United Slates
Environmental Protection
Acencv
Ollice ot Air Quality
Planning and Standards
Research Triangle Park. NC 277 1 1
HPA-454/R-94-025
October 1994
Air
& EPA
MODELING FUGITIVE DUST
IMPACTS FROM SURFACE COAL
MINING OPERATIONS - PHASE II
Model Evaluation Protocol
-------
EPA-454/R-94-025
Modeling Fugitive Dust Impacts from
Surface Coal Mining Operations - Phase II
Model Evaluation Protocol
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Technical Support Division
Research Triangle Park, NC 27711
-------
Disclaimer
This report has been reviewed by the Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, and
has been approved for publication. Any mention of trade names or
commercial products is not intended to constitute endorsement or
recommendations for use. Copies of this report are available for
a fee from the National Technical Information Service, 5285 Royal
Road, Spingfield, VA 22161.
EPA-454/R-94-025
-------
PREFACE
This model evaluation protocol was prepared by Midwest Research Institute
(MRI) and AlphaTRAC, Inc. (subcontractor) for the U.S. Environmental Protection
Agency under EPA Contract No. 68-D2-0159, Work Assignment (WA) No. I-06.
Mr. Jawad Touma is the EPA Work Assignment Manager (WAM) for the Technical
Support Division, Office of Air Quality Planning and Standards. This protocol presents
the methodology for evaluating the performance of atmospheric dispersion models in
predicting the fugitive dust impacts from surface coal mines. During the process of
developing this protocol, input regarding the objectives of the model evaluation and the
methodologies incorporated were solicited and received from the Wyoming Mining
Association and the State of Wyoming; comments received were incorporated into this
final protocol.
in
-------
CONTENTS
Preface iii
List of Figures vii
List of Tables vii
1. Introduction 1
2. Overall Approach to Model Performance Evaluation 5
2.1 Candidate modeling system components 5
2.2 Evaluation methodology 6
3. Source Representation 9
3.1 Emission sources 9
,V 3.2 Emission factors 10
^ 3.3 Source activity 15
;\ 3.4 Control efficiency 16
,Y 3.5 Geometric representation 18
1 3.6 Release height and initial vertical dispersion 26
; i 3.7 Particle size distribution 28
v
^ 4. Modeling Systems for Evaluation 31
"""
5. Observational Data Bases 33
5.1 Source activity 33
5.2 Meteorology 37
5.3 Air quality 38
5.4 Background air quality 39
5.5 Runstream preparation 40
6. Determination of Best-Performing Mode! 43
6.1 Strategy for identifying best-performing model(s) 43
6.2 Test statistic 44
6.3 Performance measures 45
6.4 Model Comparison Protocol 47
-------
CONTENTS (continued)
7.0 Evaluation of Model Overprediction 51
7.1 Overall evaluation strategy 51
7.2 Statistical evaluation of model overprediction 52
7.3 Historical data review 55
7.4 Model sensitivity analysis 56
8. References 59
VI
-------
LIST OF FIGURES
Number Page
1 Source representations for ISCST2: haul roads, haul truck loading
and dumping 21
2 Source representations for ISCSTM: haul roads, haul truck
loading and dumping—Explicit 1 (volume source for haul roads) . . 22
3 Source representations for ISCSTM: haul roads, haul truck
loading and dumping—Explicit 2 (area sources for haul roads) ... 24
4 Migrating source representation for dragline (shaded area) and
scraper operation (squares)—ISCST2 and ISCSTM 25
5 Haul road ramps from grade to pit floor 27
6 Locations of monitors at the Cordero mine 34
7 Required source data manipulation for roads and haul truck
loading/dumping 42
LIST OF TABLES
Number
1 Emission factors
2 Emission factor sets
3 Default values for Sets 1 and 2 PM-10 emission factor equations
4 Activity resolution
5 Source representation
6 Area source grid sizes
7 Composite particle size distributions
8 Modeling systems for evaluation
9 Regional background concentrations
10 Monitoring station weights (background adjusted) for composite
performance measures
Page
11
12
14
16
19
20
29
32
41
46
VII
-------
SECTION 1
INTRODUCTION
Section 234(a) of the amended Clean Air Act states the following: "Prior to any
use of the Industrial Source Complex (ISC) Model using AP-42 Compilation of Air
Pollutant Emission Factors to determine the effect on air quality of fugitive paniculate
emissions from surface coal mines, for purposes of new source review or for purposes
of demonstrating compliance with national ambient air quality standards for paniculate
matter applicable to periods of 24 hours or less, under section 110 or parts C or D of
title I of the Clean Air Act, the Administrator shall analyze the accuracy of such mode!
and emission factors and make revisions as may be necessary to eliminate any
significant over-prediction of air quality effect of fugitive paniculate emissions from
such sources. Such revisions shall be completed not later than 3 years after the date
of enactment of the Clean Air Act Amendments of 1990. Until such time as the
Administrator develops a revised model for surface mine fugitive emissions, the State
may use alternative empirical based modeling approaches pursuant to guidelines
issued by the Administrator."
In response to the Clean Air Act mandate, a two-phase program is being
conducted to evaluate the performance of emission factors and dispersion models
applicable to surface coal mining operations. In Phase I, a two-part field study was
performed to compile a comprehensive data base that could be used as a base for the
performance evaluation (Phase II).
The first part of the Phase I study consisted of a field testing program (Muleski,
et a!., 1994) performed in September-October 1992 at a large Western surface coal
mine, under Work Assignments 37 and 55 of Contract No. 68-D2-0123. The study
site was the Cordero mine within Wyoming's Powder River Basin. The majority of this
research was directed toward the validation and improvement of paniculate emission
factors for various mining operations.
-------
The second part of the Phase I study (EPA, 1994), to gather and assemble
monitoring data for dispersion model evaluation, was performed during the May-July
1993 time period at the Cordero mine, under Work Assignment No. 8 of EPA Contract
No. 68-D2-0159. The primary purpose of this effort was to compile concurrent
ambient air quality data, meteorological data, and source activity data collected during
thirty 24-h monitoring periods. This work included the following activities:
• Collection of 24-h air quality data for TSP (paniculate matter captured by the
standard high-volume air samples) and PM-iO (particulate matter nominally
10 microns and less in aerodynamic diameter) from a nine-station monitoring
network distributed in and around the Cordero mine;
• Collection of continuous on-site meteorological data (including temperature,
precipitation, wind speed, and wind direction) both above grade and inside an
active pit within the mine;
• Collection of time-resolved information about mining operations (source activity)
during three observation periods constituting each 24-hr monitoring period.3
• Estimation of hourly emission rates for all significant sources (i.e., traffic on haul
roads and equipment operations associated with topsoil, overburden, and coal
removal) operating during the monitoring periods; and
• Assembly of a comprehensive data base containing all of the above information in
a suitable electronic format.
This protocol has been prepared to define the procedure that will be used (a) to
identify the best-performing model(s) for predicting the impacts of particulate
emissions from surface coal mines and (b) to identify "significant" overprediction, if it
occurs. A "model" refers to the combination of an atmospheric dispersion model and
the required input data on source emissions and meteorology. The models of greatest
interest for predicting air pollutant concentration fields in the vicinity of open pit mines
are the current version and a new version of the Industrial Source Complex Short
aBecause the fixed source activity observation periods corresponded closely to
mine work shifts, the term "shift" is used in this report to characterize observation
periods.
-------
Term (ISCST) dispersion model, in conjunction with existing AP-42 emission factors
and revised factors developed from the Phase I source testing at the Cordero mine.
While Section 234 of the CAA is directly concerned with model overprediction, it
should be noted that model underprediction is also of concern to EPA. On balance, a
model that is unbiased is preferable to one that significantly over- or underpredicts
ambient levels. Clearly, use of an unbiased model minimizes the chance of making
errors in either direction--!.e., inadequate protection against adverse environmental
effects vs. unnecessary and costly control efforts.
During the process of developing this protocol, inputs were solicited and
received from the Wyoming Mining Association and the State of Wyoming regarding
the objectives of the model evaluation and the methodologies that are incorporated
into this protocol.
-------
This page intentionally blank
-------
SECTION 2 •
OVERALL APPROACH TO MODEL PERFORMANCE EVALUATION
This section introduces the components of the candidate modeling systems and
summarizes the methodology for model performance evaluation.
2.1 CANDIDATE MODELING SYSTEM COMPONENTS
Each candidate modeling system will consist of an atmospheric dispersion
model, an emission inventory, and a geometric representation scheme for each source
category. Two dispersion models will be evaluated: ISCST2 and ISCSTM. The latter
is a variation of ISCST2 and contains a new deposition algorithm (DEPST), an
upgraded area source algorithm (AREA-ST), and an added pit retention algorithm.
The emission inventory identifies and locates the emission sources of interest
and assigns an estimated emission rate to each source element. A calculation of the
estimated emission rate for a given source requires data on source activity,
uncontrolled emission factor, and control efficiency. The mathematical expression for
this calculation is as follows:
R = Me (1 - c) 0)
where: R = estimated mass emission rate in the specified particle size range
(mass/time)
M = source activity (activity/time)
e = uncontrolled emission factor in the specified particle size range,
i.e., mass of uncontrolled emissions per unit of source activity
(mass/activity)
c = fractional efficiency of control (dimensionless)
-------
Section 3 provides detailed information about (a) the sources to be included in
the emission inventory, (b) three sets of emission factors to be used for these sources.
(c) source activity representation, (d) control efficiency determination, and
(e) appropriate geometric source representation schemes that are compatible with
ISCST2 and ISCSTM.
Section 4 specifies the modeling systems that will be evaluated against the
observational air quality data base for PM-10 and TSP. These modeling systems
range from "base case" that best represents current practice for assessing surface
coal mine impacts, to systems that incorporate more refinements to the base case
dispersion model, emission factors, source representation, and source activity
resolution.
Section 5 describes the observational data bases generated during the thirty
24-h monitoring periods at the Cordero mine. These include the source activity,
meteorology, and air quality (PM-10 and TSP) data bases that will be used in the
model evaluation. As noted above, the source activity data are incorporated into the
emission inventory, providing the temporal resolution.
2.2 EVALUATION METHODOLOGY
The statistical methodology for model performance evaluation will be applied
separately for PM-10 and TSP. It will consist of two steps, determining the best-
performing model(s) and assessing model overprediction. The procedures for
determination of model performance are presented in detail in Section 6. The
procedures for evaluation of model overprediction are discussed in detail in Section 7.
The first step will be to identify one model (or a group of models) as the best
performing model. This evaluation will be based on pair-wise comparisons of
observed and predicted robust highest concentrations for each monitoring site, i.e., the
concentrations will be paired in space but not in time. Models will be compared using
a composite measure of performance across the nine monitoring stations, with the
stations with the greatest source impact receiving greater weighting in the composite
measure. Relative model performance will be evaluated statistically via a bootstrap
resampling procedure.
-------
The second stage of the model evaluation process will be to determine which
model or models do not significantly overpredict. The evaluation of model
overprediction will consist of three elements: (1) a statistical evaluation of model
overprediction; (2) an historical data review of paniculate concentrations observed in
the Powder River Basin; and (3) a model sensitivity analysis to assess whether the
best performing model is functioning in a reasonable manner for use in regulatory
model applications.
-------
This page intentionally blank
8
-------
SECTION 3
SOURCE REPRESENTATION
Source representation denotes the manner in which (a) the sources identified in
the emissions inventory are spatially and temporally distributed and (b) geometric
forms are used to depict the various sources. Spatial variations in emissions occur
because source locations move as the active mining area migrates; for example, such
would be the case for in-pit operations. On the other hand, temporal variations in
emissions may occur even though the source location is fixed, because of variations in
(a) source activity, (b) emission factor correction parameters, or (c) efficiency of add-
on control measures. For example, permanent haul road emission rates may vary in
time because of variations in traffic volume of haul trucks or in the moisture content of
the road surface material as caused by rainfall or watering for dust control.
This section describes the emission source components of the candidate
modeling systems that will be used as inputs for the dispersion models to be
evaluated. These are (a) emission sources to be included in the inventory,
(b) emission factors, (c) measure of source activity, (d) control efficiency, (e) geometric
representation, (f) release height and initial vertical dispersion, and (g) particle size
distribution.
3.1 EMISSION SOURCES
The following emission sources will be included in the Cordero mine emission
inventory input to the dispersion models for evaluation:
• Haul trucks traveling on unpaved haul roads
• Water trucks traveling on unpaved haul roads (to control dust emissions)
• Light-duty vehicles traveling on unpaved haul roads
• Grader travel on unpaved haul roads (for road maintenance)
• Dragline (bucket dumping—overburden)
-------
• Haul truck loading (with power shovel)
• coal
• overburden
• Haul truck dumping
• coal
• overburden
• Bulldozing (in truck loading area)
• coal
• overburden
• Scraper travel on unpaved surfaces (for topsoil removal and scoria mining)
• Wind erosion of active surface areas
• haul roads
• truck loading areas for coal and overburden
• truck unloading areas for overburden
The above sources were selected based on those categories identified in past
emission inventories of surface coal mines (e.g., Cole et al., 1985). Any additional
sources are considered to be sufficiently insignificant to be neglected in this study.
In addition to the sources within the Cordero mine property, haul trucks
traveling on the main unpaved haul road at the Caballo Rojo mine to the north
constitute a potentially significant source impacting on the air quality monitoring
stations under north wind conditions.
3.2 EMISSION FACTORS
Most of the emission factors that will be used are found in EPA's Compilation of
Air Pollutant Emission Factors, AP-42 (EPA, 1985). However, for the most important
source category (haul trucks and water trucks traveling on haul roads), a new
emission factor equation developed from Cordero source testing data will also be
used, as discussed later in this subsection. Furthermore, for heavy-duty vehicles (haul
trucks and water trucks) and light-duty vehicles traveling on haul roads, a third set of
emission factors will also be used, as derived from adjustment of directly measured
emission rates.
Table 1 summarizes the rationale for use of the three sets of emission factors.
Note that for sources other than heavy-duty and light-duty vehicles traveling on haul
roads, the Set 2 and Set 3 emission factors are identical.
Table 2 compares the set of PM-10 and TSP emission factors from AP-42
Section 8.24 (Western Surface Coal Mining), with the partially revised set of emission
factors, for the sources to be included in the inventory. The set of emission factors
found in Section 8.24 of AP-42 (Set 1) represents current practice in assessing the
impact of surface coal mines. The Set 2 emission factors consist of: a new factor for
the largest source (heavy vehicles traveling on unpaved haul roads); factors currently
10
-------
TABLE 1. EMISSION FACTORS
Terminology
Description
Rationale
Set 1
Set 2
Set3
(differs from Set 2 only for heavy-duty
vehicles [haul trucks and water trucks]
and light-duty vehicles traveling on haul
roads)
Equations found in AP-42 Section 8 24 for
all specified sources, with default values for
all correction parameters
• New equation for haul trucks
• Section 11.2 equations for haul truck
loading and dumping, dragline, light-duty
vehicles, and wind erosion
• Section 8.24 equations for scrapers,
graders and bulldozers
• Default values for all correction
parameters except for haul road surface
silt and moisture content3
• Hourly emission factor values (i.e.,
emission rate per unit of source activity)
for each road segment
• Derived from representative on-site
(Cordero) emission measurements
(uncontrolled) with adjustments for
mitigation due to hourly rainfall and shift-
resolved watering activity
Represents the commonly used predictive
equations for estimating dust emissions
from surface coal mines
Incorporates:
• Recommended improvements to the
predictive equations in Section 8.24
(except for scrapers, graders and
bulldozers), and
• Site-specific values of haul road surface
silt and moisture content in lieu of
default values
Constitutes the most accurate short-term
estimation of dust emissions from haul
roads (the predominant source category) at
Cordero
Representative values of road surface silt and moisture content for coal and overburden haul roads, coal haul ramps and access
roads will be derived from approximately 100 samples collected during the 30 days of ambient monitoring.
-------
TABLE 2. EMISSION FACTOR SETS3
Sources
Bulldozing — coal
— overburden
Draglineb
Graders
Haul trucks traveling on unpaved
roads
Haul truck loading — coal
— overburden
Haul truck unloading — coal
— overburden
Light-duty vehicles traveling on
unpaved roads
Scrapers on unpaved surface travel
Water-trucks traveling on unpaved
roads
Wind erosion
— coal loading areas
and haul roads
— overburden loading areas
and unloading areas
PM-10
14 s1 5/M1 4
075s1 5/M14
0.0016 d° 7/M° 3
0 0306 S2 °
0.0031W3'5
0.089/M0 9
Noned
Noned
Noned
2 2/M43
37x10~6s14W25
Sell
TSP
78 4 s1 2/M1 3
57s12/M13
00021 d1 '/M03
0 040 S2 5
00067w34L02
1 16/M1 2
0037
0027
0002
5.79/M4 °
2.7x10-5s13W24
Not specifically stated, but can use equations for haul
trucks traveling on unpaved roads
Noned
Noned
None, but can use factor for wind
erosion from overburden loading
areas
0087
Set 2
PM-10 TSP
14s15/M14 784s12/M13
075s15/M14 57s12/M13
k(0 0032)(U/5)1 3/(M/2)1 4
0 0306 S2 ° 0.040 S2 5
k (s/3)° 8(M/2)"° 2
k(00032)(U/5)13/(M/2)14
kfOOOSaKU/S)1 ^(M/?)1 4
k(00032)(U/5)1 3/(M/2)1 4
k(0.0032)(U/5)1 3/(M/2)1 4
k(5.9)(s/12)(S/30)(W/3)° 7(w/4)° 5
3 7x10'6s1 4W2 5 2 7x10~5s1 3W2 4
k (s/3)° 8(M/2)'° 2
n
i=1
n
Units
Ib/h
Ib/h
Ib/yd3 (Set 1)
Ib/ton (Set 2)
Ib/VMT
Ib/VMT
Ib/lon
Ib/ton
Ib/ton
Ib/ton
Ib/VMT
Ib/VMT
Ib/ton
Ib/acre/h (Set 1)
g/m2-yd (Set 2)
• Ib/acre/h (Set 1)
g/m2 yd (Set 2)
aSymbols used'
d = drop height (m)
k = correction parameter, as noted in the following footnotes
L = surface silt loading (g/rrr)
M = moisture content (%)
N = numbpr ol disturbances per year (yr~1)
P. = erosion potential corresponding to the observed fastest mile ol wind Irom the ith period
between disturbances (g/m )
P, = 58 (u* - u,')2 + 25 (u' - u,')
where' u' = Iriction velocity (m/s)
U|* = threshold Iriction velocity (m/s)
S = mean vehicle speed (mph)
s = silt content (%)
U = mean wind spppd (mph)
W = me.in vehiclp weight (ton)
w = mean number ol whppte
bFor PM-10, k = 0 35, for TSP, k = 0 74
GFor PM-10, k = 3 4, lor TSP, k = 16
dWill be assumed to be hall ol the TSP factor; based on particle
size data Irom AP-42 Section 1123
eFor PM-10, k = 0 36, lor TSP, k = 0 80.
'For PM 10, k = 0 5, lor TSP. k = 1 0
-------
contained in AP-42 Section 11.2 (Fugitive Dust Sources) for aggregate materials
handling (i.e., dragline operation and truck loading/dumping), light-duty traffic on
unpaved roads, and wind erosion; and factors currently contained in Section 8.24 of
AP-42 for bulldozing and for grader and scraper travel on unpaved surfaces.
The rationale for selecting the Set 2 emission factors for heavy vehicles (haul
trucks and water trucks) on unpaved haul roads, light-duty traffic on unpaved roads,
and scraper travel on unpaved surfaces comes from the results of the fall 1992
emission testing program at the Cordero Mine (Muleski, et al., 1994). Selection of
emission factors from Section 11.2. rather than Section 8.24, for materials handling
operations (dragline operation and truck loading/dumping) and wind erosion reflects
the strengthening of the respective Section 11.2 equations through more recent AP-42
revisions that incorporate the results of additional PM-10 emission testing.
The new (Set 2) emission factor for heavy-duty trucks traveling on unpaved
roads (haul trucks and water trucks), as developed from the 1992 emission testing
program at Cordero (Muleski, et al., 1994), is given by:
where: e = emission factor in Ib/VMT
k = correction parameter (3.4 for PM-10, 16 for TSP)
s = surface material silt content ( percent)
M = surface moisture content ( percent)
Equation 2 was developed for haul trucks and water trucks together, because
both were present on haul roads during testing. Although water trucks are not as
heavy as loaded haul trucks, they are much heavier than any other vehicles traveling
on haul roads. Also, because water is sprayed from the rear of a water truck, its
emissions reflect the before-watering conditions of the road. It should be noted,
moreover, that water trucks constituted only a small fraction of haul road traffic.
Table 3 lists the default values for correction parameters for Set 1 and Set 2
emission factors. Default values will be used for the correction parameters required
for Set 1 and Set 2 emission factors, with one exception. The exception is the use of
haul road moisture and silt content values based upon on-site measurements as
correction parameters for the new emission factor equation for heavy duty haul trucks
traveling on unpaved roads (Equation 2). In the case of the Set 1 emission factors,
the default values used will be the mean values of correction parameters specified in
Section 8.24 of AP-42. In the case of the Set 2 emission factors, default values will
be selected for the emission factor equations in Chapter 11 of AP-42 based on the
following priority: specified mean values; denominators of dimensionless correction
parameter terms in the given emission factor equation; or geometric means of
specified correction parameter ranges.
13
-------
TABLE 3. DEFAULT VALUES FOR SETS 1 AND 2 EMISSION FACTOR EQUATIONS
Source
Bulldo zing-coal
- overburden
Dragline
Graders
Haul trucks (coal, overburden, water) traveling on unpaved
roads
Coal haul ramps
Main coal haul roads
Overburden haul roads
Access roads
Haul truck loading (coal)
Haul truck loading (overburden)
Haul truck unloading (coal, overburden)
Light duty vehicles traveling on unpaved roads
Scrapers on unpaved surface travel
Wind erosion
- coal loading areas
-- overburden loading areas
- haul roads
Caballo Rojo
Default Values
Set 1
s = 8.8%, M = 10.4%
S = 6.9% M = 7.9%
d = 28.1 ft, M - 3.2%
S = 7.1 mph
n
w = 8.1 wheels, L = 40.8 g/rrr
M = 17.8%
NA
NA
M = 1 .2%
s = 16.4%, W = 53.8 tons
NA
NA
NA
(See haul trucks)
Set 2
Same as Set 1
U = 5 mph, M = 2%
Same as Set 1
M = 5.7%, s = 5.57%a
M = 5.5%, s = 2 65%a
M = 6.8%, s = 4.02%a
M = 2.1%, s = 9.18%a
U = 5 mph, M = 2%
U = 5 mph, M = 2%
U = 5 mph, M = 2%
s = 8.4%, S = 30 mph,
W = 3 tons, w = 4 wheels
Same
NA
NA
NA
(See haul trucks)
NOTES:
d = drop height (ft)
L = surface silt loading (g/m )
M = moisture content (%)
s = silt content (%)
S = mean vehicle speed (mph)
U = mean wind speed (mph)
w = mean number of wheels
W = mean vehicle weight (ton)
NA = not applicable, i.e., no default values
aSilt analysis results from the on-site measurements are presented in Appendix F; moisture analysis results are presented in the
Phase I report (EPA, 1994), Table 5-8, pg. 5-33.
-------
The Set 3 emission factors are the same as the Set 2 factors with the exception
of the emission factors for heavy-duty vehicles and light-duty vehicles traveling on haul
roads. The Set 3 emission factors for heavy-duty vehicles (haul trucks and water
trucks) and light-duty vehicles traveling on haul roads were developed from direct
source testing of these sources at the Cordero mine during Phase I. Rather than
deriving emission factor equations for these sources, representative values of the
Phase I source measurements for uncontrolled conditions were used directly in
calculating adjusted hourly emission factors. This was accomplished by multiplying
the representative (geometric mean) uncontrolled emission factors by the fractional
mitigative values that accounted for hourly precipitation and shift-resolved watering
activity. Details about this calculation procedure are provided in Appendix A.
The hourly emission factors based directly on emission measurements are
more reliable than those that could be derived using even the new emission factor
equation for haul roads (Equation 2), because the hourly factors have been adjusted
for the effects of hourly rainfalls. If equation 2 were to be used to obtain calculated
rates of a comparable level of reliability, large numbers of representative road surface
samples would be required to derive highly resolved moisture and silt correction
parameters.
For wind erosion, the Set 3 emission factor provides for hourly calculation of
paniculate emissions (for any hour with winds that exceed the threshold velocity). The
Set 3 emission factor assumes that the full erosion potential of a surface is restored
when it is disturbed by stationary or low-speed equipment operations. Because
moderate-spaced traffic on haul roads releases most of the fines generated by each
vehicle pass, the Set 3 emission factor is multiplied by 0.1 when applied to roads.
3.3 SOURCE ACTIVITY
At a surface coal mine, source activity relates to the movement of vehicles, the
transfer of excavated materials, and the exposure of disturbed surfaces to high winds.
For vehicle traffic, source activity is measured as vehicle-distance traveled. For
material transfer, the activity is simply the quantity transferred. For wind erosion,
activity should be measured in terms of (a) the amount by which the wind speed
exceeds the erosion threshold for the exposed material in question, and (b) the
frequency of mechanical disturbances of the erodible surface.
Although documentation of hourly variation in source activity is usually not
feasible, shift averages can be determined by multiple observations during each shift,
coupled with examination of shift records. This was accomplished as part of the
ambient monitoring program at the Cordero mine, using three observation periods that
corresponded closely to mine work shifts. The source activity data in the
observational data base from the study are detailed further in Section 4.
15
-------
in preparing emission inventories for model evaluation, source activity will be
represented by two levels of resolution—"shift averages for each day" and "60-day
shift averages" calculated from the daily shift values. In the first case, the shift-
average activity levels vary based upon the specific source observations for the
respective source category. In the second case, the daily values for each shift are
averaged for the 60-day period that encompasses the ambient monitoring program.
Table 4 provides the rationale for selecting these two levels of activity.
TABLE 4. ACTIVITY RESOLUTION
Terminology
"Shift averages
for each day"
"60-day shift
averages"
Description
• The activity for each source
category follows a variable
diurnal cycle (3 shifts)
• The cycle for each day is based
on daily Cordero observations of
that same category
• The activity for each source
category follows a fixed diurnal
cycle (3 shifts)
• The 60-day average activity for
each of the three shifts is based
on Cordero observations for that
source category
Rationale
• Constitutes the most time-
resolved representation of
source activity
• Takes into consideration
daily changes in location and
level of activities (e.g., north
versus south pit mining)
• Constitutes potentially more
suitable representation of
source activity when
predicting concentrations that
are not paired in time with
measured values
• Appropriate means for
projecting source activity
cycle that reflects reasonable
use of mining equipment
3.4 CONTROL EFFICIENCY
In the calculation of the emission rates, uncontrolled emissions must be
reduced to account for the effects of road watering and natural mitigation (rainfall). In
the case of road watering, a control efficiency will be assigned to each road segment
where water truck activity was noted during the observation period, except when using
the Set 2 emission factor equation for heavy duty trucks traveling on unpaved haul
roads (Equation 2). In the later case, the effects of road watering (and rainfall) are
reflected in the road surface moisture content as a correction parameter. The control
efficiencies used for road watering will be as follows:
16
-------
Watering control efficiency
Set 1 Set 2 Set 3a
Heavy-duty 50 percent to 60 percent
vehicles
Light-duty vehicles 50 percent 50 percent 60 percent
aSee Appendix A.
bEffect of watering reflected in moisture correction
parameter.
The 50 percent estimate for watering control efficiency is consistent with past
estimates for western surface coal mines. In addition, 50 percent approximates the
average control efficiencies for watering of haul roads found in the 1992 testing
program (Muleski et al., 1994):
Watering control efficiency
Coal haul roads
Overburden haul roads
Both
PM-10
52 percent
55 percent
53 percent
TSP
56 percent
21 percent
52 percent
Note, however that slightly higher efficiencies are obtained when only those controlled
emission tests that were performed within an hour of water application are included;
hence, the 60 percent value for the Set 3 factors. This is discussed further in
Appendix A.
The mitigative effect of rainfall will be assumed to apply only to unpaved haul
roads (heavy duty vehicles, light duty vehicles, graders, and wind erosion) and scraper
travel. In most cases, the effect of rainfall will be taken into account by assuming that
emissions are negligible for any hour with measurable precipitation (precipitation
greater than or equal to 0.01 inch). However, a more complex treatment of rainfall
mitigation was used in the development of the Set 3 emission factors for heavy duty
and light duty vehicles traveling on haul roads as addressed in Appendix A.
No mitigation due to rainfall will be assumed for the dragline or for truck loading
or unloading of bulk materials (coal and overburden), because of the inability of rainfall
to quickly penetrate the bulk materials that are being handled.
17
-------
3.5 GEOMETRIC REPRESENTATION
For ISCST2, roadway emissions are most appropriately represented as a string
of volume sources. The ISC User's Manual (EPA, 1992b) recommends using no
fewer than N/2 volume sources to represent a line source where N is defined as
N _ length of the line source
width of the line source
The ISCST2 accommodates square (N-S)/(E-W) oriented area sources to represent
more diffuse working areas of the mine where mined material is transferred.
Considerable latitude in choosing the size of the area sources is available to the
modeler.
Because ISCSTM allows rectangular area sources of arbitrary orientation,
elongated area sources are appropriate to represent roadway emissions, but a string
of volume sources can also be used as before. Use of elongated area sources for
roads is more convenient because it requires far fewer source elements.
Table 5 lists the geometric representations that will be used for each source
category. In the case of haul roads, two "explicit" representations will be used:
(a) strings of volume sources (ISCST2 and ISCSTM) and (b) elongated area sources
(ISCSTM). For haul truck loading and associated bulldozing activity and wind erosion,
upright square area sources will be used with ISCST2, and tilted rectangular area
sources (i.e., areas rotated in relation to the system of coordinate axes used in the
modeling) will be used with ISCSTM. All other source activities will be represented as
upright square area sources, both for ISCST2 and ISCSTM. Table 6 lists the area
source grid sizes used for each emission source category.
The representations for vehicle travel on haul roads, haul truck loading (coal
and overburden) and associated bulldozing and wind erosion, and haul truck dumping
(coal and overburden) are illustrated in the following figures:
Figure 1—Representations for ISCST2
Figure 2—Representations for ISCSTM with haul roads as volume sources
(Explicit 1)
18
-------
TABLE 5. SOURCE REPRESENTATION
Representation
Source category
Operating characteristics
ISCST2
ISCSTM
Haul roads (haul trucks,
water trucks, light vehicles,
graders and wind erosion)
Fixed routes
Explicit—Volume sources
(nominal 100 ft spacing); see
Figure 1
Explicit 1—Volume sources (nominal 100 ft
spacing); see Figure 2
Explicit 2—Rectangular area sources oriented to
road segment direction; see Figure 3
Haul truck loading and
associated bulldozing and
wind erosion, and haul
truck dumping
Mobile within definable areas that are N-S/E-W square area sources; Fixed sources oriented to bench direction; see
fixed at ends of ramps to haul roads see Figure 1
a,b
Figures 2 or 3
,b,c
Dragline
Mobile within definable area that
migrates from day to day
Migrating N-S/E-W square area Migrating N-S/E-W square area sources; see
sources; see Figure 4 Figure 4
Scraper travel
Wind erosion
Mobile within definable area that
migrates from day to day
Migrating N-S/E-W square area Migrating N-S/E-W square area sources; see
sources; see Figure 4
K
Figure 4
Definable areas of surface distur- - Same as haul truck loading and Same as haul truck loading and haul roads (above)
bance where excavation/transfer and haul roads (above)
traffic are occurring
aFigure 1: a1, a2, b1, b2, c1, c2 = North pit coal loading.
bFigure 1 for ISCST2 and Figures 2 and 3 for ISCSTM:
d = North pit overburden loading.
e = North pit overburden dumping.
f = Coal dumping.
g = South pit overburden loading.
h = South pit coal loading.
i = South pit overburden dumping.
°Figures 2 and 3: a, b, c = North pit coal loading.
kReferenced to A1...G9 and V1...Z4 (see Figure 4) grid areas. (Note that only topsoil and scoria mining operations will be modeled.)
-------
TABLE 6. AREA SOURCE GRID SIZES
o
Model
Source category
Haul roads (haul trucks,
light duty vehicles,
graders, and wind erosion)
Haul truck loading and
associated bulldozing and
wind erosion, and haul
truck dumping
Scraper travel
Dragline
Figure
Figure 3
Figure 1
Figures 2
and 3
Figures 1,2,
and 3
Figure 4
Figure 4
Identifier Operation
Roads A...Z Coal and overburden
haulage, access to work
areas and road maintenance
a1, a2, b1, b2, North pit coal loading
c1,c2
a, b, c North pit coal loading
d North pit overburden loading
e North pit overburden dumping
f Coal dumping
g South pit overburden loading
h South pit coal loading
i South pit overburden dumping
A1...G9 North pit operation
V1...Z4 South pit operation
North pit overburden removal
Size (m) ISCST2
305 m x 305
(max)3
200 x 200 x
200 x 400
200 x 200 x
200 x 200 x
200 x 200 x
200 x 200 x
200 x 200 X
200 x 200 X
305 x 305 x
305 x 305 x
Variable x
ISCSTM
x
x
X
X
X
X
X
X
X
X
X
aEach straight line road segment (30 5 m [100 ft] width) must be broken into rectangular unit areas having a maximum aspect ratio
of 10. Actual road lengths are available in the supporting data files (See Appendix D).
-------
5000
4000
3000
2000
1000
HV 1
O
X -g- 0
cr ~~-~
o
z
-1000
•2000
-3000
-4000
-5000
i i i i i i i i
_
MR/ 6
d[T%
MR' 5 V
* rr \ /
rR V
- b1|_J ""x^
\^^"T^^ * MR/ 4
V^-—""*"
MPI3
£
HVZ*a
^^Qf
.X
/'
h- !
I Mr\l 2
I
— —
1
1
I
1
j
1
• 1
MR/ J /
f
9R5'b,
-
, HV3
! 1 1 1 - 1 1 1 1
Legend
— Volume Source
D Area Source
• Ambient Air Quality
Monitor
Key
al,a2, bl,b2, d,c2,
= North Pit Coal Loading
d = North Pit Overburden
Loading
e = North Pit Overburden
Dumping
f = Coal Dumping
g = South Pit Overburden
Loading
h = South Pit Coal Loading
i = South Pit Overburden
Dumping
A4.1S SEV 0i«mip2fa 0414M
-3000 -2000 -1000 0 1000
EASTING
Figure 1. Source representations for ISCST2: haul roads, haul truck
loading and dumping.
21
-------
5000
4000
3000
2000
1000
HV 1
0
1?°
cc -'
0
-1000
-2000
-3000
-4000
-5000
•3C
i i i i i i i i
MR/ 6
MR! 5 ' \
• y^. ]
^^"*\r
\ y\ ^^—••^
yV-"~""! | MRI4
w- ^ ! *
i
| MR! 3
1 * HV2, 2a
/°'
/
; wfl/2
1
i
]
1 -
]
MflM /
1 1 1 ' • ' ! ' '
)00 -2000 -1000 0 1000
Legend
CHl.D Area Sources
Volume Source
• Ambient Air
Quality Monitor
Key
a, b, c,
= North Pit Coal Loading
d as North Pit Overburden
Loading
e = North Pit Overburden
Dumping
f = Coal Dumping
g = South Pit Overburden
Loading
h = South Pit Coal Loading
I = South Pit Overburden
Dumping
•4-1S 3EV tfuirTuf>2c 0414M
EASTING
(m)
Figure 2. Source representations for ISCSTM: haul roads, haul truck
loading and dumping—Explicit 1 (volume source for haul roads).
22
-------
Figure 3—Representations for ISCSTM with haul roads as elongated area
sources (Explicit 2)
For dragline and scraper operations, the source locations are shown in
Figure 4. The dragline is represented by one or more squares of varying size and
location for each day of operation. These squares lie within the boundary of the
irregularly shaped area (shaded) shown along the northwest portion of the north pit.
Scraper operations are represented by the square areas in the figure. Because the
State of Wyoming coordinate system (1,000 x 1,000 ft) was used to document
dragline and scraper activity during the monitoring program, a matching metric area
source grid size (305 m x 305 m) was defined.b The north pit grid system is
assigned letters A...G for the columns and numbers 1...9 for the rows. The south pit
grid system is assigned letters V...Z for the columns and 1...4 for the rows. The
specific grids that were active on any given day are contained in the observational
data bases.
For days when the threshold velocity for wind erosion is exceeded, wind erosion
emissions will be added to the grids where truck loading is occurring and to the active
haul roads. The traffic in these areas generates pulverized surface material that is far
more erodible than exposed surface material containing significant nonerodible
fractions (particles larger than about a centimeter in diameter) in areas undisturbed by
traffic. With regard to hourly wind erosion of the loading areas, it is assumed that
10 percent of an 80,000 m2 coal loading area is disturbed in any hour and 15 percent
of a 40,000 m2 overburden loading or unloading area is disturbed in any hour.
Because the exact location of an hourly activity within a loading area is unknown, the
emissions are assumed to be equally distributed over the entire area.
Not shown in Figures 1, 2, and 3 are the volume sources or elongated area
sources that will be used to represent the Caballo Rojo haul road to the north of the
Cordero property. The emission rate from this potentially significant source will be
calculated using the same techniques as applied to the Cordero haul roads. Daily
average source activity data for the Caballo Rojo haul road will be used for this
purpose, as described in Appendix B.
bThe origin of the metric coordinate system in terms of the Wyoming coordinate
system is as follows: 454,000 ft easting; 1,226,000 ft northing.
23
-------
5000
4000
3000
2000
1000
HVI
C
2
£ E °
cc — •
o
-1000
-2000
-3000
-4000
-5000
-3C
i i i i
MRI 6
MRI 5
• X
<^
VXx^l "
-
-
_
MRI t /
gQ-jl
h
1
P
x
X
•x^
.-X
s
\ i I
Vvle ~
7
MRI 4
MRI 3
* HV2, 2a
^s>^
>' '
/
MRI 2
i
-
i 1 I ' • 1 ' ' '
00 -2000 -1000
0 1000
Legend
CZD.D Area Sources
— — Elongated Area
Source
• Ambient Air
Quality Monitor
Key
a, b, c,
= North Pit Coal Loading
d = North Pit Overburden
Loading
e = North Pit Overburden
Dumping
f SB Coal Dumping
g «s South Pit Overburden
Loading
h = South Pit Coal Loading
i = South Pit Overburden
Dumping
W-1S SEV tfi*m*>1D WMW
EASTING
(m)
Figure 3. Source representations for ISCSTM: haul roads, haul truck loading and
dumping—Explicit 2 (area sources for haul roads).
24
-------
5000
4000
3000
2000
1000
HV 1
• •
o
tr
O
-1000
-2000
-3000
-4000
-5000 -
MR! 6
MRI4
MRI2
HV3.
I I
HV2,2a
I
-3000
-2000
1000
Legend
Dragline
Activity
Scraper
Activity
• Ambient
Air Quality
Monitor
•1000 0
EASTING
(m)
Figure 4. Migrating source representation for dragline (shaded area) and scraper
operation (squares)—ISCST2 and ISCSTM.
25
-------
3.6 RELEASE HEIGHT AND INITIAL VERTICAL DISPERSION
With regard to source elevation, only coal loading and coal dozing occurred at
the full pit depth of approximately 50 m during the Gordero monitoring program. The
haul road ramps that extended from grade to the pit floor will be assigned a source
reference height of 25 m. These ramps are identified by the darkened road segments
shown in Figure 5. All other operations will be assumed to occur at grade.
In using volume sources to represent haul roads, a release height of 2 m and
an initial vertical dispersion term (azo) of 3 m will be used. Both values are based on
results from the 1992 source testing program at the Cordero mine. The same values
are applicable to area sources.
A release height of 2 m approximates the level in the dust plume that equally
divides the mass flux. The 1992 source testing program (Muleski et al., 1994) found
that the maximum particulate matter concentration in the profiled haul road dust
plumes typically occurred at a height of approximately 1.5 m. However, the mass flux
(i.e., the product of concentration and wind speed) occurred at a height of
approximately 2 m.
In general, the receptors of interest in the evaluation lie far enough from the
volume sources such that adjustments to the release height have only a slight effect
on the resulting concentration estimates. For example, changing the release height by
a factor of two causes no more than a 3 percent change in modeled concentrations
(using ISCST2) at the typical source-monitor distances.
The initial vertical dispersion (azo) of 3 m was estimated using guidance
contained in the ISC2 user's guide. The guide suggests setting a20 equal to the
height of the source divided by 2.15. The 1992 source testing program
(Muleski et al., 1994) found 7 m to be a reasonable estimate of the height of haul road
emission plumes at a distance of 5 m downwind from the edge of the road.
As was the case for the release height, changes in o20 have only a slight effect
on the modeled concentrations for the receptor locations of interest. For example,
doubling the value of ozo reduces the predicted concentrations by no more than
26
-------
5000
4000
3000
2000
1000
HV 1
• •*—
0
z
x?o
CC ""
o
z
-1000
-2000
-3000
-4000
-5000
i i i i i i i i
_
MR! 6
r^
MRIS \
. ' jy
\j^3 M«M
-
MRI3
9 HV2, 2a
^
-
.
i MR! 2
2^, ramp
down from
grade to coal
loading areas
-
MRI1 /
J
— ^
i i i i HV2 . i i i
-3000 -2000 -1000 0 1000
EASTING
(m)
Figure 5. Haul road ramps from grade to pit floor.
27
-------
10 percent (using ISCST2) at the typical source-monitor distances for the Cordero
mine.
3.7 PARTICLE SIZE DISTRIBUTION
The particle size distribution of source emissions is required to develop the
necessary inputs to the deposition algorithms within both ISCST2 and ISCSTM. In the
case of ISCST2, the inputs are expressed in terms of the gravitational settling velocity
distribution and the surface reflection coefficient distribution, both of which can be
calculated from the particle size distribution.
The determination of the particle size distribution that will be used for mining
source emissions was based entirely on particle sizing data collected during the
Phase I source testing program (Muleski et al., 1994), as discussed in Appendix C. A
total of four tests for coal haul roads and one test for an overburden haul road were
conducted during the course of the field exercise. For each of the five particle sizing
tests, cyclone/cascade impactor combinations were operated at 1- and 3-m heights
(and at a nominal distance of 5 m from the roadway), providing a total of 10 measured
aerodynamic particle size distributions.
The 10 measured particle size distributions were (geometrically) averaged to
develop the composite aerodynamic particle size distribution, given in Table 7. This
composite distribution will be used to characterize emissions from all modeled
sources, because (a) haul road emissions account for more than half of the total
emissions and (b) various categories of fugitive dust sources have been shown to
exhibit similar particle size profiles, as indicated by the emission factor data presented
in AP-42 Section 11.2. Note that because the composite distribution is expressed in
terms of aerodynamic particle size, unit density (1 g/cm3) can be assigned to all
particle size fractions.
Equations (1-54) and (1-55) in Volume 2 of the ISC2 User's Guide will be used
to calculate mass median diameters and settling velocities for each particle size
subrange. Reflection coefficients will be obtained from Figure 1-7 (ISC2 User's Guide)
and the mass median diameter for each subrange.
28
-------
TABLE 7. COMPOSITE PARTICLE SIZE DISTRIBUTIONS
Particle diameter (|imA)
32a
25
20
15
10
5
2.5
1
0
Wt. % < stated
diameter
TSP PM-.10
100 - ^
61 - J
41 - 1
29 J
21 100 ^
14 67 J
9 43 1
4 19 )
0 0
Mass fraction
TSP PM-10
.39
]
> .20
J
.12
}
> .08
J
.07 .33
}
> .05 .24
J
.05 .24
}
> .04 .19
J
1.0 1.0
aThe geometric mean of the 20- to 50-(imA range usually associated with the
outpoint of the standard high-volume samples.
29
-------
30
-------
SECTION 4 •
MODELING SYSTEMS FOR EVALUATION
This section describes: (a) the modeling systems (groups of model
components) that have been selected for performance evaluation; and (b) the rationale
for selecting the sequence of modeling systems.
Each modeling system to be evaluated consists of:
• an atmospheric dispersion model;
• a set of fugitive dust emission factors;
• a set of source locations and activity levels; and
• a geometric method of representation for each source.
Table 8 shows the sequence of model "runs" proposed for evaluation of the
various modeling systems. The emission factors (Sets 1, 2, and 3), activity levels
(60-day average shift, and daily shift), and geometric methods of representation
(explicit, explicit-1, and explicit-2) identified in Table 8 have been presented and
discussed in detail in Section 3. Each run builds upon the base case by utilizing the
updated dispersion model and emission factors, more refined source activity levels,
and more refined source representations. The runs progress from the base case
using ISCST2 and the existing emission factors, 60-day source activity resolution and
volume source representation for haul roads to the most updated approaches using
ISCSTM, the updated emission factors and two alternatives for haul road source
representation.
31
-------
TABLE 8. MODELING SYSTEMS FOR EVALUATION
CO
ro
Dispersion
Run model
1 ISCST2
2 ISCST2
3 ISCSTM
4 ISCSTM
5 ISCSTM
6 ISCSTM
7 ISCSTM
8 ISCSTM
Emissior
Vehicles traveling
on haul roadsb
Set 1
SetS
Set 2
Set 2
Set 3
Set 2
Set 2
Set3
i factors8
Other sources
Set 1
Set 2
Set 2
Set 2
Set 2
Set 2
Set 2
Set 2
Activity resolution0
60-day average shift
values
Daily shift values
60-day average shift
values
Daily shift values
Daily shift values
60-day average shift
values
Daily shift values
Daily shift values
Source
representation41
Explicit
Explicit
Explicit 1
Explicit 1
Explicit 1
Explicit 2
Explicit 2
Explicit 2
1 Effect of improved ...
Base case
Emission factors
(measurement based)
Dispersion model,
emission factors
Dispersion model,
emission factors
Dispersion model,
emission factors
(measurement based)
Dispersion model,
emission factors, source
representation6
Dispersion model,
emission factors
Dispersion model,
emission factors
(measurement based)
aRefer to Tables 1 and 2.
bHaul trucks, water trucks, and light-duty vehicles.
cRefer to Table 3.
dRefer to Table 4.
"While the area source representation for haul roads may improve the ease of modeling, the effect on model performance is not known.
-------
SECTION 5
OBSERVATIONAL DATA BASES
The observational data bases are those that were generated from the intensive
air quality monitoring study conducted at the Cordero surface coal mine. The
monitoring program encompassed thirty 24-h periods (midnight to midnight) from
May 19 through July 18, 1993. Monitoring was conducted on an every-other-day
basis. Air quality was measured at a nine-station network as shown in Figure 6. The
data bases are computer files generated during preparation of the final report for the
Phase I study (EPA, 1994). A listing and brief description of these files are presented
in Appendix D.
The observational data bases were specifically developed for use in evaluating
model performance. The monitored parameters fall into three categories:
Source activity (mostly shift-resolved data),
• Meteorology (hourly data), and
• Air quality (24-h data).
5.1 SOURCE ACTIVITY
Throughout the monitoring program, the field crew collected process information
about the mining operations that were to be included in the emission inventory input
for model evaluation. Specifically, shift-resolved activity data were obtained for the
following operations:
• Haul trucks traveling on unpaved haul roads-vehicle counts
• Water trucks traveling on unpaved haul roads-vehicle counts
Light-duty vehicles traveling on unpaved haul roads-vehicle counts
• Dragline (bucket dumping—overburden)-location and cycle time
33
-------
Entrance ROM) to
- To Wvwrig SB 39
2 (2 measurements)
Figure 6. Locations of monitors at the Cordero mine.
34
-------
• Haul truck loading (with power shovel)-location
— coal
— overburden
Haul truck dumping-location
— coal
— overburden
• Scraper travel on unpaved surfaces (for topsoil removal and scoria mining)--
location
• Grader travel on unpaved surfaces (for road maintenances-location
MRI referenced its source activity observations to three periods. To the extent
practical, these periods corresponded to work shifts, as follows:
Observation
period Description
0 Period from midnight to start of day shift (6 or 7 a.m.). This period
incorporates part of the preceding day's evening shift.
1 Entire day shift (either 10- or 12-h) for the day that sampling occurred.
2 Period from start of evening shift (either 4 or 7 p.m.) to midnight stop
time for the air monitors.
In this document, the term "shift" is used to characterize the observation periods.
A grid scheme based on the Wyoming state coordinate system was used in this
study to locate emission sources within the mine, as described earlier in Section 3.5.
The major roads were stylized and segmented using aerial photographs, and the
endpoints and length of each road segment were identified. Scraper travel associated
with either topsoil removal or scoria mining was assigned to the appropriate
1,000- x 1,000-ft grid cell. The dragline location during each work shift was
referenced to the nearest quadrant of a 1,000- x 1,000-ft grid cell. The locating
coordinates of the area designations for coal loading and unloading, overburden
loading and unloading, dragline activity and scraper activity are provided in the data
file AREAS.LOC.
35
-------
Four files of source activity data were generated:
• 24HRVPH contains hourly numbers of haul trucks, water trucks, and other
vehicles passing over each of twelve road segments, for each of the thirty
24-h monitoring periods. [The "hourly" numbers represent shift values except
when it was noted that activity had ceased for part or all of a shift.]
• DRAGLINE.DAT contains (a) the number of dragline load cycles and (b) the
grid location of drop, for each shift of the thirty 24-h monitoring periods.
• SCRAPER.DAT contains the location of scraper operations by grid, for each
shift of the thirty 24-h monitoring periods.
4 GRADER.LOG contains a listing of the roads on which graders were
observed during any shift on any day.0
A separate series of files (ROADA, ROADB, ROADC,...etc., one for each road
segment) defines one or more line segments for each road and gives the x-y
coordinates (Wyoming system) for volume sources used to represent the road.
The amount of coal and overburden loaded into and dumped by haul trucks in a
particular area will be set equal to the amount transported by haul trucks over the road
that serves that area. Additional information about material flow balance calculations
is provided in Appendix D. The source activity for bulldozing (hours of operation) will
be determined by assigning one bulldozer to each power shovel used for haul truck
loading.
The source activity for wind erosion (Set 2 emission factor) will be obtained
from on-site hourly wind speed data coupled with a default value for a threshold wind
velocity defined as a fastest mile of 27 mph at a reference height of 10 m above the
surface. This value was derived from the typical mode (1 mm) of the size distribution
of surface samples collected for silt analysis. The relationship between mode and
threshold wind velocity is described in Section 11.2 of AP-42 (EPA, 1985). Assuming
CNOTE: This data file is a new file not included in the original Phase I report data
files; the information was taken from the field data sheets from Phase I.
36
-------
that the ratio of the fastest mile to the hourly mean wind speed is 1.2; an hourly mean
windspeed of 23 mph will be assumed to produce a fastest mile of 27 mph.
Throughout the 1993 ambient air quality monitoring program, samples of haul
road surface material were collected. During each day when ambient air monitoring
occurred, "moisture-tracking" samples were collected from a representative haul road
in use on that day. These samples consisted of approximately 10 incremental sub-
samples of the surface material. Each subsample was collected by broom sweeping a
randomly selected 10- x 10-in area on the road. Sampling was repeated every 15 to
20 min, so that a time profile of the surface moisture content could be obtained.
Water truck passes were noted, and sampling continued over at least one watering
cycle.
During "off days," i.e., days when monitors were not operated, standard
material samples of road surface material were collected by broom sweeping a 10-in
strip across the full width of the road. Samples were split as necessary to an
appropriate size (1 to 2 Ib). Roads were selected for sampling based on the current
level of usage.
Both types of samples underwent surface moisture analysis by measuring
weight loss upon oven drying. The results of the moisture analyses were presented in
the Phase I Report, Table 5-8, pg. 5-33 (U. S. EPA, 1994). After drying, sets of
"moisture tracking" samples were combined, and all sets were archived for further
analysis. Subsequently, the samples from the monitoring days were analyzed for silt
content by dry sieving (according to the procedures specified in AP-42). The results
of the silt analyses are presented in Appendix F.
5.2 METEOROLOGY
Data characteristics of regional meteorology were collected at Site HV-1. This
site was already equipped with a 10-m tower and associated meteorological
instruments that meet the criteria specified by EPA (1987). The station collected data
that are directly applicable to model implementation. Specified parameters monitored
at the station included wind speed and direction, standard deviation of wind direction
(oe), temperature, and precipitation. The hourly meteorological data are contained
within one file, IMLMET.DAT.
37
-------
The following approach to missing/poor quality meteorological data will be used.
Data from the Caballo Rojo mine (on adjacent property to the north) will be substituted
for a missing period if Caballo Rojo data are available. Otherwise data from the next
closest meteorological station will be used.
The ISC model also requires input data on mixing height and atmospheric
stability. Because plume dispersion from the modeled (ground-level) sources should
not be influenced significantly by reflection from the top of the mixing layer, a default
value of 3,000 m will be used for the mixing height. Calculation of hourly values of
atmospheric stability will be based on the "buffered sigma sub-theta" approach, as
described in "On-Site Meteorological program Guidance for Regulatory Modeling
Applications" (EPA, 1987). The buffering refers to the restriction of not allowing the
stability to change by more than one class from one hour to the next.
5.3 AIR QUALITY
Ambient air quality monitors for both TSP and PM-10 were installed at nine
permanent monitoring sites in and around the Cordero surface coal mine. The
locations of the nine primary ambient air monitoring sites are shown in Figure 6 along
with the types of monitors used at each site. All of the sites but one are on Cordero
property, and most lie within the permit boundary for mining activity. The "HV" sites
were those already operated by the Cordero Mine; HV-1 had been sited in an area
generally suitable for measurement of background concentration.
Each station was equipped with an elevated platform and sufficient electric
power to support at least one standard high volume sampler for TSP and one PM-10
reference sampler equipped with an inlet manufactured by Wedding and Associates.
Collocated PM-10 and TSP samplers were installed at one site (HV-2), bringing the
total samplers to ten PM-10 and ten TSP instruments. An additional continuous
monitoring instrument was also added at one of the stations (MRI-6) to provide
supplemental data on time-resolved (i.e., hourly) PM-10 concentration. The ambient
air quality monitors used in this study were operated, maintained, and calibrated in a
manner consistent with guidelines established by EPA (1977).
38
-------
The 24-h air quality data for PM-10 and TSP are contained within 30 files, one
for each of the ten pairs of PM-10/TSP samplers and each month of operation
(May-July). The files are named HV1.MAY, HV1.JUN, .. . MRI6.JUN, MRI6.JULY.
If air quality data (PM-10 or TSP) are not available at a monitoring site, that site
will be removed from the model comparison for that sampling day. Substitutions will
not be made for missing air quality data.
5.4 BACKGROUND AIR QUALITY
A critical step in this process will be the estimation of background air quality
levels, which must be subtracted from observed concentrations before comparing with
model-predicted values. A background concentration is needed for each monitored air
quality parameter (PM-10 and TSP) for each monitoring period.
There are three components to air quality levels at the monitoring sites:
1. Impacts from sources at the mine and from sources beyond the mine
property boundaries that are being modeled;
2. Impacts from sources at the mine which are not being modeled; and
3. "Regional" background: contributions from airborne particles that are
incorporated broadly in the air mass covering the region or are transported
into the region as a broad, diffuse plume from a far-distant source.
The objective is to modify the observed air quality data so that they reflect only
component 1 above, allowing a true evaluation of model performance.
A procedure for estimating regional background concentration was developed
and applied to the air quality data base. This procedure focuses on evaluation of the
lowest measured concentration for each day to see if it meets the necessary
acceptance criteria. If the lowest measured concentration meets the acceptance
criteria, it is used as the background concentration. If the lowest measured
concentration does not meet the acceptance criteria, the background concentration is
estimated. The procedure is described in Appendix E.
39
-------
The resulting regional background concentrations are shown in Table 9. As
indicated, most of the values shown correspond to the lowest measured
concentrations for the days of interest. The average ratio of PM-10 to TSP
concentration (0.66) is consistent with the findings of the Phase I study (EPA, 1994).
Figure 5-8 of that report is a plot of PM-10/TSP ratio versus TSP concentration; it
shows that as the TSP concentration decreases, its fractional PM-10 component
increases. A PM-10/TSP ratio of 0.66 corresponds to a TSP concentration of
approximately 10 (ig/m3.
5.5 RUNSTREAM PREPARATION
Runstreams for the different modeling systems listed earlier in Table 8 have a
common "ancestry." All runs may be viewed as modifications of files of source activity
and meteorological data compiled during the 1993 field study (EPA, 1994). Figure 7
illustrates how the different runstreams will be derived. Although Figure 7 addresses
only roads and truck/loading/dumping operations, the approach is analogous for all
sources.
As illustrated in Figure 7, for roadway and truck/shovel operations, the file
24HRVPH contains hourly information on the number of haul trucks, water trucks and
the vehicle passes on each of 12 roads at the mine. A program, "Program 1" of
Figure 7, will take this source activity information and combine it with emission factors
to develop an hourly emission inventory for all vehicle traffic on roads as well as truck
loading and dumping operations.
As Figure 7 shows, the next step makes use of "Program 2," which will combine
the hourly emission inventory with geographic information to produce ISCST2
runstreams with the specified temporal and spatial resolution. The final step in
generating runstreams relies on temporal averaging of the runstreams already
prepared. This is the goal of "Program 3" in Figure 7. Program 3 will average the
daily shift values of emission rates to produce a "typical" shift-based profile of
emissions, i.e., the 60-day average shift value for the monitoring period.
40
-------
TABLE 9. REGIONAL BACKGROUND CONCENTRATIONS
Background concentration (u.g/m3)
Date
5/19
5/21
5/23
5/25
5/27
5/29
5/31
6/2
6/4
6/6
6/8
6/10
6/12
6/14
6/16
6/18
6/22
6/24
6/26
6/28
6/30
7/2
7/4
7/6
7/8
7/10
7/12
7/14
7/16
7/18
PM-10
6a
8.60
5.45
6.17
9.92
9.09
7.32
4.33
4.25
4.96
6.53
8.26
9.34
10.92
6
4.07
9
5.07
11.04
12.91
6
10
5.86
5.06
7.11
9.96
15.34
9.59
6
7.50
TSP
8
16
7.90
7.87
31.25
8
13.21
8
5.23
7.79
6.27
9.98
20.28
21.75
8
4.07b
16.78
8.35
22
34.75
8
21.81
7.89
8
17.38
16.87
29
17.01
8
8
aBold italics indicates that the value is estimated (to the nearest microgram).
kTSP value set equal to PM-10 value, because lowest measured TSP concentration
(2.51 ng/m3)was more than 1 ng/m below the background concentration for PM-10.
41
-------
24HR\
24HRVPH
PROGRAM
1
Constructs daily
shift-resolved
emission rates
from emission
factors
and source
activity info.
C
Selected A
emission I
factors )
Model runs
1 to 8
PROGRAM
2
Generates
runstreams
with
time/space
resolution
PROGRAM
3
Agglomerates
emission rates
over time and
space
Figure 7. Required source data manipulation for roads and haul truck
loading/dumping.
42
-------
SECTION 6
DETERMINATION OF BEST-PERFORMING MODEL
The first stage of the model evaluation process will be to identify one model (or a
group of models) as the best performing model through a statistical evaluation
procedure that will compare monitored and modeled ambient air quality levels of both
TSP and PM-10 using procedures based on the methodology introduced by Cox and
Tikvart (1990) and later established as a protocol by EPA (1992a) for determining the
best-performing model. This procedure will be implemented via a model evaluation
software package recently developed by EPA (1993). The procedure is described in
more detail in the four subsections below. The first subsection provides an overview of
the evaluation strategy and the rationale for that strategy. The second and third
subsections provide details on the test statistic and performance measures that will be
used in the evaluation. The final subsection describes the model comparison
procedures.
6.1 STRATEGY FOR IDENTIFYING BEST-PERFORMING MODEL(S)
In defining the best-performing model(s), separate and distinct analyses will be
conducted for TSP and PM-10, and best-performing model(s) will be selected for each
pollutant. For each of the two pollutant-specific analyses, a composite performance
measure will be developed using the weighting scheme outlined in Section 6.3.
The decision to conduct separate analyses for PM-10 and TSP is based on both
program and analytical considerations. First, because model applications for PM-10
analyses are much more widespread than those for TSP analyses, there is substantial
interest in how the performance of the different models compares specifically for PM-10.
This interest provided the impetus for a separate PM-10 analysis. More importantly, the
different model scenarios handle both emission generation and particle deposition
differently. These differences may produce effects in the relative performance of the
43
-------
models that differ for PM-10 and TSP that would be masked by a composite analysis.
This potential adverse consequence outweighs possible gains in statistical efficiency that
could be gained by combining the PM-10 and TSP data for analyses.
The primary objective in developing monitor-specific weights for the composite
performance measure was to place greater weight on those monitors that were most
impacted by the emission sources contained in the model assessment. Development of
the weights involved two elements. First, personnel experienced in fugitive emission
dispersion modeling, who had a thorough understanding of the Cordero site, examined
source/receptor geometries at the site to determine whether specific monitoring sites
were appropriately sited for source impact under observed wind patterns. Then,
average observed concentrations at each monitoring site were used as an indicator of
the impact of modeled sources on the site.
6.2 TEST STATISTIC
Because model performance at maximum concentrations is a primary concern,
the test statistic that will be used for these analyses is a robust extreme value estimator
called the robust highest concentration (RHC) (Cox and Tikvart, 1990). The RHC is
preferred for this analysis because it is stable and because the distribution of RHCs
obtained via the bootstrap procedure described in Section 6.4 is not artificially bound by
the highest observation in the sample. The RHC is based on a tail exponential fit to the
upper end of the distribution of ambient concentrations using robust estimates of
percentiles calculated from rank order statistics and is calculated as follows:
RHC = X(R) + [X(R) - X(R)]logf3R2" 1 1 (4)
where: RHC = robust highest concentration
X = mean of the R - 1 largest values
X(R) = Rth largest value
For these analyses, R will be set equal to 8. This value was selected based on a
review of the calculated RHCs for background-corrected observations of both TSP and
PM-10 at the nine monitoring sites. For the different sites, the RHC generally stabilized
at R values between 6 and 11. The stable range for all sites for both TSP and PM-10
always included eight observations. Six or seven observations also could reasonably
44
-------
have been used, but because some of the analyses will be based on observations from
a single sampling station, eight observations are expected to give more stable estimates
for the bootstrap analyses than would have been obtained from six or seven
observations.
6.3 PERFORMANCE MEASURES
The foundation of the performance measures that will be used to compare the
different models is the absolute fractional bias (AFB) of the RHC obtained from the
modeled data (RHCM) relative to the RHC obtained from the monitored data (RHCO).
The AFB is calculated as:
AFB = 2
RHC0 - RHC
M
RHC0 *RHC
M
(5)
The AFB will be calculated for each of the nine monitoring sites separately for
TSP and PM-10, and a composite performance measure (CPM) based on the nine
stations will be calculated for each pollutant as follows:
9
CPM(k) = £ Wj AFB| (6)
1=1
The weights used for the composite performance measure were developed via a
combination of an engineering analysis of source/receptor geometries and calculated
estimates of source impact using observed concentrations at the monitoring sites.
Based on the review of the monitor locations, observed concentration levels and
associated wind vectors, and potential for impact of sources external to the modeling
framework, two monitoring sites (HV1 and HV2) were assigned weights of zero. HV1
was assigned zero weight because its location provides little potential for source impact
and it acts as background on most days. HV2 was assigned zero weight because it has
generally low concentrations and does have the potential to be impacted by diesel
locomotives that are not a part of the modeling framework.
45
-------
The remaining seven monitoring stations were assigned weights (w,) of the form:
w. = 1—
' 7 _ (7)
where: C, = mean background-corrected concentration over the
30-day monitoring period at station i
Separate sets of weights were calculated for TSP and PM-10 as shown in Table 10.
TABLE 10. MONITORING STATION WEIGHTS (BACKGROUND
ADJUSTED) FOR COMPOSITE PERFORMANCE MEASURES3
Monitoring
station
HV1
HV2
HV3
MRI1
MRI2
MRI3
MRI4
MRI5
MRI6
TSP
q
3.88
8.49
2.63
9.31
22.3
38.8
12.7
17.1
27.6
wi
0
0
0.020
0.071
0.171
0.298
0.097
0.131
0.211
PM-10
c,
0.45
1.42
0.99
3.65
3.45
4.73
2.38
4.00
6.31
w,
0
0
0.039
0.143
0.135
0.185
0.093
0.157
0.247
aThese weights are preliminary estimates calculated using the minimum daily
concentrations as background. When final background concentrations are
determined, average concentrations will be recalculated and weights may change
slightly.
Because the purpose of the analysis is to contrast the performance among the
possible models, the two composite performance measures (one for TSP and one for
PM-10) will be used to calculate differences in performance between pairs of models.
These differences in performance between models, called model comparison measures,
are calculated as:
MCM(AiB)(k) = CPMA(k) - CPMB(k) (8)
where: MCMA B(k) = model comparison measure for Model A versus B for
pollutant k
46
-------
CPMA(k) = composite performance measure for Model A and
pollutant k
CPMB(k) = composite performance measure for Model B and
pollutant k
If MCMA B(k) is negative, then Model A is "better" than Model B; if it is positive, Model B
is "better" than Model A for pollutant k. However, because MCMA B(k) is a random
variable subject to sampling variation, the relative performance of two models must be
evaluated statistically as described in the subsection below.
6.4 MODEL COMPARISON PROTOCOL
Because of the inherent sampling variability associated with calculating RHCO
and RHCM, MCMA B(k) may be nonzero even if Models A and B perform equally well.
Typically, statistical procedures use the standard error of MCMA B(k) to determine
whether these nonzero estimates are statistically different from zero. Because
MCMA B(k) is obtained via a complicated calculation procedure and because the
underlying sampling distributions of the observed and modeled ambient concentrations
are not fully characterized, its variance and standard error cannot be readily computed
analytically. Consequently, a bootstrap resampling procedure will be used to calculate
these standard errors. Because monitoring was conducted every second day during the
program, each monitoring day can reasonably be assumed to be independent of the
other monitoring days. A total of 1,500 bootstrap samples of size 30 will be selected by
sampling with replacement from the 30 monitoring days. For each bootstrap sample,
MCMA B(k) will be calculated for each pair of models and each of the two pollutant
specific performance measures. Because eight models are being evaluated, a total of
28 paired comparisons will be generated for each of the two composite performance
measures. The standard error of MCMA B(k), which is designated as (seABk), is simply
the standard deviation of the 1,500 bootstrap samples.
For composite performance measure k, the performance of Models A and B are
deemed to be different if the 90 percent confidence interval for MCMA B(k) does not
span zero. If the entire interval is less than zero, then Model A performs better than
Model B. If the entire interval is greater than zero, then Model B performs better than
Model A. If the interval spans zero, the performance of the models is deemed to be
equivalent. The 90 percent confidence interval was chosen as a reasonable
compromise between achieving an acceptable performance measure specific Type I
error rate (a=0.1) and being able to detect differences in models with relatively small
47
-------
sample sizes and the potential of substantial sampling variability. To provide this
performance measure Type I error rate, the 90 percent confidence limits (Cl) for a
specific pair of models for the composite performance measure for pollutant k is
calculated as:
90% CI{MCMAiB(k)} =.ckseABk 0)
where the ck are obtained from the simultaneous confidence intervals for the 28 paired
comparisons for each composite performance measure i using the procedure outlined
below.
The method of Cleveland and McGill (1984) will be used to calculate ck. For
each composite performance measure k, this method creates a 28-dimensional
rectangular hypersolid centered at the 28-tuple MCMA B(k) for the 28 combinations of A
and B obtained from the actual data. The length of the sides is 2ck • seABk.
Specifically, ck is found so that for 90 percent of the 1,500 bootstrap 28 element vectors,
- uk
(10)
SABk
where: AAB = model comparison difference measure for model pair A,B
AAB: = model comparison difference measure for model pair A,B and
bootstrap replication j
SABk = standard deviation for all the AABj values
For this analysis, we find ck for each performance measure.
jMCMAtB(k) -MCMA[Bj(k)|
-------
1. For each TSP and PM10 separately, calculate the RHC for the observed and
predicted concentrations paired by space over all data. Calculate the AFB of
the RHC with confidence limits for each monitoring site.
2. Calculate the CPM for each model for TSP and PM-10. The smaller the
CPM, the better the overall performance of the model.
3. Calculate the MCM with confidence limits for each model pair for the
composite performance measure for TSP and PM-10.
4. Tabulate the overall performance measure results and significance of the
results.
These four steps generate two primary measures that can be used to select a set
of one or more best performing models in a two step process. (Note that distinct
analyses will be conducted for each of the two pollutants [TSP and PM-10]). First, the
CPM values calculated for each model provide point estimates of model performance
that can be used to order the eight possible models. Define CPM,j, as the order
statistics for the measured CPM values where CPM,^ < CPM^) < . . . < CPM/8< and let
ModeU be the model associated with CPM^. The first step in selecting the set of best
performing models provides an ordering of the models from Model^ (the "best" model in
some sense) to Model(8) (the "worst" model). However, for any pair of models ModeU
and ModeU with kj, the performance of the two models may be indistinguishable
statistically. Consequently, the second stage of the process will be to use the MCM for
each pair of models to identify those for which the performance can be distinguished
statistically. (Those model pairs are the ones for which the confidence interval of the
MCM does not span 0.)
The following plots will be generated:
1. AFB with confidence limits for each model as a function of the site and
pollutant;
2. CPM and confidence limits for each model as a function of the pollutant; and
49
-------
3. MCM with confidence limits among the models for each performance
measure.
50
-------
SECTION 7 '
EVALUATION OF MODEL OVERPREDICTION
The second stage of the model evaluation process is to determine which model or
models do not significantly overpredict. The subsections below describe the evaluation
strategy.
7.1 OVERALL EVALUATION STRATEGY
Model overprediction is a complex concept. There are many elements to judging
whether a model overpredicts or not. Statistical evaluations can be performed which
unpair the data in time to determine if the model predicts the range of peak values
within an acceptable level of accuracy, but such evaluations are not a complete picture
of overprediction. A model can perform statistically well by averaging overpredictions
with underpredictions. In an attempt to address these complex concerns, a program of
three tiers, or elements, of evaluation has been developed. The first element consists of
a statistical evaluation with the data unpaired in time. The second element involves a
review of a 5-year historical data base that contains meteorological data and particulate
concentrations as seen in the Powder River Basin in an attempt to determine if the
concentrations measured in the 1993 field program are representative of long-term
trends in the Powder River Basin. Finally, the third tier element is a sensitivity analysis
which will determine if the model is running in a reasonable manner for use in regulatory
model applications. The subsections below describe each element of the evaluation.
More emphasis has been given to the first element, the statistical evaluation, because
objective criteria have been developed for this element. The other elements can now
only be described in more general terms, but will be specified in more detail at a latter
time.
51
-------
7.2 STATISTICAL EVALUATION OF MODEL OVERPREDICTION
7.2.1 Model Overprediction Evaluation Strategy
The strategy for evaluating statistical model overprediction has elements that are
similar to the strategy for identifying the best-performing model in that separate
evaluations will be conducted for TSP and PM-10 and statistical inference will be based
on bootstrap confidence intervals for the measures of overprediction. However, the
statistical model overprediction analysis differs from the best-performing model analysis
in four substantial areas. First, the evaluation will focus on model performance at high
concentration stations rather than across the network. The analysis will use results for
the three stations with the highest observed mean concentration for TSP and the
three stations with the highest observed mean concentration for PM-10 for the
respective pollutant-specific analyses. This analysis focuses on these high
concentration stations because they present the greatest potential for having
exceedances. The next two changes are a consequence of model overprediction being
a one directional phenomenon. The fractional bias rather than the absolute fractional
bias will be used as a measure of performance, and all confidence intervals will be one
sided rather than two sided. Finally, the statistical model overprediction analysis will
evaluate the potential for overprediction at individual sites rather than averaged across
the network. Both the point estimates of bias and confidence intervals for those
estimates will be used to define model overprediction.
7.2.2 Test Statistics and Bias Measure
The primary test statistics that will be used in the statistical model overprediction
analysis are the observed and modeled robust highest concentrations (RHCO and
RHCM) as defined in Section 6.2. For each model, these test statistics will be used to
compute the model bias measure, the fractional bias (FB), using the following equation:
FB = 2 RHC° ~
RHC0 + RHC
M
(12)
If the FB is negative, then the observed RHC is smaller than the measured RHC
indicating that the model overpredicts; conversely, if the FB is positive, then the mode!
underpredicts. The fractional bias has two desirable properties for this analysis. First, it
52
-------
is symmetric and bounded so that positive and negative values of the same magnitude
are indicative of equivalent levels of overprediction on a multiplicative scale. For
example a FB of 0.4 indicates that the model underpredicts by a factor of 1.5, while a
FB of -0.4 indicates that the mode! overpredicts by a factor of 1.5. Second, because the
FB is dimensionless, results are independent of the concentration units selected for
analysis, and results obtained for different pollutants present in substantially different
concentrations can be compared readily.
7.2.3 Model Overprediction Evaluation Protocol
The protocol used to determine statistical model overprediction is somewhat more
ambiguous than the best-performing model protocol outlined in Section 6 because until
the set of best-performing models is established, the number of models that will be
included in the analysis is unknown. However, this section will describe the general
protocol that will be implemented to determine whether each of the models in the set of
best-performing models significantly overpredicts. Under this statistical analysis
protocol, a model will be deemed to significantly overpredict if it meets two criteria
applied sequentially. First, there must be statistical evidence of overprediction at one or
more of the three sites examined. If there is statistical evidence of overprediction, the
point estimate of overprediction must exceed a level deemed to be scientifically
meaningful. Note that the protocol will be implemented separately for TSP and PM-10.
First, RHCO, RHCM, and the FB will be calculated for each monitoring station to
be used in the analysis and each model in the set of best-performing models. For TSP,
the three monitoring stations with the highest background-adjusted mean concentrations
are MRI2, MRI3, and MRI6. For PM-10, the stations are MRI3, MRI5, and MRI6. The
analysis will be based on the three highest concentration stations because using multiple
stations will provide protection against anomalies attributable to the particular geometries
of a single station and both pollutants exhibited reasonable separation between the third
and fourth highest concentrations. The fractional biases, FB^p), for specific
combinations of model (i), site (j), and pollutant (p) will be used in the statistical
component of the protocol.
Because FB^p) are random variables, nonzero values are expected even if
specific models provide unbiased estimates at a particular site. To account for this,
bootstrap confidence intervals will be developed for the fractional biases using
53
-------
procedures similar to those described in Section 6.4. However, the procedures will be
modified to generate only upper confidence bounds to address overprediction.
First, a total of 1,500 bootstrap samples will be selected as described in
Section 6.4, and the fractional bias FB^p), where k denotes the kth bootstrap sample
will be calculated. These samples will be used to calculate the standard error of the
fractional bias, denoted as se^p), for each combination of model (i), site (j), and
pollutant (p). For each pollutant p, this process yields 1,500 vectors of dimension 3L,
where lp is the number of models in the set of best-performing models for pollutant p.
The method of Cleveland and McGill again is applied to calculate the values cp such
that 90 percent of the 1,500 vectors generated by the bootstrap sampling procedures
satisfy the system of inequalities:
(13)
FB,3k(p)-FBln3(p)
se,3(P)
2,3
does not include the origin.
This procedure tests the null hypothesis that for each model in the set of "best
performing" models, the fractional bias is zero for each of the 3 sites included in the
overprediction analysis. The alternative is that the fractional bias is less than zero for at
least one site. The null hypothesis will be rejected if the joint confidence region does
not include the origin, or equivalently if the lower 90 percent confidence interval for any
single site does not include zero. By using all models to develop the initial confidence
region, the method provides an overall Type I error protection of a=0.1. At the same
time it defines overprediction in terms of the performance of models at specific sites.
54
-------
The procedure outlined above provides an assessment of statistical significance,
but effects can be significant statistically without being scientifically important. The
procedure provides overall Type I error rate protection for the general hypothesis of no
significant differences. If the overall hypothesis is rejected, the individual model/site
combinations can then be examined to identify which combinations resulted in rejection.
As a matter of scientific judgement, the model will be defined as overpredicting only if
the point estimate for this model/site pair is less than -0.67. This level indicates
overprediction by a factor of 2 at a single site. Such a level is deemed to be reasonable
for the performance of a model at a specific site,
7.3 HISTORICAL DATA REVIEW
This element of the evaluation will define trends and relationships in the observed
concentration values and meteorological data in the five-year historical data base
archived during Phase I of the study. This extensive historical data base includes air
quality (TSP and PM-10) and meteorological observations at many surface coal mines in
the Powder River Basin. A purpose of this element is to determine if the 30-day
sampling period is representative of conditions in the Powder River Basin. This element
of the evaluation includes three steps: (a) investigation of relationships between
meteorology and air quality in the Powder River Basin, (b) investigation of trends in air
quality in the Powder River Basin, and (c) comparison of the historical data to the two-
month monitoring period data.
Investigation of Relationships between Meteorology and Air Quality in the Powder
River Basin: Relationships between meteorology and air quality in the Powder River
Basin will be determined by testing several hypotheses: (a) are periods of high
concentrations restricted to periods with high wind speed conditions or do they also
occur during low wind speed conditions? (b) do periods with high concentrations occur
during extended periods when atmospheric conditions are stable? (c) do high and low
concentrations occur during periods with similar meteorological conditions regardless of
the season of the year? (d) does precipitation cause lower concentrations than
otherwise expected?
Investigation of Trends in Air Quality in the Powder River Basin: The existing
five-year historical air quality data base will be examined to evaluate annual time trends
55
-------
for the Basin as a whole, seasonal cycles, and spatial patterns of the measured
concentrations.
Comparison of Historical Data to Two-month Monitoring Period Data: The
meteorological conditions and air quality concentrations measured during the two-month
field study will be characterized to determine how representative the sample period is
relative to long conditions in the Powder River Basin. For example, do high TSP and
PM-10 concentrations observed in the two-month period occur under similar
meteorological conditions to those in the historical data base? In the evaluation
process, criteria will be developed for judgment of the representativeness of the
meteorological conditions experienced during the 1993 2-month field study for
characterizing situations which cause maximum concentrations in TSP and PM-10 in the
vicinity of large surface coal mining operations in the Powder River Basin. Limitations
within the field data sample, correspondingly limit conclusions to be reached in
assessing the performance of the air quality dispersion models during such conditions.
7.4 MODEL SENSITIVITY ANALYSIS
A sensitivity analysis will be conducted to assess whether the best performing
model (or models) is functioning in a reasonable manner. This element of the analysis
will include three steps: (a) examination of model response under various
meteorological conditions; (b) examination of source characterization input; and (c)
evaluating boundaries of model's use.
Examination of Model Response under Various Meteorological Conditions:
Trends, patterns, and relationships of high observed concentrations to various
meteorological conditions established during the five-year historical period investigation
(see Section 7.4) will be compared with trends, patterns and relationships based on
model predicted concentrations. Comparisons will be made to determine whether model
predictions of high concentrations occur for similar meteorological conditions (i.e., the
right reasons). For example, if the historical data investigation reveals that high
observed concentrations in the five-year historical period occurred during high wind
speed conditions, then the best performing model should also produce maximum
concentrations during such conditions.
56
-------
Examination of Source Characterization Input: Whether the source
characterization information requirements can be simplified without adversely affecting
model performance will be examined. For example, if the best performing model used
highly resolved source activity data (truck traffic on haul roads collected on a plant shift
basis), and an explicit source representation (road segments with specific geographical
coordinates), the impact of using less refined source characterizations will be tested.
Evaluating Boundaries of Model Use: In order to eliminate any aberrant behavior,
information from steps 1 and 2 will be used to evaluate limits of the models use. For
example, if the historical data investigation reveals that low concentrations in the Powder
River Basin have been observed to occur under low wind speed conditions but the best
performing model is not able to simulate this scenario, then limitations on the use of this
model for such conditions will be investigated.
57
-------
This page intentionally blank
58
-------
SECTION 8 •
REFERENCES
Cleveland, W. S., and R. McGill (1984). "Graphical Perception Theory, Experimentation,
and Application to the Development of Graphical Methods." J. Am. Stat. Assoc.,
79(387): 531.
Cole, C. E., B. L. Murphy, Jr., J. S. Evans, and A. Garsd (February 1985).
Quantification of Uncertainties in EPA's Fugitive Emissions and Modeling
Methodology at Surface Coal Mines. Project No. 2784-V12, TRC Environmental
Consultants Inc., Englewood, Colorado.
Cox, W. M., and J. A. Tikvart (1990). "A Statistical Procedure for Determining the Best
Performing Air Quality Simulation Model." Atmospheric Environment,
24A(9): 2387-2395.
Muleski, G. E., G. Carman, and C. Cowherd, Jr. (1994). Surface Coal Mine Emission
Factor Study. Draft Final Test Report, EPA Contract No. 68-DO-0123, Work
Assignments 37 and 55, U.S. Environmental Protection Agency, Research
Triangle Park, NC, January 17.
U. S. Environmental Protection Agency (1977). Quality Assurance Handbook for Air
Pollution Measurements, Volume II—Ambient Air Specific Methods. EPA-600/4-
77-027a, Office of Research and Development, Research Triangle Park, NC, May
(plus supplements dated January 1983 and January 1990).
59
-------
U. S. Environmental Protection Agency (1985/1986/1988/1990/1991). Compilation of A
Pollution Emission Factors, AP-42, 4th Edition. Research Triangle Park, NC.
Supplement A, October 1986. Supplement B, September 1988. Supplement C,
September 1990. Supplement D, September 1991.
U. S. Environmental Protection Agency (1987). On-Site Meteorological Program
Guidance for Regulatory Modeling Applications. EPA-450/4-87-013, Research
Triangle Park, NC.
U. S. Environmental Protection Agency (1992a). Protocol for Determining the Best
Performing Model. EPA-454/R-92-025, Research Triangle Park, NC.
U. S. Environmental Protection Agency (1992b). User's Guide for the Industrial Source
Complex (ISC2) Dispersion Models, Volume I—User Instructions.
EPA-450/4-92-008a, Research Triangle Park, NC.
U. S. Environmental Protection Agency (1993). User's Guide for the Model Evaluation
Methodology (MEM) System for Comparing Model Performance, Version 1.0
(Draft). Research Triangle Park, NC.
U. S. Environmental Protection Agency (1994). Modeling Fugitive Dust Impacts from
Surface Coal Mining Operations—Phase I. EPA-454/R-94-024, Research
Triangle Park, NC.
60
-------
APPENDIX A
DERIVATION OF SET 3 ADJUSTED EMISSION RATES
-------
APPENDIX A
DERIVATION OF SET 3 ADJUSTED EMISSION RATES
The hourly traffic emission rates are constructed with information contained in the
source activity file "24HRVPH" generated during preparation of the Cordero monitoring
report (EPA, 1994). That file contains the number of vehicle passes per hour by
• Heavy duty trucks (haul and water trucks).
• All other vehicles.
Vehicle-mile-traveled (VMT) results are obtained from multiplying vehicle passes
per hour by the road length. Emission rates are calculated by multiplying VMT with the
following representative emission factors:
Emission factors (Ib/VMT)
PM-10 TSP
Heavy-duty vehicles 6 30
Light-duty vehicles 0.13 0.72
These factors represent the geometric mean uncontrolled emission factors measured
for haul roads and light-duty traffic during the 1992 emission testing program
(Muleski et al., 1994).
Conventions followed regarding mitigating effects of natural and anthropogenic
controls are then applied. First, natural precipitation is assumed to control roadway
particulate emissions in the following way:
A-l
-------
Assumed control efficiency
Precipitation (inches) for the hour
0.2 or more during present hour 100 percent
0.01 to 0.19 during present hour . 75 percent
0.5 or more during preceding 5 hours 30 percent
1 or more during preceding 11 hours 20 percent
2 or more during preceding 23 hours 10 percent
Note that only one control efficiency due to precipitation is included,
corresponding to the highest applicable efficiency. The first three efficiencies in the
table are consistent with MRI's past findings with road watering programs. The last
two efficiencies represent our best judgment as to the effect of rainfall on active roads.
In addition to any mitigation from rainfall, 60 percent control efficiency is applied
to roadway emissions whenever water truck passes constitute part of the heavy truck
traffic on the road during the observation period. The 60 percent control efficiency
approximates the average control efficiencies for watering found in the 1992 testing
program when only those controlled emission tests that were performed within about
an hour of watering were included in the calculations, because such conditions are
more reflective of the normal Cordero watering program:
Watering control efficiency
PM-10 TSP
Coal haul roads 56 percent 57 percent
Overburden haul roads 80 percent 47 percent3
Both 63 percent 61 percent
aBased on only one controlled emission test.
Note that the control attributed to watering is added to the control attributed to natural
mitigation by rainfall. For example, if there are water truck passes on a road
(60 percent control) with 1 inch or more of precipitation during the preceding 11 hours
A-2
-------
(20 percent control), the controlled emission rate would be found as the uncontrolled
ratex (1-0.20) x (1-0.60).
A-3
-------
APPENDIX B
ESTIMATION OF EMISSIONS FROM
CABALLO ROJO MINE HAUL ROAD
-------
APPENDIX B
Calculation of coal haul truck emissions from the main haul road at the Caballo
Rojo mine utilizes the same emission factors as applicable to the coal haul roads at
the Cordero mine. Activity levels (i.e., vehicle passes over the 1.17 mi length of haul
road) are derived from the daily quantities of coal mined3 and hauled in trucks having
a capacity of 170 tons. Because coal production by work shift is not available, the
calculated average hourly emission rates do not vary within each 24-hr period.
Finally, an overall efficiency of 25% is assumed for the combined effects of
road watering and precipitation and is applied to all days uniformly. This efficiency,
which is approximately half the typical value used for regular road watering, accounts
for the fact that (a) water trucks are not operated during all shifts, and (b) other lesser
contributions from the haul road, such as the emissions from water trucks, graders
and wind erosion, are not calculated separately.
aThese data were provided to the project team by the Caballo Rojo mine.
B-1
-------
APPENDIX C
PARTICLE SIZE DISTRIBUTION
-------
APPENDIX C
PARTICLE SIZE DISTRIBUTION
The aerodynamic particle size ratios presented in Section 3.6 were derived
entirely from the results of the 1992 field testing program at Cordero
(Muleski et al., 1994). In the 1992 field testing, the primary device used for particle
sizing was a high-volume (20 acfm) sampling train that contained a cyclone
precollector and a 5-stage cascade impactor. This sampling train provided direct size
separation around the following aerodynamic particle size outpoints: 15 ^mA
(cyclone); 10.2, 4.2, 2.1, 1.4, and 0.73 (imA (5-stage impactor). For each of five
particle sizing tests, cyclone/cascade impactor combinations were operated at 1- and
3-m heights and at a nominal distance of 5 m from the roadway.
The particle sizing results from the 1992 testing are given in Table C-1, which
reproduces Table 11 (page 39) of the Revised Draft Final Test Report Surface Coal
Mine Emission Factor Study (Muleski, et al., 1994). By averaging (geometrically) the
weight percentages in each column, a representative particle size "profile" was
generated and graphed on log-probability paper (Figure C-1).
The data point for 32 |imA in Figure C-1 reflects (a) the particle size outpoint for
TSP and (b) the ratio of the coefficients (k) for TSP and PM-10 from the new emission
factor equation for haul trucks (Equation 2 in the body of this report). These
coefficients reflect the results of Cordero mass emission tests using plume profiling
towers equipped with standard high-volume samplers (for measurement of TSP) and
with reference PM-10 samplers (for measurement of PM-10), respectively.
C-1
-------
TABLE C-1. PARTICLE SIZING DATA
o
Mean 3-m
Run wind speed
Source No. (mph)
Coal haul 100X 4.2
road
(site 1 B)
102X 18.8
Watered coal 111X 18.5
haul road
(site 1)
112X 22.2
Overburden 121X 6.6
haul road
(site 4)
Geometric mean
Sample
* _ * i— 4
height
(m)
1
3
1
3
1
3
1
3
1
3
Weight percentage of total particulate less than stated size
(aerodynamic diameter)
15 (jmA
30.0
41.9
16.9
19.7
18.6
20.3
13.7
28.2
17.0
15.0
20.6
10.2 umA
21.1
32.6
14.0
16.1
14.1
14.5
11.3
23.8
11.0
8.8
15.5
4.2 umA
12.2
16.7
8.9
9.4
8.7
8.1
9.0
18.7
3.9
2.3
8.4
2.1 umA
8.6
8.2
4.3
5.5
5.4
8.1
8.3
8.5
1.3
0.8
4.7
1 .4 |jjnA
8.1
3.0
2.0
2.8
4.3
6.7
6.2
5.1
0.5
0.3
2.7
0.73 umA
6.2
3.0
1.4
1.9
4.3
5.1
4.5
3.8
0.5
0.3
2.2
-------
o
-------
APPENDIX D
SUPPORTING DATA FILES
-------
APPENDIX D
SUPPORTING DATA FILES
Primary Measurements
High-Volume Concentration Measurements
File Name: TABLE.
Description: Spreadsheet used to calculate 24-h high-volume air concentrations
at the six monitoring sites. Includes information related to filter
number, elapsed time, start and stop flow rates, etc.
Site/Month Concentrations
File Name: HV1.MAY, HV1.JUN, ... , MRI6.JUN, MRI6JUL
Description: 30 data files containing the site identification, run date, PM-10 and
TSP concentrations and status of measurements (i.e., both PM-10
and TSP OK, etc.). One file for each calendar month for each of
the ten monitoring sites.
Inter-Mountain Laboratories, Inc. (IMP Meteorological Data3
File Name: IMLMET.DAT
Description: File containing hourly surface data collected by the Cordero
meteorological station. Data includes wind speed, wind direction,
sigma theta, ambient temperature and precipitation.
Time-Resolved Dragline Activity
File Name: DRAGLINE.DAT
Description: Data file containing dragline load cycle information over each of
the three observation periods. Also contains information on
location of drop, referenced to the 1,000- x 1,000-ft grid system.
aNOTE: This file has been modified from the data file submitted with the draft
Phase I report. The current file includes the data substituted for periods of missing
data.
D-1
-------
Time-Resolved Scraper Activity
File Name: SCRAPER.DAT
Description: Data file containing information on the use of scrapers during the
three observation periods. Also contains information on location of
scraper operation, referenced to the 1,000- x 1,000-ft grid system.
Time-Resolved Truck/Shovel Activity
File Name: 24HRVPH.
Description: Source activity data file containing haul truck, water truck, and
total vehicle passes per hour on 12 different roads. This file
contains information that can be used to develop activity levels not
only for traffic on the 12 haul road segments (see Appendix A),
but also for coal and overburden truck loading and dumping. This
is done by assuming that
1. The number of coal loading operations in areas "a," "b," "c,"
and "h" in Figure 2 of the text equals one-half the number of
haul truck passes on roads A, F, G, and T, respectively.
Similarly, the number of overburden truck loadings in "d" and
"g" equals one-half the number of haul truck passes on
roads D and X.
2. Any loading of trucks (with power shovels) is accompanied by
a dozer. Thus, whenever the number of coal or overburden
loads in an area is greater than zero, dozer emissions are
assumed to be occurring in the area as well.
3. The total number of truck dumps of coal in area "f" in Figure 2
equals one-half the number of haul trucks traveling on the
permanent haul roads (M and Z). The number of truck dumps
of overburden equals one-half the haul truck passes on
roads E and V.
4. All loaded haul trucks are assumed to carry 240 tons of
material.
D-2
-------
Time-Resolved Grader Activity13
File name: GRADER.LOG
Description: Provides a list of days, shifts and road segments on which graders
were observed.
Volume Source Representations of Roads
File Name: ROADA., ROADB., ... , ROADZ., ROADX., ROADV
Description: Twelve data files, each corresponding to the 12 roads observed
for source activity. Each file consists of one or more straight line
segments used to represent the road. Start and end coordinates
can be used to depict roads as elongated area sources in the new
version of ISC2. File also contains x-y coordinates for volume
sources currently used to represent line sources in ISC2.
Coordinates referenced to Wyoming state system.
Area Source Locationb
File name: AREAS.LOC
Description: Provides the coordinates (m) for the Southwest corner (or south
corner, as applicable) of the area source designations for coal
loading and unloading, overburden loading and unloading, scraper
activity, and drag line activity.
bThis file is a new file not included in the Phase I report data files.
D-3
-------
APPENDIX E
REGIONAL PM-10 BACKGROUND CONCENTRATION
-------
APPENDIX E
REGIONAL PM-10 BACKGROUND CONCENTRATION
A procedure for determining a regional PM-10 and TSP background concentration
for each of the 30 sampling periods is described in Figure E-1. Based on the
acceptance criteria for allowable wind directions for the station with the lowest
measured concentration on a given day, 18 of the 30 sampling days had directly
measured PM-10 concentrations and 14 of the days had directly measured TSP
concentrations that were acceptable as background values (see Tables E-1 and E-2).
Therefore, it was necessary to estimate values of PM-10 and TSP background
concentration only for the remaining 12 and 16 days, respectively.
The graphical "model" for estimating PM-10 background concentration is shown in
Figure E-2, based on the 18 days with directly measured concentrations that were
acceptable as background values. It relates the background concentration to days
since significant rainfall, as a surrogate for decreasing surface material moisture
content. A "best fit" straight line through the 18 data points is also shown. As a
secondary dependent variable, the graph labels each data point with maximum daily
temperature as a surrogate for evaporation rate.
Inspection of the graphical model for PM-10 shows that when there have been no
(zero) days since rainfall, the concentration is insensitive to maximum temperature.
The sensitivity of PM-10 concentration to maximum temperature increases with
increased time since significant rainfall. This observation appears to be consistent
with the fact that increased evaporation (resulting from higher temperatures) coupled
with an absence of rainfall produces dry (dusty) surface conditions.
E-1
-------
MODEL DEVELOPMENT
1. Determine which wind directions are allowable at each monitoring station such that
impacts of "local" sources (Cordero sources or the Caballo Rojo haul road) are not
encountered.
2. For each given day, select the lowest measured concentration value as a candidate
regional background concentration.
3. If unallowable hourly average wind directions occur during < 20% of the day in question
(i.e., no more than four hourly readings), accent that day's lowest concentration as
appropriate for a regional background concentration.
4. For each day with an accepted regional background concentration, determine the
number of prior days without significant rainfall (> 0.05 in), not including the day of
interest.
5. Plot regional background concentration vs. days since rain and label each data point
with the maximum temperature for the day; add "best fit" line to the graph.
MODEL USE
6. For PM-10—If the day of interest occurs immediately after significant rainfall (i.e.,
occurs zero days after significant rainfall), use "best fit" line to estimate background
concentration to the nearest microgram per cubic meter.
For TSP— For any day of interest (without regard to time after rainfall), use the "best fit"
line to estimate background concentration to the nearest microgram per cubic meter.
Go to Step 9.
7. If the day of interest occurs one or more days after significant rainfall and has a
maximum temperature between 60° and 75°F, use "best fit" line in the "graphical
correlation" to estimate background concentration to the nearest microgram per cubic
meter.
8. If the day of interest occurs one or more days after significant rainfall and has a
maximum temperature outside the range of 60° to 75°F, estimate a background
concentration to the nearest microgram per cubic meter by inspection of the graphical
model, i.e., select a value above or below the "best fit" line that best represents the
maximum temperature of the day.
9. If the measured concentration Is less than the estimated background concentration of
the day, use the measured value as the background concentration.
Figure E-1. Steps to determine regional PM-10 background concentration.
E-2
-------
TABLE E-1. QUALIFYING VALUES FOR REGIONAL PM-10 BACKGROUND CONCENTRATION
Date
5/21
5/23
5/25
5/27
5/29
5/31
6/4
6/6
m
co 6/8
6/10
6/12
6/24
6/28
7/8
7/10
7/12
7/14
7/18
Station
HV1
HV2
HV1
HV2
HV1
HV3
HV3
HV2a
HV1
HV1
HV3
HV1
HV1
HV1
HV2
HV1
HV3
HV2a
Regional
PM-10
concentration
8.60
5.45
6.17
9.92
9.09
7.32
4.25
4.96
6.53
8.26
9.34
5.07
12.91
7.11
9.96
15.34
9.59
7.50
Wind
direction
persistence
HV
VS
MS
V
HV
S
VS
V
VS
MS
HV
MS
V
S
MS
HV
V
MS
Wind speed
Avg/Max
(mph)
14/26
16/27
8/12
10/16
10/20
10/20
18/24
10/19
25/34
6/10
10/17
13/20
17/27
9/14
14/20
15/21
23/34
10/18
Days
since
rainfall
2
0
2
4
0
2
0
0
0
1
3
1
5
1
3
5
7
0
Temp
Avg/Max
(°F)
61/78
48/55
68/54
59/67
59/72
73/61
49/61
53/60
49/57
61/73
58/68
51/60
68/81
56/66
55/65
65/80
57/65
61/69
Regression output
Constant 6.19143
Std. err. of Y est. 1.94528
R squared 0.55555
No. of observations 18
Degrees of freedom 16
X coefficient(s) 0.99789
Std. err. of coef. 0.22314
-------
TABLE E-2. QUALIFYING VALUES FOR REGIONAL TSP BACKGROUND CONCENTRATION
m
Date
5/23
5/25
5/27
5/31
6/4
6/6
6/8
6/10
6/12
6/14
6/28
7/2
7/8
7/10
Regional TSP
Station concentration
HV1
HV1
HV2
HV3
HV1
HV2a
MRI5
HV1
HV3
HV3
HV1
HV3
HV1
HV2
7.90
7.87
31.25
13.21
5.23
7.79
6.27
9.98
20.28
21.75
34.75
21.81
17.38
16.87
Wind Wind speed Days Temp
direction Avg/Max since Avg/Max
persistence (mph) rainfall (°F) Regression output
S
MS
V
S
vs
V
vs
MS
HV
HV
V
V
S
MS
16/27
8/12
10/17
10/20
18/24
10/19
25/34
6/10
10/17
8/19
17/27
15/24
9/14
14/20
0
2
4
2
0
0
0
1
3
5
5
1
1
3
48/55
54/68
59/67
61/73
49/61
53/60
49/57
61/73
58/68
58/71
68/81
66/77
56/66
55/65
Constant
Std. err. of Y est.
R squared
No. of observations
Degrees of freedom
X coefficient(s)
Std. err. of coef.
7.728353
5.464364
0.68165
14
12
4.227521
0.834001
-------
9-3
Regional PM-10 Concentration
ro
c
—^
CD
m
ID
CD
o
or
o
CQ
O
C
O
O
—^
2.
o>
-------
Inspection of the graphical model for PM-10 (Figure E-2) can be used to estimate
regional background values for the 12 days in question. The results are shown in
Table E-3. It is interesting to note that in all but three cases the estimated
background concentration and the measured lowest concentration for the specific day
agree within ±2 (ig/m3, which roughly corresponds to the 95 percent confidence
interval for collocated PM-10 measurements. This seems to indicate that suspected
local source impacts were not significant on those days. The final column of
Table E-3 shows the value of regional PM-10 background concentration actually
selected for each of the 12 days.
Similarly, the graphical model for estimating background TSP concentrations is
shown in Figure E-3. In this case, however, the concentration appears to be
insensitive to maximum daily temperature. Therefore, the best-fit linear relationship is
used in all cases to estimate background concentration.
The results of the estimating procedure for TSP background concentration are
shown in Table E-4. In all but six cases, the estimated background concentration and
the measured lowest concentration for the specific day agree within ±6 (ig/m3, which
roughly corresponds to the 95 percent confidence interval for collocated TSP
measurements. The final column of Table E-4 shows the value of regional TSP
concentrations actually selected for each of the 16 days.
E-6
-------
TABLE E-3. ESTIMATED VALUES FOR REGIONAL PM-10 BACKGROUND CONCENTRATION
Regional PM-10 concentration (fag/m3)
Date
5/19
6/2
6/14
6/16
6/18
6/22
rn 6/26
i
6/30
7/2
7/4
7/6
7/16
Station
HV1
HV2
HV2
HV2a
HV2a
HV3
HV1
MRI4
HV1
HV3
HV3
HV1
Days since
rainfall
0
0
5
0
0
3
3
0
1
3
0
0
Temp
Avg/Max (°F)
51/61
52/59
58/71
54/60
50/53
61/73
64/81
62/72
66/77
55/64
55/70
58/69
Measured
13.63
4.33
10.92
7.08
4.07
9.74
11.04
7.42
11.00
5.86
5.06
8.39
Estimated3
6
6
11
6
6
9
12
6
10
9
6
6
Estimated-
measured
-7.63
1.67
0.08
-1.08
1.93
-0.74
0.96
-1.42
-1.00
3.14
0.94
-2.39
Selectedb
6
4.33
10.92
6
4.07
9
11.04
6
10
5.86
5.06
6
a Estimated to the nearest microgram.
If there are no days since rain or the temperature is between 60° and 75°F, use the line.
Otherwise, estimate the effect of temperature based on the graph.
Equation for the line: y = (x)*(0.99789) + 6.19143.
b The lower of the estimated and measured values was selected.
-------
8-3
CQ
c
"-T
CD
m
CO
CD
ro-
O
Q)
5L 5-
-t o
5g CD
o- 3
cu =:.
o -3
3 V
i o
o O
8 01
O5-
tn-
Regional TSP Concentration
CO
ro
o
fo
oo
CO
o>
IT
CO
>
5"
\
en-
(T
01
oo
-------
TABLE E-4. ESTIMATED VALUES FOR REGIONAL TSP BACKGROUND CONCENTRATION
m
i
CD
Regional TSP concentration (^ig/m3)
Date
5/19
5/21
5/29
6/2
6/16
6/18
6/22
6/24
6/26
6/30
7/4
7/6
7/12
7/14
7/16
7/18
Station
HV3
HV3
HV3
HV2
HV2a
HV2a
HV3
HV3
HV2a
HV1
HV3
HV3
HV3
HV2a
HV3
HV3
Days since
rainfall
0
2
0
0
0
0
3
1
3
0
3
0
5
7
0
0
Temp
Avg/Max (°F)
51/61
61/78
59/72
52/59
54/60
50/53
61/73
51/60
64/81
62/72
55/64
55/70
65/80
57/65
58/69
61/69
Measured
29.23
16.75
17.98
8.41
11.05
2.51
16.78
8.35
27.77
19.06
7.89
8.32
34.67
17.01
13.95
21.14
Estimated3
8
16
8
8
8
8
20
12
22
8
22
8
29
37
8
8
Estimated-
measured
-21.23
-0.75
-9.98
-0.41
-3.05
5.49
3.22
3.65
-5.77
-11.06
14.11
-0.32
-5.67
19.99
-5.95
-13.14
Selected13
8
16
8
8
8
2.51
16.78
8.35
22
8
7.89
. 8
29
17.01
8
8
a Estimated to the nearest microgram.
According to the graph, there does not seem to be the same type of correlation with temperature as there was
with PM-10. Consequently, the line was used to estimate values in all cases. The equation is y =
(x)*{4.227521) + 7.728353.
b The lower of the estimated and measured values was selected.
-------
APPENDIX F
SILT ANALYSIS RESULTS
-------
APPENDIX F
SILT ANALYSIS RESULTS
Category
Silt, %
Date
COAL HAUL RAMPS
Road A
Average
Road T
Average
Road G
Average
Road F
Average
Overall average
2,41
3.09
9.24
14.00
9.22
5.46
4.35
6.82
1.60
1.29
7.51
3.66
3.52
2.63
3.31
5.54
3.78
9.08
4.87
9.07
2.91
5.11
6.91
5.56
5.99
6.42
6.00
5.57
7/16/93
7/6/93
7/2/93
6/28/93
6/26/93
6/24/93
5/31/93
Site
G, F, A, T)
A
A
A
A
A
A
A
7/14/93
7/2/93
6/26/93
6/22/93
T
T
T
T
6/30/93
6/14/93
6/10/93
6/6/93
5/27/93
G
G
G
G
G
6/12/93
6/10/93
5/31/93
5/29/93
5/25/93
5/23/93
5/21/93
F
F
F
F
F (composite)
F
F
Note: This is a weighted average
F-1
-------
TABLE F-1. (continued)
Category
Silt, %
Date
Site
MAIN COAL HAUL ROADS (M, 2)
Road M
Average
Road Z
Average
Overall average
4.76
3.15
2.37
1.30
3.50
3.53
0.84
3.49
1.48
3.05
1.04
2.73
5.92
3.79
2.91
1.20
2.82
1.66
1.58
3.16
1.55
1.99
2.65
7/16/93
7/16/93
7/6/93
7/4/93
6/30/93
6/12/93
6/6/93
5/31/93
5/29/93
5/27/93
5/25/93
5/23/93
5/23/93
5/21/93
5/21/93
5/19/93
7/4/93
6/26/93
6/24/93
6/22/93
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
Z
Z
Z
Z
Note: This is a weighted average
OVERBURDEN HAUL ROADS (D, E, V, X)
Road V
Average
Road D
Average
3.09
3.32
4.57
5.27
4.06
3.05
5.44
1.34
3.26
7/14/93
7/10/93
7/6/93
6/2/93
7/12/93
7/12/93
5/31/93
V (100 yds from T intersection)
V
V
V
D (water tracking composite)
D
D
F-2
-------
TABLE F-1. (continued)
Category
Road E
Average
RoadX
Average
Overall average
Silt, %
7.78
6.27
1.29
5.11
3.38
3.47
3.43
4.02
Date
7/12/93
6/14/93
5/31/93
7/8/93
6/2/93
Site
E
E
E
X
X
Note: This is a weighted average
ACCESS ROADS (B, C)
Road C
Average
Road B
Average
Overall average
13.70
3.85
8.78
9.98
9.98
9.18
7/16/93
6/1 4/93
5/27/93
C
C
B
Note: This is a weighted average
MISCELLANEOUS
5.76
4.70
7.96
7.11
12.10
3.79
4.53
7/16/93
7/6/93
6/28/93
5/13/93
6/12/93
7/1 8/93
7/16/93
Side of G
Bottom of lift A
N. pit parting haul area
Coal bench by shoven No. 6
Travel area on east side of shop
Ovb. from S. pit
Ovb. dump off E
F-3
-------
TECHNICAL REPORT DATA
frlease read Instructions on the rtvene txfore completing)
t REPORT NO.
EPA-454/R-94-Q25
4. TITLE AND SUBTITLE-
Modeling Fugitive Dust Impacts from Surface Coal
Mining Operations - Phase II. Model Evaluation Protocol
3. RECIPIENT'S ACCESSION NO.
S. REPORT DATE
October 1994
6. PERFORMING ORGANIZATION COOS
7 AUTMORlS)
C. Cowherd, G. E. Muleski, A. Caughon, D. Wallace
8, PERFORMING ORGANIZATION REPORT MC
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Midwest Research Institute
425 Volker Blvd.
Kansas City, MO 64110
10. PROGRAM ELEMENT NO.
1 !. CONTRACT/GRANT NO.
68-D2-0159
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards, TSD
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final Report
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
Technical Representative: Jawad S. Touma
18. ABSTRACT
This report is the second part of a study designed to analyze the accuracy of the Industrial
Source Complex model for application to fugitive dust sources from surface coal mining operations.
The first report, EPA-454/R-94-024 described the field monitoring program to collect data on
ambient air quality, meteorology and source activity at a surface coal mine in the Powder River
Basin in Wyoming. This report defines the procedures that will be used to identify the best-
performing model for predicting the impacts of fugitive dust sources at these mines and to assess
"significant" overprediction, if it occurs. The report describes the representation of the emission
sources in the emission inventory; the modeling systems to be evaluated; the source activity,
meteorology, air quality and background air quality data bases to be used; and the evaluation
methodology.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lOENTIFIERS/OPEN ENDED TERMS C. COSATI f-ield;Croup
Air Pollution
Air Quality Dispersion Modeling
Meteorology
Surface Coal Mines
Fugitive Dust
8. DISTRIBUTION STATEMENT
Release Unlimited
19 SECURITY CLASS iTha Rtport/
Unclassified
21. NO. OF PAGES
82
20 SECURITY CLASS iThapagt)
Unclassified
22. PRICE
KPA Pe»m 2220-t (R»y. 4 — 77} "WCvioul COITION 11
-------
1 REPORT NUMBEK
Insert the EPA report number as it appears on the covet of the publication.
2. LEAVE BLANK
X RECIPIENTS ACCESSION NUMBER
Reserved fot use by each report recipient.
TITLE AND SUBTITLE
"ule should indicate clearly and briefly the subject coverage of the report, and be displayed prominently. Set subtitle, if used, in si
ie or otherwise subordinate it to mam title. When a report is prepared in more than one volume, repeat the primary title, add vo
Tiber and include subtitle for the specific title.
5. REPORT DATE
Each report shall carry a date indicating at least month and year. Indicate the basis on which it was selected It g.. date of issue, dai
.ipi. oval, date of preparanon. etc.).
6. PERFORMING ORGANIZATION CODE
Leave blank.
7. AUTHOfl(S)
Give name(s) in conventional order {John R. Dot. J. Robert Doe. etc.). List author's affiliation if it differs from the performing org
zauon.
8. PERFORMING ORGANIZATION REPORT NUMBER
Insert if performing organization wishes to assign this number.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Give name, street, city, state, and ZIP code. List no more than two levels of an organizational hirearchy.
10. PROGRAM ELEMENT NUMBER
Use the program element number under which the report was prepared. Subordinate numbers may be included in parentheses.
11. CONTRACT/GRANT NUMBER
Insert contract or grant number under which report was prepared.
12. SPONSORING AGENCY NAME AND ADDRESS
Include ZIP code.
13. TYPE Of REPORT AND PERIOD COVERED
Indicate interim final, etc., and if applicable, dates covered.
14. SPONSORING AGENCY CODE
Insert appropriate code.
IS. SUPPLEMENTARY NOTES
Enter information not included elsewhere but useful, such as: Prepared in cooperation with. Translation of, Presented'at conference
To be published in. Supersedes, Supplements, etc.
ie. ABSTRACT
Include a brief (200 ivords or lets) factual summary of the moat significant information contained in the report. If the report vontau
significant bibliography or literature survey, mention it hen.
17. KEY WORDS AND DOCUMENT ANALYSIS
(a) DESCRIPTORS - Select from the Thesaurus of Engineering and Scientific Terms the proper authorized terms that identify the ma
concept of the research and are sufficiently specific and precise to be used as index entries for cataloging.
(b) IDENTIFIERS AND OPEN-ENDED TERMS • Use identifiers for project names, code names, equipment designators, etc. Use ofx
ended terms written in descriptor form for those subjects for which no descriptor exists.
(c) COS ATI FIELD GROUP • Field and group assignments are to be taken from the 1963 COSATI Subject Category List. Since the i
jonty of documents are multidiscipiinary in nature, the Primary Field/Group assignment(s) will be specific discipline, area of human
endeavor, or type of physical object. The appucation(s) will be cross-referenced with secondary Field/Group assignments that will foi
the primary posting!j).
18. DISTRIBUTION STATEMENT
Denote releasability to the public or limitation for reasons other than security for example "Release Unlimited." Cite any availability
the public, with address and price.
19.420. SECURITY CLASSIFICATION
DO NOT submit classified reports to the National Technical Information service.
21. NUMBER OP PAGIS
Insert the total number of pages, including this one and unnumbered pages, but exclude distribution list, if any.
22. PRICE
Insert the price set by the National Technical Information Service or the Government Printing Office, if known.
F«* 2:20.1 (R.,. 4.7?) <«.„,..) U.S. Environmental Protection Agency
Region 5, Library (PL-12J)
77 West Jackson Boulevard, 12th Floor
Chicago, !L 60604-3590
------- |