United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Trlangfe Park NC 27711
EPA-450/2-86-001
June 1987
Air
PIN/hoSIP
Development
Guideline

-------
                              ERA-450/2-86-001
SIP Development  Guideline
      U, S, ENVIRONMENTAL PROTECTION AGENCY
           Office of Air and Radiation
       Office of Air Quality Planning and Standards
      Research Triangle Park, North Carolina 27711

                June 1987

-------
This guideline is issued by the Environmental Protection Agency to assist State and local air pollution control
agencies in developing State implementation plans (SIP's) for the national ambient air quality standards for
paniculate matter. The standards for particulate matter have been revised to address particles nominally 10
micrometers and smaller in diameter ( PM10 ). Copies are available - in limited quantities - from the Library
Services Office (MD-35), U.S. Environmental Protect ion Agency, Research Triangle Park, North Carolina 27711;
or, for a fee, from the National Technical Information Service, 5285 Port Royal Road, Springfield, Virginia 22161.
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.
                                 Publication No. EPA-450/2-86-001

-------
                           TABLE OF CONTENTS

                                                                 Page

1.0  INTRODUCTION	1-1

     1.1  Overview		  1-1

     1.2  Guideline Contents	  1-2
          1.2.1  Determining the air quality status of areas....  1-3
          1.2.2  Ambient monitoring	1-3
          1.2.3  Modeling	  1-3
          1.2.4  Emission inventories	  1-3
          1.2.5  Development of control  strategies	  1-4
          1.2.6  SIP requirements and data reporting	  1-4
          1.2.7  Types of receptor models	  1-4
          1.2.8  Preliminary estimate of PM^o design
                 concentration using TSP data	  1-4
          1.2.9  Source testing	  1-5

     1.3  General SIP Approach	  1-5

2.0  DETERMINING AIR QUALITY STATUS OF AREAS	  2-1

     2.1  Policy Overview	2-1

     2.2  Demonstrating Attainment	  2-2

     2.3  Estimating Probability of Nonattainment to	  2-4
          Determine Initial SIP Requirements

     2.4  Area Categorization	  2-7

     2.5  Area Boundaries	  2-9

3.0  AMBIENT MONITORING AND DATA USAGE	  3-1

     3.1  Ambient PMio Monitoring	  3-1

     3.2  Part 58 Requirements	  3-1
          3.2.1  Quality assurance	3-1
          3.2.2  Ambient monitoring methodology	  3-2
          3.2.3  Network establishment 	  3-3
          3.2.4  Network design	  3-3
          3.2.5  Sampling interval	  3-4
          3.2.6  Sampler inlet siting	  3-5
          3.2.7  Ambient data reporting	  3-6
          3.2.8  Air pollution index	  3-7

-------
                                                                 Page

     3.3  Ambient Samplers and Ambient Data Usage	  3-7
          3.3.1  Ambient samplers in use	3-7
          3.3.2  Interim use of nonreference or equivalent
                 method samplers	  3-7
          3.3.3  Episode monitoring	  3-8

     3.4  SIP Revisions	  3-8

4.0  AIR QUALITY MODELING FOR PMiQ	,.  4-1

     4.1  Introduction	  4-1

     4.2  Considerations in Receptor Model
          Selection	4-2
          4.2.1  Particle size	.,  4-2
          4.2.2  Prior knowledge of sources and emissions ......  4-4
          4.2.3  Chemical similarity	  4-4
          4.2.4  Sources versus source categories	  4-1
          4.2.5  Particle size and time scale	,	4-5

     4.3  Screening Techniques and Refined Dispersion  Models
          for PM^o Concentrations 	,	4-7
          4.3.1  Selection of appropriate
                 dispersion models	  4-8
          4.3.2  Special considerations for PMiQ
                 dispersion modeling	4-9

     4.4  Receptor Models for Estimating PMio Concentrations ...  4-11
          4.4.1  Control strategy analyses  using receptor
                 models	4-11
          4.4.2  Preliminary analyses using receptor models  ....  4-12

     4.5  Use of Receptor and Dispersion Models in Combination  .  4-13

5.0  DEVELOPMENT OF WIQ EMISSION INVENTORIES	  5-1

     5.1  Overview	  5-1

     5.2  PMjo Emission Factors and Fractional  Multipliers 	  5-3

     5.3  PMiQ Source Data for Dispersion Models	  5-20

     5.4  Inventory Data Needed in Receptor Models	  5-29

     5.5  Condensable Particulate	,.	5-32

     5.6  Secondary Particulate	  5-32

     5.7  Use of Existing Emission Inventory	5-35

     5.8  Data Handling	5-38
                                   •n

-------
                                                                 Page
6.0  DEVELOPMENT OF CONTROL  STRATEGIES	  6-1
     6.1  Overview 		,	  6-1
     6.2  Data Base Requirements	,	6-2
     6,3  Methodologies for Determining Design  Concentrations  ..  6-2
          6.3.1  Table look-up	6-3
          6.3.2  Fitting one statistical  distribution to
                 several  years of data	6-6
          6,3.3  Using the empirical  frequency  distribution of
                 several  years of data  (graphical estimation) ..  6-7
          6.3.4  Conditional probability approach  	  6-8
     6.4  Determining Emission Limits ,	  6-8
          6.4.1  General	6-8
          6.4.2  Example for annual averages  	  6-9
          6.4.3  Considerations for 24-hour averages  	  6-11
7.0  SIP REQUIREMENTS AND DATA REPORTING	  7-1
     7.1    Introduction	7-1
     7.2    Clean Air Act Requirements	  7-1
          7.2.1 Time limits, SIP requirements	,	  7-1
          7.2.2 Regulations  in 40 CFR Part 51  	  7-2
     7.3    PMiQ SIP Development  Policy 	7-2
          7.3.1 Group I PMiQ SIP requirements  	.,..  7-2
          7.3.2 Group II PMio SIP requirements  	»	  7-3
          7.3.3 Group III PMio SIP requirements	  7-5
      7.4    SIP Content 	.,..	  7-5
      7.5    Control  Strategy Transition	  7-6
      7.6    Current Emission Regulations  	  7-7
      7.7    Surrogate Emission Regulations	  7-8
          7.7.1 Retaining existing emission limits	7-8
          7.7.2 Setting new  emission limits	,	  7-10
      7.8    PMiQ-Specific Emission Regulations 	  7-10
      7.9    Reporting Emissions  Data	  7-11
      7.10   Emissions Trading (Bubble)  Policy	  7-13
      7.11   Fugitive Dust Policy	,.  7-13

-------
                                                          Paje

APPENDIX A - TYPES OF RECEPTOR MODELS	  A-l
1.0  Chemical Mass Balance	  A-l
2.0  Factor Analysis	,	  A-2
3.0  Optical Microscopy	  A-2
4.0  Automated Scanning Electron Microscopy ...............  A-2
5.0  Other Receptor Models	  A-3
     5.1  X-Ray Diffraction (XRD)	  A-3
     5.2  Trajectory Analysis 	  A-3
     5,3  Microinventory	  A-3
APPENDIX B - PRELIMINARY ESTIMATES OF PMiQ DESIGN CONCENTRATIONS
             USING TSP DATA	  B-l
1.0  Annual NAAQS	  B-l
     1.1  Recommendations	  B-l
     1.2  Example	  B-2
2.0  24-Hour NAAQS	  B-2
     2.1  Background	  B-2
     2.2  Recommendations	,	B-4
     2.3  Example 	.,,	  B-6
APPENDIX C - GUIDELINES FOR SOURCE TESTING FOR  SIZE SPECIFIC
             PARTICULATE EMISSIONS	  C-l
1.0  Introduction	  C-l
2.0  Available Methods for Size Specific Source
     Measurements	  C-2
3.0  Selection of Sampling Traverse Points	  C-3
4.0  Sampler Selection ind Operation	  C-7
5.0  Sampling Trains	  C-9
6.0  Collection Media	  C-10
7.0  Data Reduction and Analysis	  C-13
                                   iv

-------
APPENDIX D - DETERMINATION OF PMio BACKGROUND

1.0  Introduction ........... ........... .....................  D-l
2.0  Determining Background Estimates from Measured PM^o or
     Surrogate Data ... .................. .... ........ ........  D-2

3.0  Special Considerations for Receptor Modeling ...........  D-5

4.0  Summary of Alternatives for Estimating PMiQ Background .  D-6
5.0  Example 1 - Compositing Data from Several Perimeter
     Sites ., ..... .... ...... ........ ............. . ...... ..... D-6

6.0  Example 2 - Use of Surrogate Data ................ . ..... D-8

-------

-------
          LIST OF TABLES AND FIGURES
TABLE 4-1



TABLE 4-2


TABLE 4-3

TABLE 5-1
FIGURE B-l
             SELECTING FEASIBLE SOURCE APPORTIONMENT METHODS
             BASED ON DATA AVAILABILITY AND SOURCE
             CHARACTERISTICS ......... , ................. ,.
                                                    Page
                                                    4-3
             RECOMMENDED APPROACHES FOR PMio SOURCE
             APPORTIONMENT .. ....... . ..... .
             DISPERSION MODELS APPLICABLE TO
                                                    4-6

                                     ANALYSES .....  4-9
PMio EMISSION FACTORS TO BE AVAILABLE FOR
INDICATED SOURCE CATEGORIES, PROCESSES AND CONTROL
TABLE 5-2
TABLE 5-3
TABLE 5-4
TABLE 5-5
TABLE 5-6
TABLE 5-7
TABLE 5-8
TABLE 5-9
TABLE 6-1
FIGURE 5-1
CUMULATIVE EMISSION FACTORS AND PARTICAL SIZE
DISTRIBUTION FOR A HYPOTHETICAL LIME PROCESS ....
INDUSTRIAL SOURCE COMPLEX MODEL (ISC) 	
CLIMATOLOGICAL DISPERSION MODEL (COM 2.0}........
RAM 	 	 	
SINGLE SOURCE (CRSTER) MODEL 	 	 	
MULTIPLE POINT SOURCE MODEL WITH TERRAIN
ADJUSTMENT (MPTER) 	 	 , 	
EXAMPLE OF DATA IN EPA'S PARTICULATE SOURCE
LIBRARY 	 	 	 	 	
POTENTIAL SOURCES OF CONDENSABLE PARTICULATE ....
TABULAR ESTIMATION OF PMjo DESIGN CONCENTRATIONS
EXAMPLE GRAPHICAL DISPLAY OF CUMULATIVE EMISSION
... 5-18
... 5-24
,. , 5-25
,.. 5-26
.. 5-27
... 5-28
,.. 5-31
... 5-33
.. 6-5
FACTORS VS. PARTICLE SIZE FOR A HYPOTHETICAL
PROCESS ............... . ................... ,, ...... 5-19

DISTRIBUTION OF PMio/TSP  RATIOS FOR INDIVIDUAL
24-HOUR OBSERVATIONS ............ . .......... , ...... B-5
                      VI

-------
FIGURE C-l   RECOMMENDED SAMPLING POINTS FOR CIRCULAR AND
             SQUARE OR RECTANGULAR DUCTS	 C-7

FIGURE C-2   PM10 PARTICULATE SAMPLING! TRAIN FOR NONCONDENSABLE
             PARTICULATE (MODIFIED EPA METHOD 5 TRAIN) 	 G-12

FIGURE C-3   PM10 PARTICULATE SAMPLING TRAIN FOR CONDENSABLE
             AND NONCONDENSABLE PARTICULATE (MODIFIED EPA
             METHOD 5 TRAIN) 	 C-13
                                   VII

-------
                            1.0  INTRODUCTION
1.1  Overview
     The 1977 amendments to the Clean Air Act (Act) Pub.  L,  95-91)
require the Environmental Protection Agency (EPA) to review  periodically
and, if appropriate, revise the criteria on which each national  ambient
air quality standard (NAAQS) is based along with the NAAQS themselves.1»2
In response to these requirements,  EPA revised the primary and secondary
NAAQS to apply to particulate matter in a size range defined by  the
collection characteristics of a new ambient reference method that has  a
50 percent collection efficiency (Dgg) at 10 micrometers.3'4  The material
collected by the reference method is nominally below 10 micrometers and
is referred to as "PMjo-"
     The purpose of this document is to provide guidance on  how States
can develop PMjQ control programs for attaining and maintaining  ambient
PMjQ standards.  Existing State control strategies to attain and maintain
the former NAAQS for total suspended particulates (TSP) provided for
controlling particulate matter sources; however, they must be evaluated
and revised as necessary to attain  or maintain the PMjQ NAAQS.  The
guidance herein addresses the transition that States must make from TSP
control programs to PMjQ control programs and the necessary  revisions  to
State implementation plans (SIP's)  to account for PMjQ.
     This document is intended as a starting point for anyone seeking
information, policy, and guidance on developing PMjg SIP's.   Most aspects
of developing a PMjg program from problem determination to final SIP
submittal are covered.   Certain details of program development,  however,

                                   1-1

-------
require explanation not included herein.  Reference is therefore made to
other documents for those additional details.
     Guidance for revising existing SIP's to address the review of PMjQ
under State and local p reconstruct! on review programs is not included in
this document.  It should be noted, however, that EPA's decision to
implement the revised particulate matter NAAQS via section 110 of the Act
means that EPA will not Impose Part D new source review (NSR) requirements as
a means of revising the SIP's.  Instead, the preconstruetion review of major
new and modified stationary sources which emit PMig will be carried out
largely under regulations for the prevention of significant deterioration (PSD).
The reader is referred to section IV.D in the preamble of the FEDERAL REGISTER
notice promulgating PM^o amendments to the PSD regulations for a description
of the changes that EPA has made to address PMiQ.  Additional guidance
will be forthcoming in the form of workshop and policy and guidance
memoranda to address revisions needed for State and local  preconstruction
review procedures.
1.2  Guideline Contents
     Since the size range of particles covered by the PM^Q NAAQS is
different from that included in TSP, the actions that States will  have
to take in preparing PMjg control programs are more complicated than they
would be if only the levels or averaging method of the NAAQS had been
changed.  As a result, States must look at the spatial distribution of
ambient PMjg concentrations and at source emissions from a particle size
distribution standpoint.  The major sections of this document are intended
to provide guidance on the different aspects of PMjg control strategy
                                  1-2

-------
development and SIP submittal to EPA in light of this  new focus  on  the
PM^Q size range.
     1.2.1  Determining the Air Quality Status of Areas
     The Act requires control strategies to be submitted to EPA  within
9 months of the revision of NAAQS.   Section 2.0 describes EPA's  policy
for determining when the air quality of an area is or  may be violating
the PMiQ NAAQS and the extent of the area violation.   The EPA's  policy  for
categorizing areas into three groups based on the need to revise their
SIP is also discussed.
     1.2.2  Ambient monitoring
     Each State must establish an ambient monitoring network utilizing
PM^o samplers.  Section 3,0 discusses the changes made to the ambient
monitoring requirements of 40 CFR Part 58.  These changes address PHiQ
ambient network design, various aspects of network implementation and
operation, ambient data reporting,  and the air pollution subindex for
PMjg.  Other facets of ambient monitoring are also discussed, such  as
episode monitoring and the use of data obtained by non-reference or
equivalent method samplers.
     1.2.3  Modeling
     Section 4.0 discusses modeling PMiQ with dispersion models  or  with
receptor models and describes how the two techniques can be interfaced.
The section specifies models or combinations that States should  use
depending on various considerations such as source types, modeling
objectives, and available ambient data.
     1.2.4  EmissionInventories
     Where States use dispersion models to assess the  impact of  a source
or sources on ambient PMjQ levels,  the model inputs must be in terms
                                   1-3

-------
of PMjo emissions.  Section 5.0 discusses how PMjQ emissions can be
calculated and explains In detail the type of Information needed for
various dispersion models now available as well as models now under
development.  Data needs for receptor models are also included.
     1.2.5  Development of control strategies
     Section 6.0 identifies general approaches to the use of ambient
measurements and model estimates in setting control strategy emission
limits.  It explains how design concentrations may be determined and how
control strategy demonstrations may be conducted using models.
     1.2,6  SIP requirements and data reporting
     Section 7.0 discusses EPA's policies for requiring PMjo SIP's based
on area categorization and for transferring from a control strategy for
TSP to one for PMjo-  Also discussed are PMjo emission data reporting
requirements in 40 CFR 51.322, the fugitive dust policy, and the emission
trading policy.
1.2.7  Typesof receptor models
     Appendix A lists various receptor model techniques available and
discusses each.
1.2.8  Preliminary Estimate of PMio DesignConcentration UsingTSP
       Data
     Appendix B provides a method for making a preliminary estimate of
the PM^o design concentration when only TSP data are available; however,
this estimate alone cannot be used for developing" control strategies.  In
addition to providing a preliminary estimate, it can also provide useful
information for evaluating design concentrations estimated by dispersion
modeling.
                                   1-4

-------
     1.2.9  Source testing
     One of the methods for determining a source's PHjo emissions  1s
to conduct a test which measures the amount  of emissions plus  particle
size distribution.  Appendix C discusses the state-of-the-art  for  such
source testing and provides information upon which States can  base an
emissions measurement or compliance test method for PMjo«
1.3  General SIP Approach
     States have many years of experience in developing and implementing
plans for abating particulate matter air pollution.  In general, the
activities States will engage in to attain PMip NAAQS will  not differ
radically from past activities to attain TSP NAAQS.  That is,  the  basic
approach will still be to examine air quality across the State, delineate
areas where air quality needs improvement, determine the degree of
improvement necessary, inventory the sources contributing to the problem,
develop a strategy to reduce emissions from contributing sources enough
to bring about attainment of the NAAQS, implement the strategy, and take
the steps necessary to ensure that NAAQS are not violated in the future.
     The change in the particle size range to which the standards  apply
from TSP to PHig, together with the emergence of receptor models as
effective tools in strategy development, necessitates changes  in the
specifics of particulate matter control strategy development.   The major
difference in past TSP strategies and new PHjo strategies will  be  the
need to Inventory the sources of the PMip size fraction and to focus on
control of PMjQ emissions to reduce ambient  PMjg.  This could  include
applying PMjQ emission factors to sources, in some cases testing sources
for PM^o emissions, and utilizing receptor models, where feasible  to refine
                                  1-5

-------
emission Inventories and augment dispersion models.  The contribution of
sources to ambient PM^o and the degree of emission reductions necessary
can then be determined.
                                   1-6

-------
References
1.  Air Quality Criteria for Particulate Hatter,  AP-49,  U.S.  Department
    of Health, Education, and Welfare, Washington, D.C., January 1969.

2.  Air Quality Criteria for Particulate Matter and Sulfur Oxides,
    EPA-60Q/8-82-Q29a-c, Environmental Criteria and Assessment  Office,
    Environmental Protection Agency,  Research  Triangle Park,  N.C.,
    March 1983.

3.  High Volume Method for Suspended  Parti oil ate Matter, Part 50 of
    Chapter I of Title 40 of the Code of Federal  Regulations, Appendix B,

4.  Reference Method for the Determination  of  PMjg in the Atmosphere,
    Part 50 of Chapter I of Title 40  of the Code of Federal Regulations,
    Appendix J.
                                   1-7

-------

-------
                2.0  DETERMINING AIR QUALITY STATUS OF AREAS

2.1  Policy Overview
     The Act requires that States develop SIP's that provide for timely
attainment and maintenance of NAAQS.  The method of determining whether an
area Is In attainment of the PMjo NAAQS using ambient air quality data is
described in Appendix K of 40 CFR 50.  Generally, 3 years of PMiQ data are
required to make the determination; however, data collected over shorter
periods of time may be adequate as explained in Appendix K and in the
Guldeline on Exceptions to Data Requirements for Determining Attainment: of
ParticulateHatter Standards.^ In areas where too little PMjg data are
available to demonstrate attainment of the NAAQS, EPA's policy is to use
whatever PMiQ data are available and supplement it with TSP data to determine
the probability that the area will violate the PMio NAAQS.  The EPA will
use the report. Procedures for Estimaten£ Probability of Nonattainment of a
PM^QNAAQS Using Total Suspended Particulate or PM|Q Data,2 hereafter referred
to as the "probability guideline," for calculating an area's probability of
nonattainment.  Areas will be grouped Into three categories based on their
probability of nonattainment and other factors influencing the degree to
which the existing SIP may need revisions.  Generally, Group I areas are
shown to be or have high probability of nonattainment, Group II areas are
those areas where nonattainment is uncertain, and Group III areas are shown
to be or have a high probability of attaining the NAAQS.  After the initial
grouping of areas and promulgation of the PMiQ NAAQS, only PMjo data may be
used to make definitive attainment/ nonattainment decisions in accordance
with Appendix K or reference 1, above.   The probability guideline may
                                    2-1

-------
continue to be used for planning purposes such as to assess new areas  where
TSP data Indicate a need for more extensive PMjo monitoring,
2.2   DemonstratlniAttainment
     Appendix K specifies the procedures which States can use to demonstrate
attainment of the 24-hour and annual PHjo standards at a particular location
using ambient air quality data.  To determine PMio attainment of the 24-hour
standard, EPA generally requires averaging the number of estimated 24-hour
exceedances over 3 consecutive years of representative data.   If the average
number of 24-hour NAAQS exceedances is greater than 1.0 per year, then a
nonattainment problem will be evident and States should proceed to examine
the adequacy of the existing SIP.  A potential problem of nonattainment of
the 24-hour PM]Q standard can be projected on the basis of fewer than  3 years
of data, since the criteria for nonattainment can be equivalently expressed
in terms of the total number of estimated exceedances within  a 3-year
period.  In the simplest case, if four exceedances are observed in a single
year, the average number of exceedances over a 3-year period  will exceed
1.0.  If the total number of estimated exceedances in 3 years is equal to
or greater than 3.2, then the number of exceedances averaged  over 3 years
would also exceed 1.0.  Thus,
                3.2/3 = 1.07, which rounds to 1.1
                3.1/3 = 1.03, which rounds to 1.0
     As soon as 3.2 estimated exceedances are accumulated a nonattainment
problem will be evident.  This may occur in even less than a  single year.
     As specified in Appendix K, EPA will not require adjustment of the
first observed exceedance in order to account for incomplete  sampling  if
(1) only one exceedance occurred in the calendar quarter, (2) if everyday

                                    2-2

-------
sampling is initiated as expeditiously as possible as required by 40 CFR 58.13,
and  (3) 75 percent data capture is achieved (see section 3.2.5 of this
document).  As specified in 40 CFR 58.13 for the first year of monitoring,
everyday sampling must commence within 90 days of the end of the calendar
quarter in which the first exceedance is observed and continue for 4 consecutive
calendar quarters.  In addition, if a site is already monitoring every day
and observes Its first exceedance, no adjustment for missing data will be
made to this first exceedance if 75 percent data capture is achieved.
     By not adjusting the first observed exceedance, EPA is assuming that
this first exceedance may be the only actual exceedance during the entire
year.  If, however, one or more subsequent exceedances are observed during
the transition period following the first exceedance, but prior to the
actual initiation of every day monitoring, the additional exceedances would
be adjusted to account for incomplete sampling using formula [1] of Appendix K,
according to the sampling frequency and data capture at that time.
     Days on which exceedances of the NAAQS were recorded may be flagged as
exceptional events in accordance with the Guidel1ne on the Identffic_at1o_n
and Use of Air Quality Data Affected by Exceptional Events, EPA-450/4-86-007.3
State or local control agency flagged data values may be excluded from initial
NAAQS calculations and not be considered by EPA in determining an area's
apparent nonattainment status.  The. flagged data will be excluded from
final consideration of SIP adequacy if the responsible control agency
determines in conjunction with a public review that the flagged data are
inappropriate for use.
     When the minimum 3 year data requirements specified in Appendix K are
not satisfied, attainment can still  be demonstrated, but such demonstrations
                                    2-3

-------
must be approved by the appropriate Regional  Administrator,  in  accordance
with the Guideline on Exceptions to Data Requirements for Determining
Attainment of Particulate Matter Standards.1   In particular,  If only 1 or
2 years of PMio data are available, then supplementary information provided
by modeling or other surrogate measures must  be used to demonstrate the
adequacy of the SIP to attain and maintain the standards.
     Finally, 1n the event that there has been a trend in emissions during
the attainment test period, attainment with the  standards can  be demonstrated
according to Appendix K, even though the most recent 3 years of data do  not
indicate attainment.  Either the most current representative year(s) could be
used or statistical techniques or models could be used in conjunction  with
previous years of data to adjust for trends.   Such analyses would be performed
in accordance with existing guidance and must be approved by the appropriate
Regional Administrator.4
2.3   Estimating Probability ofNonattainment to Determine Initial SIP
      Requirements
     The procedures EPA will use to categorize areas into Group I, II, or
III are discussed in section 2.4.  The schedules for submitting SIP revisions
and control strategy demonstrations for each  group are discussed in section
7.3.  If available PHjo data for an area are  adequate to demonstrate attainment
as discussed in section 2.2, then the area will be put into Group III.   On
the other hand, if the PM^Q data alone are adequate to demonstrate that  the
area is nonattainment, it will be put into Group I.  If available PMjo data
alone are not adequate to demonstrate attainment or nonattainment, EPA's
policy is to supplement available PMig data with inhalable particulate matter
data (particles with an aerodynamic diameter  less than 15 micrometers) and TSP
                                    2-4

-------
data.  These air quality data will be used In a statistical  methodology  as
described in the probability guideline to estimate the probability  that  the
PMjQ NAAQS will be violated in the area.  Initially, areas will  be  grouped
based on probabilities of nonattainment of 95 percent or greater (Group  I),
from 20 to 95 percent (Group II) and less than 20 percent (Group III).
Other factors that may influence the degree to which the existing SIP may
need revisions will also be considered in the area grouping  process.
     Also, EPA has found that some uncertainty exists in the PHjQ data
collected prior to 1987 with available PMio instruments.  A  study performed
by EPA in Phoenix has shown that in extreme situations data  collected by
Sierra Anderson SA-321A size selective PM^o instruments can  be influenced
by coarse particles to the extent that concentrations may be biased high by
up to 20 percent.  In addition, data collected with Wedding  or GMW-9000
instrument may be biased low by up to 20 percent due to soiling problems.5t6
In order to account for the uncertainty associated with such PMio concentrations
in calculating the probability of nonattainment, a zone of uncertainty or
"gray zone" of +_ 20 percent is being placed around the standard (0.8 NAAQS to
NAAQS, lower gray zone-, NAAQS to 1.2 NAAQS, upper gray zone).   By design,
the zone of uncertainty will only cause areas to move from Group I  to Group II,
or from Group III to Group II.  In particular, for SA-321A instruments,
24-hour PHio observations and annual PMip means (using all data)  within  the
upper gray zone are not to be counted as exceedances of the  respective
standards in the probability calculations.  Similarly, for the GMW-9000
instrument, 24-hour PM^g values and annual PMjQ means (using all  data) that
are within the lower gray zone are to be counted as potential  exceedances
of the respective standards in the probability calculations.

                                    2-5

-------
     If an area's nonattalnment probability using TSP data and PMiQ  data  drops
below 0.95 or rises above 0.20, as a result of PMjo data 1n the "gray  zone,"
it will be classified Group II in order to resolve the possible uncertainty
associated with the PMJQ data and to ensure that a determination is  made  as to
whether the existing SIP provides for attainment and maintenance of  the PNio
standards.  Also, PMio data in the gray zone, by itself, will  not be allowed
to move an area into Groups I or III from any other Group.  Areas classified
Group II as a result of PMjg data in the upper "gray zone" are still required
to operate air quality monitors on an everyday sampling schedule for the
first year.  This is consistent with the requirements for Group I areas as
explained in section 3.2.5.
     The following table presents the actual values associated with  the "gray
zones.
                                  TABLE 2-1
                         PK1n Gray Zone Limits(ug/m3)


Averaging
Time
Annual
24-Hours
PMio
NAAQS
50
150
Gray Zone Limits
Upper
60*
180*
Lower
40**
120**
     *Applies to readings made with a SA-321A
    **Applies to readings made with a GMH-9000
     Since the conditions under which these potential  instrument biases  were
observed are not considered typical, data collected with all  instruments
will be taken at face value when demonstrating attainment or  nonattainment
                                    2-6

-------
with the standards.  For data produced by the SA-321A (with the potential
positive bias), however, an appropriate downward adjustment will be
permitted for attainment demonstrations if influence by coarse particles
can be demonstrated.  For data produced by the GMW-9000 (with the potential
negative bias), no adjustments to observed data are necessary for demonstration
of nonattainment.  Such areas with potential exceedances of the PMjQ NAAQS
(i.e. with data in the lower gray zone) will be required to sample more
frequently for the first year (according to the Group II monitoring frequency
of every-other-day), so that true nonattainment situations will soon become
apparent.
2.4 Area Categorization
     SIP development requirements are based on an area's categorization
into Group I, II, or III of EPA's policy for PMjQ SIP development.  If
ambient PM-|p data adequate to determine attainment or nonattaiment of PMig
standards are available, they are to be used in categorizing areas.  Areas
for which adequate ambient PM-jg data are not available are initially categorized
into these groups using nonattainment probability (section 2.3) cut points
of 95 percent to distinguish between Group I and Group I! and 20 percent
to distinguish between Group II and Group III.  The EPA Regional Offices,  in
conjunction with State and local agencies may then make adjustments to
the grouping based on the adequacy of the existing particulate matter
SIP's for achieving and maintaining PM-JQ standards.
     The general criteria for group categorization are:
      Group  I  - Areas for which the existing particulate matter SIP may
                  need substantial revision to be adequate for attaining
                  and maintaining PM-jp standards.
                                    2-7

-------
      Group  II - Areas for which the existing particulate witter SIP
                  may be adequate or need only minor adjustment.
      Group III - Areas for which the existing partlculate matter SIP's are
                  believed adequate to attain and maintain the PMjo standards,
     Examples of Information other than air quality data that may warrant
moving an area from Group III to Group II, or from Group II to Group I,
are:
     0 facts showing that the current air quality is attributable to an
economic slowdown or some other temporary phenomenon rather than  the
stringency of the area's TSP SIP requirements,
     0 facts suggesting that there are few enforceable measures 1n existing
SIP's yet to be Implemented that would reduce emissions that significantly
affect air quality, and
     0 evidence that the area has an unusually high proportion of sources
In categories whose emissions typically have a higher ratio of PMio to
total particulate matter.
     Examples of information other than air quality data that may warrant
moving an area from Group I to Group II, or from Group II to Group III, are:
     ° factors suggesting that sources are not yet in compliance  with SIP
measures that, if enforced would reduce emissions that significantly
affect the area's air quality;
     "evidence that the area has an unusually high proportion of  sources
in categories whose emissions typically have a low ratio of PMio  to total
partlculate matter; and
                                   2-8

-------
     0 evldtnce that the area 1s rural 1n nature and 1s a clearly Impacted
by fugitive dust (I.e. qualifies as a rural  fugitive dust area under
the fugitive dust policy described In section 7.11),
     Examples of information that may affect EPA's classification of an
area not preliminarily classified using a nonattainment probability are:
     0 the amount and density of industrial  activity that would likely
result in significant ambient PMjQ concentrations in the area;
     0 the number and density of roadways In the area that are near activities
likely to generate significant particulate matter emissions and that are  subject
to moderate and heavy vehicle traffic; and
     0 the degree to which the existing TSP SIP will  likely limit PMiQ emissions
from these traditional and nontraditional  sources.
2.5  Area Boundaries
     Section 6.0 of the probability guideline describes several  approaches
for determining the boundaries of an air quality level  represented by one
or more monitoring stations.  In some instances the spatial extent of areas
of uniform PHjo air quality status may differ from the areas previously
defined for TSP; therefore, States may wish to use the guidance  in these
instances to more specifically define the spatial  extent of PMjo air quality
problems.
     The approaches that have been used and that are recommended for
defining area boundaries are:
     -- a qualitative analysis of the area of representativeness of the
        monitoring station, together with consideration of terrain,
        meteorology, and sources of emissions;

                                    2-9

-------
     -- spatial interpolation of air monitoring data;  and
     -- air quality simulation by dispersion modeling.
These techniques can be used singly or in combination  depending  on the
complexity of the area where a monitoring statlon(s) is  located.  More
detailed discussions of the techniques are in the probability  guideline.
                                    2-10

-------
References
1.   Guideline on Exceptions to Data Requirements  for Determining Attainment
     of Participate Matter Standards, U.S.  EPA,  RTP,  N.C.
     EPA-45Q/4-87-OQ5, April 1987.

2.   Pace, T. G., N. H. Frank, E. L. Meyer,  and  S.  F. Sleva,  Procedures for
     Estimating Probability of Nonattalnment of  a  PHin NAAQS  Using Total
     Suspended Parti cul ate or PHm Data,  U.S.  EPA,  RTP,  N.C.
     EPA-450/4-86-017, December 1986.

3.   Guideline on Identification and Use  of Air  Quality Data  Affected  by
     Exceptional Events, EPA- 450/4-86-007,  U.S.  EPA,  RTP,  N.C. July  1986.

4.   Guidance on Accounting for Trends in Parti cul ate Hatter  Emission  and
     Air Quality Data, memorandum from R.G.  Rhoads  to Regional Air Directors,
     U.S. EPA, RTP, NC.  May 11, 1987.
5.   L. Purdue et. al., Intercomparison of High-Volume PMio  Samplers  at  a
     Site with High Particulate Concentrations,  JAPCA, Vol.  36,  No.8, Aug.  1986,
     pp. 917-920.
6.   Uncertainty of PMiQ Data Collected with High-Volume Size Selective
     Inlets, Memorandum from L.J.  Purdue to Richard Rhoads,  U.S.  EPA,
     RTP, N.C. September 15, 1986.
                                    2-11

-------

-------
                  3.0  AMBIENT MONITORIN6 AND DATA USAQE

3.1  Ambient PMip Monitoring
     States must establish ambient monitoring networks to measure PM]Q just
as they measure ambient levels of other pollutants for which NAAQS have
been set.  Requirements for a PMio network are Included in 40 CFR 18 as
are the ambient monitoring requirements for other pollutants.  Existing
ambient monitoring networks measure total suspended particulate with high
volume samplers (hi-vols.)*  Data from hi-vols, as discussed previously in
section 2.0, will not allow sufficiently accurate estimates of PM-JQ levels
to determine PM]Q attainment or nonattainment with certainty.  Ambient samplers
meeting the requirements of 40 CFR 50 and 53 will be necessary to meet the
ambient monitoring requirements of Part 58 except as specified in Appendix C
of Part 58.
     The requirements for PM-JQ reference or equivalent methods with a nominal
DSQ of 10 micrometers are contained in 40 CFR 50, Appendix J, and in 40 CFR
53.  The EPA will designate reference and equivalent method samplers as
soon after promulgation of these requirements as possible.  Some provisions
are made, however, as discussed in subsection 3.3, for using other ambient
data until data from ambient PM]Q networks are available.
3.2  Part 58 Requirements
     3.2,1  Quality Assurance
     The quality assurance (Q/A) requirements for PM^Q samplers are essentially
the same as for hi-vols.  Only one minor change was made to the quality
*Described in 40 CFR 50 Appendix B of Part 50
                                    3-1

-------
assurance procedures in Appendix A of Part 58.  Section 4.2.1(a)  of Appendix A
requires only that the percent differences for paired measurements  from
collocated samplers above certain levels (20 ug/m3 for PM|0)  be calculated
and  reported.  Previously, measurements below these levels were reported.
Otherwise, Q/A procedures for PMio monitoring in both Appendices  A  and B are
identical to those for TSP monitoring.
     3.2.2  Ambient monitoring methodology
     The requirements in Appendix C of Part 58 which require  reference or
equivalent method samplers be used in State and Local Air Monitoring Stations
(SLAMS) apply to PMio with one exception.  Section 2.2 of Appendix  C allows
the  hi-vol method to be used in a PM]g station as a substitute for  a PM-JQ
reference or equivalent method sampler as long as the ambient concentrations
measured by the hi-vol are below the levels of the PM]g NAAQS.  In  such an
instance, compliance with PM]p NAAQS is assured if the hi-vol  levels are
consistently below those NAAQS; thus, there is no real need from a  compliance
standpoint to install a PMio sampler.  If TSP levels rise above PM]p NAAQS,
however, the State must install PM-jg samplers in order to measure actual
PM]Q levels to be certain PM]Q NAAQS are maintained.
     Also, section 2.2 of Appendix C requires that at TSP National  Air
Monitoring Stations (NAMS) the hi-vol be continued in operation for at
least a year after the PMio monitoring begins.  This will allow a site-specific
relationship between TSP and PMjQ data to be developed and provide  assistance
in checking validity of PMjg data.  Historical trends for ambient levels
of particulate matter can thus be estimated.
      Section 4.0 of Appendix C, which previously applied to  episode
monitoring for TSP, has been revised to apply to
                                   3-2

-------
     3.2.3  Network establ1shment
     Title 40 CFR 58.20 requires a description of the SLAMS network  be
available to the general public and be submitted to EPA approximately 6 months
after PMio NAAQS are promulgated.  The network description need not  be
Included as part of the SIP; it need only be kept updated and made available
for public inspection.  Two dates, 1 year and 2 years after promulgation,
are included in section 58.23 for completion of the PM]0 SLAMS network. By
1 year after promulgation of the NAAQS, each area within the approved
SLAMS network for which a probability of PM|g NAAQS nonattainment is greater
than or equal to 20 percent, must have at least one PM-jo sampler in  operation
which is (1) located in the area of expected maximum concentration (2)  sited
in accordance with Appendix E, (3) located as described on the station's
Storage and Retrieve! of Aerometric Data (SAROAD) identification form,  and
(4) meeting all quality assurance requirements in Appendix A pertinent  to
PMiQ.  The remaining PM-|Q samplers, including those in areas with nonattainment
probabilities below 20 percent, have until  2 years after promulgation of
the NAAQS to be fully operational and to meet the siting and quality
assurance requirements.
     The same submittal date (6 months after promulgation) also applies
for the description of the NAMS portion of the SLAMS network as required
by section 58.30.  The date by which all stations in the NAMS portion of
the SLAMS network are to be in operation and meeting Part 58 requirements
is 1 year after PM^o NAAQS are promulgated (section 58.34).
     3.2.4 Networkdesign
     Information on designing SLAMS and NAMS networks is contained in
Appendix D of Part 58.  The material is largely self-explantory and, as
                                   3-3

-------
indicated in Appendix 0, should be used jointly by States and  EPA Regional
Offices for designing ambient monitoring networks.  To account for PM]Q
network design, Appendix D has been amended to include criteria for deter-
mining the number of PM^o stations and areas in which to locate them.   In
addition to Appendix D» a document entitled Network Design and Optimum
Site Exposure Criteria for Parti.culate Matter* is available and provides
further information on designing PM]Q networks and siting ambient samplers.
     3.2.5 Sampling interval
     The sampling interval for PM]Q data collection will not necessarily
be once in 6 days as it has been for TSP.  Short-term and long-term
strategies have been added to section 58.13.  During the first full year
of sampling the short-term strategy requires daily sampling at at least one
site (the site of expected maximum concentration) in areas with a WIQ
nonattainment probability of 95 percent or greater.  Sampling  is required
every other day at least one site in areas with nonattainment  probabilities
between 20 percent and 95 percent.  Sampling in areas with a probability
below 20 percent may be once every sixth day.  The use of the  term "area"
as it applies to the required sampling frequencies of the "area" is as
follows;  (1) any urbanized area as defined by the U.S. Bureau of Census,
(2) any incorporated place such as a city or town as defined by the U.S.
Bureau of Census or group of cities or towns, and (3) any "area" designated
by the responsible air pollution control agency.  In designating these
latter "areas," the control agency should consider technical factors  such
as the types of emissions, their spatial distribution, meterology, and
topography and how these factors contribute to the uniqueness  of the
"area" thereby distinguishing it from other designated "areas."  The

                                   3-4

-------
first year of PHio da*a collection is to start no later than  the  appropriate
dates in section 58.23 and section 58.34 for completion of the  SLAMS and
NAMS networks, respectively.  The year can start,  however, as soon as the
station is put into operation.
     The long-term strategy applies after the short-term strategy ends
(in most cases after the first full year of sampling  ends).   The  sampling
interval will then vary according to the relationship of air  quality levels
to the NAAQS.  The closer the air quality levels  to the NAAQS,  the more
frequently sampling must be carried out.  A table depicting the requirements
appears in section 58.13.
     As soon as the first exceedance of the 24-hour PMfp NAAQS  occurs in
an area which historically has had a less than everyday sampling  interval,
the requirements change to everyday sampling for  at least one site (the
site of expected maximum concentration).  The State will  have up  to 90 days
after the end of the quarter in which the exceedance  occurred to  implement
everyday sampling at the site of expected maximum concentration,  and it
must be maintained for at least 4 calendar quarters,  but preferrably 1
full calender year.  If the day on which the first exceedance occurred
was flagged as an exceptional event, the State must still implement everyday
sampling, unless exempted by the Regional Administrator.
     3.2.6  Sampler inlet siting
     The criteria in Appendix E of Part 58 for siting ambient PMjQ samplers
is very similar to the previous siting criteria for hi-vol samplers.  The
new criteria for the PM^o SLAMS include microscale stations which were not
previously included for the TSP SLAMS network. An inlet height of 2 to
7 meters is specified for PM^Q microscale stations.  The  inlet  height for all
                                   3-5

-------
other scales for PM-JQ stations is 2 to 15 meters as it was previously for
all TSP SLAMS.  The spacing requirement from roadways is also different.
For microscale roadway stations, the PMiQ sampler must be between 5 to 15
meters from major roadways.  For middle, neighborhood, and urban scale
PM-|Q stations a range of separation distances are specified for each as a
function of traffic volume and the scale of representativeness.
     3.2.7  Ambient data reporting
     The reporting requirements in Part 58 for ambient data apply for PM-JQ
as they do for the other pollutants; i.e., all SLAMS data must be reported
annually in summary form and all NAMS data must be reported quarterely.
Since the regulations allow for the use of TSP monitors as a surrogate
for PHio monitors under certain conditions, the reporting requirements for
TSP data remain in Appendix F of Part 58.  The TSP data summary format in
section 2.2 of Appendix F, however, has been changed to correspond somewhat
with the format added for PM]Q data in section 2.7 of Appendix F.  The
ranges for the 24-hour summary report have been in 50 ug/m3 increments.
The format added for PM-JQ is in 25 ug/nr* increments.  A significant change
made in the reporting requirements is the provision that annual TSP data
must be reported as arithmetic means, 24-hour TSP data values must be
reported if they exceed the level of the 24-hour PM]g NAAQS, and the
sampling interval must be reported. Otherwise TSP reporting requirements
remain unchanged.
     Section 2.7 has been added to Appendix F to cover reporting of PMjg
data.  All  exceedances of the 24-hour PM]g NAAQS, the sampling interval,
and the number of 24-hour values within 30 ug/m3 ranges must be reported.
                                   3-6

-------
     3.2.8 Air poj1utlon 1 ndex
     The air pollution sublndex for TSP has been replaced by a subindex
for PMiQ.  The breakpoints for the subindex reflect the levels of the
PMiQ short-term standard.  A new table of breakpoints and a graphic display
of the breakpoints is included in Appendix S of Part 58.
3.3 Ambient SamplersandAmbientDataUsage
     3.3.1 Ambient samplers inuse
     Considerable ambient data have been collected using hi-vols, hi-vols
with size-selective inlets with a DSQ of 15 micrometers (SSIis), and dichotomous
samplers with inlets designed for DSQ of 15 micrometers.  Some data have
been collected using samplers with inlets designed for a 050 of 10 micrometers.
The PM]Q samplers cannot be designated as PMiQ reference or equivalent  method
samplers, however, until Part 50 and Part 53 requirements are promulgated,
and subsequent manufacturer test data are examined and accepted by EPA.
Thus, the PM]Q data now available have not been collected by reference  or
equivalent methods.  Therefore, nonreference or equivalent method PMiQ
data, PMjg data, or TSP data may be used for initial SIP development
purposes, subject to certain constraints.
     3.3.2  Interim use of non-reference or egulyaj[ent method samplers
     Until PMiQ reference or equivalent method samplers can be designated,
manufactured, purchased, and installed in PMiQ SLAMS, States should continue
to operate approved ambient samplers with inlets designed for a nominal
DgQ of 10 micrometers.  Approval of ambient PM}Q samplers for SIP purposes
will be made by the Environmental Monitoring Systems Laboratory, MD-77,
Environmental Protection Agency, Research Triangle Park, North Carolina 27711.

                                   3-7

-------
Data from these approved samplers will be acceptable for use in  modeling
and in determining an area's attainment status.  Use of these data  for
initial SIP classifications are discussed in section 2.4.
     3.3,3  Episode moni tori ng
     The criteria for determining which areas are required to have  air
pollution episode contingency plans ire in Subpart H of Part 51. The
sampling procedures that should be used for determining air quality levels
during air pollution episodes are similar to those used prior to promulgation
of the PM]Q NAAQS.  These procedures are identified in section 4.0  of
Appendix C of Part 58 and are further described in the document, Guideline
for PHin Episode Monitoring Methods.^  Briefly, two methods based on the
filtration principle are recommended; staggered PM]o sampling and short-term
interval sampling.  The staggered sampling procedure uses a 24-hour sampling
procedure followed by a 2-hour post-sampling filter equilibration period.
The short-term interval sampling method requires a 4-hour sampling  period
followed by a 2-hour filter equilibration period.  In addition to these
two procedures, other methods may be used provided the user demonstrates a
site-specific correlation of the alternative method with the reference method.
3.4  SIP Revisions
     That portion of the SIP providing for ambient monitoring should be
revised, if necessary, to include provisions for a PMio ambient  network.
If the ambient monitoring portion of the SIP provides for monitoring
pollutants "for which NAAQS have been set," then no revision is  necessary.
If the SIP lists pollutants, PM-|g would need to be added to the  list.
The PM]Q SLAMS network design need not be included in the SIP and a SIP
revision is not necessary as a result of any network modification.   Any

                                   3-8

-------
SLAMS network modifications, however, must be approved by the appropriate
EPA Regional Office and any NAMS modifications by EPA's Office of Air
Quality Planning and Standards.
                                   3-9

-------
Reference

1.  Koch, R.C. and H.E. Rector, Network Design and Optimum Site  Exposure
    Criteria for Participate Matter, 6EOHET Technologies,  Inc.,  Rockvllle,
    WITI  Prepa red for U.S. En vi ronmenta 1 Protection Agency, Research
    Triangle Park, NC.  EPA Contract No.  68-02-3584.   March 1983.

2.  Pace, T.6. and N.H. Frank, Procedures for Estimating Probability  of
    Nonattalnment of a PHm NAAQS Using Total Suspended Participate or
         Data, U.S. Environmental Protection Agency,
    Research Triangle Park, NC.  EPA 450/4-86-017,  December 1986.

3.  Pelton, D.J., Guideline for PMm Episode Monitoring Methods, SEOMET
    Technologies, Inc.  Rockville, MO.  Prepared for U.S.  Enviromental
    Protection Agency, Research Triangle Park,  NC.   EPA Contract
    No. 68-02-3584.  February 1983.
                                   3-10

-------
                     4.0  AIR QUALITY MODELING FOR

4.1  Introduction
     Section 51.12 of 40 CFR requires that the adequacy of a control  strategy
for attainment and maintenance of NAAQS be demonstrated by means of a disper-
sion model or other procedure which 1s shown to be adequate and appropriate
for this purpose.  The Guideline on Air Quality Models (Revised)! provides
guidance on dispersion modeling for particulate matter and lists preferred
dispersion models for this purpose.  Procedures collectively known as receptor
models are also available that examine an ambient monitor sample of particu-
late matter and the conditions of its collection to infer the types or
relative mix of sources impacting on it during collection.  Receptor models
are described briefly in 'Appendix A and the references to this section.2,3
The most widely used and accepted quantitative receptor model is the chemical
mass balance (CMB),4  The proper use of the CMS 1s described in two protocols
referenced in this section.5,6
     Three options are presented herein for estimating the air quality impact
of emissions of PMjQ using dispersion and receptor models:  (1) use of receptor
and dispersion models in combination (preferred); (2) use of dispersion  models
alone; and (3) use of two receptor models, with control strategy developed
using a proportional model (discussed in Section 6.0).  This latter approach
is only encouraged if no applicable dispersion model  is available.  Several
considerations relevant to model selection are presented in Section 4,2
(Receptor Models) and 4.3 (Dispersion Models).  The PMio is generally referred to
in the following discussion.  Where there is insufficient technical  data to
support an analysis for PMjQ with these models, the same procedures may  be
applied to TSP, except as specifically noted, as a surrogate for PMiQ.

                                    4-1

-------
4.2  Considerations In Receptor Model Selection
     Several considerations related to the nature of sources emitting
influence selection of the receptor model(s) fop SIP purposes.   These are
the availability of particle size data and the size range of the emissions
from predominant sources, prior knowledge of the sources, chemical  simi-
larity of the sources, the need to identify individual sources  vis-a-vis
source categories, and the time scale of interest.  Different factors
affecting choice of receptor models are summarized in Table 4-1 and discussed
below.  One key overriding selection consideration is that CMB  is  considered
the most advanced of the receptor methods listed in Table 4-1.   The other
methods include factor analysis (FA), automatic scanning electron  microscopy
(ASEM), and microscopy (OM).  The FA» ASEM, and OH are not generally considered
quantitative and FA requires at least 40 samples to complete an analysis.   The
reader 1s referred to the Receptor Model Technical Series (references 2, 4, 5,
6, 7, 12, 13, 16) for technical, cost, and applicability information.
     4.2.1  Particle size
     Many researchers have discussed the blmodal distribution of particulate
matter which is related to the tendency of some source categories  to emit
predominantly fine (<2-3 urn) or coarse (>2-3 urn) aerodynamic diameter
particles.  Sources which emit predominantly fine particles include those
involving the combustion of fuels (motor vehicles, boilers, field  burning,
woodstoves, etc.) and industrial processes involving combustion, chemical
reaction, or condensation of vapors.  Contributors to coarse particles in
the atmosphere include windblown dust from storage piles, agricultural
fields, etc., vehicle resuspended road dust, pollens, and fugitive emissions
from industrial process sources.  Various receptor models such  as  species
mass balance and factor analysis are well suited to analyze sources of
                                    4-2

-------
                                        TABLE 4-1

                 SELECTING FEASIBLE SOURCE APPORTIONMENT METHODS BASED ON
                       DATA AVAILABILITY AND SOURCE CHARACTERISTICS
                          Sources  Sources
             Fine  Coarse  Known   Unknown
Cheml-  Finger-   Iso-
cal     prints    lated
Simi-   Dis-      Single
larity  similar   Source
iSS
lance  (CMB)    Y      Y

sctor
ifllysis  (FA)  .  X      X
 a.b.Y
 b,X
          Air Shed
Air Shed  (Specific
(Source   Sources
Gate-     Within
gori es)   Category)
            b,Y
itomated
.anning
ectron
croscopy
kSEM) c,X X X
croscopy
iM) c»X X X
spersion
>del (DM) YY YY YY


X b,X
X b,X
YY
                                                        YY
                     YY
  YY
                                                                                     b,X
YY
- Initial use of factor analysis may be helpful.

- Useful if a source can be isolated from other similar sources by wind direction.
  Method usually cannot otherwise distinguish between sources in same category.

- Useful for fine particles larger than 1.0 urn

- Appropriate to use 1n conjunction with CMB or DM.

- Appropriate to use with DM or FA, ASEM or OM.

- Appropriate to use.
                                           4-3

-------
either fine or coarse particles.  Optical and scanning electron microscopy
are suitable for coarse and fine particles down to about 1  micrometer 1n
size but work better on coarse particles.  Host receptor models generally
work best when the sample 1s segregated by size range (e.g., fine and coarse)
because these two size ranges are associated with different types of sources.
     4.2.2  Prior knowledge of sources and emissions
     In many instances, the sources suspected of contributing to ambient
PMjo concentrations at a particular site are apparent.  However, the relative
contribution of each source is needed.  Any of the methods  discussed in
Appendix A could be used to give source apportionment information if the
sources are identifiable, provided the other requirements for using the
method are also met.  However the mass balance requires knowledge of sources
and their emission characteristics.  If some sources are unknown, FA, ASEM,
or OH might prove useful if done prior to a CMB analysis.
     4.2.3  Chemical similarity
     The availability of "fingerprints" for the sources of  interest will
often determine the optimum receptor model to use.  A fingerprint is the
characteristic chemical or morphological pattern of the emissions from a
source that is used to distinguish it from other sources.1?  Some sources
have fairly distinct fingerprints, while others do not.  Since combustion
source emissions are predominantly composed of carbon, there is very little
information upon which to differentiate among the different types of combus-
tion sources.  Some help might be gained by examining the optical properties,
by using carbon dating (C14/Q12 ratios) which can distinguish between modern
or fossil carbon (e.g., wood smoke versus fuel  oil), or by  using minor
tracer constituents (e.g., K in wood).  Other common fingerprint problems
include the difficulty in distinguishing among sources whose emissions are
                                   4-4

-------
comprised of various soil components or between flyash and soil.  Optical
properties can be useful for some situations where fingerprints are similar.
Also, X-ray diffraction  (XRD) is very useful in identifying various minerals
by examining their crystalline structure,
     4.2.4  Sources ^ersus jource categories
     Any of the receptor models listed in Table 4-1 or Appendix A can be
useful in Identifying the impact of an isolated specific source unless its
fingerprint is similar to that of background.  Likewise, the techniques can
be used to Identify many of the source categories within an airshed consistent
with the limitations Identified in the preceding discussion.  However, the
impact of specific individual sources within an airshed containing multiple
sources of the same type may not be reliably identifiable except on wind
directional samples or by a dispersion model.
     4.2.1  ^artjcl_e__si ze '_and_'_tIjne jsca 1 e
    In some cases ambient PMjQ °p PMi§ data may be available for SIP
apportionment.  In other cases, only TSP data are available.  The availa-
bility of PMio data, or PMjg, or TSP will determine to some extent the type
of source apportionment method used for PMiQ since the PMiQ signature of a
source may be significantly different from the PMj5 or TSP signature.  Another
important factor is the time scale of the nonattainment situation (annual
or 24-hour).  Table 4-2 contains recommended approaches for source apportion-
ment based on the time scale (annual or 24-hour) of the nonattainment
problem and the data base available.  The choices listed in each block of
the table are in order of preference, with the preferred approach listed
first.  These approaches are discussed in more detail  in the following
sections.
                                    4-5

-------
                                 TABLE 4-2
            RECOMMENDED APPROACHES FOR PMio SOURCE APPORTIONMENT
                      AMBIENT DATA BASE AVAILABLE
                   m~
                                                TSP
          Applicable dispersion
          and receptor model
          Applicable dispersion
          model

          Receptor methods
          (at least 2) CMB w.
          corroborating method
Applicable dispersion
model corroborated by
ASEM or optical
microscopy**
Applicable dispersion
model
 * TSP may be used as a surrogate, where PMjQ data bases are Inadequate.

** Other receptor models such as Mass Balance may be used if fine
   particle data (generally less than 2-3 micrometers) are collected
   in addition to TSP.
                                    4-6

-------
4.3  Screening TechniquesandRefinedDispersionModelsfor	
     Concentrations
     It may sometimes be appropriate to conduct a preliminary screening
study to determine likely causes of nonattainment prior to a major control
strategy development effort.  This generally is confined to use of existing
ambient samples and emissions data and usually gives only a qualitative or
"first approximation" of sources.  Such a screening study has two advantages.
First, the qualitative prioritization of sources can be used to design the
more definitive study, which is usually required for control strategy
development.  This may enable substantial cost savings, since the screening
study can be focused on specific areas of concern.  Second, 1n some cases,
the source contribution may be very clear and the screening study may be
all that is required prior to control strategy development.  This is further
discussed in the references to this section.7
     4.3.1  Sel ecti on of appropriate^ dl spersl on models
     Several publications are available that contain hand calculation
methods for arriving at preliminary  (screening) estimates of PM-|Q concen-
trations. 8,9,10  in general, these methods are based on the assumption that
all particulate matter behaves as a gas in the atmosphere, i.e.» neglecting
settling and deposition.  Although subjective values for "half-life" or
pollutant decay have been used occasionally as a surrogate for particle
removal in screening methods and also some models, such procedures are not
generally recommended for PMjQ analysis.  For screening analyses, the con-
servative assumption of negligible removal is warranted, considering the
size of the particles.
                                    4-7

-------
     Dispersion models that can be used for estimating PMiQ concentrations
are listed 1n Table 4-3.  All  models are available on UNAMAP Version  6  from
National Technical Information Service (NTIS) as PB 86-222361.   Emission
inputs for those models not considered screening techniques are discussed
1n Section 5.0.  It should be noted that only the ISC model explicitly  treats
settling and deposition of particles and can accept particle size data, in  as
many as 11 size fractions.  No model recommended for regulatory use at  this
time handles secondary particle formation or other transformations In a
manner suitable for SIP control strategy demonstrations.
     However, EPA has completed work on two models, PEH-2 (an urban model)
and MESOPUFF-II (a medium scale transport model), that can provide supporting
analyses.  These models include provision to input a settling velocity
appropriate to particle size, a deposition velocity characteristic of the
pollutant-surface interaction, and a rate (percent per hour) to describe the
transformation of primary gas pollutant to secondary particle pollutant
(e.g., sulfur dioxide to sulfate).  Since their accuracy, suitability,  and
resource requirements for regulatory applications are not well  tested,  no
specific recommendation on the use of these models is provided  here.
Nevertheless, in selected applications, they may be useful in assessing the
significance of background concentrations and of secondary sulfate particles
in the effectiveness of control strategies.  Application of these models
should follow recommendation in the modeling guideline concerning the use
of alternative models.
     4.3.2  Special considerations for PMm dispersion modeling
     The Modeling Guideline! contains guidance that should be followed  on
(1) selection of appropriate source and meteorological data for use with
dispersion models, (2) location of receptor sites, (3) selection of model
                                   4-8

-------
                                 TABLE 4-3


               DISPERSION MODELS APPLICABLE TO PMiQ ANALYSES*




1 to .24-Hour Average          Annual Average        ScreenlngTechniques**


        CRSTER                    CRSTER                   PTPLU-2

        MPTER                      MPTER                   COMPLEX I

        RAM                        RAM                     VALLEY

        ISCST                      ISCLT

                                   COM 2.0
 *For more Information concerning the applicability of these models, consult
  the Guideline on A1r_Qua_11ty Hjode^s^Reyisedj^  As noted In this document,
  these models nay also be used Tor TSP mo3eTfng analyses in conjunction with
  a suitable TSP emission inventory, as a surrogate, where PHjo data bases
  are inadequate.

**These models are considered to be screening techniques for use prior to a
  more refined analysis as outlined in the Guideline on A1r Quality'Models
  (Revised).
                                     4-9

-------
options, (4) determination of urban/rural classification, and (5)  determination
of background air quality.  With regard to background air quality  values,
there 1s now limited data that might be used to support specific values  for
PMjo background concentrations.  General guidance on the use of available  data
1s provided in Appendix D.
     Rollback and roll forward are appropriate only for preliminary analyses.
Proportional models may be used 1n conjunction with receptor modeling  If
the air quality problem 1s clearly associated with a few specific  sources.
Such procedures are discussed in Section 4.4 and 6.4.  For urban-wide  refined
analyses, CDH 2.0 or RAM may be used.  For source-specific analyses of compli-
cated sources, the ISC model is preferred to CRSTER/ MPTER because ISC is  the
only model that is capable of treating deposition, area and volume sources,
building downwash, etc.
     For those cases where no recommended technique Is available or applicable,
nonguideline modeling approaches for use In each specific situation must be
approved by the appropriate Regional Office.
     Dispersion models are more reliable for estimating longer time-averaged
concentrations (e.g., annual average) than for estimating short-term concen-
trations (e.g., 24-hour) at specific locations.H  Point source models are
reasonably reliable in estimating the magnitude of the highest concentrations
occurring some time, somewhere within an area.  Errors in highest  estimated
concentrations of +_ 10 to 40 percent are found to be typical  for sources
that can be adequately characterized.  The multiple source urban model RAM
showed no significant bias in estimating 1-hour ground level  concentrations
for the 13-station RAPS network 1n St. Louis.  The average network cumulative
frequency distributions of hourly estimated and observed concentrations
differed by no more than 30 percent over the entire concentration  range.
                                    4-10

-------
However, estimates of concentrations that occur at a specific time and site
tend to be poorly correlated with actually observed concentrations and are
much less reliable, should this performance attribute be Important 1n a
regulatory application.
4,4  Receptor Models for Estimating PHm Concentrations*
     4.4,1  Control strategy analyses using receptor models
     Receptor models tend to be well suited for source apportionment of
24-hour PM^Q samples.  However, care must be taken to ensure that the samples
analyzed are representative of the conditions causing NAAQS exceedances.
The CHB is recommended as the primary method to be used in regulatory
applications of receptor models to PM-jg data.*»12»13  There is uncertainty
in any source apportionment approach.  Therefore, if CHB is used for source
apportionment (without combining with a DM), it Is required that at least
one other receptor modeling approach be used as a corroborating analysis.
This may be FA, OM, ASEM, microinventory, trajectory analysis, XRD, or
other corroborating approach as selected from those discussed in Volume I
of the Recepto r Mode 1 Technleal Series2 or the Digest of Ambient ParticuT_at_e__
Analysis and Assessment Methods.14  it is strongly urged that either optical
microscopy or ASEM be used to corroborate CMB, along with intensive chemical
analysis (sulfate, carbon and other elements) of the samples.  The OM or
ASEM should be used instead of CMB if only TSP data are available.  It 1s
also strongly urged that the CMB be performed on size fractionated PM^Q
(into fine and coarse fractions, below and above 2.5 urn).  This greatly
*The terms "model" and "method" are used interchangeably,  even though
 analysis methods such as scanning electron or optical  microscopy are
 methods, not models,
                                    4-11

-------
increases the resolution of the techniques.  The results of receptor model
source apportionment may be used with the method discussed 1n Section 6.4
to estimate the degree of control required to demonstrate attainment at a
given monitoring site.
     For exceedances of the annual NAAQS, the selection of days on which to
perform receptor model analyses should be governed by the representativeness
of the seasons where exceedances occur.  For exceedances of the 24-hour
NAAQS consideration, emphasis should be on those days where exceedances
occurred.  This selection process Is discussed further In reference.6
    4.4.2 Preliminary analyses using receptor models
     Preliminary analyses can also be done using receptor models.  The most
common methods employed are optical  or automatic scanning elecron microscopy.
These techniques are relatively inexpensive (less than $500 per sample) and
can give a variety of information about likely sources.  They are especially
useful when only TSP data are available because they can discriminate be-
tween particles less than and greater than 10 urn.  Other methods which may
be helpful include microinventories, chemical emission inventories, and mass
balance.4,12,13  in those cases where only TSP data on glass fiber filters
are available, the CMB approach is limited because some key tracers (silicon
and aluminum) cannot be used.  However, the CMB may be a useful preliminary
procedure in areas where no steel mills or coal fired power plants are
likely contributors.  In such instances, certain chemical elements associated
with emissions from steel mills and coal fired power plants (such as iron,
Fe) may serve as useful replacements for ordinary tracers of crustal material
(e.g., Si, A!) which cannot be reliably measured on a glass fiber filter.  In
areas where steel  mills and power plants are likely sources, the microscopic

                                    4-12

-------
methods (ON and ASEM) would usually be preferred for preliminary analyses.
As suggested in Table 4-2, these are the recommended, corroborative methods
for use with dispersion modeling when only TSP data are available.
4.5  Use of Receptor and Dispersion Models 1n Combination
     Several demonstrations have been made where receptor models were used
to help evaluate the results of dispersion modeling.15  fh1s 1$ the recommended
approach for source apportionment.  It is especially useful  when the emission
Inventory used in a dispersion model is determined to be marginally adequate.
The results of the receptor model can be used to carefully scrutinize the
inventory assumed in the dispersion model to deduce whether emissions from
certain source categories appear to have been adequately characterized.   The
use of a receptor model, such as CMB, in conjunction with dispersion modeling,
is highly recommended in such situations.  Procedures for using the CMB  and
dispersion model in concert are specified in reference6 to this section.
     Guidance 1n Section 5.0 describes the kinds of data necessary  in a  PMip
emissions inventory used as input to a dispersion model for a PM^o  analysis.
For determining compliance with the annual PMjQ NAAQS, dispersion models
based on an annual PMio emission inventory and sequential or frequency dis-
tributions of observed meteorological conditions can be used.  Also, receptor
models and methods such as CMB, FA, OH and ASEM, perhaps corroborated by
microinventories, trajectory analysis, and XRD can be used on samples which
have been carefully selected to represent the annual average.  Thus, the
source contributions would reflect those which cause the annual average  to
exceed the NAAQS.
     It may be difficult to devise a PM^o short-term emission inventory  for
use in a dispersion model to characterize 24-hour episodes of high  ambient
                                    4-13

-------
particulate matter.  Therefore, analysis of the observed monitoring data  using a
receptor model is likely to be particularly useful, in concert with dispersion
model estimates.  Of the receptor models discussed, CMB, OM, or ASEH,  perhaps
corroborated by XRD or trajectory analysis are appropriate for use with
24-hour observations, with CMB the preferred method.  Factor analysis  is
limited to long-term data sets and is more useful in conjunction with  the
annual NAAQS.16  it is recommended that TSP data not be used with receptor
models for either 24-hour or annual PM^o analyses because the particles
larger than 10 urn may bias the results.  An exception to this is the use  of
ASEM or optical microscopy in association with dispersion model estimates.
This exception is permitted because ASEM and optical microscopy involve the
analysis of discrete particles where their size (relative to PMio) can be
estimated.
     Once specific 24-hour or annual average source contribution factors
are obtained, the proportioning method discussed in Section 6,4 may be used
to estimate control requirements.
                                    4-14

-------
      References


 1.   Guideline On A1r Quality Models (Revised), EPA-450/2-78-027R.  U.S.
      fn"vTroninental ProtectlonTfg'encyV lesear'ch Triangle Park, NC 27711
      July 1986.

 2.   Rec eptor Model Technlcal__Ser les, Volurne I:  0very 1 efcj^ Of Recept or Model
      AfjgTicaTTorPfo PaurtTculate SourceT^^rtTonrneritT, EPA-450/4-81-Ol6a,
      PB82-139429.  U.S. Environmental Protection Agency, Research Triangle
      Park, NC, July 1981.

 3.   Gordon, G. E., Receptor Models.  Environmental Science and Technology,
      1980, 14, 792-800.

 4.   Receptor Model TechnicalSeries,Volume II;  Chemical Mass Balance,
      EPA-450/81-016b, PB82-187345.U.S. Environmental Protection Agency,
      Research Triangle Park, NC, July 1981.

 5.   Protocol For ApplyingAnd VaildatingTheCMB.  U.S. Environmental
      Protect1 on 'Agency, Research TManglePa rk,NC.  In Preparation,

 6.   Protocol For Reconciling Differences Among^ Receptor AndDispersion Models.
      U.S. Environmental Protection AgeTicy, Research Triangle Park, NC.  In
      Preparation.

 7.   Receptor Model TechnlcaJ Series, Volume V:  Source Apportionment
      Tecjuffgues^ntf CgnsTderat1 ons In Combl_nj"M_JMJr. Ms^ Epft 450/4-84-020,
      PM5-ill524.  U.S. Environmenta 1 ProtectTon Agency, Research Triangle
      Park, NC, July 1984.

 8.   Turner, D. B., Workbook of At^iospherlc Disgerslon Estiniates,  PHS
      Publication No. 999-AP-26.U.S. Environmental Protection Agency,
      Research Triangle Park, NC, 1970.

 9.   Budney, L. E., ProceduresFor EvaluatlngAir Quality Impact Of New
      Stationary Sources, Guide] 1 nejJF^rAlj'jQual 1 ty^ Wtntenance Planning And
      Analysis Volume IQRlEPA 450/4-77-001.DTS. Environmental Protection
      Agency, Research Triangle Park, NC, October 1977.

10.   Hanna, S. R., G. A. Briggs and R. P. Hosker, Jr., Handbook on Atmospheric
      Diffusion, DOE/TIC-11223.  Technical Information Center, Department of
      Energy, Washington, DC, 1982.

11.   Rhoads, R. G., "Accuracy of A1r Quality Models," Memorandum to A1r and
      Hazardous Division Directors, U.S. Environmental Protection Agency,
      Research Triangle Park, NC, July 22, 1981.
12.   Receptor ModelJechnleal Series, Volume III (Revised);  User^s Manual
      For SpecleFMass Balance jComputer Pro|rain.  O".  FhVfronnieWal
      Protection Agency, Research Tnangle Park, NC.  In Preparation.
                                    4-15

-------
13.   Receptor Model  Technical  Series,  Volume IV:   Technical  Considerations  In
      Source Apportlpnment By ParticTiTdentlficatlon.EPA 450/4-83-Olff;
      PB84-103340.U.S. Environmental  Protection  Agency, Research  Triangle
      Park, NC» June 1983.

14.   Digest Of Ambient PartIculateAnalysis And Assessment Methods_»
      EPA-450/ 3-78-113.U.S.  Environmental Protection Agency,  Research
      Triangle Park,  NC, September 1978.

15.   Core, J. E., J. A. Cooper,  P.  L.  Hanrahan, and W. M. Cox,  Particle
      Dispersion Model Eva!yatlon;  A New Approach Using Receptor jfgtfels.
      Journal of the A1r Pollution Control Association, 1982, 32,  1142-1147.

16.  Receptor Model Technical Series. Volume VI:  A Guide To The Use  of
     Factor Analysis  And Multiple Regress1o"n~[FA/NR) Techniques  1n  Source
     Apportionment, EPA 450/4-85-007, PB86-107638.U.S. Environmental
     Protection Agency, Research  Triangle Park, NC, July 1985.

17.  Receptor Hodel Source Composition  Library, EPA 450/4-85-002,  PB85-228823.
     U,T. tnvTronmentaT ProtectIon Agency, Research Triangle Park,  NC,  1985
                                     4-16

-------
                5.0  DEVELOPMENT OF PHiQ EMISSION INVENTORIES

5.1  Ovjerjrfew
     As part of the SIP development process following promulgation of
NAAQS, emission inventories may have to be developed for PMiQ.   This section
discusses various technical considerations involving the preparation of PMio
emission inventories.  Though inventories are useful for various purposes,
the primary focus of this section is on developing or reviewing PMiQ emission
inventories to be used as input to applicable air quality dispersion models.
Such models are expected to be utilized as part of SIP control  strategy
demonstrations.  Some discussion is also included on the emission inventory
requirements of receptor models, which differ considerably from dispersion
models in the approach used to identify source/receptor relationships as well
as in the inventory data needed as input.
     This section does not repeat basic guidance on the fundamentals of
compiling emission inventories.  It is assumed that most users  of this
document will have had direct experience compiling particulate  matter or other
inventories and will find the same general procedures applicable to PMiQ
inventories.  The primary thrust of this section is to alert users to
circumstances where additional materials and different methods  are needed to
compile PM|g inventories.
     The two major components of PMiQ inventories, or inventories of any
pollutant to be used in dispersion modeling, are (1) a set of emission factors,
and (2) source data to which the emission factors are applied to estimate
emissions.  Emission factors are generally compiled from source test data and
represent emissions, before and/or after controls, that may be  expected from
a typical facility within a particular source category.  The most commonly

                                     5-1

-------
used source of emission factors 1s AP-42,  Compilation  of Air Pollutant
Emission Factors,* which 1s revised periodically to Include particle  size
data and size specific emission factors.   Emission factors 1n AP-42 are
contained 1n two separate volumes for stationary point and areas  sources
(Volume I) and mobile sources (Volume It).   The PMio emission factors for
reentrained dust from roadways are included in Volume I.  This section
describes the format of these PMip emission factors for stationary sources
and how they can be applied to the appropriate source data to develop PMjo
inventories.  Emission inventory procedures for mobile sources are presented
in a separate EPA publication.il
     The second component of the PHjo inventory, source data, consists of
stack and exhaust flow parameters, production levels,  throughputs, control
device efficiencies, etc., as well as information on source locations and
plant layouts.  The specific kinds of source data that are needed in  the
PMjo inventory depend on the model used for PM^Q simulation, with the major
differences involving (1) the levels of temporal and spatial resolution,
(2) the description and configuration of each source,  and (3) the need for
chemical composition and particle size distributions.   Thus, this section
also identifies models currently available for PMio* as well as those
anticipated to be available in the next several years, and delineates the
specific differences in emission inventory requirements of each of these
models.
     Finally, a number of specific considerations need to be made when
dealing with air quality models and PMjo inventories.   First, condensable
and secondarily formed particles may become more important since  both are
primarily within the PMio fraction.  Second, since ?MIQ is a subset of total
                                     5-2

-------
particulate matter, the question arises whether an existing participate matter
inventory, compiled for TSP analyses, can simply be modified for use 1n a PMjp
application.  Third, as in any modeling application requiring large amounts
of input, data handling becomes an important consideration.  These topics are
addressed in the latter part of this section.
5.2  PMjQ Emission Factors and Fractional Multipliers
     Any emission factor is basically a multiplier which is applied to appropriate
source activity levels, such as throughput or production rates,  to estimate
emissions.  An emission factor can be a constant, implying a linear relationship
between source activity and emissions, or can be variable, with  parameters
other than just activity levels influencing emissions.  In either case,
emissions are generally computed by the following relationship:
                Emissions factor x Activity level x Control device
                  penetration = Emissions

     In the above equation, the control device penetration factor is calculated
as (1 minus the control efficiency), which becomes unity if uncontrolled
emissions are being calculated.  This simple equation is employed regardless
if emissions are being computed for total particulate matter, PMjQ, or afW
other pollutant.  If a set of PM^Q emission factors is available, the preceding
equation becomes:
              Emission factor|p^  x Activity level x Control  device
                penetrationlpM   * EmissionslpM
                              10               10

Of course, emission factors may be utilized that already reflect the effect
of certain controls, in which cases the control device penetration factor
drops out of the above equation.

                                     5-3

-------
     Alternatively, PMiQ emissions may be computed by applying fractional
multipliers to participate matter emission estimates 1n an existing inventory  as
follows:

             (PMio fraction) x Emissions|part^cu]ate = Emissions^

An advantage of the latter approach is that PMjo fractions may be available
1n certain cases whereas PMio emission factors may not.  This approach  may
also be easier from a data handling standpoint when converting an existing
total particulate matter inventory to a PMiQ inventory.
     Thus, in compiling a PMjo Inventory for use in dispersion modeling,
one must first obtain requisite PMig emission factor or size fraction
information.  Such information is becoming available for a large number of
sources.  Table 5.1 lists the stationary source categories for which EPA has
developed PMiQ emission factor and other particle size information.  A  majority
of these factors have been distributed to State/local agencies.  Additional
PMjo information will be published in supplements to AP-42,
     The PMio fractions and emission factors are presented in tabular form
such as are shown in Table 5-2.  In the hypothetical example shown in
Table 5-2, cumulative mass percents and emission factors are shown at various
particle size cut points for a controlled process.  In this case, PMio  emissions
after controls could be calculated either by applying a factor of 0.06  Ib/ton
to the process production rate or by multiplying the existing total particulate
matter emissions for the facility by 60 percent.  A graphical display of  the
information in Table 5-2 is included, in Figure 5-1.
     Some agencies may elect to obtain their own particle size data for
certain sources.  Source testing is generally encouraged, especially for
                                     5-4

-------
                                                          Table 5.1

                                   PM10 EMISSION FACTOES TO BE AVAILABLE FOR INDICATED SOURCE
                                         CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
1.1







1.2


1.3




1.4
1.6
Source Category
Combustion-Bituminous and
Subbit ominous Coal






Combustion-Anthracite Coal


Combustion-Residual Oil


Combustion-Distillate Oil

Combustion-Natural Gas
Combustion— Bark
Process

Dry Bottom Boiler
Wet Bottom Boiler
Cyclone Furnace
Spreader Stokerl
Spreader Stoker^
Overfeed Stoker
Underfeed Stoker
Dry Bottom Boiler
Traveling Grate
Boiler
Utility Boiler
Industrial Boiler
Commercial Boiler
Industrial Boiler
Commercial Boil er
All Boilers
Wood Waste Boiler
Uncontrolled

X
X
X
X

X
X
X

X
X
X
X
X
X
X
X
CONTROL DEVICE
ESP

X
X
X

X





X






Fabric
Filter

X


X



X









Scrubber

X

X







X





X
Multiple
Cyclone

X
X

X
X

X
X



X




*l,2
Other


















tn
en
   ^Without flyash  reinjection.
   2With flyash  reinjection.

-------
                                                       Table 5.1 (continued)

                                  PMjO EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE
                                          CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
1.7
1.8

2.1

4.2


5.3
5.15
5.16




Source Category
Combustion-Lignite Coal
Comb us t i on— B ag as s e

Refuse Incineration

Surface Coating


Carbon Black
Detergent Spray Dryer
Sodium Carbonate
Manufacturing



Process
Boilers
Vibrating Grate
Stoker
Municipal
Incinerator
Spray Booth Hater
Based

Oil Furnace


Bleacher Dryer
Calciner
Dryer
Rotary Predryer
Uncontrolled
X



X



X


X
X
X
X
CONTROL DEVICE-
IS?











X
X


Fabric
Filter















Scrubber


X






X


x3

X3
Multiple
Cyclone
X










X
x3

x3
Other





X
Hater
Curtain







in
en
    ^Combined controls.

-------
                                                       Table 5,1 (continued)


                                  PM10 EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE

                                          CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
5.17




5.
5.
6.1
6.3







Source Category
Sulfuric Acid




Boric Acid
Potassium Chloride
Alfalfa Dehydrating
Cotton Ginning







Process
Absorber
Absorber 20%
Absorber 321
Secondary
Absorber
Dryer
Dryer
Drum Dryer
Battery Condenser
Lint Cleaner
Roller Gin
Bale Press
Gin Stand
Saw Gin
Bale Press
Gin Stand
Uncontrolled
X
X
X

X
X
X
X
X


X
X
X
X
X
COIiTROL DEVICE
ESP
















Fabric
Filter





X










Scrubber






x3

X
X3






Multiple
Cyclone






X3


X3






Other
















U-T
I
    ^Combined  controls.

-------

                                                                          >
                                                       Table 5.1 (continued)



                                  PM1Q EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE

                                          CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
6.4





6.18
7.1





7.2





Source Category
Feed and Grain





Ammonium Sulfate
Primary Aluminum Production





Coke Production





Process
Carob Kibble
Roaster
Cereal Dryer
Unloading and
Conveying
Rice Dryer
Rotary Dryer
Bauxite Ore
Unloading
Storage
Prebake Roof
Monitor
HSS Cell
Coal Preheat
Coal Charging
Coke Pushing
Mobile Scrubber
Car - Travel
Mode
Uncontrolled

X
X

X
X
X




X
X
X
X
X



CONTROL DEVICE
ESP



















Fabric
Filter









X









Scrubber








X




X

X


X
Multiple
Cyclone


•
















Other



















01
I
00

-------
                                                       Table 5.1  (continued)

                                    PM10 EMISSION FACTORS TO BE AVAILABLE FOE INDICATED SOURCE
                                          CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
7.2


7.3





Source Category
Coke Production (cont.)


Primary Copper





Process
Mobile Scrubber
Car Push Mode
Quenching with
Dirty Water
Quenching with
Clean Water
Combustion Stack
Multiple Hearth
Roaster and
Reverberator/
Smelter
leverberatory
Smelter
Converter
Matte Tapping
Slag Tapping
Converter Slag
and Blow
Operations
Uncontrolled

X
X
X

X
X
X
X
X
X
CONTROL DEVICE
ESP









Fabric
Filter




X
X
X


Scrubber
X








Multiple
Cyclone









Other

Baffle
Baffle







01
IO

-------
                                                       Table 5.1 (continued)


                                  PM1Q EMISSION FACTORS TO BE AVAILABLE FOE INDICATED SOUECE

                                          CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
7.4




7.5

















Source Category
.Ferroalloy Open Furnace

Production


Iron and Steel Production

















Process
50% FeSl
80% FeMn
Si Metal
FeCr (HC)
SiMn
Sintering Wlndbox
Sintering Dis-
charge Breaker
Blast Furnace
Cast house
Blast Furnace with
Local Evacuation
Hot Metal
Desulf urlzation
10F Top Blown
Furnace Melting
and Refining
Q-BOP Melting and
Refining
BOF Charging
BOF Tapping
£AF Melting and
Refining
Uncontrolled
X
X
X
X
X
X



X

X

X





X
X

X
CONTROL DEVICE
ESP



X

X

















Fabric
Filter
X
X
X




X





X





X
X


Scrubber




X
X










X

X




Multiple
Cyclone





X

















Other























CJl
I

-------
                     Table 5.1 (continued)

PM10 EMISSION FACTORS TO BE AVAILABLE PCJR INDICATED SOURCE
        CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
7.5








7.6






7.8



Source Category
Iron and Steel Production
(cont.)







Primary Lead Smelting






Secondary Aluminum



Process

EAF Melting,
Refining, and
Charging
Direct Shell
Evacuation
Open Hearth
Furnace Melting
and Refining
Blast Furnace Flue
Blast Furnace
Ore Storage
Sinter Machine
Dross Kettle
Reverberat ory
Furnace
Reverberatory
Furnace
Chi or i nation
Station
Uncontrolled








X

X
X
X
X

X

X

X
CONTROL DEVICE
ESP








X











Fabric
Filter





X



X










Scrubber




















Multiple
Cyclone




















Other





















-------
                                                       Table 5.1 (continued)
                                       EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE

                                          CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
7.10




7.11


7.13

7.15





7.
8.1


Source Category
Gray Iron Foundries




Secondary Lead Smelting


Steel Foundries

Storage Battery Production





Tinner
Asphaltic Concrete


Process
Cupola Furnace
EAF
Pouring and
Cooling
Shakeout
Blast Furance Flue
Blast Furnace
Ventilation
Casting Shakeout
Open Hearth
Grid Casting
Grid Casting and
Paste Mixing
Lead Oxide Mill
Paste Mixing
Three Process
Batch Tinner
Conventional Plant

Drum Mix Plant
Uncontrolled
X
X

X
X


X
X
X
X

X

X
X
X
X

X
CONTROL DEVICE
ESP









X










Fabric
Filter
X




X

X





X
X


X

X
Scrubber
X



















Multiple
Cyclone

















X


Other

















Spray
Tower

en
i

-------
                                                        fable 5.1 (continued)

                                   PM10 EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE
                                           CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
8.3




8.6

8.9

8.13
Source Category
Brick and Related Clay
Products




Portland Cement

Coal Cleaning

Glass Manufacturing
Process
Coal Fired Tunnel
Kiln
Sawdust Fired Kiln
Raw Material
Screening and
Grinding
Wet Kiln
Dry Kiln
Clinker Cooler
Dry Process
Thermal Dryer
Thermal
Incinerator
Furnace Exhaust
Uncontrolled

X
X

X
X
X
X
X
X
X
CONTROL DEVICE
ESP





X



X
Fabric
Filter






X
X


Scrubber







X


Multiple
Cyclone






X



Other






Gravity
Filter



01
*->
oo

-------
                                                   Table 5.1 (continued)

                              PMjg EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE
                                      CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
8.15



.


8.18







8.22

8.
8.
8.





Source Category
Lime






Phosphate Rock Processing







Taconite Ore Processing

Feldspar
Fluorspar
Ughweight Aggregate (Clay)





Process
Rotary Kiln
Product Loading
Into Open Bed
Trucks
Into Tank Truck
Glass Line Into
Tank Truck
Ball Mill
Calciner
Rotary Dryer
Rotary Dryer and
Fluidized Bed
Dryer
Roller Mill and
Ball Mill
Main Waste Gas
Stream
Ball Mill
Rotary Drum Dryer
Coal Fired Rotary
Kiln
Dryer
Reciprocating
Grate Clinker
Cool er
Uncontrolled
X


X
X

X










X



X



CONTROL DEVICE
ESP
X











X3



X3








Fabric
Filter
xa













X3



X





X
Scrubber








x3
x3


X3







X




Multiple
Cyclone
X






X
x3
X


X3

X

X







X
Other




















s. e.b




aCyclone With Baghouse.
1>S. C. - Settling Chamber.

-------
                                                       Table 5.1  (continued)



                                  PM1Q EMISSION FACTORS TO BE AVAILABLE  FOR INDICATED SOURCE

                                          CATEGORIES,  PROCESSES AND CONTROL SYSTEMS
AP-42
Section
8.



8.




8.
10.1







Source Category
Lightweight Aggregate
(Shale)


Lightweight Aggregate
(Slate)



Talc
Kraft Pulp







Process

Reciprocating
Grate Clinker
Cooler
Coal Fired Rotary
Kiln
Reciprocating
Grate Clinker
Cooler
Pebble Mill
Recovery Boiler
with DCE
Recovery Boiler
without DCE
Lime Kiln
Smelt Dissolving
Tank Vent

Uncontrolled





x



X

X

X
X

X

CONTROL DEVICE
ESP











X

X
X



Fabric
Filter


















Scrubber





X








X

X

Multiple
Cyclone


















Other



s.e,a




S.C.a







Packed
Tower
in
i
   aS.C.  -  Settling  Chamber.

-------
                                                       Table 5.1  (continued)

                                  PM1Q EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE
                                          CATEGORIES,  PROCESSES AND CONTROL SYSTEMS
AP-42
Section
10.4
11.2.1









11.2.2
11.2.3
Source Category
Wood Working Haste
Unpaved Roads









Agricultural
Aggregate Handling
Process
Belt Sander
Rural
Gravel
Dirt
Crushed Lime
Stone
Industrial
Copper Smelting
Iron & Steel
Production
Sand & Gravel
Processing
Stone Quarrying
& Processing
Taconite Mining
& Processing
Western Surface
Coal Mining
Tilling
Batch Drop
Continuous Drop
Uncontrolled

X
X

X

X
X

X
X
X
X
X
X
X
CONTROL DEVICE
ESP













Fabric
Filter













Scrubber













Multiple
Cyclone
X












Other













Ul

-------
                                                      Table 5.1 (continued)


                                 PM10 EMISSION FACTORS TO BE AVAILABLE FOR INDICATED SOURCE

                                         CATEGORIES, PROCESSES AND CONTROL SYSTEMS
AP-42
Section
11.2.5





11.2.6




Source Category
Paved Urban Roads





Industrial Paved Roads




Process
Local Streets
Collector Streets
Major Streets/
Highways
Freeways/Express-
ways
Copper Smelters
Iron and Steel
Asphalt Batching
Concrete Batching
Sand and Gravel
Uncontrolled
X
X

X

X
X
X
X
X
X
CONTROL DEVICE
ESP











Fabric
Filter











Scrubber











Multiple
Cyclone











Other











01
I

-------
                                        TABLE 5-2,  CUMULATIVE EMISSION FACTORS AND PARTICLE SIZE
01
i
00
DISTRIBUTION FOR A HYPOTHETICAL

Particle Size
(micrometer)
Total Catch
15.0
10.0
7.5
5.0
2.5
1.0
0.5


Cumulative Mass Percent Less
than Stated Size
Uncontrolled
100
70
51
42
33
13
2.8
0.9
Controlled*
100
82
60
56
50
31
11
4
LIME PROCESS


Cumulative Emission
Factors (Pound/Ton of Product)
Uncontrolled
85
59
43
36
28
11
2.5
0.8
Controlled3
0.100
0.082
0.060
0.056
0.050
0.032
0.010
0.004
            aControl Device:  Baghouse,

-------
CJ1
I
I—I
iO
                 e
                 in
                 2  n
                 m  c=
                 	

                 1
                                              0.10
                                                    T>
                                                    O


                                                    §
                                                    C/5
                                                                                               0.05  o  -5
                                                                                                        m
                                                                                                        Q
                                              0.02
                                                                                               0.01
                                                                                               0.005
                                                    3

                                                    o
                                                    CD
                                           0.5
                                                                                               0.001
 1       2.S      5  7.5 10  15


ffiRODVHAMIC PARTICLE DIAftTER, MICROHETIRS
                             FIGURE  5-1  EXAMPLE GRAPHICAL DISPLAY  OF CUMULATIVE EMISSION
                                          FACTORS VS. PARTICLE SIZE  FOR A HYPOTHETICAL  PROCESS

-------
large contributors, since facility or point-specific emission estimates  are
considered more accurate than average estimates calculated from AP-42 factors.
Any testing should be consistent with the methods described 1n Appendix  C  of
this document.
5,3  PMio Source Data for DispersionHodels
     The second major component necessary for inventory compilation,  along
with emission factors or fractional multipliers, is source data describing
the nature and level of activity at each facility or operation.  Source  data
include all Information on the nature and location of each source,  operating
rates, stack and exhaust gas parameters, and control devices employed.  The
kinds of source data that must be obtained depend on whether a particular
source can best be described as a point, area, or line source.  The degree of
detail required in a PMjo emission inventory is dictated by the dispersion
model employed for simulating air quality impacts.
     As described in more detail in section 4.0, a number of models are
currently available for assessing the air quality impact of PMjQ, differing
in many respects.  Generically, most such models can be categorized along  the
following lines:
                           dispersion vs. receptor
                           individual source vs. areawide (grid)
                           short-term vs. long-term

Dispersion models calculate ambient air concentrations primarily as
functions of source configurations, emission strengths, terrain features,
and meteorological  conditions.  Receptor models (discussed in section 5,4)
infer the relative impact of various sources on ambient air quality by

                                     5-20

-------
reconciling particle size, and shape and chemical  composition  data  of ambient
air samples with particle size, shape and chemical  composition data for
various emission sources 1n the vicinity of the ambient air measurement  site.
Individual source models evaluate the Impact of a  single source or  source
complex whereas areawide (grid) models evaluate the Impact of  numerous sources,
including area and line sources, over a larger area.  Short term models
estimate air quality levels for time periods from  1 hour to 24 hours whereas
long-term models generally predict monthly, seasonal,  or anfiual average  concentra-
tions.  The major Impacts of model choice on requisite source  data  1n the
     emission inventories are Itemized below:
     -- Area covered by the Inventory - Individual  source dispersion models
        require source data to be collected for only a single  stack, facility,
        or complex.  Areawide models require source data for all  sources
        within the defined grid system.

     -- Source configuration - Some (especially point  source)  dispersion
        models require information on individual stack heights and  building
        sizes and locations.  Many dispersion models require area sources  to
        be assigned to grid squares and some distinguish line  sources from
        area sources.

     -- Temporal resolution - Typically, the emission  rates input to most
        applicable dispersion models are expressed  in  terms of grams per
        second for point sources, grams per second  per square  meter for  area
        sources, and grams per second per meter for line sources.   Actually,
        the emission rates input to dispersion models  need not be as resolved
        as might be inferred from these units.  Ideally, the emission rates

                                     5-21

-------
        should be as resolved as the model output.  FOP example. If the model
        predicts a 24-hour average ambient air concentration, emissions
        should be Input that reflect conditions over that 24-hour period for
        best results.  In cases where maximum concentrations need to be
        predicted, maximum emission rates may need to be considered instead
        of a time-averaged rate.

        Spatial resolution - Most dispersion models require point source
        coordinates for each stack.  Area and line source emissions are
        typically assigned to grid squares and line segments, respectively,
        having arbitrary dimensions.  The scale of the coordinate systems in
        most models is at the discretion of the user.  Because of computer
        constraints and regulatory requirements, tradeoffs usually exist between
        the need for finer resolution and the overall area that can be modeled.
        In some cases, the maximum degree of spatial resolution is limited by
        the units and number of significant figures built into the model's
        software.
     -- Size fractions - Most applicable dispersion models consider the
        fraction as a single lump sum entity requiring no particle size
        resolution below 10 micrometers.  Several models have the capability
        to treat various fractions differently within the PMio fraction,
     -- C hem1 ca 1 compos i t i on - Dispersion models rarely require chemical
        composition data, although many receptor models do.
Tables 5-3 through 5-7 list various dispersion models that may be used for
simulating PMjo levels, along with specific source data requirements of each

                                     5-22

-------
model.  References 2 through 7 present a detailed description of each of
these models.  The reader should consult these user's manuals as well  as
someone with modeling experience before developing an Inventory as Input to
any of these models.
     The reader will note from the Information 1n Tables 5-3 through 5-7
that dispersion models vary considerably in their source data requirements.
As was mentioned earlier, 1t is not the Intent of this section to reiterate
guidance on the fundamentals of compiling Inventories.  Rather, this see-?.
lion's primary purpose is to highlight special considerations that may have
to be made when developing a PMio emissions data base for modeling.  Basic
guidance on compiling emission inventories is given in References 8 through 14.
Additional guidance is presented in Reference 15 on obtaining different levels
of temporal and spatial resolution, including techniques for assigning area
source emissions to grid cells.
     Information on dispersion models used for PHjo applications 1s given in
section 4,0 of this document.
                                     5-23

-------
                                  TABLE 5-3

                    INDUSTRIAL SOURCE COMPLEX MODEL (ISC)2

Tyjge - Dispersion

Application - Estimating local Impact around complex industrial sources in
     rural or urban areas.  Stack, area, and volume sources within the indus
     trial complex can be accommodated.  Area sources can include fugitive
     sources such as storage piles and slag dumps.  Volume sources can
     include sources such as building roof monitors and conveyor belts and
     may also be used for simulating line sources.

Short/long term - The ISC contains both short and long term versions called
     ISCST and ISCLT, respectively.  The ISCST maybe run for each hour of a
     year or ISCLT may be run for a year using frequency distributions of
     meteorological data and annual emissions.

               - All sources within the boundaries of the industrial complex
    _^ _
     must be included in the inventory.

Stack and exhaust data - Physical stack height, inner diameter and elevation
     above sea level, gas exit temperature, and velocity.  If the stack is
     adjacent to a building and aerodynamic wake effects are to be considered,
     the length, width, and height of the building must be known.

Npnpplnt source configurations - The horizontal dimension, elevation above
     sea level and effective emission heights are required for each area
     source or volume source,

Temporal resolution of model output - 1-, 2-§ 3-, 4-, 6-, 8-, 12- and 24-hour
     values may be selected for calculation by ISCST and annual averages are
     calculated by averaging all 24-hour concentrations.  Seasonal and/or
     annual values are calculated by ISCLT.

Actual emi ssi ons uni ts i nput to ...model_ - grams/second for stack and volume
                                          sources,
                                        grams /second-meter2 for area sources.

Spatial resolution of source data  - Source elevation above mean sea level
     and source locations witli Tespect to a user-specified origin are required
     for all sources.  Universal Transverse Mercator coordinates may be used
     to define source locations.  A plant layout drawn to scale is required
     to obtain coordinates and building dimensions to the nearest meter.

Pj^rt|cTe si _z_e_resp_1 uti on - Required if settling and deposition are to be
     considered.  Deposition is generally only important for particles
     greater than 20 micrometers in diameter.  The PMig fraction could thus
     be treated as a lump sum without any need for particle size data
     although ISC allows input of up to 11 size fractions.

Chemical composition data - None required.

                                     5-24

-------
                                  TABLE 5-4

                  CLIHATOLOGICAL DISPERSION NODEL (CDM 2.0)3.4


Jyjge - Dispersion.

Application - Estimating urban scale impact of multiple point and area
     sources distributed over a square grid system.  Line sources are handled
     as area sources.

Short/long term - Long term (seasonal/annual)

Inventory area - All sources within the boundaries of the grid system must be
     inventoried.

Stack and exhaust data - Physical stack height and inside diameter, exhaust
     gas ex11~VeTocTiy, and temperature are needed to calculate plume rise.

Nonpolnt source conf 1juration - Area source emissions are allocated to square
     grid cells.  The coordinates of the southwest corner and the width of
     each grid are  required.  Stack height (if applicable) is also required.

Temporalresolution of the model output - Seasonal or annual average
     concentrations are predicted bythe model.

Actual emission units input ta model - grams/second.  Also required are
     ratios of average daytime and nighttime emission rates to the 24-hour
     average.

Spatial resolution of source data - The scale of the coordinate system for
     1ocatihg sources is completely arbitrary.  Area source grid squares can
     be defined arbitrarily small; however, since the program will only handle
     2500 grid squares, a tradeoff exists between area source resolution and
     model coverage.  Since CDM 2.0 operates on the implicit assumption that
     area source emissions are relatively uniform, grid square sizes should
     not be selected that are too small.  Grid squares of varying sizes are
     allowed, but their side length must be an integer multiple of a common
     side length.

Particle size resoljujbloji - No particle size resolution below 10 microns is
     required,   f h e PMIQ fraction can be treated as a single lump sum.

Chemical composition data - None required.

Other - An assumed pollutant half life, may be entered into CDM 2.0, but should
     not be considered in PMjp SIP's.
                                     5-25

-------
                                  TABLE 5-5

                                     RAM_5

Type - Dispersion

Application - Estimating urban scale impact of multiple point and area
     sources distributed over a square grid system.  Line sources are handled
     as area sources.

Short/long term - The RAM is a sequential model designed to process hourly
     inputs.  The RAM calculates concentrations for 1, 2, 3, 4, 6, 8, 12 and 24
     hourly averages.  The RAM may be run for each hour of a year or by averaging
     all 24-hour concentrations.

Inventory area - All sources within the boundaries of the grid system must be
     inventoried.

Stack and exhaustdata - Stack height and inside diameter, stack gas exit
     temperature, and velocity are needed for plume rise calculations.

Nonpoint source configurations - Area source emissions are allocated to
     square grid eelIs,  The coordinates of the southwest corner and the
     side length of each grid are required.  Height of emissions 1s also
     needed, if applicable.

Temporal resolution of model output - One hour to 1 day (24-hour) average
     concentrations are calculated by RAM, and annual averages are calculated
     by averaging all 24-hour concentrations.

Actual emission units input to model - grams/second for both point and area
     sources.

Spa11al resolut1 on of source data - The scale of the coordinate system is
     completely arbitrary, allowing any level of resolution.

Particle size resglut1 on - No particle size resolution below 10 microns Is
     required.  The PMjQ fraction can be treated as a single lump sum.

Chemical composition data - None required.

Other - An assumed pollutant half life may be entered into RAM, but should
     not be considered in PMjo SIP's.
                                     5-26

-------
                                  TABLE 5-6

                        SINGLE SOURCE (CRSTER) MODELS
Type - Dispersion.
Application - Estimating local impact of a single stack or up to 19 .sources
     considered to be emitting at a single point in rural or urban areas
     where terrain is below physical stack height.

Short/long-term - The basic estimate is for a 1-hour period; multiples up
     to 24 hours may be selected and the annual mean concentration is
     calculated from the average of all hourly concentrations.

Im/ent ory area - No area, per se» is included in the inventory, because the
     sTngTi~s~fack or collection of stacks are considered to be emitted at a
     common point.

Stack and exhaustdata_ - Physical stack height and inside diameter, exhaust
     gas exit velocity, and temperature are needed to calculate plume rise.

Nonpolnt source configurations - CRSTER does not handle nonpoint sources.

Temporal resolution of model output - The model calculates 1-hour, 3-hour,
     24-hour, and annual mean concentrations.  Optional averaging times of 2, 4,
     6, 8 or 12 hours may be selected.

Actual emission units input to model - grams/second.

Spatial resolution of source data - None; all sources are considered to emit
     from the same poinC

Particle size resolution - No particle size resolution below 10 microns is
~   required.  The PMjo fraction can be treated as a single lump sum,

C henrica1 compos i 11 on data - None required.

Othe_r_ - An assumed pollutant half life may be entered into CRSTER, but should
     not be considered in PM   SIP's.
                                     5-27

-------
                                  TABLE 5-7

                   MULTIPLE POINT SOURCE MODEL WITH TERRAIN

                             ADJUSTNEKT (MPTER)?


Type - Dispersion.

Application - Estimating local impact of a number of point sources in rural
     or urban areas where terrain is below physical stack height.

Short/long term - Short-term (1-24 hours) and annual calculations can be made.

Inventory area - That area including all of the point sources (up to 250)
     that are being modeled.

Stack and exhaust data - For each stack, physical stack height and inside
     diameter, stack gas temperature and exit velocity, and stack ground
     level elevation are needed to calculate plume rise.

Nonpgint source configuration - HPTER does not handle nonpoint sources.

Temporal resolution of model output - One hour or any multiple thereof,
     including hour-by-hour concentrations for an annual average.

Actual emission units input to model - grams/second for each point source.

S pat i al^ r es o lu t i on of s qu rce data - The scale of the point source coordinate
     system is arbitrary, Tl'Towing any level of resolution.

Parti c 1 e si ze resol utl on - No particle size resolution below 10 microns  is
     required.  The PMiQ fraction can be treated as a single lump sum.
Cheml cal compos 1t1 on data - None required.

Other - An assumed pollutant half life may be entered, but should not be
     considered in PM   SIP's.
                                     5-28

-------
5.4  Inventory DataNeeded 1nReceptor Models
     The considerations In sections 5.2 and 5.3 apply primarily to
inventories developed for use in dispersion models.  Different source and
emissions data are required in receptor models.  As discussed in section 4.0,
receptor models use techniques to estimate the contributions of emission
sources based on specific characteristics of particulate matter measured at  various
receptor sites.  Receptor models should be used together with rollback procedures
or dispersion models to define source/receptor relationships.
     Because of the nature of receptor models, emission factors and mass
emission estimates are not needed as input as they are for dispersion models.
Nor, generally, are detailed source data required such as activity levels and
control techniques.*  The only source emission data that are needed for
direct use in receptor models involve similar kinds of morphological or
chemical characterizations of emissions that are made on the ambient samples
collected at the receptor site(s).  Moreover, such emission characterizations
are usually not obtained for each individual source within an area being
modeled, but typically are representative of broad classes of sources whose
particle compositions are similar (e.g., motor vehicles, oil fired boilers,
etc.).
     The major type of receptor model utilizes various chemical methods.16,17
In general, chemical methods identify source contributions by comparing the
relative amounts and temporal variability of various chemical species at
ambient sampling sites with source chemical patterns (i.e., "fingerprints").
    *Note that while detailed source and emissions  data on individual  facilities
are not required to run receptor models, these kinds  of data are needed  in
interpreting the results and in defining specific source/receptor relationships.
                                     5-29

-------
Chemical methods are comprised of mass balance techniques, which Identify the
most probable combinations of sources to explain the chemical  pattern on  a
single filter, and multl sample methods (Including factor analysts),  which
Identify the most probable linear combination of sources to explain either
the time or spatial variability in ambient chemical patterns.
     Microscopy represents another Important type of receptor  method  typically
used to confirm the results obtained from chemical methods. Optical  microscopy
relies on identification of particles collected at the receptor site  by their
size, shape, color, surface properties and birefringence (an optical  property).
Scanning electron microscopy, commonly automated by computer and supplemented
with x-ray fluorescence analysis, relies on a particle-by-particle analysis of
particle size, shape, and elemental composition.  The use of such particle
Identification techniques requires a knowledge of the physical characteristics
(e.g. shape, size, brightness, color) of particles from various source categories
or access to references that catalog particle size characteristics from various
sources.16,18,19
     Since nearly all receptor models require a knowledge of the chemistry
or morphology of source emissions, the development of a collection of emis-
sion chemistry data or particle reference samples (or both) Is often  required.
Such data may be compiled from the literature (e.g., previous  receptor model
applications); however, experience has demonstrated the value  of developing
locale-specific emission characterizations, especially for more important
source categories.
     The EPA has compiled a library of chemical and elemental  composition data
that will  provide a common basis for receptor model source characterizations.20
An example of the types of data In this library 1s shown In Table 5-8,

                                     5-30

-------
                                      TABLE 5-  8
            EXAMPLE OF DATA IN  ERA'S PARTICIPATE  SOURCE LIBRARY  (20)
SUH(
62.200
60.750
SOURCE:
s:ci
CONTROLS:

SPECIES
KUKsER
£
5
9
11
12
13
u
1$
16
17
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
37
33
40
47
ts
50
51
55
56
58
80
82
201
2C2
203
204

S?£CJES
NA,"
BE
B
F
HA
HG
AL
SI
P
S
CL
It
CA
SC
Tl
V
CR
HN
Fl
CO
Ml
C"J
2N
GA
GE
AS
Si
BR
R8
SR
ZR
AC
CO
SK
S3
CS
SA
CS
HG
PS
oc
EC
SC4
N03
STEEL PRODUCTION • OPEN HEARTH FURNANCE
3-03-009-01
ESP

******•*»*
X IY VT
NA
NA
NA
NA
NR
NR
NR
<
13.300
<
5.000
0.550
<
<
0.550
2 ."000
0.550
11.000
<
0.550
0.550
<
<
<
0.050
<
0.050
O.OSO
«
<
0.050
0.050
0.050
0.050
<
<
<
<
0.550
NA
HA
40.000
0,550
FINE
<2.5 UK
+ •
4 "
4- -
+ •
4- «
4- -
* -
4- •
4- •
4 -
+ .
* .
* .
* -
4- -
* .
4 -
* .
* *
* -
+ -
* -
4 •
4 -
4- .
+ •
4 •
4 *
* •
4 •
4 -
4 -
4 "
V .
4 »
4 .
+ .
* •
+ "
4 .
4 .
+ -
* -
4 .

•**•»*
UNC
NR
NR
KR
NR
NR
NR
NR
NR
NR
NR
NR
HR
KR
NR
HR
HR
NR
NR
NR
HR
KR
NR
NR
HR
HR
HR
HR
NR
HR
NR
NR '
HR
HR
NR
HR
HR
NR
HR
NR
NR
HR
HR
HR

*»»**#**«
X BY WT
NA
HA
HA
NA
NR
NR
NR
<
3.000
<
5.000
0.550
<
<
0.550
0.550
0.550
16.000
<
0.550
0.550
, <
<
<
0.050
<
O.OSO
0.050
•e
< ,
0.050
0.050
O.OSO
0.050
<
<
<
<
0.550
NA
NA
35.000
0.550
COARSE
2.5-10UK
*-
4. .
4 •
4- .
•4 *
4, .
4 -
4 .
4- *
4 >
4- .
. » .
4 -
4 .
4 -
+ .
* .
4- -
4 -
4 «
4 •
4 •
4 .
* .
* -
4 *
* .
4 •
4 «
4 "
4 •
4 •
4- •
4- .
4 *
4 •
4 "
4 .
4. .
* .
4 •
* .
4 •
4. .

*»«*««
UNC
HR
NR
NR
HR
NR
NR
HR
NR
NR
NR
NR
NR
NR
NR
HR
NR
NR
NR
NR
HR
NR
NR
NR
NR
HR
HR
NR
NR
NR
NR
NR
HR
HR
NR
NR
NR
HR
HR
HR
HR
HR
HR
HR
PROFILE:
RANKING:
RATING :

W#w «*»***
X BY Iff
NA
HA
NA
NA
NR
NR
HR
<
13.300

5.000
0.550

<
0.550
2.000
0.550
11.000
%
0.550
0.550

<
<
O.OSO
f
0.050 "
0.050

«
0.050
0.050
0.050
0.050
<
<
<
<
0.550
NA
NA
40.000
0.550
2S302
2121
0
TSf»
<30 UK
4"
* .
4- .
* .
4 •
* .
4, .
4- .
* •
* .
* .
4 •
+ .
* -
+ .
* .
4 •
« .
4 »
4- •
4 -
+ •
* .
4> .
+ -
4 .
* .
4, .
4- -
4 .
4- •
+ .
4 -
4 «
4 -
4 •
•» •
* "
4 >
4 -
4 .
4 .
4 •
* .


*•»••»«
UKC
KR
KR
KR
KR

KR
Nil

K?
KR

H
k.R
KR
KR
kR
h?
KR
KR
Is?
m
^
K9.

.NR
K?
kg

hi

KR

NR
KR
NR
N*
KR
HR.

VX

kS

                                                                           62.200
NOTES!   (X * ORGANIC CARBON :   EC « ELEMENTAL CARBON ;  HA - NOT ANALT2EO :  HR * HOT REPORTED
        < * LESS THAN DETECTION tIKIT
        CTHER HOTES :      SOX CARSON,  22-26X OTHER (SI, OZ, AL,  HG5 TSP AND COARSE. REF, 20
                                           5-31

-------
These data are most useful 1n the chemical methods such as the mass balance
model and factor analysis described 1n Appendix A of this guideline.   For
more Information on this participate matter source composition library,  contact the
U.S. Environmental Protection Agency, Air Management Technology Branch,  HD-15,
Research Triangle Park, North Carolina, 27711.  The appropriate references
cited in section 4.0 and Appendix A may be consulted for more details  on the
specific characterizations needed for each type of receptor model.
5.5  Condensable Particulate Hatter
     Condensable particulate matter (or condensed particulate matter,  as It
is synonymously described) can be broadly defined as material that  is  not
particulate matter at stack conditions but which condenses and/or reacts (upon
cooling and dilution 1n the ambient air) to form particulate matter immediately
after discharge from the stack.  Condensable particulate matter is  of  potential
importance because it usually is quite fine and thus falls primarily within
the PM10 fraction.21  Sources suspected of emitting significant amounts  of
condensable material are shown in Table 5-9.
     Condensable particle factors, as such, are not explicitly included  In
AP-42 for most source categories.  To the extent that EPA's Method  5 (Refer-
ence 22) captures a portion of the condensable fraction, AP-42 particulate
matter factors for certain categories that are currently based on Nethod 5  include
some condensable particles.  Some condensable particulate matter will  also  be
collected by the testing procedures outlined in Appendix C of this  guideline.
5.6  Secondary Particulate Matter
     Secondary particulate matter can be broadly defined as particles  that  form
through chemical reactions in the ambient air well after dilution and  con-
densation have occurred.  An example of this phenomenon is the formation of

                                     5-32

-------
                                 TABLE 5-9
           POTENTIAL SOURCES OF CONDENSABLE PARTICIPATE MATTER21
Stationary Sources

Alfalfa Dryers

Anode Biking Furnaces

Asphalt Plants

Asphalt Roofing
    felt saturating
    asphalt blowing

Boilers/Other Combustion
    bagasse
    coal
    lignite
    oil
    wood/bark

Charcoal Kilns

Chemical Production
    boric acid
    phosphoric acid
    potassium still
    zinc sulfate

Citrus Peel Dryers

Coke Plants

Corn Processing
    wet milling
    syrup manufacturing

Elemental Phosphorus
    electric arc furnace

Expanded Vinyl

Ferroalloy Mills

Fertilizer Plants
    ammonium nitrate
    diammonium phosphate
Fiberboard
     dryer
     press

Glass Fiber and Mineral  Hool
     curing ovens
     blow chambers

Glass Plants

Grain Dryers ,

Iron and Steel  Hills
     sinter plant
     electric arc furnace
     basic oxygen furnace
     open hearth furnace
     heat treating
     scrap steel melting
Gray Iron Foundries

Kraft Pulp and Paper Mills
     recovery boilers
     lime kiln
     smelt dissolve tank
     blow tank/hot water
       accumulator

L1me Kiln

Manure Dryers

Mineral Products    |
     gypsum
     clay dryers
     feldspar dryers
     clay kilns

Municipal Incinerators
                                     i-33

-------
                            TABLE 5-9 (CONTINUED)
Petroleum Refineries
    FCC
    catalytic regenerator
    heaters
    petroleum coke

Portland Cement Plants
    kiln
    finish mill

Primary Nonferrous Smelters
    Cu converter
    Cu electric furnace
    Cu fluid bed roaster
    Cu sinter line
    Pb sinter line
    Pb blast furnace
    No roaster
    Zn ore briquet dryer
    Zn sweat kiln
    Zn fume kiln

Residential Heating
    oil
    wood

Secondary Metal Smelters
    Al scrap furnace
    Al dross furnace
    brass and bronze furnaces
    copper furnace
    Pb furnaces
    PbOg mills
    Pb grid casting
    Pb remelt pot
    other metal furnaces

Mobile Sources

All Mobile Sources
Sewage Sludge Incinerator

Silicon Carbide Furnaces

Spray Paint Booths

Sulflte Pulp Mill
     recovery boiler
     blow tanks

Rubber Incineration

Rubber Curing Press

Textile
     nylon polymerization
     melt polymer spinning
     tenter frame
     dye beck
     heat set
     texturizing
     latex backing

Tire Buffing Operations

Wood Products
     veneer plant dryer
     resawing
                                     5-34

-------
sulfate particles in a plume from the oxidation of sulfur dioxide by  one of
several atmospheric transformation mechanisms.  Generally, secondary  particulate
matter can be distinguished from condensable particul ate matter by the time
and/or distance downwind from the stack required for formation.  Condensable
particles form in a matter of seconds in the stack exhaust due primarily to
immediate cooling and air dilution whereas secondary transformation requires
minutes, hours, or even days.
     Unlike condensable particulate matter, secondary particulate matter should
not be included in the PMio inventory as if directly emitted as particles.
Condensable particulate matter, because it forms so quickly, will likely impact
on any nearby receptor, and thus can be treated as if it were emitted as particulate
matter.  Secondary particle formation, conversely, is  a function of time
and distance downwind from the source, as well as chemical composition and
reactivity.  Since secondary particle formation is an  atmospheric phenomenon,
it should be simulated by an air quality model if it is considered to be an
important component of ambient PMio concentrations.  Precursors of secondary
particles especially SOX and NOX, need to be included  in the Inventory when  such
models are used.
5.7  Use of Existing EmissionL Liventory
     An important consideration when planning a PMio inventory is whether an
existing particulate matter Inventory can be used as a foundation or starting point.
Since many areas have compiled annual, countywide inventories of particulate
matter (e.g., for previous SIPs or for submission into EPA's NEDS or Compliance
Data System), an incentive exists for building on such inventories rather than
compiling new ones altogether.   If an existing inventory is comprehensive,
current, and accurate, much of the source data needed  in the PMjQ inventory  is
already available.
                                      5-35

-------
     The utility of the existing inventory depends on the particular air
quality model employed.  In some Instances, a conventional  Inventory of
annual, countywide particulate matter emissions will  provide most of the needed
Information.  Actually, the terms "annual" and "countywide" do not give a
complete picture of the resolution commonly inherent  in this kind of inven-
tory, at least for point sources.  First, while the emissions are typically
expressed in units of tons per year, the available operating pattern
Information for each point source generally allows the user to extrapolate
emissions for a given season, workday, and even for a given hour during a
workday.  Details on performing this kind of temporal apportionment are given
in Chapter 6 of Reference 18.
     Second, stack data are available for point sources in the annual,
countywide inventory that are sufficient for many PMjQ modeling applications.
Specifically present are stack heights, diameters, flow rates, and plume
heights, which are the major stack and exhaust gas parameters commonly  needed
by most applicable dispersion models.  (Note:  Some models also require exit
gas velocities, but flow rates can be easily converted to exit velocities by
dividing by the cross sectional area of the stack.)  Moreover, the location
of each stack is commonly specified to the nearest tenth kilometer, providing
acceptable spatial resolution for applications where  the grid system will be
comprised of relatively large grid cells.
     Thus, relating back to the inventory specifications for each model
outlined in Tables 5-3 through 5-7, 1t can be seen that, for point sources at
least, conventional inventories of particulate matter may provide sufficient temporal
and spatial resolution and stack data for a number of dispersion models such
as COM 2.0, RAH, CRSTER, and MPTER.
                                     5-36

-------
     The area sourct data available from conventional  (I.e.,  annual,
countywide) particulate matter inventories can also be useful  In some PMip  modeling
applications; however, additional data manipulation is generally needed  to
develop the requisite temporal and spatial resolution.  First, most existing
area source records do not include operating rate data comparable to  those
Included in the point source record.  Hence, to develop emissions estimates
for a period of time less than a year, the user must impose an operating
pattern for each source category.  Detailed operating  patterns for many
important area source categories are suggested in Chapter 6 of Reference 18
along with recommendations for employing them.
     Probably the most difficult aspect of using conventional  area source
data in a PM^Q inventory is that for certain applications, the emission
totals must be allocated from the county level to the  grid cell level.   This
is generally done by assuming that the distribution of a given area source
activity behaves similarly to some surrogate indicator (e.g.,  population)
whose distribution is known at the subcounty level.  For example, emissions
from residential fuel combustion might logically be apportioned according to
the distribution of dwelling units.  More specifics on gridding countywide
emissions are given in Chapter 6 of Reference 18,
     Referring to the inventory specifications in Tables 5-3  through  5-7,
the only models which could use the typical existing area source inventory  as
primary input are COM 2.0 and RAM.  The other models either do not handle area
sources or else require more detailed information.
     Several models require such specific inventory data on individual sources
or source complexes that an existing inventory may be  of little use at all.
The ISC, for example, requires input on the dimensions and juxtaposition
                                     5-37

-------
of buildings near each stack in order to consider the aerodynamic wake effects
on plumes.  The ISC also requires finer spatial resolution than is typically
available.  Receptor models are probably most unique in terms of their input
requirements.  Generally, the location of sources, stack and exhaust  data,
and even the rates of emissions are not required by receptor models.   As  an
example, the only inventory data needed by the Mass Balance (MB) model are
the chemical or elemental composition of emissions for each source category
as a function of particle size (i.e., in the <2,5 micrometer and 2.2.5 to  10
micrometer size ranges).  Particle size, shape, and composition data  will not
be contained in most inventories that have been developed for dispersion
modeling applications.
     In any case, whenever an existing particulate matter inventory is being used
to directly generate the PM^p inventory, total particulate matter emissions have
to be converted to PMio emissions.  As was discussed in section 5.2,  the  easiest
way to accomplish this is to multiply existing particulate matter estimates
by appropriate PMio fractional multipliers.  For maximum accuracy, PM^g
fractions should be applied to individual sources rather than aggregated
source categories and should account for the existence of any emission controls.
Alternatively, instead of applying PMjg fractions, PMjp emission totals can
be calculated by applying the appropriate PM^g emission factor to the existing
activity level for each source.
5.8  Data Handling
     An important consideration in any inventory compilation effort is the
potential need to develop special data handling software.  In a number of
models, data handling can be readily accomplished manually because the input
                                     5-38

-------
requirements are meager.  For certain models, however,  computerized data
handling capabilities may be desirable to compile and properly format  the
large quantity of Inventory data needed to drive the models.
     As a general rule, the amount of Inventory data that must be handled  Is
directly proportional to the number of sources being modeled.   Hence,  for  the
single source models such as CRSTER, manually preparing the Input
data In the necessary formats should present no problems.  The same will be
true for models like ISC and HPTER if only a few sources are considered.
     The primary applications where computerized data handling will prove
essential are {1} when dealing with hundreds of point sources; (2) where area
source emissions must be subcounty allocated to grid cells; (3) when numerous
line sources must be considered individually; and (4) where an existing par-
ticulate matter inventory is being converted to a PMip inventory, necessitating
a number of preprocessing computations.
     Computerized data handling is desirable when preparing PMjo inventories
for input to COM 2.0, RAM and MB, as well as MPTER if large numbers of
sources are considered.  The COM 2.0 and RAM are most likely to be applied to
large, urban scale applications involving many point and area  sources  and
large grid systems.  Moreover, existing particulate matter Inventories are
most likely to be utilized as Input for these two models, necessitating some
preprocessing to convert total particulate matter to PMiQ.   Hence, 1f  either
COM 2.0 or RAM is to be employed for PMio modeling, and if an  existing annual,
countywlde inventory will be used to provide the necessary input, sufficient
resources should be reserved for developing and running an auxiliary data
handling system that will facilitate timely processing of the  input data.
Data handling is frequently an overlooked aspect of inventory  compilation  and

                                     5-39

-------
air quality simulation modeling that can consume a great  deal  of  resources.
Thus, the inventory specialist, modeler, and programmer should all be involved
in the planning of this kind of PMig inventory effort.
                                     5-40

-------
References


1.   COTijvM atl on of Air Pollutant Emission Factors, Fourth Edition,
     Volumes I and II, U. S. Environmental Protection Agency, Research
     Triangle Park, North Carolina, September 1985 and Supplements.
     (Supplement A was issued in early 1987).

2.   Environmental Protection Agency, Industrial Source Complex (ISC) Dispersion
     Model. Second Edition, Volumes I and II, EPA 450/4-86-005a,b, U.S.
     Environmental Protection Agency, Research Triangle Park, North Carolina,
     June  1986.

3.   A. D. Busse and J. R. Zimmerman, jJser's Guide _for_ the CllmatQloqical
     Dispersion Model, EPA-R4-73-024, U.S. Environmental Protection Agency,
     Research Triangle Park, North Carolina, December 1973.

4.   Irwln, J. S., T. Chico and J. Catalano, COM 2.0 - CljmatologicaT Dispersion
     Model, EPA/600/8-029, U.S. Environmental Protection Agency,Research
     Triangle Park, North Carolina, November 1985.

5.   D. B. Turner and J. H. Novak, User's Guide for RAH, Volume I. Algorithm
     Description and Use, EPA-600/8-78-016a, U.S. Environmental Protection
     Agency, Research Triangle Park, North Caorolina, November 1978.

6.   Users Manual for Single-Source (CRSTER) Model, EPA-450/2-77-Q13, U.S.
     Environmental Protection Agency, Research Triangle Park, North Carolina,
     July  1977.

7.   T. E. Pierce and D. B. Turner, Users guide for MPTER, A Multiple Point
     Gaussian pisgerslon Al9g_rU_hm_w1th_ Qgtional Terrain Adjustment,
     EPA-6UO/8-80-016, U.S. EnvironmentalProtection Agency, Research
     Triangle Park, North Carolina, April 1980.

 8.  Procedures for Emissi_ori\ 1 nventoryPreparation, Volume I;  Emiss ion
     rni ventjpry FundamenTfaU, EPA-45dyl-8'l-OWa, U.S. Envi ronmental
     Protection Agency, Research Triangle Park, North Carolina,
     September 1981.

 9.  Pj^qceduresfor_ Em1_ss_1 OJL I nventpr/J'repa ration, Vojume II;  Point Sources,
     EPA--4i50/4-8i-026b, U.S. Environmental Protection Agency, Research
     Triangle Park, North Carolina, September 1981.

10.  Procedures for Em1_sj5lon_Jnventqrjj^Preparation, Vol ume III:  Area Sources
     EPA-45Q/4-81-Q26C, U.S. Environmental Protection Agency, Research~
     Triangle Park, North Carolina, September 1981.

11.  Procedures for Emission Inventory Preparation,Volume IV:  Mobile
     Sources7 EPA-450/4-80-Q26d,U.S. Environmental Protection Agency,
     Research Triangle Park, North Carolina, September 1981.

                                     5-41

-------
12.  Procedures for Emission Inventory Preparation, Volume V^  Bibl1ography,
     EPA-450/4-80-026e, U.S. Environmental Protection Agency, Researcfi
     Triangle Park, North Carolina, September 1981.
13.  W. H. Umason and T. F. Lahre, Procedures forjthe Preparation of
      ..                       	5E	._	
     Edition, EPA-450/2-77-028, U. S. Environmental Protection Agency,
                                           	  repc
Emission Inventories for Volatile Organic Compounds, volume t,  Second
Edition, EPA-45d/2-7?-028»U. S. Environmenl
Research Triangle Park, NC, September 1980,
14,  Example Emission Inventory Documentation For 1982 Ozone State
     TipTementation Plans (SIP's), EPA-45Q/80-033, U.S. Environmental
     Protection Agency Research Triangle Park, North Carolina, March 1981.

15.  Procedures for the Preparation of Emission Inventories for Volatile
     Organic _Comp pun ds Volume II:  Emission Inventory Requirements for"
     Photochemical Air Quality Simulation Models. EPA-450/4-79-018, U.S.
     Environmental Protection Agency, Research Triangle Park, NC,
     September 1979.

16.  J. E. Core, Receptor Model Technical Series, Volume I, Overview of
     Receptor Moder Application to Parti cul ate Source
     EPA^450/4-81-01w, U.S. Environmental Protection Agency, Research
     Triangle Park, North Carolina, July 1981.

17.  J. E. Core, Receptor Model Technical Series, Volume II, Chemical  Mass
     Balance, EPA-45Q/4-81-016b , U.S. En vTroninental Protect i on Agency ,
     Research Triangle Park, North Carolina, July 1981.

18.  Receptor Model Technical Series, Volume IV -Technical Con side rations in
     Source Apport i eminent by Parti cle Td'enti f i catTon . DraftTTf naT Report  from
     Engi neerf ng-ScienceJTo U7S. Envi ronmental Protecti on Agency,  Research
     Triangle Park, North Carolina, September 1982.

19.  W. C. McCrone and J. G. Delley, The Particle Atlas, Volume I, Principles
     and Techniques, 2nd Edition, An n Arbor Sci en ce"~P" libl i she r s Inc., Ann
     Arbor, Michigan, 1973.

20.  Receptor Model Source Composition Library.  EPA 450/4-85-002, U.S.
     Environmental Protection Agency, Research Triangle Park, North Carolina,
     November 1984.

21.  Estimation of the Importance of Condensed Parti cul ate Matter  to Inhalable
     Part icu late"!  EPA 450/4-83-012.  U.S. Environmental Protection Agency.
     Research Triangle Park, N.C, 27711.  April 1983.

22.  "Method 5 - Determination of Participate Emissions From Stationary
     Sources," 40 CFR 60, Appendix A.
                                     5-42

-------
                   6.0  DEVELOPMENT OF CONTROL STRATEGIES

6.1  Oyerview
     This section Identifies general approaches with respect to the use of
ambient measurements and model  estimates in determining the level  of control
needed to demonstrate attainment of PMiQ NAAQS.  Conceptually,  this Involves
determining the PMio design concentration for a particular site or receptor
that must be reduced to the level of the NAAQS, thereby assuring attainment.
These design concentrations are used to determine the level of  control
needed.  The guidance contained in this section focuses on participate  matter
in the specific size fraction,  PMjo, although 1t may be applied to particulate
matter in general as a surrogate where PMjQ data bases are inadequate.
     In all cases where PMjQ ambient concentrations estimated by a
dispersion model are used, the design concentration is assumed  to be the
sum of concentrations contributed by the source(s) and an appropriate back-
ground concentration.  With the PMio annual and 24-hour NAAQS,  two separate
design concentrations, one for each standard, are needed per site.  Attain-
ment of the annual NAAQS requires that the expected annual PMio concentration
be less than or equal to the level of the NAAQS.  Attainment of the 24-hour
NAAQS requires that the expected number of exceedances of the NAAQS be  less
than or equal to one per year.
     The SIP-related emission limits should be based on the NAAQS (annual or
24-hour) which result in the most stringent control requirements.   For
example, if the annual NAAQS requires more stringent control requirements
than the 24-hour NAAQS, the annual NAAQS is considered the more restrictive
standard and the corresponding emission limit(s) would be adopted,
                                  6-1

-------
6.2  Data Base Requirements
     The design concentrations for attainment of the 24-hour PNjQ NAAQS
can be based on ambient measurements of PMio, or model  estimates  of ambient
concentrations at Individual sites during 1 or more years of stable
emissions conditions.  Ideally, (1) modeling estimates  using 5 years of
National Weather Service meteorological data (or at least 1 year  of
on-s1te data), or (2) 3 years of representative air quality measurements
should be considered in determining 24-hour design concentrations.   If
more years of data with relatively unchanging emissions are available,
they also may be considered in calculating design concentrations.  The
more years of data available, the more stable the estimate of PNiQ
design concentrations.
     The preferred approach for estimating a design value is through the
use of an applicable dispersion model corroborated by receptor models, any
available TSP data (using Appendix B), and any available PMio data.  If
there is no applicable dispersion model and at least 1  complete year of
PM10 data are available,* the PM^Q data would be used to estimate the design
value; if the PMjQ data are insufficient,  the design value would be based
on Appendix B of this guideline and corroborated by the available PHiQ
data.
6.3  Hethodolog1es for Determin1ng Design Concentrations
     The annual design concentration is the expected annual  arithmetic
mean determined by the approach discussed in Appendix K, of Part  §0.  In
the simplest case, the design concentration can be determined by  averaging
3 years of monitored or 5 years of modeled PM^Q concentrations.
    *These data should meet first year sampling requirements  found  1n CFR 58.13.
                                  6-2

-------
     There are several acceptable approaches for determining appropriate
24-hour PMio design concentrations.  These approaches which are described
1n the next sections are based on monitored or modeled PMig concentrations.
They Include:  (1) a table look-up procedure; (2) fitting a statistical
distribution; (3) graphical estimation; and (4) the use of conditional
probabilities.  Each of these approaches and corresponding data usage
requirements are presented in detail in the ozone guidellne.l  The
following sections briefly summarize each of these approaches and
indicate how the technique may be applied for determining PNjQ
concentrations.
     6.3.1  Table look-up
     The 24-hour PMiQ design concentration is influenced primarily by the
few highest measured or estimated concentrations at a site.  Availability
of the highest concentrations makes it possible to construct a simple
table look-up procedure to determine the design concentration.  All
portions of the year should be adequately reflected in the measurements.
     To use the tabular approach for the 24-hour PMiQ standard, It 1s
necessary to know the total number of 24-hour PM|g concentrations at  the
site and then select the design value from among the highest concentrations.
The number of available 24-hour concentrations determines which of the
highest concentrations is chosen as the design concentration.  For
example, if a comprehensive monitoring program provides 1,095 24-hour
concentration measurements (or 3 full years of data) at a site, then
the ranks of the lower and upper bounds obtained from Table 6.1 are 4
and 3 respectively.  This means that an appropriate design concentration
                                  6-3

-------
for that site would be between the fourth-highest and third-highest
concentrations.  In using this table, the lower of the two concentrations
should be used as the design concentration, i.e., the fourth-highest
concentration.  Therefore, In this example, it suffices to know only
the four highest values during the time period,  With multiple monitoring
sites, the highest PMio concentrations at each site would have to be
considered and a design concentration established for each location.
For example, the "controlling" design concentration for an area with
seven sites, each having 1,095 values, would be the highest of the
seven fourth-highest values.
     For routine model applications with 5 full years of 24-hour
concentration estimates, the PMjg design concentration of critical
interest becomes the highest of sixth-highest concentrations for the
entire receptor network.
      The look-up procedure is basically a tabular technique for
determining what point on the empirical  frequency distribution cor-
responds to a frequency of 1/365.  By construction, the table look-up
procedure tends to provide a design concentration slightly lower than
would be derived using a continuous curve representing a theoretical
frequency distribution for PMio values.   For example, use of the
table-derived estimate might be modified by interpolation between the
                                  6-4

-------
                               TABLE 6-1
            TABULAR ESTIMATION OF PHio DESIGN CONCENTRATIONS
Number of Dally    Rank of Upper
     Values            Bound
     < 347

 348 - 695

 696 - 1042

1043 -
1

2

3
                                     Rank of Lower
                                         Bound
                                            1

                                            2

                                            3

                                            4

                                             5
                             Data Pofnt Used for
                             Design Concentration
Highest Value

Second Highest Value

Third Highest Value

Fourth Highest Value
                                  6-5

-------
third- and fourth-highest values.  However, this adds an additional  element
of calculation.  Nevertheless, if a more precise design  concentration
should be desirable, the use of interpolation formulas (Section  6.3.2}
or more simple graphical procedures (Section 6.3.3)  may  be necessary.
     For the cases which are limited to less than a  complete year of data,
(I.e., 365 observations) the maximum concentration must  generally be used
as a tentative design value.  In this case it should be  recognized that
the maximum concentration generally represents a lower-bound estimate for
the true design concentration.  In order to provide  an alternative higher
estimate for the design concentration, the extrapolated  value derived from
a fitted distribution (Section 6.3,2) can be used.*   With sparse data
sets, the tentative design concentrations defined as the maximum concen-
tration or the extrapolated concentration are quite  likely to require
further revision as more data become available.  In  addition, the failure
to adequately account for yearly variations in meteorological conditions
makes any estimate based on a single year of data very tentative.
     6.3.2  Fitting onestatisticaldistributiontoseveral  years of data
     With several years of fairly complete PMig 24-hour  air quality
measurements or model estimates, a statistical distribution  could be selected
that "fits" the data.   Information on fitting statistical distributions
can be found on pages 18-20 in the ozone guidel1ne.l»2»3  Because we are
Interested in peak concentrations, emphasis would be placed  on the top
     *An extrapolated value must be used, however,  instead of the maximum
observed concentration in order to evaluate the possibility that the 24-hour
standard 1s controlling,  using procedures described 1n Sections  6.3.2 -  6.3.4.
                                  6-6

-------
5 to 10 percent of the concentrations.  This approach, at least conceptu-
ally, provides a more stable estimate of the design concentration;  but
It also Involves additional computations and Interpretation.
     Criteria for judging reasonable fit are also given In the ozone
guideline.  The design concentration corresponds to a frequency of 1/365.
In some cases the available data will fall 1n this frequency range.  In
such cases, the fitted distribution should be consistent with the data
In this range.  With adequate data, there will be concentration datt
points on either side of the design frequency, data which can be used
as a constraint in fitting the distribution.
     When less concentration data are available, I.e., infrequent
air quality measurements, it may not be possible to "bound" the design con-
centration.  For example, If there are no measured values on the empirical
frequency distribution with frequencies less than 1/365, the estimated
design concentration will represent an extrapolated concentration.   This
extrapolated design concentration will be higher than the maximum observed
or estimated (via air quality model) value.  When this 1s the case, caution
should be exercised in the use of the extrapolated value.
     6.3.3  Using the empirical frequency distribution ofseveralyears
            of data (graphical  estimation)
     With sufficient data it may not be necessary to fit a statistical
distribution, as discussed in Section 6.3.2.  The concentration value
corresponding to a frequency of 1/365 may be read directly off a graph
of the empirical  distribution and used as the design concentration.  The
description of this approach is given on pages 25 and 34 in the ozone
guideline.
                                  6-7

-------
     6.3,4  Condi11onal probabl11 tyapproach
     While the previous methods required grouping concentration  data
from several years, this approach allows individual  years of data to be
treated separately.  This is done by fitting a separate statistical  distri-
bution to each year of data and assuming a given probability that each  year
will reoccur.  This approach is of interest when different importance
should be placed on the individual years in terms of meteorological  con-
ditions or sampling completeness.  For this reason,  this approach may also
be of interest for the annual standard.  The conditional probability approach
is somewhat theoretical in nature, but is adequately discussed in the ozone
guideline.
6.4  Determining Emission Limits
     6.4.1  General
     Once PMjo design concentrations have been established through
the use of air quality measurements or model estimates, a proportioning
method can be used at each site to estimate control  requirements for SIP
development.  This proportioning method differs from simple rollback in
that the source contributions are determined from receptor or dispersion
modeling and not directly from the emissions as in simple rollback.
Nevertheless, the method conceptually considers that the total reduction
(TR) in pollutant concentration should be:
            TR (ug/m3) = PH10 Design Concentration - PM10 NAAQS
If a design concentration is greater than the NAAQS  (TR is positive), a
reduction in PM^o emissions is required.  This might be accomplished by
                                  6-8

-------
reducing the contribution of a single source or 1t may require reduction
1n several Individual sources or source categories so that
          TR (ug/m3) - SISR1 (ug/m3)
          where ISR-j Is the individual Source Reduction desired from
          a source or source category 1.
These ISR's are generally selected based on many considerations,  Including
the technical feasibility of achieving a given emission reduction or
additional reduction at that source.  The percent reduction  1n emissions
($RE) for a source or source category Is given by
                    ISR, (ug/m3)
            %RE1 =  A^ (ug/m3)
            where AC is the ambient concentration due to the Individual
            source (i) or source category as determined through a model.
The SIP must demonstrate that the control requirements will  be adequate to
meet the NAAQS under other situations where the relative source contributions
may be different from that on the design day.  This is discussed further  in
Section 6.4.3.  If receptor models are used to determine the relative  source
contributions, the operating rates of the sources over the time period
studied must be evaluated.  If they are inappropriate for the SIP development,
the source contributions should be adjusted proportionally.
     6.4.2  ^xample for annual averages
     Assume for discussion that the annual  NAAQS for PM^Q is 75 ug/m3
and that the site with the "controlling" annual  design concentration for  the
area is 100 ug/m3.  Then the total required reduction for the area would  be;
            TR - 100 - 75 = 25 ug/m3.
                                  6-9

-------
     Next, for the site of the most restrictive design concentration, consider
the following source contributions estimated by a receptor or a dispersion model
                                              Source Category Contributions
            Source                                to Atnbi en t Cone .  ( AC )
            Steel mill roads                     15 ug/m3
            Steel mill coke ovens                10 ug/m3
            Coal storage pile                     5 ug/m3
            Urban paved roads                    20 ug/m3
            Cement plant                         10 ug/ffl3
            Background                           40
                                                100
The TR would be the sums of the individual reductions in ambient
concentration.  Assume that the following reductions in source concentration
are selected for consideration:
            Source                   Individual Source Reduction (ISR)
            Steel mill roads                    .  8 ug/m3
            Steel mill coke ovens                 5 ug/m3
            Cement plant                          8 ug/m3
            Coal storage pile                     4 ug/m3
                                           TR =  25 ug/m3

Since the TR sums to 25 ug/m3, it is assumed that the individual source
reductions, if implemented, would reduce the annual PM^Q concentration at
this "controlling site" to 75 ug/m3.  The required emission reductions
(%RE) to accomplish this are calculated for each source category.   For
example:
                                 .     .
                      =  ^cement   10
so that control requirements would be 80 percent for the cement plant.   It is
assumed that the selected emission reductions will  allow the NAAQS to be
met at the "controlling site," as well as all other monitoring or receptor
sites.  This assumption should be tested by reversing this procedure and
determining whether the anticipated reduction impacts on other sites above
                                  6-10

-------
the standard will result in attainment at those sites.  If this is not found
to be true, which is more likely in multisource situations than areas
affected by a small number of sources, individual  design concentrations
at other sites will have to be considered and the emission reductions
reassessed.  In such cases, the control strategy will be composed of the
maximum of each of the individual  source reductions calculated across
the network.
     For the proportioning method discussed here,  the background
concentration should be estimated as that portion  of the concentration which
is not attributed to the sources being investigated.  This estimate should
be based on actual observations in nonurban areas  near the boundary of the
area or on model estimates of the actual impact of the sources not under
investigation.
     6.4.3  Cp_n_sJ_d_erat1ons for 24-hour averages
     An approach similar to that discussed in the  above section can
be employed to determine control requirements for the 24-hour standard.
The PMio design concentrations and individual source contributions for each
site, as well as background concentrations, must be available.  However,
for short-term averages (e.g., 24-hour), it is likely that a single high
concentration is dominated by relatively few sources with source contri-
butions varying with meteorological  conditions.  Thus identification of
emission limits that assure the short-term 24-hour NAAQS will be met at
all sites is likely to require an  iterative and perhaps lengthy process.
This is especially true of multisource areas.
     An alternative is to develop  control  strategies with emission
limits specified for each source category, perhaps based on a preliminary
                                  6-11

-------
application of the modified rollback model.  The strategies would then  be
tested with a dispersion model  to determine if the NAAQS are met  everywhere.
Those strategies that allow the NAAQS to be met would be identified for
further consideration in preparing a set of preferred emission  limitations.
This alternative, too, could Involve an iterative procedure.  However,  it
provides a means by which factors such as the best technological, most
cost effective, and most enforceable set of emission controls can be con-
sidered.  Such a procedure might also be employed for annual concentrations,
especially where many receptor sites are of concern.
                                 6-12

-------
Reference

1.  Guideline for the Interpretation of Ozone Air Quality  Standards,
    EPA-45Q/4-79-003, U.S. Environmental  Protection  Agency,  Research
    Triangle Park, NC, January 1979.

2.  Curran, T.C. and W.M.  Cox, "Data analysis Procedures for the Ozone
    NAAQS Statistical Format," J.  Air Pollution Control Association,
    June 1980.

3.  Curran, T.C., "Data Screening  for Large Air Quality Data Sets,"
    paper no. 84-52.3, presented at the 27th Annual  Heeting  of the Air
    Pollution Control Association, San Francisco, CA,  1984.
                                  6-13

-------

-------
                 7.0  SIP REQUIREMENTS AND DATA REPORTING

7.1  Introductlori
     This section describes the requirements for SIP revisions in the
Act, EPA's PM^Q SIP development policy, the interface between PMjo and
TSP control strategies, and summarizes how PHjQ emissions data should be
reported to EPA's data bank.  To develop an acceptable control strategy
as required for a PMip SIP, a State can utilize the information summarized
1n section 4.0 of this document for carrying out appropriate modeling,
the information in section 5.0 to develop an emissions inventory for
model inputs, and the information in section 6.0 for developing control
strategies.
7.2  Clean Air Act Requirements
     7.2.1  Time 1imits, SIP requirements
     Section 110(a) of the Act requires every State to submit to EPA a
SIP which provides for implementation, maintenance, and enforcement of  each
NAAQS in each air quality control region within the State.   The SIP is
required to be submitted within 9 months of the promulgation of a new
NAAQS or the revision of existing NAAQS.  The SIP must provide for attaining
a primary NAAQS as expeditiously as practicable, but no later than 3 years
after SIP approval by EPA.  Up to an additional  2 years could be provided
for attainment if the requirements in section H0(e) of the Act are met.
     These section 110 requirements are applicable to PMjg  NAAQS.  The  PMjQ
SIP's therefore are due within 9 months of promulgation of  PMjQ NAAQS and are
to cover all areas of the State as qualified in section 2.4 of this document,

                                   7-1

-------
     7.2.2  Regulations in 40 CFR Part 51
     The regulations promulgated 1n Part 51 address the preparation,
adoption, and submittal to EPA of SIP's for implementing the  NAAQS.  These
regulations reflect the requirements set forth in the Act.  Part  51 has
recently been restructured to contain more subparts and sections  to make
it more organized and readable.  That restructuring effort  consisted only
of deleting obsolete material and reorganizing the remaining  material  and
no substantive changes were made to the regulatory requirements at that
time.  The Part 51 regulations still apply to PMip.
7.3  PMio SIP Development Policy
     As discussed in section 7.2.1, the entire State must be  covered by a
     SIP and the SIP is to be submitted within 9 months of  the date the
     NAAQS are promulgated.  However, because PMio data are not available
in all areas, EPA has placed areas into three groups.  Section 2.4 of  this
document discusses grouping procedures.  The requirements for PM]p SIP
development are described in the following sections.
     7.3.1 Group I PHio SIP requirements
     States are required to submit to EPA a revision to the particulate
matter SIP within 9 months of promulgation of PMio standards  for  Group I
areas.  The SIP revision must provide for attaining PH-jg NAAQS as
expeditiously as practicable, but no later than 3 years after the SIP  is
approved by EPA.  The 3-year attainment deadline may be extended  for up
to 2 additional years by the Administrator if the conditions  in
section 110{e) of the Act are met.  The submittal must include a  modeled
demonstration that provides for attainment and maintenance  of the PMjo

                                   7-2

-------
standards.  Portions of the SIP pertaining to prevention of significant
deterioration/new source review section and air monitoring will  also
require revision.
     The SIP must contain control  measures sufficient to demonstrate
attainment.  Provisions for studies or demonstration projects of nontrtdi-
tional particulate matter sources  and measures necessary for attainment
at a later date will not be acceptable as they were in the past  for TSP
standards,                                                :
     7.3.2 Group II PMio SIP requirements
     States are also required to submit a SIP within 9 months of promulgation
of PMiQ standards for Group II areas.  The States may submit a SIP for
these areas as required for Group  I areas, if they wish.  Otherwise,  States must
submit a SIP which contains enforceable commitments to take the  following actions.
(1)  Gather ambient PMio data, at  least to an extent consistent  with  minimum
     EPA requirements and guidance (see subsections 3.2.3 and 3.2.5)
(2)  Analyze and verify the ambient PMio data and report 24-hour PMio NAAQS
     exceedances to the appropriate Regional  Office within 45 days of each
     exceedance.
(3)  When an appropriate number of verifiable 24-hour NAAQS exceedances
     become available (see section 2.0} or when an annual  arithmetic
     mean above the level of the annual PMio  NAAQS becomes available,
     acknowledge that a nonattainment problem exists and immediately  notify
     the appropriate Regional  Office.
(4)  Within 30 days of the notification referred to in (3) above,  or  by
     the date 37 months after promulgation, whichever comes first, determine
                                   7-3

-------
     whether the existing SIP will  assure timely attainment  and  maintenance
     of the WIQ standards, and immediately notify the appropriate  Regional
     Office.
(5)  Within 6 months of the notification referred to in (4)  above  (if
     necessary), adopt and submit to EPA a PMjg control strategy that
     assures timely attainment and maintenance within a period of  3 years
     from approval of the committal SIP.
     The following factors should be considered in determining the  adequacy
of the existing SIP in item (4) above:
     (1) Air quality data, (Time is alloted for up to 3 years of PMjQ
         data to be collected if a NAAQS is not violated sooner.  At t,he
         end of that time, the available PMjo data must be examined to
         determine if attainment can be demonstrated in accordance  with
         Appendix K of 40 CFR Part 50 or the Guideline on Exceptions to
         Data Requirements jfor Determi ni ng Attai nment of Particulate
         Matter Standards in the absence of adequate PMjo data.)
     (2) Emissions data.  (The emission inventories must be  evaluated to
         determine if emissions can increase significantly because  actual
         emissions are far below allowable emissions for the area;  determine
         if sources with operating permits are not operating or  operating
         at reduced capacity and if "banked" emissions impact future air
         quality.)
     (3) The present control  strategy.  (The existing control strategy
         should be evaluated to determine if it is fully implemented; if
         it is adequately enforced; start-up,  shutdown, and  malfunction

                                   7-4

-------
         regulations are adequate to prevent circumvention of emission
         limits; and it can adequately attain and maintain the PMjQ NAAQS  1f
         the above conditions are met.  The evaluation should Include the
         use of dispersion and receptor modeling techniques where appropriate)
     The committal SIP must include an enforceable schedule with appropriate
milestones or checkpoints.  The EPA will review and act on both the committal
SIP's and control strategies submitted under step (S).  Also, revisions
required in the PSD/NSR and atr monitoring portions of the SIP must be
submitted with the committal SIP.
     7.3.3 Group III PMiQ SIP requirements
     For Group III areas, EPA will presume that the existing SIP is
adequate to demonstrate attainment and maintenance of the PM^o standards.
States are therefore only required to revise the PSD/NSR and air monitoring
portions of their SIP's within 9 months.
7.4  SIP Content
     A most important first step in developing a control strategy will be  to
inventory particulate matter sources and include all  possible contributors.
Receptor models applied to TSP samples, or PMjo samples if available, may
be useful for this analysis as discussed in section 5 of this document.
The State should include 1n its emission inventory point sources, fugitive
emission points within industrial plants, plus all area sources such as
unpaved and paved roads, unpaved parking lots, construction activity,
open lots with no vegetative cover, woodstoves, agricultural activity,
and similar sources.  As previously stated, the option to study nontraditional
sources and commit to implementing controls at a later date will no
longer be acceptable.  Information at the State level  or from EPA-funded

                                   7-5

-------
studiesl»2>3,4,§,6 should provide enough information to make determinations  of
which controls may be effective for these sources.  In determining  which
controls to employ, the State should reassess existing control  technology
requirements for traditional sources, previously approved emission  trades,
and the effectiveness of startup, shutdown,  and malfunction regulations.
7.5  Control Strategy Transition
     Particulate matter emissions from most  point sources and many  area
sources have been controlled as a result of  SIP's to implement  the  former
NAAQS for TSP.  The regulatory requirements  of an existing TSP  SIP  must
remain In effect in accordance with section  110(i) of the Act,  until  a
PMio SIP is approved by EPA.  Therefore, regulations in the existing  SIP
cannot be relaxed without a demonstration that the revision will  not
interfere with attainment or maintenance of  the PM^o NAAQS.  The  existing
regulations must continue to be enforced by  Federal and State agencies
during the period of transition from a TSP SIP to a PMip SIP.
     States will no doubt want to minimize any unnecessary disruption
caused by going from these control programs  to PMjo programs.  Therefore,
to the extent possible, States should utilize the existing control  strategy
in a TSP SIP as the basis for a PM^g program sufficient to attain and
maintain the PMio NAAQS.  The EPA expects States to build on the  current
control strategies to whatever degree necessary to demonstrate  attainment
and maintenance of PMjQ NAAQS.  This may include adopting the current
control strategy in full, if it can be shown to be sufficient for PMjQ
purposes, or adopting it in part.
                                   7-6

-------
7.6  Current Emission Regulations
     Participate matter SIP's contain emission regulations expressed  in
various terms including the following:
     — Mass of particulate matter per heat input into the process.
        Example:  0.30 pounds per million British thermal  unit.
     -- Mass of particulate matter per unit of exhaust gas.
        Examples:  0.05 grams per dry standard cubic foot.
                   0.01 grains per actual cubic foot.
                   0.02 grains per standard cubic foot.
                   0.04 pounds per ton of exhaust.
                   0.02 grains per actual cubic foot per minute.
     — Opacity of exhaust.
        Example:  20 opacity.
     — Mass of particulate matter per time period.
             i
        Examples:  144.30 pounds per hour.
                   631.00 tons per year.
                   13.44 pounds per day.
                   4.5 grams per day.
     -- Mass of particulate matter per area exposed  to wind erosion.
        Example:  5.80 pounds per acre.
     — Mass of particulate matter per mass of product input or output.
        Examples:  0.02 pounds per ton coke.
                   0.30 pounds per ton kiln feed.
                   52.00 pounds per metric ton.
                                   7-7

-------
     — Mass of particulate matter per unit of power generated.
        Example:  0.07 grains per horsepower hour.
     — Percent control of emissions.
        Example:  98.00 percent control  of particulate matter emitted due
to overspray of coating material.
     Regulations of these types for controlling all  emitted particles have
been the accepted practice for affecting ambient concentrations  of particulate
matter as measured by the high volume sampler even  though emissions as
measured by stack test methods may not be identical  to the material
measured by a high volume sampler.  Despite this discrepancy, EPA has
developed stack sampling trains now described in Methods 5 and 17 of
Appendix B to 40 CFR Part 60 which represent the present state-of-the-art
in techniques for collecting particulate matter.  Many emission  limits for
particulate matter therefore have been established  in terms of these
measurement methods.  In this document,  the term "particulate matter emissions"
is used to denote material measured by Methods 5 or 17 or by a comparable
measurement method approved by EPA in a State implementation plan.
7.7  Surrogate Emis s i on Regu1 at1ons
     This subsection discusses setting or retaining emission limits in
terms of particulate matter emissions for the purpose of controlling PMjQ
emissions as part of a PM^o control strategy.
     7.7.1  Retaining existing emissionlimits.  If a State finds that its
existing particulate matter emission limits in its  TSP SIP are sufficient
to prevent PMio NAAQS violations, there is no need  to go through a resource
intensive process of modifying the emission limits  to express them in terms
                                   7-8

-------
of PMio.*  Modeling can be used In two ways to determine that a particulate
matter emission limit OP combination of limits for several  sources  1s
acceptable for PMjo control.  First, participate matter emissions can be
modeled to determine the impact on ambient TSP levels.  If  the ambient
TSP levels determined through modeling are below the concentration  levels
of the PMiQ NAAQS, then the modeling results themselves indicate that PMio
NAAQS would not be violated.  The particulate matter emission limits  will
then be acceptable for controlling PMio emissions.  Second, modeling  can be
done with particulate matter emission rates converted to PMjo emission
rates.  If it can be assumed that for any source the PMio fraction  of
particulate matter does not vary appreciably during normal  operation  of
the source, then a certain particulate matter emission rate will represent
a certain emission rate of PMjQ for that source.  The PMjo  fraction will,
of course, vary by source and could be up to 100 percent of particulate
matter emissions.  For some sources the particulate matter  emission rate
in the TSP SIP can be converted to the corresponding emission rate  for
PMio through an emission factor (see section 5,0 of this document).  If a
State uses this corresponding PMip emission rate as input for modeling and
finds that the modeling results indicate no violation of the PMio NAAQS,
then the particulate matter emission rate will be an acceptable emission
limit for the PMio SIP.
     In either of the above cases, States may leave the existing emission
limits unchanged and may continue to conduct compliance tests with  appropriate
 *Regulations may consist of techniques such as paving a  certain  amount  of
  dirt roads or streets, street sweeping schedules,  tailings  pile spraying,
  construction activity procedures, street salting restrictions,  unpaved
  road speed limits, etc., rather than an emission limit.
                                7-9

-------
particulate matter emission measurement methods.   The PMjQ  control  strategy
analysis should Include the modeling results and  confirmation  that  the
partlculate matter emission limits were found to  be acceptable surrogate
emission limits for PMiQ,
     7.7.2  Setting new emission limits.  In the  event that a  State finds
that an emission rate lower than the current particulate matter emission
limit in the TSP SIP is necessary for PMio NAAQS  attainment and maintenance,
the PMiQ rate which is found to be necessary through modeling  can be con-
verted (through the use of an emission factor) to an equivalent partlculate
matter emission rate and expressed as such in the PMiQ SIP. The emission
limit expressed in terms of particulate matter emissions will  be acceptable
provided the modeling results are valid and the ratio of PMio  to particulate
matter emissions is not likely to drastically change.
     If it is found that substantially all of a source's emissions  are PMiQ,
then limits for that source could be expressed as either particulate matter
emission limits or PMio emission limits, and compliance testing could be
performed with either Method 5 or a method that measures only  PMio  emissions.
For such an instance, the SIP would have to indicate that the  source's
emissions are essentially all PMjo-  In any case  where an emission  regulation
is in terms other than PMiQ, such as particulate  matter emissions or opacity,
PMio levels must not increase while the indicator used in the  emission
limit remains constant.  Otherwise, the Indicator will not  be  a valid
surrogate and emission limits using the surrogate will not  be  acceptable.
7.8  PMiQ-Specif1c Emission Regulations
     Emissions limits developed to meet the NAAQS for TSP took many forms
as illustrated in section 7.6.  Compliance with these limits was determined
                                   7-10

-------
by reference method emission tests or certified visible emission observers.
The emission limits and compliance methods had no direct link to TSP as
measured by ambient monitors, however, a correlation was made through the
use of dispersion and receptor models.
     Similarly, it would be difficult, and perhaps technically infeasible, to
develop a method for measuring precisely that exhausted material that would
contribute to ambient levels of particulate matter measured by an ambient
PHio sampler.  Particles emitted from stacks are subject to agglomeration
and separation after they reach the ambient air, both of which affect
their inclusion or exclusion in ambient PMiQ.  Additional complications
arise due to the presence in emissions of condensables and various precursors
of secondarily formed particulate matter.  Regulations specifying emission
limits as part of PMio control strategies, therefore, cannot be directed
toward exactly that exhausted material that contributes to ambient PMjo-
     Rather, PMio emission limits must be directed toward reducing the amount
of PMio emitted from a source as measured by an approved compliance test
method.  Suggested procedures for measuring PM^o emitted through stacks
are discussed in Appendix C.  Further guidance is being prepared by EPA,
7.9  Reporting Emissions Data
     Emissions of pollutants for which NAAQS have been set are reported by
States to EPA in an annual report required by 40 CFR 51.322,  As a minimum,
the sources for which emissions data are to be reported are those whose
actual emissions are over 90.7 metric tons (100 tons) per year of the
pollutant of concern.  Since the indicator for both the primary and
secondary.standards for particulate matter is being changed from TSP to
    , reporting of particulate matter as required in section 51.322 will
                                   7-11

-------
simply be replaced by reporting of PMjQ emissions.  The present requirement
to report particulate matter emissions will end with the reporting of
calendar year 1987 emissions.  The reporting requirement for PMjQ emissions
will begin with the reporting of calendar year 1988 emissions.   The EPA
recognizes that time and resources are needed for States to develop the
capability to report PM^Q emissions data for EPA to develop the capability
to process, store, and retrieve the data.  States are required  to begin
the annual reporting of PMjp emissions data with calendar year  1988 data,
which are to be sent to EPA Regional Offices by mid-1989.  The  EPA plans
to provide States with the needed technical and procedural information  1n
time for States to meet that requirement.  This information includes
PMiQ-related changes to the following:  AP-42, "NEDS Source Classification
Codes and Emission Factor Listing;" the computer program for calculating
emissions that is provided routinely to those States which use  EPA's
Emissions Inventory System (EIS); and provided to other States  on request;
and procedural information such as that contained in Aeros manuals.
     The EPA has underway a project to replace the UNIVAC systems, e.g.,
NEDS, Hazardous and Trace Emissions System, that have been used for many
years to process, store, and retrieve the emissions data provided by
States.  These systems will be replaced by IBH systems under EPA's Aerometric
Inventory Retrieval System (AIRS) development program.  The current
schedule for having AIRS operational on the IBM system is March 1989.
The EPA will provide the necessary guidance to the States in time for
States to respond to the requirement for reporting calendar year 1988
PMjo emissions.
                                   7-12

-------
7'10  Emissions Trading (Bubble) Policy
     This section clarifies the effect of PMip NMQS on alternative emission
reduction options (bubbles) that have been previously approved  for TSP SIP's.
In the Initial bubble policy, as published on December 11,  1979 (44 FR 71780),
sources were warned that EPA was considering revising Its  partfculatt
matter NAAQS and that If such size-specific standards were promulgated,
some alternative approaches. Initially approved by EPA, might no longer
be adequate under the revised standards.  In a sense, the  policy Indicated
that sources which used a bubble approach to meet SIP emission  limitations
could be treated no differently from sources which did not. That is, if
additional emission reductions are required to avoid violations of the
ambient air quality standards, the State may have to revise emission limits
previously approved under the bubble policy.  In general,  bubbles
cannot interfere with a State's efforts to attain and maintain  ambient
air quality standards even if those standards are revised.   The final
Emission Trading Policy Statement promulgated December 4,  1986  (51 FR 43814)
affirms this position,
7-11 Fyjitjve Dust Policy
     The EPA will continue to implement its existing fugitive dust policy
as it was applied in rural areas violating the TSP NAAQS.   The  EPA Issued
guidance on SIP development and new source review in areas  impacted by
fugitive dust on August 16, 1977.7  This guidance, known as the Fugitive
Dust Policy, states that "urban areas should receive the highest priority
for development of comprehensive and reasonable programs to control
                                   7-13

-------
fugitive dust."  Control programs In rural areas are to "center on the
control of large existing man-made fugitive dust sources (I.e., tailings,
piles, mining operations, etc.) which 1n themselves are presently causing
violations of the NAAQS or are sources of a known toxic or hazardous
material (e.g., asbestos},"
     Another aspect of the fugitive dust policy Is that, "new sources
that wish to construct 1n rural fugitive dust areas should be allowed to
do so without the need of emission offsets, as long as they comply with
the applicable emission regulation and the impact of their emissions plus
the emissions from other stationary sources in the vicinity of the proposed
location, along with normal background, is not projected to cause violations
of the NAAQS."  The following criteria were to be used 1n defining a
rural area under the fugitive dust policy:  "(1) the lack of major industrial
development or absence of significant industrial particulate emissions
and (2) low urbanized population (i.e. eastern states <100,000-200,000 or
western states 
-------
options 1s currently being solicited.  If necessary, a revised policy
will be Issued In the future.  At that time, the categorization of areas
Initially placed 1n Group III under the existing policy will be reviewed
for compliance with the revised policy.
                                   7-15

-------
References
1.
Ident1f 1 catIon,  Assessment.  and Control  of  Fuji t1ve  Partlculate
Emissions. EPA-600/8-86-023 U.S. Environmental Protection  Agency,
Research Triangle Park,  North Carolina,  August 1986.
2.   Kinsey, J.S., P. Englehart, and A.L. J1r1k, Study of Construction
     Related Dust Control, Midwest Research Institute and ETA Engineering,
     Prepared for Minnesota Pollution Control  Agency, RosevUle,  Minnesota.
     December 1982,

3.   Port1 and Road^Dust Demonstratlon Project. Final  Report, Seton,  Johnson
     and OdeliTTnc.V"for City of Portland, Oregon, Department of Public
     Works, July 1983.

4.   Denver Pemonstratlon Studyt PEDCo Environmental, Inc., Kansas City,
     WT.  Prepared for Colorado Division of Air Pollution Control,
     Denver, Colorado.  October 1981.

5.   User's Guide:  Fugitive Dust Control Demonstration Studies,
     EPA-600/8-84-032, U.S. Environmental Protection Agency, Research
     Triangle Park, NC, January 1985.

6.   Field Evaluation of jfind Screens as a Fugitive Dust Control  Measure
     for Material Storage Piles, EPA 600/7-86-027 U.S. Environmental
     Protection Agency, Research Triangle Park, North Carolina.  July 1986.

7.   Memorandum from Edward F. Tuerk, Acting Assistant Administrator for A1r
     and Waste Management to Regional Administrators, August 16,  1977.
                                   7-16

-------
                                 APPENDIX A
                          TYPES OF RECEPTOR MODELS

     There are several major categories of receptor models which are
potentially valuable for better understanding an area's ambient particulate
problem, both PMjg and TSP, and in formulating strategies for use in State
implementation plans (SIP's).  These include chemical mass balance (CMB),
factor analysis (FA), optical microscopy (OM), and automated scanning  electron
microscopy (ASEM).  While there are other techniques 1n addition to these,
information and availability of the models or analytical  methods are limited.
The major receptor models are described briefly below and references are
provided for more detailed information. 1,2,3,4,5,6,7,8

1.0  Chemical Mass Balance
     This method compares the chemical "fingerprints" or profiles of emissions
from several source categories to the chemical composition of the sample.   It
is referred to in the literature by various names, including chemical  mass
balance, chemical  element balance, species mass balance,  and mass balance.
Weighted least squares or some other statistical routine 1s used to find the
relative proportions or mix of these sources which best "explains" or  accounts
for the composition of the sample.  Species data (chemical compounds or
elements) are required for both source and sample data.  The method 1s most
effective when the ambient samples are collected in at least two size  fractions
(larger than and smaller than 2-3 micrometers).  This method must be validated
in accordance with the protocol referenced at the end of  this Appendix.9
                                     A-l

-------
2.0  Factor^ Analysts
     Factor analysis and other multivariate statistical methods, such as
multiple linear regression, target transformation factor analysis (TTFA),  and
cluster analysis, are all variations of the least squares routine used 1n
mass balance.  However, these methods (except TTFA} require no prior assumption
of the Impacting sources and can be useful where the types of sources are
uncertain.  Like the mass balance method, this method Is most effective when
used on size segregated samples.  Factor analysis usually requires at least
40 observations to complete the analysis.
3.0  Optical Microscopy
     Particle identification by optical  microscopy is one of the first and
most widely used methods of source apportionment of coarse particles.  The
technique relies on identification of particles by their size, shape, color,
surface properties, and birefringence (i.e., an optical property).  It 1s
generally used on particles larger than 1 micrometer.  Several analysts
have been asked to provide quality assurance of their analyses and the results
have been mixed.  Optical microscopy appears to be somewhat analyst dependent
and semiquantitative.  However, it is an excellent confirmatory or supplementary
techniques.
4,0  Automated ScanningElectron Microscopy
     A computer-driven scanning electron microscope, equipped with x-ray
fluorescence capability to identify the elemental composition, provides a
particle-by-particle analysis of particle size, shape and elemental  composition.
This elemental composition and computer assisted sizing allows large numbers
                                     A-2

-------
of particles to be analyzed and, thus. Identified by source category 1n  a
relatively Inexpensive manner.
5.0  OtherReceptorMethods
     The following methods are suggested to corroborate or refine MB,  FA,  OM»
or ASEM analyses,
     5.1  X-Ray Diffraction (XRD) - This method provides direct information
on the crystalline nature of particles.   This is particularly useful  for
distinguishing among minerals.  It requires a heavy loading of coarse
particulate matter on the filter and cannot be used on ammorphous (I.e.,
non-crystalline) particles.
     5.2  Trajectory Analysis - This technique is used for tracing the history
of an air mass to determine the possible area of origin of sources of ambient
particulate.  The simplest form is the pollution rose, where only average  wind
direction is considered.  More sophisticated approaches account for the  chang-
ing wind direction with time to trace air mass in time and space for as  long
as several  days.
     5.3  Hlcroinventory - This technique is an orderly compilation of sources
near a receptor.  This is particularly useful as a finely gridded Input  to
dispersion models, and in qualitatively estimating the potential sources of
soil and other minerals to help interpret an observed mix of particulate matter
on an ambient filter.
                                     A-3

-------
References

1.  Receptor Model Technical Series. Volume I:  Overview Of Receptor Model
    Appllcation To Participate Source ApporT!o~ninent  EPff-450/4-81 -OToTaT
    U.S. Environmental Protection Agency," Research Triangle Park, NC, July
    1981.

2-  Receptor Model Technical Series, Volume II:  Chem1cal Hass_Balance*
    EPA-450/81-016b, U. S. Environmental ProtectFon Agency, Research Triangle
    Park, NC, July 1981.

3.  Receptor Hodel Technical Series, Volume III:  User's Manual For Species
    Mass Balance Computer Program. EPA 45U/4-83-014. U.S. Environmental
    Protection Agency, Research Triangle Park, NC, July 1983.

4.  Receptor Model Technical Series, Volume IV:  Technical Considerations In
    Source Apportionment By Particle Identification, EPA 450/4-83-018.UTsT
    Environmental Protection Agency, Research Triangle Park, NC, June 1983.

^ •  Digest ofjanblent Partlculate Analysis and Assessment Methods,
    EPA-450/ 1T-78-113, U. S. Environmenta1 Protect1 on Agency, Research Triangle
    Park, NC, September 1978,

6.  S. E. Gordon, Receptor Models, Environmental Science and Technology, 1980,
    14, 792-800.

7.  Receptor Model Technical Series, Volume V:  Source Apportionment Techniques
    And Considerations In Combining ThelrTse, EPA 450/4-84-020.  U.S.
    Environmental Protection Agency, Research Triangle Park, NC, July 1984.

8.  ReceptorHodel Technical Series, VolumeVI:  A Guide To The Use of
    Factor AnaTysis jruj[_Multiff e Regressi on (FA/HRj Techniques" 1 n Source
    Apportlonmentt EPA 450/4-85-007.  U.S. Environmental Protection Agency,
    Research Triangle Park, NC, July 1985.

9.  Pace, T. G., et al., Protocol For Applying And ValidatingTheCHB.  U.S.
    Environmental Protection Agency, "lesearchiriangle Park, NC, 1986
    Draft.
                                     A-4

-------
                                 APPENDIX B


     PRELIMINARY ESTIMATES OF PM10 DESIGN CONCENTRATIONS USING TSP DATA


      This appendix provides a methodology for miking a preliminary estimate

 of particulate matter (PMio) design concentrations where actual PM^p measure-

 ments are not available.  These estimates may be derived based on relation-

 ships between PMjQ and TSP, as described in the probability guideline.  The

 definitions of design concentrations used in this appendix are consistent

 with those presented in Section 6.0.  Preliminary design concentrations

 developed by this Appendix method shall not be used for developing control

 strategies.  They may be used for preliminary planning and to help evaluate

 design concentrations estimated by dispersion models.  Procedures for

 making preliminary estimates of annual design concentrations and 24-hour

 design concentrations are included.


 1.0  Annual NAAQS

      1.1  Recommendations
           (1)  Use the most recent years for which a valid annual
                arithmetic mean* exists to determine annual arithmetic
                mean TSP concentration, and

           (2)  Multiply the annual arithmetic mean TSP concentration
                determined in (1) times a factor of 0.47.**
 *A valid arithmetic mean requires at least 12 observations per quarter.  In
  the simplest case, to compute the design concentration one should use the
  average mean from the 3 most recent years of data.  However, if only 1-2
  years of valid means are available, these may also be used.

**The 0.47 factor is the 50th percent!le ratio in Table 2 of Procedures For
  Estimating Probability Of Nonattainment OfA PMjn NAAQS UsinfTotal Suspended
  Pajrticulate Or Inhalable PaVtTcutaj^JJajta.


                                     B-l

-------
          (3)  The resulting product of (I) and (2) 1s the "design concen-
               tration" for the annual arithmetic mean PM   value.
     1,2  Example
          (1)  Given;  Valid average annual arithmetic mean TSP = 140 p/m3
                       PM|n NAAQS = 50 y/m3 annual arithmetic mean.
                      Sampling 1s performed according to a systematic
                       sampling schedule.

          (2)  Solution;

               d = (0.47) (140)

               d = 66
2.0  24- hour NAAQS

     2.1  Background

          Unlike the annual design concentration, a design concentration
for Implementing the 24-hour NAAOS requires some explanation before
recommendations concerning the derivation can be given.  The 24-hour NAAQS
is met when the expected exceedance rate (EER) Is less than or equal to 1.0
per year.  When TSP measurements are the only measurements of particulate
matter available, the calculated EER is a function of:

               (1)  the relationships between PMjg and TSP as described
                    in the probability guideline,

               (?)  the observed distribution of TSP values, and

               (3)  the number of samples,

          In deriving the design concentration, one should remember that the
design concentration, d, is that value of PMjg which, when reduced to the
level of the NAAQS,   £, would result in an expected exceedance rate (EER)
of 1,0 per year.  Hence,
                                                               (1)

               where m = (l-[required reduction in d])

          Two assumptions are implicit in the definitions above:

               (1) all TSPi values in the existing distribution of TSP
                   concentrations are reduced by (1-m), and
               (2) all corresponding (PMjo)^ values are also reduced by
                   (1-m).
                                    B-2

-------
          The expected exceedance rate (EER) is calculated using equation
(2).

                        n
                     N
               EER = S  I  P(R, >   Jt  )                         (2)
                        i=l       TTSPli

               where

               N  = number of days per year (e.g., 365}
               S  = number of TSP samples
               n  = number of TSP samples with values greater than  i,  »
                    the level of the NAAQS
               RI - ratio of (PMi0)i/(TSP)i
                  = level of the NAAQS (e.g., 150 ug/m3

          and

               p^Ri>  (TSP)^ = the probability that a given (TSP}1
                                value corresponds to a (PMig)^  con-
                                centration which is greater than the
                                level of the NAAQS.

     Noting that the design concentration is defined as  that value  which
must be reduced to the level of the NAAQS such that EER  = 1.0,  and  that
for this to occur each (TSP)i value must be reduced to m (TSP)-j, equation
(2) becomes


               1.0 -365 £'    P(Ri >    £   )                 (3)
                      S  i=l         m (ISP),

     One further substitution can be made by substituting the relationship
in equation (1) into equation (3).
     This allows the design concentration "d"  to be expressed  explicitly  in
equation (4).

     To calculate the average exceedance rate  over a period  of k years,
equation (4) should be applied separately for  each year  to estimate each
annual expected exceedance rate (EER)j.   The (EER)j are  then averaged  as
shown in equation (5).
                                   B-3

-------
                1.0=  165
                where
k
z 1
j-i SJJ
n
E
J=l
P(R
                               1J
(5)
                k  = number of years considered
                Sj = number of observations  during year  j

      Equation (5) is solved iteratlvely for "d"  until the  right  hand  side
 (rhs) of the equation becomes equal to 1.0.

      2.2 Recojmendations

           Because the recommended procedure for  estimating the 24-hour
 design concentration for PM^g from TSP data is an iterative one,  it is
 recommended that the estimate be made using a computer. Software for doing
 this is available and is described in Usej^s Guide  for  PMio Probability
 Gui del i ne Software .*  For those wishing tb^ perform "We "calculations by hand,  a
 suitable procedure is described below.

           (1)  Assume an initial value of "d" twice as  high as the level
                of the PM10 NAAQS (i.e., 300 pg/nr for a NAAQS having  a level
                of 150 yg/m3);
           (2)  Using Figure B.I,** for each (TSP)^j  value greater than
                the level of "d," calculate the probability that  the
                corresponding (PMio)ij  value would be greater than this
                level, i.e.,
                P(Rij
                Note that
(a)  if (TsF)ij  <  0.14,  assume
                                                                 0.99
                        _d_                                        d
                (b) if (TSP)ij > 0.95 but £ 1.00,  assume P(Rfj  > (TSP)^}  =  0.01
 *W. P. Freas, jJser*_s Guide for
                     Probability SuldjaHne Software, U.S.
  Environmental Protection Agency,  Office of Air Quality  PIanning^and  Standards,
  Monitoring and Data Analysis Division,  Research Triangle  Park,  North
  Carolina  27711.  In Preparation.

**Figure B.I has been reproduced from Table 2 of the probability  guideline.
                                    B-4

-------
                                   K_l—  f«u«WHII-l If K SO DIVISIONS.
                                    C  KEUFFEL * ESSEH CO. HUH t* uSi
                                                            1U]J>
        4680CM3
99.99
99.9  99.8        99   98      95     90       80    70   60   50   40   30    20       10     5       2     I   O.S    0.2  0.1 0.05    0.01
1.0
0.9
0.8
0.7
0.6
(XI
I
en
0.5
0.4
0.3
0.2
0.1
















FI


GL


!R
-

E


B


.1


Exan
PM1(
Ind
" ~7

np"
/'
IV
mV
4-1

lo
r-s
id
n

I
p
u<
*

)i
F
il
*

!a
A

t
t
2
>

ribut
ios f
4-hou
i
-•--pi
•• j 1

ion (
or
r Ob
"" f.

3
S
r

f
sr
'

"vations.: — :...:.
	 *-- 	 • 	 g-
t-.««..»^..._.._._ji .».«_..
. _ *, ^ , 	 -____™. 	 ^ . _ _ ,

	 	 2---
	 _„ , 	 ^_«___™
i
— „ ..jt... — _ 	
i 	 	 	 	 	 . _ _


n


i


— i .


A





























—
—
0.0!    0.05 0.1  0.2    0.5    1
                     2       S      10       20    30   40   50   60 . 70

                         Cumulative  Frequency of
 80      90     95      98    99
Less Than  a Given Value
99.8 99.9        99.99

-------
               or
                        d                              d
               (c) if (TSPljj >_ 1.00,  assume  P(RU  >  (Tfp)ij) « 0

          (3)  Use the information 1n  step (2)  to determine the value of
the rhs of equation (5).

          (4)  (a)  if 0.99 <_ rhs <_ 1.01,  then  the  design concentration is
equal to the value of "d,"

               (b)  if rhs > 1.01, increase the value of "d" and repeat
steps (2) and (3);

               (c)  if rhs < 0.99, decrease the value of "d" and repeat
steps (2) and (3).

          (5)  If there are fewer than 48  samples in  any year, disregard
the data from that year In performing  the  preceding calculations.

     2.3  Exampl e

          Given:   (1) Three years of TSP sampling data in which 61 days
                  per year are sampled,
                  (2) Five days in which (TSP}^  was  observed to  be greater
                  than 150 ng/m3.  These values  are  500  and 240  ug/m3
                  in year 1, 400 and 350 pg/m3 in year 2,  and 300 yg/m3
                  In year 3.

          Flndr  The design concentration of the 24-hour NAAQS for PMjp.


          Solution;   (1) Assume d = 300 pg/m3, compute each p^Rij > TSP"}
                     and tabulate the information as shown below

                1      1      (TSP)j      300/(TSP)j      P(R> 300/(TSP)j)
                1      1       500          0.60              0.20
                1      2       240        >  1.00              0
                2      3       400          0.75              0.05
                2      4       300          1.00              0.
                3      5       350          0.86              0.02

                     (2) Use the Information in the last column in the
     preceding table to test equation  (5)

     1.00 f  365/, [ 1 [0.20 + 0] +  1  [0.05 + 0]  +   1  [0.02]]    (5)
     .  !             IT             IT             IT

     1.00 f 0.54
                                    B-6

-------
                    (3)  Since 0.54 < 0.99, we need to lower the  trial
design concentration, d.  For a second trial value of "d,"  halve  the  :
difference between the level of the NAAQS and trial 1.  Therefore,  the
value for "d" in trial 2 is 225 pg/m3.

                1       1      (TSP)j     225/(TSP)j      P(R> 225/(TSP)i)
1
I
2
2
3
I
2
3
4
5
500
240
400
300
350
                   0.45
                   0.94
                   0.56
                   0.75
                   0.64
                                                               0.56
                                                               0.01
                                                               0.26
                                                               0.05
                                                               0.14
     and
         1.00 = 365A £ 1 [-56 + .01]
         1.00 ? 2.03
               1  C.26  +  .05] +   1  [.1413
              6T              m
          Since 2.03 > 1.01, we need to raise the design concentration  "d"
for trial 3.  This is done by taking half the difference between  trials 1
and 2 and adding it to the design concentration in trial 2.   Hence,  the
design concentration for trial 3 becomes 225
                   i
(TSP)j
1
1
2
2
3
1
2
3
4
5
500
240
400
300
350
                365
263/{TSP),

   U.53
 > 1.00
   0.66
   0.88
   0.75

  1
(300-225)
    2     =  263  pg/ra3

     ?(R>  263/(TSP)i

          0.34
          0
          0.12
          0.02
          0.05
                                 1
         1.00 9  3  UTlQ.32 + 0)) + ~5T~ (0.07 + 0.01)  + 6l~ (0.02)]

         1.00 f 1.06

          Since 1.06 > 1.01, we must raise the design concentration  for
trial 4 by adding one-half the difference between the design  concentrations
used in trials 2 and 3 to the design concentration in trial  3.   Hence,  the
design concentration for trial 4 becomes 263 + 1/2 (263-225)  -  282 ug/m3.
          This binary search procedure is continued to derive  subsequent
trial values for "d" until finally we reach a trial value of 268 iig/m3.
                                   B-7

-------
1
1
2
2
3
1
2
3
4
5
500
240
400
300
350
                          (TSP)<     268/(TSP)i
                                        0.54
                                        1.00
                                        0.67
                                        0.89
                                        0.77
                       P(R>  268/(TSP)j

                             0.32
                             0
                             0.11
                             0.02
                             0.05
                 365   1               1                  1
          1.00 t T~ [5T (0.32 + 0) + IT (0.11 + 0.02) + "BT (0.05)]

          1.00 = 1.00

          Therefore, use d = 268 pg/ra^ as the design concentration for
     at this monitoring site.  Note that if the sample size in year 2 had
been less than 48, we would ignore the data from year 2, and equation (5)
would have been applied as shown below,
          1.00
                 365,
n   DfD   v „  d,   I        n
2   PIKU   (TSPJ.J  +  1  S
1=1                     S3 1-1
                                   B-8

-------
                                 APPENDIX C
              GUIDELINES FOR SOURCE TESTING FOR SIZE SPECIFIC
                           PARTICULATE EMISSIONS
1.0  .Introduction
     The Introduction of a size-specific, ambient participate standard may
necessitate source measurements on a slmiliar basis.  The EPA is providing
this guideline not as a mandated requirement, but as guidance for those
States that want to develop emission factors beyond those which we will
provide 1n Section 5.  This guidance may also be useful to States which
want to develop compliance methods for future PMiQ SIP emissions limits.
The material presented here is intended for the use of knowledgeable stack
testers in modifying existing techniques for gathering PMjo data.  The EPA Is
currently developing PMjo measurement methods and will provide more detailed
guidance following promulgation.
     The net contribution of a source to PHio in the ambient air arises
from both primary and secondary emissions.  Secondary emissions are defined
here as those which result from chemical or physical reactions in the
ambient air.  Primary particulate emissions are defined here to include
particles which were present in the process gas stream prior to discharge
to the atmosphere as well as modifications to those particles and new
particles formed by condensation phenomena as the process gases are cooled
upon discharge.  Testing methods which Include the contributions from
"condensation" are currently under development but are not yet generally
available (Smith et al, 1982).  The data currently available indicate that
the "condensibles" from many sources with good particulate controls may
constitute a major fraction of their present contributions to ambient air
particulate concentrations (Williamson et al., 1982).  The EPA Method 5 (40 CFR

                                    C-l

-------
60), with a filter temperature of 120°C, includes some condensibles 1n  the
material collected (EPA 450/3-81-005a).  However, further temperature
reduction in many cases results 1n substantial increases in measured emission
rates (Williamson et a!.; op cit).  The methods to be described below focus
primarily on the measurement of size fractionated particulate matter as 1t
exists at stack conditions.  However, the inclusion of those portions of
the condensibles that are currently collected as part of the EPA Method 5
catch can be readily achieved by the removal of the final filter from
the instack samplers described below and instead using a Method 5 filtration
system.  The problem of complete measurements of primary emissions including
proper handling of condensibles measurement will be addressed at a future date.
2.0  Avaljjible Methodsfor Size Specific Source Measurements
     Because the PMio standard is one based on the aerodynamic behavior of
particles, inertia! classification is the appropriate technique to use.
The likelihood of irreversible agglomeration or coalescence occuring upon
contact of particles with one another precludes the use of laboratory
particle sizing methods for the determination of PMiQ emissions from industrial
sources.  Rather, particle classification on an aerodynamic size basis  must
be accomplished while the sample is being taken.  This can be done using
either of two types of inertia!  classifiers—cascade impactors or cyclones.
If PHio emissions only (not detailed size distributions) are to be measured,
a single stage classifier would be the most desirable device to use.
     Size fractionating source samplers have been produced with as few  as
one and as many as 25 stages with each stage of the multistage devices
having a different cutoff diameter.  The samplers having the greater number
of stages provide more detailed information on the aerosol size distribution,
                                    C-2

-------
and generally have higher capital and operating costs.  Impaction type size
classifiers are much more prone to errors due to overload and particle
bounce than are cyclones; consequently, multistage operation is virtually
mandatory with them in order to insure data reliability (Harris,  1977).  A
sound theoretical basis exists for predicting the performance of cascade
impactors (Marple, 1970).  Cyclone samplers perform well  as single stage
collectors; however, no adequate theory exists at present to predict their
behavior over a wide range of operating conditions.  This limits  the use of
cyclones to those which have been demonstrated to produce the necessary
particle size cutoff by empirical calibration procedures  (Smith et al»
1982).  Cyclones with acceptable sharpness of cut have been developed and
are available from commerical vendors.  Evaluation of these devices have
demonstrated nominal 10 urn cuts over a limited range of operating conditions,
and further characterization of the devices is in progress.  Cyclone samplers
are preferred for sampling to quantify size specific mass emissions; however,
the protocol described here is generally applicable to both impactor and
cyclone sampling with details that are pertinent to only  one of the methods
noted as they arise.
 3.0  Selection of Sampling Traverse Points
     In order to obtain a representative measurement, one must obtain samples
at representative points across the duct (stack) at isokinetic rates.  In
the case of conventional particulate testing (e.g. EPA Methods 5  and 17),
this is accomplished by dividing the duct into a large number of  equal area
segments and traversing this sampling plane to obtaining  an isokinetic
sample at the center of each of these areas.  Isokinetic  sampling is achieved
by selecting a nozzle which is appropriate for the combination of the nominal

                                    C-3

-------
   flow  rate  at  which  the  sampler  1s  Intended to operate and the average duct
   gas velocity.   Compensation  for variations in the duct gas velocity 1s then
   achieved by modulating  the  sampling  rate.  This procedure cannot be used
   with  inertia!  particle  size  classifiers because changes in the sampling
   rate  also  changes the diameter(s)  at which size fractionation takes place.
   However, to require fixed sampling rates adds immensely to the complexity
   of the  testing and  subsequent data reduction and interpretation.
        A  method is currently  being investigated which permits complete traverses
   of the  sampling plane using  variable sampling rates to be made with cyclones
   and impactors 1n a  fashion similar to EPA Method 5.  It is called the emission
   gas recycle (ESR) approach.  The method uses a recirculation loop to augment
   the sample flow through the  sizing device.  This permits the operator to
   maintain a fixed flow through the  sampler while varying the sampling rate
   from  the duct by means  of compensating changes in the recirculation flow.
   Upon  satisfactory development of the E6R method, the protocol for traversing
   the sampling  plane  for  size  specific sampling will be virtually identical
•   to that of Method 5.
        An alternative procedure for  obtaining a representative multipoint
   traverse in a single  run with a fixed flow rate sampler involves changing
   the sampling  nozzle as  necessary at  each sampling point to match the sampling
   velocity to the duct  velocity within an allowable 20 percent tolerance.
   The volume of gas sampled at each  point is then made proportional to the
   volumetric flow through the  area represented by the sample point by making
   the sampling  duration at each point proportional to the gas velocity at the
   point.  This  technique will work with any velocity distribution but may be
   cumbersome to implement.
                                      C-4

-------
     Another method of obtaining an Isoklnetic sample using a fixed flow
pate sampler is to synthesize a complete sample from a multiplicity of
single point samples.  Obviously the time and expense involved in such  an
approach would be excessive if the number of samples taken equaled the
number of traverse points normally required in Method 5 compliance testing.
However, a method which utilizes this approach, called simulated Method 5
(SIM-5) is being developed.  In the SIM-5 method, one sample is obtained
from all traverse points for which the duct gas velocity falls within a
specified range (usually +_ 20 percent).  Then a complete sample can be
composed from the partial samples obtained from traverse points with similar
gas velocities.
     In the interim, while awaiting the development of the SIM-5 method, a
simple four point sampling grid may be used (Smith et al., 1982).  With a
particle size sampler located at each grid location, one sample at each
location is needed to obtain one complete measurement of the net source
emission rate (the net source emission rate being the sum of the emission
rates from each of the four quadrants).  Three complete measurements should
be made to characterize the source emission rate; thus, a total of 12 particle
size samples (3 samples at each of 4 grid locations) will usually be required
to obtain one complete measurement.  However, if the velocities at the  four
sampling points are within +_ 20 percent of one another, one sampler operating
for equal time at each of the four grid points can obtain one sample.   This
will reduce the number of samples required from 12 to 3.   The locations of
the four sampling points are shown in Figure C-l.
                                    C-5

-------
                            a/4
                        b/4
                                        4181-28
FIGURE C-1. RECOMMENDED SAMPLING POINTS FOR CIRCULAR AND SQUARE
          Off RECTANGULAR DUCTS,
                               C-6-

-------
     Implementation of all of the traverse methods described above require
obtaining velocity traverses and gas composition using EPA Methods 1,  2,  3,
and 4 followed by precalculation of sampling flow rate, nozzle selection,
etc. before sampling commences.  The total gas flow to be used In calculating
emission rates using any of the reduced point traverses must be obtained
using EPA Method 2.
4.0  Sampler Selection and Operation
     Cyclone samplers for PMio measurements are available on the comnercial
market.  Any of a number of cascade impactors are available, also.  High
flow rate (10 to 30 alpm) samplers are desirable for sampling low concentration
streams such as are anticipated herein.  If a high concentration (>500 mg/m3)
stream is to be sampled an impactor with a design flow rate of about 3 1pm
would be more useful.  The need to use different samplers for
different particle concentrations arises from a fundamental operating  limit
of impactors.  Under most circumstances, no more than about 10 mg can  be
collected on any one impactor stage; beyond this limit particle reentrainment
can invalidate the data.  As a rule this means that only about 50 mg can  be
collected in total throughout the impactor.  High flow rate impactors  reach
this load limit too rapidly for practical use in high concentration situations
(Harris, 1977).  Cyclone samplers have sufficient capacity that a single
high flow rate device will suffice for all applications.
     Sampling nozzle selection is more complex when sampling to obtain
particle size related information than when sampling for total particulate
loading.  Two major factors are added regarding nozzle selection.
     Because of size selective deposition in bends, the sampling nozzles
must be straight for use in particle sizing (Knapp, 1979; Felix and McCain,

                                    C-7

-------
1981).  Thus one cannot use "gooseneck" or other nozzles designed to turn
the sampled gas flow 90 degrees (as is done in Method 5).   The  natural
right angle orientation of the inlet of a cyclone sampler with  respect to the
sampler body and exit makes the use of straight nozzles normal  with them.
This is not the case with impactors.  One approach to the use of straight
nozzles with Impactors is to align the impactor axis with the direction of
gas flow in the duct.  However, in many sampling locations this is impractical
if not virtually impossible.  The use of a high capacity precollector having
an inlet at 90 degrees to the body eliminates this difficulty and, at the
same time, eliminates another operational  problem with Impactors.  If a
significant proportion of the particles being sampled are larger than about
10 to 15 urn, the first collection stage tends to reach its loading limit
before the succeeding stages collect enough material for reliable weighing.
The use of a high capacity precollector which has a cut larger  than that of
the first impactor stage helps alleviate the overload problem.   Several
such precollectors, some of cyclonic design and others operating on impactlon
principles, are comercially available.  Of the impaction type precollectors,
only those which have inlets and exits at 90 degrees to one another are
acceptable for this application.
     The particle size of the samplers, both cyclones and Impactors, are
dependent on a number of factors with the sampling flow rate being the only
variable which can be adjusted by the user.  However, the sample flow rate
required to obtain the PMjo cut will be dictated by the sampler used (given
the gas composition and temperature of the process stream being measured).
This means that one does not have the latitude in selecting the sampler flow
to be used that one has in simple total particulate measurements. The

                                    C-8

-------
matching of the sample Inlet velocity with the gas stream velocity  for
Isokinetic sampling must be accomplished entirely through the  cross-sectional
area of the sampling nozzle.  This means that a much larger array of  nozzles
must be available than those used In Method 5 sampling.   If the  isoklnetlc
error 1s no larger than 20 percent, the maximum error in the measured
emission rate of 10 urn particles will be about 15 percent and  the errors
for smaller particles will be lower.  Thus, the actual  error 1n  the total
concentration of particles smaller than 10 urn will be substantially lower
than 15 percent.  Thus deviations of +_ 20 percent from isokentic can  probably
be tolerated.  If sampling is to be done within + 20 percent of  isokinetlc,
an array of nozzles must be available that step by 20 percent  in diameter
from one to the next.
     If a single stage collector is used, the geometry of the  collector and
the flue gas conditions will completely dictate the sampling flow rate as
only one flow rate will produce the required size cut.   If a multistage
device is used to measure the complete size distribution, some latitude is
available in setting the flow because interpolation can  be used  to  determine
the concentration of particles in the designated size range.
5.0  Sampling Trains
     The sampling train requirements for the particle size specific method(s)
are similar to those of Methods 5 and 17 and trains designed for the  latter
applications can be used with some impactors and cyclones.   However,  Method 5
systems are not adaptable to the intermediate and low flow rate  (<20  alpm)
devices without modification.  The orifice meters used  in the  Method  5/17
systems are too large to permit accurate flow measurement and  control at
these reduced flow rates.  A substitute orifice meter,  properly  sized for
                                    C-9

-------
the specific sampler, can be used 1n place of the standard orifice  to make
a Method 5/17 system compatible with the sampler.  Alternative trains and
probes are described by Harris (Harris, 1977).  As the sampling procedure
described here 1s Intended to measure only particulate material  as  It exists
at stack conditions, the sampler must be operated Instack.  That Is, the
size fractionating device and backup filter must be mounted on the  outboard
end of the probe (replacing the usual sampling nozzle).  A typical  setup Is
Illustrated In Figure C-2»  The particle size fractionating portion of  the
system cannot be located at the exhaust end of the probe because of size
selective particle losses resulting from deposition in the probe.  A Method  5
filter system may be used at the exhaust end of the probe, with the
instack filter removed as illustrated in Figure C-3, to obtain an estimate
of the condensible emissions and to provide a method for determining the
fraction of Method 5 concentrations that fall within the defined particle
size range.
     Detailed procedures for the use of cascade impactors for industrial
source sampling are described by Harris (Harris, 1977).  Procedures for the
use of cyclone samplers are described by Smith et al,, (Smith et al, 1982).
Some of the procedures described in the latter document are specific to a
set of cyclones designed to provide a 15 urn size cut, but with the  exception
of the actual flow rates to be used, the same protocol is to be followed
for the sampling described here.
6.0  Collection Media
     In sampling operations in which the particles are not further  size
fractionated beyond the PMJ.Q cyclone, the filter specifications are the same
                                    C-10

-------
FILTER,
HOLDER
                                           IMPINGER TRAIN OPTIONAL:
                                           MAY BE REPLACED BY AN
                                           EQUIVALENT CONDENSER
                              THERMOMETERS     BY-PASS
                                               VALVE
                                                                  CHECK
                                                                  VALVE
                                            IMPINGERS IN ICE BATH
                    ORIFICE
                                       • '    ilXMMX}
                                                  MAIN    VACUUM LINE
                                                  VALVE
                MANOMETER    DRY TEST METER   AIR TIGHT PUMP
                                                                 4111-CilA
        FIGURE C-2. PMW PAftTlCULATESAMPLING TRAIN FOR NONCONDENStBLE
                  PARTICULATE (MODIFIED SPA METHOD 5 TRAIN)
                                     c-n

-------
       HEATED PROBE
SAMPLER
                                            fMPINGER TRAIN OPTIONAL:
                                            MAY BE REPLACED BY AN
                                            EQUIVALENT CONDENSER
                                                                   CHECK
                                                                   VALVE
                                             IMPINGERS IN ICE BATH
                      ORIFICE
THERMOMETERS
 9,
                                              (>Or*-£X3-
                                                   MAIN    VACUUM LINE
                                                   VALVE
                  MANOMETER    DRY TEST METER  AIR TIGHT PUMP
                                                                  41I1-446B
       FIGURE C-3. PM10 PARTICUIATE SAMPLING TRAIN FOR CONDENSIBLE AND
                 NONCONDENSIBLE PARTICULATE (MODIFIED EPA METHOD 5 TRAIN).
                                       C-12

-------
as those for EPA Methods 5 and 17,  However, 1f a cascade impactor  1s  used,
extra requirements are imposed.  A surface coating 1s required  on the
collection plates of the Impactor 1n order to Insure adequate retention  of
Impacted particles.  This coating can be In the form of a glass fiber  mat
or a grease or polymeric coating on a lightweight metal foil  (Harris,
1977),  Chemical reactions Involving gas phase constituents  of  the  process
stream, especially S02» can occur with these coatings and significantly
alter the apparent weight collected by each stage (Smith et  al,, 1975).  All
glass fiber materials used in impactor sampling must be treated to  minimize
SOg uptake.  The procedure for the treatment is given by Gushing and Smith
(Gushing and Smith, 1979),  Because the interferences are not predictable,
control runs must be made together with the actual sample runs  as the
testing takes place (Harris, 1977).
     All collection media must be dessicated prior to the initial weighing
and dessicated for a minimum of 12 hours prior to the final  weighing after
sampling.
7.0  Data Reduction and Analysis
     Ultimately, the information to be determined from a set of measurements
is the participate emission rate in the PMig size fraction together with a
measure of the uncertainty in the calculated rate.  If the traverse of the
stack required multiple single point samples, emission rates must be separately
calculated for each point and summed to obtain the total  rate.   Emission
rates should be calculated for each of the three traverses of the duct and
averaged.  The average represents the final  value to be reported.   Isokinetic
errors would be calculated for each sample in the manner prescribed by EPA

                                    C-13

-------
Method 5.  The average isokinetic error for each traverse must be less  than
_+ 20 percent for the individual traverse to be accepted.   The relaxation
from the + 10 percent requirements of Method 5 1s possible because the
errors associated with an isokinetic sampling of small  particles  are  not  as
great as those for larger particles, and Method 5 must  be able to deal  with
all sizes.
     If the sampler(s) are operated so as to provide a  size fractionation
point at the required 10 urn diameter, the particle concentration  to be  used
in the emission rate calculation is that based on the quantity of material
which passed the appropriate stage of the sampler.  If  a  single stage
sampler is used, the actual size fractionation diameter for each  sample must
be calculated based on the calibration of the sampler,  the gas volume
sampled, and the composition and physical conditions of the flue  gas.  For
a sample to be accepted, the size fractionation diameter  calculated for the
sampler from actual run conditions must be within _+ 10  percent of the
required diameter.  If the complete size distribution is  measured, the
particle concentration to be used in calculating the emission rate can  be
found by interpolation (or, within limits, extrapolation) of the  cumulative
concentration versus particle size curve generated from the data  for  each run.
Methods for accomplishing this are described by Smith (Smith et al.,  1982).
Cascade impactor data can be reduced on a single run basis using  a sophisticated
data reduction program, PADRE, which was developed under  EPA sponsorship.
This program can be accessed through a local phone link in most U.S.  cities
and there are no charges for the use of the service. Details on  the  data
reducion system can be found in "PADRE User's Guide" (EPA-600/58-84-012).
                                    C-14

-------
References

Anom.  "Code of Federal Regulations" Title 40, Chapter 1,
Sub Chapter C, Part 60.

Anom. "Control for Particulate Emissions from Stationary Sources"
Vol. 1 EPA 450/3-81-005a, September 1982.

Gushing, K.M. and Smith, W,B. "Particulate Sampling and Support:
Final Report" EPA-600/2-79-114, June 1979.

Felix, L.G. and McCain, J.D. "Errors in Recovered Particle Size Distributions
Caused by Sampling with Bent Nozzles" Paper 81-7.4 presented  at the  1981
Annual Meeting of the Air Pollution Control Association, Philadelphia, PA,  1981

Harris, D.B. "Procedures for Cascade Impactor Calibration  and Operation  in
Process Streams" EPA-600/2-77-004, January 1977.

Smith, W.B. Cashing, K.M., Johnson, J.W., Parsons, C.T., Williamson, A.D.,
and Wilson, R.R., Jr.  "Sampling and Data Handling Methods for Inhalable
Particulate" EPA-600/7-82-Q36, May 1982.
                                                                     i
Knapp, K.T. "The Effects of Nozzle Losses on Impactor Sampling."   In:
Proceedings, Advances in Particulate Sampling and Measurement (Daytona
Beach, Fl, October, 1979), W.B. Smith, Ed. U.S. EPA Report EPA-600/9-80-
004, Southern Research Institute, Birmingham, Alabama, January, 1980.
pp 101-107.

Marple, V.A.  A Fundamental Study of Inertial Impactors.  Ph.D. Thesis,
University of Minnesota, Mechanical Engineering Department, Mineapolis,
Minnesota, Particle Technology Laboratory Publication 144, 1970.   243 pp.

Smith, W.B., Gushing, K.M., Lacey, G.E., and McCain,  J.D.  "Particulate Sizing
Techniques for Control Device Evaluation" EPA-650/2-74-102-a, August 1975.

Williamson, A.D., McCain, J.D., and Parsons, S.C. "IP Sampling for Condensible
Emissions" Paper 82-61M.3 presented at the 1982 Annual Meeting of  the Air
Pollution Control Association, New Orleans, LA, June  1982.
                                    C-15

-------

-------
                                 APPENDIX D

                      DETERMINATION OF PM10 BACKGROUND

1.0  Introidyctlon
     Development of State implementation plans (SIPs) for PMjQ will  require
an estimate of PMjo background concentrations for use with either dispersion
or receptor model based control strategy development.  The term "background"
has a number of different Interpretations.  For the purpose of SIP develop-
ment, background is defined as that portion of (ambient) concentrations due
to natural sources, nearby sources other than the one(s) currently under
consideration, and unidentified sources.  General guidance on background
concentrations for all pollutants 1s provided In Guidelines on Air Quality
Models.l  The purpose of this Appendix is to outline in greater detail  the
considerations and procedures which should be used to determine background
concentrations of PMjQ.  As will  be seen, selection of an appropriate back-
ground concentration depends on the time period of concern (e.g., annual  or
24-hour), meteorological conditions of interest and the availability of data.
     In a dispersion modeling study, background should account for the
impact of source emissions not considered in the model.  Such emissions may
be either of manmade or natural origin and may be either primary (directly
emitted as particles or condensable vapors) or secondary (emitted as gases
and transformed into particles during transport).  Ideally, all  sources
should be modeled, but this is not always possible or practical  for  particu-
late matter.  In some cases, the sources may be within the agency's  jurisdiction,
but their emissions are not included explicitly 1n the model  (perhaps
because they are too small or too numerous to inventory).  In other  cases,
                                     D-l

-------
the sources may be outside the region of jurisdiction and not Identifiable.
However, all sources must be accounted for either explicitly by the node!  or
in the background value.
     Receptor modeling studies may also be used to develop control  strategy
emission limits, as outlined in Section 6.0 of this document.  Since a
receptor analysis may apportion the entire sample (Including background),
the analysis may reflect contributions from some sources which are  or
should be ascribed to background.  Section 3.0 of this Appendix discusses
this 1n more detail.
     For dispersion model studies, the background concentration is  usually
needed for the set of meteorological  conditions and time periods (e.g.,
annual, seasonal or 24-hour) associated with the exceedances of the KAAQS.
Receptor models usually require the estimate for the specific time  period
represented by the receptor analysis.  However, receptor model studies may
also be developed that represent a particular set of meteorological  condi-
tions.  This need for background to represent a particular set of meteoro-
logical conditions is discussed in Sections 3,0 and 4.0 of this Appendix.

2.0  Determining Background Estimates From MeasuredL PJJLQ Or Surrogate Data
     The PMiQ background values are most appropriately taken from measurements
at a nearby site, representative of background for the area being studied.
Such a site would ideally measure PMjo and not be influenced by (1) sources
being studied and explicitly included in a dispersion or receptor model, or
(2) sources not impacting the study area.   This usually means that  the site
be located in the vicinity of the study area.  For a site that meets these
prerequisites, the mean annual background for use in the SIP analysis would
be the average of the annual  concentrations at that monitor over the time

                                     D-2

-------
period being evaluated (usually 3 years).  Ideally, 12 valid 24-hour
samples for each quarter constitute sufficient data to calculate an annual
average.  However, for the purpose of estimating background, less data may
be used if it can be shown that an estimate by any alternate method 1s
more  reliable.
      Dispersion modelers identify specific sets of 24-hour meteorological
conditions that may lead to or be associated with high 24-hour concentrations
of particulate matter.  Also, receptor modelers often group the samples by
meteorological "regime" and consider each group separately in receptor analy-
ses.  These specific 24-hour sets of meteorological conditions are matched
with  meteorological conditions on days when measured particulate matter data
are available.  The measured data collected on days whose meteorological
conditions best match the specific conditions are then averaged to obtain
a representative background concentration.  As discussed above, care must
be taken to ensure that sources being explicitly considered are not upwind
of the "background" monitor on the days selected.
      In the absence of a single monitoring site which is clearly not
influenced by any of the sources considered explicitly in a dispersion or
receptor model, a number of judgments and adjustments may be necessary to
obtain a background estimate.  Three alternatives are available.  The first
alternate method is to exclude the impact of these sources on the monitored
background value by using a wind direction analysis.  This consists of
calculating the average background concentration for those days when the
wind direction is such that the monitor is not downwind of a significant
source.   Monitoring sites outside of a 90° sector downwind of a source may
be used to determine background concentrations.  The mean annual or seasonal
                                     D-3

-------
average concentration Is the 3-year mean of the annual  or seasonal
concentrations which are calculated 1n this manner.  If the location  of the
sources or the frequency of sampling 1s such that there are an Insufficient
number of days at a single monitor to calculate a background value  which  1s
not Influenced by those sources, background may be determined by "compositing"
the air quality concentrations from several monitoring sites located  In the
vicinity of the study area.  This composite average concentration may be
developed for the time period being studied by selecting the "upwind" site
on each monitoring day and averaging these "upwind" values (see Section 5.0
of this Appendix).  The mean annual concentration is then the average of
the annual concentrations which are calculated by the composite method.
     This procedure is greatly simplified for 24-hour cases.  The background
for a given day or set of meteorological conditions is simply the average of
the concentrations at a site or sites that are upwind on the day or days
which are selected to represent those conditions.
     As a second alternative for using measured data, background concentra-
tions may be derived using site measurements for a different day or year
than the time period being studied.  Conceptually, using data from  the same
site but during a different day or year is similar to the first alternative
above but additional care must be exercised in showing that the data  used
are representative of the background sources and climatological conditions
during the study period.
     As a third alternate, if PM^ or TSP data are available, they  may be
used as surrogates to estimate PHjo concentrations.  Use of either  TSP or
PMjs data involves the use of ratios, or multipliers, to adjust measured
background PMis or TSP to PMjQ.  Then the estimated PMiQ data would
                                     D-4

-------
be treated 1n the same manner as a PMio measurement in the preceding
paragraphs.  Surrogate data for PMjo should only be used if there  are
insufficient PMjg data for use in applying the procedures above.   In cases
where PMig data are available for part of a year and the PHjn, monitor is
collocated with TSP or PMj5 monitors, the ratio approach may be used for
the remainder of the year when no PHio data are available.  The ratio
approach is discussed in Section 6.0 of this Appendix.

3.0  Spedal Considerati ons^For Receptor Model 1 ng
     In some cases, receptor models are used for source apportionment at
urban or source oriented sampling sites.  In such cases, it may be necessary
to perform a source receptor analysis on a nearby background receptor so
that the background for specific source components can be determined and
discounted prior to using the apportionment (as outlined in Section 6.0 of
this Appendix).  This receptor analysis at a background site may  be necessary,
for example, to distinguish between locally generated soil dust and that
transported into the area.  Source apportionment analysis of the  background
samples should use the same procedures as used for the source analysis.
Then the receptor analyses from the background and study monitors  are
compared.  The sources that were identified on both monitors should be
noted and the concentations of these sources at the study monitor should  be
reduced by the amount of their impact at the background site.  In  this
adjustment process, one must carefully select the background days  for
receptor analysis to make sure they are "upwind" and are not impacted by
the sources of concern.
                                     D-5

-------
4 , 0  Summary Of Alternatives For ; Estlmatijig PHip Background
     The preferred method for background determination Is to use an  upwind
PMjo site 1n the vicinity of the study area (air shed, city, or source)  under
review, which reported data during the time period in question but which is
not influenced by the sources in question or by other sources which  do not
impact the study area.  Since such ideal measurements are not always availa-
ble, particularly for an annual  average determination, the alternate methods
discussed previously are summarized below, in order of preference.
       First          Use perimeter s1te(s) in the vicinity of the study area.
    Alternative       The effect of nearby sources that are in the modeling
                      analysis must be eliminated through the use of wind
                      direction analysis (annual or 24-hour),  A "composite"
                      background value based on several nearby sites and wind
                      direction analyses may be developed.
       Second         Use PHjo data from a different days or a different year
    Alternative       to represent the day or year being studied.
       Third          Use surrogate data (TSP or PMis),  adjusted to
    Alternative
5.0  EXAMPLE 1 - Compositing Data From Several PerimeterjSites^
     In the absence of a single appropriate PMjg background site with  a
sufficiently large data base, data from two or more sites in the vicinity  of
the study area may be combined to provide a composite estimate of background
concentration.
     For example, assume that there are two sites located on the area  perimeter
(one to the north and one to the south).  The observations from both sites
should be sorted by wind direction into three groups:

          Group   I - days when the Northern site is upwind (use Northern
                      site data for Nj. days).
                                     D-6

-------
          Group  II - days when the Southern site is upwind (use Southern
                      site data for Ng days).
          Group III - days with calms, substantial shifts 1n wind direction,
                      during the day or with winds predominantly from the
                      east or west (and both sites are upwind).   (Use average
                      of data from both sites for N3 days.)

The Group I» Group II, and Group III days are used to compile an annual
average:

                         N!            N2            N3
     annual average =    E  Group I +  2  Group II +£_. Group III
                                          NI + Ng + NS
     where N^ and Ng are the number of days where only the northern or
southern sites, respectively are upwind*  Nj is the number of days where both
sites are upwind.
     This calculation can, of course, accommodate more than two  monitoring
sites.  Also, a single site may be used if it is generally upwind of the
study area, so long as there are sufficient data to represent average upwind
concentrations after the days when the source is upwind are removed from the
data base.
                                    D-7

-------
6.0  EXAMPLE 2 - Use Of Surrogate Data
     Most regional monitoring networks contain one or more designated
background hi-vol sites and a few have PMig background sites.   However,  there
are relatively few size-fractionated parti cul ate samplers measuring PMiQ
compared with those measuring TSP; and almost  all  of the former are in urban
locations.  Therefore, States may not have direct measurements of PMjo background
available when they prepare their SIP's.  The  following is the recommended
procedure for using PHis and TSP surrogates to estimate PMio.  based on an
analysis of regional scale PMiQ, PMis and TSP  sites.
     The PM|5 measurements may be used as a surrogate for PMiQ.  The PMjg
measurement (either annual average or 24-hour) is multiplied by a constant
ratio to give an estimate of PM1Q,2  The following adjustment  applies.
          PMio = 0.85 x PMi§ (annual  or 24-hour)
     Reference 3 suggests PM^g to TSP ratios of .61 (annual) and .75 (high
24-hour).  Using the above 0.85 ratio for PMjo to PMjg, the following ratios
apply for estimating PMio background using a TSP surrogate:

          PMjo = 0.52 x TSP (annual)
               = °*6* * TSP (high 24-hour)
     The concentrations to be adjusted are obtained using the procedures  in
Section 2.0 of this Appendix.  The "adjustment"  is simply the last  step in
the process, wherein a TSP or PMi§ background estimate is adjusted  to
represent an estimate of the PMiQ background.
                                   .0-8

-------
References

  1.  Guideline on Atr Quality Models (Revised)  EPA-450/2-78-027  R.
      U.S.  Environmental Protection Agency, Research Triangle Park,  NC
      27711, July 1986.

  2.  A, K. Pollock, A, B, Hudeschewskyj and A. D. Thrall, An Examination Of
      1982-83 Participate Hatter Ratios And Their Use In The EstimatiorTOT"
      PMtn NAAQS Attainment Status, EPA 450/4-85-010.  U.S. Environmentar
      Protection Agency," OAQPS, Research Triangle Park, NC 27711,  August, 1985.

  3.  PEDCo Environmental, Inc., Esti mat ing PHiQ And FP Background Concentrations
      From TSP And Other Measurements, EPA 450/4-84-021.  U~.S.  Environmental
      Protection Agency, OAQPS, Research Triangle Park, NC 27711,  July 1984.
                                     D-9

-------

-------
                                   TECHNICAL REPORT DATA
                            fflecif rrad Insfruciions tin tht rtversr btfore coiripletingi
     RT NC
  EPA-450/2-86-001
                              |2
|3, Sf ORIENT'S ACCESSION NO.
[   PM,« SIP  Development Guideline
                                                            6. PERFORMING ORGANIZATION CODI
                                                               June 1987
                                                            6. PiRFORMIN'G ORGANIZATION REPORT NO.
   'EPPC^MiKC ORGANIZATION NAME AND ADDRESS

   Control  Programs Development Division  (MD-15)
   Office of  Air Quality Planning and Standards
   Research Triangle Park, 1C  27711
                                                            10. PROGRAM ELEMENT NO.
                                                            11. CONTRACT'GRANT NO.
1J. S?ONSOR'N3 AGENCY NAME AND A3DRESS
                                                             13. TYPE OF HEPORT AND PERIOD CQVf BED
                                                            14. SPONSORING AQENCV CODE
15, SUP'wEV.EN.TARY NC7SS
             This guideline document was  prepared to briefly describe the actions
         that must be taken by  State and  local  air pollution control agencies to
         develop SIP's that demonstrate attainment and maintenance of the PMiQ  NAAQS.
         The guideline describes  how to:  (1)  demonstrate attainment of the  PMio NAAQS,
         (2) determine the size of  an  area  exceeding the NAAQS, (3) select  a receptor
         or dispersion model, (4) prepare an  emission inventory and, (5) determine  a
         control strategy design  concentration.  The guideline also discusses ambient
         PM}Q monitoring requirements  and data  usage and EPA's policies for making
         the transition from a  SIP  designed to  protect TSP standards to one designed
         to protect PMjo standards.  References to supporting documents are given at
         the end of each section.

              This guideline does not  discuss implementation of PSD or preconstruction
         review programs for PNio-  Guidance  for implementation of these programs
         will be issued separately.
                                                 ~ = .-. T : * = E= r
                                                                            CSSA-;
                                                                                   ..« M.>_r
   PM10
   State  Implementation Plan
   Emission inventory
   Ambient monitoring
   Receptor models
   Dispersion  models
   Particulate regulations
                              (SIP)
                                                                             ,___S

-------
83 5?

Q* f>
      S3 3-

      Ill
        q en
       "22
         1
      ffl — -TJ
      II

-------