Incorporation and Evaluation of the RLINE Source Type in AERMOD For Mobile Source Applications


* O \

KWJ

*1 PRO^^

Incorporation and Evaluation of the RLINE
Source Type in AERMOD For Mobile Source
Applications

-------

-------
EPA-454/R-23-011
October 2023

Incorporation and Evaluation of the RLINE Source Type in AERMOD For Mobile Source

Applications

U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Research Triangle Park, NC

-------
Preface

This document provides details for the reformulation of the RLINE Source Type for AERMOD
version 23132 as part of the 2023 revisions to the Guideline on Air Quality Models. This
reformulation process was needed based on user feedback through the BETA release and testing
period after the AERMOD vl9191 release. The goals of the reformulation were to "harmonize"
multiple aspects of the dispersion calculations performed for the RLINE source type with
calculations performed for the other AERMOD source types, such as AREA and VOLUME.
Once this harmonization was complete, the model was reevaluated with the original evaluation
databases, and it was determined that the parameterization for the horizontal and lateral spread
coefficients would need to be adjusted to maintain model performance. This document describes
the methods and results from this reformulation process.

li

-------
Acknowledgments

The model code changes and analyses for this document were performed by WSP Environment
& Infrastructure, Inc under contract number GS-00F-314CA 68HERH20F0394. EPA thanks the
efforts of Michelle Snyder, Laura Kent, and Rebecca Miller for this project. EPA also
acknowledges the contributions of The Federal Highway Administration through an Interagency
Agreement (Agreement number 693JJ320N300057) to complete the RLINE reformulation in
AERMOD. The Office of Air Quality Planning & Standards also thanks David Heist and R.
Chris Owen of EPA's Office of Research and Development and Laura Berry of EPA's Office of
Transportation & Air Quality for their efforts in updating RLINE.

111

-------
Table of Contents

1.0 Introduction	1

2.0 Changes to the RLINE Source:	2

2.1	Wind speed	2

2.2	Harmonization with Other AERMOD Sources	2

2.3	Re-evaluation of Dispersion Coefficients Using Optimization Techniques	3

3.0 Evaluation	7

3.1	Previous Field Study Evaluations with Idaho Falls & Caltrans 99 Tracer Experiments . 7

3.1.1	Idaho Falls Roadway Study	7

3.1.2	Caltrans 99 Highway Study	9

3.2	New Field Study Evaluations with GM Sulfate Dispersion Experiment	10

3.2.1 GM Sulfate Dispersion Experiment	10

3.3	Hotspot Analyses Model Intercomparisons	16

3.3.1	Project A - PM2.5 Hot-spot Analysis	16

3.3.2	Project B: PM10 Hot-spot Analysis	22

4.0 Conclusions	28

5.0 References	29

6.0 Appendix A	30

6.1	Project A PM2.5	30

6.2	Project B PM10	35

iv

-------
List of Figures

Figure 1: The 6-panel figure shows an example of the R-squared value for multiple combinations
of the a, bs, bu, c, ds, and du coefficients. Where the "best-fit" R-squared value is shown as a red

dashed line in all panels 5

Figure 2: Idaho Falls 2009 (circles) and Prairie Grass (triangles) normalized concentration vs.
x/|L|. The solid and dashed lines represent the new az equations for stable and convective
conditions for a range of u*/Ue values which are representative of the u*/Ue values for the Idaho

Falls and Prairie Grass field studies 6

Figure 3: Idaho Falls Study layout. Source is indicated with vertical line along x/Hb = 0 (from
y/Hb = -4.5 to 4.5). Filled circles show the locations of bag samplers. North is indicated by the

direction of arrow 8

Figure 4: Modeled vs measured SF6 concentration (in ppb) using the RLINE source type for all
test days at Idaho Falls. AERMOD ver. 22112 (left four plots) and ver. 23132 (right four plots).
Each symbol represents a 15 min. average. The three lines in each plot are the 1:2, 1:1, and 2:1

lines 8

Figure 5: Caltrans 99 Highway Study layout. Tracer was released along the roadway with
samplers arrayed on either side extending from 50 m to 200 m with four additional samplers in

the median of the roadway 9

Figure 6: Modeled vs measured normalized concentrations using the RLINE source type for the
Caltrans Highway 99 tracer study for receptors located downwind of the roadway. AERMOD
version 22112 (left plot) and version 23132 (right plot). Each symbol represents a 30 min.

average. The three lines in each plot are the 1:2, 1:1, and 2:1 lines 10

Figure 7 Onsite wind speed (color) and direction by height (panels) for the average of

measurement towers 1,5, and 6 11

Figure 8 General Motors Proving Ground in Milford, MI 12

Figure 9: Survey of Sampling Area (reprint Cadle et al. 1976 Figure 4) 12

Figure 10 AERMOD v23132 run with AERMET v23132 NLCD 2011 surface characteristics
compared with GM experiment observations ([j,g/m3). Panels labeled by measurement height and

colored by tower number 15

Figure 11 AERMOD v23132 run with AERMET v23132 NLCD 2011 surface characteristics
compared with GM experiment observations ([j,g/m3). Colored by wind speed (m/s, left) and

wind direction (right) 16

Figure 12 Project layout for Project A hot spot analysis. RLINE emission source drawn as lines
with emission rate shaded (black - blue - green). Receptor locations indicated with X's and
RLINE v22112 model concentrations shaded (purple - yellow). Design concentration receptor

location indicated with red circle 17

Figure 13 Project layout for the Project A hot spot analysis. RLINE emission source drawn as
lines with emission rate shaded (black - blue - green). Receptor locations indicated with X's and
RLINE v23132 model concentrations shaded (purple - yellow). Design concentration receptor
location indicated with red circle 17

-------
Figure 14 Comparison of RLINE 8th highest 24-hour concentrations of PM2.5 ([j,g/m3) for all
receptors between source type (AREA, left column and VOLUME, right column) and AERMOD

version (top, v22112 and bottom, v23132). Model runs for Project A	19

Figure 15 Comparison of RLINE 8th highest 24-hour concentrations of PM2.5 ([j,g/m3) for all

receptors between AERMOD versions. Model runs for Project A	19

Figure 16 Comparison of RLINE 5-year annual average concentrations of PM2.5 ([j,g/m3) for all
receptors between source type (AREA, left column and VOLUME, right column) and AERMOD

version (top, v22112 and bottom, v23132). Model runs for Project A	20

Figure 17 Comparison of RLINE 5-year annual average concentrations of PM2.5 ([j,g/m3) for all

receptors between AERMOD versions. Model runs for Project A	20

Figure 18 Comparison of RLINE hourly concentrations of PM2.5 ([j,g/m3) for design
concentration dates (shape) and wind speed (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs

for Project A	22

Figure 19 Comparison of RLINE hourly concentrations of PM2.5 ([j,g/m3) for design
concentration dates (shape), wind speed (color, left), and inverse Monin-Obukhov length (color,

right) between AERMOD versions. Model runs for Project A	22

Figure 20 Project layout for the North Project B Corridor hot spot analysis. RLINE emission
source drawn as lines with emission rate shaded (black - blue - green). Receptor locations
indicated with X's and RLINE v23132 model concentrations shaded (purple - yellow). Design

concentration receptor location indicated with red circle	23

Figure 21 Project layout for the North Project B Corridor hot spot analysis. RLINE emission
source drawn as lines with emission rate shaded (black - blue - green). Receptor locations
indicated with X's and RLINE v22112 model concentrations shaded (purple - yellow). Design

concentration receptor location indicated with red circle	24

Figure 22 Comparison of RLINE 6th highest 24-hour concentrations of PM10 ([j,g/m3) for all
receptors between source type (AREA, left column and VOLUME, right column) and AERMOD

version (top, v22112 and bottom, v23132). Model runs for Project B	25

Figure 23 Comparison of RLINE 6th highest 24-hour concentrations of PM10 ([j,g/m3) for all

receptors between AERMOD versions. Model runs for Project B	26

Figure 24 Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and wind speed (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs

for Project B	27

Figure 25 Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape), wind speed (color, left), and inverse Monin-Obukhov length (color,

right) between AERMOD versions. Model runs for Project B	27

Figure 26: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and friction velocity (color) between source type (AREA, left

vi

-------
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).

Model runs for Project A	30

Figure 27: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and friction velocity (color) between AERMOD versions. Model

runs for Project A project	31

Figure 28: Comparison of RLINE hourly concentrations of PM2.5 ([j,g/m3) for design
concentration dates (shape) and surface heat flux (color) between source type (AREA, left
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).

Model runs for Project A	31

Figure 29: Comparison of RLINE hourly concentrations of PM2.5 ([j,g/m3) for design
concentration dates (shape) and surface heat flux (color) between AERMOD versions. Model

runs for Project A	32

Figure 30: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and wind direction (color) between source type (AREA, left column
and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model

runs for Project A	32

Figure 31: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and wind direction (color) between AERMOD versions. Model runs

for Project A project	33

Figure 32: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and inverse Monin-Obukhov length (color) between source type
(AREA, left column and VOLUME, right column) and AERMOD version (top, v22112 and

bottom, v23132). Model runs for Project A	33

Figure 33: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and hour of day (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs

for Project A	34

Figure 34: Comparison of RLINE hourly concentrations of PM2.5 ((J,g/m3) for design
concentration dates (shape) and hour of day (color) between AERMOD versions. Model runs for

Project A	34

Figure 35 Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and friction velocity (color) between source type (AREA, left
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).

Model runs for Project B	35

Figure 36. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and friction velocity (color) between AERMOD versions. Model

runs for Project B	36

Figure 37. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and surface heat flux (color) between source type (AREA, left

vii

-------
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).

Model runs for Project B	37

Figure 38. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and surface heat flux (color) between AERMOD versions. Model

runs for Project B	38

Figure 39. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and wind direction (color) between source type (AREA, left column
and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model

runs for Project B	39

Figure 40. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and wind direction (color) between AERMOD versions. Model runs

for Project B	40

Figure 41. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and inverse Monin-Obukhov length (color) between source type
(AREA, left column and VOLUME, right column) and AERMOD version (top, v22112 and

bottom, v23132). Model runs for Project B	41

Figure 42. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and hour of day (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs

for Project B	42

Figure 43. Comparison of RLINE hourly concentrations of PM10 ([j,g/m3) for design
concentration dates (shape) and hour of day (color) between AERMOD versions. Model runs for
Project B	43

viii

-------
List of Tables

Table 1: Comparison of RLINE oy and oz coefficient values and ranges tested	4

Table 2 Project A design concentrations (PM2.5, (J,g/m3)	18

Table 3 Project A design concentration receptors and dates used for hourly analysis	21

Table 4 Project B design concentrations (PM10, (J,g/m3)	25

Table 5 Project B design concentration receptors and dates used for hourly analysis	26

IX

-------
1.0 Introduction

The R-LINE model was originally developed by the US EPA's Office of Research and
Development as a stand-alone model dubbed the "Research Line Model", or R-LINE model. R-
LINE is based on a numerical integration scheme that approximates the line source as a series of
point sources with the model formulation described by Snyder et al., (2013). The original
formulation generally placed an emphasis on concentrations closest to the line source since this
is where maximum impacts are expected from roadway emissions. The near-surface dispersion
algorithms are based on new formulations of horizontal and vertical dispersion within the
atmospheric surface layer, details of which are described by Venkatram et al., (2013). The
dispersion curves for the POINT, VOLUME, and AREA source types in AERMOD are based on
the Prairie Grass study (Barad, 1958). The new dispersion curves in R-LINE were based on a
reevaluation of the Prairie Grass study, as well as new tracer dataset from Idaho Falls (Finn et al.,
2010), with the new formulation based on eddy diffusivity and mass conservation. The model
performance was evaluated by Heist (2013) which compared the performance for the new R-
LINE model against the AREA and VOLUME source types in AERMOD, the ADMS model,
and the CALINE3 & 4 models.

Based on the good performance of the R-LINE model, the EPA incorporated the R-LINE model
as a new source type in AERMOD in 2019. The new "RLINE" source type was directly
integrated into AERMOD version 19191 with no changes from the released version 1.2 of R-
LINE. The source was designated as a one of the first "BETA" options in AERMOD, meaning
that it had sufficient testing, documentation, and evaluations to potentially be used in a
regulatory context, with approval from the EPA Regional Office and in concurrence with the
EPA's Model Clearinghouse.

Since its release in 2019, significant testing has been conducted by the user community as well
as within EPA. Through this testing in the BETA phase of the release, several areas for model
improvement were identified. First, it was noted that several aspects of the RLINE formulation
were not well matched with the formulation of other AERMOD source types (e.g., the weighting
factor for plume meander differed slightly between RLINE and the POINT and VOLUME
sources). Second, testing with real-world scenarios, where source-receptor distances are much
greater than in the field studies used to develop and evaluate RLINE, it was noted that RLINE
concentrations deviated significantly from concentrations from the VOLUME and AREA
sources. The harmonization of the RLINE source with the existing AERMOD source types
results in changes in the dispersion calculations that have implications for the original model
formulation and evaluations. As a result, the EPA reexamined the dispersion curves originally
formulated by Venkatram et al., (2013), to refit the dispersion parameters determined previously
to the developmental field data, as well as improve the model-to-model based performance from
the real-world scenarios. This document provides details of the updated formulation, the
methodology for optimizing the dispersion coefficients, evaluation of the updated RLINE against
tracer field study data, and a model intercomparison for two real-world cases.

-------
2.0 Changes to the RLINE Source:

Modifications to the RLINE formulation occurred in three main areas: (1) Wind Speed
calculation, (2) Harmonization with AERMOD sources, and (3) Dispersion Coefficients. The
modifications were made in this order, with the wind speed and harmonization changes made
first, then the reexamination of the parameters used in the vertical and lateral dispersion
calculations. All these modifications were necessary to bring the RLINE source type into better
agreement with other AERMOD source types and simultaneously not degrade the previous
evaluation database results.

2.1 Wind speed

RLINE was developed under the assumption a vector-average wind speed would be supplied as
input (i.e., a wind speed derived from time-averaged components of the wind speed vector) to the
model. The supplied wind speed was converted to a scalar-averaged value using an approximate
relationship (Merceret, 1995), and the resulting wind speed was used in dispersion calculations
within the model. When RLINE was integrated into the AERMOD model as a new source type
wind speeds did not need to be converted because winds in the available AERMET surface files
were scalar-averaged wind speeds (i.e., speeds computed by time-averaging instantaneous wind
speeds over time). To correct this issue, parts of the RLINE code where wind speeds had been
enhanced with the following equation:

wsscalar = Jwslct0r + 2 £7 2

were removed, where WSscaiar is the scalar-averaged wind speed, WSvector is the vector-
averaged wind speed, and ov is the root-mean-square of the lateral velocity fluctuations.

RLINE selects the advecting wind speed (WSM0ST) in the dispersion calculation from a profile
generated using Monin-Obhukov Similarity Theory (MOST). The profile is adjusted using a
multiplicative factor, fws,adj-. to ensure it passes through the measured wind speed at the
reference height. Thus, /ws,ady = WSscalar(zref)/WSM0ST(zref). In addition, there is a
minimum wind speed enforced in the RLINE calculations such that WSM0ST > WSmin.
However, in previous versions of AERMOD, the value of WSM0ST could fall below WSmin for
cases when fws,adj < 1, because of the sequencing of applying fws,adj and the minimum wind
speed criteria. This sequencing has been corrected in version 23132 to ensure WSM0ST >

WSmin, even for cases where fWSAdj < 1

2.2 Harmonization with Other AERMOD Sources

To better integrate the RLINE source type within the AERMOD framework, several changes
were made to call native AERMOD functions, when possible, to calculate required parameters.
To that end, RLINE now uses the gridded value of ov used by other source types in AERMOD,

-------
rather than calculating it in RLINE's own subroutines. Though the calculation is the same, any
future changes to these calculations will be applied to all source types uniformly. Additionally,
the RLINE source type now uses native AERMOD functions to calculate the fraction of the
plume attributable to meander. This change introduces a gradual change in this fraction with
increasing distance from the source, but matches the value originally used in RLINE for near-
source calculations. The final change in this category involves the calculation of the vertical
plume width,  0.0

ay = (x + duw) ,2	L< ao

az = a — x(l + bs — (—) ^	L > 0.0

Z Ueff \ SUeff\\L\J J

U* (+ , i u* X \

= a7J~x 1 + bu7T~ in

Ueff \	eff 1^1/

Since the lateral spread is dependent on the vertical spread, the optimization of the a, bs, bu, c, ds,
and du coefficients occurred together. The optimization process that was used involved previous
databases, including Idaho Falls and Prairie Grass, and intelligently and iteratively solved for the
coefficients in the vertical and lateral spread. The optimal coefficients were found using the R
genetic algorithm "GA" package (https://cran.r-project.org/web/packages/GA/GA.pdf). This
algorithm uses an intelligent method to reduce the solution space of a multivariate set of
equations to optimize a metric, analogous to a best fit line optimizing coefficients to reduce
residuals, usually reported as an R-squared value.

The Idaho Falls database is lateral spread, sigma-y, independent; the dispersion of a line source
is only dependent on the downwind distance and the equation for vertical spread, sigma-z, given
an "infinite" line source. Thus, the Idaho Falls data was crosswind integrated to obtain
downwind concentrations as a function of downwind distance, as was done in Snyder et al.,
(2013).

3

-------
The Prairie Grass database was used to determine the coefficients for the lateral dispersion. A
Gaussian fit was used to determine the lateral spread based in the measured concentrations from
each arc at each downwind distance. Then, the lateral dispersion constants c, ds, and du
coefficients were then determined to provide the best fit.

Again, the evaluation of RLINE was performed iteratively to optimize the fit to both the Idaho
Falls and Prairie Grass data while restricting the dispersion equations to their current form.
Multiple "best-fit" metrics were explored including R-squared, NMSE, and standard error.
Ultimately, the R-squared statistics were used to optimize the fit for the lateral and vertical
spread for both Prairie Grass and Idaho Falls, as shown in the Figures below. A Genetic
Algorithm was used to optimize the combination of the six coefficients within the specified
ranges given in Table 1 below. The intent was to obtain convergence on the "optimum value",
however from the Figures below the model performance is not extremely sensitive to all the
coefficients. In particular, the model evaluations show better model performance for smaller
ranges of a, bs, bu, ds, and du. However, there is little model sensitivity to the c-coefficient.
Although, there are smaller ranges for five of the six coefficients there is not a single value
which stands out as "the" value. Thus, selection of values within these smaller ranges were
made to optimize the RLINE model performance for these datasets.

Table 1: Comparison of RLINE oy and oz coefficient values and ranges tested.

Coefficient

Original Value
(Von knl in in el ill.. (2013))

Range Tested

New Value

0.57

0.7

3.0

p
L/i

1.5

0.5-2.0

1.0

1.6

1.4

2.5

-2.5-2.5

1.5

1.0

2.0-3.5

2.5

-------
bu

0.5 10 1.5 2.0

8496-!

du

/*

#

•

#





f •



•* ••
A



1.40 1.45 1.50 1.55 1.00

Coefficient Value

2.4 2.5 2.8 2.7

Figure 1: The 6-panel figure shows an example of the R-squared value for multiple combinations
of the a, bs, bu, c, ds, and du coefficients. Where the "best-fit" R-squared value is shown as a red
dashed line in all panels.

5

-------
Stable

a = 0.7, bs = 1.5

1e+02-

O

O
£

1e+01-

03

GO

Z>

*

a

1e+00-

1e-01

Idaho Falls
Prairie Grass

Convective

a = 0.7, bu = 1

1e-02-

1e+02

O

° 1e-01

2
«5
Z)

a

1e-04







•wl























-

\ \





\ \



• Idaho Falls

\ \
\



* Prairie Grass

N



1e-01 1e+00 1e+01 1e+02
x/|Lmo|

1e+03

1e-01

1e+01
x/|Lmo|

1e+03

Figure 2: Idaho Falls 2009 (circles) and Prairie Grass (triangles) normalized concentration vs.
x/|L|. The solid and dashed lines represent the new az equations for stable and convective
conditions for a range of u*/Ue values which are representative of the u*/Ue values for the Idaho
Falls and Prairie Grass field studies.

6

-------
3.0 Evaluation

After the RLINE calculations were modified to address the wind speed calculation,
harmonization with other AERMOD sources, and optimization of the vertical and lateral
dispersion coefficients, the model was evaluated against multiple datasets. These datasets
included previous datasets which were used in the initial evaluation of RLINE (Venkatram et al.,
2013) and Snyder et al., 2013), new datasets that have been explored since the 2013 publications,
and complex hotspot analysis projects as a collaboration between EPA and FHWA.

3.1 Previous Field Study Evaluations with Idaho Falls & Caltrans 99 Tracer Experiments

As part of the original development of RLINE (Heist et al., 2013), the model was evaluated
against two tracer data sets for line sources, including the 2008 Idaho Falls roadway study (Finn
et al., 2010), which consists of 4 days of sampling at a wide array of receptors and the Caltrans
99 highway study (Benson, 1989), which consists of 14 days of sampling at 10 receptors. The
performance of the reformulated RLINE source type within AERMOD has been assessed against
these tracer studies again, and the results are presented below. Brief summaries of the studies
provide details on each of these studies.

3.1.1 Idaho Falls Roadway Study

A tracer study of dispersion from a near ground-level line source was carried out in 2008
near Idaho Falls, ID, on an open field test site designed for transport and dispersion tracer
studies (Finn et al., 2010). In this study, two parallel sites were set up, one with a noise
barrier and one without. Tracer releases were performed simultaneously at both sites. In
each case, sulfur hexafluoride (SF6) was released uniformly along a 54 m long source,
positioned 1 m above ground level, beginning 15 minutes before the first sampling period
and continuing through a 3 h experiment consisting of 12 consecutive 15-minute
sampling periods. Background levels of SF6 at the study site were measured to be
between 6 and 8 pptv, whereas measurements in the center of the grid were the order of
thousands or tens of thousands of pptv. Only data from the non-barrier site were used in
the results of the model inter-comparison presented in this document. Experimental data
are available from four separate days, capturing a wide range of atmospheric stabilities
and wind speeds... (excerpted from Heist et al., 2013).

Samplers were arrayed downwind of the site in a grid that extended from 18 m to 180 m
downwind of the source and extending from 108 m in each direction along the source from its
center (Figure 3). Sampling occurred on four tests days which were characterized as 1) near-
neutral, 2) convective, 3) weakly stable, and 4) moderate to strongly stable.

Figure 4 shows model results (using the RLINE source type) plotted against measured
concentrations for the four test days of the Idaho Falls Roadway Study for AERMOD version
22112 and 23132. Despite the changes described above in 23132, there are only small changes in
model performance observed for the Idaho Falls study, especially for the neutral and convective

-------
test days. For the weakly stable day, the highest concentrations remain relatively unchanged
though agreement between measured and modeled concentrations is reduced somewhat for lower
concentrations; however, they remain within a factor two of each other. For the strongly stable
test day, the highest overpredictions by version 22112 have been reduced by version 23132, and
now lie below the factor of two line to a greater degree.

Figure 3: Idaho Falls Study layout. Source is indicated with vertical line along x/Hb = 0 (from
y/Hb = -4.5 to 4.5). Filled circles show the locations of bag samplers. North is indicated by the
direction of arrow.

Idaho Falls, 22112

Idaho Falls, 23132

0,1 1.0 10.0

Day 3, Weakly Stable

r/v7



//	



0.1	1,0	10.0

Day 4, Strongly Stable

1.0	10.0

0.1	1.0	10.0

Day 4, Strongly Stable

1.0	10.0	0.1	1.0	10.0

Measured Concentration

Figure 4: Modeled vs measured SF6 concentration (in ppb) using the RLINE source type for all
test days at Idaho Falls. AERMOD ver. 22112 (left four plots) and ver. 23132 (right four plots).

8

-------
Each symbol represents a 15 min. average. The three lines in each plot are the 1:2, 1:1, and 2:1
lines.

3.1.2 Caltrans 99 Highway Study

A tracer study was performed in the early 1980s using SF6 released from the tailpipes of eight
specially modified automobiles traveling with traffic on Highway 99 outside Sacramento
(Benson, 1989). The study was conducted along a straight segment of the highway aligned from
northwest to southeast consisting of four lanes and a 14 m wide median. The highway carried
approximately 35,000 vehicles daily. The surrounding terrain was generally flat, and nearby land
use consisted of open park land, fields, and scattered residential developments. Eight
automobiles releasing the tracer circulated up and down a 4 km segment of the highway
beginning one hour before sampling started. Half of the modified vehicles were driven in the
right-hand lane and the other half in the left-hand lane to distribute emissions evenly across the
lanes of the highway. SF6 monitors were arrayed on both sides of the road (spaced at 50, 100 and
200 m from the center of the road) and at four locations along the median (spaced approximately
800 m apart) (Figure 5). Samplers were positioned 1 m above ground level. Samples were
collected in Tedlar bags for four consecutive 30-minute periods and analyzed using gas
chromatography. Two cup and vane anemometers were installed on a 12 m meteorological tower
near the sampling array at heights of approximately 6.5 m and 11.4 m. (excerpted from Heist et
al., 2013)

•

200 m

•

100 m

•

50 m

• •

•

-50 m

•

-100 m

•

-200 m

Figure 5: Caltrans 99 Highway Study layout. Tracer was released along the roadway with
samplers arrayed on either side extending from 50 m to 200 m with four additional samplers in
the median of the roadway.

Figure 6 shows model results (using the RLINE source type) plotted against measured
concentrations for the Caltrans 99 Highway Study for AERMOD versions 22112 and 23132. As
with the Idaho Falls results above, the results from the Caltrans 99 study show only minor

-------
differences between these two versions of the model. The most noticeable change is the
reduction in concentration for the five outlier points in the upper left part of each plot.

Caltrans 99, v22112

Caltrans 99, v23132

100

1000

10000

100

1000

10000

Measured Concentration

3.2 New Field Study Evaluations with GM Sulfate Dispersion Experiment

The previous evaluations used datasets included in the initial RLINE evaluations (Snyder et al.,
2013 and Heist et al., 2013). Two additional datasets were included in the evaluation of the new
RLINE formulation, both of which included tracer studies on roadways with measurements are
multiple distances form the roadway.

3.2.1 GM Sulfate Dispersion Experiment

The General Motors (GM) Sulfate Dispersion Experiment, conducted from late September
through October of 1975, measured vehicle emissions of sulfate in a near-road environment.
During the study, a tracer gas (SF6) was released from trucks evenly dispersed throughout the
352-vehicle fleet, which were organized into 11 packs of vehicles. On-site meteorology and
tracer gas concentrations were measured throughout the study period. Data are available for 17
days between 29 September 1976 and 30 October 1976, with about 4 samples per day between
SAM and 10AM. The study was designed to be conducted under 'worst-case' meteorological
conditions, as such, experiment days were selected to have westerly winds with low wind speeds
(Cadle et al., 1976).

-------
Figure 7 Onsite wind speed (color) and direction by height (panels) for the average of
measurement towers 1, 5, and 6.

3.2.1.1 Location Description

The GM Sulfate Dispersion Experiment study site is in Milford, Michigan on the GM Proving
Ground (Figure 8). The GM Proving Ground is located 30 km north of Ann Arbor and 50 km
northwest of Detroit in southeastern Michigan. The nearest major highways are 1-96 (7 km to the
south) and M-23 (6 km to the west). The GM Proving Ground (Figure 8) consists of 140-miles of
roadway distributed among numerous tracks, road courses, and highways. Specifically, the 10
km North-South Straightaway was chosen for this study since the track would allow high density
traffic to achieve high speeds. The North-South straightaway is located on the western edge of
the Proving Grounds shown in Figure 8 by the green and purple line. The measurement
(receptor) locations are identified by purple triangles.

On the North-South straightaway, the track consists of three 5 km long lanes in either direction,
but the turns, at either end of the straightaway, only allow for two lanes in each direction.
Therefore, this study only uses two lanes in each direction (Cadle et al., 1976).

Sampling towers were positioned on either side of the roadway at varying distances in the middle
of the straightaway section of track, as shown in Figure 9. Tower locations were purposeful to
limit the influence of surrounding trees and small hills on pollutant dispersion (Cadle et al.,

1976).

-------
Figure 8 General Motors Proving Ground in
Mil ford, MI.

I	

mt

rj.t •

.1

±U-



urnsm v *«>—

	UJIMOI »*S

1 €¦ • 7.t •
c.s-- £itti« mi

Figure 9: Survey of Sampling Area (reprint
Cadle et al. 1976 Figure 4).

12

-------
3.2.1.2 Source Characterization

SF6 was released from the exhaust pipe of 8 trucks, evenly spaced between the 32 packs of cars,
split between the two lanes. Each pack contained 11 cars. SF6 emission rates were measured for
each truck for each day of the experiment (Cadle et al., 1976).

Two roadway (RLINE) sources are used to characterize each of the north and southbound
straightaways, representing two lanes in each direction, for a total of four sources. Each lane was
defined with a length of 4,000 meters, reflecting the nearly due north/south orientation of the
sources excluding the banked turns. Each source has a width of 3.3 meters. The release height
and initial vertical dispersion parameters are set at 1.5m, based on the truck's emission release
height.

3.2.1.3 Receptors

Sampling was conducted at eight tower locations shown as the black dots on the dashed line in
Figure 9, hereon referenced as towers 1 through 8, numbered from west to east. Towers 1
through 6 contained three sampling heights which are 0.5 m, 3.5 m, and 9.5 m above ground, and
towers 7 and 8 contained one sampling height at 0.5 m above ground. Sampling points were
aligned along a 45° angle across the roadway to provide the least possible interference from the
small surrounding hills. More sampling location were placed on the west side of the track since
this was anticipated to be downwind. Figures 8 and 9 illustrate tower locations. Note, in the
modeling files, the sampling height is represented as a flagpole receptor specified at the sampling
height which is the height above the terrain at which the sample is collected. Model receptors are
located at the eight sampling towers with the respective sampling heights.

3.2.1.4 Meteorology

Site-specific meteorology was collected during the study period; however, this data only
consisted of wind speed, wind direction, and ambient temperature. Wind speed and wind
direction were measured at 1.5, 4.5, and 10.5 meters above ground at towers 1 through 6. Towers
7 and 8 only measured wind speed and direction at 1.5m. Temperature was measured at towers 1
and 6 at 1.5 meters. Note, SF6 measurement heights vary by 1 meter compared with
meteorological measurement heights at the same tower location.

The Bishop International Airport (KFNT; Weather Bureau Army Navy (WBAN) Station 14826)
was identified as the nearest permanent meteorological site and provided secondary surface and
upper air meteorological measurements throughout the study period, though on-site data was
available during all measurement periods. KFNT airport meteorological station is located at
42.967°N and 83.750°W, approximately 26.4 miles north of the center of the study area and has
an elevation of 235 meters.

-------
AERMOD/AERMET models typically use one-hour averaging times. The GM experiment was
conducted with half hour measurements. To maintain the maximum number of datapoints, two
runs were performed for each half-hour of the experiment. One run represented the meteorology
specific to the first half of the hour to represent data collected in the first half hour and a second
run with meteorology specific to the second half of the hour. During analysis, data was
regrouped into the appropriate order to be paired with the monitor data.

3.2.1.5 AERSURFACE

AERSURFACE version 20060 was used to create reproducible albedo, Bowen ratio, and surface
roughness lengths, representative of surface characteristics at the study measurement sites, for
inclusion in AERMET. 2011 land cover, impervious surface percentage, and tree canopy cover
percentage data were obtained from the National Land Cover Database (NLCD) Multi-
Resolution Land Characteristics (MRLC) Consortium. The oldest available NLCD dataset was
created in 1992, the closest date to the GM experiment study period. Post-1992 NLCD dataset
land cover categories were updated from a 21- to a 16-category system which also now includes
percentage area of grid cell that is impervious and percentage area of grid cell that is covered
with tree canopy. More information about the surface classification categories can be found in
the AERSURFACE User's Guide (EPA, 2020). Model results using the NLCD 2011 dataset
were in better agreement with GM experiment observations than the NLCD 1992 dataset (not
shown). For all AERSURFACE runs tower 5 was used as the center point, as this tower is
located in the middle of the test section and the terrain is homogeneous.

3.2.1.6 AERMET

AERMET input files include the site-specific measured wind speed, wind direction, and ambient
temperature. Since AERMET accepts only one measurement location, towers 1, 5, and 6 are
averaged together at their respective heights and assigned to tower 5's location. The
meteorological measurements were not found to vary greatly between these three towers (not
shown). In addition, the Bishop International Airport (KFNT) surface meteorological
measurements were used as the secondary data option to the on-site measurements. ISH surface
data and FSL upper air data were obtained from KFNT for AERMET processing, augmenting
the on-site measured data. Automated Surface Observing Systems (ASOS) 1-minute data was
not available for the study period. The AERMET adjust u* (ADJ U*) option was used to adjust
the surface friction velocity (u*) under low wind speed, stable conditions. AERMOD-ready
surface (.SFC) and profile (.PFL) files were generated with the most-recent version (v23132) of
the AERMET meteorological data preprocessor as of July 2023.

-------
3.2.1.7 AERMOD

AERMOD version 23132 was used for this analysis. The GM Proving Ground is located outside
of Milford, MI, which had a population of 6,175 in April 2010

(https://www.census.gov/quickfacts/fact/table/milfordvillagemichigan/PST045222). The study
site is located 30 km north of Ann Arbor and 50 km northwest of Detroit in southeastern
Michigan with low population density. Therefore, the URBAN model options were not used in
the AERMOD runs.

Elevation decreases from approximately 685 meters, above sea level, at the southern endpoints
of the roadway sources to approximately 665 meters at the northern endpoints of the roadway
sources. Over the four-kilometer-long sources a change in elevation of 20 meters is a 0.5%
grade. The non-default FLAT model option was applied to each of the AERMOD source types,
resulting in the exclusion of source and receptor elevations in concentration calculations.

3.2.1.8 Results and Discussion

Model results are shown for each measurement height in Figure 10 colored by tower location and
colored by wind speed and wind direction in Figure 11. Results suggest that the RLINE source
does a reasonable job predicting concentrations. On average the model slightly underpredicts
concentrations, however there are a few significant over predictions, greater that a factor of two.
These overpredictions are at the 1.5 m and 4.5 m heights with wind speeds less than 1 m/s and
winds out of the North to East-North-East. Thus, these could be time periods when the wind is
blowing along the roadway test section, instead of across the test section. This could suggest that
the RLINE model overpredicts when the winds are extremely light, and possibly in near-parallel
wind conditions. Further investigation is needed.

1.5 m 4.5 m 10.5 m

Figure 10 AERMOD v23132 run with AERMET v23132 NLCD 2011 surface characteristics
compared with GM experiment observations ([j,g/m3). Panels labeled by measurement height and
colored by tower number.

-------
Figure 11 AERMOD v23132 ran with AERMET v23132 NLCD 2011 surface characteristics
compared with GM experiment observations ([j,g/m3). Colored by wind speed (m/s, left) and
wind direction (right).

3.3 Hotspot Analyses Model Intercomparisons

The EPA and FHWA have coordinated on model intercomparisons of the new RLINE source
formulation to the existing AREA and VOLUME source formulations, as applied to real-world
hot-spot evaluations. These evaluations each include a large section of a highway modification
project, with an extensive receptor network to determine project design concentrations (DC). The
advantage of this type of comparison is that it exercises the models in meteorological scenarios
beyond what's available in the relatively short field studies, it uses multiple and complicated
source configurations, and has source-receptor distances and orientations beyond what is
available from the tracer datasets for mobile source emissions. The obvious disadvantage is the
lack of measurement data to make comparisons against, but the AREA and VOLUME source
types in AERMOD have been used since 2005 by the regulated community and have a proven
performance record that can serve as a benchmark against the RLINE results.

This section presents the results from two real-world PM hot spot scenarios that are used to
evaluate air quality impacts from roadway projects. The examples are referenced as Project A,
which includes a PM2.5 analysis for both the daily (98th percentile) and annual standards , and
Project B , which includes a PM10 analysis for the daily standard (6th highest concentration).

3.3.1 Project A - PM2.5 Hot-spot Analysis

Project A analyzes the intersection of two inter-state freeways, using standard receptor
placement for a typical hot-spot analysis (e.g., a receptor grid around the perimeter of the project,
starting at 5-m from the edge of the roadway), using 5-years of meteorology. Figure 12 and
Figure 13 show the project layout, with receptors colored by the design concentration, with
Figure 12 showing the results from AERMOD version 22112 and Figure 13 showing the results
from AERMOD version 23132.

16

-------
501000	502000

X

Figure 12 Project layout for Project A hot spot analysis. RLINE emission source drawn as lines
with emission rate shaded (black - blue - green). Receptor locations indicated with X's and
RLINE v22112 model concentrations shaded (purple - yellow). Design concentration receptor
location indicated with red circle.

RLINE V23132
Concentration (ng/mA3)

Emission rate (ng/m/s)

Figure 13 Project layout for the Project A hot spot analysis. RLINE emission source drawn as
lines with emission rate shaded (black - blue - green). Receptor locations indicated with X's and
RLINE v23132 model concentrations shaded (purple - yellow). Design concentration receptor
location indicated with red circle.

17

-------
Consistent with the approach for computing design concentrations for a PM-hot spot analysis, we
compute the annual average (average of each model year, averaged across the 5 model years) and
daily (98th percentile of the daily average concentrations, averaged across the 5-model years) at
each receptor. The resulting design concentration is the maximum annual and 98th percentile
concentration from all receptors for the project. Meteorological data was derived from a nearby
airport, which was decommissioned in 1995. The meteorological data for the model runs here
were the same used for the hotspot analysis conducted by the state, which was from 1990-1994,
which included using the adjusted u* option in AERMET. Project design concentrations (i.e., the
highest design concentration from all receptors) are summarized in Table 2, while Figure 14 -
Figure 17 compare the annual and daily PM2.5 design concentrations from all receptors,
comparing RLINE to the AREA and VOLUME source characterization for AERMOD versions
22112 and 23132.

Table 2 Project A design concentrations (PM2.5, (j,g/m3).

Soiiivc T\|R'

\ 22112 11X11

\ 23132 11X11

\ 22112 Aiiiuiiil A\ *i

\ 23132 Aiiiuiiil A\ *i

Kl INI

y."S

~4U

VJ"

3 Jo

VOLUME

7.69

3.30

AREA

7.97

2.93

18

-------
2.5 5.0
AREA

o.o

2.5 5.0
VOLUME

10.0

AREA	VOLUME

Figure 14 Comparison of RLINE 8th highest 24-hour concentrations of PM2.5 ([j,g/m3) for all
receptors between source type (AREA, left column and VOLUME, right column) and AERMOD
version (top, v22112 and bottom, v23132). Model runs for Project A.

RLINE V22112

Figure 15 Comparison of RLINE 8th highest 24-hour concentrations of PM2.5 ([j,g/m3) for all
receptors between AERMOD versions. Model runs for Project A.

19

-------
Figure 16 Comparison of RLINE 5-year annual average concentrations of PM2.5 ([j,g/m3) for all
receptors between source type (AREA, left column and VOLUME, right column) and AERMOD
version (top, v22112 and bottom, v23132). Model runs for Project A.









/









/



Ar





'

/A







/jj











y=c

•81V=°o0^



y







RLINE v22112

Figure 17 Comparison of RLINE 5-year annual average concentrations of PM2.5 ([j,g/m3) for all
receptors between AERMOD versions. Model runs for Project A.

20

-------
To further investigate the design concentrations, the receptors associated with the design
concentrations, for each source type, is selected. Then the dates are selected as those associated
with the highest concentrations at these receptors. DV receptors and dates listed in Table
3.Hourly concentrations and select meteorological variables are then plotted for the DV dates
(additional figures can be found in Appendix A).

For Project A, the design concentration (H8H) is the largest average concentration of the average
of the 8th highest concentration for each year at the respective receptor. Thus, the selected design
concentrations dates for Project A, are the dates associated with the design concentration
receptor's highest average concentration (one date per year). Two receptors corresponding to the
design concentrations are found (Table 3). RLINE v23132, AREA, and VOLUME source design
concentrations occur at the same receptor, where RLINE v22112 occurs at a different receptor.
The two DV receptors have the same highest average concentration dates, for each year. The
comparison for the 1-hour results for these select receptors and hours of modeling for the RLINE
source against the AREA and VOLUME sources for AERMOD versions 22112 and 23132 are
shown in Figure 18 and Figure 19.

Table 3 Project A design concentration receptors and dates used for hourly analysis.

Siuircc

Vci'skHl

kcccpinr

1 );ilcs

RLINE

V23132

500952.8, 4402905.7

1990-01-15 1993-01-04

1991-02-08 1994-12-13

1992-12-17

RLINE

V22112

500952.7, 4402880.7

1990-01-15 1993-01-04

1991-02-08 1994-12-13

1992-12-17

AREA

500952.8, 4402905.7

1990-01-15 1993-01-04

1991-02-08 1994-12-13

1992-12-17

VOLUME

500952.8, 4402905.7

1990-01-15 1993-01-04

1991-02-08 1994-12-13

1992-12-17

-------
>
LU

60	90	120	0	30	60

AREA	VOLUME

0	30	60	90	120

AREA

0	30	60	90	120

VOLUME

Wind speed (m/s

I

0

Date

¦ 1990-01-15

•	1991-02-08

*	1992-12-17

~	1993-01-04
ES 1994-12-13

>
Hi

Figure 18 Comparison of RLINE hourly concentrations of PM2.5 (ug/nr!) for design
concentration dates (shape) and wind speed (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs
for Project A.

Inverse Monin-
Obukhov length
(m1)

10.4
0.2
0.0
-0.2
W -0.4

Date

¦ 1990-01-15

•	1991-02-08
a 1992-12-17

~	1993-01-04
a 1994-12-13

RLINE v22112

RLINE V22112

Figure 19 Comparison of RLINE hourly concentrations of PM2.5 (ug/nr;) for design
concentration dates (shape), wind speed (color, left), and inverse Monin-Obukhov length (color,
right) between AERMOD versions. Model runs for Project A.

3.3.2 Project B: PMio Ftot-spot Analysis

Project B is a 10.5-mile multi-modal corridor. The analysis here models a section an interstate
freeway where a new connector to a new freeway will join the existing interstate when complete.
As with Project A, we use a standard receptor placement for a typical hot-spot analysis (e.g., a

22

-------
receptor grid around the perimeter of the project, starting at 5-m from the edge of the roadway),
using 5-years of meteorology.

Figure 20 and Figure 21 show the project layout, with receptors colored by the design
concentration, with Figure 20 showing the results from AERMOD version 22112 and Figure 21
showing the results from AERMOD version 23132.

RLINE v22112
Concentration (f.ig/mA3)

Emission rate (j.ig/m/s)

Figure 20 Project layout for the North Project B Corridor hot spot analysis. RLINE emission
source drawn as lines with emission rate shaded (black - blue - green). Receptor locations
indicated with X's and RLINE v23132 model concentrations shaded (purple - yellow). Design
concentration receptor location indicated with red circle.

23

-------
RLINE V23132
Concentration (ng/mA3)

Emission rate (ng/m/s)

Figure 21 Project layout for the North Project B Corridor hot spot analysis. RLINE emission
source drawn as lines with emission rate shaded (black - blue - green). Receptor locations
indicated with X's and RLINE v22112 model concentrations shaded (purple - yellow). Design
concentration receptor location indicated with red circle.

The form of the PM10 standard counts the number of daily averaged PM10 concentrations above
the level of the standard (150 mg/m3), with more than 3 days over three years of monitoring data
above the level of standard to be considered a violation. If the 4th highest daily averaged
concentration is less than the standard, then an area can be determined to pass. For modeling
demonstrations that typically cover a 5-year period (rather than the 3-years used with ambient
data to determine compliance), the test still allows for one concentration above the standard per
year, meaning that five concentrations can be above the level of the standard and the 6th high
value is used for the design concentration test, which is what is reported here, consistent with the
model demonstration for the PM10 analysis for this project. Meteorological data was derived
from Felts Field (SFF) general aviation relief airport in Project B, WA. The meteorological data
for the model runs here were the same used for the hotspot analysis conducted by the state,
which was from 2013-2019, which included using the adjusted u* option in AERMET. Project
design concentrations (i.e., the highest design concentration from all receptors) are summarized
in Table 4, while Figure 22 and Figure 23 compare the PM10 design concentrations from all
receptors, comparing RLINE to the AREA and VOLUME source characterization for AERMOD
version 22112 and 23132.

-------
Table 4 Project B design concentrations (PM10, (j,g/m3).

Source

V22112 H6H

V23132 H6H

RUNE

56.54

43.12

VOLUME

38.98

AREA

47.79

60-
50-

CN

40-

CN
CM
>

LU 30H

K

10-



y

A

/



a

f

p /

j

r

/









» /
* /











ill

20 30 40 50
AREA

60-
50-

CM

£ 40-



LU 30"

L

10-



' * /
j/







M

/







0/ /

I--'







y = 1 3^=+o1i

601

50-

40-

oo

T-

co
CM
>

LU 30
Z
_l
0£

20-

10

10

10 20 30 40 50 60
VOLUME









/









A







/





















/J







i 1



20 30 40 50
VOLUME

60

Figure 22 Comparison of RLINE 6th highest 24-hour concentrations of PM10 ([j,g/m3) for all
receptors between source type (AREA, left column and VOLUME, right column) and AERMOD
version (top, v22112 and bottom, v23132). Model runs for Project B.

25

-------
RLINE v22112

Figure 23 Comparison of RLINE 6th highest 24-hour concentrations of PM10 ([j,g/m3) for all
receptors between AERMOD versions. Model runs for Project B.

To further investigate the design concentrations (DC), the receptors associated with the
maximum design concentrations, for each source type, is selected. Then the dates are selected as
those associated with the design concentration concentrations at these receptors. DC receptors
and dates are listed in Table 5. Hourly concentrations and select meteorological variables are
then plotted for the DC dates (additional figures can be found in Appendix A).

For Project B, there is one DC receptor for all source types (Table 5). RLINE v23132, RLINE
v22112, and VOLUME sources highest concentration occurs on the same date, where the AREA
source occurs on a different date. Thus, there are two DC dates of interest. Table 5 summarizes
the days selected for this analysis and Figure 24 and Figure 25 show the hourly concentrations
from RLINE against the AREA and VOLUME source concentrations for the hours of interest.

Table 5 Project B design concentration receptors and dates used for hourly analysis.

Source

Version

Receptor

Date

RLINE

V23132

761001.6, 78547.1

2014-01-08

RLINE

V22112

761001.6, 78547.1

2014-01-08

AREA

V23132

761001.6, 78547.1

2014-12-07

VOLUME

V23132

761001.6, 78547.1

2014-01-08

26

-------
AREA	VOLUME

AREA	VOLUME

Figure 24 Comparison of RLINE hourly concentrations of PM10 (|jg/m3) for design
concentration dates (shape) and wind speed (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs
for Project B.

50	100	150

RLINE V22112

Wind speed (m/s)

Inverse Monin-
Obukhov length
(rtv1)

I

0.0
-0.1
-0.2

Date

¦ 2014-01-08
• 2014-12-07

Date

¦ 2014-01-08
• 2014-12-07

Figure 25 Comparison of RLINE hourly concentrations of PM10 ([jg/m3) for design
concentration dates (shape), wind speed (color, left), and inverse Monin-Obukbov length (color,
right) between AERMOD versions. Model runs for Project B.

SO	100	150	200

RLINE V22112

27

-------
4.0 Conclusions

This document describes modification to the formulation of the RLINE source in AERMOD,
which include modifications to the dispersion and meteorology used by RLINE to use similar
formulations as other AERMOD source types, bug fixes identified during the BETA release of
RLINE, and the refitting of dispersion curves to field data resulting from changes to these first
two model changes. The new model performance was evaluated against two field studies used
previously, Idaho Falls and Caltrans 99, as well a previously unevaluated database, the GM
Sulfate. For the Idaho Falls and Caltrans 99 field studies, modeled concentrations generally
decreased, though the model agreed quite well for the newly analyzed GM Sulfate experiment.
The updated RLINE formulation was also benchmarked against the existing AREA and
VOLUME source characterization for two real-world hot-spot projects. Prior to the
reformulation, RLINE design concentrations were 30-40% higher than those coming from the
other source types. Following the reformulation, RLINE design concentrations fell between the
two other source types, which would be expected given the formulation differences between the
three source types. The model formulation presented here in AERMOD version 23132 balances
performance with the evaluation datasets and the model intercomparisons with the hot-spot
cases.

28

-------
5.0 References

Barad, 1958. Project Prairie Grass. A Field Program in Diffusion. AFCRF-TR-58-235.

Benson, P.E., 1979. CALINE3—A Versatile Dispersion Model for Predicting Air Pollutant
Levels Near Highways and Arterial Streets. Interim Report, Report Number
FHWA/CA/TL-79/23. Federal Highway Administration, Washington, DC (NTIS No. PB
80-220841).

Cadle, S.H., D.P. Chock, J.M. Heuss, and P.R. Monson. 1976. "Results of the General Motors
Sulfate Dispersion Experiment."

Cimorelli, A.J., Perry, S.G., Venkatram, A., Weil, J.C., Paine, R.J., Wilson, R.B., Lee, R.F.,
Peters, W.D., Brode, R.W., 2005. AERMOD: a dispersion model for industrial source
applications. Part I: general model formulation and boundary layer characterization.
Applied Meteorology 44, 682-693.

Environ. 77, 748-756.Venkatram, A., Brode, R., Cimorelli, A., Lee, R., Paine, R., Perry, S.,

Peters, W., Weil, J., Wilson, R., 2001. A complex terrain dispersion model for regulatory
applications, Atmospheric Environment.

EPA, 2020. User's Guide for the AERSURFACE Tool. Office of Air Quality Planning and

Standards Air Quality Assessment Division. Research Triangle Park, NC. EPA-454/B-
20-008.

Finn, D., Clawson, K.L., Carter, R.G., Rich, J.D., Eckman, R.M., Perry, S.G., Isakov, V., Heist,
D.K., 2010. Tracer studies to characterize the effects of roadside noise barriers on near-
road pollutant dispersion under varying atmospheric stability conditions. Atmos. Environ.
44, 204-214.

Heist, D.K., Perry, S.G., Brixey, L.A., 2009. A wind tunnel study of the effect of roadway

configurations on the dispersion of traffic-related pollution. Atmos. Environ. 43, 5101—
5111.

Heist, D.K., Isakov, V., Perry, S., Snyder, M., Venkatram, A., Hood, C., Stocker, J., Carruthers,
D., Arunachalam, S., Owen, R. C. 2013 Estimating near-road pollutant dispersion: A
model inter-comparison. Transportation Research Part D, 25, 93-105.

Merceret, F.J., 1995. The effect of sensor sheltering and averaging techniques on wind
measurements at the Shuttle Landing facility (No. NAS 1.15: I I 1262).

Snyder, M.G., Venkatram, A., Heist, D.K., Perry, S.G., Petersen, W.B., Isakov, V., 2013.

RLINE: a line source dispersion model for near-surface releases. Atmos.

Venkatram, A., Snyder, M.G., Heist, D.K., Perry, S.G., Petersen, W.B., Isakov, V., 2013a. Re-
formulation of plume spread for near-surface dispersion. Atmos. Environ. 77, 846-855.

29

-------
6.0 Appendix A

Additional plots from PM hot spot analyses model intercomparison.

Figure 26: Comparison of RLINE hourly concentrations of PM2.5 (pg/m3) for design
concentration dates (shape) and friction velocity (color) between source type (AREA, left
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).
Model runs for Project A.

30

-------
Figure 27: Comparison of RLINE hourly concentrations of PM2.5 (pg/m3) for design
concentration dates (shape) and friction velocity (color) between AERMOD versions. Model
runs for Project A project.

AREA	VOLUME

Figure 28: Comparison of RLINE hourly concentrations of PM2.5 ([jg/m3) for design
concentration dates (shape) and surface heat flux (color) between source type (AREA, left
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).
Model runs for Project A.

31

-------
Heat Flux (W/m2)

1990-01-15

1991-02-08

1992-12-17

1993-01-04

1994-12-13

30	60	90

RLINE V22112

Figure 29: Comparison of RLINE hourly concentrations of PM2.5 (jig/m3) for design
concentration dates (shape) and surface heat flux (color) between AERMOD versions. Model
runs for Project A.

Wind Direction (°N)

¦ 300
200
100
0

Date

¦ 1990-01-15

•	1991-02-08

*	1992-12-17

~	1993-01-04
a 1994-12-13

AREA

30	60	90

VOLUME

Figure 30: Comparison of RLINE hourly concentrations of PM2.5 ([Jg/m3) for design
concentration dates (shape) and wind direction (color) between source type (AREA, left column
and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model
runs for Project A.

32

-------
LU
Z
_l
OL

°lJ

/





¦ /



~

/ Jk





y = 0 .66 * • 0 .66

'7T

Wind Direction (°N)

1300
200
100
0

Date

¦ 1990-01-15

•	1991-02-08

*	1992-12-17

~	1993-01-04
Q 1994-12-13

30	60	90

RLINE v22112

Figure 31: Comparison of RLINE hourly concentrations of PM2.5 (pg/m3) for design
concentration dates (shape) and wind direction (color) between AERMOD versions. Model runs
for Project A project.

/ i
, ~



/ ~/ /
m/ + /

*/ /







/W\a/





y i v=*o.y

30	60

VOLUME

30	60

AREA

Inverse Monin-
Obukhov length
(m1)

10.4
0.2
0.0
-0.2
-0.4

Date

¦ 1990-01-15

•	1991-02-08
A 1992-12-17

~	1993-01-04
1994-12-13

30	60	90

VOLUME

Figure 32: Comparison of RLINE hourly concentrations of PM2.5 ([Jg/m3) for design
concentration dates (shape) and inverse Monin-Obukhov length (color) between source type
(AREA, left column and VOLUME, right column) and AERMOD version (top, v22112 and
bottom, v23132). Model runs for Project A.

33

-------
/e

Hour

~ *





20





m

15





w

10





m

5

w * /













£/*/

Date

,'jT 3yS

¦

1990-01-15



•

1991-02-08



~

1992-12-17



~

1993-01-04

Jp-' y=1V«6,U

0

1994-12-13

0 30 60 90 120





VOLUME





AREA

30	60

VOLUME

Figure 33: Comparison of RLINE hourly concentrations of PM2.5 (jig/m3) for design
concentration dates (shape) and hour of day (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs
for Project A.

Hour

I

20
15
10
5

Date

¦ 1990-01-15

1991-02-08

1992-12-17

1993-01-04

1994-12-13

~

30	60

RLINE V22112

Figure 34: Comparison of RLINE hourly concentrations of PM2.5 (|ig/m3) for design
concentration dates (shape) and hour of day (color) between AERMOD versions. Model runs for
Project A.

34

-------
6.2 Project B PMio

50	100	150

AREA

200

50	100	150 200

VOLUME

50	100	150

VOLUME

Figure 35 Comparison of RUNE hourly concentrations of PM10 (jig/m3) for design
concentration dates (shape) and friction velocity (color) between source type (AREA, left
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23 132).
Model runs for Project B.

35

-------


/ /
/ /
/ X

/ /
/ /



/ /

/ /
/ /

/ /
/ /
' / ^



/

t

/

t

/

/

I

/

/

/

/

/ ¦ •

/ • /
/ ¦ •/

/ '

/

/

/

t

!

/

t

t





t /
/ / y
' - /



y = 0.69;) x +3.7
r = 0.94





Date

¦ 2014-01-08
• 2014-12-07

Friction velocity (m/s)
0.12

0.10

0.08

0.06

0.04

50	100	150

RLINE V22112

200

Figure 36. Comparison of RLINE hourly concentrations of PM10 ((j,g/m3) for design
concentration dates (shape) and friction velocity (color) between AERMOD versions. Model
runs for Project B.

36

-------
AREA	VOLUME

AREA	VOLUME

Figure 37. Comparison of RLINE hourly concentrations of PM10 (ug/m3) for design
concentration dates (shape) and surface heat flux (color) between source type (AREA, left
column and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132).
Model runs for Project B.

37

-------
Heat Flux (W/m2)

140
30
20
10
0

-10
Date

¦ 2014-01-08
• 2014-12-07

50	100	150

RLINE V22112

Figure 38. Comparison of RLINE hourly concentrations of PM10 (pg/m3) for design
concentration dates (shape) and surface heat flux (color) between AERMOD versions. Model
runs for Project B.

38

-------




CM

t- 150

CM
CM
>

w 100-

~

a:

50

200-

CM

CO 150-

T—

CO
CM
>

LU

100J

150

CM
CM
>

W 100

~

(X

50

0	50	100	150	200

VOLUME

CM

CO 150

T—

CO
CM
>

LU

100

0	50	100	150	200

VOLUME

100	150

AREA

100	150	200

AREA

Wind direction (°N)

Date

¦ 2014-01-08
• 2014-12-07

Figure 39. Comparison of RLEsfE hourly concentrations of PM10 (ag/nr1) for design
concentration dates (shape) and wind direction (color) between source type (AREA, left column
and VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model
runs for Project B.

39

-------
200-



/

/

/

/







/

/

t

t









cv, 150-
co

T—

CO
>

2 100-

~
a:





/

/

/

/

i

/









/

/

/

/

/

/

/

/

t

/

/ /

/ • '

•



/

/

/

/

1

/





•









/

/

/

/ 1







50-

t

t y
/ /

/ / >
/ /
' /
/ /9r ^ "











o-

/I









y = 0.7.>x + 3.2
r = 0.94















0

50	100	150

RLINE V22112

200

Wind direction (°N)
300

Date

¦ 2014-01-08
• 2014-12-07

Figure 40. Comparison of RLINE hourly concentrations of PM10 ((ig/m3) for design
concentration dates (shape) and wind direction (color) between AERMOD versions. Model runs
for Project B.

40

-------
AREA	VOLUME

AREA	VOLUME

Figure 41. Comparison of RLINE hourly concentrations of PM10 ([jg/m3) for design
concentration dates (shape) and inverse Monin-Obukhov length (color) between source type
(AREA, left column and VOLUME, right column) and AERMOD version (top, v22112 and
bottom, v23132). Model runs for Project B.

41

-------
AREA	VOLUME

AREA	VOLUME

Figure 42. Comparison of RLINE hourly concentrations of PM10 (ug/'m 5) for design
concentration dates (shape) and hour of day (color) between source type (AREA, left column and
VOLUME, right column) and AERMOD version (top, v22112 and bottom, v23132). Model runs
for Project B.

42

-------
200-

CM
CO

CO
CM
>

LLI

a:

150-

100-

Hour

120
15
10
5

Date

¦ 2014-01-08
• 2014-12-07

50	100	150

RLINE V22112

Figure 43. Comparison of RLINE hourly concentrations of PM10 (ug/'m5) for design
concentration dates (shape) and hour of day (color) between AERMOD versions. Model runs for
Project B.

43

-------
United States	Office of Air Quality Planning and Standards	Publication No. EPA-454/R-23-011

Environmental Protection	Air Quality Assessment Division	October 2023

Agency	Research Triangle Park, NC

-------