
The first issue to be resolved was to decide whether to apply the regression analyses to the
means and exceedances for each season separately or to each year. We examined the
exceedance data for Colorado Springs, which had the highest maximum number of annual
exceedances of 200 ppb, 69, which occurred at site 804160181 in 2000. Of these 69
exceedances, 34 occurred in the winter on January 18-20, 2000, and 35 occurred in the
summer on June 12-14, 2000. This limited analysis suggests that there is no clear pattern of
seasonality in the exceedances. We decided to apply the regression modeling to the annual
means and annual exceedances.
Table 1 describes the eight regression models fitted. As described shortly, we fitted two
distributions (normal and Poisson), two link functions (identity and exponential), and two
stratifications (all data and stratified by location). The McCurdy (1994) analysis used a normal
distribution, an exponential link, and stratified by location into Los Angeles and Not Los Angeles.
We fitted generalized linear models where the number of exceedances has a given distribution
(we fitted normal and Poisson distributions) and where the mean number of exceedances is a
given function g of the annual mean. The function g(x) is called the link function. We can also
define the link by defining the inverse link, i.e., the solution for x of the equation g(x) = y.
We fitted two link functions, an identity link g(x) = x and a logarithmic link g(x) = log(x), where
"log" denote the natural logarithm. The corresponding inverse links are the identity link, which
we also call the "linear" function, and the exponential function. Thus, the linear inverse link
models are of the form:
Mean exceedances = a + b x annual mean.
The exponential inverse link models are of the form:
Mean exceedances = exp(a + b x annual mean).
Table 1. Good ness-of -fit statistics for eight generalized linear models.
Distribution
Normal
Normal
Normal
Normal
Poisson
Poisson
Poisson
Poisson
Inverse
Link
Linear
Linear
Exponential
Exponential
Linear
Linear
Exponential
Exponential
Strata (a
separate
model is
fitted in
each
stratum)
All
Location
All
Location
All
Location
All
Location
R squared
for all data
0.033
0.244
0.066
0.401
0.025
Not Shown*
0.064
0.406
MinR
squared
among
locations
0.006
0.005
Not Shown*
0.004
MaxR
squared
among
locations
0.616
0.981
Not Shown*
0.976
Log-
Likelihood
-11527
-6065
-11438
-8734
-4737
Not Shown*
-3660
-2694
Number of
strata in
final model
1
13**
1
A <4 ***
1
Not Shown*
1
13**
* Model converged for only Cleveland, Atlanta, and "Other Not MSA" locations. Results are not shown
since the model failed to converge for the "Other MSA" location, so the overall goodness-of-fit is not
comparable to the other seven models.
** "Other MSA" includes Chicago, Detroit, Philadelphia, Jacksonville, Las Vegas, Provo, St. Louis.
*** "Other MSA" includes Chicago, Cleveland, Detroit, Philadelphia, Jacksonville, Las Vegas, Phoenix,
Provo, St. Louis.
D-4
-------
ICF International Regression Modeling of NO2 Exceedances
For each link function we fitted models using the normal distribution and the Poisson
distribution. The normal model is at best an approximation since the numbers of exceedances
must be positive or zero integers, but the normal distribution is continuous and includes
negative values. The Poisson model takes the form:
Prob(y exceedances) = (My/y!)e'M, y = 0, 1, 2, ...,
where M is the mean exceedances.
We fitted these four models (two links, two distributions) either to all the data or stratified by
location. Thus the model fitted to all the data assumes that a and b have the same value for all
site-years, and the model fitted by location assumes that a and b have the same value for all
site-years at the same location but these values may vary between locations. For the normal
models, the variance of the number of exceedances is assumed to be the same for all site-years
in each stratum. For the Poisson models, the variance equals the mean number of
exceedances.
The models stratified by location were fitted in two steps. First, each model was separately fitted
to each of the 20 locations. For several models and locations, there were problem cases where
the algorithm either failed to converge to a solution, predicted a negative slope for the annual
mean, or had only zero or one site-year with at least one exceedance. In the second case, if the
slope is negative, then the model implies that exceedances decrease when the annual mean
increases, which is unexpected and could lead to inconsistent results for projecting
exceedances to the current-standard scenario. In the third case, there would be zero degrees of
freedom and the model would be over-fitted for that location. To deal with these problem cases,
we re-allocated all the problem locations into the "Other MSA" combined location and refitted
the models. The results in Table 1 stratified by location are for the refitted models. The re-
allocated locations are listed in the footnotes.
Table 1 gives R squared and log-likelihood goodness-of-fit summary statistics. The R squared
statistic is the squared Pearson correlation coefficient between the observed number of
exceedances and the predicted mean number of exceedances. Negative predicted means are
replaced by zero for this calculation. Values close to 1 indicate a good fit and values close to
zero indicate a poor fit. For the models stratified by location, it is evident that the R squared
value has a wide range across the locations, varying from a very poor fit at some locations to a
very good fit at other locations.
For these models the log-likelihood is a better overall goodness-of-fit statistic. The log-likelihood
is defined as the logarithm of the fitted joint density function to all 4,177 site-years. The better-
fitting models are those with the highest values of the log-likelihood. (The log-likelihood can only
be used to compare different models; its value for a single statistical model is not meaningful).
Of the various normal models, the best-fitting is stratified by location and uses a linear inverse
link. Of the various Poisson models, the best-fitting is stratified by location and uses an
exponential inverse link. The Poisson models fit better than the normal models, which is to be
expected since the actual data are positive or zero discrete count data and the numbers of
exceedances are frequently zero, implying a very small mean.
We selected the Poisson exponential model stratified by location and the normal linear model
stratified by location. The estimated parameter values for these models are displayed in Tables
2 and 3, respectively.
The fitted models for the CMSA locations only are displayed in Figures 1 to 3 and are shown for
all locations in the attached file "corrplots.selected models.doc." Figure 1 and the first three
D-5
-------
ICF International Regression Modeling of NO2 Exceedances
attached plots shows the number of exceedances plotted against the annual mean. These plots
clearly show how weak the relationship between the exceedances and the annual mean is.
Figure 2 and the next three attached plots are for the Poisson exponential model, plotting
predicted versus observed exceedances. Figure 3 and the final three attached plots are for the
normal linear model, plotting predicted versus observed exceedances (negative predictions
were replaced by zero). Comparing the normal and Poisson model predictions, the normal
model tends to under-predict the higher numbers of observed exceedances.
The extensive Tables 7 and 8 at the end of this document and the attached Excel file
"predictions.selected models.xls" contain predicted values and 95 percent confidence and
prediction intervals for the number of exceedances at given mean levels. Table 7 is for the
Poisson exponential model. Table 8 is for the normal linear model. Each table gives calculated
predictions at annual mean values of 20, 30, 40, 50, 53, and 60 ppb and at the minimum, mean,
and maximum annual mean value for each location. The predicted value is the estimated mean
number of exceedances.
Tables 4 and 5 are shorter tables showing only the predictions for a mean of 53 ppb and for the
mean annual mean. The predictions for a mean of 53 ppb estimate the number of exceedances
for a hypothetical site-year with the highest annual mean concentration under the current-
standard scenario, i.e., when the highest annual mean site-year for a given location just meets
the annual standard. The predictions for a mean equal to the mean annual mean estimate the
number of exceedances for the typical "as-is" scenario, i.e., for a hypothetical site-year with an
annual mean that is the average annual mean for that location.
D-6
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 2. Parameters for Poisson exponential model stratified by location.
Location
Type
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
MSA
Location Name
Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
Cleveland-Akron, OH CMSA
Cleveland-Akron, OH CMSA
Cleveland-Akron, OH CMSA
Denver-Boulder-Greeley, CO CMSA
Denver-Boulder-Greeley, CO CMSA
Denver-Boulder-Greeley, CO CMSA
Los Angeles-Riverside-Orange County, CA CMSA
Los Angeles-Riverside-Orange County, CA CMSA
Los Angeles-Riverside-Orange County, CA CMSA
Miami-Fort Lauderdale, FL CMSA
Miami-Fort Lauderdale, FL CMSA
Miami-Fort Lauderdale, FL CMSA
New York-Northern New Jersey-Long Island, NY-NJ-
CT-PA CMS
New York-Northern New Jersey-Long Island, NY-NJ-
CT-PA CMS
New York-Northern New Jersey-Long Island, NY-NJ-
CT-PA CMS
Washington-Baltimore, DC-MD-VA-WV CMSA
Washington-Baltimore, DC-MD-VA-WV CMSA
Washington-Baltimore, DC-MD-VA-WV CMSA
Atlanta, GA
Parameter*
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
Estimate
-6.887
0.144
1.000
-14.209
0.548
1.000
-4.399
0.137
1.000
-5.628
0.181
1.000
-5.780
0.342
1.000
-6.800
0.147
1.000
-6.559
0.145
1.000
-5.081
Standard
Error
2.832
0.116
0.000
4.374
0.164
0.000
1.186
0.038
0.000
0.253
0.006
0.000
1.641
0.114
0.000
1.269
0.037
0.000
3.054
0.135
0.000
1.917
Lower
Confidence
Bound
-14.693
-0.061
1.000
-25.210
0.283
1.000
-7.182
0.070
1.000
-6.134
0.169
1.000
-9.774
0.138
1.000
-9.560
0.079
1.000
-14.610
-0.073
1.000
-9.975
Upper
Confidence
Bound
-2.757
0.430
1.000
-7.312
0.952
1.000
-2.435
0.222
1.000
-5.142
0.194
1.000
-3.068
0.606
1.000
-4.537
0.224
1.000
-2.054
0.482
1.000
-2.139
P-
value
(Chi-
square
test
that
param
eter =
0)
0.02
0.22
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.03
0.28
0.01
D-7
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 2. Parameters for Poisson exponential model stratified by location.
Location
Type
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA/CMSA
MSA/CMSA
MSA/CMSA
Not MSA
Not MSA
Not MSA
Location Name
Atlanta, GA
Atlanta, GA
Colorado Springs, CO
Colorado Springs, CO
Colorado Springs, CO
El Paso,TX
El Paso,TX
El Paso,TX
Phoenix-Mesa,AZ
Phoenix-Mesa,AZ
Phoenix-Mesa,AZ
Other MSA/CMSA
Other MSA/CMSA
Other MSA/CMSA
Other Not MSA
Other Not MSA
Other Not MSA
Parameter*
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Estimate
0.140
1.000
-4.846
0.284
1.000
-10.436
0.350
1.000
-1.568
0.106
1.000
-5.137
0.152
1.000
-4.672
0.227
1.000
Standard
Error
0.099
0.000
0.401
0.012
0.000
2.455
0.074
0.000
0.400
0.013
0.000
0.222
0.010
0.000
0.467
0.036
0.000
Lower
Confidence
Bound
-0.040
1.000
-5.675
0.261
1.000
-16.783
0.233
1.000
-2.363
0.081
1.000
-5.580
0.132
1.000
-5.654
0.158
1.000
Upper
Confidence
Bound
0.363
1.000
-4.097
0.309
1.000
-6.664
0.538
1.000
-0.798
0.131
1.000
-4.711
0.172
1.000
-3.818
0.300
1.000
P-
value
(Chi-
square
test
that
param
eter =
0)
0.16
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Using the report notation, a = "Intercept", and b = "mean." "Scale" equals 1, by definition, for this model.
D-8
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 3. Parameters for normal linear model stratified by location.
Location
Type
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
CMSA
MSA
MSA
Location Name
Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
Boston-Worcester-Lawrence, MA-NH-ME-CT CMSA
Cleveland-Akron, OH CMSA
Cleveland-Akron, OH CMSA
Cleveland-Akron, OH CMSA
Denver-Boulder-Greeley, CO CMSA
Denver-Boulder-Greeley, CO CMSA
Denver-Boulder-Greeley, CO CMSA
Los Angeles-Riverside-Orange County, CA CMSA
Los Angeles-Riverside-Orange County, CA CMSA
Los Angeles-Riverside-Orange County, CA CMSA
Miami-Fort Lauderdale, FL CMSA
Miami-Fort Lauderdale, FL CMSA
Miami-Fort Lauderdale, FL CMSA
New York-Northern New Jersey-Long Island, NY-NJ-
CT-PA CMS
New York-Northern New Jersey-Long Island, NY-NJ-
CT-PA CMS
New York-Northern New Jersey-Long Island, NY-NJ-
CT-PA CMS
Washington-Baltimore, DC-MD-VA-WV CMSA
Washington-Baltimore, DC-MD-VA-WV CMSA
Washington-Baltimore, DC-MD-VA-WV CMSA
Atlanta, GA
Atlanta, GA
Parameter*
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Estimate
-0.023
0.003
0.135
-3.259
0.176
1.755
-0.439
0.044
1.097
-3.301
0.194
4.723
-0.496
0.070
0.828
-0.230
0.013
0.407
-0.032
0.003
0.208
-0.041
0.008
Standard
Error
0.034
0.002
0.009
2.127
0.099
0.265
0.383
0.018
0.129
0.620
0.023
0.174
0.384
0.037
0.088
0.104
0.004
0.022
0.069
0.003
0.013
0.069
0.005
Lower
Confidence
Bound
-0.090
-0.001
0.119
-7.617
-0.027
1.341
-1.211
0.008
0.885
-4.519
0.148
4.402
-1.265
-0.005
0.681
-0.435
0.005
0.368
-0.167
-0.004
0.186
-0.178
-0.002
Upper
Confidence
Bound
0.043
0.006
0.156
1.098
0.378
2.436
0.332
0.080
1.408
-2.083
0.240
5.085
0.273
0.144
1.036
-0.024
0.020
0.454
0.104
0.010
0.236
0.096
0.017
P-value
(Chi-
square
test
that
parame
ter = 0)
0.49
0.17
0.13
0.08
0.25
0.01
0.00
0.00
0.20
0.06
0.03
0.00
0.64
0.35
0.55
0.11
D-9
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 3. Parameters for normal linear model stratified by location.
Location
Type
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA
MSA/CMSA
MSA/CMSA
MSA/CMSA
Not MSA
Not MSA
Not MSA
Location Name
Atlanta, GA
Colorado Springs, CO
Colorado Springs, CO
Colorado Springs, CO
El Paso,TX
El Paso,TX
El Paso,TX
Phoenix-Mesa,AZ
Phoenix-Mesa,AZ
Phoenix-Mesa,AZ
Other MSA/CMSA
Other MSA/CMSA
Other MSA/CMSA
Other Not MSA
Other Not MSA
Other Not MSA
Parameter*
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Intercept
mean
Scale
Estimate
0.226
-36.358
2.689
22.519
-2.017
0.131
0.920
-7.102
0.423
22.513
-0.100
0.013
1.098
-0.064
0.021
0.549
Standard
Error
0.022
11.812
0.674
3.123
0.440
0.024
0.098
15.545
0.557
2.274
0.051
0.003
0.015
0.049
0.006
0.018
Lower
Confidence
Bound
0.189
-60.391
1.318
17.551
-2.898
0.083
0.757
-38.177
-0.689
18.697
-0.201
0.006
1.069
-0.160
0.009
0.514
Upper
Confidence
Bound
0.277
-12.326
4.061
30.362
-1.135
0.178
1.151
23.974
1.536
27.828
0.000
0.019
1.128
0.031
0.032
0.587
P-value
(Chi-
square
test
that
parame
ter = 0)
0.00
0.00
0.00
0.00
0.65
0.45
0.05
0.00
0.19
0.00
* Using the report notation, a = "Intercept", b = "mean", and standard deviation = "Scale."
D-10
-------
ICF International
Regression Modeling of NO2 Exceedances
Observed Exceedances
50
40
30
20
10
Annual Means and Observed Exceedances of 150 ppb
CMSA Locations
•* 4
ttmfm*
10
location ""Boston "Chicago ""Cleveland "Denver
Detroit
30
Annual Mean
~Los Angeles ~Miami
40
50
60
New York Philadelphia "Washington
Figure 1. Exceedances of 150 ppb versus annual means for CMSA locations.
D-11
-------
ICF International
Regression Modeling of NO2 Exceedances
Predicted Exceedances
40
20
10
Observed and Predicted Exceedances of 150 ppb
Poisson exponential model by location
CMSA Locations
• • •
••
o
10
40
20 30
Observed Exceedances
location —Boston —Cleveland ""Denver — Los Angeles Miami ""New York "Washington
50
Figure 2. Predicted and observed exceedances for CMSA locations using Poisson exponential model.
D-12
-------
ICF International
Regression Modeling of NO2 Exceedances
Predicted Exceedances
Observed and Predicted Exceedances of 150 ppb
Normal linear model by location
CMSA Locations
• S
i:
10
40
location ""Boston
20 30
Observed Exceedances
"Cleveland "Denver ~Los Angeles Miami ~New York "Washington
50
Figure 3. Predicted and observed exceedances for CMSA locations using normal linear model
D-13
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 4. As-is and current-standard scenario predictions for Poisson exponential
model, with separate coefficients for each location.
Location
Boston
Boston
Cleveland
Cleveland
Denver
Denver
Los Angeles
Los Angeles
Miami
Miami
New York
New York
Washington
Washington
Atlanta
Atlanta
Colorado
Springs
Colorado
Springs
El Paso
El Paso
Phoenix
Phoenix
Other
MSA/CMSA
Other
MSA/CMSA
Other Not MSA
Other Not MSA
Annual
Mean
53.0
16.8
53.0
21.2
53.0
18.7
53.0
24.3
53.0
9.7
53.0
25.5
53.0
19.4
53.0
12.9
53.0
16.3
53.0
17.7
53.0
27.3
53.0
13.9
53.0
7.0
Observed
Mean
Exceed-
ances
0.019
0.019
0.455
0.455
0.389
0.389
1.403
1.403
0.182
0.182
0.092
0.092
0.030
0.030
0.057
0.057
7.346
7.346
0.295
0.295
4.469
4.469
0.079
0.079
0.081
0.081
Observed
Max
Exceed-
ances
1
1
9
9
6
6
44
44
5
5
3
3
2
2
1
1
143
143
7
7
147
147
39
39
7
7
Predicted
Exceed-
ances
2.081
0.011
1000.000
0.073
17.140
0.158
53.244
0.293
1000.000
0.086
2.737
0.048
3.038
0.023
10.242
0.038
1000.000
0.792
1000.000
0.015
56.901
3.760
18.369
0.048
1000.000
0.046
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.002
0.001
578.253
0.011
2.958
0.057
44.092
0.238
35.520
0.026
0.646
0.022
0.001
0.007
0.012
0.008
1000.000
0.528
177.602
0.001
31.702
3.221
9.388
0.040
85.717
0.028
Upper
Bound
1000.000
0.091
1000.000
0.474
99.308
0.438
64.297
0.360
1000.000
0.281
1 1 .604
0.104
1000.000
0.082
1000.000
0.181
1000.000
1.189
1000.000
0.142
102.130
4.389
35.940
0.058
1000.000
0.075
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
0
0
364
0
2
0
37
0
29
0
0
0
0
0
0
0
1000
0
156
0
26
0
7
0
75
0
Upper
Bound
1000
0
1000
1
98
1
73
2
1000
1
13
1
1000
0
1000
1
1000
3
1000
1
106
8
41
1
1000
1
D-14
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 5. As-is and current-standard scenario predictions for Normal linear
model, with separate coefficients for each location.
Location Name
Boston
Boston
Cleveland
Cleveland
Denver
Denver
Los Angeles
Los Angeles
Miami
Miami
New York
New York
Washington
Washington
Atlanta
Atlanta
Colorado
Springs
Colorado
Springs
El Paso
El Paso
Phoenix
Phoenix
Other
MSA/CMSA
Other
MSA/CMSA
Other Not MSA
Other Not MSA
Annual
Mean
53.0
16.8
53.0
21.2
53.0
18.7
53.0
24.3
53.0
9.7
53.0
25.5
53.0
19.4
53.0
12.9
53.0
16.3
53.0
17.7
53.0
27.3
53.0
13.9
53.0
7.0
Observed
Mean
Exceed-
ances
0.019
0.019
0.455
0.455
0.389
0.389
1.403
1.403
0.182
0.182
0.092
0.092
0.030
0.030
0.057
0.057
7.346
7.346
0.295
0.295
4.469
4.469
0.079
0.079
0.081
0.081
Observed
Max
Exceed-
ances
1
1
9
9
6
6
44
44
5
5
3
3
2
2
1
1
143
143
7
7
147
147
39
39
7
7
Predicted
Exceed-
ances
0.111
0.019
6.046
0.455
1.906
0.389
6.965
1.403
3.199
0.182
0.439
0.092
0.136
0.030
0.360
0.057
106.169
7.346
4.902
0.295
15.339
4.469
0.584
0.079
1.036
0.081
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.000
0.000
0.000
0.000
0.645
0.031
5.561
0.921
0.024
0.000
0.220
0.031
0.000
0.000
0.000
0.000
56.853
0.000
3.249
0.024
0.000
0.000
0.324
0.037
0.505
0.030
Upper
Bound
0.245
0.045
12.267
1.188
3.168
0.747
8.369
1.884
6.375
0.426
0.658
0.152
0.364
0.065
0.739
0.117
155.486
16.002
6.555
0.567
44.043
10.773
0.844
0.120
1.566
0.132
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
36.477
0.000
2.384
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Upper
Bound
0.412
0.289
13.612
4.198
4.490
2.648
16.360
10.703
6.871
1.871
1.272
0.897
0.608
0.443
0.957
0.514
175.862
54.709
7.421
2.172
69.369
50.219
2.752
2.232
2.238
1.161
D-15
-------
ICF International Regression Modeling of NO2 Exceedances
The 95% confidence interval gives the uncertainty of the expected value, i.e., of the average
number of exceedances over hypothetically infinitely many site-years with the same annual
mean. The 95% prediction interval gives the uncertainty of the value for a single site-year,
taking into account both the uncertainty of the estimated parameters and the variability of the
number of exceedances in a given site-year about the overall mean. All prediction intervals were
truncated to be greater than or equal to zero and less than or equal to 1,000. The maximum
possible number of exceedances in a year is the maximum number of hours in a leap year,
8,784. The maximum observed exceedances in a year was 69.
For annual means within the range of the data, the predicted numbers of exceedances are
generally within the range of the observed numbers of exceedances. The normal model
predictions tend to be lower than the Poisson model predictions. At annual mean levels above
the range of the data, the Poisson model with the exponential inverse link sometimes gives
extremely high estimates, well beyond the truncation limit of 1,000. This is mainly due to the
exponential link; each increase of the annual mean by 1 ppb increases the predicted
exceedances by a multiplicative factor of exp(b), where b > 0. The upper bounds of the normal
linear model prediction intervals are at most a more reasonable 202, but these predictions are
less reliable because the Poisson model with an exponential inverse link fits the data much
better. For the normal linear model, each increase of the annual mean by 1 ppb increases the
predicted exceedances by b ppb.
Not shown here are the results for the normal model with an exponential inverse link, which was
the model formulation selected by McCurdy (1994). That model gives roughly similar predictions
to the Poisson model with the exponential inverse link.
We can compare these predictions with the predictions for Los Angeles from McCurdy (1994)
based on 1988-1992 data. Table 6 gives the McCurdy (1994) exceedance estimates for
exceedances of 150 ppb together with our estimates for the 1995-2006 data based on the
Poisson exponential model (see Table 7) and the normal linear model (see Table 8). It is easily
seen that the McCurdy (1994) estimates agree reasonably well with our Poisson exponential
model predictions, with predicted exceedances being a little lower for annual means up to 53
ppb, but a little higher at 60 ppb. The McCurdy (1994) model predicts 75 exceedances at 53
ppb, compared to our Poisson exponential model prediction of 53 exceedances. However, the
McCurdy (1994) estimates are all much higher than our normal linear model predictions. For
example, the McCurdy (1994) model predicts 75 exceedances at 53 ppb, compared to our
normal linear model prediction of 7 exceedances. These findings are primarily due to the fact
that McCurdy also used an exponential link function.
D-16
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 6. Comparison of predicted exceedances of 150 ppb using McCurdy (1994) for 1988-
1992 data and the Poisson exponential and normal linear models for 1995-2006 data.
Annual Mean (ppb)
20
30
40
50
53
60
Predicted Exceedances of 150 ppb
McCurdy (1994)
normal
exponential model.
1988-1 992 data.
4
9
33
57
75
142
Poisson
exponential model.
1995-2006 data.
0
1
5
31
53
189
Normal linear
model. 1995-2006
data.
1
3
4
6
7
8
These analyses found a poor relationship between the annual means and the exceedances of
150 ppb, as well as frequently unrealistically high predictions of exceedances of 150 ppb for the
current-standard scenario. The uncertainty at higher exceedance threshold concentration levels
(200 to 300 ppb) would be expected to be even higher because the numbers of site-years with
non-zero exceedances are even lower (which implies a much weaker numerical relationship
between the annual mean and the annual exceedances). For example, for Los Angeles, the
maximum number of exceedances of 150 ppb was 44, but the maximum number of
exceedances of 200 ppb was only 5. Therefore we chose not to continue the regression
analyses to higher exceedance threshold concentration levels.
Table 7. Predictions for Poisson exponential model, with separate coefficients for
each location.
Location
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Cleveland
Annual
Mean
20.0
30.0
40.0
50.0
53.0
60.0
5.4
16.8
31.0
20.0
Observed
Mean
Exceed-
ances
0.019
0.019
0.019
0.019
0.019
0.019
0.019
0.019
0.019
0.455
Observed
Max
Exceed-
ances
1
1
1
1
1
1
1
1
1
9
Predicted
Exceed-
ances
0.018
0.076
0.321
1.352
2.081
5.692
0.002
0.011
0.089
0.039
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.004
0.010
0.006
0.003
0.002
0.001
0.000
0.001
0.010
0.004
Upper
Bound
0.090
0.576
17.564
661.873
1000.000
1000.000
0.175
0.091
0.801
0.358
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
0
0
0
0
0
0
0
0
0
0
Upper
Bound
1
1
14
680
1000
1000
0
0
1
1
D-17
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 7. Predictions for Poisson exponential model, with separate coefficients for
each location.
Location
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Miami
Miami
Miami
Miami
Miami
Miami
Miami
Miami
Miami
New York
New York
New York
New York
New York
Annual
Mean
30.0
40.0
50.0
53.0
60.0
14.2
21.2
28.1
20.0
30.0
40.0
50.0
53.0
60.0
6.1
18.7
36.8
20.0
30.0
40.0
50.0
53.0
60.0
3.6
24.3
50.6
20.0
30.0
40.0
50.0
53.0
60.0
5.5
9.7
16.8
20.0
30.0
40.0
50.0
53.0
Observed
Mean
Exceed-
ances
0.455
0.455
0.455
0.455
0.455
0.455
0.455
0.455
0.389
0.389
0.389
0.389
0.389
0.389
0.389
0.389
0.389
1.403
1.403
1.403
1.403
1.403
1.403
1.403
1.403
1.403
0.182
0.182
0.182
0.182
0.182
0.182
0.182
0.182
0.182
0.092
0.092
0.092
0.092
0.092
Observed
Max
Exceed-
ances
9
9
9
9
9
9
9
9
6
6
6
6
6
6
6
6
6
44
44
44
44
44
44
44
44
44
5
5
5
5
5
5
5
5
5
3
3
3
3
3
Predicted
Exceed-
ances
9.244
1000.000
1000.000
1000.000
1000.000
0.002
0.073
3.193
0.189
0.740
2.902
11.376
17.140
44.600
0.028
0.158
1.871
0.135
0.825
5.050
30.917
53.244
189.281
0.007
0.293
34.208
2.882
88.023
1000.000
1000.000
1000.000
1000.000
0.020
0.086
0.970
0.021
0.092
0.403
1.760
2.737
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
2.693
29.509
291.652
578.253
1000.000
0.000
0.011
1.490
0.074
0.438
1.201
2.426
2.958
4.659
0.004
0.057
0.925
0.104
0.713
4.632
26.439
44.092
144.681
0.004
0.238
29.084
0.636
2.282
7.591
24.900
35.520
81.274
0.003
0.026
0.380
0.007
0.052
0.211
0.507
0.646
Upper
Bound
31.732
1000.000
1000.000
1000.000
1000.000
0.092
0.474
6.845
0.482
1.251
7.014
53.350
99.308
426.973
0.186
0.438
3.786
0.174
0.954
5.505
36.154
64.297
247.629
0.011
0.360
40.236
13.069
1000.000
1000.000
1000.000
1000.000
1000.000
0.154
0.281
2.475
0.065
0.163
0.773
6.107
1 1 .604
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
2
23
184
364
1000
0
0
0
0
0
0
1
2
4
0
0
0
0
0
1
20
37
138
0
0
22
0
2
7
33
29
40
0
0
0
0
0
0
0
0
Upper
Bound
32
1000
1000
1000
1000
0
1
9
2
3
9
53
98
454
1
1
6
1
3
10
44
73
260
0
2
48
13
1000
1000
1000
1000
1000
1
1
4
0
1
2
7
13
D-18
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 7. Predictions for Poisson exponential model, with separate coefficients for
each location.
Location
New York
New York
New York
New York
Washington
Washington
Washington
Washington
Washington
Washington
Washington
Washington
Washington
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
El Paso
El Paso
Annual
Mean
60.0
9.7
25.5
42.2
20.0
30.0
40.0
50.0
53.0
60.0
6.9
19.4
27.2
20.0
30.0
40.0
50.0
53.0
60.0
3.4
12.9
26.6
20.0
30.0
40.0
50.0
53.0
60.0
6.8
16.3
34.8
20.0
30.0
Observed
Mean
Exceed-
ances
0.092
0.092
0.092
0.092
0.030
0.030
0.030
0.030
0.030
0.030
0.030
0.030
0.030
0.057
0.057
0.057
0.057
0.057
0.057
0.057
0.057
0.057
7.346
7.346
7.346
7.346
7.346
7.346
7.346
7.346
7.346
0.295
0.295
Observed
Max
Exceed-
ances
3
3
3
3
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
143
143
143
143
143
143
143
143
143
7
7
Predicted
Exceed-
ances
7.677
0.005
0.048
0.557
0.026
0.109
0.463
1.968
3.038
8.368
0.004
0.023
0.072
0.102
0.412
1.665
6.735
10.242
27.243
0.010
0.038
0.257
2.295
39.206
669.766
1000.000
1000.000
1000.000
0.054
0.792
153.247
0.032
1.075
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
1.121
0.001
0.022
0.260
0.008
0.011
0.004
0.001
0.001
0.000
0.000
0.007
0.014
0.032
0.034
0.023
0.014
0.012
0.008
0.000
0.008
0.037
1.662
33.759
526.509
1000.000
1000.000
1000.000
0.029
0.528
130.906
0.005
0.536
Upper
Bound
52.548
0.028
0.104
1.193
0.081
1.044
55.438
1000.000
1000.000
1000.000
0.256
0.082
0.366
0.327
4.953
122.647
1000.000
1000.000
1000.000
0.230
0.181
1.770
3.168
45.531
852.001
1000.000
1000.000
1000.000
0.102
1.189
179.401
0.230
2.156
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
26
523
1000
1000
1000
0
0
121
0
0
Upper
Bound
53
0
1
3
1
2
57
1000
1000
1000
1
0
1
1
5
103
1000
1000
1000
0
1
3
6
53
870
1000
1000
1000
1
3
189
1
4
D-19
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 7. Predictions for Poisson exponential model, with separate coefficients for
each location.
Location
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Annual
Mean
40.0
50.0
53.0
60.0
8.2
17.7
35.1
20.0
30.0
40.0
50.0
53.0
60.0
11.1
27.3
40.5
20.0
30.0
40.0
50.0
53.0
60.0
0.5
13.9
34.0
20.0
30.0
40.0
50.0
53.0
60.0
0.3
7.0
Observed
Mean
Exceed-
ances
0.295
0.295
0.295
0.295
0.295
0.295
0.295
4.469
4.469
4.469
4.469
4.469
4.469
4.469
4.469
4.469
0.079
0.079
0.079
0.079
0.079
0.079
0.079
0.079
0.079
0.081
0.081
0.081
0.081
0.081
0.081
0.081
0.081
Observed
Max
Exceed-
ances
7
7
7
7
7
7
7
147
147
147
147
147
147
147
147
147
39
39
39
39
39
39
39
39
39
7
7
7
7
7
7
7
7
Predicted
Exceed-
ances
35.703
1000.000
1000.000
1000.000
0.001
0.015
6.447
1.731
4.988
14.375
41 .422
56.901
119.362
0.673
3.760
15.110
0.122
0.559
2.552
11.648
18.369
53.171
0.006
0.048
1.025
0.878
8.514
82.532
799.989
1000.000
1000.000
0.010
0.046
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
11.290
95.081
177.602
757.520
0.000
0.001
3.454
1.287
4.367
10.922
24.843
31.702
55.901
0.404
3.221
11.361
0.107
0.442
1.681
6.317
9.388
23.650
0.004
0.040
0.756
0.459
2.297
11.133
53.545
85.717
256.785
0.004
0.028
Upper
Bound
112.906
1000.000
1000.000
1000.000
0.020
0.142
12.036
2.329
5.698
18.919
69.066
102.130
254.864
1.119
4.389
20.098
0.140
0.707
3.874
21.480
35.940
119.541
0.010
0.058
1.391
1.681
31.556
611.822
1000.000
1000.000
1000.000
0.025
0.075
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
11
94
156
634
0
0
1
0
1
7
21
26
56
0
0
7
0
0
0
4
7
20
0
0
0
0
1
10
57
75
226
0
0
Upper
Bound
119
1000
1000
1000
0
1
14
5
10
24
71
106
254
3
8
25
1
2
6
25
41
116
0
1
4
3
32
573
1000
1000
1000
0
1
D-20
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 7. Predictions for Poisson exponential model, with separate coefficients for
each location.
Location
Other Not MSA
Annual
Mean
19.7
Observed
Mean
Exceed-
ances
0.081
Observed
Max
Exceed-
ances
7
Predicted
Exceed-
ances
0.823
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.438
Upper
Bound
1.547
95% Prediction
Interval for
Number of
Exceedances
Lower
Bound
0
Upper
Bound
3
D-21
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 8. Predictions for Normal linear model, with separate coefficients for each location.
Location
Name
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Boston
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Cleveland
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Denver
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Los Angeles
Miami
Miami
Miami
Miami
Annual
Mean
20.0
30.0
40.0
50.0
53.0
60.0
5.4
16.8
31.0
20.0
30.0
40.0
50.0
53.0
60.0
14.2
21.2
28.1
20.0
30.0
40.0
50.0
53.0
60.0
6.1
18.7
36.8
20.0
30.0
40.0
50.0
53.0
60.0
3.6
24.3
50.6
20.0
30.0
40.0
50.0
Observed
Mean
Exceed-
ances
0.019
0.019
0.019
0.019
0.019
0.019
0.019
0.019
0.019
0.455
0.455
0.455
0.455
0.455
0.455
0.455
0.455
0.455
0.389
0.389
0.389
0.389
0.389
0.389
0.389
0.389
0.389
1.403
1.403
1.403
1.403
1.403
1.403
1.403
1.403
1.403
0.182
0.182
0.182
0.182
Observed
Max
Exceed-
ances
1
1
1
1
1
1
1
1
1
9
9
9
9
9
9
9
9
9
6
6
6
6
6
6
6
6
6
44
44
44
44
44
44
44
44
44
5
5
5
5
Predicted
Exceed-
ances
0.027
0.052
0.078
0.103
0.111
0.128
0.000
0.019
0.055
0.252
2.008
3.763
5.519
6.046
7.275
0.000
0.455
1.667
0.446
0.888
1.331
1.773
1.906
2.216
0.000
0.389
1.189
0.573
2.510
4.447
6.384
6.965
8.321
0.000
1.403
6.492
0.899
1.596
2.293
2.990
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.141
0.035
0.000
0.000
0.000
0.000
0.000
0.140
0.085
0.353
0.499
0.613
0.645
0.716
0.000
0.031
0.458
0.053
1.962
3.579
5.109
5.561
6.612
0.000
0.921
5.193
0.108
0.092
0.065
0.034
Upper
Bound
0.056
0.107
0.166
0.226
0.245
0.287
0.039
0.045
0.113
1.019
3.874
7.492
11.163
12.267
14.846
0.769
1.188
3.194
0.807
1.424
2.163
2.934
3.168
3.716
0.402
0.747
1.920
1.093
3.058
5.315
7.660
8.369
10.031
0.000
1.884
7.792
1.689
3.099
4.521
5.947
95% Prediction
Interval for Number
of Exceedances
Lower
Bound
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Upper
Bound
0.297
0.327
0.361
0.399
0.412
0.441
0.263
0.289
0.330
4.003
6.173
9.163
12.553
13.612
16.125
3.243
4.198
5.673
2.706
3.185
3.720
4.306
4.490
4.933
2.136
2.648
3.543
9.876
11.814
13.776
15.760
16.360
17.766
6.747
10.703
15.871
2.757
3.873
5.131
6.463
D-22
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 8. Predictions for Normal linear model, with separate coefficients for each location.
Location
Name
Miami
Miami
Miami
Miami
Miami
New York
New York
New York
New York
New York
New York
New York
New York
New York
Washington
Washington
Washington
Washington
Washington
Washington
Washington
Washington
Washington
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Atlanta
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
Annual
Mean
53.0
60.0
5.5
9.7
16.8
20.0
30.0
40.0
50.0
53.0
60.0
9.7
25.5
42.2
20.0
30.0
40.0
50.0
53.0
60.0
6.9
19.4
27.2
20.0
30.0
40.0
50.0
53.0
60.0
3.4
12.9
26.6
20.0
30.0
40.0
50.0
53.0
Observed
Mean
Exceed-
ances
0.182
0.182
0.182
0.182
0.182
0.092
0.092
0.092
0.092
0.092
0.092
0.092
0.092
0.092
0.030
0.030
0.030
0.030
0.030
0.030
0.030
0.030
0.030
0.057
0.057
0.057
0.057
0.057
0.057
0.057
0.057
0.057
7.346
7.346
7.346
7.346
7.346
Observed
Max
Exceed-
ances
5
5
5
5
5
3
3
3
3
3
3
3
3
3
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
143
143
143
143
143
Predicted
Exceed-
ances
3.199
3.687
0.000
0.182
0.677
0.023
0.149
0.275
0.401
0.439
0.527
0.000
0.092
0.302
0.032
0.063
0.095
0.127
0.136
0.158
0.000
0.030
0.054
0.110
0.186
0.262
0.337
0.360
0.413
0.000
0.057
0.161
17.426
44.318
71.210
98.102
106.169
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.024
0.001
0.000
0.000
0.103
0.000
0.079
0.148
0.204
0.220
0.256
0.000
0.031
0.161
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.020
0.015
0.001
0.000
0.000
0.000
0.000
0.000
0.019
7.454
24.197
38.662
52.682
56.853
Upper
Bound
6.375
7.373
0.281
0.426
1.250
0.096
0.218
0.401
0.598
0.658
0.798
0.028
0.152
0.444
0.067
0.143
0.237
0.335
0.364
0.432
0.081
0.065
0.117
0.201
0.357
0.522
0.689
0.739
0.857
0.092
0.117
0.303
27.398
64.439
103.758
143.522
155.486
95% Prediction
Interval for Number
of Exceedances
Lower
Bound
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
13.462
31.411
36.477
Upper
Bound
6.871
7.834
1.607
1.871
2.449
0.829
0.955
1.088
1.228
1.272
1.375
0.707
0.897
1.118
0.445
0.483
0.531
0.589
0.608
0.654
0.412
0.443
0.471
0.573
0.672
0.787
0.916
0.957
1.055
0.452
0.514
0.637
65.075
95.397
128.958
164.793
175.862
D-23
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 8. Predictions for Normal linear model, with separate coefficients for each location.
Location
Name
Colorado
Springs
Colorado
Springs
Colorado
Springs
Colorado
Springs
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
El Paso
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Phoenix
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Other
MSA/CMSA
Annual
Mean
60.0
6.8
16.3
34.8
20.0
30.0
40.0
50.0
53.0
60.0
8.2
17.7
35.1
20.0
30.0
40.0
50.0
53.0
60.0
11.1
27.3
40.5
20.0
30.0
40.0
50.0
53.0
60.0
0.5
13.9
34.0
Observed
Mean
Exceed-
ances
7.346
7.346
7.346
7.346
0.295
0.295
0.295
0.295
0.295
0.295
0.295
0.295
0.295
4.469
4.469
4.469
4.469
4.469
4.469
4.469
4.469
4.469
0.079
0.079
0.079
0.079
0.079
0.079
0.079
0.079
0.079
Observed
Max
Exceed-
ances
143
143
143
143
7
7
7
7
7
7
7
7
7
147
147
147
147
147
147
147
147
147
39
39
39
39
39
39
39
39
39
Predicted
Exceed-
ances
124.994
0.000
7.346
57.235
0.594
1.900
3.205
4.511
4.902
5.816
0.000
0.295
2.567
1.367
5.601
9.835
14.069
15.339
18.303
0.000
4.469
10.035
0.158
0.287
0.416
0.545
0.584
0.674
0.000
0.079
0.339
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
66.550
0.000
0.000
31.241
0.303
1.270
2.140
2.994
3.249
3.844
0.000
0.024
1.719
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.100
0.173
0.239
0.304
0.324
0.368
0.000
0.037
0.200
Upper
Bound
183.438
0.000
16.002
83.228
0.886
2.529
4.270
6.027
6.555
7.789
0.000
0.567
3.416
1 1 .546
12.546
25.027
39.591
44.043
54.495
16.406
10.773
25.696
0.216
0.401
0.593
0.786
0.844
0.980
0.003
0.120
0.477
95% Prediction
Interval for Number
of Exceedances
Lower
Bound
47.873
0.000
0.000
3.296
0.000
0.000
1.049
2.085
2.384
3.065
0.000
0.000
0.516
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Upper
Bound
202.115
31.109
54.709
111.173
2.474
3.866
5.361
6.936
7.421
8.568
0.981
2.172
4.619
47.846
51.449
57.734
66.390
69.369
76.880
46.824
50.219
58.093
2.311
2.442
2.576
2.711
2.752
2.848
2.061
2.232
2.495
D-24
-------
ICF International
Regression Modeling of NO2 Exceedances
Table 8. Predictions for Normal linear model, with separate coefficients for each location.
Location
Name
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Other Not MSA
Annual
Mean
20.0
30.0
40.0
50.0
53.0
60.0
0.3
7.0
19.7
Observed
Mean
Exceed-
ances
0.081
0.081
0.081
0.081
0.081
0.081
0.081
0.081
0.081
Observed
Max
Exceed-
ances
7
7
7
7
7
7
7
7
7
Predicted
Exceed-
ances
0.351
0.558
0.766
0.973
1.036
1.181
0.000
0.081
0.345
95% Confidence
Interval for Mean
Number of
Exceedances
Lower
Bound
0.193
0.290
0.384
0.477
0.505
0.571
0.000
0.030
0.190
Upper
Bound
0.508
0.827
1.148
1.469
1.566
1.791
0.035
0.132
0.499
95% Prediction
Interval for Number
of Exceedances
Lower
Bound
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Upper
Bound
1.440
1.669
1.910
2.161
2.238
2.421
1.024
1.161
1.434
D-25
-------
Appendix E. Technical Memorandum on Land Use and
Surface Analysis
Figures E-l to E-5 show the manually created land-use sectors around each application site; in
each case a 1.9 mile (3 km) radius circle was used. The city centers are also labeled. Data in
each case are from the NLCD92 database. Prior to the release of AERSURFACE, the user was
required to manually pull values of Bowen ratio (Po), albedo (a), and surface roughness (z0) per
season and per land-use sector from look-up tables in the AERMET User's Guide. Using the
look-up tables, values of these three surface characteristics vary by the four seasons and by eight
basic land-use categories. Furthermore, the AERMOD Implementation Guide was somewhat
ambiguous about whether Bowen ratio values should also vary with wind direction sector, as
does the surface roughness. AERSURFACE resolves these issues by providing a uniform
methodology for calculation of surface effects on dispersion; it also only varies surface
roughness by wind direction.
Before AERSURFACE, without an automated algorithm to determine land-use patterns, it was
simplest for the user to visually estimate land usage by sector. With AERSURFACE, the land-
use is automatically determined. The proximity of the meteorological site to an airport and
whether the site was located in an arid region were previously not explicitly accounted for as
they now are in AERSURFACE. Snow cover, too, is critical for determination of a, but was
largely left to user's discretion regarding its presence. With AERSURFACE, the lookup tables
have separate columns for winter without much snow and for winter with abundant snow. The
user determines if winter at a particular location contains at least one month of continuous snow
cover, and AERSURFACE will pull values of the surface characteristics from the appropriate
winter column.
We conducted a sensitivity test to evaluate the impacts of using this new tool on the present
analysis. Figure F-6 shows a sample comparison of surface roughness values at the Philadelphia
site with and without the use of AERSURFACE. In the Figure, estimated surface roughness
values using visual land-use estimations and look-up table values are shown in muted shades and
AERSURFACE values in dark shades. Monthly season definitions are the same in both cases.
However, in the AERSURFACE case, winter was specified as having a one-month period of
snow cover. Also, in the AERSURFACE case the site was specified as being at an airport.
In this case, ZQ values are much lower with AERSURFACE than with a visual estimation of land-
use. In the AERSURFACE tool, Philadelphia was noted as being at an airport, tending to
represent the lower building heights in the region and the inverse distance weighting
implemented in the tool. Thus, lower z0 values were obtained over most developed-area sectors
in this scenario. The indication that at least one month of continuous snow cover is present also
tends to lower wintertime ZQ values. In addition to these systematic differences, the automated
AERSURFACE land-use analysis for Philadelphia tended to identify less urban coverage and
more water coverage, lowering roughness values, but it also tended to identify more forest cover
and less cultivated land cover than our visual analysis, increasing some ZQ values.
Po and a also varied significantly between the scenarios. However, this was largely due to two
practical matters: First, the independence of these variables of wind direction in the
AERSURFACE case and secondly the use of monthly-varying moisture conditions in one test
case and not another. Thus we have not presented those results here.
E-l
-------
84z«rw
^, ;-V;V^-'S
^-pl'-' ^
t
•• r "•»' f ' *v 5rv'i''
> >
ia»&4'j«.ir«i
••%«-<-, ,.^
s •>5».V..«W>i
egend
Openwater
_ Lowintenatyfiesldenbai
lZH HlghlntensityResidentlal
^^| Commercial/lndustnal/Transportaion~33!38'°"N
' BareRock/SancClay
Transitonai
DecifluousForest
EveigreerForest
j MIxaJForest
Shnfcland
Grasslan*Hert)3ceous
rops
Urtjan/RecreatonalGrasses
Woodywedands
lergenW erbaceous Wetlands
T
•334ffO"N
33-3BCTM
••33'3ffO''N
Figure E-1. Land-use and sectors around the Atlanta-area surface meteorological station (KATL). Sector
borders are 43, 104, and 255 degrees from geographic North. Atlanta city center is labeled.
E-2
-------
12mAf 83 1 (HAW 838'0'W S3 8TJW
J - ™ - L. - . J. — I,.
83 O'CTW
42 3B'0"I*
42 2ffO'N
S32ffCfW 83'18'CTW 83 16OT/V 83 14'OTW 8312
I I I I |
[Legend
O pen Water
Low IntenatyResldentiai
H IghlnlEnsltyR Ktiential
CommerciaWndu5tn3l/Tran5portation
BareR
:***3?
^ ii'ut .1
Q OeciauousForest
Ev^igreerf onesl
MixMForest
•^^^( Crops
U ro an/Recreate nalGrasses
0 1.5 *3 f
I i i\i I i. ii I '
WoodyWeaands
m
| EmergenlMertjaceousWetlands
83 16'CTW 83 14'fflW S31JO'W 83 1 00' W S3 SQ'W S3 SOW S3 4'0"W
Figure E-2. Land-use and sectors around the Detroit-area surface meteorological station (KDTW).
Sector borders are 49, 117,21 7, and 322 degrees from geographic North. Detroit city center is labeled.
E-2
-------
34 tSCTU-
fr^fti* f^i^^^%**is 1^*3
I Shnjbland
GrasslantfHerbaceous
PastureWary
Crops
an/Re create nalGrasses
Figure E-3. Land-use and sectors around the Los Angeles-area surface meteorological station (KLAX).
Sector borders are 34, 96, and 275 degrees from geographic North. Los Angeles city center is labeled.
E-4
-------
OpenWater
LowlntenslyPesidentiai
HighlnEnsityResUential
CommerciaMndustrial/Transportaion
^1 BareRocwsanociay
Quairies/stripMines/GtaveiPlls
Trans iional
DeciduousForest
EvergreerFore5t
MixedForest
ShruDland
GrasslanO/Herbaceous
PaslureflHay
Crops
Urtjan/RecreatonalGrasses
WoodyWetlands
Emergenwemxeouswetlancls
Figure E-4. Land-use and sectors around the Philadelphia-area surface meteorological station (KPHL).
Sector borders are 80, 184, 262, and 312 degrees from geographic North. Philadelphia city center is
labeled.
E-5
-------
112'2'ffW
I
nzo'trw
i
f"f*£5'*zf4i:
egend
OpenWater
LowintensityResKJentiai
HlgrtlnBnsityR essential
Commercial/lndustnal/Transportation
| BarePocK^SanoClay
Qua mes/StnpM I nesrtSr
Transitonal
DeciduousForest
j^^ EvErgreerForest
~] MlxedForest
^] ShnJbland
f Grass land/Her teceous
1 | Paslur&Hay
~ Crops
urn an/Recreate nalGrasses
WoodyWeMands
EmergentH erbaceous Wetlands
0 0.5 1 2 Kilometers
i i I i i i i
Figure E-5. Land-use and sectors around the Phoenix-area surface meteorological station (KPHX).
Sector borders are 47, 153, 233, and 304 degrees from geographic North. Phoenix city center is labeled.
E-6
-------
0.7
0.6
0.5
8 0.4
O 0.3
0.2
100
150 200
Wind direction (degrees)
250
300
350
» Minimum (old) A Maximum (old) A Average (old) • Minimum (new) • Maximum (new) • Average (new)
Figure E-6. Estimated z0 Values for the Philadelphia Scenario Using Visual and AERSURFACE Land-
Use Estimations.
E-7
-------
Appendix F. Technical Memorandum on Longitudinal Diary
Construction Approach
F-l
-------
ICF
INTERNATIONAL
TECHNICAL MEMORANDUM
TO: Stephen Graham and John Langstaff, US EPA
FROM: Arlene Rosenbaum
DATE: February 29, 2008
SUBJECT: The Cluster-Markov algorithm in APEX
Background
The goals of population exposure assessment generally include an accurate estimate of
both the average exposure concentration and the high end of the exposure distribution.
One of the factors influencing the number of exposures at the high end of the
concentration distribution is time-activity patterns that differ from the average, e.g., a
disproportionate amount of time spent near roadways. Whether a model represents these
exposure scenarios well depends on whether the treatment of activity pattern data
accurately characterizes differences among individuals.
Human time-activity data for population exposure models are generally derived from
demographic surveys of individuals' daily activities, the amount of time spent engaged in
those activities, and the ME locations where the activities occur. Typical time-activity
pattern data available for inhalation exposure modeling consist of a sequence of
location/activity combinations spanning a 24-hour duration, with 1 to 3 records for any
single individual. But modeling assessments of exposure to air pollutants typically
require information on activity patterns over long periods of time, e.g., a full year. For
example, even for pollutant health effects with short averaging times (e.g., ozone 8-hour
average) it may be desirable to know the frequency of exceedances of a threshold
concentration over a long period of time (e.g., the annual number of exceedances of an 8-
hour average ozone concentration of 0.07 ppm for each simulated individual).
Long-term activity patterns can be estimated from daily ones by combining the daily
records in various ways, and the method used for combining them will influence the
variability of the long-term activity patterns across the simulated population. This in turn
will influence the ability of the model to accurately represent either long-term average
high-end exposures, or the number of individuals exposed multiple times to short-term
high-end concentrations.
A common approach for constructing long-term activity patterns from short-term records
is to re-select a daily activity pattern from the pool of data for each day, with the implicit
assumption that there is no correlation between activities from day to day for the
simulated individual. This approach tends to result in long-term activity patterns that are
very similar across the simulated population. Thus, the resulting exposure estimates are
likely to underestimate the variability across the population, and therefore, underestimate
the high-end concentrations.
F-2
-------
A contrasting approach is to select a single activity pattern (or a single pattern for each
season and/or weekday-weekend) to represent a simulated individual's activities over the
modeling period. This approach has the implicit assumption that an individual's day to
day activities are perfectly correlated. This approach tends to result in long-term activity
patterns that are very different across the simulated population, and therefore may over-
estimate the variability across the population.
The Cluster-Markov algorithm
Recently, a new algorithm has been developed and incorporated into APEX that attempts
to more realistically represent the day-to-day correlation of activities for individuals. The
algorithms first use 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 the assumption of no day-to-day correlation (i.e., re-selection for
each time period) and perfect correlation (i.e., selection of a single daily record to
represent all days).
The steps in the algorithm are as follows.
• For each demographic group (age, gender, employment status), temperature range,
and day-of-week combination, the associated time-activity records are partitioned into
3 groups using cluster analysis. The clustering criterion is a vector of 5 values: the
time spent in each of 5 microenvironment categories (indoors - residence; indoors -
other building; outdoors - near road; outdoors - away from road; in vehicle)..
• For each simulated individual, a single time-activity record is randomly selected from
each cluster.
• Next the Markov process determines the probability of a given time-activity pattern
occurring on a given day based on the time-activity pattern of the previous day and
cluster-to-cluster transition probabilities. The cluster-to-cluster transition probabilities
are estimated from the available multi-day time-activity records. (If insufficient
multi-day time-activity records are available for a demographic group, season, day-
of-week combination, then the cluster-to-cluster transition probabilities are estimated
from the frequency of time-activity records in each cluster in the CHAD data base.)
Attachment 1 presents the Cluster-Markov algorithm in flow chart format.
Evalaution against observations (Rosenbaum and Cohen 2004)
The Cluster-Markov algorithm is also incorporated into the Hazardous Air Pollutant
Exposure Model (HAPEM). The algorithm in HAPEM was tested using multi-day diary
data sets collected as part of the Harvard Southern California Chronic Ozone Exposure
Study (Xue et al. 2005, Geyh et al. 2000). In this study, 224 children in ages between 7
and 12 yr were followed for 1 year from June 1995 to May 1996, for 6 consecutive days
each month. The subjects resided in two separate areas of San Bernardino County: urban
F-3
-------
Upland CA, and the small mountain towns of Lake Arrowhead, Crestline, and Running
Springs, CA.
For purposes of clustering 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. For purposes of defining diary pools and for
clustering and calculating transition probabilities the activity pattern records were divided
by day type (i.e., weekday, weekend), season (i.e., summer or ozone season, non-summer
or non-ozone season), age (7-10 and 11-12), and gender.
Week-long sequences (Wednesday through Tuesday) for each of 100 people in each
age/gender group for each season were simulated.
To evaluate the algorithm the following statistics were calculated for the predicted multi-
day activity patterns and compared them with the actual multi-day diary data.
• For each age/gender group for each season, the average time in each
microenvironment
For each simulated person-week and microenvironment, the average of the within-
person variance across all simulated persons (The within-person variance was
defined as the variance of the total time per day spent in the microenvironment
across the week.)
For each simulated person-week the variance across persons of the mean time spent
in each microenvironment.
In each case the predicted statistic for the stratum was compared to the statistic for the
corresponding stratum in the actual diary data. The mean normalized bias for the statistic,
which is a common performance measure used in dispersion model performance and was
also calculated as follows.
(predicted - observed)
j
N i observed
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%.
Fourteen predictions have positive bias and 23 have negative bias.
F-4
-------
For the variance across persons for the average time spent in each microenvironment, the
bias ranged from -40% to +120% for any microenvironment/age/gender/season. Sixty-
five percent of the predicted variances had bias between -22% and +24%. The mean
normalized bias across any microenvironment ranged from -10% to +28%. Eighteen
predictions had positive bias and 20 had negative bias.
For the within-person variance for time spent in each microenvironment, the bias ranged
from -47% to +150% for any microenvironment/age/gender/season. Seventy percent of
the predicted variances had bias between -25% and +30%. The mean normalized bias
across any microenvironment ranged from -11% to +47%. Twenty-eight predictions had
positive bias and 12 had negative bias, suggesting some tendency for overprediction of
this variance measure.
The overall conclusion was that the proposed algorithm appeared to be able to replicate
the observed data reasonably well. Although some discrepancies were rather large for
some of the "variance across persons" and "within-person variance" subsets, about two-
thirds of the predictions for each case were within 30% of the observed value.
A detailed description of the evaluation is presented in Attachment 2.
Comparison with other algorithms (US EPA 2007)
As part of the application of APEX in support of US EPA's recent review of the ozone
NAAQS several sensitivity analyses were conducted. One of these was to make parallel
simulations using each of the three algorithms for constructing multi-day time-activity
sequences that are incorporated into APEX.
Table 1 presents the results for the number of persons in Atlanta population groups with
moderate exertion exposed to 8-hour average concentrations exceeding 0.07 ppm. The
results show that the predictions made with alternative algorithm Cluster-Markov
algorithm are substantially different from those made with simple re-sampling or with the
Diversity-Autocorrelation algorithm ("base case"). Note that for the cluster algorithm
approximately 30% of the individuals with 1 or more exposure have 3 or more exposures.
The corresponding values for the other algorithms range from about 13% to 21%.
Table 2 presents the results for the mean and standard deviation of number of days/person
with 8-hour average exposures exceeding 0.07 ppm with moderate or greater exertion.
The results show that although the mean for the Cluster-Markov algorithm is very similar
to the other approaches, the standard deviation is substantially higher, i.e., the Cluster-
Markov algorithm results in substantially higher inter-individual variability.
F-5
-------
Table 1. Sensitivity to longitudinal diary algorithm: 2002 simulated counts of
Atlanta general population and children (ages 5-18) with any or three or more 8-
hour ozone exposures above 0.07 ppm concomitant with moderate or greater
exertion (after US EPA 2007).
One or more exposures
Population
Group
General
Population
Children (5-
18)
Simple re-
sampling
979,533
411,429
Diversity-
Autocorrel
ation
939,663
(-4%)
389,372
(-5%)
Cluster-
Markov
668,004
(-32%)
295,004
(-28%)
Three
Simple re-
sampling
124,687
71,174
or more exposures
Diversity-
Autocorrel
ation
144,470
(+16%)
83,377
(+17%)
Table 2. Sensitivity to longitudinal diary algorithm: 2002 days per person with
hour ozone exposures above 0.07 ppm concomitant with moderate or greater
exertion for Atlanta general population and children (ages 5-18) (after US EPA
2007).
Mean Days/Person
Population
Group
General
Population
Children (5-
18)
Simple re-
sampling
0.332
0.746
Base case
0.335
(+1%)
0.755
(+1%)
Cluster
0.342
(+3%)
0.758
(+2%)
Cluster-
Markov
188,509
(+51%)
94,216
(+32%)
8-
Standard Deviation
Simple re-
sampling
0.757
1.077
Base case
0.802
(+6%)
1.171
(+9%)
Cluster
1.197
(+58%)
1.652
(+53%)
F-6
-------
References
Geyh, AS, Xue, J, Ozkaynak, H, and Spengler, JD. 2000. The Harvard Southern
California chronic ozone exposure study: Assessing ozone exposure of grade-school-age
children in two Southern California communities. Environ Health Persp. 108:265-270.
Rosenbaum, AS, and Cohen JP. 2004. Evaluation of a multi-day activity pattern
algorithm for creating longitudinal activity patterns. Memorandum prepared for Ted
Palma. USEPA OAQPS by ICF International.
US EPA . 2007. Ozone Population Exposure Analysis for Selected Urban Areas. EPA-
452/R-07-010 http://www.epa.gov/ttn/naaqs/standards/ozone/data/2007-
01 o3 exposure tsd.pdf.
Xue J, Liu SV, Ozkaynak H, Spengler J. 2005. Parameter evaluation and model
validation of ozone exposure assessment using Harvard Southern California Chronic
Ozone Exposure Study Data. J. Air & Waste Manage. Assoc. 55:1508-1515.
F-7
-------
ATTACHMENT 1
F-8
-------
CHAD Data Base
Demographic
Group 1
Weekday
Season 1
Demographic
Group 1
Weekday
Season 2
Demographic
Group 1
Weekend
CLUSTER
ANALYSIS
TRANSITION
PROBABILITIES
Annual Time-Activity Sequence
Figure A1 -1. Flow chart of Cluster-Markov algorithm.
F-9
-------
ATTACHMENT 2
F-10
-------
ICF
CONSULTING
TECHNICAL MEMORANDUM
TO: Ted Palma, US EPA
FROM: Arlene Rosenbaum and Jonathan Cohen, ICF Consulting
DATE: November 4, 2004
SUBJECT: Evaluation of a multi-day activity pattern algorithm for creating longitudinal
activity patterns.
BACKGROUND
In previous work ICF reviewed the HAPEM4 modeling approach for developing annual
average activity patterns from the CHAD database and recommended an approach to
improve the model's pattern selection process to better represent the variability among
individuals. This section summarizes the recommended approach. (For details see the
memorandum of July 23, 2002 from ICF Consulting to Ted Palma.)
Using cluster analysis, first the CHAD daily activity patterns are grouped into either two or
three categories of similar patterns for each of the 30 combinations of day type (summer
weekday, non-summer weekday, and weekend) and demographic group (males or females;
age groups: 0-4, 5-11, 12-17, 18-64, 65+). Next, for each combination of day type and
demographic group, category-to-category transition probabilities are defined by the relative
frequencies of each second-day category associated with each given first-day category, where
the same individual was observed for two consecutive days. (Consecutive day activity pattern
records for a single individual constitute a small subset of the CHAD data.)
To implement the proposed algorithm, for each day type and demographic group, one daily
activity pattern per category is randomly selected from the corresponding CHAD data to
represent that category. That is, if there are 3 cluster categories for each of 3 day types, 9
unique activity patterns are selected to be averaged together to create an annual average
activity pattern to represent an individual in a given demographic group and census tract.
The weighting for each of the 9 activity patterns used in the averaging process is determined
by the product of two factors. The first is the relative frequency of its day type, i.e., 0.18 for
summer weekdays, 0.54 for non-summer weekdays, and 0.28 for weekends.
The second factor in the weighting for the selected activity pattern is determined by
simulating a sequence of category-types as a one-stage Markov chain process using the
transition probabilities. The category for the first day is selected according to the relative
frequencies of each category. The category for the second day is selected according to the
category-to-category transition probabilities for the category selected for the first day. The
category for the third day is selected according to the transition probabilities for the category
selected for the second day. This is repeated for all days in the day type (65 for summer
weekdays, 195 for non-summer weekdays, 104 for weekends), producing a sequence of daily
F-ll
-------
categories. The relative frequency of the category-type in the sequence associated with the
selected activity pattern is the second factor in the weighting.
PROPOSED ALGORITHM STEPS
The proposed algorithm is summarized in Figure 1. Each step is explained in this section.
Data Preparation
Step 1: Each daily activity pattern in the CHAD data base is summarized by the total minutes
in each of five micro-environments: indoors - residence; indoors - other building; outdoors -
near road; outdoors - away from road; in vehicle. These five numbers are assumed to
represent the most important features of the activity pattern for their exposure impact.
Step 2: All CHAD activity patterns for a given day-type and demographic group are
subjected to cluster analysis, resulting in 2 or 3 cluster categories. Each daily activity pattern
is tagged with a cluster category.
Step 3: For each day-type and demographic group, the relative frequency of each day-type in
the CHAD data base is determined.
Step 4: All CHAD activity patterns for a given day-type and demographic group that are
consecutive days for a single individual, are analyzed to determine the category-to-category
transition frequencies in the CHAD data base. These transition frequencies are used to
calculate category-to-category transition probabilities.
For example, if there are 2 categories, A and B, then
PAA = the probability that a type A pattern is followed by a type A pattern,
PAB = the probability that a type A pattern is followed by a type B pattern (PAB = 1 - PAA),
PBB = the probability that a type B pattern is followed by a type B pattern, and
PBA = the probability that a type B pattern is followed by a type A pattern (PBA = 1 - PBB).
Activity Pattern Selection
For each day-type and demographic group in each census tract
Step 5: One activity pattern is randomly selected from each cluster category group (i.e., 2 to
3 activity patterns)
Creating Weights for Day-type Averaging
For each day-type and demographic group in each census tract
F-12
-------
Step 6: A cluster category is selected for the first day of the day-type sequence, according to
the relative frequency of the cluster category days in the CHAD data set.
Step 7: A cluster category is selected for each subsequent day in the day-type sequence day
by day using the category-to-category transition probabilities.
Step 8: The relative frequency of each cluster category in the day-type sequence is
determined.
Step 9: The activity patterns selected for each cluster category (Step 5) are averaged together
using the cluster category frequencies (Step 8) as weights, to create a day-type average
activity pattern.
Creating Annual Average Activity Patterns
For each demographic group in each census tract
Step 10: The day-type average activity patterns are averaged together using the relative
frequency of day-types as weights, to create an annual average activity pattern.
Creating Replicates
For each demographic group in each census tract
Step 11: Steps 5 through 10 are repeated 29 times to create 30 annual average activity
patterns.
EVALUATING THE ALGORITHM
The purpose of this study is to evaluate how well the proposed one-stage Markov chain
algorithm can reproduce observed multi-day activity patterns with respect to demographic
group means and inter-individual variability, while using one-day selection.
In order to accomplish this we propose to apply the algorithm to observed multi-day activity
patterns provided by the WAM, and compare the means and variances of the predicted multi-
day patterns with the observed patterns.
Current APEX Algorithm
Because the algorithm is being considered for incorporation into APEX, we would like the
evaluation to be consistent with the approach taken in APEX for selection of activity patterns
F-13
-------
for creating multi-day sequences. The APEX approach for creating multi-day activity
sequences is as follows.
Stepl: A profile for a simulated individual is generated by selection of gender, race (not
implemented?), age group, and home sector from a given set of distributions consistent with
the population of the study area.
Step 2: A specific age within the age group is selected from a uniform distribution.
Step 3: The employment status is simulated as a function of the age.
Step 4: For each simulated day, the user defines an initial pool of possible diary days based
on a user-specified function of the day type (e.g., weekday/weekend) and temperature.
Step 5: The pool is further restricted to match the target gender and employment status
exactly and the age within 2A years for some parameter A. The diary days within the pool
are assigned a weight of 1 if the age is within A years of the target age and a weight of w
(user-defined parameter) if the age difference is between A and 2A years. For each simulated
day, the probability of selecting a given diary day is equal to the age weight divided by the
total of the age weights for all diary days in the pool for that day.
Approach to Incorporation of Day-to-Day Dependence into APEX Algorithm
If we were going to incorporate day-to-day dependence of activity patterns into the APEX
model, we would propose preparing the data with cluster analysis and transition probabilities
as described in Steps 1-4 for the proposed HAPEM 5 algorithm, with the following
modifications.
• For Step 2 the activity patterns would be divided into groups based on day-type
(weekday, weekend), temperature, gender, employment status, and age, with
cluster analysis applied to each group. However, because the day-to-day
transitions in the APEX activity selection algorithm can cross temperature bins,
we would propose to use broad temperature bins for the clustering and transition
probability calculations so that the cluster definitions would be fairly uniform
across temperature bins. Thus we would probably define the bins according to
season (e.g., summer, non-summer).
• In contrast to HAPEM, the sequence of activity patterns may be important in
APEX. Therefore, for Step 4 transition probabilities would be specified for
transitions between days with the same day-type and season, as in HAPEM, and
also between days with different day-types and/or seasons. For example,
transition probabilities would be specified for transitions between summer
weekdays of each category and summer weekends of each category.
F-14
-------
Another issue for dividing the CHAD activity records for the purposes of clustering and
calculating transition probabilities is that the diary pools specified for the APEX activity
selection algorithm use varying and overlapping age ranges. One way to address this problem
would be to simply not include consideration of age in the clustering process, under the
assumption that cluster categories are similar across age groups, even if the frequency of
each cluster category varies by age group. This assumption could be tested by examination of
the cluster categories stratified by age group that were developed for HAPEM5. If the
assumption is found to be valid, then the cluster categories could be pre-determined for input
to APEX, while the transition probabilities could be calculated within APEX during the
simulation for each age range specified for dairy pools.
If the assumption is found to be invalid, then an alternative approach could be implemented
that would create overlapping age groups for purposes of clustering as follows. APEX age
group ranges and age window percentages would be constrained to some maximum values.
Then a set of overlapping age ranges that would be at least as large as the largest possible
dairy pool age ranges would be defined for the purposes of cluster analysis and transition
probability calculation. The resulting sets of cluster categories and transition probabilities
would be pre-determined for input into APEX and the appropriate set used by APEX for each
diary pool used during the simulation.
The actual activity pattern sequence selection would be implemented as follows. The activity
pattern for first day in the year would be selected exactly as is currently done in APEX, as
described above. For the selecting the second day's activity pattern, each age weight would
be multiplied by the transition probability PAB where A is the cluster for the first day's
activity pattern and B is the cluster for a given activity pattern in the available pool of diary
days for day 2. (Note that day 2 may be a different day-type and/or season than day 1.) The
probability of selecting a given diary day on day 2 is equal to the age weight times PAB
divided by the total of the products of age weight and PAB for all diary days in the pool for
day 2. Similarly, for the transitions from day 2 to day 3, day 3 to day 4, etc.
Testing the Approach with the Multi-day Data set
We tested this approach using the available multi-day data set. For purposes of clustering we
characterized the activity pattern records according to time spent in each of 5
microenvironments: indoors-home, indoors-school, indoors-other, outdoors (aggregate of the
3 outdoor microenvironments), and in-transit.
For purposes of defining diary pools and for clustering and calculating transition probabilities
we divided the activity pattern records by day type (i.e., weekday, weekend), season (i.e.,
summer or ozone season, non-summer or non-ozone season), age (6-10 and 11-12), and
gender. Since all the subjects are 6-12 years of age and all are presumably unemployed, we
need not account for differences in employment status. For each day type, season, age, and
gender, we found that the activity patterns appeared to group in three clusters.
F-15
-------
In this case, we simulated week-long sequences (Wednesday through Tuesday) for each of
100 people in each age/gender group for each season, using the transition probabilities. To
evaluate the algorithm we calculated the following statistics for the predicted multi-day
activity patterns for comparison with the actual multi-day diary data.
• For each age/gender group for each season, the average time in each
microenvironment
• For each age/gender group, season, and microenvironment, the average of the
within-person variance across all simulated persons (We defined the within-
person variance as the variance of the total time per day spent in the
microenvironment across the week.)
• For each age/gender group, season, and microenvironment, the variance across
persons of the mean time spent in that microenvironment
In each case we compared the predicted statistic for the stratum to the statistic for the
corresponding stratum in the actual diary data.1
We also calculated the mean normalized bias for the statistic, which is a common
performance measure used in dispersion model performance and which is calculated as
follows.
*rr, T A o 100^ (predicted - observed} _.
NBIAS = > — %
N i observed
RESULTS
Comparisons of simulated and observed data for time in each of the 5 microenvironments are
presented in Tables 1-3 and Figures 2-5.
Average Time in Microenvironment
Table 1 and Figure 2 show the comparisons for the average time spent in each of the 5
microenvironments for each age/gender group and season. Figure 3 shows the comparison
for all the microenvironments except indoor, home in order to highlight the lower values.
1 For the diary data, because the number of days per person varies, the average of the within-person variances
was calculated as a weighted average, where the weight is the degrees of freedom, i.e., one less than the number
of days simulated. Similarly, the variance across persons of the mean time was appropriately adjusted for the
different degrees of freedom using analysis of variance.
F-16
-------
Table 1 and the figures show that the predicted time-in-microenvironment averages match
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%. Fourteen predictions have positive bias and 23 have negative bias. A Wilcoxon
signed rank test that the median bias across the 40 combinations = 0 % was not significant
(p-value = 0.40) supporting the conclusion of no overall bias.
Variance Across Persons
Table 2 and Figure 4 show the comparisons for the variance across persons for the average
time spent in each microenvironment. In this case the bias ranges from -40% to +120% for
any microenvironment/age/gender/season. Sixty-five percent of the predicted variances have
bias between -22% and +24%. The mean normalized bias across any microenvironment
ranges from -10% to +28%. Eighteen predictions have positive bias and 20 have negative
bias. Figure 4 suggests a reasonably good match of predicted to observed variance in spite of
2 or 3 outliers. A Wilcoxon signed rank test that the median bias across the 40 combinations
= 0 % was not significant (p-value = 0.93) supporting the conclusion of no overall bias.
Within-Person Variance for Persons
Table 3 and Figure 5 show the comparisons for the within-person variance for time spent in
each microenvironment. In this case the bias ranges from -47% to +150% for any
microenvironment/age/gender/season. Seventy percent of the predicted variances have bias
between -25% and +30%. The mean normalized bias across any microenvironment ranges
from -11% to +47%. Twenty-eight predictions have positive bias and 12 have negative bias,
suggesting some tendency for overprediction of this variance measure. And indeed a
Wilcoxon signed rank test that the median bias across the 40 combinations = 0 % was very
significant (p-value = 0.01) showing that the within-person variance was significantly
overpredicted. Still, Figure 4 suggests a reasonably good match of predicted to observed
variance in most cases, with a few overpredicting outliers at the higher end of the
distribution. So although the positive bias is significant in a statistical sense (i.e., the variance
is more likely to be overpredicted than underpredicted), it is not clear whether the bias is
large enough to be important.
CONCLUSIONS
The proposed algorithm appears to be able to replicate the observed data reasonably well,
although the within-person variance is somewhat overpredicted.
It would be informative to compare this algorithm with the earlier alternative approaches in
order to gain perspective on the degree of improvement, if any, afforded by this approach.
F-17
-------
Two earlier approaches were:
1. Select a single activity pattern for each day-type/season combination from the
appropriate set, and use that pattern for every day in the multi-day sequence that
corresponds to that day-type and season.
2. Re-select an activity pattern for each day in the multi-day sequence from the
appropriate set for the corresponding day-type and season.
Goodness-of-fit statistics could be developed to compare the three approaches and find which
model best fits the data for a given stratum.
F-18
-------
Table 1. Average time spent in each microenvironment: comparison of predicted and observed.
Microenvironment
Indoor, home
Indoor, school
Indoor, other
Demographic
Group
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
Season
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Observed
(hours/day)
15.5
15.8
15.7
15.8
16.2
16.5
16.0
16.2
0.7
2.3
0.8
2.2
0.7
2.1
0.6
2.4
2.9
2.4
2.2
1.9
2.2
2.2
2.3
Predicted
(hours/day)
16.5
15.5
15.2
16.4
15.3
16.5
15.6
16.1
0.7
2.5
0.5
2.2
0.7
2.4
0.9
2.7
2.4
2.7
2.7
1.8
1.6
2.1
2.2
Normalized
Bias
6%
-2%
-3%
4%
-5%
0%
-3%
-1%
-1%
-9%
7%
-34%
0%
6%
13%
38%
11%
4%
-14%
13%
21%
-3%
-25%
-2%
-5%
F-19
-------
Outdoors
In-vehicle
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
1.9
3.7
2.5
4.1
3.1
3.7
2.3
3.9
2.6
1.1
1.0
1.1
1.0
1.2
0.9
1.1
0.9
2.0
3.5
2.5
4.3
2.7
5.2
2.1
4.3
2.4
0.9
0.9
1.3
0.9
1.1
0.8
1.0
0.8
4%
-2%
-6%
0%
4%
-12%
41%
-5%
9%
-7%
3%
-20%
-13%
13%
-16%
-12%
-15%
-5%
-7%
-9%
F-20
-------
Table 2. Variance across persons for time spent in each microenvironment: comparison of
predicted and observed.
Microenvironment
Indoor, home
Indoor, school
Indoor, other
Demographic
Group
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Season
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Observed
(hours/day)2
70
67
54
35
56
42
57
39
6.0
9.5
5.6
5.3
4.9
5.4
5.6
9.2
46
44
34
23
21
28
Predicted
(hours/day)2
42
60
49
30
47
38
63
42
5.2
5.9
3.8
8.2
5.5
5.3
6.0
11
32
46.
33
16
18
22
Normalized
Bias
-40%
-9%
-9%
-12%
-17%
-10%
12%
8%
-10%
-13%
-38%
-32%
53%
11%
-1%
6%
23%
1%
-30%
6%
-4%
-27%
-15%
-22%
F-21
-------
Outdoors
In-vehicle
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
33
30
17
9.3
17
8.3
22
9.0
13
10
1.9
1.8
2.5
1.5
3.5
2.8
3.2
1.3
31
30
23
6.8
18
7.6
22
9.1
29
11
2.3
1.6
4.7
1.6
4.7
2.0
5.4
1.7
-6%
0%
-12%
37%
-27%
3%
-8%
0%
1%
120%
8%
17%
24%
-11%
93%
9%
34%
-28%
69%
35%
28%
F-22
-------
Table 3. Average within person variance for time spent in each microenvironment: comparison of
predicted and observed.
Microenvironment
Indoor, home
Indoor, school
Indoor, other
Demographic
Group
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Season
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Observed
(hours/day)2
20
18
17
15
22
22
21
17
2.3
7.3
2.0
6.7
1.7
7.4
1.4
7.3
14
14
12
10
10
14
Predicted
(hours/day)2
29
23
30
24
42
25
24
24
2.4
6.4
1.5
5.8
2.1
7.6
2.9
7.8
14
18
17
13
10
15
Normalized
Bias
49%
25%
75%
64%
93%
13%
16%
38%
47%
5%
-12%
-25%
-14%
29%
3%
101%
6%
12%
-4%
30%
42%
26%
1%
7%
F-23
-------
Outdoors
In-vehicle
Boys, 11-12
MEAN
Girls, 6-10
Boys, 8-10
Girls, 11-12
Boys, 11-12
MEAN
Girls, 6-10
Boys, 6-10
Girls, 11-12
Boys, 11-12
MEAN
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
Summer
Not Summer
11
12
8.4
3.4
6.7
3.4
10
4.0
9.2
4.3
1.0
0.90
1.1
0.81
1.3
1.3
2.4
0.85
14
13
9.5
3.2
9.5
4.4
25
4.5
7.4
3.7
0.90
0.48
1.4
0.71
1.3
1.1
1.6
0.85
26%
7%
17%
13%
-3%
42%
28%
150%
11%
-20%
-15%
26%
-13%
-47%
31%
-12%
4%
-16%
-34%
1%
-11%
F-24
-------
Consolidated Human Activity Database - CHAD (CHAD)
Winter Weekday
Pattern Group
Summer Weekday
Pattern Group
Weekend
Pattern Group
Cluster 1
Pattern
Group
Cluster 1
Pattern
Cluster 2
Pattern
Cluster 3
Pattern
Group
i
r
Cluster 3
Pattern
ransition
Analysis
Transition
Probabilities
Average Winter
Weekday
0.54
Weights
Markov Selection
Individual Annual Average Activity Pattern
Figure 1. Flow diagram of proposed algorithm for creating annual average activity patterns for HAPEM5.
F-25
-------
on
re
•o 15
3
0
•c 10
=5 5
£
0.
n
/^
JJpf
/^
^S^
-X*^
/^
^s*^
**r
t-^^
^^^
s^f
• Indoor, home
• Indoor, school
inrlnnr nthpr
Outdoor
x In vehicle
0 5 10 15 20
Observed (hours/day)
Figure 2. Comparison of predicted and observed average time in each of 5 microenvironments for age/gender groups and
seasons.
c
">
•S 4
"55
3 3
o °
£_
•C 9
1
"O -1
1 1
0.
n
^^
/^
^
^
^4&
• Indoor, school
indoor, other
Outdoor
x In vehicle
012345
Observed (hours/day)
Figure 3. Comparison of predicted and observed average time in each of 4 microenvironments for age/gender groups and
seasons.
F-26
-------
80
7f)
fin
•o «;o
0) ou
+-
- 40
3 ^U
•i so
w ou
20
10
0
c
/
*/•
/
-r * ,
/'>
s
&
* girls, 6-10, summer
• girls, 6-10, winter
boys , 6-10, summer
X boys, 6-10, winter
X girls, 11-12, summer
• girls, 11-12, winter
+ boys, 11-12, summer
- boys, 1 1-12, winter
) 20 40 60 80
Observed
Figure 4. Comparison of predicted and observed variance across persons for time spent in each of 5 microenvironments for
age/gender groups and seasons.
30 • .,
"I * /
°5 X • /
- \+y
°o /
•o Z{J
$
TO -ic
D ID
E
c/) in
R
\/
^
#
f
• girls, 6-10, summer
• girls, 6-10, winter
boys , 6-10, summer
boys, 6-10, winter
X girls, 11-12, summer
• girls, 11-12, winter
+ boys, 11-12, summer
• boys, 11-12, winter
4— 1 1 1
0 10 20 30
Observed
Figure 5. Comparison of predicted and observed the average within-person variance for time spent in each of 5
microenvironments by age/gender groups and seasons.
F-27
-------
Appendix G - Exposure Risk Results for Asthmatics and
Asthmatic Children
This Appendix provides supplemental exposure and risk characterization results for two
subpopulations, all asthmatics and asthmatic children. The data are presented in series of
summary tables and figures across each of the scenarios investigated (i.e. with modeled air
quality as is and simulating just meeting the current standard), with and without modeled indoor
sources (i.e., gas stoves), for each of the potential health effect benchmark levels (i.e., 200, 250,
300 ppb 1-hour), and across three years of modeled air quality (i.e., 2001 to 2003). Repeated
exposures are presented only for the lowest potential health effect benchmark level (i.e., 200 ppb
1-hour).
G-l
-------
G. 1 All Asthmatics
Table 1. Estimated number of asthmatics in Philadelphia County exposed at or above potential health
effect benchmark levels (1 to 6 times per year), using modeled air quality (as is) and with just meeting the
current standard (std), and with and without indoor sources.
Year (AQ)
2001 (as
is)
2001 (std)
2002 (as
is)
2002 (std)
2003 (as
is)
2003 (std)
Indoor
Source
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Level
(ppb)
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
Persons with Number of Repeated Exposures
1
49796
4867
1388
10544
2584
1013
128147
49632
16805
90211
40466
15100
47652
4430
1240
9505
2276
975
133524
53367
18828
98849
43972
16693
52639
14407
6568
26120
11142
5605
132640
73387
39283
109726
65437
35948
2
19544
1414
404
2577
765
344
96119
18322
4480
51600
14362
3590
17720
1173
393
2411
778
304
102861
20737
5220
60056
16367
4389
22084
5040
1892
10007
3927
1627
1 02034
38505
16213
73489
33096
14502
3
8959
658
157
1230
413
177
70079
8523
1828
31720
6155
1595
8056
530
147
1240
332
137
77512
9855
2324
36913
7370
1950
11950
2599
887
5857
2040
778
76909
22953
9280
51133
18948
8474
4
4516
381
108
795
295
98
50253
4808
1219
19805
3225
1003
4170
274
88
706
185
59
57152
5784
1447
23238
4066
1131
7441
1577
512
3783
1261
462
58857
15416
6175
36551
12710
5654
5
2666
265
59
520
186
39
35965
3095
866
12899
2141
755
2662
166
69
401
117
49
42473
3489
925
15850
2680
766
4863
935
335
2609
777
285
44719
11101
4374
27509
8964
4098
6
1732
157
39
422
118
29
26167
2152
638
8938
1414
569
1765
127
49
323
88
49
31800
2623
648
10875
1734
510
3457
650
245
1842
550
206
34990
8499
3259
21181
6862
2935
G-2
-------
Table 2. Estimated percent of asthmatics in Philadelphia County exposed at or above potential health
effect benchmark levels (1 to 6 times per year), using modeled air quality (as is) and with just meeting the
current standard (std), and with and without indoor sources.
Year (AQ)
2001 (as
is)
2001 (std)
2002 (as
is)
2002 (std)
2003 (as
is)
2003 (std)
Indoor
Source
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Level
(ppb)
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
Percent (%) of Persons With Repeated Exposures
1
31
3
1
6
2
1
79
31
10
55
25
9
29
3
1
6
1
1
82
33
12
61
27
10
32
9
4
16
7
3
81
45
24
67
40
22
2
12
1
0
2
0
0
59
11
3
32
9
2
11
1
0
1
0
0
63
13
3
37
10
3
14
3
1
6
2
1
63
24
10
45
20
9
3
6
0
0
1
0
0
43
5
1
20
4
1
5
0
0
1
0
0
48
6
1
23
5
1
7
2
1
4
1
0
47
14
6
31
12
5
4
3
0
0
0
0
0
31
3
1
12
2
1
3
0
0
0
0
0
35
4
1
14
2
1
5
1
0
2
1
0
36
9
4
22
8
3
5
2
0
0
0
0
0
22
2
1
8
1
0
2
0
0
0
0
0
26
2
1
10
2
0
3
1
0
2
0
0
27
7
3
17
6
3
6
1
0
0
0
0
0
16
1
0
5
1
0
1
0
0
0
0
0
20
2
0
7
1
0
2
0
0
1
0
0
21
5
2
13
4
2
G-3
-------
•a t/i
0) —
V) V)
o ro
200
250
2003 AQ (as is) - with indoor soucrces
2002 AQ (as is) - with indoor soucrces
2001 AQ (as is) - with indoor soucrces
Sim ulated Year - Scenario
Potential Health Effect Benchmark Level (ppb)
Figure 1. Estimated percent of all asthmatics in Philadelphia County with at least on NO2 exposure at or
above potential health effect benchmark level, using 2001-2003 modeled air quality (as is), with modeled
indoor sources.
300
2003 AQ (as is) - no indoor soucrces
2002 AQ (as is) - no indoor soucrces
2001 AQ (as is) - no indoor soucrces
Sim ulated Year - Scenario
Potential Health Effect Benchmark Level (ppb)
Figure 2. Estimated percent of all asthmatics in Philadelphia County with at least on NO2 exposure at or
above potential health effect benchmark level, using 2001-2003 modeled air quality (as is), with no indoor
sources.
G-4
-------
100
5 5
200
Potential Health Effect Benchmark Level (ppb)
2003 AQ (std) - with indoor soucrces
2002 AQ (std) - with indoor soucrces
2001 AQ (std) - with indoor soucrces
Simulated Year - Scenario
Figure 3. Estimated percent of all asthmatics in Philadelphia County with at least on NO2 exposure at or
above potential health effect benchmark level, using 2001-2003 modeled air quality just meeting the
current standard (std), with modeled indoor sources.
~S B
ggS
Q. O C
u"K
|l|
£ 0) O
C/) >- Q)
< C £
1*1
250
Potential Health Effect Benchmark Level (ppb)
2003 AQ (std) - no indoor soucrces
2002 AQ (std) - no indoor soucrces
2001 AQ (std) - no indoor soucrces
Simulated Year - Scenario
Figure 4. Estimated percent of all asthmatics in Philadelphia County with at least on NO2 exposure at or
above potential health effect benchmark level, using 2001-2003 modeled air quality just meeting the
current standard (std), with no indoor sources.
G-5
-------
35
2003 (as is)
Estimated Number of
Repeated Exposures to
200 ppb 1-hour in a Year
2002 (as is)
Simulated Year-Scenario
2001 (as is)
Figure 5. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures at or
above 200 ppb 1-hr, using 2001-2003 modeled air quality (as is), with modeled indoor sources.
2003 (as is)
Estimated Number of
Repeated Exposures to
200 ppb 1-hour in a Year
2002 (as is)
Simulated Year-Scenario
2001 (as is)
Figure 6. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures at or
above 200 ppb 1-hr, using 2001-2003 modeled air quality (as is), without indoor sources.
G-6
-------
2003 (std)
Estimated Number of
Repeated Exposures to
200 ppb 1 -hour in a Year
2002 (std)
Simulated Year - Scenario
2001 (std)
Figure 7. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures at or
above 200 ppb 1-hour, using 2001-2003 modeled air quality just meeting the current standard (std), with
modeled indoor sources.
2003 (std)
Estimated Number of
Repeated Exposures to
200 ppb 1 -hour in a Year
2002 (std)
Simulated Year - Scenario
2001 (std)
Figure 8. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures at or
above 200 ppb 1-hour, using 2001-2003 modeled air quality just meeting the current standard (std), with
no indoor sources.
G-7
-------
G-2 Asthmatic Children
Table 3. Estimated number of asthmatic children in Philadelphia County exposed at or above potential
health effect benchmark levels (1 to 6 times per year), using modeled air quality (as is) and with just
meeting the current standard (std), and with and without indoor sources.
Year (AQ)
2001 (as
is)
2001 (std)
2002 (as
is)
2002 (std)
2003 (as
is)
2003 (std)
Indoor
Source
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Level
(ppb)
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
Persons With
1
11351
709
128
2329
393
97
36656
13543
3909
27511
11282
3440
10636
692
70
1771
158
30
38834
14855
4203
30548
12487
3736
12525
3541
1545
6724
2784
1368
37931
20044
10562
32066
18770
9547
2
3649
167
49
401
97
29
26353
4530
768
16067
3735
638
3338
139
10
315
49
10
28678
4887
947
18685
3775
670
4693
1240
423
2526
1032
355
28305
9893
4100
21662
8897
3704
Number of Repeated Exposures
3
1418
68
10
147
39
10
18272
1877
236
9890
1413
187
1439
49
0
158
20
0
20840
1978
336
11394
1288
276
2736
678
237
1515
531
208
20344
6016
2381
14938
4974
2223
4
709
49
10
98
20
10
12133
926
187
6094
500
128
800
30
0
79
10
0
14308
1086
228
7063
738
158
1712
423
138
984
335
119
15230
4088
1643
10326
3371
1496
5
424
20
0
58
0
0
8271
533
128
3757
333
109
494
0
0
10
0
0
10063
652
119
4336
493
99
1100
247
89
708
188
69
11013
2888
1211
7647
2388
1072
6
267
10
0
58
0
0
5783
295
88
2430
197
79
346
0
0
0
0
0
6996
514
79
2782
365
39
797
178
39
492
128
39
8483
2253
906
6018
1859
817
G-8
-------
Table 4. Estimated percent of asthmatic children in Philadelphia County exposed at or above potential
health effect benchmark levels (1 to 6 times per year), using modeled air quality (as is) and with just
meeting the current standard (std), and with and without indoor sources.
Year (AQ)
2001 (as
is)
2001 (std)
2002 (as
is)
2002 (std)
2003 (as
is)
2003 (std)
Indoor
Source
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Level
(ppb)
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
200
250
300
Percent
1
23
1
0
5
1
0
75
28
8
57
23
7
22
1
0
4
0
0
81
31
9
64
26
8
26
7
3
14
6
3
79
42
22
67
39
20
(%)
2
8
0
0
1
0
0
54
9
2
33
8
1
7
0
0
1
0
0
60
10
2
39
8
1
10
3
1
5
2
1
59
21
9
45
19
8
of Persons
3
3
0
0
0
0
0
38
4
0
20
3
0
3
0
0
0
0
0
43
4
1
24
3
1
6
1
0
3
1
0
43
13
5
31
10
5
With Repeated
4
1
0
0
0
0
0
25
2
0
13
1
0
2
0
0
0
0
0
30
2
0
15
2
0
4
1
0
2
1
0
32
9
3
22
7
3
Exposures
5
1
0
0
0
0
0
17
1
0
8
1
0
1
0
0
0
0
0
21
1
0
9
1
0
2
1
0
1
0
0
23
6
3
16
5
2
6
1
0
0
0
0
0
12
1
0
5
0
0
1
0
0
0
0
0
15
1
0
6
1
0
2
0
0
1
0
0
18
5
2
13
4
2
G-9
-------
30-r-^"^ M
S
18
s —
a -2 25^
X (/)
UJ _TO_
£ «" 1
0) o
l! 4"
o °
.- °
ro en
E « ,c
£ o 15-
< c
i! 1"
£ a
"• o
T3 0)
i« 5-
E
In
UJ
^—-^
S
/
. /
200
250
Potential Health Effect Benchmark Level (ppb)
2003 f\Q (as is) - with indoor soucrces
2002 I\Q (as is) - with indoor soucrces
2001 f\Q (as is) - with indoor soucrces
Sim ulated Year - Scenario
Figure 9. Estimated percent of asthmatic children in Philadelphia County with at least on NO2 exposure
at or above potential health effect benchmark level, using 2001-2003 modeled air quality (as is), with
modeled indoor sources.
300
2003 AQ (as is) - no indoor soucrces
2002 I\Q (as is) - no indoor soucrces
2001 f\Q (as is) - no indoor soucrces
Sim ulated Year - Scenario
Potential Health Effect Benchmark Level (ppb)
Figure 10. Estimated percent of asthmatic children in Philadelphia County with at least on NO2 exposure
at or above potential health effect benchmark level, using 2001-2003 modeled air quality (as is), with no
indoor sources.
G-10
-------
100
200
Potential Health Effect Benchmark Level (ppb)
2003 AQ (std) - with indoor soucrces
2002 AQ (std) - with indoor soucrces
2001 AQ (std) - with indoor soucrces
Simulated Year - Scenario
Figure 11. Estimated percent of asthmatic children in Philadelphia County with at least on NO2 exposure
at or above potential health effect benchmark level, using 2001-2003 modeled air quality just meeting the
current standard (std), with modeled indoor sources.
2*S
I ft o
: >- ot
: .E £
i — D)
: S =
250
Potential Health Effect Benchmark Level (ppb)
2003 AQ (std) - no indoor soucrces
2002 AQ (std) - no indoor soucrces
2001 AQ (std) - no indoor soucrces
Simulated Year - Scenario
Figure 12. Estimated percent of asthmatic children in Philadelphia County with at least on NO2 exposure
at or above potential health effect benchmark level, using 2001-2003 modeled air quality just meeting the
current standard (std), with no indoor sources.
G-ll
-------
2003 (as is)
Estimated Number of
Repeated Exposures to
200 ppb 1-hour in a Year
2002 (as is)
Simulated Year-Scenario
2001 (as is)
Figure 13. Estimated percent of asthmatic children in Philadelphia County with repeated NO2 exposures
at or above 200 ppb 1-hr, using 2001-2003 modeled air quality (as is), with modeled indoor sources.
2003 (as is)
Estimated Number of
Repeated Exposures to
200 ppb 1-hour in a Year
2002 (as is)
Simulated Year-Scenario
2001 (as is)
Figure 14. Estimated percent of asthmatic children in Philadelphia County with repeated NO2 exposures
at or above 200 ppb 1-hr, using 2001-2003 modeled air quality (as is), with no indoor sources.
G-12
-------
2003 (std)
Estimated Number of
Repeated Exposures to
200 ppb 1 -hour in a Year
2002 (std)
Simulated Year - Scenario
2001 (std)
Figure 15. Estimated percent of asthmatic children in Philadelphia County with repeated NO2 exposures
at or above 200 ppb 1-hr, using 2001-2003 modeled air quality meeting the current standard (std), with
modeled indoor sources.
2003 (std)
Estimated Number of
Repeated Exposures to
200 ppb 1 -hour in a Year
2002 (std)
Simulated Year - Scenario
2001 (std)
Figure 16. Estimated percent of asthmatic children in Philadelphia County with repeated NO2 exposures
at or above 200 ppb 1-hr, using 2001-2003 modeled air quality meeting the current standard (std), with no
indoor sources.
G-13
-------
United States Office of Air Quality Planning and Standards EPA-452/P-08-002
Environmental Protection Air Quality Strategies and Standards Division April 2008
Agency Research Triangle Park, NC
------- |