EPA-910/9-88-202R
United States
Environmental Protection
Agency
Region 10
1200 Sixth Avenue
Seattle WA 98101
Alaska
Idaho
Oregon
Washington
January 1991
USER'S GUIDE FOR THE
FUGITIVE DUST MODEL (FDM)
(REVISED)
USER'S INSTRUCTIONS
-------
EPA-910/9-88-202R
User's Guide for the
Fugitive Dust Model (FDM)
(Revised)
User's Instructions
By:
Kirk D. Winges
TRC Environmental Consultants, Inc.
Mountlake Terrace, WA 98043
EPA Contract No. 68-02-4399/23
EPA Technical Representative: Robert B. Wilson
Prepared for:
Region 10
U.S. Environmental Protection Agency
1200 Sixth Avenue
Seattle, Washington 98101
U S. Environmental Protection Agency
tw-on 5, Library (P!.-l-•-''
•,>,-".t Jackson Bcthtv.v:i, Uili hoor
olL 60604-3590
May, 1990
(revised January, 1991)
-------
DISCLAIMER
This report has been reviewed by the Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, and has been approved for publication as received from the
contractor. Approval does not signify that the contents necessarily reflect the views and poli-
cies of the Agency, neither does mention of trade names or commercial products constitute
endorsement or recommendation for use.
EPA-910/9-88-202R
-------
PREFACE
The Fugitive Dust Model (FDM) was developed over the past several years by Mr. Kirk
D. Winges of TRC Environmental Consultants. Parts of this development, the documentation
of the model and three performance evaluation studies, were funded by U.S EPA Region 10.
FDM was developed to meet the specific need for a regulatory air dispersion model capable of
simulating sources of fugitive dust (for example, surface mines), using theoretically sound
principles for modeling the deposition of particulate matter. While the model may undergo
formal rule-making action at a later date, at this time it does not have regulatory status as do
models listed in Appendix A of the Guideline on Air Quality Models (Revised). The Guideline
should be consulted regarding the use of alternative models such as FDM for regulatory appli-
cations.
The model and user's guide are available from EPA's electronic bulletin board system
Support Center for Regulatory Air Models
U.S. EPA, OAQPS(MD-14)
Research Triangle Park, NC 27711
(919) 541-5742, FTS 629-5742
FDM is also available from the National Technical Information Service (NTIS) as docu-
ment number PB90-215203, or with program diskette as PB90-502410. Mail your request to:
NTIS
U.S. Department of Commerce
5285 Port Royal Road
Springfield, Virginia 22161
Any questions or comments on FDM should be sent to:
Mr. Robert B. Wilson
U.S. EPA Region 10 (ES-098)
1200 Sixth Avenue
Seattle, Washington 98101-3188
(206) 553-1531, FTS 399-1531
-------
TABLE OF CONTENTS
«* T • User * s Instructions
1 . 0 INTRODUCTION ...................................... 1
2 . 0 TECHNICAL DESCRIPTION ............................. 3
3.0 USER'S INSTRUCTIONS ............................... 21
4 . 0 VALIDATION/ SAMPLE RUNS ............................ 39
REFERENCES ........................................ 41
APPENDIX A: VALIDATION STUDIES
APPENDIX B: SAMPLE INPUT/OUTPUT RUNS
APPENDIX C: RELEVANT SECTIONS FROM THE CALINE3 USER'S GUIDE
APPENDIX D: COMPLETE LISTING OF THE FDM COMPUTER PROGRAM
-------
1.0 INTRODUCTION
The Fugitive Dust Model (FDM) is a computerized air quality model specifically designed for
computing concentration and deposition impacts from fugitive dust sources. The sources may be point,
line or area sources. The model has not been designed to compute the impacts of buoyant point sources,
thus it contains no plume-rise algorithm. The model is generally based on the well-known Gaussian Plume
formulation for computing concentrations, but the model has been specifically adapted to incorporate an
improved gradient-transfer deposition algorithm. Emissions for each source are apportioned by the user
into a series of particle size classes. A gravitational setting velocity and a deposition velocity are calculated
by FDM for each class. Concentration and deposition are computed at all user-selectable receptor
locations.
The model is designed to work with pre-processed meteorological data or with card-images of
meteorological data either hourly or in STability ARray (STAR) format. FDM accepts hourly meteorological
data output from the EPA RAMMET meteorological pre-processor program. In addition to a standard printed
output, the model allows a "plotter" output file which consists of a series of records containing only the x-
coordinate, the y-coordinate and an average concentration. This series of records is printed for each
averaging time requested. The model allows printer and plotter output for 1 -hour averages, 3-hour averages,
8-hour averages, 24-hour averages and a long-term average which is the average over the entire
meteorological data base provided. Additionally, a sequential output "tape" file for post-processing with the
POSTZ program can be created. Up to 1200 receptors can be processed, and up to 121 sources can be
processed. It should be noted that while FDM has the capability of treating 1200 receptors, POSTZ can only
accept 200 receptors, thus long-term sequential uses of FDM should carefully consider the number of
receptors to be used.
The sources can be of three types: points, lines or areas. The line source and area source
algorithms are based on algorithms in the CALINE3 Model (California Department of Transportation, 1979).
-------
2
For area sources, the user supplies the coordinates of the center and the dimension in the x and y
directions. Area sources need not be square, but rather can be rectangular, up to an aspect ratio of 1 to
5 (ratio of width to length). Area sources with the length greater than five times the width must be divided
in a series of area sources, or modeled as a line source. The model divides the area source into a series
of line sources perpendicular to the wind direction. Emissions from all sources may be divided into a
maximum of 20 particle size classes.
-------
2.0 TECHNICAL DESCRIPTION
Basic Equations
The Fugitive Dust Model (FDM) is an analytical air quality model specifically designed for the
analysis of the dispersion of fugitive dust. The model incorporates a detailed deposition routine based on
the equations of Ermak (1977). The basic equations as developed by Ermak are described in the remainder
of this section. The general equation governing pollutant transport and dispersion in the atmosphere, when
the pollutant is composed of uniformly-sized particles is as follows:
- = a ft _ a*. + _j_ ^ + -J-K,& + V^L (i)
t dx * dx dx dy y dy dz z dz 9 dz
where: % = concentration (g/m3)
l^,^,,^. = eddy diffusivity in the x, y and z directions (nf/sec)
t = time (sec)
x,y,z = coordinates in three dimension space where x is parallel
with the wind direction, y is perpendicular to x and parallel
with the surface and z is perpendicular to both x and the
surface (m)
u = wind speed (m/sec)
vg = gravitational settling velocity (m/sec) where positive is in
the downward direction
To solve equation (1), several simplifying assumptions are made. First, the diffusion in the x direction is
assumed to be small compared with the advection by the wind speed in that direction. This assumption
eliminates the need consider any cases other than steady state. Second, the eddy diffusivities are assumed
to be functions only of downwind distance (as will later be seen, further constraints are placed on the eddy
diffusivity to obtain the gradient-transfer solution). The resulting diffusion equation is greatly simplified from
equation (1):
-------
.
dx
-
2 dz2
(2)
A further assumption regarding the eddy diffusivity is made which generally means that the eddy d'rffusivity
must be constant for all space and time. Actually, to obtain the solution, it is only necessary that the eddy
diffusivity have the same functional dependance on the downwind distance as vg. Since vg is generally a
constant, the eddy diffusivity in all directions is also assumed to be represented by a single parameter
denoted by K. The relationship between K and the more commonly-seen diffusion parameter, the standard
deviation of the concentration in the y and z directions, ay and oz, is as follows:
a2 (x) = -
(3)
since K is to be assumed constant with x, it is obvious that the solution to be obtained is only strictly valid
for cases where the a varies as function of downwind distance to the 0.5 power. The equation thus obtained
is:
2nayazu
2K
v,",
8K2
-(z-ji)3
2 a2,
K
erfc
v;o,
z+h
where:
X
Q
u
ay,o2 =
x,y,z
h
K
concentration in g/m3
Emission rate (g/sec)
wind speed (m/sec)
standard deviation of concentration in the y and z directions (m)
coordinates of the receptor (m)
gravitational settling velocity (m/sec)
plume centerline height (m)
eddy diffusivity (nf/sec)
Ud-Vg/2
deposition velocity (m/sec). A basic assumption in the solution of
-------
this equation is that deposition will be proportional to
concentration at the surface. The deposition velocity is the
proportionality constant (see equation (9) below).
[Note: Ermak's original paper contained some typographical errors in the equation which have been
corrected here]
With the assumption that K is constant, and the assumption from equation (3) above, we can assume:
a 2u (s~\
K- = z {*)
The following two substitutions are made for convenience:
Y =
P =
With these substitutions, equation (4) can be written in the form:
X =
_^ 2 21 r F (g-A)'l r_ U*A)a1
-« aL " * \ \Q[ 2o« J + gl 2a * \ -
The assumption regarding the behavior of K with respect to x leads to some inconsistencies when equation
(4) is used with the standard Turner ay and oz curves. The inconsistency presents itself as a failure to
conserve mass as the plume travels downwind. A numerical integration technique was used to calculate
a correction term for modifying the concentration predictions of equation (4) to ensure that conservation of
mass is obtained. The details of the development of the mass conservation correction factors are presented
in the remainder of this section.
The basic equation which defines the deposition algorithm is that:
-------
where: D = Deposition rate
Uj = Deposition velocity
X = Concentration (from equation 8)
We can define corrected concentrations as the concentrations above, multiplied by some function of x,
which we will call here q(x):
c =
D = Udq(x) C\zm0 (ID
The requirement for mass conservation can be written as follows:
o -«
Substituting, we obtain the following form of the above equation:
C tdz = Q
o
Jl •• M ™
ffndydx + uffcdydz = Q <12)
where:
-------
fcdy
(14)
Differentiating both sides of the equation by x we obtain:
uaq(x)
(15)
Reforming, we obtain a first-order differential equation in x:
= o
(16)
The above equation was solved using a third-order Runge-Kutta integration process as follows:
Let A(x)
(17)
Then:
(18)
(19)
dx
-------
8
dg(x):
uA(x)
+ u-
uA\
dx
(20)
dx
dx
(21)
(22)
dx
(23)
(24)
A variable step length (dx) was used for using the above numerical integration scheme to determine q(x).
A total of 100 steps from 1 meter to 50,000 meters was used in the numerical integration. From these 100
values of downwind distance and q(x), a least squares fit was performed for an equation of the form:
log(g(x))
b2(Iog(x))2 + b3(log(x))3
(25)
Separate values of bj.^ were calculated for the combination of 6 different wind speeds, 6 different stability
classes, 6 different particle size classes and 5 different release heights. Over 1000 values of t^.^ were
computed and entered into the FDM code.
-------
9
Treatment of Meteorological Conditions
Meteorological data can be provided to the FDM in three formats: a sequentially processed
meteorological data set using the format produced by the RAMMET pre-processor, card images of hourly
meteorological data or a statistically-produced Stability ARray (STAR). If sequential meteorological data are
used (either the RAMMET pre-processed format or card image format), the model requires average values
over the shortest averaging period for wind speed, wind direction, atmospheric stability, temperature and
mixing height. Wind speed is used directly in the above equation to determine the concentration. Wind
speed is also used in combination with temperature and atmospheric stability to determine the values for
deposition velocity if the user asks the model to compute deposition velocity (the user can alternatively enter
deposition velocities with the input stream). Wind direction is used to determine the location of the receptor
with respect to the center point of each source in a coordinate system defined with the wind direction
parallel to the x-axis. Wind directions are entered as the number of degrees clockwise from north that the
wind is coming from. This point is significant, since some air quality models (e.g. the Industrial Source
Complex Model) require wind direction to be entered as the direction towards which the wind is moving.
Atmospheric stability is used to determine the values for the standard deviations of the horizontal
and vertical plume dimension above. The atmospheric stability for each unit of meteorological data (usually
hourly values) is specified as one of six possible stability classes, using the classification scheme of Turner
(1970). The computation of the actual parameters from the Turner classification is made using the equations
and coefficients listed in the User's Guide for the Industrial Source Complex (ISC) Model (EPA, 1986). The
determination of the values for these parameters is based solely on downwind distance and stability class.
The model is generally very insensitive to values of the mixing height, since fugitive dust emissions
are usually released at ground level and reflections off the mixing height are only significant at very far
distances from the source or at elevated receptors. However, the model does consider such reflections in
the standard fashion. Equation (8) is computed at z = z+nh^ and z - z-nH,,, for even values of n starting
with 2 and progressing until concentrations are no longer significant. The values computed from these
reflections are added to the value computed from the original computation of Equation (8) to arrive at the
-------
10
total concentration at the receptor height. Some users have noted very slow computation times when very
low mixing heights are input. Users are urged to consider the computation time when selecting mixing
heights for input to the model.
It should also be noted that when the RAM MET preprocessor is used for preparing a sequential
meteorological data set, often very low values of the mixing height can be computed. When using such a
data set, the user is encouraged to examine the meteorological conditions for the highest days to determine
that meteorological conditions are realistic. FDM has been configured to automatically set the mixing height
for stable conditions to 5000. meters and to limit the mixing height for other stability conditions to no less
than 100. meters. Since long-term meteorological data sets often include some missing data, the model has
been designed to recognize missing or bad data using coding involving series of 9's. Bad or missing wind
speeds are coded as 999.9 meters per second or larger, missing wind directions are coded as 999.9
degrees clockwise from north, and missing or bad stabilities are coded as 9.
Missing data includes calm conditions which are defined as wind speeds less than 1.0 meter per second.
Missing data are treated by the model according to the EPA "calms policy (Guideline on Air Quality Models,
Section 9.3.4.2). The calms policy is based on the assumption that wind speeds less than 1.0 meters per
second are not well simulated by Gaussian Plume Models. FDM has been designed to eliminate hours with
wind speeds less than 1.0 meters per second from the analysis and the averaging computations. For the
purposes of calculating averages, if a small amount of data is missing, the model computes the average on
the basis of the data which remain. For example, if 24-hour averages are being computed, and 2 hours
within a 24-hour period are missing, the model computes the 24-hour average as a 22-hour average of the
remaining good hours. If however, the number of missing, bad or calm hours exceed a threshold value
defined separately for each hour, the model computes the average as if there were at least the threshold
number of hours, and that the missing hours produced a zero concentration and deposition rate. For
example, for 24-hour averages, the threshold is 18 good hours. Thus if 8 hours of data were missing from
a 24-hour period, the model would compute the 24-hour average concentration by adding the concentrations
for the 16 good hours and dividing the total by 18. Thresholds for the averaging times considered by FDM
-------
11
are:
Averaging time Threshold
1-hour 1
3-hour 3
8-hour 6
24-hour 18
For long-term averages, such as the annual average of an entire year of data, the model computes the
average as an average of the good hours of data without a threshold.
One important aspect of the calms policy that is not treated in FDM is the proper proceedure to
processing low wind speed hours prior to entering them in a computer model. If the threshold of the
meteorological sensor which collected the data is less than 1.0 meter per second, the calms policy calls for
the user to determine if each hour of meteorological data is greater then the threshold of the sensor, but
less than 1.0 meter per second. In that event, the user is instructed to increase the wind speed for that hour
to 1.0 meter per second so that FDM will keep the hour in the computations.
Treatment of Deposition
Equation (8) accounts for deposition through two parameters: the gravitational settling velocity and
the deposition velocity. As its name implies, the gravitational settling velocity accounts for removal of
paniculate matter from the atmosphere due to gravity. Since only the larger particles have sufficient mass
to overcome turbulent eddies, this mechanism is significant only for the larger size ranges (e.g. particles
greater than 30 micrometers). The deposition velocity accounts for removal of particles by all methods,
including turbulent motion which brings the paniculate matter into contact with the surface and allows it to
be removed by impaction or adsorption at the surface. It is known that for smaller particles the deposition
velocity is significantly different from the gravitational settling velocity, while for large particles they are
roughly the same (N'rfong and Winchester 1970). In the FDM the emission rate, Q, is divided into a user
-------
12
determined number of particles size classes (maximum of 20). Each of the classes has a unique
gravitational settling velocity and deposition velocity. The user may enter these parameters directly, or may
enter characteristic diameters for each particle size class and ask the model to compute the deposition
velocity and gravitational settling velocity. The method used by the model to compute the gravitational
settling velocities and deposition velocities is modeled after the work of Sehmel and Hodgson (1978). The
portion of the FDM computer program which calculates the gravitational settling velocity and deposition
velocity was written by Mr. Bart Croes at the California Air Resource Board (CARB) (Croes, 1987).
Key inputs to the method used by Sehmel and Croes are the roughness height and the friction
velocity. Friction velocities are calculated internally in the FDM from the wind speed and the reference
height of the meteorological data. The computer program for the computation of the friction velocity was
written by Gregory J. McRae (1977) and provided by Mr. Croes of the CARB. The roughness height is an
input parameter of the model. Figure 1 provides some typical values for the roughness height.
Each particle size class is treated separately by the model. The results for different particle size
classes are summed at the end to develop a total suspended paniculate concentration. Alternatively, the
model can compute the deposition rate. In this event, the concentrations for each particle size class are
multiplied by the deposition velocity for that particle size class and the results are summed to determine the
total deposition rate.
As noted earlier, a major difference between FDM and other available air quality models is the ability to
treat both turbulent and gravitational removal mechanisms for particles. In doing so, the gradient-transfer
deposition algorithm is used. The gradient transfer algorithm is also termed "K-theory", because turbulent
transfer processes are assumed to be proportional to the gradient or first derivative of concentration. The
proportionality constant is called the "eddy diffusivity", and is customarily denoted by the symbol "K".
A critical assumption necessary to obtain the solution shown in equation (8) is that the eddy diffusivity
be constant for all space (isotropic eddy diffusivity). Although isotropic turbulence is a common assumption
in many areas of fluid mechanics, including some atmospheric processes, it is known that a plume's rate
of dispersion changes as it moves downwind in most atmospheric dispersion processes, because eddys
-------
13
10 -
1 -
10-'-=
o
M
10
-2
10 "3-
10 "4~
10
+ Rocky Mountains
+ W. Virgtnia
+ E. Tenn.
Appalocion Mtns.
•4* Centers of cities with very toll buildings or very hilly or*
+ Centers of Large towni and cities
+ Centers of small town*
+ Average U. S. Plains
+ Outskirts of towns
-f* Many tr«#s. h«dg*j and /»w buddings
+ Uony h*d9*s
+ Few tr**s. summ«r timt
+ Isolated tr««a
+ Uncut gross
Low mtns,
fOTMt
Dens* forest
T^ IK9
+ Few tr«««, winter tin
T Cut Grass (-3 cm)
+ Natural snow lurfoce (farmlond)
+ Off-seo wind in coastal areas
Farm land
Fairly level grass plain
Desert (flat)
Large expanses of water
Calm open sea
flat or rolling ground
Based on Stull (1988)
S. American average
S. Asian average
Fairly level wooded country
S. Africa average
N. America average
U.S.S.R. average
Europe average
Australia average
Long grass («60 cm), crops
Airports (runway areas)
N. Asia average
N. Africa average
Figure 1. Typical Values for the Roughness height
-------
14
of different sizes dominate the dispersion at different times in the plume's evolution. This fact is evident in
the standard dispersion curves (sigma-y and sigma-z curves) provided by many authors and summarized
by Turner (1970). Since the FDM makes use of the standard Turner curves, a potential inconsistency exists
between the use of the constant-K assumption to develop equation (4), and then allowing K to be
represented by the standard deviations in the crosswind and vertical directions (sigma y and sigma z) in the
standard fashion. The inconsistency results in a failure of equation (8) to conserve mass as the plume
travels downwind.
To alleviate the mass conservation problem, the FDM has been corrected in the manner suggested by
Horst and Doran (1986). Concentration and deposition have been numerically integrated for a large number
of cases involving different meteorological conditions, different particle sizes and different release heights.
A numerical solution was developed to correct the concentrations so that approximate mass conservation
is obtained for all cases. In general, for particles smaller than 10 micrometers or less, the corrections are
very small, for all cases examined. However, for larger particles, at long downwind distances, the
corrections are significant. Correction factors are built into the model and the use of correction factors is
entirely transparent to the user.
Emission Rates as a Function of Wind Speed
One of the unique characteristics of fugitive dust is that often emission rates are a function of the
wind speed. The FDM has the capability to directly compute the effect of wind speed on emission rate.
For each source, the emission rate is calculated by the equation:
(26)
where: Q = emission rate in g/sec
Q, = proportionality constant
u = wind speed (m/sec)
s = wind speed dependance parameter
The emissions for every source are entered as a proportionality constant and a wind speed dependance
parameter (Q, and s). For sources which do not vary with wind speed, the emission rate is simply entered
-------
15
for Q, and s is entered as 0 (the default). However, for sources which do vary with wind speed both
parameters must be specified. Examples would include the cubic dependance on wind speed of some wind
erosion emission estimates (Woodruff and Siddoway, 1965). Also in "Compilation of Air Pollutant Emission
Factors" (EPA, 1985), many fugitive dust sources are shown to linearly depend on wind speed, such as
batch and continuous loading and unloading operations and losses from open storage stockpiles.
Treatment for Line Sources
Line sources are treated virtually the same as the CALINE3 Model (California Department of
Transportation, 1979). The code has actually been lifted from the CALINE3 Model and incorporated in the
FDM for the line source treatment. The CALINE3 line source algorithm involves the division of the line
source into a series of elements. The spacing of the division between elements is determined by the angle
of the line source with respect to the wind direction and the receptor. A line is defined which passes
through the receptor and is parallel to the wind. The area of the line source near this parallel line is divided
into many elements, which are roughly square in shape. Portions of the line source further from the parallel
line are divided into fewer elements which are longer (more rectangular). The number of elements and their
orientation depends on the geometry of the line source with respect to the receptor and the wind.
Each element is further divided into sub-elements. These sub-elements are represented in the model
as small finite line sources which are perpendicular to the wind direction. These sub-elements are
represented in the concentration equations by a cross-wind integrated version of equation (8). The finite
integral of the equation is computed using a numerical approximation for the error function. Further details
on the algorithm for line sources can be found in Appendix C, taken from the CALINE3 User's Guide
referenced above. The deposition capability has been modified to be consistent with the treatment above.
Treatment for Area Sources
Area sources are specified by the user with a center point, an x-dimension, a y-dimension, a rotation
angle and the various emission parameters from the above equations. The model computes concentrations
from the area sources by first rotating the coordinate system so that the origin is at the receptor and the
x-axis is aligned with the wind direction. Figure 2 illustrates the orientation of the receptor and a typical area
-------
16
source. The portion of the area source which is upwind (in the range of positive x values) is considered.
The area source is divided into a series of line sources oriented perpendicular to the wind direction.
There are two major options for the area source algorithm: a default 5-line integration, or a convergent
integration. In the default 5-line version, the model divides the area source into 5 lines perpendicular to the
wind direction. In the convergent version, the model begins by dividing the area into 5 lines. It computes
a concentration at each receptor for the 5-line integration, then repeats the process for using a 6-line
integration. If the results are less than 1% different at all receptors, the model uses the 6-line integration.
If, however, one or more receptors fails the 1% test, the model computes a 10 line integration and an 11-line
integration and makes the same comparison. It continues with further sub-divisions in increments of 5
(15/16, 20/21, 25/26, etc.) until convergence is obtained. The maximum allowable number of lines in the
integration is 901. If convergence has not been obtained with 901 lines in the integration, the model uses
the 901 -line integration. Since the line sources are perpendicular to the wind direction, it is possible to use
a finite integration to solve the line source integral for these lines. As with the line source sub-elements, the
cross-wind integrated form of the dispersion equation is used, with a numerical routine used for determining
the error function. It is possible for receptors to be located within area sources, but only the portion of the
area source upwind of the receptor is considered.
It should be noted that the convergent area source algorithm may require substantial additional
computation time. Run times have been observed to increase by an order of magnitude in some cases
using the convergent algorithm. The user should use discretion when selecting this algorithm, particularly
for cases involving long meteorological data bases.
The rotation angle supplied for area sources by the user requires further explanation. The geometry of
the area source is defined by the user by the specification of five parameters. The x and y coordinates of
the center point of the area source are straight-forward. Usually, the user is working with a map which
defines an area he or she wishes to model. The area may be irregular in shape and oriented in any fashion.
Typically, the user would draw a rectangle on the map to simulate part or all of the area to be modeled. The
rectangle can be oriented in any way. By drawing lines to connect opposite corners of the rectangle, the
-------
17
center point is determined and the coordinates determined. The second two parameters are the x and y
dimensions of the area source. Typically.the area source will not be located with the sides parallel to the
x and y axis of the map. The user selects one of the sides to represent the x-side (it doesnt matter which).
The length of the selected side is entered as the x dimension, and the other side dimension is entered as
the y dimension. Finally, the user determines the angle of the side selected as the x-axis with respect to
the true x-axis. This angle is the area source rotation angle and is the final parameter entered for each
source. It is important to note that the angle must be in the range of -90 degrees to plus 90 degrees.
Values larger than 90 degrees will not be accepted by the model.
The model determines the line sources used in the integration of the area source by rotating the
coordinate system so that the x-axis is parallel to the wind direction. In this rotated coordinate system, the
first step is to determine the x-extent of the area source. The x-extent is found by determining the x and
y coordinates of the four corners of the area source in the rotated system, and simply subtracting the
smallest x value from the largest x value. The actual line sources used in the integration will be
perpendicular to the wind, and hence parallel to the y axis in the rotated coordinate system. The x extent
is divided by the number of steps in the integration, and the x coordinate for each of the line sources is
determined as the center of each step. Finally, the y coordinates of the line sources are determined by
computing equations for four lines which are colinear with the sides of the area source. Each line source
will intersect the four lines of the area source sides at four distinct points (special cases are made when the
sides of the area source are parallel to the x and y axes). Since the line sources are parallel to the y axis,
the four intersection points will all have the same x value, but different y values. The y values can be sorted
and the middle two values will define a line perpendicular to the wind and within the area source.
Special Considerations when Using a STAR
If meteorological data is provided to the model in the form of a Stability Array, the model computes
concentrations in a fundamentally different manner from the other meteorological options. Instead of using
the CALINE3 algorithm as the basis for line and area sources, the model now computes concentrations as
22.5 degree sector averages, and much of the CALINE3 algorithm is moot. The sector averaging differences
-------
18
Area Source
Wind Direction
Receptor
Figure 2 Area Source Treatment in FDM
-------
19
are true for point sources as well. In fact the model calls entirely different subroutines for computations with
a STAR. The option of writing a sequential output file for post-processing is also eliminated when running
with a STAR. The same deposition algorithms are used. The main difference is that Equation (8) is
integrated in the cross wind (y) direction from minus infinity to plus infinity and the result divided evenly over
a the 22.5 degree sector referenced by one of the 16 possible wind direction categories in the STAR.
To eliminate large differences between adjacent receptors across sector boundaries, the model
smooths concentrations between adjacent sectors. The method of using the smoothing function is to
consider a receptor's placement within a sector. For example, for a point source each receptor will be
contained within only one downwind sector. The frequency used in the concentration computation for that
receptor, however, is a weighted average of the frequency of the sector the receptor is in and the nearest
other sector. The weighting is determined by how far the receptor is from the sector center. If it lies directly
on the sector center, then the frequency used is exactly that of the sector it lies in. If the receptor lies on
the boundary of two sectors, the frequency used is the average of the frequency of the two sectors. For
points in between these two extremes, a linear interpolation is used.
-------
20
(This page intentionally left blank)
-------
21
3.0 USER'S INSTRUCTIONS
Information is provided to the model in either one or two files. The first is referred to as the FDM
input file and contains information on the receptors, sources and various model switches and options. The
FDM input file also can contain the meteorological data, expressed as a series of card-images (either as a
series of 1-hour episodes, or a statistically produced STability ARray (STAR)). If, however, the user elects
to supply meteorological data in the standard pre-processed format, using the RAMMET pre-processor
program, a second file must be identified with the meteorological data.
The model was developed on an IBM-PC compatible computer, but is written in standard FORTRAN
code, and may be adapted for operation on a mainframe or other computer system. The instructions
provided here are those which apply to an IBM-PC compatible computer, running a standard Disk Operating
System (DOS). The model requirements for PC operation are a minimum of 500 K of memory and a math-
coprocessor chip (e.g. 80287, etc). An additional requirement is that the device driver, ANSI.SYS or a
compatible be installed on the machine. The ANSI.SYS file is provided with most DOS packages. To install
it, the user must make sure that the statement "DEVICE = ANSI.SYS" is present in the CONFIG.SYS file in
the user's root directory of the drive used to boot the computer. It should be noted that some operating
systems provide their own special version of the ANSI.SYS device driver. For example the commercial
software called DOUBLEDOS provides a version called "DBLDANSI.SYS". The FDM package is compatible
with any such device driver.
The version of the model supplied on the distribution diskette was compiled using the Microsoft
FORTRAN compiler. As such, it uses two special routines provided with the Microsoft compiler, which must
be deleted, or replaced if the user decides to compile the program with a different compiler. First, the model
uses a command-line interface routine called GETARG. The interface allows the user to name the input file
on the command line which begins the execution of the program. The command line interface allows the
user to "batch" run a large number of runs by simply creating a batch file with a series of command lines.
-------
22
The second Microsoft routine called by the program is the NUMARG call which counts the number of
characters in the file name. Some compilers simply treat very large negative exponents on numbers as zero.
The Lahey compiler, for example, requires calling the UNDERZERO subroutine to make numbers with large
negative exponents default to zero. The user should consult the manual for the compiler the user intends
to use for information on the treatment of underflows.
If the user does not specify a filename on the command line, the program prompts the user for the
names of the input files and output files. The input files must have been prepared prior to the operation of
the run. Directions for preparing the FDM input file are detailed in the next section of this chapter. If the
meteorological option is selected to provide the data in pre-processed format, it must be in a standard
"UNFORMATTED" file. Compilers differ in the form for "UNFORMATTED" files, thus it may be necessary to
run a separate program, also provided in this package, called "UNFORMAT" which will take a formatted file
containing the RAMMET pre-processed output and transform it to an unformatted file, suitable for the FDM
input. Contained in the diskettes which are provided is a FORTRAN program for transforming the data if
required, along with a test data set illustrating the use of this program.
Further information on the files needed and produced by FDM are provided below. For here it is
important only to know that there are two possible input files (one required and one optional) and three
possible output files (printer, plotter and sequential concentration for post processing). Once the input files
are prepared and stored on a disk drive, the FDM program is initiated by typing "FDM91028" followed by
a file name. Note that the name of the program includes the version number to prevent confusion with
previous or later versions of the program. The three letters FDM are always at the start of the program
name. The next two characters are the year of the current version, and the final 3 characters are the Julian
day (numbered day of the year from 1 to 365 or 366) of the current version. If later versions of the model
are issued they too will be executed by typing a program name of this form. Card input files are always
designated with the .IN extension, and printed output files are always use the .OUT extension. Sequential
meteorological data sets are designated by the .MET extension. Plotter output files use the .DAT extension,
while a sequential output for post processing uses the .CON extension. For example, if a run was being
-------
23
made with the card input stored in a file called RUN1.IN, and a sequential meteorological data set was
stored in RUN1.MET, the program would be started by issuing the command from the DOS prompt:
FDM91028 RUN1
The program will automatically write printed output to a file called RUN LOUT, plotter output to a file called
RUN1 .DAT (if selected), and sequential concentration output for postprocessing to a file called RUN1 .CON
(if selected). If the user fails to enter a file name after the FDM91028 command, the program prompts for
the name of the input files, and prompts for file names where the various output files are to go.
CAUTION - The FDM program will erase old files with the
same name as that specified for output, so that if the user
enters a name for the output file which already exists on
the disk drive, it will be overwritten.
3.1 The FDM Input File
The FDM input file provides the model with most, if not all, of the information needed for execution
of the program. Information is provided to the model through a series of card images which consist of a
maximum of 80 columns of data (i.e., 80 characters per line). Table 1 provides a summary of the
information needed and the format for each entry in the file. The input file contains three general types of
information:
o general values and model switches and options
o receptor locations - the locations at which concentrations and depositions are to
be calculated,
o source data, including the geometric information on source locations, sizes and
orientations, the emission information and the release heights, and
o meteorological data, if not provided by a separate sequential file.
-------
24
The first group of data includes a run title, for documentation purposes only, the selection of area source
algorithm to use: convergent or default (see Chapter 2 for a more detailed discussion of these two options),
the selection of format for the meteorological data, the selection of which output files to create, and what
items are to be contained in each output file. A specific discussion of the output options is presented below.
The general information section also includes the specification of the number of receptors and sources to
include, the number of particle size categories to use, a characteristic particle diameter for each particle size
category, a mass fraction for each particle size category, a density of the paniculate matter, and a
roughness height for the site (see Chapter 2 for more guidance and information on the selection of
roughness height).
Sample input files and output files are included on the diskette. The meaning and possible values
for each of the parameters is explained in Table 1.
The other possible input file is a RAMMET pre-processed meteorological data set. The model
requires this information to be provided in an "UNFORMATTED" file, which is discussed above. The program
UNFORMAT is provided on the diskette for converting formatted pre-processed meteorological data into
unformatted format.
3.2 The FDM output files
Output can be obtained from FDM in three formats. First, the standard output file, as contained on
the diskette, which documents all the inputs and the computed concentrations or depositions for the model.
The information to be provided in the output file is selected by the user in the input file. The options include
which averaging times to consider for the meteorological data provided. Unless directed by the user
differently, the model assumes each meteorological entry represents an hourly average of the meteorological
conditions. The model calculates concentration and deposition values for each of these entered conditions.
The user may report every value calculated, or he may choose to report only averages of groups of these
values. For example, 'rf meteorological data comprising 48-hours of hourly-averaged meteorological data
are entered by the user, the model will compute 48 1-hour average concentrations and depositions.
However, the user may not wish to have all these data reported, thus the user can direct the model to report
-------
25
only two 24-hour averages, six 8-hour averages, sixteen 3-hour averages, one 48-hour average, all 48 one-
hour averages, or any combination of the above values.
Further, for longer-term data processing, the model can be directed to sort the values computed for
each of the averaging periods and report only the top values. These top values are reported in two formats:
a top 50 table, which consists of the top 50 averages computed in the run, regardless of receptor location,
and a high two table, which consists of a table showing the two highest averages computed at each
receptor. For example, if the user directs the model to analyze a full year of meteorological data (8760 or
8784 hours) and compute 24-hour averages at 10 receptors, a total of 3650 (or 3660) 24-hour average
concentrations and depositions will be computed. With the top value output selected, the model will sort
the concentration values to find the top 50 24-hour averages computed at any receptor location and the top
two values at each receptor location. Note that the corresponding deposition rate is reported which
generally is co-located with the peak concentration value, but no actual sorting is performed of the
deposition values.
The second form of output is a "plotter" file which contains concentrations for each averaging time
selected by the user along with the coordinates for those concentrations. The format is a generic form
which simply presents the x coordinate, the y coordinate, the concentration and the deposition rate. It
should be noted that this format is ideal for input to gridding and plotting programs such as the Golden
Graphics Surfer program which can produce isopleth plots of concentration. However, the user is cautioned
that the information reported in plotter file is determined by the selections for the printed output. If the user
selects only to print top values in the printed file, the plotter file, although created by the program, will be
empty. To obtain a plotter file, the user must actually specify that every average be printed for a particular
averaging time.
The third format for the output is a sequential file of concentrations for post processing by the
POSTZ program, available on the SCRAM. POSTZ is a post-processor designed for the SHORTZ air quality
model. The SHORTZ air quality model has the capability of writing an output file of sequential
concentrations for every combination of meteorological condition, source and receptor. These tapes, on
-------
26
the IBM PC system take the form of a disk file. The FDM has been equipped with the option of writing a
tape of a format suitable for input to the POSTZ program. Much of the information on the tape is not used
by the POSTZ model, thus in many cases the FDM has been instructed to write "dummy" variables to the
tape to keep the format correct, but which do not enter in the calculation of any of the POSTZ results.
The advantage to the POSTZ post-processor option is that many alternate averaging times can be
examined, specific periods of a longer meteorological data base can be examined, the results for certain
sources can be scaled up or down, and a number of other manipulations can be performed with the data.
The POSTZ program also prepares high-five and top 50 tables which are useful for many regulatory
applications of the model.
The major disadvantages to using the POSTZ program are that the sequential tape file written by
FDM for POSTZ input can be very large, and can exceed the capacity of many typical PC hard disks. For
example, a run containing 15 sources, 200 receptors and 1 year of sequential meteorological data will write
a tape file that is over 100,000,000 bytes in length. Discretion must be exercised when selecting this option.
It is also important to note that the post-processing option is only available for concentration.
-------
27
TABLE 1
SUMMARY OF INFORMATION REQUIRED FOR FDM INPUT
Card 1 Title Card
Col Format
1-80 A80
Card 2
Col
2
Switches
Format
II
II
II
Information
Title
Information
Area source algorithm switch. Two
options are available for area
sources, the default 5-line
integration is selected by entering
a value of 1 for this parameter.
The alternate convergent algorithm
may be selected by entering a value
of 2 here. The user is cautioned
that the convergent algorithm can be
very slow in execution time and can
drastically increase run times.
Default is 1.
Meteorological Data Option Switch.
If = 1 then met. data is read from
cards (format shown below) contained
later in this input file. If = 2
then met. data is read from pre-
processed meteorological file. If =
3 then met. data is read as a STAR
contained later in this input file
(format 6F10.0). Note that the
selection of the STAR option makes
many of the later options not
applicable. Default is 1.
Plotter Output Switch. If = 1 then
no plotter file is made. If = 2
then a plotter file name is asked
for and the model writes a file with
a formatted output of x, y,
concentation, deposition for every
averaging time selected with a value
of 2 or 4 in columns 14, 16, 18, 20
or 22. Note that a value of 3 in
any of these columns will not
produce a plotter output. Default
is 1.
-------
28
8 II Print Output Switch. If = 1 then
meteorological data are not printed.
If = 2 then meteorological data are
printed. Default is 1.
10 II Post Processor Switch. If = 1 then
no post processor file is written.
If = 2, then a post processor file
is written which can be processed
with the POSTZ program to develop
High-5 tables, scale sources, or
other operations. Note that the
maximum number of receptors possible
for the post processor is 200. The
user should see the POSTZ User's
Guide for further information. Note
that post processing is performed
only for concentration — no post
processing option is available for
deposition. This option is not
available and this switch is ignored
when the met. option switch = 3.
Default is 1.
12 II Deposition Parameters Option Switch.
If = 1 then the model will compute
deposition velocity and
gravitational settling velocity
automatically on an hour by hour
basis. If = 2 then the User will
enter single values of the
deposition velocity and
gravitational velocity for each
particle size class to be used for
all hours. Default is 1.
14 II 1-Hour Output Switch. If = 1 then
1-hour average concentrations are
not printed, If = 2 then 1-hour
average concentrations are printed.
If = 3, a top 50 and high 2 table
are prepared for this averaging
time. If = 4, both the top 50, high
2 and every 1-hour average
concentration are printed. This
option is not available and this
switch is ignored when the met.
option switch = 3. Default is 1.
16 II 3-Hour Output Switch. If = 1 then
3-hour average concentrations are
-------
29
not printed. If = 2 then 3-hour
average concentrations are printed.
If = 3, a top 50 and high 2 table
are prepared for this averaging
time. If = 4, both the top 50, high
2 and every 3-hour average
concentration are printed. This
option is not available and this
switch is ignored when the met.
option switch = 3. Default is 1.
18 II 8-Hour Output Switch. If = 1 then
8-hour average concentrations are
not printed. If = 2 then 8-hour
average concentrations are printed.
If = 3, a top 50 and high 2 table
are prepared for this averaging
time. If = 4, both the top 50, high
2 and every 8-hour average
concentration are printed. This
option is not available and this
switch is ignored when the met.
option switch = 3. Default is 1.
20 II 2 4-Hour Output Switch. If = 1 then
24-hour average concentrations are
not printed. If = 2 then 24-hour
average concentrations are printed.
If = 3, a top 50 and high 2 table
are prepared for this averaging
time. If = 4, both the top 50, high
2 and every 24-hour average
concentration are printed. This
option is not available and this
switch is ignored when the met.
option switch = 3. Default is 1.
22 II Long-term Output Switch. If = 1
then average concentrations over the
entire meteorological data base
provided are not printed. If = 2
then such long term average
concentrations are printed. If = 3,
a top 50 table is prepared for this
averaging time. If = 4, both the
top 50, the long-term average
concentration are printed. This
option is not available and this
switch is ignored when the met.
option switch = 3. Default is 1.
-------
30
24
Card 3
Col
1-60
Card 4
Col
1-60
II Calms recognition switch. If = 1
then calms are recognized by the
combination of a wind speed equal to
1.0 m/sec and a repeated wind
direction from the previous hour.
If = 2 then calms are not recognized
by this combination. Regarless of
the setting of this switch, calms
will always be recognized when wind
speeds are entered less than 1.0
meter per second. bad data will
still be recognized by entering wind
speeds of 999.9 regardless of the
value for this switch. Default is
1.
STAR Data (These Cards are only read if Met. Option
Switch = 3)
Format Information
6F10.0 A total of 96 cards are read here
with the information being the
frequency of winds for each
combination of wind speed class,
wind direction class and atmospheric
stability class. Each card contains
six values corresponding to the six
possible wind speed classes. The
order of the cards is 16 cards for
the 16 possible wind direction
classes for the first stability
class, followed by the next 16 cards
for the second stability class,
followed by 16 cards for each
subsequent stability class up to the
final (sixth) stability class. The
wind direction cards are ordered
with north being first, north-
northeast being second and
proceeding clockwise until north-
northwest is entered. Stabilities
start with Turner Class A, and
proceed to Turner Class F. The sum
of all 576 values entered here
should be 1.0.
Mixing Heights for each Stability Class when using a STAR
(This Card is only read if Met. Option Switch = 3)
Format
6F10.0
Information
Six values are read here to indicate
the characteristic mixing height to
be used with each stability class
-------
31
Card 5
Col
1-60
when using a STAR for input
meteorological data. Mixing heights
should be entered in meters above
ground.
Characteristic Wind speeds for each wind speed class when
using a STAR (This Card is only read if Met. Option
Switch = 3)
Format Information
6F10.0 Six values are read here to indicate
the characteristic wind speed to be
used by the model for each of the
wind speed classes when running with
meteorological data entered in the
form of a STAR. Wind speed values
should be entered in meters per
second.
Card 6
Col
1-5
Integer Parameters
Format
15
6-10 15
11-15 15
Information
Number of Sources (maximum 121)
Number of Receptors (maximum 1200)
Number of Particle
with a maximum of 20
that in order to
deposition values or
concentrations which
accounted for, this
be set to some value
Size Classes,
allowed. Note
compute any
to compute any
have deposition
parameter must
other than 0.
16-20
15
Card 7 Real Parameters
Col Format
1-10 F10.0
11-20
21-30
F10.0
F10.0
Number of Hours of Meteorological
data to be processed in this run.
Information
ATIM - the length of time in one
unit of meteorological data entry in
minutes. Generally, this is entered
as 60.
Surface Roughness Height in cm (see
Figure 1).
SCAL - a scaling factor for all
entries involving distance. The
model assumes all entries for
coordinates or dimensions are in
meters. If the user desires to
enter some other units, he may enter
a conversion factor here such that
-------
32
when the units he has entered are
multiplied by SCAL the result will
be in meters.
31-40 F10.0 PD - the global value for density of
the particulate matter in grams per
cubic meter. Typical values range
from 1.0 to 3.0 depending on the
type of material which comprises the
particulate matter. Entered here is
a global value to be used for all
sources, unless a different value is
specified for a particular source
later by the user.
41-50 FlO.O ANHT - the anemometer height above
ground in meters. Default assumes
10 m. This value is used to correct
the wind speed to the height of the
source or to a reference height of
10 meters for some calculations in
the model.
Card 8 Meteorological Data Selection Switches. These cards are
entered only if a sequentially pre-processed
meteorological data set is being used (Meteorological
Data Option Switch = 2) . The switches allow the user to
select certain days of the sequential data set for
processing and skip the rest.
Col Format Information
1-80 8011 A series of 1's or zero's is used to
determine if a particular day from
the sequentially pre-processed
meteorological data set is to be
processed in this run. The first
number entered corresponds to day 1,
etc. A total of 366 values (4 and
1/2 cards) is needed to enter all
366 values. If a 1 is entered the
day is to be processed, if a zero is
entered the day is to be skipped.
Card 9 Characteristic Particle Diameters (not entered if the
Number of Particle Size Classes is 0)
Col Format Information
1-10 8F10.0 The average or typical diameter for
each particle size class is entered
here in micrometers (um or meters X
10 ) . [Note these values are not
multiplied by SCAL] A total of 20
particle size classes can be
-------
33
specified and a characteristic
diameter must be specified for each
particle size class used. Eight
values can be placed on each card
here. Use as many cards as
necessary to provide the number of
particle size classes specified, but
do not include any blank cards.
Card 10 General Particle Size Distribution (not entered if the
Number of Particle Size Classes is 0)
Col Format Information
1-10 8F10.0 The fraction of the emissions which
are contained in each particle size
class are entered here. A total of
20 particle size classes can be
specified and a fraction must be
specified for each particle size
class used. 8 values can be placed
on each card here. Use as many
cards as necessary to provide the
number of particle size classes
specified, but do not include any
blank cards. This card refers to a
general particle size distribution
which is used for all sources here
unless over-ridden by a specific
switch entered on each source card.
When over-ridden on the source cards
which follow, the user may specify a
specific size distribution to use
for a specific source, or may have
the mod£l assume no deposition for a
specific source.
Card 11 Gravitational Settling Velocities. This card is only
entered if the number of particle size classes is greater
than zero and the deposition parameters option switch is
set to 2. Otherwise, the model computes gravitational
settling velocities automatically. This option is only
used if the user has some reason to use specialized
gravitational settling velocities.
Col Format Information
1-10 8F10.0 The gravitational settling
velocities in m/sec are entered
here. A total of 20 particle size
classes can be specified and a
gravitational settling velocity must
be specified for each particle size
class used. Eight values can be
placed on each card here. Use as
-------
34
many cards as necessary to provide
the number of particle size classes
specified, but do not include any
blank cards.
Card 12 Deposition Velocities. This card is only entered if the
number of particle size classes is greater than zero and
the deposition parameters option switch is set to 2.
Otherwise, the model computes deposition velocities
automatically. This option is only used if the user has
some reason to use specialized deposition velocities.
Col Format Information
1-10 8F10.0 The deposition velocities in m/sec
are entered here. A total of 20
particle size classes can be
specified and a deposition velocity
must be specified for each particle
size class used. Eight values can
be placed on each card here. Use as
many cards as necessary to provide
the number of particle size classes
specified, but do not include any
blank cards.
Receptors
Format
F10.0
Information
X-Coordinate of receptors in meters,
or in units which will be converted
to meters when multiplied by SCAL
entered above.
11-20 F10.0 Y-Coordinate (units as above)
21-30 F10.0 Z-Coordinate. Note that the Z
coordinate is not a terrain
elevation, since FDM does not
simulate any rough terrain effects.
The value of. Z here is used to
represent the height of the receptor
above the ground ("flagpole
height").
Each receptor is entered on a single
card. A total of 1200 receptors may
be specified. Note, however, that
POSTZ will only accept a maximum of
200 receptors, thus if post-
processing is to be used, the number
of receptors should not exceed 200.
Card 14 Source Information. Source information is entered on a
series of cards — one card for each source. A total of
121 sources can be entered.
-------
35
Col Format Information
2 II Type of source. 1 = point source, 2
= line source, and 3 = area source.
3 II Particle size override switch. If
this switch is left blank or set to
0, the model uses the particle size
distribution specified in card 10 to
apply to this source. If, however,
this value is set to 1, the model
reads a second card (or as many
cards as necessary) , after card 14
to specify the particle size
distribution for this source. If
this value is set to 2, the model
assumes no deposition for this
source.
4-15 F12.0 Emission rate. For point sources,
the units are grams per second
(g/sec). For line sources the units
are grams per meter per second (g/m-
sec) . For area sources the units
are grams per square meter per
second (g/m -sec) . Note if this
source is a wind-speed dependant
source, the emission rate entered
here is the proportionality constant
of the wind speed dependant
expression of the form: E - Q0uw
where E is the emission rate, Q0 is
the proportionality constant, u is
the wind speed in m/sec and w is the
wind speed dependance factor.
16-20 F5.0 Wind speed dependance factor. See
the note under emission rate above.
If the source is not a function of
wind speed, leave this column blank
and enter the emission rate in
columns 3-15 as above.
21-30 F10.0 X-coordinate. For point sources,
this is the x-coordinate of the
source. For line sources, this is
the x-coordinate of one end of the
line source. For area sources, this
is the x-coordinate of the center of
the area source. In all cases the
values are in meters, or in units
which will be converted to meters
when the computer multiplies by the
value entered for SCAL above.
31-40 F10.0 Y-coordinate. For point sources,
this is the y-coordinate of the
-------
36
41-50
F10.0
51-60
F10.0
61-70
F10.0
source. For line sources this is
the y-coordinate for one end of the
line source (the same end as the
above x-coordinate) . For area
sources, this is the y-coordinate of
the center of the area source. In
all cases the values are in meters,
or in units which will be converted
to meters when the computer
multiplies by the value entered for
SCAL above.
2nd X-coordinate. For point
sources, this column is not used.
For line sources, this is the x-
coordinate for the other end of the
line source. For area sources, this
is the x-dimension of the area
source. If the area source is
rotated, one of the axes must be
selected as the x-axis, and it's
rotation angle from the true x-axis
entered below. The dimension of the
side of the area source selected to
represent the x axis is entered
here. In all cases the values are
in meters, or in units which will be
converted to meters when the
computer multiplies by the value
entered for SCAL above.
2nd Y-coordinate. For point
sources, this column is not used.
For line sources, this is the y-
coordinate for the other end of the
line source. For area sources, this
is the y-dimension of the area
source. As noted for the x-axis
above, the y dimension for rotated
area sources is the dimension of the
side of the area source selected to
be the y dimension. In all cases
the values are in meters, or in
units which will be converted to
meters when the computer multiplies
by the value entered for SCAL above.
Height of emission. The release
height for the emissions from this
source in meters, or in units which
will be converted to meters when the
computer multiplies by the value
entered for SCAL above. There is no
plume rise in FDM, thus for a source
-------
37
with plume rise, the plume rise must
be computed manually and added to
the stack height and entered here.
71-75 F5.0 Source width or area source rotation
angle. For line sources this
parameter refers to the width of the
line source in meters, or in units
which will be converted to meters
when the computer multiplies by the
value entered for SCAL above. For
area sources, this parameter is the
number of degrees that the axis of
the x dimension entered above is
rotated from zero. This angle can
have a value from -90 degrees to +90
degrees.
76-80 F5.0 Source-specific particle density.
If a value of zero is entered here,
or the space is left blank, the
model will used the globally-
specified density on Card 7 for this
source. If however, the users
wishes the source to have a
different density, it is entered
here.
Card 14A Optional Particle Size data for Source. If the particle
size switch in column 3 of the source card is set to 1,
then this card (or group of cards) is read, otherwise,
this card (or cards) is not read and should not be
included. This card (or cards) specifies the particle
size distribution for this source only and follows the
exact same format as Card 10.
Col Format Information
1-10 8F10.0 The fraction of the emissions which
are contained in each particle size
class are entered here. A total of
20 particle size classes can be
specified and a fraction must be
specified for each particle size
class used. Eight values can be
placed on each card here. Use as
many cards as necessary to provide
the number of particle size classes
specified, but do not include any
blank cards.
Card 15 Meteorological data. Meteorological data are entered
only if the met option switch is set to 1. If
meteorological data are to be entered here, each hour of
data is entered on a separate card. Note that none of
-------
38
the meteorological values are affected by the
specification of SCAL above.
Col Format Information
1-10 F10.0 Wind speed in m/sec.
11-20 F10.0 Wind direction — the direction in
degrees from north from which the
wind is coming.
25 II Stability class, where 6 values are
possible and reflect Turner classes
A-F, and 1=A, 2=B, 3=C, 4=D, 5=E and
6=F.
31-40 F10.0 Mixing Height in meters.
41-50 F10.0 Ambient Temperature in degrees
Kelvin.
-------
39
4.0 VALIDATION/SAMPLE RUNS
Three validation studies were performed using measured air quality and meteorological data. Two
of these were for major surface mining operations and include both a western surface mine and an eastern
surface mine. The third validation study was performed using data collected over a number years at the
Hanford reservation in eastern Washington and generally termed the "Hanford 67" data base. Appendix A
details the validation studies and their results. As the appendices indicates, the FDM model offers improved
performance over the currently available model for fugitive dust impact assessment, the Industrial Source
Complex Model. It should be noted that a fourth validation study was conducted with the FDM and ISC for
the Bunker Hill Smelter site in Idaho. The results of the fourth validation study were consistent with the three
presented here. TRC (1990) provides more detail on the Bunker Hill study.
Appendix B provides samples of input and output streams for the FDM Model.
Appendix C contains sections from the CALINE3 User's Guide for further documentation on the line
source algorithm in the FDM.
Appendix D (bound separately) contains a complete listing of the FORTRAN code for the FDM
Model. The version of the code contained in the appendix is that used for IBM-PC computers. Some minor
changes would be necessary to generate a mainframe computer code from the code contained in the
appendix.
-------
40
(This page intentionally left blank)
-------
41
REFERENCES
California Department of Transportation, 1979. "CALINE3 - A versatile Dispersion Model for Predicting Air
Pollutant Levels Near Highways and Arterial Streets", Office of Transportation Laboratory, Department of
Transportation, State of California, Sacramento, California 95807, No. FHWA/CA/TL-79/23.
Croes, B. A., 1987. Personal communication with Kirk D. Winges, TRC Environmental Consultants, Mountlake
Terrace, Washington.
Ermak, D. L., 1977. "An Analytical Model for Air Pollutant Transport and Deposition from a Point Source,"
Atmospheric Environment. Vol.11, pp. 231-237.
EPA, 1985. "Compilation of Air Pollutant Emission Factors, Vol I: Stationary Point and Area Sources", AP-42,
Fourth Edition, September, U. S. EPA, Office of Air Quality Planning and Standards, Research Triangle Park,
N. C.
EPA, 1987. "Industrial Source Complex (ISC) Dispersion Model User's Guide - Second Edition (Revised),
Volume I., EPA-450/4-88-002a, December 1987.
Horst, T. W. and Doran, J. C., 1984. "Experimental Evaluation of Plume Depletion Models", Fourth Joint
Conference on Applications of Air Pollution Meteorology, 16-19 Oct. Portland, Oregon, American
Meteorological Society, Boston, Massachusetts.
McRae, G. J., 1977. Computer Program for calculation of the friction velocity, Environmental Quality
Laboratory, 206-40, California Institute of Technology, Pasadena, CA 91125.
Nifong, G. D. and Winchester, J. W., 1970. "Particle Size Distributions of Trace Elements in Pollution
Aerosols," University of Michigan, Document No. COO-1705-8, August.
Sehmel, G.A. and W. H. Hodgson, 1978. "Model for Predicting Dry Deposition of Particles and Gases to
Environmental Surfaces", PNL-SA-6721, Battelle Pacific Northwest Laboratories, Richland, Washington.
Stull, Roland B. 1988. "An Introduction to Boundary Layer Meteorology", Kluwer Academic Publishers,
Boston.
TRC.1990. Task 4 Data Report: Deposition Model Evaluation, Bunker Hill RI/FS. Prepared for Dames and
Moore, 1125 17th St., Suite 1200, Denver, CO 80202, Document No. 15852-004/PD193/45030, August 3,
1990.
Turner, D. B. 1970. "Workbook of Atmospheric Dispersion Estimates," AP-26, EPA Research Triangle Park,
N.C.
Woodruff, N. P. and F. H. Siddoway, 1965. "A Wind Erosion Equation", Soil Science Society Proceedings,
pp 602-608.
-------
Fugitive Dust Model (FDM)
First Validation Study
Prepared by;
Kirk D. winges
Francis J. Gombar
Prepared for;
Region 10
U. S. Environmental Protection Agency
1200 Sixth Avenue
Seattle, Washington 98101
Project Administrator:
Robert B. Wilson
April, 1990
-------
Table of Contents
1.0 Introduction A(1)-1
2.0 Methodology and Model Inputs A(1)-3
3.0 Air Quality Modeling Results A(1)-11
4.0 Conclusions A(1)-25
References A(1)-27
-------
1.0 INTRODUCTION
The Fugitive Dust Model (FDM) was developed specifically for computing
concentrations and deposition rates of particulate matter from fugitive dust
sources. This document details a validation study of the model. Model
predictions were computed using daily emission data and on-site meteorology from
a major source of fugitive dust (a western surface mining operation), and the
results were compared with measured values for the same period. Similar
computations were performed with the current model recommended by EPA in the
Guideline on Air Quality Models -- the Industrial Source Complex (ISC) Model.
The FDM Model is designed specifically for computation of the impacts of
fugitive dust sources. It has been under development for many years in several
formats. The primary use of the model is for the computation of concentrations
and deposition rates resulting from emission sources such as open pit mining
operations or hazardous waste sites where fugitive dust is a concern. The model
contains no plume rise algorithm and is thus not aimed at handling significant
buoyant sources. It was recognized from the start that ultimate acceptance of
the model would hinge on its ability to accurately predict concentrations from
fugitive dust sources. To that end, a model validation effort was conceived
using actual fugitive dust emissions and measured particulate concentrations.
This report documents the findings of the validation exercise.
The current report is organized into three sections, in addition to this
introduction. Section 2.0 describes the methodology, and the key modeling input
-------
values such as the model layout used in the current study. Section 3.0 discusses
the modeling results and compares the values to measured values. Finally,
Section 4.0 presents the conclusions of the investigation.
-------
2.0 METHODOLOGY AND MODEL INPUTS
Both FDM and ISC are capable of predicting average concentrations of both
Total Suspended Particulate Matter (TSP) and particulate matter less than 10
micrometers in mass mean diameter (PM-10) for a variety of averaging times. The
averaging periods of interest are those for which standards or PSD increments are
in effect. In most air quality permitting investigations, the period of greatest
concern for fugitive dust impacts is the 24-hour average, since both a National
Ambient Air Quality Standard (NAAQS) and a PSD increment exist for 24-hour PM-10
and TSP concentrations. Although a similar standard exists for annual-average
concentrations, the experience gained from the conduct of air quality permitting
investigations indicates that most fugitive dust emitting projects have far more
difficulty demonstrating compliance with the 24-hour criteria than the annual
criteria. As a result, this investigation focuses on 24-hour average
concentrations. For the FDM model one version of the program deals with all
averaging times, but for the ISC model separate versions are available for
computing short- and long-term averages. The ISCST (for Short Term) version of
the ISC Model was used in this investigation.
The current validation exercise was conducted using data obtained from a
large western surface coal mining operation. The mining operation was selected
for the validation study for the following reasons:
o Mining operations are major sources of fugitive dust, and have
been the subject of numerous air quality studies and
investigations dating back to the early 1970's. Published
emission factors are available for most mining sources, and
most western air pollution agencies have had to deal with the
complex problems associated with computing mining fugitive
dust impacts.
-------
o The mining industry and the particular mining company in
question were very cooperative in providing the data and
information necessary for the model validation.
o The mine in question is a large operation and has an extensive
monitoring network for measurement of both PM-10 and TSP
concentrations at a total of 5 stations located in the
immediate area of the mine. Many have referred to the mine as
the "most monitored mine in the history of the industry".
o In addition to the air quality data, on-site meteorological
data were available for the validation investigation.
o Both the air quality and meteorological data are collected in
compliance with the full requirements of a PSD monitoring
network, including quality assurance provisions. The data are
routinely submitted to the local air pollution agency as part
of the permit for the mining operation.
Data were obtained from the mining company for a period of one entire dry
season, April through September of 1986. Since sampling was conducted on a six-
day cycle for TSP and PM-10, a total of 32 case days were available for the
validation study.
The emission inventory for the current investigation was computed for each
of the 32 case days studied. Published emission factors taken from the
literature were used in the analysis. Generally, reliance was made on the EPA's
emission factor reference, Document AP-42. The mining company provided the input
information needed to compute the emissions from the factors for each of the 32
case days. Information provided by the mining company included tonnage mined,
transported and processed (crushed) on each day, and the tonnage and locations
for disposal of the waste material removed on each day. Other general
information on the equipment in use at the mine and the schedule for each item
were also provided.
The emissions from the mining operation were divided into a total of 56
separate sources for FDM input, based on the actual layout of the mine. For the
haul roads, a total of 27 separate sections of road were identified, and the
-------
actual truck traffic identified for each section. Emissions were computed for
each section of road based on the activity on that section of road for each case
day. Figure 2-1 illustrates the location of the sources as defined by the FDM
Model.
For the ISC model runs, it was not possible to use the same emission source
layout as the FDM runs, since ISC does not have the capability to treat line
sources directly. As a result, each of the FDM line sources was broken into a
series of volume sources for ISC input. Also, ISC does not have the capability
to treat rectangular area sources, so the area source layout was revised for
square area sources only. For the ISC runs a total of 170 individual sources
were used in the modeling. The total emissions for each case day were identical
in the FDM and ISC model inputs.
Particle size distributions assumed in the modeling consisted of five
separate particle size classes: 0-2.5, 2.5-5, 5-10, 10-15 and >15 micrometers.
The modeling results were interpreted in terms of both the total suspended
particulate (TSP) concentrations (the sum of all particle size classes) and the
concentration of particles 10 micrometers and smaller in mass mean diameter (PM-
10). The PM-10 concentrations were computed as the sum of the first three
particle size classes from the modeling. Particle size distributions of emitted
sources were obtained from literature measurements of particle size distributions
in the vicinity of mining operations, including the PEDCo and MRI (EPA, 1981)
investigation of surface mine emissions in 1981, and particle size distributions
reported in EPA's Complilation of Air Pollutant Emission Factors (EPA, 1985).
A total of five air quality monitoring stations are located in the project
area. The stations are identified by number. The locations of the stations are
shown in Figure 2-1. Following are descriptions of each of the stations:
-------
* Air Quality Monitoring Station
® Point Source
— Line Source (Haul Road)
D Area Source
0.5
Scale
1.0 km
Figure 2-1
Mine Layout Used in the
FDM Validation Study
-------
o AQ-1 Located in the center of the mining operation by the
haul road to the waste dump. The station consists of two co-
located PM-10 monitors and a meteorological station.
o AQ-2 Located atop the ridge to the north of the mine. It is
often a background station, with little impact from the mine.
Equipment consists of a PM-10 monitor and a TSP monitor.
o AQ-3 Located to the east of the major mining operations.
Equipment includes a PM-10 monitor and two co-located TSP
monitors.
o AQ-5 Located to the south of the mine. It consists of a PM-
10 monitor and a TSP monitor.
o AQ-6 Located to the north west of the mining operation on a
hill. Equipment consists of a PM-10 monitor and a TSP
monitor.
Measured PM-10 and TSP data at the five monitoring stations were compiled by the
mining company and transmitted to TRC in hard copy and floppy disk format. TRC
extracted the case days from the overall particulate data and input the values
to a "spread-sheet" program for comparison with the model predictions.
The ultimate goal of this investigation was to compare the model
predictions to these measured data. The modeled concentrations, however, contain
only the contribution of the mining operation to the ambient particulate levels,
while the monitored values contain all particulates, whether from the mine or
not. The "background" contribution to the particulate loading is highly variable
and difficult to quantify. The approach taken here for estimation of background
concentrations was to scan all five measured values for each case day, and select
the lowest value as the background. The modeled concentrations, discussed in the
next chapter, are added to the background for each day to determine the total
impact for comparison to the measured values.
Meteorology is measured at several locations in the vicinity of the mine.
-------
Two candidate locations were considered for the air quality modeling study: a
monitor location near AQ-1 and a monitor location near AQ-6 (see Figure 2-1).
Ultimately the AQ-1 meteorology data were selected based on the quality of the
data and the representative nature of the wind speed and wind direction data to
the mine emission sources. Examination of Figure 2-1 shows the location of the
monitor to be central to the emitting sources at the mine.
*
Both the FDM and ISC models require information on the hourly values for
wind speed, wind direction, temperature, atmospheric stability and mixing height.
Wind speed, wind direction and temperature are measured directly by the sensors
at AQ-1. Atmospheric stability is a measure of the turbulent mixing capacity of
the atmosphere and was estimated for the current investigation from the standard
deviation of the wind direction, also recorded at AQ-1, and from the wind speed
and time of day. Stability is expressed as one of 6 classes labeled A through
F, where A is the least stable (greatest turbulent mixing) and F is the most
stable (least turbulent mixing). The conversion from standard deviation of the
wind direction, wind speed and time of day to stability is accomplished as
follows (taken from EPA, 1987):
Intital Estimate of Stability Class is provided by:
Standard Deviation Stability Class
> 22.5 A
17.5-22.5 B
12.5-17.5 C
7.5 - 12.5 D
3.75 - 7.5 E
< 3.75 F
Stability Classes are Adjusted by time of day and wind speed by:
-------
Time of Dav Initial Stabilitv Wind Speed (m/sec) Final Stab
Daytime A
3
4
6
B
4
6
C
6
D, E or F
Nighttime A
2.9
3.6
B
2.4
3.0
C
2.4
D
E
5.0
F
3.0
5.0
U < 3
s; U < 4
i U < 6
* U
U < 4
i U < 6
s U
U < 6
<: U
Any
U < 2.9
<: U < 3.6
<; U
U < 2.4
<; U < 3.0
z U
U < 2.4
<: U
Any
U < 5.0
<; U
U < 3.0
<; U < 5.0
s U
A
B
C
D
B
C
D
C
D
D
F
E
D
F
E
D
E
D
D
E
D
F
E
D
For fugitive dust impacts, results are generally very insensitive to values
used for mixing height because the emissions are released at or near the ground,
and the impacts are generally very close to the source. As a result the
emissions have little opportunity to mix vertically to the height of the mixing
layer. To provide the models with values for these required values, mixing
heights were assigned by stability class using the following general values:
-------
Stability Class Mixing Height (m)
A 1,600.
B 1,200.
C 800.
D 400.
E 10,000.
F 10,000.
-------
3.0 AIR QUALITY MODELING RESULTS
The FDM and ISC models were run for the 32 case days identified earlier and
the predicted concentrations, both PM-10 and TSP, computed as the sum of the
modeled concentration and the background as discussed earlier. The results are
presented in two formats here. First, the measured versus predicted values are
shown in Tables 3-1 through 3-4 for FDM TSP, ISC TSP, FDM PM-10 and ISC PM-10
respectively. Second, a "scatter plot" of the measured and predicted values for
these same four cases are shown in Figures 3-1 through 3-4. The scale of the
four plots are all identical.
The performance of each model is generally good for most of the days given
the usual accuracy of air quality models. However, the figures show a tendency
on the part of ISC for TSP over-predictions on a few case days. It is these case
days which are of greatest concern to regulators, since the 24-hour TSP standards
and PSD increments refer only to the highest one or two days per year. The PM-10
results are much closer for the two models. Since PM-10 represents the smaller
particles which would be expected to encounter less deposition than the TSP
particles, the FDM and ISC Model predictions would be expected to be closer for
PM-10.
A number of different techniques, including cumulative frequency plots, and
various statistical functions have been used in the past to evaluate air quality
model performance. Air quality models are frequently quoted to predict within
a factor of two, thus one means of comparison is to determine what number of the
data points are within a factor of two. For FDM, the TSP predicted results are
-------
Table 3-1. Comparison of Measured and FDM Predicted TSP Concentrations (ug/m3)
Day
Date
AQ-2
Meas. Pred.
AQ-3
Meas. Pred.
AQ-5
Meas. Pred.
AQ-6
Meas.
Pred.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
3/21/86
3/27/86
4/8/86
4/14/86
4/20/86
4/26/86
5/2/86
5/8/86
5/14/86
5/20/86
5/26/86
6/1/86
6/7/86
6/13/86
6/19/86
6/25/86
7/2/86
7/7/86
7/13/86
7/19/86
7/25/86
7/31/86
8/6/86
8/12/86
8/18/86
8/24/86
8/30/86
9/5/86
9/11/86
9/17/86
9/23/86
9/29/86
40.9
9.6
14.8
23.2
13.5
7.9
10.2
19.9
11.9
11.7
36.6
20.2
34.8
13.8
37.2
27.6
25.4
36.7
32.2
22.6
34.3
41.6
36.8
27.6
36.9
28.9
39.5
8.7
10.9
15.7
10.9
8.6
9.4
18.2
6.1
8.4
27.6
18.3
28.7
13.8
39.3
25.9
18.7
31.1
28.5
15.9
29.3
30.9
30.5
19.1
28.2
27.2
22.7
40.0
21.2
14.3
14.9
10.1
18.3
44.0
26.7
48.8
23.9
76.2
43.0
64.8
46.9
35.5
33.5
58.7
57.8
52.1
57.2
45.4
32.5
38.3
43.8
15.4
10.9
94.4
9.4
18.3
23.9
18.5
61.2
24.8
66.0
42.4
76.3
31.1
31.4
40.3
46.1
54.1
47.6 :
51.8
28.2
27.4
35.8
22.6
10.5
17.9
16.2
5.8
17.3
33.1
16.1
11.5
27.5
33.0
27.3
29.2
30.7
17.3
31.1
31.7
15.4
34.5
29.5
53.8
26.6
35.5
45.7
64.4
52.7
9.5
37.7
30.5
11.1
11.1
9.8
15.3
11.4
4.7
9.5
23.3 .
11.7
9.3
27.7
19.1
27.5
15.0
27.2
17.3
31.1
32.1 •
15.4
34.9
28.0
31.4
19.8
28.3
27.2
32.5
33.0
6.0
33.9
12.9
; 11.1
39.3
7.6
9.8
15.3
; 10.9
4.7
9.4
18.2
6.1
8.4
37.7
18.3
58.0
15.5
34.5
25.1
21.8
32.0
28.3
25.7
28.9
28.0
30.1
18.9
28.2
27.1
25.7
32.3
4.7
26.9
: 12.9
14.8
53.1
9.7 .'
13.9
15.3
10.9
4.7
9.4
18.2 .
6.1
8.4 .
27.6
18.3
46.0
13.8
45.7 ;
25.9
28.4
32.2
28.3 :
58.9 ;
30.7 :
37.9
31.6
23.0
28.2 ;
27.1
26.6
44.3 :
4.7
45.9
13.4
-------
Table 3-2. Comparison of Measured and ISC Predicted TSP Concentrations (ug/m3)
Day
Date
AQ-2
Meas. Pred.
AQ-3
Meas. Pred.
AQ-5
Meas. Pred.
AQ-6
Meas.
Pred.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
3/21/86
3/27/86
4/8/86
4/U/86
4/20/86
4/26/86
5/2/86
5/8/86
5/14/86
5/20/86
5/26/86
6/1/86
6/7/86
6/13/86
6/19/86
6/25/86
7/2/86
7/7/86
7/13/86
7/19/86
7/25/86
7/31/86
8/6/86
8/12/86
8/18/86
8/24/86
8/30/86
9/5/86
9/11/86
9/17/86
9/23/86
9/29/86
40.9
9.6
14.8
23.2
13.5
7.9
10.2
19.9
11.9
11.7
36.6
20.2
34.8
13.8
37.2
27.6
25.4
36.7
32.2
22.6
34.3
41.6
36.8
27.6
36.9
28.9
39.4
9.6
10.8
16.9
10.9
8.9
9.4
18.2
6.1
8.4
27.6
18.3
29.2
13.8
48.1
25.9
19.6
31.1
28.5
15.8
29.4
31.2
30.4
19.1
28.2
27.5
22.7
40.0
21.2
14.3
14.9
10.1
18.3
44.0
26.7
48.8
23.9
76.2
43.0
64.8
46.9
35.5
33.5
58.7
57.8
52.1
57.2
45.4
32.5
49.3
41.6
15.4
10.9
83.7
9.4
18.2
25.3
18.5
80.3
34.8
61.6
40.6
114.6
31.1
30.8
34.5
59.7
50.2
43.7
90.4
28.2
27.4
35.8
22.6
10.5
17.9
16.2
5.8
17.3
33.1
16.1
11.5
27.5
33.0
27.3
29.2
30.7
17.3
31.1
31.7
15.4
34.5
29.5
53.8
26.6
35.5
45.7
64.4
52.7
9.5
37.7
30.5
11.1
16.8
9.8
15.3
11.4
4.7
9.4
23.8
20.2
9.7
28.0
19.4
27.7
15.0
29.3
17.3
31.1
39.9
15.4
48.8
28.0
33.1
22.1
28.3
27.3
46.3
34.6
6.6
48.4
12.9
11.1
39.3
7.6
9.8
15.3
10.9
4.7
9.4
18.2
6.1
8.4
37.7
18.3
58.0
15.5
34.5
25.1
21.8
32.0
28.3
25.7
28.9
28.0
30.1
18.9
28.2
27.1
25.7
32.3
4.7
26.9
12.9
24.6
79.8
11.1 '
19.2
15.3 ..
10.9
4.7
9.4
18.2
6.1
8.4 '
27.6
18.3 :
73.5
13.8
62.6
25.9
43.2
35.1
28.3
77.3
30.8
55.1
32.0
30.5
28.2
27.1
26.6
58.8
4.7
81.1
13.5
-------
Table 3-3. Comparison of Measured and FDM Predicted PM-10 Concentrations (ug/m3)
AQ-1 AQ-2 AQ-3 AQ-5 AQ-6
Day Date Meas. Pred. Meas. Pred. Meas. Pred. Meas. Pred. Meas. Pred.
1
2
3
4
5
6
7
8
9
10
11
12
13
H
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
3/21/86
3/27/86
4/8/86
4/14/86
4/20/86
4/26/86
5/2/86
5/8/86
5/14/86
5/20/86
5/26/86
6/1/86
6/7/86
6/13/86
6/19/86
6/25/86
7/2/86
7/7/86
7/13/86
7/19/86
7/25/86
7/31/86
8/6/86
8/12/86
8/18/86
8/24/86
8/30/86
9/5/86
9/11/86
9/17/86
9/23/86
9/29/86
17.0
36.9
8.6
21.4
17.3
14.7
4.5
16.2
16.0
7.4
6.4
19.5
15.5
20.0
15.7
29.5
28.2
24.1
23.1
24.9
12.0
25.8
26.3
43.2
38.7
24.1
19.9
65.9
35.0
6.7
27.0
20.1
38.7
46.8
53.3
35.2
27.3
22.7
11.5
37.9
27.3
40.9
4.3
16.6
34.4
26.8
33.8
33.2
39.6
22.9
21.0
40.1
12.6
68.0
25.5
73.3
67.8
35.7
17.6
63.0
43.3
39.2
40.9
57.3
6.0
22.5
4.2
4.7
11.8
7.2
4.5
4.1
9.5
3.5
20.7
8.4
10.0
24.4
19.0
14.1
22.6
19.9
13.8
21.9
24.7
22.8
15.7
22.1
16.9
7.7
19.4
4.2
5.1
10.5
5.4
4.1
4.1
8.5
2.7
14.3
8.4
7.6
22.8
13.9
11.4
19.7
17.3
10.7
17.3
19.9
18.8
12.1
18.9
14.5
15.4
14.6
10.2
12.1
5.4
6.1
5.0
8.9
10.2
5.1
18.7
11.2
23.2
10.1
31.8
20.7
25.8
22.7
18.8
15.5
29.6
26.6
25.6
33.7
21.0
14.6
18.7
15.0
16.2
10.1
5.4
32.3
4.1
8.5
8.5
4.1
14.4
8.5
36.4
12.4
29.8
19.2
43.4
19.7
18.2
18.4
24.8
27.2
24.3
28.4
18.9
14.5
12.4
13.5
5.5
10.6
6.7
3.7
6.1
12.5
3.5
4.5
14.3
13.3
16.3
11.3
19.2
14.9
10.6
19.7
18.7
10.5
21.9
19.7
30.1
16.5
21.1
34.1
24.9
4.4
13.5
4.4
6.3
4.7
10.1
5.6
2.7
4.1
10.2
6.0
4.4
14.5
8.8
16.4
8.0
19.2
14.7
10.6
19.7
20.5
10.5
22.5
18.7
19.7
12.8
14.4
21.0
19.2
3.7
8.5
4.4
19.3
3.6
5.8
10.1
5.9
2.7
4.2
8.5
2.7
4.0
19.9
9.8
23.6
7.6
21.4
13.6
12.6
17.2
17.2
18.7
18.7
12.0
18.9
14.4
16.0
18.6
3.0
19.7
8.5
7.7
30.1 •
4.6
7.4
10.1
5.4
2.7
4.1
8.5
2.7
4.0
14.3
8.4
30.1
7.6
28.9
13.9
17.7
17.2
17.8
25.4
19.3
15.0
18.9
14.4
16.3
26.1
3.0
34.4
8.7
-------
Table 3-4. Comparison of Measured and ISC Predicted PM-10 Concentrations (ug/m3)
AQ-1 AQ-2 AQ-3 AQ-5 AQ-6
Day Date Meas. Pred. Meas. Pred. Meas. Pred. Meas. Pred. Meas. Pred.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
3/21/86
3/27/86
4/8/86
4/U/86
4/20/86
4/26/86
5/2/86
5/8/86
5/14/86
5/20/86
5/26/86
6/1/86
6/7/86
6/13/86
6/19/86
6/25/86
7/2/86
7/7/86
7/13/86
7/19/86
7/25/86
7/31/86
8/6/86
8/12/86
8/18/86
8/24/86
8/30/86
9/5/86
9/11/86
9/17/86
9/23/86
9/29/86
17.0
36.9
8.6
21.4
17.3
14.7
4.5
16.2
16.0
7.4
6.4
19.5
15.5
20.0
15.7
29.5
28.2
24.1
23.1
24.9
12.0
25.8
26.3
43.2
38.7
24.1
19.9
65.9
35.0
6.7
27.0
20.1
38.0
54.3
55.8
38.2
28.2
22.4
9.6
38.1
30.2
41.8
4.3
16.4
32.6
27.4
32.2
33.2
39.6
22.7
20.9
41.7
12.2
73.2
24.6
71.5
76.1
36.9
16.8
63.4
43.9
41.5
39.2
56.2
6.0
22.5
4.2
4.7
11.8
7.2
4.5
4.1
9.5
3.5
20.7
8.4
10.0
24.4
19.0
14.1
22.6
19.9
13.8
21.9
24.7
22.8
15.7
22.1
16.9
£__
10.0
19.3
4.2
5.0
10.8
5.4
3.9
4.1
8.5
2.7 :
14.3
8.4
7.6
26.0
13.8
11.8
19.7
17.3
10.6
17.3
19.7
18.8
12.1
18.9
14.6
15.4
14.6
10.2
12.1
5.4
6.1
5.0
8.9
10.2
5.1
18.7
11.2
23.2
10.1
31.8
20.7
25.8
22.7
18.8
15.5
29.6
26.6
25.6
33.7
21.0
14.6
16.0
11.6
13.9
10.1
5.4
25.5
4.1
8.5
8.2
4.1
14.4
8.4
33.6
13.9
27.1
18.1
42.2
19.7
17.9
16.0
26.6
25.1
22.6
35.1
18.9
14.5
12.4
13.5
5.5
10.6
: 6.7
3.7
6.1
': 12.5
3.5
4.5
14.3
13.3
16.3
11.3
19.2
14.9
10.6
19.7
18.7
10.5
21.9
19.7
30.1
16.5
21.1
34.1
24.9
4.4
13.5
4.4
7.4
4.7
10.1
5.6
2.7
4.1
10.1
7.5
4.4
14.5
8.8
16.5
7.9
19.2
14.9
10.6
19.7
22.3
10.5
25.3
18.7
20.1
13.4
14.5
23.4
19.8
3.6
8.5
4.4
19.3
3.6
5.8
10.1
5.9
2.7
4.2
8.5
2.7
4.0
19.9
9.8
23.6
7.6
21.4
13.6
12.6
17.2
17.2
18.7
18.7
12.0
18.9
14.4
16.0
18.6
3.0
19.7
8.5
10.9
35.7 '
4.6
7.8
10.1
5.4
2.7
4.1
8.5
2.7
4.0
14.3
8.4
37.2
7.6
33.0
13.8
18.8
17.2
17.7
27.6
19.2
15.8
18.9
14.4
16.3
28.5
3.0
41.3
8.7 :
-------
200.0
150.0
j 100.0
50.0
0.0
—i—i—I—i—i—i—i—I—I—i—I—i—I—i—i—i—r
50 100 150 200
Measured Concentration (ug/m3) Including Background
FIGURE 3-1 FDM EVALUATION FOR JSP
200.0
150.0
100.0
50.0
0.0
50 100 150
Measured Concentration (ug/m3) Including Background
200
FIGURE 3-2 ISC EVALUATION FOR TSP
-------
p
b
OJ
I
o
IJ*
o a
z
o
70
o 3
Predictod Concentration (ug/m3) Including Background
S 8 8
bob
8
b
p
o
O
TO
m
OJ
IF
o
TO
-o
z:
I
o
Predicted Concentration (ug/m3) Including Background
-* -* lo
go u o
o o o
b b b b
-------
within a factor of two of the measured results for 95 percent of the values. For
the FDM PM-10 results, the measured and predicted values are within a factor of
two for 91 percent of the values. For the ISC results, the same comparison shows
93 percent for TSP and 89 percent for PM-10.
EPA has recently been recommending a new method for evaluation of model
performance (Cox et. al., 1988). It also centers on the concept of accuracy
within a factor of two, but utilizes a more complicated comparison. There are
two steps in the evaluation procedure. First, a screening computation is
completed using two quantities, the fractional bias for the average values and
a fractional bias for the standard deviation. They are defined as follows:
FB
(OB + PR)/2
where: FB = fractional bias of the average
OB = average of highest 25 observed values
PR = average of highest 25 predicted values
S - S
TO-
(S + S )/2
o p"
where: FO = fractional bias of the standard
deviation
SQ = standard deviation of the
highest 25 observed values
S = standard deviation of the
highest 25 predicted values
The screening evaluation is performed by computing both of the above
parameters, and plotting on a special graph. The second level of analysis is
more complex. The second level is called the statistical test and involves using
the same fractional bias computation as above, but rather than using the average
and standard deviations of the observed and predicted values, the technique uses
a parameter called the robust estimate of the highest concentration (RHC). In
addition, the computation of the fractional bias is done for several averaging
-------
periods and differing meteorological conditions and the results used to compute
a composite performance measure. Finally, a statistical technique called
"bootstrapping" is used where values are extracted at random from the overall
data set to create a "sampled" data set, which is used in the computation of
these same performance measures. By conducting this random sampling many times,
the statistician can determine if differences in model performance are
statistically significant. More details on the technique can be found in Cox's
paper.
Using the screening technique, for the TSP concentrations in the current
model evaluation, the computed values for the FB for FDM was -0.018 and the FO
was -0.270. For the PM-10 concentrations the FB for FDM was -0.364 and the FO
was -0.309. For ISC, the TSP values were -0.246 for the FB and -0.559 for the
FO, while for PM-10 the values were -0.400 for the FB and -0.357 for the FO. The
values are plotted in Figures 3-5 and 3-6 for TSP and PM-10 respectively. The
box at the center of the figure is an indication of the "factor of two"
performance of the model. If the data plots within the box, then the model is
said to have performed within a factor of two. Since the current model
evaluation results for FDM show both TSP and PM-10 plot within the box, the model
performance for FDM is judged to be within the customary factor of two that EPA
and others use as a guide. Similarly, the results for ISC plot inside the box
but the values are further from the center of the box than the FDM results. As
was observed in the scatter plot comparison, the difference between FDM and ISC
is greater for TSP than for PM-10.
The second level of performance evaluation was a more complex undertaking.
The technique has been developed primarily for predicting concentrations of
sulfur dioxide or other gaseous compounds for which the data available generally
-------
O>
3
ia
m
o_
c
a
i-*-
o'
u
'70
CD
in
c
A
I
M
Avera
Fractional Bias of the Standard Deviation
I
3!
l£)
C
5
OJ
l
OI
to
o
(D
3_
3'
!?
n>
W
c
to
I
10
Fractional Bias of the Standard Deviation
i o -
I i i i i
-------
include hourly observations of S02 concentration and meteorology on a continuous
basis for a year or more. The measurement of particulate usually is done in 24-
hour integrated samples. As a result, modifications had to be made to the
statistical evaluation methods to apply them to the current application. The
modifications to the technique of Cox are summarized as follows:
o Only 24-hour values were available, thus only the only
averaging time in the evaluation was 24-hour. Cox refers to
a calculation of a "scientific" evaluation which uses 1-hour
average concentrations. This computation was dispensed with.
Given the single averaging time used here, the composite
performance measure used here was equal to the Absolute
Fractional Bias of the RHC values for the 24-hour samples.
o Since only 32 case days were examined, and since data were not
available at all stations for all of the days, it was
determined to combine all of the data into a single sampling
set for the purposes of computing the RHC, rather than
conducting the computation on a site-by-site basis as the
guidance suggests. The data sets would have been too small if
the separation of the values by site had been performed.
o The bootstrapping technique calls for the construction of a
number of trial "years" by sampling the data set. Since
sampling a six month, intermittent data set to create a full
year of data, would extend the data beyond its measurement
bounds, the sampling was performed only to create a trial set
equivalent in size to the original data set. Thus for TSP,
111 values are in the original data set and each bootstrap
sample was composed of 111 randomly-sampled values. Note that
no persistance of 3 days was used since the data are sampled
on a six-day cycle, and persistence is not relevant.
The bootstrapping analysis was completed for both TSP and PM-10 values for
both models. Although not customarily presented in this fashion, the frequency
distribution of the Fraction Bias of the RHC's calculated in the bootstrapping
analysis for TSP are shown in Figure 3-7 and for PM-10 in Figure 3-8. Note that
the figure presents the fractional bias, not the absolute fractional bias. As
the figure shows, there is a separation between the FDM values and the ISC values
for TSP, indicating a different performance. Note that the values for ISC are
all negative, while the values for FDM are closer to zero, but predominantly
-------
positive.
When the results for TSP and PM-10 are presented in the more customary
fashion for 95% confidence limits, as shown in Figure 3-9, it is clear that the
FDM Model is closer to reality than the ISC Model. However, there is
considerable overlap between the confidence limits of the two models. Note that
in Figure 3-9, the absolute value of the fractional bias is plotted.
The previous discussion would lead to the conclusion that generally, the
FDM model performs better with the data than the ISC Model. In actuality, as
earlier figures show, both models do reasonably well for the majority of data
points. However, ISC has the tendency for over-prediction on few days.
Unfortunately, it is these days which are the focus of the permitting
regulations. Most regulations concern the maximum or second highest
concentration, so the ISC over-prediction on these days causes very misleading
results in air quality permitting studies. It tends to occur under stable, low
wind speed conditions.
One of the major advantages of the FDM approach is the avoidance of these
large over-predictions. The improved prediction occurs due to the superior
deposition algorithm in the FDM Model. During low wind speed stable conditions,
the ISC model allows very high concentrations to be predicted, not reflecting the
deposition which would occur during the long travel times to the receptor. FDM
more accurately represents the behavior of particles in the atmosphere.
A(l)-22
-------
u.zu —
~
0.15 -
c
o
1
s 0.10 -
I :
_
0.05 -
-
-i
nnn -
i-r>ii
ISC
1
1, .
1
' 1
n
\l
1
f
i n
• /
^
\
-2
-1
0 1
Fractional Bias of the Average
Figure 3-7. Fractional Bias Distributions for TSP
u.zu —
0.15 -
c
5 0.10 -
|
1"
0.05 -
_
0.00 -
I-I-M t
rUM
ISC
i
i
i
1
iJ
1
il
ii
-.
> i '
1 i '
-2
-1
0 1
Fractional Bias of the Average
Figure 3—8. Fractional Bias Distributions for PM—10
-------
1.5
1.0 -
co
D
in
D
c
g
V>
o
D
CD
o
to
jQ
< 0.5
0.0
FDM TSP
ISC TSP
FDM PM-1C
ISC PM-10
Figure 3-9. Absolute Fractional Bias at 95ss Bootstrap Confidence Bounds
-------
4.0 CONCLUSIONS
The previous analysis has determined that the FDM Model performs generally
well in characterizing particulate concentrations in the vicinity of the fugitive
dust source. The ISC model also performed well for the bulk of the samples
analyzed, but failed on the high end of the statistical distribution for TSP,
leading to over-predictions of the highest and second highest concentrations,
which are the focus of many air quality regulations for short-term particulate
concentrations. The FDM Model is judged to be superior to the ISC Model in
predicting the impacts from fugitive dust sources for the data evaluated in this
study.
A(l)-25
-------
(This page intentionally left blank)
A(l)-26
-------
REFERENCES
Cox, W. M. 1988. "Protocol for Determining the Best Performing Model", U. S.
Environmental Protection Agency Report, August.
EPA, 1981. "Improved Emission Factors for Fugitive Dust from Mining Sources,
First Draft Project Report", Contract No. 68-03-2924, Work Directive No. 1,
Industrial Environmental Research Laboratory, Cincinnati, OH and Office of Air
Quality Planning and Standards, Research Triangle Park, NC.
EPA, 1985. "Compilation of Air Pollutant Emission Factors, Vol I: Stationary
Point and Area Sources", AP-42, Fourth Edition, September, U. S. EPA, Office of
Air Quality Planning and Standards, Research Triangle Park, N. C.
EPA, 1987. "On-site Meteorological Program Guidance for Regulatory
Applications", EPA Document No. EPA-450/4-87-013, OAQPS, Research Triangle Park,
NC, June, Table 6-6a and 6-6b, p 6-28.
Winges, K. D. , 1982. "Development of an Air Quality Model for Mining Fugitive
Dust", Presented at the Annual Meeting of the Air Pollution Control Association,
New Orleans, June.
A(l)-27
-------
Fugitive Dust Model (FDM)
Second Validation Study
Prepared by;
Kirk D. Winges
Francis J. Gombar
Prepared for:
Region 10
U. S. Environmental Protection Agency
1200 Sixth Avenue
Seattle, Washington 98101
Project Administrator:
Robert B. Wilson
April, 1990
-------
Table of Contents
1.0 Introduction A(2)-1
2.0 Methodology and Model Inputs A(2)-3
3.0 Air Quality Modeling Results A(2)-13
4.0 Conclusions A(2)-43
References A(2)-45
-------
1.0 Introduction
The Fugitive Dust Model (FDM) was developed as an alternative to the previously recommended
Industrial Source Complex Model (ISC) for the purposes of computing fugitive dust impacts. An initial
validation study was performed using measured air quality and meteorological data from a western
surface mining operation. The results of the first validation study were encouraging, but the number of
data points was limited. A second data base was identified for the purpose of conducting an additional
validation study for the FDM. The second data base differed in several important ways from the first data
base. The first data base was collected for the purposes of determining compliance with air quality
concentration regulations. As such, it was collected at locations which were predominantly at property
boundary locations. Since the surface mine examined in the first validation study had considerable buffer
space between the actual sources of dust and the property boundary, the source receptor distances were
on the order of 500 meters or more.
The data base for the current (second validation) study was collected for research purposes, not
for compliance purposes. As a result, the monitoring locations were very close to the sources -- in many
cases less than 100 meters away. The reported concentrations were much higher. The second validation
data set was collected at an eastern surface mining operation which also posed some unanswered
questions. Emission rates for both the first and second validation studies were based on emission factors
which had been developed at western surface mining operations, and the applicability to the higher
humidity and other different meteorology of the eastern setting was a major uncertainty of the study.
The purpose of the current document is to detail the second validation study. Like the first
validation study, reported by TRC in the FDM User's Guide, (April, 1990)1; TSP, PM-10, and
meteorological measurements were available at several locations; in addition an extensive record of
mining information was used to generate a daily fugitive emissions inventory. Daily emissions were
modeled for 120 days during the late summer and fall of 1985 and were compared with TSP and PM-10
monitoring data using the statistical techniques recommended by EPA (Cox et. al., 1988)2. Statistical
approaches have historically been used to evaluate model performance in comparing preferred regulatory
dispersion models with non-regulatory modeling approaches. Preferred regulatory models are listed in
EPA's Guideline on Air Quality Models (EPA-450/2-78-027R, including Supplement A, July 1987). This
analysis demonstrates that FDM is superior to the EPA ISCST Model (EPA-450/4-88-002) for modeling
fugitive dust sources.
-------
(This page intentionally left blank)
A(2)-2
-------
2.0 Methodology and Model Inputs
The key to any model comparison evaluation study is a well designed ambient and meteorological
monitoring study coupled to a high quality emissions inventory. This evaluation was possible because
following data were available:
Daily TSP and PM-10 measurements were made at five locations throughout the mining
complex. The measurement sites were located to provide good indications of impacts
of mining, loading and materials transport activities.
Detailed site activity information was available to quantify emissions from the variety of
sources located at the mine. A high quality, daily emissions inventory was generated for
model input.
On site meteorological data were available to define hourly fugitive emissions transport.
The data consisted of wind speed and direction, sigma theta, ambient temperature and
local rainfall.
Figure 1 depicts the locations of the five ambient monitoring sites and all fugitive dust sources as defined
for ISCST input. Both ISC and FDM can treat area sources (although they use a different method), but
one of the major differences between FDM and ISC is that FDM treats line sources directly, instead of
treating them as a series of "volume" sources as shown in the figure. The volume sources are very similar
to point sources, except the point has an initial horizontal dimension and an initial vertical dimension. The
initial horizontal dimension is shown as the diameter of the circles in Figure 1. The initial horizontal
dimension is input to the model as the initial value for the standard deviation of the concentration in the
crosswind dimension, labeled sigma-y zero (a 0). The initial vertical dimension is similarly input to the
model as the initial standard deviation of the concentration in the vertical, labeled sigma-z zero (az0). The
size of the area sources, as depicted, are approximately those of the input sources. The volume sources
(coal and overburden haul roads) are depicted in the figure, although all are defined with a oy0 of 58.0
meters and a a „ as 2.14 meters. Line sources were used to define haul roads for the FDM modeling
analysis.
The surface mining operation is located in mildly rolling terrain typical of the mid-west and east.
The mining company provided on-site activity data including daily overburden and coal removal quantities
from active mining locations. The mine included 21 active coal removal locations of various area
dimensions. The 21 coal removal sites were segmented into 63 square area sources as required by the
ISCST model. Since ISCST does not calculate impacts for line sources, all haul roads were modeled as
volume sources. Overburden haul roads were segmented into 85 volume sources and coal haul roads
A(2)-3
-------
D
D Area Sources
® Center Of Volume Sources
* Monitor
Figure 1. Source and Receptor Locations in the ISC Model
into 152 volume sources. The mine activity records provided by the company indicated which excavation
areas were active on a given day and the corresponding haul roads utilized for coal and overburden
removal. Model performance for each day was evaluated using the most representative source emissions.
The FDM model was designed to model fugitive dust from mining activities defined as area
sources and dirt roadways using the line source algorithm in CALINE33. The 153 coal haul road volume
sources input to ISCST were represented as 47 line segments carefully located to represent actual road
locations. The 85 overburden haul roads were represented by 21 line sources. All area sources were
input to FDM as defined for ISCST. Unlike ISCST, FDM can accept rectangularly-shaped area sources
whose sides can be at any angle to the X-Y axis. However, the comparison study was performed with
emphasis placed on the treatment of fugitive dust line sources and all area sources were defined
A(2)-4
-------
equivalently in both models.
Total fugitive emissions for August to December, 1985 are listed in Table 2-1. The tonnage of TSP
generated by each activity was based on total coal and overburden removal and emissions factors
reported in the EPA compilation of emissions factors AP-42. The emission estimates were derived using
the parameters specified in Table 2-2. Table 2-1 indicates that over half of the emissions are associated
with access and haul roads.
Two sets of input data for each model were generated for each modeling day, a TSP and a PM-10
input set. The particle size distribution for the TSP data sets were segregated into particle characteristics
as shown in Table 2-3.
The particle size distribution was derived from a universal distribution characteristic of western
mining (PEDCo and TRC, 1982)4. All values were calculated using the techniques specified in the user's
guide for ISCST5. PM-10 input files utilized only that portion of the input set specific to PM-10.
Virtually all of the emission factors used to estimate the fugitive dust emission rates were taken
from the EPA's Compilation of Fugitive Dust Emission Factors, Document No. AP-42. The emission factor
document does not provide PM-10 fractions for all fugitive dust emission rates. Emissions estimates for
blasting, truck loading, dozing, dragging, scrapers (travel mode), grading, light and medium vehicle traffic
and haul trucks are specified as PM-15 sized particles. Therefore, PM-10 emissions for these activities
were estimated from the western mining studies mentioned above assuming that mass fractions of
particles less than 15 microns from the universal particle distribution matched those at the site. PM-10
emissions were calculated from PM-15 emissions by multiplying by 0.646.
Fugitive dust from haul roads, overburden haulback and light and medium duty vehicle traffic was
controlled by water spraying. Control efficiencies can vary between 0.0 percent for no watering to 55
percent for twice per hour watering. Control efficiencies at the mine varied from 21.7 percent in August
to 4.8 percent in December based on watering records provided by the mining company. Roadway
watering was discontinued in the later part of the study period as natural precipitation levels increased
and dropping ambient temperatures made roadway watering less desirable.
Two template input data sets, one using the ISCST format and the second using the FDM format,
were developed including all sources emissions possible from the mine. A FORTRAN program was written
to integrate the mine activity data and define the source emissions that would occur for each day of
monitoring data. The FORTRAN code produces an FDM and an ISCST input data set for TSP and for PM-
10 for each day. The code also extracts the 24 hours of meteorological data for that particular day to be
included in the input data sets.
A background value was added to each value of TSP and PM-10 predicted by ISCST and FDM.
Background was chosen as the lowest measured value from the five monitors (usually monitoring site #5).
A(2)-5
-------
Table 2-1: Total TSP Mine Emissions; August - December, 1985
1.
2.
3 .
4.
5.
6a.
6b.
6c.
6d.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
ACTIVITY
TOPSOIL REMOVAL
SCRAPER TRAVEL: TOPSOIL
TOPSOIL REPLACEMENT
OVERBURDEN DRILLING
OVERBURDEN BASTING
OVERBURDEN REMOVAL: DRAGLINE
OVERBURDEN REMOVAL: TRUCK/SHOVEL
OVERBURDEN REMOVAL: "B" MACHINE
OVERBURDEN REMOVAL: SCRAPER/DOZER
HAUL TRUCK TRAVEL: OVERBURDEN
OVERBURDEN SHAPING: DOZERS
COAL LOADING INTO HAUL TRUCKS
HAUL TRUCK TRAVEL: COAL
COAL DUMP
COAL PILE MAINTENANCE: DOZERS
WIND EROSION
LT DUTY VEHICLE/ACCESS ROADS
ROAD MAINTENANCE: GRADERS
CRUSH, SCREEN, CONVEYING
COAL LOADOUT
AUGER
TOTAL EMISSIONS:
TSP
(TONS)
17.16
N/A
12.45
9.60
8.42
313.10
0.15
0.03
7.84
89.63
8.64
30.80
378.49
0.01
3.86
23.63
210.85
20.59
0.74
0.03
N/A
1,132.02
PERCENT OF TOTAL
1.5
N/A
1.1
0.8
0.7
27.7
0.01
NEGLIGIBLE
0.7
7.6
0.8
2.7
33.4
NEGLIGIBLE
0.3
2.1
18.6
1.8
0.1
NEGLIGIBLE
N/A
Source: TRC Environmental Consultants, Inc.
A(2)-6
-------
Table 2-2: Input Parameters Used To Estimate Emissions
MINING OPERATION
Overburden blasting
Overburden removal
dragline
Dozers - overburden
Coal loading
Haul truck: coal
Haul truck: overburden
Coal dump
Dozers coal
Light- and medium- duty
vehicle
Graders
Coal loadout
DESCRIPTION
area blasted (A)
drop distance (d)
moisture (M)
silt content (s)
moisture (M)
moisture (M)
no. of wheels (w)
silt loading (L)
no. of wheels (w)
silt loading (L)
silt content (s)
wind speed (U)
drop height (d)
moisture (M)
dump capacity (Y)
silt content (s)
moisture (M)
moisture (M)
speed (S)
silt content (s)
wind speed (U)
drop height (d)
moisture (M)
AP-42 INPUT
VALUES
3,113 m2/blast
20.0 ft
16.8%
6.9%
16.8%
15.0%
10.0
9.1 g/m2
6.0
9.1 g/m2
6.0%
5 . 46 mph
5.0 ft
15.0%
50.0 yd3
6.0%
15.0%
1.7%
7 . 1 mph
6.0%
8.9 mph
5.0 ft
15.0%
Source: TRC Environmental Consultants, Inc.
A(2)-7
-------
Table 2-3. Particle Characteristics Used in the Model Runs
DIAMETER
RANGE
(microns)
0 - 10
10 - 20
20 - 30
MASS
FRACTION
0.270
0.263
0.467
ISCST
SETTLING
VELOCITY
(m/sec)
0.00074
0.00669
0.01859
ISCST
REFLECTION
COEFFICIENT
0.93
0.80
0.72
Note: For FDM, the settling velocities are computed by the model, and a
second particle characteristic, the deposition velocity, is also computed
by the model on an hour by hour basis. The reflection coefficient is used
only by ISC.
A(2)-8
-------
Both the FDM and ISCST models require hourly values for wind speed, wind direction,
temperature, atmospheric stability and mixing height. Wind speed, wind direction and temperature are
measured directly by the sensors at monitor #1. Atmospheric stability is a measure of the turbulent
mixing capacity of the atmosphere and was estimated from the standard deviation of the wind direction
and from the wind speed and time of day. Stability is expressed as one of 6 classes labeled A through
F, where A is the least stable (greatest turbulent mixing) and F is the most stable (least turbulent mixing).
The conversion from standard deviation of the wind direction, wind speed and time of day to stability is
accomplished as shown in Table 2-4 (taken from EPA, 1987)6.
For fugitive dust impacts, results are generally very insensitive to values used for mixing height
because the emissions are released at or near the ground, and the impacts are generally very close to
the source. As a result the emissions have little opportunity to mix vertically to the height of the mixing
layer. To provide the models with the required values, a mixing height of 500.0 meters was assigned for
all hourly periods. Figure 2 is a frequency distribution of hourly meteorological data used for the model
comparison study.
A(2)-9
-------
Table 2-4. Estimation of Stability Class
Initial Estimate of Stability Class is provided by:
Standard Deviation Stability Class
> 22.5
17.5 - 22.5
12.5 - 17.5
7.5 - 12.5
3.75 - 7.5
< 3.75
Stability Classes are Adjusted by
Time of Dav Initial Stability
time of
Wind
day and
Speed
A
B
C
D
E
F
wind speed
(m/sec)
by:
Final
Stability
Daytime A
B
C
D, E or F
Nighttime A
B
C
D
E
F
3
4
6
4
6
6
2.9
3.6
2.4
3.0
2.4
5.0
3.0
5.0
U <
* U <
* U <
5; U
u <
<. u <
s U
u <
<; U
Any
U <
<: U <
s U
U <
<. U <
<; U
U <
<; U
Any
U <
* U
U <
<. U <
<, U
3
4
6
4
6
6
2
3
2
3
2
5
3
5
.9
.6
.4
.0
.4
.0
.0
.0
A
B
C
D
B
C
D
C
D
D
F
E
D
F
E
D
E
D
D
E
D
F
E
D
A(2)-10
-------
N
> 25 KNOTS
21 - 25
16 - 21
11-15
6-10
1 - 5
Peak Direction = S
Peak Frequency = 17.958
Figure 2. August 3 - December 23, 1985
A(2)-11
-------
(This page intentionally left blank)
A(2)-12
-------
3.0 AIR QUALITY MODELING RESULTS
The FDM and ISCST models were run for the 120 case days identified earlier, and the
concentrations, both PM-10 and TSP, were computed as the sum of the modeled concentration and the
background. There are a number of ways of comparing the model predictions with the measured data.
The methods can generally be separated into two distinct classes of comparison: paired and unpaired
comparisons. In paired data comparisons, a model's prediction at a particular receptor, for a particular
period of time is compared to the measured values at that same location and time. In unpaired
comparisons, a group of model predictions are analyzed to determine their own statistical properties. The
measured data are similarly analyzed, and the results of the statistical evaluations are compared.
In one sense, the model's true evaluation is based on its ability to predict actual measured
concentrations, and only paired data comparisons are indicative of this ability. However, consideration
must be given to how the model is used in a regulatory setting. The models are used to predict
compliance or non-compliance with regulatory criteria. It is not always relevant that they predict the
location or time period of a particular concentration, but rather that they generally be reliable at predicting
the levels at which a project is regulated (e.g. a maximum 24-hour paniculate concentration at any
property boundary location).
For the current analysis, both paired and unpaired data comparisons have been conducted and
will be presented. Actually, three classes of comparisons have been performed:
o first, the data are paired in time and space,
o second, the data are paired in space, but not in time, and
o finally, the data are completely unpaired.
For the paired data, the comparison is presented in two formats. First, the measured versus predicted
values are shown in Tables 3-1 through 3-5 for the TSP comparisons for sites 1 through 5 respectively,
and in Table 3-6 through 3-10 for PM-10 for sites 1 through 5 respectively. Second, a "scatter plot" of the
measured and predicted values for these same four cases are shown in Figures 3-1 through 3-4.
The tables and figures do not show the dramatic illustration of the over-prediction tendency of ISC
as compared with FDM that the first validation study had indicated. However, the same tendencies are
present in the data comparisons. Both the TSP and PM-10 figures show ISC has a greater tendency to
predict above the diagonal line of perfect prediction in the figure, while the FDM predictions are more
centered on the diagonal.
One means of comparison is to determine what number of the data points are within a factor of
A(2)-13
-------
Table 3-1. Model Results Comparison for TSP at Site 1
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Concen.
-------
Table 3-2. Model Results Comparison for TSP at Site 2
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Concen.
(ug/m3>
232.0
243.0
214.0
197.0
840.0
240.0
125.0
39.0
88.0
80.0
185.0
187.0
92.0
570.0
651.0
147.0
127.0
755.0
227.0
113.0
598.0
89.0
37.0
56.0
94.0
367.0
194.0
194.0
178.0
81.0
94.0
207.0
93.0
73.0
63.0
90.0
79.0
131.0
439.0
693.0
473.0
481.0
323.0
182.0
220.0
473.0
193.0
164.0
69.0
36.0
117.0
112.0
251.0
FDM
Predicted
Concen.
(ug/m3)
69.8
400.0
117.6
97.5
487.0
329.2
618.3
123.8
57.7
49.4
145.9
181.8
77.1
308.6
1230.5
45.5
37.7
294.2
134.9
105.7
277.1
210.5
36.3
64.3
61.3
438.1
591.6
513.8
607.5
52.1
69.9
417.9
208.1
71.0
63.0
36.0
56.0
60.3
81.9
294.7
512.4
785.9
416.4
135.4
70.7
481.7
299.4
125.8
65.0
54.8
86.2
41.6
102.8
I SCSI
Predicted
Concen.
(ug/m3)
92.0
423.5
305.2
103.0
360.5
377.3
491.5
165.3
67.7
106.8
280.5
150.3
103.6
476.4
1276.3
59.6
103.8
233.0
95.1
194.1
535.7
258.3
42.4
79.0
61.0
402.9
200.5
775.7
333.0
60.5
77.2
349.7
332.8
71.0
63.0
36.0
56.2
71.3
91.0
330.2
671.8
1270.7
591.4
147.0
72.3
325.2
236.7
147.8
117.5
86.2
95.5
72.9
227.8
Julian
Day
No.
274
275
276
278
284
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Concen.
(ug/m3)
28.0
41.0
210.0
31.0
79.0
75.0
70.0
636.0
165.0
58.0
16.0
21.0
28.0
201.0
35.0
43.0
106.0
71.0
43.0
170.0
71.0
99.0
14.0
11.0
10.0
4.0
16.0
22.0
25.0
21.0
35.0
31.0
35.0
29.0
180.0
64.0
67.0
81.0
44.0
39.0
34.0
33.0
13.0
10.0
23.0
30.0
13.0
138.0
256.0
472.0
534.0
43.0
42.0
34.0
64.0
57.0
35.0
23.0
FDM
Predicted
Concen.
(ug/m3)
78.7
61.6
107.7
59.7
125.7
57.4
401.0
96.4
67.8
81.1
16.0
15.0
31.3
27.8
53.5
49.4
64.8
122.9
35.0
200.4
36.0
111.5
9.1
6.0
13.3
4.2
74.1
233.3
24.6
85.8
61.9
222.0
116.0
705.4
534.6
246.5
44.6
52.3
37.2
46.5
24.4
72.0
11.3
11.8
18.3
34.0
13.6
31.6
35.4
39.2
32.9
89.2
118.0
247.7
295.4
70.7
57.9
49.2
I SCSI
Predicted
Concen.
(ug/m3)
178.7
113.5
262.8
39.6
118.4
60.7
227.2
73.8
100.3
43.5
16.4
15.0
44.3
50.4
88.8
58.1
95.8
134.7
35.0
248.4
36.0
70.6
9.0
6.0
20.0
4.0
138.8
208.2
29.4
102.3
55.4
306.9
75.8
411.6
507.6
176.5
46.1
72.8
52.1
49.1
24.8
104.1
11.9
16.1
19.0
31.6
14.1
31.1
42.8
40.3
38.2
185.8
117.7
256.6
379.8
106.5
89.4
100.4
A(2)-15
-------
Table 3-3. Model Results Comparison for TSP at Site 3
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Cone en.
(ug/ni3)
108.0
578.0
227.0
222.0
105.0
102.0
66.0
416.0
211.0
160.0
687.0
732.0
65.0
110.0
543.0
612.0
445.0
774.0
118.0
88.0
60.0
141.0
462.0
665.0
295.0
633.0
191.0
214.0
456.0
319.0
120.0
155.0
169.0
1344.0
785.0
1014.0
750.0
1036.0
718.0
602.0
1077.0
267.0
183.0
230.0
254.0
245.0
874.0
FDM
Predicted
Cone en.
(ug/m3)
74.9
967.9
590.6
1269.5
567.8
54.7
105.4
294.6
359.1
104.0
459.6
1258.5
169.3
249.7
382.0
492.4
753.2
880.8
432.4
37.9
67.6
83.7
514.5
839.6
1027.0
862.9
295.3
267.6
567.2
435.1
71.7
66.4
146.7
254.3
1529.1
682.4
693.4
482.1
230.1
432.6
541.9
255.2
545.2
190.5
99.2
276.7
942.7
ISCST
Predicted
Cone en.
(ug/m3)
75.7
1185.0
713.3
831.9
191.8
66.0
163.4
542.8
305.4
159.7
737.7
1866.2
226.4
446.9
476.9
734.1
1289.3
2354.6
530.7
46.4
103.3
114.0
726.5
605.3
1432.0
606.5
490.7
433.8
628.2
605.5
72.4
87.6
125.8
391.8
3195.8
1371.7
1577.8
1018.7
262.3
436.9
510.9
373.3
1011.4
310.6
160.8
572.6
2117.0
Julian
Day
No.
274
275
276
278
284
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Concen.
(ug/m3)
84.0
229.0
231.0
241.0
135.0
430.0
310.0
273.0
164.0
20.0
15.0
50.0
93.0
116.0
450.0
138.0
75.0
58.0
36.0
42.0
31.0
6.0
13.0
5.0
38.0
34.0
32.0
51.0
28.0
33.0
81.0
190.0
112.0
62.0
80.0
55.0
37.0
41.0
20.0
31.0
47.0
360.0
673.0
558.0
189.0
46.0
45.0
38.0
109.0
89.0
67.0
22.0
FDM
Predicted
Concen.
(ug/m3)
661.9
243.7
1024.2
111.0
129.9
584.5
384.3
418.6
98.1
23.0
118.6
353.7
492.8
691.7
137.0
66.7
122.2
40.1
50.4
234.1
264.8
528.4
14.0
4.0
644.9
498.6
186.6
84.5
80.6
360.7
274.0
519.0
311.4
74.6
327.1
95.3
75.9
592.9
24.0
39.3
162.2
99.1
156.9
480.8
160.3
345.8
257.7
244.2
398.2
639.2
438.3
201.1
ISCST
Predicted
Concen.
(ug/m3)
640.9
330.4
2400.9
110.6
188.6
621.7
370.6
850.6
103.1
28.8
127.2
518.1
560.8
1188.6
92.3
121.4
152.5
42.3
69.5
129.5
194.5
391.3
32.5
4.0
598.1
580.3
306.5
129.9
133.0
820.2
312.7
877.2
232.3
94.4
356.7
85.6
134.0
559.9
31.3
46.0
169.1
164.2
357.0
1137.2
263.4
791.5
325.2
393.2
796.7
1139.6
726.9
561.6
A(2)-16
-------
Table 3-4. Model Results Comparison for TSP at Site 4
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
Measured
Concen.
(ug/m3)
43.0
50.0
66.0
190.0
66.0
212.0
146.0
128.0
60.0
41.0
38.0
257.0
FDM
Predicted
Concen.
(ug/m3)
55.6
56.0
91.7
258.7
141.0
179.1
170.2
284.9
281.0
42.5
149.0
807.5
I SCSI
Predicted
Concen.
(ug/m3)
44.5
63.0
807.3
1070.7
695.1
770.4
532.5
1432.7
3501.7
43.2
456.2
3003.9
Julian
Day
No.
274
275
276
278
284
285
289
290
291
292
293
294
295
296
297
298
299
Measured
Concen.
(ug/m3)
188.0
63.0
222.0
222.0
193.0
271.0
159.0
176.0
344.0
33.0
37.0
37.0
48.0
51.0
253.0
59.0
FDM
Predicted
Concen.
(ug/m3)
181.2
187.4
140.4
154.4
75.6
285.6
133.1
163.7
281.5
21.8
96.5
134.3
48.4
32.0
303.6
32.9
ISCST
Predicted
Concen.
(ug/m3)
649.7
463.2
342.8
287.5
99.0
439.1
338.3
238.5
432.9
29.3
229.5
423.8
103.4
18.0
362.9
35.8
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
137.0
365.0
128.0
87.0
27.0
160.0
419.0
650.0
438.0
262.0
54.0
177.0
311.0
704.0
604.0
712.0
398.0
204.0
151.0
757.0
617.0
169.0
159.0
56.0
128.0
290.0
700.0
525.0
609.0
836.0
628.0
271.0
407.0
132.0
130.0
125.0
183.0
187.0
123.0
202.0
163.6
246.9
122.8
41.6
30.7
106.8
728.2
683.2
333.3
265.9
139.7
55.5
134.7
115.8
185.7
82.2
132.2
119.3
129.2
222.8
288.1
504.9
78.5
75.2
98.5
129.1
150.1
163.3
135.3
203.5
198.4
110.2
117.4
107.9
179.7
281.5
200.2
261.4
101.1
130.1
289.4
283.0
154.3
41.6
25.0
43.8
4218.7
3737.6
739.6
484.1
179.1
47.2
100.7
124.1
172.4
73.8
83.5
45.0
65.0
162.8
932.6
1126.0
226.3
89.8
228.3
301.9
243.9
251.7
142.1
426.4
378.7
161.7
137.7
235.0
482.2
799.8
327.3
732.0
121.1
163.9
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
64.0
32.0
51.0
17.0
6.0
32.0
16.0
25.0
16.0
33.0
30.0
28.0
29.0
34.0
37.0
150.0
69.0
136.0
84.0
39.0
29.0
35.0
27.0
11.0
18.0
28.0
17.0
77.0
119.0
46.0
151.0
49.0
35.0
145.0
65.0
58.0
219.0
260.7
348.4
108.6
9.7
7.7
149.3
410.0
24.6
16.2
45.5
71.9
140.9
41.7
73.6
45.6
46.7
95.7
69.1
128.7
56.1
51.1
134.0
65.5
11.5
22.1
52.4
22.1
156.8
277.5
45.7
572.2
141.8
45.4
195.1
65.2
64.7
623.9
591.5
666.4
179.6
9.7
11.3
263.6
868.8
26.4
16.5
21.0
38.9
101.5
36.1
26.0
59.8
55.0
142.2
113.8
204.9
62.3
84.9
161.9
108.3
12.3
28.4
79.5
18.6
86.9
207.1
65.2
679.2
65.4
29.0
290.4
69.1
35.2
807.5
A(2)-17
-------
Table 3-5. Model Results Comparison for TSP at Site 5
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Cone en.
(ug/m3)
33.0
43.0
82.0
76.0
62.0
86.0
54.0
31.0
44.0
52.0
63.0
77.0
74.0
74.0
48.0
27.0
38.0
43.0
109.0
80.0
59.0
34.0
40.0
69.0
61.0
68.0
62.0
55.0
55.0
65.0
76.0
87.0
71.0
64.0
36.0
64.0
56.0
77.0
64.0
60.0
93.0
94.0
92.0
69.0
77.0
27.0
104.0
31.0
46.0
32.0
39.0
42.0
FDM
Predicted
Cone en.
(ug/m3)
33.0
43.0
38.5
76.5
72.6
92.6
54.0
46.9
44.8
41.3
79.8
94.6
102.1
79.6
41.0
25.0
25.0
43.0
70.1
80.0
67.1
41.4
36.0
61.0
54.0
66.0
48.0
50.0
45.0
65.0
79.5
106.0
102.1
90.4
84.3
57.8
62.8
82.4
64.0
67.3
101.5
96.5
82.1
77.1
77.0
29.7
87.1
35.1
118.3
32.0
32.8
42.0
I SCSI
Predicted
Concen.
(ug/rri3)
33.0
43.0
38.7
78.1
94.1
119.3
54.0
99.7
56.4
49.2
102.0
131.0
174.8
99.4
41.0
25.0
25.0
43.0
70.2
80.0
82.4
62.7
38.0
61.0
54.0
66.0
48.0
50.0
45.0
65.0
83.4
134.7
146.5
142.5
195.4
61.8
82.2
98.0
64.0
99.1
131.2
100.0
82.0
109.3
77.0
32.0
143.4
43.6
327.9
32.0
32.5
42.0
Julian
Day
No.
274
275
276
278
284
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Concen.
(ug/m3)
17.0
48.0
49.0
26.0
73.0
63.0
42.0
43.0
30.0
19.0
23.0
28.0
21.0
18.0
41.0
30.0
35.0
112.0
29.0
25.0
9.0
7.0
13.0
11.0
17.0
17.0
21.0
21.0
26.0
24.0
27.0
27.0
22.0
44.0
32.0
120.0
41.0
24.0
25.0
11.0
19.0
32.0
12.0
31.0
30.0
38.0
38.0
26.0
36.0
29.0
48.0
46.0
24.0
FDM
Predicted
Concen.
(ug/m3)
43.1
66.4
73.4
26.0
99.8
78.3
51.8
43.0
30.0
18.1
37.8
52.7
27.7
18.0
72.2
26.7
54.9
81.4
47.8
31.6
9.0
6.0
23.3
43.5
16.4
16.0
21.0
25.4
32.5
22.9
26.0
24.0
22.1
57.3
36.2
32.1
43.3
26.5
25.1
18.4
20.4
29.0
12.0
31.0
30.4
39.7
30.0
26.0
36.0
29.0
53.7
44.5
45.7
I SCSI
Predicted
Concen.
(ug/m3)
80.5
130.6
127.3
26.0
146.5
101.5
77.3
43.0
30.0
20.0
88.5
124.0
49.5
18.0
121.9
28.5
85.6
122.6
90.1
35.0
9.0
6.0
39.4
78.7
16.5
16.0
21.0
37.5
46.3
24.5
26.0
24.0
22.0
75.8
43.0
32.1
57.7
29.7
25.0
29.0
25.3
30.3
12.0
31.0
30.5
53.6
30.0
26.0
36.0
29.0
66.3
50.9
110.3
A(2)-18
-------
Table 3-6. Model Results Comparison for PM-10 at Site 1
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Concen.
(ug/m3)
37.0
32.0
42.0
36.0
31.0
32.0
62.0
67.0
52.0
34.0
60.0
70.0
19.0
29.0
23.0
47.0
35.0
67.0
46.0
35.0
21.0
39.0
30.0
39.0
70.0
20.0
29.0
45.0
44.0
46.0
38.0
37.0
37.0
46.0
61.0
56.0
55.0
48.0
37.0
28.0
39.0
77.0
52.0
59.0
60.0
16.0
29.0
20.0
38.0
24.0
31.0
47.0
FDM
Predicted
Concen.
(ug/m3)
36.0
25.0
36.0
34.0
31.0
29.5
67.5
70.6
56.0
34.9
60.0
63.1
39.7
29.0
23.7
41.8
80.3
84.2
87.3
35.0
20.0
27.2
28.0
39.2
60.5
27.1
24.1
45.0
44.0
45.0
38.0
37.0
25.0
46.0
45.1
42.7
76.5
68.4
46.0
32.5
32.2
74.6
62.7
78.2
45.0
24.9
65.8
24.8
89.3
24.0
31.0
40.0
I SCSI
Predicted
Concen.
(ug/m3)
36.0
25.0
36.0
34.0
31.0
29.4
76.2
78.6
54.9
34.9
60.0
67.0
41.4
29.0
25.5
50.9
99.9
97.3
113.4
35.0
20.0
27.4
28.0
44.0
62.3
24.8
22.6
45.0
44.0
45.0
38.0
37.0
25.0
46.0
45.6
45.0
90.6
121.3
71.8
31.9
33.8
95.4
77.7
78.0
45.0
28.4
102.4
30.5
121.0
24.0
31.0
40.0
Julian
Day
No.
274
275
276
278
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Cone en.
(ug/m3)
11.0
102.0
39.0
16.0
38.0
43.0
21.0
41.0
31.0
16.0
11.0
15.0
12.0
16.0
27.0
14.0
36.0
29.0
24.0
13.0
7.0
5.0
13.0
9.0
11.0
16.0
15.0
19.0
22.0
17.0
24.0
18.0
21.0
16.0
33.0
21.0
26.0
30.0
20.0
21.0
12.0
6.0
16.0
25.0
10.0
12.0
17.0
26.0
21.0
28.0
29.0
37.0
36.0
23.0
18.0
FDM
Predicted
Concen.
(ug/m3)
27.5
33.5
37.6
14.0
38.0
45.3
33.8
41.0
24.0
13.5
16.1
25.2
12.9
16.0
120.9
14.0
43.6
43.0
27.7
17.5
6.0
3.0
29.4
39.1
13.4
13.0
13.0
23.6
38.7
19.8
22.0
18.0
21.0
14.7
31.6
28.3
26.6
30.8
23.4
24.0
8.4
7.2
15.2
25.8
10.0
12.0
16.4
24.0
21.0
23.4
26.0
47.7
35.0
21.0
45.5
I SCSI
Predicted
Concen.
(ug/m3)
28.0
36.5
39.9
14.0
38.0
54.5
38.5
41.0
24.0
13.4
16.6
28.9
13.5
16.0
139.7
14.0
41.2
78.3
28.0
20.5
6.0
3.0
46.9
51.5
15.2
13.0
13.0
21.6
36.3
19.8
22.0
18.0
21.0
15.3
36.7
31.6
27.0
31.2
27.6
28.5
10.3
7.6
15.2
26.1
10.0
12.0
16.5
24.0
21.0
23.2
26.0
48.3
35.0
21.0
55.8
A(2)-19
-------
Table 3-7. Model Results Comparison for PM-10 at Site 2
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Concen.
(ug/m3)
92.0
83.0
77.0
105.0
287.0
87.0
58.0
19.0
33.0
81.0
58.0
53.0
179.0
228.0
48.0
35.0
179.0
92.0
35.0
223.0
27.0
19.0
20.0
52.0
70.0
77.0
39.0
66.0
25.0
53.0
48.0
68.0
16.0
28.0
26.0
36.0
49.0
193.0
173.0
140.0
140.0
76.0
98.0
137.0
74.0
59.0
33.0
20.0
58.0
51.0
108.0
FDM
Predicted
Concen.
(ug/m3)
45.7
168.1
67.0
66.7
221.4
164.0
289.1
65.7
26.7
66.8
78.0
33.0
138.1
584.6
38.1
29.6
136.5
69.9
59.3
157.0
82.4
19.4
32.8
45.2
204.2
244.2
211.7
242.5
28.8
48.8
177.8
93.0
16.0
28.0
26.0
26.0
33.2
59.1
253.4
410.9
227.7
83.9
51.9
204.3
133.8
57.7
40.7
31.8
49.5
38.1
75.7
I SCSI
Predicted
Concen.
(ug/m3)
44.7
189.7
143.2
66.0
155.1
183.8
227.2
78.2
50.0
122.7
62.0
44.7
205.5
561.8
39.1
38.8
113.6
48.5
90.8
258.5
110.0
20.2
36.7
45.0
185.1
99.2
336.6
150.5
31.6
51.2
155.3
140.6
16.0
28.0
26.0
26.1
35.2
60.6
302.1
567.7
279.5
83.8
52.4
143.8
97.4
65.4
55.1
42.3
47.7
41.2
105.5
Julian
Day
No.
274
275
276
278
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Concen.
(ug/m3)
11.0
20.0
61.0
16.0
45.0
32.0
157.0
68.0
30.0
16.0
11.0
16.0
30.0
17.0
16.0
48.0
32.0
23.0
46.0
25.0
7.0
3.0
6.0
4.0
11.0
14.0
17.0
18.0
16.0
19.0
22.0
56.0
23.0
22.0
33.0
26.0
35.0
17.0
23.0
7.0
9.0
15.0
28.0
10.0
60.0
50.0
112.0
82.0
32.0
29.0
34.0
35.0
21.0
16.0
FDM
Predicted
Concen.
(ug/m3)
46.3
38.6
77.1
25.7
38.9
156.3
42.3
53.7
43.0
13.0
11.0
17.5
18.3
33.0
20.3
31.9
53.5
18.0
114.2
41.6
6.1
3.0
8.0
4.1
41.0
86.2
14.9
41.0
28.9
101.9
54.0
194.2
97.2
21.4
31.0
29.6
34.3
17.3
43.2
7.1
7.2
15.1
28.1
10.6
12.3
19.2
24.7
22.9
60.1
60.6
104.6
137.4
32.2
30.5
I SCSI
Predicted
Concen.
(ug/m3)
73.7
66.5
131.0
18.5
39.0
97.5
34.0
63.4
29.1
13.1
11.0
20.4
21.0
44.8
22.5
39.2
56.0
18.0
121.6
31.9
6.0
3.0
9.0
4.0
61.8
91.9
16.3
52.3
25.8
136.4
42.5
217.9
78.1
21.9
37.3
34.3
33.6
17.3
53.1
7.4
7.7
15.2
26.3
10.9
12.0
19.8
25.0
24.3
88.1
63.2
121.4
172.1
40.7
45.2
A(2)-20
-------
Table 3-8. Model Results Comparison for PM-10 at Site 3
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Cone en.
58.0
142.0
56.0
147.0
50.0
156.0
104.0
58.0
32.0
41.0
36.0
101.0
112.0
44.0
221.0
202.0
46.0
26.0
168.0
145.0
130.0
177.0
56.0
20.0
39.0
57.0
182.0
159.0
148.0
117.0
50.0
65.0
164.0
71.0
38.0
38.0
36.0
72.0
310.0
246.0
195.0
245.0
227.0
86.0
47.0
58.0
84.0
99.0
239.0
FDM
Predicted
Cone en.
(ug/m3)
41.5
368.8
183.7
263.0
52.2
419.2
272.9
582.7
270.9
34.7
51.5
158.1
172.1
50.6
209.7
621.5
86.3
111.6
178.1
256.2
366.0
423.3
159.9
20.5
35.3
56.9
230.2
335.4
393.6
332.1
125.8
134.0
244.2
195.0
16.5
35.7
33.5
74.8
348.3
272.6
132.9
180.0
253.6
117.0
244.3
126.6
58.0
147.5
364.4
1SCST
Predicted
Cone en.
(ug/m3)
47.9
592.4
101.7
222.0
49.7
500.1
290.8
357.4
81.9
38.0
69.8
242.6
117.7
57.5
310.2
813.5
112.7
199.1
198.9
303.7
529.5
1022.6
218.6
21.3
44.3
65.6
308.6
254.0
598.3
250.5
205.4
191.1
257.7
253.7
16.5
38.4
34.7
54.0
685.4
459.5
130.9
188.3
212.8
157.1
413.5
156.9
69.1
261.5
861.5
Julian
Day
No.
274
275
276
278
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Concen.
(ug/m3)
22.0
60.0
77.0
54.0
111.0
128.0
86.0
58.0
13.0
11.0
19.0
29.0
28.0
146.0
64.0
46.0
19.0
22.0
18.0
9.0
5.0
8.0
5.0
18.0
20.0
19.0
20.0
20.0
20.0
37.0
31.0
54.0
39.0
21.0
32.0
29.0
18.0
25.0
18.0
27.0
21.0
197.0
139.0
103.0
45.0
34.0
31.0
26.0
55.0
49.0
29.0
16.0
FDM
Predicted
Concen.
(ug/m3)
304.3
145.0
502.6
66.8
221.8
170.8
171.2
52.9
16.1
58.9
160.7
172.2
244.6
53.1
33.6
52.8
20.6
29.5
86.9
101.9
229.7
8.6
4.0
238.0
182.8
69.9
39.9
44.3
164.9
109.9
268.2
188.7
120.3
35.1
137.2
54.9
41.5
217.2
17.9
31.6
71.7
35.2
75.3
237.3
86.6
163.3
113.7
102.8
189.3
282.5
159.4
111.0
ISCST
Predicted
Concen.
-------
Table 3-9. Model Results Comparison for PM-10 at Site 4
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Concen.
(ug/m3)
36.0
36.0
35.0
71.0
47.0
103.0
87.0
27.0
26.0
76.0
49.0
48.0
89.0
60.0
41.0
20.0
47.0
99.0
250.0
119.0
84.0
18.0
52.0
72.0
170.0
181.0
225.0
97.0
57.0
59.0
163.0
165.0
42.0
49.0
42.0
26.0
43.0
99.0
152.0
180.0
232.0
124.0
93.0
143.0
29.0
37.0
36.0
29.0
52.0
40.0
63.0
FDM
Predicted
Concen.
(ug/m3)
40.7
40.2
53.0
108.9
89.2
99.0
100.5
137.4
74.6
448.8
192.4
83.0
146.1
84.0
35.4
22.2
65.8
341.0
316.4
194.5
120.4
75.1
29.6
76.6
74.2
98.7
51.9
70.9
56.3
74.7
108.5
125.0
216.6
187.7
49.5
37.6
60.6
96.4
112.1
79.4
131.7
130.4
73.9
62.8
54.8
110.5
178.8
124.1
175.0
63.4
77.5
I SCSI
Predicted
Concen.
(ug/m3)
36.7
43.2
192.8
278.9
180.0
206.2
163.8
777.0
110.7
691.2
152.7
123.8
130.5
85.6
35.3
20.0
34.8
909.7
801.6
195.4
131.3
48.7
23.8
53.8
73.5
82.8
48.7
51.3
25.0
46.0
76.7
252.1
361.2
296.3
106.4
37.6
91.5
157.1
94.8
86.9
144.3
121.9
76.8
58.6
61.1
223.2
203.3
153.0
181.1
51.2
65.7
Julian
Day
No.
274
275
276
278
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
' 312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
Measured
Concen.
(ug/m3)
23.0
26.0
50.0
43.0
89.0
38.0
56.0
55.0
16.0
11.0
16.0
17.0
18.0
59.0
20.0
113.0
22.0
31.0
17.0
24.0
7.0
4.0
12.0
9.0
14.0
16.0
17.0
19.0
13.0
21.0
23.0
23.0
29.0
62.0
24.0
35.0
27.0
31.0
19.0
26.0
12.0
9.0
16.0
28.0
10.0
41.0
26.0
28.0
24.0
50.0
32.0
28.0
38.0
25.0
32.0
FDM
Predicted
Concen.
(ug/m3)
109.0
120.0
96.7
48.2
135.4
80.1
90.0
141.6
15.8
58.3
94.3
31.2
21.2
135.7
18.3
76.9
171.4
155.3
206.8
60.8
6.4
4.2
65.1
154.2
14.6
13.1
22.5
45.6
78.4
28.6
40.9
21.3
29.6
24.6
45.6
39.3
78.4
38.0
30.9
64.9
30.8
6.7
17.1
41.9
14.4
53.5
122.6
30.4
43.6
244.9
74.2
32.8
47.2
32.6
323.5
I SCSI
Predicted
Concen.
(ug/m3)
157.7
205.7
151.3
48.9
138.0
146.0
81.4
108.6
17.7
97.5
197.5
48.0
16.0
145.5
17.3
51.7
281.7
254.1
243.1
78.0
6.2
4.8
94.4
219.9
15.3
13.2
13.0
23.1
45.9
22.6
22.0
22.7
28.4
24.4
61.0
52.2
62.6
35.1
38.9
48.9
40.5
6.8
18.5
36.9
14.3
23.7
53.1
38.0
26.6
209.7
35.2
26.0
46.0
21.1
255.6
A(2)-22
-------
Table 3-10. Model Results Comparison for PM-10 at Site 5
Julian
Day
No.
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
261
262
263
264
265
266
267
268
269
270
271
272
273
Measured
Concen.
(ug/m3)
25.0
34.0
29.0
56.0
33.0
60.0
37.0
19.0
31.0
21.0
29.0
33.0
34.0
51.0
36.0
27.0
28.0
35.0
59.0
41.0
21.0
29.0
49.0
44.0
45.0
38.0
38.0
33.0
46.0
44.0
40.0
25.0
28.0
27.0
30.0
30.0
56.0
51.0
57.0
64.0
59.0
51.0
45.0
12.0
37.0
21.0
23.0
25.0
33.0
40.0
FDM
Predicted
Concen.
(ug/m3)
25.0
34.0
29.7
56.4
41.1
65.4
37.0
32.2
31.9
23.5
39.9
45.0
57.8
55.9
35.0
27.0
28.0
35.0
59.0
32.2
23.4
20.7
45.0
44.0
45.0
38.0
37.0
25.0
46.0
45.8
51.3
34.5
46.0
57.3
27.4
35.5
60.4
58.0
64.7
66.2
59.1
58.7
45.0
13.2
47.7
22.6
73.6
24.0
31.6
40.0
ISCST
Predicted
Concen.
(ug/m3>
25.0
34.0
29.7
56.7
48.2
71.2
37.0
49.9
33.0
24.7
48.1
57.2
90.2
66.8
35.0
27.0
28.0
35.1
59.0
35.8
24.8
21.3
45.0
44.0
45.0
38.0
37.0
25.0
46.0
46.8
60.1
47.9
63.3
93.2
28.5
43.7
66.3
66.4
76.3
66.7
59.0
64.8
45.0
14.0
70.0
24.5
105.1
24.0
31.2
40.0
Julian Measured
Day Concen.
No. (ug/m3)
274
275
276
278
285
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
314
315
316
317
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
11.0
24.0
30.0
14.0
79.0
31.0
21.0
42.0
24.0
17.0
11.0
17.0
13.0
16.0
20.0
19.0
18.0
31.0
17.0
22.0
6.0
5.0
6.0
6.0
11.0
13.0
13.0
18.0
19.0
17.0
24.0
20.0
21.0
14.0
23.0
21.0
26.0
30.0
18.0
20.0
9.0
8.0
17.0
25.0
10.0
14.0
16.0
24.0
21.0
23.0
32.0
28.0
37.0
35.0
22.0
14.0
FDM
cted
icen.
i/m3)
24.1
39.0
48.8
14.0
39.7
40.4
28.0
41.0
24.0
14.2
25.1
34.7
19.1
16.0
35.4
15.1
29.2
42.9
29.2
16.0
6.0
3.0
13.5
23.0
11.2
13.0
13.0
21.7
17.6
17.6
22.0
18.9
21.0
14.0
28.7
23.8
26.1
33.7
18.6
20.0
10.8
7.9
16.4
25.7
10.0
12.0
16.2
25.6
21.0
23.0
29.0
26.0
39.5
36.9
21.0
35.3
ISCST
Predicted
Concen.
(ug/m3)
25.7
63.2
70.4
14.0
40.3
46.7
37.3
41.0
24.0
14.4
41.0
62.4
25.9
16.0
49.2
15.2
38.5
57.5
44.7
17.1
6.0
3.0
17.7
33.0
11.2
13.0
13.0
23.6
20.3
18.1
22.0
18.8
21.0
14.0
33.9
25.9
26.0
35.1
19.4
20.0
13.8
8.2
17.3
25.9
10.0
12.0
16.1
33.7
21.0
23.0
29.0
26.0
43.6
39.2
21.0
59.6
A(2)-23
-------
ISC TSP Concentration (ug/m3) Including Background
3!
O
>
^H
O
oo
"D
FDM TSP Concentration (ug/m3) Includ'ng Background
O
c
o
s:
o
O Q.
to
TJ
-------
10 4ir
10 '-
10'-
10 -
1 1—I I I I I I | 1 1—I I I I I I | 1 1—1—M I M | 1 1 I I I
10 10' 103 10
Measured PM-10 Concentration (ug/m3) Including Background
FIGURE 3-3 FDM EVALUATION FOR PM-10
10J-
10 -
i—[—r FT rrj i 1—i—i—TTTT] 1 1—i i i t r f [ 1 T i—r~n
10 103 103 10
Measured PM-10 Concentration (ug/m3) Including Background
FIGURE 3-4 ISC EVALUATION FOR PM-10
-------
two. For FDM, the TSP predicted results are within a factor of two of the measured results for 72 percent
of the values. For the FDM PM-10 results, the measured and predicted values are within a factor of two
for 73 percent of the values. For the ISCST results, the same comparison shows 63 percent for TSP and
65 percent for PM-10.
EPA has recently been recommending a new method for unpaired data comparisons (Cox et. al.,
1988). It also centers on the concept of accuracy within a factor of two, but utilizes a more complicated
comparison. There are two steps in the evaluation procedure. First, a screening computation is
completed using two quantities, the fractional bias for the average values and a fractional bias for the
standard deviation. They are defined as follows:
FB=-OB.-PR_
(OB + PR)/2
where: FB = fractional bias of the average
OB = average of highest 25 observed values
PR = average of highest 25 predicted values
S>'2
Sp>
where: FO = fractional bias of the standard deviation
SQ = standard deviation of the highest 25 observed
values
S = standard deviation of the highest 25 predicted
values
The screening evaluation is performed by computing both of the above parameters, and plotting
on a special graph. The second level of analysis is more complex. The second level is called the
statistical test and involves using the same fractional bias computation as above, but rather than using
the average and standard deviations of the observed and predicted values, the technique uses a
parameter called the robust estimate of the highest concentration (RHC). In addition, the computation
of the fractional bias is done for several averaging periods and differing meteorological conditions and
the results used to compute a composite performance measure. Finally, a statistical technique called
"bootstrapping" is used where values are extracted at random from the overall data set to create a
"sampled" data set, which is used in the computation of these same performance measures. By
conducting this random sampling many times, the statistician can determine if differences in model
performance are statistically significant. More details on the technique can be found in Cox's paper.
Using the screening technique, the results of the paired in space, but not in time, and the results
A(2)-26
-------
of the completely unpaired data comparisons are depicted in Table 3-11. The values are plotted in
Figures 3-5 and 3-6 for TSP and PM-10 respectively for the completely unpaired data comparisons. The
same screening plots for the paired in space, but not in time comparisons are shown in Figures 3-7
through 3-16. The box at the center of the figure is an indication of the factor of two" performance of the
model. If the data plot within the box, then the model is said to have performed within a factor of two.
In all of the plots, a several consistent patterns are present. These patterns are summarized as
follows:
o In all the screening analyses, the FDM model results plot closer to the center of
the box than the ISCST model results, indicating better performance for FDM than
ISCST.
o For TSP, the characteristic over-prediction of ISC, seen in the first validation study
is clearly present in most of the plots. For TSP, FDM presents a slight tendency
for over-prediction, but is generally close to the center of the box.
o For PM-10, both models exhibit a tendency for over-prediction, although in all
cases, the ISCST over-prediction is greater than the FDM over-prediction.
In the total 12 plots, only two of the FDM plots were outside the factor of two" box, while 8 of the ISCST
plots fell outside the box. Of the two that were outside of the box for FDM, in both cases, the reason for
being outside the box was a large negative value for the fractional bias of the standard deviation - in no
case, did the fractional bias of the average cause the FDM predictions to be outside the factor of two box.
Conversely, five of the twelve ISCST plots had both the fractional bias of the average and the fractional
bias of the standard deviation outside the factor of two box.
The second level of screening was a more complex undertaking. The technique has been
developed primarily for predicting concentrations of sulfur dioxide or other gaseous compounds for which
the data available generally include hourly observations of SO2 concentration and meteorology on a
continuous basis for a year or more. The measurement of paniculate matter usually is done in 24-hour
integrated samples. For the current project, the samples were done on a daily basis, thus they were
virtually continuous, but emission rates and source locations varied on a daily basis, thus modifications
had to be made to the statistical evaluation methods to apply them to the current application. The
modifications to the technique of Cox are summarized as follows:
o Only 24-hour values were available, thus only the only averaging time in the
evaluation was 24-hour. Cox refers to a calculation of a "scientific" evaluation
which uses 1-hour average concentrations. This computation was dispensed
with. Given the single averaging time used here, the composite performance
measure used here was equal to the Absolute Fractional Bias of the RHC values
for the 24-hour samples.
A(2)-27
-------
Table 3-11. Summary of the Screening Analysis Results
TSP FDM ISC
Site 1 Fractional Bias of the Average 0.128 -0.396
Fractional Bias of the Standard Deviation 0.073 -0.624
Site 2 Fractional Bias of the Average -0.101 -0.126
Fractional Bias of the Standard Deviation -0.098 -0.320
Site 3 Fractional Bias of the Average -0.138 -0.601
Fractional Bias of the Standard Deviation -0.130 -0.937
Site 4 Fractional Bias of the Average 0.250 -0.874
Fractional Bias of the Standard Deviation 0.082 -1.412
Site 5 Fractional Bias of the Average -0.059 -0.470
Fractional Bias of the Standard Deviation 0.243 -1.024
All Fractional Bias of the Average -0.090 -0.798
Fractional Bias of the Standard Deviation -0.360 -1.407
PM10
Site 1 Fractional Bias of the Average -0.136 -0.333
Fractional Bias of the Standard Deviation -0.313 -0.624
Site 2 Fractional Bias of the Average -0.430 -0.475
Fractional Bias of the Standard Deviation -0.585 -0.710
Site 3 Fractional Bias of the Average -0.624 -0.952
Fractional Bias of the Standard Deviation -0.735 -1.272
Site 4 Fractional Bias of the Average -0.385 -0.832
Fractional Bias of the Standard Deviation -0.445 -1.229
Site 5 Fractional Bias of the Average -0.090 -0.294
Fractional Bias of the Standard Deviation -0.179 -0.269
All Fractional Bias of the Average -0.561 -0.954
Fractional Bias of the Standard Deviation -0.931 -1.393
A(2)-28
-------
Fractional Bias of the Standard Deviation
Fractional Bias of the Standard Deviation
>
31 i-
id
c
CD
GJ
1
P
1 _
U) -n
do a
32 n
If 1 -
S..Q S
tt
5- m oo-
ri E «
^9. ?•
CD ^-^ « -
si i
Q. 73
CD ^
-D'E.
Q •"*•
o M
^ — h
-1
1 M-
> -• 0 - M h
1 1 1
o
a
£
. i ,
Z! ^
id
c
CD
OJ
1
1 _
T ~! O
T^ (v »
1-1. 1 -
* !Q S
5- m g^o-
rlE «
"| ! -
Q^o
Q. 70
0)
(/) W -•-
D r?
O W
(D
0
> -. O — N
1 1 1
0
3
9
a
o
i 1 <
-------
Fractional Bias of the Standard Deviation
I _
Fractional Bia. of th. Standard Deviation
J. o -»
31 i-
c
a>
GJ
1
CO
00 TI
CO 3
11. 1 -
CD => ai
CL<0 1
5' U1 S. °-
-"§- f
30. J
°o § -
-»» 3 10
o •
CO
-fi
O
-1
Tl
Si
1 Kl—
1 1 1
j *
S
t 1 1
3! M
(O
c
(D
Ol
1
co° i1^
D ~* °
|l I -
fD =J gj
5- m oo-
l| r .
-I. 3 0
0 «
•-•• c
_» **"
O
00
"^ M_
1 1 >
3
i 1 i
®
^
-------
Fractional Bias of the Standard Deviation
i
o
Fractional Bio* of the Standard Deviation
I
1 _]
~n i^
c
CD
OJ
0 ,
I
1 C
Fractional Bias <
Screening
Unpaired i
-1 m •*
*c ^*
3 c" > '
!CD -O • -
? 3
-h O «D
O D ffl
S" "
C?
Ti '
1 1 1
e
JJ
y
e
i i i
i _j
~n M
c
n>
OJ
to
CO -n
CO 3
-5 ~1 O
-f2 a> &
9.S 1 -
-1 — •
CD J CD
a_ia 5'
5'!? »o-
— I — 5-
c? *•
C"' c
"ff
M
0
-1
^
^ 10—
1 1 1
o $
i ^
I 1 I
-------
Fractional Bias of the Standard Deviation
Fractional Bias of the Standard Deviation
h
-q i-
Id
C
CD
OJ
1
!° L_
c to i1
30 a
T3 ZJ Jp
9.8 a -
3 3. „
CL 3 5-
tO 0
=> m ^
-*• o J>
O 3 «
-\
ojjif
o1
T)
i
i 1 i
O
ffl
3 -• N
I 1 i
•> 1
| _j
3! M
c
(A
OJ
1
.-> ,
c to i1
30 a
^3 * o
s.8 | '
S3. „
Q. 3 5'
iQ M
3 m a°
< ^
» S- 1 -
-* 5' S
03 •
^l
— h
O
d ,.
i
i i i
0
(/I
ffi
r
S -• K
1 1 1
-------
us"
OJ
C= C/>
DO
a> — .
CX 3
m
o
• (o
o1
i
M
Fractional Bias of the Standard Deviation
I _
Fractional Bias of the Standard Deviation
3!
ia
c
GJ
I
c: co
CL 3
m
n> o_
0%
zr w
E.
-^ in
o*
-
1
-------
Fractional Bias of the Standard Daviation
Fractional Bias of the Standard Deviation
1 1 II
M -» O -« N fo _. O — K
1
~n M
t5"
C
-1
CD
OJ
p i-
c co i1
3 P *
9.8 1 '
S3. a,
Q. 3 5'
id a
ri" o o—
=> m *-°
i <* $
CD O f
^o^ 1
0 3 -
£'§
01 w"
o1
1 1 1
o
55
w®
9
i i i
i
T| M
ua'
i=
(D
1
COO 31
3 ^ a
9. » i "
(0 2. n,
o D o'
to "
3 m a
_. < 5
3 c >
(0 Q. S -
O" 3 •
^" W
(B c
cn 5T
31
1 1 1
e
o
to
1 1 t
©
3
£
o
-------
o The bootstrapping technique calls for the construction of a number of trial "years"
by sampling the data set. Since sampling a four-month, semi-continuous data
set to create a full year of data, would extend the data beyond its measurement
bounds, the sampling was performed only to create a trial set equivalent in size
to the original data set.
The bootstrapping analysis was completed for both TSP and PM-10 for both models and was
performed for both the paired in space but not in time data sets as well as the completely unpaired data
set. Although not customarily presented in this fashion, the frequency distribution of the Fraction Bias of
the RHC's calculated in the bootstrapping analysis for TSP are shown in Figure 3-17 and 3-18 for TSP
and PM-10 for the unpaired data comparisons. Similar figures for the individual sites in the paired in
space but not in time comparisons are depicted in Figures 3-19 through 3-28. Note that the figure
presents the fractional bias, not the absolute fractional bias. The figures show the same pattern as
observed in the screening evaluation. For all cases, the FDM model predictions are closer to a fractional
bias of zero than the ISCST model predictions. The PM-10 results for FDM show a greater tendency
toward over-prediction than the TSP results, but in all cases the FDM predictions are still more accurate
than the ISC prediction.
One important aspect of the model's performance is that the improvement in prediction of FDM
not be simply an artifact of the particular cases examined, but truly representative of a statistical
superiority. In the figures, this is illustrated by a clear separation in the statistical distributions of the
bootstrapped model results. For most sites this separation is very evident in the figures. For example,
in the totally unpaired data comparisons, the separation between the FDM distribution and the ISCST
distribution is virtually complete (almost no overlap). For a few of the sites, th© separation is not as clear.
This is most evident at site 2, and somewhat evident in all of the PM-10 plots, Further examination of
these results indicates that site 2 is unusual in that a major emitting source (the primary haul road of the
mine) is extremely close to the receptor (within 100 meters). The major advantage to the FDM is the
ability to more accurately represent deposition, but for this very close receptor, little deposition has
occurred, and the improvement to be offered by FDM is not available. The same general conclusion
applies to all the PM-10 results. Although PM-10 will still undergo deposition at a mining operation due
to turbulent processes, the deposition rate is much less than for larger particles. As a result, the FDM
. predictions will be closer to the ISC predictions for PM-10 than for TSP.
The previous discussion would lead to the conclusion that generally, the FDM model performs
acceptably with the data, while the ISCST model does not. In actuality, as earlier figures show, both
models can perform reasonably well for the some of the data points. However, ISCST has the tendency
for large over-prediction on few days. Unfortunately, it is these days which are the focus of the permitting
regulations. Most regulations concern the maximum or second highest concentration, so the ISCST over-
A(2)-36
-------
i?
C
3
CD
- o
3 t-t-
1.1'
3 e.
Q.
CO
3' Q'
CO
32
">
Q CT
13 C
Q. !±.
O
O -H
S 2
p
8
p
8
Frequency Distribution
i-
P
In
I
o
g
a
a
ID
5'
ct
8.°-
?
r .
3
u?
c
-i
(D
OJ
I
32
Q CT
3 C
TJ
Q
i/j
p
8
g
Frequency Distribution
o
CD
' Q'
CO
2
O
~t 2.
p
O Ul
I I I I I I I I
-------
31
UD
OJ
I
NJ
O
1§
-I Q
03
2
~! Ul
«/
1 C/>
• Q"
~a
p
8
8
i
Frequency Distribution
p
o
p
in
p
o
c
S
CxJ
I
-* O
(/J
-
Frequency Distribution
o
p
8
-
-n a
Q |
_. O 3
"g g »
3S- 8'
Q. Q O-
o? r
-------
-D
s:
j^
o
P
8
31
to
c
OJ
I
rO
£- o
3 ii;
9.1' V
39. §
1*2
2 u>
CD rt-
i-
Frequency Distribution
p p
8 3
i i i i i i i i
p
5
6"
i
Frequency Distribution
p p
8 S
i i i i i i i i
-------
m
id
c
CD
GJ
I
rO
-i
Tl
8 8
i i i i t i i
Frequency Distribution
o
S 9- s
o-ro S.°-
§-f r J
Ol
I
p
8
c
ci>
OJ
I
M
_ Q
^- O
I I
n —
5-
w
ul
T3
Frequency Distribution
o
i i l i i
-------
Frequency Distribution
Frequency Distribution
o o o o o oooac
£
i _i
~n N
TI g
if I
n — "
5'f »
n£ 1
3 S o 5 &
i i i j j i i i i i i i i i i i i * i
I
I
l
s ~ ?> 9.
~^c *
J*.
^>-
~=!>=
^^^^^
-------
0.20
0.15 -
g
1
.a
5 0.10
§
I"
0.05 -
0.00
FDM
ISCST
-2
I
-1
0 1
Fractional Bias of the Average
Figure 3—27. Fractional Bias Distributions for TSP
Unpaired in Time for Site 5
0.20
0.15 -
5 0.10
8-
0.05 -
0.00
-2
FDM
ISCST
I
Fractional Bias of th« Average
Figure 3-28. Fractional Bias Distributions for PM-10
Unpaired in Time for Site 5
-------
prediction on these days, causes very misleading results in air quality permitting studies. It tends to occur
under stable, low wind speed conditions.
One of the major advantages of the FDM approach is the avoidance of these large over-
predictions. The improved prediction occurs due to the superior deposition algorithm in the FDM. During
low wind speed stable conditions, the ISCST model allows very high concentrations to be predicted, not
reflecting the deposition which would occur during the long travel times to the receptor. FDM more
accurately represents the behavior of particles in the atmosphere.
A(2)-42
-------
4.0 CONCLUSIONS
The previous analysis has determined that the FDM Model performs generally well in
characterizing paniculate concentrations in the vicinity of the fugitive dust source. The ISCST model also
performed well for the bulk of the samples analyzed, but failed poorly on the high end of the statistical
distribution, leading to large over-predictions of the highest and second highest concentrations, which are
the focus of many air quality regulations for short-term paniculate concentrations. The FDM Model is
judged to be superior in predicting the impacts from fugitive dust sources for the data evaluated in this
study.
A(2)-43
-------
(This page intentionally left blank)
A(2)-44
-------
REFERENCES
1. "User's Guide For The Fugitive Dust Model (FDM) (revised)",EPA-910/9-88-202r.
March, 1989.
2. Cox, W. M. "Protocol for Determining the Best Performing Model", U.S.EPA.
August 1988.
3. "CALINE3 - A Versatile Dispersion Model for Predicting Air Pollution Levels
Near Highways and Arterial Streets", FHWA/CA/TL - 79/23. California Department
of Transportation, November 1979.
3. PEDCo and TRC, 1982, "Characterization of PM-10 and TSP Air Quality Around
Western Surface Coal Mines", for U.S. EPA, Contract No. 68-0203512, June 1982.
5. "Industrial Source Complex (ISC) Dispersion Model User's Guide - Second
Edition (Revised), Vol 1.", EPA-450/4-88-002a. U.S.EPA, December 1987.
6. "On-site Meteorological Program Guidance for Regulatory Applications", EPA-
450/4-87-013. U.S.EPA, June 1987
A(2)-45
-------
Fugitive Dust Model (FDM)
Third Validation Study
Prepared by;
Kirk D. winges
Prepared for;
Region 1Q
U. S. Environmental Protection Agency
1200 Sixth Avenue
Seattle, Washington 9810],
Project Administrator:
Robert B. Wilson
April, 1990
-------
Table of Contents
1.0 Introduction A(3)-1
2.0 Methodology and Model Inputs A(3)-4
3.0 Air Quality Modeling Results A(3)-5
4.0 Conclusions A(3)-13
References A(3)-15
-------
1.0 Introduction
The Fugitive Dust Model (FDM) was developed as an alternative to the previously recommended
Industrial Source Complex Model (ISC) for the purposes of computing fugitive dust impacts. Two previous
validation studies were performed using measured air quality and meteorological data from surface mining
operation. The results of the both validation studies were very encouraging. One major uncertainty in
any air quality investigation centering on fugitive dust is the emission rate. Unlike a stack source,
emissions of fugitive dust cannot be contained and measured. Therefore, the previous two validation
studies relied on published emission factors and mine activity data to estimate emissions. While such
applications represent the real applications of such models and are very relevant in the current regulatory
setting, uncertainty is invariably introduced with the emission factor method of emission information. As
an additional validation effort for FDM, a third validation study was performed using data collected over
a number of years at the Hanford federal reservation in eastern Washington. The Hanford data may well
represent the most comprehensive study of paniculate behavior in the atmosphere in existence.
The purpose of the Hanford experiments was to quantify the dispersion and in particular the
deposition, of emitted paniculate matter. Although applicable to fugitive dust, the Hanford experiments
were not aimed at fugitive dust. A series of experiments were performed over many years where known
release rates of paniculate tracers were emitted from a point source, and concentrations measured at an
extensive array of sampling locations. The paniculate data base has been termed the "Hanford 67" data
base because experiments began at the site in 1967.
TRC has previously conducted a detailed investigation of the Industrial Source Complex (ISC)
Model using the Hanford data base, and the results of that investigation are reported in Cole (1988). We
have not attempted to detail the conduct of that investigation in the current document. Rather, the
interested reader is referred to the Cole document for more details on the adaptation of the Hanford data
base to the ISC model. In the current investigation, the input streams from the Cole runs of ISC were
adapted to make input streams for the FDM model.
Briefly, the Hanford data base consists of two general types of experimental results: single tracer
experiments and dual tracer experiments. Both were examined here, but the dual tracer experiments are
of the greatest interest in the current investigation. In the experiments, release rates of zinc-sulfide or
fluorescein, two paniculate tracers, were fixed and measured. Concentrations of these pollutants were
measured in an array of samplers arranged on arcs centered at the release point. Cross-wind integrated
concentrations formed the primary measured data used in the experiments.
In the single tracer experiments, a single paniculate tracer was released and concentrations
-------
measured and integrated. The two models were used to compute cross wind integrated concentrations
for comparisons with the measured data. Concurrent meteorological data were available for all
experiments. In the dual tracer experiments, zinc sulfide was released along with a known release rate
of a tracer gas, sulf ur-hexafluoride. The ratio of the particulate concentration to the gaseous concentration
(normalized by the respective emission rates), called the concentration ratio, is a direct measure of the
deposition rate. The ratio may be thought of as an indication of the fraction of the emitted particulate
which is still suspended at the measurement distance.
The data from total of 45 separate single tracer experiments were available, while only 6 dual
tracer experiments were available for analysis. In each experiment, concentrations were measured at from
3 to 7 downwind distances. For the purposes here, TRC examined all the data from four standard
downwind distances: 200 meters, 800 meters, 1600 meters and 3200 meters.
A(3)-2
-------
2.0 Methodology and Model Inputs
The emissions source for purposes of the FDM model runs was assumed to be a point source
at the specified release height which was generally 1 or 2 meters above the ground. The model was
modified for the Hanford analysis to compute cross-wind integrated concentrations, rather than the
external method used by Cole to develop cross-wind integrated concentrations for ISC.
Meteorological data were available for the Hanford site. Generally, a very high quality
meteorological data set was available for the Hanford site, including the measurement of roughness
height. Details of the Hanford meteorological and other site data are described in Cole report. For the
dual tracer experiments, deposition velocity and gravitational settling velocity had actually been measured
and reported by the Hanford project team. For these experiments only, the actual measured deposition
velocities were entered to the model, rather than allow the model to compute these values using the
CARB methodology.
A(3)-3
-------
(This page intentionally left blank)
A(3)-4
-------
3.0 AIR QUALITY MODELING RESULTS
The FDM and ISCST models were run for the 45 single tracer experiments and the 6 dual tracer
experiments. Since tracer particles were used, it was not necessary to estimate background
concentrations. There are a number of ways of comparing the model predictions with the measured data.
The methods can generally be separated into two distinct classes of comparison: paired and unpaired
comparisons. In paired data comparisons, a model's prediction at a particular receptor, for a particular
period of time is compared to the measured values at that same location and time. In unpaired
comparisons, a group of model predictions are analyzed to determine their own statistical properties. The
measured data are similarly analyzed, and the results of the statistical evaluations are compared.
In one sense, the model's true evaluation is based on its ability to predict actual measured
concentrations, and only paired data comparisons are indicative of this ability. However, consideration
must be given to how the model is used in a regulatory setting. The models are used to predict
compliance or non-compliance with regulatory criteria. It is not always relevant that they predict the
location or time period of a particular concentration, but rather that they generally be reliable at predicting
the levels at which a project is regulated (e.g. a maximum 24-hour paniculate concentration at any
property boundary location).
For the single tracer experiments, both paired and unpaired data comparisons have been
conducted and will be presented. For the dual tracer experiments, insufficient data are available to
compute the statistical parameters needed for unpaired data comparisons, thus only paired data
comparisons are made here.
Single Tracer Experiments
For the paired data comparisons of the single tracer experiments, the comparison is presented
in two formats. First, the measured versus predicted values are shown in Figures 3-1 and 3-2 for the FDM
and ISC models respectively. These plots are in the form of "scatter plots" of the measured and predicted
concentrations.
The figures do not show the dramatic illustration of the over-prediction tendency of ISC as
compared with FDM that the first and, to a lesser degree, the second validation studies had indicated.
Without the conduct of additional statistical analysis, the two figures appear very similar.
Air quality models are frequently quoted to predict within a factor of two, thus one means of
comparison is to determine what number of the data points are within a factor of two. For FDM, the
predicted results are within a factor of two of the measured results for 51 percent of the values. For the
For the ISCST results, the same comparison shows only 45 percent. Thus although the figures appear
A(3)-5
-------
Predicted Cross-wind Integrated Concentration (uq/m31
^ 2.
ro o
> a o-
(/)
Z
O
Predicted Cross-wind Integrated Concentration (ug/m3)
O
C
-a
c
>
O
O
m
33
3 S =
-------
the same, the paired data comparison actually shows better performance for FDM than ISC.
EPA has recently been recommending a new method for unpaired data comparisons (Cox et. al.,
1988). It also centers on the concept of accuracy within a factor of two, but utilizes a more complicated
comparison. There are two steps in the evaluation procedure. First, a screening computation is
completed using two quantities, the fractional bias for the average values and a fractional bias for the
standard deviation. They are defined as follows:
where:
(OB + PR)/2
FB
OB
PR
S -S
fractional bias of the average
average of highest 25 observed values
average of highest 25 predicted values
(S
where:
FO =
SQ =
S =
fractional bias of the standard deviation
standard deviation of the highest 25 observed
values
standard deviation of the highest 25 predicted
values
The screening evaluation is performed by computing both of the above parameters, and plotting
on a special graph. The second level of analysis is more complex. The second level is called the
statistical test and involves using the same fractional bias computation as above, but rather than using
the average and standard deviations of the observed and predicted values, the technique uses a
parameter called the robust estimate of the highest concentration (RHC). In addition, the computation
of the fractional bias is done for several averaging periods and differing meteorological conditions and
the results used to compute a composite performance measure. Finally, a statistical technique called
"bootstrapping" is used where values are extracted at random from the overall data set to create a
"sampled" data set, which is used in the computation of these same performance measures. By
conducting this random sampling many times, the statistician can determine if differences in model
performance are statistically significant. More details on the technique can be found in Cox's paper.
Using the screening technique, the results of the Hanford single tracer experiments are depicted
in Figure 3-3. The box at the center of the figure is an indication of the "factor of two" performance of the
model. If the data plot within the box, then the model is said to have performed within a factor of two.
The results are consistent with the scatter plot findings, that FDM and ISC are very similar in
A(3)-7
-------
c
o
V-*
D
Q)
Q
T>
L_
D
T3
C
O
-*-•
CO
CD
sz
in
o
QQ
"5
c
.0
-»-•
o
D
1 -
0
-1
-2-
HSCST
-2
-1 0
Fractional Bias of the Average
Figure 3 — 3. Screening Evaluation Results for Hanford
Single Tracer Experiments
-------
performance with this data set. Both models exhibit a bias toward over-prediction with the single tracer
experiments. The FDM performed slightly better than ISCST.
The second level of screening was a more complex undertaking. The technique has been
developed primarily for predicting concentrations of sulfur dioxide or other gaseous compounds for which
the data available generally include hourly observations of SO2 concentration and meteorology on a
continuous basis for a year or more. The measurement of paniculate usually is done in 24-hour samples
which are not continuous. For the current project, the samples were not done on a daily basis, thus
modifications had to be made to the statistical evaluation methods to apply them to the current
application. The modifications to the technique of Cox are summarized as follows:
o The samples varied in analysis time. Cox refers to a calculation of a "scientific"
evaluation which uses 1-hour average concentrations. This computation was
dispensed with. Given the single averaging time used here, the composite
performance measure used here was equal to the Absolute Fractional Bias of the
RHC values for each sample, regardless of sampling time.
o The bootstrapping technique calls for the construction of a number of trial "years"
by sampling the data set. Since sampling a limited, non-continuous data set to
create a full year of data, would extend the data beyond its measurement
bounds, the sampling was performed only to create a trial set equivalent in size
to the original data set.
The bootstrapping analysis was completed for both FDM and ISCST. Although not customarily
presented in this fashion, the frequency distribution of the Fractional Bias of the RHC's calculated in the
bootstrapping analysis for the single tracer experiments are shown in Figure 3-4. Note that the figure
presents the fractional bias, not the absolute fractional bias. The figures show the same pattern as
observed in the screening evaluation. The FDM model predictions are closer to a fractional bias of zero
than the ISCST model predictions.
One important aspect of the model's performance is that the improvement in prediction of FDM
not be simply an artifact of the particular cases examined, but truly representative of a statistical
superiority. In the figures, this would be illustrated by a clear separation in the statistical distributions of
the bootstrapped model results. It is clear that there is a small separation between the statistical
distributions, but considerable overlap exists. The implication is that although the FDM performed better
in the current analysis than the ISCST Model, the difference may not be statistically significant.
Dual Tracer Experiments
The dual tracer experiments differed from the single tracer experiments in that an actual measure
of the deposition function in the model is made. The prediction of gaseous and paniculate concentrations
made by the model (normalized by the emission rate) are ratioed and the compared to similar ratios for
the measured data. Paired data comparisons are shown for both models in Figure 3-5. It is clear from
A(3)-9
-------
0.20
0.15
c
o
_Q
' L_
CO
5
0.10 -
O
c
-------
Predicted Cd/CO
C
•^
CD
OJ
CJi
en
co
c
C/)
O
0)
O
c
Q
Q
O
CD
Z5
C^
^f
CO
CO
p
b
p
b"
p
ro"
p
G-T
O)
Q
CO
2.
O
CL
o P.
O O1
p
bo"
p
CD"
O"
O
p
KJ
D *
(75 g
o ^
00 S
O
O
cn
o
CD
O
CO
O
CO
-------
the figure that FDM outperforms ISCST in this comparison significantly. The ISCST data show virtually
no deposition, while the measured and FDM data clearly shown significant deposition was occurring.
There were too few values here to conduct a significant statistical analysis, but it can safely be concluded
based on Figure 3-5 that FDM performed better in the dual tracer experiments than the ISC model.
A(3)-12
-------
4.0 CONCLUSIONS
Both the FDM and the ISCST Model were evaluated using the Hanford 67 data base. In all cases
the FDM performed better than the ISCST Model. For the dual tracer experimental data, the FDM
performed significantly better than ISCST, however, for the single tracer experiments, it is uncertain
whether FDM's performance was statistically better than ISCST.
The reasons for the apparent discrepancy in performance between these two data set evaluations
are not fully understood. One possible explanation is that for the dual tracer experiments, deposition
velocity and gravitational settling velocity were provided, rather than the internal calculation of these
parameters by the FDM.
The results of the single tracer experiments, however, do not cloud the issue of FDM versus ISCST
performance in the overall data comparisons. Through 3 full validation exercises comparing the FDM and
ISC models, in every case the FDM outperformed the ISCST model. Some of the results are more
statistically significant than others, but the overall pattern is clear, that FDM, with its superior treatment
of deposition, is a better predictor of concentration than ISC for problems involving paniculate deposition.
A(3)-13
-------
(This page intentionally left blank)
A(3)-14
-------
REFERENCES
Cole, C. F. 1988. 'A Performance Evaluation of the EPA's ISC Model", Prepared for the American Mining
Congress, TRC Environmental Consultants, Englewood, Colorado, TRC Project Number 4757-R12,
September.
Cox, W. M. 1988. "Protocol for Determining the Best Performing Model", U. S. Environmental Protection
Agency Report, August.
A(3)-15
-------
APPENDIX B
SAMPLE INPUT AND OUTPUT STREAMS
-------
Sample Input Stream
B-l
-------
(This page intentionally left blank)
B-2
-------
TEST CASE
11211111
56
60.
1.25
0.0262
525.
255.
389.
443.
596.
622.
824.
1344
1554
61.
10.
10.
10.
10.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
30.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
.
10 5
1.
3.75
0.0678
1300.
1530.
1294.
1136.
388.
1103.
3191.
1939.
183.
2365.
535500E-02
680000E-01
119000E+00
102000E-01
330014E-04
337000E-08
330014E-04
337000E-08
288462E-06
337000E-08
201972E-05
337000E-08
124000E-03
337000E-08
413772E-05
337000E-08
413772E-05
337000E-08
413772E-05
337000E-08
413772E-05
337000E-08
413772E-05
337000E-08
413772E-05
337000E-08
622494E-03
622494E-03
124499E-02
674828E-03
192808E-04
192808E-04
207957E-03
110176E-03
655547E-03
655547E-03
121
0
1
1
1
0
3
0
3
0
3
0
3
0
3
0
3
0
3
0
3
0
3
0
3
0
3
0
0
0
0
0
0
0
0
0
0
24
1.
7.
0.
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
5
1704
610
610
549
488
1055
1055
1203
1203
237
237
513
513
683
683
318
318
549
549
195
195
402
402
299
299
610
610
999
1146
999
816
755
694
621
658
755
719
2.
12
0.
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.7
.0
.7
.9
.9
.9
.8
.4
.9
.3
5
.5
1536
1363.
1363.
1426.
1451.
1512.
1512.
1256.
1256.
1789.
1789.
1780.
1780.
1353.
1353.
1158.
1158.
795.
795.
1075.
1075.
597.
597.
634.
634.
207.
207.
1658.
1475.
1475.
1341.
1414.
1426.
1365.
1304.
1414.
1511.
2
C
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
2
2
1
3
5
5
5
3
8
20.
0.5820
0.0
0.0
0.0
0.0
354.0
354.0
149.0
149.0
208.0
208.0
207.0
207.0
84.0
84.0
87.0
87.0
149.0
149.0
195.0
195.0
85.0
85.0
85.0
85.0
140.0
140.0
999.7
999.7
816.9
755.9
694.9
621.8
658.4
816.9
719.3
0.0
0.0
0.0
0.0
305.0
305.0
217.0
217.0
140.0
140.0
219.0
219.0
50.0
50.0
82.0
82.0
122.0
122.0
104.0
104.0
85.0
85.0
195.0
195.0
146.0
146.0
1475.2
1475.2
1341.1
1414.3
719.3
1426
1365,
1304,
1341
1511.8
1670.3
.5
.5
.5
.1
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
B-3
-------
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
.655547E-03
.655547E-03
.550880E-04
.275440E-04
.275440E-04
.200152E-03
.200152E-03
.200152E-03
.454476E-03
.454476E-03
.454476E-03
.134000E-03
.454476E-03
.227238E-03
.227238E-03
.227238E-03
.227238E-03
.227238E-03
.227238E-03
.227238E-03
1.64
1.84
1.00
2.23
1.07
1.00
1.00
1.00
1.19
1.00
1.35
1.87
1.96
2.63
3.02
3.03
2.01
2.54
1.96
1.76
1.01
1.00
1.19
1.03
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
215.5
189.1
166.5
110.5
97.1
78.1
270.9
273.7
46.7
85.1
185.1
188.8
194.5
211.6
195.2
173.2
213.2
197.7
179.4
217.7
125.4
50.6
53.6
37.1
719
634
560
353
353
585
609
621
816
560
438
402
402
512
512
512
463
85
61
207
6
5
6
5
6
6
1
1
4
1
2
2
1
1
2
2
1
1
1
1
6
6
4
5
.3
.0
.8
.6
.6
.2
.6
.8
.9
.8
.9
.3
.3
.1
.1
.1
.3
.3
.0
.3
1670
1914
1962
1828
1828
1950
1938
1926
1341
1170
1146
1109
1109
841
841
841
804
1036
963
755
10000
10000
10000
10000
10000
10000
1400
1400
400
1400
1200
1200
1400
1400
1200
1200
1400
1400
1400
1400
10000
10000
400
10000
.3
.1
.9
.8
.8
.7
.5
.3
.1
.4
.0
.5
.5
.2
.2
.2
.7
.3
.2
.9
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
634
560
353
182
195
414
487
560
560
438
402
317
512
573
256
463
85
61
597
426
286
285
284
284
283
283
283
283
284
286
288
289
291
292
292
293
294
293
291
290
289
288
288
287
.0
.8
.6
.9
.1
.5
.7
.8
.8
.9
.3
.0
.1
.0
.0
.3
.3
.0
.4
.7
.0
.3
.7
.0
.4
.4
.0
.0
.5
.6
.0
.6
.2
.1
.8
.3
.3
.4
.7
.0
.2
.4
.0
.4
1914
1962
1828
1828
1731
1731
1743
1743
1170
1146
1109
1158
841
792
1036
804
1036
963
182
573
.1
.9
.8
.8
.3
.3
.5
.5
.4
.0
.5
.2
.2
.5
.3
.7
.3
.2
.9
.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
30.0
B-4
-------
Sample Output Stream
B-5
-------
(This page intentionally left blank)
B-6
-------
FUGITIVE DUST MODEL (FDH)
VERSION 90121
MAY, 1990
RUN TITLE:
TEST CASE
INPUT FILE NAME: test.IN
OUTPUT FILE NAME: test.OUT
PLOT OUTPUT WRITTEN TO FILE NAME:
test.DAT
CONVERGENCE OPTION 1=OFF, 2=ON
MET OPTION SWITCH, 1=CARDS, 2=PREPROCESSED
PLOT FILE OUTPUT, 1=NO, 2=YES
MET DATA PRINT SWITCH, 1=NO, 2=YES
POST-PROCESSOR OUTPUT, 1=NO, 2=YES
DEP. VEL./GRAV. SETL. VEL., 1=DEFAULT, 2=USER
PRINT 1-HOUR AVERAGE CONCEN, 1=NO, 2=YES
PRINT 3-HOUR AVERAGE CONCEN, 1=NO, 2=YES
PRINT 8-HOUR AVERAGE CONCEN, 1=NO, 2=YES
PRINT 24-HOUR AVERAGE CONCEN, 1=NO, 2=YES
PRINT LONG-TERM AVERAGE CONCEN, 1=NO, 2=YES
NUMBER OF SOURCES PROCESSED
NUMBER OF RECEPTORS PROCESSED
NUMBER OF PARTICLE SIZE CLASSES
NUMBER OF HOURS OF MET DATA PROCESSED
LENGTH IN MINUTES OF 1-HOUR OF MET DATA
ROUGHNESS LENGTH IN CM
SCALING FACTOR FOR SOURCE AND RECPTORS
PARTICLE DENSITY IN G/CM**3
1
1
2
1
1
1
1
1
1
2
1
56
10
5
24
60.
1.00
.0000
2.50
GENERAL PARTICLE SIZE CLASS INFORMATION
PARTICLE
SIZE
CLASS
1
2
3
4
5
CHAR.
DIA.
(UM)
1.2500000
3.7500000
7.5000000
12.5000000
20.0000000
GRAV.
SETTLING
VELOCITY
(M/SEC)
**
**
**
**
**
DEPOSITION
VELOCITY
(M/SEC)
**
**
**
**
**
FRACTION
IN EACH
SIZE
CLASS
0.0262
0.0678
0.1704
0.1536
0.5820
** COMPUTED BY FDM
B-7
-------
RECEPTOR COORDINATES (X,Y,Z)
525., 1300., 0.) < 255., 1530., 0.) ( 389., 1294., 0.)
443., 1136., 0.) ( 596., 388.. 0.) ( 622., 1103., 0.)
824., 3191., 0.) ( 1344,, 1939., 0.) ( 1554., 183., 0.)
61., 2365., 0.) (
B-8
-------
SOURCE INFORMATION
TYPE
1
1
1
1
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
ENTERED EMIS. TOTAL
RATE (G/SEC, EMISSION
G/SEC/M OR RATE
G/SEC/M**2) (G/SEC)
0.005355000
0.068000000
0.119000000
0.010200000
0.000033001
0.000000003
0.000033001
0.000000003
0.000000288
0.000000003
0.000002020
0.000000003
0.000124000
0.000000003
0.000004138
0.000000003
0.000004138
0.000000003
0.000004138
0.000000003
0.000004138
0.000000003
0.000004138
0.000000003
0.000004138
0.000000003
0.000622494
0.000622494
0.001244990
0.000674828
0.000019281
0.000019281
0.000207957
0.000110176
0.000655547
0.000655547
0.000655547
0.000655547
0.000055088
0.000027544
0.000027544
0.000200152
0.000200152
0.000200152
0.000454476
0.000454476
0.000454476
0.000134000
0.000454476
0.000227238
0.000227238
0.000227238
0.000227238
0.000227238
0.000227238
0.000227238
0.00536
0.06800
0.11900
0.01020
3.56316
0.00036
1.06703
0.00011
0.00840
0.00010
0.09156
0.00015
0.52080
0.00001
0.02952
0.00002
0.07522
0.00006
0.08391
0.00007
0.02990
0.00002
0.06858
0.00006
0.08458
0.00007
0.11385
0.09107
0.28226
0.06430
0.00120
0.00184
0.01479
0.01792
0.06827
0.10390
0.16932
0.05767
0.01360
0.00470
0.00513
0.05564
0.04603
0.03857
0.13988
0.05650
0.02349
0.01316
0.13175
0.01772
0.07316
0.01385
0.10074
0.01750
0.21517
0.06491
WIND
SPEED
FAC.
0.000
1.000
1.000
1.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
3.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
X1
(M)
610.
610.
549.
488.
1055.
1055.
1203.
1203.
237.
237.
513.
513.
683.
683.
318.
318.
549.
549.
195.
195.
402.
402.
299.
299.
610.
610.
1000.
1146.
1000.
817.
756.
695.
622.
658.
756.
719.
719.
634.
561.
354.
354.
585.
610.
622.
817.
561.
439.
402.
402.
512.
512.
512.
463.
85.
61.
207.
Y1
(M)
1363.
1363.
1426.
1451.
1512.
1512.
1256.
1256.
1789.
1789.
1780.
1780.
1353.
1353.
1158.
1158.
795.
795.
1075.
1075.
597.
597.
634.
634.
207.
207.
1658.
1475.
1475.
1341.
1414.
1427.
1366.
1305.
1414.
1512.
1670.
1914.
1963.
1829.
1829.
1951.
1939.
1926.
1341.
1170.
1146.
1110.
1110.
841.
841.
841.
805.
1036.
963.
756.
X2
(M)
0.
0.
0.
0.
354.
354.
149.
149.
208.
208.
207.
207.
84.
84.
87.
87.
149.
149.
195.
195.
85.
85.
85.
85.
140.
140.
1000.
1000.
817.
756.
695.
622.
658.
817.
719.
719.
634.
561.
354.
183.
195.
415.
488.
561.
561.
439.
402.
317.
512.
573.
256.
463.
85.
61.
597.
427.
Y2 HEIGHT
(M) (M)
0.
0.
0.
0.
305.
305.
217.
217.
140.
140.
219.
219.
50.
50.
82.
82.
122.
122.
104.
104.
85.
85.
195.
195.
146.
146.
1475.
1475.
1341.
1414.
1427.
1366.
1305.
1341.
1512.
1670.
1914.
1963.
1829.
1829.
1731.
1731.
1744.
1744.
1170.
1146.
1110.
1158.
841.
793.
1036.
805.
1036.
963.
183.
573.
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
WIDTH
(M)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
30.00
TOTAL EMISSIONS 7.84414
NOTE: SOME SOURCE EMISSION RATES ARE A FUNCTION OF UIND SPEED AND TOTAL IS NOT CORRECT
B-9
-------
24 HOUR AVERAGE FOR HOUR ENDING 24
CONCENTRATIONS IN MICROGRAHS/M**3
< 525., 1300.,
( 443., 1136.,
( 824., 3191.,
( 61., 2365.,
122.035)
300.847)
5.018)
3.985)
( 255.,
< 596.,
( 1344.,
(
1530.,
388.,
1939.,
36.700)
12.260)
27.336)
( 389.,
< 622.,
( 1554.,
1294.,
1103.,
183.,
79.491)
51.508)
0.042)
B-10
-------
24 HOUR AVERAGE FOR HOUR ENDING 24
DEPOSITION RATE IN MICROGRAMS/M**2/SEC
( 525., 1300., 1.242) ( 255., 1530., 0.401) < 389., 1294., 0.688)
( 443., 1136., 5.512) ( 596., 388., 0.181) ( 622.. 1103., 0.482)
( 824., 3191., 0.051) ( 1344., 1939., 0.383) ( 1554., 183., 0.001)
( 61., 2365., 0.020) (
B-ll
-------
APPENDIX C
RELEVANT SECTIONS FROM THE CALINE3 USER'S GUIDE
-------
5. MODEL DESCRIPTION
5.1 Gaussian Element Formulation
CALINE3 divides individual highway links into a series of
elements from which incremental concentrations are computed
and then summed to form a total concentration estimate for
a particular receptor location (see Fig. 1). The receptor
distance is measured along a perpendicular from the receptor
to the highway centerline. The first element is formed at
this point as a square with sides equal to the highway width.
The lengths of subsequent elements are described by the fol-
lowing formula:
EL =
Where, EL = Element Length
W = Highway Width
NE = Element Number
BASE = Element Growth Factor
PHK200, BASE=1.1
20°
-------
PHI
WIND
DIRECTION
EL = W* BASE
(NE-1)
ELEMENTS
, RECEPTOR
W = LINK WIDTH
W2 = LINK HALF WIDTH
D = RECEPTOR DISTANCE
ELEMENT NUMBER
ELEMENT LENGTH
ELEMENT CENTERLINE
DISTANCE
BASE = ELEMENT GROWTH FACTOR
( FUNCTION OF WIND ANGLE )
PHI = ROADWAY-WIND ANGLE
NE
EL
ECLD
ELEMENT SERIES USED BY CALINE3
FIGURE 1
8
-------
compromise between accuracy and computational efficiency.
Finer initial element resolution is unwarranted because the
vertical dispersion curves used by CALINE3 have been cali-
brated for the link half-width (W2) distance from the element
centerpoint.
Each element is modeled as an "equivalent" finite line source
(EFLS) positioned normal to the wind direction and centered
at the element midpoint (see Fig. 2). A local x-y coordinate
system aligned with the wind direction and originating at the
element midpoint is defined for each element. The emissions
occurring within an element are assumed to be released along
the EFLS representing the element. The emissions are then
assumed to disperse in a Gaussian manner downwind from the
element. The length and orientation of the EFLS are functions
of the element size and the angle (PHI,4>) between the average
wind direction and highway alignment (see Fig. 3). Values of
PHI=0 or PHI=90 degrees are altered within the program an
insignificant amount to avoid division by zero during the
EFLS trigonometric computations.
In order to distribute emissions in an equitable manner, each
element is divided into five discrete sub-elements represented
by corresponding segments of the EFLS (see Figs. 4 & 5). The
use of five sub-elements yields reasonable continuity to the
discrete element approximation used by the model while not
excessively increasing the computational time. The source
strength for the segmented EFLS is modeled as a step function
whose va.ue depends on the sub-element emissions. The emis-
sion rate/unit area is assumed to be uniform throughout the
element for the purposes of computing this step function.
The size and location of the sub-elements are a function of
element size and wind angle (see Fig. 6).
-------
WIND
DIRECTION
FET = RECEPTOR FETCH
YE = PLUME CENTERLINE
OFFSET
ELEMENT SERIES REPRESENTED BY
SERIES OF EQUIVALENT FINITE LINE SOURCES
FIGURE 2
10
-------
-------
WIND
DIRECTION
^ -»• £5 = SUB-ELEMENTS
EMt EN2 = SUB-ELEMENT WIDTHS
CALINE3 SUB-ELEMENTS
FIGURE 4
12
-------
WIND
DIRECTION
EGMENTED
FINITE LINE
SOURCE
ELL = EQUIVALENT LINE LENGTH
CSL * CENTRAL SUB-ELEMENT LENGTH
CALINE3 FINITE LINE SOURCE
ELEMENT REPRESENTATION
FIGURE 5
13
-------
WIND
DIRECTION
K
-Prf
SUB-ELEMENT CONSTRUCTION FOR VARIOUS
WIND ANGLES
FIGURE 6
14
-------
Downwind concentrations from the element are modeled using
the crosswind finite line source (FLS) Gaussian formulation,
Consider the receptor concentration attributable to an FLS
segment of length dy shown in Figure 7:
dC =
q.dy
exp
2CTy
zov
exp
2(7,
Where,
dC = Incremental Concentration
q = Lineal Source Strength
u = Wind Speed
H = Source Height
a = Horizontal and Vertical Dispersion
Parameters
Since o is constant with respect to y, let:
A = exp
20V
+ exp
f-U+H)*
Integrating over the FLS length yields:
C =
2 rr u (7y 0~z
dy
15
-------
Wind
direction
t t t t J
^Receptor
q = UNIFORM LINE SOURCE STRENGTH
CTy= HORIZONTAL DISPERSION PARAMETER
GENERALIZED FINITE LINE SOURCE (FLS)
FIGURE 7
-------
Note that a and a are functions of x, not y
Substituting p=y/o and dp=dy/a :
y2/
-------
CALINE3 computes receptor concentrations by approximating
the crosswind FLS equation in the following manner (see
Fig. 8):
c =
L
1=1
CNT
SGZ
k=-CNT
-(Z-H+2*k*L) \ . /-(ZiH+2*k»U
expi 5— +exP 2—
2*SGZj / \ 2*SGZj
* PDjj)
Where,
n = Total number of elements
CNT = Number of multiple reflections
required for convergence
U = Wind speed
L = Mixing height (MIXH in coding)
= az as
element
QE. = Central sub-element lineal source
strength for ith element
WT . = Source strength weighting factor for
J jth sub-element (WT, = 0.25,
WT9 = 0.75, ...) '
PDij =
Y., Y.+1 = Offset distances for jth sub-element
SGYi = a as f(x) for ith element
PD.. is calculated by use of a fifth order polynomial
' J
approximation^). Note the addition of multiple reflection
terms represented by non-zero k indices to account for
restricted mixing height (L).
18
-------
-------
The source strength weighting factor (WT.) adjusts the
J
central sub-element lineal source strength measured with
respect to the y-axis (QE) to the mean lineal source
strength for each peripheral sub-element. Because of the
uniform width of the peripheral sub-elements (EH2) and the
assumption of uniform emissions over the element, q = 0 @ y = Y, ,
q=QE/2 @ y = Y2, q = QE @ y = Y3> etc.
Therefore,
WT^QE = WT5*QE = (QE/2+0)/2 = 0.25 QE
WT2*QE = WT4*QE = (QE+QE/2)/2 = 0.75 QE
The element summation of the FLS equation is actually initi-
ated twice for each highway link specified by the user (see
Fig. 9). The computation takes place first in the upwind
direction, ending when the element limits go beyond the up-
wind length (UWL) for the link. The length of the last ele-
ment is modified to conform with the link end point.
The program then proceeds in the downwind direction until the
downwind length (OWL) is exceeded. As soon as a negative
value of fetch (FET) is encountered, the program automatically
concludes the downwind loop computations. If a receptor is
located within an element or downwind from part of an element,
only the upwind portion of the element is used to determine
the source strength.
20
-------
WIND
DIRECTION
UPWIND LOOP
(SIGN = + 1 )
DOWNWIND LOOP
(SIGN =-1)
RECEPTOR
LL
UWL
OWL
SIGN
LINK LENGTH
UPWIND LENGTH
DOWNWIND LENGTH
ECLD SUMMATION FACTOR
CALINE3 LINK-ELEMENT REPRESENTATION
FIGURE 9
21
-------
5.2 Mixing Zone Model
CALINE3 treats the region directly over the highway as a zone
of uniform emissions and turbulence. This is designated as
the mixing zone, and is defined as the region over the traveled
way (traffic lanes - not including shoulders) plus three meters
on either side (see Fig. 10). The additional width accounts
for the initial horizontal dispersion imparted to pollutants
by the vehicle wake effect.
Within the mixing zone, the mechanical turbulence created by
moving vehicles and the thermal turbulence created by hot
vehicle exhaust is assumed to predominate near the ground.
Evidence indicates that this is a valid assumption for all but
the most unstable atmospheric conditions (7_). Since traffic
emissions are released near the ground level and model accuracy
is most important for neutral and stable atmospheric conditions,
it is reasonable to model initial vertical dispersion (SGZ1)
as a function of the turbulence within the mixing zone.
Analyses by Caltrans of the Stanford Research Institute(10)
and General Motors(4_) data bases indicate that SGZ1 is in-
sensitive to changes in traffic volume and speed within the
ranges of 4,000 to 8,000 vehicles/hr and 30 to 60 mph(7.).
This may be due in part to the offsetting effects of traffic
speed and volume. Higher volumes increase thermal turbulence
but reduce traffic speed, thus reducing mechanical turbulence.
For the range of traffic conditions cited, mixing zone
turbulence may be considered a constant. However, pollutant
residence time within the mixing zone, as dictated by the
wind speed, significantly affects the amount of vertical
22
-------
U
MIXING
ZONE
THERMAL
TURBULENCE
MECHANICAL
TURBULENCE
UNIFORM EMISSIONS
-SGZ1=f (TR)
3m
TRAVELED WAY
3m
W2
W2
SGZ1 = INITIAL VERTICAL DISPERSION PARAMETER
TR = MIXING ZONE RESIDENCE TIME
CALINE3 MIXING ZONE
FIGURE 10
23
-------
mixing that takes place within the zone. A distinct linear
relationship between SGZ1 and residence time was exhibited
by the two data bases studied.
CALINE3 arbitrarily defines mixing zone residence time as:
TR = W2/U
Where, W2 = Highway half-width
U = Wind speed
This definition is independent of wind angle and element size.
It essentially provides a way of making the EFLS model com-
patible with the actual two-dimensional emissions release
within an element. For oblique winds and larger elements,
the plume is assumed to be sufficiently dispersed after trav-
eling a distance of W2 such that the mixing zone turbulence
no longer predominates.
The equation used by CALINE3 to relate SGZ1 to TR is:
SGZ1 = 1.8 + 0.11* TR
(m) (sees.)
This was derived from the General Motors Data Base. It is
adjusted in the model for averaging times other than 30
minutes by the following power 1aw(11 ) :
SGZ1ATIM = SGZ13Q* (ATIM/30)0'2
Where, ATIM = Averaging time (minutes)
The value of SGZ1 is considered by CALINE3 to be independent
of surface roughness and atmospheric stability class. The
24
-------
user should note that S G Z1 accounts for all the enhanced dis-
persion over and immediately downwind of the roadway. Thus,
the stability class used to run the model should be repre-
sentative of the upwind or ambient stability without any
additional modifications for traffic turbulence.
5.3 Vertical Dispersion Curves
The vertical dispersion curves used by CALINE3 are formed by
using the value of SGZ1 from the mixing zone model, and the
value of az at 10 kilometers (SZ10) as defined by Pasqui 11 (8.).
In effect, the power curve approximation suggested by
Pasquill is elevated near the highway by the intense mixing
zone turbulence (see Fig. 11). The significance of this
added turbulence to plume growth lessens with increased dis-
tance from the source, though, in theory, it will never
disappear. Extrapolated a curves measured out to distances
of 150 meters from the highway centerline under stable condi-
tions for both the GM and SRI data bases intersect the Pasquill
curves at roughly 10 kilometers. Beyond this point the power
curve approximation to the true Pasquill curve, which is
actually concave to the £nx axis, becomes increasingly in-
accurate. Thus, the model should not be used for distances
greater than 10 kilometers. As will be seen in the sensitivity
analysis, contributions from elements greater than 10 kilometers
from the receptor are insignificant even under the most stable
atmospheric conditions.
For a given set of meteorological conditions, surface roughness
(ZO) and averaging time (ATIM), CALINE3 uses the same vertical
dispersion curve for each element within a highway link. This
is possible since SGZ1 is always defined as occurring at a
25
-------
SZ10
SGZ1
MODIFIED CURVE
1m
W2
10km
ZO = AERODYNAMIC ROUGHNESS
ATIM = AVERAGING TIME
CLAS = STABILITY CLASS
TR = MIXING ZONE RESIDENCE TIME
X = PLUME CENTERLINE AXIS
-------
distance W2 downwind from the element centerpoint. SZ10 is
adjusted for ZO and ATIM by the following power law factors(l\)
SZ10ATIM,ZO = SZ10*(ATIM/3)°'2*(ZO/10)0-07
Where, ATIM = Averaging time (minutes)
ZO = Surface roughness (cm)
Taole 1 contains recommended values of ZO for representative
land use types(12).
The vertical dispersion of CO predicted by the model can be
confined to a shallow mixed layer by means of the conventional
Gaussian multiple reflection formulation^). This capability
was included in the model to allow for analysis of low traffic
flow situations occurring during extended nocturnal low level
inversions. Surprisingly high 8 hour CO averages have been
measured under such condi tions(13).
It is recommended for these cases that reliable, site specific
field measurements be made. The following mixing height model
proposed by Benkley and Schulman(]_4) can then be used:
MTYM 0.185*U*k
riiA" ~ £n(Z/ZO)*f
Where, U = Wind speed (m/s)
Z = Height U measured at (m)
ZO = Surface roughness (m)
k = von Karman constant (0.35)
f = Coriolis parameter
= 1.45 x 10"4 cosG (radians/sec)
0 = 90° - site latitude
27
-------
TABLE 1
Surface Roughness for Various Land Uses
Type of Surface ZU (cm)
Smooth muG flats 0.001
Tarmac (pavement) 0.002
Dry lake bed 0.003
Smooth desert 0.03
Grass (5-6 cm) 0.75
(4 cm) 0.14
Alfalfa (15.2 cm) 2.72
Grass (60-70 cm) 11.4
Wheat (60 cm) 22
Corn (220 cm) 74
Citrus orchard 198
Fir forest 283
City land-use
Single family residential 108
Apartment residential 370
Office 175
Central Business District 321
Park 127
28
-------
For nocturnal conditions with low mixing heights, wind speeds
are likely to be less than 1 M/S. Extremely sensitive wind
speed and direction instrumentation would be required for
reliable results at such low wind speeds. In order to use
CALINE3 for these conditions, measurements of the horizontal
wind angle standard deviation will be needed. The model can
then be modified to calculate horizontal dispersion parameters
based on the methodology developed by Pasqui11(Jj)) or Draxler(16)
The user is cautioned that the model has not been verified for
wind speeds below 1 M/S, and that assumptions of negligible
along-wind dispersion and steady state conditions are open to
question at such low wind speeds.
Mixing height computations must be made for each element-
receptor combination, and thus add appreciably to program run
time. As will be seen in the sensitivity analysis, the mixing
height must be extremely low to generate any significant re-
sponse from the model. Therefore, it is recommended that the
user bypass the mixing height computations for all but special
nocturnal simulations. This is done by assigning a value of
1000 meters or greater to MIXH.
5.4 Horizontal Dispersion Curves
The horizontal dispersion curves used by CALINE3 are identical
to those used by Turner(6) except for averaging time and sur-
face roughness power law adjustments similar to those made for
the vertical dispersion curves (see "ig. 12). The model makes
no corrections to the initial horizontal dispersion near the
roadway. The only roadway related alterations to the horizon-
tal dispersion curves occur indirectly by defining the highway
width as the width of the traveled way plus 3 meters on each
side, and assuming uniform emissions throughout the element.
29
-------
SY10
SY1-
1m
ZO
ATIM
CLAS
X
0V
PY1
/ZO \
ATIM
\ CLAS/
\
zo
) fJATIM
CLAS>
10km
J>
nx
AERODYNAMIC ROUGHNESS
AVERAGING TIME
STABILITY CLASS
PLUME CENTERLINE AXIS
HORIZONTAL DISPERSION PARAMETER
exp (SY1)
HORIZONTAL DISPERSION CURVE-CALINE3
FIGURE 12
30
-------
If field measurements of the horizontal wind angle standard
deviation are available, site specific horizontal dispersion
curves can be generated using the methodology developed by
Pasquill(l_5) or Draxler(]_6). CALINE3 can then be easily re-
programmed to incorporate the modified curves. This approach
is recommended whenever manpower and funding are available
for site monitoring.
5.5 Site Geometry
CALINE3 permits the specification of up to 20 links and 20
receptors within an X-Y plane (not to be confused with the
local x-y coordinate system associated with each element).
A link is defined as a straight segment of roadway having a
constant width, height, traffic volume, and vehicle emission
factor. The location of the link is specified by its end
point coordinates (see Fig. 13). The location of a receptor
is specified in terms of X, Y, Z coordinates. Thus, CALINE3
can be used to model multiple sources and receptors, curved
alignments, or roadway segments with varying emission factors.
The wind angle (BRG) is given in terms of an azimuth bearing
(0 to 360°). If the Y-axis is aligned with due north then
wind angle inputs to the model will follow accepted meteoro-
logical convention (i.e. 90° equivalent to a wind directly
from the eas t).
The program automatically sur.i 1 he contributions from each
link to each receptor. After this has been completed for all
receptors, an ambient or background value (AMB) assigned by
the user is added. Surface roughness is assumed to be rea-
sonably uniform throughout the study area. The meteorological
variables of atmospheric stability, wind speed, and wind
31
-------
ui
^
o
CO
i
O
-------
direction are also taken as constant over the study area. The
user should keep this assumption of horizontal homogeneity in
mind when assigning link lengths. Assigning a 10 kilometer
link over a region with a terrain induced wind shift after the
first 2 kilometers should be avoided. A 2 kilometer link
would be more appropriate.
The elements for each link are constructed as a function of
receptor location as described in Section 5.1 (see Fig. 14).
This scheme assures that the finest element resolution within
a link will occur at the point closest to the receptor. An
imaginary displacement of the receptor in the direction of
the wind is used by CALINE3 to determine whether the receptor
is upwind or downwind from the link (see Fig. 15).
For each highway link specified, CALINE3 requires an input
for highway width (W) and height (H). The width is defined
as the width of the traveled ,way (traffic lanes only) plus
3 meters on each side. This 3 meter allowance accounts for
the wake-induced horizontal plume dispersion behind a moving
vehicle. The height is defined as the vertical distance above
or below the local ground level or datum. CALINE3 should not
be used in areas where the terrain in the vicinity of the
highway is uneven enough to cause major spatial variability
in the meteorology. Also, the model should not be used for
links with values of H greater than 10 meters or less than
-10 meters.
Elevated highway section: nay be of either the fill or bridge
type. For a bridge, air flows above and below the source in
a relatively undisturbed manner. This sort of uniform flow
with respect to height is an assumption of the Gaussian formu-
lation. For bridge sections, H is specified as the height of
33
-------
TO
LINK 3
TO
LINK 2
-LINK 3
RECEPTOR
LINK 1
CALINE3 LINK-ELEMENT ASSIGNMENT
FIGURE 14
34
-------
/$/
// WIND
/^ DIRECTI
DISPLACEMENT
VECTOR
(XPRI, YPRI)
(XR, YR)
RECEPTOR
IF DPRI
-------
the roadway above the surrounding terrain. For fill sections,
however, the model automatically sets H to zero. This assumes
that the air flow streamlines follow the terrain in an undis-
turbed manner. Given a 2:1 fill slope (effectively made more
gradual as the air flow strikes the highway at shallower
horizontal wind angles) and stable atmospheric conditions
(suppressing turbulence induced by surface irregularities),
this is a reasonable assumption to make(17).
For depressed sections greater than 1.5 meters deep, CALINE3
increases the residence time within the mixing zone by the
following empirically derived factor based on Los Angeles
data(_3) :
DSTR = 0.72* ABS(H)0'83
This leads to a higher initial vertical dispersion parameter
(SGZ1) at the edge of the highway. The increased residence
time, characterized in the model as a lower average wind
speed, yields extremely high concentrations within the mixing
zone. The wind speed is linearly adjusted back to the am-
bient value at a distance of 3*H downwind from the edge of
the mixing zone. By this point the effect of the higher value
for SGZ1 dominates, yielding lower concentrations than an
equivalent at-grade section.
For depressed sections, the model is patterned after the
behavior observed at the Los Angeles depressed section site
studied by Cal trans (3.) . Compared to equivalent at-grade
and elevated sites, higher initial vertical dispersion was
occurring simultaneously with higher mixing zone concentra-
tions. It was concluded that channeling and eddying effects
were effectively decreasing the rate of pollutant transport
36
-------
out of the depressed section mixing zone. Lower concentrations
downwind of the highway were attributed to the more extensive
vertical mixing occurring within the mixing zone. Consequently,
the model yields higher values for concentrations within or
close to the mixing zone, and somewhat lower values than would
be obtained for an at-grade section for downwind receptors.
Except for these adjustments, CALINE3 treats depressed sections
computationally the same as at-grade sections.
It has been suggested that the model could be used for evalua-
ting parking lot impacts. If the user wishes to run the model
to simulate dispersion from a parking lot, it is recommended
that SGZ1 be kept constant at 1 meter, and that the mixing
zone width not be increased by 3 meters on each side as in
the normal free flow situation. This is because the slow
moving vehicles within a parking lot will impart much less
initial dispersion to their exhaust gases.
5.6 Deposition and Settling Velocity
Deposition velocity (VD) is a measure of the rate at which a
pollutant can be adsorbed or assimilated by a surface. It
involves a molecular, not turbulent, diffusive process through
the laminar sublayer covering the surface. Settling velocity
(VS) is the rate at which a particle falls with respect to its
immediate surroundings. It is an actual physical velocity of
the particle in the downward direction. For most situations,
a class of particles with an assigned settling velocity will
also be assured the same deposition velocity.
37
-------
CALINE3 contains a method by which predicted concentrations
may be adjusted for pollutant deposition and settling. This
procedure, developed by Ermak(J_8), is fully compatible with
the Gaussian formulation of CALINE3. It allows the model to
include such factors as the settling rate of lead particulates
near roadways (Jj)^) or dust transport from unpaved roads. A
recent review paper by McMahon and Denison(^O) on deposition
parameters provides an excellent reference,
Most studies have indicated that CO deposition is negligible.
In this case, both deposition and settling velocity adjustments
can be easily bypassed in the model by assigning values of 0 to
VD and VS.
38
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1 REPORT NO
EPA-910/9-88-202R
3. RECIPIENT'S ACCESSION NO.
4 TITLE AND SUBTITLE
USER'S GUIDE FOR THE FUGITIVE DUST MODEL (FDM)
(REVISED)
5 REPORT DATE
January 1991
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
KIRK D. WINGES
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
TRC Environmental Consultants, Inc.
21907 64th Avenue W, Suite 230
Mountlake Terrace, WA 98043
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-4399/23
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF- REPORT AND PERIOD COVERED
U.S. Environmental Protection Agency, Region X
1200 Sixth Avenue
Seattle, WA 98101-3188
14. SPONSORING AGENCY CODE
Final Report
on the EPA SCRAM (Support Center
board system. SCRAM is accessible
is.SUPPLEMENTARY NOTES Model codes are also available
for Regulatory Air Models) electronic bulletin
at 1200 or 2400 baud by dialing (919) 541-5742
16. ABSTRACT
This document provides a technical description and user's instructions for the
Fugitive Dust Model (FDM). FDM is a computerized Gaussian-plume air quality
dispersion model, specifically designed for estimation of concentrations and
deposition impacts from fugitive dust sources. The sources may be point, line,
or area sources. The model has not been designed to compute the impacts of
buoyant sources, thus it contains no plume rise algorithm. FDM employs an
advanced gradient-transfer particle deposition algorithm. Gravitational settling
velocity and deposition velocity are calculated by FDM for each of up to 20 user-
specified particle size classes. Descriptions of three performance evaluations
of FDM and the EPA Industrial Source Complex model are included in an appendix.
The user's guide may be ordered with a floppy diskette containing FORTRAN source
codes, PC executable codes, and test data sets.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI F;ield/Group
Air Pollution
Mathematical Models
Computer Dispersion Models
Fugitive Dust
18. DISTRIBUTION STATEMENT
Dispersion
Diffusion
Deposition
19. SECURITY CLASS (This Report)
Unclassified
21 NO. OF PAGES
186
20 SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
------- |