United States       Office of Air Quality       EPA-450/5-79-004
            Environmental Protection  Planning and Standards      January 1979
            Agency         Research Triangle Park NC 27711

            Air
oEPA      Studies in the
            of the Photochemical
            Oxidant Standard

-------
                            EPA-450/5-79-004
Studies in the Review
of the Photochemical
   Oxidant Standard
               by

           William F. Biller

           68 Yorktown Road
        East Brunswick, N.J. 08816
        Contract No. 68-02-2589
      EPA Project Officer: Michael Jones
            Prepared for

  U.S. ENVIRONMENTAL PROTECTION AGENCY
      Office of Air, Noise, and Radiation
   Office of Air Quality Planning and Standards
  Research Triangle Park, North Carolina 27711

            January 1979

-------
                           ACKNOWLEDGMENTS
     The author wishes to express his appreciation to Mr.  Michael  Jones
of EPA, the Project Officer for the study, for his many useful  suggestions
and guidance.  The author is particularly indebted to Mr.  Thomas Feagans
of EPA who is responsible for the underlying ideas on the  risk  assessment
methodology which are given mathematical  expression in this report.
Acknowledgment and appreciation are also  extended to Messers.  Carl  Nelson
and Ted Johnson of PEDCo Environmental for their work on the application
of the Weibull distribution to hourly average ozone concentrations  which
is basic to much of the work in this report.

-------
                          TABLE OF CONTENTS

                                                                          Page
ACKNOWLEDGMENTS                                                            iii
LIST OF FIGURES                                                             V1-
LIST OF TABLES                                                            viii
1.0  INTRODUCTION                                                           1
2.0  CHANGE FROM A PHOTOCHEMICAL OXIDANTS TO AN OZONE STANDARD              2
     2.1  Background                                                        2
     2.2  Alternative #1 - Remain with the Photochemical  Oxidant            3
          Standards
     2.3  Alternative #2 - Change to Ozone National  Ambient Air             4
          Quality Standards
     2.4  Conclusions and Recommendations                                   5
     2.5  References                                                        6
3.0  STATISTICAL FORMS OF THE PHOTOCHEMICAL OXIDANTS STANDARD               7
     3.1  Background                                                        7
     3.2  Mathematical Treatment                                           10
     3.2.1  Distribution of Hourly Average Concentrations of Ozone         10
     3.2,2  Interrelationship of Concentration Level of Standard           16
            and Expected Exceedance Rate
     3.2.3  Statistical Behavior of the Extreme Values of the              19
            Hourly Average Photochemical Oxidant Concentrations
            Observed Over a Given Period
     3.3  Nonindependence of Ambient Concentrations  of Pollutants          36
     3.4  Comparison of Statistical and Deterministic Standards            47
     3.4.2  Comparison on the Basis of Fractional  Reduction in             56
            Precursor Emissions Required to Meet Standard
     3.4.3  The Highest Second Highest Concentration                       67
     3.5  Statistical Standard Based on the Daily Maximum Hourly           72
          Average Concentration
     3.5.1  Comparison of Hourly and Daily Maximum Hour Standards          73
            When Hourly Average Concentrations Exhibit No Dependence

-------
                      TABLE OF CONTENTS (cont'd)
     3.5.2  Comparison of Hourly and Daily Maximum Standards  When
            Hourly Average Concentrations Exhibit Dependence
     3.5.3  Comparison of Deterministic Hourly and Daily Maximum
            Standards
     3.6  Conclusions and Recommendations
     3.7  References
4.0  ASSESSMENT OF RISK ASSOCIATED WITH POSSIBLE CHOICES OF A PRIMARY
     NATIONAL AMBIENT AIR STANDARD
     4.1  Background
     4.2  Underlying Principles of Risk Assessment Method
     4.3  Mathematical Description of the Method
     4.4  Obtaining the PC(C) Distributions
     4.5  Determination of PC Functions for Ozone
     4.6  Example Application of Method
     4.7  Conclusions and Recommendations
     4.8  References
APPENDIXES
     APPENDIX A:
     APPENDIX B:


     APPENDIX C:


     APPENDIX D:
SUPPORT MATERIAL FOR SECTION 3
DERIVATION OF BASIC EQUATIONS FOR ASSESSING HEALTH
RISKS ASSOCIATED WITH ALTERNATIVE AIR QUALITY
STANDARDS
EXAMPLES OF COMPOSITE HEALTH EFFECT THRESHOLD
DISTRIBUTIONS FORMED FROM SIMPLE HYPOTHETICAL
DISTRIBUTIONS FOR INDIVIDUAL THRESHOLDS
DETAILS OF CALCULATIONS PERFORMED IN SECTION 4.5
EXAMPLE APPLICATION OF METHOD
Page

 80

 88

 91
 95
 96

 96
 96
100
107
118
124
129
131


132
136


146


152

-------
                                 LIST OF FIGURES


Figure                                                                       Page

 3-1       Three Weibull  Distributions with Expected Exceedances per Year      15
          of One at a Concentration of 0.1  ppm.

 3-2       Effect of Weibull  Distribution Shape Factor on Distribution of      27
          Highest Value.

 3-3       Probability Density Function of Annual Highest Concentration        29
          for Different  Values of Weibull Distribution Shape Factor.

 3-4       Distribution of Annual  Extremal Values for a Weibull  Distribution   30
          of Hourly Concentrations Assuming Independent Hours

 3-5       Probability Density Functions for Annual  Extremal  Values for a      31
          Weibull Distribuion of  Hourly Average Concentrations
          (Assumes Independence of Hours).

 3-6       Probability of a Given  Observed Number of Exceedances as a           50
          Function of the Expected Exceedance Rate.

 3-7       Probability of a Actual Exceedances Equal to or Above a Given       51
          Number as a Function of the Expected Exceedance Rate.

 3-8       Distribution of Second  Highest Value for  Exponential  Distribution   60
          of Hourly Average Concentrations.

 4-1       Hypothetical Threshold  Probability Density Function for a Health   104
          Effect with a  20% Chance of Being Real.

 4-2       Variation in Composite  Health Effect Threshold Probability          106
          Density Functions of Two Independent Density Functions as
          Different Probabilities of Existence are  Assigned  to  the Lower
          Threshold Function.

 4-3       Hypothetical Distribution of Hourly Average Concentration and      112
          the Corresponding Distribution of the Annual  Maximum  Hourly
          Average Concentration.

 4-4       Change PC Function with Change in Standard Concentration Level.     113

 4-5       P- Function for One Year Period for Different Values  of Weibull     114
             Scale Factor, k.

 4-6       PC Function for 5 Year  Period for Different Values of Wei bull       115
             Scale Factor, k.

-------
                         LIST OF FIGURES (cont'd)


Figure                                                                      Page


 C-l      Composite Threshold Probability Densities When Health              148
          Effects Are Certain to Exist.

 C-2      Composite Threshold Probability Densities When the                 149
          Existence of One or More Health Effects is Not Certain.

 D-l      Change in Function PQ(C) with  Change in CSTD for                   153

          Exponential Distribution.

 D-2      Graphical Illustration of p. Calculation.                          159

-------
                                LIST OF TABLES
Table                                                                       Page

 3-1     Levels of Standards with Expected Exceedances per Year Equal         17
         to One Which are Equivalent to a Standard with a Level of
         0.08 ppm but with Expected Exceedances per Year Greater than
         One.

 3-2     Levels of Standards with Expected Exceedances per Year Greater      18
         than One Equivalent to Standard with Level of 0.08 ppm and
         Expected Exceedance per Year Equal  to One.

 3-3     % Risk of m Exceedances/yr for E Expected Exceedances               22

 3-4     % Risk of m or More Exceedances/yr of a Given Standard Level         23
         for E Expected Exceedances.

 3-5     Relative Positions of Concentrations Corresponding to Various       32
         Expected Exceedance Rates for Different Weibull Distributions.

 3-6     Location of the Most Probable Value of mth Highest Concentration    33
         as a Function of Number of Hours, n and Weibull Shape Factor, k.

 3-7     Location of Expected Value of the mth Highest Concentration for     35
         an Exponential Distribution Relative to the Concentration with
         an Expected Exceedance Rate m for Two Levels of n.

 3-8     Comparison of Distributions of Annual Maximum Hourly Average        41
         Concentration as Determined by Two Related Hourly Distributions.

 3-9     Effective Increase in Wei bull Constant k Needed to Give Same        43
         Distribution of Maximum Value as n is Decreased - Base Case:
         k = 1.0.

 3-10    Effective Increase in Wei bull Constant k Needed to Give Same        44
         Distribution of Maximum Value as n is Decreased - Base Case:
         k = 0.5.

 3-11    Change in Weibull  Distribution Shape Factors with Changing Time     45
         Segment in Which Hourly Average Concentrations are Collected.

 3-12    Probability of a Given Observed Number of Exceedances for           52
         Different Expected Exceedance Rates.

 3-13    Comparison of Statistical Forms with a Deterministic Standard       54
         Which Allows No More than One Exceedance per Year.
                                    vm

-------
                           LIST OF TABLES (cont'd)


Table                                                                        Page

 3-14    Cumulative Distribution of Second Highest Concentrations             58
         for 60 AQCR's.

 3-15    Location of Most Probable Value of Second Highest Concentration      61
         Relative to the Concentration Whose Exceedance Rate is Two (C2).

 3-16    Change in Percent Control Required to Meet Standard at 0.08 ppm      62
         Level with Change in Allowed Expected Exceedances.

 3-17    Change in Percent Control Required to Meet Standard at 0.08 ppm      63
         Level with Change in Allowed Expected Exceedances.

 3-18    Change in Percent Control Required to Meet Standard at 0.08 ppm      64
         Level with Change in Allowed Expected Exceedances.

 3-19    Change in Percent Control Required to Meet Standard at 0.08 ppm      65
         Level with Change in Allowed Expected Exceedances.

 3-20    Change in Percent Control Required to Meet Standard at 0.08 ppm      66
         Level with Change in Allowed Expected Exceedances.

 3-21    Position of Most Probable Highest Second High Concentration of       70
         N Years Relative to Concentration with Two Expected Exceedances
         per year.

 3-22    Density Function for Distribution of Highest Second Highest          71
         Value in N Trials of 8760 Trials Each for Wei bull Distribution
         of Hourly Averages Where k = 1, C,. = 1.0, E = 2.

 3-23    Comparison of Daily Maximum Hourly Average and Hourly Average        78
         Standards.

 3-24    Comparison of Daily Maximum Average and Hourly Average Standards     79
         with Partial Correction for Nonstationarity in Hourly Average
         Concentrations.

 3-25    Distribution of Annual Six Highest Hourly Average Concentration      81
         by Region.

 3-26    Daily Maximum Hourly Average Standard versus Hourly Average          83
         Standard.

 3-27    Distribution of Exceedances per Year for Daily Maximum Hour          85
         Standard versus Exceedances per Year for Hourly Standard

-------
                          LIST OF TABLES (cont'd)


Table                                                                        Page

 3-28    Approximate Comparison of Statistical  Daily Maximum Hour and         87
         Hourly Standards When Both Have Expected  Exceedances per Year
         of One.

 3-29    Comparison of Percent Hydrocarbon Control  Required fay Daily          89
         Maximum Hourly Average and Hourly Average  Standards.


 4-1      Subjective Probability Distributions of Lower and Upper Bounds       121
         of Weibull Shape Factor k for n at Approximately 1000 Hours
         and Maximum Concentration at Approximately 0.1  ppm.

 4-2      Subjective Probability Distributions of Lower and Upper Bounds       122
         of Wei bull Shape Factor k For n at Approximately 1000 hours
         and Maximum Concentration at Approximately 0.1  ppm.

 4-3      Subjective Probability Distributions for Upper and Lower Bound       123
         Weibull Shape Factors (k) for Distributions of Hourly Average
         Ozone Concentrations.

 4-4      Risks of an Hourly Average Concentration  Exceeding True Health       127
         Effect Thresholds in a Given Year for Different Values of a
         National  Ambient Air Quality Standard.


 0-1      Individual and Composite Health Effects Threshold Probability       155
         Density Functions Vs. Concentration.

-------
               STUDIES IN THE REVIEW OF THE PHOTOCHEMICAL OXIDANTS
                     NATIONAL AMBIENT AIR QUALITY STANDARDS
1.0  INTRODUCTION
     The Clean Air Act requires that the U.S. Environmental Protection Agency
periodically review the five National  Ambient Air Quality Standards (NAAQS).
The Office of Air Quality Planning and Standards is responsible for conducting
this review and initiated the first round with a reconsideration of the stan-
dard for photochemical oxidants.  A complex schedule of activities had to
be completed in reviewing the photochemical oxidants standard and required
the analysis of a number of issues. The contractor contributed to the analyses
performed in three areas:  1. the desirability of changing the photochemical
oxidants standard to an ozone standard;  2. the suitability of alternate ways
of expressing the oxidant standard; and  3. the assessment of risk in setting
the standard.  The last analysis combined the considerations underlying sta-
tistical forms of the standard developed in the second analysis with the con-
sideration of the uncertainty in the location of health effect thresholds.
The fundamental idea of combining these two considerations to estimate risk
is due to Mr. Thomas Feagans of The Office of Air Quality Planning and Standards.
The contractor's contribution was development of the mathematical formalism
which followed from the concept and which could be used to make actual risk
estimates.
     The report is divided into three major sections covering the areas listed
above.   Each section ends with conclusions and recommendations for the study
covered.

-------
2.0  CHANGE FROM A PHOTOCHEMICAL OXIDANTS TO AN OZONE STANDARD
2.1  Background
      When primary and secondary national ambient air quality standards were
first established for photochemical oxidants, it was proposed that the potas-
sium iodide  (KI) method be designated as the Federal Reference Method (FRM)
for measuring photochemical oxidants.  The method had been widely used for
many years in making aerometric measurements of photochemical oxidants and
measuring photochemical reactions in laboratory irradiation chambers.   However,
concern over the reliability and accuracy of the KI method led to the  adoption
of the newly developed chemiluminescent method as the FRM*- -".
      The chemiluminescent method is specific for ozone which is the major
component of photochemical oxidants.  It does not measure the remaining oxi-
dants, the most prominent of which is peroxy acetyl nitrate (PAN).  The
inability to measure all, or nearly all of the oxidants present is an   unde-
sirable feature in an FRM for photochemical oxidants.  The decision to use a
measurement method specific for ozone was made necessary by the absence of
satisfactory alternative methods.  Furthermore the prospects are poor  for
developing a satisfactory photochemical oxidants measurement method.  There
are a large number of individual compounds that potentially could be classified
as photochemical oxidant.  To date only a relatively small number have been
identified in the atmosphere or in laboratory irradiation chambers.  Developing
a practical measurement technique that would elicit a uniform response from
such a complex class of compounds would be very difficult.
      This discrepancy between the standard and the FRM has raised the question
as to whether the present standard is not, in effect, an ozone standard because
of the FRM and has led to suggestions that the standard be changed to  ozone.

-------
Two alternatives were considered in the present review of the photochemical
oxidants standards:  (1) Remain with the present standards and FRM with ozone
as a surrogate for photochemical oxidants, and (2) Change to ozone standards.
These alternatives are discussed in the following paragraphs.
2.2  Alternative #1 - Remain with the Photochemical Oxidant Standards
      Remaining with the present standards has the advantage of continuing to
emphasize the importance of protecting against the undesirable effects of all
photochemical oxidants, ozone as well as the non-ozone oxidants.
      The major difficulty is that ozone has problems as a surrogate for
photochemical oxidants.  It is not an accurate indicator of the total oxidants
present and, perhaps more important, it provides little indication of the non-
ozone oxidants present in the ambient air.  Ambient ozone concentrations can
vary from somewhat less than two-thirds to greater than 99% of the  total
photochemical oxidants concentration.  The ratio of PAN (the second most
                                                                      f2l
predominant photochemical oxidant) to ozone can vary from 1/3 to 1/150L J
from one geographic location to another.  It is likely that the ratios of
ozone to other photochemical oxidants also vary over wide ranges.
      Another difficulty with the present standard is that, aside  from ozone,
not much is known about the health and welfare effects of photochemical oxidants,
Examination of the newly revised criteria document^ -" for photochemical oxidants
shows the majority of studies to be concerned with the effects of  ozone.   While
early epidemiological  studies which were the basis for the present standard^ *
used the KI method (which is not a reliable measure for total  oxidants) to
estimate total  oxidants,  more recent studies tended  to emphasize  the use of
chemiluminescence and  other methods which are specific for ozone.

-------
2.3  Alternative #2 - Change to Ozone National  Ambient Air Quality Standards
      A change to ozone standards would resolve the major problems with the
present standards.  It also would have the added features that the Federal
Reference Method and a widespread ozone monitoring network are already in
place.  Ozone is a major atmospheric pollutant  in its own right.   It is the
dominant component in photochemical  oxidants (60 to 99%); and there is a
large and growing body of information on its health and welfare effects.  As
indicated in the previous section, the new criteria document"- -" is primarily
concerned with ozone; the clinical and toxicological  studies discussed in the
document deal almost exclusively with ozone.
      While a change to ozone standards would result in a loss of overt
emphasis on controlling photochemical oxidants  as a class, it is unlikely
to result in increased risk of adverse effects  to the public from photochem-
ical oxidants or other products from atmospheric photochemical reactions of
ozone precursors.  A study of the new criteria  document indicates that the
principle health effects associated with photochemical oxidants are also
caused by ozone alone and may be due primarily  to the presence of ozone.  For
example, the epidemiological study by Toyama and Kagawa in which both ozone
and photochemical oxidants were measured showed that ozone concentrations
correlated more strongly with observed health effects than did total oxidant
as estimated by the KI method*- \
      A national ambient air quality standard for ozone would be based upon
the effects of ozone in the ambient air.  This  means that estimations of
effects thresholds and margin of safety considerations would take into account,
to the extent allowed by available information, additive and interactive
effects with non-ozone oxidants and other compounds present.

-------
      Finally, precursor emission control measures for the reduction of ozone
in the atmosphere would also generally reduce the levels of the non-ozone
oxidants and other compounds generated by the photochemical reactions of the
          T41
precursors'-  .  For example, studies suggest that hydrocarbon control measures
designed to reduce ozone would produce a larger reduction in PAN   .   Thus,
there will be a reduction not only in the contribution from non-ozone oxidants
and related photochemical pollutants to effects primarily caused by ozone,
but also to effects, such as eye irritation and visibility reduction, which
are largely due to non-ozone oxidants and other photochemical products.  At
the anticipated concentration levels of an ozone standard, these latter effects
should be adequately controlled, based on present information.
      It is expected, therefore, that ozone ambient air standards would pro-
vide the same degree of protection as photochemical oxidants standards.  Futher-
more, it would not be necessary to make any changes in regulatory strategies
or programs as a direct result of a change from the one standard to the other.
This latter conclusion also follows in part from the fact that current strate-
gies  and programs are already based on reduction of ozone levels in  the
atmosphere.

2.4  Conclusions and Recommendations
      Given:   1. the difficulties with the present photochemical oxidants
standards;  2. the suitability of ozone as a basis for air quality standards,
3. the equivalent protection provided by an ozone standard, based on  present
knowledge, and  4. the absence of any impact on regulatory strategies and
programs, it is concluded that changing from the present primary and  secondary
Photochemical  oxidants standards to ozone national ambient air Quality standards
could lead to  benefits in administering the standard and little or no loss in
protection of the public health and welfare.

-------
2.5  References
1.  "Air Quality Criteria for Photochemical  Oxidants  and Oxidant  Precursors",
    Vols. I and II,  U.S.  Environmental  Protection  Agency,  Research Triangle
    Park, N. C. 27711.

2.  "Analysis of Photochemical  Oxidant  Mix  in  the  United States", Report  to
    EPA in preparation, PEDCo Environmental, Inc.

3.  "Air Quality Criteria for Photochemical  Oxidants",  Report No. AP-63,
    U.S. Department of Health,  Education, and  Welfare.

4.  "Control of Photochemical Oxidants  - Technical Basis and  Implications of
    Recent Findings", U.S. Environmental Protection Agency, Office of Air
    Qualtiy Planning & Standards, Research  Triangle Park,  N.C.  27711, July 15,
    1975.

5.  W.A. Lonneman, J.J. Bufalini, and R.L.  Seila,  "PAN  and Oxidant Measurement
    in Ambient Atmospheres", Environmental  Science and  Technology, Vol. 10,
    No. 4, 374 (1976).

-------
3.0  STATISTICAL FORMS OF THE PHOTOCHEMICAL OXIDANTS STANDARD
3.1  Background
     The present short-term National Ambient Air Quality Standards (NAAQS)
established by the Environmental Protection Agency all  specify a concentration
level which is not be exceeded more than once in any year at a given air
monitoring site.  Compliance with the standard is simply determined by count-
ing the number of times the  specified concentration level  is exceeded in a
calendar year at a given monitoring site.  If the number is greater than one
the Air Quality Control Region (AQCR) in which the monitor is contained is
not in compliance.  Alternatively, the second highest concentration (averaged
over the number of hours specified by the standard) observed during the year
at each monitoring site is recorded.  If any of the second highest values are
above the standard level, the level has been exceeded more than once at some
site and the AQCR is not in compliance.  The highest second high concentration
of the pollutant for all the monitoring sites within an area is considered
to be a measure of the air quality over the area relative to that pollutant.
This number can also be used as a "design value" to calculate how much emission
reduction is needed to meet the standard.  For photochemical  oxidants propor-
tional rollback, Appendix J , or the recently developed Empirical Kinetic
Modeling Approach  are used in performing this calculation.
     A single exceedance of the specified concentration level was provided for
in the standard to allow for  the  occasional  occurrence of weather conditions
and/or emission conditions leading to an unusually high peak pollutant level
and also to allow for an occasional monitoring error.  However, there are
problems with attempting in this manner to take into account the random
changes in atmospheric pollutant concentrations caused  by random fluctuations

-------
                                      8
in weather and in emissions.   For example,  if certain meteorological  conditions
leading to concentrations above the standard can occur once in a given year,
there is a definite probability they can occur more than once in other years
and, therefore, lead to violations of the standard.  It follows that  the
present deterministic standards can only be complied with on a continuing
basis if the specified concentration level  is never exceeded.  Thus,  the stan-
dard could not be exceeded even in years in which very rare adverse weather
conditions occurred.  Enforcing compliance  under such rarely occurring condi-
tions can require emissions to be reduced to levels far below levels  that
could be tolerated in more typical and  frequently occurring years.
     Another major difficulty is that the deterministic form of the present
standards does not lend itself to the development of meaningful measures of
existing air quality over an  AQCR which can then be compared with the standard.
As indicated above, the present practice is to use the annual second  highest
concentration measurement.  However, the second high value is itself  a random
variable and will vary in a given area from year to year even though  the
average level of emissions leading to the observed atmospheric concentrations
of the pollutant remains constant.  This situation can result in uneven treat-
ment from one AQCR to the next.
     Present EPA guidelines recommend that several years of data be obtained
and that the highest observed second high value be used.  This guidance is
in keeping with a strict interpretation of the deterministic standards and
gives a more stable number than the second highest value from a single base
year.  However, the highest second high of a small number of years is still
a random variable and, therefore, subject to change with fluctuations in weather
and emissions.

-------
     The way out of these difficulties is to express the standard in a
statistical form.  For example, a statistical form that would result in the
least departure from the present expression for photochemical oxidants is:

       160 yg/m  (0.08 ppm) hourly average concentration with an
      expected number of exceedances per year less than or equal to one.

By setting the maximum permitted expected or average number of exceedances
per year at one, allowance is made for occasional years in which the standard
level is exceeded more than once so long as the average number of exceedances
per year in not greater than one.
     Furthermore, the statistical standard provides a meaningful method of
comparing existing air quality with the standard.  That concentration which
has an expected exceedance rate of once per year at the existing air quality
level can be compared directly with the standard level.  This concentration,
which can be designated as C,, would be determined for each monitoring site
within the AQCR and the highest number used as a design value for calculating
the degree of emissions reductions needed to attain the standard.  This follows
because the concentration, C,, unlike the second high concentration, is not a
random variable.   Instead C-. has a single value corresponding to the average
level of the emissions giving rise to the observed atmospheric concentration
of the pollutant.
     The value of C,  for an AQCR would be estimated from one or more years
of data.    The estimate, of course, could fluctuate as the data base is
expanded or reduced.   However, the value of C,  can be estimated from the overall
distribution of hourly average concentrations,  rather than determined as a
single value as with  the second high value.   The estimate of C-,  is  less subject,

-------
                                        10
 therefore, to random fluctuations than the second high value.   It is also
 less subject to experimental   error and less influenced by missing values
 than the second high concentration.
      Other statistical  expressions could also be used.  For example, the
 standard could specify the maximum permitted value of the expected maximum
one hour average concentration in one year, or it might specify the maximum
 permitted probability that a  certain one hour concentration level will  be
 exceeded in a given year.   However, estimating the appropriate statistical
 measure of air quality in  these cases is more complex than for the standard
 based on an expected exceedance rate.
      The contractor under  EPA Contract No. 68 02 2393 investigated the  rela-
                                                       F3T
 tive merits of deterministic  and statistical standards1- J.  These studies
 were extended under the present contract.  The results are discussed in the
 following paragraphs.

 3.2  Mathematical  Treatment
      Before proceeding to  discuss the various aspects of statistical standards
 studied, it will be worthwhile to put in one place the mathematics used in the
 treatment and to discuss the  behavior of measured atmospheric  concentrations
 of ozone   (present monitoring instruments predominately measure ozone),
      3.2.1  Distribution of Hourly Average Concentrations of Ozong
           The distribution of hourly average ozone concentrations over  one
 year may be defined as follows:
           F(C)  =  Pr (CQbslC)                                            (3-1)

-------
                                     n
where F(C) may be considered the distribution function, Pr refers to the
probability of the relationship within the brackets; CQbs is an observed
hourly average concentration, and C a continuous variable representing hourly
average concentration.  The probability density function associated with F(C)
is by definition:

         f(C)  - ^                                                         C

It has the property that:
             f (c)dC  -  PrtC  < C    < C2)                                  (3-3)
         cl
     Because the standard is concerned with the behavior of the highest
hourly average concentrations in any one year a more conveneient form of the
distribution function is:
         6(C)  -  Pr(CQbs > C)                                              (3-4)
It follows from the theory of probability that:

         G(C)  =  1 - F(C)                                                  (3-5)

Strictly speaking a  21 si9n should be used in Eq.  (3-4),  but since the
function is defined as continuous there is no loss in accuracy in  using the
above more convenient definition.

-------
                                  12
                      F41
     Studies by Larson1- J and others have shown that the distributions of
short-term averages concentrations of atmospheric pollutants are described
reasonably well by lognormal distributions.  It had been observed, however,
that the lognormal distribution often did not give good fits on the high
concentration end to the photochemical oxidants hourly average concentrations.
Curran and Frank    suggested that exponential and Weibull distributions were
useful alternatives to the lognormal distribution.  Studies by PEDCo Environ-
mental, Inc.  •" of oxidant/ozone air monitoring data from a number of loca-
tions confirmed that the Weibull distribution gave a better fit than lognormal
in most cases.  The Weibull distribution expressed in the form of Eq. (3-4) is:

         6(C)  »  e"(C/6)                                                   (3-6)
The equation has two parameters, 6 and k.  The parameter 6 is called the
scale factor.  It is the concentration corresponding to G(C) = 0.368.  It,
therefore, tends roughly to place the midportion of the distribution.  The
parameter k is called the scale factor.  It tends to be a measure of the
spread of the distribution.  The larger k, the more compact the distribution.
PEDCo found that k values vary from one region to the next and may appear to
vary within an area from one year to the next.  For a full year of data
(8760 hours) the observed range is about  k = 0.5 to 2.3.  The exponential
distribution referred to above is a special case of the Weibull distribution
obtained by setting k = 1.  In this case it can be shown that 6 becomes
the mean concentration of the distribution.
The density function is found by differentiating Eq.  (3-6):

          f(C)  =  (k/S) (C/fi)1""1 e"(C/5)k                                     (3-7)

-------
                                       13
     Equations (3-6) and (3-7) can be put In a form that is particularly
convenient for dealing with a statistical standard which specifies a con-
centration level  with a specified expected exceedance rate.  From Eq.  (3-6)
it can be written that:
                                                                           (3'8)
Where E is the average or expected number of times the concentration CE
will be exceeded in n hours.   Because, in the case of photochemical  oxidant
concentrations each hour in the year is not equivalent to every other hour,
n is restricted to be a multiple of the number of hours within the year to
which the distribution applies.  Normally, this would be a multiple  of
8760, the total hours in a 365 day year.
     E is the expected number of exceedances in n hours because the  proba-
bility, 6(C), can be interpreted as the expected frequency of occurrence
of hourly average concentrations above C  that will be approached as  the
number of hours increases without limit.   If n is the number of hours in
one year and actual exceedances are observed for a large number of years
their average value would therefore approach E in Eq. (3-8) as the number
of years increased without limit.  The quantity E is, therefore, the
expected exceedance rate corresponding to n hours in a year.
     Continuing with the development of more convenient forms of Eq.  (3-6)
and (3-7):  Eq. (3-8) can be  solved for the scale factor S:

                    _ 1_
       6  *  CcOn£)  k                                                    (3-9)

-------
                                      14
and the Weibull distribution, Eq.  (3-6)  can be written:
         G(C)  = e-          E                                              (3-10)
The corresponding density function becomes:
         f(C)  .    dn) (C/CE)    e           E                           (3-11)
     Figure 3-1 shows three Weibull  distribution functions.   For each curve
C  = 0.1 ppm, E = 1, n = 8760.   Their shape factors vary from 0.5 to 2.0.
Over this range the degree of tailing in the high concentration range is seen
to vary substantially.  It is also noted that plotted on a semi logarithmic
paper the Weibull distribution with k = 1 (exponential distribution) plots
as a straight line.   Special graph papers can be constructed which will
give a straight line for any value of k (log log 1/G vs. C*-  •*.
     A distribution of hourly averages with a shape factor k just meeting
a standard expressed as
                  3
         C_TD yg/m  (or ppm) hourly average concentration with the
         number of expected exceedances per year less than or equal
         to
would then have the form:

         6{C)  = e"0       (C/CSTD)                                        (3-12)

-------
                                                         15
   10
      -1
   10-2
o
3
CT
O
s_
   10
      _ o
     ,-5
                                                                                 gaE3V;i
                                                                                       \. - ^=r^ t


                                                          	.._.._^	y_, ._.-_ „	-..:_.	 _....; ™ — 	 ~  -i^~  ~"\ '.' ^r_- \—rr~
                                                          r -"~~"L~  ]     i ~"!~'~~     _!	^r.T.r"^r^L~r   -   '  II"~'~L ~ \ ~-~ ". ~~ ~^c.n_
                                                                '               f               ',        \     ~ \
                              0.02
0.04          0.06           0.08
      Concentration (ppm)
0.10
0.12

-------
                                     16
     3.2.2  Interrelationship of Concentration Level  of Standard and
            Expected Exceedance Rate
         Given the statistical  form of the standard shown in the last section
it would appear that there is a trade-off relationship between the specified
concentration level, C_TD, and the expected exceedance rate E-T_.   That is,
E _- could be increased and CSTD lowered while maintaining the same level  of
protection .   For example, if it is assumed that the Weibull function
Eq. (3-10) adequately describes the distribution of hourly average concentra-
tions over a monitoring site and the standard concentration level  and expected
exceedance rate are given as (CSTD» ^cjn)'   Tne area around the monitoring
site is just in compliance if:
         ESTD  .   -(lnn.) (C/C)                                         (3-13)
or
                                                                            (3-14)
                        STD
     Thus, any combination of C£ and E satisfying Eq.  (3-14) could serve as
alternate equivalent specifications for the standard.   However, the inter-
relationship is a function of k.  The practical  significance of this is that
the effect of a trade-off between concentration  level  and exceedance rate
would not be uniform for all monitoring sites since k  is known to vary from
site to site. Tables 3-1 and 3-2 show the trade-off for a base case of C^yp
0.08 ppm, E<.Tn = 1.  It is seen that the relationship  is strongly influenced
by the value of the shape factor k.

-------
                                              17
       Table  3-1   Levels  of  Standards  with  Expected  Exceedances  per  Year  Equal  to

                  One  V'hich  are  Equivalent  to  a  Standard with  a  Level  of  0.08  ppm
                  but  with Expected  Exceedances  per  Year Greater than  One
   Expected  Exceedances
Equivalent Levels for Standards With
Expected Exceedances/yr = 1,  (ppm)
icc c cac i-< i UN o ta nuai u
Level = 0.08 ppm
1
2
3
4
5
6
7
3
9
10
1
5
10
15
20
25
30
35
40
45
50
*(1)= 0.5
0.080
0.094
0.104
0.111
0.118
0.124
0.130
0.135
0.139
0.144
0.080
0.118
0.144
0.162
0.178
0.192
0.205
0.216
0.227
0.237
0.2*7
0.75
0.080
0.089
0.095
0.100
0.104
0.107
0.110
0.113
0.116
0.118
0.080
0.104
0.118
0.128
0.136
0.143
0.150
0.155
0.160
0.165
0.170
1
0.080
0.087
0.091
0.094
0.097
0.100
0.102
0.104
0.106
0.107
0.080
0.097
0.107
0.114
0.119
0.124
0.128
0.132
0.135
0.138
n.141
1.5
0.080
0.084
0.087
0.089
0.091
0.093
0.094
n.095
0.096
0.097
0.080
0.091
0.097
0.101
0.104
0.107
0.109
0.111
0.113
0.115
0.116
2
0.080
0.083
0.085
0.087
0.088
0.089
0.090
0.091
0.092
0.093
0.080
0.088
0.093
0.096
0.098
0.100
0.101
0.103
0.104
0.105
0.106
(1)  k = constant  in Heibull Oistribution.  Effective hours in year taken as 8760

-------
                                       18
      Table 3-2  Levels of Standards with  Expected  Exceedances per Year
                 Greater than  One  Equivalent  to  Standard with Level of
                 0.08 ppm and  Expected  Exceedance per  Year  Equal to One.
   Expected                        Equivalent  Levels  of  Standards  (ppm)
Exceedances/yr   k^=    0.5         0.75           1           1.5
1
2
3
4
5
6
7
8
9
10
0.080
0.068
0.062
0.057
0.054
0.052
0.049
0.048
0.046
0.045
0.080
0.072
0.067
0.064
0.062
0.060
0.058
0.057
0.055
0.054
0.080
0.074
0.070
0.068
0.066
0.064
0.063
0.062
0.061
0.060
0.080
0.076
0.073
0.072
0.070
0.069
0.068
0.067
0.067
0.066
0.080
0.077
0.075
0.074
0.073
0.072
0.071
0.070
0.070
0.069
1
5
10
15
20
25
30
35
40
45
50
0.080
0.054
0.045
0.039
0.036
0.033
0.031
0.030
0.028
0.027
0.026
0.080
0.062
0.054
0.050
0.^7
0.045
0.043
0.041
0.040
0.039
0.038
0.080
0.066
0.060
0.056
0.054
0.052
0.050
0.049
0.047
0.046
0.046
0.080
"."70
0.066
0.063
0.061
0.060
0.059
0.057
0.057
O.n56
0.055
0.080
0.073
0.069
0.067
0.065
0,064
0.063
0.062
0.062
0.061
0.060
   (l)k = constant in Wei bull  Distribution.   Effective  hours  per year  taken as  8760

-------
                                     19
     Because of this strong dependence on k,ambient air quality standards
with high values of E will on a U. S. wide basis provide less protection
from high concentrations of pollutant than a standard with a low value of E.
If k were uniform across the U. S. the exchange relationship Eq. (3-13) could
be used without any degradation in degree of protection from extreme concen-
trations.
     3.2.3  Statistical Behavior of the Extreme Values of the Hourly Average
            Photochemical Oxidant Concentrations Observed Over a Given Period
     Short-term standards such as the photochemical oxidants standard are set
for those pollutants for which short-term exposures are sufficient to cause
undesirable effects.  The standards are, therefore, designed to provide assur-
ance that even the highest atmospheric concentrations observed in a one year
period will not constitute a threat to the public health.
     The short-term standards, therefore, focus on the highest concentrations
likely to be observed in a given period, generally one year.  It is of interest
then, to study the behavior of the highest hourly average concentrations.  In
particular it is desired to know the distribution and probability density
functions for the m th highest observation in n hourly average concentrations
(n=8760 for one year).  From these distributions one can calculate the median
and other fractiles, the mean, the most probable value, and moments about the
mean.
     The distribution and density functions can be readily derived if the
probability of occurrence of any hourly average concentration  level  is the
same for each hour of the year.   This is not the case for photochemical
oxidants.  Concentrations are strongly influenced by the time of day  and time
of year.  Furthermore, the concentrations on successive hours are strongly
correlated and peak hourly average concentrations on successive days  tend to

-------
                                    20
correlate as well.   Put in another way,  the underlying random or stochastic
process giving rise to the fluctuations  in oxidant concentrations is  non-
stationary and autocorrelated.   While a  start has been made"-  -* in accounting
for nonstationarity and autocorrelation  their full  effect on  the behavior
of hourly average concentrations is still  not well  understood.
     Although the results are only approximate it is  instructive to study
the statistics of the extreme values assuming the hourly average concentra-
tions are independently and identically  distributed.   The starting point
for the analysis will be the following basic relationship derived from
                  f8l
probability theory1 J.
         r.         n!      m ,„   .n-m
         Pm  =  i^mTP  d'P)
where P  is the probability that a certain event will  occur m times in n
independent trials, and where p is the probability of occurrence for a single
trial.   The collection of values, PnjPT '"  etc-> are referred to as the
binomial distribution.
     If we set G(C) = p, and take n as the number of hours in one year then
P  becomes the probability that the concentration C will  be exceeded m times
 m
in one year.
     We are also interested in the probability that a given concentration
will be exceeded m or more times in a given year.  If this probability is
designated P   it follows that:
                      m-1
           Pnrf  =  ]  '   I  Pv                                                  (3-16)
                     v=0
Where PV is given by Eq. (3-15).

-------
                                       21
     Note that In using Eqs. (3-15) and (3-16) the whole distribution need
not necessarily be defined.  The equations can be used for specific probabil-
ities.  For example, the probability for exceeding the concentration C£ is,
based on the preceeding discussion, E/n.  This quantity can then be substituted
directly into Eq. (3-16) to calculate the probability of observing m actual
exceedances in a year of a concentration level corresponding to an expected
exceedance rate of E.  Equation (3-16) can then be used to calculate proba-
bilities for m or more actual exceedances.  Since these calculations can be
useful, the equations explicitly showing the expected exceedance rates are
given here:

                   i        m      c  n~m
        P.  •  ^^T <^>  0  - fy                                       (3-17)
and
                   v=0
The term n^ is used in both equations since it specifically refers  to the
period of time over which the expected exceedance rate E occurs.  The term n
can beany number of hours and may be, for example, a multiple number of years.
     Table 3-3 shows calculated P  for a range of m and E for n^  =  n = 8760.
Table 3-4 shows calculated values for P   for both n.. and n equal to 8760 and
500.  It is seen there is little difference between the two sets  of probabili-
ties in Table 3-4.   In fact,  to a good approximation for large n  and n£ and
small E and m Eq. (3-16)  reduces to the following approximation relationship:

-------
                                22
Table 3-3  % Risk of m Exceedances/yr for E  Expected Exceedances

m E = 1 2
0 36.79 13.53
1 36.79 27.07
2 18.40 27.07
3 6.13 18.05
4 1.53 9.03
5 0.31 3.61
6 0.05 1.20
7 0.01 0.34
8 <0.01 0.09
9 0.02
10 <0.01
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
n = n = !
3
4.98
14.93
22.41
22.42
16.82
10.09
5.05
2.16
0.81
0.27
0.08
0.02
0.01
<0.01












3760
4
1.83
7.32
14.55
19.55
19.55
15.65
10.44
5.97
2.99
1.33
0.53
0.19
0.06
0.02
0.01
< 0.01











5
0.67
3.37
8.42
14.04
17.56
17.57
14.65
10.47
6.55
3.64
1.82
0.83
0.35
0.13
0.05
0.02
< 0.01










10
<0.01
0.05
0.23
0.75
1.89
3.78
6.31
9.03
11.30
12.57
12.58
11.45
9.55
7.36
5.26
3.51
2.20
1.29
0.72
0.38
0.19
0.09
0.04
0.02
0.01
<0.01

-------
                               23
Table 3-4  % Risk of m or More Exceedances/yr of a  Given
           Standard Level for E  Expected Exceedances
m = 1
                                                                  10

1
2
3
4
5
10

63.2
86.5
95.0
98.2
99.3
> 99.9
n
26.4
59.4
80.1
90.8
96.0
99.9
E = n = 8760
8.0
32.3
57.7
76.2
87.5
99.7

1.9
14.3
35.3
56.7
73.5
99.0

0.4
5.3
18.5
37.1
56.0
97.0

< 0.1
< 0.1
<• 0.1
0.8
3.2
54.2
                            =  n  =  500
1
2
3
4
5
10
63.2
86.5
95.1
98.2
99.3
>99.9
26.4
59.4
80.2
90.9
96.0
>99.9
8.0
32.3
57.8
76.3
87.7
99.7
1.9
14.3
35.3
56.7
73.6
99.0
0.4
5.2
18.4
37.1
56.0
97.2
< 0.1
< 0.1
0.1
0.8
3.1
54.3

-------
                                   24
where:
The derivation of the expression (3-18)  is given in Note 1  of Appendix A.
     Coming back to Eq.  (3-15),  if P  is obtained for m = 0 the following
simple expression results:

            Pg  -  (1 -  P)n                                                 (3-21)
or
            P0  =  (1  - G(C))n                                              (3-22)
where G(C) is the distribution function for the hourly average concentrations.
The PO of Eq. (3-22) is the probability that the concentration C will  not
be exceeded in n hours.  It is, therefore,  also the distribution function for
the highest observed concentration in n observations.    That is, if we define
F(C/ N) as the distribution function for the mth highest value in n hours
(observations) then:
                      =  PQ                                                 (3-23)
To obtain F(C/ \) for any m it is noted that all  possible cases in which the
mth highest value does not exceed a given concentration level  are contained
in those cases in which there are 0, 1, 2, 3, 	,  or m-1  exceedances.   The

-------
                                    25
probabilities for these cases are PQ, P.J, •••, and Pm_.j .   By the additive
law of probability then:
                        m-1
            F(C(m))  =  I=PV                                              (3-24)
also from Eq. (3-16)
Eq. (3-24) may be written more explicitly as:
                                      pV °-p)n"V                           (3-26)
To obtain the corresponding probability density function Eq. (3-26) is
differentiated with respect to C to yield:
            f(c> =
The median of the distribution for the mth highest concentration is obtained
by substituting G(C) for p in Eq.  (3-26) and finding the value of C which
yields F(C/_\) = °-5-  The  mean  is the expected value of C, « and by
definition is:
                      c f(c(m))dc                                           (3-28)
                    0

-------
                                     26
     The most probable value, or mode, of the mth highest value is found
by setting derivative of f(C, J equal to zero and determining the value
of C/ \ which satisfies the expression.
    (m)
     Forthe Weibull distribution the distribution of the highest value in
n hours is found by substituting Eq. (3-10) into Eq. (3-24) (or in this
case, Eq. (3-21), which yields:
The probability density function is obtained from Eq. (3-25):
                                                  n          n-1
               \  -  nk n   l\ inr ^k'1 fi ^-On-t) (C/CE)\
                  "  C~ (ln  ~)    {        1"                '
                                                                            (3-29)
                              (C/CE)k                                       (3.30)
    Figure 3-2 shows the distribution of the annual maximum value for a
Weibull distribution of hourly average concentrations, Eq.(3-29), for
several values of the shape factor k.  Each curve has the same value for
r , E, and n and, therefore, goes through the same point at C = 1.   Note,
the effect of increasing k is to narrow the range of the distribution.
Because of the term C/Cr in Eq. (3-29), and the use of C  = 1, the curves
in Fig. 3-2 are easily transformed to any other value of Cp by multiplying
the concentration axis values by the new Cr.  In effect the concentration
axis in Fig.  3-2 could be labeled relative concentration.  It is relative to
the value of C  as plotted.

-------
                                                                           27
 CT1
 O
 i.
^->
 Ul
•f~
o

 C
 o

 1-
 o

 u
 «J
 a;
 o.
 (T3
 3
-Q
 o
 O!
OJ
 I
ro

 O)
 s.
 3
 cn
                                                                                                                           CM
                                                                                                                                 LU
                                                                                                                                o
                                                                                                                           ^  O
                                                                                                                                OJ
                                                                                                                               o:
                                                                                                                           cn
                                                                                                                    —    CO
I
c
—
1 t i
cc
o
! I
U3
G3
1 1
T
C
[
CM
O
1
C
C

-------
                                     28
     Figure 3-3 shows the corresponding density functions Eq.  (3-30).   It
is seen that the most probable value of the maximum concentration  is  very
close to C = 1.  In fact, for k = 1, the most probable value of C  is  CE
for E = 1 (designated as C,).  For k less than 1,  the most probable value
is slightly below C^ and for k greater than 1  it is slightly greater  than
cr
     The cumulative distributions for the first, second,  •••-,  fifth  and
tenth highest hourly average concentrations in one year are shown  in
Fig. 3-4 and the corresponding probability density functions in Fig.  3-5
for a Weibull distribution of hourly averages with k = 1  and CE =  1,  E  *  1.
As indicated above the distributions for any other value  of CE  can be
obtained by multiplying the x-axis values by the new C^.   The maxima  in
Fig. 3-5 fall on the concentrations whose expected exceedance rates are m.
The values of these concentrations are shown in Table 3-5.
     It is shown in Appendix A that the most probable value of  the mth
highest hourly average concentration in one year is the concentration,  C^,
whose expected exceedance rate, E, is m for an exponential distribution
(Weibull distribution with k = 1).
     Table 3-6 shows the locations of the most probable value of the  mth
highest concentration as a function of the Weibull shape  factor for
n = nr = 500 and n = n  = 8760.  The lower value of n has application in
cases of nonindependence of hours and will be discussed in latter  sections.
Here it is of value because the position of the most probable value is
more sensitive to k the lower the value of n.   For k values above  0.75
the most probable value of the mth highest concentration  can be considered
equal to the concentration with an expected exceedance rate of  m with no
significant error.

-------
                                                                                   29
 3


•^»

 
c
      1-
      o
 u    u
 c    
-------
                                                                               30
.c
•r-
 i-    cn
+->    s_
 i/l    3
•r-    O
a    =
 0)
 at
•a

 0)
 a.
 OJ
•o
 o
 U.    CD
       C
 VI   •<-
 0)    E
 3    3
 i—    l/l
 IQ    1/1
  E    O
  QJ   -r-
  S_   4->
 i—   OJ
  ro   O
  3   C
  C   O
  c   o
  O    i-
        3
  C    O
  o   =
 J2

  s_
   I
  CO
   QJ
   s_

   Cl

-------
                                                                                   31

<  <   4-

 J-   >>
 O  »•"       O    OJ
 c  z   -o
 o         c
•f-  l+-    dj
-*->   O    0.
 O         QJ
 c   c   -o
 3   O    C
•M   XI    
O   ••-   <

-C
 O
 Lf)
 a>
 3
 in

-------
                                     32
Table 3-5   Relative Positions of Concentrations Corresponding to Various
           Expected Exceedance Rates for Different Wei bull Distributions
 E oected         Relative Concentration for Given Exceedance Rate
Exceedance                             Cr/C,
   Rate                                 E   '
    E
    1
    2
    3
    4
    5
   10
   25
   50
  100
k = 0.5
1.000
0.853
0.773
0.718
0.677
0.557
0.417
0.324
0.243
=1.0
1.000
0.924
0.879
0.847
0.823
0.746
0.645
0.569
0.493
=1.5
1.000
0.948
0.918
0.895
0.878
0.823
0.747
0.687
0.624
=2.0
1.000
0.961
0.938
0.920
0.907
0.864
0.803
0.754
0.702
       (l)Weibull Distribution Constants
               CE = I
               E  = 1
               n  = 8760
               k  = as given

-------
                                     33
Table 3-6   Location of the Most Probable Value of mth  Highest Concentration

            as a Function of Number of Hours,  n and Wei bull  Shape  Factor,  k
                     Concentration Level  of mth Highest  Value  Relative
                    to Concentration with Expected  Exceedance  Rate  of m

                     k =   O-5        a 1-0         =  1.5        = 2.0
      For n and n.. = 8760
                          0.977        1.000          1.003        1.003
      For n and n.- = 500
1
2
3
4
5
0.951
0.969
0.975
0.975
0.981
1.000
1.000
1.000
1.000
1.000
1.006
1.004
1.003
1.003
1.003
1.006
1.004
1.004
1.003
1.003
    10                    0.987        1.000          1.001        1.001

-------
                                   34
     The relative location of the expected or average  value  of  the mth
highest concentration is shown in Table 3-7.   The expected value  of  the
mth highest concentration is close to the concentration whose expected
exceedance rate is m.  However, it is always  distinctly higher  because
the distributions are skewed to the high side of the mode.

-------
                                     35
Table 3-7   Location of Expected Value of the mth Highest Concentration for
            an Exponential  DistributionM'Relative to the Concentration  with
                   an Expected Exceedance Rate m for Two Levels  of n
                   Relative Expected Value of mth Highest  Concentration
      Rank
n = n = 500
1.093
1.050
1.035
1.028
1.023
n = n = 8760
1.050
1.026
1.021
1.016
1.013
       1
       2
       3
       4
       5
      10                      1.014                   1.007
         (1)  Weibull  Distribution  with  k  =  1

-------
                                    36
     3.3  Norn independence of Ambient Concentrations of Pollutants
     The previous section in dealing with the behavior of time averaged concen-
trations of pollutants assumed that there existed no correlation between
the concentrations of a pollutant observed in one hour and  that observed
in another hour.  It further assumed that the underlying stochastic processes
giving rise to observed concentrations  were constant in time, or stationary.
This is not the case for most air pollutants.  In the case of ozone the
concentration in one hour tends to be dependent on concentations in previous
hours.  Furthermore, the peak concentrations observed in one day tend to
correlate with the peak concentrations observed the next day.  This correla-
tion can persist over several days.  There is a strong time dependence.
Ozone concentations over a twenty-four hour period show fairly well defined
patterns in urban areas which relate to the degree of sunlight and vehicle
use.  There can also be a dependence on day of the week.  And there is a
strong dependence on time of year, with the highest concentrations occurring
during the warmer months of the year in urban areas.
     The extent to which nonstationarity in the processes leading to observed
ambient concentrations of pollutants and autocorrelation of neighboring con-
centrations must be accounted for depends upon the form of the standard and
the information desired.  In the case of the proposed ozone standard there
is no problem in either determining compliance or degree of departure from
compliance.   In determining compliance either the average number of exceed-
ances per year of the standard or C-,,the concentration of corresponding to
an expected exceedance rate of one,needs to be determined.  In deciding  the
degree of departure from the standard,C, is needed.  The average exceedance

-------
                                    37
rate can be obtained by direct measure or from the distribution of hourly
average concentrations.  The concentration C, can be obtained from the distri-
bution.  The question of independence does not enter.
     If the standard  was  based on the expected annual maximum hourly average
concentration or the expected annual second highest hourly average concentra-
tion, the question of dependence  would have to be dealt with.  As seen in
Section 3.2.3 in order to estimate the expected maximum or expected second
high concentration it is necessary to know the distributions for the maximum
and second high concentrations respectively.  In order to derive the exact
distributions it would be necessary to fully specify the time dependence and
autocorrelation of the hourly average concentrations.  The problems of depen-
dence also becomes important in estimating the risk associated with alternative
standard levels (Section 4).
     Accounting for dependence in an exact manner would require the accurate
modeling of the nonstationarity and autocorrelation and would seem to present
formidable analytical difficulties.  The study performed under this contract
was confined, therefore, to approximate methods of dealing with the problem.
The simplest approximation, of course, is to assume dependence and apply the
expressions developed in Section 3.2.3.  An EPA study"- -" of the problem of
dependence associated with the daily maximum hourly average ozone concentra-
tion concluded that the major problem in assumption of independence was non-
stationarity.  If this is the case then the approximation can be used if
the highest concentrations in a one year period occur in a subset of the
hours in a year for which the underlying stochastic processes are relatively
stationary.   For ozone a possible subset would be the midday hours during the

-------
                                     38
summer months, for example, 11 A.M. to 4 P.M. June 15 through September 15.
This period would contain close to 500 hours.  (Aerometric data would need to
be examined to determine what size the subset needs to be to have reasonable
assurance of capturing the maximum concentrations.)  Over the June to Septem-
ber period the day to day generating processes can be  assumed reasonably
stationary.  The degree of correlation from one day to the next would be low
to moderate.  Within the four hour segment of any day the process could be
considered relatively stationary with moderate to strong correlation.  If
we assume the strong correlation within a day would not have a large effect
on the approximation, then the expressions in Section 3.2.3 can be used if
500 hours is substituted in the place of 8760 hours.
     In making the above mentioned substitution it must be remembered that
the distribution of hourly averages, G(C), now applies to the 500 hour segment
instead of the full 8760 hours in one year.  Thus, it will be necessary to test
whether the Weibull distribution is still applicable,and if it is, whether
there is a change in the Weibull constants, 6 and k.  (See Eq. (3-6)).  Since
6 is largely determined by the ambient air standard (Eq. (3-9)) the problem
is basically that of determining the effect on the shape factor k.
     The effect on k of using a subset of the 8760 hours in one year was
examined using two very approximate approaches.   In the first approach, it
was assumed that the 500 highest ozone concentrations of the 8760 hours of
measurements contained in the data set all fell within the summer subset
(approximately the five midday hours from June 15 to September 15).  This is,
of course, a very approximate assumption.

-------
                                     39
     If the n highest hours are contained within n  hours and a Weibull
distribution fits the n  hours of data, we can write:
          9(0  -
where C, is the concentration corresponding to an expected exceedance rate
of 1 in n  hours.  We can write:
         o
          J.  .  e'(1n no}  c                                              (3-32)
           o
and further that:
                               C  u
          !i  .  Jl  -(In n )  U)                                         (3-33)
          n      n              1
Thus, if the n highest values were replotted using n as  the devisor to
calculate a frequency factor for each data point,  the resulting  expression
(Eq. (3-33)) can be thought of as a three parameter Weibull, where  n /n  is
the third parameter.   Thus the effect of narrowing in on a  segment  of hours
tends to be to introduce another constant into the original  Weibull  distri-
bution.   The third parameter tends to be a displacement  factor since Eq.
(3-33) can be rewritten to give:
          r                  C-  i<    In  n  /n
          J-  -     1n n)  C(r)   -      ~]                               (3-34)

-------
                                    40
     The distribution of the maximum hourly average concentration for the
distribution given by Eq. (3-33) is:
                                       k
                      n   -(In n0)
          P0  -  (1 -    e           l   )n                                 (3-35)
It turns out that PQ calculated by Eq.  (3-35) is very close to that which
would be calculated using Eq.  (3-31):

          PQ  ,  (i _ e'(1n no> {ff) )no                                   (3-36)

that is:
          (1 - P)"0  «  (1 -P)n                                        (3-37)
     The result can also be verified by direct calculation as shown in
Table 3-8.  It is seen from Table 3-8 that for n° = 8760 and n = 500 the
distribution of the maximum value is almost identical  for either k = 1 or
0.5.  Even for n as low as 10 the approximation holds  rather well.  Thus,
if many of the highest values were included in the smaller time segment,
there would be no effective change in k, and the new distribution would be
most effectively described by a three parameter Weibull.  The more interesting
conclusion is that the nonstationarity leading to the  segmentation would have
little effect on the distribution of the highest hourly average concentration.
That is, Eq. (3-36) could be used to calculate the distribution.  While Eq.
(3-36) could not be said to hold down to n = 500 for nQ = 8760, it may apply

-------
                                   41
Table3-8   Comparison of Distributions of Annual  Maximum Hourly Average
           Concentration as Determined by Two Related Hourly Distributions
                                           Modified Weibull  Distribution
                                                                        (2)
Relative
Concentration
(c/c^
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
n =8760
0
k=1
0
0.002
0.084
0.368
0.668
0.850
0.936
0.974
0.989
0.996
0.998
0.999
1.000
1.000
n0=8760
k=0.5
0.012
0.074
0.203
0.368
0.526
0.657
0.756
0.827
0.878
0.914
0.939
0.956
0.968
0.977
n=500
k=l
0
0.002
0.083
0.368
0.668
n.850
0.936
0.97A
0.989
0.996
0.998
0.999
1.000
1.000
n=500
k=0.5
0.012
0.073
0.203
0.368
0.526
0.657
0.756
0.827
0.878
0.914
0.939
0.956
0.968
0.977
n=10
k=0.5
0.003
0.049
0.176
0.349
0.515
0.651
0.753
0.826
0.877
0.913
0.938
0.956
0.968
0.977
    (1)   G(X)  =  exp(-(ln  n^C/C.,)  ),
(2)
              =  (n/n0)  exp(-(ln  no)(C/C1)k),     nQ  =  8760

-------
                                  42
fairly well to n = 10 (the ten highest numbers).  Thus, this result is
suggestive that Eq. (3-36) gives a reasonable approximation for the dis-
tribution of the highest ozone hourly average concentrations.
     If a summer segment of selected times of the day is plotted, it probably will
not contain the 500 highest hours and therefore, will have  a higher k than
indicated by Eq. (3-33).  To investigate this point further the second
approach studied the question as to what value of k is required for an
hourly average concentration distribution for n = 500 to give an approxi-
mately equal maximum value distribution to that obtained with n = 8760.
Base case maximum value distributions were derived for n  = 8760 and
                                                        o
k = 0.5, 1.0, and 2.0.  In each case maximum value distributions for nQ = 500,
using Eq. (3-36), were calculated for a range of k values to see which k
gave the best fit to the base case.  In each case the best fit k value for
n = 500 was larger than the k for n = 8760 by a factor of about 1.4.  The
results for base case k = 1.0 are shown in Table 3-9.
     It was also of interest to repeat the above calculations for the dis-
tribution of the fifth highest concentration.  The results are shown in
Table 3-10.  They show that the modified Weibull continues to provide a good
approximation of the distribution of the fifth high.  In the case of the
unmodified Wei bull, the factor by which k must be increased to bring the
distributions into alignment increases from 1.4 to somewhat less than 1.6.
                                                         F81
     In a parallel investigation PEDCo Environmental, Inc; J examined aero-
metric ozone data from a number of monitoring sites for various subsets of
the hours in a year and compared the k values for Weibull distributions
fitted to the subsets with those obtained for the full year of data.  Table
3-11 shows the k values obtained for Heibull distributions fitted to data
from several locations for the complete set of 8760 hours and two subsets,

-------
                                         43
Table 3-9      Effective Increase in Weibull  Constant k Needed to Give Same
             Distribution of Maximum Value  as n is Decreased - Base Case:k=1.0
  Relative
Concentration
    c/c1

    0.7
    0.8
    0.9
    1.0
    1.1
    1.2
    1.3
    1.4
    1.5
    1.6
    1.7
    1.8
    1.9
    2.0
 Base Case
Distribution
 of Maximum
   n=8760
   k=1.0
   0.000
   0.002
   0.084
   0.368
   0.668
   0.850
   0.936
   0.974
   0.989
   0.996
   0.998
   0.999
   1.000
                                         Distribution of Maximum for n = 500
k=1.0
0.002
0.031
0.155
0.368
0.584
0.749
0.856
0.920
0.956
0.976
0.987
0.993
0.996
0.998
1.3
0.000
0.008
0.109
0.368
0.644
0.827
0.923
0.967
0.987
0.995
0.998
0.100


]4,2)
0.000
0.005
0.095
0.368
0.662
0.849
0.939
0.977
0.991
0.997
0.999
1.000


1,5
0.000
0.003
0.083
0.368
0.681
0.868
0.951
0.983
0.995
0.998
0.999
1.000


2.0
0.000
0.000
0.038
0.368
0.762
0.937
0.986
0.997






       (1)   Distribution of Maximum Value - [1  - exp(-(ln n)(C/C,

       (2)   Closest fit  k = 1.4;   k /k, = 1.4/1.0 = 1.4

-------
                                        44
Table 3-10    Effective Increase in Weibull Constant k Needed to Give Same

           Distribution of Maximum Value  as n is Decreased - Base Case: k=0.5
              Base Case

Relative     Distribution           Distribution of Maximum for n = 500
Concentration
c/c1
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
of Maximum
n=8760
k=0.5
0.012
0.074
0.203
0.368
0.526
0.657
0.756
0.827
0.878
0.914
0.939
0.956
0.968
0.977

k=0.5
0.063
0.145
0.252
0.368
0.478
0.575
0.658
0.726
0.781
0.825
0.860
0.887
0.909
0.927

=0.7(2>
0.019
0.085
0.211
0.368
0.521
0.651
9.751
0.825
0.878
0.915
0.941
0.959
0.971
0.980

=0.73
0.015
0.078
0.205
0.368
0.528
0.662
0.76*
0.838
0.889
0.924
0.949
0.965
0.976
O.Q83

=0.75
0.013
0.073
0.201
0.368
0.532
0.669
0.772
0.845
0.896
0.930
0.953
0.969
0.979
0.986

=0.8
0.009
0.063
0.191
0.368
0.543
0.686
0.791
0.864
0.912
0.943
0.963
0.976
0.985
0.990

= 1.0
0.002
0.031
0.155
0.368
0.584
0.749
0.856
0.920
0.956
0.976
0.987
0.993
0.996
0.998
(1)  See Table 3 for defining equation.

                            k

                            k"
                                    l   n 7
       (2) Closest fit k = 0.7;    r- = ~ = 1 .*•

-------
                                      45
Table 3-11   Change in Weibull Distribution Shape Factors with Changing Time
             Segment in Which Hourly Average Concentrations are Collected
                                                   Shape Factor, k
Site
Kansas City, Kansas
Des Moines, Iowa
Louisville, Kentucky
Memphis, Tennessee
Mamaroneck, New York
Racine, Wisconsin
Year
1975
1975
1974
1974
1975
1974
Full Year2)
1.24
1.92
0.80
1.32
0.80
1.49
May-Sept! 3)
11 AM-6 PM
--
2.04
1.66
2.28
1.31
July-Aug:
11 AM-6 PM
4.34^)
2.40
2.02
2.34
1.57
1.99(5)
  (1)   Reference 8
  (2)   8760 hours.
  (3)   1071 hours.
  (4)   434 hours.
  (5)   Poor Weibull  fit.

-------
                                     46
one containing 1071  hours and the other 434 hours.   They found  that  good  fits
of the Weibull distribution could be made to the subsets for  most  data  sets.
There was a marked increase in the k value.   They recommend a subset of 1071
(11 AM to 6 PM, May-September) hours having a good  chance of  capturing  the
highest annual hourly average ozone concentration.   Ignoring  the two cases
in the table in which a poor fit was obtained the k increased for  the 1071
hour subset by an average factor of 1.6 (range 1.06-2.08) and for  the 434
hour subset by an average factor of 1.9 (range 1.25-2.53).  These  values
are larger than suggested by the above approximate  analysis.
     In summary, it would appear that in the case of ambient  hourly  average
concentrations of ozone the distributions and related statistics for the
annual highest concentrations can be obtained by applying the assumption
of independence to a subset of hours consisting of the 1071 hours  in the
period 11 AM to 6 PM, May through June, and by applying the k values
corresponding to the subset.

-------
                                    47
3.4   Comparison of Statistical and Deterministic Standards
    In changing ambient air quality standards from their present deterministic
form to statistical forms the question arises as to how the standards compare
in the air quality provided.  A difficulty in making such comparisons is the
uncertainty in the degree to which a deterministic standard would be enforced.
As discussed earlier, strict enforcement of the present photochemical oxidant
standard logically means that the overall air quality for an Air Quality Control
Region (AQCR) must be brought the level that the 0.08 ppm hourly average con-
centration level is never exceeded.  However, it seems unlikely that this
would be done.  For example, if conditions were such that the standard level
was exceeded two or more times in one year (a violation of the standard) on
an average of once every one hundred years it is questionable as to whether
any effort would be made to further reduce precursor emissions to assure
that two or more exceedances in a single year would never occur.  The occur-
rence very likely would be downplayed as a freak meteorological  condition
untypical of the area or an  instrument error and, therefore, not worthy of
further control effort.  It is difficult to say at what frequency of violation
of the standard control officials would feel  obliged to act.   For some an
acceptable average period of reoccurrence might be as low as  ten or fifteen
years.  This difficulty stems directly from the nature of deterministic
standards which ignore the statistical  nature of ambient pollutant concentra-
tions  and needs to be kept in mind when comparing deterministic and statis-
tical  forms.

     3.4.1   Comparison Based on Return Period of Violation
     One way in which statistical standards can be compared with deterministic
standards set at the same concentration level  is through use  of  the return

-------
                                      48
period.  The return period is the average period of time separating the
repitition of an event whose occurrence is under the influence of statistical
processes.  The return period is reciprocal  of the relative frequency or pro-
bability of occurrence of the event.  For example, if theevent is the occur-
rence of two exceedances of the standard level in one year and the relative
frequency or probability of occurrence is 0.1, then the return period is
ten years.  That is, the event is expected to occur on an average of once
every ten years.
     A statistical  form of the standard fully takes into account the statis-
tical behavior of pollutant concentrations.   Thus, in setting a statistical
standard at a given concentration level it is acknowledged that there are
nonzero probabilities that the standard level will be exceeded 0, 1, 2, or
more times in the same year.  These probabilities can be estimated.  For
example, Eq. (3-15) can be used as a basis for estimating the probability
of m exceedances in one year of the level of a statistical standard and
Eq. (3-16) can be used to estimate the probabilities of m or more exceed-
ances in a one year period.  The reciprocals of these probabilities are
the return periods.  In particular, in comparing alternative statistical
forms with the present photochemical oxidant standard what is desired is the
probability and corresponding return period associated with two or more
exceedances of the  standard level in a given year, since such  an
occurrence constitutes a violation of the deterministic standard.
     In determining the probabilities it is most convenient to calculate
them as functionsof the expected exceedance rate.  In this case Eqs. (3-17)
and (3-18) are used.  Figures 3-6 and 3-7 show the probabilities of ex-
ceedances and of m or more exceedances in one year as a function of the

-------
                                   49
expected exceedance rate for a range of values of m.  The curves are inde-
pendent of the standard level and the distriubtion of hourly average concen-
trations.  They are also nearly independent of the number of effective hours
in one year.  That is, the curves for 8760 hours are essentially identical
to those for 1071 hours (see Section 3.2.3).  It will be recalled from the
previous section that the use of approximately 1000 effective hours in a
year tends to compensate for problems of nonstationarity in hourly average
ozone concentrations.
    Figures 3-6 and 3-7 illustrate  the problem of the present deterministic
standard.  Even at a low expected exceedance rate (e.g., E = 0.1) there is
a nonzero probability of 2, 3, or more exceedances.  To drive these proba-
bilities to zero it is necessary to drive the probability of one exceedance
(which is seemingly allowed by the present deterministic standard) to zero
and the probability of no exceedances to one.  Table 3-12 further illustrates
this point with selected data points from the curves in Fig. 3-6.  Each row
represents the probabilities for 0, 1, 2, 3, 4, 5 actual exceedances in a
year of a statistical standard level whose allowed expected exceedance
rate is given in the left hand column.  The proposed form of the ozone stan-
dard allows an expected exceedance rate of one.  It is seen from Table 3-12
that there is a 37% chance in any year of no violation, a 37% chance of one
violation and that in an AQCR just meeting the proposed standard there is
a 37% chance a single violation will occur in a given year and an 18% chance
there will be two violations.  By going to an expected exceedance rate of
0.1, the probability of two exceednaces can be reduced to 0.5 % or a return
period of 200 years.  The probability of one exceedance then, is forced to
9% or a return period of 11 years.

-------
 OJ
 (J
 e
 (Q
T3
 O)
 Ol
 (J
 X
i.
CU
JD
      OJ
      O
      O
      (O
     •o
      Ol
      O
      X
•Z.  T3
      01

 Ol   O

 S-   Q.
 QJ   X
 V)  UJ

O   c;


 CU
 >  <*-
•<-   o
  o   o

  >>  =3
 •!->   U.

 r—   (T3
  O
  s-
 Q-
  I
 CO
  O)
  CD
                                                                                                                                    CO
                                                                                                                                    CM
                                                                                                                                    cc
                                                                                                                                            O)
                                                                                                                                            •(->
                                                                                                                                            (C
                                                                                                                                            Oi

                                                                                                                                            (1)
                                                                                                                                            o
                                                                                                                                            c:
                                                                                                                                            (T3
                                                                                                                                            X>
                                                                                                                                            OJ
                                                                                                                                            
                                                                                                                                           O
                                                                                                                                           0)
                                                                                                                                           Q.
                                                                                                                                           X
                                                                                                                                           UJ
                                                                                                                                     CM

                                                                                                                                     O
                                        CO

                                        c:
                                                                                                         CM

                                                                                                         O

-------
 O)

•r-
CD

 <0

     
       O
       (U
       Q.
       X
      UJ

       0)
 IO    C
 3    O
•!->   ••-
 O   •!->
<    O
       C
 fO    3
      U_
 o

 >,
       
 (T5
                                                                                                                                                 QJ
                                                                                                                                                 O
                                                                                                                                                 OL>
                                                                                                                                                 OJ
                                                                                                                                                 o
                                                                                                                                                 X
                                                                                                                                                T3
                                                                                                                                                 O)
                                                                                                                                                 O
                                                                                                                                                 Q.
                                                                                                                                                 X
                                                                                                                                      C\J

                                                                                                                                      O
                                                                                                                                                           Lf)

-------
                            52
Table 3-12    Probability of a Given Observed Number of Exceedances
                      for Different Expected Exceedance Rates
                                     Probability
Expected
Exceedance
Rate
0.1
0.2
0.5
1
2
5
10
Actual
Exceedances = 0
0.905
0.819
0.606
0.368
0.135
0.007
<0.000
1
0.090
0.164
0.303
0.368
0.271
0.033
<0.00
2
0.005
0.016
0.076
0.184
0.271
0.084
0.002
3
<0.001
0.001
0.013
0.061
0.181
0.140
0.007
4
<0.001
<0.001
0.002
0.015
0.090
0.176
0.019
5
<0.001
<0.001
<0.001
0.003
0.036
0.176
0.037

-------
                                 53
     Table 3-13 shows the probability of two or more exceedances of the
standard level and corresponding return period for standards with different
allowed expected exceedance  rates.  It is seen that for the statistical
form of the standard at an expected exceedance rate of one the corresponding
deterministic form (not more than one exceedance of the standard level) with
the same concentration level will be violated on an average of about once
every four years.  If the expected exceedance rate of the statistical stan-
dard  was  set at 0.1 the present deterministic standard would be violated
on the average of about once every 200 years.  Note also that as the ex-
pected exceedance rate is increased above one the return period rapidly
falls to the point that the deterministic standard is being violated every
year.  Thus, by adjusting the exceedance rate the statistical standard can
be made to approach as close as is desired the deterministic standard in
stringency while maintaining the same concentration level.  Of course, and
as discussed in Section 3.2.2, this can also be achieved by maintaining
the exceedance rate at one and lowering the concentration level of the sta-
tistical standard.
     Table 3-13 also shows how statistical standards based on the expected
annual maximum hourly average concentration or the expected annual  second
highest concentration compare with the present deterministic standard.  Pro-
babilities for these forms are calculated by the first calculating the
expected maximum and expected second highest concentrations when the concen-
tration, C,, corresponding to an expected exceedance rate of one is equal
to one   (as done in Table 3-6) and then using Eq. (3-14) to calculate the
expected exceedance rate at the expected highest and second highest concen-
trations.   Given the expected exceedance rate the probability of two or more

-------
                                  54
Table 3-13    Comparison of Statistical  Forms with a Deterministic
            Standard Which Allows No More than One Exceedance per Year
      Expected                Probability                Return
     Exceedance             of Two or Mora               Period
        Rate                  Exceedances               (years)

        0.1                      0.005                   200.0
        0.5                      0.090                    11.1
        1.0                      0.264                     3.8
        2.0                      0.594                     1.7
        3.0                      0.801                     1.2
        4.0                      0.909                     1.1
        5.0                      0.960                     1.0
       10.0                      1.000                     1.0
Expected Max                     0.127                     7.9
Expected Second High             0.461                     2.2

-------
                                   55
exceedances in a given year is calculated from Eq. (3-18).  Following the
discussion in the previous section,1071 effective hours and a range of
Weibull shape factor, k, of 1.36 to 2.54 were used in the first part of
the calculations.  When this was done the probabilities and corresponding
return periods varied only slightly over the range of k (as noted
above, the number of hours also has little effect on the result).  Thus,
only single values of the probabilities are given for each form.  It is
seen that a standard based on the expected maximum value would lead to
violations of the present standard average of once every eight years.  It
is, therefore, significantly more restrictive than the proposed statistical
standard at the same concentraiton level.
     A standard based on the expected second highest concentration would
violate the present deterministic standard on the average of every other
year and is, therefore, significantly less restrictive than the proposed
statistical form of the standard.

-------
                                  56
     3.4.2  Comparison on the Basis of Fractional  Reduction in Precursor
            Emissions Required to Meet Standard
     Another method by which statistical  forms can be compared to determin-
istic forms of a national ambient air standard is  to determine the degree
of reduction in precursor emissions which each would require.   The difficulty
discussed at the start of Section 3.4  is particularly significant here.  If
the present deterministic standard for photochemical oxidants  were to be
fully enforced, that is, if control levels were required to an extent that
assured no exceedances of the standard level, the control levels required
would probably be significantly higher than those  on which the AQCR's have
based their present hydrocarbon control plans.  The reason for this is that
the design value for deciding level of control  required with the present deter-
ministic standard is the second highest concentration.  However, the second
highest concentration is not characteristic of the general air quality over
an AQCR since it is a statistical variable.  Even  though the levels of all
emissions which contributed to oxidant levels over an AQCR were held at a
constant level, the annual second highest concentration would vary from one
year to the next.  While it is true EPA's recommended practice is to have an
AQCR select the highest annual second high from several years of data (when
available) this does not fully allow for the range of variation which can be
experienced in the second high.  Thus, any annual  second high design value
will, to a high probability, lead to undervaluing the degree of hydrocarbon
precursor control required.
     This problem is resolved with statistical standards.  For a statistical
standard based on an expected exceedance rate E.-,  the design value would be
                                               ij
that concentration whose expected exceedance rate is E.  This concentration
has  a fixed value for a  given level of contributing precursor emissions and thus
 is  a true characteristic of the general air quality over an AOCR.  It can

-------
                                    57
 be  estimated  from  a  plot of one or more years of ambient air data for a
 monitoring  site.     The more years the better the estimate.  This estimate
 will  be  subject to statistical fluctuation for a finite set of years but
 will  be  significantly more stable than the annual second highest concentra-
 tion.
    Given the above  mentioned difficulties with the deterministic standard
 the following comparison was performed:  Use was made of data complied by
 the Monitoring and Data Analysis Division of the Office of Air Quality
 Planning and  Standards which showed observed annual second highest ozone
 concentraiton for  sixty AQCR's  having the greatest oxidant problem.  The
 second highest ozone concentrations were distributed as shown in Table 3-14.
 Accordingly, base case second high concentrations of 0.12, 0.15, 0.20, 0.25,
 0.30, and 0.47 were  choosen for study to cover the observed range.  If it
 is  assumed  these values represent the most probable values of the second
 highest  concentrations and the Weibull equation represents the distribution
 of  hourly average  concentrations, then each second high corresponds to a
 given Weibull distribution if the value of k and n are specified.  Table 3-14
 shows the relative location of the most probable value of the second highest
 concentration for  a  range of shape factors, k, and n = 8760.
      From the Table  3-14 data the concentrations, C , corresponding to
 E=  2 were calculated for each second high level  for a rangeof k values and
 n = 8760.   Given these data it was then possible to calculate the relative
 Cr-  corresponding to  each value of E from 1  through 10, 15, and 21 for each
 k for Weibull distributions corresponding to most probable second highs
of  0.12, 0.15, 0.205 0.25, 0.30 and  0.47 ppm.

-------
                                  58
Table 3-14    Cumulative Distribution of Second Highest
                   Concentrations for 60 AQCR's
          Concentration                Approx.  Cum.
           Second High                 Distribution
              ppm	                     %	
              0.12                         10
              0.15                         25
              0.20                         50
              0.30                         90
              0.47                         Highest

-------
                                   59
     Given the Cp and the second high values for each distribution the
percent hydrocarbon control required to get to C    = 0.08 with corres-
ponding number of exceedances was calculated using the proportional
roll back  method.
                     CF - 0.08
           % A HC =  -^	x 100                                       (3-38)
Background is assumed to be zero in Eq. (3-38).
     Equation (3-38) is also used to calculate the percent control  required
under the present deterministic standard by substituting the most probable
value of the second highest concentration for C •   As seen from Table 3-14
the most probable second high is very close to C£ » 2 over the range of k.
It would, therefore, give about the same percent hydrocarbon control as
CE = 2.  However, the use of this value alone gives a misleading picture of
the second high since,  as discussed, it is a statistical  variate.   To obtain
an idea of the range of percent hydrocarbon controls that could be  required
in an area of a given general air quality high and low values of tha second
highs (the 5% and 95% vlaues) corresponding to the most probable second high
were calculated and used to calculate percent control.  The distributions
for the annual second highest value were calculated using Eq. (3-26) of
Section 3.2.3.  Figure 3-8 shows the case for k * 1, n = 8760.
     Tables 3-15 through 3-20 show the calculated percent hydrocarbon control
and corresponding concentrations for the range of k and n values.   It is seen
in general that as the allowed number of expected  exceedances, E,  is increased
the percent control required decreases.  The fall  in percent control as E
increases is largest at the lowest concentration levels.  The percent control

-------
                           60
                       Figure 3-8

  Distribution  of Second Highest Value for Exponential
     Distribution of Hourly Average Concentrations

                       n=8760
—i    	1   i	y.   |
—:---!— -^r^f=-- --~i

-------
                                   61
Table 3-15    Location of Most Probable Value of Second Highest Concentration
              Relative to the Concentration Whose Exceedance Rate is Two (C-)
                   Weibull              Concentration
                   Constant             Relative to
     n = 8760 hours.
                     0.50                  0.9861
                     0.75                  0.9967
                     1.00                  1.0000
                     1.50                  1.0015
                     2.00                  1.0017

-------
ppm
o
      in
4J    4)
 fO    U
      c
•o    «»
 t-   T3
    LU
CO

4J    0)
 CO   4-)
 CO    U
s:    co
      ex
 O    X
4->   LU

T5   -O
 o,    i

•i-    o

 cr  ^
g   <

      c
 C    CO
-P    cn
 c    c
 o    ti
u    -c
      cj
4->
 c    -c
 4)    4->

 if    '5
 CO
Q.    r—
      CO
 C    >
•r-    4)

 CO
 cn
 c
 «J

o
Table 3-16
=8760 hrs
rations: k = 0.5
Average Conce
Distribution of Hour
Weibu
Ass
o
0
CO
o
in
CM
o
o
CM
CD
in
CD
Q.
CM
•*-> r-
CO O
J=
Ol
z
c •
o o
u c
0) O
VI U
I/I
4>
T3 O
4> C
4-> 
C f- 3
O Z J3
•^ *^
•U -D i.
•f C 4->
i/i CM in
o ••-
0. 0

CM
CO CO
. •
CO P"-
r-. vo

CO
c ^
O '*
CO CM
t •
o o


CO
o r-»

CO O
vo vo

o
o **•
in o
CM CM

d d




CO
o cn
• *
c o
vo m

CM
o co
c vo
CM "—
• •
0 0




o
r^ vo
B •
VO *3"
«? CO


<3-
O CM
10 CM
r"* '""
0 <=>





o
CO CO
, »
co co
CO i—

cn
^**\ f**+.
CM cn
i— O

o' o










41
^ X
o m
^^




CM
cn
•
^~
00

in
^
•a-
•
f^


t~~
n
•
CO
r«.

§§
VO
CO
•
o




CO
CO
•
CM
r***

o
in
en
CM
d




,3.
co
•
CO
VD


CM
t—
CM
CM
d





0
CO
•
^3-
in

o
f-^
f"
•
o











^^
in
en






















•
LU

CO
i_


J_
(O
>,

t-
Q)
o.

I/I
4)
U
C

U
41
d.
X
CO

CO
I/I
j?
5

^"^
E
o.
Q.
^M^
C
o
•f—
•**
u
4->
C
CO
u
c
o
u
I/I
LU


•"""**
f™-
'*'"*'








•o
0
4->
CO


^e
c>
IO
J3
^1
£

ITJ
O
•r-
•M
O
Q.
O

D.
>,
2
i/i
c
o
ja
i.
n
u
®
T>
>»
JZ

CO
C
5
4J CM
CO VC
E
o
C i

<+-
0
10
c
o
•r-
1/1
10
•^
E
CO

c
•^

c
o
•f"
4->
U '

i
J.

X
u
J_
•1—
3
CT
CO
J-
l/l
•^
<
X

**"*«
C\J
•»«-^'









-------
                             o

                             <
CO
o
 re    u
      c
•o    re
 L.   TJ
 re    oi

 c    u
 re    x
+•>   LU
to
     •o
4->    Ol
 Ol   +-*

z    a
      Q.
 O    X
•(->   LU

•O   "O
 Ol    Ol
 l_    X
•p*    O

 CT  r^
 ai   <
C£
      C
 O

 •p
 c
 o
      Ol
      cn
              i/i


             o
             vo

             00
              II
in

d

 u
              VI
              c
              o
 re



 01
 u

 o
o

 01
 cn
 re

 01
                        o
                        CO
                         in
                         CM
                         o
                         CM
 £   X
 O)
Q.  r—
      0)
 c   >
•r-   Ol

 O)
 en
 c
 re
 co
 01
 re
 o


 o


.a


-(->
 VI

a
              .a
              01
                         in


                         d
                               LU

in
CM
in
0
_
vo
in
CO
co
0
co
-
en
CM
0
CM
VO
CO
CM
CM

O
O
CM
in
r-.
vo

CO
CM
o
^.
CO
0
CO
o
_
CO
VO
Lo
CM
o
CM
o
VO
0

c
^
VC
IT


CM
0
VO
-
CM
CO
CM
o
en
in
vo
in
co
CM
0
in
!-».
in
CO
CO

o
o
CO
•3-
O

r-
CM
^-
O
CM
O
CO
VO
CM
O
CM
s
CM
CM
O
CO
in
in
en

c
CM
O
CO

0
0
0
o
en
vo
CO
in
CM
o
CO
CM
vo
in
CM
o
VO
CO
in
CM

O
CO
CO
CM

en
CO
o
en
f*.
VO
o
in
CM
o
vo
VO
CO
o
CM
o
o
CM
in
r-»
VO

o
^
in
CO
CM

en
O
CO
CO
o
o
VO
CO
CM
O
in
o
vo
CM
0
CM
O
vo
o
in
CM
vo

o
CO
co
CO
CM

CO
^
CO
o
CM
VO
vo
CO
CM
0
«a-
en
in
&
o
^.
en
8

o
^
CM
CO
CO

r«.
co
VO
co
o
^.
in
vo
CO
CM
0
in
00
in
co
en
O
CM
CO
in

o
VO
o
co
in

i-
m
in
CO
o
^
vo
CM
CM
0
^
in
en
CO
o
_
S
LO
1
o
r—
CM
co

in
r-.
CM
CO
o
r-.
vo
en
0
CM
0
o
in
•a-
•
o
VO
CM
cn
CO
1
o
^
CO
CM
O
•
«a-
CO
o
co
o
CM
cn
in
vo
cn
o
r—
LO
vo
o
en
CO
CO
co
'
O
CM
CO
CO
cn
o
•
                                                                                                                   O  CO
                                                                                                                     •    «
                                                                                                                   CO  CD
                                                                                                                   OO  CO
                                                                                                                            en
                                                                                                                            VO
                                                                                                                            00
                                                                                                                       cn   *j-
                                                                                                                   o  r»»   in
                                                                                                                   I-.  O   O
                                                                                                                   ^"  ^"   VO

                                                                                                                   odd
                                                                                                                       CO
                                                                                                                   CO  CM
                                                                                                                   co  cn
                                                                                                                   i-»  vo
                                                                                                               O
                                                                                                               co
                                                                                                                •
                                                                                                               cn
                                                                                                                       ^r   «d-
                                                                                                                   o  o   vo
                                                                                                                   O  VO   CO
                                                                                                                   CO  CM   CO
                                                                                                                     •   •     *
                                                                                                                   000
                                                                                                                       CO   VO
                                                                                                                   O  i—   i—
                                                                                                                   CO  CO
                                                                                                                   vo  vo
                                                                                                                            in
                                                                                                          o   o
                                                                                                      O  P-.   CM
                                                                                                      in  i—   CM
                                                                                                      CM  CM   CO

                                                                                                      odd
    CM
o  cn

o  co
vo  in
    VO
o  co
O  !».
CM  f—
                                                                                                                            en
                                                                                                                            CO
                                                                                                                            vo
                                                                                                                            vo
                                                                                                                            in
                                                                                                                            CM
                            CM
                      Ol O ^,

                              o
                      O U
                      u c
                      a* o
                          Ol  i-
                      "o  <_>  re
                      01  C  Of
                      *j  re  >,
                      O "O    I
                      Ol  Ol  I.
                      a. 0)  a
                      x  u  a.
                      LU  X
                                      ocroooooooo
                                      OCOCOCMCMVOCOCMCM

                                      ococoincMcnr^inco
                                      *3-   CO   CM   CM  CM  r—  r—   t—   i—
                                      coocMi^coor*,^J-CMO
                                      cocMi—   ooocncncncn
                                      f"""   r™   f™"   f™"  r^  r~~  CD   CD   CD   CD

                                      dddddddddd
                                                                                                                   o  o   o
                                                                                                                       vo   cn
                                                                                                                       in   in
                                                                                                                   vo  co   co
                                                                                                                   •a-  co   in
    CM   CM
o  o   co
in  co   cn
                                                                                                                   0  0
                                                                                                                                           •o
                                                                                                                                           o
                                                                                                                                            01
                                                                                                                                            u
                                                                                          CO

                                                                                          CO
                                                                            co   co
                                                                            CO   !-•»
                                                                            C3   CD
                                                                                                                       o
                                                                                                                   fO  CM
                                                                                                                   CO  CO
                                                                                                                   CO  CM
                                                                                                              CM
                                                                                                                •

                                                                                                              00
    CM   VO

CM  O   in


odd
                                                                                         in   o
                                                                                         r—   CM
                                                                                                        ^1
                                                                                                           cri4->

                                                                                                        O z -O
                                                                                                        •f- C 4J
                                                                                                        «/> CM  VI
                                                                                                        O    f-
                                                                                                        Q.    O
                                                                                                                   Ol
                                                                                                                   •o
                         o
                         a.
                         o
                                                                                                                                     01
                                                                                                                                     t.
                                                                                                                                     re
                                                                                                                                     01
 01
 o.

 VI
 01
 u

 re


1
 u
 X
 01

•a
 01

 u
 01
 a.
 x
 Ol

 Ol
                                                                                                                                     t
                                                                                                                                     a.
                  c
                  o
 2


 Ol
 u

 o
 u

 VI
 fc.

 u


T3

f

 ai
                                                                                                                                                co
                                                                                                                               o

                                                                                                                               i/i
                                                                                                                               c
                                                                                                                               o
                                                                                                                              •f—
                                                                                                                               I/I
                                                                                                                               VI
                                                                                                                                            u
                                                                                                                                            cr
                                                                                                                                            01
                                                                                                                                           X
                                                                                                                       r—     CM
                                                                                                                        in  in

-------
              (/I
             O
             1C
             r-.
             CO
                         O
                         co
CO
O
      I/)
4J    0)
 «    O
      c
•o    re
 J-   -O
 10       Ul
c/5
     •O
4->    
 0)    U
2:    V
      D_
 O    X
T5   T3
 0,    |

•r-    O
 3   l—

 Ol   <
o:
      c
^   »t—
 O
 i-
            in
            CM
      Ol
 c   c
 O   re
O   -C
      C5
 
 O)
 O)
 c
 rO
CO
  I
ro
 OJ
 C
 o
 re
4->
 c
 O)
 u

I

 
              V)
              •r*
              O
o
CM
                         in

                           •
                         o
                 o


                 X
                                LU
               I/I
               (/)
            o.

            CM


           o
               CM



                 <
CO
CO
cn
0
in
0
„.
in
m
CM
CO
o
^.
o
r>.
CM
o
^
CO
VC
CM
o
^
o
to
CM
vo

o
„.
CO
o
CO
CO
0
r*.
o
co
co
r*.
o
o
CO
o
0
CO
VO
0
in
CM
o
c
0
VO
o
o
CM
o

VO
•5J-
in

o
CO
CO
•—
CM
CO
^
0
0
CM
in
CO
CM
CD
^.
VO
VC
CO
CO
CM
o
0
CO
in
o
cn
o
c
5
co

CD
cn
en
<•
CO
co
O
cn
0
in
CM
o
_
VO
cn
CM
CM
o
«3-
in
CO
CO
c
cn
*
CO
CO

o
CO
r-.
cn
O
CO
en
0
^
o
CM
O
_
s
CO
CM
CM
O
_
in
in
CO
o
^
o
CO

o
CM
in
<*
O
CO
CO
o
o
co
en
VC
CM
o
CM
CO
VC
CM
o
o
in
s
0
vo
CO
CO
CO
o

o
co
CO
o
o
CO
o
o
o
VC
CO
vo
in
in
CM
O
^.
CM
VO
CO
CM
0

CM
in
£
o
CO
CO
00
CM

o
VO
r-
VO
cn
CM
en
CO
o
_
CO
VO
0
in
CM
o
^
vc
cn
o
CM
o
co
r—
in
vo
o

VO
CO
in
CM

C
_
o
co
cn
vc
CO
co
O
in
VC
vc
CM
0
0
VO
in
c
CM
0
in
o
in
CM
VC
O
O
in
CO
CO
CM

O
^
CO
cn
CO
o
o
o
10
CM
CM
O
^J.
o
VC
CM
O
CM
0
CO
cn
•a-
in
c
o
•a-
CO
CM

O
in
r--.
vo
-
in
CO
0
en
VO
CO
CM
CM
0
cn
in
o
cn
o
CO
*
CM
in
o
CO
cn
CM
-

o
CM
CM
un
vc
5
o
CM
CO
VC
CO
CM
0
cn
in
in
CO
o
cn
5
tn
CD
in
vc
CM
cn
C

c
.
CO
                                      COCOCMCMCMCMCMCMr—   i—
                                      OO^-or^-'i'CMOcor^
                                      cocMi—  i—  oooocncn
                                       oooooooooo
                          in
                          0) L.
                          o re
                                uo
                       o. at vv
                       x o ex
                      LU x
                                                                                           cn   co


                                                                                           CD   CD
                                                                                                                          CM
                                                                                                                     O   O
                                                                                                                       •     •
                                                                                                                     CO   ,—
                                                                                                                     00   00
                                                                                                                  CO
                                                                                                                   •
                                                                                                                  m
                                                                                                                  oo
                                                                                                                          vo   oo
                                                                                                                     O   r—   VO
                                                                                                                     r*.   CM   vo
                                                                                                                     «t   ^r   in

                                                                                                                     CD   O   O
                                                                                                                     CO   CM

                                                                                                                     co   o
                                                                                                                               cn
                                                                                                                               co
                                                                                                            r—   CO
                                                                                                       O  cn   •—
                                                                                                       O  vo   vo
                                                                                                       co  CM   co
                                                                                                        CO
                                                                                                        VC
                                                                                                             co
                                                                                                             co
                                                                                                             vc
                                                                                                                  co
                                                                                                             co   in
                                                                                                        CD   ^3*
                                                                                                        in   CM
                                                                                                        CM   CM
                                                                                                      •—     LU
                                                                                                      o
                                                                                                      CO      0>

                                                                                                      o      *o
                                                                                                 r—   CO
                                                                                            c   «a-   co

                                                                                            o   in   vo
                                                                                            vo   in   vo
     «S-   CM
o   cn   i—
O   I>x   -5T
CM   r-   CM

O   CD   CD
                                                                                                 «d-    co
                                                                                            p"-   in    f«^

                                                                                            vo   o    in
                                                                                                 VC   CTi
                                                                                            O  «er   O
                                                                                            in  co   co
                                                                                                                      o  o
                                                                                                 CO   CM
                                                                                            CO  1C   P>-

                                                                                            co  m   «a-
                                                                                            co  CM   «*•
                                                                                             _  vo
                                                                                             o  r*^.
                                                                                             CM  o
                                                                                                        O   O
                                                                                                                          re
                                                                                                                          O)
                   L.
                   O)
                   Q.

                   I/I
                   O>
                   O
 "W

1
 u
 X
 0)
                                                                                                                                              •o
                                                                                                                                               0
                                                                                                                                               V
                                                                                                                                               u
                                                                                                                                               re
                                                                                                                                 re

                                                                                                                                 o
                                                                                                                      o
                                                                                                                      Q.
                          V>
                          C
                                                                                                                      re
                                                                                                                      u
                                                                                                                      e
                                                                                                                     •o
        01

        re
                                                                                                                      g
                                                                                                                           VC
                   01      m
                  4->      C
                   u      o
                   Oi     •*-
                   a.     *n
                   X      V>
                   01     ••-
                                                                                                                                        I/)
                                                                                                                                        o
                          c
                          o
                         •r-
                         4J
                          U



                         I
                                                                                                                          Q.
                                                                                                                          Q.
                                                                                                                                        c
                                                                                                                                        o
                  t     n
                  C      i-
                  O)     -i-
                  U      3
                  C      O"
                  O      Ol
                  u      ^-
                                                                                                                                        in
                                                                                                                                               V)
                                                                                             Q JZ-r-
                                                                                           in   o
                                                                                           f—   CM
                                                                                             10
                                                                                             o
                                                                                            Q.
                                                                                                CTI4J
                                                                                                t-  3
                                                                                                a: xi

                                                                                                •o 'u
                                                                                                C 4->
                                                                                                CM  Vt
                                                                                                                            LU
                                                                                                                                       r-      CM
                                                                                            •O   »«  X
                                                                                             o   m  in

-------
                              <_>
                              <
 i.
 Q.
      1/1

 IO    U
      C
•o    
 01    U
z:    co
      Q.
 O    X
4-1   UJ

TJ   TJ
•f   O
 3  r-
 CT  r—
 Ol  «C
CC
 £   0,
 O
O
 C   -C
 CO   +->
 0)
a.  i—
      CO
 CO
 O)

 IB

o
en
 i
co
 co

jQ
              

             f
             o
             VO
             r^.
             00
              II
in


 n
              
              (B

             4->
              C
              O
             u

              a>
              C31
              IB

              O)
              O
stribu
                         O
                         CO
u
5
                         in
                         CM
                              U

                              <
            o
            CM
                              CJ
            in

            d
                            CM
                            M"
                              U
              V)
              I/I
                         a.

                         CM

                      VI
                      0) O
                     J=
                      o>
                     *r~
                     X

                     TJ ••
                      C  •
                      O U
                      U C
                      CO O
                     (/) O
                         to
                         CO U
                     TJ O IB
                      CO C CO
                     4J IB >>
                      O TJ    I
                      ai eo c
                      a. a> co
                      x u ex
                     UJ X
00
CO
in
en
0
^
^j-
J2
o
^
en
vo
CM
O
_
CM
VO
CM
O
^
en
CO
in
^™
o
CO
CO
CM

en
CM
CO
en
VO
«f
o
CO
CO
o
0
o
o
00
vo
0
in
CM
o
o
o
VO
o
o
CM
o
^
VO
o
in

o
CO
CO
CO
o
CM

^-
CM
oo
in
o
^.
CM
O
en
CM
o
en
S
CM
CM
O
^
00
in
CO
en
o
en
5
in

0
_
co
S

en
CO
CO
•si-
o
CO
I—
CO
CO
CM
0
^
vo
VO
co
CM
0
VO
f-
in
en
00
o
in
CO
CM

O
^
en
CM
CO

vo
00
CO
o
CM
,—
OO
l~»
CM
O
^
in
VO
CO
CM
0
CO
VO
in
in
CO
o
^
CM
en
co

o
o
oo
CM
-

CO
CO
p-
CM
0
r-.
0
CO
r-.
CM
o
en
vo
CO
CM
CM
0
_
VO
in
CM
00
c
^
*
co

0
CO
vo
CM
en
o

o
CO
CM
o
co
o
en
IO
CM
o
^.
<«3-
VO
CM
CM
O
^
in
in
0
CO
o
VO
o
in
co

o
^
in
CM
00
o

CO
0
CO
VO
o
en
en
vo
vo
CM
0
en
CO
vc
CM
CM
CM
o
en
in
r>«
o
CO
en
co
co
CO

o
CO
CM
VO
o

in
0
CO
-
o
VO
en
CO
vo
CM
o
in
co
vo
cr>
r—
CM
O
^.
in
in
0
CM
en
CO
co

o
en
CO
CM
in
o

co
o
00
0
o
co
en
o
VO
CM
o
^
CO
vo
r>.
CM
O
en
CO
in
S
o
in
CO
CO
o
CO

0
CM
CO
CM
S

in
en
o
en
CO
d
o
00
o
in
CM
0
VO
VO
CO
o
CM
C
CO
CM
in
r-.
vo
d
en
in
CO
in
CM

O
en
cr>
g
r—
.
en
CO
en
co
d
o

CM
CM
o
^.
0
vo
CM
0
CM
O
in
o
in
vo
d
en
CO
CO
CM

O
^.
Cn
en
en
0

                                           OOOOCOOOO
                                      r-   CM  CO
                                                         in  vo
                                                                                           o   o
                                                                                                                          r-»   in
                                                                                                                     o   vo   en
                                                                                                                      ...
                                                                                                                     co   •—   *r
                                                                                                                     CO   00   00
                                                                                                                          in
                                                                                                                     o


                                                                                                                     d   o   o
                                                                                                                     CO
                                                                                                                     co
                                                                                                                     vo
                                                                                                                     O
                                                                                                                     in
                                                                                                                     CM
                                                                                                                     O
                                                                                                                     vo
                                                                                                                     VO
                                                                                                                     «a-
                                                                                                                     o
                                                                                                                     in
                                                                                                                     CO
                                                                                                                     CO
                                                                                                                     CO
                                                                        co   en   o      in   o
                                                                                  i—      •—   CM
Position of
2nd High
Distribution
                                                                                                                          co  co
                                                                                                                          *t  in
                                                                                                                          en  co
                                                                                                                          CM  «5f
                                                                                                                     co   •—  vo
                                                                                                                     !*»   r^  rv
                                                                                                                          vo   
                                                                                                                     vO   ^*   CO
                                                                                                                     CO   CM   CO

                                                                                                                     odd
                                                                                                                          m  r-»
                                                                                                                          m   •—
                                                                                                                          vo   r~»
                                                                                                                          CM   CO
                                                                                                                          CM   CM
                                                                                                                          CO   00
                                                                                                                          CM   CM

                                                                                                                          d   d
                                                                                                             co   «a-
                                                                                                             en   vo
                                                                                                               •    •
                                                                                                             vo   «a-
                                                                                                             in   vo
                                                                                                                          i^*   co
                                                                                                                     o   in   vo
                                                                                                                     O   00   CM
                                                                                                                     CM   i—   CM

                                                                                                                     odd
                                                                                                                          r^.   vo
                                                                                                                          in   oo
                                                                                                                          CM   CM
                                                                                                                          ^i-   in
                                                                                                                          CO   !•>«.
                                                                                                                          en   en
                                                                                                                          co   vo
                                                                                                                          00
                                                                                                                          CM
                                                                                                                          CM
                                                                                                                          CO
                                                                                                                          CM
                                                                                                                          00
                                                                                                                     CD
                                                                                                                    •o  >«  «
                                                                                                                     o  m  in
                                                                                                                    X       en
                                                                                                                                        £
                                                                                                                                        u
                                                                                                                                        fO
                                                                                                                                        

                                                                                                                                 o
                                                                                                                                 J3

                                                                                                                                 IB
yd
                                                                                                                                              O)
                                                                                                                                              C
                                                                                                                                              10
                                                                                                                                              «*-
                                                                                                                                              0
                                                                                                                                              1/1
                                                                                                                                              I/I
                                                                                                                                             •r"

                                                                                                                                              CO
                                                                                                                                              o
                                                                                                                                              •r—
                                                                                                                                              4->
                                                                                                                                              O
                                                                                                                                       C
                                                                                                                                       O
                                                                                                                          •31   CO      I-    TJ
                                                                                                                     o   *—   tn      4->     eo
                                                                                                                     CM   i—   CO      C     t-
                                                                                                                     r—   r—   r—      CO    'r-
                                                                                                                                       0
                                                                                                                                       u
                                                                                                                                              cr
                                                                                                                                              CO
                                                                                                                                      i—    CM

-------
                               u

                               <
                          1^
                          «•
                            •
                          o
CO
o
 (O   U

•o   
-o   v
 c   u
 ITS   X
 4->  LU
 (U    -U
 0)    U

      a.
 O    X
4->    UJ

f3    -O
 01    0>

•r-    O


 at    «c
a:


"o
 k.    01
 O
(-)
 tu
 V
a.   i—
 at
 O)
 (T3
C

CM
 I
PO


 >

              T


               O


              «*-
               o

               c
               o
              •T™
              4J
               3



              T
            o
            CO
            in
            CM
            O
            CM
                   UJ
                 o
            in

              •
            o
               ai      cr
             g
             a.

             CM
         4->  i
          I/I
          at  c

          Ol
                              CM

                              "~o
               I/I
               (A
                        o  u
                        u  c
                        0)  O
                       (/) U
                           I/I
                           a> k-
                       T3  O 
                        ai  c a>
                        v ai v.
                        a. a> o>
                        X U O.
                       LU X
                                 UJ

CO
00
CO
o
PO
«3-
r—
CO
O
CM
0>
VO
0
VO
CM
O
1C
vo
CO
c

0
r»«
00
1C
m
o
CTl
in
CO
in

CM
CO
cr>
U2
O
CM
PO
cr>
cn
CM
0
o
CO
10
o
in
CM
c
o
o
vo
o
o

o
r»
1C
o
in
o
CO
PO
PO
c


CO
CO
in
o
1C
CM
CM
cn
CM
O
CM
r^
1C
<•
«3-
CM
0
O
cr<
in
in
w

o
CO
in
vc
*3-
O
r-»
PO
r^


CO
cr>
<3-
O
CM
VC
CO
CM
O
vc
VC
1C
cn
CO
CM
O
CM
CO
in
CM
CTi

O
PO
•3-
«3-
.
r—
CM
00
CM
0
1C
VC
1C
PO
CM
O
1C
1^.
if)
Ol
00

o
in
CO
CM
<3-
C
"sj-
cn
CM
PO


CO
r-s
co
o
PO
i—
cn
r-.
CM
o
i--.
U1
1C
to
CO
CM
o
r-
m
1C
CO

0
CO
CM
o
*3-
O
in
CO
CM
CM


CO
PO
PO
o
o
r—
1C
r>.
CM
o
PO
in
vc
CO
CM
O
VC
VC
in
^j-
00

o
CM
CM
CO
PO
c
l^»
!->.
CM


CO
00
CM
O
r~-
0
CO
f^
CM
0
O
in
VC
CO
CM
CM
O
CM
VC
in
CO
CO

o
VC
r—
pv.
CO
O
o
r^-
CM
o
f—

CO
in
CM
0
in
o
f>.
CM
0
r*.
^c
VC
vc
CM
CM
C
CO
in
in
i—
CO

c
i—
VC
PO
c
«5t
VC
CM
cn
^


CO
CM
CM
O
CM
0
cr
vo
CM
C
^f
«el-
1C
in
CM
CM
c
in
in
in
c
CO

c
r-.
0
in
PO
c
CO
in
CM
•-
•
CO
cn
0
*d-
d
PO
cn
1C
CM
o
CO
PO
vo
CO
CM
0
s
«e±
r^

o
CO
CO
PO
d
in
PO
CM
m
£2


CO
cn
cn
co
d
vo
CO
in
tn
CM
0
«t
CM
VO
CO
r"™
CM
O
C
PO
in
c
r^-

c
o-
r-
CO
CM
d
r>
CM
CM
^

                                                                                              o   o
                                                                                               in   o
                                                                                               i—   CM
o
CO
CO

o
r-.

d



PO
CO


0
o
PO
d



o
CO*

o
in
CM
d

c
d
o
o
CM
d




r^
U3


C
in
c




PO
PO
co


o
CM
f™"
d

o
0
CM
CO
,3.
^f
^f
^f
d


o
00
£

r^
CO
00
CM
d


vo
vo
^.
CO
d

cn
in
CO
d




cn
PO


CO
r—*
•si-
id




vo
cn
CM

r^
£
n™*
d

r-.
,3!
CO
co
in
^—
in
d


CO
VO
in

cn
CO
CM
PO
d


^
CO
d

i —
CM
d

CM
in
co
vc
CO
cn
CM
O




in
PO
ID


in
•sf
VC
d




CM
cn
PO

CO
PO
*
d























*
UJ
2!

-------
                                    67
required when the most probable value of the second high is used is close
to that for   E = 2.   The control required by the 5 and 95 percent! le
second highs cover a wide range corresponding approximately to E s 5 at
the 5% value to control levels significantly above that for  E = 1 at the
95% value.  Increasing k causes significant reduction in the effects of
increasing E values.
     In view of the discussion in Section 3.3 it would have been more correct
in the above calculations to use n » 1071 for calculating the distributions
of the annual second high concentrations.  In doing this, however, it would
have been necessary to adjust the shape factors, k, to somewhat higher values.
Since, as discussed in Section 3.3, this tends to undo the effect of going
to a lower k, the calculations were kept simple by using n = 8760 and not
changing the k values.

     3.4.3  The Highest Second Highest Concentration
     EPA's recommended practice with respect to the present deterministic
photochemical oxidants standard is that where the second high concentration
for several years are available, the highest of these values be used.  It
is, therefore, of interest to determine the most probable value of the annual
second highest concentration if more than one year of data are available.
Define:
          F(2) = P(C(2) =C)'
where C/«\ is the second highest observation of a series of n observations

-------
                                     68
     If there are N such second highest values,  by the binomial  distri-
bution it is known that the cumulative distribution of the highest of N
second high values is (assuming independence of concentrations):

          s(2) = F(2)N                                                     C3-4°)
The corresponding density function is given by:

From Eq. (3-26) the cumulative distribution of the second highest value
in n independent trials is:
          F(2) = "O'P)"" P + ^-P)"                                       (3-42)
and
                                                                           (3-43)
Following the practice in Section 3.2.1:

          p - e"(1n F} (C}                                                (3-44)

-------
                                       69
Then:
          &  = - 1. (in n?E) (C/CE)k~'  p                                  (3-45)
Therefore,
                                            1       n-2  2
                   n(n-l) (    In n/E (-)    (1-p)    p*                  (3-46)
           dC              LE          °E

Substitution of  Eqs. (3-42) and (3-46) into (3-40) and (3-41) allows the
calculation of the cumulative distribution and density function.
     Table 3-21 shows the calculated position of the most probable highest
second high concentrations for N years relative to C- _ - for a range of
k values and n = 8760.  As N increases the position of the most probable
value shifts to higher concentrations.  Table 3-21  is useful  in connection
with Tables 3-16 through 3-20.  It shows, for example, that for k = 1,
and C2 ~ 0.12, the most probable highest second high for N = 10 is 0.12 x
1.1653 = 0.140 ppm.  Reference to Table 3-18 shows the 95 percent!le second
high is 0.1447 ppm which is, as expected, just somewhat higher than the
most probable highest second high over 10 years.
     The density functions for the highest second highs for k = 1  are shown
in Table 3-22.

-------
                                   70
Table 3-21   Position of Most Probable Highest Second High Concentration
             of N Years Relative to Concentration with Two Expected
             Exceedances per Year
               Relative Position of Highest Second High Concentration
N
(Years)
1
2
3
4
5
10
K = 0.5
C = 1,
0.9861
1 .0968
1.1610
1 .2059
1.2407
1.3479
0.75
E = 2,
0.9967
1.0688
1.1094
1.1373
1.1587
1 .2239
1.0
n = 8760
1.0000
1.0531
1.0826
1.1030
1.1185
1.1653
1.5
1.0015
1 .0363
1.0554
1,0685
1.0785
1.1081
2.0
1.0017
1 .0276
1.0417
1.0514
1.0587
1.0804

-------
                                     71
Table 3-22 Density Function for Distribution of Highest Second Highest
           Value in N Trials of 8760 Trials Each for Wei bull  Distribution
           of Hourly Averages Where k = 1,   C- = 1.0,  E=2
Cone.
(ppm)
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
Max C
p Max
N = 1
0.0000
o.oooo
0.0216
1.7572
4.5396
2.6407
0.8064
0.1864
0.0382
0.0074
0.0014
0.0003
0.0000


1.0000
4.5396
2
—
-
0.0000
0.1935
3.6859
4.1479
1.5247
0.3684
0.0762
0.0149
0.0028
0.0005
0.0001


1.0531
4.8430
3
»
-
0.0000
0.0160
2.2446
4.8864
2.1619
0.5461
0.1140
0.0223
0.0042
0.0008
0.0001
0.0000

1.0826
5.0097
4
»
-
0.0000
0.0012
1.2150
5.1167
2.7250
0.7196
0.1517
0.0297
0.0056
0.0011
0.0002
0.0000

1.1030
5.1211
5
.
-
0.0000
0.0001
0.6166
5.0231
3.2199
0.8890
0.1892
0.0371
0.0071
0.0013
0.0002
0.0000

1.1185
5.2030
10
.
-
-
0.0000
0.0136
3.0016
4.8613
1.6760
0.3740
0.0740
0.0141
0.0027
0.0005
0.0001
0.0000
1.1653
5.4299

-------
                                    72
3.5  Statistical Standard Based on the Dally Maximum Hourly Average Concentration
     The discussion in Section 3 has dealt primarily with short-term
standards in which the averaging time was one hour.  It was pointed out
that much of the discussion was also applicable to longer averaging times.
The standards based on averaging times all have the common feature of divid-
ing the year into a series of contiguous segments of equal duration.  A
potentially useful alternate approach is to be only concerned with that
period during each day in which the time-averaged concentration reaches its
maximum value for the day.  The averaging time should be small compared to
twenty four hours.  For ozone the time would be one hour.  A statistical
daily maximum hour standard would be, for example:

          0.08 ppm daily maximum hourly average concentration with
             expected exceedances per year less than or equal to one.

Or the standard might be based on the maximum permitted value of the expected
highest daily maximum hourly average concentration in one year.  The standard
also could be expressed in a deterministic form.
     A daily maximum hour standard would simplify the reporting and analysis
of ambient air data.  However, it must also be acceptable from the point of
view of protecting public health.  Since a single exceedance of a daily
maximum hour standard in a year may represent one, two, or more exceedances
of an hourly standard it is clear a daily maximum standard is less stringent
than an hourly standard when both standards are set at the same level.  The
following sections explore the magnitude of this difference for both sta-
tistical and deterministic forms of the standard.

-------
                                      73
     3.5.1  Comparison of Hourly and Daily Maximum Hour Standards When
            Hourly Average Concentrations Exhibit No Dependence
          Comparison of standards based on hourly average concentrations
and daily maximum average concentrations is complicated because the lack
of independence of one hourly average concentration from  another and the
generally nonstationary character of the stochastic processes generating
ambient concentrations of pollutants.

     As a starting point it is worthwhile, however, to see how the two
types of standards would compare if complete independence and a stationary
stochastic process were assumed.  If it is also assumed the distribution
of the hourly averages is given by a Weibull  distribution, it is possible
to derive the relationships between the two standards.
     The basic hourly average concentration distribution is:
     Ph    •  e-ln i     (C/CHSTD)                                        (3-47)
where:
     "h    •  P  c>
     CHSTD *  The concentration level  of the hourly average standard.
           =  The expected number of hours per year exceeding the standard.

     nh/y  *  The e-ffectl've hours per  year.

-------
                                      74
If each hourly average concentration is independent of the others and the
stochastic process is stationary, the probability that the concentration
C will not be exceeded in n^ hours (the effective number of hours in a
day) is given by the binomial  distribution:
          P0  •  (1-Ph(C))"h/d                                             (3-48)

This probability is also the probability that the maximum observed hourly
average is less than or equal to the concentration C.  Therefore:
          Pd  •  1 - 0 - Ph(C))Vd                                       (3-49)
is the probability distribution for the daily maximum hourly average
concentration.
     If a standard concentration level and expected number of exceedances
is set for the daily hourly average concentration, can write:
                          
-------
                                      75
      n     *  The effective number of days per year.
 If ph is given by  (3-47) it follows that:
                      /  VyW
                     >\  E
(3-51)
Substituting Eq.  (3-51) into (3-50) yields:
      HSTD
In
1
- (1-
nd/y
1/nh/d"
                                               n  -1     Vk
                           In
                               h/y
                                                                           (3-52)
 Equation  (3-52) connects the parameters of the two standards when both are
 connected  by the same distribution of hourly average concentrations.  It
 allows comparisons to be made between the two kinds of standards when both
 standards  are simultaneously met.  Table 3-23 shows the corresponding
 levels of  the standards when the expected exceedances for each are the same.
 In the one case the hourly average standard level is held constant at 0.08
 ppm  for each value of the exceedance.  In the other, the daily maximum
 hourly average concentration level is held constant.  In this table the
 effective  hours per year are the full 8760 and the effective hours per day
are 24.  The calculation is done for three levels of k.

-------
                                       76
     It is seen that in the range of 1  to 20 exceedances, the difference
between the two standards is negligible for complete independence of hours
and a stationary stochastic process.
     To partially remove the effect of nonstationarity, the calculation
was repeated for effective hours per year of 600 and effective hours per
day of 5 (Table 3-24).  There is still  not much effect up to 10 exceedances.
     Thus, for the case of independent hours the two standards are essen-
tially identical for expected exceedances in the range of interest.   That
also means the risks of exceeding thresholds would be the same.  That is,
either one could be used for the same concentration level and expected
number of exceedances within reasonable limits and give the same protection.
     The analysis says that a relatively large number of exceedances would
be required before there was sufficient clustering within a day (in this
case by a purely random process) to cause the standards to begin to
separate.
     Unfortunately, the analysis cannot be stopped here because while we
have accounted for nonstationarity to some extent we probably have not
in this case, gone far enough.  For example, suppose a simple, but not
far fetched, coupling were introduced,  such as:  If one exceedance occurs
the next hour will almost always exceed the standard, while the probability
that the third hour will exceed the standard is about the same as the first
hour.  In this case the exceedances will cluster very strongly into pairs.  The
above analysis of the independence case shows that beyond that little additional
random clustering will occur if the total exceedances are not too large ( < 5).

-------
                                        77
Now in this case it can be concluded  that if the same concentration level
is set for the two types of standard the following very nearly holds:
          Ed/y  *  1/2                                                     (3-53)
In other words, the parameters of the two standards would now have to be
significantly different to provide the same protection.
     The analysis in the proceeding paragraphs shows that for complete inde-
pendence the standards  are  nearly the same over a broad range of E values.
For ozone concentrations, however, the nonstationarity is extensive.  To go
further requires working directly with ambient air data since at the present
time there is no mathematical model for the nonstationary stochastic proces-
ses generating hourly average ozone concentrations.

-------
                                           78
  Table 3-23 COMPARISON OF DAILY MAXIMUM HOURLY AVERAGE AND HOURLY AVERAGE STANDARDS
                                                   (1)
    Expected
  Exceedances^ '
       1
       5
      10
      15
      20
      50
Level  of Dally Maximum Standard When
  Hourly Standard Held at 0.08 ppm
0.5
0.0800
0.0799
0.0797
0.0795
0.0793
0.0779
1.0
0.0800
0.0799
0.0798
0.0797
0.0796
0.0789
2.0
0.0800
0.0800
0.0799
0.0799
0.0798
0.0795
       1
       5
      10
      15
      20
      50
                                Level of Hourly Standard When
                           Dally Maximum Standard Held at 0.08 ppm
0.5
0.0800
0.0801
0.0803
0.0805
0.0807
0.0822
1.0
0.0800
0.0801
0.0802
0.0803
0.0804
0.0811
2.0
0.0800
0.0800
0.0801
0.0801
0.0802
0.0805
(1)   Number of hours per year  =  8760
(2)   Number of hours per day   =    24
     Expected exceedances the same for both standards.

-------
                                            79
Table3-24 COMPARISON OF DAILY  MAXIMUM  AVERAGE  AND  HOURLY  AVERAGE  STANDARDS  WITH

         PARTIAL CORRECTION FOR NONSTATIONARITY  IN HOURLY AVERAGE CONCENTRATIONS^1
      Expected                     Level  of Daily Maximum Standard  When
    Exceedances(2)      	Hourly Standard  Held at  0.08 ppm

1
5
10
15
20
50
k = 0.5
0.0799
0.0794
0.0787
0.0777
0.0767
0.0674
1.0
0.0800
0.0797
0.0793
0.0789
0.0783
0.0734
2.0
0.0800
0.0799
0.0797
0.0794
0.0792
0.0766
                                       Level  of Hourly Standard  When
                                  Daily Maximum Standard  Held  at 0.08 ppm

1
5
10
15
20
50
k - 0.5
0.0801
0.0806
0.0814
0.0823
0.0835
0.0950
1.0
0.0800
0.0803
0.0807
0.0812
0.0817
0.0872
2.0
0.0800
0.0801
0.0803
0.0806
0.0809
0.0835
     (1)   Effective number of hours  per year =  600
           Effective number of hours  per day  =    5

     (Z)   Expected exceedances the same for both  standards.

-------
                                       80
     3.5.2   Comparison of Hourly and Dally Maximum Standards When
             Hourly Average Concentrations Exhibit Dependence
     To determine the effect of dependence of hourly average concentrations
use was made of an analysis performed by the Monitoring and Data Analysis
Division of OAQPS, at the Pollutants Strategies Branch request on data from
93 urbanized areas with populations greater than 200,000.  Data were assem-
bled on the three highest ozone readings for selected sites in each area
and the dates on which they occurred.  In a number of cases the highest
readings would duplicate and thus lead in some cases to having, in effect,
as many as the six highest readings.  The dates of each of these readings
were also recorded.  From these data it could be determined whether or not
the second highest day and the second highest hour tend to correspond to
the same reading.  The MDAD data are shown in Table 3-25 grouped according to
air quality region.  The right hand column shows that in 55% of the site-
years the second highest hour and the second highest daily maximum hour are
the same.  That is, in 552 of the cases the highest hourly average concen-
tration and the second highest concentration occur on different days.  In
45% of the cases the 3rd, or 4th, or 5th, etc. highest hourly average
corresponds to the second high day.
     It is seen in Table 3-25 that the data for each region are shown in
two groupings.  This is necessary because in many cases insufficient data
were available to determine which hourly average concentration corresponded
to the second high daily maximum hourly average.  For example, if the three
highest concentrations were all different, thru data points would have been
recorded for that site that year.  If these all occurred on the same day
then all that would be known is that the second high day corresponded to a

-------
                             in   o  vo  in
                             vo   co  in  co
                             ^-   CM
                             in   CM
                                                                                en  vo   co   i—
                                                                                *3-  en   •—   o
                                           i—   O
                                                                                                        CO
                             vo   m   CM

                             in   CM   vo  o   o
                                                                                              co
                                                                                     in       e—
                                                                                O   CM   o  co
                                                                                                        CM
                                                                                                        co
                en
                             CO   r—   CM
                             co  KT  «!f   i—  O
                             VO  CM
                                                                                vo  o  co   r»»
                                                                                 •    •    •    •
                                                                                ^t  i—  O   O
                                                                                                        VO

                                                                                                        CO
                                                                                                        CM
01
W
CO
c
o
c
cu
o

o
cu
O)
rO


CU
                00
                


                                                                                t—   CM
                                                                                         en


                                                                                         CM   O
                                                                                                        CM

                                                                                                        O
_Q C
•r- O
•M CD
 CU
••- Qi
0
in
CM
i
m
cu
!Z
(O £. II
h- -U
5^
0
w
S_ JC
IO C7)
CU •!—
>>3:
1
eu TO
4-> C
•r- O
OO U
cu
t-
3
O

4T
CJ)
•r"
4=

-o
C
CM





— —


-0 JZ
i. •(->
co  -P
in vo

in
i.
(T3
CU
>,
eu cu
•(-> S-
•r- Oi
00 -C
cu
> 01
0 S-
_Q ro
< eu
M-^
o cu
4-)
eu-r-
> 00
•1—
10*5
3
r—
o
X
c
>o
h-
(/)
I/)
eu
_J
>>
re
0
,-C
C71
•r—
*T~

-D
c
o
(J
cu
oo

•a
cu
-(->
c
•r-
5-
CL
s.
3
o
-c:

T3
s_
CO





— —


JZ JC
-l-> 4->
^r in


                                                                                                        ra
                                                                                                        cu
                                                                                                        cu

-------
                                     82
concentration less than the third highest hour.   It is possible to redis-
tribute these cases into the upper group if it is assumed the distribution
of the upper group is approximately correct.   For example, looking at the
far right column of Table 3-25, 11.49% indefinite cases have the second
high day corresponding to less than the third highest hour.   These cases
can be distributed among the 4th, 5th, 6th highest, etc.   in the grouping
above roughly in proportion to the current distribution.
     Thus, if the 1U69 is assumed to distribute entirely between the 4th,
and 5th high, 11.49 x 4.56/5.91 - 8.9%  should go to the  4th and 11.49 x
1.35/5.91 = 2.6% should go to the 5th high.  The 2.96% less than the
fourth high (Table 3-25, far right column ) can be distributed to the 5th
high.  The 1.2% greater than the 5th high can be distributed to the sixth
high.  The 1% remaining is attributed to the seventh and  higher values.
The calculation could be repeated using the newly obtained distribution.
Table 3-26 shows the result for a single iteration.  Using these data to
weight the ranking it can be calculated (Table 3-26) that the second
highest daily maximum concentration corresponds on the average to the 2.8
highest hourly concentration in a given year.
     Table 3-26 can be used to obtain an approximate comparison of at least
one statistical form of the hourly and daily maximum hour forms.  However,
it is first noted that if the forms of the standards were the expected values
of the annual maximum hourly average concentration and daily maximum hourly
average concentration, both standards would be identical.  This follows from
the fact that the maximum hour in any year is also the maximum daily maximum
hour.  This would not be true of the expected value of the second highest

-------
                                        83
Table3-26 Daily Maximum Hourly Average Standard versus Hourly Average Standard
                                                                     Per Cent
 Second Highest Dally Maximum Concentration the Same as:              of Years
        Second highest hourly concentration                             55
        Third highest hourly concentration                              23
        Fourth highest hourly concentration                             13
        Fifth highest hourly concentration                               7
        Sixth highest hourly concentration                               1
      <, Seventh highest hourly concentration                             1
                                                                       100
        Weighted Average Rank =
                  0.55x2 + 0.23x3 + 0.13x4 + 0.07x5 + 0.01x6 + 0.01x7  = 2.8

-------
                                    84
hour versus the expected value of the second highest dally maximum hour.
The equivalent daily maximum standard would have a lower expected second
high concentration.
     Where Table 3-26 can be used is in the case of a standard of the form,
X ppm with an expected number of exceedances per year less than or equal
to one.  At any given number of hourly exceedances of a given concentration
level the data in Table 3-26 can be used to estimate the distribution of
exceedances of the same concentration level by the daily maximum hour concentra-
tions.  When the annual hourly exceedances are 0 or 1, the connection is  straight-
forward.  If there are no hourly exceedances, there can be no exceedance
of a daily maximum hour standard with the same specified concentration
level.  If there is a single hourly exceedance there will be a single daily
maximum hour exceedance.
     In the case of two hourly exceedances, the data in Table 3-26 indicate
that in 45% of the years there will be a single exceedance of a daily maxi-
mum standard.  In the remaining 55% of the years there will be two exceed-
ances.  Unfortunately, the data in Table 3-26 are not sufficient to fill
out the table for more than two exceedances of the hourly average standard.
The values shown in Table 3-27 for 3 and 4 hourly exceedances are coarse
estimates.  The numbers in the single daily exceedance column, however, do
follow from Table 3-26.  For example, if there are three hourly exceedances
but only one daily maximum hour exceedance, this corresponds in Table 3-26
to cases in which the second highest day corresponds to the 4th, the 5th,
etc. highest hour.  These add to 22% of the years.  Similarily, for four
hourly exceedances the number is 9%.  These have been rounded to 20% and
10% in Table 3-27.

-------
                                             85
         Table 3-27   Distribution of Exceedances  per Year for Daily Maximum Hour
                      Standard Versus Exceedances  per Year for Hourly Standard'  '
Exceedances
per Year of
  Hourly
 Standard


     0

     1

     2

     3

   > 4
Estimated Distribution of Exceedances for
Daily Maximum Hour Standard
dances
ear = 0
1.00
0.00
0.00
0.00
0.00
1234

1.00
0.45 0.55
0.20 0.35 0.45
0.10 0.30 0.30 0.30
Weighted^
Average
Exceedances
Daily Std.
0
1.00
1.55
2.25
2.71
Relative^
Frequencey of
Exceedances
Hourly Std.
0.37
0.37
0.18
0.06
0.02
    (1)  Both standards have the same specified  concentration  level.

    (2)  For two hourly exceedances the weighted average  daily maximum

         hour exceedance rate is 0.00 x 0  +  0.45 x 1  +  0.55 x  2   =  1.55.

    (3)  For expected exceedances/yr = 1, assuming independence of hours.

-------
                                      86
     From this distribution the weighted average exceedances of the daily
maximum hour standard can be calculated for each hourly exceedance.  A
sample calculation is shown in Table 3-27.
     To compare the two standards, the next step is to assume a probability
distribution of the hourly exceedances.  The last column of Table 3-27
shows how they are distributed for an expected exceedance rate of one if
all hours of the year are independent (calculated from the binomial distri-
bution).  The corresponding expected expected exceedance rate for the daily
maximum hourly average is obtained by combining the last two columns of
Table 3-27:

   0.37 x 0 + 0.37 x 1.00 * 0.18 x 1.55 + 0.06 x 2.25 + 0.02 x 2.71 = 0.84

Thus, the daily standard is only somewhat more restrictive than the hourly
standard when the hourly expected exceedance rate is one.
     Without much loss of accuracy it can be said that a daily maximum
hour standard with an expected exceedance of once per year corresponds to
an hourly standard at the same specified concentration level with an
expected rate of 1/0.84 = 1.2 exceedances per year.
     A sensitivity analysis of the assumptions in Table 3-27 with regard
to the distribution corresponding to hourly exceedances of 3 and 4 and the
relative frequencies of exceedances of the hourly standard shows that the
last figure given is most probably in the range 1.2 to 1.4.  If it is
assumed 1.3 expected exceedances per year is the most likely value, Eq.
(3-14) can be used to calculate the equivalent concentration level of the
daily maximum hour standard if both standards have an expected exceedance
rate of one.  The comparison is shown in Table 3-28.

-------
                                         87
Table 3-28   Approximate Comparison of Statistical Daily Maximum Hour and Hourly
              Standards When Both Have Expected Exceedances per Year of One  '
                    Level  of                      Level  of
                 Hourly Standard          Daily Max. Hour Standard
                      (ppm)               	(ppm)	

                      0.06                         0.061
                      0.08                         0.082
                      0.10                         0.102
                      0.12                         0.123
                      0.14                         0.143
               (1)   Assumes distribution of hourly average concentrations
                    is represented by a Wei bull  distribution for 8760
                    hours whose shape factor is:   k = 1.25.

-------
                                      88
     While the above calculation is crude the conclusion that the two
forms of the standard do not differ much in their relative severities
and that the daily maximum hour standard is the less severe of the two
can be made with reasonable confidence.

     3.5.3   Comparison of Deterministic Hourly and Daily Maximum Standards
     The preceeding sections compared statistical forms of hourly and daily
maximum standards.  For a standard in which only one exceedance per year of
specified concentration level is permitted the results are different and
depend upon how strictly the standard is enforced.  If the restriction to
no more than one exceedance a year is strictly applied then both standards
are identical in their end result since  both (following arguments presented
in Section 3.1) will lead to an air quality in which the standard level is
never exceeded.
     In getting to this air quality level, the "design values" for the daily
maximum hour standard will initially average lower than those for an hourly
standard.  This follows from the data and analysis presented in Section 3.4
and 3.5.2.  In application of a "not to  be exceeded more than once" standard
the design value for calculating required levels of emission control is the
highest second high over a period of several base years.  If only one base
year is available,the second high for the area for that year is used.  In
this  case  the highest second highest daily maximum hourly average concen-
tration will be equal to less than that  of the highest second highest hourly
average concentration.  This situation is seen in Table 3-29 which was
obtained from the MDAD analysis referred to in the previous section and which
provided the basis for Tables 3-25 through 3-27.   In this case the highest

-------
                                       89
Table 3-29   Comparison of Percent Hydrocarbon Control Required by Daily
                 Maximum Hourly Average and Hourly Average Standards
     Highest                 % Control Hourly Max. -  % Daily Max.
  Hourly Average
   Concentrati on
       (ppm)
      0.115
      0.144
      0.174
      0.203
      0.233
      0.262
      0.292
      0.321
      0.351
      0.380
Average
(«)
1.0
4.0
7.0
4.5
4.0
2.8
1.5
1.4
1.4
1.4
Maximum
(%)
2.8
11.2
16.0
13.0
10.0
8.0
6.0
5.0
4.0
3.5
Minimum
(%)
0
0
0
0
0
0
0
0
0
0

-------
                                     90
second high for the hourly and daily maximum hour concentration for each
of the 93 urbanized areas was used to calculate the percent control of
hydrocarbon emissions needed to meet the standard.  Linear rollback was
used ( Eq. (3-38)). The results were summarized to give the average differ-
ence in calculated control level as well as the highest and lowest individual
value over a range of air quality levels (as exemplified by the highest
hourly average concentration).
     It is seen in Table 3-29 that at each level  of air quality the minimum
difference is zero.  And indeed, for many of the regions the second highest
hour and the second highest daily maximum hour are the same.  This follows
directly from the results in Table3-25which show that this situation occurs
on the average in 552 of the site years of data.   Thus, for any region as
data are collected over an increasing number of years the design values
for the two standards will eventually come together.
     Thus, in the short term a deterministic daily maximum will either
require the same level or less control than an hourly standard.  In the
long run both standards will require the same level of control.
     Notice the situation is different in statistical standards and depends
on the form.  They are the same for a standard which specifies a maximum
value for the expected value of the annual highest concentration.  However,
for an expected second high standard or standard that sets an expected
exceedance rate of a specified level the daily maximum hour standard will
be less than the hourly standard if the hourly average concentrations are
autocorrelated.  In the case of ozone there is a relatively small difference
between the standards based on expected exceedances.

-------
                                       91
3.6  Conclusions and Recommendations
     The present deterministic form of the photochemical oxidants standard
does not take into account the statistical behavior of hourly average ozone
concentrations.  As a result the only way in which strict compliance with
the standard can be achieved is by controlling precursor emissions to the
degree that the standard concentration level is never exceeded.  The standard
is, therefore, significantly more stringent than apparent from its seeming
allowance of at least one exceedance of the standard level per year.  The
uses of the second highest ozone concentration as an indicator of how far
a given Air Quality Control Region is out of compliance and as a design
value to calculate required degree of control of precursor emissions are also
misleading.  Because second high concentrations are statistical variables
they do not provide a measure of the overall air quality for a region and
will lead to an under estimation of the degree of control  of precursor emis-
sions required to gain compliance.  Use of the highest second high for
several years of data only partially corrects for this problem.
     Statistical  standards provide a way out of these difficulties by fully
taking into account the statistical  behavior of air pollutant concentrations.
They provide a clearly defined target for determining compliance and a mean-
ingful measure of existing air quality.
     The statistical  forms unlike deterministic forms lend themselves to
quantitative or mathematical  treatment.  The terms used in statistical
standards such as expected exceedance rate or expected annual  maximum
hourly average concentration  have precise meanings and are relatable to the

-------
                                      92
cumulative frequency or probability distributions of the time average con-
centrations.  With some statistical forms, such as expected annual  maximum
concentration, it is necessary to take into account autocorrelation between
concentrations close to each other in time and nonstationarity in the under-
lying stochastic processes giving rise to the random changes in concentrations.
A simple approximate method has been presented dealing with this lack of
independence.
     Comparison of statistical and deterministic standards which are set at
the same concentration level indicates the deterministic standards  to be
more stringent if the deterministic standard is strictly enforced.  An AQCR
just meeting a statistical standard allowing an expected exceedance rate
of one will violate on an average of once every four years a deterministic
standard allowing not more than one exceedance per year of the same concen-
tration level.  A statistical standard based on an expected annual  maximum
concentration will lead to violations of the above deterministic standard
on an average of once every eight years when the statistical standard is
just met.
     When the comparison is made in terms of degree of precursor emission
control, in the short term the deterministic standard could appear on the
average to be somewhat more lenient.  In the long term and if the determin-
istic standard is strictly enforced it will require greater control of
precursor emissions than comparable statistical standards.  (In the case of
the present deterministic oxidant standard the comparable statistical stan-
dard is based on an expected exceedance rate of one.  A standard based on
one expected annual maximum hourly average concentration would be somewhat,
though not strictly, comparable.)

-------
                                     93
     Comparisons between oxidant standards based on hourly and daily maximum
concentrations have led to the following conclusions.  If both standards are
deterministic in the short term it will appear the daily maximum standard
is somewhat more lenient than the hourly standard.  The difference in
apparent control of precursor emissions required for individual Air Quality
Control Regions (AQCR), in which the second high design value is based on
one to several years of monitoring data, will vary from zero to somewhat
over 15%.  In the long run both standards are equally stringent and will
require the same level of control.
     In the case of statistical forms there would be no difference in the
two forms if the hourly average concentrations were not autocorrelated.
The autocorrelation in the case of oxidant causes the daily maximum standard
to be slightly less stringent.
     It is recommended that EPA consider changing all its short-term
national ambient air quality standards from the present deterministic to
suitable statistical forms.  Its standards based on annual mean concentra-
tions could be considered as already being in a statistical form.
     Initial  guidlines have been developed by the Office of Air Quality
Planning and Standards on how to determine compliance and degree of depar-
ture from compliance with the proposed statistical standard for oxidant.
These guidlines are a good first step, but more study is needed on these
questions to fully utilize the benefits of the statistical form.
     It would also be desirable to perform a more complete investigation
of the applicability of Weibull and other distribution functions to ozone

-------
                                   94
and other pollutants.   The question of within year  and year  to year  (or
longer term) variations in air pollutants  requires  further study.
     Another major area needing investigation is the  autocorrelation of
air pollutant concentrations and nonstationarity in the  underlying stochas-
tic processes.  Improved understanding in  this  area will  allow consideration
of a broader range of statistical forms and more accurate treatment  or
aerometric data.

-------
                                     95
3.7  References

1.  U.S. Code of Federal Regulations, 40 CFR, Part 51.14, 1975.

2.  "Uses, Limitations, and Technical Basis for Procedures for Quanti-
    fying Relationships between Photochemical Oxidants and Precursors",
    U.S. Environmental Protection Agency, Research Triangle Park, N.C.,
    EPA 450/2-77-021 a, November, 1977.

3.  "Investigation of Alternative Methods of Expressing the National
    Ambient Air Quality Standard for Photochemical Oxidants", Final
    Report on Task #4 of U.S. Environmental Protection Agency Contract
    No. 68-02-2393, William F. Biller, East Brunswick, N.J., in press.

4.  R. I. Larsen, "A Mathematical Model  for Relating Air Quality
    Measurements to Air Quality Standards", U.S. Environmental Protection
    Agency, Research Triangle Park, N.C., Publication No. AP-89, November,
    1971.
5.  T.C. Curran and N.H. Frank, "Assessing the Validity of the Lognormal
    Model When Predicting Maximum Air Pollution Concentrations", Presented
    at 68th Annual Meeting of the Air Pollution Control Association,
    Boston, June 15-20, 1975.

6.  "The Validity of the Weibull Distribution as a Model for the Analysis
    of Ambient Ozone Data", Report, U.S. Environmental Protection Agency,
    PEDCo Environmental, Inc., Cincinatti, Ohio, in press.


7.  J. Horowitz and S. Barakatz, "Statistical Analysis of the Maximum
    Concentration of an Air Pollutant: Effects of Autocorrelation and
    Nonstationarity", submitted for publication to Atmospheric Environment.

8.  J. V. Uspensky, Introduction to Mathematical Probability. McGraw-Hill,
    1937, p.46.

9.  Private Communication, T. Johnson, C. Nelson, PEDCo Environmental, Inc.
    Durham, N. C.

-------
                                       96
4.0   ASSESSMENT OF RISK ASSOCIATED WITH POSSIBLE CHOICES OF A PRIMARY
      NATIONAL AMBIENT AIR STANDARD

4.1   Background
     If National Ambient Air Standards are expressed in statistical  forms it
is now possible as discussed in the proceeding sections to associate standard
levels in a meaningful way with the general levels of air quality which will
meet the standard.  The way is then paved for beginning to assess the risks
of adverse health effects to the most susceptible members of the population
associated with alternative levels of the standard.
     A quantitative assessment of these risks can be of major value to the
EPA standard setting process.   This section present a mathematical  formalism
for risk assessment.  It is based on underlying concepts developed by Mr.
Thomas Feagans of the Office of Air Quality Planning and Standards.   This
report will deal only with the mathematical treatment which was the contractor's
contribution.  A complete discussion of the method may be found elsewhere1-  .

4.2  Underlying Principles of Risk Assessment Method
     A broad perscription for setting primary National Ambient Air Quality
Standards is given by the following passage from the Clean Air Act.
     National primary ambient air quality standards... shall be ambient
     air quality standards, the attainment and maintenance of which in
     the judgment of the Administrator, based on [air quality] criteria
     and allowing an adequate margin of safety, are requisite to protect
     the public health'-1-'.
Support material to the Act clarifies this  perscription somewhat:
     An ambient air quality standard therefore, should be the maximum
     permissible ambient air level of an air pollution agent or class
     of such agents (related to a perido of time) which will protect
     the health of any group of the population

-------
                                   97
     A fundamental concept implicit in the above passages is the concept of
a health effects threshold; i.e., a concentration C_ such that if the con-
centration of pollutant, averaged over a suitable time period, is less than
CT, the pollutant will neither cause nor contribute to effects which are
detrimental to human health - even in people who are most susceptible to the
health effects at higher concentration levels.
     The provisions in the Act for an adequate margin of safety implicitly
acknowledges that the concentration of the health effects threshold will be
uncertain.   That we do not have precise knowledge on the location of health
effect thresholds is amply demonstrated in the EPA criteria documents which
discuss in detail the health effects research and which are intended to help
decide the pollutant concentrations at which health effects occur.  In the
case of photochemical oxidants, for example, numerous toxicological, clinical,
and epidermiological studies are cited which show effects at various levels
but do not directly allow a determination of the lowest concentration level of
oxidants in ambient air at which the most susceptible portion of the population
would suffer adverse health effects.  The data from different studies are
often in conflict and in cases of effects shown in animals it is not certain
the effects observed will be caused to occur in man exposed to oxidants in
ambient air.
     Fortunately, through the use of techniques developed in the field of
decision analysis, the criteria document can be used to quantify the uncer-
tainty in the location of health effect thresholds.   Roughly the procedure
would be to have highly qualified medical  people study the criteria document
in detail.  After they have absorbed all  the pertinent health data, the
technique of probability encoding would be used to quantify their subjective

-------
                                       98
judgments by eliciting a probability distribution for the location of individ-
ual health effects.  That is, they would define functions:
         DT   =  Pr (CT  < C)                                               (1)
          Ti          Ti ~
Where the quantity D   is the probability that the health effects thres-
                    TI
hold of an effect i     is less than or equal  to the concentration C.   The
method of probability encoding by which the probability distribution repre-
sented by Eq. (4-1) is determined is discussed in detail  in a comprehensive
              f3l
review article  J and its application to photochemical  oxidants is discussed
                               T41
in the report referred to above   .   The following discussion will take
Eq. (4-1) as its starting point and not discuss how it  is obtained nor how
the concept of health effects threshold is given precise meaning so that it
may be successfully encoded.  Nor will it be concerned  with the concept of
subjective probability and its utility in decision making processes.  These
matters are sufficiently treated in the references cited above.
     There is also a second source of uncertainty that  must be accounted
for in setting an adequate margin of safety.  This is the uncertainty in
the peak levels that will be reached over a time period of one or more years
when the standard is met.  As discussed in Section 3, setting National Ambient
Air Quality Standards cannot place an absolute lid on the highest pollutant
concentrations which may be observed in a given time period.  This follows
from the stochastic behavior of pollutant concentrations, and,as also discussed
in Section 3,  leads to the conclusion that standards should be expressed in
a way that acknowledges this fact.  That is, they should be expressed in suit-
able statistical forms.  Since it is not known what the highest concentration
values will be in any given period, the margin of safety again must be adequate

-------
                                      99
to give assurance that these highest concentrations will not be likely to
exceed a health effects threshold for the pollutant.   The uncertainty in
the peak concentration levels can be expressed as:
         PC  =  Pr (CMax  1
where the quantity Pp is the probability that no concentration will exceed
the concentration level  C in a given time period (e.g., 5 years).  The concen-
tration CM,  is the maximum time averaged concentration observed during the
         naX
given time period.
     It should be noted here that the concentration C appearing in Eqs.
(4-1) and (4-2) is a time averaged concentration.  The averaging time is set
by health effect considerations.  If short-term peaks cause the onset  of
health effects,then averaging times of one hour are used.  If exposure to
a concentration level over a more extended period of time is necessary, the
averaging time could be several hours, one or more days or even a year.  For
photochemical oxidants the averaging time is, as mentioned earlier, one hour.
     To assure that there is an adequate margin of safety is the same as
setting the primary standard so that the risk is adequtely small that the most
susceptible members of the population will be exposed to concentration levels
exceeding a health effect threshold in some given time period.  If the expres-
sions (4-1) and (4-2) are known either as mathematical expressions or in tabular
form this risk can be calculated for each alternative standard setting.  The
essential judgment to be made by EPA decision makers is, therefore, deciding
on the level  of risk that will  be tolerated, taking into account the seriousness
of the health effects involved.  Choosing the acceptable risk level  will  then

-------
                                   TOO
lead to choosing the corresponding alternative standard, and this standard
will, by definition, have an adequate margin of safety.
      A decision also needs to be made regarding the time period.  This
subject is discussed more completely elsewhere*- •".   In brief, it appears
reasonable that the time period minimally be keyed  into  the five year
standard review cycle.

4.3  Mathematical Description of the Method
      In the proceeding section it was pointed out  that  the risk of exceeding
a true health effects threshold in a given period of time is determined by
the uncertainty in the location of the health effects threshold (or in the
case of multiple thresholds, the uncertainty in the location of the lowest
threshold) and the uncertainty in the maximum oxidant concentration over the
given time period.  It can be shown, by application of the theory of proba-
bility, that the following equation gives the relationship between the
risks and the above uncertainties:
          =  1 -   Pc(C)pT(C)dC                                            (4-3)
where:
      R  = Probability (Risk) that a true health effect threshold,
           or the lowest of a multiple number of thresholds, is ex-
           ceeded one or more times in a given time period (e.g. one
           year, five years, etc.).
  PC(C)  = Probability that the highest observed time averaged (e.g.
           one hour, two hour, etc.) pollutant concentration does not
           exceed the concentration C in the given time period.

-------
                                    101
   P,(C) =  The probability density function for the health effect
            threshold or in the case of multiple health effects the
            function for the lowest effect (the composite density
            function).
   The derivation of Equation (4-3) is given in Appendix B.
     Equation (4-3) also holds for a more general  case in which R is defined
as the risk of m or more exceedances of a threshold in a given time period,
where m  may have the integral values 1, 2, 3, etc.  In this case PC is
redefined as the probability that the mth highest time averaged pollutant
concentration does not exceed the concentration C in the given time period.
In other words, PC is the cumulative distribution of the mth highest time
averaged pollutant concentration in the time period of interest.  Specifying
the national ambient air quality standard for a pollutant limits the range
of PC(C) functions which will satisfy the air quality requirements of the
standard.  If this limitation can be expressed quantitatively, and the
health effect threshold density functions have been determined, then for
any given specification of the standard, a range of risks associated with
that specification can be calculated from Eq. (4-3).
     In practice, it is only convenient to use Eq. (4-3) directly to
calculate risk when a single health effect with an existence probability
of one is involved.  When the risk that lowest of n health effects will
be exceeded is to be calculated and when there is  uncertainty as to whether
one or more of the effects actually occur in the sensitive population at
any attainable pollutant concentration, the following expanded version of
Eq. (4-3) is used.  (See Appendix B):

-------
                                   102
       R • 1 - J Pc(C)p° dC - (1  - ei)(l  -  e2)----(l  - en)                  (4-4)
where:
      p  = The probability density function  for the  location of the
           lowest of n thresholds  over the possible  range of concen-
           trations of the pollutant.
     e.j  = The probability that the ith health  effect actually occurs
           in the most sensitive population  in  the possible range of
           concentrations of the pollutant.
  The function p" is calculated from:
                                                                           <«>
where:
      p? =  The probability density function  for  the  threshold of the
            ith health effect assuming its  e.  = 1.   (That  is,
                  CO
                 r  o
                 J pi       >'
     In practice p! is obtained by asking  the  subject  health expert  in an
encoding session to first give his best judgment of the  value of e-j  and
then encoding him as to the location of the  threshold  assuming that  the
effect actually occurs in the sensitive population  at  an attainable  pollu-
tant concentration.  The encoding procedure  for the location of health
effects threshold gives the cumulative distributions for the health  effects

-------
                                     103
thresholds.  These functions are differentiated by numerical methods to
obtain the p^.
     The terms Q. and Q in Eq.. (4-5) are calculated from:
                       C
        Q.  =  1 - e. J  pt dC                                             (4-6)
        Q   -  Q,(J& •- Qn
The Q^ are seen to be functions of the cumulative distribution of the
                               o
probability density functions p^.
     The approach used above in treating uncertainty about the existence
of a causal relationship between the pollutant and a given health effect
is to define a density function PT assuming the effect exists and a proba-
bility e that the effect is caused by the pollutant to occur in humans.
The area under the probability density functions p° is one.  The approach
multiplies this function by e, thus reducing the area to e.  To bring the
total area to one the remaining area (1-e) is assumed to be accounted for
in a concentration range very much above the concentration range which
effectively contains p,-.  This situation is illustrated in Fig. 4-1.  The
distribution on the left is ep .  The area under this function is, therefore,
e.  The function to the right can have any shape.  However, the area under
the function must be (1-e).  A more detailed treatment of the approach is
given in Appendix B.

-------
                              104
Figure 4-1   Hypothetical  Threshold Probability Density Function
            for a Health  Effect with a  20% Chance of Being  Real
                 0.15
C a 100,000 ppm
                       Concentrations,  ppm

-------
                                      105
     When more than one independent health effect is involved,  p,. is the
distribution of the threshold for the lowest effect.  This distribution is
a composite of the distributions of the individual  thresholds and is formed
from them in accordance with Eq. (4-5).  Figure 4-2(a) shows two threshold
distributions for two hypothetical, independent health effects.  Figure
4-2(b) shows three different composite distributions formed by  assigning
different probabilities of existence to the two health effects.  A further
discussion of compositing and the effect of assigning existence probabili-
ties is given in Appendix C.
     Thus, given the Pr(C) functions corresponding to different levels
                      w
of the standard, the probability density functions p! for n independent
health effects and their corresponding e. > Eqs. (4-4) through (4-7) can
be used to calculate the range of risks associated with alternate
specifications of the ambient air quality standard.  The risks  can be
calculated for the individual health effects and for composites of two
or more of the effects in any combination.  Calculating risks for individ-
ual health effects and various combinations can be of value where the
effects differ significantly in their seriousness.
     The calculations involved in Eqs. (4-4) through (4-7) can  be most
conveniently carried out with a computer.  The function p°  can also
be calcualted by differentiating its cumulative distribution function.
The cumulative distribution is a function of the existence probabilities
e- and the cumulative distribution functions of the p?.   (See Appendix B.)
Less computational labor is involved with this method if there  is no
                                           o
interest in knowing the density functions p .
                                           i

-------
                                         106




Figure 4-2   Variation in Composite Health Effect Threshold Probability

          Density Functions of Two Independent Density Functions as Different

        Probabilities of Existence are Assigned to the Lower Threshold Function
                                                   = 0.20 ppm
                                                a. = 0.04 ppm
                                 0.2           0.3           0.4


                                      Concentration, ppm


           (a) Individual  Health Effect Threshold Probability Density Functions
                                                             0.4
0.5
                     0.2           0.3


                          Concentration,  ppm



(b)   Composite Probability Density Functions for Various Probabilities
                                                A

     of Existence are Assigned to Function with T = 0.2 ppm

-------
                                     107
4.4  Obtaining the PC(C) Distributions
     As indicated in earlier sections the Pp(C) function is the cumulative
distribution function for highest time averaged pollutant concentration
for a specified period of time.  If the risk is calculated for m or more
exceedances, it is the cumulative distribution for the mth highest concen-
tration.  To simplify the following discussion it will be assumed that the
concentration averaging time is one hour.
     The Pr(C) function for a pollutant in the air over a given region is
          u
a measure of the air quality for that region with respect to the pollutant.
Specifying a National Ambient Air Quality Standard places a limitation on
the range of PC(C) functions corresponding to air quality just meeting the
standard.  For example, if the ambient air quality standard specified an
expected (average) value of maximum hourly average concentration for  a one
year period was not to exceed a given level, this would immediately locate
the mean value of the distribution of maximum values and thus define the
concentration region in which the preponderance of maximum values must
occur.  However, depending upon the area and the control methods  used to
meet the standard, the distribution of maximum values about the mean could
be relatively narrow or spread out.  It is expected, however, that there
would be practical  limitations on the degree of spread of the distribution.
Therefore, specifying the expected maximum concentration limits the P_(C)
                                                                     0
distributions just satisfying the standard.
     In applying the risk assessment method it is necessary to determine
the range of PC(C) functions just meeting each alternate specification of
the standard.  While, in principle, this should be possible for almost any

-------
                                     108
type of standard it is more readily done for standards with statistical
forms than for standards with deterministic forms.
     The PC distribution function is related to the distribution  of hourly
average concentrations.  However, as discussed in Section 3 the presence
of correlation between hourly average concentration observed in different
hours and dependence of concentrations on time of day or period of the
year can strongly affect this relationship.  Air pollutants commonly show
this correlation and time dependence and these effects must, therefore,
be taken into account in developing suitable P~ functions.   Again, as
discussed in Section 3, taking these effects into account may make  use of
the case in which independence of hours and no time dependence  of hours
are assumed.  This case will be discussed first.
     If no correlation or time dependence of hours exists,  Eq.  (3-22)
applies, namely:
        Pc  -  (1 - G(C))                                                  (4-8)
where P- is the distribution of the highest concentration for n hours.
(n= 8760 hrs. for one year or 43,800 hrs.  for 5 years.)   The function G(C)
is defined by Eq. (3-4):
       fi(C)  -  Pr [CQbs > C]                                              (4-9)


That is, G(C) is the probability that an observed hourly average concentra-
tion is greater than C.

-------
                                   109
     If the broader definition of Pr is used, Eq. (3-26) applies, where
                                   \*
G(C) is substituted for p:
                 v=0
where p    is the distribution of the mth highest hourly average concentra-
       \s
tion for n hours.
     Thus, if the distribution function G(C) is known, the desired P-
function can be obtained from application of Eq. (4-8) or (4-10).  Studies
have found that the distributions of short-term time averaged concentra-
tions of air pollutants can usually be represented by lognormal, Wei bull,
                               T51
or gamma distribution functions   .  Of these the Weibull function provides
a good fit to photochemical oxidant air monitoring data^- •* and is convenient
to use since its G(C) function can be stated explicitly as in Eq. (3-6).
       G(C)  = e" (C/6)                                                    (4-11)
The parameter 6 is referred to as the scale factor.  It is the concentra-
tion corresponding to G(C) - 0.368.  It establishes the approximate position
of the mid-concentration values of the distribution.  The parameter k is
called the shape factor.  It tends to be a measure of the spread of the
distribution.  The larger k the more compact the distribution.  If the

-------
values of k and 6 have been determined for a given geographic area the
corresponding PC functions can then be obtained by use of Eqs. (4-8) and
(4-11) or Eqs. (4-10) and (4-11) at the given level  of air quality.
     For the risk assessment it is necessary to connect alternative
levels of the ambient air quality standard with the corresponding PC
function.  This is easily done through the Weibull distribution,  Eq. (4-11).
The proposed form for the ozone standard is:

        CSTD ppm hourly average concentration with an  expected
        number of exceedances per year less than or equal to E.

It is shown by Eq. (3-12) that for any region to which the Weibull function
applies and just meets the standard:
        6(C)  -  e -      E         ST                                     (4-12)
where:
        C<-Tn  -  level of ambient air standard.
           E  =  expected number of exceedances in n_ hours.
The term n£ is customarily the number of hours in one year or 8760 hours
The expected exceedance rate would normally be one for an air standard.
In this case Eq. (4-12) becomes:
        8(C)  -  e-9'078

-------
                                 Ill
It should be pointed out that from the point of view of the risk assessment
method developed in this report, the designation of the expected number of
exceedances is a relatively arbitrary matter and could be set at any value
that gives a convenient level for C-TD so long as the risk is the same
                    (m)
(the value of m in P~   of Eq. (4-10) has a more direct impact on health
since it directly bears on the number of exceedances of a true threshold).
For example, if it is decided that it is undesirable to have any exceedances
of a true threshold over a given time period then the Pr used in the calcu-
           (1)                                         C
lation is PC   (see Eq. (4-10)) and once an acceptable level  of risk is
chosen any combination of E and C-TD values which yields this risk value
in a given area give the same level of protection.
     The relationship between the G(C) and P~ distribution for a Weibull
distribution of hourly averages is shown in Fig. 4-3.  The combination C-,-
and E determine the general location of P~ while k determines its spread.
Figure 4-4 shows the P  functions for a one year period for a series of
                      u
alternate levels of C    with E = 1 and for the Wei bull  shape factor k = 1.
Figure 4-5 shows the effect of changing the shape factor at CSTQ = 0.1 ppm
and E = 1.  Figure 4-6 shows the effect of changing k for a PC covering a
5 year period.  It is seen form the three figures that changing CSTD
displaces the PC function over a wide range while having a relatively small
effect on its shape.  Changing k causes little actual  displacement of the
PC for a one year period but has a large effect on its shape.   The changing
k causes the P- function in Fig. 4-5 to pivot about the point (0.10, 0.368).
The effect of k on the PC for a five year period (Fig.  4-6) is still largely
in the shape of the function, but there also seems to be more displacement.

-------
                                     112
Figure 4-3  Hypothetical  Distribution of Hourly Average Concentration and the
        Corresponding Distribution of the Annual Maximum Hourly Average Concentration
                      •2     -4     .6     .8    1.0   1.2   1.4     1.6
                                    Concentration,  C/C<-Tn
                  (a)  Distribution of  Hourly Average Concentration
                 0     -2    .4    ,6    .8    1.0   1.2   1.4    1.6
                                    Concentration, (C/C$TD)
                  (b) Distribution of Annual  Maximum Hourly Average Concentration

-------
                                                              113

 O
O_
 dJ
                 O
                  II
                  O
                   II
                  UD
                  O
                    •
                  O
                   II
                    O
                    CO
                  O
 01
 3
 cr
                                                                                                                              Q-
                                                                                                                              o.
                                                                                                                             O
                                                                                                                             c
                                                                                                                             O
                                                                                                                             +J
                                                                                                                             (O
                                                                                                                              Ol
                                                                                                                              u
                                         CO
                                         C3
                                                          UD
                                                          c
CM
O
                                                              O
                                                             a.

-------
                                                                   114
 s_
 o
 u
 OJ
 O
(/I
VI
QJ
 c
 o>

 OJ
M-
 5-
 (D
Q.
 O)


 O)
 c
o
 (J
 c
 3
in
 i
 o>
 3
 CD
                      LO



                      O
                       CC


                         •

                       O
                                                                                                                      o
                                                                                                                      c
                                                                                                                      (U
                                                                                                                      o
                                                                                                                      c
                                                                                                                      o
                       o

                         «
                       c
                                                                                                                o
                                                                                                                  •
                                                                                                                c
                                                                                                                00
                                                                                                                o
                I


               o
                                 00

                                 o
CM

o

-------
                                          115
            Figure 4-6    P~  Function  for  5  Year  Period  for  Different


                             Values  of Wei bull  Scale  Factor, k
      1.01-
      0.8
      0.6
£
D-
      0.4
      0.2
        0.09
                               k = 2.0
                                     Concentration, ppm

-------
                                       116
This results from the fact that the pivot point for the distribution is
now very close to the concentration axis (0.10, 0.007).  In general,
calculated risk values will  be more sensitive to changing values of  C^jD
than to changes in k over the usual ranges of these parameters.
     From the proceeding discussion it is seen that the assumption of
independence of hours and the use of the Wei bull function to represent
the distribution of hourly average concentrations readily yield  Pr and
 (m)                                                             L
PC   functions.  The Weibull  can be used with little loss in accuracy
even where other distributions such as the lognormal distribution  provides
a better fit to the concentration data.  The primary concern in  estimating
the appropriate P  function is to have it placed properly along  the con-
                 v*
concentration axis and have the correct degree of spread.  The parameters
6 and k in the Weibull function provide wide flexibility in this regard.
As shown above, the standard  level essentially places the P- function.
The appropriate values of k can be obtained by fitting a Weibull distri-
bution to hourly average concentration data obtained from air monitoring
sites.  The range of applicable k values for standard setting purposes
should be determined by examining aerometric data in areas that  are close
to the concentration ranges of the alternate levels of the standard under
consideration.  Where this is not possible, the k values can be  determined
at existing levels of air quality and the results extrapolated to  potential
standard levels.
     While using a Weibull distribution where a loqnormal or gamma distri-
bution might be more appropriate does not appear to lead to serious errors
in the risk estimates, ignoring the possible dependence of hourly  average

-------
                                    117
concentrations can lead to significant error.  As discussed in Section 3,
internal EPA study*- -* showed that dependence of one hourly concentration
on the value of another did not lead to serious errors if independence of
hours was assumed.  It also showed that dependence of concentrations
on time of day or year can lead to PC functions which were placed lower on
the concentration axis than would be obtained assuming no time dependence.
When this time dependence can be modeled it should be possible to generate
the corresponding PC tables of functions and express them mathematically.
This was done in the EPA study for the daily maximum hourly average ozone
concentration.
     Another approach can be taken as discussed in Section 3, if the time
dependence is such that the maximum concentration tends to occur only
within some determinable period and the probability distribution of con-
centrations is approximately the same for all hours within the time period.
In this case all hours outside the time period can be excluded and inde-
pendence assumed for the hours within the period.  A Weibull  distribution
then could be fit to the hours within the time period to determine the
appropriate k values and Eqs. (4-8) and (4-12) used to calculate PC .  To
the extent that the period under consideration is also likely to contain
the mth highest hourly average concentration, P_   could be obtained
                                               w
using Eq.  (4-10).  The term n£  in Eq. (4-12) would be set equal to the
hours in one calendar year of the time period.  The term n in Eqs. (4-8)
and (4-12) would be n- times the number of years for which the risk is to
be estimated.  In the following section, this procedure is applied to
ozone.

-------
                                    118
4.5  Determination of PC Functions for Ozone

     As suggested in the proceeding sections estimating the appropriate
P- function involves:  1. determination of mathematical  function  or functions
which best describe the distribution of time averaged ambient concentrations
of the pollutant;  2. given the distribution of time averaged concentration,
the standard, and the nature of the correlation between concentrations  in
neighboring time periods and dependence upon time of day and year,  derive
sutiable P- functions and  3. estimate the range of values  of parameters
appearing in the P~ function.
     A study performed under contract for EPA involving 14  sites  scattered
around the United States and involving 22 site years of data showed that
the Weibull distribution provides an excellent fit to hourly ozone  concen-
trations   .  In only two cases did a lognormal distribution give a
superior fit.
     Ozone hourly average concentrations exhibit strong correlation and
strong dependence on time of day and time of year.  The day of the  week
also has some effect.  As indicated in Section 4-3, the correlation
between neighboring concentrations does not appear to have  an important
effect on the PC function; however, the time dependence does.
     The method of dealing with the time dependence, as discussed,  was  to
find a time period in which the time dependence was relatively constant
and which was highly likely to contain the maximum hourly average concen-
tration for the time period.  If such a time period existed and a distri-
bution could be fit to the time-averaged concentrations within this period,
then the appropriate PC function could be derived by assuming complete

-------
                                  119
independence of the time averaged concentrations.  For ozone this period
is during the midday hours of the warm months of the year for urban areas.
     In a continuation of the above mention study*- -" it was found that
Weibull distributions could be fit to data obtained between 11  AM and 6 PM
both from May through September and July through August.  It was further
shown the maximum ozone concentration had significantly more chance of
occurring during the longer of the two periods.  Therefore, this time period,
which contains 1071 hours, was used in the derivation of the P/. function.
     The form of the National Ambient Air Quality Standard proposed for
ozone is:  C<.Tn ppm hourly average concentration with an expected number of
exceedances per year less than or equal to one.  For a region whose air
quality just meets this standard the Weibull  distribution of hourly averages
for the hours 11 AM to 6 PM, May through September, would be according to
Eq. (4-12):

           G(c)=e-(lnl071)(C/CSTD)k                                     (4.u)
And the Pr function for a period of n, years would be from Eq.  (4-8)
         *"                           y
                                           ""  ny
     By substituting alternate levels, C-T_,  of the standard into Eq.  (4-15)
the Pr function for n  years needed for the risk assessment  can  be obtained.
     **               y
However, before this can be done it is necessary to estimate the range of
values of the parameter k (Weibull  shape factor) over different  regions in

-------
                                     120
the U.S.   Table 3-11  shows measured k values for the 1071  hour time period
as well as other time periods.  The range for the values based on 1071
hours is 1.31 to 2.04.
    For the risk assessment best estimates were made of the lower bound and
upper bound values of k.  This was done by probability encoding the contractor
and a PEDCo researcher involved  in the development of the  Weibull  distribu-
tions and the PC functions.  The information base was the data in Table 3-11,
plus the data developed during the Weibull distribution studies.   The data
shown in Tabel 3-11 are, by and large, for geographic regions above the range
of alternative ozone standards considered.  The Weibull studies suggest that
the k factors at the standard levels would be somewhat higher than those
shown in Table 3-11.  This factor was taken into account in the encoding.
The contractor first encoded Mr. T. Johnson of PEDCo and then himself.   The
results of the encoding are shown in Tables 4-1 and 4-2.  In encoding Mr.
Johnson  the upper and lower bounds of the variable were first encoded
                                          [3]
followed by the use of the interval method   .   The subject was  then allowed
to make slight adjustments based on viewing the plotted probability function.
When the contractor was his own subject the same approach was used but  the
final adjustments were made directly to the plotted curve.   In this way a
more completely defined curve was obtained.
    The distributions were combined by Mr. T. Feagans of EPA into a single
distribution for the lower bound and a single distribution  for the upper
bound (Table 4-3).  The median values for the location of the lower and
upper bound shape factors were 1.36 and 2.54 respectively.   Since the
range of k values varies somewhat with the standard level,  the range,
strictly speaking, should be estimated for each alternative level of the
standard.  In the case of ozone the difference is not likely to be large
enough to seriously affect the risk estimates.

-------
                                       121
Table 4-1     Subjective  Probability Distributions of Lower and Upper Bounds
              of Weibull Shape Factor k For n at Approximately 1000 Hours
             and Maximum Concentration at Approximately 0.1 ppm
Subject: Ted Johnson
PEDCo
Lower Bound k
Cum.
k Probability
1.30 0.001
1.38 0.25
1.43 0.50
1.48 0.75
1.50 0.999
Interviewer
Date:
Upper
k
2.30
2.37
2.50
2.70
3.00
: W.F. Biller
1/26/78
Bound k
Cum.
Probability
0.001
0.25
0.50
0.75
0.999

-------
                                       122
Table 4-2  Subjective Probability Distributions of Lower and Upper Bounds
           of Weibull Shape Factor k For n at Approximately 1000 hours
           and Maximum Concentration at Approximately 0.1 ppm
           Subject:  W.F. Biller             Interviewer:  W.F. Biller
                                             Date:          1/31/78
Lower
k
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
Bound k
Cum.
Probability
0.01
0.02
0.05
0.10
0.16
0.27
0.60
0.86
0.95
0.99
Upper
k
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3.0
3.1
3.2
3.3
3.4
3.5
Bound k
Cum.
Probability
0.01
0.09
0.30
0.60
0.75
0.83
0.88
0.91
0.94
0.96
0.97
0.98
0.99

-------
                                         123
Table 4-3    Subjective Probability Distributions for Upper and Lower
             Bound Weibull  Shape Factors (k)  for Distributions of Hourly
             Average Ozone  Concentrations
                   Probability
that k is
below specified
value
0.10
0.30
0.50
0.60
0,90
Lower
Bound
k
1.15
1.30
1.36
1.41
1.47
Upper
Bound
k
2.41
2.50
2.54
2.66
2.98

-------
                                      124
4.6  Example Application of Method
     Suppose that three health effects have been identified for a given
pollutant and that there is some uncertainty as to the precise location
of each of the effects.  Suppose further that the evidence for one of
the effects is very weak at the present time and there is  considerable
doubt the effect exists.  However, if it does exist,  it is a serious
effect and,therefore,it should have some bearing on the setting of the
standard.  Suppose there is also some doubt about the existence of one
of the other effects,  but in this case the effect is  not considered to
be too serious.  However, it also should be taken into account.  It will
be assumed there is no question about the existence of the third effect
which is considered to be serious.  A thorough review of the information
available on each health effect by a group of medical experts establishes
that the uncertainties in the locations of the effects are best described
by normal distributions.  The findings of the group are summarized as
follows:
Health
Effect
1
2
3
Most Probable
Location of
Threshold
(ppm)
0.15
0.20
0.30
Standard
Deviation of
Density Function
(ppm)
0.05
0.04
0.08
Probability
Effect
is Real
15
80
100
Seriousness
of
Effect
Serious
Mild
Serious
In a real situation much more information would be supplied concerning the
seriousness of the effect.  It also would not be possible to describe the
probability density function for the location of the threshold so easily.
A tabulation of the function versus concentration for each effect would
have to be developed.

-------
                                     125
      Simple inspection of the table does not provide a clear indication
as to where the level of the national ambient air standard should be
set to provide an adequate margin of safety against all three effects
simultaneously.  However, application of the procedure outlined in the
previous section gives an assessment of the medical risks involved at
each potential level of the standard and can be used, therefore, in
judging where the standard can be set to give an adequate margin of safety.
      The margin of safety requirement means that an exceedance of the
standard concentration will not, if the margin is adequate, generally
result in an exceedance of a true health effect threshold.  To apply
the approach it is first necessary to adopt a form of the standard that
provides protection appropriate to the medical effects under consideration.
Suppose the effects are such that any exposure for a short period (such
as one hour) over the thresholds are to be avoided.  In this case one
appropriate form would be:
        C^ pom hourly average with expected exceedances per year
             less than or equal to one.

This form is close to that of the current oxidant standard but differs from
that form in that its statistical nature allows it to be directly related
to yearly distributions of hourly average concentrations.
         It is assumed that a Wei bull distribution with k = 1  fits the
ambient concentrations and there is no time dependence (i.e.,  n,- = 8760 hrs.)
Also assume the risk is to be calculated for a one year period.   It follows
from the discussion in Section 4, 3 that the P~ function is:
         . (,.e-0n 87«0)(C/CSTD),
                                  87M

-------
                                     126
 Equation (4-16) plus the information supplied in the above table on the
health effects thresholds can then be used to calculate for each level
of the standard:  1.  The risk, R, that no hourly average concentration
will  exceed any of the true health effect thresholds in a given year
and 2.  The individual risks, R^,  that no hourly average concentration
will  exceed the i-th true threshold in a given year.  For the three
thresholds the risk R. .  that no hourly average concentration exceeds
                      • »J
either the i-th or j-th true threshold can also be calculated.  The
calculation of these risks for each standard level is achieved by:
      1.  Developing the appropriate Pr function for each level  at
                                      w
          which the standard may be set, in this case Eq.(4-]6).
      2.  Forming the composite distribution function p° (C) for
          all three thresholds from Eq.(4-5).
      3.  Determining R by numerically integrating the product of
          P (C) and p° (C) as indicated by Eq.(4-4).
      4.  Determining each R. by carrying out  the process in Eq.(4-4)
          for Pr(C) and pT  (normal distributions in this example).
               t         i.j
      5.  Repeating ^3 and *4 for any composite of two threshold
          functions.
When this procedure is carried out for the above three health effects the
tabulation shown in Table 4-3 is obtained (Details of the calculations are
given in Appendix D).
     It is seen that if the standard is set at 0.10 ppm, the margin of
safety is such that there is a 0.05 risk (or one chance in 20) that a
health effect threshold will be exceeded in a given year.   At the same

-------
                                       127
Table 4-4  Risks of an Hourly Average Concentration Exceeding True
           Health Effect Thresholds in a Given Year for Different
           Values of a National  Ambient Air Quality Standard
Standard
(ppm)
0.06
0.07
0.08
0.09
0.10
0.11
0.12
0.13
0.14
0.15
0.16
0.17
0.18
RMT
0.009
0.013
0.022
0.033
0.050
0.072
0.102
0.141
0.190
0.248
0.315
0.383
0.457
R1
0.007
0.010
0.015
0.022
0.030
0.040
0.050
0.062
0.073
0.085
0.096
0.106
0.115
R2
0.001
0.001
0.002
0.005
0.011
0.022
0.040
0.066
0.101
0.147
0.202
0.265
0.334
R3
0.002
0.002
0.004
0.006
0.008
0.012
0.017
0.025
0.034
0.045
0.060
0.077
0.098
Rl,3
0.008
0.013
0.019
0.028
0.038
0.052
0.067
0.085
0.104
0.126
0.149
0.175
0.202

-------
                                    128
time the risks of exceeding the individual  thresholds are 0.03 (1  in 33),
0.011 (1 in 91), and 0.008 (1  in 125).   Since the first and third  effects
are serious, this level of protection might be considered as the minimal
acceptable value by the EPA decision makers.   If so, the standrard would
be set at 0.10 ppm.  Considering, however,  that the second effect  is mild,
it is of interest to determine the probability of not exceeding either the
first or third threshold.  This probability is shown in Column 6 of the
table.  If the .05% level is still considered acceptable the standard could
be set at 0.11.  This level would still  give a 96.0% probability of not
exceeding the mild health effect.
    The above example is greatly simplified since it is primarily  intended
to show how the equations developed in the previous section are used.  In
actual application of the method other considerations would enter.  Further-
more, considerable power can be added to the method through a sensitivity
analysis of the uncertainties involved in the determination of the PC and
                                                               T41
P  functions.  These added considerations are treated elsewhere1- J.

-------
                                   129
4.7  Conclusions and Recommendations
     It is concluded that the method of risk assessment presented in this
section is a practical method for developing national  ambient air quality
standards which are in accord with the intent and spirit of the Clean Air
Act as presently written.  The method leads naturally to the use of standards
expressed in statistical forms.   It gives a proper perspective to viewing
exceedances of a standard level.  The real concern is to protect against
pollutant concentrations exceeding health effects thresholds and to set
the standard at a level which reduces the risk of such exceedances to an
acceptable level.  Exceedances of the standard level  do not necessarily
mean a health effects threshold has been exceeded and the risk that they do
can be set as low as the EPA Administrator deems necessary to protect the
public health.
    The risk assessment method also is a logical starting point for:
1. estimating the expected number of people in a population who will suffer
health effects at given air quality levels and the uncertainties in such
estimates and 2. Cost-benefit analyses.
    In those aspects of the method to which the contractor contributed there
is a need to: determine P£  functions accurately,  determine the extent to which more
accurate functions can be developed,and to obtain more complete information
on the range of P- functions experienced in different parts of the United
States.  In investigating the question of improved P- functions it will  be
necessary to further study the interdependence of time averaged concentra-
tions and the nonstationarity of the underlying stochastic  processes which
are responsible for observed pollutant levels and to  determine how these
factors affect the PC functions.

-------
                                      130
     There is also a need to better define the behavior of aerometric


data and, therefore,the P  functions at the standard level.
                        V

     It is recommended that the above mentioned problem areas be further


investigated.  In addition it would be worthwhile to extend the basic


method to estimating expected numbers of people affected at air pollution


levels above the health effects thresholds and the attendant uncertainties


and risks.

-------
                                     131
4.8  References

1. Clean Air Act, Section 109,  42 U.S.C.
2. "A Legislative History of the Clean Air Amendments of 1970",  U.S.
   Senate Committee on Public Works, 1970.
3. C.S. Spetzler and C.A.S.  Stael von Holstein,  "Probability Encoding
   in Decision Analysis", Management Science, Vol.22, No.  3, November 1975.

4. "A Method for Assessing Health Risks Associated With Alternative Air
   Quality Standards for Photochemical Oxidants",  Report by U.S.  Environ-
   mental Protection Agency, Office of Air Quality Planning and  Standards,
   in press.
5. T.C. Curran and N.H. Frank,  "Assessing the Validity of the Lognormal
   Model When Predicting Maximum Air Pollution Concentrations",  Presented
   at 68th Annual Meeting of the Air Pollution Control  Association,
   Boston, June 15-20, 1975.
6. "The Validity of the Weibull  Distribution as  a  Model  for the  Analysis
   of Ambient Ozone Data".   Draft Report to EPA by PEDCO Environmental,
   November  17, 1977.
7. Private Communication, J. Horowitz, Office of Air and Waste Management,
   U.S. Environmental  Protection Agency,  Washington, D.C.

-------
                                APPENDIX  A

                       SUPPORT MATERIAL FOR SECTION 3



1.  Derivation of Approximate Relationship (3-18)


     Starting with Equation (3-16):



          n   .    n!     ,E »m ,,  E  ,n-m
The following approximately holds for n » m:
           (n-m!)
                  «  nm                                                     (A-2
If define a constant X such that:



           X  »  !|-                                                        (A-3)





Eq. (A-l) is approximated by:
                 mm      p  n-m
           Pm - x  E  (1 - I-)                                              (A-4)
If E « n  the term in brackets can be expanded as a binomial series to

yield:
                               2!        3!
                                  132

-------
                                     133
The approximation series expression is the expansion of the exponential function.
Therefore:

               c  n-m    _^£
               it p

Substituting into the approximation (A-4) yields:
which is Expression (3-18) appearing in Section 3.

2.  Demonstration That the Concentration Corresponding to the Most Probable
    Value of the mth Highest Concentration in n Hours is the Concentration
    With an Expected Exceedance Rate of m.
     This relationship is true if the distribution of hourly average con-
centrations is described by a Weibull function with k = 1.  To find the
concentration corresponding to the most probable value of the mth highest
concentration in n hours the derivative of Eq. (3-27) is set equal to zero.
To simplify the expression use the 6 notation for the Weibull and set k = 1.
Thus:
                                             -                             (A-6)

-------
                                     134
     If the derivative of Eq.  (A-6)  is taken with respect to C and all
terms cleared that are not zero at the maximum value,  the following expres-
sion is obtained.

            *  = e" C/6                                                    (A-7)

Equation (A-7) has the same form as  Eq.  (3-8):
            I  -  .-                                                       (A-B,
Therefore:
            C° = CE                                                        (A-9)
            E  = m                                                         (A-10)
3.  Demonstration of Expression (3-37)
     Using the following series expansion:
        (1 - X)" = 1 - nX  +  HlHi
                                2!

-------
                                       135
      the right hand side of Eq. (3-35) expands to:
                       = 1 - nQp  for p very small
Similarity the right side of Eq. (3-36) also becomes approximately (1 - n p)
for small p.  Since we are concerned with the highest observations, the
value of p is much less than one and the approximation should hold closely.
Note also,this result is independent of the value k.

-------
                                      136
                            APPENDIX B
        DERIVATION OF BASIC EQUATIONS FOR ASSESSING HEALTH  RISKS
            ASSOCIATED WITH ALTERNATIVE AIR  QUALITY STANDARDS
1.0  Risk of Exceeding a True Health Effect Threshold or the  Lowest  of
     Several Thresholds
     First determine probability P where:
          P =  Probability that np_ hourly  average  concentration  exceeds
               the true health effect threshold or the lowest of several
               thresholds in a given period.
     Let:

      P-(C) =  Probability that no time averaged concentration exceeds
               the concentration C in the  given period.

      Py(C) =  Probability density function expressing the uncertainty
               in the location of the true threshold or the lowest of
               several thresholds.
     Note:

         fb
        j PT(C)dC = Probability that a true threshold or the  lowest  of
        a           several thresholds is  in the interval  a,b.

    It follows that:

     Pr(C)pT(C)dC = Probability that the true threshold or the lowest of
      w     1
                    several thresholds is  contained in the interval  dC
                    and this value is not  exceeded by any hourly average
                    concentration in the given period.

-------
                                    137
By integrating over all values of C the quantity P, defined above, is
determined:
                 oo
          P  =    Pr(C)PT(C)dC                                             (B-l)
                i  C    T

If R is the probability or risk of exceeding a health effects threshold or
the lowest of several thresholds one or more times in the given period, then:

          R  =  1 - P
          R  a  1 -    Pc(C)pT(C)dC                                        (B-2)
     Note that if the function PC is defined as the probability that no time
averaged concentration exceeds the concentration C it must also be the pro-
bability that the highest time averaged concentration observed in the time
period does not exceed C (therefore, is < C).  In other words, Pr is the
                                        ~~~                       L
cumulative distribution of the highest time averaged concentration occurring
in the given time period.
     It is further noted that if it is desired to estimate the risk of
exceeding a health effects threshold or the lowest of several  threshold
"m" or more times in a given time period, then P- becomes the  probability
that a threshold will not be exceeded more than m-1 times in the given time
period.  In this case P~ is the probability distribution of the mth highest
time averaged concentration.
2.0  Determination of Probability Density Function for the Lowest of Several
     Health Effect Threshold
     Assume n different health effects and that a probability density function
has been obtained for each.   If the function Pp(C) is known,  Eq.  (B-l) can
then be used to evaluate the individual  probabilities for each health effect.

-------
                                     138
          P.  =    Pr(C)PT (C)dC                                           (B-3)
            '     J  u    ' •{
                 o        '
     Let:
          P  ~  The probability that no hourly average concentration ex-
                ceed any of the n true thresholds in the given time period.

To evaluate this probability for all possible configurations it is assumed
that each threshold in turn is the lowest threshold.

     Differential elements of the following type can be formed:
                          *•**                 ww              OQ
     Pr(C)PT (C)dC   •   f PT (x)dx   '    f PT (x)dx---- f PT (x)dx
      L    'l            J   T2             J   T3          J   Tn
                         C                 C              C
This term is the probability that:  1. no hourly average concentration
exceeds the value C;   2. the threshold T-j is in the interval dC;  and
3. all the other thresholds are above this value.


     If:
                     CO
          Q.(C)  =  J PT (x)dx

                    C
     And:
          Q(C)   =    TTQ(C)
                     1=1  1

-------
                                    139
The above expression simplifies to:
          pr 2- p_  dc
           L WT  11
Integrating over the concentration range gives:
This integral is the probability that no hourly average concentration exceeds
any of the n thresholds in the time period and T,  is the lowest threshold.
In order to enumerate all cases, it is necessary to form the sum of integrals
of this type in which each threshold in turn is assumed to be the lowest.

Therefore:

          P  =    V t P 2- P,  dC
or:
                  oo        n
          P  -   f Pr [Q
                 J        •  i
                 0        1=1

or:
          P  =  !  P. PT dC
                   w  I
This equation is,  as would be expected,  identical  to Eq.  (B-l)  except that
P  is now a composite probability density function such that:
         p,  =  Q   n   pT /Qi                                              (B-4)
          1          1-1   Ti

-------
                                   140
Where :
                  00
       Q.(C)  =  f pT> dC                                                  (B-5)
                 C   1

and
                  n
          Q   »   n  

and:
         D   =  Pr (T, < C)
           1
It follows from probability theory that:

                 n
       1-0T  =   n 0-D- )                                                (B-7)
          1       i=l    '1
The composite probability density function is by definition:

                dD
        p    =  -1                                                        (B-8)
         '       dC

Differentiating (B-7) yields:

        dDT        n              n        dD

-------
                                     141
Since by definition:
                      CO
          1-D-   =   f PT  dC = Q.                                         (B-10)
             TI      J  T,


            Q    =   n (I-DT )
                      i      Ti

           pT    =   dDT /dC                                               (B-ll)
            1 i          i

Equation (B-9) is equivalent to Equation (B-4).

     Equations (B-7) and (B-8) provide an alternate route to obtaining
the composite density function, p_, which can then be used in Eq. (B-2)
to estimate risk.
3.0  Inclusion of Uncertainty that One or More Health Effects Exist
     For some health effects there may be uncertainty that the effect
actually occurs in humans.  It would be desirable to include this consid-
eration when considering the uncertainty in the location of the threshold
on the concentration axis.  Given uncertainty only in the location of the
threshold it has been shown that Eq. (B-2) gives the risk that the true
threshold T^ is exceeded one or more times in a given period.
     If there is uncertainty as to whether the ith effect occurs in humans,
assign the probability e. that the effect does exist.  Then choose a value
of C = u such that u is many times larger than any concentration likely to
be encountered.  In other words, u is many times beyond the concentration
range of interest.  In this case, if pi (C) is defined as the probability
                                      i-j
density function for the location of     the ith effect if it does exist.
A new overall probability density function can be written:
          PT>(C)  =  e.p°  + (i-e.) 6(C-u)          i  = 1	,  n          (B-12)

-------
                                       142

 where 5(C-u) is the Dirac delta function.  It has the property:

                     = 1   if a <_ u <_  b

                                                                           (B-13)
[  6(C-u)dC
                       0  if a > u or b < u
     In other words, if the interval of integration includes C = u the
value of the integral is unity.   If it does not include C = u the integral
is zero.  The Dirac delta function itself may be considered as zero for
all values of C except u.  At u it has an infinite value.  It is also
assumed that pZ (C) is essentially zero in the vicinity of C = u and above.
              'i
     Equation (B-12) is based on the premise that saying an effect does not
exist is mathematically equivalent to saying that the true threshold is at
a very high concentration which is above any concentration likely to be
encountered.  Thus, if the probability e can be assigned to the certainty
that it does exist, the fraction e of the total area under the probability
density curve can be assigned to the concentration range in which the effect
is thought to be located if it does occur.  The rest of the area, (1-e), can
be assigned to a range above any concentration likely to occur.
     Note also, the use of the Dirac delta function is only a matter of
convenience since it leads to a simple form of Eq. (B-12).  Any function
can be used in this outer region as long as it has the value zero in the
region of interest and has the area (1-e).  The Dirac function is simply
a convenient form for -including the desired property in the probability
density function.
     Substituting Eq. (B-12) into (B-3) gives:
        P1 = e.    PpdC  +  (1-e.)     P  S(C-u)dC
Since Pr will be essentially 1  at C = u, the above equation yields:
       v»
           GO
 P. = e.  f PrpT dC  +  (1-e.)
  i    i  j  L \.            i
                                                                           (B-14)

-------
                                   143
To find the probability  that no  hourly average concentration exceeds any
of n thresholds,  it is necessary to derive the appropriate form of the
composite threshold probability  density function.  Its general  form is
given by Eq.  (B-4)
Substituting Eq.  (B-12)  into Eq. (B-4) and rearranging gives:

                   n  e               "  1-e-
        PT  •  Q  E ITPT.  +  [Q  Eor1^0-"*
The first term on the right  side of the Eq. (B-15) is evaluated in the
ambient concentration range  of  the pollutant well below the value C = u.
In this region the functions Q. have a simple interpretation.  Substitut-
ing Eq. (B-12) into Eq.  (B-5).
        Q.    =   e.   I  PT  dC  +   (1-e^     6(C-u)dC                        (B-16)
                    c   1               c
For all  C < u
                    00

        Q.    =   e.   j  p^dC  +   (1-e.)                                     (B-17)
                    ft    1
     The second  term on  the right side of the Eq. (B-15) is evaluated in
the vicinity of  C  =  u, far above the ambient concentration range.   In this
region the behavior  of the Q. functions needs to be more carefully consid-
ered.   The problem is the behavior of the second term on the right side of
Eq. (B-16) as C  passes through u.
     Note that in  the vicinity of C = u the first term on the right side
of Eq. (B-16) is zero by definition.

-------
                                 144
Thus:
                       oo

       Qi  »  0-e..)   I  fi(C-u)dC           for C -\, u                       (B-18)

                      C



Expanding the second  term of Eq.  (B-15)  gives:




    [ Q2Q3*'"Qn(1"el)  *  QlQ3%"Qn(1"e2) *  '" + QlQ2*"Qn-l(1"en) ] 6(C"u)




which on substituting  Eq.  (B-18)  yields:


                     oo

        n fi(C-u)  [ f 6(C-u)dC  ]""  J U (1-e^
The portion of this term in large  brackets can be shown to have the proper-

ties of a Dirac delta function  (Eq.  (B-13)) as follows:


Let:
                             00

           )  =  n 6(C-u)   [ f  6(C-u)dC  ]""                                (B-19)
and:
      <|>(C)    =  f 6(C-u)dC                                               (B-20)
then:
Substituting into the bracketed terms:
              = - n
                        dC

-------
                                      145
or

                       d*
                                                                            (B-22)
To show that the derivative has the properties of the Dirac delta  function,
integrate between the limits a and b.
                         dn = $(a)n - (b)n
                       6(C-u)dC ]"  - [  f (C-u)dC                                                          (B-24)
        a
                                  = 0 if a >  uorb<  u

where b > a.

Therefore, i|>{C-u) is also a Dirac delta function.
We can, therefore, write:

                           e.
           PT   =  (Q

-------
                                    146
where:
        CO
          o
=  A   '  *
          Q.  =  en.    p° dC  +  (1-e.)
and therefore:
or alternatively:
                        C
                         PT dC
and

          Q   »  0^2	Qn                                                (B-27)

Substituting Eq.  (B-25) into Eq.  (B-l)  gives:

                  00
          P   =  I Pcp°dC  +  II   (1-e.)                                    (B-28)
              =  1  -  j PcP°dC  +  II (l-ei)                               (B-29)
The Eq. (B-25) through (B-29) are the basic working equations for estimating
risk when a multiple number of health effects thresholds are involved,  and
where one or more of the health effects may not occur in humans.

-------
                                 147
                                APPENDIX C

      EXAMPLES OF COMPOSITE HEALTH EFFECT THRESHOLD DISTRIBUTIONS FORMED
       FROM SIMPLE HYPOTHETICAL DISTRIBUTIONS FOR INDIVIDUAL THRESHOLDS


    The following simple examples will  help illustrate the relationship
of individual health effect threshold probability density functions  and
their composite functions.  Assume two health effects with probability
density distributions.   The composite density function is:

         PT  =  e!Q2pl  + 62Q1P2 + O-e^O-e^ 6 (C-u)                     (C-l)
(The superscripts on the p. are omitted for convenience.)
Assume:  1.  p,  and p? are box-like functions;   2.  P2 lies beyond p,  on
the concentration axis with no overlap;  3.  both effects are certain  to
exist.  Eq. (C-l) then becomes:
                  °
         P  =  Q2P   +  Q.^2                                               (C-2)
or
                    °°           f
         p  =  P1   J p2dC +  p2 J
  P]dC                                     (C-3)
C

-------
                                   148
This situation is shown in Figure C-l(a).   The area under each box  is 1.
     In the interval [0,a], p,   and p« are both 0,  so that the composite
distribution PT = 0.  In the interval  [a,b],  p-j = p , Q? = 1,  p  = 0.
Therefore, pT = p .    Similarily it can be shown that above point b,
PT = 0.  Thus, the plot of the  composite interval (Fig. C-l(b))is  identi-
cal to the distribution p, alone.
     This result is  expected because in this  example it is known  with
certainty that Threshold #1 is  below Threshold #2.   Thus, the  probability
that the lowest threshold is in a given interval is determined only by
the density function for Threshold #1.
     Figure C-l(c) shows the composite probability  density function for a
different situation  in which the density functions  for Threshold  #1 and
#2 are both box-like functions  and are identical.  That is, both  have  the
same distribution as shown for  p, in Fig.  C-l(c).  Since the density functions
for the individual thresholds are uniform over the  interval [a,b], the most
probable point at which the lowest of the two will  be encountered is at the
lowest end of the interval [a,b].  If there are several identical, fully
overlapping box-like threshold  distributions, the composite distribution
becomes curvilinear (Fig. C-l(d))with more of its area displaced toward the
lower end of the interval.  The area under the composite function is,  of
course, one in both Figs. C-2(c) and (d).
     Figure C-2(a) depicts a situation similar to that shown in Fig. C-l(a)
in all respects except that there is only a 30% certainty that the first
health effect exists.  In this  case the probability density function for
Threshold #1 is divided between the region in which the threshold would
be located if it does exist and a region of very high concentration above
any concentration that could be encountered in the ambient atmosphere.

-------
                                        149
                                     Figure  C-l

                       Composite Threshold Probability Densities
                        When Health Effects  Are  Certain to Exist
Ti
                          (a)
              abcdOab
                          C+                                        C
(b)
                      n = 2
                                                                 n = 3
                                                 0
                   (O
(d)

-------
                                      150

                                   Figure C-2

           Composite Threshold Probability Densities When the Existence of
                     One Or More Health Effects  is  Not Certain
Ti
   0.3p0	
         0           a        be        d7X         e        f
                                          C *                       u

                                   (a)
    0.7pr
   0.3pQ-
                                       c        d
                                         C f
                                   (b)

-------
                                   151
For graphical purposes a simple box function is shown in this region
instead of the Dirac delta function.  Both yield the same result.
The composite function can be written:
          PT  =  Q2e1p1  +  Q1P2                                           (C-4)
where:
                                                                           (C-5)
          Q2  -      P2dC                                                  (C-6)
Note the absence of a third term in Eq, (C-4) which would give the value
of PJ in the region of C = u.  This term is zero because e? = 1 and,
therefore, (l-e-|)(l-e-)  6(C-u) = 0.  The graph of the composite function
(Fig. C-2(b)) is obtained by evaluating (C-4) at each point on the concen-
tration axis.
     In the interval [0,a] both p-j and p_ are zero and, therefore, p- - 0.
     In the interval [a,b] p, = p , p? = 0.  Thus, only the first term on
the right side of the equation contributes.  Since in this interval (L = 1
and e]  » 0.3, PMJ « 0.3 pQ.
     In the interval [b,c] again both p-  and p  are zero and, therefore,
PT = 0.

-------
                                   152
     In the interval [c,d] the p-j  term in Eq. (C-4) is zero, but the second
term, Q^p2 has the value (1-0.3 x 1) x pQ = 0.7 pQ.  Therefore, PT = 0.7 pQ.
     In the interval above point d, p, * 0, p~ = 0 and the 
-------
                                    153
                              APPENDIX D

            DETAILS OF CALCULATIONS PERFORMED IN SECTION 4.5
                      EXAMPLE APPLICATION OF METHOD

The table of risks in Section 4.5 was obtained as follows:
D.I  Develop the Appropriate P- Function for Each Alternate Standard Level

     Rearranging Eq. (4-16) of Section 4, the constant in the exponential
term can be calculated for each level of the standard:
                              -In R7fin    876°
          Pf(C)  =  [1 - exp ( 7 876° C)l                                 (D-l)
           u                    LSTD
Thus:
CSTD
0.06
0.07
0.08
0.09
0.10
0.11
0.12
In 8760/CSTD
151.30
129.69
113.47
100.87
90.78
82.53
75.65
CSTD
0.13
0.14
0.15
0.16
0.17
0.18

In 8760/C$TD
69.83
64.84
60.52
56.74
53.40
50.433

Substituting a constant from this table into Eq.  (D-l) above gives the
P~(C) function corresponding to each level  of the standard.   Figure D-l
shows plots of PC(C) for various levels of  the standard generated from the
constants in the table.

-------
 c
 o
 3
JO
4J
Q




 

 «

o
o
  CJ
Q.
 O
 C

 3
 OJ

 O!


 re
  1
 Q
 OJ
 rs
 en
tn
                                                                                                  evi
                                                                     Q.

-------
                                    155
D.2  Form the Composite Distribution Function, PJ(C), For the Three Thresholds
    Equation (4-5), Section 4.2 is used to calculate the composite distri-
bution function.  In the example, theree health effects are considered.
Expanding Eq. (4-5) for three thresholds yields:
         PT(0 -  eQQP   + ep   +  eQ                              (D-2)
Note from Eq. (4-6) that:

                        o
           Qi  =    "eiQi                                                  (D-3)

where:
                   .c o
           Q!  =  j p! dC                                                  (D-4)
                  0

Thus, in addition to requiring tabulations of the threshold probability
density functions p? versus concentration, tabulations of the cumulative
distribution is also needed.  In the example all three p? are normal distri-
butions.
    From the constants provided in the small table on the first page of
Section 4.5 one can write:
         Pl°  =  2x0.05 ^  6XP HC-0.15)2 / 2 x (0.05)2)
         p°  =  3.989 exp (-200 (C-0.15)2)                                 (D-5)
The functions for p2 and p° can be developed in the same manner.

-------
                                156
                           TABLE  D-l
INDIVIDUAL AND COMPOSITE  HEALTH EFFECTS  THRESHOLD  PROBABILITY  DENSITY
                     FUNCTIONS VS.  CONCENTRATION
STANDARD
(ppm)
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
0.11
0.12
0.13
0.14
0.15
0.16
0.17
0.18
0.19
0.20
0.21
0.22
0.23
0.24
0.25
0.26
0
0.09
0.16
0.27
0.45
0.71
1.08
1.58
2.22
2.99
3.88
4.84
5.79
6.66
7.37
7.82
7.98
7.82
7.37
6.66
5.79
4.84
3.88
2.99
2.22
1.58
1.08
0.71
o
PT
T2
0.00
0.00
0.00
0.00
0,00
0.01
0.02
0.05
0.11
0.23
0.44
0.79
1.35
2.16
3.24
4.57
6.05
7.53
8.80
9.67
9.97
9.67
8.80
7.53
6.05
4.57
3.24
0
PT3
0.00
0.01
0.01
0.02
0.03
0.04
0.06
0.08
0.11
0.16
0.22
0.30
0.40
0.52
0.67
0.86
1.08
1.33
1.62
1.94
2.28
2.65
3.02
3.40
3.76
4.10
4.40
o;
0.001
0.003
0.005
0.008
0.014
0.023
0.036
0.055
0,081
0.115
0.159
0.212
0.274
0.345
0.421
0.500
0.579
0.655
0.726
0.788
0.841
0.885
0.919
0.945
0.964
0.977
0.986
«;
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.001
0.003
0.006
0.012
0.023
0.040
0.067
0.106
0.159
0.227
0.308
0.401
0.500
0.599
0.692
0.773
0.841
0.894
0.933
«;
0.000
0.000
0.000
0.000
0.001
0.001
0.002
0.002
0.003
0.004
0.006
0.009
0.012
0.017
0.023
0.030
0.040
0.052
0.067
0.100
0.106
0.129
0.159
0.189
0.227
0.264
0.309
PMT
0.01
0.03
0.05
0.09
0.14
0.21
0.31
0.45
0.64
0.92
1.27
1.77
2.37
3.14
4.05
5.07
6.09
6.99
7.64
7.83
7.82
7.30
6.44
5.41
4.31
3.33
2.50
o
P1,3
0.01
0.03
0.05
0.09
0.14
0.20
0.30
0.41
0.56
0.74
0.94
1.15
1.37
1.58
1.77
1.96
2.11
2.25
2.38
2.49
2.64
2.81
2.98
3.19
3.40
3.62
3.82

-------
         157



Table D-l  (CONTINUED)
STANDARD
(ppm) p°
0.27 0.45
0.28 0.27
0.29 0.16
0.30 0.09
0.31 0.05
0.32 0.02
0.33 0.01
0.34 0.01
0.35 0.00
0.36 0.00
0.37 0.00
0.38 0.00
0.39 0.00
0.40 0.00
0.41
0.42
0.43
0.44
0.45
0.46
0.47
0.48
0.49
0.50
0.51
0.52
0.53
0.54
0.55
0.56
0.57
0.58
0.59
0.60
o o
PT PT
'2 '3
2.16 4.65
1.35 4.83
0.79 4.95
0.44 4.99
0.23 4.95
0.11 4.83
0.05 4.65
0.02 4.40
0.01 4.10
0.00 3.76
0.00 3.40
0.00 3.02
0.00 2.65
0.00 2.28
1.94
1.62
1.33
1.08
0.86
0.67
0.52
0.40
0.30
0.22
0.16
0.11
0.09
0.06
0.04
0.03
0.02
0.01
0.01
0.00
0; ,; Q;
0,992 0,956 0.352
0,995 0.977 0.401
0.997 0.988 0.448
0.999 0.994 0.500
1.000 0.997 0.552
0.999 0.599
1.000 0.648
0.691
0.736
0.773
0.811
0.841
0.871

0.900

0.948

0.970

0.983

0.991

0.996

0.998

0.999

1.000



p°
1.89
1.45
1.18
1.02
0.92
0.85
0.80
0.75
0.70
0.64
0.58
0.51
0.45
0.39
0.33
0.28
0.23
0.18
0.15
0.11
0.09
0.07
0.05
0.04
0.03
0.02
0.01
0.01
0.01
0.01
0.00



o
P1,3
4.00
4.13
4.22
4.25
4.21
4.11
3.95
3.74
3.49
3.20
2.89
2.57
2.25
1.94
1.65
1.38
1.13
0.92
0.73
0.57
0.44
0.34
0.26
0.19
0.14
0,09
0.07
0.05
0.03
0.03
0.02
0.01
0.01
0.00

-------
                                     158
     Since the distributions are normal, their integrals as a function of
(C-C/cr) can be obtained from tables generally available in statistical text-
books.  The p. functions and integrals, Q% are tabulated for various values
of C in Table D-l of this appendix.  The composite distribution p-j- and the
semicomposite distribution p, - are also given in the table.  The function
                            i >o
p-r is calculated from the individual p^  as follows:  Take as an example
C = 0.20 ppm.  By substitution of the value at this concentration calcualte
the Q1 (0.20) from Eq. (D-3).

          Q1 (0.20) = 1 - 0.15 x 0.841 = 0.874
          Q2 (0.20) « 1 - 0.80 x 0.500 = 0.600
          Q3 (0.20) = 1 - 1.0 x 0.106  = 0.894

Note the e, are from the table on the first page of Section 4.5.  Subs ti tut-
in these values and the p? (0.20) from the table into Eq. (0-2) gives:

          P°T  (0.20) = 0.15 x 0.6 x 0.894 x 4.84 + 0.8 x 0.874 0.894 x
           9.97 + 1.0 x 0.874 x 0.600 x 2.28  =  7.82
The function p, ~ is calculated in a similar manner.
               i ,j
D.3  Determine the Risk, R, That no True Threshold Will be Exceeded
     The risk is calculated from Eq. (4-2).  Note that the product of
probabilities terms in Eq. (4-4) is zero because e^ = 1 .  For this case,
therefore, one can write:
          R  = 1     Pc (C,CSTD) p° (C)dC.                                  (D-5)

-------
                                       159
Given any potential value of the standard, the appropriate constant can
be calculated for PC as was done in the Table in Step #1.   Substituting
this constant into Eq. (C-l) a value of P  to be calculated for every
                                         L>
value of Py in Table D-l of this appendix.  The corresponding product is
then calculated and numerically integrated over the range of p°.
     For example, for a level of the standard CSTD = 0.14 ppm have:

          PC(C)  =  [1 -exp (-64.84C)]8760
Evaluating this expression at C * 0.20 ppm gives:
          Pr(0.2) *  [1 -exp (-64.84 x 0.2)]8760 = 0.980
That is, there is a 0.980 chance that the hourly average 0.2 ppm will not
be exceeded in a given calendar year.  The corresponding p° was calculated
in Step #3 to be 7.82.  The product is:
          Pr (0.2) p° (0.2) » 7.66
           w        I
This product is tabulated as concentration over the range of p°.  The product
is then numerically integrated over the range of p° using methods such as
Simpsons' rule or trapezoidal integration.  The procedure is repeated for
each level of  the  standard shown in Table 4-4, Section 4.5.
     Figure D-2 shows the procedure graphically for a standard at 0.14 ppm.
Figure D-2(a) shows the PC function.  Figure D-2(b) shows the composite
probability density function p°, and Figure D-2(c) shows their product.

-------
                                    160
      Figure D-2    Graphical  Illustration of PT Calculation
 8

 6

 4

 2

 0
1.0

 .8

 .6

 .4

 .2

  0
                           (a)
                              0.2           0.3
                        Concentration, ppm
                           (b)
                                              0.5
                0.1
    0.2           0.3
Concentration, ppm
0.4
0.5
 88

  6

  4

  2

  0
                    PCPTdC
                              0.2           0.3
                          Concentration, ppm
                                0.4
              0.5

-------
                                     161
The dashed line in Figure D-2(c) sketches in the original composite density
function. The area under the solid curve is PT, the probability of not ex-
ceeding the threshold.  The area between the dashed line and the full line
is the probability of exceeding the true threshold.  The sum of the areas,
is of course, the total area under the composite density function alone.
D.4   Determine R.
     The risk, R.., of exceeding any individual threshold is determined
proceeding in the same manner as described in Step #3.
D.5  Determine any R,- ^
     	* »j
     The risk of exceeding the lower of two thresholds,  R. .  can be
                                                         i »J
calculated by forming the composite in a manner similar to Step #2 and
then proceeding to Step #3.

-------
                                   TECHNICAL REPORT DATA
                            (Please read Instructions on the reverse before completing}
1. REPORT NO.
                                                           3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE

   STUDIES IN THE REVIEW OF  THE  PHOTOCHEMICAL
   OXIDANT STANDARD
                                                           5. REPORT DATE
                January  1979.  Date of Issue
             6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)

   William F. Biller
                                                           8. PERFORMING ORGANIZATION REPORT NO,
9. PERFORMING ORGANIZATION NAME AND ADDRESS

   William F. Biller
   68 Yorktown Road
   East Brunswick, N.J. 08816
                                                           10. PROGRAM ELEMENT NO.
             11. CONTRACT/GRANT NO.

                 68-02-2589
 12. SPONSORING AGENCY NAME AND ADDRESS
   Environmental Protection  Agency
   Research Triangle Park
   North Carolina  27711
                                                           13. TYPE OF REPORT AND PERIOD COVERED
                 Final Report
             14. SPONSORING AGENCY CODE
 15. SUPPLEMENTARY NOTES
16. ABSTRACT
       Studies were performed  in three areas for the  Office of Air Quality Planning
  and Standards to support  its review of the photochemical  oxidants standard:   1.  the
  desirability of changing  from a photochemical oxidants  to an ozone standard,  2.  the
  suitability of alternate  ways of expressing the  standard  and,  3. the assessment of
  health risks at alternate levels of the standard.   From the first study it was
  concluded that a change to an ozone standard could  lead to benefits in administering
  the standard with little  or no loss in protection of  the  public health and welfare.
  The second study concluded that national ambient air  standards should be expressed
  in a statistical form.  A statistical form is more  in accord with the  actual
  behavior of pollutant concentrations in the atmosphere.  In the third area the
  mathematical methodology  was developed for estimating the risk that health effect
  thresholds for sensitive  populations will be exceeded during a given period of  time
  at alternate levels of the ambient air quality standard.   The method quantitatively
  treats the uncertainties  in  health effects thresholds and the extreme concentration
  levels of pollutant to be experienced in the given  time period.
1 7.

j
                               KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS  C. COSATI Field/Group
   Air Pollution
   Oxidant
   Ozone
   Stantdard
   Risk
  DISTRIBUTION STATEMENT

     Release unlimited
19. SECURITY CLASS (This Report)
      NA
                           21. NO. OF PAGES
161
                                              20. SECURITY CLASS (This page)
                                                   NA
                                                                        22. PRICE
EPA Form 2220-1 (9-73)

-------

-------