-------
-------
TECHNICAL REPORT DATA
/ffmr mad Imttncttont on tHt rtrtnr btfort completiiti)
NO.
EPA/600/2-86/072
2.
4. TITLE AND SUBTITLE
Critical Review of Open Source Particulate Emission
Measurements: Field Comparison
5. REPORT OATI
August 1986
. PERFORMING ORGANIZATION CODE
7. AUTHORISt
Bobby £. Pyle and Joseph D. McCain
I. PERFORMING ORGANIZATION REPORT NO.
SoRI-EAS-85-444R
9. PERFORMINO ORGANIZATION NAME AND ADDRESS
Southern Research Institute
P. O. Box 55305
Birmingham. Alabama 35255-5305
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NOT
68-02-3696. Task 2
13. SPONSORINO AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC 27711
1JLTVPE QJ REPORT AND PERIOD CO
Task Final; 6/83 - 6785
VERED
14. SPONSORINO AGENCY CODE
EPA/600/13
19. SUPPLEMENTARY NOTES AEERL project officer is Robert C. McCrillis. Mail Drop 65.
919/54ir 2733.
16. ABSTRACT
The report gives results of a review of sampling and analytical procedures
used by various testing firms to quantify particulate emissions from open sources;
e.g.. roads and storage piles. Seven firms, who account for nearly 100 percent of
all open source data in the literature, described their current sampling and analyti-
cal procedures. Five of these firms then participated in a simultaneous side-by-side
field test on a simulated unpaved road at a major steel plant. Each firm independ-
ently measured the particulate emission concentrations produced by roadway traf-
fic. These measurements produced not only the particle-size-dependence of the
emissions but also the concentrations as functions of the distance above the road sur-
face. The results for each firm were expressed as emission factors for total particu-
late and the mass fractions of the particulate with sizes <30. <15. <10. and <2.5
micrometers diameter. Based on an analysis of the results, it was found that all five
profiling systems could produce equivalent results in terms of total emissions^, This
was not the case for emissions by particle size. The only technique of those tested
that produced reliable emission factors by particle size was inertia! sizing. ;
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
C. COSATI h Kid/Group
Pollution
Measurement
Particles
Dust
Field Tests
Sampling
Analyzing
Roads
Pollution Control
Stationary Sources
Unpaved Roads
Storage Piles
Inertia! Sizing
13B
14G
11G
14B
It. DISTRIBUTION STATEMENT
Release to Public
It. SECURITY CLASS
Unclassified
ii. NO. or PAGES
13A
M. SECURITY CLASS f7Mt*MtJ
Unclassified
22. PRICE
-------
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.
ii
-------
ABSTRACT
Current EPA emission trading (bubble) policies allow excessive emissions
from one source to be offset by improved controls on other sources within the
same industrial complex. In implementing this bubble policy, many industries
have opted to reduce particulate emissions from roads, parking lots, and
storage locations in lieu of more stringent controls on other plant operations.
The predictions of these open sources emissions based upon the existing data
for measured emission factors have been found to be subject to high
variability. Part of this uncertainty can be attributed to the variations in
source characteristics. However, a significant part of the uncertainties are
due to variations in the measuring and analytical techniques used to assess
sources upon which the emission factor data base was built. In an effort to
quantify these technique dependent 'parameters, a comparative study of the
current measuring and data reduction methods was carried out. This study
included a simultaneous, side-by-side field test with five independent testing
organizations sampling the dust emissions from a simulated unpaved road at a
major steel producing facility. Each of the participants independently
measured the particulate emission concentrations produced by roadway traffic.
These measurements resulted in not only particle-size dependence of the
emissions but also concentrations as functions of the distance above the road
surface. The results for each testing organization were expressed as emission
factors for total particulate and the mass fractions of the particulate with
sizes <30, <15, <10, and <2.5 pm diameter. Based on an analysis of the re-
sults, it was found that all five profiling systems were capable of producing
equivalent results in terms of total emissions. This was not the case for
emissions by particle size. The only technique among those tested that was
believed to produce reliable emission factors by particle size was the inertial
sizing method.
This work was submitted in partial fulfillment of EPA Contract Number
68-02-3693, work assignment 002 by Southern Research Institute under
sponsorship of the U.S. Environmental Protection Agency. This report covers a
period from 1 January 1984 to IS March 1985 and work was completed as of 10 May
1985.
ill
-------
CONTENTS
Section Page
Figures vi
Tables. xi
Acknowledgment xn
1. Introduction 1
2. Conclusions and Recommendations.. 5
3. Fugitive Measurement Methodologies 8
4. Description of Test 28
5. Particle Size Measurements 34
Methods 34
Cyclone/lmpactors 34
Stacked Filters 36
Scanning Electron Microscopy. 39
Results 41
Stacked Filter 43
Scanning Electron Microscopy. 47
Cyclone/lmpactors 52
Particle Sizing Recommendation 55
6. Exposures and Emission Factors.. 57
Methodologies 57
Sampling Xsokineticity 57
Exposure Integration 58
Field Test Results 60
Total Particulate Emission Factors 60
Exposure and Emission Recommendations 84
7. Modeling Fugitive Emissions 85
Predictive Emission Factor Equations 85
Modeling Recommendations 94
References 118
Appendix 120
-------
FIGURES
Number Title Page
1 Upwind profiling system used by EEM 10
2 Downwind profiling system used by BEN 11
3 Typical sampling head used by EEM. 12
4 Upwind profiling system used by MRI 13
5 Downwind profiling system used by MRI 14
6 Typical hi-vol sampler used by MRI (without particle sizing) 16
7 Upwind profiling system used by PEL 17
8 Downwind profiling system used by PEI 18
9 Typical stacked filter sampling head used by PEI ..19
10 Upwind sampling system used by TRC 20
11 Downwind profiling system used by TRC. 21
12 Typical sampling head used by TRC 22
13 Upwind sampling system used by USS 23
14 Downwind profiling system used by USS 24
15 Typical sampling head used by USS 25
16 Gary test site after covering road with dirt. View is looking
wes t 29
17 Physical layout of the Gary comparative fugitive emission
test site 30
18 Total suspended particulate matter by test and sampling
position as measured by standard high-volume sampler. 33
19 MRI cyclone/cascade impactor combination 35
20 Lundgren cascade impactor 37
21 Collocated Lundgren cascade impactor (left) and MRI sierra
cyclone and impactor (center)............ 38
22 PEI stacked filter profiler head 40
vi
-------
Figures (Cont.)
Number Title Page
23 Particle size distributions as measured by the various techniques..42
24a Plot of the ratio of PBI's TSP concentration to hi-vol
concentration versus the total particulate mass collected by
the hi-vol 46
24b PEI screen penetration dependence on total mass collected by
the profiler sampler. «46
25 Particle sice distribution as measured by a Lundgren cascade
impactor and as determined by the OCSBM technique from a
collocated profiler sample 48
26 Total particulate emission factors by test as measured by each
organization 64
27 Measured total exposures versus height for Test 1 for all
contractors providing total particulate exposure data.. 72
28 Measured total exposures versus height for Test 2 for all
contractors providing total particulate exposure data 73
29 Measured total exposures versus height for Test 3 for all
contractors providing total particulate exposure data 74
30 Measured total exposures versus height for Test 4 for all
contractors providing total particulate exposure data 75
31 Measured total exposures versus height for Test 5 for all
contractors providing total particulate exposure data 76
32 Measured total exposures versus height for Test 6 for all
contractors providing total particulate exposure data 77
33 Measured total exposures versus height for Test 7 for all
contractors providing total particulate exposure data 78
34 Measured total exposures versus height for Test 8 for all
contractors providing total particulate exposure data. 79
35 Measured total exposures versus height for Test 9 for all
contractors providing total particulate exposure data... 80
36 Measured total exposures versus height for Test 10 for all
contractors providing total particulate exposure data 81
37 Measured total exposures versus height for Test 11 for all
contractors providing total particulate exposure data 82
vii
-------
Figures (Cont.)
Number Title Page
38 Scatter diagrams oft the test averaged TP emission factor
measurements and that predicted by EEM's equation (filled
circles) along with the associated regression line (dotted
line)} and the values of TP emission factor both measured
and predicted by EEM (open circles) along with the
associated regression line (broken line) 91
39 Scatter diagrams of: the test averaged TP emission factor
measurements and that predicted by MRI's equation (filled
circles) along with the associated regression line (dotted
line); and the values of TP emission factor both measured
and predicted by MRI (open circles) along with the
associated regression line (broken line) 96
40 Scatter diagrams ofi the test averaged TP emission factor
measurements and that predicted by TRC's equation (filled
circles) along with the associated regression line (dotted
line); and the values of TP emission factor both measured
and predicted by TRC (open circles) along with the
associated regression line (broken line) .....97
41 Scatter diagrams of: the test averaged TP emission factor
measurements and that predicted by USS's equation (filled
circles) along with the associated regression line (dotted
line); and the values of TP emission factor both measured
and predicted by USS (open circles) along with the
associated regression line (broken line) 98
42 Scatter diagrams of: the test averaged TSP emission factor
measurements and that predicted by EEM's equation (filled circles)
along with' the associated regression line (dotted line); and the
values of TSP emission factor both measured and predicted by EEM
(open circles) along with the associated regression line (broken
line) 99
43 Scatter diagrams of: the test averaged TSP emission factor
measurements and that predicted by MRI's equation (filled
circles) along with the associated regression line (dotted
line); and the values of TSP emission factor both measured
and predicted by MRI (open circles> along with the
associated regression line (broken line) 100
44 Scatter diagrams of: the test averaged TSP emission factor
measurements and that predicted by PEI's equation (filled
circles) along with the associated regression line (dotted
line); and the values of TSP emission factor both measured
and predicted by PEI (open circles) along with the
associated regression line (broken line) 101
viil
-------
Figures (Cont.)
Humber Title Page
45 Scatter diagrams ofi the test averaged TSP emission factor
measurements and that predicted by TRC's equation (filled
circles) along with the associated regression line(dotted
line)} and the values of TSP emission factor both measured
and predicted by TRC(open circles) along with the associated
regression line (broken line).............* 102
46 Scatter diagrams oft the test averaged TSP emission factor
measurements and that predicted by OSS's equation (filled
circles) along with the associated regression line (dotted
line)) and the values of TSP omission factor both measured
and predicted by USS (open circles) along with the
associated regression line (broken line) 103
47 Scatter diagrams ofi the test averaged PM15 emission factor
measurements and that predicted by EBM's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PM15 emission factor both measured
and predicted by EEM (open circles) along with the
associated regression line (broken line) 104
48 Scatter diagrams ofs the test averaged PM1S emission factor
measurements and that predicted by MRl's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PH.. emission factor both measured
and predicted by MRI (open circles) along with the
associated regression line (broken line) 105
49 Scatter diagrams oft the test averaged PH15 emission factor
measurements and that predicted by TRC's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PH., emission factor both measured
and predicted by TRC (open circles) along with the
associated regression line (broken line) 106
50 Scatter diagrams oft the test averaged PM15 emission factor
measurements and that predicted by USS's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PMj. emission factor both measured
and predicted by USS (open circles) along with the
associated regression line (broken line). 107
51 Scatter diagr; 2 oft the test averaged PMJO emission factor
measurements and that predicted by EBM's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PM.fl emission factor both measured and
predicted by EEM (open circles) along with the associated
regression line (broken line) 108
ix
-------
Figures (Cont.)
Number Title Pagt
52 Scatter diagrams of: the test averaged PH.. emission factor
measurements and that predicted by MRI's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PMj. emission factor both measured
and predicted by MRI (open circles) along with the
associated regression line (broken line) 109
53 Scatter diagrams oft the test averaged PM1Q emission factor
measurements and that predicted by PEI's. equation (filled
circles) along with the associated regression line (dotted
line); and the values of PH.. emission factor both measured
and predicted by PEX (open circles) along with the
associated regression line (broken line)..... 110
54 Scatter diagrams oft the test averaged PM1Q emission factor
measurements and that predicted by TRC's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PH.. emission factor both measured
and predicted by TRC (open circles) along with the
associated regression line (broken line) 111
55 Scatter diagrams oft the test averaged PMJO emission factor
measurements and that predicted by USS's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PH.. emission factor both measured
and predicted by USS (open circles) along with the
associated regression line (broken line) ..112
56 Scatter diagrams oft the test averaged FP emission factor
measurements and that predicted by EEM's equation (filled
circles) along with the associated regression line (dotted
line); and the values of FP emission factor both measured
and predicted by EEM (open circles) along with the
associated regression line (broken line) 113
57 Scatter diagrams oft the test averaged FP emission factor
measurements and that predicted by MRI's equation (filled
circles) along with the associated regression line (dotted
line); and the values of PP emission factor both measured
and predicted by MRI (open circles) along with the
associated regression line (broken line) 114
58 Scatter diagrams oft the test averaged FP emission factor
measurements and that predicted by PEI's equation (filled
circles) along with the associated regression line (dotted
line); and the values of FP emission factor both measured
and predicted by FBI (open circles) along with the
associated regression line (broken line) 115
-------
Figures (Cent.)
Number Title Page
59 Scatter diagrams oft the test averaged PP emission factor
measurements and that predicted by TRC's equation (filled
circles) along with the associated regression line (dotted
line); and the values of FP emission factor both measured
and predicted by TRC (open circles) along with the
associated regression line (broken line) 116
60 Scatter diagrams oft the test averaged FP emission factor
measurements and that predicted by OSS's equation (filled
circles) along with the associated regression line (dotted
line); and the values of FP emission factor both measured
and predicted by USS (open circles) along with the
associated regression line (broken line) 117
TABLES
1 Profiling tyitti oharaattriitiQi««««««««t««««««««««»«««««. «...«.. .9
2 Average emission factors by size class* ........................... 44
3 Comparative fugitive emission test results ......................... 61
4 Average emission factors obtained by each organization ............. 62
5 Correlation coefficients of TP, TSP, PM1Q, and FP emission
factors reported by test participants ............. «« ......... 65
6 Exposure and emission values by test number and organization ....... 66
7 Size dependent constants for the predictive emission factor ........ 87
8 Size fractions of particulate as reported by MRI and
recalculated by Southern Research Institute ...................... 88
9 Average emission factors as calculated by Southern Research
Institute* «*««**'«» »90
10 Correlation and regression line coefficients for the emission
factor comparisons t * t i » t * 92
xi
-------
Several individuals and organizations contributed to the success of this
study. The efforts of those associated with the United States Steel Corporation
bear individual Mentions Victor Hordlund, Supervisor of Environmental Engineer-
ing at the Gary Works, and his staffi William Kubiak in particular for assisting
in selection and preparation of the test site; and T.P. Eckle with the USS Tech-
nical Center who participated in the comparative test at their own expense.
xii
-------
SECTION 1
INTRODUCTION
During the past decade, research has shown that particulate eaissions
from open sources such as roads and material storage piles contribute signifi-
cantly to ambient particulate matter concentrations in many areas*1 The cur-
rent EPA emission trading policy allows excessive emissions from one source to
be offset by improved control of another source within the same plant. In
implementing this bubble policy, many steel plants have agreed to reduce fugi-
tive dust emissions in lieu of tighter controls on process emissions.2 How-
ever, the efforts of several groups to develop equipment, and methods for quan-
tifying emissions from nontraditional sources such as roadways have resulted in
estimates of emission factors and emission rates with substantial variability.
Whereas it is generally agreed that emission factor estimates of process emis-
sions from ducted sources are good to ±50% of a specified measured value, pre-
dictions of open source fugitive dust emission factors may vary from measured
values by as much as an order of magnitude. These large uncertainties are due
to both differences in the measuring techniques used and the site examined.
Even for sites with closely similar physical characteristics the measured
emissions have been found to differ by as much as an order of magnitude. In an
effort to resolve these differences, the U.S. Environmental Protection Agency
contracted with Southern Research Institute (SoRI) to conduct a critical'review
of the various measurement and predictive methods and to conduct a side-by-side
field test in which testing organisations whose methods represent the principal
techniques which have commonly been applied, would simultaneously sample the
emissions from an unpaved road*
The first phase of this review was the compilation of documents that
describes in detail each of the testing methods and data reduction techniques
used in developing the existing data base. A list of these reports is shown in
the Appendix. As a result of the initial phase of this study, documents were
received from the seven organizations listed belowi
Catizone ft Martin Associates, Inc. (CMA)
63 McKee Street
East Hartford, CT 06108
(203) 269-1331
-------
Energy * Environmental Management, Inc. (BEN)
P.O. Box 71
Murrysville, PA 15668
(412) 247-5124
Energy Impact Associates (ElA)
P.O. Box 1893
Pittsburgh, PA 15230
(412) 351-5800
Midwest Research Institute (MRI)
425 Volker Boulevard
Kansas City, MO 64110
(816) 753-7600
PEI Associates, Inc. (PBI)
(Formerly PEDCo Environmental, Inc.)
11499 Chester Road
Cincinnati, OH 45246
(513) 782-4700
TRC Environmental Consultants, Inc. (TOO
800 Connecticut Boulevard
East Hartford, CT 06108
(203) 289-8631
United States Steel Corporation (OSS)
Research Laboratory
125 Jamison Lane
Monroevi le, PA 15146
(412) 372-1212
*CMA is no longer in operation.
Each of these current procedures documents (CPD) specifically addressed
the following points:
1. A description of the testing methods and data reduction schemes
employed in a specific field test conducted by that organization
for the measurement of both the total particulate (TP) and the
size distribution of that particulate resulting from open source
fugitive emissions and other ancillary measurements such as
surface loading of silt on roadway.
-------
2. A discussion of the historical development of the current
technique with particular emphasis upon any changes that have
been made and the reasons for those changes. If either the
sampling or data analysis techniques have undergone change, the
respective organization described how data obtained before the
modifications were instituted could be correlated with that
obtained afterward or detail why such correlations cannot be made
when such is the case.
3. One or more detailed test reports from recent field tests which
utilized the sampling and data reduction techniques described in
(1) above. Included in these reports were sufficient raw data to
enable the reader to reproduce the final results quoted.
4. A discussion of how the testing and data reduction procedures
>-*
described above would *e modified for use at various site
locations and under varying site conditions.
As a result of review of the CPD's, four organizations were selected to
participate in the simultaneous field comparison during Part II of this study.
These organizations were selected according to the following criteria:
1. Those organizations which have made a significant contribution to
the existing fugitive emission data base using different
methodologies.
2. A representative organization for each alternative methodology
not included in those selected by criterion number 1.
The four organizations originally selected to participate in the field compari-
son were CMA, EEM, MRI, and PEL However, before the field test could be car-
ried out CMA had to withdraw from consideration and were replaced by TRC, the
next choice in line. Another organization, USS, participated in the field
-------
comparison at their own expense and thus, a total of five organizations were
included in the field comparison, EEM, MRI, PBI, TRC, and USS. The individual
field test reports from these five organizations are also listed in the
Appendix.
Because much of the use of the emissions trading policy with respect to
roadways had been in the iron and steel industry, the desired test site was an
unpaved road within an integrated iron and steel facility. Negotiations were
undertaken with several companies with regard to possible plant access for the
tests; the United States Steel Corporation responded favorably, offering the
use of their Gary Works. The nature of the tests to be conducted made it ne-
cessary to use a moderately long, straight stretch of roadway that was oriented
more or less perpendicular to the prevailing wind and clear of local perturbing
influences for a length of a few hundred meters. It was also desirable that the
road have a moderately high traffic density for emissions generation. Unfortu-
nately, all suitable roads on the property of the Gary Works were either paved
or had been treated with chemical dust suppressant* The most suitable road
section in terms of physical layout and traffic density was a paved slag haul
road parallel and adjacent to Lake Michigan. Consequently, this road was made
to simulate an uncontrolled unpaved road by applying a 5-10 cm thick layer of
dirt to one 300 meter section. This portion of the road was located well away
from other potentially confounding sources and wind flow obstructions. Five
side-by-side test positions, each approximately 15 meters wide, were then laid
out on both sides of the road near the center of the simulated unpaved road
section. A total of eleven tests were conducted over a five-day period in June
1984. The results of these tests are the subject of the following report.
-------
SECTION 2
CONCLUSIONS AND RECOMMENDATIONS
Each of the profiling systems used in these tests exhibited both pros
and cons in terms of their versatility and ease of deployment. Consequently,
no conclusions are drawn as to the optimum mechanical design of the profiling
towers or associated hardware. However, several conclusions may be drawn
regarding the methods of sampling. These conclusions and recommendations are
discussed in detail at the ends of Sections S, 6 and 7 and are summarized in
the paragraphs below.
The data from this test series clearly indicate that there is significant
exposure at heights up to at least the 9m level. Consequently, the placement
of a sampler at the 9m or 10m (preferably 10m) elevation is highly recommended.
Also, because the maximum exposure values usually occur at a height of 1.5 to
2.0 meters, any future profiling system should include a sampler at this level.
The "ideal" profiling system would have mass samplers at 1.5, 2.5, 4.0, 6.0,
7.5, and 10.0 meters with concurrent particle sizing devices at 1.5, 4.5, and
7.0 meters. This configuration would better characterize both the exposure and
size distribution of the particulates in the plume. It was also demonstrated
that sampling isokineticity can be maintained for each sampling head on the
profiler tower. Therefore, it is recommended that each sampler be equipped
with a servo system and individual velocity sensors to provide continuous
adjustment of the flow rate based on wind speed at that elevation.
The five profiling systems were found to be capable of producing equiva-
lent results in terms of total emissions. This was not the case for particle
size distributions and emissions by particle size. Long recognized problems in
reconstituting size distributions of airborne particles from resuspensions of
collected bulk material lead to the conclusion that for fugitive emission sam-
pling of the type undertaken here the sizing should take place prior to collec-
tion (or concurrent with collection as in cyclones and impactors). The
recommended procedure for measuring the particle size distribution is MRl's
cyclone/impactor technique with some modifications. With regard to field
-------
operations, first, a further reduction in the sampling flow rate from 20cfm to
IScfm would help minimize errors from particle bounce. Alternatively, adhesive
coated substrates could be used at the current 20cfm flow rate. Second,
potential errors resulting from the possible transfer of material from the
outlet tube of the cyclone to the first stage of the impactor can be avoided by
counting only the material collected in the body of the cyclone as its catch.
The outlet tube catch would then be combined with that of the first impactor
stage. At the current 20cfm flow rate, this would result in a cyclone D50 of
22ym being used rather than the current 14|im value. Lastly, while particle
size was measured at the lower elevations (1.5 and 4.5m) an additional
cyclone/impactor unit located at a height of about 7m would provide additional
information regarding the changes in sice distribution as a function of height
within the dust plume.
A better data analysis technique than the current MRI procedure would be
that commonly used in reducing impactor data from industrial sources. In the
latter, a spline fit is made to the cyclone/impactor data in the cumulative
percentage form of the distribution. The fit is made in a manner that requires
continuity in the slope of the curve and the solution is forced to be
asymptotic to 100% at a diameter equal to the maximum diameter present in the
sample. The fitted curve is then used to interpolate or extrapolate as needed
to obtain the mass fractions in the selected size intervals. This technique
avoids the requirement of assuming a functional form for the distribution and
makes use of the complete data set rather than just two of the data points.
With regard to the exposure integration procedure, the resultant emission
factor appears to be relatively insensitive to the technique used so long as
the exposure is adequately characterized with regard to height. The most
critical area is that where the peak exposures occur (usually at 1.5 to 2. 5m).
With samplers at 1.5, 2.5, and 4.0 meters this should not be a major source of
error. However, under greatly different site conditions (road type, silt
content, etc) the samplers at the lower elevations may need to be positioned at
different heights.
-------
The utility of an emission factor predictive equation is that of predict
-ing the emissions from a particular site in lieu of actual measurements. In
order that the equation be applicable over a wide range of site locations and
conditions, it should include as many of the relevant parameters describing the
site as possible. This requires that the predictive equation be developed from
\
as large a data base as possible* The equation currently described in the
AP-42 manual was developed from a fairly broad data base using multiple linear
regression techniques. The data base has some uncertainties particularly with
regards to the particle size distributions. These uncertainties most certainly
cast some doubt upon the accuracy of the values used for the particle size
multiplier, k, in this equation. However, without an extensive evaluation of
the existing unpaved road emissions data base, there are no justifications for
invalidating the relation. This equation is probably the most reliable
predictor of unpaved road emissions currently available. Because of past
problems in the particle sizing techniques, the values used for the size
multiplier (k) are less reliable than the overall equation. It is recommended
that these values be re-evaluated in the future as the data base is expanded
with more reliable particle size information.
-------
SKTIOM 3
FUGITIVE MEASUREMENT MBTHOOOLOGIBS
All five of the participating organisations utilised variants of the
exposure profiling tower system first developed by MHZ3 to Measure emission
rates. However, there were potentially significant differences in the actual
applications of the Method with respect to such items as the number and heights
of the profiler samplers, the filter orientation within the samplers, the sam-
pling flow rates, the methods used for establishing and maintaining isokinetic
sampling conditions, and the methods used to apportion the total emissions
among the various particle sise fractions. Fractions of interest were TSP
(assumed to be the weight fraction smaller than 30 urn), PM10 (the weight frac-
tion contained in particles smaller than 10 urn), and the fine particle fraction
(defined as the weight fraction contained in particles smaller than 2.5 urn).
All particle diameters used herein are on an aerodynamic (density - ig/cm3)
basis unless specifically noted otherwise* Brief summaries of the methods used
by the participants are given in Table 1.
The profiling methods used by each of the organisations were as follows:
BEM - System composed of an upwind and downwind tower containing three and
five sampling heads, respectively, as shown in Figures 1 and 2. Bach sampling
head consisted of a hi-vol sampler motor, a vertically mounted filter, and an
inlet nossle as shown in Figure 3. The samplers on each tower were raised and
lowered vertically by a concatenated pulley arrangement. The samplers were
operated at preset fixed flow rates based on a IS-minute average wind speed at
a height of 4 m for the period immediately preceding the test. The vertical
velocity distribution of wind speed was assumed to be uniform.
MRI - System used for these tests differed somewhat from that used in the
past as indicated in Table 1. The system used here was composed of a two head
upwind tower and a five head downwind tower as shown in Figures 4 and 5. Bach
nonsising sampler head contained a horisontally mounted hi-vol filter and an
8
-------
-------
»»»-»!§
rigor* 1. Upwind profiling «y»t«« oMd by ON.
10
-------
- _..
-------
MM-III
Figure 3. Typical stapling head used by RBI.
12
-------
-------
riguc* 3. Downwind profiling «y«tt» u«^ by mi
14
-------
inlet nozzle below the filter as seen in Figure 6. The sampling heads are
raised and lowered vertically by a system of pulleys. Periodic manual flow
adjustments were based on wind speeds from a logarithm wind shear curve fit to
measured wind speeds at heights of 1.5 m and 4.5 m.
PEI - System composed of identical upwind and downwind towers each con-
taining four sampling heads as shown in Figures 7 and 8. Bach sampling head
consisted of a hi-vol motor and several circular, stacked filters and an inlet
nozzle as shown in Figure 9. Sampling heads are raised and lowered vertically
while attached to a telescoping tower. In the current tests the maximum height
of the profiler was higher than in previous tests (see Table 1). Periodic
manual flow adjustments were based on wind speed at a height of 3 m with the
vertical velocity profiles adjusted for wind shear by means of an assumed power
function.
TRC - System composed of an upwind tower containing one sampler head and a
downwind tower with five sampling heads as illustrated in Figures 10 and 11,
respectively. Each sampling head contained a hi-vol filter mounted horizontal-
ly and an air intake transformer above as shown in Figure 12. The sampling
heads were gimbal mounted to the tower and maintained their orientation as the
tower was lowered to a horizontal position. Continuous, servo-controlled flow
adjustments were based on wind speeds measured by thermal anemometers located
adjacent to each sampling inlet.
USS - System consisted of one sampling head upwind tower and a downwind
tower containing four samples as shown in Figures 13 and 14, respectively.
Each sampling head consisted of a hi-vol filter mounted horizontally and an
intake nozzle above as illustrated in Figure 15. The sampling heads were
gimbal mounted to the tower and maintained their orientation as the tower was
lowered by inclining. Continuous, servo-controlled flow adjustments were based
on wind speeds measured by thermal anemometers located adjacent to each
sampling inlet.
15
-------
MM4II
Figure 6. Typical hi-vol sampler used by MRI (without particle siting).
-------
-------
':
ttl»-tli
Figure 8. Downwind profiling system used by FBI.
18
-------
ItM-Ilt
Figure 9. Typical stacked filter sampling head used by PEL
19
-------
MIJ-411
Figure 10. Upwind sailing system used by THC.
20
-------
Figure 11. Downwind profiling system used by TRC.
21
-------
MM-ttt
Figure 12. Typical sampling head used by THC.
22
-------
Figure 13. Upwind sampling systea used by DBS.
23
-------
IOtt-
-------
-------
Three different methods were used for the determination of particle size
distributions. Both MRI and PEI used forms of inertial sizing while the
remaining three organizations used electron microscopy of the material
collected by the profiler samplers.
MRI used a combination of a specially fabricated cyclone for the removal
of large particles followed by a Sierra Model 235 Hi-volume cascade impactor.1*
Two such samplers were used, located adjacent to the profiler tower, one at a
height of 1.5 m and the other at 4.5 m. These samplers operated at fixed flow
rates of 34 cubic meters per hour (20 cfm) with intake nozzles of selected
sizes such that the samples would be obtained at near isokinetic conditions.
PEI used the stacked filter concept developed by Cahill, et al.5 In the
PEI implementation of the concept, a stainless steel inlet screen having a 30
urn pore size was used to provide a supposed nominal 30gm fractionation point.
The screen was followed by two Nuclepore filters, the first of which had a pore
size (8 um) selected to provide a 2.Sum aerodynamic diameter cutoff while the
second, which had a pore size of 0.8um, served as the final filter.
EEM, TRC, and USS all used computer controlled scanning electron micro-
scopy (CCSEM) for measuring particle size distributions.6 USS performed their
analyses while Energy Technology Consultants (ETC) p«rforiMd th«
K>th RRM AM TftC. In this ethod. *«l*«e«Kt portion* of th» fU»*r«
t>«» profilers were removed, placed between finely perforated stainless steel
pcreens, and backwashed with acetone to remove the collected particles. The
particle in acetone suspension was then agitated, and the particles were
redeposited by vacuum filtration onto 0.2gm pore size Nuclepore filters. A
thin coating of carbon was evaporated onto the redeposited specimen after
mounting on SEM sample stubs. The SEM was used in conjunction with an
automated image analysis system to detect and determine the dimensions of the
projected surface of the particles. Particle volumes were estimated by
categorizing the particle's shapes and estimating overall dimensions based on
the measurements and assigned category. X-ray fluorescence emissions were used
to assign a probable composition class to each particle from which a density
26
-------
was assigned to it. After measuring approximately 1000 particles and assigning
estimated volumes and densities to them, a physical size-weight distribution
curve was constructed. Aerodynamic diameters were also estimated for each
particle based on the assigned category, the measured dimensions, and
aerodynamic shape factors.
In addition to the profilers and sizing systems described above, SoRI
performed a limited number of particle size distribution measurements using
Lundgren cascade impactors. These samplers were collocated with various
samplers of the other participants at different times during the test program.
They can be seen hanging under the upper MRI cyclone/impactor and above the
lower MRI cyclone/impactor in Figure 5, and below the second lowest PEI sampler
in Figure 8. The Lungren impactors sampled at a fixed flow rate of 5 cubic
meters per hour (3cfm). The cross-sectional area of the inlet was masked for
each test so that the sampling would be done at near isokinetic conditions.
These samplers provided a means of obtaining data with an inertial
(aerodynamic) classifier collocated with each of the other techniques. A
description of the Lundgren cascade impactor is given in Section 5.
27
-------
SECTION 4
DESCRIPTION OF TEST
The physical layout of the simultaneous field comparison test site is
illustrated, in Figure 16. The test site was located along a portion of a paved
slag haul road adjacent to and approximately 50 meters due south of Lake Mich-
igan. The road was made to simulate an unpaved, uncontrolled, road by covering
it with a 5 to 10 cm thick layer of fine aggregate. The material used was a
nondescript mixture of clay, iron ore, and boiler ash (composition information
is available in TRC's Field Comparison report). The average silt content of
the material (fraction passing through a 200 mesh screen) was approximately 10%
as measured by all five testing parties.
The average mass concentration (TSP) as measured throughout the testing
period by standard high-volume samplers (hi-vols) positioned three meters from
the roadway edge was 13.3 mg/m3. Thus, for all intents and purposes, the test
section represented an unpaved and uncontrolled road. In the middle of this
300 meter section of treated roadway, five testing stations were laid off as
described in Figure 17. This configuration minimized the distance between
sampling systems while allowing sufficient space to each system. Through the
use of identical stations on either side of the road the upwind and downwind
locations could be swapped across the road when the wind direction reverses as
it typically does near the lake.
In an effort to minimize the confounding effects which might be produced
by small portable electric generators usually used by the sampling parties, two
high capacity diesel generators were positioned on each side of the roadway and
approximately 25 meters west of the last sampling location (No. 5 in Figure
17). This also enabled the sampling equipment to be moved from any location to
another with a minimum amount of time and effort.
The traffic over the test road was a mixture of service vehicles (pickup
trucks) and dump trucks hauling solidified slag from the blast furnaces to the
28
-------
60SO-SIJ
?igur » 15. ''icy
site ifter covaclnq toad fit ., viaw is loolting *e3t,
-------
LAKE
MICHIGAN
I
GENERATOR
DIRT
ROAD
SAMPLING
LOCATIONS
POWER
PANEL
GENERATOR
N
»0lt-4ll
Figure 17. Physical layout of the Gary comparative fugitive emission test
site.
30
-------
slag yards and returning with loads of crushed slag. Average vehicle speed in
the test section normally ran as high as 50 to 60 kph. However, after covering
the test section with aggregate the average vehicle speed was reduced, because
of poor visibility, to approximately 40 kph.
Exposure profiling measurements of the test site were carried out over the
period June 11 to June 15, 1984. In this five day period a total of 11 data
runs were accomplished. In terms of downwind direction, six tests occurred
with the wind coming out of the north (off the lake) and the remaining tests
with the wind out of the south. Background concentrations for total particu-
late (TP) varied with wind direction. When the wind came from the north, off
Lake Michigan, the background concentration at 3 meters above the ground was
approximately 100 ug/m3. During those periods when the wind was from the south
the concentration was about 600 u9/"»3 The difference was due mostly to par-
ticulate produced south of the test site where considerable industrial
activity was concentrated. This presented no problem for these tests because
the TP concentrations produced by the test road were generally several times
greater.
The procedure used during each of the eleven tests was as follows: all
parties commenced sampling at the same time, during a break in traffic flow;
sampling continued until a predetermined number of vehicles had crossed the
testing section, and sampling was discontinued simultaneously during a break in
traffic flow. Following each test run, sufficient time was allowed for each
party to recycle their equipment preparatory to the next test.
One factor of great concern during pretest planning was the possibility of
nonuniformity of particle concentrations over the length of the road test sec-
tion. If the particle concentrations were in fact nonuniform with respect to
concentration or size distribution, then comparable results could not have been
expected from the various sampling systems, even if they were functionally
identical. Any potential effect of nonuniformity of conditions along the test
zone was partially obviated by rotating the participant's locations so that
each spent one day at each of the five test positions. The lack of a reference
31
-------
-------
-------
SECTIOH 5
PARTICLE SIZE MEASUREMENT
METHODS
Three fundamentally different particle sizing methods ware utilized by
the test participants. Three of the organizations - USS, TRC, and BEN - used
Computer Controlled Scanning Electron Microscopy (CCSEM); one, MRI, used iner-
tial particle sizing devices; and one, FBI, used a combination Of forms of
sieving and inertial devices. In addition, SoRI operated collocated inertial
sizing samplers at selected locations during some of the tests. Unfortunately,
the lack of standard reference methods for particle size'distribution measure-
ment under circumstances like those encountered in roadway emissions testing
makes it difficult to arrive at unequivocal conclusions regarding the actual
distributions during these tests and in the departures from those distributions
in the data obtained by the participants.
Cyclone/lmpactors
MRI used a combination of a specially fabricated cyclone for the removal
of large particles followed by a Sierra Model 235 Hi-volume cascade impactor
for size distribution measurement. The setup is illustrated in Figure 19. The
size distribution measurements are made separately from the exposure
measurement. The analysis of the particulate material collected by their
device is by gravimetric methods. Particle collection in the cyclone takes
place in two distinct zones: the body of the device, and the outlet tube. If
the body catch is taken alone, the cyclone cut diameter is 22pm (aerodynamic
diameter); if the body and outlet tube catches are combined, the cut diameter
is reduced to 14pm. The treatment of the cyclone data with respect to the
handling of the catch(es) and the two possible cut diameters will be deferred
to the discussion of the test results. With respect to the impactor portion of
the sampling system, MRI used a total of five impaction stages (plus filter) in
the impactor producing size cuts at 0.6pm, 1.2pm, 2.0pm, 4.1pm, and 10pm. It
should be noted, however, that in past tests only the first three of the five
34
-------
-------
stages were used. The omission or inclusion of the Last two stages can have an
effect on the interpretation of the data as will be discussed later. Three
such cyclone/inpactor samplers were used, one upwind to measure both
concentration and size distribution, and two «ore downwind for sice
distribution measurement only. The latter were located adjacent to the
profiler tower, one at a height of 1.5m and the other at 4.5m. These samplers
operated at fixed flow rates of 34 cubic meters par hour (20cfm) with intake
nozzles of selected sizes such that the samples would be obtained at near
isokinetic conditions.
SoRI obtained several samples with Lundgren cascade impactors during the
tests as well. These impactors, illustrated in Figure 20, are four stage (plus
filter) inertial classifiers7,8,9, providing size fractionation at diameters of
0.44 um, 1.4 um, 4.2 urn, and 14 pm as operated in these tests. The samples
were taken quasi-isokinetically with inlet nozzle cross-sections adjusted in
accordance with the mean wind speed at the start of the test, but the flow rate
was held constant at about 5 cubic meters per hour (3 cfm). Particle
collection in this impactor takes place on a moving plastic film coated with an
adhesive. The adhesive is used to insure retention of the impacted particles.
The particle size distribution is measured gravimetrically. However, by
collecting the particles on the moving film, overlapping of particles in the
collected sample is largely avoided, permitting individual particles collected
on each stage to be easily examined by microscope. Three of these samplers
were set up collocated at various times with samplers of EEM, MRI, and FBI as
illustrated for one case in Figure 21. Although the Lundgren impactor results
are used as a basis for some comparisons through the remaining discussions of
particle size distribution, they should not be construed as reference
standards.
Stacked filters
FBI used the stacked filter concept developed by Cahill.5 In the PEI
implementation of the concept, a stainless steel inlet screen having a 30 um
pore size is used to provide a supposed nominal 30 urn fractionation point. The
screen is followed by two Muclepore filters, the first of which has a pore size
(8 urn) selected to provide a 2.5 pm aerodynamic diameter cutoff while the
second, which has a pore size of 0.8 urn, serves as the final filter. These
36
-------
-------
Figure 21. Collocated Lundgren cascade impactor (left) and MRI Sierra cyclone
and inpactor (center).
-------
stacked filter units constituted the exposure samplers on the identical upwind
and downwind profiler towers. A schematic diagram of the sampler is shown in
Figure 22.
Scanning Electron Microscopy
EEM, TRC, and USS all used computer controlled scanning electron micro-
scopy (CCSEM) for measuring particle size distributions.6 USS performed their
own analyses while Energy Technology Consultants (ETC) performed the analyses
for both EEM and TRC. In this method, selected portions of the filters from
the profilers are removed, placed between finely perforated stainless steel
screens, and backwashed with acetone (ETC) or backblown with compressed air
(USS) to remove the collected particles. In ETC's procedure the acetone par-
ticle suspension is then agitated and the particles are redeposited by vacuum
filtration onto 0.2gm pore size Nuclepore filters. USS recollects the air
suspended particles on similar Nuclepore filters for analysis. A thin coating
of carbon is evaporated onto the redeposited specimen after mounting on SEM
sample stubs.
A SEN is used in conjunction with an automated image analysis system to
detect and determine the dimensions of the projected surface of the particles.
The same hardware and software is used by both ETC and USS for their respective
analyses. Particle volumes are estimated by categorizing the particle shapes
and estimating overall dimensions based on the measurements and assigned shape
category. X-ray fluorescence emissions are used to assign a probable composi-
tion class to each particle from which a density is assigned to it. After
measuring approximately 700 to 1200 particles and assigning estimated volumes
and densities to them, a physical diameter - weight distribution curve is con-
structed. Aerodynamic diameters are also estimated for each particle based on
the assigned category, the measured dimensions, and aerodynamic shape factors.
The method by which the particles to be measured are selected and the way in
which the decision is reached that a sufficient number of particles have been
measured are considered proprietary information by both the USS and ETC;
therefore this information was not available for use in critiqueing the CCSEM
methodology used in this application. Because the distributions are heavily
skewed to the small particle end of the size spectrum in terms of number
39
-------
BRACKET
TO TOWER
1.0 In. I.D.
INLET
ORIFICE
Figure 22. FBI stacked filter profiler head.
40
-------
population, the counting statistics for the large particles, which contain the
preponderance of the mass, would be poor unless a technique like stratified
counting were used. A technique like the latter may indeed have been used,
however this could not be verified.
RESULTS
The PEI profiling system provided data in two size fractions at each
profiler level for each test. MRI'a system provided data for each test in
seven size fractions at each of two heights for a location adjacent to their
profiler. The remaining three participants used CCSEM analysis of varying
numbers of the filters from their profilers. USS performed size analyses on
all of the filters from their profiler. TRC had CCSEM analyses performed on
two or three filters from selected heights from one test each day. EEM had
analyses performed on complete profiler filter sets from two tests which were
selected as representative of typical conditions through the week.
The results from disparate methods were not found to be mutually
consistent. Comparative particle size distribution results from Test 4 (PEI
data from Test 9) are shown in Figure 23, in which marked differences can be
seen to exist among the results from the various methods. The particle size
distributions measured by a given contractor were generally similar for all
tests. Test to test variations for each contractor were smaller than the
differences resulting from using different techniques. These differences and
their probable causes will be discussed in detail in the following paragraphs.
This test was selected for illustration because it is one for which inertial
sizing devices (Lundgren impactors) were collocated with samplers for which
CCSEM size analyses were performed (EEM). The results from this test for all
participants were generally comparable to those from the other tests for which
size data were available. No results are shown for TRC because they, like EEM,
had size analyses done for only a subset of the complete test series and
samples from this date were not included in that subset. The distributions for
all three users of CCSEM size analysis were quite similar to that illustrated
in Figure 23 throughout the entire test series. The PEI data in the illus-
tration were from a different day (Teat 9), since complete measurements, which
41
-------
N
55
2
99
98
95
90
iu 80
d
2
(0
(9
i
CD
*-
w
o
cc
IU
a.
70
60
BO
40
30
20
10
6
i rn r
A 80RI)
O MRI
EEM(
U88 I
. O PEOCO TEST 9
I I I I
I IT
0.4 0.60.81.0
6 8 10
20
40 60 80 100
Figure 23. Particle cize distributions as measured by
the various techniques.
1010-411
42
-------
-------
-------
the effective cutoff diameter to a value substantially smaller than 30pm in
much the same way the the SUB pores of the Nuclepore filter produce an inertial
separation near 2.SMB. Further, as particles are collected by the screen, the
openings are reduced in sice, thus making the screen become a progressively
more efficient filter. This filtration effect will then result in depressed
values for particle concentrations and emission factors for all particle sizes
that are assumed to be passed by the screen* It is worth noting that the
openings in the weaves of fabric filters used for particulate control devices
are comparable to or larger than the 30pm openings used here.10 In Figure 24a
the PEI data obtained in this test series is used to illustrate that the
screens undoubtedly do become more efficient filters as a test proceeds.
Data from other participants, which will be discussed shortly, indicated
that the particle size distribution remained fairly constant at heights between
1.5m and 4.5m throughout the test series, leading us to expect that the ratio
of TSP concentration to total concentration would be constant as well. The
particulate concentrations obtained from the hi-vols were previously shown to
be consistent with those measured by the profilers at a height of two meters.
Since the profiler and the hi-vol sampling periods were identical, the hi-vol
filter weights can be used as surrogates proportional to the missing total
weights for the PEI profiler samples at the two meter level, and the hi-vol
concentrations as surrogates for the total concentrations that would have been
obtained by the profiler. The points plotted in Figure 24a are the ratios of
the TSP (<30um) concentrations at the two meter height from FBI's profiler to
those of the collocated hi-vols plotted versus the corresponding hi-vol catch
weight. As can be seen, the ratio declines as the filter weight, and hence the
expected total catch weight of the profiler, increases. Hashes of the sampler
screens and inlets were obtained in each of tests 8, 9, and 10; consequently,
complete data on the system catches were available for those sets. The ratios
of apparent TSP to total concentrations from the lower samplers of FBI's
profiler, where the distribution might be expected to be fairly constant, are
shown in a similar way for tests 8, 9, and 10 in Figure 24b. Again, the ratios
tended to decline as the catch weights increased. This is indicative of
decreasing screen penetration as the profiler samplers loaded up during a
sampling run. This filtration effect, together with the misassignment of the
initial particle sice cutoff of the clean screen, is the probable explanation
45
-------
-------
of why PEI's particle size distribution curve, shown in Figure 23, fell below
those of the other participants*
The catch weights of the PBX samplers were very near the detection limits
in many cases so operation with shorter sampling times in order to avoid load-
ing the screens is not feasible. Even in the absence of these problems, the
PEI system can provide, at best, questionable data with respect to PMjQ emis-
sions. The system is purported to provide sice distribution information at
only two particle sizes - 2.5pm and 30urn, - both of which are relatively far
removed from 10pm. Thus, the 10pm fraction must be estimated by means of some
assumptions regarding the shape of the distribution at each height. Even the
most commonly used distribution form, the log-normal approximation, cannot be
used in PEI's normal practice as the total concentrations are not usually
determined but only the concentrations of particles nominally smaller than 30pm
and 2.5gm.
Scanning Electron Microscopy
Turning now to the CCSEM method, Figure 25 shows a particle size distri-
bution obtained by EEM/ETC using CCSEM methodology and the distribution meas-
ured by SoRl using a collocated Lundgren cascade impactor. Six such collocated
sampling runs were carried out and each resulted in particle size distributions
like that illustrated in Figure 25. The points shown for the impactor results
for sizes smaller than 15 ym are from gravimetric determinations. The results
which are shown for the impactor for sizes larger than 15um were obtained in
two ways. Both of these involved microscopic measurement of the size distribu-
tion of the material collected on the first stage of the impactor and then
apportioning the mass collected by the stage accordingly. In the first case
SoRl personnel performed manual counts of random regions of the first stage
substrate of the Lundgren sample using optical microscopy. These counts were
made in eight size classes based on Feret's diameter. Similar measurements
were made of all of the Lundgren first stage catches with total of approximate-
ly 500 particles being measured on each. Secondly, one half of each of the
Lundgren substrates were then sent to ETC for analysis of the catches by CCSEM.
Time and cost constraints led to only one of these actually being analyzed.
47
-------
-------
-------
filter. More importantly, the probability of transfer would be low (less than
50%) for any particle larger than 50pm with the probability dropping to zero at
210um. The omission of the particles in the size classes larger than 50 to 80
micrometers obviously introduces a bias into the final reported distribution,
the severity of which depends on the relative fraction of the mass actually
contained in that fraction. In the present instance as much as 25 to 50
percent of the total mass was probably .contained in particles physically larger
than 50 to 80pm. This mass would be reassigned to smaller particles in the
CCSEM results. Thus the reported OCSEM distributions would be expected to rise
much more steeply than the true distributions in the size region above about
Returning now to the results depicted in Figure 25, we find that although
the optical and CCSEM analyses are not in complete agreement, both sets of
results show non-negligible fractions of the total catch to consist of
particles larger than 50pm, 60% and 35% respectively, with 50% and 75%
respectively being larger than 30pm. These analyses were carried out directly
on the original specimens with no particle removal and transfer from the
original substrate taking place in either case. (The actual results of the
CCSEM measurements of the substrate showed less than one percent of the mass
consisted of particles smaller than 10pm, 22% in the 10 to 30pm range, 25% in
the 30 to 50pm range, 24% in the 50 to 100pm range, 20% in the 100 to 200jjffl
range, and 9% in the 200 to 500pm range.) In contrast, only 5% of the mass on
the collocated profiler sample was assigned to particles larger than 50pm and
only 20% was assigned to particles larger than 30pm by the CCSEM analysis.
Even though the size range of the particles on the substrate was limited as
compared to that of the profiler filter, measurements of a total of about 4500
particles were needed in order to obtain reasonable statistical validity in the
data. In contrast, only about 1000 particles were measured from the profiler
filter even though the size range was presumably much larger. This raises a
question as to whether the number of particles examined in the routine CCSEM
analyses is in fact sufficient for obtaining reliable data.
The lack of large particles in the CCSEM results for all of the profiler
samples may arise from three factors. First, the analysis may not include
sufficient particles to provide valid results for the statistically rare large
50
-------
particles. Second, the screens used in the saaple transfer and preparation
procedures for the CCSBM analyses result in a bias against the inclusion of
large particles. And third, the large particles may simply be lost from the
surfaces of the filters in handling and transport after the sampling takes
place. Large particles are not strongly attached to the filters and are
commonly observed to become detached from filter surfaces. In the cases of
hi-vol filters which are folded aid placed in envelopes for transport, the
large particles are frequently found to migrate to the vicinity of the fold or
off the filter into the envelope while in transit. Thus the lack of large
particles in the CCSEM results may simply be a result of a failure in
maintaining sample integrity and representativeness throughout delivery to the
analytical laboratory. BTC reported that visual scans of the surfaces of the
EEM and TRC profiler filters from these tests revealed no particles larger than
about lOOim present on any of the samples. (In the case of the Lundgren
samples, the adhesive used on the substrates insures that none are lost.) Thus
the validity of the application of the CCSB1 technique to filter samples
containing the large particles encountered in roadway emissions testing is
highly questionable.
At the other end of the size spectrum, it is unlikely that the techniques
used to remove the particles from the filters would be effective for removing
individual small particles. Thus the bulk of the small particles removed would
be in the form of agglomerates which would be counted as a lesser number of
larger particles. This would have the effect of transferring mass from the
fine particle end of the distribution to the midrange thereby depressing the
fine particle end of the distribution and again steepening the apparent distri-
bution. Manifestation of these potential problems can be illustrated using
data taken during these tests. Six Lundgren runs were made with the impactors
collocated with EEM profiler samplers. For these runs, the total particulate
concentrations measured by the profilers and by the impactors were within 6% of
each other on the average (range of 75% to 118%). At the same time, however,
the average of the ratio of the fine particle concentration as measured by
EEM/ETC to that measured by the impactors was only 18% (range 6% to 31%). At
10pm the average ratio of the BEN concentrations to those from the impactors
was 72% (range of 32% to 109%).
51
-------
Thus aside from the difficulties in estimating particle volumes and assig-
ning aerodynamic diameters to irregular, inhonogenous particles of the types
encountered here, there appear to be fundamental difficulties at both the large
and small ends of the size spectrum in the CCSEH analyses of roadway emission
samples. These problems render the OCSBN methodology, as used by the
participants of this test, unsuitable for applications of this type.
Cyclone/lmpactors
One of the primary advantages attributed to FBI's system and to the
CCSEM particles sizing methodology is that the methods permit the particle size
determinations to be made directly on the profiler samples. Thus the size
distributions can be obtained at each height from a sample which was acquired
isokinetically. The MRI methodology on the other hand provides data obtained
directly on an aerodynamic basis using the well characterized and widely
accepted techniques of inertial separation by cyclones and cascade impactors.
The particle size cuts span the range over which data are needed with adequate
resolution. However, MRI's data are obtained at only two heights (1.5m and
4.5m) from samples that are taken under only quasi-isokinetic conditions. The
lack of true isokinetic sampling should constitute no real problem except for
tests conducted under conditions of widely varying wind velocity. However, it
would be desirable to have particle size data for at least one more height, say
about 7m.
Although MRI measures the particle size distribution by a means for which
the results are inherently on an aerodynamic basis, there are potential prob-
lems related to the operation of the actual hardware used. First, particle
collection in the cyclone takes place in two distinct zones, the main body of
the cyclone and the outlet tuba. Particles caught in the main body are fully
contained by the body cavity and cannot be transferred to succeeding parts of
the sampler by jostling or vibration during and after sampling. However, a
large part of what is considered by MRI to be the cyclone catch in reality
collects on the wall of the exit tube and transform to the impactor. There is
no way to insure that this material will remain in place during sampling or
transport and cleanup after sampling. The cutoff diameter for the cyclone body
catch is about 22ym at the nominal operating flow rate while the cutoff diame-
52
-------
ter for the combination of the body and outlet tube is about 14ym. (These cut
diameters are based on the results of recent recalibrations11 that were per-
formed on MRI's behalf and differ significantly from previously reported
values.12 The new calibrations would alter the results of PM15 and PMIQ
emissions which were reported in some of MRI's previous work.) If the combined
body and outlet catches are used, as is MRI's practice, the possibility exists
that some part of the catch will be dislodged and transferred from the plus 14
urn part of the size distribution defined by the cyclone to the 10pm to I4ym
fraction collected by the first stage of the impactor. Uncertainties from this
can be avoided by separate recovery of the material from the cyclone body and
the outlet tube and treating the outlet tube catch as part of the catch of the
first impaction stage.
A second area of potential concern in MRI's sizing methodology is related
to their treatment of the impactor backup filter data. Particle bounce has
long been a problem plaguing cascade impactor usage and the Sierra Model 235 is
no exception in this regard. In developing their roadway emissions protocol,
MRI tried a number of approaches, including greases on the collection sub-
strates, the use of the inlet cyclone to reduce overloading and bounce problems
at the first stage, and operation at reduced flow rate - approaches which had
all proven useful in other cascade impactor operations in the past. Originally
the Sierra impactor was quoted by the manufacturer as being usable at the stan-
dard hi-vol flow rate of 68 cubic meters per hour (40 cfm); however, it soon
because apparent to MRI that particle bounce and blowoff were very real prob-
lems under those conditions.
At the time these problems were being resolved the complete five-stage
impactor was being used. Because the principal error resulting from particle
bounce is to transfer material which should be collected by the upper stages of
the impactor to the backup filter, MRI began compensating for the effect of
bounce by using the lesser of two weights for the backup filter catch. These
weights being either the actual filter catch or the average of the catches of
the two previous stages. The magnitude of the bounce errors were greatly
reduced by adding the cyclone and dropping the sampling rate to 34 cubic meters
per hour (20 cfm). But in spite of the reduction in the potential errors
-------
-------
PARTICLE SIZING RECOMMENDATIONS
*
The recommended procedure for meaauring the particle size distribution is
MRI'S cyclone/impactor technique with some modifications. With regard to field
operations.
With regard to field operations, first, a further reduction in the
sampling flow rate from 20cfm to 15cfm would help minimize errors from particle
bounce. Experience has shown that the product of jet velocities, V, and DSO's
of impactor stages must be limited to values below maximums which depend on the
type of substrate used. If the jet velocities are expressed in meters per
second and the DSO's are expressed in micrometers, the limits for the product
are 15pm»m/s for glass fiber substrates and 25ym m/s if adhesive coated
substrates are used. Operation of the impactor at a flow rate of 20cfm results
in V«/D50 products in excess of the limit for the glass substrates which are
used. Operation at IScfm reduces the products to acceptable values.
Alternatively, adhesive coated substrates could be used at the current 20cfm
flow rate. Second, potential errors resulting from the possible transfer of
material from the outlet tube of the cyclone to the first stage of the impactor
can be avoided by counting only the material collected in the body of the
cyclone as its catch. The outlet tube catch would then be combined with that
of the first impactor stage. At the current 20cfm flow rate, this would result
in a cyclone D50 of 22pm being used rather than the current 14pm value.
Lastly, while particle size was measured at the lower elevations (1.5m and
4.5m) an additional cyclone/impactor unit located at a height of about 7m would
provide additional information regarding the changes in size distribution as a
function of height within the dust plume.
With respect to data analysis, a better technique than the current MRI
procedure would be that commonly used in reducing impactor data from industrial
sources.13,11* In the latter, a spline fit is made to the cyclone/impactor
data in the cumulative percentage from of the distribution. The fit is made in
a manner that requires continuity in the slope of the curve and the solution is
55
-------
forced to be asymptotic to 100% at a diameter equal to the maximum diameter
present in the sample. The fitted curve is then used to interpolate or
extrapolate as needed to obtain the mass fractions in the selected size
intervals. This technique avoids the requirement of assuming a functional form
for the distribution and makes use of the complete data set rather than just
'two of the data points. :
56
-------
SECTION 6
EXPOSURES AND EMISSION FACTORS
Exposure ia defined as the total incident particulate mass, in the
selected size range, per unit area in a vertical plane parallel to the roadway.
The integral of exposure versus height is thus equal to the particulate emis-
sions per unit length of roadway during the test. The emission factor is
obtained by dividing by the number of vehicles passing the sampler during the
test. Accurate measurement of emission rates from the resuspension of road
dust by vehicular traffic thus hinges on the exposure measurements and subse-
quent integrations.
METHODOLOGIES
All participants in the program, both during the methods review (Phase
I) and the field test (Phase II), use variants of the profiling technique first
developed by MRI1. However, there are several differences among the various
organizations in the detailed manner with which the method is employed. Spe-
cific differences include the number of samplers employed, the heights at which
exposures are measured, the approach used to obtain representative (isokinetic)
samples, the method used to measure the background aerosol, the method used to
integrate the exposure curves to obtain emission factors, and the method used
to partition the emissions into the various particle size classes of interest.
The following paragraphs discuss all of these differences except for those
related to particle size. Discussions of particle sizing are contained in
Section 5 of this report.
Sampling Isokineticity
All organizations involved in both phases of the study attempt, to
varying extents, to perform the exposure sampling in an isokinetic fashion.
USS and TRC use servo systems with individual velocity sensors at each sampling
inlet to provide continuous adjustment of the flow rate for each sampler. ElA,
EEM, and PEI sample at fixed flow rates which are set based on the wind
velocity measured during a short time period immediately prior to the start of
57
-------
sampling. EIA and FBI do alter sampling flows in the event of sustained gross
changes in windspeed. The latter three organizations (EIA, EEM, and PEI) use a
measured windspeed at a single height coupled with an assumed functional form
for wind shear to obtain the velocities for setting the flow rates of the
individual exposure samplers. Finally, MRI periodically adjusts the sampling
flows based on an assumed logarithmic vertical windspeed distribution fit to
measured speeds at two heights. Those organizations not using continuous
individual servo controls to maintain isokinetic sampling flow rates correct
their resulting exposure values by factors based on the mean inlet velocity for
each sampler, the mean calculated windspeed at each sampling height, and the
particle size distribution. Tests for which the ratio of mean sample inlet
velocity to mean windspeed falls outside the range 0.8 to 1.3 are voided by all
participants. The velocity ratio approximates the limit of the ratio of true
exposure to measured exposure so the condition imposed insures that the errors
caused by deviations from isokinetic sampling will be relatively small.
All organizations maintain the pointing angle of the sampler inlets no
greater than 30 degrees from the mean wind direction (20 degree limits for EIA
and EEM, and a 15 degree limit for PEI). No correction is made for the mean
angle of the wind with respect to the road by any of the organizations. This
causes a positive bias in the results proportional to the secant of the wind
angle with respect to the road. At a 15 degree angle the error is 3 percent
and at 45 degrees it increases to 40 percent. All of the testing organizations
void tests in which the mean angle of the wind with respect to the road is
greater than 45 degrees, so 40 percent is the greatest error which can be
introduced by ignoring the angle factor*
Exposure Integration
Integration of the measured exposures for calculating emission rates are
performed using the trapezoidal rule by BBN, PHI, and USS. BIA has also used
this method in the past, but their Phase I current procedures documents sug-
gested that in the future they would use a least squares fit of a half-gaussian
distribution to the measured exposure profile. This procedure was proposed to
eliminate difficulties in extrapolation of exposures near the ground and above
58
-------
-------
FIELD TEST RESULTS
The results obtained by the five sampling organizations during the
five-day field comparison are summarized in Table 3. As shown in this table
eleven separate tests were carried out to characterize the emissions from the
simulated unpaved road. Rotation of the testing parties through the sample
locations was carried out on a daily basis as seen in columns 2 and 3. Data in
columns 6 through 18 were extracted from the Field Comparison Final Reports
produced by each of the five testing companies. As seen in these entries, some
test parameters were either not measured or not explicitly reported by the
various organizations. Some of the data variations between contractors for a
given test are explained by the fact that occasionally some of the testing
parties were late in starting and stopping their profiler systems and conse-
quently sampled over either a different number and/or type of vehicle passes.
Also there were some variations reported in the vehicle weight estimates, but
these differences were not significant.
\
In column 24 of Table 3, the values of particulate concentrations obtained
with standard high volume samplers are tabulated. In obtaining these data,
standard hi-vol samplers were placed at each sample location. These samplers
were not rotated with the sampling parties and served as monitors of the
changes that took place throughout the test for each sampling location.
Total Particulate Emission Factors
Of the five field test participants, four purportedly measure total
particulate emission factors while the fifth, PEI, normally measures only the
fraction nominally smaller than 30 ym. The average total emission factors for
the complete test series are summarized in Table 4. For the four contractors
whose methods normally provide total emission factors. the average values were
8.34, 8.25, 8.41, and 5.73 kg/VXT. The last value (5.73 kg/VKT) is
significantly different from the other three at the 90% confidence level. TRC,
whose data produced that result, stated in their report that their values were
generated from profiler filter catches only and did not include nozzle and
inlet washes. Although the omission of the latter material could explain the
60
-------
Table 3. COMPARATIVE FUGITIVE EMISSION TEST RESULTS.
IBT US? SHE 1EST 0* Hi
B WE UGU. MM- *i*
oner
warn u«« i«> m
- - CI/S) (dMU
MO. MO. mo snj mam son ME. ME. ME.
SIB. 1DP MD. MSI- B. 9fO
OCT. «M 45-7M <75* OK KOS
OVSt MMF) (I) (I) CO (I) (I) (VI) (>
0- f 10- -11 12 13 14 IS 1* 17-
MCTOB
ISP FP
nuo ne.s
w-m.
-U 1* 20- -a 22 23 24
I
I
2
2
2
14/12/4
4/Ort
tuun
M
11
11
11
11
11
von
70
4.3 2.*
14.5
11.1
5.2 124 17.4 1.0
4.3 2.*
144
5.2 124 17*.4 14
0.17
12.2*
3.* 5.4
1*4
*4 124 214
124*
124
124
94
12.2*
3.,
1*4
8*
134
114,
104
1*4
114
94
7.f
7.f
7.,
4.7
104
1*4
114
*.»
«
11.7
*.»
10.7
*.«
4.9
11.7
*.*
10.7
t.*2
4.5
11.7
11.0
10.3
*.4
'if.
14
12.1 34.4 224
U4 Ik, U.4
12.0 10 20.0
12.1 34.4 1*4
12.1 3441 20.7
5.1 37.5 74
4.1 334 ,4 .
54 32 74
5.1 374 11.1
9.13 3744 11.7
13.44 3247 2*4
13.75 324 25.1
134 git 114
'ill 324 &
13.1 1744
40 2.03
40 ,43 34,
40 10.24 04,
40
30 04*
31 44* 142
10 *42 5.15
30 44, 3.V7
3.74 2.50
245 2.02
44, 2.4*
341 2.17
0.3
1*74 KM
241 *U3*
243 144
04101*202.7
0.31 OS40J
040 Ultt.0
0.1* 429M
8K
2121.7
0.12 44*04
0,1,211044
0.49 142M.4
04*
94
M
14
34
14
14
74
04
74
0.4
7.0
0.2
19.2
4.71 43.11 *47
4.7 43.1 *4
44 ^ 9.,
M 43.1
&' & &
194 39.7 304
S3 "it S*
04 40 204
,ii1
14.1 314 214
X 4i5 »
949 4241 »4
94 42.1 M
9.7 3t4 54
*.3 12 ,4
*4 45.1 74
*40 494* 11.4
4.3 4S4 10.2
*.l 3,4 74
14.1 2M 1,.3
15.0 2* 2,4
14.7 34.0 2*4
50 44* 24*
91 941 449
51 44* 3.70
91 4.1*0 14*
4* 4.1, 3.H
*40 241
,.04 747
244 141 0.11
141 O.N 049
0.04 2007.1
4000.1
i
2.02 141 0.924 7255.4
3.** I.W 0.1* 1U404
SO 0.22 *.4t 3.19 2.71 0.40 704M
1, ,41 14*
3* 041 4.17
40 0.1* *4*
8 » &
9.77 3.7* 0.4,2*701.9
240 2.20 04234**04.9.
34, 1.7, O.U 2,2,34
tM*
-------
-------
-------
-------
-------
-------
-------
TABLE 6. (continued)
EXPOSURE, mg/cm2
TEST 5
ORGANIZATION TRC PEI USS MRI BEN
POSITION 12345
HEIGHT, m
10 ,0.43
9 0.34 0.42
8 0.71
7.5 1.14
7 1.02
6.5 1.09
6 0.78 1.44
5 2.56
4.5 2.08 1.21
4 2.20
3 3.92 2.15
2.5
2 3.51 3.35
1.5 0.60
1 4.49
Reported
mission
Factor,
kg/VKT 5.90 4.84 2.54 5.00
SoRI
Emission
Factor,
kg/VKT 5.65 4.56 3.73 4.40
6
EEM TRC PEI USS MRI
1 2345
0.45
-0.32 1.10
1.06
3.46
0.57
2.04
2.15 4.04
2.19
4.32 6.00
3.15
3.11 6.49
3.72 5.67
5.08
4.22
5.49 4.40 7.79 11.50
5.13 4.46 7.28 10.24
NOTEi SoRI Emission Factor it from integration by the trapezoidal rule with
linear extrapolation to eero at the top of the profiler and constant
exposure from the lowest height to the ground.
68
-------
TXBLB 6. (continued)
EXPOSURE, mg/cm2
TEST 7
EEM TRC FBI OSS MRI
ORGANIZATION 12 34 5
POSITION
HEIGHT, m
10 0.88
9 0.62 1.89
8 1.46
7.5 2.33
7 1.40
6.5 2.97
6 2.78 2.91
5 2.20
4.5 5.03 3.56
4 5.71
3 3.29 5.77
2.5
2 7.76 7.49
1.5 2.35
1 4.18
Reported
Emission
Factor,
kg/VKT 11.58 5.85 12.28 7.16
SoRI
Emission
Factor,
kg/VKT 10.80 5.64 11.44 7.30
8
MRI EEM TRC PEI USS
12 3 4 5
0.03
0.15 2.41 0.74
0.47
0.55
0.98
2.03
1.73 2.54
2.74 4.46
4.44 5.21
5.89
7.73 4.40
3.36
8.39 6.99
8.89
5.50 10.40
6.26 7.46 4.54 7.81 6.59
6.89 7.05 4.49 7.47 6.60
NOTEi SoRI Emission Factor i* from integration by the trapeeoidal rule with
linear extrapolation to eero at the top of the profiler and constant
exposure from the lowest height to the ground.
69
-------
-------
TABLE 6. (continued)
EXPOSURE, tag/cm2
TEST
11
ORGANIZATION
POSITION
HEIGHT,m
10
9
8
7.5
7
6.5
6
5
4.5
4
3
2.5
2
1.5
1
USS MRI EEM TRC PEI
123 45
0.08
0.39 0.11
0.64
0.74
0.83
1.24
2.40 2.06
1.95
4.35 5.19
3.69
6.52 3.72
7.14 5.86
5.32
4.78
Reported
Emission
Factor,
kg/VKT
9.61 8.68 8.16 5.74
SoRI
Emission
Factor,
kg/VKT
8.87 8.36 7.29 5.40
NOTEi SoRI Emission Factor is from integration by the trapezoidal rule with
linear extrapolation to terp at the top of the profiler and constant
exposure from the lowest height to the ground.
71
-------
K>
C3
U
10
8
uss
0 MRI
EEM
Q TRC
TEST1
1
3579
EXPOSURE (MG/-CM2)
11 13
Figure 27. Measured total exposures versus height for Test 1 for all contractors
providing total particulate exposure data.
-------
10
8
in
f uss
O MRI
0 EEM
Q TRC
TEST 2
1
3 5 79
EXPOSURE
-------
10
8
CP
+ uss
0 MRI
EEM
H TRC
TEST 3
1
3579
EXPOSURE
11 13
Figure. 29- Measured total exposures versus heiqht for Test 3 for all contractors
providing total particulate exposure data.
-------
10
8
£ 6
u
u 4
+ uss
0 MRI
EEM
Q TRC
TEST 4
1
3 5 7 9
EXPOSURE (MG/CM2)
11 13
Figure 30. Pleasured total exposures versus heiqht for Test 4 for all contractors
providing total particulate exposure data.
-------
-------
-------
-------
-------
10
8
09
O
uss
O MRI
EEM
Q TRC
| PEI
TEST 9
e
1
3 5 7 9 11
EXPOSURE
-------
-------
10
8
CD
K>
4
1
uss
0 MRI
EEM
Q TRC
TEST 11
3 5 7 9 11
EXPOSURE (MG/CM2)
13
Figure 37. Measured total exposures versus height for Test 11 for all contractors
providing total particulate exposure data.
-------
For some size fractions and test numbers the differences between the two
entries are small, but for others they are quite significant.
The SORI recalculated values of particulate size fraction shown in Table 8
were used to define size specific average measured emission factors for each of
the 11 tests. This was accomplished in the following manner. For each of the
11 tests, the measured TP emission factor values shown in column 19 of Table 3
were averaged to obtain the values shown in column 2 of Table 9. The entries
in the other columns of Table 9 (TSP, PMj-r PMIQ' *"* **) were obtained by
multiplying the average TP value (column 1, Table 9) by the appropriate SoRI
size fraction entry in Table 8. For example, for Test Number 1, the average
emission factor for TP is, from column 19 of Table 3t (8.56 + 9.33 + 10.24 +
7.1D/4 - 8.81 kg/VKT as shown in Table 9, column 2. The average emission
factor for TSP is calculated using the SoRI Test Number 1 entry in Table 9,
column 4 and the average TP abovet (0.453) x (8.81) « 3.99 kg/VKT as shown in
Table 9, column 3. The remaining entries in this table were calculated in a
similar manner. Thus, the values shown in Table 9 were used as a best estimate
against which the results of the five predictive equations were compared.
As mentioned in Section 6, there were differences in the reported input
values as shown in Table 3. Some of the data variations between contractors
for a given test are explained by the fact that occasionally some of the
testing parties were late in starting and stopping their profiler systems and
consequently sampled over either a different number and/or type of vehicle
passes. Also there were some variations reported in the vehicle weight
estimates. In an effort to remove the effects of these differences on the
predicted values of the emission factor/ each of the predictive equations were
evaluated using the entire set of test parameters (columns 10 and 13-18 of
Table 3) for each of the 11 tests. These predicted emission factor values for
TP, TSP, PM1S, PMj0» and FP were then averaged for each test number and
equation used. For example, the average predicted TP emission factor for Test
1 using EEM's equation was determined by using each of the five sets of
83
-------
-------
SECTION 7
t
MODELING FUGITIVE EMISSIONS
One objective of this study was to correlate the various predictive
equations for unpaved roads and to identify or develop, if feasible, an equa-
tion or set of equations that best account for observed differences. As a
final step in their analysis of the field test data, each participant was to
develop a predictive equation for the test road. The results of these efforts
along with an analysis of each is presented in the following paragraphs.
PREDICTIVE EMISSION FACTOR EQUATIONS
Three of the five participants (MRI, TRC, and OSS) used the existing
predictive equation described in Supplement No. 14 of the AP-42 document.16
This equation is
(D
where E - emission factor, kg/VKT
k - particle size multiplier, dimensionless
s silt content of road surface material, % passing a 200 mesh screen
S - mean vehicle speed, km/hr
W mean vehicle weight, Mg (metric tons)
w - mean number of wheels, dimensionless
p number of days with at least 0.2S4 mm of precipitation per year,
dimensionless.
For the current series of tests the factor involving p was omitted. The values
for the particle sice multiplier k as listed in the AP-42 document are: 0.80
«30 um), 0.57 «15 ym), 0.45 «10 pa), 0.28 «5 gn), and 0.16 «2.5 u">). No
value is listed for total particulate TO. Each of those organisations using
Equation (1) derived different values for the sise multiplier k.
85
-------
The other two participants (BEN and FBI) developed their own predictive
equation based pn the parameters measured during this test series. FBI used a
multiple linear regression analysis to develop the predictive equation
B - k(*)a(m)b (2)
where m - road surface moisture content, %, and the constants k, a, and b are
dependent upon the size fraction of the particulate being considered. BEM also
used a multiple regression technique to develop the following predictive equa-
tion
B - k(.)°-139(R)-0-203(w)0-267(W)0-395 (3)
where R « ambient relative humidity, percent; and the other parameters are as
defined previously. The values of the constants in. Equations (1), (2), and
(3) for all organizations are summarized in Table 7. It is important to notice
that FBI's equation is applicable only for predicting ISP, PN10, and PP emis-
sions. Also the term containing soil moisture in Equation (2) is applicable
only for the FP fraction*
In evaluating these predictive equations, the question arises as to which
of the measured emission factors should be used for comparison. In view of the
limitations of the particle sizing techniques as discussed earlier, probably
none are truly representative of the actual road conditions with MRI's results
being the most nearly representative.
For physical reasons, the particle sizing techniques used by NRI have a
higher probability of correctly describing the actual but unknown dust size
distributions. However, because of uncertainties in MRI's data reduction proc-
ess, the "as reported" size distributions are somewhat questionable. As a
result, the raw data reported by NRI (filter weighings, cyclone washes, etc.)
were independently reduced using the cascade impactor data reduction
techniques13 described at the end of Section 5. The results of these
recalculations are tabulated in Table 8 along with the values reported by NRI.
86
-------
-------
-------
test parameters (columns 10 and 13-18 of Table 3) as'inputs to Equation (3).
The five resulting values were averaged to represent a best theoretical
approximation ta the TP emission factor for Test Number 1 using EEM's predictor
equation. A similar process was carried out for each predictive equation and
all size fractions.
In order to compare these theoretically predicted emission factors to the
best estimates of those measured in the field, as shown in Table 9, a
correlation/regression analysis) was carried out. The results of applying this
analysis to EEM's equation are summarised in Figure 38. In this scatter
diagram the predicted values of TP emission factor are plotted versus the
measured values for each of the 11 tests. In this figure there are shown two
sets of data. The filled circles correspond to the paired values of the
average predicted emission factors uoing Equation (3) and the average measured
values shown in column 2 of Table 9. The open circles correspond to the paired
values of the predicted emission factor based on only one input data set per
test, and the values measured by that specific organization, in brief, the
filled circles depict how well EEM's equation can predict emission factors
based on the average measured values for total particulate from all con-
tractors. The open circles demonstrate the degree in which EEM's equation can
predict the actual data measured by EEM. For either set of points, if the
predicted and measured values of the emission factor were in a one-to-one
correspondence, the points would all lie along the solid diagonal line. The
other lines in the figures represent the least squares linear regression lines
for the data. The dotted line is the regression line for the filled symbols
and the simple broken line that for the open symbols. The values of the least
squares correlation coefficient* (r), and the coefficients of the regression
line in Figure 38 are shown in Table 10.
Thus in Figure 38, EEM's emission factor predictive equation for TP is
moderately well correlated with the average of the values measured (r 0.68)
and also with their own measured value (r « 0.61). Also, the slopes of the
regression line for the averaged data and their own data (as shown in Table 10)
are within 54% and 80%, respectively, when compared to the one-to-one
89
-------
-------
-------
-------
correspondence line. Because of the offset in the intercept values (-3.79 and
-6.48) EEH's predictive equation would under predict at low values of the TP
emissions and over-predict at higher values.
A ainilar analysis was applied to each of the five predictive equations
and the results shown in Table 10 and Figures 39 through 60. By examining the
values in Table 10 it can be seen that for the case of the TP emission factors,
the errors in the slopes of the regression lines range from 54% to 318% for the
averaged data and from 80% to 1182% for the individual data. Thus, even though
the correlation coefficients for the averaged data are very nearly the same,
the various predictive equations will in general over-predict the measured
values by varying amounts.
Although the data base used for these statistical analyses was limited,
some general comments regarding the various predictor models can be made.
e MRI's equation consistently over-predicts the results measured here for
all size fractions considered.
e TRC's equation predicts results similar to that of MRI as might be
expected. The over prediction at TP is less .than MRI's, and the TSP is
greater than MRI's results reflecting the different values of k used
(see Table 7). This is especially true for FP fraction. Here TRC's
equation severely under-predicts the average measured results while
predicting their own data quite well as seen in Figure 59. This is due
primarily to the fact that those using CCSQl techniques simply do not
see the entire fine particulate fraction of the dust.
o USS's equation predicts results very much the same as TRC and is expli-
cable for the same reasons*
e gat's aquation appears to do a better job of predicting the average test
results. This is primarily due to the fact that the equation was de-
rived based on the current data bam* (11 tests). How well this predic-
tive equation can be applied to previous data has not been determined.
93
-------
The severe under-prediction of the FP fraction as seen in Figure 56 is
again due to the inability of the CCSEM techniques to correctly describe
the particulate size distribution.
o PEI's equation consistently under-predicts the averaged data from these
tests while giving moderately good agreement to their own dfta. This
reflects the conclusions drawn earlier regarding the stacked filter
system.
In general the results of this analysis would suggest that either the
predictive equation of BEN or that used by MRI, TRC, and USS would adequately
represent the results of this test series if appropriate values were chosen for
the constants in Table 7. The FBI relation is seen to be inadequate primarily
because the underlying TSP data base was systematically low.
MODELING RECOMMENDATIONS
The utility of an emission factor predictive equation is not necessarily
that of predicting the emissions from a particular site in lieu of actual
measurements. In order that the equation be applicable over a wide range of
site locations and conditions, it should include as many of the relevant
parameters describing the site as possible. This requires that the predictive
equation be developed from a* large a data base as possible.
Equation (1) above is just such a relation. It was developed from a
fairly broad data base using multiple linear regression techniques. The data
base has some uncertainties particularly with regards to the particle size
distributions. These uncertainties most certainly cast some doubt upon the
precision of the values used for the particle size multiplier, k, in Equation
(1). However, without an extensive evaluation of the existing unpaved road
emissions data base, there are no justifications for invalidating the relation
used by MRI, TRC, and USS.
94
-------
This predictive equation (Equation^(2)) developed by PEI is considered
inadequate because it wa« derived ugin? an invalid TSP data base. Although the
relation (Equation (3)) developed by BEM did quite well predicing the results
of this test series, the data base froa which it was derived is too small to
give it much creditability for other ait«t without extensive validation.
In summary, the predictive emission factor equation used by MRI, TRC, and
USS and as described in the AP-42 manual^* is probably the most reliable
predictor of unpaved road emissions currently available. Because of past
problems in the particle sizing techniques, the values used for the size
multiplier (k) are less reliable than the overall equation. It is recommended
that these, values be re-evaluated in the future as the data base is expanded
with more reliable particle sice information.
95
-------
|