ESTIMATION OF OZONE EXPOSURES
EXPERIENCED BY URBAN RESIDENTS USING A
PROBABILISTIC VERSION OF NEM
AND 1990 POPULATION DATA
by
Ted Johnson, Jim Capel, and Mike McCoy
International Technology Air Quality Services
South Square Corporate Centre One
3710 University Drive, Suite 201
Durham, North Carolina 27707-6208
Contract No. 68-DO-0062
JTN 830013-26
Thomas McCurdy, Project Officer
Richard B. Atherton, Project Manager
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF AIR QUALITY PLANNING AND STANDARDS
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
April 1996
-------
CONTENTS
Figures v
Tables vi
Acknowledgment xi
1. Introduction 1
2. Overview of the Methodology 5
Define study area, population-of-interest,
subdivisions of study area, and exposure period 5
Divide the population-of-interest into an exhaustive
set of cohorts 7
Develop an exposure event sequence for each cohort for
the exposure period 9
Estimate the pollutant concentration and ventilation
rate associated with each exposure event 14
Extrapolate the cohort exposures to the population-of-interest
and to individual sensitive groups 24
3. The Mass-Balance Model 32
Theoretical basis and assumptions 33
Simulation of microenvironmentai ozone concentrations 39
Air exchange rate distributions 42
Window status algorithm 45
4. Preparation of Air Quality Data 49
Selection of representative data sets 49
Treatment of missing values and descriptive statistics 68
-------
CONTENTS (continued)
5. Adjustment of Ozone Data to Simulate Compliance with Alternative
Air Quality Standards 78
Specification of AQI and estimation of baseline AQI values 79
Estimation of AQI's under attainment conditions 86
Adjustment of one-hour ozone data sets 90
Application of the AQAP's to Philadelphia 93
Special Adjustment Procedures Applied in Selected
Attainment Scenarios 101
6. Ozone Exposure Estimates for Nine Urban Areas 104
Regulatory scenarios 104
Formats of the exposure summary tables 105
Results of analyses 107
7. Initial Efforts to Validate the Exposure Model 119
The HAS data 119
The special version of pNEM/O3 120
Processing of HAS data 121
Comparison of measured and estimated exposures
by person-hour 123
Sensitivity of exposure estimates to ozone decay rate 126
8. Principal Limitations of the pNEM/O3 Methodology 130
Time/activity patterns 131
Equivalent ventilation rates 132
The air quality adjustment procedures 133
The mass balance model 134
References 136
Appendices
A. Attainment of Ozone Data to Simulate NAAQS Attainment A-1
B. Adjustment of Ozone Data to Simulate Sexex 8h NAAQS B-1
C. Sample Output of pNEM/O3 C-1
IV
-------
FIGURES
Number
1 Page from the activity diary used in the Cincinnati study 11
2 Estimated percentage of residents in Houston study area who
experience one-hour daily maximum ozone exposures above
0.12 ppm according to regulatory scenario 116
3 Estimated percentage of residents in each study area who
experience one-hour daily maximum ozone exposures above
0.12 ppm according to regulatory scenario 117
4 Estimated percentage of residents in Houston study area who
experience eight-hour daily maximum ozone exposures above
0.08 ppm according to regulatory scenario 118
-------
TABLES
Number
1 Characteristics of Study Areas 7
2 Demographic Groups Defined for pNEM/03 Analysis 9
3 Microenvironments Defined for pNEM/03 Analysis 13
4 Parameters Associated with Algorithms Used to Estimate
Ozone Concentrations in Microenvironments 17
5 Parameter Values of Lognormal Distributions Used to
Characterize Equivalent Ventilation Rate 21
6 Algorithm for Determining Upper Limit for EVR 22
7 Parameter Values for Algorithm Used to Determine Limits for
Equivalent Ventilation Rate 23
8 Estimated Fraction of Houston Workers Within Each
Home District That Commute to Each Work District 29
9 Means, Standard Deviations, and Confidence Intervals
for Estimates of k^A/V) Provided by Weschler 39
10 Distributions of Air Exchange Rate Values Used in the
pNEM/03 Mass Balance Model 43
11 Probability of Window Status for Day by Air Conditioning
System and Temperature Range 47
12 Probability of Windows Being Open by Clock Hour, Temperature
Range, and Window Status of Preceding Hour (PH) for Residences
With Central Air Conditioning 47
VI
-------
TABLES (continued)
Number
13 Probability of Windows Being Open by Clock Hour, Temperature
Range, and Window Status of Preceding Hour (PH) for Residences
With Window Air Conditioning Units 48
14 Probability of Windows Being Open by Clock Hour, Temperature
Range, and Window Status of Preceding Hour (PH) for Residences
With No Air Conditioning System 48
15 Characteristics of Ozone Study Areas and Monitoring Sites 51
16 Descriptive Statistics for 1991 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Chicago Study Area 52
17 Descriptive Statistics for 1990 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Denver Study Area 54
18 Descriptive Statistics for 1990 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Houston Study Area 55
19 Descriptive Statistics for 1991 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Los Angeles Study Area 57
20 Descriptive Statistics for 1991 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Miami Study Area 59
21 Descriptive Statistics for 1991 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the New York Study Area 60
22 Descriptive Statistics for 1991 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Philadelphia Study Area 62
VII
-------
TABLES (continued)
Number
23 Descriptive Statistics for 1990 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the St. Louis Study Area 64
24 Descriptive Statistics for 1991 Data Sets Containing
Hourly-Average Ozone Concentrations Obtained From
Selected Monitoring Sites in the Washington Study Area 66
25 Descriptive Statistics for 1991 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Chicago Study Area 69
26 Descriptive Statistics for 1990 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Denver Study Area 70
27 Descriptive Statistics for 1990 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Houston Study Area 71
28 Descriptive Statistics for 1991 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Los Angeles Study Area 72
29 Descriptive Statistics for 1991 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Miami Study Area 73
30 Descriptive Statistics for 1991 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the New York Study Area 74
31 Descriptive Statistics for 1991 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Philadelphia Study Area 75
VIII
-------
TABLES (continued)
Number
32 Descriptive Statistics for 1990 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the St. Louis Study Area 76
33 Descriptive Statistics for 1991 Data Sets Containing
Eight-Hour Ozone Concentrations Obtained From
Selected Monitoring Sites in the Washington Study Area 77
34 Baseline Air Quality Indicators for Nine Cities 83
35 Air Quality Adjustment Procedure Used to Simulate Attainment
of 1H1EX NAAQS (The Expected Number of Daily Maximum
One-Hour Ozone Concentrations Exceeding the Specified Value
Shall Not Exceed One) 87
36 Air Quality Adjustment Procedure Used to Simulate Attainment
of 8H1EX NAAQS (The Expected Number of Daily Maximum
Eight-Hour Ozone Concentrations Exceeding the Specified Value
Shall Not Exceed One) 88
37 Air Quality Adjustment Procedure Used to Simulate Attainment
of 8H5EX NAAQS (The Expected Number of Daily Maximum
Eight-Hour Ozone Concentrations Exceeding the Specified Value
Shall Not Exceed Rve) 89
38 Values for Equivalence Relationships 92
39 Determination of Adjustment Coefficients for One-Hour
NAAQS Attainment (1H1EX-120) in Philadelphia 94
40 Descriptive Statistics for Hourly-Hour Data (ppb) for Site
34-005-3001 (District 1, Philadelphia): Baseline and Attainment
of Three Ozone Standards 96
41 Determination of Adjustment Coefficients for Eight-Hour
NAAQS Attainment (8H1EX-80) in Philadelphia 97
42 Determination of Adjustment Coefficients for Eight-Hour
NAAQS Attainment (EH6LDM = 80 ppb) in Philadelphia 100
ix
-------
TABLES (continued)
Number Page
43 Number and Percent of Total Study Area Population
Experiencing One or More One-Hour Daily Maximum
Ozone Exposures Above 120 ppb at Any Ventilation Rate 108
44 Percent of Total Study Area Population Experiencing One or
More 8-Hour Daily Maximum Ozone Exposures Above
Indicated Exposure Concentrations at Any Ventilation Rate 111
45 Distributions of One-Hour Ozone Exposures (ppb) Obtained From
Ten Runs of pNEM/O3-S and From the Houston Asthmatic Study 124
46 Distributions of One-Hour Daily Maximum Ozone Exposures
(ppb) Obtained From Ten Runs of pNEM/O3-S and From
the Houston Asthmatic Study 125
47 Mean Ozone Concentrations in Microenvironments Based on
Personal Monitoring Data From the Houston Asthmatic Study
and on Exposure Estimates Obtained From 12 Runs of
pNEM/O3-S With Differing Values of Ozone Decay Rate 128
48 Multiplicative Factors Used to Determine Alternative Values
for Mean Ozone Decay Rate 129
-------
-------
ACKNOWLEDGMENT
This report describes a probabilistic version of NEM applicable to ozone
(pNEM/03). The model consists of two principal parts: 1) the "cohort exposure
program" which estimates the sequence of ozone exposures experienced by defined
population groups and 2) the "exposure extrapolation program" which estimates the
number of persons within a particular study that are represented by each cohort and
then combines this information with the cohort exposure sequences to determine the
distribution of exposures over a defined population-of-interest. Supplementary
programs are used to process air quality, population, and meteorological data for input
into each program.
All programs used in pNEM/OS were developed by IT Air Quality Services
(ITAQS) under the direction of the Ambient Standards Branch of the U.S.
Environmental Protection Agency (EPA). Mr. Ted Johnson of ITAQS developed the
general methodology for the model as described in Section 2 of this report. He also
developed the algorithms used to estimate outdoor ozone concentration, equivalent
ventilation rate, window status, and air exchange rate. In addition, Mr. Johnson
developed the algorithm used to sequence activity diary data and the algorithm used
to adjust air quality data to simulate attainment of specified ozone standards. Mr.
Johnson was the principal author of Sections 1 through 5 and Section 8 of this report.
Mr. Roy Paul served as project manager for the ITAQS effort. Mr. Paul and
Ms. Jill Warnasch were the principal authors of Section 6.
Dr. Louis Wijnberg of ITAQS developed the hourly-average mass balance
model used in the cohort exposure program. He also assisted in developing the
window status algorithm. Mr. Jim Capel wrote the cohort exposure program and a
majority of the supplementary programs. He also developed the algorithm used to
estimate cohort populations. Mr. Capel conducted the validation study described in
Section 7 and was the principal author of that section. Mr. John Carroll conducted a
review of the scientific literature concerning air exchange rates. Mr. Michael McCoy
implemented the air quality adjustment algorithm and was responsible for assembling
the population data required by the exposure extrapolation program. He also
developed the meteorological data files required by the cohort exposure program.
XI
-------
ITAQS work on this project was funded under EPA Contract No. 68-DO-0062.
Mr. Thomas McCurdy served as the EPA Work Assignment Manager and provided
guidance throughout the project. Mr. Richard Atherton was the EPA Project Officer.
The authors would like to express their appreciation to Dr. Charles J. Weschler
of Bell Communications Lab for his assistance in developing the ozone mass balance
model.
XII
-------
SECTION 1
INTRODUCTION
Within the U.S. Environmental Protection Agency (EPA), the Office of Air
Quality Planning and Standards (OAQPS) has responsibility for establishing and
revising National Ambient Air Quality Standards (NAAQS). In evaluating alternative
NAAQS proposed for a particular pollutant, OAQPS assesses the risks to human
health of air quality meeting each of the standards under consideration.1 This
assessment of risk requires estimates of the number of persons exposed at various
pollutant concentrations for specified periods of time. The estimates may be specific
to an urbanized area such as Los Angeles or apply to the entire nation.
Several researchers2-3 have recommended that such estimates be obtained by
simulating the movements of people through zones of varying air quality so as to
approximate the actual exposure patterns of people living within a defined area.
OAQPS has implemented this approach through an evolving methodology referred to
as the NAAQS Exposure Model (NEM). An early overview of the NEM methodology is
provided in a paper by Biller et al.4 From 1979 to 1988, IT Air Quality Services
(formerly PEI Associates, Inc.) assisted OAQPS in developing and applying pollutant-
specific versions of NEM to ozone,5 particulate matter,6 and CO.7 These versions of
NEM are referred to as "deterministic" versions in that no attempt was made to model
random processes within the exposure simulation.
The deterministic versions of NEM were similar in that each was capable of
simulating the movements of selected segments of an urban population through a set
of environmental settings. Each environmental setting was defined by a geographic
area and a microenvironment. The size and distribution of the geographic areas were
determined according to the ambient characteristics of the pollutant. Ambient
(outdoor) pollutant levels in each geographic area were estimated from either fixed-site
1
-------
monitoring data or dispersion model estimates. To better utilize fixed-site monitoring
data, researchers developed special time series techniques to fill in missing values
and special roll-back techniques to adjust the monitoring data to simulate conditions
under attainment of a particular NAAQS.
The population of interest in each study area was divided into an exhaustive set
of cohorts, and an activity pattern was developed for each cohort. The activity pattern
assigned the cohort to a geographic location and a microenvironment for each time
interval of a defined exposure period. In early NEM analyses, the time interval was 1
hour; in later NEM analyses, the time interval was reduced to 10 minutes. Exposure
periods varied from three months to one year. The activity patterns were based on
reviews of time use surveys. Researchers estimated the number of persons
represented by each cohort by accessing census and commuting data at the census-
tract level.
The pollutant concentration in a particular microenvironment was estimated by a
linear model which included terms relating to the ambient pollutant level and to
emission sources within the microenvironment. Researchers developed both point
estimates and distributions for the parameter values of these linear models by
conducting comprehensive reviews of the scientific literature associated with each
pollutant.
Additional details concerning the evolution of the deterministic version of NEM
are provided by Paul et al.8 Critiques of deterministic NEM are included in surveys of
exposure models by Pandian9 and Ryan.10 Two staff papers11-12 prepared by EPA
discuss the use of NEM in evaluating proposed NAAQS for CO and ozone.
In 1988, OAQPS began to incorporate probabilistic elements into the NEM
methodology and to apply the resulting model (pNEM) to the criteria pollutants. The
initial result of this work was an early version of pNEM applicable to ozone
(pNEM/O3). This model used a regression-based relationship to estimate indoor
ozone concentrations from outdoor concentrations. A report by Johnson et al.
describes this model and its application to Houston, Texas13.
-------
An advanced version of pNEM applicable to carbon monoxide (pNEM/CO) was
developed in 1991. This model marked the first time in the evolution of NEM that a
mass balance model was used to estimate indoor pollutant concentrations. The
application of pNEM/CO to Denver, Colorado, has been described by Johnson et al14.
A new version of pNEM/O3 was developed in early 1992. Unlike the earlier
version of pNEM/O3, the new model uses a mass balance model to estimate indoor
ozone concentrations. A report by Johnson et al. (February 1993)15 describes the new
version of pNEM/O3 and summarizes the results of an initial application of the model
to 10 cities.
Subsequent to the February 1993 report, ITAQS made the following
enhancements to pNEM/O3 and its input data bases.
Use of more recent (1990-91) fixed-site monitoring data for estimating
ambient ozone concentrations. The earlier analysis was based on 1981-
84 monitoring data.
An increase in the number of fixed-site monitors used to represent each
urban area.
Use of more recent (1990) census data for estimating cohort populations.
The earlier analysis used 1980 census data.
A new methodology for adjusting ambient ozone data to simulate
attainment of one-hour and eight-hour NAAQS.
Revision of the algorithm used to determine limiting values for equivalent
ventilation rate.
Development of origin/destination tables through the use of a new
commuting algorithm.
This report describes these enhancements and summarizes the results of applying the
enhanced model to nine of the ten cities included in the previous exposure
assessment. Tacoma, Washington, was excluded from the analysis because of
insufficient monitoring data.
The report is divided into seven sections. Section 2 provides an overview of
the pNEM/O3 methodology and describes in detail how the model was applied to a
-------
specific city (Houston). Section 3 describes the mass balance model incorporated into
pNEM/OS. Section 4 describes the process by which ambient ozone data sets were
selected for use in pNEM/O3. It also describes the methods used to fill in missing
values in these data sets. Section 5 presents the method used to adjust ambient
ozone data to simulate the attainment of proposed air quality standards. Section 6
provides ozone exposure estimates for each of the nine cities. Section 7 presents the
results of initial efforts to validate the new version of pNEM/O3. The principal
limitations of the model are discussed in Section 8.
-------
SECTION 2
OVERVIEW OF THE METHODOLOGY
The general NEM methodology consists of five steps.
1. Define a study area, a population-of-interest, appropriate subdivisions of
the study area, and an exposure period.
2. Divide the population-of-interest into an exhaustive set of cohorts.
3. Develop an exposure event sequence for each cohort for the exposure
period.
4. Estimate the pollutant concentration, ventilation rate, and physiological
indicator (if applicable) associated with each exposure event.
5. Extrapolate the cohort exposures to the population-of-interest and to
individual sensitive groups.
This approach has been followed in developing a probabilistic version of NEM
applicable to ozone (pNEM/03). The remainder of this section provides an overview of
the pNEM/O3 methodology. The application of pNEM/03 to Houston is used as a
means of demonstrating various features of the methodology.
2.1 Define Study Area, Population-of-lnterest, Subdivisions of Study Area, and
Exposure Period
The pNEM/O3 methodology provides estimates of the distribution of ozone
exposures within a defined population (the population-of-interest) for a specified
exposure period. The population-of-interest is typically defined as 1) all residents of a
defined study area or 2) the residents of the study area which belong to a specific
sensitive population. The study area is defined as an aggregation of exposure
districts. Each exposure district is defined as a contiguous set of census tracts or
-------
block numbering areas (jointly referred to as "census units") as defined by the Bureau
of Census for the 1990 U.S. census.
r All census units assigned to a particular exposure district were located within a
y
(^ specified radius of a fixed-site ozone monitor. The pNEM/O3 methodology is based
on the assumption that the ambient ozone concentration throughout each exposure
district can be estimated by ozone data provided by the associated fixed-site monitor.
Table 1 lists the nine study areas defined for the exposure analyses. Each
study area is associated with a major urban area. The table lists the number of
exposure districts and the exposure period for each study area. In each case, the
exposure period is defined as a series of months within a particular calendar year.
The specified months conform to the "ozone season" specified for the urban area by
EPA. The ozone season is the annual period when high ambient ozone levels are
likely to occur. Three ozone seasons appear in Table 1: January through December,
March through September, and April through October. The specified calendar year is
either 1990 or 1991, the selected year being the higher year with respect to reported
ambient ozone concentrations.
In the application of pNEM/O3 to Houston, eleven fixed-site monitors were
selected to represent ambient ozone concentrations (see Section 4). An/exposure
district jvas constructed around each monitor through the use of a special computer
program ("DIST90"). This program identified all census units having population
centroids located within 15 km of the monitor. When a census unit was paired with
more than one monitor, the program assigned it to the nearest monitor.
The sum of all census units assigned to the eleven exposure districts defined
the Houston study area. The residents of this area were designated as the principal
population-of-interest. In 1990, the study area consisted of 532 census units and
contained 2,370,512 residents16.
The Houston ozone season spans the entire calendar year. Consequently, the
Houston exposure period was defined as calendar year 1990.
-------
TABLE 1. CHARACTERISTICS OF STUDY AREAS
Study area
Chicago
Denver
Houston
Los Angeles
Miami
New York
Philadelphia
St. Louis
Washington
Number of
exposure
districts
12
7
11
16
6
12
10
11
11
1990
population8
6,175,121
1 ,484,798
2,370,512
10,371,115
1,941,994
10,657,873
3,785,810
1,706,778
3,085,419
Exposure period
Year
1991
1990
1990
1991
1991
1991
1991
1990
1991
Months
Apr-Oct
Mar-Sep
Jan- Dec
Jan-Dec
Jan- Dec
Apr-Oct
Apr-Oct
Apr-Oct
Apr-Oct
aSpecific to exposure districts which comprise the study area.
2.2 Divide the Population-of-lnterest Into an Exhaustive Set of Cohorts
In a pNEM analysis, the population-of-interest is divided into a set of cohorts
such that each person is assigned to one and only one cohort. Each cohort is
assumed to contain persons with identical exposures during the specified exposure
period. Cohort exposure is typically .assumed to be a function of demographic group,
location of residence, and location of work place. Specifying the home and work
district of each cohort provides a means of linking cohort exposure to ambient
pollutant concentrations. Specifying the demographic group provides a means of
linking cohort exposure to activity patterns that vary with age, work status, and other
demographic variables. In some analyses, cohorts are further distinguished according
to factors relating to characteristics of the residence, proximity to emission sources, or
to time spent in particular microenvironments.
-------
In the pNEM/03 analyses, each cohort was identified as a distinct combination
of 1) home district, 2) demographic group, 3) work district (if applicable), and 4)
residential air conditioning system. The home district and work district of each cohort
were identified according to the exposure districts defined for the study area. Table 2
lists nine demographic groups defined for the pNEM/03 analyses. Three of the
demographic groups are identified as workers. Each cohort associated with one of
these groups was identified by both home and work district. The remaining cohorts
were identified only by home district.
The decision to identify cohorts with respect to the residential air conditioning
system was based on the results of two supplemental analyses by ITAQS. An
analysis17 of data on window openings provided by the Cincinnati Activity Diary Study18
suggested that the time per day that windows are open in a residence is a function of
outdoor temperature and air conditioning .system, when the later is characterized as 1)
! no air conditioning, 2) room units, or 3) central air. An analysis19 of data collected by
Stock20 during a study of asthmatics in Houston suggested that indoor ozone levels
are significantly higher when windows are open than when windows are closed. For
example, the median ratio of indoor ozone to outdoor ozone for residences in the
Sunnyside section of Houston was 0.886 when windows were open and 0.088 when
windows were closed. The importance of outdoor ozone concentrations in determining
indoor ozone concentrations has also been reported by Weschler et al.21
Table 2 lists the number of cohorts in the Houston analyses that were
associated with each demographic group. Each of the six nonworking demographic
groups is associated with 33 cohorts, one for each combination of home district and
residential air conditioning system. Each of the three working demographic groups is
associated with 363 cohorts, one for each combination of home district, work district,
and residential air conditioning system. The total number of cohorts is thus (6 x 33) +
(3 x 363) or 1287.
8
-------
TABLE 2. DEMOGRAPHIC GROUPS DEFINED FOR PNEM/03 ANALYSIS
Demographic group
1.
2.
3.
4.
5.
6.
7.
8.
9.
Children 0 to 5 years
Children 6 to 13 years
Children 14 to 18 years
Workers with low probability of outdoor work
Workers with moderate probability of outdoor work
Workers with high probability of outdoor work
Nonworking adults under 35 years
Nonworking adults 35 to 54 years
Nonworking adults 55+ years
Total
Number of Houston
cohorts associated with
demographic group
33
33
33
363
363
363
33
33
33
1287
2.3 Develop an Exposure Event Sequence for Each Cohort for the Exposure
Period
In the pNEM/03 methodology, the exposure of each cohort is determined by an
exposure event sequence (EES) specific to the cohort. Each EES consists of a series
of events with durations from 1 to 60 minutes. To permit the analyst to determine
average exposures for specific clock hours, the exposure events are defined such that
no event falls within more than one clock hour. Each exposure event assigns the
cohort to a particular combination of geographic area and microenvironment. Each
event also provides an indication of respiration rate. In typical applications, this
indicator is a classification of slow - sleeping, slow - awake, medium, or fast.
The EESs are determined by assembling activity diary records relating to
individual 24-hour periods into a series of records spanning the ozone season of the
associated study area. Because each subject of a typical activity diary study provides
-------
data for only a few days, the construction of a multi-month EES requires either the
repetition of data from one subject or the use of data from multiple subjects. The
latter approach is used in pNEM analyses to better represent the variability of
exposure that is expected to occur among the persons included in each cohort.
The activity diary data used in the pNEM/OS analysis were obtained from the
Cincinnati Activity Diary Study (CADS)18. During this study over 900 subjects
completed three-day activity diaries and detailed background questionnaires. Figure 1
presents a page from the Cincinnati diary. Each subject was instructed to complete a
new diary page whenever he or she changed location or began a new activity.
The CADS data were assembled into a special data base organized by study
subject and 24-hour (midnight-to-midnight) time period. The diary records for one
subject for one 24-hour period were designated a "person-day." The CADS data base
contained 2649 person-days, each of which was indexed by the following factors:
1. Demographic group
2. Season: summer or winter
3. Temperature classification: cool or warm
4. Day type: weekday or weekend.
The demographic group index was determined by the demographic group to which the
subject filling out the diary belonged. The season and day type indices were based on
the calendar date of the person-day. The temperature classification was based on the
daily maximum temperature in Cincinnati on that date. The cool range was defined as
daily maximum temperatures below 55°F in winter and temperatures below 84°F in
summer.
A distinct EES was developed for each cohort in each of the nine study areas
by applying a computerized sampling algorithm to the CADS data base. The algorithm
was provided with the sequence of daily maximum temperatures reported for the
associated study area and exposure period (Table 1) and with the list of cohorts
10
-------
TIME
AM
PM
A. ACTIVITY (please specify)
B. LOCATION
In transit, car 01
In transit, other vehicle . . 02
Specify
Indoors, your residence .... 03
Indoors, other residence. . . 04
Indoors, office 05
Indoors, manufacturing
facility 06
Indoors, school 07
Indoors, store 08
Indoors, other 09
Specify
Outdoors, within 10 yards of
road or street 10
Outdoors, other 11
Specify
Uncertain 12
*Enter MIDN for midnight and NOON for noon. Otherwise enter four-digit
time (e.g., 0930 for 9:30 and 1217 for 12:17) and check a.m. or p.m.
Figure 1. Page from the activity diary used in the Cincinnati study.
C. BREATHING RATE
Slow (e.g., sitting) 13
Medium (e.g., brisk walk). . . 14
Fast (e.g., running) 15
Breathing problem 16
Specify
D. SMOKING
I am smoking 17
Others are smoking 18
No one is smoking 19
E. ONLY IF INDOORS
(1) Fireplace in use?
Yes 20
No 21
(2) Woodstove in use?
Yes 22
No 23
(3) Windows open?
Yes 24
No 25
Uncertain 26
11
-------
defined for the study area. The temperature data were used to assign each calendar
day in the exposure period to one of the temperature ranges used in classifying the
CADS data. To construct the EES for a particular cohort, the algorithm selected a
person-day from the CADS data base for each calendar day in the specified exposure
period according to the demographic group of the cohort and the season, day type,
and temperature classification associated with the calendar day.
Each exposure event within an EES was defined by 1) district, 2) micro-
environment, and 3) breathing rate category. The district was either the home or the
work district associated with the cohort. The home/work determination was based on
a decision rule which was applied to the activity diary data associated with the
exposure event.
Seven microenvironments were defined:
1. Indoors - residence - central air conditioning system
2. Indoors - residence - window air conditioning units
3. Indoors - residence - no air conditioning system
4. Indoors - nonresidential locations
5. Outdoors - near road
6. Outdoors - other
7. In vehicle.
As indicated in Table 3, location codes appearing in the Cincinnati activity diary data
base were used to determine the primary microenvironment location of each exposure
event (indoors - residence, indoors - nonresidential locations, outdoors- near road,
outdoors - other, or in vehicle). The indoors - residence location was subdivided into
three microenvironments according to air conditioning (AC) system: central system,
window unit(s), or none. This classification was based on the AC system specified for
the cohort's residence. For example, a cohort designated as residing in a home with
central AC would always be assigned to the microenvironment defined as "indoors -
residence - central AC" when activity diary data indicated the cohort was inside a
residence.
12
-------
TABLE 3. MICROENVIRONMENTS DEFINED FOR PNEM/03 ANALYSIS
Microenvironment
Activity diary locations assigned to
microenvironment
Indoors - residence
a) Central air conditioning system
b) Window air conditioning units
c) No air conditioning
Subject's residence
Other residence
Indoors - other
Office
Manufacturing facility
School
Store
Service station/repair facility
Other repair shop
Physical exercise facility
Auditorium
Sports arena
Museum or exhibition hall
Movie theater
Restaurant or cafeteria
Church
Shopping mall
Health care facility
Bowling alley
Bar
Other public building
Other indoor location
Indoors - not specified
Uncertain
Outdoor - near road
Within 10 yards of road
Motorcycle
Other vehicle
(continued)
13
-------
Table 3 (continued)
Microenvironment
Activity diary locations assigned to
microenvironment
Outdoors - other
Residential garage
Public garage
Parking lot
Outdoors - service station
Construction site
Residential grounds
School grounds
Playground
Sports arena or amphitheater
Park or golf course
Motorized boat
Nonmotorized boat
Other outdoor location
Outdoors - not specified
In vehicle
Car
Truck
Bus
Van
Train
Airplane
In transit - not specified
Four breathing rate categories were defined according to codes appearing in
the CADS data base: slow - sleeping, slow - awake, medium, and fast. Each
exposure event was assigned to one of these categories.
2.4 Estimate the Pollutant Concentration and Ventilation Rate Associated With
Each Exposure Event
In the general pNEM methodology, the EES defined for each cohort is used to
determine a corresponding sequence of exposures, event by event. Each exposure is
defined by a pollutant concentration and a ventilation rate indicator.
14
-------
2.4.1 Estimation of Pollutant Concentration
In the pNEM/03 analysis, the pollutant concentration during each exposure
event was assumed to be a function of the microenvironment and district associated
with the event. Consequently a continuous year-long sequence of hourly average
ozone concentrations was developed for each combination of microenvironment and
district. When an exposure event assigned a cohort to a particular combination of
microenvironment and district, the cohort was assigned the ozone concentration
specified for the corresponding clock hour in the appropriate microenvironment/district
sequence.
Each year-long sequence of hourly average ozone values for the indoor and in-
vehicle microenvironments was generated by the mass-balance algorithm described in
Section 3. Briefly, this algorithm estimated the hourly average indoor ozone
concentrations during hour h as a function of the indoor ozone concentration at the
end of the preceding hour (i.e., hour h -1), the ozone concentration outdoors during
hour h, the air exchange rate during hour h (v), and an ozone decay factor (FJ.
Values for the air exchange rate and the ozone decay factor were sampled from
appropriate distributions on a daily basis (Subsections 3.1 and 3.3). Air exchange rate
was permitted to change hourly in the three residential microenvironments depending
on whether windows were assigned a status of "open" or "closed". This assignment
was determined through the use of a probabilistic model (Subsection 3.4) in which the
status during each clock hour was assumed to be a function of AC system,
temperature range, and window status during the previous clock hour.
The outdoor ozone concentration associated with microenvironment m in district
d during hour h was determined by an expression having the general form
CoJm.d.t.s) = b(m) x Cmon(d,t,s) + e(t), (1)
where Cou,(m,d,t,s) is the outdoor (or ambient) ozone concentration in
microenvironment m in district d at time t under regulatory scenario s, Cmon(d>t,s) is the
ozone concentration estimated to occur at the monitor representing district d at time t
under regulatory scenario s, b(m) is a constant specific to microenvironment m, and
15
-------
e(t) is a random normal variable with mean = 0 and standard deviation = a(m). A
value for e(t) was selected from a normal distribution with mean = 0 and standard
deviation = o(m) every hour. The value of Cmon(d,t,s) was constant over each clock
hour.
In the application of pNEM/OS described in this report, b(m) was set equal to
1.056 for all microenvironments. A value of 5.3 ppb (0.0053 ppm) was used as the
value of o(m) for all microenvironments (Table 4). Consequently, each sequence of
hourly ozone values was generated by the expression
Cout(m,d,t,s) = 1.056 x Cmon(d,t,s) + e(t), (2)
where e(t) is a random normal variate with mean = 0 and standard deviation = 5.3
ppb.
The expression is based on the results of regression analyses performed by
ITAQS researchers on personal exposure data collected by T. Stock during the
Houston Asthmatic Study20. In these analyses, the dependent variable was five-
minute ozone concentration measured outdoors by a personal exposure monitor
(PEM). The independent variable was the simultaneous ozone concentration (hourly
average value) reported by the nearest fixed-site monitor.
An initial regression analysis of 327 paired values yielded an intercept of 0.81
ppb, a slope of 1.042, and set of regression residuals with a standard deviation of
18.5 ppb. The R-squared value was 0.544. Because the regression intercept value
was found to be non-significant (p = 0.76), a second regression analysis was
performed in which the regression line was forced through the origin (i.e., intercept =
0). This analysis yielded a slope of 1.056 and a set of regression residuals with a
standard deviation of 18.5 ppb. The residuals were found to be approximately normal
(skewness = -0.32, kurtosis = 0.87).
Attempts were made to fit more complex regression models to the Stock data.
These models included regression equations using logarithmic transformations of the
variables and regression equations which included the previous PEM value as an
16
-------
TABLE 4. PARAMETERS ASSOCIATED WITH ALGORITHMS USED
TO ESTIMATE OZONE CONCENTRATIONS IN MICROENVIRONMENTS
Parameter
b(m)
o(m)
Air exchange
rate
Ozone decay
factor
Equation (s)
containing
parameter
1
1
38
38
Microenvironment*
All
All
1-4,7
1 -4
7
Parameter value
1.056
5.3 ppb
See Table 10
Normal distribution
• Arith. mean = 4.04 h'1
•Std. dev. = I.SSrr1
• Minimum = 1.44 h"1
• Maximum = 8.09 h"1
72.0 h'1 1
"Microenvironments:
1 = Indoors - residence - central air conditioning
2 = Indoors - residence - window units
3 = Indoors - residence - none
4 = Indoors - nonresidential locations
5 = Outdoors - near road
6 = Outdoors - other
7 = In vehicle
independent variable. These alternative models were found to offer no significant
improvement in performance over the model specified above. Some of the alternative
models were found to be unstable.
The results of this analysis suggested that Equation 2 could be used as a
means of generating five-minute values of Coot(d,t,s), given that e(t) values were
selected every five minutes from a normal distribution with mean equal to 0 and
standard deviation equal to 18.5 ppb. A procedure based on this expression was
used in a previous version of pNEM/O3 to generate five-minute ozone concentrations
for the outdoor microenvironment13.
17
-------
As the new version of pNEM/03 required hourly-average outdoor ozone
concentrations rather than five-minute values, the procedure used in the earlier model
was modified so that an hourly-average value of e(t) was selected for each hour from
a normal distribution with mean equal to 0 and standard deviation equal to 5.3 ppb.
The use of a smaller standard deviation (5.3 ppb versus 18.5 ppb) for the hourly-
average e(t) terms was based on the statistical principle that the standard deviation of
the average of n values drawn from a distribution with standard deviation equal to a
will tend to have a standard deviation equal to oYm, where m is the square root of n.
As there are 12 five-minute values in one hour, the value of n is 12. The
corresponding value of m is 3.5, and 18.5 ppb/3.5 = 5.3 ppb.
The current version of pNEM/OS provides for two outdoor microenvironments:
No. 5 (outdoors - near road) and No. 6 (outdoors - other). In the pNEM/O3 analyses
described in this report, these microenvironments were treated identically; that is,
Equation 2 was used to determine the hourly ozone concentrations in each outdoor
microenvironment. This approach is likely to over-estimate ozone concentrations in
microenvironment No. 5 (outdoors - near road) because it does not account for
potential ozone scavenging by nitric oxides emitted from motor vehicles. The
magnitude of this bias is difficult to quantify because of the scarcity of research in this
area and the inconsistency of research findings. For example, a study by Rhodes and
Holland22 of a single freeway in San Diego found that downwind ozone concentrations
measured near the roadway were less than 28 percent of the ozone concentrations
measured simultaneously at more distant outdoor locations judged to be unaffected by
the roadway. However, an analysis19 of outdoor personal exposure data obtained
from the Stock study found that the average ratio of personal ozone concentration to
fixed-site ozone concentration was approximately 1.0 in areas of both low and high
traffic density.
•18
-------
2.4.2 The Air Quality Adjustment Model
In Equation 1, C^d.t.s) is the monitor-derived value for district d at time t
under scenario s. The value for this variable was determined by adjusting monitoring
data representing existing ("as is") conditions according to the equation
Cm(d.t.s) = (a)[CaOB(dlt,e)]b (3)
where Cmon(d,t,e) is the monitor-derived value for district d under existing conditions.
The multiplicative factor a and the exponent b are specific to district and scenario.
Section 5 describes the derivation of Equation 3 and provides examples of its
application to Philadelphia monitoring data.
Equation 3 requires a complete (gapless) year of hourly average C^d.te)
values for each district. These data sets were prepared by applying a special
interpolation program to the hourly average ozone data reported by each fixed-site
monitor. The interpolation program provided an estimate of each missing value. The
resulting filled-in data sets were assumed to represent existing conditions at each
monitor.
The interpolation program provides estimates of missing values through the use
of a time series model developed by Johnson and Wijnberg23. The time series model
is based on the assumption that hourly average air quality values can be represented
by a combination of cyclical, autoregressive, and noise processes. The parameter
values of these processes are determined by a statistical analysis of the reported
data.
2.4.3 Equivalent Ventilation Rate
In addition to ozone concentration, an equivalent ventilation rate (EVR) value
was estimated for each exposure event. EVR is defined as ventilation rate divided by
body surface area (BSA). Clinical research by EPA suggests that EVR exhibits less
inter-person variability than ventilation rate for a given level of exertion.24
The algorithm used to estimate EVR was employed previously in applications of
the pNEM methodology to ozone13 and carbon monoxide.14 This algorithm is based
19
-------
on an analysis25 of activity diary data provided by Dr. Jack Hackney. The data were
obtained from 36 subjects in Los Angeles who completed activity diaries identical to
those used in the Cincinnati study. The heart rate of each subject was monitored
during the period reported in the diary. Separate clinical trials were conducted to
determine a relationship between ventilation rate and heart rate for each subject.
These relationships and subject-specific BSA values were used to convert the one-
minute heart rate data associated with each diary activity to an average EVR value for
the activity. The resulting EVR estimates were then grouped by breathing rate
category (slow - sleeping, slow - awake, medium, fast). Statistical analysis indicated
that a two-parameter lognormal distribution provided a good fit to the EVR values in
each group. Table 5 lists the geometric mean and standard deviation of each fitted
distribution.
The appropriate distribution was randomly sampled to provide an EVR value for
each exposure event in the pNEM/03 simulation. EVR values were not permitted to
exceed an upper limit (EVRLIM) which varied with demographic group and event
duration. In all cases, the value of EVRLIM was set at a level estimated to be
achievable by members of the cohort who 1) exercised regularly, 2) were motivated to
attain high exertion levels, and 3) were not professional athletes. Joggers would be
included in this group; professional basketball players would not be included.
Table 6 presents the algorithm used to determine EVRLIM. The algorithm
accounts for the following research findings.
1. Ventilation rate (V^, oxygen uptake rate (VO^, and the ratio of VE to VO2
increase with increasing work rate.
2. A person's maximum VE is determined by his or her maximum oxygen
uptake rate (VO..^ and the Vg/VC^ ratio in effect under maximum
oxygen uptake conditions (MAXRATIO) such that
VEmax = (VO^ (MAXRATIO).
3- vozmax and MAXRATIO are functions of age, gender, and training,
among other factors.
20
-------
TABLE 5. PARAMETER VALUES OF LOGNORMAL DISTRIBUTIONS USED TO
CHARACTERIZE EQUIVALENT VENTILATION RATE
Age group
Children
Adults
Breathing rate
Slow-sleeping
Slow-awake
Medium
Fast
Slow-sleeping
Slow-awake
Medium
Fast
Parameter values of fitted lognormai
distribution
Geometric mean3
8.1
10.0
12.3
14.8
5.4
7.1
8.6
18.9
Geometric
standard deviation
1.60
1.46
1.44
1.62
1.22
1.36
1.34
1.92
al_iters/min per m2
4.
5.
Individuals cannot maintain oxygen uptake rates equal to
than about five minutes.
for more
For activity durations greater than five minutes (i.e., t > 5 min), the
percentage of VOa^ that can be maintained continuously (PCTVOJ
decreases as the natural logarithm of the activity duration pn(t)]
increases.
Research findings supporting these assumptions have been presented by Erb26, by
Astrand and Rodahl27, and by McArdle, Katch, and Katch.28 The algorithm itself is a
variation of "Algorithm B" proposed by Johnson and Adams.29
In determining the EVRLIM value applicable to a particular combination of
cohort and event duration, the algorithm uses estimates of VO,,,^, MAXRATIO,
SUBRATIO, and BSA specific to the demographic group associated with the cohort
(Table 7). The demographic group estimates provided in Table 7 were based on
estimates originally developed by Johnson and Adams29 for groups defined by age and
gender. Analysts defined each of the pN EM/03 demographic groups as a collection
21
-------
TABLE 6. ALGORITHM FOR DETERMINING UPPER LIMIT FOR EVR
1. Determine the demographic group of the cohort being simulated.
2. Obtain values for the following quantities from Table 7 according to
demographic group.
vozmax: maximum oxygen uptake rate
MAXRATIO: maximum ratio of ventilation rate to oxygen
uptake rate
SUBRATIO: submaximal ratio of ventilation rate to oxygen
uptake rate
BSA: body surface area
3. Determine duration of event (t).
4. If t <= 5 minutes, determine the upper limit for EVR (EVRLIM) by the
expression
EVRLIM =
5. If 5 minutes < t <= 162 minutes, determine the percentage of maximum
oxygen uptake rate that can be maintained for duration t by the expression
PCTVOana* = 116.19 - (10.06)pn(t)].
Next determine the ratio of ventilation rate to oxygen uptake rate by the
expression
RATIO = SUBRATIO +
(MAXRATIO-SUBRATIO)(PCTVO2max - 65)/35.
Finally determine EVRLIM by the expression
EVRLIM = (
6. If t > 162 minutes, determine PCTVOa^ by the expression presented in
Step 5 and EVRLIM by the expression
EVRLIM • (1.2)(VOa^(PCTVOaJ(SUBRAT10)/(100)(BSA).
22
-------
TABLE 7. PARAMETER VALUES FOR ALGORITHM USED TO DETERMINE LIMITS
FOR EQUIVALENT VENTILATION RATE
Demographic group
1 . Children 0 to 5 years
2. Children 6 to 13 years
3. Children 14 to 18 years
4. Workers with low probability of
outdoor work
5. Workers with moderate probability
of outdoor work
6. Workers with high probability of
outdoor work
7. Nonworking adults under 35 years
8. Nonworking adults 35 to 54 years
9. Nonworking adults 55+ years
Subgroup with largest
5-min EVRLIM
Sex
M
M
M
M
M
M
M
M
M
Age
5
11
15
18-24
18-24
18-24
18-24
35-44
55-64
5-min
EVRLIM
61.81
77.41
78.83
75.74
75.74
75.74
75.74
62.09
49.04
Subgroup characteristics
BSA
0.79
1.23
1.70
1.90
1.90
1.90
1.90
1.97
1.93
vo2max
0.99
2.30
3.49
3.69
3.69
3.69
3.69
3.07
2.52
MAXRATIO
41.1
34.5
32.0
32.5
32.5
32.5
32.5
33.2
31.3
SUBRATIO
35.0
26.0
22.5
21.0
21.0
21.0
21.0
22.3
23.5
ro
CO
-------
of the Johnson/Adams age-gender groups. Five-minute EVRLIM values were then
calculated for each of the age-gender groups included within each demographic group.
Table 7 lists the age-gender group associated with the largest EVRLIM value within
each demographic group. The parameter values associated with the identified age-
gender group were used to determine the EVRLIM values for the entire demographic
group.
The algorithm in Table 6 differs substantially from the EVR limit algorithm
employed in the February 1993 pNEM/O3 analysis described by Johnson et al.15
The new algorithm generally permits higher EVR values than the February 1993
algorithm.
2.4.5 Hourly Average Exposure Sequences
Algorithms within pNEM/03 provided three estimates for each exposure event:
average ozone concentration, average EVR, and the product of average ozone
concentration and EVR (ozone x EVR). These estimates were processed to produce
time-weighted estimates of ozone concentration, EVR, and ozone x EVR for each
clock hour. The result was a year-long sequence of hourly values for each of three
exposure indicators for each cohort. These sequences can be further processed to
determine cohort-specific values for various multihour exposure indicators. Examples
of such indicators include the largest eight-hour daily maximum ozone concentration
and the number of times the hourly-average ozone concentration exceeds 0.12 ppm.
2.5 Extrapolate the Cohort Exposures to the Population-of-interest and to
Individual Sensitive Groups
The cohort-specific exposure estimates developed in Step 4 of the pNEM
methodology (Subsection 2.4) were extrapolated to the general population of each
study area by estimating the population size of each cohort. Cohort populations were
estimated by the following four-step procedure. In Step 1, the number of persons
associated with each demographic group was estimated by census unit. In Step 2,
the fraction of homes falling into each of the three air conditioning categories was
24
-------
estimated by census unit. The fractions associated with each census unit were
determined using 1980 census data, as the 1990 census did not collect air
conditioning data. In cases where the boundaries of a 1990 census unit did not
coincide with 1980 census units, analysts used the fractions associated with the 1980
census unit located nearest to the 1990 census unit. In Step 3, the demographic
group populations were multiplied by the air conditioning fractions to provide an
estimate of the number of persons in each combination of demographic group and air
conditioning category. The estimation equation was
Pop(d,c/a) = F(c,a) * Pop(d,c), (4)
where Pop(d,c,a) is the population of a group associated with demographic group d,
census unit c, and air conditioning system a. F(c,a) is the fraction of housing units in
census unit c with air conditioning system a, and Pop(d.c) is the number of persons
associated with demographic group d in census unit c.
The values of Pop(d,c,a) were summed over each home district to yield
estimates of Pop(d,h,a), the number of persons in demographic group d within home
district h assigned to air conditioning category a.
The Pop(d,h,a) values provided an estimate of the population of each non-
commuting cohort residing within home district h. In Step 4, the populations of the
commuting cohorts (assumed to include all working cohorts) were determined by the
expression
Com(d,h,arw) = Pop(d,h,a) x Com(h,w)/Work(h). (5)
Com(d,hIa,w) is the number of persons in the commuting cohort associated with
demographic group d, home district h, AC system a, and work district w; Com(h,w) is
the number of workers in all demographic groups that commute from home district h to
work district w; and Work(h) is the total number of workers in home district h.
Estimates of Work(h) were developed from census data specific to each district.
Estimates of Com(h,w) were obtained from origin-destination (O-D) produced by a
. special commuting model.
25
-------
The pNEM/O3 commuting model is an enhanced version of a model developed
by Johnson et al.30 Briefly, the commuting model uses a trip duration model to
develop an O-D table. Trip duration data are collected during each census year for all
areas of the U.S. The data for 1990 are reported as the number of persons in each
census unit with one-way commute times that fall into each of the following twelve
commute duration ranges:
1. Less than 5 minutes
2. 5 to 9 minutes
3. 10 to 14 minutes
4. 15 to 19 minutes
5. 20 to 24 minutes
6. 25 to 29 minutes
7. 30 to 34 minutes
8. 35 to 39 minutes
9. 40 to 44 minutes
10. 45 to 59 minutes
11. 60 to 89 minutes
12. 90+ minutes.
The model assumes that each commute duration range can be converted into a
corresponding range of geodesic distances. Geodesic distance is defined here as the
shortest distance between two points on the globe, i.e., the distance "as the crow
files."
For example, a commuter in a large urbanized area may report that her
commute takes between 20 and 24 minutes. If the average geodesic commute speed
in the area is 0.3 kilometers per minute (km/min), the commute duration range of 20 to
24 minutes is equivalent to a geodesic distance range (GDR) of 6 to 7.2 km.
In a similar manner, each of the twelve commute distance ranges is converted
to a GDR indexed as i = 1, 2 12. The location of each census unit is represented
by its geographic centroid. If the census unit is an origin location for commuters (i.e.,
a "home" location), one can delineate twelve concentric rings centered on the centroid,
one for each GDR. The i-th GDR centered on home location h is identified as
GDR(h.i). Other useful terms are defined below.
26
-------
COM(h): Number of commuters residing in home location h
COM(h.i): Number of commuters residing in home location h who commute
to a work location in GDR(h.i)
COM(h,i,w): Number of commuters residing in home location h who commute
to work location w where work location w is in GDR(h.i)
N(h,i): Number of work locations in GDR(h.i)
Tot(h,i,w): Total number of commuters who work in work location w in
GDR(h.i)
Tot(h.i): Total number of commuters that work in GDR(h,i) (includes
commuters from all home locations).
The following method is employed to develop an O-D table.
1. Com(h.i) and N(h,i) are known. Make an initial estimate of Com(h,i,w)
using the expression
Com(h,i,w) = [Com(h,i)]/[N(hf±)] (6)
2. Calculate Tot(h,i,w) by summing Com(h,i,w) values associated with
specified w value
3. Calculate Tot(h.i) by summing Tot(h,i,w) values for all work locations in
GDR(h.i)
4. Make a new estimate of Com(h,i,w) using the expression
Com(h,i,w) = [Com(hfi)}[Tot(h,i,v))}/[Tot(h,i)} (7)
5. Go to Step 2.
In Step 1, commuters associated with a particular combination of home location and
GDR are evenly distributed across the work locations in the GDR. This step is
Iteration O. Steps 2 through 4 are repeated n times where n is determined by the
user. In each iteration of Steps 2 through 4, the commuters are redistributed across
27
-------
the work locations in the GDR in proportion to the number of workers assigned to
each work location during the last iteration.
In applying the commuting model to Houston, analysts first identified all
counties which were located within 50 km of the center of Houston. The 819 census
units located within these counties were assigned to a commute modeling zone. Each
census unit within the modeling zone was assumed to be both a potential home
location and a potential work location. Using commuting data from the 1990 census,
analysts applied the commuting model to the census units included in the modeling
zone and developed an origin-destination table. This table listed the number of
persons associated with each of the 670,761 (819 x 819) possible pairings of home
and work locations.
Analysts next defined 12 subdivisions of the commute modeling zone - one for
each of the 11 Houston exposure districts and an additional district (District #12)
containing all leftover census units. The 670,761 pairings of home and work census
units were aggregated into an origin-destination table listing the number of persons
associated with each of the 144 possible pairings of the 12 districts. This table was
used to estimate values of COM(h,w)/Work(h) - the fraction of workersjresiding in
••2? " ' ~n" " - - -- '
»> exposure district h who cQn]mutegLtg^^osjir_e_djsjrjcrw. Only persons who lived and
worked in one of the 11 Houston exposure districts were included in the exposure
assessment. Persons who lived or worked in the remaining district (i.e., District #12)
were excluded from the exposure analysis. These persons were assumed to spend a
significant part of each week in an area not included in one of the 11 Houston
exposure districts. The pNEM/O3 methodology does not provide a means for
estimating the exposures of people during periods when they are not within the
boundaries of an exposure district.
Table 8 lists the values of the quantity COM(h,w)/Work(h) determined by this
method for each of the 1.44 combinations of home and work district in Houston. Of
these, 13 combinations include District 12 as either the home or work location. As
indicated above, persons associated with these 13 entries were not included in the
exposure assessment.
28
-------
TABLE 8. ESTIMATED FRACTION OF HOUSTON
WORKERS WITHIN EACH HOME DISTRICT THAT
COMMUTE TO EACH WORK DISTRICT
District Identifier
Home
1
2
Work
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
Com (h.w)
Work (h)
0.360
0.004
0.049
0.245
0.013
0.005
0.003
0.005
0.015
0.010
0.145
0.1 06a
0.027
0.286
0.000
0.333
0.007
0.000
0.000
0.000
0.000
0.000
0.012
0.2573
District Identifier
Home
3
4
Work
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
Com fh.w)
Work (h)
0.190
0.000
0.150
0.072
0.015
0.038
0.008
0.015
0.120
0.073
0.244
0.01 9a
0.052
0.019
0.002
0.572
0.114
0.002
0.002
0.001
0.001
0.002
0.135
0.078a
District Identifier
Home
5
6
Work
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
Com fh.w)
Work (h)
0.006
0.001
0.001
0.230
0.476
0.011
0.013
0.002
0.002
0.003
0.181
0.0593
0.007
0.001
0.016
0.036
0.053
0.186
0.137
0.068
0.047
0.103
0.302
0.009a
(continued)
29
-------
Table 8 (continued)
District Identifier
Home
7
8
Work
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
f^f\n*\ /W \a A
uom (n,w)
Work (h)
0.002
0.000
0.003
0.015
0.043
0.124
0.357
0.188
0.026
0.057
0.097
0.0583
0.000
0.000
0.004
0.001
0.002
0.055
0.150
0.508
0.061
0.057
0.019
0.119s
District Identifier
Home
9
10
Work
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
^™^/>m /h \Ai\
worn in.wj
Work (h)
0.023
0.000
0.088
0.011
0.004
0.059
0.021
0.178
0.363
0.145
0.052
0.029a
0.015
0.000
0.046
0.027
0.016
0.162
0.107
0.135
0.115
0.143
0.197
0.006a
District Identifier
Home
11
12
Work
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
Com (h,w)
Work (h)
0.033
0.001
0.015
0.180
0.123
0.041
0.013
0.006
0.008
0.019
0.524
0.009s
0.041s
0.01 T
0.004s
0.098s
0.046s
0.002s
0.01 8a
0.041 a
0.0083
0.001 a
0.01 2a
0.5683
aPersons associated with this entry were not included in the exposure assessment.
30
-------
A special tabulation program in pNEM/03 combined the cohort-specific
estimates of exposure and population to produce histograms and cumulative
frequency tables for various population exposure indicators and averaging times.
Section 6 provides exposure estimates based on existing conditions in each study
area, the attainment of the current NAAQS, and the attainment of each of four
alternate NAAQS.
31
-------
SECTION 3
THE MASS-BALANCE MODEL
In the pNEM/O3 simulation, the ozone concentration in a particular
microenvironment during a particular clock hour is assumed to be constant. For
indoor and in-vehicle microenvironments, this value is determined by using a mass
balance model to calculate the average ozone concentration for the clock hour
expected under the following conditions:
1. There are no indoor sources of ozone.
2. The indoor ozone concentration at the end of the preceding hour is
specified.
3. The outdoor ozone concentration during the clock hour is constant at a
specified value.
4. The air exchange rate during the clock hour is constant at a specified
value.
5. Ozone decays at a rate that is proportional to the indoor ozone
concentration. The proportionality factor is constant at a specified value.
The mass balance model employed in these calculations is based on a generalized
mass balance model described by Nagda, Rector, and Koontz31, hereafter referred to
as the Nagda model. As originally proposed, this model assumed that pollutant
concentration decays indoors at a constant rate. For use in pNEM/O3, the Nagda
model was revised to incorporate the alternative assumption that the indoor decay rate
is proportional to the indoor concentration. The Nagda model was further revised to
incorporate ozone-specific assumptions concerning various parameter values
suggested by Weschler32 and others.
32
-------
Subsection 3.1 presents the theoretical basis for the pNEM/O3 mass balance
model and the principal model assumptions. Subsection 3.2 describes the algorithms
which were used to generate hourly values of ozone for the indoor and in-vehicle
microenvironments. Subsection 3.3 presents the procedure used to determine air
exchange rate for the mass balance model. An algorithm for simulating the opening
and closing of windows is described in Subsection 3.4.
3.1 Theoretical Basis and Assumptions
The Nagda model can be expressed by the differential equation
(3)
where C^ = Indoor concentration (units: mass/volume)
FB = Fraction of outdoor concentration intercepted by the enclosure
(dimensionless fraction)
v = Air exchange rate (1Aime)
CQU, = Outdoor concentration (mass/volume)
S = Indoor generation rate (mass/time)
cV = Effective indoor volume where c is a dimensionless fraction
(volume)
m = Mixing factor (dimensionless fraction)
X = Decay rate (mass/time)
q = Flow rate through air cleaning device (volume/time)
F = Efficiency of the air cleaning device (dimensionless fraction).
In this model, the pollutant decay rate (X) is assumed to be constant. Research by
Nazaroff and Cass33 and by Hayes34 suggests that the decay rate for ozone should be
proportional to C,,,. Consequently, the pNEM/03 mass balance equation substitutes
33
-------
the term Fd Ch for the term X/cV in Equation 8. The coefficient Fd is expressed in
units of 1Aime.
The following notational changes were made to simplify the equation:
G* — 1 C*
Fp-l-F,, (9)
va = cv. (10)
F is the "penetration factor," and V9 is the "effective volume." The resulting equation
p
is
(11)
If the three terms that are proportional to C-n are collected into one term, the equation
can be expressed as
where
^•p
(13)
It can be shown that Equation 12 has the following approximate solution:
34
-------
where
k2
and GOU, is the average value of the outdoor concentration over the interval t to t + At.
If Q&a is constant over the interval, then Equation 14 is an exact solution.
The average indoor concentration for hour h, C^ (h), is given by the expression
+ a3 (18)
where (^(h-l) is the instantaneous indoor concentration at the end of the preceding
hour, C^ (h) is the average outdoor concentration for hour h,
«! = z(h) , (19)
35
-------
a2 = (F-v/v') [1-zUz)], (20)
a3 = (-/-) [l-z(A)] , (21)
and
= i-e-v'v'. (22)
A steady-state version of the mass balance model can be developed by solving
Equation 12 under the conditions that
-£-c, = o (23)
dt **
and CQU, is constant. In this case, the mass balance equation is
0 = F vC +—- -v'Cjj,! (24)
which can be rearranged as
T7?T- (25)
36
-------
The ratio of indoor concentration to outdoor concentration is
" ../ * • <26>
Weschler32 has developed a steady-state equation for the indoor/outdoor ratio
which is expressed in his notation as
I/O = Ex/(Ex + kd(A/V)}, (27)
where I = indoor concentration, O = outdoor concentration, Ex = air exchange rate, kd
= deposition velocity, A = surface area, and V = volume. With respect to Equation 11,
Weschler's model implies that there are no indoor sources (S = zero), no air cleaning
devices (F = zero), the penetration factor is unity (Fp = 1), c = 1, and m = 1. Under
these conditions, Equation 11 becomes
(28)
and Equation 26 becomes
= TTF-' (29)
Weschler's model (Equation 27) and Equation 29 are equivalent if the following
substitutions are made:
(30)
37
-------
Cout - O (31)
v - JBL (32)
(33)
Equation 33 is a particularly useful relationship, as Weschler has identified a number
of studies which suggest that kd(A/V) is relatively constant from building to building.
He suggests that 1.0 x 10"3 sec"1 is a good general estimate of this quantity.
Weschler et al. present 14 estimates of kd(A/V) based on data obtained from
specific studies. Nine of these values are based on the observed first-order decay of
ozone in isolated rooms. The remaining five values are based on reported I/O values
and air exchange rates. Table 9 presents means and standard deviations for the first
nine estimates, for the last five estimates, and for all 14 estimates. Two-sided 95
percent confidence intervals for the means are also provided.
The values in Table 9 can be converted to units of h"1 by multiplying each value
by 3600. Expressed in these units, the mean and standard deviation for the 14
estimates are 4.04 h'1 and 1.35 h"1, respectively. A normal distribution with these
parameters was assumed to represent the distribution of Fd values for the non-vehicle
indoor microenvironments. The value of Fd was not permitted to be less than 1.44 h"1
or more than 8.09 h"1. The lower bound was based on the smallest value cited by
Weschler32 which was measured in a stainless steel room. The upper bound
corresponds to the 99.87 percentile (i.e., z = 3) of a normal distribution with mean
equal to 4.04 and standard deviation equal to 1.35. The largest value cited by
Weschler was 7.2 h"1.
38
-------
A point estimate of 72.0 was assumed for the value of Fd for the in-vehicle
microenvironment This value is based on an estimate of Fd for a single vehicle
reported by Petersen and Sabersky35. Hayes has used this value in applications of
the PAQM exposure model34.
TABLE 9. MEANS, STANDARD DEVIATIONS, AND CONFIDENCE INTERVALS
FOR ESTIMATES OF kd(A/V) PROVIDED BY WESCHLER
Parameter
Sample size
Mean, sec'1
Standard
deviation, sec"1
Two-sided 95%
confidence
interval, sec"1
Source of kd (A/V) estimate
Observed first-order
decay
9
1.133 x 10"3
0.447 x 10"3
(0.789, 1.477) x10"3
Reported
I/O values
5
1.098x 10"3
0.143 x 10"3
(0.920, 1.276) x10"3
All
14
1.121 x 10"3
0.374 x 10"3
(0.906, 1.335) x 10'3
3.2 Simulation of Microenvironmental Ozone Concentrations
Consistent with the theoretical considerations discussed in Subsection 3.1, the
following equation was used to estimate the hourly average ozone concentration in a
particular indoor or in-vehicle microenvironment during hour h:
(h) (34)
'in
out
where Cj, (h) is the average indoor ozone concentration during hour h, C-,,, (h-1) is the
instantaneous ozone concentration at the end of the preceding hour, C^ (h) is the
outdoor ozone concentration during hour h,
39
-------
z(h) , (35)
a, = (v/v7) [l-s(A>], (36)
z(h) = (l-e-v/)/v', (37)
and
(38)
The instantaneous ozone concentration at the end of a particular hour, Ch (h),
was estimated by the equation
cln(h) •kicla(h-i)^kt£(h). (39)
-in
where
jq « a"*' (40)
k, - (v/v') (1-JcJ, (41)
and v' is determined by Equation 38.
The following algorithm was used to generate a sequence of hourly-average
ozone concentrations for each combination of microenvironment and district.
1. Go to first/next day.
40
-------
2. Select value of air exchange rate for day from appropriate distribution or
use point estimate. If microenvironment is residential, select one air
exchange value for hours when windows are open and one for hours
when windows are closed. If microenvironment is a nonresidential
building or vehicle, then one air exchange rate is used for all hours of the
day.
3. Select value of decay rate (FJ for day from appropriate distribution or
use point estimate. If microenvironment is non-vehicular enclosure,
select value of Fd from normal distribution with mean = 4.04 h"1 and
standard deviation = 1.35 h'1. Value is not permitted to be less than 1.44
h"1 or more than 8.09 h"1. If microenvironment is "in vehicle", use point
estimate of 72.0 h"1.
4. Go to first/next clock hour.
5. If microenvironment is residential, use supplementary window algorithm
to determine window status for current hour (open or closed). Window
status determines which air exchange rate determined in Step 2 applies
to current hour.
6. Use Equation 34 to determine ozone concentration for current hour
based on air exchange rate specified for hour, outdoor ozone
concentration during hour, and ozone concentration at end of preceding
hour.
7. Use Equation 39 to determine instantaneous ozone concentration at end
of current hour based on air exchange rate specified for hour, outdoor
ozone concentration during hour, and instantaneous ozone concentration
at end of preceding hour. This value is saved for input into Equation 34
during the next hour.
8. If end of day, go to Step 1. Otherwise, go to Step 4.
Step 2 requires the random selection of an air exchange rate from a specified
distribution. Four enclosure categories were established for this purpose.
Residential buildings - windows open
Residential buildings - windows closed
Nonresidential buildings
Vehicles.
41
-------
A survey of the scientific literature determined that there were sufficient data available
to define distributions for only two of the four enclosure categories: "residential
building - windows closed" or "nonresidential building". In each case, a two-parameter
tognormal distribution was found to provide a good fit to the data. Point (single-
valued) estimates were developed for the remaining two enclosure categories.
Each of the two lognormal distributions was defined by the expression
AER = GM x GSDZ (42)
where AER is the air exchange rate, GM is the geometric mean, and GSD is the
geometric standard deviation. The values for GM and GSD were determined by fitting
lognormal distributions to representative data sets (Subsection 3.3). A value of AER
was selected at random from a particular lognormal distribution by randomly selecting
a value of Z from the unit normal distribution [N(0,1)] and substituting it into Equation
42. Table 10 lists the values of GM and GSD for the two lognormal distributions and
the values of the point estimates.
The distributions used to determine AER are discussed in more detail in
Subsection 3.3. Subsection 3.4 provides a description of the algorithm used to
determine window status in the residential microenvironments (Step 4),
3.3 Air Exchange Rate Distributions
A review of the scientific literature relating to air exchange rates identified 31
relevant references (list available on request). Of these, only a few were found to
contain sufficient data to construct a distribution of air exchange rates relating to a
particular building type such as residence or office. The two most useful studies were
conducted by Grimsrud et al.36 and by Turk et al.37
Residential Locations
Grimsrud et al. measured AER's in 312 residences. Reported AER values
ranged from 0.08 to 3.24. Researchers with IT Air Quality Services (ITAQS) analyzed
these data to determine which of two distributions (normal versus lognormal) better
42
-------
TABLE 10. DISTRIBUTIONS OF AIR EXCHANGE RATE VALUES USED
IN THE pNEM/03 MASS BALANCE MODEL
Enclosure category
Residential building-
windows closed
Residential building-
windows open
Nonresidential building
Vehicle
Air exchange rate distribution
Lognormal distribution
0 Geometric mean = 0.53
0 Geometric standard deviation =
0 Lower bound = 0.063
0 Upper bound = 4.47
Point estimate: 6.4
Lognormal distribution
0 Geometric mean = 1 .286
0 Geometric standard deviation =
0 Lower bound = 0.19
0 Upper bound = 8.69
1.704
1.891
Point estimate: 36
characterized the data. The lognormal distribution was found to yield a better fit, as
the data were highly skewed. The fitted lognormal parameters were geometric mean
= 0.53 and geometric standard deviation = 1.704. This distribution was used in
pNEM/03 to represent the distribution of AER's in residences with windows closed.
Upper and lower limits of 4.47 and 0.063 air changes per hour were established to
prevent the selection of unusually extreme values of AER. These limits correspond to
the substitution of Z = 4 and Z = -4 in Equation 42 when GM = 0.53 and GSD =
1.704. The upper bound is 38 percent larger than the largest reported AER (3.24).
The lower bound is 21 percent smaller than the smallest reported AER (0.08).
No comparable data bases were identified which were considered
representative of residences where windows are open. Hayes has used 6.4 as the
AER value for open windows in applications of the PAQM model.34 This value was
used in the pNEM/03 analyses presented here. This value will be replaced by a
distribution when an appropriate data base becomes available.
43
-------
Nonresidential Locations
Turk et al. measured AER's in 40 public buildings identified as schools (n = 7),
offices (n = 25), libraries (n = 3), and multipurpose buildings (n = 5). The minimum
reported AER was 0.3; the maximum was 4.1. Researchers with ITAQS fit normal
and lognormal distributions to the data for all 40 buildings and found that the
lognormal distribution produced a slightly better fit, although it had a tendency to over-
predict high values. The fitted lognormal parameters were geometric mean = 1.285
and geometric standard deviation = 1.891.
The buildings can be grouped as offices (n = 25) and nonoffices (n = 15).
Lognormal fits to these data sets yielded geometric means and standard deviations of
1.30 and 1.93 for offices and 1.27 and 1.87 for nonoffices. ITAQS performed a two-
sample t test on the two data sets and found no significant difference in the means or
standard deviations of the data. Consequently, a single lognormal distribution
(geometric mean = 1.285, geometric standard deviation = 1.891) was used in
pNEM/03 for all nonresidential buildings. To prevent the over-prediction of high AER
values, an upper bound of 8.69 was established. This value results when Z = 3 is
substituted into Equation 42 with GM = 1.285 and GSD = 1.891. This value is over
twice the largest AER value (4.1) reported for the 40 buildings and corresponds to the
99.87 percentile of the specified lognormal distribution. A lower bound of 0.19 was
also established. This value corresponds to a Z value of -3 and represents the 0.13
percentile of the lognormal distribution.
In Vehicle Locations
A point estimate of 36 air changes per hour was used for in-vehicle locations.
This value was obtained from Hayes38 based on his analysis of data presented by
Peterson and Sabersky35.
44
-------
3.4 Window Status Algorithm
The opening and closing of windows in the three residential microenvironments
was simulated by an algorithm which specified a window status (open or closed) for
each clock hour. The algorithm consisted of the following eight-step procedure.
1. Identify air conditioning system associated with cohort (central, window
units, none).
2. Go to first/next day.
3. Determine average temperature for day from supplementary file. Identify
temperature range which contains this value (below 32, 32 to below 63,
63 to 75, above 75).
4. Select random number between zero and 1. Compare random number
with probabilities listed in Table 11 for specified air conditioning system
and temperature range. Determine window status for day. If day status
is "windows open all day" or "windows closed all day", set window status
for all clock hours of day as indicated and go to Step 2. If day status is
"windows open part of day", go to Step 5.
5. Go to first/next clock hour.
6. Determine window status of preceding clock hour.
7. Select random number between zero and 1. Compare random number
with probabilities listed in Table 12, 13, or 14 for specified air
conditioning system, clock hour, temperature range, and window status
for preceding hour. If the random number is less than the specified
probability, the window will be open during the clock hour. Otherwise,
the window will be closed.
8. If end of day, go to Step 2. Otherwise, go to Step 5.
This algorithm assigns each day to one of three categories: 1) windows closed all
day, 2) windows open all day, and 3) windows open part of day. These assignments
are made according to the air conditioning system associated with the cohort and the
average temperature of the day. If the day assignment is "windows open part of day",
the algorithm assigns window status on an hourly basis for each of the 24 clock hours
in the day. These hourly assignments are made according to the 1) cohort's air
45
-------
conditioning system, 2) clock hour, 3) temperature of the day, and 4) window status of
the preceding hour. Both the daily and hourly assignments are made probabilistically
by comparing random numbers to the probabilities that the specified window status will
occur under the stated conditions.
The window status probabilities listed in Tables 11, 12, 13, and 14 were
developed through a statistical analysis of data on window openings obtained from the
Cincinnati Activity Diary Study (CADS).18 This analysis indicated that air conditioning
system, temperature, clock hour, and window status of preceding hour were
statistically significant factors affecting window status.
46
-------
TABLE 11. PROBABILITY OF WINDOW STATUS FOR DAY BY AIR
CONDITIONING SYSTEM AND TEMPERATURE RANGE
Air
conditioning
system
Central
Room units
None
Temperature
range, °F
Below 32
32 to 62
63 to 75
Above 75
Below 32
32 to 62
63 to 75
Above 75
Below 32
32 to 62
63 to 75
Above 75
Probability of window status for day
Closed all day
1.000
0.851
0.358
0.633
1.000
0.734
0.114
0.160
1.000
0.812
0.095
0.016
Open all day
0
0.009
0.343
0.167
0
0.028
0.505
0.380
0
0.011
0.672
0.823
Open part of day
0
0.140
0.299
0.200
0
0.238
0.381
0.460
0
0.177
0.233
0.161
TABLE 12. PROBABILITY OF WINDOWS BEING OPEN BY CLOCK HOUR,
TEMPERATURE RANGE, AND WINDOW STATUS OF PRECEDING HOUR (PH) FOR
RESIDENCES WITH CENTRAL AIR CONDITIONING
Clock
hour
1-3
4-6
7-9
10-12
13-15
16-18
19-21
22-24
Probability of windows being open
32°F to 62°F
PH=open
1.000
1.000
0.837
0.679
0.857
0.932
0.646
0.811
PH=closed
0.000
0.005
0.038
0.126
0.149
0.131
0.043
0.036
63°F to 75°F
PH=open
0.978
0.989
0.932
0.865
0.912
0.935
0.892
0.913
PH=closed
0.011
0.000
0.074
0.235
0.240
0.161
0.136
0.101
Above 75°F
PH=open
0.986
1.000
0.961
0.860
0.923
0.912
0.893
0.909
PH=closed
0.020
0.017
0.094
0.174
0.263
0.000
0.047
0.066
47
-------
TABLE 13. PROBABILITY OF WINDOWS BEING OPEN BY CLOCK HOUR,
TEMPERATURE-RANGE, AND WINDOW STATUS OF PRECEDING HOUR (PH) FOR
RESIDENCES WITH WINDOW AIR CONDITIONING UNITS
Clock
hour
1-3
4-6
7-9
10-12
13-15
16-18
19-21
22-24
Probability of windows being open
32°F to 62°F
PH=open
0.970
0.975
0.864
0.929
0.860
0.859
0.684
0.919
PH=closed
0.006
0.000
0.040
0.121
0.244
0.103
0.063
0.042
63°F to 75°F
PH=open
0.947
0.994
0.934
0.917
0.969
0.956
0.925
0.851
PH=closed
0.007
0.016
0.101
0.303
0.400
0.125
0.176
0.064
Above 75°F
PH=open
0.974
0.989
0.989
0.849
0.819
0.930
0.902
0.865
PH=closed
0.010
0.017
0.092
0.351
0.152
0.043
0.056
0.121
TABLE 14. PROBABILITY OF WINDOWS BEING OPEN BY CLOCK HOUR,
TEMPERATURE RANGE, AND WINDOW STATUS OF PRECEDING HOUR (PH) FOR
RESIDENCES WITH NO AIR CONDITIONING SYSTEM
/~>l_.-l-
ClOCK
hour
1-3
4-6
7-9
10-12
13-15
16-18
19-21
22-24
Probability of windows being open
32°F to 62°F
PH=open
1.000
1.000
0.950
0.889
0.923
0.848
0.609
0.684
PH=closed
0.015
0.000
0.000
0.200
0.130
0.200
0.067
0.043
63°F to 75°F
PH=open
0.974
1.000
0.868
0.933
1.000
0.964
0.909
0.800
PH=closed
0.031
0.000
0.057
0.400
0.286
0.000
0.500
0.167
Above 75°F
PH=open
1.000
1.000
1.000
0.875
0.917
0.818
1.000
0.769
PH=closed
0.000
0.000
0.000
0.500
0.000
0.667
0.200
0.500
48
-------
SECTION 4
PREPARATION OF AIR QUALITY DATA
The pNEM/03 mass balance model requires representative ambient air quality
data for each exposure district in the form of a time series containing one value for
each hour in the specified ozone season. This section describes the procedures used
to select appropriate data sets for the nine study areas. It also describes the
procedure used for filling in missing values in these data sets.
4.1 Selection of Representative Data Sets
To simplify the computer simulation, the ambient ozone concentration
throughout an exposure district was assumed to be a function of the ozone
concentration measured at a single, representative monitoring site located within the
/" district. Based on guidance from EPA, analysts defined the shape of each exposure
district by first drawing a circle of radius = 15 km with the monitoring site at the center.
If the centroid of a census unit (census tract or block numbering area) was located
within this circle, the census unit was assigned to the exposure district. If a centroid
was located within more than one circle, the census unit was assigned to the nearest
monitor. Note that the monitoring sites selected to represent a city directly determined
the location and shape of the city's exposure districts.
Researchers began the selection process by compiling a list of candidate
monitoring sites for each of the ten cities used in the February 1993 analysis.15
Chicago New York
Denver Philadelphia
Houston St. Louis
Los Angeles Tacoma (deleted)
Miami Washington
49
-------
The monitoring site list included all sites which reported data to EPA of acceptable
completeness for the ozone season of 1990 and/or 1991. The list provided an air
quality indicator (the second highest daily maximum 1-hour ozone concentration) for
each combination of site and ozone season.
A decision was made at this stage to exclude Tacoma from further analysis.
Only two ozone monitors operated in Tacoma during 1990 and 1991. These monitors
did not provide adequate geographic coverage of the Tacoma study area.
The air quality data complied for each of the remaining nine cities were
examined to determine which ozone season (1990 or 1991) was associated with
higher ozone levels across the city. The higher ozone season was designated the
exposure period for the city (Table 15). Researchers then selected a set of
representative monitoring sites to determine the location of the exposure districts for
each city (Tables 16 through 24). Sites were selected according to the following
general guidelines.
1. Select sites such that the associated exposure districts will provide good
geographic coverage of the city. Districts should include both center-city
and suburban areas with few large "holes" in the coverage of each
metropolitan area. No city should be represented by less than three
monitors. Six sites should provide adequate coverage for medium-sized
cities; ten sites should provide adequate coverage for all but the largest
cities.
2. To prevent excessive pNEM/O3 run times, no more than 16 sites should
be selected to represent a particular city. (Program run time varies
roughly as the square of the number of sites selected.)
3. Avoid sites which report unusually low ozone levels, as these may be
rural or near-road sites. Selected sites should represent typical urban
locations and provide a mix of average and high ozone levels.
Reference 39 provides a city-by-city discussion of the site selection process as
originally implemented. Researchers later determined that one of the New York
monitors was unrepresentative of ambient ozone conditions due to site location. This
monitor (identified by EPA as Site No. 36-061-0063) had been selected to represent
an exposure district centered on the southern end of Manhattan. Site No.
50
-------
TABLE 15. CHARACTERISTICS OF OZONE STUDY AREAS AND MONITORING SITES
Study area
Chicago
Denver
Houston
Los Angeles
Miami
New York City
Philadelphia
St. Louis
Washington, D.C.
Designated
exposure period
Ozone
season
Apr - Oct
Mar - Sep
Jan - Dec
Jan - Dec
Jan - Dec
Apr - Oct
Apr - Oct
Apr - Oct
Apr - Oct
Year
1991
1990
1990
1991
1991
1991
1991
1990
1991
Number of
. . a
counties
in area
7
6
5
4
2
18
13
7
13
Number of
monitoring
sites
selected
12
7
11
16
6
11b
10
11
11
Largest reported
second high daily
maximum ozone
concentration, ppb
129
110
220
310
123
175
156
125
144
aCounties are geographic areas assigned a county code by the Bureau of Census in
Summary Tape File 3 (STF3). A county is counted if any portion is within the study area.
"Monitor No. 36-061-0010 represents two exposure districts.
36-061-0063 was later judged to be unrepresentative of ground-level ozone
concentrations in this area of New York due to the site's high elevation. Consistent
with guidance from EPA, researchers selected the next nearest ozone monitor (No.
36-061-0010) to represent the Manhattan exposure district in the pNEM/O3 analysis.
Monitor No. 36-061-0010 also represents another exposure district which is centered
on the northern end of Manhattan, the actual location of this monitor.
Table 15 lists the number of ozone monitoring sites selected for each study
area. The table also indicates the largest value for the second highest daily
maximum hourly ozone concentration, reported by the selected monitors for the
indicated season. The omission of Monitor No. 36-061-0063 from the New York
Study area does not affect the value of this air quality indicator (175 ppb).
51
-------
TABLE 16. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE CHICAGO STUDY AREA
Monitor ID
17-031-0001
17-031-0032
17-031-1003
17-031-1601
17-031-4002
17-031-4003
17-031-7002
Monitor
location
Alsip
Chicago
Chicago
Lemont
Cicero
Des Plaines
Evanston
Dist-
rict
code
1
2
3
4
5
6
7
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4903
5136
4985
5136
4895
5136
4799
5136
5033
5136
4936
5136
4876
5136
Percentiles, ppb
50
19
19
28
28
19
19
28
28
18
18
23
23
30
30
90
51
50
58
59
51
50
61
60
49
49
53
52
59
58
95
61
61
69
69
63
61
71
71
60
59
63
63
69
69
99
77
77
87
87
81
81
89
89
78
78
80
80
90
90
99.5
83
83
92
92
88
87
98
97
86
86
85
86
97
96
High values,
ppb
Second
104
104
116
116
129
129
126
126
120
120
105
105
115
115
First
108
108
120
120
134
134
152
152
125
125
119
119
123
123
ro
(continued)
-------
TABLE 16 (Continued)
Monitor ID
17-031-8003
17-043-6001
17-089-0005
17-097-0001
17-097-1002
Monitor
location
Calumet City
Lisle
Elgin
Deerfield
Waukegan
Dist-
rict
code
8
9
10
11
12
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4856
5136
5100
5136
5041
5136
5011
5136
5038
5136
Percentiles, ppb
50
23
24
19
20
26
26
26
26
30
30
90
54
54
49
50
54
54.
56
56
61
61
95
64
64
59
59
63
63
67
67
71
71
99
81
81
78
78
82
82
85
85
92
92
99.5
86
86
87
87
91
90
90
90
102
102
High values,
ppb
Second
97
97
116
116
126
126
116
116
119
119
First
109
109
118
118
128
128
124
124
126
126
Ol
03
during designated ozone season.
-------
TABLE 17. DESCRIPTIVE STATISTICS FOR 1990 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE DENVER STUDY AREA
Monitor ID
08-001-3001
08-005-0002
08-005-0003
08-013-1001
08-031-0002
08-031-0014
08-059-0002
Monitor
location
Adams Co.
Arapaho Co.
Englewood
Boulder Co.
Denver
Denver
Arvada
Dist-
rict
code
1
2
3
4
5
6
7
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4322
5136
4047
5136
5036
5136
4458
5136
5063
5136
4453
5136
4908
5136
Percentiles, ppb
50
26
26
40
39
23
23
33
32
17
17
22
22
26
26
90
54
53
63
60
53
54
55
54
40
40
54
53
56
55
95
59
58
70
67
62
62
64
63
47
46
62
61
64
64
99
69
68
88
86
76
76
78
77
59
59
77
75
79
79
99.5
72
72
93
91
83
83
83
80
64
64
83
81
83
83
High values,
ppb
Second
87
87
109
109
110
110
102
102
104
104
107
107
115
115
First '"
99
99
111
110
111
111
106
106
120
120
120
120
115
115
'Number of hourly-average ozone concentrations during designated ozone season.
-------
TABLE 18. DESCRIPTIVE STATISTICS FOR 1990 DATA SETS CONTAINING HOURLY-AVER AGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE HOUSTON STUDY AREA
Monitor ID
48-201-0024
48-201-0029
48-201-0046
48-201-0047
48-201-0051
48-201-0059
48-201-0062
Monitor
location
Harris Co.
Harris Co.
Houston
Houston
Houston
Houston
Houston
Dist-
rict
code
1
2
3
4
5
6
7
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
n"
6865
8760
7689
8760
8138
8760
7970
8760
7999
8760
6941
8760
8072
8760
Percentiles, ppb
50
20
20
20
20
10
10
10
10
20
20
10
10
20
20
90
60
60
50
50
50
50
50
50
50
50
40
40
50
46
95
80
70
70
70
60
60
60
60
70
70
50
50
60
60
99
110
110
100
100
100
100
100
100
110
110
80
70
100
90
99.5
130
120
120
110
120
120
120
120
130
130
90
90
110
110
High values,
ppb
Second
220
220
160
160
200
200
210
210
200
200
140
140
180
180
First
220
220
180
180
230
230
240
240
220
220
190
190
230
230
Ol
Ol
(continued)
-------
TABLE 18 (Continued)
Monitor ID
48-201-1003
48-201-1034
48-201-1035
48-201-1037
Monitor
location
Deer Park
Houston
Houston
Houston
Dist-
rict
code
8
9
10
11
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
na
7685
8760
8098
8760
8300
8760
8086
8760
Percentiles, ppb
50
20
20
10
10
10
10
10
10
90
50
50
50
45
50
50
40
40
95
60
60
60
60
60
60
60
60
99
100
100
90
90
100
100
100
100
99.5
110
110
120
110
120
120
120
120
High values,
ppb
Second
230
230
200
200
230
230
220
220
First
-•-'.
230
230
210
210
230
230
220
220
01
en
"Number of hourly-average ozone concentrations during designated ozone season.
-------
TABLE 19. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE LOS ANGELES STUDY AREA
Monitor ID
06-037-0016
06-037-1103
06-037-1301
06-037-1601
06-037-1902
06-037-2005
06-037-4002
06-037-5001
Monitor location
Glendora
Los Angeles
Lynwood
Pico Rivera
Santa Monica
Pasadena
Long Beach
Hawthorne
Dist-
rict
code
1
2
3
4
5
6
7
8
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
8416
8760
8356
8760
8478
8760
8523
8760
8179
8760
8344
8760
8377
8760
8465
8760
Percentiles, ppb
50
20
20
10
10
10
10
10
10
26
25
10
10
20
20
20
20
90
80
80
50
50
40
40
60
60
65
64
70
70
40
40
50
50
95
110
110
70
70
50
50
80
80
80
79
100
100
50
50
60
60
99
180
180
120
110
80
80
130
130
114
112
160
160
70
70
80
80
99.5
200
200
130
130
90
90
160
160
131
128
170
170
80
80
90
90
High values,
ppb
Second
310
310
170
170
130
130
250
250
191
191
220
220
100
100
110
110
First
320
320
190
190
160
160
260
260
191
191
230
230
110
110
110
115
en
(continued)
-------
TABLE 19 (Continued)
Monitor ID
06-059-0001
06-059-1003
06-059-3002
06-059-5001
06-065-8001
06-071-1004
06-071-4003
06-071-9004
Monitor location
Anaheim
Costa Mesa
Los Alamitos
La Habra
Rubidoux
Upland
Redlands
San Bernardino
Dist-
rict
code
9
10
11
12
13
14
15
16
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
8473
8760
8358
8760
8442
8760
8492
8760
8521
8760
8408
8760
8374
8760
8514
8760
Percentiles, ppb
50
10
10
30
30
20
20
20
15
20
20
10
10
30
30
20
13
90
50
50
50
50
50
50
60
53
90
80
70
70
90
90
80
80
95
60
60
60
60
60
60
70
70
110
110
100
90
120
120
110
110
99
100
100
80
80
90
90
110
110
160
160
160
160
180
180
160
160
99.5
110
110
90
90
100
100
130
130
180
180
180
180
190
190
170
170
High values,
ppb
Second
200
200
140
140
150
150
190
190
240
240
240
240
250
. 250
240
240
First
250
250
170
170
170
170
210
210
240
240
270
270
250
250
250
250
Ol
00
aNumber of hourly-average ozone concentrations during designated ozone season.
-------
TABLE 20. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE MIAMI STUDY AREA
Monitor ID
12-011-0003
12-011-2003
12-011-8002
12-025-0021
12-025-0027
12-025-0029
Monitor
location
Broward Co.
Pompano
Beach
Dania
Dade Co.
Dade Co.
Dade Co.
Dist-
rict
code
1
2
3
4
5
6
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
=====
na
8624
8760
8664
8760
8732
8760
8470
8760
8486
8760
8576
8760
Percentiles, ppb
50
22
22
23
23
26
26
21
21
28
28
21
21
90
42
42
41
41 .
43
43
41
41
44
44
39
39
«••"-"-'
95
48
48
46
46
49
49
46
46
49
49
45
44
99
59
59
58
58
61
61
57
57
58
57
54
54
99.5
63
63
64
63
64
64
64
63
65
64
58
58
High values,
ppb
Second
93
93
91
91
95
95
123
123
90
90
85
85
First
94
94
96
96
100
100
124
124
95
95
90
90
Ol
CO
"Number of hourly-average ozone concentrations during designated ozone season
-------
TABLE 21. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE NEW YORK STUDY AREA
Monitor ID
09-001-0017
34-013-0011
34-017-0006
34-027-3001
34-039-5001
36-001-0080
36-061-0010
Monitor
location
Greenwich
Newark
Bayonne
Morris Co.
Plainfield
Bronx Co.
New York City
Dist-
rict
code
1
2
3
4
5
6
7,8
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4882
5136
5033
5136
4968
5136
4691
5136
4986
5136
4422
5136
4893
5136
Percentiles, ppb
50
29
29
18
18
24
24
39
39
19
20
12
13
14
14
90
61
60
52
52
64
64
75
73
55
55
36
36
43
42
95
75
74
67
67
81
80
88
86
69
68
47
45
58
57
99
110
110
92
92
109
108
111
111
90
90
68
67
87
87
99.5
120
118
97
97
116
116
118
118
97
96
72
72
95
95
High values,
ppb
Second
147
147
123
123
166
166
137
137
115
115
92
92
151
151
First
161
161
132
132
167
167
139
139
120
120
94
94
155
155
O)
o
(continued)
-------
TABLE 21 (Continued)
Monitor ID
36-061-0063
36-081-0004
36-085-0067
36-103-0002
36-119-2004
Monitor
location
New York City
Queens Co.
Richmond Co.
Babylon
White Plains
Dist-
rict
code
b
9
10
11
12
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4912
5136
4912
5136
4086
5136
4884
5136
4975
5136
Percentiles, ppb
50
41
41
20
20
28
29
30
30
27
27
90
82
82
57
57
67
62
67
67
62
61
95
96
95
72
. 72
81
77
81
80
78
78
99
122
122
105
105
106
103
111
110
107
107
99.5
130
130
115
115
116
111
121
120
116
116
High values,
Ppb
Second
175
175
162
162
169
169
175
175
145
145
First
177
177
174
174
178
178
217
217
152
152
aNumber of hourly-average ozone concentrations during designated ozone season.
"Originally assigned to District 8. Replaced by Monitor No. 36-061-0010.
CD
-------
TABLE 22. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE PHILADELPHIA STUDY AREA
Monitor ID
34-005-3001
34-007-0003
34-007-1001
34-015-0002
42-017-0012
42-045-0002
42-091-0013
Monitor
location
McGuire AFB
Camden
Camden
Gloucester
Bristol
Chester
Norristown
Dist-
rict
code
1
2
3
4
5
6
7
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No.
Yes
No
Yes
na
4939
5136
4998
5136
4989
5136
5001
5136
4986
5136
5085
5136
4907
5136
Percentiles, ppb
50
35
34
28
28
36
36
33
33
28
28
30
30
26
26
90
72
72
70
70
76
76
74
73
70
70
67 -
67
67
66
95
88
88
84
84
89
89
87
87
84
84
78
78
78
77
99
117
117
115
114
112
112
115
115
111
110
103
103
99
98
99.5
126
124
120
120
117
117
125
125
119
118
108
108
106
105
High values,
ppb
Second
156
156
143
143
146
146
151
151
139
139
125
125
125
125
First
156
156
148
148
149
149
151
151
144
144
135
135
127
127
(continued)
-------
TABLE 22 (Continued)
Monitor ID
42-101-0014
42-101-0023
42-101-0024
Monitor
location
Philadelphia
Philadelphia
Philadelphia
Dist-
rict
code
8
9
10
Filled
in?
No
Yes
No
Yes
No
Yes
rva
4900
5136
4786
5136
4984
5136
Percentiles, ppb
50
30
30
20
20
30
30
90
70
70
50
50
70
70
g"-—" ' :':•'<•:- —
95
80
80
70
70
80
80
99
100
100
90
90
110
110
99.5
110
110
100
100
110
110
High values,
ppb
Second
140
140
130
130
130
130
First
140
140
130
130
140
140
"Number of hourly-average ozone concentrations during designated ozone season
-------
TABLE 23. DESCRIPTIVE STATISTICS FOR 1990 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE ST. LOUIS STUDY AREA
Monitor ID
17-163-0010
29-183-1002
29-189-0001
29-189-0006
29-189-3001
29-189-5001
29-189-7001
Monitor
location
East St. Louis
St. Charles
Affton
St. Louis Co.
Clayton
Ferguson
St. Ann
Dist-
rict
code
1
2
3
4
5
6
7
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4963
5136
4587
5136
4218
5136
5038
5136
5042
5136
5026
5136
5036
5136
Percentiles, ppb
50
19
19
23
27
28
29
24
24
24
24
18
18
29
29
90
48
48
55
55
62
59
48
48
53
54
42
42
58
58
95
57
57
66
66
75
72
55
55
65
65
48
47
70
70
99
73
73
90
90
93
90
70
69
83
83
61
61
92
92
99.5
83
82
102
98
100
99
75
75
93
92
64
64
96
96
High values,
ppb
Second
116
116
125
125
120
120
99
99
125
125
75
75
130
130
First
124
124
125
125
127
127
100
100
127
127
80
80
135
135
(continued)
-------
TABLE 23 (Continued)
Monitor ID
29-510-0007
29-510-0062
29-510-0072
29-510-0080
Monitor
location
St. Louis
St. Louis
St. Louis
St. Louis
Dist-
rict
code
8
9
10
11
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
na
5008
5136
4928
5136
4830
5136
5044
5136
Percentiles, ppb
50
18
18
24
24
18
18
24
24
90
44
44
53
53
40
40
53
53
95
52
52
63
63
48
48
64
65
99
69
69
82
82
64
64
86
86
99.5
74
74
89
89
72
72
94
94
High values,
ppb
Second
96
96
108
108
100
100
117
117
First
96
96
111
111
110
110
129
129
"Number of hourly-average ozone concentrations during designated ozone season.
-------
TABLE 24. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING HOURLY-AVERAGE OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE WASHINGTON STUDY AREA
Monitor ID
11-001-0017
11-001-0025
24-031-3001
24-033-0002
24-033-8001
51-013-0020
51-059-0018
Monitor
location
Washington
Washington
Rockville
Greenbelt
Suitland-
Silver
Hills
Arlington Co.
Mt. Vernon
Dist-
rict
code
1
2
3
4
5
6
7
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
na
4928
5136
5031
5136
4881
5136
5034
5136
4997
5136
5034
5136
4897
5136
Percentiles, ppb
50
19
19
24
24
29
29
30
30
31
31
28
28
30
30
90
54
54
61
60
69
68
74
74
69
68
68
68
71
71
95
64
64
72
71
79
79
87
87
81
81
80
79
83
83
99
82
82
90
90
100
99
110
109
102
102
102
102
106
105
99.5
91
90
99
99
103
103
115
114
108
108
107
107
111
111
High values,
ppb
Second
137
137
144
144
135
135
148
148
139
139
142
142
126
126
First
147
147
148
148
137
137
153
153
144
144
148
148
142
142
o>
O)
(continued)
-------
TABLE 24 (Continued)
Monitor ID
51-059-1004
51-059-5001
51-510-0009
51-600-0005
Monitor
location
Seven
Corners
McLean
Alexandria
Fairfax
Dist-
rict
code
8
9
10
11
Filled
in?
No
Yes
No
Yes
No
Yes
No
Yes
na
4951
5136
5037
5136
4916
5136
4947
5136
Percentiles, ppb
50
33
33
27
27
22
22
33
32
90
71
71
63
63
54
54
66
66
95
86
86
73
74
65
65
77
76
99
110
109
95
95
84
84
97
96
99.5
119
116
104
101
95
94
107
106
High values,
ppb
Second
174
174
137
137
131
131
131
131
First
178
178
138
138
132
132
132
132
"Number of hourly-average ozone concentrations during designated ozone season.
CD
-------
4.2 Treatment of Missing Values and Descriptive Statistics
Hourly average ozone data reported by each site were used to estimate the
ambient ozone levels within the associated exposure district. Gaps in the hourly
average ozone data sets were filled in by using a time series model developed by
Johnson and Wijnberg23 and previously applied to hourly average ozone data by
Johnson et al.15. The model contains cyclical, autoregressive, and noise components
whose parameters are determined from a statistical analysis of the reported data.
Tables 16 through 24 provide descriptive statistics for each hourly-average data
set before and after application of the fill-in program. In general, the fill-in program
has little or no effect on the listed percentiles or high values. Whenever there is a
difference in the values for a particular percentile, the filled-in value is usually lower.
It should be noted that the data sets differ in terms of concentration resolution.
The reported ozone concentration values for all 11 Houston sites and for 15 of the 16
Los Angeles sites are rounded to the nearest 10 ppb. The data for the other seven
cities are rounded to the nearest 1 ppb. All other factors being equal, the algorithm
used to fill in missing values generally performs better when applied to air quality data
of high resolution.
Researchers also constructed a data set for each monitor listing eight-hour
running average ozone concentrations based on the filled-in data sets. These data
were used to determine each site's status with respect to various eight-hour NAAQS
under consideration by EPA. Tables 25 through 33 provide eight-hour descriptive
statistics for the monitors selected to represent each city.
68
-------
TABLE 25. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE CHICAGO STUDY AREA
Monitor ID
17-031-0001
17-031-002
17-031-1003
17-031-1601
17-031-4002
17-031-4003
17-031-7002
17-031-8003
17-043-6001
17-089-0005
17-097-0001
17-097-1002
Monitor
location
Alsip
Chicago
Chicago
Lemont
Cicero
Des Plaines
Evanston
Calumet City
Lisle
Elgin
Deerfield
Waukegan
District
code
1
2
3
4
5
6
7
8
9
10
11
12
Percentiles, ppb
50
20
28
19
28
18
24
30
24
20
26
26
31
90
46
54
46
57
45
48
55
49
45
50
52
58
95
54
63
55
66
54
57
64
58
53
58
61
66
99
69
80
71
82
70
72
83
74
70
74
77
84
99.5
75
84
76
88
75
77
86
78
79
82
83
88
High values, ppb
Second
94
106
101
108
95
93
101
90
98
106
101
104
First
95
107
101 I
109 • I
95
95 '
102
90
98
106
103
106
-------
TABLE 26. DESCRIPTIVE STATISTICS FOR 1990 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE DENVER STUDY AREA
Monitor ID
08-001-3001
08-005-0002
08-005-0003
08-013-1001
08-031-0002
08-031-0014
08-059-0002
Monitor
location
Adams Co.
Arapaho Co.
Englewood
Boulder Co.
Denver
Denver
Arvada
District
code
1
2
3
4
5
6
7
Percentiles, ppb
50
26
38
24
33
18
23
26
90
47
56
48
50
35
47
50
95
52
62
54
57
41
52
57
99
60
76
65
68
51
62
68
99.5
63
80
70
71
54
64
72
High values, ppb
Second
72
87
83
83
84
77
95
First
74
87
83
85
85
80
96
-------
TABLE 27. DESCRIPTIVE STATISTICS FOR 1990 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE HOUSTON STUDY AREA
Monitor ID
48-201-0024
48-201-0029
48-201-0046
48-201-0047
48-201-0051
48-201-0059
48-201-0062
48-201-1003
48-201-1034
48-201-1035
48-201-1037
Monitor
location
Harris Co.
Harris Co.
Houston
Houston
Houston
Houston
Houston
Deer Park
Houston
Houston
Houston
District
code
1
2
3
4
5
6
7
8
9
10
11
Percentiles, ppb
50
21
21
14
15
21
14
17
19
16
15
12
90
50
49
42
42
48
33
41
46
41
42
39
95
64
61
53
55
61
41
52
56
54
56
51
99
92
89
84
82
92
60
79
84
81
86
81
99.5
104
96
95
96
105
71
90
92
90
97
92
High values, ppb
Second
149
124
151
156
167
110
154
139
144
156
160
First
150
124
152
164
170
112
155
140
146
157
164
-------
TABLE 28. DESCRIPTIVE
CONCENTRATIONS OBTAINED
STATISTICS FOR
FROM SELECTED
1991 DATA SETS CONTAINING EIGHT-HOUR OZONE
MONITORING SITES IN THE LOS ANGELES STUDY AREA
Monitor ID
06-037-0016
06-037-1103
06-037-1301
06-037-1601
06-037-1902
06-037-2005
06-037-4002
06-037-5001
06-059-0001
06-059-1003
06-059-3002
06-059-5001
06-065-8001
06-071-1004
06-071-4003
06-071-9004
Monitor
location
Glendora
Los Angeles
Lynwood
Pico Rivera
Santa Monica
Pasadena
Long Beach
Hawthorne
Anaheim
Costa Mesa
Los Alamitos
La Habra
Rubidoux
Upland
Redlands
San
Bernardino
District
code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Percentiles, ppb
50
24
14
12
12
27
18
17
21
17
25
25
17
24
16
30
19
90
70
47
34
51.
58
62
35
46
42
47
50
50
76
61
86
74
95
95
60
41
67
69
84
42
51
52
55
59
62
97
84
110
96
99
135
85
62
97
93
120
56
67
77
71
75
90
139
124
152
135
99.5
150
92
67
111
101
130
61
76
85
76
80
100
155
134
162
146
High values, ppb
Second
181
120
86
142
155
165
82
96
119
101
97
129
194
164
197
192
First
182
120
89
146
155
166
83
99
119
102
99
132
196
165
197
192
NJ
-------
TABLE 29. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE MIAMI STUDY AREA
Monitor ID
12-011-0003
12-011-2003
12-011-8002
12-025-0021
12-025-0027
12-025-0029
. Monitor location
Broward Co.
Pompano Beach
Dania
Dade Co.
Dade Co.
Dade Co.
District
code
1
2
3
4
5
6
Percentiles, ppb
50
22
22
25
21
27
21
90
39
39
42
37
43
37
95
44
44
47
43
47
42
99
54
52
56
52
55
51
99.5
56
54
59
55
58
53
High values, ppb
Second
76
71
71
77
77
73
First
77
72
72
79
80
73
W
-------
TABLE 30. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE
EIGHT-HOUR OZONE
NEW YORK STUDY AREA
Monitor ID
09-001-0017
34-013-0011
34-017-0006
34-027-3001
34-039-5001
36-001-0080
36-061-0010
36-061-0063
36-081-0004
36-085-0067
36-103-0002
36-119-2004
Monitor
location
Grenwich
Newark
Bayonne
Morris Co.
Plainfield
Bronx Co.
New York City
New York City
Queens Co.
Richmond Co.
Babylon
White Plains
District
code
1
2
3
4
5
6
7,8
a
9
10
11
12
Percentiles, ppb
50
29
19
25
39
21
14
15
41
21
29
30
27
90
57
46
58
70
50
32
39
79
51
58
62
58
95
67
59
72
82
61
41
50
90
64
71
73
70
99
95
82
95
100
80
56
73
113
90
95
97
94
99.5
103
89
103
109
88
59
79
122
99
101
104
105
High values, ppb
Second
125
102
112
125
109
69
102
133
119
135
129
125
First
126
103
112
125
109
71
102
135
119
136
129
127
aOriginally assigned to District 8; represented by Monitor No. 36-061-0010.
-------
TABLE 31. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE PHILADELPHIA STUDY AREA
Monitor ID
34-005-3001
34-007-0003
34-007-1001
34-015-0002
42-017-0012
42-045-0002
42-091-0013
42-101-0014
42-101-0023
42-101-0024
Monitor
location
McGuire
AFB
Camden
Camden Co.
Gloucester
Bristol
Chester
Norristown
Philadelphia
Philadelphia
Philadelphia
District
code
1
2
3
4
5
6
7
8
9
10
Percentiles, ppb
50
34
28
37
33
28
30
26
31
21
27
90
67
64
71
68
64
62
60
65
49
61
95
80
76
81
80
76
72
70
76
60
72
99
107
101
103
105
100
92
92
96
79
97
99.5
114
109
107
113
104
98
98
100
86
103
High values, ppb
Second
138
129
124
135
115
113
118
125
112
116
First
141
131
125
135
116
114
118
127
114
116
-------
TABLE 32. DESCRIPTIVE STATISTICS FOR 1990 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE ST. LOUIS STUDY AREA
Monitor ID
17-163-0010
29-183-1002
29-189-0001
29-189-0006
29-189-3001
29-189-5001
29-189-7001
29-510-0007
29-510-0062
29-510-0072
29-510-0080
Monitor
location
East St.
Louis
St. Charles
Affton
St. Louis
Co.
Clayton
Ferguson
St. Ann
St. Louis
St. Louis
St. Louis
St. Louis
District
code
1
2
3
4
5
6
7
8
9
10
11
Percentiles, ppb
50
20
26
30
24
25
19
29
19
25
19
25
90
43
50
54
44
49
39
54
40
48
37
50
95
51
59
64
50
58
44
64
48
57
43
60
99
66
78
80
62
76
54
81
60
73
56
76
99.5
70
85
85
67
80
56
84
65
77
61
83
High values, ppb
Second
98
110
100
85
93
62
101
76
89
83
99
First
99
110
103
86
94
63
104
77
91
85
100
-------
TABLE 33. DESCRIPTIVE STATISTICS FOR 1991 DATA SETS CONTAINING EIGHT-HOUR OZONE
CONCENTRATIONS OBTAINED FROM SELECTED MONITORING SITES IN THE WASHINGTON STUDY AREA
Monitor ID
11-001-0017
11-001-0025
24-031-3001
24-033-0002
24-033-8001
51-013-0020
51-059-0018
51-059-1004
51-059-5001
51-510-0009
51-600-0005
Monitor
location
Washington
Washington
Rockville
Greenbelt
Suitland S.H.
Arlington Co.
Mt. Vernon
Seven
Corners
McLean
Alexandria
Fairfax
District
code
1
2
3
4
5
6
7
8
9
10
11
.___ — .
Percentiles, ppb
50
20
25
30
31
32
29
30
33
28
23
33
I — .
90
48
55
62
.68
63
61
65
66
56
50
61
95
58
64
71
79
73
72
75
77
65
59
70
99
73
79
88
96
90
91
92
97
81
75
88
99.5
78
85
93
102
94
97
99
102
89
82
96
High values, ppb
Second
120
114
113
129
124
127
110
147
115
111
110
First
120
117
113
131
125
128
112
147
115
111
111
-------
SECTION 5
ADJUSTMENT OF OZONE DATA TO SIMULATE COMPLIANCE
WITH ALTERNATIVE AIR QUALITY STANDARDS
In applying pNEM/03 to a particular study area, the analyst typically defines the
air quality conditions within the area as representing (1) baseline conditions or (2)
conditions in which the area just attains a specific National Ambient Air Quality
Standard (NAAQS). In the analyses described in this report, fixed-site monitoring data
for the years 1990 and 1991 were used to represent baseline conditions for each of
the nine study areas. Special air quality adjustment procedures (AQAP's) were used
to adjust the baseline data to simulate conditions in which each study area just attains
a specific NAAQS.
EPA identified the following NAAQS formulations for assessment:
1. One hour daily maximum - one expected exceedance (1H1EX): the
expected number of daily maximum one-hour ozone concentrations
exceeding the specified value shall not exceed one.
Standard level: 120 ppb (the current NAAQS for ozone)
2. Eight-hour daily maximum -- one expected exceedance (8H1 EX): the
expected number of daily maximum eight-hour ozone concentrations
exceeding the specified value shall not exceed one.
Standard levels: 80 ppb, 100 ppb
3. Eight-hour daily maximum - five expected exceedances (8H5EX): the
expected number of daily maximum eight-hour ozone concentrations
exceeding the specified value shall not exceed five.
Standard levels: 60 ppb, 80 ppb
A separate AQAP was developed for each of the three classes of NAAQS (1H1EX,
8H1EX, and8H5EX).
78
-------
Each AQAP consisted of the following four steps:
1. Specify an air quality indicator (AQI) to be used in evaluating the status
of a monitoring site with respect to the NAAQS of interest.
2. Determine the value of the AQI for each site within the study area under
baseline conditions.
3. Determine the value of the AQI under conditions in which the air pollution
levels within the study area have been reduced until the site with the
highest pollution levels just attains a specified NAAQS.
4. Adjust the one-hour values of the baseline data set associated with each
site to yield the AQI value determined in Step 3. The adjusted data set
should retain the temporal profile of the baseline data set.
Subsection 5.1 discusses the specification of appropriate AQI's (Step 1) and the
determination of baseline AQI values (Step 2). Subsection 5.2 presents the methods
used to estimate AQI's under attainment conditions (Step 3). Subsection 5.3
describes the procedures used in Step 4 to adjust one-hour data to simulate
significant reductions in ozone levels within a study area. More detailed descriptions
of these procedures can be found in Appendices A and B. Subsection 5.4 provides
examples in which the procedures described in Subsection 5.3 were applied to
Philadelphia. Subsection 5.5 presents an alternative procedure which analysts used to
adjust one-hour data to simulate small changes (decreases or increases) in ozone
levels within a study area. This procedure was applied to Denver, Chicago, and Miami
for selected NAAQS formulations.
5.1 Specification of AQI and Estimation of Baseline AQI Values
The following AQI's were selected for evaluating the 1H1EX, 8H1EX, and
8H5EX standards.
1H1EX: the characteristic largest daily maximum one-hour ozone
concentration
8H1 EX: the characteristic largest daily maximum eight-hour ozone
concentration
79
-------
8H5EX: the observed sixth largest daily maximum eight-hour ozone
concentration.
Note that a statistical AQI (the characteristic largest value) was specified for the
1H1EX and 8H1EX standards, whereas a deterministic AQI (the observed sixth largest
value) was used for the 8H5EX standards. Analysts elected to use statistical AQI's for
the 1H1EX and 8H1EX standards because such indicators are less affected by
anomalous high values than the corresponding deterministic AQI (the second highest
observed value). A statistical indicator was not considered necessary for the 8H5EX
standards, as the sixth highest observed value is relatively unaffected by anomalous
high values.
The characteristic largest value (CLV) of a distribution is that value expected to
be exceeded once in n observations. If F(x) is the cumulative distribution of x, then
F(x) =1-1 (43)
n
when x is the CLV.
Selection of an appropriate cumulative distribution to fit data is important in
determining a reasonable CLV. Two distributions that often provide close fits to
ambient air quality data are the Weibull and the lognormal. The Weibull distribution is
defined as
F(x) = 1 - exp [-(*)*] (44)
where 8 is the scale parameter and k is the shape parameter. The lognormal
distribution is defined as
L f-^exp (-t2/2) dt (45)
27C
where
and In x is distributed normally with mean ji and variance a2. As discussed in
previous reports, the Weibull distribution generally provides a better fit to hourly
80
-------
w= ln
15
average ozone data.
The hourly average values reported by a single monitoring site during a
specified ozone season form a time series x, (t = 1, 2, 3, ..., n). If the hourly average
time series is complete, it will contain n = (24)(N) values, where N is the number of
days in the ozone season. From this time series a second time series of daily
maximum 1-hour values can be constructed.
Assume that a Weibull distribution with parameters 8 and k provides a good fit
to the empirical distribution of hourly average values. If one disregards
autocorrelation, the value expected to be exceeded once in n = (24)(N) hours can be
estimated as
CLVOH = 5 [ln(24) (N)]i/k. (47)
This is the characteristic largest one-hour value. If we again disregard autocorrelation,
the daily maximum 1-hour value expected to be exceeded once in N days can be
estimated as
CLVOHDM = 5{-ln[l - ( J!li )1/24] }1/fc. (48)
N
This is the characteristic largest daily maximum one-hour value. For 7-month and 12-
month ozone seasons, N is equal to 214 and 365, respectively. For these values of
N, CLVOH and CLVOHDM are virtually indistinguishable in value over the range in k
values typically found in ozone data (0.6 < k < 2.5). For example, the following values
were calculated using 6 = 40 ppb.
81
-------
0.6
1.4
2.5
0.6
1.4
2.5
1428
185
94
1580
193
97
1428
185
94
1580
193
97
N. k CLVOH CLVOHDM
214
365
The CLVOH and CLVOHDM values match to the nearest ppb. Consequently,
the expression
CLVOHDM = 5 [In (24) (N} }1/k (49)
can be used as an alternative to Equation 48 for calculating CLVOHDM. The quantity
calculated by Equation 49, hearafter denoted by CLV1, was selected as the AQI to be
used in evaluating the status of a monitoring site with respect to a particular 1H1EX
standard.
A data set containing one-hour concentration values can be processed to
determine a corresponding data set containing eight-hour running average values. If a
Weibull distribution is fit to the eight-hour data, one can determine a characteristic
largest eight-hour value by the equation
CLVEH = 8[ln(24)tf) ]1/fc, (50)
where 6 and k are the Weibull parameters for the eight-hour fit. Based on the
argument made above for one-hour data, this value should be approximately equal to
the characteristic largest daily maximum eight-hour value (CLVEHDM) of the data set.
For simplicity, the term CLV8 is hereafter used to refer to the quantity calculated by
Equation 50. CLV8 was selected as the AQI to be used in evaluating attainment
status with respect to a particular 8H1EX standard.
Table 34 lists the data sets selected to represent baseline conditions in each of
the nine cities under analysis. Table 34 also provides estimates of CLV1 and CLV8
82
-------
TABLE 34. BASELINE AIR QUALITY INDICATORS FOR NINE CITIES
City
Chicago
Denver
Houston
Year
1991
1990
1990
District
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
1
2
3
4
5
6
7
8
9
10
11
Ozone concentration, ppb
CLV1
109
124
123
134
120
111
119
104
122
127
122
131
91
116
114
103
98
117
109
224
182
241
224
227
180
208
207
231
235
232
CLV8
94
107
106
114
99
97
106
92
106
111
106
111
74
94
85
86
79
78
94
162
137
161
171
179
131
165
143
154
171
167
EH6LDM
78
86
77
90
78
79
89
78
82
83
85
91
67
84
73
74
56
65
75
116
110
110
107
124
86
104
99
101
116
107
(Continued)
83
-------
Table 34 (Continued)
City
Los Angeles
Miami
New York
Year
1991
1991
1991
District
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
2
3
4
5
6
1
2
3
4
5
6
7
8a
9
10
11
12
Ozone concentration, ppb
CLV1
321
185
148
271
215
248
116
136
198
153
167
216
264
266
261
249
90
97
93
105
96
87
158
121
153
143
123
97
141
141
162
170
183
148
CLV8
207
133
99
166
162
172
85
104
121
101
100
134
209
184
215
204
74
74
72
82
80
72
135
112
133
134
113
75
108
108
131
143
140
137
EH6LDM
170
109
75
129
115
146
64
84
94
81
87
110
167
146
180
165
60
60
64
59
65
57
108
91
113
105
88
64
83
83
104
101
107
105
(Continued)
84
-------
Table 34 (Continued)
City
Philadelphia
St. Louis
Washington
Year
1991
1990
1991
District
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10
11
Ozone concentration, ppb
CLV1
167
149
153
162
145
134
135
140
131
141
124
141
131
103
122
78
124
100
114
103
119
134
135
130
143
141
143
135
169
141
134
145
CLV8
142
136
128
138
120
118
123
128
116
126
100
116
106
87
97
65
103
79
91
84
104
110
113
113
128
119
123
118
143
120
112
123
EH6LDM
116
113
111
115
107
101
102
104
90
102
73
88
87
68
81
59
87
67
80
64
86
80
88
95
106
98
100
104
102
91
85
100
"Districts 7 and 8 in New York are represented by the same ozone monitor (Monitor
No. 36-061-0010).
85
-------
based on Weibull fits to the upper two percent of each data set. These values were
used as estimates of CLV1 and CLV8 representing baseline conditions.
As previously indicated, the sixth largest daily maximum 8 hour value (denoted
EH6LDM) was used to evaluate the status of a monitoring site with respect to a
particular 8H5EX standards. Table 34 lists the baseline value of this AQI for each site
in the nine cities under analysis.
5.2 Estimation of AQI's Under Attainment Conditions
Tables 35, 36, and 37 provide the step-by-step procedures followed in
implementing the AQAP's developed respectively for 1H1EX, 8H1EX, and 8H5EX
NAAQS. In general, each AQAP assumed that the i-th ranked site (ranking
determined by baseline AQI) will undergo a change in its AQI value proportional to the
change required for the highest ranked site to exactly attain the specified standard.
The ranking assigned to a particular site under attainment conditions was determined
by the site's average ranking over five years, rather than the site's ranking under
baseline conditions. Consequently, the site ranked highest under baseline conditions
was not necessarily the highest ranked site under attainment conditions. Evaluation of
representative ozone data suggested that a site's future ranking could be better
predicted from its long-term average rank than from a single year's ranking.
Steps 1 through 4 in each table comprise the procedures used to estimate the
value of an attainment AQI value for each site in a particular city. Each attainment
AQI was converted to a corresponding characteristic one-hour largest value under
attainment (ACLV1). For 1H1EX standards (Table 35), the value of ACLV1
determined by Step 4 was used without further adjustment as the value of ACLV1
required in subsequent steps. For 8H1EX standards (Table 36), the value of ACLV8
determined in Step 4 was converted to the required ACLV1 value through the use of
an equivalence relationship (Step 5). The equivalence relationship was
ACLVl = (RATI01) (ACLV8) (54)
where RATI01 varied with city (Table 38).
86
-------
TABLE 35. AIR QUALITY ADJUSTMENT PROCEDURE USED TO SIMULATE
ATTAINMENT OF 1H1EX NAAQS (THE EXPECTED NUMBER OF
DAILY MAXIMUM ONE-HOUR OZONE CONCENTRATIONS EXCEEDING THE
SPECIFIED VALUE SHALL NOT EXCEED ONE)
1. Determine the following quantities.
CLV1(i,j): the CLV1 of i-th ranked site in City j for the "baseline" or
"start" year.
MAXCLV10): the largest CLV1 of all sites in City j for the
baseline year.
AMAXCLV1Q): the largest CLV1 value permitted under the
proposed 1-hr NAAQS.
2. Select five years prior to the baseline year and determine the value of
CLV1 (or related air quality indicator) at each site m in City j for each
year. Rank these values by city and year. Let RANK(m,j,y) indicate
the rank of site m in city j in year y. Let MEANRANK(m,j) indicate the
mean value of RANK(m,j,y) over the five years. Rank the
MEANRANK(m.j) values and let RELRANK(mj) indicate the relative
rank of MEANRANK(m.j).
3. Calculate an adjusted CLV1 for the i-th ranked site in City j by the
expression
ACLVl(i,j) = [CLVl(i,j)](AMAXCLVl(j}]/(MAXCLVl(j)}. (51)
4. If RELRANK(mj) = i, then m will be the i-th ranked site in City j under
attainment. That is,
ACLV1(m,j) = ACLV1(i,j) if RELRANK(mJ) = i.
5. The 1-hour data at Site m under attainment will be determined by
adjusting the 1-hour data at Site m in the baseline year. A Weibull
distribution fit to the adjusted data will have a CLV1 equal to
ACLV1(i,j) where i = RELRANK(mj). Subsection 5.3 provides a
method for estimating the parameters of this distribution and for
making the adjustment.
87
-------
TABLE 36. AIR QUALITY ADJUSTMENT PROCEDURE USED TO SIMULATE
ATTAINMENT OF 8H1EX NAAQS (THE EXPECTED NUMBER OF DAILY
MAXIMUM EIGHT-HOUR OZONE CONCENTRATIONS EXCEEDING
THE SPECIFIED VALUE SHALL NOT EXCEED ONE)
1. Determine the following quantities.
CLV8(i,j): the eight-hour CLV of i-th ranked site in City j for
the "baseline" or "start" year.
MAXCLV8(j): the largest CLV8 of all sites in City j for the
baseline year.
AMAXCLV8(j): the largest CLV8 value permitted under the
proposed 8-hr NAAQS.
2. Select five years prior to the baseline year and determine the value of
CLV8 (or related air quality indicator) at each site m in City j for each
year. Rank these values by city and year. Let RANK(m,j,y) indicate
the rank of site m in city j in year y. Let MEANRANK(m,j) indicate the
mean value of RANK(m,j,y) over the five years. Rank the
MEANRANK(mj) values and let RELRANK(mJ) indicate the relative
rank of MEANRANK(mJ).
3. Calculate an adjusted CLV8 for the i-th ranked site in City j by the
expression
ACLV8(i,j) = [CLV8(i,j}]{AMAXCLV8(j)]/[MAXCLV8(j}}. (52)
4. If RELRANK(mJ) = i, then m will be the i-th ranked site in City j under
attainment. That is,
ACLV8(m,j) = ACLV8(i,j) if RELRANK(mj) = i.
5. Using Equation 54, estimate the CLV1 associated with each
ACLV8(m,j) value. Denote this value as ACLV1 (m,j).
6. The 1-hour data for Site m under attainment of the 8-hr NAAQS will
be determined by adjusting the 1-hour data for Site m in the baseline
year. A Weibull distribution fit to the adjusted data will have a CLV1
equal to ACLV1(i,j) where i = RELRANK(m.j). Subsection 5.3
provides a method for estimating the parameters of this distribution
and for making the adjustment.
88
-------
TABLE 37. AIR QUALITY ADJUSTMENT PROCEDURE USED TO SIMULATE
ATTAINMENT OF 8H5EX NAAQS (THE EXPECTED NUMBER OF DAILY
MAXIMUM EIGHT-HOUR OZONE CONCENTRATIONS EXCEEDING THE
SPECIFIED VALUE SHALL NOT EXCEED FIVE)
1. Determine the following quantities.
EH6LDM(i,j): the EH6LDM of the i-th ranked site in City j for
the baseline year,
MAXEH6LDMG): the largest EH6LDM of all sites in City j for the
baseline year.
AMAXEH6LDM(j): the largest EH6LDM value permitted under the
proposed 1-hr NAAQS.
2. Select five years prior to the baseline year and determine the value of
EH6LDM (or related air quality indicator) at each site m in City j for
each year. Rank these values by city and year. Let RANK(m,j,y)
indicate the rank of site m in city j in year y. Let MEANRANK(mj)
indicate the mean value of RANK(m,j,y) over the n years. Rank the
MEANRANK(mj) values and let RELRANK(mj) indicate the relative
rank of MEANRANK(mJ).
3. Calculate an adjusted EH6LDM for the i-th ranked site in City j by the
expression
AEH6LDM(i,j) = [EH6LDM(i,j)][(AMAXEH6LDM(j)/[MAXEH6LDM(j)]. (53)
4. If RELRANK(mJ) = i, then m will be the i-th ranked site in City j under
attainment. That is,
AEH6LDM(m,j) = AEH6LDM(i,j) if RELRANK(m,j) = i.
5. Using Equation 55, estimate the CLV1 associated with each
AEH6LDM(m,j) value. Denote this value as ACLV1 (m,j).
6. The 1 -hour data for Site m under attainment of the 8H5EX NAAQS
will be determined by adjusting the 1-hour data for Site m in the
baseline year. A Weibull distribution fit to the adjusted data will have
a CLV1 equal to ACLV1(i,j) where i = RELRANK(mj). Subsection 5.3
provides a method for estimating the parameters of this distribution
and for making the adjustment.
89
-------
A similar method was employed for 8H5EX standards (Table 37). The value of
AEH6LDM determined in Step 4 was converted to the required ACLV1 value through
the use of an equivalence relationship (Step 5). In this case, the equivalence
relationship was
ACLVI = (RATI02) (AEH6LDM] (55)
where RATIO2 varied with city (Table 38).
Through these procedures, a distinct ACLV1 value was assigned to each site
for each standard under evaluation. This ACLV1 value was subsequently used to
construct an attainment one-hour data set using the procedures described in
Subsection 5.3.
5.3 Adjustment of One-Hour Ozone Data Sets
After a site's attainment ACLV1 value was determined, the baseline one-hour
data set associated with the site was adjusted hour-by-hour to create an attainment
one-hour data set. A two-stage adjustment procedure was employed. In the first
stage, the baseline one-hour data were adjusted to produce an initial attainment data
set that had the specified ACLV1 value. In the second stage, the initial data set was
"fine-tuned" to produce a final attainment data set having the exact AQI value
specified for the site.
5.3.1 Initial Adjustment for All Standards
The initial adjustment equation was
yc = (a) (xt)'° (56)
where x, was the baseline ozone concentration for hour t and yt was the attainment
ozone concentration for hour t. The terms a and b were "adjustment coefficients"
specific to the site and to the standard being attained.
The adjustment equation was based on the general assumption that Weibull
distributions would provide good fits to the one-hour data sets under baseline and
90
-------
attainment conditions. A Weibull distribution can be completely characterized through
the use of a shape parameter (k) and a scale parameter (8). The baseline values of k
and 5 were determined by applying a special maximum likelihood fitting algorithm to
each one-hour baseline data set. The attainment value of k (k1) was estimated by the
empirically-derived equation
l/k' = -0.2389 + (0.003367) (ACLVI) + (0.4726) (I/A:) (5?)
where ACLV1 was the estimated value of CLV1 under attainment conditions and k
was the baseline k value. The attainment value of 5 (§') was then determined by the
identity equation
5' = (ACL VI) /[In (n) J1'*1 (58)
where n was the number of one-hour values in the exposure period.
The unadjusted data set was treated as a time series where x, represented the
one-hour value at time t. The corresponding adjusted data set was constructed
through the use of the expression
yt = (5/) (xt/S)*/v (59)
where y, was the adjusted one-hour value at time t. This expression incorporates the
assumption that the time series y, at a site after attainment is related to the original
time series x, in such a way that 1) the rank of the one-hour value at each time t is
unchanged, 2) the x, values follow a Weibull distribution with parameters 5 and k, and
3) the y, values follow a Weibull distribution with parameters 8' and k'. These
assumptions are discussed in Appendix A. Equation 59 can be restated as Equation
56 above with the substitutions
a = (S7) /(S)*'* (60)
Jb - k/k1. (61)
91
-------
TABLE 38. VALUES FOR EQUIVALENCE RELATIONSHIPS
City
Chicago
Denver
Houston
Los Angeles
Miami
New York
Philadelphia
St. Louis
Washington
RATIO13
1.155
1.234
1.374
1.444
1.248
1.178
1.132
1.226
1.179
RATlO2b
1.441
1.453
2.091
1.846
1.513
1.436
1.367
1 .506
1.450
aRATIO1 = (ACLV1)/(ACLV8).
bRATIO2 = (ACLV1)/(EH6LDM).
5.3.2 Final adjustment for Eight-hour Standards
When applied to the 8H1EX standards, the initial adjustment procedure
described above produced a one-hour data set with a CLV1 value that exactly
matched the specified CLV1. Because the assumed relationship between CLV1 and
CLV8 was only an approximation, the CLV8 value of the adjusted data set did not
always match the attainment CLV8 value specified for the site. Consequently,
analysts made a final "fine-tuning" adjustment to the one-hour data to obtain the exact
CLV8 value specified. The following final adjustment equation was used.
Adjusted y, = (yt)(Target attainment CLV8)/(lnitial attainment CLV8) (62)
In this equation, y, is the one-hour value for hour t after the initial adjustment
procedure (Equation 56). The "initial attainment CLV8" is the CLV8 value of this data
set. The "target attainment CLV8" is the attainment CLV8 value assigned to the site
by the procedure summarized in Table 36.
A similar fine-tuning procedure was employed for the 8H5EX standards. The
final adjustment equation was
Adjusted y, = (y,)(Target attainment EH6LDM)/(lnitial attainment EH6LDM) (63)
92
-------
The "initial attainment EH6LDM" is the EH6LDM value of the site after the initial
adjustment (Equation 56). The "target attainment EH6LDM" is the attainment
EH6LDM value assigned to the site by the procedure summarized in Table 37.
5.4 Application of the AQAP's to Philadelphia
To test the reasonableness of the AQAP's described above, each was initially
applied to Philadelphia. Three attainment scenarios were evaluated:
1H1EX-120: One-hour daily maximum, one expected exceedance of 120 ppb
8H1EX-80: Eight-hour daily maximum, one expected exceedance of 80 ppm
8H5EX-80: Eight-hour daily maximum, five expected exceedances of 80 ppb.
In each case, baseline conditions were represented by filled-in 1991 ozone data
obtained from the 10 monitoring sites listed in Table 22.
5.4.1 Attainment of 1H1EX-120 Standard
The AQAP summarized in Table 35 was applied to Philadelphia for the purpose
of simulating the attainment of the 1H1EX-120 ppb standard. Table 39 presents the
results of each step. In this example, baseline conditions in Philadelphia were
assumed to be represented by 1991 ozone data as reported by the 10 monitoring
sites listed for Philadelphia in Table 22.
Analysts initiated the AQAP by fitting a Weibull distribution to the filled-in 1991
one-hour data set associated with each Philadelphia monitoring site. Each fit
produced estimates of the Weibull parameters (k and 5) and the CLV1. The largest
CLV1 for 1991 was associated with District 1 (167 ppb).
To exactly attain the specified NAAQS, the largest CLV1 must equal 120 ppb.
Consequently, Equation 51 (Step 3, Table 35) was implemented as
93
-------
TABLE 39. DETERMINATION OF ADJUSTMENT COEFFICIENTS FOR ONE-HOUR NAAQS
ATTAINMENT (1H1EX-120) IN PHILADELPHIA
District
1
2
3
4
5
6
7
8
9
10
Weibullfitto 1991 1-hrdata
k
1.69
2.21
1.96
1.81
2.28
2.23
1.93
2.14
1.74
2.26
5
46.9
56.4
51.0
49.3
56.6
51.2
44.3
51.2
38.1
54.5
CLV1
167
149
153
162
145
134
135
140
131
141
1-hr NAAQS attainment parameters3
Adjusted
CLV1
120
107
110
116
104
96
97
101
94
101
Reassigned
CLV1
107
110
116
120
104
101
94
101
96
97
k'
2.494
2.896
2.546
2.346
3.139
3.194
3.101
3.106
2.809
3.369
5'
45.27
52.44
49.95
48.09
52.51
51.60
47.06
50.62
44.74
51.31
Adjustment
coefficients
a
3.336
2.417
2.420
2.377
2.800
3.305
4.447
3.361
4.694
3.511
b
0.678
0.763
0.770
0.772
0.726
0.698
0.622
0.689
0.619
0.671
CD
aAssumes maximum CLV1 equals 120 ppb.
-------
ACLV1(i,j) = [CLV1(i,j)](120/167) = [CLV(i,j)](0.719). (64)
Applying this expression to each 1991 CLV1 produced 10 ACLVI's representing
attainment conditions. These values are listed in the column labeled "adjusted CLV1."
These values were then reassigned to the Philadelphia districts according to the five-
year ranking determined for each district. Thus, the largest adjusted CLV1 (120 ppb)
was assigned to District 4 because District 4 had the highest five-year ranking.
Similarly, the second largest adjusted CLV1 (116 ppb) was assigned to District 3
because District 3 had the second highest five-year ranking.
In this example, the five-year ranking of each site was determined by analyzing
second-high daily maximum one-hour ozone concentrations reported by the site over a
recent five-year period. Second-high daily maximum values were used in this step
rather than CLVI's because they were easier to obtain from standard EPA reports.
Analysts next used Equations 57 and 58 to estimate site-specific values for k'
and 5', the values of the Weibull parameters under attainment conditions. For District
1, the substitution of k = 1.69, ACLV1 = 107 ppb, and n = 5136 produced the
estimates k' = 2.494 and 8' = 45.27 ppb. These values were substituted into
Equations 60 and 61 to produce the values of the adjustment coefficients listed in
Table 39 for District 1 (a = 3.336 and b = 0.678).
A one-hour ozone data set representing attainment conditions was constructed
for each site by applying Equation 56 to the baseline one-hour data set for the site.
Table 40 provides descriptive statistics for'the baseline and attainment data sets
associated with District 1.
5.4.2 Attainment of 8H1EX-80 Standard
To evaluate the AQAP for 8H1EX standards, the procedure summarized in
Table 36 was applied to Philadelphia for the purpose of simulating the attainment of
the 8H1EX-80 standard. The results are presented in Table 41. As in the previous
example, baseline conditions for Philadelphia were represented by 1991 ozone data.
95
-------
TABLE 40. DESCRIPTIVE STATISTICS FOR HOURLY-HOUR DATA (PPB)
FOR SITE 34-005-3001 (DISTRICT 1, PHILADELPHIA): BASELINE AND
ATTAINMENT OF THREE OZONE STANDARDS
Statistic
Number of values
Mean
Standard deviation
Minimum
5th percentile
10th percentile
25th percentile
50th percentile
75th percentile
90th percentile
95th percentile
99th percentile
99.5 percentile
99.8 percentile
99.9 percentile
Maximum
Baseline
5136
38
25
0
4
8
19
34
51
72
87
117
124
137
143
156
Attainment of indicated standard
1H1EX-120
5136
37
18
0
9
14
25
36
48
61
69
84
88
94
97
102
8H1EX-80
5136
36
14
0
12
17
27
36
45
54
60
71
73
77
78
82
8H5EX-80
5136
34
16
0
7
13
23
34
43
54
61
74
77
82
84
89
Analysts initiated the AQAP by fitting a Weibull distribution to the filled-in 1991
one-hour data set associated with each Philadelphia monitoring site. Each fit
produced estimates of the Weibull parameters (k and 6) and the CLV1. As in the
previous example, the largest CLV1 for 1991 was associated with District 1 (167 ppb).
Analysts next estimated a baseline CLV8 for each site by fitting a Weibull
distribution to the running-average eight-hour data associated with each Philadelphia
monitoring site. The largest CLV8 was 142 ppb (District 1).
To exactly attain the specified NAAQS, the largest CLV8 must equal 80 ppb.
Consequently, Equation 52 (Step 3, Table 36) was implemented as
ACLV8(i,j) = [CLV8(i,j)](80/142) = [CLV(i,])] (0.563). (65)
96
-------
TABLE 41. DETERMINATION OF ADJUSTMENT COEFFICIENTS FOR EIGHT-HOUR NAAQS
ATTAINMENT (8H1EX-80) IN PHILADELPHIA
District
1
2
3
4
5
6
7
8
9
10
Weibull fits to 1991 data
1-hk
1.69
2.21
1.96
1.81
2.28
2.23
1.93
2.14
1.74
2.26
1-h6
46.9
56.4
51.0
49.3
56.6
51.2
44.3
51.2
38.1
54.5
CLV1
167
149
153
162
145
134
135
140
131
141
CLV8
142
136
128
138
120
118
123
128
116
126
8-hr NAAQS attainment parameters*
Adjusted
CLV8
80
77
72
78
68
66
69
72
65
71
Reassigned
CLV8
72
77
78
80
72
69
65
71
66
68
Equivalent
CLV1
82
87
88
91
82
78
74
80
75
77
1-h Weibull
parameters
k'
3.173
3.725
3.339
3.057
4.119
4.237
3.941
3.960
3.518
4.359
6'
41.45
49.01
46.44
44.89
48.41
47.08
42.70
46.75
40.60
47.06
Adjustment
coefficients
a
5.339
4.481
4.618
4.465
5.183
5.932
6.671
5.572
6.708
5.922
b
0.533
0.593
0.587
0.592
0.554
0.526
0.490
0.540
0.495
0.518
CO
"Assumes maximum CLV8 equals 80 ppb.
-------
Analysts applied this expression to each 1991 CLV8 to obtain 10 ACLVS's
representing attainment conditions. These values are listed in the column labeled
"adjusted CLV8." These values were then reassigned to the Philadelphia districts
according to the five-year ranking determined for each district. The resulting
assignments are listed in Table 41 under the heading "reassigned CLV8."
Each reassigned CLV8 was then converted into an equivalent attainment CLV1
using Equation 54 with the RATIO1 value for Philadelphia (1.132). For example, the
reassigned CLV8 for District 1 (72 ppb) was multiplied by 1.132 to produce an
equivalent attainment CLV1 of 82 ppb.
Analysts next used Equations 57 and 58 to estimate site-specific values for k1
and 6', the values of the Weibull parameters for one-hour data under attainment
conditions. For District 1, the substitution of k = 1.69, ACLV1 = 82 ppb, and n = 5136
produced the estimates k1 = 3.173 and 6' = 41.45 ppb. These values were substituted
into Equations 60 and 61 to produce the values of the adjustment coefficients listed in
Table 41 for District 1 (a = 5.339 and b = 0.533). These coefficients were then
substituted into Equation 56 to produce an initial one-hour data set approximating
attainment conditions.
The one-hour data were processed to produce a corresponding 8-hour running
average data set. A Weibull distribution was next fit to the adjusted eight-hour data
for the site to determine an initial attainment CLV8. Analysts then used Equation 62
to make the final "fine-tuning" adjustment to the one-hour data necessary to achieve
the target CLV8 specified for the site (72 ppb). The resulting one-hour data set was
assumed to represent attainment conditions for District 1. Table 40 provides
descriptive statistics for this data set. Attainment data sets were developed in a
similar manner for each of the other Philadelphia monitoring sites.
5.4.3 Attainment of 8H5EX-80 Standard
The AQAP for 8H5EX standards (Table 37) was applied to Philadelphia for the
purpose of simulating the attainment of the 8H5EX-80 standard. The results are
presented in Table 42.
98
-------
As in the two previous examples, baseline conditions for Philadelphia were
represented by 1991 ozone data. Analysts began the AQAP by fitting a Weibull
distribution to the filled-in 1991 one-hour data set associated with each Philadelphia
monitoring site. Each fit produced estimates of the Weibull parameters (k and 5) and
the CLV1. The largest CLV1 for 1991 was associated with District 1 (167 ppb).
Analysts next determined a baseline EH6LDM value for each site by first
calculating all eight-hour daily maximum concentrations in the associated one-hour
data set and then identifying the sixth largest value. The largest EH6LDM was 116
ppb (District 1).
The largest EH6LDM value permitted under the 8H5EX-80 standard is 80 ppb.
As the largest baseline EH6LDM was 116 ppb, Equation 53 (Table 37) was expressed
as
AEH6LDM(i,j) = [EH6LDM(i,j)](80/116) = [EH6LDM(i,j)](0.563) (66)
Analysts applied this expression to each 1991 EH6LDM to obtain 10 AEH6LDM's
representing attainment conditions. These values are listed in the Table 42 column
labeled "adjusted EH6LDM." Analysts next reassigned the values to the Philadelphia
districts according to the five-year ranking determined for each district. The resulting
assignments are listed in Table 42 under the heading "reassigned EH6LDM."
Each reassigned EH6LDM was then converted into an equivalent attainment
CLV1 using Equation 55 with the RATIO2 value for Philadelphia (1.367). In the case
of District 1, the reassigned EH6LDM (74 ppb) was multiplied by 1.367 to produce an
equivalent attainment CL.V1 of 101 ppb.
Analysts next used Equations 57 and 58 to estimate site-specific values for k'
and 5', the values of the Weibull parameters for one-hour data under attainment
99
-------
TABLE 42. DETERMINATION OF ADJUSTMENT COEFFICIENTS FOR EIGHT-HOUR NAAQS
ATTAINMENT (EH6LDM = 80 ppb) IN PHILADELPHIA
District
1
2
3
4
5
6
7
8
9
10
Parameters of 1991 data
1-hk
1.69
2.21
1.96
1.81
2.28
2.23
1.93
2.14
1.74
2.26
1-h6
46.9
56.4
51.0
49.3
56.6
51.2
44.3
51.2
38.1
54.5
CLV1
167
149
153
162
145
134
135
140
131
141
EH6LDM
116
113
111
115
107
101
102
104
90
102
8-hour NAAQS attainment parameters
Adjusted
EH6LDM
80
78
77
79
74
70
70
72
62
70
Rank
5
2
1
3
4
6
8
9
10
7
Reassigned
EH6LDM
74
79
80
78
77
72
70
70
62
70
Equivalent
CLV1
101
108
109
107
105
98
96
96
85
96
1-hour Weibull
parameters
k'
2.626
2.954
2.708
2.615
3.106
3.300
3.038
3.277
3.136
3.408
6'
44.62
52.24
49.37
47.10
52.64
51.16
47.38
49.88
42.89
51.15
Adjustment
coefficients
a
3.750
2.556
2.869
3.171
2.721
3.581
4.261
3.817
5.689
3.608
b
0.644
0.748
0.724
0.692
0.734
0.676
0.635
0.653
0.555
0.663
8
'Assumes maximum EH6LDM equals 80 ppb.
-------
conditions. For District 1, the substitution of k = 1.69, ACLV1 = 101 ppb, and n =
5136 produced the estimates k1 = 2.626 and 5' = 44.62 ppb. These values were
substituted into Equations 60 and 61 to produce the values of the adjustment
coefficients listed in Table 42 for District 1 (a = 3.750 and b = 0.644). These
coefficients were then substituted into Equation 56 to produce an initial one-hour data
set approximating attainment conditions.
The one-hour data were processed to produce a corresponding 8-hour running
average data set. These data were analyzed to determine an initial attainment
EH6LDM. Analysts then employed Equation 63 to make the final "fine-tuning"
adjustment to the one-hour data necessary to achieve the target attainment EH6LDM
specified for the district (101 ppb). The resulting tuned data set was assumed to
represent attainment conditions for District 1. Table 40 presents descriptive statistics
for this data set. Attainment data sets were developed in a similar manner for each of
the other Philadelphia monitoring sites.
5.5 Special Adjustment Procedures Applied in Selected Attainment Scenarios
The AQAP's described above were developed by comparing the ozone data
reported by a site in a high ozone year with ozone data reported by the same site in a
low ozone year. Consequently, the AQAP's are expected to perform best when used
to simulate a significant reduction in the ozone levels at a site. The results of an
analysis of AQAP performance by ITAQS suggested that the AQAP's described above
may produce unrealistic data sets for Denver, Chicago, and Miami when used to
simulate a small reduction in ozone levels or when used to simulate an increase in
ozone levels. For this reason, ITAQS used a different set of AQAP's for the following
combinations of study areas and attainment scenarios:
Study area Attainment scenarios
Denver 1H1EX-120 and 8H1EX-100
Chicago 1H1EX-120 and 8H1EX-100
Miami all (1H1 EX-120, 8H1 EX-80, 8H1 EX-100, 8H5EX-60, and
8H5EX-80)
101
-------
The Denver scenarios listed above required slight increases in ozone levels to exactly
meet the specified attainment conditions. The Chicago scenarios required small
increases in ozone levels. The Miami scenarios required small changes in both
directions.
In the alternative AQAP for 1H1EX-120, the procedures summarized in Table
35 were followed to the point in Step 5 where the reader is directed to Section 5.3.
The procedures in Section 5.3 were not employed to adjust the one-hour data;
instead, each value of the adjusted data set was estimated by the expression
Y, = (c)(xt) (67)
where x, was the baseline ozone concentration for hour t and y, was the attainment
ozone concentration for hour t. The value of c was determined by the expression
c = (ACLV1)/(CLV1) (68)
where ACLV1 is the characteristic largest one-hour value of the site before adjustment
and ACLV1 is the characteristic largest one-hour value assigned to the site in Step 4
to represent attainment conditions.
In a similar manner, the alternative AQAP's for 8H1EX-80 and 8H1EX-100
followed the procedures summarized in Table 36 to the point in Step 6 where the
reader is directed to Section 5.3. Again, the procedures in Section 5.3 were not
employed to adjust the one-hour data. Instead, an initial estimate of each value of the
adjusted data set was estimated according to Equations 67 and 68. In this case,
ACLV1 was the characteristic largest one-hour value assigned to the site in Step 5 of
Table 36. The adjustment procedure was completed by applying Equation 61 to the
data to make a final "fine-tuning" adjustment.
The alternative AQAP's for 8H5EX-60 and 8H5EX-80 followed the steps listed
in Table 37 to the point in Step 6 where the reader is directed to Section 5.3. The
applicable procedures in Section 5.3 were again omitted; instead, Equations 67 and 68
were employed to make an initial estimate of each value of the adjusted data set. In
this procedure, ACV1 was the characteristic largest one-hour value assigned to the
102
-------
site in Step 5 of Table 37, The adjustment procedure was completed by using
Equation 62 to make the final fine tuning adjustment.
103
-------
SECTION 6
OZONE EXPOSURE ESTIMATES FOR NINE URBAN AREAS
To illustrate the capabilities of the updated pNEM/OS methodology described in
this report, the program was applied to the nine urban areas listed earlier in Table 1.
The result of each application was a set of 36 exposure summary tables for each
regulatory scenario under evaluation. This section describes the scenarios that were
analyzed, provides a guide to the interpretation of output tables, and summarizes the
principal results of each exposure assessment.
6.1 Regulatory Scenarios
The following regulatory scenarios were examined in applying pNEM/O3
to each study area.
Baseline: Ambient ozone conditions were represented by unadjusted fixed-
site monitoring data as reported for the exposure period listed in
Table 1. These data were assumed to represent ambient
(outdoor) ozone levels typical of "as is" air quality conditions
unaffected by nearby sources.
1H1EX-120: The baseline monitoring data were adjusted to simulate the
attainment of a 1H1EX standard (see Section 5) permitting one
expected exceedance of 120 ppb (0.12 ppm). This standard is
identical to the current NAAQS for ozone.
8H1EX-100 The baseline monitoring data were adjusted to simulate attainment
of an 8H1EX standard (see Section 5) permitting one expected
exceedance of 100 ppb (0.10 ppm).
8H1EX-80 The baseline monitoring data were adjusted to simulate attainment
of an 8H1EX standard permitting one expected exceedance of 80
ppb (0.08 ppm).
104
-------
8H5EX-80 The baseline monitoring data were adjusted to simulate attainment
of an 8H5EX standard permitting five expected exceedances of 80
ppb (0.08 ppm).
8H5EX-60 The baseline monitoring data were adjusted to simulate attainment
of an 8H5EX standard permitting five expected exceedances of 60
ppb (0.06 ppm).
Section 5 describes the procedures used to adjust baseline data to simulate
attainment of 1H1EX, 8H1EX, and 8H5EX standards.
6.2 Formats of the Exposure Summary Tables
The application of pNEM/OS to a study area produced two sets of 18 exposure
summary tables - one set listing exposure estimates for the general population and
one set listing exposure estimates for children only. The general population included
all cohorts, regardless of demographic group. Children were defined as the population
subgroup containing all cohorts associated with Demographic Groups 1, 2, and 3
(Table 2).
Appendix C contains exposure summary tables for the general population and
for children obtained from an sample application of pNEM/O3 to Houston. Each set
contains one or more tables organized according to the following table formats. (Note
that the table numbers listed under each format refer to the tables in Appendix C.)
Number of people -- cumulative exposures (or doses) by EVR range
These tables list estimates by ozone concentration and EVR range. Each table
entry lists the number of people who experienced one or more ozone exposures (or
doses) during which the ozone concentration was at or above the level indicated by
the row label and the average EVR was within the range indicated by the column
heading. Separate tables provide estimates for one-hour exposures (Table 1 in
Appendix C), one-hour daily maximum exposures (Table 1A), one-hour daily maximum
doses (Table 1B), eight-hour daily maximum exposures (Table 4), and eight-hour daily
maximum doses (Table 4A).
105
-------
Number of people - cumulative seasonal mean exposures
Table 7 in Appendix C lists estimates by ozone concentration only. Each entry
lists the number of people who were associated with a seasonal mean exposure at or
above the ozone level indicated by the row label. The seasonal mean is calculated as
the average of the eight-hour daily maximum ozone exposures occurring from April to
October, inclusive.
Number of occurrences -- exposures for doses) by EVR range
These tables list estimates arranged by ozone concentration range and EVR
range. Each table entry lists the number of times a person experienced an ozone
exposure during which the ozone concentration was within the range indicated by the
row label and the average EVR was within the range indicated by the column heading.
There are separate tables for one-hour exposures (Table 2 in Appendix C), one-hour
daily maximum exposures (Table 2A), one-hour daily maximum doses (Table 2B),
eight-hour daily maximum exposures (Table 5), and eight-hour daily maximum doses
(Table 5A).
Number of occurrences -- seasonal mean exposures
Table 8 in Appendix C presents estimates by ozone range only. Each entry
lists the number of times a person experienced a seasonal mean exposure at or
above the ozone level indicated by the row label. The seasonal mean is calculated as
the average of the eight-hour daily maximum ozone exposures occurring from April to
October, inclusive.
Number of people -- highest exposures (or doses) by EVR range
Each of these tables lists estimates arranged by ozone concentration and EVR
range. Each entry indicates the number of people who experienced their maximum
ozone exposure under conditions in which the ozone concentration was at or above
the level indicated by the row label and the average EVR was within the range
indicated by the column heading. There are separate tables for one-hour daily
maximum exposures (Table 3 in Appendix C) and eight-hour daily maximum
exposures (Table 6).
Number of people -- cumulative daily maximum doses by number of days
These tables provide estimates arranged by ozone concentration and number
of days per year. Each entry lists the number of> people who experienced a daily
maximum dose at or above the indicated ozone concentration for the specified number
of days. Separate tables are provided for daily maximum one-hour doses (Table 9 in
Appendix C), daily maximum eight-hour doses (Table 10), daily maximum one-hour
106
-------
doses with EVR of 30 liters x min"1 x m'2 (Table 11), and daily maximum eight-hour
doses with EVR of 15 liters x min"1 x m"2 (Table 12).
Regardless of format, each table in Appendix C provides footnotes identifying
the study area, regulatory scenario, and the population group analyzed. The footnotes
also indicate the number of exposure districts in the study area, the first and last days
of the ozone season, and the number of days in the ozone season.
6.3 Results of Analyses
Tables 43 and 44 summarize exposure estimates for one run of the model for
the total population within each of the nine study areas according to the six regulatory
scenarios that were analyzed. Table 43 lists the percentage of each study area
population that was estimated to experience one or more daily maximum one-hour
exposures above 0.12 ppm at any ventilation rate. This ozone concentration
corresponds to the level of the current standard. Exposures above this level may
have adverse health effects. The following general statements apply to the results
presented in Table 43.
1. Of the nine study areas, Houston has the highest percentage (97.62
percent) of people experiencing one-hour daily maximum ozone expo-
sures above 0.12 ppm under baseline conditions. In the scenario where
the current NAAQS is exactly met, this percentage drops to 6.6 percent.
2. Of the nine study areas, Miami has the lowest percentage (0.59 percent)
of persons experiencing one-hour daily maximum ozone exposures
above 0.12 ppm under baseline conditions. In the scenario where Miami
exactly meets the current NAAQS, the number of persons exposed
above 0.12 ppm increases to 6.62 percent. Under baseline conditions,
the ozone levels in Miami are lower than those permitted by the current
NAAQS; consequently, the adjustment of baseline data to exactly meet
the current NAAQS produced an increase in ozone exposure.
3. In all study areas except St. Louis, the number of persons exposed to
levels above 0.12 ppm is larger under the 8H1EX-100 standard than
under the current NAAQS. In general, the 8H1EX-100 standard appears
to be less stringent than the current NAAQS with respect to this
exposure indicator.
107
-------
TABLE 43. NUMBER AND PERCENT OF TOTAL STUDY AREA POPULATION
EXPERIENCING ONE OR MORE ONE-HOUR DAILY MAXIMUM OZONE
EXPOSURES ABOVE 120 PPB AT ANY VENTILATION RATE
Study
Area
Chicago
Denver
Houston
Los
Angeles
Number of
Persons
at Risk
6,175,121
1 ,484,798
2,370,512
10,371,115
Regulatory
Scenario
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Number of
Persons
Exposed
1 ,726,506
448,269
1,191,553
0
4,434
0
195,248
166,543
733,869
0
0
0
2,314,143
156,483
670,287
35,718
165,597
0
8,454,573
117,998
1 ,020,660
27,620
173,626
0
Percent
of Total
27.96
7.26
19.30
0
0.07
0
13.15
11.22
49.43
0
0
0
97.62
6.60
28.28
1.51
6.99
0
81.52
1.14
9.84
0.27
1.67
0
108
-------
TABLE 43 (Continued)
Study
Area
Miami
New York
Philadel-
phia
St.
Louis
Number of
Persons
at Risk
1 ,941 ,994
10,657,873
3,785,810
1 ,706,778
Regulatory
Scenario
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Number of
Exposed
Persons
11,374
128,639
486,708
16
202,889
5,685
6,284,728
218,380
643,330
0
97,403
0
3,201,816
78,588
148,088
0
0
0
271,580
67,153
33,980
0
14,585
0
Percent
of Total
0.59
6.62
25.06
0
10.45
0.29
58.97
2.05
6.04
0
0.91
0
84.57
2.08
3.91
0
0
0
15.91
3.93
1.99
0
0.85
0
109
-------
TABLE 43 (Continued)
Study
Area
Washing-
ton, DC
Number of
Persons
at Risk
3,085,419
Regulatory
Scenario
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Number of
Persons
Exposed
2,251,949
61,491
136,125
0
72,516
0
Percent
of Total
72.99
1.99
4.41
0
2.35
0
110
-------
TABLE 44. PERCENT OF TOTAL STUDY AREA POPULATION
EXPERIENCING ONE OR MORE 8-HOUR DAILY MAXIMUM OZONE EXPOSURES
ABOVE INDICATED EXPOSURE CONCENTRATIONS AT ANY VENTILATION RATE
Study Area
Chicago
Denver
Houston
Los
Angeles
Regulatory
Scenario
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Number of
Persons at
Risk
6,175,121
1 ,484,798
2,370,512
10,371,115
Percentage of Population Experiencing 8-Hour
Daily Maximum Ozone Exposure Above
Indicated Concentration
0,06 ppm
96.65
82.81
94.09
72.40
83.66
4.43
77.53
80.34
96.96
68.21
65.72
2.74
99.40
64.34
82.39
33.93
59.59
4.11
96.00
23.51
32.83
16.07
13.05
1.51
0.08 ppm
33.01
13.35
30.57
0.36
4.15
0
9.28
8.81
41.05
3.40
1.48
0
90.71
8.91
26.05
1.75
5.90
0
69.77
0.86
7.87
0.17
0.99
0
0.10 ppm
1.06
0
0.12
0
0
0
0
0.10
6.09
0
0
0
52.09
0
1.08
0
0.01
0
41.06
0
1.05
0
0
0
(continued)
111
-------
TABLE 44 (Continued)
Study Area
Miami
New York
Philadel-
phia
St. Louis
Regulatory
Scenario
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Number of
Persons at
Risk
1,941,994
10,657,873
3,785,810
1 ,706,778
Percentage of Population Experiencing 8-Hour
Daily Maximum Ozone Exposure Above
Indicated Concentration
0.06 ppm
14.39
58.29
74.75
32.32
67.22
6.05
85.81
53.67
75.58
29.87
33.31
1.75
99.96
97.72
99.75
57.31
79.97
2.31
73.22
86.86
80.75
34.07
65.27
5.69
0.08 ppm
0.01
0.71
18.09
0
4.86
0
54.57
7.67
17.43
0.79
2.36
0
90.98
33.28
31.69
1.02
7.59
0
23.54
24.89
18.58
0.21
3.83
0
0.10 ppm
0
0.03
0
0
0.04
0
21.10
0
0.36
0
0
0
36.19
1.25
0.71
0
0
0
0.82
0.02
0.04
0
0
0
(continued)
112
-------
TABLE 44 (Continued)
Study Area
Washington,
DC
Composite
Regulatory
Scenario
Baseline
Current NAAQS
8H1EX-100
8H1EX-80
8H5EX-80
8H5EX-60
Baseline
Current NAAQS
8H1EX-100
8H1 EX-80
8H5EX-80
8H5EX-60
Number of
Persons at
Risk
3,085,419
Not
Applicable
Percentage of Population Experiencing 8-Hour
Daily Maximum Ozone Exposure Above
Indicated Concentration
0.06 ppm
97.67
94.14
93.33
45.08
79.02
6.00
96.00
80.34
82.39
34.07
65.72
4.11
0.08 ppm
77.99
15.30
18.67
0.45
7.29
0
54.57
8.91
18.67
0.45
4.15
0
0.10 ppm
21.53
0.13
0.27
0
0
0
21.10
0.02
0.36
0
0
0
113
-------
4. In all nine study areas, the number of persons exposed to levels above
0.12 ppm is smaller under the 8H1EX-80 standard than under the current
NAAQS. The 8H1EX-80 standard appears to be generally more
stringent than the current NAAQS with respect to this exposure indicator.
5. The exposure estimates for five of the nine study areas indicate that the
number of persons exposed to ozone levels above 0.12 is smaller under
the 8H5EX-80 standard than the current NAAQS. In the other four cities
(Houston, Los Angeles, Miami, and Washington, DC) *he number of
persons exposed under the 8H5EX-80 standard is greater than under the
current NAAQS.
6. In eight of the nine study areas, the number of persons exposed to levels
above 0.12 ppm is zero under the 8H5EX-60 standard. The one
exception is Miami, whose exposed population percentage is only 0.29
percent. This standard appears to be significantly more stringent than
the current NAAQS. Between 1.14 percent and 11.22 percent of each
study area population is exposed to levels above 0.12 ppm under the
current NAAQS.
Table 44 lists the percentage of each study area population that was estimated to
experience one or more 8-hour daily maximum exposures above 0.06 ppm, 0.08 ppm,
and 0.10 ppm, respectively, at any ventilation rate. These levels of exposure were
chosen for the analysis because they are the levels of possible alternative 8-hour
standards. Eight-hour exposures above these levels should have greater health
effects than 1-hour exposures at the same levels. The following general statements
apply to the results presented in Table 44.
1. Under baseline conditions, Philadelphia has the largest percentage of
people experiencing 8-hour daily maximum exposures above 0.06 ppm
(99.96 percent) and above 0.08 ppm (90.98 percent). When exposures
above 0.10 ppm are considered, however, Houston has the largest
percentage (52.09). The corresponding estimate for Philadelphia is
36.19 percent. The upper tail of the exposure distribution is more
extended for Houston than for Philadelphia.
2. Under the current NAAQS, Philadelphia has the largest percentage of
people experiencing 8-hour daily maximum exposures above 0.06 ppm
(97.72 percent), above 0.08 ppm (33.28 percent), and above 0.10 ppm
(1.25 percent).
114
-------
3. According to the exposure estimates listed for the 8H1EX-100 standard,
Philadelphia has the largest percentage of exposures above 0.06 ppm
(99.75 percent). Denver has the largest percentage above 0.08 ppm
(41.05 percent) and above 0.10 ppm (6.09 percent).
The estimates in Table 44 labeled "composite" are medians of the
corresponding values listed in Table 44 for the nine cities. For example, the
composite value listed in the 0.06 ppm column for the current NAAQS (80.34 percent)
is the median of the nine city-specific values listed in the 0.06 ppm column for the
current NAAQS. The composite estimates may be interpreted as representing the
exposure distributions expected in a typical city under each standard. For this typical
city, the five standards would be arranged as follows with respect to the percentage of
people with eight-hour exposures above 0.06 ppm.
1. 8H1EX-100 (82.39 percent)
2. Current NAAQS (80.34 percent)
3. 8H5EX-80 (65.72 percent)
4. 8H1EX-80 (34.07 percent)
5. 8H5EX-60 (4.11 percent)
The five standards would have the following ordering if arranged according to the
percentage of people with eight-hour exposures above 0.08 ppm.
1. 8H1 EX-100 (18.67 percent)
2. Current NAAQS (8.91 percent)
3. 8H5EX-80 (4.15 percent)
4. 8H1EX-80 (0.45 percent)
5. 8H5EX-60 (0 percent)
The orderings of these two sets of rankings are identical.
Figures 2, 3, and 4 provide graphical presentations of selected results from
Tables 43 and 44. Figure 2 shows the estimated percentage of Houston-area
residents who experience one-hour daily maximum ozone exposures above 0.12 ppm
(120 ppb) according to each regulatory scenario. Figure 3 compares all nine study
areas with respect to.the estimated percentage of residents who experience one-hour
ozone exposures above 0.12 ppm under each scenario. Figure 4 displays the
estimated percentage of Houston residents who experienced eight-hour daily
maximum ozone exposures above 0.80 ppm under each regulatory scenario.
115
-------
CD
§: 100
o
C\J
I 80
T3
8 3
0 n;
^ C 60
0 B
~m ro
"5 1
o ^ 40
"o ^
CD ^
CD 03 on
£j ^U
"c
CD
e
CD
°- 0
97.6
-
-
j
28.3
17.0
1-5 n
--•.;.•.-.•.•,-.•.-.•.-.•.-.-.•.•.-.-.-.•.-.- . .- . . ' -,y.. ?•>$$$%$£
Regulatory Scenario
D Base Case
• Current NAAQS
H8H1EX-100
H 8H1EX-80
• 8H5EX-80
H 8H5EX-60
Figure 2. Estimated percentage of residents in Houston study area who experience one-hour daily maximum ozone
exposures above 0.12 ppm according to regulatory scenario.
-------
ft
0)
Q_
100 -
2 80
•s «
0) CT3
O) --
S Q.
C CL
8 -
C\|
13 8H5EX-60
H 8H5EX-80
d 8H1EX-80
H 8H1EX-100
m Current NAAQS
D Base Case
o
D)
03
o
In
O
CD
C
0
Q
c
-2
"(/)
Z3
O
CD
CD
C
O
E
03
CD
03
!c
"o3
T3
_03
!c
Q_
q
d
c
o
-+—'
CD
C
Jz
w
03
^^^
Figure 3. Estimated percentage of residents in each study area who experience one-hour daily maximum ozone
exposures above 0.12 ppm according to regulatory scenario.
-------
-Q
0.
CL
O
00
O)
C
T3
0
100
CD
0
O CO
x (Y
LJJ "•
B
-*-•
o
(D ^
0) CD
CD
-*— *
C
CD
O
o3
CL
80
60
40
20
0
90.7
26.1
D Base Case
m Current NAAQS
B 8H1EX-100
• 8H1EX-80
M 8H5EX-80
H 8H5EX-60
Regulatory Scenario
Figure 4. Estimated percentage of residents in Houston study area who experience eight-hour daily maximum
ozone exposures above 0.08 ppm according to regulatory scenario.
-------
SECTION 7
INITIAL EFFORTS TO VALIDATE THE EXPOSURE MODEL
The validity of an exposure model such as pNEM/03 can be best evaluated by
comparing model exposure estimates for a variety of urban areas with actual personal
exposure measurements collected in the same areas under similar conditions. As the
required multi-city personal monitoring data base did not exist at the time of this study,
researchers were not able to perform a comprehensive validation effort. Validation
was limited to a comparison of pNEM/O3 estimates for Houston, Texas, with personal
monitoring data provided by Dr. Thomas Stock. These data were collected in 1981
during the Houston Asthmatic Study (HAS)20. This section provides a brief overview
of the HAS data file and summarizes the results of initial efforts to validate pNEM/OS.
7.1 The HAS Data
The main HAS data file contains time/activity data and measurements of
personal ozone exposure for 30 subjects who resided in two Houston neighborhoods
(Clear Lake and Sunnyside). The data for each subject documents one or two
daytime periods which typically began when the subject awoke in the morning and
ended around 6 pm. Personal ozone measurements were generally recorded at five-
minute intervals during this period. Each record in the data file represents a new
ozone measurement or a change in the subject's activity or location. Each record
contains the following data items:
Subject identification code
Month
Day
Year
119
-------
Hour of the day
Minute of the hour
Identification code of residence being monitored
Identification code of subject's residence
Microenvironment
Macroenvironment (either "inside study area" or "outside study area")
Air conditioning code
Smoking code
Level of exertion
Cooking code
Cooking fuel
Intensity of traffic
Some of the records contained the following information:
Duration of subject in microenvironment
Ozone concentration recorded by personal monitor
Ozone quality code
One-hour ozone concentration of Clear Lake fixed-site monitor
One-hour ozone concentration of Sunnyside fixed-site monitor
In addition, ITAQS received two data files containing the one-hour fixed-site ozone
concentrations for Clear Lake and Sunnyside. Background data were also provided to
ITAQS concerning each subject's age, occupation, and residential location.
7.2 The Special Version of pNEM/O3
ITAQS developed a special version of pNEM/O3 (designated pNEM/O3-S) and
applied it to Houston in an attempt to simulate the conditions of HAS. In this
application, the general modeling approach described in Section 2 was followed with
the following exceptions. The exposure period was changed from 1990 to the period
of the HAS diary data (June 24 to October 31, 1981). Whenever pNEM/O3-S
algorithms required fixed-site ozone concentrations, these values were determined
using fixed-site data from this 130-day period. In a similar manner, temperature and
calendar data for this period were used whenever these data were required by the
model. For example, the exposure event sequence for each cohort was generated
using temperature and weekday/weekend calendar data specific to the HAS period.
120
-------
In applying pNEM/03-S to Houston, each HAS subject was treated as a distinct
cohort. The data available for each subject were reviewed to determined the subject's
demographic group, residential air conditioning status, home district, and work district.
The study area was assumed to contain three exposure districts: 1) the Clear Lake
neighborhood, 2) the Sunnyside neighborhood, and 3) all other areas of Houston.
Ambient ozone concentrations for Clear Lake and Sunnyside were determined using
ozone data reported by fixed-site monitors in those areas. Ambient ozone
concentrations for the remaining district (i.e., all other areas) were determined by
averaging ozone data reported by all Houston monitors.
All of the remaining pNEM/O3-S algorithms were identical to the corresponding
algorithms in pNEM/O3. In particular, the algorithms used to generate exposure event
sequences, EVR values, air exchange rates, and indoor ozone concentrations (the
mass balance model) were identical to those described in earlier sections of this
report. A run of pNEM/03-S produced cohort-specific exposure files for the June 24-
to-October 31 period which could be compared directly with the HAS data.
7.3 Processing of HAS Data
The HAS data used in the validation effort consisted of the time/activity (diary)
data for each subject, personal ozone exposure data for each subject, background
data for each subject, and hourly-average ozone data obtained from fixed-site
monitors in Clear Lake and Sunnyside. These data were processed by ITAQS to
produce a special data file which could be compared directly with pNEM/O3-S
exposure estimates.
The time/activity data associated with each subject were transformed from a
series of records typically occurring at five-minute intervals to a series containing one
record for each discrete exposure event. A discrete exposure event was assumed to
begin whenever the diary data for a subject indicated a change in activity,
microenvironment, or clock hour.
The personal ozone exposure data associated with each exposure event were
averaged to produce an average ozone exposure for the event. If more than 15
121
-------
minutes of exposure data were missing for a particular event, the ozone exposure was
coded as missing. The event-specific exposures were averaged by clock hour (e.g, 2
pm to 3 pm) to produce a file containing one-hour exposures for each subject. Hours
with less than 45 minutes of ozone exposure data were coded as missing.
The hourly-average ozone data for the Clear Lake and Sunnyside fixed-site
monitors were incomplete in that ozone concentrations were missing for a number of
time intervals. Statistical analyses of the existing data for the two sites indicated a
strong correlation between the ozone concentrations reported by the sites for the
same time interval. Consequently, regression analyses were performed on the two
data sets and the resulting regression equations were used to estimate missing
values. The regression equations were
CL(t) = 10.43 ppb + (0.716)[SS(t)] (69)
SS(t) = 1.22 ppb + (0.864)[CL(t)]. (70)
In these equations, CL(t) and SS(t) indicate the one-hour ozone concentrations in
Clear Lake and Sunnyside, respectively, for time interval t. The R2 value for each
regression equation is 0.618.
Equation 69 was used to estimate missing Clear Lake values; Equation 70 was
used for missing Sunnyside values. In a few cases, values were missing in both data
sets for the same time interval. The time series procedure discussed in Subsection
4.2 was used to estimate these values.
In the validation process, each HAS subject was modeled as a distinct cohort
by pNEM/O3-S. Consequently, each HAS subject had to be classified according to
demographic group (Table 2), home district, work district (if applicable), and air
conditioning status. Data from the subject's background questionnaire were used to
determine demographic group and home district (Clear Lake or Sunnyside). The
subject's work district was determined by reviewing his or her diary data. If the
subject's activities indicated that he worked at a location coded in the diary as outside
the study area (Macroenvironment = 2), then the work district was designated as
District 3 (other Houston areas). Otherwise, the work district was assigned to the
122
-------
subject's home neighborhood. The subject's air conditioning status was determined by
examining diary entries concerning the type of air conditioning system in use during
periods when the subject was in his home microenvironment.
7.4 Comparison of Measured and Estimated Exposures by Person-hour
To evaluate the run-to-run variability in the exposure estimates produced by
pNEM/03-S, researchers executed the program ten times. Each run produced a 130-
day exposure sequence for each cohort. As discussed above, each cohort in
pNEM/O3-S corresponded to one of the HAS subjects. Consequently, researchers
were able to identify a cohort and a time period in each pNEM/O3-S output that
corresponded to a subject and a time period in the HAS data set. The hourly-average
(one-hour) ozone exposure estimates produced for each designated combination of
cohort and time period were placed in a "person-hour" file specific to the pNEM/O3-S
run which produced the estimates.
Table 45 presents selected percentiles of the one-hour exposure estimates in
the person-hour file produced by each of the 10 runs. The table also presents these
percentiles for the file containing the measured one-hour exposures of the HAS
subjects. Because of the matchup procedure, each of the 10 pNEM/O3-S files
contains the same number of one-hour values (n = 389) as the HAS file.
Table 46 is similar to Table 45 except that it presents percentiles for daily
maximum one-hour exposures. The HAS files and each of the 10 pNEM/O3-S files
contains 50 daily maximum one-hour exposures.
The run-to-run variability in the pNEM/O3-S estimates is most pronounced in
the upper percentiles and maximum values. The 99.5th percentile values in Table 45
vary from 78 ppb to 110 ppb over the 10 runs. The mean 99.5th value is 95.2 ppb.
This value is lower than the corresponding 99.5th percentile of the HAS data (120
ppb). In general, the mean pNEM/03-S percentile values in Table 45 are greater than
the corresponding HAS percentile values up to the 95th percentile and less than the
corresponding HAS percentile values above the 95th percentile.
123
-------
TABLE 45. DISTRIBUTIONS OF ONE-HOUR OZONE EXPOSURES (ppb)
OBTAINED FROM TEN RUNS OF pNEM/O3-S AND FROM THE
HOUSTON ASTHMATIC STUDY
Percentile
5
10
20
30
40
50
60
70
80
90
95
98
99
99.5
Maximum
pNEM/O3-S Runs
1
1
1
3
4
7
10
14
20
28
37
50
67
76
78
80
2
1
1
3
4
7
10
14
20
32
40
51
65
79
89
101
3
1
1
3
5
7
10
13
19
26
37
53
83
93
103
130
4
0
1
2
4
6
9
14
20
29
41
54
71
93
109
159
5
1
1
2
4
6
10
15
21
31
42
53
79
88
104
135
6
1
1
2
4
6
8
14
19
27
37
50
63
70
81
117
7
0
1
3
5
8
11
15
22
31
42
59
72
77
81
86
8
1
1
3
4
6
9
16
23
32
50
64
92
96
103
123
9
0
1
3
5
8
11
16
21
28
39
51
72
88
94
135
10
1
1
2
4
7
10
13
19
26
40
57
80
85
110
147
Mean
0.7
1.0
2.6
4.3
6.8
9.8
14.4
20.4
29.0
40.5
54.2
74.4
84.5
95.2
121.3
HAS
data
2
2
2
2
3
5
7
10
16
32
51
78
107
120
142
124
-------
TABLE 46. DISTRIBUTIONS OF ONE-HOUR DAILY MAXIMUM OZONE EXPOSURES
(ppb) OBTAINED FROM TEN RUNS OF pNEM/O3-S AND FROM THE
HOUSTON ASTHMATIC STUDY
Percentile
5
10
20
30
40
50
60
70
80
90
95
98
Maximum
pNEM/O3-S Runs
1
8
12
17
23
28
33
36
44
59
67
78
79
80
2
5
8
15
18
23
33
42
46
55
65
89
97
101
3
9
11
12
20
22
30
36
38
45
66
103
104
130
4
4
7
13
18
30
38
42
50
56
71
86
109
159
5
4
8
11
18
29
39
45
51
54
79
96
106
135
6
5
9
14
20
27
30
36
45
60
65
81
82
117
7
4
6
10
21
25
34
38
43
56
67
81
81
86
8
6
8
14
22
27
35
48
53
64
67
93
104
123
9
5
9
17
23
27
32
37
45
53
61
91
97
135
10
4
4
12
20
22
30
39
43
57
78
85
110
147
Mean
5.4
8.2
13.5
20.3
26.0
33.4
39.9
45.8
55.9
68.6
88.3
96.9
121.3
HAS
data
2
3
6
9
11
15
20
32
47
73
87
120
142
125
-------
The 98th percentile values in Table 46 vary from 79 ppb to 110 ppb. The mean
98th percentile value is 96.9. The corresponding HAS value is 120 ppb. The mean
pNEM/O3-S percentile values are greater than the corresponding HAS values up to
the 80th percentile and less than the HAS values above the 80th percentile. In
general, the results in Tables 45 and 46 suggest that pNEM/O3-S overpredicts the
HAS exposures in the range below 70 ppb and underpredicts in the range above 70
ppb.
7.5 Sensitivity of Exposure Estimates to Ozone Decay Rate
A series of exploratory analyses were conducted to identify factors contributing
to the observed differences between exposure values obtained from pNEM/O3-S and
HAS. The results of these analyses suggested that pNEM/O3-S exposure estimates
were particularly sensitive to the distribution of ozone decay rate used in the mass
balance algorithm. This effect is demonstrated in Table 47. The table presents
average ozone concentrations by microenvironment for each of 12 pNEM/O3-S runs.
In each run the mean value (4.04 h"1) used to define the distribution of the ozone
decay rate parameter (Fd) for indoor microenvironments in Subsection 3.1 was
multiplied by a factor between 0.25 and 4.0, the value of the factor varying with
microenvironment and window status (Table 48). The standard deviation of the Fd
distribution was held constant at 1.35 h"1 in each case, the value specified in
Subsection 3.1. Note that the ozone decay rate parameters were not modified for Run
No. 1, as the multipliers were set equal to 1.0.
Table 47 also presents average ozone concentrations by microenvironment for
the HAS data. Run 12 of pNEM/O3-S was judged to produce the best overall
•matchup between pNEM/O3-S and HAS microenvironmental ozone concentrations. In
this run, one set of multiplicative factors was used for the indoors - residence
microenvironment (windows open = 1.0, windows closed = 0.25) and another set
(windows open = 2.0, windows closed = 2.0) was used for all other indoor
microenvironments.
126
-------
In reviewing the results in Table 47, it is important to note that the run-to-run
variation in average ozone concentrations in the outdoor and in-vehicle
microenvironments is not a function of the specified multiplicative factors. A value for
ozone decay rate was not required by the algorithm used to estimate outdoor
concentrations; all 12 runs used the same point estimate (Fd = 72.0) for the ozone
decay rate of the in-vehicle microenvironment. It should also be noted that windows in
the indoors - other microenvironment are assumed to be closed at all times.
Consequently, the average ozone concentration in this microenvironment varies only
as a function of the multiplier applied to the windows closed case.
The results of this initial attempt to validate pNEM/OS indicate that the use of
alternative values for ozone decay rate will produce exposure estimates that more
nearly match the HAS data. In interpreting this result, it is important to note that 1)
the HAS data base only applies to two Houston neighborhoods, 2) the HAS data base
represents only 30 subjects, and 3) the alternative ozone decay rates are not
supported by the results of the literature survey discussed in Subsection 3.1. For
these reasons, researchers used the decay rate values listed in Subsection 3.1 for the
pNEM/O3 runs summarized in Section 6. ITAQS recommends that EPA acquire up-
to-date personal monitoring data for a variety of urban areas so that a more definitive
validation of pNEM/O3 can be performed.
127
-------
TABLE 47. MEAN OZONE CONCENTRATIONS IN MICROENVIRONMENTS BASED ON PERSONAL
MONITORING DATA FROM THE HOUSTON ASTHMATIC STUDY AND ON EXPOSURE ESTIMATES
OBTAINED FROM 12 RUNS OF pNEM/O3-S WITH DIFFERING VALUES OF OZONE DECAY RATE
Microenvironment
Indoors -
residence (all)
Indoors -
residence, no air
conditioning
Indoors -
residence, room
air conditioning
Indoors -
residence, central
air conditioning
Indoors - other
Outdoors
In vehicle
Mean
ozone
cone.
based on
HAS data
7.8
25.0
5.1
4.7
6.0
37.7
16.2
1
9.9
15.0
11.6
8.1
13.0
43.2
14.3
Mean ozone concentration by pNEM/O3-S Run (Table 49), ppb
2
8.0
12.6
9.1
6.3
9.7
43.3
14.3
3
6.6
10.8
8.0
5.1
7.4
43.9
14.4
4
11.8
18.7
14.5
9.3
13.2
43.2
14.3
5
12.7
20.6
15.3
9.9
13.5
43.5
14.5
6
9.0
14.7
11.7
6.8
7.5
43.1
14.4
7
10.9
18.1
13.0
8.3
7.5
43.1
14.2
8
11.7
21.1
14.2
8.4
7.4
43.6
14.5
9
8.4
14.8
10.6
6.2
4.1
43.6
14.4
10
8.2
15.0
11.3
5.7
4.1
43.0
14.3
11
11.3
21.1
15.0
7.8
4.0
43.2
14.1
12a
9.0
21.3
8.3
5.1
7.5
43.1
14.3
CO
'Run judged to provide exposure estimates most consistent with HAS data.
-------
TABLE 48. MULTIPLICATIVE FACTORS USED TO DETERMINE
ALTERNATIVE VALUES FOR MEAN OZONE DECAY RATE
pNEM/O3-S run
1
2
3
4
5
6
7
8
9
10
11
12
Multiplicative factor (m)a
Indoors-residence, no air
conditioning
Windows
closed
1.0
1.5
2.0
1.0
1.0
2.0
2.0
2.0
4.0
4.0
4.0
1.0
Windows
open
1.0
1.5
2.0
0.5
0.25
1.0
0.5
0.25
1.0
0.5
0.25
0.25
All other indoor
microenvironments
Windows
closed
1.0
1.5
2.0
1.0
1.0
2.0
2.0
2.0
4.0
4.0
4.0
2.0
Windows
open
1.0
1.5
2.0
0.5
0.25
1.0
0.5
0.25
1.0
0.5
0.25
2.0
3Mean ozone decay rate = (m)(4.04 h"1).
129
-------
SECTION 8
PRINCIPAL LIMITATIONS OF THE pNEM/O3 METHODOLOGY
The pNEM/OS methodology was developed specifically to meet the requirements of
OAQPS for a computer-based model capable of simulating the ozone exposures of
specific population groups under alternative NAAQS. In addition to meeting these needs,
the designers of pNEM/O3 have attempted to create a model which is flexible in
application and easy to upgrade. The model was deliberately constructed as a collection
of stand-alone algorithms organized within a modular framework. For this reason, analysts
can revise individual algorithms without the need to make major changes to other parts of
the model.
The structure of each algorithm in pNEM/O3 is largely determined by the
characteristics of the available input data. For example, the algorithm used to construct a
season-long exposure event sequence for each cohort is constrained by the fact that none
of the available time/activity studies provides more than three days of diary data for any
one subject. To make maximum use of the available diary data, the pNEM/O3 sequencing
algorithm constructs each exposure event sequence by sampling data from more than one
subject. The other pNEM/O3 algorithms are similarly designed to make best use of
available data bases.
In evaluating the exposure estimates presented in this report, the reader should
note that the model has a number of limitations which may affect its accuracy. These
limitations are usually the result of limitations in the input data bases. The available data
were typically collected for purposes other than use in a population exposure model.
Consequently, these data frequently represent special sets of conditions which differ from
those assumed by pNEM/O3. In these situations, analysts must exercise a certain degree
of judgement in adapting the data for use in pNEM/O3.
130
-------
This section presents a brief discussion of the principal limitations in the pNEIWOS
methodology as applied to 1990 population data. The limitations are organized according
to five major components of the model: time/activity patterns, equivalent ventilation rates,
air quality adjustment, the mass balance model, and the estimation of cohort populations.
8.1 Time/Activity Patterns
In the general pNEM/OS methodology, the exposure-related activities of each cohort
are represented by a multi-day exposure event sequence which spans a specified ozone
season. Each sequence is constructed by an algorithm which selects 24-hour (midnight-
to-midnight) activity patterns from a specially prepared database. This database contains
data from one or more time/activity studies in which subjects recorded their daily activities
in diaries.
In the application of pNEM/O3 described in this report, the time/activity database
consisted of diary data obtained from 900 subjects of the Cincinnati Activity Diary Study
(CADS). The database contained 2,649 person-days of data, an average of slightly less
than three days per subject. All 900 subjects resided in the Cincinnati metropolitan area.
Analysts used time/activity data obtained solely from the CADS subjects to
represent the activities of the general population in each of the nine study areas. Although
the algorithm which constructs exposure event sequences attempts to account for effects
of local climate on activity, it is unlikely that this adjustment procedure corrects for all inter-
city differences in people's activities. Time/activity patterns are likely to be affected by a
variety of local factors, including topography, land-use, traffic patterns, mass transit
systems, and recreational opportunities.
As discussed previously, the average subject provided less than three days of diary
data. For this reason, the construction of each season-long exposure event sequence
required either the repetition of data from one subject or the use of data from multiple
subjects. The latter approach was used in the pNEM/O3 analyses to better represent the
variability of exposure expected to occur among the people included in each cohort. The
principal deficiency of this approach is that it may not adequately account for the day-to-
day repetition of activities common to individuals. Using activities from different subjects
131
-------
may underestimate multiple occurrences of high exertion and/or outdoor exposure for
those segments of the population who engage in repetitive outdoor activities.
8.2 Equivalent Ventilation Rates
In the general pNEM/OS methodology, the EVR associated with each exposure
event is determined by an algorithm which randomly selects the value from one of the
eight lognormal distributions presented in Table 4. Each distribution is specific to age
group (children or adults) and breathing rate category (sleeping, slow, medium, or fast).
The distributions are based on EVR data obtained from two diary studies conducted by J.
Hackney and associates in Los Angeles.
A total of 36 subjects participated in the Los Angeles studies. Because of the small
sample size, the resulting EVR database may not accurately represent the variability of
EVR across the population. In addition, the database does not provide sufficient data to
adequately characterize age-specific differences in EVR. For example, none of the Los
Angeles subjects was below the age of 10 or above the age of 50. The lognormal
distributions of EVR developed for children and adults are likely to over-estimate EVR
when applied to pre-school children or older adults.
If the resulting EVR estimates are biased, they are likely to be biased high because
of the operation of the EVR limiting algorithm. This algorithm determines the maximum
EVR that can be maintained for a specified duration by a subject who is (1) male, (2)
between the ages of 18 and 24, (3) exercising regularly, and (4) motivated to reach a high
ventilation rate. The EVR value assigned by pNEM/O3 to an event of the specified
duration is not permitted to exceed this value.
The four conditions assumed by the EVR limiting algorithm do not apply to all
members of the population. Consequently, the EVR limiting algorithm may permit more
high EVR values to occur in the pNEM/O3 simulation than would occur in the actual
population. This potential bias may be corrected in future versions of pNEM/O3 by
distinguishing cohorts by gender, age, and physical conditioning. The parameters of the
EVR limiting algorithm would be varied according to these factors to yield a reasonable
upper EVR limit for each cohort.
132
-------
8.3 The Air Quality Adjustment Procedures
Section 5 presents a summary of the procedures used to adjust baseline ozone
monitoring data to simulate conditions expected when a study area just attains a
specified NAAQS. These procedures assume that 1) the Weibull distribution provides a
good fit to most ozone data, and 2) the parameters of the Weibull distribution fitting data
from a particular monitoring site will change over time in a predictable fashion. The
adjustment procedures include equations for predicting the values of the Weibull
parameters under future attainment conditions.
The prediction equations were developed through a statistical analysis of ozone
data obtained from selected monitoring sites which have experienced moderate
reductions in ozone levels during the 1980's. It should be noted that none of the
selected monitoring sites reported ozone reductions of the magnitude required to bring
Los Angeles into compliance with any one of the NAAQS under evaluation. For this
reason, the prediction equations may not produce accurate estimates for the Weibull
distribution parameters when applied to Los Angeles ozone data.
The air quality adjustment procedure is based on an assumption that the
attainment status of a particular city can be determined by a single year of monitoring
data. For example, the current status of Philadelphia is determined by ozone monitoring
data for 1991. This single year of monitoring data is then adjusted to exactly meet a
specified NAAQS. It should be noted that the pNEM/O3 approach to determining
attainment status differs somewhat from the actual method used by EPA to determine
attainment status. EPA typically examines three recent years of monitoring data for a
particular city and calculates a multi-year air quality indicator (e.g., the fourth highest
daily maximum one-hour ozone concentration for the three-year period). The air quality
indicator determined by this method is likely to differ from the air quality indicator
determined in a pNEM/O3 analysis from a single year of data. As the direction of the
difference is random, the degree of adjustment applied to a city by pNEM/O3 may be
greater than or less than the adjustment required to bring the city into compliance based
on three years of data.
133
-------
8.4 The Mass Balance Model
The pNEM/03 methodology uses the mass balance model described in Section 3
to estimate ozone concentrations for the following enclosure categories:
Residential buildings - windows closed
Residential buildings - windows open
Nonresidential buildings
Vehicles.
The mass balance model provides hourly average ozone concentrations for each
enclosure category as a function of outdoor ozone concentration, AER, and ozone
decay rate.
in the application of pNEM/OS described in this report, the outdoor ozone
concentration required by the mass balance model was always derived from fixed-site
monitoring data. These data were representative of local conditions and were
considered to be relatively reliable.
The AER values for residential buildings with closed windows were obtained from
a lognormal distribution fit to AER data from 312 residences. These data were
considered to be generally representative of housing in urban areas in the U.S.
No comparable databases were identified which were considered representative
of residences with open windows. Consequently, analysts represented this enclosure
category with a point estimate developed by Hayes45. Analysts are uncertain as to the
accuracy and general applicability of this estimate.
The AER values for nonresidential buildings were obtained from a lognormal
distribution fit to AER data from 40 buildings provided by Turk et al.44 This sample may
be too small to adequately characterize nonresidential buildings in the U.S. It should
also be noted that the Turk data are likely to represent only buildings with closed
windows. Consequently, the lognormal distribution derived from the Turk data is likely
to under-estimate the ozone exposures of people who frequently occupy nonresidential
buildings with open windows.
134
-------
A point estimate of 36 air changes per hour was used for the AER of vehicles.
This value was obtained from Hayes47 based on his analysis of data reported by
Peterson and Sabersky42 for a single vehicle. The use of a point estimate is considered
unrealistic as it does not account for varying ventilation conditions within a particular
motor vehicle or the variability in AER from vehicle to vehicle.
Analysts also used a point estimate for the ozone decay rate of vehicles. This
value was based on data from a single automobile and may be biased.
Ozone decay rates for residential and nonresidential buildings were sampled from
a normal distribution. This distribution was based on decay rate data for a relatively
small number of buildings assembled by Weschler et al.39 These data may not
adequately represent the variability of ozone decay rates among urban buildings in the
U.S.
135
-------
REFERENCES
1. Richmond, H. M. and T. McCurdy, 1988, "Use of Exposure Analysis and Risk
Assessment in the Ozone NAAQS Review Process," Paper No. 88-121.3,
presented at the 81st Annual Meeting of APCA, Dallas, Texas.
2. Ott, W. R., 1982, "Concepts of Human Exposure to Air Pollution," Environment
International. Vol. 7, page 179.
3. Duan, N., 1982, "Models for Human Exposure to Air Pollution," Environment
International, Vol. 8, page 305.
4. Biller, W. F., T. B. Feagans, and T. R. Johnson, June 1981, "A General Model for
Estimating Exposure Associated With Alternative NAAQS," Paper No. 81-18.4,
presented at the 74th Annual Meeting of the Air Pollution Control Association,
Dallas, Texas.
5. Paul, R. A. and T. McCurdy, June 1986, "Estimation of Population Exposure to
Ozone," Paper No. 86-66.2, presented at the 79th Annual Meeting of the Air
Pollution Control Association, Dallas, Texas.
6. Johnson, T. R. and R. A. Paul, 1981, "The NAAQS Exposure Model (NEM)
Applied to Particulate Matter," prepared by PEDCo Environmental, Inc. for the
Office of Air Quality Planning and Standards, U. S. Environmental Protection
Agency, Research Triangle Park, North Carolina.
7. Johnson, T. R. and R.A. Paul, 1983, "The NAAQS Exposure Model (NEM)
Applied to Carbon Monoxide," prepared by PEDCo Environmental, Inc. for the
Office of Air Quality Planning and Standards, U. S. Environmental Protection
Agency.
8. Paul, R. A., T. R. Johnson, and T. McCurdy, 1988, "Advancements in Estimating
Urban Population Exposure," Paper No. 8.8-127.1, presented at the 81st Annual
Meeting of the Air Pollution Control Association, Dallas, Texas.
136
-------
9. Pandian, M. D., 1987, "Evaluation of Existing Total Human Exposure Models,"
EPA-60Q/4-87-004, U. S. Environmental Protection Agency, Las Vegas, Nevada.
10. Ryan, P. B., 1991 "An Overview of Human Exposure Modeling," Toxicology and
Industrial Health. Vol. 1, No. 4, pp. 453-474.
11. McKee D., H. Richmond, P. Johnson, and T. McCurdy, 1984, "Review of the
NAAQS for Carbon Monoxide: Reassessment of Scientific and Technical
Information," EPA-45075-84-004, U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina.
12. McKee, D., P. Johnson, T. McCurdy, 1989, "Review of the National Ambient Air
Quality Standards for Ozone: Assessment of Scientific and Technical
Information," EPA-450/2-92-001, U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina.
13. Johnson, T. R., R. A. Paul, J. E. Capel, and T. McCurdy, 1990, "Estimation of
Ozone Exposure in Houston Using a Probabilistic Version of NEM," Paper No.
90-150.1, presented at the 83rd Annual Meeting of the Air and Waste
Management Association, Pittsburgh, Pennsylvania.
14. Johnson, T. R., J. E. Capel, E. Olaguer, and L. Wijnberg, 1992, "Estimation of
Carbon Monoxide Exposures and Associated Carboxyhemoglobin Levels in
Denver Residents Using a Probabilistic Version of NEM," prepared by IT Air
Quality Services for the Office of Air Quality Planning and Standards, U. S.
Environmental Protection Agency, Research Triangle Park, North Carolina.
15. Johnson, T., J. Capel, E. Olaguer, and L Wijnberg, 1993, "Estimation of Ozone
Exposues Experienced by Urban Residents Using a Probabilistic Version of
NEM," prepared by IT Air Quality Services for the Office of Air Quality Planning
and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
North Carolina, February.
16. Bureau of Census, 1992, "Census of Population and Housing, 1990: Summary
Tape File 3-A," Washington, D.C.
17. Johnson, T. R., 1989, Letter to Tom McCurdy, Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, June 19.
18. Johnson, T. R., 1987, "A Study of Human Activity Patterns in Cincinnati, Ohio,"
prepared by PEl Associates, Inc. for Electric Power Research Institute, Palo Alto,
available from Ted Johnson, IT Corporation, 3710 University Drive, Suite 201,
Durham, North Carolina 27707.
137
-------
19. Johnson, T. R., 1989, Letter to Tom McCurdy, Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, July 28.
y/20. Stock, T. H., D. J. Kotchman, C. F. Contant, et al., 1985, "The Estimation of
Personal Exposures to Air Pollutants for a Community-Based Study of Health
Effects in Asthmatics - Design and Results of Air Monitoring," Journal of the Air
Pollution Control Association. Vol. 35, p. 1266.
)/ 21. Weschler, C. J., H. C. Shields, and D. V. Naik, 1989, "Indoor Ozone Exposures,"
Journal of the Air Pollution Control Association. Vol. 39, p. 1562.
22. Rhodes, C. E. and D. M. Holland, 1981, "Variations of NO, NO2, and O3
Concentrations Downwind of a Los Angeles Freeway," Atmospheric Environment.
Vo. 15, p. 243.
23. Johnson, T. R. and L. Wijnberg, 1981, "Time Series Analysis of Hourly Average
Air Quality Data," Paper No. 81-33.5, presented at the 74th Annual Meeting of the
Air Pollution Control Association, Philadelphia, Pennsylvania.
24. McDonnell, W. F., D. H. Horstman, and M. J. Hazuch, 1983, "Pulmonary Effects
of Ozone Exposure During Exercise: Dose-Response Characteristics," Journal of
Applied Physiology. Vol. 54, p. 1345.
25. Johnson, T. R., L. Wijnberg, and J.E. Capel, 1990, "Review and Evaluation of
New Research Relating to Population Exposure to Air Pollution, Executive
Summary", available from Dr. Will Ollison, American Petroleum Institute, 1220 L
Street, N.W., Washington, D.C. 20005.
26. Erb, B. D., 1981, "Applying Work Physiology to Occupational Medicine,"
Occupational Health Safety. Vol. 50, pp. 20-24.
27. Astrand, P. O. and K. Rodahl, 1977, Textbook of Work Physiology, 2nd ed.
McGraw-Hill, New York, New York.
28. McArdle, W. D., F. I. Katch, and V. L. Katch, 1991, Exercise Physiology: Energy,
Nutrition, and Human Performance. Lea and Febiger, Malvern, Pennsylvania.
29. Johnson, T. R. and W. C. Adams, "An Algorithm for Determining Maximum
Sustainable Ventilation Rate According to Gender, Age, and Exercise Duration/'
available from Ted Johnson, IT Corporation, 3710 University Drive, Suite 201,
Durham, North Carolina 27707.
138
-------
rt VJ ( 30. Johnson, T. R., J. E. Capel, and D. M. Byrne, 1991, "The Estimation of
J Commuting Patterns in Applications of the Hazardous Air Pollutant Exposure
Model (HAPEM)," Paper No. 91-172.6, presented at the 84th Annual Meeting of
the Air and Waste Management Association, Vancouver, Canada.
31. Nagda, N. L, H. E. Rector, and M. D. Koontz, 1987, Guidelines for Monitoring
Indoor Air Quality. Hemisphere Publishing Corporation, Washington, D. C.
32. Weschler, C. J., H. C., Shields, and D. V., Nike, 1992, "Indoor Ozone: Recent
Findings," Tropospheric Ozone in the Environment II. editor, R. Burglund, Air and
Waste Management Association, Pittsburgh, PA.
33. Nazaroff, W.W., Cass, G.R., 1986, "Mathematical Modeling of Chemically
Reactive Pollutants in Indoor Air," Environmental Science and Technology. Vol.
20, pp. 880-885.
34. Hayes, S.R., 1989, "Estimating the Effect of Being Indoors on Total Personal
Exposure to Outdoor Air Pollution," Journal of the Air Pollution Control
Association. Vol. 39, No. 11, pp. 1453-1461.
35. Peterson, G.A. and R.H. Sabersky, "Measurements of Pollutants Inside an
Automobile," Journal of the Air Pollution Control Association, Vol. 25, No. 10, pp.
1028-1032, (October, 1975).
36. Grimsrud, D. T., M. H. Sherman, and R. C. Sondregger, 1982, "Calculating
Infiltration: Implications for a Construction Quality Standard," Proceedings of the
ASHRAE/DDE Conference: Thermal Performance of the Exterior Envelopes of
Buildings II. Las Vegas, Nevada.
37. Turk, B. H., D. T. Grimsrud, J. T. Brown, K. L. Geisling-Sobotka, J. Harrison, and
R. J. Prill, 1989, "Commercial Building Ventilation Rates and Particle
Concentrations," ASHRAE Transactions. Vol. 95, Part 1.
38.- Hayes, S. R., 1991, "Use of an Indoor Air Quality Model (IAQM) to Estimate
Indoor Ozone Levels." Journal of the Air and Waste Management Association,
Vol. 41, pp.161-170.
39. Johnson, T. R., 1993, Letter to Tom McCurdy, Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, February 18.
139
-------
APPENDIX A
ADJUSTMENT OF OZONE DATA TO
SIMULATE NAAQS ATTAINMENT
A-1
-------
INTERNATIONAL
TECHNOLOGY
CORPORATION
April 15, 1993
Mr. Thomas McCurdy
U.S. Environmental Protection Agency
OAQPS, MD-12
Research Triangle Park, North Carolina 27711
Adjustment of Ozone Data to Simulate NAAQS Attainment
Dear Torn:
Under Work Assignment II-5 of EPA Contract No. 68-DO-0062, IT Air Quality Services will
be applying a revised version of pNEM/O3 to nine urban areas. Fixed-site monitoring data
for the years 1990 and 1991 will be used to represent "as is" ambient concentrations within
each exposure district. An air quality adjustment procedure (AQAP) will be used to adjust
the "as is" data to simulate various 1-hour and 8-hour National Ambient Air Quality
Standards (NAAQS).
Historically, the AQAP's developed for NEM applications have included the following steps:
1. Specify an air quality indicator (AQI) to be used in evaluating the status of a
monitoring site with respect to the NAAQS of interest.
2. Determine the value of the AQI for each site within a study area under "as is"
conditions.
3. Determine the value of the AQI under conditions in which the air pollution
levels within the study area have been reduced until a single site just attains a
specified NAAQS.
4. Adjust the one-hour values of the "as is" data set associated with each site to
yield the AQI value determined in Step 3. The adjusted data set should retain
the temporal profile of the "as is" data set.
Regional Oilice
3710 University Drive • South Square • Corporate Center One • Suite 201
Durham. North Carolina 27707 • 919-493-3661
IT Corporation is a wholly owned subsidiary ol international Technology Corporation
-------
Mr. Tom McCurdy 2 April 15, 1993
This letter provides recommendations for implementing each of these steps.
STEP 1: SPECIFY AIR QUALITY INDICATOR
To determine compliance with the current one-hour National Ambient Air Quality Standard
(NAAQS) for ozone, a "design value" has been determined for each urban area in the United
States according to procedures specified by EPA. A common means of determining the
design value is to determine the daily maximum hourly average value ranked n+1 when the
daily maximum values for n years (n & 3) are combined and ranked from highest to lowest.
Following the directions of AQMD, ITAQS is using an alternative air quality indicator to
determine compliance with 1-hr NAAQS: the characteristic largest daily maximum one-hour
value determined from a distribution fit to ozone data reported for a single ozone season. A
similar quantity, the characteristic largest daily maximum eight-hour value, will be used to
determine compliance with proposed 8-hr NAAQS.
The Characteristic Largest Daily Maximum Value
The characteristic largest value of a distribution is that value expected to be exceeded once in
n observations. If F(x) is the cumulative distribution of x, then
F(CLV) = 1 - — (1)
n
if CLV is the characteristic largest value.
Selection of an appropriate cumulative distribution to fit data is important in determining a
reasonable characteristic largest value. Two distributions that often provide close fits to
ambient air quality data are the Weibull and the lognormal. The Weibull distribution is
defined as
F(x) = 1 - exp [-(-)*] (2)
where 6 is the scale parameter and k is the shape parameter. The lognormal distribution is
defined as
F(x) = -L, f^exp (-tV2) dt (3)
where
-------
Mr. Tom McCurdy 3 April 15, 1993
In x -
and In x is distributed normally with mean \i and variance a2. As discussed in previous
reports, the Weibull distribution generally provides a better fit to hourly average ozone data.
The hourly average values reported by a single monitoring site during a specified ozone
season form a time series x, (t = 1, 2, 3, ..., n). If the hourly average time series is complete,
it will contain n = (24)(N) values, where N is the number of days in the ozone season. From
this time series a second time series of daily maximum 1-hour values can be constructed.
Assume that a Weibull distribution with parameters 6 and k provides a good fit to the
empirical distribution of hourly average values. If we disregard autocorrelation, the value
expected to be exceeded once in n = (24)(N) hours can be estimated as
CLVOH = 6 [ln(24)(N)]1/JC (5)
This is the characteristic largest one-hour value. If we again disregard autocorrelation, the
daily maximum 1-hour value expected to be exceeded once in N days can be estimated as
CLVOHDM = 6{-ln[l - ( " )1/24]}1/*. (6)
N
This is the characteristic largest daily maximum 1-hour value. For 7-month and 12-month
ozone seasons, N is equal to 214 and 365, respectively. For these values of N, CLVOH and
CLVOHDM are virtually indistinguishable in value over the range in k values normally found
in ozone data (0.6 < k < 2.5). For example, the following values were calculated using 6 =
40 ppb.
N k CLVOH CLVOHDM
214 0.6 1428 1428
1.4 185 185
2.5 94 94
365 0.6 1580 1580
1.4 193 193
2.5 97 97
-------
Mr. Tom McCurdy 4 April 15, 1993
The CLVOH and CLVOHDM values match to the nearest ppb. Consequently, we can use the
expression
CLVOHDM - 6 [In(24) (N) ]1X* (7)
as an alternative to Equation 6 for calculating CLVOHDM. This expression is already in
place in the computer programs previously developed by ITAQS for fitting Weibull
distributions to air quality data.
For simplicity, we will refer to the quantity calculated by Equation 7 as CLV1 throughout the
remainder of this letter. CLV1 will be the AQI by which we evaluate the status of a
monitoring site with respect to a particular 1-hr NAAQS.
A data set containing one-hour concentration values can be processed to determine a
corresponding data set containing eight-hour running average values. If we fit a Weibull
distribution to the eight-hour data, we can determine a characteristic largest eight-hour value
by the equation
CLVEH = 6[ln(24)N)]1/JC, (8)
where 6 and k are the Weibull parameters for the eight-hour fit. Based on the argument
made above for one-hour data, this value should be approximately equal to the characteristic
largest daily maximum eight-hour value (CLVEHDM) of the data set. For simplicity, we will
use the term CLV8 to refer to the quantity calculated by Equation 8 throughout the remainder
of this letter. CLV8 will serve as the AQI by which we evaluate attainment status with
respect to a particular 8-hr NAAQS.
STEP TWO: DETERMINE VALUES OF AQI UNDER "AS IS" CONDITIONS
My letter of February 18, 1993 lists the data sets we have selected to represent "as is"
conditions in each of the nine cities under analysis. Tables 22 through 40 in that letter
provide estimates of CLV1 and CLV8 based on Weibull fits to the upper two percent of each
data set. These values will be our estimates of CLV1 and CLV8 under "as is" conditions.
STEP THREE: ADJUST VALUES OF AQI TO SIMULATE ATTAINMENT
We have developed a new statistical adjustment procedure for implementing this step which
does not require the use of EKMA or a similar photochemical model. The procedure is based
on patterns identified in a test data base containing multi-year ozone data.
-------
Mr. Tom McCurdy 5 April 15, 1993
The Test Data Set
Following your guidance, we selected three of the nine urban areas (Houston, Philadelphia,
and St. Louis) to be test cities for the purposes of developing a "CLV Adjustment Procedure"
(CLVAP). As part of the pNEM/O3 analysis, we had already selected 11 fixed-site monitors
to represent Houston, 10 to represent Philadelphia, and 11 to represent St. Louis. We
reviewed 1983 - 1992 ozone data reported for the sites associated with each city and selected
three years for our test data base. These years represented low, medium, and high ozone
levels with respect to the ten year period. To each set of three years we added the year
selected for use in the pNEM/O3 analysis (either 1990 or 1991). In this way, we identified
12 city-years (3 cities x 4 years) for use in developing the CLVAP.
A total of 126 site-years of ozone data were associated with these 12 city-years. We
collected 1-hour ozone data associated with each site-year, processed the data to produce a
complete data set for the specified ozone season, and fit a Weibull distribution to each data
set (upper 2 percent fit). The fitted Weibull distribution provided an estimate of CLV1.
We also developed a running 8-hour data set from each 1-hour data set. We then fit a
Weibull distribution to each data set and used the results to estimate CLV8 for the data set.
Figures 1, 2, and 3 are printouts listing statistics for Houston, Philadelphia, and St Louis,
respectively, based on the one-hour data sets. Figures 4, 5, and 6 provide corresponding
statistics for the eight-hour data sets. In following abbreviations are used in these printouts.
DISTRICT: district associated with monitor
YEAR: last two digits of calendar year
P50: 50th percentile
P90: 90th percentile
P95: 99th percentile
P995: 99.5th percentile
SECHI: second largest value
HIGH: largest value
-------
Mr. Tom McCurdy
April 15, 1993
DISTRICT YEAR P50 P90 P95 P99 P995 SECHI HIGH CLV DELTA
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
8
8
8
9
9
9
9
10
10
10
10
11
11
11
11
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
20
20
20
20
22
20
20
20
20
20
10
10
17
18
10
10
20
20
20
20
M
20
10
15
20
20
20
14
20
20
20
15
20
16
10
17
10
20
10
14
10
10
10
10
60
60
60
50
50
60
50
40
50
50
50
44
50
50
50
40
50
50
50
50
M
60
40
40
50
50
46
41
60
60
50
40
50
50
45
47
50
50
50
42
40
50
40
40
70
70
70
60
60
70
70
51
66
70
60
60
60
70
60
53
70
70
70
62
M
70
50
60
61
70
60
56
70
70
60
50
70
70
60
60
60
63
60
60
60
60
60
53
110
100
110
100
90
110
100
70
100
100
100
90
100
100
100
80
110
100
110
107
M
100
70
100
100
100
90
88
110
100
100
80
110
100
90
90
110
90
100
95
100
90
100
87
120
120
120
110
110
120
110
80
120
120
120
102
120
"no
120
93
130
110
130
120
M
110
90
110
120
110
110
101
130
110
110
90
120
110
110
'100
120
100
120
111
120
110
120
100
170
220
220
170
170
220
160
130
210
220
200
160
220
200
210
151
210
190
200
204
M
190
140
200
190
160
180
158
270
150
230
150
230
190
200
180
250
200
230
180
260
220
220
170
180
230
220
170
180
250
180
130
210
220
230
170
230
200
240
158
210
190
220
209
M
210
190
200
210
170
230
170
290
180
230
150
250
220
210
200
260
210
230
180
260
230
220
175
193
204
224
186
200
208
182
122
219
217
241
166
266
214
224
164
224
177
227
219
M
189
180
214
212
184
208
185
280
170
207
162
240
189
200
177
264
191
235
197
247
238
232
169
27.2
22.6
22.2
22.8
15.8
24.1
27.3
23.7
17.5
18.7
14.3
22.4
10.7
17.5
16.9
17.5
24.7
27.4
21.0
18.5
' M
24.9
9.5
16.1
21.5
23.2
16.5
16.3
14.7
31.3
17.3
16.5
18.4
23.4
13.0
18.9
13.5
19.2
17.3
18.4
14.7
11.7
14.9
18.9
1.13
1.00
0.95
1.05
0.87
1.02
1.16
1.34
0.87
0.90
0.78
1.10
0.69
0.88
0.85
0.99
1.00
1.18
0.93
0.89
M
1.09
0.75
0.85
0.96
1.07
0.87
0.91
0.75
1.31
0.89
0.96
0.86
1.06
0.77
0.98
0.74
0.96
0.85
0.93
0.78
0.73
0.80
1.01
Figure 1. Descriptive statistics for one-hour ozone data obtained from
monitoring sites associated with Houston exposure districts.
-------
Mr. Tom McCurdy 7 April 15, 1993
DISTRICT YEAR P50 P90 P95 P99 P995 SECHI HIGH CLV DELTA K
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
8
8
8
9
9
9
9
10
10
10
10
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
32
33
29
34
30
25
26
28
28
31
38
36
33
30
28
33
25
22
25
28
31
27
29
30
24
20
23
26
30
30
30
30
20
20
20
20
30
30
30
30
74
66
63
72
77
65
70
70
68
70
73
76
79
70
67
73
69
60
66
70
70
60
65
67
63
53
61
66
70
60
70
70
53
50
49
50
70
60
67
70
91
80
75
88
96
80
84
84
81
83
85
89
95
84
81
87
88
74
79
84
86
72
77
78
75
65
73
77
90
80
80
80
70
60
60
70
90
80
80
80
123
111
98
117
122
115
113
114
105
107
111
112
127
111
104
115
121
105
101
110
112
96
102
103
101
82
94
98
120
100
100
100
100
80
80
90
120
100
100
110
132
123
109
124
132
"131
124
120
119
114
119
117
141
120
112
125
133
117
111
118
121
108
108
108
110
87
101
"105
130
110
110
110
110
100
90
100
130
110
110
110
149
156
145
156
166
160
155
143
154
152
162
146
186
149
138
151
196
203
143
139
192
143
144
125
134
157
117
125
160
150
140
140
150
120
120
130
160
180
140
130
152
163
152
156
184
163
161
148
162
152
169
149
205
153
139
151
196
230
148
144
197
146
167
135
135
174
118
127
170
160
150
140
160
150
160
130
170
180
140
140
163
172
152
167
170
180
164
149
163
146
160
153
207
153
144
162
186
214
150
145
177
147
147
134
151
147
125
135
170
154
143
140
159
152
131
131
179
175
139
141
59.0
37.2
33.3
46.9
54.0
36.4
45.5
56.4
37.6
48.4
44.7
51.0
38.6
49.8
46.2
49.3
41.5
18.6
38.8
56.6
38.0
33.9
40.1
51.2
35.6
20.5
45.9
44.3
48.2
32.9
42.1
51.2
30.4
18.8
22.2
38.1
44.4
28.2
50.1
54.5
2.11
1.40
1.41
1.69
1.87
1.34
1.68
2.21
1.46
1.95
1.68
1.96
1.28
1.91
1.89
1.81
1.43
0.88
1.58
2.28
1.40
1.46
1.66
2.23
1.49
1.09
2.14
1.93
1.70
1.39
1.76
2.14
1.30
1.03
1.21
1.74
1.54
1.18
2.11
2.26
Figure 2. Descriptive statistics for one-hour ozone data obtained from
monitoring sites associated with Philadelphia exposure districts.
-------
Mr. Tom McCurdy 8 April 15, 1993
DISTRICT YEAR P50 P90 P95 P99 P995 SECHI HIGH CLV DELTA I
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
3
3
8
9
9
9
9
10
10
10
10
11
11
11
11
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
24
22
19
17
10
19
27
27
26
23
29
29
31
31
24
26
27
23
24
24
29
27
18
26
28
23
29
30
14
18
18
18
6
19
24
24
M
23
18
17
16
17
24
22
62
54
48
48
36
46
55
60
66
53
59
64
67
61
48
54
72
57
54
63
72
62
42
58
71
55
58
63
34
45
44
42
37
46
53
53
M
50
40
42
46
48
53
50
76
63
57
59
47
54
66
75
83
63
72
77
81
69
55
64
88
66
65
75
87
72
47
67
88
63
70
74
44
53
52
51
53
54
63
63
M
60
48
51
60
55
65
59
99
79
73
75
67
69
90
94
111
86
90
101
108
95
69
82
120
86
83
97
120
93
61
85
123
87
92
96
64
69
69
68
103
74
82
82
M
75
64
66
90
71
86
76
110
87
82
80
78
76
98
99
125
94
99
106
116
106
75
88
130
96
92
•104
133
103
64
93
133
96
96
100
69
77
74
73
112
78
89
85
M
87
72
70
99
- 79
94
80
138
108
116
102
95
120
125
114
171
132
120
124
184
157
99
106
210
144
125
120
184
144
75
114
224
136
130
119 '
89
120
96
84
177
125
108
102
M
110
100
88
145
98
117
95
149
111
124
104
104
124
125
118
195
136
127
135
195
164
100
112
212
150
127
123
188
151
80
115
243
137
135
121
93
123
96
85
183
129
111
107
M
110
110
88
148
106
129
97
149
115
124
103
113
115
141
118
187
145
131
132
178
164
103
110
188
159
122
125
192
149
78
121
201
146
124
123
96
121
100
90
187
132
114
105
M
121
103
88
145
111
119
99
36.6
31.7
21.3
33.8
17.8
19.9
29.6
51.2
31.3
23.7
36.8
50.7
31.3
23.7
26.9
37.7
38.8
19.3
33.7
51.5
37.2
29.3
32.9
36.4
35.0
24.7
43.7
52.3
22.2
17.3
27.1
34.1
20.9
15.7
37.7
42.4
M
24.0
20.5
32.9
27.2
25.2
39.7
38.7
1.53
1.66
1.22
1.93
1.16
1.23
1.38
2.57
1.20
1.19
1.68
2.24
1.23
1.11
1.60
2.01
1.36
1.02
1.66
2.43
1.31
1.32
2.48
1.79
1.23
1.21
2.06
2.52
1.46
1.10
1.65
2.22
0.98
1.01
1.94
2.36
M
1.33
1.33
2.19
1.28
1.45
1.96
2.30
Figure 3. Descriptive statistics for one-hour ozone data obtained from
monitoring sites associated with St. Louis exposure districts.
-------
Mr. Tom McCurdy 9 15' 1993
DISTRICT YZAR P50 P90 P95 P99 P995 SECHI HIGH CLV DELTA
1
1
1
«i>
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
8
8
8
9
9
9
9
10
10
10
10
11
11
11
11
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
88
90
91
85
83
90
91
85
88
90
91
85
88
90
91
85
88
90
91
20
21
21
19
24
25
21
18
18
17
14
15
17
16
15
13
22
22
21
20
M
24
14
16
21
19
17
16
22
24
19
15
20
16
16
17
15
17
15
16
15
14
12
11
51
50
50
44
45
55
49
39
46
47
42
39
41
46
42
36
50
49
48
45
M
51
33
39
46
47
41
39
52
52
46
38
49
47
41
40
42
46
42
40
40
42
39
35
62
61
64
55
55
67
61
48
74
60
53
50
55
57
55
46
65
60
61
57
M
62
41
51
59
58
52
50
66
64
56
47
62
60
54
51
56
57
56
52
54
52
51
46
89
89
92
80
80
91
89
62
84
88
84
71
86
84
82
69
97
87
92
88
M
89
60
81
93
84
79
77
100
87
84
71
92
86
81
72
87
80
86
79
87
80
81
70
99
96
104
91
92
102
96
67
96
96
95
82
102
94
.. 96
76
107
99
105
100
M
96
71
91
106
94
90
89
111
97
92
82
102
92
90
81
102
" 87
97
89
97
95
92
78
136
176
149
142
134
164
124
85
164
169
151
119
209
149
156
110
162
129
167
158
M
131
110
146
135
132
154
109
225
124
139
101
192
146
144
124
202
130
156
127
220
136
160
116
140
179
150
144
134
165
124
85
166
171
152
121
210
150
164
111
164
130
170
160
M
135
112
149
135
134
155
110
227
127
140
103
196
151
146
126
206
131
157
127
226
139
164
116
145
155
162
149
149
164
137
93
167
150
161
127
199
150
171
116
170
140
179
162
M
135
131
153
155
140
165
134
208
134
143
120
179
138
154
133
205
136
171
136
186
162
167
120
29.6
24.5
26.4
20.2
19.9
25.5
32.2
25.7
18.1
25.5
18.5
20.2
13.3
23.0
17.0
20.4
27.8
29.9
22.1
21.1
M
32.8
11.3
20.0
28.5
26.4
15.6
20.5
18.8
33.5
25.5
20.8
20.9
29.1
19.0
19.4
13.3
24.7
19.1
23.1
16.2
16.8
16.8
20.8
1.39
1.20
1.22
1.10
1.10
1. 18
1.53
1.72
0.99
1.25
1.02
1.20
0.81
1.18
0.96
1.27
1.22
1.43
1.06
1.08
M
1.56
0.90
1.09
1.30
1.32
0.94
1.17
0.92
1.59
1.28
1.26
1.03
1.42
1.05
1.15
0.81
1.29
1.01
1.24
0.90
0.97
0.96
1.26
Figure 4. Descriptive statistics for eight-hour ozone data obtained from
monitoring sites associated with Houston exposure districts.
-------
Mr. Tom McCuidy
10
April 15, 1993
DISTRICT YZAR P50 P90 P95 P99 P995 SECHI HIGH CLV DELTA
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
8
8
8
9
9
9
9
10
10
10
10
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
83
84
86
91
33
33
29
34
31
26
27
28
29
32
38
37
34
31
29
33
25
23
25
28
31
28
29
30
25
22
24
26
31
32
27
31
19
19
17
21
27
27
27
27
70
64
59
67
71
60
64
64
63
66
69
71
73
65
62
68
63
55
60
64
65
56
61
62
57
49
55
60
66
58
61
65
51
43
42
49
65
59
60
61
85
75
68
80
87
73
77
76
74
76
80
81
87
77
75
80
78
69
71
76
79
66
71
72
67
58
66
70
80
69
71
76
61
54
52
60
80
71
71
72
106
98
91
107
109
102
101
101
97
99
102
103
114
100
93
105
106
93
89
100
101
86
92
92
88
73
84
92
104
87
89
96
86
72
69
79
105
91
93
97
112
106
101
114
113
114
..112
109
107
103
107
107
128
107
99
113
112
106
97
104
107
92
97
98
96
78
88
98
111
" 92
94
100
95
81
76
86
115
100
96
103
129
129
113
138
130
132
129
129
126
114
144
124
174
120
119
135
134
159
132
115
140
117
113
113
120
125
102
118
132
119
111
125
106
101
101
112
127
131
115
116
129
129
113
141
131
133
131
131
127
114
146
125
177
120
120
135
135
161
132
116
142
120
115
114
123
126
104
118
137
119
1.13
127
106
104
102
114
129
131
115
116
135
139
123
142
133
150
144
136
136
123
140
128
189
130
122
138
143
166
132
120
141
120
119
118
123
117
109
123
140
121
116
128
119
112
106
116
139
136
120
126
58.4
41.9
43.2
50.7
64.0
36.7
42.7
47.9
41.6
53.6
45.9
57.7
32.8
50.0
46.2
50.9
49.8
22.8
35.2
61.0
44.9
38.1
47.4
49.4
36.8
23.3
43.6
42.2
48.9
38.5
46.2
45.1
36.7
24.4
23.1
30.1
53.1
33.9
48.1
51.2
2.57
1.79
2.05
2.08
2.94
1.52
1.76
2.05
1.81
2.57
1.92
2.70
1.23
2.24
2.20
2.15
2.04
1.08
1.62
3.16
1.88
1.87
2.34
2.47
1.78
1.33
2.34
2.01
2.04
1.88
2.33
2.05
1.82
1.41
1.41
1.59
2.24
1.55
2.34
2.39
Figure 5. Descriptive statistics for eight-hour ozone data obtained from
monitoring sites associated with Philadelphia exposure districts.
-------
Mr. Tom McCurdy 11 April 15, 1993
DISTRICT YEAR P50 P90 P95 P99 B995 SECHI HIGH CLV DELTA K
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
6
7
7
7
7
8
8
8
8
9
9
9
9
10
10
10
10
11
11
11
11
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
83
85
90
91
25
24
20
18
12
19
26
27
27
25
30
29
31
33
24
27
28
25
25
25
30
29
19
27
29
24
29
30
15
20
19
18
9
20
25
24
M
24
19
18
17
18
25
23
56
49
43
43
34
41
50
55
60
49
54
58
62
57
44
50
64
51
49
56
65
57
39
52
64
49
54
57
30
41
40
38
35
40
48
49
M
46
37
39
42
42
50
45
67
57
51
51
41
48
59
66
72
57
64
70
73
63
50
58
77
59
58
66
78
66
44
60
77
57
64
66
39
47
48
45
49
48
57
57
M
54
43
45
54
50
60
53
87
69
66
66
55
63
78
81
98
78
80
89
95
84
62
72
104
77
76
84
105
81
54
75
106
76
81
85
56
61
60
59
88
61
73
73
M
66
56
59
79
63
76
67
93
74
70
71
61
66
85
87
109
84
85
95
105
97
67
77
113
84
80
89
113
88
56
82
115
83
84
90
59
66
65
63
101
68
77
77
M
72
61
61
85
69
83
- 72
111
93
98
89
77
94
110
102
142
115
100
109
141
123
85
86
150
122
93
101
138
120
62
94
170
109
101
101
75
100
76
74
137
97
89
88
M
101
83
73
113
92
99
80
112
95
99
90
78
97
110
104
145
116
103
110
143
123
86
87
153
122
94
103
138
122
63
95
171
110
104
102
75
102
77
74
139
99
91
88
M
103
85
74
113
93
100
81
120
99
100
91
86
96
116
106
147
116
106
116
143
139
87
94
155
123
97
107
153
126
65
100
164
117
103
106
76
97
79
78
156
103
91
93
M
103
84
76
116
95
104
86
39.4
28.1
23.1
29.3
17.8
21.2
29.4
43.2
36.5
28.2
41.4
46.4
35.5
25.0
27.7
36.5
39.4
24.2
39.4
45.5
40.5
27.5
34.3
37.5
35.4
26.3
44.1
48.9
24.5
20.0
31.2
29.7
19.5
17.6
42.5
39.4
M
22.7
20.7
30.0
31.1
22.3
34.9
35.9
1.93
1.70
1.47
1.90
1.36
1.42
1.56
2.39
1.54
1.52
2.28
2.35
1.54
1.25
1.88
2.27
1.57
1.32
2.37
2.50
1.61
1.41
3.38
2.19
1.40
1.44
2.53
2.78
1.90
1.36
2.31
2.22
1.03
1.22
2.80
2.51
M
1.42
1.53
2.31
1.63
1.48
1.96
2.45
Figure 6. Descriptive statistics for eight-hour ozone data obtained from
monitoring sites associated with St. Louis exposure districts.
-------
Mr. Tom McCurdy 12 April 15, 1993
CLV: characteristic largest value
DELTA: 6 value of fitted Weibull distribution
K: k value of fitted Weibull distribution.
M: missing value.
The statistics listed in Figures 1 through 6 were the primary source of data used in developing
the CLVAP's discussed below.
Descriptive Statistics for CLV1 by City-Year
The CLV1 values associated with a particular city-year form an empirical distribution. We
attempted to characterize these distributions by fitting normal and lognormal distributions to
each empirical distribution. Table 1 lists (1) the arithmetic mean (AM) and standard
deviation (ASD) associated with each normal distribution and (2) the geometric mean (GM)
and standard deviation (GSD) associated with each lognormal distribution. Note that the
results listed for St. Louis, 1983, were determined after the removal of two outliers (the two
lowest values). If these outliers had been retained, the AM and ASD would have been 163.6
ppb and 36.2 ppb, respectively. The GM and GSD would have been 159.3 ppb and 1.286,
respectively.
In the listing for each city-year, Table 1 also provides the median CLV1 value, the maximum
CLV1 value (MAXCLV1), the normal z value of the MAXCLV1, and the lognormal z value
of the MAXCLV1. The normal z value was calculated as follows:
normal Z of MAXCLV1 = [(MAXCLV1) - (AM)]/ASD. (9)
The lognormal z value was calculated by the expression
lognormal z of MAXCLV1 = [ln(MAXCLVl)-ln(GM) ]/ln(GSD). (10)
Table 1 also lists the relative standard deviation (RSD), the ratio of MAXCLV1 to AM, and
the ratio of MAXCLV1 to GM.
Observed Patterns in CLV1 Data
We examined Table 1 to determine whether there were any patterns in the year-to-year
listings for each city. The following patterns were noted:
-------
Table 1
Descriptive Statistics for Characteristic Largest One-Hour Values Associated with Ozone
Monitoring Sites in Houston, Philadelphia, and St. Louis
City
Houston
Philadelphia
St. Louis
Year
1985
1988
1990
1991
Mean
1983
1984
1986
1991
Mean
1983
1985
1990
1991
Mean
n
10
11
11
11
10
10
10
10
8"
11
11
11
Characteristic largest values, ppb
Arithmetic
Mean
234.5
198.3
214.5
178.3
172.5
164.0
145.2
145.7
178.4
134.4
114.5
110.4
S.D.
29.61
19.96
20.79
26.75
15.89
21.62
12.01
12.05
20.41
18.99
17.43
14.63
Geometric
Mean
232.8
197.4
213.6
176.3
171.9
162.9
145.0
145.3
177.3
133.2
113.2
109.5
S.D.'
1.135
1.104
1.105
1.173
1.129
1.094
1.132
1.087
1.085
1.100
1.128
1.151
1.177
1.145
1.150
Median
232.0
191.0
224.0
177.0
170.0
153.5
145.5
143.0
182.5
132.0
119.0
110.0
Maxi-
mum
280
238
241
219
207
214
164
167
201
164
141
132
Relative
std. dev.
0.126
0.101
0.097
0.150
0.119
0.092
0.132
0.083
0.083
0.098
0.114
0.141
0.152
0.133
0.135
Ratio of
MAXCLV to
AM
1.19
1.20
1.12
1.23
1.19
1.20
1.30
1.13
1.15
1.20
1.13
1.22
1.23
1.20
1.20
Ratio of
MAXCLV
to GM
1.20
1.21
1.13
1.24
1.20
1.20
1.31
1.13
1.15
1.20
1.13
1.23
1.25
1.21
1.21
Z value of
maximum
Arith.
1.54
1.99
1.27
132
2.17
231
1.57
1.77
1.11
136
1.52
1.48
Geom.
1.46
1.89
1.21
1.36
2.07
2.20
1.48
1.71
1.04
1.48
135
138
'Dimensionless quantity.
bOmits two site-years.
-------
Mr. Tom McCurdy 14 April 15, 1993
1. The ASD is moderately correlated with AM (R squared = 0.474).
2. The GSD (a dimensionless quantity) is not correlated with GM (R squared =
0.029) and is relatively constant within each city. The four-year mean GSD for
each city is listed below.
Houston: 1.129
Philadelphia: 1.100
St Louis: 1.150
3. The RSD is not correlated with AM (R squared = 0.065) and is relatively
constant within each city. The four-year mean RSD for each city is listed
below.
Houston: 0.119
Philadelphia: 0.098
St Louis: 0.135
4. The MAXCLV1 is highly correlated with AM (R squared = 0.960). The
regression equation (n = 12) is
MAXCLV1 = 7.57 ppb + (1.14)(AM).
The intercept is not significant.
5. The MAXCLV1 is highly correlated with GM (R squared = 0.958). The
regression equation (N = 12) is
MAXCLV1 = 8.29 ppb + (1.14)(GM).
The intercept is not significant.
6. The MAXCLV1 is highly correlated with the median (R squared = 0.903). The
regression equation (n = 12) is
MAXCLV1 = 16.16 ppb + (1.10)(median).
The intercept is not significant.
-------
Mr. Tom McCurdy 15 April 15, 1993
7. The ratio of MAXCLV1 to AM is relatively constant within and between cities.
The four-year mean ratios are listed below.
Houston: 1.19
Philadelphia: 1.20
SL Louis: 1.20
8. The ratio of MAXCLV1 to GM is relatively constant within and between cities.
The four-year mean ratios are listed below
Houston: 1.20
Philadelphia: 1.20
St Louis: 1.21
Two Candidate Models for Adjusting CLVl's
We have developed two general models for adjusting CLVl's which are consistent with these
patterns. Model A assumes that
a) the CLVl's for a given city-year follow a lognormal distribution,
b) the GSD of the lognormal distribution is constant from year-to-year, and
c) the z value associated with the CLV1 of the r-th ranked site is constant from
year-to-year.
Model B assumes that
a) the one-hour CLVl's for a given city-year follow a normal distribution,
b) the RSD of the normal distribution is constant from year-to-year, and
c) the z value associated with the CLV1 of the r-th ranked site is constant from
year-to-year.
-------
Mr. Tom McCurdy 16 April 15, 1993
Evaluation of Model A With Respect to the Observed Patterns
To evaluate Model A, we have defined ZMAXA as the z value of the MAXCLV1 with
respect to the lognormal distribution. The relationship between MAXCLV1 and GM is
expressed by the lognormal equation
\
MAXCLV1 = (GAT) (GSD) *"***. (11)
If GSD and ZXAXA are constant from year to year as assumed, then the ratio of MAXCLV1
to GM is constant and equal to GSD2"***. This relationship is consistent with Patterns 5 and
8.
The relationship between GM and AM for a lognormal distribution is
AM = (GM)[exp(a2)]0-5, (12)
where a = In(GSD). Note that GM is being multiplied by a term that is determined solely by
the value of GSD. If GSD is constant year-to-year, then the ratio of AM to GM is constant.
Consequently, the ratio of MAXCLV1 to AM is constant. This relationship is consistent with
Patterns 4 and 7 noted above.
The median of a lognormal distribution is equal to the GM. This relationship supports
Pattern 6.
The RSD of a lognormal distribution can be expressed as
ASD/AM = RSD = [exp(o2) - I]0-5, (13)
where a = In(GSD). Note that RSD is determined solely by GSD. If GSD is constant year-
to-year, RSD will be constant year-to-year and ASD will be correlated with AM (Patterns 3
and 2). The average GSD for the 12 city-years is 1.126. Inserting this value in the equation
above yields an RSD value of 0.119. The average of the 12 RSD values in Table 7 is 0.117,
a difference of less than 2 percent.
These results suggest that Model A is consistent with all eight patterns noted in Table 7.
Evaluation of Model B With Respect to the Observed Patterns
Model B assumes that the CLV1 values are normally distributed and that RSD is constant
year-to-year. To evaluate Model B, we have defined ZMAXB as the z value of the
MAXCLV1 with respect to the normal distribution. The relationship between MAXCLVl
and AM is expressed by the normal equation
-------
Mr. Tom McCurdy 17 April 15, 1993
MAXCLV1 = AM + (ASD) (ZMAXB) .
(14)
Because ASD = (RSD)(AM), this expression is equivalent to
MAXCLV1 = (AM)[1 + (RSD)( ZMAXB)]. <15)
If RSD and ZXAXB are constant from year to year as assumed, then the ratio of MAXCLV1
to AM is constant and equal to the far-right term [1 + (RSD)(ZMAXB)]. This relationship is
consistent with Patterns 5 and 8.
The relationships among GM, AM, and the median for a normal distribution can be expressed
by the simple identity
AM = GM = median. (lfi)
If the ratio of MAXCLV1 to AM is constant year-to-year, then the ratio of MAXCLV1 to
GM and the ratio of MAXCLV1 to median are also constant year-to-year. This behavior is
consistent with Patterns 6 and 8.
Although CLVl's are by definition positive values, a normal distribution fit to these values
will always extend below zero, at least in theory. As GSD's cannot be determined for
theoretical distributions containing negative values, there is no formula for determining GSD
from the parameters of a normal distribution. Consequently, we cannot predict the behavior
of GSD from the assumptions concerning AM and ASD. There is no way that we can
evaluate Model B according to Pattern 2 (i.e., GSD is constant year-to-year).
We recommend that Model A be incorporated into the adjustment procedure. Model A does
not have the theoretical problem of negative values and is consistent with all eight patterns.
Year-to-Year Variability in Identity of Highest Ranked Site
MAXCLV1 is defined as the maximum CLV1 of a particular city for a specific year.
Associated with MAXCLV1 is a value for ZMAXA. Under Model A, ZMAXA is assumed
to remain constant from year to year, although GM may change. It should be noted that we
have not assumed that ZMAXA is associated with the same site each year. The data in
Figures 1, 2, and 3 demonstrate that the identity of the "MAXCLV1 site" can change from
year to year. For example, the MAXCLV1 site in Houston is associated with the District 8
site in 1985, the District 11 site in 1988, the District 10 site in 1990, and the District 5 site in
1991.
-------
Mr. Tom McCurdy 18 April 15, 1993
Ideally, we would like to assume that the ranking of all sites in a city with respect to CLV1
is constant from year to year. The validity of this assumption varies with city. Table 2 lists
the ranking of the sites associated with each city in each of the four years selected for
analysis. (Note that the site with the largest value of CLV1 is assigned a rank of 1.) The
table also shows the highest and lowest rank assigned to each site over the four years
included in the analysis. The mean rank is also provided.
The rankings associated with Houston display the most random behavior. Site 8 is ranked
No. 1 and No. 11. Site 5 is ranked No. 1 and No. 10. Site 6 is ranked No. 2 and No. 11.
The rankings in any one year do not provide a good indication of the rankings in another
year. For example, the Spearman rank correlation between the CLV1 values of Year A
(1985) - the year with the largest CLV1 (280 ppb) - and the CLV1 values of Years B, C,
and D is -0.1394 for Year B (1988), 0.1398 for Year C (1990), and -0.1879 for Year D
(1991). The ranks in Years B, C, and D are better predicted by the mean rank of the other
three years than by Year A alone. The Spearman rank coefficient.between the rank in each
of these years and the mean of the ranks of the other three years is 0.1216 for Year B, 0.7853
for Year C, and -0.0547 for Year D.
The least random behavior is displayed by the rankings associated with Philadelphia. For
example, the Spearman rank correlation between the CLV1 values of Year B — the year with
the highest CLV1 (214 ppb) ~ and the CLV1 values of Years A, C, and D is 0.4893 for Year
A, 0.2553 for Year C, and 0.2918 for Year D. Although these correlations are stronger than
those associated with Houston, but they are still relatively weak in an absolute sense. The
ranks in Years A, C, and D are better predicted by the mean rank of the other three years
than by Year B alone. The Spearman rank coefficient between the rank in each of these
years and the mean of the ranks of the other three years is 0.4268 for Year A, 0.4909 for
Year C, and 0.5289 for Year D.
A Procedure for Predicting Future CLVl's Under One-Hour NAAQS Attainment
At this point, the evidence suggests that the future ranking of a site under a particular one-
hour NAAQS scenario is better predicted by the average ranking of that site over a multi-year
period than by the ranking of the site in a randomly selected year. The following procedure
for predicting future CLV1 values under 1-hr NAAQS attainment conditions is based on
combining this assumption with Model A above.
1. Determine the following quantities.
CLVl(ij): the CLV1 of i-th ranked site in City j for the "baseline" or
"start" year.
-------
Mr. Tom McCurdy
19
April 15, 1993
Table 2
Ranks of Characteristic Largest One-Hour Values Associated with
Ozone Monitoring Sites Representing Exposure Districts
in Houston, Philadelphia, and St. Louis
City
Houston
Philadelphia
St. Louis
District
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
11
Rank by year*
A
10
9
7
2
6
b
8
lc
5
3
4
7
6
8
1
2
4
10
5
7
3
7
9
4
6
3
2
r
10
5
a
8
B
5
4
2
3
10
7
9
11
8
6
1
4
2
10
6
lc
8
9
5
7
3
9
10
5
1
2
3
4
7
6
8
11
C
6
10
1
5
4
11
7
8
9
2
3
3
1
2
6
4
5
10
7
9
8
3
1
2
8
5
11
4
10
7
9
6
D
4
11
8
9
1
2
5
10
6
3
7
1
4
3
2
5
9
8
7
10
6
8
5
1
6
2
4
3
10
7
11
9
Ranks over 4 years
High
4
4
1
2
1
2
5
1
5
2
1
1
1
2
1
1
4
8
5
7
3
3
1
1
1
2
2
1
7
5
8
6
Low
10
11
8
9
10
11
9
11
9
6
7
7
6
10
6
5
9
10
7
10
8
9
10
5
8
5
11
4
10
7
11
11
Mean
6.3
8.5
4.5
4.8
5.3
6.7
7.3
7.5
7.0
3.5
3.8
3.8
3.3
5.8
3.8
3.0
6.5
9.3
6.0
8.3
5.0
6.8
4.0
3.0
5.3
3.0
5.0
3.0
9.3
6.3
9.3
8.5
'Houston: A = 1985, B = 1988, C = 1990, D = 1991
Philadelphia: A = 1983, B = 1984, C = 1986, D = 1991
St. Louis: A = 1983, B = 1985, C = 1990, D = 1991.
blnsufficient data to determine rank.
"Site-year associated with largest ACLV for indicated city.
-------
Mr. Tom McCurdy 20 April 15, 1993
MAXCLV1Q'): the largest CLV1 of all sites in City j for the baseline
year.
AMAXCLVIQ): the largest CLV1 value permitted under the proposed 1-hr
NAAQS.
2. Select five years prior to the baseline year and determine the value of CLV1
(or related air quality indicator) at each site m in City j for each year. Rank
these values by city and year. Let RANK(m,j,y) indicate the rank of site m in
city j in year y. Let MEANRANK(m,j) indicate the mean value of
RANK(mj,y) over the n years. Rank the MEANRANK(m,j) values and let
RELRANK(m,j) indicate the relative rank of MEANRANK(m,j).
3. Calculate an adjusted CLV1 for the i-th ranked site in City j by the expression
ACLVl(i,j) = [CLVl(JLrj)}[AMAXCLVl(j)]/[MAXCLVl(j)].
4. If RELRANK(m,j) = i, then m will be the i-th ranked site in City j under
attainment. That is,
ACLVl(m,j) = ACLVl(ij) if RELRANK(m,j) = i.
5. The 1-hour data at Site m under attainment will be determined by adjusting the
1-hour data at Site m in the baseline year. A Weibull distribution fit to the
adjusted data will have a CLV1 equal to ACLVl(i j) where i =
RELRANK(m,j). (A method for estimating the parameters of this Weibull
distribution will be discussed later in this letter.)
The procedure assumes that the CLVl's of City j are lognormally distributed with GSD =
GSD(j) in both the baseline and attainment years. Equation 17 in Step 3 follows directly
from the Model A assumptions that the z value of the i-th ranked CLV1 is constant. We will
allow the identity of the site associated with the i-th ranked CLV1 to change, however. We
identify the i-th ranked site under attainment by first averaging the ranks of each site for five
years near to and including the baseline year. These ranks can be determined by CLV1 or by
a related air quality indicator (e.g., the observed second highest daily maximum one-hour
concentration). The site with the i-th ranked average will be assigned the i-th ranked CLV1
determined for the attainment year.
-------
Mr. Tom McCurdy 21 April 15, 1993
Application of the CLV1 Adjustment Procedure to Philadelphia
We have applied the CLV1 adjustment procedure to Philadelphia. Table 3 presents the
results. The baseline year is 1991. The five-year ranking of each site is based on second-
high daily maximum one-hour concentrations, as these values are easier to obtain than
CLVl's. District 1 has the largest CLV1 for 1991 (156 ppb). The largest CLV1 must be
reduced to 120 ppb for Philadelphia to meet the current 1-hour NAAQS. Equation 17 is thus
ACLVl(i,j) = [CLVl(i,j)] (120/156) = [ CLVl(i, j) ] ( 0 . 769) .
Applying this expression to each 1991 CLV1, we determine 10 CLVl's for the attainment
year. These values are listed in the column labeled "adjusted CLV1". We then assign these
values to the districts according to the five-year ranking determined for each district. Thus,
the largest adjusted CLV1 (120 ppb) is assigned to District 4 because District 4 has the
highest five-year ranking. Similarly, the second largest adjusted CLV1 (116 ppb) is assigned
to District 3 because District 3 has the second highest five-year ranking.
A Procedure for Predicting Future CLVS's Under Eight-Hour NAAQS Attainment
Table 4 follows the same format as Table 1 to present descriptive statistics for the eight-hour
data listed in Figures 4, 5, and 6. We examined Table 4 to determine whether there were any
patterns in the year-to-year listings for each city. The following patterns were noted:
1. The ASD is weakly correlated with AM (R squared = 0.253).
2. The GSD (a dimensionless quantity) is not correlated with GM (R squared =
0.125) and is relatively constant within each city. The four-year mean GSD for
each city is listed below.
Houston: 1.122
Philadelphia: 1.109
St Louis: 1.149
Note that these city means are almost identical to the corresponding city means
listed in Table 1 for CLVl's.
-------
Mr. Tom McCurdy
22
April 15, 1993
Table 3
Application of Adjustment/Reassignment Procedure to Characteristic
Largest One-Hour Values Associated with Monitoring Sites
Representing Exposure Districts in Philadelphia
District
1
2
3
4
5
6
7
8
9
10
1991
CLV1, ppb
167
149
153
162
145
134
135
140
131
141
1991
ranking
1
4
3
2
5
9
8
7
10
6
Five-year
ranking
4.5
3.0
2.0
1.0
4.5
7.0
10.0
6.0
9.0
8.0
Adjusted
CLV1, ppb
120
107
110
116
104
96
97
101
94
101
Reassigned attain-
ment CLV1, ppb
107
110
116
120
104
101
94
101
96
97
-------
Table 4
Descriptive Statistics for Characteristic Largest Eight-Hour Values Associated with Ozone
Monitoring Sites in Houston, Philadelphia, and St. Louis
City
Houston
Philadelphia
St. Louis
Year
1985
1988
1990
1991
Mean
1983
1984
1986
1991
Mean
1983
1985
1990
1991
Mean
n
10
11
11
11
10
10
10
10
8"
11
11
11
Characteristic largest values, ppb
Arithmetic
Mean
1763
145.8
158.3
131.2
139.8
131.4
123.1
127.5
144.2
110.4
93.8
95.7
S.D.
22.97
10.93
1531
1932
18.95
16.72
12.34
8.78
1738
14.69
14.41
12.64
Geometric
Mean
174.9
145.5
157.6
129.8
138.8
130.5
122.6
127.2
1433
1095
92.8
94.9
S.D.'
1.140
1.077
1.105
1.166
1.122
1.132
1.130
1.104
1.071
1.109
1.135
1.138
1.177
1.144
1.149
Median
174.5
140.0
162.0
133.0
137.5
126.5
121.0
127.0
150.0
103.0
97.0
94.0
Maxi-
mum
208
164
179
162
189
166
144
142
164
139
116
116
Relative
std. dev.
0.130
0.075
0.097
0.147
0.112
0.136
0.127
0.100
0.069
0.108
0.121
0.133
0.154
0.132
0.135
Ratio of
MAXCLV
to AM
1.18
1.12
1.13
1.23
1.17
135
1.26
1.17
1.11
1.22
1.14
1.26
1.24
1.21
1.21
Ratio of
MAXCLV
toGM
1.19
1.13
1.14
1.25
1.18
136
1.27
1.17
1.12
1.23
1.15
1.27
1.25
1.22
1.22
Z value of
maximum
Arith.
138
1.67
135
159
2.60
2.07
1.69
1.65
1.14
1.95
154
1.61
Geom.
132
1.61
1.28
1.44
2.49
1.97
1.63
1.60
1.07
1.85
137
1.49
'Dimensionless quantity.
bOmits two site-years.
-------
Mr. Tom McCurdy 24 April 15, 1993
3. The RSD is not correlated with AM (R squared = 0.101) and is relatively constant
within each city. The four-year mean RSD for each city is listed below.
Houston: 0.112
Philadelphia: 0.108
St. Louis: 0.135
Note that these city means are almost identical to the corresponding city means listed
in Table 1 for CLVl's.
4. The MAXCLV8 is highly correlated with AM (R squared = 0.885). The regression
equation (n = 12) is
MAXCLV8 = 15.8 ppb + (1.08)(AM).
The intercept is not significant.
5. The MAXCLV8 is highly correlated with-GM (R squared = 0.879). The regression
equation (N = 12) is
MAXCLV8 = 16.8 ppb + (1.08)(GM).
The intercept is not significant.
6. The MAXCLV8 is highly correlated with the median (R squared = 0.832). The
regression equation (n = 12) is
MAXCLV8 = 25.1 ppb + (1.01)(median).
The intercept is not significant.
7. The ratio of MAXCLV8 to AM is relatively constant within and between cities. The
four-year mean ratios are listed below.
Houston: 1.17
Philadelphia: 1.22
St Louis: 1.21
-------
Mr. Tom McCurdy 25 April 15, 1993
Note that these city means are almost identical to the corresponding city means listed
in Table 1 for CLVl's.
8. The ratio of MAXCLV8 to GM is relatively constant within and between cities. The
four-year mean ratios are listed below
Houston: 1.18
Philadelphia: 1.23
SL Louis: 1.22
Note that these city means are almost identical to the corresponding city means listed
in Table 1 for CLVl's.
These patterns are similar to those identified for the CLV1 statistics in Table 1. Consequently, these
patterns support eight-hour versions of Models A and B. We recommend that the CLV8 adjustment
procedure incorporate Model A, as this was the model selected for the CLV1 adjustment procedure.
Because CLV1 and CLV8 are highly correlated, we can expect the year-to-year rankings of CLVS's
to behave in a similar manner to those of CLVl's; that is, the future ranking of a site under a
particular eight-hour NAAQS scenario is better predicted by the average ranking of that site over a
multi-year period than by the ranking of the site in a randomly-selected year. The rankings can be
determined by either one-hour or eight-hour data, as we expect the rankings to be essentially the
same by either method.
The following procedure for predicting future CLV8 values under 8-hr NAAQS attainment conditions
follows the procedure outlined above for 1-hr NAAQS attainment.
1. Determine the following quantities.
CLV8(ij): the eight-hour CLV of i-th ranked site in City j for the
"baseline" or "start" year.
MAXCLV8(j): the largest CLV8 of all sites in City j for the baseline year.
AMAXCLV8(j): the largest CLV8 value permitted under the proposed 8-hr
NAAQS.
2. Select five years prior to the baseline year and determine the value of CLV8 (or
related air quality indicator) at each site m in City j for each year. Rank these values
by city and year. Let RANK(mj,y) indicate the rank of site m in city j in year y. Let
MEANRANK(m,j) indicate the mean value of RANK(mj,y) over the n years. Rank
-------
Mr. Tom McCurdy 26 April 15, 1993
the MEANRANK(m j) values and let RELRANK(m j) indicate the relative rank of
MEANRANK(m,j). Note: this ranking can be performed using CLVl's or related air
quality indicators.
3. Calculate an adjusted CLV8 for the i-th ranked site in City j by the expression
ACLV8(±,j) = [CLV8(±, j)}[AMAXCLV8(j)}/[MAXCLVS(j)} . (18)
4. If RELRANK(m,j) = i, then m will be the i-th ranked site in City j under attainment.
That is,
ACLV8(m,j) = ACLV8(i,j) if RELRANK(m,j) = i.
5. Using a suitable equivalence relationship, estimate the CLV1 associated with each
ACLV8(m,j) value. Denote this value as ACLVl(mj). (A recommended equivalence
relationship is provided later in this letter.)
6. The 1-hour data for Site m under attainment of the 8-hr NAAQS will be determined
by adjusting the 1-hour data for Site m in the baseline year. A Weibull distribution fit
to the adjusted data will have a CLV1 equal to ACLVl(ij) where i = RELRANK(m,j).
(A method for estimating the parameters of this Weibull distribution will be discussed
later in this letter.)
Application of the CLV8 Adjustment Procedure to Philadelphia
We have applied the procedure to Philadelphia with the assumption that an 8-hour NAAQS of 80
ppb has been attained. Table 5 presents the results using the same format as Table 3. As in the one-
hour example, the baseline year is 1991. The five-year ranking of each site is based on second-high
daily maximum one-hour concentrations, as these values were easier to obtain than CLVS's. District
1 has the largest CLV8 for 1991 (142 ppb). The largest CLV8 must be reduced to 80 ppb for
Philadelphia to meet a proposed 8-hour NAAQS. Equation 18 is thus
-------
Mr. Tom McCurdy
27
April 15, 1993
Table 5
Application of Adjustment/Reassignment Procedure to Characteristic
Largest Eight-Hour Values Associated with Monitoring Sites
Representing Exposure Districts in SL Louis
District
1
2
3
4
5
6
7
8
9
10
1991
CLV8, ppb
142
136
128
138
120
118
123
128
116
126
1991
ranking
1
3
4.5
2
8
9
7
4.5
10
6
Five-year
ranking
4.5
3
2
1
4.5
7
10
6
9
8
Adjusted
CLV8, ppb
80
77
72
78
68
66
69
72
65
71
Reassigned attain-
ment CLV8, ppb
72
77
78
80
72
69
65
71
66
68
-------
Mr. Tom McCurdy 28 April 15, 1993
ACLV8(i,j) = [CLV8(i,j) ](80/142) = [CLV8(i, j) ] (0.563 ).
Applying this expression to each 1991 CLV8, we determine 10 CLVS's for the attainment
year. These values are listed in the column labeled "adjusted CLV8". We then assign these
values to the districts according to the five-year ranking determined for each district. Thus,
the largest adjusted CLV8 (80 ppb) is assigned to District 4 because District 4 has the highest
five-year ranking. Similarly, the second largest adjusted CLV8 (72 ppb) is assigned to
District 3 because District 3 has the second highest five-year rank-ing.
STEP 4: ADJUST ONE-HOUR DATA
Simulation of One-Hour NAAQS
The CLV1 adjustment procedure discussed above provides an estimate of the CLV1
associated with each site under attainment conditions. To complete the adjustment process,
the one-hour values associated with the site must be adjusted to produce the designated
CLV1. If we assume that the adjusted data can be fit by a Weibull distribution, then we must
develop a method for estimating the parameters of this distribution.
Let 6 and k indicate the parameters of the Weibull distribution fitting the unadjusted 1-hour
data associated with a particular site. These parameters are related to the UCLV1 -- the
CLV1 of the unadjusted data — by the expression
UCLV1 = Stlnfn)]1'*. (19)
Let 6' and k' indicate the parameters of the Weibull distribution fitting the adjusted 1-hour
data. These parameters are related to the adjusted CLV1 (ACLV1) by the expression
ACLV1 = 6'[ln(n)]1/*'. (20)
We would like to find an prediction equation incorporating UCLV1, ACLV1, 6, and/or k
which can be used to estimate either 5' or k'. Given an estimate of 6', Equation 20 can be
used to estimate k'. Similarly, Equation 20 can be used to estimate 6' given an estimate of
k'.
To assist in developing the required prediction equation, we reviewed the 128 site-years of
one-hour data listed in Figures 1, 2, and 3. (As discussed above, we had previously fit a
Weibull distribution to each of the data sets.) We identified a pair of data sets associated
with each of 30 monitors such that the pair contained the highest and lowest year with respect
-------
Mr. Tom McCurdy 29 April 15, 1993
to CLV1. We assumed the high CLV1 year represented an unadjusted data set and the low
year represented an adjusted data set.
We performed a series of regression analyses in which the dependent variable was either the
6 or k value associated with the adjusted data set (i.e., 6' or k')- The independent variables
included various functions of the three parameters associated with the unadjusted data set (6,
k, and UCLV1) and various functions of ACLV1 -- the CLV1 of the adjusted data set. Based
on the results of these analyses, we are recommending that the following regression equation
be used to estimate k':
1/Jt' = -0.2389 + (0. 003367 )(ACLV1) + ( 0 .4726 ) ( 1/Jc) . (21)
Each coefficient of this equation was found to be statistically significant at the p = 0.10 level;
the over-all R squared value was 0.6993. Note that the equation assumes that ACLV1 will be
in units of ppb.
Given the ACLV1 and an estimate of k', we can estimate the corresponding 6' value by the
equation
6' = (ACLVl)/[In(n)]1/*>. (22)
We now have values for 6, k, UCLV1, 6', k', and ACLV1. If we treat the unadjusted data
set as a time series in which x, is the one-hour value at time t, we can construct a
corresponding adjusted data set by the expression
yt = (6')(V8)*/*/- <23>
where y, is the adjusted value at time t. This expression is based on the assumptions that the
time series y, at a site after attainment is related to the original time series x, in such a way
that (1) the rank of the one-hour value at each time t is unchanged, (2) the x, values follow a
Weibull distribution with parameters 6 and k, and (3) the y, values follow a Weibull
distribution with parameters 6' and k'. The latter two assumptions have already been
discussed.
The first assumption is made primarily as a means of incorporating the general seasonal and
diurnal patterns of the x, series into the y, series. We are assuming that the time series
"profile" of the x, series reflects the interactions of meteorology, precursors, and decay
mechanisms as they currently occur. In the absence of reliable information on the magnitude
and interactions of these factors under a hypothetical scenario, we are assuming the "profile"
will not change. The absolute value of the ozone concentration at each time t may change,
but the relative ranking of the values will not change. Peak values will still occur on the
same days at the same time of day. In a sense, we are assuming that the basic temporal
patterns in meteorology, precursors, and decay mechanisms will not change significantly
-------
Mr. Tom McCurdy 30 April 15, 1993
under any projected regulatory scenario. What does change is shape of the distribution of
hourly average values. Equations 21 and 22 determine the parameters of the new distribution.
Equation 23 can be restated as
*t)d (24)
c = (60 /(b)u* (25)
d = k/k'
We will refer to c and d as "adjustment coefficients."
To illustrate the adjustment procedure for one-hour values, we have applied it to the 1991
Philadelphia data sets listed in Table 6. The table lists the k value associated with the
unadjusted data set and the adjusted CLV1 expected under attainment of the current NAAQS
(120 ppb). Through the use of Equation 21, we determined a value of k' for each data set
Through the use of Equation 22, we next determined a value for 6'. These values were then
used to determine the adjustment coefficients (c and d). These values are listed in the far-
right columns of Table 6.
Simulation of 8-Hour NAAQS
The CLV8 adjustment procedure discussed above provides an estimate of the CLV8
associated with each site under attainment conditions. To complete the adjustment process,
the one-hour values associated with the site should be adjusted so that a Weibull distribution
fit to the resulting eight-hour values will yield the designated CLV8. We can simplify this
process by adjusting the one-hour values so that a Weibull distribution fit to the one-hour
values will yield a CLV1 that is equivalent to the desired CLV8. Given this CLV1 and the k
value of the Weibull distribution fit to the unadjusted one-hour data, we can use Equations 21
and 22 to determine k' and 6' for the adjusted one-hour Weibull distribution. The original
one-hour data are then adjusted using Equation 23.
In our letter of March 1, 1993, we provided a series of equivalence expressions relating
CLV8 to CLV1 (referred to as EHCLV and OHCLV, respectively, in the letter). The
expressions were based on regression analyses performed on the 96 site-years of data we will
be using to represent the "as is" (1990-91) air quajity for the nine-city pNEM/O3 analysis.
The simple expression
CLV1 = (CLVS)1-0466 (26)
-------
Table 6
Determination of Adjustment Coefficients for One-Hour NAAQS
Attainment (CLV1 = 120 ppb) in Philadelphia
District
1
2
3
4
5
6
7
8
9
10
Weibull fit to 1991 1-hr data
k
1.69
2.21
1.96
1.81
2.28
2.23
1.93
2.14
1.74
2.26
6
46.9
56.4
51.0
493
56.6
51.2
443
51.2
38.1
54.5
CLV1
167
149
153
162
145
134
135
140
131
141
1-hr NAAQS attainment parameters'
Adjusted CLV1
120
107
110
116
104
96
97
101
94
101
Reassigned CLV1
107
110
116
120
104
101
94
101
96
97
k*
2.494
2.896
2.546
2.346
3.139
3.194
3.101
3.106
2.809
3369
6'
45.27
52.44
49.95
48.09
52.51
51.60
47.06
50.62
44.74
5131
Adjustment coefficients
c
3336
2.417
2.420
2377
2.800
3305
4.447
3361
4.694
3511
d
0.678
0.763
0.770
0.772
0.726
0.698
0.622
0.689
0.619
0.671
'Assumes maximum CLV1 equals 120 ppb.
-------
Mr. Tom McCurdy 32 April 15, 1993
performed particularly well (R squared = 0.889). For CLV8 < 110 ppb, this exponential
relationship is very closely fit by the linear relationship
CLV1 = (1.213)(CLVS) . (27)
The term 1.2313 is the median ratio of CLV1 to CLV8 for the 96 site-years of data.
We examined the 96 site-specific ratios of CLV1 to CLV8 by city and found that the ratios
varied more among cities than among sites within a city. We also examined the CLVl-to-
CLV8 ratios in the three-city multiyear data base (Houston, Philadelphia, and St. Louis)
described above and found that the average city ratio was either constant year to year or else
showed a slight decreasing trend. For example, the median ratio for Houston was relatively
constant; the ratio was 1.334 for 1985, 1.370 for 1988, 1.374 for 1990, and 1.352 for 1991.
The median ratio for Philadelphia displayed an overall decline; the ratio was 1.240 for 1983,
1.249 for 1984, 1.184 for 1986, and 1.142 for 1992. We recommend that (1) a single average
CLVl-to-CLV8 ratio be applied to all sites within a particular city and (2) the ratio be
determined by a recent year of data specific to the city.
Table 7 lists the mean and median ratios for each city based on the 1990-91 nine-city data
base. The table also lists mean and median ratios for three other years for three of the cities
(Houston, Philadelphia, and St. Louis). We recommend that the flagged median ratios in this
table be used for the city-specific ratios. Each median is associated with a recent year (1990
or 1991). The medians are preferred to the means, as they are less affected by outliers.
Table 8 demonstrates the application of the adjustment procedure to Philadelphia for an eight-
hour NAAQS which permits a maximum CLV8 of 80 ppb. Under "as is" conditions the
highest CLV8 is 142 ppb. We are interested in simulating attainment of a NAAQS which
permits a CLV8 no higher than 80 ppb. Each CLV8 is first multiplied by 80ppb/142ppb as
indicated by Equation 18. These adjusted CLVS's are then reassigned according to the
relative rank of each site for the five-year period. The reassigned CLVS's are next multiplied
by 1.132, the median CLVl-to-CLV8 ratio listed for Philadelphia in Table 7. This step
provides estimates of equivalent CLVl's which are then used to determine adjustment
coefficients through the use of Equations 21 through 26.
ADJUSTMENT COEFFICIENTS FOR NINE CITIES
We have developed four sets of adjustment coefficients (c and d) for each of the monitoring
sites associated with the nine cities. The coefficients can be used to adjust the "as is" data
associated with each site to simulate attainment of the following NAAQS scenarios:
-------
Mr. Tom McCurdy
33
April 15, 1993
Table 7
Ratios of CLV1 to CLV8 for Nine Cities Based on Recent Ozone Data
City
Chicago
Denver
Houston
Los Angeles
Miami
New York
Philadelphia
St. Louis
Washington
Houston
(other years)
Philadelphia
(other years)
St Louis
(other years)
Data year
1991
1990
1990
1991
1991
1991
1991
1990
1991
1985
1988
1991
1983
1984
1986
1983
1985
1991
Number of sites
12
7
11
16
6
12
10
11
11
10
11
11
10
10
10
10
11
11
CLVl-to-CLV8 ratio
Mean
1.158
1.272
1.376
1.443
1.251
1.180
1.142
1.221
1.173
1.331
1.359
1.359
1.240
1.249
1.184
1.248
1.217
1.153
Std. dev.
0.024
0.115
0.082
0.152
0.048
0.090
0.043
0.036
0.028
0.022
0.076
0.058
0.068
0.055
0.045
0.032
0.048
0.026
Median
1.155*
1.234*
1.374*
1.444*
1.248*
1.178*
1.132*
1.226*
1.179*
1.334
1.370
1.352
1.241
1.247
1.169
1.247
1.198
1.154
'Recommended ratio for specified city.
-------
Table 8
Determination of Adjustment Coefficients for Eight-Hour NAAQS
Attainment (CLV8 = 80 ppb) in Philadelphia
District
1
2
3
4
5
6
7
8
9
10
Weibull fits
1-hk
1.69
2.21
1.96
1.81
2.28
2.23
1.93
2.14
1.74
2.26
1-hfi
46.9
56.4
51.0
493
56.6
51.2
44.3
51.2
38.1
54.5
to 1991 data
CLV1
167
149
153
162
145
134
135
140
131
141
CLV8
142
136
128
138
120
118
123
128
116
126
8-hr NAAQS attainment parameters
Adjusted
CLV8
80
77
72
78
68
66
69
72
65
71
Reassigned
CLV8
72
77
78
80
72
69
65
71
66
68
Equivalent
CLV1
82
87
88
91
82
78
74
80
75
77
1-h Weibull parameters
V
3.173
3.725
3339
3.057
4.119
4237
3.941
3.960
3.518
4.359
6'
41.45
49.01
46.44
44.89
48.41
47.08
42.70
46.75
40.60
47.06
Adjustment coefficients
c
5339
4.481
4.618
4.465
5.183
5.932
6.671
5.572
6.708
5.922
d
0.533
0.593
0.587
0.592
0.554
0.526
0.490
0.540
0.495
0.518
'Assumes maximum CLV8 equals 80 ppb.
-------
Mr. Tom McCurdy 35 April 15, 1993
1. 1-hour NAAQS: highest CLV1 in city equals 120 ppb.
2. 8-hour NAAQS: highest CLV8 in city equals 60 ppb.
3. 8-hour NAAQS: highest CLV8 in city equals 80 ppb.
4. 8-hour NAAQS: highest CLV8 in city equals 100 ppb.
With your approval, we will use these coefficients to simulate NAAQS attainment in the next
round of pNEM/O3 runs. Tables listing the coefficients will be provided in a future letter.
Draft copies of these tables can be obtained from Jim Capel.
I hope this letter is useful in explaining our proposed adjustment procedure for ozone data.
Please call me if you have any questions or comments.
Sincerely,
IT AIR QUALITY SERVICES
Ted Johnson
TRJ:jda
cc: Bill Biller
Harvey Richmond
Steve Kopp
-------
APPENDIX B
ADJUSTMENT OF OZONE DATA TO SIMULATE
5EXEX 8H NAAQS
B-1
-------
INTERNATIONAL
TECHNOLOGY
CORPORATION
June 15, 1993
IT Project No. 830013-26
Mr. Thomas McCurdy
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Mutual Building, Mail Drop -12
Research Triangle Park, North Carolina 27711
Adjustment of Ozone Data to Simulate Sexex 8h NAAQS
Dear Tom:
Under Work Assignment D-5 of EPA Contract No. 68-DO-0062, IT Air Quality Services
(ITAQS) will be applying a revised version of pNEM/O3 to nine urban areas. Fixed-site
monitoring data for the years 1990 and 1993 will be used to represent "as is" ambient
concentrations within each exposure district. An air quality adjustment procedure (AQAP)
will be used to adjust the "as is" data to simulate various 1-hour and 8-hour National Ambient
Air Quality Standards (NAAQS).
In my letter of April 15, 1993, I described AQAP's that could be used to simulate the
attainment of 1-hour and 8-hour daily maximum NAAQS which permitted one expected
exceedance of a specified ozone concentration. In this letter, I will present a recommendation
for an AQAP that can be used to simulate the attainment of 8-hour daily maximum NAAQS
that permit five expected exceedances. This letter also provides a recommendation for an
extra step to be added to the AQAP previously proposed for simulating attainment of 8-hour
daily maximum NAAQS that permit one expected exceedance.
The letter will be organized according to the same step-by-step procedure used in the previous
letter. The five steps included in the AQAP are listed below.
1. Specify an air quality indicator (AQI) to be used in evaluating the status of a
monitoring site with respect to the NAAQS of interest.
2. Determine the value of the AQI for each site within a study area under "as is"
conditions.
Regional Office
3710 University Drive • South Square • Corporate Center One • Suite 201
Durham. North Carolina 27707 • 919-493-3661
IT Corporation is a wholly owned subsidiary ot International Technology Corporation
-------
Mr. McCurdy 2 June 15, 1993
3. Determine the value of the AQI under conditions in which the air pollution
levels within the study area have been reduced until a single site just attains a
specified NAAQS.
4. Adjust the one-hour values of the "as is" data set associated with each site to
yield the AQI value determined in Step 3. The adjusted data set should retain
the temporal profile of the "as is" data set.
STEP 1: SPECIFY AIR QUALITY INDICATOR
Let "8HDM-5ExEX NAAQS" indicate an 8-hour daily'maximum NAAQS which permits 5
expected exceedances. To determine compliance with a 8HDM-5ExEx NAAQS, we must
determine an Air Quality Indicator (AQI) for the area of interest. There are at least two ways
that we can determine the AQI. Consistent with EPA guidelines, we can (1) determine the
sixth largest 8-hour daily maximum concentration associated with each monitoring site in the
area and (2) designate the largest of these values as the AQI for the area. An alternative
approach would be to (1) fit a Weibull or lognormal distribution to the daily maximum values
associated with each site, (2) determine the characteristic fifth largest 8-hour daily maximum
concentration associated with each fitted distribution, and (3) designate the largest of these
values as the AQI for the area. The latter approach is similar to the approach incorporated
into the AQAP's described in my letter of April 15, 1993. Note that the characteristic fifth
largest value is statistically equivalent to the value expected to be exceeded five times.
We recommend that the first approach be employed. The observed sixth largest daily
maximum 8-hour value should be a relatively robust statistic. We do not believe that there is
enough "noise" in the observed sixth largest values to justify the additional effort required to
fit distributions and calculate characteristic fifth highest values.
In the remainder of this letter, we will use the term EH6LDM to indicate the sixth largest
eight-hour daily maximum value.
STEP TWO: DETERMINE VALUES OF AQI UNDER "AS IS" CONDITIONS
My letter of February 18, 1993 lists the data sets we have selected to represent "as is"
conditions in each of the nine cities under analysis. We have determined the EH6LDM for
each of these 96 site-years based on the "filled-in" data set. These values are listed in
Table 1 of this letter, together with previously determined values for the characteristic largest
one-hour value (CLV1) and the characteristic largest eight-hour value (CLV8).
STEP THREE: ADJUST VALUES OF AQI TO SIMULATE ATTAINMENT
We have developed an adjustment procedure for implementing this step which depends on the
assumption of a strong relationship between CLV8 and EH6LDM.
-------
Mr. McCurdy 3 June 15» 1993
Relationship Between CLV8 and EH6LDM
Linear regression analyses of the nine-city data base described above (n = 96) produced the
following regression equations.
CLV8 = 0.38 ppb + (1.267)(EH6LDM) RSQ = 0.897 (1)
LNCLV8 = 0.107 + (1.028)(LNEH6LDM) RSQ = 0.909 (2)
In Equation 2, LNCLV8 = ln(CLV8) and LNEH6LDM = ln(EH6LDM). Note that both of the
R-squared (RSQ) values exceed 0.89, indicating a high degree of correlation between the
regressed parameters.
The intercept of each equation was found to be non-significant at the p = 0.05 level.
Consequently, the analyses were repeated with the regression lines forced through the origin.
The resulting regression equations are presented below.
CLV8 = (1.270)(EH6LDM) RSQ = 0.897 (3)
LNCLV8 = (1.052)(LNEH6LDM) RSQ = 0.908 (4)
Both RSQ values exceed 0.89. Note that Equation 3 assumes a constant ratio between CLV8
and EH6LDM, i.e., ratio = CLV8/EH6LDM = 1.270. Equation 4 can be expressed as
CLV8 = (EH6LDM)L052 (5)
For EH6LDM values between 40 ppb and 200 ppb, the power curve expressed by Equation 5
is closely fit (within 5 percent) by the linear relationship expressed by Equation 3.
In my letter of April 15, 1993, I presented a recommendation for an AQAP applicable to
CLV8. In this procedure, each CLV8 associated with a city under "as is" conditions is
adjusted to simulate attainment of a NAAQS through the use of the equation
ACLV8(i,j) = [CLV8(i,j)][(AMAXCLV80)]/[MAXCLV8(j)]. (6)
-------
Mr. McCurdy
June 15, 1993
TABLE 1
AIR QUALITY INDICATORS FOR NINE CITIES
City
Chicago
Denver
Houston
Year
1991
1990
1990
District
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
1
2
3
4
5
6
7
8
9
10
11
Ozone concentration, ppb
CLV1
109
124
123
134
120
111
119
104
122
127
122
131
91
116
114
103
98
117
109
224
182
241
224
227
180
208
207
231
235
232
CLV8
94
107
106
114
99
97
106
92
106
111
106
111
74
94
85
86
79
78
94
162
137
161
171
179
131
165
143
154
171
167
EH6LDM
78
86
77
90
' 78
79
89
78
82
83
85
91
67
84
73
74
56
65
75
116
110
110
107
124
86
104
99
101
116
107
-------
Mr. McCurdy
Table 1 (Continued)
June 15, 1993
City
Los Angeles
Miami
New York
Year
1991
1991
1991
District
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
2
3
4
5
6
1
2
3
4
5
6
7
8
9
10
11
12
Ozone concentration, ppb
CLV1
321
185
148
271
215
248
116
136
198
153
167
216
264
266
261
249
90
97
93
105
96 "•"
87
158
121
153
143
123
97
141
172
162
170
183
148
CLV8
207
133
99
166
162
172
85
104
121
101
100
134
209
184
215
204
74
74
72
82
80
72
135
112
133
134
113
75
108
145
131
143
140
137
EH6LDM
170
109
75
129
115
146
64
84
94
81
87
110
167
146
180
165
60
60
64
59
65
57
108
91
113
105
88
64
83
126
104
101
107
105
-------
Mr. McCurdy
Table 1 (Continued)
June 15, 1993
City
Philadelphia
St. Louis
Washington
Year
1991
1990
1991
District
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10
11
Ozone concentration, ppb
CLV1
167
149
153
162
145
134
135
140
131
141
124
141
131
103
122
78
124
100
114
103
119
134
135
130
143
141
143
135
169
141
134
145
CLV8
142
136
128
138
120
118
123
128
116
126
100
116
106
87
97
65
103
79
91
84
104
110
113
113
128
119
123
118
143
120
112
123
EH6LDM
116
113
111
115
107
101
102
104
90
102
73
88
87
68
81
59
87
67
80
64
86
80
88
95
106
98
100
104
102
91
85
100
-------
Mr. McCurdy 7 June 15, 1993
according to the following definitions.
ACLV8(i,j): the CLV8 of the i-th ranked site in City j under attainment of the
proposed NAAQS,
CLV8(i,j): the CLV8 of the i-th ranked site in City j for the baseline year,
MAXCLV8(j): the largest CLV8 of all sites in City j for the baseline year,
AMAXCLV8(j): the largest CLV8 permitted under the proposed NAAQS.
This expression is consistent with Model A, a model for CLV8 which we developed in the
April 15 letter. Model A assumes that
a) the CLVS's for a given city-year follow a lognonnal distribution,
b) the geometric standard deviation (GSD) of the lognormal distribution is
constant from year-to-year, and
c) the z value associated with the CLV8 of the r-th ranked site is constant from
year-to-year.
If a constant ratio exists between the values of EH6LDM and CLV8 for a city [i.e.,
CLV8/EH6LDM = constant), then the following expression should apply to EH6LDM.
AEH6LDM(ij) = [EH6LDM(ij)][(AMAXEH6LDM(j)/[MAXEH6LDMO')]. (7)
The following definitions apply.
AEH6LDM(i,j): the EH6LDM of the i-th ranked site in City j under attainment
of the proposed NAAQS,
EH6LDM(i,j): the EH6LDM of the i-th ranked site in City j for the baseline
year,
MAXEH6LDMQ: the largest EH6LDM of all sites in City j for the baseline year,
AMAXEH6LDM(j): the largest EH6LDM permitted under the proposed NAAQS.
-------
Mr. McCurdy 8 June 15> 1993
As a preliminary check of this approach, we applied it to the 1990 data for Houston. Table 2
lists the Houston EH6LDM values ranked from largest to smallest under the heading "1990
observed." The largest value [MAXEH6LDMO')] is 124 ppb. We used the equation above to
predict the EH6LDM values for 1991 under the assumption that the 1991 air quality for
Houston will just attain a NAAQS level of 108 ppb. The resulting predictions are listed in
Table 2 together with the observed values for 1991. In each case, the predicted value for
rank i is equal to 0.87 times the 1990 value for rank i. The multiplier (0.87) is equal to (108
ppb)/(124 ppb), where 108 ppb is the largest EH6LDM permitted in the adjusted data and 124
ppb is the largest EH6LDM in the baseline data. In this example, the predicted values agree
well (within 4 ppb or 5 percent) with the observed values.
We repeated the test using 1983 data for Philadelphia as the baseline data set and 1991
Philadelphia data as the data set we were attempting to predict. The results are presented in
Table 3. In this case, the predicted value is within 5 ppb of the corresponding observed value
for Ranks 1 through 9, the predicted value generally being larger than the observed value. In
the worst case (Rank 10), the predicted value differs from the observed value by 10 ppb (89
ppb vs. 79 ppb). Note that the observed value (79 ppb) appears to be an outlier, as it differs
from the next lowest observed value (89 ppb) by 10 ppb, whereas the remaining observed
values are closely spaced.
In a third test, we used 1983 data for St. Louis as the baseline data base and 1990 St. Louis
data as the data set we were attempting to predict Table 4 presents the results of this test.
In this case, each predicted value is within 5 ppb of the corresponding observed value for
Ranks 1 through 8, the predicted value generally being smaller than the observed value. In
contrast to these relatively good predictions, the differences for Ranks 9 and 10 are 17 ppb
and 15 ppb, respectively. Note that in this case, the two lowest 1983 values (68 ppb and 60
ppb) appear to be outliers, as they are widely separated from the remaining values.
-------
Mr. McCurdy
June 15, 1993
TABLE 2
PREDICTION OF 1991 EH6LDM VALUES
FOR HOUSTON GIVEN 1990 EH6LDM VALUES
Rank
1
2
3
4
5
6
7
8
9
10
11
Mean
EH6LDM, ppb
1990
observed
124
116
116
110
110
107
107
104
101
99
86
107.3
1991
Observed
108
104
100
99
97
91
89
89
85
84
74
92.7
Predicted*
108
101
101
96
96
93
93
91
88
86
75
93.5
Difference,
ppb
0
-3
1
-3
-1
2
4
2
3
2
1
0.8
Percent
difference
0
-2.9
1.0
-3.0
-1.0
2.2
4.5
2.2
3.5
2.4
1.4
0.9
*1991predicted = (1US/124) (1990 observed)
-------
Mr. McCurdy
10
June 15, 1993
TABLES
PREDICTION OF 1991 EH6LDM VALUES
FOR PHILADELPHIA GIVEN 1983 EH6LDM VALUES
Rank
1
2
3
4
5
6
7
8
9
10
Mean
EH6LDM, ppb
1983
observed
125
117
116
115
115
115
109
108
99
99
111.8
1991
Observed
112
108
102
101
101
98
97
95
89
79
98.2
Predicted"
112
105
104
103
103
103
98
97
89
89
100.2
Difference,
ppb
0
_3
2
2
2
5
1
2
0
10
2
Percent
difference
0
-2.8
2.0
2.0
2.0
5.1
1.0
2.1
0
12.7
2.0
'1991 predicted = (112/125)(I983 observed)
-------
Mr. McCurdy
11
June 15, 1993
TABLE 4
PREDICTION OF 1990 EH6LDM VALUES
FOR ST. LOUIS GIVEN 1983 EH6LDM VALUES
Rank
1
2
3
4
5
6
7
8
9
10
Mean
EH6LDM, ppb
1983
observed
119
119
117
110
109
101
96
86
68
60
98.5
1990
Observed
88
87
87
86
81
80
73
68
67
59
77.6
Predicted*
88
88
87
81
81
75
71
64
50
44
72.8
Difference,
ppb
0
1
0
-5
0
-5
-2
-4
-17
-15
-4.8
Percent
difference
0
1.1
0
-5.8
0
-6.3
-2.7
-5.9
-25.4
-25.4
-6.2
"1990 predicted = (88/I19X1983 observed)
-------
Mr. McCurdy 12 June 15, 1993
The results of these three tests suggest that the adjustment procedure is sensitive to outliers.
Because the procedure is keyed on the highest ranked value of the baseline data set, it should
perform well in predicting the higher ranked values in the attainment data set. Poor
predictions are more likely to occur in the lower ranks. The examples presented here support
this assumption, as the differences between predicted and observed attainment values are most
pronounced in the lower ranks.
Note that the large differences associated with the lower ranks were positive in Table 3 and
negative in Table 4. In the Table 3 analysis, the single outlier (79 ppb) was associated with
the attainment data. The two outliers in the Table 4 analysis (60 ppb and 68 ppb) were
associated with the baseline data set. Note that any single "universal fix" of the procedure
would tend to reduce the differences in one case while increasing the differences in the other.
Consequently, I recommend that we use Equation 7 without modification.
Equation 7 provides an estimate of the r-th ranked EH6LDM value. Following the approach
described in the April 15 letter, we will rank the monitoring sites associated with a particular
study area according to average rank calculated for each site from five recent years of ozone
data (including the year - 1990 or 1991 ~ selected for the pNEM/O3 analysis). The r-th
ranked site according to the five-year rank will be assigned the r-th ranked EH6LDM value.
Per your instructions, we will determine the five year rank through data provided by RADS.
Note that the ranking assigned to each site according to EH6LDM may differ from the
ranking assigned to the site according to CLV8.
STEP 4: ADJUST ONE-HOUR DATA
The EH6LDM adjustment procedure discussed above provides an estimate of the EH6LDM
associated with each site under attainment conditions. To complete the adjustment process,
the one-hour values associated with the site should be adjusted so that a Weibull distribution
fit to the resulting eight-hour daily maximum values will yield the designated EH6LDM.
We can simplify this process by adjusting the one-hour values so that a Weibull distribution
fit to the one-hour values will yield a CLV1 that is equivalent to the desired EH6LDM. This
short-cut requires an equivalence expression relating EH6LDM to CLV1.
Linear regression analyses of the nine-city data base described above (n = 96) produced the
following regression equations.
CLV1 = -14.3 ppb + (1.755)(EH6LDM) RSQ = 0.754 (8)
LNCLV1 = 0.110 + (1.077)(LNEH6LDM) RSQ = 0.776 (9)
-------
Mr. McCurdy 13 J"ne 15» 1993
In Equation 9, LNCLV1 = In(CLVl) and LNEH6LDM = ln(EH6LDM). Note that both of the
R-squared (RSQ) values exceed 0.75, indicating a relatively high degree of correlation
between the regressed parameters.
The intercept of each equation was found to be non-significant at the p = 0.05 level.
Consequently, the analyses were repeated with the regression lines forced through the origin.
The resulting regression equations are presented below.
CLV1 = (1.614)(EH6LDM) RSQ = 0.748 (10)
LNCLV1 = (1.101)(LNEH6LDM) RSQ = 0.775 (11)
Both RSQ values exceed 0.74. Note that Equation 10 assumes a constant ratio between
CLV1 and EH6LDM, i.e., ratio = CLV1/EH6LDM = 1.614. Equation 11 can be expressed as
CLV1 = (EH6LDM)1-101 (12)
For EH6LDM values between 80 ppb and 190 ppb, the power curve expressed by Equation
12 is closely fit (within 5 percent) by the linear relationship expressed by Equation 10.
We examined the 96 site-specific ratios of CLV1 to EH6LDM by city and found that the
ratios varied more among cities than among sites within a city. We also examined the CLV1-
to-EH6LDM ratios in the three-city multiyear data base (Houston, Philadelphia, and St. Louis)
described above and found that the average city ratio was either constant year to year or else
showed a slight decreasing trend. For example, the median ratio for Houston was relatively
constant; the ratio was 2.134 for 1985, 1.890 for 1988, 2.091 for 1990, and 1.929 for 1991.
The median ratio for Philadelphia displayed an overall decline; the ratio was 1.528 for 1983,
L622 for 1984, 1.473 for 1986, and 1.373 for 1992. We recommend that (1) a single average
CLVl-to-EH6LDM ratio be applied to all sites within a particular city and (2) the ratio be
determined by a recent year of data specific to the city.
Table 5 lists the mean and median ratios for each city based on the 1990-91 nine-city data
base. The table also lists mean and median ratios for three other years for three of the cities
(Houston, Philadelphia, and St. Louis). We recommend that the flagged median ratios in this
table be used for the city-specific ratios. Each flagged median is associated with a recent
year (1990 or 1991). The medians are preferred to the means, as they are less affected
outliers.
Given an estimate of ACLV1, the CLV1 of the adjusted data, we next predict the parameters
of the Weibull distribution that will fit the adjusted data. I presented a method for predicting
these parameters in my letter of April 15, 1993. For convenience, I have briefly summarized
this material below.
-------
Mr. McCurdy
14
June 15, 1993
TABLES
RATIOS OF CLV1 TO EH6LDM FOR NINE
CITIES BASED ON RECENT OZONE DATA
City
Chicago
Denver
Houston
Los Angeles
Miami
New York
Philadelphia
St. Louis
Washington
All Cities
Houston
(other years)
Philadelphia
(other years)
St Louis
(other years)
Data year
1991
1990
1990
1991
1991
1991
1991
1990
1991
See above
1985
1988
1991
1983
1984
1986
1983
1985
1991
Number of Sites
12
7
11
16
6
12
10
11
11
96
10
11
11
10
10
10
10
11
11
CLV1 to EH6LDM ratio
Mean
1.453
1.528
2.033
1.806
1.559
1.487
1.373
1.499
1.484
1.595
2.129
1.891
1.915
1.543
1.653
1.485
1.659
1.726
1.357
Std. dev.
0.080
0.182
0.177
0.197
0.122
0.144
0.049
0.109
0.124
0.246
0.293
0.176
0.125
0.079
0.214
0.072
0.084
0.125
0.035
Median
1.441a
1.453*
2.091*
1.846*
1.513'
1.436*
1.367*
1.506*
1.450*
1.512
2.134
1.890
1.929
1.528
1.622
1.473
1.640
1.704
1.344
"Recommended ratio tor specified city.
-------
Mr. McCurdy 15 June 15, 1993
Estimation of Weibull Parameters of Adjusted One-Hour Data
Let 6 and k indicate the parameters of the Weibull distribution fitting the unadjusted 1-hour
data associated with a particular site. These parameters are related to the UCLV1 -- the
CLV1 of the unadjusted data ~ by the expression.
UCLV1 = 6 [Info)]1* (13)
Let 6' and k' indicate the parameters of the Weibull distribution fitting the adjusted 1-hour
data. These parameters are related to the adjusted CLV1 (ACLV1) by the expression
ACLV1 = 6' [ln(ri)]w (14)
Based on the results of a series of regression analyses discussed in the April 15 letter, we are
recommending that the following regression equation be used to estimate k':
1/k' = -0.2389 + (0.003367) (ACLV1) + (0.4726) (1/k). (15)
Each coefficient of this equation was found to be statistically significant at the p = 0.10 level;
the over-all R squared value was 0.6993. Note that the equation assumes that ACLV1 will be
in units of ppb.
Given the ACLV1 and an estimate of k', we can estimate the corresponding 6' value by the
equation
6' = (ACLVl)/[ln(n)]w. (16)
We now have values for 6, k, UCLV1, 6', k', and ACLV1. If we treat the unadjusted data
set as a time series in which x, is the one-hour value at time t, we can construct a
corresponding adjusted data set by the expression
y, = (6') fr/a)* (17)
where yt is the adjusted value at time t.
Equation 17 can be restated as
y« = (c) W (18)
c = (5 W)** (19)
d = k/k'
-------
Mr. McCurdy 16 June 15> 1993
We will refer to c and d as "adjustment coefficients."
Table 6 demonstrates the application of the adjustment procedure to Philadelphia for an eight-
hour NAAQS which permits a maximum EH6LDM of 80 ppb. Under "as is" conditions the
highest EH6LDM is 116 ppb. We are interested in simulating attainment of NAAQS which
permits a EH6LDM no higher than 80 ppb. Each EH6LDM is first multiplied by 90 ppb/116
ppb as indicated by Equation 7. These adjusted EH6LDM's are then reassigned according to
the relative rank of each site for the five-year period. The reassigned EH6LDM are next
multiplied by 1.367, the median CLVl-to-EH6LDM ratio listed for Philadelphia in Table 5.
This step provides estimates of equivalent CLVl's which are then used to determine
adjustment coefficients through the use of Equations 15 through 19.
Final Adjustment of One-Hour Data
The adjustment procedure described above will produce a one-hour data set with a specified
CLV1. Note that the EH6LDM value of this data set may not exactly equal the "target"
value, as the assumed relationship between CLV1 and EH6LDM is only an approximation.
We recommend that a final "fine-tuning" adjustment be made to the one hour data so that we
obtain the exact EH6LDM specified. The following expression would be implemented.
Adjusted yt = (y,)(Observed EH6LDM)/(Target EH6LDM) (20)
In this equation, yt is the one-hour value for hour t after the Weibull adjustment procedure.
The "observed EH6LDM" is the EH6LDM value of the data set after the Weibull adjustment
procedure, and the "target EH6LDM" is the EH6LDM value assigned to the site by the
rollback procedure.
This procedure can also be used as a final step in the procedure proposed for CLV8 in my
letter of April 15, 1993. That is, the following equation would be used to adjust the one-hour
data after the Weibull adjustment step.
Adjusted y, = (y^Observed CLV8)/(Target CLV8) (21)
-------
TABLE 6
DETERMINATION OF ADJUSTMENT COEFFICIENTS FOR EIGHT-HOUR NAAQS
ATTAINMENT (EH6LDM = 80 ppb) IN PHILADELPHIA
District
1
2
3
4
5
6
7
8
9
10
Parameters of 1991 data
1-hk
1.69
2.21
1.96
1.81
2.28
2.23
1.93
2.14
1.74
2.26
* \02iUUlCO UlflAJUllIUl 1
1-hd
46.9
56.4
51.0
49.3
56.6
51.2
44.3
51.2
38.1
54.5
CLVl
167
149
153
162
145
134
135
140
131
141
ji lOLiUIVl vu U ctio CM.
EH6LDM
116
113
111
115
107
101
102
104
90
102
8-hour NAAQS attainment parameters
Adjusted
EH6LDM
80
78
77
79
74
70
70
72
62
70
Rank
5
2
1
3
4
6
8
9
10
7
Reassigned
EH6LDM
74
79
80
78
77
72
70
70
62
70
Equivalent
CLVl
101
108
109
107
105
98
96
96
85
96
1-hour Weibull
parameters
k'
2.626
2.954
2.708
2.615
3.106
3.300
3.038
3.277
3.136
3.408
6'
44.62
52.24
4937
47.10
52.64
51.16
47.38
49.88
42.89
51.15
Adjustment
coefficients
c
3.750
2556
2.869
3.171
2.721
3581
4.261
3.817
5.689
3.608
d
0.644
0.748
0.724
0.692
0.734
0.676
0.635
0.653
0555
0.663
ppb.
-------
Mr. McCurdy .. 18 June 15, 1993
With your approval, we will use these equations to make final adjustments to the data sets we
prepare for the pNEM/O3 analyses of 8HDM-lExEx NAAQS and 8HDM-8ExEx NAAQS.
This letter completes our documentation of the proposed adjustment procedures for the 1993
pNEM/O3 exposure assessments. Please call me if you have any questions or comments.
Sincerely,
IT AIR QUALITY SERVICES
Ted R. Johnson
TRJ/kdl
cc: S. Kopp
M. McCoy
J. Capel
-------
APPENDIX C
SAMPLE OUTPUT OF pNEM/O3 APPLIED TO
1990 CHILDREN AND ENTIRE POPULATION DATA
(HOUSTON, 1-HOUR, DAILY MAXIMUM 0.12 PPM STANDARD
[CURRENT NAAQS])
C-1
-------
Table 1.
Cumulative Numbers of People at Hourly 03 Exposures
during 03 Season by Equivalent Ventilation Rate
03 Level
Equalled or Equivalent Ventilation
Exceeded, ppm <15 15-24 25-29
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
Study Area =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
81227
207464
488563
489899
489899
489899
489899
489899
HOUSTON 1-HR
No. exposure districts =
First day of
Last day of
No . days in
03 season =
03 season =
03 season =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7870
185176
372450
481979
489899
489899
489899
489899
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7870
7870
22129
149948
476490
489899
489899
489899
Rate, l/min-m**2
30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
137131
298988
484709
489899
489899
RB Children
11
1
365
365
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
16348
34385
313330
444173
488509
489899
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
89097
339188
488563
489899
489899
489899
489899
489899
-------
Table 2.
Occurrences of People at Hourly Exposures
During 03 Season by Equivalent Ventilation Rate
03 Interval,
ppm
Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34
35 +
ANY
.401 +
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-. 320
.281-. 300
.261-- 280
.241-. 260
.221-- 240
.201-. 220
.181-- 200
.161-. 180
.141-- 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-- 060
.021-. 040
.001-. 020
0.000
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0,
141126.
348827.
2939799.
15896222.
83622135.
417911084.
2956488880 .
281454911.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
7870.
188311.
791203.
4049915.
22236534.
76198305.
361721235.
32139002.
0.
0.
0.
0.
0-
0.
0.
0.
0.
0.
0.
0.
0.
0.
7870.
0.
14259.
175729.
1269978.
3818606.
16871271.
1707503 .
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
146245.
359528.
1250975.
4792889-
711170.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
4932.
11416.
23351.
492558.
982068.
2355820.
383713 .
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
156866.
542070-
3756677.
20291462.
107980733.
500161038.
3342230095.
316396299.
4291515240.
Study Area = HOUSTON 1-HR RB Children
No. exposure districts = 11
First day of 03 season = 1
Last day of 03 season = 365
No. days in 03 season = 365
-------
Table 1A.
Cumulative Numbers of People at Ihr Daily Max. Exposure
During 03 Season by Equivalent Ventilation Rate
03 Level
Equalled or
Exceeded) ppm
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
Equivalent Ventilation
<15
0
0
0
0
0
0
0
0
0
0
0
0
0
0
81227
199594
488563
489899
489899
489899
489899
489899
Study Area * HOUSTON 1-HR
No. exposure districts =
First day of 03
Last day of 03
No. days in 03
season =
season =
season -
15-24
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7870
160220
335504
468162
489899
489899
489899
489899
25-29
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
13859
60356
384589
431963
431963
431963
""""
Rate, l/min-m**2
30-34
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
94040
152560
318717
337315
337315
35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
4932
22969
150984
352346
390761
390761
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
89097
339188
488563
489899
489899
489899
489899
489899
RB Children
11
1
365
365
-------
Table 2A.
Occurrences of People ai Ihr Daily Max. Exposure
During 03 Season by Equivalent Ventilation Rate
03 Interval,
ppm
Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34
35+
ANY
.401+
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-. 320
.281-. 300
.261-. 280
.241-. 260
.221-. 240
.201-. 220
.181-. 200
.161-- 180
.141-. 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-. 060
.021-- 040
.001-. 020
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
0
0
0
0
0
0
0
0
0
0
0
0
0
0
136028
255855
1785175
7084224
27442111
65976863
42400651
0
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7870
155485
442115
2166532
8055529
14303903
5606295
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
13859
54599
616321
829844
316908
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
94040
119907
372866
52230
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
0
18037
161903
294297
44756
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
143898
416272
2241149
9417432
36395771
81777773
48420840
0
178813135.
RB Children
11
1
365
365
-------
Table IB.
Cumulative Numbers of People at 1-Hr Daily Max.
During 03 Season by 1-Hr 03 and EVR.
Dose
03 Level
Equalled or Equivalent Ventilation
Exceeded, ppm <15 15-24 25-29
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
,081
.061
.041
.021
.001
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
0
0
0
0
0
0
0
0
0
0
0
0
0
0
67551
139685
402944
489899
489899
489899
489899
489899
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
145455
367666
481979
489899
489899
489899
489899
RB
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7870
7870
21729
129306
475182
489899
489899
489899
Children
11
1
365
365
Rate, l/min-m**2
30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
137131
294094
484709
489899
489899
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
16348
34385
313330
444173
474833
474833
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
75421
272384
462746
489899
489899
489899
489899
489899
-------
Table 2B.
Occurrences of People at 1-Hr Daily Max. Dose
During 03 Season by 1-Hr 03 and EVR.
03 Interval,
ppm
Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34
35+
ANY
.401 +
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-. 320
.281-. 300
.261-. 280
.241-. 260
.221-. 240
.201-. 220
.181-. 200
.161-. 180
.141-. 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-. 060
.021-. 040
.001-. 020
0.000
Study Area =
No. exposure
First day of
Last day of
No. days in
0
0
0
0
0
0
0
0
0
0
0
0
0
0
67551
166484
721625
3449658
13899509
38857485
32066341
0
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
148590
622525
2884269
13549635
33984139
26397213
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7870
0
13859
155487
1179709
2938215
3225854
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-JO
0
146245
340361
1015650
1004007
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
11416
23351
492558
897792
540805
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
75421
320006
1369425
6659010
29461772
77693281
63234220
0
178813135,
RB Children
11
1
365
365
-------
Table 3.
Number of People at Their Highest Ihr Daily Max. Exposure
During 03 Season by Ventilation Rate Categories
03 Interval,
ppm
.401+
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-. 320
.281-. 300
.261-. 280
.241-. 260
.221-. 240
.201-. 220
.181-. 200
.161-. 180
.141-. 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-. 060
.021-. 040
-001-. 020
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
81227
118367
288969
1336
0
0
0
0
0
0
0
0
0
0
0
• o
0
0
0
0
0
7870
152350
175284
132658
21737
0
0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
13859
46497
324233
47374
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
94040
58520
166157
18598
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
0
18037
128015
201362
38415
0
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
89097
250091
149375
1336
0
0
0
0
HOUSTON 1-HR RB Children
districts =
03 season =
03 season =
03 season =
11
1
365
365
-------
Table 4.
Cumulative Numbers of People at 8-Hr Daily Max. Exposure
During 03 Season by 8-Hr Equivalent Ventilation Rate
03 Level
Equalled or
_=_===_=____.
8hr
Exceeded^ ppm <15
.201+
.191
.181
.171
.161
.151
.141
.131
.121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0,000
Study Area =
No . exposure
First day of
Last day of
No . days in
0
0
0
0
0
0
0
0
0
0
0
0
91214
223911
355762
489899
489899
489899
489899
Equivalent
15-24
0
0
0
0
0
0
0
0
0
0
0
0
0
6129
15461
310180
489899
489899
489899
===== :
Ventilation
25-29
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Rate>
30-34
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
l/min-tn**2
35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
ANY
0
0
0
0
0
0
0
0
0
0
0
0
91214
226905
356162
489899
489899
489899
489899
HOUSTON 1-HR RB Children
districts =
03 season =
03 season =
03 season =
11
1
365
365
-------
Table 5.
Occurrences of People at 8-Hr Daily Max. Exposure
During 03 Season by 8-Hr Equivalent Ventilation Rate
03 Interval ,
ppm
.201 +
.191-. 200
.181-. 190
.171-. 180
.161-. 170
.151-. 160
.141 -.150
.131-. 140
.121-. 130
.111-. 120
.101-. 110
.091-. 100
.081-. 090
.071-. 080
.061-. 070
.041-. 060
.021-. 040
.001-. 020
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
8hr Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34 35+
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
91214.
264689.
582277.
9034532.
58672876.
105655918.
41536 .
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
0.
0.
0.
0.
0-
0.
0.
0.
0.
0.
0.
0.
0.
6129.
9332.
437675.
2305009.
1711943.
0.
RB Children
11
1
365
365
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.
ANY
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
91214.
270818.
591609.
9472207.
60977885.
107367866.
41536 .
178813135.
-------
Table 4A.
Cumulative Numbers of People at 8-Hr Daily Max. Dose
During 03 Season by 8-Hr 03 and 8-Hr EVR.
03 Level
Equalled or 8hr
Exceeded > ppm < IS
.201+
.191
.181
.171
.161
.151
.141
.131
.121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
Study Area =
No . exposure
First day of
Last day of
No. days in
0
0
0
0
0
0
0
0
0
0
0
0
85900
192060
352412
489899
489899
489899
489899
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
Equivalent Ventilation Rate? l/min-m**2
15-24 25-29 30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
13999
23585
335221
489899
489899
489899
RB Children
11
1
365
365
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
ANY
0
0
0
0
0
0
0
0
0
0
0
0
85900
195054
353562
489899
489899
489899
489899
-------
Table SA.
Occurrences of People at 8-Hr Daily Max. Dose
During 03 Season by 8-Hr 03 and 8-Hr EVR
03 Interval, 8hr Equivalent Ventilation Rate, l/min-m**2
ppm <15 15-24 25-29 30-34 35+
.201 +
.191-. 200
.181-. 190
.171-. 180
.161-. 170
.151-. 160
. 141- . 150
.131-- 140
.121-. 130
.111-. 120
.101-. 110
.091-- 100
.081-- 090
.071-. 080
.061-. 070
.041-. 060
.021-. 040
.001-. 020
0.000
==========
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
85900.
15Z824.
551753.
8597205.
55177394.
107362790.
41536 .
0.
0.
0-
0.
0-
0.
0.
0.
0.
0.
0.
0.
0.
13999.
10336.
721039-
3210520.
2887839.
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.
ANY
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
85900.
166823.
562089.
9318244.
58387914.
110250629.
41536 .
Study Area = HOUSTON 1-HR RB Children
No. exposure districts = 11
First day of 03 season = 1
Last day of 03 season = 365
No. days in 03 season = 365
178813135.
-------
Tabla 6.
Number of People at Their Highest 8-Hr Daily Max. Exposure
During 03 Season by 8-Hr Ventilation Rate Categories
03 Interval >
ppm
.201+
.191-- 200
.181-. 190
.171-. 180
.161-. 170
.151-. 160
.141-. ISO
.131-. 140
.121-. 130
.111-. 120
.101-. 110
.091-. 100
.081-. 090
.071-. 080
.061-. 070
.041-. 060
.021-. 040
.001-. 020
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
8hr
-------
Table 7.
Cumulative Numbers of People at 8-Hr Daily Max.
Seasonal Mean (April to October) Exposure
03 Level
Equalled or
Exceeded, ppm
.071+ 0
.066 0
.061 0
.056 0
.051 0
.046 0
.041 0
.036 0
.031 3885
.026 102243
.021 427497
.011 489899
.001 489899
0.000 489899
Study Area = HOUSTON 1-HR RB Children
No. exposure districts = 11
First day of 03 season = 1
Last day of 03 season = 365
No. days in 03 season = 365
-------
Tabla 8.
Occurrences of People at 8-Hr Daily Max.
Seasonal Mean (April to October) Exposure
03 Interval,
ppm
.071+ 0
.066-.070 0
.061--065 0
.056--060 0
.051-.055 0
.046-.050 0
.041-.045 0
.036-.040 0
.031-.035 3885
.026-.030 98358
.021-.025 325254
.011-.020 62402
.001-.010 0
0.000 0
Study Area = HOUSTON 1-HR RB Children
No. exposure districts = 11
First day of 03 season = 1
Last day of 03 season = 365
No. days in 03 season = 365
-------
Table 9.
Number of People at Daily Max Dose thai Exceed
Specified 1-HR 03 Levels 1 OP More Times per Year
03 Level
Equalled or
Exceeded, ppm 1
.401+
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
Study Area =
No . exposure
First day of
Last day of
No. days in
0
0
0
0
0
0
0
0
0
0
0
0
0
0
75421
149341
81367
0
0
0
0
0
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
123043
34829
1336
0
0
0
0
Days /
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
136602
0
0
0
0
0
Year
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
84879
4531
0
0
0
0
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
58114
16843
0
0
0
0
>s
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
66955
467189
489899
489899
489899
489899
RB Children
11
1
365
365
-------
Table 10.
Number of People at Daily Max 8-HR Dose that Exceed
Specified 8-hr 03 Levels 1 or More Times per Year
03 Level
Equalled or
Exceeded > pp
.201+
.191
.181
.171
.161
.151
.141
.131
.121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
m 1
0
0
0
0
0
0
0
0
0
0
0
0
85900
158238
97912
0
0
0
0
HOUSTON 1-HR
districts =
03 season s
03 season =
03 season =
2
0
0
0
0
0
0
0
0
0
0
0
0
0
17984
142638
0
0
0
0
RB
Days / Year
3
0
0
0
0
0
0
0
0
0
0
0
0
0
16811
52948
0
0
0
0
Children
11
1
365
365
4
0
0
0
0
0
0
0
0
0
0
0
0
0
2021
36945
0
0
0
0
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
19984
0
0
0
0
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3135
489899
489899
489899
489899
-------
Table 11.
Number of People that Exceed Specified 03 Levels
at 1-HR Daily Max Dose 1 or More Times per Year
with Ventilation Rates of 30 or Higher
03 Level
Equalled or
Exceeded , ppm 1
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
4932
16348
118366
126980
1390
0
0
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
33789
144606
8683
0
0
Days /
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
106537
33718
0
0
• — — — — — — — — — — — '•
Year
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
45147
78188
28270
28270
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1394
37753
23154
23154
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
15917
330167
438475
438475
RB Children
11
1
365
365
-------
Table 12.
Number of People that Exceed Specified 8 HR 03 Levels
at Daily Max 8-HR Dose 1 or More Times per Year
with 8 Hour Ventilation Rates of 15 or Higher
03 Level
Equalled or
Exceeded > pp
.201+
.191
.181
.171
.161
.151
.141
.131
.121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
Study Area =
No . exposure
First day of
Last day of
No . days in
m 1
0
0
0
0
0
0
0
0
0
0
0
0
0
13999
22835
74588
762
0
0
HOUSTON 1-HR
districts =
03 season =
03 season =
03 season =
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
750
147727
0
0
0
RB
Days / Year
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
79427
0
0
0
Children
11
1
365
365
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
30344
35850
0
0
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3135
34670
0
0
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
418617
489899
489899
-------
Table 1.
Cumulative Numbers of People at Hourly 03 Exposures
during 03 Season by Equivalent Ventilation Rate
03 Level
Equalled or Equivalent Ventilation
Exceeded, ppm <15 15-24 25-29
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
147640
830692
2096625
2370270
2370512
2370512
2370512
2370512
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8057
198588
814058
1354763
2227942
2370512
2370512
2370512
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8656
9845
103710
344059
1160975
1821690
2199478
2202476
Rate, l/min-m**2
30-34 35 +
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
62209
233236
662850
1341432
1918676
1942763
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
• 7442
30668
127184
855980
1296710
1730735
1740540
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
156483
'965082
2213976
2370270
2370512
2370512
2370512
2370512
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 2.
Occurrences of People ai Hourly Exposures
During 03 Season by Equivalent Ventilation Rate
03 Interval,
ppm
Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34
35+
ANY
.401 +
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-, 320
.281-. 300
.261-. 280
.241-. 260
.221-. 240
.201-- 220
.181-. 200
.161-. 180
.141-. 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-. 060
.021-. 040
.001-. 020
0.000
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
214448.
1235055,
10919964.
60889038.
345958827.
2051044177.
16061980954.
1361261946.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
8057.
201654.
1379661.
6289740.
34320249-
122913126.
592148896.
48611571.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
8656.
1189.
94651 .
369544.
2103150.
7695255.
29684368.
2625486.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
o.
0.
0.
159.
62845.
180689.
832813.
2189772.
9204118.
1045110.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
7442.
23518.
108115.
1109426.
2274776.
6007192.
679483.
0
0
0
0
0
0
0
0
0
0
0
0
0
0
231161
1445499
12480639
67837126
384324465
2186117106
16699025528
1414223596
20765685120.
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 1A.
Cumulative Numbers of People at Ihr Daily Max. Exposure
During 03 Season by Equivalent Ventilation Rate
03 Level
Equalled or Equivalent Ventilation
Exceeded, ppm <15 15-24 25-29
.401 +
.281
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
147640
822733
2094975
2370270
2370512
2370512
2370512
2370512
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8057
168994
721822
1132916
1907127
2295531
2343928
2343928
0
0
0
0
0
0
0
0
0
0
0
0
0
0
786
1975
92928
236411
903374
1454012
1531842
1531842
Rate, l/min-m**2
30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
55384
178146
396825
827282
906780
906780
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7442
17954
92471
429994
951965
1037304
1037304
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
156483
965082
2213976
2370270
2370512
2370512
2370512
2370512
Study Area = HOUSTON 1-HR
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
RB Entire Population
11
1
365
365
-------
Table 2A.
Occurrences of People at Ihr Daily Max. Exposure
During 03 Season by Equivalent Ventilation Rate
03 Interval, Equivalent Ventilation Rate
ppm <15 15-24 25-29
.401 +
.381-. 400
.361-. 380
,341-. 360
.321-- 340
.301-. 320
.281-. 300
.261-. 280
.241-. 260
.221-. 240
.201-. 220
.181-. 200
.161-. 180
.141-. 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-. 060
.021-. 040
.001-- 020
0.000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
203299
966555
6703954
28663197
121473044
341851072
306846274
213119
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8057
164072
887159
3142770
13495039
23352288
10540713
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
786
1189
91739
162180
1115292
1941939
565873
0
, l/min-m**2
30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
55225
123120
387427
702887
179458
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7442
10512
74938
433269
726738
146095
0
ANY
0
0
0
0
0
0
0
0
0
0
0
0
0
0
212142
1139417
7748589
32166205
136904071
368574924
318278413
213119
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
865236880,
-------
Table IB.
Cumulative Numbers of People at 1-Hr Daily Max.
During 03 Season by 1-Hr 02 and EVR.
Dose
03 Level
Equalled or
Exceeded, ppm
Equivalent Ventilation Rate, l/min-m**2
.<15 15-24 25-29 30-34 35+
ANY
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
.081
.061
.041
.021
.001
0.000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
101680
586789
1862210
2365190
2370512
2370512
2370512
2370512
0
0
0
0
0
0
0
0
0
0
0
0
0
0
187
158474
681126
1311877
2092130
2366671
2370512
2370512
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8656
9845
102348
297085
1076701
1773190
2071281
2071281
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
62201
232498
641882
1279366
1661165
1661165
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7442
30663
126777
854632
1285825
1693059
1693059
0
0
0
0
0
0
0
0
0
0
0
0
0
0
110523
726881
1969564
2365190
2370512
2370512
2370512
2370512
Study Area = HOUSTON 1-HR
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
RB Entire Population
11
1
365
365
-------
Table 2B.
Occurrences of People at 1-Hr Daily Max. Dose
During 03 Season by 1-Hr 03 and EVR.
03 Interval>
ppm
Equivalent Ventilation Rats, l/min-m**2
<15 15-24 25-29 30-34
35 +
ANY
.401 +
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-. 320
.281-. 300
.261-. 280
.241-. 260
.221-. 240
.201-. 220
.181-. 200
.161-. 180
.141-. 160
.121-. 140
.101-. 120
.081-. 100
.061-. 080
.041-. 060
.021-. 040
.001-. 020
0.000
0
0
0
0
0
0
0
0
0
0
0
0
0
0
101776
660165
3652933
18623289
86085760
275771153
313180932
213119
0
0
0
0
0
0
0
0
0
0
0
0
0
0
187
161422
1075555
4452618
22291881
58553853
54755841
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8656
1189
93289
305018
1887247
5913687
6918633
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
62042
179769
756883
1777552
2457127
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7442
23226
102430
1048656
1950034
2163357
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
110619
830377
4907045
23663124
112070427
343966279
379475890
213119
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
865236880.
Entire Population
11
1
365
365
-------
Table 3.
Number of People at Their Highest Ihr Daily Max. Exposure
During 03 Season by Ventilation Rate Categories
03 Interval,
ppm
.401 +
.381-. 400
.361-. 380
.341-. 360
.321-. 340
.301-. 320
.281-. 300
.261-. 280
.241-. 260
.221-- 240
.201-. 220
.181-- 200
.161-. 180
.141-. 160
.121-. 140
.101-. 120
-081-.100
.061-. 080
.041-. 060
.021-. 040
.001-. 020
0.000
Equivalent Ventilation Rate, l/min-m**2
<15 15-24 25-29 30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
147640
675093
1272242
275295
242
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8057
160937
552828
411094
774211
388404
48397
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
786
1189
90953
143483
666963
550638
77830
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
55225
122762
218679
430457
79498
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7442
10512
74517
337523
521971
85339
0
ANY
0
0
0
0
' 0
0
0
0
0
0
0
0
0
0
156483
808599
1248894
156294
242
0
0
0
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 4.
Cumulative Numbers of People at 8-Hr Daily Max. Exposure
-During 03 Season by 8-Hr Equivalent Ventilation Rate
03 Level
Equalled or 8hr
Exceeded > ppm <15
.201 +
.191
.181
.171
.161
.151
.141
.131
-121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
0
0
0
0
0
0
0
0
0
0
0
1467
211171
600390
1524093
2365702
2370512
2370512
2370512
Equivalent Ventilation Rate, l/min-m**2
15-24 25-29 30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
6129
17517
461142
1061471
1161075 .
1161075
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8642
29025
35600
35600
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
382
382
382
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
ANY
0
0
0
0
0
0
0
0
0
0
0
1467
211171
603384
1525115
2365702
2370512
2370512
2370512
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 5.
Occurrences of People at 8-Hr Daily Max. Exposure
During 03 Season by 8-Hr Equivalent Ventilation Rate
03 Interval, 8hr
ppm <15
.201 +
.191-. 200
.181-. 190
.171-. 180
.161-. 170
.151-. 160
.141-. ISO
.131-. 140
.121-. 130
.111-. 120
.101-. 110
.091-. 100
.081-. 090
-071-. 080
.061-. 070
.041-. 060
-021-.040
.001-. 020
0.000
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1467.
209729.
605843 .
226S535.
32194775.
254465331.
568691373.
146507.
Equivalent Ventilation Rate, l/min-m**2
15-24 25-29 30-34
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
6129.
11388.
675509.
3501047.
2424252 .
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
8642.
22184.
6786.
0.
0.
0.
0.
0.
0.
0.
0,
0.
0.
0.
0.
0.
0.
0.
0.
0.
382.
0.
0.
35+
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1.
0.
0.
ANY
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1467.
209729.
611972.
2276923.
32878926.
257988945.
571122411.
146507.
865236880.
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 4A.
Cumulative Numbers of People at 8-Hr Daily Max. Dose
During 03 Season by 8-Hr 03 and 8-Hr EVR.
03 Level
Equalled or 8hr
Exceeded, ppm <15
.201 +
.191
.181
.171
.161
.151
.141
.131
.131
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
0
0
0
0
0
0
0
0
0
0
0
1416
199893
529910
1503914
2356499
2370512
2370512
2370512
Equivalent
15-24
0
0
0
0
0
0
0
0
0
0
0
0
0
13999
25394
560608
1099214
1222681
1222681
Ventilation Rate, l/min-m**2
25-29 30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
10796
38292
47680
47680
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
382
382
382
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
13
13
13
ANY
0
0
0
0
0
0
0
0
0
0
0
1416
199893
532904
1505653
2356499
2370512
2370512
2370512
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table SA.
Occurrences of People at 8-Hr Daily Max. Dose
During 03 Season by 8-Hr 03 and 8-Hr EVR
03 Interval, 8hr Equivalent
ppm <15 15-24
.201 +
.191-. 200
.181-. 190
.171-. 180
.161-. 170
.151-- 160
.141-. ISO
.131-. 140
.121-- 130
.111-. 120
.101-- 110
.091-. 100
.081-- 090
.071-. 080
.061-. 070
.041-. 060
.021-. 040
.001-. 020
0.000
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1416.
198S02.
461221.
2200165.
30672652.
244107723.
577018505.
161449.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
13999.
12145.
1045991.
4802454.
4485519.
786.
Ventilation Rate, l/min-m#*2
25-29 30-34 35+
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
10801.
30311.
12846.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
382.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
13.
0.
0.
ANY
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
1416.
198502.
475220.
2212310.
31729444.
248940883 .
581516870.
162235.
865236880,
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 6.
Number of People at Their Highest 8-Hr Daily Max. Exposure
During 03 Season by 8-Hr Ventilation Rate Categories
03 Interval, 8hr
ppm <15
.201 +
.191-. 200
.181-. 190
.171-. 180
.161-. 170
.151-. 160
.141 -.ISO
.131-. 140
.121-- 130
.111-. 120
.101-. 110
.091-. 100
.081-. 090
.071-. 080
.061-. 070
.041-. 060
.021-. 040
.001-. 020
0.000
0
0
0
0
0
0
0
0
0
0
0
1467
209704
389219
923703
841609
4810
0
0
Equivalent
15-24
0
0
0
0
0
0
0
0
0
0
0
0
0
6129
11388
443625
600329
99604
0
Ventilation Rate, l/min-m**2
25-29 30-34 35+
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
8642
20383
6575
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
382
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
ANY
0
0
0
0
0
. 0
0
0
0
0
0
1467
209704
392213
921731
840587
4810
0
0
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 7.
Cumulative Numbers of People at 8-Hr Daily Max.
Seasonal Mean (April to October) Exposure
03 Level
Equalled or
Exceeded} ppm
.071+ 0
.066 0
.061 0
.056 0
.051 0
.046 0
.041 0
.036 0
.031 6134
.026 219654
.021 1434125
.011 2370512
.001 2370512
0.000 2370512
Study Area = HOUSTON 1-HR RB Entire Population
No. exposure districts = 11
First day of 03 season = 1
Last day of 03 season = 365
No. days in 03 season = 365
-------
Table 8.
Occurrences of People at 8-Hr Daily Max.
Seasonal Mean (April to October) Exposure
03 Interval,
ppm
.071+ 0
.066--070 0
.061-.065 0
.056-.060 0
.051-.055 0
.046-.050 0
.041-.045 0
.036-.040 0
.031-.035 6134
.026-.030 213520
.021-.025 1214471
.011-.020 936387
.001-.010 0
0.000 0
Study Area = HOUSTON 1-HR RB Entire Population
No. exposure districts = 11
First day of 03 season = 1
Last day of 03 season = 365
No. days in 03 season - 365
-------
Table 9.
Number of People at Daily Max Dose that Exceed
Specified 1-HR 03 Levels 1 or More Times per Year
03 Level
Equalled or
Exceeded? ppr
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
.101
-081
.061
.041
.021
.001
0.000
n 1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
110427
527223
507817
9728
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
96
185360
424359
25082
0
0
0
0
Days / Y
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
14139
391327
20881
0
0
0
0
ear
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
159
307331
40632
0
0
0
0
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
144643
120541
0
0
0
0
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
194087
2148326
2370512
2370512
2370512
2370512
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
-------
Table 10.
Number of People at Daily Max 8-HR Dose that Exceed
Specified 8-hr 03 Levels 1 or More Times per Year
03 Level
Equalled or
Exceeded > ppm
.201 +
.191
.181
.171
.161
.151
.141
.131
.121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
1
0
0
0
0
0
0
0
0
0
0
0
1416
199868
438842
698881
6149
0
0
0
2
Q
0
0
0
0
0
0
0
0
0
0
0
25
48453
439091
7859
0
0
0
Days / Year
3
0
0
0
0
0
0
0
0
0
0
0
0
0
43046
232740
101795
0
0
0
4
0
0
0
0
0
0
0
0
0
0
0
0
0
2563
73090
79995
0
0
0
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
57884
26662
0
0
0
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3967
2134039
2370512
2370512
2370512
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season -
Entire Population
11
1
365
365
-------
Table 11.
Number of People that Exceed Specified 03 Levels
at 1-HR Daily Max Dose 1 or More Times per Year
with Ventilation Rates of 30 or Higher
03 Level
Equalled or
Exceeded , ppr
.401 +
.381
.361
.341
.321
.301
.281
.261
.241
.221
.201
.181
.161
.141
.121
• .101
.081
.061
.041
.021
.001
0.000
n 1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7601
91317
286699
609506
539354
373931
373931
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
776
43613
306847
285437
281467
281467
Days /
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
157
130165
196323
263861
263861
Year
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
168
54866
173985
154285
154285
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
49908
100998
93416
93416
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
16318
429263
868350
868350
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season -
Entire Population
11
1
365
365
-------
Table 12.
Number of People that Exceed Specified 8 HR 03 Levels
at Daily Max 8-HR Dose 1 or More Times per Year
with 8 Hour Ventilation Rates of IS or Higher
03 Level
Equalled or
Exceeded > ppm
.201 +
.191
.181
.171
.161
.151
.141
.131
.121
.111
.101
.091
.081
.071
.061
.041
.021
.001
0.000
1
0
0
0
0
0
0
0
0
0
0
0
0
0
13999
24644
238922
183062
101027
101027
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
750
187363
77747
86930
86930
Days / Year
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
106292
85307
79230
79230
4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
33142
154601
85014
85014
5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3150
88395
113740
113740
>5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
349
510102
756740
756740
Study Area = HOUSTON 1-HR RB
No. exposure districts =
First day of 03 season =
Last day of 03 season =
No. days in 03 season =
Entire Population
11
1
365
365
------- |