FIELD TESTING AND EVALUATION OF DUST DEPOSITION AND ...

FIELD TESTING AND EVALUATION OF DUST DEPOSITION AND
REMOVAL MECHANISMS: FINAL REPORT
Vic Etyemezian
Jack Gillies,
Hampden Kuhns,
Djordje Nikolic
and
John Watson
DRI
755 E. Flamingo rd
Las Vegas, NV
(702) 895-0569
vic@dri.edu
John Veranth,
Raed Laban,
and
Gauri Seshadri
University of Utah
SLC, UT
and
Dale Gillette
NOAA
RTP, NC
Report Prepared for:
The WESTAR Council
9 Monroe Pkwy, Suite 250
Lake Oswego, OR 97035
01/01/03

ACKNOWLEDGEMENTS
The support and assistance of Bob Lebens (WESTAR), the project manager for
this study, were essential for the completion of this work and greatly appreciated. The
prompt reviews of the draft report by Duane Ono (Great Basin Unified APCD) and Mark
Scruggs (National Park Service) were very helpful and we are indebted for the
contribution of their time. The authors additionally thank Christopher Biloft, Michael
Brown (Los Alamos National Laboratory), M. Nelson (University of Utah), Dragon Zajic
(Arizona State University), Sean Ahonen, Mark Green, Judith Chow (Desert Research
Institute), Marc Pitchford (NOAA), Geary Schwimmer and Dave Miller (NASA), Clyde
Durham (FT. Bliss), and Thompson Pace (EPA) for contributing to the completion of this
work through equipment loans, comments on the manuscript, and assistance in data
analysis.
This work was funded primarily by the WESTAR Council. Experiments at the Ft.
Bliss military facility and Dugway Proving Ground were funded by DoD SERDP
contracts CP1190 and CP1191. Field testing was made possible through the cooperation
of the United States Army at the Dugway Proving Ground and at Ft. Bliss, the DoE
NNSA Chemical and Biological National Security Program, and the Mock Urban Setting
test participants. USEPA Southwest Center for Environmental Research and Policy
projects A-01-7 and A-02-7 provided data for supplemental analysis.
iii

EXECUTIVE SUMMARY
The Western States Air Resources Council (WESTAR) contracted the Desert
Research Institute (DRI) and the University of Utah (U of U), through subcontract, to
perform an investigation of a model that has been proposed by Dale Gillette of NOAA
(Countess, 2001) to account for near-source deposition of dust emissions from unpaved
roads. This work was motivated by the well-documented disagreement between estimates
of road and other geological dust obtained by emissions inventory methods and the actual
amount of inorganic minerals observed on filter samples at ambient monitoring sites
(Watson and Chow, 2000). This WESTAR project leveraged two existing DoD-funded
efforts (CP 1190, CP 1991) to increase the utility of data collection and analysis in the
context of meeting the objectives of this study.
The work presented in this report drew on two field studies with robust datasets,
three modeling approaches, and a review of the mechanisms of deposition and dispersion
and their parameterization in air quality models. The first field test took place at the Ft.
Bliss military facility near El Paso, TX in April, 2002. The setting was representative of
the desert Southwestern United States. Atmospheric conditions ranged from near-neutral
to moderately unstable. The second field test was conducted at the Dugway Proving
Ground in September, 2001 as part of the Mock Urban Experiment. The setting was
representative of areas with large roughness elements, similar to suburban tract homes.
Atmospheric conditions were stable. Common to both field tests were towers that were
instrumented with highly time-resolved particle concentration monitors at multiple
heights. This experimental setup allowed for calculation of horizontal dust flux
approximately 100 m downwind of the unpaved road test strip on a per vehicle basis.
The three models evaluated were the Gillette Box Model, a Gaussian plume
model similar to that used in EPA’s ISC, and a numerical solution to a simplified version
of the Atmospheric Diffusion Equation (ADE). Predictions from all three models were
compared with the data from the field studies.
There were six major conclusions from the work documented in this report.
First, the hypothesis that fugitive dust particles emitted from unpaved roads can
deposit to an appreciable extent within several hundred meters downwind of the road
(Watson and Chow , 2000; Countess, 2001) is well-founded, though the fraction of
particles removed varies greatly based on setting and atmospheric stability. In arid
regions such as the Southwestern United States, and for daytime emissions, the
transportable fraction of unpaved road dust emissions is likely to be near unity.
Second, for the purposes of modeling the transport and removal of dust particles
near an unpaved road source, the Gaussian-style models, such as EPA’s ISC3, provide an
imperfect, but reasonable preliminary approach. Comparison of modeled and measured
concentration profiles for desert, daytime conditions indicated that the ISC approximately
simulates the shape of the dust plume from an unpaved road. Similar comparison for
fugitive road dust emissions collected under stable conditions, amidst large roughness
elements, indicated that the ISC model may under estimate the deposition of dust in some
cases. The deposition velocity used by the model is difficult to verify with
measurements, and may be incorrect by a factor of two or more. This means that any
model used, especially in the absence of additional field data, would provide a rough
v

approximation at best. Solution of the Atmospheric Diffusion Equation did not provide
useful information for the first 1,000 or so meters downwind of the road. This was
attributed to the violation of some basic assumptions in the underlying mixing length
theory.
Third, the Gillette Box Model approach is elegant in its simplicity, but the
assumptions behind the model are probably, in general, not valid for unpaved road dust
emissions. Specifically, the assumptions required for the accurate parameterization of
dispersion are not met. Thus, the Box Model may be used as a “back of the envelope”
test for the importance of particle removal downwind of an unpaved road source, but it is
not likely to give reliable quantitative data in all cases.
Fourth, the height of the wake generated behind a moving vehicle has an effect on
the fraction of particles removed as the plume travels downwind, though this effect is
only significant for stable conditions. The turbulent wake of the vehicle increases in
vertical extent approximately linearly with the vehicle height. Considering that the main
processes, dispersion and deposition, are not well-quantified, inclusion of the turbulent
wake height in current modeling attempts appears to be of secondary importance
according to preliminary tests.
Fifth, measurement of emission factors for seven different vehicles with gross
weights varying from 1,622 kg to 23,636 kg were at odds with the silt-based emission
factors suggested in AP-42 (USEPA). The Ft. Bliss measurements indicated that
emission factors were highly dependent on vehicle speed. Linear regression of emission
factor vs. speed for each of the seven vehicles resulted in R 2 values ranging from 0.37 to
0.95, with six of the seven vehicles giving values greater than 0.70. In contrast, the AP-
42 emission factor formulation does not consider vehicle speed. According to the
measurements at Ft. Bliss, for passenger-sized vehicles, the AP-42 appears to
overestimate emissions for vehicle speeds under 35 mph. Emission inventories often
assume an average unpaved road travel speed of 20 or 25 mph. Under those conditions,
use of AP-42 would result in an overestimate of PM 10 emissions by 50% to 100% based
on the field study results.
Sixth, prior to applying a correction to existing emissions inventories, additional
field studies are recommended. In view of the good agreement between the ISC and data
collected in desert, daytime conditions but the poor agreement between the ISC and
measurements under stable conditions with large roughness heights, some selected field
studies that span a range of land cover and atmospheric stabilities are recommended. In
addition, there are a number of other possible sources of error that may result in inflated
emission inventories (Watson and Chow, 2000; Countess, 2001) that should be
considered as a source of the discrepancy between PM 10 dust emission inventories and
ambient concentrations. One finding of the present study is that the AP-42 tends to over
predict emissions of PM 10 dust for smaller, passenger-sized vehicles, especially when the
vehicles are traveling at speeds less than 35 mph.
Watson and Chow (2000) documented the discrepancy between emission
inventories for PM 10 fugitive dust and the source attribution of ambient filter samples.
Their analysis indicated that the amount of geologically-derived PM 10 found in the air is
much smaller than would be expected based on the emission inventories and dispersion
vi

models. Countess (2001) summarized eleven shortcomings in the current treatment of
fugitive dust emissions. Though not all eleven are directly relevant to the work in this
report, we address several of them below.
The removal of dust in the area near a source has been hypothesized to be
substantial because concentrations of particles downwind of the source rapidly attenuate
to the background. Watson and Chow (2000) and Countess (2001) both cite an earlier
study (Watson et al., 1996) where the concentration of PM 10 dust was found to decrease
by 90% only 50 m downwind from the source. The modeling effort in this study showed
that much of the decrease in concentration - indeed for the conditions at Ft. Bliss, most of
the decrease - is due to dispersion in the vertical direction. Therefore, using the
magnitude of the PM 10 concentration downwind of a dust source to estimate the removal
of particles will suggest a much higher deposition rate than is achieved in reality. This
does not invalidate the correct assumption that some of the PM 10 emitted will be removed
as the dust plume travels downwind of the source. However, the removal occurs to a
smaller degree and over larger distances than previously speculated, especially for the
open, sparsely vegetated terrain that much of the American southwest is comprised of.
The results of this study support the first finding put forth by Countess (2001),
“Not all suspendable particles are transported long distances”. However, the Ft. Bliss and
Dugway studies show a large range of near-field removal rates. In the rangeland setting
of Ft. Bliss, little removal of suspended dust was observed at a downwind distances of
100 m. This result was supported with a modeling effort that used a Gaussian approach
similar to the ISC model (USEPA, 1995). Nighttime experiments at the Mock Urban
Setting at the Dugway Proving Grounds suggested that under stable atmospheric
conditions and large roughness elements (cargo containers) 85% of the PM 10 may be
removed by deposition 95 m downwind of the unpaved road. This enormous range of
removal rates emphasizes that it is not appropriate to apply a single correction factor to
all fugitive dust emissions as a means of accounting for near-field particle removal.
Though not documented, the community of scientists and professionals have, in the last
several years, been circulating the idea that if fugitive dust emissions were divided by a
factor of four, then the discrepancy between emissions and ambient measurements of
geological PM 10 would disappear. While it is possible that this is true on an average
basis (i.e. over large spatial domains), it is unlikely that this factor of four is applicable to
every combination of air shed, land use distribution, and atmospheric conditions. Each
combination of setting and meteorological conditions should be considered separately in
a modeling framework that makes use of the known physics of particle dispersion and
deposition.
Related to the last point, the removal of PM 10 dust particles in the “impact zone”
immediately after emission, at the junction of the road edge and the first row of
vegetation, should be examined more closely. The work presented in this report only
considers deposition of particles from “above” the vegetative canopy. It does not
consider that particles may be removed to an appreciable extent by a process similar to
filtration when the dust plume is forced through vegetative elements.
As one of the recommendations from his second finding, “Regional-scale vertical
flux is smaller than the local-scale fugitive dust flux”, Countess (2001) suggests testing
the Gillette box model and modifying the model parameters if appropriate. In this study,
vii

the box model was examined, from a theoretical and practical standpoint. The analysis
indicated that the model is not practical for wide spread use because the parameterization
of dispersion is reliant on assumptions that cannot always be met in reality.
Related to the third finding from Countess (2001), “Fugitive dust emission factors
need to be appropriate”, the emission factors measured at Ft Bliss with high timeresolution
instruments suggest that AP-42 may overestimate PM 10 emissions from
unpaved roads. This overestimate was due to the lack of accounting for vehicle speed
and to the use of the average weight of all vehicles in calculating the emission factors.
All in all, this report concludes that there is no single, “silver bullet” solution to
the problem of reconciling fugitive dust emission inventories and ambient source
contribution estimates. Progress has been achieved in the quantification of the near-field
removal of dust particles and work in that area should continue. There are still
uncertainties about the AP-42 emission factors for unpaved road dust that have resurfaced
in this study and must be resolved.
viii

TABLE OF CONTENTS
1. INTRODUCTION 1-1
1.1 Background 1-1
1.2 Objectives 1-1
1.3 Approach 1-1
1.4 Report Organization 1-2
2. BACKGROUND 2-1
2.1 Dry Deposition 2-2
2.1.1 The Model of Slinn (1982) 2-4
2.1.2 Dry Deposition in the ISC3 2-12
2.1.3 The Model of Raupach (1999) 2-13
2.2 Dispersion in the atmosphere 2-14
3. MEASUREMENT OF DEPOSTION VELOCITY 3-1
3.1 Methods 3-1
3.1.1 Deposition on Flat Substrates 3-1
3.1.2 Deposition on Artificial Vegetation 3-4
3.1.3 Flat Substrate Method Testing 3-5
3.1.4 Vegetative Method Testing 3-5
3.2 Results: Deposition velocities measured by SEM and gravimetric techniques
3-6
3.3 Discussion 3-11
3.3.1 Comparison of results with the model of Slinn (1982) 3-11
3.3.2 Results in the context of the resistance model for deposition 3-13
3.4 Conclusions 3-14
4. MODELS USED TO STUDY PARTICLE REMOVAL BY DRY DEPOSITION
DOWNWIND OF AN UNPAVED ROAD 4-1
4.1 Presentation of Models 4-1
4.1.1 The one-dimensional Atmospheric Diffusion Equation (ADE) 4-1
4.1.2 The Integrated Source Complex Model 3 (ISC3) 4-4
4.1.3 The Gillette Box Model This section of the text was provided by Dr. Dale
Gillette (NOAA) with some minor adjustments by DRI 4-6
ix

4.2 Inter-comparison of the ADE, ISC3, and the Gillette Box Model 4-10
4.2.1 Analysis of the Gillette Box Model 4-11
4.2.2 Comparison of results for particle removal between predictions from ADE,
ISC3, and the Gillette Box Model 4-14
4.3 Summary 4-20
5. DETERMINATION OF PARTICLE REMOVAL RATES, EMISSION
FACTORS, AND INJECTION HEIGHTS ON AN UNPAVED ROAD IN AN ARID
SETTING 5-1
5.1 Methods 5-1
5.1.1 Upwind/Downwind Towers 5-1
5.1.2 Sonic Anemometer Tests 5-7
5.2 Results 5-10
5.2.1 Dispersion and deposition downwind of the unpaved road at Ft. Bliss 5-10
5.2.2 Comparison of Measured Emission Factors with AP-42 Emissions 5-17
5.2.3 Injection Height 5-20
5.3 Summary 5-24
6. DETERMINATION OF PARTICLE REMOVAL RATES UNDER STABLE
CONDITIONS IN A MOCK URBAN SETTING 6-1
6.1 Experimental Methods 6-1
6.2 Data Analysis Methods 6-2
6.2.1 Wind Speed Models 6-3
6.2.2 Vertical Dust Concentration Models 6-3
6.2.3 Calculation of Horizontal Flux 6-4
6.3 Results 6-4
6.3.1 Wind Profile 6-5
6.3.2 Dust Concentration 6-5
6.3.3 Horizontal Dust Flux 6-6
6.4 Discussion 6-8
6.4.1 Comparison of measurements with model results 6-8
7. SUMMARY AND CONCLUSIONS 7-1
7.1 The findings of this study in relation to recent work 7-5
8. REFERENCES 8-1
x

TABLE OF FIGURES
Figure 2-1. Development of a dust plume downwind of an unpaved road...................... 2-1
Figure 2-2. Illustration of resistance analogy to deposition. The numbers on the right side
of the figure give an order of magnitude estimate of the physical depth of each layer.
The arrow going through r g indicates that gravity only acts in the downward
direction. The equations for flux in the figure do not include the effect of gravity. 2-3
Figure 2-3. Deposition velocity vs. aerodynamic particle size using Slinn's (1982) model.
See Table 2-1 for parameters used........................................................................... 2-8
Figure 2-4. Contributions of Brownian motion, Interception, and inertial impaction to the
total removal efficiency. See Table 2-1 for parameters used................................... 2-8
Figure 2-5. Variation of Slinn (1982) deposition velocity with estimated parameters. a.
with assumed length scales of vegetation; b. with assumed rebound coefficient b; c.
with assumed viscous to total drag ratio c v /c d . See Table 2-1 for base case.......... 2-10
Figure 2-6. Variation of Slinn (1982) deposition velocity with wind conditions that obey
a logarithmic profile as in Equation 2-9. a. with roughness height z 0 while u * and d
held constant; b. with friction velocity u * while z 0 and d held constant c. with
assumed fraction of canopy height equal to turbulence scale at the top of the canopy
while u * and z 0 held constant. See Table 2-1 for base case ................................... 2-11
Figure 3-1 The experimental setup of the passive deposition on flat surfaces for
examination by electron microscopy. Three-axis (corresponding to the stream-wise,
crosswind, and vertical components of wind direction i.e. +x, +y, +z, -x, -y and –z
directions) passive collection substrates were mounted on a ring stand. Standard
right hand coordinates are used with the positive x-axis downwind perpendicular to
the road. A GRIMM inlet was placed right next to it to monitor the dust
concentration. The substrates were placed 1 or 2m high and 5 or 10m away from the
unpaved road............................................................................................................ 3-2
Figure 3-2. The experimental set up for passive deposition on artificial vegetation (Fir
Christmas garland). The plastic foliage was wrapped around a PVC frame, prewashed
and installed above a clean plastic tub placed 5 m away from the road and 1
m above grade. A GRIMM inlet was placed alongside it to monitor the dust
concentration............................................................................................................ 3-4
Figure 3-3. The deposition velocities calculated for the flat polycarbonate substrates (17,
18, 19) and the artificial vegetative samples (F002, C002) are compared to the
literature values (Slinn 1982). Slinn values are for wind velocity of 5 m/s and
gamma of 3.5. The deposition velocity predicted by the Slinn model is based on the
horizontal area, but the deposition velocity for the polycarbonate surfaces has been
calculated in terms of projected area. ..................................................................... 3-8
Figure 3-4. Directional variation in deposition velocity normal to a flat surface. Data
given are mean and one standard deviation, normalized by the deposition velocity on
the horizontal surface facing up (PZ) averaged for the range 0.5-7.5 m (physical
diameter) based on four determinations (SEM 12, 17, 18, 19). The vertical surfaces
(NX, NY, PY) show enhanced deposition due to impaction. Deposition on surface
facing down (NZ) and away from the road (PX) is presumed to be due to turbulence.
.................................................................................................................................. 3-9
xi

Figure 3-5. Directional variation in deposition velocity normal to a flat surface as a
function of particle size. Data are the mean of four determinations (SEM 12, 17, 8,
19), normalized by the deposition velocity on the horizontal surface facing up (PZ).
The plot indicates no consistent size dependence for particles less than 10 µm in
diameter.................................................................................................................... 3-9
Figure 3-6. . The grand average deposition velocity (SEM 17, 18, 19) for the SEM
samples versus the particle size. The dashed line represents one standard deviation
of the three determinations. Deposition velocity and standard deviation of replicates
data outside the 0.5 m to 5 m range are not shown due to uncertainty in the
accuracy of the data. For d5 m, the results are uncertain due to inlet line loss.
................................................................................................................................ 3-10
Figure 3-7. Variation in particle concentration between sampling days and between
instruments with different inlet configurations. SEM 12 data were obtained from a
GRIMM with a direct sample inlet and SEM 17, 18, 19 data were obtained from a
GRIMM with a 2.5 m long PVC tubing attached to its sample inlet. For diameters
less than 5 m (physical), both the inlets agree but at particle diameters greater than
5 m, the GRIMM with the long inlet tube exhibited sampling losses. Particle
losses affect deposition velocity calculations. ....................................................... 3-10
Figure 3-8. Deposition velocities calculated from SEM particle count for a dry
polycarbonate membrane and a clear tape surface that were collected simultaneously
(SEM 18). From 5 m to 10 m both the substrates give similar particle counts. The
count is increased (and therefore deposition velocity is increased) for particles d20 m) that fell off the dry substrate (data not shown). .................... 3-11
Figure 3-9. Measured deposition velocities on polycarbonate substrates and deposition
velocities calculated based on the boundary layer resistance r b only vs. particle
aerodynamic diameter. The average u * for the sample periods represented by the
measured data and used for the theoretical prediction was 0.35 m/s..................... 3-14
Figure 4-1. Control Volume for Fugitive Dust Model Depicting Vertical and Horizontal
Fluxes....................................................................................................................... 4-6
Figure 4-2. Representation of the Gillette Box Model in the context of the resistance
model to transport. The dashed horizontal line effectively divides the resistance r a
into two smaller resistances, r a1 and r a2 . The equations for fluxes in the figure do not
include the effect of gravity. .................................................................................. 4-12
Figure 4-3. The total aerodynamic resistance from the ground up to the height indicated
on the x-axis. R a calculated after Byun and Dennis (1995). L is the Monin-Obukhov
length and indicates stability. (100,000 > ⏐L⏐ > 10,000 near neutral, 10,000 > L
> 10 stable, 10 > L > 0 very stable, 0 > L > -100 very unstable, -100 > L >
-10,000 unstable). U * = 0.2 m/s and z 0 =0.01 m in all cases.............................. 4-13
Figure 4-4. Fraction of particles remaining in suspension vs. time since emission in the
near field (0

Figure 4-6. The fraction of particles remaining in suspension and the Concentration at a
height of 1 meter above ground level vs. time for neutral atmospheric conditions
according to the ADE and ISC3 models. z 0 =0.01 m, D p =8 µm, u * = 0.3 m/s in all
cases. ...................................................................................................................... 4-17
Figure 4-7. Fraction of particles remaining in suspension for a given downwind distance.
Panels a through e correspond to numerical solutions to the atmospheric diffusion
equation with z 0 =0.01 m, D p =8 µm. under various conditions of atmospheric
stability. The downwind distance is estimated based on the wind speed at 10 m
assuming a logarithmic profile and z 0 = 0.01 m. Panel f shows the fraction
remaining according to the box model for various assumed values of A G ............ 4-18
Figure 4-8. Fraction of particles remaining in suspension vs. time since emission for 8
µm particles falling under the influence of gravity in the absence of any dispersion
(u * =0). .................................................................................................................... 4-19
Figure 5-1. Schematic of equipment setup at Ft. Bliss during the April, 2002 field
campaign. The vertical and horizontal components are drawn on different scales. 5-2
Figure 5-2. Example of time series of DustTrak PM 10 concentrations after the passage of
a vehicle on the unpaved road at Ft. Bliss. The arrows in the figure illustrate the start
and stop times estimated for a baseline reading and the dust plume passing through
the downwind towers DT_1, DT_2, and DT_3. ...................................................... 5-4
Figure 5-3. Frequency distribution and cumulative distribution of u * for the 110 15-
minute intervals corresponding to periods when testing was ongoing at Ft. Bliss. The
u * values are based on a roughness height, z 0 equal to 0.005 m.............................. 5-5
Figure 5-4. Examples of comparison between measured wind speed and wind speed
calculated from curve fitting of Equation 2-9. a. light winds (u * =10 cm/s), b.
medium winds (u * = 32 cm/s), c. high winds (u * = 54 cm/s). The roughness height,
z 0 is equal to 0.005 m in all cases. ........................................................................... 5-5
Figure 5-5. Comparison between collocated spinning cup and sonic anemometers. a.
Example time series from 4/18/02; b. regression of 1-second-averaged and 1-
minute-averaged data from 4/18/02; c. regression of 1-second-averaged and 1-
minute-averaged data from 4/19/02; d. regression of 1-second-averaged and 1-
minute-averaged data from 4/20/02......................................................................... 5-9
Figure 5-6. Example of W-velocity as measured by the sonic anemometer in the vicinity
of a vehicle pass. The vertical lines indicate the start and end times for the two
background intervals and the influence interval. ................................................... 5-10
Figure 5-7. Comparison of concentration profiles from the passage of two convoys with
the concentration profiles at equivalent friction velocities based on the average from
multiple passes of a moving point source. The line source is approximated by the
passage of two convoys on two separate occasions. The moving point source
average concentration profiles were calculated as the integrated concentration from
Equation 5-5. Concentrations are normalized to the value at 1.26 meters in all cases.
................................................................................................................................ 5-12
xiii

Figure 5-8. Concentration profiles measured at the three towers downwind of an unpaved
road and concentration profiles predicted by the ADE model for equivalent
downwind distances a. for 32 passes with u * between 0.15 and 0.25 m/s and b. for 15
passes with u * between 0.45 and 0.55 m/s. Tower distances from edge of road: T1 -
7 m, T2 – 50 m, T3 – 100 m. Modeled profiles are based on neutral atmospheric
conditions. Both measured and modeled concentration have been normalized to the
concentration at 1.26 m.......................................................................................... 5-13
Figure 5-9. Normalized measured and modeled gaussian concentration profiles at a.
Tower 2, 50 meters downwind of unpaved road and b. Tower 3, 100 meters
downwind............................................................................................................... 5-14
Figure 5-10. Comparison of ISC3 gaussian concentration profile and profile from
numerical solution of Atmospheric Diffusion Equation at multiple downwind
distances. Solutions for the ADE are for neutral conditions................................. 5-15
Figure 5-11. Particle removal rates measured at Ft. Bliss at tower 100 m downwind of an
unpaved road. a. Comparison of horizontal flux among the three downwind towers.
Data are averages over 27 individual passes when the wind direction was within 15
degrees of perpendicular to the road. Concentrations measured with DustTrak
monitors with PM 10 inlets. Fluxes calculated according to Equation 5-1 and
normalized to value at Tower 1. Vertical bars indicate standard errors; b.
Comparison of particle size distributions at Towers 1 and 3 measured by GRIMM
1.108 OPC with model predictions. Data are averages over 137 individual passes. X-
axis shows particle aerodynamic diameter calculated by assuming a density of 2.6
g/cm 3 and geometric mean diameter representing each size bin. Particle
concentrations at Towers 1 and 3 were each normalized by concentration of 3.9
micron particles. Y-axis represents ratios of normalized concentrations. ............ 5-16
Figure 5-12. Example of emission factor dependence on vehicle speed. The vertical bars
are the standard deviations among the three downwind towers. ........................... 5-18
Figure 5-13. The ratio of emission factors calculated according to AP-42 (USEPA, 1999)
and those measured at Ft. Bliss. The upper line is the ratio based on a 7% silt
content, while the lower line assumes 4% silt content. The top panel shows the ratios
when all the vehicles in Table 5-2 are considered. The bottom panel shows the same
data, but only for the passenger vehicles. .............................................................. 5-20
Figure 5-14. Turbulence generated vs. height above ground for three vehicle types. The
x-axis represents the difference between the vertical component of turbulence
generated by the passing of the vehicle and the background fluctuations in vertical
velocity normalized by the speed of the vehicle. The horizontal lines represent the
background turbulence standard deviation. The dotted lines represent hand drawn
curves to fit the data. The dotted line corresponding to the compact car was drawn
based on the assumption that the height of the wake plume is proportional to the
height of the vehicle (see Figure 5-15). ................................................................. 5-21
Figure 5-15. Estimate of turbulent wake height vs. physical vehicle height. The dotted
line represents a zero-intercept regression on the data from the Cargo Van and the
Moving Truck. The hollow circle is an estimate of the wake height for the Compact
Car based on the regression. .................................................................................. 5-21
xiv

Figure 5-16. Comparison of fraction of particles removed according to the ISC3 model
for various values of the injection height under a. moderately unstable conditions; b.
neutral conditions; c. very stable conditions; and d. same as c. but x-axis extends to
10 km. The injection height is manifested in the model by setting the initial value of
σ z to 0.5×IH. .......................................................................................................... 5-23
Figure 6-1. Plan view of the MUST site showing the array of shipping containers
representing buildings and the location of the measurements. (1) 2-D sonic
anemometers at 4, 8, 16 m; (2) DustTraks at 0.9, 1.7, and 3.7 m; (3) 3-D sonic
anemometer at 1.6 m; (4) DustTraks at 1.8, 4.6, 9.1, and 18.3 m and 2-D sonic
anemometers at 4, 8, 16, 32 m. ................................................................................ 6-1
Figure 6-2. Wind Equation Fit. Both the logarithmic (dashed line) and power law (solid
line) equations give a reasonable fit to the 15-minute average wind speed
measurements (circles)............................................................................................. 6-6
Figure 6-3. Dust concentration versus time as measured by DustTraks. Vehicle passes
generated well-defined spikes at 3 m horizontal and 1 m vertical from the road
(bottom). The spikes were broader but still well defined 95 m downwind and 4.6 m
above grade (middle). Most vehicle trips did not cause a noticeable spike at 95 m
downwind and 18.3 m above grade (top). Note that full scale for the top graph and
middle graph are 1/1000 and 1/50 of full scale for the near road measurements
respectively. ............................................................................................................. 6-7
Figure 6-4 Collocated DustTraks. Five collocated DustTraks showed large differences in
individual readings but little difference between the instruments when averaged over
the entire experiment. Left: Histogram of maximum minus minimum divided by the
average for sets of five coincident readings. Middle: outlier box plot of the same
data. Right: a typical time trend showing poor correlation between individual
readings.................................................................................................................... 6-8
Figure 6-5. Comparison of fraction of PM 10 particles remaining in suspension 95 m
downwind of an unpaved road during a nighttime test on Dugway Proving Grounds,
UT. The white circles represent the fraction of PM 10 associated with each size bin.
Mass fractions associated with each size bin were estimated from Ft. Bliss data from
a different experiment. The gray circle is the fraction remaining in suspension
according to the methods described in the previous section. The black squares and
triangles correspond to modeled removal rates under very stable conditions. ...... 6-10
xv

xvi

LIST OF TABLES
Table 2-1. Parameters used in the Slinn (1982) model for estimating the deposition
velocity over a bean crop. ........................................................................................ 2-9
Table 3-1 Cumulative screen analysis for the soils samples collected before and after the
sampling period........................................................................................................ 3-1
Table 3-2. Log of all flat substrate samples..................................................................... 3-7
Table 3-3. Log of all field vegetative samples................................................................. 3-8
Table 3-4. Deposition velocity calculations for vegetative samples using different areas:
1. Deposition velocity calculation using total area:.............................................. 3-12
Table 3-5. Deposition Velocity calculated for vegetative samples................................ 3-13
Table 4-1. Wind speed at 10 meters corresponding to given friction velocity and
roughness height .................................................................................................... 4-15
Table 5-1. Test dates, times, vehicles, data recovery, and conditions during the FT. Bliss,
April 2002 field campaign. ...................................................................................... 5-6
Table 5-2. Summary of vehicle emission factors measured at Ft. Bliss and Comparison
with emission factors calculated as specified in AP-42 (USEPA, 1999) for two
values of silt content (4% and 7%); moisture content was assumed to be 0.2%. The
speed dependence of the measured emission factors is given in the third column and
represents the slope of a linear regression between measured emission factor (g/vkt)
and vehicle speed (mph); the R 2 value for the regression is given in the fourth
column. For cases where a particular vehicle was tested more than once, the average
of the two tests is used. All emission factors are reported in grams per vehicle
kilometer traveled (g/vkt). ..................................................................................... 5-19
Table 6-1. Screen analysis of test road surface material. Average of two soil samples. 6-2
Table 6-2 Meteorological data from station DPG08, approximately 1 km south of test
site, for September 26, 2002 01:00-03:00 MDT...................................................... 6-4
Table 6-3. Mean ± one standard deviation wind and concentration data for the entire 1.5
hr test period with 44 vehicle trips. Wind standard deviation based on variation
between 15-minute averages. Concentration standard deviation based on variation
between individual vehicle trips. Pulse area is the time-integrated concentration for
the interval corresponding to one vehicle pass. ....................................................... 6-5
xvii

xviii

1. INTRODUCTION
1.1 Background
The Western States Air Resources Council (WESTAR) contracted the Desert
Research Institute (DRI) and the University of Utah (U of U), through subcontract, to
perform an investigation of a model that has been proposed by Dale Gillette of
NOAA(Countess, 2001) to account for near-source deposition of dust emissions from
unpaved roads. This work was motivated by the well-documented disagreement between
estimates of road and other geological dust obtained by emissions inventory methods and
the actual amount of inorganic minerals observed on filter samples at ambient monitoring
sites (Watson and Chow, 2000). This WESTAR-funded project leveraged two existing
DoD funded studies (CP 1190, CP 1991) to increase the utility of data collection and
analysis in the context of meeting the objectives of this study.
1.2 Objectives
Broadly, the goal of this study was to provide insight into the magnitude of PM 10
dust removal by deposition close to the emission source. The analysis was intended to
focus on emissions from unpaved roads, though the results may be pertinent to other
sources of fugitive dust. The specific objectives of the study were:
1. Provide recommendations for feasibility of the Gillette Box Model as a
tool for correcting emissions inventories for fugitive dust from unpaved
roads and suggest possible modifications that may improve model
performance.
2. Compare the results of field studies with the Box Model and with the
algorithm used in ISC3.
3. Provide recommendations for modeling PM 10 dust deposition and
removal and suggest areas to target in future work
1.3 Approach
To achieve the stated objectives, this project drew on two field studies with robust
datasets, three modeling approaches, and a review of the mechanisms of deposition and
dispersion and their parameterization in air quality models. The first field test took place
at the Ft. Bliss military facility near El Paso, TX. The setting was representative of the
desert Southwestern United States. Atmospheric conditions ranged from near-neutral to
moderately unstable. The second field test was conducted at the Dugway Proving
Ground as part of the Mock Urban Experiment. The setting was representative of areas
with large roughness elements, similar to suburban tract homes. Atmospheric conditions
were stable.
The three models evaluated were the Gillette Box Model, a Gaussian plume
model similar to that used in EPA’s ISC, and a numerical solution to a simplified version
of the Atmospheric Diffusion Equation (ADE). Predictions from all three models were
compared with the data from the field studies.
1-1

1.4 Report Organization
This report is divided into six main Chapters. Chapter 2 reviews the literature
with respect to models for particle deposition and turbulent dispersion. Pilot-scale
measurement of deposition to surrogate surfaces and comparison of results with theory
are summarized in Chapter 3. In Chapter 4, the three models considered in this report are
introduced and analyzed. The assumptions of the Box Model and the characteristics of
the ADE and ISC are examined in this Chapter. The Ft. Bliss field experiments are
presented in Chapter 5. In addition to the horizontal flux and particle concentration
profiles, this Chapter details the measurement of vehicle-induced turbulence and its effect
on particle removal downwind of the source. Model predictions are compared with
measurement and used to characterize the effect of the “injection height”. We summarize
The Dugway Proving Ground experiments in Chapter 6 and analyze the related model
results. A summary of findings and major conclusions is provided in Chapter 7.
1-2

2.BACKGROUND
Fugitive dust is emitted from an unpaved road when a vehicle passes and disturbs
the soil. A cloud of dust is raised behind the vehicle and begins to travel downwind.
Initially, the dust cloud is very dense, sometimes to the point of being opaque, but with
travel downwind, the cloud is dispersed by mechanical mixing at the ground and by
buoyant eddies in the atmosphere. Particles suspended in the cloud can be removed by
interaction with the vegetative cover in the downwind fetch of the unpaved road.
Figure 2-1 schematically illustrates the progression of a dust plume emitted
behind a vehicle and advected downwind over a vegetative cover. In the region (A)
where the dust cloud first meets the vegetation, dust particles may be removed by
individual vegetation elements. Though possibly significant, none of the models or
measurements discussed in this report is geared towards quantifying the extent of
removal in this “impact zone”. Some work in the area of particle removal by windbreaks
may be applicable to this region (Raupach and Leys, 1999), though the specific
formulation would have to be adjusted for vegetative covers with a long fetch (as
opposed to a windbreak that is only several meters deep).
Region C: Far Downwind
50 m < x < ∞
Region B: Near Source
5 m < x < 200 meters
Region A: Impact Zone
0 < x < 20 m
Point of
Emission
Figure 2-1. Development of a dust plume downwind of an unpaved road
Very far downwind (Region C in the figure), the dust plume approaches a steady
concentration profile that does not change very much with transport downwind. In
between the “impact zone” and the “Far downwind” region, the concentration profile of
dust particles is changing with transport distance; specifically, the dust plume is
expanding in the vertical direction. The models discussed in this report are most
concerned with this region, where it is necessary to simulate the expansion of the dust
plume in the vertical direction simultaneously with the removal of particles at the ground
by the vegetative cover. These two processes are turbulent dispersion and dry deposition,
respectively.
2-1

2.1 Dry Deposition
There are a number of models available for dry deposition of particles. Three of
them are presented below. Because it is widely cited and because it has been suggested
for use in the Gillette Box Model, the formulation provided by Slinn (1982) is discussed
at some length. One of the difficulties of using Slinn’s model is that the parameters
required are numerous and difficult to specify. Simpler methods that are based on similar
reasoning, but requiring fewer parameters, are used in the Industrial Source Complex
utility (USEPA, 1995) and by Raupach (1999).
Dry deposition is frequently modeled in analogy to resistance in an electrical
circuit. The flux to the ground is assumed to be equal to the concentration measured at a
given height multiplied by the mass transfer coefficient, which is also dependent on
height. The mass transfer coefficient is called the deposition velocity and is given the
symbol v d . Thus, according to the resistance model
F = C( z)
v ( z)
d
Equation 2-1
where F is the flux of material depositing to the ground (g/m 2 s) and C(z) is the heightdependent
concentration. The deposition velocity is given as the inverse of three
resistances in series. That is,
1
vd
=
ra
+ rb
+ rc
Equation 2-2
where r a is the resistance to aerodynamic transport through the surface layer, r b is the
resistance to transport through the “quasi-laminar sublayer”, and r c is the resistance to
collection on vegetative elements (The subscripts a, b, and c are unrelated to the Regions
in Figure 2-1). Figure 2-2 shows the resistance model to deposition. It is built on the
assumption that close to the ground (i.e. from the top of the surface layer down), the flux
of a depositing species is constant with height. This is analogous to the flow of electrons
through a circuit where the number of electrons that leave the positive terminal (top of
the surface layer) must be matched by the number arriving at the negative terminal (the
deposition surface). Related to the assumption of constant flux, the model also assumes
that the system is approximately at steady-state. That is, the concentration profile through
the surface and quasi-laminar layers is not changing significantly over time. This
corresponds strictly only to region C in Figure 2-1, though it may also hold
approximately for region B.
It is usually assumed that for particles with aerodynamic diameters less than 10
µm, r c is very small. That is, once a particle impacts on a collector, it sticks to that
collector and is permanently removed from the atmosphere. Deviations from this ideal
are addressed by Slinn (1982) as summarized in section 2.1.1.
2-2

Planetary Boundary Layer (PBL)
F = F(z)
~200-2000 m
C = C SL
r a
Surface Layer
F = F s =V da [C SL -C QL ]=(1/r a ) [C SL -C QL ]
~2-20 m
r g
r b
C = C QL
Quasi-laminar sublayer
F = F s =V db [C QL -C 0 ]=(1/r b ) [C QL -C 0 ]
C = C 0 =0
~0.1-1 mm
r c
r c assumed = 0 for particles if particle rebound is
negligible
Figure 2-2. Illustration of resistance analogy to deposition. The numbers on the right side of the
figure give an order of magnitude estimate of the physical depth of each layer. The arrow going
through r g indicates that gravity only acts in the downward direction. The equations for flux in the
figure do not include the effect of gravity.
In addition to turbulent transport through the surface and quasi-laminar layers,
particles are also influenced by gravity. This effect is accounted for in the resistance
model by assuming that gravitational settling represents another resistor (r g ) acting in
parallel with those shown in Figure 2-2. With the surface resistance r c equal to zero and
the inclusion of gravity, the equation used to describe the deposition velocity is
v 1
d
=
v
r + r + r r v
+
a
b
a b
g
g
Equation 2-3
where v g is the gravitational settling velocity and is equal to 1/r g . V g is a well-known
function of the particle size and the characteristics of the surrounding fluid (air). For
particles with aerodynamic diameter less than about 20 µm, it is given by (e.g. Seinfeld
and Pandis, 1998)
v
g
=
1
18
2
D ρ gC
p
p
µ
c
Equation 2-4
2-3

where D p is the particle diameter, ρ p is the particle density, g is the gravitational constant,
µ is the viscosity of air, and C c is the Cunningham slip correction factor,
C
c
2λ
1.1D
= + + −
p
1 1.257 0.4exp
D
2 λ
p
Equation 2-5
with λ equal to the mean free path of air molecules.
2.1.1 The Model of Slinn (1982)
In a seminal paper, Slinn (1982) described the framework for a model to predict
particle dry deposition to vegetative canopies. The model is based on the mass balance
expression
∂ ∂C
=
z
Κ
∂ ∂z
( c αu) C
d
ξ c
Equation 2-6
where K is the turbulent diffusivity in the vertical direction, C is the mass or number
concentration of particles, c d is the average drag coefficient for vegetation, α is the
surface area of vegetation per unit volume, u is the mean wind speed, and ξ c is the
canopy element collection efficiency for particles.
The LHS of Equation 2-6 represents the flux of particles through a horizontal
plane located at a height z above the ground while the RHS is the removal rate of
particles by dry deposition. It is implied in the form of Equation 2-6 that turbulent
dispersion only serves to bring particles from high above the canopy towards the ground.
Practically, this means that the particle concentration profile above the vegetative canopy
is well-developed and at steady state (i.e. not changing in time). This situation
corresponds strictly only to Region C in Figure 2-1. However, if the deposition rate
calculated is assumed to apply only locally (i.e. not constant in the downwind direction),
then we may extend the applicability of Equation 2-6 into Region B.
The final form of the deposition velocity provided by Slinn (1982) includes a term
to account for gravitational settling:
v
d
= v
g
+ C
D
uh
ur
1
+
ur
ξ +
1−ξ
ξ tanhγ
ζ
−1
Equation 2-7
where C D is the overall drag coefficient for the canopy, u h is the wind speed at the top of
the canopy (h being equal to the canopy height), u r is the wind speed at some reference
height z r above the canopy, ξ is the location-independent canopy collection efficiency,
and γ is a parameter that characterizes the wind profile within the canopy.
2-4

C
D
u* =
ur
The overall drag coefficient C D is given by
2
Equation 2-8
where u * is the friction velocity which may be estimated if the velocity profile above the
canopy is approximately logarithmic with either
u −
= *
z d
u(
z)
ln
κ z0
or
u z)
= u
*
(
r
u zr
−
− ln
κ
z −
( h − l)
( )
h − l
Equation 2-9
Equation 2-10
where κ is the von Karman constant and is equal to 0.4, d is the displacement height, z 0 is
the roughness height, h is the canopy height, and l is the length scale of turbulence at the
top of the canopy. The displacement height d can be thought of as the height above which
the velocity profile is expected to have a logarithmic shape. Under some conditions, the
displacement height and the roughness height z 0 may be difficult to estimate. For this
reason, it may be preferable to use Equation 2-10, which utilizes characteristics of the
canopy, namely the canopy height and the length scale of turbulence at the top of the
canopy. Slinn suggests the approximation
d = h − l .
The ratio u h /u r in Equation 2-7 is given by Equation 2-9 with h-d=l
Equation 2-11
u
u
h
r
=
u *
l
ln
κur
z
0
Equation 2-12
or by Equation 2-10:
u
u
h
r
=
u z
− ln
u κ
1
*
r
r
( h − l)
−
l
.
Equation 2-13
2-5

The particle collection efficiency ξ is given by
ξ = EB = ( E + E E )B
B IN
+
IM
Equation 2-14
where E B , E IN , and E IM are the removal efficiencies due to Brownian diffusion, particle
interception, and inertial impaction, respectively and E is the total removal efficiency due
to the three processes operating concurrently. The factor B (≤1) represents a reduction in
capture efficiency due to particles bouncing off of surfaces.
E
B
The Brownian diffusion component can be estimated from
cv
=
Sc
cd
− 2 3
Equation 2-15
where c v /c d is the ratio of viscous to total drag (equal to the sum of viscous drag and form
drag) and assumed to be 1/3 by Slinn and Sc is the Schmidt number (Sc =ν/D, ν is the
kinematic viscosity of air and D is the Brownian diffusivity of the particle given by the
Stokes-Einstein relation). Note that Brownian diffusion is expected to be a negligible
pathway for deposition for particles with diameters greater than 1 µm.
E
IN
c
=
c
The interception term E IN is calculated from
v
d
F
s
D
p
Dp
A
+
s
2
+
( 1−
F )
s
D
p
Dp
A
+
L
2
Equation 2-16
where F s is the fraction of the total interception due to small collectors in the canopy such
as hairs, A s is a characteristic width of the small collectors, A L is a characteristic radius of
the large collectors such as blades, stalks, and needles.
E
IM
The Impaction term E IM is equal to
2
Sta
1+
Sta
=
2
where St a is an “average” Stokes number defined by
Equation 2-17
2-6

St
a
vg
u
g
=
cA
L
*
Equation 2-18
with c being an unspecified constant that is expected to be approximately equal to unity.
B = exp − b
The factor B is estimated from
( )
St a
where b is another unspecified constant.
Equation 2-19
The factor γ used to represent the wind conditions within the canopy can be
approximated by
( u u )
*
γ ≅
κ
or
h
h
h − d
Equation 2-20
1 2
c d
α
γ ≅ h .
κ l
Equation 2-21
Note that the height-dependent α is related to the bulk Leaf Area Index (LAI), defined as
the ratio of upward facing leaf surface to ground surface in a canopy, according to
1
LAI =
2
0
h
α ⋅ dz .
Equation 2-22
The parameters that are used for an example of deposition to a bean crop are
given by Slinn (1982) and summarized in Table 2-1. Figure 2-3 demonstrates the classic
v-shaped deposition velocity curve as a function of particle size for unit density particles
(ρ p =1 g/cm 3 ); particles that are less than 0.1 µm in diameter deposit primarily due to
Brownian motion; particles in the 1 µm to 10 µm size range deposit by impaction;
intermediate-sized particles (0.1 µm to 1 µm) are removed by a combination of Brownian
motion, interception, and impaction. Figure 2-4 shows the relative contributions of E B ,
E IN , and E IMP to the total collection efficiency E T for the example of flow over a bean
crop. Though not shown in Figure 2-4, the gravitational settling term v g dominates the
2-7

1E+01
1E+00
vd (cm/s)
1E-01
1E-02
1E-03
1E-03 1E-02 1E-01 1E+00 1E+01 1E+02
Dp (microns)
Figure 2-3. Deposition velocity vs. aerodynamic particle size using Slinn's (1982) model. See Table 2-1
for parameters used.
1E+01
Etotal, Ebrownian, Einterception,
Eimpaction
1E+00
1E-01
1E-02
1E-03
E_total
E_Brownian motion
E_Interception
E_Impaction
1E-04
1E-04 1E-03 1E-02 1E-01 1E+00 1E+01 1E+02
Dp (microns)
Figure 2-4. Contributions of Brownian motion, Interception, and inertial impaction to the total
removal efficiency. See Table 2-1 for parameters used.
RHS of Equation 2-3 for particles larger than 10 µm and for particles between 1 µm and
10 µm under low wind conditions.
It is clear from examination of Table 2-1 that use of Slinn’s model requires the
specification of a number of parameters that may not be available. For example, the
parameters F s , A s , A L , c, b, and c v /c d are generally unavailable for a wide range of
vegetative covers. These parameters are roughly estimated by a best-guess approach.
Figure 2-5 illustrates the sensitivity of the model with respect to uncertainties in these
parameters. The Figure is confined to the vicinity of the particle size range relevant for
PM 10 dust. F s and A s only affect the deposition velocity through the interception term E IN
(Equation 2-16). Since interception is a minor pathway for super micron particles, the
effect of varying these two parameters is small. In contrast, changing the size of A L from
1 mm to 0.5 mm measurably improves the impaction efficiency and the overall collection
efficiency. It is unclear how the coefficient b should be chosen for a given canopy.
Rebound for particles with diameter less than ≈ 5 µm appears to be a minor
consideration. The effect of increasing the value of b from 2 to 4 is to decrease the overall
deposition velocity for 10 µm particles by about 30%. Varying the ratio of the viscous
drag to the total drag between 0.25 and 0.45 exerts little effect on the deposition
velocities for particles in the 1 – 10 µm size range.
2-8

Table 2-1. Parameters used in the Slinn (1982) model for estimating the deposition velocity over a
bean crop.
Parameter Description Equation used Value in Base
Case
Fraction of total interception by
User-specified 1%
F s
small collectors such as vegetative
hairs
A s
Characteristic width of small
collectors such as vegetative hairs
User-specified
10 -5 m
(10 microns)
Characteristic radius of large
collectors such as grass blades,
User-specified
10 -3 m
(1 mm)
A L
stalks, needles, etc
Unknown numerical factor used in
User-specified 1
c
Equation 2-18 and assumed to be
near unity
Numerical factor used to calculate
User-specified 2
b particle rebound in Equation 2-19
Ratio of viscous to total drag
User-specified 1/3
c v/c d
within the canopy
h Height of the canopy User-specified 1.18 meters
Fraction of canopy equal to
User-specified 20%
F l
characteristic Eddy size
u * Friction velocity User-specified 0.25 m/s
Roughness height User-specified or estimated from knowledge of z r, u r, 0.054 m
z 0
h, and l
Wind speed at reference height z r User-specified or estimated from u *, z 0, and d and 1.94 m/s
u r
Equation 2-9 or from u *, z r, u r, h, and l, and Equation
2-10
Wind speed at canopy height h Estimated from l and z 0 and Equation 2-12 or from u r, 0.81 m/s
u h
u *, h, and l and Equation 2-13
d Displacement height User-specified or estimated from h and F l and Equation 0.94 m
2-11
l Characteristic Eddy size at the top
Estimated from F l and h
0.24 m
γ
of the canopy
Parameter that describes the wind
speed profile within the canopy
expected to be between 2 and 5
User-specified or estimated from u *, u h, h, and d and
Equation 2-20 or from c d, l, h, and α and Equation
2-21
3.8
Slinn (1982)
suggests using
γ=3. However,
this assumption
for γ is not
consistent with
Equation 2-20
The deposition velocity is sensitive to the assumed wind conditions. The
roughness height, z 0 , and the displacement height, d, are mathematical tools used to fit
the wind speed to a logarithmic profile such as in Equation 2-9; they are not directly
measurable quantities. In general, z 0 varies from 10 -5 m for smooth ice to 10 m for a
high-density urban area (Seinfeld and Pandis, 1999). Vegetative covers span the range
between 0.01 m for cut lawn to 1 m for a tree-covered area. The displacement height is
related to the canopy height, but can only be roughly estimated without field
measurement of the wind profile above the canopy. Slinn (1982) makes the
approximation that the displacement height is equal to 80% of the canopy height and that
the sum of d and the characteristic Eddy size, l, is equal to the canopy height (i.e. l=0.2h).
U * is also a “fitted” quantity and specifying its value requires some knowledge of the
wind speed profile. Figure 2-6 shows that there can be significant deviations in v d when
slightly different values for z 0 , u * , and l are assumed. Thus, without accurate data for
these parameters, the deposition velocity may only be known to within a factor of about
two or three.
2-9

1E+01
Fs = 1%; As = 10 microns; AL=1 mm
Fs = 0.1%; As = 10 microns; AL= 1 mm
1E+00
Fs = 1%; As = 10 microns; AL=0.5 mm
vd (cm/s)
1E-01
1E-02
1E-03
1E-01 1E+00 1E+01 1E+02
Dp (microns)
a.
1E+01
1E+00
b=1
b=2
b=4
vd (cm/s)
1E-01
1E-02
1E-03
1E-01 1E+00 1E+01 1E+02
Dp (microns)
b.
1E+01
1E+00
cv/cd=0.25
cv/cd=0.33
cv/cd =0.45
vd (cm/s)
1E-01
1E-02
1E-03
1E-01 1E+00 1E+01 1E+02
Dp (microns)
c.
Figure 2-5. Variation of Slinn (1982) deposition velocity with estimated parameters. a. with assumed
length scales of vegetation; b. with assumed rebound coefficient b; c. with assumed viscous to total
drag ratio c v /c d . See Table 2-1 for base case.
2-10

1E+01
1E+00
z0 = 4.4 cm
z0 = 6.4 cm
z0 = 8.4 cm
vd (cm/s)
1E-01
1E-02
1E-03
1E-01 1E+00 1E+01 1E+02
Dp (microns)
a.
1E+01
1E+00
u* = 0.2 m/s
u* = 0.25 m/s
u* = 0.3 m/s
vd (cm/s)
1E-01
1E-02
1E-03
1E-01 1E+00 1E+01 1E+02
Dp (microns)
b.
1E+01
1E+00
l = 0.15 h
l = 0.2 h
l = 0.3 h
vd (cm/s)
1E-01
1E-02
1E-03
1E-01 1E+00 1E+01 1E+02
Dp (microns)
c.
Figure 2-6. Variation of Slinn (1982) deposition velocity with wind conditions that obey a logarithmic
profile as in Equation 2-9. a. with roughness height z 0 while u * and d held constant; b. with friction
velocity u * while z 0 and d held constant c. with assumed fraction of canopy height equal to turbulence
scale at the top of the canopy while u * and z 0 held constant. See Table 2-1 for base case
2-11

Aside from the error associated with improperly specifying parameters, it is
unclear if the assumptions in Slinn’s model apply to sparsely vegetated landscapes where
inter-plant distances may be several times the height of individual plants. The model was
derived for a “horizontally-homogeneous” canopy, where quantities are dependent only
on the distance z above the ground.
2.1.2 Dry Deposition in the ISC3
For a stable atmosphere, the aerodynamic resistance r a is given by (Byun and
Dennis, 1995):
1 zr
ra = ln
+ 4. 7
κu*
z0
z
L
Equation 2-23
while for an unstable atmosphere, it is given by
r
a
1
=
κu
ln
( 1+
16z
)( )
( )( )
r
/ L −1
1+
16z0
/ L + 1
1+
16zr
/ L + 1 1+
16z0
/ −1
* L
Equation 2-24
where L is the Monin-Obukhov length scale (L > 0 stable conditions, L < 0 unstable
conditions, ⏐L⏐>> 1 neutral conditions). The reference height z r is taken to be the
larger of 1 m and 20×z 0 .
1995):
r
b
=
The quasi-laminar resistance is given by (Pleim et al. 1984; Slinn, 1982; USEPA,
1
−
( 2 3 −3
St
Sc + 10 ) u *
Equation 2-25
where Sc is the Schmidt number (Sc =ν/D, ν is the kinematic viscosity of air and D is the
Brownian diffusivity of the particle given by the Stokes-Einstein relation) and St is the
Stokes number in a slightly different form than Equation 2-18
St =
( v g)
g
υ
2
u *
Equation 2-26
The term containing the Schmidt number in Equation 2-25 accounts for Brownian
motion while the term containing the Stokes number accounts for inertial impaction.
Interception is not explicitly accounted for in this formulation, and neither is particle
rebound. However, it is clear from Figure 2-4 that inertial impaction is the dominant
2-12

pathway for deposition for particles with diameters relevant to dust emission, i.e. 1 – 10
µm in diamter.
This model for dry deposition requires knowledge of u * , z 0 , and L.
2.1.3 The Model of Raupach (1999)
Unlike the multi-layer, multi-resistance approach characterized by Equation 2-3, Raupach
(1999) has proposed that a single layer with multiple conductances can, under some
conditions, adequately represent deposition according to
v = v + G + G
d
g
IM
B
Equation 2-27
where v d and v g are defined as before and G IM and G B are the bulk conductances for
impaction and Brownian motion, respectively. Raupach (1999) uses the portion of the
bulk aerodynamic conductance G aM (G aM = u * 2 /u r ) associated with viscous drag to scale
G B and the portion associated with form drag to scale G IM . Thus,
G = a
IM
f
E
IM
( G f )
aM
form
Equation 2-28
and
G
B
v
B
[ G ( − f )]
= a E 1
aM
form
Equation 2-29
where E IM and E B are the particle impaction and the Brownian diffusion efficiencies,
respectively, f form is the fraction of the total drag that is form drag and is assumed to be
approximately 75% (the remainder is viscous drag) and a v and a f are numerical
parameters of order unity that account for inter-element sheltering. Raupach (1999)
suggests using the values 8 and 2, respectively based on a best-fit to Chamberlain’s
(1967) wind tunnel data. As in the earlier examples, the efficiency of removal by
Brownian motion is given by the Schmidt number raised to the –2/3 power:
−2 3
E B
= Sc
E IM is given by a form similar to that in Slinn (1982) (see Equation 2-17)
Equation 2-30
E
IM
Std
=
Std
+
p
q
Equation 2-31
2-13

where p and q are numerical factors estimated to be 0.8 and 2, respectively from earlier
work (Peters and Eiden, 1992; Bache, 1981). St d is a form of the Stokes number
St
d
2
=
( v g)
g
d
c
U
c
Equation 2-32
with U c equal to the flow velocity around the vegetation elements and d c equal to a
characteristic width of canopy elements upon which particles impact. Raupach (1999)
suggests that U c is “characterized” by the friction velocity u * . Note that if U c is assumed
to be equal to u * and d c is set equal to 2A L , then Equation 2-32 and Equation 2-18 are
equivalent.
With the approximations stated, this model only requires the specification of three
parameters, namely u * , u r , and d c .
2.2 Dispersion in the atmosphere
Whereas deposition serves to permanently remove particles from the air stream,
turbulent dispersion causes particles to dynamically redistribute through the mixed layer
in the atmosphere. Dispersion causes pockets of air with comparatively high
concentrations of particles to be diluted with pockets of air that have lower
concentrations.
There are a number of mathematical tools to describe dispersion in the
atmosphere. Most commonly, in analogy to mixing length theory for momentum
transport, it is quantified as a dispersion coefficient multiplied by a concentration
gradient. Thus, the amount of particles crossing a horizontal plane is given by an equation
of the form
F
z
= −K
zz
∂C
∂z
Equation 2-33
where F z is the flux of particles at a height z and is equal to the vertical (z-direction)
dispersion coefficient K zz (m 2 /s) multiplied by the negative of the vertical concentration
gradient (g/m 4 ). Similar equations are used to describe dispersion in the two horizontal
directions utilizing the coefficients K xx and K yy which may be close in magnitude to one
another but can be quite different from K zz . Horizontal dispersion may be ignored if the
source is approximated by an infinitely long line (e.g. a long road).
In the surface layer (nominally 5 – 50 m above ground), where the flux of
momentum to the ground is constant with height, K zz is well-specified. At greater
heights, measurements are more difficult, and the functional form of K zz is inferred from
wind tunnel tests, theory, or numerical simulations.
2-14

An alternative method for modeling diffusion is to specify the spread of a
pollutant as a function of downwind distance and atmospheric stability. This is the
approach used by a number of Gaussian Plume Models, including the ISC3 (USEPA,
1995). The concentration of an airborne species is assumed to be normally distributed in
each of the three ordinal directions, with standard deviations of σ x , σ y , and σ z For a
continuous line source, dispersion in the horizontal plane can be ignored. The parameter
σ z has been specified by a number of authors including Gifford (1961), Klug (1963),
Turner (1969, 1970), and Martin (1976). The ISC3 model uses the formulation of Turner
(1970). The Gaussian approach is useful because of its simplicity in implementation.
However, it has its limitations, most notably, that the assumption of a normally
distributed concentration profile may be too simple to adequately account for the range of
land use and atmospheric conditions that can be encountered.
Alternative methods for quantifying turbulent dispersion are available from the
literature (e.g. Venkatram, 1993; Du and Venkatram, 1997; Schopflocher and Sullivan,
2002), some of which may be more accurate than mixing-length or Gaussian models.
2-15

2-16

3. MEASUREMENT OF DEPOSTION VELOCITY
The goal of this WESTAR-funded, University of Utah subproject was to estimate
the magnitude of particle deposition from vehicle-generated dust clouds downwind of an
unpaved road. In this chapter, we discuss the methods and results of direct measurement
of deposition velocity v d . Measurements were conducted in the “impact zone” (refer
Figure 2-1). The measured values of v d are compared to values previously reported in the
literature and analyzed in the context of the resistance model for deposition.
Two types of surfaces for particle deposition, flat substrates and artificial
vegetation, were analyzed with microscopic and gravimetric techniques. Comparison of
these measurements with prior work suggests that there are challenges to bridging direct
measurements with canopy-wide deposition velocities.
3.1 Methods
All the deposition velocity testing was conducted during April 2002 at Ft. Bliss
TX, on a spur road south of Loyalty Lane at GPS coordinates 13 3 70 544 E, 35 28 119
N. The field experiments were conducted in cooperation with the SERDP-funded
projects CP1190 and CP1191. The site is arid with sparse vegetation cover and exposed
sand. The prevailing winds were from the west and nearly perpendicular to the road.
Road surface samples were collected and size distribution was measured by standard
sieve analysis (refer Table 3-1).
Table 3-1 Cumulative screen analysis for the soils samples collected before and after the sampling
period
Facility & GPS Units/
Sample Site Location
Well Test
Road
13 3 70 544 E
35 28 119 N
DRI/JMV ID Notes Silt %
+ 3/8 + 4 + 20 + 30 + 50 + 100 + 200 - 200
mesh mesh mesh mesh mesh mesh mesh mesh
Heavy sand deposits
K408-1A from previous day’s 7.0 1.00 1.00 1.00 0.97 0.85 0.64 0.36 0.07
windstorm
Well Test
Road
13 3 70 544 E
35 28 119 N
K419-TR-1
After extensive driving
compared to previous 4.1 1.00 0.99 0.98 0.87 0.81 0.59 0.28 0.04
samples
Dust Concentration was measured using a GRIMM (Dust Monitor Series 1.108®,
GRIMM Technologies Inc., Douglasville, GA), which recorded particle-count
concentration for 16 size ranges logging every six seconds. This is an optical instrument
that determines particle size and count by light scattering. The sample inlet was placed
next to the surfaces used for particle deposition sampling. A 2.5 m long PVC tube
connected the inlet to the GRIMM instrument, which was placed with the data-logging
computer in a sealed box. One test was conducted using a different GRIMM that had a
direct sample inlet and that was programmed to report calculated mass. Quality checks
for the GRIMM included an internal calibration cycle and cleaning of sample inlet head.
The data from the GRIMM were used to obtain a particle number size distribution timeaveraged
over the duration of each dust deposition experiment.
3.1.1 Deposition on Flat Substrates
A series of tests was conducted to measure the passive deposition of vehicle
generated dust particles on flat surfaces placed near the road. The technique involved
3-1

counting by scanning electron microscopy (SEM) the number of particles deposited on
polycarbonate membranes which were exposed using a three-axis support stand (Figure
3-1). Six samples corresponding to the stream-wise, crosswind and vertical directions
(+x, -x, +y, -y, +z, and –z) could be exposed simultaneously. The +x direction was
downwind perpendicular to the road and the +z direction was vertical upward. The
substrates were exposed to dust from vehicles driving on the test road at speeds from 5 to
50 mph. Alternative locations, exposure durations, and substrate materials were tested as
discussed in method development, below. The best results were obtained on
polycarbonate substrates (Millipore type GTTP, Fisher Scientific) located 5 m downwind
of the road and exposed for 15 vehicle trips.
6-Direction Passive Deposition
Polycaronate
Polycarbonate
membrane on on
Carbon Tape
Petri Dish Detail
Typical for 6
Support Stand
with Clamps
Grimm
Figure 3-1 The experimental setup of the passive deposition on flat surfaces for examination by
electron microscopy. Three-axis (corresponding to the stream-wise, crosswind, and vertical
components of wind direction i.e. +x, +y, +z, -x, -y and –z directions) passive collection substrates
were mounted on a ring stand. Standard right hand coordinates are used with the positive x-axis
downwind perpendicular to the road. A GRIMM inlet was placed right next to it to monitor the dust
concentration. The substrates were placed 1 or 2m high and 5 or 10m away from the unpaved road.
The size and number of particles deposited on the flat substrates were determined
by quantitative electron microscopy. The flat substrates were mounted on double-sided
carbon tape in a disposable 50 mm diameter Petri dish. Different substrate types were
tested side by side in a single Petri dish. After exposure in the field the Petri dish covers
were replaced and the set of samples was boxed for shipment to the laboratory. To
prepare for microscopy, a section of the substrate assembly was cut from the dish and the
bottom layer of conductive double-sided carbon tape was used to mount the sample on an
aluminum SEM stub (Hitachi S-450®, Redding, CA) for gold coating. All samples were
evaluated qualitatively for dust loading using the Hitachi S3000N® scanning electron
microscope (Hitachi, Pleasanton, CA) located in the Material Science and Engineering
3-2

Department at the University of Utah. The selected filters that passed QA underwent full
quantitative microscopy to provide the images for particle counting.
SEM images for counting were taken in a non-overlapping structured grid pattern
at three magnifications (100X, 500X, 2500X). By combining data from multiple
magnifications both the infrequent, large, high-mass particles and the ubiquitous, less
massive sub-micron particles could be quantified. Particles d>20 µm were measured on
100X images, particles 5

Note that all the size bins measured by electron microscopy and by the GRIMM
are physical diameters and not the aerodynamic diameters. In addition, Equation 3-2
requires that the concentration time average be calculated over the same interval as the
exposure time.
3.1.2 Deposition on Artificial Vegetation
Dust deposition on artificial vegetation (AV) surfaces placed near the edge of the
road was measured by collecting the deposited material, separating the particles by size,
and weighing the material. Plastic "fir" Christmas garland was used since real vegetation
is subject to artifacts from natural variation, biogenic debris, and moisture changes. The
pre-washed AV was wrapped around a 17”x19” PVC frame and mounted on a support
frame above a clean plastic tub (20"x 14"= 0.18 m 2 ) that was placed downwind of the
road, Figure 3-2. The "fir" consisted of a central wire with close-spaced needle-covered
branches. The multiple wraps of the garland around the frame created a dense but porous
obstruction to the dust flow. Total surface area of the branches and needles was 3.7 m 2
compared to the horizontal projected area of the assembly outline, which was 0.18 m 2 ,
and the frontal projected area above the tub, which was 0.26 m 2 . For manufactured AV,
total surface area could be determined accurately by measuring dimension components
(needles, branches) and multiplying by number of components per length.
Artificial
Vegetation
Grimm
Collection
Tub
Figure 3-2. The experimental set up for passive deposition on artificial vegetation (Fir Christmas
garland). The plastic foliage was wrapped around a PVC frame, pre-washed and installed above a
clean plastic tub placed 5 m away from the road and 1 m above grade. A GRIMM inlet was placed
alongside it to monitor the dust concentration.
A control tub of similar dimensions was placed next to the AV to measure particle
settling on horizontal surface in the absence of vegetation. During sampling runs, the AV
was exposed to the vehicle-generated dust. Exposure ranged from 1 hour at 5 m from the
road to 8 hours at 50 m from the road. After sampling, the assembly was covered and
taken indoors, where the deposited material was collected in two fractions: the dry
material dislodged by shaking the vegetation and the material removed by spray washing
with distilled water.
3-4

In the laboratory, the deposited material was weighed then separated into four size
bins (d 30 µm, 10 d < 30 µm, 3 d < 10 µm and d < 3 µm) by gravity settling in
water, based on the Stokes velocity difference between various size particles (see
Equation 2-4). The dry and wet fractions collected in the field were analyzed separately
and the results were added to get the total mass deposited. The size-separated fractions
were dried and weighed and the amount of deposition on the AV in each size range was
calculated by mass balance. The gravity separation method was validated in preliminary
experiments using surrogate mixtures of particles with known size and the separations of
the field samples were run in duplicate after splitting the sample using a riffler. The
quality of the size separations obtained from the field samples was also verified using
SEM.
For the artificial vegetation (AV) experiments, the deposition velocity was
determined using Equation 3-2 expressed on a mass basis. The GRIMM particle number
data were converted to particle mass. The amount deposited was determined from
weighing the 3 d < 10 µm and d < 3 µm wet separated fractions. The deposition
velocity was calculated based on the horizontal projected area of the collector outline.
3.1.3 Flat Substrate Method Testing
This study of field measurement of vehicle-generated dust deposition required
testing to evaluate alternative procedures, materials, sampling locations, sampling times,
and data analysis methods. Twenty experiments measuring particle deposition on flat
substrates were collected using various locations and numbers of vehicle trips. Samples
were collected 1 or 2 m above grade at 5 and 10 m from the unpaved road. In addition, a
blank sample was collected 100 m upwind from the road.
Four substrates were tested to determine which surface gave the best data:
1. Dry Polycarbonate filters (Isopore Membrane filter, 47 mm diameter,
0.2 µm GTTP, Millipore)
2. Oiled Polycarbonate filters, created by spraying DOW 316 (Slipicone®
316 release agent, Dow Corning Corp., Midland, MI) on the filters and drying them
3. Carbon tape (Catalog # 16073®, 8 mm, Ted Pella Inc, Redding, CA)
4. Double sided clear tape (Scotch ®, 3M, St. Paul, MN)
3.1.4 Vegetative Method Testing
Six experiments were conducted to evaluate deposition velocities by the artificial
vegetation method. A combination of different distances from the road (5 and 50 meters)
and different artificial vegetation substrates (Fir Garland and artificial Ivy) were used.
Deposition velocity was calculated using different areas in Equation 3-2:
horizontal, total and projected. The horizontal projection area was used because it
compares to what Slinn used in his calculations. It measures the enhancement of removal
provided by vegetation. It is also easy to measure and is applicable to flux studies. The
frontal projection area of the vegetation was used since it relates to the SEM flat substrate
experiments. It represents the area of direct contact with the bulk of the dust cloud. The
total area represents all surfaces available for deposition. Since not all exposed surfaces
3-5

will achieve equal deposition rates due to shadowing and other issues, the overall
deposition velocity calculated using this area would be low.
Logs of all the flat substrate samples and the artificial vegetation experiments,
including samples not completely analyzed due to method or QA issues, are given in
Table 3-2 and Table 3-3 respectively. This includes method development tests, samples
where data were lost, samples that did not pass QA, and spare field samples that were not
analyzed. The incremental cost of collecting spare field samples during the campaign
was low compared to the cost of microscopy and image analysis.
3.2 Results: Deposition velocities measured by SEM and gravimetric techniques
Microscopy counting of particles deposited on flat substrates gave a deposition
velocity based on projected area ranging from 0.3 cm/s for submicron particles to 2 cm/s
for 3 µm particles. The artificial vegetation suspended above a tub gave a deposition
velocity based on horizontal surface area of about 10 cm/s. Deposition velocities by
electron microscope particle counting of flat substrates and by gravimetric analysis of the
deposits on artificial vegetation are presented in Figure 3-3. These values are also
compared to predictions for dry deposition into a forest canopy in a theoretical study by
Slinn (1982). The model prediction curve plotted in Figure 3-3 corresponds to a wind
velocity of 5 m/s and a value of gamma (a measure of wind profile in the canopy) of 3.5,
which represent the middle of the range considered by Slinn. A major difference is that
Slinn's model calculates vertical flux from the atmosphere to a horizontal area of forest
canopy while the data from this study are deposition from a horizontally moving cloud
impinging on the projected area of a surface. Figure 3-4 gives the directional variation in
the deposition velocities on flat surfaces (polycarbonate filter) normalized by the
deposition velocity measured on the horizontal surface facing up (+z). The flux on the +z
surface corresponds to the deposition velocity measured by Slinn. In addition,
gravitational settling would be expected to deposit particles on this surface only. The
vertical plane surfaces (-x, +y, -y) show enhanced deposition due to impaction caused by
wind and vehicle wake whereas the deposition on surfaces facing down and away from
the road (+x, -z) is presumed to be due to turbulence. Figure 3-5 shows the directional
deposition velocities over the size range studied.
Complete analysis was done on four samples (dry polycarbonate filters). They
were SEM ID 12, 17, 18 and 19. SEM 17, 18 and 19 were 5 m away from the road and
SEM 12 was 10 m away from the road. A grand average deposition velocity (17, 18, 19)
curve versus each size bin plot is given in Figure 3-6. The particle concentration from
the GRIMM measured coincident with the four SEM samples has been plotted in Figure
3-7. There appear to be systematic differences between the size distribution measured
with a direct inlet (SEM 12) and the size distributions measured with an inlet tube.
Particle losses increase the calculated deposition velocity, but data for d < 5 µm appear to
be valid.
Testing of alternative methods increased the chance of success for this experiment
and provided data on materials and sampling strategies that can be used to design future
3-6

Table 3-2. Log of all flat substrate samples
Date SEM ID #
Location
(x, y)
# Of
Trips
Vehicle type
03/26/02 1N 10, 2 10 Chevy Impala
Comments
03/26/02 1F 40, 2 20 Chevy Impala Very lightly loaded
03/27/02 2N 10, 2 10 Chevy Impala
Only PM10 mass, no
concentration by size
03/28/02 3N 10, 2 10 Chevy Impala Very lightly loaded
03/28/02 3F 40, 2 20 Chevy Impala Very lightly loaded
04/08/02 1– (+x, +y, +z)
-100, 1
-
Patriot
Battery
04/08/02 2 5, 1 40 Voyager
Missile
04/11/02 3 5, 1 17 Dodge Caravan
04/12/02 4 5, 1 20 Voyager
04/12/02 5 5, 2 20 Voyager
Decided not to use oiled filters
for counting
Decided not to use oiled filters
for counting
-X filter lost, oiled base
formed cake-like structure
Decided not to use oiled filters
for counting
Decided not to use oiled filters
for counting
04/12/02 8 5, 1 20 Voyager Very lightly loaded
04/13/02 1- (-x, -y, -z) -100, 1 -
1 hr upwind
control
04/16/02 13 - - Voyager Not analyzed
Decided not to use oiled filters
for counting
13-1 - 0 - Filter pores clearly visible
13-2 - 0 -
Oil forms a cracked cake-like
base
13-3 (-x) 5, 1 1 - Very lightly loaded
13-4 (-x) 5, 1 5 - Loading greater than 13-3
13-5 (-x) 5, 1 5 -
Cake base formed, Particles
barely distinguishable
13-6 (-x) 5, 1 25 - Loading greater than 13-4
04/17/02 7 5, 1 50 2 ½ ton truck
04/17/02 14 5, 1 50 2 ½ ton truck
Decided not to use oiled filters
for counting
Carbon tape seems to be a
good substrate for future SEM
experiments
04/17/02 12 10, 2.3 50 2 ½ ton truck Deposition velocity calculated
04/18/02 15 -100, 1
1 hr
control
Deposition velocity calculated
04/18/02 16 5, 1 5 Hemmet Spare – not analyzed
04/18/02 17 5, 1 25 Hemmet Deposition velocity calculated.
04/19/02 18 5, 2 15
04/19/02 19– (+x, +y, +z) 5, 1 15
04/19/02 20 50, 1.5 15
18 wheel Flatbed
military truck
18 wheel Flatbed
military truck
18 wheel Flatbed
military truck
Deposition velocity calculated.
Deposition velocity calculated.
Spare – not analyzed
studies. Four different substrates were tested for flat surface sampling. The reported
deposition velocities are based on SEM examination of dry polycarbonate filter
membranes. The oiled filters developed a cracked surface film and it was difficult to
distinguish the particles. The sample collected on carbon tape indicated that this might
be a suitable substrate. The clear tape has high collection efficiency for large size
particles, but it is difficult to distinguish the smaller size particles from artifacts on the
adhesive surface. A comparison of particle counts done on a clear tape and a dry filter
sample is given in Figure 3-8.
3-7

Table 3-3. Log of all field vegetative samples
Sample ID Date
Location Total Run Time Comments
(x, y) - m (min)
F = Fir C = control
F001 04/08/02 5, 1 20 Problem in lab processing
5, 1 52 Unable to open GRIMM data
C001 04/08/02 5, 1 52 Unable to open GRIMM data
F002 04/12/02 5, 1 460 Deposition velocity result
C002 04/12/02 5, 1 460 Deposition velocity result
F003 04/17/02 5, 0.75 194 Gravimetric data suspect
C003 04/17/02 5, 0.75 194 Gravimetric data suspect
F004 04/17/02 50, 0.75 185
04/18/02 50, 0.75 330
I001 04/18/02 5, 0.75 220
Deposition velocity result
No GRIMM data for d

16
Normalized Deposition Velocity
11
6
1
-4
-9
NX NY PY NZ PX
Direction
Figure 3-4. Directional variation in deposition velocity normal to a flat surface. Data given are mean
and one standard deviation, normalized by the deposition velocity on the horizontal surface facing up
(PZ) averaged for the range 0.5-7.5 m (physical diameter) based on four determinations (SEM 12,
17, 18, 19). The vertical surfaces (NX, NY, PY) show enhanced deposition due to impaction.
Deposition on surface facing down (NZ) and away from the road (PX) is presumed to be due to
turbulence.
100
Normalized Deposition Velocity
10
1
0.1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Physical Diameter - microns
NX NY NZ PX PY PZ
Figure 3-5. Directional variation in deposition velocity normal to a flat surface as a function of
particle size. Data are the mean of four determinations (SEM 12, 17, 8, 19), normalized by the
deposition velocity on the horizontal surface facing up (PZ). The plot indicates no consistent size
dependence for particles less than 10 µm in diameter.
3-9

100
Deposition velocity - cm/s
10
1
0.1
0.3
0.4
0.5
0.65
0.8
1
1.6
2
3
4
5
7.5
10
15
20
Physical Diameter - microns
Grand Avg of all the filters excluding SEM 12
Stdev
Figure 3-6. . The grand average deposition velocity (SEM 17, 18, 19) for the SEM samples versus the
particle size. The dashed line represents one standard deviation of the three determinations.
Deposition velocity and standard deviation of replicates data outside the 0.5 m to 5 m range are
not shown due to uncertainty in the accuracy of the data. For d5 m, the results are uncertain due to inlet line
loss.
1.E+10
Concentration - N/m3
1.E+08
1.E+06
1.E+04
1.E+02
1.E+00
0.3
0.4
0.5
0.65
0.8
1
1.6
2
3
4
Physical Diameter - microns
5
7.5
10
15
20
DRI - 12 Utah - 17 Utah - 18, 19
Figure 3-7. Variation in particle concentration between sampling days and between instruments with
different inlet configurations. SEM 12 data were obtained from a GRIMM with a direct sample inlet
and SEM 17, 18, 19 data were obtained from a GRIMM with a 2.5 m long PVC tubing attached to its
sample inlet. For diameters less than 5 m (physical), both the inlets agree but at particle diameters
greater than 5 m, the GRIMM with the long inlet tube exhibited sampling losses. Particle losses
affect deposition velocity calculations.
3-10

1.E+05
Deposition Velocity - cm/s
1.E+03
1.E+01
1.E-01
1.E-03
0.1 1 10 100
Physical Diameter - microns
Clear Tape
Dry Filter
Figure 3-8. Deposition velocities calculated from SEM particle count for a dry polycarbonate
membrane and a clear tape surface that were collected simultaneously (SEM 18). From 5 m to 10
m both the substrates give similar particle counts. The count is increased (and therefore deposition
velocity is increased) for particles d20 m) that fell off the dry substrate (data not shown).
Deposition velocity for the vegetative samples was calculated using three
different area counts, given in Table 3-4. Using the horizontal projection area resulted in
a higher deposition velocity for the AV and control combination when compared to the
control alone. The first calculation given in Table 3-4 used the entire area of the AV plus
the area of the tub underneath it. The calculated deposition velocity for d< 3 µm using
this area was 0.7 cm/s compared to 8.3 cm/s for the control. The second deposition
velocity calculation used the frontal projection area of the AV plus the area of the tub
underneath. The resulting deposition velocity for d< 3 µm using this area was 6.5 cm/s
vs. 8.3 cm/s for the control. The final calculation used the horizontal AV projected area
plus the area of the tub (essentially the area of the tub). The resulting deposition velocity
for d

Table 3-4. Deposition velocity calculations for vegetative samples using different areas:
1. Deposition velocity calculation using total area:
Sample ID
Size Range
(physical)-m
Particle
Concentration-g/m 3 Gravimetric Wt. - g Area – m2 Time - sec Velocity–cm/s Comments
F001 d

are deposition from a horizontally moving cloud impinging on the projected area of a
surface. Slinn reasoned that because momentum transfer from the wind to ground is
higher over a forest, then downward particle transfer should also be higher, and
developed a model that used the wind velocity profile in the forest canopy as a parameter.
The experiments in this study are more closely analogous to the deposition predicted by
models of single fiber filtration efficiency.
Table 3-5. Deposition Velocity calculated for vegetative samples
Sample ID
Location
(x, y) - m
Size Range Particle
(physical)- µ Concentration- g/m 3 Gravimetric Wt. - g Area – m2 Time - sec
F002 5, 1 d

polycarbonate surfaces and those calculated for similar conditions according to Equation
3-3 and Equation 2-25. The difference between measured and predicted deposition
velocities is quite large, ranging from one order of magnitude for large particles to two
orders of magnitude for small particles. This difference is expected to be manifested in
the overall deposition velocity (I.e. considering both r b and r a ) since the resistance to
transport through the surface layer r a is very small compared to r b . That is, under the
atmospheric conditions of Ft. Bliss, the deposition velocity is controlled almost entirely
by the transport through the quasi-laminar sublayer.
100
Deposition Velocity (cm/s)
10
1
0.1
Vd=1/rb calculated
Ft Bliss
Expon. (Ft Bliss)
0.01
0 1 2 3 4 5 6 7 8 9 10 11
Aerodynamic Dp (microns)
Figure 3-9. Measured deposition velocities on polycarbonate substrates and deposition velocities
calculated based on the boundary layer resistance r b only vs. particle aerodynamic diameter. The
average u * for the sample periods represented by the measured data and used for the theoretical
prediction was 0.35 m/s.
There is a subtle point that explains some of the differences between the measured
and modeled deposition velocities. The measured values in this study represent
deposition to a surrogate surface (artificial plant or flat substrate). It is likely that the
density of the vegetation in the canopy, characterized perhaps by the fraction of ground
area covered by plants should enter in the calculation of overall deposition to the canopy.
That is, the surrogate surfaces used in this study may adequately simulate the removal of
particles by individual vegetative elements, but do not account for the density of those
elements. Some work in this area has been done by Raupach and Leys (1999), though
testing of their hypotheses against data is not complete.
3.4 Conclusions
An important point, directly relevant to this WESTAR project as a whole, is that
the quantification of the deposition velocity is not a straightforward matter. Most models,
including the ones discussed in (Chapter 2) are only tested against limited data sets, most
notably those of Chamberlain (1967). One reason for this is that direct measurement of
deposition velocities is quite difficult, and each of the available techniques has some
significant shortcomings as noted by Dabberdt et al. (1993). Though use of surrogate
surfaces for deposition is a common approach (e.g. Wu and Davidson ,1992; Etyemezian
et al., 1999), it also suffers from shortcomings. Most notably, the surrogate surface is a
3-14

single element that is intended to induce similar transfer of material (momentum,
particles, gases, etc) as an entire canopy. While there have been recent novel approaches
that attempt to measure deposition in-situ (e.g. Gillies et al., 2001), at present there
remain substantial uncertainties in the true value of deposition velocities. The solution to
the difficulty of quantifying deposition is outside the scope of the present study. We note
simply that values of v d calculated by any of the methods described in Chapter 2 may be
erroneous in some cases by a factor of two or more, based on the input parameters alone.
The results of this portion of the study, which utilized surrogate surfaces for
deposition, indicate that measured deposition velocities are difficult to relate to
theoretical formulation. This is illustrative of the ongoing challenge in reconciling data
with theory in the field of dry deposition.
3-15

3-16

4. MODELS USED TO STUDY PARTICLE REMOVAL BY DRY
DEPOSITION DOWNWIND OF AN UNPAVED ROAD
One of the objectives of this study was to provide an analysis of the Gillette Box
Model in terms of its utility in determining the removal of dust particles downwind of an
unpaved road. This was accomplished by considering two other models, one based on
the Atmospheric Diffusion Equation (ADE), and the other based on the Gaussian plume
approximation. The former provides the underlying basis for many regional air quality
models, while the latter is used by the EPA’s ISC3 dispersion utility. In this chapter, we
present the ADE, Gaussian, and the Gillette Box Models. Simulations using all three
models are compared to one another. Special consideration is given to the assumptions
of the Box Model, and the conditions for their validity.
4.1 Presentation of Models
4.1.1 The one-dimensional Atmospheric Diffusion Equation (ADE)
The Atmospheric Diffusion Equation (ADE) describes the transport and removal
of airborne material. It is derived from a mass balance on a control volume (CV), where
species (gases or small particles) are allowed to diffuse in and out of the CV by the
turbulent motions of the atmosphere, move through the CV by advection, and react
chemically. Invoking the K-theory for turbulent dispersion, the full Eulerian equation is:
∂c
∂c
∂ ∂c
∂ ∂c
∂ ∂c
+ u = K
xx + K
yy
+ K
zz + S
∂t
∂x
∂x
∂x
∂y
∂y
∂z
∂z
Equation 4-1
where c is the concentration of the species of interest, u is the velocity in the x-direction,
K ii are the turbulent diffusivities in the x, y, and z directions, and S is a chemical reaction
term. For an approximate continuous line source oriented in the y-direction, the net
transverse dispersion is zero, and the concentration at any one location is invariant with
time. In the x-direction, for any appreciable value of u (u > 1 m/s), advection is much
greater than dispersion. Dust particles do not react chemically, and the source/sink term
S is also set to zero. In summary,
K
∂c
∂c
∂c
∂ ∂c
= 0 , = 0, u >> K
xx ,
S = 0
∂y
∂t
∂x
∂x
∂x
yy
.
Equation 4-2
reducing Equation 4-1 to
∂c
∂ ∂c
u = K
zz .
∂x
∂z
∂z
Equation 4-3
4-1

At this point, it is possible to remove the dependence of c on x if it is assumed that
the wind speed u is nearly constant with height and by expressing u as dx/dt. This casts
the ADE in a Lagrangian frame where the concentration c is dependent only on the height
z and the time since emission, yielding the one-dimensional, time dependent ADE:
∂c
∂
= K
∂t
∂z
∂c
∂z
zz
.
Equation 4-4
Note that the time derivative in Equation 4-4 is not the same as that in Equation 4-1.
Specifically, in the latter equation δc/δt is evaluated at a stationary point in space,
whereas in the former, it is implied that the derivative is evaluated in a frame of reference
that moves downwind with the plume.
The downwind distance at a given time can be estimated by multiplying the time
with the wind speed at a 10 m height, obtained, for example, from a logarithmic profile
(e.g. Equation 2-9). Values of K zz as a function of height have been provided by a
number of investigators. The functional forms used in this model are
Unstable Conditions (Lamb and Duran, 1977):
K
zz
w z
*
i
4
1
3
4
z z z
2.5
κ
1
− 15 → 0 ≤ < 0.05
zi
L zi
2
z z
0.021+
0.408
+ 1.351
↵
zi
zi
3
4
z z z
= −
4.096
+ 2.560
→ 0.05 ≤ < 0.6
zi
zi
zi
z z
0.2exp6
−10
→ ≤ <
0.6 1.1
z
i
zi
z
0.0013 → ≥ 1.1
z
i
Equation 4-5
Neutral Conditions (Shir, 1973):
8 fz
K zz = κ u
−
*
z exp
u*
Equation 4-6
Stable conditions (Businger and Arya):
4-2

κu
( )
*
z
8 fz
K zz
=
exp
−
0.74
+ 4.7 z L u*
Equation 4-7
where z i is the mixing height, L is the Monin-Obukhov length scale for turbulence, u * is
the friction velocity, f is the Coriolis parameter, κ is the Von Karman constant (0.41), and
w * is the convective velocity scale given approximately by
w
1
3
zi
*
= u*
− L
(L is negative for unstable conditions).
Equation 4-8
To solve Equation 4-4, it is necessary to specify boundary conditions on z and
initial conditions for t=0. The top of the modeling domain is set high enough (z top > 2×z i )
so that there is no net flux of particles. That is,
∂c
K
zz
= 0 , z = z top
.
∂z
Equation 4-9
At the ground, the downward flux of particles by dispersion must equal the removal by
deposition
∂c
K
zz
= vd
c, z = 0
∂z
Equation 4-10
where v d is calculated from Equations 2-3, 2-25, and 2-26 with r a set to zero, since the
depositing particles are already at the ground (z=0).
The initial condition is simply that the concentration of particles is uniform up to
a specified injection height. That is
c z)
= c , z < IH
(
0
c(
z)
= 0, z ≥ IH
Equation 4-11
In practice, the transition across z=IH is made mathematically smooth to avoid
difficulties in the numerical solution. With these initial and boundary conditions, the
ADE is solved numerically at 1,000 time steps spaced logarithmically from 1 second to
10,000 seconds.
4-3

4.1.2 The Integrated Source Complex Model 3 (ISC3)
The ISC3 Short-Term dispersion model (ISC3 hereafter) is built on the
approximation that a plume dispersing in the atmosphere assumes a shape similar to a
Gaussian distribution (USEPA, 1995). The ISC3 utility allows for simulation of point,
area, and volume sources with the aid of hourly meteorological data. Point and volume
sources are treated in essentially the same way; volume sources are assumed to be point
sources that originated at some distance upwind. Area sources are treated as multiple
point sources. Line sources can be approximated by using multiple area or volume
sources. The ISC3 also includes a number of optional parameters that can be used for
adjusting the height of a plume (e.g. in the case of hot or high-speed stack gases),
accounting for building downwash, and dispersion in complex terrain. For simulating a
road dust plume generated by the movement of a vehicle on an unpaved road, the volume
source approach is most directly applicable; accordingly, we will restrict the presentation
of the ISC3 model to volume sources and omit material that is not directly applicable to
unpaved road dust emissions.
The gaussian representation of the concentration profile in the ISC3 model is
actually a result of an approximate analytical solution to the Atmospheric Diffusion
Equation under certain atmospheric conditions. It works best for sources that are high
enough above ground that the dispersion parameter (the equivalent of K zz in Equation
4-3) can be assumed roughly constant. The applicability and limitations of Gaussian
approaches to dispersion modeling have been considered by other investigators (e.g.
Seinfeld and Pandis, 1999) and an in depth discussion is omitted here.
The basic equation that is solved for the ground-level concentration C (µg/m 3 )
downwind of a point source is
C =
−6
( 1×
10 )
QVD
y
exp − 0.5
2πu
sσ
yσ
z
σ
y
2
Equation 4-12
where
Q = pollutant emission rate (g/s)
V = a vertical term that includes the effects of ground reflection, dry deposition, and
vertical mixing
D = chemical decay term (assumed equal to 1 since road dust is non-reactive)
σ y , σ z = standard deviation of vertical and lateral concentration distribution
u s = mean wind speed at release height.
For a continuous line source of non-reactive material (i.e. D=1); Equation 4-12 may be
integrated over -∞ < y < +∞ to obtain
4-4

C =
−6
( 1×
10 )
QV
2π
u s
σ
z
Equation 4-13
where the dot over the Q indicates that the source strength is expressed in terms of a unit
crosswind distance (i.e. (g/s)/m). The most important parameter in the ISC3 model with
regard to a continuous line source is the vertical standard deviation σ z. In the ADE,
turbulent dispersion is calculated from the vertical gradient of the concentration whereas
in the ISC3, σ z. completely specifies the concentration distribution and therefore, also the
vertical gradient. This is accomplished through the vertical term V in Equation 4-13
which, in the absence of deposition is specified as
V
z
r
− h
= exp−
0.5
σ
z
2
z
r
+ h
+ exp−
0.5
σ
z
2
+
[.....]
Equation 4-14
where z r is the height at which the concentration is to be evaluated, and h is the initial
height of the release. To account for the reflection of the plume at ground level, the first
two terms on the RHS of the equation actually represent two sources, one located at h,
and one at –h. For road dust, it may be assumed that the release height of the plume is
the ground. The release height is different from the initial vertical depth of the plume
(discussed below). The unspecified bracketed term on the far RHS of Equation 4-14 is
used to account for multiple reflections between the ground and the mixing height. This
term can be ignored for a ground-level release provided that the analysis does not proceed
too far downwind (i.e. σ z.

When a vehicle passes over a road, the plume generated behind the vehicle has a
discrete depth in the vertical direction. This depth is a measure of the initial breadth of
the plume and is different from the plume release height. To account for the fact that the
dust plume is initially dispersed by the turbulent wake of the vehicle, a “virtual” distance
is added to the value of x in Equation 4-15. For example, the virtual distance x 0 is
calculated by solving the equation for the initial value σ z0 . Since σ z is the standard
deviation of the concentration in the vertical direction assuming a gaussian (Normal)
distribution, then 95% of the plume is initially below the height 2×σ z0 . Therefore, we
may approximately define σ z0 as one half the “injection height” or the height of the
influence of turbulence generated in the wake of a vehicle. The concept of the “injection
height” is discussed in detail in Chapter 5.
4.1.3 The Gillette Box Model This section of the text was provided by Dr. Dale
Gillette (NOAA) with some minor adjustments by DRI
4.1.3.1 Formulation and solution
A “lumped control volume” approach was used by Gillette to gain understanding of
fugitive dust sources. This case considers dust generated from a road surface. By letting
the road be directed into and out of the page while wind is directed from right to left,
two-dimensionality or symmetry in the direction into the page was invoked. That is,
fluxes are equal into and out of the direction into the page. Figure 4-1 shows the
geometry of the control volume. A dirt road exists at the right side of the control volume.
To the left of the road is a surface that is grass or shrub-covered and does not emit
particles. The ceiling of the control volume is the surface of primary interest as to
vertical flux.
dm/dt up
Wind
dm/dt ceil
dm/dt ambout
dm/dt ambin
dm/dt road
X
X=X o
dm/dt depos
dm/dt ceil
dm/dt ambin
dm/dt out
dm/dt road
∆Z
Z=0
∆L=1
Figure 4-1. Control Volume for Fugitive Dust Model Depicting Vertical and Horizontal Fluxes.
The quantities shown in the figure are as follows:
4-6

oad.
M is the mass of particles in the control volume (CV),
dm up /dt is the mass per unit time passing out vertically at the top of the CV,
dm depos /dt is the mass per unit time depositing to the floor of the CV,
dm ambout /dt is the mass per unit time passing out of the CV through the left wall,
dm ambin /dt is the mass per unit time passing into the CV from the right wall,
dm road /dt is the mass per unit time emitted from the road, and
dm ceil /dt is the mass per unit time passing out of the ceiling directly above the
A conservation of mass equation can be written for the control volume shown in
the figure as:
dM/dt + dm up /dt + dm depos /dt + dm ambout /dt - dm ambin /dt - dm road /dt + dm ceil /dt = 0
Equation 4-16
where, M is the mass of particles in the control volume (CV),
dm up /dt is the mass per unit time passing out vertically at the top of the CV,
dm depos /dt is the mass per unit time depositing to the floor of the CV,
dm ambout /dt is the mass per unit time passing out of the CV through the left wall,
dm ambin /dt is the mass per unit time passing into the CV from the right wall,
dm road /dt is the mass per unit time emitted from the road, and
dm ceil /dt is the mass per unit time passing out of the ceiling directly above the
road.
A simplifying assumption is that of steady state emissions, that is, dM/dt = 0. For this
case:
dm up /dt = - dm depos /dt - dm ambout /dt + dm ambin /dt + dm road /dt - dm ceil /dt
4.1.3.1a
road
Equation 4-17
Relationship of mass per unit time to the horizontal mass flux from the
A subvolume of the control volume is shown in Figure 4-1 as that part that
contains the road but no area downwind of the road. The left wall of this part of the
control volume is the surface through which all of the dust emitted from the road passes.
That is, we specify that none of the road dust is part of the ceiling flux (dm ceil /dt). For the
condition of steady state, a description of the conservation of mass for this part of the
control volume is:
dm road /dt = dm out /dt - dm ambin /dt + dm ceil /dt
Equation 4-18
where dm out /dt is the horizontal mass per unit time passing out through an imaginary wall
just left of the left edge of the road reaching from the surface to the top of the control
4-7

volume, and dm ceil /dt is the mass per unit time passing through the ceiling of the control
volume Because the road dust is usually the overwhelming source of dust, that is,
dm road /dt >> dm ambin /dt + dm ceil /dt , the horizontal flux dm out /dt for these conditions is
approximately:
dm road /dt = dm out /dt
which is the horizontal mass flux of dust from the road.
Equation 4-19
4.1.3.1b Deposition at the floor and vertical flux of dust through the ceiling of the
control volume downwind of the road
The total loss of material diffusing vertically through the ceiling and depositing
on the floor of the control volume to the left of the road may be calculated by first
calculating the effective concentration at position x to the left of x 0 (i.e., the left edge of
the road). We use the equation:
V
dC(
x)
dx
=
− C(
x)[
V
∆z
d
+ K]
Equation 4-20
where, V is the wind speed that carries the dust through the control volume (CV),
C(x) is the concentration of dust mass at position x in the CV,
x is the horizontal position in the CV, increasing to the left,
z is the vertical position in the CV, increasing from the floor to the ceiling,
V d is the deposition velocity, and
K is a coefficient having the dimensions of velocity.
A solution to Equation 4-20 is:
dmroad
= Vd
+ K
C(
x)
dt
exp−
x − x
V∆z∆L
V∆z
o
[ ] ( )
Equation 4-21
Multiplying C(x) by V d L (i.e., the deposition velocity times the unit length of the road)
and integrating with respect to x from x = x 0 to x = gives:
dm
depos
dt
V
=
d
dm
dt
( V + K )
d
road
Equation 4-22
4-8

By making the approximation that dm ambout /dt - dm ambin /dt 0 and ignoring dm ceil /dt we
may rewrite Equation 4-17 as:
dmup
dm dm
road depos
= −
dt dt dt
Equation 4-23
Watson (personal communication, 2000) stated that the condition dm ambout /dt - dm ambin /dt
0 was observed in the field for a distance x of about 200 meters in the absence of any
intervening dust sources.
4.1.3.1c Ratio of vertical flux of road dust into the atmosphere to the horizontal
flux of road dust
The ratio of the mass per unit time emitted through the ceiling of the CV
downwind of the road (dm up /dt) to the mass emitted from the road (dm road /dt) is called the
“transportable fraction” and represents the fraction of emitted road dust that can be
considered to be regionally transported. This fraction Φ is given in Equation 4-24.
dmup
dt
Φ = =
dm
1
−
road
dt
V
K
( V + K ) ( V + K )
d
d
=
d
Equation 4-24
The value of K was derived by using data of Porch and Gillette (1977) and
(Gillette, 1974). Porch and Gillette (1977) provided data on fast-response concentrations
of diffusing dust simultaneously taken with fluctuation vertical wind speeds at the same
location. Analysis of the high-speed data showed that the aerosol flux could be expressed
approximately as 0.04u * [C] where [C] is the mean mass concentration. Analysis of
gradients of dust concentration in wind-eroding fields (Gillette, 1974) showed that the
aerosol flux could be expressed as approximately 0.07u * [C]. The mean of the
coefficients of [C] for these analyses gave a value of 0.06 u * which is the suggested
value for K.
Using Equation 4-24 with K equal to 0.06 u * , yields the “transportable fraction”
of fugitive dust emitted from the road. Practical application of Equation 4-24 can use
values of V d chosen to represent particle size and environmental conditions, for example,
those of Slinn (1982). Values of u * are chosen by the user to represent the environmental
conditions of interest.
4.1.3.2 Discussion
The expression given in Equation 4-24 may explain part of why observed
concentrations are smaller than that predicted by regional scale models that do not correct
for the transportable fraction of emitted dust but rather use the entire amount of dust
emitted by roads. However, there are limitations in applying this model to natural
4-9

situations. Some of these limitations are listed here with a detailed discussion given in a
later section.
1. If the flux through the ceiling above the road dm ceil /dt is not negligible. This
implies that the thickness of the first layer of a regional air quality model is
important to the transport of dust.
2. If the flux upwind of the road dm ambin /dt is not negligible compared to the flux
from the road. If the flux upwind from a source is not negligible, a regional
model must accommodate carrying the dust from upwind to downwind.
3. If the approximation of a well mixed dust concentration from the ground to the
top of the control volume is not a good description of the natural situation.
4. If the assignment of the flux through the ceiling of the control volume as KC(x) is
different from the diffusion taking place in the natural situation.
5. If the assignment of the deposition as V d C(x) is different from the deposition
taking place in the natural situation.
6. If steady state conditions not observed.
7. If the downwind flux dm ambout /dt is not of the order of dm ambin /dt within a few
hundred meters to a kilometer, the result of the model would not be applicable to
most regional scale regional models. Since the grid-scale of regional models is of
the order of a kilometer, dm ambout /dt should be of the order of dm ambin /dt within a
few hundred meters. The downwind flux equilibration distance may vary
depending on local surface variables. This may have some implications for the
minimum spatial resolution of a regional dust transport model. Perhaps in
modeling, the distance could be varied as a function of the soil and vegetative
cover—which in turn implies the need of good geographic data bases of such
information.
8. The analysis above was done for the wind perpendicular to the road. For wind at
an angle to but not parallel the road, the analysis would be almost the same, but
with a wider “effective road width.” For an infinitely long road parallel to the
wind, however, the assumptions of the analysis would be violated and the solution
would not be useful
4.1.3.2a
Weaknesses of the Box Model
1. The model continues to pass dust mass through the CV ceiling even though realworld
situations exist where no vertical flux takes place or there is vertical flux
from above the ceiling into the CV.
2. The model assumes dust to be well mixed in the control volume even though there
are situations where dust continues to lie close to the ground (possibly caused by
atmospheric stratification.)
4.2 Inter-comparison of the ADE, ISC3, and the Gillette Box Model
Since the model Gillette has proposed is mechanistically different from the ADE
and the ISC3, we begin with a discussion of the assumptions in the Box Model.
Modeling results from the ADE and the ISC3 are then compared with those obtained
from the Box Model. Note that except were noted, all simulations for deposition apply to
particles with aerodynamic diameters of 8 µm. PM 10 is the mass of all particles with
4-10

diameters less than 10 microns. Discussing model results for all particle sizes requires
more space in the text than is warranted. Eight microns is close to the Mass Mean
Diameter (MMD) of dust particles in the PM 10 size range near the point of emission.
Thus, the behavior of 8 micron particles, at least in the area close to the source is a
reasonable surrogate for PM 10 as a whole.
4.2.1 Analysis of the Gillette Box Model
It is useful in analyzing the Gillette box model to consider the assumptions made
in deriving that model. Invoking the resistance analogy, we consider the deposition and
dispersion of particles contained in a box with very small vertical extent equal to ∆z and
at an arbitrary height that falls somewhere in the surface layer. This is shown
schematically in Figure 4-2. We can describe the rate of change of the concentration in
the box as
∂C
=
∂t
( F + F )
dep
∆z
Box−SL
Equation 4-25
where F Box-SL is the flux of particles out the top of the box and F dep is the deposition flux
towards the ground. Replacing the fluxes with the concentration multiplied by the
appropriate resistance and noting that the wind speed V is equal to dx/dt, we obtain
C ra
V
=
∂x
1
+ r + r r v
+ v
( C − C ) + − v ( C − C )
∆z
1
ra
∂
g 0 Box
1 b a1
b g
2
g
SL
Box
Equation 4-26
where the variables in the equation have been defined in Figure 4-2. If we assume that the
concentration at the ground C 0 is equal to zero (i.e. there is no particle rebound and the
surface resistance is zero) and that the concentration at the top of the surface layer C SL is
equal to zero then Equation 4-26 reduces to
∂C
− C
V =
∂x
Box
[ v + K ]
d
∆z
G
Equation 4-27
with v d equal to the deposition velocity given in Equation 2-3 and the dispersion
coefficient K G given by
K
G
1
=
r
a2
− v
g
Equation 4-28
4-11

where the subscript “G” is used to indicate that the dispersion coefficient is defined as in
the Gillette box model with units of m/s. Equation 4-27 is identical to Equation 4-20, and
forms the basis for the Gillette Box model. Note that the gravity resistance has been
included for the transport of material through the top of the box model (i.e. Equation
4-28).
Planetary Boundary Layer (PBL)
F = F(z)
C SL Surface Layer2
r a2
F Box-SL =V da2 [C SL -C Box ]=(1/r a1 ) [C SL2 -C SL1 ]
C Box ∆z
r Surface Layer1
a1 F dep = F QL = F Box-SL1 =V da2 [C Box -C QL ]=(1/r a2 ) [C Box -C QL ]
C QL
Quasi-laminar Layer
F r dep = F Box-SL1 = F QL =V db [C QL -C 0 ]=(1/r b ) [C QL -C 0 ]
b
C 0 = 0
Ground or Vegetative Surface
r c r c assumed = 0 for particles if particle rebound is
negligible
r g
Figure 4-2. Representation of the Gillette Box Model in the context of the resistance model to
transport. The dashed horizontal line effectively divides the resistance r a into two smaller resistances,
r a1 and r a2 . The equations for fluxes in the figure do not include the effect of gravity.
In arriving at Equation 4-27 we have made three assumptions that warrant closer
inspection. First, we have assumed that there is no resistance to deposition at the ground
or vegetative surface (r c =0, C 0 =0). This assumption is frequently invoked for particles
that are 10 µm in diameter or less, and we accept it as being reasonable (Seinfeld and
Pandis, 1999; Raupach et al., 1999). Second, we have assumed that the concentration
profile through the surface layer is sufficiently well-developed that F box-SL is constant
from the box to the top of the surface layer. The validity of this assumption depends on
the atmospheric stability. Under unstable conditions, most of the resistance to
aerodynamic transport resides near the surface (see Figure 4-3). When conditions are
stable, the resistance to transport does not drop off as dramatically with height.
Therefore, assuming that the constant flux layer is instantly well-developed is probably a
reasonable approximation for unstable conditions, less so for neutral conditions, and not
valid for stable conditions.
The magnitude of K G cannot be estimated for a stable atmosphere since r a does
not converge on a maximum value. Under unstable conditions, r a converges on a
maximum value, and we can in principle calculate K G . However, since r a is dependent on
height above ground, we must make an assumption about which height to use. Selection
4-12

of such a height is somewhat arbitrary. For example, in the formulation of the box model,
it is suggested that K G = Au * with the value of A equal to 0.06. Neglecting the gravity
term for the moment, we can use Equation 4-28 to infer the corresponding value for r a2
1
r a 2
= =
Au
*
1
0.06u
*
Equation 4-29
If we now use Equation 2-24 to infer the height z at which K G was calculated, we find
that we need to specify the roughness height z 0 and the Monin-Obukhov length L. That is,
our specification of K G (0.06u * ) is not only dependent on the height z, but is also
dependent on the stability parameters of the flow. Thus, in general, our formulation for
K G is very approximate and only reasonably applicable under neutral to unstable
conditions.
100
aerodynamic resistance r a from the
ground up to indicated height (s/m)
90
80
70
60
50
40
30
20
10
0
very unstable L=-2
very unstable L=-10
unstable L=-100
slightly unstable L=-1,000
slightly unstable L=-10,000
neutral L=-100,000
slightly stable L=1,000 stable L=100
very stable L=10 very stable L=2
0 2 4 6 8 10
height above ground (m)
Figure 4-3. The total aerodynamic resistance from the ground up to the height indicated on the x-
axis. R a calculated after Byun and Dennis (1995). L is the Monin-Obukhov length and indicates
stability. (100,000 > ⏐L⏐ > 10,000 near neutral, 10,000 > L > 10 stable, 10 > L > 0 very
stable, 0 > L > -100 very unstable, -100 > L > -10,000 unstable). U * = 0.2 m/s and z 0 =0.01 m in
all cases.
Third, we have assumed that the concentration at the top of the surface layer (C SL ) is so
small compared to the concentration in the box (C Box ) that it is effectively zero (i.e. C SL -
C Box ≈ -C Box in Equation 4-27). Again, this approximation is only reasonably valid under
unstable conditions when most of the resistance to transport within the surface layer is
confined to the lowest few meters. Even then, the assumption only holds over a limited
distance downwind. At some point, particles that were originally lofted to a high
elevation are driven back down towards the ground. This occurs because particles are
removed at the surface and the concentration gradient that originally (near the source)
drove the vertical dispersion reverses direction as the concentration of particles at the
surface decreases.
4-13

4.2.2 Comparison of results for particle removal between predictions from ADE,
ISC3, and the Gillette Box Model
The results of numerically solving the Atmospheric Diffusion Equation are
segregated by atmospheric stability and shown in Figure 4-4. The fraction of particles
remaining in suspension increases as the atmospheric conditions vary from very stable to
very unstable. For all atmospheric stability conditions, the fraction remaining in
suspension decreases with increasing friction velocity. This is a very important point.
Increasing the friction velocity not only enhances turbulent dispersion, but it also
increases the impaction of particles and therefore the deposition velocity. Table 4-1 gives
approximate values for the wind speed at 10 meters corresponding to the friction
velocities shown in Figure 4-4. For a roughness height of 0.01 m and friction velocity of
0.5 m/s, the equivalent 10 m wind speed is 13.8 m/s (30 mph), a value that corresponds to
a fairly windy day at most locales. For low to moderate winds in arid to semi-arid areas
(i.e. u * < 0.5 m/s), the fraction of particles remaining in suspension after 500 seconds is
greater than 70% except for the very stable case. Note however that in the case of a
stable atmosphere, the wind speed is unlikely to exceed a few meters per second, since at
higher wind speeds, mechanical mixing at the surface would disallow the formation of a
stable layer to any appreciable height (i.e. more than a few tenths of a meter).
The trend of reduced deposition with increasing instability is consistent with
expectation. During stable conditions, dust particles stay close to the ground for longer
periods of time. Therefore, they are removed more quickly. In contrast, during unstable
conditions, particles are lofted to higher elevations quickly by buoyant eddies. This
means that those particles are not immediately available for deposition to ground
surfaces. The concentration profiles are compared among different conditions of
atmospheric stability in Figure 4-5. It is clear from the figure that unstable conditions
result in the rapid vertical dispersion of particles.
The removal of dust in the area near a source is often assumed to be significant
because concentrations of particles downwind of a source rapidly attenuate to the
background. This is an erroneous deduction. Watson and Chow (2000) and Countess
(2001) both cite an earlier study (Watson et al., 1996) where the concentration of PM 10
dust was found to decrease by 90% only 50 m downwind from the source. The
concentration of particles may decrease rapidly with downwind distance, but it is
incorrect to assume that the decrease is due solely to deposition. This concept is
illustrated in Figure 4-6. The figure clearly shows that while concentrations decrease
rapidly downwind of the source, the actual fraction of particles remaining in suspension
may be quite high. For example, according to the ISC3 model, under neutral conditions,
the concentration at a height of 1 meter 200 meters downwind of the road is 6% of its
value near the road (within a few meters). However, 90% of the particles are still in
suspension at the same downwind distance. Clearly, the decrease in concentration can be
due primarily to the vertical mixing of particles and is not necessarily due to deposition.
4-14

1
0.9
0.8
1
0.9
0.8
Fraction remaining
0.7
0.6
0.5
0.4
0.3 ADE u*=0.1
ADE u*=0.3
0.2
ADE u*=0.5
0.1 ADE u*=0.8
ADE u*=1.2
0
0 50 100 150 200 250 300 350 400 450 500
Time since release (s)
Fraction remaining
0.7
0.6
0.5
0.4
0.3 ADE u*=0.1
ADE u*=0.3
0.2
ADE u*=0.5
0.1 ADE u*=0.8
ADE u*=1.2
0
0 50 100 150 200 250 300 350 400 450 500
Time since release (s)
a. very stable (L=10 m) b. stable (L=1,000 m)
1
0.9
0.8
1
0.9
0.8
Fraction remaining
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ADE u*=0.1
ADE u*=0.3
ADE u*=0.5
ADE u*=0.8
ADE u*=1.2
0 50 100 150 200 250 300 350 400 450 500
Time since release (s)
Fraction remaining
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ADE u*=0.1
ADE u*=0.3
ADE u*=0.5
ADE u*=0.8
ADE u*=1.2
0 50 100 150 200 250 300 350 400 450 500
Time since release (s)
c. neutral |L|>100,000 d. unstable (L=-1,000 m)
Fraction remaining
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ADE u*=0.1
ADE u*=0.3
ADE u*=0.5
ADE u*=0.8
ADE u*=1.2
0 50 100 150 200 250 300 350 400 450 500
Time since release (s)
e. very unstable (L=-10 m)
Figure 4-4. Fraction of particles remaining in suspension vs. time since emission in the near field
(0

200
180
160
140
Very Stable
Stable
Neutral
Unstable
Very Unstable
Height z (m)
120
100
80
60
40
20
0
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
Concentration Normalized to Initial Release
200
a. t = 10 seconds
180
160
140
Very Stable
Stable
Neutral
Unstable
Very Unstable
Height z (m)
120
100
80
60
40
20
0
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Concentration Normalized to Initial Release
200
b. t=30 seconds
180
160
140
Very Stable
Stable
Neutral
Unstable
Very Unstable
Height z (m)
120
100
80
60
40
20
0
-0.1 0.0 0.1 0.2 0.3 0.4 0.5
Concentration Normalized to Initial Release
C. t=100 seconds
Figure 4-5. Concentration profiles according to the solution of the atmospheric diffusion equation at
a. t=10 seconds , b. t=30 seconds, c. t=100 seconds. U * = 0.3 m/s in all cases.
4-16

Fraction of Particles Remaining in Suspension or
Normalized concentration
1
0.8
0.6
0.4
0.2
0
ADE-Normalized concentration
ADE-Fraction in suspension
ISC-Normalized concentration
ISC-Fraction in suspension
0 100 200 300 400 500 600 700 800 900 1000
Distance Downwind (m)
Figure 4-6. The fraction of particles remaining in suspension and the Concentration at a height of 1
meter above ground level vs. time for neutral atmospheric conditions according to the ADE and ISC3
models. z 0 =0.01 m, D p =8 µm, u * = 0.3 m/s in all cases.
The fraction of particles removed is shown for multiple downwind distances in
Figure 4-7 a-e. Results are shown for both the ISC method and the ADE. For the ADE,
the distances are computed by multiplying the wind speed at 10 meters by the time in
seconds (The one-dimensional ADE is only dependent on time). The wind speed at 10
meters is in turn based on the assumption of a logarithmic wind profile with z 0 = 0.01 m.
One interesting result that is common to both models is that when the friction
velocity is 0.1 m/s, the removal of particles is greater than when it is 0.3 m/s; however, as
u * continues to increase, the removal of particles also increases. The key to
understanding this behavior is that the deposition velocity for particles also increases with
friction velocity. For values of u * that are higher than a certain threshold (near 0.3 m/s
for the conditions of Figure 4-7) the greater mixing that occurs is more than offset by the
greater deposition rate, resulting in a net increase in particle removal. The opposite is
true for values of u * that are less than the threshold.
Both ADE and ISC3 models indicate that under neutral to unstable conditions, the
fraction of particles remaining in suspension for values of u * ≤ 0.5 m/s is greater than
78% at a downwind distance of 1 km and greater than 59% at a downwind distances of
10 km. Even for a very high friction velocity (u * =1.2 m/s) the fraction of particles
remaining in suspension 1 km downwind is greater than 50%. Note that in arid regions,
very high values of u * initiate wind blown dust storms, dwarfing the emissions of dust
from an unpaved road. For example, the unpaved road tests at Ft. Bliss were suspended
on several occasions because of windblown dust events. These events occurred when u *
was greater than 0.6 m/s, approximately equivalent to a wind speed of 14 m/s at a height
of 10 m.
4-17

Fraction Remaining
Fraction Remaining
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ISC
ISC
ISC
ISC
ISC
ISC
ISC
ISC
ISC
ISC
ADE, u*=0.1
ADE, u*=0.3
ADE, u*=0.5
ADE, u*=0.8
ADE, u*=1.2
100 500 1000 5000 10000
Distance Downwind(m)
Fraction Remaining
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ISC
ISC
ISC
ISC
ISC
ADE, u*=0.1
ADE, u*=0.3
ADE, u*=0.5
ADE, u*=0.8
ADE, u*=1.2
100 500 1000 5000 10000
Distance Downwind(m)
a. very stable (L=10 m) b. stable (L=1,000 m)
ADE, u*=0.1
ADE, u*=0.3
ADE, u*=0.5
ADE, u*=0.8
ADE, u*=1.2
100 500 1000 5000 10000
Distance Downwind(m)
Fraction Remaining
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ISC
ISC
ISC
ISC
ISC
ADE, u*=0.1
ADE, u*=0.3
ADE, u*=0.5
ADE, u*=0.8
ADE, u*=1.2
100 500 1000 5000 1000
Distance Downwind(m)
c. neutral |L|>100,000 d. unstable (L=-1,000 m)
Fraction Remaining
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
ISC ADE, u*=0.1
ISC ADE, u*=0.3
ISC ADE, u*=0.5
ISC ADE, u*=0.8
ISC ADE, u*=1.2
100 500 100 500 100
Distance Downwind(m)
e. very unstable (L=-10 m)
Fraction Remaining
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
u*=0.1
u*=0.3
u*=0.5
u*=0.8
u*=1.2
A=0.04 A=0.08 A=0.12 A=0.2 A=0.3
Dispersion Parameter in Box Model
f. Box model for various values of the
parameter A G and V d (1 m)
Figure 4-7. Fraction of particles remaining in suspension for a given downwind distance. Panels a
through e correspond to numerical solutions to the atmospheric diffusion equation with z 0 =0.01 m,
D p =8 µm. under various conditions of atmospheric stability. The downwind distance is estimated
based on the wind speed at 10 m assuming a logarithmic profile and z 0 = 0.01 m. Panel f shows the
fraction remaining according to the box model for various assumed values of A G .
The degree of particle removal is greater under stable conditions than neutral and
unstable conditions. This is intuitive, since in the absence of buoyant mixing, particles
are likely to remain closer to the ground where they can be removed by impaction and
4-18

gravitational settling. Stable conditions cannot exist when the friction velocity is high
because mechanical mixing does not allow for stratification to occur to any appreciable
extent (there is always a small stable layer near the ground at night, though it may only be
a few centimeters in depth when wind speeds are high). Therefore, conservatively
assuming that u * does not exceed 0.5 m/s when conditions are stable, the removal of
particles 1 km downwind of the unpaved road is not likely to be greater than about 50%.
There is an exception to this result that occurs when the friction velocity is nearly zero
(i.e. very little vertical mixing and downwind transport) and particles settle to the ground
under the influence of gravity. In this case, the dispersion models presented here are not
applicable, and the fraction of particles remaining in suspension decreases linearly with
time since emission (e.g. Figure 4-8). It seems unlikely that there would be a significant
amount of motor vehicle traffic on unpaved roads during nighttime stable conditions.
Thus, though removal of particles under stable conditions is considered for completeness,
we note that for most emissions from unpaved roads, conditions are probably neutral to
unstable.
1.2
Fraction remaining in suspension
1
0.8
0.6
0.4
0.2
0
0 200 400 600 800 1000 1200
Time since Release (s)
Figure 4-8. Fraction of particles remaining in suspension vs. time since emission for 8 µm particles
falling under the influence of gravity in the absence of any dispersion (u * =0).
Comparing the ADE and ISC model predictions, the predictions of fraction
remaining in suspension are very similar, though some systematic differences exist. Very
near the source (first few hundred meters), the assumptions of the Atmospheric Diffusion
Equation are violated, because vertical gradients in concentration are large compared to
the absolute values of concentration. This results in a significant over prediction of the
initial dispersion of the plume. In the ISC model, the vertical extent of the plume is
specified as a function of downwind distance (through the parameter σ z ) and is therefore
not dependent on the steepness of the concentration gradient. The difference in initial
downwind dispersion between the ADE and the ISC is examined further in the next
section. Another reason for the difference between ADE and ISC predictions is that the
ADE accounts for dispersion explicitly through the friction velocity u * . That is, the
4-19

downwind concentration profile is dependent on u * . In contrast, the ISC model
prescribes the concentration profile based on downwind distance only. In practice, this
means that the ISC does not allow for alteration of the concentration profile due to
variations in u * . This point is illustrated in the next chapter.
Figure 4-7 f. shows the fraction remaining in suspension according to the box
model for various values of the parameter A in Equation 4-29. The deposition velocity
for the box model was computed assuming a height of 1 meter and neutral stability. Note
that the box model gives a result that is independent of distance downwind of the source.
For all values of A, the box model captures the general behavior of particle deposition to
a surface; the greater the dispersion rate (as characterized by A and u * ), the smaller the
fraction of particles that deposit. However, comparison of Figure 4-7f. with the other
five panels in the figure (ADE and ISC3 model results) underscores the basic difficulty of
using the box model: Though in its derivation the box model does not depend on the
height of the box or the downwind distance, the parameter A used in the model depends
on both. A also depends on atmospheric stability.
It is clear from Figure 4-5 that for a line source, the concentration profile through
the surface layer is dynamic, becoming flatter in the vertical direction with downwind
distance. This means that the concentration at the top of the surface layer (C SL in the
previous section) is neither constant in time, nor is it negligibly small compared to the
concentration in the box (C Box ). Thus, the formulation for dispersion in the Box Model is
not valid for a plume that is expanding in the vertical direction with concurrent decay in
the concentration gradients within the surface layer.
In a dust storm (Gillette, 1974), the ground acts like an area source, which
behaves quite differently in the context of the Box Model from a line source for three
reasons. First, an area source is horizontally homogeneous. Therefore, the vertical
concentration profile is, on average, invariant with downwind distance. Second, the
ground is acting as a strong source and the net flux of material is always in the upward
direction. This means that we do not have to account for particles re-entering the Box
through the ceiling. Third, because in a dust storm, the ground is a very strong source,
the concentration near the ground is much higher than it is aloft (i.e. C Box >> C SL ), or at
least, the ratio C Box /C SL is constant over large distances in the horizontal. For these three
reasons, dispersion in a dust storm can be estimated based only on the concentration near
the ground (as in Equation 4-27), without considering the concentration profile aloft. For
a dust plume that is expanding while traveling downwind, the assumptions behind
Equation 4-27 do not hold as well.
4.3 Summary
For unpaved road dust emissions, the Box Model provides an order of magnitude
estimate of the dust particle removal due to deposition. For a more accurate assessment, a
model that accounts for changes in the concentration profile with downwind distance is
required. Among the models considered in this section, the algorithm of the ISC3 is most
suited for simulating near-source dispersion. The ADE performs poorly in the near field,
4-20

though it is expected to improve at larger downwind distances (500 - 1,000 m). In the
next two chapters, the ISC3 and ADE models are compared with field data.
4-21

4-22

5. DETERMINATION OF PARTICLE REMOVAL RATES,
EMISSION FACTORS, AND INJECTION HEIGHTS ON AN
UNPAVED ROAD IN AN ARID SETTING
As part of an ongoing DoD-funded contract, leveraged by this WESTAR project,
field experiments were performed at the Ft. Bliss military facility near El Paso TX.
Towers instrumented with particle monitors were used to measure the vertical
concentration profiles and horizontal flux of PM 10 dust particles emitted from an unpaved
road. These data provided both an estimate of emission factors for a number of vehicles
and an estimate of the fraction of PM 10 emissions that is regionally transportable. During
the same field campaign, the turbulence behind moving vehicles was examined because
of a hypothesis that the spatial extent of the wake may have a strong influence on the
transportable fraction of emitted dust particles. In this Chapter, the results of the Ft. Bliss
field tests are presented. Those results are compared with predictions from both the ADE
and the ISC models.
5.1 Methods
5.1.1 Upwind/Downwind Towers
From April 11 through 24, 2002 unpaved road emissions experiments were
conducted at Ft. Bliss. The procedure was based on an upwind/downwind technique that
has been used by other investigators (e.g., Gillies et al., 1999; Cowherd, 1999). Three
towers were set up collinearly and perpendicular to a 1,000 m section of unpaved road on
the open range of Ft. Bliss, TX, which was oriented in a north-south direction. Historical
meteorological data indicated that winds at this time of year in this area were
predominantly from the west.
The three towers were all downwind of the road at distances of 7 m, 50 m, and
100 m. The tower closest to the road was 6 m high with the other two being 12 m in
height(Figure 5-1). Each downwind tower was instrumented with four DustTraks (Model
8520, TSI Inc., St. Paul, MN) that were spaced logarithmically in the vertical direction.
The DustTrak is a portable, battery-operated, laser-photometer that uses light scattering
technology to determine mass concentration in real-time. In applied studies of dust
emissions from playa surfaces and partially vegetated surfaces Houser and Nickling
(2001), and Nickling et al. (1997), respectively, found these instruments to be superior to
the traditional method of collecting suspended sediment on filters. Etyemezian et al.
(submitted) and Kuhns et al. (2001; submitted) used DustTraks to measure PM 10
emission from paved and unpaved roadways using an on-board vehicle measurement
system.
The tower at 7 m downwind (DT_1) had measurement positions at 0.76, 1.28,
2.66, and 5.18 m above ground level (AGL). The second tower (DT_2) at 50 m had
measurement positions at 1.25, 2.6, 5.7, and 12.2 m. The third tower at 100 m (DT_3)
had the same measurement positions as DT_2 with an additional sampling location at 0.4
m. The DustTraks were equipped with PM 10 inlets and measured particle concentrations
at intervals of 1 second. Four anemometers, one wind vane, and one temperature probe
were mounted on the DT_3 tower in order to characterize the local meteorological
conditions. During the course of the study four GRIMM model 1.108 particle size
5-1

analyzers (GRIMM Tech. Inc., Atlanta, GA) were deployed at various locations and
different measurement periods to provide information on the particle size characteristics
of the emitted dust.
-DT
-SA
-DT
Top View
Legend
DT_3
DT_2
DT_1
1220
DT: DustTrak
GR: GRIMM
SA: Sonic Anemometer
: Wind Vane
: Cup Anemometer
: Laptop computer
1220
Trailer with
visibility
equipment
Generator
200 meters
LIDAR 3,000
meters
-DT
-GR
-DT
-GR
-DT
570
570
517
-DT
-GR
-DT
-DT-GR
260
260
125
40
-DT
-DT
-GR
125
-DT
-GR
266 128
76
-DT
-DT
10,000
5,000
700
Figure 5-1. Schematic of equipment setup at Ft. Bliss during the April, 2002 field campaign. The
vertical and horizontal components are drawn on different scales.
The concentration data from the DustTraks, particle size distribution data from the
GRIMMs, and meteorological data were collected in 1-second intervals on a PC located
at the base of each tower running a custom-designed LabView (National Instruments)
data acquisition program. The PC clocks were set to a master clock at the beginning of
each day to ensure the data collected from the towers could be merged and synchronized
into a master database. The data files were created in Microsoft Access readable format.
At the beginning of each day, the zero baselines were set for each DustTrak on the
flux towers with a HEPA filter. Analysis of the DustTrak data indicated that baseline
drift over the course of a day, generally less than 50 µg/m 3 , did not affect all instruments
equally. Thus, a baseline for each instrument was calculated approximately every 20
minutes over the course of the day from background concentrations measured during
periods when there were no dust emissions from the unpaved road. PM 10 concentrations
measured during periods when the tower was influenced by a passing dust plume were
corrected with the most recent baseline value that was determined. This ensured that the
resulting concentration was due only to road dust emissions generated by passing
vehicles.
A portable filter sampler with a PM 10 inlet (Aerometrics, MiniVol) was collocated
with a DustTrak at 1.5 m on DT_1. The filter sampler was operated for 6 hours during
testing. The Concentration of PM 10 measured with on the filter was compared to the
average concentration measured by the DustTrak as a means of calibrating the DustTrak
light scattering measurement with a mass measurement. The results indicated that the
5-2

PM 10 measured by the DustTrak must be multiplied by 1.35 ± 0.15 in order to obtain an
equivalent PM 10 mass-based measurement. The DustTrak-based emission factors
presented in this report have been corrected to reflect an equivalent mass-based
measurement (i.e. the DustTrak signal has been multiplied by 1.35).
PM 10 emissions fluxes were calculated for each downwind tower using the
assumption that each DustTrak represented the PM 10 concentration over a height that
spanned half the distance to the next lowest monitor to half the distance to the next
highest monitor. For example, the lowest monitor at 0.76 m AGL on DT_1 was
considered representative of the PM 10 concentration from the ground surface to 1.02 m
AGL. The second monitor at 1.28 m AGL represented the PM 10 concentrations from
1.02 m agl to 1.94 m agl, and so on. The time series of PM 10 concentrations were
examined for each DustTrak at each location. Each peak in concentration was associated
with an individual vehicle pass and each pass was visually assigned a start and stop time.
Figure 5-2 shows an example of a concentration time series and the determination of peak
start and stop times. The emissions factor per vehicle pass for each downwind tower was
calculated using the sum of the 1-second PM 10 fluxes with the equation:
EF =
endofpeak
startofpeak
cos(θ
)
4
i=
1
u C ∆z
∆t
i
i
i
Equation 5-1
where EF is the emissions factor of PM 10 in grams per vehicle kilometer traveled, θ is the
angle (1-second measurement) between the wind direction and a line perpendicular to the
road, i is one of the four positions (five for T_3) of the monitors on the tower, u i is the
average wind speed in m/s over the interval represented by the i th monitor (1-second
measurement), C i is the baseline-corrected PM 10 concentration in mg/m 3 as measured by
the i th monitor over the period ∆t (1-second measurement), ∆z in m is the vertical interval
represented by the i th monitor, ∆t is equal to 1 second. Note that if multiplied by the
length of the unpaved road, the bracketed term following the first summation symbol
would be the amount of PM 10 that passes through a vertical plane in an interval of time
equal to ∆t (1 second).
The friction velocity, roughness height, and displacement height (u * , z 0 , and d,
respectively) where estimated from the vertical distribution of wind speeds. The wind
profile was assumed to be logarithmic as in Equation 2-9 with the displacement height, d,
equal to zero. This is a reasonable assumption considering the sparseness of vegetation at
Ft. Bliss and that d is expected to be much less than the height of the first anemometer
(1.25 m), making it difficult to accurately determine (since z-d≈ z). In order to obtain
meaningful estimates of the bulk quantities u * and z 0 , data were averaged over 15-minute
intervals to obtain the vertical distribution of wind speeds. There are 110 15-minute
intervals that corresponded to times during which sampling was in progress. For each 15-
minute interval and at each height where the wind speed was measured, the friction
velocity, u * was calculated from an initial guess for z 0 and the measured values of u(z),
yielding four calculated values of u * . These four values were averaged and the average
friction velocity was used to obtain a velocity according to Equation 2-9 for each of the
5-3

four heights. The root mean sum of squares (RMS) was then computed from the relative
difference between the calculated velocities and the measured velocities. The sum of the
RMS over all 110 intervals was used as a weighting function. The final estimate for z 0
was obtained by minimizing the weighting function. Note that this method is superior to
performing a regression for each 15-minute interval, since it provides a single value for z 0
that is applicable to all intervals.
20
15
DustTrak Reading (mg/m3)
10
5
Vehicle passes by DT_1
0
Baseline
DT_1 DT_2
DT_3
-5
18:02:10 18:02:27 18:02:44 18:03:01 18:03:19 18:03:36 18:03:53 18:04:11
Time
Figure 5-2. Example of time series of DustTrak PM 10 concentrations after the passage of a vehicle on
the unpaved road at Ft. Bliss. The arrows in the figure illustrate the start and stop times estimated
for a baseline reading and the dust plume passing through the downwind towers DT_1, DT_2, and
DT_3.
The final value of 0.005 m obtained for z 0 lies between those expected for “level
desert” (0.001 m) and “lawn” (0.01 m) (Seinfeld and Pandis, 1999). Figure 5-3 shows the
distribution of u * values over the 111 intervals. Ninety-five percent of the values of u *
fell between 10 and 50 cm/s, with values between 30 and 40 cm/s occurring most
frequently (35%). Examples of how the wind speed fitted to a logarithmic vertical profile
compares with the measured wind speed are given in Figure 5-4 for conditions
corresponding to the lightest, most frequent, and heaviest winds observed over the 110
15-minute intervals. In general, the logarithmic fit is reasonable for all cases, though the
wind speed at 12.2 meters is overestimated in the case of light winds. This may be caused
by small differences in the resistance to motion among the four cup anemometers used.
The fact that in Figure 5-4a the measured wind speed at 12.2 meters is slightly less than
that at 5.7 meters supports this hypothesis.
5.1.1.1 Emission Tests
Road dust emissions were created by having a test vehicle travel back and forth
along the roadway for a number of passes. This ranged from a minimum of 20 to a
maximum of 102. The mean number of vehicle passes was 49. The test vehicles traveled
at set speeds of 16, 24, 32, 40, 48, 56, 64, 72, and 81 km/hr. Not all vehicles attained the
highest speeds for safety considerations. After two passes (one heading south and one
returning north) at the same speed, the vehicle speed was increased incrementally to the
5-4

100%
90%
80%
Relative Frequency (probability)
Cumulative %
Relative Frequency
70%
60%
50%
40%
30%
20%
10%
0%
0 -10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 > 60
u* (cm/s)
Figure 5-3. Frequency distribution and cumulative distribution of u * for the 110 15-minute intervals
corresponding to periods when testing was ongoing at Ft. Bliss. The u * values are based on a
roughness height, z 0 equal to 0.005 m.
u*= 10 cm/s
u*= 32 cm/s
u*= 54 cm/s
14
u(z) measured
u(z) fitted
14
u(z) measured
u(z) fitted
14
u(z) measured
u(z) fitted
12
12
12
Height (z) in meters
10
8
6
4
Height (z) in meters
10
8
6
4
Height (z) in meters
10
8
6
4
2
2
2
0
12 16 20 24
u(z)/u *
0
12 16 20 24
u(z)/u*
a. b. c.
0
12 16 20 24
u(z)/u*
Figure 5-4. Examples of comparison between measured wind speed and wind speed calculated from
curve fitting of Equation 2-9. a. light winds (u * =10 cm/s), b. medium winds (u * = 32 cm/s), c. high
winds (u * = 54 cm/s). The roughness height, z 0 is equal to 0.005 m in all cases.
5-5

highest attainable value and then decreased incrementally to the minimum value. This
cycle was then repeated as time allowed. Test vehicles paused for approximately one and
a half minutes between passes to allow for the dust plume to clear all three downwind
towers. The schedule and test conditions are summarized in Table 5-1.
Table 5-1. Test dates, times, vehicles, data recovery, and conditions during the FT. Bliss, April 2002
field campaign.
Date
4/11/2002
4/11/2002
4/12/2002
4/12/2002
4/15/2002
4/15/2002
4/16/2002
4/16/2002
4/16/2002
4/17/2002
4/18/2002
4/19/2002
4/22/2002
4/23/2002
4/24/2002
4/24/2002
Time
13:44 -
14:51
15:11 -
16:06
14:41 -
15:21
15:35 -
16:13
10:38 -
12:37
12:48 -
13:30
10:57 -
12:26
13:53 -
15:35
17:18 -
19:20
10:22 -
12:36
10:56 -
13:15
11:06 -
14:29
9:47 -
10:08
13:38 -
15:11
10:50 -
12:33
13:41 -
15:16
Equipment
status
DT3 only, No
wind
direction data
DT3 only, No
wind
direction data
DT3 only
DT3 only
DT1, DT2,
DT3
DT1-partial,
DT2, DT3
DT1, DT2,
DT3
DT2, DT3
DT1, DT2,
DT3
DT1, DT2,
DT3
DT1, DT2,
DT3
DT1-partial,
DT2, DT3
DT1, DT2,
DT3
DT1-partial,
DT2, DT3
DT1, DT2,
DT3
DT1, DT2,
DT3
Vehicle
HUMVEE
Dodge
Caravan
Dodge
Caravan
Dodge
Neon
Ford Taurus
U-Haul
Dodge
Caravan
5-ton truck
1979 Chevy
van
(TRAKER)
LMTV
Hemmet
22
Freightliner
Dodge
Neon
22
Freightliner
1979 Chevy
van
(TRAKER)
Vehicle
speeds
(km/hr)
8, 16, 24, 32,
40, 48, 56,
64, 72
16, 32, 48,
64, 81
Total
number of
valid passes
26
Wind conditions
Light and variable with small
westerly componenet
40 South -West
48 30 Light and variable
16, 24, 32,
48, 64, 81
16, 32, 48,
64, 81
16, 32, 48,
64
16, 32, 48,
64, 81
16, 32, 48
,64
16, 24, 32,
40, 48, 64
16, 32, 48,
64, 81
16, 32, 48,
64
16, 24, 32,
40, 48, 56
16, 32, 48,
64
16, 24, 32,
40, 48, 56
16, 24, 32,
40, 48, 56
HUMVEE 16, 64 2
12 Light and variable
23 South -West
11 South -West
29 South -West
27 South -West
46 South -West
16 South -West
16 South -West, with gusts
11 South -West, with gusts
0 South: Test suspended
18 West, with gusts
0
Light winds from the North.
No valid data obtained
Winds light and variable. Data
recovery poor.
The unpaved road at Ft. Bliss was 1 kilometer in length and oriented in the northsouth
direction, with the three towers lined perpendicular to and east of the road. Thus,
when winds did not have a strong westerly component, the dust plume generated by
passing vehicles did not travel in the direction of the towers. The transport time between
the road and the far downwind tower (100 m) was estimated to be approximately 30
seconds. One-second wind direction data were averaged for the 30 seconds prior to plume
impact at the far downwind tower. If the average direction fell in the quadrant spanned
by the northwest and southwest vectors, the corresponding vehicle pass was considered
valid. Vehicle passes corresponding to wind directions that were not in the northwestsouthwest
quadrant were considered suspect and not included in subsequent data analysis.
Note that for westerly winds, a 30-second transport time corresponds to a wind speed of
3.3 m/s while for southwesterly or northwesterly winds, it corresponds to 4.7 m/s. The
5-6

average wind speed measured at 12 meters over all emissions tests was 6.2 m/s. Thus, a
transport time of 30 seconds is probably, on average, a conservative overestimate.
The wind speed and direction were measured only at the far downwind tower
(DT_3). Therefore, for using Equation 5-1, it was necessary to use wind speeds and
directions measured at DT_3 for DT_1 and DT_2. The possibility of introducing errors
by this method was examined. The horizontal PM 10 fluxes were calculated for DT_3 only
and averaged over the test vehicle speed. This procedure was performed twice, once with
wind data that corresponded to the time of the passing of the dust plume, and once with
wind data that was retarded by 30 seconds. For example, the flux at 13:30:00 was
calculated using wind data measured at 13:30:00 and also with wind data measured at
13:29:30. The purpose of this exercise was to subject the flux calculations for DT_3 to
the same uncertainties that the data from DT_1 and DT_2 would be experiencing. In
comparing the two resultant sets of horizontal fluxes, no significant difference were
found (slope = 1.01, R 2 =0.98, n=53, intercept was forced to 0), indicating that the use of
DT_3 wind direction and wind speed to calculate horizontal fluxes of PM 10 at DT_1 and
DT_2 would not introduce errors.
5.1.2 Sonic Anemometer Tests
A three-dimensional sonic anemometer (“A” style probe, applied Technologies
Inc) was used to assess the effect of vehicle size on the initial distribution of the dust
plume generated. The anemometer was first collocated with the cup anemometer and the
wind vane at the 12.2 m height on DT_3 for three sample days (4/18/02 – 4/20/02). The
sonic anemometer was set to measure the U, V, and W components (corresponding to the
x, y, and z directions) of the velocity at a frequency of 10 Hz. Figure 5-5 shows the
comparison between wind speeds measured with the cup anemometer and a sonic
anemometer; part a. shows a time series; parts b. through d. show comparisons of 1-
second averages and 1-minute averages. In general, the cup and the sonic anemometers
track each other well, though the sonic anemometer gives smaller values of wind speed as
indicated by the slopes of the regressions in Figure 5-5 b-d (0.91 – 0.94). The significant
improvement in R 2 values between the 1-second (0.84-0.89) and 1-minute averaged data
(≥ 0.98) is to be expected since the cup anemometer has a nominal response time of 1
second (approximately the time required to register 1/3 of the difference between changes
in wind speed) while the sonic anemometer response is practically instantaneous.
On 4/21/02 between 19:44 and 22:15 a series of tests were conducted to assess the
magnitude of the turbulent wake behind a passing vehicle. The tests occurred at night
because under stable nighttime atmospheric conditions, the turbulence generated by a
vehicle can be more readily identified with a sonic anemometer than when the
atmosphere is unstable and the background turbulence level is much higher. The
downwind tower closest to the unpaved road was moved to the edge of the road. Two
markers were placed on the unpaved road, one at a distance of 2 meters from the edge of
the road (and the tower) and the other at a distance of 6 meters. The sonic anemometer
was mounted on the tower at a height of 0.9 meters above ground level (AGL) and
protruded 1.1 meters towards the road (i.e. approximately 1 meter from the near marker).
Three vehicles with considerably different profiles were tested: a Dodge Neon, a 1979
Chevy cargo van, and a 24-foot moving truck. Each vehicle was driven through the two
markers on the road once heading north and once heading south at 16, 32, 48, and 64
5-7

km/hr for a total of eight passes per vehicle. The time, accurate to within two seconds,
was recorded whenever the vehicle passed immediately in front of the sonic anemometer.
After all three vehicles completed this cycle, the sonic anemometer was moved to 2.4
meters and then to 5.5 meters AGL. The driving pattern (three vehicles, four speeds, two
passes at each speed) was repeated at each of those two heights.
Under stable atmospheric conditions, natural fluctuations in the vertical
component of velocity (W) are expected to be small. Thus, W was chosen to assess the
vertical extent of the turbulent wake generated by the passage of the vehicles. This was
done by making use of Reynold’s decomposition
W ( t)
= Wave + W '
Equation 5-2
where W(t) is the instantaneous velocity at time t, W ave is the average component of the
vertical velocity, and W’ is the fluctuating component of the velocity. Upon inspection of
the data, it became apparent that the effect of the passing vehicle could be seen from 7
seconds prior to the recorded vehicle pass time to 15 seconds after the recorded time.
This was not the case for every vehicle pass, but this “influence” interval (-7 to +15
seconds) was large enough to be inclusive of all vehicle passes. W ave in Equation 5-2 was
obtained by averaging the W-component of the velocity for 30 seconds prior to the
beginning of the interval influenced by the vehicle (i.e. –37 to –7 seconds) and for 30
seconds following the interval (+15 to +45 seconds). These two 30-second intervals were
considered “background” conditions. The background turbulence was quantified by
taking the average and standard deviation of J b over each 0.1 second measurement
obtained during the entire “background” period where
J
b
=
( W t)
−W
) 2
(
b ave
1
Equation 5-3
and the subscript b indicates that this operation is performed for the background intervals.
The turbulence generated by the passing of the vehicle is likewise quantified by obtaining
the average of J i
J
i
=
( W t)
−W
) 2
(
i ave
Equation 5-4
1 As an aside, note that J can be thought of as the vertical component of the turbulence
kinetic energy (KE) which is customarily defined as
2 2
KE = U ' + V ' +
W
2
'
where U and V represent the x- and y- component velocities.
5-8

where the subscript i indicates that this operation is performed for the influence interval (-
7 to + 15 seconds). Figure 5-6 shows an example of the data resulting from the passage of
a vehicle and the corresponding background and influence intervals.
12
10
Wind speed (m/s)
8
6
4
2
Sonic anemometer
Cup anemometer
0
9:49:41 9:50:24 9:51:07 9:51:50 9:52:34 9:53:17
Time of Day
a.
1-Second Wind Speed Measured
with Sonic Anemometer (m/s)
1-Second Wind Speed Measured
with Sonic Anemometer (m/s)
1-Second Wind Speed Measured
with Sonic Anemometer (m/s)
20
15
10
5
y = 0.93x
R 2 = 0.89
4/18/02
0
0 5 10 15 20
20
15
10
5
1-Second Wind Speed Measured with Cup Anemometer (m/s)
y = 0.94x
R 2 = 0.84
4/19/02
0
0 5 10 15 20
30
25
20
15
10
5
1-Second Wind Speed Measured with Cup Anemometer (m/s)
y = 0.91x
R 2 = 0.89
4/20/02
0
0 5 10 15 20 25 30
1-Second Wind Speed Measured with Cup Anemometer (m/s)
b.
c.
d.
1-Minute Wind Speed Measured
with Sonic Anemometer (m/s)
1-Minute Wind Speed Measured
with Sonic Anemometer (m/s)
1-Minute Wind Speed Measured
with Sonic Anemometer (m/s)
20
15
10
5
y = 0.94x
R 2 = 0.99
4/18/02
0
0 5 10 15 20
20
15
10
5
1-Minute Wind Speed Measured with Cup Anemometer (m/s)
y = 0.94x
R 2 = 0.98
4/19/02
0
0 5 10 15 20
20
15
10
5
1-Minute Wind Speed Measured with Cup Anemometer (m/s)
y = 0.91x
R 2 = 0.99
4/20/02
0
0 5 10 15 20
1-Minute Wind Speed Measured with Cup Anemometer (m/s)
Figure 5-5. Comparison between collocated spinning cup and sonic anemometers. a. Example time
series from 4/18/02; b. regression of 1-second-averaged and 1-minute-averaged data from 4/18/02; c.
regression of 1-second-averaged and 1-minute-averaged data from 4/19/02; d. regression of 1-secondaveraged
and 1-minute-averaged data from 4/20/02.
5-9

0.8
Vertical velocity W (m/s)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
Background
interval #1:
-37 to -7
seconds
Influence
interval: -7
seconds to
+15 seconds
Recorded
pass time
Background
interval #2:
+15 to +45
seconds
00.5 10.0 19.5 29.0 38.5 48.0 57.5 07.0 16.5 26.0 35.5
Time (seconds)
Figure 5-6. Example of W-velocity as measured by the sonic anemometer in the vicinity of a vehicle
pass. The vertical lines indicate the start and end times for the two background intervals and the
influence interval.
5.2 Results
5.2.1 Dispersion and deposition downwind of the unpaved road at Ft. Bliss
For a ground-level release, it is as important to accurately represent dispersion as
deposition. The two processes act simultaneously and the removal of particles downwind
of a dust source is strongly dependent on both. The parameterization used for dispersion
in the box model was discussed in Chapter 4. This section examines the ability of the
numerical solution of the Atmospheric Diffusion Equation (ADE) and the Gaussian
distribution used in the ISC3 model (USEPA, 1995) to represent the concentration profile
downwind of a ground-level source. The results of the Ft. Bliss experiments are
compared with predicted ADE and ISC concentration profiles. This comparison is only
applicable to the distance between the unpaved road and the furthest downwind tower
(100 m). Data for comparison of concentration profiles over longer distances may be
obtained from a long-range scanning LIDAR that was operating concurrently with the
tower measurements. However, those data were not available at the time of the writing
of this report.
Both the ADE and the ISC3 are designed to handle continuous release sources
and consequently, downwind concentrations that do not vary with time (steady-state).
The experiments at Ft. Bliss examined the downwind plume from a vehicle that traversed
an unpaved road, first in one direction, then in the other. The result is that the time series
of the concentration measured at the downwind towers more resembles an instantaneous
puff release than a continuous source (see Figure 5-2). To compare the Ft. Bliss data
with the models, the concentration time series was integrated for each vehicle pass and
each location on the tower. That is
5-10

k
C int
=
peakend
k
C
peeakstart
dt
Equation 5-5
where C k int is the integral of the concentration C k at tower location k. For the purpose of
comparison, it was necessary to use a common scale for concentration. This was
accomplished by normalizing the concentrations to a common value for each tower.
While the location of the particle monitoring instruments (DustTraks) varied slightly
from one tower to the next, all towers were equipped with one monitor at a height of 1.26
m. Thus, the integrated concentrations for each individual pass and at each height on a
tower were normalized by the integrated concentration at 1.26 m for that tower. The
concentration profile from each of 109 passes was placed in a category according to the
value of the friction velocity, u * , associated with that pass. The normalized concentration
profiles were then averaged within each friction velocity category. There were four
categories in all for u * , corresponding to bins centered at 0.2 m/s, 0.3 m/s, 0.4 m/s, and
0.5 m/s. For example, all concentration profiles with a u * between 0.15 m/s and 0.25 m/s
were lumped together under the u * =0.2 category.
The validity of using Equation 5-5 to estimate the concentration profile from a
line source based on multiple passes of a moving point source (one vehicle traversing the
road) was assessed. On two occasions during the field campaign, large military convoys
traveled on the road near the towers. The convoys, consisting of tens of vehicles spaced
approximately 10 meters apart, are an excellent approximation to a true line source.
Figure 5-7 shows a comparison between the concentration profiles from a line source and
those from multiple moving point sources. Allowing for some scatter, especially in the
line source curves where only one data point was used for each convoy, the two sets of
profiles are very similar for both the 50 m (T2) and 100 m (T3) downwind towers. This
indicates that the treatment of an unpaved road as a line source is reasonable at the scale
of 100 meters or so. At larger distances, the observed differences between a true line
source and multiple moving point sources is expected to be small.
The integrated concentration profile for three towers at multiple downwind
distances from the road (7, 50, and 100 m) are shown in Figure 5-8 for two values of u * .
All tests were performed on sunny days after 10:00 AM. Therefore atmospheric stability
ranges from neutral to moderately unstable. As expected, the concentration profile is
steep near the road at the first tower. The profile becomes less steep as particles mix
vertically while being transported further downwind past the second and third towers.
The profiles for the u * =0.3 m/s case indicate a greater degree of dispersion than those for
the u * =0.5 m/s. This seems counter-intuitive since it is expected that higher values of
friction velocity would enhance dispersion. The data shown in Figure 5-8 do not
contradict that expectation; higher values of friction velocity also imply higher ambient
wind speeds and shorter transport times. Thus, while on a time basis, dispersion
increases with friction velocity, on a distance basis, the opposite appears to be true, at
least for the conditions of the experiment at Ft. Bliss.
5-11

14
12
10
Puff_average T2 u*=0.2
Convoy T2_u*=0.19
Puff_average T2 u*=0.5
Convoy T2_u*=0.47
Height z (m)
8
6
4
2
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalized Concentration
a. Tower 2 (50 m)
14
12
10
Height z (m)
8
6
4
Puff_average T3 u*=0.2
Convoy T3_u*=0.19
Puff_average T3 u*=0.5
Convoy T3_u*=0.47
2
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalized Concentration
b. Tower 3 (100 m)
Figure 5-7. Comparison of concentration profiles from the passage of two convoys with the
concentration profiles at equivalent friction velocities based on the average from multiple passes of a
moving point source. The line source is approximated by the passage of two convoys on two separate
occasions. The moving point source average concentration profiles were calculated as the integrated
concentration from Equation 5-5. Concentrations are normalized to the value at 1.26 meters in all
cases.
Figure 5-8 also shows the predicted concentration profile for each tower resulting
from the numerical solution of the one-dimensional Atmospheric Diffusion Equation. It
is clear from the figure that the ADE significantly over predicts dispersion in the nearfield.
This in turn means that the ADE is expected to under predict deposition over the
same distance scale. The ADE depends on mixing length theory for its validity. In
particular, the equation adequately describes dispersion provided that the length scale for
changes in the mean concentration is much larger than the length scale for turbulent
transport (Seinfeld and Pandis, 1999). This condition does not hold true for the region
near an isolated source such as an unpaved road, with the result that dispersion is
overestimated near the unpaved road.
5-12

Height z (m)
14
12
10
8
6
T1_data
T2_data
T3_data
T1_ADE
T2_ADE
T3_ADE
4
2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Normalized Concentration
a. u * = 0.3
Height z (m)
14
12
10
8
6
T1_data
T2_data
T3_data
T1_ADE
T2_ADE
T3_ADE
4
2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Normalized Concentration
b. u * = 0.5
Figure 5-8. Concentration profiles measured at the three towers downwind of an unpaved road and
concentration profiles predicted by the ADE model for equivalent downwind distances a. for 32
passes with u * between 0.15 and 0.25 m/s and b. for 15 passes with u * between 0.45 and 0.55 m/s.
Tower distances from edge of road: T1 - 7 m, T2 – 50 m, T3 – 100 m. Modeled profiles are based on
neutral atmospheric conditions. Both measured and modeled concentration have been normalized to
the concentration at 1.26 m.
It is interesting that the Gaussian model used in the ISC3 better emulates the
concentration profiles observed in the experiment, especially since the gaussian plume
equations are actually approximate analytical solutions to the Atmospheric Diffusion
Equation. This is illustrated in Figure 5-9 where normalized concentrations at Towers 2
and 3 are plotted on the same graph as the predicted gaussian concentration profile. The
value of σ z in the ISC3 model is based only on atmospheric stability and downwind
distance. Thus, friction velocity is not represented. This is partly why experimental
conditions corresponding to low values of u * are better represented by the “moderate” to
“very” unstable curves. Higher values of u * correspond to “slight” to “moderate”
5-13

Height z (m)
14
12
10
8
6
T2 u*=0.2
T2 u*=0.3
T2 u*=0.4
T2 u*=0.5
T2_ISC_neutral
T2_ISC_slightly unstable
T2_ISC_moderately unstable
T2_ISC_very unstable
4
2
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalized Concentration
a. Tower 2 (50 m)
Height z (m)
14
12
10
8
6
T3 u*=0.2
T3 u*=0.3
T3 u*=0.4
T3 u*=0.5
T3_ISC_neutral
T3_ISC_slightly unstable
T3_ISC_moderately unstable
T3_ISC_very unstable
4
2
0
0 0.2 0.4 0.6 0.8 1 1.2
Normalized Concentration
b. Tower 3 (100 m)
Figure 5-9. Normalized measured and modeled gaussian concentration profiles at a. Tower 2, 50
meters downwind of unpaved road and b. Tower 3, 100 meters downwind.
unstable conditions. The difference between the ADE and the ISC3 is that in the latter,
the dispersion parameter (embodied in σ z ) is allowed to vary with downwind distance,
while in the former (embodied in K zz ) the dispersion is only allowed to vary with height.
In practical terms, this means that the ISC3 has a correction factor built into it to account
for the shortcomings of the ADE close to a source. Figure 5-10 shows the downwind
progression of the concentration profiles according to the ADE and the ISC3 model. The
results for the ADE are shown for neutral conditions at two values of u * and those for the
ISC3 are shown for three stability conditions ranging from neutral to very unstable. At
far downwind distances (i.e. > 500 m), the profile for the ADE curve lies between
moderately unstable and neutral profiles from the ISC3. This is consistent with the
expectation that at further from the source, the assumptions of the ADE are better met
than in the immediate vicinity of the source.
5-14

Horizontal fluxes of PM 10 measured at each of the three downwind towers during
the 2002 Ft. Bliss experiments appear in Figure 5-11a. The horizontal flux was
calculated according to Equation 5-1 using PM 10 concentrations measured with
DustTraks at multiple heights on each of the downwind towers. For each pass, the
horizontal flux measured at each tower was normalized by the average flux for the three
towers. Because of the great variability among fluxes for individual vehicle passes, the
columns in the figure represent averages of the normalized flux over 27 passes when the
wind direction was known to be within 15 degrees of the vector perpendicular to the road.
The vertical bars are standard errors. Figure 5-11a indicates that there is no measurable
reduction in PM 10 horizontal fluxes for at least 100 meters downwind of the unpaved
road. The fluxes measured at T1 and T2 (50 m) are nearly identical, while those
measured at T3 (100 m) are slightly higher, but within two standard errors of T1. This
indicates that there is no statistically significant difference (α=0.05) between the
horizontal fluxes measured at any of the three towers and that for the conditions at Ft.
Bliss, the removal of PM 10 particles over a distance of 100 m downwind of the road is
negligible.
250
200
ISC3_neutral
ISC3_moderately_unstable
ISC3_Very_unstable
ADE_u*=0.3
ADE_u*=0.3
250
200
ISC3_neutral
ISC3_moderately_unstable
ISC3_Very_unstable
ADE_u*=0.3
ADE_u*=0.3
hieght z (m)
150
100
hieght z (m)
150
100
50
50
0
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Normalized Concentration
Normalized Concentration
a. 50 m b. 100 m
250
200
ISC3_neutral
ISC3_moderately_unstable
ISC3_Very_unstable
ADE_u*=0.3
ADE_u*=0.5
250
200
ISC3_neutral
ISC3_moderately_unstable
ISC3_Very_unstable
ADE_u*=0.3
ADE_u*=0.5
hieght z (m)
150
100
hieght z (m)
150
100
50
50
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Normalized Concentration
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Normalized Concentration
c. 500 m d. 1,000 m
Figure 5-10. Comparison of ISC3 gaussian concentration profile and profile from numerical solution
of Atmospheric Diffusion Equation at multiple downwind distances. Solutions for the ADE are for
neutral conditions.
5-15

Horizontal flux of PM10 by DustTrak normalized to the FLux at
Tower1
1.5
1.4
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
T1 (7 m) T2 (50 m) T3 (100 m)
a. Comparison of horizontal flux among the three downwind towers.
Relative concentration of particles in size range
normalized to mass in 3.9 micron size bin
1.2
1.0
0.8
0.6
0.4
0.2
0.0
Normalized concentration at 100 m (T3) divided by initial value
(T3) of ratio of mass in bin to mass in 3.9 micron size bin
ISC prediction (u* =0.35 m/s)
ADE prediction (u* = 0.35 m/s)
3.9 5.6 7.2 9.9 14 19.7
Particle Diameter (microns)
b. Comparison of particle size distributions at Towers 1 and 3
Figure 5-11. Particle removal rates measured at Ft. Bliss at tower 100 m downwind of an unpaved
road. a. Comparison of horizontal flux among the three downwind towers. Data are averages over 27
individual passes when the wind direction was within 15 degrees of perpendicular to the road.
Concentrations measured with DustTrak monitors with PM 10 inlets. Fluxes calculated according to
Equation 5-1 and normalized to value at Tower 1. Vertical bars indicate standard errors; b.
Comparison of particle size distributions at Towers 1 and 3 measured by GRIMM 1.108 OPC with
model predictions. Data are averages over 137 individual passes. X-axis shows particle aerodynamic
diameter calculated by assuming a density of 2.6 g/cm 3 and geometric mean diameter representing
each size bin. Particle concentrations at Towers 1 and 3 were each normalized by concentration of
3.9 micron particles. Y-axis represents ratios of normalized concentrations.
The finding that there is no measurable removal of PM 10 particles can be further
supported by examining the particle size distributions measured at T1 and T3. Figure
5-11b shows the ratio of particle concentrations measured at the same height (2.66 m) on
both towers. The figure represents data from 137 individual passes. The x-axis shows
the mean geometric diameter for the size bin measured. Aerodynamic diameters have
5-16

Table 5-2 enumerates the speed-dependent emission factors estimated from the
measurements at Ft. Bliss alongside those determined according to Equation 5-6.
Samples of road dirt were obtained both prior to and at the end of the field campaign.
The silt contents for those samples were 7% and 4%, respectively. The higher silt
content at the beginning of the study reflects two phenomena. First, on the day prior to
sample collection, there was a windstorm that may have deposited some fine soil material
onto the unpaved road. Second, the field tests themselves may have resulted in the
depletion of fine soil material from the road. Thus, during the field tests, it is reasonable
to assume that the silt content on the unpaved road was somewhere in between 4% and
7%. The AP-42 factors calculated using these two values as lower and upper bounds
appear in Table 5-2.
04/23_22 Freightliner
EF (g/vkt)
2500
2000
1500
1000
y = 38.397x
R 2 = 0.9458
Average
Linear (Average)
500
0
0 10 20 30 40 50 60
Speed (mph)
Figure 5-12. Example of emission factor dependence on vehicle speed. The vertical bars are the
standard deviations among the three downwind towers.
In general, the AP-42 factors overestimate PM 10 emissions for vehicles traveling
at low speeds compared to the measurements. This is illustrated in Figure 5-13, where
the ratios of the AP-42 values to measured values are plotted as a function of vehicle
speed. The top panel in the figure summarizes the data for all the vehicles in Table 5-2.
The figure illustrates that, depending on the assumed silt content, for speeds
approximately less than 20 mph, AP-42 overestimates PM 10 emissions while for speeds
greater than 25 mph, AP-42 underestimates emissions. It is noteworthy that most
emissions inventories assume that vehicles traveling on unpaved roads average speeds of
either 20 or 25 mph. This suggests, that at least from the standpoint of emissions
inventories, use of the data from the Ft. Bliss experiments would give results similar to
the AP-42. However, the vehicles in this study are on average much heavier than one
would expect to find on rural unpaved roads (industrial haul roads may be an exception).
The bottom panel of Figure 5-13 shows that if only passenger-sized vehicles are
considered, then the AP-42 may substantially overestimate PM 10 emissions for speeds up
to about 35 mph. In fact, in the 20 or 25 mph range, the AP-42 values are 1.5 to 2 times
those measured at Ft. Bliss. This suggests that the AP-42 emission factors may be
positively biased for passenger-sized vehicles. This bias can be a source of discrepancy
5-18

etween PM 10 emission inventories for crustal material and the amount of such material
found on ambient filter samples.
Table 5-2. Summary of vehicle emission factors measured at Ft. Bliss and Comparison with emission
factors calculated as specified in AP-42 (USEPA, 1999) for two values of silt content (4% and 7%);
moisture content was assumed to be 0.2%. The speed dependence of the measured emission factors
is given in the third column and represents the slope of a linear regression between measured
emission factor (g/vkt) and vehicle speed (mph); the R 2 value for the regression is given in the fourth
column. For cases where a particular vehicle was tested more than once, the average of the two tests
is used. All emission factors are reported in grams per vehicle kilometer traveled (g/vkt).
vehicle/ speed
GVWR
(kg)
Emission EF
(g/vkt/ mph) R 2 mph
Potential 10
EF
15
mph
Dodge Neon 1,622 9.8 0.37 82 123 164 205 246 287 328 369 410 247 387
Dodge
Caravan Test
1 2,495 17.5 0.85
Dodge
Caravan Test
1 2,341 15.5 0.70
Dodge
Caravan
Average 2,418 16.5 138 208 277 346 415 484 553 623 692 294 460
Ford Taurus 2,124 9.7 0.90 81 122 163 203 244 285 325 366 407 275 431
Humvee Test
1 2,445 10.0 0.93
Humvee Test
2 2,445 12.5 0.76
Humvee
Average 2,445 11.3 94 141 189 236 283 330 377 424 472 291 456
EF
20
mph
EF
25
mph
EF
30
mph
EF
35
mph
EF
40
mph
EF
45
mph
EF
50
mph
5-ton 9,818 76.0 0.82 637 956 1275 1593 1912 2230 2549 2868 3186 508 795
LMTV 8,060 30.0 0.80 252 377 503 629 755 880 1006 1132 1258 469 735
Freightliner
M915A4 Test
1 23,636 38.8 0.92
Freightliner
M915A4 Test
2 23,636 38.4 0.95
Freightliner
M915A4
Average 23,636 38.6 324 485 647 809 971 1133 1295 1456 1618 722 1130
Average for
all vehicles 230 345 460 574 689 804 919 1034 1149 401 627
Average for
small
vehicles 99 149 198 248 297 347 396 446 495 277 433
AP-
42
4%
AP-
42
7%
In addition to the neglect of vehicle speed, some of the inconsistencies in the AP-
42 emission factors for PM 10 dust from unpaved roads come from the “lumping” of the
vehicle weights in Equation 5-6. Emission factors calculated in this way are not selfconsistent.
For example, consider three scenarios a. 1,000 vehicles weighing 1 ton each,
b. 1,000 vehicles weighing 10 tons each, and c. a mix of 500 1-ton and 500 10-ton
vehicles. According to Equation 5-6, the relative emissions of PM 10 for the 1,000
vehicles in each of these three scenarios are respectively 1, 2.51, and 1.97. This goes
against the logical result that one would expect, namely that the emissions in case c.
should be equal to the average of those in a. and b (1.25).
5-19

6
Ratio of AP-42 Emission Factor to
Measured Emission Factor
5
4
3
2
silt content = 7%
1
Silt content = 4%
0
10 15 20 25 30 35 40 45 50
Speed (mph)
a. All vehicles in Table 5-2
6
Ratio of AP-42 Emission Factor to
Measured Emission Factor
5
4
3
2
1
silt content = 4%
silt content = 7%
0
10 15 20 25 30 35 40 45 50
Speed (mph)
b. same as a, but only considering Dodge Neonb, Dodge Caravan, Ford Taurus, and
Humvee (GMC)
Figure 5-13. The ratio of emission factors calculated according to AP-42 (USEPA, 1999) and those
measured at Ft. Bliss. The upper line is the ratio based on a 7% silt content, while the lower line
assumes 4% silt content. The top panel shows the ratios when all the vehicles in Table 5-2 are
considered. The bottom panel shows the same data, but only for the passenger vehicles.
5.2.3 Injection Height
5.2.3.1 Estimates of injection heights
The results of the nighttime wake turbulence tests are shown in Figure 5-14. Because of
the inherent variability in the data, it was necessary to combine the passes at all speeds
for each vehicle. This was accomplished by normalizing the square root of the difference
between the vehicle-influenced vertical fluctuations in velocity and the background
fluctuation with the vehicle speed. The figure illustrates that the size of the turbulent
wake behind a vehicle increases with physical size. The 24-foot moving truck creates a
measurable wake up to a height of approximately 6 meters. The smaller cargo van wake
5-20

8
7
6
Compact
Cargo Van
Box Truck
height (z)
5
4
3
2
1
0
-0.005 0 0.005 0.01 0.015 0.02 0.025
(j iave -j bave ) 1/2 /Vehicle speed
Figure 5-14. Turbulence generated vs. height above ground for three vehicle types. The x-axis
represents the difference between the vertical component of turbulence generated by the passing of
the vehicle and the background fluctuations in vertical velocity normalized by the speed of the
vehicle. The horizontal lines represent the background turbulence standard deviation. The dotted
lines represent hand drawn curves to fit the data. The dotted line corresponding to the compact car
was drawn based on the assumption that the height of the wake plume is proportional to the height of
the vehicle (see Figure 5-15).
Turbulent Wake height (m)
7
6
5
4
3
2
1
0
Compact Car
estimated from
regression
y = 1.7x
Cargo Van
0 0.5 1 1.5 2 2.5 3 3.5 4
Vehicle Height (m)
24-foot Moving
Truck
Figure 5-15. Estimate of turbulent wake height vs. physical vehicle height. The dotted line represents
a zero-intercept regression on the data from the Cargo Van and the Moving Truck. The hollow circle
is an estimate of the wake height for the Compact Car based on the regression.
only goes up to about 2.5 meters. The compact car is apparently too small to cause a
measurable wake, even at the minimum setting for the anemometer height (0.9 meters).
The dotted line drawn for the compact car is based on a linear fit of the wake heights of
the box truck and the cargo van vs. their physical heights (3.2 and 2.0 meters,
respectively). The compact car is 1 meter in height (measured from the top of the trunk
5-21

to the ground) giving an approximate wake height of 1.7 meters according to this
regression (see Figure 5-15).
Turbulence is a surrogate for mixing. The wake height measured in this portion
of the study is important because it gives an indication of the height to which the dust
plume is mixed behind a vehicle traversing an unpaved road. As a first approximation,
we may assume that the dust emitted behind a vehicle is well-mixed up to the height of
the turbulent wake (approximately 1.7 times the height of the vehicle itself). This
“injection” height may play an important role when determining the fraction of
particulate matter that deposits close to the road vs. the fraction that is transportable; as a
rule, the further a particle is from the ground initially, the less likely it is to deposit in a
given period of time. As an aside, the injection height also poses a lower limit on the
height that can be used in Gillette’s model of a hypothetical box.
The approximation for the injection height suggested here is quite crude and
simplistic. The height of the wake is likely to depend on a number of parameters, most
notably, the atmospheric stability, the shape of the vehicle, and the angle of the ambient
wind with respect to the direction of vehicle travel. The analysis is based on data
obtained under stable conditions. Under unstable conditions, turbulent mixing is not
inhibited by buoyancy forces. Therefore, it is likely that the effective injection height will
be larger than under stable conditions. For the present purpose, it is sufficient to note that
the height to which a dust plume is thoroughly mixed in the wake of a vehicle is
dependent on the size of the vehicle, increasing in a roughly linear relationship with
height.
Detailed discussion of vehicle wake characteristics is outside the scope of this
project. Hosker (1982) provides a thorough literature review of flows around bluff
bodies. More recently, Rao et al. (2002) have measured wake properties with a trailer that
was drawn by a cargo van and equipped with multiple sonic anemometers. Hider et al.
(1997) have examined a modeling approach for particles dispersing in a vehicle wake.
Though preliminary and based only on a single vehicle type, their data suggest that the
turbulent kinetic energy in the wake of the vehicle is greatest at approximately the vehicle
height and approaches background levels at twice the vehicle height. This is qualitatively
consistent with our assumption of a wake plume that is completely mixed from the
ground up to 1.7 times the vehicle height.
5.2.3.2 The effect of injection height on removal of particles downwind of an
unpaved road
The injection height is least important under unstable atmospheric conditions, and
most important under stable atmospheric conditions. This is because under unstable
conditions, the dust plume is lofted up high quickly anyway, more or less regardless of
the starting height. Under stable conditions, the extent of vertical mixing is retarded by
buoyancy, and a lower injection height allows the particles to be closer to the ground for
a longer period of time, thereby enhancing deposition.
According to the ADE model, overall, the effect of injection height on particle
removal is quite small. The reason for this was discussed earlier in section 5.2.1, namely
that the dispersion near the source (nominally less than 1,000 m downwind) is greatly
5-22

overestimated by the ADE since some of the basic model assumptions are violated in that
region.
The ISC model more accurately reflects the degree of dispersion that occurs in the
first 1,000 or so meters downwind of the unpaved road source. Recall that the injection
height of the dust plume can be accounted for in the model through the initial dispersion
parameter σ z , assumed to equal one-half of the injection height, and a “virtual” upwind
distance (see section 4.1.2). Comparison of downwind removal rates (Figure 5-16)
indicates that for unstable conditions, the injection height has a non-trivial, but small
effect. For very stable conditions, the injection height has a large effect, approaching a
factor of two difference in estimated removal rates at 1,000 m downwind of the source.
The difference between the removal of particles for two plumes with different injection
heights is confined to the near-source region. That is, the difference does not continue to
grow past about 1,000 meters downwind (Figure 5-16). With this in mind, and noting as
before that most unpaved road dust emissions are likely to occur in the daytime, when
conditions are neutral to unstable, the effect of the injection height is secondary to the
uncertainties associated with deposition velocities and dispersion parameters.
1.2
1.2
Fraction of particles remaining in
suspension
1
0.8
0.6
0.4
0.2
IH=1
IH=2
IH=4
IH=6
Fraction of particles remaining in
suspension
1
0.8
0.6
0.4
0.2
IH=1 m
IH=2 m
IH=4 m
IH=6 m
0
0
0 200 400 600 800 1000
0 200 400 600 800 1000
Distance Downwind (m)
Distance Downwind (m)
a. Moderately unstable (Class B) b. Neutral (Class D)
1.2
1.2
Fraction of particles remaining in
suspension
1
0.8
0.6
0.4
0.2
IH=1
IH=2
IH=4
IH=6
Fraction of particles remaining in
suspension
1
0.8
0.6
0.4
0.2
IH=1
IH=2
IH=4
IH=6
0
0 200 400 600 800 1000
Distance Downwind (m)
c. Very stable (Class F)
0
0 2000 4000 6000 8000 10000
Distance Downwind (m)
c. Same as c. except x-axis extends to 10
km
Figure 5-16. Comparison of fraction of particles removed according to the ISC3 model for various
values of the injection height under a. moderately unstable conditions; b. neutral conditions; c. very
stable conditions; and d. same as c. but x-axis extends to 10 km. The injection height is manifested in
the model by setting the initial value of σ z to 0.5×IH.
Another aspect of the injection height that is not examined here is its effect on the
removal efficiency for particles in the “impact” region. This region (labeled Region A in
Figure 2-1) may play an important role in the removal of particles, especially if the
injection height is lower than or comparable to the vegetative canopy height. None of the
5-23

models analyzed in this study is suited for estimating the interactions of particles with
obstructions in the immediate vicinity of the source (i.e. at the junction of the unpaved
road shoulder and the first row of vegetation downwind of the road). For estimating
particle deposition in this region, it is preferable to use a technique similar to the one
presented by Raupach and Leys (1999), who have developed a method to estimate the
removal of particles by windbreaks. Their analysis is applicable to windbreaks of finite
depth, but it may be possible to extend it to entire canopies with some modifications.
5.3 Summary
Field experiments at the Ft, Bliss military facility indicated that for particles
covered by the PM 10 size range, removal by deposition was not measurable at a distance
of 100 m. This was confirmed by comparison with predictions by the ISC and the ADE
models. Based on these findings, it was concluded that under neutral to unstable
conditions in sparsely vegetated arid shrubland, which may comprise a large portion of
the southwestern United States, removal of particles in the near source region by
deposition is not an important consideration.
The ADE model was found to over predict dispersion very close to the source (<
100 m), but the ISC concentration profiles fit the data reasonably well. One shortcoming
of the ISC algorithm is the lack of ability to represent varying friction velocities.
The turbulence behind moving vehicles was measured to quantify the height of
the wake, termed the “injection height”. Using very simple analysis, it was determined
that the size of the wake varies roughly linearly, by a multiple of 1.7, with the height of
the vehicle. The injection height was found to have a large effect on the removal of
particles downwind of an unpaved road, though only for stable atmospheric conditions.
Comparison of emission factors measured at Ft. Bliss and those calculated
according to the AP-42 guidance document (USEPA, 1999) indicated that by not
accounting for vehicle speed, the AP-42 estimates are likely to be too high for vehicles
traveling at low speeds. When only passenger-sized vehicles were considered, the Ft.
Bliss data suggest that at 20 to 25 mph (a speed that is commonly assumed for unpaved
road travel in emission inventories), AP-42 over predicts PM 10 emissions by 50%-100%
compared to the measurements. When larger vehicles were included in the analysis, the
AP-42 factors were in line with those measured for travel speeds of 20 to 25 mph. One
reason cited for this apparent dependence on the distribution of vehicle size was the use
of an aggregate-average weight in the AP-42 equation. The “lumped” weight approach is
not self-consistent and is likely to result in inventories that overestimate the contribution
of unpaved road dust to ambient PM 10 concentrations. This source of error should be
addressed as part of the effort to reconcile emission inventories for geologically-derived
PM 10 with the amount of crustal material found on filters measuring ambient
concentrations.
5-24

6. DETERMINATION OF PARTICLE REMOVAL RATES UNDER
STABLE CONDITIONS IN A MOCK URBAN SETTING
As part of a DoD-funded study, the University of Utah measured vehiclegenerated
road dust under stable atmospheric conditions at a flat site in the Utah west
desert, and the horizontal flux of dust with the aid of interpolating functions that were fit
to the wind speed and dust concentration measurements at discrete heights. The
measured horizontal flux of dust and particle removal by deposition were compared to
predictions by the one-dimensional Atmospheric Diffusion Equation (ADE) and the
Industrial Source Complex (ISC) models.
6.1 Experimental Methods
The field measurements were conducted at the Dugway Proving Ground, Tooele
County, Utah in collaboration with the Mock Urban Setting Test (MUST), an
atmospheric dispersion experiment funded by the Defense Threat Reduction Agency
(Biltoft, 2001; Yee and Biltoft, 2002). The site is located at 40° 12.6' N, 113° 10.6' W
and is 1310 m above sea level. The configuration, shown schematically in Figure 6-1,
consisted of a 10 by 12 array of 2.5 m high by 2.4 m wide by 12.2 m long rectangular
cargo shipping containers simulating buildings. The area surrounding the test site has a
slope of 0.5 m / km rising to the south. Sand dunes 4-6 m high are 1 km to the north.
Vegetation is a thin cover of brush (shadscale and black greasewood) 0.5 to 1 m high.
The soil is classified by the Natural Resources Conservation Service (2000) as Skumpah
silt loam and has 15% silt content. Screen analysis of a sample from the test road is
provided in Table 6-1.
180 m
9 8 7 6 5 4 3 2 1 0
A
B
C
D
176 m
4
E
F
G
H
I
J
95 m
30 m
1
2 3
Y
Unpaved Test Road
X
True
North
K
3 m
L
Figure 6-1. Plan view of the MUST site showing the array of shipping containers representing
buildings and the location of the measurements. (1) 2-D sonic anemometers at 4, 8, 16 m; (2)
DustTraks at 0.9, 1.7, and 3.7 m; (3) 3-D sonic anemometer at 1.6 m; (4) DustTraks at 1.8, 4.6, 9.1,
and 18.3 m and 2-D sonic anemometers at 4, 8, 16, 32 m.
6-1

Table 6-1. Screen analysis of test road surface material. Average of two soil samples.
Cumulative
+4 mesh 1.00
-4 mesh 0.88
-18 mesh 0.81
-30 mesh 0.70
-50 mesh 0.52
-100 mesh 0.40
-200 mesh 0.16
Silt Content 15.7%
The vehicle activity was designed to approximate a line source where discrete
puff releases create a uniform cloud in the crosswind (y) direction. A 1994 Ford pickup
truck (3900 kg maximum GVW) was driven on a graded, native soil road running parallel
to the upwind edge of the container array. At 1-1.5 minute intervals, the vehicle was
driven at 9 m s -1 (20 mph) along a test section of road that extended beyond the end of
the container array in each direction.
Dust concentration was measured using seven portable DustTrak analyzers (TSI,
Inc. St Paul, MN) with PM 10 inlets. The DustTrak uses a 90° light scattering laser diode
sensor and has a range of 0.001 - 100 mg/m 3 . Reported mass concentration was based on
the factory calibration. Flow rate, zero on filtered air, and the instrument clock were
checked daily.
Instrument locations are shown in Figure 6-1. The near-source dust concentration
measurement consisted of DustTraks at 0.9, 1.7, and 3.7 m above grade located 3 m from
the edge of the road or about 4.5 m from centerline of vehicle travel. Two DustTraks
were mounted on a stepladder placed in the open aisle between containers and the highest
DustTrak was offset parallel to the road and located 1.2 m above the top of the first row
of shipping containers. The downwind dust concentration was measured at a 32 m high
tower located at the center of the container array. The DustTraks were 95 m from the
centerline of vehicle travel at 1.8, 4.6, 9.1, and 18.3 m above grade. The dust
concentration versus time data were treated as a series of 44 measurements of the puff
generated by a discrete vehicle trip using both the peak concentration value for each trip
(mg m -3 ) and the time integrated area under the concentration curve (mg sec m -3 per
trip).
Sonic anemometers provided time-resolved wind data upwind of the container
array, between the test road and the first container, and in the center of the array.
Ambient conditions were monitored at a permanent meteorological station (DPG08)
located about 1 km from the test site.
6.2 Data Analysis Methods
The horizontal flux of dust is the product of the dust concentration times the wind
speed integrated from ground level to the top of the dust cloud. For a line source, the
6-2

units are particulate matter mass per length of road per vehicle trip. Defining the
coordinate axes as x perpendicular to the road, y parallel to the road and z vertical, and
assuming dust concentration and wind profile do not vary parallel to the road, the mass
flux per length of road passing through a plane at constant distance from the road is:
F dust =
z=∞
t=tmax
z=0 t=0
C(x, z,t) u(x, z,t) dt dz
Equation 6-1
where C is the dust concentration (mg/m 3 ), u (m s -1 ) is the wind component
perpendicular to the y, z plane, and tmax is the trip interval or other averaging time.
The dust flux calculation approach used in this portion of the study was to fit
plausible interpolation functions to the measurements and then numerically integrate
using the interpolation function values evaluated at each height step. This algorithm has
the advantage of being mathematically well defined and providing insight into the
sensitivity of the results to wind speed and concentration vertical profiles.
6.2.1 Wind Speed Models
An empirical power law equation can give a reasonable approximation to the
wind profile over a limited range of height. This gives
u = u ref
z
z ref
P
Equation 6-2
where u ref is the wind speed measured at height z ref . The power law interpolation
function has only one adjustable parameter, extrapolates to zero velocity at ground level,
and is mathematically convenient.
6.2.2 Vertical Dust Concentration Models
A first-order exponential decay model was used for dust concentration
C(z) = A exp( −Bz)
Equation 6-3
where A and B are empirical fitting parameters. An exponential equation for dust
concentration was derived by Goosens (1985) by assuming that the change in eddy
diffusivity with height is a power law function. Coefficients were obtained by a leastsquares
fit to the time-averaged field data.
Stepwise expressions for wind speed and concentration were compared to the
interpolation function method. For the stepwise method, the value of the bottom
measurement was used for the interval from ground level to the midpoint between the
bottom and second measurements where there was a discontinuous change to the new
6-3

value which was used up to the midpoint between the second and third measurement, and
so on.
6.2.3 Calculation of Horizontal Flux
The power law wind profile is appealing since analytical solutions exist for the
integral of a power law for wind with either an exponential or a Gaussian function for
dust concentration. For the exponential dust concentration case:
F dust =
C(z) u(z) dz =
∞
( Ae − Bz)
z
u ref
0
z ref
P
− P Γ( P +1)
dz = Au ref z ref
B ( P+1)
Equation 6-4
where Γ is the gamma function. Using a mathematical model that results in a converging
definite integral avoids any uncertainty about the upper limit of integration, z max . The
integral of a power law for wind, times a power law for dust concentration has an analytic
solution, but does not converge. For this study, the horizontal dust flux was also
calculated by trapezoidal rule numerical integration of Equation 6-1 using alternative
interpolation equations to model the time averaged wind speed and concentration as a
function of height. Variable step size (starting at ∆z = 0.1 m) was used to provide higher
resolution near the ground where both concentration and wind speed change rapidly with
height.
6.3 Results
The data were collected during two vehicle activity periods between 01:00 and
02:30 Mountain Daylight Time on September 26, 2001. The clear sky, nighttime, desert
conditions resulted in a stable atmosphere during the sampling period. Wind speed was
generally between 1 and 5 m s -1 , which was sufficient to move the dust cloud, and the
wind direction relative remained within 45° of perpendicular to the container array and
road. General meteorological data are summarized in Table 6-2. Table 6-3 gives the
wind speed and dust concentration mean and standard deviation.
Table 6-2 Meteorological data from station DPG08, approximately 1 km south of test site, for
September 26, 2002 01:00-03:00 MDT.
Average Range
Temperature 294 K 293 - 297 K
Relative Humidity 23% 20-26%
Barometric Pressure
866.54 mbar
24 hr- Precipitation 0
Wind at 10 m 3.0 m s -1 15 m s -1 gust
Wind Direction from True North 98-186°
6-4

Table 6-3. Mean ± one standard deviation wind and concentration data for the entire 1.5 hr test
period with 44 vehicle trips. Wind standard deviation based on variation between 15-minute
averages. Concentration standard deviation based on variation between individual vehicle trips.
Pulse area is the time-integrated concentration for the interval corresponding to one vehicle pass.
Wind
Height
m
Upwind
m s -1
4 3.42 ± 0.51 2.22 ± 0.20
8 4.07 ± 0.50 3.44 ± 0.32
16 4.72 ± 0.52 4.59 ± 0.25
32 4.89 ± 0.31
Concentration
Near Measurement
3 m from road
Height
m
Pulse Area
mg s m -3 per trip
32 m Tower
m s -1
Peak Height
mg m -3
Far Measurement
95 m from road
Pulse Area Peak Height
mg s m -3 per trip mg m -3
0.9 302 ± 171 38.9 ± 21
1.7 144 ± 92 19.9 ± 11
1.8 24.3 ± 17 0.72 ± 0.52
3.7 77.8 ± 56 10.3 ± 7.2
4.6 10.7 ± 7.2 0.41 ± 0.31
9.1 3.18 ± 3.1 0.21 ± 0.17
18.3 0.35 ± 0.1 0.019 ± 0.016
6.3.1 Wind Profile
Figure 6-2 shows the wind speed versus height with the logarithmic and power
law interpolation equations superposed on the 15-minute average data points. Both
equations fit the upwind data within the variation of the measurements. Based on the
field measurements, the fitted coefficients for Equation 6-2 upwind and at the center of
the array are:
Upwind
Center of Array
U ref at z ref, m/s 3.42 2.22
Z ref, m 4 4
P 0.234 0.414
6.3.2 Dust Concentration
Typical dust concentrations versus time data are shown in Figure 6-3.
Measurement locations are identified by the x, z coordinates in meters where x is
downwind from the road and z is height above grade. Near-source dust clouds from
individual vehicle trips appear as clearly defined spikes in the particle concentration
6-5

measured at 3,1. Downwind the dust clouds were lower in magnitude and had longer
duration. It was still possible to resolve peaks corresponding to the individual vehicle
trips as shown by the time trend for the 95, 4.6 monitor. The trend for the highest
measurement, 18.3 m above grade, shows that this instrument was above the top of the
dust cloud for most trips. A floodlight allowed observation of the dust, which confirmed
that the visible cloud was less than half the height of the 32 m tower. The background
dust concentration was approximately 10µg m -3 during periods with no vehicle activity.
Based on field measurements, the fitted coefficients Equation 6-3 near the road and at the
center of the array are:
Near (3 m)
Center of Array
A, mg/m3 594.9 41.15
B 0.725 0.289
Upwind
Array Center
35
35
30
30
25
25
20
20
15
15
10
10
5
5
0
0 2 4 6 8
m/s
0
0 2 4 6 8
m/s
Figure 6-2. Wind Equation Fit. Both the logarithmic (dashed line) and power law (solid line)
equations give a reasonable fit to the 15-minute average wind speed measurements (circles).
6.3.3 Horizontal Dust Flux
The horizontal dust flux decreased to less than 15% of the near-source value as
the cloud passed over and through the container array. The horizontal dust flux passing
through vertical planes 3 and 95 m from the vehicle activity was calculated using the
power law interpolation function for the wind speed and the exponential equation for the
dust concentration. The analytical integration of the flux, Equation 6-4, provides the
following:
6-6

Near (3 m) Center of Array Ratio
Horizontal dust flux, mg/m trip 1990 261 13.6%
Numerical integration of the flux using alternative models for wind speed, (power
law, logarithmic, and step), and for dust concentration, (exponential, power law, and
step), gave similar results. The ratio of downwind to upwind dust flux ranged from 9% to
15% for the different cases. The numerical integration shows that most of the dust flux is
below 5 m. The observation of a significant decrease in the flux ratio at 3 and 95 m
appears robust, but the calculated absolute values vary between the different assumptions
of how the wind speed and dust concentration vary near the ground. The calculated nearroad
dust flux ranges from 1.1 to 4.4 g per vehicle meter traveled which can be compared
to 1.8 g per vehicle meter traveled obtained using the correlation for silt fraction and
vehicle weight from AP-42.
x = 95, z = 18.3
0.1
mg / m 3
0.075
0.05
1
0.025
0
2:14 2:16 2:18 2:20 2:22 2:24
2
x = 95, z = 4.6
mg / m 3
1.5
3
1
4
0.5 1 2
5 6
7
0
2:14 2:16 2:18 2:20 2:22 2:24
x = 3, z = 1
mg / m 3
100
75
50
25
1
2
3
0
2:14 2:16 2:18 2:20 2:22 2:24
Local Time
4
5
6
7
Figure 6-3. Dust concentration versus time as measured by DustTraks. Vehicle passes generated
well-defined spikes at 3 m horizontal and 1 m vertical from the road (bottom). The spikes were
broader but still well defined 95 m downwind and 4.6 m above grade (middle). Most vehicle trips did
not cause a noticeable spike at 95 m downwind and 18.3 m above grade (top). Note that full scale for
the top graph and middle graph are 1/1000 and 1/50 of full scale for the near road measurements
respectively.
Figure 6-3 shows large variations in peak dust cloud concentrations even though the
upwind conditions and vehicle travel were relatively constant. A large part of this
variation is due to non-uniformity in the vehicle dust cloud.
Figure 6-4 shows the results of a separate experiment where 5 DustTraks were
collocated parallel to a road (Seshadri, 2002). Over the entire experiment, 18 vehicle
6-7

trips, the average readings for the five instruments differed by only 0.06 mg/m 3 which
was not statistically significant (P>0.77), and which is small compared to peak readings
of 20-50 mg/m 3 . The time trend shows differences in peak heights and in the arrival time
of the peak. The normalized range for sets of simultaneous readings, defined as
(maximum - minimum)/average, had a mean of 0.96 with a maximum of 3.5, as shown in
the distribution plot. This type of short duration random variation within the dust cloud is
consistent with turbulent mixing.
3.5
3.0
2.5
Range
Average
0.7
0.6
0.5
2.0
1.5
1.0
0.5
0.0
mg/ m 3
0.2
0.1
0.0
14:09 14:10 14:11
Figure 6-4 Collocated DustTraks. Five collocated DustTraks showed large differences in individual
readings but little difference between the instruments when averaged over the entire experiment.
Left: Histogram of maximum minus minimum divided by the average for sets of five coincident
readings. Middle: outlier box plot of the same data. Right: a typical time trend showing poor
correlation between individual readings.
6.4 Discussion
This study supports the hypothesis proposed by Watson, Chow et al. (2000) that
not all dust initially suspended by vehicles is transported far downwind. The
measurements reported in this study were made under stable atmospheric conditions at a
flat site with large solid roughness elements between the test road and the downwind
measurement location. Atmospheric stability and the terrain irregularities have not been
variables in most studies of unpaved road dust. Additional field studies under a range of
atmospheric and terrain conditions will be needed to determine when deposition is
significant and to develop models for predicting the magnitude.
More accurate measurements of dust flux will require better characterization of
dust concentration and wind speed near the ground. The analytical interpolation
equations and the empirical results of this study can be used to determine the optimum
locations of wind speed and dust concentration instruments for future fieldwork.
6.4.1 Comparison of measurements with model results
The experiment at the Dugway Proving Ground, UT, was performed late at night
when the atmosphere was stable. The wind speed varied between 1 and 5 m/s, resulting
in an average friction velocity of 0.23 m/s (z 0 =0.1 m) upwind of the container array, and
0.39 m/s (z 0 =0.71 m) at the center of the container array. This situation presents a unique
set of circumstances where although the atmosphere is stable to very stable, a high
6-8

friction velocity is made possible by the abrupt change in land use from the desert shrub
surroundings to an array of cargo containers. Under these conditions, high stability
concurrent with fairly high friction velocity (therefore also deposition velocity), the
removal rate for particles is expected to be significant, even over the short distance of 95
m between the unpaved road and the instrumented tower.
Figure 6-5 compares the measured PM 10 removal rate from the Dugway
experiments with the estimated removal rate from the ISC and the ADE models. The
PM 10 DustTrak instruments used in the field do not provide an estimate of the particle
size distribution. Therefore, removal rates based on measurements with those
instruments are applicable to the entire fraction of particles with diameters less than about
10 microns. Since the models use specific deposition velocities for each of the particle
sizes, comparison of measured and modeled values requires some assumptions about the
particle size distribution during the Dugway experiments. For the purposes of
illustration, the average particle size distributions obtained during the Ft. Bliss
experiments (Chapter 5) was used. The white circles in Figure 6-5 represent the fraction
of the total PM 10 mass that is associated with particles in the stated size bin (sum is equal
to unity). The gray circle is the measured PM 10 removal rate 95 meters downwind of the
unpaved road. The black symbols correspond to the removal rates for the stated particle
sizes estimated from the ISC and ADE models. The modeled overall removal rate for
PM 10 (“Total” in the figure) was calculated by multiplying the fraction of the mass
associated with each size by the removal rate predicted for that size and then summing
the values for all four sizes shown.
The previously noted trend of higher removal rates predicted by the ISC than the
ADE is readily apparent in the figure. According to the ISC, 56% of the 9.9 micron
particles and 22% of 7.2 micron particles deposit within the first 95 meters of the road.
Removal rates for smaller particle sizes are negligible. The model predicts that 30% is
deposited over the same distance when the entire PM 10 size fraction is considered. While
significant, and clearly larger than the removal rates for the neutral to unstable conditions
at Ft. bliss, the modeled removal of PM 10 falls short of the measured values (87%
removal) by a considerable amount. The reason for this disparity is not known. The large
roughness elements in the Dugway container array were intended to simulate an urban
area. Perhaps the parameterization of the deposition velocity (Equation 2-26) used for
the case of vegetative canopies does not readily extend to other land uses such as urban
settings. In particular, the air flow around bluff, impermeable bodies may deviate
significantly from what occurs in a comparatively porous vegetative canopy. Related to
this point, it is possible that the dispersion parameter (σ z in the case of the ISC) does not
adequately represent the channeling of the airflow through the spaces between the
containers or the stagnation regions in their wake. Despite the models substantial under
prediction of PM 10 removal, some solace can be found in noting that most unpaved roads
are not located in urban areas and that it is likely that the majority of traffic on unpaved
roads occurs during the daytime when the atmosphere is either neutral or unstable.
6-9

Fraction of particles of stated size remaining in
suspension 95 m downwind of unpaved road
1.2
1
0.8
0.6
0.4
0.2
0
ISC prediction (u*=0.4 m/s, very stable)
ADE prediction (u*=0.4 m/s, very stable)
Mass Fraction Associated with Particle Size (based
on Ft. Bliss size distribution)
Measured suspended fraction remaining
3.9 5.6 7.2 9.9 Total
Aerodynamic Diameter (microns)
Figure 6-5. Comparison of fraction of PM 10 particles remaining in suspension 95 m downwind of an
unpaved road during a nighttime test on Dugway Proving Grounds, UT. The white circles represent
the fraction of PM 10 associated with each size bin. Mass fractions associated with each size bin were
estimated from Ft. Bliss data from a different experiment. The gray circle is the fraction remaining
in suspension according to the methods described in the previous section. The black squares and
triangles correspond to modeled removal rates under very stable conditions.
6-10

7. SUMMARY AND CONCLUSIONS
Six major conclusions were formed from the models and field data examined in
this report. We summarize these conclusions below and discuss their ramifications with
respect to recent and ongoing work in the next subsection.
First, the hypothesis that fugitive dust particles emitted from unpaved roads can
deposit to an appreciable extent within several hundred meters downwind of the road
(Watson and Chow , 2000; Countess, 2001) is well-founded, though the fraction of
particles removed varies greatly based on setting and atmospheric stability. Two recent
studies with robust datasets were examined. The field study at Ft. Bliss provided
information on dust particle removal in a sparsely vegetated terrain, under neutral to
unstable atmospheric conditions. The experiments at the Dugway Proving Ground in
Utah occurred under nighttime stable conditions with large roughness elements
downwind of the unpaved road. These two studies provide upper and lower bounds for
the transportable fraction of PM 10 dust emissions.
A field study set at the Ft. Bliss military installation in El Paso, TX took place in
April 2002. An unpaved road, located in a desert open range, sparsely peppered with
small shrubs, was used to conduct a series of emissions tests. Three towers, 9 to 12 m in
height, were placed downwind of the road at distances of 7, 50, and 100 m. A variety of
vehicles were driven back and forth along the road. DustTrak particle monitors located at
several heights along the towers were used to capture the PM 10 concentrations due to the
passing dust plumes. Wind speed and direction were recorded at the far downwind tower
(100 m). The horizontal flux of PM 10 particles was calculated for each of the three
towers. No measurable differences were found between the horizontal flux at the road
and the flux 100 m downwind. This indicated that the removal of PM 10 by deposition
was not occurring to an appreciable extent. Modeling studies based on individual particle
sizes confirmed this finding and further suggested that even at 1,000 m downwind, the
fraction of particles removed is less than 25% at high wind speeds (u * > 0.5 m/s). At
lower, more common wind speeds (u * < 0.5 m/s), the fraction of particles removed at
1,000 m was less than 20% for neutral conditions and less than 10% for very unstable
conditions.
Vehicle generated fugitive dust transport was examined in a field study at the
Dugway Proving Ground, UT. A 10 by 12 array of 2.5 m high by 2.4 m wide by 12.2 m
long cargo shipping containers was used to simulate urban terrain. The containers were
arranged with their widest dimension against the prevailing winds and were spaced 14
meters apart. An unpaved road upwind of the container array was traversed with a
vehicle under nighttime stable atmospheric conditions. Two tower arrays, one near the
road and one 95 m downwind were instrumented with DustTrak PM 10 monitors.
Calculation of the horizontal dust flux using a variety of mathematical fits to the wind
and concentration profiles indicated that by the time the plume from the road reached the
95 m tower, approximately 85% of the PM 10 had deposited.
The two field tests described span a large range of conditions with regard to
particle removal. For the Ft. Bliss tests, dispersion of the fugitive dust plume was
efficient because of the neutral to unstable atmosphere. In contrast, the stable atmosphere
during the Dugway tests retarded near-source dispersion, keeping the dust plume close to
7-1

the ground, where particles have a better chance to deposit. In addition, the areas of
stagnation on both the windward and leeward sides of the containers may have enhanced
the removal rate of PM 10 particles.
The conditions during the Ft. Bliss test may be more true to everyday reality than
the Dugway tests. This assertion is supported by two observations. 1.) Unpaved road
dust is more likely to be an air quality concern in arid, open-range regions characterized
by much of the American Southwest, and 2.) It is likely that the majority of travel on
unpaved roads occurs during the day when the atmosphere is either neutral or unstable.
With these observations in mind, it seems that in areas were unpaved road dust is of
greatest concern, the removal of particles by deposition is smallest in magnitude. This
goes against a recent movement to divide fugitive dust emissions by a factor of 2-4 in
order to account for deposition within several hundred meters of the road. We note
however that the conditions at Ft. Bliss are not representative of every area in the United
States that is concerned with road dust. There may be some instances where there is
thick vegetative cover very close to an unpaved road (e.g. farms). This possibility is
revisited later in the text.
Second, for the purposes of modeling the transport and removal of dust particles
near an unpaved road source, the Gaussian-style models, such as EPA’s ISC3, provide an
imperfect, but reasonable preliminary approach. Dispersion and deposition are the two
main processes that must be accounted for in any model used. The ISC model was
compared to the one-dimensional Atmospheric Diffusion Equation (ADE) and to data
obtained at Ft. Bliss. The ADE relies on similarity theory for the parameterization of
dispersion. Based on vertical concentration profiles measured as part of the Ft. Bliss
experiments, it was determined that ADE over predicts dispersion in the region very close
to the source (i.e. 500 m downwind or less) and that the ADE is not suited for modeling
over short downwind distances. The ISC model uses a distance-based dispersion
parameter that depends on atmospheric stability (σ). Comparison of predicted
concentration profiles with those observed at Ft. Bliss indicated that the ISC captures the
approximate shape of the dust plume. However, whereas the measured profiles varied in
steepness at different values of the friction velocity u * , the ISC profile was invariant.
This is because the friction velocity is not included in the parameterization of σ.
Several different models for dry deposition were compared. They produced
deposition velocities that were in agreement within a factor of two or so. The
formulation for deposition velocity in the ISC model has the advantage that it only
requires knowledge of three parameters: The friction velocity u * , the roughness height z 0 ,
and the Monin-Obukhov length L. U * and L may be estimated from commonly-measured
meteorological parameters. Z 0 can be inferred from existing land use databases (e.g.
EPA’s BELD database). In contrast, more complex formulations, including the widely
cited Slinn (1982) model, require many parameters as input, some of which are difficult
to obtain.
It is difficult to determine how well models for deposition velocity reflect realworld
values. This is largely because these models have not been widely-tested with
field data. In most cases, adjustable parameters are made to fit Chamberlain’s (1967)
wind tunnel tests for deposition on grass surfaces. As part of this study, the deposition of
particles to two different types of surrogate surfaces was measured. Comparison with the
7-2

predictions of Slinn’s (1982) model and the one used in the ISC3 showed that both
models tend to under predict the measured deposition by at least a factor of ten. This
discrepancy was attributed, partly, to the difficulty of relating measurable field deposition
velocities to parameterized model results. The accurate measurement and modeling of
particle deposition velocities continues to pose a challenge to scientists, even after four
decades of research. For the purposes of this report, it is sufficient to note that there may
be substantial (factor of two or more) uncertainties in the specification of deposition
velocity for use in a model.
Despite the previously-noted shortcomings, both in the deposition and the
dispersion portions of the ISC model (they are not unique to the ISC), it predicted the
results for the Ft. Bliss tests fairly well. Specifically, the ISC predicted that at 100 m
downwind of the road, there are no measurable reductions in the horizontal flux of dust
particles in the size range covered by PM 10 (Aerodynamic diameter of 10 microns or
less). However, the model suggested that for larger particles, D p =19.7 µm, the fraction
remaining in suspension would be reduced to about 90% over the same distance.
The ISC did not perform as well for the tests at the Dugway Proving Ground.
While the model indicated that the fraction of PM 10 remaining in suspension 95 m
downwind was 70%, the measured values were closer to 15%. It was hypothesized that
the disagreement between the model and the measurements was due to the unusual terrain
consisting of large containers. Perhaps deposition in the stagnation zones around those
containers is not adequately represented in the modeled deposition velocity.
Alternatively, it is possible that the channeling of particles between the containers under
stable conditions is not adequately represented by the ISC dispersion parameter.
The similarity between the concentration profiles measured at Ft. Bliss and those
predicted by the ISC is encouraging. The further agreement between the modeled and
measured values of near-field particle removal support the use of the ISC. The same
cannot be said for the experiments at the Dugway Proving Ground. Nevertheless,
considering the uncertainties associated with deposition velocities and the paucity of field
data, we cannot expect any model to provide better than an approximate estimate of the
transportable fraction. Therefore, in the foreseeable future, the ISC provides a
reasonable, though imperfect approach. An additional consideration is that the ISC has
been standardized by the US EPA, making it facile to disseminate to the air quality
community if so needed.
Third, the Box Model approach is elegant in its simplicity, but the assumptions
behind the model, specifically those relating to dispersion, are probably not valid for
unpaved road dust emissions. Gillette (Countess, 2001) proposed a Box Model to
estimate the transportable fraction of unpaved road dust PM 10 emissions. In the model,
dust particles deposit at the ground and disperse out of the top of the box at a rate that is
proportional to the concentration in the box (assumed to be uniform with height). For
deposition, this is a reasonable assumption, especially since the deposition velocity is not
too variable with height for neutral to unstable conditions (i.e. drawing on the resistance
analogy, r a

particles through the top of the box musts always be in the upwards direction. These
conditions are closely met during a dust storm when the ground is a strong source, and
the vertical concentration profile does not vary over a large distance in the horizontal.
They are only met under neutral to unstable conditions for the case of emissions from
unpaved roads, and even then, only over very short downwind distances.
Fourth, the height of the wake generated behind a moving vehicle has an effect on
the fraction of particles removed as the plume travels downwind, though this effect is
only significant for stable conditions. Preliminary measurements suggested that the
vertical extent of the turbulence induced in the lee of a passing vehicle scales
approximately linearly with vehicle height. According to the ISC model, a larger fraction
of particles is removed as the dust plume travels downwind for small values of the
“injection height” than for larger values. This effect is more pronounced under stable
conditions when the dust plume is not rapidly expanded in the vertical direction. Overall,
in view of the uncertainties in quantifying deposition velocities, the “injection height” is
probably a secondary consideration.
Fifth, measurement of emission factors for seven different vehicles with gross
weights varying from 1,622 kg to 23,636 kg were at odds with the silt-based emission
factors suggested in AP-42 (USEPA). The Ft. Bliss measurements indicated that
emission factors were highly dependent on vehicle speed. Linear regression of emission
factor vs. speed for each of the seven vehicles resulted in R 2 values ranging from 0.37 to
0.95, with six of the seven vehicles giving values greater than 0.70. In contrast, the AP-
42 emission factor formulation does not consider vehicle speed. As a result, the AP-42
tends to over predict PM 10 dust emitted from unpaved roads when the vehicle speed is
less than about 20 to 25 mph. In addition, the use of an “average” vehicle weight in the
AP-42 equation tends to introduce a positive bias for small vehicles. According to the
measurements at Ft. Bliss, for passenger-sized vehicles (large, heavy trucks excluded),
the AP-42 appears to overestimate emissions for vehicle speeds under 35 mph. Emission
inventories often assume an average unpaved road travel speed of 20 or 25 mph. Under
those conditions, use of AP-42 would result in an overestimate of PM 10 emissions by
50% to 100% based on the field study results. This discrepancy, if it can be shown to
hold at locations other than Ft Bliss, would explain some of the incongruencies between
PM 10 dust emission inventories and the amount of crustal material found on filter samples
(Watson and Chow, 2000).
Sixth, prior to applying a correction to existing emissions inventories, it is
important to verify if a substantial amount of PM 10 is removed near the source and if it is
possible that there are other sources of error in the emissions estimates for unpaved road
dust. The conclusions of this report are based, in part, on two field tests that represent the
opposite extremes of atmospheric and land use conditions. In one case, the terrain
consisted of small roughness elements and the atmosphere was unstable. In the other
case, the roughness elements were building-sized and the atmosphere was stably
stratified. It is important to compare model data, ISC or other models, to measurements
performed under a range of conditions. Considering the uncertainties in deposition
velocities and dispersion parameters, model predictions must be taken with a grain of salt
until field data are available for comparison. The two field studies summarized in this
report utilized highly time resolved particle monitors operating at multiple heights. This
7-4

allowed for the comparison of downwind concentration profiles and horizontal fluxes of
PM 10 dust with model predictions. A similar approach is recommended for future studies
with the addition of a LIDAR instrument to quantify dispersion at long downwind
distances.
It is particularly important to obtain data for multiple types of terrains. The ISC
model performed well for sparsely vegetated desert terrain; for a mock urban area, the
model under estimated particle removal by a factor of 2.5. It is unclear from these results
how well the ISC model performs when the downwind fetch is comprised of, for
example, corn crops. Related to this, the models considered in this report only account
for the deposition of particles after the emitted dust plume has passed the first several
rows of vegetative cover. That is, we have not considered what happens at the interface
between the unpaved road and the first row of vegetation encountered downwind of the
road. Classic dispersion models are not suited for representing this “impact region”. The
efficiency of particle removal in this region is unknown, largely due to a paucity of data.
Preliminary work performed in this area (Raupach and Leys, 1999) indicates that the
filtering of particles by windbreaks may be significant. This hypothesis should be
examined experimentally.
Futhermore, at the time of the writing of this report, it is still unclear if the
discrepancy between PM 10 emission inventories and the fraction of PM 10 found in
ambient samples can be attributed solely to the near field deposition of particles. There
are a number of other possible sources of error that may result in inflated emission
inventories (Watson and Chow, 2000; Countess, 2001). One finding of the present study
is that the AP-42 tends to over predict emissions of PM 10 dust for smaller, passengersized
vehicles, especially when the vehicles are traveling at speeds less than 35 mph.
This may frequently be the case on small rural unpaved roads or driveways.
Additionally, the accuracy of the activity levels that are applied to unpaved road dust
inventories is a great source of uncertainty. There is a paucity of traffic counts and
accurate vehicle speed estimates for unpaved roads. This shortage of data makes intuitive
sense, since unpaved roads are unpaved, precisely because they are not used often enough
to warrant attention from community planning associations.
7.1 The findings of this study in relation to recent work
Watson and Chow (2000) documented the discrepancy between emission
inventories for PM 10 fugitive dust and the source attribution of ambient filter samples.
Their analysis indicated that the amount of geologically-derived PM 10 found in the air is
much smaller than would be expected based on the emission inventories and dispersion
models. Countess (2001) summarized eleven shortcomings in the current treatment of
fugitive dust emissions. Though not all eleven findings are directly relevant to the work
in this report, we address several of them below.
The removal of dust in the area near a source has been hypothesized to be
substantial because concentrations of particles downwind of the source rapidly attenuate
to the background. This is an erroneous deduction. Watson and Chow (2000) and
Countess (2001) both cite an earlier study (Watson et al., 1996) where the concentration
of PM 10 dust was found to decrease by 90% only 50 m downwind from the source.
While this may be true, it is incorrect to assume that the decrease is due solely to
7-5

deposition. Much of the decrease in concentration - indeed for the conditions at Ft. Bliss,
most of the decrease - is due to dispersion in the vertical direction. Therefore, using the
magnitude of the PM 10 concentration downwind of a dust source to estimate the removal
of particles will suggest a much higher deposition rate than is achieved in reality. This
does not invalidate the correct assumption that some of the PM 10 emitted will be removed
as the dust plume travels downwind of the source. However, the removal occurs to a
smaller degree and over larger distances than previously speculated, especially for the
open, sparsely vegetated terrain that much of the American southwest is comprised of.
The results of this study support the first finding put forth by Countess (2001),
“Not all suspendable particles are transported long distances”. However, the Ft. Bliss and
Dugway studies show a large range of near-field removal rates. In the rangeland setting
of Ft. Bliss, little removal of suspended dust was observed at a downwind distance of 100
m. This result was supported with a modeling effort that used a Gaussian approach
similar to the ISC model (USEPA, 1995). Nighttime experiments at the Mock Urban
Setting at the Dugway Proving Grounds suggested that under stable atmospheric
conditions and large roughness elements (cargo containers) 85% of the PM 10 may be
removed by deposition 95 m downwind of the unpaved road. This enormous range of
removal rates emphasizes that it is not appropriate to apply a single correction factor to
all fugitive dust emissions as a means of accounting for near-field particle removal.
Though not documented, the community of scientists and professionals have, in the last
several years, been circulating the idea that if fugitive dust emissions were divided by a
factor of four, then the discrepancy between emissions and ambient measurements of
geological PM 10 would disappear. While it is possible that this is true on an average
basis (i.e. over large spatial domains), it is unlikely that this factor of four is applicable to
every combination of airshed, land use distribution, and atmospheric conditions. Each
combination of setting and meteorological conditions should be considered separately in
a modeling framework that makes use of the known physics of particle dispersion and
deposition.
Related to the last point, the removal of PM 10 dust particles in the “impact zone”
immediately after emission, at the junction of the road edge and the first row of
vegetation, should be examined more closely. The work presented in this report only
considers deposition of particles from “above” the vegetative canopy. It does not
consider that particles may be removed to an appreciable extent by a process similar to
filtration when the dust plume is forced through vegetative elements. Cowherd and Pace
(2002) and He et al. (2002) have initiated work in this area. Those authors suggest that
the fraction of particles removed in the impact zone is proportional to the blockage of the
wind by the vegetation. This is applied to canopies exhibiting up to 50% blockage. At
higher blockages, the impinging dust plume is assumed to be diverted over the
vegetation. This conceptual approach warrants pursuit. However, prior to being used to
correct road dust emission inventories, it requires testing against real data and analysis in
the context of particle physics and previous work (e.g. Raupach and Leys, 1999).
As one of the recommendations from his second finding, “Regional-scale vertical
flux is smaller than the local-scale fugitive dust flux”, Countess (2001) suggests testing
the Gillette box model and modifying the model parameters if appropriate. In this study,
the box model was examined, from a theoretical and practical standpoint. The analysis
7-6

indicated that the model is not practical for wide spread use because the parameterization
of dispersion is reliant on assumptions that cannot always be met in reality.
Related to the third finding from Countess (2001), “Fugitive dust emission factors
need to be appropriate”, the emission factors measured at Ft Bliss with high timeresolution
instruments suggest that AP-42 may overestimate PM 10 emissions from
unpaved roads. This overestimate was due to the lack of accounting for vehicle speed
and to the use of the average weight of all vehicles in calculating the emission factors.
All in all, this report concludes that there is no single, “silver bullet” solution to
the problem of reconciling fugitive dust emission inventories and ambient source
contribution estimates. Progress has been achieved in the quantification of the near-field
removal of dust particles and work in that area should continue. There are still
uncertainties about the AP-42 emission factors for unpaved road dust that have resurfaced
in this study and require resolution.
7-7

7-8

8. REFERENCES
Bache, D.H. (1981). Analyzing Particulate Deposition to Plant Canopies. Atmos Env.,
13: 1881 – 1887.
Biltoft, C., 2001. Customer Report for Mock Urban Setting Test, Dugway, UT,
Meteorology & Obscurants Division, West Desert Test Center. Defense Threat
Reduction Agency, Alexandria, VA.
Byun, D.W., and R. Dennis (1995). Design Artifacts in Eulerian Air Quality Models:
Evaluation of the Effects of Layer Thickness and Vertical Profile Correction on
Surface Ozone Concentrations. Atmos Env. 29: 105-126.
CEPA, 1998a. AP-42, Compilation of Air Pollutant Emission Factors - Vol 1. Stationary,
Point, and Area Sources, Chapter 13.2.2 Unpaved Roads, Washington, D.C., U.S.
Environmental Protection Agency.
Chamberlain, A.C. (1967). Tranport of Lycopedium Spores and Other Small Particles to
Rough Surfaces. Proc. Roy. Soc., 208A: 45-70.
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. (1999). Profiling data for open fugitive dust sources. Prepared for U.S.
Environemntal Protection Agency, Emissions Factor and Inventory Group, Office
of Air Quality Planning and Standards, RTP, 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.
Dabberdt, W.F., Lenschow, D.H., Horst, T.W., Zimmerman, P.R., Oncley, S.P., and A.C.
Delany (1993). Atmosphere-Surface Exchange Measurements. Science, 260:
1472-1481.
Dai, W., C.I. Davidson, V. Etyemezian, and M. Zufall (2001). Wind Tunnel Studies of
Particle Transport and Deposition in Turbulent Boundary Flows. Aerosol Science
& Technology, 35:887-898.
Du, S., and A. Venkatram (1997). A Parametrization of Verical Dispersion of Ground-
Level Releases. J. App. Meteorology, 36: 1004-1015.
Etyemezian, V., C.I. Davidson, S. Finger, M. Striegel, N. Barabas, and J. Chow (1997).
Vertical Gradients of Pollutant Concentrations and Deposition Fluxes on a Tall
Limestone Building. J. Am. Inst. Conservation, 37: 187-210.
Etyemezian V., H. Kuhns, J. Gillies, M. Green, M. Pitchford, and J. Watson (2002).
Vehicle based road dust emissions measurements (I): Methods and Calibration.
(submitted).
8-1

Gifford, F.A. (1961). Use of Routine Meteorological Observations for Estimating
Atmospheric Dispersion. Nucl. Safety, 2: 47-51.
Gillette, D. A. (1974). On the production of soil wind erosion aerosols having the
potential for long range transport., J. de Recherches Atmospheriques, 8: 735-744.
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 PM 10
emissions from unpaved roads. JAWMA, 49:3-16.
Gillies, J.A., N. Lancaster, W.G. Nickling, and D. Crawley (2000). Field determination
of drag forces and shear stress partitioning effects for a desert shrub (Sarcobatus
vermiculatus, Greasewood). Journal of Geophysical Research, Atmospheres 105
(D20): 24871-24880.
Goossens, D., 1985. The granulometrical characteristics of a slowly-moving dust cloud,
Earth Surface Processes and Landforms, 10:353-362.
He, S., J. Ching, D. Gillette, W. Benjey, T. Pace, and T. Pierce (2002). 2002 AAAR
annual meeting.
Hider, Z.E., Hibberd, S., and C.J. Baker (1997). Modelling Particulate Dispersion in the
Wake of a Vehicle. Journal of Wind Engineering Vol 67&68 :733-744.
Hosker, R.P. Jr. 1984. Flow and Diffusion near Obstacles. In Atmospheric Science and
Power Production Report for United States Department of Energy. Darryl
Randerson, editor.
Houser, C. A., and W.G. Nickling (2001). The Emission and Vertical Flux of Particulate
Matter

Report prepared for the Great Basin Unified Air Pollution Control District, April
24, 1997.
Peters, K., and R. Eiden (1992). Modelling the Dry Deposition Velocity of Aerosol
Particles to a Spruce Forest. Atmos Env, 26A: 2555-2564.
Pleim, J., A. Venkatram, and R. Yamartino (1984). ADOM/TADAP Model Development
Program. Volume 4. The Dry Deposition Module. Ontario Ministry of the
Environment, Rexdale, Ontario.
Porch, W. and D. A. Gillette (1977). A Comparison of Aerosol and Momentum Mixing
in Dust Storms Using Fast-Response Instruments. J. Appl. Meteor., 16:
1273-1281, 1977.
Rao, K.S., Gunter, R.L., White, J.R., and R.P. Hosker. Turbulence and Dispersion
Modeling near Highways. Atmos Env. 36: 4337-4346.
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.
Schopflocher, T. P., and P.J. Sullivan (2002). A Mixture Model for the PDF of a
Diffusing Scalar in a Turbulent Flow. Atmos Env., 36: 4405-4417.
Sehmel, G.A. (1973). Particle Resuspension from an Asphalt Road Caused by Car and
Truck Traffic. Atm. Env., 7: 291-309.
Seinfeld, J.H., and S.N. Pandis (1998). Atmospheric Chemistry and Physics: From Air
Pollution to Climate Change. Wiley Interscience. New York, USA.
Seshadri, G., 2002. Fugitive Dust Transport from Unpaved Roads in Arid Climates. MS
Thesis, Department of Chemical and Fuels Engineering, University of Utah, Salt
Lake City, Utah.
Shir, C.C. (1973). A preliminary Numerical Study of Atmospheric Turbulent Flows in the
Idealized Planetary Boundary Layer. J. Atmos Sci., 30: 1327-1339.
Slinn, S.A., and W.G.N. Slinn (1980). Prediction for Particle Deposition on Natural
Waters. Atmos Env., 14: 1013-1016.
Slinn, W. G. N. (1982). Predictions for particle depositions to vegetative canopies,
Atmos. Env. 16: 1785-1794.
Turner, D.B. (1970). Workbook of Atmospheric Dispersion Estimates. PHS Publication
No 999-AP-26. U.S. Department of Helath, Education and Welfare, National Air
Pollution Control Administration, Cincinnati, OH.
USEPA, (1995). User’s Guide for the Industrial Source Complex (ISC3) Dispersion
Models Vollume II: Description of Model Algorithms. US Environmental
Protection Agency, Office of Air Quality Planning and Standards: Emissions,
Monitoring, and Analysis Division, RTP, NC.
USEPA (1999). Compilation of Air Pollutant Emission Factors – Vol I, Stationary point
and area sources. Report N0 AP-42, 5 th ed. Prepared by US Environmental
Protection Agency, Research Triangle Park, NC.
8-3

Venkatram, A. (1993). The Parametrization of the Vertical Dispersion of A Scalar in the
Atmospheric Boundary Layer. Atmos Env., 27A: 1963-1966.
Watson, J.G., J.C. Chow, J.A. Gillies, H. Moosmuller, C.F. Rogers, D. Dubois, and J.C.
Derby (1996). Effectiveness demonstration of fugitive dust control methods for
public unpaved roads and unpaved shoulders on paved roads. Report No. 685-
5200.1F. Prepared for San Joaquin Valley Unified Air Pollution Control District,
Fresno, CA, by Desert Research Institute
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.
Wu, Y.L., C.I. Davidson, D.A. Dolske, and S.I. Sherwood. 1992. Dry Deposition of
Atmospheric Contaminants: The Relative Importance of Aerodynamic, Boundary
Layer, and Surface Resistances. Aerosol Science and Technology 16: 65-81.
Yee, E. and Biltoft, C. A., 2002. Concentration fluctuation measurements in a plume
dispersing through a large rectangular array of obstacles, Boundary Layer
Meteorology In Press.
8-4