-------
the required input formats (using the RAMMET and STAR programs,
respectively). These models also allow for the calculation of impacts at
multiple receptors, a capability needed to determine the spatial extent of the
nonattainment area.
An essential prerequisite to the use of dispersion models in SIP
development is the availability of reliable background concentration
estimates. As shown in Section 5.0, the existing example urban area data base
contains sufficient data to produce such estimates.
Since there are dispersion models available that are applicable to the
nonattainment problem, these models should be used in the SIP development
analyses. Dispersion models would not be used only if there were no
applicable models available.
6.3 Receptor Models Selected for the Example Urban Area
Upon review of the available data base (see Tables 5-1- and 5-2), and the
criteria for receptor model applicability (see Table 6-1 and the
aforementioned receptor modeling documents), the CMB model was selected for
the example urban area analyses for SIP development. Although the data base
is not' all-encompassing, it does contain particle composition data in both the
fine and coarse fractions for many ambient samples and several site-specific
sources suspected of being important contributors to the nonattainment
problem. Therefore, useful CMB analyses can be performed when site-specific
source signatures are supplemented with other representative source signatures
from the EPA Source Composition Library (U.S. EPA, 1984a).
The availability of particle size data will enable a distinction to be
made between the impacts of sources of fine and coarse particles. However,
the lack of data for certain species (e.g., carbon and sodium) may result in
the need for supplemental filter analyses to distinguish between otherwise
-50-
-------
similar sources. The availability of many ambient samples makes it possible
to average the results of a number of CMB analyses to obtain an estimate of
long-term (e.g., annual) source contributions.
Since the CMB receptor model is applicable and the necessary data are
available, the first (and preferred) SIP development option of using
dispersion and receptor models in combination is appropriate for the example
urban area. A receptor model would not be used only if there were no
applicable model and necessary data available.
6.4 Preliminary Analyses
An efficient approach to SIP development modeling should include
preliminary or screening modeling analyses to obtain a qualitative assessment
of the cause(s) of nonattainment prior to embarking on any large scale
modeling analyses. This step can include both dispersion and receptor
modeling. The screening analyses performed to address the nonattainment
problems identified in the example urban area in Section 4.0 are described
below.
6.4.1 Screening Dispersion Modeling
Screening modeling was performed to determine if any of the point sources
in the emission inventory are likely to have large impacts at the monitoring
sites identified as having a high probability of nonattainment. The selection
of the point sources to be subject to the screening modeling was based on the
source's TSP emission rate (Q), an exponential term which included stack
height (H.) and distance (D) from the monitor. Those sources with the
greatest QH./D ratios were modeled using standard screening procedures (U.S.
EPA, 1981c) to determine maximum 1-hour average impacts. The screening
modeling results showed that no single point source or group of point sources
-51-
-------
can be expected to be the primary cause(s) of nonattainment. However, the
largest contributors were found to be coal-fired power plants.
6.4.2 Receptor Modeling
Potentially useful preliminary receptor modeling methods include optical
microscopy, automated scanning electron microscopy, microinventories, chemical
emission inventories, factor analysis and CMB. Of these, only the chemical
emission inventory and CMB can be performed using the available data base.
For the example urban area, one of the high concentration ambient samples was
selected for each of the three monitors with available chemical composition
data, and a CMB analysis was performed to obtain an estimate of source
contributions. The source signatures used for this analysis were all obtained
from the EPA Source Composition Library (U.S. EPA, 1984a) and included the
types of point sources identified in the screening dispersion modeling
analyses. Also included was a road dust source signature, since this type of
source is expected to be a major contributor in urban areas.
The preliminary CMB results indicated that road dust is the major
contributor to the PMi o at all three monitors where receptor model data were
available in the example urban area. However, additional analyses using
site-specific source signatures would be required for most sources to obtain
quantitative results for each type of source.
6.5 Comprehensive Analyses
Comprehensive modeling analyses will generally be required in urban areas
for several reasons. Urban area nonattainment problems often are caused by a
complicated mix of many sources and source types such that identification of
the contributing sources and quantification of their impacts will be
difficult. The development of cost-effective control strategies will depend
-52-
-------
on an ability to reliably attribute most of the particulate matter to specific
sources or source types. The selection of cost-effective controls may require
resolution of minor source or source type contributions if such controls
cannot be applied to the major contributors. Comprehensive dispersion
modeling analyses may also be needed to define the boundaries of the
nonattainment area and test control strategy effectiveness.
For the example urban area, the preliminary dispersion and receptor
modeling analyses showed that more thorough analyses were required to obtain
definitive results. They also provided valuable insight into a likely cause
of the nonattainment problem, i.e., fugitive dust from roads and other
sources. Therefore, subsequent efforts focused on quantifying the emissions
and impacts of both coal-fired power plants (based on the screening dispersion
model results) and fugitive dust sources.
-53-
-------
7.0 DEMONSTRATING RELIABLE MODEL PERFORMANCE FOR THE EXAMPLE URBAN AREA
The results of air quality models form the foundation of most control
strategy decisions. The availability of reliable modeling results is critical
for developing effective and efficient control strategies. Therefore, the
performance of dispersion and receptor models must be evaluated prior to use
in control strategy development. This section describes procedures for
evaluating selected air quality models (dispersion and receptor) for use in
PMio SIP development. The first subsection covers general considerations
while the remainder of the chapter addresses the evaluation methods used for
the example urban area. The topics covered for the example urban area
include: data preparation; derivation of background concentrations;
comparisons of observed to dispersion modeled concentrations; and
reconciliation of receptor and dispersion model results.
7.1 General Considerations
7.1.1 Development of PM10 Emission Inventories
Evaluation of dispersion models in terms of their abilities to reliably
predict ambient PMio concentrations requires a PMio emission inventory.
The PMio SIP Development Guideline (U.S. EPA, 1987a) contains PMio
emission factors and PMio/TSP fractional multipliers that may be applied to
existing TSP inventories. Further information on PMio emission factors are
contained in AP-42 (U.S. EPA, 1985a and 1986c). Many of the procedures for
developing a PMio emission inventory are the same as those which nave been
employed in the past for TSP inventory development. As with TSP, the four
main types of PMio sources are point, area, mobile and fugitive. In areas
exceeding the 24-hour NAAQS, adding temporal resolution to a PMio emission
inventory may be important. Improving the existing methods of estimating the
emissions of sources contributing condensable and secondary aerosol is also of
-55-
-------
greater concern for PMio emission inventories. Finally, resources must be
allocated for software development tasks which are inevitable in emission
inventory compilation.
Evaluation of receptor models requires reliable information on the
chemical/physical characteristics of the particles emitted by major
contributors to ambient PMio concentrations. The type of source
characterization needed is dependent on the specific receptor oriented
approach undergoing evaluation (U.S. EPA, 1984b). For the purpose of this
discussion, the use of the CMB model will be assumed. Emission composition
data are readily available for many source types. The EPA Source Composition
Library (U.S. EPA, 1984a) is the principal repository for existing source
composition profiles or "fingerprints". Although the fingerprints in the
source library may be used for CMB modeling, more reliable results are
obtained using airshed-specific source composition data developed by
collecting and chemically analyzing emission samples from local sources.
Because the CMB model is strongly dependent on the conservation of relative
composition assumption, more reliable CMB results can be obtained by
collecting source samples in a manner which minimizes deviations from this
assumption.
7.1.2 Compilation of Ambient PMio Data
Dispersion model performance evaluations are based on comparisons between
measured and modeled concentrations. However, the lack of a large historical
PMio data base will inhibit the evaluation process for many areas. For
locations for which only sparse sets of PMio observations exist, it will be
necessary to derive ambient and background PMio concentrations from
historical TSP data. Transformation of TSP concentrations to PMio
concentrations is accomplished most reliably by applying a site specific
-56-
-------
PMio to TSP ratio. In the absence of a site specific ratio, AQCR, State,
regional or national average TSP to PMio ratios may have to be applied.
Further suggestions on the use of TSP measurements as a surrogate for PMio
are given in Appendix D of the PMio SIP Development Guideline (U.S. EPA,
1987a).
As in the case of emissions data, the use of ambient air quality data in
PMio SIP development often requires a substantial amount of data
preparation. This may include data processing to obtain data for use in: 1)
dispersion model evaluation comparisons; 2) receptor model input; 3)
background concentration determinations; and 4) PMio/TSP concentration ratio
calculations.
7.1.3 Prioritization of Monitoring Sites
In general, model evaluation analyses should employ as many monitoring
sites as possible although particular emphasis should be placed on results for
the sites within the boundaries of the nonattainment area. Model credibility
is enhanced by the correct prediction of concentrations at both attainment and
nonattainment monitors. Nevertheless, monitors should not be included if they
are located outside the territory covered by the emission inventory used in
the model.
7.1.4 Meteorological Data
Dispersion models require input of meteorological data. For example, RAM
requires input of hourly wind direction, wind speed, atmospheric stability,
temperature, and mixing height data, while CDM 2.0 requires input of a
joint-frequency distribution of wind direction, wind speed, and atmospheric
stability in the STAR (stability array) format. In addition, model
performance evaluation requires identification of critical meteorological
-57-
-------
conditions. For dispersion modeling, the issue of critical meteorological
conditions pertains primarily to the 24-hour PMio NAAQS. Two sets of
critical conditions are of concern. One set is that which produces the
highest observed concentrations and the other is that which produces the
highest modeled concentrations.
For receptor modeling, critical meteorological conditions must be
considered with respect to the annual average NAAQS. In general, receptor
modeling is applied to a limited subset of the available ambient samples in
order to develop an estimate of the annual average source contributions. This
subset should be chosen to be representative of the distribution of
meteorological conditions which occur at the site (U.S. EPA, 1984b). To
ensure the representativeness of the subset, probability sampling should be
employed to select the subset from the population of available ambient
samples. Two probability sampling techniques which have been used to select
ambient samples for receptor modeling analysis are random and stratified
sampling.
7.2 Data Base Preparation for the Example Urban Area
A 'considerable amount of data manipulation was required to prepare the
data base required for the model evaluation analyses. This subsection
describes, for the example urban area, the procedures used to prepare the
emissions, ambient, and meteorological data and the selection of receptor
locations employed in the analyses.
7.2.1 Emissions Data
The starting point for the example urban area PMio emission inventory
was the existing annual average TSP emission inventory described in Section
5.1. The procedure employed to conduct the TSP to PMio transformation
followed the recommendations found in the PMio SIP Development Guideline
-58-
-------
(U.S. EPA, 1987a). Two types of inventories were prepared for model input;
one comprised of major point sources and the other containing area sources
(which included all other source types). The existing inventory contained
data for 3867 point sources within the modeling region. Selection of 261
major point sources was accpmplished through the use of a screening procedure
based on the QHe/D ratios described in Section 6.3.1. Source-specific
PMio/TSP ratios were then applied to the major point sources to complete the
annual average point source PMio emission inventory. All other sources
(i.e., area, fugitive, mobile, and minor point sources) were placed in the
area source inventory. A general PMio/TSP emission ratio was applied to the
minor point sources based on the average ratio found for the major point
sources. Other sources in the area source inventory were scaled with
source-specific PMio/TSP emission ratios. This completed the compilation of
the PMio emission inventories as required for the annual average model
evaluation.
The annual average PMio emission inventories were then used to derive
temporally resolved PM10 emission inventories. The temporal resolution was
needed in order to evaluate the importance of seasonal, daily (e.g.,
t
weekday/weekend) and diurnal emission rate variations.
To evaluate the dispersion model with respect to estimating 24-hour
average PMio concentrations, hourly emission rates were calculated for each
of the 61 TSP sampling days (every sixth day) in the test year (1982) selected
for the example urban area. Equation 7-1 below was used to derive the hourly
emission rate for each source for each of the 61 TSP sampling days:
PMio(hour) = PMioA*(SF/NDS)DF*HF*252 (7-1)
-59-
-------
where: PMi0A = actual annual PMio emission rate (TPY),
SF = fraction of PMioA in a given season,
NDS = number of days in a particular season,
DF = 7/number of days per week the source was operative (e.g., DF
= 7/5 = 1.4 if the source was operative 5 days per week,
Monday through Friday; and DF = 0 for Saturdays and Sundays),
HF = I/number of hours/day source is operative (e.g., if the
source is operative 8 hours per day, HF = 1/8 for the 8
operative hours, and HF = 0 for the 16 inoperative hours.
Note, a starting time of 7-8 AM was assumed for sources that
operated for less than 16 hours per day.), and
252 = (2000*453.59)/3600 to convert tons/hour to g/sec.
Numerous sources of information were consulted in order to glean appropriate
values for SF, DF, and HF for each source in the inventories. Obtaining and
applying these values was a labor intensive process which required
considerable software development.
In order to evaluate the CMS receptor model, emission composition data
were required. This requirement was fulfilled by the collection and analysis
of emission samples from seven of the largest local point sources (Dzubay
T.G., et al., 1987). A size-segregating dilution sampler was used to collect
the stack emissions. In addition to the point sources, road dust and soil
samples were collected from several locations, resuspended onto filters and
chemically analyzed. X-ray fluorescence was employed to determine the trace
element composition of the source samples. Chemical profiles for other
sources were obtained from the EPA Source Composition Library (U.S. EPA,
1984a).
7.2.2 Ambient Air Quality Data
A limited amount of data preparation was needed to produce the PMio data
used in the RAM and CDM 2.0 model evaluation analyses for the example urban
area. The starting point in preparing these data was the file provided by EPA
which contained TSP concentrations every sixth day for 1982 at the 59
operating monitoring sites in the example AQCR. Of these 59 sites, only 17
(see Figure 3-3) were located sufficiently within the boundaries of the area
-60-
-------
source inventory to be suitable for use in dispersion model evaluation
analyses. An airshed specific PMio/TSP ratio (0.8) was applied to the TSP
data to estimate PMio concentrations. This ratio was the average of the
PMio/TSP ratios observed at four of the six monitoring sites where
collocated samples were collected during the one-month intensive sampling
program in the example urban area (Sites A, B, C, and E in Figure 3-3). Data
from the other two monitoring sites were not used because of excessive
fugitive dust influence at one site and invalid PMio and TSP data at the
other site (Sites D and F, respectively, in Figure 3-3).
Software was developed to transform the receptor model data received from
EPA into the format required by the CMB model. This software made it possible
to readily perform CMB modeling with any of the samples collected at the three
PMio monitors.
7.2.3 Meteorological Data
The data preparation for the example urban area began with the file of
hourly NWS meteorological data plus mixing heights for all of 1982. Using a
combination of the RAMMET preprocessor program and appropriate software, the
hourly data for the every-sixth-day air quality monitoring schedule were
extracted for use in RAM. Additional software was used to read the RAM data
and create annual and seasonal Day/Night STAR data (based on the 61 TSP
sampling days) for use in CDM 2.0.
7.2.4 Receptor Locations
The only receptor locations used in the model evaluation analyses
consisted of TSP monitor site locations. During 1982, a sufficient number of
samples was collected to calculate annual statistics at 44 of the 59 operating
TSP monitoring sites in the example AQCR. These 44 sites are shown in Figure
3-1. Of these 44, 17 (Figure 3-3) were located at least 2.5 Jon inside the
-61-
-------
edges of the 42.5 by 42.5 km area source inventory grid. Only these 17 were
used as receptors in the dispersion model evaluation analyses.
7.3 Derivation of Background Concentrations
Background concentrations are essential to dispersion model evaluation
analyses. In dispersion modeling analyses, background concentrations are
generally considered to be those concentrations caused by sources that are not
included in the emission inventory used by the model. Therefore, dispersion
model evaluation analyses are performed by comparing measured minus background
concentrations to modeled concentrations.
The derivation of the desired background concentrations began with the
identification of monitoring sites that could be used for that purpose. In
order to do this, the 44 TSP monitoring sites that obtained statistically
representative data in the AQCR in 1982 were plotted on a map of the study
area (see Figure 3-1). Potential background monitoring sites were defined as
those located outside the urban center of the 42.5 by 42.5 km area source
inventory grid. Also excluded from consideration as background sites were
those not located in generally rural or undeveloped areas. Of the 44 sites,
12 were selected as being likely to record background concentrations under at
least some meteorological conditions. These 12 background sites are indicated
in Figure 3-1.
The RAM model required 24-hour average background concentrations for each
of the 61 TSP sampling days, while the CDM 2.0 model needed the 61-day average
background concentration. The 24-hour average background concentration for a
specific day was taken as the average concentration from the subset of the 12
background monitoring sites which were upwind of the example urban area for
that day. The annual average background concentration was calculated as the
mean of the sixty-one 24-hour average values.
-62-
-------
7.4 Model Operation
Prior to the use of the models, the selection of model options was
required. This section describes the options employed for the RAM, CDM 2.0,
and CMB modeling analyses which were conducted.
The RAM model (U.S. EPA, 1987f) was operated with the control options set
to:
IPOL = 4, only particulate matter considered
MUOR = 1, urban mode
Z = 0, no consideration of receptor height
IOPT(38) = 1, the regulatory default option for urban area
applications
HANE = 10.
Area source input parameters were as follows:
FH = 0.5, which assumes that area source heights were
comprised of equal contributions from physical stack
height and plume rise.
XLIM = 61, this value is approximately the length of a diagonal
within the 17 by 17 area source grid. Thus, it was
assumed that no integration tables would be necessary at
distances greater than 61 km.
NHTS = 3
HINT(3) = 4.60, 9.10, 13.70
BPH =6.85 and 11.40
The CDM 2.0 model (U.S. EPA, 1985c) was operated with the control options
set to:
NGRAD = 0
FAC = 0.5
RCEPTZ =0.0
NDEF = 1, the regulatory default option for urban area
applications
DELR = 200
RAT = 2.5
CV = 1000
-63-
-------
HL(1)-HL(6) = 2100, 1400, 1400, 1400, 1050, 700 with mean
values from Holzworth (1972) mixing height tables
U(1)-U(6) = 1.50, 2.46, 4.47, 6.93, 9.61, 12.52
XG = 464.5
YG = 4400
TOA =12.6 from local climatological data (LCD) summary
TXX = 2500
DINT = 10
In contrast to dispersion models, the interactive nature of the CMB model
does not facilitate the compilation of a list of parameters which are adjusted
once and then remain in effect for an entire set of model executions. The
adjustments made in the CMB model take the form of selecting the set of source
profiles and chemical features which will be included in the mass balance
regression calculation.
For the example urban area, the selection process was conducted in an
iterative fashion by systematically adding and deleting sources and features
from the calculation until a "best" fit solution was obtained. The
determination of this fit was guided by a combination of summary statistics
reported by the model together with the analyst's understanding of the airshed
under consideration (see U.S. EPA, 1987d).
7.5 Comparison of Observed and Dispersion Modeled Concentrations
The model runs described in the previous section generated a large amount
of data with which the performance of the dispersion models can be assessed
with respect to the example urban area. In this section, the performance of
the dispersion models is examined by comparing the model results with the
PMio concentrations derived from the TSP measurements. In Section 7.6,
dispersion model performance is evaluated further by comparing the receptor
and dispersion model results.
-64-
-------
7.5.1 Annual Average Modeling
Table 7-1 shows the initial comparisons between the measured minus
background and modeled PMio concentrations predicted by CDM 2.0 and RAM in
the example urban area. The table is based on the 17 selected TSP sites, and
uses PM10 values derived from the measured TSP concentrations. CDM 2.0
comparisons are presented for four approaches to incorporating temporal
variability into the emission inventory including: 1) seasonal and diurnal
variability; 2) seasonal variability; 3) diurnal variability; and 4) constant
inventory (i.e., no temporal resolution). Table 7-1 shows that on the basis
of correlation coefficient, slope and intercept, CDM 2.0 performs reasonably
well for all versions of temporally resolved inventories. However, all four
sets of CDM 2.0 modeling results underpredict the observed PMio. Seasonal
variability is seen to have negligible influence on the predictions while
diurnal variability provides the largest underpredictions. The lower
predictions provided by the diurnally variable inventory are due to higher
daytime emission rates coinciding with the higher daytime winds speeds which
cause greater dilution and result in reductions in overall average
concentrations. Despite the large underpredictions associated with the
diurnally dependent emission inventories, diurnal variability was retained in
the inventories for further annual average modeling because the concept is
physically justified and was shown to have a significant influence on the
modeling results.
7.5.2 Short-Tenn Modeling
The comparisons of RAM predicted and measured minus background PMio
concentrations shown in Table 7-1 indicate that the results of the RAM model,
in general, follow the same trends exhibited by the CDM 2.0 model results. As
was found for CDM 2.0, the RAM results underpredict the measured minus
-65-
-------
CO
CO
UJ
^_
»-t
CO
a
UJ
K
UJ
^j
Ul
CO
fx
UJ
z
l-
ae
o
u.
CO
_j
at
in
I-H
CO
_J
z
z
o
^K
3
<
UJ
_J
UJ
oc
a
0
CM
i
*"*
^j
l-l
l-l
l-l
5
a
£
ai a
to
01 X
4>> *>
« -
ac .-
c 2
O IO
71 u
in 10
r
vo
Ul
o«
vO
Ul
vO
Ul
vO
Ul
en
vO
in
Ol
|X
^m
IX
^m
r*
fm
|X
^_
"t?
<4-
U i
§ CO
0 » VO
VO Ul
CM *
^
Ul PI
CM - CM
IN. CM CM
O PI
CM Ul
*- Ul CM
00 O
« *
PI PI CM
PI «
CM i CM
fx Ul
VO CO Ol
. vo
f) ^f
vO 0
^*> ^o
\o
« a
tN UJ
Ul O PI
to
^-
u
c £
o
3 'E a
« N» 01
r- U
01 01 U
u a oi
- vo
CM CM
vo
CM ' CM
vO
CM CM
- vO
CM CM
w- VO
CM CM
VO
CM CM
VO
CM CM
VO
CM CM
vo
CM CM
o
c
§
a
j<
u
', £
i i
a 3
JC VI
u «
S S,T
a a a
en ^ 01 ^* ^ PI vo
*- vo v in PI in pi
vO ^ PI CN PI PI CM
vo t ui PI in PI
a\ t f r» o PI oo
vO in in * in PI
M? ^ O IN FN p| Ul
.- i£ in in PI in PI
vo « CM r» IN. pi \o
~ vo ui in PI ui PI
in i i ii ii
CM 1 1 II II
CM 1 1 II II
CM 1 1 II II
-
Ul 1 1 II II
CM 1 1 II II
CM 1 1 II II
CM 1 1 II II
n
o E
O P* ^ »
t- E Ol ~
a >. u
.M <-> i a 3
-~ u n -o 5 vi -o
« * JS « N- «
o. X« X CM u CM w
E a a
a a a a a "a a a u a
> > 3 > > « > > * >
g
Ni^
s
Q.
1
Ul
01
a
g
CE
U
a
u
|O
CD
c
-66-
-------
background concentrations, and hourly variability (which for RAM is in essence
equivalent to diurnal variability for CDM 2.0) is the only characteristic of
emission inventory variability which significantly influences the modeling
results. Although the RAM results follow the same patterns as the CDM 2.0
results, the overall performance of RAM, as indicated by the correlation
coefficients, slopes, intercepts and ratios of measured minus background to
predicted PMio concentrations, is much poorer than CDM 2.0. However, with
short-term modeling, which is conducted in reference to the 24-hour average
NAAQS, overall model performance is not as important as the ability to
correctly predict elevated impacts. Furthermore, a short-term model must
provide estimated concentrations which agree in magnitude and location with
elevated observations. The lower right-hand corner of Table 7-1 summarizes
the performance of the RAM model with respect to predicting elevated impacts.
On average, the elevated RAM results greatly underpredict the measured
elevated concentrations. However, the addition of temporal resolution to the
inventory does not appear to influence the performance of RAM with respect to
predicting elevated measured concentrations. Therefore, the development of a
temporally variable emission inventory is not necessary for short-term
modeling.
The RAM model evaluation analyses also revealed that the RAMMET processor
program occasionally may produce unrealistically low mixing height estimates
and subsequent overpredictions of area source impacts. Should this problem
arise the EPA Regional Meteorologist should be consulted to develop
appropriate corrective action.
In summary, the initial comparisons between modeled and observed PMio
concentrations show that CDM 2.0 and RAM underpredict the measured values for
the example urban area. The largest relative underprediction was noted for
the RAM model when measured minus background and predicted elevated
-67-
-------
concentrations were compared. In addition, temporal variability was shown to
be important for annual average CDM 2.0 modeling but of negligible influence
with respect to the significant indicators of RAM model performance. On the
basis of the initial comparisons between predicted and measured minus
background concentrations, a tentative conclusion was made that the
underpredictions by both models were most probably due to misspecified
emission rates in the inventory. In the next section, a more detailed
understanding for the underpredictions is derived by comparing the receptor
and dispersion model results.
7.6 Comparison of Receptor and Dispersion Model Results
The source contribution estimates provided by CMB and dispersion modeling
generally are not easily compared in the formats generated by the models. To
facilitate logical comparisons, the dispersion model results are usually
regrouped. The regrouping proceeds by combining the impacts of point sources
associated with similar processes and separating the area source impacts into
their principal components on the basis of the emission inventory. For the
example urban area, the regrouping process provided 14 source categories for
comparing receptor and dispersion model results.
Table 7-2 shows the average receptor/dispersion model comparisons for
three sites in the example urban area. The averages were formed from the same
set of sampling periods for each of the sites. The RAM model was used to make
the comparisons because, unlike CDM 2.0, RAM permitted the use of the
meteorological conditions which were measured during the sampling periods
selected for CMB analysis. However, any misspecifications in dispersion model
input which are identified from the comparisons of RAM with CMB should be
applicable to CDM 2.0 because both dispersion models rely on essentially the
same assumptions and input data.
-68-
-------
Table 7-2 shows several similarities between the results of the CMB and
RAM models including: 1) reasonable agreement between RAM and CMB for many
sources at all three sites; 2) an equal unexplained portion of the observed
mass by both models; 3) reasonable agreement between background and secondary
sulfate; and 4) excellent agreement between CMB and RAM for the impact of the
antimony source at site A (see Figure 3-4) which is the site nearest to the
source. Table 7-2 also shows several substantial differences between the
results of the RAM and CMB models. These differences were examined and then
reconciled following the eight-step procedure described in the Protocol for
Reconciling Differences Among Receptor and Dispersion Models (U.S. EPA,
1987b). In the following discussion, the differences will be identified and
their reconciliation summarized on a case-by-case basis.
Case 1: CMB estimates of the impact of crustal material are
significantly greater than those predicted by RAM. What is
the cause of the disagreement and how can the difference be
reconciled?
A thorough review was conducted of the CMB modeling procedure and input data
with respect to estimating crustal material contributions. The review
indicated that the CMB model provided very reliable estimates of crustal
material impacts. A review of the RAM model inputs identified two errors in
the emission inventory which could have substantially influenced the crustal
material impact estimates. The errors included: 1) the emission factor for
construction activity was low by a factor of four; and 2) the emission factor
used for unpaved road dust was low by a factor of two. The errors were then
corrected and resulted in a 50 percent increase in the overall area source
emissions.
Case 2: CMB estimates of the impact of coal combustion are
significantly greater than those predicted by RAM. What is
the cause of the disagreement and how can the difference be
reconciled?
-69-
-------
TABLE 7-2
INITIAL COMPARISONS OF THE AVERAGE RAM AND CMB MODEL RESULTS
(Contributions expressed as percents of the measured averages)
Source Type
Oil Combustion
Coal Combustion
Oil Refineries
Incinerators
Antimony Source
Secondary Metals
Iron & Steel
Chem. Mfg.
Other Point Sources
Other Area Sources
Mineral Processing
Road Dust & Soils
Mobile Sources
Background
Secondary S04
Total (+ Background)
Total
Unexplained
RAM
CMB
RAM
CMB
RAM
CMB
RAM
CMB
RAM
CMB
RAM
CMB
RAM
CMB
RAM
RAM
RAM
RAM
RAM
CMB
RAM
CMB
CMB
RAM
CMB
RAM
CMB
Site A
.9
.8
1.3
5.9
.4
2.2
.2
.3
1.0
1.4
.2
1.0
- .2
.02
.06
.5
6.0
.2
11.3
19.
5.5
3.2
55.
46.
83.
80.
17.
20.
Site B
.6
1.0
.1
1.3
1.0
7.3
.4
1.0
.1
.02
.03
1.3
.2
.05
.1
.1
3.5
.1
6.7
14.
3.2
3.3
63.
49.
80.
77.
20.
22.
Site F
.1
1.6
.7
1.0
.08
0
.05
2.5
.003
.03
.03
.8
.2
.08
.08
.6
5.0
.2
9.0
23.
4.6
3.8
62.
49.
83.
82.
18.
19.
-70-
-------
A review of the CMB parameters associated with coal combustion
substantiated the coal combustion contribution estimates provided by the CMB
model. Reviewing the RAM inputs related to coal combustion revealed two
causes for the differences between CMB and RAM. First, several estimates were
available for the emission rate for the largest coal combustion source in the
airshed. In the RAM modeling, the lowest estimate of the emission rate had
been employed. The inventory was changed to include the highest emission rate
estimate for this source because RAM was underpredicting CMB and the CMB
results had been judged to be reliable. Second, a portion of the difference
between CMB and RAM was found to be an artifact of the method used to group
the source contributions for comparisons and not a problem with one of the
modeling approaches. Specifically, the impacts from residential coal
combustion emissions were reported as part of the "other area sources"
category in Table 7-2.
Case 3: At sites A and B (see Figure 3-3), there are several
sampling periods for which CMB and RAM predicted elevated
impacts from oil refineries. For these periods, the CMB
estimates are much greater than the RAM estimates. What are
the causes of the disagreements and how can the differences
be reconciled?
A review of the CMB parameters associated with refineries revealed that
the source composition profile associated with refineries lacked any
distinguishing features which resulted in a high detection limit for the CMB
model in terms of estimating refinery impacts. Because of the high detection
limit, the CMB estimates of refinery contributions were re-examined in detail
for each site and sampling period combination^ > With one exception, the CMB
results were found to be consistent with the supporting data. The one
exception was a high impact which was estimated for one sampling period at
site B. An error was found in the modeling procedure used for this sample and
therefore the CMB model was re-run for the sample using the correct modeling
procedure.
-71-
-------
Because the remaining CMB estimates were shown to be valid, the
reconciliation process shifted focus to examining the RAM input data related
to oil refineries. Seven major refineries were identified in the emission
inventory. Only two of these were located upwind during the periods for which
CMB and RAM predicted elevated impacts. As was found to be true for the coal
combustion source, several emission rate estimates existed for each of the two
upwind refineries. Once again, the low end of the emission rate scale had
been used in the evaluation modeling. Because RAM was underestimating CMB,
the emission inventory was adjusted to include the higher emission rate
estimates for the two refineries.
Case 4: There was reasonably good overall agreement between CMB and
RAM for the incinerator source category. However, for some
of the sampling periods at site F (see Figure 3-3), the CMB
model predicted significant contributions from incinerators
while the RAM model results showed impacts close to zero.
What are the causes of the disagreements and how can the
differences be reconciled?
A thorough review of the CMB input data showed the incinerator impact
estimates to be very reliable due to the highly distinctive source profile
associated with incinerator emissions. Next, the RAM input data were reviewed
and three major incinerators were identified. The emission rates in the
inventory corresponded well with the capacities and operating conditions of
the plants. This agrees with the fact that the CMB and RAM results were in
reasonable agreement when known incinerators were upwind of the monitoring
sites. However, the CMB model also predicted significant impacts at site F
from incineration during periods when none of the inventoried sources were
upwind of the site. Therefore, the conclusion was made that during these
periods, incinerators not in the inventory were responsible for the impacts
estimated by the CMB model. This implied that the contributing incinerators
were either outside the area covered by the inventory or else within the
inventoried territory but absent from the inventory. This matter was not
-72-
-------
pursued further for the example urban area. In the case of a real SIP
development, a more definitive resolution would be required. For this study,
the incinerator emission rates were left unchanged because the available
evidence did not justify any alterations which would have brought CMB and RAM
into closer agreement.
7.7 Comparison of Observed Concentrations to Final Dispersion Model Results
Following reconciliation of the RAM and CMB modeling results, selected RAM
and CDM 2.0 model evaluation analyses were repeated using the revised source
emission rates obtained during the reconciliation process. Table 7-3 shows
the results of the post-reconciliation evaluation analyses and provides
comparisons with the initial evaluation analysis results which were originally
discussed in Section 7.5. The left-hand side of Table 7-3 shows that the
revised inventory produced little overall improvement in the performance of
the CDM 2.0 model. The most significant change is the replacement of the
previous underprediction with an overprediction of approximately equivalent
magnitude (~5 ug/m3) suggesting that the revised inventories contained
increases that were too large. The right-hand side of Table 7-3 shows a
substantial improvement in RAM model performance due to the inventory
revisions. Much better agreement now exists between the predicted and
measured minus background combined-site average PMio concentrations. More
importantly, there is now excellent agreement between the RAM model-predicted
and the measured minus background concentrations for the short-term averages
which are critical to the development of reliable control strategies for the
24-hour NAAQS.
In summary, the model evaluation analyses demonstrated that the
receptor/dispersion model reconciliation process improved the performance of
the dispersion models. Although CDM 2.0 overpredicts, RAM provides excellent
-73-
-------
TABLE 7-3
FINAL CDM 2.0 AND RAM MODEL EVALUATION ANALYSIS RESULTS
FOR THE 17 SELECTED SITES
CDM 2.0
Initial Final
Number of 17 17
Samples (n)
Correlation (r) .763 .736
Slope (m) .603 .820
Intercept (b) 6.1 10.0
Avg. Background* 21 21
Avg. Measured-Background 26 26
(ug/m3)
Avg. Predicted (ug/m3) 22 32
Avg. Max. Measured-Background
(ug/m3)
Avg. Max. Pred. (ug/m3)
Avg. 2nd Max. Measured-Background
(ug/m3)
Avg. 2nd Max. Pred. (ug/m3)
Avg. lst-5th Max. Measured-
Background (ug/m3)
Avg. lst-5th Max. Pred. (ug/m3)
RAM
Initial
956
.260
.159
14.6
21
26
19
64
52
57
37
53
36
Final
956
.259
.225
21.3
21
26
27
64
69
57
53
53
49
* Background Range: 5-44 ug/m3
-74-
-------
agreement with measured minus background concentrations using the same
emission inventory. This difference in model performance must be ascribed to
the different model algorithms rather than the emission inventory. When the
post-reconciliation CDM 2.0 and RAM model evaluation analyses are considered
together, further changes to the emission inventory are neither justified nor
necessary. For the annual average control strategy modeling, the predicted
impacts were calibrated using the procedures described in the CDM 2.0 User's
Guide (U.S. EPA, 1985c) and the regression coefficients in Table 7-2 to ensure
that future annual impacts are not overpredicted.
-75-
-------
8.0 CONTROL STRATEGY EVALUATIONS FOR THE EXAMPLE URBAN AREA
8.1 Overview of the Design Concentration Concept
Control strategy development is the process of preparing a verifiable plan
showing the level of control needed to demonstrate attainment of the PMi0
NAAQS. The concept of design concentrations assumes a central role in control
strategy development. A design concentration is that PMio concentration
which a control strategy must be capable of reducing to the level of the
appropriate NAAQS. In other words, design concentrations function as the
basis or reference point from which the necessary level of controls are
determined. Therefore, the process of control strategy development begins
with determining appropriate design concentrations.
PMio control strategies must address the 24-hour and annual average
PM10 NAAQS. Therefore, two sets of design concentrations must be
established for receptors which show a high probability of nonattainment with
respect to the 24-hour and annual average NAAQS. Emission limits must be set
to provide for the attainment of both standards. This is accomplished by
developing a control strategy on the basis of the standard which produces the
more stringent emission limits. In general, a confident assessment of which
standard will provide the stricter limitations cannot be made prior to control
strategy testing. In these cases, a reasonable procedure is to make a
systematic assessment as to which standard is more restrictive and design the
control strategy with reference to that standard. Next, the control strategy
is evaluated in terms of providing for demonstrable attainment of both NAAQS.
If attainment of both standards is demonstrated, then the correct NAAQS was
used as the basis for control strategy development.
Ambient measurements or model estimates may be employed to determine
design concentrations. If model estimates are used, the design concentration
is taken as the sum of the modeled source impacts plus the background.
-77-
-------
Development of the annual average design concentration is relatively
straightforward. If .one or more years of ambient PMio measurements are
available for a site, the design concentration is the average of the observed
annual averages. Similarly, if one or more years of annual average PMio
modeling have been performed, the design concentration is simply the average
of the annual averages.
In contrast to the annual average case, establishing a design
concentration which is appropriate for the 24-hour NAAQS can be considerably
more complicated. The complications arise from the statistical nature of the
24-hour NAAQS which limits the expected number of exceedances of the NAAQS to
one or less per year. The PMio SIP Development Guideline (U.S. EPA, 1987a)
gives four procedures for calculating 24-hour design concentrations including:
1) table look-up; 2) projections from statistical distributions; 3) graphical
estimation; and 4) conditional probabilities. For further details on the four
procedures see the PMio SIP Development Guideline (U.S. EPA, 1987a).
Following the PMio SIP Development Guideline (U.S. EPA, 1987a), the
design concentrations for the example urban area were determined from the
results of dispersion model simulations performed with five years of
meteorological data. The average of the CDM 2.0-produced annual averages for
each of the five meteorological years was used for the annual average design
concentration at each modeled receptor. The table look-up procedure was
applied to the results of 1827 individual days modeled by RAM to determine the
24-hour design concentrations. Since the model provided a continual record of
the PMio concentrations for five years, the design concentration for each
receptor was simply the sixth highest modeled plus background PMio
concentration at each receptor.
-78-
-------
8.2 Establishing Baseline and Projected Emissions
Precise determination of baseline and projected emissions is critical to
developing a defensible control strategy. Baseline emissions are the current
emission levels of the sources within the geographical area covered by the
control strategy. Projected emissions refer to expected future emission
levels. Three types of sources are of importance in developing emission
projections; 1) existing regulated sources which are currently emitting
significantly below allowable levels; 2) new major regulated sources; and 3)
unregulated sources whose emission levels may change in response to general
economic development and/or population growth.
Detailed information on forecasting techniques applicable to developing
emission projections is contained in the Guidelines for Air Quality
Maintenance Planning and Analysis, Volume 7; Projecting County Emissions (U.S.
EPA, 1975).
For the example urban area, maximum allowable emissions and appropriate
growth factors were employed to develop a projected emission inventory for
future years. No new major sources were included in the inventory.
8.3 Preparation of Dispersion Model Input Data for the Example Urban Area
Emissions, meteorological and receptor data sets were prepared for use in
the dispersion model analyses required for control strategy development.
Input data were prepared for both the RAM and COM 2.0 models. Three different
data sets were prepared for input to RAM. These data sets were prepared for
use in analyses designed to identify the most cost-effective and reliable
method of using RAM in control strategy development. The preparation of the
three types of model input data required for RAM and COM 2.0 is described
below.
-79-
-------
8.3.1 Emissions Data
The emission inventory used for the RAM and CDM 2.0 control strategy
modeling was the projected inventory described above. For RAM, that inventory
contained temporally constant emission rates for all 261 point and 289 area
sources. For one of the RAM data sets, a smaller point source inventory was
created by combining similar collocated point sources. Similar sources were
defined as those sources with similar stack heights and plume buoyancies as
determined by calculating their "K" values as follows (U.S. EPA, 1973):
K = (H * T * V)/Q <8-l)
where:
H = stack height
T = exhaust gas temperature
V = exhaust gas flow rate
Q = emission rate
Similar sources were then modeled as one source using the stack parameters
of the original source with the lowest "K" value and the sum of the emissions
from each original source. This reduced the number of point sources from 261
to 144.
For CDM 2.0, diurnal emission rate variability was simulated for all
sources through the use of the input parameters YD and YN, which were set to
1.28 and 0.8, respectively, based on the temporal variability of the emissions
in the inventory.
8.3.2 Meteorological Data
Two sets of meteorological data were prepared for the control strategy
modeling with RAM. The first data set contained 36 days of hypothetical
meteorological data. For each day, each hour of data represented one of the
critical combinations of wind speed and atmospheric stability conditions
-80-
-------
commonly used in screening dispersion modeling (U.S. EPA, 1981c). The 24
combinations of hourly wind speed and atmospheric stability conditions used
for each day follow:
Pasquill/Gifford Atmospheric
Wind Speed (m/s) Stability Class
1 A,B,C,D,E,F
3 A,B,C,D,E,F
5 B,C,D,E,F
7 C,D
10 C,D
15 C,D
20 D
Each of the 36 days of hypothetical meteorological data was assigned a
different wind direction (which was then used for all the hours in the given
day). The wind direction for each day differed in 10 degree increments. For
each hour of each day, the ambient temperature was set to 293 degrees Kelvin
and the mixing height was set to 500 meters.
The second set of meteorological data needed for the control strategy
modeling with RAM was the hourly data from 1980 to 1984. The five years of
data were prepared for model input using the RAMMET processor program.
Separate files of STAR data were developed for each of the five years
(1980-1984) and were then used for the CDM 2.0 control strategy modeling
analyses. These files were created by applying appropriate software to the
corresponding years of RAM format hourly data.
8.3.3 Receptors
The selection of appropriate receptor locations is critical to control
strategy development modeling. Receptors are needed wherever there is the
potential for PMio nonattainment. Receptor grids of varying densities can
be used to determine the location and boundaries of nonattainment areas.
-81-
-------
However, multisource modeling for large receptor grids is expensive. For the
example urban area, a more efficient approach was developed to identify
locations with the greatest potential for PMio nonattainment. The approach
involved applying the EPA procedures designed for estimating the probability
of PMio nonattainment (U.S. EPA, 1986a) to the 1982 to 1984 TSP data
recorded at the 17 TSP monitoring sites located within the 42.5 by 42.5 km
area source grid. This process produced a 95 percent probability of
nonattainment at 3 monitoring sites (9, 14 and 17 in Figure 3-4) and a 20 to
95 percent probability of nonattainment at 5 monitoring sites (4, 5, 7, 10 and
16 in Figure 3-4). A receptor grid with 2 km spacing was then placed around
these 8 monitoring sites, and these receptors (a total of 74) were used in the
initial input for the CDM 2.0 control strategy modeling. (Initially, 1 km
spacing was employed, but the spatial variation in modeled concentrations was
negligible, primarily because the area sources were modeled as 2.5 km squares.)
8.4 Derivation of Background Concentrations
The background concentrations needed for control strategy development
differ somewhat from those needed for dispersion model evaluation. For model
evaluation, background values were needed for the specific days being
modeled. In contrast, the background concentrations needed for control
strategy development must be reasonable estimates of concentrations that may
occur in the future. For an annual standard, the future concentrations should
represent an average year, but for the 24-hour standard the future
concentrations should represent maximum concentrations that may occur on any
given day. This section describes the derivation of such estimates for the
example study area.
-82-
-------
There are a number of available methods for determining PMio background
concentrations. Several are described in the Guideline on Air Quality Models
(Revised) (U.S. EPA 1986b) and in Appendix D of the PMio SIP Development
Guideline (U.S. EPA, 1987a). For the example urban area, the background
concentration estimates used in the model evaluation analyses were based on
the average concentrations observed from among a set of upwind monitoring
sites identified as background sites.
In the model evaluation analyses referred to above, 24-hour average TSP
background concentrations were determined on the basis of the TSP sampling
conducted in 1982 at the set of 12 background sites. The appropriate average
concentration recorded within the upwind subset of the 12 sites was used as
the 24-hour average TSP background concentration. In order to derive PMio
background concentrations for control strategy development, this process was
repeated using 1980, 1981, 1983 and 1984 TSP data at the same set of 12
designated background sites. For each year, this procedure produced sets of
61 24-hour average TSP background concentrations. The resulting 305 values
provide . a detailed set of estimates of 24-hour average TSP background
concentrations under a variety of meteorological conditions. The maximum
value in this data set provides a good estimate of the maximum future 24-hour
average TSP background concentration and the average of all the values
provides an estimate of the annual average TSP background concentration.
The procedure described above produced maximum 24-hour and annual average
TSP background concentrations of 85 and 31.3 ug/m3, respectively. PMio
background concentrations were then obtained by assuming that the 0.8
PMio/TSP concentration ratio observed in the study area also applies to
background. Applying this assumption yields maximum 24-hour and annual
average PMio background concentrations of 68 and 25.0 ug/m3,
respectively.
-83-
-------
8.5 Modeling Projected Source Emissions
The next task in control strategy development was modeling using the
projected source emissions to determine the location, degree and spatial
extent of future nonattainment in the example urban area. The primary
objective of the modeling was to determine the design concentrations that must
be reduced below the NAAQS level. Another objective was to confirm that the
EPA procedures for estimating the probability of PMio nonattainment can be
used to identify nonattainment area locations. A third objective was
establishing the NAAQS exceedance boundaries. The CDM 2.0 and RAM modeling
performed to meet these objectives is described below.
8.5.1 Modeling for the Annual Average NAAQS
Annual average modeling was performed using CDM 2.0 and the data input
described in Section 8.3. All 550 sources were modeled using the projected
emissions (with diurnal variability) of the post-reconciliation inventory.
Modeling was performed separately using each of the five years of STAR
meteorological data. The 1982 STAR data were used first to calculate
concentrations at the 74 receptors selected previously (see Section 8.3.3).
Additional receptors (2 km spacing) were then added near any receptors where
concentrations in excess of the NAAQS were indicated. The modeling and adding
of receptors was continued using the other four years of STAR data to
establish the boundaries of the NAAQS exceedance area. In two areas where
point sources produced significant impact gradients, other receptors (1 km
spacing) were also added as necessary to obtain better estimates of maximum
concentrations. The concentrations from the five modeled years were averaged
to determine the attainment status of each receptor.
-84-
-------
The average results from the five years of CDM 2.0 modeling analyses are
shown in Figure 8-1. The figure shows the complete grid of 107 receptors and
the average modeled PMio concentrations corrected for the overprediction
bias noted in the CDM 2.0 model evaluation analyses {see Section 7.7). The
corrected values were obtained by applying the final CDM 2.0 slope and
intercept values shown in Table 7-3 to the modeled PMio concentrations.
In order to achieve compliance with the annual PMio NAAQS, all the
concentrations at the receptors in Figure 8-1 must be reduced to less than
25 ug/m3 (the difference between the 50 ug/m3 annual NAAQS and the 25
ug/m3 background concentration). The isopleth line shows the boundaries
of the annual average PMio exceedance area. There are four hot spot
receptors where the annual average PMio concentrations reach maximum
levels. The four hot spot receptors are located in the northeast
(36 ug/m3), center (39 ug/m3), upper southern (59 ug/m3), and
lower southern (70 ug/m3) areas of the grid. To make this example task
more manageable, it was hypothesized that control strategies which result in
the NAAQS being attained at the four hot spot receptors will also produce
attainment at all the other exceedance area receptors. The four hot spot
concentrations plus background were therefore established as the design
concentrations for the annual average PMio NAAQS.
8.5.2 Modeling for the 24-Hour NAAQS
For typical AQCR's, the use of refined short-term dispersion models to
define design concentrations appropriate for the 24-hour NAAQS is potentially
a very expensive process. The high costs are produced by the large number of
computations needed to perform a comprehensive determination of the maximum
predicted 24-hour PMio concentrations at many receptors throughout the
example urban area. For example, a comprehensive analysis of only a portion
-85-
-------
Open rectangles enclose hot spot receptors.
Solid squares indicate monitoring sites.
YMAX - 35.1
025 p25 024 024 o25 O24
"25 - -^^
O22 cj26 Q29 o28 O28 Q27 ~D2^ O25
022 023 023
023 024 /D26 Q31 Q32
023
O30 n30 O30 O28
031 031 o35 031
17* r-zn
023 o27 o29 Q31 O39 n35 Q34 Q31 Q28 o26/ Q25
Q23
O22
D27 o29 a33 n33
9
O33 p32 Q35 O35
- "10
Q30 C329 o30 Q30
028 o29 Q37
023
022
Q23
025
D22 023
D25 O24
024 023
D22 021 D21
YMIN = 10.9
Figure 8-1. Corrected 5-year average COM 2.0-modeled PM,n concentrations (yg/m3).
-86-
-------
of the example urban area would require the calculation of 2.6 x 109 hourly
concentrations. This estimate is based on five years (43,800 hours) of RAM
modeling for 550 sources and 107 receptors. In general, only a minute
fraction of these predicted concentrations will exceed the NAAQS. Therefore,
a modeling approach which identifies the elevated predicted concentrations as
reliably as a comprehensive analysis but in a more computationally efficient
manner, has the potential for providing substantial cost reductions. Any such
efficiency modeling approach would have to rely on some subset or sample of
the populations of 43,800 hours, 550 sources and 107 receptors.
Two general types of sampling could be applied to form a subset of the
hours, sources and receptors populations; probability sampling and directed
sampling. Probability sampling is best suited for determining typical or
average properties of a distribution. However, identifying the elevated
concentrations is a search for outliers or the upper tail of a distribution
and therefore, probability sampling is an inappropriate basis on which to
design an efficiency modeling approach. On the other hand, directed sampling
does have considerable potential with respect to efficiently defining the
elevated concentrations. Therefore, a directed modeling efficiency analysis
was developed, tested and used for the example urban area.
The RAM model efficiency analysis was performed in four steps (see
Figure 8-2). The objectives of the first three steps were to 1) identify
potential receptors in excess of the NAAQS, 2) identify critical days of
meteorological data, 3) define NAAQS exceedance area boundaries and establish
design concentrations.
In the first step of this analysis, the full inventory of 550 sources was
modeled using the 36 days of hypothetical meteorological data described in
Section 8.3.2. The model results provided maximum 1-hour average
concentrations at a grid of 95 receptors which were selected because they were
-87-
-------
RECEPTORS IN AREAS POTENTIALLY
IN EXCESS OF THE NAAQS
IDENTIFIED BY ANNUAL MODELING
OR NONATTAINMENT
PROBABILITY ANALYSES
CLUSTERS OF
SOURCES
(COMBINED POINT
SOURCES)
STEP 1
IDENTIFY CLUSTERS
OF CRITICAL
RECEPTORS
ALL
SOURCES
HYPOTHETICAL
METEOROLOGY
5 YEARS OF
METEOROLOGICAL
DATA
STEP 2
IDENTIFY CRITICAL
METEOROLOGICAL
' DAYS
STEP 3
DEFINE BOUNDARIES OF
AREAS IN EXCESS OF
NAAQS AND DESIGN
CONCENTRATIONS
ALL
CRITICAL
RECEPTORS
HOT SPOT
RECEPTORS
5 YEARS OF
METEOROLOGICAL
DATA
STEP 4
CONFIRM
DESIGN
CONCENTRATIONS
Figure 8-2. Directed Modeling Approach for 24-Hour Average Design Concentrations.
-88-
-------
within or near the area predicted by CDM 2.0 to be in excees of the NAAQS.
Estimates of maximum 24-hour average concentrations were then obtained by
multiplying the maximum 1-hour average concentrations by a factor of 0.4 (U.S.
EPA, 1981c).
Based on an NAAQS of 150 ug/m3 and the maximum 24-hour average
background concentration of 68 ug/m3, potential nonattainment areas may
exist wherever 24-hour average modeled impacts exceed 82 ug/m3. Potential
receptors in excess of the NAAQS were therefore defined as those receptors
where the maximum 1-hour average concentrations exceeded 205 ug/m
(82/0.4). Based on these assumptions, the initial RAM analysis results were
used to identify areas where additional receptors were needed to establish the
boundaries of the potential nonattainment area. A total of 38 such receptors
were required to complete the RAM modeling with hypothetical meteorological
data.
The results of Step 1, the hypothetical meteorological data modeling, are
shown in Figure 8-3. The figure shows the entire 133 receptor grid and the
maximum 1-hour average impact at each receptor. The figure also shows an
isopleth line depicting the areas where a 150 ug/m3 24-hour average PMio
NAAQS may be exceeded. In Figure 8-3, there are three hot spot areas where
maximum 1-hour average PMio concentrations are predicted to occur. These
three areas are the same as those predicted by the CDM 2.0 modeling analysis
(northeast, central, and southern hot spot areas).
The second step in the RAM model efficiency analysis focused on the three
hot spot areas discussed above. Within each of these areas, a cluster of
approximately 20 receptors was selected for further RAM modeling. For each
cluster of receptors, the 25 combined point and 25 area sources with the
greatest impact potential within the cluster were identified. For the
combined point sources, impact potential was determined as a function of
-89-
-------
Open rectangles enclose hot spot receptors.
Solid squares indicate monitoring sites.
YMAX - 39.5
Q1S0 0190 0192 0200
a 194
0169 Q185 Q186 Q184
0168 Q154 0179 Q164 Q190 o207 jarf83 D247 D252
0149 D191
0217 o181 o187 Q192
7
Q144
O232 Q256 Q223 Q185 D177
D182 0240/0214 Q185
J33610)
196 Q414 Q369
J5
0257 n681 o327 dj90
9 17
D168 0175 0169 D164 o207 o193 0187 o2^b O254 o683 Q292
"10
OlTB o175 Q172 g207 Q201 o266 o210 Q191 o246 o235 o22& o179
205
0170 0167X0229 O212|o25ll Q187 Q232 D26/4 Q168 o160 Q174 o195
oO^ *^
D189^0<97 0252 O228 O212
0180 W212 Q254 0204-B396 C
Q195 O204 o198 Q193
YMIN =6.5
Figure 8-3. Maximum 1-hour average RAM-modeled PM, concentrations produced using
hypothetical meteorological data
-90-
-------
emission rate, distance from the receptor and stack height. The selected area
sources were those enveloping or nearest the receptors. RAM modeling analyses
were then performed for each of the three groups of sources and receptors
using all five years of meteorological data. Based on these analyses, the 120
days that produced any of the five highest modeled 24-hour average PMio
concentrations were defined as critical days of meteorological data.
The third step of the RAM model efficiency analysis was performed using
the 120 critical days of meteorological data identified in the first two steps
as well as a limited number of receptors and all 550 sources. The receptors
included the three clusters of receptors discussed above plus an additional
15, which were placed as needed to establish the NAAQS exceedance area
boundaries.
Figure 8-4 shows the 6th highest 24-hour average PMio concentrations
calculated by RAM at the grid of 77 receptors using the 120 critical days of
meteorology. The figure also shows isopleth lines drawn to represent the
NAAQS exceedance area boundaries. Based on a 24-hour average PMio
background concentration of 68 ug/m3, the isopleth lines on Figure 8-4
show the spatial extent of the 24-hour average PMio nonattainment areas.
Figure 8-4 shows three hot spot areas in excess of the NAAQS. These areas
are the same as those identified in the first step of the efficiency
analyses. Hot spot receptors are evident in each of the three NAAQS
exceedance areas shown in Figure 8-4. In the northeast and central hot spot
areas, the maximum modeled 24-hour average PMio concentrations are 120 and
89 ug/m3, respectively. In the southern ,,hot spot area, two maximum
concentrations, 163 and 518 ug/m3 are apparent.
To evaluate the reliability of the directed modeling efficiency analysis,
a comprehensive analysis was conducted using the RAM model to calculate
concentrations at the 40 receptors within and surrounding the three hot spot
-91-
-------
Open rectangles enclose hot spot receptors.
Solid squares indicate monitoring sites.
TMAX - 33
062 Q67
D72
Q75
H4
i
061
064 068 069 Q71 074
5_
079 076
O61 Q71 Q70 Q72 D74 Q74
068 Q77
077 o78
16
072 071 Q90 Q79 o80 o72
D74 Q67
D/3
G68 Q73
| 089 |
4
D80 Q76 O70
O63 O68 O72 D81 Q79 08! D78
065 063
063
060
074
076
067
Q70 Q69
YMIN * 11
Figure 8-4.
Sixth highest 24-hour average RAM-modeled PM,Q concentrations produced
by 120 critical days of meteorological data wnich were selected by
modeling source groups and receptor clusters (ug/m3).
-92-
-------
areas shown in Figure 8-4. This modeling was performed using all five years
of meteorological data and all 550 sources. The comprehensive analysis
confirmed the boundaries of the NAAQS exceedance area produced by the directed
modeling approach and the magnitude of the sixth-highest concentrations at
three of the four hot spots shown in Figure 8-4. A larger sixth-highest
concentration was obtained only at the northeast hot spot receptor, and its
value was 126 ug/m3 versus the 120 ug/m3 calculated previously.
Therefore, the directed modeling approach produced definitive estimates of the
NAAQS exceedance area boundaries and the maximum locations of the
concentrations needed as design concentrations for the 24-hour average PMio
NAAQS. Since the first three steps in the directed modeling approach slightly
underestimated the sixth highest concentration at one of the four hot spot
receptors, the final step in any directed modeling analysis must be the
confirmation of the design concentrations at the individual hot spot monitors
using all five years of meteorological data.
On the assumption that emission reductions which produce concentrations at
or below the NAAQS at the hot spots will also produce attainment at all other
receptors, the four hot spot concentrations plus background were assigned as
the design concentrations for the 24-hour average PM10 NAAQS.
8.6 Control Strategy Selection
The design concentrations developed in the previous section indicate that
there is a PMio nonattainment problem for both the 24-hour and annual
averaging periods. Integrated planning is required to develop control
strategies with respect to the two averaging periods. As discussed
previously, the controlling standard is that standard for which the greater
emission reductions are required. The controlling standard cannot be reliably
established until control strategy testing is performed. However, for the
-93-
-------
purpose of control strategy development, a tentative determination of the
controlling standard was obtained by compiling source contribution listings
for each averaging period at the four critical receptors. Based on these
listings, proposed control strategies were derived by focusing on the sources
and source groups with the largest percentage impacts and greatest potential
for emission reduction.
Tables 8-1 to 8-4 present the annual and 24-hour average PMio source
contributions, proposed controls, and controlled contributions for the four
critical receptors identified in the example urban area. Table 8-1 pertains
to the northeast hot spot receptor and shows that nearby area sources and a
coke production plant are the largest contributors to the annual and 24-hour
nonattainment problems at this site. The table demonstrates that the controls
required to achieve compliance with the annual standard equal or exceed those
required for the 24-hour standard. Therefore, the annual standard governs the
controls required at the northeast hot spot receptor. The proposed control
strategy includes emission reductions of 75 percent at the coke plant and 25
percent for area sources. The proposed area source emission reductions could
be provided by implementing a variety of control measures including 1) street
sweeping; 2) limiting track-out from construction sites; and 3) paving
frequently traveled unpaved roads.
Table 8-2 shows that nearby area sources are primarily responsible for the
annual and 24-hour concentrations in excess of the NAAQS at the center hot
spot receptor and that as above, the annual standard dictates the level of
required controls. Most of the area sources will require the same 25 percent
emission reduction as was needed for the northeast hot spot receptor with the
exception of area source 145 which surrounds the center receptor. For this
area source, a 60 percent reduction in impact is required. While attempting
to develop control measures to achieve this large reduction, three unusual
-94-
-------
1
OO
UJ
_J
m
<
»-
V^
z
o
t-
oe.
\- oe
z o
UJ 1
u o.
o u
<_) UJ
QC
IS t-
K£
UJ I/I
a
UJ O
Of t-
UJ I/)
< UJ
ae t-
§§
z z
« UJ
>-
a
z ae
< o
u.
< t/1
3 UJ
Z 1-1
Z O
< UJ
t-
Ul <
X OS
K t-
V)
o
I- _l
1/1 2
Z H-
O Z
PS
m a
1-1 UJ
ae t/i
»- o
z a.
SS
a.
UJ
I*
IE
i/i
0
K
U
Ol
^
t
a
X
i
^
(M
^j
111
U
£
O
Z
£
X
i/i
X
a:
C
o o
»
V. -^ 01
4J U ^
C *J
o c
U 0
o
o i/i
OJ ^«
=
u c ^
c o
3
eS
01 «-
O J3 E
U H- ^
jj u a
C *> =
0 c
U O
u
o vi
OJ -
VI O
Si
o e
u o
0. U
I/I
o c
o 01 o
*J r *J
0>
« u C 5
<_) C 0 ^
3U
VI
at u
U 01
M
o 3
M Z
ai
U VI
Jo.
H-
OO ^ ^ 00 fH «*>
"~ 1
f\
t
vo «
0
4J
3
3
O
U
Ot AJ
-X C
U »-
10 o
CO 0.
i LO p^ o
=1
i i in tsi i
i i <« a\ f
*l s| 5
oo
eo
m in in ui m
oj o 4J te
o ,,
u
3
1
a.
-95-
-------
f^
1
00
UJ
<
^
VI
z
o
HI
f~
ce
t-
x ce
Ul O
U K
Z CL
O ui
O U
z ce
O
l-C I
IO O
Q (/I
Ul t
oo
< X
ce
ui ce
> UJ
z
DC U
=30
O
I X
w i
CM
ce
0 0
z u.
I/I
_J Ul
< HI
Z Ul
Z I-
ce
UJ H*
X l/>
1-
es
j_
m
t- a
3 UJ
CD
(M
t/1
OJ
£
a
X
i
X
.*
^
i
ce
41
O
id
4)
>
<
1_
R
41
>
1
in
o
CM
i
(J
I/I
C
TJ O
41 f ^
'- 3"e
o a \
ill
C J->
o c
U 0
u
O (rt
41 r-
VI O
O V.
O C
I- O
o. o
VI
o c
41 O
f~ ^J '^
O 3"
U J3 £
4J <
c u oi
o ^> ;:
0 C ~-
c o
i/i
c
a o
*I1
Sf"^
*i u S
o c £
u o
o
D V>
Ur-
vi a
o u
o c
U 0
a. u
vi
o c
0 41 O
IB O 3n
U U JS g
-? IB S
m u c '
u e a
= 0
M
41 U
U 41
11
z
4)
U VI
U 41
3 ex
o >>
ao * fiinaoii! KOICMI »
\O CM CM i i i inl ^
i 4) X H ** ' i i **l »<1 '
i c in in m i i t mj ml i
O CM CM CM CM! CM)
Z
\O CM «** i t i CMI an m
1 1 ^
o Ch i^neorx^r-jr-j o
in m -- *B o o CM eJ -4 e
CM i-j CM| m
I 41 !»<»*»<»«»< MM I
i c i in o in in in 10! inj i
O CM M3 CM CM CM CMJ CM|
Z
o on iiAf^o^ao ^l*?] ^
to «n ^ in o CM rJ kcJ *
«M f )«>( ID
^
1 T- CM ««« \O IO w (8
< ^. o ^
4J
3
(/I
o
O
Ok 4,)
^w c id
U «- 41 ^>
SO k. f-
0. < <
-96-
-------
CO
1
t- ae
< o
ae t-
h- a.
Z UJ
UJ <_)
(_) UJ
z ae
O
o t-
zS
O I/I
t/i H-
ui o
a x
UI Z
O oe
< UJ
ae z
UJ h-
> S
trt
ae
3 ae
o ui
z a.
i a.
« 3
0 Z
ae
-1 O
3"
Z I/I
Z UJ
* 3
UJ UJ
Z 1-
ae
o t-
1/1 -j
SS
M (-
K Z
3 O
2°
ae a
K- UJ
Z 01
o o
u a.
UJ K
II Q^
£
s a
0 Z
a
s
01
a
1C
u
01
"*
^,
c
z
1
~
V
Ol
n
c
1
^
*
I/I
I
<
K
01
C
^
01
j
u
n
Ol
i
ui
a
CM
V
2
^
VI
c
o o
Ol * -^
il^
i r '
C ^J
0 C
U 0
U
O vt
01 f
VI O
at
eg
a. u
VI
-o c
01 O
"3 3"
t£ E
C v. at
0 « 2
0 C
c o
T,S
£5 -v
a u n
§4J Z
C ->
u o
u
0 VI
Z
01
U VI
O X
ai«- «ni«-ii *1'~( "
\O rt ^ 1 \O 1 ' ' 1 ^"1 ^
1 1 ~ -
i o o i i uiiunii UT( S3 i
\O v<9 1 N (M C\4 Nl
CO VOV0IO «tlv0ll UTI «r) IM
v0 f^ f^ 1 ui i ii 1 ^"n rt
1 - 1 1 «M
r« 10 aa ~ loooaooo r oovflla o fM o o ON «j r^
i MM oj i i ** ** M M M M i
i va«a q i iinuiminuil it i
O\ O> ^ fM »M CM fM >
BO U i-
aa Q. < <
Ol
c
o
c
id
z
01
o
c
o
u
<
-97-
-------
.
t/»
z
o
h-(
^ tt
< 0
CK t-
i- a.
Z Ul
Ul U
<_) U
8"
O t-
o
O co
CO f-
Ul O
a x
Ul Z
0 ae
< Ul
ae x
Ul t-
^ ^3
< o
CO
ae
3 ae
CM
Ul
O X
Z t-
ae
-i a
< u.
3
Z CO
Z Ul
u
Ul Ul
ae
O K
*- CO
CO -I
38
1- Z
3 a
a u
ae a
h- ui
Sco
0
U a.
i! °-
§i
CO <
o
iE
OJ
o
10
01
^
^
c
x
1
CM
V!
Qi
^
a
i
f
^i
x
CO
4
ae
01
a
ca
u
01
>
<
1_
(O
01
>-
1
in
e
CM
E
c
c
o o
01 f ~~
IP
*j ^ S
C JJ
o c
0 0
u
13 in
01
in o
0 1.
a &
o c
1- O
a. u
in
o c
01 O
S-O'E
*j f ^
C U 01
O -J -
0 C
c o
3<_>
in
c
o o
01 *^
O J3 E
«. -^ -V
*J U ra
C *J _
0 C ~
U 0
u
V in
01 r-
VI O
Si;
o c
u o
a. u
o c
o oj o
01 r- ->
* *O 3 E
< c ^ S
r- a * ~-
IO 0 C
u c o
30
in
0> L.
U 4)
3 E
0 §
z
«l
g^
^
(/) h-
KO iiiiii i i i | m i fsij v v
\ iiiiiiv0\0vOvOi^O^*9< C] i linui m in! i
eo iiiiiivO\o^^-iirjr*i ON ui i CM ^- r^| coi so
vO iiiiiimfo^r^-ivcsii ^ ^ i ^- <*Mm| ao
1 * 1 1 in
CMCM p-cnn ON a* o^r^in inj voi in
in vvv vvoleOfM ON* oo
I M *«»« M »«>«»* X »*»« M M M M g I J$**J? J5 J3 !
ON(J>OO>a»a>CT> o CM CM CM CM! CM!
ocMC^cj\a>an ON ^- « « r> rs j CM
in CM CM invcMCMCMaoao> MM « CM CM! ON in
«M | in »H >H O">
U t-
01 at
o o
*> jj
^ a
CO CO
j;
g
u
a 4J
j< C J
c
ra
a
L.
ca
w
0)
ae
in 01
JJ U
C «-
|u.
C
Ol 1
a! a
c
*j
Ol 4?
Z co
>- X
10 C
u O
1z
S. C
u <
< a
-98-
-------
characteristics of this area source were identified. These unusual
characteristics are discussed below.
The three distinguishing features of area source 145 are:
1) The center hot spot and monitoring site 4 are contained within the
area source. Site 4 was the site where the greatest
overpredictions of measured concentrations were obtained in the
model evaluation analyses.
2) The emission rate for this area source was a factor of two greater
than the next largest area source in the inventory. A high level
of construction activity associated with this area source was
responsible for the elevated emission rate.
3) The emission release height modeled for this source was found to
be too low for the types of emissions in this area source.
Upon further investigation, the construction activity emission rate was found
to be in error and was corrected. In addition, an appropriate value was
inserted for the release height. These changes together with a 25 percent
emission reduction for this area source provided the necessary 60 percent
impact reduction at the center hot spot due to area source 145.
Table 8-3 shows that an iron ore handling facility and nearby area sources
are the primary contributors to concentrations in excess of the NAAQS at the
upper .southern hot spot receptor and that the required emission limits are
those associated with the annual standard. The proposed control strategy
includes emission reductions of 96 percent for the iron ore facility and 25
percent for area sources.
Table 8-4 shows that the three source types primarily responsible for
concentrations in excess of the NAAQS at the lower southern hot spot receptor
are a chemical manufacturing plant, an antimony smelting and fire retardant
manufacturing facility and nearby area sources. For this receptor, the
24-hour and annual standards require the same levels of emission controls.
Specifically, emission reductions of 96 percent for the chemical plant and
-99-
-------
antimony facility are proposed together with 25 percent reductions for area
source emissions.
8.7 Control Strategy Testing
To provide an efficient mechanism for testing the above set of control
strategies, the following plan was developed. The control strategy testing
modeling was performed using an "impact offset" approach in which the sources
with proposed controls were modeled in two configurations simultaneously. In
the first configuration, the sources under consideration were modeled with
negative emissions to reflect uncontrolled conditions, and in the second
configuration these sources were modeled with positive emissions to reflect
proposed control conditions. The modeling results thus provided the n.et
negative change in impacts produced by the emission controls. These negative
impacts were then summed with the uncontrolled impacts calculated previously
for-all sources.
The COM 2.0 and RAM dispersion models were used to estimate PMio
concentrations at all the receptors in the area in excess of the NAAQS shown
in Figure 8-4. All five years of meteorological data (1980 to 1984) were used
in this modeling. When the results of the RAM and CDM 2.0 modeling showed
continued nonattainment at the hot spot receptors, further control measures
were developed. The RAM and CDM 2.0 modeling was repeated until compliance
with the 24-hour and annual average PMio NAAQS was achieved at all the
previously identified receptors in excess of the NAAQS.
-100-
-------
9.0 INDUSTRIAL SOURCE EXAMPLE
9.1 Introduction
This section describes an example SIP development process for a monitoring
site located within an industrial area. Measurements at this site clearly
demonstrate exceedances of the 24-hour and annual average PMio NAAQS.
However, the causes of the elevated concentrations of PMio and the
corresponding control strategy are not as evident. Based strictly on the
emission inventory, a large steel mill and traffic-generated resuspended road
dust appear to be potentially large contributors to observed PMio
concentrations. However, the level of confidence associated with emission
inventory-supplied source contribution estimates is insufficient for reliable
control strategy development. Therefore, an investigation, which included
data collection and modeling efforts, was conducted in order to apportion the
source impacts with the level of confidence necessary for making control
strategy decisions.
The industrial site PMio SIP development example presented in this
section was compiled primarily from experience gained from working on an
individual site. The location of the site is unimportant to the objective of
the example and is therefore not identified. In addition, the data base used
for this example was modified to better illustrate the SIP development process.
9.1.1 Overview of the Source Apportionment Study
A combined receptor and dispersion modeling study was implemented with the
objective of identifying and quantifying the impacts of the sources which
contributed to the violations of the PMio NAAQS recorded at the monitoring
site. To supplement the emission inventory in providing a basis for
completing the data gathering efforts, pollution concentration roses were
-101-
-------
prepared using the two most recent years of data for the eight TSP monitoring
sites located near the PMio monitoring station. The pollution roses
indicated that the major sources of TSP were located in the quadrant south of
the PMio monitor. This suggested that additional emphasis should be placed
on characterizing steel mill related contributions because the steel mill is
the largest source of particulate emissions located nearby and south of the
monitoring site (see Figure 9-1).
9.1.2 Data Collection Tasks
This section describes the data gathering tasks which were conducted in
order to satisfy the input requirements of the receptor and dispersion
models. The starting point of the data gathering effort was reviewing the
available data which consisted primarily of: 1) ambient PMio samples
collected for one-year on an every sixth day schedule using quartz fiber
filters and a SSHV sampler; and 2) hourly measurements of wind direction, wind
speed and temperature at the PMio monitoring station. In addition, for
one-month a dichotomous sampler was in operation at the monitoring site and
collected several 24-hour coarse and fine fraction samples on Teflon filters.
The results of the PMio monitoring program provided an annual average PMio
concentration of 65 ug/m3 as well as six exceedances of the 24-hour
average PMio standard of 150 ug/m3.
The initial data collection effort performed for this investigation was a
site visit to compile a microinventory and obtain bulk samples from local
emission sources. The results of the microinventory were combined with
published emission factors to calculate emission rates for input to a
dispersion model.
As part of the site visit, bulk samples of material were collected from
six sources in the general vicinity of the site which had been identified as
-102-
-------
Numbers represent fugitive
dust volume sources.
PAVED STREETS
UNPAVED AREAS
COAL STORAGE PILES
RAILROAD TRACKS
1/4 MILE
4OO FT
SCALE
STEEL MILL
Figure 9-1. Schematic Diagram of the Industrial Source Example Study Area.
-103-
-------
potentially important contributors to PMi0. The sources which were sampled
included: a steel mill (blast furnace, coke oven, and basic oxygen furnace);
two road shoulders (one near the monitoring site and one near an inoperative
lead plant); and an agricultural processing plant (potash, corn gluten pellets
and soybeans). For each sample an aliquot was then aerosolized in a dust
chamber and collected onto a quartz filter by a SSHV sampler and onto Teflon
filters by a dichotomous sampler (U.S. EPA, 1984b).
Application of receptor modeling required the chemical characterization of
the source and ambient samples. Analysis costs limited the number of filters
which could be analyzed. The receptor model input requirements mandated the
analysis of all the resuspended source samples. Two subsets of the ambient
filters were selected for analysis. Subset A contained the filters from 20
sampling days which were selected to be representative of the annual average
conditions at the monitoring station. The average PMio concentration
calculated from the filters in subset A was 67 ug/m3, which compares well
with the annual average calculated from all the filters (65 ug/m3).
Subset B contained the samples from the six days on which exceedances of the
24-hour average NAAQS «rere recorded. This subset was chosen to guide control
strategy development related to the 24-hour standard.
Two multi-elemental characterization techniques were applied to the source
and ambient filters. X-ray fluorescence (XRF) was performed on the Teflon
filters and Plasma Emission Spectroscopy (PES) was employed for the quartz
filters. The elemental carbon content of the samples was determined by
optical attenuation. In addition to the chemical characterization procedures,
optical microscopy was applied to several of the ambient samples. Optical
microscopy, which provides reliable particle identifications and
semi-quantitative source contribution estimates, was used as a QA check on the
-104-
-------
receptor modeling source identifications and to aid in the receptor/dispersion
model reconciliation process.
9.2 Data Preparation
9.2.1 Dispersion Modeling Data Preparation
The inventory of all the point sources in the two counties surrounding the
PMio monitoring site was obtained from the State agency responsible for
local air pollution control regulations. A total of 882 point sources were
included in the inventory. In order to reduce the point source inventory to a
more manageable size, the following set of emission-size versus distance from
the receptor criteria were developed:
Distance (Kilometers)
<5.0
5.0<10.0
10.CK20.0
>20.0
Emission Rate (grams/second)
>0.01
>0.1
>1.0
>3.0
Application of the above criteria on a composite plant basis resulted in an
inventory of 140 point sources for input to the dispersion model. The point
source inventory was modified for use in modeling PMio impacts by applying
source specific PMio/TSP emission ratios to the TSP emission factors which
were originally contained in the inventory.
The State inventory did not contain area or fugitive dust sources. As a
partial solution for this deficiency, the fugitive dust sources compiled in
the microinventory were included as volume sources in the input to the
dispersion model. Emission rates for the fugitive dust sources were estimated
with respect to PMio emissions (see U.S. EPA, 1985a and 1986c).
The hourly wind direction, wind speed, and temperature measurements
collected on-site were used directly for model input. Also required for
-105-
-------
iispersion modeling were Pasqui11/Turner atmospheric stability data and mixing
height estimates. The calculations to obtain the stability data used for
model input were performed with the CRSTER preprocessor program using the
on-site wind speed measurements in combination with ceiling and cloud cover
data obtained from the nearby NWS station as well as the latitude and
longitude of the PMio monitoring site. Mixing heights were calculated by
applying the CRSTER preprocessor to NWS data from the nearest surface and
upper air stations.
For this study, background concentrations were defined as that portion of
the measured ambient levels that is not attributable to emissions within the
study area. To estimate the appropriate background concentrations, a
pollution rose was developed using data from TSP monitoring stations located
outside the study area. Data were used only from the days having persistent
winds blowing into the study area from the direction of the background TSP
stations. An annual average TSP concentration of 26 ug/tn3 can be
attributed to sources outside the study area. Application of the
site-specific PMio/TSP ratio (0.57) yields an annual average background
PMio concentration of 15 ug/ra3. In addition to the annual average,
background concentrations were also estimated as a function of wind direction
for use in the 24-hour average modeling.
9.2.2 Receptor Modeling Data Preparation
Receptor modeling was performed for this study using the Chemical Mass
Balance (CMS) model (U.S. EPA, 1987c). The CMB model requires an input file
containing the measured ambient concentrations of the elements for which the
samples were analyzed. This requirement was fulfilled by transforming the
results of the ambient filter analyses into the format specified by the CMB
model. In addition, the CMB model requires a file containing the source
-106-
-------
compositions reported as the elemental mass fractions. A source composition
file in CMB specified format was compiled. The file contained the elemental
composition of the resuspended local sources as well as a number of source
profiles which were extracted from EPA's Source Composition Library (U.S. EPA,
1984a).
9.3 Model Evaluation Analyses
9.3.1 Dispersion Modeling Procedure
The Industrial Source Complex Short-Term (ISCST) model (U.S. EPA, 1986d)
was used for this investigation. ISCST was used because it is applicable to
industrial sources and it contains several features that provide increased
efficiency to the source apportionment process. Two of the features of ISCST
which proved to be advantageous were its ability to: 1) model the
microinventoried fugitive dust sources as volume sources; and 2) calculate the
combined impact for selected groups of sources. This latter feature greatly
decreased the manipulations which were necessary to transform the dispersion
and receptor modeling results into a format which provided a logical basis for
comparisons. Values for the user-selectable parameters of ISCST were
determined in accordance with standard regulatory practice (Guideline on Air
Quality Models (Revised), (U.S. EPA, 1986b).
The emission inventory was used to develop a tabulation of the types of
materials emitted by each source. The many types of emitted materials were
then associated with one of twenty general categories because ISCST is limited
to predicting impacts for a total of twenty source groups. Each of the 140
point sources and 25 volume sources was assigned a code corresponding to one
of the twenty categories.
-107-
-------
9.3.2 Receptor Modeling Procedure
Two receptor oriented approaches were applied for this study: the CMB
model and the optical microscopy technique. CMB modeling analyses were
performed on the filters contained in subsets A and B described in Section
9.1.2. The modeling procedure consisted of obtaining a "best" fit solution by
systematically adding and removing sources and elements from the mass balance
calculation in accordance with the Protocol for Applying and Validating the
CMB Model (U.S. EPA, 1987d). Termination of the iterative fitting process at
the "best" fit solution was determined on the basis of a series of summary
statistics reported by the model as well as the analyst's understanding of the
airshed.
Two SSHV samples, two coarse fraction dichotomous samples and one fine
fraction dichotomous sample underwent optical microscopic analysis to confirm
the CMB source identifications. In addition, the results of the particle
counting performed as part of the microscopic analysis were used to calculate
semi-quantitative source contribution estimates.
9.3.3 Comparison of Receptor and Dispersion Model Results
The source contribution estimates provided by the CMB and ISCST models for
subset A (annual average) and subset B (samples violating the 24-hour NAAQS)
are compared in Table 9-1. The results of the two models display reasonable
agreement in two categories: 1) the background estimates used by ISCST are
approximately equal to the secondary sulfate values estimated by CMB; and 2)
resuspended road dust is listed as a major source of PMio by both methods.
In addition, the similarities and diff erences''between CMB and ISCST are
consistent between subsets A and B. Aside from the relatively few
similarities which have been enumerated, many differences between the CMB and
ISCST results are evident. These differences were examined and then.
-108-
-------
TABLE 9-1
COMPARISON OF INITIAL EVALUATION RESULTS BY ISCST AND CMB FOR
SUBSETS A AND B (ug/m3).
Source Categories
Oil & Gas Combustion
Wood-fired Boilers
Coal Combustion
Coking Operations
Blast Furnace
Basic Oxygen Furnace
Coal Handling
Agricultural Prod. Handling
Paint Production
Cement & Limestone
Sand & Bentonite
Aluminum Production
Zinc Processing
Oil Refining
Fertilizer
Tire Production
Motor Vehicle Exhaust
Road Dust & Soil
Secondary (NH4)2S04
Background
Total
Measured
Subset
ISCST
1.6
0.2
0.6
1.4
M
0.3
1.1
1.9
1.2
0.8
0.6
M
15.0
45.4
67.0
A
CMB
__ ,
3.2
2.8
8.6
12.0
3.8
2.4
30.0
15.0
77.8
67.0
Subset
ISCST
2.0
0.8
1.2
4.0
J,oj
0.6
1.6
0.8
1.9
1.7
0.4
0.6
1.6
0.8
2.2
> 34.8 <
32.0
89.0
162.0
B
CMB
5.6
8.0
23.2
3.6
10.4
2.4
79.2
27.4
189.8
162.0
-109-
-------
reconciled following the eight step procedure described in the Protocol for
Reconciling Differences Among Receptor and Dispersion Models (U.S. EPA,
1987b). In the following Section, the differences between the CMB and ISCST
source contribution estimates will be identified and their reconciliation
summarized on a case by case basis. A more detailed description of the
example can be found in the aforementioned protocol (U.S. EPA, 1987b).
9-3.4 Reconciliation of CMB and ISCST Results
Case 1: A significant disagreement exists between the CMB and ISCST
estimates of the coal combustion impacts. What is the cause
of this disagreement and how can the difference be
reconciled?
The first step taken in reconciling the coal combustion impact estimates
was to review the ambient and source composition data for errors which could
potentially invalidate the CMB results. This review uncovered no obvious
errors in the CMB input data related to coal combustion. The next step was to
examine the comparisons between CM, CMB and ISCST which are shown in Tables
9-2 and 9-3. These tables show that OM predicts coal combustion impacts in
reasonable agreement with CMB and larger than ISCST. The emission inventory
was then reviewed with respect to coal combustion sources. The review
identified four major coal combustion facilities. For the closest of these
sources to the receptor, the emission inventory contained an erroneously high
value for the efficiency of the emission controls. A correct value was
obtained and ISCST was then re-run with the corrected inventory providing much
closer agreement between CMB and ISCST with respect to coal combustion impacts.
Case 2: The combined contributions of motor vehicle exhaust and
resuspended road dust, as estimated by CMB and ISCST,
disagree by over a factor of two. What is the cause of the
disagreement and how can the difference be reconciled?
-110-
-------
1/1
J
a.
Z
a:
GO
°
Z
a.
en
00
2
"
"~
ex
CO
0
t-
o
X
t
ce
0
14.
in
>-
^
VI
u
a
^j
UJ
a
s
M
^O
OC
U
O-
u. f- U
Ej a
g >
-3 X 00
(/) Z
VI I/to
c
o
... u
4J tV O U-
15 in
a x o o
Z v, !tj b 2.
0,-.-
0 Z
c
o
- -a vi
U i U
0) 01 ra
a.-o a
vi o E
*- Z w
O
Ol t/l
i X
41 ^~
0 0)
Z u
u
09 3
u«5i
>
x
a. i/i
o at
u -*
VI U
o o
u a
u ot
Z «
u
10
I/I U
u u
0) 3
..X O
V) «/)
0
lOMSIII'tf'lll11'1
icnvoiiiaoiiiiiii
L VO 1 1 VO 1 CM 1 1 1 1 1 1 1
lOOIlONlenlllllll
«n .
IIOI*'*'.OIIIIIII
i lOjicMvotni i i i i i i
r »
OieOQ-vDCMIOOOl 1 1 » 1
i »r ui zzinua. iiata
(_)<_)<- X0.4JXC
e- I 1 X C i-
i t-za>« t ta « xg
01 a, . a, oj tauuE ia3
3 3 4J 4J 4J O 01 0. S r->
U. U. I/I I/I (XI (J < «/l O (J <
VI
§ s
4J -a 01
VI O C
3 C C UOO)
g o o a. vi *j 4J
8 13 L i- o y "c
u vi vi i i ia 3 E o
331 1 U -O ->- 4J
VI A -O 1 1 3 O -1 C
385' ' ^ij £|
0 f^ U 4J C * *^
o to .x a> -giac3
>-po o 4j oatiatvoiiar-
o5u <_)>V) >_> < a. u -j to <
IIIIIIIOOCMIOI
iiiiiiioinitni
"*"
i i i i i i 1001*1
1 1 1 1 1 1 1. (Ml^l
CM
iiiiiiifl-tMimi
iiiiiiivinirvi
CM CM
I I 1 I I lOfMOOCMeni
t i i i i i» (McninCMi
(M CM
1 1 1 1 1 ICMOa*TCMeni
1 1 1 1 1 ICMCMf^tniAl
.
1 1 1 1 1 ICMOCMIV0I
i i i i i i en in i r» i
* CM
x*9«cflCMIV.9l ICMI IO
CMftf'imiaoi irMi i--
cn en
T
4J O
vi co
3 CM
to -*
S, V
X X
1 1 1 1 1 1 1 UJ 1 X 1
1 1 1 1 1 1 1 4J 1 ~ 1
1 1 1 1 1 1 1 O) 0) I 1
Illlllli OJIXI
*- I- U
ft 4J 10
§1/1 -a
c
u o
*J (8 U
30) 0)
< Z (/I
I/I
4J
C C
EC O
o
A VI U I/I
ia u o in >
u. <_> uj z vi
o
0) g *D <.<5 U
at u 3 c ai
C * " *" I/I <
i i i ^ i l k. v 2 1- i
i l i C i l ea X i
III-.-II -*-UUI
1 1 Itt-l IUUU-*-lOI
ai «»-- o> -o
ac u-u- u- o c
i ia ea ta'o u
-.- u u u >- ai
O K »- I- ca cy>
VI
vi
VI X
i i X
U *> 0) 4J
i- ia c u i i ii
E u- eu -o i i 4J i i c
£ X 01 N O 1 1 VI 1 13
o u ac u 3 o
0) a. o u
C i- Ol
cviulviuu ia u
*.**.O)^0)>* ' O 10
N z ac z u. i- ac ca
. i »*> i i
r* l in i i
O^ vO
09 OO 1 II
OO O> 1 1 1
00 17*
U) 1 1 1 CM
en i i i en
ON V0
r
r^ ao i ii
Ch O> 1 1 1
CfA QA
1 « 1 1
cn i in i i
ui to
VO 1 1 1 CM
en l i i en
*) vo
.^ »
00 00 W CM CM
* on in in en
o o> vO vO vO
. .. ^
i i l l i
i i l l i
i i l i i
i i i i i
l l l i l
l l 1 l i
l 1 1 l 1
l l l l i
e
GL O
JjE
^£"X
a. a. o to
o-o -o-o
0* 0> 0) 01
u u u u
f-3333
IO I/I VI VI VI
*> IO cO »8 IO
-111-
-------
1
a
1- O
IO U
u in
c. O
4-1 k.
a. u
O £
X
c
o
- -o
in oj
k. <
OI 0
Q.TJ
in c
^ y
a
4-1
O
£i
"5
a
o
OX
S u
./ *-
vi JL
coo
4, jj
in o
V. £
IO O
o -
u a
4J
01 0
c s.
u. >-
a
>
S
Ul
in
o
10
t
I-H
01 in
c ai
«- 0.
i- X
01 r-
0 01
tD 3
X O
^
X
a in
O OI
u >-
in u
o o
k. 01
U 01
- 4J
(O OI
u u
£!
O 01
OI
c
"SI VI
o a
0 3
Z O
*- 01
VI U
V. k.
01 3
a. o
VI I/)
O
1 vO 1 1 1 O 1
i o i i i in i
1 00 1 1 1 1 1
, VO 1 1 I , 1
I I 1 I 00 VO 1
1 1 1 1 O 1
« 1 CM VO O 1
O 1 *- O 1
O 1 00 vO 00 GO 1
« i ^» o CM in i
i i as «r o «s- i
1 I - 01 <) M 1
«r i i i i i i
O 1 1 1 1 1 1
4J
« oi
0. 10
c
k. k.
O V
1 0, I U C
it. i
- k. £ 3 X
OJ £ < "" °
1 4J O
O r- 01 VI -^
o e Or- a
1515
o o
J«J
u
§X k.
0, X
a Q. oi
U 4J .X
m k. o
(-J 10 4J O
Ok k.
- « -
f\ JQ r- i- 1
! S^zz !
r- t- X I**
01 0> « «
3 3 4J 4J *J
U. u. 01 O1O1
c
o
4J
Sec
f55
4J 4J
U in in i i
33i i
in .0 .a i i
3SS '
u u
O r- «l1u
>- e ie j< a*
»- O O O 4J
O3 U U01
i i i
i i i
i i i
i i i
i i i
i i i
O vO 1
CM 1
CM \O 1
CM *>J 1
0 « 1
m n i
1 O 1
i i
i i i
i i i
i i i
i i i
a
^B
o
a||
f 4J
O "O f
n u £
X
r- Id
10 k. k.
O OI Q.
O <01
4-*
U
3
ultural Pro
Products
U 4J
gl|
u < at
i i i i
i i i i
i i i i
i i i i
. i i i
i i i i
i i i i
itii
i < i i
i i i i
i i i i
iiii
I 1 \O I
1 1 O 1
i i i i
i i i i
i i i i
i i i i
a
c
ot "in
c in
o * o
IO 1 TJ O-
^|g
1 II
U O<
01
t. Limeston
Bentonlte
num
c -«-
91 -o -o E
issl
O-IOK
1
1 1
1 ,
1 |
I I
I I
1 I
I 1
I I
1 1
I I
I |
1 1
1 1
1 I
1 i
1 I
1 1
Chemicals
u u
C VI
NZ
i i
t ,
i i
1 '
i i
, ,
i i
i i
i i
i i
i i
i i
i i
, |
i i
i i
i i
i i
,f
i "e
1 Ik
£
r*
O
in
4J
ery Catalys
Refinery
C *
-r- U
%z
OCX
i i o vo i oe i
1 1 1 CM O 1 O I
CM
1 1 1 10 1 1 i in i
1 1 vO IO W I » 1
i i o in i ^ i
1 1 O CM rs| 1 O I
1 1 f O 1 '
^
1 1 CM « CM 1 e 1
1 1 CM 0 1 00 1
1 1 1 1 CM 1 1 O
i i i i i i r~
*r ^m
4J O
VI (/I
3
< Z (/)
VI
E** c
c o
.a in o in
M k. o vi
k. id k. -i- o
tk o ui z vi
0
01 S ^ tf k.
k. 5 c w
II IO X '
i i o o o-<- 1 O I 1
CM n
co crv i i i
-2 ' ' '
.- 1 1 1 00
00 1 t 1 O
O"i O> I i I
ON O^ i I I
^ »_
CM 1 CM 1 1
O 1 O 1 1
00 1 1 1 00
O 1 1 1 O
r^ r^
CSI O^ CM - 00
o o> o o o
vO *^ co u) r^
Illll
11111
11111
11111
11111
Illll
Illll
11111
e
^
a. o
o
TJ£
a» in o
Z*Z -^i/5
a. a. o 01
o -a -a -a
01 01 01 01
k. k. k. k.
-3333
ie in in in in
4J id id *
-------
The first step taken to reconcile the disagreement between CMB and ISCST
with respect to the combined impact of vehicle exhaust and road dust was to
re-examine the comparisons between OM, CMB and ISCST. In this case, OM was
found to agree very well with ISCST. The CMB input data were then evaluated
for potential errors related ,to estimating road dust and vehicle exhaust. The
coarse fraction filter on which the resuspended road dust source sample was
collected appeared to have lost a substantial fraction of the sample between
the time it was weighed and the time it was analyzed. This potential problem
was identified by performing the following test: 1) convert the mass fractions
of the major elements to mass fractions of those elements as their assumed
oxides (i.e., convert mass fraction of Si to mass fraction of SiOz); 2) sum
the mass fractions of the major species reported as oxides; 3) compare the sum
of the mass fractions with a value of one. The carbon content of the road
dust was known to be approximately five percent. Therefore, the sum of the
oxides of the major species would be expected to equal approximately 0.9.
When this test was performed on the results of the analysis of the coarse
fraction road dust sample, the sum of the mass fractions of the oxides was
approximately 0.45. This indicated that the mass of the filter was high by a
factor of two. The filter was then re-weighed and the gross filter mass was
found to be lower than the measurement which was made immediately after
resuspension. The probable cause of the difference was the loss of particles
from the overloaded coarse fraction filter. The mass determined during the
re-weighing was used to revise the road dust source composition profile. The
revised mass fractions were higher than the original mass fractions and would
therefore decrease the CMB estimated road dust impacts.
The CMB model was re-run with the revised road dust profile and good
agreement was now found between CMB and ISCST with respect to the combined
impacts of road dust and vehicle exhaust. The CMB estimated impacts of the
-113-
-------
other sources remained unchanged by the revised road dust profile. Finally,
the other source samples were re-weighed and no other cases of significant
mass loss were identified.
Case 3: CMB estimates wood-fired boilers are a much larger
contributor to PMio than does ISCST. What is the cause of
the disagreement and how can the difference be reconciled?
The first step in reconciling the CMB and ISCST estimates for the
wood-fired boiler impacts was to review the CMB input data. The review did
not identify any obvious errors in the data. Next, the OM results were
examined. In this case, OM was of little assistance because OM did not
identify any impact from wood-fired boilers while CMB and ISCST both did. The
emission inventory was then evaluated in terms of wood-fired boilers and
revealed that an erroneous emission rate was listed in the inventory for one
of the wood-fired boilers. Therefore the inventory was revised. ISCST was
then re-run and good agreement was found between CMB and ISCST with respect to
the wood-fired boiler source category.
Case 4: There is a big disagreement between CMB and ISCST for steel
mill related impacts (i.e., coking operations, blast"furnace
and basic oxygen furnace source categories). What are the
causes for the disagreements and how can the differences be
reconciled?
The steel mill source is of additional interest due to the fact that there
is a big disagreement between CMB and ISCST for each of the three main PMio
emitting activities within the steel mill. This disagreement results in CMB
predicting that the steel mill is the largest industrial source of the PMio
levels observed at the receptor site while ISCST predicts that the steel mill
is a relatively minor source. This discrepancy will have a big impact on
control strategy development and therefore must be reconciled very
conclusively.
-114-
-------
The first step in the reconciliation process was to review the CMB input
data. The review did not identify any obvious anomalies in the CMB input
data. In fact, the steel mill related source profiles were judged to be of
very good quality due to the close agreement which was found between the PES
analysis of the resuspended samples which were collected on quartz fiber and
the XRF analysis of those collected on Teflon. The next step was to examine
the impacts estimated by OM for the steel mill. In this case, OM was found to
agree much better with CMB than with ISCST. The emission inventory was then
reviewed with respect to steel mill emissions. The stack emission rates were
found to be in good agreement with emission rates for similar activities at
other steel mills. However, the review revealed that the emission inventory
did not contain any emission factors associated with fugitive emissions from
the steel plant. This omission was viewed as a potentially serious deficiency
and therefore a site visit was conducted to re-assess the fugitive emission
rates at the steel mill. Substantial fugitive emissions were found to be
associated with the coking operations, blast furnace and basic oxygen
furnace. ISCST was then re-run with the new fugitive emissions for the steel
plant included in the inventory. This resulted in very good agreement between
CMB and ISCST with respect to the steel mill contribution.
9.3.5 Post-Reconciliation Comparison of CMB and ISCST Results
The previous section presented the reconciliation of the CMB and ISCST
source impact estimates and resulted in a number of revisions to the input
data used by the models. Following reconciliation, CMB and ISCST were re-run
for subsets A and B using the revised data. Very good agreement now exists
between the source impacts estimated by CMB and ISCST for subsets A and B.
ISCST was then run using five years of meteorological data to obtain estimates
of the annual average PMio and the six highest values. The results provided
-115-
-------
by ISCST are shown in Table 9-4. ISCST predicts that both the annual average
and 24-hour PMio NAAQS will be exceeded at the monitoring station. Steel
mill related activities are the major contributors to PMio at the receptor
site. The second largest contributor is resuspended road dust. The six
highest predicted PMio concentrations all occur during persistent south
winds. The next section will discuss the development of control strategies to
bring the receptor site into attainment of the PM10 NAAQS.
9.4 Control Strategy Development
The first step in control strategy development was determining the annual
and 24-hour average PMio design concentrations for the monitoring station
that must be reduced to the level of the appropriate NAAQS. For this study,
the annual average design concentration was calculated as the mean of the
PMio concentrations recorded during the one-year PMio sampling program (65
ug/m3). To determine the 24-hour average design concentration, the table
look-up procedure (U.S. EPA, 1987a) was used in combination with the results
of the five years of ISCST simulations. This approach required the use of the
sixth highest modeled PMio value exceeding the 24-hour NAAQS to be used as
the design concentration. Table 9-4 shows the 24-hour average design
concentration (183 ug/m3) and gives the corresponding source contributions.
After establishing the design concentrations, emission limits must be set
on the basis of the NAAQS (annual or 24-hour average) which requires the most
stringent set of controls. For this study, the total reductions in PMio
concentrations required to attain the annual and 24-hour average NAAQS were 23
percent and 18 percent, respectively. Table 9-4 indicates that the relative
contributions of the major sources are similar for the annual average and
24-hour average design concentrations. This suggests that adoption of a set
of emission limits designed to attain the annual standard should in turn bring
-116-
-------
TABLE 9-4
SOURCE CONTRIBUTIONS ESTIMATED BY ISCST USING 5 YEARS OF
METEOROLOGICAL DATA, (ug/m3)
Annual Sixth-Highest
Source Categories Average 24-hr Avg. Concentration
Oil & Gas Combustion 1.4 1.7
Wood-fired Boilers 3.0 6.0
Coal Combustion 2.0 7.5
Coking Operations 8.7 27.6
Blast Furnace 10.1 38.0
Basic Oxygen Furnace 2.7 14.8
Coal Handling 0.3 0.4
Agricultural Prod. Handling 0.9 0.4
Paint Production 0.1 1.1
Cement & Limestone 1.6 2.1
Sand & Bentonite 1.0 0.7
Aluminum Production 0.1 0.2
Zinc Processing 0.6 0.6
Oil Refining 0.3 2.1
Fertilizer 0.1 0.9
Tire Production 0.4 1.5
Motor Vehicle Exhaust )
> 12.7 45.4
Road Dust & Soil )
Background 15.0 32.0
Total 61.0 183.0
-117-
-------
the monitoring station into compliance with the 24-hour standard. To ensure
that this was the case, emission limits were developed to attain the annual
standard and then were evaluated with respect to the annual and 24-hour
standards.
The total reduction (TR) to achieve the annual standard is given by:
TR (ug/m3) = PMio Design Concentration - PMio NAAQS
which for this example leads to:
TR = 65 - 50 = 15 ug/m3.
From Table 9-4, on an annual average basis the contributions of the major
sources to PMio are seen to be:
Source Contribution to PMio
Wood-fired Boilers 3.0 ug/m3
Coal Combustion 2.0 ug/m3
Coking Operations 8.7 ug/m3
Blast Furnace 10.1 ug/m3
B.O.F. 2.7 ug/m3
Road & Soil Dust 12.7 ug/m3
& Veh. Exhaust
On the basis of available technology, cost-effectiveness and
enforcability, the following set of reductions in source contributions were
derived:
Source Individual Source Reduction
Wood-fired Boilers 1.0 ug/m3
Coking Operations 5.0 ug/m3
Blast Furnace 6.0 ug/m3
B.O.F. 1.0 ug/m3
Road & Soil Dust 2.0 ug/m3
& Veh. Exhaust
TR = 15.0 ug/m3
-118-
-------
The percent emission reductions corresponding to the above list of absolute
source reductions are:
Source Percent Emission Reduction
Wood-fired Boilers 33%
Coking Operations . 57%
Blast Furnace 59%
B.O.F. 37%
Road & Soil Dust 16%
& Veh. Exhaust
The total reduction to achieve the 24-hour NAAQS is:
TR = 183 - 150 = 33 ug/m3 .
From Table 9-4, the contributions of the major sources to the PMio
concentration on the sixth highest modeled day are:
Source Contribution to PMio
Wood-fired Boilers 6.0 ug/m3
Coal Combustion 7.5 ug/m3
Coking Operations 27.6 ug/m3
Blast Furnace 38.0 ug/m3
B.O.F. 14.8 ug/m3
Road & Soil Dust 45.4 ug/m3
& Veh. Exhaust
Application of the percent emission reductions derived to attain the annual
standard to the sixth highest day contributions provides the following
individual source reductions:
Source Individual Source Reductions
Wood-fired Boilers 2.0 ug/m3
Coking Operations 15.7 ug/m3
Blast Furnace 22.4 ug/m3
B.O.F. 5.5 ug/m3
Road & Soil Dust 7.2 ug/m3
& Veh. Exhaust
TR = 52.8 ug/m3
-119-
-------
This demonstrates that the emission limits which were derived for the annual
average case should also ensure the attainment of the 24-hour NAAQS. The next
section will describe the testing of the proposed control strategy with
respect to the annual and 24-hour NAAQS.
9.5 Control Strategy Testing
To test the control strategy, the proposed emission reductions were
applied to the emission inventory and the ISCST model was re-run with five
years of meteorological data. The average of the five predicted annual
averages was 49.6 ug/m3 and only two exceedances of the 24-hour NAAQS were
predicted for the five year period. The control strategy was therefore
sufficient to achieve the attainment of the annual and 24-hour average PMio
NAAQS.
-120-
-------
10.0 REFERENCES
1. Dzubay, T.G., R.K. Stevens, G.E. Gordon, I. Olmez, A.E. Sheffield, and
W.J. Courtney, 1987: "A Composite Receptor Method Applied to Philadelphia
Aerosol", submitted to ES&T.
2. Engineering-Science, 1984: Development of an Emission Inventory for Urban
Particle Model Validation in the Philadelphia AQCR, prepared for U.S. EPA,
ESRL.
3. Holzworth, G.C., 1972: Mixing Heights, Wind Speeds, and Potential for
Urban Air Pollution Throughout the Contiguous United States, AP-101. U.S.
EPA, Research Triangle Park, NC 27711.
4. NEA, Inc., 1982: Philadelphia Airshed Aerosol Study Summer, 1982 -
Sampling Site Characterization and Source Inventory Survey, prepared for
U.S. EPA, ESRL.
5. PEDCo Environmental, Inc., 1983: The 1982 Philadelphia Aerosol Field
Study Data Collection Report, prepared for U.S. EPA, ESRL.
6. U.S. EPA, 1973: Guide for Compiling a Comprehensive Emission Inventory
(Revised).
7. U.S. EPA, 1975: Guidelines on Air Quality Maintenance Planning and
Analysis Volume 7; Projecting County Emissions.
8. U.S. EPA, 1981a: Receptor Model Technical Series, Volume I, Overview of
Receptor Model Application to Particulate Source Apportionment, EPA
450/4-81-0163.
9. U.S. EPA, 1981b: Receptor Model Technical Series, Volume II, Chemical
Mass Balance, EPA 450/4-81-016b.
10. U.S. EPA, 1981c: Regional Workshops on Air Quality Modeling; A Summary
Report, EPA 450/4-82-015.
11. U.S. EPA, 1983: Receptor Model Technical Series, Volume IV, Summary of
Particle Identification Techniques, EPA 450/4-83-018.
12. U.S. EPA, 1984a: Receptor Model Source Composition Library, EPA
450/4-85-002.
13. U.S. EPA, 1984b: Receptor Model Technical Series, Volume V, Source
Apportionment Techniques and Considerations in Combining Their Use, EPA
450/4-84-020.
14. U.S. EPA, 1985a: Compilation of Air Pollutant Emission Factors, AP-42.
15. U.S. EPA, 1985b: Receptor Model Technical Series, Volume VI, A Guide to
the Use of Factor Analysis and Multiple Regression (FA/MR) Techniques in
Source Apportionment, EPA 450/4-85-007.
-121-
-------
16. U.S. EPA, 1985c: CDM 2.0 Climatological Dispersion Model, EPA
600/8-85-029.
17. U.S. EPA, 1986a: Procedures for Estimating Probability of Nonattainment
of a PMio NAAQS Using Total Suspended Particulate or PMio Data, EPA
450/4-86-017.
18. U.S. EPA, 1986b: Guideline on Air Quality Models (Revised), EPA
450/2-78-027R.
19. U.S. EPA 1986c: Supplement A to Compilation of Air Pollutant Emission
Factors, AP-42.
20. U.S. EPA, 1986d: Industrial Source Complex (ISC) Dispersion Model User's
Guide, Second Edition, Volume 1, EPA 450/4-86-005a.
21. U.S. EPA, 1987a: PMio SIP Development Guideline, EPA 450/2-87-001.
22. U.S. EPA, 1987b: Protocol for Reconciling Differences Among Receptor and
Dispersion Models, EPA 450/4-87-008.
23. U.S. EPA, 1987c: Receptor Model Technical Series, Volume III, (Revised)
User's Manual for Chemical Mass Balance Model, EPA 450/4-83-014R.
24. U.S. EPA, 1987d: Protocol for Applying and Validating the CMS Model, EPA
450/4-87-010.
25. U.S. EPA, 1987e: User's Guide for PMio Probability Guideline Software
Version 2.0.
26. U.S. EPA, 1987f: User Guide for RAMSecond Edition.
-122-
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-450/4-87-012
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Example Modeling to Illustrate SIP Development
for the PM1Q NAAQS
5. REPORT DATE
Mav 1Q87
May
6. PERI
FORMING ORGANIZATION CODE
7. AUTHOR(S)
Michael Anderson, Richard DeCesar,
Richard Londergan, and Edward Brookman
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
TRC Environmental Consultants, Inc.
800 Connecticut Boulevard
East Hartford, Connecticut 06108
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-3886
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
OAQPS, MDAD, SRAB (MD-14)
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This document provides an illustration of the use of modeling techniques for SIP
development for the PMjo NAAQS. Available dispersion and receptor modeling tech-
niques are applied and their results are compared. Analyses applicable to both
the annual and 24-hour PMio NAAQS are demonstrated. The application of models in
control strategy development is described. The examples provided include an urban
area problem and a problem caused by an industrial source with fugitive dust.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COS AT i Field/Group
PM10
PM1Q NAAQS
PM-JQ SIP Development
Dispersion Modeling
Receptor Modeling
Parti oil ate Matter Standard
18. DISTRIBUTION STATEMENT
19. SECURITY CLASS (This Report)
21. NO. OF PAGES
131
20. SECURITY CLASS (This page)
22. PRICE
EPA Form 2220-1 (R«v. 4-77) PREVIOUS EDITION is OBSOLETE
-------
-------