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

-------