This paper describes comparisons between AERMOD and ISCST3 for three industrial sites. These sites are a refinery, a gas compressor station, and a Portland Cement plant.

1. Modeling Software for EHS Professionals
Comparison between AERMOD and ISCST3 using
Data from Three Industrial Plants
Paper #536
Prepared By:
Russell F. Lee
Abby Goodman
Arron Heinerikson
BREEZE SOFTWARE
12700 Park Central Drive,
Suite 2100
Dallas, TX 75251
+1 (972) 661-8881
breeze-software.com

2. 1
Comparison between AERMOD and ISCST3 using Data
from Three Industrial Plants
99-your paper number536
Russell F. Lee
MeteorologistConsultant, 17 Cobbleridge Court, Durham, NC 27713-9493
Abby Goodman
Trinity Consultants, 9401 Indian Creek Parkway, Suite 970, Overland Park, KS 66210
Arron Heinerikson
Trinity Consultants, 9401 Indian Creek Parkway, Suite 970, Overland Park, KS 66210
ABSTRACT
In 1991, the American Meteorological Society (AMS) and the United States Environmental
Protection Agency (EPA) initiated a formal collaboration for the purpose of introducing recent
scientific advances into applied dispersion models. A working group (AMS/EPA Regulatory
Model Improvement Committee, AERMIC) of three AMS scientists and four EPA
meteorologists was formed to facilitate this collaborative effort. The result of the effort is
AERMOD (AERMIC Model). The U.S. Environmental Protection Agency is expected to
propose AERMOD as a replacement for ISCST3 for most regulatory modeling applications.
It is of considerable interest to those industries being regulated to learn, not only how accurate
AERMOD is, but also how AERMOD compares with the current regulatory model, ISCST3, for
the complexities of real industrial sites. What will be the impact of changing from ISCST3 to
AERMOD for these industrial sites? This paper describes comparisons between AERMOD and
ISCST3 for three industrial sites. These sites are a refinery, a gas compressor station, and a
Portland Cement plant.
The results of the comparison show that AERMOD may calculate either higher or lower than
ISCST3. AERMOD and ISCST3 results may be fairly similar for some modeling scenarios, and
quite different for others. Differences between the models exceeded a factor of two in one case in
this study.
INTRODUCTION
AERMOD is a new, advanced plume dispersion model that the U.S. Environmental Protection
Agency intends to propose for regulatory use. AERMOD is intended to replace ISCST3 for most
modeling applications. AERMOD is the result of an effort to incorporate scientific knowledge
gained over the last three decades into regulatory plume models. Comparisons made with field
studies, some using tracer data, indicate significant improvements in the ability of the model to
represent measured concentrations1
.

3. 2
It is of considerable interest to those industries being regulated to learn not only how accurate
AERMOD is, but also how AERMOD compares with the current regulatory model, ISCST3, for
the complexities of real industrial sites. What will be the impact of changing from ISCST3 to
AERMOD for these industrial sites? This paper describes comparisons between AERMOD and
ISCST3 for three industrial sites. These sites are a refinery, and a gas compressor station,. and a
Portland Cement plant.
MODEL DESCRIPTIONS
The Industrial Source Complex Short-term Model (ISCST3) is the current model recommended
by the EPA for estimating near-field (less than 50 kilometers) air quality impacts from most
industrial source types for regulatory applications. ISCST3 is the most recent version of the
ISCST model, which was first released in the mid-1970s. This model is a Gaussian plume model
whose underlying dispersion theory is essentially that of D.B. Turner2
, based on the work of
F. Pasquill3
and F.A. Gifford4
.
[Russ]AERMOD is a new air quality model whose development was initiated under the joint
sponsorship of the American Meteorological Society (AMS) and the U.S. Environmental
Protection Agency. AERMOD is designed for estimating near-field impacts from a variety of
industrial source types. As such, it is being proposed as a replacement for ISCST3. The
formulation of AERMOD is described in detail in the draft U.S. EPA report AERMOD–
Description of Model Formulation.5
There are several important differences between ISCST3 and AERMOD. In ISCST3, the
atmosphere is characterized as one of six stability classes, based on wind speed and an estimate
of the net radiation arrived at from cloud cover observations and (in daytime) solar elevation
angle. The plume is assumed to have specific horizontal and vertical dimensions that are a
function of only stability class and downwind distance. The plume is assumed to behave the
same regardless of height. AERMOD, on the other hand, makes use of two continuous stability
parameters, the Monin-Obukhov length, which indicates the relative strengths of mechanical and
buoyant effects on turbulence, the friction velocity, which is a measure of mechanical effects
alone, i.e., wind shear at ground level. The effects of albedo (surface reflectivity of sunlight),
Bowen ratio (an indication of the moisture available for evaporation), and surface roughness on
atmospheric turbulence is accounted for by AERMOD. The Pasquill dispersion curves, upon
which ISCST3 is based, are for a “sampling time of about 3 minutes” and are “based on wind
fluctuation statistics over open downland6
” (i.e, rolling grassland). AERMOD predicts 1-hour
average concentrations by design. In AERMOD, modern planetary boundary layer theory is used
to scale turbulence and other parameters to the height of the plume, while ISCST3 scales only
wind speed to stack height. Furthermore, AERMOD allows for a non-Gaussian probability
density function (pdf) to characterize the plume, while ISCST3 uses the Gaussian pdf for all
cases. The two models differ, also, in the handling of mixing heights. The ISCST3 system
(specifically, the meteorological preprocessor, PCRAMMET) accepts a single afternoon mixing
height and linearly interpolates hourly values, assuming the mixing height at sunrise is zero (for
rural cases). The AERMOD system (specifically, the AERMET meteorological preprocessor)
derives hourly mixing heights based on the morning upper air sounding and the surface
meteorology.

4. 3
DATA BASES
Three industrial sites representative of typical refineries, gas compressor stations, and Portland
cement plants are included in the analysis. Each facility is evaluated for one regulated air
pollutant for the averaging periods corresponding to the National Ambient Air Quality Standards
(NAAQS). All modeling utilizes four to five years of surface measurements and upper air data
from representative weather stations.
Refinery
The refinery being modeled in this analysis contains approximately 20 process heaters and 2
boilers, with stack heights ranging from 13 to 41 meters. Most of the process heaters are
primarily fired on refinery fuel gas with natural gas backup. The boilers are dual fueled with
refinery fuel gas and fuel oil.
As with all refineries, there are structures near the emission sources that result in significant
downwash effects. These structures include pipe racks, tanks, and process equipment like
distillation towers. The vicinity of each source has been evaluated to include all structures that
could result in downwash effects.
The refinery, located in a rural, semi-desert area in the south central U.S., was evaluated
assuming flat terrain, with ground cover typical of “desert shrubland.” Seasonal values of albedo,
Bowen ratio, and surface roughness required by AERMOD were obtained using the tabulated
values from the AERMET User’s Guide7
, for “desert shrubland,” assuming average moisture
conditions for this kind of area. An albedo of 0.45 was used for winter, 0.30 for spring, and 0.28
for summer and autumn. Bowen ratios used were 6.0 for winter and autumn, 3.0 for spring, and
4.0 for summer. Surface roughness lengths used were 0.15 for winter, and 0.30 for spring,
summer, and autumn. Cartesian receptor grids of several densities were used to calculate the
ground level concentrations of SO2 from the refinery. The grid contained 100 meter spacing for
receptors outside of the property boundary out to 1 kilometer. Beyond 1 kilometer, the receptor
spacing was 500 meters. Discrete receptors were placed at 50-meter intervals along property
boundaries.
Meteorological surface and upper air data for the model runs were from the nearest National
Weather Service (NWS) station for the period 1989–1992. It was originally intended to include
the year 1993 to cover a full 5-year period. However, the 1993 data did not include cloud cover,
which is needed by both ISCST3 and AERMOD.
Ground level concentrations of SO2 were calculated for each of the four years for the high-
second high 3-hr average, high-second high 24-hr average, and highest annual average. This
corresponds to the averaging times used for the National Ambient Air Quality Standard for SO2.
Natural Gas Compressor Station
The natural gas compressor station evaluated consists of a dehydrator and three natural gas
compressors.. The stacks were each 12.8 meters in height. The effect of nearby 9-meter tall
buildings on downwash was included in the analysis.
The pipeline compressor station is located in a rural area, in flat terrain in the central U.S., with

5. 4
surrounding ground cover characterized as “grassland.” Seasonal values of albedo, Bowen ratio,
and surface roughness required by AERMOD were obtained using values recommended in the
AERMET User’s Guide7
, for “grassland,” assuming average moisture conditions for this kind of
area. An albedo of 0.60 was used for winter, 0.18 for spring and summer, and 0.20 for autumn.
Bowen ratios used were 1.5 for winter, 0.4 for spring, 0.8 for summer, and 1.0 for autumn.
Surface roughness lengths used were 0.5 for winter, 0.3 for spring, 0.4 for summer, and 0.5 for
autumn. A 2-kilometer by 2-kilometer grid with 50-meter spacing was used to calculate the
annual ground level concentrations of NOx from the compressor station.
Meteorological surface and upper air data for the model runs were from the nearest National
Weather Service (NWS) station for the period 1987–1991.
Ground level annual average concentrations of NOx were calculated for each of the five years.
The National Ambient Air Quality Standard for NOx is based on the highest of the maximum
annual average concentration for each of the five years.
Portland Cement Plant
The Portland cement plant being modeled contains approximately 35 dust collectors (bagfilters),
a preheater/precalciner kiln system, two diesel fired generators, and various fugitive dust sources
including conveyor transfer points, wind erosion from storage piles, and vehicle activity on
unpaved roads.
As with most cement plants, there are structures near the emission sources that result in
downwash effects. These structures include a preheater/precalciner tower, raw material, clinker,
and cement storage silos, and numerous processing buildings. All structures at the site have been
included to ensure that downwash effects are appropriately accounted for.
The cement plant is located in a rural area, in hilly terrain in central U.S., with surrounding
ground cover characterized as “grassland.” Seasonal values of albedo, Bowen ratio, and surface
roughness required by AERMOD were obtained using values recommended in the AERMET
User’s Guide7
, for “grassland,” assuming average moisture conditions for this kind of area. An
albedo of 0.60 was used for winter, 0.18 for spring and summer, and 0.20 for autumn. Bowen
ratios used were 1.5 for winter, 0.4 for spring, 0.8 for summer, and 1.0 for autumn. Surface
roughness lengths used were 0.5 for winter, 0.3 for spring, 0.4 for summer, and 0.5 for autumn.
A Cartesian receptor grid, with terrain height specified at each grid point, was used to calculate
the ground level concentrations of PM10. The grid utilized 100 meter spacing for receptors
outside of the property boundary out to 3 kilometers. Discrete receptors were placed at 50-meter
intervals along property boundaries.
Meteorological surface and upper air data for the model runs were from the nearest National
Weather Service (NWS) station for the period 1990–1994.
Ground level particulate (PM-10) concentrations were
[Abby and Arron]calculated for the average of each year’s high-fourth high (H4H) 24-hour
concentration and for the average of the highest annual average concentration for each year, as
corresponding to the national PM-10 standards.

6. 5
ANALYSIS AND RESULTS
[Russ]This study compares modeled results, using ISCST3 and AERMOD, for a refinery, a
natural gas compressor station, and a cement plant. These sources are modeled, with minor
variations, as they would be for a permit application, as described above.
The model results for the refinery show AERMOD and ISCST3 predictions for the 3-hour high-
second high to be fairly closely matched. AERMOD predictions ranged from 11% below those
of ISCST3 to 13% above (Figure 1). ISCST3 predicted the highest 3-hour high-second high
concentration to occur in 1992, while AERMOD expected the highest concentration to be in
1991. For these concentrations, AERMOD predicted 6% lower than ISCST3. In both models,
this value was clearly dominated by a nearby flare. Both models predicted the 3-hour high-
second high values to occur just north of a flare, when winds were from the south. ISCST3
predicted this value to occur from 200 meters to 500 meters north of the source, depending on
the year, while AERMOD consistently predicted the high-second high to occur 200 meters north
of the flare for all years. AERMOD was provided values of albedo, Bowen ratio, and roughness
length appropriate for desert shrubland. This implies a higher albedo and Bowen ratio than is the
case for the grassland experiments on which ISCST3 dispersion formulation was based. A higher
albedo means that less heat is retained from the sunlight and, therefore, less heat is available to
develop turbulence. This is compensated for by a higher Bowen ratio (meaning drier land surface
and less heat lost to evaporation) and the plant’s more southerly location (relative to the latitude
of the experiments on which ISCST3 was based) makes more the heat is available for turbulence.
Similarly, the results of the 24-hour high second-high calculations shows AERMOD results
ranging from 10% lower to 29% higher than those of ISCST3 (Figure 2). In this case, ISCST3
expected the highest concentrations to occur in 1991 and AERMOD in 1992. AERMOD
predicted only 2% higher than ISCST3 for the highest of the 24-hour values. Although the
concentration values are not much different, AERMOD places the location about 300 meters
north of the flare, while ISCST3 places it about 900 meters to the north.
For the annual average case, AERMOD predicts the highest concentrations ranging from 34% to
48% higher than ISCST3 (Figure 3). Both models predicted the highest concentrations in 1990,
with AERMOD predicting 34% higher than ISCST3 for that case. Interestingly, ISCST3 also
predicted the highest annual value to occur 900 meters north of the flare, while AERMOD placed
the peak at 400 meters north of the flare. The fact that AERMOD predicts higher than ISCST3
for the annual average is not surprising. The authors have noted in several model evaluation
studies involving Gaussian models, a common, but not universal, tendency for Gaussian models
to underpredict long-term averages. In model evaluations conducted with AERMOD, this
problem does not seem to occur with AERMOD.
The results for the natural gas compressor station show AERMOD predicting annual average
concentrations from a factor of 2.2 to a factor of 3.2 higher than ISCST3 (Figure 4). Both models
predicted the highest annual concentration to occur in 1990, with AERMOD predicting a factor
of 2.2 higher than ISCST3 for that case. AERMOD places the location of the maximum rather
close to the stacks, about 128 meters to the north, while ISCST3 places the maximum 214 meters
to the north. These computations are almost certainly affected by building downwash effects.
AERMOD currently uses the downwash algorithm that is contained in ISCST3. However, that

7. 6
does not mean it will behave similarly in both models because of the differences in the way
dispersion is treated. The albedo, Bowen ratio, and surface roughness lengths used in the
AERMOD runs are those appropriate for grassland. This is similar to the ground cover for the
experiments on which the ISCST3 dispersion formulation were based. It would be a useful future
exercise to determine 1) whether the high ratio between AERMOD and ISCST3 results would
carry over to the 3-hour and 24-hour averaging times, and 2) whether the large ratio is in any
way related to the building downwash treatment.
The results for the cement plant show comparable predictions between the two models for the
regulatory high-fourth high concentration, with AERMOD predicting about 21% higher than
ISCST3. However, AERMOD places the location of the highest values near the loading silo and
dust collectors, while ISCST3 places the location between the quarry and a haul road. The
cement plant was also modeled using albedo, Bowen ratio, and surface roughness lengths
appropriate to grassland, and for a location in the mid-latitudes, comparable to the ground cover
and latitude of data on which the ISCST3 dispersion algorithms were based.
The results for the 5-year averaged annual averages from the two models are nearly identical in
value, but much different in location. As with the high-fourth highest 24-hour concentrations
noted above, AERMOD located the highest values near the loading silo and dust collectors
(point and volume sources), while ISCST3 placed the highest values between the quarry and a
haul road (area sources). Although the treatment of terrain was included in this modeling study,
it did not prove to be a significant factor in the cement plant analysis, since the maximum
concentrations were all adjacent to plant, and at about the same elevation.
CONCLUSIONS
These three cases give examples of the kinds of differences that can occur between the
predictions of AERMOD and those of ISCST3 in a regulatory application. AERMOD may
predict higher or lower than ISCST3 for any given modeling scenario. In the case of the cement
plant used in this study, the differences exceeded a factor of two. This study illustrates that in
some cases, at a complex industrial site, the maximum concentrations calculated by the two
models may be dominated by the same source, but located at a slightly different distance from
the source. In other cases, the maximum concentrations may be dominated by an entirely
different source, as is the case with the cement plant in this study. AERMOD may give
substantially different results than ISCST3 for a given modeling scenario; however, one cannot,
a priori, predict whether AERMOD will give higher or lower concentrations, or even, in some
cases, whether the same source will be identified as providing the greatest contribution to the
maximum concentration.
[Russ]
ACKNOWLEDGEMENTS
[Abby and Arron—we could acknowledge data sources here, if the sources want to be
acknowledged]
REFERENCES
Formatted

8. 7
1. Paine, R.J.; Lee, R.F.; Brode, R.; Wilson, R.B; Cimorelli, A.J.; Perry, S.G.; Weil, J.C.;
Venkatram, A.; Peters, W.D. Model Evaluation Results For AERMOD, draft document, U.S.
Environmental Protection Agency, 1998; Available from U.S. EPA Web Site
http://www.epa.gov/scram001/.
2. Turner, D.B. Workbook of Atmospheric Dispersion Estimates; U.S. Public Health Service,
Cincinnati, Ohio, 1967; PHS Publ. No. 999 AP-26.
3. Pasquill, F. The Estimation of the Dispersion of Windborne Material, Meteor. Mag. 1961,
90(1063), 33-49.
4. Gifford, F.A., Jr Atmospheric Dispersion Calculations using the Generalized Gaussian Plume
Model, Nuclear Safety, 1960, 2(2), 56-59, 67-68.
2.5.Cimorelli, A.J.; Perry, S.G.; Venkatram, A; Weil, J.C.; Paine, R.J.; Wilson, R.B.; Lee, R.F.;
Peters, W.D. AERMOD—Description of Model Formulation; draft document, U.S.
Environmental Protection Agency, 1998; Available from U.S. EPA Web Site
http://www.epa.gov/scram001/.
3.6.Pasquill, F. Atmospheric Diffusion, 2nd
Edition; Ellis Horwood Limited, Chichester, England,
1974, pages 367-368.
4.7.U.S. EPA. Revised Draft User’s Guide for the AERMOD Meteorological Preprocessor
(AERMET), U.S. Environmental Protection Agency, 1998; Available from U.S. EPA Web
Site http://www.epa.gov/scram001/.

9. 8
FIGURES
Figure 1. A comparison of ISCST3 and AERMOD calculations of high second-high 3-hourly
concentrations at a refinery.
Figure 2. A comparison of ISCST3 and AERMOD calculations of high second-high 24-hourly
concentrations at a refinery.
0
50
100
150
200
250
300
350
400
450
500
1989 1990 1991 1992 Maximum
Year
HighSecond-High3-hourConcentration(g/m
3
)
ISCST3
AERMOD
0
20
40
60
80
100
120
140
160
1989 1990 1991 1992 Maximum
Year
HighSecond-High24-hourConcentration(g/m
3
)
ISCST3
AERMOD

10. 9
Figure 3. A comparison of ISCST3 and AERMOD calculations of annual average
concentrations at a refinery.
Figure 4. A comparison of ISCST3 and AERMOD calculations of annual average
concentrations at a natural gas compressor station.
0
5
10
15
20
25
30
35
40
1987 1988 1989 1990 1991 Maximum
Year
AnnualAverageConcentration(g/m
3
)
ISCST3
AERMOD
0
5
10
15
20
25
30
35
40
45
50
1989 1990 1991 1992 Maximum
Year
AnnualAverageConcentration(mg/m
3
)
ISCST3
AERMOD

11. 10
Figure 5. A comparison of ISCST3 and AERMOD calculations of high-fourth high 24-hour
concentrations, averaged over five years, and the annual average concentrations, averaged over
five years (corresponding to the new PM-10 standards), at a cement plant.
0
5
10
15
20
25
30
35
40
24-hr H4H Annual
Averaging Time
5-YearAverageoftheValuesforEachYear
ISCST3
AERMOD