cxEPA
             United States
             Envirbnmental Protection
             Agency
            Office of Air Quality
            Planning and Standards
            Research Triangle Park NC 27711
EPA-450/4-87-008
March 1987
             Air
Protocol For
Reconciling
Differences Among
Receptor And
Dispersion Models

-------
                                          EPA-450/4-87-008
                                          March 1987
      PROTOCOL FOR RECONCILING
DIFFERENCES AMONG RECEPTOR AND
           DISPERSION MODELS
                         By

                  Air Management Technology Branch
                 Monitoring And Data Analysis Division
              U.S. ENVIRONMENTAL PROTECTION AGENCY
                   Office of Air and Radiation
               Office of Air Quality Planning and Standards
                  Research Triangle Park NC 27711

-------
This report has been reviewed by the Office of Air Quality Planning And Standards, U S Environmental
Protection Agency, and approved for publication. Any mention of trade names or commercial products is not
intended to constitute endorsement or recommendation for use.
                                         EPA-450/4-87-008

-------
                             TABLE OF CONTENTS

                                                                      Page
List of Figures	  1v
List of Tables	   v
Introduction	   1
CMB-DM Reconci 1 i ati on	   3
Protocol Step 1:  Compare CMB and DM Results	   3
     Selection of Sampling and Analysis Periods
     for Analysis and Comparison	   4
     Background and Secondary Particulate Matter	   5
     Grouping of Sources	   6
     Comparison of .Results	   8
Protocol Step 2:  Verify Input Data in Both Models and Rerun
     If Necessary	   8
Protocol Step 3:  Recompare Results	   8
Protocol Step 4:  Refine CMB Model Inputs	   8
Protocol Step 5:  Recompare Results	   9
Protocol Step 6:  Refine Dispersion Model Inputs and Rerun	   9
Protocol Step 7:  Recompare Results and Evaluate the Dispersion Model  11
Protocol Step 8:  Final Model Estimates as Basis for Control Strategy  12
Exampl e	  12
Acknowl edgments	  14
References	  15
Appendix A - Example Application of Protocol in an
                  Industrial Area	  A-l
                                    iii

-------
                              LIST OF FIGURES
Number                                                              Page

  1        Information Flow for Source Apportionment
           and .Control Strategy Development	   14
                                     iv

-------
                               LIST OF TABLES
Number                                                             Page

 A-l      Types of Emissions Assigned to Each Source
          Group in the Dispersion Modeling Analysis ...............  A-8

 A-2      Comparison of Initial  Results by ISCST and
          CMB for Subsets A and  B (ug/m3) .........................  A- 9

 A-3      Comparison of the Receptor and Disperson Model
          Results for the October 1, 1983 PMio Samples ............  A-12
 A-4      Comparison of the Receptor and Dispersion Model
          Results for the October 25, 1983 PMjQ Samples ...........   A-13

 A-5      Comparison of Final  Validation Results by ISCST
          and CMB for Subsets  A and B (ug/m3) .....................   A-19

 A-6      Source Contributions Estimated by ISCST Using
          5 Years of Meteorological Data (ug/m3) ..................   A-20

-------
INTRODUCTION
     The PMm SIP Development Guideline recommends that both Chemical Mass
Balance (CMB) and Dispersion Models (DM) be used to determine the relative
importance of various sources in contributing to ambient PM^g concentrations,
provided the models are applicable.!  Dispersion model  applicability is
discussed in Section 11.2.2 of the Guideline on Air Quality Models and in
the Interim Procedures for Evaluating Air Quality Models.2*3  The CMB
applicability is discussed in the Protocol  for Applying and Validating the
CMB Model .4
     Additionally, the PM^Q SIP Development Guideline provides for the use of
other receptor analyses to corroborate CMB and DM estimates.   These other
methods include optical or automated scanning electron microscopy (OM,
ASEM), microinventory, factor analysis (FA), multiple linear  regression
(MLR), and trajectory analysis (TA), as discussed an Appendix A to the
SIP Development Guideline.  Use of several methods is desirable, because it
yields estimates of source contributions from several perspectives, based
on different assumptions and data.  This approach greatly enhances the
likelihood that the nature of an ambient PM^g problem is correctly understood.
However, systems which rely on different measurements or use  of different
estimation parameters will likely yield different results. The differences
may be negligible, but occasionally the differences may be large enough to
require that they be reconciled.
     The protocol summarized in this report consists of an eight-step procedure
for examining and reconciling differences in model estimates  in a systematic,
justifiable way.  This procedure is illustrated in Appendix A.  The emphasis
is on CMB and DM reconciliation, primarily because the SIP Development

-------
Guideline identifies  CMB  as  the principal receptor modeling approach  for
use in preparing SIP's  for PM^o*  Nevertheless, this protocol  is  intended  to
be general  enough so  that the  results of other methods  (e.g.,  "OM,"  "ASEM,"
or "MLR") could be reconciled  in the same manner.  (In  fact, OM is used  in
the example in Appendix A.)
     There are at least three  generic cases which require model differences
to be resolved (reconciled).  The first case is when the results  strongly
suggest that at least one of the models has a significant error.  The second
case is when the model  estimates are not disparate, but the overall  uncertainty
is judged to be large.  Resources devoted to further refinement of the models
and data bases might  significantly  reduce the uncertainties.   The third  case,
a very common one, includes  situations where one model  is better  for  some
source types, and another is better for other source types.  Refinement  of
the final source contribution  estimates should be cognizant of the suitability
of each model to the  suspect sources.  Trijonis reports that the  literature
assumes that CMB models are  more reliable than dispersion models  for  quanti-
fying the impact of source categories (i.e., steel mills, power plants,
woodstoves, road dust).5   According to Trijonis, this may be because  the
CMB is tied directly  to the  ambient data that are being apportioned.   The
CMB is 1imi ted to apportionment of  source categories unless there is  only
one source of that category  upwind  on the sampling day.  In contrast,  the
DM can deal explicitly  with  emissions from single, identifiable sources
within the same source  category.  However, in order to  do so,  the DM  must
build on several potentially uncertain inputs (emissions data, meteorological
data, and the transport-diffusion-transformation-deposition mechanisms).
Thus, the CMB and DM  are  complementary in their approach to source apportion-
ment and using both can reduce the  limitations of each  alone.

-------
 CMB-DM RECONCILIATION
     The  CMB-DM  reconciliation process is based on the premise that the
 receptor  and  dispersion models provide independent estimates of source
 contributions,  rely  on different data bases, have different underlying
 assumptions,  and have different strengths and weaknesses.  Generally, other
                                                                       %
 receptor  methods are used  first, then the CMB is performed and, finally, the
 DM  is run.  Figure  1 depicts the flow of information if this order is maintained.
 At  each step, the results  of the other analyses are used to corroborate the
.ongoing analysis.  The use of other receptor methods to corroborate the CMB
 is  shown  on Figure 1 for completeness but is discussed elsewhere.^
     "The  reconciliation of CMB-DM results should follow an eight-step protocol
 involving comparison, reverification of input data, refining the inputs to
 both models,  and rerunning the model(s) if necessary.  The reconciliation
 protocol  described in the  remainder of this document assumes that both the
 dispersion model  and CMB have been determined to be applicable and that
 preliminary results  have been obtained for both models.  The reconciliation
 process ends  with control  strategy development using the dispersion model
 unless the dispersion model is found to be inconsistent with the majority
 of  the physical  data.  In  such case, the CMB along with other receptor models-
 would be  used with a proportional model to develop a control strategy.1

 Protocol  Step 1:  Compare  CMB and DM Results
     There are  several issues that must be resolved in order to compare CMB
 and DM results:   (1) consideration of suitable periods for sampling and
 analysis, with  particular  emphasis on source emission variability and

-------
             Source
             Information
                  I
Recommended
Approach
Alternate   |
Approach    j
    Ambient
    Sample
Other
Data
                         t

Other
Receptor
Methods
h
1
1
1
1
1
1
r
i
T
D
1
r
f
CMB
Model
i
ispi
Mo
~ 1
l_ 	
1
1
srsion '
del 1
                                   I
                                   I
                              Apportionment
                                    by
                             Dispersion  Model
                      Apportionment
                            by
                      Other Methods
                                  i
Control  Strategy
       by
Dispersion Model
Control Strategy
       by
Proportional Model
          FIGURE 1.    Information  Flow For Source Apportionment
                       And Control  Strategy Development

-------
meteorological conditions; (2) the issues of secondary participate matter
and background; and (3) grouping of the sources.  These issues are discussed
in the following paragraphs.
Selection of Sampling Periods for Analysis and Comparison.    Attainment of
the NAAQS is generally calculated based on ambient data.  Ideally, for
comparison purposes, the DM would be run for the actual  monitoring period
and CMB analyses would be available for each day during that same period.
It is preferable that the CMB be performed on days selected randomly during
the period.  However, this ideal is virtually never achieved.for numerous
reasons of cost and practicality.  For example, in assessing potential
causes for not meeting the annual NAAQS, the DM may have to be run using
meteorological data representing different years than those on which non-
attainment is based.  In that case, it would be necessary to perform CMB on
a subset of days which are representative of the measured concentrations
during the time period covered by the DM but also representative of the
overall meteorology during that time period.  In addition,  the number of
samples to be compared must reflect the seasonal differences in measured
concentrations.  In CMB applications relating to the annual  NAAQS, a minimum
of five samples in each quarter are considered necessary to obtain representative
results.
     For violations of the 24-hour NAAQS, it *is preferable  to apply the CMB
model on the days on which "exceedances" of the NAAQS were  observed.  In this
case, one would analyze and compare all observations greater than the level
of the NAAQS.  In order to obtain representative results, a minimum of five
samples (e.g., the five highest values) should be compared.   If there are
fewer than five observed "exceedances" of the NAAQS, the five highest

-------
values overall  should be used  so that  the  analysis  is  based  upon  a repre-
sentative number of days.   If  a receptor model  study were  undertaken for a
short time period,  it must  be  shown  that the period covered  was  generally
representative  of the types of source  activity  and meteorology which are
associated with exceedances measured outside of the receptor model  study.
The source operating levels must be  similar and the dispersion conditions
and wind deviation  during the  shorter  period must be such  that the same
sources would be expected to contribute to measured concentrations.   The
receptor data are particularly useful  if exceedances are measured during
the receptor study.
     Unfortunately, a 24-hour  emissions inventory appropriate for specific
days may not be available or obtainable.   Inventories  usually contain
annual average  emissions.   Thus, associating "dates" with  the dispersion
model results,  which are based on annual emissions, could  be risky.   In
this case, an "aggregate" comparison of the analysis results is  recommended.
To make an "aggregate" comparison, select  a subset of  dispersion  model
24-hour estimates that has  dispersion  meteorology similar  to that which
occurred on the CMB analysis days (e.g., days with similar wind  velocities
and precipitation/cloud cover).  Then, compare  the source  contributions for
each CMB analysis wi.th the  average-source  contributions for  the  corresponding
subset of dispersion modeled days.
Background and  Secondary Particulate Matter.    In many cases, the receptor
model is used for source apportionment at  urban or source  oriented sampling
sites, and the  apportionment will include  not only the "urban" contribution
but also the background component which has been transported to the  urban

-------
area from elsewhere.  In such cases, it is usually necessary to perform an
analysis on a nearby background receptor so that the background or regional
contributions to ambient PM observed at urban or source-oriented sites can
be distinguished.  This receptor analysis at a background site is
necessary, for example, to distinguish between locally generated soil  dust
and that transported into the area.  Receptor analyses from the background
and other ambient monitors are then compared.  Source categories that  were
identified on both urban and background monitors should be noted and estimated
contributions of these sources at the study (urban)  monitor should be
reduced by the amount of their impact at the background site.   In this
process, one must carefully select the background days for receptor analysis
to make sure they are "upwind" and are not impacted  by local  sources,  and
adjustments must be made, to the CMB results prior to CMB-DM comparisons.
Also, this reduction in urban contributions must be  made carefully when
background sources are chemically similar to sources in the urban area
(e.g., soil and resuspended street dust), especially if the background site
has considerable localized sources.
Grouping of Sources.   CMB is usually only useful for identifying source
categories contributing to ambient PM^Q.  In contrast, DM can  identify
contributions from individual sources.  Thus, source contributions from the
CMB and the DM must be grouped in such a way that contributions from similar
sources can be assessed.  The DM results are usually regrouped because the
source chemical characteristics used in the CMB cannot usually be so done.
Therefore, for the purpose of comparing DM results with CMB results, sources
considered by the DM must be grouped into larger "categories"  or source
groups.

-------
     For example, perhaps  a monitored  sample was  obtained  from a  coking
operation.  From a literature review or from experience, it  is determined
that the following operations at  a  steel making facility have  emissions
chemically similar to the  coke sample:
     Quenching
     Pushing
     Charging
     Finished Coke Storage Pile
     The "coke" source contribution calculated by the CMB  would conceivably
represent all of these specific sources.  The DM  output should group  all  of
these sources together into a "coking" group; alternatively, the  DM estimates
for these sources could be manually grouped in a  separate  tabulation  after
the model is run.
     Although groups are often arbitrarily selected, for purposes of  CMB-DM
model reconciliation, a little judicious planning can make the comparison
of CMB with DM results fairly straightforward.  This can enhance  confidence
in the models themselves if good  agreement is found.
     Grouping of DM sources is performed as follows:
     (a)  Review the detectable CMB source contribution categories to determine
          what they really include  or  represent.  This should  be  documented
          with the best available references.  For example, a  "sea salt"  source
          might include salt from ocean spray and also salt which was spread
          for snow control.6
     (b)  Review the Emission Inventory (El) and  decide which  DM  sources
          collectively comprise each CMB source category.  Also,  identify
          any unmodeled sources (e.g., continental dust, sea salt, etc.)
          that might be represented by the CMB source category.
                                    8

-------
     (c)  Combine the DM sources, into groups that are consistent with CMB
          source categories.
     (d)  Identify any sources which were significant contributors in the
          CMB analysis but which were not in the DM emission-inventory.
     (e)  Add these sources to the El as appropriate.
   * (f)  Rerun the DM.
     (g)  Make sure that the CMB has considered all nontrivial  sources which
          were identified in the DM inventory.   Rerun the CMB  if necessary.
Comparison of Results.   The grouping of sources described above allows  a
comparison of the DM and CMB estimates of the percentage contributions to
the total calculated PM^g mass.  First, establish an interval  of + or -
30 percent around the DM results and + or - one standard error around the CMB
results.  If these intervals overlap, no further refinement is generally
warranted.  However, if the DM results imply a  source group is a major contributor
and the CMB results imply that it is not (or vice versa), the  results must be
reconciled, even if the comparison criterion (above) is met.  If further
refinement of the models or inputs are needed to meet the criterion,  the
procedure for doing so is described in the following section.   Note that the
+ or - 30.percent interval about the DM is only for use in this intercomparison;
it is not intended as a confidence interval for the DM.

Protocol  Step 2:  Verify Input Data in Both Models and Rerun if Necessary
     This step ensures that the difference in CMB and DM results are not due
to trivial or inadvertent problems with data entry.  It should focus  on  the
data entry process to verify that the models were run using the intended

-------
data.  The data review should  include the ambient and  source  profile data,
emissions data and meteorology.   If  all data  are correct,  skip  to Step 4.

Protocol  Step 3:   Recompare Results
     Any  differences found  in  model  estimates  due to data  entry should' now be
eliminated.  Revised model  results can now be  compared,  as  discussed in
Protocol  Step 1.   If the  comparison  criteria  are not met,  the next  step is to
reexamine model inputs, initially focusing on  the CMB  model.

Protocol  Step 4:   Refine  CMB Model Inputs
     This is an indepth review of the CMB model inputs,  focused on  those
source groups where large differences between  CMB and  DM source category
contribution estimates occurred.  Because of  the validation that the CMB
application should have received, it is unlikely that  there are any substantive
CMB model input problems  which were  not apparent from  the  preliminary analyses.
     However, insights available through the  Dispersion  Model analysis may
highlight new problems or confirm suspicions  that the  modeler may have already
had about weaknesses in the CMB that could not be previously  substantiated.
Also, it  is appropriate to  review the results  of other receptor analyses
at this point.  This review may require modifying the  model inputs  or addressing
col linearity.4
     All  measurements have  some imprecision or uncertainty  about them, and
there will be some inherent underlying "noise" level in  the data set which
cannot be reduced by a reasonable allocation  of time and resources.  The
user should review the CMB  validation in light of the  DM results.4  If any
weaknesses become apparent, address  them to the extent time and resources
permit, focusing  on the source groups with disparate estimates. -Then rerun

                                     10

-------
the CMB model, if any refinements were made.   It is not sufficient to know

just that a revision to a parameter will  "fix" a discrepancy.   Technical

justification is required to document all  changes that are made.


Protocol Step 5:  Recompare Results

     Any remaining differences found between  CMB and DM model  estimates  are

best addressed by closer examination of the dispersion model.   If the comparison

criteria in Step 1 are met, skip to Step 7.


Protocol Step 6:  Refine Dispersion Model  Inputs and Rerun

     This is an indepth review of the dispersion model, focusing  on those

source groups (e.g., fugitive dust sources) for which large differences  occur

between DM and CMB.  This review should include (1) the appropriateness  of

emissions data, which are based on the identification of sources  and their

locations, activity rates, emission factors and release heights;  (2) meteor-

ological data; "and (3) underlying assumptions in the model itself and its

dispersion, transformation, and removal algorithms.  Reference 7  gives examples

of factors that primarily affect individual source-to-monitor  impacts:

     o  Erroneous emission data caused by  such items as the omission
        of unknown sources from the modeling  or the use of inaccurate
        throughput information.

     o  The use of inappropriate emission  rates, such as the use  of
        total particulate matter emission  rates which include  particle
        sizes larger than PMjg that settle out prior to impacting the
        monitoring site and are not collected by a PMig sample if they
        do reach the receptor.

     o  Incorrect information concerning daily source operating
        parameters.  For example, while a  source may operate at
        45 percent capacity on an annual  basis, its actual mode
        of operation may be at 90 percent  capacity for 50 percent
        of the days in a year.
                                     11

-------
     o  Neglect or incorrect  consideration  of  downwash  from tall
        stack sources.
     o  Neglect or incorrect  consideration  of  particles resuspended
        by the wind or  mechanical  action.
     o  Building interference causing  source-to-receptor (i.e.,
        source-to-monitor)  geometry to be incompatible  with
        Gaussian dispersion assumptions.
     o  Local meteorology  differing from that  modeled.   A typical
        problem is wind direction  shift or  channeling caused by
        buildings or topographic features.
     In reference 7, Anderson suggests that these  emissions and  meteorological
data problems can often be satisfactorily resolved by:
     (a)  identifying significant  sources or source  groups from  the CMB
results;
     (b)  assigning a "level  of confidence" to emission and meteorological
parameter estimates for each  significant source or source group,  based on
engineering estimates.   The confidence level information needed  for this
analysis may be found in Volume V  of the Receptor Model  Technical  Series.7
Ideally this analysis should  be completed before the CMB and dispersion -
models are run.
     (c)  revising the  emission and meteorology parameter estimates, as
appropriate.  It is not sufficient to  know  just that a  revision to a parameter
will "fix" a discrepancy.   Technical justification is required to  document
all changes that are made.
     The pattern of discrepancies  at different receptors may also  aid in
identifying potential inventory errors.  For example, large under- or over-
prediction by the dispersion  model at  a receptor very close to a source may
suggest possible errors in  source  receptor  geometry.  Large overprediction
at receptors far from a source would suggest an erroneous particle size
                                      12

-------
distribution in emission factors.  Experience with dispersion models and
emission inventories will  play an important role in diagnosing potential
inventory problems.
     The dispersion model  is then rerun, using the revised input parameters.

Protocol Step 7:  Recompare Results and Evaluate the Dispersion Model
     The revised dispersion model results are again compared with the CMB
results to determine whether further modifications to the dispersion model
or meteorological inputs are warranted.
     Anderson provides examples of factors that are not typically addressed
by the dispersion model  that may cause systematic biases in model results.
These factors are addressed in various nonguideline models that are currently
available.
     o  Heat island effects which cause the actual near-source
        dispersion of elevated emissions to be greater than that
        modeled.
     o  Sea/land breeze and convergence zone effects.
     o  Effects caused by the fumigation of tall stack emissions.
     o  Effects caused by the development of thermal  internal
        boundary layers (TIBL) over areas with varying surface
        heating characteristics.
     Remedies to these problems might include acquisition of additional
meteorological data, choice of different model options or modification of
dispersion algorithms to meet local disperson conditions and source char-
acteristics.  Such problem specific model modifications may prove a time and
cost effective alternative to additional sampling and analysis.  These remedies
should be applied to the extent permitted by time and resource constraints.
                                     13

-------
Protocol  Step 8:   Final  Model  Estimates as Basis for Control  Strategy
     The  reconciliation  process should help improve both the  initial DM and
CMB. results so that  the  differences between models are  resolved.   If this
is the case, it is recommended that the DM model be used for  control strategy
development.  If,  however,  it  is clearly evident that the  dispersion model
results are inconsistent with  the majority of the physical  data  and cannot
be made consistent through  justifiable modifications to the input  data, the
CMB estimates should be  used as the basis for control strategy development.
It is better to use  the  results of the method believed to  provide  the  best
results,  rather than average the results of two or more methods.

EXAMPLE
     Appendix A describes the  application of this protocol  to an industrial
area.  The example includes the use of preliminary analyses including  pol-
lution rose and microinventory, dispersion modeling using  the Industrial
Source Complex-Short Term Model (ISCST), and receptor modeling using the
Chemical  Mass Balance (CMB) and optical microscopy (OM).
                                    14

-------
ACKNOWLEDGMENTS

     The primary authors of this document are Thompson G. Pace, P.E., of
the U.S. EPA, and Dr. John G. Watson of the Desert Research Institute.
Messrs. Richard DeCesar and Michael  Anderson of TRC Environmental  were
largely responsible for the Example  in Appendix A.  Dr. John Trijonis
provided valuable conceptual framework at the onset of this protocol
development.  The following co-authors participated in a workshop  to
review and revise the document in May 1986 in San Francisco:  Dr.  Kit
Wagner, Mr. Bart Croes, Mr. Duane Ono, Dr. Hal  Javitz, Mr.  Mike Naylor,
Dr. Judith C. Chow, Dr. David Maughan, Dr. Edwin Meyer, Mr. Luke Wijnberg,
Mr. Ken Axetell, Mr. John Core, Mr.  Pat Hanrahan, Dr.  Ron Henry, Mr.  Bong
Mann Kim, Mr. Chung Liu, and Dr. Andy Gray.

     Several others provided review  and comment outside of  the workshop.
The reviews by Dr. Glen Gordon, Mr.  William Cox, Mr.  Chuck  Lewis,
Dr. Thomas D. Dzubay, Mr. Robert K.  Stevens, Mr. Johnnie Pearson,
Dr. Richard DeCesar, Dr. Sylvia Edgerton, and Mr. Edward Lillis are
greatly appreciated.  The typing and revisions  by Ms.  Cathy Coats  and
Ms. Jo Harris are much appreciated.
                                     15

-------
 REFERENCES

 1.  U.S. EPA.  PMTn SIP Development Guideline,  EPA 450/2-86-001,
     Research Triangle Park,  NC27711.February, 1987.

 2.  U.S. EPA,  Guideline on Air Quality Models  (Revised).   EPA 450/2-
     78-027R.  U.S. EPA, Research  Triangle Park,  N C   2"7711.

 3.  U.S. EPA,  Interim Procedures  for Evaluating  Air Quality  Models,
     EPA-450/4-85-023, U.S. EPA,  Research  Triangle Park,  NC27711,
     September  1984.

 4.  U.S. EPA,  Protocol  for Applying and  Validating the CMB Model,
     EPA 450/4-87-010, U.S. EPA,  Research  Triangle Park,  N.C.  27711,  May, 1987.

 5.  Trijonis,  J.  "Model Reconciliation,"  Special Report  prepared  for U.S.
     EPA and TRC Environmental Consultants under  Contract  68-02-3886, Work
     Assignment No. 13,  Research  Triangle  Park, NC  27711,  September  1985.

 6.  U.S. EPA,  Receptor  Model  Source Composition  Library,  EPA 450/4-85-002,
     U.S. EPA,  Research  Triangle  Park, NC  27711, November 1984.

 7.  U.S. EPA,  Receptor  Model  Technical Series, Volume V:   Source  Apportionment
     Techniques and Considerations in Combining Their  Use,  EPA-450/4-84-020.
     U.S. EPA,  Research  Triangle  Park, NC  27711, July 1984.

 8.  Anderson,  M.  K.,  Richard Decesar, R.  Longergan and E.  Brookam, Example
     Modeling To Illustrate SIP Development For The New Particulate Matter
     TiflAQS, Prepared by  TRC Environmental  for the U.S. EPA under contract
     68-02-3886, work  assignments  37, 48 and  52,  Draft, in  Preparation.

 9.  OeCesar, R. T. personal  communication to T.  Pace, February 1987.

10.  U.S. EPA,  Receptor  Model  Technical Series, Volume III  (Revised):
     Chemical Mass  Balance User's  Manual,  U.S.  EPA, Research  Triangle
     Park, NC 27711.In preparation.
                                     16

-------
      APPENDIX A - EXAMPLE APPLICATION OF PROTOCOL IN AN INDUSTRIAL AREA

INTRODUCTION
     This section describes an example application of the protocol  for a
nonattainment site located within an industrial  area.  The example  was
adapted from an actual environmental case study  and modified where  necessary
to enable its use with this protocol.^  All  modifications were cognizant of
what could and would have actually been done in  the reconciliation  had time
and resources permitted.9  in this example,  the  causes of the elevated concen-
trations of PMio are not apparent due to the large number of point  and area
sources located relatively close to the monitoring site.  Based strictly on
the emission inventory, stack and fugitive dust  emissions from a large steel
mill as well as traffic generated resuspended road dust appear to be potentially
large contributors to the particulate matter loadings observed at the monitoring
site.  An investigation, which includes ambient  and source data collection,
optical microscopy (OM), pollution rose (a form  of trajectory analysis),
dispersion modeling (DM) and chemical mass balance (CMB) modeling efforts,
was conducted in order to apportion the source impacts with the level  of
confidence necessary for making control strategy decisions.  The results are
reconciled using the eight-step process previously described.
     The industrial site PM^g source apportionment example presented in this
section is compiled primarily from experience gained from assessing source
impact at a single monitoring site.  The purpose of this example is to Illus-
trate use of the eight-step reconciliation protocol described in this document.
Therefore, the identity of the site and details  of the OM, CMB, and DM analyses
are unimportant to the objective of the example  and are not presented in
detail.
                                     A-l

-------
OVERVIEW OF THE SOURCE APPORTIONMENT STUDY
     A combined receptor and  dispersion modeling study was implemented  with  the
objective of identifying and  quantifying impacts of the sources  contributing
to violations of the PM^g NAAQS  recorded at a monitoring site.   To  supplement
the emission inventory, pollution concentration roses were prepared using  data
for eight TSP monitoring sites located near the PM^Q monitoring  site.   The
pollution roses clearly indicate that the major sources of TSP are  located in
the quadrant south of the PM^Q monitor.  That is, a preponderance of the high
24-hour average TSP measurements correspond with days having  resultant  winds
from the south.  This preliminary receptor analysis suggests  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.

Receptor Modeling Data Collection Tasks
     The ambient data base consisted primarily of:  (1) ambient  PM^g samples
collected for 1 year on an every-6th-day schedule; and (2) hourly measurements
of wind direction, wind speed, and temperature at the PM^g monitoring station.
The PM^Q samples were collected  on quartz fiber filters using a  size-selective
high volume (SSHV) sampler.   In  addition, for 1 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 PMig monitoring
program provided an annual  average PM^g concentration of 65 ug/m^ as well
as four exceedances of the 24-hour average PM^g standard of 150  ug/m^.
                                    A-2

-------
     A site visit was made to compile a microinventory of emission sources.
The microinventory procedure consisted of determining the nature, location
and spatial extent of all fugitive dust sources within one-quarter mile of
the PM^g monitoring station.  The information was then used to calculate the
emission rates for input to a dispersion model.  Source types included unpaved
roads, paved streets, railroads, and two coal storage piles.  Emission rates
were determined using published emission factors together with the source
area and traffic volumes.
     As part of the site visit, samples of emitted materials were collected
from six sources in the general vicinity of the site which had been identified
as potentially important contributors to PM^g.  The samples were subsequently
analyzed to provide source profiles for use in the CMB model.  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).
     Two subsets of the ambient PM^o filters  were selected for analysis as
discussed above.  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 PM^o concentration calculated from the filters
in subset A was 67 ug/m^, which compares well with the annual average
calculated from all the filters (65 yg/m3).  Subset B contained the
samples from the 4 days on which exceedances  of the 24-hour average NAAQS
were recorded plus the highest nonexceeding day (a total of 5 days).   This
subset was chosen to guide control strategy development related to the 24-hour
standard.
                                     A-3

-------
     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 following species were analyzed:  Na, Mg,  Al, Si,  K,"
Ca, Ti, V, Mn,  Fe, Zn, Br, Pb, EC and $04.  The elemental  carbon  (EC) content
of the samples  was determined by the optical  attenuation analysis  method.  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 receptor modeling source  identifications
and to aid in the receptor/dispersion model reconciliation process.
     Receptor modeling was performed for this study using  the Chemical  Mass
Balance (CMB) model.4'10   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 analysis into the format  specified by the CMB model.   In
addition, the CMB model requires a file  containing the source 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.6 The emission  inventory
and the microinventory were used to  select sources for consideration in the CMB.

Dispersion Modeling Data  Preparation
     The emission inventory of all the point  sources in the two counties
surrounding the PMjo monitoring site was obtained.  A total of  140 point sources
                                    A-4

-------
were included in the inventory.  The point source inventory was modified for
use in modeling PMjg impacts by applying source specific PMjg/TSP emission
ratios to the TSP emission factors which were originally contained in  the
inventory.
     The Industrial Source Complex Short-Term (ISCST)  model was used  for this
                                               %
investigation.  ISCST was used because it is applicable to the dispersion con-
ditions found in the study area, and it contains several features that improve
the source apportionment process.  Two potentially advantageous features of
ISCST are the ability to: (1) model microinventoried fugitive dust sources as
volume sources; and (2) calculate the combined impact  for selected groups of
sources.  This latter feature greatly decreases the manipulations necessary
to transform dispersion and receptor modeling results  into comparable  formats
providing basis for comparisons.
     Two receptor oriented approaches are utilized in  this example:   the CMB
model and the optical microscopy technique.  CMB modeling analyses were per-
formed on the data obtained from filters contained in  subsets A and B  described
in Section 2.1.  The modeling procedure consists of obtaining a solution
using the CMB procedures found in reference 4.
     Two SSHV samples, two coarse fraction dichotomous samples, and one fine
fraction dichotomous sample (ambient samples) 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 are
used to calculate semi-quantitative source contribution estimates.
                                     A-5

-------
     As part of the CMB modeling procedure, efforts are made to  develop
information which could be  used to confirm or refute the validity  of  the CMB
results.  This information  can be classified into four categories:
(1) applicability of the model to the situation; (2) evidence  of adherence  or
deviation from model assumptions by examination of the summary statistics
                                    %
resulting from the use  of model assumptions; (3) stability  of  source  impact
estimates with respect  to minor changes in the CMB model fitting parameters;
and (4) comparisons with preliminary analyses (e.g., pollution rose,  OM  and
microinventory analyses).  By evaluating the CMB results on the  basis of the
above four categories,  the  CMB analyses are determined to be valid.
     Background concentrations are 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 for the  dispersion
model, an eclectic 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 average TSP concentration of 26 yg/m^ was
determined to be attributed to sources outside the study area.   Application
of the site-specific PMjo/TSP ratio (0.57) yields an average background
concentration of 15 ugm3.  Background concentrations were also estimated
for each wind sector using  this method.
*The term "eclectic pollution  rose" is used to describe a rose developed  by
 compositing the data  from several monitors surrounding the urban  area  such
 that the eclectic rose reflects concentrations only when the wind is blowing
 into the urban area.
                                    A-6

-------
THE EIGHT-STEP RECONCILIATION PROCESS
Step 1 - Compare DM and CMB Results
     The dispersion model simulation results for time periods corresponding
to the PM^o sampling schedule at the monitoring station were compared with
measured PM^g air quality data.   This comparison suggests that the ISCST is
underpredicting PM^g concentrations at the monitoring site.  Comparison of
the predicted PMig w^tn tne measured PM^g minus estimated background PM^g
concentrations provided an indication that the model  was underpredicting.
     The source contributions from the DM and CMB were grouped in such a
way that they could be compared.  This was actually done to some extent before
the ISCST was run by examining the emission inventory.  The emission inventory
contained the name of each plant and a description of each emission source.
This information was used to develop .a preliminary tabulation of the types of
materials (e.g., specific source chemistry) emitted by each source.  The
many types of emitted materials  were then associated  with 1 of 17 general
categories which were determined to be consistent with the source profiles
used in the CMB analysis.  Each  of the 140 point sources and 25 volume
sources was assigned a code corresponding to 1 of the 17 categories.  Some
overlap and "juggling" of categories was inevitable but a reasonably good
preliminary grouping was made.  The results are presented in Table A.I.
Adjustment to this preliminary grouping may be necessary in some cases.
     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 A-2.  The basic oxygen furnace and steel  blast
furnace groups were combined for the same reason.  The results of the two
                                    A-7

-------
                                  TABLE A-l
                  TYPES  OF  EMISSIONS ASSIGNED TO EACH SOURCE
                  GROUP  IN  THE DISPERSION MODELING ANALYSES
Group Number                    Type of Emission
      1                     *   Oi 1 and Gas Combustion
      2                         Coal Combination
      3                         Wood-fired Boilers
      4                         Agricultural Products Handling
      5                         Coking Operations
      6                         Coal Handling
      7                         Steel Blast Furnace and BOF
      8                         Paint Production
      9                         Cement and Limestone
     10                         Motor Vehicle Exhaust and Lead  Processes
     11                         Sand and Bentonite
     12                         Aluminum Production
     13                         Zinc Processing
     14                         Tire Production
     15                         Oil. Refining
     16                         Fertilizer
     17                   •      Road Dust, Soil, and Motor Vehicle Exhaust
                                       A-8

-------
                             TABLE A-2
         COMPARISON OF INITIAL RESULTS BY ISCST AND CMB FOR
                      SUBSETS A AND B (ug/m3)
                                   (Representing
                                   Annual  Average)
                                       Subset  A
(Average of 5 Highest
24-hr Exceedances)
    Subset B
Source Categories
Oil & Gas Combustion
Wood-fired Boi lers
Coal Combustion
Coking Operations
Blast Furnace3
Basic Oxygen Furnace3
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)2$04
Background
Total
Measured
ISCSI
1
0
0
1
- 0

0
1

1
. 1

0
0


15


15
45
67
.6
.2
.6
.4
.6

.3
.1
—
.9
.2
--
.8
.6
— .
--
.1

—
.0
.4
.0
CMB

3
2
8
12
3










2
3U
15

77
67
..
.2
.8
.6
.0
.8
—
—
--
--
--
--
—
—
--
--
.4
.0
.0
--
.8
.0

(1
(1
(4
(3
(1










(0
(7
(2



ISCS

.4)
.1)
.2)
.5)
.5)










.8)
.2)
.5)



2.
0.
1.
4.
2.

0.
1.
0.
1.
1.
0.
0.
1.
0.
2.
2.
31.
-
32.
89.
162
T
0
8
2
0
o

6
6
8
9
7
4
6
5
8
2
6
2
-
0
0
.0
CMB

5
8
23
33
10










2
79
27

189
162
..
.6
.0
.2
.6
.4
—
--
--
__
__
--
—
—
--
_-
.4
.2
.4
--
.8
.0

(2.1)
(3.2)
(8.7)
(8.9)
(2.7)










(0.9)
(20.3)
(5.0)



a Estimated as a single source by the ISCST because  of  lack  of
  detail  in the emission inventory.

NOTE:  Numbers in parentheses  indicate one standard  error  in the
       CMB analysis.
                                   A-9

-------
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 PMjo by both methods.   In  addition,  the similarities and differences
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.   For example,  ISCST underpredicts
the measured mass and CMB overpredicts  the  measured mass.   Furthermore, ISCST
assigns contributions to several  source categories which were  not identified
by the CMB method.  Finally, CMB  and ISCST  estimated  substantially different
impacts for wood-fired boilers, coal  combustion, the  road  dust and motor
vehicle exhaust combination and the  steel mill  related  emissions from coking
operations, blast furnace, and basic oxygen furnace.  Unfortunately, Table
A-2 does not reflect conclusions  drawn  from any  OM analysis,  because the OM
analysis is available for only 2  days.
     As a result of this comparison, four cases  were  identified (based on
comparison model results as discussed in Step  1  of the  Protocol) where additional
model refinements may be necessary:
Case 1.   The combined contributions of motor  vehicle exhaust  and resuspended
road dust, as estimated by CMB and ISCST, were  in major  disagreement.
Case 2.   A significant disagreement exists between the  CMB and ISCST estimates
of the coal combustion impacts.
Case 3.   CMB estimates wood-fired boilers  to  be a much  larger contributor to
PM10 than does ISCST.
                                     A-10

-------
Case 4.   There is a major disagreement between CMB and ISCST for steel  mill
related impacts (i.e., coking operations, blast furnace and basic oxygen
furnace source categories).
     Each case is discussed below under the appropriate step in the reconciliation
process.
Steps 2 and 3 - Verify Input Data, Rerun Models and Recompare Results
     A review of the input data for both models indicated that they were
indeed using the data that were intended.  None of the four cases were  related
to inadvertent errors in data input.

Step 4 - Refine CMB Model  Specification
     Operationally, all four cases identified in Step 1 are reviewed at  this
point.  Only case 1 was found to be related to the CMB inputs.  Thus, discus-
sion of the other three cases is included under step 6.  Tables A-3 and  A-4
compare the source impacts estimated by ISCST, CMB and optical microscopy  for
the 2 days, October 1, 1983 and October 25, 1983, when OM analysis was  performed.
Case 1.   The combined contributions of motor vehicle exhaust and resuspended
road dust, as estimated by CMB and ISCST, were in major disagreement.
     The first step taken  to reconcile the disagreement between CMB and  ISCST
with respect to the combined impact of vehicle exhaust and read dust was to
                                                              •
re-examine the comparisons between OM, CMB, and ISCST.  This is shown in Tables
A-3 and A-4 for the 2 days for which OM data are available.  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
                                     A-ll

-------
                                                                   TABLE  A-3

                         COMPARISON OF THE RECEPTOR AND DISPERSION MODEL RESULTS FOR THE OCTOBER 1, 1983 PM,0 SAMPLES
       Dispersion Modeling
         Source Groups
                          Optical Microscopy
                          Source Categories
                          CMB Modeling
                          Source Types
                                                                                             Modeled Concentrations
                  Coarse  Fine
Dispersion  SSHV  Dichot Dichot
 Modeled  Optical Optical  by           Fine  Coarse
 Impacts  Micros. Micros. Diff.  SSHV  Dichot Dichot
            (A)    (B)   (A-B)   CMB    CMB    CMB
r\>
Oil & Gas Combustion
Hood Combustion
Coal Combustion

Coke
Steel

Coal
Agricultural Products
Paint Products
Cement, Limestone
Lead
Sand, Bentonite
Aluminum
Zinc
Misc. Chemicals
Refinery Catalysts
Misc. Refinery
Fertilizer
Tire Products
       Road Dust
       Background

       Total
       Measured PM».$
       Measured PM2.$ -1o
       Measured Dichot. PM|0
       Measured SSHV PM,0
Fuel Comb., Carbon

Fuel Comb..Part.Pyr.Coal
St.Mill, Part. Pyr. Coal
Steel Mill, Coke
Steel Mill

Coal Handling
Agr. Prod. Handling
Spray Painting
Cement Handling

Clay Handling
Alumina Handling
                                 Oil Refining
                         Traffic,  Tire Fragments
                         Traffic,  
-------
I
i—*
CO
                                                                   TABLE A-4


                        COMPARISON OF THE RECEPTOR AND DISPERSION MODEL RESULTS FOR THE OCTOBER 25, 1983 PM,0 SAMPLES


                                                                                              Modeled Concentrations (yiq/m1)
Dispersion Optical
Modeled Microscopy
Dispersion Modeling
Source Groups
Oil & Gas Combustion
Wood Combustion
Coal Combustion
Coke
Steel
Coal
Agricultural Product
Paint Products
Cement, Limestone
Lead
Sand. Bentonite
Aluminum
Zinc
Misc. Chemicals
Refinery Catalysts
Misc. Refinery
Fertilizer
Tire Products
Road Dust
Background
Total
Measured PM* . $
Measured PMj . s-io
Measured Dichot. PMio
Measured SSHV PM,0
Optical Microscopy
Source Categories "•
Fuel Comb. , Carbon
Fuel Comb. .Part. Pyr. Coal
St. Mill, Part. Pyr. Coal
Steel Mill, Coke
Steel Mill
Coal Handling
Agr. Prod. Handling
Spray Painting
Cement Handling
Clay Handling
Alumina Processing
Oil Refining
Traffic, Tire Fragments
Traffic, 
-------
between the time it was weighed and  the time  it was  analyzed.   Further
evaluation of the sample indicated that this  was  indeed  true.   Thus,
the correct mass was used to recompute the  road dust profile.
     The CMB model  was rerun 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
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.

Step 5 - Recompare  Results After Rerunning  CMB
     The CMB model  was rerun, the results were recompared and  no significant
difference in model estimates remained.

Step 6 - Refine Dispersion Model Inputs and Rerun
Case 2.   A significant disagreement exists between  the  CMB and ISCST estimates
of the coal combustion impacts.
     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 re-examine the comparisons between OM, CMB, and ISCST which are shown in
Tables A-3 and A-4.  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
                                    A-14

-------
these sources to the receptor,  the emission inventory contained a very high
value for the efficiency of the emission controls.   Discussions with  the
plant personnel revealed that the inventory contained an  erroneous value and
a correct value was obtained.  ISCST was then rerun  with  the corrected inventory
and much closer agreement was found between CMB and  ISCST with  respect to
coal combustion impacts.
Case 3.   CMB estimates wood-fired boilers are a much larger contributor to
PM10 than does ISCST.
     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. There were no
obvious problems found for the  two wood-fired boiler sources in the inventory.
Therefore, a site visit was conducted to each of the two  plants.   The site
visits identified the possibility that one of the wood-fired boilers, located
at a lumber mill, could be emitting particulate matter at a much  higher rate
than was listed in the inventory.  Therefore, emission testing  was conducted
at the mill.  The emission testing demonstrated that the  emission inventory
had severely underestimated the emission rate of the wood-fired boiler located
at the lumber mill.  ISCST was  then rerun with the revised emission rate for
the lumber mill boiler and good agreement was found  between CMB and ISCST
with respect to the wood-fired  boiler source category.
                                     A-15

-------
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).
     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
PMjo emitting activities  within the steel mill.  This  disagreement results  in
CMB predicting that  the steel  mill is the largest industrial  source of the
PMjo 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  carefully,
     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
plasma emission spectroscopy analysis of the resuspended  samples which were
collected on qoartz  fiber and  the XRF analysis of those collected on  Teflon.
The next step was to examine the OM estimated impacts  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  fugitive emissions
from the steel plant.  This  omission was viewed as a potentially serious
deficiency; and, therefore,  a  site visit was conducted to reassess the stack
and fugitive emission rates  at the steel mill.  The results  of the stack
testing were in good agreement with the original stack related emission
                                    A-16

-------
rates.  However, substantial fugitive emissions were found to be associated
with the coking operations, blast furnace, and basic oxygen furnace.  ISCST
was then rerun 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.

Step 7 - Recompare Results and  Evaluate the Dispersion Model
     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 Step 4, the CMB was rerun and after Step
6, the ISCST was rerun for.subsets A and B using the revised data.  The
results obtained using the revised data are shown in Table A-5.  Good
agreement now exists between the source impacts estimated by CMB and ISCST
for subsets A and B.  Comparison of the ISCST results for subset A with the
observed minus background concentrations is also significantly improved.
ISCST was then run using 5 years of meteorological  data to obtain estimates
of the annual average PMjg and  the six highest values.  The results provided by
ISCST are shown in Table A-6.  ISCST predicts that  both the annual average
and 24-hour PM^g NAAQS will be  exceeded at the monitoring station.  Steel  mill
related activities are the major contributors to PMjo at the receptor site.
The second largest contributor  is resuspended road  dust.  The highest predicted
PMlQ concentrations all occur with persistent south winds, which further
substantiates the steel mill contributions.
                                     A-17

-------
Step 8 - Final  Model  Estimates



     The dispersion model  results  have  been  improved  after extensive interaction



with the CMB results.  Thus,  the dispersion  model  results  in  Table A-6 would



be used for control strategy  development.
                                    A-18

-------
                           TABLE A-5

  COMPARISON OF FINAL VALIDATION RESULTS BY ISCST AND CMB FOR
                    SUBSETS A AND B (ug/m3)
                                (Representing  Annual)
                                       Subset  A
(Avg.  of  5  24-hr  runs)
      Subset  B
Source Categories
Oil A Gas Combustion
Wood-f i red Boi lers
Coal Combustion
Coking Operations
Blast Furnace3
Basic Oxygen Furnace3
Coal Handling
Agricultural Prod. Handling
Paint Production
Cement & Limestone
Sand & Bentonite
Aluminum Production
Zinc Processing
Oil Refining
Fertilizer
Tire Production
Motor Vehicle Exhaust
Koad Dust X Soil
Secondary (NH4)2S04
Background
Total '
Measured
ISCST
1
3
2
10
14

0
1

1
1

0
0


15


!5>
68
67
.6
.5
.3
.1
.9

.3
.1
--
.9
.2
—
.8
.6
--
--
.1

—
JD
.4
.0
CMB

3
2
8
11
3










2
15
15

62
67
..
.4
.6
.4
.6
.8
—
--
--
--
__
--
--
—
--
--
.2
.1
.5

.6
.0

(1
(1
(3
(2
(1










(0
(3
(2




.6)
.4)
.7)
.5)
• 7)










.6)
.4)
.7)



ISCST
2
6
9
19
41

0
1
0
1
1
0
0
1
0
2
2
31

32
157
162
.0
.8
.2
.0
.0

.6
.6
.8
.9
.7
.4
.6
.6
.8
.2
.6
.2
--
.0
.0
.0
CMB
.
5.
8.
22.
31.
10.
-
-
_
_
_
_
-
-
.
.
2.
39.
28.
-
148.
162.
.
5
2
0
6
4
-
-
-
_
_
-
-
-
-
_
6
2
6
-
1
0

(2.2)
(3.0)
(8.5)
(6.9)
(2.6)










(0.8)
(8.1)
(4.4)



a estimated as a composite by the ISCST

NOTE:  Numbers in parentheses indicate one standard  error  in  the
       CBM analysis.
                              A-19

-------
                       TABLE  A-6

SOURCE CONTRIBUTIONS ESTIMATED  BY  ISCST  USING 5 YEARS OF
              METEOROLOGICAL  DATA  (ug/m^)
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
Background
Total
Annual
Average
1.4
3.0
2.0
8.7
10.1
2.7
0.3
0.9
0.1
1.6
1.0
0.1
0.6
0.3
0.1
0.4
12.7

15.0
61.0
24-hr 6th Highest
Concentration
1.7
6.0
7.5
27.6
38.0
14.8
0.4
0.4
1.1
2.1
0.7
0.2
0.6
2.1
0.9
1.5
45.4

32.0
183.0
                         A-20

-------
                                   TECHNICAL REPORT DATA
                            (PLasc rnJ Instructions an the rcveni before compltlinxi
1. REPORT NO.
   EPA-450/4-87-008
                                                            3. RECIPIENT'S ACCESSION NO.
4. TITLE ANOSUBTITLE
   Protocol For Reconciling Differences Among
     Receptor And  Dispersion Models
             5. REPORT DATE
                  March 1987
             6. PERFORMING ORGANIZATION CODE
7. AUTMOR(S)
   Air Management  Technology Branch
   Monitoring And  Data  Analysis Division
                                                            8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
                                                            10. PROGRAM ELEMENT NO.
   Office Of Air  Quality Planning And Standards  (MD 14)
   U.  S.  Environmental  Protection Agency
   Research Triangle,  NC  27711
              11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
                                                            13. TYPE OF REPORT AND PERIOD COVERED
                                                            14. SPONSORING AGENCY CODE
IE. SUPPLEMENTARY NOTES
   Project Officer:   Thompson G. Pace
16. ABSTRACT
   The PM,n SIP  Development Guideline,  EPA-450/2-86-001, prepared  by U. S. EPA,
   recommends that  both dispersion models  (DM)  and receptor models,  primarily  the
   Chemical Mass  Balance (CMB) model, be  used to determine the  contributing sources.
   When two or more independent methods,  such as DM and CMB, are used to develop
   estimates for  the contributing sources,  differences are bound to  occur.

   This orotocol  consists of an eight step  procedure for examining and reconciling
   differences in model estimates an a  systematic, justifiable  way.   The steps outline
   a procedure for  intercomparison of results,  model refinement, recomparison  and
   development of final model estimates.
17.
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
                                                                          c. COSATI Field/Group
   State  Implementation Plan
   Receptor Model ing
   Dispersion  Modeling

   PM10
   Model  Reconciliation
   Protocol
18. DISTRIBUTION STATEMENT
                                               19. SECURITY CLASS (This Keportj
                            21. NO. OF PAGES
                               42
                                               20. SECURITY CLASS (This page I
                           22. PRICE

EPA Form 2220-1 (R«v. 4-77)   PREVIOUS EDITION IS OBSOLETE

-------