PB86-23U89
EPA/600/7-86/027
July 1986
FIELD EVALUATION OF WINDSCREENS
AS A. FUGITIVE DUST CONTROL MEASURE
FOR MATERIAL STORAGE PILES
By
Robert A. Zimmer
Kenneth Axetell, Jr.
Thomas C. Ponder, Jr.
PEI Associates, Inc.
14062 Denver West Parkway
Golden, CO 80401
Contract No. 68-02-3995
Task ID Ho. 15
EPA PROJECT. OFFICER: Dale L. Harmon
AIR AND ENERGY ENGINEERING RESEARCH LABORATORY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
Air and Energy Engineering Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
NOTICE
This document has been reviewed in accordance with
U.S. Environmental Protection Agency policy and
approved for publication. Mention of trade names
or commercial products does not constitute endorse-
ment or recommendation for use.
-------
ABSTRACT
EPA completed an in-house study designed to determine changes in
windspeed (not changes in emissions) due to windscreens. A wind tunnel
was used to determine the optimal windscreen porosity, size, and location
for control of fugitive dust emissions from storage piles. Before this
information could be, applied to the design of windscreens, it wai> necesaary
to conduct the- field study described in this report to validate the wind
tunnel studies with respect to windspeed changes, and to determine .".he
relationship between changes in windapeed and changes in fugitive t'ust
emissions. The field study suggests that the optimum windscreen design
parameters are porosity - 50 percent; height • l.OU; width » 5.0D; and
distance - 2.OH for a conical pile of height H and diameter D. Analysis
of the field data shows that emission rates were directly related to
windspeed and inversely related to moisture content of the pile surface.
These relationships held regardless of the particle size fraction consid-
ered.
-------
Figures
CONTENTS
Paoe
1.0 Introduction end Summary 1-1
1.1 Ir:troduct ion 1-1
1.2 Summary 1-2
2.0 Previous Studies 2-1
2.1 Wind.-oeed and Particle Uptake 2-1
2.2 Windscreens as a Fugitive Dust Control Measure for
Storage Piles 2-2
2.3 The Bilhnan Study 2-5
30 Field Sampling 3-1
3.1 Study Design 3-1
3.2 Field Sampling Program 3-6
4.0 Quality Assurance ' 4-1
4.1 Quality Assurance Plan 4-1
<.2 Sampling Procedure for Critical Measurements 4-11
4.3 Equipment Calibration 4-18
5.0 Objective 1— Verification of Wind Tunnel Wind Speed Data 5-1
5.1 Windspeed Comparisons Between Wind Tunnel and Field
Testing for an Unscroeneri Pile . 5-1
5.2 Windspeed Control Effectiveness 5-8
6.0 Objective 2- -Comparison of >/indspeed Reductions and
Participate Control Eff iciercies 6-1
6.1 RAM-! Particulate Data 6-1
6.2 Exposure Profiler O.»ta • 6-5
6.3 Factors Other than Windscreen Affecting Emission Rates 6-14
7'.0 Objective 3- -Development §jf Windscreen Design Parameters 7-1
7.1 Screen Height, Length, and Distance from Pile 7-1
7.2 Wind Direction Relative to Screen 7-9
References R-l
Appendix A A-l
-------
FIGURES
Page
Win-: velocity pattern above a mown field during a
17 m/sec wind blowing at right angles to a 4.9 m
high wood fence 122 m long of 50% porosity • 2-3
2-2 U/U about conical pile for'no w'r.dbreak case 2-1
2-3 Windspeed reduction factor for the 65% porous wind-.
break of height 0.5H and length l.OD placed 1H from
the conical pile base. 2-9
3-1 Test site 3-9
3-2 Test plot layout ' 3-10
3-3 Wind sensor placement, uncontrolled pile 3-13
3-4 View of sampling array from upwind of pile 3-17
5-1 Composite u/u values for an unscreened pile, field
testing compared to wind tunnel data 5-4
5-2 Composite u/u values for an unscreened pile-field
testing and wind tunnel data 5-6
5-3 Windscreen control efficiencies for position 1H by
screen length and screen height 5-17
5-4 Windscreen control efficiencies for position 2H by
screen length and screen height 5-18
5-5 Windscreen control efficiencies for position 3H by
screen length and screen height b-19
6-1 Windspeed reduction versus particulate reduction 6-7
6-2 Scatter plot of emission rate vs. Windspeed 6-17
6-3 Scatter plot of emission rate vs. moisture content 6-13
6-4 Scatter plot of emission rate vs. silt content 6-19
7-1 Evaluation of screen height 7-5
7-2 Evaluation .of screen length . 7-6
7-3 Evaluation of screen-to-pile distance 7-7
-------
Figures continued
7-4 Scatter plot of participate- emission reduction vs.
wind direction deviation from perpendicular 7-10
7-3 Scatter plot of participate reduction vs. wind
direction by screen length 7-12
TABLES
Page
Windscreen impact for various windbreak cases—Billman
study (percent reduction) . 2-10
3-1 Variables to be tested 3-7
4-1 Precision, accuracy and completeness objectives 4-1
4-2 Precision, accuracy and completeness objectives 4-1
4-3 Screened pile windspeed sensor precision test 4-4
4-4 Exposed pile windspeed sensor precision test 4-5
4-5 Screened pile windspeed sensor precision test 4-6
4-6 Exposed pile RAM-1 instrument precision test 4-7
4-7 Exposi-re profiler sampler audit results
July 29, 1985 4-8
4-8 Exposure profiler sampler audit results
August 28, 1985 . 4-9
4-9 Instrument flow rate audit results 4-10
4-10 Summary of chart recorder audits of computer data
inputs 4-13
3-2 Precision, accuracy and completeness objectives 3-20
5-1 U/U values for the unscreened pile with winds
175-184° 5-3
5-2 U/U values for the unscreened pile 105-264° wind
directions 5-5
5-3 Screen Effectiveness, (1- u/u ) 5-9
VII
-------
Tables continued
5-4 Maximum measured windscreen reduction factor
[1 - (u/u )] - screen efficiency 5-13
6-1 Evaluation of average windspeed reduction and
particulate reduction (RAM-1) by test 6-3
6-2 Upwind RAM-1 concentrations • 6-4
6-3 Comparison of windspeed and total particulate reduc-
tions • 6-6
6-4 Calculation of emission rates by particle size range 6-11
6-5 Impact of windspeed reductions on particle size' emis-
sion reductions 6-13
6-6 MLR runs to identify variables that affect emission
rates 6-15
6-7 MLR runs with emission data by particle size range 6-22
7-1 Stepwise MLR to evaluate windscreen design parameters 7-3
vm
-------
1.0 INTRODUCTION AND SUMMARY
1.1 INTRODUCTION
The Air and Energy Engineering Research Laboratory (AEERL) has instituted
a cccrdiT^ program to develop control technology for fugitive particulate
sources. A major source of fugitive particulate emissions is storage piles.
The AEERL has identified windscreens as a promising control technique for this
source. However, before this technology can be effectively applied, applica-
tion criteria need to be developed. These criteria include: (1) screen
porosity, (2) screen distance to pile, (3) screen width, and (4) screen
height. Answers are needed to these and other related questions before the
use of windscreens can be optimized.
AEERL and the Environmental Sciences Research Laboratory (ESRL) have
completed an inhouse study (Billman 1985), using the ESRL wind tunnel,
designed to determine changes in windspeed (not changes in emissions) due to
windscreens. Experiments were conducted to determine the optimal windscreen
*
porosity, size and location for storage-pile fugitive dust emission control.
In order for this information to find application in the design of \ind-
screens, it is necessary to conduct a field study to validate the wind tunnel
studies with respect to windspeed changes, and to determine the relationship
between changes in windspeed and changes in fugitive dust emissions.
More specifically, the three objectives of this study are:
1-1
-------
(1) To verify that the data collected in the wind tunnel with respect to
changes in windspeed are accurate under field conditions.
(2) To determine the relationship between changes in windspeed and
changes in particulate emissions by particle size.
(3) To develop windscreen design parameters.
The remainder of this section presents a summary of the results of the
study. Section 2 presents an overview of previous studies rn windscreens.
Section 3 contains a description of the field sampling for the present study.
Section 4 contains the analysis results for Objective l--Verification of Wind
Tunnel Wind Speed Data. The analyses for Objectives 2 and 3 are presented in
Sections 5 and 6.
1.2 SUMMARY
The Air and Energy Engineering Research Laboratory (AEERL) and the
Environmental Sciences Research Laboratory (ESRL) have completed an inhouse.
study (Billman 1985), using the ESRL wind tunnel, designed to determine
changes in windspeed (not changes in emissions) due to windscreens. Experi-
ments were conducted to determine the optimal windscreen porosity,- size and
location for storage-pile fugitive dust emission control. In order for this
information to find application in the design of windscreens, it was necessary
to conduct a field study to validate the wind tunnel studies with respect to
windspeed changes, and to determine the relationship between changes in
windspeed and changes in fugitive dust emissions.
Previous studies have yielded contradictory results concerning the
relationship between particle emissions and windspeed. Similar contradictions
were found in the twc studies performed to investigate reductions in dust
concentrations due to the use of windscreens. The Billman study and the study
described herein are the first laboratory and field studies which attempt to
measure windspeed or particulate reductions at or near a pile surface.
1-2
-------
The Billman study simulated, in a wind tunnel, the effect of a windscreen
on reducing windspeed on the surface of a storage pile. The scale model
storage pile used was 11 cm tall and nas covered with gravel having diameters
less than 4 mm. A variety of windscreen parameters were- evaluated during the
study and' isotachs of 'windspeed and windspeed reduction were presented both
for unscreened and screened piles. Based on the results, Billman calculated
inferred emission reductions assuming that the change in emission rate was
proportional to the cube of the windspeed. No emission measurements were made
during the study.
The present study was a field exercise to evaluate the results of the
Billman study under actual conditions. The basic sampling protocol used was
to measure windspeed and particulate concentrations on two identical storage
piles simultaneously. One pile was control led with a windscreen and one had
no windscreen. The control efficiency is then simply the difference between
corresponding values for each pile. Instrumentation for each pile consisted
of anemometers, RAM-1 monitors and exposure profiler samplers.
The first objective of the study was to compare the wind tunnel data with
the windspeed data collected in the field. The comparison raci two major
elements: comparison of the windspeed isctachs for an unscreened pile and
comparison of the windspeed isotachs on screened piles by screen configura-
tion.
For the unscreened pile, composite u/u values (windspeed at pile sur-
face/windspeed at the maximum height of the pile) were calculated for 10°
incoming wind direction cohorts. The computerized data base developed for
this analysis consisted of five minute average windspeed data, stratified by
incoming wind direction. As the wind direction moves around the pile, the
1-3
-------
stationary sensor locations were.effectively shifted .to new positions relative
to the isotach plots in Figure 3-3.
The results of the analysis showed a good comparison between wind tunnel
and field data for the front of the pile. However, the area where the u/u
ratio is _> 1 was substantially larger in the field data. The highest ratios
were found on the backside of the pile. The field data suggests that the high
windspeed flew lines not only extend around.to the back of the pile but are
reinforced in some fashion. The basic question relates to the comparability
of the idealized wind tunnel experiment to the real-world situation evaluated
••%
in the field. In general, the results from the two studies show good
agreement for the front of the pile. However, there are some additional
physical processes that still need to be investigated and explained.
For screened piles, the wind tunnel data were presented as a series of
i
isotach lines in the form of 1 - (u/u ) for windspeeds with (u) and without
(u ) a windscreen. Similar isotachs could not b<* developed for the field
data, as only four data points were obtained on each pile. However, manipu-
lation and analysis of the data obtained during the study yielded several
conclusions. Windspeed reduction was greatest for perpendicular screen
orientations. A 2.0-pile-height distance and a 1.25-pile-height screen height
were found to be most effective. For aperpcndicular winds, a 3.0-pile-
diameter screen width was more effective than a narrower screen. For
perpendicular winds, on the other hand, the 1.5-pile-diameter screen width was
the most effective. In the lee of the pile, negative control efficiencies
were recorded.
In comparison to the Billman study, both studies found the taller wind-
screens to be most effective. Billman found a 3.0-pile-height distance to be
1-4
-------
more effective than 1.0-pile-height diameter. A 2.0-pile-height distance was
not evaluated. This study found a 2.0-pile-height distance to be more
effective than either a 1.0- or 3.0-pile-height-distance. .Both studies found
a 1.5 screen diameter length to be more effective than a 1.0-screen-diameter
length. Both studies recorded negative screen efficiencies in the lee of the
pile, but the field study showed this result to a much greater extent. In
general, the wind tunnel efficiencies were higher than those measured in the
field.
The second objective of the study was to compare windspeed reductions and'
particulate control efficiencies. Due to problems with the RAM-1 data only
the total particulate data were used for this analysis.
Average windspeed reductions were compared with particulate emission
reductions for 42 one-hour tests taken with the profilers. It was found that
a highly significant relationship exists between windspeed and particulate
emission reductions, and the relationship is approximately linear with a slope
less than one. Also, there appear to be Instances where windspeed on the
front of the pile is reduced but emissions actually increase as a result of
higher windspeeds on the back of the pile. j
The total particulate (TP) data were disaggregated into discrete particle
size ranges based on laser diffraction analysis of selected filters. The
resulting percentages of net weight by size range were multiplied by the TP
emission rate to obtain emission rate by particle size range. These data
along with the corresponding windspeed data were subjected to regression
analysis. Slopes of regression curves for the two largest particle size
ranges showed an emission reduction almost equal to windspeed reduction. The
smallest particle size ranges showed no significant relationship.
1-5
-------
Further regression analyses showed a strong linear relationship between
TP emission rate and windspeed and a strong inverse relationship between TP
emission rate and moisture content of the' pile. Approximately the same
relationship was observed between the two v-ariables and emission rate regard-
less of the size fraction considered.
The final objective of the study was to develop windscreen design para-
meters. In terms of screen length, it appears that screen lengths of
5.0-pile-diameter are appropriate for permanent or semi-permanent
installations. Given the wind direction variations that occur in real
situations the 1.0- to 1.5-pile-diameter lengths tested in the wind tunnel are
probably too short. The 2.0-pile-height screen-to-pile distance was found to
be optimum. This distance yielded slightlv nreater emission reductions than
either the 1.0- or 3.0-pile-height distance. Both the wind tunnel study and
this study showed that the 0.5-pile-height windscreen height was not as effec-
tive as screens of 1.0-pile-heights. Also, a screen height of 1.0-pile-height
is nearly as effective as higher screens. In general it appears thst the
optimum design parameters are: height - 1.0-pile-height; width =
5.0-pile-diameters; and distance = 2.0-pile-heights.
The field study has helped to identify several important areas for
further investigation. Although the wind tunnel study and the field study are
in general agreement for the front of the pile, there was one significant area
where the results are contradictory. The field study showed that large
portions of the back of the pile had windspeeds higher than the reference
windspeed. This observation was reinforced by the particulate emission data.
There were a large number of tests where negative emission reductions were
noted for the screened pile. This basic result is in direct conflict with the
1-6
-------
bulk of the wind tunnel data. Although Billman did find some negative reduc-
tions, the field study showed negative reductions as large as 40 percent.
There must be some ongoing physical process or processes that has not
been adequately investigated in this study. The results to date raise
questions on the applicability of windscreens for reducing emissions from
storage piles. Prior to recommending windscreens as a control measure, it is
imperative that the observed relationship between the use of windscreens and
emission rate be .investigated further.
1-7
-------
2.0 PREVIOUS STUDIES
2.1 WINOSPEED AND PARTICLE UPTAKE
Several relationships between wind speed and particle emission rate are
found in the literature. Bagnold (1941) suggested that the particle emission
rate is proportional to the cube of the wind speed. Gillette (1978a),' in a
wind tunnel test of the effects of sandblasting, wind speed, soil crusting,
and soil surface texture on wind erosion, showed that the soil particle flux
is proportional to the cube of the friction velocity (u*), where u^ is deter-
mined from the mean velocity profile over a horizontal surface,
U - U* In Z ,
2o
where U is wind speed at height z, z is the surface roughness length, and k
is von Karman's constant (^0.4). Blackwood and Wachter (1978) suggested that
the storage pile emission rate, Q (mg/s), may be expressed as
where c is a constant, u is wind speed (m/s), pb is bulk density (g/cm ), s is
2
pile surface area (cm ), and PE = Thorntwaite's precipitation-evaporation
index (Thorntwaite, 1931).
Field tests with portable, open-floored wind tunnels indicated that
threshold speeds, given in terms of threshold friction velocity (u^)t, are
typically 0.2 to 2 m/s depending upon the type of material (Gillette, 1978b;
Gillette et al., 1980; and Cowherd et al., 1979). In other field tests,
threshold speeds of about 10 m/s at a height of 15 cm above a coal pile
2-i
-------
surrace were estimated based upon the onset of visible particle uptake (Cow-
herd, 1982; Cuscino et al., 1983). Extrapolating these speeds to a 10 n
reference height from the velocity profile, implies that very high mean wind
speeds (e.g. 20 m/s) are needed for erosion at the surface (z=0) to commence.
Hence, Cowherd .(1982) suggested that strong wind gusts, not the mean wind,
cause erosion.
In the above relationships for particle emission, emission rate is
independent of time. However, unless an unlimited supply of erodible
particles is present, erosion will be time dependent. Erosion rate has been
observed to decrease with time (e.g. Cowherd et al., 1979). Cowherd (1982)
suggested that erosion rate is proportional to the amount of erodible material
remaining and that a given storage pile has an "erosion potential" equal to
the total quantity of erodible material present on the surface prior to
erosion.
Conclusions that can be derived from these studies are that:
(1) Particle emissions are related to windspeed, either directly or at a
power of the windspeed. There is a threshold windspeed under which
no erosion occurs, although results are contradictory.
(2) Emissions are limited by the amount of erodible material available.
2.2 WINDSCREENS AS A FUGITIVE DUST CONTROL MEASURE FOR STORAGE PILES
The use of windscreens has been proposed for reducing fugitive dust
emissions from active and inactive piles. Studies of windscreen effectiveness
have been performed on reduction in windspeeds, thereby theoretically reducing
emissions, and direct measurement of emission reductions.
Results of reduction in windspeed velocity caused by a porous wood fence
are shown in Figure 2-1 (Carnes anrt Drehmel 1982). Reductions in windspeed
2-2
-------
I
VgfTICAt. $£CTK>M
ALONG £ OF
IQH fOH 30N 40H SOU SOU
DISTANCE IN FCNCC HCISMTS
I r I
6ROUND
tHMO aCMDINGS AT
tt/3flA90VC 68OUNO
SOU
SOU
OtSTAMCf Ht ffftCC HCItHTS
Figure
-------
velocity of 60 percent were measured at a distance of 10 screen heights.
(Th.is does not necessarily mean a corresponding reduction in fugitive dust
emissions.) Windsoeec reductions downw.'nd of other types of windscreens
were measured by TP.C-trwironmental Consultants, Inc. (Carnes and Drehmel
1982). Using a 65 percent permet-.ble windscreen, with windspeeds of 3.0
m/s3c., wind reductions of 70 percent were measured immediately downwind, and
wind reductions of 40 percent were measured 14 heights downwind. For a 50
percent permeable windscreen, windspe«?d reductions were comparable adjacent to
the fence, but the reductions were smaller further downwind.
Reductions in fugitive particulate emissions were measured by TRC
as well as reductions in windspeed. Total suspended particulate (TSP)
emissions were sampled with high volume samplers (hi-vols). Testing was
performed on a flyash pile. The study concluded that the windscreen was
effective both API reducing wind velocity approximately 66 percent under
ordinary conditions and peak gusts by approximately 58 percent, and in
reducing TSP and inhalable particulate (IP) concentrations downwind by an
average of 75 percent and 60 percent, respectively.
PEDCo (1984) studied windscreens using RAM-1 aerosol monitors and wind-
speed sensors interfaced with a portable computer to give real-time data
results. The analysis indicated that the windscreen did not produce signifi-
cant reductions in concentrations in the less than 10 micrometer respirable
size range. The screen did reduce windspeeds by the amount anticipated, but
this did not result in commensurate reductions in particulate concentrations
coming from the pile.
An explanation for the windscreen's performance was that wind erosion
emission rates in the less than 10 micrometer size range were fairly constant
at windspeeds above a threshold of about 7 mph (hourly average). The
2-4
-------
additional emissions associated with high wind erosion losses at high wind-
speeds were larger particles that were not detected by the RAM-l's. The wind-
screen may be effective in stopping or reducing'the movement of these large
particles, but many of them do not stay airborne because of their relatively
large s^ze, so they present less f a threat of offsite exposure.'
In summary, all studies are in fair agreement about reductions in wind-
speed caused by windscreens. Only two studies have measured reductions in
dust concentrations as opposed to reductions in windspeed. The TRC study
found reductions in the TSP size range of 60 to 75 percent. The PEDCo study
of particles in the less than 10 micrometer size respirable range showed no
consistent benefit from the windscreen, buc acknowledged that positive control
efficiencies of larger size particles were likely.
This contradiction in findings between the TRC study that measured less
than 30 micrometer particles, and the PEDCo study which measured less than 10
micrometer particles suggests that particle uptake may respond to windspeed
changes differently according to particle size.
No study, laboratory or field base, has attempted to measure windspeed
reduction or particulate reductions at or near a pile surface before the
3illman Study (1985) and the! field study described herein.
2.3 THE BILLMAN STUDY
The Billman study (1985) simulated, in a wind tunnel, the effect of a
windscreen on reducing windspeed on the surface of a storage pile.
The experiment was conducted in the EPA Meteorological Wind Tunnel, a
low-speed, open-return tunnel having a test section 2.1 m high x 3.7 m wide x
18.3 m long. A neutrally stratified simulated atmospheric boundary layer was
2-5
-------
generated by a 15.3 cm high trip fence placed 22.3 cm from the test section
entrance. Gravel roughness composed of pebbles having typical diameters of 1
cm covered the tunnel floor downstream of the fence. The boundary layer was
characterized by a depth of approximately 1 m, a roughness length (z } of 0.1
mm, and a friction velocity (u*) of 0.048U . The model pile had to be small
enough to be within the surface layer but large enough to construct windbreaks
of height the same order as the pile height and to facilitate measurements.
The results was a model pile 11 cm high (37° slope and base diameter of 29.2
cm). The pile could not be roughened with the same gravel as that covering
the floor of the tunnel because the 1 cm gravel was too large with respect to
the pile size. Gravel having diameter less than 4 mm was used instead.
Heated thermistor beads were mounted directly on the pile to measure wind-
speed. Nine thermistor., were mounted on the simulated pile 2 to 3 mm above
the surface. Actual windbreak material could not be used due to scale prob-
lems. Nylon mesh screen was used, with the type of screen being selected
after wind tunnel testing of wind porosity.
Figure 2-2 shows the top view of the pile with contours of normalized
windspeed, u/u , where u is the windspeed measured at the pile surface, and u
is the incoming windspeed at the equivalent full scale height of 10 m. The
areas of maximum wind speed ar? near the top of the upwind face but toward the
sides of the pile. A high speed region (u/u > 0.75) is on the upstream face,
extending from near the crest down both sides. The area of minimum wind speed
is in the lee near the top of the pile with regions of low wind speed extend-
ing down the pile on both sides of the centerline. High speeds along the pile
sides are expected because the flow is accelerating around the pile. The flow
2-6
-------
flo
0.
.4
Figure 2-2 g/gr about conical pile for no windbreak case.
2-7
-------
separ. ies on the lee side, resulting in a region of low-speed recirculating
flow.
Windscreen/pile variables tested by Billman were (Billman i985):
c Pile shape—conical and oval
0 Screen porosity--50 and 65 percent
0 Screen height—0.5, 0.75, 1.0, 1.25 and 1.5 pile heights
° Screen length--!.0 and 1.5 pile diameters
° Screen position—on pile, 1.0 and 3.0 pile heights
0 Screen orientation—perpendicular, ± 20 and ± 40 degrees to wind
direction
An example plot of windspeed reductions is shown in Figure 2-3. Wind-
speeds in the example were reduced 40 percent over most of the pile face, with
a small area of 60 percent reduction.
Windspeed reductions are summarized in tabular form in Table 2-1. When
the reduction was averaged over the entire pile surface, values ranged from 21
to 51 percent. Considering the reduction in maximum values, windspeed
reductions ranged from 17 to 94 percent.
Since no changes in emissions were measured, Billman calculated inferred
emission reductions assuming that the change in emission rate is proportional
to the cube of the windspeed, and that all windspeeds exceed the erosion
threshold level. The latter assumption results in a maximum predicted impact.
The calculated inferred emission reductions ranged from 66 to 99 percent.
Concerning design parameters, conclusions reached based on area average
windspeed reduction were:
° Screen porosity—The 50 percent porocity was more effective.
0 Screen height—The 0.5 height was less effective than the 1.0 and
1.5 heights. The latter two heights showed similar effectiveness
except for the 50 percent porosity screen at 3.0 heights downwind,
where the 1.5 height was slightly more effective.
0 Screen length—Screen length made little difference in most cases.
This is due to the perpendicular flow of wind to the screen in the
experiment. The greater length did provide increased effectiveness
when winds were not perpendicular to the screen.
2-8
-------
20
Figure 2-3 Wind speed reduction factor for the 651 porous windbreak
of hefght 0.5H and length l.OD placed 1H from the conical pile base.
2-9
-------
TABLE 2-1. WINDSCREEN IMPACT FOR VARIOUS WINDBREAK CASES—
BILLMAN STUDY (percent reduction)
Screen to
pile distance
Screen
length
Screen Height
65% Porosity
0.5 H
1.0 H
1.5 H
50% Porosity
0.5 H
1.0 H
1.5 H
Area Average Windspeed Reduction
1.0 H
3.0 H
1.0 D
1.5 0
1.0 D
1.5 D
26
25
21
22
45
42
48
47
45
43
51
51
36
36
27
26
60
60
57
62
60
58
62
70
Maximum Windspeed Reduction
1.0 H
3.0 H
1.0 D
1.5 D
1.0 D
1.5 D
.91
.93
.91
.94
.55
.59
.54
.56
.56
.60
.50
.52
- .90
.93
.82
.86
.31
.34
.37
.27
.39
.42
.25
.17
Calculated Inferred Emission Reduction
1.0 H
3.0 H
1.0 D
1.5 D
1.0 D
1.5 D
74
72
66
67
88
85
91
90
86
82
92
91
82
80
76
76
97
97
97
98
95
95
98
99
1
Assumes that the change in emission rate is proportional to the cube of the windspeed.
-------
Screen position—At higher windscreen heights, the 3.0 pile height
distance was generally more effective than a 1.0 pile height
distance.
2-11
-------
3.0 FIELD SAMPLING
3.1 STUDY DESIGN
A detailed study design was set forth in a 1985 sampling report.
Reiterating from Section 1.0, the three study objectives were:
(1) To verify that the data collected in the wind tunnel with respect to
changes in windspeed are accurate in field conditions.
(2) To determine the relationship between changes in windspeed and
changes in particulate emissions by particle size.
(3) To develop windscreen design parameters.
By way of overview, the basic sampling protocol was to measure two iden-
tical storage piles simultaneously, one controlled with a windscreen and one
without. The control efficiency is then simply the difference between cor-
responding values for each pile. To meet objective 1 (windspeed reductions),
several anemometers were placed on each pile to measure windspeed. The field
results were compared to the values in previously cited Figures 2-2 and 2-3,
and Table 2-1. To meet objective 2 (windspeed/emissions), emissions were
measured by exposure profiling and RAM-1 aerosol monitors. To meet objective
3 (design parameters), the change in emissions data developed for objective 2
were used.
3.1.1 Pile/Windscreen Configuration
Reiterating, the basic test protocol was to establish two identical
piles, and to sample around the two piles simultaneously when one was
controlled with a windscreen and the other was not. A critical parameter in
3-1
-------
such a test protocol is that the piles be identical initially and throughout
the test period with respect to dust emitting characteristics. The piles were
constructed >ut of the same highly credible material, and were exactly the
same shape. After initial construction, both piles were sampled with RAM-1
mon.itors with both piles uncontrolled. The RAM-1 monitors output real-time
concentration data that can be used to instantaneously determine if both piles
are emitting dust in a similar manner.
It was anticipated that as testing began and continued over time, that
the uncontrolled pile would begin to emit dust at a different rate than the
controlled pile. To the extent that this occurred, the control efficiency
data derived from simultaneous comparative testing would be inaccurate. This
problem was overcome in three ways. These were:
(1) Outside of the eight hour test period, the windscreen on the con-
trolled pile was dropped. Consequently, for 16 of every 24 hours,
both piles were subject to the sanw erosional forces. This aided in
keeping the two piles similar.
(2) On a daily basis, at the beginning of each test day, instantaneous
RAM-1 measurements and windspeed measurements were made. Real-time
computerized five-minute averages were compared. If the pile
emission rates were ± 10 percent, the piles were considered to be
emitting at the same rate. If the difference was greater, the pile
emitting at the lower rate was raked to expose new soil. Comparar
tive readings were again taken until the values reached the desired
comparative level. A similar procedure was used with the windscreen
sensors to insure a ± 10 percent value for corresponding sensors on
each pile.
(3) At the beginning and end of the testing and after every 25 tests, a
complete test was run with all instrumentation in place and without
the wind screen. The results from the tests will show the overall
comparability of the piles.
3.1.2 Variables Tested
Because one of the objectives of the sampling was to verify the wind
tunnel testing, it was appropriate to analyze similar variables. The EPA-
3-2
-------
sponsored wind tunnel study (Billman 1985) contained examinations of the
following variables:
1. Pile shape—conical and oval
2. Screen porosity—50 and 65 percent
3. Screen height--0.5, 0.75, 1.0, 1.25 and 1.5 pile heights
4. ' Screen length--!.0 and 1.5 pile diameters
5. Screen position—on pile, 1.0 and 3.0 pile heights
6. Wind direction/Screen orientation—perpendicular, ± 20 and ± 40
degrees
These variables represent 360 total combinations. Each combination required
several repetitive tests.
It was estimated that 75 to 100 field test pairs could be completed with
available project resources. This range in tests allowed for an average
number of 3 tests per day, with the lower number representing test days lost
to rain or unmanageable winds. Therefore, it was apparent that not all wind
tunnel results could be verified in the field.
Tie choice of how many and which combinations to test was determined
using the following considerations:
1. How many tests for each combination are required to produce statis-
tically significant results?
2. Are results for certain of the variables already conclusive based on
the wind tunnel testing and other field testing?
3. Which variables most represent typical potential industrial applica-
tions?
3.1.2.1 Number of Tests Required—
The purpose of this subsection is to estimate the number of test values
of control efficiency that will be required to establish the mean control
efficiency with a predetermined precision and confidence. A control ef-
ficiency value v .Actually composed of two separate tests, one for an uncon-
trolled condition and the other for the controlled condition. Previous
3-3
-------
studies have shown that uncontrolled emission rates for the dust producing
activities are not normally distributed (PEDCo/KRI 1934 ). Consequently,
controlled emission rates and control efficiencies are .probably not normally
distributed either. Therefore, Stein's relatively simple two-stage method for
estimating required sample sizes cannot be properly applied. A similar method
for estimating sample size, based on the assumption that uncontrolled and
controlled emission rates are each Icgnormally distributed has been derived in
a itcent study (PEDCo 1984s). In addition to the assumption of lognormality,
the derivation also assumed that the relative standard deviations of the
uncontrolled and controlled data sets (untransformed) are equal. With the
latter assumption, the standard deviation from tests taken under a previous
EPA-sponsored windscreen study can be used i.o estimate expected variance in
the test data for this study (PEDCo 1984).
The equation derived in a recent study for estimating sample size is:
n =
In K
where:
n> = number of control efficiency values (CE), equal to number of uncon-
trolled or controlled tests
t = tabled t-value for specified confidence level end n-1 degrees of
freedom
s = estimate of population standard deviation (In-transformed), obtained
from previous testing
K = ratio of upper limit value to lower limit value for confidence
interval around (1-CE)
The estimate of standard deviation (of In-transformed values) was ob-
tained from sampling of a topsoil stockpile performed by PEDCo in 1984 (PEDCo
1984). It was felt that this operation was a reasonable approximation of the
proposed field tests in this study. The calculated value for the standard
deviation was 0.35 and the value of K selected was 3.
3-4
-------
Using Equation, 1 and trial substitutions of t-values with n-1 degrees of
freedom and 90 percent confidence, the required number of control efficiencies
•values of a control option can be calculated to be 4 as follows:
v/fn= (2.828)(2.!32)(0.35)
1 n 3
n = 3.7
This value represents the number of control efficiency values required. In
other words, a total of 4 paired field tests (4 control, 4 no-control) need to
be taken for each control evaluated.
3.1.2.2 Variables for which Conclusive Data Exist--
•^
Of the six variables listed in Section 2, the only variabl.e for which
data are reasonably documented a:iJ consistent is screen porosity. Data
(Carnes and Drehirel 1981; Lawrence 1983; PEDCo 1984) indicate that a 50
percent porosity is more effective than 65 percent porosity presumably because
an optimum balance of shielding and low turbulance is achieved. In addition,
the Billman (19B5) wind tunnel study also verified that a 50 percent porosity
screen was superior to a 65 percent porosity screen. In^refore, a 65 percent
screen porosity was not tested.
3.1.2.3 Typical Industrial Applications—
Typical piles found in industries such as the st-?el, cement, aggregate
and power industries come in all sizes and shapes. Piles may be conical or
rectangular in shape.
Pile heights vary but usually do rot exceed 30 feet. On 30 foot tall
piles, screens as tall as the pile, or 1.5 times the pile height are difficult
to install, maintain and move, and therefore, a screen height of less than one
pile height would be desirable if it was effective. Another option used on
3-5
-------
large tall piles with a flat top is to place the screen on top of the pile to
shield the flat top.
Base dimensions may be as small as a few feet, or as large as several
hundred feet in the case of coal piles for the power "Industry, or waste piles
for other industries. For smaller piles, pile screen lengths of 2 and 3 times
the pile base are feasible. For large piles, a 1.0 diameter screen length is
more feasible.
With regard to wind direction/screen orientation, windscreens are most
feasible installed perpendicular to the predominant wind direction. However,
when wind directions vary, angles of 20 and 40 percent are likely and very
often exceeded. For small screens, screens can be purchased in eight foot
heights on portable stands. In other applications, standards are placed in
cement slabs or other movable platforms. The slabs can be moved by forklift
so that the wind direction/screen orientation is more correct.
3.1.2.4 Variables to be Tested—
As shown in this section approximately 18 to 25 variable combinations can
be examined. Variables to be tested are indicated in Table 3-1. Twtnty-five
variable combinations are identified. In addition, 4 tests were performed
with both piles uncontrolled for quality assurance related reasons.
3.2 FIELD SAMPLING PROGRAM
3.2.1 Test Plot Layout
The field site was located on a privately owned farm in the Wichita,
Kansas area. Wichita, Kansas had the desirable characteristics of relatively
high speed winds with a predominent direction. The 24-acre field is located
in a rural area about 7 miles northwest of the Wichita Mid-Continent Airport.
3-6
-------
TABLE 3-1. VARIABLES TO BE TESTED
Screen to pile
distance
Ih
2h
3h
Screen
length
1.5D
3D
50 .
1.50
3D
5D
1.50
3D
5D
Screen height
0.5h
X
X
0
X
X
0
0
X
0
l.Oh
X
X
0
X
X
X
X
X
X
1.25h
X
X
0
X
X
0
X
X
X
l.SOh
0
0
0
X
X
X
0
X
X
h = pile height
D = pile diameter
x = combination to be tested
3-7
-------
It is level except for a gully that projects into the middle of the field from
a stream bed that forms the southern and eastern boundaries of the field. The
western edge of the field is bounded by a paved road. The northern edge
(downwind) is bounded by an unpaved road.
The entire field was covered with grass, which grew to a height of 2 to 6
inches. There were no continu?lly active particulate sources in ttie upwind
direction (south) from the field, just additional pastures and fields with
mature crops. However, for the first few tests there was some construction
activity at a bridge located south of the site. This activity did not occur
beyond the first few tests. Also, there were no tall windbreaks within
one-half mile to the south. Trees that grew along the stream at the south end
of the field only extended 10 to 15 ft above field level and were at least 500
ft distant from the sampling area. A sketch of the site is shown in Figure
3-1.
The storage piles were constructed identically from dried, shredded
topsoil. The piles were conical in shape, with a height of 8 feet and a base
diameter of 25 feet.
A detailed test plot is shown in Figure 3-2. The piles were located 150
feet apart. The instrument trailer was located 75 feet downwind cf the piles.
Screen widths up to five pile diameters were accommodated with this layout.
Since all downwind instrumentation were located on the pile, a wind direction
shift of 90 degrees from perpendicular would be required for cross-contamina-
tion. Test abort protocol called for test cessation when winds average
greater than 30 degrees from perpendicular for a five minute period. There-
fore, this test plot layout and the test abort protocol eliminated cross-conta-
mination.
3-8
-------
_l
Q
ce
x
8
£
13th St.
GATE
UNCONTROLLED
A PILE
APPROX. 500 ft
TO LOM TREES
CONTROLLED
PILE
TREES
STREAM
BED
Figurt 3-1. Test site
3-9
-------
Predominate
Wind
Direction
-125'-
H-25'-
•50'-
125'
O o in
• • •
in <*><-!
Ill
•3H
•2H
•1 H
O
Controlled
Pile
Screen
Upwind
SainpUng
Array
O
Uncontrolled
Pile
0 20 40 60 Feet
Figure 3-2. Test Plot Layout
3-10
-------
3.2.2 Sampling Equipment and Deployment
3.2.2.1 Windspeed and Wind Directiori--
l-.'indspeed was monitored with several MET ONE Wind Speed Sensors, Model
14a. This sensor has an accuracy of ± 0.25 mph, a starting .threshold of 1.0
mph, and a temperature operating range of -50CC to +70°C, ever a range of 0 tc
100 mph. The sensor is a rot?ting cup assembly with a pulsed output. The
output is directed through a wind speed trarjlator module that converts the
signal to a standardized analog voltage. This signal is translated to a
digital signal through the use of an analog to digital converter. This signal
was then processed by a personal computer.
Wind direction was monitored with a MET ONE Wind Direction Sensor Model
24a. The instrument has a threshold of i mph and an accuracy of i,5 degrees.
The signal is input to * translator module and an analog to digital converter
for computer processing.
A total of IP «mdspeed sensors were used. One sensor was located upwind
at a height of 8 feet, corresponding to the height of the storage pile. A
logarithmic wind speed profile was assumed for lower heights. This assumption
was based upon standard references as well as previous PEI field experience
testing windscreens and storage piles (PEDCo 1984).
Placement of the sensors was guided by study objective 1, i.e. verifica-
tion of wind tunnel testing. It was desirable to obtain wind speed measure-
ments at the same locations as in the wind tunnel testing (Billmon 1985).
However, the wind tunnel testing included 108 wind speed measurement loca-
tions, nine sensors at a time. Pile rotation to 12 positions yielded 108
measurements. Such a protocol was impractical for a -Tieid test because of the
3-11
-------
equipment requirement and because wind direction in the field was not fixed cis
in a wind tunnel test.
A total of nine wind speed sensors were deployed downwind of the screen,
five on the no control pile, and 4 on the control pile. The sensors were set
at a fixed position on the pile, about 6 inches above the surface of the pile,
ana perpendicular to the ground. The positions were set relative to the
prevailing wind direction, and the positions remained fixed over the eight-
week test period. The positions on the pile were set in order to be able to
evaluate Figure 5.1, 6.3, 6.6 and 6.8 of the Billman (1985) report. Ser.sor
placement for the uncontrolled pile is shown in Figure 3-3. The sensor
locations are superimposed over Figure 5.1 from the Billman (1985) report.
Because the isotach lines are symmetrical, and because of the inability to
fully instrument each pile, sensors were only placed on one-half of the pile.
Sensor placement on the controlled pile were in exactly the same locations as
the uncontrolled pile, except location 1 was not used.
In order to position identical instruments at the same relative locations
on each pile, a true north and true south point was determined for the base of
each pile. Then, a string was run across the peak of the pile connecting tne
two points. Vertical distances were measured along the string, while hori-
zontal dimensions were measured perpendicular to the string. Tape measure-
ments were accurate to less than one inch.
While it is possible that the instrumentation interfered with the flow
field around the piles, since both piles were instrumented exactly the same,
identical changes occurred on each pile. In order to make such measurements,
instruments must be placed on the pile even though the measurement systems may
slightly interfere with what is being measured.
3-12
-------
1
flow
0.6
0.4
u/ur about conical piIt for no windbreak case.
Figure 3-3. Wind Sensor Placement, Uncontrolled Pile
-------
A wind direction sensor was placed upwind. The height of the sensor was
eight feet, the height of the storage.pile. The data were used to determine
the angle of the wind to the winds'creen.
3.2.2.2 Particulate—
Particulate was measured with 2 devices, a total particulate exposure
profiler head, and a model RAM-1 aerosol monitor, manufactured by GCA
Environmental Instruments; Bedford Massachusetts.
Profiler Head—
The exposure profiler heads consisted of an adjustable flowrate
high-volume motor, a filter holder, and a cylindrical intake nozzle which was
oriented directly into the wind during testing. The filter media was the
standard glass fiber high-volume filter. Since the sampler collects all
ambient particles non-d^scriminately, the emission data obtained represented
total particulate (TP).
The sampling heads were operated at a near Isokinetic flowrate so as not
to skew the particle size distribution of the collected sample. This design
was potentially difficult since the pile and windscreen induced wind currents
would not follow the standard logarithmic profile, and would change with
changing wind direction and windscreen height. This problem was overcome by
mounting rotating cup anemometers near each profiling head mounted on the
profiling tower. The anemometers sent data to the Apple computer. Windspeeds
for each sampling height were averaged from the computer every ten minutes,
and sampler flow rates were checked and adjusted accordingly to maintain a
near isokinetic flowrate.
This exposure profiler head has been used in numerous Environmental
Protection Agency (EPA) emission factor studies and has a long field history.
Quality assurance procedures are wsll documented and reproducibility is
excellent. The heads were calibrated to actual '"'ield conditions.
3-14
-------
It. was desirable to obtain particle size data from the exposure profiler.
filters. These data permitted a determination of windscreen control ef-
ficiency by particle size. Alternative methods to obtain the size distribu-
tion from the filter were optical microscopy, scanning electron microscopy and
laser diffraction. All methods share the same two weaknesses, i.e. material
must be removed from the filter, and physical size data must be converted to
aerodynamic size data. The most reliable and cost efficient of the methods is
laser diffraction, using the Microtrac Particle Size Analyzer manufactured by
Leeds and Northup. This device outputs particle size distributions in up to
13 particle size classes over a range of 1 to 175 micrometers.
Project resources were adequate to use laser diffraction on one filter
from each tower per test. Each sample was then subjected to the particle
sizing analysis.
The laser diffraction technique requires a relatively large amount of
mass for analysis. This placed two requirements on the sampling. They were:
(1) That the profiler heads be located'very close to the source to
collect the maximum amount of materiel.
(2) That the upwind samples be combined for a single days testing. If
sampling is conducted in an area with very low background concen-
trations, this combination will not compromise the data to an
unacceptable level.
There are two basic methods to perform exposure profiling sampling of
dust, emissions, as a point source, or as a line source. Sampling the pile as
a point source requires that both a horizontal and vertical plume profile be
obtained. This would require eight to twelve sampling heads per pile, an
unreasonable number of samplers in light of the desired 100 to 125 test pairs.
If the source is considered to be a line source, only a vertical profile is
required. If a pile is to be considered a line source, however, the concept
3-15
-------
of a line source must be extended to include a "bent line source". To make
this assumption, it must be assumed that the area in the shaded 0.8 area en
previously sited Figure 3-3 is emitting at a relatively uniform rate across
its longest dimension.
A second issue was where to place the samplers relative to the pile.
Options were: on the pile; immediately downwind of the pile; or >10 feet
downwind of the pile. Four factors influenced this decision. They were:
0 Laser diffraction, to be used for particle sizing, requires a
relatively large sample.
0 Exposure profile calculation assumptions require that the plume be
sampled before the largest particle sire of interest has fallen out
of the plume. Total particulate data are of interest.
0 There is an area of wind eddy behind the pile for an unknown dis-
tance where the plume behaves abnormally.
0 Tc satisfy study objective 2, i.e. determine the relationship
between changes in windspeed and changes in emissions, it is desir-
able to associate a specific oust measurement with a specific wind
sensor. The further the sampling head is located from the wind
Sensor, the mere difficult the association becomes.
These four factors all directed that the samplers be placed on the pile.
In order to sample the pile as a bent line source with the exposure
profiling technique, 4 exposure profilers were mounted on an 11 foot tower.
The tower was located on the pile, 10 feet behind the midline of the pile, and
2 feet to the side of the top of the. conical pile (Figure 3-4). Profiler head
heights were 4, 6, 8, and 11 feet off the ground. Samplers A and B were
associated with windspeed sensor 2.
RAM-1 Monitor—
The RAM-1 monitor is a portable sampler for respirable particulate. Its
measurement is based on detection of near-forward scattered electromagnetic
radiation by particles passing through the optical chamber. Air flow is
2-16
-------
Q D
11'
8'
6'
•25'
* Hind Sensor
O Profiler Head
D RAM-1
Note: Profiler heads behind pile at 4 and 6 feet heights.
Figure 3-4. View of Sampling Array from Upwind of Pile
3-17
-------
maintained at -i constant rate of about 2 liter/min. The sampler has a DrQ of
10.microns. A pu'ised semiconductor light-emitting diode generates a narrow-
band signal; after passage through the sample, the radiation is detected by a
silicon photovoltaic-type diode'with integral preamplifier. Maximum sampling
time is 32 seconds (other options are 0.5, 2 .jnd 8 seconds). The instrument
outputs an anaTog signal, which when used with an analog to dicital converter,
outputs a digital signal suitable for use with a computer.
Independent evaluation of the RAM-1 has shown reproducibility error of 3
to 5 percent and average comparisons with low volume sampler gravimetric
readings of 0.90 to 1.20.
A total of five RAM-1 monitors wer? used. One was located at the eight
foot height upwind. Two others were located at each pile. Again to satisfy
study objective 2, it was desirable to associate each RAM-1 monitor with a
specific /vindspeed monitor. The locations of the RAM-1 monitors are also
shown in Figure 3-4. RAM R-l can be associated with windspeed measurements
from windspeed sensor 2. RAM R-3 can be associated with wind sensor 3.
3.2.2.3' Independent Variables-
Independent variables monitored were:
0 Temperature
0 Pile surface silt content
c Pile surface moisture
Soil samples were taken from the pile hy removing the top i inch cf soil
in a vertical strip of 1 x 48 inches from the middle of the pile.
Samples of soil were stored briefly in their airtight containers, then
reduced with a sample splitter (riffle) to about. 1 kg. The final split
samples were placed in a tared metal par,, weighed, and dried in an oven at
HOT for 24 hours. The dried samples were reweighed and the moisture content
3-18
-------
calculated as the weight loss divided by the original weight of the sample.
The dried samples were stored in airtight containers until they could be
sieved.
Sieving of these samples was done with mechanical dry sieves. The
portion of the material passing a 200 mesh screen is defined as the silt
content (>75 ym). The nest of tared sieves was placed on ? conventional
shaker for 15 irnnutes. Each sieve was then weighed to determine the distri-
bution of material and the silt content. For 10 percent of the samples, both
halves of the final split were analyzed for moisture and silt content. This
duplication allows determination of the reproducibility of the methods.
Temperature data was collected because of previous experience that the
RAM data becomes inaccurate at temperatures greater than 105°F.
No data related to dispersion (e.g. cloud cover and solar intensity) were
collected because no dispersion calculations were required.
3-19
-------
4.0 QUALITY ASSURANCE
4.1 QUALITY ASSURANCE PLAN
A detailed quality assurance (QA) plan was prepared prior to starting
the field sampling. The report was used as a guideline throughout the field
sampling, data analysis, and data evaluation portions of the project.
Principal elements of the QA procedures are contained and reviewed in this
section.
4.1.1 Precision, Accuracy, and Completeness Objectives
Objectives for precision, accuracy, and completeness as stated in the QA
plan are shown in Table 4-1. Table 4-2 shows the criteria for precision,
accuracy, and completeness met during the sampling phase of the project.
Precision is defined as a measure of mutual agreement among individual
measurements of the same property under similar conditions. It was difficult
to establish the precision for the exposure profiler samples used in this
study. Precision is normally determined from measurements taken with a pair
of collocated instruments. In general, for particulate measurements, the
siting guidelines require that the samplers be separated by at least 2 meters.
For the test procedures scheduled for this study location, differences of only
several inches could yield drastically different results. Consequently,
precision could not be properly evaluated within the constraints of any
specific test.
In the case of the RAM-1 and windspeed sensors, a special array of paired
instruments of each type was erected once during the study. These instruments
were deployed to measure ambient conditions in close proximity to each other
4-1
-------
TABLE 4rl. PRECISION, ACCURACY AND COMPLETENESS OBJECTIVES
Measurement
Parameter
RAK-1
Exposure
prof i ler
heads
Windspeed
Wind
direction
Reference
Appendix A
EPA-600/4-
77-027a
May 1977
Appendix B
Appendix B
Experimental
Conditions
Ambient Air
Ambient Air
Ambient Air
Ambient Air
Precision
Std. Dev.
± 15%
N/Aa
± 5%
M/Aa
Accuracy
± 5%
± 7%
± 5%
± 5°
Completeness
> 80%
> 80%
> 80%
> 80%
N/A - not applicable - see text
TABLE 4-2. PRECISION, ACCURACY AND COMPLETENESS RESULTS
Measurement
Parameter
RAM-1
Expos ive
profiler
heads
Windspeed
Wind
direction
Precision
Std. Dev.
2.2%
N/A
3.9%
N/A
Accuracy
<3.9%
<6.3%
±5.0%
0.0%
Completeness
20.0%
85.4%
98.1%
100.0%
N/A - not applicable - see text
4-2
-------
in a location removed from the test area and away from local obstructions and
interferences. Since the computerized data capture occurred at 1-minute
intervals only a short exposure was .necessary. Precision test data are
displayed in Tables 4-3 through 4-6. For these data, both the windspeed
sensors and the RAM-1 instruments agreed to within 5 percent cf full scale.
Sensors and instruments showing the greatest differences in recorded values
were cleaned and recalibrated to improve comparability.
For the exposure profiler heads, no precision determinations were made.
In the ambient air, collocated samplers would need to operate for up to 8
hours to obtain ah adequate mass for analysis. It was not feasible to remove
two instruments from the sampling array on a regular basis to perform such a
comparison. In the plume downwind of the storage piles, any comparisons of
closely located samplers would be meaningless due to concentration variations
within the plume. No precision estimates of wind direction were obtained
since only one wind direction sensor was deployed.
The accuracy detenninations for the exposure profiler heads were obtained
from single point flow checks performed twice during the study using a
standard orifice calibration kit different from the one used for calibration.
Similarly for the RAM-1 instruments, a standard bubble tube was used to audit
the.RAM-1 instrument flowrates once during the field portion of the project.
Exposure profiler sampler flow checks are summarized in Table 4-7 and 4-8.
RAM-1 audit flow rate results are displayed in Table 4-9. All audited values
for the RAM-1 were within 3.9 percent of the original calibration. For the
profilers, all values were within 6.3 percent.
For windspeed accuracy, the manufacturer's stated accuracy of i 5 percent
was used. Wind direction accuracy was determined using a compass. The
4-3
-------
TABLE 4-3. SCREENED PILE WINDSPEF.D SENSOR PRECISION TEST
PEI ASSOCIATES, INC.
WINDSCREEN MODEL VERIFICATION
Location: SCREENED PILE
Test Description: SENSOR TEST
Date: 07-22-85
1
TiM
15:00
15:05
ISMO
15:15
15:20
15:25
15:30
15:35
15:40
15:45
15:50
15:53
ilipud
TiM
0
5
10
IS
20
25
30
35
40
45
50
55
I —
0
0
2322
2053
0
0
0
0
0
0
0
0
• RM2
-) (U6/H2)
0
0
2
12
19
0
0
0
1147
413?
4127
4097
RAM
(U6/B3)
0
0
0
14
0
0
0
0
1043
4456
4428
4409
FHS2
IFT/H)
367
287
336
208
414
320
371
347
239
388
328
299
W54
-------
TABLE 4-4. EXPOSED PILE WINDSPEED SENSOR PRECISION TEST
PEI ASSOCIATES, INC.
WINDSCREEN MODEL VERIFICATION
Location: EXPOSED PILE
Test Description: SENSOR TEST
Date: 07-22-85
Elipud
TIM TiM
13:00
13:05
15:10
13:13
13:20
13:23
15:3C
13:33
13:40
13:43
ISiSO
13:55
feu
ton ma
0
3
10
15
20
23
30
35
40
43
50
55
RMU RM3 PMS1 PNS3 PUSS PNS7 PVS9 HHS5
(U6/N3) (US/N2) (FT/HI (FT/HI (FT/N) in/Hi (FT/N) (FT/HI
0
0
59
66
0 .
0
0
0
240
2762
2805
2820
729
2820
0
0
0
0
0
0
0
0
2094
4430
4433
4446
1284
4446
382
315
369
215
424
360
345
465
248
459
404
334
362
465
377
311
339
199
416
350
344
414
227
410.
349
313
338
416
385
328
343
203
425
357
322
372
212
379
347
316
332
425
416
331
368
201
435
321
302
372
214
423
316
319
335
435
358
325
316
195
428
298
325
331
173
404
354
278
315
428
382
327
341
193
447
304
338
332
195
395
315
296
322
447
WS8 Wll (MM NSPD
(FT/N) (FT/HI (U6/K3) (FT/NI
0
0
C
0
0
0
0
0
0
0
0
0
0
0
258
225.
257
131
376
244
279
269
109,
325
270
223
247
376
1218
926
460
1448
1141
1913
114
102
2455
5880
SOT4
5912
2289
5912
319
234
355
270
493
413
328
483
271
717
395
428
392
717
UIR
(SEC)
4
4
2
2
3
3
2
2
2
3
3
3
3
4
TiM it Niiwa 15:55 15:55 15:35 15:20 15:20 15:20 15:20 15:20
12:20 15:55 13)43 15:00
4-5
-------
TABLE 4-5. SCREENED PILE RAM-1 INSTRUMENT PRECISION TEST
PEI ASSOCIATES, INC.
WINDSCREEN MODEL VERIFICATION
Location: SCREENED PILE
Test Description: RAM TEST
Date: 07-22-85
— i(AN2 MM PIB2 MB4 PNS6 HS8 IWSS HMS8 Nil UttW USPO NOIR
TIM Tin ( ) (U6/H2) (U6/H3) (FT/N) (FT/HI (FT/MI in/Hi (FT/11) (FT/HI
-------
TABLE 4-6. EXPOSED PILE RAM-1 INSTRUMENT PRECISION TEST
PEI ASSOCIATES, INC.
WINDSCREEN. MODEL VERIFICATION
Location: EXPOSED PILE •
Test Description: RAM TEST
Date: O7-22-85
Elipttd AMI rW3 PK1 P»S3 PW5 P«7 MK9 HW1 HKS8 Hill Uttfl HSPD KOIR
TIM TIM (U6/flJ) IU6/N2) (FT/HI (FT/11) (FT/HI (FT/HI (FT/HI (FT/HI (FT/HI (FT/4 <«/«) (FT/HI (SEC)
16:43
16:30
14:55
0
5
10
0
2
2
30
33
30
423
211
634
404
Ml
564
412
2i6
593
418
206
572
362
215
538
383
203
564
0
0
0
303
146
484
43
41
41
436
206
563
J
3
3
feu 2 32 423 396 407 399 372 384 0 313 42 402 3
2 35 634 584 593 572 538 564 0 489 43 563 3
TIM it 8ii IM> :<>:53 16:50 16:33 16:33 16:53 16:33 16:55 16:55 16:55 16:45 16:55 16:45
4-7
-------
TABLE 4-7. EXPOSURE PROFILER SAMPLER
AUDIT RESULTS, JULY 29, 1985
Sampler .
. 1
2
3
4
5
6
7
R
9
Flow Rate
. CFM
35.99
37.32
33.76
35.23
33.39
35.56
34.18
34.10
35.23
Audit Flow
. CFM
37.58
38.10
36.03
36.29
35.52
37.84
35.78
36.03
37.58
Difference
CFM
-1.59
-0.78
-2.27
-1.06
-2.13
-2.28
-1.60
-1.93
-2.35
Percent
Difference
-4.22
-2.05
-6.30
-2.92
-6.00
-6.03
-4.47
-5.36
-6.25
4-d
-------
TABLE 4-8. EXPOSURE PROFILER SAMPLER
AUDIT RESULTS, AUGUST 28. 1985
Sampler'
1
2
3
4
5
6
-»
/
8
9
Flow Rate
CFM
44.43
37.87
27. 61
29.16
47.05
43.15
30.42
27.97
27.06
Audit Flow
CFM '
44.71
38.61
26.91
28.45
47.46
44.08
31.31
26.37
28.45
Difference
CFM
-0.28
-0.74 •
0.70
0.71
-0.41
-0.92
-0.89
1.60
-1.39
Percent
Difference
-0.63
-1.92
2.59
2.48
-0,87
-2.10
-2.84
6.08
-4.90
4-9
-------
TABLE 4-9. INSTRUMENT FLOW RATE
AUDIT RESULTS
Site: Wichita
Date: 07-29-84
Auditor: K Rosbury
Audit Device: Bubble Tube (500 cc)
Sampler
UWRM S/N 1393
RAM2 S/N 1302
RAM1
RAM3 S/N 1230
RAW S/N 1394
Time
14.84 sec
14.79 sec
15.39 sec
15.30 sec
15.64 sec
I 5. 57. sec
15.14 sec
15.22 sec
15.33 sec
15.32 sec
Flow Rate
2025 cc/min
1955 cc/min
1922 cc/rain
1976 cc/min
1958 cc/min
Diff
25.0 cc
-5.0 cc
-78.0 cc
-24.0 cc
-42.0 cc
% Diff
1.25 %
-0.25 %
-3.90 %
-1.20 %
-2.10 %
4-10
-------
accuracy of the sensor was determined during the audit on July 29, by
comparing the.displayed directional value, on the computer, to a compass
reading. During the audit, there was no difference between the compass and
computer read-ing.
The use of the RAM-1 aerosol monitors was discontinued after test 18 and
data collected between tests 11 and 18 were voided when the instruments failed
to meet quality control requirements. Data froir, the instruments were only
used for eleven of the fifty-five tests resulting in a 20 percent data
capture.
Because of a high moisture content on the pile surface, eight tests were
performed without the exposure profiler samplers. This lowered the
completeness of the data for the exposure profiler heads to 85.4 percent.
Some wihdspeed data was voided because of sensor malfunctions. This deletion,
however, resulted in less than a 2 percent data loss. Data completeness for
the windspeed sensors was 98.1 percent. No wind direction data were lost,
resulting in 100 percent data completeness for wind direction.
A.2 SAMPLING PROCEDURES FOR CRITICAL MEASUREMENTS
4.2.1 Daily Procedures
Each test day a wind forecast was obtained from the National WeatHer
Service (NWS) station at the Wichita Mid-Continent Airport, which was
located about 7 miles from the site. If appropriate winds were forecast, the
upwind and downwind sampling arrays were deployed appropriately for the
expected wind direction. Next, the RAM-1 instruments were electronically
zeroed and spanned. The windspeed sensors were also zeroed and electronically
spanned. The wind direction instrument was steadied and aligned to magnetic
north with a compass.
4-11
-------
After all instruments and samplers were checked for calibration drift
criteria, a short comparative test was made to check pile comparability and
sensor precision. Windspeed sensors located on the pile surface and the RAM-1
monitors were run for short periods (10 to 15 minutes) before the screen was
put in place to compare instrument readings between paired sensors on the two
piles. Windspeed sensors whose reading did not agree within 1C percent of
each other were checked for free cup rotation and were cleaned and lubricated
as necessary. RAM-1 instruments whose readings showed noticeable differences
were checked for correct calibration factors, re-spanned, and the calibration
•%
factors were updated as necessary.
The windscreen was erected as needed for the test and sampling commenced
when ambient windspeed exceeded 6 mph. A soil sample was taken from the
surface of each pile for later moisture and silt, analyses. The appropriate
i
flowrat.e for the isekinetic exposure profile heads was determined based on
ambient windspeed and the samplers were set accordingly at the start of the
test. Computerized data capture for the RAM-1 instruments and the wind-
speed/wind direction sensors was begun simultaneously. The exposure profiler
heads were started individually. All instruments were started within a 3-5
minute time span.
During the 1-hour test, the computerized data collection required little
attention. Two of 24 input signals were monitored each tost by temporarily
hooking the channels into a strip chart recorder. The computer monitor
indication was compared to the trace on the recorder to ensure that each
channel was performing correctly. In all cases, the computer indication and
the strip chart trace agreed within 5 percent of full scale. A summary of
chart recorder audits of computer data unputs is displayed in Table 4-]0.
4-12
-------
TABLE 4-10. SUMMARY OF CHART RECORDER AUDITS OF COMPUTER DATA INPUTS
Date
07-18
07-18
07-18
07-18
07-18
07-18
07-24
07-27
07-27
07-27
07-27
07-27
07-30
07-30
07-30
07-30
07-30
07-30
07-30
07-31
07-31
08-02
08-02
08-02
08-02
08-02
08-02
Test
6
6
7
7
9
9
11
12
13A
13A
13B
13B
15A
15B
15B
16
16
17
17
18
18
19A
19A
19B
19B
20
20
Screened/
Exposed
Exposed
Exposed
Screened
Exposed
Screened
Screened
Exposed
Screened
Screened
Screened
Screened
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Screened
Screened
Screened
Screened
Screened
Screened
Screened
Screened
Channel
UWRM
HW11
RAM4
PWS3
RAM2
WSPD
RAM3
RAM4
RAM4
PWS2
HWS8
UWRM
PWS9
RAM3
WSPD
HW11
RAM3
PWS7
UWRM
RAM3
PWS2
HWS5
PWS8
PWS6
PWS4
PWS2
HWS8
Time
1530-1540
1545-1555
1715-1730
1735-1745
1025-1040
1045-1055
1530-1540
1400-1410
1510-1520
1525-1535
1645-1655
1630-1640
1546-1556
1649-1659
1710-1725
1840-1850
1800-1810
1934-1944
1948-1958
1421-1431
1435-1445
1325-1340
1354-1404
1447-1500
1506-1520
1623-1633
1638-1652
Computer
Average
14.5 ug/ro3
826.5 ft/mi n
51.3 ug/m3
1224.5 ft/mi n
49.7 ug/ra3
1016 ft/min
5 ug/m3
3.5 ug/m3
3.0 ug/m3
403.5 ft/min
123 ft/min
12.5 ug/m3
610.5 ft/min
5.5 ug/m3
1106 ft/min
738 ft/min
6 ug/m3
MO ft/min
9 ug/m3
11.5 ug/m3
253.5 ft/min
301.7 ft/min
365.5 ft/min
570 ft/min
295 ft/min
500 ft/min
593.0 ft/min
Chart
Average
70.0 wg/m3
880 ft/min
200 ug/m3
1320 ft/min
160 ug/m3
1232 ft/min
200 ug/m3
120 ug/m3
120 ug/m3
440 ft/min
264 ft/min
100 ug/m3
616 ft/min
70 ug/m3
1144 ft/min
704 ft/min
80 ug/m3
264 ft/min
74 ug/m3
55 ug/m3
176 ft/min
334.4 ft/min
352.0 ft/min
616 ft/min
264 ft/min
528 ft/min
616 ft/min
Difference
55.5 ug/m3
53.5 ft/min
148.7 ug/m3
95.5 ft/min
110.3 ug/m3
216 ft/min
195.0 ug/m3
116.5 ug/m3-
117.0 ug/m3
36.5 ft/min
141 ft/min
87.5 ug/m3
5.5 ft/min
64.5 ug/m3
38.0 ft/min
34.0 ft/min
74 ug/m3
34 ft/min
65 ug/m3
43.5 ug/m3
77.5 ft/min
32.7 ft/min
13.5 ft/min
46 ft/min
31 ft/min
28 ft/min
23 ft/min
% Diff.
79.3 %
6.08 %
74.4%
7.23%
68.9%
17.5%
97.5%-
97.1%
97.5%
8.30%
53: 4%
87.5%
0.89%
92.1% '
3.32%
4.61%
92.5%
12.9%
87.8%
79.1%
30.6%
9.79%
3.69%
7.47%
10.5%
5.3%
3.73%
% Full
Scale
0.28%
1.22%
0.74%
2.17%
0.55%
4.91%
0.98%
0.58%
0.59%
0.83%
3.20%
-0.44%
0.13%
0.32%
0.86%
0.77%
0.37%
0.77%
3.25%
2.18%
1.76%
0.74%
0.31*
1.04%
0.70%
0.64%
0.52*
I
OJ
continued
-------
lable 4-10 (continued)
Date
08-03
08-03
08-06
08-06
08-08
08-08
08-08
08-09
08-09
08-12
08-16
08-16
08-16
08-16
08-16
08-16
08-16
08-16
08-16
08-16
08-16
08-26
08-26
08-26
08-26
08-L'7
08-27
08-27
08-27
08-27
08-21
08-21
08-21
08-21
08-21
Test
22
?3A
23B
24
25A
25A
25B
26
27
28A
28B
28B
29
29
30A
30B
30B
31
31
32
32
34A
34A
34B
34B
35
35
36
36
36
54
54
55A
89
90
Screened/
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Screened
Screened
Exposed
Exposed
Exposed
Screened
Screened
Screened
Screened
Screened
Screened
Screened
Screened
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Exposed
Channel
WSPD
HWS5
HW11
HWS8
HW11
PWS7
PWS1
PWS7
PWS5
HWS5
PWS7
PWS3
WSPD
WDIR
PWS1
PWS3
HW8
PWS2
PWS4
HWS8
HW11
HWS8
HUH
PWS6
WSPD
HWS5
WDIR
PSW5
WSPD
PSW3
WSPD
PWS1
PWS3
PWS9
PWS7
Time
1605-1620
1717-1733
1445-1455
1642-1655
1225-1237
1240-1250
1403-1436
1217-1236
1355-1412
1345-1350
1130-1155
1200-1219
1315-1335
1336-1340
1505-1520
1615-1630
16JO-1655
1750-1815
1815-1830
1915-1930
1930-1950
1322-1332
1345-1400
1500-1515
1516-1529
1035-1050
1050-1110
1206-1225
1226-1240
1250-1300
1355-1410
1415-1430
1453-1510
1647-1705
1740-1755
Computer
Average
1143 ft/min
904 ft/ml n
409.5 ft/min
695 ft/min
1004.5 ft/min
612.0 ft/min
495.5 ft/min
588.7 ft/min
949 ft/min
856 ft/min
422.8 ft/min
904.4 ft/min
719 ft/min
192.4 deg -
4ia.3 ft/min
815 ft/min
770.6 ft/min
293.2 ft/min
304.7 ft/min
491.3 ft/min
610.5 ft/min
411.5 ft/min
34,3 ft/min
428 ft/min
340.7 ft/min
555.7 ft/min
167 deg
358.7 ft/min
537 ft/min
489 ft/min
950.7 ft/min
522.7 ft/min
853.3 ft/min
618.5 ft/min
618.7 ft/min
Chart
Average
1232 ft/min
968 ft/min
528 ft/min
704 ft/min
1056 ft/min
7040 ft/min
528 ft/min
616 ft/min
968 ft/min
880 ft/min
440 ft/min
968 ft/min
836 ft/min
196.4 deg
440 ft/min
880 ft/min
836 ft/min
440 ft/min
396 ft/min
616 ft/min
704 ft/min.
440 ft/min
396 ft/min
484 ft/min
484 ft/min
616 ft/min
163.6 deg
396 ft/min
660 ft/min
528 ft/min
1012 ft/min
528 ft/min
924 ft/min
616 ft/min
660 ft/min
Difference
89 ft/min
64 ft/min
118.5 ft/min
9 ft/min
51.5 ft/min
92 ft/min
32.5 ft/min
27.3 ft/min
19.0 ft/min
24.0 ft/min
17.2 ft/min
63.6 ft/min
177 ft/min
3.97 deg
21.7 ft/min
65.0 ft/min
65.4 ft/min
146.8 ft/min
91.3 ft/min
124.7 ft/min
93.5 ft/min
28.5 ft/min
53 ft/min
56 ft/min
143.3 ft/min
60.3 ft/min
3.36 deg
37.3 ft/min
123 ft/min
39 ft/min
61.3 ft/min
5.3 ft/min
70.7 ft/min
2.5 ft/min
41.3 ft/min
% Diff.
7.22%
6.61%
22.4%
1.287,
4.882
13.1%
6.16%
4.43%
1.96%
2.73%
3.91%
6.57?;
14.0%
2.02%
4 . 93%
7.39% .
7.82%
33.4%
23.]%
20.2%
13.3%
6.48%
13.4%
11.6%
29.6%
9.79%
2.01%
9.42%
18.6% .
7.39%
6.06%
1.0%
7.C5%
0.40%
6.26%
% Full"
Scale
2.02%
1.45%
2.69%
0.20%
i.17%
2.1%
0.74%
0.62%
0.43%
0.55%
0.39%
1.45%
2.66%
0.73%
0.49%
1.48%
1.49%
3.34%
2.08%
2.83%
2.13%
0.65%
1.20%
1.27%
3.26%
1.37%
0.62%
0.85%
2.80%
0.89%
1.39%
0.12%
.1.61%
0.06%
0.94%
I
-e»
-------
A manual flow system was used for each profiler head to maintain nearly
isokinetic flow. Such a system has been used for every profiler determined
fugitive dust emission factor presently in AP-42. The 10-15 min. average
adjustment period has been found to be adequate. As noted in a previous
section, anemometers were collocated with the profiling heads at each height
of the towers. It was necessary to have anemometers at each heioht since the
pile influenced wind flow and the wind profile could not be assumed to be
lognormal. The anemometer signals were input to the on-site computer. Every
10 minutes during testing, an average windspeed was obtained for each of the
eight sampling heads, and flows on the profiling heads were adjusted
accordingly. Isokinetic ratios (windspeed/inlet velocity) varipd from 0.2 to
1.3. The extremely low ratios occurred under low windspeeds. The samplers
could be adjusted for windspeeds down to about 750 ft/min. For lower
windspeeds the samplers had to be run superisokinetic. Wichita was selected
as a test location because of the area's high persistent winds. It was
planned that normally testing would begin when the wind averaged approximately
750 ft/min. Since adverse weather conditions prevailed (excess rain,
northerly winds, and low windspeeds), and because more than two weeks of data.
were lost as a result of equipment damaged by lightning, every effort was made
to complete as many tests as possible. Numerous tests were run below the
minimum windspeed threshold since other test conditions were favorable (wind
direction and soil moisture content).
At the end of each test, the RAM-1 instruments and meteorological sensors
were placed in a standby mode until the beginning of the next test. The
exposure profiling heads were brought into the trailer where all filter
recovery activity took place. The exposed filters from the samplers were
4-15
-------
removed, logged into the field log book ani new filters were installed. The
samplers were then redeployed for the next test.
At the end of the day, the RAM-1 instruments were ruulintfly recalibrated
to check for drift during the day. Any drift over 2 percent required that
corrective action be taken during the data reduction. All sampler filters,
computer printouts and disks, and log books were stored in the locked trailer
overnight. If rain or high winds were forecast overnight, the storage p^es
were covered with tarpaulins prior to leaving the site.
4.2.2 Sample Handling
Only two types of samples were handled in this study. The first type
consisted of standard glass fiber hi-vol filters. All handling procedures
conformed to the standard operating procedures (SOP) for ambient TSP moni-
toring (EPA 1977). Filters were equilibrated at a constant temperature and at
a relative humidity of less than fifty percent for 24 hours before weighing.
Every tenth filter was reweighed. No filter weights differed by more than 5
mg. from the original weight after 24 additional hours in the controlled
temperature and humidity environment. The balance used for filter weighing
was checked for accuracy with class S weights during each weighing session.
Data records for these fijters were maintained in two locations. First, the
field data sheet, for each test contained all aspects of the test conditions
plus the filters used for each test. Second, a separate filter log book was
maintained to record the filters used each day.
The exposed filters remained within the field trailer u;itil they were
hand carried back to the laboratory for gravimetric and laser diffraction
analyses.
The second type of sample generated during the study was the soil sample
taken from each storage pile. A sample of the soil of each pile was taken
4-16
-------
prio1" to each test by removing the top half inch of soil in a vertical strip
of 1x48 inches from the front mid section of the pile. Samples were stored
briefly in a clean, airtight sample jar, then reduced with a sample spliter
(riffle) to about 1 kg. The final split samples were placed in a tared metal
pan, weighed, and dried in an oven at 110°F for 24 hours. The dried samples
were reweighed and the moisture content calculated as the weight loss divided
by the original weight of the sample. For ten percent of the samples taken a
moisture content for both sides of the final split was determined and the
results.for the two portions were then compared. In all cases, the moisture
analysis of both splits of the same original sample agreed within 5 percent.
The silt content, that portion of the sample passing through a 200 mesh
screen after being shaken for 15 minutes, was then determined for each soil
sample. The duplicate moisture analysis samples were compared for silt
content. All duplicate samples analyzed for silt content compared within ten
percent. This duplication of analysis allowed a QC determination of the
reproducibility of the method.
4.2.3 Data Records
A number of separate data records and log books were maintained by the
field team. Separate logs were maintained for filters, soil samples, equip- .
ment calibrations and maintenance, and other notes on events that affected the
testing. In addition, a computer-generated printout of the test results, a
field data sheet and a magnetic disk were obtained for each test. The field
records were coded with unique identification for each test. The field data
sheets were developed specifically for the testing and contained all relevant
support data for each test. The field supervisor had the responsibility for
maintenance of these records, and reviewed all records on a daily basis to
4-17
-------
tneir completeness and accuracy. These records remained within the
field trailer or in the custody of the field supervisor at all times. When
not in the personal custody of the supervisor, they were in the locked trail-
er. . ,
4.3 EQUIPMENT CALIBRATION
• All equipment used in the study was in proper working order at the outset
of the study. At the beginning of the testing, the exposure profile heads
were calibrated on-site according to accepted SOP for the high volume method
(Section 2.2.2 of the QA Handbook, EPA 1977). Once during the study, single
point flow checks of each instrument were performed.
The RAM-1 instruments had been calibrated against a primary standard
within 6 months prior to the testing. Then, at least twice a day during the
testing, the samplers were electronically zeroed and spanned.
The exposure profiler heads were calibrated with a standard orifice
calibration kit. The flows for the RAM-1 instruments were calibrated with a
bubble tube. Each of the calibration devices was traceable to a primary
standard within 6 months of this study.
4-18
-------
b.O OBJECTIVE 1--VERIFICATION OF WIND TUNNEL WIND SPt'ED DATA.
The first .objective of the study was to compare the wind tunnel dota
(Billmar. 1985) with the windspeed data collected in the field. The comparison
had two major elements. They were:
1. Comparison of the windspeed isotachs on an unscreened pile.
2. Comparison of the w.ndspeed isotachs on screened piles by screen
configuration.
5.1 WINDSPEED COMPARISONS BETWEEN WIND TUNNEL AND FIELD tESTING FOR AN
UNSCREENED PILE
As noted in Section 2, the Bill man (1985) windspeed data are presented as
a set of dontours of normalized windspeed, u/u , where u is the windspeed at
the pile surface, and u is the windspeed at the maximum height of the pile
measured in the absence of the pile. Previously cited Figure 2-2, token from
the Billrran (1985) report, summarized the wind tunnel data for windflow on an
unscreened pile.
In order to directly compare the field data to previously cited Figure
2-2, the following cata manipulations of the unscreened pile data were re-
quired:
1. Prepare a data bose with 5-minute average data, of windspeeds when
the incoming wind direction value was ±5° of perpendicular to the
sampling array. The wind direction restriction is necessary because
Figure 2-2 wind tunnel data was derived from a perpendicular wind
direction.
2. For each sampler location, prepare u/u values.
3. Summarize u/u values derived from the field for comparison to the
wind tunnel data.
5-1
-------
A computerized data base was prepared consisting of f iv~ minute average
windspeed data, stratified by incoming wind direction in 10° cohorts. Using
the 175 to 185C cohort data base for the unscreened pile,'u/u values were
calculated. These data are shown in Table 5-1. Incoming wirdspeed varied
from 207 to 1172 feet/minute.
The five composite u/u values from the field measurements are overlayed
on the isotach lines from the wind tunnel data in Figure 5-1. Based on the
results shown in Table 5-1 several preliminary conclusions can be drawn.
0 The field u/u values obtained on the front side of the pile match
extremely welT with the wind tunnel data.
° The field ratios on the back of the pile are much higher than the
wind tunnel data. The reason is not readily apparent. The field
data suggest that the isotach lines curve around the pile much more
in the field'than in the wind tunnel.
c At positions 2 through 5, the u/u values appear to be related to
incoming windspeeds. At the loweP windspeeds, the ratios are
higher. The ratios decrease as the windspeed increases. This
phenomena was not investigated in the wind tunnel.
With the data developed to this point, it is not possible to construct
isotach lines from the field data, because the samplers were deployed at only
five locations on the pile.
Composite u/u values were then calculated for the other 10° wind direc-
tion cohorts as shown in Table 5-2. As the wind direction moves around the
I
pile, the sensors locations are effectively shifted to new positions. This
same approach was used by Billman in the original wind tunnel study. However,
in the field study the wind direction was varied rather than the pile orien-
tation. Utilizing the entire data base yields a total of 80 data points. As
only )60 degrees of the compass were sampled, a substantial portion of the
compass is left unresolved. Some of the data points are plotted in Figure
5-2. The data are plotted over the isotachs from the wind tunnel study.
5-2
-------
TABLE 5-1. U/U VALUES FOR THE UNSCREENED PILE
WITH WINDS 175-184°
Incoming
U'indspeed,
ft/mi n
0-299
300-399
400-499
500-599
600-699
700-799
800-899
900-999
>1000
Mean Value
Position 1
u/ur
0.50
0.47
0.59
0.66
0.45
0.46
0.50
0.49
0.52
0.53
N
3
O
L
6
13
9
12
8
3
20
76
u/ur
X
0.94
0.89
1.17
1.25
0.96
0.87
0.94
0.92
0.98
—
Position 2
u/ur
1.33
1.28
1.03
1.07
1.11
1.11
1.10
1.07
1.05
1.09
N
3
2
6
13
9
12
8
3
20
76
u/ur
X
1.22
1.17
0:94
0.98 "
1.01
1.01
1.01
0.98
0.96
—
Position 3
u/ur
0.96
0.96
0.84
0.93
0.85
0.79
0.80
0.80
0.61
0.84
N
3
2
6
13
9
12
8
3
20
76
u/ur
X
1.14
1.14
1.00
1.11
1.01
0.95
0.95
0.95
0.96
—
Position 4
u/ur
1.23
1.02
0.72
0.99
0.98
1.01
0.92
0.90
0.87
.94
N
3
2
6
13
9
12
7
3
15
70
u/ur
X
1.31
1.09
0.77
1.05
1.04
1.G7
0.98
0.96
0.93
Pas it ion 5
u/ur
J.15
0.99
0.85
0.94
0.96
0.98
0.93
0.95
0.92
.95
N.
3
2.
6
13
9
12
7
3
-18
73
u/ur
X
1.21
1.04
0.89
0.99
1.01
1.03
0.98
1.00
0.97
—
For position locations, see Figure 5-1
N = Number of data points
* = Mean value of u/u
-------
flow
P.4
1 =
3 =
4 =
5 -
0.53
1 .05
0.84
0.94
0.95
Figure 5-1 Composite u/ur values for an unscreened
pile field testing compared to wind tunnel data.
5-4
-------
TABLE 5-2. U/U VALUES FOR THE UNSCREENED PILE
105°-264° WIND DIRECTIONS
Wind Direc-
tion 10°
Cohort
105-114
115-124
125-134
135-144
145-154
155-164
165-174
175-1S4
185-194
195-204
205-214
L'15-??4'
225-234
235-244
245-254
255-264
Position 1
N
2
2
9
19
35
69
77
76
79
1C1
47
25
I7
3
1
2
X
0.94
1.01
0.82
0.68
0.65
0.56
0.46
0.53 .
U.57
0.60
C.67
0.82
0.88
1.0?
1.02
1.00
Position 2
N
2
2
9
19
35
69
77
76
79
101
47
25
19
3
1
5
X
1.25
1.13
1.08
'1.08
1.13
1.07
1.09
1.09
1.11
1.08
1.11
1.16
1.25
1.17
1.24
1.23
Position 3
N
2
2
9
19
35
69
77
76
79
99
47
25
19
3
1
5
X
0.84
0.84
0.72
0.71
0.83
0.81
0.82
0.84
0.88
0.86
0.38
1.02
1.07
C.98
0.97
1.00
. Posi tion 4
N
/*
c
2
9
19
33
61
73
70
62
94
47
25
19
3
1
5
X
1.55
1.30
1.36
1.34
1.35
1.21
1.07
0.94
0.74
0.45
0.33
0.20
0.28
0.23
0.42
0.87
Position 5
N
2
2
9
19
35
69
77
73
72
99
47
25
19
3
2
5
X
1.12
0.95
1.05
1.08
1.14
1.06
l.CO
0.95
0.82
0.64
0.19
0.39
0.38
0.33
0.26
0.42
N = Number of data points
X = Mean Value of u/u
5-5
-------
0.4
0.4
0.4
0.4
0.4
Figure 5-2. Composite u/u values for an unscreened
pile-field testing and wind tunnel data.
5-0
-------
Those data <105C and ^265° are not included in the figure. The data were
outside the acceptable winu direction limits for the testing. Hence, they
were not assigned to a specific wind direction cohort. Also, only those wind
directions with N>5 are plotted (55 data points) as it was felt that 5 or less
data points in a particular 10C wind direction cohort was insufficient to
calculate an average u/u ratio given the variability in the data.
As can be seen in Figure 5-2, the data on the front of the pile match
reasonably well for u/ur <0.8. However, the area where the ratio is _>1.0
appears to be larger than wat found in the wind tunnel. The field data for
the back side of the pile yielded significantly higher U/L. ratios than the
wind tunnel study. In fact, the highest ratios measured during the field
testing occurred on the back of the pile. The testing suggests that the high
wind speed flow lines not only extend around to the back of the pile, but are
reinforced in seme fashion.
The differences noted between this study and the Billman study can be
attributed to a number of factors. (1) Ambient wind speeds and direction
measured in the field are much more variable than that observed in the wind
tunnel resulting in higher turbulence. (2) Actual pile configuration and
composition during the field testing may not have been comparable to the
idealized scale model pile used in th? wind tunnel. This factor may be
significant. (3) Experimental equipment usea in the two studies may not have
been comparable. It is unknown to what extent the wind sensors correspond to
the laboratory thermistors. (4) Experimental errors between the two studies
mey not have heeri of comparable magnitude. (5) The presence of reentrained
dust from the pile surfece may have sorre effect in the measurements. For
example, it may be that the kinetic energy inherent to the entrained particles
-------
from the front of the pile can be transferred to the wind sensors behind the
pile, the.reby yielding higher apparent windspeeds.
In general, the results from the two studies are comparable for the front
of the pile. There are some additional ongoing physical effects on the back
of the pile that still need to be investigated and explained.
5-2 WINDSPEED CONTROL EFFECTIVENESS
The Billman (1985) windspeed data for screened piles are presented as a
series of isotach lines. The isotach lines are presented in the form of 1 -
(u/u ), where u and u are windspeeds with and without a windbreak.
The field data were also manipulated into the 1 - (U/UQ) format and are
shown as Table 5-3. Data are also stratified in the table by incoming wind
direction and screen configuration. Data could not be presented as a series
of isotach line like the Billman report, because only four data points on the
pile were obtained (in contrast to 108 data points in the Billman study).
Ideally, much of the field data and the Billman wind tunnel data would
have been derived from directly comparable wind direction/windscreen confi-
guration combinations. The situation did not occur. Almost all the wind
tunnel data was gathered for the condition of perpendicular winds. Of the
screen combinations tested by PEI, four were directly comparable to wind
tunnel work. Other combinations tested by Billman, e.g. one screen width
configurations, were thought to be inappropriate for real world field ap-
plication where wind directions are constantly changing.
No perpendicular winds were recorded during field testing of 3 of 4
combinations. Wind tunnel testing with aperpendicular winds was limited to
one pile diameter width screens. PEI did not test the one pile diameter width
screen for reasons previously stated.
5-8
-------
TABLE 5-3.
SCREEN EFFECTIVENESS, (1- u/u0)
Incoming
Uind
Direction
< 105
105-114
115-124
125-134
135-144
'•
145-154
155-164
j
1
165-174
Screen
Configura-
tion2
1-3.0-1.25
1-3.0-1.25
2-1.5-1.25
1-3.0-1.25
1-3.0-1.25
2-3.0-1.0
1-3.0-1.25
2-1.5-1.0
2-3.0-1.0
2-5.0-1.0
2-1.5-1.25
1-3.0-1.0
1-1.5-1.25
1-3. 0-1. 25
2-1.5-1.0
2-3.0-1.0
2-5.0-1.0
3-3.0-0.5
2-1.5-1.25
2-3.0-1.25
NONE
1-3. 0-1. G
1-1.5-1.25
1-3.0-1.25
2-1.5-1.0
2-5.0-1.0
3-3.0-0.5
3-5.0-1.0
2-1.5-1.25
2-3.0-1.25
2-1.5-0.5
HONE
1-1.5-1.0
1-3.0-1.0
1-1.5-1.25
No.
Pts.
3
•1
1
2
4
4
3
2
7
2
5
2
1
3
7
1
3
5
6
2
5
8
5
4
6
7
5
11
4
14
2
3
2
2
3
Position
2
0.034
0.056
0.438
0.042
0.129
0.122
0.172
0.016
0.074
0.358
0.433
0.259
0.337
0.304
0.077
0.168
C.329
0.151
0.373
0.066
0.253
0.281
0.286
0.190
0.174
0.455
0.11?
0.462
0.442
0.408
0.283
0.317
0.286
0.425
0.478
No.
Pts.
3
1
1
2
4
4
3
2
7
O
4.
5
2
1
3
7
1
3
5
6
2
5
8
5
4
6
7
5
11
4
14 '
2
3
2
2
3
Position
3
0.265
0.184
0.209
0.181
0.226
0.018
0.235
-0.115
-0.024
0.306
0.086
0.333
0.219
0.365
-0.023
0.039
0.306
0.044
0.089
0.579
-0.141
0.331
0.244
0.28D
0.089
0.441
0:006
0.166
0.175
0.391
-0.049
-0.098
0.302
0.5C4
0.407
No.
Pts.
3
1
1
2
4 '
4
3
2
7
2
5
2
1
3
7
i
3
5
6
2
5
8
5
4
6'
7
5
11
5
14
2
3
2
2
3
Position
4
0.265
0.221
0.142
0.220
0.300
0.125
0.252
-0.030
0.054
0.270
0.136
b
0.272
0.361
0.102
0.043
0.199
0.105
-0.003
-0.059
-0.084
b
0.178
0.209
0.022
0.310
0.003
0.078
0.009
0.105
-0.042
0.104
b
b
0.416
No.
Pts.
3
1
1
' 2
4
4
3
2
7
2
5
2
1
3
7
1
3
5
6
2
5
8
5
4
6
7
5
11.
4
14
O
t_
3
2
2
3
Position
5
0.128
0.078
0.188
0.106
0.171
0.093
0.151
-0.061,
0.018
0.244
0.116
0.232
0.198
0.274
0.047
0.077
0.256
0.106
0.074
0.204
-0.052
0.255
0.132
0.160
0.048
0.359
0.027
0.086
0.004
0.163
-0.039
0.147
0.207
0.421
0.327
? Distance, H - Width, D - Height, H
No data
-------
TABLE 5-3 (continued)
Incoming
Wind
Direction
165-174
175-184
•185-194
195-204
Screen
Configura-
tion
1-3.0-1,25
2-3. 0-1. C
2-5.0-1.0
3-3.0-0.5
3-3.0-1.0
3-5.0-1.0
2-1.5-J.P5
2-3.0-1.25
2 1.5-1.0
2-1.5-0.5
NONE
1-1.5-1.0
1-3.0-1.0
2-3.0-0.5
2-3.0-1.0
2-5.0-1.0
3-3.0-0.5
3-3.0-1.0
3-5.0-1.0
2-1.5-1.25
2-3.0-1.25
2-1.5-1.0
2-1.5-0.5
NONE
1-3.0-0.5
1-1.5-1.0
1-3.0-1.0
1-3.0-1.25
2-1.5-0.5
2-3.0-0.5
2-3.0-1.0
3-3.0-0.5
3-1.5-1.0
3-3.0-1.0
2-1.5-1.25
2-3.0-1.25
2-1.5-1.0
NOf4E
1-1.5-0.5
1-3.0-0.5
1-1.5-1.0
1-3.0-1.0
1-1.5-1.25
2-1.5-0.5
No.
Pts.
3
4
15
7
4
6
1
5
8
13
4
5
1
3
7
6
9
15
5
5
6
7
3
4
8
10
7
4
1
8
4
5
3
12
5
1
8
3
11
9
4
3
6
11
Position
2
0.384
0.471
0.479
0.238
0.465
0.655
0.759
0.456
0.343
0.329
0.288
0.38i
0.372
0.265
0.516
0.490
0.232
0.525
0.635
0.645
0.533
0.416
0.430
0.191
0.230
0.49S
0.507
0.434
0.399
0.232
0.561
0.306
0.375
0.594
0.685
0.513
0.561
0.130
0.197
0.224
0.456
0.533
0.628
0.280
No.
Pts.
3
4
15
7
4
6
1
5
8
13
4
5
1
3
7
6
9
15
5
5
6
7
3
4
8
10
7
4
1
8
4
5
3
10
5
1
8
3
11
9
4
3
6
9
Position
3
0.508
0.443
0.457
0.106
0.221
0.452
0.513
0.417
0.188
0.041
-C.040
0.421
0.443
0.264
0.509
0.496
0.129
0.312
0.434
0.361
0.601
0.263
0.140
-0.075
-0.326
0.479
0.630
0.525
0.372
0.246
0.578
0.236
0.270
0.468
0.488
0.381
0.415
0.005
0.346
0.365
0.517
0.656
0.588
0.232
No.
Pts.
3
4
15
7
4
6
1.
5
8
13
4
5
1
3
7
6
9
15
5
5
6
7
3
4
8
4
1
8
4
5
3
10
5
1
8
3
11
9
6
11
Position
4
C.359
0.137
0.295
0.002
0.070
0.273
0.169
0.107
0.061
-0.006
-0.129
b
b
-0.738
0.186
0.217
-0.024
0.075
0.249
0.166
0.195
-0.039
0.060
-0.150
Q-.4Q5
b
b
0.154
-0.858
-1.173
0.133
-0.102
-0.216
0.051
0.118
-C.132
0.064
-0.234
-3.044
0.119
b
b
0.880
-1.151
No.
Pts.
•3
4
15
7
4
6
1
5
8 .
13
4-
5
1
3
6
9
15
5
5
6
7
3
4
8
10
7
4
1
8
4
5
3
10
5
1
8
3
11
9
4
3
6
11
Position
5
0.369
0.288
0.361
0.104
0.146
0.385
0.338
0.166
0.145
,0.066
0.003
-0.068
b
0.097
0.376
0,361
0.101
0.218
0.334
0.240
0.109
0.148
0.199
0.042
0.237
-0.284
0.3oi
0.258
-0.060
0.019
0.403
0.160
0.010
0.288
0.317
-0.243
0.294
0.036
0.034
0.130
0.169
0.421
0.411
-0,137
continued
5-lu
-------
TABLE 5-3 (continued)
Incoming
Wind
Direction
195-?04
205-214
215-224
225-234
235-244
245-254
255-?64
> 265
Screen
Configura-
tion
2-3.0-0.5
2-3.0-1.0
3-3.0-0.5
3-1.5-1.0
2-1.5-1.25
2-1.5-1.0
NONE
• NONE
1-1.5-0.5
1-3.0-0.5
1-1.5-1.25
2-3.0-1.25
2-1.5-0.5
2-3.0-0.5
2-3.0-1.0
3-1.5-1.0
NONE
1-1.5-0.5
1-1.5-1.25
2-3.0-1.25
3-1.5-0.5
2-3..0-0.5
NONE
2-1.5-1.0
2-1.5-1.25
2-3.0-0.5
NONE
2-1.5-1.0
2-1.5-1.0
2-3.0-0.5
2-1.5-1.0
2-3.0-0.5
2-1.5-1.0
2-1.5-1.25
2-3.0-0.5
No.
Pts.
8
6
6
13
4
11
6
4
9
6
2
3
9
4
2
7
14
4
2
Position
2
0.274
0.563
0.291
0.494
0.688
0.638
0.108
0.049
0.252
0.217
0.677
0.463
0.397
0.311
0.592
0.495
0.031
0.100
0.664
3 ! 0.365
1 i 0.427
1
10
5
2
4
2
1
1
1
2
3
7
1
11
0.232
0.012
0.066
0.638
0.183
0.093
0.053
0.025
0.242
0.081
0.096
-0.379
0.537
0.311
No.
Pts'.
8
6
6
13
4
11
6
4
9
'6
2
3
9
4
2
7
14
4
0
i..
3
1
1
1C
5
2
4
2
1
1
1
2
3
7
1
11
Position
3
0.149
0.632
'0.233
0.419
0.509
0.543
0.005
0.012 '
0.413
0.361
0.667
0.554
0.334
0.126
0.672
0.477
0.020
0.35,6
C.693
0.504
0.416
0.078
-0.007
0.145
0.658
-0:020
0.104
0.387
0.049
0.017
0.212
0.041
0.166
0.3^4 •
o.no
No.
Pts.
8
6
6
13
4
11
6
4
9
6
2
3
9
4
*^
L.
7
14
4
2
3
1
1
10
5
2
4
2
1
1.
1
2
3
7
1
11
Position
4
-1.266 .
-0.050
-0.244
-0.145
-0.016
0.018
-0.231
-0.760
-2.071
-0.097
0.952
-0.685
-0.774
-1.030
0.005
-0.251
-2.297
-0.809
0.933
-0.780
-0.289
-0.383
-2.827
-0.253
0.0
-0.019
-1.416
0.147
0.272
0.357
0.555
0.256
0.360
-0.031
0.160
No.
Pts.
8
6
6
13
4
11
6
• 4
9
6
2
3
9
4
2
7
14
4
2
3
1
1
10
5
2
4
2
1
1
1
2
3
7
1
11
Position
5
-0.098
0.428
0.106
0.185
0.323
0.389
-0.095
0.007
0.051
0.074
0.381
0.059
-0.165
-0.189
0.499
0.206
-0.064
-0.057
0.505
-0.201
0.194
-0.198
-0.133
-0.194
0.642
-0.46
-0.027
-0.364
-0.135
-0.055
0.141
-0.141
-0.150
-0.060
-0 035
5-11
-------
A portion of'the data shown in Table 5-3 were refonr.axed in Table 5-4 to
indicate the maximum windscreen wind reduction by incoming windspeed and
screen configuration. These data are also shown in Figures 5-3 through 5-5.
In interpretation of these data, the following can be said:
0 Efficiencies were greater when winds were near perpendicular to the
screen. This is true for all screen heights and screen widths.
0 Maximum efficiencies D W H* Incoming U'D
1.25 H height .759 2/1.5/1.25 170°
l.OC H height .672 2/3.0/1.00 210°
0.50 H height .430 2/1.5/0.50 180°
* D = distance, W = width, H = height
c The 2 pile height distance was the most efficient.
0 At rea~ perpendicular winds, the 1.5 diameter width screen was the most
effective. For aperpendicular wind direction, 3.0 D screen width was
more efficient than 1.5 D screen width.
0 The 1.25 pile height screen was the most effective.
0 The location of maximum efficiency was almost always position 2, except
where winds were >±35° from perpendicular.
" In the lee (backside) of the pile, the windscreen produced a negative
control efficiency that was often significant. This was evident at
position 4 when the wind direction was >180°, and at position 5 when
winds were greater than 205°.
0 In comparison of the field work to the Billman (1985) work:
Only one windscreen efficiency can be compared because configura-
tions were rot identical.
5-12
-------
TABLE 5-4. MAXIMUM MEASURED WINDSCREEN REDUCTION FACTOR
[1 - (u/u )] = SCREEN EFFICIENCY
Incoming WO = <105
Position:
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Position
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Position
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Position
Length:
1.5 3.0 5.0 1.5 3.0 5.0 1.5 3.0
.265O/4)1
WO = 115-124
: 1 2 3
1.5 3.0 5.0 1.5 3.0 5.0 1.5 3.0
.220(4)
WD = 125-134
: 1 2 3
1.5 3.0 5.0 1.5 3.C 5.0 1.5 3.0
.125(4)
.300(4)
WD = 135-144
: 1 2 3
1.5 3.0 5.0 1.5 3.0 5.0 1.5 3.0
5.0
5.0
5.0
5.0
Height:
0.50 .
1.00 .016(2) .074(2)- .358(2)
1.25 .252(4) .433(2)
1.50
Pile position with highest windspeed reduction
continued
5-13
-------
TA'BLE 5-4 (continued)
Incoming WD = 145-154
Posi t ion
Lenoth:
Heiylit:
0.50
1.00
1.25
1.50
Incoming
Position
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Position
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Posi tion
Length: .
Height:
0.5C
1.00
1.25
1.50
Incoming
Position
Length:
: 1
1.5 3.0
.333(3)
.337(2) .365(3)
WD = 155-164
: 1.
. 1.5 3.0
.331(3)
.286(2) .285(2)
WD = 165-174
: 1
1.5 3.0
.302(3) .504(2)
.478(2) .508(3)
WO = 175-184
: 1
1.5 3.0
.421(3) .443(3}
WD = 185-194
: 1
1.5 3.0
5.0 1.5 .
.077(2)
.373(2)
5.0 1.5
.283(2)
.174(2)
.442(2)
5.0 1.5
.329(2)
.343(2)
.759(2)
5.0 1.5
.430(2)
.416(2)
.645(2)
5.0 1.5
2
3.0 5 ..0
.329(2)
.579(3)
2
3.0 5.0
.455(2)
.408(2)
2
3.0 5.0
.471(2) .479(2)
.456(2)
2
3.0 5.0
.265(2)
.516(2) .496(3)
.601(3)
2
3.0 5.0
3
1.5 3.0
• .151(2)
3
1.5 3.0
.112(2)
3
1.5 3.0
.238(2^
.465(2)
.655(2)
3
1.5 3.0
.232(2)
.525(2)
3
1.5 3.0
5.0
5.0
.462(2)
5.0
5.0
.635(2)
5.0
Height:
0.50 .230(2)
1.00 .499(2) .630(3)
1.25 .525(3)
1.50
continued
.399(2) .246(3)
.561(2) .578(3)
.685(2) .513(2)
5-14
.306(2)
375(2) .594(2)
-------
TABLE 5-4 (continued)
Incoming WD = 195-2C4
Position
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Position
Length:
Height:
0.50
1.00
1.25
1.50
Incoming
Positior,
Length:
Height:
0.50
1.00
1.25'
1.50
Incoming
Position
Length:
Heiqht:
C.50
1.00
1.25
1.50
Incoming
Position
Length:
Height:
0.5C
!.CO
1.25
1.50
1.5
.346(3)
.517(3)
.62€(2)
1 ,
3.0 5.0
.365(3)
.656(3)
...
1.5
.280(2)
.638(2)
.688(2)
2
3.0
.274(2)
.632(3)
3
5.0 1.5' 3.0 5.0
.291(2)
-.494(2)
WD, = 205-214
;
1.5
.413(3)
.677(2)
1
3.0 5.0
.361(3) .
1.5
.397(2)
2
3.0
.311(2)
.672(3)
.554(3)
3
5.0 1.5 3.0 5.0
!*95(2)
W) - 215-224
:
,.5
.356(3)
.693(3)
1
3.0 5.0
1.5
7
(_
3.0
.504(3)
3
5.0 1.5 3.0 , 5.0
WD = 225-234
:
1.5
i
3.0 5.0
1.5
.145(3)
.658(3)
2
3.0
.183(2)
3
5.0 1.5 3.0 5.0
WO = 255-264
;
1.5
1
3.0 5.0
1.5
2
3.0
.256(4}
3
5.0 1.5 3.0 5.0
continued
5-15
-------
TABLE 5-4 (continued)
Incoming WD = >265
Position: 1 2 3
Length: 1.5 3:0 5.0 U> ' 3.0 M LI 3.0 5.0
Htight:
0.50 .311(2}
1.00 .360(4) ,
1,25
1.50
5-16
-------
0.3
OJ
OJ
i 0.6
5
> 0.5
c.
o
5 0.4
u
3
•o
(U
0.3
(U
-------
0.8
a 0.6
3 0.5
•o
c
r. 0.4
u
3
•o
0.3
4)
<->
n o
0.2
0.1
1.25H
l.OH
I
I
100 120 140 160 180 200 220 240 260
Degrees
Figure 5-4. Windscreen control efficiencies for
Position 2H by screen length and screen height.
5-10
-------
0.8
0.7
0.6
J°-5
o
'Z 0.4
•o
O)
*~ 0.3
c
O)
O)
S 0.2
•o
c
0.1
1
1.25H
0.5H...
0.5H
1
1
1
1
1
1
100 120 140
160 180 200
Degrees
220 240 260
Figure 5-5. Uindscreen control efficiencies for
°osi:cion 3H by screen length and screen height.
5-19
-------
Isotach diagrams similar to Billmans' wdrk could not be constructed
because only four data points were available.
Like Billmans findings, tfie taller windscreen was more effecti/e.
Billman found a three pile diameter distance to be more effective
than d one pile diameter distance. A two pile diameter distance was
not tested. The field study indicated the two pile diameter dis-
tance to be the most effective.
Billman found a 1.5 diameter length screen to be more effective than
a 1.0 diameter screen. The field study indicated that the 1.5
diameter length screen produced a higher reduction than a wider
screen when winds were perpendicular, but not when they were aper-
pendicular.
Billmar recorded seme negative screen efficiencies in the lee of the
pile, but to a much lesser extent than the field data indicated.
Although difficult to compare exactly, the wind tunnel appears to
have produced efficiencies 10 to 20 percent higher than the field
date on the front of the pile at position 2. In the one directly
comparable test, wind tunnel values were about 40 percent higher
than field data.
5-20
-------
SECTION 6.0 OBJECTIVE 2—COMPARISON OF WINDSPEED REDUCTIONS
AND PARTICIPATE CONTROL EFFICIENCIES
6.1 RAK-1 PARTICL'LATE DATA
The first comparison of windspeed reductions versus particulate control
efficiencies caused by the windscreen was with the KAM-1 data because of the
potentially large data set generated by the RAM-1 samplers and because these
samplers were essentially collocated on the piles with the windspeed sensors.
The other available particulate data fcr windspeed versus particulate reduc-
tions were the profiling data, which were generated such that only a single
control efficiency value'was available for each test period and it. applied to
the entire pile rather than a specific location on the pile's surface.
RAM-1 data collection was suspended after Test 18 because of continued
difficulty getting the instruments to work properly. A post-test quality
assurance check of data for the first 18 tests indicated a sharp reduction in
quality of data after Test 11 when the readout scale was changed (from 0-20
n,g/m3 to 0-2 mg/m3) in an attempt, to measure the very low particulate concen-
tration?, associated with periods of low windspeed. Data for Tests 12 trough
18 were disqualified as a result of this QA review.
For Tests 1 through 11, collocated windspeed and RAM-1 data were avail-
able in 5 minute increments fcr two different positions on each pile. These
positions, labeled 2 and 3 in Figure 3-3, corresponded to expected locations
of maximum wind erosion on the faces of the piles. However, average windspeed
6-1
-------
reductions were first compared with corresponding particulate re
-------
TABLE 6-1. EVALUATION OF AVERAGE WINDSPEED REDUCTION
AND PARTICULATE REDUCTION (PAM-1) BY TEST
Test
<
lBa
2
3
4
5
6
7
8
9
10
11
Time
period
1300-1340
1415-1515
1130-1230
1305-1405
1200-1240
1400-1500
1515-1600
1700-1800
1800-1900
1005-1100
1515-1600
1500-1555
Wind
dir., °
217
227
204
213
203
194
190
186
161
192
163
206
Windspeed reduction
Location
2
.02
.04
.36
.37 '
.36
.36
.50
.44
.36
.62
.30
.64
Location
3
.03
.06
.21
.16 '
.21
.?.?.
.44
.40
.30
.50
.36
.65
Participate
reduction
Location
2
.50
.57
-.78
'.75
.52
.71
-2.50
-9.25
-7.00
-.63
-8.00
-1.11
Location
3
.50
.39
.77
-.24
.30
ND
-1.42
-.50
-2.20
-.86
.82
-9.29
No windscreen for this test. Reductions at both locations should be C.
6-3
-------
TABLE 6-2 UPWIND RAM-1 CONCENTRATIONS
Test
IA
IP
2
3
4
j>
6
7
8
9
10
11
Upwind
CPP.C. , vg/m3
202
140
4?
40
45
28
11
29
20
26
28
20
Windspeed
ft/KiiP
170^
1745
!582
1777
1316
1383
1044
1040
C19
1066
be«
982
mph
19.4
19.8
17.8
20.2
15.0
15.7
11. P
11.8
9.3
12.1
6.4
11.2
6-4
-------
RAM-1 could contribute to this problem. .Regardless of the cause of the
problem, it was concluded -that the RAM-1 data could not be used in the
analysis.
6.2 EXPOSURE PROFILER DATA
6.2.1 Total Participate
The next step wos to compare overage windspeed reductions with panicu-
late emission reductions for entire tests as measured by the exposure pro-
^ilers. There were 42 valid'tests in the data set. Two alternative sets of
windspeed values could potentially be the most appropriate to estimate overall
pile windspeed reduction, depending on incoming wind direction during the
test. Location ? (see Figure 3-3} was nec.r the crest of the pile in the area
of maxin;utr. erosion; location 3 was about inidway between the crest «nvel of 0.015, ulope of C.P^l, ^nd
y-intercrpt of -0.15C. This indicated a significant relationship between the
two variables that was nearly one to one. The data pairs for this regression
flrp plotted in Figure 6-1.
If the linear regression was forced through zero (no wirc'speed rec:.rtion
results ir no TP reduction), the correlation improved to 0.417, R2 was 0.174,
6-5
-------
TABLE 6-3. COMPARISON OF WINDSPEED AND TOTAL PARTICIPATE REDUCTIONS
Test
IB
o
3
4
5
6
•^
/
9
'10
li
12
138
14
15A
15B
16
17
J9A
i9B
20
21A
21B
22
21A
23C
24
T5A
25B
26
27
res
29
3C/»
30B
31
32
33
34A
3"P
35
36
105
Wind
dir."
228
204
213
2C3
194
190
186
192
163
206
166
209
212
204
203
201
187
200
102
208
196
179
172
169
157
157
190
178
196
203
203
187
185
180
173
16C
186
170
168
162
167
166.
Best
sensor
2
2
2 .
2
2, 3
2, 3
2, 3
2, 3
2, 3
2
2, 3
2
2
2
/-»
?
2, 3
2
2, 3
2
2, 3
2, 3
2, 3
2, 3
2
2
2, 3
-2, 3
?a- 3
2a
?
vd
2
2
£
c
2
2
o
^
•5
.1
2
Wspd, ft/mi n
screen
1679
951
1039
791
1128, 810
630, 460
682, 515
5t2, 364
37?, 329
312
310, 204
198 •
477
661
641
493
446, 327
256
398, 301
343
473, 33?
556, 432
587, 438
645, 479
230
426
771, 658
744, 616
950, 854
577
424
375
410
433
313
305
174
214
198
310
256
408
, exposed •
1755
1480
1653
1228
1451, 1257
1129, 920
1134, 918
1174, 966
578, 469
869
46?, 356
314
1018
1020
679 .
580
583, 433
294
456, 320
342
1102, 884
1119, 851
1173, 846
1201, 861
380
455
1084, P38
948, 693
1091, 882
1009
721
609
6fiO
620
545
356
289
275
272
410
380
389
Wspd
rdn.
.043
.357
.37]
.356
.289
.471
.419
.564
.327
.641
.378
.369
.531
.352
.056
.150
.240
.122
.093
-.003
.598
.498
.491
.453
.395
.064
.252
.163
.080
.428
.412
.384
.384
.302
.426
.143
.398
.222
.272
.244
.326
-.04?
T. Partic. , Ib/m
screen
9.82 .
1.45
2.92
1.04
1.26
.54
.46
.38
.02
.02
.03
.01
.04
.03
. .03
.01
.03
.03
.01
.01
.04
.06
.07
.20
.02
.02
.29
.15
.04
.17
.09
.01
.04
.02
.01
.01
.04
.01
.01
.01
0
0
exposed
8.36
1.34
2.34
.98
2.52
.66
.49
.47
.04
.04
.03
.01
.06
.02
.03
.01
.08
.02
.01
.01
.04
.34
.91
2.15
.04
.03
. .19
.10
.03
.12
.13
.02
.03
.03
.01
.01
.03
.01
.01
.01
0
0
T.
Partic.
rdn.
-.175
-.082
-.250
-.058
.490
.180
.051
.197
.453
.471
-.061
.371
.394
-.188
.101
.029
.588
-.596
-.111
.180
.108
.825
.928
.908
.486
.192
-.509
-.555
-.431
-.403
.290
.358
-.599
.301
.357
.063
-.147
.391
-.071
.218 .
.000
.000
Peginning with Test 27, the sensor at location 3 on the screened pile oegan
giving erroneous readings. Therefore, only data from, location 2 were used for
subsequent tests.
6-6
-------
Figure 6-1. W'.'ndspeed Reduction Versus Particulate Reduction
6-7
-------
significance level was 0.005, and t.he slope was 0.466. This relationship was
more significant than the one above aii*j. indicated TP reductions slightly less
than half of the corresponding WS reductions.
There were 15 negative TP reductions in the 42 tests'(see Figure 6-1). A
negative TP reduction meant that a higher emission rate was measured on the
screened pile than en the exposed pile. Many of these tests in which the
screen appeared to increase emissions could be the result cf differences that
were less than the measurement error for exposure profiling, but several of
the differences in emission rate were large enough that these measurements
probably reflected real occurrences of increased emissions. This observation
agreed with the findings in Section 5 and in the Billmnn study that the
windscreen actually produced increased windspeeds en the lee side of the pile
during several tests.
A preliminary examination of windscreen parameters (distance, height,
length) did not isolate any specific design parameter closely associated with
the negative TP reductions. This topic is examined in greater detail in
Section 7.
If the negative TP reductions were assumed to be due to measurement
errors and set at zero, the correlation was virtually unchanged at C.404, R2
was C.163,'significance level was C.C10, the slope was 0.626, and the
y-intercept was C.175.
It was observed that many of the tests were taken at windspeeds too low
to cause wind erosion and during other tests the moisture content of the pile
was so high that erosion would not occur even with high windspeeds. Twenty of
these tests with negligible emissions were eliminated to see whether the
windspeed-emission rate relationship was stronger during tests with wind
6-8
-------
erosion. The results of .this regression analysis were: R - 0.287, R2 =
O.C82, significance level = 0.184, slope = 0.859, and y-intercept = -0.122.
According to this analysis, the tests'with negligible emissions did not appear
to be distorting the calculated relationship between the two variables.
If the relation between WS and TP reductions is not linear, the correla-
tion between these variables would be improved by transforming one of the
variables to its natural logarithm (In) form. When US was In-transformed, the
results were:. R = 0.36?, R2 = 0.131, and significance level = 0.022. This
result indicated that WS 'and TP reductions were approximately linearly related
rather than to some power.
From the above regression analyses, it can be concluded that a highly
significant relationship exists between windspeed and particulate emission
reductions, and that the relationship is approximately 'linear with a slope
less thcin^ ere (one percent reduction in windspeed results in less than one
b
percent reduction in particulate .emissions). Also, .there appear to be in-
stances in which windspeed on the front face of the pile is reduced by the
windscreen but emissions from the pile actually increase as a result of higher
windspteds on the back side of the pile. In general, windscreen configuration
should not affect the relationship between windspeed and emission reduction.
6.'2.2 Particle Size Data
The emission rates of particles within several size ranges were deter-
mined by selecting a heavily loaded filter from each profiling test and
subjecting it to laser diffraction analysis. The resulting percentages of net
sample weight by particle size range were then multiplied by the TP emission
rate for the test to get emission rate by particle size range. The effective-
ness of the windscreen in reduring emissions for each size range was then
6-9
-------
calculated as (1 - ER /ER ) for that range, as shown in Table 6-4.
Next, the participate reductions by size range (from the last five
•
columns on Table 6-4) were compared with corresponding windspeed reductions in
the same manner as with the TP reduction data, i.e., regression analysis. The
results of these analyses are summarized in Table 6-5. The entire sequence of
regressions performed for TP was not repeated for the size fractions. The
regular binary regression and the regression forced through zero were done for
each size range; but removal of negative emission reductions, removal of tests
with negligible emissions, and In transformation were not performed.
None of the particle size ranges had as significant a relationship with
windspeed as TP did, partially because of the smaller data sets available with
the particle size data (19 tests instead of 42). Particle size data could not
be obtained on many filters because of their light mass loadings. The parti-
cle size emission reductions that had the highest correlations with windspeed
reduction were 30-62 urn and 62-176 um. The two small particle size ranges
both had poor correlations. Slopes of regression lines for the two particle
sizes that were reasonably significant were higher than slopes for TP, indi-
cating an emission reduction (in those size ranges) almost equal to windspeed
reduction. No explanation was apparent for differences in variation with
windspeed for the different particle sizes.
The s?me frequent negative emission reductions were observed in all
particle size ranges as for TP (see Table 6-5). This is a good indication
that increased emissions were actually occurring as a result of the screen,
rather than anomalous results from sampling or lab analysis errors. The
negative reductions did not all occur in the same tests, or in the tests that
had neaative TP reductions.
6-10
-------
TABLE 6-4. CALCULATION OF EMISSION RATES BY PARTICLE SIZE RANGE
Test
18
IE
2
2
3
3
4
4
5
5
6
6
7
7
9
9
14
14
17
17
?1A
21 A
21B
21B
22
22
Pile
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
Sample wt. by particle size, *
175
24.0
31.4
13.6
9.2
11.1
22.8
17.4
18.1
28.2
21.2
35.6
34.2
22.1
20.3
29.8
27.1
40.4
37.2
47.6
16.6
37.2
17.2
40.2
22.5
43.2
Em. rate by particle size,
Ib/meter
<10
.10
.11
.03
.05
.04
.04
.04
.02
.02
.03
.02
.01
.01
.01
.01
.02
0
0
0
0
0
.01
0
.03
0
10-30
.70
.79
.12
.23
.32
.18
.07
.07
.21
.10
.03
.01
.02
.01
.01
.03
0
0
.01
0
0
.03
0
.06
0
30-62
2.13
1.P6
.32
.38
.63
.72
.27
.26
.49
.29
.10
.04
.05
.05
.07
.05
.01
.01
.01
.01
.01
.06
.01
.23
.01
62-176
3.42
3.99
.70
.67
1.09
1.31
.43
.50
1.09
.57
.27
.31
.30
.29
.24
.18
.02
.02
.02
.03
.02
.18
.03
.39
.03
>176
2.01
3.08
.18
A3
.26
.67
117
.19
.71
.27
.24
.18
.11
.09
.14
.10
.03
.01
.04
.01
.02
.06
.02
.20
.03
Pdr by screen by size
<10
-.10
-.67
0
.50
-.50
.-50
0
-1.0
0
0
1.00
1.00
10-30
-.13
-.92
- .44
0
.52
.67
.50
-2.0
0
o .
1.00
1.00
30-62
.13
-.19
-.14.
.04
.41
.60
0
.29
0
0
.63
.96
62-176
-.17
.04-
-.20
-.16
.48
-.15
.03
.25
0
.33
.83
.92
>176
-.53
.28
-J.58
-.12
.66
.25
.18
.29
.67
-1.0
.67
.85
-------
TABLE 6-4 (continued)
Test.
23A
23A
23B
23B
25 A
25A
2SB
25B
27
27
28B
28B
30A
30A
12
12
26
26
Pile
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
E
S
Sample wt. by particle size, %
<10
1.3
2.5
0
4.6
3.6
2.1
5.2
0.5
1.6
0.3
2.0
2.1
0.4
0.2
1,0
2.0
0.7
10-30
7.7
4.5
3.8
9./
8.3
2.3
7.1
2.3
2.5
4.7
7.1
0.2
7.5
0.7
2.1
1.3
1.8
30-62
22.7
9.9
16.3
11'. 5
10.8
17.5
11.9
16. &
14.4
16.1
10.0
13.8
11.6
12.6
8.8
22.0
15.3
62-175
39.1
42.4
49.3
37.2
50.6
59.9
58.0
69.5
51.7
43.2
54.4
48.5
47.2
49.9
35.9
62:0
63.4
>175
29.2
40.7
.30.6
36.0
26.7
18.2
18.8
10.9
29.8
35.7
26.5
35.4
33.3
36.6
52.2
12.7
18.8
Em. rate by particlf size,
Ib/meter
<10
.03
.01
0
.01
.01
0
' .01
0
0
0
0
0
0
0
0
0
0
10-30
.17
.01
0
.02
.i;2
C
.01-
.01
.01
.01
.01
0
0
0
0
0
0
30-62
.49
.02
.01
.0?
.03
.02
.02
.02
.02
.02
.01
.01
.01
.01
0
.01
.01
62-176
.84
.08
.02
.07
.15
.06
.09
.08
.09
.06
.05
.01
.02
.01
.01
.02
.02
>176
.62
.08
.01
.07
.08
.02
.03
.01
.05
.04
.02
.01
r *
t
.01
.02
0
.01
Rdn by screen by r.i/e
<1C
.67
C' -
— oo
0
C
0
0
10-30
.94
0 '
— oo
0
0
0
0
.
30-62
.96
-.50
0
0
.50
0
1.0
62.176
.90
-1.J4
-.50
-.13
.17
-1.0
0
>176
.87
-.14
-.50
-5.00
.50
0
-1.0
Note: E - exposed (no windscreen), S = screened
-------
TABLE 6-5. IMPACT OF WINDSPEED REDUCTIONS ON PARTICLE
SIZE EMISSION REDUCTIONS
Dependent
variable
Emission rdn. in
size range:
<10 um
10-30 um
30-62 um
62-176 um
>176 um
TP
Regression line
forced through
zero
<10 urn
10-30 um
30-62 um
•62-176 um
>176 um
TP
Regression against windspeed reduction
R
.166
-.018
.301
.426
.269
.372
.194
.147
.572
.263
.075
.417
R2
.028
.000
.091
.181
.072
.138
.038
.022
.327
.069
.006
.174
S.ignif.
level
.510
.945
.210'
.069
.266
.015
.425
.547
.008
.263
.755
.005
Slope
.674
-.099
.971
1.860
1.368
.841
.233
.245
.688
.370
.121
.466
y-int.
-.195
.152
-.124
-.651
-.545
-.150
0
0
0
0
0
0
6-13
-------
6.3 FACTORS OTHER THAN WINDSCREEN AFFECTING EMISSION RATES
6.3.1 Total Participate
The approach used to identify external variables that affected emission
rates from the piles was multiple linear regression (MLR) analysis. Only test
data from the unscreened pile were utilized, since these emission rates were
not altered by the presence of the windscreen.
The variables included in the analysis were windspeed, moisture content,
and silt content of surface material. Particle size was also a variable in
that MLR analyses were run with four different sets of emission rates
(<10 urn, <30 urn, <62 urn, and TP) as dependent variables to examine the effects
of the external variables on different size ranges of particles. Analyses
with size fractions are presented in the next subsection of this report. Kind
direction was not considered to he a variable because the sampler inlets were
pointed directly into the wind on each test and the profiling tower was placed
in approximately the same location downwind of the pile.
Initially, all 42 tests with exposure profiling data were included in the
MLR run. As shown in Table 6-6, :-*ie multiple-R2 with all three variables was
0.481 and the MLR equation was significant at the 0.000 level. In other
words, the| three specified variables explained 48.1 percent of the variation
in tP emission rates arid the probability that the three variables were not
related to emission rate was less than 0.05 percent.
From the initial run, windspeed and moisture content appeared to be
highly significant variables, but silt content was marginal. Silt had a
negative coefficient, indicating a decrease in emissions with increased fine
6-14
-------
TABLE 6-6. MLR RUNS TO IDENTIFY VARIABLES THAT AFFECT EMISSION RATES
Description
of MLR Run
1. Initial
2. Eliminate low
WS and high
MC tests
3. Eliminate
silt
4. In-transformec
variables
5. In-trans.'ormed
6. In-transformed
back to full
data set
No. of
Tests
42
22
22
22
22
42
Indep.
Variable
wndspd
moist
silt
wndspd
moist
silt.
wndspd
mcist
wndspd
moist
silt
wndspd
moist
wndspd
moist
Simple
Correl .
.614
-.352
-.312
.719
-.677
-.371
.719
-.677
.629
-.871
-.405
.629
-.871
.486
-.730 .
Partial
Correl .
.535
-.292
-.100
.284
-.204
-.103
.278
-.242
.032
-.569
-.116
.030
-.604
.387
-.668
Signif. of
var. in egn.
.000
.018
.018
.083
.205
.514
.005
.008
.84f.
.000
.629
.785
.000
.000
.000
Coeff.
0.002
-0.401
-0.061
0.002
-1.261
-0.080
0.004
0,081
0.158
-2.067
-0.489
.214
-2.108
1.118
-1.346
Inter-
cept
0.091
0.812
-2.537
1.068
-0.4C6
-6.572
Mult.
R2
.481
-.570
.507
.763
.760
.663
Adjusted
R?
.439
.498
.458
.726
.736
.667
Signif.
of eqn
."00
.002
.001
.000
.000
.000
o>
-------
particles in the pile, and a significance of 0.4C2. Also, when the variables
were placed in the equation stepwise, the root mean square error (RKSE) term
increased when, silt was adced as the third variable.
•
As part of the regression analysis, each of the independent variables was
plotted against TP emission rate. The resulting graphs are shown in Figures
6-2 through G-4. It was observed that tests with windspeeds less than about
900 ft/nnn (10.2 mph) and tests with moisture contents greater than abcot 1.5
percent all had very low emission rates.
Since these appeared to be thresholds beyond which the variable no longer
reduced emission rates, some of the test data in these insensitive zones were
eliminated from the data set to bee if correlation could be improved for the
smaller data set. The cut points arbitrarily used we,re 500 ft/min (5.7 mph}
and 2.0 percent moisture content (M-C). This eliminated 20 of the &•? tests.
None of the eliminated tests had a TP emission rate greater than 0.08
Ib/meter.
The MLR results for the smaller data set (run 2) were slightly better,
with a multiple-R2 of 0.570 and a significance level of 0.002 (see Table 6-6).
Again, the silt variable was not highly significant and had a negative coef-
ficient. It should probably not be included in an MLR predictive equation.
Without the silt variable, multiple-R2 was 0.507 and the significance level
improved to O.OOi. The MR equation was: ER = 0.48 WS - 2.31 MC - 2.537
Ib/meter.
Most current emission factors for fugitive dust sources adjust for the
effect of external variables by use of multiplicative correction parameters.
For instance, the most widely used emission factor equation for unpaved roads
is:
6-16
-------
Figure 5-2. Scatter Plot of Emission Rate vs. Windspeed.
6-17
-------
Figure 6-3. Scatter Plot of Emission Rate
vs. Moisture Content
6-18
-------
• -. • = c-r : •:•:•: .T, ; - :,Ti'j.Ti
. r.'. =. >.;••:•.•* --r. . .T...-T;
Figure 6-4. Scatter Plot of Emission Rate
vs. Silt Content
6-19
-------
E = 5.9 (kM-M
in which k, s, S, W, w, and P are correction parameters to adjust for signifi-
cant variables such as silt content, vehicle speed, vehicle weight, and days
per year with measurable precipitation. In order for a multiple regression
equation for storage pile emissions to be in a mul tipl icetive form, the
variables to be included must be transformed to their natural logarithms (In)
and the MLR rerun.
When the windspeed and moisture data were transformed, the resulting MLR
results (with the smaller data set) were greatly improved, as shown in Table
•H
6-6. The multiple-R? was 0.760 and the significance level was 0.000. The
individual variables were significant at the 0.784 level for WS (not signifi-
cant) and at the 0.000 level for MC.
The regression coefficients of the transformed WS and MC terms were Or214
and -2.108, respectively, and the best fit equation was:
ER = 0.0078 (WS)°'21(MC)"2-1
The -2.1 power agrees quite well with exponents in other fugitive dust emis-
sion factor equations, which generally range from -1.3 to -2.0. The 0.21
power for windspeed, however, indicated that emissions were much less than
lineraly related to windspeed, in contrast to the cubic relationship cited in
Section 2 of this report. WS was not significant in the equation, so it is
net like'iy that its exponent is meaningful.
If the entire data set of 42 tests were used with In transformation, the
percent of variation explained (multiple-R2) was lower at 0.683, but the
significance level of the overall ecuation and the two individual variables
were 0.000. With the addition of the data points that were insensitive to
6-20
-------
changes in windspeed and/or moisture content, the coefficients became 1.118
for WS and -1.346 for MC and the best fit equation was:
Efi = O.Q0012 (WS)1-12(MC)"!'35
The exponents in this equation agree better with previous published work, and
this equation is mere stable over the entire range of WS and MC values tested.
Silt data were input as their In-transformed values into the above two
MLR runs. Results were essentially the same as with the untransformed data:
the In silt variable did not increase the Multiple-R2, it had a relatively low
simple correlation with emission rate, it had about a 0.40 significance level,
and it varied inversely with emission rate.
An independent statistical review of the data was performed and ib
reported in Appendix A. Their analyses indicated an exponent of 2.53 for WS
which agrees well with previous fugitive dust control research which suggests
an exponent of 3.
6.3.2 Particle Size Data
The MLR analysis described above for total particulate was also done for
emission rates in the particle size ranges of <10, <30, and <62 ym. Emission
rate data by size range, obtained by laser diffraction of the filters were
previously presented in Table 6-4. The whole sequence of MLR runs was not
repeated, just those runs with In-transformed variables and the full data set.
However, the In silt variable was included in one run for each size range to
determine whether silt in the pile had an effect on emission rates of smaller
particles. The results of the particle size MLR analyses are summarized in
Table 6-7.
The multiple-R2 were greatest for the smallest size range and consis-
tently decreased with larger particle sizes. This result could be interpreted
6-21
-------
TABLE 6-7. MLR RUNS WITH EMISSION DATA BY PARTICLE SIZE RANGE"
Dependent
variable
Total participate
Total participate
<10 pm
<10 pm
<30 pm
<30 pm
<62 pm
<62 pm
Indep.
variable
wndspd
moist
silt
wndspd
moist
wndspd
moist
silt
wndspu
moist
wndspd
moist
silt
wndspd
moist
wndspd
moist
silt
wndspd
moist
Simple
correl .
.486
-.730
-.324
.486
-.730
.646
-.831
-.462
.646
-.831
.592
-.811
-.499
.592
-.811
.560
-.821
-.476
.560
-.821
Partial
correl .
.373
-.636
-.021
.387
-.668
.258
-.552
- .088
.285
-.596
.196
-.541
-.150
.235
-.602
.159
-.574
-.130
.193
-.631
Signif. of
var. in eqn.
.000
.000
.824
.000
.000
.031
.000
.^37
.018
.COO
.124
.000
.233
.071
.000
.212
.000
.305
.134
.000
Coeff.
(expon. )
1.104
-1.334
-.141
1.118
-1.346
.019
-.019
-.011
.021
-.019
.115
-.143
-.143
.135
-.153
.331
-.537
-.440
.391
-.569
Mult.
R2 .
.684
.683
.780
,772-
.735
.713
.728
.711
Adjusted
R?
.658
.667
.743
.748
.691
.683
.683
.681
Signif.
of eqn.
.000
.000
.000
.000
.000 .
.000
.000
.000
en
-------
to mean that windspeed and moisture content have a greater or more predictable
effect on emissions of small particles, which certainly agrees with the
theoretical disc;'r.sions in Section 2.
The significance of the equations for the size fractions was just as high
as for TP (0.000) even though a smaller data set was used. Size data were not
available for some of the tests because very light filter loadings did not
permit the laser diffraction method to be used.
The exponents for WS and MC in the MLR equation were smallest for the
<10 um size range and were consistently larger as the particle size range
increased. An explanation for this is not apparent, but this same variation
has been observed with other fugitive dust emission factor equations that have
been derived separately for different size ranges (e.g., coal mining factors).
Moisture was the variable with the highest simple and partial correlation
in every particle size range. It was also the most significant variable in
every MLR equation. Windspeed was highly significant in the TP and <10 um
runs, but only marginally significant (0.071 to 0.212) for the <30 um and
<62 um runs. Silt was not a significant variable for any of the size ranges
in addition to not being significant in the TP run. This was an unexpected
finding, and indicates that wind erosion rates are relatively i-.dependent of
available silt-sized material near the surface or that the material sampling
procedure does not accurately reflect the size distribution of material
available for erosion.
In summation, two external variables—windspeed and moisture content--
explained an extremely high percent of variance in emission rates from the
unscreened storage pile. Approximately the same relationship was observed
between the two variables and emission rate regardless of the size fraction of
6-23
-------
emissions used in the analysis. Some interrelationship was anticipated,' in
that fractional emission rates were calculated as a percent of total mass flux
rate for each test, providing a fixed upper limit value for the emissions for
each test. The additional information furnished by the particle size MLR runs
was that the size distribution of emissions from test to test had to be quite
consistent in order to achieve the similar MLR results.
Also, several data subsets and additional external variables <.vere exa-
mined in different MLR runs. The relationship between windspeed, moisture
content, and emission rate demonstrated good stability in these different
runs.
6-24
-------
SECTION 7.0 OBJECTIVE 3—DEVELOPMENT OF WINDSCREEN
DESIGN PARAMETERS
7.1 SCREEN iFIGHT, LENGTH, AND DISTANCE FROM PILE
For the three windscreen design parameters that were varied by test, the
field test results should provide better information than wind tunnel data for
optimizing design values. This is because conditions such as wind direction
variation, surface moisture content, and crusting could be incorporated in
field testing but not in wind tunnel studies.
The statistical test employed was stepwise MLR. This procedure identi-
fied which variables had the closest relationship to (and presumably the
greatest effect on) emission rate reduction. The three windscreen varia-
bles—height, length, and distance from pile--were entered along with exo-
genous variables such as windspeed, surface moisture content, and silt con-
tent. Stepwise MLR also removed the effect of one or more variables from the
data set (by modifying values of the dependent variable) so that the effect of
remaining variables could be examined better.
Originally, these analyses were to be performed with the sub-10 um and
sub-30 um data sets. However, the smaller number of tests available (19
versus 42) and the generally good agreement between analyses done, with TP and
those done with specific particle sizes resulted in a change to TP emission
reductions as the dependent variable in all these MLR runs. The full data set
and the subset of 22 tests in which wind erosion losses were probable because
of high winds and dry soil surface were both subjected to the MLR analysis.
7-1
-------
Results of the two stepwise MLR runs are presented in Tat^e 7-1. In both
cases, screen length and screen height were the'twc most significant varia-
bles, in particular, screen length seemed to be highly correlated with
emission reductions. According to the coefficients in the final equations,
each increase of one pile diameter would cause emission reductions to improve
by 13 to 16 percent. The corresponding coefficients when screen length was
regressed alone with TP emission reduction were 22 and 28 percent. Screen
height had much lower correlation and was a less significant variable in the
equations than screen length, but it had a consistent coefficient in the range
of 38 to 45 percent reduction in emissions for each increase of one pile
height in the screen height. Of course, these regression equations are only
applicable over the range of the test data, which was 1.5 to 5 pile diameters
for screen length and 0.5 to 1.25 pile heights for screen height.
As indicated in Table 7-1, the screen-to-pile distance did not appear to
be related to ^mission reductions. The test range of '.0 to 3.0 pile heights
distance was a narrower range than used in previous field testing or in the
wind tunnel work. It was an attempt to "fine tune" this design parameter, but
for MLR analysis i „ resulted in no definitive findings.
None of the exogenous variables had such an overriding effect on relative
emission rates with and without the screen that they obscured the impacts of
changes in screen parameters. By including these variables in the MLR, the
relatively small effects of these variables were taken into account rather
than acting as interferences in the direct comparison of test results.
In the study design, four separate tests with the same combination of
screen length, height, and distance were planned. This would have allowed the
plotting of emission rates as a function of each variable separately (e.g.,
7-2
-------
TABLE 7-1 . STEPWISE MLR TO EVALUATE WINDSCREEN DESIGN PARAMETERS
Description
of run
All 42 tests
22 tests with
high winds
and low
moisture
Indepen.
Variable
Screen length
Screen height
Moisture
Windspeed
Screen distance
Silt
Screen length
Screen height
Windspeed
Moisture
Silt
Screen distance
Simple
correl .
.474
.293 •
.100
.046
.277
.186
.532
.355
.243
.186
.373
.226
Signif .
of var.
in final
eqn.
.016
.171
.240
.312
.512
.823
.089
.301
.383
.515
.603
.777
Coeff.
.13
.38
.07
.0002
.03
.0057
.16
.45
.0004
.19
.04
.02
Mult.
RZ
.225
.242
.260
.274
.283
.283
.284
.305
.316
.332
.342
.345
RMSE
.345
.345
.346
.347
.350
.355
.377
.380
.387
.392
.400
.410
Sign-;f .
of eqn.
.000
.007
.017
.038
.052
.063
.002
.021
.050
.096
.184
.238
7-3
-------
screen height) with the other two parameters being held constant (e.g., only
tests with screen length of 3 diameters and pile distance of two heights).
However, with the shortfall in total tests--100 were planned—not enough data
points were generated to carry out this plan. Curves have been generated
which eliminate one of the other two variables. These are presented as Figure
7-1, 7-2 and 7-3, which look at screen height, screen length, and screen-to-
pile distance, respectively.
From these plots, it did not appear that screen height had an effect on
emission reductions in the range of heights tested. Also, the pronounced
scatter in the data is accentuated by this type of presentation. It is
unfortunate that none of the 1.5-pile-height screen height tests specified in
the study design were conducted. Screen length had an obvious effect on
emissions when tests of 5.0 diameter screen length were included (upper right
graph in Figure 7-2). The advantage of a 3.0-pile-diameter screen over a 1.5-
pile-diameter screen was not apparent in the other two graphs. This could
tentatively lead to a conclusion that a screen length longer than 3.0
diameters is needed to get consistent emission reductions. Given the >40°
variations in wind direction that occurred during some of the one-hour tests
periods, the 5.0-pile-diameter length seems reasonable for a pennanejnt or
semi-permanent installation.
Although the MLR results did not reveal any trend in emission rates with
screen distance from the pile, the plots in Figure 7-3 appear to show that the
2.0-pile-height distance is superior to either the 1.0- or the 3.0-pile-height
distances. Again, the scatter in these data points makes conclusive findings
difficult.
By combining the results of the MLR and graphic analyses, the following
7-4
-------
8C-
6C.
*-4C"
o
3
-O
v/)
v/)
-6C
»
*
.5 1.0 1.5
Screen Ht. •
ALL 1.5L SCREENS
• 1 H DIST.
. 2 H DIST.
* 3 H DIST.
**
•o
o>
6C
Art
4C
d
1/1
in
l.-2
UJ
.5
1.0
Screen Ht.
1.5
Al 1 3L Screens
« 1 H DIST.
" 2 H DIST.
» 3 H DIST.
Figure 7-1. Evaluation of Screen Height.
7-5
-------
IOC
80
6C
.
c
o
1 20
T3
d)
i-
| C
i/l
l/)
|-2C
LU
-4C
-60
8C
6C
** 4C
r 80
0.5 H
*! 60
• §
S 40
O)
• § 20
»• I/-.
1 o
• *
; -20
-40
*
*
•
l.OH ' t
A
•
A
* 1 *
*
A
-
. A , * , ,
012345 "OU01 2345
Screen Length Screen Length
1.25 H
•
:
• 1 H DIST
2C
O)
s-
c C
o ^
5-2q
-4C
-60
* 2 H DIST
• 3 H DIST
—i—f—i—£
Screen Length
Figure 7-2. Evaluation of Screen Length
7-6
-------
60-
40-
c
o
- 2d
•§ d
i/)
s -2d
-4C
-6C
IT
*
2 3
Screen-to-Pile Cist,
ALL 1.5L SCREENS
* 0.5 H
* 1.0 H
« 1.25 H
C
o
60
40
20
TD '
£U
§ o
= -20
-40
-60
U
*
•4-
ALL 3L SCREENS
* 0.5 H
* "1.0 H
» 1.25 H
2
Screen-to-Pile Dist.
Figure 7-3 Evaluation of Screen-to-Pile D stance
7-7
-------
design conclusions could be drawn:
0 Screen length—This parameter does not optimize at a certain length.
In general, the longer the screen (or in the case of a curved
screen, the greater the arc covered) the lower the emissions should
be.
? Screen height—The large reductions in emissions estimated by MLR
(38 to 45 percent per pile height) are not apparent in the plots.
It should be emphasized that the MLR results were not highly
significant for screen height. Screens higher than the pile appear
to be only marginally beneficial.
0 Screen-to-pile distance—Previous studies had shown distances of
1.0- to 3.0-pile-heights to be better than shorter or longer
distances. Within that range, the. 2.0-pile-heights distance may be
1 the optimum distance.
These results confirmed conclusions of Billman's wind tunnel study for
screen height and distance but differed with respect to design of screen
length. The wind tunnel study found that increasing length from 1.0- to 1.5-
pile-diameters caused only slight improvement in screen efficiency. No
greater lengths could be tested in the wind tunnel. Therefore,
recommendations were that the screen should be at least as long as the pile.
With wind direction variations that occur in real situations, such a short
screen does not seem to be an optimum design.
The Billman study found that the 0.5-pile-height windscreens were not as
effective as hiyher screens, but that screens as high a.s the pile (1.0-pile-
height) were nearly as effective as higher ones. This is essentially
identical to the field study findings. In the wind tunnel, the effect of
7-8
-------
screen-to-pile distance was found to be interrelated with screen height. A
screen of 1.0-pile-height placed at a distance of 3.0-pile-heights away from ,
the pile caused greater windspeed reduction' on the Windward face of the pile
but less reduction in the lee of the pile than the same screen at a 1.0-pile-
height distance. The net result -as approximately equivalent reductions at
screen-tc-pile distances OT 1.0- and 3.0-pile-heights. The 2.0-pile-height
distance was not tested in the wind tunnel. In the field, the 1.0- and
3.0-pile-height distances had about the same efficiency, with the
2.0-pile-height distance arguably having slightly greater emission reductions.
7.2 WIND DIRECTION RELATIVE TO SCREEN
•For tests in which the resultant wind direction was not perpendicular to
the windscreen, some reduction in control efficiency would be expected as the
pile might become exposed at the end of the screen. When wind direction
deviation from perpendicular was plotted against total particulate emission
reduction associated with the screen, the line of best fit had a slope of
-0.012 and an intercept of 0.356 (see Figure 7-4). In other words, the
average control efficiency of 35.6 percent for the screen positioned perpen-
dicular to the wind was, reduced by 1.2% for every degree deviation from
perpendicular in the wind direction.
The correlation coefficient for the data in Figure 7-4 was -0.300, and
the probability that the two variables were not related was 0.164. The
correlation was relatively low because of other variables in the tests that
also affected emission reductions—in particular, the length, distance and
height of the windscreen.
The original study design specified that four repetitive tests be taken
for each windscreen configuration (combination of length, distance and
7-9
-------
PART1C. REDUCTION VS. WIND D1R. REL. TO PILE
X < 20 ) by Y < 3 ) = *
+ + ^ + + +
1 + *
I » .1
I •» I
1 I
i * . I
:8 ' 4
i i
l . i
i
i
.0 + •» +
i 1
I * * I
,1 • I
p , . I . I
A .4 « •»
R I * I
7,1* * I
1 • I i
C 'I * I
.2 . » » • - •» i
R I I
O.I * 'I
N I I
! I
0 * +
! « I
I * I
! . I
-.4
I « I
I * I
-. o + •(
t + 4 + + 4
0 10 20 30 40 50
NET 01R, • FROM PERPENDICULAR
Number of points Below minimum (X) : 0 (Y) : 0
Number o-f points above maximum (X) : 0 (Y) : 0
Number o-f points with missing values (X) : 0 (Y) : 19
Number of points plotted : 23
Figure 7-''.. Scat*.-r plot of participate emission reduction
vs wind direction deviation from perpendicular.
7-10
-------
height). However, because of the reduced number of tests completed and the
further reduction caused by tests with negligible emissions (due to low wind
speeds and/or high moisture content), separate curves of emission reduction
versus wind direction could jot be drawn for each configuration. If only
screen length is considered as a variable, the series of curves shown in
Figure 7-5 can be produced. The three curves are almost parallel, indicating
a similar impact on emission races over a wide range of screen lengths. The
slopes of these estimated curves are about -1.2% for a short screen, -1.6% for
a medium screen, and -2.0% for a long screen. The individual curves would not
be highly significant because of the few data points used in their derivation.
Figure 7-5 helps to explain the negative emission reductions, or in-
creased emission rates which were observed for several tests. Generally,
these tests had short sere-ens and oblique winds that may have resulted in
localized turbulence and increased surface erosion compared with conditions on
the unscreened pile. The two outlier data points, for test 25B and 30A, were
attributed to low emission rates on both piles during these tests.
A separate statistical review was performed by Brian Aldershof and David
Ruppert of the Statistics Department of the University of North Carolina at
Chapel Hill. This review,, entitled "Statistical Revision of Field Evaluation
of Windscreens as a Fugitive Uust Control Measure for material storage p'iies
is contained in the appendix.
7-11
-------
1.0
» 22 LG
• 23 A
LG
LG = long
MO = medium
SH = short
Long Screen
(5.0D)
Med. Screen
(3.00)
Short Screen
(l.OD)
IB .
None
-0.4 -
-0.6
25 B SH
* 30 A MO
J.
I
10 20 30 40
Net Direction, ° From Perpendicular
50
Figure 7-5. Scatter plot of participate reduction
vs. wind direction by screen length.
7_ 1 7
-------
REFERENCES '
Bagnold, R..A. 1941. The Physics of Blown Sand and Desert Dunes. Methuen,
London.
Billman, B.J., and S.P.S. Arya. 1985. Windbreak Effectiveness for Storaye
Pile Fugitive Dust Control--A Wind Tunnel Study. EPP-6UU/3-85/U5y (NTIS
PB85-243848). U. S. Environmental Protection Ayency, Kesearch Triangle Park, NC,
Slackwood, T.R. and R. A. Wachter. 1978. Source Assessment: Coal Storaye
Pi'.os. EPA-6UO/*-78-004k (NTIS PB284297;. U.S. Envi ror.mental Protection
Agency, Cincinnati, OH.
Games, D. and O.C. Drehmel. 1082. The Control of Fugitive Emissions Using
Windscreens. In Proceedings: Third Symposium on Transfer and Utilization of
Particulate Control Technology. Volume IV. EPA-6iJO/9-82-OU5d (NTIS PB83-
149617), U.S. Environmental Protection Agency, Research Triangle Park, NC.
Cowherd, C., Jr., R. Bonn, and T. A. Cuscino. 1979. Iron and Steel Plant
Open Source Fugitive Emission Evaluation. EPA-60U/2-79-1U3. (NTIS PB29y38b).
U.S. Environmental Protection Agency, Research Triangle Park, NC.
Cowherd, C., Jr. 1982. Emission Factors for Wind Erosion of Exposed Aggregates
at Surface Mines. Proc. 75th APCA Annual Meeting. Paper 82-15.5.
Cuscino, T., (i.E. Muleski, and C. Cowherd, Jr. 1.983. Iron and Steel Plant
Open Source Fugitive Emission Evaluation. EPA-600/2-83-110 (NTIS P884-110568)
Environmental Protection Agency, Research Triangle Park, NC.
Gil'stte, D. I978a. A Wind Tunnel Simulation of the Erosion of Soil: Effect
of Soil Texture, Sandblasting, Wind Speed, and Soil Consolidation on Dust
Production. Atmos. Environ. 12:1735.
Gillette, D. 1978t>. Tests with a Portable Wind Tunnel for Determining Wind
Erosion Threshold Velocities. Atmos. Environ. 12:2309.
Gillette, D.A., J. Adams, A. Endo, and D. Smith. 1980. Threshold Velocities
for Input of Soil Particles into the Air by Desert Soils. J. Geophys. Res.
85(C10):5621, October 20, 198U.
Lawrence, R. 1983. The use of Fabric Windscreens for Fugitive Dust Control.
Presented at the International Specialty Conference on Fugitive Dust Issues in
the Coal Use Cycle.
PEDCc Environmental, Inc. 1984. Fugitive Dust Control Techniques at
Hazardous Waste Sites, Interim Technical Field Sampling Report No. 3. Control
of Storage Pile Emissions with Windscreens and Chemical Dust Suppressants.
Draft final report prepared for U.S. Environmental Protection Ayency,
Municipal Environmental Research Laboratory, Cincinnati, OH.
R-l
-------
PEDCo Environmental, I,.c. 1984a. Cost-Effectiveness of Dust Suppressants on
Sjrface Coal Mine Haul Roads. Prepared for U.S. dureau of Mines, Twin Cities
Research Center, Minneapolis, Minn.
PEJCc Enviromnental, Inc. and Midwest Research Institute. 1984. Improved
Emission Factors for Fuyitive Dust from Western Surface Coal Mines.
EPA-6UO/7-84-048 (NTIS PB84-1708U2). U.S. Environmental Protection Ayency,
Industrial Environmental Laboratory, Cincinnati, Ohio 45268.
Thornthwaite, C.W., 1931. The climates of North America according to a new
classification. The Geographical Review. 21:633.
U.S. Environmental Protection Ayency. 1977. Quality Assurance Handbook for
Air Pollution Measurement Systems. Volume II - Ambient Air Specific Methods.
EPA-600/4-77-027a (NTIS PB 273618). Environmental Monitoriny Systems
Laboratory, Research Trianyle Park, NC.
R-2
-------
APPENDIX
STATISTICAL REVIEW OF
FIELD EVALUATION OF WINDSCREENS
AS A FUGITIVE DUST CONTROL MEASURE
FOR MATERIAL STORAGE PILES
A-l
-------
Statistical Review of
•Field Evaluation of Windscreens
as a Fugitive Dust Control Measure
for Material Storage Piles
Prepared by
Brian Aldershof and David Ruppert
Statistics Department
University of NC-Chapel Hill,
Chapel Hill, NC 27514
Contract No. 68-02-3858
EPA Project Officer: Gary L. Johnson
Air and Energy Engineering Research Laboratory
Research Triangle Park, North Carolina 27711
Prepared for
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
May 1986
A-2
-------
TIM focus of our analysis of the windscreen data was twofold:
to develop a modal relating total particulate emission to
windspeed and soil moisture content, and to determine the
relationship between various windscreen parameters and particulate
reductions. The analyses suggest optimal windscreen
configurations as well as some overall guiding principles in
windscreen design.
MoUellino total particulate omissions.
examination of plots of particulate emissions against
windspeed and moisture content of the soil suggest that a power
model is probably appropriate. Since this is supported by
theoretical results as well, our analysis proceeded directly to
power models. Of the 42 tests given in table 5.3. only 39 were
used in the analysis. Test 32 was eliminated because thsre was no
information given on soil moisture content. Tests 36 and 105 were
eliminated because the 0 particulate emissions did not allow log
transformations. The simplest model was the power model:
ER a pl(WS)/3*(MC)^Jl fc
where CR is the particulate emission rate in Ib/m*. US is the
windspeed in ft/min. NC is the moisture content as percent of
weight, and 4 is the error term. The multiplicative error term
was assumed to be log-normally distributed. This model was
analyzed using ordinary least squares regression after log
transforming both sides of the equation. The resulting
A-3
-------
coefficients are /Jt= 6.U. x 10~*. p,a 2. S3. /J,= -o.r» (See Table
l). Plots of the residuals confirmed that the assumption that the
error terms are log-normally distributed was reasonable. These
values agree well with previous fugitive dust control research
which suggest that /}, should be about 3 and /}, between -j.3 and
-2.0. The resulting R2 for this model is 0.7^ and the R2
adjusted for the degrees of freedom is 0.785. The model and all
the regression coefficients are significant at the 0.001 level.
The Akaike's information criterion (AIC)1for this model is -&6.37.
A plot of the residuals against the predicted values of
particulate emissions suggest a slight quadratic relationship.
Examination of the plots of the residuals against the windspeed
suggest that this can best be modelled using another term to
account for differing effects of windspeed beyond a threshold
value of 10OO ft/sec. The model examined was:
ER = /*
where US* = US for US > 100O and US* = iOOO for US < iOOO. The
model was examined by taking logarithms of both sides of the
equation and then applying ordinary least squares. Again, the
error term i«as assumed to be log-normal ly distributed.
The AIC is a relative measure of goodness-of-fit. Models that
minimize the AIC fit the data well. The AIC favors models with a
small number of parameters and. high log-likelihoods.
A-4
-------
lisa resultant '"" **' IS:
log(ER) » log(/»t) * /J,log(US) * />,log(NC)
+ /j«(log(US) - log(iooo)] + e' for US > 10OO
log(ER) 3 iog(/J,) + /J. log(US) * /j,iog(NC) > &' for US < 1000
mere were 17 tasts with windspeeds greater than 1OOO ft/mm. The
resulting least squares coefficients are /J,» 5.8 x 10~ , /},» 1.7S.
/J,= -O.43, and /J«» S.18 (Sea Table l). The R2 for this model is
2
0.833 and t.te adjusted R is 0.819. The overall model is
significant at the 0.001 level. Each of /j, and ft, are significant
at the O.OOl level. /}, is significant at the O.lS level and pA is
significant at the 0.01 level. The AIC for this model is -73.02.
In several respects the second model Tonrnf better than tha
first. The R . adjusted R . and AIC all seem to indicate that the
second model more adequately summarizes the data. A threshold
effect of windspeed on many large particles also seems plausible.
However, there were not enough tests (17) done in high windspeeds
to estimate /J« accurately. The standard error for /J« was 1.86.
While the model seems to fit very well within the range tested.
some caution oust be taken in extrapolating ".he model to higher *
winospeeds.
Regardless of which model is chosen the important feature is
the very high exponent of windspeed in the models. Both models
agree reasonably viell with the cubic relationship found in the
previous research on fugitive dust control. The importance of
this'cubic relationship is that in aany regards the conparison of
windspead reductions at a single sensor and total particulate
reductions is irrelevant. Comparison of the windspeed reductions
A-5
-------
at sensor 2 and sensor 3 snows that the variability In wlndspeed
Is very high across the pile. Since particulate emissions are
proportional to the cube of wlndspeed. small overall reductions in
windspeed are easily outweighed by large local uihdspeeds created
by turbulence caused by the screen. This observation Is supported
by the many (14) tests where reduced windspeed at either or both
sensors was accropanied by an increase in particulate emission.
An important characteristic for an effective windscreen is that it
does not create turbulence.
Modelling the effect of windscreen configuration on particulate
reduction.
Regression analysis was next employed to determine the
factors causing reductions in particulate output associated with
the windscreen. Again, several models are suggested. The
dependent variable used in the analysis was 1 - (ER /ER ).
v scr unscr'
This dependent variable can also be expressed as
(ER ^^ - ER )XER and should be interpreted as a
v unscr scr' unscr
proportional reduction in particulate emission rate. Only 37 of
the 42 tests were originally used in the analysis. Test ib and 26
were eliminated because windscreens were not used in these tests.
Test 32 was excluded because of its missing value for moisture
content. Tests 34 and 105 were eliminated because of their 0
particulate emission rates.
Regression using windspeed. moisture content, and three
screen dimensions as independent variables showed that the only
A-6
-------
significant factors affecting partlculate reductions ware
windspeed measured on the unscreened pile. length of the
•windscreen, and height of the -windscreen. The analysis continued
by considering only these three independent 'variables. Test 32.
which was originally excluded from the analysis because of its
missing value for moisture content, was then included. In all. 38
tests were used In the analysis, of partlculate reductions.
Examination of the plots of the data suggested log trans/orm&tions
of the independent variables. Maximizing the log-likelihood with
respect to the Box-Cox trans format ion exponent. A. suggested that
the dependent variable should not be transformed at all. The
resulting model is:
1 - (ER /ER ) = ft. + /Jtln(US) + fl,ln(L) + fl ln(H) + t
scr unsc • 4 v
where L is length of the windscreen in pile diameters and H is
height in pile diameters. The resulting least squares
coefficients are /Jx- -1.S9. /},= 0.22. /3,= 0.34. /3 » 0.38 (See
Table 2). R2 for this model is 0.274 and the adjusted R2 is
0.210. The overall model is significant at the 0.01 level. The
regression coefficient p, is significant at the 0.002 level. /J, at
0.07. /3, at 0.02. /»4 at O.O4. The AIC for the model is -72.9.
This model is rather complicated but its importance is in the
positive coefficients for height and length. Within the ranges
tested, the greater the height and length, the wore effective was
the windscreen. Also, greater reductions in particulate emissions
were found with higher winds, confirming that windscreens are
particularly effective and necessary with high winds.
Examination of interaction terms suggested a better, if
A-7
-------
conceptually more difficult, model. The same dependent variable
VMS ragressed on the untransformed values of wlndspeed. length.
and height and their interactions. The best model using these
terms was found to be:
x .UST) + fc
where HT is the mean-centered height (ie. H - ff) . UST Is the
mean-centered wlndspeed. and US x H is the interaction between
the mean-centered wlndspeed and mean-centered height . The least
squares values for the coefficients ire /3,= 0.33. /3,= 0.24. and
/J,= 0.43. /J4a3.4 x 10*4. /J,»l.O x IO~ ""(See Table 2). This model
fit the data better than the previous model. R Increased to
0.354. the adjusted R2 decreased to 0.275. and the A 1C improved to
-74.22. The overall model was significant at the 0.01 lex-el. The
coefficient /J, is significant at 0.3. ft, at 0.02. /J, 'at 0.0&. /J«
at 0.05. /}, at 0.05.
Mathematically, the second model is probably better than the
first. The R and AIC are Improved in the second model and the
significant interaction should not be ignored. The interpretation
of the interaction between height and wlndspeed can be misleading.
For fixed values of L and US . particulate reduction is a linear
function of H with estimated slope:
/J, •»• £4(WSr) a 0.43 •»• (3.4 X 10~4) (wV)
This slope increases with US . Thus, low windspeeds lessen the
effectiveness of increased height .n reducing relative particulate
output. in fact, the estimated slope is negative for large
negative values of US suggesting that low windscreens ar<«
A-8
-------
actually acre efficient than high screens at low ulndspeeds.
However, since high windspeeds cause much greater particulata
emission ratis. it is such raore desirable to have a windscreen
that is efficient at high windspeeds rather than low windspeeds.
TO further Investigate the interaction, a hypothesis test was
performed by testing:
H, : ftt + /5,(WS) > 0
vs. Ht : /3, * /3,(US) < 0
This tests the null hypothesis .I0) that increasing the height of
the screen at z-iro wlndspood causes a decrease in particulate
emission versus the alternative hypothesis (H,) that increasing
the height causes a decrease in particulate output. The
t-statistic for this test is -O.507. which is not significant.
There is no evidence to suggest that a low screen is really more
effective in low windspeeds than a high screen. It is clear that
increasing the height of the windscreen at low windspeeds does not
Increase Its efficiency.
Our analysis did not show any effect of distance between the
pile and the windscreen. Distance obviously has some 35feet (a
reasonable sized screen two blocks from the pile will probably not
reduce particulate emissions), but within the range of values
studied there does not soem to be any preferable distance.
The analysis of the first model (the non-linear model without
the interaction) suggested that increased length and height
reduced particulate emissions similarly. Since it is probably
less expensive to build a screen a unit longer than a unit higher.
building a long screen is more cost effective in reducing
A-9
-------
particular emissions than building a high one. The second model
(the linear node! with the interaction) summarizes the data fairly
well, but leati* to some difficult conclusions about screen design.
Apparently, the most efficient screen per unit area would be.
inf initesmally high and Infinitely long. This ridiculous
conclusion is the result of :r.e limited range of heights studied.
The important .conclusion from the interaction model is that if
high winds are encountered on the pile, high screens are acre
efficient than long ones.
The original report presents only a linear model of
particulate reduction and does not examine interactions. This
linear model is not satisfactory since plots of the residuals
reveals that it does not fit the error structure of the data well.
Addition ->f the interaction cenn helps in understanding the
dynamics of particulate reduction caused by the windscreen.
additional
Section 5.2.1 concludes that "windscreen configuration should
not affect the relationship between windspeed and emission
reduction". In fact, ••his is not true as is suggested by the
emission reduction model using the interaction between height of
the windscreen and windspeed.
We did not do any analyses using the particle size data.
Section 5.2.2 concludes that the negative particulate reductions
in all the particle size ranges confirm that the increased
emissions found in the total particulate data resulted from actual
A-10
-------
Increases caused by the screen. This does not follow because
these *re not independent observations. Total particulate
emissions are probably highly correlated with emissions in every
size range which are also probably highly correlated among
themselves.
The equation for total particulate output given in the
original report is ER = 0.0078 («S)°'21(MC)~2*1. Our equation.
based on the same data set. is ER = &.2xlO~9 (US)2'53(MC)~°'9<'.
This discrepancy rnnmT to be causad. by an error in the original
report. Apparently, the un trans formed ER was regressed on the log
transformed independent variables. The rcodel that results from
this.regression is :
exp(ER) =« ^(w^^MC)^ t
not
ER a /Jt(USy°*(MC/Jl &
as reported. This exponential model does not fit the data as well
as the power mxJel. It appears that a similar error occurs in the
analysis of the particle size data. The dependent variables in
Table 5.7 are actually exp(ER ) rather than ER
part* part
Ue did not find that wind direction significantly affected
particulate reductions. While it yinm* obvious that wind
directions '•hat are very ^perpendicular to the screen must
undermine its effectiveness, this WAS not seen in the narrow range
of wind directions tested.
Ue feel that the rather simple multiple regression models
presented here suanarlze the data quite well. The analysis
presented here is not Intended to replace all of the analysis in
A-ll
-------
Chapters S and 4 of tfw original report. Rathwr. wa foal that
ttMsa models supplement th« longar analyses In tt« original
report, as well as emend «mme of them.
A-12
-------
Table i; Modelling Total Particuiate Output
Model: ER = ftt (US)2(MC)» t
Coefficient value (se) ' P-value
/J, 4.14 x 10~* (3.1 x 10~8} 0.001
ft, ' 2. S3 (0.27) 0.001
ftt -0.99 (0.21) 0.001
adj. R2=0.78S AIC = -<,8.37
Model: ER 3 /J, (VS)/J*(MC)/J*(UsVlOOO)^
-------
Table 2; Modelling the effect of windscreen configuration on
Mc.de 1:1-
Coef f icient
fti
3,
ft,
R2a0.274
Dart Iqulata
(ER /ER ) » ft.
v $cr unsc' ^'
.Value (se)
-1.S9 (0.74
0.22 (O.ll)
0.34 (0.13)
0.38 (0.17)
adj. R2»0.2
reduct i.ons,
+ ptln(US) + p,ln(L) + /J^lnJH) * e.
P-va lue
) 0.002
0.07
0. 02
0.04
10 AIC--72.9
Model: 1 - (ERscr/ERunsc) - />, + P,(L)
Coefficient Value (se) P-value
/), -o. 14 (0. iS) 0.33
flj 0. 11 (0.05) 0.02
/J, 0.43 (0.21 ) C.06
0, 3.4 x 1O~4 ( 1 . t> x 10"4) 0.05
fl, 1.0 x 10~3 (5.2 x 10~4) O.OS
R2=0.3S adj. R2=0.27S AIC»-74.2
-------
TECHNICAL REPORT DATA
(Pleat rtad /mtotietiutu on tHt nvtnt btfort eomptttingt
. REPORT NO.
EPA/600/7-86/027
3. RECIPIENT'S ACCESSION NO.
PBS fi 2 3 T 2 8 9 /AS
». TITLE ANOSUBTITLE
Field Evaluation of Windscreens as a Fugitive Dust
Control Measure for Material Storage Piles
6. REPORT DATE
July 1986
«. PERFORMING ORGANIZATION CODE
1. AUTHORIS;
Robert A. Zimmer. Kenneth Axetell. Jr.. and
Thomas C. Ponder. Jr.
a. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
PE1 Associates, Inc.
14062 Denver West Parkway
Golden. Colorado 80401
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-3995. Task 15
12. SPONSORING AGENCY NAME ANO ADDRESS
EPA, Office of Research and Development
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
3. TYPE OF REPORT ANO PER I Op CD
Task Final; 4/85 - 6/85
VEREO
14. SPONSORING AGENCY CODE
EPA/600/13
19. SUPPLEMENTARY NOTES AEERL project officer is Dale L.
541-2429.
Harmon. Mail Drop 61. 919 /
16. ABSTRACT
The report gives results of a field study to validate earlier wind tunnel
studies with respect to windspeed changes, and to determine the relationship be-
tween changes in windspeed and changes in fugitive dust emissions. (The earlier
studies were to determine changes in windspeed—not changes in emissions—due to
windscreens, and utilized a wind tunnel to determine the optimal windscreen poro-
sity, size, and location for control of fugitive dust emissions from storage piles.)
The field study suggests that the optimum windscreen design parameters are: poro-
sity = 50%, .height = 1.0 H. width = 5.0 D, and distance = 2.0 H for a conical pile of
height H and diameter D. Analysis of the field data shows that emission rates were
directly related to windspeed and inversely related to moisture content of the pile
surface. These relationships held regardless of the particle size fraction considered,
KEY WORDS ANO DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Ftttd/Gioup
Pollution
Dust
Leakage
Particles
Windbreaks
Protectors
Storage
Wind Velocity
Pollution Control
Stationary Sources
Particulate
Fugitive Dust
Windscreens
Storage Piles
13 B
11G
14G
04B
15E
3. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (Thit Report/
Unclassified
21. NO. OF PAGES
20. SECURITY CLASS (Thisptgt)
Unclassified
22. PRICE
EPA Perm 2220-1 (»-7J)
------- |