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.

1. Environmental solutions delivered uncommonly well
Sensitivity of AERMOD to AERMINUTE-
Generated Meteorology
Paper No 2012-A-421-AWMA
Prepared By:
George J. Schewe, CCM, QEP – Principal Consultant
Abhishek Bhat, PhD – Consultant
TRINITY CONSULTANTS
1717 Dixie Highway
Suite 900
Covington, KY 41011
(859) 341-8100
trinityconsultants.com
June 19, 2012

2. 2
ABSTRACT
Recent modeling studies in support of regulatory permitting, SO2 nonattainment modeling,
and other AERMOD Model applications have switched to the use of AERMINUTE-enhanced
meteorological data sets. Surface meteorological data collected by the National Weather
Service (NWS) are often used as the source of meteorological data for AERMOD. Over the
past several years the use of NWS data resulted in a high incidence of calms and variable
wind conditions as reported for the Automated Surface Observing Stations (ASOS) now in
use at most NWS stations since the mid-1990’s. In the coding used to report surface
observations beginning July 1996, a calm wind is defined as a wind speed less than 2 knots
and is assigned a value of 0 knots. The ASOS system also truncates values from 2.1 to 2.9
knots to 2 knots and thus, all values 2.9 or lower are set to a calm. AERMOD currently
cannot simulate dispersion under calm or missing wind conditions. To reduce the number
of calms and missing winds in the surface data, archived 1-minute winds for the ASOS
stations may be used to calculate hourly average wind speed and directions. EPA released
a processor due perform these calculations, namely, AERMINUTE (latest version is 11325).
These AERMINUTE generated wind speeds and wind directions may be used to supplement
the standard NWS archive of hourly observed winds processed in AERMET. This paper
presents the results of applying meteorological data sets with and without the
supplemental AERMINUTE data. Air concentrations for various source types are modeled
representative if an industrial facility point, area, and volume sources representing stacks,
transfer points, storage piles and roads. Meteorology from various U.S. regions is used
along with the land use characteristics for each airport. Comparisons of the concentrations
for various source types and meteorological data sets are made.
INTRODUCTION
The AERMOD Model1 was introduced to the regulatory dispersion modeling community in
the late 1990s. AERMOD was developed specifically by the AMS/EPA Regulatory Model
Improvement Committee (AERMIC) to employ best state-of-practice parameterizations for
characterizing the meteorological influences on dispersion. Section 4.2.2.b of the Guideline
on Air Quality Models (GAQM), Appendix W, 40 CFR Part 512 states that AERMOD is the
recommended model for “a wide range of regulatory applications in all types of terrain”
thus, AERMOD is the primary refined analytical technique for modeling traditional
stationary sources. Provided along with the AERMOD Model are a number of preprocessors
for preparing data sets applicable to running the AERMOD algorithms for transport,
dispersion, convective boundary layer turbulence, stable boundary layer, terrain
influences, building downwash, and land use. These are AERMAP, AERSURFACE, and
AERMET. AERMAP is used to process elevation data from digitized data sets to generate
elevations of receptors, sources, and structures as well the critical height for each receptor.
AERSURFACE3 uses land use land cover (LULC) data to calculate albedo (reflectivity of the
earth’s surface), Bowen Ratio (ratio of sensible to latent heat), and the surface roughness
parameter (related to the height of obstructions but more of a measure of the height above
ground where the wind speed approaches zero) which can vary on an annual, seasonal, or
monthly basis for one or up to twelve sectors around a site. AERMET4 is the meteorological

3. 3
data processor that uses a combination of surface observation data from the National
Weather Service (NWS), upper air data from NWS, onsite data if available and meeting
prescribed collection and quality assurance criteria, and albedo, Bowen Ratio, and surface
roughness parameters from AERSURFACE.
Prior to about 1995, surface data measured and archived by the NWS used a threshold
velocity for the wind instruments of 2 knots (about 1 m/s). EPA adjusted for this lower
limit of wind speed by making any value below 2 knots equal to a calm in AERMET. The
NWS still uses 2 knots as the threshold velocity but also truncates up to 2.9 knots to 2 thus,
making a wider range of calms. This has been noted in recent data sets where periods of
calm up to 15-20% are common. Thus, true low wind speeds are not being considered and
the number of hours in the data set is much reduced.
Rather than accepting the archived NWS wind speed and direction as the best
representation of each hour, the AERMINUTE5 program reprocesses these 1-minute
readings to a lower threshold and does not truncate the values. These 1-minute values are
then averaged for all values that are considered valid readings resulting in a one-hour wind
speed and wind direction.
While it is certainly beneficial to have these values filled in and have fewer calm periods, it
is also of concern that potentially low wind speeds will be used in AERMOD. Of concern is
the fact that AERMOD has not been shown to perform well in low wind speed conditions
and the number of these conditions has increased. This may become more apparent in
terms of the ambient air concentrations generated with AERMINUTE data sets and
compared to AERMET without AERMINUTE. For the remainder of this paper the following
will apply:
AERMET will refer to data sets based on using straight NWS ISHD (integrated
surface hourly data format) TD3505 data for the surface meteorology as available
from the National Climatic Data Center (NCDC) along with upper air soundings in
the FSL format as obtained from the National Oceanic and Atmospheric
Administration website http://www.esrl.noaa.gov/raobs/.
AERMINUTE/AERMET will refer to data sets based on using 1-minute running 2-
minute average winds from NOAA at ftp://ftp.ncdc.noaa.gov/pub/data/asos-
onemin/ along with NWS ISHD (integrated surface hourly data format) TD3505 data
for the surface meteorology as available from NCDC along with upper air soundings
in the FSL format as obtained from the NOAA website
http://www.esrl.noaa.gov/raobs/.
METHODOLOGY
Study locations were defined in several areas of the U.S. including Gainseville, Florida,
Orangeburg, South Carolina, Harrisburg, Pennsylvania, Fargo, North Dakota, and Cape
Girardeau, Missouri. The diversity of these locations insured that no one location with its
specific climatological characteristics would dominate the analysis or influence the results.

4. 4
The year of data selected was 2006 which gave a mix of whether the ice free winds (IFW)
instruments were being used or not (different sites have changed over to sonic
anemometers at different times). The specific sites used in the comparisons in AERMET
are shown in Table 1.
Table 1. Sites Used AERMET Vs. AERMINUTE/AERMET Comparisons
Surface/Upper Air Sites Surface Data
Ice Free Winds
Start Date
Upper Air Data
Harrisburg, Dulles KMDT, NWS 14711, ISHD August 22, 2008 KIAD, NWS 93734, FSL
Cape Girardeau, Springfield KCGI, NWS 03935,
3280VB
December 16, 2006 KSGF, NWS 13995,
6201FB
Fargo, Aberdeen KFAR, NWS 27530,
CD144
September 26, 2006 KABR, NWS 14929, FSL
Orangeburg, Charleston KOGB, NWS 53854, ISHD April 29, 2009 KCHS, NWS 13880, FSL
Gainesville, Jacksonville KGNV, NWS 12816, ISHD March 9, 2007 KJAX, NWS 13889, FSL
After these sites were selected, both the standard NWS suface data and upper air files were
obtained. Some surface data sets were based on older formats CD144 and 3201VB while
the remainder was in the ISHD format (TD3505-Full). Upper air data was primarily in the
FSL format with one file in the 6201FB format. AERMET can and does read and process
any of the above formats. The latest version (11059) of AERMET was used to process these
data sets into AERMET output format suitable for processing in AERMOD.
For the AERMINUTE/AERMET processing the latest version of AERMINUTE (dated 11325)
was used to process the 1-minute data as input to AERMET for each of the meteorological
stations processed for use in AERMOD. The 1-minute wind data was obtained from the
NCDC’s online ftp directory (website above) in the TD6405 format which is compatible
with the AERMINUTE program. The downloaded data consists of text files; each text file
contains data for one station-month. The 1-minute wind data consists of running sequential
2-minute average winds that are reported every minute at each ASOS station. The archived
1-minute winds contained in the downloaded text files were used by AERMINUTE to
calculate hourly average wind speed and direction which was then used to supplement the
standard archive of hourly observed winds in the surface data. Because each hour could
have up to 60 1-minute winds, these could be averaged to determine the winds and thereby
reducing the number of calms, variable winds and missing data.
The AERMINUTE preprocessor requires the start and end month and year of the data being
processed as well as whether or not the station is part of the Ice Free Winds (IFW) group.
The IFW group date refers to start of use at the ASOS site of sonic anemometers instead of
cup and vane anemometers (which may have icing problems) to measure winds. If the
station is part of the IFW group during the data period being processed by AERMINUTE,
then the IFW installation date must be entered into the program. In this analysis each IFW
date was entered into AERMINUTE to consitently report this date even if not applicable in
2006. The NWS website http://www.weather.gov/ops2/Surface/documents/IFW_stat.pdf
was used to determine if the stations were part of the IFW group and their respective
installation dates. AERMINUTE gives an option to include data files of standard NWS
observations in order to compare the non-quality controlled 1-minute winds from the 1-

5. 5
minute data files against the quality controlled standard observations. This comparison
was performed only for the Harrisburg, Orangeburg, and Gainesville sites which had ISHD
data.
The combination of the data sets described above was processed by AERMINUTE to
produce the necessary hourly wind speed and direction file for merging with the NWS
surface and upper air data. All other inputs to AERMET were set using regulatory default
options (like random winds), airport specific coordinates and time, dates of processing, and
airport specific roughness parameters, albedo, and Bowen ratio as generated by
AERSURFACE (using the 12 standard 30 degree sectors). Ten sets of meteorological data
were processed including five sets through AERMET and five sets through
AERMINUTE/AERMET. Comparisons of the wind roses and the information pertaining to
calms and average wind speeds is presented in Figures 1-10. Table 2 presents the
comparison of average wind speeds and number of calms.
Figure 1. KCGI AERMET

6. 6
Figure 2. KCGI AERMET/AERMINUTE
Figure 3. KGNV AERMET
Figure 5. KFAR AERMET
Figure 4. KGNV AERMINUTE/AERMET

7. 7
Figure 6. KFAR AERMINUTE/AERMET
Figure 7. KORG AERMET
Figure 9. KMDT AERMET
Figure 8. KORG AERMINUTE/AERMET

8. 8
Figure 10. KMDT
AERMINUTE/AERMET
Table 2. Comparison of AERMET vs. AERMINUTE/AERMET Wind Speeds
Surface/Upper Air
Sites
AERMET Calms, %
AERMET Average
Wind speed, m/s
AERMINUTE/AER
MET Calms, %
AERMINUTE/AER
MET Average Wind
Speed, m/s
Harrisburg, Dulles 23.5% 3.33 10.89% 3.41
Cape Girardeau,
Springfield
22.8% 3.28 2.48% 3.43
Fargo, Aberdeen 5.72% 4.79 1.18% 5.03
Orangeburg,
Charleston
23.18% 2.81 12.1% 2.90
Gainesville,
Jacksonville
23.63% 2.85 12.56% 2.93
As can be seen in making a qualitative comparison between the AERMET and
AERMINUTE/AERMET data sets, more lower wind speeds occur when the 1-minute data is
considered. Also Table 2 shows that the number of calms decreases when considering the
1-minute data, sometimes in dramatic fashion as in the case of Cape Girardeau where the
number of calms dropped from 22.8% to 2.48%. Interestingly, wind speeds increased
slightly at each NWS site when considering the 1-minute data. Thus, on one hand the
number of low wind speeds increased (which means more low wind speeds) while overall
the average wind speed increased.
Of considerable interest that this use of 1-minute data in AERMET has is on the calculations
that take place in AERMOD. To determine if any differences in concentration estimates

9. 9
results, a number of sources ranging from tall stacks to short stacks and area and volume
sources were examined. Table 3 presents the source types reviewed in this analysis.
Included with each stack configuration was an influencing building that could cause
downwash. Most of these scenarios were derived from EPA test files for the new AERMOD,
Version 12060 found at website
(http://www.epa.gov/ttn/scram/dispersion_prefrec.htm#aermod).
Table 3. Source Types and Parameters
Source Type Height, m Diameter,m Temp, K Velocity, m/s Emissions, g/s
Stack 65 5.0 425 15.0 500
Stack 35 2.4 432 11.7 100
Stack 20 1.5 350 7.5 10
Stack 10 0.5 325 4.0 2.5
Area 1 20 = Length 20 = Width - 0.001
AreaCircle 10 40 - - 0.1
Volume 1 20 = zo 20 = yo - 0.1
These sources were modeled using AERMOD (Version 12060) for averaging periods of 1
hour, 3 hours, 8 hours, 24 hours, and annual for each NWS site and for each AERMET and
AERMINUTE/AERMET set. A receptor grid was set such that fence line receptors were
positioned around the sources at 175m from the center at about a 65m spacing. All other
receptors were set up in a Cartesian grid at spacings of 50m out to 500m, 100m out to
1500m, and 250m out to 3500m. All regulatory default options were exercised in AERMOD
except that flat terrain was used.
RESULTS
Comparisons of the ambient concentrations between averaging times and source types
were made to facilitate the determination of the differences that the meteorological data
sets could have on an analysis. The concentrations estimated using the straight AERMET
meteorological data were divided by the concentrations estimated using the AERMET plus
AERMINUTE meteorological data to determine the ratio of the difference in concentration
caused by the change. Ratios less than 1.0 indicate an expected increase or and ratios
greater than 1.0 indicate an expected decrease in concentration, while a ratio equal to 1.0
indicates no expected change in concentration. Similarly, ratios of 2.0 and 0.5 indicate a
halving or doubling of the concentrations, respectively.
Figures 11 though 16 present comparisons for the various source types and averaging
periods. As can be seen from Figures 11-17, the model results for the short term
concentrations using the AERMINUTE/AERMET data sets are greater than for those just
using the AERMET data with no 1-minute winds (ratios less than 1.0). This is expected
given that a greater number of hourly values will be available for AERMOD processing and
winds may be lower because of the increased sensitivity of sonic anemometers (but this
would only affect Fargo met data sets as that was the only station to have sonic
anemometers installed in a portion of the 2006 data set). The combination of these two
factors results in many hours originally reported as calm being replaced with non-calm

10. 10
wind data, and occasionally with wind speeds that are less than 1 m/s. This is apparent in
terms of the number of calms shown in Table 2. Also shown is that the
AERMINUTE/AERMET data sets give somewhat lower concentrations for longer averaging
periods.
Figure 11. All Sites, 65 m Stack
Figure 12. All Sites, 35 m Stack
0
0.5
1
1.5
2
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
STACK 65
Fargo
Annville
OrangeB
Gainsville
Cape G
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
STACK 35
Fargo
Annville
Cape G
OrangeB
Gainsville

11. 11
Figure 13. All Sites, 20 m Stack
Figure 15. All Sites, Area
Figure 17. All Sites, Volume
Figure 14. All Sites, 10 m Stack
Figure 16. All Sites, AreaCircle
0
0.2
0.4
0.6
0.8
1
1.2
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
STACK 20
Fargo
Annville
CapeG
OrangeB
Gainsville
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
AREA
Fargo
Annville
Cape G
OrangeB
Gainsville
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
VOLUME
Fargo
Annville
CapeG
Orange B
Gainsville
0
0.2
0.4
0.6
0.8
1
1.2
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
STACK 10
Fargo
Annville
Cape G
Orange B
Gainsville
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 hr 3-hr 8-hr 24-hr Annual
Q(met)/Q(metmin)
AreaCircle
Fargo
Annville
Cape G
OrangeB
Gainsville

12. 12
While for the taller stacks, 65m and 35m the results across all sites is similar, more
disparity is shown in the ratios of the concentrations across the averaging periods for the
shorter stacks, 20m and 10m but rather consistently giving higher concentrations using the
1-minute data. Another comparison of the results is shown in Figures 18 and 19 for just
the two extremes of averaging time, namely, 1-hour and annual. In these figures the
sources are compared to one another to discern the affect of the 1-minute data by source
type. As can be seen in Figure 18, the Area source was affected the most with
concentrations ranging from about 0.2 to 0.7 for the AERMET concentrations verssu the
AERMINUTE/AERMET concentrations. The same infrmation for the circular area source,
CIRC was not as dramatic but still in the range of 0.6-1.0, with the best agreement between
data sets in Fargo. This was expected given the higher average wind speed in Fargo (5.03
m/s) than other sites and the low frequency of calms before the 1-minute data was even
considered.
Figure 18. All Sites, All Sources, 1-Hr Figure 19. All Sites, All Sources, Annual
0
0.2
0.4
0.6
0.8
1
1.2
STACK
65
STACK
35
STACK
20
STACK
10
AREA CIRC VOL
Q(met)/Q(metmin)
1-hr by Source Type
Fargo
Annville
Cape G
OrangeB
Gainsville
0
0.2
0.4
0.6
0.8
1
1.2
1.4
STACK
65
STACK
35
STACK
20
STACK
10
AREA CIRC VOL
Q(met)/Q(metmin)
Annualby Source Type
Fargo
Annville
Cape G
Orange B
Gainsville

13. 13
CONCLUSIONS
While it is not the interest of these authors to make such comparisons as above in the
interest of saying one data set is better, or one data set is less or more conservative. Each
data set has its merits with that of AERMINUTE to help define hourly winds for hours that
in the past may have had a calm or no calculation performed. The possible downside to
using the 1-minute data is exacerbating the known AERMOD problem in calculating
concentrations for low wind speed conditions. Thus, the main conclusion to this analysis is
that the user should be aware of the possible differences in concentrations for various
source types and for various averaging periods.
REFERENCES
1. User’s Guide for the AMS/EPA Regulatory Model - AERMOD. U.S. Environmental
Protection Agency, Research Triangle Park, North Carolina. Revised September
2004.
2. Guideline on Air Quality Models. Appendix W to 40 CFR Parts 51 and 52.
FederalRegister, November 9, 2005. pp. 68217-68261. 2005.
3. AERSURFACE Users Guide, U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina, 2008.
4. User’s guide for the AERMOD Meteorological Preprocessor (AERMET), EPA-454/B-03-
002, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina.
Under Revision, November 2004.
5. AERMINUTE User’s Instructions, U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina. Date given on File December 20, 2011.
KEYWORDS
AERMOD, AERMINUTE, dispersion, meteorology, calms