Technical Support Document (TSD) for
Replacement of CALINE3 with AERMOD for
Transportation Related Air Quality Analyses
-------
-------
EPA-454/B-16-006
December, 2016
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 Assessment 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
2.1 CALINE3 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 17
3.3 Summary of findings and recommended model 17
4. Acknowledgements 19
5. Additional information 19
References 20
Appendix A 21
Results from comparison of AERMOD and CAL3QHC for CO hot-spot screening for highway and
intersection projects 21
-------
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 - Intersection Configuration and Receptor Placement (U. S. EPA, 1995) 22
Figure 6 - Representation of Intersection in AERMOD (Links defined as LINE sources) 32
Figure 7 - Close-up of Eastbound Approach and Queue Lanes as LINE Sources in AERMOD 33
Figure 8 - Close-up of Eastbound Approach and Queue Lanes as VOLUME Sources in AERMOD Error!
Bookmark not defined.
-------
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 - Vehicle Volume and Average Vehicle Speed by Roadway Link 23
Table 4 - CO Emissions by Vehicle Type 24
Table 5 - CAL3QHC Initial Input Values 29
Table 6 - CAL3QHC Derived Values for Queue Links 29
Table 7 - AERMOD LINE Source Input Values 30
Table 8 - AERMOD VOLUME Source Input Values 31
Table 9 - Receptor locations for CO screening runs 32
Table 10 - Modeled Concentrations of CO (CAL3QHC vs. AERMOD) 35
-------
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 (80 FR 45340- 45387, July 29, 2015). 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 refined
model for mobile source dispersion modeling.
2. Background
The version of Appendix W that was published in 2005 (70 FR 68218- 68261, Nov. 9, 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
(N02), 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 (S02), 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 N02 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 CAL3QHCthat 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, until this final rule, 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 have been hosted on EPA's dispersion model website.
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. In the
final rule, CALINE3 has been delisted from Appendix W as a preferred model for refined analyses, but
CAL3QHC is available for CO screening until guidance has been developed for CO screening with
AERMOD.
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 N02) 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
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.
-------
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).
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 N02 concentrations as part of the 2008 Risk and Exposure Assessment for revision of the primary
N02 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, 2015)).
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 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 the final
revised 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. In 2005 when Appendix W was
promulgated, 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 provided justification for removing CALINE3 from the list in Appendix A in the 2016 Appendix W
final rule.
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 is applied to estimate vehicular
emissions. There can be significant uncertainties for model evaluation 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 for model evaluation, 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 uncertainties 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.
In (Heist, et al., 2013), a model inter-comparison was conducted, 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 RUNE, 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 - Model Performance Statistics from
the Idaho Falls Study and Table 2.
Table 1 - Model Performance Statistics from the Idaho Falls Study. Source; (Heist, et al., 20131.
Model
FB (0 is best)
NMSE (0 is best)
R (1 is best)
FAC2 (1 is best)
CALINE4
0.42
1.94
0.76
0.59
AERMOD - volume
0.38
1.26
0.84
0.59
AERMOD - area
0.32
1.25
0.82
0.59
ADMS
0.36
1.14
0.88
0.70
RLINE
0.23
0.96
0.85
0.73
Table 2 - Model Performance Statistics from the CALTRANS §9 Study. Source: (Heist, et al., 20131,
Model
FB (0 is best)
NMSE (0 is best)
R (1 is best)
FAC2 (1 is best)
CALINE3
0.25
2.26
0.29
0.45
CALINE4
0.19
0.86
0.47
0.68
AERMOD - volume
0.15
0.28
0.77
0.78
AERMOD - area
0.13
0.31
0.72
0.76
ADMS
0.09
0.20
0.78
0.85
RLINE
0.05
0.34
0.75
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 performed 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.
-------
QQ plot for Idaho Falls SF6 study
100
75
CD
W
¦O
-------
15000
CD
LL
CO
Xt 10000
oj
-------
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, 2015).
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, prior to the 2016 Appendix W final rule, AERMOD and CAL3QHCR had both been
allowed for refined analyses. Note that having the capability to internally parameterize queuing
emissions, like CAL3QHCR has, 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 CAL3QHCR to determine
ambient impacts.
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 replaced CALINE3 with AERMOD for all mobile source applications.
This 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 for refined modeling include:
• PM hot-spot analyses
• CO hot-spot analyses
• PM SIP attainment demonstrations
• PSD applications (PM, S02, N02, Pb, CO)
-------
• N02 near-road monitor siting and other potential future applications
4. Acknowledgements
The authors would like to acknowledge the intra-agency workgroup, specifically contributions from EPA
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 N02 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
AERMOD and CAL3QHC for CO hot-spot screening example
As noted in the 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,
a comparison 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. Version 15181 of AERMOD and version 04244 of CAL3QHC
were used for this model-to-model comparison.
The basis for these comparisons is modeled emissions for a two-way timed signalized intersection
(under capacity), adapted from Example 1 in the CAL3QHC User's Guide (U. S. EPA, 1995). The example
consists of a 4-lane, two-way main street intersecting a 2-lane, one-way local street. The main street
runs north and south with two northbound lanes and two southbound lanes. The local street runs from
west to east. The configuration of the intersection and receptor placement is illustrated in Figure 5,
taken from the CAL3QHC User's Guide. Traffic lanes were modeled as both LINE2 sources and lines of
adjacent VOLUME sources in AERMOD, for comparison to CAL3QHC.
2 The LINE source option in AERMOD uses the same dispersion parameterization as the AREA source option, but
with simplified inputs for rectangular area sources and was specifically added to AERMOD to aid in modeling for
transportation projects.
-------
pa wm
¦K^
M: *
«*e »
C-*V Hi
©
ft
fcrf «,5» k'»* * r^N
T
u
(¦<«««/ ; T"
S V, 4t -»
j L
J-
i*i
111
10. -:
t
*dHTt-
«*f *
C*S.»J
e
—
<-«.0i
0>
cam
.(ID 05
IDU. S'ttt*
C13iC Co
¦t: »
{-lie. -11"
BUBWt . « »>C'W
1
.t*D, -IB*
t r~
* ~_
«t: z
C-«a.-»r,
e
«IC ¦»
oj| ilt »ipr« * 'B*vfvtro *** £?c l
-L-
¦fT ¥
f-4* ¦ ",r
e
¥ ,
s !
i
I »®tf PLC*
IrfHw # **» - JO s
fltC 1
<4S,*-*5:
e
r-W^-r
* ! *
Figure 5 - Intersection Configuration and Receptor Placement (U. S. EPA, 1995)
-------
Emissions Modeling
MOVES2014a was used at the project scale to estimate emissions from this roadway configuration for a
single hour during 2016. MOVES runs were performed in the Inventory mode to produce total CO
emissions for the hour, then post-processed using the MySQL script, "CO_CAL3QHC_EF.sql" to generate
emissions in units of grams/vehicle-mile (g/veh-mi) for free-flow links and grams/vehicle-hour (g/veh-hr)
for queue links. The MySQL script is available in the MOVES post-processing pull-down menu. The
MOVES run specifications for this example include the following:
• Scale: Project scale, Inventory mode;
• Time Span: One hour: 2016, January, 7am-8am, Weekday;
• Geographic Bounds: Washtenaw County, Ml;
• Vehicle/Equipment: All valid fuel/vehicle combinations included;
• Pollutants/Processes: All running CO processes;
• Road Type: Urban Unrestricted (i.e., representing urban arterial roads); and
• Output: By link and hour.
Local inputs provided to the Project Data Manager include:
• Links: Vehicle volume and average speed are consistent with those published for Example 1 in
the CAL3QHC users' guide. Each intersection approach was modeled as two links: one link with
an average speed representing free-flow traffic, and one link with an average speed of 0 miles
per hour (mph) representing idling traffic (queue link). Each intersection departure was
modeled as single link. Thus, three links in each direction were modeled for a total of nine links
representing 6 approach lanes (3 links), 6 departure lanes (3 links), and 6 queue lanes (3 links).
All links were modeled at 0% grade. Vehicle volume in vehicles per hour (veh/hr) and average
speed for each road link are listed in Table 3.
Table 3 - Vehicle Volume and Average Vehicle Speed by Roadway Link
Street
Link
Vehicle Volume
(veh/hr)
Average Speed
(mph)
Main St. Northbound
(2 lanes)
Approach
1500
20
Queue
1500
0
Departure
1500
20
Main St. Southbound
(2 lanes)
Approach
1200
20
Queue
1200
0
Departure
1200
20
Local St. Eastbound
(2 lanes)
Approach
1000
20
Queue
1000
0
Departure
1000
20
-------
• Link Source: For all links, the national default distribution of VMT was used as a representative
fleet mix, generated from a separate 2016 national scale run.
Fleet mix: 8.6% Light Duty, 91.4% Heavy Duty.
• Source Type Age Distribution: The 2016 national default age distribution was used for all vehicle
types.
• Meteorology: A temperature of 30 degrees F and 70% humidity was used.3
• l/M: No l/M program.
Using the input specifications above, MOVES2014a produced the following emission rates for the
example intersection:
• Free-flow rate, for approach and departure links (at 20 mph): 5.11 g/veh-mi
• Idle rate, for queue links: 20.45 g/veh/hr.
Table 4 lists the total CO emissions, by vehicle type (Light Duty vs Heavy Duty), for the intersection.
Table 4 - CO Emissions by Vehicle Type
Vehicle type
CO Emissions (g/hr)
Heavy duty vehicles (8.6 %)
659
Light duty vehicles (91.4%)
7,004
Total
7,663
Air Quality Modeling
For the air quality modeling, the following combinations of source characteristics were included:
• Urban Dispersion (urban population of 1,000,000 used in AERMOD)
• All lanes at grade, flat terrain
• 6 Free Flow Lanes, (2 north bound lanes, 2 south bound lanes, and 2 east bound lanes with each
pair of lanes further divided into approach lanes and departure lanes with respect to the
intersection)
Each pair of free flow approach lanes was modeled as a single link in CAL3QHC, as were each
pair of departure lanes. Similarly, each pair of free flow approach lanes was modeled as a single
LINE source in AERMOD, as were each pair of departure lanes. Each individual approach lane
was modeled as a single line of adjacent VOLUME sources in AERMOD, as was each individual
departure lane.
• 6 Queue Lanes (2 north bound lanes, 2 south bound lanes, 2 east bound lanes)
Each pair of queue lanes was modeled as a single link in CAL30HC, a single LINE source in
AERMOD, and a single line of adjacent VOLUME sources in AERMOD.
3 This input does not affect CO emission rates for running emission processes, which are the only processes
occurring in this intersection example.
-------
• Lane Width = 10 feet (3.05 meters, m)
• Lane Length (approach + departure) = 2000 ft (610 m)
• Receptor Height = 6 ft (1.83 m)
• Surface Roughness = 1.25 m
• 36 wind directions, modeled every 10 degrees (10-360 degrees)
Source Characterization
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 (AG)
o 6 Free Flow Links (1 approach and 1 departure link in each of 3 directions)
¦ Link Width = (# lanes X 10 ft) + 20 ft, i.e. 2 x 10 ft +20 ft = 40 ft.
¦ Link Length = 1000 ft
o 3 Queue Links (1 in each of 3 directions)
¦ Link Width = # lanes x 10 ft, i.e., 2 x 10 ft = 20 ft
¦ Link Length = 1000 ft
¦ Saturation Flow Rate = 1600 veh/hr/lane (default)
¦ Signal Type = pre-timed (default)
¦ Arrival Type = average progression (default)
o Link Height = 0 ft
o Release Height: based on weighted emissions by vehicle mix (discussed below)
o Initial Vertical Dimension: based on weighted emissions by vehicle mix (discussed below)
• AERMOD:
o Flat terrain, source elevation = 0 m
o Each link in CAL3QHC modeled as a single LINE source in AERMOD
o Each free flow link in CAL3QHC modeled as two lines of VOLUME sources in AERMOD
(one per lane)
o Each queue link in CAL3QHC modeled as a single line of adjacent VOLUME sources in
AERMOD (one per pair of lanes)
LINE Sources
o 6 Free Flow LINE Sources (1 approach and 1 departure LINE in each of 3 directions)
¦ LINE Source Width = (# lanes X 10 ft) + 20 ft, i.e. 2 x 10 +20 ft = 40 ft (12.2 m)
¦ LINE source Length = 1000 ft (304.8 m)
o 3 Queue LINE Sources (1 LINE in each of 3 directions)
¦ LINE Source Width = (# lanes X 10 ft), i.e. 2 x 10 ft = 20 ft (6.1 m)
¦ LINE Source Length: based on traffic volume and capacity of approach (discussed
below)
VOLUME Sources
-------
Individual VOLUME sources have an equal length and width, and the size of all volume
sources that comprise a lane are equal. The number of VOLUME sources that comprise a
lane is dependent on the user specified width of the VOLUME source and the length of
the lane.
o 12 Free Flow Lines of VOLUME Sources (2 approach and 2 departure lines of VOLUME
sources in each of 3 directions, see Figures 6-8))
¦ VOLUME Source Width = lane width X 2, i.e., 10 ft (3.05 m) X 2 = 20 ft (6.1 m)
¦ Total Length of Line of VOLUME Sources = 1000 ft (304.8 m)
o 3 Queue Lines of VOLUME Sources (1 line of VOLUME sources in each of 3 directions)
¦ VOLUME Source Width = (# lanes X 10 ft), i.e. 2 x 10 ft = 20 ft (6.1 m)
¦ Total Length of Line of VOLUME Sources: based on traffic volume and capacity
of approach (discussed below)
Some of the input requirements for AERMOD are not directly known from the information provided for
Example 1 in the CAL3QHC user's guide, including: the calculated length of the queue lanes represented
as LINE or VOLUME sources in AERMOD, the release height, the initial vertical dispersion of the release,
and the initial lateral dimension and dispersion of the release (VOLUME sources). In CAL3QHC, the
queue length entered is the actual lane length, and the emission rate for queue links is specified in units
of g/veh-hr; whereas, the emission rates for free flow links are specified in g/veh-mi. In AERMOD, LINE
source emission rates are specified in units of grams/second-sq. meter (g/s-m2) and VOLUME source
emission rates are specified in grams/second (g/s). For queue links, CAL3QHC adjusts the queue lane
length and traffic volume based on the approach volume specified and the capacity of the approach.
With those adjusted parameters, CAL3QHC derives an equivalent emission factor expressed in g/veh-mi,
consistent with free flow links, and computes total emissions for each link in g/hr. The queue link
lengths calculated by CAL3QHC were then used in AERMOD to define the queue link geometry in order
to have equivalent emissions between the two model runs (see Tables 6 and 7 for more details).
AERMOD LINE Source : Emission Rate Conversions
To convert CAL3QHC emission rates for free flow links, expressed in g/veh-mi, to equivalent emission
rates for AERMOD LINE sources in g/s/m2, CAL3QHC emission rates were multiplied by the vehicle traffic
volume (# vehicles/hour) and link length (mi) to get g/hr. The product was then divided by 3600 s/hr
and the area of the LINE source to get an emission rate in units of g/s-m2. This is demonstrated in
equations 1 and 2.
Free Flow Vehicle Emission Rate:
[grams/vehicle-mile] * [# vehicles per hour] * [link length in miles] = [grams/hour] (Eq. 1)
Free Flow LINE Source Emission Rate:
[grams/hour] + [3600 seconds/hour] + [link area in sq. meters] = [grams/second/sq. meters](Eq. 2)
To facilitate consistency for this comparison of CAL3QHC with AERMOD, the length of the LINE sources
that represent queue links were obtained from the CAL3QHC output. The emission rates for the
CAL3QHC queue links, in g/veh-hr, were converted for AERMOD LINE sources by multiplying the rate by
-------
the number of vehicles per hour (adjusted by CAL3QHC) and the fraction of red light time. This product
was then divided by 3600 s/hr and by the link area. These calculations are shown in equations 3 and 4.
Queue Vehicle Emission Rate:
[grams/vehicle-hour] * [# vehicles] * [red light fraction] = [grams/hour] (Eq. 3)
Queue LINE Source Emission Rate:
[grams/hour] + [3600 seconds/hour] + [link area in sq. meters] = [grams/second/sq. meters](Eq. 4)
AERMOD VOLUME Source: Emission Rate Conversions
To convert CAL3QHC emission rates for to equivalent emission rates for individual AERMOD VOLUME
sources in g/s, the product of Equation 1 (for free flow links) or Equation 3 (for queue links) was divided
by 3600 s/hr to get the total emission rate for the link in g/s (Equation 5). For free flow links (approach
and departure lanes), the quotient from Equation 5 was then divided by 2 to get the total emissions for
a single approach or departure lane (i.e., the line of VOLUME sources that comprise a lane) since the
CAL3QHC free flow links each represent a pair of lanes (Equation 6). The emission rate for each
individual VOLUME source that comprises a free flow approach or departure lane or a pair of queue
lanes was obtained by dividing the total emission rate for the line of VOLUME sources (Equation 5 or
Equation 6) by the number of individual VOLUME sources that make up the lane as shown in Equation 7.
Link Emission Rate: [grams/hour] + [3600 seconds/hour] = [grams/second] (Eq. 5)
Free Flow Lane Emission Rate: [grams/second] + 2 = [grams/second] (Eq. 6)
VOLUME Source Emission Rate: [grams/second] + # volume sources = [grams/second] (Eq. 7)
AERMOD LINE and VOLUME Source: Initial Vertical Dimension (Szinit)
When defining a LINE or VOLUME sources in AERMOD, the user has the option for LINE sources to
specify the initial vertical dimension (Szinit), in meters. This input is required for VOLUME sources. This
value was calculated as 1.7 multiplied by the average vehicle height (light duty = 1.53 m, heavy duty =
4.0 m), weighted by the contribution of emissions from light duty (91.4%) and heavy duty (8.6%)
vehicles. The weighted vehicle height for this example is 2.96 meters. The weighted height was then
divided by 2.15, per the AERMOD user's guide for elevated releases. The final value for the Szinit is 1.38
meters. These calculations are demonstrated in equations 8 through 11.
Light Duty: 1.53 m * 1.7 = 2.6 m (Eq. 8)
Heavy Duty: 4.0 m * 1.7 = 6.8 m (Eq. 9)
Weighted Vehicle Height: (0.914 * 2.6 m) + (0.86 * 6.8 m) = 2.96 m (Eq. 10)
Initial Vertical Dimension (Szinit): 2.96 m -r 2.15 = 1.38 m (Eq. 11)
-------
AERMOD LINE and VOLUME Source: Release Height
The release height, per hot-spot guidance, was computed as 1/2 of the weighted vehicle height
(equations 8 through 10) as shown in equation 9.
Re/ease Height: 0.5 * 2.96 m = 1.48 m (Eq. 9)
AERMOD VOLUME Source: Initial Lateral Dimension (Syinit)
Another required input for VOLUME sources is the initial lateral dimension (Syinit), in meters. Syinit for
adjacent volume sources, per the AERMOD user's guide, is calculated as the VOLUME width divided by
2.15 as in Equation 10.
Initial Lateral Dimension (Syinit): VOLUME source width (m) -r 2.15 = m (Eq. 11)
The full details of the source characteristics for CAL3QHC and AERMOD are provided in Table 5, Table 6,
Table 7, and Table 8. The receptor locations, as defined in Example 1 of the CAL3QHC user's guide are
provided in Table 9. Table 5 and Table 6 show the differences in the values input into CAL3QHC
(Table 5) and how the model adjusts several of the input values for the queue links (Table 6), as
discussed above, such as the link length and traffic volume and derives an emission factor. Figure 6
through Figure 8 illustrate the intersection represented as LINE and VOLUME sources in AERMOD. The
eastbound free-flow link/lanes and queue link/lanes are shaded in these two figures for reference.
-------
Table 5 - CAL3QHC Initial Input Values
Link
XL(1)
(ft)
YL(1)
(ft)
XL(2)
(ft)
YL(2)
(ft)
Link
Length
(ft)
Link
Length
(mi)
Link
Width
(ft)
VPHL
(veh/hr)
EFL
(g/veh-mi)
(g/veh-hr)
Avg. Signal
Cycle Length
(s)
Avg. Red
Time Length
(s)
Clearance
Lost Time
(s)
NB Approach
10
-1000
10
0
1000
0.1894
40
1500
5.11
-
-
-
NB Queue
10
-10
10
-1000
1000
0.1894
20
1500
20.45
90
40
3
NB Depart
10
0
10
1000
1000
0.1894
40
1500
5.11
-
-
-
SB Approach
-10
1000
-10
0
1000
0.1894
40
1200
5.11
-
-
-
SB Queue
-10
10
-10
1000
1000
0.1894
20
1200
20.45
90
40
3
SB Depart
-10
0
-10
-1000
1000
0.1894
40
1200
5.11
-
-
-
EB Approach
-1000
0
0
0
1000
0.1894
40
1000
5.11
-
-
-
EB Queue
-20
0
-1000
0
1000
0.1894
20
1000
20.45
90
50
3
EB Depart
0
0
1000
0
1000
0.1894
40
1000
5.11
-
-
-
Table 6 - CAL3QHC Derived Values for Queue Links
Link
Link
Link
Total
XL(1)
YL(1)
XL(2)
YL(2)
Length
Length
Width
VPHL
EFL
Emissions
Link
(ft)
(ft)
(ft)
(ft)
(ft)
(mi)
(ft)
(veh/hr)
(g/veh-mi)
(g/hr)
NB Approach
10
-1000
10
0
1000
0.1894
40
1500
5.11
1,451.70
NB Queue
10
-10
10
-238.5
228.5
0.0433
20
49
100
212.05
NB Depart
10
0
10
1000
1000
0.1894
40
1500
5.11
1,451.70
SB Approach
-10
1000
-10
0
1000
0.1894
40
1200
5.11
1,161.36
SB Queue
-10
10
-10
141.2
131.2
0.0248
20
49
100
121.76
SB Depart
-10
0
-10
-1000
1000
0.1894
40
1200
5.11
1,161.36
EB Approach
-1000
0
0
0
1000
0.1894
40
1000
5.11
967.80
EB Queue
-20
0
-165.4
0
145.4
0.0275
20
61
100
167.98
EB Depart
0
0
1000
0
1000
0.1894
40
1000
5.11
967.80
Values altered or derived by CAL3QHC are highlighted in yellow.
Total: 7,663.54
-------
Table 7 - AERMOD LINE Source Input Values
LINE
LINE
Release
Emission
Total
Xs(l)
Ys(l)
XL(2)
Ys(2)
Zs
Length
Width
Height
Szinit
Rate
Emissions
LINE
(m)
(m)
(m)
(m)
(m)
(m)
(m)
(m)
(m)
(g/s-m2)
(g/hr)
NB Approach
3.0
-304.8
3.0
0.0
0.0
304.8
12.2
1.48
1.38
1.085E-04
1,451.70
NB Queue
3.0
-3.0
3.0
-72.7
0.0
69.6
6.1
1.48
1.38
1.387E-04
212.05
NB Depart
3.0
0.0
3.0
304.8
0.0
304.8
12.2
1.48
1.38
1.085E-04
1,451.70
SB Approach
-3.0
304.8
-3.0
0.0
0.0
304.8
12.2
1.48
1.38
8.681E-05
1,161.36
SB Queue
-3.0
3.0
-3.0
43.0
0.0
40.0
6.1
1.48
1.38
1.387E-04
121.76
SB Depart
-3.0
0.0
-3.0
-304.8
0.0
304.8
12.2
1.48
1.38
8.681E-05
1,161.36
EB Approach
-304.8
0.0
0.0
0.0
0.0
304.8
12.2
1.48
1.38
7.234E-05
967.80
EB Queue
-6.1
0.0
-50.4
0.0
0.0
44.3
6.1
1.48
1.38
1.727E-04
167.98
EB Depart
0.0
0.0
304.8
0.0
0.0
304.8
12.2
1.48
1.38
7.234E-05
967.80
Total:
7,663.54
-------
Table 8 - AERMOD VOLUME Source Input Values
VOLUME Line
Xs(l)
(m)
Ys(l)
(m)
XL(2)
(m)
Ys(2)
(m)
Zs
(m)
VOL Line
Length
(m)
#of
VOLUME
Sources
Source
Width
(m)
Release
Height
(m)
Szinit
(m)
Syinit
(m)
Emission
Rate
(g/s-m2)
Total
Emissions
(g/hr)
NB Approach-1
3.0
-304.8
3.0
0.0
0.0
304.8
50
6.1
1.48
1.38
2.84
0.20163
725.85
NB Approach-2
1.5
-304.8
1.5
0.0
0.0
304.8
50
6.1
1.48
1.38
2.84
0.20163
725.85
NB Queue
3.0
-3.0
3.0
-72.7
0.0
69.6
11
6.1
1.48
1.38
2.84
0.05890
212.05
NB Depart-1
4.6
0.0
4.6
304.8
0.0
304.8
50
6.1
1.48
1.38
2.84
0.20163
725.85
NB Depart-2
1.5
0.0
1.5
304.8
0.0
304.8
50
6.1
1.48
1.38
2.84
0.20163
725.85
SB Approach-1
-4.6
304.8
-4.5
0.0
0.0
304.8
50
6.1
1.48
1.38
2.84
0.16130
580.68
SB Approach-2
-1.5
304.8
-1.5
0.0
0.0
304.8
50
6.1
1.48
1.38
2.84
0.16130
580.68
SB Queue
-3.0
3.0
-3.0
43.0
0.0
40.0
7
6.1
1.48
1.38
2.84
0.03382
121.76
SB Depart-1
-4.6
0.0
-4.6
-304.8
0.0
304.8
50
6.1
1.48
1.38
2.84
0.16130
580.68
SB Depart-2
-1.5
0.0
-1.5
-304.8
0.0
304.8
50
6.1
1.48
1.38
2.84
0.16130
580.68
EB Approach-1
-304.8
-1.5
0.0
-1.5
0.0
304.8
50
6.1
1.48
1.38
2.84
0.13442
483.90
EB Approach-2
-304.8
1.5
0.0
1.5
0.0
304.8
50
6.1
1.48
1.38
2.84
0.13442
483.90
EB Queue
-6.1
0.0
-50.4
0.0
0.0
44.3
7
6.1
1.48
1.38
2.84
0.04666
167.98
EB Depart-1
0.0
-1.5
304.8
-1.5
0.0
304.8
50
6.1
1.48
1.38
2.84
0.13442
483.90
EB Depart-2
0.0
1.5
304.8
1.5
0.0
304.8
50
6.1
1.48
1.38
2.84
0.13442
483.90
Total: 7,663.54
-------
Table 9 - Receptor locations for CO screening runs
Receptor Name
x(ft)
Y (ft)
z(ft)
X (m)
Y(m)
Z (m)
REC 1 (SE CORNER)
45.
-35.
6.00
13.72
-10.67
1.83
REC 2 (SW CORNER)
-45.
-35.
6.00
-13.72
-10.67
1.83
REC 3 (NW CORNER)
-45.
35.
6.00
-13.72
10.67
1.83
REC 4 (NE CORNER)
45.
35.
6.00
13.72
10.67
1.83
REC 5 (E MID-MAIN)
45.
150.
6.00
13.72
-45.72
1.83
REC 6 (W MID-MAIN)
-45.
150.
6.00
-13.72
-45.72
1.83
REC 7 (N MID-LOCAL)
-150.
35.
6.00
-45.72
10.67
1.83
REC 8 (S MID-LOCAL)
-150.
-35.
6.00
-45.72
-10.67
1.83
Queue Lanes
Single LINE Source
(Eastbound)
Free-flow Lanes
Single LINE Source
(Eastbound Approach)
Receptor
-400 -350 -300 -250 -200 -150 -100 -50
I I I I I
50 100 150 200 250 300 350
Meters
Figure 6 - Representation of Intersection in AERMOD (Links defined as LINE sources)
-------
Free-flow Lanes
Single LINE Source
(Eastbound Approach)
Queue Lanes
Single LINE Source
(Eastbound)
'Receptor
i ii | I I i ]ri i | I I 11 r i i | i i i | r i i | i i i | i i i | i i i | i r i | i i I | I r i | i i i |i i i | ii i | i ii | i i i| i i i | I i i | I I I | I i i | I I I | I ii | I
-75 -70 -65 -60 -65 -50 -45 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 35 40
Meters
Figure 7 - Close-up of Eastbound Approach and Queue Lanes as LINE Sources in AERMOD
Meteorology
The meteorological parameters accepted by CAL3QHC are minimal and include wind speed, wind
direction, mixing height, and Pasquill-Gifford (P-G) stability class (1-6). CAL3QHC also requires an
estimate of the surface roughness length, independent of wind direction. The surface roughness length
was set to 1.25 m for both CAL3QHC and AERMOD to represent an urban setting.
CAL3QHC is used for screening analyses which focuses on using "worst-case" meteorology to estimate
the worst possible 1-hour concentrations. The example in the CAL3QHC User's Guide, which is
consistent with EPA's current guidance for CO screening analyses (U. S. EPA, 1992), consists of a 1000 m
mixing height, a 1 m/s wind speed, a stable atmosphere (P-G stability class 5). However, these
assumptions are somewhat unrealistic. For near 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. Thus,
a mixing height of 1000 m could not physically occur in the atmosphere with the accompanying low
wind and stable conditions. Nonetheless, 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.) In contrast, AERMOD is sensitive to mixing
height, as well as other boundary layer parameters, and it is important that the boundary layer
parameters are consistent with the measured meteorological conditions.
For these examples, a range of meteorological conditions were modeled. One hour of representative
meteorology was developed for each of the six stability classes and replicated for 36 wind directions by
varying the wind direction every 10 degrees from 10 to 360 degrees. Separate model runs were
performed for each of the stability classes. The meteorology was created using the MAKEMET tool
provided with AERSCREEN (U. S. EPA, 2011). MAKEMET generates an array of realistic meteorological
conditions for physical parameters that are typically observed (e.g. wind speed, temperature, and cloud
cover) and computes the boundary layer parameters required by AERMOD (e.g., mixing height, Monin-
Obukhov length (L), and surface friction velocity) from the physical parameters for each set of
conditions. Because MAKEMET does not derive or report the stability class required by CAL3QHC,
Golder's4 method for relating L to stability class was used to assign a stability class to each unique set of
meteorological conditions generated by MAKEMET. For all but stability class 4, a single set of conditions
was selected from the MAKEMET as representative of the stability category. Two sets of conditions
were selected for stability class 4 as explained below.
Within each stability class, a combination of the value of L and the mixing height, if needed, was used as
the basis for selecting representative meteorology. With the exception of stability class 4, the selection
was made by taking the set of conditions for which the value of L is closest to the computed median
value of L within the stability class. If this resulted in more than one set of conditions, then from that
subset the selection was made by taking the set of conditions for which the mixing height is closest to
the computed median mixing height within the stability class.
For all stability classes except 4, the values of L, by definition, are either all positive or all negative.
However, neutral stability is characterized by either large positive values of L or large negative values of
L (i.e., as L gets farther from zero). Because L in neutral conditions can be both positive and negative,
the median value of L across all neutral conditions would likely be close to zero which does not
represent neutral conditions. Therefore, two sets of neutral conditions were selected, one for which L is
positive and one for which L is negative. Rather than using the median value of L as the criteria for
selection, the 10th percentile was used for negative values of L and the 90th percentile was used for
positive values of L. Subsequently, if this resulted in more than one unique set of conditions, then the
median value of the mixing height was applied as described before. In addition to the seven CAL3QHC
4 Golder, D., 1972. Relations among stability parameters in the surface layer. Boundary Layer
Meteorology 3, 47-58.
-------
and AERMOD model runs for a single stability category, an eighth AERMOD run was performed using all
of the meteorology generated with MAKEMET.
Modeling results
The results from these comparisons are shown in Table 10. For the model-to-model comparisons, the
worst-case conditions of those modeled proved to be stability class 6 with a wind speed of 0.5 m/s and a
mixing height of 156 meters (i.e., stable, low wind speed, and low mixing height). In this case, AERMOD
predicted a concentration that is about half the concentration predicted by CAL3QHC. The lowest
concentrations occurred under neutral stability with a high wind speed (6.0 m/s) and high mixing height
(2874 m). In this case, AERMOD predicted a higher concentration. For the other conditions, there is a
mix of results. In some cases, AERMOD predicted a higher concentration than CAL3QHC and in others, a
lower concentration. The modeled concentrations for stability class 3 (slightly unstable) are nearly
identical. Note that the mixing height for each of these model runs is greater than 100 meters, where
CAL3QHC's sensitivity to mixing height is not significant. Modifying the mixing height in the CAL3QHC
control file for any of these cases is expected to have little if any effect on the maximum modeled
concentration, whereas AERMOD is more sensitive to the mixing height but also takes into account
other boundary layer parameters whose values reflect the measured physical parameters from which
they are computed.
Table 10 - Modeled Concentrations of CO (CAL3QHC vs. AERMOD)
Wind
Speed
(m/s)
Mixing
Height
(m)
P-G
Stability
Class
Monin-
Obukhov
Length
(m)
CAL3QHC
(ppm)
CAL3QHC
(ng/m3)
AERMOD
LINE
(ng/m3)
AERMOD
VOLUME
(g/m3)
Ratio
AERMOD
LINE/CAL3
Ratio
AERMOD
VOL7CAL3
0.5
766.
1
-5.8
0.90
1030.5
945.2
578.6
0.92
0.56
2.0
803.
2
-33.2
0.20
229.0
378.4
287.4
1.65
1.26
2.0
1439.
3
-153.5
0.40
458.0
458.7
397.0
1.00
0.87
6.0
2874.
4
-3257.1
0.10
114.5
171.1
156.1
1.49
1.36
2.5
1993.
4
6748.5
0.40
458.0
292.6
210.2
0.64
0.46
1.0
470.
5
114.1
0.90
1030.5
815.6
495.3
0.79
0.48
0.5
156.
6
27.9
2.20
2519.0
1279.9
732.6
0.51
0.29
All Meteorology
1713.8
1012.2
-------
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-454/B-16-006
Environmental Protection Air Quality Assessment Division December, 2016
Agency Research Triangle Park, NC
------- |