Clark County (Nevada)
Paved Road  Dust Emission Studies in Support of Mobile
                   Monitoring Technologies
                  Rodney Langston, and Russell S. Merle Jr
  Clark County Department of Air Quality and Environmental Management (DAQEM)
                500 S Grand Central Parkway P.O. Box 555210
                      Las Vegas, Nevada 89155-5210

          Vic Etyemezian, Hampden Kuhns, John Gillies, and Dongzi Zhu
                       Desert Research Institute (DRI)
                           755 E Flamingo Road
                      Las Vegas, Nevada 89119-7363

                       Dennis Fitz and Kurt Bumiller
               Center for Environmental Research and Technology
                      University of California, Riverside
                          1084 Columbia Avenue
                      Riverside, California 92507-0434

                      David E. James and Hualiang Teng
                Department of Civil and Environmental Engineering
                       University of Nevada, Las Vegas
                             P.O. Box 45-4014
                           4505 Maryland Parkway
                        Las Vegas, Nevada 89154-4015
                            December 22, 2008

-------
                              EXECUTIVE SUMMARY

                            Clark County (Nevada)
  Paved Road Dust Emission Studies in Support of Mobile Monitoring
                                  Technologies
1.   Background

     a.    Need for Alternative to AP-42 Methodology

    The  Las Vegas  Valley in Clark County,  Nevada, has been  classified as  a serious
nonattainment area for the federal fine particulate matter (PMio) National Ambient Air Quality
Standards (NAAQS).   The June 2001 PMio  State Implementation Plan (SIP) specifically
addressed improvement of paved road dust emission characterization because of the importance
of paved road dust as a major category in the PMio emission inventory.

    The  SIP  contained  a research  commitment  to  explore  the  feasibility of a  more
comprehensive sampling system using vehicle-based mobile monitoring  for development of
improved paved road  emissions inventories.   The intention was to overcome the limitations of
the AP-42 methodology, which made it impractical to represent all of the classes and subclasses
of roadways.  The required road surface sampling is time-consuming and potentially hazardous
because of the need to block traffic lanes.  In  addition there are serious issues related to the
number of samples needed to represent spatial and temporal variations across roadway networks.
It became clear that the challenges related to the successful maintenance of conformity made it
imperative that an alternative approach to measuring and estimating paved road dust emissions
be developed.

    Beginning in 1999, Clark County undertook a series of field studies to investigate alternative
ways of estimating PMio emissions in the form of surface dust entrained from paved roads. A
new vehicle-mounted  mobile sampling technology was tested in comparison with the traditional
AP-42  method and its associated road surface sampling.  In addition, the plume flux profiling
method, which was the basis for development of the AP-42 emission factor equation  for public
paved roads, was used to calibrate the mobile monitoring technology.

    Two versions of the mobile monitoring technology were tested—TRAKER and SCAMPER.
Both technologies involve  on-board sampling of the dust plume generated by a test vehicle.
Both use continuous PMio particle monitors in conjunction with GPS systems, so that dust plume
concentrations  can be  mapped on  to  the road  system  traveled  by the test  vehicle.   The
SCAMPER samples the plume in the wake of the  test vehicle.  The TRAKER I and II test
vehicles sample the plumes from the front wheel wells of the respective vehicles. TRAKER II
has  a dilution system  to provide for use on unpaved roads.  All three units have samplers that
monitor the  PMio concentration in  front of the vehicle so that "background" PMio can be
subtracted.

-------
    Early in this program, it was decided that the test vehicles would travel at the normal traffic
speeds (25 to 45 mph) on the selected paved roadway network.  In addition, the weights of the
TRAKER  and  SCAMPER test  vehicles  were  closely  matched  and  provided  a good
representation of the fleet average vehicle weight in the Las Vegas Valley.

    In principle, two essential mathematical calculations are involved in the mobile monitoring
technology:  (1) conversion of the particle monitor reading (minus background) to the net PMio
concentration, and (2) conversion of the net PMio concentration in the test vehicle plume to the
equivalent PMio emission factor for the test vehicle.  Note that in this study, these two steps were
combined so that the average particle monitor reading was converted directly to the equivalent
PMio emission factor.

     b.   Previous Related Studies (Phases I through III)

    The field study reported in this document is Phase IV of a series of studies that began in
2004.  The Phase I study entailed a two-day field effort utilizing a 107-mile sampling route.  The
purpose of the study was to determine the feasibility of vehicle-based mobile  sampling system
for use in Clark County  to  better characterize paved-road emissions and to develop real-time
emissions of PMio for emissions inventory use.  The sampling  route was designed to include
worst-case silt-impacted roads and best-case clean roads in order to evaluate the detection limits
of the two systems. A total  of sixteen AP-42 silt samples were  also collected on the sampling
route.  Phase I demonstrated the feasibility of using vehicle-based mobile sampling systems as an
alternative to conventional AP-42 paved-road emissions estimating methods.

    The Phase  II study, which was completed in early 2005, was an expansion of the Phase I
study.  It entailed four days of sampling with the SCAMPER and TRAKER systems on a 103-
mile travel route.   The route included the five classes  of roadways (local,  collector, minor
arterial, major arterial, and freeway) and four political jurisdictions in the Las Vegas Valley.  The
route passed through developing areas, older  established neighborhoods, and newer planned
communities that were completely built-out. The developing areas included a cross section of
incomplete road infrastructure (e.g. unpaved road  shoulders) and deposition sources such as
vacant lots and construction activities.  The built-out areas had completed road infrastructure,
with few vacant lots, and little construction activity.  The route also included a cross section of
soil  classifications based  on Clark  County's Paniculate  Emission  Potential   (PEP)  soil
classification system.   The  sampling route included ten historical AP-42 sampling sites  and
eleven new sites that had not previously been sampled using AP-42 methodology.

    The Phase  III study utilized only the SCAMPER system and  the AP-42  methodology for
sensitivity and variability analysis.  The study occurred over seven consecutive days in late 2005
and utilized three  sampling routes.   Road infrastructure,  adjacent land use (e.g. vacant land,
residential, etc) and sources of deposition were  comprehensively  mapped prior to the  study.  The
first  sampling route (industrial route) was dominated by industrial haul roads with heavy silt
loadings and was used to  determine the precision of the SCAMPER unit.  This route included
local, collector and arterial roads. The second route (transitional route) was a 7.3-mile course in a
transitional development area in the  Las  Vegas Valley  that included a mix of  commercial,
residential, rural residential and vacant land.   The third route (developed community route)

-------
consisted of a 12.6-mile course traversing a newly developed planned community and contained
local, collector and arterial roads. In addition to providing baseline measurements for  fully
developed roadways with minimal silt deposition sources, this route was used to evaluate the
sensitivity of the SCAMPER unit.

2.   Phase IV Study

    The Phase IV study was  also directed to evaluating mobile monitoring technologies in
comparison  with the  traditional  AP-42  methodology,  but in  a controlled  measurement
environment that included restricted vehicle movement, controlled vehicle speeds and controlled
road surface material loadings.  This was accomplished by dedicating half of a divided roadway
as the test course for the 5-day field study. The specific study objectives were as follows:

    •  Comparison  of SCAMPER and TRAKER  system  measurements with  emission
       measurements using a downwind flux tower.
    •  Determination  of the relationship between  roadway silt loading and SCAMPER and
       TRAKER measurements at several standard vehicle speeds (25, 35 and 45 mph).
    •  Comparison of SCAMPER and TRAKER measurements to AP-42 emission estimates.
    •  Characterization of road surface silt depletion rate as  a function of the number of vehicle
       passes.
    •  Characterization of quantified emissions vs. quantified silt loading mass.
    •  Data assessment and review for recommendations  on performance specifications for
       vehicle-mounted mobile sampling systems.

    The test road consisted of two lanes of a four-lane divided highway, with a curbed median
and roadsides (Veterans Memorial Boulevard)  in Boulder City, Nevada.  The test road  segment
was oriented southeast  to northwest as shown in Figure 1.   All of the normal road traffic was
diverted to the southeast-bound lanes, allowing the two northwest-bound lanes and the stabilized
median  area to  be  utilized  exclusively for the five-day study.  This  diversion provided for
dedication of the test lanes to this project for the entire study period, so that only SCAMPER and
TRAKER vehicles traveled the test lanes, except for the spreader and sweeper that applied and
removed dirt from the travel course.

    The test course consisted of an acceleration zone to reach the desired test vehicle  speed, a
silt zone (approximately 1/2 mile), followed by a deceleration zone. Except for the first two
measurement sets, where test vehicles traversed the test course in both directions, a layer of silt
was applied to the curbside lane (measuring on average 13'5" in width), and the test  vehicles
traveled over that outer lane in a northwest direction.  In the course of a test series, the silt
loading was measured at designated locations at both ends of the test road segment. At the end
of a test set, the road was swept before the next silt layer was applied.  Typically one sampling
tower was  located on  the curbside adjacent  to the test lane.   It  was anticipated  that these
controlled traffic and  measurement parameters would enhance  the  quality of the tower flux
measurements compared to previous paved road dust studies.

-------
                             Figure 1-1. Layout of Test Course
                             AP42 Sampling Zone North
                             120 Feet
                                                 Profiling Zone
                                                 2206 Feet
Legend

^^| AP42 Sampling Zone

|    | Silt Buffer Zone Day2-5

|    | Profiling Zone Day2-5

|	| Acceleration Zone

|	| Deceleration Zone Day2-5

  —  Approximate Tower Locations Day2-5

  <   42" Cone

  1   Light Poles

jtjjvjijBy Drainage Channel

   - Curb

     Roadway

     Lane Lines

     Sidewalks
  200 100 0
                                       AP42 Sampling Zone South
                                       118 Feet
                                                                                 Acceleration Zone
                                                                                 662 Feet
                                                    Silt Buffer Zone,
                                                    87 Feet
  Note: The silt zone, with total length 2770 feet, includes Silt Buffer Zones, AP42 Sampling Zones, and Profiling Zone.

3.   Test Methods (Equipment, Operation, and Data Reduction)

    Particle  concentration  measurements  formed  the  basis  for  the  mobile  monitoring
technologies as well as the roadside emission flux measurements.  A continuously recording
particle monitor (DustTrak Model 8520, TSI Inc., Shoreview MN) was the basic instrument used
to log 1-sec PMio readings in all  cases.  Because the DustTrak operates  on a light-scattering
principle, a collocated mass-based reference monitor was used to correct the DustTrak readings
to equivalent PMio mass-based concentrations, as described below.

     a.    Flux Tower Method

    A "master" tower was erected downwind of the road  (approximately 5  m from centerline of
test vehicle travel path) and aligned perpendicular to road. The trailer-mounted, 9  m-high tower
was instrumented with DustTraks at five heights above the ground (0.7, 2.1, 3.4, 6.4,  and 9.8 m).
At one of the heights (3.4 m), a  DustTrak with a PM2.5 impactor inlet was collocated with the
PMio monitor.   The master tower  also  included an EPA-approved  reference PMio  monitor
(TEOM Model  1400a, R&P) at a height of 2.3 m, to be used for conversion of PMio measured
with the tower-mounted monitors  into mass-based PMio  concentrations. However,  note that a

-------
different but comparable conversion factor from a laboratory chamber study was actually used in
this study.   A wind vane was  mounted  at the top of the tower, and cup anemometers  were
positioned on the tower to monitor wind  speed as a function of height.  All data from the PM
samplers and meteorological instruments were telemetered and logged in 1-second intervals by a
laptop located on the master tower.

     b.   AP-42 Method

    As shown in Figure 1, two zones of the course, called "south" and "north," were designated
for silt recovery as input to the AP-42 emission factor equation for public paved roads. The
"near" end of the south AP-42 sampling zone was established 201.7 meters (662 feet) from the
start of the  course. This distance  was selected so that the mobile  monitoring vehicles could
complete the acceleration portion  of their pass before entering the south soil sampling  zone.
Both AP-42 sampling zones were approximately 36.6 meters (120 feet) long.  The "far" end of
the north AP-42 sampling zone was established about 500 feet from the end of the course, to
allow for deceleration just before the gradual curve in the roadway.

    Two different plot layouts were used  during the empirical study to collect soil samples from
the test course. An array of seven full-size plots, with 2.4 meter (8-foot) spacing between the
plots was laid out at each zone of the driving course. A full  size (3.3 meter  long x 4.1 meter
wide) plot was used to measure silt loading at the beginning and end of most  of the test series.
The 3.3 meter (10 foot) plot length was consistent with EPA  recommendations. The 4.1 meter
(13.5  foot) width was  selected to recover soil from the edge of the asphalt (at the start of the
concrete gutter) to the edge of the opposite edge of the test lane.  For experiments evaluating the
effects of vehicle passes on surface silt depletion, 0.61 meter (2 foot) wide "Quickie-Strips" were
laid out in the zones between the full-size plots.

    Canister vacuum cleaners with hard-floor inlets were used to recover applied soil from the
roadway sites into pre-tared vacuum bags. Three soil recovery techniques were used during the
study:

    •   Soil from one  large heavily soiled plot would be recovered  into one pre-tared vacuum
       bag.
    •   Soils from two lightly soiled  large plots, sampled at  the same time (before or after a
       particular vehicle pass) would be accumulated into one vacuum bag.
    •   Soil from a series of Quickie  strips, sampled in sequence after a specific vehicle pass
       would be accumulated into one vacuum bag.

    Road dust emission factors were then  calculated for the silt loadings using the AP-42
emission factor equation:

                                E=k (sL/2)0'65 (W/3)1'5 - C

where   E  =  particulate emission factor (having units matching the units of k)
         k  =  base emission factor for particle size range and units of interest
         sL  =  road surface silt loading (grams per square meter) (g/m2)

-------
        W  =  average weight (tons) of the vehicles traveling the road
        C  =  emission factor for 1980's vehicle fleet exhaust, brake wear and tire wear

A weight  of 2.88 tons,  based on the  arithmetic average  of the reported weights of the three
mobile source vehicles (SCAMPER 2.5 tons, TRAKER I 3.4 tons, and TRAKERII
2.75 tons) was used to calculate the AP-42 emission factors from the silt loadings.

     c.    SCAMPER Technology

    The  SCAMPER  determines PM  emission rates from  roads  by measuring the  PMio
concentrations in front of and in the wake of the test vehicle using DustTrak monitors. As a first
approximation, the concentration difference (mg/m3) is multiplied by the vehicle's frontal area
(3.66  m2)  to obtain an emission factor in units of mg/m. The particle monitor for the vehicle
wake is mounted on a small trailer with a flat bed, so that the vehicle wake was disturbed as little
as possible.  The inlet for the wake monitor, which is 10 ft behind the rear of the vehicle, allows
sampling as  isokinetically as possible over the full range of vehicle speeds. A GPS determines
vehicle location and speed, and a PC collects 1-sec data from GPS and PMio measuring devices.

     d.    TRAKER Technologies

    TRAKER I is comprised of a van that is equipped with three exterior steel pipes acting as
inlets  for the onboard instruments.  Two of the pipes are located behind the left and right front
tires and are used to measure emissions from the tires.  The third pipe is the inlet for background
air and runs along the centerline of the van underneath the body and extends through the front
bumper. The background measurement is used to correct the measurements behind the tires for
fluctuating dust and exhaust emission contributions  from other vehicles on the road.  Separate
DustTraks are connected to each of the left and right inlet lines as well as on the middle inlet
line. A central computer collects all the data generated by the onboard monitors as well as GPS
coordinates,  and vehicle  speed and acceleration with a 1-second frequency.

    The  TRAKER II  inlet  lines are configured  so that  on  unpaved  roads,  where  PMio
concentrations behind the front tires could exceed the particle monitor upper limit (150 mg/m3),
clean  air can be mixed with air from the wheel well inlets in a controlled manner to achieve a
desired amount of dilution.  Instead of an onboard sampling plenum as in TRAKER I, a 10-cm
diameter external pipe is used to channel/dilute inlet flow into a  manifold with connections to
particle monitors.  The  circular inlets  used currently on TRAKER I are replaced by flattened
manifolds on TRAKER II.

4.   Study  Design

    The test conditions for the Phase IV study are summarized in Table 1. Further explanation of
these conditions is provided in the following paragraphs.

-------
Table 1-1.  Study Measurement Conditions
Q
M
-w
i/
VI
1

2

3


4


5


6


7


8


9
10

11


12


13

.8
s
Q
9/11
9/11
9/11
9/11
9/11
9/11
9/12
9/12
9/12
9/12
9/12
9/12
9/13
9/13
9/13
9/13
9/13
9/13
9/13
9/13
9/13
9/14
9/14
9/14
9/14
9/14
9/14
9/14
9/14
9/14
9/15
9/15
9/15
Approximate Time (local)
11:55-13:15
13:35
13:52-14:18
14:30
15:17-26:30
17:00
9:15
10:15-11:00
11:05
13:00
13:35-14:40
15:00
9:00
9:40-10:25
11:09
12:15
12:45-13:35
14:00
14:45
15:20-16:15
17:00
8:00
8:40-9:20
9:20-9:50
10:05
10:25-11:20
11:30
12:30
13:10-14:05
14:30
8:00
8:30-11:15
11:30
>.
j*
u
<
Test: Baseline road conditions - No Sweep,
No silt
Sweep
Test: After Sweeping, No silt applied
Silt applied to test road
Test: After application of silt, 35 mph
Sweep
Silt applied to test road
Test: After application of silt, 45 mph
Sweep
Silt applied to test road
Test: After application of silt, 25 mph
Sweep
Silt applied to test road
Test: After application of silt, 45 mph
Sweep
Silt applied to test road
Test: After application of silt, 25 mph
Sweep
Silt applied to test road
Test: After application of silt, 45 mph
Sweep
Silt applied to test road
Test: Depletion of silt resulting from
vehicle passes
Test: Measure emissions prior to sweeping
Sweep
Test: Measure emissions after sweeping
Sweep
Silt applied to test road
Test: Speed tests
Sweep
Silt applied to test road
Test: Speed tests
Sweep
Vehicles used
All test vehicles
Street Sweeper
All test vehicles
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
SCAMPER Only
All test vehicles
Street Sweeper
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Nominal speed (mph)
35
NA
35
NA
35
NA
NA
45
NA
NA
25
NA
NA
45
NA
NA
25
NA
NA
45
NA
NA
35
35
NA
35
NA
NA
25-45
NA
NA
25-45
NA
Total passes/passes per vehicle
60/20
NA
30/10
NA
27/9
NA
NA
30/10
NA
NA
42/14
NA
NA
30/10
NA
NA
30/10
NA
NA
36/12
NA
NA
10/10
12/4
NA
30/10
NA
NA
27/9
NA
NA
84/28
NA

-------
     a.   Passes, Runs, and Sets

    Phase IV consisted of a total of 13 test sets,  each encompassing different  experimental
conditions. Most sets consisted of approximately 30 vehicle passes, and each pass was identified
separately by the type  of mobile  sampling technology used and the time it passed by the flux
tower.   A run  typically  consisted of three  successive passes,  one by each  mobile sampling
technology.

    The first test set was performed prior to any sweeping or application of soil/silt to the roads,
and is representative of the natural condition of the road.  The second set was performed after the
road had been cleaned by a street sweeper. Sets 3-9 consisted of applying a controlled amount of
soil/silt to the road prior to the first  pass.  During these sets, the road was swept before each
soil/silt application.  There was no soil/silt application for Sets  10  and 11, but the road was still
swept between  the sets.  Vehicle speed  was held constant at 25, 35, or 45  mph for Sets 1-11.
Prior to Sets 12 and  13, soil/silt was applied to the road and the  speeds of the vehicles varied in
cycles from 25 to 35 to 45 mph and back from 45 to 35 to 25 mph. Each speed was held for one
run (one pass of each mobile technology).

    During each pass within a set, emissions  on the master tower were recorded along with the
signal of the particular mobile technology.   In most cases, silt samples were taken for AP-42
calculation at the beginning and end of each set.

     b.   Silt loadings

    An area soil was selected  for application to the test course with a measured silt fraction
(14%) approximating the 65th  percentile of 35 sieved road dust silt samples taken from all three
roadway categories in calendar years 2005-2006. The soil  was passed though a 1.18 mm sieve
opening during collection to remove gravel and vegetative matter. A fertilizer drop  spreader was
used for soil application with a constant pull  speed of 10 mph.  Prior to the first application of
soil, a group of preliminary measurements (sets 1 and 2) by the  mobile PMio sampling vehicles
were used to characterize the PMio emission rates of the natural road soil before  and after the
road was swept.   The  silt loading values along with other study design details  are found in
Table 2.

     c.   Quality Assurance

    Quality assurance  focused on flow and  concentration measurement within the operating
ranges  of the DustTrak and reference monitors, filter and bag handling and weighing for mass-
based sampling, and suitability of wind conditions for each plume passing the flux tower.

-------
                   Table 1-2  Silt Loadings and Other Test Conditions
Date
9/11/06
9/11/06
9/11/06
9/12/06
9/12/06
9/13/06
9/13/06
9/13/06
9/14/06
9/14/06
9/14/06
9/14/06
9/15/06
Set#
1
2
3
4
5
6
7
8
9
10
11
12
13
Experiment
Name
Pre-Sweep
Post-Sweep
Apply silt #1
Apply silt #2
Apply silt #3
Apply silt #4
Apply silt #5
Apply silt #6
Apply silt #7 -
Depletion,
SCAMPER only
Continuation of
silt #7-
Depletion, all
vehicles
Post-sweep
Apply silt #8 -
strong winds
Apply silt #9 -
strong winds
Start
Pass ID
1
63
93
140
170
212
243
273
309
319
334
365
392
End
Pass ID
60
92
139
169
211
241
272
308
318
331
364
391
476
Nominal
Drive
Speed
(mph)
35
35
35
45
25
45
25
45
35
35
35
Repeat
25,35,45,
45,35,25
cycle twice
Repeat
25,35,45,
45,35,25
cycle 4 1/2
times
Applied Soil
Loading
(gram/rrr)
N/A
N/A
6.16
17.17
16.58
4.99
4.70
7.63
7.78
7.78
N/A
17.61
28.47
Avg.
Recovered
Silt Loading,
(gram/m2)
0.17
N/A
0.75
2.48
3.17
0.88
0.74
1.14
0.80
0.80

2.55
2.31
    Two factors were used to determine if a specific tower flux measurement associated with an
individual vehicle pass was valid. First, the one-second wind direction over the duration of the
three intervals associated with a mobile monitor pass (pre-peak background, peak, and post-peak
background) was examined.  In cases where the average wind direction over the three intervals
was within 45 degrees of the perpendicular line drawn between the tower and the road segment
and the wind speed was  relatively constant (i.e. holding at >  1 m/s from the same general
direction), the wind direction was considered valid. If the wind direction was always less than
75 degrees from the perpendicular, the wind speed was relatively constant, and fluctuations in
wind direction did not exceed 30 degrees, the wind direction was considered valid.  In all other
cases, wind conditions were considered to invalidate the horizontal flux measurement.

    The second factor in determining the validity of a  specific tower measurement was the noise
level of the baseline PMio concentration.  During periods of high wind, non-traffic dust clouds
often passed by the  flux tower (especially true on the  last two days of testing).  These high and
spurious concentrations of PMio prevented subtraction of a baseline value from the plume impact
concentration.  In other cases, the passage of a large vehicle on the  south side of Veterans
Memorial Highway would sometimes result in a temporary spurious baseline reading.  The entire
time series of data from the flux tower was examined  to flag periods when the baseline was too
noisy for a measurement. Those data were considered  invalid.

-------
5.   Data Analysis

     a.   Data Averaging

    To compare PMio tower flux measurements with AP-42 silt methodology and mobile system
measurements,  data  were  averaged  by  measurement  set.   For  each  set  all  tower flux
measurements  were  averaged together  regardless  of the test  vehicle.   Thus,  tower flux
measurements  represent average fluxes for all vehicles.   This was to ensure that all methods
examined would be calibrated (or compared in the case of AP-42) against the same standard and
that results from future measurements can be compared using a  common basis. A minimum
criterion of 10 valid vehicle passes was applied to the tower flux average value.

    DRI combined the following data sets (using Vehicle Pass_ID as a common variable) into a
master Excel database that was used for joint data analysis:

    •  UNLV AP-42 emission factor data, averaged north and south for each pass,
    •  Tower mass emission rate data, averaged for each pass,
    •  SCAMPER, TRAKER I and TRAKER II mobile technologies data, averaged for each
       pass.

The Excel® database (containing date,  time, vehicle Pass_ID, vehicle speed, silt loadings and silt
loading uncertainties) and  AP-42 emission factors and  emission  factor uncertainties  were
transmitted to all cooperating agencies for data analysis.

    The  TRAKER signal was averaged over the full test route, rather than only using values
obtained  near the  master flux tower. It was found that there is a good correlation (R2 =  0.82)
between  the pass-averaged TRAKER I  signal  and the TRAKER I  signal averaged over data
points that correspond to measurements within 50 m of the master tower.  The SCAMPER data
were collected at 1-sec intervals, and the front DustTrak value was  subtracted from the rear value
to yield a net value in mg/m3. Pass averages were calculated from the net values calculated at 1-
sec intervals. The  background correction was generally small, and negative emission rates were
not encountered.

     b.   Conversion of Particle-Monitor Readings to PMIO Concentrations

    Several cross-comparisons were performed to determine the ratio between the DustTrak
reading and the PMio mass-based concentration measured by a collocated reference sampler.
First, the DustTrak located at 2.1 m on the master flux tower was compared to  the  TEOM
measurements  at 2.3 m, also located on the master tower.  The correlation between the DustTrak
and TEOM on the master tower was quite noisy, but showed that DustTrak values would have to
be multiplied by a factor of 2.8 ± 0.6 to obtain mass-equivalent PMio.

    Second, controlled laboratory tests were  used to more  accurately obtain a relationship
between the DustTrak measurements and mass-based measurements. For this purpose,  a  well-
mixed  chamber was constructed, within which  the same  silt material that was used in the field
study was injected and  suspended. Measurements of the PMio inside the  chamber were made

-------
with the DustTrak as well as filter  samples.   These tests generated a DustTrak correction
multiplier of 2.4, which was chosen for use in this program.  The in-lab measurements resulted
in a higher  correlation, due to the fact that in the field the DustTrak and  TEOM were only
nominally collocated, whereas in the lab the two instruments sampled a well mixed volume of
air.

    In prior studies with the SCAMPER, the response of the  rear DustTrak was compared to
mass determined by a collocated filter  sampler on the trailer of the vehicle. The average response
factor based on a linear regression  was approximately 3. Given the scatter of the data, this is in
general agreement with the correction factor of 2.4 cited above.

    It should be noted that in this study the factor of 2.4 was used only to correct tower sampler
readings to mass-based PMio concentrations.
6.   Results and Conclusions

     a.   Calibration Factors

    Calibration  factors  were  developed  for  each  mobile monitoring  technology.    Each
multiplicative factor represented the ratio of the PMio emission factor from the flux tower to the
raw  mobile  monitor  reading (mg/m3).   For  each  mobile monitoring technology, a single
calibration factor was developed for each test set, using average tower flux values and  average
mobile  monitor readings.   Then  regression analysis  of the  individual factors  was used  to
calculate an  average  calibration factor  for each technology.  These factors are  presented  as
coefficients in the following equations for the PMio emission factor (EF):

       TRAKERI                 EF = 0.54*TI         [correlation of 0.57]

       TRAKERII                EF = 0.92*TII        [correlation of 0.75]

       SCAMPER                 EF = 20*SC          [correlation of 0.47]

Each  coefficient is used  as  a multiplier to  convert the mobile monitor  (DustTrak)  reading
(mg/m3) to the equivalent PMio emission factor (g/vkt).

     b.   Initial Emission Decay

    In the context of the present study, the test data indicate that dust emissions occur  under a
different regime during the first 9 vehicle passes than in ensuing passes. Since for a paved road,
the volume of vehicles is  generally much higher than  9 per day, the first 9  passes after silt
material application probably do not reflect the regime under which real-world dust emissions
occur.  It is more likely that the latter passes (greater than 9) more accurately reflect the slower,
steadier emissions of PMio road dust that occurs on paved roads.

-------
    The  TRAKER I signal  decay with vehicle passes matches AP-42 silt loading decay in
Sets 5, 8, and 10 for cases of constant vehicle speed. However, TRAKER I measured emissions
also showed, in sets 12 and 13, clear vehicle travel speed dependence that is not accounted for in
the current AP-42 emission factor equation. The rising and falling TRAKER I signals in Sets 12
and 13 are a result of systematically varying vehicle speeds first rising from 25 to 35 to 45 mph,
then declining  from  45 to 35 to 25  mph. Silt loadings in Set 12 declined throughout the
experiment, even though TRAKER I  emissions increased with increasing vehicle speed.  Silt
loadings in Set 13 declined rapidly to a steady state value, while TRAKER I emissions fluctuated
regularly with rising and falling vehicle speed.  TRAKER II and SCAMPER signals showed
similar behavior.

     c.    MM Technologies vs. AP-42 Methodology

    Two conclusions can be made from the test results obtained in this study, when comparing
mobile monitoring technologies with the AP-42 methodology:

    •  The  calibrated mobile methods measured  emission  factors  that were about  1.5 times
       higher than found with the AP-42 methodology when higher silt loadings were applied
       to the test road.
    •  The mobile methods tracked each other quite well under most conditions.

    The first conclusion appears to reflect a different  silt mobilization process, which occurred
as a result of silt being distributed on top the embedded road surface aggregates and hence being
more easily entrained by vehicle mechanical and aerodynamic shear.  In contrast the aged silt
found on most roads is more likely to be embedded between the road surface aggregates.

    Throughout this  field  study,  the implementation  advantages   of  mobile  monitoring
technologies were evident. The mobile monitoring technologies were found to provide for much
easier representation of spatially distributed roadway emission characteristics, while eliminating
the need to divert traffic.  The one limiting factor for mobile monitoring was high winds which
made the monitored plume  concentration difficult to  differentiate from  higher than normal
background levels.

-------
This page intentionally left blank

-------
                               TABLE OF CONTENTS

1.0   INTRODUCTION	1
      1.1  Study Objectives	1
      1.2  Study Design Overview	2

2.0   BACKGROUND	3
      2.1  EPA AP-42 Development and Limitations	3
      2.2  Clark County Background with AP-42	5
      2.3  Paved Road Phase I-Phase III	6

3.0   METHODOLOGY	8
      3.1  Experimental Design	8
           3.1.1 Route Selection	8
      3.2  Soil Selection and Application	10
           3.2.1 Soil Sampling Site Selection	10
           3.2.2 Soil Excavation and Packaging	11
           3.2.3 Soil Characterization	12
           3.2.4 Soil Application	12
      3.3  Horizontal Flux Tower	13
      3.4  EPA Method AP-42	16
           3.4.1 Plot Layout	16
           3.4.2 Vacuum Soil Recovery Methods	19
           3.4.3 Field Soil Application History	20
      3.5  Mobile Technologies	21
           3.5.1 SCAMPER	21
           3.5.2 TRAKERI	26
           3.5.3 Inlet configuration	27
           3.5.4 TRAKERII	29

4.0   QA/QC	32
      4.1  Horizontal Flux Tower	32
      4.2  EPA Method AP-42	38
           4.2.1 Field Balance Mass Calibration	38
           4.2.2 Road Plot Marking Uncertainty	38
           4.2.3 Sieve Analysis Calibration	39
      4.3  Mobile Technologies	40
           4.3.1 SCAMPER	40
           4.3.2 TRAKERI	40
           4.3.3 TRAKERII	43

5.0   DATA HANDLING	43
      5.1  Horizontal Flux Tower	43
      5.2  EPA Method AP-42	45
           5.2.1 Organizing Bag Data	45
           5.2.2 Organizing AP-42 Emission Factor Data	45

-------
      5.3  Mobile Technologies	47
          5.3.1  SCAMPER	47
          5.3.2  TRAKERI	47
          5.3.3  TRAKERII	58

6.0    RESULTS	61
      6.1  Short-Term Emission Factor Decay and Silt Loading Depletion	63
      6.1.1 Silt Loading Depletion	63
      6.2  Comparison of Horizontal Flux Tower Emission Factors to EPA Method AP-42.... 67
      6.3  Comparison of Horizontal Flux Tower Emission Factors to Mobile Technologies
          Emission Factors	72
          6.3.1  TRAKERI	72
          6.3.2  TRAKERII	74
          6.3.3  SCAMPER	76
      6.4  Comparison of Calibrated Mobile Technologies Emission Factors to EPA Method
          AP-42 Emission Factors to measured PM10 Horizontal Flux Tower Values	79
      6.5  Comparisons of SCAMPER "First Principles" EF With TRAKER and AP-42	83

7.0    DISCUSSION	83
      7.1  Real World Precision and Reproducibility	83
          7.1.1  UCR Paved Road Phases II & III for DAQEM	83
          7.1.2  DRI Studies—Clark County Phase II, Lake Tahoe and Idaho	84
      7.2  Applying Phase IV Results in Real World Conditions- Explanation of Higher
          EF'sinPhaselV	88
      7.3  Advantages of Mobile Technologies	89
      7.4  Paved Road Dust Emission Inventory Development	90

8.0    CONCLUSIONS	92
      8.1  Conclusions	92
      8.2  Recommendations	93

9.0    References	95

10.0  Acknowledgments	102

-------
                                 LIST OF FIGURES

Figure 1-1.  Layout of Test Course	v
Figure 2-1.  Map of Clark County 2/14/05 - 2/17/05 sampling route	7
Figure 3-1.  Phase IV Route Map (Veterans Memorial Blvd, Boulder City)	10
Figure 3-2.  Photograph of Master (left) and Satellite (right, not used in present study) Towers
            Showing Locations of DustTrakPMio Monitors	14
Figure 3-3.  Schematic of Field Sampling Layout	15
Figure 3-4.  Phase IV Veterans Memorial Drive Plot Layouts	18
Figure 3-5.  Isokinetic Inlet Schematic Diagram	23
Figure 3-6.  Photographs of the Front and Rear of the SCAMPER	25
Figure 3-7.  TRAKER Influence Monitors Measure the Concentration of Particles Behind the
            Tires	26
Figure 3-8.  TRAKER Vehicle and Instrumentation	28
Figure 3-9.  TRAKER II.  Vertical Inlet Pipe Near the Passenger-Side Door is Used to Sample
            Background Air for the Right Side Inlet	30
Figure 3-10. Schematics and Dimensions of TRAKER II	31
Figure 4-1.  Scatter Plot of DustTrak PMio Average Concentrations and TEOM PMio
            Measurements	33
Figure 4-2.  The Resuspension Chamber Used to Establish the Relationship between the
            DustTrak-derived PMio and PMio derived by Gravimetric  Analysis	35
Figure 4-3.  Relationships Between Gravimetrically Determined Average PMio and Average
            PMio	36
Figure 4-4.  Scatter Plot of DustTrak Monitor Outfitted With PM2.5 Inlet versus DustTrak With
            PMio Inlet	37
Figure 4-5.  Comparison of Filter-Based PM2.s Mass Measurement With DustTrak Outfitted
            With PM2.5 Inlet	38
Figure 4-6.  TRAKER Coefficient of Variation Expressed as a Percentage for left and right PMio
            DustTrak  signals as a Function of Speed	41
Figure 5-1.  Illustration of Portions of Flux Plane Represented by DustTrak and Wind
            Instruments at Each Height	44
Figure 5-2.  Schematic of GPS Data Points on Top of Street Layout	51
Figure 5-3.  Relationship Between TRAKER I Signal Averaged Over Entire Pass (route length)
            and TRAKER I Signal Only Within 50 m of Master Tower	52
Figure 5-4.  TRAKER  Signal (left and right) Averaged Over Entire Pass (pass-avg.) and
            Averaged Over Only the Portion of Test Road in the Vicinity of Flux Tower
            (Tower-avg.)	52
Figure 5-5.  Ratio of Background (middle) Inlet PMio Concentration to  Average of Left and
            Right Tire Inlet PMio Signals for TRAKER I Passes	53
Figure 5-6.  Time Series of Ratio of Pass-Averaged TRAKER I Right to Left Inlet Signal Ratios
            and Pass-Averaged Wind Speed (m/s)	53
Figure 5-7.  TRAKER I Signal Normalized to First TRAKER I Pass of  the Measurement Set for
            Sets 4, 5, 6, 7, and 8	55
Figure 5-8.  Normalized TRAKER I Decay Curve for Sets  5 and 7 (25 mph measurement) and
            Hypothesized Aerodynamically Suspendable, Mechanically Suspendable, and Total
            Suspendable (aerodynamic plus mechanical) Road Dust Decay Curves	56
                                                                                   in

-------
Figure 5-9.  Normalized TRAKER I Decay Curve for Sets 4, 6, and 8 (45 mph measurement) and
           Hypothesized Aerodynamically Suspendable, Mechanically Suspendable, and Total
           Suspendable (aerodynamic plus mechanical)  Road Dust Decay Curves	56
Figure 5-10. Division of Set 13 Into Four Cycles, With Each Cycle Comprised of Six Passes for
           TRAKER 1	57
Figure 5-11. Speed Response of TRAKER I Signal	58
Figure 5-12. Relationship Between TRAKER II Signal Averaged Over Entire Pass (route
           length) and TRAKER II Signal Only Within 50 m of Master Tower	59
Figure 5-13. TRAKER II Ratio of Average of Left and Right Tire Inlet PMio Concentrations to
           Background  (middle) InletPMio Concentration	60
Figure 5-14. TRAKER II Time Series of Ratio of Pass-Averaged Right to Left Inlet Signal
           Ratios and Pass-Averaged Wind Speed (m/s)	60
Figure 5-15. Speed Response of TRAKER II Signal	61
Figure 6-1.  Time Series of Pass-Averaged Horizontal Tower PMio flux (g/vkt), Silt-estimated
           AP-42 Emission Factor (g/vkt), TRAKER I, TRAKER II, and SCAMPER raw
           signals (mg/m3)	63
Figure 6-2.  Silt Depletion With Increasing Vehicle Passes	64
Figure 6-3.  Comparison of Averaged AP-42 Emission Factors, in gram/VMT, Computed From
           Silt Loadings for First Nine Passes, Compared to AP-42 Emission Factors for
           Remaining Passes	65
Figure 6-4.  Comparison of TRAKER I Signal and Average North-South Silt Loading for All
           Vehicle Passes	66
Figure 6-5.  Comparison of SCAMPER Signal and Average North-South Silt Loading for All
           Vehicle Passes	67
Figure 6-6.  Time Series of Horizontal PMio Fluxes Measured With Tower Measurement System
           for Different Test Vehicles	68
Figure 6-7.  Tower-Based PMio Emission Factors versus AP-42 Silt-Based Emission Factors ..71
Figure 6-8.  Time Series of Measured Horizontal PMio Flux on the DRI Tower System and the
           Pass-Averaged TRAKER I Signal for Passes When the Horizontal Flux
           Measurement was Valid	73
Figure 6-9.  PMio Emission Factors versus TRAKER I Average Signal	74
Figure 6-10. Time Series of Measured horizontal PMio flux on the DRI Tower System and the
           Pass-Averaged TRAKER II Signal for Passes When the Horizontal Flux
           Measurement was Valid	75
Figure 6-11. PMio Emission Factors versus TRAKER II Average Signal	76
Figure 6-12. Time Series of Measured Horizontal PMio Flux on the DRI Tower System and the
           Pass-Averaged SCAMPER  Signal for Passes When the Horizontal Flux
           Measurement was Valid	78
Figure 6-13. PMio Emission Factors versus SCAMPER Average Signal	79
Figure 6-14. Emission Factors (g/vkt) For All Valid Passes	80
Figure 6-15. Comparison of Set-Averaged Emission Factors (g/vkt)	81
Figure 6-16. Set Averaged TRAKER IEF, TRAKER IIEF, and AP-42 Silt-Based EF Plotted
           Against SCAMPER EF	82
Figure 6-17. Set Averaged TRAKER II EF, SCAMPER EF, and AP-42 Silt-Based EF Plotted
           Against TRAKER I EF	82
Figure 7-1.  TRAKER I Calibrations	86
                                                                                  IV

-------
Figure 7-2. Emission Factors (g/vkt) From Phase II Clark County Study	87
Figure 7-3. Scatter Plot of TRAKER IEF (g/vkt) versus AP-42 Silt-Based EF (g/vkt)	88

-------
                                       LIST OF TABLES

Table 1-1.   Study Measurement Conditions	viii
Table 1-2    Silt Loadings and Other Test Conditions	x
Table 3-1.   Summary of Data Used to Determine 50* Percentile Silt Loading Value for
            Collector Roadways	11
Table 3-2.   Summary of Applied Silt Loadings During Phase IV Controlled Field Study—
            Veterans Memorial Boulevard. Boulder City, NV	21
Table 5-1.   Example of Vehicle Pass_ID Data	48
Table 6-1.   Summary of Tests During Field Study (9/11/06 - 9/15/06)	62
Table 6-2.   Summary of Observed Silt Decay With Increasing Number of Vehicle Passes	64
Table 6-3.   Summary of Measured PMio Horizontal Fluxes	69
Table 6-4.   Summary of Equivalence Multipliers Between Mobile Measurement Systems
            and PMio Emission Factors Assuming that the Raw Signal for the Mobile
            Systems is Linearly Related to Measured Emission Factors	77
                                                                                   VI

-------
1.0 INTRODUCTION

The Las Vegas Valley in Clark County, Nevada, is classified as a serious nonattainment area for
federal fine paniculate matter (PMio) National Ambient Air Quality Standards (NAAQS).  Clark
County submitted a PMio State Implementation Plan (SIP) for this nonattainment area in June of
2001.  As part of the SIP development, Clark County contracted with a consultant to collect 24
silt samples representative of Clark County roadways for estimating PMio paved road emissions.
The  silt measurements were significantly higher than EPA default values, and public works
officials  from four  agencies  and other stakeholders  asserted that the Clark County  SIP
overestimated PMio emissions  from paved roadways.  Clark County committed to conducting
quarterly silt sampling through  the end of 2006 as part of the now federally approved PMio SIP.
Sampling is ongoing and the current AP-42 data base includes sampling from the spring of 2000
through the spring of 2006.  The PMio SIP also contained a research commitment to explore the
feasibility of vehicle-based  mobile sampling systems for development of improved  paved road
emissions inventories.
During this timeframe, Clark County has seen substantially improved air quality for the
pollutant, particularly from the year 2004 forward.  Visually, it  also appears that Las Vegas
Valley roads have become cleaner, in part due to tightened controls on construction site track-out
and an increased  emphasis on enforcement, implemented in early 2003.  However, statistical
analysis performed by UNLV under contract has generally not shown  statistically  significant
declines in paved  road emission factors during this timeframe using silt sample data and AP-42
emission estimation methods.  These results have reinforced Clark County's belief that the paved
road emissions inventory developed using AP-42 methods for the PMio SIP overestimates actual
emissions. In addition, silt measurements are time consuming, expensive, and frequently require
the alteration of roadway traffic patterns and  increased traffic congestion while samples are
being procured.

Initial work utilizing vehicle-based mobile sampling systems in Clark County occurred in  1999
as part of PMio SIP development.  The test  results showed even higher emission rates than
corresponding AP-42 calculations and were not considered realistic.  In addition, the need to
complete an approvable PMio SIP was urgent and EPA approval of this new method was very
unlikely based  on work completed at that time.  Phase I of the current research  effort was
initiated in 2004 and  Phase II was completed in early 2005.  Fieldwork for Phase III occurred in
late 2005. Objectives for Phase IV are described below.

1.1 Study Objectives

    1   Evaluate  precision  of  all  measurement methods  under  controlled  conditions:
       Measurement  methods  include measurements  from  the  tower  sampling  array,
       SCAMPER measurements, TRAKER measurements, and road silt measurements using
       AP-42 sampling methodology.  Additional ancillary measurements include weights of
       silt material  applied to test area, wind speed  data, wind direction data,  and relative
       humidity data.

-------
    2   Evaluate validity of original AP-42 emissions factor estimates: Compare measured
        tower emissions to AP-42 emissions calculated from silt loadings using the AP-42
        equation.

    3   Calibrate mobile technologies systems to the tower emissions factors: Comparison of
        SCAMPER  and TRAKER  system  measurements with  external sampling  array
        measurements  in  a  controlled measurement environment,  with  defined  vehicle
        movement, controlled speeds, and controlled road material loadings.

    4   Compare mobile technologies emissions factors to predicted AP-42 emissions factors:
        Determine relationships between roadway  silt loading and measured SCAMPER and
        TRAKER particulate emissions under controlled  conditions  (standard vehicle speeds
        and  weight).  Compare  SCAMPER/TRAKER  measurements to AP-42  emission
        estimates under controlled conditions.

    5   Compare mobile technologies measurements: Comparison of SCAMPER to TRAKER
        measurements  estimates under controlled measurement  conditions, including defined
        vehicle movement controlled speeds, and controlled road material loadings.

    6   Data  assessment and  review  for  recommendations on performance  specifications:
        Assess data for accuracy and precision of vehicle-mounted mobile sampling systems
        and  compare with other  measurement  methods.  Prepare recommendations for  the
        utilization of vehicle-mounted mobile sampling systems into AP-42.

    7   Characterization of silt depletion rate: Assess by number of vehicle passes with defined
        vehicle speeds and weight.

1.2 Study Design Overview

The five-day study included testing two vehicle-mounted mobile sampling systems, SCAMPER
and TRAKER, under controlled road conditions.  One SCAMPER and two TRAKER systems
were utilized in this study.  Comparative external measurements included horizontal PMio flux
measurements with multiple samplers on a nine-meter tower and  AP-42 silt sampling.  Study
objectives included a comparison of tower  emissions measurements to  SCAMPER/TRAKER
measurements, a  comparison of SCAMPER to TRAKER  measurements, and  AP-42  silt
measurements/emission estimates under controlled conditions.

The sampling area consisted of two lanes of a four-lane divided  highway with curbed median
and curbed roadsides (see Figure 3-1).  All road traffic was diverted to the southeast-bound
lanes, allowing the two northwest-bound lanes and the stabilized median area to be utilized
exclusively for the five-day  study.   This diversion  allowed the research team to limit vehicle
passes between the external tower samplers to SCAMPER and  TRAKER vehicles, with  the
tower located either on the median between the test  area and adjacent traffic or on the sidewalk
on the test side of the road.  It was  anticipated that these controlled traffic and measurement
parameters would enhance the quality of the horizontal PMio flux measurements compared to
previous paved road dust studies.

-------
Controlled road silt loading conditions were created through the application of known quantities
of material onto the measurement section of the test area. The applied material approximated the
sand and silt/clay percentages historically sampled on paved roads in the Las Vegas Valley.  The
test area was of sufficient length to allow for measurement at constant speeds of up to 45 miles
per hour.
2.0 BACKGROUND

2.1 EPA AP-42 Development and Limitations

The  United States  Environmental Protection Agency (EPA) published a document entitled
Compilation of Air Pollutant Emission Factors (AP-42) beginning in 1972.  Since AP-42s
inception as a tool for regulators, permit writers, and environmental planners, many have used
this tool to account for  emissions of air pollutants from a variety of sources in the human
environment.  EPA periodically reviews and updates the emission factors available in AP-42 to
meet the needs of state and local air pollution control programs and industry.  It wasn't until the
late 70's that EPA, and others started looking at emissions from paved roads.  Prior to this time,
much of the work with respect to roadways was focused on unpaved roads.  Prior to the March
1993  research findings1,  AP-42 contained two   sections  concerning paved road fugitive
emissions.  One of the early attempts  to characterize paved road dust was addressed by EPA in
1983 with the  inclusion of a Section 11.2.6 Industrial Paved Roads, and was slightly modified in
1988. Section  11.2.5, Urban Paved Roads, was first drafted in 1984 using the test results from
public paved roads and was included in the AP-42, 4  Edition documentation in  1985.   The
emission factors included in Sections 11.2.5  and 11.2.6 were never quality rated "A" through
"E."  The  updates proposed with the March 1993  report assumed there were no distinctions
between public and industrial  roads or between controlled  and non-controlled test.   These
assumptions evolved into a single emission factor equation for all paved roads.

In July  1993, the AP-42 Section 13.2.1 (Paved Roads) was published to help better characterize
the paved road dust source. The quantity of dust emissions from vehicle traffic on a paved road
could be estimated using the following empirical expression:

                                 E=k (sL/2)0'65 (W/3)1'5                      Equation 1.1

where  E  =  particulate emission factor (having units matching the units of k)
        k   =  base emission factor  for particle size range and units of interest
        sL =  road surface silt loading (grams per square meter) (g/m2)
        W =  average weight (tons) of the vehicles traveling the road

This equation was slightly modified in 2004 to account for vehicle exhaust, tire and brake wear.
In the most recent  version the quantity of particulate emissions from re-suspension of  loose
material on   the  road   surface   due   to   vehicle  travel   on  a   dry   paved  road  is
1 U.S. Environmental Protection Agency, Emission Factor Documentation for AP-42, EPA Contract No. 68-DO-
0123, MRI Project No. 9712-44 dated March 8, 1993.

-------
estimated by using the following empirical expression.

                                E=k (sL/2)0'65 (W/3)1'5 - C 5                     Equation 1.2

where    E  =  particulate emission factor (having units matching the units of k)
         K  =  base emission factor for particle size range and units of interest
         sL =  road surface silt loading (grams per square meter) (g/m2)
         W =  average weight (tons) of the vehicles traveling the road
         C  =  emission factor for 1980's vehicle fleet exhaust, brake wear and tire wear

The  AP-42 equation variable for weight of vehicle is defined as the average weight of all
vehicles  traveling the road.  EPA did not intend that separate weights of vehicles be used to
calculate a separate emission factor for  each vehicle weight class.  Instead, only one emission
factor is  calculated to represent the  "fleet"  average weight of all vehicles traveling the road or
road network. The particle size multiplier  (k) above varies with aerodynamic size range.  The
emissions factors for the exhaust,  brake wear and tire wear are for a 1980's vehicle fleet (C), as
calculated by EPA's MOBILE6.2 model.

The AP-42 paved road emissions equation is an arithmetic equation based on 65 tests conducted
in the early 1990s.  The test included measurements of vehicles moving at speeds of 10 to 55
miles per hour. The equation is intended for estimating emissions from free flowing traffic and
is not intended to estimate emissions for stop  and go traffic.  Where road specific silt loading
factors are utilized, the EPA assigns a  quality rating of "A" provided the silt loadings mean
vehicle weight and vehicle speeds  fall within the following parameters:

Silt loading:                0.02 - 400 g/m2 [0.03 - 570 grains/square foot  (ft2)]
Mean vehicle weight:        2.0-42 tons
Mean vehicle speed:         6-88 kilometers per hour  (kph) [10 -55 miles per hour
                           (mph)]

Where the EPA recommended  default  silt  loadings are used in place of locally measured  silt
loadings, the  quality rating is reduced by one level (e.g. "B").  The EPA provides default values
for High  ADT and Low ADT roads.  Each of these two ADT classes has default silt loading for
normal conditions and worst-case conditions.

The assumptions,  limitations, and silt loading data collection requirements  needed to utilize the
equation  considerably diminish  the accuracy of emissions inventories for paved road emissions.
Urban areas,  where a majority of vehicle travel occurs in most airsheds, typically do not have
free flowing traffic.  Vehicle speeds  have been shown to exert substantial influence on road dust
emission rates, but the equation lumps all  speeds from 10 to 55 mph into one  emissions rate.
Speeds above 55 mph, which  may comprise a  significant component of the  vehicles miles
traveled in an airshed; are not accounted  for at all, introducing additional error into the emissions
estimates.

The  determination of correct silt loading  values for each class of  roadway and subclass of
roadway  is the most serious limitation of the  AP-42 methodological approach.  Road  silt

-------
sampling is expensive, time consuming, and dangerous. As a result, only a few silt samples can
be collected in each  sampling quarter.   Each sampling point is therefore  used to represent
hundreds if not thousands of miles of roadways.  This limitation prevents emission inventory
developers from  obtaining  a statistically  valid  number of  silt  samples  for the  roadways
represented. Moreover,  because  of traffic  congestion and  safety  concerns,  department of
transportation officials may not allow any sampling on some roadway classes such as freeways
and major arterials. As  a result, the silt loading data is always suspect for any paved road dust
emissions estimate.

The inherent limitation on the feasible amount of silt sampling makes it impossible to accurately
estimate future emissions from projected  growth in vehicle miles traveled. This arises because
sufficient silt loading data is not available to develop separate emissions rates for built-out areas
and developing areas.  Therefore, emissions for all future increases in vehicle miles traveled must
be estimated using current emissions rates. This  straight-line projection for future paved road
dust emissions  is  at variance  with observed real  world  conditions  and  can  doom  any
transportation conformity finding for an airshed experiencing substantial growth.

In summary,  the limitations  of the arithmetically derived  AP-42  paved road dust emissions
equation  combined with the infeasibility  of collecting sufficient silt loading data to accurately
represent all classes and subclasses of roadways  make all  current paved road dust emissions
inventories highly  suspect.   The  increased  traffic  congestion  and  personal  safety  issues
associated with developing better silt loading  data further reduce the  utility  of the current road
dust emission estimating methodology.   Finally,  the challenges  related  to  the  successful
maintenance of conformity make it imperative that an alternative approach to measuring  and
estimating paved road dust emissions be developed.

2.2 Clark County Background with AP-42

The Las Vegas Valley in Clark County, Nevada, is classified as serious nonattainment for federal
fine paniculate matter (PMio) National Ambient Air Quality Standards (NAAQS).  Clark County
submitted a PMio State  Implementation Plan (SIP) for this nonattainment area in June of 2001.
As part of the SIP development,  Clark County contracted with a consultant to collect 24 silt
samples representative of Clark County  roadways for estimating PMio paved road  emissions.
The silt measurements  were significantly higher  than EPA default values,  and public  works
officials  from four  agencies and  other stakeholders asserted that the Clark  County  SIP
overestimated PMio emissions from paved roadways.   Clark County committed to conducting
quarterly silt sampling through the end of 2006 as  part of the now federally approved PMio SIP.
Sampling  is ongoing and the current data base  includes sampling  from the spring of 2000
through the spring of 2006. The PMio SIP also contained a research commitment to explore the
feasibility of vehicle-based mobile sampling systems for development of improved paved road
emissions inventories.
During this timeframe, Clark County has seen substantially improved air quality for the
pollutant,  particularly from the year 2004 forward.  Visually, it also appears that Las Vegas
Valley roads have become cleaner, in part due to tightened controls on construction site track-out
and an increased emphasis on enforcement, implemented in early  2003.  However,  statistical

-------
analysis performed by UNLV under contract has generally not shown statistically  significant
declines in paved road emission factors during the 1999  through 2006 timeframe using silt
sample data and  AP-42 emission estimation methods.   These results have  reinforced Clark
County's belief that the paved road emissions inventory developed using AP-42 methods for the
PMio SIP overestimates actual emissions.  In addition, silt  measurements are time consuming,
expensive, and frequently require the alteration of roadway traffic patterns while samples are
being procured.

Initial work utilizing vehicle-based mobile sampling systems in Clark County  occurred in 1999
as part of PMio SIP development.   The test results showed even higher emission  rates than
corresponding AP-42 calculations and were not considered realistic.   In  addition, the need to
complete an approvable PMio SIP was urgent and EPA approval of this new method was very
unlikely based on work completed at that time. Clark County DAQEM submitted the SIP using
the current AP-42 methodology, and initiated a research effort to develop better methods to
characterize paved road PMio emissions.  Phase I of the current research effort was initiated in
2004 and Phase II was completed in early 2005.  Fieldwork for Phase III  occurred in late 2005
with augmentation work occurring in early 2006.

2.3  Paved Road Phase I-Phase III

The  Phase I  study entailed a two-day field study utilizing a 107-mile sampling route.  The
purpose of the study was to determine the feasibility of vehicle-based mobile sampling system
for use in Clark County to better characterize paved-road  emissions and to develop real-time
emissions of PMio for emissions inventory use.  The sampling route was designed  to include
worst-case silt-impacted roads and best-case clean roads in order to evaluate the detection limits
of the two systems.  The route was further designed to include all political jurisdictions in the
Las Vegas Valley.  Several deviations from the original sampling route were required due to road
closures resulting  from road construction.   An effort was made to note road infrastructure
conditions and deposition sources during sampling using notepads and "wrist watch time."  A
total of sixteen AP-42 silt samples were  also collected  on the sampling  route.    Phase I
demonstrated the feasibility of using vehicle-based mobile sampling systems as an alternative to
conventional AP-42 paved-road emissions estimating methods.

The  Phase II study entailed four days of sampling on  a 103-mile sampling route. The Phase II
sampling route was designed to  include a number of parameters.  The route included the five
classes of roadways (local, collector,  minor arterial, major arterial,  and freeway) and four
political jurisdictions in the Las Vegas Valley. Consideration was given to development patterns
in the Las  Vegas Valley and  the final  sampling route  included developing  areas,  older
established neighborhoods, and newer planned communities  that were completely built-out. The
developing areas included a cross section of incomplete road infrastructure (e.g. unpaved road
shoulders) and deposition sources such as vacant lots and construction activities.  The built-out
areas included  completed  road  infrastructure,  with few vacant  lots,  and little  construction
activity.   The final route  also included  a cross section of  soil  classifications based on Clark

-------
County's Particulate Emission Potential (PEP) soil  classification system .  The sampling route
included ten historical AP-42 sampling sites and eleven new sites that had not previously been
sampled using AP-42 methodology.  Relative humidity was measured during sampling at each
AP-42 site.  Specific road conditions and sources were not mapped or recorded during the study.
The study was delayed for two weeks due to rain.  The sampling route is shown in Figure 2-1.
Staff  from Maricopa County, U.S.  EPA Region 9 and U.S. EPA observed the field study.
Limited notes on road infrastructure  and silt deposition sources were made during development
of the sampling route.

            Figure 2-1.  Map of Clark County 2/14/05 - 2/17/05 sampling route.
The Phase III study utilized only the  SCAMPER and AP-42 emissions estimates. This study
focused on development of specific  emission  factors for specific conditions and to assess
measurement variability. A comparison of SCAMPER data to AP-42 emissions estimates was a
second component of this study. To accomplish these objectives, the study occurred over seven
consecutive days and utilized three sampling routes. Road infrastructure, adjacent land use (e.g.
2 Geotechnical and Environmental Services, Inc., Presentation of Final Versions of Deliverables for Re-Evaluating
and Updating the Particulate Emission Potential Map and Soil Classification for Dust Mitigation Best Management
Practices Manual for Clark County, dated September 26, 2003.

-------
vacant land, residential, etc) and sources of deposition were comprehensively mapped prior to
the study.  In order to better evaluate site conditions during the  study, a video camera was
mounted externally on the driver's  side of the vehicle.  The video camera was linked to the
SCAMPER GPS clock and camera sound was wired to a microphone located inside the vehicle
to permit the operators to record comments and observations while operating the system.

The first sampling route (industrial route) was dominated by industrial haul roads with heavy silt
loadings and was used to  determine the precision of the SCAMPER unit.  This route included
local, collector and arterial roads.  This route was sampled for most of day one of the study.
The second route (transitional route) was a 7.3-mile track in a transitional area in the Las Vegas
Valley.  Development in the area is a mix of commercial, residential, rural residential and vacant
land.  Paved roads range from fully improved with sidewalks, curbs and gutters to unimproved
with unpaved shoulders on both sides.   Sources of deposition included road construction,
residential construction, vacant land used for storing fill soil, and vacant land with no active use.
The area also has some of the highest PEP (Paniculate Emission Potential) soils in the Las Vegas
Valley.  The transitional sampling area route was sampled for four consecutive days, including
the weekend.  This allowed a comparison of weekday and weekend paved road  emission rates.
The third route (developed community route) consisted of a 12.6-mile track traversing a newly
developed  planned  community and contained  local, collector and  arterial  roads.  This route
contained fully developed road infrastructure that was not impacted by any observable sources of
silt deposition. The route included  local, collector,  and arterial streets, all of which contained
very light silt loadings.  In addition to providing baseline measurements for fully developed
roadways with minimal silt deposition sources,  this route was used to evaluate the sensitivity of
the SCAMPER unit. Measurements were taken on this route for two full days. Relative humidity
was measured  during sampling at each AP-42 site and at a nearby DAQEM monitoring site. The
study was coordinated with the cities of Las Vegas and North Las Vegas to insure that none of
the streets were swept within three days prior to sampling.
3.0 METHODOLOGY

3.1 Experimental Design

3.1.1  Route Selection

Based on experience with previous studies and the sampling characteristics of the SCAMPER
and TRAKER systems, DAQEM developed the following criteria selection of a study site:

    1.  The  micro scale  prevailing wind direction must be  roughly perpendicular  to the
       road direction at the study site.

   2.  The  study  site   cannot  have  trees,  buildings,  or  other  obstructions   in  close
       proximity to the roadway.

   3.  The   study  site  must  not   have   significantly  elevated  topography  in  close
       proximately to the roadway on either side.

-------
   4.  The study site must have a four-lane road divided by a median and the traffic conditions
       must  make it  feasible  to  block  off two  of the  lanes on one  side  of  the
       median during the study.

   5.  The study site must be located  where there are no significant  sources  of PMio that
       may cause elevated PMio concentrations at the site during the study.

   6.  The study site must have  an uninterrupted travel distance  of at least %  of a mile.
       Meteorological data from various sources was consulted to establish the road directional
       parameters for candidate sites. The requirements for no wind obstructions and particulate
       sources generally limited candidate sites to somewhat rural  areas,  whereas a majority of
       the roads in these areas did not meet the four lane and median  separation  criteria.  Where
       all road  and wind direction criteria  were met,  traffic volumes generally precluded
       blocking  two travel  lanes.  After evaluating  all available  sites  in Clark County,  the
       Veterans  Memorial Highway site in the City of Boulder City, Nevada, was the only  site
       found that met all of the study criteria.

The study was conducted in the City of Boulder  City, Nevada, on Veterans Memorial Highway,
immediately west of Buchanan Boulevard. The  sampling area consisted of two  lanes of a four-
lane    divided   highway   with   curbed   median   and  curbed  roadsides.     Details
are shown in the  study plot plans and are also described below:

   1.  During the five study days, all road traffic was diverted to the southeast lanes, allowing
       the two northwest lanes and the stabilized curbed median area to be utilized exclusively
       for the five-day study. This allowed us to limit vehicle passes next to the external tower
       sampler to SCAMPER and TRAKER vehicles. These controlled traffic and measurement
       parameters enhanced the quality of the external source emissions measurements
       compared to previous paved road dust studies.

   2.  The tower sampling array was located either on the median  or on the sidewalk areas  and
       was moved to achieve optimal orientation with the prevailing winds and sampling lane.
       Relocation of the tower position was logged throughout the study.

As shown in Figure 3-1, the course ran in a northwesterly direction  approximately 4551' from
the intersection of Buchanan and Veterans Memorial  Hwy in the northwest-bound travel lanes.
The 4551' course was divided into sections for testing purposes. The sections are described as
follows:

       Entire Length of Study Area: 4551'

       Acceleration Zone (Southern End of Course):  662'

       Deceleration Zone (Northern End of Course):  1119'

       Profiling  Zone/Sampling Zone: 2206'

-------
       AP-42 Sampling Zones:  1 IS'(south) located after acceleration zone and 120' (north)

       before deceleration zone at each end of the profiling-sampling zone, for a total of 238

       feet.


         Figure 3-1.  Phase IV Route Map (Veterans Memorial Blvd, Boulder City)

                                                                 Legend

                                                                    | AP42 Sampling Zone
                                                                     | Silt Buffer Zone Day2-5
                                                                 |    | Profiling Zone Day2-5
                                                                    	| Acceleration Zone
                                                                 |	| Deceleration Zone Day2-5
                                                                  _1 Approximate Tower Locations Day2-5
                                                                   <  42" Cone
                                                                   '  Light Poles
                                                                 l^^^ij Drainage Channel
                                                                     Curb
Deceleration Zone
1119 Feet
                 AP42 Sampling Zone North
                 120 Feet
                                                                     Roadway

                                                                     Lane Lines

                                                                     Sidewalks
       200 TOO  0
       Note: The silt zone, with total length 2770 feet, includes Silt Buffer Zones, AP42 Sampling Zones, and Profiling Zone.
3.2  Soil Selection and Application

3.2.1 Soil Sampling Site Selection

The 50th percentile silt content for collector roadways sampled in Clark County in 2005 and 2006
was used as a target value for silt content for selection of a candidate soil to be applied to the
road surface for the Phase IV controlled study. Data summarizing the 50*  percentile calculations
are shown in Table 3-1. The 50*  percentile silt content value for collector roads was 13%.

UNLV, in collaboration with Clark County DAQEM staff, surveyed four candidate field sites in
southern  metropolitan Clark County in July of 2006.  Three candidate sites, in southwest Las
Vegas,  were not  selected because either the  silt content  was incorrect,  or because permission
could  not be  obtained  from  either the  US Bureau  of Land Management or from  private
landowners for large-scale excavation.
                                                                                             10

-------
   Table 3-1. Summary of Data Used to Determine 50  Percentile Silt Loading Value for
                                 Collector Roadways
QTR-Year
3™Q-2005
4tnQ-2005
1SIQ-2006
3™Q-2005
4tnQ-2005
1SIQ-2006
3™Q-2005
4tnQ-2005
2naQ-2005
3raQ-2005
2naQ-2005
4tnQ-2005
1stQ-2006
2naQ-2005
3raQ-2005
geometric
mean
10th
percentile
50th
percentile
90th
percentile
UNLV
Site
24
24
23
23
23
15
15
15
5
5
2
2
1
1
1




Site
Modifier



















DAQEM
location name
Pabco
Pabco
Burkholder
Burkholder
Burkholder
lone
lone
lone
Washburn
Washburn
Marion
Marion
Gowan
Gowan
Gowan




DAQEM
Roadway
Classification
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector
Collector




Plot
Number


4


4






4






Percent
Gravel
25.4
17
11
20.1
7
16
13.3
5
15.6
2.1
14.3
6
8
24.9
17
11.4
5.4
14.3
23.0
Percent
Sand
68.9
76
77
75.6
80
70
75.7
83
55
7.6
49.2
78
79
61.8
78.5
67.4
51.5
75.7
79.6
Percent Silt
SClay
5.7
7
12
4.3
13
14
11
12
29.4
90.3
36.5
16
13
13.3
4.5
13.1
5.0
13.0
33.7
* Gravel-sand boundary was 2.00 mm
* Sand-silt boundary was 75 microns

A 21.9 kilogram sample of soil from a site located at Sunset Park, designated UNLV Road Dust
site 29 (wet sieve) or 32  (dry sieve),  in Wind Erodibility  Group (WEG) 2, at an elevation of
1,988 feet, latitude N36° 3.792', longitude W115° 6.748' (Garmin eTrex®, WGS 84 datum) was
collected on August 4, 2006. A 675  gram sample was sieved on August 11, 2006 and was found
to be predominantly sand,  with a 14% silt content.

A  second  group of samples were  collected from (60 meters)  200 feet west of the original
sampling site on August 23, 2006, designated as UNLV sites 38 and 39, at latitude N36° 03.777'
and longitude W115° 06.824'.  Volumetric soil moistures were found to range from 0.0% to
0.5%. Results of sieve analyses for  silt content were similar to the first sample, and the decision
was made  to use this sandy WEG 2 deposit as the source material for the Phase IV controlled
study.

3.2.2  Soil  Excavation and Packaging

On Wednesday, September 6, 2006, a team of Clark  County DAQEM and UNLV personnel,
assisted by staff from Clark County Department of Parks and Recreation, excavated soil from the
Sunset Park site.  The excavation location was at latitude N36° 3.782' and longitude  W115°
6.770', a location in between the two original soil collection sites.
                                                                                    11

-------
A 0.38 cubic meter (0.50 cubic yard) bucket loader was used to remove soil from the site and
deposit it in a loose pile. Soil was excavated to a depth of about 0.40 meters (18 inches). Round-
end hand shovels were used to excavate soil from the pile and pour it through 30.1-centimeter
(12 inch) diameter 1 mm sieves placed on top of tared plastic 19-liter (5-gallon) paint buckets.
Three sets of 1 mm sieves and buckets were used in parallel  to speed the bulk sieving process.
The sieves and buckets were vigorously rocked from side to side to agitate fine soils through the
sieve opening. Loose conglomerates of soil remaining on top of the sieves were hand-crushed to
pass them through the sieves.  Rocks,  twigs, and other debris were  shaken off the sieves and
placed in a spoils pile at one side of the excavation site.

Tared and total bucket weights with soil were recorded on a calibrated Sunbeam Freightmaster®
150 scale to the nearest 0.1 kilogram and were logged into a bound laboratory notebook.

After total  (tare + soil) bucket weight was calculated,  each  bucket  was covered with a tight-
fitting snap-down lid and moved to the bed of a pickup truck for transport to the Phase IV study
site.

Fifty (50) covered buckets of sieved  soil were prepared in  this manner.  They were then all
simultaneously transported to the storage yard of the DRI Solar facility on Adams Boulevard in
Boulder City, Nevada, and stored  outside for four days until September 11, 2006, when the soil
samples were applied to the Phase  IV road site.

3.2.3  Soil Characterization

A soil sample with a mass of about 700 grams was extracted from each of six soil buckets with a
trowel during the excavation process, sealed in plastic cash bags,  and transported to Ninyo and
Moore, the geotechnical company  contracted to perform  soils analysis, on September 6, 2007 for
sieve analyses.  Sampled soil masses were measured with a  calibrated Sunbeam model  78411
postal scale. Every tenth bucket, corresponding to Bucket numbers 1,  11, 13, 17, 28 and 39, was
sampled for soil (buckets were not filled in numerical order).  Soil moistures were measured with
a Dynamax HH2 TDR volumetric moisture meter. Values ranged from 1.9 to 4.1 volume%.

Ninyo and Moore sieved these samples, using a sieve  stack consisting of number 16 (1.18 mm),
30 (0.600 mm), 50 (0.300 mm), 100 (0.150 mm) and 200 (0.075 mm) mesh sieves, and an eight-
minute shake time, to determine silt contents.  This non-AP-42 sieving technique was used only
for recovered  field soil samples that were collected before  the Phase IV AP-42 field  study.
Results using this method showed that the average silt fraction for the excavated soil was 14.3%.

3.2.4  Soil Application

Soil from 15 buckets (about 340 kilograms,  or 750  pounds) was poured into  a  12-foot wide
Gandy 10T series fertilizer  drop  spreader at the Phase IV  empirical study field site  on the
morning of 9/11/2006.

The Gandy spreader was then driven to the Veterans Memorial Boulevard (VMB) site.
                                                                                     12

-------
Prior to the first application of soil a group of preliminary measurements by the mobile
sampling vehicles were used to characterize the PMio emission rates of the natural road soil on
the VMB site before and after two  sweeper passes. Soil was first applied from the Gandy
spreader at about 1120 in the morning of 9/11/2006 after 92 vehicle passes had been completed.

During the five days of the study, the spreader pull speed was kept constant at approximately 5
meters/second (16 kilometer/hour or 10 miles  per hour) over an 844.3 meter length of the course
(2770 feet). The spreader was  pulled by a Dodge MaxiVan on the first day while a large garden
tractor was used on subsequent days. Spreader soil application was driven by geared wheel that
turned an agitating feeder at a rate that is proportional  to ground speed. The rate of application by
the spreader is controlled by adjusting the size of the  diamond-shaped openings that feed soil to
the ground surface. The opening was held constant  for  each set.  Opening size was varied for
different sets to apply soil at different loadings to the test site.

Soil was applied from 27 meters (87 feet) before the start of the southern AP42 sampling zone to
72.8 meters (239 feet) after the end of the northern AP42 sampling zone.

3.3 Horizontal Flux Tower

The flux  of  PM  downwind  of the  test  roadway  emissions  was quantified using  a  flux
measurement  technique similar to  that described in previous work by Gillies  et al (2005).  A
"master" tower was erected downwind of the road (Between 4 and 6 m from centerline of test
vehicle travel path) and aligned perpendicular to road (Figure  3-2, Figure 3-3).  The trailer-
mounted, 9 m-high tower was instrumented with DustTraks (Model 8520, TSI Inc., Shoreview
MN) configured to measure PMio at five heights above the ground surface (0.7, 2.1, 3.4, 6.4, and
9.8 m).  At one of the heights (3.4 m), a DustTrak equipped with a PM2.s  impactor inlet was
collocated with the PMio DustTrak.  The master tower also included a TEOM (R&P, Model
1400a),  which measures PMio at a height of 2.3 m.  The TEOM sampling inlet was nominally
collocated with one of the PMio inlet-equipped DustTrak monitors (at 2.1 m above ground level).

The DustTrak monitor measurement is based  on light scattering  of particles which  is dependent
on the particle  size-distribution and the  optical properties of the  emissions. The TEOM was
intended to help account for  differences between optical based measurement and mass based
measurements.  These data were  used  to confirm supplemental,  controlled  measurements
conducted in  a resuspension  chamber and described below.  This allowed for conversion of
emission factors measured with the tower-mounted DustTraks into mass-based emissions factors
(see Section 4.1). A wind vane was mounted at the top of the tower and one cup anemometer was
approximately collocated with each pair of DustTrak  samplers.  All data from the PM samplers
and meteorological instruments were telemetered and logged in 1-second intervals by a laptop
located on the master tower.
                                                                                     13

-------
Figure 3-2. Photograph of Master (left) and Satellite (right, not used in present study)
     Towers Showing Locations of DustTrak PMio Monitors. For present study,
 only one PMi.s inlet-equipped DustTrak was used on the master tower at a height of
                           3.4 m above ground level.
                                                                              14

-------
Figure 3-3.  Schematic of Field Sampling Layout. The gray star shows the location of the
 master tower on 9/11/06 and the black star shows the location of the master tower from
                                   9/12/06 - 9/15/06.
                     Section of Veterans Memorial Blvd, Boulder City
Legend

   Lislil I'oJe

   Drainage

   Cutter
                              a. View of entire test section
                     b. Close-up of section where tower was located
                                                                                     15

-------
3.4 EPA Method AP-42

3.4.1  Plot Layout

Two zones of the course, called "south" and "north" were designated for silt recovery during
controlled study.

The "near" end of the south AP-42 sampling zone was established 228.3 meters (749 feet) from
the start of the course, as defined by  the intersection of Veterans Memorial Dr.  and Adams
Drive.  This distance was selected so that the mobile technologies vehicles could complete the
acceleration portion of their  pass before entering the  soil sampling zone.  The south AP-42
sampling zone was 36  meters (118 feet) long. The "far" end of the north AP-42 sampling zone
was 36.6 meters (120  feet) established 1358 feet (Deceleration Zone and "North" Silt Buffer
Zone) from the end of the course, just before the gradual curve in the roadway. GPS coordinates
of the "near"  and "far" corners of the sampling zones were measured  using an un-corrected
Garmin E-trex Global Positioning  System receiver.  Distances  were  also measured with a
measuring wheel.

Seven 3.3  meter long x 4.1 meter wide  (10 foot long x 13.5 foot-wide) plots were laid out in the
south and north zones for soil  recovery (Figure 3-4).  Each of the AP-42  sampling  plots was
separated  by a 2.4  meter (eight-foot)  buffer zone. The  buffer zone was used to  allow field
personnel and equipment to access the plots without disturbing the sampled area.

Two different plot layouts were used during the empirical study to collect soil samples:

(1) A full size  3.3 meter long x 4.1 meter wide plot, with an area of 12.5 square meters was used
to estimate soil and silt loading at the beginning and end of most of the mobile technologies
sampling experiments.  A 3.3 meter (10 foot) plot length was selected to remain consistent with
recommended  clean road plot length  on page 7 of Appendix C.I, Procedures for Bulk Sampling
of Surface Loading (US EPA 1993a) A 4.1  meter (13.5 foot) width was selected to  recover soil
from the edge of the asphalt (at the start of the concrete gutter) to the line dividing the eastern
and western northwest-bound travel lanes on Veterans Memorial Boulevard.

An array of seven (7) numbered full-size plots, with 2.4 meter (8-foot) spacing between the plots
was laid out at each end (zone) of the driving course. Layout was established by first setting up a
string rectangle consisting of colored surveyor's twine wrapped around gravel-filled  cans. The
3.3 meter and 4.1 meter lengths were different colors, and were tied to form a rectangle with an
uncertainty of +/- 0.05 meters.  White  surveyors paint  was used to establish the corners of the
rectangles.  The surveyors' twines were pulled tight around the gravel-filled cans, and then 5.1
cm (2-inch) masking tape was applied from a roller dispenser to match the perimeter established
by the colored surveyors' twine.

(2) For experiments evaluating the effects of vehicle passes on applied soil depletion, 0.61 meter
(2 foot) wide "Quickie-Strips" (Etyemezian, personal communication, 2006) were laid out in the
zones between the full-size plots.  Quickie-strip locations were marked  on the concrete gutter
and on the lane dividing line with white spray painted dots  spaced every 2 feet apart. Painted
                                                                                      16

-------
lines or masking tape were not used to indicate boundaries of the Quickie Strips. Quickie Strip
samples were also collected inside unused full-size plots, when needed. Although sampled in the
"buffer" zones between AP-42 plots, the Quickie-strip samples were not collected in areas where
there had been foot traffic, as the seven plots and, when needed, Quickie strips  in the buffer
zones were sampled in a progression from the near to far ends of the course (south zone) or far to
near ends of the course (north zone).
                                                                                      17

-------
                                Figure 3-4. Phase IV Veterans Memorial Drive Plot Layouts

(a)  Schematic south zone plot layout (not to scale). Start of course is to left of Plot  1. Plots sacrificially sampled in ascending
numerical order from 1 to 7, moving from left to right. Spaces between plots are eight-foot buffer zones for personnel and equipment
access. Shaded plots indicate already sampled.
(b) Schematic north zone plot layout (not to scale). End of course is to right of Plot 1. Plots sacrificially sampled in ascending
numerical order from 1 to 7, moving right to left. Spaces between plots are eight-foot buffer zones for personnel and equipment access
(c) Example south zone quickie-strip plot layout (not to scale). Dotted lines show partitioning of un-used buffer zones or un-used plots
into Quickie Strips for silt depletion sampling

  1
                                                                                                                        18

-------
3.4.2  Vacuum Soil Recovery Methods

One Hoover Model S3636 Wind Tunnel Plus®  and two Hoover Model  S3639 Wind Tunnel
Plus® canister vacuum cleaners, rated at 12 amperes, were used to recover applied soil from the
roadway sites. The vacuum cleaners were connected by 50-foot or 100-foot 14-gauge extension
cords  to portable 3750-watt 120-volt Coleman generators.  In cases where  two samples were
required at one point in time, two vacuum cleaners were simultaneously operated in parallel at
the south zone of the site, and the northern vacuum cleaner would  sample two test plots in
sequence.

Soil samples were recovered into pre-tared (to +/-  1 gram using the Sunbeam 78411 postal scale)
Hoover Type S Allergen Canister vacuum bags, model 4010100S.  To determine the tare mass of
the bags, the empty Hoover bags were removed from their plastic liner bags, weighed in the
laboratory to within +/- 1 gram, labeled with a bag number and a tare mass, and replaced back in
their plastic bags for interim storage until used in the field.

Vacuum hose-to-bag connections were  sealed with low-density, high compression white foam
polyethylene weather-stripping to minimize leakage of collected sample. New  secondary motor
filters were installed at the start of the study. They were cleaned every morning by removing and
knocking the dust off. They were replaced every two days at a point when  knocking the filter
could not remove visible discoloration from soil.

Hoover Hard Floor Tools were used for soil recovery.  Brushes on  the Hard Floor tools are
known to wear out quickly on asphalt. The most rapid wear occurred on the brush closest to the
wand  connection, with this brush worn down from about 9 mm to about 3 mm after 1/2 day's use
in the field. Floor tools were replaced when visible wear of the brush below 3 mm was observed,
typically every 1/2 day.

For full-size  (12.5 square meters,  135  square feet) plots, two sets of twine  wrapped around
gravel-filled soup cans were used to visually partition the full-size plot into thirds across the
direction of travel. Each  partition was  vacuumed twice with a curb-to-gutter vacuum stroke.
After  the curb-to-gutter vacuum strokes had been completed, the  twine dividers were realigned
along the direction of travel. Each partition was vacuumed twice with a front-to-back vacuum
stroke. A total  of four vacuum strokes were passed  over each portion of the vacuumed  plot,
consisting of two curb-to-gutter strokes and two front-to-back strokes.  Four vacuuming passes
had been previously shown to  recover 95-98%  of applied mass on  asphalt surfaces (UNLV
unpublished data).

For Quickie-strip plots (area 2.51 square meters or 27 square feet), the hard floor tool was passed
back and forth twice over each strip (Figure 2), first on the  /^ of the plot nearest  the curb, starting
from the curb side towards the center of the road, and then on the 1A of the plot nearest the lane
divider, starting at the lane  divider and vacuuming  towards the  curb.  Quickie-strip plots,
comprised of five subsections of a standard plot, were not as well-marked as standard plots, so
side to side variations in the swept width of the Quickie-strips were larger than they were for the
                                                                                     19

-------
full-size plots. As a result, the absolute and relative uncertainty in the width of the Quickie-strip
is larger than for the full-size plot.

Three soil recovery techniques were used during the study.

(1) One plot per bag (Individual). Soil from one large heavily soiled plot would be recovered into
one pre-tared bag, the bag would be weighed, sealed with plastic film to prevent leakage, and
then placed in a labeled large brown 25 cm x  35 cm (10" x 14") office envelope. The envelope
would then be held closed with its brass clasp.  The  date and time of the collection would be
noted on the bag and on the log sheets.

(2) Two large plots per bag (Cumulative).  Soils from two lightly soiled  large plots,  sampled at
the  same time (before  or after a particular vehicle pass) would be accumulated into one tared
vacuum bag. The vacuum bag would be removed from the vacuum cleaner, weighed by one of
the portable balances after the first soil recovery, and  then reinstalled in the vacuum  cleaner for
sampling the second plot. After plot sampling was completed, the bag would be removed, sealed
with film, placed in a labeled large brown office envelope and held  in a sealed plastic storage
container until needed for silt sampling analysis by Ninyo and Moore. The following formulae
were used calculate the individual plot weights and silt loadings.

Silt mass plot 1 = (Ninyo and Moore silt fraction) x (Ninyo and Moore  silt mass) x (Bag mass
after plot 1 - Bag tare mass) / (Net mass for plot 1 + plot 2)

Silt mass plot 2 = (Ninyo and Moore silt fraction) x (Ninyo and Moore  silt mass) x (Bag mass
after plot 2 - Bag mass after plot 1) / (Net mass for plot 1 + plot 2)

(3)  Multiple small plots  per bag (Cumulative). Soil  masses from a series of Quickie strips,
sampled in sequence after a specific vehicle pass.

Filled bag masses were recorded in the field after each vacuuming using the Pelouze  SP5 and
Sunbeam 78411 field scales.  Scales were kept shaded from direct sun and measurements were
made either inside a large  plastic  storage box  or inside  a closed 12-passenger cargo van  to
minimize effects of wind shake.

3.4.3  Field Soil Application History

The native road dust on Veterans Memorial Boulevard was first sampled by the AP-42 recovery
technique before any passes were made by the mobile technologies vehicles.

Emissions from the native road soil were then measured by the mobile technologies sampling
vehicles (DRI TRAKER I,  TRAKER II, and UCR SCAMPER) and the DRI tower.   After a
series of 60 mobile technologies sampling passes, a PM-efficient sweeper was driven twice over
the  site to remove native road dust.  Another 30  sampling passes by the mobile technologies
vehicles then took place.
                                                                                     20

-------
Soil from the Gandy spreader was first applied after vehicle pass 92. Pass 93 was the first mobile
technologies measurement using the applied soil.

A summary of the applied soil loadings, vehicle passes and speeds is shown in Table 3-2.

 Table 3-2.  Summary of Applied Silt Loadings During Phase IV Controlled Field Study—
                    Veterans Memorial Boulevard. Boulder City, NV
Date
9/11/06
9/12/06
9/12/06
9/13/06
9/13/06
9/13/06
9/14/06
9/14/06
9/15/06
Set#
3
4
5
6
7
8
9,10
12
13
Nominal
Drive Speed
(mph)
35
45
25
45
25
45
35
varying
varying
Spreader
Setting
15
30
30
15
15
20
20
30
35
Netwt
Applied,
Ibs
45
117
113
34
32
52
53
120
194
Spreader
Path
Length,
ft
2977
2775
2775
2775
2775
2775
2775
2775
2775
Applied
Soil
Loading
(gram/m )
6.16
17.17
16.58
4.99
4.70
7.63
7.78
17.61
28.47
Avg.
Recovered
Silt
Loading,
(gram/m2)
0.75
2.48
3.17
0.88
0.74
1.14
0.80
2.55
2.31
3.5  Mobile Technologies

3.5.1  SCAMPER

The SCAMPER determines PM emission rates from roads by measuring the PM concentrations
in front of and behind the  vehicle using real-time sensors.  In this study, the concentration
(mg/m3)  is  found by subtracting the  background concentration (front  sampler)  from  the
concentration measured by  the  rear sampler.   As a first approximation, the concentration
difference (mg/m3) can be multiplied by the vehicle's frontal area (in this case, 3.66m2) and by a
DustTrak  calibration factor to obtain an emission factor in units of mg/m. The vehicle frontal
area is defined  as the vehicle width at the  highest part of the vehicle multiplied by the overall
height at the highest part (no correction made for ground clearance). In previous SCAMPER
studies, a  reference sampler was collocated with the rear sampler in order to  find a DustTrak
calibration factor to convert from concentration-based readings to a mass-based emission factor.

This SCAMPER includes five major components:

       (1) PMio Sensors

          Thermo  Systems  Inc. (TSI Incorporated) Model 8520 DustTrak optical  PM sensors
          with PMio inlets are used. These  sensors are based on the principle that the amount of
          light scattered by particles is related to the particle concentration.  Since the efficiency
          of light scattering depends on particle  size, the response  of the sensor depends on the
          particle-size  distribution. Particles less than approximately 0.1 um  diameter are  not
          detected. The instruments are calibrated at the  factory using NIST reference material
          8632  Ultrafine Test Dust, more commonly know  as  "Arizona Road  Dust".  The
                                                                                     21

-------
  manufacturer recommended measurement range is from 0.001 to 100 mg/m3, although
  the instrument will generate readings up to 150 mg/m3 with less reliability.  The time
  constants are selectable from 1-60 seconds; the 1-second time constant is used on the
  SCAMPER. An impactor supplied with the instrument is used as a PMio size-selective
  inlet.

(2) Sampling Inlet

  An inlet for the real-time PM sensors was used that allowed sampling as isokinetically
  as possible over the full range of vehicle speeds. Figure 3-5 shows the design of the
  inlet. Stainless  steel tubing (1A inch OD, 3/16 inch ID) is used to connect the sample
  inlet to  the one  end of a hollow cylindrical pleated paper  filter element (1.7 cm
  diameter, 5.0 cm  long) and from the other end to the DustTrak (the sampled air is not
  filtered,  but travels from one end of the  hollow cylinder to the other). The stainless
  steel tubing is attached to the metal end caps  of the filter element using "JB Weld".
  The filter element and attached tubing are contained in  a 1  inch PVC pipe with a cap at
  each end with a PVC "T" in between. Each cap is drilled and tapped for a 1A inch pipe
  fitting. A Swagelock® male connector ( Vi inch pipe  x Vi inch tubing) that has  been
  drilled through  with a 1A inch drill is screwed into each end cap. An end cap  assembly
  is slid  over each  piece  of stainless  steel  tubing and onto  the  PVC pipe.  The
  Swagelock® tubing fittings are then tightened to seal the tubing within the PVC pipe
  assembly. The overall length of the PVC pipe section is 33 cm.

  To slow the flow to the sample flow rate of the DustTrak without creating a virtual
  impactor, excess  air is pulled through the  outside of the cylindrical filter through the
  arm of the PVC "T" with a vacuum pump that maintains the bulk air speed at the inlet
  equal to the speed of the air going past the inlet. The flow rate of the vacuum pump is
  adjusted by the data logging PC to produce a reading of zero pressure on the gauge.
  When the pressure equals zero, there is no pressure drop from the probe inlet to the
  tubing that leads to the DustTrak.  This condition creates a  no-pressure-drop inlet;
  therefore, the sampled air stream has the same energy as the ambient air stream.
                                                                              22

-------
                    Figure 3-5. Isokinetic Inlet Schematic Diagram
        % inch OD
        metal tubin
                                      Filter
             ,% inch OD
PVC Pipe   /metal tubing
To DustTrak
                                                                      Sample Inlet
                                                                       Static Pitot
                                                                       Tube
                                    To Vacuum Pump
                  Detail of Flow Splitting Section
       1.7 L/min
                  tiff +  f  f f f
                                                              ;
                                                            3'
                    1.7+x
      (3) Sampling Trailer

        To determine PMio concentrations in the vehicle wake, a DustTrak was mounted on a
        small trailer.  The  trailer  has a flat  bed four  feet wide and  six  feet  long,  this
        configuration  chosen such that the vehicle wake would be  disturbed as  little as
        possible. In addition, the trailer holds the bypass flow system. The trailer has a three
                                                                                      23

-------
         foot extension on the hitch to place the DustTrak in a position ten feet behind the
         vehicle, which was shown to be representative of the PMio concentrations in the wake
         and yet be safe to operate on public roads (Fitz, 2001).

       (4) Position Determination

         A Garmin GPS Map76  global positioning system was used to determine vehicle
         location and speed.

       (5) Data Collection

         A PC was used to collect data from GPS and PMio measuring devices. Data was stored
         as one-second averages. The PC also was used to automatically adjust the sample inlet
         bypass flow to  maintain isokinetic particle  sampling using a  10-second running
         average of vehicle speed based on the GPS.

Figure 3-6 shows front and rear photographs of the SCAMPER.  The tow vehicle is a 2006 Ford
Expedition with a custom trailer using an extended hitch.
                                                                                     24

-------
Figure 3-6. Photographs of the Front and Rear of the SCAMPER
                                                                  25

-------
3.5.2  TRAKERI

The principle behind the TRAKER system is illustrated in Figure 3-7.  The concentration of
airborne particles is monitored through inlets that are mounted near the front tires of a vehicle.
These particle sensors are influenced by the road dust generated through the tire contacting the
road surface.  A background measurement  of particle concentrations is obtained simultaneously
at a location on the vehicle farther away from the tires. The difference in the signals between the
influence monitors  and the background monitor is related to the amount of road dust generated:
                                       T - T -
                                       L  - L
Equation 3.1
where T is the "raw" TRAKER signal, TT is the particle concentration measured behind the tire
(average of left and right), and TB is the background concentration.

 Figure 3-7. TRAKER Influence Monitors Measure the Concentration of Particles Behind
             the Tires. A background monitor is used to establish a baseline.
TRAKER I is comprised of a van that has been equipped with three exterior steel pipes acting as
inlets for the onboard instruments (Figure 3-8a).  Two of the  pipes are located behind the left
and right front tires and are used to measure emissions from the tires. The third pipe runs along
the centerline of the van underneath the body and extends through the front bumper.  This pipe is
the inlet for background air.  Dust and exhaust emissions from other vehicles on the road can
cause fluctuations in the particle concentration  above  the road  surface.   The  background
measurement is used to correct the measurements behind the tires for those fluctuations.

The three exterior pipes  enter the cargo compartment of the van through the underbody. Each
pipe then goes into a plenum/manifold; the plenum can be used to distribute the sample air to up
to five instruments (Figure 3-8c).  For the present study, one TSI DustTrak with PMio inlet was
operated  at each of the left and right inlet lines as well as on  the middle inlet line. A central
computer collected all the data generated by the onboard DTs as well as  GPS coordinates, speed,
and acceleration with 1-second frequency (Figure 3-8d).

All DustTrak monitors used for the study were calibrated by the manufacturer within 12 months
of their use. Prior to each day of measurement, flows on the DustTraks were checked to ensure
                                                                                     26

-------
they were within manufacturer specifications and the instruments were "zeroed" with an inline
HEPA filter as specified by the manufacturer.
3.5.3  Inl et confi gurati on

Unlike gases, particles have inertia; as a result, the sampling of particles through an inlet results
in some particle losses to inlet surfaces.  These losses could be due to the diffusion of particles
toward inlet walls or the impaction/settling of particles upon inlet walls.   Diffusion is  a
phenomenon that governs the motion of very small particles (less than 0.1 |j,m).  Since road dust
is composed primarily of larger particles (greater than 0.3 |j,m), diffusion is not an important
consideration for  TRAKER.   Impaction and  gravitational  settling, however, are  important
processes for sampling particles with aerodynamic diameters greater than 1 |j,m.  Gravitational
settling can be minimized by reducing the amount of time a particle spends in the inlet lines (e.g.,
by increasing the speed of the flow).  On the other hand, particle impaction can be minimized by
reducing the speed of the flow turns within the inlet lines.

The inlet lines, visible in Figure 3-8a, are 19 mm (3/4") in diameter and 2.3 m (7.5') long for the
tire lines and 3.7 m (12') long for the background line. The influence inlets on the right and left
are in slightly different positions with respect to the tires.  On  the right, the inlet is 165 mm
(6.5") above the ground, 50 mm (2") behind the tire, and 63 mm (2.5") in (toward the center of
the vehicle) from the outside edge of the tire.  On the left, the inlet is 165 mm (6.5") above the
ground, 63 mm  (2.5") behind the tire,  and 63 mm (2.5")  in  from the outside edge of the tire.
Because  of the  vehicle's  configuration,  it is not possible  to avoid bends in  the inlet lines.
However, the bends have been kept as shallow  as possible in order to minimize  losses of
particles to the inlet walls.  Each of the inlet lines feeds into a 600 mm (20") long torpedo-shaped
plenum (Figure 3-8c).  All particle sampling  instruments are connected through the plenum via
short non-conductive tubes that are in turn attached to 20 mm (8") long steel tubes that extend
into the body of the plenum. Flowrates through the inlets, developed with a high vacuum pump,
are 75 liters per minute (1pm), corresponding to an inlet face velocity of 4 meters per s (mps) and
0.3 mps in the plenum.  Rotameters connected to each of the inlet lines are used to ensure that
the flows through the inlets remain within 10% of the desired value.  An independent rotameter
equipped with stopper is used at the inlet lines to verify the readings of the onboard rotameters.
Noting that in the seven years of experience using TRAKER I, the flowrate through the inlets has
never drifted by more than a few percent of the  desired  value  over the  course of a day, the
operator  of the  TRAKER can  periodically  check flows by examining  the readouts  on the
rotameters in the vehicle's rear-view mirror.
                                                                                      27

-------
 Figure 3-8.  TRAKER Vehicle and Instrumentation: (a) Location of inlets (right side and
 background shown); (b) Generator and pumps mounted on a platform on the back of the
van; (c) Two sampling plenums (bottom), a suite of DustTrak particle monitors (top right),
 and three rotameters used for ensuring proper flows through the two plenums; and (d) a
 dashboard-mounted computer screen used to view the data stream and a GPS to log the
                        TRAKER's position every 1 second.
                    c.
                                                           b.
                                                           d.
                                                                               28

-------
3.5.4  TRAKERII

In addition to the TRAKER I test vehicle described above, DRI also employed a prototype of a
modified unit (TRAKER II). There are two major design differences between TRAKER I and
TRAKER II.  First, TRAKER II (Figure 3-9 and Figure 3-10) uses low pressure-drop blowers
to pull sample air in from behind the front tires and from the background instead of the high
vacuum pump utilized by  TRAKER I.  This substantially reduces the power requirements of
TRAKER II compared to TRAKER I and allows for the modified unit to be powered by onboard
DC batteries that are recharged by the vehicle's alternator. Second, the TRAKER II inlet lines
are configured so that on unpaved roads, where PMio concentrations behind the front tires could
exceed the DustTrak instrument's upper limit (150 mg/m3), clean air can be mixed with air from
the tire inlets in a controlled manner to achieve a desired amount of dilution.

There are also other minor differences between TRAKER I and  TRAKER II.  For example, a)
the inlets behind  the front tires in  TRAKER II are located farther behind  the tire than in
TRAKER I; b) Instead of an  onboard sampling plenum as in TRAKER I,  a 10  cm diameter
external pipe is used to channel/dilute inlet flow and instruments can sample the air within that
pipe through small manifolds  located on the floor of TRAKER II; c) The circular inlets used
currently on TRAKER I are replaced by flattened manifolds on TRAKER II.  Aside from these
differences, TRAKER II is based on the same basic principle of operation as the TRAKER I.

In the present study, the use of TRAKER II is intended to obtain preliminary data for assessing if
changes in design have achieved the desired  outcome or if additional  changes are needed.  Like
TRAKER I, TRAKER II was  outfitted with  PMio DustTraks on the left and right tire inlets as
well as on the  "Background" inlet, which in  the case of TRAKER II resides above and slightly
behind the driver-side and passenger-side doors (See Figure 3-9).

The electric blowers in the inlet pipes were turned on and fixed at a flowrate of 10  1pm.  Within
each inlet line, the flow rate is measured in 200 ms intervals by a small pitot tube  attached to a
pressure transducer (Dwyer Instruments, Vi" of water max).  An onboard laptop computer adjusts
the power to the blower motor to maintain the flow at 10 1pm with a frequency of 200 ms.

As with TRAKER I,  DustTrak monitors were zero- and flow-checked at the beginning of each
sampling day.  In  operation, the DustTrak instruments extract particle-laden air from within the
pipe that runs  along the underside of the vehicle through non-conductive tubing.  Optionally,
TRAKER II can be equipped with  other instruments such as filter samplers and particle size
analyzers through additional sample ports on  the inlet pipe.  A GPS unit in TRAKER II provides
geospatial coordinates vehicle  speed, acceleration, and wheel angle.  These data, along with 1-
second DustTrak  measurements from the three inlet lines (left, right, and background) are
displayed in real-time and logged by the laptop computer for subsequent analysis.
                                                                                    29

-------
Figure 3-9. TRAKERII. Vertical inlet pipe near the passenger-side door is used to sample
                        background air for the right side inlet.
                                      a.  side view
                                   b. inlet close-up
                                                                                  30

-------
                 Figure 3-10.  Schematics and Dimensions of TRAKERII
     Real Time
     Control and
     Data
     Acquisition
     Computer
                              a. Functional TRAKER Diagram
                             b. Dimensions - Not drawn to scale
     5.7 cm
7.6 cm
       2.5 cm
      5.7 cm
                 19 cm
             9.6 cm
                         10 cm
                  1.9 cm
11.5cm
 2cm
10cm
                                                                                      1cm
                            Bern

                           TTcrn
               c. Inlet, Top View
d. Inlet, Side View

-------
4.0 QA/QC

4.1 Horizontal Flux Tower
Horizontal  fluxes of PMio (units of grams PMio per vehicle kilometer traveled - g/vkt) were
calculated using data from the master tower.  Level 0 data validation involved  ensuring that
instruments were operating properly and data were recorded correctly.  This included cross-
referencing the data recovered from computer files with dates and times of operation noted in
field notebooks. Additionally, whenever new wire connections were made or modified or any
part of the data acquisition was modified (change of communication ports on data acquisition
system,  replacement  or exchange of DustTrak monitors, etc), the data files  were spot-checked
against the instrument visual  display to  ensure that readings in the data files corresponded to
instrument labels.

Level I  validation required visual as well as automated inspection of the data.   The measured
PMio concentrations at multiple heights, wind speeds, and wind direction were plotted with one-
second resolution.  In addition, the vehicle  passage times that  were  manually noted by field
personnel and verified with GPS data onboard TRAKER I and TRAKER II were also plotted on
the same graph.

Two factors were used to determine  if a specific flux measurement associated  with a specific
vehicle  pass  was valid.   First, the one-second wind direction over the duration of the three
intervals - pre-peak background, peak, and post-peak background was examined.  In cases where
the average wind direction over the three intervals was within 45 degrees of the perpendicular
line drawn  between the tower and the road segment  and the wind speed was relatively constant
(i.e. holding  at > 1 m/s from  the same  general direction), the wind direction was considered
valid.  In cases where the  average wind direction  was outside  of this 90-degree window (45
degrees  in each direction about the perpendicular), one-second data were examined.  If the wind
direction was always less than  75 degrees from the perpendicular, the wind speed was relatively
constant, and fluctuations in wind direction did not  exceed 30 degrees, the wind direction was
considered valid. In all other cases, wind conditions were considered to invalidate the horizontal
flux measurement.

The second factor  in determining the validity of a specific tower measurement was the noise
level of  the baseline PMio concentration.  During periods  of high wind, wind-entrained dust
clouds often passed by the flux tower (especially true on 9/14/06 and 9/15/06). These high and
spurious concentrations of PMio rendered the baseline from which peak values are estimated
extremely noisy.  In  other cases, the passage of a large vehicle on the south side of Veterans
Memorial Highway would sometimes result in a temporary spurious baseline reading. The entire
time series  of data from the flux tower was examined to flag periods when the baseline was too
noisy for a measurement.  Those data were considered invalid.

Note that an individual dust plume from a moving vehicle may exhibit a high degree of spatial
heterogeneity, owing to the turbulent nature of air flow in the wake of a moving vehicle. Thus,
an actual plume consists of clouds of dust interspersed with comparatively clean background air.
This is especially true close to the road; PMio concentrations become more spatially continuous
and smooth as the plume advects and disperses downwind.  For the present study, in certain
                                                                                     32

-------
cases,  baseline noise levels  and the wind  direction over the expected peak period  were
acceptable.  However, a visible peak associated with the passage of a vehicle was not always
clearly discernible. In those cases, the measurement was considered valid and the PMio flux was
calculated and reported.  Though these cases could result in near-zero or negative fluxes, which
are not physically reasonable, it is important to retain  these measurements to avoid biasing the
data. Estimation of peak duration (whether or not peak  was visible) is discussed in Section 5.1.

DustTrak Mass Correction
PMio measurements with the DustTrak were compared to two types of mass-based PMio
measurements.  First, the DustTrak located at 2. 1 m on the master flux tower was compared to
the TEOM measurements at the same height, also located on the master tower.  Second, in-lab
tests were used to more accurately obtain a relationship  between the DustTrak measurements and
mass-based measurements. The correlation between the DustTrak and TEOM on the master
tower is quite noisy, but shows that DustTrak values would have to be multiplied by a factor of
2.8 ± 0.6 to obtain mass-equivalent PMio.  (See Figure 4-1)
   Figure 4-1. Scatter Plot of DustTrak PMio Average Concentrations and TEOM
Measurements.  Both measurements were collected at nearly the same height (2.1 m height)
  on the master flux tower.  Red dot shows averages for all sets of measurements over the
                                  course of the study.
           250
           200 -
        o  15°"
        b
        o
        I  10°
           50 -
AvgOfFirstOfMass Cone

Linear (AvgOfFirstQfMass
Cone)
y =2.8158x
R2= 0.1219
                        10
             20         30         40

                    PM10(DT)
      50
60
In the laboratory, we constructed a chamber in which the soil material that was used to seed the
road at the Boulder City site (See section 3.2) was injected and suspended.  The "resuspension
chamber"  was constructed from a modified medium volume sampler plenum (the DRI SGS-
sampler) (Gertler et al., 1993).  The dimensions  of the cone shaped aluminum  plenum and
sampling configuration are provided in Figure 4-2. The resuspension technique involves the
                                                                                    33

-------
following steps. A small amount (~ 0.5 g) of the soil is placed in a 250 ml Erlenmeyer flask that
is connected via Tygon tubing to the high-pressure air line in the laboratory.  The valve is opened
and the dust is suspended and injected into the top of the plenum by the high speed jet of air
(Figure 4-2). At the bottom of the sampling plenum, through specially designed ports equipped
with O-ring seals, dust-laden air is sampled through two Teflon filter holders (Savillex, 47 mm).
This is accomplished with pumps (URG, model URG-3000-02Q) that draw 5 1pm through each
of the filter holders. The chimneys of the filter holders are outfitted with in-line PMio impactors
(Airmetrics), similar to those used on MiniVol samplers (Airmetrics). [Note that the Airmetrics
PMio impactors are not regarded as primary reference  instruments.]  The flow rates (5 1pm) are
set using calibrated rotameters.  One of the Teflon filter holders houses a 47 mm Teflon filter.
The other filter holder is used to channel the dust-laden air (already having passed through the
PMio size-selective inlet in the chimney of the  filter holder) to two DustTrak samplers via
conductive tubing.  One DustTrak sampler is equipped with the manufacturer's PMio inlet while
the  other is equipped with  the manufacturers PM2.s inlet.   This configuration ensures  that the
DustTraks "see" the same dust-laden air that goes through the Teflon filter.  A zero-air filter is
attached to the top of the sampling plenum to allow for through flow of clean room-air through
the plenum to mix with the dust-laden air in the plenum.

Measurements with the resuspension chamber were completed within two weeks of the field
study.  Two DustTraks were randomly selected form the set of units that were used in the field
study.  Five target mass loadings were generated that spanned the ambient measured values at the
test site as recorded by the  DustTraks.  The one-second DustTrak data measured were used to
guide the target  mass loadings.  When the estimated target mass  was reached the test was
terminated.   The one-second particle concentration measurements obtained with the  DustTraks
were used to calculate a time-integrated average, which was then compared with the average PM
concentration obtained using the filter based gravimetric method for each target concentration.

The  manipulation  of the resuspension  chamber  testing was done  manually.  Lab  personnel
opened and closed all valves and started and stopped pumps manually.  All  timing of tests was
determined by the elapsed  time as  recorded by  the laptop computer recording the DustTrak
instruments. A sampling interval was defined by the amount of time  elapsed between the period
when the injected dust was first recorded by the  DustTraks (a quick and noticeable rapid rise
above background) until the program indicated that the target mass  loading had been reached.
For the tests reported here,  the target mass loadings were reached within 600 - 1300 seconds.
All  valves and pumps were closed or stopped within a few  seconds after  reaching the target
mass.  The mass concentration of PM for a sampling interval  was determined by the difference
between the post- and pre-weighed Teflon-filter membrane masses,  the  measured flow rate (5
1pm) and the elapsed time of the test (seconds).   Filters were weighed on a microbalance with
precision of 0.001 g.

Comparisons of  DustTrak measured  values  and  mass  based PMio  obtained  using this
resuspension method are shown in Figure 4-3. The top panel of the Figure shows the results for
the soils used in this study.  For comparison, the bottom-panel shows results for a soil collected
from Yuma, Arizona.  The  differences in the slopes indicate that  this mass  correction factor is
specific to the type of soil being examined.
                                                                                    34

-------
Figure 4-2. The resuspension chamber used to establish the
relationship between the DustTrak-derived PMio and
            derived by gravimetric analysis.
                                                   0.5 m
                                                   0.2m
                                                   0.56m
                                                                  35

-------
Figure 4-3. Relationships Between Gravimetrically Determined Average PMio and
 Average PMio as Measured With the DustTrak for the Boulder City Study (top
 panel) and a Separate Study Carried Out for a Desert Soil Collected at the Yuma
                       Proving Ground, Yuma AZ

                         PM10(grav)vsPM10DT
   35.000

   30.000 -
"I
1 25.000 -

•= 20.000 -
 15.000

° 10.000 -
5
Q.
   5.000 H

   0.000
            Las Vegas Study
                   y = 2.4024X
                   R2 = 0.8396
                                      6
                                PM10(DT)(mg/m3
                                                       10
                                       12
                        PM10(grav)vsPM10(DT)
      140

      120

      100

   I  30
   o
   i  60
   0.
       40

       20
          Desert Soil, Yuma AZ
                   y = 1.473x
                   Ff = 0.9875
               10
                    20
30
40     50

PM10(DT)
60
70
80
90
                                                                       36

-------
Results of this laboratory experiment showed that the DustTrak-measured PM2.5 is highly
correlated with the PMio measurements (PM2.5 = 0.501 PMio , R2 = 0.955) (Figure 4-4). The
relationship between gravimetric mass concentration and DustTrak concentrations are also quite
good. For PMio filtered mass concentration versus DustTrak we observed the relationship PMio
(gravimetric) = 2.4 ± 0.2 x PMio (DustTrak) with a correlation coefficient (R2) of 0.84 (Figure
4-3). For the PM2.5 the relationship between gravimetric and DustTrak derived mass
concentrations was PM2.5 (gravimetric) = 0.7 x PM2.5 (DustTrak) with a correlation coefficient
(R2) of 0.891 (Figure 4-5).

Based on these two sets of collocated tests, one conducted in the field and the other in the lab, we
chose a DustTrak correction multiplier of 2.4  corresponding to the in-lab measurements. Noting
that  the uncertainty  in the  regression  between the DustTrak  and  the TEOM in the field
encompasses this value (2.8 ± 0.6),  the in-lab measurements were chosen for correcting  the
DustTraks because the correlation was much  better than in the field. This was likely due to the
fact that in the field, the DustTrak and TEOM were only nominally collocated whereas in the lab,
the two instruments were sampling a well mixed controlled volume of air.
 Figure 4-4. Scatter Plot of DustTrak Monitor Outfitted With PM2.5 Inlet versus DustTrak
 With PMio Inlet. Both instruments sampled silt material from the Phase IV tests that was
                      resuspended in a specially designed chamber.
             5  -

            ^
            O)
            B
            10
            0.
            &  H
             0
                                     DTPM2.5vsDTPM10
                  DTPM2.5 = 0.501 DTPMl
                  R2 = 0.9554
  0
                0
4       6        8       10      12
   DT PM10 (mg/m3) (avg of 2 DTs)
14
                                                                                   37

-------
    Figure 4-5.  Comparison of Filter-Based PMi.s Mass Measurement With DustTrak
Outfitted With PMi.s Inlet. Both instruments sampled silt material from the Phase IV tests
                  that was resuspended in a specially designed chamber.
         4.500
  4.000 -

  3.500 -

_ 3.000 -
ro
& 2.500 -

S 2.000 -

a 1.500 -

  1.000 -

  0.500 -

  0.000 -
              0
                   y = 0.7112x
                   R2= 0.8907
                                PM2.5 (grav) vs PM2.5 (DT)
                              2          3
                                     PM2.5 (DT)
4.2 EPA Method AP-42

4.2.1  Field Balance Mass Calibration

Calibration of all postal measurement scales was carried out with Rite-O-Weigh® brass weights
meeting ASTM Class 6 adjustment tolerances.

The Sunbeam 78411 postal scale has a readability of ±1  gram and was found to read within
1 gram of the true weight from 0 gram to 200 grams, and within 2 grams of the true weight from
200 grams to 1,000 grams.

The Pelouze SP5 Postal scale has a readability of +/1 gram was found to read within 1  gram of
the true weight from 0 grams though 1,000 grams.

The Sunbeam  Freightmaster®  150  scale for soil  sample  excavation  has  readability  to
±0.1 kilogram. It was calibrated with the Rite-O-Weigh  brass weights over the 0.1-kilogram to
4.0-kilogram range and found to deviate less than 0.2 kilogram.

4.2.2  Road Plot Marking Uncertainty

Full size roadway plots 10 feet (3.05 meters) long by 13.5  feet (4.12)  meters wide were marked
with 3.05  meter and 4.12  meter  string lengths were different colors, and were tied to form a
                                                                                    38

-------
rectangle with an uncertainty of +/- 0.05 meters (5 centimeters). Corners were squared so that
the string was taut with standard building bricks, and then 2-inch masking tape was applied from
a roller dispenser to match the perimeter established by the colored surveyors' twine.  The tape
perimeter  was then marked with white surveyors paint  and the tape was removed.  White
surveyor's paint spots are laid out at one foot (0.305 meter) intervals across the road way at each
end of the 3.047 meter long plot to delineate the area to be vacuumed.

"Quickie strip" roadway plots 2 feet (0.610  meter) long by 13.5 feet (4.115 meters) wide were
laid out between the full size plots. White surveyor's paint was used to mark the  corners of the
quickie strip plots.  Painted lines or masking tape were not used to indicate boundaries of the
quickie strips. As a result, vacuum path width for the quickie strips, guided only by the eye of the
operator from the inside curb to the lane divider, tended to deviate by up to 1/6  of the 30 cm
(12 inch) width of the Hard Floor tool, or about 5.0 cm or 2 inches. This deviation in path width
results in a proportionately larger single sample uncertainty in the vacuumed area of the quickie-
strip plots compared to the vacuumed area of the full size plots.

4.2.3  Sieve Analysis Calibration

Collected  soil samples were held in sealed plastic  containers for  three weeks  in a  climate-
controlled laboratory at UNLV. Ninyo and Moore's laboratory in Las Vegas, Nevada, performed
sieve analyses. Sieves  are manufactured to  ASTM standard E-ll:87 and to AASHTO M-92.
Sieves are calibrated annually by a calibration laboratory following ATM Manual 32. All sieved
masses are determined to ±0.1 gram on a calibrated electronic balance.

The eight-inch (20.3 cm) sieve stack recommended in AP-42, Section 13.2.1, Appendix C.2. (US
EPA  1993b), consisting  of sieve numbers  3/8 inch, 4  mesh, 10 mesh,  20 mesh, 40 mesh,
100 mesh, 140 mesh, and 200 mesh, plus pan,  was used to sieve all recovered soil samples. A
standard sieve time of  10 minutes was used, per AP-42 13.2.1 Appendix C.2. The sieves were
agitated on a Tyler Ro-Tap® RX-29 mechanical Test  Sieve shaker, operating at a fixed  speed of
278 ±10 revolutions per minute with 150 taps ±5 taps per minute. Silt masses were reported as
the mass passing the number 200 (75 micron) sieve.

Upon review  of AP-42 methods for minimum soil required sample masses  (Appendix  C.2, US
EPA 1993b, page 7), where "100 to 300 grams may be sufficient when 90% of the  sample passes
a No. 8 (2.36 mm) sieve," soil masses for simultaneous parallel bags from the same  sampling
location and vehicle pass were combined for sieving to make total sieve masses exceeding
100 grams, if individual bag masses were less than 50 grams.

Sieving analyses by Ninyo and Moore were "blind" in that they did not know the location or
expected composition of the recovered soil samples.

All sieving work was completed by the end of October 2006.  Ninyo and Moore transmitted soils
data  back to UNLV as multi-page PDF files, with one page for each sample. Each page of the
PDF file contained results for one sample, organized by UNLV site identification number.
                                                                                     39

-------
4.3  Mobile Technologies

4.3.1  SCAMPER

The zero response and flow rate of each DustTrak was recorded at the beginning and end of each
day.  In prior studies the response of the rear DustTrak was compared to mass determined by
collocated  filter samples.  The  average response factor  based on  a linear  regression  was
approximately 3. Given the scatter of the data, this is in general  agreement with the correction
factor  described previously.  The response  of the  DustTraks was therefore less than when
calibrated using Arizona Road Dust. This most likely is due the PMio behind the SCAMPER
consisting of a greater fraction of larger particles than the Arizona Road Dust. The mass-specific
light scattering response drops rapidly with increasing particle size for particles larger than lum
diameter, thus  a  small  change  in  the particle-size  distribution can change the response
significantly.

The  data acquisition system recorded  all data digitally  at one-second intervals. Data  was
downloaded from the PC and entered into an Excel worksheet where all of the calculations were
made. Quality control data such as inlet pressure and various voltages were also entered into the
master worksheet in addition to GPS location, time, speed, and DustTrak values.

Data was validated to Level 0 and then Level  1 status from QC pressure and voltage data,
logbook entries, and by observing time  series, to determine if the results made  physical  sense.
The data was flagged as follows in the Excel worksheet:

       0 or blank: valid data
       1: missing or erroneous
       2: DustTrak on filtered air for zero check- not moving control
       3: DustTrak on filtered air for zero check-moving control
       J: DustTrak values not changing for 30 seconds of more
4.3.2  TRAKERI

The DustTrak instrumentation onboard the TRAKER vehicle has a resolution of 1 ug/m3.  Thus,
the smallest measurable difference in concentration between the tire and the background monitor
locations is 1 ug/m3. This corresponds approximately to a single-point minimum detection limit
equivalent to an emission factor of 0.0005 g/VKT (0.0008 g/VMT) for paved roads, meaning
that any 1 s measurement can be resolved to within this value only. In practice, emission factors
from real roads are generally higher than 0.01 g/ VKT (0.016 g/VMT).  At the other end of the
measurement range, DustTrak readings above 150 mg/m3 are not reliable.  This corresponds to
an emission factor for PMio of approximately 75 g/VKT (120 g/VMT).  Again, in practice, 20
g/VKT (32 g/VMT) represents an upper limit to paved road PMio emissions.
Figure 4-6 shows the TRAKER coefficient of variation calculated from the left and right
DustTrak signals as a function of vehicle speed.  The coefficient of variation is a measure of the
relative precision and is equal  to  the standard deviation of the measurement divided by the
                                                                                      40

-------
average of the measurement. In the figure, the measurement corresponds to multiple passes on
the same 1-mile stretch of road (Etyemezian et al., 2003). The figure shows that the precision of
the measurement improves with increasing vehicle speed. The precision is 84% at 5 m/s, 30% at
9 m/s, and approximately 10% above 14 mps.  Note that most TRAKER measurements occur at
speeds greater than 9 m/s (approximately 20 mph).  The poor precision at low speeds is probably
due to the influence of fluctuating ambient winds on the flow regime behind the front tires.  As
the vehicle speed increases, such fluctuations  become less important compared to the speed of
the vehicle.

     Figure 4-6. TRAKER coefficient of variation expressed as a percentage for left
     and right PMio DustTrak signals as a function of speed. The data represent left
     and right PMio DustTrak signals averaged over a 1-mile stretch of road near
     Boise, Idaho (Etyemezian et al., 2003). The coefficient of variation provides an
   estimate of the precision and is equal to the standard deviation of a measurement
                               divided by the average.
               100
                90 -

                80 -


             35  70-
             c
             o
             «  60 -

             1
             •5  50 -
             o
             g  40 -

             O  30

                20 -

                10 -

                0
                                             15        20
                                        Vehicle Speed (mps)
The vehicle speed can become important in moderate to high winds. If the TRAKER is not
moving fast enough, crosswinds and fluctuations in the ambient winds can lead to unsteady flow
conditions between the front tire and the inlet.  To avoid this possibility, a minimum speed of 5
m/s is required to consider a data point valid.  Acceleration/deceleration criteria (<0.7 m/s ) are
also applied to the TRAKER measurement.  During periods of high acceleration, the flow regime
around the inlets may be transient; during periods of deceleration, dust from the brakes may
influence the particle concentrations behind the front tire.  Note that in the prior  work of
Etyemezian et al. (2003a, 2003b) and Kuhns et al. (2001) the criterion for acceleration was 0.5
m/s2.  Relaxation of the criterion for the present study should not affect the measurement quality
significantly since the original criterion was set to be overly conservative.

In addition, the wheel angle must be less than 3  degrees with respect to the vehicle body. This is
to ensure that the orientation of the inlets with respect to the front tires is not changing over the
                                                                                     41

-------
course of the measurements.  The vehicle  speed, acceleration, and wheel angle are calculated
from the time derivatives the 1-second GPS coordinates. The criteria shown in Table 4-1 are
based on empirical observations and statistical analyses  of the TRAKER measurement under a
variety of driving regimes. These criteria are applied to the one-second data prior to any further
aggregating or averaging.  They are conservative and intended to ensure that the measurements
used in this study are valid.

           Table 4-l.Validity criteria applied to each 1 s TRAKER data point.
Parameter
Speed
Acceleration
Deceleration
Wheel Angle
Criterion
>
<
<
<
Threshold
5 m/s - paved roads
(-11 miles/hr)
0.7 m/s2
(~ 1 .3 miles/hr/s)
0.7 m/s2
(~ 1 .3 miles/hr/s)
3 degrees with respect to
the vehicle body
Description
Minimize disturbances due to
ambient winds.
Lateral shear during acceleration
and transient airflow around the
TRAKER inlets render TRAKER
measurements during times of high
acceleration unreliable.
Applying the brakes releases dust
particles and may result in false
high road dust readings.
Turns cause the front wheels to
form an angle with the vehicle
body. This in turn changes the
orientation of the TRAKER inlets
with respect to the front tires. Data
associated with sharp turns are not
valid.
Level 0 validation was performed by examining the DustTrak and GPS time series for the entire
study.  The data were examined  for completeness and correspondence with known sampling
times.  GPS data were checked by mapping coordinates from the GPS receiver on a spatially
referenced GIS  map. Any documented deviations in flow rate or procedure were examined to
ensure that they did not affect data quality. For the entire study, all instruments were found to be
logging as expected and no deviations from normal operating procedure were noted. In addition,
the DustTrak zero-check on all days indicated that there was not significant instrument drift from
day to day (i.e.,  correction required was less than 3
Level I validation included examination of the time series for each pass that was completed
through the test course. We looked for sudden jumps (spikes or troughs) in the DustTrak record
as well as in the GPS time series.  In TRAKER I, the DustTrak samples air from a plenum with
an approximate residence time of 2 seconds. Thus, spikes in PMio concentration that appear for
only one second are considered suspect data. No such data were found for the present study.
Level II validation included examining relationships between the signals on the left and right
TRAKER inlets as well  as  over the course of a measurement  set.   The ratio of PMio
concentrations measured behind the tires to those  measured at the background (bumper) inlet
was also examined to ensure that the TRAKER signal was substantially above background.
                                                                                    42

-------
4.3.3  TRAKERII

Noting that the use of TRAKER II as part of this study was experimental and that this updated
version of TRAKER I has not been as extensively characterized, TRAKER II data were handled
in a manner  similar  to TRAKER I.  The same  speed, acceleration and  wheel angle criteria
applied to TRAKER I (Table 4-1) were also applied to TRAKER II on a one-second basis.
Level 0 validations included ensuring that all instruments were operating and logging data during
the measurement period.  Level I validation included examination of time series of DustTrak
concentrations and GPS data. Time series of the flow rates through the left and right inlets were
also examined for deviation from the fixed value of 10 1pm. It was discovered during this
examination that for all of 9/11/06 and a portion of 9/12/06 the flowrate through the inlets was
not being properly maintained at 10 1pm, but rather was held at 6 1pm. This problem was
attributed to a glitch in the software that controls the TRAKER II data acquisition system and
repaired in the field.  In summary, TRAKER II passes with Pass IDs of 170 and higher were
considered level I valid whereas those with Pass ID lower than 170 were considered invalid.

Level II validation was conducted as part of the data analysis for this study and the outcomes of
that effort are summarized in a later section along with other study findings.
5.0 DATA HANDLING

5.1 Horizontal Flux Tower

Horizontal PMio fluxes were calculated from the tower data for all individual passes that met the
validation criteria outlined  in  Section 4.1.   This measurement is similar in principle to the
upwind/downwind technique employed in previous work (e.g. Cowherd, 1999) with one major
difference in its practical application. As the name implies, the upwind/downwind method relies
on measuring the horizontal flux of PMio through the upwind side of the road and the downwind
side of the road separately. The flux on the upwind side is subtracted from the flux measured on
the downwind side in  order to  determine the net contribution of horizontal PMio flux from the
road.  The technique used in this study employs only one tower that is located on the downwind
side.  Since the source of PMio road  dust was intermittent and associated  with the passage of
individual vehicles through the  test road, periods when there was no vehicle activity through the
test road where considered to represent the background horizontal PMio flux. In this sense, these
periods  correspond  to the  "upwind" measurement.    Similarly,  times when the tower was
impacted by the passage of a test vehicle correspond to the "downwind" measurement.

As mentioned in that section, the two factors that were used to determine whether a data point
was valid or not were the wind  direction/speed and the background noise level, determined from
periods with no influence from  any of the test vehicles. The general approach for calculating the
horizontal PMio flux was to assume that the master tower was located in a flux plane parallel to
the road and  that the multiple  vertical  measurements of wind  speed, wind  direction,  and
DustTrak PMio concentrations  each represented a discrete  section  of the tower height.  This is
illustrated in Figure 5-1 which  shows instruments mounted at 0.7, 2.1, 3.4, 6.4, and 9.8 m above
ground level representing the sections  spanning 0-1.4, 1.4 - 2.75,  2.75 - 4.9, 4.9 - 8.1, and 8.1
                                                                                    43

-------
- 12 m, respectively.  Following the method of Etyemezian et al. (2004), each DustTrak was
assumed to  represent the concentration of PMio over a distance that spanned halfway to the
DustTrak location above and halfway to the DustTrak location below.

   Figure 5-1. Illustration of Portions of Flux Plane Represented by DustTrak and Wind
  Instruments at Each Height. The dots shoe the instrument locations and the horizontal
      lines show the height range that the instruments represented in calculating flux.

                               16
                             0)
                             s
                               10
                                8
                             o>  6
                            '53
                            I
An emission factor (EF, g/km) for each vehicle pass was calculated using the equation:
                  EF = a
 5   tend     /
y  y\uti-(cti
Z_j  jL-i  t,i  \ t,i
-c
xlOOO
Equation 5.1
where: /' refers to the vertical section represented by the DustTrak height, t is the time (sec), t begin
is the peak start time, tend is the peak end time, u is the wind speed (m sec"1), C is the measured
concentration (g m"3), Co is the background concentration over the period tbegm - tend (g m"3), and
//is the height of the section of the flux plane represented by position /',  $is the angle of the 1-
sec wind direction relative to the flux plane, and a is a constant used to convert DustTrak-
measured PMio concentrations to mass equivalent PMio and has a value  of 2.4 (See Section 4.2).
An example calculation is provided in an Appendix to this report.

In some  cases, DustTrak concentration peaks were clearly discernible and associated with the
known passage of a vehicle.  In practice, this required that DustTrak concentrations departed
                                                                                     44

-------
from baseline values on multiple DTs within 15 seconds of the passage of a test vehicle in front
of the master tower.  In those cases where peaks were clearly associated with the test vehicle, the
peak curves were divided into three intervals.  The first interval corresponded to the background
PMio concentration prior to the peak and included the  10-30 second period that ends with the
peak start time.  The second interval was bounded by  the peak start and stop times (giving the
values of tbegin and tend\ which were determined visually as the instance when any of the tower-
mounted DustTraks  began exhibiting a peak in concentration to the instance when  all  of the
tower-mounted DustTraks exhibited a return to baseline concentration values.  The third interval
corresponded to  the background PMio concentration after the end of the peak and included the
10- to 30-second period after the peak stop time.  The first and third  intervals were aggregated to
estimate the baseline average PMio concentrations (Co  in Equation 1) for each DustTrak and the
noise level (standard deviation) exhibited by the background signal.  For cases where a peak was
not clearly discernible, the peak duration was assumed  to span 20 seconds that were centered on
the recorded vehicle passage time.
Horizontal fluxes calculated using Equation 1 yielded an emission factor in units of gram
per kilometer traveled for every time a test vehicle passed through the test course and wind
conditions and background PMio levels  were  considered  acceptable for  providing a valid
measurement.

5.2 EPA Method AP-42

5.2. 1  Organizing Bag Data

Soil sample bag data, consisting of a bag number, its assigned UNLV site number, date and time,
tared mass, and final mass were entered into a the MS Access® database. This database was used
to organize and print out bag identification data in tables were transmitted with the soil samples
to Ninyo and Moore's geotechnical laboratory for soil sieve analysis.

5.2.2  Organizing AP-42 Emission Factor Data

Returned silt masses from AP-42 sieving conducted by Ninyo and Moore were manually entered
into the Access® bag database.

The Access® database table was then exported to an Excel® database to facilitate calculation of
AP-42 Emission  Factors. The silt  recovery time that most closely matched  the  time of a
particular  vehicle pass identification number, taken from the DRI Excel  vehicle Pass_ID and
time database, was  used to match silt recovery to a mobile technologies event.   An entry was
made in the AP-42 Excel® database to indicate if the silt recovery had taken place before or after
the vehicle  Pass_ID.   Where available, separate silt mass  values were  entered  for each
corresponding Pass_ID for both the south and north zones. Silt mass data were  then converted
into silt loadings by dividing by the corresponding plot area in square meters. Uncertainties in
individual silt loadings were computed using root-mean  square (RMS)  error analysis  of the
uncertainty in the silt mass and the uncertainty in the plot area.
                                                                                      45

-------
AP-42 emission factors were then calculated for the silt loadings using the AP-42 emission factor
equation.
                              EF = k * (SL/2)0'65 (W/3)1'5 - C,
                                                                  Equation 5.2
where:
EF =
k  =
          SL =
          W  =
          C  =
the computed AP-42 PMio emission factor in gram/VMT or gram/VKT
the coefficient for PMIO, with values of
   7.3 gram035-mL30/(VMT-tonl.5) or
   4.6 gram035-mL30/ (VKT-tonl.5)
silt loading in gram/m2 calculated from field measurements,
a fleet average vehicle weight in U.S. short tons, and
the brake and tire wear correction factor, with values of:
   0.2119 gram/VMT, or
   0.1317 gram/VKT.
A weight of 2.88 tons, based on the arithmetic average of the reported weights of the three
mobile source vehicles (SCAMPER 2.5 tons, TRAKER I 3.4 tons, and TRAKER II 2.75 tons)
was used to calculate the AP-42 emission factors from the silt loadings.

Uncertainties in the  individual emission factors were computed using root-mean square error
analysis of the uncertainty in silt loading.  Fleet vehicle weight was assumed to be known
exactly, with an uncertainty of zero.

In cases where multiple silt loading measurements, in the north or south, were available for a
particular Pass_ID, the average north or south silt loading measured for that pass was used to
compute the  AP-42 emission factor.   Standard deviations of the north and south silt loadings
were calculated, and for each zone, the larger value of the individual RMS silt uncertainty or the
plot-to-plot silt standard deviation was used in a root mean square computation of the AP-42
emission factor uncertainty.

Averages  and standard deviations of the silt loading and  AP-42 Emission factors  for each
Pass_ID were computed from the combined north and south zone data, where available.  The
larger uncertainty of the RMS error calculation or the north-south standard deviation was used as
the uncertainty of the AP-42 emission factor measurement.

The Excel® database containing date, time, vehicle Pass_ID, vehicle speed, silt loadings and silt
loading uncertainties, and AP-42 emission factors and emission  factor uncertainties was then
transmitted to all cooperating agencies for data analysis.

5.2.3  Unification of Data Sets

DRI combined the following data sets using Vehicle Pass_ID as a common variable into a master
Excel database that was used for joint data analysis:

     (1) UNLV AP-42 emission factor data, averaged north and south for each pass,
     (2) Tower mass emission rate data, averaged for each pass,
                                                                                      46

-------
     (3) SCAMPER, TRAKER I and TRAKER II mobile technologies data, averaged for each
        pass.

5.3 Mobile Technologies

5.3.1  SCAMPER

The data acquisition system for the SCAMPER collects GPS and digital DustTrak values once
per second and stores them in a folder by hour of the day.  These data were then merged into an
EXCEL spreadsheet  for post-processing.  The one-second data from both the  front and rear
DustTraks were corrected for the  average zero response (from the beginning and end of each
set), and then the front concentration was subtracted from the rear. The result was multiplied by
the frontal area of the Ford Expedition (3.66m ), to yield the emission factor in mg/m. All data
with a flag of 1 (missing or erroneous data) were  removed from the data that were submitted.
The master Excel worksheet shows all the calculations and all flags.

Data for the test track were selected from the GPS coordinates of the test track boundaries and
the heading  of SCAMPER.  The test track was divided into southern and northern segments to
facilitate comparisons with the AP-42 silt sampling conducted  on those ends of the test track.
The following coordinates were used for boundaries:

Location           Latitude      Longitude
SouthEnd         35.94798333  -114.8470833
North End          35.95065     -114.8546667
Middle:            35.9493303   -114.851408

The data were checked to insure that flags 2 and  3 (for QC checks conducted away from the
Sampling Zone) were removed in this process. No "J" flags (concentration unchanged for 30 sec)
were found in this data set. There were occasional periods when the GPS did not  report data,
most likely due to interferences in the sight path to a satellite. In these cases the cell was filled
with the average of the position before and the position after. The  same was done for speed and
PMio emission rate. Averages and  standard deviations of this emission rate data were calculated
for the southern end, northern end,  and full track each test pass.

Concentration units (mg/m3) were used in the "calibration" with the tower performed  by DRI.
These units were derived by dividing the emission rate originally reported for the full  test track
by the frontal area of the Expedition (3.66m2).

5.3.2  TRAKER I

Following validation of individual  one-second TRAKER I data, several steps were taken to align
and aggregate the  data points for data analysis.  First, the  GPS time  stamp was retarded 3
seconds and linked to the DustTrak data using the  retarded time.  This was done to  account for
the discrete amount of time (3 seconds) that it takes for the air at the inlets of the TRAKER I to
move through the inlet lines and plenum and the DustTrak sampling nozzle. That is,  data logged
by the DustTrak at time to corresponds  to the dust that was channeled to the inlet of the
TRAKER (either behind a tire or through the bumper at time to - 3 seconds.
                                                                                     47

-------
Next the TRAKER signal was calculated for all valid data points using the equation:
T -
L -
                                              V 2
                                            L,tJ< z
Equation 5.3
Where Tt is the TRAKER signal in mg/m3 at time t and CR, CL,  and CB are the concentrations
(mg/m3) respectively measured at the right, left, and middle (background) inlet.  The quantity T
in Equation 1 is the main entity that is provided by the TRAKER measurement system and is the
"raw" TRAKER signal.

Next, only data that correspond to the test  route were selected for analysis.   This was
accomplished by imposing limits on the latitudes and longitudes of the  GPS coordinates as well
as the direction of  travel of the vehicle (See  Figure 5-2).   After extracting only data that
correspond to measurements along the test route, each data point was associated with a Pass ID
number common to  all study participants.  Depending on the speed of travel on the test  route,
between 28 and 57 points were associated with each Pass ID that was  assigned to TRAKER I.
An example of the raw vehicle Pass ID data is shown in Table 5-1. Pass durations are about 1.5
minutes at 35 mph intervals between successive vehicle passes within a given Run ID.
   Table 5-1. Example of Vehicle Pass_ID Data.  Pass durations are about
        35 mph intervals between successive vehicle passes within a given
                                       1.5 minutes at
                                      Run ID.
Date
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
9/11/2006
Set ID
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Test type
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Pre-sweep
Run ID
1
1
1
1
1
1
2
2
2
2
2
2
3
3
3
Pass ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Vehicle
UC
TR1
TR2
UC
TR1
TR2
UC
TR1
TR2
UC
TR1
TR2
UC
TR1
TR2
Speed
(mph)
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
Drive
Direction
N
N
N
S
S
S
N
N
N
S
S
S
N
N
N
Time
(Local)
11:56:20
11:57:32
12:02:49
12:04:25
12:05:53
12:07:09
12:08:49
12:10:18
12:11:42
12:13:23
12:15:04
12:16:26
12:18:00
12:19:27
12:20:25
Exact
time?
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
There are 468 total passes in the database that covers the five days of Phase IV experiments.
Data in Table 5-1 are shown only for the first 15 passes of Set 1.

Experimental Set_ID numbers describe different experiments that took place during the Phase IV
experiments.  Each Set number describes a different  experimental condition. Usually, each Set
ID number describes  a unique combination of applied silt loading and mobile technology vehicle
speed.
                                                                                    48

-------
A summary of the Pass_ID numbers that correspond to each Set_ID in the Phase IV study is
shown in Table 5-2.

Table 5-2. Summary of Set_ID's and corresponding Pass_ID's for the Phase IV study
Date
9/11/06
9/11/06
9/11/06
9/12/06
9/12/06
9/13/06
9/13/06
9/13/06
9/14/06
9/14/06
9/14/06
9/14/06
9/15/06
Set#
1
2
3
4
5
6
7
8
9
10
11
12
13
Experiment
Name
Pre-Sweep
Post-Sweep
Apply silt #1
Apply silt #2
Apply silt #3
Apply silt #4
Apply silt #5
Apply silt #6
Apply silt #7
- Depletion,
one vehicle
Apply silt
#7- all
vehicles
Post-sweep
Apply silt #8
- strong
winds
Apply silt #9
- strong
winds
Start
Pass ID
1
63
93
140
170
212
243
273
309
319
334
365
392
End
Pass ID
60
92
139
169
211
241
272
308
318
331
364
391
476
Nominal
Drive
Speed
(mph)
35
35
35
45
25
45
25
45
35
35
35
Repeat
25,35,45,
45,35,25
cycle
twice
Repeat
25,35,45,
45,35,25
cycle 4
1/2 times
Applied
Soil
Loading
(gram/m2)
N/A
N/A
6.16
17.17
16.58
4.99
4.70
7.63
7.78
7.78
N/A
17.61
28.47
Avg.
Recovered
Silt
Loading,
(gram/m2)
0.17
N/A
0.75
2.48
3.17
0.88
0.74
1.14
0.80
0.80

2.55
2.31
     Pass IDs 61, 62, 242, 328, 332, 333, 358, and 449 do not correspond to test vehicles used in
     this study.

To facilitate comparison among the  different measurement systems (TRAKER II, SCAMPER,
Tower measurements,  and  silt measurements), all real-time data were aggregated by vehicle
pass.  For the remainder of data analysis, pass-averaged TRAKER signals are used. That is, the
TRAKER signal  (Equation 5.3) was averaged over all real-time data points acquired during a
specific Pass ID, and the resulting average value was used to represent the TRAKER I signal for
that Pass  ID.  Each pass corresponded to a linear distance of approximately 760 m (distance that
spans northern and southern locations where AP-42 measurements were performed). The effects
of this assumption/simplification were examined by comparing pass-averaged TRAKER I
signals to the averages of data points that correspond only to measurements taken within 50 m of
the master tower (See Figure 5-2b). Figure 5-3 shows that there is a good correlation  (R2 =
0.82) between the pass-averaged TRAKER I signal and the TRAKER I signal averaged only
                                                                                  49

-------
over  data  points  that  correspond  to  measurements within  50  m  of the  master tower.
Nevertheless, the relationship does exhibit substantial noise (note: in log scale) indicating that a
number of factors can change over the length of the test route including the road dust loading and
the portion of the lane where the driver is driving the vehicle.

Figure 5-4 shows a time series of pass-averaged TRAKER I signal over the whole length of the
test road section (between the longitudes: -114.854849 and -114.847239) as well as averages in
the vicinity of the DRI flux tower (between longitudes of -114.853817  and -114.852524, See
also Figure 5-2).  The "Tower-averaged" TRAKER signals tend to exhibit  more pass-to-pass
variability than the "pass-averaged" signals. This is to be expected since the former are averages
over a smaller number  of individual one-second measurements (5-12) compared to the  latter
which include many more data points (25 - 60).  Larger numbers  of data points in the average
mitigate variations in driving technique and road dust distribution on the test road surface.  In
some cases, (e.g. Set 3 and Set 12), the "pass-averaged" TRAKER I signal is consistently higher
than the tower-averaged signal, indicating that silt was probably not applied uniformly  over the
length of the test road.   In  addition, prior to application  of silt (i.e. Sets 1 and 2), there are
substantial  differences between the TRAKER I signal over the entire test road length and the
TRAKER I signal  in the vicinity of the tower.  This  is true for both eastbound and westbound
travel.  This suggests that the "natural" condition  of Veterans Memorial Highway around the
area of the  measurements consists of a high degree of spatial variability with respect to road dust
emissions.   However, overall, agreement between TRAKER I signal averaged over  the two
different lengths of road is quite good (See Figure 5-3).

In applying Equation 5.3  to obtain the TRAKER I signal, two observations are worth noting.
First, the value  of CR and  CL were substantially  higher  than  CB  for  all TRAKER  I passes
(Figure 5-5). This indicates that the "influence" measurements behind the two front tires were
able to resolve a signal  substantially above background (minimum of a factor of 10) for  even the
cleanest road conditions encountered over the duration of the study.

Second, the signals (concentrations) measured behind the right and left tires were not  equal over
the course of the study.  While the ratio of the right to left  signal fluttered about unity for many
of the test  passes,  for some measurement sets (11 and 13), the right signal was  considerably
higher than the left  signal and for other measurements sets (1 and 2) the  opposite was true.
Figure  5-6 shows  a time series of the ratio of the TRAKER I right inlet signal to the left inlet
signal.  The vertical lines in the Figure indicate the beginning of a new measurement set and the
gray squares indicate passes  along the same test route in the eastbound (instead of the primarily
used westbound) direction. Note that eastbound passes were conducted in the lane adjacent to
the one  where westbound passes were completed. The figure shows that the ratio of right to left
inlet signals can vary substantially.  This variation does not appear to be caused by moderate
cross-wind (< 6 m/s or < 13  mph), but rather by variations in the distribution of road dust
material on the road as well  as variations in where  the vehicle tires are with respect to the lane
(i.e. where  the driver guides  the vehicle with respect to previous passes and drivers of other test
vehicles).  Having  noted these asymmetries, the actual PMio emissions are a combination of the
signals from both sides of the vehicle. Thus, using the average of the left and right signals as is
done in Equation 1  is appropriate for estimating road dust emissions.
                                                                                      50

-------
Figure 5-2. Schematic of GPS Data Points on Top of Street Layout, a. All TRAKERI GPS
 data and b. only data that correspond to the test route or data collected within 50 m of the
        master tower (black dots). The gray cross shows the approximate location
                                of the master tower.
                                 a. all TRAKER data points
      b. Data filtered for validity and location on test route and/or within 50 m of master tower.
                                                                                51

-------
 Figure 5-3. Relationship Between TRAKERI Signal Averaged Over Entire Pass (route
            length) and TRAKER I Signal Only Within 50 m of Master Tower
                  1000
               t
                e
                  100 -
                   10 -
                o
                H   0.1
                  0.01
                    0.01
                                                      Rz = 0.8195
                               0.1          1         10        100

                                  Pass-averaged TRAKER signal (mg/ni3)
                                                                       1000
 Figure 5-4. TRAKER Signal (left and right) Averaged Over Entire Pass (pass-avg.) and
Averaged Over Only the Portion of Test Road in the Vicinity of Flux Tower (Tower-avg.)
                   —•— TRl_Run_ave_TRAKER_signal_mgm3
                    D Run ave eastbound
T Rl _T ower_ave_T RAKER_signal_mgm3
Tower ave eastbound
                             3    4    5    6  7   8   10  11   12
               0     50    100    150    200    250    300    350    400     450     500
                                                                                           52

-------
Figure 5-5. Ratio of Background (middle) Inlet PMio Concentration to Average of Left and
                   Right Tire Inlet PMio Signals for TRAKERI Passes
               10000
               1000
                100
                 10
                                                 *    .
 T-^


            11
         ^Mr
                   0    50    100    150    200   250    300   350   400    450   500
                                            Pass ID

 Figure 5-6. Time Series of Ratio of Pass-Averaged TRAKER I Right to Left Inlet Signal
 Ratios and Pass-Averaged Wind Speed (m/s). Squares denote passes where travel was in
 the eastbound direction. Vertical lines represent times when the road was swept and silt
  was applied, while double vertical lines represent times when the road was swept only.
              Numbers at the top correspond to different measurement sets.
                    Ratio of Right to Left Inlet Signals    D Eastbound Passes
-*— Tower_ave_Wind_Speed_ms
                        2    3
                                           6   7    8   10
                                                              •   .  .
                                                              ., ./  •' in ^b
                                                             • ^* '  *, m ill i>
                                                                             -- 10
               0      50
                                                                              12
                                                          350     400    450    500
                                                                                    53

-------
Sets 4,  5, 6, 7,  and 8 were analyzed further to examine the rates  of decay as  measured by
TRAKER I.  Sets 1, 2, and 3 were not associated with uniform road  silt coverage.  Sets 9-13
were  associated with rather variable loadings, owing perhaps to the redistribution  of road
material by high winds.  For each of Sets 4- 8, the first TRAKER measurement was taken to be
the baseline reading and a measure of the amount of measurable road dust at the beginning of the
set - i.e. immediately after silt material was laid on the road. Thus, all measurements within the
same set were divided by this value to normalize the rate of decay across the different sets.
Figure 5-7 shows the normalized decay curves for Sets 4-8.  The solid  black circles and
triangles in the figure represent the average normalized decay for sets completed at a  25 mph
measurement speed (Sets 5 and 7) and Sets  completed at a 45 mph measurement speed (Sets 4, 6,
and 8), respectively. In examining the average decay curves for the  25 mph  Sets (Figure 5-8)
and the 45 mph sets (Figure 5-9) separately, it appears that in both cases, the  decay rate can be
described by two separate decay processes.

We hypothesize a conceptual  mechanism  for the reduction in  TRAKER-measured road dust
emissions  over the course of a measurement set.  The roadway surface is not completely smooth
and there  are pits and protrusions on  even  the  smoothest asphalt surfaces.   When road silt
material is placed onto the road by the spreader, a portion of the material nestles into the  pits and
a portion settles on protrusions in the asphalt.  The suspendable material that is associated with
the protrusions is more exposed than the material that is nestled in the pits.  We hypothesize that
aerodynamic  forces generated by the  passage of the vehicles are able to influence the road dust
associated with the protrusions and entrain a portion of that road  dust. In contrast, road dust
material that  is nestled in the pits is protected somewhat from aerodynamic stress generated by
the movement of the test vehicles through  the air above the surface.  Road dust material in the
pitted portions can only be entrained through contact with (or more generally influence from) the
tire surface.  With this conceptual model, we can propose a mathematical  reconstruction of the
removal of road dust from the driving  surfaces.

First, we assume that road dust placed on the road surface is either associated with protrusions in
the road and  is referred to in our model as "aerodynamically suspendable" road dust (RDA)  or
nested into the pits of the asphalt surface  and referred to  as "mechanically suspendable" road
dust (RDM).  These two categories sum to the total suspendable road dust (RDT):

                                   RDT = RDA + RDM                       Equation 5.4


                                RDT0 =  RDA, + RDM0 = 1                    Equation 5.5


Equation 5-5 above reflects that  at time 0, before any vehicles traverse  the tests course, the
normalized sum of RDA and RDM is unity. Second, assume that the decay curves for RDA and
RDM are  first-order.  In words,  this  means that each time a test vehicle  passes over the road
surface; some percentage of  the RDA and  some  percentage of the RDM  are suspended.
Mathematically, this is written as

                             RDA(X) = RDA. • EXP(-aaero • X)                 Equation 5.6
                                                                                     54

-------
                           RDM(X) = RDM, • EXP(-amech • X)
Equation 5.7
where X is the number of test vehicle passes after road silt material has been applied, aaero is the
coefficient of decay for aerodynamically suspendable road dust, and amech is the coefficient of
decay for  mechanically suspendable road dust. These two decay coefficients can be thought of
as the fraction of  suspendable  road dust (either aerodynamically or  mechanically)  that is
removed (suspended) each time a vehicle passes over the road surface.

The data shown in Figure 5-8 and Figure 5-9 fit this hypothesized conceptual model quite well.
The implications of this model may be quite important for road dust management practices.  In
the context of the present study,  these data indicate that dust emissions occur under a different
regime during the first 9 vehicle passes than in ensuing passes.  Since for a paved road, the
volume of vehicles  is generally much higher than  9, the first  9 passes after  silt material
application probably do not reflect the regime under which real-world dust emissions occur. It is
more likely  that the latter passes (greater than 9) more accurately reflect the slower,  steadier
emissions  of PMio road dust that occurs on paved roads.  Note that this observation does not
depend on whether  or not our earlier hypothesis regarding the separation of "aerodynamically
suspendable" road dust (RDA) and "mechanically suspendable" road dust (MDA) is  deemed
physically plausible.  It is clear from Figures 5-7,  5-8 and 5-9  that the rate of road dust
emissions  changes after the first 9  (or so) vehicle passes.  This phenomenon is seen  not just
through the TRAKERI measurements, but from the results of all of the measurement techniques,
namely AP-42 silt,  SCAMPER,  TRAKER II, and tower horizontal PMio  flux measurements
(Illustrated in Chapter 6 of this report).

  Figure 5-7. TRAKER I Signal Normalized to First TRAKER I Pass of the Measurement
Set for Sets 4, 5, 6,  7, and 8.  The black circles and triangle represent averages for speeds of
                         25 mph (circles) and 45 mph (triangles)
          1.2
                                                               «— Set 4-45mph
                                                               *— Set 5-25mph
                                                               »  Set 6 - 45 mph
                                                               x— Set 7-25mph
                                                               «— Set8-45mph
                                                                 - Sets 5,7-25mph
                            10      15      20     25      30

                                Passes After First TRAKER Pass in Set
                                                                35
                                                                       40
                                                                               45
                                                                                      55

-------
 Figure 5-8. Normalized TRAKERI Decay Curve for Sets 5 and 7 (25 mph measurement)
  and Hypothesized Aerodynamically Suspendable, Mechanically Suspendable, and Total
          Suspendable (aerodynamic plus mechanical) Road Dust Decay Curves
                             RDT = RDA + RDM
                                                       - Sets 5,7-25mph
                                                       All_suspendable
                                                       Aerodynamically _suspendable
                                                       Mechanically _suspendable
                                    = 0.75.Exp(0.38.X)


                               RDM = 0.25 . Exp(0.029 . X)
                                   10       15      20      25      30

                                   Passes After First TRAKER Pass in Set
                                                                          35
       Figure 5-9.  Normalized TRAKER I Decay Curve for Sets 4, 6, and 8 (45 mph
      measurement) and Hypothesized Aerodynamically Suspendable, Mechanically
           Suspendable, and Total Suspendable (aerodynamic plus mechanical)
                                Road Dust Decay Curves
                                            -•-Sets 4,6,8-45mph

                                            -S- Modeled-All_suspendable

                                            -H—Modeled-Aerodynamically_suspendable

                                            —*—Modeled-Mechanically_suspendable
                                            DDDDDDDDDDDDD D [J—

                                   10       15      20      25      30

                                   Passes After First TRAKER Pass in Set
35
In prior work, it was observed that the TRAKER I signal was dependent on the speed of travel
on the road that was being measured.  Those speed response relationships summarized in prior
                                                                                       56

-------
work were obtained by traversing the same section of road several times at varying travel speeds
(15 - 60  mph).  The underlying  assumption behind those tests was that the test road was
essentially unaffected by the passage of the TRAKER I and provided a constant "loading" of
road dust, allowing us to isolate the effect of traversal  speed on the TRAKER I signal (i.e. road
"dirtiness" was constant throughout speed tests). In several of the prior studies, it was found that
the TRAKER I signal for a given, time-invariant test road, was approximately proportional to the
speed of traversal raised to the third power (cubed).

It  is instructive to  extract a similar speed response  relationship from  the present study for
comparison. However, there are some complicating factors.  First, the only full set of speed tests
were completed on the last day of the field study during Set 13. Second, owing to the high winds
on that  day, the ratio of right to left inlet signals was quite variable (See Figure 5-6).   Third,
owing to the nature  of the field study, the road dust loadings were constantly changing over the
course of the Set 13  measurements.  In order to extract speed response information comparable
to the speed tests reported in earlier work, it was necessary to account for  these three non-
idealities.   Set 13 TRAKER I passes were separated  into 4 complete cycles, with each cycle
consisting of 2-25 mph, 2-35 mph, and 2-45 mph passes (See Figure 5-10).  Using only the
TRAKER I signal from the right side of the vehicle  (the side sheltered from  direct southerly
crosswinds which were  prevalent during  Set 13), the TRAKER I signal  from the two 25 mph
measurements within each cycle were averaged and assumed to reflect the average condition of
the roadway over the cycle. The two 35 mph measurements within each cycle were averaged
together as were the  two 45 mph measurements.  To account for cycle-to-cycle changes in road
conditions, these averages were normalized to the 25 mph average for each cycle. The results of
this normalization for each of the four cycles appear in Figure 5-11 as do the normalized data
averaged over all four cycles. A least-squares power-fit to the 4-cycle average suggests that the
TRAKER I signal for Set 13 data approximately obeys  a cubic (regression  exponent = 3.1)
relationship with speed, though we note that there are some differences from cycle to cycle.
     Figure 5-10. Division of Set 13 Into Four Cycles, With Each Cycle Comprised of
                                Six Passes for TRAKER I
           <
           &
              100
               10
               1 --
              0.1
                                                                      50

                                                                      45

                                                                      40

                                                                     - - 35
Pass-average
Right
TRAKER
signal

Speed
 30

- - 25

- 20

- 15

- 10

- 5
                                                                      0
                 0     10    20    30    40    50    60    70
                                    Passes Since Silt Applied
                                                          80
                                                               90
                                                                    100
                                                                                      57

-------
Figure 5-11. Speed Response of TRAKERI Signal.  Figure shows the TRAKERI signal at
 each speed normalized to the average signal at 25 mph in the same cycle. Data are shown
             for 4 consecutive cycles as well as the average value for all cycles.
              10
               9 -
               ™
               / -
            I  6^
            ^  4 -
               1 -
 «  Cyclel
 n  Cycle2
 A  Cycle3
 x  Cycle4
 •  Averages_all_Cycles
	Power (Averages_all_Cycles)
y = 4E-05x3'1237
 R2 = 0.9998     A
                            10
                                  15
                                       20
                                             25    30
                                           Speed (mph)
                                                         35
                                                               40
                                                                     45
                                                                           50
5.3.3  TRAKERII

Data  alignment and aggregation  for TRAKER  II were  conducted almost identically as for
TRAKER I. Starting with all valid 1-second data, the GPS time was retarded by 3 seconds and
then re-associated with DustTrak  data.   The TRAKER II 1-second signal was calculated with
equation 1 for each valid data point.  Only data associated with measurements on the test route
were  considered for further analysis and each of those data points was linked with a Pass ID.
Pass-average values were calculated from the 1-second data points for further data analysis.

As with TRAKER I, some differences were evident between the TRAKER II signal  averaged
over an entire pass and the signal averaged only over data points corresponding to measurements
within 50 m of the master tower (Figure 5-12).  Also, the concentrations measured in the left and
right  tire  inlets were  always  substantially  higher than  those measured  in the background
(Figure 5-13)

Examination of the ratio of the right to left inlet signals  indicated that overall, the signal from the
left side was higher than the signal from the right side (Figure 5-14).  Unlike the near-unity
values for Sets 4 to 8  exhibited by the TRAKER I data (Figure 5-6), TRAKER II data suggest
that for almost all passes, the signal from the left side was higher than from the right (less than
unity ratio).  Moreover, the ratio is much more variable for TRAKER II.  There are several
possible reasons for this.  First, the cargo bay in TRAKER II was heavily loaded on the left side
with tools and equipment with an approximate mass of 300 kg.  This may have resulted with a
higher signal on the left.  Second, the inlets for  TRAKER II  are further behind the tire than
TRAKER I, resulting perhaps  in a generally noisier  signal.   Third,  the  signal values from
TRAKER II were consistently lower than TRAKER I, also perhaps contributing to greater noise
                                                                                     58

-------
in the ratio for right to left inlet signals. Fourth, TRAKER II was operated by a different driver
than TRAKER I and it is possible that small differences in the  paths that the vehicle tires
followed could have caused higher signals on the left side of TRAKER II compared to the right
side.  The difference between the left and right signals in TRAKER II deserves further attention
in future work.  However, for the present study, we note again that the road dust emissions from
the  TRAKER II will be a combination of the emissions  from the left  and right sides of the
vehicle  and that it is appropriate to apply Equation 1 to  obtain a representative TRAKER II
signal for the whole vehicle.

Finally,  Figure 5-15 shows the speed response of the TRAKER II signal (same as Figure 5-11
for  TRAKER I).  The relationship between speed and TRAKER II signal is close to the cubic
relationship (exponent of speed term is 3.3 according to regression) exhibited  by TRAKER I.
 Figure 5-12.  Relationship Between TRAKER II Signal Averaged Over Entire Pass (route
           length) and TRAKER II Signal Only Within 50 m of Master Tower
                 1000
               H
               "S
                  100 -
                   10 -
                   1 -
                  0.1 -
                 0.01
y = 0.9619x
R2 = 0.858
                    0.01
                             0.1        1        10        100

                                Run-averaged TRAKER II signal (mg/m3)
                                                                  1000
                                                                                     59

-------
      Figure 5-13.  TRAKERII Ratio of Average of Left and Right Tire Inlet
           Concentrations to Background (middle) Inlet PMio Concentration
                  10000
                  1000
                   100
                    10
                      ft
                                    *!


                              t 1 1
                              I ' '
                             » ,1 1
1
                                                                > t
                                                                .v
1 J at  »I r T '  '.	
\* Ik  ;  ' t I i  |i
if  \ | ! | \ |t l|
i   ^V  ]/VfU  ;
      i     Wf
                          r
                          50
                               100
                                    150
                                         200
                                              250
                                             Pass ID
                                                    300
                                                         350
                                                              400
                                                                   450
                                                                        500
Figure 5-14.  TRAKER II Time Series of Ratio of Pass-Averaged Right to Left Inlet Signal
Ratios and Pass-Averaged Wind Speed (m/s).  Vertical lines represent times when the road
 was swept and silt was applied, while double vertical lines represent times  when the road
     was swept only. Numbers at the top correspond to different measurement sets.
             ?  i
                         Ratio of Right to Left Inlet Signals
                                                 -«- Tower ave Wind Speed ms
                                   Ivr
                                         =jx
          XX

          1UJ

                           100    150    200    250    300    350    400    450    500

                                         Pass ID
                                                                                   60

-------
Figure 5-15. Speed Response of TRAKERII Signal. Figure Shows the TRAKERII Signal
  at Each Speed Normalized to the Average Signal at 25 mph in the Same Cycle.  Data are
        shown for four consecutive cycles as well as the average value for all cycles.
              10
               9 -
               7 -
               6 -
            S -5 -
•a  41
s
.2  3
c
               H
        « Cyclel
        a Cycle2
        A Cycle3
        x Cycle4
        • Averages_all_Cycles
       	Power (Averages_all_Cycles)
                                                  y = 2E-05x"z8
                                                   R2 = 0.9708
                                                         X

                                                        't
                           10
                                 15
                                       20
                                             25     30
                                          Speed (mph)
                                                         35
                                                              40
                                                                    45
                                                                          50
6.0 RESULTS

The test types, times,  vehicles involved,  and number of vehicle passes are summarized in
Table 6-1.  The total numbers of traversals through the test course for each test vehicle were:
TRAKER I - 154, TRAKER II - 152, SCAMPER - 162. The distribution of nominal speeds at
which measurements were conducted was approximately: 25 mph - 24%, 35 mph - 47%, and 45
mph - 29%. Except for the first two sets of measurements, where test vehicles  traversed the test
course  in both directions, vehicles traversed the course from the  eastern  end of Veterans
Memorial Highway by the Command Center (CC) towards the west/northwest.
Results  from  AP-42  silt measurements  and the three mobile systems  used in  this study
(TRAKER  I, TRAKER H, and SCAMPER)  are discussed in  individual sections  below.   In
addition to providing data summaries, those sections assimilate the different methods for  road
dust emission estimation with horizontal PMio tower flux data.  In the case of AP-42 silt
sampling, this provides a basis for comparing the  AP-42 methodology to  emission  factors
measured on-site.  In the cases of the mobile systems, the horizontal flux measurements which
represent an  independent measure  of PMio  emission  factors  are used to  calibrate the three
systems used as part of this study.

One important finding deserves discussion here since it  applies to all the mobile systems as well
as the AP-42 silt  sampling.  It was noted when examining the time series of the  tower flux,
TRAKER I, TRAKER II, SCAMPER, and even the AP-42 silt measurements that the application
of silt material to the test road section led to an initial surge in PMio emissions.  This  can be  seen
in Figure 6-1 where the pass-averaged time series for all of these data sets are  plotted.  Starting
with measurement Set  3  - the first instance when silt  was applied to the test road - the first
several passes in the set exhibit comparatively very high road dust emissions or mobile  system
                                                                                    61

-------
raw  signals.   Subsequently,  emissions  begin to stabilize  at a  lower though not necessarily
constant value.  Measurement Sets 12 and 13 deviate somewhat from this pattern because during
those sets, the travel speeds of the test vehicles were varied over the course of the Sets.  We alert
the reader at this time that for comparing the signals from the mobile systems to those measured
on the horizontal flux tower, the first 9 vehicle passes will not be  considered for sets where road
silt material was applied to the  surface. The justification for this was provided in  an earlier
section (Section 5.3.2).

           Table 6-1. Summary of Tests During Field Study (9/11/06 - 9/15/06)
Q
"S
X
1

2

3


4


5


6


7


8


9
10

11


12


13

v
«
Q
9/11
9/11
9/11
9/11
9/11
9/11
9/12
9/12
9/12
9/12
9/12
9/12
9/13
9/13
9/13
9/13
9/13
9/13
9/13
9/13
9/13
9/14
9/14
9/14
9/14
9/14
9/14
9/14
9/14
9/14
9/15
9/15
9/15
Approximate Time (local)
11:55-13:15
13:35
13:52-14:18
14:30
15:17-26:30
17:00
9:15
10:15-11:00
11:05
13:00
13:35-14:40
15:00
9:00
9:40- 10:25
11:09
12:15
12:45-13:35
14:00
14:45
15:20-16:15
17:00
8:00
8:40-9:20
9:20-9:50
10:05
10:25-11:20
11:30
12:30
13:10-14:05
14:30
8:00
8:30- 11:15
11:30
>-,
"jj-
u
<(
Test: Baseline road conditions - No Sweep,
No silt
Sweep
Test: After Sweeping, No silt applied
Silt applied to test road
Test: After application of silt, 35 mph
Sweep
Silt applied to test road
Test: After application of silt, 45 mph
Sweep
Silt applied to test road
Test: After application of silt, 25 mph
Sweep
Silt applied to test road
Test: After application of silt, 45 mph
Sweep
Silt applied to test road
Test: After application of silt, 25 mph
Sweep
Silt applied to test road
Test: After application of silt, 45 mph
Sweep
Silt applied to test road
Test: Depletion of silt resulting from
vehicle passes
Test: Measure emissions prior to sweeping
Sweep
Test: Measure emissions after sweeping
Sweep
Silt applied to test road
Test: Speed tests
Sweep
Silt applied to test road
Test: Speed tests
Sweep
Vehicles used
All test vehicles
Street Sweeper
All test vehicles
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
SCAMPER Only
All test vehicles
Street Sweeper
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Tractor/spreader
All test vehicles
Street Sweeper
Nominal speed (mph)
35
NA
35
NA
35
NA
NA
45
NA
NA
25
NA
NA
45
NA
NA
25
NA
NA
45
NA
NA
35
35
NA
35
NA
NA
25-45
NA
NA
25-45
NA
Total passes/passes per vehicle
60/20
NA
30/10
NA
27/9
NA
NA
30/10
NA
NA
42/14
NA
NA
30/10
NA
NA
30/10
NA
NA
36/12
NA
NA
10/10
12/4
NA
30/10
NA
NA
27/9
NA
NA
84/28
NA
                                                                                      62

-------
    Figure 6-1. Time Series of Pass-Averaged Horizontal Tower PMio flux (g/vkt), Silt-
 estimated AP-42 Emission Factor (g/vkt), TRAKER I, TRAKERII, and SCAMPER raw
    signals (mg/m3). Vertical lines represent times when the road was swept and silt was
     applied, while double vertical lines represent times when the road was swept only.
              Numbers at the top correspond to different measurement sets.
1
~5fc
      I
          1000
          100
          0.01
         0.001
                  - - Tower 1 Flux (g/vkt)	TRAKER I
                    50     100     150    200     250     300
                                             Pass ID
                                                           350
                                                                  400
                                                                         450
                                                                               500
6.1  Short-Term Emission Factor Decay and Silt Loading Depletion

6.1.1  Silt Loading Depletion

Figure 6-2 shows a typical pattern of silt loading depletion for Set 12, at a low initial applied silt
loading of 0.6 g/m  depleted at cyclically varying vehicle travel speeds of 25, 35, and 45 mph.
Silt loading undergoes a rapid decay to about for the first nine passes, and then stabilizes at a low
constant value that is about one-third of the initial value.
                                                                                     63

-------
                 Figure 6-2.  Silt Depletion With Increasing Vehicle Passes
                Sept 14 - Set 12, Nominal applied silt loading - 0.6 gram/m , Varying vehicle speed
      2.50
      0.50
      0.00
        360
               365
                     370
                           375
                                  380
                                        385    390     395
                                             Pass ID
                                                           400
                                                                 405
                                                                        410
                                                                              415
                                                                                    420
This  pattern  was  observed  in five of the  nine data  sets  for  which sufficient  silt  loading
information is available. Results are summarized in Table 6-2.
  Table 6-2. Summary of Observed Silt Decay With Increasing Number of Vehicle Passes
Set
4
5
6
7
8
9
10
11
12
13
Initial
Loading
(gram/m2)
2.5
2.3
0.6
0.5
0.7
0.7
0.3
0.2
0.6
1.1
Vehicle
Speed
(mph)
45
25
45
25
45
35
35
35
varying
varying
Decay First
Nine Passes?
N/A
Yes
No
Yes
Yes
Yes
Yes
No
Yes
Yes
Ratio Last Pass Avg./ first
9 Pass Averages

0.55
0.87
0.89
0.41
0.63
0.57
2.11
0.47
0.23
Comments
decay observed, but only 2
data points


first 6 passes


9 passes total

Strong cross winds at end of
experiment
Strong cross winds
throughout experiment
                                                                                            64

-------
A comparison of AP-42 Emission factors computed separately for the first nine passes and for
the remaining vehicle passes (Figure 6-3) shows that AP-42 emission factor values for the first
nine passes, were (with the exception of Run 11) higher than values for the remaining passes.

The rapid decay  in silt loading  over the first  few passes lends support to the  DRI/UCR
hypothesis that two separate  mechanisms, aerodynamic (first  nine  passes)  and mechanical
(subsequent passes) may be responsible for suspending PMio from paved road surfaces.

 Figure 6-3.  Comparison of Averaged AP-42 Emission Factors, in gram/VMT, Computed
    From Silt Loadings for First Nine Passes, Compared to AP-42 Emission Factors for
                                   Remaining Passes

                      Phase IV - Average AP42 EF +/-1 standard deviation vs
                                     Set number
                                     D initial 9 passes D remaining passes
9.0-
8.0-
7.0-
6.0-
5.0-
4.0-
3.0-
2.0-
1.0-






I

fl



f


















.
-




•



ft


fi


123456







fl










!


t.

fl


1







789
Set number



ft




L J
TF




rFi
I

10 11



i

12








t
i

13
     *Error bars are ± one standard deviation.

Signals from the mobile technologies systems also showed high initial decay within several
experimental sets. Figure 6-4 compares TRAKER I signal to AP-42 silt over all observed
experimental runs.
                                                                                     65

-------
 Figure 6-4.  Comparison of TRAKERI Signal and Average North-South Silt Loading for
                                   All Vehicle Passes
            1000
         I
            0.01
                           100
                                 150
                                       200     250    300
                                            Pass ID
                                                          350
                                                                400
                                                                       450
                                                                             500
The TRAKER I signal decay with vehicle passes matches AP-42 silt loading decay in Sets 5, 8,
and 10 for cases of constant vehicle  speed. However, TRAKER I measured emissions also
showed, in sets 12 and 13, clear vehicle travel speed dependence that are not accounted for in the
current AP-42 emission factor equation. The rising and falling TRAKER I signals in Sets 12 and
13 are a result of systematically varying vehicle speeds first rising from  25 to 35 to 45 mph, then
declining from 45 to 35 to 25 mph. Silt loadings in Set 12 declined throughout the experiment,
even though TRAKER I emissions increased with increasing vehicle speed. Silt loadings in Set
13 declined rapidly to a steady state value, while TRAKER I emissions  fluctuated regularly with
rising and falling vehicle speed.

TRAKER II and SCAMPER signals showed similar behavior. The SCAMPER signal is plotted
alongside silt loading in Figure 6-5.
                                                                                    66

-------
 Figure 6-5.  Comparison of SCAMPER Signal and Average North-South Silt Loading for
                                   All Vehicle Passes
           S  o.l
             0.01
                      50
                            100
                                   150
                                         200    250     300
                                              Pass ID
                                                                   400
                                                                         450
                                                                                500
The SCAMPER signal tracks decay in AP-42 silt loading with vehicle speed in Sets 5, 8, and 10
for cases of constant speed. However, just as in the case for TRAKER I, SCAMPER measured
emissions showed,  in  sets 12  and  13, clear vehicle  travel  speed dependence that are not
accounted for in the current AP-42 emission factor equation.

6.2 Comparison of Horizontal Flux Tower Emission Factors to EPA Method AP-42

Plumes from  point and  line  sources  are  often  modeled  as  exhibiting  smooth, Gaussian
concentration distributions. This type of representation has been adequate over long spatial and
time scales, where a dispersive  force from random turbulent eddies are allowed to proceed for
long periods and average out.  In practice, individual, non-steady plumes such  as from a point
puff or a moving line source  are quite erratic and the  instantaneous  spatial distribution of
concentration does not at  all resemble a Gaussian profile.  Furthermore, owing to the random
nature of plume dispersion, the  flux measured at a point in space is likely to vary considerably
from one event (e.g. passage of a vehicle) to the next.  This can be seen in Figure 6-6 where
individual tower flux measurements associated with the passage of the test vehicles are plotted.
The figure (Note log y-axis scale) shows that individual flux measurements exhibit substantial
pass-to-pass variability.
                                                                                     67

-------
 Figure 6-6. Time Series of Horizontal PMio Fluxes Measured With Tower Measurement
                           System for Different Test Vehicles
          1000
          0.01
         0.001
                   50
                         100
                                150
                                      200
                                            250

                                           Pass ID
                                                   300
                                                               400
                                                                      450
                                                                            500
The inherent variability of tower flux measurements requires that data be aggregated (averaged)
over several replicate measurements in order to filter out some of the measurement noise.  In the
case of the present study, this poses a slight challenge because the road dust loading on the test
road was not constant over the course of the field study and indeed was changing over the course
of a single set of measurements.  This can be seen quite clearly in (See Figure 6-2) where, as the
number of vehicle passes within a measurement set increases, the signals from the three mobile
systems decrease, indicating decay in road  dust loading over time.  (Please refer to Figure 5-8
and Figure 5-9 in Section 5.3.2) The observed decay pattern suggests  that there are two modes
for this decay.  During the first several vehicle passes after silt is applied to the surface, road dust
loading appears to diminish quickly.  Earlier, we termed this "aerodynamically suspendable"
road dust. After 9 or so vehicle passes, the road dust loading decreases much more slowly as the
"mechanically suspendable" material is all  that remains on the test road surface.  As discussed
earlier, for the purpose of reporting emissions from the different test vehicles used in this study,
we consider only the horizontal PMio fluxes for times when the number of vehicles passing over
the road after silt application was greater than 9 (Note that this does not affect Sets 1  and 2  when
silt was not applied to the  surface).  This serves to both  mitigate the  large range of emissions
factors that were measured (if first 9 passes are included) as well as separate the "mechanically
suspendable" road dust from the "aerodynamically suspendable" road dust - the former being
more likely to prevail on well traveled roads.

The average horizontal fluxes (emissions) by  measurement set, and test vehicle are reported in
Table  6-3.  With some  set-to-set variation in the emissions magnitude, in general  all  three
vehicles exhibit approximately the same emissions within  the standard  error of the measurement
set. If averaged over all valid horizontal flux measurements, mechanically suspended PMio dust
                                                                                      68

-------
fluxes  are 4.1 ± 0.7,  5.0 ±  1.2, and 5.0 ± 2.0 g/vkt for TRAKER I, TRAKER II, and UCR
SCAMPER, respectively - not statistically significant differences.

 Table 6-3. Summary of Measured PMio Horizontal Fluxes. Data shown are averages for
    all passes following the ninth pass after silt application. Standard errors shown are
based on the standard deviation divided by the square root of the number of measurements
                                included in the average.
Set
1
2
3
4
5
6
7
8
9
10
11
12
13
TT?T
..^, TRI TRI
Valid
Flux ave Std err
Count (g/vkt) (S/vkt)
7 1.32 0.62
5 1.53 0.58
13 3.04 1.50
7 5.53 1.19
6 10.53 6.13
2 2.13 1.04
1 2.05 NA
8 6.40 1.48
NA NA NA
4 1.18 1.11
5 3.70 3.78
3 0.40 3.08
9 5.47 2.25
TRTT
,yr. TRII TRII Std
Valid
Flux ave err
Count ^ (S/vkt)
4 1.66 1.28
5 0.94 0.70
12 1.91 1.05
7 9.44 6.39
8 4.51 1.40
2 7.90 5.88
1 0.29 NA
5 4.57 0.69
NA NA NA
4 2.94 0.43
5 4.53 3.32
3 1.31 0.63
10 11.03 6.26
UCR UCR
Valid Flux
„, Std err
Flux ave
Count (g/vkt) (g/Vkt)
7 0.59 0.36
5 0.72 0.65
12 1.89 0.42
7 11.64 4.51
8 2.23 0.41
1 0.99 NA
2 3.24 2.86
4 3.02 1.27
1 11.04 NA
4 3.32 3.60
4 6.46 6.24
2 2.28 0.50
9 13.88 5.77
vArH if Allstd
Valid Flux
T-I err
Flux ave
Count (g/vkt) (g/Vkt)
18 1.11 0.38
15 1.06 0.36
37 2.30 0.63
21 8.87 2.57
22 5.32 1.80
5 4.21 2.42
4 2.21 1.36
17 5.07 0.82
1 11.04 NA
12 2.48 1.18
14 4.79 2.33
8 1.21 1.07
28 10.16 2.96
"NA" indicates that either there were no valid flux measurements during the indicated period or silt was not applied
   to test road prior to measurements.

Silt measurements were conducted at various points in time over the course of measurement sets
and within measurement sets (Please refer to Table 6-1).  Full silt sampling (as  opposed to
"quickie strips") was primarily conducted at the beginning and end of measurement sets. The silt
sample procured at the beginning of measurement sets where silt was applied to the road contain
significant  fractions of  "aerodynamically suspendable" road  dust.   This can be seen  in
Figure 6-1 in Section 6.0 where it is clear that at the beginning of those measurement sets, the
rate of decay of silt loading is high compared to later periods (i.e. after the first 9 vehicle passes).
Silt samples procured at  the end of those  measurement  sets represent, in principle, the lowest
emission factors  of the  measurement set.   Referring again to  Figures 5-8  and  5-9  in
Section  5.3.2, the rate of decay of "mechanically suspendable"  road dust is much lower than that
of "aerodynamically suspendable" road dust.  If for the purposes of the present effort, we accept
the decay  rates shown for mechanically suspendable" road dust (Figures 5-8 and  5-9  in
Section  5.3.2), namely an exponential decay of-0.029 X, where X is the number of passes since
silt application, then the  difference in mechanically suspendable  road dust between X=10 and
X=25 is about a factor of two. Considering that PMio emission fluxes from consecutive passes
can vary by an order of magnitude or more (Figure 6-6), the error introduced by assuming that
the silt  sample  procured at the end  of the measurement  set represents  all passes where
"mechanically suspendable" road dust was dominant  (i.e. from > 9 passes after silt until set
completion) is acceptably small.
                                                                                      69

-------
To compare PMio tower flux measurements with AP-42 silt methodology and mobile system
measurements,  data  were  averaged by  measurement set.   For  each  set all  tower  flux
measurements were  averaged together  regardless of  the  test  vehicle.   Thus,  tower  flux
measurements represent average fluxes for all vehicles.  This was to ensure that  all methods
examined would be calibrated (or compared in the case of AP-42) against the same standard and
results from  future  measurements  can be compared  using  a common basis.   In examining
Table 6-3 (three rightmost  columns), it is clear that the number of valid flux  measurements
varied from set to set.  A minimum  criterion of 10 valid vehicle passes was applied  to the tower
flux  average value.  This invalidated sets 6, 7,  9, and  12. In addition, data from  set 13 were
considered invalid because wind speeds were very high during that period and neither the mobile
systems nor the tower flux  measurement system measurements are trustworthy at  high winds.
The remaining valid sets for comparison were 1, 2, 3,  4, 5, 8, 10, and 11.  These measurement
sets were used to compare AP-42 silt-based emission factors estimated from the AP-42 emission
factor equation (See Section 5.2.2 for full equation) to PMio emission factors measured with the
horizontal flux tower. Silt measurements at the end of a set were available for Sets 3-13. Thus,
the measurement sets that remained for comparison between the AP-42 methodology and the
tower data were 3, 4, 5,  8,  10, and 11.

Comparison of AP-42 silt based emission factors and  set-averaged PMio emission factors are
shown in Figure 6-7. The solid line in the Figure represents  a least-square linear fit to the data
with a zero intercept while the dashed line represents a power  law  fit.  The power law fit appears
to accommodate the data better than the linear fit (R2 =  0.33 compared to -0.22). In general, AP-
42 estimated emission factors appear to be substantially lower than measured tower-based
emission factors for all measurements sets by about 40%.
                                                                                    70

-------
    Figure 6-7. Tower-Based PMio Emission Factors versus AP-42 Silt-Based Emission
   Factors. Solid squares represent emission factors that are averages of all valid tower
  measurements for sets 3, 4, 5, 8,10, and 11. AP-42 data shown are averages of the north
  and south sample measurements procured at the end of the measurement sets. The solid
 line in the Figure represents a least-square linear fit to the data with a zero intercept while
  the dashed line represents a power law fit. A one-to-one line is included in the Figure for
     comparison. X and Y error bars represent standard errors which are based on the
 standard deviation of individual measurements within the measurement Set divided by the
           square root of the number of measurements included in the average.
         1000
      w
          0.1
             0.01
0.1
    1
AP-42 g/vkt
10
100
We hypothesize in Section 7.2 that an altered distribution of freshly applied road silt on a low
roughness experimental road surface increased mobile PMio emission factors compared to AP-42
     emission factors.
(1) On the Phase IV road  surface,  soil was freshly-applied and had not yet been swept by
repeated vehicle passes into the "pits" between asphalt-embedded aggregate "protrusions," as
would occur on normally traveled road surfaces. As a result,  for the same silt loading, a greater
proportion of the freshly applied silt would be located on the "protrusions" of the road surface,
and would be less sheltered  from conditions of applied mechanical or aerodynamic shear than is
the case for a well-traveled road where road silt has been generated by natural processes.

(2) The road surface used in this experiment was recently paved, is very smooth, and is in better
condition than the normally traveled road surfaces studied in  earlier phases of this project.  The
road surface "pits" were therefore shallower and the silt that is deposited in the valleys would be
less sheltered than would normally be the case on a well-traveled road with silt generated by
natural processes.
                                                                                    71

-------
The combined effects of 1) and 2) are to make the freshly-applied PMio on the experimental road
more "exposed" to suspension during conditions of mechanical vehicular shear and moderate
vehicular aerodynamic than the amount of more "sheltered" PMio mobilized into the air from a
normally traveled, rougher typical road surface.

Compared to the moderate  shears developed by vehicles, vacuum cleaners apply much higher
shears during AP-42 silt recovery (Bettancourt Rodriguez, 2006). Silt recoveries of greater than
99% were observed after four vacuum cleaner head passes (Rodrigues, 2006) on both smooth
and rough road  surfaces. As a result, both silt recoveries and calculated AP-42 PMio emissions
factors would not be  as  sensitive  to silt  distribution or road surface condition as  mobile
technologies emission factors.

When simultaneously  measuring AP-42 emissions factors and mobile  technologies  emission
factors that are sensitive to roughness and silt spatial distribution on a smooth road with freshly
applied silt, we  hypothesize that, compared to what would be observed on a well-traveled road,
mobile technologies PMio emissions factors would increase relative to AP-42 emissions factors.

Recommendations of experiments that could be performed to test this hypothesis are proposed in
Section 8.2

6.3 Comparison of Horizontal Flux Tower Emission Factors to Mobile Technologies
    Emission Factors

6.3.1 TRAKERI

Figure 6-8 shows the pass-averaged PMio emission factor measured by the tower system and the
pass-averaged TRAKER signal for  cases where both data sets were valid.  Overall, the flux
measurement and the TRAKER signal track reasonably well, though on a point-to-point basis,
the relationship between the two measurements is somewhat noisy.  To compare PMio tower flux
measurements with AP-42 silt methodology  and mobile system  measurements, data were
averaged by  measurement  set.  For each set  valid,  tower flux measurements for all passes
excluding  the first 9 following silt  application were averaged together regardless of the test
vehicle.  A minimum criterion of 10 valid vehicle passes was applied to the tower flux average
value. This invalidated sets 6,  7, 9, and 12. In addition, data from set 13 were  considered invalid
because wind speeds were very high during that period.  The remaining valid sets for comparison
of TRAKER signal to PMio flux were 1, 2, 3, 4, 5, 8, 10, and 11.
                                                                                     72

-------
 Figure 6-8. Time Series of Measured Horizontal PMio Flux on the DRI Tower System and
 the Pass-Averaged TRAKERI Signal for Passes When the Horizontal Flux Measurement
                                       was Valid
             1000
            0.001
                      50
                            100
                                   150
                                         200    250    300
                                              Pass ID
                                                            350
                                                                  400
                                                                        450
                                                                              500
Comparison of set-averaged TRAKER I data and set-averaged PMio emission factors are shown
in Figure 6-9. The solid line in the Figure represents a least-square linear fit to the data with a
zero intercept while the dashed line represents a power law fit.  The power law fit appears to
accommodate the leftmost data point better than the linear fit, though we note that the linear fit
provides a  better R2 value (0.57 compared to 0.48).  However,  it is unknown whether the
leftmost data point is an outlier. The white squares also shown in the figure were collected on a
road near Lake Tahoe, California as part of an earlier study (Kuhns  et al., 2004). Whereas these
earlier data are not fully comparable owing  to a slightly different field setup, they tend to
indicate that the  linear fit (or a near-linear fit)  to  the  data from the present study is more
reasonable  than the power law fit which exhibits  an exponent of  0.38.  Of course,  without a
mechanistic understanding of the road dust emission process, there is no a  priori  reason to
anticipate a specific form for the equation that best represents a calibration  of TRAKER I.  In the
absence of further information, we assume for simplicity that the TRAKER I signal is related to
      emission factors through the simple linear relationship:
                                     EFW =0.54xT
                                                           Equation 6.1
where:  EFio =

        T   =
the PMio mass emission factor from the tower data for all the vehicles used as
test vehicles in the present study, and
the TRAKER signal defined simply as the background corrected average of
the concentrations measured behind the left and right tires (Equation 5.3).
                                                                                     73

-------
 Figure 6-9. PMio Emission Factors versus TRAKERI Average Signal. Solid squares are
  data from the present study and represent emission factors that are averages of all valid
    tower measurements for sets 1, 2, 3, 4, 5, 8,10, and 11.  TRAKER I data shown are
averages of TRAKER I passes during the respective set. Averages include only passes after
the ninth pass following silt application for sets when silt was applied to the test road.  The
   solid line in the Figure represents a least-square linear fit to the data from the present
   study with a zero intercept while the dashed line represents a power law fit. The white
  squares are data collected during an earlier study near Lake Tahoe, California. X and Y
    error bars represent standard errors which are based on the standard deviation of
  individual measurements within the measurement Set divided by the square root of the
                   number of measurements included in the average.
        1000
         100 -
      w
            0.01
0.1               1
          TRAKER I (mg/m3)
10
100
6.3.2 TRAKERII

Figure 6-10 shows the PMio horizontal fluxes and the TRAKER II signal averaged by pass when
both measurements were valid.  As with the TRAKER I data, the two measurements tend to
follow  each other, though not consistently owing to the noise  that  is inherent  to  both
measurements,  especially the tower fluxes.   As with  the  TRAKER  I data,  to obtain  a
correspondence between tower measured PMio emission factors and the TRAKER II signal, we
compared set-averaged tower data to set averaged TRAKER II signal. Only Sets with at least 10
valid tower measurements  corresponding to "mechanically  suspendable"  road dust (i.e. more
than 9 passes after silt application) were considered (1, 2, 3, 4, 5, 8, 10, and 11).  Although the
TRAKER II data for pass IDs lower than 170 were considered of suspect  validity because of a
malfunction in the  inlet  flow control, they have been included in the comparison shown in
Figure 6-11. If not included, only  a few points for comparison would be available. Thus, the
                                                                                 74

-------
relationship between the TRAKER II signal and the PMio emission factors should be considered
preliminary.
       	                             	                                        r\
Unlike TRAKER I, the power law fit for TRAKER II provides a substantially higher R  value
than the simple linear fit (0.90 compared to 0.75). It would be interesting as additional research
becomes available to re-examine the relationship between  the TRAKER II  signal and  tower
measured emission factors.  For the purposes of comparison with TRAKER I and SCAMPER
(below), we propose to use the same simple linear form that was presented for TRAKER I in
Equation 1 above, namely,
                                   EFW = 0.92xTn
                       Equation 6.2
where 7}/ is the TRAKER II signal.
 Figure 6-10. Time Series of Measured horizontal PMio flux on the DRI Tower System and
 the Pass-Averaged TRAKER II Signal for Passes When the Horizontal Flux Measurement
                                      was Valid
         1000
         100
          10
          0.1
         0.01
                     ^ - - Tower 1 Flux (g/vkt)
-TRAKER II Signal (mg/m3)
                                                  10
                                                      11
                  50
                        100
                              150
                                    200    250    300
                                         Pass ID
                                                      350
                                                            400
                                                                  450
                                                                         500
                                                                                  75

-------
  Figure 6-11. PMio Emission Factors versus TRAKERII Average Signal. Solid squares
represent emission factors that are averages of all valid tower measurements for sets 1, 2, 3,
 4, 5, 8,10, and 11. TRAKER II data shown are averages of TRAKER II passes during the
 respective set. Averages include only passes after the ninth pass following silt application
  for sets when silt was applied to the test road. The solid line in the Figure represents a
 least-square linear fit to the data with a zero intercept while the dashed line represents a
   power law fit. X and Y error bars represent standard errors which are based on the
 standard deviation of individual measurements within the measurement Set divided by the
          square root of the number of measurements included in the average.
        1000
      w
          0.1
            0.01
6.3.3  SCAMPER
0.1               1

         TRAKER II (mg/m3)
10
100
Figure  6-12  shows the time  series  of pass-averaged  net  (rear - front  DustTrak signal)
SCAMPER signal and PMio horizontal flux measurements when both types of measurements
were valid.  As with TRAKERs I and II, the UCR SCAMPER follows the general trend of
emission factors captured by the tower system.  For comparing the SCAMPER signal to PMio
emission factors measured by the tower, only Sets with at least 10 valid tower measurements
corresponding  to "mechanically suspendable" road  dust  (i.e. more  than 9  passes after silt
application) were considered (1, 2, 3, 4, 5, 8, 10, and 11).  Set averaged PMio emission factors
are plotted against set-averaged SCAMPER signal in Figure 6-13.  As with TRAKER I and
TRAKER  II, we show both a linear fit and a power law fit in the Figure.  Similar to TRAKER I,
there was  no benefit in terms  of R2 values in a power law fit (0.40) over a linear fit (0.47).
Assuming  a linear relationship between PMio emission factors and the SCMAPER signal, the
following empirical equation can be used to relate the two quantities:
                                    EFW = 20xSC
                                             Equation 6.3
                                                                                  76

-------
where SC is the SCAMPER signal.

In the SCAMPER the net signal is multiplied by the frontal area of the tow vehicle (maximum
height * maximum width), 3.66 and the DustTrak "calibration factor". The later is determined
from PMio filter sampling collocated with the rear-mounted DustTrak. Due to a leak in the PMio
sampler during this  study, we did not determine a calibration factor. In previous  studies
conducted in Clark County  NV and in Maricopa County AZ  the average factor  has been
measured as 3.4 with an estimated uncertainty of 1. Therefore  the emission factor based on this
method is given by:

                                    EFjo = 12 x SC                         Equation 6.4

This is within a factor of two of the value determined by the tower measurements and given the
scatter in both data sets, they are in reasonable agreement.

It is interesting to note the multipliers for the different mobile  systems that are needed to obtain
the same emission factors (Table 6-4), especially in  the context  of the distance of the mobile
measurement from the  road dust source.  The inlets of TRAKER I are located closest to the
vehicle's front tires.  In TRAKER II, the distance between the  inlet and the vehicle front tires is
almost twice that of TRAKER I. For SCAMPER, the distance between the "influence" DustTrak
mounted on the trailer behind the  vehicle and the vehicle  tires is more than an order of
magnitude that of TRAKERs I and II. These simple observations  suggest that the differences in
the signals from these three mobile systems are closely related to the distances  between where
the "influence" measurement is  taken compared to the  locations of the tires.

  Table 6-4. Summary of Equivalence Multipliers Between Mobile Measurement Systems
    and PMio Emission Factors Assuming that the Raw Signal for the Mobile Systems is
                    Linearly Related to Measured Emission Factors
System
TRAKER 1
TRAKER II
SCAMPER
Raw Signal (mg/mj) Multiplier to get PM10
Emission Factor (g/vkt or g/vmt)
0.54 (0.86)
0.92(1.5)
20 (32)
                                                                                    77

-------
Figure 6-12.  Time Series of Measured Horizontal PMio Flux on the DRI Tower System and
 the Pass-Averaged SCAMPER Signal for Passes When the Horizontal Flux Measurement
                                       was Valid
           1000
            100
            10
            o.i
         I °-01
          0.001
                         ^ - - Tower 1 Flux (g/vkt)
-UCR Signal (mg/m3)
                                                              12
                                                                      13
                    50
                          100
                                 150
                                       200
                                             250     300
                                            Pass ID
                                                                400
                                                                       450
                                                                             500
                                                                                      78

-------
  Figure 6-13.  PMio Emission Factors versus SCAMPER Average Signal.  Solid squares
represent emission factors that are averages of valid tower measurements for sets 1, 2, 3, 4,
  5, 8,10, and 11. SCAMPER data shown are averages of SCAMPER passes during the
 respective set. Averages include only passes after the ninth pass following  silt application
  for sets when silt was applied to the test road. The solid line in the Figure represents a
 least-square linear fit to the data with a zero intercept while the dashed line represents a
   power law fit. X and Y error bars represent standard errors which are  based on the
standard  deviation of individual measurements within the measurement Set divided by the
          square root of the number of measurements included in the average.
          1000
           100
        •59
        w
            10
            0.1
                               I   I  I
              0.01
0.1               1

          SCAMPER (mg/m3)
10
100
6.4 Comparison of Calibrated Mobile Technologies Emission Factors to EPA Method AP-
    42 Emission Factors to measured PMio Horizontal Flux Tower Values

Figure 6-14 shows a time  series comparison of pass-averaged emission factors using the five
different methods.  The Figure shows direct PMio horizontal flux measurements with the tower
system, emission factors estimated from silt measurements and use of AP-42 equations, and
calibrated  emission factors from the  three mobile  systems,  TRAKER I, TRAKER II, and
SCMAPER. The mobile system emission  factors are calculated by multiplying the respective
pass-averaged signals (in mg/m3) by the appropriate calibration factors discussed in Section 6.3
(Equations 1 - 3). The Figure illustrates how well the mobile systems track one another and to a
lesser extent, the horizontal flux tower measurements.  It also shows that the silt-based AP-42
method tends to underestimate the measured emission factors and does not respond to changes in
emission factors that appear to be related to vehicle speed (see for example the speed test cycles
in Set 13 measurements).
                                                                                  79

-------
    Figure 6-14. Emission Factors (g/vkt) For All Valid Passes.  Tower data are direct
  measurements, AP-42 data are based on silt measurements and use of AP-42 equations,
 SCAMPER, TRAKER I, and TRAKERII data are based on the regression between those
    mobile systems and measured PMio tower fluxes (using Equations 1-3 in Section 6.).
           1000
                                              SCAMPER      TRAKER I   - - - TRAKER II
          0.001
The current approved AP-42 PMio emission factor equation does not include speed as a factor in
estimating PMio emissions.  The  equation assumes an equilibrium  silt  loading, SL, that is
determined by rates of removal by mechanical and aerodynamic shear that are opposed by rates
of creation and deposition from road, brake  and tire wear,  and atmospheric  and hydrologic
transport and vehicle track-out.  Equilibrium  silt loadings are known to be lower on roadways
with higher average daily traffic (ADT), and higher ADT's are usually accompanied by higher
average speeds.

In this experiment, freshly applied silt on the road surface was not in equilibrium, and was
progressively depleted by successive vehicular passes. Rapid depletion was observed in both the
first 9 passes of the mobile technologies data and in the "quickie strip" AP-42 silt sampling.
Additionally,  effects of varying vehicular speed can be clearly observed in sets 12 and 13  (from
Pass_ID 360 onwards) in Figure 6-14, where mobile technologies vehicle speeds were increase
from 25 mph to 45  mph and then decreased back to 25 mph  over several cycles. Al three mobile
technologies  emissions  factors consistently  increased with  increasing vehicle speed, and
decreased with decreasing speed.
It is illustrative to examine the estimates of emission factors from the different mobile systems,
tower measurements, and AP-42  silt based  method  on a  set-averaged basis.   As  was done
previously, during  sets when silt was applied to the road surface, we include  in the set average
only data from passes after the ninth pass following silt application.  Figure 6-15 shows the
estimated emission factors using the calibrated  mobile  systems (SCAMPER,  TRAKER I,
                                                                                    80

-------
TRAKER II) and AP-42 equations that utilize on-site silt measurements.  Overall, 1) mobile
methods measured higher emission factors when higher silt loadings were applied, and 2) the
mobile methods  track each other quite well.  The silt-based  AP-42 emission factor method
captures some of the variability exhibited by the mobile systems, but agreement of AP-42 with
mobile systems is not as good as agreement among mobile systems. The same information is
shown as scatter plots of TRAKER I, II, and silt based EF versus SCAMPER EF in Figure 6-16
and TRAKER II,  SCAMPER, and silt-based EF in Figure 6-17.

    Figure 6-15.  Comparison of Set-Averaged Emission Factors (g/vkt). Figure shows
  averages over sets with valid data for mobile systems calibrated against PMio tower flux
 measurements as described in Equations 1-3 of Section 6.2 and  silt-based emission factors
  using AP-42 equations.  Averages include only passes after the ninth pass following silt
   application for sets when silt was applied to the test road.  AP-42 emission factors are
          calculates using measured silt loadings at the end of a measurement set.
      No Valid
       Data
                                     567
                                          Set Number
                                                          10
                                                                11
                                                                     12
                                                                           13
                                                                                  81

-------
Figure 6-16. Set Averaged TRAKERI EF, TRAKERII EF, and AP-42 Silt-Based EF
                           Plotted Against SCAMPER EF
                 O TRAKER I EF
                 D TRAKER II EF
                 A AP-42 EF
                — -Linear (TRAKERII EF)
                - - - Linear (AP-42 EF)
                	Linear (TRAKER I EF)
                                                                [y=0.7356x°
                                                                |R2 = 0.49281
                                   345
                                       SCAMPER EF (g/vkt)
Figure 6-17. Set Averaged TRAKER II EF, SCAMPER EF, and AP-42 Silt-Based EF
                           Plotted Against TRAKER I EF
o

/
& f.
1> 6 "

HH
LLl ^
M"1 J
ts
^f
^4

PH
H T
M J

^•^^
X!l>^^
<2^


iy= 1.0187xi X
1 ? ' x
|R2 = 0.78251 X
i i x •
	 X "
A X ","""""" "i
X - * V - 0.793x •
X J,
X ,-' R' = 0.8186.
X ,' ........
X .* 	 "
X . ' ^» 	
x" ,-•' >^ •
X ' ~,^^
,' M.-* ^^^ v = 0.6523x
_ . - 'I^x^^"^ R2 = °-3765
^ —
*^^ •
•

                                  345
                                    TRAKER I EF (g/vkt)
                                                                                    82

-------
6.5 Comparisons of SCAMPER "First Principles" EF With TRAKER and AP-42

By first principles, the emission  factor can be calculated by multiplying the average net
concentration in the plume by the area of the plume swept out by the vehicle. Although detailed
plume concentration data are not  available from this study, the location of the SCAMPER'S rear
DustTrak has been shown to be  representative of the average concentration within the plume,
and the plume height and  width  have been shown to be approximately the frontal area of the
vehicle  (Fitz, 2001). Thus, multiplying the frontal area by the net PM concentration gives an
approximate emission rate. The  rear DustTrak's central position tend to give values of PM
concentrations that are higher than the average, while the plume dimensions have been shown to
be somewhat greater than  the frontal area. These two factors tend to cancel one another in the
multiplication process.

Based on the tower calibration, the net SCAMPER PMio concentration is multiplied by a factor
of 20 to convert mg/m3 to g/vmt. Using the  first principle approximation, the net SCAMPER
concentration is multiplied by 3.66 m , the frontal area of the Ford Expedition and multiplied by
the factor of 2.4 described in Section 4.1 to account for the discrepancy between DustTrak and
filter-based data.

As shown in Figure 6-16,  the regression  of "calibrated"  SCAMPER emission  factors with the
AP-42 emission factors yielded a slope of 0.65 with an R  of 0.63. Using the  "first principle"
SCAMPER emission factor, the SCAMPER EF shown in the figure should be divided by 20 and
then multiplied by  3.66 and 2.4; a factor 0.44 should therefore be applied.  The correlation
coefficient, R2, would remain essentially the  same. Multiplying the SCAMPER values by 0.44
would increase the slope of the regression with the AP-42 emission factor to 1.48 (0.65/0.44).
This  approach, without using a calibration,  therefore gives results  within a factor  of two
compared to the tower calibration approach.  Given the potential errors in the tower technique
and AP-42  measurements, the results for both approaches are therefore approximately equivalent
when comparing with the AP-42 emission factors.
7.0 DISCUSSION

7.1 Real World Precision and Reproducibility

7.1.1  UCR Paved Road Phases II & III for DAQEM

In the DAQEM's Phase II evaluation of mobile emissions from paved roads the  SCAMPER
system was used to characterize PMio emission rates on a single 120 mile long test route in Las
Vegas, NV. Tests  were conducted February 14-17, 2005, with one traverse of the route per day.
Emission rates for speeds less than 10 mph were excluded, as we would  not expect a well-
developed plume behind the SCAMPER vehicle. The results showed that PMio emission rates
were  generally near zero except when occasional "hot  spots"  were encountered,  which is
consistent with previous measurements. The daily average PMio emission rates for the routes
were  0.086,  0.105,  0.040 and  0.012 g/VKT (0.14, 0.17, 0.064, and 0.019 g/VMT) for
February 14th, 15th,  16th, and 17th, respectively.  Due to likely  enforcement activities  after the
                                                                                     83

-------
second measurement day, the precision of the measurement approach could not be quantified.
The two initial days suggest that the precision is approximately 10%. The emission rates for the
first two days were approximately a factor of two lower than those measured in the summer of
2004 during phase I. The test route, however, was different than the summers and there are likely
to also be seasonal differences that affect emission rates.

In the DAQEM's Phase III evaluation of mobile emissions from  paved roads the SCAMPER
system was used to characterize PMio emission rates from road loops in the Las Vegas area.  One
of the primary objectives of this study was to determine measurement uncertainty.  This was done
by making consecutive measurements over a loop of roads.   One loop was  short with high
emission potential roads in an industrial area so that a large number of traverses could be made.
Two longer loops were chosen to be more representative of emission potential  of roads in the
area. High PMio emission rates were expected from one of the longer loops, while low rates were
expected from  the other. The measurements were also used to compare the SCAMPER results
with AP-42 silt sampling, and evaluate diurnal variations of the emission factors.

The results showed that PMio emission rates met the loop expectations and were generally  low
except when "hot spots" were encountered, which is consistent with previous measurements. We
concluded that the measurement uncertainty, based on the coefficient of variation for each loop,
was approximately 25%. The PMio emission rates did not change significantly during the course
of the day, but on the high emission longer loop the rates dropped by a factor of two over the
weekend. The  comparison with AP-42 silt sampling showed good correlation (R2 = 0.86) with
the SCAMPER segment results, which were three times lower. The SCAMPER data however
were not calibrated to actual mass measurement. The calibration factor, based on a limited (8)
number of filter samples was  approximately  2, which compares well with the value of 2.4
reported here.  Applying this factor, the SCAMPER and AP-42 silt PMio emission rates were
equivalent  well within experimental  uncertainty.   Since SCAMPER  directly  measures  PM
emission rates, it is likely to be a more direct and accurate measure of PM emissions from roads.

7.1.2 DRI Studies—Clark County Phase II, Lake Tahoe and Idaho

The study  reported here is the latest in a series of TRAKER studies that started in 1999 when a
passenger vehicle was outfitted with sample tubes behind the front tire. That earlier study in Las
Vegas, Nevada, reported by Kuhns et al. (2001), was the "proof of concept" for the TRAKER
idea.  Since then a number of research efforts have been completed using the TRAKER in the
Treasure Valley in Idaho (Etyemezian et  al., 2003a, 2003b; Kuhns et al, 2003), near El Paso,
Texas (Kuhns  et al., 2005;  Gillies et al., 2005), in the  vicinity  of Lake Tahoe on  both the
California and Nevada sides (Gertler et al., 2006), and again in Las Vegas, Nevada (Etyemezian
et al., 2006).

The study near El Paso, Texas, involved the  use of a horizontal PMio flux tower to directly
measure the PMio emissions from an unpaved road and correlate those measurements with the
TRAKER  signal.  Three important findings came out of that study. First, it was found that the
PMio emission factor for a vehicle traveling on an unpaved road was directly proportional to the
speed of the vehicle as well as its weight  (Etyemezian et  al., 2003a;  Gillies et al., 2005).  This
was tested for speeds ranging from 5 to 45  mph and vehicle sizes ranging from a small passenger
                                                                                    84

-------
vehicle (Dodge Neon) to a 22-wheeled tractor-trailer.  Second, it was found that for the same
paved road, the TRAKER signal increased with speed.  Specifically, the TRAKER signal was
proportional to a constant multiplied by  the TRAKER travel speed raised to the third power.
Third, it was found that for unpaved roads, the PMio emission factor scaled with the cube root of
the raw TRAKER signal.  In summary,  it  was found that the TRAKER signal could be related to
PMio road dust emissions from unpaved roads using the Equation:

                                      EF = kT1'3                           Equation 7.1

where EF is the emission factor (g/vkt), k is the constant that relates emissions to the TRAKER
signal and is approximately 0.33  (ag=1.5),  and T is the  TRAKER  signal as  defined in
Equation 5.3 in Section 5.3.2. This provided the fit shown in Figure 7-1 for the solid circles.

For the  Treasure Valley Road Dust Study, Etyemezian et al., (2003b) used TRAKER I data
collected over 150 miles of roads near Boise, Idaho over two  seasons to assemble a PMio paved
and unpaved road dust emission inventory.  At the time of that study, the TRAKER I had not
been calibrated against an independent measure (such as horizontal flux towers) on  a paved road.
Therefore,  those  authors extrapolated the unpaved  road calibration to obtain preliminary
estimates of emissions from Treasure Valley Roads.  It was clear from the relative  magnitude of
road dust emissions in the emissions inventory that the unpaved road calibration was providing
unreasonably high values for PMio emission factors.  This was reinforced during the Lake Tahoe
Study (Gertler et al., 2006), when TRAKER I was operated on a paved road segment that was
also outfitted with a horizontal tower flux emission measurement system.  This resulted in three
data points (shown as open squares in Figure 7-1) that were clearly not in line with the unpaved
road calibration used in the Treasure Valley Study.

It  is  worth noting that up until the present study, emission factors reported for TRAKER  I
measurements were based on calibration  of the TRAKER I primarily on unpaved roads.  In the
absence of a paved road calibration,  those earlier calibrations from an unpaved road were
extrapolated to measurements on paved roads.  The present study provides a direct paved road
calibration for the TRAKER I (and TRAKER II).

In the present research effort,  TRAKER I - along with SCAMPER and TRAKER II  - was
extensively operated on a paved road in  conjunction with horizontal tower flux measurements.
The  results of this  study, shown  in Figure 7-1  as gray  circles,  along with the Lake  Tahoe
measurements (open squares), indicate that the relationship between the TRAKER signal and
PMio emission factors on paved roads  is quite different from unpaved roads.  This shows that
earlier  emissions  estimates obtained with the TRAKER I  (using unpaved road  calibration
extrapolated to paved  roads) were substantially higher than emissions that would have results
from using a paved road calibration (See Figure 7-1).
                                                                                    85

-------
Figure 7-1.  TRAKERI Calibrations. Open circles show data collected from unpaved road
 calibration near El Paso, Texas (Etyemezian et al., 2003a).  Open squares show later data
   collected on a paved road near Lake Tahoe in California (Gertler et al., 2006).  Closed
 circles are data collected on paved road from the present study. Dashed line is best linear
                  fit to data from current study and Lake Tahoe study.
       -S3
       o
       -i—i
       o
       CD
       in
10000


 1000


  100


   10


    1


  0.1
           0.01
               0.01
O El Paso Unpaved Road
D Lake Tahoe Paved Road
O Clark County Paved Road
x Paved Road - all
™ ™ Paved Road Best Linear Fit
Power (El Paso Unpaved Road)





























































i
















H
*












^

/
r^
\










= 0
?
—




t=*


1 	

JH*

^
.5406
.\/Z(j



'l-



X
0


^
^
7
*y
1







-1

H =










H«H*















^




















'



















h-





































^1^2
«— j_












-r
L












	












v =
F

















TL
^
|2 _ Q

















*
81

















H
k.
4


















3485
6











               0.1
1          10       100
  Traker Signal (mg/m3)
1000
10000
As part of an earlier phase (Phase II) of the Clark County research effort, the TRAKER I was
used to measure road dust emission potential over a road  circuit  (~ 100  miles) on four
consecutive days in February, 2005.  Researchers from UNLV were also collecting silt samples
for AP-42 based emissions estimation from points along the road circuit over the same period.
Two important findings resulted from the Phase II study that is relevant to the present effort.
First, Etyemezian et al. (2006), reported that over the 645 separate road segments that constituted
the road circuit, the precision of the  TRAKER I measurement system was better than 20% for
62% of the road segments and the precision was better than 50% for 96% of the road segments.

Second, the data collected as part of Phase II were  re-processed using the relationship between
the TRAKER signal and paved roads that has resulted from the present study (namely, Equation
1 in Section 6.3). Where data were  available from both the TRAKER I measurement and silt
samples collected from UNLV, the emission factors measured by TRAKER I were compared to
the emission factors estimated from silt measurements and application of the AP-42 equations.
Emission factors using these two methods  are shown side by  side in Figure 7-2. For the
majority of the streets  where both measurements  were completed,  the  TRAKER I emission
factors using the paved road calibration obtained from the present study are substantially lower
than the silt based emission factors calculated using the AP-42 equations. Two exceptions are
Sapphire Light and Hardin, both of which were heavily loaded with soil. Combined with the
information provided  in Figure 6-15 in Section 6.4, these data point  to a preliminary trend. It
appears that for heavily loaded roads, mobile measurement systems such as TRAKER I provide
                                                                                   86

-------
higher emission factor estimates than silt-based methods. This seems to be true for most of the
Phase IV measurements (with mobile system emission factors in the range of 2 - 7 g /vkt) as
well as the Sapphire Light and Hardin  roads measured in Phase II of the Clark County Study
(with mobile emission factors around 10 g/vkt).  In contrast, for lightly loaded roads (Emission
factors less than 1 g/vkt); the mobile systems appear to provide  a lower estimate of emission
factors than silt based methods.

Figure 7-3 shows the same information in scatter plot format.  The regression between AP-42
silt based  emission  factor estimates and TRAKER I emission factor estimates exhibits a poor
correlation (R2 ~ 0). This is in contrast to the regression of silt-based methods against TRAKER
I from the Phase IV study, where the relationship is not one to one, but does exhibit at least a
weak correlation (R2 = 0.37, See Figure 6-17, Section 6.4).  There are two possible reasons for
this difference between the Phase IV study and the real-World conditions of the Phase II study.
In the Phase IV study, the same parent road material was used for silt application for all tests and
silt was applied to the entire roadway test section more or less homogeneously. In contrast, in
the real World,  the road material that can result in  road  dust may be of quite  variable
composition (in terms of size distribution at least).  Furthermore, there are  likely to be rather
large  differences in road dirt loading  over  several kilometers of the same street.   These
differences cannot be captured by what is essentially a single point silt sample.

 Figure 7-2. Emission Factors (g/vkt) From Phase II Clark County Study.  Data are shown
     for streets where both TRAKER I and silt-based measurements were conducted.
   TRAKER I emission factors were calculated using the paved road calibration resulting
                   from the present study (See Equation 1 Section 6.2).

                   Phase II AP42-TRAKER I Emission factor comparison, Spring 2005
10.000 ,r
s
> 1 nnn
2 T
D)
£ 0.100-
O -_'-_
o
eo 0.010,:
E !;
LU ;
n nn-i -

• AP-42 silt-based EF DTRAKER I EF w/ Paved Road Calibration


ri








-



'


-i







-i






-i



P




-

r








r

-,

-i 1





-i






-i



-i




-i




-i












—

—
-i
-

-
                                      co
                                            m
                                            Street Name
                                                                           to  >
                                                                                     87

-------
   Figure 7-3.  Scatter Plot of TRAKERI EF (g/vkt) versus AP-42 Silt-Based EF (g/vkt)
        20
         18
         16
         14
         12
      S  10
                         y = 0.1019x+ 1.289
                            R2 = 0.0034
                                              10      12
                                        AP-42 EF (g/vkt)
                                                            14
                                                                   16
                                                                          18
                                                                                 20
7.2 Applying Phase IV Results in Real World Conditions- Explanation of Higher EF's in
    Phase IV

We propose a working hypothesis about the cause of the shift in the relationship between AP-42
emission factors and mobile technologies emission factors, which we call the "differential silt
mobilization hypothesis", or DSMH, for short. We will attempt to use DSMH to explain why, in
Phase II, AP-42 EF's were higher than mobile EF's compared Phase IV, where AP-42 EF's were
lower than mobile  EF's.  We believe  that these observations are  caused  by both  different
availabilities of silt for resuspension in Phase II and Phase IV, and by the higher amount of shear
applied to mobilize silt in the AP-42 method compared to mobile technologies methods.

       (a)   The Phase IV experiment was conducted on a road surface in excellent condition
            with very  low physical  roughness. In comparison, the  roadways sampled during
            Phase II exhibited a variety of roughness, but are thought to have generally higher
            physical roughness, and have more highly worn pavements than the Phase IV site.
       (b)   Silt deposited from natural processes on well-traveled  road surfaces  tends to be
            swept into the pits of the road  surface, between the protrusions caused by aggregate
            embedded in the asphalt binder.
       (c)   Measurements by Rodriguez-Bettancourt (2006)  showed that aerodynamic  shear
            applied by a conventional vacuum cleaner head during AP-42 silt recovery is likely
            to be one to three orders of magnitude higher than the  shear applied by vehicles.
                                                                                     88

-------
The combined result of (a), (b), and (c) makes it more difficult to mobilize silt into the air when
it is embedded in the pits of a normally-traveled, medium-rough road surface during conditions
of moderate shear applied by  vehicle tires and aerodynamic wakes compared to  the greater
degree of mobilization resulting from conditions of higher shear applied by a vacuum cleaner
head  during  AP-42  sampling. As a  result,  on natural, rougher  road   surfaces,  a  higher
mobilization of the silt fraction by a vacuum cleaner would  lead to  a higher AP-42  emission
factor, for  the same amount of silt loading, than the emission factors observed by mobile
technologies vehicles.

In contrast, during the Phase IV experiment, freshly applied road silt was more evenly distributed
between the smaller "protrusions" and "pits" on a  smoother, well-sealed road surface, and was
therefore easier  to mobilize in conditions of moderate  applied vehicle shear than a similar
loading on  a normally traveled paved road. As a result, observed mobile  EF's would be higher,
for the same silt loading, on the Phase IV experimental road surface, than they were  on the
normally traveled paved roads that were measured in Phase II. AP-42 vacuumed EF's would be
similar, since four vacuum passes have been shown to recover greater than 99% of applied road
silt  on both smooth  and rough road surfaces, and  the  vacuum exerts a  very  high level of
aerodynamic  shear.  The result is  that mobile technologies EF's  are  hypothesized to have
increased relative to AP-42 EF's on the Phase IV surface compared to the Phase II surface.

This hypothesis would also explain why, in conditions of heavy soil loading, such  as Hardin and
Sapphire Light in Phase II, as well as for the Veterans Memorial Boulevard loadings in Phase IV,
mobile technologies  emissions factors were higher than for AP-42,  because, under these
conditions,  there is a large amount of silt on top of the protrusions that can be easily suspended

7.3  Advantages of Mobile Technologies

Real-time vehicle  mounted mobile sampling systems provide a number  of very significant
improvements over the  current AP-42  paved road dust emissions estimating equation.  The
mobile sampling systems are not subject to many  of the assumptions  and limitations applicable
to the AP-42 equation, including the requirement for free flowing traffic, speed ranges between
10 and 55 mph, the need to block lanes of traffic for silt sampling, the ability to  sample  on all
road functional classes, and the ability to  collect a large number of measurements over a short
time period.

Mobile sampling systems can effectively sample on congested urban streets where traffic is not
free flowing, whereas  the  AP-42  emissions equation is predicated on free  flowing traffic.
Applying AP-42 emissions estimating methodology to roadways with  heavily congestion results
in unknown but potentially significant errors.  The GPS linked data collection system utilized in
the mobile  sampling systems allow the operator to easily exclude data points collected below a
specified de minimus threshold speed, typically set at 10 mph.

Mobile sampling systems are speed independent  and can accurately measure emissions at all
non-de minimus  (>10 mph)  speed  ranges, including speeds above 55  mph.  By comparison, the
AP-42 emissions equation is not validated for vehicle speeds above 55 mph.
                                                                                    89

-------
Mobile sampling systems provide a safer method of measuring paved road dust emissions. The
mobile sampling systems can operate without the need for lane closures and the associated public
safety risk and increased traffic congestion.

Mobile  sampling systems can accumulate  paved road emissions data much faster and more
economically than the AP-42 emissions equation methodology.  The mobile sampling systems
provide a means of sampling significant percentages of the entire road network in an airshed or
nonattainment area.   The abundance of data developed  with  the  mobile sampling systems
approach allows for the development of specific emission factors for many criteria known to
affect the  paved road  dust emission rate.  These include, in addition to road  functional
classification, road infrastructure development and land use type and development.  Impacts of
specific silt deposition sources may also be evaluated.  These detailed breakdowns will allow SIP
developers to prepare more complete and representative emissions inventories for the paved road
dust source category.  The benefits of more robust emission factor information would be even
more profound for air regulatory agencies and MPOs developing future emissions projections for
this source category.  The mobile sampling systems ability to provide much larger data sets will
allow SIP  planners and MPOs to develop far more detailed and realistic  projected emissions
estimates for future year paved  road dust emissions.

7.4 Paved Road Dust Emission Inventory Development

The AP-42, Section 13.2.1 Paved Roads - Background Documentation, sets forth test results for
selected functional classes of  roadways.  This  documentation does not address  utilization  of
emissions factors or development of emissions inventories.  State and local transportation agency
nomenclature for functional road classification may very  slightly from place to place,  but a
typical breakout of functional road classifications is as follows:

    •  Freeway
    •  Major Arterial
    •  Minor Arterial
    •  Collector
    •  Local

In addition, certain other classes  such  as freeway on-ramps and off-ramps, industrial roads, and
alleys may also be included in an MPO's functional road classification and transportation model.

Most agencies vested with state implementation plan development responsibilities use the AP-42
default silt loading values provided in  the AP-42 document in lieu of acquiring current local silt
measurements.   This  significantly degrades the quality and confidence levels of the AP-42
derived  paved road dust emissions estimates.  For those entities that make local silt loading
measurements when  developing emission  factors  for  paved  road  dust  emissions,  the
measurements are typically confined to minor arterial, collector, and local roads.  Public works
agencies typically will not issue encroachment permits for sampling on heavily congested  major
arterial roads.  Less heavily congested major arterial  roads  are sometimes sampled,  but these
results in biased emissions estimates  as the samples are not  representative  of the most heavily
traveled major arterial roads. State departments of transportation seldom allow silt sampling  on
                                                                                      90

-------
congested urban freeways due to severe traffic disruptions and related safety hazards of traffic
flow disruption.  Emission inventory developers must therefore rely on silt loading data for
freeways which may be decades old and may have originated from freeways located in another
state.

Once silt sampling and analysis is complete, the AP-42 paved road equation is used to establish a
vehicle miles traveled (VMT) based emission rate for each functional  road class.  This VMT
emission rate for each functional  road class is  then  applied to the regional road network to
determine the total emissions from that functional road class.  This step is repeated for each
functional road  class represented in the road network.  The sum of emissions from all functional
road classes provides the total road network emissions.  VMT values for each functional road
class are obtained from the transportation model utilized by the MPO.

The VMT by  functional road  class approach  would also be utilized with emissions  data
developed  using  near real time vehicle  mounted mobile  sampling systems.   The primary
difference between the mobile sampling system  and AP-42  approach would be the number of
data points used and the percentage of that road network that could  be represented by sampling.
Where cost and time constraints inherent in the AP-42 method limit sampling to a few hundred
feet of the road network, the mobile systems allow sampling of many miles of the road network.
It is also feasible to make multiple repeat measurements of road segments to allow assessment of
week day and weekend emission rates using the mobile sampling systems.

The largest constraints on  road  network emissions characterization  with mobile sampling
systems are the transportation models. Given the ability to sample many miles of roads using a
mobile system,  it is feasible to develop emission factors for subclasses for each functional road
class.   One sub classification might reflect the presence  or absence of paved shoulders, curbs
and gutters. Clark County research has shown that, other factors remaining the same, emission
rates are higher on roads without paved shoulders, curbs and gutters.  Another sub classification
that may affect a road's emission  rate are adjacent or nearby land uses.  Industrial land uses
typically contain more sources of road silt deposition than commercial or residential land uses.
Another matrix that might be applied to each functional road  class is some quantification of
construction activity occurring in the vicinity of each road segment.

Transportation models are typically not currently set up to break out these sub classifications of
functional road  class. As a result, the potential refinement of the functional  road class emission
factors may not provide additional benefits with the current transportation models.  The DAQEM
is currently exploring the feasibility of using functional sub classification emission factors with
the Regional Transportation Commission of Southern  Nevada (RTC). Development of a more
comprehensive  library of emission  factors could potentially provide significant improvements to
present and future year emission inventories.

The DAQEM will continue to work with the RTC (the MPO for Clark County) to determine the
appropriate sub class emission  factors that can be used  in conjunction with the TransCAD
transportation model to develop the most refined paved road dust emission inventory.  It may be
necessary to pre-process data inputs for the model  in order to utilize certain sub class emission
factors.  For example, the current road network data set may not include complete information on
                                                                                      91

-------
existing curb and gutter infrastructure.  Curb and gutter location information can be developed
using Geo Span  digital street imagery and then coded into the model input data.  Model VMT
outputs will then allow use of the correct sub class emission factor.

Once an optimal set  of  sub  class emission factors are identified,  a  sampling plan will be
developed for determining the emission factor for each  road sub classification using a real time
vehicle mounted mobile sampling system.  This sampling plan will provide the basis of the
emission inventory  improvement plan  for the  paved road dust source category and will be
submitted to EPA Region 9 for review and concurrence.

Upon receipt of concurrence from EPA Region 9,  DAQEM  will complete a formal  scope of
work for field measurements, data  processing  and  analysis, and report  preparation.   The
department will  then  acquire the services  of a qualified consultant  utilizing standard  county
business practices.

Following completion of  field measurements and acceptance  of the study report and data, the
DAQEM will work with the RTC to develop a new "clean sheet" emissions inventory for paved
road dust emissions.  Incorporation of this inventory into the PMio Maintenance Plan will allow
Clark County to develop improved future year  PMio projections and  transportation conformity
budgets.
8.0 CONCLUSIONS

8.1 Conclusions

In this study, controlled measurements of PMio road dust emissions were completed on a test
road in Boulder City, Nevada.  Well-characterized parent soil was spread onto the test road
surface at the beginning of most measurement sets. Silt samples were procured at the beginning
and end of each measurement set as  well as during the measurement set in  some  cases.
Simultaneously, three mobile road dust measurement systems were used to traverse the test road:
SCAMPER, TRAKER I, and TRAKER II. These mobile systems were used both to measure the
potential for road dust emissions and to serve as road dust sources. Horizontal flux of PMio was
measured using an instrumented tower system  to obtain an independent  measure of the PMio
emission factors from travel on the test road section.  The tower measurements were considered
as the  standard for comparing the other four measurement methods (three mobile methods and
silt method).

It was clear from examining the data from both the horizontal flux tower and the mobile systems
that after the application of soil to  the test road, the first nine or so vehicle passes resulted in
     emissions that were

       (a)   much higher than subsequent passes and
       (b)   apparently caused by a different mechanism than subsequent passes.
                                                                                    92

-------
In comparisons of mobile and  silt systems to horizontal tower measurements, the first nine
vehicle passes were omitted as  they likely represented a very short-lived mechanism for road
dust emissions that would not be prevalent on a well traveled real road.

Averages  of PMio emission factors  measured with the tower system  were calculated on  a
measurement set basis along with comparable averages for mobile systems.  A simple linear fit
appeared to be adequate for describing the relationship between the mobile systems' raw signal
and the emission factors measured by the tower system.  The raw signals for all three  mobile
units were calculated as the PMio concentration at a location that is influenced by the road dust
generated by the vehicle minus  the background PMio concentration.  All three mobile systems
correlated reasonably well with  the tower measurements  (with R2 values ranging from 0.47 to
0.75).  To obtain PMio emission factors, it was found that the TRAKER I, TRAKER  II, and
SCAMPER raw signals required multiplication by 0.54, 0.92, and 20, respectively.

Silt measurements were used to calculate emission factors following the equations provided in
AP-42.  Those emission factors were then compared to the  tower data as well as to  emission
factors obtained with the calibrated mobile systems.  The mobile systems agreed well  with one
another - not surprising since they were all  calibrated against the same tower data - and showed
reasonable correlation with silt-based emission factors.

In general,  silt based  measurements  resulted  in slightly lower  emission  factors  than those
measured by the tower  and mobile systems. In contrast, when the  same tower based calibration
was applied  to TRAKER I data acquired on a wide range of Clark County roads as part of an
earlier phase of this research effort, and compared to AP-42 emissions factors derived from silt
measurements obtained from those same roads over the same sampling period, the TRAKER I
measurements generally provided much lower emission factors than emission factors calculated
from the silt measurements.

As  described in Section 7.2, we believe  that this  shift in the relationship between  mobile
technologies EF's and AP-42 EF's is caused by differential silt mobilization, which occurred as
result of a greater proportion of the applied silt loading being distributed on a the  tops of the
embedded road surface  aggregates, and hence being more easily entrained by vehicle mechanical
and aerodynamic shear  from the  Phase IV experimental road surface, compared to the less easily
entrained silt more likely to be embedded between the  road surface aggregates on the Phase II
road surfaces.

8.2 Recommendations

Vehicle mounted mobile sampling systems avoid many  limitations  of the  current AP-42 method
for estimating road dust emissions.  These  limitations led Clark County to conduct the Phase I
through IV  field  measurement  studies to  validate  the effectiveness of the mobile  sampling
systems.  These studies augmented  six-years of extensive AP-42 silt sampling and analysis for
emissions inventory development. As a result of this effort, the DAQEM concluded that real-
time based vehicle mounted mobile  sampling systems provide superior and a more flexible
approach for developing SIP emissions inventories.   These systems provide similar advantages
for inventorying emissions from stabilized unpaved haul roads and other public and  private
                                                                                    93

-------
unpaved roads.  In addition to SIP emissions inventory development, these systems provide a
preeminent method for  measuring road dust emissions at  stationary  sources for permitting
purposes.

DAQEM has discussed approval of real-time based vehicle mounted mobile sampling systems
for SIP emissions inventory development with EPA Region 9 and EPA OAQPS.  Both offices
have indicated the need for a peer review process prior to  a regional or OAQPS approval.
DAQEM is seeking regional (Region 9) approval to utilize vehicle mounted mobile sampling
systems to develop the paved road dust emission inventory for the County's PMio Maintenance
Plan. As part of  Clark  County's  evaluation of the real-time based vehicle mounted mobile
sampling systems, DAQEM informally contacted  a number  of state and  local air regulatory
agencies, many of which have  expressed support for this  alternative method of emission
inventory development.  This alternative method was also discussed with Metropolitan Planning
Organizations (MPOs), all of whom were interested.

Following the presentation of Clark County's conference paper at the 16*  Annual International
Emissions Inventory Conference, Clark County worked with project contractors to further refine
the study findings and develop a formal research report.  A number of air regulatory agency,
MPO staff, and research scientists  have agreed to participate in the peer review.   Following
completion of the peer review process,  Clark County will request EPA Region 9 approval of
real-time vehicle-mounted mobile sampling systems as a locally approved method for use in the
Clark County's PMio Maintenance Plan.

Clark County's  mandates do not require EPA OAQPS approval of the real-time based vehicle
mounted mobile sampling system as an approved (alternative) AP-42 method, Clark County may
indirectly benefit  from  improved characterization of the paved  road  dust sources by  other
regulatory agencies if this were  to  occur.  Clark County  DAQEM  will provide technical
assistance as  requested, to other state and federal agencies such as BLM and DOD, MPOs and
organizations such as WRAP, NACAA, WESTAR, who may wish to pursue AP-42 federal
reference method approval through OAQPS.
                                                                                    94

-------
9.0  REFERENCES

U.S.  Environmental  Protection Agency,  Emission Factor Documentation  for  AP-42, EPA
Contract No. 68-DO-0123, MRI Project No. 9712-44 dated March 8, 1993.
Geotechnical and Environmental Services, Inc., Presentation of Final Versions of Deliverables
for Re-Evaluating and Updating the Paniculate Emission Potential Map and Soil Classification
for Dust Mitigation Best Management Practices Manual for Clark County, dated September 26,
2003.
Etyemezian V., H. Kuhns, J. Gillies, M. Green,  M. Pitchford,  and J. Watson (2003). Vehicle
based  road  dust emissions  measurements  (I):  Methods  and  Calibration.  Atmospheric
Environment 37: 4559-4571.
Etyemezian V., H. Kuhns, J. Gillies, J. Chow, K.  Hendrickson, M.  McGown  and  M. Pitchford
(2003b). Vehicle based road dust emissions measurements (III):  Effect of speed, traffic volume,
location, and season on PMio road dust emissions. Atmospheric Environment 37: 4583-4593.
Etyemezian, V., H. Kuhns, and  G. Nikolich (2006).  Precision and repeatability of the TRAKER
vehicle-based paved road dust emission measurement. Atmospheric Environment 40: 2953-2958.
Gertler A,  H.  Kuhns, M. Abu-Allaban, C. Damm, J.  Gillies, V. Etyemezian, R. Clayton, D.
Profitt (2006). A Case Study of the Impact of Winter Road Sand/Salt and Street  Sweeping on
Road Dust Re-entrainment. Atmospheric Environment 40 (31): 5976-5985.
Gillies, J.A.; V. Etyemezian,  H. Kuhns,  D.  Nikolic, and D.A. Gillette (2005). Effect of vehicle
characteristics on unpaved road  dust emissions; Atmospheric Environment 39,  2341-2347.
Kuhns, H., Etyemezian, V.,  Landwehr, D.,  MacDougall, C.,  Pitchford, M.,  and  M.  Green.
(2001). Testing Re-entrained  Aerosol  Kinetic Emissions from  Roads (TRAKER): A New
Approach to Infer Silt Loading on Roadways. Atmospheric Environment 35: 2815-2825.
Kuhns H., V. Etyemezian, M. Green, Karin Hendrickson, Michael  McGown, and Kevin Barton,
Marc  Pitchford (2003) Vehicle-based  road  dust emissions  measurement  (II):  Effect of
precipitation, winter time road  sanding, and  street sweepers on PMio fugitive dust emissions
from paved and unpaved roads.  Atmospheric Environment 35: 4572-4583
Kuhns, H.,  M.-C. O.  Chang, J.C. Chow, V. Etyemezian, L.-W.  A. Chen, N. Nussbaum,  S.K.K.
Nathagoundenpalayam, D. Trimble,  S. Kohl, M. MacLaren, M.  Abu-Allaban,  J.  Gillies,
A. Gertler,  C.  Damm, C. Denney, C. Gallery, and J. Skotnik (2004).  DRI Lake Tahoe  Source
Characterization  Study:  Final Report.   Prepared  for  California Air Resources   Board,
1001 I Street, Sacramento, CA 95812, Oct. 2004.
Kuhns, H., V. Etyemezian, J. Gillies, D. DuBois, S. Ahonen, and D. Nikolic (2005).   Spatial
Variability  of Unpaved Road Dust PMio Emission Factors near El Paso, Texas. Journal of the
Air and Waste Management Association.55: 3-12.
Etyemezian, V., H. Kuhns, J. Gillies, M. Green, M. Pitchford and J.  Watson  (2003). "Vehicle-
based road  dust emission measurement:  I - methods and calibration." Atmospheric Environment
37(32): 4559-4571.
                                                                                   95

-------
Kuhns, H., V. Etyemezian, D. Landwehr, C. MacDougall, M. Pitchford and M.  Green (2001).
"Testing Re-entrained Aerosol Kinetic Emissions from Roads (TRAKER): a new approach to
infer silt loading on roadways." Atmospheric Environment 35(16): 2815-2825.
Countess,  R. (2001).  Methodology  for  Estimating Fugitive  Windblown  and  Mechanically
Resuspended Road  Dust Emissions Applicable for Regional  Scale Air Quality Modeling.
Prepared for the Western Governors Association by Countess Environmental, Westlake Village,
CA, April, 2001
Cowherd,  C. (1984). Paved Road Particulate Emissions: Source Category Report. Report No.
EPA-600/7-84-077.  Prepared by  U.S.  Environmental  Protection Agency, Research  Triangle
Park, NC, USA.
Cowherd,  C. (1999).  Profiling Data for  Open Fugitive  Dust Sources. Prepared for U.S.
Environmental Protection Agency, Emission Factors and Inventory Group,  Office  of Air Quality
Planning and Standards, Research Triangle Park, NC,  by Midwest Research Institute, Kansas
City, MO.
Cowherd,  C., and T. Pace (2002). Potential role of vegetative ground cover in the removal of
airborne fugitive dust.  Presented at the  Annual meeting of the Air and Waste Management
Association,  Baltimore, MD, June 2002.
Etyemezian V., Kuhns,  H., Chow, I, Gillies, I, Green,  S. Kohl, and M. Pitchford (2001). The
Treasure  Valley Road  Dust  Study: Final  Report. Prepared  for  the Idaho Department of
Environmental Quality, Boise, Idaho by the Desert Research Institute, Las Vegas, NV. February,
2002.
Etyemezian V., Kuhns, H., Chow, J., Gillies, J., Green, M., Hendrickson, K., McGown, M., and
M. Pithford  (2003a). Vehicle  Based  Road Dust Emissions Measurement (I):  Methods and
Calibration. Manuscript submitted as companion paper to Atmospheric Environment.
Etyemezian,  V., Gillies, J., Kuhns, H., Nikolic, D., Watson, J., Veranth, J., Laban, R., Seshadri,
G., and D. Gillette (2003b).  Report prepared for the WESTAR Council, Lake  Oswego, OR,
USA by DRI, Las Vegas, NV, USA, January 2003.
Etyemezian,  V., H.  Kuhns,  G.  Nikolich,  and K.  Graham (2004).  The Las Vegas Road Dust
Emissions Technology Assessment: Final Report. Prepared by the Desert Research Institute, Las
Vegas NV for the Clark County Department of Air Quality and Environmental Management, Las
Vegas, NV, July, 2005.
Gillette, D. (2002). The Long-distance  "Transportable Fraction" of the Vertical  Flux of wind-
transported dust. Proceedings of ICAR5/GCTE-SEN Joint Conference, Jeffrey A. Lee  and Ted
M. Zobeck editors, Lubbock, Texas, USA Publication 02-2 p. 34.
Gillies, J.A.; Watson, J.G.; Rogers, C.F.;  Dubois, D.; Chow, J.C.; Langston, R.; and Sweet, J.
(1999). Long Term Efficiencies of Dust Suppressants to Reduce PMio Emissions  from Unpaved
Roads. JAWMA, 49:3-16.
Kuhns, H., Etyemezian, V., Landwehr, D., MacDougall, C., Pitchford, M.,  Green, M. (2001).
Testing Re-entrained Aerosol Kinetic Emissions from Roads (TRAKER):  A New Approach to
Infer Silt Loading on Roadways. Atmospheric Environment Vol  35: 2815-2825.
                                                                                   96

-------
Kuhns, H., Etyemezian, V.,  Gillies, J., Green,  M., Hendrickson, K., McGown, M., and M.
Pitchford (2003). Vehicle Based Road Dust Emissions Measurement (II): Effect of Precipitation,
Wintertime Road Sanding,  and Street Sweepers  on PMio Fugitive Dust Emissions from Paved
and Unpaved Roads. Manuscript submitted as companion paper to Atmospheric Environment.
Moosmuller, H., Gillies, J.A., Rogers, C.F., Dubois, D.W., Chow, J.C., Watson, J.G., and R.
Langston (1998). Particulate Emission Rates for Unpaved Shoulders along a Paved Road. J. Air
& Waste Manage. Assoc., Vol 48: 398-407.
Raupach, M.R., and F.L. Leys (1999). The Efficacy of Vegetation in Limiting Spray Drift and
Dust Movement.  Report Prepared for the  Department of Land and Water  Conservation,
Gunnedah, Australia by CSIRO, Canberra, Australia.
Sehmel, G.A. (1973). Particle Resuspension from an Asphalt Road Caused by Car  and Truck
Traffic. Atmospheric Environment, Vol 7: 291-309.
U.S. EPA (1995). Compilation of Air Pollutant Emission Factors. Report No. AP-42. Prepared
by the U.S. Environmental Protection Agency, Research Triangle Park, NC, USA.
U.S.EPA (1998).   National air pollutant emission trends,  procedures  document,  1900-1996.
Report No. EPA-454/R-98-008. Prepared by U.S. Environmental Protection Agency, Research
Triangle Park, NC, USA.
U.S.EPA (1999). Compilation of air pollutant emission factors - Vol. I, Stationary point and area
sources.  Report No.  AP-42, 5th  ed.   Prepared by U.S. Environmental Protection  Agency,
Research Triangle Park, NC, USA.
U.S.EPA (2002). User's Guide to MOBILE6.0 Mobile Source  Emission Factor Model. Report
No. EPA-420/R-02-001. Prepared by U.S. Environmental Protection Agency, USA.
Venkatram, A. (2000). A Critique of Empirical Emission Factor Models: A Case Study of the
AP-42 Model for Estimating PMio Emissions from Paved Roads. Atmospheric Environment, Vol
34: 1-11.
Venkatram, A. (2001). Response to Comments by Nicholson. A Critique of Empirical Emission
Factor Models: A Case Study of the AP-42 Model for Estimating PMio Emissions from Paved
Roads. Atmospheric Environment Vol 35: 187.
Watson, J.G., and J. Chow (2000).  Reconciling Urban Fugitive Dust Emission Inventory and
Ambient Source Contribution  Estimates: Summary of Current Knowledge and Needed Research.
DRI Document No. 6110.4F.  Prepared for the U.S. Environmental Protection Agency, by Desert
Research Institute, Reno, NV.  May,  2000.
Watson,  J.G.,  Chow, J.C.,  Pace,   T.G.  (2000).   Fugitive dust emissions.   Air  Pollution
Engineering Manual, Davis, W.T., Ed.  Van Nostrand Reinhold, New York, NY, pp. 117-134.
Watson, J.G., and J. Chow (2000).   Reconciling Urban Fugitive Dust Emissions Inventory and
Ambient Source Contribution  Estimates: Summary of Current Knowledge and Needed Research.
DRI Document No. 6110.4F.  Prepared for the U.S. Environmental Protection Agency, by Desert
Research Institute, Reno, NV.  May,  2000.
Countess,  R. (2001).  Methodology for Estimating Fugitive  Windblown and  Mechanically
Resuspended  Road  Dust Emissions Applicable for Regional Scale  Air  Quality  Modeling.
                                                                                  97

-------
Prepared for the Western Governors Association by Countess Environmental, Westlake Village,
CA, April, 2001.
Veranth, J. M., G. Seshadri and E. Pardyjak (2003). Vehicle-generated fugitive dust transport:
Analytic models and field study. Atmos Env 37(16): 2295-2303.
Raupach, M.R., N. Woods, G. Dorr, J.F. Leys, and H.A. Cleugh (2001).  The Entrapment of
Particles by Windbreaks. Atmos. Env. 35: 3373-3383.
Lamb, R.G., and D.R. Duran (1977). Eddy Diffusivities Derived from a Numerical Model of the
Convective Boundary Layer. Nuovo Cimento, Ic: 1-17.
Shir,  C.C.  (1973). A  preliminary Numerical  Study  of Atmospheric Turbulent Flows  in the
Idealized Planetary Boundary Layer. J. Atmos Sci., 30:  1327-1339.
Venkatram,  A. (1993). The Parametrization of the Vertical Dispersion of A Scalar  in the
Atmospheric Boundary Layer. Atmos Env.,  27A: 1963-1966.
U.S. EPA, (1995). User's Guide for the Industrial Source Complex (ISC3) Dispersion Models
Volume II: Description of Model Algorithms. US Environmental Protection Agency, Office of
Air Quality Planning and Standards: Emissions, Monitoring, and Analysis Division, RTF, NC.
Seinfeld, J.H., and  S.N. Pandis (1997).   Atmospheric  Chemistry and  Physics: From Air
Pollution to Climate Change.  Wiley Interscience. New York,  USA.
Willeke, K., and M. Xu (1992). Impaction and rebound of particles from surfaces. J. Aerosol Sci.
23, Supplement: S15-S18.
Gillies,  J.A., J.G. Watson, C.F. Rogers, D. Dubois, J.C.  Chow,  R.  Langston,  and J. Sweet.
(1999).  Long term efficiencies of dust suppressants to reduce PMio emissions  from unpaved
roads. J. Air Waste Manage. Assoc. 49:3-16.
Cowherd, C.  (1999).  Profiling  data for open fugitive dust  sources.  Prepared for U.S.
Environmental Protection Agency, Emissions Factor and Inventory Group, Office of Air Quality
Planning and Standards, RTF, NC, by Midwest Research Institute, Kansas City, MO.
Niu, J., B.M.K.  Lu,  and T.C.W. Tung (2002). Instrumentation Issue in  Indoor Air Quality
Measurements: The Case with Respirable Suspended Particulates. Indoor Built Environment 11:
162-170.
Moosmuller, H., W.P.  Arnott, C.F. Rogers, J.L. Bowen, J.A. Gillies, W.R. Pierson, J.F. Collins,
T.D. Durbin, and J.M. Norbeck  (2001).   Time resolved characterization of diesel particulate
emissions. 1. Instruments for particle mass measurement. Environ. Sci. Technol.  35: 781-787.
Chung, A., D.P.Y. Chang, M.J. Kleeman, K.D. Perry, T.A. Cahill, D. Dutcher, E.M. McDougall,
and K. Stroud (2001).  Comparison of real-time instruments used to monitor airborne particulate
matter. J. Air & Waste Manage. Assoc. 51: 109 - 120.
Etyemezian, V., J. Gillies, H. Kuhns, D. Nikolic, J. Watson, J. Veranth, R. Labban, G. Seshadri,
and D.  Gillette  (2003).   Field  Testing and  Evaluation  of Dust Deposition  and Removal
Mechanisms: Final Report. Report Prepared for WESTAR Council, Lake Oswego, OR, by DRI,
Las Vegas, NV. January 2003.
Etyemezian, V.,D. Gillette, J.  Gillies,  H. Kuhns, D. Nikolic,  J. Veranth, and J. Watson (2003).
      Emissions Factors for Unpaved Roads: Correction for Near-Field Deposition.  PM AAAR
                                                                                    98

-------
2003: Atmospheric Sciences, Exposure, Health and Welfare Effects, Policy.  Poster. American
Association for Aerosol Research, Pittsburgh, PA, March 31 - April 4, 2003.
Rodrigues, Geisa Bettancourt, 2006. Improved Techniques  for estimation of PMio emissions.
Master's thesis, University of Nevada, Las Vegas, NV. 171 pp.
US EPA 2006. AP42 Emissions Factors. Miscellaneous Sources, Section  13.2.1. Paved Roads.
US Environmental Protection Agency, Research Triangle Park, NC, November, 2006. 15 pp.
US EPA 1993a. AP42 Emissions Factors. Miscellaneous Sources, Section  13.2.1. Paved Roads.
Appendix  C.I.  Procedures for Sampling  Surface/Bulk Dust Loading.  US  Environmental
Protection Agency, Research Triangle Park, NC, November, 2006 13 pp.
US EPA 1993b. AP42 Emissions Factors. Miscellaneous Sources, Section  13.2.1. Paved Roads.
Appendix C.2. Procedures for Laboratory Analysis of Surface/Bulk Dust Loading Samples. US
Environmental Protection Agency, Research Triangle Park, NC, November, 2006 9 pp.
Fitz, Dennis R (2001). Final Report, Measurements of PMio and PM2.5 Emission Factors from
Paved Roads in California, Prepared for the California Air Resources Board Monitoring and
Laboratory Division, Contract No. 98-723.
Venkatram, Akula and Fitz, Dennis R (1998). Phase 1 Final Report, Measurement and Modeling
of PMio and PM2.5 Emissions from Paved Roads in California.  Prepared for the California Air
Resources Board Monitoring and Laboratory Division, Contract No. 94-336.
Gertler,  A.W., Coulombe, W.G., Watson, J.G., Bowen, J.L. and Marsh, S.,  1993. Comparison of
PMio concentrations  in  high- and  medium-volume  samplers  in  a  desert  environment.
Environmental Monitoring and Assessment, 24: 13-25.
Etyemezian, V., H. Kuhns, J. Gillies, S. Uppapalli, G. Nikolich, and D.  Zhu (2008).  Clark
County Alternative Technologies for Paved Road Dust Emissions Measurement: DRI Draft Final
Report.  Report Prepared  for Clark County Department of Air Quality and Environmental
Management 500 S. Grand Central Pky. P.O. Box 551741 Las Vegas, NV 89155-1741.
Fitz, D.R. (2001) Measurements on PMio and PM2.5 Emission Factors from Paved Roads in
California. Final Report to the California Air Resources Board  under Contract No 98-723, June
2001.
                                                                                   99

-------
10.0GLOSSARY
Term
Set
Run
Pass
TRAKER I
TRAKER II
Silt application
Aerodynamically
suspendable
Mechanically
suspendable
Tower-based
measurements
PM10 horizontal
flux
PM10 emission
factor
TRAKER signal
Explanation
Measurements were organized into sets, runs, and passes. A measurement
set consisted of a series of tests conducted with the mobile sampling systems
under a specific set of conditions. The conditions specified include whether or
not and how much silt material is applied to the road surface, the speed of
travel of the mobile sampling systems through the test course, and the
purpose of the measurements (e.g. uncover the rate of road dust material
depletion overtime).
Measurements were organized into sets, runs, and passes. Within each
measurement set, the three mobile sampling systems would go through the
test course in turn. A run refers to three consecutive measurements through
the test course with one measurement associated with each the TRAKER I,
TRAKER II, and SCAMPER.
Measurements were organized into sets, runs, and passes. A pass refers to
the completion of a single test vehicle through the test course. The Pass ID is
an integer index used to uniquely identify the vehicle that passed through the
test course, the measurement set, and the time of the pass (the time that the
vehicle passes the tower measurement system).
Testing Re-Entrained Aerosol Kinetic Emissions from Roads I. This is a
vehicle based, mobile sampling platform developed at DRI that measures the
amount of dust suspended behind the vehicle's front tires.
Fundamentally similar to TRAKER I with some modifications of the inlet
configurations behind the front tires and some software improvements.
Refers to the intentional spreading of soil material on the test road surface in
order to simulate different degrees of road "dirtiness". The material applied is
not exclusively composed of silt, but rather represents soils in Southern
Nevada.
Refers to emissions of road dust through aerodynamic entrainment. This
usually occurred during the first 9 times that a vehicle passed through the test
course following silt application. Measurements associated with aerodynamic
entrainment (first 9 passes) were not included in the calibration procedures
where tower-based PM10 emission factors were compared to data from the
mobile sampling systems.
Refers to emissions of road dust through mechanical entrainment.
Immediately after silt application, the dominant emission process was
aerodynamic entrainment. After 9 vehicle passes, emissions occur under a
long-lived mechanical regime.
Measurements of the horizontal flux of PM10 road dust using a vertical tower.
By measuring wind speed, wind direction, and PM10 concentrations at several
different heights above the ground, it is possible to numerically integrate the
mass of PM10 crossing a vertical plane that is parallel to the test road.
Tower-based measurements provide PM10 horizontal flux. By measuring wind
speed, wind direction, and PM10 concentrations at several different heights
above the ground, it is possible to numerically integrate the mass of PM10
crossing a vertical plane that is parallel to the test road. Since the emitted
particles are moving with the wind, this is referred to as the horizontal flux.
For the purposes of this study, "PM10 horizontal flux" and "tower-based PM10
emission factor" are synonymous.
For the purposes of this study, "PM10 horizontal flux" and "tower-based PM10
emission factor" are synonymous.
Refers to the background-corrected PM10 concentrations measured behind
the front tires of the TRAKER vehicle. The TRAKER signal is calculated by
                                                                     100

-------
      Term
                             Explanation
                   obtaining the average of the PM10 concentrations measured through the inlets
                   located behind the left and right front tires and then subtracting the
                   background PM10 concentration from this value. The background PM10
                   concentration is measured at the front bumper of TRAKER I and through a
                   chimney located near the  roof of TRAKER II.	
Set average
Applies to mobile systems as well as tower-based measurements. When
averaging over a set, data from individual passes that comprise the
measurement set are averaged. For tower-based measurements, set
averages consist of all valid PM10 horizontal flux measurements obtained
within the measurement set regardless of the test vehicle.  For some sets,
where silt was intentionally applied to the road surface, the set average does
not include data associated with the first 9 passes after silt application (See
"aerodynamically entrainable").  For mobile system measurements, set
averages consisted of all valid passes of the specific sampling vehicle (i.e.
TRAKER I set averages are based only on TRAKER I passes). As with the
tower data, for some measurement sets, the first 9 passes after silt
application were excluded from the average.	
Pass average
Average of data associated with the passage of a specific test vehicle through
the test course.  For tower-based measurements, the pass-average is
obtained from the horizontal PM10 flux measured at a single location along the
test course. For mobile measurement systems, the pass average
encompasses all the 1-second data collected over the duration of the pass
(i.e. from the time the test vehicle crosses the beginning of the test course till
the vehicle crosses the end marker of the test course).	
                                                                                          101

-------
11.0 ACKNOWLEDGMENTS

The  authors gratefully acknowledge the field  and technical assistance provided by Sebastian
Uppapalli, and George Nikolich. We would also like to thank James King, Robert Powell, Bias
Kavouras, and Alex Nikolich for logistical support during the field study. Also, we acknowledge
Kurt Bumiller for operating the  SCAMPER and preparing the data for that portion of the report,
and Deborah Hart for assisting with field study safety coordination and course design.   In
addition, we would like to thank Rebecca Kies, student intern at Midwest Research Institute, for
her editing contributions.
                                                                                    102

-------