United States Office of Air Quality EPA-450/4-79-012
Environmental Protection Planning and Standards June 1979
Agency Research Triangle Park NC 2771 1
An Empirical Approach
for Relating Annual TSP
Concentrations to
Participate Microinventory
Emissions Data
and Monitor Siting
Characteristics
-------
EPA-450/4-79-012
An Empirical Approach for Relating
Annual TSP Concentrations to Particulate
Microinventory Emissions Data and Monitor
Siting Characteristics
by
Thompson G. Pace
Air Management Technology Branch
Office of Air Quality Planning and Standards
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise, and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
June 1979
-------
This report is issued by the Environmental Protection Agency to report
technical data of interest to a limited number of readers. Copies are available
free of charge to Federal employees, current contractors and grantees,
and nonprofit organizations - in limited quantities - from the Library Services
Office (MD-35), U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina 27711.
Publication No. EPA-450/4-79-012
n
-------
TABLE OF CONTENTS
Page
LIST OF TABLES iv
LIST OF FIGURES V
INTRODUCTION 1
MICROINVENTORY METHODOLOGY 3
PRELIMINARY SURVEY 9
FINAL SURVEY AND COMPUTATIONS 9
DEVELOPMENT OF THE EMPIRICAL TECHNIQUE 13
FORM OF THE METHOD 13
EVALUATION OF METHOD 16
POTENTIAL APPLICATIONS AND EXAMPLES 24
Calculation of a City-Specific Regression
(Equation 2) 24
Calculation of Relative Source Impacts from
Equation 1 25
Control Strategy Applications 28
LIMITATIONS OF THE METHOD 28
References 31
Appendix A . A-l
Appendix B B-l
Appendix C C-l
Appendix D D-l
-------
LIST OF TABLES
Table No. Page
1 Example of Area Source Inventory Around
TSP Monitoring Sites 7
2 Information to be Obtained in Preliminary Survey 10
3 Land Use Categories and Classification Criteria 12
4 Summary of Multiple Regression Analysis 18
5 Comparison of Predicted and Observed
Concentrations in Three Urban Areas 22
6 Example Tabulation of the Predicted and Relative
Impacts of POINT, AREA, LOCAL and VISPLUME Source
Categories at Two Sites in Philadelphia 27
-------
LIST OF FIGURES
Figure No. Page
1 Standard sectors used in Microinventory
Analysis for Area Sources 5
2 Microinventory grid system recommended for use
with dispersion models 6
3 Sketch of the one-quarter mile radius around an
example site 8
4 Plot of predicted versus observed concentration
for regression 17
5 Plot of predicted versus observed concentration
in Philadelphia 21
-------
INTRODUCTION
A recent nationwide study of Total Suspended Particulate (TSP) indicated
that an average of around one-third of the TSP found on the high-volume sampler
comes from industrial types of sources. Another one-third of the TSP is
attributed to background or distant natural and man-made sources, while the
remaining one-third is due to nontraditional sources such as fugitive dust and
area source combustion products. This study suggested that these nontraditional
sources have their greatest impact in their immediate vicinity while industrial
process sources seem to have their greatest influence over a broader scale of
several miles because of their higher effective stack height. Several other
2 3
studies noted similar findings. ' In the past, emission inventories have
frequently been compiled for use in atmospheric dispersion models to assess
the impact of various source categories and control strategies for these
sources. Spatial resolution of these inventories has typically been limited
to 1 or 2 km square grids, even in the most densely populated or industrialized
areas. Also, many of the source categories are allocated to these grids by
population or other demographic indicators. Such allocations do not neces-
sarily reflect a true emissions profile of the area immediately surrounding
each monitor. One modeling effort apportioning vehicle related particulate
emissions from major paved roads into 1 km grids, greatly improving the model's
4
predictive capability. In another study, the effect of the particulate
emissions associated with a given roadway was shown to diminish rapidly as the
distance away from a roadway increased, indicating the importance of improving
our knowlege of sources within several hundred feet of a monitoring site.
Several studies have attempted to describe the area around monitoring sites in
detail using pictures, sketches, scaled maps, site descriptions and/or emission
inventories.1'2'3'6'7
-------
Procedures for developing and applying a microinventory are described
herein. A microinventory is an inventory and orderly compilation of sources
in the vicinity of a high volume sampler. Its purpose is to consolidate
information on source emissions or activity rates, monitor siting, air quality
data and land use characteristics into a consistent format and data base. The
purpose of the microinventory is to assist in identifying the sources of
particulate matter in close proximity to a monitoring site. Such information
can subsequently be used to interpret observed differences in air quality at
various monitoring sites. Emissions from point sources within five miles and
area source emissions within one mile are inventoried. Documentation of these
emissions becomes most precise in the area closest to the monitor. Monitor
siting characteristics such as height and distance from sources within 200 feet
are recorded as a part of the microinventory.
An empirical approach in which monitor siting characteristics and
microinventory information are related to measured annual mean TSP concentra-
tions using multiple regression analysis has been developed. The regression
uses variables which consider the emissions and distances to both point and
area sources and the effects of traffic and dust on nearby streets upon
measured TSP concentrations. The purpose of this report is to describe:
1) the methodology used in compilation of a microinventory and 2) the devel-
opment and uses of an empirical approach for relating observed annual TSP
levels to microinventory and monitor siting information.
-------
MICROINVENTORY METHODOLOGY
The term "microinventory" has been used to describe the procedure of
estimating annual particulate emissions in the area surrounding hi-volume
o
sampler sites. The methodology for the procedure has evolved from a fairly
simple genesis into a detailed compilation of sources and their distances and
9
directions from monitors.
A microinventory must be conducted over a geographical area large enough
to include the location of most of the sources believed or assumed to have
significant effect on TSP air quality. This geographic area will necessarily
be different for elevated point sources than the ground level or rooftop
emission height of area sources because of the greater dispersion potential
for emissions from the taller stacks characteristic of the major point sources,
The distances suggested for a microinventory are five miles radius for
point sources and one mile radius for area sources. The following rules of
thumb are suggested for point sources: a) all sources within one mile radius
be included, b) from one to three miles distance, only those sources greater
than or equal to 100 tons per year actual emissions be included and c) from
three to five miles distance, only those sources greater than or equal to
ZSO tons per year be included in the microinventory. These rules of thumb
were selected after review of Gaussian dispersion model applications and
theory to include the significant sources impacting air quality. Point
sources are further described in the microinventory as to their distance and
direction from the monitor. This information provides additional perspective
on the sources and their potential for influencing TSP concentrations.
-------
For area sources, the large number of small sources, diversity of source
shapes and broad land area coverage would make such a listing impractical if
not impossible. Considering the practical constraints, an inventory technique
has been developed which places the area sources into annular sectors in four
directions and three distances from the monitor as depicted in Figure 1.
(Figure 2 suggests a grid system for use with dispersion models which require
Cartesian coordinate systems). Within the center section (Sector 1), the
major sources within 200 feet of the monitor are described as to their dis-
tance and direction from the monitor. Sources within the quarter-mile radius
(Sector 1) are considered to be sufficiently close to the site that they would
affect the monitor under most wind conditions and thus, no compass directions
were recorded. Separate tabulations are made for area sources within each of
the nine sectors. An example of an area source inventory around a TSP site is
given in Table 1. Also, Figure 3 is a sketch of the quarter-mile area around
the monitor.
Fugitive dust sources are generally the dominant area sources. These
include dust from paved roads, unpaved roads and parking lots, cleared areas,
construction, rail yards, agricultural tilling, etc. Industrial processes
smaller than the point source cutoff of 25 tons per year actual emissions can
also be listed. Also included are area combustion sources such as space
heating, small boilers and incinerators which were not included in the point
source inventory, and vehicle exhaust. A gridded inventory using demographic
allocation methods is generally used for stationary combustion sources, and
these are apportioned into the nine sectors by land area.
-------
km
0 0.40 0.80
SCALE I \-
0 0.25 0.50
miles
1.6
1.0
Figure 1. Standard sectors used In Microinventory Analysis
for Area Sources
-------
9
9
8
8
9
5
5
4
4
5
1
1
4
8
6
2
1
1
3
2
2
3
3
7
6
6
7
7
(
SCALE
(
3 0.5
I
1
)
km
1.0
0.62
miles
2.(
j
1
1.2
Figure 2. Microlnventory grid system recommended
for use with dispersion models
-------
to
OI
-p
GO
en
£=
S-
o
c:
O
Q-
OO
O
o
c:
cu
c:
CD
O)
o
3
o
t/0
re
cu
O
CD
Q.
E
10
X
K
$
Cd
D
in
U
o;
D
O
in
r-
ro
W
-si
o
DQ
"H
>i
c
jj
V-i
0
4J
O
(9
U)
>•
ft
in
C
o
-H
tn
tn
•H
s
rH
ro
4J
O
H
cr>
00
r-.
vr>
m
^
ro
1
4-»
•rH O
> 4J
•H ro
-U ^
U
<
>,
IH
VD [rH
CN
in
o
m
o
in
O
in
O
rH
O
rH
O
i
O i
tjl
C.I
jj
'0
u
Source
z
o
M
H
COMBUS
1-1
O
rH
0
CN
•
0
a
o
a
t
cn
«3"
cn
rH
3
in
T3
C
Comm/I
t/i
M
O
4-J
13
l-l
Incine
vj
•H
Rail/A
ro
^3"
vD
in
ro
rH
CN
•
ro
r.
vo
r-H
O
(-J
CN
VD
O
CN
O
VO
O
vO
O
cn
•
rH
EH
^
kO
o
rH
X
•^T
•
r~
en
u
i'i
3
rj
x:
X
cu
0
4J
3
<
CO
w
in
10
u
C)
o
ct
A
INDUS
vO
ro
rH
0
rH
VD
^
O
rH
rH
IT,
CU
U
M
3
0
to
T
in
Q)
in
to
0)
U
O
M
ft
•o
C
M
l_
*.
E--
in
•J
D
s?
FUGITI
-.'
kO
r- In
r^.
0
in
o
o
ro
CN
1-1
o
CN
^
•
o
V
o
ro
00
r~
in
T!
r4
ro
>,
T)
*
Railro
j~\
ro
CO
O
^
rH
TT
rH
VD
in
cn
ro
CN
in
rH,
vD
ro
•q-
rH
VD
ro
vD
ro
vD
.
O
rH
fn
^
>
3
O
rH
X
in
.
M1
cn
tn
jj
01
cu
u
ij
r/i
•D
CD
>
(O
Q.
j
CN
i— 1
CN
.
f— I
H
^C
>
•>tf
O
X
r-
CM
tn
T.
M
O
M
•c
Unpave
tr
D
c
u
ro
rH
.
o
in
i\3
a>
M
i«
-a
Cleare
c
c
A.I
u
3
Constr
1
|
1
1
01
)u
3
4J
rH
Agricu
"J
ro
CN
t
ro
U
m
vD
rH
cn
rd
CU
n
rO
a
Storag
CN
•*r
CN
•^r
r~
o
CN
rH
oo
•
••31
U
ro
•^r
m
CN
10
j_>
O
rH
IT>
,v;
ft
t)
Unpave
£
4J
JH
O
U
U
0)
tn
>,
^HlO
n3
ij
rj
m
r~
£i
, i
.
|
VO
in
fN
Si
in
c
o
•H
(/)
tn
•rH
s
f^
ft
frl
nj
u
cu
Pi
^
00
vD
0
^r
vo
CN
z
o
t~H
CH
t/)
1
O
u
VD
ro
rH
vO
CN
O
O
rH
o
rH
10
DJ
10
CN
•^
in
CN
r-t
Ifl
4J
O
H
-------
1,7. Unpaved pkg lot, 0.1 ac, 10 cars
2. Cleared area, 0.1 ac
3,6. Unpaved pkg lot, 0.1 ac, 5 cars
4. Unpaved storage area, 0.2 ac
5. Unpaved pkg lot, 0.3 ac, 25 cars
8. Unpaved pkg lot, 1.7 ac, 35 trucks
9. Unpaved pkg lot, 1.5 ac, 150 cars
10. Unpaved storage area, 2.0 ac
11. Unpaved storage area, 8.1 ac
One-quarter mile radius around
Scale 1" = 400'
12. Unpaved pkg lot, 0.8 ac, 20 cars
13. Unpaved pkg lot, 0.7 ac, 50 cars
14. Unpaved storage area, 3.4 ac
15. Unpaved storage area, 2.3 ac
16. Unpaved road, 400 ft, 5 cars
17. Unpaved road, 300 ft, 10 cars
18,19. Unpaved alley, 750 ft, 5 cars
20. Unpaved road, 675 ft, 5 cars
21. Unpaved road, 600 ft, 5 cars
6402 East 37th site.
Figure 3. Sketch of the one-quarter mile radius around an example site.
-------
The emission factors used in the inventory should be consistent with
published EPA factors. The factors or formulas used to estimate fugitive
dust emissions in a recent microinventory of sites in Kansas City, along with
a summary of these calculations is given in Appendix A. These emission factors
were applied to activity factors such as VMT (vehicle miles traveled), acres
of tilled land, etc., to compute the emission estimates for each sector. A
description of the procedure recommended for use in deriving the information
needed to compile a microinventory follows. The procedure consists of two
steps: 1) a preliminary survey, and 2) a final survey and computations.
PRELIMINARY SURVEY
The procedure for determining the locations of the sources and their
activity levels should be carefully developed to minimize time required in the
field. A preliminary survey is suggested if the sites to be inventoried are
not located in the same area as the agency or person performing the inventory.
This is a one-to two-day visit to the sites to be inventoried. Table 2 indi-
cates the information obtained during the preliminary survey.
FINAL SURVEY AND COMPUTATIONS
Following this preliminary survey, a grid of the road network within
Sector 1 is drawn to scale using a map such as the U.S. Geological Survey 7
1/2 minute quadrangle. Then the distances and directions from the sites to
the point sources are calculated. Aerial photos can be used as one means of
locating cleared areas or other fugitive dust sources, or alternatively, a
visual drive through can be made. The final survey visit consists of a
detailed visual inspection of the vicinity of the site. The condition of the
-------
TABLE 2. Information to be Obtained in Preliminary Survey
U.S. Geological Survey Maps of the Area
(from a local engineering supply store in most areas)
Land Use Maps
(from local Council of Governments - COG)
Traffic Data
(from local COG, or city or State transportation agency)
TSP Annual Air Quality Data and Site UTM Coordinates
(State/local Air Pollution Agency)
Gridded Inventory of Area Combustion Sources
(State/local Air Pollution Agency)
Point Source Inventory Including Annual TSP Emissions
over 25 TPY and UTM coordinates
(State/local Air Pollution Agency)
Precise Location of Hi-vol Including Height and Distance
to all Major Sources (such as roads, parking lots, etc.,
within 60 meters)
(obtained during site tour)
Photographs of Site and Surrounding Area
(obtained during site tour)
Aerial Photographs (optional)
(usually available through COG)
10
-------
streets (e.g., presence of paving, curbs, etc.) is described. The location
and activity level of unpaved or gravel parking, for example, is estimated by
the best available means (talk with owner or attendant, or estimate lot capacity
and turnover rate). The locations of very close point sources (as calculated
from UTM coordinates) are confirmed. The mile radius area is checked for
omissions from the point source inventory. Any obviously unusual character-
istics that may not have been considered accurately in the area source allo-
cation scheme (such as a concentration of coal-or wood-fired space heaters and
fireplaces) are noted, and appropriate adjustments should be made to the
inventory. The final survey usually requires two persons for about one-third
to one-half day in the field per site, depending upon the complexity of the
area and the number of fugitive sources. All of this information is then used
to estimate an annual emission inventory for each of the sectors in Figure 1.
Predominant land use in each sector and in the several mile area sur-
rounding the inventory area can be estimated by visual inspection or, if
warranted, through land use maps. For further accuracy, the 1/4 mile radius
sector (Sector 1) can be divided into quadrants and these classified according
to land use. Such information may be useful for future planning and projection
activities. Table 3 indicates the land use categories and classification
criteria which were found to be useful.
An example microinventory for a site in Kansas City, Missouri, is in
Appendix B. It includes a site description, area source inventory, one mile
map, quarter-mile sketch, and photographs. Appendix C suggests uses for the
microinventory which are not directly related to the empirical technique which
will be discussed next.
11
-------
TABLE 3 Land Use Categories and
Classification Criteria
CATEGORIES
CRITERIA
Undeveloped
Agriculture
Light residential
Dense residential
Suburban commercial
Central commercial
Light industry
General industry
Heavy industry
Airport
No significant activity; includes
parks and pasture land
Active farming
Four or less dwelling units/acre
Greater than four dwelling
units/acre
Retail businesses in strip
development or shopping
centers
Central business district
Metal fabrication, warehousing,
trucking, etc.
Controlled industrial processes
Steel mills, foundries, coking,
etc.
Municipal airport
-------
DEVELOPMENT OF THE EMPIRICAL TECHNIQUE
This technique represents an approach to quantifying the relationship
between participate emissions, monitor siting characteristics, and annual
ambient measurements of Total Suspended Particulates (TSP) using a micro-
inventory data base. It requires a moderately intensive data gathering effort
to quantify the emissions and environment in the immediate vicinity (i.e., up
to five miles) of high volume sampler sites.
The empirical method is a regression equation which relates monitoring
site characteristics and microinventory information to observed annual mean
TSP concentration. The data base used in the development of the method included
data for 79 sites in four urban areas (Portland, Oregon, Bimingham, Alabama,
and the St. Louis and Kansas City metropolitan areas). ' ' '
FORM OF THE METHOD
The form of the method is a multiple linear regression equation with
annual geometric mean TSP concentration as the dependent variable and four
independent variable terms. The independent variables are: 1) AREA, annual
area source particulate emissions within one mile (including fugitive dust and
influence of monitor height); 2) POINT, annual point source emissions within
five miles; 3) LOCAL, local sources (those traffic-related sources within
200 feet) and 4) VISPLUME, an indicator variable for the presence of a visible
plume of resuspended dust resulting from passing traffic on the nearest streets
(within 200 feet). In addition, the differences in concentrations among urban
areas due to areawide factors (e.g., natural and transported background) not
included in the microinventory were estimated by assigning a "city effects"
13
-------
classification variable to each urban area. These are also independent
variables in the equation. These variables were selected after examining the
correlation matrix for approximately 30 potential variables, as listed in
Appendix D.
As can be seen from the appendix, many different data transformations
involving height of the monitor, distance of the sources, etc., were attempted.
The final selection of variables was based on the degree of intercorrelation
between each of these variables and average ambient air quality. Partial
correlation coefficients were developed, holding the effects of other vari-
ables constant, and those selected gave the highest partial correlations to
average air quality. Also, consideration was given to the interdependence
among the independent variables in arriving at the final selection. For
instance, it may have been desirable to consider individual categories of area
source emissions, such as paved roads, unpaved roads, and others separately.
However, the relatively high intercorrelation between such variables necessi-
tated the exclusion of one of these variables from the regression model.
Multiple regression was then employed to derive the coefficients for each of
the terms in the equation and to obtain the statistical values used to judge
15
the accuracy of the regression equation.
The multiple correlation coefficient, (r), with the four variables and
city effects constants, K., in the regression equation was 0.876 using all 79
data sets; none were eliminated as outliers. Therefore, the seven variables
explained 77 percent of the variance (r2) in measured concentrations.
14
-------
Fhe regression equation is given by:
where AVGAQ
AREA
AVGAQ = 0.00451 (AREA) + 0.00096 (POINT) +
50.5 (LOCAL) + 18.6 (VISPLUME) .+
K. ("city effects" variables)
Predicted annual geometric mean, yg/m3
Ai
HGT
POINT
(1)
AT
25
0.0324 HGT
0.16
A6+A7+A8+A9
0.6084
The constants .0324, .16 and .6084 are the square of
the radius in miles to the area weighted center of the
annular ring defined by the sectors in the numerator.
Total area source emissions in sector i, ton/yr
(sector locations are shown in Figure 1)
Height of sampler, ft
n PSEM.
£ -mr- (WWF)
n =
PSEMi
Di
WWF =
LOCAL -
ADTj
DIS1
VISPLUME =
Number of point sources within 5 miles
Emissions from point source i, ton/yr
Distance to point source i, mile
(lower limit of D. is 0.5)
Wind weighting factor, computed as annual wind direction
frequency of occurrence (%} in quadrant where source is
located, divided by 2b, dimensionless
In ADT
In APT?
/HGF+DIS22
= Average daily traffic on nearby road i, veh/day
= Distance to road i, ft
(upper limit of DIS. is 200)
0 if no visible plume of resuspended dust results from
passing traffic on paved roads, 1 if there is a plume.the
Care should be taken to assure that the conditions on
street are typical of year-round conditions and not a
short term dust loading, for example.
15
-------
K. (CITY EFFECTS)^ Classification variables to account for different
1 background concentrations in each of the cities.
The "city effects" for each city were found to be:
Portland, 35.6 yg/m3
St. Louis, 50.6 ug/m3
Kansas City, 57.2 y
Birmingham, 50.3 ug/m3
The predicted versus observed concentrations are plotted from equation 1 in
Figure 4.
Regression equations can be developed from the data sets in any one city
and their use will be discussed later. For example, the equation derived with
just the 32 Kansas City sites would be:
AVGKC = 0.0054(AREA) + 0.00057(POINT) + (2)
55.2(LOCAL) + 10.6(VISPLUME) + 57.6
Similar equations were not calculated for the other cities because they had
too few sites.
EVALUATION OF METHOD
Three separate evaluations of the regression model (Equation 1) were
performed: 1) by using the statistics provided by the multiple regression
program; 2) by testing stability of the coefficients in the equation; and 3)
by testing the method's predictions with a new data set of emissions and air
quality data collected at 28 sites in three other areas (Philadelphia,
Pittsburgh and New England).
The summary statistics from the multiple regression analysis with 79 site
data set are presented in Table 4. The correlation matrix revealed no signi-
ficant intercorrelations between any of the independent variables used in the
16
-------
20 40 60 80 100 120 140 160
PREDICTED CONCENTRATION,,
180
Figure 4. Plot of predicted versus observed
concentration for regression
17
-------
TABLE 40 SUMMARY OF MULTIPLE REGRESSION ANALYSIS
Variable
AREA
POINT
LOCAL
VISPLUME
Constant
Regression
Coefficient
0.00451
0.00096
50.5
18.6
*
Std Error of
Regression
Coefficient
0.00082
0.00014
12.0
4.9
Computed
F Value
30.4
45.7
17.6
14.6
* The constant can be thought of as the value of the "city
effects" variable. For instance, for a site in Portland,
the intercept could be thought of as 35.6 yg/m3.
18
-------
regression. The t-test was used in testing the regression coefficient for
each of the independent variables to determine whether each was significant.
For the four variables and city effects constants and the 79 observations, F
values in Table 4 indicate that all the coefficients are highly significant.
The standard errors for the coefficients of the variables AREA, POINT,
LOCAL and VISPLUME were only about 20 percent of the coefficient values. The
standard error of the regression equation was 13.6 pg/m3, about 17 percent of
the average measured concentration. The relative errors for all but three of
the 79 predicted concentrations were less than 28 percent. It should be
emphasized again that none of the observations from the four cities were
removed before performing the regression analysis, although some were sus-
pected to contain anomalies. It was felt that these "anomalies" may represent
"real" but unusual situations which might be encountered by users of this
model.
The stability or sensitivity of coefficients in the equation was tested
by running two additional regression analyses, each with 75 percent of the
sites (selected at random) from the initial analysis. This was done to assure
that there were not a small number of sites which might be responsible for
determining the value of the coefficients. The AREA and POINT coefficients
were found to be very stable, the LOCAL coefficient fairly stable, and the
VISPLUME coefficient fairly sensitive to the specific sites included in the
analysis. It was shown that excluding just a few influential sites from the
regression could affect the VISPLUME coefficient by as much as 15 percent.
However, VISPLUME was judged sufficiently stable and important to remain
included in the regression equation.
19
-------
In an effort to independently test the fundamental relationships in the
empirical method, microinventories were obtained in Philadelphia, Pittsburgh
and several sites in the New England area. These microinventories were then
used to calculate values for the variables AREA, POINT, LOCAL and VISPLUME at
each site. The product of the variable values and the variable coefficients
for each variable were added together to give a "predicted" concentration at
each site. These "predictions" for the sites in each area were compared with
average annual mean concentrations at the sites to give an indication of the
extent to which the regression explained the variance among the concentrations
in each area.
Since equation 1 was developed empirically for four specific areas, it
may not be appropriate to use it as a predictive model in other areas. This
is because other factors, such as special terrain, or climate, may influence
the absolute values of the concentrations it predicts. When applying the
model to other areas, it is preferable to use it in a relative sense. This
will be discussed later.
An example comparison of predicted and observed concentrations is given
in Figure 5. In calculating the predicted concentrations in Figure 5, only
the LOCAL, AREA, POINT and VISPLUME terms of equation 1 were used. Table 5
summarizes the correlation coefficients (between predicted and observed) and
city effects constants (K) for each of the three test case areas. In Table 5,
the city effects constants, K, were calculated as the average differences
between (observed concentration - predicted concentration) at each site in the
area.
20
-------
140
120
en
.E
en
=L
100
oc
2
LU
o
2
O
Q
a:
LU
80
60
SL
t>
40
o
o
o
20
20 40 60 80
PREDICTED CONCENTRATION,
100
120
Figure 5. Plot of predicted versus observed
concentration in Philadelphia Using
LOCAL, AREA, POINT and VISPLW1E terms
In Equation 1
-------
TABLE 5. COMPARISON OF PREDICTED AND OBSERVED
CONCENTRATIONS IN THREE URBAN AREAS
K. Average Difference
City
Philadelphia5
Pittsburgh0
New England°'d
No. of
Sites
10
7
11
r r2
0.95 0.90
.94 .88
.93 .86
between Predicted ai
Observed, yg/m3
58
70
34
a It represents the fraction of the variation in concentrations between
sites, which is explained by the model.
Reference 17
0 Reference 18
Reference 19
22
-------
The "city effect" represents that portion of the concentration which is
assumed to be relatively constant across the study area and is not accounted
for by the AREA, POINT, LOCAL and VISPLUME terms of the equation. This portion
is thought to have several distinguishable components. The largest component
of K is probably background, the portion of ambient concentrations which
cannot be reduced by controlling emissions from local manmade sources. Back-
ground in turn has many components, including windblown dust from natural
surfaces, biological debris, and long-range transport of both primary and
secondary aerosols. Background for an urban area is usually measured at a
nonurban site near the urban area which is unaffected by nearby emission
sources and represents the air which flows into an area.
Two other probable components of the constant value could be collectively
described as "urban excess," in that these components would not be accounted
for in a measured nonurban background concentration. The first of these is
locally emitted sulfates, nitrates, and organics that form secondary aerosols
in the atmosphere. Long-range transport would include additional secondary
aerosols that would be included in the background value described above. The
other component would be the collective impact of area and small point sources
located in the urban area but outside the one mile radius microinventory area.
Since the impacts of the above sources vary considerably from city to
city but to a much smaller degree for different sites within the same urban
area, the constant should also vary depending on the city. The method for
determining this city-specific correction to equation 1 is discussed further
TO
in the report which describes the validation of this empirical technique.
23
-------
POTENTIAL APPLICATIONS AND EXAMPLES
There are two applications of the regression study which will be dis-
cussed in more detail. These include the predictions of source impacts from
the calculation of city-specific regressions (e.g., Equation 2) and the
calculation of relative source impacts from Equation 1.
Calculation of a City-Specific Regression (Equation 2)
The most desirable way to apply the above analysis is to estimate the
regression equation coefficients for a specific city using the four regression
variables (LOCAL, POINT, AREA, VISPLUME) and data for sites in that city. To
do this, one would calculate values for each of the four variables using the
microinventory collected at a number of sites in an urban airshed. Then a
multiple regression would be performed to estimate the regression coefficients
for the specific city involved. No firm guidance can be given on how few or
how many sites are necessary to perform such a regression because the vari-
ability of the data and representativeness of the sites will be different in
different situations. It may be possible to perform the regression using as
few as 12 or 15 sites, if there is good agreement in the data and if the
regression coefficients appear stable. Such a decision (on the number of
sites required) will require inspection and evaluation of the preliminary
regression runs and standard statistical texts should be consulted. If such a
regression were performed, the equation could be used as a predictive model
within the range of the observations.
An example of a city-specific regression was given in Equation 2 for data
in Kansas City. The coefficients of the other variables are slightly different
from those in Equation 1, indicating that some factors influencing the source-
receptor relationship in Kansas City may be slightly different from the "average"
24
-------
conditions in the four cities used to develop equation 1. Several factors
could effect the values of city-specific regression coefficients, the most
likely being dispersion meteorology. Other factors might be the presence or
absence of major point source (or fugitive dust) oriented sites or just random
differences in the most probable values, exacerbated by the smaller size of
the data set.
Calculation of Relative Source Impacts from Equation 1
Unfortunately, it may not always be possible to calculate a city-specific
regression equation. In these cases, it would be desirable to use Equation 1
in some manner to assist in the analysis of the TSP problem in that. area.
Therefore, it is suggested that the variables in Equation 1 (AREA, POINT,
LOCAL, and VISPLUME) be interpreted only in a relative sense to estimate
source impacts. It would be inappropriate to use Equation 1 to predict con-
centrations at sites in cities other than the ones used in the data base
(St. Louis, Birmingham, Portland and Kansas City). (Predictions were calculated
for sites in Philadelphia, Pittsburgh and New England as a means of illus-
trating the validity of equation 1, but should not be interpreted as "yg/rn3"
concentrations for control strategy development or other uses.)
An example of the calculation of the relative impacts of the variables
LOCAL, AREA, POINT, and VISPLUME are discussed below. First, the variable
values are calculated for each site. Then the coefficients (from Table 4) are
multiplied times the variable values to give the calculated concentrations at
each site which are associated with each variable. As an example, the cal-
culation of the contribution from LOCAL sources at a site is given below.
25
-------
LOCAL = •'" ""[1 + ^-^L
where: ADTl = 20 = average daily traffic on Allegheny Avenue,
vehicle/day
DISj = 20 = distance from monitor to Allegheny Avenue,
feet
ADT2 - 4900 = average daily traffic on Delaware Avenue,
vehicle/day
DIS2 = 65 = distance from monitor to Delaware Avenue,
feet
HGT = 15 = height of monitor, feet
LOCAL = .25 (calculated variable value)
Then the product of the coefficient and the LOCAL variable value is
calculated. From Equation 1, the coefficient of LOCAL is 50.5. Calculating
the product of the variable value, .25, and the coefficient, 50.5, gives a
contribution due to LOCAL sources at the site of 12.6 yg/m3.
The contributions of the LOCAL, POINT, AREA, and VISPLUME source categories
at each site are similarly calculated as the first step in calculating relative
concentrations. These predictions are then summed and the percentage contri-
butions of each source category are calculated. When these percent contribu-
tions are compared, they represent the relative contributions of each source
category to the ambient annual concentrations. The sum of the predicted
concentrations may be compared with the actual observed concentration at a
site. The difference is the amount of the observed which is unaccounted for
by the LOCAL, POINT, AREA and VISPLUME terms. Table 6 gives an example of the
predicted and relative impacts and amount unaccounted for at two sites in
Philadelphia.
26
-------
TABLE 6. EXAMPLE TABULATION OF THE PREDICTED AND
RELATIVE IMPACTS OF POINT, AREA, LOCAL
AND VISPLUME SOURCE CATEGORIES AT TWO
SITES IN PHILADELPHIA
Site
Allegheny Avenue
Predicted, yg/m3 Relative,
Broad and Soruce
Predicted, yq/m3 Relative,
LOCAL
VISPLUME
POINT
AREA
12.4
0
7.8
22.1
29
0
18
53
28.5
0
.44
21 .9
56
0
1
43
Total Predicted
42.3
50.8
Annual Average
Unaccounted for
by LOCAL, AREA
POINT, VISPLUME
Terms
107
65
116
65
27
-------
Control Strategy Applications
The method outlined above can provide much useful information for the
development of control strategies in urban areas. The major limitation of the
method is that it cannot be used to predict a "hot spot" concentration in an
area where there are no monitors, such as can be done with a dispersion model
unless the hot spot can be identified by some other means and a microinventory
developed for that area. However, it provides much information about the
relative impacts of source categories on measured annual concentrations.
Additional information on the sources which comprise the AREA and POINT
concentrations can be inferred from the emissions inventories. If desired,
the POINT, or AREA relative impacts could be apportioned among the major
source categories which are found in the microinventory and were used in
computing these variable values. The LOCAL and VISPLUME variables are sug-
gestive of resuspended fugitive dust and vehicle exhaust emissions problems as
well as monitor siting considerations.
LIMITATIONS OF THE METHOD
As with any technique for assessing data, there are certain limitations
which must be recognized. These limitations include the range of variable
values which can be used, as well as meteorological, geographic and other
considerations.
Since this is a linear regression model, and since the model was "fit" to
a limited range of variable values, it should only be used with variable
values which are in the range of those in the 79 site data base. The minimum
and maximum values for each of the variables are as follows:
28
-------
Minimum Maximum
AREA 0 12,850
POINT 0 74,150
LOCAL 0 .95
There is some indication that the LOCAL relationship may result in some-
what high estimates for very low traffic levels (i.e., less than 1000 vehicles
per day). This should be considered when interpreting the data. It may be
justified to delete such roads from the analysis if, upon visual inspection,
the roads appear to have very low levels of dust in the traffic lanes.
Another limitation concerns the minimum distance from point sources and
the confounding effect of plume rise. Very few sites in the data base had a
substantial volume of industrial emissions within one-half mile of the moni-
tor. It can be shown from dispersion model calculations that high stacks and
buoyant plumes may not have a significant effect on nearby monitors. There-
fore it is suggested that no point sources closer than one-half mile be
included in the POINT estimate. Analysis should not be attempted for sites
which have a substantial number of stack emission points within one-half mile.
Some discretion must be used in applying this method to highly abnormal
situations such as very dry climates, areas with excessive windblown dust or
areas with very poor dispersion climatology. City-specific models may ame-
liorate some of these problems. Also, the method purposely limits the geo-
graphic area of consideration to one mile for area sources and five miles for
point sources. Experience has shown that, while much of the particulate
burden does come from outside of this area, most of this comes from outside of
1 2
the urban area and not from, say a five-to-ten mile distance. '
29
-------
As with any technique using emissions data, the accuracy of the inventory
is very important. The microinventory procedure assures that sources within
the prescribed limits are at least considered. However, it does not ensure
the accuracy of these emission estimates. Faulty emission factors, seasonal
variations in emission rates and process fugitive emissions are but a few of
the factors that can contribute to an inaccurate emissions inventory.
30
-------
References
1. David A. Lynn, et al., National Assessment of the Urban Participate
Problem. Volume I: Summary of National Assessment. EPA-450/3-76-024,
U.S. Environmental Protection Agency, Research Triangle Park, N.C.,
July 1976.
2. F. A. Record and Robert M. Bradway, Philadelphia Particulate Study,
Draft Report, GCA-TR-78-02-G, GCA/Technology Division, Bedford,
Mass., February 1978.
3. Analysis of Probable Particulate Non-attainment in the Kansas City
AQCR, EPA Contract No. 68-02-1375, U.S. Environmental Protection
Agency, Research Triangle Park, N.C., February 1976.
4. Personal communication, T. G. Pace and Brock Nicholson, Chief Engineer,
North Carolina Air Quality Section, Raleigh, N.C., June 1977.
5. T. G. Pace, "The Relative Impact of Vehicle-related Particulate on
Particulate Concentrations and Rationale for Siting Hi-vols near
Roadways," U.S. Environmental Protection Agency, Research Triangle Park,
N.C., April 1978.
6. Fugitive Dust Survey and Inventory, Final Report, EPA Contract
No. 68-02-2093, U.S. Environmental Protection Agency, Research Triangle
Park, N.C., October 1977.
7. Characterization of Particulate Sources Influencing Monitoring Sites
in Region VIII Non-attainment Areas, Draft Report, EPA Contract
No. 68-02-1375, U.S. Environmental Protection Agency, Research
Triangle Park, N.C., June 1976.
8. Fugitive Dust in Kansas and Nebraska, EPA Contract No. 68-02-0044,
U.S. Environmental Protection"Agency, Research Triangle Park, N.C.,
February 1974.
9. Thompson G. Pace, Kenneth Axetell and Robert Zimmer, "Microinventories
for TSP," Proceedings APCA Specialty Conference on Emission Factors and
Inventories, Anaheim, California, November 1978.
10. Compilation of Air Pollutant Emission Factors, AP-42,
U.S. Environmental Protection Agency, Research Triangle Park, N.C.,
May 1978.
11. TSP Source Inventory around Monitoring Sites in Selected Urtan
Areas - Kansas City, FEDCo Environmental under Contract No. 68-02-2603,
Task Order 15, to U.S. Environmental Protection Agency,
Research Triangle Park, North Carolina 27711. To be published.
31
-------
12. TSP Source Inventory around Monitoring Sites in Selected Urban Areas-
St. Louis. PEDCo Environmental under Contract No. 68-02-2603, Task
Order 15, to U.S. Environmental Protection Agency, Research Triangle Park,
North Carolina 27711, December 1978.
13. TSP Source Inventory around Monitoring Sites in Selected Urban Areas-
Portland, Oregon'PEDCo Environmental under Contract No. 68-02-2603,
Task~0rder ]S, to U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina 27711, December 1978.
14. TSP Source Inventory around Monitoring Sites in Selected Urban Areas-
Birmingham, Alabama. PEDCo Environmental under Contract No. 68-02-2603,
Task Order 15, to U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina 27711, December 1978.
15. Norman H. Nie, C. H. Hull, J. G. Jenkins, Karin Steinbrenner, and
D. H. Bent, Statistical Package for the Social Sciences, Second
Edition, McGraw-Hill Book Company, New York, 1975.
16. George W. Snedecor and William G. Cochran, Statistical Methods, Sixth
Edition, Iowa State University Press, Ames, Iowa.
17. TSP Source Inventory around Monitoring Sites in Selected Urban Areas-
Philadelphia, Pennsyl vania~PEDCo Environmental under Contract
No. 68-02-2603, Task Order 15, to U.S. Environmental Protection Agency,
Research Triangle Park, N.C., December 1978.
18. Validation of Empirical Model for Estimating TSP Annual Concentrations,
PEDCo Environmental under Contract No. 68-02-2603, Task Order 21, to
U.S. Environmental Protection Agency, Research Triangle Park, North Carolina,
January 1979.
19. Fletcher G. Shivers, Inventory Development for Evaluation of Measures
for the Control of Nontraditional Sources of Particulates in Southern
New England. Draft Final Report, GCA Technology Division, Bedford,
Massachusetts under U.S. Environmental Portection Agency Contract
No. 68-02-2607, September 1978.
32
-------
APPENDIX A
RAILROAD YARDS
Most of the emissions in -rain yards appear ~o be from
wind erosion rather than from train movement. Therefore, an
adaptation of the U.S. Department of Agriculture's wind
erosion equation was used as the basis for deriving emission
factors for railroad yards. The modified equation estimates
annual suspended particulate emissions due to wind erosion
from exposed soil surfaces:0
S = a I K C L1 V (eq.1)
where S = emission factor, ton/acre/yr
a = portion of total wind erosion losses
that would be measured as suspended
particulate, estimated to be 0.025
I = soil erodibility, ton/acre/yr
K = surface roughness factor
C = climatic factor-
L1 = unsheltered field width factor
,V = vegetative cover factor
In this equation, K, C, L1, and V are all dimension-
less. An average erodiblity (I) of 56 ton/acre/yr was
assumed to be applicable for the soil in the railroad yards.
The area around railroad yards generally has a fairly smooth
surface which is equivalent to a X value of 1.0. The climatic
factor for the Kansas City area is 0.10. An average width
of 1000 ft was assumed for railroad yards, since this is not
a critical variable in the equation. The L' value equiv-
alent to L = 1000 ft was determined from a nomograph in
which soil type is also a variable--0.64. There is no
vegetation in rail yards so V = 0 and V = 1.0. Also, these
areas are generally treated with oil or chemicals to reduce
A-l
-------
dust and weed problems. Therefore, the emissions have been
estimated with the above values and the following assump-
tions:
° no vegetation
0 sandy soil (I = 56)
0 half the area is stabilized, with an
emission rate of 20 percent of the
unstabilized area
0 train traffic increases the wind
erosion emissions by 20 percent
Therefore, the resulting emission factor is:
E = (0.025) (56) (1.0) (0.1) (0.85) (1.0) (l- * °'2) (1.2)
= 0.1 ton/acre/yr
REENTRAINED DUST FROM PAVED STREETS
The most recent emission factors for estimating emis-
sions from paved roads were obtained from a draft of AP-42.
The recommended emission factors are 5.0 gm/VMT for clean
streets. These factors include an exhaust emission factor
and are not corrected for days with rain.
Based on regional estimates for the percentage of heavy
duty gasoline and diesel trucks, an exhaust emission factor
of 0.6 gm/VMT was calculated from published emission factor
data. This factor was subtracted from the paved road emis-
sion factors and the resulting values were corrected for
rainfall.
Rainfall data available for 1974 indicate that there
were 86 days with at least 0.01 inch of rain. Based on this
information, the reentrained dust emission factors were
A-2
-------
-educed by approximately 24 percent. The corrected emission
*ac<-o*-s are: 3.4 gm/VMT for clean streets and 14.8 gm/VMT
for commercial streets.
•JNPAVED ROADS
Manv different groups have investigated emissions from
unpaved roads. All the studies show emission rates usually
in the range of one to 20 Ib/VMT, and show that the rate is
hiahly dependent on vehicle speed. After reviewing these
studies, EPA selected the following emission factor for
publication in AP-42:
£= (0.6) (0.81) (s) (S/30) (^..,) (eq.2)
where E = emission factor, Ib/VMT
0.6 - fraction of total particulate
less than 30 u diameter
s = percent silt of road surface
material
S = average vehicle speed, mph
w = mean annual number of days with
0.01 inch or more of rainfall
The percent silt on gravel road surfaces is about 12
percent based on data presented in the EPA publication,
Development of Emission Factors for Fugitive Dust Sources.
For graded and drained road surfaces, no aggregate material
is applied to the roadbed so it is composed of compacted
native soil. The fine material originally on the surface is
probably rapidly removed by turbulence from passing vehicles
or by wind and water erosion forces. The remaining stable
surface is composed of sand- and pebble-sized particles,
with dust being generated primarily by the continuing mechan-
ical breakdown of these particles as a result of traffic.
A-3
-------
It is assumed that the percent of silt-sized particles on a
seasoned dirt road surface is approximately the same as that
ror gravel/ or 12 percent. The number of days with 0.01
inch or more of rainfall is 86. Assuming a vehicle speed of
20 mph, the average emission factor is 3.0 Ib/VMT.
CLEARED OR EXPOSED AREAS
The modified wind erosion equation described earlier
was used for estimating emissions from cleared or exposed
areas. An average erodibility (I) of 38 ton/acre/yr was
assumed for the soil types common to this area. The cleared
areas are generally left with a fairly smooth surface, which
is equivalent to a K value of 1.0. The climatic factor for
the Kansas City area is 0.10. An average field width of
1000 ft was assumed for all the cleared areas, since this is
not a critical variable in the equation. The L' value
equivalent to L = 100 ft was determined from a nomograph in
which soil type is also a variable—0.64. There is no
vegetation on cleared areas, so V = 0 and V = 1.0.
Based on these input data, the average emission rate
from cleared areas would be 0.06 ton/acre/yr.
CONSTRUCTION
The emission factor presented in AP-42 of 1.2 ton/acre/
mo was corrected for the climatic conditions in Kansas City
by use of the equation:
E = : 7 (eq.3)
(PE/50)
where PE = 95 for Kansas City
E = 0.33 ton/acre/mo
A-4
-------
This value is appropriate for residential and commercial
construction with excavation and regrading, and includes the
effect on emission reduction of moderate watering of the
site.
AGRICULTURE
Total emissions from agriculture are calculated from
the sum of tilling and wind erosion emissions. The factor
for tilling from AP-42 is:
E= (0.8)(1.4)s/(PE/50)2 (eq.4)
where E = emission factor, Ib/acre/run
0.8 = fraction of total particulate less
than 30 u diameter
s = percent silt in surface soil
PE = precipitation-evaporation index
An average of three tilling operations per year is assumed,
and a silt content of 40 percent. Therefore, total tilling
losses are 37.2 Ib/acre.
The modified wind erosion equation described earlier is
recommended for determining the wind erosion component of
agricultural emissions. All variables in the equation
except C are dependent on either the soil type or crop grown
on each field. However, this component is lower in mag-
nitude than tilling losses in the Kansas City area, so the
following average values were assumed:
A-5
-------
I = 47
K = 0.6
C = 0.1
L = 2000 ft, L' = 0.73
V = 500 Ib/acre, V = 0.14
E = 0.007
Therefore, the combined agricultural losses are 0.026
ton/acre/yr.
AGGREGATE STORAGE AREAS
The emission factor for this source category was based
on sampling performed in the Kansas City and Cincinnati
areas. The emission factor (E) is:
E = 13.2 Ib/acre/day
Assuming activity for six days a week, the annual emission
factor is:
E = 2.1 ton/acre/yr
UNPAVED PARKING LOTS
If the average number of vehicles using a parking lot
daily and the average distance of travel in the lot can be
estimated, these data can be converted to VMT/yr and then
multiplied by the emission factor for unpaved roads to
determine unpaved parking lot emissions. With an average
speed of 10 mph in the parking lots, the emission factor is
1.5 Ib/VMT. Usage of most lots for only five days per week
was assumed in calculating annual VMT.
A-6
-------
UNPAVED STORAGE AREAS
No factor was found for this source category. By
comparison with other source categories, it was estimated
that emissions per acre from this source were greater than
for a cleared area (0.06 ton/acre/yr) or railroad yards
(0.1), but not nearly as high as aggregate storage (2.1) or
construction (4.0). A value of 0.2 ton/acre/yr was used for
this minor source category.
GRAIN ELEVATORS
A value of 1.80 ton/silo/yr was developed from file
data for a previous PEDCo study of particulate air quality
in rural Kansas and Nebraska. This same value was applied
here for sources for which NEDS data were not available.
INCINERATORS
The emission factor for multiple chamber industrial/
commercial incinerators from AP-42 was used in this inven-
tory. That value is 7.0 Ib/ton of refuse incinerated.
A-7
-------
Emission Factors and Formulas Used to Calculate
Fugitive Dust Emissions in Microinventory
Source category
Emission factor
Units
Railroad yards 0.1
Reentrained dust
paved streets 3.4
exhaust 0.6
Unpaved roads 3.0
Cleared or exposed areas 0.06
Construction 0.33
Agriculture 0.03
Aggregate storage 2.1
Unpaved parking lots 1.5
Unpaved storage areas 0.2
Grain elevators 1.8
Incinerators 7.0
ton/acre/yr
gm/vrrrb
gm/VMT
Ib/VMT
ton/acre/yr
ton/acre/mo
ton/acre/yr
ton/acre/yr
Ib/VMT
ton/acre/yr
ton/silo/yr
Ib/ton of refuse
"Other potential sources of fugitive dust which were not inventoried include:
Mining
Quarrying
Materials Handling
Snow and Ice Control
Feed Lots
Fairgrounds, etc.
The emission factor given in Supplement 8 of Reference 9 will be 5.6 gm/VMT.
This was "adjusted" to account for days with precipitation since it was
assumed that resuspended dust would be suppressed on those days. Also, this
factor was modified to exclude auto and diesel exhaust emissions.
A-8
-------
APPENDIX B
DESCRIPTION OF SITE
SAROAD code - 26 2380 005 HOI
Location - 6402 East 37th,
Kansas City, Missouri
Monitor height - 35 ft, on roof of
fire station
Land use, by sector -
la light: industry
Ib light residential
Ic undeveloped
Id light residential
2 light industry
3 general industry
4 undeveloped
5 light residential
6 light industry
7 light industry
8 dense residential
9 dense residential
10 general industry
11 dense residential
12 dense residential
13 dense residential
Localized sources, within 200 ft of monitor -
Source Distance Description
Stadium Drive
Unpaved parking lot
Unpaved alley
Fremont
ft
55
105 ft
100 ft
60 ft
7870 ADT, dirty, uncurbed
0.1 acres, 10 cars
5 cars
75 ADT, dirty, uncurbed
Air quality data -
Year
1977
1976
1975
1974
1973
Annual geometric mean, ug/m"
85
89
86
89
101
No. of samples
53
54
47
51
B-l
-------
-
.
•• . H-,;
^ .^c.
-\ ••...,?,— >•;•« ;; •..;-
' *
.
"
^nU'/ ^
One mile radius around 6402 East 37th site
B-2
-------
1,7. Unpaved pkg lot, 0.1 ac, 10 cars
2. Cleared area, 0.1 ac
3,6. Unpaved pkg lot, 0.1 ac, 5 cars
Unpaved storage area, 0.2 ac
Unpaved pkg lot, 0.3 ac, 25 cars
Unpaved pkg lot, 1.7 ac, 35 trucks
Unpaved pkg lot, 1.5 ac, 150 cars
. Unpaved storage area, 2.0 ac
. Unpaved storage area, 8.1 ac
One-quarter mile radius around
4.
5.
8.
9.
10
11
Scale 1" = 400'
Unpaved pkg lot, 0.8 ac, 20 cars
Unpaved pkg lot, 0.7 ac, 50 cars
Unpaved storage area, 3.4 ac
Unpaved storage area, 2.3 ac
Unpaved road, 400 ft, 5 cars
Unpaved road, 300 ft, 10 cars
18,19. Unpaved alley, 750 ft, 5 cars
20. Unpaved road, 675 ft, 5 cars
21. Unpaved road, 600 ft, 5 cars
6402 East 37th site.
12,
13.
14.
15.
16.
17.
B-3
-------
Viewed to the northeast.
Viewed to the northwest.
Sampler at 6402 East 37th.
B-4
-------
POINT SOURCE SUMMARY
Site: 6402 East 37th
Plant Emission Distance Compass
number level, t/yr from site, mi direction,0
26 48 .39 180
25 893 1.7 35
24 98 2.4 25
23 •- 43 2.7 20
22 1,233 3.5 10
16 249 4.3 5
15 41 4.4 0
13 341 4.7 350
14 327 4.8 0
9 46 5.1 330
B-S
-------
CO
O
LO
OS
4J
r-
ro
W
•H
CO
i-l
X
c
o
4-1
0
4J
u
o
rQ
en
Q
• H
W
-H
rH
05
O
CT>
CO
v£>
LTI
^
CO
rH
>i
•*-*
•rH O
•H 03
4J M
u
rfj
^
'M
o ;
VD jrH
01 CO
in
0
m
0
m
t
O
in
o
rH
O
rH
d
rH
O
rH
d
OJ
o
a
o
a
rH
CT1
<3<
CTl
rH
Ol
3
^-4
cji ' !
O.1
d
^
o
^r
d
m
d
a
E
0)
rH
CO
o
CO
rH
a>
^)
U-l
ra
c
M
^\
|
0
o
{/)
Vi
0
(TJ
J^
OJ
c
u
c
M
1
ro
in
n
rH
OJ
n
r-~
.
vO
rH
1
l_t
•H
^
"*\
rH
Oj
K
0
c^
OJ
VO
0
(N
O
o
U>
d
CA
rH
1
kD
O
rH
>i
,
r-
en
J..)
t')
a
r;
X<
X
0)
0
4J
3
£'
ro
O
rH
V
JOT kD
i Sj->
•~* 1
O
j j
rH
W
01
U
V-l
o
w
^
(D
Cl
(/I
in
(U
U
O
GL|
•o
c:
M
<•
^~4
f-O
^)
Q
W
J^>
M
^H
M
O
a
t.
•TO
r^
o
CO
o
r-~
rH
•
OJ
^.
o
n
vo
ro
vo
O
rH
H
\p
u
05
CO
^
t/1
rQ
^j
05
^
TJ
O
0
rH
X
m
.
*3"
CTl
to
JJ
Q)
£
rH
OJ
rH
H
^
^
O
rH
X
•
OJ
t/1
Ti
rj
O
4Jl 1
CO
M T3
rH
•H
nj
a
0)
ro
^l
T!
G)
^>
d
C
D
Cr
0)
C
u
m
, — i
•
0
(/)
Oj
ai
M
OJ
'O
GJ
Vj
01
rH
U
C
OJ
n
1
O 0'
X)
O
3
V-i
JJ
,—
JM| 3
C
O
U
u
en
«N
m'
o
0)
vD
, rH
cn
H5
01
OJ
d)
CT.
VJ
O
4J
CO
^1
!
^r
r-
0
OJ
rHi
CO
u
05
m
OJ
10
jj
0
rH
^Tl
'a,
t!
a)
ai
c<
C
i O,
Ifl
•°
wl
CD
(N
O!
O
CTl
oo
"Z
D< O
m
a) Ico
co
r>
al
M
M
£
B-6
-------
APPENDIX C
OTHER APPLICATIONS OF MICROINVENTORIES
Two uses of the micro-inventory method have been suggested, which do not
require use of the empirical method. These are intersite comparisons and
dispersion modeling.
Intersite Comparison of Inventories—The inventories for sites within a specific
city can be compared to one another to indicate the differences in the nature
of the sources near the monitors. Such a tabulation is made for two sites in
Kansas City in TableC-1. This table summarizes the emissions around a site
which is dominated by point sources (101 Lou Holland) and around a site domi-
nated by fugitive dust sources (1517 Locust). Since both sites report roughly
comparable high annual averages, it is reasonable to assume that the Lou Holland
concentrations are due primarily to the point sources and the Locust Street
site is primarily influenced by fugitive dust sources. Obviously, such a
clear pattern is not always so apparent, and a detailed analysis of the data
at many such sites may be necessary before making conclusive inferences about
the magnitude of the impact of these different types of sources. This could
be done using a dispersion model or through the use of a multiple linear
regression to explain effects of the variables used.
In another example, shown in Table C-2,, both sites have relatively small
industrial emissions. However, the area source inventory indicates great
differences in the fugitive dust emissions within the nearest 1/4 mile. In
this example, the Missouri Avenue site is dominated by localized sources,
whereas the Ann Avenue site is not so dominated. This difference in nearby
C-l
-------
TABLE C-l. Comparison of Point and Area Source Inventories at Sites
in Kansas City with Conparable Air Quality Concentrations
EMISSIONS ADD SOURCES NEAR MONITORING SITES a
101
AREA SOURCES
1/4 mile
1/4 to 1/2 mile
1/2 to 1 mile
POINT SOURCES
1 mile
1-2 miles
2-3 miles
AVERAGE CONCENTRATION
Lou Holland Drive
1.1
22.5
353.2
853
2266
4061
1517 Locust Street
87.1
196.1
769.7
28
805
1356
1975-76 Average Annual
Geometric Mean, yg/m3 93 102
a. Emissions in tons per year
TABLE C-2. Comparison of Local Fugitive Dust Source Contributions and Average TSP Concentrations
at Two Sites in Kansas City with Comparable Point and Area Source Emissions
'EMISSIONS AND SOURCES WITHIN 1/4 MILE, TPY *•
AREA SOURCES
Combustion
Industrial
Fugitive Dust **
LOCAL FUGITIVE DUST
Distance, ft.
Traffic, veh/day
405 Missouri Avenue
18.5
0.0
74.4
SOURCES **
65
46,220
619 Ann Avenue
5.6
0.0
22.4
130
800
AVERAGE CONCENTRATION *** 118 69
* Total area source emissions within 1 mile were comparable for the two sites
as were total point source emissions within 2 miles, tons per year
** Fugitive Dust includes dust resuspended by traffic
*** Average of annual geometric mean concentration in 1974 and 1975 at each site, ng/m3
— C-2
-------
sources and emissions could explain much of the difference in air quality
levels between the sites. As mentioned earlier, dispersion or regression
models could be used to help quantify these effects.
Dispersion Modeling Using Improved Emission Inventory Data Base—As mentioned
earlier, emission inventory input to dispersion models are generally conducted
using grid sizes of 1 or 2 km square. The emissions from many source categories
are "allocated" to these grids using population or other indicators. The
studies cited earlier indicate the importance of nearby sources in explaining
TSP concentrations and support the need for more accurate inventory methods
and data in the vicinity of particulate monitors.
The microinventory technique directly addresses the problems associated
with "allocation" procedures through the use of detailed site surveys and
inventories based on site inspections. Also, the technique can easily be
adapted to a Cartesian coordinate system (from polar coordinates), resulting
in a small, detailed grid system for use in dispersion models.
It has been established that distance from source to receptor is an
important consideration in explaining TSP concentrations. Thus, it follows
that a smaller grid size would provide greater resolution of source-receptor
relationships.
A rectangular microinventory grid, such as is depicted in Figure 2, is
recommended for use in dispersion modeling. It closely corresponds to the
polar gridding technique in Figure 1. Such a grid system should be incor-
porated into the grid system for the urban area being modeled to provide
better source-receptor resolution in the immediate vicinity of the monitoring
sites being used in model applications. In converting the polar inventory to
C-3
-------
rectangular grids, a simple and consistant method would be to grid the area
surrounding the site into .5 km square grids. The closest four grids would
be assigned the same emission density as inventoried in sector 1. The sur-
rounding ring of .5 km square grids would be assigned the same density as
sectors 2 through 5. The outer grids would be divided into 1 km square grids
and would be assigned the same density as sectors 6 through 9. Obviously, if
one were performing a microinventory specifically for dispersion model in-
ventory improvement, it should be made using the grid system in Figure 2.
C-4
-------
APPENDIX D
Variables Screened for Inclusion in Regression
The following variables were screened by computing simple correlations with
Average Annual Geometric Mean Concentration computed for up to five year's
data.
1. Height of monitor (HGT)
2. Distance of monitor to each of nearest two roads (DIS1 ,DIS2)
3. Average daily traffic on each of nearest two roads (ADT1 ,ADT2)
4. The following terms combining traffic, height and distance to
nearest roads.
a)
b)
c)
In ADT1 +
/ DIS1Z+HGTZ
-L
/ ADT1
/DIS1Z+HGTZ
ADT1
In ADT2
/DIS22+HGTZ
/ ADT'2
v/DIS2z+HGT2
ADT2
5. The following terms combining point source emissions (PS), distance
to each point source (DPS), stack height (STHT) and wind direction
weighting factor (WWF)
a) PS^ (wwf)
DPS
b) PS (wwf)
DPS2
PS(wwf) (STHT < 60 ft)
$
d) (STHT > 60 ft)
e) PS (wwf)
DPS (In(STHT))
f) PS (wwf)
DPS^ (In(STHT))
D-l
-------
6. The following terms combining the area source emissions for combustion
and small industrial sources (ACI01 thru ACI09); paved roads (APR01 thru
APR09); and other fugitive sources (AOF01 thru AOF09) where 01 thru 09
are sector designations.
09
a) ACI = z ACIn (wwf)
n=oi
09
b) APR = I APRn (wwf)
n=oi
09
c) ADF = i AOFn (wwf)
n=oi
d) ATOT = ACI + APR + AOF
e) ATOT1 = AGIO! + APR01 + AOF01
05 05 05
f) ATOT25 = z ACIn + I APRn + Z ADFn
n=02 n-02 n=02
09 09 09
g) ATOT69 = z ACIn + z APRn + Z ADFn
n=oe n=oe n=oe
h) ATOT1H = ATOT1 (jj|T)
i) ATOT25H = ATOT25 (jj|T)
j) ATOT69H = ATOT69 (^|T)
k) ATOTD = ATOT1 ATQT25 ATOT69
Dl D25 D69
where Dl , D25 and D69 are the distances to the area weighted
centers of Sectors 1 , Sectors 2-5 and Sectors 6-9
1) ATOTDS = ATOT1 ATOT25 ATOT69
~DP~ D252
Presence of visible plume of resuspended dust after passage of
car on a road, VISPLUME.
D-2
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-450/4-79-012
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
An Empirical Approach for Relating Annual TSP Concentra-
tions to Participate Microinventory Emissions Data and
Monitor Siting Characteristics
5. REPORT DATE
May 1979
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Thompson G. Pace
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Air Management Technology Branch
Office of Air Quality Planning and Standards
Environmental Protection Agency
Research Triangle Park, NC 27/11
10. PROGRAM ELEMENT NO.
?AA635
11. CONTRACT/GRANT NO.
none
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Final Report
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Procedures for developing and applying an empirical procedure for relating micro-
inventory data to annual TSP data is described herein. A microinventory is an inven-
tory and orderly compilation of sources in the vicinity of a high volume sampler. Its
purpose is to consolidate information on source emissions or activity rates, monitor
siting, air quality data and land use characteristics into a consistent format and data
base. The purpose of the microinventory is to assist in identifying the sources of
particulate matter in close proximity to a monitoring site. Such information can sub-
sequently be used to interpret observed differences in air quality at various monitorin
s i tes.
An empirical approach in which monitor siting characteristics and microinventory
information are related to measured annual mean TSP concentrations using multiple
regression analysis has been developed. The regression uses variables which consider
the emissions and distances to both point and area sources and the effects of traffic
and dust on nearby streets upon measured TSP concentrations. The purpose of this repor
is to describe: 1) the methodology used in compilation of a microinventory and 2) the
development of uses of an empirical approach for relating observed annual TSP levels
to microinventory and monitor siting information.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Particulate Matter
Modeling
Monitor Siting
Data Analysis
18. DISTRIBUTION STATEMENT
Unlimited
19 SECURITY CLASS (ThisReport/
Unclassified
21. NO. OF PAGES
57
20 SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
-------
TJ O
OJ CD -*•
83 *•
QJ_ O
<^
O^
~D 4;
9 ^
C
8
0! tD
£ 5
Q) C/l
3?OO
S **
------- |