AERMOD:

Latest Features and Evaluation

Results.

1
 EPA-454/R-03-003

June 2003

AERMOD:
Latest Features and Evaluation Results

U.S. Environmental Protection Agency

Office of Air Quality Planning and Standards

Emissions Monitoring and Analysis Division

Research Triangle Park, NC 27711

2
DISCLAIMER
This paper has been reviewed by the Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, and has been approved for publication. Mention of trade
names or commercial products does not constitute endorsement or recommendation for use.

3
Acknowledgments
The Agency wishes to acknowledge AERMIC (the American Meteorological
Society/Environmental Protection Agency Regulatory Model Improvement Committee),
members of which have given a considerable amount of time, energy and dedication over the last
10 years to develop the AERMOD air dispersion modeling system:

R. J. Paine, ENSR Corporation

A. Venkatram, College of Engineering, University of California at Riverside

J. C. Weil, Cooperative Institute for Research in Environmental Sciences,

University of Colorado

R. B. Wilson, U.S. Environmental Protection Agency, Region X

S.G. Perry1, Atmospheric Sciences Modeling Division,

Air Resources Laboratory, EPA/ NOAA

R. F. Lee, Consultant, Meteorologist

A. J. Cimorelli, U.S. Environmental Protection Agency, Region III

W. D. Peters, US Environmental Protection Agency, OAQPS

In addition, Mr. Roger Brode of MACTEC Federal Programs, Inc (formerly known as Pacific
Environmental Services Inc) has provided considerable talent and support to the AERMOD
project and has conducted the analyses on which this report is based.

This report is based on an Air and Waste Management Association paper, No 69878, presented
on June 26, 2003 in San Diego, CA at the Annual 2003 Meeting by Robert Paine, a member of
the AERMIC, at the AB-3a session (Applications of Short-Range Dispersion Modeling –
Innovative Approaches and Lessons Learned).

This has been funded in part by the following U.S. EPA purchase orders: 1D-6635-NTSA, 1D-
6634-NTSA, 1D-6633-NTSA. Warren Peters was the Project Officer.

1
On assignment to the Atmospheric Research and Exposure Assessment Laboratory, U.S.
Environmental Protection Agency.

4
ABSTRACT
AERMOD is an advanced plume model that incorporates updated treatments of the boundary
layer theory, understanding of turbulence and dispersion, and includes handling of terrain
interactions. The model was formally proposed by EPA in April 2000 as a replacement for the
ISCST3 model. Several model enhancements were made as a result of public comment,
including the installation of the PRIME downwash algorithm. The latest version of the model,
version 02222, has been placed on EPA’s web site for beta test purposes. This paper reviews the
latest features and updated evaluation results for AERMOD version 02222.

5
INTRODUCTION
In 1991, the American Meteorological Society (AMS) and the United States 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
five EPA scientists was formed for this collaborative effort. AERMIC members and participants
in model code development and testing are listed as the authors of this paper.

For many years now, we have known that an update to EPA’s basic regulatory models is needed
(e.g., see Weil1). Responding to this need, AERMIC was formed in 1991 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 attention. This
broadened scope resulted in the development of a complete replacement for EPA’s Industrial
Source Complex Short-Term model version 3 (ISCST3)3 by: 1) adopting ISCST3's input/output
computer architecture; 2) updating, where practical, antiquated ISCST3 model algorithms with
newly developed or current state-of-the-art modeling techniques; and 3) insuring that all
processes presently modeled by ISCST3 will continue to be handled by the AERMIC Model
(AERMOD).

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 ISCST3 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. An initial version of the model subjected
to a “developmental” model evaluation with five databases was released to the public prior to the
Sixth EPA Modeling Conference in August 1995. After this release and receipt of public
comments, AERMIC conducted additional (independent) evaluations and made some further
improvements in response to public comments. A formal peer review4 of AERMOD was
conducted in 1998, and an independent model evaluation5,6 consisting of five additional
databases was completed in 1998.

The complete AERMOD modeling system consists of two pre-processors and the dispersion
model itself. The AERMOD 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 AERMOD mapping program
(AERMAP) is a stand-alone terrain pre-processor, which is used to both characterize terrain and
generate receptor grids for AERMOD.

6
PROPOSED RULEMAKING
On April 21, 2000, the EPA proposed7 to replace ISCST3 with AERMOD version 99351. EPA
described this new model as an advanced dispersion technique that incorporates state-of-the-art
boundary layer parameterization techniques, convective dispersion, plume rise formulations, and
complex terrain/plume interactions. Relative to ISCST3, AERMOD as proposed contained 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). Table 1 provides a more extensive list of the comparison features
between AERMOD and ISCST3.

AERMOD as proposed did not incorporate newly developed building downwash algorithms that
were developed independently and concurrently for the PRIME model (as installed in ISC-
PRIME8), sponsored by EPRI. At the 7th USEPA Modeling Conference held in June 2000 and in
the written comments provided afterward, many public comments focused upon the need to
enhance AERMOD with the advancements in building downwash treatments that PRIME offers.

IMPLEMENTATION OF PRIME IN AERMOD

There were several issues involved with implementing9 the PRIME downwash algorithms into
the AERMOD model. The PRIME algorithm as implemented in ISC-PRIME was designed to
use vertical profiles of wind and temperature that are consistent with the ISCST3 profiles,
whereas AERMOD generates vertical profiles of wind and temperature based on similarity
scaling and can also incorporate a full profile of measurements. PRIME was implemented in
AERMOD to use the AERMOD meteorological profiles.

The ISC-PRIME model uses ambient turbulence intensities based on PG stability class to
determine the distance at which the wake turbulence intensity has decayed to ambient levels, and
also uses Pasquill-Gifford (PG)-based dispersion beyond the wake. The PRIME algorithm was
implemented in AERMOD to use ambient turbulence intensities based on the AERMOD
profiles.

The more significant issues were related to the use of a non-Gaussian probability distribution
function (PDF) for the vertical dispersion in the convective boundary layer (CBL) in AERMOD,
and AERMOD’s treatment of the direct, indirect and penetrated plumes in the CBL. The ISC-
PRIME model uses a Gaussian vertical distribution for both convective and stable conditions,
consistent with the ISCST3 model.

7
Table 1. Comparison of dispersion model features: AERMOD vs. ISCST3.

Feature ISCST3 AERMOD (version 02222) Comments

Types of sources modeled Point, area, and volume sources Same as ISCST3 Models are comparable
Plume Rise Uses Briggs equations with In stable conditions, uses Briggs equations with AERMOD is better because in stable
stack-top wind speed and winds and temperature gradient at stack top and half- conditions it factors in wind and temperature
vertical temperature gradient way to final plume rise; in convective conditions, changes above stack top, and in unstable
plume rise is superposed on the displacements by conditions it accounts for convective updrafts
random convective velocities and downdrafts
Meteorological Data Input One level of data accepted An arbitrarily large number of data levels can be AERMOD can adapt multiple levels of data to
accommodated various stack and plume heights
Profiling Meteorological Only wind speed is profiled AERMOD creates profiles of wind, temperature, and AERMOD is much improved over ISCST3 in
Data turbulence, using all available measurement levels this area
Use of Meteorological Data Stack-top variables for all Variables measured throughout the plume depth AERMOD treatment is far more advanced than
in Plume Dispersion downwind distances (averaged from plume centerline to 2.15 sigma-z that of ISCST3; accounts for meteorological
below centerline; changes with downwind distance) data throughout the plume depth
Plume Dispersion: General Gaussian treatment in Gaussian treatment in horizontal and in vertical for AERMOD’s unstable treatment of vertical
Treatment horizontal and vertical stable conditions; non-Gaussian probability density dispersion is a more accurate portrayal of
function in vertical for unstable conditions actual conditions
Urban Treatment Urban option either on or off; Population is specified, so treatment can consider a AERMOD provides variable urban treatment
no other specification available; variety of urban conditions; sources can individually as a function of city population, and can
all sources must be modeled be modeled rural or urban selectively model sources as rural or urban
either rural or urban
Characterization of Choice of rural or urban Selection by direction and month of roughness AERMOD provides the user with considerably
Modeling Domain Surface length, albedo, and Bowen ratio, providing user more options in the selection of the surface
Characteristics flexibility to vary surface characteristics characteristics
Boundary Layer Parameters Wind speed, mixing height, and Friction velocity, Monin-Obukhov length, AERMOD provides parameters required for
stability class convective velocity scale, mechanical and convective use with up-to-date planetary boundary layer
mixing height, sensible heat flux (PBL) parameterizations; ISCST3 does not
Mixed Layer Height Holzworth scheme; uses Has convective and mechanical mixed layer height; AERMOD’s formulation is significantly more
interpolation based upon convective height based upon hourly accumulation advanced than that of ISCST3, includes a
maximum afternoon mixing of sensible heat flux mechanical component, and in using hourly
height input data, provides a more realistic sequence
of the diurnal mixing height changes

8
Feature ISCST3 AERMOD (version 02222) Comments
Terrain Depiction Elevation at each receptor point Controlling hill elevation and point elevation at each AERMOD’s terrain pre-processor provides
receptor, obtained from special terrain pre-processor information for advanced critical dividing
(AERMAP) that uses digital elevation model (DEM) streamline height algorithms and uses digital
data data to obtain receptor elevations
Plume Dispersion: Plume Based upon 6 discrete stability Uses profiles of vertical and horizontal turbulence Use of turbulence-based plume growth with
Growth Rates classes only; dispersion curves (from measurements and/or PBL theory); variable height dependence rather than that based upon
(Pasquill-Gifford) are based with height; uses continuous growth functions rather stability class provides AERMOD with a
upon surface release than a discrete (stability-based) formulation substantial advancement over the ISCST3
experiments (e.g., Prairie treatment
Grass)
Plume Interaction with If plume centerline is above lid, Three plume components are considered: a “direct” The AERMOD treatment avoids potential
Mixing Lid: convective a zero ground-level plume that is advected to the ground in a downdraft, underpredictions suffered by ISCST3 due to its
conditions concentration is assumed an “indirect” plume caught in an updraft that reaches “all or nothing” treatment of the plume;
the lid and eventually is brought to the ground, and a AERMOD’s use of convective updrafts and
plume that penetrates the mixing lid and disperses downdrafts in a probability density function
more slowly in the stable layer aloft (and which can approach is a significant advancement over
re-enter the mixed layer and disperse to the ground) ISCST3
Plume Interaction with The mixing lid is ignored A mechanically mixed layer near the ground is AERMOD’s use of a mechanically mixed layer
Mixing Lid: stable (assumed to be infinitely high) considered. Plume reflection from an elevated lid is is an advancement over the very simplistic
conditions considered. ISCST3 approach
Building Downwash Combination of Huber-Snyder New PRIME downwash algorithm installed AERMOD benefits from the technological
and Scire-Schulman advances offered by the PRIME model
algorithms; many
discontinuities

9
To address these issues, the AERMIC committee adopted an approach that defines two plume
“states”, one corresponding to a plume that is influenced by building downwash, and the other
corresponding to a plume that is not influenced by building downwash. AERMOD models the
“wake state” plume using the PRIME algorithms with the adaptations described above, and
models the “non-wake state” plume using the regular AERMOD algorithms for a source without
building downwash. The contributions from the two plume states are combined using a
weighting factor that is a function of the receptor location relative to the building wake.

For a receptor located within the wake region, the AERMOD model uses the concentration
calculated by the PRIME algorithm, and the model transitions to the AERMOD estimate
(without downwash) beyond the wake region. The lateral and vertical extents of the wake region
are defined internally by the PRIME algorithm. For purposes of transitioning to the AERMOD
estimate, the longitudinal extent of the wake region is defined as the maximum of 15R and the
distance where wake turbulence intensity decays to the ambient level, where R is the wake
length scale and is a function of the building dimensions. Beyond the wake region, the total
concentration is calculated as follows:

χ TOTAL = γ χ PRIME + (1 − γ ) χ AERMOD

(1)

The weighting function, γ, is equal to 1.0 within the wake region, and beyond the wake region is
calculated as follows:

⎛ − ( x − σ xg )2 ⎞ ⎛ − ( y − σ yg )2 ⎞ ⎛ − ( z − σ zg )2 ⎞
γ = exp ⎜ ⎟ exp⎜ ⎟ exp⎜ ⎟
⎜ 2σ 2 ⎟ ⎜ 2σ 2 ⎟ ⎜ 2σ 2 ⎟
⎝ xg ⎠ ⎝ yg ⎠ ⎝ zg ⎠ (2)

where:

x= downwind distance of receptor from upwind edge of the building;

y= lateral distance of receptor from building centerline;

z= receptor height above stack base, including terrain and flagpole;

σxg = maximum of 15R and the distance to transition from wake to ambient turbulence;

σyg = lateral distance from building centerline to lateral edge of the wake at receptor location;
and

σzg = height of the wake at the receptor location.

For applications involving terrain effects and building downwash, the AERMOD component is
calculated with the standard terrain treatment, and the PRIME component is calculated with the
minimum terrain weighting factor of 0.5, since the wake region is considered to be near neutral
due to the building-enhanced turbulence. The use of the receptor height above stack base in the
calculation of the vertical component of γ indicates that if the terrain is within the wake, then the

10
PRIME component should dominate. However, if the plume is rising above the wake and terrain
extends above the wake, then the AERMOD component should become imprtant.

During the developmental evaluation of AERMOD with PRIME, preliminary results indicated a
tendency for the model to overpredict during light wind convective conditions. The PRIME
algorithm includes a test on the trajectory angle of the rising plume to determine if the plume
will escape the effects of the building. If the trajectory of the plume falls below 45 degrees from
horizontal before the plume rises above the top of the wake, then the plume is subjected to
building downwash influences. The light wind convective conditions for the Bowline data were
evaluated to determine a “best fit” for this critical trajectory angle based on the normalized mean
square error, and a best fit was found for a critical angle of 20 degrees. Based on this result,
PRIME was implemented in AERMOD using a critical angle of 20 degrees to determine if wake
effects apply.

OTHER MODIFICATIONS TO AERMOD

Many commentors supported the implementation of AERMOD as an improved and advanced
dispersion model. Some of the public comments (in addition to those advocating the installation
of the PRIME advancements into AERMOD) led to additional improvements to AERMOD
version 02222, released by EPA as a beta test version and described in an updated Model
Formulation Document10.

• An option to use representative measurements of delta-T and wind speed in lieu of cloud
cover in AERMET for characterizing boundary layer parameters in stable conditions has
been included. This option is new and has not been extensively tested by the user
community.

• AERMAP was modified to remove its dependence upon the terrain domain for determining
controlling hill heights for each receptor. This change involved the concept that a terrain
feature has an influence zone on the surrounding area up to about 10 hill relief heights. For
each receptor, terrain features more distant than 10 hill relief heights are not considered in
AERMAP.

• AERMAP was modified to be able to convert receptor, source and elevation coordinates
from North American Datum (NAD) of 1927 and other datums to NAD 1983 using the
United States Geological Survey (USGS) sanctioned program, NADCON version 2.1.

• AERMET was modified to read Forecast Systems Laboratory (FSL) upper air and Hourly
United States Weather Observations (HUSWO) surface meteorological data formats.

• A correction was made to AERMOD to avoid elevated concentrations for terrain below stack
base from the virtual image source. This fix also addressed some of the public comments
regarding anomalous AERMOD concentrations from a hypothetical stack located on the top
of a hypothetical terrain feature.

• In AERMOD, the urban mode was modified to allow the user to input the urban roughness
length as an optional parameter. The latest version of AERMOD also includes an adjustment

11
 to the friction velocity and the Monin-Obukhov length for urban stable cases, by equating the
"convective" sigma-w based on the urban "convective" velocity scale with the mechanical
sigma-w based on the friction velocity evaluated at a height of 7 times the urban roughness
length.

• The reference urban mixing height for a reference population of 2,000,000 was changed from
500 m to 400 m to better match literature references and observed data.

• The minimum layer near the ground for calculating effective parameters was changed from
the lowest 2 m to 5 m to avoid problems for high roughness length applications.

• Enhancements were made to AERMOD based on the ISCST3 model version 00101. These
included

1) the use of globally allocatable arrays for data storage;

2) expanded data structures to allow for output of concentration and deposition in a single
model run (for use when deposition algorithms are added to AERMOD);

3) EVENT (individual period) processing for short-term culpability analyses;

4) post-1997 PM10 processing;

5) TOXICS option enhancements such as optimizations for area sources, the Sampled
Chronological input Model (SCIM) option, and Season and Hour-of-Day (SEASONHR)
output file option;

6) explicit treatment of multiple-year meteorological data files and ANNUAL averages;

7) the SHRDOW and SHRDOW7 options for specifying emissions that vary by Season,
HouR-of-Day, and Day-of-Week; and

8) improved data structures for field length and filename lengths.

• Other minor corrections and/or adjustments were made to the AERMOD code due to public
comments or user/beta-test comments. They are documented in the comments sections of the
model source codes. Some of the more notable changes involve the following features:

¾ The meander feature that has the most effect in very light winds is now applicable for
both stable and unstable conditions, has been removed from the PRIME downwash
component of the model, and combines "plume" and "pancake" components of
concentrations rather than just blending the lateral dispersion term.

¾ There is a modification to the potential temperature gradient profile for extrapolating

above the highest measurement height for cases with observed temperature profiles.

¾ A modification was made to the upper limit on the integration for the critical dividing
streamline height.

12
In addition to the full-scale version of AERMOD, an ad-hoc group within EPA and the states is
working to develop a screening version of AERMOD, AERSCREEN. When completed,
AERSCREEN will generate screening meteorological data appropriate for the site in question
and execute the AERMOD model to provide a conservative estimate of concentration impacts.

Another ad-hoc group, largely from the state of West Virginia, is working on an objective
method to use digitized land use data to develop input values to AERMET for roughness length,
albedo, and Bowen Ratio. This technique, referred to as AERSurface, will help to improve the
consistency of the formulation of this input to AERMET.

MODEL EVALUATION DESIGN

The model evaluation was designed to provide diagnostic as well as descriptive information
about the model performance. The procedures used were designed to address the following
questions:

• Does AERMOD provide good predictions for the “right” reasons (a model physics
evaluation)?

• How well does AERMOD predict the peak ground-level concentrations that are used to
assess compliance with air quality regulations (an operational performance evaluation)?

• Is AERMOD’s performance significantly better than that of other applied models, such as
ISCST3, HPDM (Hanna and Paine11), RTDM (Paine and Egan12), and CTDMPLUS13,14?

The ISCST3, RTDM, and CTDMPLUS models are currently approved by the EPA for general
use in regulatory applications (“Appendix A” models). AERMOD is being proposed for use in
place of these models, although CTDMPLUS would still be available for applications involving
a well-defined hill or ridge. HPDM was developed by EPRI as a state-of-the-art model for use in
simple terrain. Comparison of AERMOD performance to that of HPDM is useful as a
benchmark. Other advanced models such as SCIPUFF15 and ADMS16 were considered when
this model comparison effort was initiated. However, these models were not released to the
public at that time and thus were not included.

The average model error (or “residual”) examined over a broad range of input variables is used
to evaluate the model physics. The residual is examined by plotting the ratio of the model
prediction to observed values for data paired in time as a function of various model input
variables (e.g., distance, wind speed, and mixing height). Residual plots can be examined for
partial data sets such as for stable or unstable conditions. If a significant trend is observed in the
predicted-to-observed ratio as a function of the abscissa variable, then the model physics
associated with or responsible for this feature can be further examined.

Operational performance of models for predicting compliance with air quality regulations,
especially those involving a peak or near peak value at some unspecified time and location, can
be assessed with quantile-quantile (Q-Q) plots (Chambers et al.17). Q-Q plots, are created by
sorting by rank the predicted and the observed concentrations from a set of predictions initially
paired in time and space. The sorted list of predicted concentrations are then plotted by rank

13
against the observed concentrations also sorted by rank. These concentration pairs are no longer
paired in time or location. However, the plot is useful for answering the question, “Over a
period of time and over a variety of locations, does the distribution of the model predictions
match those of observations?” Scatterplots, which use data paired in time (and / or space),
provide a more strict test, answering the question: “At a given time and place, does the
magnitude of the model prediction match the observation?” It is the experience of model
developers (e.g., Weil, et al.18 and Liu and Moore19) that wind direction uncertainties can and do
cause disappointing scatterplot results from what are otherwise well-performing dispersion
models. Therefore, the Q-Q plot instead of the scatterplot is a more pragmatic procedure for
demonstrating model performance of applied models. Venkatram20 makes a cogent argument for
the use of Q-Q plots for evaluating regulatory models.

In addition to the residual and Q-Q plots, the difference between AERMOD and ISCST3 was
assessed with a robust test statistic (robust highest concentration, or RHC21) that represents a
smoothed estimate of the highest concentrations, based on a tail exponential fit to the upper end
of the concentration distribution. With this procedure, the effect of extreme values on model
comparison is reduced. The RHC statistic is reported for each of the 10 databases. The robust
highest concentration is given by:

⎛ 3n − 1⎞
RHC = χ {n} + ( χ − χ {n}) ln⎜ ⎟
⎝ 2 ⎠
(3)

where n = Min (mo, m), mo is the number of values used to characterize the upper end of the
concentration distribution, m is the number of values exceeding a specified threshold value, χ is
the average of the n - 1 largest values, and χ{n} is the nth largest value. In this evaluation, the
value of mo was taken to be 26 except for databases with a limited sample size (for which mo was
taken to be 11).

Highlights of the evaluation results for the proposed regulatory version of AERMOD were first
presented by Paine, et al.5. As with Paine, et al., the evaluation results here include selected
residual plots and Q-Q plots to address the model performance issues noted above. In the
AERMOD evaluation, many of the statistical tests and comparisons with observations were
applied to analyze the performance of the model and various model algorithms. The observed
peak concentration for a given arc of samplers was compared to the predicted arc maximum. The
comparisons included time and downwind-distance pairings to challenge the model components.
Residual plots (predicted/observed, paired in time and downwind distance) of concentration
estimates were used to judge whether AERMOD was performing correctly and was yielding
better results (than existing models). Generally, residuals were plotted as a function of distance,
although residuals versus other parameters, such as friction velocity, Monin-Obukhov length,
and mixing height, proved to be extremely valuable diagnostic tools. In addition, Q-Q plots were
used to examine the ability of the model to reproduce the distribution of observed concentrations
over a wide range of environmental conditions. Reproducing the measured distribution,
particularly the high concentration end, is important in regulatory applications.

14
For the intensive tracer data sets (Prairie Grass, Kincaid SF6, Indianapolis, and Tracy),
concentration residuals of the form <Cp/Co> were plotted as a function of downwind distance
(arc distance) for each of two stability regimes, convective and stable. Here, Co is the maximum
observed concentration and Cp the maximum predicted concentration on an arc at a given time.
The brackets, <Cp/Co>, indicate the mean of the ratio. These data were paired in time and
downwind distance. For the tracer databases, observations and predictions corresponded to
maximum concentrations on each arc of samplers, rather than at each individual sampler, to
eliminate the effect of wind direction uncertainties on the evaluation results. The use of arc
maxima was possible due to the dense coverage of samplers along each arc. For the other
nondownwash data sets (Kincaid SO2, Lovett, Baldwin, Clifty Creek, Martins Creek, and
Westvaco), where the sampler array was not sufficiently dense to arrange the data in arcs,
residual plots by distance were not meaningful. In contrast to the tracer studies (where a Co, Cp
pair are available for each arc-distance of each time period), the long-term databases have only a
single Co, Cp pair selected (for each time period) as the maximum observed and predicted
concentrations, respectively in the entire receptor array.

For the tracer databases, results for 1-hour averages are reported (with the Prairie Grass 10-
minute measurements used). For the long term SO2 data sets, 3-hour, 24-hour, and annual results
are reported. All of the observed concentrations for the long-term databases are subject to an
additional uncertainty related to the estimate of background concentration that is subtracted from
the monitored observations. In addition, it should be realized that SO2 monitors typically have a
detection limit on the order of 6 ppb (16 mg m-3), and baseline (zero) drifts of up to 10 ppb (26
mg m-3). Concentrations below the detection limit are typically set to half of the limit (8
mg m-3), even though they may actually be zero. Baseline drift is generally ignored. Therefore,
the uncertainties due to the combined errors in SO2 measurements from the detection limit
treatment, ignored baseline drifts, and background concentration estimates reflect on the
reliability of the observed concentrations, particularly for annual averages. Peak short-term
averages are not affected significantly because the uncertainty is typically a small percentage of
the reported value.

MODEL EVALUATION RESULTS FOR NON-DOWNWASH DATABASES

The combined developmental and independent performance evaluation5,6 of AERMOD involved
four short-term tracer studies and six conventional long-term SO2 monitoring databases in a
variety of settings. The purpose of the evaluation studies was to be sure that AERMOD had
been tested in a variety of types of environments for which it will be used. Compared with many
widely used applied models, AERMOD has been subjected to a large degree of testing with these
evaluation databases.

The evaluation of AERMOD was accomplished in two phases. The first phase, the
“developmental evaluation,” was performed concurrently with the development of the model. As
each model feature was tested, a relevant portion of the developmental evaluation was repeated
with five databases to identify any problems that might have been introduced with that feature.
Because the model could have been inadvertently biased by particular characteristics of the
developmental databases, a second phase, the “independent evaluation,” was conducted using
five additional data sets.

15
AERMOD is intended to handle a variety of pollutant source types, including surface and
buoyant elevated sources, in a wide variety of settings such as rural and urban as well as flat and
complex terrain. With this in mind, data were selected from five diverse field studies for the
developmental evaluation. Due to space limitations, maps of the various sites are not provided
in this report, but can be found on the Internet in Paine, et al.5. A brief description of these data
sets is provided below.

Developmental Evaluation (No Downwash)

The Prairie Grass study (Barad22; Haugen23) used a near-surface, non-buoyant tracer release in a
flat rural area in Nebraska. This study involved a tracer of SO2 released at 0.46 m above the
surface. Surface sampling arrays (arcs) were positioned from 50 m to 800 m downwind.
Meteorological data included the 2-m level wind direction and speed, the root-mean-square wind
direction fluctuation, and the temperature difference (∆T) between 2 m and 16 m. Other surface
parameters, including friction velocity, Monin-Obukhov length, and lateral plume spread were
estimated. Wind, turbulence, and temperature were obtained from a multi-leveled instrumented
16-m meteorological tower. A total of 44 ten-minute sampling periods were used, including
both convective and stable conditions.

The Kincaid SF6 study (Liu and Moore19; Bowne et al.24) involved an elevated, highly-buoyant
tracer release in a flat rural area of Illinois. Two intensive measurement periods each lasting six
weeks were conducted during the spring and summer of 1980 and 1981. During these periods,
approximately 200 monitors were placed on arcs ranging from about 500 m to 50 km downwind
of the single 187-m stack and provided 1-hour averaged concentration samples for a total of
1,801 arc-hours. Meteorological data included wind speed, direction, and temperature from a
tower instrumented at 2, 10, 50, and 100 m levels, and nearby National Weather Service (NWS)
data. Estimates of lateral plume spread were obtained from the sampling arcs.

The Indianapolis study (Murray and Bowne25) consisted of an elevated, buoyant tracer (SF6)
released in a flat-terrain urban to suburban area from a single 84-m stack. Data are available for
approximately a four- to five-week period with 177 monitors providing 1-hour averaged samples
along arcs from 250 m to 12 km downwind for a total of 1,297 arc-hours. Meteorological data
included wind speed and direction, sigma-theta on a 94-meter tower; and wind speed, ∆T (2m -
10m) and other supporting surface data at three other 10-m towers. Observed plume rise and
estimates of plume sigma-y are also available from the database.

The Kincaid SO2 study (Liu and Moore19; Bowne et al.24) was conducted at the same location as
the Kincaid SF6 study. It involved a buoyant, continuous release of SO2 from a 187-m stack in
rural flat terrain. The study included about six months of data between April 1980 and June 1981
(a total of 4,614 hours of samples). There were 30 SO2 monitoring stations providing 1-hour
averaged samples from about 2 km to 20 km downwind of the stack. The meteorological data
were the same as in the Kincaid tracer study.

The Lovett Power Plant study (Paumier et al.26) consisted of a buoyant, continuous release of
SO2 from a 145-m tall stack located in a complex terrain, rural area in New York State. The data
spanned one year from December 1987 through December 1988. Data were collected from 12
monitoring sites (ten on elevated terrain and two near stack-base elevation) that were located

16
about 2 to 3 km from the plant. The monitors provided hourly-averaged concentrations. The
important terrain features rise approximately 250 m to 330 m above stack base at about 2 to 3 km
downwind from the stack. Meteorological data include winds, turbulence, and ∆T from a tower
instrumented at 10 m, 50 m, and 100 m. National Weather Service surface data were available
from a station 45 km away.

Independent Evaluation (No Downwash)

The independent evaluation of AERMOD initially employed the first three databases described
below. Results for two additional databases were added in response to peer review comments.

The Baldwin Power Plant (Hanna and Chang27) is located in a rural, flat terrain setting of
southwestern Illinois and has three identical 184-m stacks aligned approximately north-south
with a horizontal spacing of about 100 m. There were 10 SO2 monitors that surrounded the
facility, ranging in distance from two to ten km. On-site meteorological data was available
during the study period of 1 April 1982 through 31 March 1983 and consisted of hourly-
averaged wind speed, wind direction, and temperature measurements taken at 10 m and wind
speed and wind direction at 100 m.

The Clifty Creek Power Plant is located in rural southern Indiana along the Ohio River with
emissions from three 208-m stacks during this study. The area immediately north of the facility
is characterized by cliffs rising about 115 m above the river and intersected by creek valleys. Six
nearby SO2 monitors (out to 16 km from the stacks) provided hourly averaged concentration
data. Meteorological data from a nearby 60-m tower covered the two-year period from 1
January 1975 through 31 December 1976, although only the data from 1975 were used in this
evaluation. This database was also used in a major EPA-funded evaluation of rural air quality
dispersion models in the early 1980s28.

The Martins Creek Steam Electric Station is located in a rural area along the Delaware River on
the Pennsylvania/New Jersey border, approximately 30 km northeast of Allentown, PA and
95 km north of Philadelphia, PA. The area is characterized by complex terrain rising above the
stacks. Sources included multiple tall stacks ranging from 122 to 183 m in height. The seven
SO2 monitors29 were located on Scotts Mountain, which is about 2.5 - 8 km southeast of the
Martins Creek facility. On-site meteorological data covered the period from 1 May 1992
through 19 May 1993. Hourly temperature, wind speed, wind direction, and sigma-theta
(standard deviation of the horizontal wind direction) at 10 m were recorded from an
instrumented tower located in a flat area approximately 2.5 km west of the plant. In addition,
hourly multi-level wind measurements were taken by a sodar located approximately three km
southwest of the Martins Creek station.

The Westvaco Corporation’s pulp and paper mill in rural Luke, Maryland is located in a complex
terrain setting in the Potomac River valley (Strimaitis et al.30). A single 183-m buoyant source
was modeled for this evaluation. There were 11 SO2 monitors surrounding the facility, with
eight monitors well above stack top on the high terrain east and south of the mill at a distance of
800 - 1500 m. Hourly meteorological data (wind, temperature, and turbulence) were collected
between December 1980 and November 1991 at three instrumented towers: the 100-m Beryl
tower in the river valley about 400 m southwest of the facility; the 30-m Luke Hill tower on a

17
ridge 900 meters north-northwest of the facility; and the 100-m Met tower located 900 m east-
southeast of the facility on a ridge across the river.

The Tracy Power Plant (DiCristofaro et al.31) is located 27 km east of Reno, Nevada in the rural
Truckee River valley completely surrounded by mountainous terrain. A field tracer study was
conducted at the power plant in August 1984 with SF6 being released with the moderately
buoyant plume from a 91-m stack. A total of 128 hours of data were collected over 14
experimental periods. Stable atmospheric conditions were dominant for this study. On-site
meteorological data (wind, temperature, and turbulence) for Tracy were collected from an
instrumented 150-m tower located 1.2 km east of the power plant. The wind measurements from
the tower were extended above 150 meters using a Doppler acoustic sounder and temperature
measurements were extended with a tethersonde.

Evaluation Results
For completeness, EPA re-ran the evaluation of the 10 non-downwash databases. The earlier
evaluation results indicated that AERMOD shows superior performance relative to ISCST3 over
all of the databases tested. The newest evaluation results are very similar to the results reported
by Pain et al.5. A summary of the robust highest concentration prediction results is provided in
Table 2. More complete results5,6 that include Q-Q plots and some residual plots for the
previous evaluation are available and would show very little change for the updated evaluation.

Table 2. Summary of AERMOD evaluation results – nondownwash databases (previous results5

in parentheses).

Database Ratio of Modeled/Observed Robust Highest Concentrations*

Prairie Grass (SO2) AERMOD: 0.89 (0.87) (1-hr avg)
Flat, grassy field (Nebraska, USA) ISCST3: 1.50 (1-hr avg)
Kincaid (SF6) AERMOD: 0.77 (0.76) (1-hr avg)
Flat, rural (Illinois, USA) ISCST3: 0.68 (1-hr avg)
Kincaid: (SO2) AERMOD: 0.98 (1.01) (3-hr avg)
Flat, rural (Illinois, USA) ISCST3: 0.56 (3-hr avg)

AERMOD: 0.94 (0.97) (24-hr avg)

ISCST3: 0.45 (24-hr avg)

AERMOD: 0.30 (0.30) (annual peak)

ISCST3: 0.14 (annual peak)
Baldwin (SO2): AERMOD: 1.24 (1.31) (3-hr avg)
Flat, rural (Illinois, USA) ISCST3: 1.43 (3-hr avg)

AERMOD: 0.97 (1.02) (24-hr avg)

ISCST3: 1.14 (24-hr avg)

AERMOD: 0.97 (0.97) (annual peak)

ISCST3: 0.63 (annual peak)
Indianapolis (SF6) AERMOD: 1.11 (1.20) (1-hr avg)
Flat, urban (Indiana, USA) ISCST3: 1.30 (1-hr avg)
Clifty Creek (SO2) AERMOD: 1.05 (1.25) (3-hr avg)
Moderately hilly terrain, rural (Indiana, ISCST3: 0.98 (3-hr avg)
USA)

18
Database Ratio of Modeled/Observed Robust Highest Concentrations*
AERMOD: 0.67 (0.72) (24-hr avg)
ISCST3: 0.67 (24-hr avg)

AERMOD: 0.54 (0.54) (annual peak)

ISCST3: 0.31 (annual peak)
Tracy (SF6): AERMOD: 1.04 (1.07) (1-hr avg)
Hilly terrain, rural (Nevada, USA) ISCST3: 2.81 (1-hr avg)
CTDMPLUS: 0.77 (1-hr avg)
Martins Creek (SO2): Hilly terrain, rural AERMOD: 1.12 (1.06) (3-hr avg)
(Pennsylvania/New Jersey, USA) CTDMPLUS: 4.80 (3-hr avg)
ISCST3: 7.25 (3-hr avg)
RTDM: 3.33 (3-hr avg)

AERMOD: 1.78 (1.72) (24-hr avg)

CTDMPLUS: 5.56 (24-hr avg)
ISCST3: 8.88 (24-hr avg)
RTDM: 3.56 (24-hr avg)

AERMOD: 0.78 (0.74) (annual peak)

CTDMPLUS: 2.19 (annual peak)
ISCST3: 3.37 (annual peak)
RTDM: 1.32 (annual peak)
Lovett (SO2) AERMOD: 1.03 (1.00) (3-hr avg)
CTDMPLUS: 2.36 (3-hr avg)
ISCST3: 8.20 (3-hr avg)

AERMOD: 1.01 (1.00) (24-hr avg)

CTDMPLUS: 2.02 (24-hr avg)
ISCST3: 9.11 (24-hr avg)

AERMOD: 0.85 (0.78) (annual peak)

CTDMPLUS: 1.71 (annual peak)
ISCST3: 7.49 (annual peak)
Westvaco (SO2): AERMOD: 1.06 (1.08) (3-hr avg)
Hilly terrain, rural (Maryland/Virginia, CTDMPLUS: 2.14 (3-hr avg)
USA) ISCST3: 8.50 (3-hr avg, estimated*)

AERMOD: 1.07 (1.14) (24-hr avg)

CTDMPLUS: 1.54 (24-hr avg)

AERMOD: 1.59 (1.64) (annual peak)

CTDMPLUS: 0.93 (annual peak)
*Notes:

1. The Robust Highest Concentration (RHC) is a statistical estimator for the highest concentration. It is
determined from a tail exponential fit to the high end of the frequency distribution of observed and
predicted values. The number of points used for the fit is arbitrary, but usually ranges between 10
and 25.

2. The estimated 3-hour ISCST3 result for Westvaco is derived from the EPA Complex Terrain Model
Development study (Strimaitis et al.30) in which several models, including CTDMPLUS and
COMPLEX-I (now part of ISCST3), were evaluated.

19
The overall model evaluation results for AERMOD version 02222 with nondownwashing
databases can be summarized as follows, taking one composite (geometric mean) ratio of
predicted to observed RHC value for short-term averages at each site, and also taking the annual
average ratio at sites with year-long databases:

ƒ 1.03 is the overall predicted-to-observed ratio for short-term averages (with a range among
sites from 0.76 to 1.35).

ƒ 0.73 is the overall predicted-to-observed ratio for annual averages (with a range among sites
from 0.30 to 1.64).

While the predicted-to-observed ratios did not vary substantially for AERMOD between simple
and complex terrain sites, there was a large change in the average ratio for ISCST3 : 0.96 for
simple terrain and 6.4 for complex terrain.

MODEL EVALUATION RESULTS FOR DOWNWASH DATABASES

A developmental evaluation of the AERMOD model with PRIME added was conducted on four
developmental databases prior to its application to four independent databases. Paine32 describes
these databases and others that were originally considered for the EPRI PRIME evaluation study.
The developmental databases (described below) included one half of the days selected at random
from a full year of data for the Bowline power plant database located on the Hudson River near
Haverstraw, New York, the Millstone power plant located on the Connecticut coast, the Duane
Arnold Energy Center (DAEC) located in eastern Iowa, and the Alaska North Slope field study
near Prudhoe Bay, Alaska. The Bowline Point database was used in both the developmental and
evaluation evaluations because it was the only full year database, and more complete
development testing of both ISC-PRIME and AERMOD version 02222 required the use of half
of this database.

The main purpose of testing with the developmental databases was to assure that the AERMOD
predictions were consistent with the ISC-PRIME predictions for stack-receptor combinations
dominated by building downwash. However, as noted earlier, one correction was made to
AERMOD to adjust the threshold trajectory angle of the rising plume that determines whether
the plume will escape the effects of the building. As noted below, the evaluation results for the
developmental databases included both underpredictions and overpredictions for both AERMOD
and ISC-PRIME. However, no attempt was made in the developmental phase of testing to
further adjust the downwash algorithm.

Developmental Evaluation (Downwash)

The Bowline Point site33, located in the Hudson River valley in New York State, is shown in
Figure 1 (topographic map). The electric utility site included two 600-MW units, each with an
86.9-m stack and a dominant roof tier with a height of 65.2 m high in a rural area. There were
four monitoring sites as shown in Figure 1 that ranged from about 250 to 850 m from the stacks.
Hourly emissions data was determined from load data, coal analyses, and site-specific
relationships between loads and fuel consumption. Meteorological data was obtained from a

20
100-m tower at the site. This site was also used as an independent evaluation database with the
entire year included.

Figure 1. Bowline Point Study Area (SO2 Releases)

The Millstone nuclear power plant is located on the Connecticut coast, near Niantic. The model
evaluation database34 features 36 hours of SF6 emissions from a 48-m reactor stack and 26 hours
of Freon emissions from a 29-m turbine stack. Exit temperatures were close to ambient (about
295K) with exit velocities of about 10 m/s for both the reactor stack (48.3 m) and the three
turbine stacks (29.1 m). These stacks were associated with 45-m and 28-m building tiers,
respectively. The monitoring data consisted of three arcs at 350, 800 and 1,500 m.
Meteorological data were available from an on-site tower at the 10-m and 43-m levels. There
was about an even split between stable and unstable hours, with mostly on-shore winds and
fairly high wind speeds. There were only 3 stable hours with wind speed less than 4 m/s, and the
majority were above about 7 m/s and several above 10 m/s. Figure 2 shows the layout of the
study area.

21
Figure 2. Millstone Study Area (SF6 and Freon Releases)

The Duane Arnold Energy Center (DAEC) is located in rural Iowa, located about 16 km
northwest of Cedar Rapids. It is located in a river valley with some bluffs on the east side.
Terrain varies by about 30 m across the receptor network with the eastern half of the
semicircular receptor arcs being flat and the western half elevated. The tracer study35 involved
SF6 releases from two rooftops (46-m and 24-m levels) and the ground (1-m level). Building
tiers for the rooftop releases were 43 and 24 m high, respectively. The 1-m and 24-m releases
were non-buoyant, non-momentum, while the 46-m release was close to ambient, but had about a
10 m/s exit velocity. The number of tracer release hours was 12, 16 and 11 from the release
heights of 46 m, 24 m, and 1 m, respectively. There were two arcs of monitors at downwind
distances of 300 and 1000 m (see Figure 3). Meteorological data consisted of winds at 10, 24,
and 50 m. The meteorological conditions were mostly convective (30 out of 39 hours), with
fairly light wind speeds. Only one hour had a wind speed above 4 m/s (4.6) , and almost half of

22
the hours were less than 2 m/s.

Figure 3. DAEC Study Area (SF6 Releases)

The Alaska North Slope tracer study36,37 (see Figure 4) involved 44 hours of buoyant SF6
releases from a 39-m high turbine stack. Tracer sampler coverage ranged over seven arcs from
50 to 3,000 m downwind. Meteorological data, including wind speed, wind direction,
temperature, sigma-theta, and sigma-w, were available from an on-site tower at the 33-m level.
Atmospheric stability and wind speed profiles were influenced by the smooth snow-covered
tundra surface with negligible levels of solar radiation in the autumn months. All experiments
(44 usable hours) were conducted during the abbreviated day light hours (0900 – 1600). Wind
speeds taken at the 33-m level during the tests were less than 6 m/s during one and part of
another test, between 6 and 15 m/s during four tests, and in excess of 15 m/s during three tests.
Stability conditions were generally neutral or slightly stable.

23
Figure 4. Depiction of Alaska North Slope Oil Gathering Center Turbine Stack, Meteorological
Tower (X), and Camera Locations Used to Visualize Plume Rise43

Independent Evaluation (Downwash)

Besides the full year of Bowline Point data (described above), the independent building
downwash evaluation databases consist of the American Gas Association (AGA) tracer
experiments, the Experimental Organic Cooled Reactor (EOCR) study, and the Lee Power Plant
wind tunnel study. Previous model evaluation results for these databases have been reported by
Paine and Lew38.

The AGA experiments39 occurred during spring and summer 1980 at gas compressor stations in
Texas and Kansas. At each test facility, one of the gas compressor stacks was retrofitted to
accommodate SF6 tracer gas emissions. In addition, stack height extensions were provided for

24
some of the experiments (with the normal stack height close to 10 m). The stack height to
building height ratios for the tests ranged from 0.95 to 2.52. There were a total of 63 tracer
releases over the course of the tests, and the tracer samplers were located between 50 and 200 m
away from the release point (see Figure 5). An instrumented 10-m tower was operated at both
experimental sites. The tracer releases were generally restricted to daytime hours. Stability
classes range from neutral to extremely unstable, except for three hours that were slightly stable.
Wind speeds range from 2 to 11 m/s over the 63 hours.

Figure 5. Plan View of the Locations of Tracer Samplers at Site 1, AGA Field Study
(SF6 Releases)

The EOCR study40 involved the simultaneous release of three tracer gases (SF6, F12, and Freon-
12B2) at three levels around the Experimental Organically Cooled Reactor (EOCR) test reactor
building at the Idaho National Engineering Laboratory in Southeast Idaho. The terrain was flat
with low-lying shrubs. The main building was 25 m high with an effective width of 25 m. The
tracer releases typically occurred simultaneously, and were conducted during 22 separate time
periods. Tracer sampler coverage was provided at eight concentric rings at distances of about
50, 100, 200, 400, 800, 1200, and 1600 m from the release points (see Figure 6). The stability
classes ranged from stable to unstable. The 10-m wind speeds for the cases selected ranged from
3 to 8 m/s.

25
 Figure 6. Terrain Map Featuring the Entire EOCR Grid with the Source at the Grid Center
(SF6 Releases).

Arcs are at distances of about 40, 80, 200, 400, 800, 1200, and 1600 m.

The Lee Power Plant wind tunnel study41 featured releases from steam boiler stacks with a
common height of 64.8 m, affected by a building tier with a height of 42.6 m. The world’s
largest fluid modeling study chamber at Monash University in Australia was used for these
experiments (see plan view in Figure 7). Stable conditions were simulated by using an inverted
model of the facility that was suspended from the ceiling of the tunnel. A stably stratified layer
was developed along the tunnel by heating the inflowing air, and a buoyant plume was simulated
by using a negatively buoyant gas mixture. A stable potential temperature lapse rate of
0.035 deK/m was modeled with a stack-top real-world equivalent wind speed of 7 m/s with
several wind directions being tested. In neutral conditions, stack-top speeds (at the 64.8-m level)

26
ranged in real-world equivalents from 5 to 40 m/s. There were 78 combinations of wind
direction, wind speed, and plume buoyancy tested for the neutral cases, and 14 combinations for
the stable cases. The tracer sampler coverage included ground-level concentrations at six
distances ranging from the cavity zone to beyond the wake (150-900 m). Since the actual Lee
Power Plant area is rural, the models were run with a rural source characterization. The EPRI
model evaluation38 considered both urban and rural source representations because of the
enhanced turbulence levels present in the wind tunnel. Consistent with the EPRI model
evaluation, the wind tunnel observations were adjusted from an assumed 5-minute duration to a
full hour using a time-dependent 1/5 power law42.

Figure 7. Plan View of the Lee Power Plant Model and Nearby Buildings Showing the Power
Station Units and the Zero Reference Position Used in the Monash Wind Tunnel Tests

Evaluation Results
A summary of the robust highest concentration prediction results for the downwash databases is
given in Table 3. Summaries of the results for each downwash database are provided below.

Table 3. Summary of AERMOD evaluation Results – downwash databases.

Database Ratio of Modeled/Observed Robust Highest Concentrations*

Bowline Point (buoyant, SO2) AERMOD: 1.14 (3-hr avg)
Hudson River Valley, New York ISC-PRIME: 1.23 (3-hr avg)

AERMOD: 1.43 (24-hr avg)

ISC-PRIME: 1.42 (24-hr avg)

27
Database Ratio of Modeled/Observed Robust Highest Concentrations*
AERMOD: 1.50 (annual avg)
ISC-PRIME: 1.35 (annual avg)
Alaska North Slope (buoyant, SF6) AERMOD: 1.06 (1-hr avg)
ISC-PRIME: 1.49 (1-hr avg)
Duane Arnold Energy Center AERMOD: 0.69 (1-hr avg; 46-m release)
(nonbuoyant, SF6) (Iowa) ISC-PRIME: 0.76 (1-hr avg; 46-m release)

AERMOD: 0.25 (1-hr avg; 24-m release)

ISC-PRIME: 0.29 (1-hr avg; 24-m release)

AERMOD: 0.51 (1-hr avg; 1-m release)

ISC-PRIME: 0.38 (1-hr avg; 1-m release)
Millstone Nuclear Power Plant AERMOD: 0.44 (1-hr avg; 46-m release)
(nonbuoyant, SF6) (Connecticut) ISC-PRIME: 0.41 (1-hr avg; 46-m release)

AERMOD: 1.32 (1-hr avg; 29-m release)

ISC-PRIME: 1.42 (1-hr avg; 29-m release)
American Gas Association (buoyant, AERMOD: 0.92 (1-hr avg)
SF6) ISC-PRIME: 0.76 (1-hr avg)
Texas and Kansas)
Experimental Organic Cooling AERMOD: 1.72 (1-hr avg)
Reactor (nonbuoyant, SF6) (Idaho) ISC-PRIME: 1.69 (1-hr avg)
Lee Power Plant (buoyant, wind AERMOD: 0.51 (1-hr avg; neutral cases; rural)
tunnel) ISC-PRIME: 0.49 (1-hr avg; neutral cases; rural)

AERMOD: 2.50 (1-hr avg; stable cases; rural)

ISC-PRIME: 2.11 (1-hr avg; stable cases; rural)

28
AERMOD and ISC-PRIME had a similar evaluation outcome for the full-year Bowline Point
database, featuring buoyant steam electric plant releases, with no significant differences in model
performance. The 3-hour Q-Q plot is shown in Figure 8, and the 24-hour Q-Q plot is shown in
Figure 9. For each averaging time, both models exhibit a modest overprediction tendency.

Figure 8. Q-Q Plot for Bowline Point 3-Hour Averages (SO2)

Bowline 3-hr Q-Q Plot (χ/Q) - 87m Stack

1000
Pred. Conc. (µg/m3)

100

10

AERMOD
ISC3P
1
1 10 100 1000
3
Obs. Conc. (µg/m )

Figure 9. Q-Q Plot for Bowline Point 24-Hour Averages (SO2)

Bowline 24-hr Q-Q Plot (χ/Q) - 87m Stack

1000
Pred. Conc. (µg/m3)

100

10

AERMOD
ISC3P
1
1 10 100 1000
Obs. Conc. (µg/m3)

29
Both models also had a similar evaluation outcome for two non-buoyant release heights (with
two different tracer gases) at the Millstone Nuclear Power Plant. Figure 10 shows that both
models overpredict by close to a factor of 2 for the 29-m Freon releases, but underpredict by
about a factor of 2 for the 46-m SF6 releases (see Figure 11).

Figure 10. Q-Q Plot for Millstone 1-Hour Averages for 29-m Releases (Freon)

Millstone Freon 1-hr Q-Q Plot (χ/Q) - 29m Stack

100
Pred. Conc. (µs/m3)

10

AERMOD
ISC3P

1
1 10 100
Obs. Conc. (µs/m3)

Figure 11. Q-Q Plot for Millstone 1-Hour Averages for 46-m Releases (SF6)

Millstone SF6 1-hr Q-Q Plot (χ/Q) - 48m Stack

100
Pred. Conc. (µs/m3)

10

AERMOD
ISC3P
1
1 10 100
3
Obs. Conc. (µs/m )

30
The Duane Arnold Energy Center also featured similar evaluation outcomes for AERMOD and
ISC-PRIME for the non-buoyant releases. Figure 12 shows a 1-m release Q-Q plot, in which
both models generally underpredict, but AERMOD has less of an underprediction tendency.
Both models underpredict for the 24-m releases (see Q-Q plot in Figure 13), with AERMOD
showing a larger underprediction tendency except for the highest concentrations. The Q-Q plot
in Figure 14 shows both models with peak concentrations that are nearly unbiased for the 46-m
releases.

Figure 12. Q-Qt for DAEC 1-Hour Averages for 1-m Releases (SF6)

DAEC SF6 1-hr Q-Q Plot (χ/Q) - 1m Stack

1000
Pred. Conc. (µs/m3)

100

10

AERMOD
ISC3PT
1
1 10 100 1000
Obs. Conc. (µs/m3)

Figure 13. Q-Q Plot for DAEC 1-Hour Averages for 24-m Releases (SF6)

DAEC SF6 1-hr Q-Q Plot (χ/Q) - 24m Stack

1000

100
Pred. Conc. (µs/m3)

10

1
AERMOD
ISC3P
0.1
1 10 100 1000
Obs. Conc. (µs/m3)

31
Figure 14. Q-Q Plot for DAEC 1-Hour Averages for 46-m Releases (SF6)

DAEC SF6 1-hr Q-Q Plot (χ/Q) - 46m Stack

1000

100
Pred. Conc. (µs/m3)

10

1
AERMOD
ISC3PT
0.1
0.1 1 10 100 1000
Obs. Conc. (µs/m3)

The modeling results for the buoyant releases at the Alaska North Slope experimental site are
shown as a Q-Q plot in Figure 15. The AERMOD predictions are nearly unbiased (but slightly
conservative), while the ISC-PRIME results are more than a factor of 2 high except for the
highest predictions. This result shows some of the biggest performance differences between
AERMOD and ISC-PRIME among all seven downwash evaluation databases.

Figure 15. Q-Q Plot for Alaska North Slope 1-Hour Averages for 39-m Releases (SF6)

Alaska North Slope 1-hr Q-Q Plot (χ/Q) - 39m Stack

10
Pred. Conc. (µs/m3)

0.1

AERMOD
ISC3P
0.01
0.01 0.1 1 10
Obs. Conc. (µs/m3)

32
The Q-Q plot for the AGA experiments is shown in Figure 16. This plot shows that the
AERMOD predictions parallel the 1:1 line over a larger concentration domain than the ISC-
PRIME predictions. For subsets of the modeling cases, AERMOD shows a significant
improvement especially for the cases involving stack-to-building height ratios greater than 1.25.
Residual plots of prediction concentrations as a function of stability indicate consistently less
biased AERMOD results over all stability types.

Figure 16. Q-Q Plot for AGA 1-Hour Averages (SF6)

AGA 1-hr Q-Q Plot (χ/Q)

1000
Pred. Conc. (µs/m3)

100

10

AERMOD
ISC3P
1
1 10 100 1000
3
Obs. Conc. (µs/m )

The Q-Q plot for the EOCR experiments is shown in Figure 17. Although the AERMOD
predictions parallel the 1:1 line over a larger portion of the concentration domain, the two curves
are nearly identical for the peak concentrations.

33
Figure 17. Q-Q Plot for EOCR 1-Hour Averages (SF6)

EOCR 1-hr Q-Q Plot (χ/Q)

10000

1000
Pred. Conc. (µs/m3)

100

10

1
AERMOD
ISC3P
0.1
0.1 1 10 100 1000 10000
Obs. Conc. (µs/m3)

The combined stable and neutral case Q-Q plot for the Lee power plant wind tunnel experiments
is shown in Figure 18. This figure shows underpredictions for the peak concentrations, in
contrast to the EPRI evaluation results38 with urban source characterization.

Figure 18. Q-Q Plot for Lee Wind Tunnel Data

Lee Wind Tunnel 1-hr Q-Q Plot (χ/Q)

100

10
Pred. Conc. (µs/m3)

0.1
AERMOD
ISC3P
0.01
0.01 0.1 1 10 100
3
Obs. Conc. (µs/m )

34
The overall model evaluation results for AERMOD version 02222 and ISC-PRIME with
downwashing databases can be summarized as follows, taking one composite (geometric mean)
ratio of predicted to observed RHC value for short-term averages at each site:

ƒ 0.97 is the overall predicted-to-observed ratio for short-term averages using AERMOD.

ƒ 0.94 is the overall predicted-to-observed ratio for short-term averages using ISC-PRIME.

CONCLUSIONS
Several enhancements to AERMOD have been made as a result of the public comments received
by EPA after the Notice of Proposed Rulemaking was issued on April 21, 2000. The most
notable change to AERMOD was the incorporation of the PRIME downwash algorithm. Other
model changes address complex terrain implementation issues, urban dispersion issues,
NADCON coordinate conversions, use of site-specific delta-T and wind speed data instead of
cloud cover, issues regarding terrain below stack base, ISCST3 updates, and various other
responses to beta test comments. The resulting beta test version is referred to as version 02222.

This AERMOD version has been re-run for the 10 non-downwash evaluation databases, with
minor differences in the results from those previously reported. The overall short-term ratio of
AERMOD version 02222 predicted/observed RHC concentrations is 1.03, averaged over the
non-downwash databases, with improvements over the ISCST3 results, especially for complex
terrain situations.

Seven downwash databases, divided into developmental and independent evaluation phases,
were used for the AERMOD evaluation. Comparisons with ISC-PRIME performance indicate
similar results for most databases, with occasional notable improvements, such as for the Alaska
North Slope database. The overall short-term ratio of AERMOD version 02222 predicted/
observed RHC concentrations is 0.97 averaged over the downwash databases. These results
were comparable in performance to those of ISC-PRIME, as expected.

REFERENCES
1. Weil, J.C. “Updating applied diffusion models,” J. Clim. and App. Meteor., 24(11): 1111-
1130. (1985)

2. Weil, J. C. Updating the ISC model through AERMIC. Preprints, 85th Annual Meeting of Air
and Waste Management Association, Air and Waste Management Association, Pittsburgh, PA.
(1992)

3. U. S. Environmental Protection Agency. User’s guide for the Industrial Source Complex
(ISC3) dispersion models. Volume II: Description of model algorithms. EPA-454/B-95-003b,
120 pp. [NTIS PB95-222758.] (1995)

4. United States Environmental Protection Agency. Compendium of

http://www.epa.gov/scram001/7thconf/aermod/dockrpt.pdf. Accessed March 2002.

35
5. Paine, R.J, R.F. Lee, R. Brode, R.B. Wilson, A.J. Cimorelli, S.G. Perry, J.C. Weil, A.
Venkatram, and W. Peters. Model Evaluation Results for AERMOD.
http://www.epa.gov/scram001/ (accessed December1998)

6. Paine. R., R. Lee, R. Brode, R. Wilson, A. Cimorelli, S. Perry, J.Weil, A. Venkatram, W.

Peters. AERMOD: Model Formulation and Evaluation Results. Preprints, 92th Annual Meeting
of Air and Waste Management Association, Air and Waste Management Association, Pittsburgh,
PA. (1999)

7. United States Environmental Protection Agency. Notice of Proposed Rulemaking (April 21,
2000). http://www.epa.gov/scram001/7thconf/information/nprappw.pdf. (accessed May 2000)

8. Schulman, L.L., D.G. Strimaitis, and J.S. Scire. Development and Evaluation of the PRIME
Plume Rise and Downwash Model. J. Air & Waste Management Assoc., 50, 378-390. (2000)

9. Brode, R. Implementation and Evaluation of PRIME in AERMOD. Paper J2.12 presented at

the 12th Joint Conference on Applications of Air Pollution Meteorology (with A&WMA),
Norfolk, VA. American Meteorological Society, Boston, MA (2002)

10. Cimorelli, A. J., S. G. Perry, A. Venkatram, J.C. Weil, R. J. Paine, R. B. Wilson, R. F. Lee,
W.D. Peters, R.W. Brode, and J. O. Paumier. AERMOD: Description of Model Formulation ,
http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf. (accessed December 2002)

11. Hanna, S.R. and R.J. Paine. Hybrid Plume Dispersion Model (HPDM) Development and
Evaluation. J. Appl. Meteor., 28, 206-224. (1989)

12. Paine, R.J. and B.A. Egan. User’s Guide to the Rough Terrain Diffusion Model (RTDM),
(Rev. 3.20). NTIS # PB88-171467/REB. U.S. Environmental Protection Agency, Research
Triangle Park, NC. (1987)

13. Perry, S.G., D.J. Burns, L.H. Adams, R.J. Paine, M.G. Dennis, M.T. Mills, D.J. Strimaitis,
R.J. Yamartino, and E.M. Insley. User's Guide to the Complex Terrain Dispersion Model Plus
Algorithms for Unstable Situations (CTDMPLUS) Volume I: Model Description and User
Instructions. EPA-600/8-89-041. U.S. Environmental Protection Agency, Research Triangle
Park, NC (NTIS PB89-181424). (1989)

14. Perry, S. G. CTDMPLUS: A dispersion model for sources near complex topography. Part I:
Technical formulations. J. Appl. Meteor., 31, 633–645. (1992)

36
15. Santos, L., R. I. Sykes, P. Karamchandani, C. Seigneur, F. Lurmann, R. Arndt, and N.
Kumar. Second-Order closure integrated puff (SCIPUFF) model with gas and aqueous phase
chemistry and aerosols - Paper 6.8. Preprints, 11th Joint Conference on the Applications of Air
Pollution Meteorology with the A&WMA, American Meteorological Society, Boston, MA.
(2000)

16. Carruthers, D.J., R.J. Holroyd, J.C.R. Hunt, W.S. Weng, A.G. Robins, D.D. Apsley, D.J.
Thomson, and F.B. Smith. UK-ADMS – a New Approach to Modelling Dispersion in the
Earth’s Atmospheric Boundary Layer. In Proceedings of the Workshop: Objectives for Next
Generation of Practical Short-Range Atmospheric Dispersion Models, May 6-8, 192, Riso,
Denmark, pp. 143-146. (1992)

17. Chambers, J. M., Cleveland, W. S., Kleiner, B., and Tukey, P. A. Chapter 3: Comparing
Data Distributions. Graphical Methods for Data Analysis. (Bell Laboratories). Wadsworth
International Group and Duxbury Press. (1983)

18. Weil J.C, Sykes and Venkatram A. Evaluating air-quality models: Review and outlook. J.
Appl. Met., 31, p 1121-1144. (1992)

19. Liu, M. K., and G. E. 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,
CA. (1984)

20. Venkatram, A., R. W. Brode, A. J. Cimorelli, J. T. Lee, R. J. Paine, S. G. Perry, W. D. Peters,

J. C. Weil, and R. B. Wilson. A complex terrain dispersion model for regulatory applications.
Atmos.Environ., 35, 4211-4221. (2001)

21. Cox, W. and Tikvart, J. A Statistical Procedure for Determining the Best Performing Air
Quality Simulation Model. Atmos. Envir., 24A, 2387-2395. (1990)

22. Barad, M. L., (Ed.). 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, 439 pp. (1958)

23. Haugen, D. A. (Editor). Project Prairie Grass, A field program in diffusion. Geophysical
Research Paper, No. 59, Vol. III. Report AFCRC-TR-58-235, Air Force Cambridge Research
Center, 439 pp. (1959)

24. Bowne, N. E., R. J. Londergan, D. R. Murray, and H. S. Borenstein. Overview, Results, and
Conclusions for the EPRI Plume Model Validation and Development Project: Plains Site. EPRI
Report EA-3074, Project 1616-1, Electric Power Research Institute, Palo Alto, CA, 234 pp.
(1983)

25. Murray, D. R., 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)

37
26. Paumier, J. O., S. G. Perry, and D. J. Burns. CTDMPLUS: A dispersion model for sources
near complex topography. Part II: Performance characteristics. J. Appl. Meteor., 31, 646–660.
(1992)

27. Hanna, S.R. and J.C. Chang. Hybrid Plume Dispersion Model (HPDM) Improvements and
Testing at Three Field Sites. Atmos. Envir., 27A, 1491-1508. (1993)

28. TRC. Evaluation of Rural Air Quality Simulation Models. EPA-450/4-83-003 (NTIS #
PB83-182758), prepared for Environmental Protection Agency, Research Triangle Park, NC.
(1982)

29. TRC. Air Quality Model Performance Evaluation and Comparison Study for Martins Creek
Steam Electric Station. TRC Project No. 14715-R61. TRC Environmental Corporation,
Windsor, CT. (1994)

30. Strimaitis, D. G., R. J. Paine, B. A. Egan and R. J. Yamartino. EPA Complex Terrain Model
Development: Final Report. Contract No. 68-02-3421, U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina. (1987)

31. DiCristofaro, D. C., D. G. Strimaitis, B. R. Green, R. J. Yamartino, A. Venkatram, D. A.

Gooden, T. F. Lavery and B. A. Egan. EPA Complex Terrain Model Development: Fifth
Milestone Report - 1985. EPA-600/3-85-069, U. S. Environmental Protection Agency, Research
Triangle Park, North Carolina. (1985)

32. Paine, R.J. Protocol for Evaluating Building Downwash Models. Prepared for the Electric
Power Research Institute Palo Alto, CA. ENSR Document Number 2460-026-430. (1995)

33. Schulman, L.L and S.R. Hanna. Evaluation of Downwash Modifications to the Industrial
Source Complex Model. JAPCA, 36: 258-264. (1986)

34. Bowers, J.F. and A.J. Anderson. An Evaluation Study for the Industrial Source Complex
(ISC) Dispersion Model. Report EPA-450/4-81-002. NTIS #PB81-176539. United States
Environmental Protection Agency Office of Air Quality Planning and Standards. (1981)

35. Thullier, R.H. and R.L. Mancuso. Building Effects on Effluent Dispersion from Roof Vents at
Nuclear Power Plants. EPRI Report NP-1380. Research Project 1073-1. Electric Power Research
Institute, Palo Alto, CA. (1980)

36. Guenther, A., B. Lamb, and E. Allwine. Building Wake Dispersion at an Arctic Industrial
Site: Field Tracer Observations and Plume Model Evaluations. Atm. Env., 24A: 2329-2347.
(1989)

37. Guenther, A. and B. Lamb. Atmospheric Dispersion in the Arctic: Winter-Time Boundary
Layer Measurements. Boundary-Layer Meteor., 49: 339-366. (1990)

38. Paine, R.J. and F. Lew. Results of the Independent Evaluation of ISCST3 and ISC-PRIME.
Prepared for EPRI under Contract WO3527-02. EPRI, Palo Alto, CA. (1997)

38
39. Engineering Science. Field Validation of Atmospheric Dispersion Models for Natural Gas
Compression Stations. Report No. PR-133, Prepared for the American Gas Association. (1980)

40. Start, G.E., N.F. Hukari, J.F. Sagendorf, J.H. Cate, and C.R. Dickson. EOCR Building Wake
Effects on Atmospheric Diffusion. NUREG/CR-1395. National Oceanic and Atmospheric
Administration, Idaho Falls, ID. (1981)

41. Melbourne, W.H. and J. Taylor. Wind Tunnel Studies of Plume Dispersion from the Lee
Power Plant Power Station. TR-135274: Draft Final Report. Prepared for Electric Power Research
Institute, Palo Alto, CA. (1994)

42. Turner, D.B. Workbook of Atmospheric Dispersion Estimates. PHS Publication No. 999
AP-26. U.S. Department of Health, Education, and Welfare, Public Health Service, Cincinnati,
OH (NTIS No. PB-191482). (1969)

KEYWORDS
AERMOD

ISCST3

Dispersion Model

Model Evaluation

PRIME

Building Downwash

39
 TECHNICAL REPORT DATA
(Please read Instructions on reverse before completing)

1. REPORT NO. 2.
3. RECIPIENT'S ACCESSION NO.

EPA-454/R-03-003

4. TITLE AND SUBTITLE

5. REPORT DATE June 2003
AERMOD: Latest Features and Evaluation Results

6. PERFORMING ORGANIZATION CODE

7. AUTHOR(S) 8. PERFORMING ORGANIZATION REPORT NO.

10. PROGRAM ELEMENT NO.

9. PERFORMING ORGANIZATION NAME AND ADDRESS See below. 11. CONTRACT/GRANT NO.

12. SPONSORING AGENCY NAME AND ADDRESS 13. TYPE OF REPORT AND PERIOD COVERED

U.S. Environmental Protection Agency Final Technical Report

Office of Air Quality Planning and Standards

14. SPONSORING AGENCY CODE

Emissions Monitoring and Analysis Division

EPA/200/04
Research Triangle Park, NC 27711
15. SUPPLEMENTARY NOTES Air and Waste Management Association paper No 698878 presented in June 2003.

16. ABSTRACT

AERMOD is an advanced plume model that incorporates updated treatments of the boundary layer theory,
understanding of turbulence and dispersion, and includes handling of terrain interactions. The model was
formally proposed by EPA in April 2000 as a replacement for the ISCST3 model. Several model
enhancements were made as a result of public comment, including the installation of the PRIME downwash
algorithm. The latest version of the model, version 02222, has been placed on EPA’s web site for beta test
purposes. This paper reviews the latest features and updated evaluation results for AERMOD version 02222.

17. KEY WORDS AND DOCUMENT ANALYSIS

a. DESCRIPTORS b. IDENTIFIERS/OPEN ENDED TERMS c. COSATI Field/Group

Model Evaluation, AERMOD, ISCST3, PRIME, Air Dispersion Model

building downwash

21. NOOF PAGES 40

18. DISTRIBUTION STATEMENT Release Unlimited 19. SECURITY CLASS (Report Unclassified

20. SECURITY CLASS (Page Unclassified 22. PRICE

EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION IS OBSOLETE

40
_______________________________________________________________________________________________________________________________________

United States Office of Air Quality Planning and Standards Publication No. EPA-454/R-03-003

Environmental Protection Emissions Monitoring and Analysis Division June 2003

Agency Research Triangle Park, NC 27711

_______________________________________________________________________________________________________________________________________

41