Technical Support Document (TSD) for
Replacement of CALINES with AERMOD for
Transportation Related Air Quality Analyses
-------
-------
EPA-454/B-15-002
July 2015
Technical Support Document (TSD) for Replacement of CALINE3 with AERMOD for
Transportation Related Air Quality Analyses
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Analysis Division
Air Quality Modeling Group
Research Triangle Park, North Carolina
-------
-------
Preface
This document provides a comparison of CALINE3 and AERMOD, including an analysis of the scientific
merit of each dispersion model, a summary of existing model evaluations, and the presentation of
additional testing by EPA.
-------
Contents
Preface 5
Contents 6
Figures 7
Tables 8
1. Introduction 9
2. Background 9
2.1CALINE3 history and status 9
2.2 AERMOD history and status 10
3. Model selection 11
3.1 Model inter-comparison studies 12
3.2 Regulatory applications for mobile sources 14
3.3 Summary of findings and recommended model 17
4. Acknowledgements 18
5. Additional information 18
References 19
Appendix A 20
Results from comparison of AERMOD and CAL3QHC for CO hot-spot screening for highway and
intersection projects 20
-------
Figures
Figure 1 - QQ plot of Model Performance for Idaho Falls Study, based on (Heist, et al., 2013) 15
Figure 2 - RHC vs FB Model Performance Statistics for Idaho Falls Study, based on (Heist, et al., 2013).. 15
Figure 3 - QQ plot of Model Performance for CALTRANS 99 Study, based (Heist, et al., 2013) 16
Figure 4 - RHC vs FB Model Performance Statistics for CALTRANS 99 Study, based (Heist, et al., 2013). . 16
Figure 5 - Layout of sources and receptors for CO screening tests 25
Figure 6 - Close-up of receptor locations for CO screening tests 26
-------
Tables
Table 1 - Model Performance Statistics from the Idaho Falls Study. Source: (Heist, et al., 2013) 13
Table 2 - Model Performance Statistics from the CALTRANS 99 Study. Source: (Heist, et al., 2013) 13
Table 3 - Summary of MOVES emissions for CO example 20
Table 4 - Link dimensions for CO screening runs 23
Table 5 - Link emissions for CO screening runs 23
Table 6 - Receptor locations for CO screening runs 24
Table 7 - Results of CAL3QHC and AERMOD CO screening tests 27
-------
1. Introduction
The proposed revisions to EPA's Guideline on Air Quality Models, published as Appendix W to 40 CFR
Part 51, include the proposal to remove CALINE3 for mobile source applications from Appendix A and
replace it with AERMOD. This document provides the technical details supporting this proposed change,
including the scientific merit of each dispersion model, summary of existing model evaluations, and
presentation of additional testing by EPA used to determine appropriate application of these options as
part of the proposal for AERMOD to be the required model for mobile source dispersion modeling.
2. Background
The current version of Appendix W, published in 2005, addresses modeling mobile sources, with specific
recommendations for each criteria pollutant. AERMOD is currently EPA's recommended near-field
dispersion model for regulatory applications. In addition, for carbon monoxide (CO), CAL3QHC is
recommended for screening and CALINE3 for free flow situations. For lead (Pb), CALINE3 and CAL3QHCR
are identified for highway emissions, while for nitrogen dioxide (NO2), CAL3QHCR is listed as an option.
No models for mobile emissions are explicitly identified for coarse particulate matter (PM10), fine
particulate matter (PM2.5), or sulfur dioxide (SO2), though CALINE3 is listed in Appendix A as
appropriate for highway sources for averaging times of 1-24 hours.
2.1 CALINE3 history and status
The first CALINE line model was initially developed in 1972, with a focus on predicting CO concentrations
near roadways (Benson, 1992). CALINE2 was developed in 1975, porting CALINE to FORTRAN and adding
formulations for depressed roadways (Benson, 1992). CALINE3, which was developed in 1979 (Benson,
1979), updated the vertical and horizontal dispersion curves, reducing, but not eliminating, the over-
predictions occasionally seen in CALINE2 (Benson, 1992). CALINE3 also updated the available averaging
time, parameterized vehicle-induced turbulence, replaced the virtual point source with a finite line
source, and increased the number of links capable in the model. CALINE3 was replaced by CALINE4 in
1984 (Benson, 1984), with further modifications to the lateral plume spread and vehicle induced
turbulence, the addition of intersections, and limited chemistry for NO2 and PM. Unlike CALINE3,
CALINE4 is not open source, such that the model code is not publically available, and thus does not meet
the requirements in Appendix W for consideration as a preferred model. The CALINE models are
Gaussian plume models, and though changes were made to the dispersion curves with each version, the
dispersion curves are based on the Pasquill-Gifford (P-G) stability classes. The P-G stability classes do
not reflect state of the science: the ISC dispersion model was also based on the P-G stability classes, and
EPA replaced the ISC model with AERMOD in EPA's 2005 revision to Appendix W. Section 2.2 includes
additional detail about how stability is defined in AERMOD.
In the late 1980s, CALINE3 was modified to automate estimates of vehicle queue lengths at
intersections, resulting in the CAL3QHC screening model (U. S. EPA, 1995). In the early 1990s, further
modifications were made to CAL3QHC to update traffic queuing and signaling based on the 1985
Highway Capacity Manual, increasing the number of links and receptors, and to add multiple wind
directions to facilitate screening analyses (U. S. EPA, 1995). CAL3QHC was developed primarily for CO
hot-spot analyses, computing hourly concentrations using "worse case" meteorology, which can then be
scaled to an 8-hour average to estimate compliance with the CO National Ambient Air Quality Standard
(NAAQS).
-------
Shortly after the development of CAL3QHC, additional work was done with the model to allow more
refined estimates (rather than screening estimates) of emissions from roadways. The CAL3QHCR model
is based on CAL3QHC, but has several modifications, including the ability to run 1-year of hourly
meteorology, additional capabilities related to queuing and signalization, the addition of PM to the hard-
coded pollutant options, incorporation of the mixing height algorithms from ISCST2, the ability to vary
emissions by hour of the week, and the ability to calculate averages longer than 1-hour (Eckhoff &
Braverman, 1995). The model was developed for situations when the screening, worst-case estimates
from CAL3QHC indicated potential exceedances of the standard and more refined estimates were
required. It should be noted that with the incorporation of the ISCST2 mixing height algorithms,
CAL3QHCR has undergone modifications from the dispersion in CALINE3 and CAL3QHCRthat have not
been reviewed with the same rigor and detail that was conducted for the other two models (Eckhoff &
Braverman, 1995). As a result, there is some question as to the equivalency of CAL3QHCR to CALINE3
and CAL3QHC for identical model scenarios. Even so, CAL3QHCR has been listed in text of Appendix W,
but not as a preferred model in Appendix A. CALINE3 was originally developed jointly by the Federal
Highway Administration (FHWA) and the CA Department of Transportation (Caltrans). EPA sponsored
much of the work to develop CAL3QHC and CAL3QHCR in the 1990s. The model codes are hosted on
EPA's dispersion model website (hSBj//www,.,e,Ba,^ Brefrecjitm).
The CALINE3-based models present some challenges when used for mobile source modeling. Current
pre-processed meteorological data cannot be used with these models; the most recent pre-processed
meteorological data available for them is from the 1990s. Furthermore, applying the CAL3QHCR model
for the 24-hour and annual PM NAAQS requires multiple runs to represent a sufficiently long
meteorological data period. For example, where a project-sponsor has off-site meteorological data, one
AERMOD run is needed, in contrast to 20 CAL3QHCR runs. The CALINE models can model line sources
only, which limit their application to highways and intersections.1 They cannot be used for any other
type of mobile source modeling, such as modeling a project that involves a parking lot or a freight or
transit terminal. The use of the queuing algorithm for intersection idle queues is no longer
recommended as EPA's MOVES emission factor model now accounts for changes in such activity.
2.2 AERMOD history and status
The AMS-EPA Regulatory MODel (AERMOD) was developed over a 10-year period jointly by the
American Meteorological Society (AMS) and EPA through the AERMOD Model Improvement Committee
(AERMIC). In 2005, AERMOD was promulgated as EPA's preferred dispersion model for most inert
pollutants (plus NO2) as part of revisions to Appendix W. The model reflects state of the science
formulation for Gaussian Plume dispersion models (Cimorelli, et al., 2005). One of the major updates in
AERMOD versus the previous preferred dispersion model, ISCST3, was the transition from the usage of
P-G stability class based parameterizations of dispersion coefficients. As detailed in (Cimorelli, et al.,
2005), state of the science models like AERMOD use a planetary boundary layer (PBL) scaling parameter
to characterize stability and determine dispersion rates based on Monin-Obukhov (M-O) similarity
profiling of winds near the surface. AERMOD's performance was evaluated with 17 field study databases
that represent a large variety of source types, local terrain, and meteorology (Perry, et al, 2005).
1 Based on implementation since 2010, some PM hot-spot analyses have been completed with CAL3QHCR,
although the majority of such analyses have been based on AERMOD.
-------
AERMOD was found to be superior to ISCST3 for the majority of the situations modeled.
AERMOD includes options for modeling emissions from volume, area, and point sources and can
therefore model the impacts of many different source types, including highways, intersections,
intermodal terminals, and transit projects. In addition, EPA conducted a study to evaluate AERMOD and
other air quality models in preparation for developing EPA's quantitative PM hot-spot guidance, and the
study supported AERMOD's use (Hartley, Carr, & Bailey, 2006). To date, AERMOD has already been used
to model air quality near roadways, other transportation sources, and other ground-level sources for
regulatory applications by EPA and other federal and state agencies. For example, EPA used AERMOD to
model concentrations of nitrogen dioxide (NO2) as part of the 2008 Risk and Exposure Assessment for
revision of the primary NO2 NAAQS (U. S. EPA, 2008). Also, other agencies have used AERMOD to model
PM and other concentrations from roadways (represented as a series of volume or area sources) for
regulatory purposes, including Clean Air Act transportation conformity analyses Current pre-processed
meteorological data based on AERMET is available for AERMOD from state air agencies, and the model
offers efficiencies in calculations needed for the 24-hour and annual PM NAAQS (only one run is needed
with site-specific meteorological data in contrast to four runs for CAL3QHCR; one run would be needed
with data from off-site, in contrast to 20 runs for CAL3QHCR, (U.S. EPA, 2013)).
As EPA's preferred model, AERMOD has undergone continuous updates and developments in order to
improve its performance for particular source types, meteorological conditions, and terrain features as
well as to keep the model up to date with state of the science parameterizations for dispersion
modeling. One of the major actions of the EPA's proposed revisions to Appendix W is to formally adopt
many enhancements made over the past 10 years into AERMOD (version 15181). EPA is committed to
continuing to update the AERMOD modeling system to keep it a state of the science dispersion model
and to incorporate updates and advancements, as scientifically appropriate, in accordance with the
needs of regulatory stakeholders and the broader modeling community. The preamble for this proposed
modification to Appendix W and the supporting technical support documents describe the numerous
modifications that have been made to AERMOD over the last decade as well as provide details on the
scientific basis and model evaluations that have been conducted to continually improve the AERMOD
modeling system.
3. Model selection
Section 3.1.1 of the current Appendix W (also section 3.1.1 of the proposed Appendix W) states, "When
a single model is found to perform better than others, it is recommended for application as a preferred
model and listed in Appendix A." Appendix A lists the models that EPA has determined can be used
without any further justification for the particular application they have been identified. There are
several requirements for a "preferred model" to be listed in Appendix A (section 3.1.1 of Appendix W),
including that the model is written in a common programming language; the model is well documented;
test datasets are available for model evaluation; the model is useful to typical users; there are robust
model-to-monitor comparisons; and the source code is freely available. At the time of the current
Appendix W's promulgation in 2005, there had been no inter-comparisons between AERMOD and
CALINE3 with sufficient merit to modify the status of CALINE3 as the preferred model for mobile source
-------
applications. However, since 2005, there have been notable model inter-comparisons for AERMOD and
CALINE3, as described below, that warrant removing CALINE3 from the list in Appendix A.
3.1 Model inter-comparison studies
There are several types of model inter-comparison studies that are applicable for mobile source
modeling. There are model sensitivity tests that compare model simulations for matching
meteorological and emissions scenarios, but lack the ambient monitoring data to evaluate model
performance. Alternatively there are studies for which ambient concentration measurements are
available along with meteorological data for the measurement site, but emissions are parameterized in
some fashion. Typically, traffic counts are used, and an emissions model (e.g., EPA's Motor Vehicle
Emission Simulator, or MOVES, model) is applied to estimate vehicular emissions. There can be
significant uncertainties in these studies based on errors in the traffic counts, uncertainty in the
emission profiles, and estimates that must be made to distribute emissions among different vehicles
types, ages, etc. The best studies, however, are field studies based on metered emissions, usually the
release of a passive tracer, with little or no background concentrations. These studies generally
eliminate the uncertainties in the emissions and other model input and allow for the best evaluation of
model performance.
When dealing with inert pollutants, a Gaussian dispersion model will operate in the same way regardless
of pollutant. While CAL3QHC and CAL3QHCR are hard-coded to convert the input emissions to mixing
ratios of CO (or concentrations of PM for CAL3QHCR), the dispersion parameterizations in these models
would apply for any pollutant. Therefore, the models' performance can be examined accurately using
another inert pollutant such as a passive tracer, as is done in the field studies discussed here.
(Heist, et al., 2013) conducted a model inter-comparison based on data from two field studies that had
known, metered emissions of inert SF6 tracers. SF6 is an inert pollutant used as the passive tracer in the
studies. The first field study, CALTRANS 99, was conducted along Highway 99 outside Sacramento, CA.
CALTRANS 99 used eight automobiles outfitted with SF6 emission units. The automobiles completed
circuits of a section of highway during periods when meteorological conditions were favorable, i.e.,
winds were blowing from the highway to the monitors. SF6 monitors were placed perpendicular to the
roadway at 50, 100, and 200 meters (m), with monitors along the roadway median. A total of 14 days of
samples were collected for CALTRANS 99. The second field study, carried out in Idaho Falls, ID, was
conducted in an open field with SF6 released uniformly along a 54 m long source meant to replicate
emissions from a roadway. A grid of 56 monitors were placed downwind of the source at distances
ranging from 15-180 m. Data was collected on a total of four days, representing a range of atmospheric
stabilities and wind speeds. Both field studies had on-site meteorological measurements.
(Heist, et al., 2013) used these two field studies to evaluate model performance for several dispersion
models to determine their ability to model concentrations from roadway emissions in the near-field. The
models included AERMOD, CALINE3 and CALINE4, the Atmospheric Dispersion Modelling System
(ADMS), which is the UK's preferred dispersion model for regulatory purposes, and RLINE, a research
model specifically for roadway sources that is being developed by EPA's Office of Research and
Development (ORD). Four statistical measures were computed to benchmark each model's ability to
-------
replicate the monitored concentrations. These measures were the fractional bias (FB), normalized mean
square error (NMSE), the correlation (R), and the fraction of estimates within a factor of two of the
measured value (FAC2). These results are summarized in Table 1 and Table 2.
Table 1 - Model Performance Statistics from the Idaho Falls Study. Source: (Heist, et al., 2013).
Model
CALINE4
AERMOD - volume
AERMOD - area
ADMS
RLINE
FB (0 is best)
0.42
0.38
0.32
0.36
0.23
NMSE (0 is best)
1.94
1.26
1.25
1.14
0.96
R(lisbest)
0.76
0.84
0.82
0.88
0.85
FAC2(lisbest)
0.59
0.59
0.59
0.70
0.73
Table 2 - Model Performance Statistics from the CALTRANS 99 Study. Source: (Heist, et al., 2013).
Model
CALINE3
CALINE4
AERMOD - volume
AERMOD - area
ADMS
RLINE
FB (0 is best)
0.25
0.19
0.15
0.13
0.09
0.05
NMSE (0 is best)
2.26
0.86
0.28
0.31
0.20
0.34
R(l is best)
0.29
0.47
0.77
0.72
0.78
0.75
FAC2(lisbest)
0.45
0.68
0.78
0.76
0.85
0.78
In general, the performance statistics indicate that the CALINE models are the worst performing for both
field studies (also see Figure 5 and Figure 9 in (Heist, et al., 2013)). However, it should be noted that
these metrics were computed for all modeled concentrations, rather than for the highest concentrations
only. Regulatory models are generally needed to replicate the highest concentrations and, as a result,
model evaluations for regulatory models typically focus on statistics for the highest concentrations (the
highest 25 is the most common, (Cox & Tikvart, 1990)). The need to replicate only the highest
concentrations also means that performance of regulatory models is generally not based on pairing
modeled concentrations in time and space. Instead, all concentrations are ranked from highest to lowest
and compared independent on the timing and location. Figure 1 and Figure 3 show the quantiles plot, or
QQ-plot, typically used to show model performance for ranked concentrations. From these plots, it can
be seen that CALINE has the worst performance at the highest concentrations for both field studies and
severely underestimates concentrations in Idaho Falls and overestimating concentrations in CALTRANS
-------
99. Based on these results, AERMOD appears to perform the best of all the dispersion models, being
closest to the 1:1 line for the highest concentrations. When only the top 25 concentrations are
considered, the FB and RHC are clearly better for AERMOD than CALINE. Figure 2 and Figure 4 show the
ratios of modeled RHC to observed RHC vs FB for the field studies for the highest 25 concentrations only.
A perfect model would have a FB of 0 and a ratio of modeled RHC to observed RHC of 1. For Idaho Falls,
RLINE and AERMOD with both volume and area sources have virtually identical performance. For
CALTRANS99, ADMS and AERMOD with both volume and area sources have very similar performance,
with AERMOD volume sources performing best.
3.2 Regulatory applications for mobile sources
The current and future needs for mobile source modeling have evolved beyond the uses outlined in the
2005 version of Appendix W. For example, Pb modeling for mobile sources is no longer needed, as
leaded gasoline is no longer used in the U.S. Currently, mobile source modeling for regulatory needs
occurs primarily for CO, PM10, and PM2.5 hot-spot modeling for mobile source conformity analyses.
Due to the background levels, emissions factors, types of projects modeled, and the shorter time period
covered by the CO NAAQS, screening modeling involving conservative, worst-case modeling is
exclusively done for CO analyses. Refined analyses involving actual meteorology with best estimates of
emissions are conducted for PM10 and PM2.5. Because of the complex nature of PM emissions, the
statistical form of each NAAQS, and the need to consider temperature effects throughout the time
period of a year, EPA believes that quantitative PM hot-spot analyses need to be completed using the
refined analysis procedures described in EPA's quantitative PM hot-spot guidance (U.S. EPA, 2013).
For CO screening analyses, CAL3QHC has been exclusively used for the past several decades with refined
CO hot-spot modeling being completed in limited cases. Currently, EPA's MOVES emission model is used
to estimate vehicular emissions for CO modeling (except in California, where EMFAC, short for EMission
FACtor, is used). These emission models can be used to determine emission rates for free-flow traffic
and rates for idle traffic (i.e., traffic in a queue at an intersection). Emissions from free-flow and idle
traffic are input to CAL3QHC, along with the signalization and geometries of the intersection.
For PM10 and PM2.5, AERMOD and CAL3QHCR have both been allowed for refined analyses. Although
AERMOD does not have the capability to internally parameterize queuing emissions like CAL3QHCR
does, it is not needed because queuing emissions are already accounted for by MOVES (and EMFAC in
California). As noted in EPA's quantitative PM hot-spot guidance, CAL3QHCR's queuing algorithm should
not be used in PM hot-spot analyses. These emissions, along with the geometries of the project, and
meteorological data are input into AERMOD and CAL3QHCRto determine ambient impacts.
-------
QQ plot for Idaho Falls SF6 study
100-
75-
variable
RLINE
A AERMOD.area
• AERMOD.volume
+ CALINE4
EI ADMS.R
* ADMS.M
i i i
25 50 75
Measured SF6
100
Figure 1 - QQ plot of Model Performance for Idaho Falls Study, based on (Heist, et al., 2013).
FB vs RHC for Idaho Falls SF6 study
3-
II
o
I
C£
o-
I
-:
Model
• ADMS
A AERMOD_area
• AERMOD_volume
-\- CALINE4
M RLINE
-1
FB, n=25
Figure 2 - RHC vs FB Model Performance Statistics for Idaho Falls Study, based on (Heist, et al., 2013).
-------
QQ plot for Caltrans 99 SF6 study
15000-
CO
-D 10000
J)
(D
-a
o
5000-
0-
I
0
1000 2000 3000
Measured SF6
+
ISB +
4000
variable
RLINE
A AERMOD.A
• AERMOD.V
+ CALINE4
EI CALINE3
* ADMS
Figure 3 - QQ plot of Model Performance for CALTRANS 99 Study, based (Heist, et al., 2013).
FB vs RHC for CALTRANS 99 SF6 study
LO
II
O
o-
i
-;
-1 o 1
FB, n=25
Model
• ADMS
A AERMOD_area
• AERMOD_volume
+ CALINE3
M CALINE4
RLINE
Figure 4 - RHC vs FB Model Performance Statistics for CALTRANS 99 Study, based (Heist, et al., 2013).
-------
3.3 Summary of findings and recommended model
As discussed in section 3.1.1 of Appendix W, EPA should only list a preferred model in Appendix A when
it is "found to perform better than others." In the 2005 update to Appendix W, no comparison was
made between AERMOD and CALINE3 to assess which model actually performed better for mobile
source applications. However, since that time, model inter-comparison studies now provide strong
evidence that AERMOD is the best performing model relative to CALINE3 (and CALINE4) for mobile
source applications. Specifically, EPA has found that:
• The dispersion modeling science used in CALINE3 is very outdated (30 years old) as compared to
AERMOD, RLINE and other state-of-the-science dispersion models. CALINE3 is based on the
same dispersion science underlying the ISCTS3 model, which EPA replaced with AERMOD in
2005 as the preferred regulatory dispersion model for inert pollutants.
• The model performance evaluations presented by (Heist, et al., 2013) represent the best model
comparison for AERMOD, CALINE3 and CALINE4 to date. This study used metered emissions of
an SF6 tracer and concurrent near-road measurements to serve explicitly as a platform for
evaluating mobile source models. The results showed that CALINE3 and CALINE4 were the worst
performing models of the 5-model comparison for the two available field studies (Idaho Falls
and CALTRANS 99) when considering all modeled and monitored concentrations, paired in time
and space.
• Additional analysis of the data from (Heist, et al., 2013) was conducted by EPA in the context of
regulatory use of models. This analysis focused on the highest concentrations (i.e., top 25
concentrations), which are most relevant for regulatory purposes, and typically the focus of
performance evaluations of regulatory models. This additional analysis showed that not only
were CALINE3 and CALINE4 the worst performers, but that AERMOD was the best performing
model of the group.
• As described in more detail in Appendix A below, CALINE3 is insensitive to changes in mixing
height which provides further support for the replacement of this model with AERMOD. For
surface releases like roadways, low winds, stable conditions and a low mixing height are
expected to result in the worst case concentrations because they are kept close to the ground.
The recommendations in the 1995 CAL3QHC User's Guide result in assumptions that are
somewhat contradictory and unrealistic.
In addition to the evidence about model performance, CALINE3, CAL3QHC, and CAL3QHCR have several
limitations related to the model input that make them more difficult than AERMOD to use for refined
modeling:
• Meteorological pre-processors for the CALINE3 models are only available for older
meteorological data sets. As a result, newer, higher resolution meteorological data, that is
more representative of actual wind conditions cannot readily be used. In contrast, pre-
processed meteorological data from AERMET is available from state air agencies for use in
AERMOD.
• For CAL3QHCR, only 1 year of meteorological data can be used in each model run. For
-------
refined PM10 and PM2.5 analyses, this requires multiple model runs to cover a 5-year
modeling period with resulting model output data from up to 20 model runs that must be
separately post-processed to obtain the necessary results.
Based on the data available, AERMOD is the best performing model for mobile source applications.
Additionally, AERMOD is not limited by the practical usability issues especially in terms of most
recent and improved model inputs data inputs that are not available with the CALINE3 models. As a
result of these factors, EPA has proposed to replace CALINE3 with AERMOD for all mobile source
applications. This proposed change also promotes greater commonality and consistency in air
quality modeling analyses for EPA regulatory applications. For mobile sources, regulatory situations
in which AERMOD would be used now and in future include:
• PM hot-spot analyses
• CO hot-spot analyses
• PM SIP attainment demonstrations
• PSD applications (PM, SO2, NO2, Pb, CO)
• NO2 near-road monitor siting and other potential future applications
4. Acknowledgements
The authors would like to acknowledge the intra-agency workgroup, specifically contributions from staff
in the Office of Research and Development, the Office of Transportation and Air Quality, and Regions 5
and 8.
5. Additional information
Data for the analyses presented in this TSD can be obtained by contacting:
Chris Owen, PhD
Office of Air Quality Planning and Standards, U. S. EPA
109 T.W. Alexander Dr.
RTP, NC 27711
919-541-5312
owen.chris@epa.gov
-------
References
Benson, P. (1979). CALINE3-a versatile dispersion model for predicting air pollutant levels near highways
and arterial streets. FHWA-CA-TL-79-23: CA DOT, Sacramento, CO.
Benson, P. (1984). CALINE4-a dispersion model for predicting air pollutant concentrations near
roadways. California Department of Transportation, Sacramento, CA: FHWA-CA-TL-84-15.
Benson, P. (1992). A review of the development and application of the CALINE3 and CALINE4 models.
Atm. Env., 379-390.
Cimorelli, et al. (2005). AERMOD: A Dispersion Model for Industrial Source Applications. Part I: General
Model Formulation and Boundary Layer Characterization. J. App. Meterol., 682-693.
Cox, W., & Tikvart, J. (1990). A Statistical Procedure for Determining the Best Performing Air Quality
Simulation Model. Atmos. Environ., 24, 2387-2395.
Eckhoff, P., & Braverman, T. (1995). Addendum to the User's Guide to CAL3QHC Version 2.0 (CAL3QHCR
User's Guide). OAQPS, RTP, NC.
Hartley, W., Carr, E., & Bailey, C. (2006). Modeling hotspot transportation-related air quality impacts
using ISC, AERMOD, and HYROAD. Proceedings of Air & Waste Management Association
Specialty Conference on Air Quality Models. .
Heist, et al. (2013). Estimating near-road pollutant dispersion: A model inter-comparison. Trans. Res.
Part D, 93-105.
Perry, et al. (2005). AERMOD: A Dispersion Model for Industrial Source Applications. Part II: Model
Performance against 17 Field Study Databases. J. App. Meterol., 694-708.
U. S. EPA. (1992). Guideline for Modeling Carbon Monoxide from Roadway Intersections, EPA document
number EPA-454/R-92-005. Research Triangle Park, NC: Office of Air Quality Planning and
Standards.
U. S. EPA. (1995). User's guide to CAL3QHC version 2.0: A modeling methodology for predicting pollutant
concentrations near roadway intersections (revised). OAQPS, RTP, NC: EPA-454/R-92-006R.
U. S. EPA. (2008). Risk and Exposure Assessment to Support the Review of the NO2 Primary National
Ambient Air Quality Standard, EPA document # EPA-452/R-08-008a. RTP, NC.
U. S. EPA. (2011). AERSCREEN User's Guide. RTP, NC 27711, EPA-454/B-11-001: Office of Air Quality
Planning and Standards.
U.S. EPA. (2013). Transportation Conformity Guidance for Quantitative Hot-Spot Analyses in PM2.5 and
PM10 Nonattainment and Maintenance Areas . Retrieved from Transportation and Climate
Division, Office of Transportation and Air Quality, EPA-420-B-13-053.
-------
Appendix A
Results from comparison of AERMOD and CAL3QHC for CO hot-spot screening for
highway and intersection projects
As noted main document, AERMOD is already used for PM10 and PM2.5 hot-spot analyses. However, for
CO hot-spot analyses, CAL3QHC is currently the primary air quality model used. Therefore, some
comparisons of CO screening scenarios are presented here to illustrate the differences between
AERMOD and CAL3QHC for these types of analyses and to illustrate how AERMOD can be used for CO
screening purposes in hot-spot analyses.
The basis for these comparisons is modeled emissions from a simple one-mile highway segment,
consisting of four lanes, two in the northbound and two in the southbound directions. MOVES2014 was
used at the project scale to estimate emissions from this roadway in the year 2015. Runs were done in
both the Inventory mode to produce total CO emissions in the hour, and in the Emission Rates mode to
produce a CO rate per vehicle-mile. The following choices were made in MOVES:
• Each lane was assumed to have 2000 vehicles traveling during the hour, i.e., 4000 vehicles in
each direction per hour, for both the highway and the arterial. This amount of vehicles is close
to capacity for the length of road, to be conservative (i.e., produce a higher level of emissions).
• A temperature of -10°F and humidity of 50% was assumed, to be conservative because CO
emissions increase at colder temperatures.
• The average speed on the highway was assumed to be 74 mph.
• All valid combinations of gasoline, diesel, ethanol, and CNG capable vehicles were chosen, and
EPA used a typical mix of vehicles for each facility type.
• The age distribution of vehicles was based on EPA's age distribution calculator (default
information), for 2015.
• No I/M program was assumed.
• Default fuel parameters for Washtenaw County, Michigan were used.
MOVES2014 produced the following emissions:
Vehicle type Highway emissions, each direction (i.e., two lanes)
Heavy duty vehicles 1791 g (10.2%)
Light duty vehicles 15,747 g (89.8%)
Total 17,538 g
For the air quality modeling, the following combinations of source characteristics were included:
• Urban Dispersion (urban population of 1,000,000 used in AERMOD)
• Free Flow Lanes, At Grade (TYP=AG)
• 4 Lanes, (2 north bound lanes, 2 south bound lanes)
• Lane Width = 12 ft (3.66 m), Lane Length = 5280 ft (1 mile)
• Surface Roughness = 0.01 m
• Highway: 10.2% emissions from heavy duty vehicles, 89.8% from light duty vehicles
-------
• 36 wind directions modeled, every 10 degrees
There are several settings that are unique to each model, in particular AERMOD has more source-
characterizations options than CAL3QHC. The model-specific settings are summarized as follows:
• CAL3QHC
o Link Type = At Grade
o Link Width = (# lanes X 12 ft) + 20 ft, i.e. 4 x 12 +20 ft = 68 ft
o Link Height = 0 ft
• AERMOD:
o Each link modeled as a single LINE source
o Source Elevation = 0 m
o LINE Width = (# lanes X 3.66 m) + 6 m, i.e. 4 x 3.66 + 6 = 20.64 m, or 67.72 ft
o Release Height: based on weighted emissions by vehicle mix
o Vertical Dispersion Coefficient: based on weighted emissions by vehicle mix
The source characterization for AERMOD was modified to meet recommendations outlined for PM hot-
spot modeling (e.g., source width and initial dispersion parameters, av and az). The full details of the
source characteristics and receptor locations are given in Table 3, Table 4, Table 5, and Table 6.
Receptors were placed at a height of = 1.83 m (6 ft) for both models.
CAL3QHC is used for screening analyses which focuses on using "worst-case" meteorology to estimate
the worst possible 1-hour concentrations. For the examples provided here, the "worst case"
meteorology was taken from the original CAL3QHC runs, which consists of a 1000 m mixing height, a 1
m/s wind speed and high stabilty (P-G stability class 5). The selection of this meteorological combination
is consistent with the CAL3QHC User Guide and EPA's current guidance for CO screening analyses (U. S.
EPA, 1992). However, these assumptions are somewhat contradictory and unrealistic. For surface
releases like roadways, low winds, stable conditions and a low mixing height are expected to result in
the worst case concentrations because emissions are kept close to the ground. These conditions
typically occur during nighttime, as mixing heights and turbulence are generally higher and during the
day due to solar heating of the surface. The mixing height of 1000 m used in this scenario is much too
high for "worst case" concentrations and in fact could not physically occur in the atmosphere with the
accompanying low wind and highly stable conditions. Despite this, the 1000 m mixing height is
recommended in the 1995 CAL3QHC User's Guide:
"Mixing height should be generally set at 1000 m. CALINE-3 sensitivity to mixing height is
significant only for extremely low values (much less than 100 m)."
(As noted above, the fact that CALINE3 is insensitive to changes to the mixing height provides further
support for the replacement of this model with AERMOD.) Despite the unrealistic meteorological
conditions used in the CAL3QHC, they were replicated as closely as possible for the base case in
AERMOD. The AERMOD meteorology was created using the MAKEMET tool provided with AERSCREEN
(U. S. EPA, 2011) to find met conditions corresponding to the CAL3QHC "worst-case" stability and mixing
height. These met scenarios corresponded to wind speeds of 10 m/s. The 10 m/s winds were then
replaced with 1 m/s to approximately match the CAL3QHC "worst-case".
-------
In addition to the CAL3QHC "worst-case" scenario, several additional meteorological scenarios were
created with MAKEMET that were then, as closely as possible, converted to equivalent meteorology for
use in CAL3QHC. In the first case, the unsubstituted wind-speed meteorology from the "worst-case" was
used, evaluating a stable condition, with high wind speeds and a high mixing height. The other scenarios
are meant to represent a range of meteorological conditions, including representative nighttime
meteorology, with low mixing height, wind speeds, and high stability (a closer representation of the
"worst-case"), and a moderately unstable daytime condition, with moderate wind speeds and mixing
heights. In contrast to the base case meteorology, the meteorology generated with MAKEMET are
meteorological conditions that could actually be observed in the atmosphere. The additional
meteorological scenarios also highlight how responsive AERMOD is relative to CAL3QHC.
The results from these tests are shown in Table 7. For the base (and unrealistic) case, AERMOD has
lower concentrations than CAL3QHC. For the other scenarios, AERMOD has higher concentrations. As
shown in the results above, in some cases CALINE3 will underpredict, while in others, CALINE3 will
overpredict. In contrast, AERMOD is expected to most accurately estimate concentrations. Thus, the
higher concentrations predicted by AERMOD for these alternative scenarios are expected to be
reasonable and more accurate than CAL3QHC.
-------
Scenario
4
Lane-
Scenario
4
Lane-
CAL3QHC
INK
Northbound
Southbound
AERMOD- LINE Source
SrcID
NORTH01
SOUTH01
Links
TYP
AG
AG
(1 Source
Xsl
3.
-3
(m)
7
7
XL1 (ft)
12
-12
YL1 (ft)
-2500
-2500
XL2 (ft)
12
-12
YL2 (ft)
2500
2500
Per Link)
Ysl (m)
-762
-762
0
0
Xs2
3.
-3
(m)
7
7
Ys2 (m)
762.0
-762.0
Zs(m)
0
0
Table 4 - Link dimensions for CO screening runs
CAL3QHC Links
Scenario
4 Lane-
AERMOD-
Scenario
4 Lane-
LNK
Northbound
Southbound
LINE Source (3 Equal
VPHL
(veh/hr)
4400
4400
Area, Equal
SrcID Emis (g/s-m2
NORTH01
SOUTH01
1.734E-03
4.059E-04
EFL (g/veh-mi) HL (ft)
4.38 0
4.38 0
Length Sources Per Link)
) Rel Ht (m) Width (m)
1.55 13.3
1.55 13.3
WL (ft)
44
44
Szinit (m)
1.515
1.515
Table 5 - Link emissions for CO screening runs
-------
Receptor Name
REC 10R
REC20R
REC30R
REC55R
REC80R
REC 105R
REC 130R
REC 155R
REC 180R
REC295R
REC 10L
REC20L
REC30L
REC55L
REC SOL
REC 105L
REC 130L
REC 155L
REC 180L
REC295L
x(ft)
50.40
60.40
70.40
95.40
120.40
145.40
170.40
195.40
220.40
335.40
-50.40
-60.40
-70.40
-95.40
-120.40
-145.40
-170.40
-195.40
-220.40
-335.40
Y(ft)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
z(ft)
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
6.00
X(m)
15.36
18.41
21.46
29.08
36.70
44.32
51.94
59.56
67.18
102.23
-15.36
-18.41
-21.46
-29.08
-36.70
-44.32
-51.94
-59.56
-67.18
-102.23
Y(m)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Z(m)
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
1.83
-------
9-
-1«B -BOO -60Q -MO -20G
200 43) KH 300 1000 1200 1403 1600
X-Direction [m]
Figure 5 - Layout of sources and receptors for CO screening tests
-------
.£ «:
-35 -30 -75 -70 -65 -50 -oo -50 -lo -4C -35 -JO -25 -20 -15 -10 -o 0 5 10 15 20 25 30
X-Direction [m]
W 15 M 55 &D 65 70 75 BO
Figure 6 - Close-up of receptor locations for CO screening tests
-------
Modeled Concentrations of CO (CAL3QHC vs. AERMOD)
Model Scenario
CAL3QHC Meteorology
CAL3QHC: Mix Ht = 1000 m, Ws = 1 m/s, Stability = 5
AERMOD: Mix Ht = 992 m, Ws = 1 m/s, L = 572.5 m
CAL3QHC Meteorology (w/ MAKEMET Ws)
CAL3QHC: Mix Ht = 1000 m, Ws = 10 m/s, Stability = 5
AERMOD: Mix Ht = 992 m, Ws = 10 m/s, L = 572.5 m
Highly Stable (Night)
CAL30HC: Mix Ht = 57 m, Ws = 1 m/s, Stability = 6
AERMOD: Mix Ht = 57 m, Ws = 1 m/s, L=3.3m
Moderately Unstable (Day)
CAL30HC: Mix Ht = 645 m, Ws = 2 m/s, Stability = 2
AERMOD: Mix Ht = 645 m, Ws = 2 m/s, L = -3.4m
CAL3QHC
(ppm)
2.1
0.2
2.7
0.5
CAL3QHC
fog/m3)
2,404.5
229.0
3,091.5
572.5
AERMOD
fog/m3)
706.0
563.0
4,088.7
714.6
Table 7 - Results of CAL3QHC and AERMOD CO screening tests
-------
-------
United States Office of Air Quality Planning and Standards Publication No. EPA- 454/B-15-002
Environmental Protection Air Quality Analysis Division [July, 2015]
Agency Research Triangle Park, NC
------- |