Emission Factor Documentation for AP-42
Section 13.2.2
Unpaved Roads
Final Report
For U. S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Emission Factor and Inventory Group
EPA Purchase Order 7D-1554-NALX
MRI Project No. 4864
September 1998
-------
Emission Factor Documentation for AP-42
Section 13.2.2
Unpaved Roads
Final Report
For U. S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Emission Factor and Inventory Group
Research Triangle Park, NC 27711
Attn: Mr. Ron Myers (MD-14)
Emission Factor and Inventory Group
EPA Purchase Order 7D-1554-NALX
MRI Project No. 4864
September 1998
-------
NOTICE
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency under Contract No. 68-D2-0159 and Purchase Order No. 7D-1554-
NALX to Midwest Research Institute. It has been reviewed by the Office of Air Quality Planning and
Standards, U. S. Environmental Protection Agency, and has been approved for publication. Mention of
trade names or commercial products does not constitute endorsement or recommendation for use.
11
-------
PREFACE
This report was prepared by Midwest Research Institute (MRI) for the Office of Air Quality
Planning and Standards (OAQPS), U. S. Environmental Protection Agency (EPA), under Contract
No. 68-D2-0159, Work Assignment No. 02 and Purchase Order No. 7D-1554-NALX. Mr. Ron Myers
was the requester of the work.
Approved for:
MIDWEST RESEARCH INSTITUTE
Roy Neulicht
Program Manager
Environmental Engineering Department
Jeff Shular
Director, Environmental Engineering
Department
September 1998
111
-------
iv
-------
TABLE OF CONTENTS
Page
1. INTRODUCTION 1-1
2. SOURCE DESCRIPTION 2-1
2.1 SOURCE CHARACTERIZATION 2-1
2.2 EMISSIONS 2-1
2.3 HISTORY OF THE UNPAVED ROAD EMISSION FACTOR EQUATION
IN AP-42 2-1
2.4 EMISSION CONTROL TECHNOLOGY 2-3
3. GENERAL DATA REVIEW AND ANALYSIS PROCEDURES 3-1
3.1 LITERATURE SEARCH AND SCREENING 3-1
3.2 METHODS OF EMISSION FACTOR DETERMINATION 3-1
3.2.1 Mass Emission Measurements 3-1
3.2.2 Emission Factor Derivation 3-4
3.3 EMISSION DATA AND EMISSION FACTOR QUALITY RATING SCHEME
USED FOR THIS SOURCE CATEGORY 3-8
4. REVIEW OF SPECIFIC TEST REPORTS 4-1
4.1 INTRODUCTION 4-1
4.2 REVIEW OF SPECIFIC DATA SETS 4-1
4.2.1 Reference 1 4-1
4.2.2 Reference 2 4-2
4.2.3 Reference 3 4-2
4.2.4 Reference 4 4-3
4.2.5 Reference 5 4-5
4.2.6 Reference 6 4-6
4.2.7 Reference 7 4-6
4.2.8 Reference 8 4-7
4.2.9 Reference 9 4-8
4.2.10 Reference 10 4-8
4.2.11 Reference 11 4-9
4.2.12 Reference 12 4-10
4.2.13 Reference 13 4-10
4.2.14 Reference 14 4-11
4.2.15 Reference 15 4-12
4.2.16 References 16-19 4-13
4.3 DEVELOPMENT OF CANDIDATE EMISSION FACTOR EQUATION 4-14
4.3.1 Validation Studies 4-23
4.4 DEVELOPMENT OF DEFAULT VALUES FOR ROAD SURFACE MATERIAL
PROPERTIES 4-27
4.5 SUMMARY OF CHANGES TO AP-42 SECTION 4-30
4.5.1 Section Narrative 4-30
4.5.2 Emission Factors 4-32
5. RESPONSES TO COMMENTS ON THE DRAFT SECTION 5-2
v
-------
-------
LIST OF FIGURES
Figure Page
2-1. Average control efficiencies over common application intervals for chemical
dust suppressants 2-5
3-1. Normal probability plot for PM-10 unpaved road emission factors 3-5
3-2. Normal probability plot for logarithms of PM-10 unpaved road emission factors 3-6
4-1. PM-10 residuals (log-scale) versus PM-10 emission factor (log-scale) 4-34
4-2. PM-10 residuals (log-scale) versus silt content (log-scale) 4-35
4-3. PM-10 residuals (log-scale) versus moisture content (log-scale) 4-36
4-4. PM-10 residuals (log-scale) versus average vehicle weight (log-scale) 4-37
4-5. PM-10 residuals (log-scale) versus average vehicle speed (log-scale) 4-38
4-6. PM-10 residuals (log-scale) versus average number of wheels (log-scale) 4-39
4-7. PM-10 residuals (log-scale) versus average vehicle speed <15 mph 4-40
4-8. PM-10 residuals (log-scale) versus average vehicle speed >15 mph 4-41
4-9. PM-30 residuals (log-scale) versus PM-30 emission factor (log-scale) 4-42
4-10. PM-30 residuals (log-scale) versus surface silt content (log-scale) 4-43
4-11. PM-30 residuals (log-scale) versus surface moisture content (log-scale) 4-44
4-12. PM-30 residuals (log-scale) versus average vehicle weight (log-scale) 4-45
4-13. PM-10 residuals (log-scale) versus average vehicle speed (log-scale) 4-46
4-14. PM-10 residuals (log-scale) versus average number of wheels (log-scale) 4-47
vii
-------
LIST OF TABLES
Table Page
3-1. QUALITY RATING SCHEME FOR SINGLE-VALUED EMISSION FACTORS . . . 3-10
3-2. QUALITY RATING SCHEME FOR EMISSION FACTOR EQUATIONS 3-11
4-1. SUMMARY INFORMATION - REFERENCE 1 4-48
4-2. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 1 . . . 4-48
4-3. SUMMARY INFORMATION - REFERENCE 2 4-48
4-4. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 2 . . . 4-49
4-5. SUMMARY INFORMATION - REFERENCE 3 4-50
4-6. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 3 . . . 4-50
4-7. SUMMARY INFORMATION - REFERENCE 4 4-51
4-8. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 4 . . . 4-52
4-9. SUMMARY INFORMATION - REFERENCE 5 4-54
4-10. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 5 . . . 4-55
4-11. SUMMARY INFORMATION - REFERENCE 6 4-56
4-12. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 6 . . . 4-57
4-13. SUMMARY INFORMATION - REFERENCE 7 4-59
4-14. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 7 . . . 4-59
4-15. SUMMARY INFORMATION - REFERENCE 8 4-60
4-16. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 8 . . . 4-61
4-17. SUMMARY INFORMATION - REFERENCE 9 4-64
4-18. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 9 . . . 4-64
4-19. SUMMARY INFORMATION - REFERENCE 10 4-65
4-20. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 10 . . 4-65
4-21. SUMMARY INFORMATION - REFERENCE 11 4-66
4-22. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 11.. 4-67
4-23. SUMMARY INFORMATION - REFERENCE 12 4-68
4-24. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 12 . . 4-69
4-25. SUMMARY INFORMATION - REFERENCE 13 4-70
4-26. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 13 . . 4-71
4-27. SUMMARY INFORMATION - REFERENCE 14 4-72
4-28. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 14 . . 4-73
4-29. SUMMARY INFORMATION - REFERENCE 15 4-76
4-30. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 15 . . 4-77
4-31. RESULTS OF CROSS-VALIDATION 4-78
4-32. PREDICTED VS. MEASURED RATIOS FOR NEW UNPAVED ROAD
EQUATION USING REFERENCE 15 TEST DATA 4-78
Vlll
-------
EMISSION FACTOR DOCUMENTATION FOR AP-42 SECTION 13.2.2
Unpaved Roads
1. INTRODUCTION
The U. S. Environmental Protection Agency (EPA) publishes the document Compilation of Air
Pollutant Emission Factors (AP-42) as its primary compilation of emission factor information.
Supplements to AP-42 have been routinely published to add new emission source categories and to update
existing emission factors. AP-42 is routinely updated by EPA to respond to new emission factor needs of
EPA, State and local air pollution control programs, and industry.
An emission factor is a value that attempts to relate the representative quantity of a pollutant
released to the atmosphere with an activity associated with the release of that pollutant. Emission factors
usually are expressed as the weight of pollutant divided by the unit weight, volume, distance, or duration of
the activity that emits the pollutant. The emission factors presented in AP-42 may be appropriate to use in
a number of situations, such as making source-specific emission estimates for area wide inventories for
dispersion modeling, developing control strategies, screening sources for compliance purposes, establishing
operating permit fees, and making permit applicability determinations. The purpose of this report is to
provide background information from test reports and other information to support revisions to AP-42
Section 13.2.2, Unpaved Roads.
This background report consists of five sections. Section 1 includes the introduction to the report.
Section 2 gives a characterization of unpaved road emission sources and a description of the technology
used to control emissions resulting from unpaved roads. Section 3 is a review of emission data collection
and emission measurement procedures. It describes the literature search, the screening of emission data
reports, and the quality rating system for both emission data and emission equations and methods of
emission factor determination. Section 4 details how the revised AP-42 section was developed. It includes
the review of specific data sets, a description of how candidate the emission equation was developed, and a
summary of changes to the AP-42 section. Section 5 presents the AP-42 Section 13.2.2, Unpaved Roads.
Throughout this report, the principal pollutant of interest is PM-10—particulate matter (PM) no
greater than 10 pimA (microns in aerodynamic diameter). PM-10 forms the basis for the current National
Ambient Air Quality Standards (NAAQS) for particulate matter. PM-10 thus represents the particle size
range that is of the greatest regulatory interest. Because formal establishment of PM-10 as the standard
basis for the NAAQS occurred in 1987, many earlier emission tests (and in fact the current version of the
unpaved road emission factor) have been referenced to other particle size ranges, such as,
TSP Total Suspended Particulate, as measured by the standard high-volume (hi-vol) air sampler. Total
suspended particulate, which encompasses a relatively coarse size range, was the basis for the
previous NAAQS for PM. Wind tunnel studies have shown that the particle mass capture
efficiency curve for the hi-vol sampler is very broad, extending from 100 percent capture of
particles smaller than 10 micrometers to a few percent capture of particles as large as
100 micrometers. Also, the capture efficiency curve varies with wind speed and wind direction,
relative to roof ridge orientation. Thus, the hi-vol sampler does not provide definitive particle size
information for emission factors. However, an effective cutpoint of 30 pim aerodynamic diameter
is frequently assigned to the standard hi-vol sampler.
1-1
-------
SP Suspended Particulate, which is often used as a surrogate for TSP, is defined as PM with an
aerodynamic diameter no greater than 30 /iinA. SP may also be denoted as "PM-30."
PM-2.5 PM with an aerodynamic diameter no greater than 2.5 /iinA.
The EPA promulgated new PM NAAQS based on PM-2.5, in July 1997.
Because of the open source nature of unpaved roads, ambient particulate matter samplers are
usually most applicable to emission characterization of this source category. Nevertheless, one may adapt
traditional stack source sampling methods to unpaved roads. In that case, "total PM" refers to the amount
of PM collected in EPA Method 5 plus EPA Method 202 sampling trains. "Total filterable PM" denotes
the filter catch in the Method 5 train. Similarly, "PM-10" refers to the sum of the catch in EPA Method
201A and Method 202 trains, while "filterable PM-10" corresponds to the filter catch in Method 201A.
1-2
-------
2. SOURCE DESCRIPTION
2.1 SOURCE CHARACTERIZATION1
Particulate emissions occur whenever vehicles travel on unpaved roads. Dust plumes trailing
behind vehicles on unpaved roads are a familiar sight in rural areas of the United States. Many industrial
areas also have active unpaved roads. When a vehicle travels an unpaved road, the force of the wheels on
the road surface causes pulverization of surface material. Particles are lifted and dropped from the rolling
wheels, and the road surface is exposed to strong air currents in turbulent shear with the surface. The
turbulent wake behind the vehicle continues to act on the road surface after the vehicle has passed.
2.2 EMISSIONS12
The emission of concern from unpaved roads is particulate matter (PM) including PM less than
10 microns in aerodynamic diameter (PM-10) and PM less than 2.5 microns in aerodynamic diameter
(PM-2.5). The quantity of dust emissions from a given segment of unpaved road varies linearly with the
volume of traffic. Field investigations also have shown that emissions depend on correction parameters
that characterize (a) the condition of a particular road and (b) the associated vehicle traffic. Parameters of
interest in addition to the source activity (number of vehicle passes) include the vehicle characteristics (e.g.,
vehicle weight), the properties of the road surface material being disturbed (e.g. silt content, moisture
content), and the climatic conditions (e.g., frequency and amounts of precipitation).
Dust emissions from unpaved roads have been found to vary directly with the fraction of silt in the
road surface material. Silt consists of particles less than 75 /im in diameter, and silt content can be
determined by measuring the proportion of loose dry surface dust that passes through a 200-mesh screen,
using the ASTM-C-136 method.
2.3 HISTORY OF THE UNPAVED ROAD EMISSION FACTOR EQUATION IN AP-42
The current version of the AP-42 unpaved road emission factor equation for dry conditions has the
following form:1
E = Emission factor, pounds per vehicle-mile-traveled, (lb/VMT)
k = Particle size multiplier (dimensionless)
s = Silt content of road surface material (%)
S = mean vehicle speed, miles per hour (mph)
W = mean vehicle weight, ton
w = mean number of wheels (dimensionless)
The AP-42 discusses how Equation 2-1 can be extrapolated to annual conditions through the
simplifying assumption that emissions are present at the "dry" level on days without measurable
(2-1)
where:
2-1
-------
precipitation and conversely, are absent on days with more than 0.01 in. (0.254 mm) of precipitation.
Thus, the emission factor for annual conditions is:
E = k 5.9
jl A w
12 j 30/ 3
0.7
0.5
365-p
365
(2-la)
where all quantities are as before and:
p = number of days with at least 0.254 mm (0.01 in.) of precipitation per year
The particle size multiplier "k" for different particulate size ranges is shown below.
Aerodynamic Particle Size Multiplier (k) for Equation 2-1
<30(jma
1.0
<30(jm <15(jm <10(jm <5(jm
0.80 0.50 0.36 0.20
<2.5(jm
0.095
"Stoke's diameter
The earliest emission factor equation for unpaved roads first appeared in AP-42 in 1975. The
current version of the emission factor equation appeared in 1983 as part of Supplement 14 to the third
edition of AP-42.
The earliest version of the unpaved road emission factor equation included the first two correction
terms shown in Equation 2-1 (i.e., silt content and mean vehicle speed). However, the data base for that
version was limited to tests of publicly accessible unpaved roads travelled by light-duty vehicles and had a
small range of average travel speeds (30 to 40 mph).3 Subsequent emission testing (especially roads at iron
and steel plants) expanded the ranges for both vehicle weight and vehicle speed. In 1978, a modified
equation that included silt, speed, and weight was published in an EPA report.4 In 1979, the current
version (Equation 2-1) was first published;5 it incorporated a slight reduction in the exponent for vehicle
weight and added the wheel correction term.
Although the emission factor equation for unpaved roads has been modified over the past 20 years,
all versions have important common features. All were developed using multiple linear regression of the
suspended particulate emission factor against correction parameters that describe source conditions. The
silt content has consistently been found to be of critical importance in the predictive equation. The first
version of the predictive equation (and each subsequent refinement) included a roughly linear (power of 1)
relationship between the emission factor and the road surface silt content.3
In addition to the unpaved road emission factor equation discussed above, other studies have been
undertaken to model emissions from unpaved road vehicular traffic. For example, the 1983 background
3 Note that during the 1970's, the exponent for the silt content was rounded to unity because of the greater
computational ease. Recall that this equation predated inexpensive calculators with "x to the y"
capability.
2-2
-------
document for this section of AP-42 lists three other candidate emission factor equations.6 Equation 2-1 was
recommended over the other candidates on the basis of its wider applicability.
Additional studies addressed emissions from restricted classes of unpaved roads. In particular, a
1981 report included separate emission factors for (a) light-to medium-duty traffic, and (b) haul trucks on
unpaved roads for use at western surface coal mines.7 Neither equation bore resemblance to the generic
unpaved road emission factor (Equation 2-1). A 1991 study (described in Section 4 of this report)
addressed emissions due to relatively high-speed traffic on publicly accessible roads in Arizona.2
Furthermore, in response to Section 234 of the Clean Air Act Amendments, the western surface coal
mining emission factors were reexamined.8'9 Results from that study are also described in Section 4.
2.4 EMISSION CONTROL TECHNOLOGY11011
Controls to reduce particulate emissions from unpaved roads fall into three general categories as
follows: source extent reductions, surface improvements, and surface treatment. Each of the categories is
discussed below.
Source extent reductions limit the amount of traffic to reduce particulate emissions. The emissions
directly correlate to the vehicle miles traveled on the road. An example of limiting traffic is restricting road
use to certain vehicle types. The iron and steel industry, for example, has instituted some employee busing
programs to eliminate a large number of vehicle passes during shift changes.
Surface improvements offer a long term control technique. Paving is a surface improvement that is
a highly effective control, but can be cost prohibitive especially on low volume roads. From past
experience, paving has an estimated 99 percent control efficiency for PM-10. Control efficiencies
achievable by paving can be estimated by comparing emission factors for unpaved and paved road
conditions. The predictive emission factor equation for paved roads, given in AP-42 Section 13.2.1,
requires estimation of the silt loading on the traveled portion of the paved surface, which in turn depends on
(a) the intensities of deposition processes that add silt to the surface, and (b) whether the pavement is
periodically cleaned.
Other surface improvements include covering the road surface with a new material of lower silt
content. For example a dirt road could be covered with gravel or slag. Also, regular maintenance practices,
such as grading of gravel roads, help to retain larger aggregate sizes on the traveled portion of the road and
thus help reduce emissions. The amount of emissions reduction is tied directly to the reduction in surface
silt content.
Surface treatments include control techniques that require reapplication such as watering and
chemical stabilization. Watering increases the road surface moisture content, which conglomerates the silt
particles and reduces their likelihood to become suspended when a vehicle passes over the road surface. The
control efficiency of watering depends upon (a) the application rate of the water, (b) the time between
applications, (c) traffic volume during the period, and (d) the meteorological conditions during the period.
Chemical stabilization suppresses emissions by changing the physical characteristics of the road
surface. Many chemical unpaved road dust suppressants form a hardened surface that binds particles
together. As a result of grinding against the improved surface, the silt content of loose material on a highly
2-3
-------
controlled surface may be substantially higher than when the surface was uncontrolled. Thus, the predictive
emission factor equation for unpaved roads usually cannot be used to estimate emissions from chemically
stabilized roads.
Although early studies of unpaved road dust control showed a strong correlation between efficiency
and the silt content of the surface material, this correlation was based on the very high (e.g., >90 percent)
control efficiencies and very low silt values typically found over the first few days after application.
Because these conditions represent only a small, restricted portion of the range of possible conditions
encountered during a control application cycle, the high degree of correlation was misleading.
Later study of long-term control indicated no significant correlation between silt content and control
efficiency. In addition, fairly high (-50 percent) control efficiencies were found to occur with silt contents
at or above the uncontrolled level. Because of these findings, attention turned to the use of the amount of silt
per unit area (i.e., "silt loading") as a performance indicator.
A long-term study of the performance of 4 chemical dust suppressants of interest to the iron and
steel industry was conducted through EPA in 1985. This study found that although emission factors varied
over an order of magnitude, the silt loading values varied over two orders of magnitude, and did not appear
to follow a specific trend with time. Furthermore, the results for the different suppressants tended to be
clustered together; this indicated that the various suppressant types did not affect silt loading in the same
way.
The control effectiveness of chemical dust suppressants depends on the dilution rate, application
rate, time between applications, and traffic volume between applications. Other factors that affect the
performance of dust suppressants include the vehicle characteristics (e.g., average vehicle weight) and road
characteristics (e.g., bearing strength). The variabilities in the above factors and in individual dust control
products make the control efficiencies of chemical dust suppressants difficult to calculate. Past field testing
of emissions from controlled unpaved roads has shown that chemical dust suppressants provide a PM-10
control efficiency of about 80 percent when applied at regular intervals.
Because no simple relationship of control efficiency with silt or silt loading could be found to
successfully model chemical dust suppressant performance, other types of performance models were
developed based on the amount of chemical applied to the road surface. Figure 2-1 presents control
efficiency relationships for petroleum resins averaged over two common application intervals, 2 weeks and
1 month.10
2-4
-------
Ground Inventory
(liters/square meter)
0.25
0.5
0.75
0.25
0.5
0.75
Note: Averaging periods (2 weeks or 1 month)
refer to time between applications
2 weeks
1 month
Total Particulate
2 weeks
-1 month
Particulate <10 |imA
0.05
0.1
0.15 0.2 0.25 0
(gallons/square yard)
Ground Inventory
0.05
o.l
0.15
0.2 0.25
Figure 2-1. Average control efficiencies over common application intervals for chemical dust suppressants.
-------
REFERENCES FOR SECTION 2
1. Compilation of Air Pollutant Emission Factors, AP-42, Fifth Edition, U. S. Environmental
Protection Agency, Research Triangle Park, NC, January 1995.
2. Unpaved Road Emission Impact, Final Report, Arizona Department of Environmental Quality,
Phoenix, AZ, March 1991.
3. Development of Emission Factors for Fugitive Dust Sources, EPA-450/3-74-037, Office of Air
Quality Planning and Standards, U. S. Environmental Protection Agency , Research Triangle Park,
NC, June 1974.
4. Fugitive Emissions from Integrated Iron and Steel Plants, EPA-600/2-78-050, Office of Research
and Development, U. S. Environmental Protection Agency , Research Triangle Park, NC, March
1978.
5. Iron and Steel Plant Open Source Fugitive Emission Evaluation, EPA-600/2-79-103, Office of
Energy, Minerals, and Industry, U. S. Environmental Protection Agency , Research Triangle Park,
NC, May 1979.
6. Fugitive Dust Emission Factor Update for AP-42, EPA Contract No. 68-02-3177, Assignment 25,
Industrial Environmental Research Laboratory, U. S. Environmental Protection Agency , Research
Triangle Park, NC, September 1983.
7. Improved Emission Factors for Fugitive Dust from Western Surface Coal Mining Sources, U. S.
Environmental Protection Agency , Research Triangle Park, NC, EPA Contract No. 68-03-2924,
Assignment 1, July 1981.
8. Review of Surface Coal Mining Emission Factors, EPA-454/R-95-007, Office of Air Quality
Planning and Standards, U. S. Environmental Protection Agency, Research Triangle Park, NC, July
1991.
9. Surface Coal Mine Emission Factor Study, U. S. Environmental Protection Agency, EPA Contract
No. 68-D2-0165, Assignment 1-06, Research Triangle Park, NC, January 1994.
10. Fugitive Dust Background Document and Technical Information Document for Best Available
Control Measures, EPA-45 0/2-92-004, Office of Air Quality Planning and Standards, U. S.
Environmental Protection Agency, Research Triangle Park, NC, September 1992.
11. Control of Open Fugitive Dust Sources, EPA-68-02-4395, Assignment 14, Office of Air Quality
Planning and Standards, U. S. Environmental Protection Agency, Research Triangle Park, NC,
1988.
2-6
-------
3. GENERAL DATA REVIEW AND ANALYSIS PROCEDURES
3 .1 LITERATURE SEARCH AND SCREENING
To reduce the amount of literature collected to a final group of references from which emission
factors could be developed, the following general criteria were used.
1. Emissions data must be from a primary reference.
a. Source testing must be from a referenced study that does not reiterate information from previous
studies.
b. The document must constitute the original source of test data. For example, a technical paper
was not included if the original study was contained in the previous document. If the exact source of the
data could not be determined, they were eliminated.
2. The referenced study must contain test results based on more than one test run.
3. The report must contain sufficient data to evaluate the testing procedures and source operating
conditions.
A final set of reference materials was compiled after a thorough review of the pertinent reports,
documents, and information according to these criteria.
3.2 METHODS OF EMISSION FACTOR DETERMINATION2
Fugitive dust emission rates and particle size distributions are difficult to quantify because of the
diffuse and variable nature of such sources and the wide range of particle size involved including particles
which deposit immediately adjacent to the source. Standard source testing methods, which are designed for
application to confined flows under steady state, forced-flow conditions, are not suitable for measurement of
fugitive emissions unless the plume can be drawn into a forced-flow system. The following presents a brief
overview of applicable measurement techniques.
3.2.1 Mass Emission Measurements
Because it is usually impractical to enclose open dust sources or to capture the entire emissions
plume, only the upwind-downwind and exposure profiling methods are suitable for measurement of
particulate emissions from most open dust sources.3 These two methods are discussed separately below.
The basic procedure of the upwind-downwind method involves the measurement of particulate
concentrations both upwind and downwind of the pollutant source. The number of upwind sampling
instruments depends on the degree of isolation of the source operation of concern (i.e., the absence of
interference from other sources upwind). Increasing the number of downwind instruments improves the
reliability in determining the emission rate by providing better plume definition. In order to reasonably
define the plume emanating from a point source, instruments need to be located at two downwind distances
3-1
-------
and three crosswind distances, at a minimum. The same sampling requirements pertain to line sources
except that measurement need not be made at multiple crosswind distances.
Net downwind (i.e., downwind minus upwind) concentrations are used as input to dispersion
equations (normally of the Gaussian type) to back calculate the particulate emission rate (i.e., source
strength) required to generate the pollutant concentration measured. Emission factors are obtained by
dividing the calculated emission rate by a source activity rate (e.g., number of vehicles, or weight of material
transferred per unit time). A number of meteorological parameters must be concurrently reported for input
to the dispersion equations. The test report should describe what constitutes acceptable meteorological
conditions.
At a minimum, the wind direction and speed must be recorded on-site and should remain within
acceptable ranges. When the upwind/downwind technique is applied to unpaved roads, the test report must
describe the mean angle of the wind relative to the road centerline.
As part of a sound test methodology, source activity parameters should be recorded, including the
vehicle weights and vehicle speeds. The surface material at the test location (specifically, its silt and
moisture contents) should also be characterized following guidance of AP-42 Appendicies C.l and C.2.
While the upwind-downwind method is applicable to virtually all types of sources, it has significant
limitations with regard to development of source-specific emission factors. The major limitations are as
follows:
1. In attempting to quantify a large area source, overlapping of plumes from upwind (background)
sources may preclude the determination of the specific contribution of the area source.
2. Because of the impracticality of adjusting the locations of the sampling array for shifts in wind
direction during sampling, it cannot be assumed that plume position is fixed in the application of the
dispersion model.
3. The usual assumption that an area source is uniformly emitting does not allow for realistic
representation of spatial variation in source activity.
4. The typical use of uncalibrated atmospheric dispersion models introduces the possibility of
substantial error (a factor of three according to Reference 4) in the calculated emission rate, even if the
stringent requirement of unobstructed dispersion from a simplified (e.g., constant emission rate from a single
point) source configuration is met.
On an even more fundamental level, typical traffic volumes on unpaved roads are far too low to
represent the road as a steady, uniformly emitting line source for dispersion analysis purposes. A far better
representation (but one which, unfortunately, is not available at this time) would view the unpaved road
source as a series of discrete moving point sources.
Just as importantly, it is not clear that "cosine correction" used to account for the effect that an
oblique wind direction has on line sources is applicable to the case of an unpaved road. As the plume is
released, dispersion occurs in all three cartestian coordinate directions. Only dispersion in the direction
3-2
-------
parallel to the plume centerline would be negligible. Depending on the direction a vehicle is traveling, an
oblique wind would appear to dilute or "concentrate" the plume mass seen by the samplers, as compared to
the case of a perpendicular wind. Correction for each plume depends upon the magnitude and direction of
the wind relative to vehicle velocity vector.
The other measurement technique, exposure profiling, offers some distinct advantages for source-
specific quantification of fugitive emissions from open dust sources. The method uses the isokinetic
profiling concept that is the basis for conventional (ducted) source testing. The passage of airborne
pollutant immediately downwind of the source is measured directly by means of simultaneous multipoint
sampling over the effective cross section of the fugitive emissions plume. This technique uses a mass-
balance calculation scheme similar to EPA Method 5 stack testing rather than requiring indirect calculation
through the application of a generalized atmospheric dispersion model. As with other testing methodologies,
source activity must be recorded as part of a sound exposure profiling program.
For measurement of nonbuoyant fugitive emissions, profiling sampling heads are distributed over a
vertical network positioned just downwind (usually 5 m) from the source. If total particulate emissions are
to be measured, sampling intakes are pointed into the wind and sampling velocity is adjusted to match the
local mean wind speed, as monitored by anemometers distributed over heights above ground level.
Note that, because the test method relies on ambient winds to carry emissions to the sampling array,
acceptance criteria for wind speed/direction are necessarily based on antecedent monitoring. That is, the
immediate past record is used to determine acceptability for the current or upcoming period of time. As a
practical matter, this means that wind monitoring must be conducted immediately before starting an
exposure profiling test. The test methodology must also present what guidelines govern stopping/suspending
a test for unacceptable wind conditions. For example, testing should be suspended if the angle between the
mean wind direction and the perpendicular to the road centerline exceeds 45 ° for two consecutive 3- to 10-
min averaging period. Similarly, testing should be suspended if the mean wind speed falls below 4 mph or
exceeds 20 mph for more than 20 percent of the test duration.
The size of a sampling grid needed to conduct exposure profiling tests of an unpaved road depends
on several factors, including size/speed of the vehicles traveling the road; expected wind speed; width of the
road; and the sampler separation distance from the road. Particulate sampling heads should be
symmetrically distributed over the concentrated portion of the plume containing roughly 90 percent of the
total mass flux (exposure). In general, the best way to judge the sampling height is to view the plumes being
generated from vehicle passes over the road. Past field studies using exposure profiling also provide a good
means to establish the necessary size for the sampling grid.
Grid size adjustments may be required based on the results of preliminary testing. To be reasonably
certain that one is capturing the entire plume, one needs to demonstrate that the concentration (or, more to
the point, the mass flux) decreases near the top of the sampling array. As a practical matter, this means
that individual samplers be deployed so that results can be compared from one height to the next.
Specifically, use of a manifold to (a) collect air samples at different heights but (b) to route the emissions to
a common duct for measurement cannot provide direct evidence of the sufficient height of the sampling
array.
3-3
-------
Use of dispersion algorithms to determine sampling heights suffers from the same limitations as
noted earlier in connection with the upwind/downwind method. That is, typical traffic volumes on unpaved
roads are far too low to represent the road as a steady, uniformly emitting line source for dispersion
purposes. Just as importantly, it is not clear that "cosine correction" used to account for the effect that an
oblique wind direction has on line sources is applicable to the case of an unpaved road.
To calculate emission rates using the exposure profiling technique, a conservation of mass approach
is used. The passage of airborne particulate (i.e., the quantity of emissions per unit of source activity) is
obtained by spatial integration of distributed measurements of exposure (mass/area) over the effective cross
section of the plume. The exposure is the point value of the flux (mass/area/time) of airborne particulate
integrated over the time of measurement.
3.2.2 Emission Factor Derivation
Usually the final emission factor for a given fugitive source operation, as presented in a test report,
is derived simply as the arithmetic mean of the individual emission factors calculated from each test of that
source. Frequently, test reports present the range of individual emission factor values.
Although test reports often present an arithmetic mean emission factor for a single specific source, it
is important to recognize that the population of all unpaved road emission factors is better characterized as
log-normally than as (arithmetic) normally distributed. That is to say, the logarithms of the emission factor
are themselves normally distributed. This can be seen in Figures 3-1 and 3-2, which present normal
probability plots for both a set of PM-10 unpaved road emission factors and the logarithms of the factors.
Note that the plot of the log-transformed data results in a straight line, which indicates normality. In
Figures 3-1 and 3-2 the ordinate (y-axis) is sometimes termed the "z-score." The z-score is found by
ranking the data in ascending order and dividing each value's rank by the total number N of data points:
Proportion = (RANK - 0.5)/N
The z-score represents the value of the standard normal distribution (i.e., mean equal to 0 and a standard
deviation of 1) whose cumulative frequency equals the proportion found. In practical terms, a sample from
a normally distributed population will exhibit a reasonably straight line in this type of plot.
To characterize emissions from unpaved roads, one could use the geometric mean emission factor
(i.e., the arithmetic mean of the log-transformed data). However, attempting to characterize emissions from
data spanning several orders of magnitude, from extremely large mine haul trucks to light-duty vehicles on
county roads, with a single valued emission factor would be futile. Alternatively, one could construct a
series of different single-valued mean emission factors, with each mean corresponding to a different category
of unpaved roads. For example, one might derive a factor for use with passenger cars on rural roads,
another factor for haul trucks, and a third for plant traffic at an industrial facilities. This "subcategory
mean" approach, as applied to emissions from unpaved roads, has several drawbacks.
The approach ignores the similarities in the dust-emitting process between subcategories of unpaved
road travel. Despite the contrast in scale between haul trucks and small vehicles, the general physical
process is the same. The vehicle's tires interact with the surface material, directly injecting particles into the
atmosphere while at the same time pulverizing the material. Furthermore, the passage of the vehicle results
3-4
-------
EXPECTED
VALUE
NORMAL PROBABILITY PLOT, N
0 10 20 30
LBVMT
4 CASES WITH MISSING VALUES EXCLUDED FROM PLOT
Figure 3-1. Normal probability plot for PM-10 unpaved road emission factors.
3-5
-------
EXPECTED
VALUE
NORMAL PROBABILITY PLOT, N - 212
2
1
0
-1
-2
-3
4 CASES WITH MISSING VALUES EXCLUDED FROM PLOT
Abscissa consists of natural logarithm of emission factor in Ib/vmt.
Figure 3-2. Normal probability plot for logarithms of PM-10 unpaved road emission factors.
-------
in a wake which also entrains particulate matter. Admittedly, the intensity of any process will depend on
many factors, such as: vehicle weight, number of wheels, tread design, tire footprint pressure, clearance
height, vehicle speed. The approach undertaken in this study (as described later in this section) attempts to
capture the essential traffic differences in a few easily quantified vehicle parameters.
Beyond variations in vehicle scale, unless one devises many different classifications, the
"subcategory mean" technique cannot capture important regional or other differences. For example, an
emission factor applied throughout the United States for passenger cars on rural roads would necessarily
smear any differences in emissions between arid western states and those in the wetter, eastern part of the
country. Beside "east" and "west," one could also distinguish between: improved/unimproved and
well/poorly maintained road surfaces. No matter how many classifications are chosen, partitioning emission
test data into finely divided categories reduces the amount of data available to develop each factor. The
practical result from this fine subdivision is to lower the confidence in any result obtained from the analysis.
As an alternative to a single valued mean, an emission factor may be presented in the form of a
predictive equation derived by regression analysis of test data. The general method employed in regression
anlaysis is to first examine the physical forces that affect the dependent variable, to construct an empirical
model reflective of those forces, then to use regression to provide a best fit. Such an equation
mathematically relates emissions to parameters which characterize those measurable physical parameters
having the most affect on the emissions. Possible parameters considered may be grouped into three
categories:
1. Measures of source activity or energy expended (e.g., the speed, number of wheels, and weight
of vehicles traveling on an unpaved road). As a practical matter useful vehicle-related parameters should be
observable at a distance under normal traffic conditions. Most secondary parameters such as tire size,
pressure, etc., are correlated with gross vehicle characteristics such as vehicle weight as related to the type
of vehicle (light duty automobile, tractor trailer, etc.).
2. Properties of the material being disturbed (e.g., the content of suspendable fines in the surface
material on an unpaved road or the moisture content of the surface material).
3. Climatic parameters (e.g., number of precipitation-free days per year during which emissions
tend to be at a maximum).
An emission factor equation is useful if it is successful in "explaining" much of the observed
variance in emission factor values on the basis of corresponding variances in specific source parameters.
This enables more reliable estimates of source emissions on a site-specific basis. In general, an equation's
success in explaining variance is gauged by the R-squared value. If an equation has an R-squared value of
0.47, then it is said to "explain" 47 percent of the variance in the set of emission factors.
It should be noted, however, that a high value of R2 may sometimes prove misleading in developing
an emission factor equation for a particular data set. For example, an equation may be "fine tuned" to the
developmental data set by including an additional correction parameter, but in a manner that is contrary to
the physical phenomena of the dust generation process. This was illustrated in a field study conducted for
the Arizona Department of Environmental Quality (as described in Section 4) that found that inclusion of
moisture and silt content as correction parameters would require that they enter into the equation in a
3-7
-------
manner opposite to common sense. That is to say, emissions would increase with increasing moisture
content and would decrease with increasing silt content. In that instance, it is important to recognize that the
goal of an emission factor equation is not to provide a near-perfect fit to the emission measurements in the
developmental data base, but rather to provide reasonably reliable estimates of emissions for situations
where no test data are available.
A generic emission factor equation is one that is developed for a source operation defined on the
basis of a single dust generation mechanism that crosses industry lines. Clearly, vehicle travel over unpaved
roads is not only a common operation in almost all industries but also represents a general, public source of
particulate emissions.
Unpaved road source conditions encompass extreme variations. For example, average vehicle
weights on unpaved roads (ranging from country roads to mining haul roads) easily span two orders of
magnitude. Furthermore, there is also a wide range in surface material properties. Values for silt and
moisture content from the available test data span one and two orders of magnitude, respectively. Not
surprisingly, these correction parameters (like the emission factor values) are better characterized by a log-
normal rather than (arithmetic) normal distribution.
Furthermore, normal and log-normal distributions appear to fit other vehicle-related variables
(speed and number of wheels) equally well. Because standard tests of significance assume normal parent
populations, regression of log-transformed data is far more appropriate than regression of untransformed
values. The log-linear regression results in a multiplicative model.
To establish its applicability, a generic equation should be developed from test data obtained in
different industries. As will be discussed in Section 4, the approach taken to develop a new unpaved road
equation has been to combine (to the extent possible) all emission tests of vehicles traveling over an unpaved
surface. The combination is made without regard to previous groupings in AP-42. In particular, tests at
surface coal mines are combined with tests of unpaved roads within other industries and tests of publicly
accessible unpaved roads.
3 .3 EMISSION DATA AND EMISSION FACTOR QUALITY RATING SCHEME USED FOR THIS
SOURCE CATEGORY12 5
As part of the analysis of the emission data, the quantity and quality of the information contained in
the final set of reference documents were evaluated. The uncontrolled emission factor quality rating scheme
used for this source category represents a refinement of the rating system developed by EPA for AP-42
emission factors. The scheme entails the rating of test data quality followed by the rating of the emission
factor(s) developed from the test data, as described below.
In the past, test data that were developed from well documented, sound methodologies were viewed
equally and assigned an A rating. Although side-by-side studies would better define the differences in
precision between upwind/downwind and profiling methodologies, historical experience has granted a
greater degree of confidence in the ability of profiling to characterize the full particulate emissions plume.
In this document, test data using sound, well documented profiling methodologies were assigned an A rating.
Test data using sound, well documented upwind/downwind methodologies were assigned a B rating.
3-8
-------
In evaluating whether an upwind-downwind sampling strategy qualifies as a sound methodology, the
following minimum test requirements are used. At least five particulate measuring devices must be operated
during a test, with one device located upwind and the others located at two downwind and three crosswind
distances. The requirement of measurements at crosswind distances is waived for the case of line sources.
Also wind direction and speed must be monitored concurrently on-site.
For upwind/downwind testing, it is generally assumed wind speed and direction are constant. To
maintain a likeness of constant conditions, the downwind sampler should be shut down when the wind speed
drops below 75 percent or raises above 125 percent of the predetermined design speed for periods longer
than 3 minutes. Once the wind speed has returned to the acceptble range of 90 percent to 110 percent for
2 minutes, the downwind sampler should be restarted. Samplers should also be shut down when the wind
direction varies by 10° or more from the predetermined design direction for longer than 3 minutes. Once the
wind direction has returned to the acceptable range for two minutes, the samplers should be restarted.
General procedure includes shutting down the upwind sampler during the same periods the downwind
samples are shut down.5
The minimum requirements for a sound exposure profiling program are the following. A one-
dimensional, vertical grid of at least three samplers is sufficient for measurement of emissions from an
unpaved road. At least one upwind sampler must be operated to measure background concentration, and
wind speed must be measured on-site.
As an alternative to discrete downwind sampling units, a manifold system comprising several
sampling points may be used. The mass collected at different heights is ducted to a common tube where
stack sampling methods can be applied. A fundamental difference between the use of discrete samplers and a
manifold is the need in the latter case to demonstrate plume capture. In other words, the discrete sampling
approach directly demonstrates that concentration (or, more to the point, the mass flux) decreases near the
top of the sampling array. Because the manifold approach, on the other hand, integrates samples collected
at different heights, it cannot provide direct evidence of plume capture. Should the manifold approach be
adopted, a minimum of 4 sampling heights should be used for unpaved road testing. In addition, the test
report must address the issues related to capture of the entire plume. Furthermore, because wind speed
increases with height, the test report must also discuss isues of how intake velocities at different points were
selected and controlled to account for the variation in mass flux due simply to wind speed.
For a sound exposure profile operation, several test parameters must remain in predetermined
ranges including wind direction, wind speed, precipitation, and source conditions. Mean wind direction
during sampling should remain within 45° of perpendicular to the path of the moving point source for
90 percent of the 10 min averaging periods. The mean wind speed should not move outside of the 4 to 20
mph range more than 20 percent of the sampling period. Rainfall must not ensue during the equipment set-
up or during sampling for uncontrolled conditions. The predetermined criteria for source conditons (e.g.,
uncontrolled surface conditions, change from normally maintained road, unusual traffic, truck spill) should
be maintained.
Neither the upwind-downwind method nor the exposure profiling method can be expected to
produce A-rated emissions data when applied to large, poorly defined area sources, or under very light and
variable wind flow conditions. In these situations, data ratings based on degree of compliance with
minimum test system requirements were reduced one letter.
3-9
-------
It is critically important in either the upwind/downwind or exposure profiling method that the
unpaved road is uniformly emitting along the length of the road. In practical terms, this generally requires
that
* The road is straight or very gently curving over a distance that is much greater than the distance
to the downwind samplers.
* Vehicles do not typically start or stop moving in the general vicinity of the sampling array.
* In the case of heavy-duty vehicles, there is no need to downshift or otherwise cause substantial
diesel emissions near the test site.
It is also important to note that neither upwind-downwind nor exposure profiling interfere with
plume development or dispersion by forcing or blocking the air flow. Instead, the PM travels "naturally due
to vehicle wakes and ambient winds toward the sampling array
After the test data supporting a particular single-valued emission factor are evaluated, the criteria
presented in Table 3-1 are used to assign a quality rating to the resulting emission factor. The collection and
reporting of activity and process information such as road surface silt content, moisture content, and average
vehicle weight are also considered in the evaluation. These criteria were developed to provide objective
definition for (a) industry representativeness and (b) levels of variability within the data set for the source
category. The rating system obviously does not include estimates of statistical confidence, nor does it reflect
the expected accuracy of fugitive dust emission factors relative to conventional stack emission factors. It
does, however, serve as a useful tool for evaluation of the quality of a given set of emission factors relative
to the entire available fugitive dust emission factor data base.
TABLE 3-1. QUALITY RATING SCHEME FOR SINGLE-VALUED EMISSION
FACTORS
Code
No. of test
sites
No. of tests
per site
Total No. of
tests
Test data
variability3
Adjustment for
EF ratingb
1
>3
>3
-
3
>3
-
>F2
-1
3
2
>2
>5
2
>5
>F2
-2
5
-
-
>3
3
>F2
-3
7
1
2
2
>F2
-3
8
1
2
2
>F2
-4
9
1
1
1
-
-4
"Data spread in re
ation to central value. F2 denotes factor of two.
bDifference between emission factor rating and test data rating.
Minimum industry representativeness is defined in terms of number of test sites and number of tests
per site. These criteria were derived from two principles:
3-10
-------
1. Traditionally, three tests of a source represent the minimum requirement for reliable
quantification.
2. More than two plant sites are needed to provide minimum industry representativeness.
The level of variability within an emission factor data set is defined in terms of the spread of the
original emission factor data values about the mean or median single-valued factor for the source category.
The fairly rigorous criterion that all data points must lie within a factor of two of the central value was
adopted. It is recognized that this criterion is not insensitive to sample size in that for a sufficiently large
test series, at least one value may be expected to fall outside the factor-of-two limits. However, this is not
considered to be a problem because most of the current single-valued factors for fugitive dust sources are
based on relatively small sample sizes.
Development of quality ratings for emission factor equations also requires consideration of data
representativeness and variability, as in the case of single-value emission factors. However, the criteria used
to assign ratings (Table 3-2) are different, reflecting the more sophisticated model being used to represent
the test data. As a general principle, the quality rating for a given equation should lie between the test data
rating and the rating that would be assigned to a single-valued factor based on the test data. The following
criteria are used to determine whether an emission factor equation has the same rating as the supporting test
data:
1. At least three test sites and three tests per site, plus an additional three tests for each independent
parameter (P) in the equation.
2. Quantitative indication that a significant portion of the emission factor variation is attributable to
the independent parameter(s) in the equation.
TABLE 3-2. QUALITY RATING SCHEME FOR EM
ISSION FACTOR EQUATIONS
Code
No. of test sites
No. of tests per
site
Total No. of tests3
Adjustment for EF
ratingb
1
>3
>3
>(9 + 3P)
0
2
>2
>3
>3P
-1
3
>1
-
<3P
-1
aP denotes the number of correction parameters in the emission factor equation.
bDifference between emission factor rating and test data rating.
Loss of quality rating in the translation of these data to an emission factor equation occurs when
these criteria are not met. In practice, the first criterion is far more influential than the second in rating an
emission factor equation, because development of an equation implies that a substantial portion of the
emission factor variation is attributable to the independent parameter(s). As indicated in Table 3-2, the
rating is reduced by one level below the test data rating if the number of tests does not meet the first
criterion, but is at least three times greater than the number of independent parameters in the equation. The
rating is reduced two levels if this supplementary criterion is not met.
The rationale for the supplementary criterion follows from the fact that the likelihood of including
false relationships between the dependent variable (emissions) and the independent parameters in the
3-11
-------
equation increases as the ratio of the number of independent parameters to sample size increases. For
example, a four parameter equation based on five tests would exhibit perfect explanation (R2 = 1.0) of the
emission factor data, but the relationships expressed by such an equation cannot be expected to hold true in
independent applications.
REFERENCES FOR SECTION 3
1. Procedures for Preparing Emission Factor Documents, EPA-454/R-95-015, Office of Air Quality
Planning and Standards, U. S. Environmental Protection Agency, Research Triangle Park, NC,
May 1997.
2. Emission Factor Documentation for AP-42, Section 11.2.5 and 11.2.6, Paved Roads,
EPA-68-D0-0123, Assignment 44, Office of Air Quality Planning and Standards, U. S.
Environmental Protection Agency, Research Triangle Park, NC, March 1993.
3. Fugitive Dust Emissions Factor Update for AP-42, EPA 68-02-3177, Assignment 25, U. S.
Environmental Protection Agency, Research Triangle Park, NC, 1970.
4. Workbook of Atmospheric Dispersion Estimates, AP-26, U. S. Environmental Protection Agency,
Research Triangle Park, NC, 1970.
5. Protocol for the Measurement of Inhalable Particulate Fugitive Emissions from Stationary
Industrial Sources, EPA Contract 68-02-3115, Task 114, Process Measurements Branch, Industrial
Environmental Research Laboratory, Environmental Protection Agency, Research Triangle Park, NC,
March 1980.
3-12
-------
4. REVIEW OF SPECIFIC TEST REPORTS
4.1 INTRODUCTION
A total of 12 field test reports were identified as sources of either potentially directly useful data on
PM-10 emissions from unpaved roads or data that could be used to interpolate the necessary PM-10
information. These reports are described in Section 4.2.
4.2 REVIEW OF SPECIFIC DATA SETS
Profiling methodologies are generally used for these tests and include the following test parameters:
(a) downwind test equipment should be located approximately 5 meters from the source, (b) background
equipment should be placed approximately 15 meters upwind of the source, (c) wind direction should remain
within 45° of perpendicular to the path of the moving point source for 90 percent of the 10 min averaging
periods during testing, (d) mean wind speed should not move outside of the 4 to 20 mph range more than
20 percent of the sampling period, (e) and no wind flow disturbances should exist immediately upwind or
downwind of the testing location. When following standard testing methodologies some vehicle heights may
exceed the height of the sampling equipment typically about 7 m; however, the fact that the emissions
originate at the road surface and the emission plume density can be characterized as decreasing with height
indicates the total plume can be estimated. Vehicle heights are not generally reported in the source test
reports. Analysis for silt content and moisture content of the road surface follow methodologies described in
Appendix C.l and Appendix C.2 of the AP-42. Variations from these generally accepted test parameters or
any other nontraditional testing parameters are discussed within the individual test report reviews.
For this study, a well documented report not only discussed the test methodology but also included
source condition and activity information. With each report description both a summary of all reported
particulate sizes and individual PM-10 test data are presented. From these test reports, all uncontrolled tests
and all water tests were included in the emission equation development unless noted otherwise. Chemical
stabilizers were not included in the emission equation development discussed in Section 4-3.
4.2.1 Reference 1
Midwest Research Institute. "Letter Report of Field Tests. Road Sampling." for Washoe County
District Health Department. Reno. NV. August 1996.
This letter report presents results of sampling of an unpaved road and a paved road in Washoe
County, Nevada, in May and June of 1996. The study was undertaken to provide site-specific PM-10 test
data to supplement a yearlong road surface sampling program. Also, the study supported ongoing EPA
reviews of the PM-2.5 fraction of PM-10 emissions from paved and unpaved roads.
Exposure profiling was employed downwind to measure particulate emissions. For the unpaved
road tests, three hi-vol samplers each fitted with a cyclone preseparator were located downwind of the test
road at heights of 1, 3, and 5 m. Reference method PM-10 samplers were located upwind and downwind of
the roadway as well. Road widths were not reported. Wind speed was also recorded at heights of 1, 3, and
5 m.
4-1
-------
Four unpaved road tests and three paved road tests were completed. The unpaved road tests used
only lightweight captive vehicles at low vehicle speeds. Although the testing methodology was sound, the
conciseness of the letter report warranted a "B" rating of the test data. Table 4-1 presents summary test
data and Table 4-2 presents detailed test information.
4.2.2 Reference 2
Midwest Research Institute. "Improvement of Specific Emission Factors (BACM Project No. IV'
for South Coast AOMD. California. March 1996.
This study developed improved particulate emission factors for construction activities and paved
roads in western States. Sampling results for PM-10 are reported from testing in June and July, 1995, at
three construction sites located in Nevada and California. Also, surface silt loading measurements were
taken from paved roads in four separate areas in Nevada and California.
Exposure profiling was employed for the emission measurements. The downwind profiling arrays
contained three high volume air samplers fitted with cyclone preseparators at heights of 1, 3, and 5 m. One
high volume air sampler with a cyclone preseparator measured upwind concentrations at a 2 m height.
Warm wire anemometers, located at heights of 1 and 5 m, measured wind speed. Road widths were not
reported.
The unpaved road testing focused on particulate emissions from scraper travel and light-duty
vehicles. Six uncontrolled scraper tests and three uncontrolled light duty vehicle tests were completed. In
addition, watering was utilized as a control for two controlled scraper tests. The test data were assigned an
"A" rating. Table 4-3 presents summary test data and Table 4-4 presents detailed test information.
4.2.3 Reference 3
Air Control Techniques. "PM10. PM2.5. and PM1 Emission Factors for Haul Roads at Two Stone
Crushing Plants." for National Stone Association. Washington. D.C.. November 1995.
This test program presents the results of sampling at two stone crushing plant quarries in August
1995. This study was undertaken to accurately measure PM-10, PM-2.5, and PM-1 emissions from a
controlled haul road at a stone quarry. Testing occurred at Martin Marietta's Garner and Lemon Springs
quarries in North Carolina.
The study used what was termed "an upwind-downwind profiling technique." The test approach
relied on the use of a manifold to sample at several heights (up to 30 feet), which constitutes a profiling
method. Downwind samples were drawn (approximately isokinetically) into 10 sample nozzles 8 to
10 inches in diameter that joined a single downcomer connected to an 18 in. horizontal duct. The vertical
sampling occurred approximately 3 m downwind of the source. The system maintained a total gas flow rate
of approximately 2,500 acfin. Sampling occurred along the 18 in. horizontal duct using EPA Method 201A
for in-stack measurements of PM-10. Particle distribution measurements were collected with a cascade
impactor and a nephelometer. Upwind measurements were made using a hi-vol sampler at a height of 15 ft,
a cascade impactor, and a nephelometer placed only a few meters upwind. The roads were 30 ft wide at
4-2
-------
both test sites. Analysis included polarizing light microscopy (PLM) that measured particles of combustion
products. Wind direction was required to be ±60° of perpendicular to the line source.
Three emission tests were completed at both Garner and Lemon Springs. All samples were
considered controlled through water application during the test periods. Road watering occurred
approximately every 2.5 to 3 hours. The amount of water applied per unit road surface area is not stated.
Table 4-5 presents summary test data and Table 4-6 presents detailed test information. Emissions are
presented in Table 4-5 as reported in the study; however, the emissions calculation in the study did not
adjust for combustion product particles in the upwind measurements. For the development of the AP-42
emission equation, all particulate matter was factored into the emissions.
Although the sampling methodology varied from the more common exposure profiling methods, it
was judged satisfactory to capture and measure a representative mass emission from the road. As a result,
the Lemon Springs test was assigned an "A" rating. At the Garner test location, a large rock wall that stood
immediately behind the downwind sampling site may have interrupted natural wind flows and/or created a
local recirculation event. The potential wind obstruction accounted for a "B" rating of the test data at the
Garner quarry.
4.2.4 Reference 4
Midwest Research Institute. "Surface Coal Mine Emission Factor Study." for U. S. EPA.
January 1994.
This test report presents results of sampling during September and October 1992 at a surface coal
mine near Gillette, Wyoming. This study was undertaken to address issues identified in the Clean Air Act
Amendments of 1990 regarding the potential overestimation of the air quality impacts of western surface
coal mining. The principal objective was to compare PM-10 field measurements against available emission
factors for surface coal mines and revise the factors as necessary.
The study focused on characterizing particulate emissions from line sources such as haul roads and
scrapers at a surface mining site. Four haul road sites (No. 1, IB, 2, and 4) and one scraper site (No. 5)
were characterized using downwind exposure profilers for PM-10 fitted with cyclone preseparators, a
Wedding PM-10 sampler, and two hi-vol samplers for TSP. The exposure profiling arrays consisted of four
samplers located from 1 m to 7 m in height. Upwind concentrations were monitored with a Wedding PM-10
sampler and one cyclone preseparator. Wind direction at one height (3 m) and wind speed at three heights
(1 m, 3 m, and 5 m) were recorded at the downwind sites. Additional sampling studies included measuring
the near-source particle size distributions using a combination cyclone preseparator and a cascade impactor.
At the five sites a total of 36 PM-10 emission tests were completed. A majority of the tests
(34 PM-10 tests) were performed on haul roads. The road width was not reported. The haul road tests
spanned a large range of wind speeds from 4.5 mph to 22 mph. Approximately half of these tests were
controlled by use of water/surfactant. The water/surfactant provided a control efficiency from 40 to
70 percent for PM-10 and from 30 to 60 percent for TSP. A summary of emissions data is presented in
Table 4-7 and detailed test information is presented in Table 4-8. The test data were assigned a rating of A.
The report included adequte detail and the methodology meets the requirements for a sound exposure
profiling system.
4-3
-------
The study also presented an evaluation of the performance of emission factor models in predicting
independent emission test data. An emission factor developed specifically for haul roads in the surface coal
mining industry (see Equation 4-1) was compared against the "generic" AP-42 unpaved road emission factor
(Equation 2-1). The Fourth Edition of AP-42 (September 1988) presented the following PM-30 emission
factor for haul trucks in Section 8.24, "Western Surface Coal Mining:'"
E30 = 0.0067 (w)3.4 (L)0.2 (4-1)
where:
E30 = TSP emission factor (lb/vmt)
w = mean number of wheels
L = road surface silt loading (g/m2)
In addition, the performance of an emission factor developed specifically for light-/medium-duty
traffic at surface coal mines was also compared against that of the generic model. Section 8.24 in the
Fourth Edition of AP-42 (September 1988) presented the following equation (Equation 4-2) for estimating
PM-30 emission from light-/medium-duty traffic on unpaved roads at surface coal mines.
E30 = 5.79 / (M)4.0 (4-2)
where:
E30 = TSP emission factor (lb/vmt)
M = road surface moisture content (%)
It is important to note that, when Equation 2-1 was applied to independent emission test data, the
generic emission factor performed as well as or better than emission factors developed specifically for the
mining industry. For haul trucks, Equation 4-1 severely underpredicted the measured emission factors. On
average, Equation 4-1 underpredicted the independent test data by a factor greater than 5. In contrast,
Equation 2-1 tended to overpredict the independent test data, but by a factor of less than 2 on average.
Equation 2-1 also performed reasonably well (within 20 percent on average) when applied to
independent tests of light-duty traffic emissions. Although the AP-42 light/medium duty factor provided
reasonably accurate (within a factor of 2) estimates in two of three cases, the industry-specific factor
overpredicted a third independent test result by a factor of 20. In summary, then, the generic AP-42
emission factor performed at least as well as the industry-specific factors on average and performed
substantially better in terms of extreme over/underprediction. As will be discussed in Section 4.3, these
findings led to combining emission tests collected over a broad range of source conditions into a single large
data set for emission factor development.
4-4
-------
4.2.5 Reference 5
Entropy. "PM10 Emission Factors for a Haul Road at a Granite Stone Crushing Plant." for
National Stone Association. Washington. D.C.. December 1994.
This test report presents test data from measurements at a granite quarry in Knightdale, North
Carolina. The testing program occurred in October 1994 and focused on PM-10 emissions from an unpaved
haul road.
The testing protocols followed what the report termed a "push-pull method." Four 36-inch diameter
circulating fans were utilized on the upwind side of the road and large hoods were located downwind to
capture particulate emissions. Two sets of two hoods stacked vertically were located side-by-side. A set of
hoods consisted of two hoods each four ft high by seven ft wide with one located 2 ft and the other seven ft
above the ground. The road width was 40 ft. Emissions captured in a set of hoods were drawn through a
common 12 inch duct and sampled for PM-10 using EPA Method 201A. One hi-vol PM-10 ambient
sampler was located upwind of the circulating fans. Wind speed and wind direction were also monitored.
Three controlled tests and four uncontrolled tests were performed. All seven tests utilized both sets
of hoods and the results from both sets were averaged for the emission factor calculations. Testing was
discontinued when wind speeds exceeded 3 mph. Controlled tests utilized water as the dust suppressant.
For the controlled tests, watering occurred on average every 3.6 hr. The water application rate in terms of
volume of water applied per unit road surface area was not reported. Table 4-9 presents summary test data
and Table 4-10 presents detailed test information.
The push-pull method as described in Reference 5 does not correspond directly to any of the test
methods presented in Section 3 of this report. Furthermore, the data reported provide strong evidence that
some basic premises underlying unpaved road testing were not met. For example, in three of the seven
tests, the concentrations measured by the side-by-side hood differed by a factor of 5 to 7, strongly
suggesting either a lack of precision in the testing methodology or that the road under consideration could
not be reasonably represented as a uniformly emitting line source.
There are additional concerns about operational features of the push-pull method. Reference 5
describes wind directions up to 80 ° from perpendicular as acceptable and testing was interrupted if the
wind velocity exceeded 3 mph. Testing under low-speed winds or winds with very oblique directions
promotes the passage of PM-10 over the short sampling array. In other words, the wind speed/direction
acceptance criteria established for the push-pull method actually promote incomplete plume capture, thus
resulting in a low bias in the reported emission factors.
Because of the deviations from established acceptable sample methodology and the lack of precision
of the push-pull method, the quality highest rating the data could receive (following guidance given in EPA-
454/R-95-015, Procedures for Preparing Emission Factor Documents') is "C." Nevertheless, because the
operational parameters associated with the method would bias results low, a final quality rating of "D" was
assigned.
4-5
-------
4.2.6 Reference 6
Midwest Research Institute. "Unpaved Road Emission Impact."for Arizona Department of
Environmental Quality. March 1991.
This study performed field sampling on Arizona rural roads in Pima, Pinal, and Yuma counties.
The study also recommended a mathematical model to estimate emissions from unpaved rural roads for arid
and semiarid regions, based on a review of historical data as well as Arizona-specific field sampling results.
Particle emission sizes of interest in this study were TSP and PM-10. Contrary to expectation, the
examination of the historical data base did not find a systematic underprediction of emissions from unpaved
roads in the arid portions of the Western United States.
Exposure profiling formed the basis of the measurement technique used at the Arizona sampling
sites. For this study, two downwind arrays were deployed 5 m from the road. Each array had three
sampling heads located at heights of 1, 3, and 5 m. One downwind unit was fitted with cyclone
preseparators. The other downwind unit was equipped with cyclones for half the sampling periods and with
standard high volume roofs for the other sampling periods. In addition, one pair each of high volume and
dichotomous samplers were operated at a 100 ft downwind distance. No road widths were reported.
Upwind measurements were obtained with a vertical array containing two sampling heads, a standard hi-vol
sampler, and a dichotomous sampler. Wind speed was measured with warm wire anemometers at two
heights (1 and 5 m), and wind direction was measured at a single height.
Vehicle passes were controlled during testing periods and three vehicle speeds were tested (35, 45,
and 55 mph). The test data were assigned an "A" rating. Table 4-11 presents summary test data and
Table 4-12 presents detailed test information. The report examined how well the data developed in the field
tests agreed with the current version of the AP-42 emission factor.
Although the AP-42 equation provided reasonably accurate results when applied to the field tests
conducted in this study, another emission factor model was developed. This was justified in the report by
differences between typical traffic conditions in Arizona and the basis of the existing AP-42 emission factor.
Common travel speeds on rural unpaved roads in Arizona generally fall outside the range of values in the
AP-42 model's underlying data base. As a result of the numerous industrial road tests, the data base
generally reflected heavier vehicles than are common on rural roads.
4.2.7 Reference 7
Midwest Research Institute. "Roadway Emissions Field Tests at US Steels Fairless Works." for
U.S. Steel Corporation. May 1990.
This testing program focused on paved and unpaved road particulate emissions at an integrated iron
and steel plant near Philadelphia, Pennsylvania, in November 1989. Exposure profiling was used to
characterize one unpaved road (Site "X") located near the center of the facility and used principally as a
"shortcut" by light-duty vehicles.
Two tests were conducted using a profiling array, with sample heights from 1.5 m to 6.0 m, that
measures downwind mass flux. A high-volume, parallel-slot cascade impactor was employed to measure the
4-6
-------
downwind particle distribution and a hi-vol sampler was utilized to determine the downwind TSP mass
fraction. Road width was not reported. The upwind particle size distribution was determined with a
standard high-volume/impactor combination.
Unpaved roads at the plant had been treated with dust suppressant several years before the test
program started. As a result, only controlled unpaved road emissions were tested. In other words, this test
program did not produce data that could be used for an uncontrolled unpaved road emission equation. The
control efficiencies for PM-10 were estimated to be 80 to 90 percent. Control efficiencies for TSP were
estimated at 70 percent to 80 percent for the unpaved road chemical suppressants. Table 4-13 presents
summary information and Table 4-14 presents detailed test information.
4.2.8 Reference 8
Midwest Research Institute. "Evaluation of the Effectiveness of Chemical Dust Suppressants on
Unpaved Roads." for U. S. EPA. EPA-600/2-87-102. November 1987.
This study obtained data on the control effectiveness of common dust suppressants used in the iron
and steel industry. Tests were conducted from May through November, 1985, at LTV's Indiana Harbor
Works in East Chicago, Indiana, and at Armco's Kansas City Works in Missouri. The testing program
measured control performance for five chemical dust suppressants including two petroleum resin products
(Coherex® and Generic 2), a emulsified asphalt (Petro Tac), an acrylic cement (Soil Sement), and a calcium
chloride solution.
The exposure profiling methodology was utilized for all testing. The downwind exposure profiler
contained sampling heads at 1.5, 3.0, 4.5, and 6.0 m. Particle size distribution was determined both upwind
and downwind with high volume cascade impactors. Wind speed was monitored at two heights and wind
direction was monitored at a single height. Road width was not reported.
A total of 64 tests were completed with seven uncontrolled tests and 57 controlled tests.
Suppressants tested at Indiana Harbor Works were initially applied as follows: Petro Tac at 0.44 gal/yd2,
Coherex® at 0.56 gal/yd2, and calcium chloride at 0.25 gal/yd2. All five suppressants were tested at the
Kansas City Works facility and were initially applied at the following rates: Petro Tac at 0.21 gal/yd2,
Coherex® at 0.21 gal/yd2, Soil Sement at 0.16 gal/yd2, Generic at 0.14 gal/yd2, and calcium chloride at
0.24 gal/yd2. A rating of "A" was assigned to the data. Testing followed an acceptable methodology and
the test report was reasonably well documented.
Total particulate, IP, PM-10, and PM-2.5 were measured during this study. A control efficiency of
50 percent or greater was measured for all chemicals tested. Reapplication of the suppressant resulted in a
notably higher level of control. A cost-effectiveness comparison found little variation between suppressants
under the test conditions with the exception of a nonfavorable comparison of calcium chloride. Table 4-15
presents summary test data and Table 4-16 presents detailed test information.
The report also discussed the development of models to estimate the control efficiency of different
chemical dust suppressants. As was discussed at the end of Section 2, various suppressants do not appear
to affect the road surface characteristics in the same way. As a result, this makes performance models
based on surface physical parameters unfeasible.
4-7
-------
4.2.9 Reference 9
Midwest Research Institute. "Fugitive Emission Measurement of Coal Yard Traffic at a Power
Plant." for Confidential Client. December 1985.
This study included seven tests of controlled, unpaved surfaces and four tests of uncontrolled,
unpaved surfaces at a power plant. Airborne particle size fractions of interest in this study are total
particulate, TSP, IP, PM-10, and PM-2.5. A section of road within the facility's coal yard was tested in
August 1985. The road was a permanent ramp up the main stockpile and is used by scrapers for both
stockpiling and reclaiming operations.
Particulate emissions were characterized using three downwind exposure profilers, each consisting
of four profiling heads at heights of 1.5, 3.0, 4.5 and 6.0 m. (The use of three profiling systems allowed
continuous testing after water application by staggering the operation of the samplers.) Three high-volume,
parallel-slot cascade impactors equipped with cyclone preseparators were used to characterize the downwind
particle size distribution at a height of 2.2 m. One cyclone/impactor combination was used to characterize
the upwind particle size distribution and total particulate concentration. Wind speed was measured with
warm-wire anemometers at two heights (3 and 6 m) and wind direction was measured at a single height
(4.5 m). Also, incoming solar radiation was measured with a mechanical pyranograph. Road width was not
reported.
For the controlled tests, the road and surrounding areas were watered for approximately 30 minutes
before the start of air sampling. Water was applied to the surface in two passes with a total mean of
0.46 gal/yd2 (which is equivalent to approximately 0.08 in. of precipitation). The watering was found to
provide effective control for 3 to 4 hours with 35 vehicle passes/hr. The control efficiency for TSP and
PM-10 averaged 74 and 72 percent over 3 hours, respectively. The control efficiency closely correlated to
the surface moisture content, with a higher moisture content increasing the control efficiency. A summary
of the emissions data is presented in Table 4-17 and detailed test information is presented in Table 4-18.
Because testing followed an accepted test methodology and the results were reasonably well documented,
data were rated "A."
4.2.10 Reference 10
Midwest Research Institute. "Critical Review of Open Source Particulate Emission Measurements-
Part II - Field Comparison." for Southern Research Institute. August 1984.
This report presents test results from a June, 1984, test at U.S. Steel's Gary Works in Gary,
Indiana. The study was conducted to compare exposure profiling methodologies as used by five independent
testing organizations to characterize fugitive emissions originating from vehicular traffic. The source tested
was a paved road simulated as an unpaved road through the addition of exceptionally high road surface
loading (600,000 lb/mile).
An exposure profiler with 5 sampling heads (located at heights of 1.5, 3.0, 4.5, 6.0, and 7.5 m) was
used to characterize downwind emissions. Particle sizing was determined using cyclone/impactors located
alongside the exposure profiler. Particle sizes of interest in this study included total particulate (TP),
<30 /iin. <15 /iin. <10 //m , and <2.5//m in aerodynamic diameter. One cyclone/impactor and one cyclone
4-8
-------
were deployed upwind for background measurements. Warm wire anemometers measured wind speed at
two heights (1.5 and 4.5 m). The road was reported to be 30 ft wide.
The material used to cover the road surface was a mixture of clay, iron ore and boiler ash.
Reasonably good agreement was found between the AP-42 unpaved road model (Equation 2-1) and the
emission data collected for the simulated unpaved road. However, the report noted that this was a surprising
result for a number of reasons. First, the material (a mixture of clay, iron ore and boiler ash) used to
simulate the surface is not typical of unpaved roads. There were also concerns about the homogeneity of
the material spread over the five test sections. These problems were further complicated by the fact that the
source conditions were not at a steady-state. Instead, the surface loading (mass of material per unit area)
steadily decreased throughout the week of emission testing.
4.2.11 Reference 11
Midwest Research Institute "Size Specific Particulate Emission Factors for Uncontrolled Industrial
and Rural Roads" for U. S. EPA. January 1983.
This study reports the results of testing conducted in 1981 and 1982 at industrial unpaved and
paved roads and at rural unpaved roads. Unpaved industrial roads were tested at a stone crushing facility in
Kansas, a sand and gravel processing facility in Kansas, and a copper smelting facility in Arizona. The
rural unpaved road testing occurred in Colorado, Kansas, and Missouri. The study was conducted to
increase the existing data base for size-specific particulate emissions. The following particle sizes were of
specific interest for the study: IP, PM-10, and PM-2.5.
Exposure profiling was utilized to characterize particulate emissions. Five sampling heads, located
at heights of up to 5 m, were deployed on the downwind profiler. A standard hi-vol sampler and a hi-vol
sampler with a 15 /-/m size selective inlet (SSI) were also deployed downwind. In addition, two cyclone
impactors were operated to measure particle size distribution. A hi-vol sampler, a hi-vol sampler with an
SSI, and a cyclone impactor were utilized to characterize the upwind particulate concentrations. Wind
speed was monitored with warm wire anemometers. No road width was reported.
A total of 18 paved road tests and 21 unpaved road tests were completed. The test data were
assigned an "A" rating. Eleven industrial unpaved road tests were conducted as follows: five unpaved road
tests at the stone crushing plant, three unpaved road tests at the sand and gravel processing plant, and three
unpaved road tests at the copper smelting plant. For rural unpaved roads, six tests were conducted on roads
with a crushed limestone surface in Kansas, four tests were conducted on dirt roads in Missouri, and two
tests were conducted on gravel roads in Colorado. Rural road tests only measured emissions from light duty
vehicles at speeds from 25 to 40 mph. The industrial road tests were conducted with medium duty vehicles
at the stone crushing and copper smelting plants and heavy duty vehicles at the sand and gravel processing
facility. Table 4-21 presents summary test data and Table 4-22 presents detailed test information.
4-9
-------
4.2.12 Reference 12
Midwest Research Institute. "Iron and Steel Plant Open Source Fugitive Emission Control
Evaluation." for U. S. EPA. August 1983.
This test report centered on the measurement of the effectiveness of different control techniques for
particulate emissions from open dust sources in the iron and steel industry. The test program was performed
at two integrated iron and steel plants, one located in Houston, Texas, and the other in Middletown, Ohio.
Water and petroleum resin (Coherex®) were used to reduce emissions from traffic on unpaved roads.
Control techniques to reduce emissions from paved roads and coal storage piles were also evaluated.
Particle emission sizes of interest in this study were total particulate (TP), IP, and PM-2.5.
The exposure profiling method was used to measure unpaved road emissions at Armco's
Middletown Iron and Steel plant. For this study, one downwind profiler with four or five heads located at
heights of 1 to 5 m was deployed. Two high volume parallel slot cascade impactors samplers, one at lm
and the other at 3m, measured the downwind particle size distribution. A standard hi-vol sampler and an
additional hi-vol sampler fitted with a size selective inlet (SSI) were located downwind at a height 2 m. One
standard hi-vol sampler and two hi-vol samplers with SSIs were located upwind for background collections.
The road width was not reported.
Nineteen unpaved road tests for controlled and uncontrolled emissions were performed. Testing
included 10 runs of heavy-duty traffic (>30 tons) and 9 runs of light-duty traffic (<3 tons). Six heavy duty
traffic tests were controlled and four were uncontrolled, whereas, the light-duty traffic had five controlled
tests and four uncontrolled tests. The testing methodology was assigned an "A" rating, although a lack of
reported moisture data downgraded the report to a "B" rating. Uncontrolled and watered tests were used in
the exploratory development described in Section 4.3; however, due to the lack of reported moistures the
data were not included in the final emission factor equation. Table 4-23 presents summary test data and
Table 4-24 presents detailed test information.
For heavy-duty traffic, a 17 percent solution of Coherex® in water applied at a rate of 0.19 gal/yd2,
provided an average control efficiency of 95.7 percent for TP, 94.5 percent for IP, and 94.1 percent for
PM-2.5 over a 48 hr period. Water was applied at a rate of 0.13 gal/yd2 and, 'A hour after application, was
found to decrease emissions by 95 percent for all particles. Control efficiencies 4.4 hours after the water
applications were 55.0 percent for TP, 49.6 percent for IP, and 61.1 percent for PM-2.5.
A 17 percent solution of Coherex® in water was the only control applied during testing for the light-
duty traffic. The Coherex® solution was applied at a rate of 0.19 gal/yd2 and, 51 hr after application,
provided a control efficiency of 93.7 percent for TP, 91.4 percent for IP, and 93.7 percent for PM-2.5.
4.2.13 Reference 13
Midwest Research Institute. "Extended Evaluation of Unpaved Road Dust Suppressants in the Iron
and Steel Industry."for U. S. EPA. October 1983.
This study centered on the reduction of particulate emissions for various dust suppressants used on
unpaved roads in the iron and steel industry. Long-term control effectiveness of the dust suppressants was
4-10
-------
determined through testing at iron and steel plants located in East Chicago, Indiana and Kansas City,
Missouri. Water, an emulsified asphalt, and a petroleum resin were the dust suppressants used. Particle
emission sizes of interest in this study were TSP, IP, PM-10, and PM-2.5.
The exposure profiling method was used to measure unpaved road emissions at the Jones and
Laughlin's (J&L's) Indiana Harbor Works and Armco's Kansas City Works. For this study, one downwind
profiler, with four sampling heads at heights of 1.5 to 6 m, was deployed during all testing. High volume
cascade impactors located at heights of 1.5 and 4.5m measured particle sizes. A high volume cascade
impactor was also used to characterize the upwind particle distribution. Warm-wire anemometers at two
heights monitored wind speed and a wind vane monitored horizontal wind direction. Road width was not
reported.
Twenty-nine controlled and uncontrolled unpaved road tests were performed in this study. Three
uncontrolled tests and eight controlled tests were conducted at J&L's Indiana Harbor Works; and three
uncontrolled tests and 15 controlled tests were completed at Armco's Kansas City Works. All tests have
been assigned an "A" rating. Only uncontrolled tests and controlled tests using water were utilized in the
emission factor equation development. Table 4-25 presents summary test data and Table 4-26 presents
detailed test information.
The three controlled conditions in this study included a 20 percent solution of emulsified asphalt
(Petro Tac) applied at 0.7 gal/yd2, water applied at 0.43 gal/yd2, and a 20 percent solution of petroleum
resin (Coherex®) applied at 0.83 gal/yd2 followed by a repeat application of 12 percent solution 44 days
later.
The control effectiveness was reported as the number of vehicle passes that occurred as the control
efficiency decayed to zero. The initial asphalt emulsion application had an estimated lifetime of
91,000 vehicle passes for PM-10, the initial petroleum resin application had an estimated lifetime of
7,700 vehicle passes for PM-10, and the water application had an estimated lifetime of 560 vehicle passes
for PM-10. Also, a reapplication of the petroleum resin had an estimated lifetime of 23,000 vehicle passes
for PM-10.
4.2.14 Reference 14
Midwest Research Institute. "Improved Emission Factors for Fugitive Dust From Western Surface
Coal Mining Sources" for IJ S Environmental Protection Agency. Cincinnati. OH. July 1981.
This study was conducted to develop emission factors for major surface coal mining activities
occurring in the western United States. Results are reported of testing conducted in 1979 and 1980 at three
surface coal mines located in Wyoming, North Dakota, and New Mexico. Sampling was conducted on the
following mining operations: drilling, blasting, coal loading, bulldozing, dragline operations, haul trucks,
light- and medium-duty trucks, scrapers, graders, and wind erosion of exposed areas. Particulate sizes
measured include, TSP, IP, and PM-2.5.
Exposure profiling was used to measure emissions from line source activities such as vehicle traffic
on unpaved roads and from scraping and grading. Comparisons of data from profiling and upwind-
downwind methods were made for scrapers and haul roads. A modified exposure profiling methodology was
4-11
-------
utilized for blasting emission measurements, and a wind tunnel was used to measure wind erosion emissions.
Area source emissions such as coal loading were tested with an upwind/downwind methodology.
The exposure profiling method used a downwind profiler with four sampling heads located at
heights of 1.5 to 6.0 m. A standard hi-vol sampler (2.5 m), a hi-vol sampler fitted with a cascade impactor
(2.5 m), and two dichotomous samplers (1.5 and 4.5 m) were located downwind. Dust fall buckets were
placed upwind and downwind at a height of 0.75 m to measure the particle deposition. Upwind
concentrations were measured with one dichotomous sampler and one standard hi-vol sampler, both located
at a height of 2.5 m. Wind speed was measured with warm wire anemometers downwind at heights of
1.5 and 4.5 m. Road widths were not reported.
A total of 256 tests were performed in the study. Fifty-six of the tests were used in the development
of the AP-42 emission factor equation. The source activity distribution for unpaved road tests was as
follows: 20 uncontrolled haul road tests, 8 controlled haul road tests, 10 uncontrolled light- and medium-
duty vehicle tests, 2 uncontrolled light- and medium-duty vehicle tests, and 15 uncontrolled scraper tests.
Table 4-27 presents summary test data and Table 4-28 presents detailed test information.
4.2.15 Reference 15
Midwest Research Institute. "Fugitive Particulate Matter Emissions." for U.S. EPA. April. 1997.
This test report describes the results of field measurement and other data collection activities that
were undertaken in late 1995 and early 1996. The study focused on the determination of PM-10 and
PM-2.5 components of fugitive dust emissions from representative paved and unpaved roads at four
geographic locations in the United States (Kansas City, MO; Reno, NV; Raleigh, NC; and Denver, CO.)
Although, an emphasis was placed on the estimation of the PM-2.5 fraction of the emissions from unpaved
and paved roads, this study only reports PM-10 emission factors and PM-2.5/PM-10 ratios.
Exposure profiling was employed to measure particulate emissions. As is general practice with
profiling methods, the downwind sampling equipment was placed 5 m after the emission source and the
upwind sampling equipment was placed 10 m before the source. For the unpaved road PM-10 tests, a high-
volume air sampler equipped with a cyclone preseparator was utilized. A high-volume sampler equipped
with cyclone preseparators and parallel-slot, five-stage cascade impactors collected particle sizing
information. Also, dichotomous samplers were operated for particle sizing measurements. Wind speed was
monitored by wind odometers at three heights and wind direction was recorded with a wind instrument.
State-of-the-art equipment was employed for particle sizing at two of the unpaved road locations;
however, at the Raleigh, North Carolina location, an Amhurst Aerosizer Particle sizer failed because of a
power supply problem. At the Kansas City, Missouri location, MRI personnel operated a DustTrak Aerosol
Monitor light scattering instrument.
Thirteen uncontrolled unpaved road tests at three locations were completed as follows: five tests in
Kansas City, four tests in Reno, and four tests in Raleigh. Testing was completed using lightweight captive
vehicle traffic operated at a speed of 30 mph. This study recommends a PM-2.5/PM-10 particle size
adjustment factor of 0.15 for unpaved roads. The test data were assigned an "A" rating and were used as
4-12
-------
part of the PM-10 validation study discussed in Section 4.3.1 of this report. Table 4-29 presents summary
test data and Table 4-30 presents detailed test information.
4.2.16 References 16-19
Illinois State Water Survey—AWMA/APCA Publications. 1988-1989
Approximately 36 other unpaved road tests have been reported in a series of three APCA/AWMA
papers. These tests employed a exposure profiling method to characterize emissions from captive traffic on
several rural roads near Champaign, Illinois. A conversation19 with the project manager confirmed that
there is no test report that describes the methodology and results for the tests.
Twenty-one tests are reported in Reference 16, with the experimental methodology being described
in an earlier APCA paper (Reference 18). The main interest in Reference 16 is the set of emission factors
developed through exposure profiling. Sampling made use of three dichotomous samplers located at 1.55,
3.05, and 4.88 m. (Note that the sampling heights are different from those given in the paper [Reference 18]
describing the methodology.) The stacked samplers were located at a distance of 20 m from the road.
Reference 18 notes that wind speed and direction were continuously monitored, but no other details are
available. No dates are given for the tests.
Captive traffic was used to generate emissions from unpaved, limestone roads. Single tests at each
of three travel speeds (25, 35, and 45 mph) were conducted in each experiment. A total of 8 experiments
(denoted as 7 and 9 through 14) are reported in Reference 16. Although the only two road identification
codes are reported, it is not clear whether the tests were conducted at the same location and thus constitute
replicate samples.
In each of the 21 cases analyzed, the emission factors were calculated by assuming a linear profile
for exposure values. Thus, the maximum exposure 20 m downwind from the road distance is assumed to
occur at ground level even though the wind speed (and thus exposure) vanishes at ground level. This leads
to a systematic high bias in the emission factors reported.
Surface samples were collected "periodically" from the roads. All tests reported in a single
experiment are associated with a single silt value. When samples were not available for the day that
emission testing occurred, values are interpolated. Sample collection and analysis methods were not
specified.
An additional fifteen tests were conducted in 1988 and are reported in Reference 17. In those tests,
a fourth dichotomous sampler was included in the sampling array 20 m from the roadway. Sampling
spanned 1.5 to 6.1 m, but individual sampling heights are not reported. Wind speed was monitored on-site
at a 1.5 m height. Those measurements were combined with 10-m wind data from an off-site meteorological
station to develop a logarithmic profile for calculation purposes.
A total of 4 experiments (15 through 18) are reported in Reference 17. With the exception of
experiment 15, all consisted of an individual test at each of 4 captive vehicle speeds: 25, 35, 45 and 55
mph. Experiment 15 examined emissions at speeds of 25, 45 and 55 mph.
4-13
-------
The 1988 tests were associated with a great deal of surface sampling. Three different samples were
collected before and after every 100 vehicle passes. As opposed to Reference 16, separate silt values are
reported with each test in an experiment.
Two sets of surface samples were considered. The first set was generally collected in the same manner as
described in AP-42, Appendix C. 1. Contrary to AP-42 Appendix C.2, however, these samples were not
oven-dried prior to sieving. A second set of samples focused on the tracks and ruts formed by the captive
traffic. The paper does not compare the results from the two different sets of samples.
Two roads were tested - one with limestone and the other with glacial gravel. Experiments 16 and
17 were conducted on the limestone road and on consecutive days; these constitute replicate measurements.
Experiment 14 was conducted on the limestone road, but it is not know whether at the same location as
experiments 16 and 17. Experiment 18 was conducted at the glacial road.
Although specific data reduction methods are not described, it is assumed that a linear profile was
used to characterize exposure values. As noted earlier, this would lead to maximum exposure at ground level
and to a systematic high bias in the emission factors reported.
Because supporting documentation could not be obtained, the data were not available for the
development of an emission factor equation.
4.3 DEVELOPMENT OF CANDIDATE EMISSION FACTOR EQUATION
For unpaved roads, an emission factor equation has been found to be successful in predicting
particulate emissions at different sites with varying source parameters. This section describes the
development of the emission factor equation that will be proposed for the updated AP-42 Unpaved Road
section.
Various road surface and vehicle characteristics are likely to have an impact on the particulate
emissions from unpaved roads. Those parameters most likely to influence the particle emissions, while at
the same time are able to be measured in a practical manner, are considered for the emission equation
development. The possible parameters may be grouped into three categories: (a) measure of source activity
(b) properties of the material being disturbed and (c) climatic parameters.
The measure of source activity includes the speed and weight of the vehicles traveling on the
unpaved road. This category would also include the number of wheels of the vehicles in contact with the
unpaved road. Subparameters that affect the particle emissions might also be considered; however, cost
conscience efforts and clarity considerations for potential emission equation users have narrowed in-depth
reviews of these subparameters. These subparameters may include the following: the turbulence created by
the aerodynamics and clearance of the individual vehicle traveling on the unpaved road; the unique
characterisics of the tire such as width, pressure, and tread design; angle of wheels compared to vehicle
thrust; and wheel slippage over the unpaved road surface. Also, if extensive detailed traffic data were
available for 15,000+ vehicle passes in the current data set, it would be possible to consider the relation of
emissions of tangential wheel velocity compared to vehicle speed.
4-14
-------
The properties of the material being disturbed includes moisture content and the content of the
suspendable fines in the surface material. Although difficult to characterize within the magnitude of the
available data, emissions could potentially be affected by interactions between dust particles of different
physical characteristics. Conditions of the unpaved road may also be considered such as the characteristics
of the road base (e.g., compacted, hardbase, washboard). Difficult to characterize variability in road
conditions and resultant complexity of the emission equation were considered as basis for not including the
road base characteristics in the emission factor equation.
Climatic characterization is generally reflected by the precipitation-free days per year on which
emissions tend to be at a maximum. The radiant energy of the sun may be important when determining the
control efficiency of watering, and in effect the average moisture content of the surface material. Direct
moisture measurements are appropriate in this case.
The parameters readily measureable and applicable to a general unpaved road equation include
surface silt content, surface moisture content, mean vehicle weight, mean vehicle speed, and mean number of
wheels. Discussion of the analysis of these parameters continues later in this section.
The development of a revised unpaved road emission factor equation was built upon findings from
the reviewed data sets. First, the decision was made to include all tests of vehicles traveling over unpaved
surfaces. For example, tests of scrapers in the "travel mode" between cut and fill areas were included.
Also, tests of very large off-road haul trucks used in the mining industry were also included in the
developmental data set. On the other hand, graders blading an unpaved road were not included because of
the low speed and the additional road surface disturbance involved. This decision had the effect of greatly
expanding the historical data base. Not only are far more data available, but the data encompass a wider
range of vehicle weights and travel speeds.
The decision to composite the data sets was based on findings from Reference 4, which dealt with
the western surface coal mining industry. Remarks made in Section 4.2.4 bear mention here as well.
Reference 4 found that the "generic" unpaved road emission factor model currently contained in AP-42
(Equation 2-1 in this document) performed at least as well in predicting emissions from both haul trucks
and light-duty vehicles as did emission factors developed specifically for the industry under consideration.
Next, the decision was made to add tests of watered roads to tests of uncontrolled roads, because
moisture content is also affected by natural mitigation resulting from climatic factors. Chemically
controlled unpaved roads were not included because those treatments cause lasting physical changes to the
road surface. A review of the measurable physical characteristics (silt content and moisture content) of
chemically controlled unpaved roads found no identifiable trends. Reference 8 examined the historical data
base and concluded that a general control estimation method based on surface characteristics was not
feasible.
The inclusion of both uncontrolled and watered roads was based on findings in the Reference 4
study. That study and a later review included moisture as a potential correction parameter in developing a
predictive equation for unpaved roads. It was found that both the old (Reference 14, circa 1980) and new
(Reference 4, 1992) haul truck data could be successfully fitted with one equation that applied to both
watered and uncontrolled surfaces. The decision was also supported by a similar approach taken in
4-15
-------
developing the current AP-42 paved road equation. In that case, controlled and uncontrolled tests were
combined.
Inclusion of watered surfaces in the data base recognizes a fundamental difference in how the
addition of water controls emissions (as opposed to the addition of other types of suppressants). First, the
addition of water is a short-term control measure and is similar to the effect of rain. In addition, it causes no
permanent change in the road surface characteristics. To an extent, one could argue that a road subject to
frequent rain is no different than a road which is routinely watered.
Finally, the decision was made to focus on PM-10 emission tests. Because Equation 2-1 was
developed earlier than the 1987 promulgation of the PM-10 NAAQSs, that factor did not focus on the
particle size range of current regulatory interest. Combining data sets emphasizes the basic physical process
of dust generation by vehicle traffic on unpaved roads. In keeping with that view, it is reasonable to expect
that emission factors for different size fraction resemble one another. The approach requires that the models
developed for different particle size ranges be "consistent," in the sense discussed below.
As a first step, the "developmental" data base was prepared from the test reports discussed in the
previous section, with the following exceptions:
1. No test data were included from Reference 5. As noted earlier, these data were rated "D."
2. No data were included from Reference 7, because the unpaved road considered had been
previously treated with a chemical dust suppressant. Also, individual tests of chemical dust suppressants in
other references were not included.
Finally, some additional preparation of the data base was required. For example, References 12 and
14 did not present PM-10 emission factors; values were developed by log-normal interpolation of the
PM-15 and PM-2.5 ratios to total particulate emissions. In addition, References 1, 12, and 13 did not report
individual surface moisture contents. However, because silt content is determined after oven drying, the
necessary information was readily available for Reference 1, which was being prepared at the same time that
the current work was being undertaken. In Reference 13, some individual tests had moisture contents
reported and a few additional tests were associated with moisture contents as well. Those tests for which
moisture data were reported were included in the development data set. Furthermore, the data from
Reference 3 had been corrected for "combustion particulate" content (although upwind concentrations had
not). Using information contained in the report, "total" PM-10 emission factors (i.e., without regard to
chemical composition) were calculated for inclusion in the developmental data set. (An ASCII data file
containing the developmental data set is provided in the file D13502B.ZIP located on EPA's CHIEF
web site under Draft AP-42 Sections.
Model development relied on the stepwise linear regression routine contained in the SYSTAT,
Version 4 set of statistical routines. The default level of significance used by SYSTAT for a variable to
"enter" the stepwise linear regression was 0.15 (15 percent). In this context, "level of significance" refers to
the probability of making a so-called Type I error. The possibility of making this kind of error arises
because we are dealing with samples drawn from a parent population. That is to say, under the default
setting, samples drawn from two completely independent populations would be found to have a significant
4-16
-------
relation purely due to chance 15 times out of 100. The 15 percent level of significance was used for
exploratory data analysis; refined analysis relied on specifying a 5 or 10 percent significance level.
Standard statistical tests of significance assume normal parent populations. Because unpaved road
emission factors and key correction parameters are log-normally distributed, the regression analysis needs to
rely on log-transformed data. This results in a multiplicative model, which is the form of the current AP-42
emission factor predictive equation.
Stepwise multiple linear regression was used to develop a predictive emission factor equation from
the data set. Five potential correction parameters were included:
1. Surface silt content, s;
2. Surface moisture content, M;
3. Mean vehicle weight, W;
4. Mean vehicle speed, S; and
5. Mean number of wheels, w.
In addition to the emission factor and correction parameter values, the data base also contained
codes indicating:
1. Whether the test was of an uncontrolled or a watered surface;
2. The type of road;
a. publicly accessible unpaved road
b. unpaved travel surface at an industrial facility
c. "simulated" unpaved road
3. The predominant type of vehicle traveling the road;
a. Light or medium-duty vehicles;
b. Haul trucks;
c. Scrapers in the travel mode; and
d. Heavy-duty, over-the-road trucks.
For the initial analyses, the data base was sorted by whether the test represented uncontrolled or
watered conditions and by the type of road (industrial vs. public unpaved road). There were two main
objectives in this step. The first objective was to determine simply whether the different portions of the data
base could be successfully combined. The second objective was to determine whether an emission factor
model resulting from the large combined data would be consistent. The term "consistent" refers to
(a) whether or not the same basic set of correction parameters could be used to estimate emission levels and
(b) whether or not the relationships were similar between different subsets in the data base.
For example, suppose that stepwise regression of one portion (I) of the data base (e.g.,
uncontrolled industrial roads) showed that emissions were highly dependent on variable X but independent
of variable Y. If stepwise regression of another portion (II) of the data base, on the other hand, indicated
that emissions were very dependent upon Y but not on X, then the results for the two portions would not be
viewed as consistent. The consistency in the relationships between independent and dependent variables is
also important. To continue the example, suppose that regression of portions I and II both showed that the
emission levels depend on variable X. If, however, for portion I, emissions depended on the 0.5 power of X
4-17
-------
while in portion II, emissions varied with the second power of X, then the relationships would again be
viewed as "inconsistent."
Given that the individual sets within the data base do not necessarily contain many test results,
evaluation of consistency cannot always follow hard and fast rules. For example, one would reasonably
expect that the emissions from watered tests would depend on the surface material moisture content. The
lack of a discernible relationship between moisture and emissions from the uncontrolled tests in the data
base would not necessarily indicate inconsistency. Furthermore, determining how "close" two relationships
are, requires considerable judgment as well. For example, both a power of 0.86 and power of 1.1 indicate a
roughly linear relationship.
The analysis began by stepwise regression of only the 160 uncontrolled tests in the data base, using
the potential correction parameters of silt, weight, speed and number of wheels. Note that moisture content
was not included. In this case, mean vehicle weight entered the regression first, and surface silt content on
the second step. This first regression was roughly equivalent to repeating how the current AP-42 unpaved
road emission factor was derived. Unlike the past, however, the effort focused on PM-10. The resulting
emission factor for PM-10 exhibited an almost linear (power of 1) relationship with silt content.
Furthermore, emissions were shown to follow a "less-than-linear" relationship with vehicle weight, although
the exponent was roughly half of that contained in the current AP-42 equation (Equation 2-1).
Next, uncontrolled and watered tests were considered separately, but this time with moisture content
included as a potential correction parameter. For the 137 uncontrolled tests, weight and silt were again the
first two variables to enter the regression. The exponents for both these variables were consistent with the
values obtained for only the uncontrolled tests. However, two additional variables entered the stepwise
regression in this case. Surface moisture content entered on the third step and mean vehicle speed on the
fourth.
Inclusion of speed was somewhat tentative, in that its level of significance was just slightly greater
than 10 percent. The default significance level for a variable to enter the regression was 15 percent. If the
requirement for a variable to enter had been tightened to the 10 percent level of significance, speed would
not have entered the relationship.
For the 43 watered tests, only two correction parameters entered the regression—silt and weight.
The powers for silt and weight were reasonably consistent with the results obtained when the uncontrolled
tests were considered separately. The reasonably consistent relationships for both silt and weight suggested
that the two uncontrolled and watered portions of the data base could be successfully combined.13
When both uncontrolled and watered tests were considered as one data set, weight and silt again
entered first and second, with moisture entering on the third step. Wheels would enter the equation if the
level of significance were relaxed to 20 percent; however, for this analysis at the 10 percent level of
significance wheels are not included. Speed entered on the fourth iteration. The resulting emission factor
equation has the form
b The relationships for both of these variables are also reasonably consistent with the relationships in the
current AP-42 model (Equation 2-1).
4-18
-------
E = k s085 Wa50 S0-32 / M0-29
(4-3)
where k is a constant of proportionality.0 The R2-value (0.354) for Equation 4-3 indicates that the model
explains approximately 35 percent the variation in emission factors.
An alternative to Equation 4-3 results from tightening the significance requirement, from 10 percent to
5 percent, for a variable to enter the regression. In this case, speed does not enter the equation, and the
equation has the form:
E = k s082 W0-46 / M0-28 (4-4)
This equation has a R2-value of 0.345, which is only slightly less than Equation 4-3.
Equations 4-3 and 4-4 represent the two candidate PM-10 emission factor equations considered in this
study. Initially, preference was given to Equation 4-3 because the inclusion of speed was viewed as
providing additional predictive accuracy for instances involving very slow or very fast traffic. Equation 4-3
was initially chosen and validation of that model proceeded.
However, in the validation of Equation 4-3, it was found that almost no additional predictive
accuracy was achieved and that the equation did not permit actual estimates of the effects of speed
reduction. The inclusion of speed was highly dependent on the data set being used. For example, exclusion
of only one or two low-speed tests from the data resulted in speed not entering the regression at even the
15 percent level of significance. On the other hand, dropping those tests had no effect on the other terms in
the model. Thus, the four-parameter model (Equation 4-3) appeared to be relatively unstable.
Furthermore, past testing studies have found that, when all other road/traffic parameters are held
constant, emissions depend on a higher power of mean vehicle speed than the 0.32 value given in
Equation 4-3. In Reference 6 and other older studies designed to assess the influence of vehicle speed on
PM emissions, powers between 1 and 2 have been found. Note, however, that those studies were able to
separately consider different speeds by supplying "captive" traffic during testing. In other words, the testing
organization supplied essentially all the vehicular traffic during the field exercise to tightly control source
conditions. This is a "parametric approach" that is the only systematic way to isolate the effect of
individual source parameter on emission levels. In practical terms, such an approach is restricted to roads
that (a) have relatively little "natural" traffic and (b) are traveled by mostly light-duty vehicles.
The captive traffic approach to systematically examine the effect of vehicle speed is in pointed
contrast to how most tests in the data base were conducted. Most tests were conducted on roads at which
c Working versions of the emission factor equation are presented. In this context, the term "working"
refers to factors that require that weight be expressed in tons, speed in mph, and silt and moisture
contents in percent. Furthermore, the emission factor must be expressed in lb/VMT. In this case, the
constant of proportionality has a complicated set of dimensions. The model recommended later in
Equation 4-5 has been "normalized" by dividing, for example, weight by a default vehicle weight of 3
tons. In that case, the constant of proportionality has the same dimensions as the emission factor itself
and can be readily converted from one set of units to another.
4-19
-------
the traffic could not be tightly controlled by the testing organization. Because data from many studies have
been assembled and because most tests do not rely on "captive" traffic, it is not possible to isolate the effect
of speed on emissions. Without the benefit of captive traffic, it is not surprising that weight and speed are
highly intercorrelated in the data set. Furthermore, speed and emissions are not significantly correlated in
the developmental data set. In fact, there is a negative (although not significant) correlation between
emission factor and speed.
It is crucially important to keep in mind that predictive accuracy is the goal of any emission factor
equation. With this in mind, the predicted-to-actual ratios for Equation 4-3 were compared to those for
Equation 4-4. The summary statistics follow:
Equation 4-3
(with speed term)
Equation 4-4
(no speed term)
Minimum
0.104
0.100
Maximum
30.1
27.4
Geometric Mean
1.02
0.986
Geometric Std. Dev.
2.74
2.71
(Note that geometric rather than arithmetic statistics are used here. The reason for this choice is explained
in Section 4.5.1). In comparing the two sets of statistics, it is clear that the inclusion of a speed term in
Equation 4-3 lends almost no additional accuracy.
In summary, the following emission factor equation is recommended for estimating PM-10
emissions from vehicles traveling over unpaved surfaces:
E10 = 2.6 (s/12)0-8 (W/3)a4/(M/0.2)a3 (4-5)
where:
E10 = PM-10 emission factor (lb/VMT)
s = surface material silt content (%)
W = mean vehicle weight (tons)
M = surface material moisture content (%)
Note that the "normalizing factors" of 12 percent silt and 3 tons are the same as for the current
AP-42 model. This allows one to compare the leading term of 2.6 lb/VMT in Equation 4-5 to the factor of
2.1 lb/VMT inherent in the current version of the unpaved road predictive model.d (The selection of
0.2 percent to normalize the moisture term follows from the specification of a default value. See
Section 4.4).
d That is, the leading value of 5.9 (in Equation 2-1) times the aerodynamic particle size multiplier of 0.36
for PM-10.
4-20
-------
To the extent practical, the development of emission factor equations for other the PM size ranges
followed that for PM-10. That is to say, the preferred approach was to develop a stepwise regression of the
available test data. For PM-30 (used as a surrogate for TSP), stepwise regression of the 65 uncontrolled
emission test data led to the following result:
E30 = k s0-97 W0-52 / M0-45 (4-6)
where all variables are the same as before and E30 denotes the PM-30 emission factor in lb/vmt. The
R2-value for the above factor is 0.49 and the equation compares well with the intermediate and final results
for PM-10. In contrast to PM-10, however, vehicle speed did not enter the stepwise regression for PM-30.
When both uncontrolled and watered PM-30 tests were considered, the same three variables—silt
and moisture contents, and mean vehicle weight—again entered the stepwise regression of the 92 test date.
With the inclusion of the tests of emissions from watered surfaces, the only noticeable change in exponents
was a slight reduction in the power for silt content. Because of the consistency between the watered/
uncontrolled tests and between the PM-10/PM-30 results, the following emission factor equation is
recommended for PM-30:
E30 = 10 (s/12)0-8 (W/3)0-5 / (M/0.2)04 (4-7)
The PM-30 emission factor is clearly consistent with the factor for PM-10 (Equation 4-5). Both
factors involve the same three independent variables, each raised to essentially the same power. In contrast
to PM-10, vehicle speed did not enter any of the stepwise regressions of PM-30 test data.
Model building efforts for PM-2.5 initially followed the same procedures as for PM-10 and PM-30.
That is, stepwise linear regression of 77 uncontrolled PM-2.5 emission test data led resulted in three
variables entering the equation
E30 = k s0-67 W0-21 / M017 (4-8)
where all variables are the same as before and E2 5 denotes the PM-2.5 emission factor in lb/vmt
Note that, again, the same three variables entered the stepwise regression: silt content, mean vehicle weight
and moisture content. Although the power to which the silt term is raised is reasonably comparable to the
exponents in the PM-10 and PM-30 factors, the two remaining exponents are only half those in the other
emission factor equations. More troubling is the fact that a low R2 value for the equation implies that only
8 percent of the variation in emission levels is explained by the equation. Furthermore, when the watered
tests are added to PM-2.5 developmental data set, two more variables—mean vehicle speed and number of
wheels—now enter the stepwise regression. The R2 for the equation is again low at a value of 0.23. In other
words, even with five variables, the regression-based PM-2.5 factor appears to be disappointingly poor in
terms of predictive ability.
Because of the failure of stepwise regression to produce a suitable PM-2.5 emission factor equation,
the significant difference from the PM-30 and PM-10 equations, the potential for the five variable PM-2.5
equation to result in a value exceeding the PM-10 equation under some circumstances, and the low R2 for
the three variable equation that is reasonably comparable to the PM-10 and PM-30 equation, an alternative
4-21
-------
approach was taken. In this case, a PM-2.5 factor was developed by scaling the PM-10 model
(Equation 4-5) by the measured PM-2.5/PM-10 in the available data base:
Geometric mean ratio
of PM-2.5 / PM-10
Uncontrolled (n = 108)
0.140
Watered (n=20)
0.196
Overall (n=128)
0.148
No significant difference was found between the ratios for watered versus uncontrolled conditions,
so the overall mean was applied. Furthermore, no significant correlation (at the 5 percent level) was found
between PM-2.5/PM-10 ratio and emission factor, silt, moisture, weight, speed, or number of wheels.
In summary, for the three PM size fractions of greatest interest, the following emission factor
equation is recommended for inclusion in AP-42:
E = k (s/12)a (W/3)b/(M/0.2)c (4-9)
where: k, a, b and c are empirical constants given below and
E = size-specific emission factor (lb/vmt)
s = surface material silt content (%)
W = mean vehicle weight (tons)
M = surface material moisture content (%)
The parameters for size-specific emission factors in Equation 4-9 are given below:
Empirical constant
PM-2.5
PM-10
PM-30
k
0.38
2.6
10
a
0.8
0.8
0.8
b
0.4
0.4
0.5
c
0.3
0.3
0.4
Based on the rating system given in Section 3.5, both the PM-10 and PM-30 emission factors would
be rated "A" by strictly following the decision rules presented there. However, because the predictive
equation was developed to span a very broad range of source conditions and has an R2 of only 0.34, a
lowering of the quality rating is appropriate. The PM-10 and PM-30 emission factors are rated "B."
Because the factor is based on scaling the PM-10 factor, the PM-2.5 factor is downgraded 1 letter. Thus
the PM-2.5 factor carries a quality rating of "C."
4-22
-------
It is important to note that the overall performance of any emission factor improves when it is
applied to a number of sources within a specific area. This is an important distinction between fugitive dust
sources and the "stack" ("point") emission sources (such as utility boilers) commonly discussed by AP-42.
That is to say, an area being inventoried typically contains no more than a handful of the stack-type sources
which use a specific emission factor. Furthermore, stack sources are far better defined and steady in terms
of operating conditions (feed rate, air flow, etc.). In contrast to a handful of stack sources, an inventoried
area may contain dozens of unpaved travel surfaces, each with very different vehicle characteristics that
change with hour of the day, seasonally, etc. In that case, the performance of an emission factor in
accurately predicting emissions from a single, isolated source should not form a central focus. Instead, one
should be most concerned about how well the factor performs in estimating the total (or average) emission
from the entire set of sources over time periods of interest.
4.3.1 Validation Studies
A series of validation studies were undertaken to examine the predictive accuracy of the various
emission factors recommended in the preceding section. Validation focused on the PM-10 model.
This section discusses the performance of the model primarily in terms of the predicted-to-measured
ratio:
As a practical matter, because of the log-linear regression used to develop the emission factor models, the
log of the predicted-to-measured ratio is identical to the "residual" or error term:
residual = log(predicted) - log(measured) = log(predicted-to-measured)
Throughout this section, summary statistics are presented in terms of geometric mean and standard
deviation. This follows directly from the use of log-linear regression. Furthermore, use of the geometric
mean is clearly more appropriate to describe ratios than the arithmetic mean for the following reason.
Unlike the arithmetic average, the geometric clearly represents the tendency of the ratio. To illustrate this
point, consider the following 10 hypothetical ratios:
emission factor predicted by model
measured emission factor
Case
1
2
3
4
5
6
7
Predicted-to-measured
0.678
1.48
2.76
0.885
0.754
0.248
1.87
0.126
1.76
3.15
Measured-to-Predicted
1.47
0.68
0.36
1.13
1.33
4.03
0.53
7.94
0.57
0.32
8
9
10
Arithmetic mean
Geometric mean
1.37
0.95
1.84
1.05
4-23
-------
By using the arithmetic mean of the predicted-to-measured ratio of 1.37, one could argue that the
predictions were about 37 percent higher than the measured. This leads to a natural suspicion that the
measured values were roughly 37 percent lower than the predictions. However, it is seen that the arithmetic
mean of the measured-to-predicted ratio is in fact 1.84 which is greater than 1.37. On the other hand, the
geometric mean has the property that it is equal to the inverse of the mean for the inverse ratio.
In addition, because of the log-linear regression, the residuals are log-normally distributed. For this
reason, logarithmic plots of the residuals are presented.
The first two PM-10 validations used the data base assembled for developing the model. The first
made use of a cross-validation analysis of the PM-10 data set. In this approach, each data point is
eliminated one at a time. The regression obtained from the "reduced" data base is used to estimate the
missing data value. In this way, a set of "n" quasi-independent observations is obtained from the data set of
"n" tests.
The PM-10 cross-validation (CV) shows that the model is fairly accurate for a very broad range of
source conditions. Table 4-31 indicates that, although the model may slightly under- or overpredict
individual emission factors in some specific subset of the data base, the general agreement is quite good.
The CV analysis further found that, for the quasi-independent estimates of the measured emission factors:
1. 52 percent are within a factor of 2;
2. 73 percent are within a factor of 3;
3. 90 percent are within a factor of 5; and
4. 98 percent are within a factor of 10.
Again, recall that, because a facility typically contains numerous roadway segments, each with its
own vehicle mix, one is most concerned about how well the factor performs in estimating the total (or
average) emission. Thus, even though the above-cited statistics suggest that, for example, there is
approximately a 30 percent probability of over- or underestimating emissions by a factor of 3 for an
individual roadway segment, there is a substantially lower chance of making the same level of error for
emissions from the totality of roadways under consideration at a facility. Computation of an exact
probability would depend on: (a) the number of individual segments under consideration and (b) the relative
contribution of each segment to the total PM emissions. Note that item (b) is a relatively complicated
function of the emission factor, the vehicle traffic and the road segment length .
To illustrate the increased confidence, a series of simple random drawings of 5 tests from the
developmental data set was made. Comparing the sum of the measured and the estimated emissions is
analogous to a hypothetical situation in which plant contains 5 road segments, each with the same length and
same number of vehicle passes. In 1000 repetitions of the random draw of 5 from the developmental data
set, the following was found for the sum:
1. 73 percent were within a factor of 2;
2. 92 percent were within a factor of 3; and
3. 99.6 percent were within a factor 5.
In this illustration, one would have only and 8 percent chance of over- or underestimating total emissions by
a factor of 3.
4-24
-------
Plots of the residuals versus individual PM-10 emission factor, silt, moisture, weight, speed and
wheels are presented in Figures 4-1 through 4-6, respectively. In examining the PM-10 residuals (i.e., the
error between individual predicted and measured observed emission factors), it was found that Equation 4-9
tends to overpredict the lowest and underpredict the highest measured factors. In other words, the model
appears to have a systematic bias at the extremes of the parent data base. This tendency is to be expected of
any model developed from regression techniques.
The only other significant relationship found for the residuals in the PM-10 cross-validation
involved the tendency of the equation to overpredict emissions for very slow speeds. The equation does not
exhibit any bias for mean vehicle speeds 15 mph and higher. Figures 4-7 and 4-8 present separate residual
plots for average vehicle speeds below and at 15 mph or higher, respectively. For the 19 tests conducted
with an average speed less than 15 mph, Figure 4-7 suggests overprediction by approximately 80 percent.
In contrast, at speeds higher than 15 mph (and especially for speeds 45 to 55 mph) the residuals are
symetrically distributed about the line of perfect agreement.
The finding that the equation overpredicts for very slow speeds also influences how to account for
the emission reduction due to speed control. This overprediction suggests that speed reduction has a near
linear effect on emissions. That is to say, for an approximately 50 percent reduction (i.e., from 30 mph to
less than 15 mph) in speed, the emission factor is roughly 50 percent lower than expected (i.e., overpredicted
by about 80 percent). This is consistent with the linear reduction based on the current AP-42 factor
(Equation 2-1). As discussed in Section 4.5, a linear effect for speed reduction is included in the revised
AP-42 section.
A second validation of the PM-10 factor reserved approximately 20 to 25 percent of the data base
for validation purposes. Test data were randomly selected for inclusion in either the "development" or the
"validation" data set. Two separate random selections were performed. The development data set is used to
develop the relationship which is used to estimate tests in the validation set. The first development set led to
the following predictive equation for PM-10:
E = 2.8 (s/12)0-78 (W/3f44 / (M/O.2)035 (4-10)
and Development Set 2 led to the following equation for PM-10:
E = 2.7 (s/12)aso (W/3f43 / (M/O.2)026 (4-11)
Note that both development sets led to equations very similar to that in Equation 4-5. When the two
models were used to predict data that had been withheld for validation, the following summary statistics
resulted:
Ratio of predicted to measured
Validation set
No. of cases
Minimum
Maximum
Geo. mean
Geo. std.dev.
1
n = 41
0.123
29.3
0.926
2.92
2
o
II
0.125
6.58
1.27
2.63
4-25
-------
Unlike the quasi-independent estimates obtained in the cross-validation, the above truly represent
independent applications of an emission factor model developed through stepwise regression technique. For
that reason, this validation leads to a slight bias in the resulting estimates, underpredicting in the first set by
7 percent and overestimating by roughly 30 percent in the second. Nevertheless, the spread (variation) in the
estimates is quite comparable to that found in the cross-validation and the estimates generally agree well
with the measured values in the validation data set.
A final PM-10 validation study involved nine emission tests that had not been formally reported
when the study began (Reference 15). Table 4-32 shows the results of the comparisons of predicted to
measured PM-10 emission factors. Predictions based on both Equation 4-5 and the current AP-42 equation
are considered. In general, agreement is quite good for the new unpaved road equation.
Validation of the PM-30 and PM-2.5 emission factors was also undertaken. For the PM-30, a
cross-validation similar to that performed for PM-10 led to results very comparable to those found earlier.
Figures 4-9 through 4-14 present the residuals from the PM-30 cross-validation. Interestingly, there was no
significant relationship between the residuals and speed for the PM-30 equation. In other words, unlike the
PM-10 equation, the PM-30 equation does not appear to systematically overpredict at very slow travel
speeds.
In the PM-30 cross-validation, the following results were found comparing the predicted to
measured values,
1. 50 percent were within a factor of 2;
2. 72 percent were within a factor of 3; and
3. 96 percent were within a factor of 5.
Remarks made earlier in connection with PM-10 bear repeating here. Recall that, in general, one is
more interested in how well the factor performs in estimating the total (or average) emission from several
roadway segments within a facility. In this way, there is considerably greater accuracy in the total emission
estimate than might be inferred from the above statistics. As in the case of PM-10, consider the example of
comparing the measured and predicted sums in random draws of five from the data set. In 100 realizations,
1. 83 percent were within a factor of 2;
2. 98 percent were within a factor of 3; and
3. All were within a factor of 5.
Note that the estimate for the total is substantially "tighter" than that for the individual road
segment.
Because the result for PM-2.5 in Equation 4-9 was not developed by stepwise regression, a different
type of validation was undertaken. In this case, the estimate based on Equation 4-9 was directly compared
to the measured emission factor contained in the data. Because PM-2.5 data were not used directly to
develop a regression-based model, the comparisons already represent essentially independent applications of
Equation 4-9. That is to say, there was no need to eliminate tests on a point-by-point basis and repeatedly
use stepwise regression to develop quasi-independent estimates.
4-26
-------
In comparing the Equation 4-9 estimates to the measured emission factors in the PM-2.5 data set, it
was found that, for individual test results,
1. 44 percent were within a factor of 2;
2. 68 percent were within a factor of 3; and
3. 78 percent were within a factor of 5.
Again, greater accuracy results when the predictive equation is applied to a set of roadway segments
to estimate total emissions. As discussed in connection with the PM-10 and PM-30 validations, an
illustration is provided by summing the emissions from five randomly selected tests from the data set. In
100 realizations of the random draw of five tests,
1. 62 percent were within a factor of 2;
2. 78 percent were within a factor of 3; and
3. 90 percent were within a factor of 5.
In summary, then, the validation found that Equations 4-5, -7 and -9 provide reasonably accurate
estimates of the PM-10, -30, and -2.5 emissions from an individual roadway. As noted throughout this
section of the document, one has substantially greater confidence when the predictive models are applied to a
set of roadways contained at a specific facility.
4.4 DEVELOPMENT OF DEFAULT VALUES FOR ROAD SURFACE MATERIAL PROPERTIES
As noted earlier, all previous versions of the AP-42 unpaved road emission factor have included the
road surface silt content as an input variable. The predictive equations recommended in the last section are
no exception. AP-42 Section 13.2 has always stressed the importance of using site-specific input
parameters to develop emission estimates. Recognizing that not all users will have access to site-specific
information, AP-42 has included methods to allow readers to determine default values appropriate to their
situation.6
* Table 13.2.2-1 currently in AP-42 contains default silt information for various applications. As
part of this update, the table was modified to (a) include updated information on construction sites and log
yards and (b) reformat the information for publicly accessible roads. Item (a) was a relatively
straightforward process. On the other hand, item (b) required a thorough reexamination, as described
below.
In order to develop default information for publicly accessible unpaved roads, a data set of available
silt and moisture contents was assembled. The 78 data points were collected either as part of a field
emission testing program or as input necessary to prepare emission inventories. Note that several of the
e The inclusion of the surface moisture content as an input variable is not considered to represent an undue
burden on the users of AP-42. In particular, the methods presented in AP-42 Appendix C.2 require oven
drying before sieving. In other words, determination of the silt content of a road surface sample requires
that the moisture content of the sample also be determined. Thus, users of AP-42 who have already
determined site-specific values for road surface silt content should have corresponding moisture content
information available as well.
4-27
-------
inventory-type samples were aggregated from subsamples collected from different road segments within
some portion of the study area.
Data are classified as being from either an "eastern" or a "western" location, based on the common
distinction between "pedalfer" and "pedocal" soils. For pedalfer soils common in the eastern U.S.,
precipitation exceeds evaporation. Conversely, evaporation is greater than precipitation in the West and the
soils are termed "pedocal." The 97th meridian is roughly coincident with the dividing line between pedalfer
and pedocal soils.
Also, to the extent practical, data were classified as being from a "gravel" or "dirt" type of unpaved
road surface. In this context, "dirt" refers to a road surface constructed from soils in the general vicinity of
the site without a crushed aggregate (stone, slag, etc.) being incorporated. Similarly, "gravel" refers to
surfaces in which aggregate material has been incorporated, regardless of whether the aggregate is crushed
stone or some other material (such as slag or scoria).
Statistical analysis of the data set was undertaken to examine whether significant differences exist
between the characteristics of eastern vs. western and gravel vs. dirt roads. Because the available data set
had not been developed for this use, i.e., specifically to explore how unpaved road surface characteristics
vary because of different road surface materials or different locations in the country, the data set contains
unequal subsets of data. The 78 data points are distributed as shown below:
Location
Surface type East West
Dirt 10 14
Gravel 15 31
Unknown 0 8
The unequal sample sizes make it difficult to efficiently examine differences. First, the choice of
statistical tests becomes limited. Generally, the most powerful methods to examine treatment and interaction
effects rely on having equal number of observations per cell. On an even more fundamental basis, there is a
question whether the available data represent a reasonably representative, random sample from the set of all
publicly accessible unpaved roads. That assumption would underlies any statistical test undertaken.
Because of the data limitations, a series of pairwise comparisons such as,
1. Eastern gravel vs. eastern dirt roads;
2. Eastern vs. western roads; and
3. Gravel vs. dirt roads.
were undertaken to determine if there existed significant differences in either moisture or silt content. The
small-sample comparison of means test was used with the level of significance set at 10 percent. When
appropriate, a one-sided alternative hypothesis was used. For example, one could reasonably expect, on an
a priori basis, that on average
1. Gravel roads have lower silt contents than dirt roads; and
2. Moisture contents are lower in the western U.S. than in the East
4-28
-------
When there was no a priori reason available, a two-sided alternative hypothesis was selected. For example,
there was no reason to suspect that the set of eastern gravel roads would have higher silt contents than
gravel roads in the west. In that case, the alternative hypothesis selected was that the mean silt contents for
eastern vs. western gravel roads are not equal.
Given the limitations on the available data set, it is not particularly surprising that the pairwise
comparisons led to somewhat contradictory findings. For example, although the data set indicated that
eastern dirt roads had a higher average moisture content than eastern gravel roads, that result was not
duplicated for western roads or for roads overall. Similarly, gravel surfaces were found to have a lower
mean silt content than dirt when (a) only eastern roads and (b) all roads were compared. That is, no
significant difference was found for silt contents between western gravel and dirt roads. Results from the
pairwise comparisons are summarized below. In the table, "S" and "M" indicate that a significant different
(10 percent level of significance) in the mean value of the silt and moisture content, respectively, was found
in the comparison.
Comparison of gravel vs. dirt Comparison of East vs. West
East S M Gravel
West -- Dirt ~ M
Overall S ~ Overall
In keeping with the findings summarized above, it was decided to provide separate default silt
values for gravel and dirt roads, for use throughout the United States (i.e., no distinction between east and
west).
Mean
Silt Content
Gravel Roads 6.4 percent
Dirt Roads 11 percent
Specification of an appropriate default moisture content for a dry road proved more problematic.
The overall mean moisture content in publicly accessible road data set was found as 1.1 percent. Although
this value potentially could have provided the default, it was believed that 1.1 percent did not adequately
represent the extremes of the data set. The data base contained moisture contents approximately 0.1 to
0.3 percent for roads even in what are not considered "dry" parts of the nation. For example, four samples
collected for an emission inventory of Grants Pass, Oregon, ranged from 0.14 to 0.38 percent in moisture
content, with a mean value of 0.24 percent. The four Raleigh, North Carolina ("BJ") tests presented in
Table 4-32 are associated with moisture contents between 0.07 and 0.1 percent. (In fact, the Raleigh test
series provided the lowest moisture contents in the entire data set. By comparison, moisture contents for the
desert [the Arizona, Palm Springs and Reno tests in References 6, 1 and 2, respectively] ranged from 0.17
to 0.48 percent.)
This situation is not surprising since the moisture content of the surface material of an unpaved road
is very dynamic. The moisture content is affected by a number of meteorological and physical parameters
that vary considerably with time and by location. For urban roads, rain is the primary meteorological event
which adds moisture to the road surface. The frequency, duration, and quantity of rain are important
aspects which determine the moisture content on any day and the long term average moisture content. The
4-29
-------
average annual number of rain days in the U.S. ranges from about 20 to over 200 with a variation in annual
rainfall from less than 4 inches per year to over 100 inches per year. The primary meteorological
parameters that affect the evaporation of moisture from the road surface include solar radiation,
temperature, dew point, and wind speed. The Class A pan evaporation is a reasonable indicator of the
evaporation potential. The variation in the annual Class A pan evaporation varies from about 25 inches per
year to over 120 inches per year. Some physical parameters which affect the moisture content of the surface
material include the amount and size distribution of the loose surface material and vehicle traffic on the
road. The amount and size distribution of the loose surface material would affect the maximum amount of
water that the surface material is capable of holding. Vehicle traffic enhances the evaporation of moisture
from the road surface due to the increase in surface air movement. The presence of trees and other natural
and man made formations may affect the moisture balance of the road surface material. As a result, the
selection of any single default moisture content would introduce significant bias for all but a few locations in
the U.S.
In the interest of encouraging AP-42 readers to collect site-specific data, a reasonably conservative
(worst case) value of 0.2 percent was selected for the default dry condition moisture content. This moisture
content value is higher than approximately 20 percent of all the publicly accessible uncontrolled road data
set. It should be noted that this moisture value is not the average moisture content of the road surface
material but is the minimum moisture content following an extended period without water additions to the
road surface.
Even though the default moisture value may be viewed as conservative, the default should not
generally lead to unacceptable emission estimates. This is due to the fact that moisture is raised to such a
low power (0.3 and 0.4) in the predictive emission factors. When the 0.2 percent default is substituted for
the site-specific moisture content for the 43 publicly accessible road tests in the PM-10 data set, all but four
results are within a factor of 2 of the estimate based on the site-specific value. At most, use of a default
value of 0.2 resulted in an estimate 2.5 times greater. Furthermore, on average, the increase in estimated
emission factor was only 12 percent when the default was substituted for the site-specific moisture content.
4.5 SUMMARY OF CHANGES TO AP-42 SECTION
4.5.1 Section Narrative
The major revisions to AP-42 Section 13.2.2, Unpaved Roads, are as follows:
1. Text surrounding the emission factor equation was revised to reflect the new equation and
provide more background information on how the equation was derived. Reference to the PM-15 size
fraction has been removed.
2. The discussion on defaults and quality ratings was substantially expanded. In particular, there is
a description of the model's performance when used to predict emissions from very slow-moving traffic and
a presentation of a default value for moisture content.
3. The extrapolation to annual conditions (incorporating natural mitigation) has been revised to
reflect the variables contained in the new equation. Readers who are interested in finer temporal and spatial
resolution are directed to the background reports area of the CHIEF web site
(http://www.epa.gov/ttn/chief/ap42back.html). An alternative procedure for estimating emissions on a
monthly basis is available as a spreadsheet file. Information required to use this procedure includes hourly
precipitation, humidity and snow cover data, and monthly Class A pan evaporation data.
4-30
-------
It is emphasized that neither the simple assumption underlying the annual estimates or the more
complex set of assumptions underlying the use of the alternative procedure have been verified in any
rigorous manner.
4. Section 13.2.2.3, "Controls," was re-organized and re-written. The section now begins with an
overview of three basic control methods (vehicle restrictions, surface improvement, and surface treatment).
Extensive new material was added to address the effect of speed reduction and watering on fugitive dust
emissions from unpaved roads. A new method for "prospective" analysis based on the alternative procedure
for estimating emissions using hourly precipitation data and Class A pan evaporation data was added.
Slight revisions were made to the material presented for chemical unpaved road dust suppressants.
5. The revised Table 13.2.2-1 is as follows [bold indicates additions, strikeouts indicate deletions]:
4-31
-------
Table 13.2.2-1. TYPICAL SILT CONTENT VALUES OF SURFACE MATERIAL
ON INDUSTRIAL AND RURAL UNPAVED ROADS3
Industry
Road Use Or Surface
Material
Plant
Sites
No. Of
Samples
Silt Content (%)
Range
Copper smelting
Iron and steel production
Sand and gravel processing
Stone quarrying and
processing
Taconite mining and
processing
Western surface coal
mining
[Construction sites
[Lumber sawmills
Rural roads
Municipal roads
Municipal solid waste
landfills
Plant road
Plant road
Plant road
Material storage area
Plant road
Haul road
[Haul road to/from pit
Service road
Haul road [to/from pit]
Haul road [to/from pit]
[Plant] Access road
Scraper route
Haul road
(freshly graded)
Scraper routes
Log yards
Gravel/crushed
limestone
Dirt
Unspecified
Disposal routes
1
19
1
1
2
4
1
1
3
2
3
2
7
2
7
3
3
135
3
1
10
20
8
12
21
2
10
5
20
2
9
32
26
20
16-19
0.2- 19
4.1 -6.0
2.4- 16
5.0- 15
5.0-15
2.4-7.1
3.9-9.7
2.8- 18
4.9-5.3
7.2 - 25
18-29
0.56-23
4.8-12
5.0- 13
1.6-68
0.4- 13
2.2-21
[Publicly accessible roads
Gravel/crushed
limestone
Dirt (i.e., local material
compacted, bladed, and
crowned)
46
24
0.10-15
0.83-68
References 1,5-16.
4-32
-------
4.5.2 Emission Factors
Analysis of the test data exhibited an emission factor equation appropriate for average conditions.
The equation no longer contains speed and mean number of wheels as parameters. The current data base
shows a correlation of emissions to the surface moisture content, which was added as a parameter. The
annual precipitation is now considered only when the emission factor equation is annualized for a particular
source. As with the old equation, the new equation allows for the emission calculations of different particle
sizes (PM-2.5, PM-10, and PM-30) with the use of appropriate constants. The old Section 13.2.2
Equation (1) is presented below (striked out) followed by the new Section 13.2.2 Equation (1).
Old Equation (1) e - k(5.9) (s/12)(S/30)(W/3)*^-(w/4>^(365-p/365)
where:
e = emission factor (lb/vmt)
k = particle size multiplier (dimensionless)
s = silt content of road surface material (%)
S = mean vehicle speed, (miles per hour [mph])
W = mean vehicle weight, megagrams (Mg) (ton)
w = mean number of wheels
p = number of days with at least 0.01 in. of precipitation per year
Aerodynamic particle size multiplier
Constant PM-2.5 PM-10 PM-15 PM-30
k (lb/VMT) 0^95 036 036 O^ft
New Equation (1) E - k(s/12) (W/3)
(M/0.2C)
where k, a, b
and c are empirical constants given below
E = size-specific emission factor (lb/vmt)
s = surface material silt content (%)
W = mean vehicle weight (tons)
M = surface material moisture content (%)
Constants for Equation 1 based on the stated aerodynamic particle size:
Constant
PM-2.5
PM-10
PM-30
k (lb/VMT)
0.38
2.6
10
a
0.8
0.8
0.8
b
0.4
0.4
0.5
c
0.3
0.3
0.4
Quality rating
C
B
B
4-33
-------
SYSTAT INFLUENCE PLOT N = 180 R= -.807
LRESID
0
0 0 0
0 0 0
0
0 0
0
0 0 0 00 00
0 0 000
000
00 0
0 0 0
0 00 0
00 0
0 0 0 0
0
00 0 0
000 000
00
0
0 0
00
0 0 0
0 0
0 0
0
0
00
0
0
0 00
0 0 0
0000
00
00 0
0 000
0 0 0
0 0
0 0
0 00
000
0
00 0
0
0
0 0 0 0
0 0 0 0
0 0 0 0 0 0 0
0 00 0
0 0
0 0
0
0 0
00
0 1
LEF10
Figure 4-1. PM-10 residuals (log-scale) versus PM-10 emission factor (log-scale).
4-34
-------
SYSTAT INFLUENCE PLOT N = 180 R= .009
0 00
3 00
00 0
2
LSILT
Figure 4-2. PM-10 residuals (log-scale) versus silt content (log-scale).
4-35
-------
SYSTAT INFLUENCE PLOT N = 180 R= -.005
LRESID
1
1
1
1
I 1
0
0
0
3
0
2
0
0 0
0
0
0
0
0
0
0
0
0 0
0
0
0
0 0 0
1
0
0
0 0
0 0
0
0
0 0
0 0
0
00
0
00
0 00
0
0 0
00
00 0
00
0
0
00 0
0 0 00 0 0
0
0
0
0 0
0
0
0
0
0 0 0
0 0
0
0
0
0
0 0
0
0
0
0
0
00 0 00
00 0
0 0 0
00 0 0 0 0 0 0
0
0
00
0 000 0
0
0 0 0
0
0
0
0
00 0
0 0 0
00
0 0
-1
0
0 0
0 0
0
0 0 0 0 0 0
0
0
0
0
0 0
0
0
0 0
0
0
-2
0 0
0
-3
i
1
1 1
1 1 1
-4 -3 -2 -1 0
LMOIST
SYSTAT INFLUENCE PLOT N = 180 R= -.005
Figure 4-3. PM-10 residuals (log-scale) versus moisture content (log-scale).
4-36
-------
SYSTAT INFLUENCE PLOT N = 180 R= -.005
LRESID
4
1
1
1
0
1 1
0
1
1
0
3
0
2
0
0
0
0
0
0 0
0
00
0
0
0
0
0 00
0
0
0
1
0 00
0
0
0
0
00
0
00
0
000 0
0 0
0 0
00
0
0 0
0 000
0 00 0
0
0
0 00
0
0
0
0 00
0
0
0
0
0
0 0 0
0
0
0 0
0
0
0 0 0
0
0 0
00
0
0
0
00
000
0
0
0
0 00
00
0
0
0
00
0
0 0 0
0
0
0
00
0
0 0
00
00 0 0
0
0
0
-1
0
00 0
0 0 0
0
000
0
0 0
0
0
0
0
0
0
0
0
-2
0
0
-3
t
1
1
1 1
3
LTONS
Figure 4-4. PM-10 residuals (log-scale) versus average vehicle weight (log-scale).
4-37
-------
SYSTAT INFLUENCE PLOT N = 180 R= -.143
LRESID
4
l
1
0
0
1
1 1
0
3
0
2
0
0
0
0
0 0
0
0
0
0
00
0
0
00
1
0
0
00
00
0
0
0
0 0
0
0
0
0
0
0
0
0
0
0
00 00
0
0
0
0
0
0
0 0
0 0
0
0
0
00
0
0
0
0
00
0 0
0
0
0
0
0
00
0
0
00
0 0
0
00
0
0 0
0
0 0
0
00
0
0
000
00
00
0
0
0
00
0
0
0
00
000
0
0
0
00
-1
0
0
0
0 0
00
0
0
00
0
0
0
00
0
0
0
0
0
0
-2
00
0
-3
1
1 1
1 2 3 4 5
LMPH
Figure 4-5. PM-10 residuals (log-scale) versus average vehicle speed (log-scale).
4-38
-------
SYSTAT INFLUENCE PLOT N = 180 R= .051
0
0
0
0
0
0
0
0 0
0
0 0
0 0
0
0
0 0
0 0
0 0
0
0 0
0 0
0
0
00
0
0 0
0
0
0 0
0 00
00 00
0 0
0 0
0
0 00
0 0
00 0
0 00
0 0 0 0
0 0
2.0
LWHLS
Figure 4-6. PM-10 residuals (log-scale) versus average number of wheels (log-scale).
4-39
-------
SYSTAT INFLUENCE PLOT
LRESID
Figure 4-7. PM-10 residuals (log-scale) versus average vehicle speed <15 mph.
4-40
-------
SYSTAT INFLUENCE PLOT N =
161
R=
-.034
1 i
0
I
1
1
0
0
0 0
0
0 0
0
0
0
0 0
0 0
0
0
00
0
0
0
0 0
0
0
0
0 0
0
0
0
0 0 0 0
0
0
0 0
0
0 0
00
00
0 0
0
0
0
0
0
0 0 0
0
0 0
0
0
0
0
00 0
0
00
0 00
0
00
0
00
0
0
0 0
0
0
0
0
0
0 0
0
0
000
0
0
0
0
0
00
00 0
0
00
0 0
0
00
0
0
0 0
00
0
00
0
0
0
0
00
0
0
0
0
0
0
0
0
0
0
0
0
0
1 1
0
1
1
1
¦
0 10 20 30 40 50 60
Figure 4-8. PM-10 residuals (log-scale) versus average vehicle speed >15 mph.
4-41
-------
SYSTAT INFLUENCE PLOT
92
~r
o o
0
0
0
00
0 0
00 0
0
0 0 0 0
0 0
0 0
0
0
0 0
0
0 0
0 0 0 0
0 0
0
0 0 0
0 0
2
LEF
Figure 4-9. PM-30 residuals (log-scale) versus PM-30 emission factor (log-scale).
4-42
-------
SYSTAT INFLUENCE PLOT N
0 0
0 00
0 0
0 00
0 0
0 0
2
LSILT
Figure 4-10. PM-30 residuals (log-scale) versus surface silt content (log-scale).
4-43
-------
SYSTAT INFLUENCE PLOT
U 0
00 0
0000
LMOIST
Figure 4-11. PM-30 residuals (log-scale) versus surface moisture content (log-scale).
4-44
-------
SYSTAT INFLUENCE PLOT N = 92 R= -.003
LRESID
1 I
1
0
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
00
0
0
0
0
0
0 0
0
0
0
0
0
0
0
00
0
0
0
0
0
0
0
0
00
0
0
0
00
0
0
0
0
0
00
0
0
0
0
0
0
0
0
0 0
0
0 0
0
0
0
0
0
0
0 0
0
0
0
0
1
1 1
0
1
0
1
1
0 1 2 3 4 5 6
LTONS
Figure 4-12. PM-30 residuals (log-scale) versus average vehicle weight (log-scale).
4-45
-------
SYSTAT INFLUENCE PLOT N = 92 R= .078
LRESID
0 0 0
0 00
00
0 0
0 0
0 0 0
0 0 0
0 0 0 0 0
0 0 0 00
0 0 0
0 0 0
0 0 0 0 0 0
0 0 00
0 0 0 0 0
0 0 0
0 0 0
0 0 0 0 0
0 0 0 0
0 0
0 0 0 0 0
0 0 0
0 0
Figure 4-13. PM-10 residuals (log-scale) versus average vehicle speed (log-scale).
4-46
-------
SYSTAT INFLUENCE PLOT N = 92 R= .102
LRESID
1
1
1
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 0
0
0
0
0
0
0 0 0
0
0
0
0
0 0
0
0
0
0
0
0
0
0
0 0
0
0
0
0
0
0
0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
1
'
1.8
LWHLS
Figure 4-14. PM-10 residuals (log-scale) versus average number of wheels (log-scale).
4-47
-------
TABLE 4-1. SUM
MARY INFORMATION - REFERENCE 1
Control
method
Test
run
Test
date
No.
of
tests
PM-10 emission factor, lb/VMT
Operation
State
Geom. mean
Range
Unpaved road
None
BK1-
BK4
Nevada
5/96
4
0.820
0.309-2.65
Paved road
None
--
Nevada
5/96
3
0.0025
0.0022-0.0028
1 lb/VMT = 281.9 g/VKT
TABLE 4-2. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 1
Unpaved
road test
runs
PM-10
emission
factor,
lb/VMT
Duration,
min.
Meteorology
Vehicle information
Mean
vehicle
speed,
mph
Silt,
%
Moisture
%
Temp.,
°F
Avg.
wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight,
ton
Mean
No. of
wheels
BK-1
0.375
59
72
6.0
138
1.5
4
15
7.2
0.48
BK-2
0.309
29
70
6.5
150
1.5
4
15
5.2
0.44
BK-3
1.48
47
70
6.6
100
2.0
4
15
5.9
0.45
BK-4
2.65
27
71
6.6
80
2.0
4
15
6.6
0.38
TABLE 4-3. SUMMARY INFORMATION - REFERENCE 2
Operation
Control
method
Unpaved
road test runs
State
Test
date
No.
of
tests
PM-10 emission factor, lb/VMT
Geom. mean
Range
Scraper
None
BA1-BA2
Nevada
6/95
2
8.19
6.05 -11.1
Scraper
None
BA3-BA6
California
6/95
4
0.838
0.550-1.32
Scraper
Watering
BA8-BA9
California
6/95
2
0.174
0.090-0.340
Light duty
None
BA10-BA12
California
7/95
3
7.24
3.33-12.5
4-48
-------
TABL]
E 4-4. DETAILED INFO
RMATION FOR UNPAVED ROAD TESTS - REFERE]
VCE 2
Vehicle information
Unpaved
road test
runs
PM-10
emission
factor,
lb/VMT
Duration,
min
Temp.,
°F
No. of
vehicle
passes
Mean
vehicle
weight,
ton
Mean No.
of wheels
Mean
vehicle
speed,
mph
Silt, %
Moisture,
%
BA-1
6.05
43
91
19
54.8
4.2
00
00
7.69
1.16
BA-2
11.1
22
91
12
58.5
4.0
9.5
7.69
1.16
BA-3
1.32
40
74
17
86.5
4.0
14
6.04
7.41
BA-4
0.580
40
74
17
86.5
4.0
14
6.04
7.41
BA-5
1.17
56
74
14
77.0
4.0
14
6.04
7.41
BA-6
0.550
56
74
16
77.0
4.0
14
6.04
7.41
BA-8
0.340
13
70
42
86.7
4.1
16
4.11
4.14
BA-9
0.090
16
70
74
79.6
4.1
16
3.35
5.69
BA-10
3.33
29
105
32
2.8
4.3
25
15.5
0.27
BA-11
9.10
35
105
29
2.0
4.0
25
15.5
0.27
BA-12
12.5
28
105
31
2.0
4.1
25
15.5
0.27
4-49
-------
TABLE 4-5. SUMMARY INFORMATION - REFERENCE 3
No.
of
tests
PM-10 emission
factor, lb/VMT
PM-2.5 emission factor,
lb/VMT
PM-1 emission factor,
lb/VMT
Operation
Control
method
Tests
State
Test
date
Geom.
mean
Range
Geom.
meanb
Rangeb
Geom.
meanb
Rangeb
Stone quarry
Haul truck
Watering
G-DWb
North Carolina
8/95
3
0.195
0.006-1.60
0.109
0.027-0.441
0.092
0.063 -0.136
Stone quarry
Haul truck
Watering
S-DW
North Carolina
8/95
3
1.37
0.490-2.99
0.353
0.137-1.32
0.059
0.015-0.360
1 lb/VMT = 281.9 g/VKT
Emissions reported are said to include noncombustible particles only. Upwind measurements were not adjusted for noncombustible particles in
report calculations.
bNegative emissions reported at Garner location are not included in range or geometric mean calculation.
TABLE 4-6. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 3
Meteorology
Vehicle information
Unpaved road test
runs
PM-10 emission
factor, lb/VMT
Duration,
min
Temp., °F
Avg. wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight, ton
Mean No.
of wheels
Average
vehicle
speed, mph
Silt, %
Moisture, %
G-DW-M201A-2
0.0061
356
88
4.66
204
NAa
NAb
18.55
7.22
5.96
G-DW-M201A-3
1.60
360
85
6.21
245
NAa
NAb
18.55
6.73
3.65
G-DW-M201A-4
0.76
360
86
6.35
200
NAa
NAb
18.55
8.23
9.68
S-DW-M201A-1
2.99
240
91
4.99
128
NAa
NAb
16.87
6.65
3.97
S-DW-M201A-2
0.49
300
90
3.69
250
NAa
NAb
16.87
9.81
6.44
S-DW-M201A-3
1.74
360
79
6.53
168
NAa
NAb
16.87
6.48
4.59
aMean vehicle weight not available - Estimated = 52 tons for AP-42 development.
bMean number of wheels not available - Estimated = 6 wheels for AP-42 development.
-------
TABLE 4-7. SUMMARY INFORMATION - REFERENCE 4
Uncontrolled TSP
emission factor,
lb/VMT
Uncontrolled PM-10
emission factor,
lb/VMT
Controlled TSP
emission factor,
lb/VMT
Controlled PM-10
emission factor,
lb/VMT
Operation
Location
State
Uncontrolled
test runs
Test date
No. of
tests
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Haul road
Summary
1, IB, 2
and 4
Wyoming
BB2-16, BB29-
34, BB36,
BB44-48
9/92-
10/92
42
31
0.49-95.1
5.5
0.08-15.6
15
4.64-
84.2
2.6
0.83 -
13.0
Coal Haul
Road
Site 1
Wyoming
BB2,3,10,11
9/92-
10/92
6
42
20.2-95.1
6.1
2.86-13.6
-
-
-
-
Coal Haul
Road
Site IB
Wyoming
BB6-8,
BB12-16,
BB45,BB48,
10/92
24
14
0.40-
20.2
3.6
0.08-6.52
10
4.64-
18.0
2.2
0.93 -
4.25
Coal Haul
Road
Site 2
Wyoming
BB33,34
10/92
4
46
44.4-47.9
7.3
5.70-9.48
17
10.2-
27.3
2.4
0.83 -
6.66
Overburden
Haul Road
Site 4
Wyoming
BB29,31,36,44
10/92
8
72
1.27-84.2
13
0.25-15.6
57
38.4-
84.2
5.8
2.61 -
13.0
Scrapper
Site 5
Wyoming
BB46,47
10/92
2
-
-
9.5
8.17-11.0
-
-
-
-
lb/VMT = 281.9 g/VKT
-------
TABLE 4-8. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 4
Site
Run
PM-10
emission
factor,
lb/VMT
Control
measure
Duration,
min.
Meteorology
Vehicle
Silt, %
Moisture,
%
Vehicle
passes
°F
Wind
speed
mph
Mean
vehicle
weight,
ton
Avg. No.
of wheels
Mean
vehicle
speed,
mph
1
BB-2
10.8
None
30
35
61
14.43
131
5.54
36.4
10.7
1.08
1
BB-3
13.6
None
25
35
61
14.43
131
5.54
36.4
10.7
1.08
IB
BB-6
4.67
None
55
40
74
9.68
200
5.80
22.7
3.57
1.19
IB
BB-7
6.51
None
66
45
74
9.60
200
5.73
22.4
3.57
1.19
IB
BB-8
5.20
None
29
18
79
8.68
220
5.56
21.2
3.78
1.01
1
BB-10
3.26
None
88
57
80
18.06
160
5.47
27.5
3.08
1.17
1
BB-11
1.79
None
89
57
80
18.02
160
5.47
27.5
3.08
1.17
IB
BB-12
1.49
None
58
50
73
14.29
155
5.80
22.6
2.24
1.09
IB
BB-13
1.49
None
60
50
73
14.26
155
5.80
22.6
2.24
1.09
IB
BB-14
2.62
None
80
44
59
9.88
92.0
5.18
22.9
3.32
1.77
IB
BB-15
4.37
None
64
41
62
11.39
183
5.66
21.3
2.05
1.39
IB
BB-16
5.18
None
63
51
62
10.01
178
5.57
22.1
2.05
1.39
IB
BB-17
1.63
Watering
79
50
65
12.73
169
5.48
24.6
2.08
1.80
IB
BB-18
4.25
Watering
93
71
65
9.92
184
5.97
23.0
1.34
1.29
IB
BB-19
3.13
Watering
67
47
65
8.15
192
5.74
22.8
1.25
1.45
IB
BB-20
2.69
Watering
53
41
68
7.98
175
5.66
24.3
3.89
1.40
IB
BB-21
1.81
Watering
82
32
78
8.11
218
5.75
22.8
1.76
2.00
IB
BB-22
1.38
Watering
36
32
82
4.54
161
5.50
24.3
1.70
2.50
IB
BB-23
0.940
Watering
52
33
87
7.55
181
5.70
22.6
1.90
4.10
IB
BB-25
1.24
Watering
62
40
60
18.17
207
5.70
19.2
3.82
4.00
IB
BB-26
2.97
Watering
79
63
66
13.51
183
5.65
21.8
2.45
4.40
IB
BB-27
3.86
Watering
72
42
69
12.05
244
5.81
19.5
2.72
1.89
4
BB-29
15.6
None
37
21
65
5.86
283
5.90
18.8
19.2
3.78
-------
Site
4
2
2
2
4
IB
IB
4
4
4
2
4
5
5
IB
IB
TABLE 4-8. (continued)
Run
PM-10
emission
factor,
lb/VMT
Control
measure
Duration,
min.
Meteorology
Vehicle
passes
Wind
speed
mph
Vehicle
Mean
vehicle
weight,
ton
Avg. No.
of wheels
Mean
vehicle
speed,
mph
Silt, %
BB-31
9.34
None
37
22
65
5.18
271
6.09
20i
19.2
BB-33
5.70
None
92
32
61
13.72
153
5.44
29.2
3.02
BB-34
9.45
None
72
36
63
12.24
170
6.06
28.6
4i
BB-35
6.65
Watering
87
32
60
8.27
173
5.44
28.0
3.71
BB-36
14.2
None
44
21
69
4.63
286
6.00
19.3
12.9
BB-38
3.22
Watering
50
43
53
22.71
141
5.26
22.0
1.57
BB-39
1.70
Watering
45
40
53
22.52
137
5.25
21.8
1.44
BB-40
2.62
Watering
78
40
45
12.24
271
6.05
21.2
4.79
BB-41
5.66
Watering
97
51
45
1U
267
5.92
22.3
6.48
BB-42
13.0
Watering
70
36
44
11.63
275
5.94
22.0
9.48
BB-43
BB-44
BB-46
BB-47
BB-45
BB-48
0.810
0.25
11.0
8.16
0.0782
0.120
Watering
None
None
None
None
None
48
105
89
44
75
50
25
200
32
14
322
381
62
69
80
80
53
53
14.11
9.01
10.13
5.31
9.93
7.71
164
2.00
63.0
65.0
2.00
2.00
5.52
4.00
4.06
4.00
4.00
4.00
30.4
30.0
15.5
18.0
30.0
30.0
1.78
1.82
12.7
14.0
1.95
1.95
-------
TAB]
LE 4-9. SUMM
ARY INFORMATION -
REFERENCE 5
Control
method
Test
date
No. of
tests
PM-10 emission factor,
lb/VMT
Operation
Tests
State
Geom. mean
Range
Stone quarry
haul truck
Watering
W-201A-1 to
W-201A-3
North Carolina
8/95
3
0.112
0.0553-0.217
Stone quarry
haul truck
None
D-201A-1 to
D-201A-4
North Carolina
8/95
4
1.74
0.528-4.70
1 lb/VMT = 281.9 g/VKT
4-54
-------
TABLE 4-10. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 5
Unpaved road test
runs
PM-10
emission
factor,
lb/VMTa
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of vehicle
passes
Mean vehicle
weight, tonb
Mean No. of
wheels0
Average
vehicle
speed, mphb
W-201A-1
0.116
330
69
2.7
190
52.5
NA
16.94
5.86
5.59
W-201A-2
0.055
360
63
1.1
192
52.5
NA
16.94
7.35
6.31
W-201A-3
0.217
180
57
1.0
95
52.5
NA
16.94
7.19
5.87
D-201A-1
0.528
70
62
2.3
33
52.5
NA
16.94
8.54
2.22
D-201A-2
1.57
120
72
1.6
72
52.5
NA
16.94
7.34
1.19
D-201A-3
2.34
90
73
1.3
57
52.5
NA
16.94
9.25
1.31
D-201A-4
4.70
120
62
2.1
78
52.5
NA
16.94
11.03
0.83
aEmission Factors are average of left hood and right hood concentrations.
bMean vehicle weight and average vehicle speed were a representative sample applied to entire testing
period.
cMean number of wheels not reported, estimated mean from truck description = 6.
-------
TABLE 4-11. SUMMA
RY INFORMATION - REFERENCE 6
Operation
Control
method
Test run
State
Test
date
TSP emission factor, lb/VMT
PM-10 emission factor, lb/VMT
No. of
tests
Geom.
mean
Range
No. of
tests
Geom.
mean
Range
35 mph rural road
None
AZ
Arizona
5/90
3
3.40
3.19-3.86
9
0.735
0.497 - 1.43
45 mph rural road
None
AZ
Arizona
5/90
3
4.59
3.56-5.94
9
1.26
0.777 - 2.97
55 mph rural road
None
AZ
Arizona
5/90
3
6.73
5.35-9.24
9
1.70
0.969 -2.88
1 lb/VMT = 281.9 g/VKT
I
Lh
On
-------
TABL]
E 4-12. DETA
[LED INFORMATION FOR UNPAVED ROAD TESTS
- REFERENCE 6
Unpaved road
test runs1
PM-10 emission
factor, lb/VMT
Duration, min.
Avg.
wind,
mph
Vehicle information
Silt, %
Moisture, %
No. of
vehicle
passes
Mean vehicle
weight, ton
Mean No. of
wheels
Mean vehicle
speed, mph
AZ-01
0.780
21
4.9
53
1.9
4.0
45
11
0.2
AZ-02
b
21
4.9
53
1.9
4.0
45
11
0.2
AZ-03
0.920
22
6.0
55
1.9
4.0
45
11
0.2
AZ-04
0.880
22
6.0
55
1.9
4.0
45
11
0.2
AZ-05
1.35
71
4.2
62
1.9
4.1
55
11
0.2
AZ-06
1.46
71
4.2
62
1.9
4.1
55
11
0.2
AZ-01
0.970
31
4.8
54
1.9
4.0
55
11
0.2
AZ-08
b
31
4.8
54
1.9
4.0
55
11
0.2
AZ-09
0.500
97
5.9
172
1.9
4.0
35
11
0.2
AZ-10
b
97
5.9
172
1.9
4.0
35
11
0.2
AZ-11
0.670
96
3.9
178
1.9
4.0
35
11
0.2
AZ-12
0.630
96
3.9
178
1.9
4.0
35
11
0.2
AZ-21
0.810
42
8.2
98
1.6
4.0
45
7.4
0.22
AZ-22
0.920
42
8.2
98
1.6
4.0
45
7.4
0.22
AZ-23
1.16
47
5.0
50
1.6
4.0
45
7.4
0.22
AZ-24
b
47
5.0
50
1.6
4.0
45
7.4
0.22
AZ-25
1.55
27
5.4
51
1.6
4.0
55
7.4
0.22
AZ-26
b
27
5.4
51
1.6
4.0
55
7.4
0.22
AZ-27
2.01
39
7.4
77
1.6
4.0
55
7.4
0.22
AZ-28
2.01
39
7.4
77
1.6
4.0
55
7.4
0.22
AZ-29
0.730
50
7.0
153
1.6
4.0
35
7.4
0.22
AZ-31
0.630
82
4.0
105
1.6
4.1
35
7.4
0.22
AZ-32
b
82
4.0
105
1.6
4.1
35
7.4
0.22
AZ-33
0.650
46
6.4
134
1.8
4.0
35
7.4
0.22
AZ-41
1.03
96
3.8
155
1.6
4.1
35
4.3
0.17
AZ-42
0.680
96
3.8
155
1.6
4.1
35
4.3
0.17
AZ-43
1.43
76
3.7
107
1.6
4.0
35
4.3
0.17
AZ-44
b
76
3.7
107
1.6
4.0
35
4.3
0.17
AZ-45
1.28
48
3.9
72
1.6
4.0
55
4.3
0.17
-------
TABLE 4-12. (continued)
Unpaved road
test runsa
PM-10 emission
factor, lb/VMT
Duration, min.
Avg.
wind,
mph
Vehicle information
Silt, %
Moisture, %
No. of
vehicle
passes
Mean vehicle
weight, ton
Mean No. of
wheels
Mean vehicle
speed, mph
AZ-46
b
48
3.9
72
1.6
4.0
55
4.3
0.17
AZ-47
2.88
97
3.0
35
1.6
4.0
55
4.3
0.17
AZ-48
2.62
97
3.0
35
1.6
4.0
55
4.3
0.17
AZ-49
2.97
72
5.2
36
1.6
4.3
45
4.3
0.17
AZ-50
2.57
72
5.2
36
1.6
4.3
45
4.3
0.17
AZ-51
1.91
115
5.0
45
1.6
4.0
45
4.3
0.17
AZ-52
b
115
5.0
45
1.6
4.0
45
4.3
0.17
aTest runs include simultaneously collected samples (ex. AZ-01 and AZ-02). Tests AZ-1 through 12, AZ-21 through -33, and AZ-41 through -52 conducted in
Pinal, Pima, and Yuma Counties, respectively.
bTSP emission factor.
I
Lh
00
-------
TABLE
-13. SUMMARY INFORMATION - REFERENCE 7
Controlled TSP emission factor,
lb/VMT
Controlled PM-10 emission factor, lb/VMT
Operation
Location
State
Test dates
No. of tests
Geom. mean
Range
Geom. mean
Range
Vehicle traffic
AU-X (Unpaved road)
PA
11/89
2
0.61
0.39-0.96
0.16
0.14-0.18
Vehicle traffic
Paved road
PA
11/89
6
0.033
0.012-0.12
0.0095
0.0009-0.036
Vehicle traffic
Paved road
PA
11/89
4
0.078
0.033-0.30
0.022
0.0071-0.036
lb/VMT = 281.9 g/VKT.
TABLE 4-14. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 7
Unpaved road -
test runs
PM-10
emission
factor, lb/VMT
Control method
Duration, min.
Meteorology
Vehicle information
Silt content, %
Temp., °F
Wind, mph
No. of vehicle
passes
Mean vehicle
weight, ton
Mean vehicle
speed, mph
AU-X-1
0.14
Chemical
168
62
8.7
110
3.9
25
3.3
suppressant
AU-X-2
0.18
Chemical
71
60
6.5
101
2.1
26
4.1
suppressant
-------
TABL
4-15. SUMMARY INFORMATION - REFERENCE 8
TSP emission
factor, lb/VMT
IP emission factor,
lb/VMT
PM-10 emission
factor, lb/VMT
PM-2.5 emission factor,
lb/VMT
Operation
Control
method
Test runs
State
Test date
No. of
tests
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Heavy-duty
traffic
None (U)
AP
Indiana
5/85 & 8/85
4
10.3
2.20-
37.6
1.21
0.064 -
7.91
2.55
0.575 -
6.42
0.408
0.156-
0.791
Heavy-duty
traffic
Calcium
chloride (C)
AP
Indiana
5/85 & 8/85
1
1.26
1.26
-
-
-
-
-
-
Heavy-duty
traffic
Petro Tac (P)
AP
Indiana
5/85 & 8/85
5
2.59
0.645-
7.70
0.305
0.076-
1.46
0.193
0.048-
1.08
0.066
0.019-
0.369
Heavy-duty
traffic
Coherex (X)
AP
Indiana
5/85 & 8/85
5
4.68
0.653-
21.3
0.776
0.108-
4.26
0.564
0.078-
3.20
0.079
0.011-
0.766
Heavy-duty
traffic
None (U)
AQ
Missouri
9/85, 10/85, &
11/86
2
6.67
5.68-
7.84
1.47
1.25-
1.72
1.00
0.851-
1.18
0.180
0.153-
0.212
Heavy-duty
traffic
Calcium
chloride (C)
AQ
Missouri
9/85, 10/85, &
11/86
6
2.09
0.211-
17.5
0.279
0.032-
3.87
0.144
0.008-
2.98
0.418
0.102-
0.922
Heavy-duty
traffic
Generic (G)
AQ
Missouri
9/85, 10/85, &
11/86
11
3.05
1.27-
14.5
0.728
0.397-
2.46
0.546
0.279-
2.03
0.118
0.029-
0.724
Heavy-duty
traffic
Petro Tac (P)
AQ
Missouri
9/85, 10/85, &
11/86
5
4.84
2.57-
11.9
0.781
0.387-
2.26
0.572
0.283-
1.78
0.134
0.064-
0.582
Heavy-duty
traffic
Soil Sement
(S)
AQ
Missouri
9/85, 10/85, &
11/86
11
1.63
0.200-
6.78
0.265
0.050-
1.08
0.176
0.014-
0.816
0.053
0.009-
0.148
Heavy-duty
traffic
Coherex (X)
AQ
Missouri
9/85, 10/85, &
11/86
9
2.14
0.208-
10.5
0.282
0.034-
1.42
0.182
0.017-
1.11
0.104
0.013-
0.334
1 lb/VMT = 281.9 g/VKT.
-------
TABLE 4-16. DETAILED INFORMAT
ON FOR UNPAVED ROAD TESTS - REFERENCE 8
Unpaved road test
runs
PM-10
emission
factor,
lb/VMT
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of vehicle
passes
Mean vehicle
weight, ton
Mean No. of
wheels
Avg. vehicle
speed, mplV
AP2-P
0.0479
128
70
11
68
27
12.3
15
1.9
0.46
AP2-X
-
128
70
7.6
68
27
12.3
15
<0.05
0.50
AP2-C
-
128
70
4.2
65
28
11.8
15
2.7
1.2
AP2-U
6.42
127
70
4.2
8
33
7.0
15
8.1
0.64
AP3-P
0.124
119
70
11
50
29
7.08
15
2.6
0.36
AP3-X
0.0780
119
70
8.5
50
29
7.08
15
<0.05
1.4
AP3-C
-
119
70
8.5
50
29
7.08
15
4.3
1.4
AP3-U
4.47
119
70
6.2
10
37
5.2
15
8.3
1.1
AP5-P
1.08
84
73
2.6
34
28
13.9
15
6.1
0.12
AP5-X
3.20
82
73
3.9
34
28
13.9
15
11
0.14
AP6-P
0.178
59
75
2.0
51
26
17.4
15
6.8
0.13
AP6-X
1.38
56
75
3.7
51
26
17.4
15
10
0.08
AP6-U
-
46
75
3.7
51
26
17.4
15
7.3
0.10
AP7-P
0.231
104
72
0.92
87
26
13.5
15
11
-
AP7-X
0.293
109
72
1.6
90
26
13.4
15
12
-
AP7-U
0.575
87
72
1.6
85
25
13.4
15
6.0
-
AQ1-U
0.851
64
82
8.4
50
10
6.0
15
7.0
1.5
AQ1-G
0.887
66
82
8.4
50
10
6.0
15
7.6
1.5
AQ1-S
0.201
75
82
8.4
50
10
6.0
15
0.6
0.94
AQ1-X
0.809
75
82
8.4
50
10
6.0
15
15
1.2
AQ2-U
1.18
69
82
8.7
68
9.8
5.9
15
7.0
1.5
AQ2-G
1.04
82
82
8.7
68
9.8
5.9
15
7.6
1.5
AQ2-S
0.158
85
82
8.7
68
9.8
5.9
15
0.6
0.94
AQ2-X
0.504
82
82
8.7
68
9.8
5.9
15
15
1.2
AQ3-P
0.401
105
75
11
76
9.7
5.9
15
3.1
1.8
AQ3-G
0.329
52
75
9.0
19
9.3
5.8
15
6.8
1.5
-------
TABL
4-16. (continued)
Unpaved road test
runs
PM-10
emission
factor,
lb/VMT
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of vehicle
passes
Mean vehicle
weight, ton
Mean No. of
wheels
Avg. vehicle
speed, mpha
AQ3-S
0.135
50
75
9.0
19
9.6
5.9
15
1.5
1.1
AQ3-X
0.103
47
75
9.0
19
9.6
5.9
15
12
1.6
AQ4-G
2.03
22
75
11
50
24
6.0
15
6.8
1.5
AQ4-S
0.440
28
75
10
50
24
6.0
15
1.5
1.1
AQ4-X
0.585
22
75
12
50
24
6.0
15
12
1.6
AQ4-C
0.451
33
75
13
50
24
6.0
15
-
-
AQ5-P
1.78
21
63
5.9
34
24
5.9
15
5.0
1.1
AQ5-G
0.497
20
63
5.9
34
24
5.9
15
10
1.3
AQ5-S
0.816
29
63
5.9
34
24
5.9
15
4.4
0.99
AQ5-C
2.98
20
63
5.9
34
24
5.9
15
12
1.4
AQ6-P
0.568
18
75
5.0
44
24
6.0
15
5.0
1.1
AQ6-G
0.812
28
75
5.0
36
24
6.0
15
10
1.3
AQ6-S
0.646
23
75
5.0
36
24
6.0
15
4.4
0.99
AQ6-C
2.43
23
75
5.0
36
24
6.0
15
12
1.4
AQ7-P
0.283
30
64
6.5
50
24
6.0
15
3.6
1.2
AQ7-G
0.390
25
64
6.5
48
24
6.0
15
7.0
1.2
AQ7-S
0.284
28
64
6.5
50
24
6.0
15
2.9
0.95
AQ7-X
0.929
28
64
6.5
50
24
6.0
15
6.7
-
AQ8-P
0.536
22
70
5.0
36
24
6.0
15
3.6
1.2
AQ8-G
0.401
16
70
5.0
34
24
6.0
15
7.0
1.2
AQ8-S
0.422
17
70
5.0
34
24
6.0
15
2.9
0.95
AQ8-X
1.11
17
70
5.0
34
24
6.0
15
6.7
-
AQ9-G
0.282
110
64
6.5
125
10
6.0
15
.76
0.95
AQ9-S
0.0145
110
64
6.5
125
10
6.0
15
1.2
0.77
AQ9-X
0.0200
62
64
6.5
79
10
6.0
15
1.1
0.78
AQ9-C
0.0084
267
64
6.5
125
10
6.0
15
1.6
2.1
-------
TABLE 4-16. (continued)
Unpaved road test
runs
PM-10
emission
factor,
lb/VMT
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of vehicle
passes
Mean vehicle
weight, ton
Mean No. of
wheels
Avg. vehicle
speed, mpha
AQ10-G
0.279
138
61
6.6
200
7.6
5.3
15
2.9
1.3
AQ10-S
0.0340
134
61
6.6
200
7.6
5.3
15
-
-
AQ10-X
0.0168
129
61
6.6
200
7.6
5.3
15
-
-
AQ10-C
0.0204
133
61
6.6
200
7.6
5.3
15
-
-
AQ11-G
0.422
127
55
8.7
250
6.5
5.0
15
2.9
1.3
AQ11-S
0.0848
127
55
8.7
250
6.5
5.0
15
-
-
AQ11-X
0.0255
130
55
8.7
250
6.5
5.0
15
-
-
AQ11-C
0.0161
130
55
8.7
250
6.5
5.0
15
--
--
aTests at AQ were conducted with captive traffic and vehicles were operated at 15 mph. For test runs, control methods were c
following codes: C = calcium chloride, G = Generic, P = Petro Tac, U = uncontrolled, S = Soil Sement, X = Coherex.
esacribed with the
On
OJ
-------
TABLE 4-17. SU
MMARY INFORMATION - REFERENCE
39
Operation
Test
runs
State
Test
date
No. of
tests
TSP emission factor,
lb/VMT
IP emission factor,
lb/VMT
PM-10 emission
factor, lb/VMT
PM-2.5 emission
factor, lb/VMT
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Uncontrolled tests -
Scraper Travel
Controlled Tests -
Scraper Travel
AN24-
AN25
AN21-
AN23
Michigan
Michigan
8/85
8/85
4
7
51
10
41 -64
2.1 -37
34
9.2
28-43
1.5-27
26
5.3
22-33
1.2-21
7.7
1.6
6.3 - 10
.47-7.2
lb/VMT = 281.9 g/VKT
TABLE 4-18. DETAILED INFORMATION
OR UNPAVED ROAD TESTS - REFERENCE 9
Unpaved
road test
runs
PM-10
emission
factor,
lb/VMT
Control
method
Duration,
min.
Meteorology
Vehicle information
Silt,
%
Moisture,
%
Temp., °F
Avg.
wind,
mph
No. of
vehicle
passes
Mean vehicle
weight, ton
Avg. No. of
wheels
Mean vehicle
speed, mph
AN21U
1.90
Watering
46
84
3.8
75
49
4
13
8.9
7.3
AN21X
1.20
Watering
57
84
3.7
59
49
4
15
8.9
8.7
AN21Y
6.70
Watering
81
84
3.9
99
49
4
16
8.9
3.5
AN22U
21.0
Watering
56
81
4.1
49
49
4
17
5.9
2.3
AN22Y
11.0
Watering
61
79
3.7
45
49
4
17
5.9
3.1
AN23U
7.30
Watering
35
77
3.1
40
49
4
16
8.4
3.6
AN23Y
4.80
Watering
15
72
2.1
20
49
4
16
8.4
3.4
AN24U
27.0
None
23
82
7.1
20
49
4
18
7.7
1.7
AN24Y
22.0
None
23
82
7.1
20
49
4
18
7.7
1.7
AN25U
33.0
None
12
83
6.8
10
49
4
20
7.7
1.7
AN25Y
30.0
None
12
83
6.8
10
49
4
20
7.7
1.7
AN21U = Site "AN" test no. 21 at station "U."
-------
TABLE
-19. SUMMARY INFORMATION - REFERENCE 10
TP emission
factor, lb/VMT
TSP emission
factor, lb/VMT
IP emission
factor, lb/VMT
PM-10 emission
factor, lb/VMT
PM-2.5 emission
factor, lb/VMT
Operation
Control
method
Test run
State
Test
date
No. of
tests
Geom.
Mean
Range
Geom.
Mean
Range
Geom.
Mean
Range
Geom.
Mean
Range
Geom.
Mean
Range
Heavy-duty
traffic
None
AL 1, 3,
4, 7, 8,
9, 12
Indiana
6/84
6
10.4
7.16 -
15.9
4.66
3.69 -
7.13
3.20
2.65 -
4.82
2.46
2.02 -
3.75
0.781
0.618 -
1.23
Light/
Medium
duty traffic
None
AL 2, 6,
10, 11
Indiana
6/84
4
4.61
2.54 -
6.88
2.13
1.75 -
2.88
1.39
1.12 -
2.02
1.09
0.860 -
1.58
0.377
0.274 -
0.524
lb/VMT = 281.9 g/VKT
TABLE 4-20. DETA
LED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 10
Unpaved road
test runs
PM-10 emission
factor, lb/VMT
Duration,
min
Meteorology
Vehicle information
Silt, %
Temp., °F
Avg. wind,
mph
No. of vehicle
passes
Mean vehicle
weight, ton
Mean No. of
wheels
Mean vehicle
speed, mph
AL-1
7.16
40
64
6.2
40
22
12
19
11.1
AL-2
3.05
55
64
6.3
31
7.7
5.2
20
11.1
AL-3
7.90
24
80
7.6
41
28
14
19
10.6
AL-4
13.3
24
80
9.2
41
27
13
20
10.6
AL-6
4.04
20
80
9.0
42
7.1
4.7
20
10.6
AL-7
9.36
29
73
5.4
42
28
14
17
11
AL-8
8.12
31
73
4.8
40
33
16
18
11
AL-9
3.65
44
59
11
67
31
15
25
6.9
AL-10
3.27
37
59
12
50
9.0
5.6
20
6.9
AL-11
5.60
30
59
14
50
11
6.3
20
6.9
AL-12
7.80
25
60
6.0
39
32
15
16
10.3
-------
TABLE 4-21. SUMMARY INFORMATION - REF
TP emission factor,
lb/VMT
IP emission factor,
lb/VMT
PM-10 emission factor,
lb/VMT
PM-2.5 emission factor,
lb/VMT
Operation
Type
Control
method
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Rural roads
Crushed
Limestone -
Light duty
None
21.9
17.9-27.0
3.84
3.17-4.99
2.17
1.75-3.09
0.334
0.300-0.407
Rural roads
Dirt -
Light duty
None
28.6
11.1-42.1
3.42
2.83-4.18
1.60
0.951-1.99
0.293
0.090-0.507
Rural roads
Gravel - Light
duty
None
6.70
5.43-7.96
1.25
1.10-1.39
0.835
0.713-0.957
0.366
0.251-0.481
Copper smelter
Medium duty
vehicle
None
8.99
7.62-10.0
2.57
2.21-2.97
1.67
1.46-1.91
0.317
0.283-0.370
Stone crushing
Medium duty
vehicle
None
25.0
9.36-35.2
7.1
3.20-9.67
--
2.15-5.83
4.17
2.15-5.83
Sand and
gravel
Heavy duty
vehicle
None
11.1
8.28-15.3
3.92
3.35-4.44
2.73
2.34-3.26
0.742
0.620-0.982
;rence 11
1 lb/VMT = 281.9 g/VKT
-------
TABLE 4-22. DETAILED INFORMATION FOR UNPAVED ROAD T
ESTS - R]
EFERENCE 11
Run No.
PM-10
emission factor,
lb/VMT
Industrial category
Type of traffic
Avg.
wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight,
ton
Mean
No.
wheels
Mean
vehicle
speed,
mph
Silt
content,
%
Moisture
content, %
U-l
9.13
Rural roads crushed limestone
Light duty
8.28
125
1.9
4.0
35
9.5
0.25
U-2
3.09
Rural roads crushed limestone
Light duty
7.61
105
1.9
4.0
35
9.1
0.3
U-3
1.75
Rural roads crushed limestone
Light duty
2.46
101
1.9
4.0
35
7.7
0.27
U-4
1.87
Rural roads crushed limestone
Light duty
7.16
102
1.9
4.0
25
8.6
0.4
U-5
1.97
Rural roads crushed limestone
Light duty
11.6
107
2.3
4.0
25
9.2
0.37
U-6
--
Rural roads crushed limestone
Light duty
13.2
51
1.9
4.0
30
--
--
AB-1
12.1
Rural roads dirt
Light duty
13.2
94
2.3
4.0
25
35.1
3.9
AB-2
0.950
Rural roads dirt
Light duty
6.49
50
2.3
4.0
25
16.7
4.5
AB-3
1.99
Rural roads dirt
Light duty
8.50
50
2.3
4.0
25
16.8
3.2
AB-4
1.86
Rural roads dirt
Light duty
11.2
50
2.3
4.0
25
5.8
3.1
AE-1
0.710
Rural roads gravel
Light duty
9.62
46
2.1
4.0
40
5.0
0.26
AE-2
0.960
Rural roads gravel
Light duty
11.2
22
1.8
4.0
35
5.0
0.26
AA-1
2.15
Stone crushing
Med. duty
4.70
55
11
5.0
15
13.7
0.4
AA-2
0.940
Stone crushing
Med. duty
2.46
24
13
4.4
15
15.3
0.34
AA-3
0.090
Stone crushing
Med. duty
4.92
34
10
4.0
10
10.5
0.84
AA-4
4.52
Stone crushing
Med. duty
8.05
56
14
5.6
10
15.6
2.1
AA-5
5.83
Stone crushing
Med. duty
9.40
56
13
5.0
10
15.6
2.1
AC-1
1.63
Copper smelting
Light duty
4.25
51
2.2
4.8
10
19.1
0.07
AC-2
1.46
Copper smelting
Light duty
5.37
49
2.1
4.0
10
15.9
0.07
AC-3
1.91
Copper smelting
Light duty
6.93
51
2.4
4.3
10
16
0.03
-------
TABLE 4-23. SUM1V
[ARYIN
FORMATION - REFE]
RENCE 12
TP emission factor,
lb/VMT
IP emission factor,
lb/VMT
PM-2.5 emission factor,
lb/VMT
Operation
Control
method
Location
State
Test date
No. of
tests
Geom.
mean
Range
Geom.
mean
Range
Geom.
mean
Range
Heavy-duty traffic
None
E
Ohio
11/80
3
132
129-133
30.5
25.9 - 33.5
8.35
7.74 - 8.84
Heavy-duty traffic
Coherex
C
Ohio
11/80
4
5.04
3.35-8.17
1.48
1.18-2.04
0.439
0.274 - 0.594
Heavy-duty traffic
Watering
E
Ohio
11/80
3
28.9
8.27- 99.3
4.94
0.992 -25.8
1.07
0.219-5.46
Light-duty traffic
None
B
Ohio
7/80
4
11.7
9.98-14.2
2.69
1.05-4.25
0.731
0.245 - 1.27
Light-duty traffic
Coherex
B
Ohio
10/80
5
0.636
0.089 - 1.23
0.226
0.061 -0.384
0.0628
0.0318 -0.0945
1 lb/VMT = 281.9 g/VKT
-p^
i
On
00
-------
TABLE 4-24. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 12
Site
Unpaved
road test
runs
PM-10
emission
factor,
lb/VMTa
Type
Control
Duration,
min.
Meteorology
Vehicle information
Silt, %
Temp., °F
Avg. wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight, ton
Avg. No. of
wheels
Mean
vehicle
speed,
mph
E
F-68
25.1
Heavy-duty
None
17
50
7.4
21
22
5.9
20
14
E
F-69
20.6
Heavy-duty
None
13
50
7.9
14
53
10
20
-
E
F-70
25.3
Heavy-duty
None
13
50
8.2
10
53
10
20
16
E
F-65
0.70
Heavy-duty
Watering
57
60
6.4
64
53
10
20
4.5
E
F-66
3.53
Heavy-duty
Watering
20
60
5.5
41
54
9.0
25
-
E
F-67
19.4
Heavy-duty
Watering
17
55
9.5
30
54
9.8
25
5.1
C
F-59
--
Heavy-duty
Coherex
125
50
9.3
61
19
9.3
16
5.4
C
F-60
--
Heavy-duty
Coherex
123
50
8.2
84
46
9.2
22
5.4
c
F-63
--
Heavy-duty
Coherex
107
50
5.2
118
54
7.7
18
2.5
c
F-64
--
Heavy-duty
Coherex
121
50
6.5
136
54
7.8
15
-
B
F-28
0.750
Light-duty
None
45
78
1.6
101
3
4
15
-
B
F-29
3.34
Light-duty
None
34
79
6.2
50
3
4
15
-
B
F-30
2.40
Light-duty
None
17
79
6.2
50
3
4
15
-
B
F-31
3.10
Light-duty
None
40
80
3.5
33
3
4
15
-
B
F-40
--
Light-duty
Coherex
133
50
4.0
300
3
4
25
0.015
B
F-41
--
Light-duty
Coherex
100
50
5.1
255
3
4
25
0.075
B
F-42
--
Light-duty
Coherex
128
50
7.0
294
3
4
25
0.99
B
F-43
--
Light-duty
Coherex
120
50
8.5
300
3
4
25
-
B
F-44
--
Light-duty
Coherex
55
50
9.1
200
3
4
25
1.8
aPM-10 emission factor calculated from logarithmic interpolation of PM-15 and PM-2.5 data.
-------
TABLE 4-25. SUMMARY INFORMATION - REFERENCE 13
Operation
Control
method
Test run
State
Test date
No. of tests
TSP emission
factor, lb/VMT
IP emission factor,
lb/VMT
PM-10 emission
factor, lb/VMT
PM-2.5 emission
factor, lb/VMT
Geom.
Mean
Range
Geom.
Mean
Range
Geom.
Mean
Range
Geom.
Mean
Range
Heavy-duty
traffic
None
AG1-3
Indiana
6/82
3
18.1
12.0-
23.4
3.80
1.38 -
7.47
3.05
1.34-
5.55
0.384
0.117-
0.994
Heavy-duty
traffic
Petro Tac
AG4-11
Indiana
6/82
8
3.39
0.963-
8.88
0.366
0.015-
2.24
0.282
0.035-
1.54
0.080a
0.0154 to
0.259
Heavy-duty
traffic
None
AJ1-3
Missouri
9/82
3
16.4
13.8-
21.4
3.79
2.94-
5.15
2.86
2.14-
4.17
0.694
0.498 -
0.915
Heavy-duty
traffic
Watering
AJ4-6
Missouri
9/82
3
1.77
0.255-
5.81
0.340
0.086-
0.781
0.242
0.051-
0.563
0.191
0.122-
0.272
Heavy-duty
traffic
Coherex
AJ7-18
Missouri
9/82
12
2.79
0.384-
16.6
0.42
0.047-
3.57
0.233
0.006-
2.23
0.076a
0.0049 to
0.449
1 lb/VMT = 281.9 g/VKT
''Only included test runs with reported measurements.
-------
TABLE 4-26. DETAILED INFORMATION FOR UNPAVED ROAD TESTS -
REFERENCE 13
Unpaved
road test
runs
PM-10
emission
factor,
lb/VMT
Duration,
min.
Meteorology
Vehicle information
Silt,
%
Moisture
%
Temp.,
°F
Avg.
wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight
, ton
Mean No. of
wheels
Mean vehicle
speed, mph
AG-1
1.34
31
71
4.2
27
27
9.8
15
7.5
0.59
AG-2
5.55
106
69
7.4
30
25
7.3
17
5.8
0.33
AG-3
3.82
99
70
5.8
22
28
6.6
16
7.2
0.27
AG-4
0.097
107
52
2.7
79
23
9.2
15
0.28
-
AG-5
0.248
128
69
4.8
120
32
10
14
0.29
-
AG-6
0.035
166
87
6.6
160
30
13
15
5.0
-
AG-7
0.136
202
71
2.2
84
34
10
16
4.9
-
AG-8
0.610
100
70
3.2
93
31
9.1
14
5.3
-
AG-9
1.54
75
69
6.3
31
28
6.1
13
8.2
-
AG-10
1.11
76
65
3.4
49
31
8.1
13
8.5
-
AG-11
0.335
62
74
2.6
62
26
5.8
14
13
-
AJ-1
4.17
48
77
3.3
45
54
6.0
15
6.3
-
AJ-2
2.62
46
76
2.0
47
52
6.0
15
7.4
-
AJ-3
2.14
50
80
4.2
50
50
7.1
15
7.7
-
AJ-4
0.060
79
90
6.1
86
48
6.1
15
4.9
5.1
AJ-5
0.560
67
85
5.6
71
50
6.0
15
5.3
2.0
AJ-6
0.493
46
78
4.4
49
48
5.9
15
-
-
AJ-7
0.490
90
66
3.6
68
49
5.9
15
1.9
-
AJ-8
0.022
89
70
5.8
120
34
7.2
15
5.5
-
AJ-9
1.05
126
69
5.3
120
50
6.4
15
7.1
-
AJ-10
1.49
50
62
2.8
44
29
6.0
20
6.1
-
AJ-11
0.904
65
65
3.1
61
27
6.0
19
4.3
-
A J-12
2.23
68
61
7.7
60
44
6.0
21
5.7
-
AJ-13
0.006
190
57
8.2
150
38
6.0
18
ND
-
A J-14
0.183
240
42
12
250
56
6.0
22
0.034
-
AJ-15
0.313
131
49
00
00
107
54
6.0
17
1.6
-
AJ-16
0.098
140
55
4.9
140
32
6.0
23
2.1
-
AJ-17
0.066
125
65
7.9
120
34
6.0
20
1.5
-
AJ-18
0.373
119
43
5.0
115
31
6.0
22
1.7
-
4-71
-------
TABLE 4-27. SUMMARY INFORMATION - REFERENCE 14
Operation
Control
method
Test Run
State
Test date
TSP emission factor,
lb/VMT
No. of
tests
Geom.
mean
Range
IP emission factor,
lb/VMT
Geom.
mean
Range
PM-2.5 emission factor,
lb/VMT
Geom.
mean
Range
Haul Truck
Haul Truck
Light/Medium
Duty Truck
Light/Medium
Duty Truck
None
Watering
None
CaC12
J9-J12,J20,J21,
K1,K7,K9-K12,
K26,L1,L3,L4,
Pl-3, P5
K6,K8,K13,P4,P
6-P9
J13,J18,J19,K2,
K3,K4,K5,P11,P
12,P13
J7,J8
North
Dakota,
Wyoming,
New Mexico
Wyoming,
New Mexico
North
Dakota,
Wyoming,
New Mexico
North Dakota
1979-80
1979-80
1979-80
1979-80
20a
10
10.8
2.97
2.94
0.35
0.70 - 73
0.60 - 8.4
0.60 - 9.0
ND-0.35
5.54
1.51
1.79
0.34
0.32-42
0.40-4.1
0.33 -6.6
ND-0.34
0.23
0.09
0.119
0.09
0.02-2.88
0.05-0.16
0.03 - 1.5
ND-0.09
lb/VMT = 281.9 g/VKT
aHaul Truck uncontrolled tests listed in report text =19 and watered tests = 9, however data tables list 20 uncontrolled and 8 watered tests.
bTest Run 17 was reported as a nondetect (ND). Geometric Mean was calculated using only the detected test.
-------
TABLE 4-28. DETAILED INFORMATION FOR UNPAVED ROAD TESTS - REFERENCE 14
Unpaved
road test ran
PM-10
emission
factor
lb/VMTa
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight, tons
Mean No.
of wheels
Mean
vehicle
speed, mph
J-6
~
67
76.1
0.9
39
--
--
--
7.9
5.4
J-9
4.6
51
82.94
4.8
41
65
8
19.3
9.4
3.4
J-10
14.1
52
87.8
4.4
45
60
7.7
19.3
9.4
2.2
J-ll
9.4
48
86.9
4.2
40
60
9.9
20
8.2
4.2
J-12
4.9
49
80.06
0.8
19
99
9.5
15
14.2
6.8
J-20
2.9
49
73.4
2.5
23
125
10
16.8
11.6
8.5
J-21
3.1
26
77
1.6
14
110
9.3
15
-
-
K-l
1.6
86
58.28
6.2
65
63
6.1
32.9
7.7
2.2
K-6
0.6
177
64.04
3.4
84
89
7.4
34.8
2.2
7.9
K-7
1.6
53
74.3
2.6
57
24
4.9
34.2
2.8
0.9
K-8
0.8
105
50.54
5.7
43
65
6.3
36
3.1
1.7
K-9
2
89
53.6
5
63
74
6.7
29.2
4.7
1.5
K-10
1.5
65
51.08
5
40
69
6.6
36
7.7
2
K-ll
1.5
64
54.5
5.2
50
73
6.5
30
8.9
2
K-12
2
58
59.9
5.4
43
95
7.3
36
11.8
2.3
K-13
0.3
73
39.2
3.7
78
64
6.6
31.7
1.8
2.7
L-l
0.2
92
33.26
1.9
57
95
8.8
26.1
13
7.7
L-3
27.7
47
55.76
6.5
26
107
9.3
20
13.8
4.9
L-4
20.9
48
56.48
6.1
32
86
8.3
20
18
5.1
P-l
11.3
57
95
3.8
15
79
8.5
26.7
4.7
0.4
P-2
2
95
80.6
1.8
10
42
7.2
26.1
4.7
0.4
P-3
6.3
89
80.6
3.8
18
94
9.7
31.1
4.1
0.3
P-4
1.2
135
80.6
3.7
48
55
7.6
31.7
2
0.3
P-5
3.4
108
89.6
2.8
38
47
7.1
31.1
3.1
0
P-6
0.7
112
84.2
2.2
48
25
5.6
31.7
2.8
2.9
P-7
2.3
95
84.2
2.5
35
61
7.6
31.1
2.4
1.5
-------
TABLE 4-28. (continued)
Unpaved
road test ran
PM-10
emission
factor
lb/VMTa
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight, tons
Mean No.
of wheels
Mean
vehicle
speed, mph
P-8
1.2
103
84.2
3
49
47
7.5
29.2
7.7
15.3
P-9
1.4
142
80.6
3.7
48
58
8.7
31.1
1.6
20.1
J-l
2.48
87
73.94
2.8
63
50
4.1
19.3
8.9
5.7
J-2
2.09
34
77
1.4
33
53
4
19.3
23.4
2.3
J-3
16.3
51
84.92
1.3
35
54
4.1
24.2
15.8
4.1
J-4
0.963
52
68
1.1
30
36
4
20
14.6
1.5
J-5
5.8
60
85.1
1.4
14
70
4
18
10.6
0.9
K-15
4.54
13
41
3.9
6
46
4
28
K-16
10.3
41
47.84
2.6
10
64
4
30
25.2
6
K-17
20.9
18
53.6
4
31
57
4.1
23
25.2
6
K-18
10.7
37
55.58
2.6
30
66
4
25
25.2
6
K-22
2.92
110
41
3
20
45
4
31.7
21.6
5.4
K-23
6.61
43
42.98
4.6
20
54
4
28
24.6
7.8
L-5
115
14
38.3
8.6
20
53
4
21.1
21
L-6
51.3
22
39.56
9.4
15
50
4
20
21
P-15
-
43
89.6
1.6
4
42
4
16.2
7.2
1
P-18
0.714
33
80.6
3.9
18
64
4
10
7.2
1
J-7
-
59
82.94
1.1
104
7
4.2
25
3
3.6
J-8
0.27
68
86
1.6
160
3
4
25
3
3.6
J-13
3.22
26
77.9
2.9
59
2.2
4
25
10.1
1
J-18
5.32
21
79.7
3.7
34
2.6
4
25
8.8
1.1
J-19
3.69
31
80.24
3.6
70
2.3
4.1
25
8.2
0.9
K-2
0.195
55
46.94
5.5
150
2.3
4
35
4.9
1.6
K-3
0.242
58
53.78
4.8
150
2.4
4
35
4.9
1.6
K-4
0.225
67
61.16
3.1
150
2.4
4
35
5.3
1.7
K-5
0.351
68
68.72
4.3
150
2.4
4
35.9
5.3
1.7
-------
TABL
: 4-28. (continued)
Unpaved
road test run
PM-10
emission
factor
lb/VMTa
Duration,
min.
Meteorology
Vehicle information
Silt, %
Moisture, %
Temp., °F
Avg. wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight, tons
Mean No.
of wheels
Mean
vehicle
speed, mph
P-ll
2.56
73
95
5.8
100
2
4
42.5
5.5
0.9
P-12
2.94
60
95
5.2
125
2
4
43.1
5.5
0.9
P-13
2.52
55
84.2
4.2
100
2
4
43.1
5.5
0.9
aPM-10 emission factors were calculated from the PM-15 and PM-2.5 data using
ogarithmic interpolation.
-------
TABLE 4-29. SUMIV
[ARY INFORMATION - REFERENCE
3 15
Operation
Control
method
Tests
State
Test date
No. of
Tests
PM-10 (<10 Mm)
Emission Factor
(lb/VMT)*
Geom.
Mean
Range
Lightweight
None
BG
Missouri
11/95 to
5
0.352
0.0884-1.12
vehicle
12/95
Lightweight
None
BJ
North Carolina
4/96
4
1.15
0.851-1.31
vehicle
Lightweight
None
BK
Nevada
5/96
4
0.819
0.309-2.63
vehicle
lb/VMT = 281.9 g/VKT
* Study reports a PM-2.5/PM-10 ratio of 0.15
4-76
-------
TABLE 4
i-30. DETA]
LED INFORMATION FOR UNPAVED ROAD TESTS - REFE
RENCE 15
Meteorology
Vehicle information
Unpaved road
test runs
PM-10
emission
factor,
lb/VMT
Duration,
min
Temp., °F
Avg.
wind,
mph
No. of
vehicle
passes
Mean
vehicle
weight,
ton
Mean No.
of wheels
Average
vehicle
speed,
mph
Silt. %
Moisture,
%
BG-1
0.503
85
60
4.2
110
2
4
30
7.2
0.93
BG-2
0.925
125
60
11.6
330
2
4
30
6.22
0.65
BG-3
1.12
84
65
12.2
300
2
4
30
6.07
0.54
BG-4
0.118
102
57
6.0
306
2
4
30
7.56
1.38
BG-5
0.0884
88
62
4.0
320
2
4
30
7.97
1.12
BJ-1
1.24
92
84
10.2
257
2
4
30
4.01
0.1
BJ-2
1.28
115
84
10.5
261
2
4
30
2.9
0.1
BJ-3
0.851
115
84
14.6
247
2
4
30
4.26
0.07
BJ-4
1.31
82
84
16.4
251
2
4
30
3.70
0.09
BK-1
0.372
59
72
5.0
138
2
4
30
7.2
0.48
BK-2
0.309
29
70
5.6
150
2
4
30
5.24
0.44
BK-3
1.49
47
70
6.5
100
2
4
30
5.88
0.45
BK-4
2.63
27
71
6.5
80
2
4
30
6.55
0.38
-------
TABL
E 4-31. RESULTS OF CROSS-VALIDATION
Uncontrolled/
watered
No. of
cases
Ratio of quasi-independent estimate to
measured emission factor
Type of vehicle/road
Geo. mean
Geo. std. dev.
Haul trucks
U
39
0.98
2.44
W
34
1.10
2.49
Overall
73
1.03
2.45
Light-medium duty/traffic on
industrial roads
U
29
1.09
2.85
Light-medium duty/traffic on
public roads
U
Overall
43
72
0.97
1.02
2.36
2.54
Heavy duty/traffic on
industrial roads
U
3
1.28
1.39
Scrapers in travel mode
U
23
0.82
3.62
w
9
1.00
5.13
Overall
32
0.87
3.93
TABLE 4-32. PREDICTED VS. MEASURED RATIOS FOR NEW UNPAVED ROAD EQUATION
USING REFERENCE 15 TEST DATA
Run
Silt, %
Moisture, %
Weight,
tons
Speed,
mph
No. of
wheels
Measured
PM-10
emission
factor,
lb/VMT
Ratio of Predicted to
measured
Equation 4-5
Current
AP-42
BJ-1
4.01
0.10
2
30
4
1.23
0.88
0.43
BJ-2
2.90
0.10
2
30
4
1.29
0.65
0.30
BJ-3
4.26
0.07
2
30
4
0.840
1.51
0.67
BJ-4
3.70
0.09
2
30
4
1.32
0.80
0.37
BG-1
7.20
0.93
2
30
4
0.503
0.95
1.89
BG-2
6.22
0.65
2
30
4
0.925
0.95
0.89
BG-3
6.07
0.54
2
30
4
1.12
0.81
0.71
BG-4a
7.56
1.4
2
30
4
0.118
6.95
8.44
BG-5a
7.97
1.1
2
30
4
0.088
10.3
11.9
aThese tests were conducted during misty conditions.
4-78
-------
REFERENCES FOR SECTION 4
1. Muleski, G., Midwest Research Institute, Letter Report of Field Tests, MRI Project No. 4470, "Road
Sampling," for Washoe County District Health Department, Reno, Nevada, August 1996.
2. Improvement of Specific Emission Factors (BACM Project I), South Coast AQMD, South Coast
AQMD Contract No. 95040, March 1996.
3. PM-10, PM2.5, PM-1 Emission Factors for Haul Roads at Two Stone Crushing Plants, National
Stone Association, Washington, D.C., November 1995.
4. Surface Coal Mine Emission Factor Study, U. S. Environmental Protection Agency, EPA Contract
No. 68-D2-0165, Assignment 1-06, Research Triangle Park, NC, January 1994.
5. PM-10 Emission Factors for a Haul Road at a Granite Stone Crushing Plant, National Stone
Association, Washington, D.C., December 1994.
6. Unpaved Road Emission Impact, Arizona Department of Environmental Quality, Phoenix, AZ, March
1991.
7. Roadway Emissions Field Tests at US Steel's Fairless Works, U.S. Steel Corporation, Fairless Hills,
PA, USX Purchase Order No. 146-0001191-068, May 1990.
8. Evaluation of the effectiveness of Chemical Dust Suppressants on Unpaved Roads, funded by LTV
Steel Company, Inc., prepared for U. S. Environmental Protection Agency, Office of Research and
Development, Washington, D.C., EPA No. 600/2-87-102, November 1987.
9. Fugitive Emission Measurement of Coal Yard Traffic at a Power Plant, Midwest Research Institute
for Confidential Client, December 1985.
10. Critical Review of Open Source Particulate Emission Measurements - Part II - Field Comparison,
Southern Research Institute, Birmingham, AL, August, 1984.
11. Size Specific Particulate Emission Factors for Uncontrolled Industrial and Rural Roads, U. S.
Environmental Protection Agency, Research Triangle Park, NC, EPA Contract No. 68-02-3158,
Assignment 12, January 1983.
12. Iron and Steel Plant Open Source Fugitive Emission Control Evaluation, U. S. Environmental
Protection Agency, Research Triangle Park, NC, EPA Contract No. 68-02-3177, Assignment 4,
August 1983.
13. Extended Evaluation of Unpaved Road Dust Suppressants in the Iron and Steel Industry, U. S.
Environmental Protection Agency, Research Triangle Park, NC, EPA Contract No. 68-02-3177,
Assignment 14, October 1983.
14. Improved Emission Factors for Fugitive Dust from Western Surface Coal Mining Sources, U. S.
Environmental Protection Agency, Research Triangle Park, NC, EPA Contract No. 68-03-2924,
Assignment 1, July 1981.
4-79
-------
15. Fugitive Particulate Matter Emissions Final Draft Report, U. S. Environmental Protection Agency,
Research Triangle Park, NC, EPA ContrctNo. 68-D2-0159, Assignment 4-06. January 1997.
16. A. L. Williams, G. J. Stensland, and D. F. Gatz, Development of a PM10 Emission Factor from
UnpavedRoads, APCA Publication 88-7IB.4, Presented at the 81st Annual Meeting of APCA, Dallas,
TX, June 19-24, 1988.
17. A. L. Williams and G. J. Stensland, Uncertainties in Emission Factor Estimates of Dust from
Unpaved Roads, AWMA Publication 89-24.6, Presented at the 82nd Annual Meeting & Exhibition,
Anaheim, CA, June 25-30, 1989.
18. W. R. Barnard, G. J. Stensland, and D. F. Gatz, Development of Emission Factors for Area and Line
Source Fugitive Dust Emissions: Implications of the New PM-10 Regulations, Transactions of PM-
10 Implementation of Standards, ed. by C. V. Mathai and D. H. Stonefield, APCA publication TR-13,
p 326, 1988.
19. Telephone communication between G. Stensland, Illinois State Water Survey, Champaign, Illinois, and
G. Muleski, Midwest Research Institute, Kansas City, Missouri, July 11, 1997.
4-80
-------
5. PROPOSED AP-42 SECTION
Summaries of comments on the proposed AP-42, Section 13.2.2 Unpaved Roads, and responses to these
comments are presented on the following pages. The final AP-42 section is available as a seperate file.
5-1
-------
IDAHO Department of Environmental Quality
Memorandum dated December 31. 1997 from Val Bohdan of Idaho DEO to Ron Myers
of EPA (attached)
3) The backup studies cited in this draft appear to have no representation from the
cement/ concrete industry—a significant number of which exist in the State of Idaho.
By contrast, much data backup originated as studies of the coal and steel
industries—none existing in Idaho. This raises the question of eventual
appropriateness of the proposed emissions formulas in terms of country regional fit.
4) The issue of control efficiency factoring which can be afforded by vehicular speed
reduction is very confusing and needs to be resolved more clearly. Logically, speed
reduction, especially on the lower range, needs to be incorporated as an inducement for
emissions reduction.
5) We would like to suggest the use of clear statements in the writing of the whole
Section 13.2.2. Instead of "should be," the direction needs to be "do this" or "use this"
in order to give some assurance to the eventual user of this section.
6) Notwithstanding the derivation process for the formulas, it seems logical that the
formulas should be mathematically simplified for the eventual user. Use of negative
exponentials can intimidate those not acquainted with higher math and thus should be
avoided simply by placing the exponent as a positive number in the denominator.
Moreover, combining of numerical constants should be carried out as far as possible,
again to assist the eventual ease for the user.
7) Concerning the default moisture content value of 0.5%: For Southern Idaho, much
of which is considered "high desert" area, the 0.5% default value is probably correct.
We glean this value from the 1996-1997 "Pocatello Road Dust Study" of moisture
content which was performed, however, on an unpaved MgCl treated local road. The
lower end of this study indicated the moisture content value of 0.6%.
Specific Comments:
Comments pertaining to the Section 13.2.2 Draft
(a) Page 13.2.2-3: the first equation should be simplified to become
E = (k/7.03) (s)08 (W/3)b (1/MC) (1)
Notice that "a" factor has been replaced by 0.8. This is proper since the value for "a"
is the same for all the particulate sizes considered in this equation. Also notice that the
MRI agrees that wider representation of different industries would be extremely
beneficial in developing a truly generic equation applicable to all situations.
Nevertheless, one must use the data available to develop emission factors.
Unfortunately only limited data are available for cement/concrete industries. Please
see also the response to the Portland Cement Association. Additionally, text has been
added to the background document to more fully describe the approach taken here to
capture the essential features of the emission process with a few readily obtained
variables.
The linear reduction in emissions due to decrease in vehicle speed was not clearly
expressed in the draft AP-42 section. This will be corrected in the final version.
Suggested wording changes will be considered in conjunction with suggestions made
by other reviewers.
These suggestions need to be considered in conjunction with comments made by other
reviewing organizations.
Note that the default value of 0.2% will be incorporated at the "normalizing" factor for
moisture. This value was selected based upon the available moisture content data
available from uncontrolled publicly accessable unpaved roads. The 20th percentile
moisture content value was selected to represent a typical minimum value exclusve of
natural mitigation. See also response to Minnesota Pollution Control Agency
comment.
Wording/organization suggestions will be considered in conjunction with other
comments and suggestions received. To avoid the use of a negative exponent, one
could also write Equation 1 in the draft section as
E = (k/7.03) (s)08 (W/3)b / (M/l)c
(1)
-------
value of superscript "c" is positive. Thus, Table 13.2.2-2 supporting Equation 1 needs
also to be changed: eliminate row "a" (since the constant remains the same for all
particulate sizes), and change the minus sign (-) for "c" factor to a plus sign (or better
yet, no sign in front of it at all).
(b) Page 13.2.2-4: At the bottom of this page, the formula should be changed to read
S/15 not (15-S/15) if you intend the factor to drop linearly from 15 to zero vehicular
speed "S". It needs to be stated more clearly that the emissions factor remains constant
at speeds above 15 mph.
In Table 13.2.2-3, (Range of Source Conditions for Equation 1) on this page, we
recommend that column headings also contain the appropriate letter symbols (s, W, S,
and M—in that order) from Equation 1. This will aid all users, especially the
infrequent users.
(c) Page 13.2.2-5; The equation on this page should also be simplified to become
E = (k/7.03) (s)08 (W/3)b (l/M^) [(365-p)/365] (2)
The issues identified in Paragraph "a" (just above) also apply to this equation.
Moreover, your use of the term M/l appears overly simplistic and should be shortened
to just M.
Page 13.2.2-6; Insert the word "directly" at the sign of * in the third sentence from the
bottom, which reads: "Although vehicle speed does not appear * as a parameter, it is
obvious..."
(d)Page 13.2.2-8; The second paragraph (control efficiency afforded by speed
reduction) is very confusing and should be either clarified or deleted. The use of a
power factor for vehicular speed "S" is very misleading and counters earlier
statements. However, the power factor S3 2 may best represent the emissions factor
relationship for speeds below 15 MPH. If that is the case, then it should be so stated.
A simple graph may be the best way to explain and clarify this point.
Attachment to Review and Comments on the Draft AP-42 Section 13.2.2. Unpaved
Roads From the Idaho Division of Environmental Quality (IDEO)
The following are the comments/suggestions compiled by the Technical Services
Bureau, Air and Hazardous Waste Section, Idaho DEQ, in response to the invitation to
comment on the Draft AP-42 Section 13.2.2. The cover letter addresses the AP-42
draft section whereas the following comments are more broad-based and address the
background document and the overall methodology for the study.
In response to the Minnesota Pollution Control Agency comments and evaluation of
the moisture data from publicly accesable roads, the normalizing factor for moisture
will be changed to 0.2%. See also the comments from the North Carolina DNR.
As noted above, the expression in the draft section was in error and will be corrected.
MRI agrees that change might be useful to infrequent readers of AP-42 and the change
will be made in the final version.
The "1" serves to non-dimensionalize the M term. The normalization allows one to
more readily convert the emission factor expression from one set of units to another.
This includes units for both the dependent (i.e., emission factor) and the independent
variables. Please see the footnote '"c" in connection with Equation 4-1 of the
background document.
This change will be made in the final versions of both the background document and
the AP-42 section.
MRI agrees that the paragraph as presented can be confusing. The discussion about
speed being raised to a power between 1 and 2 refers to tests conducted of captive
traffic and will be removed and/or revised in order to improve clarity. At present,
however, there is no good technical basis for the use of a 3/2 power relationship.
-------
General Comments:
In making such sweeping changes to a set of equations which govern the emission
estimation process from a major source category for the next decade(s), more testing
and studies are warranted. The much touted ease of use is achieved by sacrificing the
fine dependencies afforded by specific governing parameters, such as number of wheels
and speed. The moisture term is a definite improvement but can be already enhanced
in its application and by reference from other studies already performed. It is strongly
recommended that this equation be implemented in a test-mode for one or two years
before finalizing it. This would allow more time to analyze and study the effects of
these proposed changes.
1.
What were the basic guidelines used to select studies used in the background
document? The IDEQ is aware of two other studies, performed in Idaho with
guidance from the Midwest Research Institute (MRI) that meets established
screening criteria, which could have been used as background information for
developing this emission factor. As those studies were conducted in Idaho, they
would have provided some regional representation, a more extensive database,
and made the factors more robust and applicable to regions like Idaho.
2.
The studies chosen have no representation from the cement/concrete industry.
Are the differences accounted by the silt content adequate to characterize
emission factor dependence on significant parameters? The cement/concrete
industry constitutes a significant number of sources in Idaho.
3.
The document seems to primarily focus on PM-10. Is there a similar study
planned for PM-2.5 to decipher the relationships between significant parameters
that contribute to fine particle emissions? This is especially relevant in light of
the fact that geologically derived material and agricultural impacts contribute to
regional contributions of fine particles from studies in the west. This is also an
issue of focus since the promulgation of the new PM-2.5 standards in mid-1997.
4.
There appears to be a preference to test unpaved roads in iron and steel industries
in the east and coal industry in the west. Are these thought to be major
contributors of emissions from this source category? Is there any test that was
reviewed from unpaved roads in agricultural rural areas? IDEQ feels that such
information is key to have in the database as most western states have
agriculturally-dependent areas from which emissions have to be quantified, as
accurately as possible, if any sort of control scenario is desired to be achieved.
MRI agrees that more tests ~ especially for the PM-2.5 size fraction ~ would be
extremely beneficial.
The two studies mentioned in the comment were directed to paved roads. The first was
a surface sampling program and no emission test data were collected. The second
study involved a yearlong road surface sample collection together with a one-time
paved road emission testing program (April 1997). Only one unpaved surface was
sampled and no unpaved road emission testing was performed. The background
document under review considered only test reports with unpaved road emission test
data.
Clearly, one must rely on the available historical emission test data in order to develop
the candidate emission factor expressions. MRI agrees that additional testing in many
industries and parts of the country would be very beneficial.
Once again, one must rely on the historical data available and most data referenced
PM-10. MRI agrees that more testing focused on PM-2.5 is very much needed to
improve the estimation methods for all fugitive emission sources.
The Arizona DEQ study considered three unpaved roads in rural portions of that state,
and two sites were within the immediate vicinity of active agricultural lands. As noted
above, although one must rely on the historical data base, collection of additional
emission test data from many different situations would be very beneficial in later
updates to this section of AP-42.
-------
5.
The IDEQ is aware of several studies to characterize emissions from paved and
unpaved roads by the Washington State University in Pullman from 1994 to 1997
using tracers (The Measurement of Roadway PM-10 Emission Rates using Tracer
Techniques, Washington State Department of Transportation, Technical Report #
WAR 397.1). This study had important findings related to road emissions
compared to relative humidity. There seems to be no mention of the same.
6.
The Columbia Plateau PM-10 study reports a number of wind erosion studies,
and techniques to address them. Specifically, the soil erosion factor, and the
surface roughness factor, are mentioned as key parameters for wind erosion.
Would this also not be a major factor in emissions from unpaved roads? (See
related comment beginning of next section).
7.
As there seem to be key omissions in the literature search conducted, to compile
the database for the study, IDEQ is skeptical as to the comprehensiveness and
soundness of the proposed equation to adequately provide an accurate emission
factor for every region in the country.
8.
IDEQ is also concerned that the use of this forum is to review and provide
comment is instituted at a stage later than at which key directional changes to the
study can be implemented. What procedures are followed at each phase of the
study to ensure participation and encourage input from state and local agencies,
to make the study more robust and applicable to all regions? This process would
also foster confidence in the final product.
Specific Comments:
Chapter 2. Background Document:
1.
Is it not intuitive that over time, over a given surface area, that the suspendable
particulate loading would decrease (by advection, carry-out, etc.), provided new
material is not significantly added to the road surface (relates to erosion factors)?
Is there, then, any decay factor, or parameter (added or planned) to be added to
the equation as a correction for this effect? The effect of not having this
correction would be an assumption that constant surface loading is available for
re-suspension over an infinite amount of time resulting in gross overestimates-as
compared to realistic measurements.
How is the effect of relative humidity in the friction layer of the planetary
boundary layer on characteristics of suspended particles accounted for? Although
This report was issued to the Washington State DOT in March 1996 and the federal
DOT forwarded it to EPA in November 1996. As a result, the test report was not
available when the AP-42 update project began during the late summer of 1996.
(It should be noted that there is no discernible trend in Table 4 of the WSU study
between the 3 sets of paired unpaved road emission factors and relative humidity.)
The emission factor developed and recommended for inclusion in AP-42 deals with
traffic-generated PM, which is an ongoing emission source for active roads rather than
the occasional wind erosion of the surface. Because traffic causes emissions even in
the absence of wind, it is not intuitive that the parameters presented in the comment
are applicable to the emission factor under consideration. See also the response to first
comment under "Specific Comments," below.
Please see responses to comments 1 and 5.
No, it does not seem intuitive that an actively used unpaved road will lose its dust
emitting potential over time. Instead, the surface is continually ground by passing
vehicles. Although there is only a limited amount of data available, emission tests
conducted on the same uncontrolled road (References 8 and 13 in the background
document) from one year to the next do not provide evidence of diminished emission
potential (due to traffic over the roadway).
The draft section does not include a direct treatment of relative humidity (RH). During
the 1980s, attention was directed to use of a relative humidity term in road predictive
emission factor equations. In one version of the unpaved model, RH was raised to a
-------
there may be no measurable precipitation on the ground surface, high relative
humidity associated with high pressure events and associated interventions may
result in decreased circulation events in the surface friction layer closest to the
ground and cause suppression of dust, as in a fog with some precipitable water
content.
Chapter 3. Background Document:
•
In the last paragraph of page 3-7 the comments suggest, that tests from various
sources have been combined to derive the new equation. This approach suggests
that a large amount of testing was conducted to come up with gross average. As
explained elsewhere in the document, a mathematical fit needs not always imply
a reality fit. A log-normal distribution conveniently encompasses a wide range.
This approach is good as screening criteria but not for further refined purposes as
is applied from the AP-42 for permitting, PSD, and SIP purposes. For refined
purposes, an industry-by-industry equation should be considered. Although the
final equation may or may not differ much, the approach makes the study more
robust and increases user confidence as the database would be broad. At the very
least, a comparative study should be undertaken to establish the applicability and
usefulness of industry specific equations.
Chapter 4. Background Document:
It is interesting to note that tests continue to be accepted as approved even as the
emission factor values spread over 2-3 orders of magnitude without further
investigation as to this extensive spread. The final calculations of emissions and
the discretion, as to which order of magnitude to choose, is left to the field
operator or engineer in the absence of any further supporting documentation on
application of such ranges of values. In a practical regulatory sense this scenario
leaves emissions from certain categories in"grey areas."[underline added; see
response]
Please correct the table columns in Table 4-8.
The comment on page 4-20 that Equation 2-1 performed as well in estimating
emissions as did factors for specific sources in the coal industry could also mean
that the specific industry factors were somehow biased. It does not necessarily
mean the general Equation 2-1 is adequate and correct. It seems a fundamentally
gross over-generalization to then lump all the tests, in all studies reviewed, to
positive power of about 4, whereas in another version the same organization found RH
dependence at a power of -0.2. Furthermore, the WSU unpaved road results are
inconclusive with respect to the relationship between unpaved road emissions and RH.
Recall that the emission factor presented in the draft section references dry conditions.
Clearly, misty conditions should result in lower observed emissions; nevertheless, there
are insufficient data to determine the mathematical relationship. EPA has drafted
additional guidance to better account for the effects of precipitation within the AP-42
section.
As noted above, the development of an emission factor makes use of the data available.
Under the ideal situation, one could have sufficient information to develop industry-
specific factors for use in different regions of the country. MRI would welcome the
opportunity to work with a broader data base that spans many more industries;
however, these types of tests simply are not available. To the best of our knowledge,
the only industry-specific unpaved road emission factors recommended in AP-42
pertain to western surface coal mining. As a result of this update of the unpaved road
section, the emission factors in the western surface coal mining AP-42 section for haul
trucks and for light duty trucks are being replaced with the equations developed during
this effort. As part of Section 234 of the 1990 CAAA, a thorough comparison of the
generic (i.e., Chapter 13) unpaved road expression with the industry-specific equations
was undertaken. The background document summarizes the findings that, when
applied to independent data (i.e., not used in the development of the models), the
generic expression performed as well or better than the industry-specific factors.
The intent of the comment is unclear to MRI. The 2 to 3 orders of magnitude spread
in the overall data base is directly attributable to the wide range of underlying source
conditions (e.g., vehicle weight, road surface texture, etc.). Should the comment refer
to individual test reports, that type of spread might result if one were to compare
controlled and uncontrolled test results. (In particular, the intent of the last 2
sentences (underlined) is especially unclear to MRI.) On the other hand, when one
considers roads under comparable source conditions, there is considerably less spread.
Wind speed will be placed under the "meteorology" heading instead of "vehicle."
Please see response to Chapter 3 comment (above).
-------
come up with one large data set for the emission factor development. Is this the
only specific industry factor test that provided the impetus to lump all the test
data?
It is not clear whether reference 12 was used in the final equation development as
it did not have moisture content or PM-10 factors listed. What is the exact
meaning of "data was used in the expanded data analysis, they were not included
in equation development"?
If as mentioned in page 4-26 the effect of speed could not be isolated due to
unavailability of speed segregated data ... (s)uch data should probably be
obtained to study the effects of speed on emission factors. This leads to the
conclusion that if a model does not simulate reality to some extent then, perhaps,
the fundamental assumptions that went into creating the model are flawed, and
are unable to be verified. It could lead to serious errors if the equation is used in
this manner. The speed correction factor seems like an extreme ad hoc measure
to solve this problem.
Different size fractions may have different influences and effects, as related to the
determined significant parameters, in that multiplication of PM-10 emission
factors by appropriate size fraction would only be applicable as a rule-of-thumb
calculation.
It is interesting to note that a high measure of reliability is established using
equation 4-5, as established by Table 4-32 without inclusion of speed in the
equation! It is also particularly worrisome that the emissions increase with
decreasing speed, [underline added; see response] This table also demonstrates
the effect of high humidity (misty conditions) on the suppression of emissions.
The attached graph demonstrates the effect that speed multiplier will have on the
emission factor. The emission increases linearly with decreasing speed from 15
mph to 0 mph, and also causes an anomaly of having emissions from a stationary
vehicle with a 'B' rating! The text implies the need for an inverse effect. So, the
multiplier has to be inversed, as mentioned in the cover letter.
What is the rationale for using 12, 3, and 1 as the norms' for silt content, mean
vehicle weight, and moisture content, respectively?
MRI agrees that, as written, the background document is confusing on this point. That
portion of the document will be rewritten to clearly explain that although Reference 12
was not used in development of the final emission factor equation, its data were used in
those analysis that did not directly reference moisture content as a potential correction
parameter (as in the second full paragraph on page 4-24 of the background document).
MRI agrees that collection of additional test data can only strengthen the validity of
estimation methods. Nevertheless, as pointed out several times, one is forced to work
with the data sets that are available.
MRI agrees that there can be substantially different mechanisms involved in the
reentrainment of particle sizes other than PM10. The reduced quality ratings for
"scaled" emission factor equations reflect that concern. (See also responses to
comments made by the Minnesota and North Carolina state agencies.)
Because all tests in Table 4-32 were conducted with a travel speed of 30 mph, it is
unclear what is meant by the underlined portion of the comment.
As noted elsewhere, this term will be corrected in the final version.
The reference silt and weight values are the same as those used to normalize the old
unpaved road factor. The moisture content of 1% was selected because it corresponds
approximately to the geometric mean value for uncontrolled tests in the data set.
However, MRI expects to revise the final equation with a normalizing value of 0.2%
which is the same as the default value. This change should help ensure that water
addition is not "double counted." (See also the response to a Minnesota Pollution
Control Agency comment.)
-------
Chapter 5. Proposed AP-42 Section:
It is possible for the end-user of the equation to obtain daily precipitation totals
and relative humidity readings from the National Weather Service (NWS), Local
Climatological Data (LCDs). It should be made feasible to incorporate short-
term relative humidity and precipitation data into daily or hourly estimates for
emissions. Annual data can then be very accurately totaled from this equation.
This approach is preferred to the national precipitation data map provided.
The number of samples in determining silt content values in the table should be
at least 10 or more to provide an adequate level of confidence in the data.
As mentioned elsewhere, EPA has indicated its plans to include additional discussion
in the final version of the AP-42 section on how to incorporate more finely resolved
precipitation data in emission estimates for public roads. Two methods are provided to
accomodate local climatological information. One method provides a very simplistic
but directionally correct method that has been used for many years to accomodate long
term differences in the average moisture content of the road surface material. Another
method accomodates more variables that are believed to result in changes in the road
surface moisture content. This additional method requires hourly data on the quantity
of precipitation, humidity and snow cover as well as monthly data on the evaporation
potential (Class A pan evaporation and average traffic volume).
MRI agrees that more confidence should be placed on values based on more samples,
but believes that it is important to provide the sparse industry-specific information that
is available. State agencies should encourage site specific collection and analysis of
road surface material to better characterize the silt and moisture content of roads. If
state agencies have more surface material data, they are encouraged to forward that
information to EPA for inclusion in Table.13.2.2-1
-------
MINNESOTA POLLUTION CONTROL AGENCY
Letter dated October 29. 1997 from Michael J. Sandusky of Minnesota Pollution
Control Agency to Ronald E. Myers of EPA, (attached)
Table 1 summarizes the findings of the MPCA staff in a thorough review on statistical
analysis of the emission data provided by the EPA. (see table in attached comments
for footnotes, etc.)
Table 1. Empirical Constants from Statistical Analysis of Uncontrolled Particulate Emission Factors
PM-2.5 PM-10 PM-15 PM-30
Constant Draft
MPCA
Draft
MPCA
Draft
MPCA
Draft
MPCA
k, lb/VMT 0.24
3.57
1.6
1.72
2.4
3.41
5.3
6.08
a 0.8
0.67
0.8
0.77
0.8
0.72
0.8
0.97
b 0.4
0.24
0.4
0.43
0.4
0.29
0.5
0.52
c -0.3
-0.55
-0.3
-0.24
-0.3
-0.06
-0.4
-0.45
Cases ?
77
180
141
?
77
92
65
R-squared ?
0.125
0.345
0.384
?
0.255
?
0.512
Adj. R-sq ?
0.089
?
0.371
?
0.224
?
0.488
Q. Rating B
?
A
?
B
?
A
?
Regression ?
Forced
Stepwise
Stepwise
?
Forced
Stepwise
Stepwise
The fitting constants' quality ratings, the potential dual role of road surface moisture
content, the annual adjustment for precipitation, and the disappearance of vehicle
speed are major concerns to the MPCA. We believe, however, that the PM10 emission
factor equation (lb/VMT) with the fitting constants is acceptable from the statistical
standpoint.
Quality Rating Scheme
Emission Factor Documentation for AP-42 Section 13.2.2 (Draft Report) describes in
Section 3.3 emission data and emission factor quality rating scheme used for unpaved
roads source category. It states, "(t)he uncontrolled emission factor quality rating
scheme used for this source category represents a refinement of the rating system
developed by EPA for AP-42 emission factor. The scheme entails the rating of test
data quality followed by the rating of the emission factor(s) developed from the test
data...."
The quality control and quality assurance efforts in the development of emission
factors for this source category are important. However, we believe that the final
quality rating, as seen in Table 1 for PM-10, should also be more related to the
goodness of fit of the regression model. In plain words, we think the ratings of A and
B in Table 1, should be lower, e.g., C and D.
To further explain our concern with factor ratings, let's look at another rating and the
The MPCA re-evaluated the different emission factors presented in Equation 1 and
Table 13.2.1 of the draft AP-42 section. Several items should be noted:
1. The expressions for PM-30 do not agree because MPCA regressed only the 65
uncontrolled emission tests whereas the expression recommended for inclusion in
AP-42 is based on both the 65 uncontrolled as well as the 27 watered tests. Note,
however, that the two expressions in MPCA's Table 1 are essentially identical in terms
of the "fitting constants." Thus, had only uncontrolled tests be considered in the
development, the resulting PM-30 expression would not be substantively different from
the recommended equation.
2. Similarly, the MPCA's PM-10 expression also is based on some subset of the total
data sets used by MRI. Although it could not be confirmed from the information
presented, it appears that the MPCA expression is again based on the uncontrolled test
data. As was the case with the PM-30 factors, the MPCA's results indicate that no
substantive difference in the form of the PM-10 would be expected if MRI had
considered only uncontrolled tests in the AP-42 update.
3. The differences between the MPCA and MRI expressions for PM-2.5 and PM-15
stem the fact that MPCA developed their expression from a regression analysis while
the background document describes how the draft versions were scaled against the
PM-10 expression. Page 4-28 of the background document discusses MRI's stepwise
regressions of PM-15 and -2.5 data and the decision to scale emission factors against
the result for PM-10.
MRI agrees that the quality ratings should be dropped one letter when the emission
factor is applied to a specific test road. The background document and draft AP-42
section will be revised to reflect this decision. (See also the response to comments
from the North Carolina DNR.) The revisions will also discuss how the overall
performance of the emission factor improves when it is applied to a number of roads
within a specific area. This is an important distinction between fugitive dust sources
and the type of combustion emission source mentioned in the comment. That is to say,
-------
assumptions we make about it. An emission factor rating of A is given to the S02
emission factor for No. 6 oil fired, normal firing utility boilers in the current AP-42
Table 1.3-1. People in the regulatory and regulated communities are very confident in
using such an emission factor. Now, when an emission factor rating of A is given to
the uncontrolled PM-10 emission fitting constants, it has some profound implications.
First, it implies that the predicted uncontrolled PM-=10 emission for unpaved roads
from the regression model is the best (true), however, it also implies it is directly
comparable to that of the S02 emission factor for No. 6 oil fired, normal firing utility
boilers in the current AP-42 Table 1.3-1 (not true). People using these factors, who
tend to take a number out of a table without carefully reading the context, will assume
these factors are of equally high quality. Second, when people realize that less than 40
percent of the total variance in the emission data is explained by the regression model
(see PM-10 column in Table 1) and rating A still is given to the regression model, they
are going to seriously doubt the reliability of all the emission factors from AP-
42—stack emissions and fugitive emissions.
We believe that people can be satisfied with the notion that, because of inherent
variability, fugitive emission factors can never achieve the same level of quality rating.
Therefore, we would urge you to lower the factor ratings associated with the proposed
AP-42 for unpaved roads.
Road Surface Material Moisture Content
The efficiency of water application to control particulate emissions is not analyzed
statistically in this study, although equation (3) is presented in the Draft AP-42 for
estimating control efficiency for water applications. Input parameters for this equation
include water application parameters and pan evaporation rate, all of which to a great
extent determine road surface material moisture content.
There is a potential for double-counting the road surface material moisture content and
watering control efficiency. If road surface material moisture content resulted from a
control technology application, the road surface material moisture content before the
application should be used to establish the regression equation with fitting constants
shown in Table 1. We would like confirmation from the EPA that this was done
correctly.
The inclusion of road surface material moisture content makes sense in reflecting the
reality, if data collection to establish the equation in Table 1 was done correctly.
However, users of the equation still may double count the moisture contribution by
using post-application moisture value in the equation to predict uncontrolled emissions
and adding control efficiency due to water application to get the "controlled" fugitive
emissions. Of course, we realize that each regulatory agency just needs to guard
against dual use of moisture.
Table 2 presents moisture content data associated with PM-10 emissions, uncontrolled,
watered, and the combined data set. There is a significant overlap between the
a facility being inventoried typically contains no more than a handful of the stack-type
source mentioned. Furthermore, the stack sources are far better defined and steady in
terms of operating conditions. On the other hand, a facility may contain dozens of
unpaved travel surfaces, each with very different vehicle characteristics that change
with hour of the day, seasonally, etc. In that case, the performance of an emission
factor in accurately predicting emissions from a single source is not necessarily the
central issue. Instead, one is interested in how well the factor performs in estimating
the total (or average) emission from the entire set of sources over time periods of
interest. It should be noted that for many sources of particulate matter, the
performance of AP-42 emission factors applied to individual source is not significantly
different than the predictive capability of the unpaved road equation. The emission
factor ratings are more a function of the number of emission tests supporting the
emission factor than on the inherent variability of the emissions from the source being
characterized.
MRI agrees. Please also see the response to the first "General Observation" made in
the North Carolina comments
EPA has drafted additional guidance to better account for the effects of precipitation
within the AP-42 section. This material ~ which provides a means of using the hourly
precipitation values that are readily available ~ will be included in the final versions of
both the background document and the AP-42 section.
As noted in response to one of the Idaho DEQ's comments, the normalizing factor for
moisture will be changed to 0.2% (i.e., the default value) and the definition of Mdry in
Equation 2 will be expanded to ensure that this references uncontrolled conditions.
The moisture contents for the 137 "uncontrolled" tests in the development data set all
reference dry conditions (i.e., without any artificial watering or rainfall for a minimum
of 24 hours). For the "watered" tests, the moisture content reported represents a time-
averaged value of moisture during the test period. Thus, the appropriate value to
substitute (for inventorying purposes) in Equation 1 would be the average moisture
content during the watering cycle. If Equation 2 were used, then the appropriate value
for Mdry would be the uncontrolled moisture content.
Note that Table 2 in the MPCA comments averages over a variety of different
industries, road surface types, etc. More meaningful comparisons would result by
-------
uncontrolled data and the watered data, suggesting the difficulty in preventing dual
usage of moisture from happening.
Table 2. Road Surface Material Moisture Content for PM10 Emission Data
Description Uncontrolled Watered
Combined Data Set
Number of valid observations
145
37
182
Missing observations
27
4
31
Mean
1.611
4.751
2.249
Standard deviation
2.049
4.099
2.879
Skewness
1.786
2.17
2.621
Range
8.5
19.8
20.1
Minimum
0
0.3
0
Maximum
8.5
20.1
20.1
Annual Adjustments for Precipitation
Section 2.4 of the Draft Report (page 2-4) indicates the control efficiency of watering
depends upon (a) the application rate of the water, (b) the time between applications,
(c) traffic volume during the period, and (d) the meteorological conditions during the
period. This suggests the annual simplifying assumption (365-p)/365, which reflects
only first term, is an over simplification on the effects of natural precipitation, which is
equation (2) in the draft AP-42 Section 13.2.2.
In our experience with mining operations, 0.01 inches of precipitation in a 24-hour
period cannot achieve 100 percent control of particulate emissions from unpaved roads.
A multi-tier approach would be better such as minimal control for 0.01 inches,
moderate control for 0.10 inches, near-maximum control for 0.50 inches, and
maximum control for 1.00 inches or more. This could be done by developing four
maps similar to Figure 13.2.2-1 using current monthly climatological data such as that
in the enclosed Climatological Data, Minnesota, February 1997.
Vehicle Speed
Section 4.3 of the Draft Report (page 4-27) states, "it is obvious to any one who has
driven on an unpaved road that vehicle speed affects emissions, with faster vehicles
generating more dust than slower ones. For this reason, it was decided to incorporate
the findings of the captive traffic studies into the AP-42, independent of the emission
factor equation." Unfortunately, the corresponding section of the draft AP-42 Section
13.2.2 (page 13.2.2-8) is unclear on how this should be calculated.
The MPCA staff did confirm the apparent difficulty with vehicle speed in our
statistical analysis of the data file, unpaved.dat (July 31, 1997). We are unable at this
point of time to propose any better way of dealing with this variable in a statistically
acceptable manner. As for the emission factor adjustment for vehicle speed reduction
matching uncontrolled and watered tests from in Tables 4-5 through 4-28 in the
background document. However, MRI shares MPCA's concern that the effect of
moisture might be "double-counted " and, as noted in a previous response, will expand
the discussion of Equation 2 in Section 13.2.2 to ensure that Mdry clearly references
uncontrolled conditions.
As noted elsewhere in the comment log, EPA plans to incorporate additional guidance
in the use of more finely time resolved precipitation data.
Material drafted by EPA includes use of both current hourly rainfall totals as well as
antecedent precipitation.
This portion of the AP-42 will be revised to more clearly define the linear decrease in
emissions with a decrease in travel speed. As noted elsewhere in the comment
response log, the linear reduction in emissions was mistakenly expressed in the draft
AP-42 section and will be corrected in the final version.
The 50 mph value was used solely for illustration purposes. The numerical example
will be expanded to more fully describe the estimation process.
-------
in the draft AP-42 Section 13.2.2, we strongly suggest that some examples be provided
to clarify how this adjustment should be calculated for regulatory purposes. The text
on page 13.2.2-8 alludes to a 30 percent reduction in emissions for a vehicle speed
reduction from 50 mph to 35 mph; however, it is unclear why 50 mph is the
appropriate reference vehicle speed when (1) the proposed emission factor equation
lacks any reference vehicle speed, and (2) the SYSTAT regressions indicate vehicle
speed adds little to the Revalues.
-------
NATIONAL STONE ASSOCIATION
Techncial Comments Concerning Sections 4.2.3 and 4.2.5 of the Report Entitled.
"Emission Factor Documentation for AP-42. Section 13.2.2 Unpaved Roads (DraftV
(attached)
3. COMMENTS CONCERNING SECTION 4.2.3
3.1 Adequacy of the Testing Methodology
The first sentence of paragraph 2 of Section 4.2.3 makes an implied statement that the
methodology was not adequate.
"The study used an upwind-downwind profiling technique that varied from the
more commonly used exposure profiling method. "
A similar statement was included in the fourth paragraph of Section 4.2.3. This
statement goes on to declare that a large rock well created unrepresentative testing
conditions.
"At the Garner test location, a large rock wall that stood immediately behind the
downwind sampling site may have interrupted natural wind flows and/or created
a local recirculation event. The potential wind obstruction and the variation in
methodology from common exposure profiling methods accounted for a "B "
rating of the test data at the Garner quarry. The Lemon Springs test was
assigned an "A " rating. "
It is apparent that MRI has assigned a "B" rating to this test report due to the presence
of the "large rock wall" and due to the testing methodology. NSA objects to these
statements and to the "B" rating.
The clearly expressed intent of the NSA sponsored studies was to evaluate fugitive
particulate emissions from quarry haul roads. A major fraction of a quarry haul road at
stone crushing plants is in the quarry pit that varies in depth from 50 feet to more than
300 feet.
One of the testing locations selected for this test program was a portion of the haul
road at the Garner, NC quarry of Martin Marietta. As shown in the photographs
included with the test report, this location was approximately 100 feet below the top of
the quarry and next to a "large rock wall." The Garner site is highly representative of
quarry haul roads in the stone crushing industry. The other test location selected for
this test program was at the top of the Lemon Springs, NC quarry of Martin Marietta.
This site is representative of the portion of the quarry haul road outside of the quarry
pit. NSA believes that the selection of these two sites was technically correct and
justifiable.
As a basis for this response, recall that emission source testing requires one to first
isolate and then quantify the PM contribution from the source. This is spelled out
more completely in the following responses.
Issues of pit trapping notwithstanding, the source testing procedure chosen by NSA
and its contractors would require them, at a minimum, to
a) determine what portion of the downwind particulate is due to the source and
what is due to "background"
b) ensure that the source contribution is not sampled more than once
c) demonstrate that the entire plume is accounted for in a calculation scheme to
determine net mass passing through the measurement plane
d) relate the net mass passage to some meaningful measure of source activity to
obtain an emission factor
The tested road may indeed be representative of roads at stone crushing plants, but the
test site must allow one to isolate the source contribution in order to characterize
emissions. These are separate issues.
-------
There is, in fact, air recirculation due to the close proximity of the face of the quarry
wall to the downwind side of the quarry haul road. This is the natural wind flow
condition that exists in a deep quarry pit, and it must be taken into account during
emission factor testing. This recirculation condition makes the emission profiling
technique referred to by MRI difficult to apply for the following reasons.
• The haul road and its "shoulder" are not sufficiently wide for the fifteen meter
upwind and five meter downwind spacing of the monitoring instruments.
• The downwind particulate matter concentration does not necessarily approach
ambient levels at the 21 foot elevation. Accordingly, there is no clear limit to the
concentration profile integration.
Due to the proper selection of the test sites at the Garner and Lemon Springs quarries,
the emission factor data are highly representative of stone crushing plant haul roads.
The "B" rating is entirely inappropriate for the Garner tests. Exclusive use of the
"commonly used emission profiling technique" outside of the quarry, where there was
sufficient room for the monitoring towers would have clearly been non-representative
of quarry pit haul roads.
3.2 Adherence to the Test Program Protocol
NSA and its contractor, Air Control Techniques, P.C., fully adhered to the test
protocol. The first version of this protocol was submitted by NSA to EPA on May 8,
1995. Based on EPA comments, the protocol was revised and resubmitted by NSA on
July 20, 1995. Both of these versions included the following statement.
"Due to the short distances between the downwind side of the haul road and the
edge of the quarry cliff, the ambient PM-10 monitors may be influenced by PM-
10 emissions from the quarry itself or PM-10 particles formed due to the turbulent
eddies that exist at the edge of the cliff."
This comment was included in a section of the protocol explaining why the "commonly
used emission profiling technique " was not applicable. NSA believes that this
statement also clearly indicates our intent to test in the quarry pit itself, not just on the
upper portion of the quarry haul road. During an extended negotiation in the three
month period prior to the beginning of these tests in late August 1995, EPA personnel,
at no time, indicated that the proposed test location in the quarry pit or the testing
methodology described in the July 20, 1995 version of the protocol was inadequate.
The tests were conducted under the belief that EPA personnel had every opportunity to
review the testing approach and that all EPA concerns had been fully satisfied.
Accordingly, NSA is surprised that MRI has taken the position on behalf of EPA that
Given the recirculation, any number of things can occur that prevent one from
isolating and quantifying the source contribution. For example, the upwind samplers
may be impacted by the plume, resulting in too high a background concentration being
subtracted out and biasing the calculated emissions low. On the other hand, if the
plume circulates in the general vicinity of the samplers, the downwind samplers may
repeatedly collect PM from the same vehicle pass, thus biasing the results high. The
best one could hope for would be that the recirculation equally impacts both the
upwind and downwind samplers to the same extent. Even in that case, however, it is
problematic as to how one would attribute the net mass to a suitable measure of source
activity if the PM from one vehicle pass is sampled repeatedly.
At the upper boundary of the plume, the concentration should approach not necessarily
an "ambient" level, but the background concentration. Also, if there is "no clear
limit," then substantial plume mass would pass over the top sampler. The calculation
scheme based on a fixed height (of 28.5 ft) may or may not account for the additional
emissions. (See also the comment below on meaning of "ambient.")
The "representativeness" is based on grade, physical setting and other
geometrical/location criteria . Nevertheless, for testing purposes, the basic issue of
source isolation must be addressed independently.
How are PM-10 particles formed due to the eddies? In the quoted section, does
"ambient" refer to background samplers? If so, how would particles formed downwind
(i.e., due to eddies) influence the background sampler? Does the protocol address how
to deal with these influences?
MRI functions as an independent contractor and certainly does not purport to speak
directly for EPA. MRI's comments on the test method and the sites chosen are based
on a review of the test report and results presented therein. (Note that the test report
does not include the protocol in the list of references and ~ to the best of MRI's
knowledge ~ the protocol is not mentioned in the test report.) MRI neither received
nor was ever asked to review a copy of the test protocol. Had we reviewed the protocol,
at a minimum, questions would have arisen about effects mentioned in the quotation.
-------
the Garner tests should be rated "B" due to the test location and the test methodology.
NSA have done everything in our power to work in a fully cooperative manner with
EPA. Furthermore, we have conducted these tests in complete adherence to the test
protocols. The rating of "B" for the Garner test is completely inappropriate.
3.3 Water Application Rates
The second sentence of paragraph 3 of Section 4.3.2 of the MRI report states the
following:
"Specific water application rates were not reported, although the watering is
said to have occurred approximately every 2.5 to 3 hours. "
Appendix D of the emission test report for Garner and Lemon Springs (pages 100
through 124) specifically lists the exact time that every haul truck, water truck, pickup
truck, tractor, car, and van passed the sampling assembly. This MRI comment seems
to imply that Air Control Techniques omitted an important variable and was careless
in test documentation. This is not correct.
NSA and Air Control Techniques, P.C. have fully reviewed the May 8, 1995 and July
20, 1995 test protocols submitted to EPA prior to the tests. It is clear in these protocols
that we did not intend to record the water application rates. Furthermore, it was not
our intent to analyze the data in any manner that might involve EPA's wet
suppression efficiency equation. To our knowledge, this is the only equation that uses
the water application rates as an independent variable. Accordingly, we are surprised
that MRI has taken the position that we failed to include these data. This MRI
criticism is even more surprising considering that MRI and EPA have not included
water application rate data in the revised haul road equations. If the water application
rate data had been present, it is clear that it would have been ignored by MRI and EPA.
This MRI criticism is clearly unnecessary.
NSA would like to emphasize that we adhered fully to the revised test protocol that we
submitted to EPA more than a month before the tests began. At no time during the
pretest negotiations did EPA personnel request these data. NSA requests that MRI's
criticism regarding the water application rate data be removed from their document.
4. COMMENTS CONCERING SECTION 4.2.5
4.1 The Use of Colocated Push-Pull Hoods
Paragraph five of Section 4.2.5 states the following:
"The 'push-pull' method used for this study is not considered an accepted
methodology for measuring open source particulate emissions."
Paragraph 4 of Section 4.2.5 states the following:
The term "rate" is used to refer not only to the time between watering but also to the
amount (volume) of water applied per unit area. The statement that rates were not
reported is simply a remark based on the completeness of the report.
Had MRI reviewed the protocol, another item that would have been raised is
measurement of "rates" (in both the time and volume senses).
MRI will revise the background document to clearly state that the volume of water
applied per unit road area was not reported.
-------
"The low sampling height at relatively low wind conditions used for this test
program potentially allows the particulate plume to pass over the sampling device
without capture."
After reviewing the Entropy emission test report (Reference 5), NSA and Air Control
Techniques, P.C. believe that the emission factor calculation procedures have not been
clearly described, and we understand how MRI could have misinterpreted these results.
Actually, the "push-pull" method described in the Entropy emission test report is a
straight-forward adaptation of the upwind-downwind concentration monitoring often
used for measurement of fugitive dust emissions. Entropy did not calculate the
emissions based solely on the quantity of air captured by the hoods. It was also not
necessary for the hoods to capture 100% of the haul road emissions in order to
facilitate an accurate measurement of the downwind concentration. It is clear from the
sample emission factor calculation shown on page 12 of the Entropy report that the
average wind velocity (not the hood capture velocity) through the entire testing zone
was used to calculate the emission factor. Accordingly, this test used a conventional
upwind/downwind concentration measurement technique.
Entropy used the hoods simply to gather a sufficient gas stream sample to measure the
downwind concentration. As shown in Figure 2-3 of the Entropy report, the hoods
were located approximately 1 meter from the side of the haul road. This is
considerably closer than the 5 meter position used in MRI tests. Accordingly, there is
considerably less vertical dispersion from the point of dust release next to the haul road
surface to the monitoring site in the Entropy tests as compared to MRI tests. Due to
the extremely close position of the Entropy hoods, a representative sample of the
downwind concentration was obtained, (underline added; see response)
NSA and Air Control Techniques, P.C. do not believe that significant quantities of dust
escaped over the top of the hoods. Almost all of the particulate matter is emitted close
to the road surface. This belief is consistent with the particulate matter emission
mechanism described in draft Section 13.2.2.1 of AP-42, "Particles are lifted and
dropped from the rolling wheels, and the road surface is exposed to strong air currents
in turbulent shear with the surface. " The hoods used at Knightdale extended up to ten
feet above the road surface, and smoke tracer tests confirmed that during truck passage,
the large majority of the emissions remained at less than the 10 foot elevation and were
sampled by the hoods. It should also be noted that hoods were located immediately
adjacent to a 60 foot cliff that was part of the quarry pit wall. The 60 foot cliff less
than 4 meters from the edge of the haul road also precluded the use of an emission
profiling tower located 5 meters from the haul road.
It should also be noted that the fans on the upwind side of the haul road were used to
enhance particle capture and reduced vertical dispersion of the plumes from the wakes
of the haul road trucks. These fans increased the average wind speed across the road
surface and drove the particulate toward the hoods.
As applied since at least 1972, the "conventional" upwind/downwind (UW/DW)
technique utilizes atmospheric dispersion models along with measured net
concentrations at a single height to back-calculate an emission rate. In conventional
UW/DW sampling, the plume is intended to pass over the sampling device and so
complete capture is not an issue. Terms such as "capturing the fugitive emissions"
and "capture hoods" are used throughout the Entropy report. On the other hand, we
could find no reference to any dispersion [diffusion] model that would be a core feature
of "conventional" UW/DW.
The calculation scheme described on pages 11 and 12 of the test report relies on the
"area of the sampling array." The scheme described here is very reminiscent of the
roof monitor and quasi-stack measurement approaches to fugitive emissions. There is
no resemblance to "conventional" UW/DW method.
The intent of the underlined portion of the comment and the meaning of
"representative sample" are unclear to MRI. If the intention is to demonstrate
reasonably complete capture, the comment immediately preceding this one is
contradictory by stating 100% capture was not necessary. On the other hand, if
vertical dispersion is "considerably less," how would that situation affect emission
factors back-calculated in a conventional UW/DW method?
Although "tests utilizing smoke" are mentioned on page 4 of the test report, there is no
discussion of how or what type of tests were conducted. Where was smoke released ~
at the height of the rolling wheels, top of truck, or at the surface? Was the smoke
mixed with the dust plume as the particles are dispersed in the wake of the vehicle?
What is meant by "immediately adjacent?" The cliff is not evident to MRI in Figure
2-3? What is the orientation of the 60 ft cliff with respect to the hoods? In any event,
does the presence of the cliff aid in the capture or is recirculation likely?
-------
4.2 Adherence to the Emission Test Protocol
The "push-pull" upwind-downwind concentration test procedure used at the
Knightdale quarry was first proposed in a series of meetings attended by EPA
personnel and NSA personnel in the fall of 1993. It was described in an emission
testing protocol dated December 3, 1993 and submitted to the Emission Measurement
Branch by NSA. EPA personnel did not raise any objections to this test procedure over
the ten month period preceding the test program. The only comments received was a
telephone call from Dr. Chatten Cowherd or MRI on the first day of testing. NSA and
Air Control Techniques believe that more than an adequate opportunity was provided
to EPA and MRI to review the test procedure and raise any issues necessary. It was
clearly unreasonable to delay the comments for over ten months and then raise issues
after the equipment was set-up and testing was underway. It is also unreasonable to
declare that the testing procedure is not an accepted methodology.
4.3 Co-located Hoods
Paragraph 4 of Section 4.2.5 states the following:
"The co-located hoods showed an order of magnitude difference between the left
and right hoods in the concentrations sampled in three out of seven tests. "
It is important to note that the side-by-side hoods were not used in a co-located
manner. The emissions data from the two sets of hoods were combined. This is
entirely different than the procedures used for co-located ambient monitors. The term
"co-located" was not used in the Entropy report.
The term "order of magnitude" means a factor of 10. A review of the left and right
hood concentrations at Knightdale indicates that MRI is exaggerating with respect to
these differences. The data shown in the table [below have been taken from Entropy
Table 3-3. One of the tests (Uncontrolled Run 4) was factor of seven different, and two
of the tests (Controlled Runs 1 and 2) were approximately a factor of five different.
Left Hood Right Hood
Concentration Concentration Difference
Test qrains/DSCF qrains/DSCF
Left/Right
Controlled Run 1 1.05E-04 2.06 E-05 5.1
Controlled Run 2 1.35 E-04 2.83 E-05 4.7
Controlled Run 3 2.99 E-04 1.85 E-04 1.6
Uncontrolled Run 1 5.94 E-04 2.83 E-04 2.1
Uncontrolled Run 2 1.29 E-03 1.37 E-03 0.94
Uncontrolled Run 3 2.18 E-03 2.53 E-03 0.86
Uncontrolled Run 4 7.38 E-04 5.18 E-03 0.14
NSA and Air Control Techniques, P.C. have reviewed the Entropy data and believe
that the difference is caused primarily by the location of the left hood relative to an
intersection of two haul roads and the quarry pit haul road near the test site. It was
sometimes necessary for haul road trucks to stop and idle while another vehicle passed
As mentioned earlier, MRI functions as an independent contractor. MRI's comments
on the test method and the sites chosen are based solely on review of the test report and
results presented therein. The phone call mentioned in the comment was placed at the
request of EPA, who asked MRI to provide a "courtesy" review of the overall approach
on short (i.e., <24 hr) notice. To the best of our knowledge, MRI never received a full
copy of the protocol. In any event, MRI was never asked to provide formal written
comments.
MRI used the term "co-located" to indicate that the two set of hoods were in very close
proximity and Entropy never employed the term in their report. The point being made
in the background document was that the test data indicate a non-uniformly emitting
source. The importance of a uniformly emitting source would be even more important
for a conventional upwind/downwind sampling approach because of the need to apply
a dispersion model to the source.
The data are taken from Entropy Tables 3-4 and 3-5 rather than 3-3. MRI used "order
of magnitude" in the sense of "how many places left of the decimal point." Admittedly,
this may be less than technically precise and more of "colloquial" use of the term. In
any case, factors of 5 to 7 are still surprising high and indicative of a non-uniformly
emitting source.
This emphasizes the importance of being able to isolate the source under consideration
from the influence of other nearby (upwind) PM sources. Would idling emissions be
collected by the upwind samplers? Were diesel emissions from the vehicles passing the
array sampled or did these emissions pass above the 10 ft high array at a distance of
-------
through the intersection. The stopping point for vehicles exiting the pit and
approaching the primary crushers was close to the left hood. Air Control Techniques,
P.C. believes that the high concentrations observed in the left hoods during the first
two runs were due to the capture of these idling emissions.
NSA and Air Control Techniques can not find any indications of the possible cause for
the difference in the Left and Right Hood during Uncontrolled Run 4. However, we do
not believe that Uncontrolled Run 4 should be treated as an outlier and discarded.
Also, it should be noted that more than a factor of seven variability was described in
many of the references used by MRI in developing the proposed unpaved road
equation. The following examples illustrate the extent of differences in these other
tests.
Variability of Particulate Emission Factor Data
(MRI Conducted Emission Factor Tests)
MRI
Reference Run #
Lbs/VMT DifferenceSilt, %
Moisture, %
2 BA-9
0.09
3.35
5.69
BA-3
1.32
14.6
3.04
7.41
4 BB-47
78.2
14.0
5.11
BB-46
8.14
9.6
12.7
4.88
8 AQ7-G
0.39
7
1.2
AQ6-C
2.43
6.23
12
1.4
All three studies were conducted by MRI, and all three sets of runs were conducted at
similar moisture and silt levels as indicated in the table above. MRI chose not to
discuss the factor of 6 to 14 variability in their test runs but was highly critical of the
factor of five to seven variability in the Entropy data. In fact, variability is a common
problem in the large majority of fugitive emission testing projects.
4.4 Recirculation Air Flow
The fourth paragraph of Section 4.3.5 states the following.
"Strong evidence of recirculation of emissions to the upwind sampler is provided
by the fact that the upwind concentrations increased by roughly an order of
magnitude from the controlled to the uncontrolled tests. "
There is no technical basis for the criticism. The upwind concentrations increased
"... roughly an order of magnitude... " because the upwind ambient air sampler had to
be located close to a portion of the unpaved quarry haul road (see Figure 1). During
the uncontrolled tests, this section of the road was not watered.
lm away from the road? Are there additional PM or source activity components not
included in the emission factors reported? If additional PM emissions were sampled
and not subtracted out as background, then one would expect (all other things being
equal) that the factors would be biased high. However, controlled runs 1 and 2 have
the two lowest factors reported of the 3 controlled tests considered at Knightdale.
MRI's original remark had nothing to do with the emission factors reported. Even so,
we cannot let this comment pass without noting that in NSA's table :
• Runs B A-9 and B A-3 should not be compared because, although both are tests of
scrapers in transit,
1. the two tests were conducted at different sites;
2. more importantly, one was a test of controlled emissions while the other
was a test of uncontrolled emissions.
• Runs AQ7-G and AQ6-C are not comparable because they were conducted on
surfaces treated with different chemical dust suppressants.
• Table 4-8 of the AP-42 background document contains a mistakenly converted
emission factor for run BB-47. In the original test report, the emission factors for
runs BB-46 and BB-47 are given as 3100 and 2304 g/VKT [11.0 and 8.1
lb/VMT], respectively. Entries in Table 4-8 in the background document will be
corrected. (The correct values were included in the developmental data base.)
In the interest of isolating the source contribution, why wasn't the upwind section
watered?
-------
Air Control Techniques has recalculated the uncontrolled emission factors by ignoring
the contribution of the upwind dust concentrations to the measured downwind
concentrations. By taking this approach, the data are biased to higher-than-true levels.
It is apparent that the revised emission factors (ignoring upwind dust concentrations)
are only slightly higher than the emission factors reported in the test report. The order
of magnitude increase in the ambient air concentrations upwind of the test location did
not have a significant impact on the reported uncontrolled emission factors as indicated
in the table below.
Recalculated Emission Factors Based on Zero Upwind Dust Concentration
Upwind Original PM10 Revised PM10 % Difference
Concentration Emission Factor Emission Factor in Emission
Factors,
Revised /Original
Uncontrolled 1
2.28
E-04
0.528
1.10
2.08
Uncontrolled 2
2.28
E-04
1.57
1.89
1.20
Uncontrolled 3
2.28
E-04
2.34
2.59
1.11
Uncontrolled 4
1.75E-04
4.70
5.01
1.07
Except for one of the four runs, ignoring the contribution of the upwind air
concentration entirely results in an increase of only 7% to 20% in the calculated
emission factor.
It is important to note that a quarry haul road has an entirely different configuration
than a public unpaved road and haul roads at iron and steel plants. The quarry haul
road inherently has a swirl pattern necessary to allow heavy duty trucks to descend
several hundred feet into the pit. Furthermore, there must be one or more approach
roads to allow the heavy duty trucks, graders, and water trucks to reach the swirling
quarry pit road. In most quarries, an ideal upwind ambient air monitoring site is hard
to find due to the complex road pattern in a compact industrial site. Air Control
Techniques believes that Entropy properly selected a monitoring site and accurately
measured the actual upwind dust concentration approaching the portion of the haul
road tested. There is no basis for the "... recirculation " criticism expressed by MRI.
4.5 Testing Was Discontinued During Certain Wind Conditions
The third sentence of the third paragraph of MRI Section 4.2.5 states the following.
"Testing was discontinued when speeds exceeded 3 miles per hour. "
This statement is a misinterpretation of the comments and data provided in the
Entropy report. As stated in the Entropy report: "Furthermore, the test was delayed if
winds in excess of 3 miles per hour shifted and came from the North or East. " As
indicated in Figure 1, the hoods were located directly west of the portion of quarry pit
What reason is there that the emission factors monotonically increased over the four
uncontrolled test runs? (There is only a 6% probability of this occurring by chance
alone.) How long had watering been suspended?
Note that the last column represents a ratio, rather than the percent difference shown
in the column heading.
Note that the revised factors again increase monotonically. Again, how long had the
watering been suspended?
As before, the issues of "representativeness" are based on geometry and physical setting
criteria. As mentioned throughout, isolation of the source contribution is critical to
successful source testing.
Testing under higher winds in the "proper" direction would help ensure more complete
capture by the hoods, while testing under low-speed winds or winds with very oblique
directions (up to 80 degrees off perpendicular, according to page 8 in the test report)
would encourage material to pass over/around the hoods. What is the reason that
-------
haul road tested. The testing was conducted whenever the winds were from the west or
northwest. Furthermore, testing was conducted during all low wind speed conditions
(<3 m) because the upwind side fans generated a west-to-east air flow of approximately
3 mph. Accordingly, the testing contributed during all conditions when the air was
flowing in the proper direction.
The testing was interrupted whenever there were strong winds that were not in the
proper direction. The testing was restarted when the winds shifted back to the
acceptable direction. Winds from the north or east that exceeded 3 mph would have
caused a bias to lower-than-true emissions because the hoods were not in a proper
downwind orientation during these time periods. The procedures used by Entropy
were correct. Furthermore, these procedures are entirely consistent with those used by
MRI in tests of unpaved roads, [underline added; see response]
testing would be delayed under the very conditions that enhance complete capture?
Also, what is the basis for the very broad acceptance criterion for wind direction?
Again, this allows testing under the conditions of very poor capture.
Where was the Weather Wizard unit deployed? What height was the monitoring unit?
The last sentence (underlined) in the comment is entirely mistaken. MRI's acceptance
criteria is not at all similar to that used in the Entropy study. Had criteria "consistent"
with MRI's ranges been used, the underlined question in the above response would not
have been asked.
-------
PORTLAND CEMENT ASSOCIATION
Letter of November 14. 1997 from Garth J. Hawkins to Ronald E. Myers. USEPA (attached)
The Portland Cement Association (PCA) has the following comments on the September 1997 draft
version of the following U.S. Environmental Protection Agency (EPA) report:
Emission Factor Documentation forAP-42, Section 13.2.2 Unpaved Roads (the "AP-42
IJnpaved Road Document"),
PCA appreciates the opportunity to review this document.
All portland cement manufacturing facilities require large amounts of limestone and other naturally
occurring materials such as slate, shale, etc. Because of this fact, each cement plant operates quarries
and crushing operations to provide these materials to the manufacturing facility, and therefore,
constructs and maintains unpaved haul roads for the transportation of these materials.
The quarries are developed so that the most efficient transportation as possible of raw materials from
the source to the cement plant can be accomplished. To move the volume of limestone and other
materials required by the manufacturing facility, only large dump trucks or similar vehicles are used,
and the trucks are operated at fairly consistent speeds from the quarry operation to the crushing and
screening machinery. Smaller vehicles, such as pickup trucks or cars, are a limited percentage of the
vehicles traveling the unpaved roads within the facility.
Due to the availability of limestone and similar materials, the unpaved roads at the quarry and
manufacturing facility are constantly constructed and maintained with the raw materials being
extracted. Overall, cement plants are very similar to limestone quarries that provide crushed stone to
the road-building and construction industries.
Although several studies of unpaved roads related to the stone industry are included in the AP-42
Unpaved Road Document, some very dissimilar industries are also included in the development of
the emission factor equations. Industries such as coal mining, copper smelting, and the iron and
steel industry may required different types of vehicles, have variations in the traffic patterns, and use
other materials in the construction of their unpaved haul roads. For example, multiple types of
aggregate may be used at the above industries due to the lack of the availability of road-building
materials.
The emission factor equations in the AP-42 Unpaved Road Document are also dependent on data
collected from unpaved roads used by pickup trucks and cars. The use of these vehicles results in
great variations in possible dust generation due to the differences in tires, vehicle speeds, and vehicle
aerodynamic effects.
Therefore, PCA requests that the EPA consider including the emission factor equations developed by
the National Stone Association (NSA) in the AP-42 Unpaved Road Document. PCA believes that the
NSA equations are more representative of the unpaved roads found at a cement facility. The
inclusion of the NSA equations will allow a cement manufacturing facility to select the equation that
best represents the possible emissions form the haul roads related to its operations. For your
reference, a copy of the cover page of the report summarizing the NSA findings is attached.
The National Stone Association (NSA) emission factors utilize the
mathematical form of a predictive equation developed from tests of very
large haul trucks at western surface coal mines. That is to say, the
factors that PCA requests be considered are in fact based on source
relationships that the PCA describe as "dissimilar" to portland cement
industry. (See also the discussion of the NSA equation in the responses
to Air Control Techniques, P.C. comments below.)
-------
AIR CONTROL TECNIQUES, P.C.
Letter of November 24. 1997 John Richards and Todd Brozell to Ron Myers of US
EPA (attached)
1. Applicability of the Draft Unpaved Road Equation to Stone Crushing Plants
We believe that the predictive equation developed based strictly on emission factor tests
at stone crushing plants is a better predictor of PM-10 and PM-2.5 emissions than the
general emission factor equation for all types of unpaved roads. This position is
consistent with the following statement included on page 3 of the Fifth Edition of AP-
42.
"If representative source-specific data cannot be obtained, emissions information
from .. .actual test data from similar equipment, is a better source of information
for permitting decisions than an AP-42 emission factor. When such information
is not available, use of emissions factors may be necessary as a last resort."
The predictive equations developed based on NSA sponsored tests at stone crushing
plants located at Knightdale, Garner, and Lemon Springs, NC are shown below as
Equation 1 and 2.
EpM-io = (s/3)°-8(M/2)"0-9 Equation 1
EpM-2.5 = 0.25(s/3)08(M/2)-09 Equation 2
Where:
EpM-io = PM-10 Emissions, Lb./VMT
Epm-2 5 = PM-2.5 Emissions, Lb./VMT
s = Silt content, %
M = Moisture content, %
The use of the precipitation factor from Section 13.2.2 can be used to adapt this
equation for predicting annual emissions. This results in Equations 3 and 4.
Epm-io = (s/3)08(M/2)"09[(365-p)/365] Equation 3
Epm-2.5 = 0.25(s/3)08(M/2)"09[(365-p)/365] Equation 4
We believe that these equations are more representative of the PM-10 and PM-2.5
emissions from stone crushing plant haul roads for the following reasons:
• All tests were conducted on quarry haul roads representative of the stone crushing
industry.
• One of the three tests was conducted in the quarry pit.
• The vehicle weights and speeds during the test program were representative of the
The quote from AP-42 applies to situations in which an emission test result is to
applied to a different source at the same facility.
Even though Equations 1 through 4 in the comment reference stone crushing plant
roads, several points should be noted about those factors and how they were developed.
Those points are raised in the following paragraphs.
Equation 1 is presented as Equation 16 in a May 1996 report prepared for the National
Stone Association (NSA) entitled "Review of the EPA Unpaved Road Equation and its
Applicability to Haul Roads at Stone Crushing Plants." Because that report contains
the recurring theme that the AP-42 unpaved road emission factor lacks a firm technical
basis for application to pit roads, the report presents no discussion of the technical
basis for the recommended Equation 1. The report only states that "it was necessary to
change the exponents concerning the moisture content and to adjust one of the
constants" in an equation developed for western surface coal mines. Just how that
change and adjustment were made is never discussed. A preliminary analysis of the 13
reported Knightdale, Garner and Lemon Springs tests (using the emission factors,
moisture and silt contents reported) clearly shows that neither simple nor multiple
linear least-squares regression was used. Just what is the technical basis for the
"modification?"
Other points to note about Equations 1 through 4 in the comment:
• Combining the Knightdale and the Garner/Lemon Springs data sets mixes two
types of data. The May 1996 report explains that that Garner/Lemon Springs
emission factors have "subtracted] out the combustible particulate and organic
particulate that were obviously not emitted from the road." (The test report,
however, describes a correction only for "combustion particles resulting from
diesel exhaust" and implies in the example calculation that organic material is
included.) In any event, the Knightdale factors did not undergo this correction
and, just as importantly, the corrections were not made in the upwind
concentration measurements. (Recall from the background document that this
-------
stone crashing industry.
• The silt and moisture contents of the road surfaces were representative of the stone
crashing industry.
• The surface characteristics of stone crashing plant haul roads are different from
other types of unpaved roads due to the frequent watering, the compaction caused
by the heavy duty tracks, and the high degree of road maintenance provided by
plant operators.
A comparison of Equation 1 with the measured PM-10 emission factors at the three
stone crashing plants is shown in Figure 1. [See figure in attached comments.] The
R2 correlation coefficient for this equation is approximately 59%. A comparison of the
measured PM-10 emission factors with the draft unpaved road equation is shown in
Figure 2. [See figure in attached comments.] The R2 correlation coefficient is 54%,
slightly lower than for NSA's Equation 1. This means that the NSA equation explains
the variability of the data slightly better than the EPA equation.
The EPA unpaved road equation appears to have a significant bias to higher-than-
observed PM-10 emissions for stone crashing plants having high haul road moisture
levels. This bias is indicated by the intercept of the linear regression line with the y-
axis at a value of approximately 2.0 lbs/VMT. We believe that this bias is due to the
fact that the material present in the silt and stone crashing plants is inherently more
wettable than the silt present on rural unpaved roads (e.g., clay), western surface coal
mines (e.g., coal dust and clay), and iron and steel plants (e.g., slag). Use of the new
omission leads to a systematic low bias in the emission factors.)
• It is unknown what, if any, other culling/clean-up of the data sets may have been
performed. For example, of the three Garner tests, one test has negative emission
factors reported for both PM-2.5 and PM-1 and another test has EPM.! > EPM_2 5 >
EpM-10-
• Despite questions about the origin of Equation 1, it does reference back to the May
1996 report to NSA. There is, however, no indication as to how Equation 2 came
to be. Presumably, it was scaled from Equation 1 using the PM-2.5/PM-10 data
from the tests conducted for NSA. Because the Entropy test program (reference 5
in the background document) reports only PM-10 factors, we assume that only the
six Garner and Lemon Springs tests were used to scale Equation 1 to PM-2.5.
However, one of those tests resulted in a negative PM-2.5 emission and another
implied a PM-2.5-to-PM-10 ratio of more than 100%. Assuming those test results
were not used, the remaining ratios (58% at Garner and 8.2%, 28%, and 76% at
Lemon Springs) do not yield the value of 0.25 implicit between Equations 1 and 2.
The figures are misleading in several ways. For example, the R2 value shown in Figure
2 pertains to the least-squares best fit line between a subset of the measured and
predicted emission factors. Also, because of the multiplicative form of both the AP-42
and NSA equations, the more appropriate plot (and correlation) for each figure would
be log-log in nature. The R2 shown is not the same as a multiple R-squared value for a
regression-based predictive equation of a multiplicative form. Even more importantly,
direct comparisons of R-squared values is misleading unless one also considers the
number of "degrees of freedom." In addition to the R-squared value, the number of
observations and the number of independent variables determine the "level of
significance" for a predictive model. Because it is unclear how the NSA factor was
derived, it was not possible to assign a meaningful level of significance for the NSA
expressions.
It also appears that values plotted in Figure 1 only -70% of what is directly calculated
using Equation 1. Consider, for example, the fourth uncontrolled test at Knightdale
(the far right-hand data point in Figures 1 and 2). From Table 3-6D in the Entropy
test report, the silt is 11.03% and the moisture content is 0.83%. In that case, Equation
1 leads to an estimated value over 6 lb/vmt which is 50% higher than the value shown
on Figure 1. What are the predicted values and what silt and moisture contents were
used?
Some bias results simply because the Garner and Lemon Springs tests have undergone
"correction for combustibles and organic material." In that case, a higher value from
the draft AP-42 equation (which includes exhaust and other components found
downwind of the roadway) is certainly to be expected. Also, recall that although
downwind samples were adjusted, no corresponding adjustment was made to the
upwind samples. That omission results in a systematic low bias in the resulting
emission factors.
-------
unpaved road equation may penalize the operators of stone crushing plants that are the
most conscientious in maintaining high moisture levels on their haul roads.
The emission factor data obtained in the NSA sponsored tests appear to be more
representative of PM-10 and PM-2.5 emissions from stone crushing industries. This is
indicated by the more reasonable form of the relationship shown between the predicted
and observed emission factor data shown in Figure 1.
2. General Comments
Road Surface Moisture Levels
We believe that the EPA draft equation in its present form underestimates the benefits
of moisture. Extrapolation of the curve defined by the equation to the 20% moisture
level yields predicted PM-10 emission factors in the range of 1.0 lbs/VMT as shown in
Figure 3. [See figure in attached comments.] Air Control Techniques, P.C. believe
that the new equation overpredicts PM-10 emissions at high moisture levels.
The curve generated by the equation should approach very low emission factor values
at 20 percent moisture levels. The particulate emissions from essentially all unpaved
road surfaces should be very low at this very high moisture level. The mathematical
form of the equation should be reviewed to determine if there is a more appropriate
exponent for moisture that provides a better representation of emissions from highly
moist unpaved road surfaces.
Despite the apparent deficiencies at high moisture levels, the equation appears to have
the proper form for low moisture levels. As indicated in Figure 3, the predicted
emissions have an asymptotic relationship with moisture at levels below approximately
0.3%. We have observed the same relationship in tests conducted for the National
Stone Association.
Precipitation Factor
We agree with the inclusion of the precipitation factor, [(365-p)/365] in Equation 2 of
Draft Section 13.2.2, and with the statement that, "... all roads are subject to some
natural mitigation because of rainfall and other precipitation." However, it would be
helpful to add a statement that the precipitation days should include all days that the
road surface is covered by snow or ice, irregardless of the amount of precipitation
occurring on each specific day.
Although one may argue about the form of and procedure used to develop the revised
AP-42 unpaved road equation, the background document describes how the predictive
model was developed. In this way, arguments and discussion can proceed with all
parties on equal footing. On the other hand, the procedures and data that result in
Equations 1 through 4 have not clearly been presented. Even ignoring issues of
negative emission factors and mixed types of data, it is still not possible to recreate the
results reported. In fact, simply calculating the values in Figure 1 using Equation 1
was unsuccessful. Given the undocumented procedure used to develop the predictive
models, unilateral claims about the reasonableness of the method are simply not
supported.
MRI agrees that 15% represents a reasonable estimate of surface moisture content
above which essentially no road dust is emitted. On the other hand, extensive watering
should have no effect on emissions due to exhaust or any material entrained from the
truck's load, undercarriage, etc.
Recall that the Garner and Lemon Springs data have had at least diesel exhaust
removed from the reported emission factors. Furthermore, the adjustment
systematically biased emission factors low by not correcting the upwind background
samples.
Emissions should increase as the surface moisture content decreases, but it is also
reasonable that each road has some "effective lower limit" for its surface moisture
content. In that case, the asymptotic behavior in Figure 3 would not be observed, but
instead emissions would follow a flatter portion. In other words, once a road is "dry,"
becoming "bone dry" would not greatly increase emission levels.
As mentioned elsewhere in the response log, EPA has drafted additional guidance to
better account for the effects of precipitation within the AP-42 section.
-------
Vehicle Speed and Other Factors
It is apparent in the Emission Factor Documentation report and in the draft Section
13.2.2 that the EPA and MRI authors are not entirely confident in the form of the new
unpaved road equation. For example, the following statement is included in Section
1.2.2.3.
"Although vehicle speed does not appear as a correction parameter, it is obvious
to anyone who has driven on an unpaved road that (visible) emissions increase
with vehicle speed."
Air Control Techniques, P.C. agrees with this comment regarding the importance of
the speed factor. Furthermore, we believe that there are a number of other important
factors that have a direct and significant impact on PM-10 and PM-2.5 emissions. A
partial list of these factors include the following.
• Vehicle road clearance and the associated magnitude of the turbulent wake as a
function of the vehicle speed
• The tire tread characteristics with respect to the tendency to pick-up and entrain
particles into the turbulent wake of the vehicle
• The tire tangential velocity with respect to the tendency to release particles from
the tire into the turbulent wake of the vehicle
• The actual pressure exerted by the vehicle tire on the road surface that causes
pulverization of silt particles to form PM-10 and PM-2.5 particles
• The grindability of the silt particles
• The extent of compaction of the road surface under various wet suppression and/or
natural precipitation conditions
• The extent to which tailpipe exhaust contributes to particle entrainment into the
turbulent wake of the vehicle
Obviously, neither EPA nor NSA has the budget necessary to accurately analyze the
possible impact of all of these important variables. Accordingly, Air Control
Techniques, P.C. recommends that EPA conduct a fundamental particle formation and
emission study using modern computational fluid dynamic modeling (CFD)
techniques. These are "First Principle" models that are being actively used in a wide
variety of aerospace design projects, automotive design projects, process equipment
design projects, and air pollution control equipment optimization projects. We have
had the opportunity to work on a number of projects involving CFD, and we are very
impressed with the capability and accuracy of this technology. CFD would provide an
economical way to provide a sound technical basis to the unpaved road equation. For
too long, this equation has been based simply on layer after layer of empirical studies
concerning only a few of the important variables affecting emissions. There is now a
readily available technology to provide improved emission factor equations.
The statement pertains to dust generated by individual vehicle passes over a road while
the recommended emission factor equation references emissions due "fleet average"
conditions over a road. MRI believes that the statement does not connote a lack of
confidence in the equation but rather implies that a) that every road probably has a
fairly narrow range of "natural" average speeds and b) there are insufficient test data
available to fully define the influence of average speed on emissions.
MRI agrees that there are many factors that can influence emission levels from vehicle
travel over unpaved surfaces. However, two points must be reiterated:
1. Many of the factors listed (and, for that matter, other potentially important
variables) are highly intercorrelated. For example, speed is inversely correlated
with weight; and tire tangential velocity, tread design, and footprint pressure are
all interrelated. In developing a phenomenological model from available
empirical data, inclusion of highly intercorrelated independent variables is
usually not appropriate.
2. Related to the previous item, it would be necessary to obtain emission data under
tightly controlled conditions to fully address factors of the type listed. In the case
of tread design, for example, one would ideally want at least duplicate tests of 2
or 3 different tread designs on the same trucks driven by the same operators at the
same speed over the same road. Even so, because tread design potentially affects
the "steady-state" road surface properties, one would also need to allow the road to
"condition" itself to each design over a period of at least several days or weeks.
Even assuming one could achieve extensive experimental control over a "real-
world" source, one would still need to contend with test conditions beyond
control, such as antecedent meteorology.
The comments regarding CFD are interesting, but it is not clear how such an approach
could be "operationalized." For example, one could use CFD to determine the near-
source air flow field for analysis of the trajectory of an individual particle released
tangentially from a tire. Similarly, one might simulate air flow due to the turbulent
wake that mixes entrained particles. However, the feasibility of CFD would depend on
the analyst's ability to specify initial and boundary conditions that would be used
relative to the entrainment of particles? From what point along the tire is the particle
released? How would that change with size of a particle? A much more thorough
prospectus of how CFD could be used is necessary before one can reach the conclusion
of the last sentence in the comment.
-------
NORTH CAROLINA DEPARMENT of Environment and natural Resources
Letter dated October 22. 1997 from Jim Southerland to Ronald E. Myers (with
attached marked-up copies of background document and draft AP-42 section)
Table 13.2.2-1 could use some additional clarity. For example, "yard area" should
clearly state that this is the storage area. "Haul" and "Access" should clearly indicate
that these are to the pit or wherever. Is "mean" in the header an arithmetic or
geometric variety? Can more definition be given to the road surface "dirt?" Again,
additional explanation of what the new information in the table are as opposed to old,
etc.. should be added to provide clarity to the user who might be familiar with using
the old tables in separate sections, [underline added; see response]
Page 3: The first paragraph does not seem to describe satisfactorily what was done.
Additional detail and clarity with a reference to the further discussions in the
background report might be helpful. Also on same page, I suggest writing out each
equation (PM-30, PM-10, PM-2.5) separately for clarity. Footnote meaning or
equivalence of PM-30, and drop PM-15 as it has little relevance/meaning. I do not
believe these resulting equations technically merit the "A" and "B" ratings and should
be downgraded at least a letter due to the statistics in the background report and
personal judgment.
Page 5: The discussion talks about defaults but stops short of a "presumptive default"
equation or expression for crude approximation. Since this is likely to be done
anyway, I suggest providing such an equation with calculated extremes that can occur
if applied without regard to real input data.
Page 8: The first full paragraph discusses collecting new road samples after 6 months
of use. I sincerely doubt that anyone will likely do this. It is difficult to even get a
facility to take samples at all to estimate emissions.
Page 10: The section does not explain "Class A pan evaporation," and it should.
Some other word changes recommended on enclosed copies.
Page 12: How does one determine "ground inventory?" is there a rule of thumb for
default?
Background Report
Page 1-1: The Second Edition of AP-42 was published in 1972. The earlier "Duprey"
edition was in 1968 or 1969. Earlier versions of similar documents were issued in
1965 or so. However, I don't believe fugitive dusts were addressed until the Third
Edition, or perhaps a supplement to the Second or Third Edition.
In general, the suggested wording changes will be incorporated into the draft section.
As pertains to the underlined portion of the comment, note that only the western
surface coal mining section would be affected ~ in fact, the appropriate change has
drafted and sent to the EPA work assignment manager. Also note that the current
version of Table 13.2.2-1 already includes road material information for surface
mining.
Reference to the background document will be added on page 13.2.2-3. Quality
ratings will be re-evaluated in conjunction with suggestions made by other reviewers
(most notably Minnesota Pollution Control Agency).
As mentioned elsewhere in the response log, the predictive equation will re-written
with a normalizing value of 0.2% for moisture and text will be added to clearly
indicate that 0.2% is the default value.
The recommendation concerns speed controls. Although it may be difficult to
convince a facility to collect any samples, this seems to be a reasonable request if a
facility claims control credit for speed reduction.
Text will be added to better explain Class A pan evaporation and its use in the
prospective analysis.
Additional text will be added the example in Table 13.2.2-4 to supplement the
explanation of ground inventory given in item 1 on page 13.2.2-12. Because Figure
13.2.2-2 is used either to estimate the effectiveness of an existing control application
plan or for planning a program to meet a certain efficiency level, it does not seem that
a default value is necessary.
The statement is based on page 1-1 of EPA-454/R-95-015, Procedures for Preparing
Emission Factor Documents. The paragraph will be rewritten to remove the date
reference.
-------
In the definitions section, "filterable particulate" should be included for completeness.
I would suggest dropping the IP or PM-15 as it is not now used and could be
confusing.
In Section 3, measurement methods are discussed. However, the "stone association"
method seems avoided somewhat. Since it has been used and the data evaluated, it
should be included in the descriptions. Here and in Section 4, the evaluations seem a
bit biased against data not collected by MRI. Their data may be better or not, but
"outside" tests seem more rigorously critiqued than the other tests. Comments may be
valid, but need to be equal and balanced in presentation so as to not give this
impression. For example, "unacceptable" is a judgment given without any
documentation or reasons. Also, it is not reasonable that road widths and such basic
information not included in test reports, even by the same contractor, are not
recoverable in some fashion.
Filterable particulate will be added. In addition, the material will be rewritten to
follow a "PM-x" format.
MRI will expand Section 3 to indicate that both the upwind-downwind and exposure
profiling methods doe not interfere with plume development/dispersion by forcing or
blocking the flow. Furthermore, as evidenced in the National Stone Association
comments, the Knightdale test report did not clearly establish how emission factors
were developed. However, MRI believes that the background document was lenient in
the assignment of quality ratings to the Garner and Lemon Springs test data. For
example, consider that
• It is unclear what run the example calculation on page 17 refers to. The example
states "Run Number G-UW-M201A-3 8/15/95." However, the end result of the
example is an emission factor that corresponds to the reported value for run "G-U-
AMB-2 / G-DW-M201A-2" in Table 3-12.
• The example calculation also based the emission factor on 204 vehicle passes, but
does not imply where that information is to be found. Table 3-7 gives 95 and 122
loaded truck passes during Runs G-UW-AMB-2 and G-UW-AMB-3, respectively.
(Apparently, the traffic counts given in Appendix D are used.)
• It remains unclear why, if one were to correct the downwind concentration for mineral
content, etc., one would not also make the same correction for the background
concentration. To not do so creates an "apples and oranges" situation, systematically
biasing the results to lower than actual emission levels of mineral and organic
particulate.
• Issues of upwind composition notwithstanding, there are also questions about how the
size distributions were used to correct for combustion particulate vs. stone dust. The
data used in the correction are based on microscopy, but no mention is made of
translating the number-based distribution to a mass-based distribution that would be
needed to made the correction.
• Surprisingly little discussion is offered for some unusual results reported. For example,
in the three tests conducted at the Garner site, there is a ratio of 300 between the highest
to lowest emission factors. Nothing is said about this. Assuming that the same types of
trucks traveled at roughly the same speed over the same road during the 3 tests would
lead one to the conclusion that the reproducibility of the measurements is not very good.
Also, no discussion is offered for findings of negative PM2.5/PM1 emission factors nor
of a PM1 emission factor being greater than a PM2.5 factor, which in turn is greater
than the PM10 factor.
In spite of the above, the Garner and Lemon Springs testing programs were still assigned B
and A ratings. Given the issues raised about recirculation and source isolation in response to
NSA's comments, it appears that the quality rating for Garner was even more lenient than
originally believed.
-------
Page 4-29 and thereabouts: Would it not make sense to view the data bases for PM-30,
PM-10 and PM-2.5 separately and independently? There may likely be forces (e.g.,
static) acting upon the different sized particles that would best be represented by this
treatment. With the statistics presented on page 4-30 and 31. the "A" rating on page
4-29 does not seem warranted, [underline added; see response]
Mid-page 4-37: "0.5 percent" seems to materialize out of the air. Explain "pan
evaporation" and its relevance on the next page.
General Observations
There continues to be a generally insufficient level of information and detail for
confidently estimating emissions from fugitive dust sources of all types. This includes
information which would assist in relating sources more closely with their ambient
impact. The parameters upon which the emissions should be based are fairly intuitive
and the existing equations seem to address those. However, there is a gap of
acceptance of these emissions as being part of the "real world" of sources which are
emitting into the ambient air and for which we are comfortable with emissions being
well correlated with their ambient impact. The complexity of resulting equations
generally precludes a majority of facilities from estimating their emissions in this
manner. The availability of a simple, stable, defensible and usable (user friendly)
computerized model to accomplish this would be of assistance, but perhaps be only a
partial solution. It might be helpful to develop several (based on aridity, soil
characteristics, etc.) models which could represent different parts of the country and
types of facilities and make the calculations simpler, although somewhat more crude.
Facilities and agencies are somewhat geared to permit conditions, so this might
provide a means to categorize further the estimation of emissions, application of
controls and operations.
Reading the section, I could not help but wonder if some future reviews and updates
should not address this problem a little differently. For example, would an approach to
separate the mechanical lifting forces and the air turbulent forces in the analysis be
productive? Also, for PM-10 and PM-2.5,1 doubt if it is still appropriate to look at just
silt analysis. I am sure silt is still a crude and somewhat commonly available indicator,
but the size particles being simulated are so much smaller than silt that one can not
help but wonder if there is not a finer delineation within "silt" that is necessary before
a determination of this sector can be appropriately made.
The background document notes that all the PM data sets were originally analyzed
separately and independently. The problem arose in that the resulting factor for PM-
2.5 was not consistent with the result for PM-10/PM-30 and had only limited
predictive accuracy. Also, note that the statistics for the underlined portion of the
comment deal only with hypothetical data and not with any emission factor developed
in the background document. Rather, these are only hypothetical data that serve
illustrate why a geometric mean is more appropriate for the ratio-based statistics.
Following a reevaluation of the public road data base the default surface moisture
content and thus the moisture normilization parameter was revised to 0.2%. The
background document will be modified to more fully explain why a value of 0.2% is
recommended. As noted above and elsewhere in the response log, the moisture
normalization in Equation 1 of Section 13.2.2 also will be changed to 0.2% with an
explanation that it is a default value.
MRI agrees that fugitive sources are indeed a unique class of emissions unto
themselves. In essence, this type of source is defined by what it is not (i.e., not directly
through a stack or vent). Nevertheless, fugitive sources are pervasive throughout
industry. Admittedly, in an ideal situation, one would have sufficient information to
develop industry-specific factors for use in different regions of the country. However,
one is always forced to work with the data that are available. Over the past 20 years,
emission estimation methods have relied on similarities in the basic emission process
over the broad range of source conditions throughout different industries.
Again, in an ideal situation, one would have access to data that clearly delineate
emissions from wakes, tire/road interactions, etc. Nevertheless, the practical
constraints on developing this type of information are overwhelming, as discussed in
response to an Air Control Techniques comment.
-------
This report on fugitives from unpaved roads does not sufficiently show the comparison
of old parameters and results with the newer ones. I recommend that each estimation
process, including those for aggregate operations, coal mines, paved roads, etc., be
examined in a case study comparison approach so the reader can view them side by
side and evaluate the impacts of the revisions. One is understandably reluctant to
adapt and apply a new set of numbers without having some concern about and
evaluation for what this will do to the existing data structure and integrity built up over
the previous years of application. A clear concise comparison detailed in the
background report and summarized in the sections themselves would facilitate this
level of confidence. A cross reference to any applicable (EIIP) estimation methods
would be helpful.
MRI agrees that such a side-by-side comparison would be useful. Nevertheless, MRI
believes that regular AP-42 users are in the best position to conduct such a study in
order to provide information most applicable to their particular situation.
------- |