SOME APPLICATIONS OF MODELS
TO AIR TOXICS IMPACT ASSESSMENTS
by
Daniel J. McNaughton
Marshall A. Atwater
Richard J. Londergan
TRC Environmental Consultants, Inc.
800 Connecticut Boulevard
East Hartford, CT 06108
EPA Contract No. 68-02-3886
Prepared for
Monitoring and Data Analysis Division (MD-14)
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
May 1986
-------
DISCLAIMER
"This report has been reviewed by the Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, and approved for publica-
tion as received from TRC Environmental Consultants, Inc. Approval does
not signify that the contents necessarily reflect the views and policies
of the U.S. Environmental Protection Agency, nor does mention of trade
names or commercial products constitute endorsement or recommendation for
use. Copies of this report are available from the National Technical
Information Service."
-------
ACKNOWLEDGEMENT
This report is extensively based on work conducted by D. McNaughton, M.
Atwater and R. Londergan of TRC Environmental Consultants, Inc., East
Hartford, CT. That work was originally funded by the Department of the
Interior's (DOI) CERCLA 301 Project through an Interagency Agreement (IAG No.
RW14931395-01-0) with the Environmental Protection Agency; S. Coloff was the
DOI Project Officer. The work was performed under Contract No. 68-02-3886
with D. Layland as the EPA Project Officer. The original report was prepared
for use by authorized officials conducting natural resource damage assessments
under the Comprehensive Environmental Response Compensation and Liability Act,
Section 301(c) and 43CFR Part 11, Natural Resources Damage Assessments. The
original report is available as:
Type B Technical Information Document
Application of Air Models to Natural Resource Injury
Department of Interior
Washington, D.C.
-11-
-------
PREFACE
Air toxics are of increasing concern to Federal and State air pollution
control agencies. As with criteria air pollutants (e.g., S02, TSP, CO,
Oj, NOX» Pb), ambient impact assessments for air toxics frequently must be
based on dispersion models. However, air toxics present unique problems in
mathematically simulating the emissions characteristics and the atmospheric
transport, transformation, and removal of these pollutants. While models are
available for many toxic pollutants and emissions situations, frequently they
are not widely known or tested.
The purpose of this report is to identify models that are available for
toxics impact assessments and factors that should be considered in selecting
models for specific applications. There is no claim as to the merits of
individual models or that the list of models is comprehensive. This report
only provides information that may be considered useful to air pollution
control programs concerned with air toxics and should not be construed as
providing regulatory guidance.
-111-
-------
TABLE OF CONTENTS
SECTION PAGE
ACKNOWLEDGEMENT ii
PREFACE iii
1.0 INTRODUCTION
2.0 AIR TOXICS RELEASES AND MODELS 3
2.1 Atmospheric Impacts . . . 3
2.2 The Role of Models in Air Toxics Impact
Assessments 4
2.3 Characteristics of Models 4
2.3.1 Source Types and Emissions Modules 4
2.3.2 Dispersion Models 7
2.3.3 Chemical Transformation and Deposition Models . . 8
3.0 APPLICATIONS OF MODELS TO AIR TOXICS IMPACT ASSESSMENTS 9
3.1 Alternative Data Sources 9
3.2 Model Selection 9
4.0 AVAILABLE MODELING TECHNIQUES 19
4.1 Complete Models 19
4.1.1 Environmental Protection Agency Models on the
UNAMAP System 19
4.1.2 Alternative Models 26
4.2 Model Components and Alternative Formulations ... 33
4.2.1 Source/Emission Models 34
4.2.2 Chemical Conversions 37
4.2.3 Deposition 37
4.2.4 Plume Rise 40
4.2.5 Peak Concentration Levels 40
4.2.6 Downwash 41
5.0 MODEL EVALUATIONS 42
5.1 Model Limitations and Uncertainties - Models for
Neutrally Buoyant Emissions 42
5.1.1 Factors Limiting Model Accuracy 43
5.1.2 Operational Uncertainty 45
5.1.3 Reliability of Model Components 50
5.2 Model Limitations and Uncertainties - Air Toxics
Models 51
6.0 SUMMARY 53
-IV-
-------
REFERENCES PAGE
R.I MODELS 56
R.2 FIELD EXPERIMENTAL PROGRAMS 64
R.3 MODEL EVALUATIONS 67
LIST OF FIGURES
FIGURE PAGE
5-1 Comparison of Highest Observed and Predicted Relative
Concentration Values for Paired One-Hour Concentrations . . 47
5-2 Cumulative Frequency Distributions of the Highest Observed and
Predicted Relative Concentration Values for One-Hour
Concentrations Using the CRSTER Model 48
LIST OF TABLES
TABLE PAGE
1-1 Potential Components of Source and Initial Dispersion Models 2
3-1 Information Requirements for Air Toxics Impact Assessments . 10
3-2 Decision Tree to Select Models for Air Toxics Impact
Assessments 13
4-1 Characteristics of Alternative Models 27
4-2 Source Characterization 38
-v-
-------
1.0 INTRODUCTION
The assessment of air toxics impacts requires analyses of a wide range of
pollutant release types and atmospheric phenomena over varying time and
distance scales. Simulation of all aspects of such releases is not possible
using a single model or modeling approach. Often, the issues addressed in
modeling assessments are at the limits of our current knowledge of atmospheric
dispersion phenomena as related to source release characteristics.
Models used for air toxics impact assessments consist of components for
simulation of source emissions, transport and dispersion, chemical
transformations, and deposition. Table 1-1 lists a sample of components that
might be considered in simulating the emission of material and initial
dispersion of pollutants. Requirements for simulation of transport,
dispersion, and deposition are not included in this table. These may require
significantly more detail than is commonly considered. This suggests the need
for caution in applying the techniques and the need for expert assistance in
the more complicated assessments.
This report consists of five technical sections. Section 2.0 provides a
summary description of air toxics releases and characteristics of models.
Section 3.0 identifies a method for selection of appropriate modeling
techniques by outlining the components of assessment methods for
characterizing different types of releases. Section 4.0 provides a catalog of
available modeling techniques. Section 5.0 presents a summary of previous
performance evaluations of models and a statement on potential modeling
uncertainties. Section 6.0 provides a summary and general comments on
modeling techniques. An extensive list of references provides additional
information on models.
-1-
-------
TABLE 1-1
POTENTIAL COMPONENTS OF SOURCE AND INITIAL DISPERSION MODELS
Structure Effects
Flow obstructions
Multiple stack plume rise modifications
Stack tip downwash
Wake cavities and effects
Plume Rise
Buoyant
Momentum jet
Moist
Directional
Flares
Time dependent
Evaporation/Vaporization
Pool spreading
Flashing
Aerosol formation (two phase)
Chemical Transformations
Vapor overflow
Pool vaporization (heat/mass transfer)
Initial Dispersion
Cold gases
Heavy molecular weight gases
Transitional buoyancy gases
Buoyancy induced initial mixing
Gravity spreading (slumping)
Plume liftoff
Particle settling
Uncontrolled Releases (fires)
Buoyant plume rise
Chemical formation
-2-
-------
2.0 AIR TOXICS RELEASES AND MODELS
This section summarizes types of impacts associated with air toxics
releases. The second subsection discusses the role of models in impact
assessments. The third subsection discusses the types of models available.
2.1 Atmospheric Impacts
Air toxics releases can have either prolonged (chronic) or acute impacts
on public health and welfare. A prolonged impact is one that does not have an
immediate effect on health or welfare, but is the result of accumulated
exposure. These impacts are generally related to continuous chemical releases
over long periods which cause a persistent low level' concentration. Such a
long-term impact might result from a situation such as the annual accumulation
of a chemical which originates as a slowly evolving volatile emission from a
material storage area or material carried in windblown soil particles. An
acute impact is one related to short-term, high concentrations of a pollutant
resulting in an immediate effect on health or welfare. These short-term
impacts are the result of a single event or chemical release causing a
one-time concentration. This type of impact might include a plant upset in
which an air toxic is released over a period of minutes.
Spatial scales of interest vary by chemical. In general, for long-term
impacts, concentration levels may be lower and source receptor distances
longer. For short-term impacts, high concentrations are typically of greatest
interest and as a result the source/receptor separation distance is generally
small. Also, different time scales in combination with the need to simulate
source emissions, transport and dispersion, and potential chemical
transformations and deposition, require a number of different modeling
techniques. These techniques are summarized in Section 2.3.
-3-
-------
2.2 The Role of Models in Air Toxics Impact Assessments
The role of models in impact assessments in most cases is to contribute to
the general body of knowledge on ambient concentrations. Model selection for
assessments is dependent on processes being simulated and data availability.
The following sections discuss results generated by different types of
modeling requirements to simulate physical and chemical processes of
interest. Data availability is crucial in determining the uncertainty in
information. Poor data will not support good models and, as a result, a weak
data base may require that a simplified model be used. From simplified
models, less precise conclusions are usually drawn for the impact assessment.
2.3 Characteristics of Models
Complete dispersion models are those which simulate all important aspects
of contaminant behavior from the source to the receptor. Such models are
generally constructed from collections of component modules or submodels.
Submodels simulate individual processes affecting the fate of pollutants but
do not provide the entire answer to the source/receptor relationship.
The basic model components for short- and long-term impacts can be
characterized as source and emissions, transport and dispersion, and chemical
transformation and deposition modules.
2.3.1 Source Types and Emissions Modules
Sources are typically distinguished as being point, line, area or volume
sources, with emissions specified in units of rate, rate per line length, or
rate per area. Assessment of air toxics impacts often requires complex
descriptions of sources through emission modules. Some of the characteristics
of sources identified in this study are as follows:
-4-
-------
Point Sources - Point sources of air toxics are frequently characterized
as being continuous (steady state) or time dependent. Traditional air quality
models use continuous point source parameters in determining steady state
solutions to diffusion equations. Source inputs to the models include a
constant emission rate and source release height, and parameters for
simulating plume rise (i.e. exhaust flow rate, exit temperature and stack
diameter). Phenomena simulated are typically limited to stack tip downwash
for low velocity releases, buoyancy induced dispersion, and momentum plume
"rise equations for high velocity, neutral buoyancy plume rise. Models in the
Guideline on Air Quality Models (U.S. EPA, 1978) are adequate to simulate
these releases.
Air toxics impact studies could potentially require simulation of time
dependent emissions, emissions and plume rise from fires and flares, and
directional releases. Time dependent releases are often from pressure drops
associated with leaks and venting from pressure vessels and pipelines.
Emissions modules for establishing the release rates for pipelines are
available (e.g. Hanna and Munger, 1983; Blewitt, 1985). Emission rates from
fires are not generally described in the literature due to the difficulty in
identifying the general chemical composition of combustion products. Plume
rise for fires and flares can be determined by buoyant and jet plume rise
equations. The simulation of directional plume rise is most important for
high momentum jets and is sometimes simulated by estimating the vertical
momentum component.
Line Sources - Line sources are most commonly used to describe mobile
source effects. Some of the applications for which line source descriptions
are applicable for air toxics are the spraying of agricultural pesticides and
herbicides and vapor releases over berms.
-5-
-------
Area Sources - Typical area source applications for air quality modeling
include simulations of large areas of poorly defined sources and areas of
fugitive dust generation due to industrial and agricultural activity or wind
blown dust. Area sources in air toxics modeling may be of the following types:
emissions for small or poorly defined sources located in urban
areas or industrial complexes (e.g. fugitive emissions).
evaporation from liquid spills in confined areas (e.g. dikes).
vaporization from spills of liquefied gases in confined areas.
heavy gas leaks in confined areas.-
evaporation from liquids and liquified gases spreading on water.
emissions from waste disposal operations, (e.g. landfills, land
treatment, surface impoundments, waste water treatment units).
Modeling volatile liquids and liquefied gases in unconfined spills on
water provide the most complicated cases as they require simulation of both
spreading and evaporation in a time dependent emissions module. These modules
are described in more detail in Section 4.2.1.
Volume Sources - In air quality modeling, volume sources are seldom
considered in explicit solutions to diffusion equations. Sources are
sometimes defined to represent the volume generated in the wake zones of flow
obstructions such as buildings. Downwash models are important in air toxics
applications, but more common is the requirement to define an initial cloud
resulting from the rapid vaporization of liquefied gases. In refrigerated
liquefied gas spills on land, rapid vaporization results from soil surface
heating; vaporization models show a decrease in vaporization rate with time.
For pressurized liquefied gases, a rapid vaporization (flashing) due to
adiabatic decompression occurs. For example, in a liquid chlorine spill from
a typical pressurized storage vessel, 20 percent of the gas may be flashed.
The turbulence of this gas entrains air to build an almost instantaneous cloud
-6-
-------
of perhaps ten times the volume of the gas. The situation is more complex if
a pressurized storage vessel is damaged below the liquid level. In this
instance, rapid boiling causes the generation of a gas and aerosol cloud which
could include the major portion of liquid in the tank.
2.3.2 Dispersion Models
Air quality dispersion models for assessment purposes are typically of the
continuous emission source, Gaussian dispersion type that handle buoyant or
neutrally buoyant gases and aerosols. For air toxics applications, two
modifications to this approach are sometimes needed. First, simulations for
acute impacts may require instantaneous or puff solutions to diffusion
equations and second, negatively buoyant plumes may require simulation using a
gravity spreading or slumping model.
Continuous emission source dispersion models, as represented by the EPA
Guideline on Air Quality Models, incorporate model components including:
- point, line and area equations
- momentum and buoyant plume rise
- building and stack tip downwash
- variations in averaging periods
- multisource and multiple pollutant capabilities
- limited terrain and deposition capabilities
Neutrally buoyant instantaneous sources are simulated with an alternative
and generally accepted puff solution to the diffusion equations. Models are
available representing time dependent sources by sequences of steady state
puffs or plume elements for simulating dispersion in temporally and spatially
changing wind fields. Both continuous and instantaneous models are used in
conjunction with slumping models to simulate heavy gas dispersion.
-7-
-------
2.3.3 Chemical Transformation and Deposition Models
Chemical transformations and deposition can be very important in air
toxics impact modeling when the initial generation of toxic byproducts and the
materials losses en route between a source and receptor are being simulated.
The initial chemical reactions are typically determined on a chemical specific
basis as part of an emission specification. For example, several
tetrachloride compounds react with water vapor in air to form HC1 droplets.
An initial assumption for modeling may be that the reactions go to completion
prior to dispersion. Modeling of chemical transformations downwind might also
take the form of a linear transformation rate which, in simplified Gaussian
models, would involve an exponential loss (or gain) term.
Deposition losses are difficult to simulate due to our limited current
understanding of this phenomena. Estimates of wet deposition require a
further understanding of whether the scavenging takes place in the subcloud
layer or by incorporation of the contaminant in the cloud system. Proper
specification of precipitation rate is very important.
Dry deposition in simplified models is often represented by an exponential
decrease in concentration at a rate determined by a gravitational settling
velocity (large particles) or a deposition velocity (gases and fine
aerosols). These parameters although called velocities are simply the ratios
of mass flux of contaminant to the local concentration and are determined
empirically. The parameters are site and pollutant specific and little
information is available for air toxics applications.
-8-
-------
3.0 APPLICATIONS OF MODELS TO AIR TOXICS IMPACT ASSESSMENTS
The sub-sections which follow discuss the use of dispersion models in air
toxics impact assessments. Section 3.1 discusses sources of input data.
Section 3.2 describes the basis for model selection.
3.1 Alternative Data Sources
Data for dispersion modeling must be representative of the conditions
which govern emissions and transport. Dispersion simulations are subject to
uncertainties (section 5.1) even when the appropriate data are used. Poor
selection of data or data errors increase these uncertainties. As a result,
data selection can be as important as model selection.
Data typically required for air toxics modeling are listed in Table 3-1.
On-site meteorological data are preferred for modeling analyses. Alternative
data sources follow:
Source data:
- review of process data for similar incidents
- determination of chemical characteristics from general references
- examination of reports of the behavior of similar chemicals
- evaluation of processes: mass balances
Meteorological data:
- identification of local representative observation sites
- data collection from the National Climatic Data Center
- review of historical data bases for feasible worst case
meteorological conditions
3.2 Model Selection
The description of model characteristics provided in Section 2.0 suggests
the complexity which accompanies an impact assessment for air toxics
emissions. This section provides techniques for selection of models. Models
considered here are of an intermediate complexity in that they provide
quantitative results but at the same time are less sophisticated than current
-9-
-------
TABLE 3-1
INFORMATION REQUIREMENTS FOR
AIR TOXICS IMPACT ASSESSMENTS
Release
- source type (stack, flare, uncontrolled spill on water, etc.)
- chemical characteristics
- release characteristics
- release height
- release rates
- visible cloud dimensions
- operational characteristics/description of the source
- duration
Meteorology
- wind speed
- wind direction
- stability parameters (wind variation, lapse rate, etc)
- day/night
- mixing height
- temperature
- cloud cover
- date
- precipitation
Site Characteristics
- obstacles to the flow at the release site
- spill surface (land/water)
- characteristics of dispersion route (terrain, snow surface,
roughness, etc.)
General Description
- time sequence of release events
-10-
-------
research grade models. This level of model complexity was selected to provide
a category of techniques useful to a large number of users. Research grade
models are recommended when specific source attribution estimates are required
in combination with a thorough understanding of processes involved. Use of
those models is beyond the intent or scope of this report.
The basis for model selection in this guideline is a decision tree. The
decision tree attempts to identify model components required to simulate a
particular effect and those components that should be included in complete
dispersion models. The decision tree considers simulation of various
materials released through different mechanisms (e.g., liquefied gases
released on land in confined areas versus a pressurized liquefied gas tank
failure). Since many of the physical principles involved in dispersion are
shared among release types, the same dispersion models are selected when
conditions allow. The decision tree directs the user to subsections of
Section 4.0 where complete models or modules are described.
Table 3-2 presents the decision tree used in model selection.
Instructions are provided which carry the user through the model selection
scheme by considering relevant questions on source and release type, data
evaluation and selection, selection of a dispersion model or submodule, and
selection of parameters to run the model.
The decision tree in Table 3-2 is used by following the sequence of
instructions listed. The tree begins with a decision on whether the impact is
associated with long- or short-term releases. The selection establishes a
pathway of subsequent decisions and instructions. Instructions through the
tree are of three types. The first type is a directional instruction to
proceed to another numbered instruction and has the form of a statement "go to
(number)". The second form of instruction is an implicit continuation which
simply means that if there is no directional instruction, proceed to the next
-11-
-------
numbered instruction. The third type of instruction directs the user to
specific subsections of Section 4.0 where some model or module is described.
After identifying the information in Section 4.0, the user returns to the
instruction number to continue the path.
-12-
-------
TABLE 3-2
DECISION TREE TO SELECT MODELS
FOR AIR TOXICS IMPACT ASSESSMENTS
DETERMINATION OF RELEASE TYPE
Characterize the impact as either short- or long-term.
1. Short-term - go to 9
2. Long-term - go to 3
LONG-TERM MODELING APPROACHES
3. It is assumed that impacts associated with long-term exposures involve
sources with normal releases, persistent fugitive releases, or fugitive
releases from either evaporation of volatile material or generation of
particles as wind blown dust. Select emission factors based on the
following categories:
3a. Resuspension {wind erosion) generation of particles (Section 4.2.1).
3b. Evaporation of volatile materials (Section 4.2.1).
3c. Specified releases of materials from controlled sources or process
fugitives. Use measurements of emissions or process estimates.
4. Evaluate available data from the period of emissions. Dispersion models
typically require information on wind speed, wind direction and
atmospheric stability representative of the source. In addition, data on
atmospheric mixing height are very important if the source/receptor
separation distance is in excess of a few kilometers. For long-term
impacts, data can be supplied to models as sequences of short-term
(hourly) data or as joint frequency distributions of wind direction and
speed and stability. Lack of on-site data requires an analysis of
off-site data resources for representativeness. Feasibility studies
determining order of magnitude estimates are possible with judiciously
selected or worst case wind data.
5. Select model. EPA has recommended models to simulate dispersion of
neutral buoyancy gases in the atmosphere. These models simulate
dispersion from point, area and line sources with options to include
multiple receptors and sources, varying averaging periods and emissions.
Options in the models include building wake effects, momentum and buoyant
plume rise equations, terrain effects and deposition. The models are
typically steady-state Gaussian models using straight line trajectories
which limits their range of applicability. Variable trajectory models
may be needed for assessments over large source/receptor separation
distances or, in areas of locally varying wind fields (e.g. areas
influenced by terrain obstacles or local circulations; Section 4.1.2).
-13-
-------
TABLE 3-2
(Continued)
DECISION TREE TO SELECT MODELS
FOR AIR TOXICS IMPACT ASSESSMENTS
5. (Cont.)
Model selection from the UNAMAP series of models involves an assessment
of available information in relation to model inputs. Major differences
in models are in the categories of:
- terrain inputs
- single versus multiple sources
- source types
- frequency distribution 'versus sequential inputs for
meteorological data.
6. Select model input parameters and review model assumptions. Model
documentation should be reviewed to determine if all required variables
are available and that model assumptions realistically describe the
nature of the air toxics impact.
7. Modify models if necessary. Modifications to the model may include:
7a. Chemical transformations (Section 4.2.2).
7b. Alternative deposition techniques (Section 4.2.3).
7c. Alternative plume rise equations (e.g., flares: Section 4.2.4).
8. Perform model simulations.
STOP
-14-
-------
TABLE 3-2
(Continued)
DECISION TREE TO SELECT MODELS
FOR AIR TOXICS IMPACT ASSESSMENTS
SHORT-TERM MODELING APPROACHES
9. Determine the probable form of the released material:
9a. Gas, particles and/or aerosol - go to 10
9b. Liquid* - go to 14
9c. Liquefied gases - go to 20
9d. Dense gas - go to 23
10. Determine if the released material should be simulated as a continuous,
or an instantaneous, source. For gas releases, unlike liquids, the
distinction between instantaneous and continuous releases is often
clear. Various definitions are available to distinguish between the two
classes but a simple rule of thumb might be to consider a release
instantaneous if the release time is much less than the travel time
between the source and the receptor.
lOa. If releases are continuous - go to 15.
lOb. Go to 11 for instantaneous releases.
11. Instantaneous source models allow simulation of dispersion from point,
area and volume sources using steady state solutions to diffusion
equations. Area and volume sources are represented by virtual point
sources or by spatial integrations of the equations. These expressions
simplify the representation of building wake effects, and initial
dilution for pressurized releases.
In situations of complex wind fields and/or large suspected
source/receptor distances variable trajectory models are advised if
sufficient wind data are available.
lla. Straight line models - Section 4.1.1
lib. Variable trajectory models-Section 4.1.2
12. Modify the models. Assumptions in the selected models should be
evaluated for the specific application. Modifications may be required to
simulate factors such as deposition.
13. Perform an analysis. Instantaneous source models typically available
provide semi-empirical results in the form of mean concentrations. In
realistic concentration fields, peak concentrations can exceed this mean
instantaneous value by a large factor (see Section 4.2.5).
STOP
*Liquids are defined as fluids having boiling points above ambient
temperatures at ambient pressures for this application.
-15-
-------
TABLE 3-2
(Continued)
DECISION TREE TO SELECT MODELS
FOR AIR TOXICS IMPACT ASSESSMENTS
RELEASES OF LIQUIDS
14. Simulations of liquid spills are accomplished with standard dispersion
models after accounting for the generation of vapors by evaporation.
14a. Instantaneous unconfined spills on water (Section 4.2.1) - gravity
spreading of the liquid continues until halted by evaporation. Go to
11.
14b. Continuous spills on water (Section 4.2.1) - gravity spreading of
the liquid establishes a steady state spill radius controlled by
evaporation rate and spill rate. Go to 15.
14c. Confined spills on land (Section 4.2.1). Go to 15.
15. Evaluate available data for the period of emissions. Dispersion models
typically require information on wind speed, wind direction and
atmospheric stability representative of the source. In addition, data on
atmospheric mixing height are very important if the source/receptor
separation distance is in excess of a few kilometers. Lack of on-site
data requires an analysis of other off-site data resources for
representativeness. Feasibility studies determining order of magnitude
estimates are possible with judiciously selected or worst case wind data.
16. Select model. EPA has recommended models to simulate dispersion of
neutral buoyancy gases in the atmosphere. These models simulate
dispersion from point, area and line sources with options to include
multiple receptors and sources, varying averaging periods and emissions.
Options in the models include building wake effects, momentum and buoyant
plume rise equations, terrain effects and deposition. The models are
typically steady-state Gaussian models using straight line trajectories
which limits their range of applicability. If impact is being estimated
for large source/receptor separation distances or in areas of locally
varying winds it may be necessary to consider variable trajectory models
(Section 4.1.2).
Model selection from the UNAMAP series of models requires an assessment
of available information in relation to required model inputs. Major
differences in models are in the categories of:
- terrain inputs
- single versus multiple sources
- source types
17. Select model input parameters and review model assumptions. Model
documentation should be reviewed to determine if all required variables
are available and that model assumptions realistically describe the
impact.
-16-
-------
TABLE 3-2
(Continued)
DECISION TREE TO SELECT MODELS
FOR AIR TOXICS IMPACT ASSESSMENTS
18. Modify models if necessary. Modifications to the model may include:
18a. Chemical transformations (Section 4.2.2).
18b. Alternative deposition techniques (Section 4.2.3).
18c. Alternative plume rise equations (e.g., flares: Section 4.2.4)
19. Perform model simulations.
STOP
LIQUEFIED GAS SPILLS
20. Liquefied gases represent the most complicated source of air toxics due
to the effects of pressurization and/or refrigeration required for
storage.
Simulation of these gases involves models which are currently under
development and may not be adequately tested for impact assessments
(Section 4.1.2). As an approximation, the first step in analysis is a
determination of whether the gas is pressurized or refrigerated. If the
gas is both refrigerated and pressurized, the most immediate effects will
result from the pressurization and this factor would be simulated either
concurrently or before the effects of the refrigeration are considered.
20a. Pressurized liquefied gas - go to 21
20b. Refrigerated liquefied gas - go to 22
PRESSURIZED GASES
21. Models for pressurized gases should distinguish the source of the gas
spill (see sub-models for pressurized gases in Section 4.2.1).
21a. Spills above the liquid level of a tank, a flashing module will
estimate the fraction of material immediately vaporized due to
adiabatic decompression. This gas is often treated as an
instantaneous release and is simulated as in Step 11 if the
buoyancy is near neutral with respect to air. Heavy gases
resulting from partial refrigeration, cooling by decompression or
those with high molecular weights should be simulated using Step
23.
21b. Spills below the liquid level of a storage vessel will be
simulated as in step 21a except for a provision to estimate gas
droplet formation due to rapid boiling in the vessel. The added
mass is often assumed to vaporize rapidly and it is added to the
initial plume volume prior to simulation as a neutral gas under
Step 11 or a dense gas under Step 23.
-17-
-------
TABLE 3-2
(Continued)
DECISION TREE TO SELECT MODELS
FOR AIR TOXICS IMPACT ASSESSMENTS
22. Simulation of refrigerated liquefied gas vaporization is described in
Sections 4.2.1. Dispersion modeling of the gas is in the category of
22a. Negatively buoyant gas clouds - go to 23, or
22b. Buoyant or neutrally buoyant source models - go to 11
DENSE GAS DISPERSION
23. Models for dense gas dispersion are described in Section 4.1.2. Typical
models simulate initial phases of dispersion by slab models where plume
spreading is a result of gravitational forces until plume heating (in the
case of a cold gas) and entrainment of environmental air dilute the plume
sufficiently that atmospheric turbulence takes over as the dominant force.
in dispersion.
24. Simulations of dense gas dispersion can follow general procedures
outlined for short-term exposures, but models specific to negatively
buoyant gas clouds (Section 4.1.2) should be substituted for standard
Gaussian dispersion models from the UNAMAP series.
24a. continuous pool evaporation - go to 15
24b. Initial or instantaneous vaporization - go to 11
-18-
-------
4.0 AVAILABLE MODELING TECHNIQUES
Air toxics impact assessments require a wide variety of modeling
techniques. This section is provided to identify some of these techniques and
to summarize model components. Section 4.1 describes complete models used
previously in assessment studies. The models included were derived primarily
for air quality applications, but models for chemical and fuels safety
assessments are also included. Section 4.2 reviews the techniques used in
simulating some of the complex aspects of air toxics dispersion.
4.1 Complete Models
The decision tree in Table 3-2 attempts to include all aspects of
dispersion modeling to assure that releases of differing types are simulated
using a complete complement of appropriate techniques. The method of analysis
uses complete models designed to simulate a particular type of release. In
keeping with this idea, the following two subsections describe complete models
currently available for simulations. The first section describes EPA models
in the UNAMAP modeling system. The second section describes models which were
identified through the literature search. The latter category includes
instantaneous and heavy gas dispersion models.
4.1.1 Environmental Protection Agency Models on the UNAMAP System
EPA provides numerous models for air quality simulations through the
User's Network for Applied Modeling of Air Pollution (UNAMAP). Currently
UNAMAP Version 5 provides FORTRAN programs and model users manuals for over
thirty models and data processors supporting the models. UNAMAP models are
divided into two classes, guideline and non-guideline. Guideline models have
been evaluated by EPA and deemed appropriate for simulations in regulatory
applications as reported in the Guideline on Air Quality Models (EPA, 1978).
-19-
-------
The Guideline (currently under revision) also provides descriptions and
suggestions for a variety of model related topics such as model and data
selection, uncertainty of models, regulatory applications of models and
details such as simulating chemical transformations, deposition and plume
rise. Non-guideline models in the UNAMAP system are typically those
undergoing development or evaluation, those with duplicate capabilities or
those with features not yet endorsed by EPA for regulatory applications.
Models in UNAMAP are generally Gaussian dispersion models for continuous
emissions sources. As a class, they are single or multisource and include
plume rise equations. Some of the models include provisions for downwash,
terrain considerations, deposition through settling, and trapping by inversion
layers. Also identified are models which simulate variable trajectories to
show the effects of temporally and spatially varying wind fields. For air
toxics assessments, the UNAMAP models are appropriate for simulations of
continuous or guasicontinuous releases of materials that behave in the
atmosphere as neutrally buoyant tracers. These conditions cover a majority of
air toxics releases.
The following are brief descriptions of the guideline and non-guideline
models appropriate to air toxics assessments:
Guideline Models:
RAM Gaussian-Plume Multiple Source Air Quality Algorithm - This short-term
Gaussian steady-state algorithm estimates concentrations of stable
pollutants from urban point and area sources. Hourly meteorological data
are used. Hourly concentrations and averages over a number of hours can
be estimated. Briggs plume rise is used. Pasguill-Gifford dispersion
equations with dispersion parameters developed for urban areas are used.
Concentrations from area sources are determined using the method of Hanna,
that is, sources directly upwind are considered representative of area
source emissions affecting the receptor. Special features include
determination of receptor locations downwind of significant sources and
determination of locations of uniformly spaced receptors to ensure good
area coverage with a minimum number of receptors.
-20-
-------
Reference - Turner, D.B, and Novak, J.H., 1978: User's Guide for RAM.
Vol. I. Algorithm Description and Use. EPA-600/8-78-016A (NTIS Accession
Number PB-294 791), Vol. II. Data Preparation and Listings.
EPA-600/8-78-016B {NTIS Accession Number PB-294 792). U.S. Environmental
Protection Agency, Research Triangle Park, NC. (November 1978).
CRSTER - This algorithm estimates ground-level concentrations resulting
from up to 19 collocated elevated stack emissions for an entire year and
prints out the highest and second-highest 1-hr, 3-hr, and 24-hr
concentrations as well as as the annual mean concentrations at a set of
180 receptors (5 distances by 36 azimuths). The algorithm is based on a
modified form of the steady-state Gaussian plume equation which uses
empirical dispersion coefficients and includes adjustments for plume rise
and limited mixing. Terrain adjustments are made as long as the
surrounding terrain is physically lower than the lowest stack height
input. Pollutant concentrations for each averaging time are computed for
discrete, non-overlapping time periods (no running averages are computed)
using measured hourly values of wind speed and direction, and estimated
hourly values of atmospheric stability and mixing height.
References - Monitoring and Data Analysis Division, 1977: User's Manual
for Single-Source (CRSTER) Model. U.S. Environmental Protection Agency,
Research Triangle Park, NC EPA-450/2-77-013. (NTIS Accession Number
PB-271 360).
CDM - The Climatological Dispersion Model determines long-term (Seasonal
or Annual) quasi-stable pollutant concentrations at any ground level
receptor using average emission rates from point and area sources and a
joint frequency distribution of wind direction, wind speed, and stability
for the same period.
Reference - Busse, A.D., and Zimmerman, J.R., 1973: User's Guide for the
Climatological Dispersion Model. Environmental Monitoring Series,
EPA-R4-73-024, (NTIS Accession Number PB-227-346). U.S. Environmental
Protection Agency, Research Triangle Park, NC. 131 pp. (December 1973).
CDMQC - This algorithm is the Climatological Dispersion Model (CDM)
altered to provide implementation of calibration, of individual point and
area source contribution lists, and of averaging time transformations.
The basic algorithms to calculate pollutant concentrations used in the CDM
have not been modified, and results obtained using CDM may be reproduced
using the CDMQC.
Reference - Brubaker, K.L., Brown, P., and Cirillo, R.R., 1977: Addendum
to User's Guide for Climatological Dispersion Model. Prepared by Argonne
National Laboratory for the U.S. Environmental Protection Agency, Research
Triangle Park, NC. EPA-450/3-77-015. (NTIS Access Number PB-274 040).
(May 1977).
-21-
-------
MPTER - MPTER is a multiple point-source Gaussian model with optional
terrain adjustments. MPTER estimates concentration on an hour-by-hour
basis for relatively inert pollutants (i.e., S02 and TSP). MPTER uses
Pasquill-Gifford dispersion parameters and Briggs plume rise methods to
calculate the spreading and the rise of plumes. The model is most
applicable for source-receptor distances less than 10 kilometers and for
locations with level or gently rolling terrain. Terrain adjustments are
restricted to receptors whose elevation is no higher than the lowest stack
top. In addition to terrain adjustments, options are also available for
wind profile exponents, buoyancy induced dispersion, gradual plume rise,
stack downwash, and plume half-life.
Reference - Pierce, T.E. and Turner, D.B., 1980: User's Guide for MPTER:
A Multiple Point Gaussian Dispersion Algorithm with Optional Terrain
Adjustment. EPA-600/8-80-016, (NTIS Accession Number P880-197 361). U.S.
Environmental Protection Agency, Research Triangle Park, NC. 239 pp.
(April 1980).
BLP - BLP (Buoyant line and point source dispersion model) is a Gaussian
plume dispersion model designed to handle unique modeling problems
associated with aluminum reduction plants, and other industrial sources
where plume rise and downwash effects from stationary line sources are
important. POSTBLP and BLPSUM are related postprocessors in this system.
Reference - Schulman, L.L., and Scire, J.S., 1980: Buoyant line and point
source (BLP) dispersion model user's guide. Document P-73048. Prepared
for the Aluminum Association, Inc. by Environmental Research and
Technology, Inc., Concord, MA. (NTIS Accession Number P881-164 642).
(July 1980).
- Addendum/Supplemental Information for BLP. 2 pp. (December 1982).
(Distributed as part of UNAMAP, Version 5, Documentation.)
ISCST - The industrial source complex short term model is a steady- state
Gaussian plume model which can be used to assess pollutant concentrations
from a wide variety of sources associated with an industrial source
complex. This model can account for settling and dry deposition of
particulates, downwash, area, line and volume sources, plume rise as a
function of downwind distance, separation of point sources, and limited
terrain adjustment. Average concentration or total deposition may be
calculated in 1-, 2-, 3-, 4-, 6-, 8-, 12-, and/or 24-hour time periods.
An "N"- day average concentration (or total deposition) or an average
concentration (or total deposition) over the total number of hours may
also be computed.
References - Bowers, J.F., Bjorklund, J.R., and Cheney, C.S., 1979:
Industrial Source Complex (ISC) Dispersion Model User's Guide, Volumes 1
and 2. EPA-450/4-79-030, EPA-450/4-79-031. (NTIS Accession Number
P880-133 044, P880-133 051), Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency, Research Triangle Park, NC.
(December 1979).
-22-
-------
- Addendum/Supplemental Information to the Industrial Source Complex
Model. 20 pp. (December 1982). (Distributed as part of the
UNAMAP, Version 5, Documentation.)
ISCLT - The industrial source complex long term model is a steady-state
Gaussian plume model which can be used to assess pollutant concentrations
from a wide variety of sources associated with an industrial source
complex. This model can account for settling and dry deposition of
particulates, downwash, area, line and volume sources, plume rise as a
function of downwind distance, separation of point sources, and limited
terrain adjustment.
ISCLT is designed to calculate the average seasonal and/or annual
ground-level concentration or total deposition from multiple continuous
point, volume and/or area sources. Provision is made for special x, y
receptor points that may correspond to sampler sites, points of maxima or
special points of interest. Sources can be positioned anywhere relative
to the grid system.
References - Same as ISCST (above).
CALINE3 can be used to estimate the concentrations of non-reactive
pollutants from highway traffic. This steady-state Gaussian model can be
applied to determine air pollution concentrations at receptor locations
downwind of "at-grade," "fill," "bridge," and "cut section" highways
located in relatively uncomplicated terrain. The model is applicable for
any wind direction, highway orientation, and receptor location. The model
has adjustments for averaging time and surface roughness, and can handle
up to 20 links and 20 receptors. It also contains an algorithm for
deposition and settling velocity so that particulate concentrations can be
predicted.
Reference - Benson, Paul E. "CALINE3 - A Versatile Dispersion Model for
Predicting Air Pollutant Levels Near Highways and Arterial Streets."
Interim Report, Report Number FHWA/CA/TL-79/23, Federal Highway
Administration, 1979.
Non-Guideline Models:
TEM8 - TEM8 (Texas Episodic Model) is short-term, steady-state Gaussian
plume model for determining short-term concentrations of non-reactive
pollutants.
Reference - Staff of the Texas Air Control Board. User's Guide to the
Texas episodic model. Texas Air Control Board, Permits Section, 6330
Highway 290 East, Austin, TX 78723. (NTIS Accession Number P880-227 572).
TCM2 - TCM2 (Texas Climatological Model) is a climatological steady-state
Gaussian plume model for determining long-term (seasonal or annual)
arithmetic average pollutant concentrations of non-reactive pollutants.
-23-
-------
Reference - Staff of the Texas Air Control Board. User's Guide to the
Texas Climatological Model (TCM). Texas Air Control Board, Permits
Section, 6330 Highway 290 East, Austin, TX 78723. (NTIS Accession Number
P881-164 626).
PAL - Point, Area, Line Source Algorithm. This short-term Gaussian
steady-state algorithm estimates concentrations of stable pollutants from
point, area, and line sources. Computations from area sources include
effects of the edge of the source. Line source computations can include
effects from a variable emission rate along the source. The algorithm is
not intended for application to entire urban areas but for smaller scale
analysis of such sources as shopping centers, airports, and single
plants. Hourly concentrations are estimated and average concentrations
from 1 hour to 24 hours can be obtained.
References - Petersen, W.B., 1978: User's Guide for PAL - A
Gaussian-Plume Algorithm for Point, Area, and Line Sources.
EPA-600/4-78-013. (NTIS Accession Number PB-281 306). U.S. Environmental
Protection Agency, Research Triangle Park, NC. (February 1978).
- Addendum/Supplemental Information for PAL, HIWAY2, and RAM. 5 pp.
(December 1980).
PTPLU - PTPLU is a point source Gaussian dispersion screening model for
estimating maximum surface concentrations for 1-hour concentrations.
PTPLU is based upon Briggs plume rise methods and Pasguill-Gifford
dispersion coefficients as outlined in the workbook of atmospheric
dispersion estimates. PTPLU is an adaptation and improvement of PTMAX
which allows for wind profile exponents and other optional calculations
such as buoyancy induced dispersion, stack downwash, and gradual plume
rise. PTPLU produces an analysis of concentration as a function of wind
speed and stability class for both wind speeds constant with height and
wind speeds increasing with height. Use of the extrapolated wind speeds
and the options allows the model user a more accurate selection of
distances to maximum concentration. PTPLUI is the interactive version of
this model.
HIWAY2 - HIWAY2 is a batch and interactive program which computes the
hourly concentrations of non-reactive pollutants downwind of roadways. It
is applicable for uniform wind conditions and level terrain. Although
best suited for at-grade highways, it can also be applied to depressed
highways (cut sections). HIWAY2 is intended as an update to the hiway
model. HIWAY2I is the interactive version of this model.
References - Petersen, W. B., 1980. User's guide for HIWAY2: A highway
air pollution model. EPA-600/8-80-018. (NTIS Accession Number P880-227
556). U.S. Environmental Protection Agency, Research Triangle Park, NC.
70 PP. (May 1980).
- Rao, S.T., and Keenan, M.T., 1980: Suggestions for
improvement of the EPA-HIWAY Model. J. Air Pollution
Control Assoc., 30, 6, 247-256.
-24-
-------
- Addendum/Supplemental Information for PAL, HIWAY2, and
RAM. 5 pp. (December 1980).
COMPLEX I - Complex I is a multiple point source code with terrain
adjustment. It is a sequential model utilizing hourly meteorological
input. It assumes a normal distribution in the vertical and a
uniform horizontal distribution across a 22.5 degree sector.
Reference - There is no users guide for Complex I, and EPA has no
plans to develop one as of December 1980. (Since Complex I is based
upon MPTER, the user guide for MPTER is useful. Also note the
differences from MPTER given in comment statements in the first few
pages of the Complex I source code).
SHORTZ - SHORTZ is designed to calculate the short-term pollutant
concentrations produced at a large number of receptors by emissions
from multiple stack, building, and area sources. SHORTZ uses
sequential short term (usually hourly) meteorological inputs to
calculate concentrations for averaging times ranging from 1 hour to 1
year. The model is applicable in areas of both flat and complex
terrain, including areas where terrain elevations exceed stack-top
elevations. The program requires random-access mass storage
capability. An associated compatible meteorological data processor
is METZ.
References - Bjorklund, J.R., and Bowers, J.F., 1982: User's
Instructions for the SHORTZ and LONGZ Computer Programs, Volumes I
and II. EPA-903/9-82-004A and B. (NTIS Accession Number P883-146
092 and P883-146 100). U.S. Environmental Protection Agency, Middle
Atlantic Region III. Philadelphia, PA. (November 1982).
LONGZ - LONGZ is designed to calculate the long-term pollutant
concentration produced at a large number of receptors by emissions
from, multiple stack, building, and area sources. LONGZ uses
statistical wind summaries to calculate long-term (seasonal or
annual) average concentrations. The model is applicable in areas of
both flat and complex terrain, including areas where terrain
elevations exceed stack-top elevations. The program requires random-
access mass storage capability.
References - Same as SHORTZ (above).
MESOPUFF - MESOPUFF is a variable trajectory regional-scale Gaussian
puff model especially designed to simulate the air quality impacts of
multiple point sources at long distances. Highly user-oriented,
MESOPUFF provides a range of flexible options. It is designed to be
driven by user-specified meteorological scenarios, of arbitrary
duration, constructed by a suitable meteorological preprocessor,
MESOPAC. It outputs spatially-gridded concentration arrays averaged
over arbitrary time intervals of one hour or more and is designed to
-25-
-------
be coupled to a postprocessor, MESOFILE, to provide additional
graphical and statistical analyses. Routines are provided for:
plume rise, plume growth, fumigation, linear conversion of SOz and
S04, and dry deposition of S02 and S04.
References - Bass, A., Benkley, C.W., Scire, J.S., and Morris, C.S. ,
1979: Development of MESOSCALE Air Quality Simulation Models:
Volume 1. Comparative Sensitivity Studies of Puff, Plume, and Grid
Models for Long-Distance Dispersion. EPA-600/7-80-058 (NTIS
Accession Number P880-227 580) U.S. Environmental Protection Agency,
Research Triangle Park, NC. (September 1979).
- Benkley, C.W. , and Bass, A., 1979: Development of
MESOSCALE Air Quality Simulation Models, Volume 3.
User's Guide to MESOPUFF (MESOSCALE PUFF) Model. EPA
600/7-80-058) U.S. Environmental Protection Agency,
Research Triangle Park, NC. (September 1979).
- Addendum/Supplemental Information for MESOPUFF. 25 pp.
(December 1982). (Distributed as part of UNAMAP,
Version 5, Documentation.)
4.1.2 Alternative Models
Dispersion for continuous chemical releases can in general be simulated
with the continuous models described in the previous section. Table 3-2
indicated situations in which these models would be inappropriate or at least
incomplete for a simulation of air toxics emissions. Table 4-1 summarizes the
capabilities of a selection of complete models. The table is organized to
identify components which are involved in the models under general categories
of source, dispersion, chemistry, and deposition components. In addition,
details are available about the availability and characteristics of the
computer programs. Symbols for the table are described in the footnote.
Complete models identified in the table as having codes which are generally
available for assessment are described further following the table.
-26-
-------
TABU 4-1
CHARACTERISTICS OF ALTERNATIVE MODELS
NJ
MODEL
Source
Point
Area
Line
Plume Rise
Oownwash
Pool Evaporation
Land
Water
Pool Spread
Flashing
Liquid
Liquefied Gas
Gas/Aerosol
Dispersion
Gravity Spread
Continuous
Instantaneous
Straight Line
Variable
Trajectory
Cloud Heating
Chemistry
Deposition
Wet
Dry
Model
Available Code
Available Manual
Type
ALWAS
X"
X
X
X
.
-
.
-
-
.
-
-
X
-
X
X
X
.
-
-
X
X
X
X
c
Tox-
Sereen
X
X
-
X
-
-
-
-
-
-
-
-
X
-
X
-
X
-
-
X
X
X
X
X
c
SLAB
-
X
-
-
-
.
-
-
-
-
-
X
X
X
X
_
X
-
X
-
.
-
X
_
c
INPUFF
X
X
-
X
-
-
-
-
.
-
-
-
X
-
X
-
X
X
-
-
-
-
X
X
c
SRO
DENZ
-
X
-
-
-
-
X
X
X
X
-
X
-
X
X
X
X
.
X
-
-
-
X
X
c
IEPA
X
-
-
-
-
X
X
-
-
-
s
s
s
-
X
-
X
-
-
-
-
-
-
X
N
Fuming
Acid
X
_
-
-
-
X
X
-
-
-
X
-
X
-
X
.
X
-
-
-
-
-
NS
NS
C
OB/OG
X
-
-
-
-
-
-
-
-
-
-
-
X
-
X
-
X
-
-
-
-
-
-
-
N
EIOSVIK
-
X
-
-
-
-
-
-
-
-
-
-
X
X
X
X
X
-
X
-
-
-
NS
NS
NS
DOESB
X
-
-
X
-
X
X
-
-
-
X
-
X
-
X
-
X
-
-
-
X
X
NS
NS
N
EPA
Puff
X
X
-
-
-
-
-
-
-
-
-
-
X
-
-
X
X
-
-
-
-
-
X
X
PC.N.C
Van
111 den
-
X
-
-
-
-
-
-
-
-
-
-
X
X
-
X
X
-
-
-
-
-
NS
NS
NS
AGA
-
X
X
-
-
X
X
-
-
-
-
X
X
-
X
X
X
-
-
-
X
X
N
SPILLS"
X
X
-
X
.
X
X
-
-
X
X
X
X
-
x
X
X
X
-
-
-
-
X
X
C.PC
TOXCOP
X
_
-
-
-
X
X
-
-
-
X
.
X
-
X
-
X
-
-
-
-
-
NS
NS
NS
HEGADIS
-II
X
X
-
-
-
-
-
-
-
.
-
-
X
X
X
X
X
-
X
-
-
-
NS
NS
NS
SRI
PUFF
X
_
-
-
-
-
-
-
-
.
-
-
X
-
X
X
X
X
-
-
-
-
X
X
PC
Germel-
es and
Drake
X
X
-
-
-
X
-
X
X
_
-
X
-
X
X
X
X
-
X
-
-
-
NS
NS
NS
DEGAOIS
-
X
-
-
-
-
-
-
-
-
-
-
X
X
X
X
X
-
X
-
-
-
X
X
c
Chlorine
Institute
X
X
-
X
-
X
X
-
-
X
-
X
X
X
X
X
X
-
-
-
-
-
-
X
N
ICARIS 1J »n AAR version of SPILLS
"Symbols
X Included; - not Included
C Computer
N Nomogrwi/cilcuUtor
NS not specified
PC minicomputer
S specified
-------
U.S. Environmental Protection Agency
Model: ALWAS
General Description:
ALWAS is the air, land, water analysis system designed to simulate
the fate of airborne toxic materials on the land and surface
water. The major component is the DiDOT, Dispersion/Deposition of
Toxics, model which simulates the source of toxic material in
terms of emission rates from point and area sources; dispersion,
using a continuous Gaussian model; and wet and dry deposition with
a submodel. The dispersion model was developed and is used to
simulate TSP and any other chemical. An option of the model
allows interaction of volatile organics and TSP to simulate
surface absorption. Dispersion in the model is simulated using
techniques from EPA-approved models.
Dry deposition is simulated using a surface depletion model and
wet deposition is simulated using a scavenging ratio technique.
Wet deposition calculations are provided only to estimate the flux
of material from the surface rather than providing an estimate of
plume depletion en route between the source and the receptor. The
model simulates concentration and deposition values by use of a
sequence of hourly data. Modifications are discussed in the
user's manual to simulate instantaneous source releases. Inputs
to the model include source and meteorological data and chemical
specific parameters for wet and dry deposition.
Reference: Tucker et.al. (1984)
U.S. Coast Guard
Model: DEGADIS
General Description:
DEGADIS, the Dense Gas Dispersion Model has been developed for
inclusion in the Hazard Assessment Computer System (HAGS) to
simulate dispersion of heavy gas releases. The model is a
modified version of the HEGADIS-II model, Colenbrander (1980).
DEGADIS uses as input information on vapor generation rate and
initial source area to build a vapor cloud by considering
entrainment and cloud heating. The cloud has a cylindrical core
which is horizontally homogeneous and edges which decrease
exponentially in the horizontal. The model includes a gravity
spreading or intrusion equation to simulate horizontal spread,
vertical and horizontal mixing and equations for energy balances
and mass uptake. Equivalent parameters are determined relating
the cylinder height and width to equivalent mixing parameters from
which concentration estimates can be made. It allows a smooth
transition from gravity spreading to passive dispersion as the
energy level of the cloud due to buoyancy approaches that of
existing atmospheric turbulence.
Reference: Havens and Spicer (1985)
-28-
-------
U.S. Environmental Protection Agency
Model: TOX-SCREEN
General Description:
TOX-SCREEN is a multimedia screening model for long term
assessment of transport in the air, soil and surface water. The
air pathway model is a simplified Gaussian model for point sources
and a simple box model for area sources. The dispersion model
results for a given wind direction are assumed to stay constant
for the month (i.e. short term concentrations) are simulated by a
source depletion model and an exponental decay term.
Precipitation scavenging is estimated using a washout ratio and
monthly average precipitation.
Reference: Bicknell et.al. 1985
U.S. Environmental Protection Agency
Model: INPUFF
General Description:
INPUFF is a single source Gaussian dispersion model which allows
dispersion calculations for stationary and moving sources of
neutrally buoyant materials in air. INPUFF uses a Gaussian puff
dispersion equation in stationary or temporally and spatially
varying wind fields provided by the user. The model simulates
emissions from a single source at up to 25 receptors for up to 144
meteorological periods of length from minutes to an hour. Puff
positions are determined by trajectory calculations. Options in
the model allow estimation of plume rise; wind speed at release
height; position from a moving source; and, from wind field input,
the effects of temporally and spatially varying winds including
wind fields representative of pollutant transport in complex
terrain.
As a Gaussian model, INPUFF is limited to steady state simulations
of neutral tracer dispersion without chemical transformation
and/or depositional losses. Area sources are not directly
simulated although use of initial dispersion parameters and
virtual point source concepts would allow approximations. The
model is not part of an emergency response system.
Reference: Petersen et al. (1984)
-29-
-------
Safety and Reliability Directorate (SRD) , U.K.
Model; DENZ
General Description:
DENZ is a heavy gas dispersion model which draws on and includes
components in other models as options in explaining heavy gas
releases. For pool evaporation, DENZ uses the SPILL model of the
SRD described in the following subsection. For dispersion, the
model uses the formulation of Cox and Roe for gravity spreading of
heavy gas with top entrainment and a standard Gaussian dispersion
model for passive tracers. The model and manual include provision
for source simulations from refrigerated and pressurized vessels
and the heavy gas or slumping model includes terms for slumping,
cloud heating, air entrainment in the cloud. The dispersion model
is a Gaussian puff model.
DENZ provides estimates of concentrations, areas, and doses as
well as cumulative probabilities of exposure (for toxic gases).
Input to the model includes control records, source information
and parameter constants and meteorological data for the site under
study.
Reference: Fryer and Kaiser (1979)
Illinois Environmental Protection Agency
General Description:
The Illinois EPA uses a set of equations based on a continuous
Gaussian equation for ground level point sources for estimating
downwind evacuation distances. In addition to the dispersion
model, the equations also provide estimates for pool evaporation
and specifications for continuous and instantaneous discharges of
gas or volatile leaks on land and water. The evaporation rate is
determined only by the vapor pressure of the material released.
The model is very simple in its specification of release amounts
and simulation of dispersion. Its use is intended to provide a
rough estimate of evacuation corridors.
Reference: Kelty (1984)
-30-
-------
U.S. Environmental Protection Agency
Model: Instantaneous Puff
General Description:
The EPA puff model is an instantaneous Gaussian point source model
for ground level sources. It includes provisions for using
instantaneous horizontal dispersion parameters and specified
initial dispersion rates which give it the ability to simulate
area sources by virtual point source techniques. Methods are
included for dose calculations and approximating concentrations
for time periods other than those for which the initial
calculations were performed.
The puff model is limited as a Gaussian model to simulations of
dispersion for neutral buoyancy releases from instantaneous
sources.
Reference: Petersen (1982)
American Gas Association
Model: LNG Spills in Dikes
General Description:
The AGA has provided a model for LNG spills on flat and sloped
dike floors which considers vaporization, dike filling by liquid
and vapor and dispersion. Dispersion calculations are performed
using a continuous Gaussian line source model without a
parameterization for heavy gas spreading. The vaporization model
uses previous studies to estimate the boiling rate in conjunction
with a series of equations for liquid releases and the geometry of
liquid spreading on flat and sloped floors. Estimates are
included for the time required before bermed areas are filled by
vapors and overflow causing downwind transport.
The model includes a mix of assumptions which make it in some ways
general and in some ways very specific to LNG. Gravity spreading
of vapor is neglected and dispersion is simulated using a Gaussian
(neutral buoyancy) model, but the boiling model is specific to LNG.
Reference: AGA (1978)
-31-
-------
Shell Development Company
Model: SPILLS
General Description:
SPILLS is a widely used model for simulation of liquid spills or
gas releases on land and the subsequent dispersion of the gases or
vapors. The dispersion model is a Gaussian puff model which uses
continuous plume dispersion parameters and has the capability of
simulating elevated sources and inversion trapping. Area sources
are simulated using a virtual point source formulation with the
initial dispersion parameters. Source models are of several
types. For routine stack emissions, the model uses Briggs plume
rise equations. For liquid spills, the model differentiates among
continuous leaks and instantaneous spills of a liquefied gas or a
liquid. All spills are assumed to be bounded. For continuous
leaks, the spill rate is calculated and used as the emission rate
with a pool area calculated to provide an area source. For
instantaneous sources of liquefied gases, the flashing of gas is
calculated, the area of the pool defined and boiling estimated as
a result of conductive soil heating and convective heating from
the air to provides a time dependent rate. For liquids, a mass
transfer model is used. The model is unsteady in that emissions
and meteorological conditions change. Puffs released under
different conditions are integrated at the receptors of interest.
The model was prepared for simulating releases of 36 chemicals but
is appropriate to others. It is limited in its ability to
simulate peak instantaneous concentrations and does not include
provisions for pool growth and gravity spreading.
References:
- Fleischer (1980)
- Kricks et al. (1983)
- Pan' et. al. (1983)
U.S. Coast Guard
Model; HAGS
General Description:
HAGS is the Hazard Assessment Computer System which contains 18
models describing chemical spills and dispersion. The system is
under revision after reviews of early models indicated that
significant errors in the models could exist (Tebaugh, personal
communication, 1985). One model under revision is the heavy gas
dispersion model for which a new model DEGADIS has been developed.
-32-
-------
References:
Colonna et. al. (1984)
Potts (1981)
Harding et. al. (1978)
SRI, International
Model: SRI PUFF
General Description:
The SRI PUFF model is a microcomputer based dispersion model for
simulation of unsteady emissions in temporal and spatially varying
wind fields. The model calculates non-divergent wind fields from
multiple stations, estimates plume positions, determines the puff
release rate to assure a continuous plume simulation, and
estimates concentrations with the dispersion equation.
References:
- Ludwig (1983; 1984)
- Ludwig et. al. (1977)
4.2 Model Components and Alternative Formulations
Section 2.0 included discussion of some of the differences in models that
would be required to simulate air toxics releases. In Section 3.0, a decision
tree (Table 3-2) was presented which defines certain modules which should be
included in models given various types of chemical releases. Information on
complete models in Table 4-1 can be reviewed to determine if the models can be
used to evaluate air toxics emission releases. If the models are
inappropriate, the missing or inconsistent component can be modified to
provide the appropriate model with a minimum of modifications. The following
subsections describe some areas of potential changes.
-33-
-------
4.2.1 Source/Emissions Modules
Air toxics releases generally fall into long-term continuous or short-term
categories. Often continuous releases can be simulated using generally
accepted dispersion models with point or area source terms. Emission rates
are determined from field measurements, parameterizations or theoretical
models. Short-term releases more often occur as a result of accidental
releases or batch type processes for which source estimates are poorly defined
and difficult to determine.
Selection of emission parameters for soil related dusts which may have a
toxic component can be guided by emission factor relationships developed for
the EPA(1984). Equations have been developed for paved and unpaved roads due
to wind and vehicular traffic, agricultural tilling, and aggregate handling
and storage. Equations .consider wind erosion, material types, mechanical
action, and soil moisture to determine an emission rate for different
conditions. Cox et. al. (1977) discuss an alternative theoretical model for
dust emission rate.
Other continuous types of emissions are those due to stack emissions and
the volatilization of contaminants on the ground or from storage lagoons. The
latter will be discussed with information on short-term releases.
Short-term releases are considered to be of five types:
Emissions from stationary sources
Short-term releases from stationary sources are defined in terms
of emissions and source parameters required for the models in
Section 4.1. Emissions data can be determined from measurements
or site specific knowledge of the combustion characteristics or
process involved. Short-term releases often occur as a result of
relief venting as a safety measure. The characteristics of the
relief vent or valve can often be obtained from the source in
question. Characteristics of the emissions and other source
parameters needed for estimates of plume rise may be available as
design parameters but the nature of plant upsets as excursions
from normal operations suggests the need for a process evaluation
following short-term release. Pressure blowdown will be discussed
as another category.
-34-
-------
Evaporation of liquids
Evaporation of liquids has been simulated with a number of models
which will be presented later in this section. Evaporation models
are typically derived for evaporation from confined pools or
confined spills on land. The first element of the estimate is a
determination of the liquid release rate. In general standard
fluid dynamical equations are used with simplifying assumptions.
The Chlorine Institute (1982) uses an expression in which release
rate of the liquid is proportional to the area of the release, a
discharge coefficient for orifices and nozzles and the square root
of the product of the pressure on the fluid, its density and the
gravitational acceleration. Such models are generally not unique
nor tested. Models described later in this section describe the
release rate in greater detail.
The basic pool evaporation equations simulate "mass transfer by
considering the vapor pressure of the liquid, the area of the
pool, the energy balance and a mass transfer coefficient. The
general form of the equations is discussed by Ille and Springer
(1978).
The pool temperature and mass transfer coefficients represent the
primary differences among models and incorporate terms such as the
energy balance in the pool and the gas characteristics (e.g.
diffusivity). Models can be made time dependent for expanding
pools. Shaw and Briscoe (1978) review models for spreading
liquids on land. Spreading of liquids on water follows density
intrusion models developed for oil spills. The Coast Guard HACS
models describe source models of this type.
Vaporization of liquefied gases (cryogenic)
Cryogenic liquefied gases have been studied due to the interest in
LNG. Pool models for land spills and water spills have been
developed based on measurements of LNG vaporization and
theoretical models for heat conduction. Heat conduction from the
ground is considered the most important heat source and models for
land spills are solutions of one dimensional heat conduction
equations. A common form of equation for vaporization rate per
unit area is given by Shaw and Briscoe (1978).
Other models consider additional heat sources such as the latent
heat of fusion and solar insolation. Total emission rates are
given by a number of different assumptions including:
- for unconfined continuous spills, vaporization equals liquid
release rate
- the area of vaporization is equal to the area of confinement.
- the spill area can be determined from the pool spreading speed.
-35-
-------
Spills on water use similar models with density intrusion models
to represent liquid spreading and vaporization based on heat
transfer which is enhanced by convection in the water. An example
model for continuous spills on water is given by (Shaw and
Briscoe, 1978).
Most data available are for LNG experiments on both land and
water. In many cases, the data and the models are not easily
transferable because the data are given in terms of LNG regression
rate. Differences between LNG and other chemical vaporization
models include primarily a question of what are the most important
variables to be considered in energy balances and whether ice (and
related latent heating) is formed by spills on water.
Pressurized liquified gas spills
The accidental, short-term release of pressurized gases is one of
the most difficult types of release to simulate. Since the gas is
stored under pressure, the boiling point is raised to a level
above ambient temperature. On release of the pressure, by damage
to the pressure vessel or piping, the pressure is reduced and the
boiling point is lowered to the value for ambient atmospheric
pressure. With the rapid drop in pressure, a significant portion
of the liquid is adiabatically flash vaporized to gas and the
temperature of the remaining pool or tank volume is reduced to the
boiling point. For chlorine, the amount of gas flashed is 30 to
25 percent of its initial volume depending on the initial storage
temperature. The amount can be determined through thermodynamic
calculations.
If the tank damage is above the liquid level, gas will escape to
the atmosphere and the remaining liquid will evaporate as heat is
obtained from its surroundings. This gas release rate is
relatively small. If the tank damage is below liquid level or if
there is a complete and rapid failure, large portions of this
initial volume will be entrained as very small droplets into the
vapor from the vigorous boiling. The size of the droplets is
small enough to avoid gravitational settling and their evaporation
in the cloud-entrained air reduces the cloud temperature to the
boiling point. This reduction can be sufficient to make the cloud
negatively buoyant with respect to air. Complex two-phase flow
models are available for releases from pipes. In the case of
catastrophic tank failures little is known about the entrainment
of droplets and it is sometimes assumed that the entire tank
contents are flashed or entrained into the initial cloud. Kansa
et al. (1983) report an attempt to simulate the transitional
density of an ammonia vapor/droplet cloud given an initial
estimate of the liquid fraction.
* Pipeline ruptures
Models have been developed to estimate the source rate of gas
released from a pipeline leak or failure (Hanna and Munger, 1983;
Blewitt, 1985). The models consider the characteristics of the
gas and opening and the mass available between the failure and
block valves to provide a time dependent emission rate.
-36-
-------
Vapor discharge rates can be calculated by standard equations. For
example, the Chlorine Institute (1982) presented an equation for vapor
discharge.
Table 4-2 identifies and lists the characteristics of source models
identified in this study. The majority of available source models are for
pool evaporation of specific chemicals such as propellants. A major source
for models of vaporization for a large range of source conditions is the U.S.
Coast Guard HACS system.
4.2.2 Chemical Conversions
Chemical modules have traditionally been included in air quality models to
simulate photochemistry related hydrocarbon, NOX and oxidant emissions.
Losses of material have also been included in models by a simple exponential
decay parameter based on a constant conversion rate. Few models discussed
here have chemical conversion/reaction modules, although an understanding of
chemical and combustion products may be very important in air toxics impact
assessments.
4.2.3 Deposition
Deposition from plumes may be important both because of the potential
impact of the materials deposited and the loss of material from the plume.
Dry deposition is controlled by the transfer characteristics of the
atmosphere, the chemical and physical characteristics of the depositing
material, and the characteristics of the surface. Atmospheric factors govern
the rate at which pollutants are transferred through the atmosphere to the
surface. The pollutant characteristics determine if the pollutant will act as
an aerosol or gas and whether it reacts with the surface to enhance the
deposition rate (e.g. if a gas is soluble in water). The surface
-37-
-------
TABLE 4-2
SOURCE CHARACTERIZATION
I
U)
CD
Characteristics
land
water
continuous
instantaneous
confined
unconf ined
pressurized
cryogenic
gas
liquid
pipeline failure
tank leak
tank failure
Model
pool evaporation
steady state
time dependent
flashing
aerosol
generation
plume rise
empirical
semiempirlcal
analytical/
numerical
chemical reactions
USCG/
Tana
X
-
X
-
X
-
-
-
X
-
-
-
X
-
X
-
X
-
-
X
-
-
USAF/
Clexell
X
-
X
-
X
-
-
-
-
X
-
-
-
X
X
-
-
-
-
-
X
-
X
HYSDEC/
SJiea
X
X
-
X
-
-
-
-
X
-
-
-
X
X
-
-
-
-
-
-
X
-
Alberta/
Miiuaier
X
X
-
-
-
X
-
X
-
X
-
-
-
X
-
-
-
-
-
-
X
-
Toronto/
Sliy.ec
X
-
X
-
X
X
-
-
-
X
-
-
-
X
X
X
-
-
-
-
-
X
USAF/
-Mahler. U
X
-
X
-
X
-
-
-
-
X
-
-
-
X
X
-
-
-
-
X
-
-
ARMY/
X
-
X
-
X
-
-
-
-
X
-
-
-
X
X
-
-
-
-
-
-
X
USAF/
Hie
X
-
X
-
X
-
-
-
-
X
-
-
-
X
X
-
-
-
-
-
-
X
Monsanto
Hi'
X
X
X
X
X
X
-
-
-
X
-
X
X
X
X
X
-
-
-
-
-
X
Reid
_1UGL
X
-
X
-
X
-
-
X
-
X
-
-
-
X
-
X
-
-
-
-
-
X
UKSRD
Shaw
X
X
X
X
X
X
-
X
-
X
-
-
-
X
-
X
-
-
-
-
-
X
Input s/Parameteri 23 t Ions
area
wind speed
stability/
turbulence
heat exchange:
insolation
substrate
ice formation
air: convective
radiative
evaporation
substrate type
roughness
pool temperature
pressure
vapor pressure
_
-
-
-
. -
-
-
-
-
X
X
-
-
-
-
X
-
-
X
X
-
-
-
-
X
X
-
X
-
-
-
-
-
-
X
X
-
-
-
-
-
X
-
X
X
X
-
-
-
-
-
-
-
X
X
X
X
-
-
X
-
X
-
-
X
X
-
X
X
X
X
-
X
X
-
X
-
X
X
X
X
X
X
X
-
X
X
-
X
X
X
-
-
X
-
X
X
-
-
-
X
-
-
-
-
X
-
X
-
-
X
-
-
-
X
-
-
-
-
X
X
X
-
X
-------
characteristics help determine the meteorological parameters as well as
determining the mode of deposition.
Dry deposition parameters or velocities are typically determined by
experimentation. Deposition velocity is a parameter made up of a ratio of
deposition flux to the surface to the ambient pollutant concentration. In the
simplest case, a flux can be determined by multiplying an ambient
concentration times a deposition velocity. Deposition velocities,
particularly for exotic chemicals, are poorly defined. McMahon and Denison
(1979) and Sehmel (1980) have reviewed available data.
The simplest approach to deposition estimates is an exponental decay model:
C = Co e
[vd t"
AZ
where vd is the deposition velocity, t is time and AZ is a characteristic
depth often given as
ir
AZ =
2
Wet deposition parameterizations can also use the simple exponential decay
parameterization with either of two parameters, a scavenging coefficient,
which is simply a fractional scavenging rate, or a washout ratio (W) (the
ratio of pollutant concentration in precipitation to that in air). For
scavenging coefficients(A):
C = Co e (-At)
where C is the resultant concentration after the scavenging of initial
concentration, C0. This concentration can be calculated using a standard
-39-
-------
model without scavenging. For washout ratio W, a scavenging coefficient can
be formed by
WP
A oc
H
where P is the precipitation rate, H, the depth over which scavenging occurs.
The proportionality can be removed by using density corrections to provide A
in the proper units.
Scavenging coefficients and washout ratios for many materials are reported
by McMahon and Denison (1979). Dana et al (1984) describe a program of
measurements specifically designed to evaluate scavenging parameters for a
limited number of air toxic pollutants based on solubilities.
4.2.4 Plume Rise
Buoyant and momentum plume rise equations are included in most of the EPA
preferred models. These equations developed by Briggs are appropriate to most
types of air toxic pollutant releases. A summary by Briggs (Randerson, 1984)
provides guidance for plume rise estimates for special cases such as plume
rise from multiple sources, stack tip downwash, and rise of moist plumes
(latent heating effects).
4.2.5 Peak Concentration Levels
Most models for air toxics impact assessments are simplified analytical
solutions to advection/diffusion equations representing mean concentrations
over some averaging time. Briggs (1973) reports power laws that relate
maximum concentration relative to 30-min values and averaging times with
exponents ranging from -2/3 to -1/6 depending on atmospheric stability. Some
-40-
-------
American Gas Association tests (1967, 1974) indicate peak-to-mean ratios of
about 2 or 3.
Care must be exercised in interpreting peak-to-mean ratios. Determina-
tions should be made as to whether the peak values are being ratioed to the
maximum value along the centerline of the plume or to an average value off the
centerline of the plume. Ramsdell and Hinds (1971) studied a plume from a
continuous source and reported fluctuations based on 38-sec time intervals.
They observed peak-to-mean ratios greater than 5 less than 1/2% of the time at
the plume centerline, whereas ratios of 5 or greater occurred more than 6% of
the time near the edge of the mean plume. Terrain effects and atmospheric
stability can strongly influence these ratios.
4.2.6 Downwash
Aerodynamic downwash of plumes can be very significant in increasing
near-source pollutant concentrations. A limited model adjustment for building
downwash is found in ISC, an EPA model described in the previous section.
Downwash effects include reduction of plume height, potential recirculation of
pollutants in wake cavities, and enhanced initial mixing. The scope of the
effects varies significantly with the nature of the obstruction. Hosker
(Draxler, 1984) provides a good summary of flow disturbances and the modeling
assumptions used in cases with simple geometry.
-41-
-------
5.0 MODEL EVALUATIONS
Models for air toxics impact assessments have not been widely evaluated
for specific applications. Modeling can involve simulations of phenomena for
which appropriate experimental data bases are not available. Most air toxics
simulations depend on standard dispersion models either recommended for use by
the EPA or which have received implied scientific approval by their continued
use over long periods. These models are primarily those developed for the
simulation of neutral buoyancy pollutant releases. The predictability of
these models is discussed in Section 5.1. Models for the dispersion of
pollutants with transitional or negative buoyancy have received less scrutiny
due to the dearth of evaluation experiments and the specificity of their
applications and testing. Evaluation of these models is discussed in Section
5.2.
5.1 Model Limitations and Uncertainties - Models for Neutrally Buoyant
Emissions
Dispersion models contain many simplifications and generalizations,
relative to actual plume behavior for a given source at a given time. The
pollutant concentrations predicted by a model should be regarded as estimates,
subject to error and uncertainty. It is a fairly simple task to enumerate
factors which limit a model's accuracy, but it is much more difficult to
quantify model uncertainty. Results from a number of model evaluation
studies, however, provide a means of characterizing model reliability and of
identifying critical areas for future model improvement.
Issues relating to model performance of particular concern for air toxics
impact assessments include the estimating of peak (short-term) concentrations.
-42-
-------
the spatial distribution of concentration patterns, estimating concentration
frequency distributions (mean, median, and extreme values), applications
involving complex terrain, performance for very unstable (Class A) dispersion
conditions, and performance for negatively buoyant emissions.
5.1.1 Factors Limiting Model Accuracy
The assumptions and simplifications built into a model impose basic
limitations on prediction accuracy. A number of these model assumptions are
not strictly satisfied in many model applications. Beychok (1979) provides a
relatively clear and thorough discussion of the Gaussian model and the
conditions for which it is appropriate. Assumptions which are often important
limitations include the following:
Meteorological conditions are assumed to be constant during a
given hour. The effects of any systematic change or trend in wind
speed, wind direction, or stability conditions during an hour are
not described by the model.
Winds and turbulence are assumed to be the same at all locations
throughout the boundary layer. The effects of wind speed or wind
direction shear, or of changes in turbulence with height, are not
considered within the Gaussian formulation.
The basic model averaging time is assumed to be long, in
comparison to the time scale of turbulent atmospheric motion and
in comparison to the transport time from source to receptor.
Source emissions characteristics are assumed constant over the 1
hour averaging time.
Pollutant mass is conserved within the Gaussian formulation.
Processes which add or remove mass, such as deposition, decay, or
chemical transformations, are assumed to be of secondary
importance.
A number of models (built on the Gaussian framework) contain provisions for
treating conditions which are not consistent with the assumptions noted
above. "Retrofit" treatment of the effects of wind shear, gravitational
settling and deposition, for example, should be recognized as greatly
-43-
-------
simplified approaches to describing these complex phenomena. Numerical
models avoid many of these shortcomings and incorporate fewer assumptions.
These models, however, are resource and data intensive.
Complex terrain applications pose special difficulties for simple models.
Assumptions of straight-line flow ignore the effect of terrain obstacles and
can lead to incorrect predictions of plume location and unrealistic
concentration estimates, particularly for peak short-term values.
For very unstable conditions, the spatial and temporal scale of
atmospheric turbulence increases and assumptions of homogeneous turbulence or
long averaging times are generally not satisfied. These conditions also
present a formidable challenge to modelers.
Meteorological Factors. For many locations, meteorological conditions are
generally consistent with Gaussian model assumptions. For some conditions,
however, model predictions should be considered suspect. At very low wind
speeds, for example, several model assumptions break down. Pronounced wind
direction meander, on a 10- to 30-minute time scale, produces similar
problems, as does large wind direction shear (particularly for elevated
sources).
Geographical Factors. The Gaussian formulation is consistent with
homogeneous terrain/surface roughness and surface radiative properties.
Regions where terrain and/or surface conditions are highly variable or undergo
an abrupt transition pose considerable difficulties. Examples of problem
areas include coastal regions, mountainous regions, rural/urban transition
areas, and even forested versus cleared regions. Different sets of dispersion
coefficients have been developed for urban and rural regions; how conditions
in a given region compare to those for which the coefficients were developed
will also influence model reliability.
-44-
-------
5.1.2 Operational Uncertainty
Considering the number of factors which contribute to model uncertainty,
the prospects of assessing reliability by calculating their combined effects
are quite discouraging. A more practical method of assessing model
reliability is to compare model predictions with observed air quality. A
number of model evaluation studies have involved systematic comparisons of
predicted and observed concentration values for a variety of source types and
meteorological conditions. Draxler (Randerson, 1985) provides a description
of data bases which can be used in evaluations of neutral buoyancy dispersion
models.
Such operational tests of model performance are, of course, subject to the
limitations of measurement uncertainty and experimental design. Differences
between model.predictions and observations may be due to uncertainties in
model inputs or measured concentrations, in addition to model deficiencies.
For most model evaluation studies, care is taken to obtain the best available
data sets in order to minimize these effects.
Differences may also result from the natural "noise level" produced by the
random fluctuations which characterize atmospheric turbulence. The Gaussian
model predicts "expected values" based on a postulated probability
distribution of concentrations, but the measured value for a given event
represents only a single sample from this distribution. Such "inherent model
uncertainty" will set a lower limit to the discrepancies between observed and
predicted values.
Operational model evaluation studies have been conducted using data bases
ranging from atmospheric tracer dispersion experiments to long-term air
quality monitoring programs for actual pollutant sources. Tracer experiments
offer the benefits of idealized, well-controlled conditions, while studies for
actual sources test the models under "real-world" conditions.
-45-
-------
Londergan, et al. (1980) assembled an archive of historical tracer
experiments representing a wide range of dispersion conditions and evaluated
available regulatory models for urban and rural applications. Most of the
data sets represented either near-ground or elevated, fixed-height releases.
In a second study, Schulman and Scire (1982) evaluated the EPA models RAM and
ISC using a data set of S02 measurements taken around an industrial source
complex at Midland, Michigan. Comparisons of observed and predicted
concentrations for a specific time and location showed large scatter and low
correlation. Schulman and Scire found that comparisons of peak concentration
values, regardless of time or location, showed better agreement between
predictions and observations. The meteorological conditions associated with
peak predictions often did not match those for peak observed values.
The Electric Power Research Institute's Plume Model Validation and
Development project (Bowne, et.al. 1983) represents an extensive program of
field experiments and evaluation studies .for tall-stack buoyant plumes.
Tracer releases from actual power plants were combined with air quality
measurements on the ground and aloft to document plume behavior and resulting
concentrations for two coal-fired power plants. The first site, the Kincaid
plant in central Illinois, is a flat, rural location. Over 300 hours of
tracer measurements were made and over 8 months of S02 measurements were
acquired with a 28-station network of monitors.
Figure 5-1 illustrates the comparison of maximum hourly tracer
concentrations predicted by the CRSTER model (an EPA Guideline model) with
values measured by the 200-sampler tracer array. The scatter between
predicted and observed concentrations is readily apparent. The observed and
predicted cumulative distributions of tracer concentrations are illustrated in
Figure 5-2. While systematic differences are evident, differences between the
-46-
-------
600 -
500 -
400 -
c:
o.
100 -
200
300
400
OBSERVED (10~9 s/m3)
Figure 5-1. Comparison of highest observed and predicted relative concentrations (x/Q;
concentration/emission rates) values for paired one-hour concentration .
averages from the CRSTER continuous Gaussian Point Source Model showing
the typical scatter in paired comparisons.
-47-
-------
Oi
o:
o
t
z:
LJJ
C_)
CsL
UJ
O.
UJ
99.9 U
99 h
95 h
80 h
50
20 h
5
1
10
PREDICTED
120
240
360
480
OBSERVED AND PREDICTED
RELATIVE CONCENTRATION
(X/Q, 10"9 s/m3)
Figure 5-2. Cumulative frequency distributions of the highest observed and predicted
X/Q values for one-hour average concentrations using the CRSTER model.
The plots show more agreement in frequency distributions than was found
in one-to-one pairings in the previous figure.
-48-
-------
distributions are generally smaller than those indicated for event-by-event
comparisons.
EPA has sponsored a series of systematic model performance evaluation
studies for regulatory models. An AMS workshop on model validation (Fox,
1980) recommended a statistical approach for validation studies. A group of
seven rural models (Londergan, et al., 1982) were evaluated following this
approach, using data from a six-station SOz network around the Clifty Creek
power plant in Indiana. Six urban models (Minott, et.al., 1983) were then
evaluated using one year of S02 data (13 stations) from the St. Louis RAPS
study. A group of eight complex terrain models were evaluated (Wackter and
Londergan, 1984) using two data sets, including tracer experiments conducted
at Cinder Cone Butte, an isolated hill in Idaho, plus measurements taken
around the Westvaco Luke paper mill in western Maryland.
All of these studies have indicated similar limitations for models in
present use. Comparisons between observed and predicted concentrations paired
in time show large scatter and low correlation (generally 0.2 or less).
Measurements for actual sources show more scatter than those for idealized
tracer experiments. The meteorological conditions associated with peak
observed and predicted values often do not match. For rural and urban
applications, peak predicted concentrations (regardless of time) generally
matched observed peak values within a factor of 2. For complex terrain,
however, the Complex I model (currently recommended by EPA as a screening
technique) over-predicted peak concentrations for Westvaco by a factor of 10.
For long-term average concentrations, model performance varies with source
and terrain characteristics. For distributed low-level sources in
uncomplicated terrain, model predictions often agree with observed
concentrations within 20 to 30 percent. This level of performance has been
achieved in urban studies, such as Minott, et al (1983). For elevated
-49-
-------
point-source emissions, regulatory models have been found to underpredict
long-term average values substantially. This problem is particularly evident
in complex terrain settings.
5.1.3 Reliability of Model Components
An understanding of the reliability of individual model components and of
the sensitivity of model predictions to these components is an essential first
step in efforts to improve current models. Several studies aimed at
developing improved models have examined components such as plume rise and
horizontal and vertical dispersion coefficients.
The EPRI Plume Model Validation and Development project undertook a
systematic "diagnostic model validation" (Liu, et al., 1982) of the basic
components of models applied to tall-stack plumes, including a model
sensitivity analysis which identified plume rise and vertical dispersion
coefficients as the variables which influence predicted ground-level
concentration most strongly (and contribute most to model uncertainty). Based
on remote-sensing (lidar) measurements of plume height and dimensions aloft at
the Kincaid site, both the plume rise and or algorithms were found to
contain systematic biases, with predictions differing from observed values by
an average of 20 to 40 percent.
In a follow-up study using the historical tracer data archive, Londergan
et al., (1982) compared predicted and observed dispersion coefficients for
near-ground release tracer experiments. One important finding from these
comparisons is the critical role which stability classification plays in the
prediction of dispersion coefficients. For horizontal dispersion, prediction
schemes using measured horizontal turbulence to estimate cy were more
reliable than the standard Turner method based on wind speed and solar
radiation.
-50-
-------
When the uncertainties in individual model components are examined
separately, it is sometimes surprising that Gaussian models perform as well
they do. The interdependence of different components, through formulas which
depend upon the same model inputs {such as stability class), often leads to
compensating errors which reduce the effect on predicted concentrations. For
example, the EPRI Kincaid study found that both plume rise and vertical
dispersion were generally over-predicted, and the effects of these errors (on
ground-level concentrations) were partially offsetting.
5.2 Model Limitations and Uncertainties - Air Toxics Models
Much of the previous discussion focused on difficulties and uncertainties
associated with models for relatively well behaved gases or, at least, well
controlled sources. These models assume that steady-state release and
dispersive conditions exist and that the dispersing material behaves like a
neutral buoyancy tracer. From these models, mean concentrations are
calculated. Point-by-point comparisons of observations and predictions
indicate that model performance is limited. Modeling for most air toxics
applications uses the same capabilities and assumptions identified for air
quality models and is therefore subject to the same level of uncertainties.
Other modeling applications require consideration of a list of additional
model assumptions and constraints to cover such phenomena as:
- heavy gas dispersion
- time varying release rates
- effects of warming cold gas clouds
- liquid evaporation and liquified gas vaporization
Each constraint adds a different set of requirements to modeling and the
potential for increased uncertainties.
-51-
-------
Model evaluation for dense gas releases has been attempted. The
evaluations are often chemical specific and suffer from inadequate data
bases. These problems occur because generation and dispersion of many
dense/toxic gases is scale dependent which requires testing of actual
chemicals at full or at least large scale. Often the number of tests is so
limited that variations of even the main model parameters cannot be thoroughly
examined through final test results.
The references listed in Reference Sections R.2 and R.3 provide a
description of some experimental programs and previous model evaluations.
Results for the roost part are limited to tests and models for heavy gas
dispersion, particularly LNG. Experiments with LNG were difficult to control,
particularly at the source. In many cases, they were of a scale too small to
demonstrate gravity spreading. In some larger experiments, gravity spreading
was identified but gravity effects dominated atmospheric turbulence effects
only on limited occasions. Models were compared to these data with the result
that the simple models conservatively overpredicted the distance of the
maximum impact zones. The modules were not designed nor were they effective
in simulating high frequency fluctations in pollutant concentrations.
-52-
-------
6.0 SUMMARY
This report deals with air toxics impact assessments. The primary
assessment tools considered are models to simulate releases of air toxics and
subsequent dispersion. The scope of the report is very broad in that impact
assessments may be needed for a large variety of pollutants over time periods
from a few minutes to many years. To provide a more manageable perspective,
the report is bounded by considering only dispersion models of the scale which
could reasonably be expected to provide clear impact assessments in the event
of air toxics pollutant emissions from a single source. This is not meant to
imply that the models always provide unambiguous assessments, but rather that
the models as a class represent dispersion over scales typically
representative of air toxics impacts. The resulting collection of models is
most generally representative of dispersion over distances up to fifty
kilometers from the release point.
Model descriptions are prepared by combining information on EPA models,
developed for regulatory applications, with models for unique applications
developed for chemical modeling. In general, EPA models are identified in
this report where the model assumptions are appropriate for air toxics
assessments. The literature review provides descriptions of additional models
in the following categories:
- source models for estimating emissions for toxic and flammable
materials spills (particularly pool evaporation models)
- dispersion models tied to multimedia models for impact assessments.
- models not limited by standard assumptions of typical Gaussian
models (e.g. models for instantaneous sources, models for
depositing chemicals)
- models for the disperson of chemicals which are negatively buoyant
with respect to air or those with transitional buoyancy.
-53-
-------
A general conclusion from the literature survey is .that no single,
available model can completely simulate all air toxics impacts. The variety
of chemicals and release scenarios requires a variety of models. Development
of non-traditional dispersion models has been sporadic in response to needs
related to individual chemicals or industries. For example, models for heavy
gases were oriented initially toward liquefied gaseous fuels and were
developed at a time of oil shortage.
A second and related result of the literature survey is that the models
currently identified for air toxics impact assessments are inadequately
evaluated. Studies assessing air quality models have been progressing with
mixed results, but very little data for non-traditional model evaluations is
generally available. This finding is related to the specificity of
experiments to individual chemicals (most notably LNG). Also, there are
difficulties in conducting experiments for pollutants that may be toxic,
flammable, depositing, and/or chemically reactive. Findings from available
studies have indicated that for many chemicals, dispersion behavior is scale
dependent and pollutant specific. These findings indicate that experimental
programs must be of large size and must be performed with the chemicals of
interest. The costs and logistical problems associated with such large-scale
programs have often been prohibitive.
-54-
-------
REFERENCES
The following references are provided to support information presented in
the report and to provide additional information on available models. The
reference list is divided into three sections to identify information on
modeling, field experiments used to evaluate models, and model evaluations.
The latter section is beneficial since evaluation of models identified herein
is beyond the scope of the report.
-55-
-------
REFERENCES.1
MODELS
1. American Gas Association, 1974: LNG Safety Program, Interim Report on
Phase II Work. Report to the American Gas Association Project IS-3-1 by
Battelle Columbus Laboratories.
2. Arthur D. Little, Inc., 1978: Simplified Methods for Estimating Vapor
Concentration and Dispersion Distances for Continuous LNG Spills in Dikes
with Flat or Sloping Floors. Prepared for American Gas Association.
Report X50978.
3. Ashworth P. 1982: A Dispersion Model for Sinker Liquids Spilled into
Waterways. 1982 Hazardous Materials Spills Conference. Milwaukee pp.
404-413.
4. Association of American Railroads Research & Test Department, 1984:
Industrial Chemical Accident Response Information System (ICARIS).
Revised 1985. Washington, DC.
5. Battelle Institut e.V., Frankfurt, 1978: Risk Assessment Study for an
Assumed LNG Terminal in the Lysekil Area Prepared for the Swedish Energy
Commission, Stockholm, Sweden.
6. Bicknell B.R., S.H. Boutwell, D.B. Watson, 1985: Testing and evaluation
of TOX-SCREEN Model, U.S. EPA - EPA-600/3-85-001.
7. Bleeker, D.E., 1984: A Real-Time Air Dispersion Modeling System Final
Report, Sierra Geophysics, Inc., Redmond, WA. USAF Report
AFESC/ESL-TR-83-63.
8. Blewitt, D.N., 1985: Computerized Modeling Rupture Design Analysis for
Sour Gas Pipeline Safety Analysis. Standard Oil Company of Indiana.
Chicago, IL.
9. Bowers, J., J. Bjorklund, C. Cheney 1979: Industrial Source Complex (ISC)
Dispersion Model User's Guide Vol. I. US EPA report EPA-450/4-79-030.
10. Briggs, G.A., 1973: Diffusion Estimation for Small Emissions. Air
Resources Atmospheric Turbulence and Diffusion Laboratory, National
Oceanic and Atmospheric Administration, Oak Ridge, TN.
11. Chan, S.T., and D.L. Ermak, 1983: Recent Progress in Modeling the
Atmospheric Dispersion of Heavy Gases over Variable Terrain Using
Three-Dimensional Conservation Equations. Prepared for the I.U.T.A.M.
Symposium on Atmospheric Dispersion of Heavy Gases and Small Particles,
The Netherlands.
12. Chlorine Institute. 1982: Calculating the Area Affected by Chlorine
Releases. The Chlorine Institute Pamphlet 74, New York.
13. Clewell, H.J., 1983: A Simple Formula for Estimating Source Strengths
from Spills of Toxic Liquids. U.S.A.F., AFESC/ESL-TR-83-03.
-56-
-------
14. Colenbrander, G.W., 1980: A Mathematical Model for the Transient Behavior
of Dense Vapor Clouds. 3rd International Symposing on Loss Prevention and
Safety. Base, Switzerland.
15. Colonna, G.R., 1982: Hazard Assessment Computer System (HACS) - Models
Review and Validation. Hazardous Materials Spills Conference. Milwauke.
16. Colonna, G.R., F.T. Dodge, T.B. Morrow, J.C. Buckingham and J.A. Havens,
1984. Hazardous Materials Spill Conference, Nashville, TN.
17. Cox, R.A., D.E. Shillitan, P.J. Comber, and D.H. Slater, 1977: Novel
Methods for Assessment of Air Pollution in Planning Situations Where
Conventional Dispersion Models are Inapplicable. 4th International Clean
Air Congress, Tokyo.
18. Cox, R.A., and R.J. Carpenter, 1980: Further Development of a Dense Cloud
Dispersion Model for Hazard Analysis. Heavy Gas and Risk Assessment, S.
Hartwig Ed., D. Reidel Publishing.
19. Dainis, G.A. Ill and R.C. Reid, 1981: The Boiling and Spreading of
Liquefied Natural Gas on Water. Prepared for Gas Research Institute,
Chicago, IL.
20. Dana, M.T., R.N. Lee, and J.M. Hales, 1984: Hazardous Air Pollutants:
Wet Removal Rates and Mechanisms., U.S. EPA. EPA-600/3-84-11.
21. de Nevers, N. , 1984: Spread and Downslope Flow of Negatively Buoyant
Clouds. Atm. Envir, 18 (10): 2023-2027.
22. Department of Defense, 1980: Methodology for Chemical Hazard Prediction,
Explosives Safety Board, Technical Paper-10. DDESB-TP-10-CHG-3.
23. Dodge, F.T., J.C. Buckingham, and T.B. Morrow, 1982: Analytical and
Experimental Study to Improve Computer Models for Mixing and Dilution of
Soluble Hazardous Chemicals. U.S.C.G. Report CG-D-1-93.
24. Duke, J., 1985: Estimating Downwind Impact Distances from Fuming Acid
Spills. 78th Annual Meeting of the Air Pollution Control Association,
Detroit, Michigan, Paper 85-25B.1.
25. Dumbauld, R.K., J.R. Bjorklund, H.E. Cramer and R.A. Record, 1970:
Handbook for Estimating Toxic Fuel Hazards, Final Report Prepared for
George C. Marshall, Space Flight Center, National Aeronautics and Space
Administration, Marshall Space Flight Center, Alabama. GCA
Report TR-69-16N.
26. Eidsvik, K.J., 1980: A Model for Heavy Gas Dispersion in the Atmosphere.
Atmospheric Environment, 14:769-777.
27. Eidsvik, K.J., 1981: A Model for Heavy Gas Dispersion in the Atmosphere.
Atmospheric Environment, 15(7):1163-1164.
-57-
-------
28. England, W.G., L.H. Teuscher, L.E. Hauser, and B.F. Freeman, 1978:
Atmospheric Dispersion of Liquefied Natural Gas Vapor Clouds Using SIGMET,
A Three-Dimensional Time-Dependent Hydrodynamic Computer Model.
Proceedings of the 1978 Heat Transfer and Fluid Mechanics Institute held
at Washington State University, Pullman, WA.
29. Fleischer, M.T., 1980: SPILLS: An Evaporation/Air Dispersion Model for
Chemical Spills on Land. Shell Development Company, Houston, TX.
30. Flothmann, D. and H. Nikodem, 1980: A Heavy Gas Dispersion Model with
Continuous Transition from Gravity Spreading to Tracer Diffusion. Heavy
Gas and Risk Assessment. S. Hartwig (ed). D. Reidel Publishing.
31. Frame, G.B., 1980: Determination of Risk from A Chlorine Spill or Major
Release. Chemistry in Canada, 32(8):27-29.
32. Fryer, L.S., and D. Kaiser, 1979: DENZ - A Computer Program for the
Calculation of the Dispersion of Dense Toxic or Explosive Gases in the
Atmosphere. United Kingdom, Atomic Energy Authority, Culcheth,
Warrington. SRD R 152.
33. Georgakis, C. , J. Congalidis, and G.C. Williams, 1979: Model for
Non-Instantaneous LNG and Gasoline Spills.
34. Germeles, A.E. and E.M. Drake. 1975: Gravity Spreading and Atmospheric
Dispersion of LNG Vapor Clouds. 4th International Symposium on Transport
of Hazardous Cargo be Sea and Inland Waterway. Jacksonville, FL.
35. Gudiksen, P., R. Large, M. Dickerson, T. Sullivan, L. Rosen, H. Walker,
G.B. Boeri, R. Caracciola, and R. Fiorenza, 1982: Joint Research and
Development on Toxic Material Emergency Between ENEA and LLNL: 1982
Progress Report. Lawrence Livermore Laboratory, Livermore, CA.
36 Hannum, J.A., 1981: Air Dispersion Modeling S & EPS Workshop Held at
Panama City. Chemical Propulsion Information Agency, Laurel, MD.
37. Harding, R.V., M.C. Parnarouskis, and R.G. Potts, 1978: The Development
and Implementation of the Hazard Assessment Computer System (HACS).
Control of Hazardous Materials Spills Conference Miami.
38. Havens, J. , 1979: A Description and Assessment of the SIGMET LNG Vapor
Dispersion Model. USCG Report CG-M-3-79.
39. Havens, J.A., and T.O. Spicer, 1985: Development of an Atmospheric
Dispersion Model for Heavier-than-Air Gas Mixtures Final Report Prepared
for Commandant (GFCP-22F/TP64), U.S. Coast Guard, Washington, D.C.
40. Havens, J.A. and P.J. Schreus, 1984: Evaluation of 3-D Hydrodynamic
Computer Models for Prediction of LNG Vapor Dispersion in the Atmosphere,
Annual Report prepared for Gas Research Institute, Chicago, IL.
41. Ille, G. and C. Springer, 1978: The Evaporation and Dispersion of
Hydrazine Propellants from Ground Spills. Final Report. Air Force Civil
and Environmental Engineering Development Office, Florida. CEEDO-TR-78-30.
-58-
-------
42. Johnson, C. , D. Latimer, R. Bergstrom, and H. Hogo, 1980: User's Manual
for the Plume Visibility Model (PLUVUE). U.S. Environmental Protection
Agency EPA 450/4-80-0032.
43. Jones, P.H. 1981: Planning Emergency Response Systems for Chemical
Accidents. European Cooperation Environmental Health Aspects of the
Control of Chemicals - Interim Document 1. World Health Organization,
Copenhagen.
44. Kahler, J.P., R.G. Curry and R.A. Handler, 1980: Calculating Toxic
Corridors. Air Weather Service, USAF AWS/TR-800/003.
45. Kaiser, G.D., and B.C. Walker, 1978: Releases of Anhydrous Ammonia from
Pressurized Containers - The Importance of Denser-Than-Air mixtures.
Atmospheric Environment, 12:2289-2300.
46. Kansa, E.J., H.C. Rodean, S.T. Chan, and D.L. Ermak, 1983: Atmospheric
Dispersion of Ammonia: An Ammonia Fog Model. Lawrence Livermore National
Lab, Livermore, CA.
47. Kansas State University, 1981: A Community Model for Handling Hazardous
Materials and Transportation Emergencies, Prepared for Department of
Transportation, Washington, DC.
48. Kelty, J., 1980. Hazardous Materials Response Guide Illinois
Environmental Protection Agency, Springfield, IL.
49. Kelty, J. 1984: Calculation of Evacuation Distances During Toxic Air
Pollution Incidents. Atmospheric Emergencies: Existing Capabilities and
Future Needs. Transportation Research Record 902. National Academy of
Science. Washington, D.C.
50. Kricks, R.J., S. Pan, and T. Minnich, 1983: Air Quality Modeling of
Chemical Spills: Determination of the Thermophysical Properties of
Chemicals Not Included in the Data Base of the Shell SPILLS Model. 75th
Annual Meeting of the Air Pollution Control Association, Atlanta.
51. Kunkle, B.A., 1984: An Evaluation of the Ocean Breeze/Dry Gulch
Dispersion Model. Air Force Geophysics Laboratory, Hanscom AFB, MA.
52. Kunkle, B.A. 1983: A Comparison of Evaporative Source Strength Models for
Toxic Chemical Spills. USAF. Air Force Systems Command AFGL-TR-83-0307.
53. Ludwig, F.L. 1983: Information for Users of the SRI Puff Model. SRI,
International. Menlo Park, CA.
54. Ludwig, F.L. 1984: Transport and Diffusion Calculations During Time and
Spare Varying Meteorological Conditions Using a Microcomputer. JAPCA, 3_1
(3): 5-10.
55. Ludwig, F.L., L.S. Gasiovek and R.E. Ruff, 1977: Simplification of a
Gaussian Puff Model for Real Time Minicomputer Use. Atm. Envir. , 11:
431-436.
-59-
-------
56. Marrs, J.T., E.G. Stenmark, and F.V. Hansen, 1983: Toxic Corridor
Prediction Programs. Atmospheric Emergencies: Existing Capabilities and
Future Needs. Transportation Research Record 902. National Academy of
Science. Washington, D.C.
57. Massachusetts Institute of Technology Department of Chemical Engineering,
1978: Confined Boiling Rates of Liquefied Petroleum Gas on Water Prepared
for the National Science Foundation and U.S. Department of Energy,
Washington, D.C. HCP/P4548-01.
58. McMahon, T.A. and P.J. Denison, 1979: Empirical Atmospheric Deposition
parameters: A Survey. Atm. Envir., 13: 571-585.
59. Morgan, D.L., E.J. Kansa, L.K. Morris, 1983: Simulations and Parameter
Variation Studies of Heavy Gas Dispersion Using the SLAB Model - Condensed
Atmospheric Dispersion of Heavy Gases and Small Particles. The Hague,
Netherlands. (UCRL - 90150).
60. Pan, S.C., R.J. Kricks, and T.R. Minnich, 1983: Air Quality Modeling of
Chemical Spills: Sensitivity Analyses of Thermophysical Property
Parameter used as Input to the Shell SPILLS Model. 76th Annual Meeting of
the Air Pollution Control Association, Atlanta.
61. Pepper, D.W., 1980: Automated Emergency Meteorological Response System.
National Petroleum Refiners Association Meeting, Philadelphia.
62. Peterson, W. , 1982: Estimating Concentrations Downwind from an
Instantaneous Puff Release, EPA 600/3-82-078.
63. Petersen, W.B., J.A. Catalano, T. Chico and T.S. Yuen, 1984: INPUFF - A
Single Source Gaussian Puff Dispersion Algorithm. User's Guide, United
States Environmental Protection Agency. EPA-600/8-84-027.
64. Pickett, E.E., R.G. Whiting, and L.H. Kocchiu, 1982: Detection and Impact
Prediction of Hazardous Substances Released to the Atmosphere. The
Science of the Total Environment, 2_3:141-149. Elsevier Scientific
Publishing Company, Amsterdam.
65. Potts, R., 1981; Hazard Assessment Computer System Reports. USCG-D-74-81,
USCG-D-75-81, USCG-D-76-81
66. Potts, R.G., 1981: Development of a HACS User Interface Module. Arthur
D. Little, Inc., Final Report prepared for Department of Transportation,
Washington, D.C.
67. Potts, R., 1981: Hazard Assessment Computer System HACS/UIM Users'
Operation Manual, Volume III. Arthur D. Little, Inc. Final Report
prepared for Department of Transportation, United States Coast Guard,
Washington, D.C.
68. Raj, P.K. and R.C. Reid, 1978: Underwater Release of LNG. Control of
Hazardous Material Spills Conference. Miami.
-60-
-------
69. Raj, P.K. , and R. C. Reid, 1978: Fate of Liquid Ammonia Spilled onto
Water. Environmental Science & Technology, 12(13):1422-1425, American
Chemical Society.
70. Ramsdell, J.V. and W.T. Hinds, 1971: Concentration Fluctuations and
Peak-to-Mean Concentration Ratios in Plumes from a Ground-Level Continuous
Point Source. Atmospheric Environment. 5(7):483.
71. Ramsdell, J.V., O.F. Athey and C.S. Glantz, 1983 MESOI Version 2.0: An
Interactive Mesoscale Lagrangian Puff Dispersion Model with Deposition and
Decay, NUREG/CR-3344, U.S.N.R.C.
72. Randerson, (ed.) D. , 1984: Atmospheric Science and Power Production.
U.S. Department of Energy, DOE/TIC-27601.
73. Rausch, A.H., N.A. Eisenberg, and C.L. Lynch, 1977: Continuing
Development of the Vulnerability Model. Enviro Control, Inc., Rockland,
MD.
74. Reid, R.C. and K.A. Smith, 1978: Behavior of LPG on .Water. Hydrocarbon
Processing: 117-121.
75. Reinhart, R., J. Piepers and L.H. Toneman, 1980: Vapor Cloud Dispersion
and the Evaporation of Volatile Liquids in Atmospheric Wind fields - I.
Theoretical Model. Atmospheric Environment, 14:751-758.
76. Rottman, J.W., J.C.R. Hunt,, and A. Mercer 1985. The Initial and Gravity-
Spreading; Phases of Heavy Gas Dispersion: Comparison of Model with Phase
I Data. Journal of Hazardous Materials, 11; 261-279.
77. Schnatz, G. and D. Flothmann, 1980: A K-model and Its Modification for
the Dispersion of Heavy-gases. Heavy Gas and Risk Assessment, S. Hartwig,
(ed), D. Reidel Publishing.
78. Sehmel, G.A., 1980: Particle and Gas Dry Deposition: A Review. Atm.
Envir., 14: 983-1011.
79. Sehmel, G.A., R.N. Lee, and T.W. Horst, 1984: Hazardous Air Pollutants:
Dry Deposition Phenomena, U.S. EPA. EPA-600/3-84-114.
80. Sethuraman, S., G.S. Raynor and R.M. Brown, 1982: Variation of Turbulence
in a Coastal Thermal Internal Boundary Layer Third Joint Conf. in Applic.
of Air Pollution Meteorology, AMS, Boston.
81. Shaver, O.K., E.E. Harten, Jr., R.L. Berkowitz, and T.J. Rudd, 1982: Post
Accident Procedures for Chemicals and Propellants. Systems Technology
Laboratory, Inc., Report F04611-80-C-0046, Arlington, VA.
82. Shaw, P., And F. Briscoe, 1978: Evaporation of Hazardous Liquids on Land
and Water. U.K. Atomic Energy Authority, Safety and Reliability
Directory, SRD-R-100.
83. Shieh, C.M., 1978: A Puff Pollutant Dispersion Model with Wind Shear and
Dynamic Plume Rise. Atm. Envir., 12: 1933-1938.
-61-
-------
84. Shen, T.T., 1982: A Simplified Method for Estimation of Hazardous
Emissions from Waste Lagoons. 75th Annual meeting of the Air Pollution
Control Association, New Orleans, Louisiana.
85. Sherman, C.E., 1978: A Mass Consistent Model for Wind Fields Over
Complex Terrain, J. Appl. Met., 17: 317-319.
86. Simmons, J. , 1974: Risk Assessment of Storage and Transport of Liquefied
Natural And LP-Gas. Science Applications, Incorporated, Prepared for
Office of Radiation Programs, Springfield, VA. EPA-520/3-75-015.
87. Slade, D.H., 1968: Meteorology and Atomic Energy 1968, U.S.A.E.C,
TID-24190.
88. Smith, D.G., 1978, Turbulent Dispersion Around a Building in the Natural
Wind and Effluent Reentry. Ph.D. thesis, Harvard Univ. (School of Public
Health, Boston, MA).
89. Soldat, J.K., 1976: Methodology for Calculation of Radiation Doses in
the Environs from from Nuclear Fuel Cycle Facilities. Pacific Northwest
Laboratory. BNWL-2075. Richland, WA.
90. Stiver, W. and D. MacKay, 1983: Evaporation Rates of Chemical Spills.
Environment Canada 1st Technical Chemical Spills Seminar, Toronto.
91. Tang, I.N., W.T. Wong, H.R. Munkelwitz, M.F. Flessner, 1980: Sulfuric
Acid Spills in Marine Accidents. Sixth International Symposium on the
Transport of Dangerous Goods by Sea and Island Waterways. Tokyo.
92. Thompsen, E.S., 1984: Evacuation Distances for Spills of Hazardous
Chemicals. Atmospheric Emergencies: Existing Capabilities and Future
Needs. Transportation Research Record 902. National Academy of
Science. Washington, D.C.
93. Tucker, W.A., A.Q. Eschenroeder, and G.C. Magel, 1984: Air,Land, Water
Analysis System (ALWAS): A Multi-media Model for Toxic Substances. U.S.
EPA. EPA-600/3-84-052.
94. U.S. Environmental Protection Agency, 1984: Compilation of Air Pollutant
Emission Factors, 3rd ed, AP-42.
95. U.S. Environmental Protection Agency, 1978. Guideline on Air Quality
Models, EPA-450/2-78-027.
96. Van Ulden, A.P., 1974: On the Spreading of Heavy Gas Released Near the
Ground. 1st International Loss Symposium. The Hague, Netherlands:
211-226.
97. Wada, R.Y. 1983: Atmospheric Problems from Hazardous Material Spills in
the San Francisco Bay Area. Atmospheric Emergencies: Existing
Capabilities and Future Needs. Transportation Research Record 902.
National Academy of Sciences. Washington, D.C.
-62-
-------
98. Webster, R.D., R.L. Welsh, P.K. Terkonda, 1978: AIRMOD-A General Program
for the Rapid Assessment of Airborne Pollutants, Interim Report.
Construction Engineering Research Laboratory, Champaign, IL.
99. Whitacre, C.G., and M.M. Myriski, 1982: Computer Program for Chemical
Hazard Prediction. U.S. Army Report ARCSL-TR-82014.
100. Whittaker, J. 1976: A Risk Approach to Land Use in the Vicinity of Sour
Gas Facilities, Alberta Energy Resource Conservation Board.
101. Wilson, D.J. 1981. Along-wind Diffusion of Source Transients. Atmos.
Environ., 15: 489-495.
102. Wu, J.M. and J.M. Schroy, 1979: Emissions from Spills. Specialty
Conference on Control of Specific (Toxic) Pollutants. Air Pollution
Control Association.
103. Zeman, 0., 1982: The Dynamics and Modeling of Heavier-than-Air Cold Gas
Releases. Atm. Environ. 16 (4): 741-751.
-63-
-------
REFERENCES.2
FIELD EXPERIMENTAL PROGRAMS
104. Colenbrander, G.W. and J.S. Puttock, 1983: Dense Gas Dispersion
Behavior: Experimental Observations and Model Developments.
International Symposium on Loss Prevention and Safety, Harrogate, U.K.
105. Colenbrander, G.W. and J.S. Puttock, 1983, Maplin Sands Experiments
1980: Interpretation and Modeling of Liquefied Natural Gas Spills onto
the Sea. IUTAM Symposium on Atmospheric Dispersion of Heavy Gases.
Delft, The Netherlands.
106. Dancer, A.K., E.M. Drake and R.C. Reid, 1977: Boiling of Liquid Nitrogen
and Methane on Water. The Effect of Initial Water Temperature. Int. J.
Heat. Mass Transfer, 20: 177-180.
107. Directorate of Labor of the Ministry of Social Affairs: Experiments with
Chlorine. Voobuay, Netherlands.
108. Drake, E.M., A. Jeje, and R.C. Reid, 1975: Transient Boiling of
Liquefied Cryogens on a Water Surface. Vol. I. Nitrogen, Methane and
Ethane. Int. J. Heat Mass Transfer, 18: 1361-1368. Vol. II Light
Hydrocarbon Mixtures, Int. J. Heat Mass Transfer, 18: 1369-1375.
109. Duffey, A.R, D.N. Gideon, and A.A. Putnam, 1974: Dispersion and
Radiative Experiments. Section C. LNG Safety Program, Interior Report of
Phase II work. American Gas Association. AGA No. M19874. Arlington, VA.
110. Ermak, D.L., H.C. Goldwire, W.J. Hogan, R.P. Koopman and T.G. McRae.
1982: Results of 4(
Laboratory UCRL-87640.
1982: Results of 40 m3 LNG Spills onto Water. Lawrence Livermore
111. Ermak, D.L., R.P. Koopman, T.G. McRae, W.J. Hogan, 1982: LNG Spill
Experiments: Dispersion, RPT, and Vapor Burn Analysis. LLL UCRL-87608.
112. Hall, D., 1979: Further Experiments on a Model of an Escape of Heavy
Gas. Warren Spring Laboratory, Stevenherge (UK) LR-312 (AP).
113. Humbard-Basset R. and A. Mantel 1985: Penetration of the Atmosphere by
Flammable Mixtures as a Result of Spillage of LNG on the Ground. Gaz de
France, Processed by NASA (N85-17432).
114. Koopman, R.P., H.C. Goldwire, and T.G. McRae, 1983: Large-scale
Hazardous Gas Release Experiments. JANNAF, Safety and Environmental
Protection Subcommittee Meeting, LLL, Livermore, CA.
115. Koopman, R.P., T.G. McRae, H.C. Galdwire, D.L.Ermak and E.J. Kansa.
1984: Results of recent Large-Scale NH3, and N204 Dispersion
Experiments. Third Symposium on Heavy Gases and Risk Assessment.
Wissenschaftszentrum, West-Germany.
-64-
-------
116. Lewellen, W.S., R.I. Sykes, S.F. Parker, 1985; Comparison of the 1981
INEL Dispersion Data with Results from a Number of Different Models.
USNRC NUREG/CR-4159.
117. Maybark, J., K. Yoshida, and R. Graner 1978: Spray Drift from
Agricultural Pesticide Applications. J. Air Pollution Control
Association, 28 (10): 1009-1014.
118. McNerney, J. , D. Towson, and W. Henderson, 1966: Nitrogen Tetroxide
Evaporation Rate Studies, E ATM 511-1, Edgewood Arsenal (DTIC AD488566).
119. McQuaid, J., 1985: Heavy Gas Dispersion Trials at Thorney Island.
Journal of Hazardous Material, 11.
120. McRae, T.G., H.C. Goldwire, and R.P. Koopman, 1984. The Evaporation and
Gaseous Dispersion of Large-Scale Releases of Nitrogen Tetroxide. JANNAF
S & EPS Annual Meeting. Las Cruces. UCRL 89687.
121. Morgan, D.L. 1984: Dispersion Phenomenology of LNG Vapor in the Burro
and Coyote LNG Spill Experiments 1984 Winter Annual ASME Meeting, New
Orleans. UCRL - 91741.
122. Nickola, P.W., J.V. Ramsdell, Jr., and J.D. Ludwick, 1970, Detailed
Time-Histories of Concentrations Resulting from Puff and Short-Period
Releases of an Inert Radioactive Gas, Report BNWL-1272, Battelle Pacific
Northwest Laboratories. Richland, VA.
123. Picknett, R.G. 1978: Field Experiments on the Behavior of dense Clouds.
Report IL1154/78/1 Chemical Defense Establishment, Porton Downs.
124. Picknett, R.G. 1981. Dispersion of Dense Gas Puffs Released in the
Atmosphere at Ground Level, Atmospheric Environment, 15: 509-525.
125. Puttock, J.S. and G.W. Colenbrander, 1985: Thorney Island Data and
Dispersion Modeling, J. Haz. Mat., 11: 381-397.
126. Puttock, J.S., D.R. Blackmore, G.W. Colenbrander 1982: Field Experiments
on Gas Dispersion. J. Haz. Mat., 6: 13-41.
127. Reid, R.C. and R. Wang, 1978: The boiling rates of LNG on typical dike
floor materials. Cryogenics: 401-404.
128. Rodear, H.C. 1983: Effects of a spill of LNG on Mean flame and
Turbulence under Low Wind Speed, Slightly Stable Atmospheric Conditions.
IUTAM Symposium Atmospheric Dispersion of Heavy Gases and Small
Particles, The Hague, The Netherlands.
129. Welker, J.R. 1982: Vaporization, Dispersion, and Radiant Fluxes from LPG
Spills. US DOE Report DOE/EV/07020-1.
-65-
-------
130. Witcofski, R.O. and J.E. Chirinella, 1982: Experimental and Analytical
Analyses of the Mechanisms Governing the Dispersion of Flammable Clouds
Formed by Liquid Hydrogen Spills. Work Hydrogen Energy Conference IV,
Pasadena, CA.
-66-
-------
REFERENCES.3
MODEL EVALUATIONS
131. Alp, E., R.B. Caton, R.V. Portelli, S.G. Guerin, A. Mitchell, and C.
Doherty, 1983: A Comparison of Conventional Spill Air and Water
Dispersion Models 1st EPA Technical Symposium on Chemical Spills.
132. Balentine, H.W. and M.W. Eltgroth, 1985: Validation of a Hazardous Spill
Model using N204 and LNG Spill Data, APCA annual meeting - Detroit
Paper. 85-25B.1.
133. Bander, T.J. 1979: Literature Applicable to Descriptions of Diffusion of
Uncontrolled Natural Gas Releases Containing Hydrogen Sulfide, Battelle
Report to the Energy Resources Conservation Board, Calgary Alberta.
134. Bass, A. and D.G. Smith, 1980 National Environmental Studies Project.
Atmospheric Dispersion Modeling for Emergency Preparedness. Atomic
Industrial Forum.
135. Beychok, M.R., 1979: Fundamentals of Stack Gas Dispersion, Milton R.
Beychok, Irvine, CA.
136. Bovme, N.E., R.J. Londergan, D.R. Murray, H.S. Borenstein, 1983:
Overview Results and Conclusions for the EPRI Plume Model Validation and
Development Project: Plains Site, EA-3074, Electric Power Research
Institute, Palo Alto, CA, 1983.
137. Fox, D.G., 1981: Judging Air Quality Model Performance, Bulletin
American Meteorology Society, 62^ (5).
138. Fabrick, A., 1982: CHARM: An Operational Model for Predicting the Fate
of Elevated or Surface Releaes of Dense or Buoyant Clouds. 3rd Joint
Conference on Air Pollution Meteorology, San Antonio, TX.
139. Hanna, S.R. and B. Munger, 1983:. A Survey of Emergency Models and
Applications - Part I: Models Developed for Hazardous Spills
Applications. Paper 83-261, 76th Annual Meeting of the Air Pollution
Control Association. Atlanta.
140. Havens, J.A., 1977: Predictability of LNG Vapor Dispersion from
Catastrophic Spills onto Water: An Assessment USCG report. Office of
Merchant Marine Safety.
141. Lewellen, W.S., and R.I. Sykes, 1982: A Scientific Critique of Available
Models for Real-time Simulations of Dispersion. U.S. NRC NUREG/CR-4157.
142. Lewellen, W.S., R.I. Sykes, S.F. Parker 1985: Comparison of the 1981
INEL Dispersion Data with Results from a number of Different Models U.S.
N.R.C NUREG/CR-4159.
143. Liu, M.K. and G.E. Moore, 1982: Diagnostic Validation of Plume Models at
a Plains Site, Pub. No. 82162, Systems Applications, Inc., San Rafael, CA.
-67-
-------
144. Londergan, R.J., J.J. Mangano, N.E. Bowne, D.R. Murray, H. Bornestein,
1980: An Evaluation of Short-Term Air Quality Models Using Tracer Study
Data, API Report No. 4333, American Petroleum Institute, Washington, D.C.
145. Londergan, R.J., D.H. Minott, D.J. Wackter, T.M. Kincaid, D.M. Bonitata,
1982. Evaluation of Rural Air Quality Simulation Models,
EPA-450/4-83-003, U.S. Environmental Protection Agency, Research Triangle
Park, NC.
146. McNaughton, D.J. and C.M. Berkowitz, 1980: Overview of U.S. Research
Activities in the Dispersion of Dense Gases. Heavy Gas and Risk
Assessment. S. Hartwig (ed.) D. Reidel Publishing.
147. McRae, T.G., 1985: Analysis and Model/Data Comparisons of Larger-Scale
Releases of Nitrogen Tetroxide. U.S.A.F. ESL-TR-85-06..
148. Minott, D.H., R.J. Londergan, D.J. Wackter, R.R. Fizz, 1983. Evaluation
of Urban Air Quality Simulation Models, EPA Publication No.
EPA-450/4-83-020, U.S. Environmental Protection Agency, Research Triangle
Park, NC.
149. Ohmstede, W.D., R.K. Dumbauld, G.G. Worley, 1983: Ocean Breeze/Dry Gulch
Equation Review. U.S.A.F. Reporr AFESC/ESL-TR-83-05.
150. Reithmuller, M.L. (1983): Critical Confrontation of Standard and More
Sophisticated Methods for Modeling the Dispersion Ion Air of Heavy Gas
Clouds. Commission of the European Communities. Luxenburgh.
EUR-8423-EN.
151. Schulman, L.L., J.S. Scire, 1981: Technical Evaluation of the ISC Model
at an Industrial Complex, API Report No. 4341, American Petroleum
Institute, Washington, DC.
152. Stoner, R.R., B.L. Orndorff, and T.E. Pierce, 1984 Meteorological Aspects
of Emergency Preparedness and Response: An Overview.
153. Wackter, D.J., R.J. Londergan, 1984: Evaluation of Complex Terrain Air
Quality Simulation Models, EPA-450/4-84-017, U.S. Environmental
Protection Agency, Research Triangle Park, NC.
154. Woodward, J.L., J.A. Havens, W.C. McBride, and J.R. Taft. 1982: A
comparison with Experimental Data of Several Models for Dispersion of
Heavy Vapor Clouds. J. of Haz. Mat., 6: 161-180.
-68-
-------
TECHNICAL REPORT DATA
(Please read Instructions or, the reverse before completing)
1. REPORT NO.
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Some Applications of Models to Air Toxics Impact
Assessments
5. REPORT DATE
Mav IQSfi
May
i. PERFI
ORMING ORGANIZATION CODE
7. AUTHOR(S)
Daniel 0. McNaughton, Marshall A. Atwater,
Richard J. Londergan
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
TRC Environmental Consultants, Inc.
800 Connecticut Blvd.
East Hartford, CT 06108
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-3886
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
OAQPS, MDAD (MD-14)
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This report identifies models that are available for toxics impact assessments and
factors that should be considered in selecting models for specific applications.
There is no claim as to the merits of individual models or that the list of models
is comprehensive. This report only provides information that may be considered
useful to air pollution control programs concerned with air toxics and should not
be construed as providing regulatory guidance.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Air Toxics
Mathematical Modeling
Meteorology
Air Quality
Impact Assessment
18. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
68
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
------- |