United States
Environmental Protection
Agency
Industrial Environmental
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S7-84-077 Aug. 1984
v>ERA Project Summary
Paved Road Paniculate
Emissions -- Source Category
Report
Chatten Cowherd, Jr. and Phillip J. Englehart
This study entailed an extensive field
testing program to develop emission
factors for particulate emissions
generated by traffic entrainment of
paved road surface particulate matter.
The emission sampling procedure used
in this program provided emission
factors for the following particle size
ranges: £ 30, 15, 10, and 2.5 fan
aerodynamic diameter. Testing was
performed at sites in the Kansas City*
and St. Louis (MO) areas. These sites
represented significant urban paved
road emission sources in the following
land use categories: commercial/
industrial, commercial/residential.
expressway, and rural town.
The measured inhalable particulate
(IP--^ 15 fan aerodynamic diameter)
emission factors ranged from 0.06 to
8.8 g/VKT (vehicle km traveled).
Lowest emissions were measured for
the expressway category: highest
emissions were measured for the rural
town category- About 90% of the IP
emissions consisted of particles £ 10
fan in aerodynamic diameter, and about
50% of the IP emissions consisted of
particles <, 2.5 /urn in aerodynamic
diameter.
Correlation analysis of particulate
emissions with parameters characteriz-
ing the source conditions showed the
existence of a relatively strong positive
relationship between intensity of
emissions and roadway surface silt
loading. This relationship was used as
the basis for deriving predictive
emission factors for each particle size
range. The equation for IP emissions
was found to represent measured IP
more accurately over a much larger
range of values than does the AP-42*
single-valued factor.
To facilitate the use of these particle
size specific equations in developing
emission inventories, a classification
system of mean or typical silt loadings
as a function of roadway category was
derived. These mean silt loadings were
then inserted into the respective
emission factor equations to derive a
matrix of emission factors for specific
roadway categories and particle size
fractions.
This Project Summary was developed
by EPA's Industrial Environmental Re-
search laboratory. Research Triangle
Park, NC, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering infor-
mation at back).
Introduction
As early as 1976, receptor oriented air
quality assessments showed traffic-
entrained particulate from paved roads to
be a major cause of non-attainment of air
quality standards for total suspended
participates (TSP) in urban areas.
However, only a few field programs (all
completed by 1977) had tried to directly
measure dust emissions from urban
streets. Moreover, these programs were
seriously limited in the measurement of
aerodynamic particle size characteristics.
"In this Summary, failure to specify either Kansas or
Missouri after "Kansas City," implies both cities.
*U.S. EPA report AP-42, Compilation of Air Pollutant
Emission Factors. Third Edition (NTIS PB 275525)
July 1977.
-------
This study was aimed at developing
size-specific paniculate emission factors
for urban paved roads, based on
expanded field testing of representative
sources. The resulting emission factors
would provide for the development of
effective strategies for the attainment
and maintenance of the TSP standards,
as well as the anticipated standard for
particles £10 yum in aerodynamic
diameter.
The emission sampling procedure used
in this program provided emission factors
for the following particle size ranges:
TSP = Total suspended paniculate
matter (as measured by a
standard 'high-volumesamp-
ler) consisting of particles <.
30 yum in aerodynamic diame-
ter.
IP = Inhalable paniculate matter
consisting of particles £15
fjm in aerodynamic diameter.
PM-10 = Paniculate matter consisting
of panicles £10 >um in
aerodynamic diameter.
FP = Fine paniculate matter consist-
ing of particles £ 2.5 yum in
aerodynamic diameter.
Results are presented for winter testing
in the Kansas City, MO, area and spring
testing in areas of St. Louis, MO, and
Granite City, IL
Sampling Site Selector
Eight candidate sampling areas in
Kansas, Missouri, and Illinois were
identified by the EPA as likely
representative sites for the field study.
These areas represented a range of
typical road, traffic, geographical, and
environmental conditions within residen-
tial, commercial, and industrial land uses.
Each sampling area contained a TSP
monitoring site providing historical air
quality data.
Three major criteria were used to
determine the suitability of specific sites
within the designated areas, for sampling
of road dust emissions by the exposure
profiling technique:
1. Adequate space for sampling equip-
ment.
2. Sufficient traffic and/or surface
loading so that adequate mass
would be captured on the lightest
loaded collection substrate during
a reasonably short sampling time.
3. A wide range of acceptable wind
directions, taking into account (a)
the street orientation relative to
the predominant wind directions
for the locality, and (b) upwind
obstacles (houses, buildings, or
trees) to free wind flow.
Although roads with light traffic were not
considered, they probably do not
contribute substantially to total
emissions of traffic entrained dust in
urban areas.
Based on the above criteria, seven sites
were selected for this testing program:
Kansas City Area - three sites
7th Street in Kansas City, KS (commer-
cial/industrial)
Volker Boulevard/Rockhill Road in
Kansas City, MO (commercial/residen-
tial)
4th Street in Tonganoxie, KS (rural town)
St. Louis, MO - two sites
I-44 (expressway)
Kingshighway (commercial/residential)
Granite City, IL - two sites
Madison Street (commercial/residential)
Benton Road (commercial /residential)
Sampling Equipment
A variety of sampling equipment was
utilized in this study to measure
paniculate emissions, roadway surface
paniculate loadings, and traffic charac-
teristics.
'The basic emission sampling
equipment included an isokinetic expo-
sure profiling system with four sampling
heads positioned at 1 - to 4-m heights. In
addition, high-volume samplers, each
fitted with a size selective inlet (SSI) and a
parallel-slot cascade impactor (Cl), were
placed at 1-and 3-m heights to determine
the respective IP mass fractions of the
total paniculate emissions and the
corresponding particle size distributions.
The five-stage cascade impactors had, at
a flow rate of 40 scfm (1133 L/min), 50%
efficiency cutpoints at 7.2,3.0,1.5,0.95,
and 0.49 //m aerodynamic diameter. The
impactor substrates were greased to
reduce particle bounce. A standard high-
volume air sampler was operated at a
height of 2 m. Normally, these sampling
devices were positioned 5 m from the
downwind edge of the road.
The basic upwind equipment included
SSIs and a standard high-volume air
sampler. In the Kansas City testing, two
SSIs at heights of 2 and 4 m were used to
obtain the IP concentration of upwind
paniculate matter. In the St. Louis testing,
the primary upwind equipment included a
high-volume air sampler and an SSI/CI
with greased substrates.
Samples of the dust on the roadway
surface were collected during the source
tests. To collect this surface dust, it was
necessary to close each traffic lane for
about 15 min. Normally, an area that was
3 m by the width of a lane was sampled.
For each test, collection of material from
all travel lanes and curb areas (extending
to about 25 to 30 cm from the curbing)
was attempted. A hand-held portable
vacuum cleaner was used to collect the
roadway dust. The attached brush on the
collection inlet was used to abrade
surface-compacted dust and to remove
dust from the crevices of the road surface.
Vacuuming was preceded by broom
sweeping if large aggregate was present.
Characteristics of the vehicular traffic
during the source testing were deter-
mined both automatically and manually.
The characteristics included: (a) total
traffic count, (b) mean traffic speed, and
(c) vehicle mix.
Total vehicle count was determined
using pneumatic-tube counters. To
convert the axle counts to total vehicles,
vehicle mix was determined visually over
1-min intervals every 15 min during the
source testing. The vehicle mix
summaries recorded vehicle type,
number of vehicle axles, and number of
vehicle wheels. From this information,
the total axle counts were corrected to the
total number of vehicles by type.
The speed of the freely flowing traffic
was taken to be the posted speed limit of
the roadway test section. As a check,
speeds of the vehicles were determined
occasionally using a hand-held radar
gun. The weights of the vehicle types
were estimated by consulting automobile
literature and distributors of medium-
duty and semi-trailer trucks.
Sampling and Analysis
Procedures
The sampling and analysis procedures
used in this study were subject to quality
control (QC) guidelines which met or
exceeded the requirements specified by
EPA. As pan of the QC program for this
study, sampling and analysis procedures
were audited routinely, to demonstrate
that measurements were made within
acceptable control conditions for panicu-
late source sampling and to assess the j
source testing data for precision and
accuracy. Audit items included gravi-
-------
metric analysis, f lowrate calibration, data
processing, and emission factor
calculation.
Prior to equipment deployment, a
number of decisions were made as to the
potential for acceptable source testing
conditions. These decisions were based
on forecast information obtained from the
local U.S. Weather Service office. A
specific sampling location was identified
based on the anticipated wind direction.
Sampling would be initiated only if the
wind speed was forecast between 4 and
20 miles per hour (6 and 32 km/hr).
Sampling was not planned if there was a
high probability of measurable precipita-
tion (normally > 20%) or if the road
surface was damp.
Sampling usually lasted 4 to 6 hr.
Occasionally sampling was interrupted
because of unacceptable meteorological
conditions and then restarted when con-
ditions were suitable. The unacceptable
meteorological conditions most
frequently encountered consisted of light
winds (below 4 mph or 6 km/hr) and in-
sufficient angle (< 45 degrees) between
mean (15-min average) wind direction
and road direction.
The vertical distributions of exposure
(i.e., the product of plume concentration
and mean wind speed) were numerically
integrated to calculate emission factors.
The size selective inlet/cascade impactor
sampler combinations provided reliable
point concentrations for IP and finer
particle size fractions. Plume height was
determined by extrapolation of the
vertical profile of total paniculate concen-
tration as measured by the MRI exposure
profiler.
Test Results
Table 1 summarizes, by land use
category, the emission factor data and the
corresponding source characteristics. As
can be seen, the smallest emission
factors were measured in the freeway
category, which also had the lowest
surface silt loadings. The highest
emission factor was measured in the
rural town category which showed a cor-
respondingly high surface silt loading.
Multiple Regression Analysis
The source tests were evaluated
according to established QA criteria for
exposure profiling. Seven of the nine
Kansas City tests met all of the QA
criteria, while only three of the ten tests
conducted in the St. Louis/ Granite City
area met the QA criteria. The spring test-
ing in particular, was hampered by
unseasonably light winds. Wind speed for
four of the ten spring tests did not meet
the minimum wind speed criterion of 4
mph (6 km/hr).
Stepwise multiple linear regression
(MLR) was used to evaluate independent
variables for possible use as correction
factors in a predictive emission factor
equation. Because it was desirable to
have multiplicative rather than additive
correction factors in the emission factor
equations, all independent and depend-
ent variable data were transformed to
natural logarithms before being entered
in the MLR program.
The independent variables evaluated
initially as possible correction factors
were silt loading (g/m2), total loading
(g/m2), average vehicle speed (km/hr),
and average vehicle weight (Mg). The
rationale for including measures of
roadway paniculate loading stems from
findings of an earlier program that
indicated that the magnitude of roadway
emissions was directly related to
variations in surface loadings. The
vehicle parameters—mean weight and
speed—were included largely by analogy
to a predictive emission factor equation
for unpaved roads, although it was
recognized that the dust generation
mechanism for paved roads may differ
from that for unpaved roads. The
moisture content of the road surface
particulate was not included as a
correction parameter because of the
difficulty of collecting a sample without
altering its moisture content.
The resulting MLR equation after
normalization to a typical value for silt
loading was:
Table 1. Mean Emission Factors and Source Characterization Parameters
Land Use Category
No. of Emission Factor (g/VKT)
Tests <.15fim <,1Q fim ^.2.5 urn
Silt
Loading
Vehicle Vehicle
Speed Weight
(km/hr) (Mg)
Commercial/Industrial
2.43
2.07 1.31
0.29
48
4.1
Commercial/ Residential
Expressway
Rural Town
10
4
1
0.94
0.14
8.77
0.80
0.13
6.96
0.46
0.066
1.42
0.54
0.022
2.5
53
89
32
2.1
4.0
2.0
*IP= 254 <-blr)0-8 (1)
where:
elP = IP emission factor, g/VKT
sL = Silt loading of road surface par-
ticulate matter, g/m2.
This equation explained 73% of the
variation on the emission factors. The
MLR data set did contain data from all the
land use categories sampled during the
field program.
The emission factor equation was
found to predict the MLR series test data
with a precision factor of 2.0. The
precision factor (f) for an emission factor
is defined such that the 68% confidence
interval for a predicted value (P) extends
from P/f to Pf. The precision factor is
determined by exponentiating the
standard deviation of the differences
(standard error of the estimate) between
the natural logarithms of the predicted
and observed emission factors.
The precision factor may be interpreted
as a measure of "average" error in
predicting IP emissions from the
regression equation. Assuming that the
actual IP emission factors are normally
distributed about the regression line,
about 68% of the predictions are within a
factor of 2. The effective outer bounds of
predictability are determined by expo-
nentiating twice the standard error of the
estimate. The resultant estimate of
predictive accuracy, in this case 4.0, then
encompasses about 95% of the
predictions.
To put the precision factor of the IP pre-
dictive emission factor equation emission
factor into perspective, two comparisons
were undertaken utilizing the single-
valued emission factor found in EPA's
AP-42. However, before valid compari-
sons could be made, it was necessary to
convert the AP-42 single-valued factor
(which represents TSP emissions) to an
approximate IP emission factor. This was
done by multiplying the AP-42 value by
0.4, which is the mean ratio of net IP
(downwind minus upwind) to net TSP
concentrations as determined from the
data collected in this study.
The first comparison involved calcula-
ting a precision factor for the AP-42 data
set. The resulting value of 2.1 is a
measure of the ability of the single-
valued factor to represent the 40 pieces
-------
of data which were averaged originally to
produce the AP-42 factor. The second
comparison involved calculating a preci-
sion factor using the single-valued AP-42
factor to represent the MLR data set, as
collected in this study. This comparison
yielded a precision factor of 4.4.
The most important conclusion that
can be drawn from these comparisons is
that the emission factor equation, though
far from ideal, does predict IP emissions
more accurately over a much greater
range of values than does the AP-42
single-valued factor over a considerably
smaller range of data values correspond-
ing to the AP-42 data set. Furthermore,
applying the single-value AP-42 factor (to
represent the wide range of IP emissions
from paved roads as measured during this
program) yields a precision factor that is
more than double (4.4 vers'us 2.0) that
associated with the predictive equation.
This ability of the predictive equation to
more accurately represent variations in IP
emissions is directly attributable to the
relatively strong relationship between
roadway surface silt loading and IP
emissions.
Predictive emission factor equations
for the PM-10 and FP particle size
fractions were developed using the same
procedure as that applied in developing
the equation for IP. Derivation of TSP
emission factors for use in developing a
predictive equation required different
initial calculations, since only two TSP
samplers (one upwind, one downwind)
were operated during the measurement
phase of the program. In essence, the
initial calculation involved multiplying
the IP emission factor for each run in the
MLR data set by the corresponding net
ratio of TSP to IP concentration as
measured by appropriate samplers. This
procedure assumed that the TSP/IP ratio
was constant over the vertical extent of
the plume.
The general form of the emission factor
equations, applicable to the additional
particle size fractions, was the same as
Equation 1:
Table 2. Paved Road Emission Factor
Equation Parameters (by particle
size fraction)
e =
sL
0.5
(2)
The base emission factor coefficient (k),
exponent (P), and precision factor for
each size fraction are listed in Table 2.
Note that the tendency for the power
term in the equation to increase witn
larger particle size fraction is generally
consistent with the previous paved road
equation in which silt loading to the 1.0
power was employed to account for
variations in TSP emissions.
Particle Size
Fraction
TSP
IP
10 fjm
FP
k (g/VKT)
5.87
2.54
2.28
1.02
P
0.9
O.8
0.8
0.6
Precision
Factor3
2.4
2.0
2.2
2.2
Represents the interval encompassing 68%
of the predicted values.
Emissions Inventory
Applications
For most emissions inventory
applications involving urban paved roads,
silt loading will probably not actually be
measured. Therefore, to facilitate the use
of the previously described equations, it
was necessary to characterize silt
loadings according to a parameter(s)
more readily available to developers of
emissions inventories. After examination
and analysis of silt loading and traffic data
collected during relevant MRI sampling
programs, as well as surface loading data
gathered in connection with an extensive
study of urban water pollution, the
decision was made to characterize
variations in silt loading based on the
roadway classification system shown in
Table 3. This system generally
corresponds to the functional
classification systems employed by
transportation agency personnel; and
thus the data necessary for emissions
inventory—number of road miles per road
category and traffic counts— should be
easily obtainable.
Table 3. Paved Roadway Classification
Roadway Type
Average
Daily
Traffic No. of
(ADT) Lanes
Freeways/Expressways
Major streets/highways
Collector streets
Local streets
> 50,000 >4
> 10,000 > 4
500-10,000 2"
<500 2b
' Total roadway width 2 32 ft (9.75 m).
b Total roadway width < 32 ft (9.75 m).
The data base made up of 44 samples col-
lected and analyzed according to the
procedures outlined above, may be used
to characterize the silt loadings for each
roadway category. These samples,
obtained during field sampling programs
over the past 3 years, represent a broad
range of urban land use and roadway
conditions. Geometric means for this
data set are broken out by sampling
location (i.e., city) and roadway category
in Table 4.
Table 5 presents the emission factors
broken out by roadway category and
particle size. These were obtained by
inserting the above mean silt loadings
into the emission factor equations with
parameters defined in Table 2. These
emission factors can be utilized directly
for emission inventory purposes. It is
important to note that the current AP-42
paved road emission factors for TSP
agree quite well with those developed in
this study. For example, those cited in
connection with previous testing were
conducted at two roadway sites in the
major street and highway category.
Those tests yielded a mean TSP emission
factor of 4.3 g/VKT versus 4.4 g/VKT as
determined from the data presented here.
Summary and Conclusions
The purpose of this study was to quan-
tify size-specific particulate emissions
generated by traffic entrainment of paved
road surface particulate matter. Paved
road source testing was performed at
sites representing significant emission
sources within a broad range of urban
land-use categories.
The measured inhalable particulate
emission factors spanned two orders of
magnitude (0.06 to 8.8 g/VKT). Lowest
mean emissions were measured for the
expressway category; highest mean
emissions were measured for*the rural
town category. About 90% of the IP
emissions consisted of particles £ 10/um
in aerodynamic diameter, and about 50%
of the IP emission consisted of particles •£
2.5 fjm in aerodynamic diameter.
Correlation analysis of IP emissions
with parameters characterizing the
source conditions showed the existence
of a relatively strong positive relationship
between intensity of emissions and
roadway surface silt loading. This
confirms the findings of earlier testing.
Regression analysis of a subset of
acceptable (MLR) test runs was used to
derive a predictive IP emission factor
equation which explained 73% of the
variation in the emission factors.
This predictive equation has an associ-
ated precision factor of 2.0 in relation to
the MLR data set. By way of comparison,
the AP-42 single-value factor (corrected
to represent IP emissions) has a precision
factor of 2.1 for its data set and a preci-
sion factor of 4.4 for the MLR data set,
which spans a much larger range of
values than the AP-42 data set. There-
fore, the predictive equation, though far
-------
Table 4. Summary of Sift Loadings for Urban Paved Roadways (g/m2) '
Roadway Category
Local
Cityi
Baltimore
Buffalo
jranite City (IL)
Kansas City
St. Louis
Overall
1
1.
1.
-
1.
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
HI\I
------- |