EPA-450/4-74-012
  September 1974
  (OAQPS No. 1.2-030)
       GUIDELINES FOR AIR QUALITY
  MAINTENANCE PLANNING AND ANALYSIS
                 VOLUME 11 :
          AIR QUALITY MONITORING
             AND DATA ANALYSIS
mmmmmm

mmlmmg

                                            mm
wr-
   U.S. ENVIRONMENTAL PROTECTION AGENCY
      Office of Air and Waste Management
 jiV  Office of Air Quality Planning and Standards
<$), Research Triangle Park, North Carolina 27711


-------
                                        EPA-450/4-74-012

                                      (OAQPS No.  1.2-030)
      GUIDELINES FOR  AIR QUALITY

MAINTENANCE  PLANNING  AND ANALYSIS

                  VOLUME 11 :

         AIR QUALITY MONITORING

             AND  DATA ANALYSIS
                        Prepared by

                     the GCA Corporation
                   in partial fulfillment of
              Task Order No. 1, Contract No.  68-02-1478

                   Program Element No. 2AH137
                 ENVIRONMENTAL PROTECTION AGENCY
                Office of Air and Waste Management
             Office of Air Quality Planning and Standards
             Research Triangle Park, North Carolina 27711
                      September 1974

-------
                         OAQPS GUIDELINE SERIES

The guideline series of reports is being issued by the Office of Air Quality
Planning and Standards (OAQPS) to provide information to state and local
air pollution control agencies;  for example, to provide guidance on the
acquisition and processing of air  quality data and on  the planning and
analysis requisite for the maintenance of air quality.  Reports published in
this series will be available - as supplies permit - from the Air Pollution
Technical Information Center,  Research Triangle Park, North Carolina
27711; or, for  a nominal fee, from the National Technical Information Ser-
vice,  5285 Port Royal Road, Springfield, Virginia 22151.
This report was furnished to the Environmental Protection Agency by
the GCA Corporation, Bedford, Massachusetts, in partial fulfillment of
Task Order No. 1,  Contract Number 68-02-1478.  Prior to final preparation,
the report underwent extensive review and editing by the Environmental
Protection Agency and other concerned organizations.  The contents
reflect current Agency thinking and are subject to clarification, procedural
change, and other minor modification prior to condensation for inclusion in
Requirements for Preparation, Adoption, and Submittal of Implementation
Plans (40 CFR"Part 51) .
                     Publication No.  EPA-450/4-74-012
                       (OAQPS Guideline No .  1.2-030)

-------
                            FOREWORD

    This document is the eleventh in a series comprising Guidelines for Air
 Quality Maintenance Planning and Analysis.  The intent of the seriesTsTto
 provide State and  local agencies with information and guidance for the prepa-
 ration of Air Quality Maintenance Plans required under 40 CFR 51. The volumes
 in this series are:

    Volume 1:   Designation of Air Quality Maintenance Areas
    Volume 2j_   Plan Preparation
    Volume 3^_   Control Strategies
    Volume 4:   Land Use and Transportation Consideration
    Volume 5j_   Case Studies ir^ Plan Development
    Vokjme 6^   Overview of Air Quajjty Maintenance Area Analysis
    Volume 7:   Projecting County Emissions
    Volume 8:   Computer-Assisted Area Source Emissions Gridding
                Procedure
    Volume 9:   Evaluating [ndirect Sources'
    Volume 10:   Reviewing New Stationary Sources
    Volume II:   Air Quality Monitoring and  Data~Analysis
    Volume 12:   Applying Atmospheric Simulation Models toA[r Quajjty
                Maintenance Areas

    Additional volumes may be issued.

    All  references  to 40 CFR Part 51  in this document are to the regulations
as amended through July  1974.
                                   in

-------
                                PREFACE

The Monitoring and Data Analysis Division of the Office of Air Quality
Planning and Standards has prepared this report entitled "Guidelines
for AQM Planning and Analysis, Air Quality Monitoring and Data Analysis
(Vol. 11)," for use by the Regional Offices of the Environmental
Protection Agency and by State and local air pollution control agencies,
This report draws upon the information contained in the previously
issued air quality monitoring documents listed below and applies them
toward air quality maintenance planning:
Guideline No.
   1.2-008
   1.2-012
   1.2-013
   1.2-014
   1.2-015
   1.2-018
   1.2-019
   Date          Title
Aug. 1974     Guidelines for the Interpretation
              Qf Air Quality Standards
Jan. 1974     Guidance for Air Quality Monitoring
              Network Design and Instrument Siting
              (Draft revision)
May 1974      Procedures for Flow and Auditing of Air
              Quality Data
Feb. 1974     Guidelines for the Evaluation of Air
              Quality Trends
Feb. 1974     Guidelines for the Evaluation of Air
              Quality Data
May 1974      Designation of Unacceptable Analytical
              Methods of Measurement for Criteria
              Pollutants
July 1974     Air Quality Monitoring Site Description
              Guideline
Adherence to the guidance presented in this report will, hopefully,
lead to acquisition of more useable and mutually compatible data by all
States and Regions and will also facilitate the implementation of State
air monitoring programs which are compatible with the goal of air
quality maintenance.
                               iv

-------
                             CONTENTS
List of Figures
List of Tables

Section
  11.1
  11.2
  11.3
  11.4
  11.5
  Appendix A
Introduction
Network Configuration
Sampling Instrumentation Selection
Data Processing and Summarization
Data Analysis
Statistical Tables
                                                 Page
                                                  vi
                                                  vii
  1
  8
 42
 49
 73
101

-------
                            LIST OF FIGURES

—                                                              Page
 1      Estimated distance from an elevated point source to       30
        the maximum ground level concentration

 2      Graphs of seasonal patterns for various pollutants        61
        at a particular site

 3      Frequency distribution - TSP (Philadelphia-1969)          64

 4      Frequency distribution - log of TSP data                  65
        (Philadelphia-1969)

 5      Eight-hour average violations determined by counting      80
        procedure

 6      Eight-hour average violations as highest non-over-        80
        lapping values

 7a,b   Subtleties involved in using non-overlapping values       81

 8      Suspended particulate data from Tucson, Arizona           87

 9      Suspended particulate data from New Haven,                 97
        Connecticut

10      Quality control approach applied to particulate
        data from Denver,  Colorado                                99
                              vi

-------
                            LIST OF TABLES
No.                                                              Page
 1      1972 Pollutant-Method-Stations Summary                    45
 2      Suggested Reporting Accuracy for Raw Data                 52
 3      Minimum Detectable Limits for Selected Measurement        56
        Techniques
 4      Number of Hours Above Oxidant Standard by Month and       62
        Time of Day (1971 Data)
 5      Maximum and Second High  Values (Philadelphia-1969)        68
        for Various Sampling Schemes
 6      Geometric Means, Medians, and 90th Percentile Values       69
        for Sampling Data of Table 5.
 7      Probability of Selecting Two or More Days When            77
        Site is Above Standard
                               vii

-------
                             SECTION 11.1
                             INTRODUCTION

The purpose of this document is to provide the states with planning in-
formation and preliminary guidance for the preparation and implementa-
tion of a monitoring system compatible with the goal of air quality
maintenance and the need for the development of Air Quality Maintenance
Plans required under 40 CFR 51 as amended on June 18, 1973.  In order
to satisfy this objective, background information on the air quality
maintenance planning process, as well as summary descriptions of cur-
rent monitoring techniques, is included herein.  However, the major
thrust of this document is to provide instruction on monitoring to
the state and local agencies that are charged with the maintenance of
air quality.

General guidelines have been issued which are intended to provide  a
basic rationale for the development and evaluation of air monitoring
networks.  These earlier guidelines are largely  subjective but do
provide a good understanding of  the generally applicable considera-
tions.  Much  of the guidance provided by  these earlier publications  is
repeated herein.  Areas where new guidance  is presented  include:   loca-
tion of samplers, emphasis of  the monitoring network, and  data analysis

The development and implementation of a network  must by  necessity  in-
volve a trade-off between what  is considered desirable  from  a  strictly
technical point of view  and what is  feasible with  the available  re-
sources.  An ideal network will, in  almost  all instances,  require  more
resources  than are currently  available.   Since air  quality maintenance

-------
planning is a continuing, long-range activity, this document encourages
the development of the "ideal" network.  While the implementation of
such a network would not, in most cases, be possible immediately, it
is felt that the time scales involved allow for the eventual develop-
ment of a network which fully meets the needs of air quality maintenance
planning.  This document does not address the development of interim
networks due to the wide variations of situations which occur in main-
tenance planning; however it is felt that this document provides suf-
ficient background information on the needs of air quality maintenance
planning to establish the basis for making judgemental decisions for
the most appropriate interim network.  The basic difference between the
interim monitoring network and the ideal is that the interim network
has fewer and perhaps less sophisticated instruments.  Designers of
the network should attempt to maximize the effectiveness of the interim
network through careful selection of sampling sites, scheduling of
variable sampling frequencies, and possible use of mechanical (inte-
grated) as well as automatic (continuous) samplers.  Careful considera-
tion of the basic requirements of the monitoring network will often
allow for the selection of one sampling site which may satisfy more
than one need of the maintenance planning activities.

As with any guideline document, it is expected that the procedures pre-
sented will aid the goal of continuity and uniformity in data gathering
and handling.  It is the nature of the scope of this document that this
uniformity is applicable to not only the data which is normally col-
lected through a monitoring network, but also the handling and inter-
pretation of data for the use of air quality maintenance planning.
Again, since air quality maintenance planning is a continuing process,
this document more heavily emphasizes these goals.

11.1.1  MONITORING IN GENERAL

The development of an air quality monitoring network includes determ-
ining the number and location of sampling sites, selecting appropriate

-------
instrumentation,  determining frequency of sampling,  and following estab-
lished instrument siting criteria.   Experience and judgment are essential
in this process,  especially for determining the number and location of
sampling sites because mathematical models or other methods may not be
entirely reliable or, in some instances, may not be available.  In
addition, the lack of necessary resources often requires the esta-
blishment of a minimum network which is based upon this judgmental
expertise to best cover the monitoring needs under these constraints.
General guidelines for the development of an appropriate network are
given in  Guidance for Air Quality Monitoring Network Design and
Instrument Siting  (OAQPS No. 1.2 - 012).

The selection of the methods for measurement of air pollutants requires
less judgement.  Officially approved methods (federal reference me-
thods) are described in appendices to 40 CFR Part 50.  In addition,
any method which can be demonstrated to be "equivalent" to the reference
method may be approved.  Further guidance is provided through the
 Designation of Unacceptable Analytical Methods of Measurement for
Criteria Pollutants  (OAQPS No. 1.2 - 018).

Specific procedures have also been established for the processing of
air quality data for input to the National Aerometric Data Bank.  The
guidelines for data handling,  Procedures for Flow and Auditing of
Air Quality Data  (OAQPS No. 1.2 - 013), provide information and sug-
gestions concerning data flow, editing, validation, correction pro-
cedures and certification, verification, and statistical flagging
techniques.

The analysis of the data provided by the air quality monitoring network
is less structured due, in part, to the varied analytical and statis-
tical techniques which are available for the interpretation of the
data.  Such techniques range from the basic conventions for handling
air quality data, summarizing it, determining characteristic patterns
and making inferences from the data, as presented in  Guidelines for

-------
 the Evaluation of Air Quality Data  (OAQPS No.  1.2  -  015),  to the ac-
 tual assessing and interpreting  of trends  based on  statistical  metho-
 dologies,  as  given in  Guidelines  for  the  Evaluation  of Air Quality
 Trends  (OAQPS No.  1.2 -  014).

 These four basic  concerns of  any air quality monitoring network -
 network design, instrument selection,  data processing,  and  data anal-
 ysis - are separately discussed  in this  document with regard  to the
 needs for  air quality maintenance  planning.  While  these concerns
 already have  some  basis due to their prior implementation,  it is
 necessary  to  have  an  understanding of  the  activities  involved in  air
 quality maintenance planning  and the likely impact  these activities
 will have  on  monitoring needs.  A  comprehensive understanding of  the
 considerations involved can be derived from reviewing the primary
 documents  in  this  series.  A  summary of  these considerations  is given
 below.

 11.1.2  MONITORING FOR AIR QUALITY MAINTENANCE

 11,1.2.1  The Air Quality  Maintenance Planning Process

All  states, pursuant  to 40 CFR51.12(e), are required to identify areas
 that have  the potential for exceeding any National Ambient Air Quality
 Standard (NAAQS) due  to the current air quality and/or  the projected
 air  quality due to growth over the subsequent 10-year period.  Such
 areas  are called Air Quality Maintenance Areas (AQMA's), and may be
 identical  to counties, urban areas, Standard Metropolitan Statistical
Areas  (SMSA's) or other boundaries.  At 5-year intervals, the area
identification shall be reassessed  to determine if additional areas
should be designated as AQMA's.  EPA reviews the information supplied
by the States and the Administrator issues an official list of-the
designated AQMA's.

-------
For each area identified by the Administrator, the State must submit an
analysis of the impact on air quality of the projected growth and
development over the subsequent 10-year period.  In addition, the State
must prepare and submit an Air Quality Maintenance Plan (AQMP) which
demonstrates that, within each AQMA, the national standards will be
prevented from being exceeded over the 10-year period from the date
of plan submittal.  Such a plan is considered a revision to the state
implementation plan and includes any control strategy revisions and/or
other measures that may be necessary to insure that the projected
growth and development will be compatible with the maintenance of the
national standards throughout the 10-year period.  The States are also
required to review and revise the plans where necessary at 5-year
intervals.

11.1.2.2  Particular Emphases of Monitoring

Throughout the maintenance planning process the emphasis is on determ-
ining and projecting air quality.  Therefore, monitoring, which
provides hard data on the actual air quality, is important in providing
the basis from which projections are made and models are calibrated.
In addition, the data from the monitoring network will be used during
the preparation and implementation of the AQMP to assess the impact
of the control strategies aimed at controlling the growth of emissions
and therefore the degradation of the air quality.

With regard to air quality maintenance activities, the first task of
the States must be the designation of the AQMA's.  As this responsi-
bility will have been performed for the first time as of the publica-
tion of this document, these guidelines concentrate on the establish-
ment of a monitoring network useful for the identification of new
AQMA's when considering the designation of AQMA's.  Basically, such
a system is directed at the determination of air quality trends.
Since the development of  an adequate monitoring network for an

-------
area which is not currently an AQMA requires much the same type of
analysis as for an AQMA for which a maintenance plan has been developed,
and therefore the analysis completed, the order of presentation in
this document is first those areas which are not currently AQMA's but
may be considered "potential AQMA's" and then those areas which are
designated AQMA's.  It is felt that this format allows for a better
understanding of the considerations involved with minimum repetition.

Once an area has been designated as an AQMA and a control strategy
has been implemented for the pollutant(s) of interest, the air quality
monitoring network necessarily shifts from establishing the trend,
which leads to the plan preparation, to the assessment of the impact of
the control strategy.  This provides the basis from which the determina-
tion of the achievement of the goals of the AQMP can be made.  This
allows for more stringent control measures to be developed and im-
plemented if the need arises.  In addition, the air quality data serves
as the basis for the review of new stationary and indirect sources.

11.1.2.3  General Changes in Monitoring Due to Maintenance Planning

Several changes in the monitoring network are assumed to occur as a
result of the need for monitoring for air quality maintenance planning.
Generally, these changes are directed at the size of the network and
the location of the instruments.  The extent of these changes will
depend upon how well the currently established network fulfills the
criteria presented in this document.

The size of the network and location of the instruments for main-
tenance planning are naturally interrelated.  Given the special con-
cern which AQMA's have within the overall implementation planning
mechanism, the data required, and hence the monitoring necessary, may
be significantly more than is now operational or even planned in a given
maintenance area.  Air quality maintenance planning requires that certain
areas, especially controlled or encouraged growth areas, be monitored  to
establish the trend of air quality.  Other sites which may be of special
interest in the maintenance plan deserve monitoring.
                                  6

-------
The other major change in monitoring due to air quality maintenance
activities involves the analysis of the data.  More attention needs to
be directed at the reasons for variations or trends in the data, es-
pecially as they relate to other parameters.  Essentially, many of the
same techniques as currently employed will continued to be used with
either an increase or change in emphasis.

Other areas which are stressed in this document are the result of the
long-term nature of air quality maintenance planning.  Such planning
allows for more standardization of monitoring networks, data handling,
and data analysis.  Of particular interest is the phasing in of new
standard equipment and the more complete use of improved data handling
techniques.

-------
                             SECTION 11.2
                         NETWORK CONFIGURATION

The configuration of an air quality monitoring network involves two dis-
tinct elements:  the number of sampling sites of various types, and their
geographical location.  Under varying circumstances, decisions on the
two elements can be made in either order; an overall number of sensors
or sites may be selected, based on a criterion such as resource availa-
bility, and then distributed geographically, or specific sites may be
selected first, based on a criterion such as the need for the data, with
the aggregate number of sites then being just the total number of sites
selected.

Historically, as the national control effort under the Clean Air Act
began, emphasis was on developing not only new networks but also the
resources, both human and monetary, to operate them.  Consequently, the
first approach was necessarily taken; networks were sized directly or
indirectly in relation to the resources available, and the sites were
distributed with as much consideration of sources, topography, etc., as
was possible.  These monitoring networks, were then, and still are
being expanded or modified in the light of increasing knowledge, ex-
perience and funding.

However, in the case of planning an air quality monitoring network for
purposes of long-term maintenance of ambient air quality standards, the
emphases are necessarily different and the second approach is the more
appropriate one.  Rather than being primarily concerned with resource
limitations, maintenance network planning is involved with a long-term,

-------
on-going, network configuration, and resource concerns will affect pri-
marily the length of time required for the network to evolve into its
ultimate configuration.  Rather than being constrained by knowledge
deficiencies, pollution control officials can now understand with some
degree of precision what decisions they will be faced with in the future
and what air quality data will be needed to assist in making those
decisions.  Consequently, the emphasis in this section will be on
designing a network configuration to meet specific air quality main-
tenance needs, and the aggregate size of the monitoring effort will be
determined as a result of meeting the various specific needs.

11.2.1  SCALES OF SENSOR PLACEMENT

Determining the location of air quality.monitoring sites, including the
location of the sensor or sampling probe itself, is a decision process
that ranges over a very wide range of geographical scales.  In the past,
with air quality monitoring networks existing primarily in urban areas,
the decisions involved could be conveniently summarized by defining
three selection processes, each of which operates in a different size
scale, and each of which depends on somewhat different criteria:
    •  Neighborhood Selection - the choice of the areas within
       the city where sites should be located - operating in a
       scale of many blocks or a few kilometers, and primarily
       determined by data needs.
    •  Site Selection - the choice of a specific site within
       the neighborhood to be monitored - operating in a
       scale of tens of meters, and largely determined by
       questions of accessibility, security, etc.
    •  Probe location - the selection of the precise loca-
       tion of the sensor or inlet probe - operating in a
       scale of a few meters or smaller, and determined by
       concerns of sample representativeness and standardiza-
       tion.
These three categories are derived largely from concerns with urban net-
works; monitoring for maintenance purposes is of course the same in many
respects, but involves the addition of one larger-scale decision category

-------
which has been labeled Community Selection.  This is meant to involve
the selection of, from among all areas in an AQCR, those portions where
monitoring is required for air quality maintenance purposes.  While in
some AQCR's, this may be no more than the control urban area, in others
it may, in addition, involve outlying growth and recreational areas,
areas where major new industrial or resource development activity is
expected, and other areas of non-traditional concern.

These four categories form the framework for the discussion of designing
a network configuration for air quality maintenance purposes.  Community
selection and neighborhood selection, which are primarily dependent on
data and information needs, are considered rather thoroughly in the next
two subsections.  The areas of site selection and probe placement, which
are less a function of data needs, are then considered somewhat more
generally.

11.2.2  COMMUNITY SEIECTION

The selection of communities for increased monitoring attention is
basically a function of the growth of the area, whether it exists in a
potential or a designated AQMA.  The task, however, is much easier for
designated AQMA's as most of the analysis of growth has already been
done and the development of the maintenance measures has identified
those communities which will be the primary target of the AQMP.  For
the purpose of this discussion, the division between potential and
designated AQMA's is made to allow for this different degree of required
analysis as well as providing a logical separation of the two activities
which may have varying priorities to different states.

11.2.2.1  Potential AQMA's

After the designation of AQMA's at each 5-year interval,a review of those
areas within each state which were not so designated should be made.  This
review would identify those other areas which, while not expected to
                                 10

-------
exceed the NAAQS within the subsequent 10-year period, have the potential
for violating one of the standards within 15 years.  Such a determination
would be based upon similar growth projections as used in designating the
AQMA's except that the time period under evaluation would be 15 instead of
10 years.  Ample consideration should also be given to those natural re-
source areas which, while not currently being exploited perhaps due to
economic reasons, may be opened up when social or economic constraints are
removed.

A region may be determined to be a potential air quality maintenance area
if the projections of growth indicate that a national air quality stan-
dard will be violated within the area by 1990.  For the purpose of this
discussion, the term "potential AQMA" refers to not only those areas
which have not yet been designated AQMA's for the current 10-year main-
tenance planning work but also any presently designated AQMA which may
exceed a standard by 1990 for which pollutant it is not now designated.

Initial Selection of Potential AQMA's - Criteria for  the initial designa-
tion of areas as potential AQMA's are whether the recent analysis for
AQMA designation indicated that the area will be nearing the standard
within 10 years, the area is currently undergoing a high growth rate which
is expected to continue, the area is expected to undergo rapid expansion
after the current 10-year period, or new and major development, either
residential or industrial, is forecasted to occur after 1985.  The re-
source development areas in  the western states come under the  latter
criterion.

In addition, it is possible  to establish boundary conditions which can
serve to automatically exclude from consideration as  a potential AQMA
a particular area.  Basically, such criteria for exclusion are functions
of the current regulations controlling emissions of individual pollutants
as well as expectations of normal growth rates.  By pollutant, these
exclusion criteria are:
                                 11

-------
     1.  Particulate matter:  any standard metropolitan statis-
         tical area (SMSA) undergoing normal growth which is
         not projected to exceed any of the particulate NAAQ's
         in 1985.
     2.  Sulfur dioxide: any SMSA undergoing normal growth,
         not currently an AQMA for SCL.
     3.  Carbon Monoxide:  any SMSA undergoing normal growth.
     4.  Photochemical oxidants:  any SMSA undergoing normal
         growth.
     5.  Nitrogen dioxide:  any SMSA undergoing normal growth.
The SMSA's were chosen above for providing means for the elimination
from concern of those areas that are already developed, that is areas
with already defined industrial development characteristics.  New
growth in areas not previously monitored is more likely to cause pro-
blems of maintenance than growth in SMSA's where new, controlled sources
will be replacing older, less controlled ones.  In addition, the emis-
sions from automobiles are generally expected to be leveling off
around the 1985 to 1990 period with little if any increase by 1990
under normal growth conditions so that any SMSA which was not already
under a maintenance plan for the motor vehicle pollutants for 1975 to
1985 will not be providing a maintenance problem for 1990.

The SMSA, alone or in its entirety may serve as a desirable geographic
area for designation as potential AQMA.  However, it is not always the
area which may be most applicable.  For instance, projections of emis-
sions for cities, counties or townships within the SMSA may be possible
to calculate, in which case it would be desirable to designate these
sub-SMSA areas.  In other cases, the projected growth in emissions may
be expected to occur around the fringe of the SMSA, in which case the
designation may be more desirable if it included that fringe area with
or without the SMSA in whole or in part.  Many of the potential AQMA's
                                12

-------
will be found to be areas completely apart from any SMSA's, occurring
in the more rural areas where natural resources are to be developed.

In deciding upon the actual area to be designated a potential AQMA some
consideration should be given to the difficulty of the eventual manage-
ment of control programs within an AQMA.  It is easier to designate by
the names of the existing areas (political or non-political) than to
delineate an area by listing roads, rivers, other topographical features,
or latitude-longitude coordinates which constitute the boundaries of
the area.  Designation by currently defined areas, however, does not mean
that the subsequent detailed analysis of the potential AQMA and possible
monitoring network must apply to the entire AQMA as originally designated-
the analysis and network  could be restricted to selected problem areas
within the AQMA.

In addition, one should be aware of possible relationships between the
potential AQMA's and the areas to be chosen under the forthcoming
regulations concerning significant deterioration.  For instance, if the
significant deterioration regulations provide that some (probably urban)
areas are permitted to deteriorate up to the secondary national ambient
air quality standard, these areas will probably be the same areas as
the potential AQMA's.  Therefore, it might be appropriate to designate
an area large enough to allow for the proper amount of desired growth.

In deciding upon the particular boundaries of a  potential  AQMA,  the
following factors  should  be  considered.
         The potential AQMA should include all of the territory
         which shares a common air envelope and a common aggrega-
         tion of sources.  This will usually be an urbanized area
         plus some adjoining areas which are now undeveloped but
         which are expected to develop in the next 10 years or so.
         It may include satellite communities which are now
         separated from the central urbanized area but will, in
         10-20 years, become part of the central urbanized area
         and thus share the air resource.
                                13

-------
     2.  Use of areas previously designated by agencies of various
         kinds may have merit in that a data base may be available
         and a proliferation of "regions" can be avoided.   Examples
         are regional planning areas; State designated planning
         areas; transportation planning areas; etc.

     3.  Emission control and other air conservation measures neces-
         sary to maintain air quality standards in the urbanized and
         developing parts of major urban centers may be quite strin-
         gent.  Application of such stringent measures in isolated
         or undeveloped areas may not be advantageous.  Thus, in-
         clusion of large rural areas in a potential AQMA may not be
         desirable.

     4.  Design and implementation of air conservation measures will
         involve certain governmental agencies.  Common boundary
         lines for potential AQMA's and one or some  combination of
         jurisdictional areas of implementing agencies may have merit
         from an operational point of view.

     5.  Long-range transport of pollutants is a matter of concern.
         It is also true that if ambient air standards are maintained
         near an aggregation of sources, such standards will also
         usually be maintained at more distant locations.   Therefore,
         it may not be necessary to include those areas on the
         periphery of an aggregation of sources in order to assure
         maintenance of standards at locations distant from the
         aggregation of sources.

     6.  The influence of topography and geography on dispersion of
         pollutants and on overall community growth  patterns should
         be considered.

     7.  When selecting potential AQMA's, preparation of air
         quality projections and development of any needed monitoring
         networks will need to be based on presently available land
         use, transportation and other plans because ot time con-
         straints.  It may be, however, that new general regional
         development plans will be prepared in the future because of
         air quality considerations or other reasons.  The potential
         AQMA selection would desirably be compatable with any such
         future community planning activity.
A non-exhaustive list of types of areas which might, be used for designa-

tion include:


     AQCR's

     SMSA's


                                 14

-------
     Urbanized Areas
     Communities
                      Cities
     Groupings of:   ^ Townships
                      Boroughs

Potential AQMA Analysis for Community Selection - The decision for com-
munity selection within the potential AQMA may be on a best guess basis
depending upon the guidelines presented later; however, for a more ef-
fective network it is recommended that an indepth investigation of the
projected air quality situation be conducted.  Descriptive analysis for
selecting communities would proceed along the general lines described
below concerning analytical procedures.  The tasks to be performed here
have a different purpose than those outlined above where it was only nec-
essary to identify potential AQMA's on the basis of non-specific designa-
tion criteria.  These tasks go beyond that and quantitatively evaluate the
air pollution problem in each potential AQMA for the period 1975 to 1990.
The tasks are:

     1.  Determine baseline emissions for each pollutant for which
         the potential AQMA was designated
         a.  By source category
         b.  By location as required by EPA models.
     2.  Identify principal sources (baseline and projected to 1990).
     3.  Acquire all necessary data to determine growth in emissions
         from 1975 to 1990 by source category and location for each
         pollutant.  This would involve acquiring data on:
         a.  Past trends
         b.  Planned and projected economic and demographic growth
         c.  Projected control technology
         d.  Present and future regulations for new and existing sources
         e.  Expected industrial and residential development.
                                 15

-------
     4.  Project a detailed emission inventory for 1980 to 1990 by
         source category for each pollutant.
     5.  Project 1980 to 1990 air quality using calibrated diffusion
         models as provided by EPA.  Use these models to:
         a.  Analyze the impact of indirect sources
         b.  Analyze the impact of new sources.
     6.  Determine which parts of the potential AQMA's are problem
         areas and require increased monitoring attention.  (A
         problem area is any portion of potential AQMA where the
         above analysis indicates any standard may be violated at
         any time between 1980 and 1990.  If the standard  is violated
         by 1985, the area should be designated as a real AQMA and
         an air quality maintenance plan should be prepared for it.)

The models, emission factors, growth projection techniques, etc., suitable
for performing the needed analysis can be found in the respective guide-
line documents listed in the bibliography.  Guidance on the analysis, as
well as much of the data base and methodology applicable to each state,
would also be available from the personnel involved in the current ac-
tivities relating to the designation of AQMA's and the preparation of the
air quality maintenance plans therefor.  A short review of the considera-
tions involved are presented below.

Emission Baseline - While air quality is the final determining factor for
the selection of the community, the emission inventory, and changes there-
in, is the only real base from which the air quality can be projected.  In
order to estimate emissions between the time standards are attained and
1990, it is necessary to determine emissions at the time that standards are
attained.  Some state implementation plans (SIP*s) contain these projections
of emissions and these can be used where available.  If not available,
these attainment date emissions can be calculated by the method presented
in the Manual for Analysis of State Implementation Plan Progress.  Regula-
tions which are currently in existence should be used to project emissions.
NEDS emission data should be used as the basis for these emissions pro-
jections.  Other emissions data that may be locally available may be used
                                 16

-------
in lieu of or in addition to the NEDS data; such data must be entered into the
NEDS at the next semiannual update.  The update is to be accomplished as
prescribed in APTD-1135, Guide for Compiling a Comprehensive Emissions
Inventory.

Emission Projections - Using best judgement, constrained by data avail-
ability limitations, demographic-economic  indicators would be selected  as
representative of each pollutant-source category combination.  Normally
these would be population, employment, and earnings by industrial category.
Percentage growth rates for these indicators are determined for the  inter-
val  of interest, baseline year to 1990, for application to the corre-
sponding pollutant-source category.  Detailed  instructions for the develop-
ment of growth factors  and the projection  methodology are contained  in  the
EPA  guideline document, Projecting County  Emissions  (Volume  7 of this
series).  For the purpose of  community  selection  from potential AQMA's,
the  Level 1  approach presented in  this  document would be  appropriate.
Should resources permit, however,  the projections  could be refined
using the Level 2 or Level  3  approach.

Air  Quality  Baseline -  Several of  the models  presented below for use in
predicting air  quality  require the use  of  air quality at  the time of im-
plementation of existing  regulations.   It  is  the  major purpose of this
review  to define an adequate  monitoring network so that  later evaluation
 of the  area  for actual  AQMP development will  be on a firmer  air  quality
basis.   For  current evaluation  of  potential AQMA's,  however, certain
 assumptions  may be  necessary. As  with  emissions,  the  SIP's  may  contain
 projections  of  air  quality at the  time  of  full SIP implementation,  and
 these air quality  values  can  be  used.   For cases where  air quality  pro-
 jections are not  contained in the  SIP,  it  may be assumed that the NAAQS
 will not be  exceeded.   Alternatively,  recent  (1972-1973)  air quality data
 may be projected  to 1975 and  hence to 1990, making proper adjustments for
 growth and scheduled abatement actions.
                                  17

-------
Air Quality Projections - The projected emissions are now used to drive
an air quality projection model to estimated 1980 and 1985 air quality.
The following are types of models which may be used for the pollutant
indicated:

     •   Particulate Matter - Atmospheric Diffusion Model
     •   Sulfur Oxides      - Atmospheric Diffusion Model
     •   Nitrogen Oxides    - Appendix J Relationships (Rollback
                              Model for Texas and California AQMA's)
     •   Hydrocarbons       - Appendix J Relationships (Rollback Model
                              for Texas and California)
     •   Carbon Monoxide    - Supplements to the Indirect Source Review
                              guidelines (Volume 9 of this series) for
                              new sources;  rollback model described in
                              the AQMA designation guideline (Volume 1)
                              for existing  sources.

Acceptable models for projecting air quality are described in Volume 12
of this guideline document, Applying Atmospheric Simulation Models to Air
Quality Maintenance Areas^ .  Locally available models may also be used.

Community Selection - Each portion of a potential AQMA which is projected
to exceed one of the national ambient air quality standards by 1990 should
be considered as a community in which an evaluation of the monitoring net-
work is needed.  Different portions which may have similar causes for the
violation and are adjacent may be considered as one community.  If, how-
ever, the pollutants vary or the causes of the high concentrations are
different, they should be individually considered as communities.

If the projected air quality would not be adequately measured and/or
represented by the current monitoring network, the network should be
revised and/or expanded to provide the coverage needed.  The implementa-
tion of the revised network should occur at least three years before the
development of the next set of AQMP's so that the basis for the air quality
projections will be well founded.  Considerations appropriate to the review

                                 18

-------
of the network  include whether the spatial configuration of the monitors
would be appropriate to that of the pollution and whether the type(s) of
monitors are compatible with the pollutant(s) under consideration.
The following national ambient air quality standards should be considered
in designating communities in which standards may be exceeded:
    Pollutant
           Primary
     Secondary
Particulate matter   (a)
Sulfur dioxide
Carbon monoxide
Photochemical
oxidants
Nitrogen dioxide
     75 p,g/m3, annual
     geometric mean
             n
     260 (Jg/m ,  second highest
     24-hr average per year
     80 (ig/m^, annual
     arithmetic mean
     365 ^g/m »  second highest
     24-hr average per year
10 mg/m3, second highest 8-hour
(b)
(a)

(b)
 150 ug/m3, second
 highest 24-hr
 average per year
 1300 |ig/m3, second
 highest 3-hour
 average per year
average per year
160 p.g/m3, second highest 1-hour average per year
100 |j.g/m3, annual arithmetic average
For carbon monoxide, assume that the 1-hour standard will be maintained if
the 8-hour standard is maintained.  As in the original SIP's, a demonstra-
tion of achieving the oxidant standard will imply that the hydrocarbon
standard is also achieved.

11.2.2.2  Designated AQMA's

As areas currently designated as air quality maintenance areas are already
to be receiving extensive analysis of the projected air quality situation
for the development of applicable control measures, the degree of review,
as discussed above for potential AQMA's, will not be necessary solely for
the basis for the establishment of a monitoring network, especially the
community selection thereof.  In addition, the purpose of the monitoring
network for designated AQMA's is different then for the potential AQMA's.
                                19

-------
Whereas the monitoring networks in the potential AQMA's are directed at
establishing a sufficient data base for future air quality maintenance
planning, the design of a monitoring network in a currently designated
AQMA is contingent upon the maintenance plan itself.  The network must be
able to monitor the air quality sufficiently to assure that the control
measures implemented in the AQMP are providing the desired degree of
control.

With regard to air quality maintenance planning at the community level,
there are basically only two types of areas which relate directly to con-
trol measures and should therefore receive particular attention in network
designing:  controlled growth areas and activity centers.

Controlled Growth Areas - In the course of the preparation of the air
quality maintenance plan for a currently designated AQMA, certain por-
tions of the AQMA will be found to either be presently exceeding a NAAQS
or be projected to do so by 1985.  An appropriate control measure for
maintenance of air quality, especially for areas presently in violation,
would be to discourage or control increased growth in these areas.  Growth
may be controlled by doing some form of emission density zoning, refusal
to provide any increase in services (e.g., sewer systems), or other re-
gional planning measures.

The monitoring network in such an area should be re-evaluated to deter-
mine its overall adequacy and reliability.  Normally, a properly designed
network would already be operative in such an area.

Activity Centers - In a similar vein, many air quality maintenance plans
are expected to encourage development in other areas.  These "activity
centers" are normally proposed either as an alternative area where growth
may still occur away from the controlled growth areas or as an integral
part of a regional plan which is directed at the concept of de-centralized,
sub-compartmentialized services or a defined transportation system.
                                20

-------
The selection of such activity centers as communities deserving increased
monitoring attention naturally follows from the need to also control
growth in these areas.  Care must be exercised in promoting and distri-
buting the growth in these communities so that the standards are not
similarly exceeded in these areas.  The air quality monitored by the
community network plays a major role in this control by indicating trends
in air quality due to recent activity and by providing a proper base from
which the impact of continued development can be projected.

Other Non-Planned Communities - Most of the control measures proposed in
the maintenance plan will not be solely applicable to the general growth
and development of selected communities.  Instead, they will be generally
applicable throughout the designated AQMA (e.g., emission standards) or
major portions thereof  (e.g., transportation control measures for an
SMSA).  In addition,  some measures will apply  only to a couple of sources
in a defined area, especially in resource development areas.

In the latter case,  such defined areas are definitely communities which
should be selected for  a monitoring network or the evaluation of the
currently operating  one.  The placement of the monitors would depend pri-
marily on the sources involved and meteorological conditions, as discussed
under neighborhood selection.

For  the other areas  throughout the AQMA which  do not  come  under  specific
growth restrictions,  the  criteria for  selecting a community as  one  which
deserves  increased monitoring attention  are whether  a NAAQS was  projected
to be exceeded by  1985  before the implementation of  the  control  measures
 or  a NAAQS  is  currently projected to be  exceeded in the  area,  under the
AQMP control measures,  by 1990.   If  the  community comes  only under  the
                                21

-------
second criterion, it should be monitored as described below for a potential
AQMA.

11.2.3  NEIGHBORHOOD SELECTIONS

Neighborhood selection involves the further delineation of the network
configuration by selecting those smaller areas within a community which
are  the most deserving for the locations of the monitors.  Normally, this
selection process operates on a scale of a quarter to a few square miles
(if  the community is large or the number of samplers small) and depends
primarily on the type of data needed.

For  the purpose of monitoring for air quality maintenance, these data
needs are again different for potential AQMA's and designated AQMA's.
The  following discussion highlights these differences and discusses the
proper considerations for the selection of a neighborhood based upon
maintenance considerations.

11.2.3.1  Potential AQMA's

The primary purpose for establishing a monitoring network in a potential
AQMA is the creation of a data base adequate for the eventual analysis
required for the designation of the area as an AQMA and, most importantly,
the preparation of an air quality maintenance plan.  As the standards,
and  the plans for the maintenance thereof, revolve around the highest
concentrations measurable in an area, this becomes the determining fac-
tor in neighborhood selection.

The analysis needed for neighborhood selection in potential AQMA's is
essentially similar to that done for the selection of the community:  de-
termine base line emissions,  determine growth factors,  project emissions,
determine projected air quality.   The scale of the analysis,  completed
                                 22

-------
 for  the  community  selection,  especially  the modeling  aspects, will  de-
 termine  the  amount of  additional  effort  that  is  necessary  for the proper
 selection of the neighborhood.  If  the analysis  was completely  on a
 macroscale basis,  it will  be  necessary to  undertake the  re-evaluation of
 the  community on a mesoscale  basis  to provide a  sufficient understanding
 of the air quality situation.   If,  however, enough detail  is accounted
 for  in the preliminary analysis for community selection, it would be  pos-
 sible to use those results for  neighborhood selection.

 The  criterion for  such a distinction is  whether  the analysis provided
 estimates of concentrations on  a  scale capable of allowing the  selection
 of several high and low points  within the  community.  This may  be de-
 termined either directly from the air quality modeling or  the emission
 densities, or indirectly via  isopleths derived from one  of these two.

 Since the eventual strategy options must be chosen on the  basis of  the
 control  of standard violations, the neighborhoods selected for monitor-
 ing  sites should,  theoretically, be any  individual portions of  the  com-
 munity which are projected to exceed the standards.   This  is perhaps
 even more desirable for potential AQMA's than for existing ones due to
 the  need for eventual  strategy  formulation.   It  is expected that, more
 and  more,  the  control  measures  selected  in future air quality maintenance
 plans will be  directed at  individual problem  solving.   Instead of deter-
 mining a  control strategy which has broad  application to the whole AQMA,
 or even  major  portions thereof, individual measures will be tailored  to
 the  individual problem areas.   Therefore,  it  is necessary  to have a
 thorough  understanding of  the air quality  in  the particular problem
 area of  interest.

 General Neighborhood Selection Criteria  -  These considerations,  along
with the  need  for  an understanding of the  total air quality situation
 in a potential AQMA, provide the basis for the following general criteria
 under which  a neighborhood would be selected  for monitoring:
                                  23

-------
      1.   The neighborhood is  projected to exceed a NAAQS  by 1990;  or
      2.   The  neighborhood contains  the  highest projected popula-
          tion of  the community; or
      3.   The  neighborhood is projected  to undergo  the most  rapid
          industrial growth between  1980 and 1990;  or
      4.   The  neighborhood is projected  to be that  portion of the
          community with  the highest air quality  level below the
          standard of interest.
It must be realized that many of the above criteria rely upon knowledge
of 15-year projections of occurrences within a neighborhood.  While such
planning is rarely accomplished at this stage, some best guess estimate
of expected activity is possible.

The above criteria require that at least two neighborhoods within a
community be selected for monitoring.  The minimum network would apply
to a community where the standard is projected to be exceeded in only
one neighborhood which is also the highest, industrial growth region and
the second highest projected level occurs in the highest populated area,
or vice versa.

Pollutant Specific Neighborhood Selection Criteria - The development of
a larger than minimum network would also follow along these same general
guidelines; however, some consideration of the characteristics of the
pollutants can be used to provide more pollutant specific guidelines.
These are developed and discussed below.

Particulates - Violations of the total suspended particulate standards
are most likely to be initially occurring in isolated sections of a com-
munity due to one large source or a combination of smaller ones.  In ad-
dition, these are likely to occur in industrial areas, with a lesser
chance of the sources being from the residential sector.  As these are
individual occurrences with minimal degree of correlation among them,
the selection  of the neighborhoods for monitoring should parallel that
of the general criteria presented above.
                                 24

-------
 Carbon monoxide -  Projected  occurrences  of  carbon monoxide  standard
 violations are  likely to be  either  due to a high  growth  rate  in VMT
 in a business district or a  particular type of  development  activity
 which would contain large sources of  CO.  While indirect sources  such
 as interchanges and shopping centers  are likely to  also  allow for the
 violation of the CO standard,  such  sources  are  seldom predictable on a
 more than 10-year  basis.

 While the development activities containing new large sources of  CO
 would allow for the same  general criteria considerations as for partic-
 ulates above, the  high concentrations of CO as  a  result  of  significant
 increases in VMT are  likely  to be much more dispersed over  the general
 community and may  appropriately enough involve  several neighborhoods.
 Normally  one CO monitor placed at the highest projected  level of  CO
 would suffice;  however, in the few  cases where  the  CO standard may be
 exceeded  in an  area greater  than 25 square  kilometers (about  10 square
 miles), monitors should be added at rate of 1 + j|  where A is the area
 of the region (in  kilometers) which violates the  standard.  The actual
 selection of the neighborhoods for  the location of  the monitors should
 be done on the  basis  of optimizing  the highest  levels which would be
 measured  and the spatial  distribution of the monitors.

 Photochemical oxidants  -  The criteria for selection of neighborhoods for
 monitors  of  photochemical oxidants would also be  similar  to that  for CO
 as  this is  also  a pollutant which comes about primarily  as a result of
motor  vehicles.   For  large areas which exceed the standards, however,
 the degree of monitoring  need not be as detailed.   Current guideline
documents  for the preparation of an air quality maintenance plan allow
for the aggregation of emissions on a county-wide basis.   Therefore, for
the purpose of monitoring oxidants,  the number of monitoring sites could
        A
be 1 + —. (This relationship, as the one above, is based on judgmental
experience derived  from the development of a sample air  quality mainte-
nance plan.)  The large point sources of  hydrocarbons would receive the
same attention as given under the general criteria.
                                 25

-------
Nitrogen dioxide - The criteria for selection of neighborhoods for mon-
itors of NC>2 are similar to those for photochemical oxidants.  The
neighborhoods, and eventually the sites selected, would be generally
the same as for photochemical oxidants when monitoring NC>2 from motor
vehicles.

11.2.3.2  Designated AQMA's

The selection of neighborhoods for designated AQMA's is much more struc-
tured and requires less analysis than for potential AQMA's for the same
reasons as under the community selection process.  The primary purpose
of monitoring in a designated AQMA is to ensure that the control measures,
which were deemed necessary for the maintenance of air quality, are
having the desired effects.  Due to different land uses within the AQMA,
the impact of the control measures on the individual communities will
vary.  In addition, certain communities (e.g., activity centers, con-
trolled growth areas) will have particular measures which are applicable
only to the individual community.  Despite these variations, certain cha-
racteristics of the different types of control measures which may be
applied help to provide some delineation of the neighborhoods which should
be monitored.

For the purpose of this discussion, the control measures are divided into
three basic categories: emission regulations for stationary and indirect
sources, transportation control measures, and land-use planning.  General-
ly the discussion of  the criteria for neighborhood  selection under each
of these categories is applicable to all communities; individual  applica-
tions to a particular  type of community are presented as needed.   Since  a
major objective of current monitoring is the evaluation of  the progress
made in attaining and maintaining the desired air  quality,  sampling
stations are often already strategically situated  to  facilitate  evalua-
tion of the implemented control  tactics.  The development of  a monitoring
system  for a designated AQMA will often  then just  require the  evaluation
                                 26

-------
 of the current monitoring  system in light  of the  air  quality maintenance
 plan with changes  and  additions  made as  it appears  necessary.   As
 throughout this report,  any use  of  the criteria or  discussion thereon
 must be tempered with  the  experience and characteristics within the
 individual regions.

 Monitoring for Emission  Regulations - The  selection of  neighborhoods  for
 the  monitoring of  the  impact of  emission regulations  depend  partially
 upon the  types of  regulations imposed and  the  sources to which  these
 regulations apply.  The  types of emission  control measures which are
 listed in Volume 2 of  the  Guidelines for Air Quality  Maintenance Planning  are;
     A.  New Source Performance Standards
     B.  Revision of Existing SIP Control Measures
     C.  Phaseout or Prohibition  of  Emission  Sources
     D.  Fuel Conversion
     E.  Energy Conservation and  Utilization
     F.  Combination of Emission  Sources
     G.  Special  Operating  Conditions
     H.  Stack  Height Regulations
     I.  Control  of Fugitive  Dust  Sources
 Full  explanations  and discussions of these measures are given in Volume
 3, Control Strategies, of  the above  guidelines.

 For  the purpose  of the discussion of neighborhoods selection, it is
 useful  to group  these types  of measures into three main categories:
 controls directed at significantly large contributors (B, C, D, E, G, H,
 I) and controls which are generally applicable (A, B, E, F, I).  There
 is obviously overlap as similar types of measures can be applied to both
 large and small  sources.

The reason for the  above  division lies entirely in the need for the
 selection of a limited number of neighborhoods due to constraints on
the extent of the monitoring network.  Whereas the state or local air

                                 27

-------
pollution control agency is likely to be charged with the responsibility
of general ambient monitoring, many large sources, especially new ones,
may be required to provide their own monitoring capable of determining
the adequacy of control measures.

Significantly Large Sources - Those individual sources which received
special attention in the air quality maintenance plan due to their degree
of contribution to the total air quality burden also deserve to have
increased attention when reviewing or establishing a monitoring network.
As most individual sources make their largest contribution to the ambient
concentrations within a reasonably short distance from the stack, the
neighborhood for location of sampling sites is likely to be the one in
which the source is located or an adjoining one.   The final selection
of the neighborhood and the number of sites which should be monitored
will depend upon the meteorology of the area, the source strength, the
other sources in the area, and the population exposure potential.

These factors are best decided on a case by case basis, however, certain
general comments can be made to direct the consideration given to the
selection of the neighborhood.
    1.  Monitor(s) should be placed in the neighborhood to
        record the maximum likely concentration which will
        occur as a result of the individual source.
    2.  If another adjoining neighborhood is heavily populated
        and also receives a not insignificant contribution
        from the source of interest, a monitor should also be
        located there.
    3.  The above selection of the neighborhoods would rely
        principally upon the use of meteorological data
        to determine the most frequent wind direction and
        speed which would account for the high levels.
        The review of data should include the determination
        of possible seasonal influences on emission of
        contaminants and specific meteorological conditions.
    4.  A monitor should not be located in a neighborhood which
        would receive significant contributions from other nearby
        sources so as to confuse the interpretation of the data.

                                28

-------
    5.  The use of maximum short-term (1-24 hour) concentra-
        tions for neighborhood selection allows more validity
        than long-term concentrations due to averaging effects.
    6.  Some consideration should be given to compromising
        the optimum location of two monitors for different
        pollutants to allow for the establishment of a
        station containing the two monitors.
    7.  For isolated sources, a minimum of three sampling
        sites is suggested.  Two should be along the most
        frequent downwind direction(s) and one would be
        along the direction that is predominantly upwind.
A reasonable estimate of the distance to the maximum concentration from
the source can be derived from Figure 1.  The distance is a function of
the effective source height and atmosphere stability; effective source
height is the sum of the physical stack height and the estimate plume
rise.  The Briggs plume rise or other appropriate plume rise equations
may be used.

It should be remembered that certain sources, i.e., those under sup-
plementary control strategies, are required to perform their own moni-
toring.  In addition, many other sources, especially new indirect
sources, may be required to perform sufficient monitoring before and
after construction to demonstrate that standards are not being exceeded
as a result of the installation.

General Emission Regulation Monitoring - The selection of neighborhoods
for the location of monitors for the review of the progress  of the
maintenance plan involves decisions regarding the distribution of sam-
plers within the region and expected impact of the general emission
regulations inherent in the plan.  Such a network should consist of
stations that are situated primarily to document the trends in the high-
est pollution levels, to measure the exposure to the population, to
measure the pollution generated by the specific types of sources which
are controlled under the maintenance plan, and also to allow for the
monitoring of air quality in basically uncontrolled areas to provide a
baseline from which to understand the fluctuations in measurements.
                                 29

-------
u>
o
                                              PASOUILL STABILITY CLASSES
                                                GROUND-LEVEL CONCENTRATION,  x7/0'(m-z)
                    Figure  1.   Estimated distance  from an elevated point source  to  the maximum

                                ground level  concentration

-------
The selection of the neighborhoods which takes into account the above
consideration requires a fairly thorough understanding of the air quality
throughout the area.  It is assumed that much of this understanding will
come about as a result of the re-inventorying of sources and modeling of
air quality needed for the development of an air quality maintenance plan.
As the plan must also provide an estimate of the expected impact of the
measures formulated therein, much of the analysis for neighborhood selec-
tion will be started.  Working from this basis, it is possible to develop
a sampling network applicable to the monitoring of the progress of the
maintenance plan.

The extra degree of analysis to be applied to the definition of the net-
work depends upon the funding and time constraints for the establishment
of the network.  The extent of the network will also be contingent upon
the funding constraints.  Working within these constraints, the demonstra-
tion of the establishment of an adequate network should quantitatively or
qualitatively be able to show that the following conditions are met:
     1.  Current areas with highest concentrations are being
         monitored.
     2.  Areas projected to have the highest concentrations
         under the AQMP are being monitored.
     3.  Areas which are expected to be undergoing the most
         rapid growth are being monitored.
     4.  Sections of the AQMA which have the highest population
         density and/or total population are being monitored.
     5.  At least a few areas which are not expected to have
         concentrations significantly affected by the regula-
         tions are being monitored.  This should include at
         least one area with a currently low concentration
         (background) and one area with a moderate to high
         concentration of the pollutant of interest.
     6.  The monitors are sufficiently dispersed to provide
         overall coverage of the AQMA •
                                 31

-------
Monitoring for Transportation Measures - If the air quality maintenance
plan for the designated AQMA contains any measures for controlling the
total emissions from the transportation sector or relies on those in the
state implementation plan or a subsequent transportation control plan,
then the sampling network should also be geared to monitor their impact.
However, where the controls on the emissions are directly on the sources
themselves (e.g., FMVCP) or assume a generally applicable reduction in
VMT (e.g., gas rationing, carpooling) then the monitoring of the impact of
the measures would be done under the same conditions discussed above for
other generally applicable emission regulations.

Those measures which, on the other hand, call for reductions and increases
in VMT in particular corridors of the AQMA deserve to be monitored on an
individual basis.  Such measures include the establishment of a highway
network which may redirect the flow of traffic or a mass transit system
which would reduce the travel in particular corridors or sections of the
AQMA.   The monitoring of these controls should include not only air quality
but also travel related parameters, such as VMT, to provide an adequate
measure of the actual effect the various systems are having.  A description
of the network adequate to determine the changes in these transportation
related parameters is beyond the scope of this guideline.

The number of samplers which are needed in any particular corridor or area
depends upon the size of that area and its expected air quality.  Normally
it would suffice to have one monitor which is measuring the air quality in
a particular corridor or area due to the broad scale type of emissions.
For corridors which are axial to an area of high concentrations (e.g., a
central business district), a monitor located in each corridor and one at
the hub would be appropriate.  Where a highway network may be directed at
rerouting the traffic away from areas of high concentration (e.g., a
business district by-pass), monitoring should occur both along the by-pass
and in the area which is being routed around.  If individual corridors are
part of the transportation system only as a convenience to residents or
                                 32

-------
for the control of the air quality in an area to which that corridor leads
or by-passes and there is no potential problem with the air quality within
the corridor itself, it would not be necessary to monitor that corridor.

The actual selection of the neighborhood for the placement of air quality
samplers which will be monitoring the changes due to these transportation
controls depends primarily on the neighborhoods in which these controls
will be operating.  For central areas which are the object of the control
measures, it would, in most cases, be sufficient to monitor in the neighbor-
hood where the concentrations are projected to be the highest.  The selec-
tion of the neighborhood for monitoring of the air quality due to the
corridor control measures would normally best be consistent with the mid-
point of the corridor.  If this midpoint places the neighborhood within
three miles of the neighborhood selected for monitoring the central
area due to a short corridor length, then it may be better to monitor the
concentrations which correspond to the emissions at the beginning of the
corridor.  Monitoring at the outermost point of the corridor would also be
appropriate if significant activity was expected to be generated at this
point due to a mass transit system.

The monitors for measuring the air quality which results from a bypass
should be in any neighborhood where the standard has the potential to be
exceeded as a result of the bypass.  If the standard could be violated
along the bypass, then those neighborhoods with the highest potentials
should be monitored.  Normally the concentrations in these areas will cor-
respond to the emissions from interchanges where maximum traffic levels
will occur.

Modeling activities need not be part of the analysis in determining the
neighborhoods for the location of these monitors.  For carbon monoxide,
the neighborhoods selected will be one of the neighborhoods in the area or
corridor of interest due to the local nature of CO (e.g., the midpoint of
the corridor would exactly determine the neighborhood).  For nitrogen
                                  33

-------
oxides and photochemical oxidants, the neighborhoods selected would be
located a distance away from the actual corridor or area.   The exact dis-
tance and direction would depend upon the most frequent wind direction and
the average wind speed.

Monitoring for Land-Use Planning - If the air quality maintenance plan for
the AQMA also incorporates certain land-use planning measures as means of
maintaining the air quality, then these measures should receive some
attention from the monitoring network.   Land-use and planning measures as
listed in Volume 2 of the Guidelines for Air Quality Maintenance Planning are:

     A.  Emission Allocation Procedures
     B.  Regional Development Planning
     C.  Emission Density Zoning
     D.  Zoning Approvals and Other Indirect Regulatory Controls
     E.  Transportation Controls
     F.  Emission Charges
     G.  Transfer of Emission Source Location
     H.  Indirect Source Review
     I.  Environmental Inpact Statements (EIS's)

Full explanations and discussions of these measures are given in Volume III,
Control Strategies, of the above guidelines.

Transportation Controls have already been discussed above as a separate
program for monitoring due to their importance and broad applicability to
AQMA's.  Of the remaining measures, two main categories appear:  Those
measures which are real planning measures (A, B, C, D, G) and those mea-
sures which are review processes for ensuring the standards will not be
violated (H, I).  Emission  charges  are not currently expected to be an
acceptable measure for implementation and is therefore not considered.
                                  34

-------
This division allows for the discussion of these two separate categories
each of which have different monitoring requirements.  The primary dif-
ference is that the review measures are individual source oriented while
the general planning measures are more community or neighborhood oriented.

General Planning Measures - Of the measures presented above which are con-
sidered actual planning measures, regional development planning is the
only one which is necessarily a wide-ranging planning measure.  While emis-
sion allocation procedures may well be on a community basis, this is much
closer to the neighborhood type scale of the other planning measures.  For
these reasons, the selection of neighborhoods for the latter planning
measures is much better defined than under regional development planning.

Theoretically, measures which are neighborhood oriented need to be moni-
tored in that neighborhood or one downwind from it.  Practically, however,
this would require a prohibitively large network with monitors in every
neighborhood in an area in which zoning was a primary control measure.
This problem may be alleviated by monitoring in selected neighborhoods for
a period of time and then moving the location of the monitors to other
neighborhoods, or by having elaborate modeling activities which are ade-
quate to interpolate concentrations between monitoring sites based on
known air quality, meteorological conditions, and source strengths.  In
addition, certain areas which may be zoned for low emissions or none at
all (e.g., green belt)  may be excluded from the monitoring requirement
though the impact of green belts would be nice to know.

For those areas for which regional development planning is used to maintain
the air quality, certain activity centers and controlled growth areas, as
well as a possible transportation system,are likely to be established. These
measures have already been discussed under community selection and trans-
portation controls.  The selection of the neighborhoods for monitoring
growth in these communities should basically be a function of the degree
of growth incentives or restraints within each neighborhood.
                                35

-------
 Review Processes - Where indirect source review and environmental impact
 statements are used as a means of ensuring that a particular project will
 not cause a violation of the standard,  the monitoring required is obviously
 source oriented.   The considerations  are then the same as  discussed above
 under the monitoring for regulations  which apply to significantly large
 sources.   Particularly in this case,  the monitoring can be expected to be
 a short term operation where the  ambient air  quality around the  proposed
 source is monitored for a period  before  and after the construction of the
 source.   The monitoring may  also  be required  to be performed by  the source
 itself.

11.2.4  SITE SELECTION AND PROBE  PLACEMENT

The  final  selection of  the site for a sampler  and  the  exact  placement of
the  probe  rely upon the microscale influences  of buildings which may
exist  in  the neighborhood and  the need for  power and  security  for  the
equipment.   These  are problems which are  independent of  the  considera-
tions  of monitoring  needs for maintenance planning  and  therefore come
under  the  same guidelines as for  a regular  sampling network.   Such  guide-
lines  are  given in Section 4 of Guidance  for Air Quality Monitoring
Network Design and  Instrument Siting (OAQPS No. 1.2 - 012).  The follow-
ing  discussion is  basically drawn from this report.

11.2.4.1  General Considerations

In the selection of  a particular  site for a single  sampler or  a complex
station, it is essential that the sampler(s) be situated to yield data
representative of  the location and not be unduly influenced by the imme-
diate surroundings.  Little definitive information is available con-
cerning how air quality measurements are affected by the nearness of
buildings, height  from ground and the like.  There are, however, general
guidelines that can be provided based on operational experience:
                                 36

-------
   1.  Avoid sites where there are restrictions to air flow
       in the vicinity of the air inlet—such as adjacent to
       buildings, parapets, trees, etc.

   2.  Avoid sampling sites that are unduly influenced by
       down-wash  from a minor local source or by reentrain-
       ment of ground dust, such as a  stack located on the
       roof of a  building where the air inlet is located
       close to ground level near an unpaved road.  In the
       latter case, either elevate the sampler intake above
       the level  of maximum ground turbulence effect or
       place the  sampler intake away from the source of
       ground dust.

   3.  The monitoring site should be generally inaccessible
       to the public and should have adequate security,
       electricity, and plumbing.

   4.  Uniformity in height above ground  level is  desirable.
       Roof-top  samplers should be utilized in moderate  to
       high  density areas  (in  terms of structures).  Ground-
       level  samplers should be utilized  in low  or sparse
       density  areas  (in terms  of  structures).

    5.  For  CO or N02 monitoring,  samplers should not be
        located  in the median  area  of multilane highways.

11.2.4.2   Pollutant Specific Considerations


Sulfur dioxide  can be  considered to be  rather  well mixed  near the ground

at receptors not overly  affected by specific  point sources.   Therefore,

either ground or roof-top (one to three stories)  sampling is recommended.


Similarly, TSP is usually well mixed within the first few hundred feet
above the ground, but only roof-top sampling is recommended to avoid
the influence of possible reentrainment of particulates close to ground

level.


In contrast, the distribution of CO across a neighborhood consists  of

many more areas of peak levels, one at each street or major traffic

center, with areas of quite low levels between.  On a microscale level,

the variability means that one must consider site  locations for CO  in

scales of plus or minus a few meters,  rather than plus or minus


                               37

-------
thousands of meters as might be the case with S02 and particulates.  For
the same reason, height from the ground of the air inlet is more restric-
tive than for the other pollutants.  It is desirable, however, to sam-
ple as close as possible to the breathing zone within practical consid-
erations (i.e., proper exposure, security from vandalism, minimizing
surface effects, etc.).

The strong dependence of carbon monoxide concentration upon distance
from its source was illustrated in a recent field survey.  In this
study it was found that the concentrations experienced by pedestrians
exceeded those measured by the fixed air monitoring station, while
concentrations at randomly selected locations throughout the survey
grid were less than those at the monitoring station.  More specifically,
the data indicated that average concentrations determined by the monitor
would be reduced to near the urban background level by moving the
monitor approximately 200  feet farther back from the street.  For peak
CO sampling within street canyons, the side of the street which is
opposite the side facing the roof-top-level  winds is more likely to
experience the highest concentrations.

These variations basically require that two alternative locations be
selected, one for the 1-hour standard to which the pedestrian is exposed
and another for the 8-hour standard which better represents the ex-
posure of most individuals.  In most cities, the 1-hour standard is not
violated or likely to be if the 8-hour standard is maintained; there-
fore, such double monitoring is not normally necessary as long as the
urban background level is being adequately monitored.  The urban
background site for CO should be utilized to measure the maximum area-
wide concentrations to which the general population is exposed.  Thus,
either roof top or ground-level sampling in urban or suburban areas is
recommended.  This station should not be adjacent to major thorough-
fares (not closer than 50 feet from the street curb) to rule out the
influence of localized peaks due to roadway traffic.
                                38

-------
In the case of the reactive secondary pollutants (oxidants and NCL), the
best sampling locations are, in most cases, away from the sources which
emit the necessary precursors (and contribute to the reaction pro-
cesses).  Thus, the use of emission density and land use maps are not
always helpful in determining sampling site locations.  They can, how-
ever, be used in conjunction with information on the direction and
magnitude of prevailing mid-morning winds to provide approximate
sampling locations.  There are no well established meteorological dis-
persion models presently available for selecting areas of expected
maximum concentration for the secondary pollutants.  Probably high
concentration areas for these pollutants are based on: (1) available
information on the reaction kinetics of atmospheric photochemical
reactions involving hydrocarbons, nitrogen oxides, and oxidants; (2)
the diurnal variation in pollutant concentrations; (3) the distribution
of primary mobile sources of pollution; and, (4) meteorological fac-
tors.

In general, the maximum concentrations are indicated to occur between
5 and 15 miles downwind from the downtown or area of heavy traffic
density.  However, if the winds are light and variable, high levels
may occur in the vicinity of the pollutants emissions such as the cen-
ter city.  The location of good N02 and oxidant sampling sites is a
difficult process and in many cases is based largely on intuition or
trial and error.  The use of mobile N0~ and oxidant samplers could be
helpful in locating areas of concentration.

A minimum distance away from major traffic arteries and parking areas
of 100 meters is specified for the oxidant monitoring site
because NO emissions from motor vehicles consume atmospheric ozone.
NO,, is considered both as a primary stationary source pollutant and as
a secondary pollutant and air monitoring stations for this pollutant
should be located consistent with the respective station location
guidelines.  Differences in horizontal and vertical clearance dis-
tances are based on increased probability of reaction between reactive
gases and vertical surfaces.
                                39

-------
11.2.5  SITE DESCRIPTIONS

A complete and accurate description of a monitoring station, including
the monitoring instruments, is extremely useful in validating, editing
and interpreting ambient air quality data.  The location of the sta-
tion helps to establish what kind of levels are expected and how these
are supposed to change over time.  The identification of the instru-
ments provides the basis upon which the reliability of the data, es-
pecially at  the end points of the instrument's sensitivity, can be
determined.  The type of information which should be supplied for the
stations and the instruments is outlined in the  Air Quality Monitoring
Site Description Guideline   (OAQPS No. 1.2-019).  The type of informa-
tion  called  for and the  specific  items of  the  description which relate
particularly to maintenance  activities are discussed below.

The general  description  of the  station,  for all pollutants,  should
include the  objective  for monitoring each  pollutant  at that  site, the
type  of monitoring  station  (mobile or  stationary), the location of  the
station, nearby pollutant  sources, and the heating and air  conditioning
requirements.  This later  information  is  needed because  many heating
and cooling systems generate one  or more  of the criteria pollutants.
Of  special interest to maintenance planning is the objective for moni-
toring  at  a particular  site.  Special  attention needs  to be  paid to
those monitors  placed with  the  objective  of trend  analyses  and  of
assessing  the maintenance  of the  NAAQS.   It would  be helpful if  the
specific  objective, e.g.,  monitoring  area of  controlled  growth  or
activity center, were  indicated.  Some of this information may  be  pro-
vided by giving  the location of the  station.

Special information concerning  the  stations  for continuous  monitors
 includes the manifold  design, manifold composition,  and  the electrical
 requirements.   The former  information helps  to indicate  whether the
manifold will react with the constituents of  the  air sample so  that the
 composition of  the air in the maniford will  change.   The latter

                                40

-------
information  is simply for determining  if an adequate power supply  is
being received by all instruments.  Other special information that
should be provided concerns the actual siting positions for primary
and secondary pollutant stations and the air inlets.  None of this
special  information  is felt to be especially applicable to monitoring
for air  quality maintenance planning.

Descriptions of the  instruments are also no different for maintenance
monitoring than for  any other reason.  The basic information which
should always be provided include the identification of the manufacturer,
trade name,  and model of the instrument, the pollutant being monitored,
the SAROAD codes used by NA.DB to store the data from each monitoring
instrument used at each site, the measuring principle (e.g. colori-
metric,  nondispersive infrared, etc.), whether it is manual or  instru-
mental,  and  the actual techniques used in monitoring.  For continuous
instruments, the performance specifications should also be given.

Once a site  description for a station has been prepared and submitted to
the EPA  Regional Offices, it should not be considered to be invariable.
Instead, care should be taken to provide periodic updates of the infor-
mation included in the site description.  Under normal circumstances,
this should  occur once a year; however, if significant changes  occur
(e.g. new sources or new samplers), these should be recorded immediately
so that  any  variations in the data will not be misunderstood.

With regard  to maintenance planning, changes which reflect implementation
of a component of the air quality maintenance plan should be recorded.
In most  cases these changes occur on a scheduled basis and may  be  pro-
jected in the yearly site description update.  Estimates, qualitative
or quantitative, of the impact these control measures would have on a
particular site would help in understanding the trend of the data  at
that site.  In areas which are being monitored as potential AQMA's, it
will probably be necessary to update on a yearly basis only, unless
significant changes occur.

                               41

-------
                             SECTION 11.3
                  SAMPLING INSTRUMENTATION SELECTION

11.3.1  NEED FOR CONTINUITY

Selecting the proper sampling instruments for a monitoring network
which is directed toward the air quality maintenance planning activities
involves the same considerations as for any other air quality monitor-
ing network.  One consideration which deserves more attention in this
area, however, is the incorporation of reference methods.

Since air quality maintenance planning is a continuing process and is
not a special case-study type of project which has a defined beginning
and end, it is important to establish a proper sampling methodology
and procedure which may be continued with few minor interruptions or
variations over time.  Therefore care should be taken in the selection
of instruments which are to be used in the monitoring network.  Only
reference methods or approved alternatives, as discussed below, should
be used.

Where it is necessary to use unapproved methods, due perhaps to
previous acquisition of the equipment, they should be confined as much
as possible to networks in potential AQMA's.  It is much more likely
that the stations themselves will be changed in a potential AQMA than
in a designated AQMA for which an air quality maintenance plan
establishes the need for monitoring for at least 10 years;  in addition,
the unapproved methods will probably be sufficient to establish the
needed trend data for the potential AQMA.  If an unapproved method is
                                42

-------
used as part of the AQMP monitoring network in a designated AQMA, it
should be noted and care paid to replacing'it at the earliest possible
convenience.

11.3.2  APPROVED SAMPLING METHODS

A good summary of the sampling methods which may be used and discussion
of those methods which are not acceptable  is given in the guideline
document on the  Designation of Unacceptable Analytical Methods of
Measurement for Criteria Pollutants  (OAQPS No. 1.2-018).  The following
is extracted from that document to provide an understanding of the
categories of sampling methods.  Reference should be made to the
document itself for a more thorough understanding.

11.3.2.1  Categories of Analytical Methods

Methods for measuring air pollutants fall  into one of three categories:
(1) approved, (2) unacceptable, and (3) those methods which are neither
approved nor unacceptable (unapproved).  At present, the only officially
methods are the federal reference methods  described in appendices to 40
CFR Part 50, originally promulgated on April 30, 1971 (36FR8186) with
the National Ambient Air Quality Standards (NAAQS).  This Federal
Register also introduced the concept of an "equivalent method ," which
is any method which can be demonstrated to be "equivalent" to the
reference method.  Thus, unapproved methods may become approved only
by demonstrating equivalence to the reference method.

Those methods designated as unacceptable are not equivalent to the
                                       s
reference methods because they are known to yield measurements of poor
accuracy and reliability.  They are considered to be obsolete.  In each
case, suitable analytical methods which produce measurements of greater
reliability are available to replace the unacceptable methods.
                               43

-------
 11.3.2.2  Reference  and Equivalency Regulations

Regulations governing the procedures and criteria by which
unapproved methods may be determined to be equivalent have been pro-
posed in the Federal Register on October 12, 1973 (38 FR 28438) as a
new Part 53.  Pending revision based on comments from interested per-
sons, the new regulations, when finally promulgated, will require that
a method must be tested according to prescribed procedures and meet
certain prescribed specifications to be approved as an equivalent
method.  In essence, manual methods must demonstrate a consistent
relationship to the reference method in side-by-side measurements of
ambient air.  Automated methods (automatic air analyzers) must demon-
strate such a consistent relationship as well as meet certain perfor-
mance specifications.  The new regulations will also cover reference
methods which are automated methods (i.e. CO and Oxidants).  An
analyzer must meet prescribed performance specifications before it can
be determined to be approved as a reference method.

Unapproved methods must be tested according to the prescribed procedures
and submitted with an application for approval to the Quality Assurance
and Environmental Monitoring Laboratory (QAEML) of the National
Environmental Research Center (NERC) in Research Traingle Park, North
Carolina.  Approved methods are to be published in the Federal Register.
The regulations will apply to S02> CO and Oxidants corrected for S0?
and N02-

 11.3.2.3  Acceptability of Analytical Methods

Table 1  lists those  analytical methods for which data were submitted by
the States in 1972.  The individual methods have been listed as
 "approved," "unapproved" and "unacceptable."  Use of methods designated
 "unacceptable" should be discontinued as soon as possible.  Data
derived  from those methods will not be accepted or used by the NADB
after September 1974.

                               44

-------
Table 1.  1973 POLLUTANT-METHOD-STATIONS SUMMARY
Pollutant code
TSP
CO



so2











NO 2











11101 91
42101 11
12
21

42401 11
13
14
15
16
31
33
91

92
93

42602 11
12
13
14
71
72
84
91
94
95
96

Method
Hi-Vol (FRM)a
NDIR (FRM)
Coulometric
Flame ionization

Colorimetric
Conductimetric
Coulometric
Autometer
Flame Photometric
Hydrogen Peroxide*3
Sequential Conductimetric
West Gaeke-sulfamic
acid (FRM)
West-Gaeke bubbler
Conductimetric bubbler

Colorimetric
Colorimetric
Coulometric
Chemiluminescence
J-H bubbler (orifice)
Saltzman
Sodium Arsenite (orifice)
J-H bubbler (frit)
Sodium Arsenite (frit)
TEA
TGS

No. of
stations
3602
278
2
10
290
89
108
172
1
29
38
6
1510

11
0
1964
136
14
10
8
14
5
26
995
456


1664
Percent
of total
100
96
0
4
100
5
6
9
0
1
2
0
77

0
0
100
8
1
1
0
1
0
1
60
28


100
Approved Unapproved
xa
X

X

X
X
X
xb
X
xb
X
X




X
X
X
X


X

X
X
X

Unacceptable


X











X
X





X
X

X





-------
                       Table 1 (continued).   1973  POLLUTANT-METHOD-STATIONS  SUMMARY
Pollutant code
Photochemical
0 44101 11
(Ozone) 13
14
15
51
81
82
44201 11
13

Method

Alkaline KI instrumental
Coulometric0
Neut KI colorimetric
Coulometric
Phenolphthalin
Alkaline KI bubbler
Ferrous oxidation
Chemi luminescence (FMR)
Coulometric0

No. of
stations

10
10
89
22
3
79
91
131
1
436
Percent
of total Approved Unapproved Unacceptable

2
2 Xc
21 X
5 X
1
18
21
30 X
0 Xc
100

X



X
X
X



aFRM = Federal Reference Method.
^These methods should be reported under method code 42401 13.
GThese methods should be under method code 44101 15.

-------
For SO ,  CO and Oxidants corrected for S02 and N02> unapproved methods

may be used until the Equivalency Regulations are promulgated.  After

promulgation of those regulations and additional approved methods

become available, unapproved methods may be used only until they can be

replaced with approved methods, and not later than 5 years after pro-

mulgation after which time only approved methods are to be used.  For

NO  and hydrocarbons corrected for methane, guidance for selecting

adequate automated methods may be found in the forthcoming EPA
Environmental Monitoring Series Document  (EPA-650/4-74), Guidelines for

Determining Performance Characteristics of Automated Methods  for

Measuring Nitrogen Dioxide and Hydrocarbons Corrected for Methane in

Ambient Air.


Until these regulations and guidelines become available, the  following

guidance should be considered:

     • TSP - The hi-vol method is the  federal reference method
       lor total suspended particulates.  Since  the air quality
       standard  is defined by the method, the hi-volume sampler
       is the only acceptable method.  No procedures for deter-
       mining equivalency of alternate methods have been de-
       veloped,  so all  other methods  are  to be considered un-
       acceptable.

     • Carbon Monoxide  - The nondispersive  infrared (NDIR)  is
       the federal reference method  for CO.  Automated analyzers
       based on  other principles have  not yet been tested with
       respect to equivalency, and  are therefore unapproved.

     • Sulfur Dioxide - The manual West-Gaeke -  sulfamic acid
        (24-hour  bubbler) method  is  the federal reference
       method for S02-  The other manual  methods listed are un-
       acceptable.   The similarly named 'VJest-Gaeke" method
        (SAROAD method code 42401  92)  is not  equivalent  to  the
       reference method (SAROAD method code  42401  91).  Since
       no continuous method has  yet  been  tested  for equivalency,
       they  are  classified as unapproved.

     • Nitrogen  Dioxide - The manual NASN bubbler  method  is the
       federal reference method  for  N02«   However, in  the  June
       8,  1973 issue of the Federal  Register (38 FR 15174), it
       was proposed  that the NASN method  be  withdrawn  as  the
       reference method and a  new one designated after  testing
                               47

-------
of proposed candidate methods.  Although the method was
not officially withdrawn, the problems with variable
collection efficiency and NO interferences are such
that it must be considered unacceptable.  All other
methods, both manual and continuous, have been classi-
fied as unapproved.

Photochemical Oxidants (Ozone) - The reference method
for photochemical oxidants is a continuous chemilumi-
nescent method based on the gas-phase reaction of
ozone with ethylene.  This method is specific for
ozone.  All other methods listed in Table 1 are total
oxidant methods.  Six of these methods for total oxi-
dants are being designated unacceptable.  While the
remaining automated methods are not being designated
unacceptable, strong consideration should be given to
replacing them with the reference method.

Total Hydrocarbons Corrected for Methane - This cate-
gory is unique in that, while hydrocarbons corrected
for methane is a criteria pollutant, the Ambient Air
Quality Standard is only a guide for achieving the oxi-
dant standard.  A gas chromatographic flame ionization
technique is the federal reference method for hydro-
carbons corrected for methane, but this method is dif-
ficult and expensive to use.  Other methods are now
becoming available and, as mentioned before, guidance
for selection of adequate automated methods may be
obtained in the EPA Environmental Monitoring Series
document (EPA-650/4/74), Guidelines for Determining
Performance Characteristics of Automated Methods for
Measuring NOp and Hydrocarbons Corrected for Methane
in the Ambient Air.
                        48

-------
                              SECTION 11.4
                   DATA PROCESSING AND SUMMARIZATION

11.4.1  DATA PROCESSING EVALUATION

Data processing, as discussed in this section, involves the basic pro-
cedures needed for ensuring a smooth flow of air quality data from the
point of primary recording of the data, by an instrument or in a lab, to
its ultimate storage in some format or system from which it may be later
retrieved for analysis.  While this is directed at the handling of the
data and not the eventual analysis, which is considered in the following
section, some analysis is essential to provide validation of the data
entries into the storage system.  As with the selection of instruments,
the actual processing of the data from a monitoring network that is
directed toward the air quality maintenance planning activities involves
the same considerations as for any other air quality monitoring network.

The evaluation of current monitoring networks which must occur in light
of the maintenance activities provides the opportunity for the similar
evaluation of the data processing procedures presently being employed.
The establishment or expansion of monitoring networks may be increasing
the amount of data which must be handled, and the data processing systems
must be able to adequately cope with any increases.  In addition to
evaluating the capacity of the current data processing system, the ade-
quacy of the validation processes  in the system should be determined.
                                 49

-------
This  latter point,  the validation of data, deserves special attention
when  considering air quality maintenance planning.  One of the major
reasons  for this document is to aid in the establishment of a monitoring
network  which will  provide air quality data adequate for a basis from
which projections and long-range control strategies can be formulated.
If  the data is not  reported accurately, due either to equipment mal-
function or human error, then much of the effort expended in establish-
ing and  operating the network is wasted.

The discussion of data processing which follows,attempts to provide
guidance as to how  a data processing system should operate.  It is
directed more at the establishment of proper procedures in handling
and validation than at current practices, which are considered in the
Guideline "Procedures for Flow and Auditing of Air Quality Data" (OAQPS
No. 1.2-013).  It is recommended that these procedures be followed by
the states, as they are the bodies which are responsible for the desig-
nation of AQMA's and preparation of AQMP's based on this data.

11.4.2  DATA PROCESSING AT THE STATE LEVEL

11.4.2.1  Data Handling

Data handling refers to the basic mechanics of processing monitoring
data  from the sensor to the data storage system and the preparation of
routine summaries in one or more standard formats.   Included in this
category is the activity of preparing, on appropriate SAROAD data forms
or in other SAROAD-compatible format,  the quarterly air quality data
reports to EPA that are required by regulation.  The actual character-
istics of the data handling system depend upon the requirements of the
state.  Generally,  the more data that  needs to be processed,  the more
computerization and automation will be employed.  Therefore,  the data
handling systems can be categorized as either totally manual,  semi-
automatic (partially computerized),  or totally automated (including
telemetering of data).

                                50

-------
It is not the intent of this document to present guidelines on the

mechanics of a data handling system, especially as optimum systems are

most likely different for each state.  However, some generally applicable

comments regarding the type of air quality data which should be handled

can be made.

    a.  All air quality data obtained by the State, which satis-
        fies the criteria established for monitoring network ad-
        equacy, should be processed.  This includes data from all
        operational stations in the maintenance planning network,
        any other state or local network, special studies, in-
        dustrial monitoring, or private citizen monitoring.

    b.  It is desirable that the data be representative of a con-
        secutive 3-month period.  For continuous 24-hour data, it
        is desirable that there be at least five data points in
        the quarter, with at least 2 months being reported and a
        minimum of two data values in the month with the least
        number of data values reported.

    c.  Data must represent an interval of 1 hour or greater -
        shorter interval data must be averaged over a clock hour

    d.  Note should be made of whether the data is coming from
        a site for which an updated description has been made.

    e.  Data must be representative of the conditions of the
        site for the period of time specified; modification of
        the environment in which the site is located must be
        reported on the site description forms.

    f.  The number of significant figures that are meaningful
        for a particular air quality measurement is limited by
        the methodology employed.  To use more significant
        figures than is warranted by the sensitivity of the
        analytical procedure adds no real information and can
        often be misleading.  Table 2 presents the suggested
        reporting accuracy for raw data for various pollutants.
        While the conventions apply to the raw data it is also
        useful to specify the accuracy of geometric and annual
        means.  For simplicity, the general convention is that
        all means be reported to one more significant digit than
        the raw data.
                                51

-------
Table 2.  SUGGESTED REPORTING ACCURACY FOR RAW DATA
Pollutant
Suspended particulate matter
Benzene soluble organic matter
Sulfates
Nitrates
Ammonium
Sulfur dioxide
Nitrogen dioxide
Nitric oxide
Carbon monoxide
Total oxidants
Total hydrocarbons
Ozone
Methane
Number of decimal places
ug/m3
0
1
1
1
1
0
0
0
1
0
1
0
1
ppm
_
-
-
-
-
2
2
2
0
2
1
3
1
                       52

-------
 11.4.2.2   Data Validation

 The  type  of  analysis  that occurs in data processing is directed at  the
 individual pieces of  data as opposed  to a set of data as discussed  in
 the  following section.  Data auditing is that portion of the data proces-
 sing procedure that attempts to detect and correct errors.  This is not
 considered to occur at a distinct step in the flow of data but rather
 at many small steps.  Data auditing has two major components:  editing
 and  validation.  Editing refers primarily to the handling of the data
 where it  is  important to make sure that the proper data is recorded on
 the  proper forms.  This has already been briefly mentioned under data
 handling.

 On the other hand, data validation refers to the actual examination of
 the  individual data at specific steps in the processing system to de-
 termine where anomalous values may have entered.  It is not generally
 a major aspect of manual data processing operations, due in part to the
 small volume of data being handled and because the data will often be
 processed  entirely by a single individual.  However, data validation
 is of much greater significance with semi-automatic and automatic data
 handling  systems which involve a complex series of steps involving both
 human and  computer operations.   There are major laboratory or field
 operational errors, data entry or copying errors made by field or cler-
 ical personnel,  bad data from malfunctioning instrumentation; all of
 these will likely proceed through the processing beyond the point at
which they were made,  to be caught and fixed at some later point, or
 else not to be fixed at all, leaving a number which is not what it
 ought to have been.

 The only validation system which would have the slightest chance of
catching the  vast majority of mistakes of such diverse nature would be
 a system of 100  percent double  checking of every operation.   Such a
 system,  is, however,  not only prohibitively expensive, but also really
                                53

-------
impossible in the sense that the many operations in the field are not
susceptible to repetition for the sake of checking.

The types of errors which are likely to occur may be divided into two
groups, even though they are really just the opposite ends of the same
continuum.  There are small errors in data that are detectable only if
the true value is known, because they are really quite close to the true
value and do not seem "out of line".  Then there are the gross errors,
which are detectable because they produce data which deviates very much
from what is expected.

An error in the  former group is the more difficult to detect.  Because it
is never really  known what the "true value" should be, a system capable
of detecting small errors, e.g., a 28 in the midst of the data which
should really be a 31, cannot be established.  One must approach this
type of error purely by prevention, by designing the sampling, analysis,
and data processing procedures to minimize the chance for error, spot-
checking each operation to avoid systematic errors, and then relying  on
the knowledge that a few random errors of small magnitude will generally
"average out."   Systematic  errors will  of  course not  average  out.

The primary concern of procedural data validation must be those errors
which are relatively large in magnitude, i.e., those that produce num-
bers that are "outlying" in some sense, either extremely large, ex-
tremely small, or perhaps extremely different  from their neighbors.   It
is important to  remember that the data cannot be judged by what the
true value is but rather only by what it is expected to be.

What is most easily done is to set  an upper limit above which all data
is marked by the computer  (or data handler) as possibly in error; for
particulate data and  some gases a lower limit  is also wise.  For con-
tinuous data recorded as hourly averages or less, a criterion of change
between adjacent data values is also appropriate.  These boundary limits,
                                54

-------
beyond which data are singled out, may be based upon the past data for
that pollutant and that station, etc., setting aside a fixed percentage
of the data, or they may be standard for the entire system.  The former
is by far preferable, being a more even-handed, rational, way of doing
this operation and is discussed later, while the latter, being easier
to program into a computer system, is by far more common.

Minimum Validation Process - For performing upper limit validation
checks on a total network basis, the selection of the upper limits must
be such that legitimate excursions above the expected are not unneces-
sarily flagged out.  The following list illustrates some computerized
hourly validation checks under consideration:

               CO                               100   ppm
               S02                                2   ppm
               Ozone (Total Oxidant)              0.7 ppm
               Total Hydrocarbons                10   ppm
               Non-methane Hydrocarbons           5   ppm
               N02                                2   ppm
               NO                                 3   ppm
               NO                                 5   ppm
                 x                                        3
               Total Suspended Particulate     2000   ^ig/m

In cases where the reported pollutant measurements are below  the limit
of detection for the analytical procedure,  the reported  number  should
be viewed as representing a range  from zero to the minimum detectable.
However, in order  to use such data in computing annual summary  statis-
tics  such as geometric means it is convenient to have a  convention in-
dicating what value  should be substituted  for a measurement below the
minimum detectable.  As a general  rule, each value below the  minimum
detectable is replaced by a value  approximately equal to one-half the
minimum detectable.  Table 3 indicates selected minimum  detectable
limits used by the National Aerometric Data Bank  (NADB)  for various
analytical methods.  A complete listing may be obtained  from  the
                                55

-------
                       Table 3.   MINIMUM DETECTABLE  LIMITS  FOR SELECTED MEASUREMENT TECHNIQUES
Pollutant
Suspended particulate
Nitrate
Sulfate
Carbon monoxide
Sulfur dioxide
Total oxidants
Collection
method
Hi-vol
Hi-vol
Hi-vol
Instrumental
Gas bubbler
Instrumental
Analysis method
Gravimetric
Reduction-diazo coupling
Colorimetric
Nondispensive infra-red
West-Gaeke sulfamic acid
Colorimetric neutral KI
Units
, 3
ug/m
ug/m
, 3
ug/m
, 3
ug/m
, 3
ug/m
, 3
ug/m
Minimum
detectable
1.0
0.05
0.5
0.575
5.0
19.6
o\

-------
National Air Data Branch, EPA, Research Triangle Park, N.C. 27711.
The mid-point substitution was selected after examining the statistical
distribution of the data.  It should be noted that in comparing data
over several years, a standard minimum detectable should be used unless
it has changed by an order of magnitude.

In preparing summary statistics, if more than 25 percent of the obser-
vations are less than the minimum detectable no statistics are computed
from the data.

Having selected certain data for further investigation (by maximum
upper limit validation), the field and laboratory personnel (in the
case of data from mechanized sampling equipment) or the instrumenta-
tion staff (in the case of data from automatic instrumentation) then
would go back to the record cards, strip charts, or other records to
either supply the correct value, if in fact an error has been detected,
or to offer an explanation of why the value, though extreme, is prob-
ably correct.  This second look by the sampling personnel is more often
rewarding in the case of continuous instruments, where automatically-
recorded data has probably been processed from the instrument to the
first preliminary computer tabulation with little human scrutiny be-
yond the station operator's maintenance of the instruments themselves.
In the case of intermittent data, the laboratory personnel have prob-
ably used their experience to verify the numbers, at least subcon-
sciously, while they were recording the data, so that second looks are
less frequently fruitful here.

A system of this type, combining computer flagging of extremes with
manual verification of the flagged values, will catch only the most
blatant instrument errors, laboratory errors, and clerical misrecord-
ings.   Since it is desirable to perform validation on many data points
between these upper and lower extremes, which may be inapplicable to
particular sites, further sophistication in this area is needed.
                               57

-------
Desirable Validation Process - A desirable sophistication in the data
validation process is the use of comparison values based on the air
quality history of the specific site in question.  This is generally
not done at present, but should increasingly become the general prac-
tice as monitoring networks assume a permanent form and accumulate
larger bodies of historical data, as is the case for maintenance planning.

Such a technique simply involves using a set of comparison values that
reflect the actual air quality at the site in question, rather than
values that reflect only the extremes of all possible data from all
possible types and locations of sites.  The comparison values listed
previously, for instance, are quite extreme values, far beyond what
typical urban stations would ever experience.  They need to be this
high, however, in order to be unquestionably above the maximum levels
of those very few sites in the most extremely polluted locations.
Thus, for most stations, their use would be appropriate only for
automatically discarding values as obvious extreme errors, and not
for selecting out the few highest values for study and validation.

This difficulty is avoided by the use of a set of comparison values
specific to each site.  These can take into consideration not only
the general level of the air quality at the site, but also seasonal
variation patterns and, in the case of continuous stations, daily and
weekly variation patterns.  This ability to take into consideration
the regular variations of air quality over time is very important in
computerizing such processes.  As an example, a human data analyst
would react quite differently to an abrupt change in CO levels if it
occurred between 6 and 7 a.m. at the time of the morning rush hour
than if it occurred between 2 and 3 a.m. when CO levels are typically
stable.  In designing a system to computerize these judgments, the
comparison values must either be adjustable for different hours of
the day, or else must be so large to accommodate the morning peak that
they could miss significant smaller changes at other times.
                               58

-------
 Examples of the types of tests then can be used to identify potentially

 anomalous values as a step in the addition of the new data to the file
 are given below:

     • Values that are larger than the arithmetic mean of the
       data by some preassigned factor (such as 2).

     • Values that are some factor,  say 1.5 times larger than
       the 99th  percentile  of the  observed  data

     • Hourly values that differ from adjacent values by more
       than some preassigned ratio,  suggesting some  abrupt
       change in baseline or a transient interference.

     • Chebyschev type tests, wherein values that are more than
       four standard deviations away from the mean are to be con-
       sidered suspect.

     • Detection of any values that  are larger by some  factor
       (e.g.,  1.5)  than the theoretical expected  value  of the 99th
       percentile of the  distribution under question.

     • The finding that the average  of K >  5 successive values
       falls  outside the  (n + 3ff)  limit,  where p.  and  a2 are  the
                             /K
      mean and variance, respectively,  of  the distribution
       under  question.

 The  difference  in these  types  of  tests  should be noted.   In the first

 four,  the  assigned  percentile  is  estimated from  the  data, whereas

 in  the latter two  it  is  theoretically  obtained.   The  sensitivity  of

 these  latter  tests  can be  determined  analytically from the  frequency
 distribution.
11.4.2.3  Data Summarization


Characteristic Data Patterns - Before summarizing any data, some thought

must be given to the characteristics of the raw data.  This is partic-

ularly true for pollutants that have strong seasonal and diurnal pat-

terns that will affect the interpretation of the data.  For example,

the maximum hourly oxidant value for a year based on 4,000 observations

could have completely different meanings, depending upon whether the
                               59

-------
observations were made primarily during the winter or the summer.  This
section presents examples of some of these patterns.  The study of
these patterns can frequently be an important analysis in itself, since
they usually provide important insight into the behavior of the pol-
lutant.  An awareness of these patterns also provides a means for
screening the data for anomalous values.  It should be noted that
while the following discussion is general in nature, the character-
istic pattern at a given site is a function of local emissions and
meteorology and, as a consequence, characteristic patterns may be
specific to that site or locality.

Seasonal patterns - Figure 2 displays graphs of monthly averages for
various pollutants at a particular site.  Superimposed on these graphs
is a smoothed curve representing the seasonal patterns in the data.
Although the intensity of the seasonal pattern for a particular pol-
lutant may vary from site to site depending upon local factors, the
qualitative patterns are generally consistent.  Oxidants tend to have
seasonal maxima in the summer, while the other gaseous pollutants
usually have maxima in the winter.  S0_ usually has a more distinct
pattern than in this example, though particulate levels are frequently
as erratic as here.  A knowledge of the seasonality of a pollutant can
provide useful information for interpreting the data since it suggests
the season in which maximum concentrations would be expected.

Diurnal patterns - In addition to seasonal patterns some pollutants
also have pronounced diurnal patterns.  These patterns may be due to
factors such as solar radiation, traffic density, etc., which in-
fluence pollution levels.

Table 4 summarizes the 1971 oxidant data for the Downtown Los Angeles
site operated by Los Angeles County Air Pollution Control District.
The number of times that the national oxidant standard was exceeded
is presented by month and hour of the day.  The marginal totals indi-
cate both the diurnal pattern and the seasonal pattern.

                              60

-------
                    SUSI'KHnEU PAMTtCULATK
                     Hl-VOL GHAVIMKTRIC
                     VG/CU MBtKR (25 C)
57 '  5* * 5»  '  60 '  61 ' 62  '63  * 64  ' 65  ' 66 ' 67  ' 68  I 69
                                                                               CARBON MONOXIDE
                                                                      INSTRUMENTAL HOHUIEV KRS1VE
                                                                              NG/CU HKTKH (25 C)
                                                                 63 ' 64 '  65  ' 66 ' 67 ' 68 '  69 ' 70 * 71
                            SULFUR DIOXIDt
                            BMTAL COUDtJCTIMBTRIC
12!
 75
                          UG/CU METER  (25 C)
    63   64   65   66   67   68   69  '70  ' 71 ' 72'
                                                                  100
                                                                  JJU
                                                                                  OXIDES OF NITROGEN
                                                                               INSTRUMENTAL COLORIMETKIC
                                                                                  OG/CO METER (25 C)
                                                                      68 ' 69 '  70 ¥7T* 72
                            TOTAL onoMrra
                             COLOKIMRUC HBOTML XI
                          OG/CT) METER (25 C)
    «3  ' 64  ' 65  ' 66 ' 67  ' 68  ' 69
                                                                  Sflf

                                                                  4QO

                                                                  1/^ft

                                                                  ZAA
                                                                                     NITRIC OXIDE
                                                                                INSTRUMENTAL COLORIKETP^C
                                                                                  UG/CU METER (25 C)
                                                                                7y  ' 71  • 72
             Figure  2.   Graphs  of  seasonal patterns  for various  pollutants  at a
                           particular site
                                                  61

-------
Table 4 .  NUMBER OF HOURS ABOVE OXIDANT STANDARD  BY MONTH AND TIME OF DAY (1971 DATA)
M 1
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec
Total by
hour
23456789 10


1 1
4

129
2 13
2 8
3 6
2


1 10 43
11

1
1
6
3
9
19
17
10
7


73
N
1
4
3
8
4
12
18
16
10
5
1

82
1
2
4
3
8
4
12
15
16
10
9
1

84
2
2
4
2
7
3
11
11
7
6
6


59
345
3
3
1
731
1 1
621
4 1
3 1
1
2


31 7 3
Total by
6 7 8 9 10 11 month
8
16
12
44
16
65
83
70
46
31
2
0
393

-------
Frequency distributions -  One characteristic pattern of air quality
data that is particularly important becomes apparent after examining
some frequency distributions.  Many quantities are assumed to have a
distribution that is symmetric about the average, such as the normal
distribution.  Figure 3 shows the frequency distribution of total sus-
pended particulate data from Philadelphia.  It is apparent that this
distribution is not symmetric.  However, Figure 4 shows the frequency
distribution for the logs of this same data.  The distribution is more
symmetric and can be better approximated by a normal curve.  Data hav-
ing this property is said to be log-normally distributed, and this is
a common assumption regarding air quality data. •L>2

Preparing Data Summaries -  In planning or preparing a summary of air
quality data, perhaps the most important step is to first define the
purpose of the summary.  The usual use of these summaries is to simply
describe typical and peak levels.  This section discusses several basic
statistics that can be used for this purpose.  The first two subsec-
tions discuss the treatment of typical and peak values.  The third
discusses the range of the data.

Indicating typical values - This section discusses the arithmetic mean,
the median, and the geometric mean as indicators of typical values.  The
geometric mean and the median are  frequently used in air pollution
studies because of certain properties of the log-normal distribution.
In choosing the appropriate statistic, the purpose of the summary must
be considered.  While all three may indicate typical values, if the
purpose of the summary is to compare the data to the National Ambient
Air Quality Standards, then the standard suggests the appropriate
statistic.  Another statistic commonly used to indicate typical values
is the mode, which is the value that occurs most frequently.  The use
of the mode is not discussed here since it is usually of little value
in summarizing air quality data.  For example, the mode for oxidant
would be near the minimum detectable due to the large number of very
low values throughout the night.

                               63

-------
ON
50


40




30



20



10
                                ON
                                CO
                                    ON  ON ON
                               ON  ON
                               OO  ON
                                 I    I   I   I   I   I    I
                                O  O  O  O  O  O  O
                                C*"l  >^  lO  \O  f**»  00  ^^
ON
O
i-l
1
O
O
r-l
1-1
i-l
O
r-l
i— 1
CM
r-l
1
O
CM
,-1
ON
r-l
1
O
CO
I-l
ON
r-4
O
r-l
r-l
O
r-l
O»
NO
r-4
1
O
r-l
r-l
1
O
r-l
00
r-l
O
oo
I-l
ON
ON
r-l
O
ON
r-l
ON
0
CM
1
O
o
CM
ON
CM
1
O
CM
ON
CM
CM
1
O
CM
CM
ON
CO
CM
1
O
CO
CM
ON
-*
CM
0
CM
O
CM
A
                               Figure 3.  Frequency distribution - TSP  (Philadelphia-1969)

-------
                  40
                   30   -
                   20   -
                   10   -
Ul
                                    CT\
                                                    00   ON
    ON   ON  ON  ON  ON

    O   t—i  csl  ro  <)•
                                                                                      ON  ON

                                                                                      r~-  00
ON  ON   ON

O  .-I   CS
                                                                                                                      in
                                    men  nrom
                                     ii    iii
                                    oo  ooo
                                    n  vor^-oo
m^-j  fr»  ^-j-  in   vo   r^-   oo   ON  o  »—^   c^   
                                                                                                  m  m   m  m
                                                                                                                      in
                           Figure  4.   Frequency distribution  - log of TSP data  (Philadelphia-1969)

-------
    Arithmetic Mean
        Given a set of n observations, say X.., X-, •••, Xn, the
arithmetic mean is simply       n
                                   x
                             n     Xi
When the term "average" is used the arithmetic mean is usually what is
meant .

    Median
        The median is the middle value of the data, that is, the value
that has half the data above and half below.  If the data is ranked in
order of magnitude so that
X, < X0 ... < X  , then the median is X ...  if n is odd, and
 1 —  2.     —  n                      ITTJL
                                       2
   + X   , ,\     if n is even.
        The median is a convenient  statistic  that is not  influenced
by changes in  the extremely high or low values of the distributions,
as would be the  arithmetic mean.

    Geometric  Mean
        Given  a  set  of n  observations, say  Xn, X0,  ..., X ,  the
                                    I        j.   z.         n
geometric mean is g  =  (X^Xg.. -\)    •    Since this probably is  the
least  intuitive  of the statistics presented,  it  is worthwhile to dis-
cuss it in more  detail.

        If  a  distribution is  symmetric,  such as  the  normal distribu-
 tion,  then  the expected  value of the arithmetic  mean and  median are
 identical.   However, for a log-normally  distributed  variable, it is
 the  expected  value  of the geometric mean that approximates the ex-
 pected value  of the  median.   Therefore,  since air pollutants often
 have a distribution that is approximately log-normal, the geometric
                               66

-------
mean has become commonly used as a convenient method of summarizing
the data; for total suspended particulate matter, the annual standards
are expressed as geometric means.
        As an alternate computational formula, it should be noted that
log g = -
 M                           n
V1                      1   V
 )   log x,  or g = EXP  -    >
i_j        i             n   L_i
i=l                    I    i^l
                                                 log x
Indicating maximum values - As in the previous section, the purpose
of the summary is the critical factor in determining the appropriate
statistic.  Maximum values may be indicated by listing the maximum
and/or the second highest value.  The second highest value is impor-
tant because compliance with the short-term air quality standards is
determined by this value.  However, there are other statistics  that
are useful for indicating maximum values.  The principal difficulty
in using  the second highest value is that it does not allow for dif-
ferences  in sample sizes.  For example, if two monitoring devices are
side by side and one operates every day of the year, while the  other
operates  only every 6th day, it would be expected that the second
highest value for the everyday sampler would be higher than that for
the other, even though they both monitored the same air.  Table 5
illustrates how the second high value may vary depending upon dif-
ferent sampling frequencies based upon total suspended particulate
data from a Philadelphia site that sampled daily.

To allow  for this dependence upon sample size, various percentiles
are sometimes used to indicate maximum values.  For example, the 99th
percentile might be used for hourly data, while the 90th might  be ap-
propriate for daily measurements.  By using a percentile value  rather
than an absolute count of samples, allowance is made for sampling fre-
quencies  that differ from site to site and year to year.  Table 6
indicates the 90th percentile for the sampling schedules used in
Table 5.
                               67

-------
Table 5.  MAXIMUM AND SECOND HIGH VALUES (PHILADELPHIA-1969)
          FOR VARIOUS SAMPLING SCHEMES
Sampling Schedule
Everyday
Every Sixth Day
n
n
n
n
n
Every Fifteenth Day
n
n
n
n
n
n
n
ii
n
n
n
n
n
n
Observations
365
61
61
61
61
61
60
25
25
25
25
25
24
24
24
24
24
24
24
24
24
24
Maximum
325
219
195
244
215
325
239
205
325
239
219
234
201
215
195
188
195
160
244
215
179
238
Second Highest
244
215
171
238
211
234
205
176
207
191
196
165
198
211
183
173
169
154
199
201
171
205
                          68

-------
Table 6.  GEOMETRIC MEANS,  MEDIANS,  AND 90TH PERCENTILE VALUES  FOR SAMPLING DATA OF TABLE 5 .
Sampling Schedule
Everyday
Every Sixth Day
n
ii
n
n
ii
Every Fifteenth Day
ii
n
n
n
n
n
n
H
n
n
n
n
n
n
Observations
365
61
61
61
61
61
60
25
25
25
25
25
24
24
24
24
24
24
24
24
24
24
Geometric Mean
102.6
99.8
95.2
113.6
107.2
106.4
94.7
100.2
114.6
125.0
104.9
100.8
99.8
104.4
102.4
92.1
100.8
92.0
104.6
107.2
94.1
99.6
Median
97
105
93
113
101
105
94
111
121
130
96
105
90
98
99
95
96
88
97
109
94
98
90th Per cen tile
171
162
155
188
177
171
158
175
178
189
192
148
190
177
171
143
162
140
186
173
162
165

-------
 Indicators  of spread -  In addition to an indication of typical and
 peak values,  it is also frequently desirable to have a measure of how
 variable  the  data is -  did it fluctuate  widely or were all values
 fairly uniform?  The customary statistics for this purpose are the
 arithmetic  standard deviation and  the geometric standard  deviation,
 used in conjunction with the  arithmetic  and  geometric means respec-
 tively.   Ranges or percentiles could  also be used depending upon the
 desired use of the summary, but they  are not discussed.   The basic
 formulas  for  the  arithmetic and the geometric standard deviations are
 given below.
            Lfit Xj,  X^i  •••> X  be  a set  of n observations.
 Then the  arithmetic standard  deviation is:
                  n                              n
               11  V      — 2l  1/2         —   1  V
               •*•   \    f     \ *»|  A/ ^    ,     ^^   I  \    v1
               n  L   (V*> 1       Where X = n  /     i
                 1-1                            £l
 and  the geometric  standard deviation  is
                      n
        s   = EXP  F| V   (In X. - In g)2] 1/2
         °        L   /  i        *•         J

where g is  the geometric mean.

 11.4.3.  SUBMITTING DATA TO THE NADB

Beyond the  data processing and validation  the states  conduct  for  their
own purposes,  they  have a responsibility to provide  the data  to EPA
for entry into the  National Aerometric Data Bank.  This subsection
briefly describes the procedures used by EPA  to accept and process
this data.  Although these procedures include some data screening
and validation, it must be emphasized that these efforts can  in no way
be viewed as supplanting any portion of the data validation effort re-
quired of the  states.  The major contributions to quality control of
air quality data handling must necessarily be preventative, and there
is no choice other  than the states' conducting these  efforts.  The
                              70

-------
EPA procedures are designed only to protect the integrity of the

data bank; they cannot really contribute to the quality of the data

for the states' use.


The procedures used in entering data are at present (mid-1974) still
settling into ultimate form, as the new data reporting requirements are

implemented the first few times.  During this transition, it is ex-
pected that the EPA Regional Offices will be assuming increasing re-

sponsibility with respect to the screening and validation of data

before it is entered into the data bank.


The initial steps involved in submitting data are:

    1.  The State agency submits air quality data to the ap-
        propriate EPA Regional Office as part of the State
        Implementation Plan reporting procedures.  These re-
        ports, which are forwarded on a quarterly basis,
        contain both the air quality data and any new site
        descriptions for the State's air monitoring station
        The data may be sent in more frequently than quar-
        terly if desired, but must be submitted to the Re-
        gional Office in SAROAD format on either coding  forms,
        punched cards, or magnetic tape.  Data for all opera-
        tional stations as described in the SIP's, beginning
        with that used in plan preparation, must be submitted.
        It is strongly encouraged that all reliable data
        satisfying  the criteria established for monitoring
        network adequacy also be submitted, including data
        from the stations established for air quality main-
        tenance purposes.

    2.  The NEDS/SAROAD contact in the Regional Office arranges
        for keypunching of forms, if necessary, and then
        mails the data to the MDAD's National Air Data Branch
        in card or  tape form.

    3.  Air Quality data submitted to the National Air Data
        Branch should have the  following characteristics:

        a.  Data must be coded  in SAROAD format.

        b.  Data values less than the monitoring minimum de-
            tectable sensitivity should be reported as a
            "zero"  value.
                               71

-------
        c.   It is desirable  that the  data  be  representative  of
            a consecutive  3-month period  for  which at least
            75 percent of  the  data values  are valid.   (Values
            below the minimum  detectable  sensitivity  are con-
            sidered valid.)  However, if  the  validity cri-
            teria are not  met, the data should still  be sub-
            mitted, particularly for  evaluation of maximum
            value standards.   For intermittent 24-hour data,
            there should be  at least  five  data points in the
            quarter, with  at least 2  months being reported
            and a minimum  of two data points  in the month
            with the least number of  data  value reported.

        d.   Data must represent an interval of 1-hour or
            greater -- shorter interval data  must be  averaged
            over a  clock hour.

        e.   Data must be representative of the conditions of
            the site for the period of time specified; modifi-
            cation of the  environment in which the site is
            located must be  reported  to the MDAD by the State
            and/or the Regional Office.

    4.   Data will then be  processed using  the SARDAD  edit
        program, and investigation and correction of  poten-
        tial errors accomplished by the Regional Office in
        conjunction with the state involved.

    5.   Corrected data are submitted  to the NADB for  file
        updating.
11.4.4  REFERENCES


1.  Larsen, R. I., "A Mathematical Model for Relating Air Quality
    Measurements to Air Quality Standards," U. S. Environmental Pro-
    tection Agency, Office of Air Programs, Research Triangle Park,
    North Carolina, OAP Publication No. AP-89, 1971.

2.  Hunt, W. F., Jr., "The Precision Associated with the Frequency of
    Log-Normally Distributed Air Pollutant Measurements," J. Air Poll.
    Control Assoc., Vol.22, No. 9, p.687, 1972.
                                72

-------
                             SECTION 11.5
                             DATA ANALYSIS

The ultimate purpose of monitoring is of course the use of the data
gathered.  The analysis of data gathered for air quality maintenance
purposes utilizes the same general approaches and statistical tech-
niques that are applicable to analysis of any data, although the em-
phases and some of the interpretive conventions are necessarily differ-
ent.  This section considers the conventions involved in interpreting
data with respect to the NAAQS in the context of maintenance planning
and analysis, the general thrusts and emphases of data analysis re-
quired by the maintenance planning process, and finally, some of the
analysis techniques of primary utility in such efforts.

11.5.1  DATA INTERPRETATION CONVENTIONS

Because the National Ambient Air Quality Standards (NAAQS) and other
promulgated rules and regulations are usually expressed in results-
oriented phrasing, there have arisen specific questions about the
detailed interpretation of data with respect to the NAAQS or other
regulations.   Many of these questions have been addressed in the EPA
Guideline Document "Guidelines for the Interpretation of Air Quality
Standards," (OAQPS No. 1.2-008), and in a recent publication by Curran
and Hunt.   Some of these raise substantive policy issues, while others
are more a matter of simply assuring uniformity, and some arise pri-
marily in programming computerized analysis routines, where questions
that are normally a matter of human judgment must be quantified in
advance.  Guidance on these issues has been provided in the past for use
in an implementation planning framework; the discussion herein concerns
these issues in an air quality maintenance context.
                               73

-------
11.5.1.1  Geographic 'Aggregation of Air Quality Data

In using and studying air quality data, it is possible to either con-
sider each monitoring site separately or to average the data over a
specific geographic area.  When considering the data in relation to the
NAAQS, each monitoring site in an AQCR must be considered individually
in determining whether or not the AQCR is in violation of the standards.
This policy applies not only in the process of implementation planning
for the achievement of the standard, but also in evaluating the main-
tenance of the standards over the years.

The basic reason behind this policy is simply that the NAAQS were de-
fined to protect human health and welfare.  The presence of one moni-
toring site within an AQCR violating any given standard indicates that
receptors are being exposed to possibly harmful pollutant concentra-      :
tions.  This question arises most often when the concentrations in
excess of the standards values at a single monitoring station result
from the effect of a small, nearby source that is insignificant in
terms of the total emission inventory, or when the station in violation
is so located that the probability of individuals being exposed for
prolonged periods is negligible.  Such circumstances do not mitigate
the stated interpretation, since NAAQS are generally interpreted as
being set to protect health and welfare regardless of the population
density.  Although air quality improvement should be most stressed in
areas of maximum concentrations and areas of highest population expo-
sure, the goal must be ultimately achieving and maintaining the stan-
dards in all locales.  Data from monitoring sites are the only available
measure of air quality and must be accepted at face value.  Attention
is thus focused on the selection of monitoring sites in terms of the
representativeness of the air they sample.  This is discussed in more
detail in the guideline series document entitled "Guidance for Air
Quality Monitoring Network Design and Instrument Siting,"  (OAQPS No. 1.2-
012).  Good station siting should minimize this type of problem, and
                                74

-------
consideration should be given to the relocation of monitoring stations
that do not meet the guideline criteria.

In the case of projecting air quality for purposes of designing an air
quality maintenance plan, however, the aggregation of data within homo-
geneous geographical areas is frequently necessary and desirable.
Since emissions are necessarily aggregated and projected for areas as
large as counties or larger, perhaps for an entire AQMA or potential
AQMA, it is only sensible to project the air quality on a scale neither
much finer nor much coarser than the emissions data with which it will
be related, so long as the air quality is sufficiently uniform that an
average of several stations represents the area with sufficient accu-
racy for 10- and 15-year projection purposes.

This degree of uniformity can be recognized in situations where no
single site is consistently, year after year, significantly higher than
the others, but rather where various sites are roughly similar in level,
with the relative magnitude of the levels at the various sites changing
randomly from year to year or season to season because of minor meteor-
ological and source activity factors.  This would be experienced, for
instance, in an AQMA or potential AQMA that consists of a number of
towns or small cities within a large relatively undeveloped county.

If, in contrast, one of the towns or cities were consistently higher
than the others, its data should not be averaged in with other towns;
rather, the reason for its higher levels should be sought, separate
emission projections should be made, and a control strategy tailored
to its particular problem should be developed.  The extreme example of
this latter situation is of course an AQMA including both a major urban
area and its outlying areas.  In this case, the air quality data should
not be aggregated for the center city and the outlying areas any more
than one should expect that the same control strategy could be uni-
formly applied over such disparities of conditions.  In such cases,
fortunately, it is often possible to make separate projections of

                               75

-------
emissions, etc., for such sub-areas because more finely-divided data
bases are available.

11.5.1.2  Frequency of Violation of 24-hour Maximum Standards

The chances of detecting violations of 24-hour maximum standards depend
considerably upon the frequency with which the samplers are operated.
In view of this, there are questions about how data obtained from inter-
mittent monitoring should be interpreted.

Ideally, continuous monitoring of all pollutants would be conducted.
However, except for those pollutants specified in Federal regulations,
EPA does not currently require continuous monitoring.  Thus, one is
left with either (1) predictive equations employing data from partial
annual coverage, or (2) the data collected through partial annual cov-
erage.  It has been EPA policy that noncompliance will not be declared
based on predicted frequencies, because the accuracy of predictive
equations is not well established.  However, the detection of excursions
over the short-term standards should be a major consideration in deter-
mining the sampling frequency.  For assistance in making these judge-
ments, the following table gives the probabilities of sampling on two
or more days on which excursions have occurred for different numbers
of actual excursions above the standard and different sampling fre-
quencies.  The underlying assumption in determining these probabilities
is that excursions above the standard occur randomly over the days of
the year.  This is, of course, an oversimplification, but is sufficient
for the purposes of this discussion.

11.5.1.3  The Use of Running Averages for Short-Term Standards

The NAAQS for CO and S02 include 8-hour and 3-hour averages, respec-
tively.  With continuous monitoring data, there are various ways these
averages could be defined and the second highest average chosen.  For
                                76

-------
Table 7.  PROBABILITY OF SELECTING TWO OR MORE DAYS WHEN SITE IS ABOVE
          STANDARD
                              Sampling frequency - days per year
         Actual number
         of excursions

               2
               4
               6
               8
              10
              12
              14
              16
              18
              20
              22
              24
              26
61/365
122/365
0.03
0.13
0.26
0.40
0.52
0.62
0.71
0.78
0.83
0.87
0.91
0.93
0.95
0.11
0.41
0.65
0.81
0.90
0.95
0.97
0.98
0.99
0.99
0.99
0.99
0.99
183/365

 0.25
 0.69
 0.89
 0.96
 0.99
 0.99
 0.99
 0.99
 0.99
 0.99
 0.99
 0.99
 0.99
                                77

-------
maintenance planning purposes as well as implementation planning, com-
pliance with these standards should be judged on the basis of running
averages starting at each clock-hour.  The second highest average
should be determined so that there is one other non-overlapping value
that is at least as high as the second highest value.  Although this
seems relatively straightforward, the following discussion indicates
some of the subtleties involved.

The use of running averages to determine compliance with specific air
quality standards necessitates that the number of values above the
standard be evaluated on the basis of non-overlapping time periods.
That is, any two values above the standard must be distinct and not
have any common hours.  This can be achieved by a relatively straight-
forward counting procedure.  For example, in the case of CO, an 8-hour
average can be associated with each clock hour of the calendar year.
Then values above the 8-hour standard are counted sequentially begin-
ning with the first 8-hour average of the year.  Each time a violation
is counted, the next seven 8-hour values are ignored, and the counting
procedure resumes with the eighth 8-hour average.  This counting pro-
cedure results in the maximum number of non-overlapping violations of
the 8-hour standard.

This count is all that is needed to evaluate compliance with the 8-hour
standard because the standard is not to be exceeded more than once
per year; and, therefore, any count value greater than one is suffi-
cient to indicate non-compliance.  However, for maintenance planning
purposes, it is also desirable  to employ  the second highest 8-hour
average to indicate the magnitude of the problem.  There are several
ways to define the second highest value,  and three possible definitions
will be indicated here in order  to briefly discuss  their consistency
with the counting procedure  described above.   The three definitions
considered for the second highest value  are:   (1) the  second highest
8-hour value of those counted as being  above the  standard,  (2)  the
second highest 8-hour value  that does not overlap the maximum 8-hour
value,  and  (3) the maximum  second highest non-overlapping  8-hour average.
                                78

-------
Annotated graphs of 8-hour CO are used to facilitate the discussion of
the consequences of each definition.  For example, Figure 5 illustrates
that the first definition underestimates the magnitude of the problem
because the counting procedure may count the first time the standard
is exceeded and bypass the peak values.  Therefore this definition is
inadequate.

Although the second definition is intuitively appealing, Figure 6
illustrates that in some cases there could be two violations of the
standard, and yet the second highest value that does not overlap the
maximum is less than the standard.  This can only occur in marginal
cases in which the standard is only exceeded during one 15-hour period
in the year and that the maximum value occurs in the middle of this
interval.  Figure 7 shows another case in which this definition
produces the peculiarity that a higher CO value may lower  the second
highest value.

In order to avoid these inconsistencies, it becomes necessary to define
the second highest value as the maximum second highest non-overlapping
value.  What this means is that there  is one 8-hour value  that is
greater than or equal to the maximum second highest value  and that
these two values are not overlapping.  It is important to  recognize
that the maximum second highest value  may overlap the maximum 8-hour
value.  However, as shown in Figure 6, there is still one  other 8-hour
non-overlapping value that exceeds  the maximum second high.

With these subtleties in mind, it is considered appropriate  to use  the
maximum second highest non-overlapping value as the second high.  In
this way,  the magnitude of the problem is properly  assessed; for
quantitative planning the second high  value is also always consistent
with the number of violations.  This definition of  the second highest
value is also consistent with the approach used in  determining control
strategies on the basis of the roll-back equation.  It is  this maximum
                                79

-------
                                                                      STANDARD
 I
 oo
2 3456789  10
12  13  14 15  16 17 18 19 20 21  22 23  24
Using  the  counting procedure,  the violations are counted at hours 3 and
12,  as indicated  by the x's.   Note that the peak values do not occur
at these points.


Figure 5.   Eight-hour average  violations determined by counting procedure
    14

    12

    10
            M
                                            ''III
                                                        1
                                                                     STANDARD
       I  2  3 4 5 6 7 8 9 10 II  12  13  14  15  16  17  18  19 20 21  22  23  24
There are two non-overlapping violations at hours 3  and  12, and these
are detected by the counting procedure.   However, the  true  maximum
occurs at hour 8 and  the second  is below the  standard.   However, in
this case the maximum second highest would  be V2, which  is  above the
standard.  Although M2 overlaps  the maximum,  M,  there  is one 8-hour
average, namely Vj, that is at least as  high  as  V2 and the  two time
periods are disjoint.


Figure 6.  Eight-hour average violations as highest  non-overlapping values
                                 80

-------
   14
   12
   10
      M  I  I  M I  I  I   I   I   I   I   I   I   I   I   I   |   |
                                                                     STANDARD
      I  Z 3 4 S 6 7 8 9 10  II  12  13  14  IS  16  17  18  19  20 21  22 23 24
  10
                                                                    STANDARD
      123456789 10
                                                 18  19 20 21  22 23 24
In Figure  7 a the maximum value  is  12,  as  is the second highest value.
However, in Figure  7b  the maximum  is now  14,  and the second highest
non-overlapping is below the  standard.  Ihus,  the second highest non-
overlapping value can  actually  be  lowered by  having more high values.
It should be noted that in both of the  above  cases the maximum second
highest non-overlapping value is 12.


Figure 7a,b.   Subtleties involved in  using non-overlapping values
                                  81

-------
second highest value that must be maintained below the standard in
order to satisfy the requirement that the standard not be exceeded
more than once per year.

11.5.2  DATA ANALYSIS NEEDS FOR MAINTENANCE

The discussion of the considerations for the selection of communities
and neighborhoods for monitoring networks, as presented in Section 11.2
relied primarily on the type of data which was desired to be assimilated.
Obviously, the type of data which was wanted was a direct result of
the use to which it was to be put.  Therefore a basic understanding
of the concepts behind the establishment of a data gathering system
has already been established.

This discussion summarizes these concepts and provides a review of
how the data will actually be used.  Just as the differences in the
needs for data assimilation for potential AQMA's and designated AQMA's
required the separate presentation of these two categories, the parallel
need for different analysis of this data also requires their separate
presentation at this point.

11.5.2.1  Designation of AQMA's

The monitoring network for potential AQMA's, described in Section 11.2
is one which allows for a better analysis of the area under question
for the next iteration of AQMA designations in 1980, and the subse-
quent maintenance plan preparation if warranted, by the establishment
of an appropriate data base.  To fulfill both of these objectives it
is necessary to analyze the air quality and other data to determine
what the trend is in air quality and how this compares with the changes
in other parameters of interest.  This level of comparison may either
be on a visual trend-versus-trend basis or may actually be directed
at the determination of a quantitative relationship.
                              82

-------
Trend Comparison - In order to project air quality on a 10-year basis,
it is necessary to have a good understanding of how the air quality can
be expected to change due to the growth of certain sectors of the econ-
omy and the emission reduction planned under the state implementation
plans.  At this time, such an understanding is insufficient.  A review of
the trends of air quality and the parameters of interest may at least
provide an estimate of the changes which are likely to occur and may even
provide a quantitative methodology which is applicable to the region of
interest.  The parameters of interest which are suggested for considera-
tion are population, employment, and industry earnings.  The more finely
these parameters are divided to correspond to the pollution being mea-
sured, the more likely good relationships can be determined.

If the emission levels are also being monitored, then it is likely
that the air quality data would be compared to the emission levels and
the emission levels would be compared to the parameters of interest.
This type of review leads the way to the calibration of modeling tech-
niques while providing a more appropriate quantity (i.e., emissions)
to be projected from the socioeconomic data.

Special attention in the review could be paid to the impact on air
quality of any new sources, regulations, controlling emissions, or
special meteorological conditions which occurred during the period
of interest.  Such events could significantly change the air quality
without affecting the socioeconomic data thereby leading to confusion
when trying to make comparisons.  Consideration of these impacts also
provides a good basis from which the impact of possible maintenance
measures could be projected.

Other Analyses - As mentioned above, the data provided by the monitor-
ing network may provide sufficient information to allow for a quali-
tative comparison of the air quality trends with the appropriate para-
meters which could allow for a semi-quantitative interpretation of
future air quality based on the projections of the parameters.  Such
                               83

-------
a trend comparison may also lead the way to a more quantitative deter-
mination of an air quality or emission growth projection methodology
which is applicable to the area of interest.

11.5.2.2  Monitoring of the AQMP

The monitoring network described for the designated AQMA's was pri-
marily formulated around the need for reviewing the progress of its
maintenance plan.  This network provides the data necessary for the
establishment of a trend from which the general success of the program
can be determined and for the continuing review processes which may
be inherent in the plan.

Establishment of Trend - By reviewing the air quality data provided by
the monitoring network, it is possible to establish a trend in the air
quality.  For the purpose of determining the real success of the plan,
it is most helpful to do trend analysis on  the individual stations as
well as the overall air quality situation.

The trend in the air quality at each station should be  compared with
that projected to occur under the AQMP.  Where the trend  is not that
which was projected to occur, whether it is more  advantageous  or not,
the reasons for  the differences should be investigated.   The review
of both situations allows  for not only the  identification of problem
areas where new, more  stringent or  different controls are needed but
also  for  the better understanding of those  measures which were im-
plemented and are having a greater  than expected  effect.  Either of
these reviews may provide  a better  basis  for estimating the projected
impact  of any new measures which may need  to be  applied.

Review  Processes - The air quality  data  and the  trends  thereof also
serve  as  the basis  for any review processes which may be required  in
 the maintenance  plan.   This  review  may be  needed for  new source con-
 struction (both  direct and indirect),  for  special operating  conditions

                               84

-------
 (supplementary control strategies), or for emergency actions needed
 due to periods of air quality alerts.  The latter two of these require
 an immediate response type of network where minimum analysis is
 possible which the review for new source construction would rely more
 heavily on  the trends in the area and air quality modeling.

 11.5.3  TREND ANALYSIS TECHNIQUES AND APPLICABILITY

 The detection, verification, and assessment of trends in the ambient air
 quality, as measured by the monitoring network, is the single most im-
portant data analysis requirement, being significant not only for monitor-
 ing the maintenance of the standards in designated AQMA's, but also for
 obtaining guidance relative to the possible designation of additional
 areas.  To fulfill this need, there are a variety of techniques avail-
 able, ranging from the simple to the sophisticated, and in some cases
 applying best in slightly different situations.  In this section, a
 variety of techniques will be described, and the purpose and applica-
 bility of each discussed.

 11.5.3.1  Visual Techniques

 When performing a trend analysis, it is extremely desirable to actually
 look at the data in a graphic form.  Since the raw data are commonly
 quite variable, plots are usually made of quarterly or annual averages.
 If there is a clear trend in the air quality, its determination may be
 intuitively obvious in such a plot, and no more may be needed.

 If there is still significant variability in the statistics in such a
 plot, however, the trend may not be obvious; in this case, it is de-
 sirable to have a smoothed trend line through the data, and it is im-
 portant that such a trend line be determined in an objective way.  This
 can be most simply done by calculating a moving average of the observa-
 tions.  This will provide a smoother and simpler representation of the
                                85

-------
original data, as in Figure 8.  For quarterly averages, a moving annual
average consisting of four quarterly averages will eliminate the seasonal
fluctuations and remove much of the random variation as well.  When con-
sidering annual values over several years, a 3-year moving average will
smooth out much of the year-to-year variation. In specific instances other
averaging schemes may be considered; the selection of the appropriate
moving average is somewhat subject to personal judgment and experience.
When employing the moving average, the first and last values at the be-
ginning and end of the data series are usually omitted.

At least one much more elaborate smoothing technique, the Whittaker-
Henderson smoothing formula, has been previously applied in displaying
trends in air quality data, but this approach is sufficiently more
laborious to apply and generally requires the use of a computer.

The Whittaker-Henderson formula is  a finite-difference  formula  drawn
from the field of actuarial science.  It calculates the smoothed curve
as a balance between the smoothest possible line (a straight line) and
the best-fitting line (a line through each data point); by changing a
weighting parameter in the formula, the balance between smoothness and
good fit can be modified to provide plots that indicate seasonal patterns
as well as plots that demonstrate long-term trends.  These smoothed plots
make excellent, objective, diagnostic tools for examining data in search
of trends.  They do not, however, provide a means of objectively demon-
strating the significance of the trends that may be found;  for that, a
statistical significance test, such as described in the next subsection,
is required.
                                 86

-------
     150
     125
     100
      75
oo
      50
      25
                                                                                 V  ANNUAL  GEOMETRIC MEAN
                                                                                 O  THREE  YEAR AVERAGE
       0
         60
61
62
63
64
                                                         65
                                               66
                                              67
                                              68
                                               69
                                                                                                        70
71
                                                           YEAR
                            Figure  8.  Suspended particulate data  from Tucson, Arizona

-------
 11-5.3.2  Significance Technique^

 These  techniques  consider  the  statistical  significance  of  the  correlation
 coefficient  between  a  series of  pollutant  observations  or  summary  statis-
 tics with  the  sequence  in which  they were  observed.  If there  is an  over-
 all trend  upward  or  downward,  this  type of technique assesses  the
 probability  that it may be a  sampling anomaly  from a real- life situation
 which  does not have  a real trend.   This is  done by judging the trend in
 the light  of the  variability of  the data,  so that a real trend will be
 much  less  likely  to be found  significant  if the series  of  data show
 great variability than if  the  data  series  is  rather  smooth.   Because
 of this, it  is desirable to use  a data  series  that has  no  extraneous
 variability;  for  a  pollutant with strong  seasonal patterns,  for instance,
 one should either use  annual average values or  use the  averages for  each
 quarter  as a  separate  series.

 Two different  types  of  these correlation significance procedures are
 presented.  The first is nonpar ame trie, meaning no further mathematical
 assumptions are necessary.  It seeks a consistently changing series,
 either up  or down.   The second is parametric, requiring  the additional
 assumption, frequently encountered, that either the data or their  log-
 arithms  are normally distributed.  It is sensitive to a  constant absolute
 or percentage change.  In both approaches,  the time interval between ob-
 servations is not considered,  so that missing observations can be  ig-
 nored .

Daniel's Test for  Trend -  In order to utilize this procedure, at least
four observations  should be available.   Given observations  X , ..., X
and their corresponding relative  ranks  R^),  ...,  R(X ) , the test sta-
tistic is the Spearman Rank Correlation Coefficient:
                            P-l
                                    n(n2-!)
                                 88

-------
                    2
where T = Z[R(X.)-i] , that is, the summed squares of the differences
between each value's rank and its sequential order, i, in the series of
n observations.  The absolute value of p is compared with a critical
value w   in a table, if n < 30, or with w  = X / Vn-1, if n > 30, where
X  is the p quantile of a standard normal random variable.  If |p| > w ,
then a trend is declared significant at the a = 2p significance level.
A positive value of p indicates an upward trend, while a negative value
of p indicates a downward trend.  It can be noted that the estimate of
the Spearman rank correlation coefficient p is merely the usual product
moment correlation  of  the ranks of  the observations with  the order in
which the observations were  taken.

Example  applications of Daniel's  test - The following  table provides  the
annual geometric means, their  relative values,  and  the index over time,
for part of the Tucson TSP data from Figure 8.
xi
R(Xi)
i
128
8
1
118
7
2
80
3
3
89
5
5
70
1
5
78
2
6
96
6
7
88
4
8
 If  ties had occurred,  the ranks could be determined by averaging  the
 ranks  among the  tied observations, or preferably by utilizing the data esti-
 mates  to  the next  available place, even if it is not a significant digit.
     T  = Z [R(Xi)-i]2
       = (8-1)2 + (7-2)2 + (3-3)2 + (5-4)2 +  (1-5)2 + (2-6)2
        + (6-7)2 + (4-8)2
       =49+25+0+1+16+16+1+16
       = 124
       _ i      6T
     p  - I -    ^
            n(n -1)
       = -0.476
                                  89

-------
 The  0.90 quantile  of  the  Spearman test  statistic  is  0.5000  for  n  = 8.
 Since  the absolute value  of  the  calculated  rank correlation coefficient
 (0.476)  is less  than  this value,  the  correlation  would  not  be accepted
 as being significantly  different  from zero,  even  at  the a = 0.20  level.
 This means that, based  on probability theory,  there  is  at least a  20 per-
 cent chance  that the  observed  downward  trend in the  8 years' data  is a
 sampling anomaly from a situation where  there  is  really no  trend;  i.e.,
 no correlation between  the position of  a year  in  the series  and the relative
 geometric mean particulate level  for  that year.

 Daniel's test is,  like  other such tests, sensitive to the number  of
 data points  involved.   When  the  test  is  applied to the  12 years of
 data 1960 to 1971,  the  calculated coefficient  is  p = -0.769.  The  ab-
 solute value 0.769  is greater  than the  0.995 quantile for n = 12.
 Therefore,  the 12-year  trend would be clearly  classified as  signifi-
 cantly different from zero (with  1 percent  chance of error)  and, be-
 cause of the negative correlation, as being  downward.   When  the test
 is applied  to the  4-year  period 1968  to  1971,  the coefficient is + 0.80,
 which is significant  only at the  0.20 level.   Although  the  upward  pat-
 tern for the 4 years  seems quite  clear,  the  test  is  just not very  sen-
 sitive for  such short periods  of  data.

 The nonparametric  correlation  technique  is primarily useful  for clas-
 sifying  the  temporal  pattern as upward or downward and  for indicating
 the consistency of  the  pattern by the statistical significance  level.
 It is not very useful for  studying the data  to find a trend because of
 its lack of  sensitivity,  although it could be  used in a computerized
 screening  situation to  select  out only the clearest and most obvious
trends.   Recent  publications  by Faoro and Frank provide excellent ex-
                    -•                  23
amples  of the use of* this methodology. '

Student  t-Test  for  Linear Trend in a Sequence of Normal Variables  - In
contract with the previous technique,  the theory underlying this ap-
proach  requires  assumptions about the distributional form of the data,
                                 90

-------
 specifically that the data or their logarithms are normally distributed.
 With participate data, the latter assumption is usually chosen.
 Let X.,  i = l,n  be a sequence of observations or their logarithms.
 Then the test statistic is
           	  	 s\
 r*
v a  -  c
                         where   c = r-r- (n -1)
                                     *-£.
The calculated T is compared to the p   quantile of Student's t sta-
tistic  with n-2 degrees of freedom provided in Table A-3.  If  |T| > t,
then the trend is declared significant at the a = 2 (1-p) significance
level.  A positive value of T indicates an upward trend, while a nega-
tive value of T indicates a downward trend.

Example Application of Student t-Test - Again using as an ex-
ample the particulate data from Tucson, Arizona for 1964 to 1971, we
have:
                        I
                        / V
                         n-2 Vc  /3/ VCT -c/3

                      -    (n2-!) =    (64-1) = 5.25

                      _   1     [-3.5 In (128) -2.5 In (118) -1.5 In (80)
                        8(5.25)   -0.5 In (89) + 0.5 In (70) + 1.5 In (78)
                                 + 2.5 In (96) + 3.5 In (88)]
                      = -1.985/42 = -.047
                                91

-------
2          2 -     ^  - °-
                             /6 V5.25 (-0.047/ J .037-5.25 (0.002))

                      = -1.65

This value lies between the 0.90 and 0.95 quantile of the student's t
statistic.  Therefore, the trend can be considered significant at the
0.20 level but not the 0.10 level.  Thus both the Spearman and the para-
metric correlation techniques failed to detect a trend during 1964 to
1971 because of the year-to-year variability in the annual estimates.

Considering the entire 12-year period, the value of the nonparametric
test statistic T is -3.810.  This is significant at the 0.01 level, and
the  trend can therefore be classified as significantly downward.  The
corresponding value of the test statistic T for the 4-year interval
1968 to 1971 is + 2.15.  This is only significant at the 0.20 level.
Again, note the similarity between these results and those obtained
by using the simpler non-parametric analogue.

Testing for Proportion of Standards Violations -  This technique is
useful to test for a trend in the occurrence of extreme values or other
short-term statistics.  A chi-square test compares the percent of ob-
servations above a given threshold concentration, such as a 1-hour
standard, between two time periods.  It is desirable to consider inde-
pendent observations.  Therefore, for hourly data one should consider
at most one observation per day, e.g., the maximum observation per day
or the observation of a particular hour.  In general, observations
derived by intermittent sampling can be considered independent.

The data are arrayed in a table and labeled as shown below; the  test
should not be used if there are less than five observations in any of
the four cells.
                                92

-------
                       No. Obs • ^tandard   No. Obs > Standard
     TIME PERIOD I

     TIME PERIOD II
a
c
b
d
nl
n2
                                                                 N
Let p1 = b/n- be the proportion of observations in time period I that
are above the standard.  Similarly, P2 = d/n2 for time period II.

One can test (i) for any change between the two time periods, disre-
garding whether it is an increase or a decrease; i.e., PJ^ = P2» or (ii)
for a specific direction of change between the two time periods, say
The  test  statistic  T is defined  as:
                                  (ad-bc)
                         nxn2  (a+c)  (b+d)

 Consider  a  change  to  have  occurred if  T exceeds  the  chi- square  statis-
 tic  of Table A-4 at the  1  - a  quantile with  1  degree of  freedom.

 If only one direction of change  is of  interest,  for  example  "has  there
 been improvement", then  consider improvement to  have occurred if  T ex-
 ceeds the chi-square  statistic at the  (1 - 2a) quantile.   In either
 case the  significance level is approximately a.

 Example Chi- Square Test  -  The  following table  represents the number of
 days on which  the maximum  1-hour oxidant concentration exceeded the
 1-hour standard at a  particular  location during  two  periods, 1964 to
 1967 and  1968  to 1971.

-------
                        < standard
> standard
1964-1967
1968-1971
662
714
154
111
nx = 816
n = 825
                   TOTALS   1376
     265
1641
                                 (ad-bc)
                     T =
                              (a+c) (b+d)
                       _ (1641) [(662) (111) - (154) (714)]
                              (816) (825) (1376) (265)
                         8.9
At the level of significance of 0.05, the calculated value of  T
exceeds the tabulated statistic at 0.90 quantile = 2.706.  It can
therefore be concluded that short-term oxidant levels have signifi-
cantly decreased in recent years at the particular site.

11.5.3.3  Quantitative Techniques

Although the significance techniques just discussed provide an objective
probability judgment about whether an air quality trend is real, they
do not provide any quantitative estimate of how much the pollutant
                                                         3
levels are increasing or decreasing, say in terms of ^g/m  per year.
Such information could be gained, for example, by scaling a plot of
the data.   However, this introduces again an element of subjectivity,
as well as being impractical for extensive application.  Preferably,
objective estimates of the slope of the trend should be made using re-
gression techniques.

Regression techniques involve choosing a simple model to represent the
trend and then estimating the parameters of the model to fit the data
as well as possible.  The two models to be discussed here are a simple
linear model, corresponding to a constant absolute change from year to
                                94

-------
year, and an exponential model, corresponding to a constant percentage
change from year to year.

Simple Linear Model - To estimate a constant absolute change, b, cor-
responding to the model X = a + bT, use as an estimate of the slope b:
where pollutant concentration X. exists at time T^.
The estimate of "a" is  a = X - bT.

The use of this algebra will of course produce an estimate of a slope
even if the trend would not be seen as real by one of the significance
tests.  If there is little real trend, the estimate of the slope will
be small, but only  by chance will it be precisely zero.  Thus there is
a need to test, in  the manner of the significance techniques, whether
the estimated slope is significantly different from zero, given the
variability of the  original data.

The statistical significance of the estimate of b as compared with an
assumed value b. can  be tested by computing
                  B  =  (b-b.)  v  S2/E  (T.T)2

where  s2  =  {Z  (X-X)2 -  [Z  (T-T) X]2 /  [Z  (T-T)2]}  /  (n-2)
and comparing  B with the Student's t  statistic, t, at  the p quantile,
with n-2  degrees of  freedom.  If |fi| >  t then the  rate of change is
significantly  different than b. at the a  = 2p  signficance level.  If
b. is  chosen as zero,  the  test amounts to a  test of  whether there is
any real  trend.  In  a  similar manner,  a confidence interval can be
                           s\
created about  the estimate b.  The interval  is defined as
                                95

-------
^       2      — 2
b + t  s /£ (T-T) .  This interval contains the "true" rate of change,
b, with probability 1 - cc.
Exponential Model - To estimate the percent rate of change, r, corres-
                           T
ponding to the model X = ar , calculate and test the significance of
log(r) by substituting log(X.) for X. in the formulae of the previous
section.  The rate of change is usually presented as a change of (r-1)
x 100 percent per unit of time; e.g., r = 1.17 would mean 17 percent per
year.
Examples of Regression Application - The above regression techniques are
applied to the TSP data for New Haven, Connecticut, in Figure  9.

The estimates of absolute and percentage rates of change are presented
for the time intervals 1960 to 1971, 1964 to 1971, and 1968 to 1971.

                         Rates of Change
                          3
            Absolute  (ug/m )	Percent
0.26
-2.24
+7.00
1960-1971
1964-1971
1968-1971
+0.27
-3.46
+9.26
This again demonstrates  that  the choice of time interval can play an
important role in  the determination of an estimated rate of change.

11.5.3.4  Quality Control Approach

Another approach to trend analysis that is really a significance type
technique applied sequentially, has been adopted from the field of
quality control;  it is utilized in EPA's computerized Plan Revision
Management System (PRMS), and is particularly suited to the situation
of air quality maintenance.^ This quality control approach abstractly
                               96

-------
150
125
100
 75
 50
 25
                                                              I
                                                                   V  ANNUAL GEOMETRIC MEAN

                                                                   O  THREE YEAR AVERAGE

                                                                           I         I
   60
61
62
63
64
65
   66
YEAR
67
68
69
70
71
                        Figure 9.  Suspended particulate data from New Haven, Connecticut

-------
involves making a significance test at each period of time in a series,
testing the observed air quality value against a prediction.  This
involves first calculating t-test confidence intervals about the
predicted air quality value, based on the observed standard deviation,
and using them to test whether the measured air quality is significantly
different.

The primary value of this approach is visual; as seen in Figure 10,
plotting the predicted and observed air quality and the confidence
limits together permit the visualization of trends toward or away from
the predicted value even before they are confirmed by exceeding the
confidence limits.   It is precisely this "advance warning" feature of
these techniques that make them attractive in the field of process
quality control monitoring.  In Figure 6, the line labelled "predicted
geometric mean" was calculated based on the Colorado State Implementa-
tion Plan.  The fourth quarter of 1970 is the starting point for both
the observed and predicted annual geometric means because 1970 was used
as the base year in the implementation plan.  From that point on, both
the observed and predicted geometric means increase through the first
quarter of 1973; then the predicted geometric mean decreases until it
achieves the NAAQS in the first quarter of 1976.  However, the observed
geometric mean exceeds the upper bound in the first quarter of 1972 and
continues to exceed the upper bound through the third quarter of 1973.
The observed value exceeded the upper confidence limit as early as the
first quarter of 1972, and the trend would have been apparent even
before that, although the analysis system was not available at that
time.

The quality control approach has obvious applications to air quality
maintenance planning.  The observed ambient data in an AQMA can be
monitored against the NAAQS, for example, or the trends in the air
quality of potential AQMA's can be observed relative to their possible
                                98

-------
  00
  i
     150
     140
     130
     120
     100
      90
      80
      70
              OBSERVED
              GEOMETRIC
              MEAN
                                            UPPER BOUND
                                            ON PREDICTED
                                            GEOMETRIC MEAN
                         PREDICTED
                         GEOMETRIC
                         MEAN
           1  2  3  4 5 6 7 8 9 10  12  14  16  18  20   22  24
                               11  13  15  17   19  21   23  25
            1970    1971
1972    1973    1974    1975
1976
Figure 10. Quality control approach applied to particulate data from
           Denver, Colorado.
                              99

-------
designation.  For best utilization, the approach does require compu-

terization, preferably with graphic output, so that it is a significant

effort.  However, as air quality maintenance planning becomes more

institutionalized over the years, some such capability should be ob-

tained as soon as possible.


11.5.4.  REFERENCES
1.  Curran, T. C. and Hunt, W. F.,  Jr., "Interpretation of Air Quality
    Data with Respect to the National Ambient Air Quality Standards,"
    Presented at the 67th Annual Meeting of the Air Pollution Control
    Association, Denver, Colo., 1974.  (Accepted for publication
    J.A.P.C.A.)

2.  Faoro, R., "Trends in Concentrations of Benzene-Soluble Suspended
    Particulate Fraction and Benzo(a) Pyrene Determined by Data from
    the National Air Surveillance Network," Presented at the 67th
    Annual Meeting of the Air Pollution Control Association, Denver,
    Colo., 1974.  (Accepted for publication J.A.P.C.A.)

3.  Frank, N., "Temporal and Spatial Relationships of Sulfates, Total
    Suspended Particulates and Sulfur Dioxide," Presented at the 67th
    Annual Meeting of the Air Pollution Control Association, Denver,
    Colo., 1974.

4.  Hunt, W. F., Jr. and Curran, T. C., "An Application of Statistical
    Quality Control Procedures to Determine Progress in Achieving the
    1975 National Ambient Air Quality Standards," 28th Annual Technical
    Conference Transactions, American Society for Quality Control,
    Boston, Mass., 1974.
                                100

-------
    APPENDIX A





STATISTICAL TABLES
     101

-------
Table A-l.
/i
4
5
6
7
8
9
10
II
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
For n
p = .900
.8000
.7000
.6000
.5357
.5000
.4667
.4424
.4182
.3986
.3791
.3626
.3500
.3382
.3260
.3148
.3070
.2977
.2909
.2829
.2767
.2704
.2646
.2588
.2540
.2490
.2443
.2400
greater than
QUANTILES OF THE SPEARMAN
.950
.8000
.8000
.7714
.6786
.6190
.5833
.5515
.5273
.4965
.4780
.4593
.4429
.4265
.4118
.3994
.3895
.3789
.3688
.3597
.3518
.3435
.3362
.3299
.3236
.3175
.3113
.3059
.975

.9000
.8286
.7450
.7143
.6833
.6364
.6091
.5804
.5549
.5341
.5179
.5000
.4853
.4716
.4579
.4451
.4351
.4241
.4150
.4061
.3977
.3894
.3822
.3749
.3685
.3620
30 the approximate quantiles
w .^- 	
.990

.9000
.8857
.8571
.8095
.7667
.7333
.7000
.6713
.6429
.6220
.6000
.5824
.5637
.5480
.5333
.5203
.5078
.4963
.4852
.4748
.4654
.4564
.4481
.4401
.4320
.4251
of p may
TEST STATISTIC3
.995


.9429
.8929
.8571
.8167
.7818
.7455
.7273
.6978
.6747
.6536
.6324
.6152
.5975
.5825
.5684
.5545
.5426
.5306
.5200
.5100
.5002
.4915
.4828
.4744
.4665
be obtained
.999



.9643
.9286
.9000
.8667
.8364
.8182
.7912
.7670
.7464
.7265
.7083
.6904
» .6737
.6586
.6455
.6318
.6186
.6070
.5962
.5856
.5757
.5660
.5567
.5479
from
where xv is the p quantile of a standard normal random variable obtained from
Table 1.

SOURCE.  Adapted from Glasser and Winter (1961), with corrections.
  " The entries in this table are selected quantiles wv of the Spearman rank correlation
coefficient p when used as a test statistic. The lower quantiles may be obtained from (In-
equation
                                   wp =  -*!_„
The critical region corresponds to values of p smaller than (or greater than) but not includ-
ing the appropriate quantile. Note that the median of p is 0.
                                102

-------
         Table  A-2.   QUANTILES  OF  THE  STANDARD
                        NORMAL DISTRIBUTION*1
w»
-3.7190
-3.2905
-3.0902
-2.5758
-2.3263
-2.1701
-2.0537
- .9600
- .8808
- .7507
- .6449
- .5548
- .4758
- .4395
- .4051
- .3408
- .2816
- .2265
- .1750
- .1264
- .0803
- .0364
-.9945
-.9542
-.9154
-.8779
-.8416
-.8064
-.7722
-.7388
-.7063
-.6745
-.6433
-.6128
-.5828
-.5534-
-.5244
-.4959
/'
.0001
.0005
.001
.005
.01
.015
.02
.025
.03
.04
.05
.06
.07
.075
.08
.09,
.10
.11
.12
.13
.14
.15
.16
.17
.18
.19
.20
.21
.22
.23
.24
.25
.26
.27
.28
.29
.30
.31
wp
-.4677
-.4399
-.4125
-.3853
-.3585
-.3319
-.3055
-.2793
-.2533
-.2275
-.2019
-.1764
-.1510
-.1257
-.1004
-.0753
-.0502
-.0251
.0000
.0251
.0502
.0753
.1004
.1257
.1510
.1764
.2019
.2275
.2533
.2793
.3055
.3319
.3585
.3853
.4125
.4399
.4677
.4959
l> w.
.32 .5244
.33 .5534
.34 .5828
.35 .6128
.36 .6433
.37 .6745
.38 .7063
.39 .7388
.40 .7722
.41 .8064
.42 .8416
.43 .8779
.44 .9154
.45 .9542
.46 .9945
.47 .0364
.48 .0803
.49 .1264
.50 .1750
.51 .2265
.52 .2816
.53 .3408
.54 .4051
.55 .4395
.56 .4758
.57 .5548
.58 .6449
.59 .7507
.60 .8808
.61 .9600
.62 2.0537
.63 2.1701
.64 2.3263
.65 2.5758
.66 3.0902
.67 3.2905
.68 3.7190
.69
/'
.70
.71
.72
.73
.74
.75
.76
.77
.78
.79
.80
.81
.82
.8.1
.84
.85
• .86
.87
.88
.89
.90
.91
.92
.925
.93
.94
.95
.96
.97
.975
.98
.985
.99
.995
.999
.9995
.9999

SOURCE.  Abridged from Tables 3 and 4, pp. 111-112, Pearson and Hartley (1962).
  o The entries in this table are quanliles wv of the standard normal random variablv M
selected so P(W £ *„) - p and P(W >Wf)-l-p.
                                 103

-------
Table A-3.  QUANTILES OF THE STUDENT'S t DISTRIBUTION

-------
Table A-4.   QUANTILES OF A CHI SQUARE DISTRIBUTION WITH
            ONE DEGREE OF FREEDOM
      Quantile, p
W
.750
.900
.950
.975
.990
.995
.999
1.323
2.706
3.841
5.024
6.635
7.879
10.830
                       105

-------
                              TECHNICAL REPORT DATA
                        (Please read Instructions on the reverse before completing)
 1. REPORT NO.

 EPA 450/4-74-012
           3. RECIPIENT'S ACCESSION" NO.
4. TITLE AND SUBTITLE
 Guidelines  for AQM Planning &  Analysis
 Air Quality Monitoring & Data  Analysis  (Vol  n
           5. REPORT DATE
              September  1974
           6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)
                                                  8. PERFORMING ORGANIZATION REPORT NO.
9 PERFORMING ORGANIZATION NAME AND ADDRESS

  GCA/Technology Division
  Burlington Road
  Bedford, Massachusetts  01730
           10. PROGRAM ELEMENT NO.

              2AH137
           11 CONTRACT/GRANT NO.
                                                     68-02-1478
 12. SPONSORING AGENCY NAME AND ADDRESS
 Environmental Protection Agency
 Office of  Air Quality Planning  & Standards
 Monitoring and Data Analysis Division
 Research Triangle Park, N.C.  27711
           13. TYPE OF REPORT AND PERIOD COVERED
           Rpt.  for Task  1;  7/74-9/74
           14. SPONSORING AGENCY CODE
15 SUPPLEMENTARY NOTES
16. ABSTRACT
 This report  contains guidance  concerning monitoring and air quality
 data analysis related to air quality maintenance.   Topical areas
 covered  include:   network design,  instrument  siting,  acceptable
 instrumentation,  monitoring site description, air  quality trend
 evaluation,  air quality data evaluation, interpretation of air quality
 as it relates national ambient air quality standards and procedures
 for validating, editing, and screening air quality data.

 The quantity,  type, temporal,  and geographical distribution of air
 quality  data necessary for establishing baseline air quality levels
 is presented.
                           KEY WORDS AND DOCUMENT ANALYSIS
               DESCRIPTORS
 Design &  Instrument Siting of Air
   Quality Monitoring Networks
 Air Quality  Data Evaluation
 Air Quality  Data Validation
 Baseline  Air Quality Data
                                       b.IDENTIFIERS/OPEN ENDEDTERMS
Air Quality Mainte-
 nance Guidance
Federally Required
 Air Monitoring
Monitoring Pursuant t
 State Impbmentation
 Plans
                         COS AT I Field/Group
  DISTRIBUTION STATEMENT
 Release Unlimited
                                       19 SECURITY CLASS (This Report)
                                        None_
20. SECURITY CLASS (This page}
 None
                       21. NO. OF PAGES
                           113
                       22. PRIOE
if :A Form 2220-1 (9-73)
                                       106

-------
ENVIRONMENTAL PROTECTION AGENCY
    Technical Publications Branch
      Office of Administration
 Research Triangle Park, N.C. 27711
        OFFICIAL BUSINESS

 AN EQUAL OPPORTUNITY EMPLOYER
     POSTAGE AND FEES PAID
ENVIRONMENTAL PROTECTION AGENCY
          EPA - 335

     THIRD CLASS BULK RATE
                               Return this sheet if you do NOT wish to receive this material r^
                               or if change of address is  needed . ~].  (Indicate change, including
                                   code.)
                                   PUBLICATION  NO.  EPA-450/4-74-012
                                             (OAQPS  NO.  1.2-030)

-------