"This study presents a comparison of the pollutant concentration predictions from the
AERMOD and ISC air dispersion models in the context of
fugitive storage tank emissions at a bulk petroleum storage terminal."
Comparison of Two Dispersion Models_A Bulk Petroleum Storage Terminal Case Study - BREEZE AERMOD
1. Modeling Software for EH&S Professionals
Comparison of Two Dispersion Models:
A Bulk Petroleum Storage Terminal Case Study
Prepared By:
Anthony J. Schroeder
BREEZE SOFTWARE
12770 Merit Drive
Suite 900
Dallas, TX 75251
+1 (972) 661-8881
breeze-software.com
January 1, 2004
2. 1A
ABSTRACT
In 65 FR 21506 (dated April 21, 2000), the United States Environmental Protection
Agency (U.S. EPA) proposed revisions to its air dispersion modeling guidance, found in
40 CFR 51, Appendix W (“Guideline on Air Quality Models”) that included replacing the
Industrial Source Complex (ISC) model with the American Meteorological Society
(AMS) / EPA Regulatory Model (AERMOD) as the regulatory default model for state
and federal permitting applications. It is expected that when this switch is finally
promulgated, there will be an interim period during which results from either ISC or
AERMOD will be considered acceptable. For this reason, it is both interesting and useful
to explore differences in pollutant concentrations predicted by each model in a variety of
industrial contexts.
This study presents a comparison of the pollutant concentration predictions from the
AERMOD and ISC (ISCST3 and ISC-PRIME) air dispersion models in the context of
fugitive storage tank emissions at a bulk petroleum storage terminal. Data are presented
that shows that ISC consistently predicts higher overall and higher maximum pollutant
concentrations when compared with AERMOD in this particular situation. This trend is
most pronounced when using a volume source to simulate fugitive tank emissions and
least pronounced when using an area source.
Predicated concentrations can vary for different facility configurations, in regions of
differing terrain, and for different meteorological data sets. For this reason, this study
should be viewed as an example of one application of these two dispersion models and
not as a general treatment of predictions resulting from these models in all applications.
INTRODUCTION
In 65 FR 21506 (dated April 21, 2000),1
the United States Environmental Protection
Agency (U.S. EPA) proposed revisions to its air dispersion modeling guidance, found in
40 CFR 51, Appendix W (“Guideline on Air Quality Models”). One of the more far-
reaching revisions included is replacing the Industrial Source Complex (ISC) model with
the American Meteorological Society (AMS) / EPA Regulatory Model (AERMOD) as
the regulatory default model for state and federal permitting applications. It is expected
that when this switch is finally promulgated, there will be an interim period during which
results from either ISC or AERMOD will be considered acceptable. For this reason, it is
3. 2
both interesting and useful to explore differences in pollutant concentrations predicted by
the models.
The move in the regulatory community to replace ISC with a more advanced model was
driven by the fact that the standard short-term version of ISC (ISCST3) has several
theoretical deficiencies, including a poor characterization of building downwash and
terrain features. AERMOD (along with the Plume Rise Model Enhancement (PRIME)
algorithm) was developed in response to these deficiencies and is generally regarded as
yielding a more theoretically accurate treatment of atmospheric dispersion. This does not
necessarily mean that predicted concentrations from AERMOD will always be lower than
those predicted by ISC, however. Previous studies have shown that predicted
concentrations can vary for different source types and in regions of differing terrain
complexity.
At the same time that AERMOD was under development, refinements were made to ISC
in an attempt to improve its deficiency regarding building downwash through the
incorporation of the PRIME algorithm. Using ISC with the PRIME algorithm (referred
to collectively as ISC-PRIME) generally results in an improved representation of
building downwash effects.2
On April 15, 2003, the majority of the proposed provisions of 65 FR 21506 were
promulgated with the issuance of 68 FR 18449.3
The promulgation of AERMOD as the
new regulatory default dispersion model was not included in this document, however.
The general consensus of commenters on the promulgation of AERMOD was that a
version of the PRIME downwash algorithm should be included with the regulatory
default version of AERMOD in order to substantially differentiate it from ISC-PRIME.
For this reason, U.S. EPA decided to temporarily postpone the promulgation of
AERMOD.
On September 8, 2003, U.S. EPA released a Notice of Data Availability in the Federal
Register4
in which it was announced that two new papers were added to the public docket
that assess the performance of the most recent version of AERMOD, which includes the
PRIME downwash algorithm. Also in this Notice of Data Availability, U.S. EPA states,
“…it appears that the modified AERMOD is ready to be incorporated into the
Guideline…”. At the time of the writing of this paper, there is every indication that
AERMOD will be promulgated as the regulatory default short-range dispersion model
within a matter of months or weeks.
The following sections provide a brief overview of ISC and AERMOD, as well as the
PRIME downwash algorithm. Next, a case study comparison of ISC- and AERMOD-
predicted 1-hour, 24-hour, and annual average pollutant concentrations associated with
fugitive emissions from a large storage vessel at a bulk petroleum terminal will be
presented. Emissions from storage vessels generally come from a series of vents near the
top of the tank and around the top of its perimeter that allow product vapors to escape to
the atmosphere at ambient temperatures and with no initial vertical velocity. This type of
emission source may potentially be represented in both ISC and AERMOD in a number
4. 3
of ways. The first manner in which this source is to be represented consists of a single
emission point located at the top of the tank. The second consists of an area source that
covers the top portion of the tank. Finally, the emission source may be represented as a
volume source that is located near the top of the tank. The use of each of these source
types results in the employment of differing dispersion algorithms in both ISC and
AERMOD. Dispersion algorithms for the same source type also vary between the two
models. For this reason, this case study will explore the differences in concentrations
predicted by the two models using each of the source types listed above.
It should once again be noted that previous studies have shown that predicted
concentrations can vary for different source types and in regions of differing terrain
complexity. For this reason, this study should be viewed as an example of one
application of these two dispersion models and not as a general treatment of predictions
resulting from these models in all applications.
MODEL DESCRIPTIONS
ISCST3
ISCST3 is a steady-state Gaussian plume dispersion model with a minimum one-hour
time step that was developed specifically to support the U.S. EPA regulatory modeling
programs. The concept of steady-state essentially means that for each hour of the
modeled period, downwind concentrations are calculated as if the meteorological
conditions are the same throughout the entire domain and have been the same for the
entire hour. Due to its steady-state nature, ISCST3 is best used to predict pollutant
concentrations within 50 kilometers of point, area, and volume sources. ISCST3 has
been the workhorse of U.S. EPA regulatory models since it was first made available to
the public in final form in early August 1995.
In the following case study, modeling with ISCST3 is performed using the regulatory
default option, which includes stack heights (for point sources) adjusted for stack-tip
downwash, buoyancy-induced dispersion, and final plume rise. Ground-level
concentrations occurring during “calm” wind conditions are calculated by the model
using the calm processing feature. Regulatory default values for wind profile exponents
and vertical potential temperature gradients are used since no representative on-site
meteorological data are available. Rural dispersion coefficients are used in these cases.
For the point sources runs, the PRIME downwash algorithm is used. Downwash is not
calculated in the model algorithm used for area and volume sources, so the regulatory
default version of ISCST3 is used in these runs (i.e., no PRIME downwash).
Different algorithms are used by ISCST3 to compute atmospheric dispersion for the three
different types of sources. Because of this, different versions of the ISCST3 program are
run for each of the cases. In the point source runs, the version dated 01228, including the
PRIME algorithm, is used. In both the area and volume source runs, the version of
ISCST3 dated 02035 is used.
5. 4
Throughout the remainder of this paper, both ISCST3 and ISC-PRIME will be referenced
as ISC.
AERMOD
AERMOD is also a steady-state Gaussian plume dispersion model with a minimum one-
hour time step. It also has the ability to predict pollutant concentrations resulting from
point, area, and volume sources and, as with ISC, due to its steady-state nature, it is best
used to predict pollutant concentrations within 50 kilometers of the source.
As in the ISC runs, modeling with AERMOD is performed using the regulatory default
option in the following case study. The regulatory default option includes stack heights
(for point sources) adjusted for stack-tip downwash and the use of the calm processing
feature to predict ground-level concentrations during “calm” wind conditions.
A major difference between ISC and AERMOD is seen in the simulation of boundary
layer processes. Accurate simulation of boundary layer processes is important in
dispersion modeling because this is the region of the atmosphere where most mixing and
dispersion occurs. Whereas ISC uses relatively simple vertical profiles of wind and
temperature gradients, AERMOD’s treatment of the boundary layer is more complex
(and more realistic). Surface land-use information is used in the AERMET data
processor along with hourly meteorological data to produce more realistic profiles of
parameters that affect boundary layer dispersion.
As previously stated, the PRIME downwash algorithm has been incorporated into
AERMOD and is used to predict downwash for the point sources runs. Once again,
downwash is not calculated in the model algorithm used for area and volume sources, so
the version of AERMOD without PRIME downwash is used in these runs.
As with ISC, different versions of AERMOD are run for each of the source types. In the
point source runs, the version dated 03273 is used along with the PRIME algorithm. In
the area source runs, the version of AERMOD dated 03273 is used. Finally, in the
volume source runs, the version dated 03273 is also used.
More information and in-depth discussions on both ISC and AERMOD are available
from U.S. EPA.5, 6
MODEL INPUT DATA
Emission Source and Tank Data
Frequently, the effects of fugitive emissions from petroleum storage tanks are sought for
regulatory purposes outside of terminal fencelines. In this case study, the effects of
fugitive emissions of a generic pollutant (representing gasoline or any other petroleum
product) from one tank at a petroleum storage terminal are studied. The tank is assumed
to release one ton per year of fugitive emissions into the atmosphere.
6. 5
Table 1 summarizes the source parameters used in the modeling analyses for each
emissions source case.
Table 1. Source Parameters Used in the ISC and AERMOD Modeling Analyses.
* The diameter, velocity, and temperature are set to these values in order to simulate a release with little or no plume
rise.
The location of the tank from which emissions are being analyzed is depicted in Figure 1.
The relative location of the terminal fenceline, as well as the locations of other tanks
located at the terminal, are also depicted in Figure 1. Structures located within close
proximity to the emission source can contribute to downwash; for this reason, the
accurate placement and simulation of structures near the emission source can
dramatically affect modeled results.
Receptor Grids and Terrain
Three different grids of receptors are defined in order to provide a detailed mapping of
ground level, off-property concentrations in the areas immediately surrounding the
storage terminal. These grids cover a region extending one kilometer (km) from all edges
of the facility fenceline. The first grid (boundary) contains 25-meter (m) spaced
receptors along the fenceline. Next, a second grid (tight) contains 25-m spaced receptors
extending approximately 100 m from the fenceline. Finally, a third grid (fine) contains
100-m spaced receptors extending approximately 1.0 km from the fenceline. In many
regulatory cases, receptor grids must extend five or even 10 km from the facility
fenceline. In almost all cases, however, the highest modeled concentrations are located
within one km of the facility fenceline, so only this inner region is examined in this case
study. The locations of the receptor grids relative to the facility are shown in Figure 2.
CASE
RELEASE
HEIGHT
(FT)
STACK
DIAMETER
(M)
EXIT
VELOCITY
(M/S)
EXIT
TEMP.
(°F)
SOURCE
AREA
(FT
2
)
INITIAL
LATERAL
DISPERSION
COEFFICIENT
DIMENSION
(FT)
INITIAL
VERTICAL
DISPERSION
COEFFICIENT
DIMENSION
(FT)
Point
Source
Case*
24 0.001 0.001 Ambient -- -- --
Area
Source
Case
24 -- -- -- 491.1 -- --
Volume
Source
Case
24 -- -- -- -- 5.81 11.16
7. 6
The receptor, source, and tank elevations input to the models are extracted from USGS
1:24,000 scale (7.5-minute series) topographical maps of the area surrounding the
terminal. Elevations were determined electronically by processing Digital Elevation
8. 7
Model (DEM) files published by USGS through the National Geospatial Data
Clearinghouse. The elevations of the DEM points immediately surrounding each
receptor are examined, with the highest values conservatively selected to represent the
receptor, source, and tank elevations. In this case, simple terrain surrounds the terminal.
Meteorological Data
All runs are made with five consecutive years (1987-1991) of meteorological data from a
single measurement location. The meteorological data used in this study for both ISC
and AERMOD runs consists of hourly surface and 12-hourly upper air observations taken
at a nearby National Weather Service (NWS) site.
9. 8
The meteorological data used in AERMOD runs is further processed using the AERMET
preprocessor to incorporate land-use information into the AERMOD input files. The land
within three km to the northeast and east of the terminal is used primarily for agricultural
purposes. The land within three km to the south and west of the terminal is primarily
used for urban housing purposes. As discussed in the Model Descriptions Section, land-
use characteristics are taken into account for AERMOD to produce profiles of
meteorological parameters in the atmospheric boundary layer, but not for ISC.
ANALYSIS AND RESULTS
Point Source Case
Comparisons of ISC- and AERMOD-predicted concentrations for the point source case
using the 1-hour, 24-hour, and annual averaging periods are shown in Figures 3, 4, and 5.
Each data point plotted on these figures represents a comparison of ISC and AERMOD
maximum modeled concentrations for the specified averaging period at a particular
receptor. Each figure contains ISC versus AERMOD comparison points for each of the
receptor grids (boundary, tight, and fine) and each of the meteorological data years (1987
– 1991).
In the runs using an emission point source, ISC predictions are generally higher than
AERMOD predictions. This trend is especially evident in the 1-hour average cases
(Figure 3) and the 24-hour average cases (Figure 4) where the majority of the comparison
points fall to the right of the center line (the center line represents perfect agreement
between the two models). For the annual averaging period, there is more agreement
between ISC- and AERMOD-predicted concentrations. For the 1-hour and 24-hour
averaging periods, as modeled concentrations increase, ISC-predicted concentrations
become more consistently higher than AERMOD predictions. This point is particularly
interesting in light of the fact that the highest modeled concentrations are most important
in most regulatory modeling applications.
11. 10
The maximum concentrations (on any of the three grids and for any of the meteorological
data years) for each model and averaging period in the point source case are presented in
Table 2. This format presents a comparison of ISC and AERMOD results in a manner
that is very relevant to regulatory applications. As stated previously, in many regulatory
applications, the reviewing agency is only concerned with the highest predicted
concentration anywhere on the grid and for any of the meteorological data years. As
shown in the table, the maximum concentration predicted by ISC is higher than those
predicted by AERMOD for all three averaging periods. These observations agree with
those made using the comparison plots above.
12. 11
Table 2. Maximum Predicted Pollutant Concentrations for the Point Source Case.
AVERAGING
PERIOD MODEL GRID YEAR
MAXIMUM
CONCENTRATION
(µg/m3
)
1-Hour ISC Boundary 1987 275.80
1-Hour AERMOD Tight 1989 204.98
24-Hour ISC Tight 1990 44.63
24-Hour AERMOD Tight 1990 39.52
Annual ISC Boundary 1988 2.51
Annual AERMOD Boundary 1989 2.01
Area Source Case
Comparisons of ISC- and AERMOD-predicted concentrations for the area source case
using the 1-hour, 24-hour, and annual averaging periods are shown in Figures 6, 7, and 8.
These figures contain data for the same grids and meteorological data years as the point
source case.
The results in the area source case are consistent with those seen in the point source case.
Once again, ISC predictions are generally higher than AERMOD predictions, especially
in the 1-hour average cases (Figure 6) and the 24-hour average cases (Figure 7) where
13. 12
again, the majority of the comparison points fall to the right of the center line. For the
annual averaging period, there is more agreement between ISC- and AERMOD-predicted
concentrations. In the 24-hour and annual averaging period plots, the trend of consistent
higher predictions by ISC is less evident for higher predicted concentrations.
As was done in the point source case, the maximum concentration (on any of the three
grids and for any of the meteorological data years) for each model and averaging period
in the area source case are presented in Table 3. As shown in the table, the maximum
concentrations predicted by ISC are higher than those predicted by AERMOD for the 1-
and 24-hour averaging periods. In this case, however, the maximum annual average
concentration is higher in the AERMOD runs than in the ISC runs.
14. 13
These observations agree with those made for the area source case using the comparison
plots above. While Figures 6 – 8 indicate that AERMOD predicts lower pollutant
concentrations than ISC overall, the results in Table 3 indicate that where predictions of
concentration are highest (and most important to the regulatory community), there is little
difference in results between AERMOD and ISC for this area source case.
Table 3. Maximum Predicted Pollutant Concentrations for the Area Source Case.
AVERAGING
PERIOD MODEL GRID YEAR
MAXIMUM
CONCENTRATION
(µg/m3
)
1-Hour ISC Boundary 1990 73.50
1-Hour AERMOD Boundary 1990 71.45
24-Hour ISC Tight 1988 11.66
24-Hour AERMOD Boundary 1990 11.55
Annual ISC Boundary 1990 1.13
Annual AERMOD Boundary 1990 1.50
Volume Source Case
Comparisons of ISC- and AERMOD-predicted concentrations for the volume source case
using the 1-hour, 24-hour, and annual averaging periods are shown in Figures 9, 10, and
11. These figures contain data for the same grids and meteorological data years as in the
point source and area source cases.
The results in the volume source case are consistent with, but more pronounced than,
those seen in the point and area source cases. Once again, ISC predictions are generally
higher than AERMOD predictions, especially in the 1-hour average cases (Figure 9) and
the 24-hour average cases (Figure 10) where in this case, nearly all of the comparison
points fall to the right of the center line. For the annual averaging period, there is more
agreement between ISC- and AERMOD-predicted concentrations for lower concentration
predictions, but there is little agreement between the models for higher concentration
predictions. For the 1-hour and 24-hour averaging periods, as modeled concentrations
increase, ISC-predicted concentrations become more consistently higher than AERMOD
predictions, as was the trend in the point source case.
16. 15
As was done in the point and area source cases, the maximum concentrations (on any of
the grids and for any of the meteorological data years) for each model and averaging
period in the volume source case are presented in Table 4. As shown in the table, the
maximum concentrations predicted by ISC are considerably higher than those predicted
by AERMOD for all three averaging periods. These observations agree with those made
for the volume source case using the comparison plots above.
Table 4. Maximum Predicted Pollutant Concentrations for the Volume Source Case.
AVERAGING
PERIOD MODEL GRID YEAR
MAXIMUM
CONCENTRATION
(µg/m3
)
1-Hour ISC Boundary 1990 371.67
1-Hour AERMOD Boundary 1990 64.44
24-Hour ISC Boundary 1987 41.19
24-Hour AERMOD Tight 1990 12.83
Annual ISC Boundary 1990 3.62
Annual AERMOD Boundary 1990 1.22
17. 16
CONCLUSIONS
The recent proposed and pending updates to the federal guideline for air quality modeling
(40 CFR Part 51, Appendix W) include provisions through which AERMOD will likely
replace ISC as the regulatory default air dispersion model for U.S. EPA regulatory
purposes. It is important, therefore, to understand differences in predicted AERMOD and
ISC pollutant concentrations for a variety of industrial facility types and also a variety of
emission source types.
A comparison of the pollutant concentration predictions from the AERMOD and ISC air
dispersion models in the context of fugitive storage tank emissions at a bulk petroleum
storage terminal in simple terrain is presented here. Data resulting from this study show
that, in this context, ISC consistently predicts higher overall and higher maximum
pollutant concentrations when compared with AERMOD. This trend is most pronounced
using a volume source to simulate fugitive tank emissions and least pronounced using an
area source.
It should once again be noted that predicted concentrations could vary for different
facility configurations, in regions of differing terrain, and for different meteorological
data sets. For this reason, this study should be viewed as an example of one application
of these two dispersion models and not as a general treatment of predictions resulting
from these models in all applications.
ACKNOWLEDGMENTS
The author would like to thank Jeff DeToro, who served as a Trinity peer reviewer for
this work.
REFERENCES
1. Federal Register notice, 65 FR 21506, April 21, 2000.
2. U.S. EPA, ``Comparison of Regulatory Design Concentrations: AERMOD vs.
ISCST3, CTDMPLUS, ISD-PRIME.'' Office of Air Quality Planning and Standards,
Research Triangle Park, NC 27711; EPA Report No. EPA-454/R-03-002, July 2003.
3. Federal Register notice, 68 FR 18449, April 15, 2003.
4. Federal Register notice, 68 FR 52934, September 8, 2003.
5. U.S. EPA, “User’s Guide for the Industrial Source Complex (ISC3) Dispersion
Models”, Office of Air Quality Planning and Standards, Research Triangle Park, NC
27711, Report No. EPA-454/B-95-003a, September 1995.
18. 17
6. U.S. EPA, “AERMOD: Latest Features and Evaluation Results”, Office of Air Quality
Planning and Standards, Research Triangle Park, NC 27711, Report No. EPA-454/R-
03-003, July 2003.