EPA/600/A-96/055
March 20,1996	A988
Current Progress In the AERMIC Model Development Program
Alan J. Cimorelli
U. S. Environmental Protection Agency, Region 3
841 Chestnut Street
Philadelphia, PA 19107
Steven G. Peiry1
Atmospheric Sciences Modeling Division
Air Resources Laboratory/National Oceanic and Atmospheric Administration
Research Triangle Park, NC 27711
Russell F. Lee2
Atmospheric Sciences Modeling Division
Air Resources Laboratory/National Oceanic and Atmospheric Administration
Research Triangle Park, NC 27711
Robert J. Paine
ENSR Corporation
35 Nagog Park
Acton, MA 01720
Akula Venkatram
College of Engineering
University of California at Riverside
Riverside, CA 92521
Jeffrey C. Weil
Cooperative Institute for Research in Environmental Sciences
University of Colorado
Boulder, CO 80309
Robert B. Wilson
U. S. Environmental Protection Agency, Region 10
1200 Sixth Avenue
Seattle, WA 98101
'On assignment to the National Exposure Research Laboratory, U. S. Environmental
Protection Agency.
2On assignment to the Office of Air Quality Planning and Standards, U. S, Environmental
Protection Agency.

-------
A988
INTRODUCTION
Background
In 1991, the American Meteorological Society (AMS) and the Environmental Protection Agency (EPA)
initiated a formal collaboration with the designated goal of introducing recent advances in boundary layer
meteorology into regulatory dispersion models. A working group (AMS/EPA Regulatory Model
Improvement Committee, AERMIC) of three AMS and four EPA scientists was formed for this
collaborative effort. AERMIC members are listed as the authors of this paper.
For many years now, we have known that a comprehensive overhaul of EPA's basic regulatory models is
needed (e.g., see Weil1). Responding to this need, AERMIC was formed to update EPA models with
current state-of-the-art Planetary Boundary Layer (PBL) parameterizations. The early efforts of
AERMIC are described by Weil2. As we went through the design process and considered the nature of
present regulatory models, AERMIC's goal became more comprehensive. In addition to improving how
regulatory models characterize the PBL, we decided that other areas such as terrain interactions and
surface releases needed immediate attention. This broadened scope is best expressed in AERMIC's
present objective which is to develop a complete replacement for EPA's Industrial Source Complex
model version 3 (ISC3)? by: 1) adopting ISC3's input/output computer architecture; 2) updating, where
practical, antiquated ISC3 model algorithms with newly developed or current state-of-the-art modeling
techniques; and 3) insuring that all processes presendy modeled by ISC3 will continue to be handled by
the AERMIC Model (AERMOD). A detailed description of the areas, within the ISC3 model, that are
being improved by AERMOD can be found in Perry, et al.4
In developing AERMOD, we have strived to follow certain design criteria to yield a model with desirable
regulatory attributes. We felt that the model should: 1) be robust in estimating regulatory design
concentrations (i.e„ provide reasonable estimates under a wide variety of conditions with minimal
discontinuities); 2) be easily implemented (user friendly, reasonable input requirements and computer
resources), as is the current ISC3 model; 3) be based on state-of-the-art science that captures the
essential physical processes while remaining fundamentally simple; and, 4) accommodate modifications
with ease as the science evolves.
We chose a phased approach in developing AERMOD. Relative to ISC3, AERMOD currently contains
new or improved algorithms for: 1) dispersion in both the convective and stable boundary layers; 2)
plume rise and buoyancy; 3) plume penetration into elevated inversions; 4) treatment of elevated, near-
surface, and surface level sources; 5) computation of vertical profiles of wind, turbulence, and
temperature; and 6) the treatment of receptors on all types of terrain (from the surface up to and above
the plume height). Terrain handling is done with a simple approach while still considering the dividing
streamline concept in stably-stratified conditions. Where appropriate the plume is modeled as either
impacting and or following the terrain. High priority for future efforts include new or improved
algorithms dealing with building down wash and both wet and dry deposition.
The complete AERMOD modeling system consists of two pre-processors and the model itself. The
AERMIC METeorological preprocessor (AERMET) is a stand-alone program which provides
AERMOD with the information it needs to characterize the state of the surface and mixed layer, and the
vertical structure of the PBL. The AERMIC MAPping program (AERMAP) is a stand-alone terrain pre-
2

-------
A988
processor which is used to both characterize terrain and generate receptor grids for AERMOD. Id
addition to the full scale version of AERMOD, we are developing a screening version. When completed,
the screening version will operate as an option within the AERMOD code.
Model Development Process
The process we are following in the development of AERMOD includes, in the following sequence: 1)
initial model formulation; 2) developmental evaluation; 3) peer review and beta toting; 4) revised model
formulation; 5) performance evaluation and sensitivity testing; and 6) submission to EPA's Office of Air
Quality Planning and Standards (OAQPS) for consideration as a regulatory model.
Starting with the ISC25 code, we built the initial formulation of AERMOD by replacing many of ISC's
modules with new or more current formulations. For certain processes (e.g., terrain treatment)
AERMOD was coded with more than one formulation to facilitate testing of various ideas during
development. The initial formulation of AERMOD is summarized in Perry, et al.4 Once formulated, we
then test the model (i.e. the developmental evaluation) against a variety of field measurements in order to
improve and/or replace its algorithms, and provide a basis for selecting formulation options.
We are using five data bases in the developmental evaluation (also referred to as the Phase I Evaluation).
Three of the data bases are event-based tracer releases, while the other two each contain up to a full year
of continuous SO, measurements. The data bases cover both elevated and surface releases, complex and
simple terrain, and both rural and urban boundary layers. We present below a summary of the data bases
and some results from the developmental evaluation. For a detailed description of the developmental
evaluation see Lee, et al.6 To date, this evaluation has resulted in many revisions to AERMOD, which we
discuss below. At the present time, we are nearing completion of the developmental evaluation.
Both peer review and beta testing are included in the model development plan. An internal EPA peer
review of the AERMOD model formulation (AERMIC7) and evaluation is in progress. This will be
followed by an external peer review which will be conducted prior to the performance evaluation. Beta
testers have been selected from among federal, state, and private sector users. In addition, a preliminary
version of the model and its documentation are available to the public in the Sixth Modeling Conference
docket, and through the OAQPS TTN (Office of Air Quality Planning and Standards Technology
Transfer Network) electronic bulletin board system-
Based on the results of the developmental evaluation and comments from peer reviewers, beta testers,
and the Sixth Modeling Conference, we are constructing a final version of AERMOD. This final version
will then be subjected to a comprehensive performance evaluation (also referred to as the Phase E
Evaluation), which is designed to assess how well AERMOD's concentration estimates compare against a
variety of independent data bases.
The major purpose of the performance evaluation is to assess the adequacy of AERMOD for use in
regulatory decision making. As a regulatory model, the operational performance evaluation must be
designed to focus on how well the model predicts concentrations at the high end of the concentration
distribution. The design details of the performance evaluation appear in AERMIC.8 At this time, we
intend to evaluate AERMOD against at least five independent data bases (three in flat terrain and two in
complex terrain), each containing at least one full year of continuous SOz measurements. AERMOD's
3

-------
A988
performance will be compared against appropriate EPA Guideline models following the procedures in
EPA's "Protocol for Determining the Best Performing Model*." Once the performance evaluation is
completed, we intend to submit AERMOD to OAQPS for possible inclusion in the Guideline on Air
Quality Models10. The results of the performance evaluation will be used by OAQPS to decide what
proposal should be made regarding the regulatory status of AERMOD. Should AERMOD replace ISC3?
If not, what role, if any, should AERMOD play?
In addition, a sensitivity analysis is being designed which will help us determine the degree of precision
and accuracy needed for input data; thereby providing a basis for developing guidance for regulatory
implementation. In addition, we will use the results from this analysis to examine the stability of
AERMOD's estimates to small changes in the input data. As stated in the design criteria above, we are
committed to developing a model that produces robust results and minimizes discontinuities.
Basic Model Structure
Design Overview. In this section, we give a very general overview of the most important features of
AERMOD. As a replacement for ISC3, AERMOD will be applicable to rural and urban areas, flat and
complex terrain, surface and elevated releases, and multiple sources (including, point, area and volume
sources). Every effort is being made to avoid model formulation discontinuities wherein large changes in
calculated concentrations can result from insignificant changes in input parameters.
AERMOD is a steady-state plume model. In the Stable Boundary Layer (SBL), the concentration
distribution is assumed to be Gaussian in both the vertical and horizontal. In the Convective Boundary
Layer (CBL), the horizontal distribution is assumed to be Gaussian, but the vertical distribution is
described with a bi-Gaussian probability density function (p.d,f.) to accommodate observed11-12 vertical
concentration distributions that are skewed.
Also, in the CBL, AERMOD is designed to treat the phenomenon of "plume bumping," whereby a
portion of plume mass, released from a buoyant source, hugs the top of the boundary layer before
becoming mixed into the CBL. In addition, AERMOD also tracks any plume mass which penetrates an
elevated stable layer allowing it to re-enter the boundary layer when appropriate.
AERMOD incorporates, with a new simple approach, current concepts about flow and dispersion in
complex terrain. We have designed this approach to be physically realistic and simple to implement while
avoiding the distinction, made by ail other regulatory models, among simple, intermediate and complex
terrain. As a result, AERMOD removes the regulatory need for defining complex terrain regimes; all
terrain is handled in a consistent manner.
One of the major improvements which AERMOD brings is its ability to characterize the PBL through
both surface and mixed layer scaling. AERMOD constructs vertical profiles of required meteorological
variables based on measurements and extrapolations of those measurements using similarity (scaling)
relationships. Vertical profiles of wind speed, wind direction, turbulence temperature, and temperature
gradient are estimated using all available meteorological observations. Although we designed AERMOD
to operate without the need to collect extensive on-site data, all evaluations to date have included a
complete complement of on-site meteorological measurements. Since it has been AERMIC's goal to
4

-------
A988
develop a model which works well using readily available meteorological data (such as National Weather
Service data), we will be testing the model's veracity with degraded (reduced from complete on-site)
meteorological data sets during the performance evaluation.
Unlike existing regulatory models, AJERMOD accounts for the vertical inhomogeneity of the PBL. We
accomplish this by "averaging" the parameters of the actual PBL into "effective" parameters of an
equivalent homogeneous PBL. With these effective parameters, AERMOD accounts for the
inhomogeneity of the PBL, in an averaged sense.
Structure of the Modeling System. As explained above, AERMOD is constructed with one main
program (AERMOD) and two pre-processors (AERMET and AERMAP). The major purpose of
AERMET is to calculate boundary layer parameters for use by AERMOD. The meteorological
INTERFACE, internal to AERMOD, uses these parameters to generate profiles of the needed
meteorological variables. In addition, AERMET passes all meteorological observations to AERMOD.
Surface characteristics in the form of albedo, surface roughness and Bowen ratio, plus standard
meteorological observations, are input to AERMET. AERMET then calculates the PBL parameters:
friction velocity («-), Monin-Obukhov length (£,), convective velocity scale (w.), temperature scale (0.),
CBL height (z,), SBL height (ft), and surface heat flux (H) These parameters are then passed to the
INTERFACE where vertical profiles are calculated, from similarity expressions, for wind speed («), wind
direction, lateral and vertical turbulent fluctuations (<7„ 
-------
A988
In this section, we describe those changes that have occurred since our original discussion of AERMOD
in Percy, et al.4 Except where necessary for clarity, we have not included those portions of the model that
have remained unchanged since the initial formulation. Therefore, in order to obtain a complete
description of the present model this paper should be read in conjunction with Perry, et al.4. It is
important to note that at the time of this writing the technical formulation of AERMOD is still subject to
revision.
General Structure of AERMOD Including Terrain Handling
The general form of AERMOD's concentration equation has not changed since its inception. AERMOD
assumes that the plume dispersion near terrain is characterized by two states. The concentration at a
receptor, located at a position (x,>',z), is the weighted sum of two concentration estimates: one for which
the plume trajectory is horizontal (i.e., the "horizontal plume state" - representing plume material below
the dividing streamline) and the other for which the plume travels over the terrain (i.e., the "terrain
responding state" - representing plume material above the dividing streamline). The relative weighting of
the two terms depends on: 1) the degree of atmospheric stability; 2) the wind speed; and 3)the plume
height relative to terrain. In flat terrain, the concentration equation reduces to the form for a single
plume. The general form for the total concentration at any receptor is:
The two terms in eq. (1) correspond to the contributions from the "horizontal" and "terrain-responding"
plume states. The coefficient/is a weighting factor which relates to the fraction of plume material and z^
is an "effective" receptor height which is defined below.
In AERMOD, Hc does not relate to the geometry of a specific hill. As such, it is conceptually different
from the traditional context in which it is used. AERMOD uniquely defines Hc for each receptor. The
height scale (traditionally the height of the hill being modeled), used for calculating Hc in AERMOD, is
based on the general nature of the terrain within the modeling domain and the location of the specific
receptor for which it is defined. The terrain height scale (hc) is described in detail in Perry, et al.4.
In the initial formulation of AERMOD, we developed two options for defining/and z^in eq. (1). The
two possible formulations for terrain are:
C^x,yX> =/C(x,y^) + (l
(1)
Option 1:
J = 
-------
A988
Option 2:
/ « 0.5-(l + )
and z^= (z~zt)
(3)
where: hp is the plume height; z. is the height of the terrain; and, z is the receptor height.
The CBL's Three Plume Model
To treat the interactions of plume material with the height of the CBL, AERMOD uses a three plume
approach: direct, indirect, and penetrated plumes. The "direct source (at the stack) describes the
dispersion of plume material that reaches the ground directly via downdrafts. The "indirect" source,
located above the CBL, is included to treat the zero-flux condition at z=z, for material that initially rises
to the CBL top in updrafts and is returned to the ground by downdrafts; this material does not have
sufficient buoyancy to penetrate the stable air aloft That is, mass which reaches the height of the CBL,
but does not penetrate the stable layer aloft, is permitted to hug the top of the CBL until the amount of
entrained air is sufficient to allow the plume to mix downward. The "penetrated" source describes the
dispersion of plume material that initially penetrates the elevated stable layer but can re-enter the CBL.
We have made slight changes in the mathematical formulation of the three plume mode since Perry, et al."
(see AERMIC7 for a complete description).
Dispersion
The standard deviations for both the lateral and vertical concentration distributions (<7V and o.
respectively) result from the combined effects of: ambient dispersion (aa); dispersion induced by plume
buoyancy (ab); and, enhancements from building effects (crj. Combining these effects, we produce the
following general expression for oy or oc
Ambient dispersion (ff?a3I) is known to vary significantly with height; having its strongest variation near
the earth's surface. Unlike present regulatory models, we designed AERMOD to account for this height
variation. In our original formulation, both and were taken directly from Taylor's statistical
theory of dispersion14. In the following sections eqs. (5), (8), (11), and (12) each represent the complete
expression, used in our original formulation, to calculate o.jX! and for the CBL and SBL respectively.
Changes to these expressions have been made based on our analysis of the Prairie Grass data. We now
have separate expressions for dispersion from surface and elevated sources. In the following sections we
will describe our present formulation for and first for the CBL and then for the SBL.
(4)
1
-------
A988
Ambient Lateral Dispersion (a,,) in the CBL. In AERMOD, the ambient portion of the lateral
dispersion is composed of an elevated and surface portion. The elevated part (o^ follows directly from
Taylor14 as:
O-xfU
=
i +
0.5*
UT,
Ly
in
(5)
where ctv is the lateral turbulent fluctuation; x is the downwind distance; U is the wind speed; and, TLy is
the lateral Lagrangian time scale.
The surface portion, ays, which we developed empirically from the Prairie Grass data is written as:
o..x/U
a = —
>* t
1 +78
Uz
03
(6)
 0.1 z,.
8
-------
A988
The coefficient ah was developed to insure a smooth transition between the SBL and CBL. That is, we
designed ab such that in the neutral limit (L- ®), the SBL and CBL expressions for a. are equal. The
form of is again from Taylor with the assumption that Tu = <*> during convective conditions. The bt in
eq. (8) results directly from the assumed bi-Gaussian p.d.f. for the vertical CBL.
The expression used for aa (the surface portion) is taken from Venkatram
is

H
for	<; 0.1
Z;
and,
0.t = 0.0
far
H,
Z,
> 0.1
(9)
We determined the values of the coefficient bc and the exponent a empirically from the Prairie Grass data.
In the present version of the model, they are 0.5, and 1.0, respectively.
The total ambient portion of the vertical dispersion can now be written as:
a,
0- x
' U
+ or.
(10)
Ambient Lateral Dispersion (o^) in the SBL. As with the CBL, the ambient portion of the lateral
dispersion in the SBL is composed of an elevated and surface portion. The elevated part (op is given by
the following16:
ayt = MAX
a
0.05; —
U
(11)
The surface portion, a>T, is given by the same empirical expression presented above for the CBL (i.e., eq.
(6)) with Zi (the height of the CBL) replaced by h (the SBL height). As with the CBL, the interpolation
formula, eq. (7), is used to insure a smooth transition between the surface and elevated components of
9
-------
A988
Ambient Vertical Dispersion (a^ in the S6L. The ambient portion of the vertical dispersion is also
composed of an elevated and surface portion. The elevated part (o^) is given by the following:
o„,x/U
0~ = t \
i + o-5*
VTU
m	(12)
where.
T = 	_**	; and (j>. (z/L) = 0.74 + 4.7(z/L)	riT>
u 1.3 A (z/L)aw	h	{li)
The surface portion of the ambient vertical dispersion ou is taken from Venkatram15.
0 =
2
71 U
( JtA ~m
1+0.7-	(14)
As with the CBL, interpolation in the form of eq. (7) is used to weight the surface and elevated portions
of the ambient vertical dispersion (i.e., eqs. (12) and (14)).
Reflection From The Top of the SBL.
The original formulation of AERMOD included reflections from the top of the SBL. The reflecting
surface was set equal to h or the plume height, whichever was larger. Results from the Lovett evaluation
showed that performance improved if these reflections were eliminated. Therefore, we do not include
reflection from the top of the SBL in the present version of AERMOD.
Urban Dispersion
In our original formulation of AERMOD we did not explicitly account for the difference between the
urban and rural boundary layers. However, in order for us to achieve acceptable comparisons with the
Indianapolis data it was necessary to consider these effects. By adding an additional anthropogenic
contribution (50 watts/m2)17 to the surface energy balance, AERMOD interpreted the PBL as convective,
and selected algorithms accordingly.
Inhomogeneity in the Boundary Layer
AERMOD, unlike existing regulatory models, is designed to treat the effects on dispersion from vertical
variations in wind and turbulence. This treatment is primarily needed to properly handle surface releases
and to provide a mechanism by which the penetrated source can re-enter the CBL. The algorithms in
AERMOD function under the assumption that the atmospheric boundary layer is vertically homogeneous
(single values of the meteorological parameters represent the layer). Therefore, we designed a method
to "convert" the inhomogeneous values (as measured or estimated) into equivalent (representative)
10
-------
A988
homogeneous values. AERMOD uses equivalent values for wind speed, and as needed
throughout the computations. We refer to these equivalent values as effective parameters (o^).
In our original approach to inhomogeneity (see Peny, et al.4), we basically interpolated (with downwind
distance) between two observations of plume behavior 1) near the source, plume dispersion is
dominated by meteorological variables near the release point; 2) when the plume later disperses through
the depth of the mixed layer, it is reasonable to assume that plume behavior is governed by
meteorological variables averaged through the layer. The interpolation had an exponential form which
was controlled by travel time and Tu.
This original approach has subsequently been revised since we were unable to adequately describe what
was observed in the Prairie Grass experiment The revised approach, used in the present version of the
model, is described below.
In our current formulation, the effective parameters are determined by averaging their values over that
portion cf the layer between hjx) (plume height) and z, (the height of the receptor above ground) that
contains plume material. The layer through which <% is calculated is controlled by a.(xr) (where xr is the
distance from source to receptor) and is bounded by hp(x) and zr
Since cr.fxj depends on the effective values of cc, u, and Tu the plume size is estimated through a series
of iterations. We use oJhp(x)), u(hp(x)) and T,Jhp(x)) as initial values in the calculation of o.(xr). We
then use o.(xr) to determine the layer over which ajxj^ ufxj^and T^x,)# are calculated. This process
is continued for a number of iterations. The number of iterations depends on convergence and
computational considerations. At the end of the iterative process, we calculate	and
TtJxJtfOver the final layer.
We then calculate a# from the following expression:
h
(15)
where:
hp(x),	if kp(x) < zr
MAX{[hp(x) -2.15az(xr)lzJ, if hp(x) >
(16)
hp(x),	if hp(x) > zr
MIN{[h(x^ +2.15 
-------
A988
PBL Height. Since our original formulation we have made two changes to AERMOD which are
designed to prevent temporal discontinuities in the growth of the PBL. First, at the time of transition
from a stable to a convective boundary layer (i.e., at sunrise), we prevent the PBL from artificially
collapsing by requiring *, to be greater than or equal to the value of h during the last hour of the previous
nocturnal period. Secondly, we avoid sudden (and unrealistic) drops in h for those hours that experience
a large decrease in wind speed by controlling its time evolution as follows7:
dh h
— = —	, where z = (3/i/w	(17)
dt z
This technique performs a temporal smoothing on the original SBL height (h4) as calculated by
Nieuwstadt18.
Turbulence Parameterizations. Based primarily on peer review comments, we have revised the manner
in which turbulence is parameterized. In our original formulation for the SBL, ajh) and ajk), the
turbulence at the top of the SBL, were based on their values at the surface. Since the surface is generally
decoupled from higher layers, we have adopted a formulation based on parameterized turbulent intensity.
This revised approach is presented below.
Turbulence in the SBL. For L > 0, we develop the vertical profile of cr%. by linearly interpolating between
the value of h	(20)
The vertical profile of ow in stable conditions takes on the following new form20:
flo.
aj(z) = MAX ja^ exp
.08z
h
; 0.05; U(h)i.
for all z
(21)
12
-------
A988
where the expression we use for (i.e., o^=-l.7u^) is unchanged; and, i. (the vertical component of
the turbulent intensity) we set equal to 0.016 after Briggs21 (its parameterized value for PGT "F*
stability).
The exponential from allows some turbulence above h due to occasional nocturnal downbursts, but
effectively gives lower vertical turbulence that approaches zero as the height increases above h. The
minimum value of 0.05 m/s is based upon a variety of measurement programs. The minimum formulation
involving the vertical component of the turbulence intensity is consistent with the decoupling of
turbulence in the vertical that is also used in eq. (19).
Turbulence in the CBL. For L<0 and z
2/3
1.7
• + 1.6 •
_z
• wC . for — s 0.1


v Z'-,
zi
and,
al = | 1.7 --j • «.2+ 0.35• w? , for 0.1 s - s 1.0	Q
Eq. (24), which represents a change from our original formulation, was revised to insure continuity
between the surface layer and mixed layer formulations (i.e., eqs. (24) and (25) respectively); whereas,
eq. (25) has remained unchanged. As with aw is linearly interpolated between z, and 1.22;. In our
original formulation aJ1.2Zj)=.01 crfa), whereas in the present version oj 1.2Zi)=U(zJL where L is set
equal to 0.016, its value for PG class "F" after Briggs22.
Temperature Gradient. In unstable conditions (L < 0), the potential temperature gradient (dd/dz) is
13
-------
A988
assumed to be zero for all heights within the convective mixed layer. In our original formulation we set
dd/dz equal to 0.005°K/m for z > Z,. However, based on typical lapse rates found at the top of the
"mixed layer" and above the "interfacial layer," we presently assume that d 6/dz=0.01° KJm for
z,10m. In addition we assume a minimum value for
dd/dz of 0.002°K/m for all z>J0m, Paine24.
Although unreported in Perry, et al.4, AERMOD develops the vertical profile of potential temperature
from its estimate of the temperature gradient
Summary. In the past two years, since AERMOD was initially formulated, we have made significant
revisions to certain areas of the model based primarily on the developmental evaluation. Although the
fundamental structure of the model has not changed, revisions, which were discussed above, have been
made in the areas of: 1) dispersion; 2) turbulence and temperature profiling; 3) treatment of the urban
boundary layer; 4) SBL reflections; 5) growth of the PBL height and, 6) vertical inhomogeneity.
Furthermore, prior to the start of the performance evaluation, we intend to: 1) develop a generalized
treatment for die urban boundary layer, and 2) make all final algorithm selections in areas where options
now exist.
RESULTS OF THE DEVELOPMENTAL EVALUATION
AERMOD contains many algorithms that are new to routine regulatory modeling and although in most
cases these algorithms are based on existing published work, their comparison against field data within
the AERMOD framework must be tested. Since AERMOD is intended to handle pollutants from a wide
variety of source types in a variety of modeling situations, it is important to challenge the model as much
as possible in the development process. Where performance is poor, improved approaches have been
included and tested. While some of the improvements to AERMOD noted previously are the result of
peer review comments or simply further consideration by the AERMIC committee, most improvements
are the result of unacceptable model performance during the developmental evaluation phase of the
project. The results shown here are those provided by the most recent version of AERMOD which
includes the revisions described in this paper.
The Data Bases
Five data bases were selected for the developmental evaluation.
1.The	Prairie Grass data base (Barad25) involves a near-surface, non-buoyant S02 release, in a rural area,
with flat terrain. Surface sampling arrays were positioned in arcs from 50m to 800m downwind of the
source. Both convective and stable conditions are included.
2.	The Kincaid SF6 data base (Liu, et al.26) involves an elevated, buoyant release in a rural area with flat
terrain. Approximately 200 SF6 monitors were placed in arcs from about 500m to 50km downwind of
14
-------
A988
the single 187m stack.
3.	The Indianapolis SF6 data base (Murray, et al.27) involves an elevated, buoyant release in an urban area
with flat terrain. Data were obtained from 177 SF6 monitors in arcs from 250m to 12km downwind of an
84m stack.
4.	The Lovett Power Plant S02 data base (Paumier et al.28) involves an elevated, buoyant release in a
rural area with complex terrain. This one-year data set involves a 145m stack and 12 S02 monitoring
sites on terrain features rising 250 to 330m above the stack base. The monitors are generally 2 to 3km
downwind of the stack.
5.	The Kincaid SO, data base (Liu, et al.26) involves an elevated, buoyant release in a rural area with flat
terrain. There were 30 S02 monitoring stations from about 2km to 20km downwind of the 187m stack.
This data base contains a total of 248 days of valid meteorological observations.
Evaluation Results
As pointed out by Lee et al.6 the purpose of the developmental evaluation is primarily diagnostic, that is
to identify and correct deficiencies of the model during development Highlights of die evaluation with
the above mentioned data bases are presented here using one of the more general tools of our analysis,
quantile-quantile (Q-Q) plots of modeled and measured concentrations (all concentrations have been
normalized by emission rates). Q-Q plots are simple pairings of predicted concentrations, ranked highest
to lowest, with observed concentration, ranked in the same manner. If the ranked distributions are
identical, then all points lie on the x = y line. Q-Q plots are an effective method for comparing the
distributions of two data sets. They are very useful for assessing the performance of regulatory models
since they provide an easy comparison between the high end of the model concentration distribution and
the high end of the observations. To assist us in judging AERMOD's relative performance, we have
inducted comparisons with ISO in the Q-Q plots. The fractional bias, FB, is also used to evaluate
AERMOD. FB provides a quantitative measure with which to compare the models and is defined as:
™mean = 2 (Cj-g/(c; + c;)	(26)
where Cp is the predicted concentration and C0 is the observed concentration with the overbar indicating
(for our analysis) the mean of the top 25% of the data values in each set Note that with this formula,
negative FB indicates underprediction and positive FB indicates overproduction. For example, FB = 0.67
is a factor of two overprediction while FB = 1.0 is a factor of three and FB = 1.33 is a factor of five;
similarly, -0.67 is an underprediction by a factor of two.
At the time of this writing, we have completed our initial analysis with the three tracer studies (Prairie
Grass, Kincaid SF6, and Indianapolis) and the two full year data bases (Lovett and Kincaid SOz).
Analyses with these data bases will continue until we are satisfied with the performance of the model. At
that time we will finalize the model and conduct the performance evaluation. The three tracer-study
comparisons have been reported in some detail by Lee, et al.6 Only a brief overview (focusing on the
results of regulatory interest) will be presented here. In addition, since we have now completed our initial
15
-------
A988
analysis of the two full year data bases, we have included a discussion of the results of these preliminary
comparisons. All of the tracer comparisons have been stratified by stability (stable or convective) and
daxa base; the full-year comparisons include those for one-hour, three-hour, and twenty-four-hour
averages.
Prairie Grass. As reported in Lee, et ai.6 for convective conditions at Prairie Grass, the Q-Q plot of
Figure 1 shows that the distribution of AERMOD predictions compare well against the distribution of
observations, with the highest observations being underpredicted by a small amount. Almost the entire
distribution is within a factor of two for this unpaired comparison. The fractional bias (FB) based on the
mean of the top 25% of the distribution is -0.301 (underprediction within a factor of about 1.35). In
contrast to AERMOD, ISC3 shows a tendency to overpredict by about a factor of two at the high end of
the distribution (Figure 1). The FB for ISC3 is 0.610.
For stable conditions at Prairie Grass, both ISC3 and AERMOD perform well as indicated by the Q-Q
plot of Figure 2. Both models provide an upper end distribution which follows that of the observations;
however, ISC3 has a slight overpredietion tendency (FB = 0.321) while AERMOD is slightly
underpredicting (FB = -0.158). In general we have concluded that the present version of AERMOD,
which includes previously discussed changes, adequately simulates the observations in this rural, flat
terrain, surface release data base. The areas of the model that were improved as a result of our
comparison with the Prairie Grass data included: the treatment of vertical inhomogeneity; and, a specific
approach for surface dispersion which included development of an empirical relationship for ar
Kincaid, SF6. Comparison between AERMOD's and ISC3's performance, during convective conditions,
is shown in Figure 3 for the Kincaid data. This tracer data base is characterized by a buoyant, elevated
gaseous release. The model estimates are compared against surface level peak concentrations. The Q-Q
plot of Figure 3 shows a good match between die distribution of AERMOD estimates and the
observations (at least over the upper portion) with a poorer comparison over the less interesting and less
important lower end the distribution. The dropoff of the distribution at lower concentrations may be
related to the relative uncertainty in the observations for low concentrations. The FB for AERMOD
during convective conditions is -0.028 (essentially unbiased on average over the top 25%). ISC3 shows
similar performance to AERMOD at the upper end of the distribution as seen in Figure 3 (FB of top 25%
= -0.188). However, ISC3 does not match the overall distribution quite as well as AERMOD. Based on
the good comparisons we obtained with the Kincaid SF6 data, we have made no notable changes to our
original CBL formulations.
Traditionally, worst case surface-level impacts from elevated buoyant releases in flat terrain have been
found during convective conditions, where the plume is brought quickly to the ground. Therefore, the
Kincaid study focused primarily on daytime conditions. However, there is a limited number of stable
cases in the data base against which we examined the performance of both AERMOD and ISO. For
these cases both models performed poorly. We attempted to determine the cause of this poor
performance but we have yet to find an adequate explanation. Since the stable data at Kincaid
represented only a very small portion of the distribution of expected stable conditions at this site, we
concluded that it would not be productive to continue the analysis. As a result, we have not used these
comparisons to reformulate the model.
16
-------
A988
Indianapolis. Comparison between AERMOD's and ISO's performance for data at Indianapolis is
shown in Figure 4 for convective conditions and Figure 5 for stable conditions The Q-Q plot for
convective conditions shows a remarkable match between model and measurements. The FB of 0.045
confirms this. The reader should note that we found similar favorable comparisons in convective
conditions, using the original formulation. ISO also performed very well at Indianapolis. For
convective conditions, the FB was found to be 0.076.
The model also performed quite well during stable conditions after we reformulated AERMOD to
account for the anthropogenic effects that an urban boundary has on the surface energy balance. The
Indianapolis area is urban in nature. In stable conditions, where these effects can dominate, AERMOD
estimates concentrations well within a factor of two (Figure 5) of the observations
(FB = -0.145). ISC3 (also within a factor of two with a FB = 0.476) shows a slight tendency for
overprediction. ISO accounts for urban effects by replacing the rural PGT dispersion curves with the
Briggs urban curves.
Analysis of model performance using the data during nighttime "stable" conditions at Indianapolis is one
of the most obvious areas where the developmental evaluation process worked well for AERMIC. At
Indianapolis the original model performed poorly against the surface concentrations during stable
conditions because of its incorrect characterization of the urban boundary layer. Our initial formulation
allowed the boundary layer in Indianapolis to develop a strong stable stratification when, in fact, the
urban nature of the area rarely would allow this condition to arise. Consequently, plume material was
often not being appropriately mixed in the vertical. As indicated by the current results, the modification
(addition of anthropogenic heat) was necessary.
Lovett (Full-Year) SO,. We have completed our initial analysis of AERMOD's performance against the
full-year SO: data base collected at the Lovett Power Plant. This plant is located in the complex
topography of the Hudson River Valley in New York State. Complex terrain effects are handled by this
model in a manner that is totally novel to regulatory models. The two-state model as described above
and in Perry, et al."1 has been coded with two options. The initial performance of each of these options
has been examined with this data base. Preliminary comparisons (for one-, three-, and twenty four-hour
averages) of AERMOD against ISC3 and the Lovett observations are shown in Figures 6,7, and 8.
For the one-, and three-hour averages, both options of AERMOD are performing well (within a factor of
two) in reproducing the distribution of the observations at Lovett. Option 1 has a tendency to
underestimate, while Option 2, showing a similar absolute bias, has a tendency towards overprediction.
ISC3, for the shorter averaging times, overpredicts the observations by a factor of three to four.
We have found very little bias in the comparisons of the observed and predicted 24-hour averages, with
AERMOD's Option 2. However, predictions using AERMOD's terrain Option 1, were found to
underestimate observations, in general, by a factor of two. Furthermore, ISC3 overpredicts the observed
concentrations, for this averaging period, by a factor of two to three.
Although these comparisons are encouraging, we are performing an analysis of the models sensitivity to a
wide variety of source-receptor-terrain relationships in order to build additional confidence in these
methods. If similar good performance is found for both terrain options, we will examine each option in
17
-------
March 20. 1996
A988
the Phase II performance evaluation.
Kincaid, S02. As with Lovett, the Kincaid SO; data provides an extensive temporal record which
allows us to evaluate predictions of the longer term averages. Our initial analysis of the one-, three-, and
twenty four-hour average concentrations have been completed and the results are presented in Figures 9.
10, and 11. For the one-, and three-hour comparisons AERMOD reproduces the observed distribution
very well. For the same averaging times. ISC3 is showing a tendency towards underprediction although
still generally within a factor of two.
For the twenty four-hour averages, both models provide noticeable underpredictions. With the exception
of the highest few points, which are close to the one-to-one line. AERMOD "s predictions fall around a
factor of two below the observations. Whereas. ISC3's predictions fall generally around a factor of four
below the observations.
Summary In summary. AERMOD has gone through considerable modification as a result of the
developmental evaluation process and will likely continue to do so over the remaining few months of
this phase of the project. For the surface release data (Prairie Grass) AERMOD shows a slight
underprediction tendency in both connective and stable conditions while ISC3 has a small
overprediction tendency. With the rural elevated release tracer data (Kincaid SF6) AERMOD exhibits
insignificant bias in convective conditions while performing poorly against the few available stable
cases. ISC3 performs similarly to AERMOD with this tracer data base. With the urban data, both
models perform well with only small biases for all conditions.
For the full year data bases. AERMOD compared well with the observations, particularly for the shorter
averaging times.
DISCLAIMER
This paper has been reviewed in accordance with the U. S. Environmental Protection Agency's peer and
administrative review policies and approved for presentation and publication. Mention of trade names
or commercial products does not constitute endorsement or recommendation for use.
18
-------
-BLANK PAGE-
19
-------
A988
10,000
~
Hi
_i
UJ
a
o
1,000
100
10
AERMOD
10
100	1,000
OBSERVED
10,000
Figure 1, Quantile-quantile plot of predicted versus observed arc-maximum concentrations for
AERMOD and ISC3 based on Prairie Grass convective data.
100,000
Q
w
a
O
10,000
1,000
100
100
& * iAERMOD
1,000	10,000
OBSERVED
100,000
Figure 2. Quantile-quantile plot of predicted versus observed arc-maximum concentrations for
AERMOD and ISC3 based on Prairie Grass stable data.
20
-------
A988
0.1
o
Ul
_l
tu
o
o
s
lAHRMOD
0.01
ISCST3
0.001
0.1
0.001
0.01
1
OBSERVED
Figure 3. QuantUe-quantile plot of predicted versus observed arc-maximum concentrations for
AERMOD and ISC3 based on Kincald SF6 convectivc data.
10
o
tu
—!
ui
o
o
s
ISCST3
AERMOD
0.1
0.01
10
0.01
1
0.1
OBSERVED
Figure 4. Quantile-quantile plot of predicted versus observed arc-maximum concentrations for
AERMOD and ISC3 based on Indianapolis convective data.
21
-------
A988
10
o
ui
_i
ui
a
o
0.1
0.01
AERMOD
0.01
0.1
10
OBSERVED
Figure 5. Quantile-quantile plot of predicted versus observed arc-maximum concentrations for
AERMOD and ISC3 based on Indianapolis stable data.
o
ui
_i
UJ
a
o
100
10
0.1
<0
AERMOD OPTION
AERMOD OPTION t\
0.1
10
OBSERVED
Figure 6. Quantile-quantile plot of predicted versus observed maximum one-hour
concentrations for AERMOD and ISC3 based on the Lovett data base. Options 1 and 2 refer to
the two complex terrain modeling approaches presently in AERMOD.
22
-------
A988
10
ISCST3
O
Ul
_l
ui
Q
O
s
1AERMOD OPTION 2>
JAERMQD OPTION 1
0.1
1
0.1
10
OBSERVED
Figure 7. Quantile-quantile plot of predicted versus observed maximum three-hour
concentrations for AERMOD and ISC3 based on the Lovett data base. Options 1 and 2 refer to
the two complex terrain modeling approaches presently in AERMOD.
10
ISCST3
Ul
a
Q
£
AERMOD OPTION 2
0.1
AERMOD OPTION 1
0.01
10
0.01
0.1
1
OBSERVED
Figure 8. Quantile-quantile plot of predicted versus observed maximum twenty-four hour
concentrations for AERMOD and ISC3 based on the Lovett data base. Options 1 and 2 refer to
the two complex terrain modeling approaches presently in AERMOD.
23
-------
A988
A E R M O D ] £,
Q
Ul
-J
UJ
o
O
s
0.1
ISCST3
0.01
0.1
1
0.01
OBSERVED
Figure 9. Quantile-quantile plot of predicted versus observed maximum one-hour
concentrations for AERM0D and ISC3 based on the Kincaid S02 data base.
1
0.1
AERMOP
1SCST3
0.01
0.01
0.1
OBSERVED
Figure 10. Quantile-quantile plot of predicted versus observed maximum three-hour
concentrations for AERMOD and ISC3 based on the Kincaid S02 data base.
24
-------
A988
0.1
o
ui
_i
ui 0.01
o
o
s
0.001
AERMOD
0.001
0.01
OBSERVED
0.1
Figure 11. Quantile-quantile plot of predicted versus observed maximum twenty-four hour
concentrations for AERMOD and ISC3 based on the Kincaid S02 data base.
25
-------
A988
REFERENCES
1.	J.C. Weil, "Updating applied diffusion models,** J. aim, and App. Meteor.. 24(11); 1111-1130
(1985).
2.	J.C. Weil, "Updating the ISC model through AERMIC," in Proceedings 85th Annual Meeting
of Air and Waste Management Association. 92-100.11. Air and Waste Management Association,
Pittsburgh, 1992.
3.	U.S. Environmental Protection Agency, User's Guide for the Industrial Source Complex
rises') Dispersion Models f revised) Volume I - User Instructions EPA-454/b-95-003a, U.S.
Environmental Protection Agency, Research Triangle Park, NC, 1995.
4.	S.G. Perry, AJ. Cimorelli, R.F. Lee, RJ. Paine, A. Venkatram, J.C. Weil, and R.B. Wilson,
"AERMOD: A dispersion model for industrial source applications," in Proceedings 87™ Annual
Meeting Air and Waste Management Association. 94-TA23.04, Air and Waste Management
Association, Pittsburgh, PA, 1994.
5.	U.S. Environmental Protection Agency, User's Guide for the Industrial Source Complex
fISC21 Dispersion Models. Volume II - Description of Model algorithms. EPA-450/4-92-008b,
U.S. Environmental Protection Agency, 1992.
6.	R.F. Lee, S.G.Perry, AJ. Cimorelli, R.J. Paine, A. Venkatram, J.C. Weil and R.B. Wilson,
"AERMOD - the developmental evaluation," Preprints of the Twenty-First NATO/CCMS
International Technical Meeting on Air Pollution Modeling and Its Application, Baltimore, MD,
U.S.A., 1995.
7.	AERMIC. Formulation of the AERMIC MODEL (AERMOD) (DRAFT). Regulatory Docket
AQM-95-01, AMS/EPA Regulatory Model Improvement Committee (AERMIC), 1995.
8.	AERMIC. Design Plan for the AERMOD Final fPhase TO Evaluation. Regulatory Docket
AQM-95-01, AMS/EPA Regulatory Model Improvement Committee (AERMIC), 1995.
9.	U.S. Environmental Protection Agency, Protocol for Determining the Best Performing Model.
EPA-454/R-92-025, U.S. Environmental Protection Agency, Research Triangle Park, 1992.
10.	Code of Federal Regulations, 40CFR, Part 51, App. W, Guideline on Air Quality Models, (as
modified by FR60, No. 153, pp 40465-40474, August 9, 1995), (1995).
11.	G.E. Willis, and J.W. Deardorff, "A Laboratory study of dispersion in the middle of the
convectively mixed layer," Atmos. Env.. 15: 109-117 (1981).
12.	G.A. Briggs, "Plume dispersion in the convective boundary layer. PartH: Analyses of
CONDORS field experiment data," J. Applied Meterol.. 32:1388-1425 (1993).
13.	S.G. Perry, DJ. Burns, L.H. Adams, RJ. Paine, M.G. Dennis, M.T. Mills, D.G. Strimaitis,
RJ. Yamartino, E.M. Insley, User's Guide to the Complex Terrain Dispersion Model Plus
26
-------
A988
Algorithms for Unstable Situations fCTDMPLUSt Volume 1: Model Description and User
Instructions. EPA/600/8-89/041, U.S. Environmental Protection Agency, Research Triangle Park,
NC, 1989, 196pp.
14.	G.I. Taylor, "Diffusion by continuous movements," Proc. London Math. Soc.. Scr.2(20):196-
211(1921).
15.	A. Venkatram, "Vertical dispersion of ground-level releases in the surface boundary layer,"
Atmos. Environ.. 26A:947-949 (1992).
16.	J.S. Hanna, "Lateral turbulent intensity and plume meandering during stable conditions," L
aim. App. Meteor- 22:1424-1430 (1983).
17.	S-R. Hanna and J.S. Chang. Modification of the Hybrid Plume Dispersion Model (HPDMt
for Urban Conditions and Its Evaluation Using ;he Indianapolis Data Set: Volume III: Analysis of
I Irhan Boundary Laver Data. Report Number A089-1200.IH, EPRI Project RP-02736-1,
Prepared for the Electric Power Research Institute, Palo Alto, CA, 1991.
18.	F.T.M.Nieuwstadt, "The steady-state height and resistance laws of the nocturnal boundary
layer: Theory compared with Cabauw observations," Bound. Layer Meteor.. 20:481-485 (1981).
19.	S.R. Hanna and R.J. Paine, "Hybrid Plume Dispersion Model (MPRM) development and
evaluation," J. Appl. Meteor.. 28:206-224 (1989).
20.	G.A. Briggs, Atmospheric Sciences Modeling Division, Air resources Laboratory/NOAA,
Research Triangle Park, NC, peisonal communication, 1995.
21.	G.A. Briggs, Diffusion Estimation for Small Emissions. ATDL conL File 79, ATDL, Oak
Ridge. TN (1973).
22.	J.A. Businger, J.C. Wyngaard, Y. Izumi and EJF. Bradley, "Flux-profile relationships in the
atmospheric surface layer. J. Atmos. Sci. 28:181-189 (1971).
23.	R.B. Stull. An Introduction to Boundary Laver Meteorology. Kluwer Academic Publishers.
Boston, MA, 1988,666pp.
24.	RJ. Paine, "Comparison of observed profiles of winds, temperature, and turbulence with
theoretical proceedings of the international results," in Specialty Conference: The Role of
Meteorology in Managing the Environment in the 90's. 9-2, Air and Waste Management
Association, Pittsburgh, PA, pp395-413 (1993).
25.	M.L.Barad. "Project Prairie Grass, a field program in diffusion," Geophysical Research
Papers. No. 59, Vols, I and II, Report AFCRC-TR-58-235, Air Force Cambridge Research
Center, (1958).
26.	M.K. I Ju and G.F.. Moore. Diagnostic validation of plume models at a plains site. EPRI
Report No. EA-3077, Research Project 1616-9, Electric Power Research Institute, Palo Alto,
27
-------
A988
CA, (1984).
27. D.R. Murray and N.E. Bowne, Urban power plant plume studies. EPRI Report No. EA-
5468, Research Project 2736-1, Electric Power Research Institute, Palo Alto, CA, (1988).
28. J.O. Paumier, S.G. Perry and DJ. Burns, "CTDMPLUS, A dispersion model for sources
near complex topography. Part II: Performance characteristics." J. Appl. Meteorol. 31:646-660,
(1992).
28
-------
TECHNICAL REPORT DATA
. REPORT KO.
EPA/600/A-96/055
2.

. TZTZ£ AND SUBTITLE
talent progress in the AERMIC model development program
5.REPORT DATE
S PERFORMING ORGANIZATION CODE
AUTHORCS)
3IMORELLI, Alan J. 3 VENKATRAM, Akula
>ERRY, Steven G. 'WEIL, Jeffrey C.
¦JEE. Russell F. 'WILSON, Robert B.
?AINE, Robert J.
SJPERFORMSiG ORGANIZATION REPORT NO.
. PERFORMING ORGANIZATION NAME AND ADDRESS
JSEPA, Region 3 'College of Engineering
41 Chestnut Street University of California at Riverside
'hiladclphia, PA 19107 Riverside, CA 92521
Same as Block 12 "Coop. Institute for Research in Env. Sciences
University of Colorado
SPA/OAQPS/EMAD/MD-14 Boulder, CO 80309
LTP.NC 27711
ENSR Corporation TJSEPA, Region 10
5 Nagog Park 1200 Sixth Avenue
tcton. MA 01720 Seattle, WA 98101
IOPROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
2. SPONSORING AGENCY NAME AND ADDRESS
JATIONAL EXPOSURE RESEARCH LABORATORY
>FFICE OF RESEARCH AND DEVELOPMENT
J.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
I3.TYPE OF REPORT AND PERIOD COVERED
Symposium Paper
l*. SPONSORING AGENCY CODE
EPA/600/9
5. SUPPLEMENTARY NOTES
6. ABSTRACT
fbe American Meteorological Society and the Environmental Protection Agency fonned a combined working group to develop '
mproved algorithms for regulatory air dispersion models. This workgroup has developed the AERMOD model which is a steady
late, plume model intended to be applied in those regulatory situations currently modeled with the Industrial Source Complex
i/Iodel (ISC3). This model development project is moving into its final phases and the purpose of this paper is to examine the
echmcal basis for the most recent AERMOD algorithms. In addition, there is a discussion of the performance of the model in
elation to measurements at five research field studies over a wide variety of meteorological conditions and source configurations.
7.	 	KEY WORDS AND DOCUMENT ANALYSIS
u DESCRIPTORS
hJDEN"TIFIi_.iS/ OPEN ENDED TERMS
e.COSAH



18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS {This Report)
UNCLASSIFIED
2I.NO, OF PAGES

20. SECURITY CLASS (This Page)
UNCLASSIFIED
22. PRICE
x:\abstract\aermic. wpd
-------