United Sl.iti.-s
Hnviioiiniciil.il Protection
Agency
Oll'ice of An Quality
Planning aiul Standaids
Reseaich Tiun-lc Paik. NC 277 1 1
EPA-454/R-99-021
July 1999
EPA DISPERSIOP^ MODELING OF TOXIC
POLLUTANTS IN URBAN AREAS
GUIDANCE, METHODOLOGY AND APPLICATIONS
-------
EPA-454/R-99-021
AIR DISPERSION MODELING OF TOXIC POLLUTANTS
IN URBAN AREAS
GUIDANCE, METHODOLOGY AND EXAMPLE APPLICATIONS
Emissions, Monitoring and Analysis Division (MD-]4)
Office of Air Quality Planning and Standards
U S. Environmental Protection Agency
Research Triangle Park, NC 27711
July 1999
U.S. Environmental Protection Agency
Region 5, library (PL-12J)
77 West Jackson Boulevard. 12th Floor
Chicago, II 60604-3590
-------
NOTICE
This report has been reviewed by the office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, and has been approved for publication. Any mention of trade
names or commercial products is not intended to constitute endorsement or recommendation.
-...-. '.-.•< J'^ro
ii
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY vii
1. GUIDANCE FOR URBAN AIR TOXICS ANALYSES 1
1.1 INTRODUCTION 1
1.2 MODELING METHODOLOGY 2
1.2.1 Model Features 2
1.2.2 Model Options 2
1.2.3 Averaging Period 3
1.2.4 Receptors 3
1.2.5 Terrain 4
1.2.6 Meteorological Data 4
1.2.7 Chemistry' 5
1.2.8 Background Concentrations 6
1.2.9 Model Evaluation 6
1.2.10 Study Limitations 7
1.3 SOURCE DEFINITIONS 8
1.3.1 Modeling Domain 8
1.3.2 Emission Inventor}' Definitions 8
1.3.3 Source Characterization for ISCST3 9
1.3.4 Spatial and Temporal Distribution and Characterization of Area and
Mobile Source Emissions 11
1.3.5 Default Source Parameters 12
1.3.6 Source Parameters for Deposition Calculations 14
1.3.7 Pollutants 14
1.3.8 Source Grouping 14
1.3.9 Quality Assurance 14
1.4 MODEL OUTPUT FOR ANALYSIS 15
2. CASE STUDY MODELING METHODOLOGY 16
2.1 INTRODUCTION 16
2.2 MODELING METHODOLOGY 17
2.2.1 Model Selection 17
2.2.2 Modeling Options 18
2.2.3 Receptor Locations 19
2.2.3.1 Receptor Sampling Strategy* 19
2.2.3.2 Treatment of Terrain Influences 19
2.2.4 Meteorological Data 20
2.2.4.1 Selection of Surface and Upper Air Stations 20
2.2.4.2 Meteorological Parameters for Deposition Calculations 20
2.2.4.3 Meteorological Preprocessing 21
in
-------
2.2 MODELING METHODOLOGY (continued)
2.2.5 Emission Characteristics 22
2.2.5.1 Determination of Background Concentrations 22
2.2.5.2 Default Source Parameter Values 22
2.2.5.3 Area and Mobile Source Modeling 23
2.2.5.4 Source Parameters for Dry Deposition Calculations 23
2.2.5.5 Source Grouping 25
2.2.6 Model Evaluation 26
2.3 OVERVIEW OF EMISSION INVENTORIES 27
2.3.1 Phoenix, Arizona Inventory 27
2.3.2 Houston, Texas Inventory 28
2.4 APPROACH USED TO ESTIMATE ATMOSPHERIC SECONDARY
FORMALDEHYDE PRODUCTION 31
2.4.1 Introduction 31
2.4.2 Simulation Specific Input 32
2.4.3 Analysis of Results for Use with ISCST3 33
2.4.4 Results 35
2.4.5 Conclusions 35
2.5 OVERVIEW OF MODELING RESULTS 40
2.5.1 Phoenix, Arizona Modeling Results 40
2.5.2 Houston, Texas Modeling Results 40
2.6 SUMMARY AND CONCLUSIONS 42
3. REFERENCES 43
APPENDIX A CASE STUDY FOR PHOENIX, ARIZONA A-l
APPENDIX B CASE STUDY FOR HOUSTON, TEXAS B-l
APPENDIX C PROPOSED METHODS FOR SELECTING RECEPTOR SAMPLES
FOR THE APPLICATION OF THE ISCST3 DISPERSION MODEL
TO URBAN AREAS C-l
APPENDIX D DEFAULT STACK PARAMETERS OBTAINED FROM OZONE
TRANSPORT ASSESSMENT GROUP FOR SUBSTITUTION OF
MISSING DATA D-l
APPENDIX E PARAMETERS RELATING TO THE FATES OF SELECT
ATMOSPHERIC POLLUTANTS E-l
APPENDIX F
ADDITIONAL INFORMATION ON HOW TO USE OZIPR F-l
IV
-------
LIST OF TABLES
Table 1.2-1 Sources of Meteorological Data 4
Table 2.3-1 Toxic Air Pollutant Emissions for Phoenix, Arizona, Based on Year 1993 30
Table 2.3-2 Toxic Air Pollutant Emissions for Houston, Texas, Based on Year 1993 30
Table 2.5-1 Highest Annual Average Concentrations from All Sources Combined for
Phoenix, Arizona, Based on 5 Modeled Years 1987 - 1991 41
Table 2.5-2 Highest Annual Average Concentrations from All Sources Combined for
Houston, Texas, Based on 5 Modeled Years 1987 - 1991 41
-------
LIST OF FIGURES
Figure 2.4-1: Formaldehyde concentrations for prototypical summer day in Houston Texas, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde), FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde) 36
Figure 2.4-2: Formaldehyde concentrations for prototypical autumn day in Houston, TX, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde), FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde) 37
Figure 2.4-3: Formaldehyde concentrations for prototypical winter day in Houston, TX, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde), FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde) 38
Figure 2.4-4: Formaldehyde concentrations for prototypical spring day in Houston, TX, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde). FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde) 39
VI
-------
EXECUTIVE SUMMARY
The 1990 Clean Air Act Amendments (CAAA) Section 112(k) requires EPA to reduce
urban air toxics. EPA is devising a broad strategy for reducing risks posed by air toxics from all
sources in urban areas, which is discussed in the Draft Integrated Urban Air Toxics Strategy
(U.S. EPA, 1999b). In order to fully understand the air toxics problem in urban areas, it is
necessary to know the concentrations of air toxics to which people are exposed; however, air
monitoring data are scarce and limited. Another means for understanding the air toxics problem
is to estimate hazardous air pollutant (HAP) air concentrations through the use of dispersion
models, relying on emissions measurements or estimates. Because urban areas can vary greatly
in terms of air toxics, sources, meteorology, and the legal enforcement options provided by State
and local programs to address air toxics, State and local measures will be needed to reduce urban
air toxics risks. This document was prepared to provide:
guidance for modeling urban area impacts of air toxics
• a demonstration of a methodology for modeling air toxics for use in city-specific analyses
• two example applications of city-specific air toxics modeling applications.
The first two sections of the document present generalized guidance and the overview of
two modeling applications. Section 1 presents guidance for conducting air toxics dispersion
modeling for an urban area. Section 2 presents highlights from two case studies of applying the
modeling guidance in urban areas: Phoenix, Arizona and Houston, Texas. In these examples the
impacts of five potential cancer causing air toxics were examined: benzene, 1,3-butadiene,
formaldehyde, polycyclic organic matter (POM), and hexavalent chromium (chromium VI).
Appendix A and B present more detail on the Phoenix, Arizona and Houston, Texas case studies,
respectively. These appendices also include detailed documentation of the emission inventories
used in the modeling analyses, as well as. detailed summaries of the results of the modeling
analyses. The approach for preparation of a mobile source toxic emission inventory used for
Houston is currently being updated from the approach used here. Therefore, persons conducting
subsequent analyses should contact OAQPS for the current emission inventor}' preparation
guidance.
The results of the modeling analyses show some significant differences between the two
cities that were studied. For Phoenix, the mobile sources were clearly the dominant source of
emissions for four of the five pollutants, the exception being hexavalent chromium, which
occurred mostly from cooling towers and other major sources. The mobile source emissions also
exhibited the strongest temporal variations, reflecting the diurnal patterns in road traffic, as well
as, some influences of meteorology on emission estimates. These patterns in the emission
inventory for Phoenix are also evident in the modeling results for that city. Since the majority of
emissions were from mobile sources, they were distributed over the entire domain, with some
spatial variability evident based on surrogate factors such as population. As a result, there was
little evidence in the modeling results of localized "hot spots" (sharp gradients in concentrations
over a relatively short distance) within the Phoenix domain.
While the Houston inventory showed emissions from mobile sources that were
comparable in magnitude to Phoenix, major source (also called point source) emissions from
Houston were significantly higher than major source emissions for Phoenix. Benzene emissions
from major sources were almost 100 times higher for Houston than for Phoenix, and
vn
-------
1,3-Butadiene emissions were about 700 times higher. As a result of this, the modeling analysis
for Houston does exhibit some significant "hot spots" associated with some of the larger sources
of emissions for certain pollutants.
These applications illustrate a methodology that may be applied to similar urban-wide
analyses of point and area sources of air toxics. The use of a plume model (a modified version of
the Industrial Source Complex Short Term 3 - ISCST3) has certain advantages over puff and grid
models in terms of less stringent input data requirements, and plume models require less
computational resources for long term exposure analyses. Further insight into the applicability
of such models on this scale of analysis may be gained in future studies by comparing modeled
concentrations to monitored concentrations, a task begun in the study for Houston.
vni
-------
1. GUIDANCE FOR URBAN AIR TOXICS ANALYSES
1.1 INTRODUCTION
The 1990 Clean Air Act Amendments (CAAA) Section 112(k) requires EPA to reduce
urban air toxics. EPA is devising a broad strategy for reducing risks posed by air toxics from all
sources in urban areas, which is discussed in the Draft Integrated Urban Air Toxics Strategy
(U.S. EPA, 1999b). In order to fully understand the air toxics problem in urban areas, it is
necessary to know the concentrations of air toxics to which people are exposed; however, air
monitoring data are scarce and limited. Another means for understanding the air toxics problem
is to estimate hazardous air pollutant (HAP) air concentrations through the use of dispersion
models, relying on emissions measurements or estimates. Because urban areas can vary greatly
in terms of air toxics, sources, meteorology, and the legal enforcement options provided by State
and local programs to address air toxics, State and local measures will be needed to reduce urban
air toxics risks. This document was prepared to provide:
• guidance for modeling urban area impacts of air toxics
• a demonstration of a methodology for modeling air toxics for use in city-specific analyses
two example applications of city-specific air toxics modeling applications.
Section 1 provides guidance and recommendations on specific issues for urban-wide
analyses of air toxics. Urban areas contain major sources and numerous smaller, area sources.
As a result modeling analyses for large numbers of air toxics sources posses special challenges.
Since most modelers are more familiar with modeling applications for a single facility, this
section should help provide guidance to transition from single facility applications to more
complex urban-wide applications.
Section 2 provides an overview of two applications of the Industrial Source Complex
Short Term 3 (ISCST3) model to urban-wide studies. ISCST3 was applied to the Phoenix,
Arizona and Houston, Texas urban areas. Section 2 contains information that is general to both
applications, while later sections present more detailed information and results of the analyses.
Appendix A covers the Phoenix, Arizona study, while Appendix B covers the Houston, Texas
study.
The guidance section begins with Section 1.2 covering the modeling methodology;
Section 1.3 discusses the emissions inventory; Section 1.4 discusses modeling output/analysis.
-------
1.2 MODELING METHODOLOGY
The recommended plume dispersion model for use in estimating urban-wide
concentrations of toxic air pollutants is the ISCST3 model. Justification for selecting ISCST3 is
provided in Section 2.2.1, where the needs for the case studies are discussed. This section
describes some of the most important details needed to apply ISCST3 for these types of
applications. Other sources of detailed guidance are listed below.
For general information on air quality modeling, consult Appendix W to CFR Part 51-
Guideline on Air Quality Models.
For information on how to use the ISCST3 model, consult the ISC3 model user's guide to
ISC Model (U.S. EPA, 1995c).
For information on how to preprocess the meteorological data for input in ISCST3,
consult the PCRAMMET user's guide (U.S. EPA, 1996b) and the MPRM user's guide
(U.S.EPA,1996a).
All of the items listed above can be obtained from EPA's SCRAM web site at
http: //www. ep a. go v/scram 001.
Information on the "Integrated Urban Air Toxics Strategy" developed under the authority
of Section 112(k) and 112(c) of the Clean Air Act is obtained from EPA's web site at
http: //www. epa.gov/ttnuatw 1.
1.2.1 Model Features
Key features of the ISCST3 dispersion model that are useful for urban-wide air toxics
applications include:
handles multiple point, area, and mobile sources
incorporates building downwash effects
• includes an urban dispersion option
contains considerable flexibility for specifying receptor locations and for grouping of
source impacts
• includes algorithms to treat the effects of elevated and/or complex terrain
• treats the effects of deposition of gaseous and particulate emissions
• includes an option to vary emissions by season and hour-of-day
• includes an option to treat atmospheric transformations by exponential decay.
1.2.2 Model Options
The regulatory default model option in ISCST3 should be selected for urban-wide
applications. More information about the default option parameters can be found in the ISCST3
user's guide (U.S. EPA, 1995c).
The option to vary the emissions by season and hour-of-day should be selected, unless the
objectives of the application or the form of the emissions inventory data dictate otherwise.
-------
For best estimates, the use of the wet and dry deposition and plume depletion options
should also be selected. Note that the selection of the plume depletion option will significantly
increase model run time. To utilize the wet and dry deposition option, the model requires
additional data for the meteorological and chemical parameters. For meteorological input, the
user should consult the meteorological data preprocessors' user's guides (see above). Deposition
estimates are very useful in multi-pathway exposure assessments. Information on chemical
parameters may be obtained from the technical literature.
The urban modeling option should normally be selected. However, this guidance
assumes that the sources are in an urban area and thus the urban option for ISC3 should be
selected. To determine if the modeling domain is urban, apply the criteria in Section 8.2.8 of the
Guideline on Air Quality Models. If the result of this analysis shows that the area to be modeled
is rural, this guidance should not be used, and any toxics modeling should be done in close
consultation with the EPA Regional Office.
The exponential decay option should be selected when half-life values are available for
the air toxics under consideration.
1.2.3 Averaging Period
The ISCST3 model computes an hourly concentration for each receptor. Other averaging
periods, e.g., 3-hour, daily, seasonal and annual can also be aggregated (U.S. EPA, 1995c). The
averaging period selected is based on the intended use. For chronic (long-term) exposure studies,
annual average air concentrations are generally needed. Some exposure studies require seasonal
average air concentration estimates. Shorter term ambient concentrations are usually needed for
determining acute exposure.
1.2.4 Receptors
A receptor is any location where ambient concentration estimates are calculated.
Receptors are usually placed in "ambient air"' which is outside of inaccessible plant property.
The ISCST3 model requires the coordinates of the specified receptors. Receptor locations should
be selected based on a case-by-case determination with expert judgement on the needs of the
study. Often, receptors are selected at coordinates provided in the census data (census block,
census block groups or census tracts). The procedure in Appendix C, Section C.7, Step-by-Step
Guidelines for Using Sampling Method C, is recommended. Analysts should supplement Step 3
of those guidelines by selecting "automatic"(i.e., arbitrary) receptor points (termed "certainty
points" in Appendix C) where, for example, there are likely to be very high emission gradients or
other important source/receptor relationships that otherwise might lead to extreme concentration
outliers affecting the variance of the sample mean. Method C is appropriate when there is no
requirement to: (1) specifically estimate the maximum concentration, (2) estimate concentrations
in ambient air exclusively, (3) estimating concentrations at locations over water, or (4) estimating
concentrations on building roofs or within wakes or cavities.
In the two example studies below, receptor selections were based on the input needs of
the Hazardous Air Pollutant Exposure Model (HAPEM4) (U.S. EPA, 1999a). The receptor
points were defined as the centroids of census block groups (see Appendix C). Census data and
-------
urban land use information can be used to identify locations (potential receptors) where
individuals live, work, attend school, and spend time in recreation.
For many studies, the number of receptors selected will be very large, and since model
run time is proportional to the number of receptors, unreasonable model run times can occur. For
instance, the decision to place a receptor at each census block centroid will result in many more
receptors than if they were placed at the centroid of the census block group. Following the
guidelines in Appendix C should minimize such problems.
1.2.5 Terrain
Terrain elevation at each source and receptor is required input for ISCST3. Digitized
terrain elevation data are available from U.S. Geological Survey (USGS) maps. Source (stack)
elevation is usually provided in the inventory. For many urban areas, terrain can be assumed to
be flat and source and receptor elevations set to zero. Where the urban area is in mountainous
terrain, terrain effects are important. First, the impact of individual plumes on elevated terrain
results in higher air concentration (through placing the receptor at the correct higher air
concentration, vertical location within the plume and estimating the impaction of the plume upon
intervening terrain). Second, wind channeling due to terrain can cause higher air concentrations.
The ISCST3 model does not address wind channeling effects other than if these effects are
captured by the available meteorological data. If the urban area contains complex terrain features
that are expected to significantly affect the modeled concentrations, a dispersion model that
handles such situations should be selected from those listed in the Guideline on Air Quality
Models (40CFR51).
1.2.6 Meteorological Data
The 1SCST3 model requires meteorological data consisting of hourly surface and upper
air observations. Surface and upper air meteorological data files may be obtained from the
National Climatic Data Center (NCDC) through their web site at http://www.ncdc.noaa.gov.
Alternatively, several CD's are available which contain national and international meteorological
data. These CD's are SAMSON, HUSWO, INSWO, and the Radiosonde Data of North
America, Table 1.2-1 presents the years covered by each CD.
Table 1.2-1 Sources of Meteorological Data
Source - CD
SAMSON
HUSWO
INSWO
Radiosonde Data of N.A.
Type
Surface
Surface
Surface
Upper Air
Years Covered
1961-1990
1990-1995
1982-1997
1946-1996
Reference
NCDC, 1993
NCDC. 1997a
NCDC, 1998
NCDC. 1997b
If the INSWO CD is used and wet deposition estimates are required, an additional file is
necessary. The TD-3240 precipitation data file can be obtained from NCDC through their web
-------
site. A cross reference is also available from NCDC that is useful in matching TD-3240 station
identifications to National Weather Service (NWS) station numbers.
In selecting surface and upper-air stations, consult with your State/Regional
meteorologist for the most applicable stations for your area. The closest stations may not be the
most representative due to the influences of terrain or water bodies.
Meteorological data must be preprocessed before use in ISCST3. Mixing heights can be
computed using surface (SAMSON or HUSWO) and upper air data (Radiosonde Data of North
America) via the mixing height program provided on the SCRAM web site. PCRAMMET and
MPRM preprocessors use surface and mixing height data as input to create ISCST3 input
meteorological files. PCRAMMET was developed for use with NWS data, while MPRM is used
primarily for processing on-site meteorological data. In urban areas, on-site meteorological data
are not often available.
The meteorological data preprocessor MPRM should be used to prepare the input files
necessary for applying the gas dry deposition algorithm in ISCST3. Values for additional
parameters needed in applying the gas dry deposition algorithms for the two case study cities are
presented in Sections 2.2.4.2 and 2.2.5.4. MPRM can also be used for setting up a
meteorological data file for ISCST3 to use in estimating both particle dry deposition, and gas and
particle wet deposition. PCRAMMET does not contain the algorithms for setting up a file to
support gas dry deposition, although it does prepare a meteorological data file for use in
estimating particle dry and wet deposition and gas wet deposition. Finally, PCRAMMET can
accept data directly from the SAMSON and HUSWO CD's, while MPRM has not been updated
to read the HUSWO data. Neither preprocessor has yet been updated to read in the INSWO data.
The MPRM and PCRAMMET meteorological data preprocessors can occasionally
produce very low mixing heights (less than 10 meters) based on the twice-daily values from the
mixing height data file and the interpolation scheme used to provide hourly values of mixing
height. Anomalously low mixing heights in the NCDC data files may be associated with a mid-
day cold frontal passage. While the occurrence of very low mixing heights is more likely for the
rural mixing heights than for the urban mixing heights, due to differences in the interpolation
routines, low mixing heights may occur for both rural and urban conditions. The application of a
very low mixing height with a near-surface level area source can produce anomalously large air
concentrations due to the treatment of limited mixing effects in the ISCST3 model; expert
judgement is needed to determine the minimum mixing height for a given urban area. For the
two example studies, a minimum value of 100 meters was applied to the hourly mixing heights
produced by MPRM to avoid this anomaly from influencing the results. For urban areas,
building heights will limit the lower mixing heights and the 100-meter value was considered the
upper limit to the minimum value for the depth of the well-mixed boundary layer in a large urban
area(Sutton, 1953).
1.2.7 Chemistry
The ISCST3 model provides concentration estimates due to primary emissions and has a
limited capability to consider atmospheric transformations by exponential decay (half-life). Some
pollutants (e.g., formaldehyde, acetaldehyde, and acrolein) are also formed in the atmosphere due
-------
to reactions among other pollutants (i.e., formed by secondary reactions). Thus, in addition to
estimating concentrations due to primary emissions, an estimate of concentrations based on
secondary reactions is usually needed and should be added to the ISCST3 output. EPA's OZIPR
screening model (Gery and Crous, 1991) may be used to estimate the secondary transformation
of pollutants. Appendix F contains additional detailed guidance on the use of OZIPR, and
Section 2.4 contains details about the application of OZIPR to estimate secondary formaldehyde
formation in Houston, TX.
1.2.8 Background Concentrations
Background air quality includes pollutant concentrations due to natural sources, nearby
sources other than those under consideration, and unidentified sources. For typical exposure
assessments, background concentrations should be added to the modeled concentrations to
provide total ambient air concentrations for estimating exposure. Air quality data from a HAP
monitoring network in the vicinity of the analysis area are often used to establish background
concentrations. Also, background concentrations of some air toxics may be found in the
literature.
In the absence of measured or other reported values, the following approach for
estimating background concentrations should be used. An expanded point source inventory
should be obtained for an area surrounding each city from the National Toxic Inventory' (NTI).
The domain for this expanded point source inventory should extend approximately 50 km
beyond the domain of the inventory being explicitly modeled in the analysis. The NTI point
sources should be modeled to estimate background concentrations within the modeling domain
as a function of wind direction. The modeled background concentration should be based on an
average concentration computed from a coarse grid, about every 5 kilometers, across the
modeling domain. These direction-specific background concentrations should then be added to
the modeled concentrations by a post-processor that also reads the meteorological data to obtain
the appropriate wind direction.
1.2.9 Model Evaluation
Monitoring data can be used to check the validity of the modeled concentration estimates
or determine background concentrations. Ideally, the monitoring and modeling data should span
the same time period. Air toxics monitoring data are available from EPA's Aerometric
Information Retrieval System (AIRS) web site at http://www.epa.gov/airs. In most instances,
ambient data are collected at a frequency of one in six days. A variety of statistical tests can be
used to compare modeled with observed estimates. How the model estimates compare to annual
average monitored data is useful for determining the suitability for inputting the estimates into
HAPEM4 (U.S. EPA, 1999a). For comparisons in urban areas, there are many uncertainties in
all facets of the comparison effort. Thus a factor of two agreement between modeled and
observed values is considered to be very good. Appendix B contains a discussion of the
comparison of modeled and monitored air concentrations for Houston, TX.
-------
1.2.10 Study Limitations
As part of the conclusions in an urban-wide air toxics modeling study report, the
limitations of the modeling effort should be clearly stated. The important limitations of the
ISCST3 model are provided in the User's Guide (U.S. EPA, 1995c). Limitations due to data
availability and other factors should also be described.
-------
1.3 SOURCE DEFINITIONS
1.3.1 Modeling Domain
The urban area domain should be selected based on case-by-case determination with
expert judgement. The urban area domain can include a city center or multiple counties. It
should be carefully defined because the larger the modeling domain, the greater the number of
sources and receptors to be considered and thus the greater the required computational resources.
Guidance in the draft Integrated Urban Air Toxics Strategy (U.S. EPA, 1999a) should be
consulted.
1.3.2 Emission Inventory Definitions
In many applications the initial step of the urban-wide air modeling process will be the
assembly of the emissions inventory. Depending upon the end use of the analysis, this inventory'
may include:
• point sources - releases that can be attributed to individual stacks or release points,
• area sources - releases that cannot be attributed to individual stacks or release points, and
• mobile sources - releases attributed to engine emissions, both on and off-road.
Information about the most recent National Toxics Inventory (NTI) and documentation are
available from EPA's web site at ftp://www.epa.gov/pub/EmisInventory/nti_96.
Major/Area Source Emissions
According to Title I, Section 112(a) of the CAA, a "major source" is any stationary
source (including all emission points and units located within a contiguous area and under
common control) of air pollution that has the potential to emit, considering controls, 10 tons or
more per year of any HAP or 25 tons or more per year of any combination of HAPs. An "area
source" is any stationary source of HAPs that does not qualify as a major source. Area sources
are also defined as emission sources that are too small and numerous to inventory individually.
Area and mobile source emissions are not attributed to a specific location. Instead, they are
calculated as county-wide aggregated emissions. For example, all dry cleaners' emissions are
summed to a single number that aggregates the emissions of all facilities in a county, instead of
by individual dry cleaner facilities.
Major (Point) Source Emissions
For the purposes of the inventory used for air toxics analysis, major sources should be
considered as point sources, meaning sources for which a location is known. This clarification is
meant to distinguish point sources from area or mobile sources.
Mobile Source Emissions
Typically, mobile source emissions are split into on-road and off-road components.
On-road mobile sources are those vehicles certified for highway use and to applicable emission
-------
standards. They include cars, trucks (light-duty, such as pick-ups, sports utility vehicles,
minivans; and heavy-duty, typically in the form of a semi tractor-trailer rig), buses, and
motorcycles. They may be fueled using gasoline, highway diesel fuel, or alternative fuel (e.g.,
CNG, LPG, electricity). They do not include off-road equipment that is occasionally on
highways in order to move to the work location, such as most types of construction equipment
and agricultural equipment.
Off-road is a term that covers a diverse collection of engines, equipment, and vehicles
within the mobile source realm. Also referred to as "non-road" or "off-highway," the off-road
category includes recreational equipment, airport service equipment, industrial/commercial
equipment, agricultural equipment, construction equipment, oil and mining equipment, lawn and
garden equipment, logging equipment, and recreational and commercial marine vessels. Though
dealt with separately in the Clean Air Act, locomotives and aircraft can also be considered
categories of off-road engines.
1.3.3 Source Characterization for ISCST3
Generating the source inventory for modeling is intertwined with the creation of the
pollutant inventory'. Each emissions source and the constituents each source emits must be
specifically identified. For ISCST3 dispersion model, each source will need to be classified as a
point, area, volume, or line source. Building the source inventory usually begins with mapping
the locations of emission sources, receptors and the study domain.
The ISCST3 model can accommodate a large number of sources and receptors, however,
an optimum configuration is needed in order to minimize computer resources. Because source
inputs van,' with the type of source modeled, an important first step in creating the inventory' is to
identify each source of emissions as a point, area, volume, or line source. With the source types
established, the appropriate model inputs can be determined. The following subsections describe
the various source types and associated inputs for modeling.
Point Source Characterization
Point sources involve the release of emissions from a well-defined stack or vent, at a
known temperature and flow rate. Consequently, characterizing point sources for modeling is
fairly straightforward. The basic model inputs for any point source are: stack height above
ground level; inside diameter at stack exit; gas velocity or flow rate at stack exit; gas temperature
at stack exit; building dimensions, and emission rate. The location of the source will also need to
be defined in terms of the model receptor grid used.
Area Source Characterization
Area sources are sources of air toxic pollutants that are emitted at or near ground level
(e.g.,landfills, waste lagoons, evaporation and settling ponds, etc.). The sizes of these sources
can range from a few square meters in the case of settling ponds, to a few square kilometers or
larger in the case of landfills. Emissions from area sources are assumed to be of neutral
buoyancy. Therefore, plume phenomena such as downwash and impaction on elevated terrain
features are not considered relevant for modeling area sources. The emission rate for area
-------
sources is in units of mass per unit time per unit area [e.g., g/(s-meters squared[m2])]. It is an
emission flux rather than an emission rate. As an example, assume the pollutant emission rate
from a small lagoon is 150 g/s. The dimensions of the lagoon are 10 m by 20 m (total area is 200
m2). If this source were modeled as a single, square area source, then the modeled emission flux
would be 0.75 g/s-m2 (150 g/s - 200 m2).
For dispersion modeling, the important parameters used to characterize area sources are
location, geometry, and relative height. If the area source is not at ground level, a height for the
source may be entered (for example, a non-zero value would typically be entered for the height
of a land fill). If the release height of the source is greater than approximately 10 m, it should
probably be modeled as a volume source.
The 1996 emissions inventory provides a county-wide emissions rate for area sources.
Thus, the actual location of area sources is not available. In this situation, county-wide
emissions must be distributed to locations in the county (see Section 1.3.4).
Volume Source Characterization
There are two basic types of volume sources: surface-based or ground-level sources that
may also be modeled as area sources, and elevated sources. The effective emission height of a
surface-based volume source, such as a surface rail line, is usually set equal to zero. An example
of an elevated volume source is an elevated conveyor with an effective emission height set equal
to the height of the conveyor. A source may be defined as a volume source for modeling when
its emissions can be considered to occur over a certain area and within a certain depth of space.
At refineries, fugitive exhaust from on-site structures such as tanks, or a treatment facility may
be modeled as a volume source. A roadway over which contaminated soil is hauled may also be
modeled as a series of volume sources. As with area sources, emissions from volume sources are
assumed to be of neutral buoyancy.
The important parameters used to characterize volume sources for dispersion modeling
are location and initial lateral and vertical dimensions. The particular model user's guide will
have instructions on defining the initial lateral and vertical dimensions of the source. The length
of the side of the volume source will need to be known, as will the vertical height of the source,
and whether it is on or adjacent to a structure or building. Generally, the north-south and east-
west dimensions of each volume source must be the same. For refined modeling, the location is
simply expressed by a single east-west (X) and north-south (Y) coordinate.
Available emissions inventories, e.g., the 1996 NTI, do not contain sufficient information
to use the volume source feature in the ISCST3 model.
Line Source Characterization
Line sources are typically used to represent roadways. Basic model inputs are the overall
source length, width, and height. Emissions may be entered in units of grams per meter per
second.
10
-------
Line sources may also simply be modeled as a series of area or volume sources. In the
case of a long and narrow line source, it may be impractical to divide the source into N volume
sources, where N is given by the length of the line source divided by its width. Dividing the
length of the line source by its width effectively splits the line source into a string of squares (for
example, if the length of the line source was 100 m, and the width was 5 m, then the line source
could be split into twenty, adjacent square volume sources). An approximate representation of
the line source can be obtained by placing a smaller number of volume sources at equal intervals
along the line source (for example, for the line source of length 100 m and width 5 m, a total of
10 square volume sources separated from one another by 5 m could be defined). With this
option, the spacing between individual volume sources should not be greater than twice the width
of the line source. A larger spacing can be used, however, if the ratio of the minimum source-
receptor distance and the spacing between individual volume sources is greater than about 3.
Typically mobile sources are modeled as line sources. Using the county-wide mobile
sources emissions from the 1996 NT1, mobile source emissions can be modeled as area sources
by distributing them within the county (see Section 1.3.4). A better, but more resource
intensive, approach is to allocate on-road mobile sources to actual roadways using Geographical
Information System (GIS) software. These emissions can be modeled as line sources. Emissions
from off-road mobile sources can be modeled as area sources.
1.3.4 Spatial and Temporal Distribution and Characterization of Area and Mobile Source
Emissions
Area and mobile source emissions are usually provided on a county-wide basis. These
emissions must be allocated correctly to smaller areas contained in the modeling domain for use
by ISCST3. Analysts should divide the urban area into two-by-two kilometer grids and
apportion county wide emissions to these grids. Choose the allocation carefully since it will
affect the accuracy/reliability of the concentration estimates. For example, assign emissions
from large area sources such as landfills and airports to their actual locations in these grids rather
than averaging them over the entire domain. Note that resources often preclude assigning
emissions from sources with numerous locations, such as dry cleaners, gasoline stations, etc., to
specific locations. For these sources, apportion county level emissions to the 2 kilometer grid
cells within each county using surrogate distribution data, such as residential population, land
use, or any other parameter whose distribution is known for the 2 kilometer grid cells. County
wide emissions should be allocated to the grids based on the proportion of each cell's surrogate
value of the total county surrogate. Tools such as GIS can be utilized to assign the county wide
emissions from these sources to each grid cell.
For on-road mobile sources, GIS techniques can be used to estimate the relative length of
major highways (roadway miles) in each county and grid cell. The ratio of roadway length in the
grid cell to the roadway length in the county can be applied to the emissions rate. On-road
mobile emissions are then assigned to grid cells in proportional amounts. Tables and maps
should be produced to ensure that the allocation is performed properly. In gridding emissions in
this manner, a pattern of highly variable emissions density in the modeling domain is established.
The more comprehensive this emission allocation effort becomes, the sharper the gradient in the
modeled concentrations.
11
-------
Due to the time consuming nature of allocating emissions, EPA is developing a pre-
processor Emissions Modeling System for Hazardous Air Pollutants (EMS HAP), to assist in this
effort. Model code and a user's guide will be made available on the SCRAM website. EMS
HAP will provide recommended spatial and temporal surrogates for area and mobile source
emissions.
The area source algorithm in the ISCST3 model treats emissions as being uniformly
distributed over each area source grid cell and allows for concentration estimates to be made
within the area source itself. The area and mobile sources should be modeled using a nominal
release height above ground, and an initial vertical dispersion value (sigma-z) to account for the
fact that the area and mobile source emissions have some initial release height and initial depth,
rather than being emitted passively from the ground surface. Expert judgement on the value used
for this initial depth is needed to reflect local conditions (obstructions, proximity of receptors to
roadway, etc.). For the example studies, the nominal release height was 2 meters and the initial
vertical dispersion value (sigma-z) was 1 meter.
The ISCST3 model has the capability to address the temporal variations in emissions if
the sources do not operate at the same rate for every hour of the year. For point sources,
information on operating schedule or data from continuous emissions monitors can be input into
the ISCST3 model. For area and mobile sources, temporal profiles are recommended, see for
example (U.S. EPA, 1995b).
1.3.5 Default Source Parameters
Besides the emission rate, the parameters needed to model emissions from point sources
include source location coordinates, physical release height, stack diameter, and stack exit
velocity and temperature. Since most modeling analyses include a large number of sources over
a relatively large area, it is inevitable that there will be gaps in the data for some of the sources.
It is necessary to determine values for all the missing source characteristics, substitute them, and
document the substitutions before the sources can be modeled. For national scale applications.
using default source parameters is the only economical option.
The AIRS data base should be the primary source for identifying substitutions for missing
source locations and missing stack parameters. For point sources with missing data that are not
included in the AIRS data base, values for these missing data fields must be substituted before
modeling those sources. An additional source for default values of stack height, stack diameter.
exit velocity or flow rate, and exit temperature is the Ozone Transport Assessment Group
(OTAG) defaults (OTAG, 1998), which are based on averages calculated by Source
Classification Code (SCC). A list of the default parameters obtained from OTAG is provided in
Appendix D. For any point sources which still have missing parameters and for which default
parameters can not be identified from the OTAG data, the following conservative values are
recommended for use in air toxics modeling analyses:
Stack height 10 meters
Stack diameter 1 meter
Exit temperature ambient
Exit velocity 1 meter/second
12
-------
If the actual location of the point source is missing, and it can not be identified from the
AIRS data base, then the source should be assigned coordinates based on other available data
(e.g., comparisons with other inventories, consultation with EPA Regional Office, etc.).
Since inventories available for analysis may not contain building dimension information,
and since building downwash influences can significantly increase concentrations for receptors
located close to the point source, the following approach may be used to set default values of
building height and building width in the ISCST3 model. Default building dimensions of Hb =
0.625 * Hs and Htt = 2 * Hb (where Hb is building height, Hw is building width and Hs is stack
height) may be used for stack heights of less than or equal to 65 meters, with a minimum
building height of 3.05 meters, representative of a one-story structure. The value used for Hb
places the stack height just above the Schulman-Scire criterion, except for stack heights that are
less than about 4.6 meters, which is 1.5 times the minimum building height of 3.05 meters. The
application of the Schulman-Scire downwash algorithm is therefore limited to the shorter stacks
for which it is more likely to be applicable. The use of the Huber-Snyder downwash algorithm
for stacks that are taller than 4.6 meters also avoids the potential for unrealistically increasing
predicted impacts for these stacks based on relatively arbitrary building information, which could
occur if the Schulman-Scire algorithm were to be applied to those stacks. For stack heights of
greater than 65 meters, assume no building downwash occurs, since stacks of that height are
likely to satisfy good engineering practice (GEP) stack height requirements to avoid building
downwash influences.
Point source inventories, such as those available from AIRS, typically include either a
stack height and stack gas exit parameters (temperature and flow rate), or a plume height for a
given source. Sources with plume heights rather than stack heights and exit parameters can be
considered non-buoyant releases (e.g.. from isolated vents), where the release height is equal to
the stated plume height. Therefore, for those point sources that include only plume height
information and no stack parameters, the plume height should be taken as the release height, and
other stack parameters should be set to zero to give no plume rise. Since these sources are likely
to be from building vents or similar emission points, a building height equal to the stack height
should be assumed, with building width equal to twice the building height. This automatically
triggers the ISCST3 building downwash algorithm, and is a conservative approach.
1.3.6 Source Parameters for Deposition Calculations
The ISCST3 model is capable of estimating wet and dry deposition rates of both gases
and particles. While calculating deposition, the model also calculates the depletion of the
deposited fraction from the plume during transport, resulting in a less conservative, more precise
estimate of air concentrations. Calculating wet deposition requires additional meteorological
data relating to precipitation and scavenging coefficients (U.S. EPA, 1995c). Values for the
required deposition input parameters to ISCST3 should be obtained from the literature. The
values used in the studies for Houston, Texas, and Phoenix, Arizona, below can be considered
appropriate for those air toxic pollutants.
13
-------
1.3.7 Pollutants
The ISCST3 model is run for one pollutant at a time. The number of pollutants should be
carefully defined to minimize resources. Section 112 of the Clean Air Act lists 188 hazardous air
pollutants (HAPs). The draft Integrated Air Toxics Strategy has identified 33 HAPs that are of
primary concern in urban areas (U.S. EPA, 1999b).
1.3.8 Source Grouping
In some emission inventories, large industrial sources (e.g., paper mills, refineries, etc.)
are grouped together so that hundreds of individual release points are assumed to exit from a few
groups of stacks. If an inventory does not contain the individual location and release parameters.
the analyst has little choice but to model the source as a group. Ambient concentrations from
such facilities should be viewed with caution, especially at nearby receptors. If one assumes that
ground level releases (e.g., leaks from pumps, seals or compressors, spilled liquids that form a
puddle and then evaporate, lagoons, etc.) exit through an elevated stack, ground level
concentrations will be underestimated.
From a post-analysis viewpoint, by grouping similar sources (e.g., mobile emissions), the
analyst can more easily look at the impact of different source types. These groupings can be
further subdivided into on-road and off-road mobile source groupings. ISC provides methods for
grouping sources for these purposes.
1.3.9 Quality Assurance
Point source emissions must include all source parameters needed for input in the
ISCST3 model: stack coordinates and source release parameters (stack height, temperature, exit
velocity and diameter). Defaults values should be substituted for missing parameters using the
guidance above.
Point source locations should be verified using GIS. Questionable locations should be
identified. Large emission sources (e.g., greater than 10 tons/yr) should be verified where
possible. A useful source of data for stack location is the AIRS database and, in particular, the
AIRS Facility Subsystem (AIRS/AFS) which is a computer-based repository of information
about airborne pollution. General information about the AIRS database is available at the
Internet web site: http://www.epa.gov/airs/airs2.html. Although some hazardous air pollutant
data are included, AIRS/AFS primarily houses data for criteria pollutants submitted by the
States. The facility information includes data on emissions, process, control, stack, location, etc.
14
-------
1.4 MODEL OUTPUT FOR ANALYSIS
The ISCST3 model output includes concentration or deposition estimates for the various
averaging periods. Annual average concentrations are used in an exposure model like HAPEM4
to estimate inhalation exposure as individuals move among different microenvironments such as
from their homes to their work or school throughout the day. Model output can be further
divided to show the impacts from the different types of sources, e. g., major, area, and onroad
and off-road mobile. Estimates of deposition are used in multi media models.
The data analysis requirements for an air toxics modeling application are likely to be
extensive. Large numbers of sources need to be modeled at a large number of receptor points
that are representative of the exposure regimes/microenvironments and populations found in an
urban area. As noted above, annual averages and the contributions of various sources/groupings
are required, as a minimum.
15
-------
2. CASE STUDY MODELING METHODOLOGY
2.1 INTRODUCTION
Air quality simulation models have a long history of use by the EPA in providing
pollutant concentrations for use in specifying emission limits and assessing control strategies.
The Guideline on Air Quality Models (40CFR51) was established to promote consistency in the
use of models within the air management process. In an urban air toxics study, modeled
concentrations are compared to health/exposure bench mark levels. The use of existing modeling
tools for an urban air toxics study poses special challenges due to the large geographical scale in
urban areas, the large number and variety of sources to be modeled and the variety of pollutants
to be considered.
In this portion of the report the modeling methodology outlined in Section 1 is applied to
two urban areas: Phoenix. Arizona and Houston, Texas. The pollutants modeled are benzene,
1.3-butadiene, formaldehyde, polycyclic organic matter (POM), and hexavalent chromium. This
section addresses the selection of air quality model(s), describes the modeling options and
modeling domains, selection of receptor locations, selection of meteorological data, data analysis
requirements and important aspects of the emissions inventory. Also presented is the application
of a scheme for estimating secondary formaldehyde formation and an overview of the model
results.
This section of the report presents a discussion of the modeling methodology, while
Appendix A and B present the case studies following this methodology for Phoenix, Arizona and
Houston, Texas, respectively. The appendices include descriptions of the emission inventories
used in the case studies, as well as, more detailed results of the air dispersion modeling analyses.
16
-------
2.2 MODELING METHODOLOGY
2.2.1 Model Selection
The extent to which a specific model is suitable for the evaluation of source impact
depends on several factors that include the meteorological and topographic complexities of the
area; the level of detail and accuracy of the data base (i.e., emissions inventory), and the
resources available.
There are a number of design criteria which need to be satisfied in order yield an
acceptable modeling study of toxic pollutants. For the air dispersion model, for example, these
include:
1. readily available/public domain/endorsed by EPA
2. represents state-of-modeling practice
3. applicable to urban areas and irregular terrain
4. capable of handling point, area and mobile sources
5. capable of accounting for dry deposition of pollutants
6. capable of treating atmospheric chemical transformations - pollutant chemistry
7. capable of accounting for pollutant emissions that vary by season and hour-of-day
8. ability to group source types for assessing impact
9. capable of providing annual average concentration estimates (as well as shorter time
averages)
10. computationally efficient
11. good performance - estimated vs. observed concentrations.
The Gaussian plume model is a widely used technique for estimating the impacts of
nonreactive pollutants because of its good performance against field measurements, and because
it is computationally efficient relative to other types of models, such as grid and puff models.
The ISCST3 dispersion model includes the capability of handling multiple point, area.
and mobile sources, incorporates building downwash effects, includes an urban dispersion
option, and also contains considerable flexibility for specifying receptor locations and for
grouping of source impacts. The ISCST3 model also includes algorithms to treat the effects of
elevated and/or complex terrain, and the effects of dry and wet deposition of gaseous and
particulate emissions. The ISCST3 model includes an option to vary emissions by season and
hour-of-day, which was useful in meeting one of the design criteria for this modeling analysis,
since the available emissions inventories reflect variations in emission rates by season and hour-
of-day as inputs. This temporal resolution has also been selected for the model outputs based on
the needs of a typical long term exposure assessment.
The model used for this study was the EPA Industrial Source Complex Short Term
(ISCST3) dispersion model. The ISCST3 model is a steady-state Gaussian plume model which
can be used to assess pollutant impacts from a wide variety of sources such as multiple point,
area and mobile sources. This model was selected for this application to demonstrate what can
be done with off the shelf modeling tools. For this modeling application annual and seasonal
average concentrations, by hour-of-day, were generated. This selection for the temporal
17
-------
resolution of the modeling results was based on the type of data that might be needed for use in a
typical long term human exposure assessment. The ISCST3 model is applicable to receptors
within about 50 km from the source and does not directly simulate the effects of pollutant
chemistry (i.e., chemical transformation and reactivity).
Another approach considered for this study was the use of a photochemical model, such
as the EPA Urban Airshed Model (UAM) (see Guideline on Air Quality Models). However, this
approach had the following drawbacks: a) four of the five pollutants listed in this study were not
treated specifically within UAM in the simulation of photochemical reactions (the exception is
formaldehyde); b) UAM was designed for use during the summer months, hence the photolysis
rates affecting some reactions leading to conversion of benzene, 1,3-butadiene, and POM for
winter are not well characterized; c) UAM could not currently account for the effects of particle
deposition, which were needed to treat hexavalent chromium; and d) there was little experience
running UAM for an entire year (UAM normally is used for one to three day episodic periods).
UAM also could not adequately account for pollutant source apportionment (required for
developing control strategies).
For benzene, 1,3-butadiene and POM, the use of the simpler ISCST3 model was justified
since UAM did not contain photochemical reactions for these pollutants. For hexavalent
chromium. UAM did not handle particulate deposition, while ISCST3 does include a particle
deposition algorithm. While UAM did contain photochemical reactions for formaldehyde, due to
the dissimilarity between the two model's input and output, the time and effort required to use
both the ISCST3 and the UAM models for this study was judged to be prohibitive. Also,
considerable time and resources would have been needed to obtain annual estimates from UAM.
Therefore, except for episodic concentrations of formaldehyde and for determining contributions
from secondary formation of formaldehyde, ISCST3 was the preferred model.
Since ISCST3 did not address effects of secondary transformation for pollutants such as
formaldehyde, a screening level photochemical model (OZIPR) was used to estimate the
magnitude of secondary formaldehyde formations. A simplified approach to estimate secondary
formaldehyde production for the ISCST3 model is described in Section 2.4.
2.2.2 Modeling Options
The regulatory default option of ISCST3 was selected for all modeling runs performed.
This option specifies that the following will be used: final plume rise, buoyancy-induced
dispersion, stack-tip downwash, calms processing routine, default wind profile exponents and
default vertical potential temperature gradients. The modeling option to consider the influence of
elevated and complex terrain was also selected for use for the Phoenix analysis, while flat terrain
was assumed for the Houston analysis. The basis for these selections for treatment of terrain is
discussed in more detail in Section 2.2.3.2.
Since the majority of the modeling domains for each city consists of urban land use
categories, the urban modeling option was selected for all modeling runs, following the guidance
contained in Section 8.2.8 of the Guideline on Air Quality Models (40CFR51).
-------
Some of the toxic pollutants considered in this analysis are photochemically reactive, and
the exponential decay option in the ISCST3 was utilized. The half-life values contained in
Appendix A and B were used.
2.2.3 Receptor Locations
2.2.3.1 Receptor Sampling Strategy
The ISCST3 model incorporates numerous options for specification of receptor locations.
including options for defining grids of receptors and/or discrete receptor locations in a cartesian
and/or a polar coordinate system. These options provide the user with considerable flexibility in
defining receptor locations for a particular application. Since the purpose of this analysis was the
application of dispersion models to the assessment of human exposure to toxic pollutants on an
urban scale, a study was undertaken to evaluate possible strategies for selecting receptor
locations. The outcome of this study, which is documented in Appendix C, addresses the
practical considerations of modeling a large number of sources spread over an urban area, as well
as, the spatial resolution and sampling needs of a typical long term exposure assessment. The
sampling strategy was relatively easy to implement, and significantly reduced the computing
requirements of the analysis (by about a factor of five), while providing reasonable estimates of
the mean and variance of the air concentrations (exposures) within the domain. The sampling
strategy/procedure (Method C) in Section C.7 of Appendix C was applied to determine receptor
points in both the Phoenix and Houston domains.
2.2.3.2 Treatment of Terrain Influences
The ISCST3 model may be run without terrain influences, i.e, flat terrain, or alternatively,
the ISCST3 model will adjust the plume heights by the receptor elevation above or below stack
base to account for the effects of elevated and complex terrain. The ISCST3 User's Guide (U.S.
EPA, 1995c) contains information for handling terrain. The flat terrain option was used for the
Houston analysis, while the elevated and complex terrain options in ISCST3 were used for the
Phoenix analysis. The terrain within the Houston modeling domain is relatively flat with
maximum height variations of about 50 feet. Given that a significant portion of the emissions for
these pollutants is from area sources, and the ISCST3 model ignores terrain influences for area
sources, the flat terrain assumption is considered adequate for Houston. While most of the
Phoenix area averages about 1,100 to 1,300 feet above mean sea level (MSL), there are
significant variations in terrain heights, up to about 2,700 feet MSL, within the modeling
domain. Since the Phoenix modeling domain includes significant terrain features where high
concentrations from stable plume impaction from elevated point sources may be of concern,
terrain influences were addressed.
The terrain elevation for each receptor location modeled for Phoenix was determined
from a file containing the 1-degree Digital Elevation Model (DEM) data, which provides terrain
elevations from the U. S. Geological Survey (USGS) at a horizontal resolution of about 70-90
meters. The 1-degree DEM data can be downloaded for free from the USGS site on the World
Wide Web at http://edcw\vw.cr.usgs.gov/glis/hyper/guide/l_dgr_denfig/indexlm.html. The
selection of 1-degree data for this analysis was based on the cost and availability factors and the
fact that the resolution is considered more than adequate relative to the many other uncertainties
19
-------
in the analysis. Stack base elevations for point sources in the Phoenix inventory were also
determined from the 1-degree DEM data in a the same manner as receptor elevations.
2.2.4 Meteorological Data
2.2.4.1 Selection of Surface and Upper Air Stations
The ISCST3 model requires hourly surface observations of wind speed, wind direction,
ambient temperature, and stability category, in addition to mixing heights derived from twice-
daily upper air soundings as meteorological inputs. The mixing height data, processed by the
National Climatic Data Center (NCDC), and the hourly surface data for major National Weather
Service (NWS) stations are currently available for most cities for years up to 1991 from EPA's
SCRAM Internet web site at http://www.epa.gov/scram001. The appropriate data for the five
most recent years of readily available data were obtained from SCRAM for each city. The
stations and years of data used for each city are summarized below (the numbers in parentheses
are each station's identification number):
City Surface Station Upper Air Station Years
Houston Houston (12960) Lake Charles, LA (03937) 1987-1991
Phoenix Phoenix (23183) Tucson, AZ (23160) 1987-1991
The selection of the surface stations was based on the only available first-order NWS
station for each city. The selection of the upper air stations for mixing heights was based on the
station considered to be the most representative for each city. For Phoenix, the Tucson upper air
station is located about 100 miles away, while the next nearest upper air station is located at
Albuquerque, NM, which is over 300 miles away. For Houston, the Lake Charles upper air
station is located about 135 miles away, while the upper air station at Victoria, TX is about 120
miles away. However, the Victoria station was relocated to Corpus Christi, TX in January 1990,
and 1989 upper air data is missing from SCRAM for both Victoria and Corpus Christi. Since
Victoria is located about the same distance inland from the Gulf of Mexico as both Houston and
Lake Charles, and the distance from Houston to Victoria is comparable to the distance from
Houston to Lake Charles, both stations would be equally representative for use with the Houston
surface data. However, given the fact that the Victoria station was moved to Corpus Christi
resulting in a gap for 1989, and the fact that Corpus Christi is located nearer to the Gulf coast,
Lake Charles was considered to be the better choice for use with Houston. This selection also
corresponds with the recommendation of the Texas Natural Resource Conservation Commission
(TNRCC) for modeling in Harris County, where Houston is located (TACB, 1992).
2.2.4.2 Meteorological Parameters for Deposition Calculations
Several additional meteorological parameters are needed as inputs to the Meteorological
Processor for Regulatory Models (MPRM) in order to implement the dry deposition algorithms
in the ISCST3 model for particulate and gaseous emissions. Additional parameters related to wet
deposition were not needed, since wet deposition was not included in the analysis. The
additional dry deposition parameters are listed below:
20
-------
Albedo
Bowen Ratio
Roughness Length (measurement site)
Roughness Length (application site)
Minimum Monin-Obukhov Length
Surface Heat Flux (fraction of net)
Anthropogenic Heat Flux
Leaf Area Index
These parameters were estimated on a seasonal basis for this analysis, since many of the
parameters will vary significantly by season, and the modeling analysis is designed to produce
average concentrations by season. Based on a review of the guidance provided in Section 3.3 of
the MPRM User's Guide (U.S. EPA, 1996a) for specifying these parameters, the following
values were selected for this analysis:
Phoenix:
Winter
Summer
Fall
Albedo
Bowen Ratio
Roughness Length (measurement site) (m)
Roughness Length (application site) (m)
Minimum Monin-Obukhov Length (m)
Surface Heat Flux (fraction of net)
Anthropogenic Heat Flux (W/rn2)
Leaf Area Index
0.20
2.0
0.15
1.00
50.0
0.25
10.0
0.5
0.14
2.0
0.15
1.00
50.0
0.25
10.0
0.5
0.16
4.0
0.15
1.00
50.0
0.25
10.0
0.5
0.18
4.0
0.15
1.00
50.0
0.25
10.0
0.5
Houston:
Winter
Summer
Fall
Albedo
Bowen Ratio
Roughness Length (measurement site) (m)
Roughness Length (application site) (m)
Minimum Monin-Obukhov Length (m)
Surface Heat Flux (fraction of net)
Anthropogenic Heat Flux (W/m:)
Leaf Area Index
0.20
1.5
0.15
1.00
50.0
025
10.0
1.0
0.14
1.0
0.35
1.00
50.0
025
10.0
1.0
0.16
2.0
0.15
1.00
50.0
0.25
10.0
1.0
0.18
2.0
0.15
1.00
50.0
0.25
10.0
1.0
The Bowen ratio values for Phoenix reflect dry conditions for an urban area, while the
Bowen ratio values for Houston reflect average conditions for an urban area. Since the surface
meteorological data used in the analysis are from major airports, it is assumed that the
measurements are taken from well-sited instruments, away from major obstructions, with a
nominal surface roughness length for the measurement site of 0.15 meters. For the application of
the ISCST3 model in the selected urban areas, the roughness length at the application site was set
at 1.0 meter.
2.2.4.3 Meteorological Preprocessing
The MPRM program was used to preprocess the meteorological data for use with the
ISCST3 model. Both the MPRM (U.S. EPA, 1996a) and PCRAMMET (U.S. EPA, 1996b)
21
-------
meteorological preprocessors can be used to preprocess NWS surface and mixing height data for
use with the ISCST3 model. However, PCRAMMET does not allow for specifying temporal
(e.g., seasonal) or spatial variations of the surface parameters identified in the previous section.
Also, PCRAMMET does not support the additional parameters needed to utilize the dry
deposition algorithm for gaseous pollutants. These additional parameters are leaf area index
(input by the user), and incoming solar radiation (calculated by MPRM). An estimate of
minimum mixing depth for both study areas was determined based on guidance in Section 1.2.6.
2.2.5 Emission Characteristics
The air toxics emission used for this analysis were extracted from the air toxics emissions
inventory prepared by the State of Arizona and the State of Texas. These emissions inventories
were already gridded (e.g., spatially and temporally allocated) and could not be changed. These
aspects are not further discussed.
2.2.5.1 Determination of Background Concentrations
As the guidance in Section 1.2.8 above indicates, background concentrations should be
added to the modeled concentrations to provide total concentration/exposure. However, no
background concentrations of the five modeled chemicals were available for Houston, nor were
POM or hexavalent chromium for Phoenix. Resource limitations did not allow detailed analyses.
For the simulation period, background concentrations of zero were assumed for all five pollutants
in this analysis. This assumption creates uncertainties about the magnitude of the total estimated
concentration.
2.2.5.2 Default Source Parameter Values
Besides the emission rate, the parameters needed to model emissions from point sources
include source location coordinates, physical release height, stack diameter, and stack exit
velocity and temperature. Since the modeling analysis included a large number of sources over a
relatively large area, it was inevitable that there were gaps in the data for some of the sources. It
was therefore necessary to determine default values of the necessary source characteristics to be
substituted for missing data before the sources could be modeled. This was a more significant
problem for the Phoenix inventory, since the only information provided besides the emission rate
was the grid cell containing the point source, and the facility and stack IDs. Source parameters
were available for nearly all of the Houston point and major sources.
Since the Phoenix point source data were identified by facility and stack IDs, the AIRS
data base was used as the primary source to identify substitutions for missing source locations
and missing stack parameters. For point sources with missing data that were not included in the
AIRS data base, default values of the missing data fields were substituted before modeling those
sources. The default values of stack height, stack diameter, exit velocity or flow rate, and exit
temperature were taken from the OTAG defaults, as described in Section 1.3.3. The procedures
for compensating/substituting for missing or erroneous stack parameters are described in Chapter
3, Part 1, Section 3 of the OTAG Technical Supporting Document (OTAG, 1998). For any point
sources which had missing parameters and for which default parameters could not be identified
from the OTAG data, the values given in Section 1.3.3 above were used in the modeling analysis.
22
-------
If the actual location of any point source had been missing, and it could not be identified
from the AIRS data base, then the source would have been assigned coordinates within the grid
cell on a random basis (see Section 1.3.5 above). None of the sources from the Phoenix or
Houston inventories fell into this category. The guidance in Section 1.3.5 concerning building
dimensions was followed.
2.2.5.3 Area and Mobile Source Modeling
The area source emissions that were provided as part of the emission inventory for each
city were spatially allocated to rectangular grid cells of varying sizes, depending on the city. The
grid size for Houston was 2 km x 2 km, and the grid size for Phoenix was 4 km x 4 km. The area
source algorithm for the ISCST3 model was used to physically model these area source
emissions, which is equivalent to modeling the emissions as being uniformly distributed over
each area source grid cell. The ISCST3 area source algorithm also allows for air concentration
estimates to be calculated within the area source itself. The guidance concerning values for
initial dispersion, Section 1.3.4 above, was followed.
2.2.5.4 Source Parameters for Dry Deposition Calculations
ISCST3 is capable of estimating wet and dry deposition rates of both gases and particles.
While calculating the deposition, the model also calculates the depletion of the deposited fraction
from the plume, resulting in a less conservative, more precise, estimate of air concentrations. In
this analysis only the dry deposition algorithms were selected. Chemical-specific scavenging
coefficients were not available for the gaseous pollutants in order to estimate wet deposition.
The neglection of wet deposition, which requires additional meteorological data related to
precipitation, results in a conservative air concentration estimate. Dry deposition of particles was
modeled for hexavalent chromium, with dry deposition of gases modeled for the other four
pollutants.
In order to apply the gas dry deposition algorithm in the ISCST3 model to the gaseous
pollutants of interest (benzene, 1,3-butadiene, formaldehyde and POM), several additional
parameters (see ISC Keywords, U.S. EPA, 1995c) must be specified:
Molecular diffusivity in air (Diff) was obtained from Fletcher, et al., 1997.
The solubility enhancement factor (Alphas) (also referred to as Alpha*) is used when
applying the deposition algorithm over wet surfaces such as moisture on vegetation due to
precipitation or water bodies. No major water surfaces were included in the respective modeling
domains. Since a value for this aqueous phase dissociation is not available for the pollutants in
this study, a value of 1.0 for SO2, suggested in the ISCST3 model User's Guide was used.
The reactivity parameter (Reac) is the scaling factor for "stickiness" of the pollutant and
is pollutant specific. In the absence of observed data in this study, it was set to 10 (a moderate
value).
The mesophyll resistance (Rsubm) can be set to zero for soluble compounds (e.g.
formaldehyde, maximum water solubility = 550000 mg/L), while non-soluble compounds (e.g.
23
-------
naphthalene, maximum water solubility = 30 mg/L) are set to a high value i.e. 100. Values for
maximum water solubility may be found in chemical engineering handbooks and various
publications, such as the Air/Superfund National Technical Guidance Study Series' air emission
model (U.S. EPA, 1993).
Henry's Law coefficient (Henry) is used when applying the deposition algorithm over
wet surfaces. It is a measure of the vapor/water pardoning of a compound. This dimensionless
value is obtained by dividing the Henry's Law constant (values based on Fletcher, et al., 1997) by
the gas constant, R, (8.314 Pa-m3/mol-K) times the ambient temperature, T, (assumed to be
^
Benzene:
Molecular diffusivity in air (Diff) = 0.0912 cm2/sec
Alphas (Alpha*) = 1.0
Reactivity parameter (Reac) = 10.0
Mesophyll resistance (Rsubm) = 10.0
Henry's Law coefficient (Henry) = 0.24
Maximum water solubility - 1,780 mg/L
Henry's Law constant = 543 Pa-nr/mol
1,3-Butadiene:
Molecular diffusivity in air (Diff) = 0. 1 1 58 cnr/sec
Alphas (Alpha*) ' =1.0
Reactivity parameter (Reac) = 10.0
Mesophyll resistance (Rsubm) = 20.0
Henry's Law coefficient (Henry) = 2.95
Maximum water solubility - 735 mg/L
Henry's Law constant = 7,180 Pa-nrVmol
24
-------
Formaldehyde:
Molecular diffusivity in air (Diff) = 0.1698 cm2/sec
Alphas (Alpha*) = 1.0
Reactivity parameter (Reac) = 10.0
Mesophyll resistance (Rsubm) = 0.00
Henry's Law coefficient (Henry) = 0.00
Maximum water solubility = 550,000 mg/L
Henry's Law constant = 0.032 Pa-m3/mol
POM can be defined in a number of ways, e.g., a group of 16-PAH's (polycyclic aromatic
hydrocarbons). Since a precise definition of POM was not provided in the emissions inventor}'.
parameter values used here are for a surrogate compound, naphthalene. Note, there is uncertainty
about this approach. Values for benzo(a)pyrene, or another PAH, could have been used if
available. These values, based on naphthalene were used for POM:
Naphthalene:
Diffusivity in air (Diff) = 0.0590 cm2/sec
Alphas (Alpha*) = 1.0
Reactivity parameter (Reac) = 10.0
Mesophyll resistance (Rsubm) = 100.00
Henry's Law coefficient (Henry) = 0.02
Maximum water solubility = 30 mg/L
Henry's Law constant = 48.6 Pa-m3/mol
The fifth pollutant modeled, hexavalent chromium, is emitted in a particulate form. To
apply the dry deposition algorithm for hexavalent chromium, a particle density and particle size
distribution was input to the model for each source. Particle size distributions are provided in
AP-42 for many source categories that typically emit particulate matter (U.S. EPA, 1995a). As
an example, for chrome electroplating facilities (SCC 3-09-010-18), Section 12.20 of AP-42
provides the following particle size distribution for uncontrolled emissions: 6.9 percent for
particles less than O.S^um; 60.8 percent for particles between 0.5 and 2.4/^m; 14.9 percent for
particles between 2.4 and 8/0/j.m; and 17.4 percent for particles larger than 8.0/^m. This
distribution was used for chrome electroplating facilities, with a particle density of 1.0 gm/cm?.
Similar distributions were determined for other major categories of hexavalent chromium
emissions, based on review of the inventories.
2.2.5.5 Source Grouping
Source grouping was used to allow for tracking and comparison of impacts for various
source types, such as major sources versus area and mobile source types. All source types were
combined into three source groups: major/point sources, area sources, and mobile sources. There
were a few small point sources for which no coordinates were available; these were placed in the
area source group. Modeled concentrations were generated for all sources combined, as well as,
for each of these groups of sources.
25
-------
2.2.6 Model Evaluation
As a preliminary, exploratory model evaluation analysis, available ambient monitoring
data in the Houston area were obtained and compared with modeled estimates. No suitable
ambient monitoring data were available for Phoenix. Appendix B contains a discussion of the
comparison study. Resource permitting, further investigations are warranted.
26
-------
2.3 OVERVIEW OF EMISSION INVENTORIES
This section provides an overview of the emission inventories used in the case studies.
Section 2.3.1 discusses the Phoenix, Arizona inventory, and Section 2.3.2 discusses the Houston,
Texas inventory. Each section describes the sources of the emissions data, the temporal and
spatial resolution of the emissions data, summarizes the data processing that was performed to
get the data into model ready format, and provides brief summaries of the total emissions by
pollutant and source category. Each of the inventories is described in more detail in the
respective case study appendix.
2.3.1 Phoenix, Arizona Inventory
The Phoenix inventory used in this study was extracted from an inventory that was
developed as part of the Arizona HAP research program by the Arizona Department of
Environmental Quality (ADEQ), and the Arizona Department of Health Services (ENSR, 1995).
The original inventory included gridded emissions of 163 pollutant species on a 4 km-by-4 km
grid resolution covering the regions around Phoenix, Tucson, Casa Grande, and Payson, Arizona.
The gridded inventory incorporated all source categories, including major, area and mobile
sources. Temporal variations of the emissions on a season by hour-of-day basis were also
included in the inventory. Table 2.3-1 summarizes the emissions information. Detailed
information about the Phoenix emission inventory can be found in Appendix A, Section 2.
The Phoenix inventory consisted of gridded emission estimates on a 4km-by-4km grid
resolution. A total of 850 grid cells was used to cover the modeling domain. The data consisted
of separate files by season for each pollutant, with each file containing seasonal average emission
estimates by hour of day. The original data also included emission estimates by source
classification code (SCC) for three-level categories. For the five pollutants modeled in this
analysis, the original raw emissions data files consisted of about 66 Mb of data. The process of
converting the Phoenix emissions data to model-ready format for the ISCST3 model involved the
following steps for each pollutant, which were accomplished using several utility programs that
were written in the Fortran programming language:
1. Defining a unique source ID for each grid cell based on the i- and j-cell values in
the data files;
2. Extracting the hour-of-day emission estimates from the four seasonal files for a
particular pollutant;
3. Summing the emissions by source category to the total emissions for the
particular source (i.e., grid cell);
4. Converting the emissions data to the proper units for model input; and
5. Writing out the source locations, physical parameters, and season by hour-of-day
emission rates in the proper format for model input.
27
-------
These basic steps applied to both the major source inventory for Phoenix and the area source
inventor>'. However, the raw data files for major sources included all pollutants within a single
data file for each season, while the area source data were in separate data files for each pollutant.
Minor adjustments to the utility programs to process the data were therefore needed for each type
of source.
The physical source characteristics used to model the area and mobile source emissions
are described in Section 2.2.5.3. Since the raw data files did not include physical source
characteristics for major sources, additional processing was needed to prepare these data for
input to the model. The general approach taken for this is described in Section 2.2.5.2. The
approach was implemented by means of a Fortran utility program that read the raw data files,
attempted to match the facility and source ID with data from AIRS, assigned source locations
and physical stack parameters based on data in the AIRS data base, or applied default values if
no other data were available.
2.3.2 Houston, Texas Inventory
The Houston inventory' used in this study included multiple components. Major source
emissions of the selected pollutants for the Houston area were contained in a spreadsheet. The
area source emission inventory was taken from the inventory developed as part of the Houston
Area Source Toxic Emissions (HASTE) project (Radian, 1995 a and b). It consisted of several
components including total annual emission estimates by pollutant and source category for
Harris County, gridded values of various activity factors, such as population, number of dry
cleaners, etc., and an association between activity factors and source categories that could be
used to spatially allocate the pollutant emissions across the domain. The activity factors, and
therefore the area source emissions derived from them, were gridded with a 2km-by-2km grid
resolution for the Harris County domain.
Table 2.3-2 summarizes the emissions information. Note that in Table 2.3-2, major and
area source emissions are estimates for 1993, and mobile source emission estimates are for the
base year 1990. It is assumed that mobile source emissions did not change appreciably from
1990 to 1993, since no fuel additives were introduced in Houston during that period. More
detailed information about the emissions can be found in Appendix B, Section B.2.
As with the Phoenix inventory, several steps were involved in converting the emissions
inventory data as provided to a model-ready format for the ISCST3 model. However, there were
significant differences in the type and structure of data provided for each city that placed
different requirements on this conversion process. As noted above, the Houston inventory
consisted of three main components: major sources; non-mobile area sources; and mobile
sources. Each of these components had somewhat different processing requirements. For major
sources, the source locations and physical release parameters were provided for nearly all
sources, and required simple reformatting to the input format required for the ISCST3 model
after extracting the data from the spreadsheet file. The more difficult task in preparing the major
sources was in separating the emissions by pollutant and applying toxic speciation factors to
certain sources that were listed as emitting only kerosene, fuel oil, etc. These toxic fractions
varied by source category, and matching the sources provided to the source categories had to be
accomplished manually based on the information available.
28
-------
The area source data provided from the HASTE project required several steps to prepare
them for input to the ISCST3 model. The gridded activity factor data that were used to spatially
allocate the toxic emissions across the Houston domain were provided in the form of RAM
model (40CFR51) input files. A separate RAM model file was provided for each of the 17
activity factors used in the analysis. These activity factors were provided in absolute units per
grid cell, such as number of people per grid cell, number of dry cleaners per grid cell, etc. The
gridded activity factors were combined into a single file that contained the relative activity factor
for each grid cell, and the grid cells were converted to source coordinates for area sources to be
used in the model input files. In addition, some activity factors were combined for use in
spatially allocating certain source categories. A utility program was written in Fortran to
combine data from the gridded activity factors with total emissions by source category into
model-ready inputs for ISCST3.
The mobile source inventory for Houston, provided by EPA's Office of Mobile Sources
(OMS) consisted of gridded VOC emission estimates by season and hour-of-day for on-road and
off-road mobile sources by four-level SCC. The raw data consisted of separate files by season
for both on-road and off-road sources. The OMS also provided toxic fractions for benzene,
1,3-butadiene, formaldehyde, and POM to be applied to the VOC emissions. The process of
converting the mobile source data to model-ready inputs was similar to that applied to the
Phoenix area source data, with the additional step required of applying the toxic fractions by
source category and pollutant. The on-road and off-road emissions were also combined to reduce
the number of area sources (i.e, grid cells) that needed to be modeled.
29
-------
Table 2.3-1 Toxic Air Pollutant Emissions for Phoenix, Arizona
Based on Year 1993
Source Category
Mobile
Sources
On-Road
Off-Road
Total Mobile
Area Sources
Major Sources
TOTAL
Pollutant Emissions (Mg/yr)
Benzene
1,003.0
727.0
1,7300
79.5
128
1.822.3
13-Butadiene
100.0
165.0
265.0
3.5
0.4
268.9
Formaldehyde
605.0
392.0
997.0
20.4
7.1
1,024.5
POM™
195.0
34.4
229.4
29.2
__(3)
258.6
CrVI'11
-J3>
__<3)
-<3>
0.082
0.028
0.11
(1 ) POM = Polycyclic Organic Matter. Individual constituents of POM vary. For mobile source. POM is
defined as sum of 16 chemicals. For area & mobile sources, data provided by TNRCC do not define POM
constituents. Data reported as POM by TNRCC are used here.
(2) CrVI = Hexavalent Chromium
(3) Not Applicable
Table 2.3-2 Toxic Air Pollutant Emissions for Houston, Texas
Based on Year 1993
Source Category
Mobile
Sources
On-Road
Off-Road
Total Mobile
Area Sources
Major Sources
TOTAL
Pollutant Emissions (Mg/vr)
Benzene
971.8
2664
1,238.2
116.9
1,080.7
2,435.8
1.3-Butadiene
159.2
69.4
228.6
4.9
386.5
620.0
Formaldehyde
439.5
1563
595.8
59.0
75.7
730.5
POM01
0.13
0.22
0.35
1.70
0.00
2.05
CrVI(2)
__<_')
(3)
__<3)
2.2
11.9
14.1
(1) POM = Polycyclic Organic Matter. Individual constituents of POM vary. For mobile source, POM is
defined as sum of 1 6 chemicals. For area & mobile sources, data provided by TNRCC do not define POM
constituents. Data reported as POM by TNRCC are used here.
(2) CrVI = Hexavalent Chromium
(3) Not Applicable
30
-------
2.4 APPROACH USED TO ESTIMATE ATMOSPHERIC SECONDARY FORMALDEHYDE
PRODUCTION
2.4.1 Introduction
Formaldehyde, one of the five frequently emitted pollutants in this analysis, is the only
one of the five formed in significant amounts in the atmosphere. It is necessary to estimate the
amount of formaldehyde formed and destroyed by subsequent reactions in the atmosphere.
ISCST3 does not contain a chemical reaction module to handle the transformations of
formaldehyde or other chemicals that occur after emission from a source. These transformations
are often termed "secondary reactions" and the results "secondary production." A simplified
approach to estimate annual average secondary formaldehyde production for the ISCST3 model
is described below. The option of using a refined photochemical model was explored but
rejected because of: 1) the expense of using a photochemical model; 2) the inability of current
models to provide annual average estimates; and 3) data limitations.
An annual average estimate of secondary formation of formaldehyde was approximated
using EPA's research oriented version of the Ozone Isopleth Plotting Package (OZIPR), an air
quality screening model (Gery and Crouse, 1991) which can use very complex chemical
mechanism, in this case the Carbon Bond 4 (CB-4) mechanism, to explore the photochemistry of
a well mixed column of air traveling along a trajectory1. The column of air extends from the
ground surface to the top of the mixed layer, the height varying throughout the day. Initial
chemical species concentrations are changed via dilution by air aloft as the mixed layer height
increases, by time dependent pollutant emissions along the trajectory path, and by the model
photochemistry. A reaction mechanism for use in OZIPR has been adapted to distinguish
between formaldehyde due to primary sources (emissions), and formaldehyde formed in situ
(secondary production). Estimates based on OZIPR are added by a post processor to the initial
ISCST3 model estimates on a seasonal basis and then summed for the annual average. This
simplified approach provides a "ball park" estimate of secondary formaldehyde formation.
The complete process consists of the following:
1. Obtaining the input variables necessary to run OZIPR: meteorological data,
chemical species background concentrations, and chemical species hourly
emissions;
2. Running OZIPR;
3. Using the OZIPR results to estimate secondary formaldehyde concentrations for
use with the ISCST3 model results; and
4. Applying the derived secondary values to the ISCST3 modeled concentrations
based on primary formaldehyde emissions to approximate total formaldehyde
concentrations.
1The OZIPR model may be obtained from EPA's Office of Research and Development, National Exposure
Research Lab. in Research Triangle Park, North Carolina.
31
-------
The analysis is specific to both a city and season. For Houston, Texas, a prototypical day
was determined for each season: winter, spring, summer, and autumn. Seasonally averaged
hourly meteorological data were used. For example, Houston's 8 am temperature values for each
spring day were averaged to simulate the 8 am temperature value; 9 am, 10 am, and so on,
averages were created for the meteorological data. The emissions data for each season were
handled in a like fashion.
2.4.2 Simulation Specific Input
Appendix B describes in detail how to run the OZIPR model for an application similar to
the Houston one. It assumes a user has little familiarity with the model. To run OZIPR for
determining total formaldehyde the following is needed:
1. Location (latitude, longitude);
2. Date (to be used in zenith angle calculation);
3. Meteorology for the period of the simulation;
4. Background concentrations of chemical species; and
5. Hourly emissions throughout the run.
Input is defined for a prototypical day for each season. The prototypical day is defined
using seasonally averaged hourly meteorology, solar radiation, and pollutant emissions. The date
selected (used in zenith angle calculations) for the prototypical day for a given season was the
day that fell in the middle of the season (i.e. July 16 for summer). Seasons for this purpose are:
Winter: December, January, February
Spring: March, April, May
Summer: June, July, August
Fall: September, October, November
The time period selected for a simulation cannot exceed 24 hours. The initial time is
selected so the simulation does not extend into more than one period of daylight. For example, a
24-hour run starts before sunrise and goes through one consecutive period of daylight.
Seasonally averaged meteorology is extracted from local or state records. It is possible to
approximate the hourly values using long term climatological data sets that include morning and
afternoon values for temperature, relative humidity, or mixing height; linearly interpolating
between given values. Mixing heights can be entered at two points in time using the model's
DILUTION option. Morning and afternoon mixing heights are from Holzworth (1972).
Minimum and maximum temperatures and morning and afternoon relative humidities were
obtained from a climatic atlas (NCDC, 1984).
Background concentrations values for NO, NO2, O3, and NO3 for this run used using 6 am
concentrations from a previous day's simulation as noted in Appendix F. However, these values
32
-------
are lower than expected. For example, NOx was initialized to 0.04 ppb which then results in an
initial O3 value of 4.5 ppb instead of an expected value of 100 ppb. It is uncertain at this time
how final model results would differ if different initialization values were used. The user should
consult with a State modeling expert for advice in values to use in initialization.
2.4.3 Analysis of Results for Use with ISCST3
Results from the OZIPR giving both the primary and secondary formaldehyde for the
location, and for a prototypical day for each season, are shown in Figures 2.4-1 through 2.4-4.
The plots shown are for a 24-hour run from 6 am until 5 am the following morning, hours 6-29.
Note the morning maximum of primary formaldehyde, FORM, around 10 am in all four seasonal
simulations. The time of the maximum in formaldehyde emissions for these simulations is 4 pm
in the afternoon, but it is the photochemistry that defines the shape of the curve and depletes the
primary formaldehyde in daylight hours. As the primary emissions of formaldehyde are the
same for all seasons, the differences in the FORM curves are due to meteorology and
photochemical depletion, and are not large.
It should be noted that the secondary formaldehyde, FRMS, contributes a much greater
share of the total formaldehyde. FRMS also exhibits greater seasonal differences, influenced by
both seasonal emissions of precursors and by meteorology.
These results, presented in Figures 2.4-1 through 2.4-4, are used to define a seasonally
dependent scaling factor which relates the amount of secondary to primary formaldehyde at
10 am. The values for 10 am are chosen, as this corresponds to a time of day when primary
formaldehyde values are relatively high for the day. The 10 am seasonally dependent scaling
factor (SF) is then the 10 am secondary formaldehyde concentration from OZIPR divided by the
10 am primary formaldehyde concentration from OZIPR:
OZIPR [10 am Secondary FORM]
SF = - -
OZIPR [10 am Primary FORM]
This scaling factor is then used with the ISCST3 model results for the prediction of
secondary formaldehyde from the primary formaldehyde in the ISCST3 model. (The scaling
factor calculated for Houston, TX in summer was 6.74.)
ISCST3[10 am Secondary FORM] = SF * ISCST3[10 am Primary FORM]
Using a scaling factor for each hour is not recommended, as large errors can be
introduced. The 10 am value was chosen for its minimal SF. For hours other than 10am, the
primary formaldehyde simulated by OZIPR is lower, and any uncertainty in this value will be
amplified in computing an hourly SF value. This method keeps in check amplification of high
outlier values. Instead, the OZIPR model results for secondary formaldehyde are used to
determine the amount of secondary formaldehyde relative to that at 10 am for each hour. The
result is an hourly fraction value (FR), calculated for each hour, which can be used to determine
the values of secondary formaldehyde with respect to the calculated 10 am value. For example,
the 6 am fraction would be:
33
-------
FR( 6 am) =
OZIPR [6 am secondary FORM]
OZIPR [10 am secondary FORM]
Using the approximated value for ISCST3 10 am secondary formaldehyde, and fractions
relating the 10 am secondary formaldehyde for all hours, the final calculations for approximated
ISCST3 secondary formaldehyde can be made. For example, the 6 am value for ISCST3
secondary formaldehyde would be:
ISCST3 [ 6 am secondary FORM] = ISCST3 [10 am secondary FORM]* FR(6 am)
A data file was created for each season for use with ISCST3 which contains the scaling
factors for each season, and the hourly fractional values for each season:
HOUSTON TEXAS SUMMER:
USE HOUR 10 DOMAIN AVERAGE PRIMARY FORMALDEHYDE MULTIPLY BY
6.73699 TO GET 10 AM SECONDARY FORM = FRMS10AM
FRMS(10AM)=(10AM Primary FORM)*(SF summer)
FRMS(HR)=FRMS(10AM)*FR(HR)
^FRMS(IOAM)
^FRMS(IOAM)
;FRMS(10AM)
^FRMS(IOAM)
^FRMS(IOAM)
:FRMS(10AM)
:FRMS(10AM)
^FRMS(IOAM)
^FRMS(IOAM)
TRMS(IOAM)
= FRMS(10AM)
^FRMS(IOAM)
= FRMS(10AM)
= FRMS(10AM)
^FRMS(IOAM)
^FRMS(IOAM)
FRMS(IOAM)
:FRMS(10AM)
FRMS(IOAM)
FRMS(IOAM)
FRMS(IOAM)
FRMS(IOAM)
FRMS(IOAM)
FRMS(IOAM)
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
FRMS
600.000
700.000
800.000
900.000
1000.000
1100.00
1200.00
1300.00
1400.00
1500.00
1600.00
1700.00
1800.00
1900.00
2000.00
2100.00
2200.00
2300.00
2400.00
2500.00
2600.00
2700.00
2800.00
2900.00
1.44
1.24
1.10
1.01
1.00
1.04
1.09
1.13
1.17
1.25
1.28
1.29
1.28
1.28
1.31
1.33
1.35
1.36
1.38
1.40
1.41
1.42
1.43
1.43
34
-------
Total formaldehyde, secondary plus primary, was calculated as following:
FRMT(HR)=FRMP(HR)+FRMS(HR)
2.4.4 Results
Time series plots were generated for total formaldehyde, the sum of the primary and
secondary formaldehyde. A comparison of the time series plots for the primary formaldehyde
and the total formaldehyde in Figures 2.4-1 through 2.4-4 shows that the general diurnal pattern
remains the same with an hourly increase of 0.4 - 0.6 /ug/m3 when the secondary is added. The
isopleth maps in Appendix B also show a slight increase in the spatial distribution of the total
formaldehyde levels within the study area. The maximum concentration increased to 2.5 yug/m?
from 2.1 /wg/m3.
2.4.5 Conclusions
The OZIPR model has been used to approximate the secondary formation of
formaldehyde in Houston, Texas by its relationship to the amount of primary formaldehyde using
the CB-4 mechanisms. In all cases, the secondarily formed formaldehyde was the greater
contributor to total formaldehyde. The resulting approximation for secondary formaldehyde was
applied to ISCST3 model results uniformly throughout the Houston domain.
Because there is no direct chemical relationship between the primary ar.'d secondary
formaldehyde, relating these species empirically can only be applied in a small area where the
local chemistry can be well defined. Using OZIPR requires information on both biogenic and
anthropogenic emissions, which might not readily be on hand. The model is not very flexible
and it was not trivial to use for this application. For example, because OZIPR is a 24-hour
model, running for additional time requires setting up a new input file based on the model output.
This includes initializing of all species represented in the model. The OZIPR results were based
on the conditions for the geographic area of Houston and will not necessarily apply to other
areas.
35
-------
Formaldehyde Concentration
61
L -
Data
FORM
FRMS
TOT FORM
Figure 2.4-1: Formaldehyde concentrations for prototypical summer day in Houston Texas, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde), FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde).
36
-------
Formaldehyde Concentration
fin:
hour
Data
FORM
FRMS
TOT FORM
Figure 2.4-2: Formaldehyde concentrations for prototypical autumn day in Houston, TX, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary7
formaldehyde), FRMS (secondarily produced formaldehyde), and TOTJFORM (total
formaldehyde).
37
-------
Formaldehyde Concentration
4 -i
c_
0 t r r*
L J I I
Data
FORM
FRMS
.TOT FORM
Figure 2.4-3: Formaldehyde concentrations for prototypical winter day in Houston, TX, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde), FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde).
38
-------
Formaldehyde Concentration
Data
.FORM
FRMS
TOT FORM
Figure 2.4-4: Formaldehyde concentrations for prototypical spring day in Houston, TX, at
approximate steady state. OZIPR with hour of simulation vs ppb of: FORM (primary
formaldehyde), FRMS (secondarily produced formaldehyde), and TOT_FORM (total
formaldehyde).
39
-------
2.5 OVERVIEW OF MODELING RESULTS
This section provides an overview of the modeling results obtained from the two case
studies for Phoenix, Arizona and Houston, Texas. A summary of the annual average
concentrations and of the temporal concentration variations is provided. More detailed
summaries of the modeling results are provided in the appendix for each city.
2.5.1 Phoenix, Arizona Modeling Results
Table 2.5-1 summarizes the highest annual-average impacts for each modeled pollutant.
The impacts listed in this table are the combined impacts of all source categories (major, area and
mobile) in the Phoenix area. The largest contributions to the total impacts for all pollutants,
except hexavalent chromium, were due to the mobile sources. The area sources contributed
significantly less than the mobile sources, and the major sources contributed the least. For
hexavalent chromium, for which there were no mobile sources, the major and the area sources
contributed equally to the highest total impacts. The contributions from the individual source
categories, as well as further discussions on the modeling results, including isopleths of
concentrations, can be found in Appendix A.
2.5.2 Houston, Texas Modeling Results
Table 2.5-2 summarizes the highest annual-average impacts for each modeled pollutant.
The impacts listed in this table are the combined impacts of all source categories (major, area and
mobile) in the Houston area. For benzene, 1,3-butadiene and POM, the largest contributions to
the total impacts were due to the major sources, with the area and the mobile sources
contributing only small fractions. For formaldehyde, the largest contributions were from the
mobile sources, while the major and the area source contributions were significantly smaller. For
hexavalent chromium, for which there were no mobile sources, the area sources account for
almost all of the highest total concentration, while the major sources have an insignificant
contribution. The contributions from the individual source categories, as well as further
discussions on the modeling results, including isopleths of concentrations, can be found in
Appendix B.
40
-------
Table 2.5-1 Highest Annual Average Concentrations from All Sources Combined for
Phoenix, Arizona Based on 5 Modeled Years 1987 - 1991
Pollutant
Benzene
1,3-Butadiene
Formaldehyde
POM(2)
Chromium VI(3)
Highest Annual Average
Concentration (^g/m3)
1.37
0.17
0.62
0.18
0.0001
Receptor Location (1) (X,Y) (meters)
(396936, 707195)
(396936, 707195)
(393553, 707047)
(396936, 707195)
(395455.699171)
(1) Receptor locations are in Universal Transverse Mercator (UTM) coordinates for Zone 15.
While modeling, the first digit of the Y coordinate (North UTM) was removed.
(2) POM = Polycyclic Organic Matter
(3) Hexavalent Chromium
Table 2.5-2 Highest Annual Average Concentrations from All Sources Combined for
Houston, Texas Based on 5 Modeled Years 1987 -1991
Pollutant
Benzene
1,3-Butadiene
Formaldehyde
POM<2)
Chromium VI<3)
Highest Annual Average
Concentration|^g/m3)
10.41
26.17
2.13
0.004
0.11
Receptor Location (1) (X,Y) (meters)
(296660, 299970)
(281902,287136)
(275162,319329)
(305619, 292378)
(305619, 292378)
(1) Receptor locations are in Universal Transverse Mercator (UTM) coordinates for Zone 15.
While modeling, the first digit of the Y coordinate (North UTM) was removed.
(2) POM = Polycyclic Organic Matter
(3) Hexavalent Chromium
41
-------
2.6 SUMMARY AND CONCLUSIONS
This section provides an overall summary of the study, including the major conclusions,
based on the emissions data and the modeling results presented in the case study sections of this
report.
The results of the modeling analyses show some significant differences between the two
cities that were studied. For Phoenix, the mobile sources were clearly the dominant source of
emissions for benzene, 1,3-butadiene, formaldehyde and POM; the mobile source emissions for
these pollutants were an order of magnitude higher than area source emissions and two orders of
magnitude higher than major source emissions. For hexavalent chromium, however, the
emissions were due to cooling towers and other major sources. The mobile source emissions
also exhibited the strongest temporal variations, reflecting the diurnal patterns in road traffic, as
\\ell as some influences of meteorology on emission estimates. These patterns in the emission
inventory for Phoenix are also evident in the modeling results for that city. The mobile sources
were clearly the largest contributors to the overall highest impacts. Also, since the majority of
emissions were from mobile sources, they were distributed over the entire domain, with some
spatial variability based on surrogate factors such as population. As a result, there was little
evidence in the modeling results of localized "hot spots" (sharp gradients in concentrations over a
relatively short distance) within the Phoenix domain.
While the Houston inventory showed emissions from mobile sources that were
comparable to Phoenix, major source emissions from Houston were significantly higher than
major source emissions for Phoenix. Benzene emissions from major sources were almost 100
times higher for Houston than for Phoenix, and 1,3-Butadiene emissions were about 700 times
higher. Also, the major source emissions of benzene, 1,3-butadiene and POM were the same
order of magnitude as the corresponding mobile source emissions. As a result of this, the major
sources were found to be the largest contributors to the overall highest impacts in Houston for
these pollutants and the modeling analysis for Houston does exhibit some significant "hot spots"
associated with some of the larger sources of emissions.
These analyses illustrate a methodology that may be applied to similar urban-wide
analyses of area sources. The use of a plume model has certain advantages over puff and grid
models in terms of less stringent input data requirements, and plume models are also more easily
applied to long term exposure analyses. Limited comparisons with modeled data were made
where monitored data were available and are presented in Appendix B. Further insight into the
applicability of such models on this scale of analysis may be gained in future studies involving
more extensive comparisons of modeled concentrations and monitored values.
Although some of the factors that affected the pollutant impacts were identified as part of
this study, there are other factors that were not fully analyzed. Therefore, it may be beneficial to
conduct further studies. For example, a study of the affect of the temporal variations in
emissions when coupled with temporal variations in meteorology could provide further insight
into the observed concentration patterns.
42
-------
3. REFERENCES
40CFR51. Guideline on Air Quality Models. Appendix W to 40CFR51.
ENSR Consulting and Engineering, 1995. Arizona Hazardous Air Pollution Research Program
Volumes 1 & 2. Arizona Department of Environmental Quality, Phoenix, Arizona.
Fletcher, K., Strommen, M., and Kamens, R., 1997. Final Report on Parameters Relating to the
Fates of Selected Atmospheric Pollutants. Department of Environmental Sciences and
Engineering, UNC Chapel Hill.
Gery, M. W. and Grouse, R. R.,1991. User's Guide for Executing OZIPR, U. S. Environmental
Protection Agency, Research Triangle Park, NC.
Holzworth, G. C., 1972. Mixing Heights, Wind speeds, and Potential for Urban Air Pollution
Throughout the Contiguous United States, AP-101, U. S. Environmental Protection
Agency, Research Triangle Park, NC.
OTAG (Ozone Transport Assessment Group), 1998. Technical Supporting Document. The
OTAG Final Report is available on the Internet at
hup:/y\vww.epa.gov/ttn/rto/otag/fmalrpt/.
NCDC (National Climatic Data Center), 1984. Comparative Climatic Data for the United States.
Asheville, NC.
NCDC (National Climatic Data Center), 1993. Solar and Meteorological Surface Observation
Network, Version 1.0. Asheville, NC.
NCDC (National Climatic Data Center), 1997a. Hourly United States Weather Observations.
Asheville, NC.
NCDC (National Climatic Data Center), 1997b. Radiosonde Data of North America. Asheville,
NC.
NCDC (National Climatic Data Center), 1998. International Surface Weather Observations.
Asheville, NC.
Radian Corp., 1995a. Development of the Houston Area Source Toxics Emissions (HASTE)
Inventory. Prepared for Texas Natural Resources Conservation Commission.
Radian Corp., 1995b. Air Quality Dispersion Modeling for the Houston Area Source Toxic
Emissions (HASTE) Project. Prepared for Texas Natural Resources Conservation
Commission.
Sutton, O. G., 1953. Micrometeorology. McGraw-Hill, New York, NY
43
-------
TACB (Texas Air Control Board), 1992. Air Quality Modeling Guidelines. Texas Air Control
Board, Austin, Texas.
U.S. EPA (Environmental Protection Agency), 1993. Air/Superfund National Technical
Guidance Study Series, Models for Estimating Air Emission Rates from Superfund
Remedial Actions. Office of Air Quality Planning and Standards, Research Triangle
Park,NC. EPA-451/R-93-001.
U.S. EPA (Environmental Protection Agency), 1995a. Compilation of Air pollutant Emission
Factors - Volume I: Stationary Point and Area Sources. AP-42. Office of Air Quality
Planning and Standards, Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency), 1995b. The Development and Improvement of
Temporal Allocation Factor Files. Joint Emissions Inventory Oversight Group, Prepared
for the Office of Air Quality Planing and Standard, prepared by Air and Energy
Engineering Research Lab, Washington, DC. EPA-600/R-95-004.
U.S. EPA (Environmental Protection Agency), 1995c. User's Guide for the Industrial Source
Complex (ISC3) Dispersion Models. Office of Air Quality Planning and Standards,
Research Triangle Park, NC. EPA-454/B-95-003b.
U.S. EPA (Environmental Protection Agency), 1996a. Meteorological Processor for Regulatory
Models (MPRM) User's Guide. Office of Air Quality Planning and Standards. Research
Triangle Park. NC. EPA-454'B-96-002.
U.S. EPA (Environmental Protection Agency), 1996b. PCRAMMET User's Guide. Office of
Air Quality Planning and Standards, Research Triangle Park, NC. EPA-454/B-96-001.
U.S. EPA (Environmental Protection Agency). 1999a. The Hazardous Air Pollution Exposure
Model (HAPEM4 ) User's Guide, Draft, Office of Air Quality Planning and Standards,
Research Triangle Park, NC.
U.S. EPA (Environmental Protection Agency), 1999b. Integrated Urban Air Toxics Strategy to
Comply with Section 112(d), 112(c)(3) and Section 202(1) of the Clean Air Act; Draft
Notice. Office of Air Quality Planning and Standards, Research Triangle Park, NC.
44
-------
APPENDIX A
AIR DISPERSION MODELING OF TOXIC POLLUTANTS
IN URBAN AREAS
CASE STUDY FOR PHOENIX, ARIZONA
-------
APPENDIX A - TABLE OF CONTENTS
A. CASE STUDY FOR PHOENIX, ARIZONA A-1
A.I INTRODUCTION A-1
A.2 SUMMARY OF EMISSION INVENTORY FOR PHOENIX A-2
A.3 OVERVIEW OF MODELING RESULTS A-4
A.3.1 Annual Average Modeling Results A-4
A.3.1.1 Concentration Contributions by Source Category A-4
A.3.1.2 Isopleths for Benzene A-5
A.3.1.3 Isoplethsfor 1,3-Butadiene A-5
A.3.1.4 Isopleths for Formaldehyde A-5
A.3.1.5 Isopleths for Polycyclic Organic Matter (POM) A-6
A.3.1.6 Isopleths for Hexavalent Chromium A-6
A.3.2 Modeling Results for Average Maximum Concentrations
by Hour-of-Day A-6
A.3.3 Maximum Seasonal Average Concentrations A-7
A.3.4 Pollutant Decay A-7
A.4 SUMMARY AND CONCLUSIONS A-40
A.5 REFERENCES A-41
A-i
-------
APPENDIX A - LIST OF TABLES
Table A.2-1 Toxic Air Pollutant Emissions for Phoenix, Arizona Based on Year 1993 .... A-3
Table A.3-1 Highest Annual Average Concentrations from All Sources Combined for
Phoenix, Arizona Based on 5 Modeled Years 1987 -1991 A-7
Table A.3-2 Source Category Contributions to Total Annual Average Concentrations
for Phoenix, Arizona A-8
Table A.3-3 Half-Life Decay Values A-9
A-ii
-------
APPENDIX A - LIST OF FIGURES
Figure A.3-1. Modeled Receptor Location Urban Area Source Modeling, Phoenix, Arizona
A-10
Figure A.3-2. Population Density, Phoenix, Arizona A-l 1
Figure A.3-3. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, All Sources (1987-1991) A-12
Figure A.3-4. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, Major Sources (1987-1991) A-13
Figure A.3-5. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, Area Sources (1987-1991) A-14
Figure A.3-6. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, Mobile Sources (1987-1991) A-15
Figure A.3-7. Isopleths of Annual Average Concentrations, Phoenix, Arizona
1,3-Butadiene, All Sources (1987-1991) A-16
Figure A.3-8. Isopleths of Annual Average Concentrations, Phoenix, Arizona
1,3-Butadiene. Major Sources (1987-1991) A-17
Figure A.3-9. Isopleths of Annual Average Concentrations. Phoenix, Arizona
1,3-Butadiene, Area Sources (1987-1991) A-18
Figure A.3-10. Isopleths of Annual Average Concentrations, Phoenix, Arizona
1,3-Butadiene, Mobile Sources (1987-1991) A-19
Figure A.3-11. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde, All Sources (1987^991) A-20
Figure A.3-12. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde, Major Sources (1987-1991) A-21
Figure A.3-13. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde. Area Sources (1987-1991) A-22
Figure A.3-14. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde, Mobile Sources (1987-1991) A-23
Figure A.3-15. Isopleths of Annual Average Concentrations, Phoenix, Arizona
POM, All Sources (1987-1991) A-24
Figure A.3-16. Isopleths of Annual Average Concentrations, Phoenix, Arizona
POM. Area Sources (1987-1991) A-25
Figure A.3-17. Isopleths of Annual Average Concentrations, Phoenix, Arizona
POM, Mobile Sources (1987-1991) A-26
Figure A.3-18. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Hexavalent Chromium, All Sources (1987-1991) A-27
Figure A.3-19. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Hexavalent Chromium, Major Sources (1987-1991) A-28
Figure A.3-20. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Hexavalent Chromium, Area Sources (1987-1991) A-29
Figure A.3-21. Urban Area Source Modeling, Phoenix, Arizona, Benzene Average Maximum
Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological Data .... A-30
Figure A.3-22. Urban Area Source Modeling, Phoenix, Arizona, 1,3-Butadiene Average
Maximum Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data A-31
A-iii
-------
Figure A.3-23. Urban Area Source Modeling, Phoenix, Arizona, Formaldehyde Average
Maximum Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data A-32
Figure A.3-24. Urban Area Source Modeling, Phoenix, Arizona, POM Average Maximum
Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological Data .... A-33
Figure A.3-25. Urban Area Source Modeling, Phoenix, Arizona, Hexavalent Chromium
Average Maximum Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data A-34
Figure A.3-26. Urban Area Source Modeling, Phoenix, Arizona, Benzene Maximum Seasonal
Average Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data A-35
Figure A.3-27. Urban Area Source Modeling, Phoenix, Arizona, 1,3-Butadiene Maximum
Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data A-36
Figure A.3-28. Urban Area Source Modeling, Phoenix, Arizona, Formaldehyde Maximum
Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data A-37
Figure A.3-29. Urban Area Source Modeling, Phoenix, Arizona, POM Maximum Seasonal
Average Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data A-38
Figure A.3-30. Urban Area Source Modeling, Phoenix, Arizona, Hexavalent Chromium
Maximum Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data A-39
A-iv
-------
A. CASE STUDY FOR PHOENIX, ARIZONA
A.I INTRODUCTION
This appendix documents the modeling methodology employed by the EPA in estimating
ambient air concentrations of selected toxic pollutants for the city of Phoenix, Arizona. The
pollutants modeled were benzene, 1,3-butadiene, formaldehyde, polycyclic organic matter
(POM), and hexavalent chromium. The modeling study serves as an example of guidance on the
application of dispersion models to the assessment of exposure to toxic pollutants on an urban
scale.
A-l
-------
A.2 SUMMARY OF EMISSION INVENTORY FOR PHOENIX
Five pollutants were included in this modeling study; benzene, 1,3-tmtadiene,
formaldehyde, polycyclic organic matter (POM) and hexavalent chromium. The emissions for
these pollutants were extracted from an air toxics emission inventory developed by the Arizona
Department of Environmental Quality and the Arizona Department of Health Services as part of
the Arizona Hazardous Air Pollution Research Program (ENSR, 1995). The original inventor)',
which was based on the emissions for the year 1993, included gridded emissions of 163 species
of pollutants on a 4km-by-4km grid resolution covering the regions around Phoenix, Tucson,
Casa Grande, and Payson, Arizona. The gridded inventory incorporated all source categories,
including point, area and mobile sources. Temporal variations of the emissions on a season by
hour-of-day basis were also included in the inventory.
A subset of the original inventory was selected to cover the Phoenix area for this analysis.
This subset consisted of a rectangular domain extending from 349 kilometers to 485 kilometers
Easting and 3,663 kilometers to 3,763 kilometers Northing in Universal Transverse Mercator
(UTM) coordinates in UTM Zone 12.
The combined total 1993 emissions of the five modeled pollutants was 3,374 megagrams
(3,719 tons) in the Phoenix metropolitan area. The largest of these five pollutants was benzene
(1,822 Mg or 2,008 tons), followed by formaldehyde (1,024 Mg or 1,129 tons), 1,3-butadiene
(269 Mg or 296 tons), POM (258 Mg or 285 tons) and hexavalent chromium (0.11 Mg or 0.12
ton). A breakdown of emissions of each pollutant by source category (major, area and mobile) is
presented in Table A.2-1.
As shown in Table A.2-1, the mobile sources account for the largest emissions for all
pollutants except hexavalent chromium. Area sources (i.e., small stationary sources which are
too numerous and diverse to be counted as individual point sources) contributed much less than
mobile sources, and major sources contributed less than 1% of the total pollutant emissions,
where applicable, except for hexavalent chromium (25%). Within the mobile sources, the light-
duty gasoline vehicles and trucks and the gasoline-powered lawn and garden equipment were
among the largest contributors.
There is uncertainty about the POM and hexavalent chromium emissions. POM may
have been undercounted in the Phoenix study because emissions of POM from gasoline-powered
off-road sources were not reported. The study reported emissions of hexavalent chromium from
cooling towers (which are being phased-out), but did not report emissions from chrome-plating
facilities or surface coating operations.
A-2
-------
Table A.2-1
Toxic Air Pollutant Emissions for Phoenix, Arizona
Based on Year 1993
Source Category
Mobile
Sources
Area Sources
On-Road
Off-Road
Total Mobile
Major Sources
TOTAL
Pollutant Emissions (Mg/yr)
Benzene
1,003
727
1.730
79.5
12.8
1.822.4
1,3-
Butadiene
100
165
265
3.5
0.4
269
Formaldehyde
605
392
997
20.4
7.1
1,024.5
POM(1)
195
34.4
229
29.2
_J3|
258
CrVI(2)
__(3)
-_<3>
__(3)
0.082
0.028
0.11
( 1 ) POM = Polycyclic Organic Matter. Individual constituents of POM vary. For mobile source, POM is
defined as sum of 1 6 chemicals. For area & mobile sources, data provided by TNRCC do not define POM
constituents. Data reported as POM by TNRCC are used here.
(2) CrVI = Hexavalent Chromium
(3) Not Applicable
A-3
-------
A.3 OVERVIEW OF MODELING RESULTS
This section presents and discusses the results of the modeling analysis conducted for the
sources located in Phoenix, Arizona. As discussed earlier, the five pollutants that were included
in this study are benzene, 1,3-butadiene, formaldehyde, polycyclic organic matter (POM), and
hexavalent chromium.
A total of 927 sources of benzene were modeled using the ISCST3 model. This number
consisted of 87 major sources and 840 area/mobile sources. For 1,3-butadiene, a total of 859
sources were modeled including 19 major sources and 840 area/mobile sources. For
formaldehyde, a total of 888 sources were modeled including 48 major sources and 840
area/mobile sources. For POM, a total of 589 sources were modeled all of which were
area/mobile sources. For hexavalent chromium, a total of 236 sources were modeled including
28 major sources and 208 area sources. There were no mobile sources for hexavalent chromium.
A total of 356 receptors were modeled. Figure A.3-1 shows the modeled receptor
locations. The modeling was conducted using the five year period 1987 through year 1991
National Weather Service surface meteorological data from Phoenix Airport with mixing heights
from Tucson, Arizona. Figure A.3-2 shows the population density for Phoenix.
Both the annual average concentrations as well as the seasonal average concentrations by
hour of day were calculated. Results for both of these averaging periods, and a discussion on the
contributions to the total annual average concentrations from each source category (major, area
and mobile) are presented below for all pollutants modeled.
The effects of pollutant decay on predicted concentrations were included in this analysis
for three modeled pollutants; 1,3-butadiene, formaldehyde, POM. Hexavalent chromium was not
modeled with decay due to its particulate nature, and benzene was not modeled with decay due to
its long half-life. The modeling results presented below for these pollutants are based on
seasonally variable (cold vs warm) half-lives for each pollutant. A discussion on the effects of
decay is also presented in this section.
It should be noted that low mixing heights, i.e., less than 100 meters, occurred in less than
2.5 percent of all the hours during the five year period that was modeled. Thus the need to adjust
the hourly values up to 100 meters arose quite infrequently.
A.3.1 Annual Average Modeling Results
Table A.3-1 presents the highest annual average concentrations for each of the five
pollutants. The corresponding receptor locations are also listed in the table. The listed
concentrations represent the combined total concentrations of all modeled sources for each
pollutant.
A.3.1.1 Concentration Contributions by Source Category
Table A.3-2 presents the contributions of each source category (major, area and mobile)
to the highest annual concentrations listed in Table A.3-1. As can be seen from this table, the
A-4
-------
mobile sources have the largest contribution to the total impacts. The next largest contribution is
from the area sources. The major sources account for only a small fraction of the total impacts.
For comparison purposes, Table A.3-2 also presents the highest concentrations of all
pollutants due to each of the source categories separately. These results show that, out to the
three source categories, the mobile sources cause the highest concentrations, the area sources
cause the next highest concentrations and the major sources have the smallest concentrations.
However, given that the receptor placement for this study was based on population density, it is
possible that there may not be a receptor close enough to a large source to "capture" its
maximum impacts.
A.3.1.2 Isopleths for Benzene
Figures A.3-3 through A.3-6 show the isopleths of concentration for benzene.1 Figure
A.3-3 shows the isopleths for the highest total concentrations due to all sources. Figures A.3-4,
A.3-5, and A.3-8 show the isopleths for the highest concentrations for each source category, i.e.,
major, area and mobile, respectively. Similar to the results shown in Table A.3-2, the major
source category is the largest contributor (accounting for two-thirds or greater contribution) to
the total concentrations. By contrast, the mobile and area sources contribute only a small
fraction (2% or less) in most of the modeling domain.
A.3.1.3 Isopleths for 1,3-Butadiene
Figures A.3-7 through A. 3-10 show the isopleths of concentration for 1,3-butadiene.
Figure A.3-7 shows the isopleths for the highest total concentrations due to all sources. Figures
A.3-8. A.3-9, and A.3-10 show the isopleths for the highest concentrations for each source
category, i.e., major, area and mobile, respectively. Similar to the results shown in Table A.3-2,
the mobile source category is the largest contributor to the total concentrations (accounting for
two-thirds or greater contribution in the central part of the domain and almost 100% contribution
near the edges of the domain). By contrast, the major sources contribute only a small fraction
(accounting for 1 % or less contribution) in most of the modeling domain.
A.3.1.4 Isopleths for Formaldehyde
Figures A.3-11 through A.3-14 show the isopleths of concentration for formaldehyde.
Figure A.3-11 shows the isopleths for the highest total concentrations due to all sources. Figures
A.3-12, A.3-13, and A.3-14 show the isopleths for the highest concentrations for each source
category, i.e., major, area and mobile, respectively. Again, the mobile source category is the
largest contributor to the total concentrations (accounting for two-thirds or greater contribution).
By contrast, the major sources contribute only a small fraction (accounting for 0.5% or less
contribution) in most of the modeling domain.
1Isopleth contours should be viewed with caution because this shape is also dependent on the software
package used.
A-5
-------
A.3.1.5 Isopleths for Polycyclic Organic Matter (POM)
Figures A.3-15 through A3-17 show the isopleths of concentration for POM. Figure
A.3-15 shows the isopleths for the highest total concentrations due to all sources. Figures A3-16
and A3-17 show the isopleths for the highest concentrations for each source category, i.e., area
and mobile, respectively. Note that there were no major sources for POM. Again, the mobile
source category contributes a majority of total concentrations, accounting for approximately two-
thirds or greater contribution. The area sources contribute approximately one-third or less to the
total concentrations.
A3.1.6 Isopleths for Hexavalent Chromium
Figures A.3-18 through A3-20 show the isopleths of concentration for hexavalent
chromium. Figure A3-19 shows the isopleths for the highest total concentrations due to all
sources. Figures A.3-19 and A.3-20 show the isopleths for the highest concentrations for each
source category, i.e., major and area, respectively. Note that there were no mobile sources for
hexavalent chromium. A comparison of Figures A3-18 and A.3-20 shows that the major sources
contribute the most to the total concentrations in most of the modeling domain. However, as was
shown in Table A.3-2, both the major and the area source categories contribute equally to the
highest total hexavalent chromium concentration.
A.3.2 Modeling Results for Average Maximum Concentrations by Hour-of-Day
The modeling results for the annual average maximum concentrations by hour-of-day
from the entire modeling domain are presented in a series of figures. These values represent the
maximum concentration of each of the four seasonal averages, averaged over the five year
period. These figures show the temporal variation of the annual average concentrations. A
figure was generated for each pollutant showing the concentrations due to all sources and for
each source category (major, area and mobile).
Figures A.3-21 through A.3-25 shows the temporal variation curves for benzene, 1-3
butadiene, formaldehyde, POM, and hexavalent chromium, respectively.
With the exception of the figures for the major source category, all figures show a distinct
peak during the morning hours (7-8 a.m.) and a distinct peak during the evening hours (7-8 p.m.).
The peaks generally occur at the same time for all four seasons. There is also a slight increase in
concentrations observed between the hours of 12 noon and 3 p.m. Although a detailed study
would be necessary to determine the exact reason(s) for the patterns seen, the most likely reason
is the temporal variation of emissions from mobile sources, as well as, related to meteorological
conditions during these hours, as well as. For example, mobile source emissions are higher
during the morning and evening rush hours, and will contribute to the morning and evening
peaks. Since a majority of the emissions are from low-level releases, the peaks may also be
related to an increase in concentrations due to more stable atmospheric conditions at night.
For the major sources, although there is a general trend that shows high concentrations
during the morning and evening hours, as compared to the rest of the day, the high
A-6
-------
concentrations consist of several peaks. Further analysis would be required to explain these
occurrences in detail.
A.3.3 Maximum Seasonal Average Concentrations
From Figures A.3-26 through A.3-30 showing seasonal averages, it can also be noted
that, generally, the concentrations of all pollutants during winter and fall seasons are higher than
spring and summer seasons. This may be associated with the seasonal variations in the operation
of certain types of sources. For example, residential wood combustion occurs primarily during
winter with practically no wood combustion during the summer. The only exceptions to this are
the major sources which do not show any apparent seasonal trend.
Similar to the annual average concentrations, the total seasonal average concentrations
are also dominated by mobile sources. For example, a comparison of Figure A.3-21 and A.3-22
shows that the mobile sources contribute approximately 75% to the observed evening peak. By
contrast, major sources contribute only a small fraction to the total concentrations.
A.3.4 Pollutant Decay
As noted above, pollutant decay was modeled for 1,3-butadiene, formaldehyde and POM.
Pollutant decay was not modeled for benzene because of its long half life or for hexavalent
chromium because of its particulate nature. Decay has been defined seasonally (cold versus
warm). With the study area being located in the southern latitudes, winter has been designated
the cold season with spring, summer and fall designated warm seasons. Appropriate half-life
decay values, shown in Table A.3-3 have been assigned to pollutants using these criteria.
TABLE A.3-1 Highest Annual Average Concentrations from All Sources Combined for
Phoenix, Arizona Based on 5 Modeled Years 1987 -1991
Pollutant
Benzene
1,3-Butadiene
Formaldehyde
POM(2)
Chromium IV(3)
Highest Annual Average
Concentration (ug/m3)
1.37
0.17
0.62
0.18
0.0001
Receptor Location (1> (X,Y)
(meters)
(396936,707195)
(396936,707195)
(393553, 707047)
(396936,707195)
(395455,699171)
(1) Receptor locations are in Universal Transverse Mercator (UTM) coordinates for
Zone 15. While modeling, the first digit of the Y coordinate (North UTM) was
removed.
(2) POM = Polycyclic Organic Matter
(3) CrVI = Hexavalent Chromium
A-7
-------
TABLE A.3-2 Source Category Contributions to Total Annual Average Concentrations
for Phoenix, Arizona
Pollutant
Benzene
1 ,3-Butadiene
Formaldehyde
POM(3)
Chromium VI(3)
Concentrations for AH Sources combined (ug/m3)
Highest Total
Concentrations
1.37
0.17
0.62
0.18
0.00015
Contributions to the Total
Concentrations (l)
Major
0.0046
0.000056
0.00041
N/A(4)
0.000074
Area
0.031
0.002
0.017
0.025
0.000074
Mobile
1.34
0.17
0.604
0.15
N/A(4)
Highest Concentrations for Individual
Source Categories<2) (ug/m3)
Major
0.13
0.0047
0.0034
N/A(4)
0.000074
Area
0.033
0.0021
0.0178
0.026
0.000114
Mobile
1.34
0.17
0.604
0.15
N/A(4)
(1) Represent the contributions from the individual source categories to the listed total concentrations for all sources combined.
(2) Represent the highest concentrations due to the individual source categories. The locations of these concentrations may be
different from the locations of the highest total concentrations for all sources combined.
(3) POM = Polycyclic Organic Matter, Chromium VI = Hexavalent Chromium
(4) N/A = Not Applicable
A-8
-------
Table A.3-3 Half-Life Decav Values
Pollutant
Benzene (1)(3)
1,3-Butadiene"'
Formaldehyde
-------
tjlflflflflm If
MARICOPA COUNTY
YAVAPAI COUNTY
Figure A.3-1. Modeled Receptor Location Urban Area Source Modeling, Phoenix, Arizona
A-10
-------
MARICOPA COUNTY
YAVAPAI COUNTY
LEGEND
E! County boundary
El Major Roads
Population Density
CD < 20 per sq.km.
EH 20 - 400 per sq.km.
n 400 - 1000 per sq.km.
• 1000-2000 per sq.km.
H > 2000 per sq.km.
10 KM 20 KM
Figure A. 3-2. Population Density, Phoenix, Arizona
A-ll
-------
MARICOPA COUNTY
El County boundary
. .
Do.51-0.7f
Q 0.79-1.07
YAVAPAI COUNTY
Figure A.3-3. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, All Sources (1987-1991)
A-12
-------
MARICOPA COUNTY
GILA
COUNTY
^ii^^m&m MSFfr-r-^
r^25miim$3^^:- •,
;jx-.: 'c'!M|pliSgy^j
^^. -v-r^v%^^>;.^t±2I
----...a--. '. -'• —• ,-v-, ; ~'--''' r-
•"--•--•-••"*""' VAVAPAI COUNTY
LEGEND
ESI County boundary
E3 Major Ro«
-------
MAMCOPA COUNTY
YAVAPAI COUNTY
LEGEND
13 County boundary
C3 Major RoMb
D<«.t239
O 9.0239- 0.0282
D 0.0282 -0.0305
E3 0.0305 -0.0313
E3 0.0313 -0.0323
S> 0.0323
ItKM MKM
Figure A.3-5. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, Area Sources (1987-1991)
A-14
-------
MARICOPA COUNTY
YAVAPAI COUNTY
Figure A.3-6. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Benzene, Mobile Sources (1987-1991)
A-15
-------
MAMCOPA COUNTY
0 County boundary
K9 ftUjor RoMb
Concentration. Bated
-t.it
E3>».u
YAVAPAI COUNTY
Figure A.3-7. Isopleths of Annual Average Concentrations, Phoenix, Arizona
1,3-Butadiene, All Sources (1987-1991)
A-16
-------
MARICQPA COUNTY
YAVAPAI COUNTY
LEGEND
B3 County boumlarjr
CooccntniUon. nK>m3
D<0.«M«3
D«.WMM3-«.OOM5
Qd.OMM5-t.OM10
ED 0.00010 -0.00017
E3 0.00017 -0.00039
> 0.00039
MM. CtmcfmtrtOtm »
I.IWMM ^(m3
I* KM MKM
Figure A.3-8. Isopleths of Annual Average Concentrations, Phoenix, Arizona
1,3-Butadiene, Major Sources (1987-1991)
A-17
-------
GILA
COUNTY
PIMA COUNTY
MAWCOPA COUNTY
County boundary
Major RoMlt
ConcCTtratkm. ugfrnJ
D<«.«OM
D«.000t-0.0011
D •.6011-0.0015
0.0015-0.0018
S 0.0018-0.0020
> 0.0020
YAVAPAI COUNTY
Figure A.3-9. Isopleths of Annual Average Concentrations, Phoenix, Arizona
1,3-Butadiene, Area Sources (I987-T99I)
A-18
-------
MARICOPA COUNTY
E3 County boundary
Concentration. ujto3
D
-------
PIMA COUNTY
GILA
COUNTY
MARICOPA COUNTY
ESI County boundary
0 Major Ro«
-------
MARICOPA COUNTY
YAVAPAI COUNTY
LEGEND
0 County bouiMlarr
Concentnitkm. oa/m3
D 0.0002-0.0004
D 0.0004-0.0008
O 0.0008-0.0011
E3 OLOOH -0.0018
E3> iooi8
ItKM »KM
Figure A.3-12. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde, Major Sources (1987-1991)
A-21
-------
MARICOPA COUNTY
YAVAPAI COUNTY
LEGEND
EJ County boundury
Concentration. ugto3
Do.OM-0.OM
Qft.OOR.M13
G3 0.013 -MW
00.0U- 0.018
E3 >•.»!»
MITT!
I0KM 2»KM
Figure A.3-13. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde, Area Sources (1987-1991)
A-22
-------
MARICOPA COUNTY
YAVAPAI COUNTY
LEGEND
09 County boundary
CooctntnUkm. ughn3
DOJI-IJS
Q«i35-«.48
•iCMM
It KM MKM
Figure A.3-14. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Formaldehyde, Mobile Sources (1987-1991)
A-23
-------
GILA
COUNTY
PIMA COUNTY
MARICOPA COUNTY
O County boundary
E3 Major Roads
D
-------
MARICQPA COUNTY
YAVAPAI COUNTY
LEGEND
0 County boundary
E3 Major R«Hfc
Concentration. mhn3
D 0.005 -0.012
Q 0.012 -0.018
Q 0.018 -0.022
C3 0.022 -0.026
E3> 0.026
•L02641
1«KM MKM
Figure A.3-16. Isopleths of Annual Average Concentrations, Phoenix, Arizona
POM, Area Sources (1987-1991)
A-25
-------
MARICOPA COUNTY
El County boundary
G3 Major RaMk
YAVAPAI COUNTY
Figure A.3-17. Isopleths of Annual Average Concentrations, Phoenix, Arizona
POM, Mobile Sources (1987-1991)
A-26
-------
GILA
COUNTY
MARICOPA COUNTY
YAVAPAI COUNTY
LEGEND
0 County boundary
Conccntratioii. u»hn3
D o.oooi j
•i(MOI4*
I«KM MKM
Figure A.3-18. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Hexavalent Chromium, All Sources (1987-1991)
A-27
-------
MAWCQPA COUNTY
YAVAPAI COUNTY
LEGEND
0 County boundary
^ Major Rmdk
Concentration. mha3
<«.MMM2
D6LOOON2-MMM5
CDt.OOMl-t.MM2
E3t.twta-t.MM5
>t.MM5
IMOV74
I0KM WKM
Figure A.3-19. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Hexavalcnt Chromium, Major Sources (1987-1991)
A-28
-------
MARICOPA COUNTY
YAVAPAI COUNTY
LEGEND
0 County boundar
Concentration. ughn3
D< 0.00002
D 0.00002 -0.00003
D 0.00003-0.00005
00.00005-0.00008
00.00008-0.00011
E3> o.oooii
•.000114
10 KM ZtKM
Figure A.3-20. Isopleths of Annual Average Concentrations, Phoenix, Arizona
Hexavalent Chromium, Area Sources (1987-1991)
A-29
-------
Predicted Concentration (ug/m3)
^^ ••*•
CD QTQ
S 5
£ ?
D >•
c/i 2
<*
cr
65
>
o
• «
oo g S
Y> c »
"-1 g c/3
VO s o
-03
2g S
2. » 5?
« S ?
III
5T o 5'
w .5 w
ft **
890-^
— ^ BT
O a- o
M 35
-------
Predicted Concentration (ug/m3)
69
or >
8 "8 §
S S 3
II
s" 3 S'
S. aero
o o -
69 0 hfl
as
or 2
01
0)
!••••
O
o
g;
(5*
-------
Figure 3-40: Urban Area Source Modeling, Phoenix, Arizona, Formaldehyde
Average Maximum Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
8 10 12 14 16 18 20 22 24
All
Area
Major
Mobile
Figure A.3-23. Urban Area Source Modeling, Phoenix, Arizona
Formaldehyde Average Maximum Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
A-32
-------
1.0
0.8
CO
o 0.6
1
0>
o
o
o
•o
0.4
0.2
0.0
8 10 12 14 16 18 20 22 24
All
Area
Mobile
Figure A.3-24. Urban Area Source Modeling, Phoenix, Arizona
POM Average Maximum Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
A-33
-------
Predicted Concentration (ug/m3)
>
U)
0)
-------
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Winter
Spring
Summer — Fall
Figure A.3-26. Urban Area Source Modeling, Phoenix, Arizona
Benzene Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
A-35
-------
CO
o
V- >
I
CD
O
O
O
s>
Q_
0 2 4 6 8 10 12 14 16 18 20 22 24
Winter
Spring
Summer — Fall
Figure A.3-27. Urban Area Source Modeling, Phoenix, Arizona
1,3-Butadiene Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
A-36
-------
CO
c
o
I2
O
o
s
T:
0246
i i i i ii
8 10 12 14 16 18 20 22 24
Winter
Spring
Summer — Fall
Figure A.3-28. Urban Area Source Modeling, Phoenix, Arizona
Formaldehyde Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
A-37
-------
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24
Winter
• Spring
Summer
Fall
Figure A.3-29. Urban Area Source Modeling, Phoenix, Arizona
POM Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
A-38
-------
0.0006
0.0005-
co
"§> 0.0004-
c
o
I
0 0.0003 H
c
o
o
S 0.0002
0.0001 -
0.0000
I I
T I I
0 2 4 6 8 10 12 14 16 18 20 22 24
_,. ._ VA/;«.*^J_
Winter
• Spring
Summer
Fall
Figure A.3-30. Urban Area Source Modeling, Phoenix, Arizona
Hexavalent Chromium Maximum Seasonal Average Concentrations by Hour of Day
AH Sources, 1987-1991 Meteorological Data
A-39
-------
A.4 SUMMARY AND CONCLUSIONS
This appendix presents the results of a study conducted in support of an air quality impact
analysis for five toxic air pollutants emitted from major, area and mobile sources located in
Phoenix, Arizona. In this study, the ambient concentrations attributable to these sources were
estimated through the application of the ISCST3 dispersion model.
Both the annual average concentrations, as well as the seasonal average concentrations by
hour of day, were estimated. The results of the modeling study show that a majority of the total
air concentrations for Phoenix can be attributed to the mobile sources. The area sources were
found to be the next largest contributors. The major sources were found to contribute
insignificantly to the total concentrations.
A study of the hour-by-day variations of seasonal average concentrations showed that the
concentrations are higher during the morning and evening hours, and that the concentrations are
generally higher during winter and fall seasons as compared to spring and summer seasons.
In order to appropriately explain the observed patterns in concentrations, it is
recommended that further detailed studies be conducted. The detailed studies should focus on an
analysis of the temporal variations in emissions for various types of sources and their
contributions to the predicted total concentrations. The studies should also take into
consideration the effects of the meteorological conditions. It is also recommended that these
studies be conducted using at least the same five years of meteorological data as used in this
report to ensure continuity. Long term trends could be identified using additional years of data.
A-40
-------
A.5 REFERENCES
ENSR (Consulting and Engineering), 1995. Arizona Hazardous Air Pollution Research Program
Volumes 1 & 2. Arizona Department of Environmental Quality, Phoenix, Arizona.
A-41
-------
APPENDIX B
AIR DISPERSION MODELING OF TOXIC POLLUTANTS
IN URBAN AREAS
CASE STUDY FOR HOUSTON, TEXAS
-------
APPENDIX B - TABLE OF CONTENTS
B. CASE STUDY FOR HOUSTON, TEXAS B-l
B.I INTRODUCTION B-l
B.2 SUMMARY OF EMISSION INVENTORY FOR HOUSTON B-2
B.3 OVERVIEW OF MODELING RESULTS B-4
B.3.1 Annual Average Modeling Results B-4
B.3.1.1 Concentration Contributions by Source Category B-5
B.3.1.2 Isopleths for Benzene B-5
B.3.1.3 Isopleths for 1,3-Butadiene B-5
B.3.1.4 Isopleths for Primary Formaldehyde B-5
B.3.1.5 Isopleths for Poly cyclic Organic Matter (POM) B-6
B.3.1.6 Isopleths for Hexavalent Chromium B-6
B.3.1.7 Isopleths for Total (Primary and Secondary) Formaldehyde
' '.....B-6
B.3.2 Modeling Results for Average Maximum Concentrations
by Hour-of-Day B-6
B.3.3 Maximum Seasonal Average Concentrations B-7
B.3.4 Pollutant Decay B-7
B.4 PRELIMINARY ANALYSIS OF AIR QUALITY DATA AND MODELED
ESTIMATES B-43
B.5 SUMMARY AND CONCLUSIONS B-49
B.6 REFERENCES B-50
B-i
-------
APPENDIX B - LIST OF TABLES
Table B.2-1 Toxic Air Pollutant Emissions for Houston, Texas Based on Year 1993 B-3
Table B.3-1 Highest Annual Average Concentrations from All Sources Combined for
Houston, Texas Based on 5 Modeled Years 1987 - 1991 B-7
Table B.3-2 Source Category Contributions to Total Annual Average Concentrations
for Houston, Texas B-8
Table B.3-3 Half-Life Decay Values B-9
Table B.4-1 Statistics for Predicted and Observed Concentrations B-44
Table B.4-2 Model Prediction Statistics by Source Type B-45
B-ii
-------
APPENDIX B - LIST OF FIGURES
Figure B.3-1. Modeled Receptor Locations Urban Area Source Modeling, Houston, Texas
B-10
Figure B.3-2. Population Density, Houston, Texas B-l 1
Figure B.3-3. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, All Sources (1987-1991) B-12
Figure B.3-4. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, Major Sources (1987-1991) B-13
Figure B.3-5. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, Area Sources (1987-1991) B-14
Figure B.3-6. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, Mobile Sources (1987-1991) B-15
Figure B.3-7. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, All Sources (1987-1991) B-16
Figure B.3-8. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, Major Sources (1987-1991) B-17
Figure B.3-9. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, Area Sources (1987-1991) B-18
Figure B.3-10. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, Mobile Sources (1987-1991) B-19
Figure B.3-11. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, All Sources (1987-1991) B-20
Figure B.3-12. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, Major Sources (1987-1991) B-21
Figure B.3-13. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, Area Sources (1987-1991) B-22
Figure B.3-14. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, Mobile Sources (1987-1991) B-23
Figure B.3-15. Isopleths of Annual Average Concentrations, Houston, Texas
POM, All Sources (1987-1991) B-24
Figure B.3-16. Isopleths of Annual Average Concentrations, Houston, Texas
POM, Area Sources (1987-1991) B-25
Figure B.3-17. Isopleths of Annual Average Concentrations, Houston, Texas
POM, Mobile Sources (1987-1991) B-26
Figure B.3-18. Isopleths of Annual Average Concentrations, Houston, Texas
Hexavalent Chromium, All Sources (1987-1991) B-27
Figure B.3-19. Isopleths of Annual Average Concentrations, Houston, Texas
Hexavalent Chromium, Major Sources (1987-1991) B-28
Figure B.3-20. Isopleths of Annual Average Concentrations, Houston, Texas
Hexavalent Chromium, Area Sources (1987-1991) B-29
Figure B.3-21. Isopleths of Annual Average Concentrations, Houston, Texas
Total Formaldehyde (Primary and Secondary), All Sources (1987-1991) B-30
B-iii
-------
Figure B.3-23. Urban Area Source Modeling, Houston, Texas, 1,3-Butadiene Average
Maximum Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data B-32
Figure B.3-24. Urban Area Source Modeling, Houston, Texas, Primary Formaldehyde Average
Maximum Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data B-33
Figure B.3-25. Urban Area Source Modeling, Houston, Texas, POM Average Maximum
Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological Data B-34
Figure B.3-26. Urban Area Source Modeling, Houston, Texas, Hexavalent Chromium Average
Maximum Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data B-35
Figure B.3-27. Urban Area Source Modeling, Houston, Texas, Total Formaldehyde Average
Maximum Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data B-36
Figure B.3-28. Urban Area Source Modeling, Houston, Texas, Benzene Maximum Seasonal
Average Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data B-37
Figure B.3-29. Urban Area Source Modeling, Houston, Texas, 1,3-Butadiene Maximum
Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data B-38
Figure B.3-30. Urban Area Source Modeling, Houston, Texas, Primary Formaldehyde
Maximum Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data B-39
Figure B.3-31. Urban Area Source Modeling, Houston, Texas, POM Maximum Seasonal
Average Concentrations by Hour of Day, All Sources, 1987-1991 Meteorological
Data B-40
Figure B.3-32. Urban Area Source Modeling, Houston, Texas, Hexavalent Chromium
Maximum Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data B-41
Figure B.3-33. Urban Area Source Modeling, Houston, Texas, Total Formaldehyde Maximum
Seasonal Average Concentrations by Hour of Day, All Sources, 1987-1991
Meteorological Data B-42
Figure B.4-1. Box Plot of Houston Benzene 1987-1991 Modeled vs. 1993-1994
Monitored B-46
Figure B.4-2. Scatter Plots of Monitored Benzene Values (ug/m3) B-47
Figure B.4-3. Scatter Plots of Modeled Benzene Values (ug/m3) B-48
B-iv
-------
B. CASE STUDY FOR HOUSTON, TEXAS
B.I INTRODUCTION
The purpose of this appendix is to document the modeling methodology employed by
EPA in estimating ambient air concentrations of selected toxic pollutants for Houston, Texas.
The pollutants modeled were benzene, 1,3-butadiene, formaldehyde, polycyclic organic matter
(POM), and hexavalent chromium. The modeling study serves as an example of guidance on the
application of dispersion models to the assessment of exposure to toxic pollutants on an urban
scale.
B-l
-------
B.2 SUMMARY OF EMISSION INVENTORY FOR HOUSTON
The Houston inventory used in this study included multiple components. Major source
emissions of the selected pollutants for the Houston area were contained in a spreadsheet. The
area source emission inventory was taken from the inventory developed as part of the Houston
Area Source Toxic Emissions (HASTE) project (Radian 1995 a and b). It consisted of several
components including total annual emission estimates by pollutant and source category for
Harris County, gridded values of various activity factors, such as population, number of dry
cleaners, etc., and an association between activity factors and source categories that could be
used to spatially allocate the pollutant emissions across the domain. The activity factors, and
therefore the area source emissions derived from them, were gridded with a 2km-by-2km grid
resolution for the Harris County domain. The modeling domain extends from 214 kilometers to
316 kilometers Easting and from 3,266 kilometers to 3,342 kilometers Northing in Universal
Transverse Mercator (UTM) coordinates in UTM Zone 15 (see Figure B.3-1).
Since the area source inventory developed under the HASTE project did not include
mobile source emissions, this component was obtained separately from the U.S. EPA Office of
Mobile Sources (OMS). The mobile source inventory provided by OMS included VOC
emissions that were gridded with the same 2km-by-2km resolution as the HASTE area sources,
and toxic fractions that could be applied to the VOC emissions to obtain emission estimates for
the pollutant species of interest in this study. Since the VOC emission estimates included
averages by season and hour-of-day, this temporal resolution was also applied to the toxic
emissions from mobile sources. Since the benzene toxic fractions for some mobile source
categories varied depending on whether the emissions were from exhaust or evaporative
emissions, OMS also provided a data file containing exhaust versus evaporative fractions for
VOC emissions by source category and county within the modeling domain1.
Emissions of benzene, 1,3-butadiene, formaldehyde, POM, and hexavalent chromium
from the Houston modeling domain totaled 3,803 megagrams (Mg), or 4,192 tons for the base
year of 1993. The largest of these five pollutants was benzene (2,436 Mg or 2,685 tons),
followed by formaldehyde (731 Mg or 806 tons), 1,3-butadiene (620 Mg or 683 tons), POM
(2.05 Mg or 2.26 tons) and hexavalent chromium (14.1 Mg or 15.5 ton). Table B.2-1
summarizes the information.
The HASTE report assumed no emission controls on electroplater chromium emissions.
However, TNRCC indicated that these sources were using emission controls. Therefore,
chromium emissions for the electroplater source category was reduced to 2.345 tons per year.
For POM emissions, TNRCC suggested only using reported POM emissions and not the sum of
16 definition for POM.
•"•The approach for preparation of a toxics mobile source emission inventory is currently being updated
from the approach used here. Guidance on the preparation of emission inventories for toxics air quality modeling is
being prepared by OAQPS. For more information concerning this guidance contact the Info CHIEF help line at
(919)541-5285.
B-2
-------
In contrast to the Phoenix inventory described in Appendix A, Houston major sources
accounted for the largest portion of emissions for 1,3-butadiene, and were a much more
significant component of total emissions for formaldehyde. The mobile source emissions for
Houston were slightly less than for Phoenix for benzene, 1,3-butadiene, and formaldehyde
(ranging from about 15 to 40 percent lower), but were about 650 times lower than Phoenix for
POM emissions. The much lower POM emissions for Houston can not readily be explained.
The Houston inventory shows almost 150 times as much hexavalent chromium emissions
as reported for Phoenix, with a similar ratio of emissions from area sources versus major sources.
The majority of chromium emissions for Houston are from metal plating facilities, with cooling
towers also contributing a significant portion. Hexavalent chromium use in cooling towers is
being phased out.
Table B.2-1 Toxic Air Pollutant Emissions for Houston, Texas
Based on Year 1993
Source Category
Mobile
Sources
On-Road
Off-Road
Total Mobile
Area Sources
Major Sources
TOTAL
Pollutant Emissions (Mg/yr)
Benzene
971.8
266.4
1,238.2
116.9
1,080.7
2435.8
1,3-Butadiene
159.2
69.4
228.6
4.9
386.5
620.0
Formaldehyde
439.5
156.3
595.8
59.0
75.7
730.5
POM(I)
0.13
0.22
0.35
1.70
0.00
2.05
CrVI(2)
..(3)
-<3>
__(3)
2.2
11.9
14.1
(1) POM = Polycyclic Organic Matter. Individual constituents of POM vary. For mobile source, POM is defined
as sum of 1 6 chemicals. For area & mobile sources, data provided by TNRCC do not define POM
constituents. Data reported as POM by TNRCC are used here.
(2) CrVI = Hexavalent Chromium
(3) Not Applicable
B-3
-------
B.3 OVERVIEW OF MODELING RESULTS
This section presents and discusses the results of the modeling analysis conducted for the
sources located in Houston, Texas. As discussed earlier, the five pollutants that were included in
this study are benzene, 1,3-butadiene, formaldehyde, polycyclic organic matter (POM), and
hexavalent chromium.
A total of 4609 sources of benzene were modeled using the ISCST3 model. This number
consisted of 1432 major sources and 3177 area/mobile sources. For 1,3-butadiene, a total of
3372 sources were modeled including 195 major sources and 3177 area/mobile sources. For
formaldehyde, a total of 3324 sources were modeled including 147 major sources and 3177
area/mobile sources. For POM, a total of 3177 sources were modeled all of which were
area/mobile sources. For hexavalent chromium, a total of 1296 sources were modeled including
20 major sources and 1276 area sources. There were no mobile sources for hexavalent
chromium.
A total of 573 receptors were modeled. Figure B.3-1 shows the modeled receptor
locations. The modeling was conducted using the five year period 1987 through 1991 National
Weather Service surface meteorological data from Houston Airport with mixing heights from
Lake Charles, Louisiana. Figure B.3-2 shows the population density for Houston.
Both the annual average concentrations as well as the seasonal average concentrations by
hour of day were calculated. Results for both of these averaging periods, and a discussion on the
contributions to the total annual average concentrations from each source category (major, area
and mobile) are presented below for all pollutants modeled.
The effects of pollutant decay on predicted concentrations were included in this analysis
for three modeled pollutants; 1,3-butadiene, formaldehyde, POM. Hexavalent chromium was not
modeled with decay due to its particulate nature, and benzene decay was not modeled due to the
HAPs long half-life. The modeling results presented below for these pollutants are based on
seasonally variable (cold vs warm) half lives for each pollutant. A discussion on the effects of
decay is also presented in this section.
It should be noted that low mixing heights, i.e., less than 100 meters, occurred in less than
1.4 percent of all the hours during the five year period that was modeled. Thus the need to adjust
the hourly values up to 100 meters arose quite infrequently.
B. 3.1 Annual Average Modeling Results
Table B.3-1 presents the highest annual average concentrations for each of the five
pollutants. The corresponding receptor locations are also listed in the table. The listed
concentrations represent the combined total concentrations of all modeled sources for each
pollutant.
B-4
-------
B.3.1.1 Concentration Contributions by Source Category
Table B.3-2 presents the contributions of each source category (major, area and mobile)
to the highest annual concentrations listed in Table B.3-1. As can be seen from this table, the
major sources have the largest contribution to the total impacts for benzene (98.8%), 1,3
butadiene (99.9%) and hexavalent chromium (100%). Mobile sources have the largest
contribution to the total impact for formaldehyde (100%) and area sources have the largest
contribution to the total impact for POM (100%).
For comparison purposes, Table B.3-2 also presents the highest concentrations of all
pollutants due to each of the source categories separately. These concentration values follow the
same trend seen in the contributions for all sources combined as stated above.
B.3.1.2 Isopleths for Benzene
Figures B.3-3 through B.3-6 show the isopleths of concentration for benzene.2 Figure
B.3-3 shows the isopleths for the highest total concentrations due to all sources. Figures B.3-4,
B.3-5, and B.3-6 show the isopleths for the highest concentrations for each source category, i.e.,
major, area and mobile, respectively. Similar to the results shown in Table B.3-2, the major
source category is the largest contributor (accounting for 98.8%) to the total concentrations. By
contrast, the mobile and area sources contribute only a small fraction (2% or less) in most of the
modeling domain.
B.3.1.3 Isopleths for 1,3-Butadiene
Figures B.3-7 through B.3-10 show the isopleths of concentration for 1,3-butadiene.
Figure 3-7 shows the isopleths for the highest total concentrations due to all sources. Figures
B.3-8, B.3-9, and B.3-10 show the isopleths for the highest concentrations for each source
category, i.e., major, area and mobile, respectively. Similar to the results shown in Table B.3-2,
the major source category is the largest contributor to the total concentrations (accounting for
99.9%). By contrast, the mobile and area sources contribute only a small fraction (accounting for
0.1% or less contribution) in most of the modeling domain.
B.3.1.4 Isopleths for Primary Formaldehyde
Figures B.3-11 through B.3-14 show the isopleths of concentration for primary
formaldehyde. Figure B.3-11 shows the isopleths for the highest total concentrations due to all
sources. Figures B.3-12, B.3-13, and B.3-14 show the isopleths for the highest concentrations
for each source category, i.e., major, area and mobile, respectively. The mobile source category
is the largest contributor to the total concentrations (accounting for greater than 99.9%). By
contrast, the major and area sources contribute negligibly (accounting for 0.1% or less
contribution) in most of the modeling domain.
2Isopleth contours should be viewed with caution because this shape is also dependent on the software
package used.
B-5
-------
B.3.1.5 Isoplethsfor Polycyclic Organic Matter (POM)
Figures B.3-15 through B.3-17 show the isopleths of concentration for POM. Figure
B.3-15 shows the isopleths for the highest total concentrations due to all sources. Figures B.3-16
and B.3-17 show the isopleths for the highest concentrations for each source category, i.e., area
and mobile, respectively. Note that there were no major sources for POM. The area source
category contributes a majority of total concentrations (accounting for 99.8%) contribution. The
mobile sources contribute 0.2% or less to the total concentrations.
B.3.1.6 Isopleths for Hexavalent Chromium
Figures B.3-18 through B.3-20 show the isopleths of concentration for hexavalent
chromium. Figure B.3-18 shows the isopleths for the highest total concentrations due to all
sources. Figures B.3-19 and B.3-20 show the isopleths for the highest concentrations for each
source category, i.e., major and area, respectively. Note that there were no mobile sources for
hexavalent chromium. A comparison of Figures B.3-19 and B.3-20 shows that the major sources
contribute the most to the total concentrations in most of the modeling domain.
B.3.1.7 Isoplethsfor Total (Primary and Secondary) Formaldehyde
Figure B.3-21 shows the isopleths of concentration for total formaldehyde. Figure B.3-21
shows the isopleths for the highest total concentrations due to primary (all sources) plus
secondary formaldehyde. A comparison of Figures B.3-11 and B.3-21 shows the isopleths to be
the same, however, the concentrations have increased with the maximum concentration
increasing by 16.6%.
B.3.2 Modeling Results for Average Maximum Concentrations by Hour-of-Day
The modeling results for the annual average maximum concentrations by hour-of-day for
the entire modeling domain are presented in a series of figures. These values represent the
maximum concentration of each of the four seasonal averages, averaged over the five year
period. These figures show the temporal variation of the annual average concentrations. For
each pollutant, figures were generated for the total concentrations due to all sources and for each
source category (major, area and mobile).
Figures B.3-22 through B.3-27 show the temporal variation curves for benzene, 1-3
butadiene, formaldehyde, POM, hexavalent chromium and total formaldehyde, respectively. All
figures show a distinct diurnal patterns. Benzene and 1,3 butadiene and hexavalent chromium,
whose main constituents are major sources, show maxima occurring during the night hours (6
p.m. - 8 a.m.) and a distinct peak during the early morning hours (1-6 a.m.). POM
concentrations, are dominated by primarily mobile sources, and to a lesser extent area sources.
The diurnal pattern follows the same pattern as that of benzene and 1,3 butadiene. For the major
sources, there is a general trend that shows high concentrations during the night hours, as
compared to the rest of the day. Again, further analysis would be required to explain these
occurrences in detail.
B-6
-------
Formaldehyde follows the same general pattern, however, there are distinct peaks at 7
a.m. and 6 p.m.. The major contributor to formaldehyde concentrations is mobile sources. These
peaks may be attributed to increased motor vehicle activity during peak rush hour. The peaks
generally occur at the same time for all four seasons. Although a detailed study would be
necessary to determine the exact reason(s) for the patterns seen, the most likely reason is the
temporal variation of emissions from mobile sources, as well as, related to meteorological
conditions during these hours. For example, mobile source emissions are higher during the
morning and evening rush hours, and will contribute to the morning and evening peaks. The
peaks may also be related to an increase in concentrations due to more stable atmospheric
conditions at night.
B.3.3 Maximum Seasonal Average Concentrations
From Figures B.3-28 through B.3-33 showing seasonal averages, it can also be noted that,
generally, the concentrations of all pollutants during winter and fall seasons are higher than
spring and summer seasons. This may be associated with the seasonal variations in the operation
of certain types of sources, or the fact that the half lives for all pollutants modeled are shorter
during warm conditions versus cold conditions. For example, residential wood combustion
occurs primarily during winter with practically no wood combustion during summer. The only
exceptions to this are the major sources which do not show any apparent seasonal trend.
B.3.4 Pollutant Decay
As noted above, pollutant decay was modeled for 1,3-butadiene, formaldehyde and
POM. Pollutant decay was not modeled for hexavalent chromium because of its particulate
nature, and for benzene due to its long half-life. Decay has been defined seasonally (cold versus
warm). With the study area being located in the southern latitudes, winter has been designated
the cold season with spring, summer and fall designated warm seasons. Appropriate half-life
decay values have been assigned to pollutants using these criteria, as presented in Table B.3-3.
Table B.3-1 Highest Annual Average Concentrations from All Sources Combined for
Houston, Texas Based on 5 Modeled Years 1987 -1991
Pollutant
Benzene
1,3-Butadiene
Formaldehyde
POM(2)
Chromium VI<3)
Highest Annual Average
Concentration (ug/m3)
10.41
26.17
2.13
0.004
0.11
Receptor Location (1) (X,Y) (meters)
(296660, 299970)
(281902,287136)
(275162,319329)
(305619, 292378)
(305619,292378)
(1) Receptor locations are in Universal Transverse Mercator (UTM) coordinates for Zone 15. While
modeling, the first digit of the Y coordinate (North UTM) was removed.
(2) POM = Polycyclic Organic Matter
(3) Chromium VI = Hexavalent Chromium
B-7
-------
Table B.3-2 Source Category Contributions to Total Annual Average Concentrations
for Houston, Texas
Pollutant
Benzene
1,3-Butadiene
Formaldehyde
POM(3)
Chromium VI(3)
Concentrations for All Sources combined (ug/nr1)
Highest Total
Concentrations
10.41
26.17
2.13
0.0038
0.11
Contributions to the Total Concentrations0'
Major
10.29
26.15
0.0029
N/A (4)
0.11
Area
0.029
0.002
0.0041
0.0038
0.00016
Mobile
0.097
0.04
2.13
0.00002
N/A(4)
Highest Concentrations for Individual Source
Categories'2' (ug/m3)
Major
10.29
26.15
0.21
N/A(4)
0.11
Area
0.28
0.011
0.03
0.0038
0.017
Mobile
0.76
0.63
2.13
0.0023
N/A(4)
(1) Represent the contributions from the individual source categories to the listed total concentrations for all sources combined.
(2) Represent the highest concentrations due to the individual source categories. The locations of these concentrations may be different from the locations of
the highest total concentrations for all sources combined.
(3) POM = Polycyclic Organic Matter, Chromium VI = Hexavalent Chromium
(4) N/A = Not Applicable
B-8
-------
Table B.3-3 Half-Life Decay Values
Pollutant
Benzene <1)(3)
l,3-Butadiene(1)
Formaldehyde (l)
POM(2)
Half-Life (hours)
Cold Season (Winter)
1560
8
6
6
Half-Life (hours)
Warm Seasons (Spring, Summer, Fall)
144
2
2
1.2
( 1 ) From Volume I, Appendix E, Table E. 1
(2) From Volume I, Appendix E, Table E.3, using the shortest reported value
(3) Due to long half-life value, pollutant decay was not used for this pollutant.
B-9
-------
>1AAAAm
WALLER
UNTY
i i
MONTGOMERY flOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FORT BEND COUNTY 4 A
3280000m
GALVESTON COUN
LEGEND
El Courty boundary
* Receptors
+ Monitors
10KM MKM
Figure B.3-1. Modeled Receptor Locations Urban Area Source Modeling, Houston, Texas
B-10
-------
MONTGOMERY (JOUNTY
LIBERTY
COUNTY
FORT BEND COUNTY
UNTY\ / GALVESTON COUN
LEGEND
l/Yl County boundary
El Ma,jor Roads
Population Density
C3 < 20 per sq.kra.
EH 20 - 400 per sq.km.
D 400 -1000 per sq.km.
H 1000 - 2000 per sq.km.
• > 2000 per sq.km.
10 KM 20 KM
Figure B. 3-2. Population Density, Houston, Texas
B-ll
-------
Wv v LIBERTY
COUNTY
MONTGOMERY ttOUNTY
FOJRT BEND COUNTY
LEGEND
ESI County boundary
E3 Mfjor Roadt
D«J5-15»
I«i40CI
It KM It KM
Figure B.3-3. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, All Sources (1987-1991)
B-12
-------
WALLER
COUNTY
MONTGOMERY flOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FOfcT BEND COUNTY
Figure B.3-4. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, Major Sources (1987-1991)
LEGEND
E3 County boundary
Ed M^jor Roads
Concentration, ugfai3
D 0.09 -0.16
00.U-OJ4
0 0.34 - 0.50
Mn. C
-------
WALLER
COUNTY
MONTGOMERY flOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
23 County boundirj
Concentration, m/m3
D< 0.016
G 0.01* -0.029
Q 0.029 -0.046
00.046-0.057
E3 0.057 -OO3
MML Cmcentrdioa
1»KM
20 KM
Figure B.3-5. Isopleths of Annual Average Concentrations, Houston, Texas
Benzene, Area Sources (1987-1991)
B-14
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY dOUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
B3 County boundary
E3 Major Roadc
Concentration. ug/m3
D
-------
WALLER
COUNTY
MONTGOMERY flOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
23 Caantj boundary
0 Major Road.
Concentration. ughn3
D 0.05 -0.08
Q 0.08 -0.13
00.13-0.18
H 0.18 -0.5*
Mn.
241735
1*KM
20 KM
Figure B.3-7. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, All Sources (1987-1991)
B-16
-------
LIBERTY
COUNTY
MONTGOMERY COUNTY
HARRIS COUNTY
LEGEND
E3 County boundary
E3 Major Roads
Concentration. ughn3
D< 0.017
D 0.017- 0.035
G] 0.035-0.089
CD 0.089-0.133
E3 0.133 -0383
EH > 0.383
MM. CoBccnlrilka =
2&1492 l«lM3
1»KM
2* KM
Figure B.3-8. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, Major Sources (1987-1991)
B-17
-------
WALLER
COUNTY
MONTGOMERY BOUNTY
LIBERTY
,\ COUNTY
COUNTY
HARRIS
FOfcT BEND COUNTY
LEGEND
E59 County boundary
E) M^jor Ro«d*
Concentration. ughn3
D< 0.0004
D 0.0004 -0.0008
Q 0.0008 -O.OOH
G3 0,001* -0.0024
0 0.0024 -0.0089
E3> 0.0089
•.01129*
1»KM
2* KM
Figure B.3-9. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, Area Sources (1987-1991)
B-18
-------
WALLER
COUNTY
MONTGOMERY COUNTY
LIBERTY
A COUNTY
HARRIS COUNTY
FOJRT BEND COUNTY
LEGEND
C3 County boundary
El Major Road*
Concentration. nn/mJ
D < O.OJ7
D 0.027- 0.042
13 0.056 -0.067
00.067 -0.091
Mas. C«0ecBtrifkn
«.«2705
1*KM
2* KM
Figure B.3-10. Isopleths of Annual Average Concentrations, Houston, Texas
1,3-Butadiene, Mobile Sources (1987-1991)
B-19
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY «OUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
E3 County boundary
0 Major Road*
Concentration, ug>m3
a«».M
O 0.09 -0.13
CD 0.13 -0.18
00.18-0.20
MM. C«Ka>tr*tioa
2.13332 acta3
I»KM
2* KM
Figure B.3-11. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, All Sources (1987-1991)
B-20
-------
MONTGOMERY SOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
LEGEND
E! County boundary
Concentration, ugfm3
D< 0.003
Do.W3-0.OW
Qo.ttM-t.M4
0 0.014 -OJ04
E3 0.024-0.053
£3 > 0.053
•OM27
It KM
2«KM
Figure B.3-12. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, Major Sources (1987-1991)
B-21
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY flOUNTY
HARRIS CQUOTY
FORT BEND COUNTY
LEGEND
E3 County boundary
E3 Major Roads
Concentration. u>fm3
D 0.031
Max. Ctncattrmtiam m
•.034*94
1«KM
20 KM
Figure B.3-13. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, Area Sources (1987-1991)
B-22
-------
LIBERTY
COUNTY
MONTGOMERY ffOUNTY
HARRIS COUNTY
, ,
FORT BEND COUNTY
Figure B.3-14. Isopleths of Annual Average Concentrations, Houston, Texas
Primary Formaldehyde, Mobile Sources (1987-1991)
LEGEND
01 County boundary
01 Major RoMfc
Concentration. ughn3
D<0.«7
D«.07-o.n
D 0.11 -0.14
00.14-0.17
2.127K
1»KM
MKM
B-23
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY ffOUNTY
HARRIS COUNTY
FORT BEND COUNTY
RAZC
UNTY
LEGEND
El County boundary
Concentration. ug 0.0030
Max. Cmtetatrmtitm =
•.OO3I12
1»KM
» KM
Figure B.3-15. Isopleths of Annual Average Concentrations, Houston, Texas
POM, All Sources (1987-1991)
B-24
-------
WALLER
COUNTY
MONTGOMERY dOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
El County boundary
153 Major Road*
Concentration, ngfrnJ
n 0.0029
••OtJTM
1»KM
20 KM
Figure B.3-16. Isopleths of Annual Average Concentrations, Houston, Texas
POM, Area Sources (1987-1991)
B-25
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY ttOUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
E3 County boundary
Concentration. m/m3
D< 0.00003
D 0.00003 -0.00005
CD 0.00005 -0.00007
13 0.00007 - 0.00008
El 0.00008 -0.00018
E3> 0.00018
1»KM
MKM
Figure B.3-17. Isopleths of Annual Average Concentrations, Houston, Texas
POM, Mobile Sources (1987-1991)
B-26
-------
WALLER
COUNTY
MONTGOMERY dOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
E3 County boundary
(39 Major Roadi
Concentration. ughn3
D< 0.001
D 0.001-0.002
Q 0.002-0.004
00.004-0.006
00.006-0.017
E£3 > 0.017
Max. Con
11W41
1«KM
MKM
Figure B.3-18. Isopleths of Annual Average Concentrations, Houston, Texas
Hexavalent Chromium, AH Sources (1987-1991)
B-27
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY flOUNTY
HARRIS COUNTY
FOjRT BEND COUNTY
LEGEND
E3 County boundary
d Major Road*
Concentration. ujhn3
D< 0.000*
Dtt.OOW-0.OOK
QO.OOK- 0.0035
O 0.0035 -0.0054
E3 0.0054- 0.0158
E3> 0.0158
•.1M2S
1«KM
MKM
Figure B.3-19. Isopleths of Annual Average Concentrations, Houston, Texas
Hexavalent Chromium, Major Sources (1987-1991)
B-28
-------
WALLER
COUNTY
MONTGOMERY ttOUNTY
LIBERTY
COUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
E3 County boundary
E3 Major Road*
Concentration. uaJm3
D 0.0077
It KM
MKM
Figure B.3-20. Isopleths of Annual Average Concentrations, Houston, Texas
Hexavalent Chromium, Area Sources (1987-1991)
B-29
-------
WALLER
COUNTY
LIBERTY
COUNTY
MONTGOMERY COUNTY
HARRIS COUNTY
FORT BEND COUNTY
LEGEND
E3 County boundary
E3 Major Road*
Concentration. vgfaa3
-0.60
L5S03
It KM 2*KM
Figure B.3-21. Isopleths of Annual Average Concentrations, Houston, Texas
Total Formaldehyde (Primary and Secondary), All Sources (1987-1991)
B-30
-------
W
CD ore
a c
N -«
n n
O h-ri
rt ^^
E^ >• V
{/) rt K>
§ 3 '
5 ore G
« rt "i
rt 531 CT
* ff g
•-* s
SO =• >
S3
o o st
QTQ B B
S* «" «P
» a* —
— «< K
O « 2
a o s
65 5 |
"
Predicted Concentration (ug/m3)
D)
o
CT
=:
00
ro
10
I I I I
I I I
I I I I
J L
I II I
-------
60
50
CO
"^ A*
3 40
1
g 30
o
o
o
20
2
o_
10
8 10 12 14 16 18 20 22 24
All
Area
Major
Mobile
Figure B.3-23. Urban Area Source Modeling, Houston, Texas
1,3-Butadiene Average Maximum Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-32
-------
8 10 12 14 16 18 20 22 24
A II
All
A ____ _
Area
Major
Mobile
Figure B.3-24. Urban Area Source Modeling, Houston, Texas
Primary Formaldehyde Average Maximum Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-33
-------
0.010
0.008
n
o 0.006
I
§
o
0 0.004
i
0.002
0.000
All
Area
Mobile
8 10 12 14 16 18 20 22 24
Figure B.3-25. Urban Area Source Modeling, Houston, Texas
POM Average Maximum Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-34
-------
Predicted Concentration (ug/m3)
CO
a
n
x
(TO
w
>o
= 3
si
wi i; ts
- 2 S
>— as
? 2 i
h-> H^
\o x o
ip
<•»• 3 5^
2 o S
2 « °
III
MA (V M
Q2. B 5
O ^ :J^
& B
o
0)
J2.
o
-------
Predicted Concentration (ug/m3)
o -* ro w
I I I I I I I I I I _ I __ I __ I _ I _ I _ L_J _ I
Cd
H
o
O (TO
J*
>• r& u»
K>
g5"!^1
I! i
'I |
VO ffi >•
»,-„!-«
•-42'*
i, » »
^o £?* o
N-i 3 e
.^ c i
2 g S
ff n §
^ 2 e
0 Q.
o>
o n
n
o-S f
S" ^ ere
s. " ««
e B
ro
o>
oo
ro
o
ro
ro
ro
1 - 1 - 1 - L
-------
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Winter
* Spring
Summer
Fall
Figure B.3-28. Urban Area Source Modeling, Houston, Texas
Benzene Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-37
-------
60
50-
co
40
o
I
§ 30
c
o
O
.> 20
T3
-------
CO
•I—>
CD
O
c.
o
o
0)
?
Q.
8 10 12 14 16 18 20 22 24
Winter
Spring
Summer — Fall
Figure B.3-30. Urban Area Source Modeling, Houston, Texas
Primary Formaldehyde Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-39
-------
0.010
0.008-
§ 0.006
I
I
O
^ 0.004
8>
Q_
0.002 -
0.000
Winter
U
0 2 4 6 8 10 12 14 16 18 20 22 24
Spring
Summer — Fall
Figure B.3-31. Urban Area Source Modeling, Houston, Texas
POM Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-40
-------
CO
g
'to
0.8
0.7-
0.6-
0.5
§ °-4
c
o
O
~o n Q —
QJ U.O
CD
0.2 H
0.1-
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24
Winter
Spring
Summer — Fall
Figure B.3-32. Urban Area Source Modeling, Houston, Texas
Hexavalent Chromium Maximum Seasonal Average Concentrations by Hour of Day
All Sources, 1987-1991 Meteorological Data
B-41
-------
H
o
i *
SLw
a c
n -i
BT »
3-
* B 8
ro
(30
^
to
oo
-J K
^ O »
\0 g 05
vo £. O
- "I
2 « o.
I- n 2.
(^ § B*
S n »9
*> 5
a S.
o o
B p
o- H
«-< rt
P: £
o w
o
§
5'
(Q
CO
I
0)
0)
Predicted Concentration (ug/m3)
oo
ro
00
ro
o
-------
B.4 Preliminary Analysis of Air Quality Data and Modeled Estimates
As a preliminary, exploratory model evaluation analysis, available air quality data in the
Houston area were obtained and compared with model estimated concentrations to determine
how the model was performing. Ideally, the monitored and modeling data would be available for
the same time period; however, in this exercise, the modeling period (1987-1991) did not
coincide with the period (1993 -1994) for which monitored data exists or the emissions data
period (1993 for point and area sources and 1990 for mobile sources). Although this disparity
precludes day by day comparisons, annual averages of the monitored and modeling values were
compared as a gross check on the model adequacy.
A review of available ambient air quality data in the Houston area indicated that
1,3-butadiene and benzene were monitored by the State of Texas at four locations in Harris
County, Texas in 1993 and 1994. The location of the four monitored sites is shown in Figure
B.3-1. Data were reported as 24-hour daily average values. For 1,3-butadiene, observed values
were below the minimum detectable limit at many sites for many consecutive days. Thus,
1,3-butadiene data were excluded from further analysis.
Table B.4-1 provides various statistics for predicted and observed concentrations. Model
receptors were located at the same coordinates as each of the four monitors. The number of
modeled values represents the number of daily values in a 5-year meteorological data period,
1987 - 1991 (89 days X 5 years plus one leap day). The number of monitored observations
ranges from 97 to 114, while modeled values are 1826. The mean observed benzene values
range from 3.7 to 6.5 ug/m3 and for predicted values from 0.78 to 1.58 ug/m3. Thus, on average,
the modeled values under predict observed values at all four locations.
Figure B.4-1 provides graphical depictions of the observed to predicted values at the four
monitoring sites for benzene. The figure shows that there is more variance in observed data than
that in the modeled values. To investigate the degree of correlation between the monitored
values, scatter plots comparing observed values among all benzene monitored sites are shown in
Figure B.4-2 and show a slight positive association between the daily values between monitors
26 and 64. However, although the monitored values are in relatively close proximity to each, the
degree of correlation among the monitored data appears low. This indicates that the monitors are
influenced by proximity to nearby sources and micro meteorological conditions.
To investigate the degree of correlation between the modeled values, scatter plots
comparing modeled concentrations at the same benzene monitoring sites are shown in Figure
B.4-3. Similar to the monitoring data comparisons, these figures show that the degree of
correlation among the modeled data at these locations is as low as that for the monitored data.
Such a low correlation suggests that point sources dominate the variability in concentrations.
Thus, the concentrations at these sites are influenced by proximity to nearby sources and
micrometeorological conditions.
To further investigate this assumption, model prediction statistics by source type were
calculated and are shown in Table B.4-3. These data also show that the predominant impact on
B-43
-------
concentrations at these sites is from point sources. These data are also consistent with Table
B.3-2 that show that the predominant impact for benzene is from point sources in Houston.
In conclusion, results show that modeled concentrations are lower than observed values
for benzene and that proximity is important when dealing with point sources. Again, this is a
preliminary, exploratory, analysis not to be confused with a model evaluation exercise.
Table B.4-1 Statistics for Predicted and Observed Concentrations*
Benzene (ug/m3)
Monitor
Site
26
64
803
1035
Predicted
Observed
Predicted
Observed
Predicted
Observed
Predicted
Observed
N
1826
98
1826
89
1826
114
1826
97
Max.
3.2
10.11
5.90
21.25
3.14
114.27
7.60
37.07
Min.
0.05
0.02
0.08
0.02
0.09
0.02
0.17
0.02
10th
Percentile
0.28
1.66
0.20
1.17
0.30
1.59
0.49
1.36
Mean
0.78
3.70
0.99
3.95
0.84
6.51
1.58
5.84
Median
0.69
3.58
0.68
3.15
0.75
3.79
1.35
4.55
90th
Percentile
1.40
6.37
2.23
7.64
1.49
11.11
2.94
8.12
*Note: Model predictions are for all available days in 1987-1991 while monitored values are available only for
selected days during 1993-1994.
B-44
-------
Table B.4-2 Model Prediction Statistics by Source Type
Benzene (ug/m3)
Source
Type
Area
Mobile
Point
All
Monitor
Site
26
64
803
1035
26
64
803
1035
26
64
803
1035
26
64
803
1035
Max.
0.175
0.285
0.255
0.609
1.227
1.322
1.241
1.353
3.139
5.285
2.412
6.927
3.218
5.898
3.143
7.597
Min.
0.002
0.008
0.013
0.086
0.022
0.067
0.041
0.057
0.000
0.000
0.000
0.000
0.054
0.076
0.094
0.169
10th
Percentile
0.007
0.019
0.032
0.166
0.056
0.137
0.092
0.131
0.148
0.000
0.055
0.030
0.279
0.197
0.301
0.489
Mean
0.035
0.060
0.079
0.265
0.127
0.269
0.182
0.232
0.621
0.659
0.580
1.088
0.784
0.988
0.841
1.585
Median
0.032
0.055
0.075
0.267
0.108
0.245
0.167
0.216
0.511
0.281
0.510
0.891
0.689
0.684
0.746
1.347
90th
Percentile
0.067
0.108
0.131
0.361
0.220
0.435
0.287
0.339
1.237
1.869
1.182
2.393
1.396
2.229
1.486
2.938
B-45
-------
120
100
80
60
20
H*
Site 26
Site 64
Site 803
Site 1086
o Modeled * Monitor*
Figure B.4-1. Box Plot of Houston Benzene 1987-1991 Modeled vs. 1993-1994 Monitored
B-46
-------
Motor 26
40
35
30
25
»
6
B
5
0
Motor 64
40
35
30
25
20
E
V
S
0
1
Motor 28 M. Ifator 64
y
/
/
/
/
. /
A • /
zjp. .' '
j. • • • _ t •
• •
0 5 062025303540
Montr 035
Figure B.4-2. Scatter Plots of Monitored Benzene Values (ug/m3)
B-47
-------
9to2tM.9k64
9k M
SIM C35 «1 Sli 28
1234567690
TT"
9
SUM
0
9
6
7
6
5
4
S
2
1
0
SkKBM.StoM
TTT'1 T t~ri 1 T"f I T T~T1 TT
4567
SkM
9
6
7
6
5
4
3
2
t:
0
Sto t95 » Sto 64
Sib 095
sura
o
9
6
7
6
5
4
3
2
I;
O1
Figure B.4-3. Scatter Plots of Modeled Benzene Values (ug/m3)
T IT | IT T T"pTT
23
T-prTTT-J I Tf
45
SfeOB
' " I
0
B-48
-------
B.5 SUMMARY AND CONCLUSIONS
This appendix presents the results of a study conducted in support of an air quality impact
analysis for five toxic air pollutants emitted from major, area and mobile sources located in
Houston, Texas. In this study, the ambient concentrations attributable to these sources were
estimated through the application of the ISCST3 dispersion model.
Both the annual average concentrations, as well as the seasonal average concentrations by
hour of day, were estimated. The results of the modeling study show that a majority of the total
concentrations for Houston can be attributed to the major point sources. The mobile sources
were found to be the next largest contributors. The area sources were found to contribute least to
the total concentrations.
A study of the hour-by-day variations of seasonal average concentrations showed that the
concentrations are higher during the morning and evening hours, and that the concentrations are
generally higher during winter and fall seasons as compared to spring and summer seasons.
A preliminary comparison between available air quality data in Houston area with model
estimated concentrations showed that on average the model predictions are lower than monitored
values. Due to the differences in the meteorological, emissions and monitored value periods,
additional analysis is needed to verify this conclusion.
In order to appropriately explain the observed patterns in concentrations, it is
recommended that further detailed studies be conducted. The detailed studies should focus on an
analysis of the temporal variations in emissions for various types of sources and their
contributions to the predicted total concentrations. The studies should also take into
consideration the effects of the meteorological conditions. It is also recommended that these
studies be conducted using at least the same five years of meteorological data as used in this
report. Long term trends could be identified using additional years of data.
B-49
-------
B.6 REFERENCES
Radian Corp., 1995a. Development of the Houston Area Source Toxics Emissions (HASTE)
Inventory. Prepared for Texas Natural Resources Conservation Commission.
Radian Corp., 1995b. Air Quality Dispersion Modeling for the Houston Area Source Toxic
Emissions (HASTE) Project. Prepared for Texas Natural Resources Conservation
Commission.
B-50
-------
APPENDIX C
PROPOSED METHODS FOR SELECTING RECEPTOR SAMPLES
FOR THE APPLICATION OF THE ISCST3 DISPERSION MODEL
TO URBAN AREAS
-------
APPENDIX C - TABLE OF CONTENTS
C. PROPOSED METHODS FOR SELECTING RECEPTOR SAMPLES FOR THE
APPLICATION OF THE ISCST3 DISPERSION MODEL C-l
C.I INTRODUCTION C-l
C.2 INITIAL EVALUATION OF SAMPLING METHODS A AND B C-3
C.2.1 Sampling Objectives C-3
C.2.2 Population of Interest C-3
C.2.3 Finite Population Survey Sampling versus Continuous Spatial
Sampling C-4
C.2.4 The General Sampling Design C-4
C.2.5 Defining the Strata C-4
C.2.6 Sample Size Determination C-10
C.2.7 Selection of Samples by Method A C-16
C.2.8 Selection of Samples by Method B C-16
C.2.9 Decision to Perform Sensitivity Analyses C-20
C.3. SUMMARY OF SAMPLING METHODS A AND B C-21
C.3.1 Step-by-Step Procedure C-21
C.3.2 Limitations of Methods A and B C-22
C.4 SENSITIVITY ANALYSES OF SAMPLING METHODS A AND B C-23
C.4.1 Evaluation of Sample Sizes Calculated by Methods A and B C-24
C.4.2 Effects of Reducing Nominal Sample Sizes C-28
C.5 REFINEMENTS TO SAMPLING METHODOLOGY BASED ON FINDINGS
OF SENSITIVITY ANALYSES C-36
C.5.1 Proposed Refinements C-36
C.5.2 Revised Methodology Applied to Annual Average Data C-37
C.5.3 Application of Method C to Data from Specific Hourly/Seasonal
Combinations C-39
C.6 APPLICATION OF SAMPLING METHOD C TO HOUSTON BENZENE
CONCENTRATIONS C-48
C.6.1 Houston Study Area C-48
C.6.2 Defining the Strata for Method C C-49
C.6.3 True Sampling Distributions C-53
C.6.4 Method C Applied to Annual Average Data with Emissions-Defined
Strata C-53
C.6.5 Method C Applied to the Hourly/Seasonal Data with Emissions-Defined
Strata C-57
C.6.6 Results Using HAPEM-Like Strata C-63
C.6.7 Forced Receptors - Certainty Units C-63
C.6.8 Summary Discussion C-71
C.7 STEP-BY-STEP GUIDELINES FOR USING SAMPLING METHOD C C-73
C.8 REFERENCES C-77
C-i
-------
APPENDIX C - LIST OF TABLES
Table C.2-1 The Distribution of Emissions Values Across the Census Block Groups Within
Each Stratum C-l 1
Table C.2-2 Number of Grid Cells and Total Population for Each Stratum C-l 3
Table C.2-3 Sample Sizes Selected for Each Stratum in the Central Region C-l 5
Table C.4-1 Comparison of Sample Size Calculation Results Using Concentration Values
versus Using the Emissions Values C-25
Table C.4-2 Results of 200 Monte Carlo Samples for Sampling Methods A and B Using
Sample Sizes as Calculated in Table 2-3 C-27
Table C.4-3 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a
Nominal Size 30 Drawn from Each Stratum C-29
Table C.4-4 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a
Nominal Size 25 Drawn from Each Stratum C-30
Table C.4-5 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a
Nominal Size 20 Drawn from Each Stratum C-31
Table C.4-6 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a
Nominal Size 15 Drawn from Each Stratum C-32
Table C.4-7 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a
Nominal Size 10 Drawn from Each Stratum C-33
Table C.4-8 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a
Nominal Size 5 Drawn from Each Stratum C-34
Table C.5-1 Results of 200 Monte Carlo Samples for Sampling Method A When Applied to
Annual Average Data C-38
Table C.5-2 Descriptive Statistics for the Winter, 7 a.m., Concentration Values and Results of
200 Monte Carlo Samples Using Sampling Method C C-40
Table C.5-3 Descriptive Statistics for the Winter, 12 p.m., Concentration Values and Results
of 200 Monte Carlo Samples Using Sampling Method C C-41
Table C.5-4 Descriptive Statistics for the Winter, 12 a.m., Concentration Values and Results
of 200 Monte Carlo Samples Using Sampling Method C C-42
C-ii
-------
Table C.5-5 Descriptive Statistics for the Summer, 7 a.m., Concentration Values and Results
of 200 Monte Carlo Samples Using Sampling Method C C-43
Table C.5-6 Descriptive Statistics for the Summer, 12 p.m., Concentration Values and Results
of 200 Monte Carlo Samples Using Sampling Method C C-44
Table C.5-7 Descriptive Statistics for the Summer, 12 a.m., Concentration Values and Results
of 200 Monte Carlo Samples Using Sampling Method C C-45
Table C.6-1 Sampling Results for Annual Average Houston Benzene Concentrations Using the
Emissions-Defined Strata C-56
Table C.6-2 Sampling Results for Winter 7 am Houston Benzene Concentrations Using the
Emissions-Defined Strata C-58
Table C.6-3 Sampling Results for Winter 12 pm Houston Benzene Concentrations Using the
Emissions-Defined Strata C-59
Table C.6-4 Sampling Results for Summer 7 am Houston Benzene Concentrations Using the
Emissions-Defined Strata C-60
Table C.6-5 Sampling Results for Summer 12 pm Houston Benzene Concentrations Using the
Emissions-Defined Strata C-61
Table C.6-6 Sampling Results for Summer 12 am Houston Benzene Concentrations Using the
Emissions-Defined Strata C-62
Table C.6-7 Sampling Results for Annual Average Houston Benzene Concentrations Using the
HAPEM-Like Strata C-64
Table C.6-8 Sampling Results for Winter 7 am Houston Benzene Concentrations Using the
HAPEM-Like Strata C-65
Table C.6-9 Sampling Results for Winter 12 pm Houston Benzene Concentrations Using the
HAPEM-Like Strata C-66
Table C.6-10 Sampling Results for Summer 7 am Houston Benzene Concentrations Using the
HAPEM-Like Strata C-67
Table C.6-11 Sampling Results for Summer 12 pm Houston Benzene Concentrations Using the
HAPEM-Like Strata C-68
Table C.6-12 Sampling Results for Summer 12 am Houston Benzene Concentrations Using the
HAPEM-Like Strata C-69
C-iii
-------
Table C.6-13 The Sampling Distribution for Stratum 3 of the Emissions-Defined Strata and
Stratum 9 of the Hapem-like Strata after the Outlying Block Group at Utm Coordinates
283.735 East, 3289.09 North Is Forced into the Sample C-70
Table C.7-1 Example Sas Code3 for Generating a Sample of Receptor Points Using
Method C C-74
Table C.7-2 Sample Inputs for Step 5 Calculations (Exposure District No. 1 of Hypothetical
Study Area) C-75
C-iv
-------
APPENDIX C - LIST OF FIGURES
Figure C.2-1 Contour map of the Phoenix study area showing cell emission totals and 4 km
by 4 km emissions inventory grid cells C-6
Figure C.2-2 Contour map of the Phoenix study area showing cell emission totals and the
locations of all block group centroids C-7
Figure C.2-3 Strata constructed for application of sampling procedure to central region of
Phoenix C-8
Figure C.2-4 Strata constructed for application of sampling procedure to central region of
Phoenix with superimposed centroids of all block groups C-9
Figure C.2-5 Location of block groups selected by Method A C-17
Figure C.2-6 Location of block groups selected by Method B C-19
Figure C.6-1 Contour map of the Phoenix region indicating cell emission totals C-50
Figure C.6-2 Emissions-defined strata based on Phoenix contour map presented in
Figure 6-1 C-51
Figure C.6-3 HAPEM-like strata defined for Phoenix region C-52
Figure C.6-4 Distribution of block-group centroids over emissions-defined strata C-54
Figure C.6-5 Distribution of block-group centroids over HAPEM-like strata C-55
C-v
-------
C. PROPOSED METHODS FOR SELECTING RECEPTOR SAMPLES FOR THE
APPLICATION OF THE ISCST3 DISPERSION MODEL
C.I INTRODUCTION
The U.S. Environmental Protection Agency (EPA) is preparing tools to estimate the
exposures to five toxic pollutants [benzene, butadiene, formaldehyde, hexavalent chromium, and
polycyclic organic matter (POM)] for two cities (Houston, and Phoenix). It is likely that the
Hazardous Air Pollutant Exposure Model (HAPEM) (Johnson, et.al., 1996) will be used to
develop these estimates. This model requires input data showing the spatial pattern of outdoor
concentrations for a particular pollutant across the specified study area. To achieve this
objective, researchers apply a dispersion model to an emissions data base for each study area to
produce estimates of pollutant concentration at selected "receptor points." These concentrations
are then averaged over a set of user-defined "exposure districts." These districts are used in the
HAPEM exposure assessments as potential locations for homes and work places. Exposures
within each district are estimated as a function of (1) the average outdoor pollutant concentration
determined for the district and (2) the microenvironment occupied by the exposed individual.
Because HAPEM uses census data in computing exposures, it is useful to define HAPEM
districts as aggregates of census units such as blocks, block groups or census tracts. To better
relate these districts to dispersion model estimates of outdoor pollution levels, analysts typically
use the centroids of selected census units as the receptor points for the dispersion model runs. As
the computational time required for a dispersion model run increases with the number of receptor
points used, it is advantageous to minimize the number of census units used as receptor locations.
This goal can be accomplished by (1) defining a population of possible receptor points that
provides a good characterization of the residential patterns in the study area and then (2) using
statistical techniques to select a representative sample from this population that meets the
minimum requirements of the analyst.
Researchers developed two candidate schemes (Method A and Method B) for selecting
this sample which can be generalized to other pollutants in other study regions. Section C.2 of
this report describes these methods and provides examples of the application of each method to
one pollutant (1,3-butadiene) in one study area (Phoenix). Section C.3 provides a step-by-step
procedure for implementing each method and summarizes the principal limitations of Methods A
andB.
Researchers performed a series of sensitivity analyses using the Phoenix butadiene data to
test the statistical assumptions underlying Methods A and B. Section C.4 summarizes these
analyses, identifies Method B as the superior method, and concludes that both methods tend to
select samples which are larger than necessary to achieve a specified set of sampling goals.
Section C.5 presents Method C, a revised version of Method B, which is capable of achieving the
sampling goals using a significantly smaller sample. To illustrate the generalizability of Method
C, researchers applied it to annual average data and to data representing various combinations of
time of day and season. Section C.5 presents results of these analyses and discusses how the
method may be used in future exposure analyses.
C-l
-------
Analysts noted that the Phoenix results were representative of a region dominated by area
sources and may not be indicative of results obtained for a region dominated by point sources.
Consequently, the analysis was repeated using benzene data for Houston, an area with a greater
density of point sources. Section C.6 presents the results of applying Method C to Houston
benzene data based on two alternative schemes for defining sampling strata. Step-by-step
guidelines for using Method C are presented in Section C.I.
C-2
-------
C.2 INITIAL EVALUATION OF SAMPLING METHODS A AND B
This section presents the objectives to be met by the sampling approach and provides two
alternative methods for accomplishing these objectives. The application of each method to
butadiene in Phoenix is described, and the results of these applications are compared.
C.2.1 Sampling Objectives
The sampling methods were developed to accomplish the following objectives:
• Develop a set of sampling strata (geographic zones) which completely covers the
designated study area. These strata may later be considered as potential exposure
districts for the HAPEM model.
• Estimate the minimum sample size required to estimate the mean outdoor
concentration (expressed as an annual average concentration) for each stratum to
within 10 percent of the true mean concentration with 95 percent confidence.
• Choose a sample which is representative of the outdoor concentrations at
population centers (as represented by the geographic centroids of census block
groups).
In accomplishing these objectives, researchers assumed that the sample representing each
combination of pollutant and study area would be selected independently.
C.2.2 Population of Interest
Researchers defined the study population as consisting of all census block groups (BGs)
in the designated study area according to the 1990 U.S. census. In the example considered here,
the study area consisted of all BGs in a rectangular region surrounding Phoenix, Arizona. The
universal transverse mercator (UTM) coordinates of the corners of this rectangle are listed below:
Corner UTM Zone UTM (east) UTM (north)
Northeast 12 485 km 3763 km
Southeast 12 485km 3663km
Southwest 12 349km 3663km
Northwest 12 349km 3763km
The study area contained approximately 13,600 square kilometers of land area and 1894 BGs.
The 1894 BGs were assumed to constitute the sampling frame for the study area (i.e., the total
population of available sampling units). The target population for this study is the current
population of Phoenix.
C-3
-------
C.2.3 Finite Population Survey Sampling versus Continuous Spatial Sampling
Although outdoor pollutant concentration is a continuous variable which is distributed
spatially over a region, researchers assumed that the HAPEM exposure assessment would be
concerned only with those areas in the region actually inhabited by people. Consequently,
researchers decided to choose a survey sample from a finite population of points (the geographic
centroids of the BGs) rather than using continuous-variable spatial sampling methods. In
essence, the goal of the selection process was a sample that represented the outdoor pollutant
concentrations "observed" by the population, rather than one representing the overall spatial
pattern of outdoor pollutant concentrations. The population was assumed to be clustered at
points located at the centroids of BGs.
In developing the sampling methods described below, researchers treated the outdoor
concentration determined by the dispersion model for a particular receptor as
an attribute of the population subgroup residing within the associated census unit.
C.2.4 The General Sampling Design
The proposed sampling design is a stratified random sample from a finite population.
According to this general approach, the study region is first divided into a number of smaller
geographic zones or strata. This procedure helps to account for the spatial variation of the data.
Next, a sample is taken from within each of the strata. Two methods can be used to obtain this
sample. The first method draws a simple random sample without replacement from within each
stratum. The second method stratifies each stratum further into 4 kilometer by 4 kilometer grid
cells, and then takes a random sample from within each grid cell. In each method, analysts used
computer-generated random numbers to draw random samples from the sampling frame. The
following subsections describe how researchers applied each method to the sample task
(butadiene in Phoenix) and provide a comparison of the results of each method.
C.2.5 Defining the Strata
EPA compiled an emission inventory for each combination of pollutant (benzene,
butadiene, formaldehyde, hexavalent chromium, or POM) and study area (Houston or Phoenix).
To assist future HAPEM exposure assessments, analysts will apply a version of the ISCST3
dispersion model to each emission inventory using local meteorological data for the associated
study area. Each run of the dispersion model will produce estimates of outdoor pollutant
concentrations at a set of receptor points defined by the user. As the spatial pattern in
concentration estimates is likely to be similar to the spatial pattern in the emissions data, it is
reasonable to assume that the area-source emissions data in each inventory will provide a
reasonable basis for defining the sampling strata. [Note that this assumption may not be valid for
point-source emissions data. In such cases, the approach discussed in Section C.5 (Sampling
Method C) is likely to produce superior results.]
In compiling the emissions database for Phoenix, researchers defined the emissions
inventory area by a rectangular grid containing 850 cells (34 cells by 25 cells). Each cell
C-4
-------
measured 4 km by 4 km. The grid included all of the Phoenix metropolitan area and some
adjoining rural areas. Using the emissions inventory data, estimates were developed for
emissions from various area-source types for each cell in the grid. (Point-source emission
estimates were not included in the database.) The data from the various source types within each
cell were summed to provide a total emissions value for the cell. The grid cell totals for
butadiene emissions ranged from 0 to 5097.2 kg/yr with a mean of 315.5 kg/yr and a standard
deviation of 854.9 kg/yr. Note that these data may be revised, as analysts are currently refining
the emission estimates for Phoenix.
Figure C.2-1 presents an emissions contour map of the Phoenix region indicating the cell
emission totals together with the 4 km by 4 km grid cells. The same map with the locations of
the block group centroids is provided in Figure C.2-2. From the latter map, it can be seen that a
very high proportion of the population is located in areas of relatively high butadiene emissions.
Because a large majority of the sampling units (i.e., the block group centroids) are located
in the center of the region, the study region was divided into two large subregions — the "central"
region and the "outer" region. Researchers noted that points in the outer region were too widely
spaced to attribute any type of "stratum mean" concentration to them. In a sense, each of these
points (or in some cases, small clusters of points) formed its own stratum. Consequently,
researchers recommended that every point in the outer region be included in the sample. These
points can be treated as separate exposure districts or aggregated into small clusters at a later
stage in the exposure assessment process.
In the central region, each 4 km x 4 km grid cell was assigned a classification
corresponding to the total emissions for that cell. The classification categories were high
(greater than 4000 kg/yr), moderate (between 1000 and 4000 kg/yr), or low (less than 1000
kg/yr) emissions. The "breakpoint" values of 1000 and 4000 were chosen somewhat
subjectively, but were considered reasonable as they were approximately equal to the 25th and
75th percentiles, respectively, of the distribution of emissions over all sampling units.
After classification, the grid cells of the central region were placed into sampling strata
according to two objectives:
(1) each stratum would contain cells with similar emission levels and
(2) each stratum would be a contiguous collection of cells with no stratum being
completely surrounded by another stratum.
The 12 resulting strata are shown in Figure C.2-3. Here, Strata 1 and 2 are "high" strata, Strata 3
through 7 are "moderate" strata, and Strata 8 through 12 are "low" strata. Figure C.2-4 shows the
distribution of sampling units across the 12 central strata.
C-5
-------
o
03
o
s
C5
O
•=• cvj
5 8.
03
§
(D
CO
s.
360
380
400 420
UTM (easting)
440
460
480
Figure C.2-1 Emissions contour map of the Phoenix study area showing cell emission contours
of butadiene and 4 km by 4 km emissions inventory grid cells.
C-6
-------
o
«o .
CO
o
2
co
I §
o
O
o
eo
o
oo
eo
1000
360
380
400 420
UTM (easting)
440
460
480
Figure C.2-2 Emissions contour map of the Phoenix study area showing cell emission contours
of butadiene and the locations of all block group centroids.
C-7
-------
O)
c
5
D
o
CO .
CO
0
CM .
co
o
CO
o
CO
o
O) .
o
«> .
CO
to •
m
1
J
I
2 3
,.,.., 1 1
380 400 420
UTM (easting)
440
Figure C.2-3 Strata constructed for application of sampling procedure to central region of
Phoenix.
C-8
-------
o
co
n
o
CM
CO
o
V™
N.
g
£
CO
CD
CO
CO
CD
CO
380
400 420
UTM (easting)
440
Figure C.2-4 Strata constructed for application of sampling procedure to central region of
Phoenix with superimposed centroids of all block groups.
C-9
-------
Table C.2-1 shows the distribution of the emissions values over the sampling units (i.e.,
the BGs) within each stratum. As expected from the definition of the classification categories,
the means for Strata 1 and 2 are between 4000 and 5000 kg/yr, the means for Strata 3 through 7
are all near 2000 kg/yr, and the means for Strata 8 through 12 are approximately 500 kg/yr. The
standard deviations for the high and low strata all fall between 165 and 430 kg/yr, while the
standard deviations for the moderate strata range from 715 to approximately 890 kg/yr.
Table C.2-2 lists the number of grid cells and the total population within each stratum.
The results in this table indicate that the 305 grid cells of the central region contain almost 98
percent of the study area population. The high and moderate strata (Strata 1 through 7) are also
the most densely populated areas. Almost 25 percent of the study area population are located in
the areas of highest butadiene emissions (Strata 1 and 2), and approximately 80 percent of the
population are located in areas of at least moderate (> 1000 kg/yr) butadiene emissions.
C.2.6 Sample Size Determination
In selecting a sample from a defined population, analysts typically begin the process by
estimating the sample size needed to estimate a particular population parameter with some
specified degree of precision. To meet the second objective (see Section C.2.1), researchers
determined that the mean of each sampling stratum in the central region should be estimated to
within 10 percent of the true mean with 95 percent confidence.
In calculating the sample size required to achieve a specified sampling objective, it is
often necessary to obtain a reasonable approximation of the expected precision of the estimate
being produced. This approximation could be an educated guess, an estimate obtained from prior
studies, or an estimate obtained using a surrogate variable which has similar distributional
characteristics to the variable under study. Because the pollutant concentrations to be estimated
by the dispersion model are a function of the emissions data provided to the dispersion model,
the concentration estimates are likely to be roughly proportional to the emissions data. Analysts
incorporated this assumption of proportionality into the sample size calculations discussed
below.
The emission values in each stratum have a mean (UQ) and a standard deviation (OQ). If
one can assume that the outdoor pollutant concentration, C, in each stratum is proportional to the
emissions for that stratum, Q, then C = kQ for some constant k. (In reality, k is not a constant
but is a function of meteorological parameters as well as the total emissions for the entire study
region). If C = kQ, then the mean concentration (uc) is equal to kuQ and the standard deviation
of the concentrations (oc) is equal to koQ. The coefficient of variation, y, is defined as the ratio
of the standard deviation to the mean. Thus, under the assumption of proportionality, the
coefficients of variation for the emissions and associated outdoor concentrations in each stratum
are equal,
C-10
-------
Table C.2-1 The Distribution of Emissions Values Across the Census Block Groups Within Each Stratum
Sampling
area
All BGs
BGsin
Outer
Region
BGsin
Central
Region
Stratum 1
Stratum 2
Stratum 3
Stratum 4
Stratum 5
Stratum 6
Number
of block
groups in
area
1894
33
1861
457
75
128
201
100
246
Distribution of emission values, kg/yr
Mean
2666.7
52.2
2713.1
4587.8
4233.0
2206.8
2706.3
1952.4
2221.6
Standard
deviation
1512.9
77.8
1485.3
428.9
300.3
773.6
887.7
861.6
715.7
Skew-
ness3
-0.02
1.72
-0.01
-0.37
0.64
0.69
-0.07
1.16
0.49
Kur-
tosisa
-1.15
2.92
-1.15
-1.16
-1.61
-0.45
-0.89
-0.22
0.08
Mini-
mum
0.82
0.82
19.94
3774.6
3991.1
1163.1
868.8
1141.1
711.0
25th
Percentile
1620.8
1.60
1644.1
4285.4
3991.1
1734.1
2229.3
1278.1
1768.1
Median
2552.8
4.75
2609.7
4675.6
4040.3
1967.0
2609.7
1644.1
2168.4
75th
Percen-
tile
3991.1
89.5
3991.1
5031.8
4641.6
2629.4
3434.8
1822.5
2453.4
Maxi-
mum
5097.2
319.6
5097.2
5097.2
4641.6
3650.0
3924.5
3551.5
3620.9
C-ll
-------
Table C.2-1 The Distribution of Emissions Values Across the Census Block Groups Within Each Stratum (continued)
Sampling
area
Stratum 7
Stratum 8
Stratum 9
Stratum 10
Stratum 1 1
Stratum 12
Number
of block
groups in
area
365
39
65
86
46
53
Distribution of emission values, kg/yr
Mean
2513.7
387.8
416.6
511.4
594.1
672.9
Standard
deviation
756.0
309.6
165.5
415.7
406.1
230.0
Skew-
ness3
-0.22
0.33
-1.00
1.71
0.36
-0.97
Kur-
tosis'1
-0.80
-1.46
-0.03
2.51
-0.35
0.23
Mini-
mum
535.2
34.6
19.9
31.0
34.5
28.5
25th
Percentile
1929.8
70.9
300.5
257.7
204.2
430.1
Median
2552.8
331.3
508.3
427.6
633.0
732.2
75th
Percen-
tile
3134.7
715.7
537.1
683.2
779.2
890.5
Maxi-
mum
3648.2
851.7
621.6
1633.3
1561.2
904.9
aDimensionless.
C-12
-------
Table C.2-2 Number of Grid Cells and Total Population for Each Stratum
Sampling area
Entire study area
Outer region
Central region
Stratum 1
Stratum 2
Stratum 3
Stratum 4
Stratum 5
Stratum 6
Stratum 7
Stratum 8
Stratum 9
Stratum 10
Stratum 1 1
Stratum 12
Number of
grid cells
850
545
305
13
3
12
14
6
12
21
69
24
43
48
40
Number of
block
groups
1894
33
1861
457
75
128
201
100
246
365
39
65
86
46
53
Number of
populated
grid cells
206
29
177
13
3
12
14
6
12
21
21
16
31
19
9
Total
Population
2,138,258
44,021
2,094,237
411,318
116,821
205,531
223,354
138,373
247,370
380,968
73,510
71,240
93,570
71,770
60,452
Percentage of
total population
100.0
2.1
97.9
19.2
5.5
9.6
10.4
6.5
11.6
17.8
3.4
3.3
4.4
3.4
2.8
C-13
-------
kac
Consequently, it is reasonable to use the emissions data as a surrogate for the concentration data
in estimating the sample size required to estimate the mean concentration in each stratum.
The general formula for calculating the sample size, n, for a simple random sample of
each stratum is
n = [(r/Za,2Yc)2 + (l/N)]-', (2)
where r is the relative precision desired, Za/2 is the 100(1 - a/2) percentile from the standard
normal distribution (1.96 for a = 0.05), YC is the unknown coefficient of variation for the
concentrations, and N is the total number of sampling units in the stratum.2 The goal of the
sample is a precision of +/- 10 percent with 95 percent confidence. Consequently, the value of r
is 0.10, the value of a is 0.05, and the corresponding value of Za/2 is 1.96. Practically, n should
always be an integer, and, to assure the proper coverage probability, should always be rounded
up. In addition, this sample size formula is based on the Central Limit Theorem, which generally
holds for samples of at least 30, regardless of the underlying distribution. Therefore, if any
calculated sample size was less than 30, a sample of size 30 was applied.
In the above sample size formula, YC wiH be estimated by YQ under the assumption of
proportionality. Note from the formula that if YQ substantially underestimates yc, the calculated
sample size will be less than required for the desired amount of precision. However, if YQ
overestimates jc, oversampling will result, which is generally not a concern.
The sample size formula does not consider potential spatial correlations between
population density and emission rates. The goal of the sample selection is to estimate the mean
concentration at a specified set of locations (the census block centroids) and not to estimate the
spatial trend in concentrations; consequently, it is not necessary to account for spatial
correlations in the calculation of sample size or variance.
Table C.2-3 contains the sample sizes for each stratum calculated using Equation 2. The
total sample size required is 470. The calculated sample sizes for Strata 1, 2, and 12 were all less
than 30. Therefore, to meet the requirements of the Central Limit Theorem, samples of size 30
were imposed on each these strata. All other sample size calculations resulted in values between
30 and 45, with the exception of Stratum 10 where a sample of 64 is required.
C-14
-------
Table C.2-3 Sample Sizes Selected for Each Stratum in the Central Region
Sampling area
Stratum 1
Stratum 2
Stratum 3
Stratum 4
Stratum 5
Stratum 6
Stratum 7
Stratum 8
Stratum 9
Stratum 10
Stratum 1 1
Stratum 12
Total for central
region
Total
number of
block
groups, N
457
75
128
201
100
246
365
39
65
86
46
53
1861
Coefficient of
variation, yQ
0.0935
0.0709
0.3506
0.3280
0.4413
0.3222
0.3008
0.7983
0.3973
0.8129
0.6836
0.3418
-
Required
sample size,
n
30a
30a
35
35
43
35
32
34
32
64
37
30a
437
Sizes of samples taken
Method A
30
30
35
35
43
35
32
34
32
64
37
30
437
Method B
30
30
35
35
43
36
34
36
33
72
38
30
452
"The calculated sample sizes for Strata 1, 2, and 12 were less than 30.
C-15
-------
Two methods for obtaining samples of the specified sizes are proposed. The following
two subsections describe these methods and present the results of applying each method to the
sample task.
C.2.7 Selection of Samples by Method A
In Method A, the analyst randomly selects the specified number of BGs from each
stratum without regard to location of grid cells within the stratum. An unbiased estimate for the
mean concentration at the block group centroids can be constructed for each stratum using the
stratum sample mean Y. The variance for this estimate is calculated by the equation
N-n °c (3)
Var(Y) =
N n
where o2c is the variance of the concentrations within the stratum.2
For the test application (butadiene in Phoenix), analysts used Method A to select a
stratified random sample of 470 BGs from the sampling frame of 1894 BGs. Of these 470 BGs,
437 were selected from the 12 strata in the central region as indicated in Table C.2-3. The
remaining 33 BGs in the sample include all of the BGs located in the outer region. As previously
discussed, analysts determined that the BGs in the outer region were too widely dispersed to be
aggregated into multi-BG strata.
Figure C.2-5 shows the locations of the centroids of the selected BGs. In the proposed
HAP EM analysis, these BGs can be used to construct as many as 45 "exposure districts," 12
corresponding to the 12 strata in the central region and 33 corresponding to the 33 individual
BGs in the outer region. (The number will be less than 45 if analysts decide to aggregate groups
of BGs in the outer region). The outdoor pollutant concentration in each district would be
determined by averaging the dispersion model estimates for centroids of the BGs assigned to
each district.
C.2.8 Selection of Samples by Method B
As indicated by Figure C.2-5, random clustering of data can occur when Method A is
used to draw the sample. There is no guarantee that a simple random sample will adequately
cover the study area. To improve coverage, the first-stage strata can be further stratified using
the 4 km by 4 km grid cells within each stratum. A random sample of block groups can then be
taken from each grid cell. The sample sizes determined above (listed as "required sample size"
in Table 2-3) can be divided among the grid cells using a proportional allocation. The number of
BGs sampled from grid cell h would be calculated as
nh = nNh/N, (4)
C-16
-------
o
o
CO
o
^J-
r-~
eo
CM
O
o
co
360
380
400 420
UTM (easting)
440
460
480
Figure C.2-5 Location of block groups selected by Method A.
C-17
-------
where n is the total sample size in the particular stratum, Nh is the total number of BGs in cell h,
and N is the total number of BGs in the stratum. This result is then rounded to the nearest
integer. To guarantee that all populated grid cells are included in the sample, however, at least
one BG must be sampled from each populated cell. Therefore, if nh as calculated above would
normally be rounded down to 0 (e.g. nh = 2(35) / 246 = 0.28), it is rounded up to 1. Note that this
rounding scheme may actually result in larger sample sizes being drawn from some strata than
those calculated for Method A. Unbiased estimates of the mean concentrations in each stratum
can be constructed using a weighted sample mean of the form
(5)
h=l
where L is the total number of grid cells in the stratum, Nh is the number of block group
centroids in cell h, y~h is the sample mean concentration in cell h, and N is the total number of
BGs in the stratum (Thompson, 1992). Note that this formula places more weight on the more
densely populated grid cells. The variance for this estimate is
Var(y) =
N.
h
N
o,
n.
(6)
where oh2 is the concentration variance for grid cell h. However, oh2 will not be _
estimable for cells from which only one sample is taken. Consequently, estimates of Var(y) may
be too low.
For the example application (1,3-butadiene in Phoenix), analysts used Method B to select
a stratified random sample of 485 BGs (Figure C.2-6). Of these, 452 were selected from strata in
the central region as shown in Table C.2-3. The remainder were selected by taking all 33 BGs in
the outer region. Consistent with the results of Method A, the BGs selected by Method B can be
assigned to 45 districts, 12 corresponding to the 12 strata in the central region and 33
corresponding to the individual BGs in the outer region. The outdoor pollutant concentration in
each district would be determined by using the above formula to compute a weighted average of
the dispersion model estimates for centroids of the BGs assigned to each district.
The locations of the BGs for one sample selected by Methods A and B are displayed in
Figures C.2-5 and C.2-6, respectively. Both samples show good coverage of the study region,
although some random clumping of data is present in both samples. Thus, analysts anticipate
that Method A will provide results very similar to Method B.
C-18
-------
o
(O
CO
I1-
CO
=
o
o
o
co
o
co
ID
CO
360
380
400 420 440
UTM (easting)
460
480
Figure C.2-6 Location of block groups selected by Method B.
C-19
-------
C.2.9 Decision to Perform Sensitivity Analyses
Based on these observations, researchers noted that Method A may be preferable to
Method B because (1) it is easier to apply, (2) it requires a smaller sample size, and (3) it
provides an unbiased estimate of the variance.
Section C.3 provides step-by-step procedures for implementing Methods A and B.
Before making a final endorsement of one of these candidate sampling methods, researchers
conducted a series of analyses to test the statistical assumptions underlying Methods A and B.
Section C.4 summarizes these analyses, identifies Method B as the superior method in terms of
statistical performance, and concludes that both methods tend to select samples which are larger
than necessary to achieve the specified set of sampling goals. Sections C.5 and C.6 present
Method C, a revised version of Method B, which is capable of achieving the sampling goals
using a significantly smaller sample.
C-20
-------
C.3 SUMMARY OF SAMPLING METHODS A AND B
This section summarizes the step-by-step procedure for drawing samples using Methods
A and B, the two methods proposed in the previous section. It also provides a summary of the
principal limitations associated with these methods.
C.3.1 Step-By-Step Procedure
1. Align the census population data and the emissions data for the particular
pollutant and city. For the Phoenix data, this step was implemented by first
converting the block group centroid coordinates from latitude and longitude to
UTM coordinates, and then converting the UTM coordinates to X and Y
coordinates corresponding to the emissions grid.
2. Define the sampling strata. This step requires both subjective and objective
decisions on the part of the sampler. For the Phoenix data, the study region was
first divided into two large subregions, the central region and the outer region,
based on the density of the population within each of these regions. (If the
populations in the other cities in this study are more uniformly spread over the
study region, this step may not be necessary). Next, contour maps of the
emissions data were drawn with isopleths separating areas of "high", "moderate",
and "low" total emissions. For Phoenix, analysts subjectively chose isopleth
levels of 1000 kg/yr and 4000 kg/yr for butadiene. Each of the emissions grid
cells was then categorized as "high", "moderate", or "low", corresponding to the
emissions level within that grid cell. Smaller contiguous strata were formed by
joining together neighboring grid cells of the same emissions category. These
strata were further broken into smaller strata to avoid the "doughnut effect", in
which one stratum is completely surrounded by another stratum. (The commuting
algorithms of the HAPEM model do not allow for "doughnut shaped" exposure
districts.) Additionally, extremely large strata were divided into smaller strata to
help maintain the spatial characteristics of the concentration values.
3. Determine the required sample sizes. Each block group centroid was assigned an
emissions value equivalent to the emissions value of the grid cell into which it
fell. The distribution of emissions values across the block groups was then
tabulated for each stratum. In particular, the coefficient of variation was
determined for each stratum. Under the assumption of proportionality between
the emissions and concentrations within each stratum, the sample size formula
from Section C.2 was then utilized to determine the minimum sample size
required to meet certain specifications. A minimum sample size requirement of
30 for each stratum was imposed to meet the large sample requirements necessary
for the Central Limit Theorem.
4. Select the sample using one of the two proposed methods. Under Sampling
Method A from Section C.2, simple random samples of block group centroids
C-21
-------
were chosen from within each stratum to act as receptor points. Under Sampling
Method B, each stratum was further stratified into grid cells, and a random sample
of block group centroids was chosen from within each populated grid cell in each
stratum. Simple random number generators were utilized to choose these
samples, without replacement.
This approach is applicable to each combination of study area and pollutant.
C.3.2 Limitations of Methods A and B
The sampling methods described above have the following limitations.
• The analyst must create a separate set of strata for each pollutant included in the
analysis. Consequently, a distinct sample of block groups must be drawn for each
pollutant.
• The method for defining the strata presented above will not be adequate when the
pollutant under consideration is dominated by point-source emissions rather then
area-source emissions.
• In determining sample size, emissions should not be used as a surrogate for the
concentrations when point sources dominate the emissions for a particular
pollutant.
These limitations will be addressed further in the development of a revised sampling method in
Section C.5.
C-22
-------
C.4 SENSITIVITY ANALYSES OF SAMPLING METHODS A AND B
Previous sections of this appendix describe two methods for choosing a sample which is
representative of the outdoor pollutant concentrations at population centroids. In both methods,
the study region is first divided into a set of sampling strata which completely covers the study
area. In sampling Method A, a simple random sample is taken from within each stratum for the
purposes of estimating the mean concentration of the stratum. In sampling Method B, each
stratum is further stratified into grid cells, and a weighted estimator is used to estimate the mean
concentration of the stratum. One sampling objective was to estimate the mean outdoor pollutant
concentration for each stratum to within 10 percent of the "true" mean concentration with 95
percent confidence. Note that here the "true" mean concentration would be the value obtained
from the dispersion model estimates at each centroid within the stratum.
In calculating the sample sizes required to meet the sampling objectives for a simple
random sample, analysts had to make two assumptions about the distribution of the
concentrations. First, it was assumed that the coefficient of variation for the emission values was
approximately the same as the coefficient of variation for the concentration values within a
particular stratum. This is the "assumption of proportionality" discussed in previous sections.
Second, it was assumed that the sample size for each stratum was "large enough" for the Central
Limit Theorem to hold; i.e., the sampling distribution of the sample mean should be
approximately normal.
Analysts conducted a series of sensitivity analyses with the following three general
objectives in mind:
• To check the validity of the sample size assumptions.
• To compare sampling Method A with sampling Method B with regards to the
accuracy and precision of the estimates.
• To identify deficiencies in the proposed sampling methods and develop
modifications to address these deficiencies.
It will be shown through the sensitivity analyses that the methods previously proposed
adequately meet the sampling objectives. However, modifications to these methods will be
proposed to increase the efficiency and reduce the number of assumptions required.
Researchers obtained dispersion model estimates of the butadiene concentrations at each
of the block group centroids in the Phoenix study region for 1991. Estimates were obtained for
annual average concentrations as well as for 96 hourly/seasonal average concentrations. The
sensitivity analyses will first focus on the annual average data. Subsequent analyses will then be
performed on specific combinations of hour and season. Since all current analyses are being
performed specifically on butadiene concentrations in the Phoenix study area, broad
generalizations of these results to all pollutants in all cities must be approached cautiously.
C-23
-------
C.4.1 Evaluation of Sample Sizes Calculated by Methods A and B
Table C.4-1 contains the true means and standard deviations of the annual average
concentrations, by strata, as well as a comparison of the sample size calculations using the
emissions values and the concentration values. The true mean concentrations for strata 1 and 2
are "high" relative to the other strata, the means for strata 3 through 7 are "moderate", and the
means for strata 8 through 12 are "low". Recall that strata 1 and 2 were defined in Section C.2 as
high emissions strata, strata 3 through 7 were defined as moderate emissions strata, and strata 8
through 12 were defined as low emissions strata. The strata mean concentrations range from
about 0.04 to almost 0.18 u/m3. Additionally, the standard deviations of the concentrations
within each stratum follow a similar pattern to that observed for the emission values in
Section C.2. That is, the "high" and "low" strata tend to vary less than the "moderate strata".
Equation 2 in Section C.2 was used to calculate the sample size required to meet the
sampling objectives for a simple random sample. Recall that Za/2 is 1.96 and r is 0.1 in this
formula. Also recall that the result, n, should always be rounded up to the next positive integer.
The coefficient of variation for the concentrations, yc> is shown in the fifth column of Table
C.4-1, and is simply defined as the standard deviation divided by the mean. The coefficient of
variation for the emissions values, YQ, which was used to approximate yc in tne sample size
calculations of Section C.2, is shown in column 7 of this table, and the ratio of these two values
is presented in the ninth column. The sample sizes calculated from Equation 2 using yc ar>d YQ
are shown in columns 6 and 8 of Table C.4-1, respectively. The sample sizes in column 8 are the
same as those presented in Table C.2-3, with the exception that sample sizes less than 30 were
not rounded up here. The ratio of these sample sizes is shown in column 9.
In every stratum, YQ exceeds YO which indicates that the required sample size was
overestimated in each stratum when using the distribution of the emissions values to approximate
the distribution of the concentration values. Sometimes, the use of YQ led to an estimated sample
size which was actually more than twice the sample size required to meet the objectives. As
discussed in Section C.2, oversampling is generally not a concern in estimation problems. On
the other hand, oversampling results in longer dispersion model runs. Therefore, it may be
beneficial to reduce the amount of oversampling. This point will be addressed further below.
The results presented in the last column of Table C.4-1 indicate that using the emissions
values to represent the concentration values in calculating sample sizes would have been
adequate, although somewhat inefficient, for the annual average 1,3-butadiene data. Analysts
used Monte Carlo sampling methods to compare sampling Methods A and B with regards to the
accuracy and precision of the estimated mean concentrations as well as to assess the validity of
the normality assumption required for the sample size formula. Using the sample sizes presented
in Table 2-3, analysts drew 200 samples of block group centroids using each method. The mean
concentration of each stratum was then estimated by the concentrations at the block group
centroids in each of the 200 samples. Recall that for Method A, the usual sample mean is used as
the estimator, while for Method B the weighted sample mean in Equation 5 is used. Here, each
estimator will simply be referred to as a sample mean.
C-24
-------
Table C.4-1 Comparison of Sample Size Calculation Results Using the Concentration Values
Versus Using the Emissions Values
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Total
number of
block
groups, N
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive statistics and sample size calculations for
concentration values
True
mean
(u/m3)
0.1786
0.1601
0.1033
0.1191
0.0895
0.1137
0.1175
0.0540
0.0496
0.0549
0.0434
0.0503
0.1211
Standard
deviation
0.0137
0.0066
0.0233
0.0261
0.0271
0.0228
0.0230
0.0212
0.0119
0.0160
0.0182
0.0166
0.0462
Coefficient of
variation, yc
0.0767
0.0415
0.2259
0.2190
0.3025
0.2008
0.1956
0.3928
0.2391
0.2916
0.4185
0.3293
-
Calculated
sample size,
nc
3
1
18
17
27
15
15
24
17
24
28
24
213
Sample size calculations for
emissions values (see Table 2-3)
Coefficient of
variation, YQ
0.0935
0.0709
0.3506
0.3280
0.4413
0.3222
0.3008
0.7983
0.3973
0.8129
0.6836
0.3418
-
Calculated
sample size,
"Q
4
2
35
35
43
35
32
34
32
64
37
25
378
Ratio of
coefficients of
variation, YC /
YQ
0.82
0.59
0.64
0.67
0.69
0.62
0.65
0.49
0.60
0.36
0.61
0.96
-
Ratio of
required
sample
sizes,
nc/nQ
0.75
0.50
0.51
0.49
0.63
0.43
0.47
0.71
0.53
0.38
0.76
0.96
0.56
C-25
-------
Each of the 200 Monte Carlo samples resulted in estimates of the true strata means. The
distribution of these estimates (for each stratum) across the 200 samples is known as the
sampling distribution. If the mean of the sampling distribution is equal to the true mean, then the
estimator is unbiased. In addition, the variance of the sampling distribution can be estimated
from these 200 samples. The square root of the variance, the standard error of the estimator,
gives one an idea of the "spread" of the sampling distribution around the true mean.
Table C.4-2 contains the results of the 200 Monte Carlo samples obtained by applying
sampling Methods A and B to the annual average butadiene data. Included in this table are the
estimated means of the sampling distribution and the estimated standard errors of the sampling
distribution for both sampling methods. Statistics for evaluating the attainment of the sampling
objective and for testing the normality assumption are also provided.
Comparing the true means in Table C.4-1 with the estimated means in Table C.4-2
reveals that both sampling methods produce unbiased estimates of the mean concentration, as
anticipated by analysts. However, the standard errors are consistently much smaller under
Method B than under Method A. Typically, estimators from stratified random samples (e.g.,
Method B) have lower variances than those from simple random samples (e.g., Method A).
Therefore, analysts anticipated that Method B would result in more precise estimates than
Method A. However, it was not anticipated that Method B would outperform Method A to the
degree observed here. As can be seen from Table C.4-2, the standard errors from Method A are
generally between 2 and 7 times higher than those from Method B.
The reason for the large differences in precision between the two types of estimators as
seen here is fairly obvious. The variance of an estimator from Method A is a function of the
overall stratum concentration variance, as shown in Equation 3 in Section C.2. The variance of
an estimator from Method B is a function of the variances within each of the grid cells within
each stratum, as shown in Equation 6. Therefore, the large differences in precision are mostly
attributable to the small within-cell concentration variances relative to the overall stratum
concentration variances.
Recall that one of the primary sampling objectives was to estimate the true mean
concentration to within 10 percent with 95 percent confidence. Table C.4-2 contains the
proportion of sample means from the 200 Monte Carlo samples which were within 10 percent of
the true strata concentrations for both sampling methods. Both methods easily met the sampling
objectives. For Method A, at least 198 of the 200 (99 percent) sample means from each stratum
were within 10 percent of the true mean. For Method B, every sample mean met the objective.
Equation 2 in Section C.2 is used for calculating the required sample sizes for simple
random samples. This equation requires that the distribution of the unweighted sample mean
from a simple random sample be at least approximately normal. To validate this assumption, the
p-values from Shapiro-Wilks tests of the normality of the sampling distributions for Method A
are included in Table C.4-2. In general, a p-value larger than 0.05 indicates that the normality of
C-26
-------
Table C.4-2 Results of 200 Monte Carlo Samples for Sampling Methods A and B Using Sample Sizes
as Calculated in Table C.2-3
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Sampling Method A
Sample
size drawn
from each
stratum
30
30
35
35
43
35
32
34
32
64
37
30
Estimated
mean of the
sampling
distribution
(u/m3)
0.1786
0.1601
0.1032
0.1191
0.0892
0.1134
0.1174
0.0540
0.0495
0.0550
0.0434
0.0505
Estimated
standard
error of the
sample
mean
0.0023
0.0010
0.0035
0.0039
0.0031
0.0033
0.0039
0.0013
0.0015
0.0010
0.0013
0.0020
Proportion of
sample means
within 10% of
the true mean
1.0
1.0
0.990
0.990
0.995
1.0
1.0
1.0
1.0
1.0
1.0
0.995
Sampling Method B
Sample
size drawn
from each
stratum
30
30
35
35
43
36
34
36
33
72
38
30
Estimated
mean of the
sampling
distribution
(u/m3)
0.1786
0.1602
0.1032
0.1191
0.0895
0.1136
0.1175
0.0540
0.0496
0.0549
0.0434
0.0503
Estimated
standard
error of the
sample
mean
0.0010
0.0006
0.0008
0.0010
0.0006
0.0010
0.0010
0.0004
0.0008
0.0003
0.0002
0.0005
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
P-value for
the Shapiro
- Wilks test
of normality
0.7733
0.7413
0.1547
0.9949
0.8108
0.6466
0.4202
0.2152
0.7461
0.8829
0.2394
0.5321
C-27
-------
sampling distribution cannot be rejected. None of the p-values in this table indicate that the
normality assumption should be questioned for samples of at least 30. Thus, the minimal sample
size of 30 which was imposed on all strata is adequate. However, it will be shown below that it
may be possible to reduce this minimal requirement.
C.4.2 Effects of Reducing Sample Sizes
Tables C.4-3 through C.4-8 contain similar statistics to those in Table C.4-2 for 200
Monte Carlo samples of nominal sizes 30, 25, 20, 15, 10, and 5, respectively. The sample sizes
are "nominal" due to the rounding scheme employed in sampling Method B. Recall from
Section C.2 that at least one block group centroid must be sampled from each populated grid cell
under this method. Therefore, while the nominal sample size is always the size of the sample
actually selected under Method A, samples selected under Method B will tend to be larger. In
fact, under Method B, the minimum sample size drawn from any particular stratum equals the
number of populated grid cells within that stratum. Recall that a "populated" grid cell was
defined as one which contains at least one block group centroid.
Four of the p-values from the Shapiro-Wilks test of normality presented in Tables C.4-3
through C.4-8 are significant at the 0.05 level (Table C.4-3, stratum 1; Table C.4-4, stratum 2;
Table C.4-6, stratum 4; and Table C.4-8, stratum 3). However, it is well known that the Shapiro-
Wilks statistic is overly conservative (i.e., it rejects the hypothesis of normality too often).
Therefore, in addition to performing these tests, normal probability plots and histograms were
examined for each Monte Carlo sample. The results of these tests, as well as the plots, indicate
that the sampling distribution is at least approximately normal in all strata for sample sizes as
small as 5. Therefore, these data indicate that the minimal sample size requirement of 30
imposed in Section C.2 could be lowered to 5. This result only effects the sample sizes for strata
1, 2, and 12, as the calculated sample sizes from the other strata were all greater than 30.
However, this would result in a total sample size requirement of 382 block group centroids being
drawn from the central region under Method A, rather than 437 as calculated in Section C.2.
This is a reduction of 55 block groups (13 percent) in the overall sample size.
Analysts also observed from Tables C.4-3 through C.4-8 that, although both sampling
methods result in unbiased estimates for the true mean concentration, Method B continues to
provide significantly more precise estimates than Method A, even when smaller sample sizes are
used. The standard error of the mean estimator from Method A is generally 2 to 7 times higher
than that of Method B for each stratum in all of these tables. Note that this comparison is not
entirely fair, as larger samples are usually drawn using Method B, and larger samples generally
lead to more precise estimates. However, the difference in standard errors is so great here that it
is reasonable to conclude that Method B results in a much more precise estimator than Method
A. As previously mentioned, the high precision of Method B is primarily a result of the low
within-cell concentration variances observed in this data.
C-28
-------
Table C.4-3 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a Sample of Nominal Size 30
Drawn from Each Stratum
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Sample
size drawn
from each
stratum
30
30
30
30
30
30
30
30
30
30
30
30
Sampling Method A
Estimated
mean of the
sampling
distribution
(H/m3)
0.1783
0.1601
0.1034
0.1193
0.0891
0.1136
0.1174
0.0539
0.0495
0.0549
0.0435
0.0503
Estimated
standard
error of the
sample
mean
0.0025
0.0010
0.0035
0.0045
0.0044
0.0039
0.0044
0.0019
0.0015
0.0023
0.0018
0.0020
Proportion of
sample means
within 10% of
the true mean
1.0
1.0
0.995
0.990
0.950
0.990
0.990
0.995
1.0
0.970
0.980
0.990
Sampling Method B
Sample
size drawn
from each
stratum
30
30
33
31
31
30
31
34
33
41
33
30
Estimated
mean of the
sampling
distribution
(u/m3)
0.1786
0.1602
0.1033
0.1191
0.0896
0.1137
0.1175
0.0539
0.0497
0.0549
0.0434
0.0503
Estimated
standard
error of the
sample
mean
0.0010
0.0006
0.0008
0.0009
0.0008
0.0011
0.0011
0.0005
0.0008
0.0007
0.0004
0.0005
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
P-value for
the Shapiro -
Wilks test of
normality
0.00 18a
0.2204
0.7251
0.8368
0.6738
0.8122
0.7118
0.5961
0.2951
0.4006
0.7102
0.5466
"The results from a normal probability plot and a histogram indicate that the sampling distribution is at least approximately normal.
C-29
-------
Table C.4-4 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a Sample of Nominal Size 25
Drawn from Each Stratum
Stratum
1
2
• 3
4
5
6
7
8
9
10
11
12
Sampling Method A
Sample
size drawn
from each
stratum
25
25
25
25
25
25
25
25
25
25
25
25
Estimated
mean of the
sampling
distribution
(u/m3)
0.1784
0.1602
0.1033
0.1182
0.0895
0.1137
0.1180
0.0541
0.0495
0.0551
0.0434
0.0502
Estimated
standard
error of the
sample
mean
0.0026
0.0011
0.0042
0.0053
0.0046
0.0040
0.0044
0.0026
0.0018
0.0028
0.0023
0.0027
Proportion of
sample means
within 10% of
the true mean
1.0
1.0
0.990
0.975
0.930
0.985
0.990
0.960
0.985
0.950
0.945
0.935
Sampling Method B
Sample
size drawn
from each
stratum
25
25
26
25
25
26
28
29
29
38
29
25
Estimated
mean of the
sampling
distribution
(u/m')
0.1786
0.1600
0.1032
0.1192
0.0894
0.1136
0.1176
0.0540
0.0496
0.0548
0.0434
0.0503
Estimated
standard
error of the
sample
mean
0.0011
0.0007
0.0011
0.0012
0.0009
0.0012
0.0011
0.0006
0.0009
0.0007
0.0005
0.0006
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
P-value for
the Shapiro -
Wilks test of
normality
0.9279
0.0408a
0.3877
0.6035
0.1756
0.7960
0.5219
0.3657
0.8800
0.5963
0.3380
0.9830
I'he results from a normal probability plot and a histogram indicate that the sampling distribution is at least approximately normal.
C-30
-------
Table C.4-5 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a Sample of Nominal Size 20
Drawn from Each Stratum
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Sampling Method A
Sample
size drawn
from each
stratum
20
20
20
20
20
20
20
20
20
20
20
20
Estimated
mean of the
sampling
distribution
(u/m3)
0.1786
0.1600
0.1035
0.1190
0.0896
0.1136
0.1174
0.0541
0.0496
0.0549
0.0432
0.0505
Estimated
standard
error of the
sample
mean
0.0029
0.0012
0.0046
0.0058
0.0056
0.0049
0.0050
0.0035
0.0022
0.0029
0.0030
0.0030
Proportion of
sample means
within 10% of
the true mean
1.0
1.0
0.985
0.970
0.880
0.980
0.975
0.855
0.975
0.930
0.945
0.935
Sampling Method B
Sample
size drawn
from each
stratum
20
20
20
21
20
21
23
27
24
36
24
22
Estimated
mean of the
sampling
distribution
(u/m')
0.1783
0.1600
0.1031
0.1189
0.0897
0.1138
0.1175
0 0539
0.0497
0.0548
0.0434
0.0503
Estimated
standard
error of the
sample
mean
0.0012
0.0009
0.0011
0.0014
0.0012
0.0015
0.0015
0.0007
0.0012
0.0008
0.0007
0.0006
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
P-value for
the Shapiro -
Wilks test of
normality
0.0686
0.0654
0.3180
0.1508
0.1115
0.5078
0.2064
0.1491
0.4034
0.1935
0.9480
0.5187
C-31
-------
Table C.4-6 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a Sample of Nominal Size 15
Drawn from Each Stratum
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Sampling Method A
Sample
size drawn
from each
stratum
15
15
15
15
15
15
15
15
15
15
15
15
Estimated
mean of the
sampling
distribution
(u/m3)
0.1782
0.1601
0.1037
0.1189
0.0892
0.1148
0.1179
0.0538
0.0497
0.0551
0.0431
0.0505
Estimated
standard
error of the
sample
mean
0.0036
0.0016
0.0061
0.0066
0.0064
0.0057
0.0062
0.0042
0.0030
0.0034
0.0041
0.0036
Proportion of
sample means
within 10% of
the true mean
1.0
1.0
0.940
0.955
0.830
0.935
0.960
0.800
0.885
0.890
0.695
0.840
Sampling Method B
Sample
size drawn
from each
stratum
15
15
16
16
15
18
21
25
21
33
21
17
Estimated
mean of the
sampling
distribution
(u/m3)
0.1785
0.1601
0.1034
0.1192
0.0895
0.1136
0.1176
0.0539
0.0495
0.0549
0.0433
0.0504
Estimated
standard
error of the
sample
mean
0.0016
0.0010
0.0013
0.0018
0.0014
0.0016
0.0017
0.0009
0.0013
0.0009
0.0008
0.0009
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
P-value for
the Shapiro -
Wilks test of
normality
0.0689
0.7642
0.8156
0.0032a
0.6856
0.3086
0.5515
0.0971
0.4799
0.6217
0.1890
0.4908
''The results from a normal probability plot and a histogram indicate that the sampling distribution is at least approximately normal.
C-32
-------
Table C.4-7 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a Sample of Nominal Size 10
Drawn from Each Stratum
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Sample
size drawn
from each
stratum
10
10
10
10
10
10
10
10
10
10
10
10
Sampling Method A
Estimated
mean of the
sampling
distribution
(u/m3)
0.1790
0.1601
0.1039
0.1195
0.0907
0.1145
0.1175
0.0538
0.0494
0.0549
0.0433
0.0501
Estimated
standard
error of the
sample
mean
0.0044
0.0020
0.0070
0.0076
0.0079
0.0072
0.0076
0.0055
0.0032
0.0047
0.0054
0.0050
Proportion of
sample means
within 10% of
the true mean
1.0
1.0
0.845
0.895
0.705
0.870
0.885
0.650
0.910
0.760
0.605
0.665
Sampling Method B
Sample
size drawn
from each
stratum
13
10
13
14
11
12
21
22
17
31
20
12
Estimated
mean of the
sampling
distribution
(u/m')
0.1785
0.1603
0.1033
0.1190
0.0897
0.1138
0.1174
0.0541
0.0496
0.0549
0.0435
0.0501
Estimated
standard
error of the
sample
mean
0.0016
0.0013
0.0016
0.0017
0.0014
0.0022
0.0015
0.0014
0.0018
0.0012
0.0009
0.0011
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
0.995
1.0
1.0
1.0
P-value for
the Shapiro -
Wilks test of
normality
0.6434
0.7803
0.2148
0.7887
0.1144
0.3483
0.4185
0.6295
0.6832
0.3455
0.6603
0.7422
C-33
-------
Table C.4-8 Results of 200 Monte Carlo Samples for Sampling Methods A and B with a Sample of Nominal Size 5
Drawn from Each Stratum
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Sampling Method A
Sample
size drawn
from each
stratum
5
5
5
5
5
5
5
5
5
5
5
5
Estimated
mean of the
sampling
distribution,
(u/m1)
0.1790
0.1603
0.1042
0.1187
0.0904
0.1136
0.1176
0.0541
0.0503
0.0549
0.0428
0.0501
Estimated
standard
error of the
sample
mean
0.0062
0.0029
0.0106
0.0125
0.0115
0.0094
0.0110
0.0085
0.0051
0.0073
0.0072
0.0077
Proportion of
sample means
within 10% of
the true mean
0.995
1.0
0.665
0.630
0.560
0.790
0.690
0.460
0.660
0.555
0.435
0.515
Sampling Method B
Sample
size drawn
from each
stratum
13
6
12
14
6
12
21
21
16
31
19
9
Estimated
mean of the
sampling
distribution,
(H/m3)
0.1786
0.1602
0.1033
0.1189
0.0895
0.1135
0.1174
0.0539
0.0495
0.0549
0.0435
0.0503
Estimated
standard
error of the
sample
mean
0.0016
0.0017
0.0016
0.0017
0.0023
0.0021
0.0014
0.0014
0.0021
0.0011
0.0011
0.0016
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
0.995
1.0
1.0
0.995
P-value for
the Shapiro -
Wilks test of
normality
0.2042
0.4813
0.0080a
0.2860
0.8763
0.5681
0.0643
0.0541
0.5721
0.3890
0.1403
0.4410
The results from a normal probability plot and a histogram indicate (hat the sampling distribution is at least approximately normal.
C-34
-------
Tables C.4-3 through C.4-8 also indicate the proportion of sample means from the 200
Monte Carlo samples which actually fall within 10 percent of the true concentration mean for
each stratum. In Table C.4-3, at least 95 percent of the sample means are within 10 percent of
the true mean under Method A. Under Method B, every sample mean fell within 10 percent of
the true mean. However, as the sample size drawn from each stratum decreases in the
subsequent tables, the proportion of sample means from sampling Method A which meet the
sampling objective decreases. When the sample sizes fall below those presented in Table C.4-1,
the proportion of sample means within 10 percent of the true mean generally falls below 0.95 for
Method A. Under Method B, this proportion is at least 0.995 in all cases.
Although the results of this section show that both methods described in Section C.2
would have been adequate for the annual average butadiene data, it is clear that Method B is
superior to Method A. Method C, a revised version of Method B which significantly improves
the sampling process, will be presented in the next section.
C-35
-------
C.5 REFINEMENTS TO SAMPLING METHODOLOGY BASED ON FINDINGS OF
SENSITIVITY ANALYSES
As analysts observed in the previous section, the sampling methods of Section C.2 of this
appendix produced adequate results for the annual average 1,3-tmtadiene data. However, as was
noted in Section C.5, Method B produces significantly more precise estimates than Method A.
Additionally, the methods used for computing required sample sizes tend to cause significant
oversampling, resulting in less efficient dispersion model runs. Refinements to the above
sampling methodology are proposed in this section which address both of these issues, resulting
in a more general sampling method.
The results of the previous section demonstrate that Method B produces significantly
more precise estimates than Method A, but that both methods produce unbiased estimators. In
this sense, Method B is a better sampling method than Method A. Therefore, the focus in this
section will be on revising sampling Method B exclusively.
C.5.1 Proposed Refinements
Recall that Table C.4-8 contains the results from 200 Monte Carlo samples of "nominal"
size 5 from each stratum. However, with the exception of stratum 2, each of the strata contains at
least 5 populated grid cells. Therefore, due to the rounding scheme employed in Method B,
taking a sample of "nominal" size 5 results in choosing one centroid from each populated grid
cell within each stratum for the Phoenix study area. In stratum 2, which contains only 3
populated grid cells, two centroids were chosen from each grid cell.
The results presented in Table C.4-8 suggest that taking one centroid from each grid cell
would produce unbiased and very precise estimators. However, it would not be possible using
this method to estimate the within-cell variances. Consequently, it would not be possible to
estimate the variance of the weighted sample mean (see Equation 6, Section C.2). Without an
estimate of the variance of the estimator, researchers would have no indication of the precision of
their estimator.
Analysts suggest the following modifications to Sampling Method B. Randomly choose
two block group centroids from each of the populated grid cells within each stratum. If there is
only one centroid in a grid cell, it is included in the sample automatically. The weighted sample
mean in Equation 5, Section C.2, can be used as before to estimate the mean concentration within
each stratum. With two centroids being selected from each grid cell, it is also possible to
estimate the variance of the weighted sample mean. Researchers can simply substitute the
within-grid-cell sample variances for the true variances in Equation 6 to obtain an unbiased
estimate.
This revised version of Method B (hereafter referred to as "Method C") has several
advantages over the methods suggested in Section C.2. First, an unbiased estimate of the
variance of the sample mean can be formed, which was not the case for the original Method B.
Second, analysts are required to make fewer assumptions about the distribution of the pollutant
C-36
-------
concentrations. Since a fixed number of centroids are being selected from within each grid cell,
it is not necessary to calculate the required sample sizes. Thus, the following assumptions that
were used in calculating sample sizes are no longer required: (1) the proportionality between the
emission values and the concentration values and (2) the normality of the sampling distribution.
Another, and perhaps the most important, advantage of Method C is that the sample size
is not pollutant dependant. That is, regardless of the pollutant being analyzed, researchers could
use the same number of receptor points for the dispersion model. In fact, it may be possible to
use the same receptor points for each pollutant, which would be beneficial to the researchers
performing the dispersion model runs.
The original strata served to divide the study area into subregions, or "exposure districts".
In addition, the strata played a large role in the sampling methods described in Section C.2 of this
appendix. However, under Method C, the original strata themselves play no role in the actual
sampling. Researchers can perform the dispersion model runs at the chosen receptor points
before the study region is divided into the smaller exposure districts. In fact, the concentration
values themselves could be used to create the exposure districts, rather than using the emissions
values as a surrogate.
C.5.2 Revised Methodology Applied to the Annual Average Data
Analysts drew 200 Monte Carlo samples from each stratum using the revised sampling
method (Method C) proposed above. Each of the 200 Monte Carlo samples resulted in estimates
of the true strata means. The distribution of these estimates (for each stratum) across the 200
samples is known as the sampling distribution. In addition, the variance of the sampling
distribution can be estimated from these 200 samples. The square root of the variance, the
standard error of the estimator, gives one an indication of the "spread" of the sampling
distribution around the true mean.
Table C.5-1 contains the results of the 200 Monte Carlo samples for the annual average
butadiene data. For strata 1 through 7, the sample sizes are equal to twice the number of
populated grid cells within the strata. That is, all of the populated grid cells within the most
heavily populated strata contain at least 2 block group centroids. In each of the more sparsely
populated strata, strata 8 through 12, there is at least one grid cell which contains only one block
group centroid. Thus, the sample sizes taken from these strata are all less than twice the number
of populated grid cells. The total required sample size from the central region is 312 centroids, a
reduction of 125 from the 437 centroids suggested in Section C.2. Including the 33 centroids in
the outer region, the total number of receptor points for the dispersion model run in the Phoenix
study area would then be 345.
C-37
-------
Table C.5-1 Results of 200 Monte Carlo Samples for Sampling Method Ca
When Applied to the Annual Average Butadiene Data
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Sample size drawn
from each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated mean of
the sampling
distribution
(u/m3)
0.1786
0.1602
0.1034
0.1190
0.0896
0.1136
0.1174
0.0539
0.0496
0.0549
0.0433
0.0501
-
Estimated standard
error of tlbe sample
mean
0.0010
0.0016
0.0011
0.0011
0.0036
0.0(615
O.OD11
O.OO08
0.0014
0.0007
O.W07
O.CM309
-
Proportion of
sample means
within 10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
-
Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-38
-------
The results in Table C.5-1 indicate that this method provides unbiased estimates of the
strata mean concentrations, as expected. A comparison of the standard errors in Table C.5-1 with
those for Method A in Table C.4-2 reveals that Method C provides, in general, more precise
estimators than Method A, even with a large reduction in sample size. Every sample mean from
the 200 Monte Carlo samples was within 10 percent of the true stratum mean.
C.5.3 Application of Method C to Data from Specific Hourly/Seasonal Combinations
In addition to the annual average concentrations, dispersion model estimates were
obtained for the average concentration at each hour of the day during each season of the year. In
all, there are 96 such hourly/seasonal combinations (24 hours x 4 seasons). Some of these
combinations were found to have concentration estimates which varied to a much greater extent
than the annual average concentrations. Other hourly/seasonal combinations had concentration
estimates which varied little. In order to assess the generalizability of the revised sampling
method, analysts applied the method to several hourly/seasonal combinations representing
concentration estimates with differing levels of variance.
Various population activities and meteorological patterns throughout the day have effects
on the levels of pollutant concentrations and the degree to which they vary. For this reason,
researchers chose to test Method C on a time during rush hour, 7 a.m., a time at mid-day, 12
p.m., and a time at night, 12 a.m. Researchers found that the rush hour concentrations tended to
be higher than average and vary greatly, the mid-day concentrations tended to be very similar to
the annual average data, and the night time concentrations tended to be low and vary little.
Researchers also found that the variability of the concentration estimates was similar in the
winter and fall seasons as well as in the spring and summer seasons. Method C was applied to
the 6 combinations of two seasons (winter and summer) and three times of day (7 a.m., 12 p.m.,
and 12 a.m.).
Descriptive statistics and results from 200 Monte Carlo samples for each of the 6
hourly/seasonal combinations are presented in Tables C.5-2 through C.5-7. Analysts noted the
following points concerning the descriptive statistics in these six tables:
• The winter concentrations tended to be substantially higher than the
corresponding summer concentrations at 7 a.m. and 12 p.m. At 12 a.m., the
results from the two seasons were relatively similar.
• The 7 a.m. mean concentrations were higher than the other two hours in both
seasons. This pattern is probably due to rush hour traffic. The winter mean
concentrations at 7 a.m. ranged from 0.1021 u/m3 in stratum 8 to 0.5516 u/m3 in
stratum 1. The summer mean concentrations at 7 a.m. ranged from 0.0502 u/m3
in stratum 11 to 0.2417 u/m3 in stratum 1.
C-39
-------
Table C.5-2 Descriptive Statistics for the Winter, 7 a.m., Concentration Values and Results of 200 Monte Carlo Samples
Using Sampling Method C"
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Number of block
group centroids
within stratum
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive Statistics
Mean
concentration
(u/m3)
0.5516
0.4681
0.2649
0.3930
0.3015
0.4971
0.3328
0.1021
0.2158
0.2671
0.1032
0.2079
0.3863
Concentration
standard
deviation
0.0772
0.0260
0.0714
0.0764
0.0898
0.0814
0.0859
0.0724
0.0775
0.1042
0.0536
0.0767
0.1580
Coefficient
of variation
0.1400
0.0556
0.2695
0.1943
0.2978
0.1637
0.2580
0.7094
0.3590
0.3901
0.5188
0.3689
-
Sampling Results (Method C)
Sample size
drawn from
each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated
mean of the
sampling
distribution
(u/m3)
0.5513
0.4683
0.2654
0.3927
0.3011
0.4972
0.3326
0.1020
0.2159
0.2668
0.1033
0.2079
-
Estimated
standard error
of the sample
mean
0.0052
0.0073
0.0046
0.0045
0.0060
0.0052
0.0044
0.0030
0.0066
0.0039
0.0022
0.0063
-
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
0.995
-
•"Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-40
-------
Table C.5-3 Descriptive Statistics for the Winter, 12 p.m., Concentration Values and Results of 200 Monte Carlo Samples
Using the Sampling Method Ca
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Number of block
group centroids
within stratum
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive Statistics
Mean
concentration
(u/mj)
0.1585
0.1316
0.0763
0.1136
0.0670
0.0977
0.0922
0.0351
0.0409
0.0600
0.0356
0.0441
0.1019
Concentration
standard
deviation
0.0119
0.0063
0.0185
0.0285
0.0219
0.0203
0.0238
0.0142
0.0069
0.0125
0.0122
0.0133
0.0440
Coefficient
of variation
0.0751
0.0476
0.2427
0.2508
0.3273
0.2078
0.2577
0.4026
0.1700
0.2083
0.3430
0.3013
-
Sampling Results (Method C)
Sample size
drawn from
each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated
mean of the
sampling
distribution
(u/m3)
0.1585
0.1316
0.0763
0.1136
0.0669
0.0978
0.0923
0.0351
0.0409
0.0600
0.0356
0.0440
-
Estimated
standard error
of the sample
mean
0.0008
0.0014
0.0008
0.0010
.00011
.00013
0.0008
0.0006
0.0009
0.0005
0.0003
0.0009
-
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
-
"Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-41
-------
Table C.5-4 Descriptive Statistics for the Winter, 12 a.m., Concentration Values and Results of 200 Monte Carlo Samples
Using the Sampling Method Ca
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Number of block
group centroids
within stratum
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive Statistics
Mean
concentration
(u/m3)
0.0068
0.0047
0.0034
0.0061
0.0042
0.0049
0.0048
0.0030
0.0034
0.0042
0.0033
0.0038
0.0051
Concentration
standard
deviation
0.0006
0.0003
0.0006
0.0030
0.0005
0.0006
0.0008
0.0008
0.0003
0.0010
0.0014
0.0004
0.0017
Coefficient
of variation
0.0914
0.0698
0.1886
0.5018
0.1177
0.1179
0.1671
0.2527
0.0978
0.2303
0.4223
0.1021
-
Sampling Results (Method C)
Sample size
drawn from
each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated
mean of the
sampling
distribution
(u/m3)
0.0067
0.0047
0.0034
0.0061
0.0042
0.0050
0.0047
0.0030
0.0034
0.0042
0.0033
0.0038
-
Estimated
standard error
of the sample
mean
0.0001
0.0001
0.0001
0.0006
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
-
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
0.920
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
-
•'Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-42
-------
Table C.5-5 Descriptive Statistics for the Summer, 7 a.m., Concentration Values and Results of 200 Monte Carlo Samples
Using the Sampling Method Ca
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Number of block
group centroids
within stratum
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive Statistics
Mean
concentration
(u/m')
0.2417
0.2312
0.1324
0.1787
0.1371
0.1958
0.1421
0.0507
0.0672
0.1243
0.0502
0.0955
0.1686
Concentration
standard
deviation
0.0311
0.0078
0 033 1
0.0324
0.0417
0.0387
0.0340
0.0285
0.0234
0.0311
0.0245
0.0327
0.0617
Coefficient
of variation
0.1288
0.0370
0.2498
0.1815
0.3045
0.1977
0.2392
0.5626
0.3485
0.2505
0.4884
0.3422
-
Sampling Results (Method C)
Sample size
drawn from
each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated
mean of the
sampling
distribution
(H/m3)
0.2416
0.2313
0.1325
0.1786
0.1368
0.1959
0.1421
0.0507
0.0673
0.1242
0.0503
0.0955
-
Estimated
standard error
of the sample
mean
0.0016
0.0027
0.0015
0.0017
0.0025
0.0018
0.0016
0.0010
0.0022
0.0013
0.0011
0.0022
-
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
-
"Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-43
-------
Table C.5-6 Descriptive Statistics for the Summer, 12 p.m., Concentration Values and Results of 200 Monte Carlo Samples
Using Sampling Method Ca
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Number of block
group centroids
within stratum
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive Statistics
Mean
Concentration
(u/m3)
0.1280
0.1058
0.0596
0.0823
0.0495
0.0632
0.0772
0.0264
0.0282
0.0229
0.0250
0.0249
0.0776
Concentration
standard
deviation
0.0119
0.0081
0.0175
0.0238
0.0205
0.0162
0.0220
0.0107
0.0067
0.0092
0.0093
0.0074
0.0389
Coefficient
of variation
0.0926
0.0763
0.2940
0.2888
0.4143
0.2570
0.2853
0.4037
0.2384
0.4013
0.3734
0.2958
-
Sampling Results (Method C)
Sample size
drawn from
each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated
mean of the
sampling
distribution
(u/m3)
0.1280
0.1058
0.0596
0.0823
0.0494
0.0633
0.0773
0.0264
0.0283
0.0229
0.0250
0.0248
-
Estimated
standard error
of the sample
mean
0.0008
0.0011
0.0006
0.0008
0.0009
0.0009
0.0007
0.0005
0.0007
0.0004
0.0002
0.0004
-
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
-
"Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-44
-------
Table C.5-7 Descriptive Statistics for the Summer, 12 a.m., Concentration Values and Results of 200 Monte Carlo Samples
Using Sampling Method Ca
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
Total
Number of block
group centroids
within stratum
457
75
128
201
100
246
365
39
65
86
46
53
1861
Descriptive Statistics
Mean
concentration
(u/m<)
0.0046
0.0062
0.0054
0.0068
0.0059
0.0032
0.0035
0.0040
0.0023
0.0032
0.0045
0.0059
0.0045
Concentration
standard
deviation
0.0010
0.0004
0.0007
0.0037
0.0010
0.0003
0.0006
0.0006
0.0005
0.0012
0.0012
0.0009
0.0019
Coefficient
of variation
0.2157
0.0671
0.1396
0.5459
0.1644
0.0834
0.1714
0.1550
0.2127
0.3828
0.2697
0.1503
-
Sampling Results (Method C)
Sample size
drawn from
each stratum
26
6
24
28
12
24
42
29
26
50
30
15
312
Estimated
mean of the
sampling
distribution
(H/m3)
0.0046
0.0062
0.0054
0.0069
0.0059
0.0032
0.0035
0.0040
0.0023
0.0032
0.0045
0.0059
-
Estimated
standard error
of the sample
mean
0.0001
0.0002
0.0001
0.0009
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
-
Proportion of
sample
means within
10% of the
true mean
1.0
1.0
1.0
0.785
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
-
Sampling Method C consists of drawing two BG centroids at random from each populated grid cell.
C-45
-------
• The 12 p.m. winter concentrations ranged from 0.0351 u/m3 in stratum 8 to
0.1585 u/m3 in stratum 1. The 12 p.m. summer concentrations ranged from
0.0229 u/m3 in stratum 10 to 0.1280 u/m3 in stratum 1. The 12 a.m. winter
concentrations ranged from 0.0030 u/m3 in stratum 8 to 0.0068 u/m3 in stratum 1.
The 12 a.m. summer concentrations ranged from 0.0023 u/m3 in stratum 9 to
0.0068 u/m3 in stratum 4.
• The concentration standard deviations were significantly higher for both seasons
at 7 a.m. than at the other two hours. The winter values varied substantially more
than the summer values at 7 a.m. At the other two hours, the concentrations from
the two seasons displayed about the same degree of variation.
• At 7 a.m. and 12 p.m., the concentration means for both seasons followed a
similar pattern as the annual average data with regards to the high, moderate, and
low strata. However, at 12 a.m., the mean concentrations were very nearly
uniform over the study area.
• The concentration values from stratum 4 varied considerably more than those of
the other strata at 12 a.m. This important finding will be discussed further below.
Coefficients of variation are presented in Tables C.5-2 through C.5-7 merely for
completeness. They actually play no role in the revised sampling method. However, comparison
of these coefficients of variation with those for the emissions values in Table C.4-1 reveals that
sampling Method A would not meet the sampling objective of estimating the mean to within 10
percent with 95 percent confidence in several instances. For example, the coefficient of variation
in stratum 4 for the winter, 12 a.m., data is 0.5018, which is much higher than the 0.3280 from
the emissions values. This difference would result in the under-estimation of the required sample
size by almost half. In most other cases, the coefficient of variation for the emissions is still
higher than that of the concentrations, which would result in oversampling, as before.
The results of 200 Monte Carlo samples using Method C are also presented in Tables
C.5-2 through C.5-7. Each of the 200 samples included the same block group centroids for all
combinations of hour and season. Because the estimated mean of the sampling distribution is
approximately equal to the true mean in all cases, analysts determined that the method produces
unbiased results.
In almost every case, all, or nearly all, of the 200 sample means are within 10 percent of
the true mean concentration. The only two exceptions both occur in stratum 4 at 12 a.m. For the
winter data, 191 of 200 (92 percent) sample means met the 10 percent objective, while for the
summer data, only 157 out of 200 (78.5 percent) sample means met the 10 percent objective. As
discussed above, the standard deviation of the concentrations in stratum 4 at 12 a.m. was
considerably higher than those of all other strata for both seasons. This relatively large variance
was apparently caused by the variance within one particular grid cell (UTM coordinates 397
Easting, 3700 Northing), perhaps due to a point source within this grid cell. For the winter,
concentrations ranged from 0.0064 to 0.0395 u/m3, while the summer concentrations ranged
C-46
-------
from 0.0056 to 0.0430 ^/m3 within this grid cell. The variances in the other grid cells of stratum
4 were not considerably different from each other. Because the concentrations are generally low
at 12 a.m., and because the population tends to be indoors at this hour, this particular decrease in
precision may not significantly affect population exposure estimates based on these estimates of
ambient air quality.
Method C has been shown to be superior to the other sampling methods proposed in this
report in several respects:
• The assumptions proposed in Section C.2 are not required for this method.
• The "exposure districts" can be defined after the dispersion model runs have been
completed.
• The method produces unbiased estimates with higher precision based on smaller
sample sizes. This results in faster dispersion model runs.
Although the proposed method worked very well for 1,3-butadiene in the Phoenix study
area, it is possible that the corresponding estimates for other pollutants of interest may not
exhibit the same degree of statistical precision. By estimating the variance of the estimator,
researchers can obtain information regarding the precision of the estimator. One quick way to
check that the sampling objectives are being met is to form an interval by taking the estimator
plus and minus two standard errors (the square root of the variance). If the width of this interval
is less than 20 percent of the estimate, then researchers can be somewhat assured that the
objectives are being met. If the precision requirements are not being met, larger sample sizes
may be necessary.
In evaluating the results of applying Method C to 1,3-butadiene in the Phoenix study
area, analysts noted that the butadiene emissions in Phoenix were dominated by area sources. To
determine whether Method C would perform well for a pollutant emitted largely by point
sources, the method was subsequently applied to benzene emissions in Houston, an area where
almost half of the total benzene emissions derive from point sources. Section C.6 summarizes
the results of this application.
C-47
-------
C.6 APPLICATION OF SAMPLING METHOD C TO HOUSTON BENZENE
CONCENTRATIONS
Section C.5 of this appendix describes sampling Method C, a stratified random sampling
method for obtaining receptor points for use in the ISCST3 dispersion model. The results
presented in Section C.5 demonstrate that this method works adequately for estimating the mean
butadiene concentrations at the census block group centroids in a number of pre-defined
exposure districts in Phoenix. However, the butadiene emissions in Phoenix were dominated by
area sources, and the question remains as to how well Method C would work for a pollutant
which is produced largely by point sources. In this section, the quality of the estimates obtained
from sampling Method C is examined for Houston benzene concentrations, where almost half of
the total emissions derive from point sources.
Section C.2 of this appendix describes a method for defining sampling strata to be used as
exposure districts. However, the results of the analyses described in Section C.5 suggest that the
exposure districts can actually be defined after the sample has been drawn and the dispersion
model runs have been completed. This approach would enable researchers performing
population exposure assessments to define the exposure districts to best suit their needs.
Analysts "test" this claim in this section using two sets of sampling strata; the first set will be
defined using the emissions values, as in Section C.2, and the second set will be defined as
exposure districts used in the HAPEM model might be defined.
As in Section C.5, the sampling methods will be applied to the annual average benzene
concentrations and to specific hourly/seasonal combinations. The same combinations of two
seasons (winter and summer) and three times of day (7 a.m., 12 p.m., and 12 a.m.) that were used
in Section C.5 for 1,3-butadiene in Phoenix will be considered here. However, complete data
were not available for the winter - 12 a.m. combination, so that particular combination will not
be considered here.
Unlike the results presented in Section C.5, the method of Monte Carlo sampling will not
be used in this section. Instead, exact variances of the sample mean will be calculated using
Equation 6 in Section C.2. These variances will then be used, together with the normal
probability distribution, to estimate the probability that the sample mean will lie within 10
percent of the true mean for each stratum.
C.6.1 Houston Study Area
The Houston study area consists of all census block groups in a rectangular region
surrounding Houston, Texas, according to the 1990 U.S. census. The UTM coordinates of the
comers of this rectangle are listed below:
C-48
-------
Corner UTM Zone UTM (east) UTM (north)
Northeast 15 316km 3342km
Southeast 15 316km 3266km
Southwest 15 214km 3266km
Northwest 15 214km 3342km
The study area contains approximately 7,752 square kilometers of land area and 2265
BGs. The 2265 BGs were assumed to constitute the sampling frame for the study area (i.e., the
total population of available sampling units). The target population for this study is the current
population of Houston. (See Section C.2 for an explanation of why survey sampling methods
were chosen over spatial sampling methods).
C.6.2 Defining the Strata for Method C
For convenience, analysts utilized the emissions inventory grid to divide the study area,
consistent with the approach used for Phoenix. However, it should be noted that Method C does
not require the use of the emissions inventory grid. Unlike the grid used in Phoenix, which was
comprised of 4 km by 4 km cells, the emissions inventory grid for Houston is comprised of 2 km
by 2 km cells. This would require a sample size of 1330 receptor points under Method C, which
would result in unsatisfactory dispersion model run times. Since the 4 km by 4 km grid cells
worked well in Phoenix, analysts chose to aggregate the 2 by 2 grid cells into 4 by 4 grid cells for
Houston, which reduced the sample size required for Method C to 573 receptor points. Note that
the grid cell size being used here is in no sense "optimal", but it does work adequately for the
data analyzed thus far. It is not clear how an "optimal" grid cell size could be determined
without the use of trial-and-error methods, and would undoubtedly depend on the data at hand.
The results of previous analyses suggest that Method C will perform well even when the
sampling strata are defined after the dispersion model runs are completed. The only restriction
on the strata is that they must be comprised of aggregations of the original grid cells. As a "test"
of this assertion, analysts created two sets of sampling strata. The first set was created following
the methods outlined in Section C.2 using the emissions values for each grid cell. Breakpoint
values of 500 and 2000 were subjectively chosen to define these strata. Because the block group
centroids were more uniformly located over the Houston study area than they were in Phoenix,
analysts determined that it was not necessary to first divide the region into outer and central
regions. Figure C.6-1 presents a contour map of the Phoenix region indicating the cell emission
totals together with the 4 km by 4 km grid cells for the Houston study area. Figure C.6-2 shows
the 17 strata which were created using the contour map as a guide, as in Section C.2. These
strata shall hereafter be referred to as the emissions-defined strata.
The other set of strata were defined in a manner similar to the method commonly used to
define exposure districts for HAPEM. These strata were created by aggregating grid cells, and
are thus formed by concentric squares rather than the "bulls-eye" pattern of concentric circles
which is typically used in HAPEM. Figure C.6-3 presents the 17 strata which were formed in
this manner, and which will hereafter be referred to as the HAPEM-like strata.
C-49
-------
220
240
260
UTM (easting)
280
300
Figure C.6-1 Benzene emissions contour map of the Phoenix region.
C-50
-------
o
r}-
co
co
o
C\J
o
CO
O
f 8
t «
H CO
o
CO
CM
CO
16
17
13
_L
10
15
12
14
It
220
240
260 280
UTM (easting)
300
320
Figure C.6-2 Emissions-defined strata based on Phoenix contour map presented in Figure C.6-1.
C-51
-------
o
TT
3
o
CM
CO
CO
O)
o
o
oo
C\J
o
15
16
11
10
12
13
14
17
220
240
260 280
UTM (easting)
300
320
Figure C.6-3 HAPEM-like strata defined for Phoenix region.
C-52
-------
Figures C.6-4 and C.6-5 show the distribution of the BG centroids over the emissions-
defined strata and the HAPEM-like strata, respectively.
C.6.3 True Sampling Distributions
In Section C.5, analysts used Monte Carlo sampling to determine the distributional
properties of the sample mean for each stratum. This time-consuming process is unnecessary,
however, since the distributional properties of estimators from stratified random samples are well
known (Thompson, 1992). The estimators are unbiased (i.e., the mean of the estimator is equal
to the value being estimated). Furthermore, the true benzene concentrations are available for all
of the block groups in Houston. Consequently, the true variances are known. Therefore, the true
variances (or standard errors) of the sample means can be calculated using Equation 6 from
Section C.2. Assuming that the true distributions of the sample means are at least approximately
normal, analysts can then estimate the probability that the sample mean will be within 10 percent
of the true mean for each stratum.
C.6.4 Method C Applied to Annual Average Data with Emissions-Defined Strata
Table C.6-1 contains the descriptive statistics and sampling distributions for the annual
average benzene data using the emissions-defined strata. As noted previously, using Method C
with 4 km by 4 km grid cells requires 573 receptor points for Houston. The annual average
stratum means range from 0.19 ug/m3 in Stratum 16 to 1.28 ug/m3 in Stratum 2. The
concentration standard deviations range from 0.04 in Strata 16 and 17 to 2.65 in Stratum 3; the
standard deviations in strata 2 and 3 are each more than two times higher than that of any other
stratum. For most of the strata, the probability is approximately 1.0 that the sample mean will be
within 10 percent of the true mean. In fact, this probability is less than 95 percent only for strata
1 and 3. The problems encountered in these two strata will be discussed next.
For the annual average data, two block groups have concentrations which stand out as
outliers. The first grid cell is located (UTM coordinates 297 east, 3300 north) in stratum 2 of the
emissions-defined strata. The annual average concentration at this point is 10.31 ug/m3, while
the mean of stratum 2 is only 1.28 ug/m3 and the next highest concentration in this stratum is
1.75 ug/m3. However, this block group is in a grid cell by itself and would be included as a
receptor point in every sample. Therefore, this receptor point does not contribute to the variance
of the sample mean, and does not cause a problem. However, if the grid over the study area were
to be shifted, this may no longer be true, and this particular outlier would then pose a precision
problem. In this case, the discussion which follows for the other outlier would apply here as
well.
The other outlier is located at UTM coordinates 2834 east, 3289 north in stratum 3 of the
emissions-defined strata. The annual average concentration at this point is 36.14 ug/m3, while
the next highest concentration is 3.30 ug/m3 and the stratum mean is 1.15 ug/m3. This block
group is not located in a grid cell by itself, and therefore does contribute heavily to the variance
of the sample mean. This block group does cause a problem with the precision of the sample
C-53
-------
CO
CO
o
a
CO
en
O
CO
g
220
240
260 280
UTM (easting)
300
320
Figure C.6-4 Distribution of block-group centroids over emissions-defined strata.
C-54
-------
co
co
CO
O)
Z
o
I- CD
O
co
C\J
CO
• ••%
. * •••_••••
. •;.•::•..•:•;'*
wEPs^Vv*-®"
*"»
-------
Table C.6-1 Sampling Results for Annual Average Houston Benzene Concentrations
Using the Emissions-Defined Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
ofBGsin
stratum,
N
62
20
179
356
325
146
52
160
123
143
148
70
39
256
62
45
79
2265
True mean
concentration,
Hg/m3
0.46
1.28
1.15
0.66
0.52
0.48
0.51
0.40
0.36
0.44
0.43
0.36
0.28
0.57
0.22
0.19
0.20
0.54
True standard
deviation of
concentrations,
ug/m3
0.42
2.10
2.65
0.09
0.07
0.05
0.11
0.08
0.06
0.10
0.20
0.11
0.05
0.12
0.09
0.04
0.04
0.81
Sampling Distribution for Method C
Sample
size, n
18
12
28
22
32
18
12
55
34
43
52
38
22
48
54
43
42
573
True
standard
error of
sample
mean
0.0449
0.0130
0.4377
0.0103
0.0094
0.0056
0.0257
0.0046
0.0049
0.0081
0.0136
0.0101
0.0050
0.0062
0.0025
0.0003
0.0025
-
Probability sample
mean is within 1 0%
of true mean3
0.7000
1.0000
0.2079
1.0000
1.0000
1 .0000
0.9524
1.0000
1.0000
1.0000
0.9984
0.9996
1.0000
1.0000
1.0000
1.0000
1.0000
-
" Assuming approximate normality of the sampling distribution. Values listed as 1.0000 have been rounded to four
decimal places.
C-56
-------
mean, as is obvious in the results for Stratum 3 in Table C.6-1. The estimated probability that
the sample mean will be within 10 percent of the true mean is only 0.21 in this stratum. One
possible remedy for this problem will be discussed in the subsection "Forced Receptors -
Certainty Units" below.
The lack of precision observed in Stratum 1 was not caused by any one outlying block
group concentration. Instead, it was caused by the relatively high concentration variance within
one particular grid cell (UTM coordinates 302 east, 3290 north). The only obvious pre-sampling
solution for this problem would be to use smaller grid cells, which would, of course, result in
larger sample sizes. The objective that the mean concentration should be estimated to within 10
percent with 95 percent confidence was not met in Stratum 1. Instead, the mean concentration
can be estimated to within approximately 19 percent with 95 percent confidence.
C.6.5 Method C Applied to the Hourly/Seasonal Data with Emissions-Defined Strata
Tables C.6-2 through C.6-6 contain descriptive statistics and sampling distributions for
the concentrations from the five different hourly/seasonal combinations for the emissions-defined
strata. Analysts noted the following points concerning the descriptive statistics in these five
tables:
• The winter and summer benzene concentration patterns were very similar, which
was not the case for butadiene in Phoenix.
• The 7 a.m. and 12 a.m. benzene concentrations tended to be high relative to the 12
p.m. concentrations. The 12 a.m. concentrations were very similar to the 7 a.m.
concentrations, an unexpected result.
• The 7 a.m. winter concentrations ranged from 0.15 ug/m3 in stratum 16 to 1.34
ug/m3 in stratum 2. The 7 a.m. summer concentrations ranged from 014 ug/m3 in
stratum 16 to 1.78 ug/m3 in stratum 3. The 12 p.m. winter concentrations ranged
from 0.18 ug/m3 in stratum 15 to 0.78 ug/m3 in stratum 2. The 12 p.m. summer
concentrations ranged from 011 ug/m3 in stratum 17 to 0.58 ug/m3 in stratum 2.
The 12 a.m. summer concentrations ranged from 013 ug/m3 in stratum 16 to 2.48
ug/m3 in stratum 2.
• The concentration standard deviations for all strata were similar for all time
periods.
In the majority of the strata, the probability that the sample mean will be within 10
percent of the true stratum mean is approximately 1.0 for all hourly/seasonal combinations.
There are a few situations where this probability is less than 95 percent. However, as will be
shown below, in most of these cases the sampling method performed reasonably well.
C-57
-------
Table C.6-2 Sampling Results for Winter 7 Am Houston Benzene Concentrations
Using the Emissions-Defined Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
ofBGsin
stratum,
N
62
20
179
356
325
146
52
160
123
143
148
70
39
256
62
45
79
2265
True mean
concentration
(ug/m3)
0.47
1.34
1.23
0.77
0.70
0.56
0.52
0.40
0.48
0.61
0.55
0.39
0.30
0.58
0.26
0.15
0.26
0.62
True standard
deviation of
concentrations
(Hg/m3)
0.39
2.14
1.65
0.11
0.10
0.05
0.15
0.10
0.08
0.13
0.19
0.12
0.06
0.13
0.09
0.05
0.05
0.58
Sampling Distribution for Method C
Sample
size, n
18
12
28
22
32
18
12
55
34
43
52
38
22
48
54
43
42
573
True
standard
error of
sample
mean
0.0448
0.0211
0.2685
0.0127
0.0127
0.0075
0.0341
0.0060
0.0066
0.0123
0.0140
0.0138
0.0039
0.0083
0.0028
0.0005
0.0030
-
Probability sample
mean is within 10%
of true mean3
0.7069
1.0000
0.3537
1.0000
1.0000
1.0000
0.8702
1.0000
1.0000
1.0000
0.9999
0.9954
1.0000
1.0000
1.0000
1.0000
1.0000
-
a Assuming approximate normality of the sampling distribution. Values listed as 1.0000 have been rounded to four
decimal places.
C-58
-------
Table C.6-3 Sampling Results for Winter 12 Pm Houston Benzene Concentrations
Using the Emissions-Defined Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
of BGs in
stratum,
N
62
20
179
356
325
146
52
160
123
143
148
70
39
256
62
45
79
2265
True mean
concentration
(ug/m3)
0.31
0.78
0.75
0.63
0.53
0.53
0.50
0.45
0.37
0.39
0.38
0.36
0.33
0.53
0.18
0.26
0.21
0.49
True standard
deviation of
concentrations
(ug/m3)
0.29
1.20
1.41
0.07
0.07
0.04
0.07
0.07
0.08
0.06
0.12
0.11
0.06
0.07
0.05
0.05
0.06
0.44
Sampling Distribution for Method C
Sample
size, n
18
12
28
22
32
18
12
55
34
43
52
38
22
48
54
43
42
573
True
standard
error of
sample
mean
0.0345
0.0083
0.2339
0.0072
0.0091
0.0054
0.0166
0.0041
0.0054
0.0055
0.0069
0.0091
0.0062
0.0057
0.0013
0.0003
0.0025
-
Probability sample
mean is within 10%
of true mean2
0.6297
1.0000
0.2517
1.0000
1.0000
1.0000
0.9972
1.0000
1.0000
1.0000
1.0000
0.9999
1.0000
1.0000
1.0000
1.0000
1.0000
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-59
-------
Table C.6-4 Sampling Results for Summer 7 Am Houston Benzene Concentrations
Using the Emissions-Defined Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
ofBGsin
stratum,
N
62
20
179
356
325
146
52
160
123
143
148
70
39
256
62
45
79
2265
True mean
concentration
(Hg/m3)
0.87
1.73
1.78
0.91
0.86
0.61
0.60
0.47
0.67
0.99
0.70
0.44
0.25
0.66
0.27
0.14
0.38
0.79
True standard
deviation of
concentrations
(ug/m3)
0.52
3.15
1.71
0.15
0.12
0.05
0.21
0.13
0.14
0.21
0.25
0.18
0.06
0.17
0.12
0.05
0.13
0.70
Sampling Distribution for Method C
Sample
size, n
18
12
28
22
32
18
12
55
34
43
52
38
22
48
54
43
42
573
True
standard
error of
sample
mean
0.0657
0.0235
0.2657
0.0188
0.0158
0.0074
0.0393
0.0085
0.0079
0.0208
0.0171
0.0189
0.0065
0.0105
0.0029
0.0005
0.0050
-
Probability sample
mean is within 10%
of true meana
0.8162
1.0000
0.4973
1.0000
1.0000
1.0000
0.8731
1.0000
1.0000
1.0000
1.0000
0.9788
0.9999
1.0000
1.0000
1.0000
1 .0000
-
" Assuming approximate normality of the sampling distribution. Values listed as 1.0000 have been rounded to four
decimal places.
C-60
-------
Table C.6-5 Sampling Results for Summer 12 Pm Houston Benzene Concentrations
Using the Emissions-Defined Strata
Stratum
1
2
3.
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
of BGs in
stratum,
N
62
20
' 179
356
325
146
• 52
160
123
143
148
70
39
256
62
45
79
2265
True mean
concentration
(ug/m3)
0.32
0.58
0.49
0.39
0.31
0.30
0.29
0.25
0.19
0.22
0.20
0.26
0.21
0.34
0.17
0.13
0.11
0.30
True standard
deviation of
concentrations
(ug/m3)
0.31
0.83
1.28
0.05
0.06
0.04
0.04
0.05
0.05
0.06
0.11
0.07
0.04
0.05
0.04
0.03
0.03
0.39
Sampling Distribution for Method C
Sample
size, n
18
12
28
22
32
18
12
55
34
43
52
38
22
48
54
43
42
573
True
standard
error of
sample
mean
0.0398
0.0073
0.2142
0.0057
0.0076
0.0045
0.0092
0.0033
0.0040
0.0055
0.0086
0.0070
0.0040
0.0037
0.0009
0.0003
0.0020
-
Probability sample
mean is within 1 0%
of true meana
0.5727
1.0000
0.1792
1.0000
1.0000
1.0000
0.9987
1.0000
1 .0000
0.9999
0.9809
0.9998
1.0000
1.0000
1 .0000
1.0000
1.0000
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-61
-------
Table C.6-6 Sampling Results for Summer 12 Am Houston Benzene Concentrations
Using the Emissions-Defined Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
ofBGsin
stratum,
N
62
20
179
356
325
146
52
160
123
143
148
70
39
256
62
45
79
2265
True mean
concentration
(Hg/m3)
0.83
2.48
1.66
0.73
0.56
0.51
0.72
0.39
0.35
0.47
0.47
0.52
0.36
0.82
0.57
0.13
0.18
0.67
True standard
deviation of
concentrations
(ug/m3)
0.55
2.50
5.49
0.14
0.09
0.05
0.34
0.11
0.07
0.15
0.73
0.12
0.09
0.40
0.27
0.06
0.05
1.63
Sampling Distribution for Method C
Sample
size, n
18
12
28
22
32
18
12
55
34
43
52
38
22
48
54
43
42
573
True
variance of
sample
mean
0.0773
0.0549
0.9144
0.0174
0.0116
0.0081
0.0725
0.0062
0.0059
0.0156
0.0633
0.0111
0.0055
0.0195
0.0095
0.0004
0.0030
-
Probability sample
mean is within 1 0%
of true mean3
0.7149
1.0000
0.1438
1.0000
1.0000
1.0000
0.6764
1.0000
1.0000
0.9972
0.5386
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-62
-------
In all cases, the sample mean from stratum 3 has very low precision. The same block
group which had an unusually large annual average concentration has a relatively large
concentration at every hourly/seasonal combination. It is exclusively this outlier which causes
the precision in stratum 3 to be so low; this point will be explored further in a later subsection.
The results from Tables C.6-2 through C.6-6 indicate that the sample mean from stratum
1 fails to meet the precision objective in all cases. The cause of this lack of precision is the same
as was discussed previously for the annual average data. The sample mean will not be within 10
percent of the true stratum 1 mean at least 95 percent of the time for any of these hourly/seasonal
combinations. However, in all of these cases, the sample mean will be within 25 percent of the
true mean at least 95 percent of the time.
The sample mean from stratum 7 fails to met the precision objectives for the winter 7
a.m., summer 7 a.m., and summer 12 a.m. data sets. In all cases, however, the sample mean from
stratum 7 will be within 20 percent of the true mean at least 95 percent of the time. The sample
mean from stratum 11 for the summer 12 a.m. data will be within approximately 27 percent of
the true mean approximately 95 percent of the time.
C.6.6 Results Using HAPEM-Like Strata
The set of HAPEM-like strata, shown in Figure C.6-3, are being used to demonstrate the
flexibility available in sampling Method C. It was claimed in Section C.5 that Method C would
produce precise estimates, no matter how the grid cells were aggregated into strata (i.e., exposure
districts). The results in this section are being presented in support of this claim.
Tables C.6-7 through C.6-12 contain descriptive statistics and sampling distributions for
the annual average concentrations and the five different hourly/seasonal combinations for the
HAPEM-like strata. In almost every case, the sample mean will be within 10 percent of the true
stratum mean concentration at least 95 percent of the time (this probability is actually close to 1.0
in most cases). The outlying block group concentration described in the previous subsection is
now located in stratum 9. The sample mean from this stratum has low precision for all data sets;
this point will be explored further in the following subsection. The only other case where the
level of precision is lower than prescribed is in stratum 17 for the summer 12 a.m. data. Here,
the sample mean is within 10 percent of the true mean with approximately 85 percent confidence.
C.6.7 Forced Receptors - Certainty Units
In the previous subsections, one very extreme outlier was consistently noted in the
Houston benzene concentrations. The UTM coordinates for the block group centroid where
these high concentrations were observed are 2834 east, 3289 north. This point is marked with an
asterisk in Figures C.6-4 and C.6-5. Table C.6-13 contains the benzene concentrations observed
at this centroid in each of the data sets. The stratum means for both types of strata are also
included in this table for comparative purposes. The concentration observed at this one block
group is always between 13 and 45 times higher than the mean concentration of the entire
stratum! There are no other block groups within 1 km of this block group. There are 5 block
C-63
-------
Table C.6-7 Sampling Results for Annual Average Houston Benzene Concentrations
Using the HAPEM-Like Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
ofBGsin
stratum,
N
90
108
72
148
277
62
94
283
255
55
76
141
185
53
67
82
217
2265
True mean
concentration
(ug/m3)
0.55
0.52
0.48
0.50
0.68
0.44
0.43
0.50
0.91
0.34
0.36
0.37
0.69
0.23
0.23
0.20
0.48
0.54
True standard
deviation of
concentrations
(ug/m3)
0.05
0.05
0.05
0.07
0.09
0.08
0.10
0.07
2.24
0.12
0.10
0.08
0.26
0.21
0.07
0.05
0.74
0.81
Sampling Distribution for Method C
Sample
size, n
8
16
16
16
16
27
32
31
32
30
31
41
46
51
49
44
87
573
True
standard
error of
sample
mean
0.0091
0.0084
0.0060
0.0094
0.0133
0.0059
0.0113
0.0098
0.3069
0.0119
0.0130
0.0062
0.0208
0.0005
0.0029
0.0025
0.0156
-
Probability sample
mean is within 1 0%
of true mean"
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.9999
1.0000
0.2322
0.9961
0.9949
1.0000
0.9991
1.0000
1.0000
1.0000
0.9980
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-64
-------
Table C.6-8 Sampling Results for Winter 7 am Houston Benzene Concentrations
Using the HAPEM-Like Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
of BGs in
stratum,
N
90
108
72
148
277
62
94
283
255
55
76
141
185
53
67
82
217
2265
True mean
concentration
(ug/m3)
0.56
0.53
0.50
0.62
0.75
0.45
0.43
0.68
1.01
0.39
0.36
0.51
0.83
0.27
0.21
0.27
0.55
0.62
True standard
deviation of
concentrations
(ug/m3)
0.07
0.06
0.07
0.08
0.10
0.09
0.14
0.12
1.40
0.13
0.11
0.11
0.26
0.20
0.09
0.06
0.74
0.58
Sampling Distribution for Method C
Sample
size, n
8
16
16
16
16
27
32
31
32
30
31
41
46
51
49
44
87
573
True
standard
error of
sample
mean
0.0158
0.0108
0.0095
0.0128
0.0145
0.0074
0.0165
0.0132
0.1879
0.0169
0.0145
0.0076
0.0258
0.0008
0.0022
0.0030
0.0151
-
Probability sample
mean is within 10%
of true mean"
0.9996
1.0000
1.0000
1.0000
1 .0000
1.0000
0.9915
1.0000
0.4095
0.9780
0.9865
1.0000
0.9988
1.0000
1.000
1.0000
0.9998
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-65
-------
Table C.6-9 Sampling Results for Winter 12 pm Houston Benzene Concentrations
Using the HAPEM-Like Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
ofBGsin
stratum,
N
90
108
72
148
277
62
94
283
255
55
76
141
185
53
67
82
217
2265
True mean
concentration
(ug/m3)
0.58
0.53
0.53
0.54
0.65
0.43
0.46
0.52
0.66
0.32
0.40
0.35
0.52
0.19
0.29
0.21
0.36
0.49
True standard
deviation of
concentrations
(Ug/m3)
0.04
0.05
0.05
0.05
0.06
0.08
0.07
0.07
1.18
0.10
0.07
0.06
0.12
0.10
0.07
0.06
0.43
0.44
Sampling Distribution for Method C
Sample
size, n
8
16
16
16
16
27
32
31
32
30
31
41
46
51
49
44
87
573
True
standard
error of
sample
mean
0.0103
0.0088
0.0056
0.0083
0.0085
0.0059
0.0079
0.0098
0.1640
0.0105
0.0082
0.0056
0.0120
0.0006
0.0036
0.0024
0.0107
-
Probability sample
mean is within 1 0%
of true mean3
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.3111
0.9977
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.9993
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-66
-------
Table C.6-10 Sampling Results for Summer 7 am Houston Benzene Concentrations
Using the HAPEM-Like Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
of BGs in
stratum,
N
90
108
72
148
277
62
94
283
255
55
76
141
185
53
67
82
217
2265
True mean
concentration
(Mg/m3)
0.63
0.61
0.59
0.71
0.90
0.49
0.53
0.81
1.33
0.41
0.36
0.77
1.25
0.27
0.18
0.42
0.79
0.79
True standard
deviation of
concentrations
(ug/m3)
0.07
0.08
0.07
0.14
0.16
0.14
0.19
0.14
1.49
0.16
0.11
0.21
0.45
0.24
0.08
0.15
1.06
0.70
Sampling Distribution for Method C
Sample
size, n
8
16
16
16
16
27 .
32
31
32
30
31
41
46
51
49
44
87
573
True
standard
error of
sample
mean
0.0131
0.0142
0.0102
0.0182
0.0223
0.0144
0.0231
0.0158
0.1845
0.0214
0.0111
0.0096
0.0441
0.0005
0.0038
0.0049
0.0212
-
Probability sample
mean is within 10%
of true mean3
1.0000
1.0000
1.0000
0.9999
1.0000
0.9994
0.9790
1.0000
0.5298
0.9417
0.9989
1.0000
0.9955
1.0000
1.0000
1.0000
0.9998
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-67
-------
Table C.6-11 Sampling Results for Summer 12 pm Houston Benzene Concentrations
Using the HAPEM-Like Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
of BGs in
stratum,
N
90
108
72
148
277
62
94
283
255
55
76
141
185
53
67
82
217
2265
True mean
concentration
(ug/m3)
0.34
0.33
0.30
0.32
0.40
0.29
0.26
0.30
0.41
0.24
0.23
0.19
0.31
0.18
0.16
0.11
0.25
0.30
True standard
deviation of
concentrations
(ug/m3)
0.03
0.03
0.04
0.05
0.05
0.06
0.05
0.06
1.08
0.06
0.06
0.05
0.11
0.10
0.06
0.03
0.33
0.39
Sampling Distribution for Method C
Sample
size, n
8
16
16
16
16
27
32
31
32
30
31
41
46
51
49
44
87
573
True
standard
error of
sample
mean
0.0073
0.0051
0.0048
0.0070
0.0073
0.0049
0.0041
0.0081
0.1503
0.0078
0.0058
0.0046
0.0073
0.0003
0.0024
0.0020
0.0127
-
Probability sample
mean is within 10%
of true meana
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.9998
0.2164
0.9979
0.9999
1.0000
1.0000
1.0000
1.0000
1 .0000
0.9554
-
* Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-68
-------
Table C.6-12 Sampling Results for Summer 12 am Houston Benzene Concentrations
Using the HAPEM-Like Strata
Stratum
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Total
Descriptive Statistics
Number
of BGs in
stratum,
N
90
108
72
148
277
62
94
283
255
55
76
141
185
53
67
82
217
2265
True mean
concentration
(ug/m3)
0.58
0.58
0.51
0.56
0.77
0.64
0.46
0.53
1.27
0.57
0.45
0.36
0.94
0.64
0.22
0.18
0.70
0.67
True standard
deviation of
concentrations
(ug/m3)
0.05
0.07
0.06
0.10
0.13
0.16
0.19
0.09
4.62
0.21
0.35
0.10
0.58
0.95
0.12
0.05
1.08
1.63
Sampling Distribution for Method C
Sample
size, n
8
16
16
16
16
27
32
31
32
30
31
41
46
51
49
44
87
573
True
standard
error of
sample
mean
0.0118
0.0089
0.0071
0.0115
0.0222
0.0218
0.0228
0.0125
0.6416
0.0140
0.0421
0.0068
0.0393
0.0012
0.0030
0.0030
0.0484
-
Probability sample
mean is within 10%
of true meana
1.0000
1.0000
1.0000
1.0000
0.9995
0.9968
0.9582
1.0000
0.1572
1.0000
0.7177
1.0000
0.9832
1.0000
1 .0000
1.0000
0.8530
-
a Assuming approximate normality of the sampling distribution.
decimal places.
Values listed as 1.0000 have been rounded to four
C-69
-------
Table C.6-13 The Sampling Distribution for Stratum 3 of the Emissions-Defined Strata and Stratum 9 of the HAPEM-Like
Strata after the Outlying Block Group at UTM Coordinates 283.735 East, 3289.09 North Is Forced into the Sample
Data set
Annual Average
Winter, 7 am
Winter, 1 2 pm
Summer, 7 am
Summer, 12 pm
Summer, 12 am
Concentration
in outlying BG,
ug/m"
36.14
22.69
19.32
23.04
17.47
73.93
Emissions-Defined Stratum 3
Sampling Distribution
Stratum Mean
Concentration
(ug/m')
1.15
1.23
0.75
1.78
0.49
1.66
True standard
error of sample
mean
0.0450
0.0499
0.0268
0.0821
0.0118
0.0894
Probability
sample mean is
within 10% of
true mean"
0.9897
0.9864
0.9949
0.9699
1 .0000
0.9363
HAPEM-Like Stratum 9
Sampling Distribution
Stratum Mean
Concentration
(ug/m3)
0.91
1.01
0.66
1.33
0.41
1.27
True standard
error of sample
mean
0.0285
0.0322
0.0177
0.0510
0.0078
0.0597
Probability sample mean
is within 10% of true
mean3
0.9986
0.9983
0.9998
0.9910
0.9979
0.9669
' Assuming approximate normality of the sampling distribution.
C-70
-------
groups within 2 km of this block group, and none of them has an annual average concentration
higher than 3.3 ug/m3.
Because this outlying block group is not in a grid cell by itself, it has great influence on
the variance of the sample mean. Recall that the variance of the sample mean (Equation 6)
depends only on the within grid cell variances. The presence of an outlier of this magnitude in
any grid cell will substantially increase the variance within that grid cell, which will then inflate
the true variance of the sample mean.
Unfortunately, there is not a simple method that would help to alleviate the problems
caused by such an outlier. One way to eliminate the influence from this one point would be to
force it into the sample. In survey sampling terminology, any sampling unit which is included in
the sample automatically is known as a "certainty unit." Certainty units do not contribute to the
variance of the sample mean because they are included in every sample, and are not selected
randomly. In effect, under Method C, any block group which is located in a cell either alone or
with only one other block group is a certainty unit. Therefore, treating an outlier as a certainty
unit would be equivalent to treating it as if it were contained within its own grid cell. Of course,
treating this outlier as a certainty unit would require researchers to have the foresight to know,
before drawing a sample, that this point would probably have extremely high concentrations.
While this is probably not a realistic suggestion in general, it may have been possible for
researchers to predict before hand that this particular point would have high concentrations; there
are 108 point sources of benzene within 1 km of the outlying block group. Identifying this type
of situation before hand would mean looking at the location of every block group in the frame
with respect to its proximity to point sources, which would be a very time consuming operation.
Furthermore, it is not recommended that researchers include points as certainty units without
good justification.
Forcing the outlying block group into the sample would eliminate the precision problems
which were observed in stratum 3 of the emissions-defined strata and stratum 9 of the HAPEM-
like strata. The sampling distribution obtained from these strata after this outlier is forced into
the sample are shown in Table C.6-13. In all cases, the standard errors of the sample mean are
substantially reduced. Forcing this block group into the sample would increase the sample size
in each of these strata by 1. Thus, the results in Table C.6-13 are for sample sizes of 29 and 33
from stratum 3 and stratum 9, respectively.
C.6.8 Summary Discussion
The results of this section indicate that sampling Method C works reasonably well, even for an
air pollutant which is produced largely by point sources. In most cases, the method produces
estimates which adequately satisfy the sampling objectives prescribed in Section C.I. In fact, in
most cases the estimates are actually "too precise", which indicates that larger grid cells and
smaller sample sizes could be used. However, "optimizing" the size of the grid cell would have
to be accomplished by trial-and-error methods for each pollutant, and it is not clear how this
could be done accomplished to obtaining estimates of the grid cell variances. For the two
pollutants analyzed to date (1,3-butadiene and benzene), precise estimators could have been
C-71
-------
obtained by choosing only one receptor point per grid cell. This would have substantially
reduced the sample size. However, this sample size reduction would make it impossible to
obtain an unbiased estimate for the standard error of the sample mean. It is necessary for
researchers to have such an estimate in order to have an indication of the precision of their
estimates.
C-72
-------
C.7 STEP-BY-STEP GUIDELINES FOR USING SAMPLING METHOD C
This section summarizes the step-by-step procedure for drawing samples of receptor
points using sampling Method C, as proposed in Section C.5, and then defining exposure
districts for use in HAPEM and similar exposure assessment models. Steps 1 through 3 below
define the method for obtaining the sample of receptor points. Exposure districts are created in
Step 4, and estimates of the mean and variance of the ambient concentration across each district
are calculated in Step 5. Note that Steps 4 and 5 can be performed after the sample of receptor
points has been selected.
1. Define the study area and obtain a listing of all census block group centroids
within the study area. The user will need to have the UTM coordinates of all of
the block group centroids within the study area.
2. Define a grid over the entire study area. In this report, analysts used a regular
grid of 4 km by 4 km cells, created using the emissions inventory grid. This grid
does not need to be regular, however, and the cell width of 4 km is not necessarily
the "optimal" grid size.
3. Randomly choose 2 block group centroids from each of the populated grid cells.
If there is only one centroid in a grid cell, it is included in the sample
automatically. This step can be performed prior to defining the "exposure
districts" (see the next step). Recall that a "populated" grid cell was defined as
one which contains at least one block group centroid. Table C.7-1 contains
example SAS code for performing Steps 2 and 3.
4. Define the "exposure districts". This step can be performed using the emissions
values in the same manner as the sampling strata were defined in Section C.2, or it
can be accomplished using the actual concentration values after the sampling has
been completed and the dispersion model has been run. Consequently, Method C
provides the researchers performing the exposure assessments with greater
flexibility in defining the exposure districts. The individual grid cells can be
aggregated in any way desirable to form the exposure districts, given that the final
exposure districts are formed simply by aggregating grid cells.
5. Obtain estimates for mean concentrations in each exposure district. This step can
be accomplished using Equation 5. In addition, unbiased estimates of the
variances of these sample means should be obtained using Equation 6.
This approach is independent of the pollutant and is applicable to all cities of interest. The same
set of receptors could be used for each pollutant within a given city.
C-73
-------
Table C.7-1 Example SAS Code8 for Generating a Sample of Receptor Points
Using Method C
DATA ONE;
INFILE 'BGS.DAT'; *File containing block group information;
INPUT EAST NORTH; *May need to define additional variables;
U = RANUNI(-l); *Generates uniform (0,1) random number;
*** Creates 4 km by 4 km grid over area with southwest corner at UTM coordinates 214 east, 3266 north, as in
Houston. Assigns each bg to a grid cell;
X = FLOOR((EAST - 214)/4);
Y = FLOOR((WEST - 3266)/4);
ID = X* 100 + Y; *Creates pseudo id for each grid cell;
PROC SORT;
BY ID U;
***Now, two data sets will be created. The first will contain the block group with the smallest random number
from each grid cell. This block group will be included as a receptor point. The second data set will contain all
the rest of the block groups. Then, the block group with the next highest random number is selected, and the two
data sets are merged together into the final data set;
DATA TWO THREE;
SET ONE;
BY ID;
IF (FIRST.ID) THEN OUTPUT TWO;
ELSE OUTPUT THREE;
DATA FOUR;
SET THREE;
BY ID;
IF (FIRST.ID);
DATA FIVE; *The final data set containing selected receptors;
SET TWO FOUR;
alt is assumed that the user has a file ('BGS.DAT') which contains the listing of all block group
centroids in the study area. This file should contain at least the UTM coordinates of the block
group centroids.
C-74
-------
To illustrate the procedures to be used in making these calculations, we have constructed
a hypothetical example in which analysts used the sample SAS code in Table 7-1 to implement
Steps 1 through 3 with respect to Study Area X. The effort produced a grid of 4 km by 4 km
cells over the study area and then randomly selected a maximum of two BGs from each grid cell.
In Step 4, the analysts defined 20 exposure districts as contiguous collections of grid cells based
on a review of the emissions data.
Exposure District No. 1 was defined as the aggregation of 10 populated grid cells
(identified as h = 1 through 10 in Table C.7-2). Table C.I-2 lists values for
Nh = total number of BGs in cell h, and
nh = number of BGs selected from cell h.
For each grid cell, Nh > 2. Consequently, analysts were able to randomly select two BGs from
each cell (i.e., nh = 2).
Step 5 was the only step in the methodology which required statistical calculations.
Each selected BG was associated with a concentration value obtained from the dispersion model
run. The mean (yh) and standard deviation (oh) statistics listed in Table C.7-2 for each cell were
calculated from these values using the standard formulas. The mean concentration for the entire
exposure district (y) was then calculated by the
expression
y N h=1
in which
L = total number of grid cells in the exposure district = 10, and
N = total number of BGs in the exposure district =171.
The variance of y was calculated by the expression
Var(y) = V -^
A
tf
2
AT - nu
/2 ft
h
2
h
"»
in which all terms have been previously defined. Making the indicated substitutions from Table
7-2, the analysts obtained y = 0.596 and var(y) = 0.003 for Exposure District No. 1. The same
procedure was then applied to each of the other exposure districts in the hypothetical study area.
C-75
-------
Table C.7-2 Sample Inputs for Step 5 Calculations
(Exposure District 1 of Hypothetical Study Area)
Grid cell ID (h)
1
2
3
4
5
6
7
8
9
10
Total number of
BGs in grid cell (Nh)
16
18
15
21
14
17
18
16
19
17
Number of selected
BGs in grid cell (nh)
2
2
2
2
2
2
2
2
2
2
Mean of
concentrations
estimated for
selected BGs
(7h)> ^g/m3
0.59
0.40
0.32
0.39
0.62
0.63
0.85
0.95
0.60
0.65
Standard deviation
of concentrations
estimated for
selected BGs (oh),
ug/m3
0.20
0.24
0.10
0.35
0.29
0.29
0.28
0.16
0.22
0.16
C-76
-------
C.8 REFERENCES
Johnson, T., Warnasch, J., McCoy, M., Capel, J., and Riley, M. 1996. Developmental Research
for the Hazardous Air Pollutant Exposure Model (HAPEM) as Applied to Mobile Source
Pollutants. IT Air Quality Services, Durham, North Carolina. February.
Thompson, S. K. 1992. Sampling. Wiley, New York.
C-77
-------
APPENDIX D
DEFAULT STACK PARAMETERS
OBTAINED FROM OZONE TRANSPORT ASSESSMENT GROUP
FOR SUBSTITUTION OF MISSING DATA
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
101001
101002
101003
101004
101005
101006
101007
101008
101009
101010
101012
101013
102001
102002
102003
102004
102005
102006
102007
102008
102009
102010
102011
102012
102013
102014
102999
103001
103002
103003
103004
103005
103006
103007
103009
103010
103012
103013
105001
105002
201001
201002
201007
201008
201009
Stack
Height
(ft)
332
483
410
252
308
185
134
398
182
357
212
342
151
165
217
95
74
68
134
194
102
67
80
176
125
171
25
141
153
104
98
60
75
61
82
42
126
74
40
32
55
55
38
28
30
Stack
Diameter
(ft)
6.23
15.91
19.05
12.18
13.78
10.14
8.80
15.69
5.43
15.55
8.24
13.58
7.74
7.08
11.63
5.18
3.96
4.27
6.94
8.13
4.76
3.49
6.56
6.72
6.01
8.03
1.20
6.05
6.79
4.89
5.19
4.13
3.73
3.20
3.99
1.90
6.81
5.94
2.07
1.40
7.47
7.32
3.28
3.28
12.58
Exit
Temperature
(°F)
423
299
271
343
394
348
413
329
365
315
337
333
356
387
283
413
395
396
480
361
380
361
208
328
430
456
205
416
394
381
417
383
382
325
380
346
282
433
294
449
647
651
375
949
791
Exit
Velocity
(ft/s)
52.83
63.79
79.75
53.60
66.36
41.27
55.69
67.72
55.83
36.25
62.75
66.38
29.45
35.18
35.62
30.13
29.80
27.27
31.78
34.50
34.54
23.55
28.31
43.60
30.78
43.47
0.00
18.33
25.46
30.24
25.72
30.45
23.94
19.06
49.59
34.99
36.85
27.02
27.47
32.58
69.49
67.07
31.19
89.45
152.12
D-l
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
201900
202001
202002
202003
202004
202005
202009
202010
203001
203002
203003
203010
204001
204002
204003
204004
288888
301001
301003
301005
301006
301007
301008
301009
301010
301011
301012
301013
301014
301015
301016
301017
301018
301019
301020
301021
301022
301023
301024
301025
301026
301027
301028
301029
301030
Stack
Height
(ft)
125
32
33
16
30
151
40
31
30
28
57
26
50
0
67
37
52
91
101
78
60
26
51
52
43
36
80
82
28
43
82
62
48
85
34
97
38
140
65
136
48
89
106
102
107
Stack
Diameter
(ft)
1.31
2.26
1.92
0.69
1.70
7.52
3.74
0.87
1.60
1.87
1.38
2.15
14.39
0.00
8.82
1.83
2.81
1.46
5.99
2.81
3.39
1.00
6.76
2.47
1.57
1.42
1.39
2.98
3.58
1.24
5.46
2.59
2.33
4.28
1.76
2.69
2.39
4.90
3.51
2.42
2.47
3.15
3.55
4.97
5.84
Exit
Temperature
(°F)
131
552
729
414
635
374
689
674
786
706
700
393
291
0
449
348
334
270
405
464
480
82
156
117
153
100
214
346
102
149
120
154
180
251
127
117
171
145
178
96
158
139
118
118
128
Exit
Velocity
(ft/s)
0.00
86.47
75.22
26.50
21.35
53.81
87.90
28.23
78.14
95.23
69.81
16.91
39.09
0.00
83.83
31.64
67.34
52.03
75.24
45.00
29.53
26.06
24. 98
15.72
15.85
42.37
20.15
72.30
13.68
21.02
41.38
66.47
38.73
37.37
14.34
13.13
25.43
50.05
31.40
34.69
27.04
47.72
56.52
38.30
56.58
D-2
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
301031
301032
301033
301034
301035
301038
301039
301040
301041
301042
301045
301060
301066
301070
301091
301099
301100
301112
301120
301121
301124
301125
301126
301127
301130
301132
301133
301137
301140
301152
301153
301156
301157
301158
301167
301169
301174
301176
301181
301190
301195
301197
301202
301205
301206
Stack
Height
(ft)
97
162
67
49
93
112
66
110
87
132
36
50
0
42
54
0
49
104
46
52
100
46
34
74
63
49
40
34
54
210
42
43
45
34
50
45
53
0
34
72
0
74
48
43
52
Stack
Diameter
(ft)
0.96
3.83
1.59
2.43
2.35
2.06
1.46
7.78
1.00
3.28
0.00
1.85
0.00
2.41
1.44
0.00
3.54
5.12
3.24
2.61
0.00
2.60
0.81
1.92
1.42
6.01
0.77
0.90
22.39
3.50
1.66
1.29
34.10
0.34
3.31
0.85
10.14
0.00
0.87
6.69
0.00
10.78
1.64
1.82
13.58
Exit Exit
Temperature Velocity
(°F) (ft/s)
151
719
229
259
146
234
642
147
203
146
109
124
0
139
294
0
272
580
178
150
77
129
107
283
75
274
552
89
425
0
567
95
309
130
176
131
344
0
122
438
0
460
122
116
322
32.
36.
34.
12.
28.
66.
1.
70.
26.
44.
0.
15.
0.
7.
30.
0.
39.
44.
37.
27.
0.
31.
0.
44.
53.
4.
22.
9.
12.
0.
2.
1.
0.
103.
16.
0.
34.
0.
18.
7.
0.
27.
24.
54.
8.
26
45
23
60
60
21
97
06
24
86
00
71
00
61
32
00
83
02
31
38
00
43
00
62
05
97
44
04
02
00
01
53
03
27
02
94
47
00
16
48
00
69
48
92
75
D-3
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
301210
301211
301250
301251
301252
301253
301254
301258
301301
301303
301304
301305
301800
301810
301820
301830
301840
301870
301875
301885
301888
301900
301999
302001
302002
302003
302004
302005
302006
302007
302008
302009
302010
302012
302013
302014
302015
302016
302017
302018
302019
302022
302026
302030
302031
Stack
Height
(ft)
27
85
36
45
55
38
73
48
31
51
70
30
27
34
15
30
42
24
29
31
26
94
45
39
85
160
43
74
36
66
63
87
161
133
67
52
84
79
69
40
35
39
38
63
72
Stack
Diameter
(ft)
2.76
84.06
2.51
4.10
0.98
0.33
6.74
4.15
1.86
2.50
3.40
0.79
2.64
4.51
2.98
2.61
2.11
1.62
3.02
2.16
2.21
7.89
2.04
3.70
2.43
3.75
1.58
2.94
3.84
2.95
2.43
2.42
4.18
5.93
3.04
1.58
5.52
6.32
0.00
3.12
1.61
2.24
3.42
2.69
7.00
Exit
Temperature
(°F)
146
400
189
235
201
104
253
187
93
130
126
141
115
107
82
153
189
107
100
287
104
790
184
157
690
210
68
89
79
113
86
130
87
151
156
109
251
213
350
263
101
68
71
156
80
Exit
Velocity
(ft/s)
0.93
0.14
16.98
27.67
48.36
0.95
22.41
6.41
11.79
48.35
55.49
63.70
3. 65
24.52
9.22
0.49
11. 66
1.76
0.26
10.53
28.37
26.13
23.07
35.74
44.81
33.14
21.00
44.59
38.50
44 .02
39.16
21.48
25.84
182.54
21.79
46.45
43.72
53.64
O.OC
24.32
48.20
21.82
9.96
58.96
22.16
D-4
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
302032
302033
302036
302038
302040
302888
302900
302999
303000
303001
303002
303003
303005
303006
303007
303008
303009
303010
303012
303014
303023
303024
303030
303888
303900
303999
304001
304002
304003
304004
304005
304006
304007
304008
304009
304010
304020
304022
304049
304050
304888
304900
304999
305001
305002
Stack
Height
(ft)
43
59
23
0
58
104
68
46
49
56
144
204
261
79
95
150
119
132
134
80
112
33
97
74
155
37
50
52
49
69
29
50
47
41
25
71
57
32
0
29
36
66
55
37
34
Stack
Diameter
(ft)
1
3
1
0
2
2
3
1
1
4
4
8
9
9
16
7
9
8
2
3
4
3
4
7
6
2
3
3
3
2
2
6
4
2
1
2
8
2
0
0
3
3
2
3
4
.42
.68
.55
.00
.04
.00
.05
.92
.66
.53
.23
.18
.25
.49
.46
.40
.52
.32
.00
.23
.71
.40
.11
.66
.43
.63
.33
.54
.93
.45
.09
.00
.46
.67
.08
.50
.24
.17
.00
.86
.48
.43
.54
.45
.44
Exit Exit
Temperature Velocity
(°F) (ft/s)
318
117
235
0
251
117
222
132
139
140
294
298
184
299
142
459
455
89
0
206
107
72
135
112
640
181
429
367
169
208
164
97
214
281
538
151
258
193
0
183
155
593
294
268
230
26
40
23
0
53
27
36
19
17
47
59
26
27
98
0
28
24
35
0
0
54
112
44
60
51
34
30
40
53
39
73
45
38
30
8
11
30
50
0
24
34
33
19
44
50
.01
.71
.51
.00
.47
.72
.26
.60
.37
.85
.51
.15
.53
.70
.00
.31
.17
.21
.00
.00
.06
.22
.84
.93
.65
.51
.51
.51
.49
.88
.15
.58
.61
.84
.60
.58
.99
.06
.00
.75
.15
.37
.89
.07
.71
D-5
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
305003
305004
305005
305006
305007
305008
305009
305010
305011
305012
305013
305014
305015
305016
305017
305018
305019
305020
305021
305025
305026
305030
305032
305033
305040
305100
305101
305102
305103
305104
305105
305150
305888
305900
305999
306001
306002
306003
306004
306005
306006
306007
306008
306009
306010
Stack
Height
(ft)
41
91
44
86
105
38
51
72
31
58
85
90
56
68
54
38
60
29
36
41
32
29
49
0
34
103
52
72
38
34
58
41
26
58
57
107
129
94
112
15
49
34
35
142
61
Stack
Diameter
(ft)
2.72
5.59
2.87
3.99
5.05
2.03
3.00
5.27
3.13
3.77
2.61
4.77
2.65
5.12
6.03
4.00
3.95
3.45
1.35
2.16
0.95
1.83
1.70
0.00
4.94
1.88
2.43
1.85
4.00
1.83
1.74
1.05
1.89
4.29
2.36
5.28
6.50
6.96
11.00
11.33
10.93
11.97
13.16
14.49
4.23
Exit
Temperature
(°F)
329
403
244
167
167
150
160
98
81
231
143
399
216
214
186
403
105
86
80
107
73
120
70
77
146
91
78
82
77
101
79
176
117
384
197
577
428
480
705
151
224
115
198
1168
235
Exit
Velocity
(ft/s)
42.46
26.88
19.69
47.86
39.50
42.55
52.08
50.47
25.09
46.46
32.14
45.81
39.15
39.41
42.61
66.58
37.34
55.33
0.00
57.35
6.52
38.11
52.87
0.00
40.74
54.62
56.01
26.41
3.28
43.73
60.05
49.40
13.98
48.86
48.52
22.39
60.46
59.65
28.48
18.64
58.06
10.98
24.46
22.95
15.59
D-6
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
306011
306012
306014
306016
306099
306100
306888
306999
307001
307002
307003
307004
307005
307007
307008
307011
307013
307020
307030
307888
307900
307999
308001
308005
308006
308007
308008
308009
308010
308900
308999
309001
309002
309003
309006
309010
309011
309015
309016
309020
309025
309030
309040
309060
309888
Stack
Height
(ft)
49
91
113
199
189
24
28
31
174
105
150
53
19
56
58
48
42
34
41
33
97
40
41
60
36
34
31
291
35
35
37
28
27
35
0
32
33
33
36
29
35
30
24
29
39
Stack
Diameter
(ft)
3.08
5.50
10.41
2.50
3.67
2.53
52.11
8.79
5.54
2.63
3.60
5.24
4.06
3.64
2.89
1.07
6.67
3.39
2.91
3.04
4.45
3.68
2.23
1.28
2.23
2.16
2.07
2.04
1.70
1.67
2.00
2.52
2.09
2.11
0.00
2.77
2.13
2.92
2.47
3.50
2.40
2.23
1.72
2.73
1.42
Exit Exit
Temperature Velocity
(°F) (ft/s)
755
305
787
1400
872
142
149
255
204
190
138
138
120
174
148
200
145
87
101
144
231
115
92
74
127
116
110
76
124
287
122
183
86
127
0
110
116
102
90
142
468
71
97
77
122
30.07
35.04
41.03
40.74
58.46
7.30
5.36
22.19
50.59
45.76
42.44
10.80
9.60
48.15
34.81
0.04
27.24
34.18
45.28
29.66
40.02
24.10
40.78
44.45
32.85
37.04
54.51
54.12
45.27
21.12
34.29
27.57
36.32
6.07
0.00
35.18
41.93
30.52
19.21
14.36
36.22
14.23
58.45
39.47
27.12
D-7
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
309900
309999
310001
310002
310004
310888
311001
312999
313010
313030
313065
313070
313900
313999
314009
314010
314011
314015
314999
315010
315020
320999
330001
330002
330003
330004
330005
330888
360001
385001
390001
390002
390004
390005
390006
390007
390008
390009
390010
390012
390013
399900
399999
401001
401002
Stack
Height
(ft)
39
36
92
40
30
9
20
33
30
68
40
32
28
35
20
18
40
0
36
45
47
27
40
35
18
40
0
29
40
27
65
110
116
69
62
159
95
79
48
140
124
74
41
26
30
Stack
Diameter
(ft)
2.26
2.00
4.71
4.38
1.87
7.37
0.00
2.27
2.00
1.41
3.05
1.83
0.95
3.17
2.20
1.50
3.75
0.00
4.65
1.50
3.49
2.29
2.80
3.11
3.56
3.28
1.00
3.00
2.00
16.29
7.00
8.86
6.26
5.19
4.10
7.29
5.77
3.96
4.01
7.12
7.34
3.38
4.35
2.39
2.42
Exit Exit
Temperature Velocity
(°F) (ft/s)
318
154
241
632
587
98
77
111
77
86
251
723
313
99
85
575
93
0
99
70
163
97
248
193
77
209
95
108
600
83
1850
302
440
339
373
585
280
194
309
354
342
327
162
125
92
27.52
28.88
11.22
49.50
18.23
13.67
0.00
16.64
63.66
28.95
22.28
23.00
11.37
33.40
28.50
37.73
31.13
0.00
27.02
3.77
10.32
31.97
51.66
48.85
2.99
72.58
0.00
29.81
63.66
25.67
1.30
41.91
32.77
53.89
36.95
13.57
51.88
41.09
39.50
36.82
28.19
48.06
18.83
23.83
29.21
D-8
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
401003
401004
401005
401010
401888
401999
402001
402002
402003
402004
402005
402006
402007
402008
402009
402010
402011
402012
402013
402014
402015
402016
402017
402018
402019
402020
402021
402022
402023
402024
402025
402026
402099
402888
402900
402999
403001
403002
403003
403010
403011
403012
403888
403999
404001
Stack
Height
(ft)
56
0
45
0
38
10
38
37
32
33
34
39
38
40
40
61
38
32
43
56
34
85
45
45
33
34
27
46
32
36
40
45
0
35
55
38
29
41
0
30
43
59
22
34
29
Stack
Diameter
(ft)
2.72
0.00
1.30
0.00
2.47
0.00
3.97
2.82
2.23
2.58
2.77
2.88
2.25
2.26
2.59
2.78
3.12
2.67
2.89
3.78
1.33
3.66
2.58
2.81
3.61
3.17
2.62
2.54
4.96
3.74
2.92
2.02
0.00
3.08
4.06
2.94
3.55
2.97
0.00
4.10
9.87
6.92
4.17
6.01
3.50
Exit Exit
Temperature Velocity
(°F) (ft/s)
101
0
120
0
124
70
125
178
123
92
116
112
163
254
133
187
199
138
277
124
420
119
299
476
75
106
88
88
74
78
116
153
0
116
370
112
92
73
77
96
82
355
108
76
97
26.
0.
44.
0.
31.
0.
26.
29.
35.
28.
31.
33.
34.
25.
37.
34.
34.
1.
42.
28.
36.
38.
29.
50.
44.
33.
54.
40.
65.
31.
36.
16.
0.
26.
37.
35.
1.
0.
0.
1.
0.
4.
2.
6.
0.
57
00
36
00
75
00
08
93
92
05
51
41
25
71
70
50
22
46
80
19
81
29
41
86
20
32
83
76
05
18
42
99
00
30
78
31
57
01
00
10
51
62
42
06
84
D-9
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
404002
404003
404004
405001
405002
405003
405004
405005
405006
405007
405008
405888
406001
406002
406003
406004
406888
407004
407008
407016
407020
407032
407036
407040
407044
407048
407052
407056
407060
407064
407068
407076
407080
407084
407158
407172
407176
407180
407208
407220
407228
407232
407816
407820
407832
Stack
Height
(ft)
25
54
20
37
32
29
36
38
37
28
18
26
20
21
12
12
15
26
27
27
47
28
29
27
32
44
27
28
34
31
31
35
36
25
0
34
31
35
33
37
29
22
45
54
18
Stack
Diameter
(ft)
3.
1.
3.
2.
2.
2.
2.
3.
3.
2.
0.
]_ _
2.
11.
0.
0.
2.
1.
1.
3.
8.
1.
5.
2.
1.
5.
2.
2.
4.
0.
2.
3.
2.
2.
0.
2.
3.
22.
2.
1.
3.
0.
11.
7.
1.
08
53
19
32
41
13
43
15
45
45
69
73
27
24
32
31
91
46
36
40
86
94
27
06
94
95
74
88
09
75
18
27
56
48
00
47
30
35
52
78
04
00
39
22
28
Exit Exit
Temperature Velocity
(°F) (ft/s)
121
95
87
327
244
178
297
198
237
73
75
106
123
92
66
68
119
131
107
90
389
124
115
116
266
238
83
89
113
79
111
108
93
135
0
131
78
72
73
73
85
77
504
556
112
2
1
20
39
40
40
38
42
36
13
0
44
10
4
0
0
9
0
29
1
0
0
1
0
1
0
0
1
5
0
0
0
0
0
0
0
0
0
0
11
0
0
10
8
0
.15
.21
.27
.88
.76
.56
.93
.54
.76
.63
.00
.20
.98
.75
.10
.00
.97
.01
.97
.16
.15
.07
.55
.12
.76
.07
.41
.69
.91
.01
.03
.08
.08
.22
.00
.13
.01
.02
.09
.52
.39
.00
.90
.15
.01
D-10
-------
Appendix D.
DEFAULT STACK PARAMETERS OBTAINED FROM OZONE TRANSPORT
ASSESSMENT GROUP FOR SUBSTITUTION OF MISSING DATA
sec
407848
407860
407864
407872
407999
408999
490001
490002
490003
490004
490005
490900
490999
501001
501002
501004
501005
501006
501007
501900
502001
502002
502003
502005
502006
502900
503001
503002
503005
503006
503007
503008
503900
625400
Stack
Height
(ft)
8
34
16
16
28
20
70
33
24
11
33
52
34
167
0
0
93
3
18
79
62
19
83
64
25
64
58
51
91
38
48
19
75
86
Stack
Diameter
(ft)
0.36
0.57
0.17
0.34
3.40
3.73
2.61
1.89
19.37
2.63
2.13
9.72
2.58
5.79
0.00
0.00
3.98
0.00
2.67
4.62
2.87
5.11
3.28
2.91
0.70
4.02
3.33
2.78
4.19
3.24
2.57
2.40
3.36
4.83
Exit Exit
Temperature Velocity
(°F) (ft/s)
81
87
100
216
169
213
148
120
451
81
170
1232
122
462
0
69
395
77
111
487
760
1317
174
762
96
418
652
439
526
593
238
81
811
177
0.
0.
0.
40.
0.
3.
26.
38.
9.
0.
19.
31.
26.
50.
0.
0.
37.
0.
0.
17.
31.
12.
0.
39.
0.
49.
32.
25.
35.
63.
9.
28.
24.
38.
00
00
00
53
98
95
00
82
43
04
57
76
35
91
00
00
28
00
24
05
25
98
00
01
00
08
29
03
02
71
21
15
05
70
D-ll
-------
APPENDIX E
PARAMETERS RELATING TO THE FATES OF
SELECT ATMOSPHERIC POLLUTANTS
-------
APPENDIX E - TABLE OF CONTENTS
E. PARAMETERS RELATING TO THE FATES OF SELECT ATMOSPHERIC
POLLUTANTS E-1
E.I GAS-PHASE ORGANIC REACTIONS E-1
E.I.I Photolysis of Formaldehyde and Acetaldehyde E-2
E.I.2 Summary of Atmospheric Half-lLves E-2
E.2 REACTIONS OF POLYCYCLIC AROMATIC HYDROCARBON (PAH) .. . . E-4
E.3 WET AND DRY DEPOSITION E-5
E.4 GAS-PARTICLE PARTITIONING E-7
E.5 REFERENCES E-14
E-i
-------
APPENDIX E - LIST OF TABLES
Table E.I Parameters Used in Determining the Atmospheric Fate of Hazardous Gas Phase
Pollutants E-8
Table E.2 Average 24 Hour OH, NO3, and O3 Concentrations in a Moderately Polluted
Atmosphere E-l 1
Table E.3 Estimation of Photo-induced Decay Rate Constants for PAHs Under Summer and
Winter Conditions E-l 1
Table E.4 Comparison of the Cumulative Deposition of Particles During 100 Seconds
by Diffusion and Gravitational Settling E-l 1
Table E.5 Particle Phase Size Distributions of High Molecular Weight PAH at 298K E-12
Table E.6 Particle Phase Size Distributions of Trace Metals at 298K, latm Pressure E-12
Table E.7 Predicted Partitioning of PAHs Based on Equation 21 Model E-12
Table E.8 Predicted Partitioning of Dioxins and Dibenzofurans Based on Equation 22
Model E-13
Table E.9 Partitioning Potential of Other Hazardous Air Pollutants E-13
E-ii
-------
APPENDIX E
PARAMETERS RELATING TO THE FATES OF SELECT ATMOSPHERIC
POLLUTANTS
This appendix contains a summary of a literature survey conducted by Fletcher, et al.
(1997) The purpose of this literature survey was to provide best estimates of some parameters
used in the determination of the fates of selected hazardous air pollutants (HAPs). These data
would then serve as inputs to EPA's Industrial Source Complex Short Term Model (ISCST3).
An estimate of the atmospheric half-lives for gas-phase organic compounds reacting with
hydroxyl radicals (OH), the nitrate radical (NO3), and ozone (O3) is presented first. The
photolysis of formaldehyde and acetaldehyde is then discussed. This is followed by a discussion
of polycyclic aromatic hydrocarbon (PAH) half-lives as developed from reactions in sunlight.
Parameters used in calculating wet and dry deposition velocities, namely particle density and
size, molecular diffusivities in air and water, and Henry's law constants are then presented.
Finally, particle-gas phase distribution for semi-volatile contaminants (PAHs and trace metals
such as chromium) is discussed.
E.I GAS-PHASE ORGANIC REACTIONS
Second order reaction rate constants at 298K for 22 volatile organic compounds from
reactions with OH, NO3, O3 are summarized in Table E.I. References used to generate this table
are given at the end of this report. Temperature dependent OH rate constants exist for a number
of the compounds in Table E.I. Generally, reaction rates will show an exponential temperature
dependence over a narrow temperature range and can be estimated by the Arrhenius equation
(Atkins, 1990)
-E /RT (Eq. 1)
k=A*e *
where k is the rate of reaction, Ea is the activation energy for the reaction (J mol"1), R is the gas
constant (8.31 J K"1 mol"1), and T is the temperature of the system in Kelvin. The coefficient A is
a preexponential factor that is independent of, or only slightly dependent on, the temperature of
the system. If the temperature dependence of the preexponential factor is not small, typically
with small activation energies, a different form relating temperature to the reaction rate needs to
be employed (Finlayson-Pitts, 1986)
In this form of the Arrhenius equation the preexponential factor is separated into a temperature-
independent constant, B, and a temperature-dependent constant, Tn, where the exponent n is a
number chosen to best fit the data and T is the temperature in Kelvin. Table E. 1 lists
temperature-dependent rate equations for the reaction of several hazardous contaminants with the
E-l
-------
OH radical. Although temperature dependent equations of the rate constant for reactions with the
NO3 radical and O3 exist, they are not provided in this report.
E. 1.1 Photolysis of Formaldehyde and Acetaldehyde
Rate constants for the photolysis of formaldehyde and acetaldehyde were estimated by
integrating the product of the absorption cross section, o (A), the quantum yield, (j) (A), and the
actinic flux, J (A), at each wavelength, A, (Finlayson-Pitts, 1986)
The photolysis rate constants listed in Table E.I were calculated at 40° latitude for wintertime
(January 1) and summertime (June 1) conditions for three times of the day: 9 am, 12 noon, and 3
pm. These values are based on absorption cross section, quantum yield, and the actinic flux data
available in the literature (Atkinson, 1989; Atkinson, 1997; and Finlayson-Pitts, 1986).
E.I.2 Summary of Atmospheric Half-lives
From the rate constant data in Table E.I, half-lives were computed for each pollutant.
For pollutants, here referred to as [A], that react with oxidants OH, NO3 or O3, the half-live (t,/2)
can be calculated by (Atkins, 1990)
ln(2) (Eq. 4)
1/2 k
A
where kA is the second order rate constant. For the gas-phase reactions of interest in this report,
[B] is the concentration of the oxidizing species, and fluctuates according to temperature and
level of contamination in the atmosphere by the oxidant. Average concentrations of oxidants,
[B], in a relatively polluted atmosphere are given in Table E.2. For a first order reaction such as
photolysis, the reaction rate depends only on the concentration of A. The equation to calculate
the half-life of a first order reaction is simply (Atkins, 1990)
In (2)
ti2 = —r— (Eq. 5)
Reaction with OH radicals is the primary loss pathway for the majority of the compounds
considered here. The NO3 radical readily photolyzes and will have a relatively low steady-state
concentration during daylight hours. Overnight it can be assumed that all OH radicals are
reacted, and the steady-state concentration is approximately zero. However, NO3 concentrations
increase overnight, and this becomes the main route of decay of atmospheric contaminants. O3
E-2
-------
can react with gas-phase compounds during the day and night, however, a closer inspection of
the ozone rate constants in Table E. 1 reveals that these reactions are, in general, extremely slow
and will usually have little effect on the overall half-lives of atmospheric contaminants.
The overall half-lives for selected HAPs were calculated from equation 6 and are listed in
Table E.I. It is cautioned that formaldehyde and acetaldehyde are also produced via
photochemical processes and hence only introducing a decay constant will underestimate
formaldehyde and acetaldehyde concentrations (Finlayson-Pitts, 1986)
1 1111
= + + +
1/2, overall 1/2, OH 1/2, N03 1/2, 03 1/2,phot
(Eq. 6)
E-3
-------
E.2 REACTIONS OF POLYCYCLIC AROMATIC HYDROCARBON (PAH)
PAHs on particles react via photo-induced processes with gas phase O3, NO2, nitric acid
(HNO3), and dinitrogen pentoxide (N205), however, reactions in sunlight are the fastest. In the
late spring through the early fall months, when the daily average total solar radiation ranges from
0.25 to 0.4 calories cm"2 min"1, photolysis reactions will be more important than O3-PAH
reactions. During the winter months, with a daily average O3 concentration of 0.02, O3 reactions
may become more important.
Estimates of the first order rate constants of PAH decay in this report were developed
from outdoor chamber studies of PAHs on wood soot particles (Kamens, 1988). From these rate
constants half-lives were estimated for summer and winter time conditions and are tabulated in
Table 3. In these chamber experiments a reduction in the rate of reaction is typically observed
after two half- lives. Although this is not always the case, reducing the rate constants by a factor
of 2 after two half-lives of reaction is recommended. Given the current very limited data base,
losses due to PAH reaction with O3 were not included in half-life estimates. Because a reduction
in the reaction rates entered in ISCST3 is not possible after two half-lives, ISCST3 users should
use the rates and half-lives as listed in Table E.3.
The rates of photolysis of PAHs are highly dependent upon the types of particles onto
which the PAH adsorbs. For combustion particles such as wood soot or diesel particles, the half-
life of benzo-a-pyrene (BaP) is predicted here. However, if the particle were dark fly ash, the
half-life of BaP is on the order of a few days rather than a few hours (Atkinson, 1990). The rate
constants and half-lives presented in this report are applicable to atmospheres that have
significant wood and diesel combustion particle emissions. We would expect that atmospheres
dominated by coal power plants and incinerator emissions would have PAH half-lives which are
considerably longer than those presented here (Behymer, 1988; Pennise, 1996; and Wehry,
1990).
E-4
-------
E.3 WET AND DRY DEPOSITION
Wet and dry deposition are important sinks for atmospheric pollutants such as PAHs and
trace metals bound to aerosol particles. Wet deposition is generally considered to be a more
efficient removal process relative to dry deposition (Steiger, 1989). Key parameters which
influence wet and dry deposition in the ISC are particle densities and diameters, Henry's law
constants, and molecular diffusivities of selected hazardous air pollutants.
Particle size. Particle size is a major factor in determining deposition velocities due to
the gravitational settling velocity. For example, although a larger percentage of PAHs are
associated with the smaller, high surface area fine particles (generally greater than 75%), the
deposition velocities of particle-bound PAHs are overwhelmingly controlled by the gravitational
settling velocity of the larger coarse particles (Suman, 1989).
Table E.4 illustrates the relation between particle size and cumulative deposition
(Finlayson-Pitts, 1986). Gravitational settling velocity is much more important than diffusion for
larger particles. The fine particles remain suspended in the atmosphere, can possibly travel much
longer distances, and participate in a large number of atmospheric reactions. In the absence of
wet deposition, submicrometer particles (<1.0 um) have atmospheric residence times that vary
between 100 - 1000 hours while particles with diameters greater than 1 um have much shorter
residence times on the order of 10-100 hours (Finlayson-Pitts, 1986).
Henry's law values. The rate of volatilization of the contaminant is dependent on the
value of its Henry's law constant (HLC) and is controlled by its molecular diffusion through air.
Henry's law values on the order of 10"7 atm m3 mol"1 indicate that the substance is relatively
nonvolatile. Henry's law values between 10"5 and 10"3 atm m3 mol"1 indicate that both the gas and
aqueous phases play a significant role in determining into which medium the compound will
partition. When HLC are relatively high (greater than 10"3 atm m3 mol"1) the compound is only
slightly soluble (Lyman, 1996). Table E.I lists Henry's law constants for several atmospheric
pollutants.
Molecular diffusivity. Molecular diffusion is the net transport of a molecule through a
gaseous or liquid medium by Brownian motion. The molecular diffusivity has units of cm2 sec"1.
A comparison of experimentally determined molecular diffusivities to calculated diffusivities for
this report gave deviations of less than 10% in most cases (Schwarzenbach, 1993). Table E.I lists
the molecular diffusivity of select atmospheric pollutants at 298K in both air and water, Da]r and
Dw, respectively, as calculated by simplified equations which only require the molecular mass,
m, to be known (Schwarzenbach, 1993)
0-
2.7x10-* (Eq.8)
E-5
-------
Particle size distributions, Miguel and Friedlander (1978) measured the concentration of
high molecular weight PAHs (specifically BaP and coronene) adsorbed onto several different
sizes of particles for two temperatures. Table E.5 quantifies the high molecular weight PAH
concentration to particle size distribution.
Trace metal distribution. Relatively little work has been published concerning the
partitioning of trace metals, or more specifically the partitioning of chromium, onto particles.
Whitby (1977) reported approximate total particulate mass size distributions which have been
used by modelers to estimate chromium and other trace metal particle size distributions (U.S.
EPA, 1997). Such a distribution is given in Table E.6. Despite the lack of data related to
chromium particle size distributions, the assumed values used by modelers closely match
experimental results of Steiger et al. (Steiger, 1989) for the distributions of lead and vanadium in
the particle phase.
E-6
-------
E.4 GAS-PARTICLE PARTITIONING
The phase in which a compound exists in the atmosphere is largely dependent on its
vapor pressure. Compounds with low volatility, such as large PAHs (6 rings), exist almost
exclusively in and on particulate matter (i.e., in the "particle phase"), whereas highly volatile
compounds, such as small PAHs (2 rings, i.e., naphthalene), remain mostly in the gas phase.
However, semi-volatile compounds, which have ambient vapor pressures of approximately 10"
to 10"7 torr, demonstrate significant partitioning between the gas and particle phases. The most
straightforward model for predicting the partitioning of PAHs was presented by Yamasaki (1982)
log Cg =log - = --+B (Eq-9)
C /TSP K T
P P
where Cg is the gas-phase concentration (ng/m"3), Cp is the particle-phase concentration (ng/m"3),
TSP is the concentration of total suspended particulates (ng/m'3), Kp is the equilibrium partition
coefficient, T is the temperature (K), and A and B are empirically determined parameters. Table
E.9 shows the values of A and B determined by Yamasaki for samples taken in Osaka, Japan
(17). In order to determine the percent of PAH mass present on particulate matter, a value of 25
ug m"3 was used for TSP. This value represents a reasonable urban particle load (U.S. EPA,
1993).
Yamasaki's model can only be used when the values of A, B, Kp, and T are known.
Unfortunately the coefficients A and B have not been determined for many hazardous pollutants.
Therefore, a slightly more complicated model for predicting gas-particle partitioning has been
proposed by Pankow and Bidleman (1992)
o
logK =mrlogpL + br (Eq 10)
where p is the sub-cooled liquid vapor pressure (torr) for the compound of interest at a given
temperature, and n\ and br are empirical parameters. This model requires the determination of p
values at different temperatures. Various methods for estimating p values have been proposed
(1,100). Values of mr = -1.15 and br = -9.70 have been reported by Pankow and Bidleman (1992)
for dioxins and dibenzofurans. The resulting Kp values, and the percent in the particle phase at
298K and TSP = 25 ug/m~3 are shown in Table E.7. Values of p were taken from Eitzer and
Kites (1989).
Table E.8 classifies other hazardous air pollutants between those that have the potential to
partition between the gas and particle phases and those that exist almost exclusively in the gas
phase. This classification is based on the hazardous air pollutants' vapor pressure and a
simplified partitioning model (U.S. EPA, 1993).
E-7
-------
Table E.I Parameters Used in Determining the Atmospheric Fate of
Hazardous Gas Phase Pollutants
Compound
ace t aldehyde
acrolein
acrylamide
aery lonit rile
benzene
bis(2-
ethylhexyl)
phthalate
1,3 -
butadiene
carbon
tetrachloride
chloroform
1,4-
dichlorobenze
ne
1,1-
dichloroethen
e
1,2-
dichloropropa
ue
k0H
1 .6x10 J1
2 .OxlO'11
nr
4 . 8x10 12
1.2xlO'u
nr
6 .7x10'"
<5xlO'16
1.0x10-"
3 .2x10'"
S.lxlO-12
S4 .4x10
temp dependent kOH
5.6xlO-12*exp(310/T)
240 < T < 530
nr
nr
nr
7 . 57xlO-12*exp(-529/T)
T <, 325
nr
1.48xlO'11*exp(448/T)
250 < T < 425
-------
Table E.I Parameters Used in Determining the Atmospheric Fate of
Hazardous Gas Phase Pollutants (continued)
Compound
1,3-
dichloroprope
ne
ethyl
acrylate
ethylene
dibromide
ethylene
dichloride
ethylene
oxide
formaldehyde
hydrazine
methyl
chloride
quinoline
styrene
1,1,2,2-
tetrachloroet
hane
tetrachloroet
hylene
]r
J*-OH
nr
nr
2 .3xl(Tu
2 .2x10 "
SxlCT12
1 .0x10 ]1
6.5x10 "
4 .3x10 14
nr
SxlO'12
nr
1 .7x10'"
temp dependent kOH
nr
nr
nr
nr
nr
8.8xlO-12*exp(25/T)
240 < T < 300
nr
1.8xlO-12*exp(-1115/T)
240 < T < 300
nr
nr
nr
9.4xlO-12*exp(-1200/T)
300 < T < 420
^N03
nr
nr
nr
nr
nr
5,8xlO-16
nr
io-17
nr
1. 5x10 13
nr
-------
Table E.I Parameters Used in Determining the Atmospheric Fate of
Hazardous Gas Phase Pollutants (continued)
Compound
1,1,2-
trichloroetha
ne
trichloroethy
lene
vinyl
chloride
*OH
3x10-"
2.2xlCr12
6.6x10 12
temp dependent kOH
5 .0x10 "*exp(445/T)
230 < T < 420
1.63±0.22* (T/300)2 M*
exp[ (70 + 55) /T]
295 < T < 850
1 . 14x10 "12*exp( 1045 /RT)
299 < T < 426
]r
^NOj
nr
2.9x10-"
4xlQ-16
k 03
nr
<5xlO 20
nr
V
^phot
winter
na
na
na
summer
na
na
na
tl/2
winter
267 d
28 d
11 d
summer
22 d
3 d
1 d
dif fusiv-
ity in air
0.065
0.0644
0 .1054
dif fusiv
-ity in
water
8.46x10-'
8.36xlO-6
1.43x10 5
Henry' s
law
constant
Pa mj mol"1
901
95.5
8207a
"nr" = no recommended value.
"na" = does not react in sunlight or no experimental data reported.
"nc" = not computed.
a at low ionic strengths and 293-298 K.
E-10
-------
Table E.2 Average 24 Hour OH, NO3, and O3 Concentrations
in a Moderately Polluted Atmosphere
radical
OH
NO3
03
temperature (K)
concentration (molecules cm"3)
summer winter
1.2 xlO6
2.4xl08
l.lxlO12
324.8
1.0 xlO5
1.2 xlO8
6.8x10"
284.3
reference
48
48
49
50
Table E.3 Estimation of Photo-induced Decay Rate Constants for PAHs
Under Summer and Winter Conditions
PAH
cyclopenta( c , d)pyrene
benz(a)anthracene
chrysene and triphenylene
benzo(b)fluoranthene
benzo(k)fluoranthene
benz(a)pyrene
indeno(l,2,3-cd)pyrene
benzo(ghi)perylene
CpC
BaA
Chry
BbF
BkF
BaP
Ind
BshiP
Summer
k^Csec'1) t,.,'(hr)
1.66xlO"4
8.8xlO'5
3.3xlO'5
2.62xlO"5
3.35X10'5
7.62xlO'5
4.65xlO'5
3.73 xlO'5
1.2
2.2
5.8
7.3
5.7
2.5
4.1
5.2
Winter
k^fsec'1) t,,(hr)
3.23xlO'5
8.5xlO'6
2.63X10'6
4.27xlO'6
3.7xlO'6
8.52xlO'6
1.65xl06
2.95xlO~6
6.0
22.7
73.0
45.1
52.0
22.6
116.7
65.4
tr2 = In^/k^,. Computed rate constants were divided by a factor of two to account for PAH formed on higher
temperature particles.
Table E.4 Comparison of the Cumulative Deposition of Particles During 100 Seconds
by Diffusion and Gravitational Settling3
Diameter
urn
0.001
0.01
0.1
1.0
10
100
Cumulative Deposition
Diffusion Gravitational Settling
(number cm'2) (number cm'2)
2.5
0.26
2.9E-2
5.9E-3
1.7E-3
5.5E-4
6.5E-5
6.7E-4
8.5E-3
0.35
31
2500
a Assume unit particle densities and deposition onto a horizontal surface from unit aerosol concentrations. Adapted
from Finlayson-Pitts and Pitts (1986).
E-ll
-------
Table E.5 Particle Phase Size Distributions of High Molecular Weight PAH at 298K
Compound
High molecular weight PAH
(MW >220 g mol'1)
Particle Diameter (um)
0.26
70-75%
0.26-1.0
10-15%
1.0-4.0
5-10%
>4.0
5%
Table E.6 Particle Phase Size Distributions of Trace Metals at 298K, latm Pressure
Compound
Trace metals
Particle Diameter (um)
0.1
20%
0.3
50%
1.0
20%
3.0
5%
10.0
5%
Table E.7 Predicted Partitioning of PAHs Based on Equation 21 Model
Compound
phenanthrene &
anthracene
methylphenanthrene
& methylanthracene
fluoranthene
pyrene
benzo(a)fluorene &
benzo(b)fluorene
chrysene,
benz(a)anthracene,
& triphenylene
benzofluoranthene
benzo(a)pyrene &
benzo(e)pyrene
number
of rings
3
3
4
4
4
4
5
5
A
4117
3365
4421
4183
4554
5826
5693
4864
B
21.45
18.46
21.52
20.52
21.49
24.89
23.24
19.99
Kp (m3 ug-1)
2.3xlCT5
6.7xlO'5
2.0x1 Q-4
3.4xlO'4
6-lxlO'4
4.5xlO'3
7.2x1 0-2
2.1x10-'
percent of PAH mass
associated with
particles at 298K and
TSP = 25 ug m'3
0.057
0.17
0.51
0.84
1.50
10.1
64.1
84.0
E-12
-------
Table E.8 Predicted Partitioning of Dioxins and Dibenzofurans
Based on Equation 22 Model
No. of chlorine atoms
4
5
6
7
8
p (ton) at 298K
7.2xlO-7tol.7xlO'6
I.lxl0-7to4.8xl0-7
3.1xlO-8to9.8xlO'8
7.6xlO'9to 1.4xlO'8
1.9xlO-9to2.0xlO-9
Kp(m3ng-')
2.3xlO'3 to 8.6x10-"
2.0xlO-2to3.7xlO-3
8.6xlO-2to2.3xlO-2
4.3x10'' to 2.1x10-'
2.1 to 2.0
percent of compound
mass hassociated with
particles at 298K and TSP
= 25 ug m"3
5.5 to 2.1
33.4 to 8.4
68.3 to 36.4
91. 6 to 84.3
98.2 to 98.1
Table E.9 Partitioning Potential of Other Hazardous Air Pollutants
Hazardous air pollutants that may partition between
gas and particle phases or exist exclusively in the
particle phase
Hazardous air pollutants that exist almost exclusively in
the gas phase
arsenic compounds
beryhum compounds
cadmium compounds
chromium compounds
manganese compounds
nickel compounds
lead compounds
mercury compounds
bis(2-ethylhexyl)phthalate
coke oven emissions
acetaldehyde
acreolein
acrylonitrile
benzene
1,3-butadiene
carbon tetrachloride (tetrachloromethane)
chloroform (trichloromethane)
ethylene dichloride (1,2-dichloromethane)
formaldehyde
methylene chloride (dichloromethane)rrichloroethylene
tetrachloroethylene (perchloroethylene)
vinyl chloride (chlorothene)
1,4-dichlorobenzene (p-dichlorobenzene)
ethylene dibromide (1,2-dibromoethane)
1,1,2,2-tetrachloroethane
acrylamide
1,3 -dichloropropane
1,1 -dichloroethene (vinyhdene chloride)
1,2-dichloropropane (propylene dichloride)
ethyl acrylate
ethylene oxide
hydrazine
methyl chloride (chloromethane)
quinoline
styrene
1,1,2-trichloroethane
E-13
-------
E.5 REFERENCES
Atkins, P. W., 1990. Physical Chemistry 4th ed, W. H. Freeman and Company, New York.
Atkinson, R., Baulch, D. L., Cox, R. A., Hampson, Jr., R. F., Kerr, J. A., Troe, J., 1989. J. Phys.
Chem. Ref. Data 18,881.
Atkinson, R., 1990. Final Report to California Air Resources Board, Contract No. A732-107
March.
Atkinson, R., Baulch, D. L., Cox, R. A., Hampson, Jr., R. F., Kerr, J. A., Rossi, M. J., 1997. J.
Phys. Chem. Ref. Data 26, 521.
Behymer, T. D., Kites, R. A., 1988. Environ. Sci. and Technol. 22, 1311.
Eitzer, E. D., Kites, R. A., 1989. Environ. Sci. Technol. 23, 1389.
Finlayson-Pitts, B. J., Pitts, Jr, J. N., 1986. Atmospheric Chemistry: Fundamentals and
Experimental Techniques, John Wiley and Sons, New York.
Fletcher, K., Strommen, M., Kamens, R., 1997. Final Report on Parameters Relating to the Fates
of Select Atmospheric Pollutants. Prepared for Jawad S. Touma, U. S. Environmental
Protection Agency, Office of Air Quality Planning and Standards, Research Triangle
Park, NC.
Kamens, R. M., Guo, Z., Fulcher, J. N., Bell, D. A., 1988. Environ. Sci. & Technol. 22, 103.
Lyman, W. J., Reehl, W. F., Rosenblatt, D. H., 1996. Handbook of Chemical Property
Estimation Methods, American Chemical Society, Washington, DC.
Miguel, A. H., Freidlander, S. K., 1978. Atmos. Environ. 12, 2407.
Pankow, J. F., Bidleman, T. F., 1992. Atmos. Environ. 26A, 1071.
Pennise, D., Kamens, R. M., 1996. Environ. Sci. Technol. 30, 2832.
Schwarzenbach, R. P., Gschwend, P. M., Imboden, D. M., 1993. Environmental Organic
Chemistry John Wiley and Sons, New York.
Steiger, M., Schultz, M., Schwikowski, M., Naumann, K., Dannecker, W., 1989. J. Aerosol Sci.
20, 1229.
Suman, D., 1989. Aerosol Sci and Technol. 10, 131.
E-14
-------
U. S. EPA (Environmental Protection Agency), 1993. Simultaneous Control ofPM-10 and
Hazardous Air Pollutants: Rationale for Selection of Hazardous Air Pollutants as
Potential Paniculate Matter or Associated with Paniculate Matter at Source Conditions',
Research Triangle Park, NC. EPA/452-R-93-013.
U. S. EPA (Environmental Protection Agency), 1997. Personal Communication with Dr. Russ
Bullock. Office of Research and Development. Research Triangle Park, NC.
Wehry, E. L., 1990. "Adsorption and photochemical transformations of PAHs sorbed on coal fly
ash." Atmospheric Processes of Organic Toxic Pollutants and Their Role in current
Environmental Issues Workshop, Canada Institute for Research in Atmospheric
Chemistry.
Whitby, K. T., 1977. Atmos. Environ. 12, 135.
Yamasaki, H., Kuwata, K., Miyamoto, H., 1982. Environ. Sci. Technol. 16, 189.
E-15
-------
APPENDIX F
ADDITIONAL INFORMATION ON HOW TO USE OZIPR
-------
APPENDIX F - TABLE OF CONTENTS
F. ADDITIONAL INFORMATION ON HOW TO USE THE OZIPR F-1
F.I THE OZIPR INPUT FILE F-1
F.2. SAMPLE SIMULATION!: HOUSTON TEXAS , SUMMER F-10
F.3 RUNNING ADDITIONAL 24 HOUR SIMULATIONS F-12
F.4 REFERENCES F-18
APPENDIX F - LIST OF EXHIBITS
Exhibit F.I { SAMPLE INPUT FILE houtx2a.inp } F-13
Exhibit F.2 { SAMPLE CB-4 REACTIVITIES FILE cb4_form.rea } F-14
Exhibit F.3 { SAMPLE BOUNDRY FILE cb4sum.BOUND_a } F-15
Exhibit F.4 { SAMPLE EMISSIONS (MASS) FILE cb4sum.MASS } F-16
Exhibit F.5 { SAMPLE METEOROLOGY FILE summer.met } F-17
F-i
-------
APPENDIX F
ADDITIONAL INFORMATION ON HOW TO USE THE OZIPR
F.I THE OZIPR INPUT FILE
To run OZIPR, one needs an input file, filename.inp, which is a text file consisting of a
series of OPTIONS and COMMANDS. (See Sample Input File.) The program is started by
typing "oziprfilename.inp". A standard output file, filename.out is created.
The input file can contain the names of external files to be included in the text of the
input file. The character "@" followed by the name of the external file to be included will read
the external file text line by line into its position in the input file. There are three required
external data sets that are included at the beginning of the input file: a chemical mechanism, a
zenith set, and reactivities. Any of the input can be isolated in external files if so desired, or
simply included in the input file.
General rules for the input file are:
6. Usually, the first four letters of a command option name are significant.
7. Comments can be included following the first character " ! " or bracketed between
(}or().
8. Top level commands are followed by a " > ", the option commands, and ends
with a " < ".
9. Case is not important for this version.
Sample files are included at the end of this section.
A line by line description of the sample input file follows.
Line 1 of the sample input file is:
@cb4jsox_form.mec
The chemical mechanism external file includes information about the number of carbons
for some organic species, and the list of reactions with rate constants used for the model run. For
the purpose of determining secondarily produced formaldehyde, the chemical mechanism must
differentiate between the primary formaldehyde and the secondary formaldehyde produced in
situ. Secondarily produced formaldehyde is added as a new species, FRMS, while the symbol
used for formaldehyde, FORM, now will refer to only primarily produced formaldehyde.
F-l
-------
The number of carbons for organic species is included under "cnum", and must include
the new species FRMP:
cnum = par, 1.0, eth, 2, ole, 2, tol, 7,
xyl, 8, form, 1, ald2, 2, nr, 1,
frmp, 1;
The chemical mechanism used in this study is the standard Carbon Bond IV mechanism
with minor modifications to allow it to differentiate between primary and secondary
formaldehyde production. This includes:
1. In the reaction equations that produce formaldehyde (formaldehyde on the right
side of the equation), the symbol FORM is changed to FRMP in reactions 45, 46,
49, 50, 51, 56, 57, 58, 59, 60, 61, 62, 70, 71, 76, and 77 in the reactions list.
2. For each of the reactions where formaldehyde is depleted (formaldehyde on the
left side of the equation), similar equations must be added that differ only by
depleting FRMP instead of FORM. These additional equations can simply be
added at the bottom of the list of chemical reactions in the mechanism file, as
equations 86-90.
Line 2 of the sample input file is:
@jspec640.zen
The second line refers to a zenith angle file, specific for the Carbon-Bond Mechanism IV
chemical mechanism file, and is used in interpolation of photochemical rates. This file will not
need to be edited.
Line 3 of the sample input file is:
@cb4_form.rea
This line refers to an external file containing the fractional speciations of VOC species and is
also specific to the Carbon-Bond Mechanism IV chemical mechanism. This file will need to be
edited to include FRMP in background air. This file can also be edited if it is desirable to alter
the VOC speciation.
There are 4 columns under reactivity. The first is the organic species name. The last
three columns indicate the fraction of total VOC that comprises each species in: 1) the initial
mixture ad/or emissions (VOC entered under CALC, EMIS, or MASS); 2) the aloft VOC
(VOCALOFT under TRANSPORT); and 3) the transported surface-layer VOC (VOCSURFACE
under TRANSPORT). Initial VOC concentrations will be defined as VOCSURFACE air (in this
case), and therefore initial speciation is defined in the last column of the file. The FRMP to
FORM ratio assumed for a 6AM initialization is 9:1. To approximate the ratio for a different
F-2
-------
initialization hour, look at the seasonal tables of hourly FORM and FRMP for Houston. The
initial ratio of FRMP to FORM will not be crucial, because of the relatively short lifetime of
formaldehyde.
BOUNDARY >
REACTIVITY =
par,
eth,
ole,
tol,
xyl,
form,
frmp,
ald2,
nr,
0.5640,
0.0370,
0.0350,
0.0890,
0.1170,
0.0210,
0.0000,
0.0520,
0.0850,
0.4980,
0.0340,
0.0200,
0.0420,
0.0260,
0.0700,
0.0000,
0.0370,
0.2730,
0.4980,
0.0340,
0.0200,
0.0420,
0.0260,
0.0070,
0.0630,
0.0370,
0.2730;
Line 4:
MODIFY> ACC = 0.0001 ;<
This line modifies the accuracy in the tolerance of the numerical integration routine.
Lines 5-14:
TITLE > HOUSTON, TX Design Day Sim SUMMER <
PLACE>
CITY = HOUSTON, TX;
LAT = 29.8,
LONG = 95.2,
TZone = 5.0,
Year = 1988,
Month = 7,
Day = 16; <
TIME> 0600, 2900 <
These lines define the place and time for the run. TITLE is the title chosen for the run, up to 72
characters. PLACE is used in determining the zenith angle of the sun throughout the day. CITY
name can be up to 24 characters. LATITUDE is latitude in degrees north, and LONGITUDE is
the longitude in degrees west. TZONE is the local time zone, Eastern Daylight time = 4, Pacific
Daylight Time = 7. YEAR, is the year, MONTH is the number for month, and DAY is the day
of the month. TIME is the range of hours to run the simulation. In this example, the model will
run from 6 am on July 16 until 6 am on July 17.
F-3
-------
Line 16:
@summer.MET
Meteorology is defined in an external include file which includes pressure, mixing heights, air
temperature, and air moisture. This file should be edited for meteorology specific to the region
and time of year.
In this example, the hourly mixing heights were entered explicitly using the MIXING option:
MIX[24] =
598., 697., 797., 897., 996.,
1096., 1195., 1295., 1395., 1395., 1395., 1395., 1395., 1392., 1278.,
1106., 924., 738., 576.,
567., 562., 558., 562., 566.;
As these are needed at the beginning and end of each hour, the number of mixing heights should
equal one plus the number of hours for the simulation. If the number of values is less than this,
the last value entered will be used for the remaining hours.
Temperature is entered in a similar manner:
TEMP[24,K] =
296.7,298.0,300.1,301.8,303.1,
304.1, 304.9, 305.3, 305.5, 305.5, 305.1, 304.5, 303.6, 302.3, 301.1,
300.3,300.0,299.0,298.5,
298.0, 297.6, 297.2, 296.9, 296.7;
As for the mixing height, the number of values should equal one plus the number of hours for the
simulation. The units for temperature are entered as "K" for Kelvin, "C" for Centigrade, and "F"
for Fahrenheit.
Pressure is entered as a constant value of 29.75 inches of mercury for this example:
pres[in] = 29.75;
In this example, air moisture is entered as relative humidity:
RH[24] =
92.0, 86.5, 81.0, 75.5, 70.0,
64.5, 59.0, 60.8, 62.7, 64.5, 66.3, 68.2, 70.0, 71.8, 73.7,
75.5, 77.3, 79.2, 81.0,
82.8, 84.7, 86.5, 88.3, 90.2;
F-4
-------
The hourly values in average relative humidity are entered for the beginning and end of each
hour.
DILUTION option. In this example, the hourly mixing heights were entered explicitly using
the MIXING option:
MIX[24] =
598., 697., 797., 897., 996.,
1096., 1195., 1295., 1395., 1395., 1395., 1395., 1395., 1392., 1278.,
1106., 924., 738., 576.,
567., 562., 558., 562., 566.;
As these are needed for the beginning and end of each hour, the number of mixing heights
should equal one plus the number of hours for the simulation.
Temperature is entered in a similar manner:
TEMP[24,K] =
296.7, 298.0, 300.1, 301.8, 303.1,
304.1, 304.9, 305.3, 305.5, 305.5, 305.1, 304.5, 303.6, 302.3, 301.1,
300.3,300.0,299.0,298.5,
298.0, 297.6, 297.2, 296.9, 296.7;
The number of values input should equal one plus the number of hours for the simulation. If
the number of values is less than this, the last value entered will be used for the remaining hours.
The units for temperature are entered as "K" for Kelvin, "C" for Centigrade, and "F" for
Fahrenheit.
In this example air moisture is entered as RH (relative humidity):
RH[24] =
92.0, 86.5, 81.0, 75.5, 70.0,
64.5, 59.0, 60.8, 62.7, 64.5, 66.3, 68.2, 70.0, 71.8, 73.7,
75.5, 77.3, 79.2, 81.0,
82.8, 84.7, 86.5, 88.3, 90.2;
The hourly values in average relative humidity are entered for the beginning and end of each
hour. The total number of values equals the total simulation hours plus 1. If wnum is less than
this, the last value entered will be used for the remaining hours.
In line 17 of the input file, boundary conditions are defined in an external include file for the
sample input file:
Line 17:
m cb4sum.BOUND f
F-5
-------
Boundary conditions include the assumed ratios of NO2/NOx in any NOx entered as initial
conditions or as emissions. Also included are deposition velocities for selected chemical species,
and initial concentrations assumed for selected chemical species.
The TRANSPORT option allows predetermining the concentrations of selected species for
both surface and aloft air. The top level command here is BOUNDARY:
BOUND >
conditions....
IFRACTION_NO2, is the NO2/NOx fraction for the NOx in controllable emissions at the start
of the simulation.
IFRaction NO2 = .05;
Hourly deposition velocities in cm/s for NO2, O3, HNO3, H2O2, and PAN are entered using
the DEPOSITION option and the format:
DEPO [24] =
NO2 = .24, .36, .48, .54, .60,
.60, .60, .60, .60, .60, .54, .48, .36, .24, .24,
.24, .24, .24, .24,
.24, .24, .24, .24, .24,
TRANSPORT is used to define the initial concentrations for O3, NOx, VOC, and CO that are
due to transport.
TRANSPORT =
O3SURFACE = .021, { W=.021, SP=.033, SU=.024, A=.026 }
03ALOFT = 100,0.04,
VOCSURFACE = .000,
VOCALOFT = .030,
NOXSURFACE = .000,
NOXALOFT = .002,
COSURFACE = .000,
COALOFT = .500;
Surface layer concentrations entered under TRANSPORT can be used for the species as
initial conditions. Values are in ppm (ppmc for VOC). Input for O3ALOFT includes two
values, the first of which is the value in meters above which height the O3 ALOFT value should
be used. The second value is the O3 ALOFT concentration in ppm.
F-6
-------
VOC species entered under TRANSPORT will be speciated as designated in the boundary
conditions for REACTIVITY. These are species specific to the chemical mechanism used and
are not defined in this external file, but are defined in the external reactivity file, cb4_form.rea .
This external file was included in line 3 of the input file. There are 4 columns under reactivity.
The first is the organic species name. The last three columns indicate the fraction of total VOC
that comprises each species in the initial mixture ad/or emissions (VOC entered under CALC,
EMIS, or MASS), the aloft VOC (VOCALOFT under TRANSPORT), and the transported
surface-layer VOC (VOCSURFACE under TRANSPORT). So when initializing VOC using
VOCSURFACE, it is the last column under REACTIVITY that is accessed. From the external
file cb4_form.rea:
BOUNDARY >
REACTIVITY =
par,
eth,
ole,
tol,
xyl,
form,
frmp,
ald2,
nr,
0.5640,
0.0370,
0.0350,
0.0890,
0.1170,
0.0210,
0.0000,
0.0520,
0.0850,
0.4980,
0.0340,
0.0200,
0.0420,
0.0260,
0.0700,
0.0000,
0.0370,
0.2730,
0.4980,
0.0340,
0.0200,
0.0420,
0.0260,
0.0070,
0.0630,
0.0370,
0.2730;
Note that REACTIVITY in its external file is nested in its top level command, BOUNDARY.
Finally, the boundary file also can contain initial concentrations of species other than CO,
and O3 using the INITIALIZE option. If initial concentrations of O.O are desired, it is not
necessary to initialize any species this way. However, this option allows you to input realistic
non-zero concentrations for initializing a simulation. In this example, we have initialized the
concentrations of all species to equal the 6 a.m. concentrations from a previous day's simulation:
INIT =
no2 = 0.000044041,
no = 0.000000121,
o = 0.000000000,
no3 = 0.000000007,
isop - 0.000964346,
nr = 0.068530180,
apin = 0.000000000,
unkn = 0.000000000;
F-7
-------
In line 18 of the input file, an external file which defines the hourly emissions is included.
Line 18:
@ cb4sum.MASS_f
In this file, the hourly emission mass for up to 10 species is entered in units of kg/km2 per
hour. The species name is followed by its molecular weight in brackets. An initial
concentration is also entered (in units of ppm). For species other than VOC, NOX, and CO, the
initial concentration specified here is added on to concentrations specified in the INIT option.
After the concentration, hourly emission values are entered for each hour in the format:
MASS[24] >
VOC [ 14.50] = .5530,
3.936, 4.894, 6.044, 7.217, 7.746,
8.077,8.318,8.377,8.759,9.120,
9.282,9.198,7.868,6.221,5.035,
4.201,3.613,3.368,2.780,
2.528, 2.370, 2.432, 2.520, 2.883,
Line 19:
After adding the emissions via the external file, we finish with the CALCULATE option.
In this option, the initial concentrations due to emissions for VOC, NOX, and CO are input.
In this example we have already specified initial concentrations under the BOUNDARY options,
so we have set them equal to zero here.
VOC = 0.000;
NOX = 0.000;
CO = 0.000;
Also included under CALCULATE is the PRINT option, which is used to select species
concentrations to be output.
For this example, we output only concentrations for 5 species:
PRINT[FULL] =
NAMES [5] = form, frmp, O3, NO, NO2
CDUMP[5] = form, frmp,O3, NO, NO2
NODUMP;
-------
There are many other options available for use with OZIPR. However, these are not necessary
for this application and are not reviewed here. For further information, the Users Guide (Gery
and Grouse 1991) should be consulted.
F-9
-------
F.2. SAMPLE SIMULATION 1: HOUSTON TEXAS , SUMMER
The analysis will be specific to both a given city and season. Houston, Texas has been used
in the development of this process. A prototypical day has been determined for each season:
winter, spring, summer, and autumn. Seasonally averaged hourly meteorology data for each city
are used. For example, Houston's 8 AM temperature values for each spring day are averaged to
create the 8 AM temperature value used in the simulation. The emissions data for each season
are handled in a like fashion.
Files needed for this simulation include the input file:
houtxZa.inp
There are three external files specific to the cb-4 mechanism that must be made available:
cb4jsox_form.mec (the chemical mechanism file)
jspec640.zen (the zenith angle file)
cb4_form.rea (the reactiities file)
In addition, additional input for the simulation is included in these other external files:
summer.MET (meteorology)
cb4sum.BOUND_a ( boundary conditions )
cb4sum.MASS (emissions)
Local meteorology information is included in the file "summer.MET". Houston temperature
and mixing height values were taken from 5 year averages of the hourly data. (1987-1991).
Morning and afternoon relative humidities from "Comparative Climatic Data for the United
States" (NCDC, 1984) were used and data for all hours were linearly interpolated from the two
values.
Background concentrations are entered in the external file "cb4sum.BOUND_a". The
background concentrations for O3, VOC, NOX, and CO used in this simulation are from 1997
Photochemical Ambient Monitoring Site (PAMS) data. These species are initialized by entering
them as surface values of transported air:
03SURFACE = .024,
VOCSURFACE = .553,
NOXSURFACE = .000,
COSURFACE = .911,
Emissions for the simulation are entered in "cb4sum.MASS_a". For this simulation, biogenic
emissions for VOC and NOX were provided by running PC-Beis2 for Harris County, Texas
using temperature data from the ISCST3 model inputs. This was run for July 14, 1990. The PC
F-10
-------
Beis2 program and associated files can be obtained from the EPA SCRAM web pages
(http://www.epa.gov/scram001/index.htm). Daily anthropogenic emission values for NOX, CO,
and reactive organic gases (ROG) were extracted from the OTAG base emissions inventory for
Harris County Texas.
Once all the input is gathered, the model is begun by typing "ozipr.xfilename.inp", or in this
case, "ozipr.x houtx2a.inp".
F-ll
-------
F.3 RUNNING ADDITIONAL 24 HOUR SIMULATIONS
The results for determining secondary formaldehyde based on one 24 hour run can be very
dependent on initial conditions, and perhaps not provide the best estimate for secondary
formaldehyde. Additional consecutive 24 hour simulations may be performed to the point where
the 24 hour curve for formaldehyde values (primary and secondary) remains approximately the
same from day to day. The output file, "filename.oul", contains information about the species
concentrations throughout the run. The model can be run for several 24 hour interactions, or
until both primary and secondary formaldehyde concentrations have approximately the same
concentrations for each species at the beginning and end of the 24 hour run. In order to set up a
simulation to continue from the end of the last simulation, there will be some differences in the
in input. The boundary file in Simulation 1, cb4sum.BOUND_a, can be adapted for a second
run with the boundary file cb4sum.BOUND_b using the final species concentrations in the
output file, "houtx2a.out". First, the VOCSURFACE and NOXSURFACE concentrations under
TRANSPORT must be set to 0.0, as their constituents (organic species, NO and NO2) will be
initialized individually. O3 and CO from the output file are used to initialize O3 SURF ACE and
COSURFACE under TRANSPORT. All other species will be initialized under the INIT option
(also part of the boundary conditions). A new input file, houtx2b.inp, should be created that
includes the external boundary file cb4sum.BOUND_b. This process of using the final species
concentrations as input to a subsequent run can be continued for additional 24 hour periods.
F-12
-------
Exhibit F.I { SAMPLE INPUT FILE houtx2a.inp }
@/home/wendy/models/ozipr/mec/cb4jsox_form.mec
@/home/wendy/models/ozipr/mec/jspec640.zen
@/home/wendy/models/ozipr/mec/cb4_form.rea
TITLE > HOUSTON, TX Design Day Sim SUMMER <
MODIFY> ACC = 0.0001 ;<
PLACE>
CITY = HOUSTON, TX;
LAT = 29.8,
LONG = 95.2,
TZone= 5.0,
Year = 1988,
Month = 7,
Day =16;<
TIME> {100, 2400} 0600, 2900 <
@/home/wendy/models/ozipr/sims/summer.MET
@/home/wendy/models/ozipr/sims/cb4sum.BOUND_a
@/home/wendy/models/ozipr/sims/cb4sum.MASS_a
CALO
VOC = .000; { W=302, SP=427, SU=553, A=.507 }
NOX = .000; { W=.001, SP=.001, SU=000, A=000 }
CO =.000; { W=.716SP=..721 SU=.911, A=896 }
PRINT[FULL] =
NAMES[5] = form,frmp,o3,no,no2,
CDUMP[5] = form,frmp,o3,no,no2,
NODUMP;
{
PRINT[FULL] =
NODUMP;
END.
F-13
-------
Exhibit F.2 { SAMPLE CB-4 REACTIVITIES FILE cb4_form.rea }
boundary >
reac =
par,
eth,
ole,
tol,
xyl,
form,
frmp,
ald2,
nr,
0.5640,
0.0370,
0.0350,
0.0890,
0.1170,
0.0210,
0.0000,
0.0520,
0.0850,
0.4980,
0.0340,
0.0200,
0.0420,
0.0260,
0.0700,
0.0000,
0.0370,
0.2730,
0.4980,
0.0340,
0.0200,
0.0420,
0.0260,
0.0070,
0.0630,
0.0370,
0.2730;
{reactivity = site specific reactivity of the 6-9 am mix as nmoc fractions}
t — —
< {boun}
F-14
-------
Exhibit F.3 { SAMPLE BOUNDRY FILE cb4sum.BOUND_a }
BOUND> { SUMMER }
DEPO [24] =
NO2 = .24, .36, .48, .54, .60,
.60, .60, .60, .60, .60, .54, .48, .36, .24, .24,
.24, .24, .24, .24,
.24, .24, .24, .24, .24,
O3 = .30, .50, .60, .70, .80,
.80, .80, .80, .80, .80, .70, .60, .50, .30, .30,
.30, .30, .30, .30,
.30, .30, .30, .30, .30,
HNO3 = 2.60, 2.60, 3.00, 3.30, 3.50,
3.50, 3.50, 3.50, 3.50, 3.50, 3.30, 3.20, 3.00, 2.60, 2.60,
2.60, 2.60, 2.60, 2.60,
2.60, 2.60, 2.60, 2.60, 2.60,
H2O2 = 1.60, 1.70, 1.80, 1.90, 2.00,
2.00, 2.00, 2.00, 2.00, 2.00, 1.90, 1.80, 1.70, 1.60, 1.60,
1.60, 1.60, 1.60, 1.60,
1.60,1.60,1.60, 1.60,1.60,
PAN = .24, .36, .48, .54, .60,
.60, .60, .60, .60, .60, .54, .48, 36, .24, .24,
.24, .24, .24, .24,
.24, .24, .24, .24, .24;
IFRaction_NO2 = .050;
TRANSPORT =
O3 SURF ACE = .024, { W=.021, SP=. 033, SU=024, A=. 026 }
O3ALOFT = 100,0.04,
VOCSURFACE = .553, { W=.302, SP=427, SU=553, A=507 }
VOCALOFT = .030,
NOXSURFACE = .000, { W=.001, SP=.001, SU=.000, A=000 }
NOXALOFT = .002,
COSURFACE = .911, { W=.716 SP=..721 SU=.911, A=.896 }
COALOFT = .000;
{INITial Concentrations = }
F-15
-------
Exhibit F.4 { SAMPLE EMISSIONS (MASS) FILE cb4sum.MASS }
{cb4sum.MASS}
MASS[24] >
VOC [ 14.50] = .5530, {from Allan, beis,anthro}
3.936, 4.894, 6.044, 7.217, 7.746,
8.077, 8.318, 8.377, 8.759, 9.120,
9.282, 9.198, 7.868, 6.221, 5.035,
4.201,3.613,3.368,2.780,
2.528, 2.370, 2.432, 2.520, 2.883,
NO [30.00]= .0000000001, {from Allan, beis}
.0070, .0075, .0083, .0097, .0109,
.0115, .0125, .0129, .0129, .0125,
.0121,.0118, .0118, .0112, .0106,
.0103, .0100, .0100, .0094,
.0078, .0078, .0076, .0072, .0072,
NOX[ 46.00]= .0000000001, {from Allan, anthro}
4.702, 5.496, 5.896, 6.343, 6.650,
6.734, 6.802, 6.842, 6.984, 7.350,
7.451, 7.317, 6.686, 5.657, 4.751,
4.200,3.814,3.596,3.194,
2.940, 2.772, 2.864, 3.080, 3.604,
ISOP[ 68.13]= .0000010, {from Allan, beis}
0.484,0.822,1.122,1.899,2.567,
2.999, 3.585, 3.809, 3.692, 3.239,
2.669, 1.492, 0.133, 0.000, 0.000,
0.000, 0.000, 0.000, 0.000,
0.000, 0.000, 0.000, 0.000, 0.088,
CO [28.00]= .91100, {from Allan anthro}
13.995, 19.234, 21.852, 24.906, 27.148,
28.435, 29.318, 29.916, 31.692, 34.627,
35.833, 35.150, 29.333, 21.523, 15.068,
11.170, 8.331, 6.781, 4.949,
3.544, 2.094, 2.720, 3.570, 6.899;
F-16
-------
Exhibit F.5 { SAMPLE METEOROLOGY FILE summer.met }
MET> { start at proper hour, see TIME above }
MIX[24] =
{ 567., 562., 558., 562., 566., 598., 697., 797., 897., 996.,}
598., 697., 797., 897., 996.,
1096., 1195., 1295., 1395., 1395., 1395., 1395., 1395., 1392., 1278.,
1106., 924., 738., 576.,
567., 562., 558., 562., 566.;
TEMP[24,K] =
{ 298.0, 297.6, 297.2, 296.9, 296.7, 296.7, 298.0, 300.1, 301.8, 303.1,}
296.7, 298.0, 300.1, 301.8, 303.1,
304.1, 304.9, 305.3, 305.5, 305.5, 305.1, 304.5, 303.6, 302.3, 301.1,
300.3, 300.0, 299.0, 298.5,
298.0, 297.6, 297.2, 296.9, 296.7;
PRES[IN] = 29.75;
RH[24] =
{ 82.8, 84.7, 86.5, 88.3, 90.2, 92.0, 86.5, 81.0, 75.5, 70.0,}
92.0, 86.5, 81.0, 75.5, 70.0,
64.5, 59.0, 60.8, 62.7, 64.5, 66.3, 68.2, 70.0, 71.8, 73.7,
75.5, 77.3, 79.2, 81.0,
82.8, 84.7, 86.5, 88.3, 90.2;
F-17
-------
F.4 REFERENCES
Gery, M. W. and Grouse, R. R., 1991. "User's Guide for Executing OZIPR", U.S. Environmental
Protection Agency, Research Triangle Park, NC.
NCDC (National Climatic Data Center), 1984. "Comparative Climatic Data for the United
States", Asheville, NC.
F-18
-------
TECHNICAL REPORT DATA
(Please read Instructions on reverse before completing)
1 REPORT NO 2
EPA-454/R-99-021
4 TITLE AND SUBTITLE
Air Dispersion Modeling of Toxic Pollutants in Urban
Areas; Guidance, Methodology and Example
Applications
i AUTHOR ( S I
9 PERFORMING ORGANIZATION NAME AND ADDRESS
EC/R Incorporated
1129 Weaver Dairy Road
Chapel Hill, NC 27514
i: SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Emissions, Monitoring & Analysis Division
Research Triangle Park, NC 27711
15 SUPPLEMENTARY NOTES
EPA Work Assignment Manager: Jawad S.
3. RECIPIENT'S ACCESSION NO
5. REPORT DATE
July 1999
6 PERFORMING ORGANIZATION CODE
8 PERFORMING ORGANIZATION REPORT NO
10 PROGRAM ELEMENT NO
11 CONTRACT/GRANT NO
EPA Contract No. 68D98006
13 TYPE OF REPORT AND PERIOD COVERED
Final Report
14 SPONSORING AGENCY CODE
Touma
I€ ABSTRACT
This report provides guidance on how to use the Industrial Source Complex Short Term
(ISCST3) model to estimate concentrations of air toxic pollutants for urban-wide
analyses. Urban areas contain major sources and numerous smaller, area sources. As a
result modeling analyses posses special challenges. Section 1 provides guidance and
recommendations on specific issues for urban-wide analyses of air toxics. Section 2
provides an overview of two applications of the Industrial Source Complex Short Term
(ISCST3) model to urban-wide studies. Appendices A and B provide demonstration of the
methodology for two example applications.
KEY WORDS AND DOCUMENT ANALYSIS
a DESCRIPTORS
Air Pollution
Air Quality Dispersion Models
Meteorology
Air Toxics
Urban Area Modeling
18 DISTRIBUTION STATEMENT ,, . ,
Release Unlimited
b IDENTIFIERS/OPEN ENDED TERMS c COSATI Field/Group
.
' *
13 ' SECURITY CLASS (Report) , ^ 21. NO OF PAGES
' Unclassified 284
L ' r • **
' • • '
20 SECURITY CLASS (Page! 22. PRICE
Unclassified
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
------- |