EPA-650/4-75-018-0
                 EVALUATION
        OF  THE  MULTIPLE  SOURCE
GAUSSIAN  PLUME  DIFFUSION  MODEL  -
                    PHASE I
                        by

             Robert C. Koch and George E. Fisher

                     Geomet, Inc.
                    15 Firstfield Road
               Gaithersburg, Maryland 20760
                 Contract No. 68-02-0281
                   ROAP No. 21 ADO
                Program Element No. 1AA009
             EPA Project Officer: D. Bruce Turner

                 Meteorology Laboratory
            National Environmental Research Center
          Research Triangle Park, North Carolina 27711
                     Prepared for

          U.S. ENVIRONMENTAL PROTECTION AGENCY
           OFFICE OF RESEARCH AND DEVELOPMENT
                WASHINGTON, D. C. 20460

                      April 1973

-------
                        EPA REVIEW NOTICE
I
I
I
This report has been reviewed by the National Environmental Research                 .
Center - Research Triangle Park, Office of Research and Development,                 I
EPA, and approved for publication.  Approval does not signify th^t the                 *
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial                      I
products constitute endorsement or recommendation for use.                           I
                   RESEARCH REPORTING SERIES

Research reports of the Office of Research and Development, U.S. Environ-             I
mental Protection Agency, have been grouped into series. These broad
categories were established to facilitate further development and applica-
tion of environmental technology.  Elimination of traditional grouping was               I
consciously planned to foster technology transfer and maximum interface                I
in related fields.  These series are:

          1.  ENVIRONMENTAL HEALTH EFFECTS RESEARCH                          |

          2.  ENVIRONMENTAL PROTECTION TECHNOLOGY

          3.  ECOLOGICAL RESEARCH                                               ]

          4.  ENVIRONMENTAL MONITORING

          5.  SOCIOECONOMIC ENVIRONMENTAL STUDIES                            I

          6.  SCIENTIFIC AND TECHNICAL ASSESSMENT REPORTS                     '
          9.  MISCELLANEOUS                                                     ,

This report has been assigned to the ENVIRONMENTAL MONITORING
series. This series describes  research conducted to develop new or
improved methods and instrumentation for the identification and quanti-
fication of environmental pollutants at the lowest conceivably significant                ^
concentrations.  It also includes studies to determine the ambient concen-
trations of pollutants in the environment and/or the variance of pollutants                !
as a (unction  of time or meteorological factors.                                         <
This document is available to the public for sale through the National
Technical Information Service, Springfield, Virginia 22161.

                 Publication No. EPA-650/4-75-018-a

-------
   GEOMET Report Number EF-186

           April 1973
EVALUATION OF THE MULTIPLE-SOURCE
  GAUSSIAN PLUME DIFFUSION MODEL
          PHASE I Report
               for

 Environmental Protection Agency
     Research Triangle Park
      North Carolina 27711
              under

   Contract Number 68-02-0281



               by

         Robert C. Koch
        George E. Fisher

-------
 I


 I


 I                               TABLEOFCONTENTS



 I
              1.0   INTRODUCTION                                                      1

 ง            2.0   MODELS                                                            2

 _                 2.1   Features  of  the Multiple-Source  Gaussian  Plume
 •                      Diffusion Model                                              2
 •                 2.2   The  Simplified Gifford-Hanna Model                          18

 •            3.0   MODEL COMPARISONS                                               21

                   3.1   Comparisons  Between Calculations Using Mean and
                        (Variable  Emission  Rates  for SCIM                           21
                   3.2   Comparisons  Among  Ten Model Variations Using
                        New  York  City Data                                         42

 •            4.0   CONCLUSIONS                                                      91

              5.0   REFERENCES                                                       94

 I                 APPENDIX  A - An Algorithm for Estimating Area  Source
                               S02  Emission Rates

 I                 APPENDIX  B - Method of  Estimating the Height of the
 *                             Mixing Layer

                   •APPENDIX  C - Development of a Method  of Estimating Emission
                               Rates for  Use in the Simplified Gifford-Hanna
                               Model

 |                 APPENDIX  D - Frequency  Distributions  for SCIM  Calculations
                               Using 3 to 96 Hour Sampling Intervals


 I


 I


 I


I


I


I

-------
I
                                 INDEX   OF   FIGURES
             Figure                                                              Page
I

I

I

I
                1      Location of St. Louis Observing Stations Used in
m                     Validation Analysis                                        23
                2      Frequency Distribution of Calculated and Measured
                       Two-Hour Concentrations for St. Louis Station #13          25
™              3      Frequency Distribution of Calculated and Measured
                       Two-Hour Concentrations for St. Louis Station #15          25
                4      Frequency Distribution of Calculated and Measured
                       Two-Hour Concentrations for St. Louis Station #17          26
I              5      Frequency Distribution of Calculated and Measured
                       Two-Hour Concentrations for St. Louis Station #23          26
•              6      Frequency Distribution of Calculated and Measured
                       Two-Hour Concentrations for St. Louis Station #33          27

                7      Frequency Distribution of Calculated and Measured
                       Two-Hour Concentrations for St. Louis Station #4           27
I

—              8      Frequency Distribution of Calculated and Measured
•                     Two-Hour Concentrations for St. Louis Station #10          28

                9      Frequency Distribution of Calculated and Measured
•                     Two-Hour Concentrations for St. Louis Station #12          28

               10      Frequency Distribution of Calculated and Measured
m                     Two-Hour Concentrations for St. Louis Station #28          29

               11      Frequency Distribution of Calculated and Measured
_                     Two-Hour Concentrations for St. Louis Station #36          29

"             12      Location of Chicago TAM Stations Used in Validation
                       Analysis                                                   30

|             13      Frequency Distribution of Calculated and Measured
                       One-Hour Concentrations for Chicago TAM #1                 31

•             14      Frequency Distribution of Calculated and Measured
*                     One-Hour Concentrations for Chicago TAM #2                 31

               115      Frequency Distribution of Calculated and Measured
                       One-Hour Concentrations for Chicago TAM #3                 32

_                                                                       (Continued)

•                                            -111-


I

-------
 I
 I
I
I

I
                           INDEXCFFIGURES    (Continued)
 •            Figure

 I

 I

 I

 I
                16       Frequency Distribution of Calculated and Measured
                        One-Hour Concentrations  for Chicago JAM #4                  32

                17       Frequency Distribution of Calculated and Measured
                        One-Hour Concentrations  for Chicago TAM #5                  33

                18       Frequency Distribution of Calculated and Measured
                        One-Hour Concentrations  for Chicago TAM #6                  33

                19       Frequency Distribution of Calculated and Measured
                        One-Hour Concentrations  for Chicago TAM #7                  34

                20       Frequency Distribution of Calculated and Measured
                        One-Hour Concentrations  for Chicago TAM #8                  34

•              21       Locations of Air Quality Measurements in the
™                      Immediate Vicinity of New York City                         45
               22       Frequency Distribution of SCIM Calculated and Measured
                        One-Hour Concentrations for New York City Station #0       51
                123       Frequency  Distribution of  SCIM Calculated and Measured
                        One-Hour Concentrations for New York City Station #1       51

                24       Frequency  Distribution of  SCIM Calculated and Measured
•                      One-Hour Concentrations for New York City Station #3       52
                25       Frequency Distribution of SCIM Calculated and Measured
                        One-Hour Concentrations for New York City Station #10      52
I
               26      Frequency Distribution of SCIM Calculated and Measured
M                     One-Hour Concentrations for New York City Station #14      53

               27      Frequency Distribution of SCIM Calculated and Measured
                       One-Hour Concentrations for New York City Station #17      53

•             28      Frequency Distribution of SCIM Calculated and Measured
                       One-Hour Concentrations for New York City Station #27      54

|             29      Frequency Distribution of SCIM Calculated and Measured
                       One-Hour Concentrations for New York City Station #28      54

•                                                                       (Continued)
                                               -IV-

-------
             INDEX   OF   FIGURES   (Continued)
Figure                                                              Page

  30      Frequency Distribution of SCIM Calculated and Measured
          One-Hour Concentrations for New York City Station #31       55

  31      Frequency Distribution of SCIM Calculated and Measured
          One-Hour Concentrations for New York City Station #36       55

  32      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #0       56

  33      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #1       56

  34      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #3       57

  35      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #10       57

  36      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #14       58

  37      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #17       58

  38      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #27       59

  39      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #28       59

  40      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #31       60

  41      Frequency Distribution of GHM Calculated and Measured
          One-Hour Concentrations for New York City Station #36       60

  42      Variations in Measured and Calculated Mean One-Hour  SOp
          Concentration by Hour of the Day                           67

  43      Variations in Measured and Calculated Mean One-Hour  S02
          Concentrations by Temperature                              69

                                                            (Continued)
                                  -v-

-------
             INDEX   OF   FIGURES   (Continued)
Figure
  44      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #0        72

  45      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #1        72

  46      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #3        73

  47      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #10       73

  48      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #14       74

  49      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #17       74

  50      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #27       75

  51      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #28       75

  52      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #31        76

  53      Frequency Distribution of SCIM Calculated and Measured
          24-Hour Concentrations for New York City Station #36       76

  54      Frequency Distribution of GHM Calculated and  Measured
          24-Hour S02 Concentrations for New York  City  Station #0    77

  55      Frequency Distribution of GHM Calculated and  Measured
          24-Hour S02 Concentrations for New York  City  Station #1     77

  56      The  Frequency Distribution of 24-Hour SOp Concentrations
          for  the Two Sets  of GHM Calculations and Measured
          Values  for New York City Station  #3                        78

  57      The  Frequency Distribution of 24-Hour S02 Concentrations
          for  the Two Sets  of GHM Calculations and Measured
          Values  for New York City Station  #10                       78

                                                            (Continued)

                                  -vi-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
              INDEX   OF   FIGURES   (Concluded)
Figure                                                              Page

  58      The Frequency Distribution of 24-Hour SCL Concentrations
          for the Two Sets of GHM Calculations and Measured
          Values for New York City Station #14                       79

  59      The Frequency Distribution of 24-Hour S02 Concentrations
          for the Two Sets of GHM Calculations and Measured
          Values for New York City Station #17                      ,79

  60      The Frequency Distribution of 24-Hour SOp Concentrations
          for the Two Sets of GHM Calculations and Measured
          Values for New York City Station #27                       80

  61      The Frequency Distribution of 24-Hour S02 Concentrations
          for the Two Sets of GHM Calculations and Measured
          Values for New York City Station #28                       80

  62      The Frequency Distribution of 24-Hour S02 Concentrations
          for the Two Sets of GHM Calculations and Measured
          Values for New York City Station #31                       81

  63      The Frequency Distribution of 24-Hour S02 Concentrations
          for the Two Sets of GHM Calculations and Measured
          Values for New York City Station #36                       81
                                 -vn-

-------
                     INDEX   OF   TABLES


Table

  1       Comparisons of Calculated Mean S0? Concentrations
          Using SCIM with St. Louis and Chicago Data Sets            35

  2       Comparison Statistics for the Distribution of
          Calculated Hourly S02 Concentrations Using SCIM with
          St. Louis and Chicago Data Sets                            37

  3       Comparisons Between Calculated and Measured SCL
          Concentrations Using Mean and Variable Emission
          Rates in the CIGP Model                                     39

  4       Variations in Mean Concentrations for Selected
          Classifications                                            41

  5       Model-to-Measurement Comparisons of 1-Hour NYC S(L
          Concentrations by Mean,  Standard Deviation, RMSE,
          MAE and Error at Maximum Measurement                       61

  6       Model-to-Measurement Comparisons of 1-Hour NYC SCL
          Concentrations by Correlation Coefficient,
          Variance and Regression  Coefficients                       62

  7       Comparisons of 1-Hour S0? Concentrations by Hour
          of the Day and Day of the Week                             65

  8       Comparisons of 1-Hour SCL Concentrations by Temperature,
          Wind Speed and Stability Classes                           66

  9       Model-to-Measurement Comparisons of 24-Hour NYC S(L
          Concentrations by Mean,  Standard Deviation, RMSE,
          MAE and Error at Maximum Measurement                       82

 10       Model-to-Measurement Comparisons of 24-Hour NYC S02
          Concentrations by Correlation Coefficients,
          Variance and Regression  Coefficients                       83

 11       Summary of Mean Error and RMSE Associated with
          Various Sampling Intervals for Proportionate
          Stratified Sampling in SCIM                                86

 12       Model Comparisons for Annual  Mean S0? Concentrations
          Using New York City Data                                   88

                                                            (Continued)


                                -viii-

-------
 I
 I
 I                            INDEXOFTABLES   (Concluded)
 I            Table                                                               Page
 —             13       Model  Comparisons for Annual  Mean Particulate
 •                      Concentrations Using New York City Data                     90
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
I
I
-ix-

-------
Section 1.0



INTRODUCTION

-------
I
I
                                           Section 1 .0
                                           INTRODUCTION
I                     This report describes the work accomplished during Phase  I of
             a research program undertaken by GEOMET, Incorporated for the Environ-
Q           mental  Protection Agency  to  further evaluate and document a Gaussian
M           Plume Urban Diffusion Model.  The analysis  has concentrated on validation
             of a steady-state model which uses sampled  chronological input data.
I           The model was developed for  EPA by GEOMET in previous work  (Contract
             Number  CPA 70-94).
|                     The model has been compared with  three other models (using
_           the same data base).  In  future work it will be modified to improve its
™           computational efficiency  and to make it compatible with the Implementation
•           Planning Program (IPP).   Procedures for preparing emission and meteoro-
             logical data for use in the model will also be enumerated and evaluated.
J           The use of the statistical portion (Larsen  transform) of the Air Quality
_           Display Model (AQDM) and  alternative procedures to estimate short-term
•           maximum concentrations will be examined.  EPA personnel will be trained
•           to use  the model, and a user's manual will  be provided,

I

I

I

I

I

I

-------
Section 2.0



  MODELS

-------
                               Section 2.0
                                 MODELS

          The models studied in this report include three variations of
the multiple-source, Gaussian plume, meteorological diffusion model.  Two
of the models, the Air Quality Display Model (AQDM) and the Climatological
Dispersion Model  (COM), are primarily designed to calculate long-term
mean concentrations.  The third model, developed by GEOMET under previous
EPA sponsorship,  is designed to calculate both the long-term mean concen-
tration and the frequency distribution of short-term concentrations using
selected chronological data.  The frequency distribution is determined by
concentrations calculated for a statistically selected set of short-term
periods.  Representative meteorological characteristics and simulated
time-dependent emission characteristics are determined for each selected
period.  The other two models use mean emission characteristics and a
specified set of  combinations of meteorological characteristics (wind
direction, wind speed and stability).  To determine the long-term mean,
the calculations  for each combination of meteorological conditions are
weighted by the relative frequency of occurrence of the combination.
          In addition to the three Gaussian plume models, a simple model
based on the work of Gifford and Hanna is included.

2.1  FEATURES OF  THE MULTIPLE-SOURCE GAUSSIAN PLUME DIFFUSION MODEL
          The details and assumptions involved in the development and
use of the Gaussian plume model  as applied to multiple point sources has
been described by many authors (e.g., Calder 1970).  They are briefly
outlined here for convenient reference.

                                   -2-

-------
          It is assumed that any long-term period may be represented by
a series of quasi-steady-state periods during each of which the following
"Gaussian plume" equation describes the concentration at a ground level
receptor from an elevated point source.
                  y(x)oz(x)
                            exp
where
           X  = concentration at the receptor (yg/m3)
            Q = pollutant emission rate at the source  (yg/sec)
            u = wind speed (m/sec)
        a (x) = horizontal diffusion parameter (m)
         J
        a (x) = vertical  diffusion parameter (m)
            h = effective height of the source (m)
            y = crosswind distance between source  and  receptor  (m)
            x = downwind  distance from source to  receptor (m).

          The above equation requires that the value of seven variables
be known for each possible pair of source and receptor coordinates.   The
information requirements  are greatly simplified by smoothing the  point
sources into an area source, since the variables  of nearby sources  are
                                  -3-

-------
closely related.  The concentration at a single receptor from all

sources is
          XA = ^7  J     ual^a'l^ 6*P<-  VZK-V-y)  +[-J-TriV 
-------
I

I
                       It is also of interest to determine the maximum concentration
J           which will occur during a single 3-hour or 24-hour period of a year.  This
             problem is not directly approached in this report but will be a subject
•           to be studied in the second phase of this program.
M                     The six features of the multiple-source Gaussian plume diffusion
             model which vary in the model's variations studied in this report are:

™                     •    Method of defining quasi-steady-state conditions and
                            their frequency of occurrence over a long-term period
•                     •    Method of performing the integration operation in
                            Equation (2)
|                     •    Method of defining emission rates applicable to a quasi -
                            steady-state period
•                     •    Method of estimating the effective source height
                       •    Method of estimating a  and a  values
I                     •    Method of estimating wind speed for use in Equations (1)
                            and (2).
I
             2.1.1  The Sampled Chronological  Input Model  (SCIM)
V                     The SCI Model is the more detailed of the three variations.
             The long-term period of interest is defined to be composed of a sequence
9           of equal increment, quasi-steady-state periods of one hour duration.  A
ซ           sampling procedure is used which selects one hour from a specified
             sampling interval and guarantees that all hours of the day are equally
•           represented in the selected sample.  Calculations are made for the
             selected quasi-steady-state periods, each of which is characterized by
J           a specific wind speed, wind direction, atmospheric stability class, and
             whatever parameters are required to characterize the time dependence of
I
I
I
                                                -5-

-------
source emission rates (e.g., temperature, month of the year,  hour of the
day).  The  objective of this procedure is to minimize bias  due to diurnal
and seasonal  variations in the model  inputs, and to allow the long-term
characteristics to be estimated  by a  reduced number of calculations.  All
selected  quasi-steady-state concentrations are weighted  equally in
determining the long-term mean concentration.
          The integration operation required to calculate concentrations
from area sources is performed by  using the "narrow plume"  assumption.
Mathematically, it requires that q(x)  be defined such that
                                                 I           _
           a ixi i      	 --•  i  ,-•- a .v). W = V2* q(x).
             IXJI      .---.   'II     rrlVili"    • —  -i \" / ซ        /_ป
            r)Jy                LI     V ;JJ                       (5)

This will  be  generally true as long  as  the spatial distances between
variations  in  area-source emission rates q(x,y) are large compared to
the horizontal  diffusion parameters  a  (x).  It is assumed that

                               ?(x)=*q(x,o).                          (6)
As a result,  Equation (2) becomes
                                                                       (7)
This equation  may be evaluated by  use  of the trapezoid rule.'
* A slight variation of the trapezoid rule is being used in which small increments in x are gradually
  increased with increasing x to a uniform increment.  This allows the larger variations in the integrand
  which tend to occur with small, rather than large x,  to be more accurately evaluated.
                                    -6-

-------
 I
 I
                        The following equations are used to estimate emission rates
 •            for New York City area sources as a  function of temperature and hour of
 •            the day.
                                 q  - qA  [A,  +  B,  (T,  -  T)]
I
I
I
I
 a
              where
 •                      q  =  area  source  emission  rate  for  ith  hour  of
                            the day  with temperature T (yg/sec/m2)
 |                     qfl  =  annual mean  area  source emission rate
                            (yg/sec/m2)
                 IA. ,B.,T.  =  empirical  parameters  for ith hour  of the
                        1    day
T = temperature (ฐF).
 •                      If  sufficient  data  are  available, point  source emissions may
              be  generated  for  each  hour  using  a  special algorithm.  Otherwise, they
 •            are treated as  constant.  For New York City, all point sources were
              treated  as having a  constant  emission rate.  A description of algorithms
 •            used for estimating  emissions for St. Louis and Chicago was previously
 •            reported by Koch  and Thayer (1971).
                        Plume rise for point sources is estimated using the following
 •            equations adapted from Briggs (1969).  For stable  conditions (Turner-
              Pasquill  stability class 5),
I
                                 ฐ-33                      M
               - 0 Q I 	ฐ    I                           \*)
                 *'y v 0.03034 u
                        -7-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
where
           AH  =  plume  rise  (m)

           T   =  ambient  air temperature  (ฐK)
           a

           u  =  wind speed  at stack  height  (m/sec)
                        F  =  2.45   1  -
 (10)
where
          T  = stack gas  temperature  (ฐK)

          V  = stack gas  exit speed (m/sec)

          D  = stack diameter (m)  ,


For neutral and unstable  conditions (Turner-Pasquill stability classes

1 through 4),
                             -i  7c  c
                        AH  =  J'/5  *"
                                   O.33   0.67

(11)
where
                            /2.16Fฐ-4Hฐ-6, Hl305l

                            I 67.3F0'4      , H > 305/
(12)
                        H = stack height  (m)  .
The effective source height is
                         h  =  H  +  AH.
(13)
                                  -8-

-------
          The integration in Equation (7) is repeated over three effective
area source heights.  For New York City, the area source emission rates
are assigned a height as follows:

                                        IPP Emission Inventory
          Assigned Height (m)            Reported Height (m)
                  15                        10.0 to  20.1
                  30                        25.0 to  39.9
                  70                        50.0 to 100.0

          Estimates of a  and az are based on curves published by
McElroy (1969) for the Turner-Pasquill stability classes.  The curves
were extrapolated to distances less than 100 m using the Pasquill-Gifford
curves as a guide as previously described by Koch and Thayer (1971).  The
a  values are also modified to accommodate the influence of the mixing
layer ceiling as described by Koch and Thayer (1971).
          Wind speeds are determined as a function of effective height h
as follows:
where
          u-, = reported wind speed (m/sec)
          h-, = observing height (m)
           k = wind profile parameter.
                                  -9-

-------
I
*
             For New York City, k was assigned the following values on the basis of
             the Turner-Pasquill stability class:
                       Turner-Pasquill Stability
                                Class
                              1, 2, or 3                        0.1
                                  4                             0.1!
                                  5                             0.2

                       The selection of these values was governed by the results of a
             previous analysis over New York City (Singer et al.  1970).
I
1                                4                             0.15

1

I
             2.1.2  The Climatological Dispersion Model
•                     The Climatological Dispersion Model (COM) was developed as a
•           revised form of a model first proposed by Martin and Tikvart (1968).  The
             model, described in detail by Calder (1971), is similar to the Air Quality
•           Display Model (Section 2.1.3) in the way quasi-steady-states are defined
             to estimate a long-term mean.  There are 576 quasi-steady-state conditions
m           defined by six stability classes, six wind speed classes, and 16 wind
m           direction classes.  Each combination is weighted by its relative frequency
             of occurrence in calculating the long-term mean.
•                     Calder (1971) has shown that the evaluation of Equations (1)
             through (4) can be greatly simplified by applying the narrow plume assump-
•           tion to both point and area sources in a slightly different form than
M           presented above for the SCIM.  For point sources, the assumption implies
             that the computations for the contributions of a point source to a

I

I
f                                             -10-

-------
I
I
I
I
1
1
1
I
I
I
I
I
I
I
1
I
I
I
I
 receptor location need  only  consider wind direction class inputs which

 include the direction from the  source to the receptor.
                                N .
                                E  fjs"vVYpj>
                                                  (15)
 where
             (x )• = mean concentration at a receptor due  to  the  ith
               p     point source
                Q. = source emission rate (yg/sec)
              9— _  —
            I  ~" ft I
                N,  =
                Xi

                Ni
angular width of wind direction  class for which
the N. input combinations  are  defined (radians)

number of equal  wind direction classes

distance from source to  receptor (m)

number of input  combinations  defined  for direction
9. (direction from source  to  receptor)

relative frequency of occurrence of jth combination
of wind speed and stability classes
S(Xi,hi;uj,Pj)  - /f [u.gz(l.,p.)-] exp {-
                              (X.L2XL)
                                                                      (16)
                                   -11-

-------
I

I
             where
                       X, = distance such that <>Z(XL,P.) = ฐ-47Lj
•                     L. = mixing ceiling attributed to jth input combination (m)
                       P. = stability class for jth input combination
                         1J
                       u. = wind speed for jth input combination.
                         J
|           Linear  interpolation is used between XL and 2XL.  Equation (16) permits
_           the  effect  of the mixing layer depth to be treated in a form recommended
"           by Pasquill.  This effect is treated in the SCIM by modifying oz-
•                     For area sources, the  computation for each set of inputs  is
             equivalent  to the SCIM except that q"(x) is defined:
I
                               q(x) = JT /        q(x.e) de                        (17)
                                         ,      i+6k/2


•            where

fl                   q(x,e)  =  area  source emission  rate
                         e.  =  median  wind direction of class containing direction
Aj                            from  source to  receptor.

_            The  integral is  evaluated by the trapezoid rule using an increment of
*            22.5ฐ for  e.   The integrals indicated by Equation (2) are then evaluated
•            using Equation (7) and  the trapezoid  rule using an increment of 100 m for
              X.   The model  was applied with  ek equal to 22.5ฐ corresponding to 16 equal
J            wind direction classes.

1

I

I

-------
I
I
I
                        A uniform effective area  source  height  is  assumed.   Plume  rise
              for point sources is estimated by the Briggs  equations  as  follows:
                                                                 0.5
•                      t    For stable  conditions  (Turner-Pasquill  stability  class  5),

                                          |xf =  1.00323  u(  Ta   }
                                                        \ 36/32 /

ฃ                           where
                                 xf = distance to  final  rise, m
•                                 u = wind  speed at  stack  level, m/sec
                                 IT  = ambient  air  temperature,  ฐK
                                   a
                                 — = vertical  potential temperature  gradient,  ฐK/m
I                                    JO.02, Pasquill  Category E
                                      |0.035,  Pasquill Category F

                             (1)  Distance from source  to receptor < xf,
I                                             ,    1/3  2/3
                                          AH =  -I^-t	*	
                                                     u
                             where  F is  given by Equation  (10)
|                               x = distance  from source  to receptor
V                           (2)  Distance from source  to receptor >_ x.c, use
                                 minimum  of AH, and  AHL,
I
                                                v
                                  AH]  =  2.4       a
                                                       (ao
                                                       *l
                                                       dZ
                                                           1/3
                                                -13-

-------
I
1
                       t    For neutral  and stable conditions  (Turner-Pasquill
m                          stability classes 1, 2, 3, and 4),
I/  , fi  Fl/3  x2/3
f  1-6  F     x -   >  x < 3.
                                                         5 X*
                            where X*=/14F^'F<55
                                       | 34 F^/t3,  F >. 55,

                       The diffusion parameter a  is calculated as  power functions
|           of distance using stability dependent parameters  a, b,  and c.   The
^           functions are based on the Pasquill-Gifford values, but the stability
*           classes are modified to characterize  an urban atmosphere.   The stability
•           classes are defined using the Turner-Pasquill rules as  in  the  SCIM.
             During daytime hours, the assigned stability class is  made one class more
jj           unstable (except the most unstable class);  during nighttime hours, when
—           the most stable class occurs, one class less stable is  assigned.   As a
*           result, only four stability classes are used.  The a  values are  further
•           modified using the virtual  source concept with an assumed  initial  value
             of 30 m for sources with an effective height of 20 m or less.   The initial
g           value is decreased linearly with increasing source height  to zero for
             effective heights which exceed 50 m.


I

I

I

I

-------
           The mixing layer  depth is treated as  a  function of stability.
The neutral  stability class is  subdivided into  day and night subclasses.
The height is determined as follows:

            Stability Class*                      Mixing Layer Depth (L)
          Very unstable (1)                  1.5 Times Mean Afternoon Maximum
          Unstable (2)                      Mean Afternoon Maximum
          Slightly Unstable (3)                Mean Afternoon Maximum
          Day, Neutral (4)                  Mean Afternoon Maximum
          Night, Neutral (4)                 0.5 Times (Mean Afternoon Maximum
                                         plus 100 m)
          Stable (5)                        100 m

  *  Numbers in parenthesis indicate Turner-Pasquill stability class.


           The wind speeds are determined for each  source as a function
of the  effective source height  using a wind profile power law as  in  the
SCIM.

2.1.3   The Air Quality Display  Model
           The Air Quality Display Model (AQDM)  is  based more directly on
the ideas of Martin and Tikvart than the preceding COM.  It was imple-
mented  for EPA as a component of an air quality planning tool known  as
the Implementation Planning Program (IPP) by the TRW Systems Group  (1969).
In this model, the quasi-steady-states used to  determine a long-term
mean are  defined by 480 combinations of 16 wind direction classes, six wind
speed classes, and five stability classes for which the joint frequency
distribution  is known.  A mean  value is assigned for each class and  the
calculations  for each combination are averaged  by  weighting them  by
their relative frequency of occurrence.
                                    -15-

-------
          The concentrations from area sources, Equation (2), are eval
uated by subdividing the area into grid squares (not necessarily of
equal size) and replacing each grid square by a virtual point source.
The virtual point source definition which is employed is:
                                   ,/JTi
                             -
                                      . _ _
                           o ~ 2 tan  (11.25ฐ)

where
          X  = virtual distance (m)
          A1 = area of an area source grid square lying within a
               22.5ฐ sector upwind of the receptor (m2).

The distance from the source to the receptor is assumed to be XQ + X.
which replaces X.. in Equation (15).  However, the az values in Equation (16)
are evaluated using X.. .  Equation (15) is used to evaluate the contribu-
tions from point and area sources.  An additional modification of
Equation (15) is used to associate each source with the two neighboring
wind direction classes whose medians  (e.  and 0k+1) define an interval which
includes the direction (e) from the source to the receptor.  The emission
rate Q.. in Equation (15) is multiplied by
                                  and
                         Vl ' 6k    Vl " ek
for use with wind classes k and k+l, respectively.  For area source grid
squares, Q. is reduced by multiplying by A'/A where A is the area of the
grid square and A1 is defined above.
                                  -16-

-------
          Emission rates are treated as constant.  For point sources, the
plume rise is estimated by a stability dependent modification of Holland's
equation.
            Ah = Y>  [l.5 + 0.00268 PD  -y-^j ]  (1.4 - 0.1 S)          (19)
where
          Ah = plume rise (m)
           V = stack gas exit speed (m/sec)
           D = stack exit diameter (m)
           u = wind speed (m/sec)
           P = atmospheric pressure (mb)
          T  = stack gas exit temperature (ฐK)
          "T = atmospheric temperature (ฐK)
           a
           S = Turner-Pasquill stability class (1  to 5).

Since the area source is represented by a set of points, one for each grid
square, an area source height may be defined for each area grid square.
          The oz values are determined by functions which were fitted to
the Pasquill-Gifford curves as for COM.  The Turner (1964) definitions of
the stability classes are used except that class 4 values are used for
class 5 occurrences.  Mixing layer depths are defined by stability class
as in COM, except that class 4 is not subdivided.   Instead, three-fifths
of the class 4 stability occurrences are treated as having the daytime
class 4 depth and two-fifths as having the nighttime class 4 depth.
                                  -17-

-------
          No vertical  variation in wind speed  is  computed  as with SCIM
and COM.

2.2  THE SIMPLIFIED GIFFORD-HANNA  MODEL
I

I

I

™                     A simple model was selected to compare with  the  three  variations
fl           of the Gaussian plume model.  The simplified version of the  Gifford-Hanna
             model recently described by Hanna (1971) was used for  this purpose.  The
I           model is

I                                           X • C a                                (20)

Q           where

I                     x = concentration at a receptor location (yg/m )
                                                                             2
                       Q = area source strength surrounding  the receptor  (yg/m /sec)
9                     u = wind speed (m/sec)
m                     C = dimension!ess constant.

—           Recently, Gifford (1972) recommended the following values for C:

                       Atmospheric Stability                  C
I                            Stable                         600
"                            Neutral                        200
                             Unstable                         50

             For a long period, a value of 225 was  recommended.  The  constant C is
             related to other parameters in the original  Gifford-Hanna model as follows
                     c=Vrl^raxl    j^                       (2i,
                                 -18-

-------
I
            where
I
I
                      N = number of grid squares (AX on a side) from receptor
                          to edge of area source
                    a,b - parameters of the vertical diffusion function (a )
                          of the Gaussian plume equation
                         -  avb
                      a   =  aX  .
•           Empirical  estimates  of  C  were  reported  by Gifford  and  Hanna  (1973)  based
             on  particulate  and SCL  data  for  44  U.S.  cities.  The estimates were based
                                  I2
             on  long-term  mean concentrations.   For  particulates, the  empirical  con-
•           stant  ranged  from a  low of 57  to a  high of  634.  The mean value was 219.
             The estimates of C reported  for  SCL data varied  from 7 to 218 with  a mean
l|           of  50.  Gifford and  Hanna suggested that the  lower values for SCL result
             from the  large  fraction of S02 emitted  by tall stacks.  The  validation
Q           analysis  in this study  used  C  =  50  for  S02  emissions and  C = 225 for
—           particular emissions, for both long- and short-term concentrations.
™           No  attempt was  made  to  further relate C to  stability conditions.
fl                     Methods of estimating  Q from  emission  inventory data are  dis-
             cussed and compared  in  Appendix  C.  The method used in comparisons  with
j|           other  models  was the following:
•                     1.  Determine annual emissions applicable to 1  km  squares.
                       2.   For each  receptor  location, get the average emission rate
                           for s2  1  km  squares centered on each receptor location,
                           as required  by the criteria below.
I
I
I
I
I

-------
    Pollutant              Criteria (ton/km2/day)(1)            Use Q = Q* where s is


     SO             (a) Q* > 0.701 and Q* > 0.512                   9km
       2                y            3/

                    (b) Q* < 0.701 and Q* < 0. 512                  37km


                    (c) Neither (a) nor (b)                         31 km


   Particulates         (a) O* >0.128 and Q* > 0.082                   3km
                         o            28

                    (b) 0*< 0.128 and Q*< 0.082                  39km
                         5            28

                    (c) Neither (a) nor (b)                         19km


 (1)                                2
    Q*  is the average emission rate from the s  grid squares whose center grid square contains the

    receptor (s must be an odd integer).





It should be noted  that the grid size criteria were developed by corre-



lating the  effective emission rate for different grid  sizes with SOp



measurements which were later used in the  validation study.   This  results



in an  advantage  to  the Gifford-Hanna model  over other  model  results used



in this  study.



          The wind  speed used in this model  is based on  the same input



as that used in  the  other models,  the three-hourly airport observation



measured  at LaGuardia Airport.   However, the wind speed was  assumed



constant  with height.
                                     -20-

-------
   Section 3,0
MODEL COMPARISONS

-------
                              Section 3.0
                           MODEL COMPARISONS

          Calculations using the four models described in Section 2.0
have been compared against each other and against measured values.  Each
model was run in its normal mode.  In addition, certain simplifications
were made by averaging the inputs used for the model.  The model compar-
isons which follow include consideration of 10 different variations of
models and inputs.
          Two model comparison tasks were carried out.  The first task
was to compare calculations for SCIM which were obtained in a preceding
program (Contract No. CPA 70-94, "Validity and Sensitivity of the Gaussian
Plume Urban Diffusion Model") with calculations made using the same model
and data set, except that mean rather than time-dependent emission rates
were used for all sources.  These calculations involved a 3-month data
period for St. Louis and a 1-month data period for Chicago, for sulfur
dioxide emissions.
          The second task was to compare 10 different combinations of
variations in input and model with each other and with measured values
using a 1-year data set for New York City.  The comparisons include
3-hourly, 24-hourly, and annual concentrations of sulfur dioxide emissions,
and annual concentrations of particulate matter emissions.
3.1  COMPARISONS BETWEEN CALCULATIONS USING MEAN AND VARIABLE EMISSION
     RATES FOR SCIM
          Calculations of S02 concentrations made with the SCIM were
previously compared with measurements obtained for a 1-month period in
                                 -21-

-------
Chicago and a 3-month period in St. Louis.  In these calculations emis-
sion rates were varied from hour to hour as a function of production and
other activity factors and temperature-dependent space heating require-
ments.  A new corresponding set of calculations has been made using the
same model inputs for each hour except for the emission rates.  At each
location (St. Louis and Chicago) a new emission rate was computed for
each source by taking the mean of the values attributed to each hour of
the data period.  The previously obtained calculations are referred to
as variable Q calculations; the new calculations, mean Q calculations.
          The St. Louis data set consists of hourly estimates of source
strengths, source height, and meteorological conditions for an 89-day
period (December 1, 1964 to February 28, 1965).  The data includes SCL
emission rates for 1200 squares (5000 ft on a side) and 51 stacks.  Hourly
estimates of wind speed and direction are the vector average of measure-
ments from the three wind measuring stations and the TV tower shown in
Figure 1.  The Pasqui11-Turner stability classification is based on
hourly airport observations reported by Lambert Field (Figure 1).  A
wind profile power law is included based on measurements from three heights
on the TV tower.  The mixing layer depth for each hour is based on hourly
interpolation of the mean layer depth estimated from OOZ and 12Z radio-
sonde soundings for Columbia, Missouri and Peoria, Illinois.  More details
concerning the data set are given in a previous report (Koch and Thayer 1971).
Measured 2-hour S02 concentrations are included for 10 locations (Figure 1).
As a result, two corresponding hourly calculations for each location must
be averaged to compare with the measured concentrations.  The cumulative
                                -22-

-------
                                                         ฎ  Wind Measuring Station



                                                             Sampler Station
Figure 1 .   Location of St. Louis Observing Stations Used in Validation Analysis
                               -23-

-------
frequency distribution of calculated (mean Q and variable Q) and measured
2-hour concentrations of S0ซ for each receptor location are shown in
Figures 2 through 11.
          The Chicago data sets consists of hourly estimates of inputs
similar to the St. Louis data set for the period January 1  to 31, 1967.
However, the SCL emission rates include three estimates for each of
600 squares (1 mile on a side) and one estimate for each of 85 stacks.
The wind speed and direction are the vector average of eight measure-
ments (one at each TAM location in Figure 12).  The Pasqui11-Turner
stability classification is based on hourly airport observations at
Midway Airport (Figure 12).  The wind profile power law is  determined
as a function of the stability classification.  The mixing  layer depth
is derived from radiosonde soundings for Green Bay, Wisconsin and
Peoria, Illinois.  Measured 1-hour SOp concentrations are included for
eight locations shown in Figure 12.  The cumulative frequency distribu-
tion of calculated (mean Q and variable Q) and measured 1-hour concen-
trations of S02 for each receptor location are shown in Figures 13
through 20.
3.1.1  Model-To-Model Comparisons
          Tests were made to determine if the mean and frequency distri-
bution of S02 concentrations calculated for a receptor location are
significantly different between using mean Q and variable Q inputs.
Comparison results for calculated mean values are shown in  Table 1.   The
principal  result to be noted in Table 1  is that at all  of the eight  Chicago
stations and 8 of the 10 St.  Louis stations, the difference in the two means
                                -24-

-------
                                                   T
                                                             ซ'*
                                                                        Key

                                                                    i  Variable
                                                                      Mean
                                                                    i  Measu
                                                                            •etf-
                                           100
1,000
                                                                       10,000
                           Concentration  (iig/m )

Figure 2. Frequency Distribution of Calculated and Measured Two-Hour Concentration* for St. Louis Station #13
                                                   7W
                                           1
                                             {-
   • Variable
   x Mean Q
                                                                    o MeasLred
10                      100
                                                                  1,000
                       10,000
                 ,          Concentration (yg/m  )

Figure 3.  Frequency Distribution of Calculated and Measured Two-Hour Concentratiom for St. Louis Station #15
                                      -25-

-------
HI
3
a-
100
90
80
70
60
50
40
30
20
10
0



100
Qfl
80
70
60

40
30
20
in
0





































































--^^T.













^













^












/
z&-








/
A
1
1 1
W A








W
/l
V










ซf













*T












u* -A V *




Key
ii Varia
( Mean
> Measu












)le
3
-ed












3




1 10 100 1,000
Concentration (vg/m )
Figure 4. Frequency Distribution of Calculated and Measured Two-Hour Concentrations for St. Louis Station #17























































,, -"',












/,
S












7











/
/









f
J.
i /
/ i /
, /








/
r
t' /
C /AJ
//
/








^
'
/,
7











9^
S
K











f
*' S*
~?
















Key
i Varia
( Mean
> Measu












ale
3
red












3


















10,000
...to














10
TOO
                                                                                        1,000
10,000
                                      '         Concentration

                   Figure 5.  Frequency Distribution of Calculated and Measured Two-Hour Concentrations for St. Louis Station #23
                                                            -26-

-------
1
1
Cumulative Frequency of Occurrence (%)
— 'r-oco-l^cncn^Jcoioo
ooooooooooo
•
1
1
1
15 100
,
s 90
80
1
? 60
• 50
40
1
• o 20
i :
i
i
i













__ -^.













^~-













^-*


























/
_, 4











/
//
' /
( )










/,
/











/
f



*<







/
/




s




- -
/
/
tt G
/
^




/





'•?
ฐ
/
/]
/
/
/

;
i






^
/
/'











^
/












i=













jS

















Key
i Varia
) Mean i
( Measu












1e i
ed



























































10 x 100 1,000 10,000
• Concentration (wg/m )
Figure 6. Frequency Distribution of Calculated jnd Muaiurcd Two-Hour Concentrations for St. Loull Station #33













_ 	














^i- —














-^













-^


























/

/^












?












/













//
/
y









f
/!
/'
/
7







/
//
//
rf
/








A
/











$












?












2*v












--, ,ป^Q




Key
.1 .
( Mean
) Moasu














\
-ed







_ป





















































| 10 100 1.000 10,000
Concentration (wg/m )
Hgure 7. frequency Distribution ol Calculated and Mealured 'rwo-41our Concentration! fur St. LouU SUtlou #4
-27-

-------
Cumulative Frequency of Occurrence (%)
— 'roco-^tS'CT^JCoioo
ooooooooooo
I
wm
I
•
Ia? 100
7
u on
1^ 80
o BU
8
u_ 70
o
Iu 60
c
s
cr en
U-
 ฐ
ฃ
* 30
1
5 ?n
110
n




























^ _ —





































































.^a*


























/













/





ฃ





/
/











/
/
/







 Mean (
c Measui






>^-





le I
ed























































1 10 100 1,000 10,000
* Concentration (pg/m }
Figure 8. Frequency Distribution of Calculated aftd Measured Two-Hour Concentrations for St. Louis Station #10

































































































M *



























^-*













^












/
^
; >•











/
^












/
/
/











/
^ ,
';
//
f







i
\
t
i
/
'

*•







//
v
™











\f,
/v
/












'ป
/
f












^
'












^ - ^
if *
>'












\< •




Kej
i Varia
c Mean
i Heasi












ble
P
red












Q




























































^ 1 10 100 1.000 10,000
™ • Concentration (ug/m )
Figure 9. Frequency Diitrlbutlon of Calculated and Measured Two-Hour Concentration! for St. Loull Station (12
I -28-
1

-------
   I
   I
   I
   I
|
I
I
I
I
I
I
I
I
I
I
I
I
                                  90

                                  80

                                  ป

                                  60

                                  50


                                  4ฐ
                                  30

                                  20



                                   0
                                                                                         100
                                                                         Concentration (ug/m )
                         1,000
                        10,000
                                             Figure 1U.  Frequency Distribution of Cakui.ilcJ .uul Measured Two-lluur Concelltrntldiu far St. l.oull Sl.itloil fM
                                                                         •**
                                                                                                            2*
                                Key
                              Var1 a
                              Mean
                             neasu ^ea
                                                               10
100
1,000
10,000
                                                                          Concentration (w9/m )

                                              Figure tl, Frequency Pistribmion of Calcu Met! and Measured Two-Hour Concentrations for St. LouU Station #36
                                                                                     -29-

-------
1
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
                                                                LAKE MICHIGAN
                           Midway Airport    A ,-
Figure  12.  Location of Chicago TAM Stations Used in Validation Analysis
                              -30-
                                                                                    Ind.

-------
I
I
I
I
I
I
I
I
I
I
1
I
f
I
I
I
I
I
                      o
                      *t-
                      0
100

 90

 80

 70

 60


 50


 40

 30

 20


 10

  0
                           100

                            90

                            80



                            "
                            60

                            50


                            40

                            3ฐ

                            20
                                                            /
                                                                                                               Key
                                                                                                             Varlal
                                                                                                             Mean (
                                                                                                             Msia^UI
                                                                                                                    e  (
                                                        10
                                                                                 100
                                                                              1,000
                      10,000
                                                                  Concentration  (pg/m  )

                                        Figure 13.  Frequency Dlitribution of Calculated and Meuured One-Hour Concentration! for Chicago TAM ปl
                                            7
                                                                                       Key
   "VaT
    Mean
    Heasu
                                                         10
                                                        100
,000
10,000
                                                                  Concentration  (vg/m  )

                                        Flgurt! H.  Fn'< nrnry Piftrlbutinn of C.i ru .ited ind Meanired One-Hour Concentration! for Chicago TAM K2
                                                                               -31-

-------
100
90
80
70
60
50
40
30
20
10
0
1
100
90
80
70
60
50
40
30
20
10
0





































































—








































	













i 	
_^—













ซ — -
.^













=S























J
f










/








I
/,
/
///
/J






j/
/
/
4
//
/
/








-L
/
7










1
f/











1












>
/












Bf
- ^
/












/



Key
i Varia
: Mean
> Heasu












i>le
5
red












3































10 100 1,000
Concentration (yg/m)
Figure IS. Frequency Distribution of Calculated and Measured One-Hour Concentrations for Chicago TAM #3













„ *













V*
— *•













*~
=.






















































-













-
.. 	











^,

_ — '
	











--^

^, — •
^•~-













>•




-






























/
/
/
/
f







^
//
//
//,
//
'/
f
/





I
/
7
I
/
/








,>•
7

/
T












/












/













/












f













r/*



Key
<' Varia
; Mean
i Measu












tie
1
•ed












!














































10,00
-j at * G














                  10
                                          100
1,000
10,000
                             Concentration  (yg/ra )
Figure 16. Frequency Distribution of Calculated and Measured One-Hour Concentrations for Chicago TAM 14
                                       -32-

-------
**     ioo

QJ

.i     90

2!

5     80
U
o

i-     70
o


&     60


-------
100

 90


 80


 70


 60




 40


 30


 20


 10

  0
                                              Z
/n
                                                                        Key
                                                                    ii Varia
                                                                      Mean
                                                                                       le
                                                                      Measu
                                                                                      •elT
                 10                        100
                            Concentration  (ug/m )
                                                                           1,000
' Figure 19.  Frequency Distribution of Calculated and Measured One-Hour Concentrations fur Chicago TAM 07
                                                                                                   10,000
100


 90


 80

 70

 60


 50


 40

 30


 20


 10

  0
                                    ,(.
              7
                                                                         Key
                                                                      -VarTa5Tฅ
                                 ( Mean
                                 h-fteast
                                                                            3
                   10
                                                    100
                                                                              1,000
                                                     10,000
                            Concentration  (yg/m)

   Figure 20.  Frequency Distribution of Calculated and Measured Onc*Hour Concentrations for Chicago TAM #8
                                       -34-

-------
 I
 I
 I
 I
 I
 I
 I
 1
 I
 I
 t
1
 I
 I
I
 I
t
 I
1
a
 td
Q

 a
 3

u
t!
 rt
 CO
•*M

I
to
•s
'S
s
D
to
 bo
•B
 CM
O
to
 n
 ซ
 (U
U
"3
U


I
s
o
u
nO
 nt

H





o
•-a

ฃ
3
rt
a.
fi
c3
'rt
OH








CO

a
I
*-H
Ti
0









 M DO a oo
Stfl rt rt rt
CJ CJ O O
S 2 2 S S
u u o u o


in
O
o
o
V




CM
1-H
o
'



CM
^



tx
CM
1






tx
in

o
CO


ซ


o
CO
rt
CJ
"
U


in
8
o
V




in
o
en
i



CM




in
tx
i






00

CO
oo


tx


a
rt
O
•P
O


in
O
O
O
V




s
tx




CM




O
CM
1






CO


-------
is highly significant.  Clearly, the method by which hourly emissions are
estimated has an important impact on the mean estimated for a 1-month
or 3-month period.  It is interesting that at the eight Chicago stations,
the mean Q estimates are all less than the variable Q estimates.   At the
St. Louis stations, the three largest differences are the result  of high
mean Q estimates.  Since these differences are all attributable to the
emission algorithm, it is clear that the algorithm can act either to
increase or decrease the mean calculated value over a long-term period.
The non-linear correlation between changes in emission rates and  changes
in the rate of diffusion through the atmosphere is indicated by these
results.
          Since the means of the two sets of calculations are significantly
different, it is not surprising that the distributions about these means
are different.  The standard deviation of the calculations for each
station is shown in Table 2.  Except for one St.  Louis station, the cor-
responding standard deviations for calculations using the two types of
emission rates are significantly different at the one percent level based
on the F-test statistic.   The distributions were  also compared by means  of
the nonparametric Kolmogorov-Smirnov statistic.  All  the distributions
are significantly different at the five percent level using this  statistic,
and all but two are significantly different at the one percent level.  The
results of the Kolmogorov-Smirnov test are the more conclusive because it
is likely that the distributions are not normal as is assumed in  the F-test.
                                 -36-

-------













ซ* a
^6
rO
rt
ฃ""














0
•43
CO
'43
g ^
1 S
D

ง
1
CO
> -~ '43
O K^
o '"' to
^
'o
A



o-
0)

ro ^
^_ rt
ti 'C
_O rt
Q
1
XI
rt O'
^ S
(U
^




a
o
•43
a
to





•B
rt

^






OOOOOOOOOO OOOOOOOO
OOOOOOOOOO OOOOOOOO












muifqtxTj.oic\j 81 81 Si 8> 81 8> Si
	 	 . i3S3SS2S2
tocotototototot/Jtoto (JUUUUOUU

-37-
























o
60
rt
o
U

O
*+-t
13
O
P.
a
o
E
rt
a
rt
M
'3
3
.
T3
(U
d
S
S
,3
rt
K,
o
X
a)
fci
S
+J
U.
bo
cซ
1
(U
(U
(— t
4J
a
u
o
I
o
a
a
a
S
5
^
•M
rt
u

1
•K

(U
rt
a
o
•43
rt

TJ
t4
•S
a
5
W
^
ro


-------
I

*           3.1.2  Model -To -Measurement Comparji sons
•                      Having determined that tjhere is a significant difference between
             calculations made  using  hourly variations in emission rates and calcula-
•           tions made using mean values, it is of interest to determine whether one
_           gives significantly  better agreement with corresponding measured values.
™           Table 3  shows  some comparison statistics.  For the St. Louis data, the
•           use  of variable emission rates yields a calculated mean closer to the
             measured mean  at 7 of 10 stations.  For the Chicago data, the variable
•           emission rate  calculations are closer at only 3 of 8 stations.  Three
             sets of  paired comparisons statistics are shown in Table 3, including
*•           mean absolute  error  (MAE), root-mean-square error (RMSE), and correlation
•           coefficient.   The  variable emission rate calculations show better agree-
             ment with  measured values based on MAE and RMSE than the mean emission
•           rate calculations  for St. Louis stations.  For Chicago stations the
             reverse  is true.   Using  correlation coefficients, the variable emission
w           rate calculations  show the better agreement with measured values at both
m           locations.  Except for correlation coefficients, the mean emission rate
             calculations are best for Chicago stations and the variable emission
•           rate calculations  are best for St. Louis stations.
                        The  model -to-model comparisons shown in Tables 1 and 2 indicate
•           that the model is  sensitive to hourly variations in emission rates such
m           as are represented by the algorithm which has been employed.  However,
             the  validation results suggest that the algorithm has some, but limited,
ff           skill in representing these hourly variations.  Another possibility is
             that other sources of error are so large that they mask the effects of
|           changes  in emission  rates.  Since the emission algorithm uses only
                                              -38-

-------
I
I
I
I
1
I
I
I
I
I
I
t
I
I
I
1
I
I
1
    "8
 CM Jj

W  tj
•o  „

 งซ
    a
T3  O
S  'g
11
_O  W
"rt  
 tu  -a


 1  ซ

'rt  2

 fi  g1
a  s
r>

1
a *J rt
1 g H
43 .S; H
rt O >
"""* *n
O o
U U O

4J

o
rt
a 8 1
rt B S
U W ฃ
O j3
3 o* o
OS t/j ^
rt
S
O
1 1
-3 S
0 <- >
5 g
a w
rt
HI O
2 n
rt
(U
S
o
tu

4^
rt
•a
rt
g O
s ง
2
1
M
O
o
•13
o
1


%.
i
o



oo
TH
TH
O



TH
CM



*
CM
O
CM

S
ro
T-i





(M
CO



\Q
o\
rH


ro
1
tx

VO
in
TH


ro

|
to

* •*
00 O
T* Ol
^ ro
O O



g R
O rt
o o
1

* *
•* ro
ro >O





CM O
O oo
CM rt

* *
tx -^
CO rt
TH





oo oo



# &
00 rt
rt 00
TH TH



00 •*

tx CM
ro rt
rt CM


in tx
TH TH
CO CO
to to

# *
rt ro
•* Cft
CM ro
O O



00 rt
rt O
O O


X
S5 S




•*
rt in
TH TH

*
CM ro
TH




.ฃ,
CM ro



*
O CM
CTl tQ
TH



*
Tjl CO

O\ ro
00 tx



ro ro
CM ro
•3 "3
3 3
ซ +J
to to

\r ซ
ro in
CJi ro
CM CM
O 0



Rtx
CM
O rt
O O


*
rt 8
CM



^
in CM
O rt
TH ro

* *
m CM
rt O
rt CM





t*"* ^
ro CM


^L ^L
TH tX
^* o
TH CM



ro tx
co to

to to
N. ro
TH CO


^ O
TH
w in
'3 '3
& &

* *
O CM
rf ro
O O



in 3t!
ro O
o o
1

^j.
in ro
Oi m




*
Cl 00
•

*
O
O



8
CM
O


*
CM
CM





00
CO
TH

1






xj*
ro


^
QHj
CO




cj

i



\Q
CO
•3
3
•


tx
VO
0



in
in
o



o
o




jฃ
TH
10


o





y.
tx
CM



tx





ft
CM

CO
ro



TH

a
s
2
o

*
00
TH
0



TH
tx
O
o



o
ro




*
TH
TH

o
a\





#
ro
00


^.
tn





a

TH
•vH


03

a
s
2
u

* *
S R
ro ro
0 O



CM m
CM rt
O 0



80
ro




•* *
ro O
CM •*
CM rt

0 0
CM O
CM CM




•# •*
rt (O
tx CTl


•*
01 in
IX rt
ro ro



ง ฃ

CM CO
TH CM
CO ^"*


ro Tf

a a
s s
S S

•$(•
tx O
IX 00
CM rt
0 O



m 01
in tx
ro rt
o o



o o
in o




•st •$
in tx
tx m


o o
oo in





* *
in tx
•* CM



oo oo
tM in
TH



oo ro

CM CO
IO CM



in 
                                                                                                                                             rt

                                                                                                                                             tu
                                                                                                                                             tJ
                                                                                                                                             a
                                                                                                                                             rt

                                                                                                                                             tu
                                                                                                                                            T3
                                                                                                                                             •  K
 i    a  o
rt    ^  8

1    II
                                                                                                                                                SSH
                                                                                                                                                 o
                                                                                                                                             >,   5
                                                                                                                                             rt  -O
      ฃ  a
      I*  •*

      i  i
      s  1
      4)  ซ
      43  C
      H  <
    CM  ro

-------
temperature, time of day and day of the week as variables, it is possible
to examine the validity of the model with different combinations of these
variables and to determine whether there is any detectable bias.  This
has been done in Table 4 for classes of hour of the day, day of the week
and temperature.  With regard to hour of the day, the St, Louis measure-
ments show two peaks, one in late morning and one in late evening.  The
calculations using variable emission rates also show two peaks, but they
precede the measured peaks by one class.  The amplitude of the calculated
cycles, although small, agrees with the amplitude of the measured cycles.
Calculations using a mean emission rate show a single unlikely peak
occurring during the early morning hours.  In the Chicago data, the
measurements also show two peaks, although the second peak in the early
evening is very weak.  The calculations using variable emission rates
are in phase with the Chicago measurements but the amplitude of the
cycle is greatly magnified.  The single cycle of the calculations using
mean emission rates agrees more closely with the Chicago measurements.
          There does not appear to be any systematic variation in the
measured or calculated class means by day of the week.  However, both
measured and calculated class means show a trend from high to low values
with increasing temperature.  The trend is more pronounced in the Chicago
measurements than in the St. Louis measurements.  The temperature depen-
dence is exaggerated in the calculations based on variable emission rates.
It appears that many sources were incorrectly assumed to have a high
degree of temperature dependency.
                                -40-

-------
Table 4.  Variations in Mean Concentrations for Selected Classifications





Classification
Hours
00-04
04-08
08-12
12-16
16-20
20-24
Day of the Week
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday
Temperature
<15F
15F to 24. 9F
25F to 34. 9F
35F to44.9F
45F to 54. 9F
55F to 64. 9F
>65F





No. of
Cases

1578
1595
1613
1480
1540
1585

1294
1330
1376
1367
1338
1377
1309

806
1795
2748
2099
1359
495
89
Mean 3- Month
Concentration (jig/m )
Over 10 St. Louis
Stations

Measured

145
161
165
128
153
173

154
166
164
141
153
148
155

161
150
154
168
144
139
124
Calculated
Variable Q

148
167
140
123
175
154

141
157
159
148
151
130
173

217
202
161
129
95
73
84
MeanQ

209
180
138
132
185
188

164
171
180
153
176
155
208

178
191
181
163
156
138
132




-KT ,,(
No. ol
Cases

908
896
893
904
897
685

836
850
703
707
716
742
901

738
1000
2365
848
309
180
15
Mean 1-Month
Concentration (/ig/rrr)
Over 8 Chicago
Stations

Measured

79
113
106
92
98
95

87
88
124
102
110
91
82

128
95
89
102
85
73
44
Calculated
Variable Q

152
213
123
103
154
133

123
140
229
140
121
109
160

280
155
122
125
95
46
23
Mean Q

86
101
69
64
84
86

72
76
96
90
72
66
96

93
81
82
72
95
52
36
                             -41-

-------
          In view of the sensitivity shown by the model-to-model  compar-
isons, the variable emission rate is a desirable approach where pollution
emissions have significant time-dependent variations.   It appears that
the emission algorithms used for St. Louis and Chicago are introducing
significant errors, and that data must be sought which can be used to
make more reasonable estimates of the variations in emission rates.

3.2  COMPARISONS AMONG TEN MODEL VARIATIONS USING NEW  YORK CITY DATA
          A validation study of the three Gaussian plume models and  the
simplified Gifford-Hanna model, described in Section 2.0, was carried
out using data applicable to the vicinity of New York  City for 1969.
There are various ways of preparing input for use in these models.   In
particular, the emission rates, stability conditions or  wind speed may
be assumed to be constant throughout the data period or  varied from
hour to hour.  Altogether  10 combinations of models and data condition-
ing were analyzed.  The material which follows describes the combinations,
describes the New York City data, presents the method  of analysis, presents
and discusses the results, and summarizes the findings.

3.2.1  Description of the Ten Model and Data Conditioning Combinations
          For the SCIM, area source emissions and the  meteorological
conditions of atmospheric stability and height of the  mixing layer
(grouped together) are treated either as varying from  hour to hour or
                                -42-

-------
as being constant throughout the data period.   Three combinations of
input data conditioning were analyzed, including:

          •    Area source emission rates, atmospheric stability and
               height of the mixing layer variable
          •    Area source emission rates constant, but atmospheric
               stability and height of the mixing  layer variable
          •    Area source emission rates, atmospheric stability and
               height of the mixing layer constant.

          For the simplified Gifford-Hanna Model  (GHM), area source
emissions and wind speed are treated as both varying from hour to hour
or as both being constant throughout the data  period.  In addition, the
calculated concentration at a receptor due to  point sources (as estimated
by SCIM with variable atmospheric stability and mixing layer height) are
either added or not added to the GHM calculations.  This results in four
variations of this model, including:
          •    Constant area source emission rates and wind speed,
               without point sources
          •    Variable area source emission rates and wind speed,
               without point sources
          •    Constant area source emission rates and wind speed,
               with point sources
          t    Variable area source emission rates and wind speed,
               with point sources.
          Calculations for COM, which treat atmospheric stability and
height of the mixing layer as either both variable or both  constant,
were furnished by Mr. D. B. Turner of the Division of Meteorology,
EPA/NERC/RTP.  Statistical results of model-to-measurement  comparisons
                                 -43-

-------
I
              for  these  calculations  are  included  in  this report for comparison with
V            the  other  models.   Calculations  for  AQDM  (no variations) also were
              furnished  by  Mr. Turner and are  included  for comparison.

              3.2.2   Description  of the Data
ฃ                      Three  types of data for  New York City which were used to test
_.            the  validity  of  the models  are described  here, including air quality
*            measurements,  pollutant emission information and meteorological measure-
A            ments.   The air  quality and meteorological measurements were made during
              1969, and  the emission  information is applicable to  1969.

              3.2.2.1  Air  Quality Data
•                      The  air quality measurements  consist of annual mean concentra-
              tions of SO-,  and particulate aerosols and hourly concentrations of S00.
                        12                                                         2
              The  annual mean  concentrations were  measured at 127  locations in the
ฃ            New  York - New Jersey - Connecticut  Air Quality Control Region,  S02 was
              measured at 75 locations and particulate  aerosols were measured at
If            114  locations.   Hourly  mean concentrations of SOg were measured at
              39 locations  in  New York City.   The  10  locations shown in Figure 21,
ฃ            representing  different  geographic  characteristics, were selected for use
_            in this  analysis.  All  127  of the  locations measuring annual concentra-
*            tions were used.  The locations  in the  immediate vicinity of the New York
•            City limits are  shown in Figure  21.

m            3.2.2.2  Emission Data
                        The  emission  information consists of:  (1) emission inventory
•            data prepared  for the Implementation Planning Program (IPP) for the
•                                            .44,

1

-------
I
I
t
I
I
I
I
I
I
I
I
I
t
I
I
I
I
I
I
                                                                  Annual Concentrations Only
                                                                              4-Hour SO  and Annual

                                                                                     Concentrations
                                                                                                     4530
                                                                                                     4520
                                                                                                     4510
                                                                                                     4500
                                                                                                     4490
                                                                                                     4480
      570                     580                    590                     600

Figure 21.  Locations of Air Quality Measurements in the Immediate Vicinity of New York City
                                       -45-

-------
New Jersey - New York - Connecticut AQCR, and (2) time and temperature
dependent weighting factors which when multiplied by the annual  emission
rate of area sources provide an estimate applicable to a particular hour.
          The IPP data includes 675 point sources and 871 area sources.
For each point the available information includes Transverse Mercator
Projection Coordinates (TMPC), Standard Industrial Code (SIC), annual
S02 and particulate emission rates, stack height, stack diameter, stack
gas exit velocity and stack gas exit temperature.  Each area source is
                                                                2
defined for a specified square area varying between 1 and 100 km .   The
available information includes the TMPC of the southwest corner, the
size of the area, the annual S02 and particulate emission rates, and
the effective height of the emissions.
          The weighting factor used to determine area source emissions
for a particular hour is a function of temperature and time of day; its
development is described in Appendix A.  The area source emission rate
algorithm is

                              ซ1 = QA Ki
where
          Q. = emission rate for the ith hour of the year (yg/irfysec)
          Kn- = [(l-Fh) J.L1 + Fh (T.-T.) I.KL2] for T. < T.
                                                                       (22)
          Kn- = [(l-Fh) J.L^  for T. >. T.
                                                          p
          0. = mean annual area source emission rate (yg/m /sec)
          F,  = 0.29 = fraction of emissions attributed to space  heating
                      requirements
                                 -46-

-------
I
                        Ji  = base level  emission coefficient for hour i  (day/hr)
•                      TJ  = threshold temperature for hour i  (ฐF)
                        T.  = observed temperature for hour i (ฐF)
I                      Ii  = temperature sensitivity factor for  hour i  (Index/ฐF)
•                       K  = constant H  0.0002337 yr/Index hr
                        LI  = unit conversion = 24 hr/day
I                      L2  = unit conversion = 8760 hr/yr.

•            The  numerical  values of the terms J., T.  and I.  are given  in  Appendix A.
              A  value  of K.  was  computed for each 3-hour temperature observation
I            obtained for  1969.

|            3.2.2.3   Meteorological Data
_                      The meteorological measurements consist  of  3-hourly synoptic
™            weather  observations reported for LaGuardia Airport and twice daily
A            radiosonde observations reported for Kennedy Airport,   The observations
              from the 3-hourly weather  report for LaGuardia Airport which  were used
|            in this  study were:

I                      •    Wind direction (nearest 10ฐ azimuth)
                        0    Wind speed  (observing height of 6.1 m)
|                      •    Temperature
ป.                      t    Cloud cover (tenths)
                        •    Height of cloud ceiling.

•
m

I

I
              The first observation period of each day was 0100 EST.   Twelve hours for
              which the reported wind speed was calm or less than 1 m/sec were excluded

-------
from the data set.  The validation was limited to the use of 3-hourly
observations since this is what is routinely available from NOAA's
National Weather Records Center in Ashville.  As a result, the validation
findings indicate what can be expected from routine as opposed to research
use of the models tested.
          The radiosonde observations were used to estimate the height
of the surface-based mixing layer.  The details of the parcel method
used to define the height are described in Appendix B.  The height
derived for the morning sounding was assumed to be applicable to the
hours of 0100, 0400 and 0700.  The height derived for the afternoon
sounding was assumed to be applicable to the hours of 1300, 1600, 1900,
and 2200.  An average of the morning and afternoon values was used for
1000.  If a sounding was missing, the height assigned to the previous
24-hour period was used.  An annual mean mixing height was determined
by averaging all the 3-hourly values.  Heights in excess of 5000 m were
treated as being 5000 m.  The resulting annual mean was 979 m which is
about midway between the annual morning and afternoon mixing layer
heights recently reported by Holzworth (1972) for New York City.
          Two other annual mean meteorological values which were used
in this study were 5.1852 m/sec for wind speed and neutral (Pasquill-
Turner class 4) for atmospheric stability.
3.2.3  Comparison Statistics
          Model-to-measurement comparison statistics were obtained for
each of the 10 model and data conditioning combinations.  The statistics
are listed and compared in the following sections.  All the statistics
                                -48-

-------
I


              are standard and require no special explanation.   Error as used in these
•            statistics refers to calculated minus measured concentration.

•            3.2.4  Results of Model -toJIeasurement Comparisons

                        The ability of all four models to estimate long-term concentra-

|            tions accurately is evaluated.   Only the SCIM and GHM are appropriate

_            for testing for short-term concentration estimates.   The short-term

'            concentration calculations, which are compared with  measured hourly mean
•


m
concentrations of S0pป are presented first.


3.2.4.1  One-Hour SOz Comparisons for SCIM and GHM

          Using SCIM and GHM, the SOp concentration was calculated at

each of 10 receptor locations for every third hour of the year,  for which

a measured concentration was available, beginning 0100 on January 1, 1969.

For each hour selected, SCIM calculations were made in three ways:


          •    With variable area source emissions (using Equation (4)
               for Ki) and variable stability and mixing layer height
               (variable Q, S, and H)

          t    With mean area source emissions (K-j=l) and variable
               stability and mixing layer height [mean Q, variable S and H)

          t    With mean area source emissions, stability (Pasqui11-Turner
               class 4) and mixing layer height (979 m) (mean Q,  S,  and H).


          The GHM calculations were made in  two ways:


          •    With variable area source emissions and wind  speed,  and
               without the inclusion of any  contributions from point
               sources

          •    With variable area source emissions and wind  speed,  and
               with point source contributions as calculated by  SCIM
               added to the basic GHM calculations.
                                              -49-

-------
          The frequency distribution of 1-hour concentrations for each of
the three types of SCIM calculations and for the measured values are shown
on log-normal graphs for each of the 10 stations in Figures 22 through 31.
The corresponding frequency distributions for the two types of GHM calcula-
tions and for measured values are similarly plotted in Figures 32-41.  The
curves are close to straight lines; therefore, it is reasonable to expect
that they may be approximated by log-normal distribution functions.  This
hypothesis will be tested during Phase II.  The GHM graphs clearly show that
the model calculations with point source concentrations added give better
agreement with the frequency distribution of measured concentrations at 8 of
the 10 New York City stations.  Of the other two (Stations #0 and #1), only
Station #0 shows better agreement with calculations without point source con-
centrations.  The SCIM frequency distributions based on variable S and H
differ very little and agree best with the measured frequency distributions
at Stations #14, 17, 27, 28, and 31.  Only at Station #0 does the SCIM fre-
quency distribution based on mean Q, S and H agree best with measured values.
          A summary of model-to-measurement comparison statistics is given
in Tables 5 and 6.  Table 5 includes a comparison of the means and standard
deviations of measured and calculated concentrations for each station.  The
root-mean-square errors (RMSE), the mean absolute errors (MAE), and the error
(calculated minus observed) concentration corresponding to the measured
maximum concentration are also included for each set of model calculations.
Table 6 shows correlation coefficients and linear regression characteristics
for model-to-measurement comparisons.  The tabulated regression characteristics
of measured on calculated concentrations include the variance attributed to
the regression relationship and the two slope and intercept regression
coefficients.
                                  -50-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
























































































X














'














X










r/^
/
//
//
/
T






/
m
ff
T









/
0










/
/
j











/
r

i
/

//














— ff
^
4






I ft
7
r
/





Key









- 8 Observed
• Variable Q, S,
x Mean Q, :>, H




t



lean Q, Va



riabl












H
e H











S





























































RS
80
70
60
•in
^n
in
?n
is
10
5
10
                               100
                                                            1,000
10,000
                                         Concentration
 Figure 22. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations for New York City Station #0
                                                    i—|-~r-iX2>rr
                                                                           Key
                                                                8  Observed
                                                                t  Variable Q, S,  H
                                                                x  Mean  Q, S, H
                                                                +  Mean  0. Variab  e H. S
     99


     98



     95


     90


     85

     80


     70


     60

     50

     40


     30


     20

     15

     10
10                             100               •           1,000                         10,000


                                           Concentration  (yg/m3)


  Figure 23.  Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations lor New York City Station #1
                                                  -SI-

-------
                                                             t77
      99

      98


      95


      90


      85

      80


      70


      60

      50

      40

      30


      20

      15

      10

                                                                  Key
                                                          6  Observed
                                                          •  Variable Q, S, H
                                                          x  Mean  Q,  S, H
                                                          +  Mean  Q,  Variable H,  S
10
                             100               •          1,000

                                        Concentration  (ug/mj)
10,000
 Figure 24.  Frequency Distribution of SC1M Calculated and Measured One-Hour Concentrations for New York City Station #3
     99

     98


     95


     90

     85

     80


     70


     60    re

     50    ซ
           o
     40    ^
           n
     30    S

           3
     20    ?>
           fD
     15    ^
           **
     10  -
                                                  //
                                                 ??
                                                 /

                                                                 Key
                                                          6  Observed
                                                          •  Variable  Q,  S, H
                                                          x  Mean Q, S, H
                                                             Mean Q, Variable H, S
10
                             100               •          1,000

                                        Concentration  (pg/m3)
10,000
 Figure 25.  Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations for New York City Station #10
                                                -52-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
 I
 I
 I
 I
 I
 I
                                 I
                                                         Pr
                                                                   Key
                                                            ซ  Observed
                                                            •  Variable Q.  S.  H
                                                            x  Mean  q,  S, H
                                                               Mean q, Variable H,  S
                                                                    99

                                                                    98



                                                                    95


                                                                    90    g


                                                                    85    I
                                                                          OJ
                                                                    so    ฃ
                                                                          o5

                                                                    70    ^


                                                                    60    S
                                                                    50

                                                                    40


                                                                    30
                                                                                                      o
                                                                                                      -t>
                                                                                                      o
                                                                    20    =
                                                                          n
                                                                    15    „
                                                                          w
                                                                    10    "
10                            100               •          1,000                         10,000

                                          Concentration (tjg/m3)


 Figure 26.  Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations for New York City Station #14
                ty
7
                                                   f- --
                             fr
r
                                                                 Key
                                                         8  Observed
                                                         •  Variable O.S,  H
                                                         x  Mean q, s; H
                                                            Mean q, Variable H, S
99

98


95
                                                                    50
                                                                    30


                                                                    20

                                                                    15

                                                                    10
                                                                                                     j
                                                                                                     I
                                                                                                     4
                                                                                                     o
10                             100            '             1,000                         10,000

                                          Concentration (\ig/m3)

 Figure 27. Frequency Distribution of SCIM Calculated und Measured One-Hour Concentratlonl for New York City Station 117
                                                  -53-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
 I
 I
 I
                                                               Key
                                                       8   Observed
                                                       i   Variable q,  S. H
                                                       x   Mean  Q, S, H
                                                           Mean q, Variable H,  S
                                              99


                                              98


                                              95


                                              90


                                              85

                                              80
                                                                                                     o
                                                                                                70   31
                                              50

                                              40


                                              30


                                              20

                                              15

                                              10
                                                                                                     o
                                                                                                     -h

                                                                                                     O
                                                                                                     i
10                            100              •           1,000                         10,000

                                        Concentration  (pg/m?)


Figure 28. Frequency Distrlbutlซn of SCIM Calculated and Measured One-Hour Concentration! tor New York City Station 027
                                             -*~T
3
                                                             Key
                                                       0  Observed
                                                       t  Variable q, S,  H
                                                       x  Mean q.  S. II
                                                          Mean 0.  Variable H. S
99


98



95



90    o


85    ^

80    2
      <
      n

70    3
      (D

60    I

50    Q
      O
40    •"
      o
      n
30    9
 10                            100               '           1,000                         10,000

                                           Concentration (pg/m3)

  Figure 29. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentration! fcr New York City Station ป28
                                                  -54-

-------
I
I
I
I
I
I
I
I
I
I
I
 I
 I
 I
 I
 I
 I
 I
 I





























T











/,
y













^












/
\'
— T


/





/
/
7x
L/7
/ \. fi.
V /
t
9*
I





/ f
J
Y
I/
y1









ni
/

^/^










r

//
~fy











//
'/

























Key
' Observed
• Variable C
x Mean Q, S
<• Mean Q. Vc


!.HS,
riabl


H
e H











S


















	 95
90 *"*
|
rt-
on -•*
5
_. 70 J
i*
- - 50 **
o
n
in c
- - JU ^
3
-.20 o
- 15 ~ป
_. 10 -
-. 5

10 100 • 1,000 10,000
Concentration wg/m3
FiRiire 30. Frequency Distribution of SCIM Calculated and Meaiured One-Hour Concentration! tor New York City Station *31










^














^
J

.
I










/^
>*

X











^

/










/
'/

/




1





' J[

I
/







/j
•t
J;











y
f
/











//
/
^
F^






e o
• V
x N
t M



//
t
















Key
bserved
ariable Q, S
ean Q, S, II
ean Q, Vartat


. H
>1e H











S

































95
90
• 7W o
85 1

3
- 70 -n
fin c
- - 50 ฃ
0
_ - 40 -*
o
o
20 1
15 S
ปซ
-.10 ~"
- . 5
.. 2
10
                                                                                            10,000
                               100                •          1,000

                                          Concentration  (yg/m3)


Figure 31. Frequency Diminution of SCIM Calculated and Meaiured One-Hour Concentration! for New York City Station #36
                                                    -55-

-------
























































































^

I









- 85 —
- 80 ~
ฃ
- 70 ?
3
_Q
--50 1
o
- 40 -
n
_.30 S
T
rt>
-.20 S
S
_. 15 -~
ซ
10
-- 5
. •>
10 100 1,000 10,000
Concentration (ug/m3)
Figure 32. Frequency Distribution of GHM Calculated and Measured One-Hour Concentrations for New York City Station #0
•









































































y













^
^^
X/







J.


If
7
in
7
i





//
i
if










L
f/\
/












j






















^
'/








i
+
•


V









Ob
Mi
Mi











Key
served
th Point S
thout Poin












ource
t Sou






























s
rces











95
90
85 ง
Ol
3
70 -n
3
A
--60 =
3
so C
o
AO "**
o
o
-. 30 S
3
20 3
S
15
ซ
10 *"*
5
_. 2
10
                                100
                                                               1,000
                                                                                               10,000
                                            Concentration




Figure 33.  Frequency Dlitrlbutlo^ of CUM Calculated and Measured One-Hour Concentration* for New York City Station 41
                                                   -56-

-------
                                                                       Key
                                                           6  Observed
                                                           +  With  Point Sources
                                                           •  Without Point  Sources
                                     99

                                     98


                                     95


                                     90

                                     85

                                     80


                                     70


                                     60

                                     50

                                     40

                                     30


                                     20

                                     15

                                     10
 10
                               100
                                                            1,000
                               10,000
                                           Concentration  (ug/m3)


 Figure 34. Frequency Distribution of CUM Calculated and Measured One-Hour Concentrations for New York City Station ป3
                                     y.
                                                                     Key
                                                         6   Observed
                                                         +   With Point Sources
                                                         •   Without Point  Sources
                                                                                                99
                                    98



                                    95


                                    90


                                    85

                                    80


                                    70


                                    60

                                    50

                                    40


                                    30


                                    20

                                    15

                                    10
                                                                                                      ฃ>
                                                                                                      C
                                                                                                      ro

                                                                                                      I

                                                                                                      o
10
                              100
1,000
                                                                                          10,000
                                         Concentration  (pg/m3)


 Figure 35.  Frequency Distribution of CUM Calculated and Measured One-Hour Concentration! for New Yolk City Station #10
                                                 -57-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
 I
 I
 I
 I
 I
                                                    b*
                                                                       Key
                                                           8  Observed
                                                           +  With Point Sources
                                                           •  Without Point Sources
                                                                                                  99


                                                                                                  98



                                                                                                  95


                                                                                                  90


                                                                                                  85

                                                                                                  80


                                                                                                  70


                                                                                                  60

                                                                                                  50

                                                                                                  40


                                                                                                  30


                                                                                                  20

                                                                                                  15

                                                                                                  10
10
                              100               '            1.000

                                           Concentration (ug/m3)
                                                                                            10,000
Figure 36. Frequency Distribution of CHM Calculated and Measured One-Hour ConcentradoDj for New York City Station #14
                              100
                                                 '            1,000

                                           Concentration  (pg/m3
                                                                                           10,000
Figure 37.  Frequency Distribution of CHM Calculated and Measured One-Hour Concentrations for New York City Station #17
                                                 -58-

-------
1
"
1
1
1
1
1
1
]
1

1
1
1
1
1
1
1
1
1
1











































X


























/
//
A
n






/

//
///
///
iff
' if
/





L
V
n

p








I
//
V












/
/
r













f








9
*













Key
Observed
With Point
Without Po





















Sour
int S


ces
ourc












es

















99

- 95
90
85
- on
- 70

- 50

30
20
15
- 10
-- 5
•>
10 100 ' 1,000 10,000
Concentration (ug/m3)
Figure 38. Frequency Distribution of CHM Calculated and Measured One-Hour Concentrations for New York City Station #27
- ^ A OU



























/jr<
1











/
/
/
/









t
1
4









1
/
1 i
t
f
y




t
/
J
' 1 1
' j j
^ / /
y (J

f
/





7
1 1
/
r











/
'













































































Key
8 Observed
+ With Point Sources
• Without Point Sources



























- • 95
- 90
- - 85

- . 70
- . fin
- 50
40
- 30
_. 20
-. 15
_. 10
5
. . J
o 100 • 1,000 10,000
Concentration (yg/m3)
Figure 39. Frequency Distribution of CHM Calculated and Measured One-Hour Concentrations for New York City Station ป28
-59-
   o
o
-*

o

-------
1
I-L
1
1
1
1
1
1
1
I_

1






I
•



1






















































(
J
f
1 /
<{'



/
/
/
/ /
//
' //
/ //
//
/ f
f f
f
./ .
/

/
/
*

/
//

r








r
Wt

?











i
/













7









9
+
•








































Key
Observed
With Point
Without Po


Soure
nt Sc


es
>urc



es



ซ. 99

	 95
-.90 r>
- 85 ฃ•
a>
--80 2
S
--70 ?
n
J3
- - fin S
--50 ซ
0
-.40 2
o
n
-.30 c
2
-.20 o
- . 15 "•
ซ
-.10 ""
	 5
	 1
LO 100 ' 1,000 10,000
Concentration (yg/m3)
Figure 40. Frequency Distribution of CHM Calculated and Measured One-Hour Concentration! for New York City Station *31
•
i/v on











/
J
//
1 \








/
1
1,
J
/ซ•








/
f

/
/

X1




/
A
/



/
^ .
/

/
/
^ _
xT



/
1
1
f
}
/
//
/" f
/ 1
' ฎ

/
ri/







//
/
c/
™










I
//
r












7
























ซ Ob
+ W1
• Wi











Key
served
th Point Sot
thout Point











irces
Sourc











.es

















- 95
- 90 <•>
85 -•
- • 03 Q,
80 2
•80 s
- 70 ?
J
- 50 5
o
-ti
- 40 o
S
- 30 5
- 20 2
- 15 C1
. 10
5

10 100 • 1,000 10,000
• Concentration pg/m3)
Figure 41 . Frequency Dlrtilbudon of Cl IM Calculated and McniuKcl One-Hour Conccntntloiil br Niw Yolk City Station 136
• -60-
1

-------
 a ซ
 o c

 '•ฐ
 O

 U
O
=F  ง

^  w
M-l
 o  t5

 a  S

 g  w
u
s PS
   -

ง!

si
rA T3
 ™ -5 i?
1 c? o- ซ 1
1 g S o
C w Co r* t-j
^ 2 QJ fl] T3 ^3
^ *|5 5- "5J *"> ''j!
| D u u a 3
2 oo oo oo O O
(cui/2r/) ui3aj^
in TH a\ (Ti co CM
O CM CO TH CM O

00 IX >H ID O TH
o a* oo & vo CM

oo o rji tn TH TJI
oo oo oo tn T^i m

CTl in CM tx TH TH
CM in in  oo CTI

CTI in o^ tx ^* co
TH fo CM 00 00 O

tx CO CO CO 00 VD
CM CM Tf tx VO tx

CM 00 tx CO CO T*
CO CM CM TH CM CM
CM 00 00 CO CO TH
O^ oo O VO rx oo

tx in CM oo m o
in O TH Tj< CM Tjl
O TH TH O O C\
CM CM CM tx O O
CM Tf Tl< TH CM CM
*^L
t>> ,• ffi jT
(> > Vi 'g "Q
Q) ^ •ป A* ri
"B > S S 'S 'S
! o o o ss
A 00 00 00 O O
(gui/STf)
TH ' T$I in TJ* TJ<

M m vo TJ in

oo tf> oo &i tf
00 C\ CM O O

TH oo m T* ix
m vo vo TJI m

.N rv) tn  Tj< CM CM CM
~ ฃ"
O > oo -g g"
ป • • Bซ B
2 2 2 2 S
D)-H HH LlJ JZ!
U U S K
oo oo oo (J O
( /ปttt /Sn)
JOJkfrT 3Jl?lA^V
tX VO O TH T*
CO T^l iH TH O

o^ vo o oo 01
in vo TH o o\

CM oo vo at m
TH TH C> tx tx

VO ffl 00 CO tx
^* in TH o fJi

CM vo in co oo
in in CM o o\

VO O tx tx OO
Tjl VO Tjl fO TH

VO IX IX tx CO
in tx in o TH
CM CM CM CM CM
00 T* 00 tx VO
CM in vo vo vo
CM N TH TH TH

CM O\ CM TH TJI
TH CM TH TH TH
co a\ oo T* o
CO 00 00 I/) VO
CM CM TH TH TH
^ > "if
m ~ • o. a
3 o o g 'g
lillf
"5! "5 ^| M ป^i
0 D D i i
oo oo oo O O
(pUi/3T/) 10113
s
o\
CM
tx
O
VO
CO
oo
oo
tx
TH
ซ
TH
in
CO
TH
TH
tx
VO
CO
CM
tx

O
in
CO
TH
CM
CM
mm Measured (ug/m3)
ฃ
I

TH VO VO O Tj.
T* o\ o in TH
i i i i i
CO CM >H O\ TH
1 1 1 TH |
tx tx CM tx tx
CM CM CO 00 00
1 1 1 1 1
Ix TJI (M VO CM
vo co m vo o
1
in CM co o m
Tj" in in co CM
i i i i i
SOO OO T* CM
tX TH T* CM
1 | TH TH TH
CM CM TH ft 00
STH TH CM in
in in TJI jo
VO VO VO VO VO
1 1 1 1 1
CO O^ O) O co
vo vo tx 5> tx
^J* ^J* I/) ^ T^(
1 1 1 1 1
TH 00 O ** CM
in o "* co 05
co in cji TH oo
1 1 1 TH |
1
•* oo vo co co
co in ^ tx CM
oป CM in co TH
I TH TH TH
1 1 1
C? t> IA 'g g~
2 9 S Q
1 3 | | g
D D D a a
on oo oo O O
pamstfaj/v umot
-prej/\[ JE 10113
                                                -61-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
  .
u
  .

B
2 fi
C rt
rt *r^

a >
o
U j-T
S ฃ
II
3 J2
4, g

Z











*o

TH

00

tx

tx

•*

O

CO

i— 1

O




K

CO
GO
TH
T-H
00
(f)
T— 1
(M
o
TH
CO
^O
T-H
^
T-1
vo
T-l
R
T— 1
Ol
00
TH
CO
i— I
rH
CM
a
(M
iber of Comparisons
tt
3

vo Oi oo co co
CM TH TH CO TH
O O O O O
TH 00 Cl TH T^
CM TH TH CM CM
O O 0 0 O
tx CM in a\ rj<
TH TH O TH TH
0 O 0 0 0
01 -* in O co >O TH
CM T-I t-i TH CM
O O 0 0 O
o o o •* vo
ro CM CM ro co
O O O O O
CO O co VO ซ3
T}< CM i-l Tf ^J<
0 0 O 0 0
S"
K. ">~~
^ J X •g'
- ^ - s _
0 > w -S K
a) ซ ~ & o
3 o o- 3 -5
oi n a ง &
•fl rt rt J3 J!
I ^ ^ S s
b E s $ $
2 2 S S S
1 — 1 ป— 1 1 — 1 [^ Lj
U u o K X
w oo oo U O
juapi^aoo
UOPB-[3JJCO
8 S 8 3 S
o o o o o
s s s s ง
o o o o o
CO CM O •* CM
O 0 O O 0
0 O 0 0 0
00 CM CM CTi CTl
O O O O O
O 0 0 0 0
SCM CM 
O > w -g y
o ^ . a, a
3 o o- 3 -g
rt fl fl g &
•fi rt rt .s rj
J3  ฃ 2 3 3
ง ง ง s s
D u u K x
oo t/j oo O U
cioissaj3a>j jB3ur[
Xq aowsuEA
jo uoponpa-^j
CM on -H CO CO
o o o o o
CO K O CO TH
CM T-I uo in vo
0 O O 0 O
o Tji 10 oo in
CM O O CO CO
O O 0 0 O
1-1 oo in oo CM
TH O TH TH CM
o o o o o
CM 00 T-I CM TH
TH O CM Tj< Tj<
O O O O 0
CM O tx O CTv
CM TH TH in ^f
o o o o o
^K
E>~
oo" 4 ffi J-
M rt •ป fl
o > w -5 e"
5, . . 6, s
3 o o- g •ฃ
|| I 5 il
> ^ E ^ ^
SUss
D u u pc ffi
00 00 00 O O
aui-j uoissai8a-a
jo ado^s
in T-t c^ fj, co
CM CO CM 00 CO

TH CM TH T-i -^
K K 10 in m

O\ T-I co O fx
TH (M CM O O

tx CM CO 00 CTi
in K 10 CM TH

O U5 Tj< 00 TH
tX K CX Tjl I^H

TH 00 CM CO O
tri (Ti oo tx in

O K tx Oi TH
VO O O tx tx

rf tx G^ tx tx
m 10 m -^t1 CM

TH TH CJ\ O O
oo 01 rx co CM

o oo in oo co
TM m in •* co

ฃ
^"'~~
w" ,; X 5-
•> ซ* ซ ซ
o > w 'S F
u . ~ 6, s
SO-O-^-S
•e S S I
| •S 3 2 ^
> S S ฃ u
S 2 S 2 S
o u u x x
00 00 00 O O
auji uotsssiSa^j
jo jdaojajuj
                                 -62-

-------
          Several general results may be noted by comparing the statistics
for the three SCIM calculations.  The results obtained using variable Q,
S, and H are very similar to the results obtained using mean Q, variable  S,
and H inputs; however, higher correlations and slightly smaller errors are
observed using variable Q.  When the variable Q, S, and H results are
compared with the mean Q, S, and H results, the differences are more strik-
ing.  The calculated mean is closer to the measured mean at 7 out of 10
receptors using the variable inputs, while the root-mean-square error is
smaller at all receptors using the mean inputs.  The correlation coefficients
are higher at all receptors using the variable inputs.  The frequency dis-
tributions of errors for each set of calculations was also examined.  It
was found that the variable input calculations produce some large overpre-
diction errors not found with mean input calculations.  In view of the
similar results obtained using mean and variable Q with variable S and H,
these large errors must be due to inaccurate estimates of the atmospheric
stability or the mixing layer height.
          The GHM results are slightly, but consistently, better for the
calculations with point source concentrations than for the calculations
without point sources.  The correlation coefficients with point sources
are higher at 7 of the 10 stations; the RMSE is smaller at 6 stations;
the error at the minimum measured concentration is smaller at 9 stations.
          The results in Tables 5 and 6 suggest that SCIM with Variable Q,
S and H and GHM with point sources gave the best overall results.   Of these
two, SCIM produced better results for estimating the mean and the maximum
measured concentration.  GHM produced smaller RMSE's and better correla-
tion as indicated by the correlation coefficient and the regression charac-
teristics.
                                  -63-

-------
 I

                         In order to get more insight into the validation results, the
 I            results were stratified by classes of five different parameters.  For
               each parameter, the mean of each class was determined for each set of
 |            model calculations and for the measured values.  The class means for
 •            hour of the day and day of the week are shown in Table 7.  The class
               means for temperature, wind speed, and stability are shown in Table 8.
 H                      With regard to day of the week, observations for Sunday were
               available for only 1 station (#0).  The means for Sunday were made
 |            comparable to other days of the week by multiplying the Sunday mean by
 _            the ratio of the annual mean for all stations to the annual mean for
 *            Station #0.  None of the six columns in Table 7 show any important
 •            variation by day of the week.
                         The results by hour of the day are listed in Table 7 and plotted
 g            in Figure 42.  With small exceptions all five sets of calculated results
               follow the general trend of the measured values.  Calculations using
 •            SCIM with mean Q, S, and H underestimate the magnitude of the cycle of
 •            measured values while the other four sets of calculations overestimate
               the cycle.  This result suggests there is a need to include model param-
 •             eters with diurnal variations, and that the diurnal variation is easily
               overestimated.  Clearly, the diurnal cycle of stability, mixing layer
 I             depth and emissions as represented in SCIM exaggerate the diurnal cycle
 m             of measured values.  The 6HM calculations which do not include a stability
               parameter show a better correspondence to the diurnal  cycle of measured
 I             values.  The addition of point sources in GHM calculations, while not
               changing the amplitude of the diurnal cycle, brings it into better agree-
 I             ment with measured values.

 I                                              -64-

I

-------
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
1
 I
 I
 I
 I
 I
1
                     Table 7. Comparisons of 1-Hour SC>2 Concentrations by

                              Hour of the Day and Day of the Week

Classification
Hour
01
04
07
10
13
16
19
22
Day
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday*
Number of
Cases
2143
2116
2111
2077
2134
2146
2146
2141

1796
3191
3350
3362
3193
1710
412
Measured
(Mg/m3)
209
237
303
258
223
214
231
226

232
242
232
233
238
240
(215)
SCIM (Mg/m3)
Variable
Q, S, H
269
311
444
223
166
163
226
247

239
245
229
264
241
306
(270)
MeanQ,
Variable S and H
341
372
336
223
191
179
219
268

253
257
243
272
252
308
(269)
MeanQ,
S andH
169
183
190
168
142
136
137
154

154
155
149
158
152
182
(152)
GHM (Mg/m3)
Without
Points
156
180
302
188
139
139
176
171

171
175
165
184
176
206
(178)
With
Points
194
221
352
236
180
177
211
209

210
216
205
224
217
254
(218)
*  Sunday values were observed only for Station #0.  The tabulated values were normalized by multiplying

   the Sunday mean by the ratio of the mean of all station measurements to the mean of all Station #0
   measurements.
                                           -65-

-------
I
I
I
I
I
I
I
I
I
I
I
I
1
1
I
I
I
I
1
Table 8.  Comparisons of 1-Hour SC>2 Concentrations by

    Temperature, Wind Speed and Stability Classes



Temperature (ฐF)
<15
15 to 24.9
25 to 34. 9
35 to 44. 9
45 to 54.9
55 to 64.9
>65
Wind Speed f m/sec)
<1.5
>1.5 to 2.0
>2.0 to 2.5
>2.5 to 3.0
>3.0 to 4.0
>4.0 to 5.0
>5.0 to 6.0
>6.0 to 8.0
>8.0 to 10.0
>10.0
Stability
(Turner-Pasquill)
1
2
3
4
5

Number
of Cases

36
511
2595
2573
2401
3316
5582

55
724
728
989
2570
2901
2952
4699
1271
271


17
559
1784
11667
2987

Measured
(jUg/m3)

363
322
299
306
279
210
167

226
258
254
278
252
228
230
228
231
230


143
213
235
236
248
SCIM (jug/m3)
Variable
Q, S, H

248
316
268
290
258
261
225

937
799
506
446
325
232
210
146
114
106


162
374
284
188
482
Mean Q,
Variable S, H

146
180
196
238
258
315
295

1054
893
544
467
338
246
211
146
103
83


255
434
300
185
533
Mean
Q, S, H

132
123
131
147
154
170
179

625
402
285
248
194
159
135
107
82
67


332
273
201
133
218
GHM (/ig/m3)
Without
Points

288
281
230
221
185
162
141

680
456
342
288
221
166
148
123
111
114


316
239
193
160
247
With
Points

320
306
256
255
223
202
194

763
515
396
335
266
210
190
158
136
132


340
302
248
198
284
                      -66-

-------
I
I
I
I
I
1
500
400
            300
        bo

        a
        o
        o
        e
        o
        O
            200
            100
                                                         Measured
                                                         SCIM  (Variable  Q, S, H)
                                                     & SCIM  (Mean Q, Variable  S,  H)
                                                         SCIM  (Mean Q, S.  H)
                                                         GHM (.without points
                                                         SHM (with points)
                                            10      13      16

                                              Hour of the Day
                                                                22
            Figure 42.  Variations in Measured and Calculated Mean One-Hour SO2 Concentration by Hour of the Day
                                                  -67-

-------
I

                        Temperature-dependent trends are tabulated in Table 8 and shown
•            graphically in Figure 43.  The importance of the temperature-dependent
              emission algorithm is suggested by the observation that the two sets of
•            calculations based on mean Q run counter to the trend of the three sets
m            of calculations using a variable Q and the measured values.
                        All five sets of calculated results show an important wind
•            speed dependence in Table 8.  This dependence is not evident in the
              measured concentrations.  It is difficult to explain the measured results,
•            since they are contrary to results observed at other locations such as
m            Chicago, and St. Louis (Koch and Thayer 1971).   Normally, if the wind speed
              is increased one would expect the pollutant concentration to decrease
ff            because the pollutant is diluted by a larger volume of air.  However,
              for this particular data some compensating effect occurred, such as a
jj            decrease in diffusion or an increase in emissions.  Clearly, the compen-
g            sating effect is not accounted for in the models.
*                      Stratifying the calculated and measured concentrations by
•            stability class (Table 8) also shows some systematic errors in the models.
              The results for class 1 are inconclusive due to the small number of cases,
f            less than two per station.  For the remaining four stability classes, all
_            five model  results show the same pattern in varying magnitudes.   The
™            means decrease from class 2 to class 4 and then increase for class 5.
•            The magnitude of the variations are especially large for the two SCIM
              results with variable stability inputs.  The pattern contradicts the
•            pattern of measured concentrations which show a small  consistent increase
              from class 2 to class 5.  Since wind speed tends to be light with very
•            stable and unstable conditions, it is likely that the difference between

I
1

-------
1
I
I
I
1
1 40ฐ
1
™ 300
ro
f ^
• 1
ฃ
O
I.P-*
•M
a
ts
s
o
f<5
200
1
f
100
I
1 .
1
Figu
•
1




































4






E
<ง













\

^
t--—-.
/
/



S
s
3 —














\
^
ฃ
1



15
/\

~~~~-^














•>
%
\
\


<^
il

)— *















^
^s
V









>-=

Ik
\T
V

^


^^







il


r"















— c
^

i

k


^



^V-
V
H-
ฎ-
A
•_









i.
"N
"x
\
i
JK
\


) —





	
—












k.
\
^













k
*\
V

--$.



\
r'













I/
— ^
\
^
^

><^













^

/
•x

[r
r^
^


Key









\
S>

S.
\
s

— 4
V<$

^













J




>
N
)



































































ฃL Measured
XX ซ
y^- •.
-ra ซ
^S> S
A r
^r t
• r
t
iCIM (Variable Q, S, H)
CIM (Mean Q, Variable S, H)
CIM (Mean Q, S, H)
.HM (without points)
.HM (with points)

10 20 30 40 50 60 70
Temperature, ฐF
re 43. Variations in Measures and Calculated Mean One-Hour SC^ Concentrations by Temperature
-69-

-------
I

^*            measured values and the three sets of calculated means without a stability
•            dependent parameter (SCIM with mean Q, S and H, and the two GHM's) are
              due to the effects of wind speed on the model calculations.  Although
J            the use of a mean wind speed is shown in Section 3.2.4.4 to improve
_            annual mean concentrations calculated using the GHM with point sources
^            added, this seems undesirable because it amounts to suppressing a measured
•            variation in atmospheric conditions.  It is more desirable to find modi-
              fications to the models or to other model inputs which represent the
I            compensating effects and leave the model compatible with results for other
              locations.
•                      In view of the interrelationships between the parameters considered
*            above, it is not possible to isolate the sources of error.  However, it
              appears that some hypotheses can be drawn which are worthy of future study.
•            The rate of change of mean calculated concentration with temperatures for
              the two best performing models (SCIM with variable Q, S and H, and GHM
m            with points) is less than the rate of change of mean measured concentra-
•            tion with temperature.  These models also show a larger mean diurnal varia-
              tion than the measured values.  These results suggest that the emission
•            algorithm overestimates the diurnal factor and underestimates the fraction
              of emissions which are , temperature-dependent.  Another hypothesis is that
•            SCIM overestimates diffusion effects associated with changes in atmospheric
M            stability.  This might be due to the inadequacy of the meteorological  mea-
              surements based on the Turner-Pasquill index to represent stability varia-
M            tions.  Of course, there are other possible sources of error which have


-------
              not been  investigated.   These include  the  effective  height  of  the  area
|            source  emissions,  the  allocation  of  emissions  to  geographical  areas  in
_            the emission  inventory,  and  the diurnal  variations in  the mixing layer
^            height, to mention a few.

9            3.2.4.2  Twenty-four Hour  SO? Comparisons  for  SCIM and GHM
•                      The calculated and measured  concentrations were averaged for
              each 24-hour  period and  compared.  The frequency  distributions for each
•            of the  three  sets  of SCIM  calculations and the set of  measured values
              are shown in  Figures 44  through 53 by  receptor location.  These figures
I            agree with the results of  the preceding figures for  hour concentrations.
M            They suggest  that  both sets  of calculations based on variable  S and  H
              inputs  more closely approximate the  distribution  of measured concentra-
I
              tions  than  the  calculations  based on mean S and H  inputs.  The frequency
              distribution  of 24-hour  SCL  concentrations for the two sets of GHM
B            calculations  and the measured  values at each of 10 New York City sampling
_            stations  are  shown  in  Figures  54 through 63.  Of the two model calcula-
*            tions  the frequency distribution of calculations using point sources more
•            closely coincides with the frequency distribution of measured values at
              7  of 10 stations.   For two stations (#0 and #1), the calculations without
ฃ            points show better  agreement with measured values.  The measurements at
—            the other station (#3) are about equally matched by both sets of calcula-
'            tions.
•                     Statistical comparisons of the five sets of model calculations
              with measured values are given in Tables 9 and 10.  As with the hourly
•            concentrations,  the 24 hourly  concentrations show that, of the three sets

I

f

-------
                                                                                                 99

                                                                                                 98


                                                                                                 95
                                                                                                85

                                                                                                80


                                                                                                70


                                                                                                60

                                                                                                50

                                                                                                40


                                                                                                30


                                                                                                20

                                                                                                15

                                                                                                10

                                                                  Key

                                                                Observed
                                                                Variable  Q, S, H
                                                             x  Mean Q, S,  H
                                                                Mean Q, Variable  H, S
10
                               100
                                                                                           10,000
                                                            1,000

                                         Concentration  (ug/m3)


Figure 44. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station 10
10
                               100

























































j

If











u
1












/,














^s

t
/











t
/
/








/
/ <6
/m
' /I
7
r
t






//
/
ซ
/







\
\
i
i
7











'
^
^
'O











/

/









1

/
y







8
X

?

1








Ob:
Vat
Me;
Mec

//
>/








Key
erved
•iable Q,
n Q, S, H
n Q, Vari












5, H
able t












<. s










































































QQ
95
90 _
85 1
D*
on r+
70 -n
f.n c
50 Q
o
40 ^*
o
o
30 g
^ 3
0
15 "
X
10 ~
5

                                                             1,000
                                                                                           10,000
                                           Concentration  (yg/m3)

 Figure 45. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for Hew York City Station #1
                                                  -72-

-------
I
I
I
I
I
I
f
I
1
I
1
I
1
I
1
I
I
t
I
                                               t
                                                        i\
                                                                   Key
                                                         9  Observed
                                                         •  Variable Q, S,  H
                                                         x  Mean  Q, S, H
                                                         +  Mean  Q, Variable H, S
                                                                                         n
                                                                                             99


                                                                                             98



                                                                                             95


                                                                                             90


                                                                                             85

                                                                                             80


                                                                                             70


                                                                                             60

                                                                                             50


                                                                                             40


                                                                                             30


                                                                                             20

                                                                                             15

                                                                                             10
10
                              100
                                                                                         10,000
                                                         1,000

                                       Concentration  (yg/m3)


Figure 46. Frequency Distribution of SCIM Calculated and Meatured 24-Hour Concentration! lot New York City Station 13
                                                                      Key
                                                            8 Observed
                                                            • Variable Q, S,  H
                                                            x Mean  Q,  S, H
                                                            + Mean  Q,  Variable  H, S
                                                                                             99

                                                                                             98


                                                                                             95


                                                                                             90


                                                                                             85

                                                                                             80


                                                                                             70


                                                                                             60

                                                                                             50

                                                                                             40

                                                                                             30


                                                                                             20

                                                                                             15

                                                                                             10
10                            100                          1,000                         10,000


                                         Concentration (pg/m3)

 Figure 47.  Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentration* for New York Ctty Station #10
                                                 -73-

-------
I
I
t
I
I
I
f
I
1
I
I
I
I
I
1
I
I
I
I
                               100
                                                                                            10,000
                                                            1,000


                                          Concentration (ng/m3)


Figure 48. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station #14
                            A/
                            i
                                                            Observed
                                                          •  Variable Q,  S, H
                                                          x  Mean Q, S,  H
                                                          +  Mean Q, Variable  H,  S
99


98



95


90    o


85    H,
      oป
80    2
      n>

70    ^


60    ง

50   %
      o
40    ~*
      o
      o
30    ฃ


20    =
      o>
15   _

10   ""
10                             100                          1,000                         10,000


                                          Concentration (yg/m3)


Figure 49.  Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentradoni for New York Ctty Station #17
                                                  -74-

-------
                                                                  Key
                                                          8   Observed
                                                              Variable  Q,  S, H

                                                          *   Mean Q, S,  H

                                                              Mean Q, Variable H.  S
                                                                                                99



                                                                                                98




                                                                                                95




                                                                                                90    o



                                                                                                85    |

                                                                                                80    ?
                                                                                           70



                                                                                           60


                                                                                           50


                                                                                           40



                                                                                           30



                                                                                           20


                                                                                           15


                                                                                           10
                                                                                                      3
                                                                                                     JD
                                                                                                      C
                                                                                                      ro
10
                              100
                                                                                          10,000
                                                       1,000



                                      Concentration (tjg/rn ^



: SO.  Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations forNew York City Station #27
                     ^
                          4-
                                                 m
                                                                 Key
                                                         e  Observed

                                                         •  Variable Q,  S,  H

                                                         x  Mean Q, S, H

                                                         +  Mean Q, Variable H, S
                                                                                            99


                                                                                            98
                                                                                                90     r>



                                                                                                ป     I1

                                                                                                80     ?
                                                                                            70
                                                                                                50
                                                                                                30

                                                                                                       o
                                                                                                       -*

                                                                                                       o
                                                                                            20    8

                                                                                            15    S
                                                                                            10    "^
 10                            100                   .       1,000                        10,000



                                          Concentration (pg/m3)



  Figure 51,  Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentration! forNew York City Station #28
                                                  -75-

-------
I
I
I
I
I
I
I
I
I
I
I
I
t
I
I
I
I
I
I
                                                              9  Observed
                                                              ป  Variable  Q, S, H
                                                              x  Mean Q, S,  H
                                                              +  Mean Q, Variable  H, S
                                                                                             10,000

                                            Concentration


 Figure 52. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station #31



























/
/












/
/
/











/
/
7-











-H*
A

/










/
/
//
/





/
I
1
I /
f Jy
//
/ i
$
> - -
r





/
7
//
///
/
i








7
/














/f
Y













'/








9
t
X
t


,7
//









Ubse
Varf
Mean
Mean











Key










rved
able Q, S, H
Q. S, H
Q, Variable H,














S






























0 100 1,000















10,

QS
90
DC

70


40
30
20
1 C
10
5

000
                                                                                                        •<
                                                                                                        o
                                         Concentration (pg/m3)


Figure 53. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station #36
                                                   -76-

-------
1
I
1
1
1
1
I
1
1
I
I
1
1
1
1
1
1
1
1
t
1
















0
Figure 54.



































































































/
r
J i
/ 1

100
Frequency Distribution of CHM




































































/








r/
/
/ /
//
1 j
1
I





/
i
/i
7
[









/I
I
?











I












!f\
1













1
Concentration (yg/
Calculated and Measured 24-Hour
- -- - -fd- <•>





4
/
//
//





1
I

II"


/
//
il
/
i







!
(I
1











'/
/












/






































































Key
% Observe
+ With PC
t Without


d
int Sour
Point S




ces
ource




s


- 99

- 95
90
85

- 70

-60
50
-- DU v
40
- 30
20
- 15
10 '
5

000 10,000
ป3)
SO Concentradoni for Nซw York City Station 10
	 . QQ






















































Key



•


Observed
With Poin
Without P


t Sou
Dint


rces
Sour


ces











95

- 85 1.
DJ
on rf
IV
- 70 -n

00 fj
- 50 •<
o
o
o
- 30 S
T
15 "
- 10 ^
. 5
. 2
0 100 1,000 10,000
Concentration (pg/m3)
ire 55. Frequency Distribution of CUM Calculated and Measured 24-Hour SO,, Concentration! for New York City Station 11
-77-

-------
1
1
1
I
1
1
1
1
1
1
1
1
1
1
1
I
t
1
1































0






























_














~





































































of
1
ft
// \
///

-.- -.





'/
1
VI
1
/
r



—


/
/ j
7
f











A
It
t










I
1


























PI





































































Key
9 Obse
* With
• With


rved
Poin
3.Ut P


t So
j1nt


ur(
Sc


.es
>ur


ces


99

- - 95
-. 90
o
-.85 I
- - 80 rt
1
--70 -„
,n *
n
- - 50 D
-. AO ฐ
o
30 S
20 1
15 8
..10 =
. . 5

100 1,000 10,000
Concentration (ug/m3)
Figure 56. The Frequency Distribution of 24-Hour SO2 Concentrations for the Two Sets
of CHM Calculation and Measured Values for New York City Station 13
. ^ A oa
























































//







'


^
/
/ ;
//
/









/

/









I
J
*











it
ji
f











k
1









/
i /
f







%


^
1







Key
Observed
With Pol
Without












nt So
Point
































urces
Sources






















95
90 o
85 —
03 o.
sn 2
80 1
70 7
en ' S
50 ^
O
40 •*
0
o
30 0
-i
3
20 g
' 8
15 ~-
M
10 ~"
5

0 100 1,000 10,000
Concentration (yg/ra3)
Figure 57. The Frequency Distribution of 24-Hour SO2 Concentrations for the Two Sets
of CHM Calculations and Measured Values for New York City Sudon 110
-78-

-------
1
1
1
1
1
1
1
1
1
I
I
t
t
1
1
1
1
















0









































































1










I
/

1 1
J

:
Figure 38. Tl
of CM
. . _


































































/
/
i



J
/.
II
1 I
II


1 4
1 I
T y
? f
I
I

LOO
(
ie Frequency I
A Calculatlonl



/
/
/
/ /
/ //
III
ur SOj Conceatradona, foi the Two Sect
el for New York City Station 114
on


















9














!
































Key
Observed
With Point
Without Pol


Sourc
nt Sc


es
urc


es






















. 95
90
• o
. 85 i
80 ฃ
70 -n
,0 !
60 |
50 4J
. 40 ฐ
,0 i?
20 a
o
. 15 "
. 10 ^
s

10 100 . 1,000 '• 10,000
t Concentration (pg/m3)
Figure 59. The Frequency Distribution of 24-Hour SOj Concentration forth* Two Sett
of CHM Calculation! and Measured Values for New York City Station 117
I
1

-------
I
I
I
I
I
I
I
I
1
I
t
I
I
I
I
I
I
I
I
10

                                              TL
                                                                   Key
                                                         t  Observed
                                                         +  With  Point Sources
                                                         •  Without Point  Sources
      99


      98


      95


      90

      85

      80


      70


      60

      50

      40

      30


      20

      15

      10
                                                                                                       2
                                                                                                       ง
                               100                          1,000

                                             Concentration (ug/m3)
10,000
                    Figure 60.  The Frequency Dlftrtbutloii of 24-IJour SOj Concentnttou.fcn tin Two Sea
                          of CHM Calculationi and Meuured Valuei for New Yolk City Station *27




























	 if














J











1
1
7^
-4






/

7
—r ~ 4-
1 t
4 -1
7
/(_
I


II
1!
I

1 1 j

/
I









/
/•











1
t






































8
+
•











Key




















Observed
With Point Sources
Without Point Sourc






0 wo i.ooo











es








- - 80

- - 60
- • 50






10,000
                                                                                                     3
                                                                                                     J3
                                                                                                     3
                                         Concentration  (ii9/m3)

                   Figure 61.  The Frequency Dlltrltuition of 24-Hour SOj ConccntnUonl fcr tha TWo Sซn
                        of CUM Calculatloni anJ Meaurcd Valuel tor New York City Station 128
                                                -80-

-------
                                           r
      99


      98



      95


      90


      85

      80


      70


      60

      50

      40


      30


      20

      15

      10
 10
10

                                                                    Key
                                                           8   Observed
                                                           +   With Point Sources
                                                           •   Without Point Sources
                               100
                                                             1,000
                                             Concentration
                  Figure 62.  The Frequency Distribution of 24-Hour SO2 Concentration! for.the Two Sell
                        of CHM Calculation! uil Meaiured Value) tor New York City Station ป31
                               100                          1,000

                                            Concentration  (ug/m3)


                  Figure 63.  The Frequency Distribution of 24-Hour SOj Concentrationl forth* Two Sett
                       of CUM Calculation! and Measured Value! forNcw York City Station 136
 10,000






















1
1
1
I
1







1
1
1




/
™





/
/




4
f
I
P



,
'






/

/



1
/












/
'





;


f
J






/
j
/
f /
/
/
/
/





/
It

1
I










r
1



















































8
+
•











Key
Observed
H1th Point
Without Poi












Sourc
nt So






















es
urces



































- - 95
- 90
85
80
- . 70

50
. . 40
. . 30
. . 20
. . 15
10
- - 5
.. 2
10,000
                                                                                                       c
                                                                                                       t
                                                                                                       3
                                               -81-

-------
Table 9. Model-to-Measurement Comparisons of 24-Hour NYC SC>2 Concentrations
  by Mean, Standard Deviation, RMSE, MAE and Error at Maximum Measurement


Number of Comparisons
^
4
DO
a
S

a
\n
a
rt *-^
(tj QJcf)
I 3 6
•* 3 -5.
ฃ CT< 60
U \J\ w
o 3ป
ps

SnP
f— 1 s

< A
a ซ-
rt O
dj c


.1 "ซ
X 3 ^~.
rt ซro^
rt S ~J0
t-> P ^>
fc 3
w S
Measured
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)

Measured
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
SCIM (Variable Q, H, S)

SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)

GHM (with points)
Maximum Measured (/ig/m )
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
NYC Station Number
0
364
303
421
440
283
307
350

163
241
258
106
119
121
247
314
192
143
150
183

242
161
113

121
917
118
41
-528
-555
-443
1
213
212
281
289
159
196
227

139
276
276
106
83
92
261
284
164
124
121
164

199
129
84

87
933
-203
-267
-619
-715
-628
3
257
292
367
384
246
268
310

163
246
274
112
111
115
274
309
194
190
184
175

206
140
135

134
1313
-1112
-1129
-1208
-1125
-1117
10
268
420
314
324
226
375
429

220
188
186
99
159
166
271
300
313
226
227
196

224
232
157

165
1613
-1025
-1166
-1300
-804
-749
14
266
229
170
180
93
106
131

98
137
150
47
43
45
133
148
161
151
126
106

118
138
126

103
607
-161
-142
-499
-485
-444
17
285
194
193
197
101
130
169

89
150
148
58
52
64
150
161
137
111
95
112

115
109
86

72
674
-451
-433
-605
-520
-480
27
281
185
211
211
118
128
156

98
160
159
60
52
55
158
179
126
108
95
107

125
100
83

72
547
-235
-286
-380
-249
-223
28
233
126
99
102
49
66
90

74
115
110
30
25
31
120
123
108
91
81
84

90
83
68

58
446
-284
-318
-384
-352
-333
31
283
186
225
229
122
99
158

76
213
202
63
40
72
212
204
105
115
87
113

122
82
96

64
474
-359
-375
-375
-376
-316
36
245
144
169
178
102
37
96

80
141
156
69
14
57
141
162
100
130
100
94

106
75
107

76
413
-301
-286
-308
-326
-314
                                     -82-

-------
Table 10. Model-to-Measurement Comparisons of 24-Hour NYC SO2 Concentrations
          by Correlation Coefficients, Variance  and Regression Coefficients

Number of Comparisons
fi -y
11
g iH
3d
MM
S s s J
0 a j^a

-------
of SCIM calculations, the calculations based on variable Q, S and H show
the highest correlation with measured values and the closed agreement
between calculated and measured means.  The smallest root-mean-square
errors occurred with the mean Q, S and H calculations.  As with the hourly
concentrations, the 24-hourly concentrates show that using variable S and H
leads to a higher frequency of large overprediction errors and fewer and
slightly smaller underprediction errors.  The 24-hour comparisons also
support the finding that using variable Q and mean Q give similar results
with slightly better results (smaller errors, larger correlations) occur-
ring with variable Q.
          These results indicate that one possible approach to improving
SCIM is to reexamine the means by which stability classifications and
mixing layer heights are determined.  The present method introduces more
error (indicated by MAE and RMSE) than the use of annual mean values;
however, there is some skill associated with the use of hourly determinations
indicated by the better correlation with measured values.  Some improvement
in overall results might be expected from using variable Q and mean S and H
input.  This combination was not tested.
          A statistical summary of the model-to-measurement comparisons
for 24-hour mean GHM concentrations is given in Tables 9 and 10 for calcula-
tion without and with point source contributions.  As with the 1-hour con-
centrations, the 24-hour concentrations show that calculations with point
source concentration give slightly but consistently better results over the
10 stations.  The calculations with point sources have larger correlation
coefficients at 7 stations, smaller RMSE's at 8 stations, and calculated
means closer to measured means at 8 stations.

                                 -84-

-------
I
I
                         As  with the 1-hour comparisons,  the 24-hour comparisons  show
g             that the  variable Q,  S and H version of SCIM and  the  GHM with  point
               sources produced the  best results.   When these two are compared  the
•             results are mixed.  The means,  mean errors and correlation  coefficients
M             are similar for both  models. SCIM  produced smaller errors  in  estimating
               the maximum measured  concentration, and GHM produced  smaller RMSE's and
•             better regression line characteristics.   Favorable regression  line
               characteristics include a slope near unity and an intercept near zero.
               3.2.4.3   Selection of a Sampling Interval  for Calculating Annual Mean
•                      Concentrations with SCIM
                         It  was previously found that  many characteristics of the
•             frequency distribution of hourly concentrations over  a long-term period
               can be estimated by calculations for a  small  sample of the  hourly  periods
I             included  in the long-term period (Koch  and Thayer 1971).  The  annual mean
m             is  a characteristic which is particularly  well -suited | to statistical
               sampling.   A  sampling procedure found to be successful  is one  in which
f|             one hour  out  of a set sampling  interval  is selected.   However, the
               selection procedure is set up so that all  hours of the day  are equally
|             represented in  the selected sample.  The success  with which the  annual
_             frequency distribution can be represented  by various  size sampling
™             intervals is  shown for 10 New York  City locations in  Appendix  D.   With
•             input data available  for every  3 hours,  Table 11  shows  the  mean  error and
               root-mean-square error in estimating an  annual mean concentration  with
I             various size  sampling intervals.  The error is based  on  results  obtained
—             for 10 New York City  locations.   An error  is  defined  as  the annual  mean


I

I

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
calculated using a  stated  sampling interval less the annual mean calcu-

lated using all available  input data (one set every three hours).  On

the basis of the results shown in Table 11 it was decided to use a

sampling interval of  39 hours in estimating annual mean concentrations

with SCIM for New York City.   It may be noted that between a sampling

interval of 36  hours  and 48 hours, the results in Table 11 suggest a

sharp upturn in the RMSE associated with the use of proportionate

stratified sampling as described here.
       Table 11. Summary of Mean Error and RMSE Associated with Various Sampling

                Intervals for Proportionate Stratified Sampling in SCIM
Sampling Interval
(Hours)
3
6
9
12
18
24
36
48
72
96
Mean Error
(Mg/m3)
0.0
-4.8
4.3
-0.1
-2.1
-3.8
1.8
-6.7
6.0
-7.9
Root-Mean-Square Error
(jig/m3)
0.0
5.5
4.8
5.0
5.5
5.7
6.1
11.6
16.5
29.7
3.2.4.4  Annual Mean  $02 Concentrations Using Ten Model and Data
         Conditioning Combinations

          The 10 model  and  input data combinations being compared were

described in Section  3.2.1.   Each combination was used to calculate

annual mean S02 concentrations  for 75 sampler locations in the New York

New Jersey - Connecticut AQCR for the year 1969; however, the four com-

binations using the GHM were only used to calculate concentrations at
                                 -86-

-------
I

              71 sampler locations.  Although the four omitted locations are located
I            near the perphery of the AQCR, it is estimated that their omission does
              not have an important bearing on the overall  evaluation.
I                      A statistical summary of the results obtained for each model
I              and data conditioning combination is given in Table 12.  The mean errors
                                3
              vary from -62 yg/m  for the GHM, using mean emission rates and wind speed
                                                   o
I              and without point sources, to 76 yg/m  for the AQDM.  The smallest mean
                             3
              error of 3 yg/m  occurred for the COM with variable stability and mixing
I            layer height.  The COM also produced the smallest root-mean-square error
—            (RMSE) and mean absolute error (MAE).  Two of the combinations, COM with
™            mean stability and mixing layer height and AQDM, had mean errors, RMSE's
•            and MAE's which were notably higher than any of the others.  Next to
              COM, the GHM combinations with point sources  and the two  SCIM combinations
•            with variable stability and mixing layer height produced  very similar
              RMSE's and MAE's, which are slightly larger than those for COM.  It may
•            be noted that GHM combinations underestimated while the SCIM combinations
•            overestimated, on the average.  All combinations produced high correla-
              tions with the measured concentrations; however, the 0.89 value for
•            AQDM led the rest by a slight amount.  A desirable regression relationship
              of measured on calculated values would have a slope of unity and an
•            intercept of zero.  No one combination is clearly superior to the others.
m            It may be noted that the smallest intercepts  and slope closest to
              unity occur for different SCIM combinations.   The error for the maximum
I            measured value shows best results occurred for two of the GHM combinations
              and one COM combination.
I
I
I
-87-

-------
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
I
I







rt
Q
^*
•rt
U
*
u
O
^
ฃ
Q)
^
If
•1-*
I

a
ง
Q

Q
U
CM
f~\
8
rt
(U
^
<-t
I
2
C/3
h— <
U
to







Statistic
in
tx

R

in



f^





ฃ




tx


T-l
^^





in



R

R








| Number of Comparisons
in
ro

m
CO

in
ro
TH

o
TH




o
TH


o
TH


o

TH




ro
1— 1


in
ro

in
ro
rH







1 Mean Measured (jjg/m )
O tx
CM
00 ro
CO
TH


TH tx
CM

CM ฃ;
TH 1




tx CO
O co
TH |


ซ
1


00 C*J
is. *o
1




oo tx
00 •*<
1


CM lx
o in



in T|*

in -^





^^
ro
R
Root-Mean-Square Error (/xg/i
Mean Absolute Error (jug/m^)
CM CM
r-t ro
TH ro
i
2 !ง
TH TH
1
tx O
OO TH
l ro

ro ฎ
ro O
TH CM
l



ro in
TH TH
1

O CM
TH TH
1

TH T*
tx O
TH TH
1



o\ o
•H* irt
TH
1

<& CM
o ^o
1
TH s
7 rt






ro -~*
Largest Negative Error (jxg/m
Largest Positive Error (/xg/m
,

CM

^
CTk
ro

CM
ro




g
CM


CM
ro
CO


^
[ซ^
CM




8
CM


oo
vo
CM

TH
00
CM







ro
3
f
Pi
S S
O
** o
OS tx
o
*
00 t^
0

88 S
•
O



CO O
00 t"N-
•
o

CM tx
CO >O
0

CM tx
00 VQ
•
O



CM OO
00 VO
0
O

ro CTi
oo to
O
S ^
o"







Correlation Coefficient
Reduction of Variance (%)
TH
o
8
o
in

o

S
•
0



R
,
o

o

o

m
co
*
o



oo
,
O

o
*
o
o
o







Slope of Regression Line
o

in
ro


ro


TH
to





in



ซ•
^


T^4
f^





oo



TH
CM

O
CM








Intercept of Regression Line
O ro
in TH
ro
C^ TH
un o
CO TH

O CM
in TH
CO TH

oo in
ro




in o
00 rH
ro l


in rH
OO TH
ro l


m m
OO tx
ro l




in eft
oo T)<
ro TH


in oo
oo m
CO 1

in tx
00 Tf
ro l


co"'
B
""So
ซj
•CS
Maximum Measured (jxg/m^)
Error for Maximum Measured
-88-

-------
 I

              3.2.4.5  Annual Mean Particulate Concentrations Using Ten Model and Data
 _                     Conditioning Combinations
 ™                      Annual mean particulate concentrations were calculated for about
 •            113 sampling locations in the New York - New Jersey - Connecticut AQCR
              using each of the 10 model and data conditioning combinations described
 •            in Section 3.2.1.  A statistical summary of the model-to-measurement
              comparisons obtained for each combination is given in Table 13.  A back-
 "            ground concentration of 35 yg/nr* was added to all model calculations.
 •                      The smallest mean errors were produced by the two COM combina-
              tions.  However, the mean errors were small for all 10 combinations,
                                         13                                       3
              varying from -24 to 22 yg/m  compared to a measured mean of 82 yg/m .  The
              same relatively small RMSE and MAE were obtained for the two SCIM combina-
 •            tions and the COM combination which use variable stability and mixing
 •            layer height.  The largest positive error was notably lower for the SCIM
              and SCIM (Q") combinations than for any of the others.  This resulted in
 •            a notably smaller error range.  Except for the SCIM (07,S",FT ) combination,
              for which correlation coefficient was low, all the correlation coefficients
 I            were similar.  The highest correlation coefficients occurred for the SCIM
 •            and SCIM (Q") combinations.  The SCIM and SCIM (qi) also produced the best
              regression of measured on calculated values, having both the smallest
 •            intercepts and the slopes most nearest to unity.  The smallest errors
              for the maximum measured concentration occurred for the AQDM and COM (S'JL)
 |            combinations.

I

I

|                                             -89-

I

-------
  I
  I
  I
  I
  I
  I
  I
  I
  I
  I
 I
 I
 I
 I
 I
 I
 I
 I
I
si

'2.
2
Q
O
Q
g o
rt <
Q
ฃ
0 ^ ง
•S So
O tJ3 P-ซ
1 S'M
D o i2l ^
1
•B
a s
ง x
}
0 U
g
CJ
Q) S I^-T*


"a O |O
o -"
a
ฃH ^
rt
PL, ,-^
a - S
n) S I5C
Q> KH *
5 U Ito
•^ oo •>
^ 12;
a
8 2 -
-8 Q |O

SOO
B ง
o U
U oo
I— (
"CJ — — — —
1

ro
T— 1
At
TabL
Statistic
CO
TH
TH

ro
T-l
T-l

CO
T-l
TH



CM
T-l

T— 1
T-l



CM



CM
TH







TH
I—I



TH
T-l

TH







Number of Comparisons
oo


CM
00


CM
oo




CM
00

CM
00



CM
00



CM
00







TH
00




00


00







I Mean Measured (yg/m3)
00
oo


it


CM
O
*""*



ro


O



S



CM
a\







oo
in




vo


01







12
Mean Calculated (yg/m )
vo


oo


8




rt
CO

2



CM
 W
vo m
TH |

0) •*
vo m
TH 1

ซr
s
*6ซ6
"••ซ
**2T
Maximum Measured (yg/m^)
Error for Maximum Measured
-90-

-------
Section 4.0
CONCLUSIONS

-------
                              Section 4.0
                              CONCLUSIONS

          Conclusions (1-5) regarding the use of the Sampled Chronological
Input Model (SCIM), a multiple-source Gaussian plume model, to estimate
short-term SO,, concentrations (e.g., 1-hour and 24-hour concentrations) are
based on model-to-measurement comparisons for 1 month of Chicago, 3 months
of St. Louis and 1 year of New York City (NYC) data.  The model was analyzed
using NYC data for three types of inputs, including:  (1) variable emission
rates, stability classifications and mixing heights (variable Q, S, H),
(2) mean emission rates and variable stability classifications and mixing
heights (mean Q, variable S, H), and (3) mean emission rates, stability clas-
sifications and mixing heights (mean Q, S, H).  The model was analyzed  using
St. Louis and Chicago data for the first two types of input.   For comparison
purposes, an analysis was also made of the use of the simplified Gifford-
Hanna Model (GHM).

          1.  Comparing the results for the three types of input to SCIM,
it is concluded that:
          e    Use of a mean, rather than a variable, emission rate may
               either increase or decrease the root-mean-square error
               (RMSE) at a receptor but will decrease the correlation
               with measurements (observed at 10 of 10 St.  Louis receptors
               for 2-hour concentrations, 5 of 8 Chicago receptors for
               1-hour concentrations, and 10 of 10 NYC receptors for 1-hour
               and 24-hour concentrations).
          •    Based on comparisons using NYC data and mean emission rates,
               the use of a neutral stability classification  and a mean
               mixing height will decrease the RMSE at a receptor (observed
               at 9 of 10 receptors for 1-hour concentrations and 8 of  10
               receptors for 24-hour concentrations).
                                  -91-

-------
          •    Based on NYC comparisons, the combined use of a mean emis-
               sion rate and mixing height and a neutral stability classi-
               fication will decrease the correlation with measurements but
               will also decrease the RMSE (observed at 10 of 10 receptors
               for 1-hour concentrations and 7 of 10 receptors for 24-hour
               concentrations).


          2.  In evaluating GHM, it was concluded that adding point source
contributions (i.e., calculated using SCIM) to GHM calculations improved
the results for this model.  The RMSE was smaller at 6 of 10 NYC receptors,
the correlation coefficient was higher at 6 of 10 receptors, and the standard
deviation of calculated concentrations was closer to the standard deviation
of measured concentrations at all 10 receptors.

          3.  Comparing SCIM and GHM based on NYC data, SCIM produced the
least annual mean error at 6 of 10 receptors, the closer agreement between
standard deviations of calculated and measured concentrations at 5 of 10
receptors, the least error in estimating the maximum measured concentration
at 6 of 10 receptors, and the highest correlation coefficient at 3 of 10
receptors; GHM produced the least RMSE at all 10 receptors.

          4.  There is a need to improve the input data used with the multiple-
source Gaussian plume type of model, particularly atmospheric stability infor-
mation, since the model is very sensitive to the rather gross changes in
stability which are routinely introduced.  SCIM calculations on the average,
greatly overestimated concentrations associated with Turner-Pasquill stability
classes 2 and 5.

          5.  Calculations based on a NYC emission algorithm developed in
this report, particularly when applied with GHM, generally agree with diurnal
and temperature dependent trends in measured S0ฃ concentrations.  Further
improvements in this algorithm are desirable but require more detailed
information.


          Conclusions (6-8) regarding the use of several versions of the

multiple-source Gaussian plume model and GHM to estimate long-term mean con-

centrations of S02 and particulates are based on model-to-measurement com-

parisons for the same data periods and locations.


          6.  The use of variable emission rates for SCIM and GHM are not
able to demonstrate any conclusive improvement in model validity over the
use of mean emission rates.  It is inferred that this result is  due to the
failure to properly treat other causes of variance, such as those associated
with atmospheric stability.
                                  -92-

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
          7.  Based on results for NYC, the Climatological  Dispersion
Model (COM) and SCIM versions of the multiple-source Gaussian plume model
produce a smaller station-to-station RMSE than the Air Quality Display
Model (AQDM) version (i.e., RMSE's of 52 and 59,  respectively, compared
to 92, with an overall mean of 135 ng/m3 of SC^;  RMSE's of  22 and 22 com-
pared to 36 with a overall mean of 82 yg/m^ of parti culates) .
          8.   Although the NYC validation statistics  for GHM,  COM,  and
SCIM are similar for SC^, GHM results for particulates have  a much  higher
station-to-station RMSE than do COM and SCIM (i.e., RMSE of  60  compared
to 22, with an overall mean of 82 yg/m ) .
                                  -93-

-------
Section 5.0
 REFERENCES

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
                              Section 5.0

                               REFERENCES
Blade, E. and E.F. Ferrand.  1969.  Sulfur Dioxide Air Pollution on New
  York City:  Statistical Analysis of Twelve Years.   Journal of the
  Air Pollution Control Association, Volume 19, Number 11, pp. 873-878.

Briggs, G.A.  1969.  Plume Rise.  AEC Critical Review Series.  U.S. Atomic
  Energy Commission, Division of Technical Information.  (Available as
  TID-25075 from NTIS.)

Calder, Kenneth L.  1970.  "Some Miscellaneous Aspects of Current Urban
  Pollution Models"  from Stern (ed.), Proceedings of Symposium on
  Multiple-Source Urban Diffusion Models^U.S. Environmental Protection
  Agency, Air Pollution Control Office, Research Triangle Park, North
  Carolina.

Calder, K.L.  1971.  A Climatological Model for Multiple Source Urban
  Air Pollution.  Proceeding 2nd Meeting of the Expert Panel  on Air
  Pollution ModelingTNATO Committee on the Challenges of Modern Society,
  Paris, France.

Chamot, C.  et al.  1970.  A Computerized Air Pollution Data Management
  System_APICS.   ANL/ES-CC-006, Argonne National Laboratory,  Argronne,
  Illinois.

Gifford, F.A.   1972.   Applications of a Simple Urban Pollution Model.
  Preprints Conference on Urban Environment and Second Conference on
  Biometeorology, Philadelphia, October 31-November 2, 1972.pp. 62-63.
  American  Meteorological Society.  Boston, Massachusetts.

Gifford, F.A.  and S.R. Hanna.  1973.   Modeling Urban Air Pollution.
  Atmospheric  Environment 7, pp. 131-136.

Haltiner, G.J. and F.L. Martin.  1957.   Dynamical  and Physical Meteorology.
  McGraw-Hill, New York.

Hanna, S.R.  1971.  A Simple Method of Calculating Dispersion from Urban
  Area Sources.   Journal of the Air Pollution Control Association,
  Volume 21, Number 12, pp. 774-777.

Holzworth,  G.C.   1972.  Mixing Heights, Wind Speeds, and Potential  for
  Urban Air Pollution Throughout the  Contiguous United States.AP-101,
  U.S. Environmental  Protection Agency, Office of Air Programs,
  Research  Triangle Park, North Carolina.  (Available from GPO.)

Koch, R.C.  and S.D. Thayer.   1971.  Validation and Sensitivity Analysis
  of the Gaussian Plume Multiple-Source Urban Diffusion Model.
  GEOMET Report  Number EF-60.  GEOMET,  Inc., Rockville, Maryland.
                                 -94-

-------
I

            Martin, Delance 0.  and Joseph A.  Tikvart.   1968.   A General  Atmospheric
              Diffusion Model  for Estimating ting the Effects  of One or More Sources on
•            Air Quality.   Presented at 61st Annual Meeting  of the Air Pollution
•            Control  Association. St.  Paul,  Minnesota.June 1968.APCA-68-148.
            IMcElroy, J.L.  1969.   A Comparative Study of Urban and Rural  Dispersion.
              Journal  Applied  Meteorology 8.  pp. 19-31.
_          Roberts, J.J.  et al.   1970.   Chicago Air Pollution Systems Analysis
I            Program:  A Multiple-Source Urban Atmospheric Dispersion ModeTi
*            ANL/ES-CC-007, Argonne National Laboratory, Argonne, Illinois.

I

I

I

I

I

I

i

I

I

I

I

I

I

I
NOAA Environmental Data Service,  1968.  Climatic Atlas of the United
  States, U.S. GPO, Washington, D. C.

Singer, I.A., Y.B. Lee, C.M. Nagle and D.H. Slade.   1970.  Comparative
  Studies of Urban and Rural Wind Climatology.  Proceedings of Seminar
  Wind Loads on Structures, ed. A.N.L. Chiu, Held at University of
  Hawaii, Honolulu, Hawaii, October 19-24, 1970, pp. 39-53.  (Available
  from NTIS, PB203449.)

Saucier, W.J.  1955.   Principles of Meteorological  Analysis.   The University
  of Chicago Press, Chicago, Illinois.

TRW Systems Group.  1969.  Air Quality Display Model.   Contract Number
  PH 22-68-60.  Prepared for Department of HEW, Public Health Service
  Consumer Protection and Environmental Health Service, National Air
  Pollution Control Administration, Washington, D.  C.   (Available from
  CFSTI as PB189194.)

Turner, D.B.  1964.  A Diffusion Model for an Urban Area.  Journal
  Applied Meteorology 3, pp. 83-91.

Turner, D.B.  1968.  The Diurnal and Day-to-Day Variations of Fuel
  Usage for Space Heating in St. Louis, Missouri.  Atmospheric
  Environment, Volume 2, pp. 339-351.

Turner, D.B., J.R. Zimmerman, and A.D. Busse.  1972.  An Evaluation
  of Some Climatological Dispersion Models.  Proceedings of the Third
  Meeting of the Expert Panel on Air Pollution Modeling, October 2-3.
  1972, Paris, France.Published by Air Pollution  Technical  Information
  Center, Office of Air and Water Programs, Environmental Protection
  Agency, Research Triangle Park, North Carolina, 27711.
                                 -95-

-------
  I
  I
  I
  I
  I
  I
  I
  I
  I
 I
 I
 I
 I
 I
 I
 I
 I
 I
I
                        Appendix A

AN ALGORITHM FOR ESTIMATING AREA SOURCE SOg EMISSION RATES

-------
                               Appendix A
                                          more crj

A.I  BASIS FOR NEW YORK CITY HOURLY S00 EMISSION RATE ALGORITHM
I

I
•                 AN ALGORITHM FOR ESTIMATING AREA SOURCE S00 EMISSION RATES

I
                      An algorithm has been developed for estimating SOp emissions in
g          New York City from area sources.  It is designed to use the emissions data
_          prepared for the Air Quality Implementation Planning Program (IPP).   The
1
•          IPP area source data consists of a single annual source emission rate for
•          each grid square.  The area source algorithm developed here is  cast  in a
            general framework so that it can be applied to any large city where  IPP
•          area source data, or its equivalent, are available.  The area source
            emissions are assumed to be due primarily to residential and commercial
•          space heating requirements and to be functions of the ambient air tempera-
•          ture and of hour of the day.
                      It is recognized that annual  emissions occur as a function of
I          both diurnal and seasonal  influences resulting from climate and weather,
            and from patterns of human activities.   The weakness of previous attempts
I          to estimate these effects  directly from emission information is illustrated
m          in Table 4 of the main body of this report.  Another source of  data, namely
            SOp concentration measurements from centralized downtown sampling stations,
•          was used in this study to  describe the  annual  distribution of emissions.
            The obvious disadvantage of using this  data to model  emissions  is that
|          systematic variations due  to meteorological effects such as atmospheric
—          stability and wind speed will be included both in the emissions estimates
            and in the diffusion model.   However, the double effect may be  small  since
            it is only the consistent  long-term effects which are being represented  in
                                   A-l

-------
I
            the emissions.  For instance, the average wind speed in summer (8 mph) is
I          only slightly less than the average wind speed in winter (11 mph) (NOAA
            Environmental Data Service 1968).  Also, the consistent diurnal variation
I          in atmospheric stability from unstable in the afternoon to stable in the
_          early morning is reduced in the city by increased roughness and thermal
            emissions.
•                    Data from a study by Blade and Ferrand (1969) of the S02 concen-
            trations in New York City were used to formulate an emission algorithm.
|          They analyzed a large set (12 years of almost continuous hourly observa-
—          tions) of SOp concentrations.  They summarized the data by hour of the
•          day, day of the week, and week and month of the year.  They also invest! -
•          gated the functional relationship between normal temperature and SOp con-
            centration.  Blade and Ferrand present diurnal curves of the mean hourly
0          SOp levels for each calendar month.  There are peaks in the concentrations
            at 7 a.m. and 7 p.m.; these peaks being most pronounced in the winter
I
•          season.  The area source algorithm described here is based on values
•          extracted from these 12 diurnal  curves.  No discrimination is made between
            weekdays and weekends since only a small difference was found in Blade and
•          Ferrand's study.
                      The first step in the  formulation of the algorithm was to find
•          the mean hourly SC^ concentration and temperature for each hour of the day
•          for each month of the year.  Climatological temperature data from LaGuardia
            Airport for the period 1962-1971  were used to obtain mean hourly tempera-
•          tures for each month.   A graph was prepared for each hour of the day of
            the 12 mean monthly SOp concentrations  versus the respective climatological
I                                             A-2
I

-------
  I
  I
            temperatures.   As  examples,  the two graphs  for Hours  0  and  9  are  shown
  I         together in Figure A-1.
  _                   Excluding the  values  for the  four warm season months  (June-
  '         September), the SOp concentrations closely  approximate  a linear function
 M         of temperature. A mean  of the  hourly concentrations  for the  four warm
            season months  was  used to  represent a "base" level.   A  straight line was
 M         fitted to the  remaining  eight data points.   The  temperature at  which the
  ,          line  intersected the base  level was taken to be  the "threshold" tempera-
 •         ture  at which  space heating  would  just  be required for  that hour.  The
 •         slopes of the  straight lines, the  base  levels,  and the  threshold  tempera-
            tures were plotted as  a  function of hour of the  day and fitted  by smooth
 •         curves.   The adjusted  values of base level, threshold temperature and
            slope read from the curves for  each hour of the  day are shown in  Table A-1.
 •         The base level  values  have been normalized  by dividing  by the sum of the
 m         24 values. Each slope value gives the  rate of change of SO 2 concentration
            with  temperature.   This  is assumed to be equal to the rate  of change of
 •         space heating  emissions  with temperature.   The base level concentration is
            assumed  to be  proportional to the  non-temperature dependent emissions.
 •         For the  present, the units of the  emission  rates  (both  space heating and
 m         non-temperature dependent  sources)  are  undefined  and  will be designated as
            index values.   The index values were determined  in terms  of measured SOg
 I          concentrations  but will  be multiplied by an appropriate scaling factor
            later.   The 24-hour threshold temperatures  are shown  in Figure  A-2, where
 |          for comparison  two climatologically normal  days are included, one repre-
            senting  a fall  or  spring day with  a mean temperature  of 65ฐF and  the other
•

I                                            A- 3
I

-------
1
1
1
1
1
1
1
I
I

1



1


1

1
I
1
1
1
1
1








a,
a
a
o
4-i
K

-------
 I
 I
 I
 I
 1
 I
 I
 1
 I
 I
 I
 1
 1
 1
 I
 I
 I
I
1
Table A-l.  Numerical Values for Constants in the New York City Emission Algorithm
Hour
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Base Level
Emission Factor
(J.) (day/hr)
0.0428
0.0417
0. 0399
0. 0399
0.0431
0. 0491
0.0543
0. 0550
0. 0526
0.0484
0.0421
0. 0389
0. 0365
0. 0343
0. 0340
0. 0340
0. 0343
0. 0351
0. 0368
0. 0386
0. 0406
0.0421
0.0428
0.0431
Temperature Sensitivity
Threshold (T.)(ฐF)
56
55
55
55
56
58
59
61
63
64
65
65
65
65
65
65
65
65
65
65
65
64
62
60
Temperature Sensitivity
Factor I. (Index/ฐF)
0.0161
0.0137
0.0137
0.0143
0.0181
0.0350
0. 0590
0. 0702
0. 0593
0. 0477
0. 0436
0.0411
0.0411
0.0416
0.0418
0.0433
0.0480
0. 0518
0. 0547
0. 0558
0.0544
0.0500
0. 0340
0. 0203
                                    A-5

-------
I
I
t
I
f
I
1
I
I
I



||


I


t

—




I






1


I


                   50
                   40
                    30
                                                                            Janu
                                                                   s
                                                                       Norn al 6J
ฐFDiy
                                                                                     Threi hold Tern jerat ire
ry
    ean
                       0    24   6    8    10   12   14   16   18   20   22   24

                                                  Hour


                        Figure A-2.  Threshold Temperature Profile  and Representative Diurnal Temperature Profiles
                                                               A-6

-------
I

            a January day with a mean temperature of 31 ฐF.  For a typical spring or
fl          fall day requiring minimum space heating (e.g., a mean daily temperature
            of 63ฐF), the 65ฐF line in Figure A-2 may be lowered by 2ฐF.  This would
j[          indicate that space heating would be required for the morning hours 0600
^          to 1000 and for the evening hours 1900 to 2200.  For progressively colder
™          days, the space heating requirement will gradually extend to all hours of
j|          the day.
                      The relationship between S02 concentrations and temperature
:j|          which is illustrated in Figure A-l is assumed to relate area source emis-
            sions from space heating and temperature.  The temperature dependent emis-
™          sions have the following form:
I
"          where

ซ              (Qu).  . = area source emission rate for hour i, day of year j,
                     >J   due to space heating (Ib/hr)
•                   Q. = total annual area source emission (Ib/yr)
                     FH = fraction of the total area source emission which is
flf                        temperature dependent
                     T. = threshold temperature for hour i (ฐF) (refer to Table A-l)
•                 T.  . = observed temperature for hour i, day of year j (ฐF)
w                  i >J
                     I-j = temperature sensitivity of emissions for hour i (Index/ฐF)
•                        (refer to Table A-l), and
                      K = index scaling factor (yr/Index hr).
I


-------
The parameter FH represents the fraction of the annual emissions which
are temperature dependent.  Presumably, these emissions are primarily due
to space heating requirements.  The constant K is a scaling factor which,
for a specific set of I. values and the climatological mean hourly tem-
peratures for each month, is defined as
                         •24     12
                 K = 1 "
                        -i
E  I   E  \  (Tn-  -  T    )                       (A-2)
       k=l   k    n    1)k J
where
         N.  = number of days in month k, and
       T. .  = mean temperature for hour i, month k.
        1 ,K
          The base level area source emissions, assumed the same for each
day of the year, show a diurnal variation which is incorporated into the
following non-temperature dependent portion of the algorithm.

                        QB< = QA (1 - FH) J./365                        (A-3)
where
        QB  = base level area source emission rate for hour i (Ib/hr),
          i   and
         J. = base level emission factor for hour i (day/hr) (refer to
          1   Table A-l).
          The J. values taken from Table A-l already have been scaled.
They represent the ratio of emissions during the ith hour to a total day's
emissions.
                                   A-8

-------
I
I
I
•            Q,  4 = 0,
                       The  complete  form of the  algorithm for  the  area  source  emission
             rate  is  then

                      ,   + QH     =  Qfl  (1  -  FH)  J./365,  (For  T.  <  T.   '
                      3i     Hi,j     A        H    '             '  ~   ')J
                                                                                     (A-4)
*           where
I
                     Q.  .  =  area  source  emission  rate  for  hour  i,  day  of  the year  j
                      1ปJ    (Ib/hr).
                      The  numerical  values  of J., T.,  and  I. are given in Table A-l.
•           Using  the climatological  data for LaGuardia Airport, the constant K is
             equal  to 0.0002337 yr/Index  hr.
I                    To complete  the algorithm in a completely objective manner, the
m           P.,  factor can  be  derived  from the same 12-year data sample used to derive
             the parameters  J., T., and I..  The temperature-dependent portion of the
fl           sum of all S02  concentration observations  can be evaluated using the monthly
             mean hourly temperatures  derived from LaGuardia; the average annual sum
0           for the 12-year period is  exactly the value of 1/K, or 4278.64 Index hr/yr.
^                    The  non-temperature-dependent portion of the SO^ concentration
*           observations is the remainder, which is an average of 10,413.45 Index hr/yr.
•           Therefore, the  FH factor  is defined as

                                IF  =      4278.64 	= Q 2g
                                rH    4278.64 + 10,413.45   u'^'

I
•                                             A-9

1

-------
 I
*
I
             It  may be  noted  that  this  differs  greatly from 0.80 which  was  previously
 1
             judged  to  be  appropriate  (Roberts  et  al.  1970).
 •                    The sources  which  contribute  to FH  are  difficult  to  identify
             since many effects  are represented, including space  heating of residences,
 |          space heating of  commercial  establishments, industrial  temperature-dependent
^          operations, hot water  requirements, etc.   This value of F,,  = 0.29  is  used
             in  all  applications  of the algorithm  to the New York City experiments.
flj          For comparison purposes,  GHM calculations  using variable wind  speeds  and
             emission rates were  made  with emissions based on  FH  = 0.29  and a previously
P          assumed value of  0.80.  The  resulting calculations are  plotted as  a func-
 _          tion of temperature  intervals along with  measured values in Figure A-3.
 ™           The results suggest  that  the use of Fu =  0.29 gives  a slope (and a result-
                                                  n
•           ing temperature sensitivity)  which is very nearly parallel  to  the  measured
             values.

             A.2 COMPARISONS  WITH  OTHER  HOURLY EMISSION RATE ALGORITHMS
•                     It  is informative  to compare the preceding area source emission
             algorithm  with the area source algorithms  used in the application  of  the
H           GEOMET  diffusion  model  to the St. Louis and Chicago  sites (Koch and Thayer
M           1971).  Each  of the  three algorithms  are  tailored to the New York  City
             climate (LaGuardia), and  put  in the form
                                            Qi = Qj^                               (A-5)
            where
                     Q. = area source emission rate for hour i  (Ib/hr)
                     QT = annual area source emission (Ib/yr)
                                               A-10

-------
 I
 1
 I
 I
1
1
I
I
f
I
O
0
O
      600
      500
      400
      300
       100
                     N
                            .80
                         F^ ^0. 2 >
>^
                                         S
                                              \
               Msasur
                                                     id
         :14     15-24    25-34    35-44    45-54   55-64     65

                                    Temperature Interval (ฐF)


          Figure A-3.  Mean GHM Calculated and Measured Hourly SQ^ Concentrations as a Function

                                      of Observed Temperature
                                               A-n

-------
I

•                   K.  = algorithm coefficient for hour i  which  can  be  a  function
                          of temperature, hour of the  day,  and climatological
m                        variables (yr/hr).
_          The three algorithms  are evaluated for days  of  three  different daily mean
•          temperatures having diurnal  temperature ranges  typical  of New  York  City.
tf          The three daily mean  temperatures  considered are  70ฐ, 50ฐ, and 30ฐF.  The
            variable K.'s,  plotted  against hour of the day, are compared for each of
I          the three algorithms  and for each  of the three  types  of days.
                      For the St. Louis  experiment, detailed  information was available
™          for the area source emissions. The complex  algorithm included the  contri-
•          butions from residential, commercial, river  vessel, automotive, railroad,
            backyard burning, and industrial sources.  For  comparison purposes, only
•          two components  of the algorithm are examined here - the residential and
            commercial  sources.  These two components  are based primarily  on the  study
V          by Turner (1968)  of the fuel usage for space heating  in St. Louis during
f          a winter season.   Turner developed methods to determine the rate of fuel
            use from residential  and commercial space  heating sources as a function
•          of temperature, hour  of the  day, and day of  the week.
                      The residential component of the algorithm  is composed of a con-
•          stant Base Residential  S02 Emission Rate,  and a Residential Heating S02
            Emission Rate which is  a function  of temperature  and  time of day
I
•                          r         FRDW               Dr(t)  (1  -  FR)     i
ง                 QR " QRT [24(365 D^ + D(1-FR))  +  24(365  DWFR  +  D(tl-FR))J

1
I
1                                            A~12
I
                                                                                   (A-6)

-------
I

             where
*                    QD = residential SO, emission rate (Ib/hr)
                       K                 c
•                   QRT = annual residential SOg emission (Ib)
                      FR = 0.2 = summer day fuel consumption as fraction of average
m                         winter day fuel consumption
                      DW = average winter degree day (= 32. 9ฐ F day/day)
|                 Dr(t) - 65 - T(t) - Ar(t) (ฐF)
                    T(t) = temperature at time t (ฐF)
                   Ar(t) = residential correction factor (Table A-2)
 I
 _                      D  =  annual  degree  days  (=  4977  degree  days).

             The commercial  component of  the St.  Louis algorithm is  composed of a Base
 •           Commercial S0ซ Emission  Rate which is  a function of hour  of the day, and
 •           a Commercial Heating  S02 Emission Rate which  is  a function of temperature
             and hour  of  the day.

                             f      FSDWFc(t)               Dc(t)  (FW  '  FS}      1
 _                 0=0           ^ ฃ-       —., -  - - +	I- „ - - K-. ฃ  ,. -n.        (A-7}
 •                 WC   wct  24(365 DUF<.  +  D(FU-FซJ)    24(365 DUF- + D(FU-F,J )\      VM /;
^M                           L        Wo       Wo              Wo       WoJ

fl           where
m                     Qc =  commercial  S02 emission  rate (Ib/hr)
                     Q . =  annual  commercial S02  emission  (Ib)
I                     F_ =  fraction of annual  fuel  consumption used  in  summer
                       FW =  fraction of annual  fuel  consumption  used  in  winter
*                 Fc(t) =  commercial  diurnal  variation factor (Table A-2)
*                 Dc(t) =  65 - T(t)  - Ac(t) (ฐF)
                   Ac(t) =  commercial  correction  factor (Table  A-2).

ป                                             A-13

I

-------
   I
   I
   t
   I
   I
   I
   I
  I
  I
  I
  I
 I
 I
 I
 I
 I
 I
 I
I
Table A-2.  Numerical Values for Constants in the St. Louis Algorithm
Hour
Ending
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
IS
16
17
18
19
20
21
22
23
Ac(t)
(Weekday)
13.19
13.32
13.23
12.54
10.43
5.64
-1.75
-8.04
-11.69
-13.91
-12. 94
-12.43
-12. 53
-12.39
-11.19
-9.62
-7.88
-4.37
0.56
4.55
6.62
9.08
10.41
11.53
Ar(t)
(Weekday)
9.11
11.11
10.61
9.69
8.54
7.08
3.13
-2.15
-7.32
-7.61
-8.85
-8.44
-7.46
-6.73
-6.25
-5.11
-4.08
-3.17
-2.41
-0.77
-0.01
2.56
3.22
5.33
Fc(t)
0.20

0.20

0.20

0.20

1.96

1.82

1.75

1.69

1.62

1.48

0.68

0.20

                            A-14

-------
Values for F~ and FW are determined as follows:
          Twenty percent of the annual commercial fuel usage  is  attributed
to hot water requirements and is distributed uniformly throughout  the year.
Therefore, F,,,,, the fuel required for hot water in a  three-month season,
is defined as
                      FHW = 0.2  (1/4 years) = 0.05.
Assuming that the summer season requires no space heating
The remaining 80 percent of the fuel usage is used for space heating; the
amount of this used in the winter season is proportional to the ratio of
the number of degree days in the winter to the total annual degree days,
so
                                           D
                          FW = 0.05 + 0.80
where
         DW = degree days for the winter season = 2886, therefore,
                    Fw = 0.05 + 0.80 () = 0.514.
                                  A-15

-------
I
•                    For the  Chicago  study, an estimate of the annual  S02 emission
m          rate was obtained  for each of  three classes of emitters.   These include:
_                    •    Low-rise  residential structures consisting of 19 or less
•                         dwelling  units  (Class I)
f
•
                          High-rise residential  structures consisting of 20 or
                          more dwelling units, and  commercial and institutional
                          buildings (Class  II)
                          Industrial  plants  not  large enough to be treated as
                      It    industrial plants not  large enougn to
                           individual point emitters (Class III).

•          In addition to the  annual emission rates, estimates regarding diurnal
            variations in  emission  rates were generated in Argonne's extensive  study
|          of Chicago emission data  (Roberts et al. 1970, Chamot et al.  1970).
            Algorithms for estimating diurnal variations in emission rates are  given
            by Equations  (A-8) , (A-9), and (A-10).  In the equations below, Turner's
            temperature correction  factors have been incorporated into the Argonne
            algorithms as  proposed  by Koch and Thayer (1970).

                      1.   Residential or commercial low-rise (Class I)

'                           f FH    (1 - FJ (65 - T - AR) U.]
-                      QI  = \ms + - "-W. - — J QIT               (A-8>
                      2.   Residential or commercial high-rise (Class II)
                               f F     (1 - FH) (65 - T - Ajl
                           Q2  - [urn + - -vฎ - -J
                      3.   Industrial  (Class III)

 1
 •                                          A-16
 I

-------
I
I
            where
 j                    Q,  =  emission  rate  of  SCL  for  Class  I emitters  (Ib/hr)
                      Q2  =  emission  rate  of  S(L  for  Class  II emitters  (Ib/hr)
 •                    Q3  =  emission  rate  of  S02  for  Class  III emitters  (Ib/hr)
 f                   Q,y  =  annual  SIX,  emission for Class I emitters (Ib/yr)
                     Guy  ~  annual  S0?  emission for Class II emitters  (Ib/yr)
 •                   Q3j  =  annual  SCL  emission for Class III emitters  (Ib/yr)
                      FH  =  0.2  =  fraction  of annual  residential/commercial fuel
                           usage  attributed to hot water requirements
I
                     U. = fraction average hourly fuel usage associated with ith
•                        hour
                     H.  =  hours of  fuel  usage per day  (hr/day)
                       •j
ง                   Ui =  1.5  (T  < 5ฐF,  hours 4 and 5),(5ฐF < T < 65ฐF, hours 6 and 7)
I
                     U. = 1.0 (T < 5ฐF, hours 6 to 22),(5ฐF < T < 65ฐF, hours 8 to 22)
                      1          """
                     U. = 0 (T < 5ฐF, hours 0 to 3 and 23),(5ฐF < T < 65ฐF, hours 0
                      1   to 5 and 23),(T >_ 65ฐF, all  hours)
•                   H. =  19  (T <_ 5ฐF)
                      J
I
                     H. = 17 (T > 5ฐF).
                      J
—          In Equations  (A-8) and  (A-9) , 20 percent of the annual emissions are attrib-
*          uted to hot water requirements and distributed evenly over the year.  The
•          remaining emissions for Classes I and II are attributed to space heating
            requirements  and are allocated on the basis of outside air temperature
f          deficit below 65ฐF.  Only the first term is applicable in these equations
            when the outside air temperature is over 65ฐF.  Equation (A-8) includes a
ป          "janitor" factor U  to account for "hold fire" periods after 10 p.m. and
I
•                                            A-17
I

-------
for a 50 percent increase in the burn rate during the first two early
morning start-up hours (starting at 4 a.m. when temperature is <_ 5ฐF
and 6 a.m. otherwise).
          The three sets of algorithms represented by equation sets (A-4),
(A-6 and A-7), and (A-8, A-9, and A-10) and referred to as the "New York
City," "St. Louis," and "Chicago" algorithms respectively, were evaluated
for New York City climate on three days of different mean temperature.
For the purpose of comparison, the F,, factor in equation set (A-4) was
assigned a value of 0.8.  Values of K. in Equation (A-5) for each of the
three algorithms are presented in Figures A-4 through A-6.  Numbers in
parentheses refer to 24-hour summations of K.. values.  The two New York
City K.J values may be added together to give the total K. representing
both base level and space heating emissions.  The curves of K. for St. Louis
and Chicago cannot be added together since the K.. 's are coefficients of
different emission rates (QT).  The general  shapes of the K.  profiles are
similar for the three algorithms, each having a minimum value during the
pre-dawn hours, and a maximum value between  the hours of 6 and 10 a.m.
The New York City algorithm is closely approximated by the Chicago (Argonne)
Residential or Commercial Low-Rise Profile.   The space heating profile
for New York City has a second, weaker maximum in the K.. profile in the
early evening hours.
                                   A-18

-------
0   2    4   6   8    10   12   14   16   18  20  22  24





„
1
•r-t
*
tf
(U
'o
O
O
fi
.2
CO
CO
1





Temperature ,



.04
.03
.02
.01
.00
.04
.03

.02
.01
.00

.04
.03
.02
.01
.00
20
30
40
50
60
70
80























	

























• ^






































iฃ~





.*••*






~~~~*-











s
r?





—





-x^






x-




~ c
tf




1^
>
^** *i












. hia




)mm
*.esi<





- R
• ป....






^^^^





>e Le




;rcia
enti;




Indu
aside






^•••M





/el (i




(O.C
1(0.




trial
ntial






— — •





1. U55




28)
JOS)




U. ฃ/
or Cc






. — -





)










4)
mine





J^—


















rcial





— —

Ml





SI





a




High







;w Y





LOU





IICA<




Rise







)RK





(S





X>




(0.0







CIT*
















8)

























































                 8   10   12   14   16   18  20   22  24
                       Hour
        Figure A-4.  Area Source Algorithms for Mean Daily Temperature of 70 F

                                        A-19

-------
                                                            (E.S.T.)
0246
              10   12   14   16   18   20  22   24
.04
.03
.02
.01
ฃ .00
>.
•ft
^
1 -04
ฐo
ฃ .03
As
s*





• — -~







\.,






*- -











"^








Mll^





-~rปป



^/
t ^*






^*







• -^. .





•"=



^ R'
^S5S






- — .







—- '





_x




isidei








•• —






.^"





/^
^




tia.1
•^•^•H






—






/
/-





Zonii

X._



)T Co
L/C.
*





ป- •**





5pac
— .
/




aerci
/
•ta^^



nme
Resi
•^

	



r^--






: Hea

'\




al(0
Resi<
^



cial
lenti

*C











ting
iase




243)
enti:




,ow !
ilor
~Inc







NEW


0.22
,eve]

ST.



1(0.

CHK


Use(
Zomt
'O. ?-:
ustri;







YOR


>)
(0.0

.GUI



189)

:AGC


0.29
tierci
K)
1(0.







KCI



55)

)








')
a.1 Hi
>74)







7Y















;hRi
























ie








              6    8    10   12  14   16   18   20   22  24
                            Hour
                                                            (E. S. T.)
Figure A-5.  Area Source Algorithms for Mean Daily Temperature of 50*F
                              A-20

-------
0   2    4    6   8    10   12   14  16   18  20   22  24
.04
.03
.02
.01
Emission Coefficient, K. (yr/hr)
DOOO OOOOO C
•ป N UJ iP O I-* tO OJ S C
.00
PH 20
O
8 30
3
S 40
4)
H 50
60
70
80





	




i^-S





	 	


_ •










_^->*




—





^^^f


—







f
/
f
/
/
f




/
L^




xx










A
l





s
/
^


A
\

/



— i— i







%
\
~\




-""•^




^^•^KH
*^ "^ *




*•" ••














	 ^



^\
^




=ป-ซ







	 ,





— —•ป.




' 	 ,





	







*





- 	 ^




	 ^
•*ป..




"• 1 '•












/"
">|



J
^
^>V




	






^-





omul
>ปป



^ I
-
-. /
*V. _


	 	 -^






SN>
\%

^"


ercia
Lesid



esidt
— ~^


— --


•






^ 5

\
V
^^^~^


L(O.;
intia
-*v


ntial

- R
— :


•K=






pace


- B


164)
(0.4



or C

eside
Hi?
** ]







NEW

He a


ise L
ST.


41)


CHIC
>mm

ntial
hRi
ndust








YOB

ing(


vel
jOUI





:AGC
:rcia
(n

orC<
e(0.
rial(








KCI

). 59.


0.05







Low
fi4.8\

>mm<
569)
127'








7Y

)


5)







Rise

rcia!
)

































0246
8   10   12  14   16   18   20  22   24
          Hour
       Figure A-6.  Area Source Algorithms for Mean Daily Temperature of 30ฐF
                                     A-21

-------
                     Appendix B



METHOD OF ESTIMATING THE HEIGHT  OF  THE  MIXING  LAYER

-------
                               Appendix B
          METHOD OF ESTIMATING THE HEIGHT OF THE MIXING LAYER

          Mixing layer heights were estimated for New York City using
radiosonde data collected at Kennedy Airport in New York City.   The data
were obtained from the NOAA National Weather Records Center in  Asheville,
North Carolina on two magnetic tapes.  One tape contained standard level
data (height, temperature, and relative humidity for 50 mb intervals
starting with 1000 mb).   The other tape contained significant level  data
(height, pressure, temperature, and relative humidity of significant
points identified by the radiosonde observers).  These two data sets were
merged to form a single  chronological file of height-ordered measurements.
The following seven steps were used to determine the mixing layer  height
for each observation time.
          1.  Read and store the height,  pressure,  temperature,  and
relative humidity of each data level.
          2.  Convert all relative humidities  to mixing ratios  using
the following equations (Saucier 1955):
                              M = 0.01  U  S                               (B-l)

                                  J— ซ0.622 |                        (B-2)

                                       7.5  T
                         E = 6.11  (10)T + 237'3                          (B-3)
                                   B-l

-------
where
          M = mixing ratio
          U = reported relative humidity (percent)
          S = saturation mixing ratio
      f ซ 1 = correction factor for departure from ideal gas laws
          E = saturation vapor pressure of water (mb)
          P = reported pressure (mb)
          T = reported temperature (ฐC).
          3.  Find the maximum mixing ratio for the observation time (M ).
                                                                       A
          4.  Find the mixing condensation level by the following equa-
tions (Saucier 1955):
                          Z  =
                             _ 1000
                               "O
ro -  Do>
(B-4)
where
         Z  = mixing condensation level (m)
         T  = ground level temperature (ฐC)
         D  = ground level dewpoint (ฐC).

In order to account for evaporation of dew during the early morning, it
is assumed that the mixed atmosphere will  contain moisture equal to that
indicated by M .   D  is determined from M  by means of Equations B-2 and
B-3 using S = XM  and T = D :             x
                X          0
                     n  - 	'U it). I I  lU.O^JJ                    (n c\
                     un               f   R~T     n                    \ฐ-s)
237.3 log1Q
7.5 - log1Q
" M P
X 0
6.11 (0.622)
f MxPo '
[6.11 (0.622)_
                                   B-2

-------
where
         P  = ground level pressure.


          5.  Using the reported data levels to define layers, find the
layer (Zj_i to Z-j) containing the top of the mixing layer.   The top of
the mixing layer is identified by the parcel method.   When  the reported
vertical temperature profile exceeds the temperature  of a parcel  lifted
from the surface by 1ฐC, this is assumed to be the top of the mixing
layer (Zm).  The layer containing the mixing layer height is identified
by testing if


                              T. iTr + 1                                (B-6)


where


          T. = temperature of ith level (ฐC)

          J: = temperature of parcel lifted to ith level  (ฐC).


The parcel temperature is calculated as follows:


                       T: = T.^ + Y' (Z.  - Z._.,)                        (B-6)
-u.uuyo,
_n nnQfl

L. <_ L
1 ' Rd (Ti-l + 273)
L^ Si-l
h c R, (T.', + 273) z
p v i-l
                                                                        (B-7)
                                                 '  Zl  " ZG
where
          ' = temperature lapse rate (ฐC/m, Haltiner and Martin  1957)

          L = 2500 = latent heat of vaporization (joules/g)
                                   B-3

-------
       S.", = saturation mixing ratio of parcel  lifted to (i-l)th  level,
              estimated from Equations B-2 and B-3 using P.  -j  and  T.',

         Rd = 0.287 = gas constant for dry air (joule/g/ฐC)

         c  = 1.003 = specific heat of dry air at constant pressure
          p   (joule/g/ฐC)

         R  = 0.461 = gas constant for water vapor (joule/g/ฐC).


          6.  Estimate the height of the mixing  layer  by linear  interpo-
lation as follows:
            7    Z    ,          -
            Zm-Zi-l        IT,  - Tp  -  (T.,,
          7.   Enter Z  on the output data  file
                                  B-4

-------
                             Appendix C

DEVELOPMENT OF A METHOD OF ESTIMATING EMISSION RATES  FOR USE IN THE
                   SIMPLIFIED GIFFORD-HANNA MODEL

-------
                               Appendix C
  DEVELOPMENT OF A METHOD OF ESTIMATING EMISSION RATES FOR USE IN THE
                     SIMPLIFIED GIFFORD-HANNA MODEL
          The mean area source emission rate used in the Gifford-Hanna
model  (GHM) is to be a mean value over n by n 1 km grid cells centered
on the grid cell containing the receptor.  An appropriate value for n
is required.
          On the average, the New York City receptors considered in this
study  are located in areas of relatively high area source emissions of
both S02 and particulates.  The mean SOp emission decreases steadily from
                                         2
an average maximum value of 0.747 tons/km /year for the 1 km grid cell con-
                                     2
taining the receptor to 0.319 tons/km /year for a square with sides of
79 km  (maximum size considered).  Likewise, for particulates a value of
             2
0.136  tons/km /year is obtained for the average 1 x 1 km square, and a
                      2
value  of 0.059 tons/km /year for the 79 x 79 km averaging area.   Clearly
then the mean value of the calculated concentrations will vary by a factor
of 2 depending on the choice of the averaging area.
          In a forthcoming paper, Turner et al. (1972)  in applying GHM
used a 10 km radius for determining (p for both S0ซ and particulate emis-
sions.  They considered circles of 3, 5, 10, 20, 30, and 40 km radii
                                                             2
centered over the receptor location.  If the center of a 1  km  grid cell
was within the circle, it was included in the average;  if the center was
outside, it was not included.  Their selection  was based on the  best
linear correlation of the measured annual  concentration at  the receptor
with the mean annual  emissions over the different sized circles.
                                   C-l

-------
          The GEOMET analysis based on squares instead of circles confirms
Turner's findings.  Forty different squares were tested, ranging from
1 x 1 km to 79 x 79 km.
          A maximum correlation of 0.84 was obtained between SCL concen-
tration measurements and a 31 x 31 km square emission averaging area; for
particulate concentration measurements, a maximum correlation of 0.65 was
obtained for all squares with sides ranging from 15 to 57 km.
          A second analysis of the area source emission data was made
after dividing the receptors into subsets of low and high measured con-
                                    3                    3
centration.  Mean values of 135 yg/m  for SO^ and 82 yg/m  for particu-
lates were used to divide the sample.  The correlations between measured
concentrations and Q* as a function of the size of the averaging square
were markedly different for the two data subsets.  The correlation coef-
ficients and the Q* values for 1 to 73 km squares are plotted in Figures C-l
and C-2, for S02 and particulates, respectively.  Correlations for those
receptors measuring large concentrations were maximum for small averaging
areas based on emissions near the receptor and fell off rapidly with
increasing size of the averaging area.  The receptors with low concentra-
tions, however, tended to have a maximum correlation with emission means
computed from large-sized squares.  Also, the cases of low concentration
had mean emission values which either increased or remained constant with
increasing area size.   These results indicate that large measured concen-
trations are primarily due to nearby sources, and that low measured
concentrations are affected by sources at large distances from the recep-
tor.  It might also be hypothesized that the pattern of correlation
coefficients shown for the divided sets is associated with selecting
                                   C-2

-------
I
      Correlation
            r
          Mean
          Annual
          Area
          Source
     (Tons/km2/day)
I-U
.9
.8
.7
.6
.5
• 4
• 3
-2
.1
0-0
1.2
1.1
1.0
.9
.8
.7
.6
.5
.4
.3
.2
.1
0.0


\^
•
/







^'N




X"





x' -



X
/"
^ •







V
\



>v/









- — -
\










\


TOT
X..







^ — •

-V
\









HI(
/
\


^L :
— • — .



,L01



'

— — .
'~










\
"X

;AMP
"ซป.



- . -





_ •
— " 	






. — —

-'
^"—






LC
—
—





— . —
\-.

S




Key : Receptoi
LOW <135
HIGH >135



^

LE
-™**^



. • -





\





• • — *"









• 	 .


• —







\

— "ซ-,


• -



. 	 /
• 	

:GH
•v
X.




^'i'U'J
\,
*


\



'AL
^^
~~


^_



fciAMI:
^^
~~~ *





: Concentration
pg/m3
yg/m3







^

• -







^

• — ^

• ~"







^

— — .










^
-^ป.





LE

"^^



^











^
^9M.






^^,
"~"


^^













*-~~.







. \















--
—







>^^ •



--.











-^
—— 	







	 -.



-s.












-~-~.



1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 72
I.U
.9
.8
.7
.6
.5
•4
.3
.2
.1
0.0
12
1.1
1.0
.9
.8
.7
.6
.5
.4
.3
.2
.1
0.0
                                     Side  of averaging square(km)

                                   Figure C-l. Mean SO Area Source Emissions
                                                  C-3

-------
I
I
I
I
I
I
I
I
       Correlation
    Mean
    Annual
    Area
    Source
(Tons/km2/day)
I.U
.9
.8
.7
.6
i -5
.4
.3
.2
.1
0.0
.20
.1 8
.16
.14
.12
.10
.08
.06
.04
.02
0.00
1




/
^
/







^-\



\


\
•
\






— --'











\



^-


"*" • __









\









"^


^-^_

/ LC











"~ —









-.


r
__/_

w









._

^^









/


COTA












. -

71
"**•






• -

GH
— -

Key:





_ /


- —

Rec
LOW
HIGH


met
^~

L S^









I
^.

iMPL
~^~









-
E
1 	 _








/

LOW





eptc
<82
>82




" -.

1 	 — — .







TOT








)r C


^L ฃ


•^ • -





once


AMP


. . -





ntr
yg/m3
yg/m3






	 .










"^











-

— . — .






LE


• •- .


- -

atio






	

•^~ m










	


— -


n

























• —












^














- —












ป•









	 ~N



~x












ซ*.










^-



\












^
	 -




5 f 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 7,
Side of averaging square (km)
Figure C-2, Mean Particulate Area Source Emissions
IJU
.9
.8
.7
.6
.5
.4
.3
.2
.1
0.0
.20
.18
.16
.14
.12
.10
.08
.06
.04
.02
0.00
                                               C-4

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
averaging square sizes which yield the highest  values of mean emission


rates.


           Based on the preceding analysis,  the  following five different


averaging routines for the  area source emissions  were defined and  tested:


                                                                             2
           1.   Use only the  single area source emission value for the 1 km
grid  cell  containing the  receptor (designated 1 x 1  in Tables C-l  and C-2).


           2.   Use an averaging area approximating the 10 km radius used by
Turner,  i.e., 19 x 19 km  square centered on the grid cell containing the
receptor (designated 19 x 19 in Tables C-l  and  C-2).


           3.   Use the area  for which the highest  linear correlation was
obtained,  i.e., 31 x 31 km  for S0? and 19 x 19  km for particulates (desig-
nated 31  x 31 in Table C-l  and 19 x 19 in Table C-2).


           4.   Search increasingly large squares until the maximum  area
source emission value is  obtained (designated Q*  in  Tables C-l and C-2).


           5.   Use the following Q* magnitude and  pollutant oriented cri-
teria (designated Q* in Tables C-l and C-2):



    Pollutant                         Criteria                      Averaging Square
                                2                      2
     SO       (a) Q* >0. 701 ton/km /day and Q* >0. 512 ton/km /day     (a)  9x 9km
       ฃ            y                     j /
                                2                      2
              (b) Q* <0.701 ton/km /day and Q* <0. 512 ton/km /day     (b) 37 x 37 km

              (c) Neither (a) nor (b)                                (c) 31 x 31 km
                                2                      2
  Particulates   (a) Q* >0.128 ton/km /day and Q* > 0. 082 ton/km /day     (a)  3x 3km
                   3             2        28            2
              (b) Q*<0.128 ton/km /day and Q* < 0. 082 ton/km /day     (b) 39 x 39 km
                   3                     28
              (c) Neither (a) nor (b)                                (c) 19 x 19 km
  (1)                                2
     Q* is the average emission rate from the s  grid squares whose center grid square contains the
     receptor (s must be an odd integer).



Each of the five above  procedures has been evaluated by comparing GHM cal-


culations  with corresponding  measured values.  For the  particulates there


would only be four procedures since the procedures (2)  and (3) above  yield
                                     C-5

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
 I
 I
I
Table C-l. SO Experiments
Statistic
Number Cases
Mean Measured
Mean Calculated
RMSE
Mean Error
Mean Absolute Error
Largest Negative Error
Largest Positive Error
Error Range
Least Error
Linear Correlation
Reduction of Variance
Slope, Least Squares
Intercept, Least Squares
Max. Measurement
Error at Max. Measurement
Skewness
Kurtosis
Frequency Distribution of
Calculated Minus Measured
By Percentiles
100.0
99.5
99.0
Experiment
1 x 1
68
142.31
75.62
98.69
-66.69
84.80
-199.91
174.76
374.67
-4.07
0.733
0.537
0.539
101.58
385.00
146. 55
0.96
1.59

174. 76
174.76
174. 76
19 x 19
68
142.31
65.06
90.13
-77.25
78.63
-208.29
46.71
255.00
-9.77
0.817
0.667
1.186
65.14
385.00
-208.29
-0.36
0.02

46.71
46.71
46.71
31 x 31
68
142.31
56.43
100.45
-85.88
86.40
-273.60
17.76
291.36
-16.25
0.831
0.690
1.766
42.63
385.00
-273.60
-0.84
1.07

17.76
17.76
17.76
<&
68
142.31
96.99
90.13
-45.32
70.09
-171.57
350.41
521.98
-4.60
0.738
0.545
0.504
93.44
385.00
95.36
2.31
9.28

350.41
350.41
350.41

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Table C-l.  SO Experiments (Concluded)
Statistic
Frequency Distribution of
Calculated Minus Measured








By Percentiles
95.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
5.0
1.0
0.5
0.0
Experiment
1 x 1


100. 47
9.61
-24. 99
-38.11
-46. 93
-61.65
-92. 33
-111.38
-122.15
-151.52
-169.96
-199.91
-199.91
-199.91
19 x 19


-19.27
-26. 69
-37. 15
-46. 06
-53. 83
-65. 30
-85. 18
-103.42
-119.90
-138. 50
-160.95
-208. 29
-208. 29
-208. 29
31 x 31


-21.51
-26. 22
-37. 47
-51.13
-65. 75
-78.13
-97. 74
-109. 36
-122.83
-156.51
-186. 22
-273. 60
-273. 60
-273. 60
Q*A


74.98
21. 71
-20. 30
-31.39
-38. 84
-42. 68
-55. 00
-77. 61
-107.24
-123.48
-146. 90
-171.57
-171.57
-171.57
Q*
VB


-4.02
-14.36
-27. 36
-37. 36
-46. 97
-56.04
-70. 19
-85. 93
-105. 29
-122.83
-135.25
-1 70. 71
-170.71
-1 70. 71
                C-7

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Table C-2.  Particulate Experiments
Statistic
Number Cases
Mean Measured
Mean Calculated
RMSE
Mean Error
Mean Absolute Error
Largest Negative Error
Largest Positive Error
Error Range
Least Error
Linear Correlation
Reduction of Variance
Slope, Least Squares
Intercept, Least Squares
Max. Measurement
Error at Max. Measurement
Skewness
Kurtosis
Frequency Distribution of
Calculated Minus Measured
By Percentiles
100.0
99.5
99.0
Experiment
1 x 1
101
83.73
97.07
73.37
13.34
47.04
-81.44
319.38
400. 83
-0.33
0.486
0.236
0.138
70.30
169. 00
148. 48
1.99
4.62

319.38
319.38
300. 60
19 x 19
101
83.73
83.46
33.44
-0.27
27.53
-79. 67
75.35
155.02
-0.16
0.632
0.399
0.338
55.52
169.00
-10.68
0.37
-0.37

75.35
75.35
71.52
31 x 31
101
83.73
78.00
26. 77
-5.73
22.14
-72. 41
41.27
113.68
0.07
0.603
0.364
0.425
50.61
169. 00
-47. 63
-0.12
-0.65

41.27
41.27
40.06
QA
101
83.73
105. 21
66.65
21.48
42.65
-66. 49
324. 85
391.34
1.54
0.640
0.410
0.195
63.24
169.00
150.00
2.14
6.45

324. 85
324. 85
281.42
Qฃ
101
83.73
97.34
63.23
13.60
38.55
-66. 49
324. 85
391.34
0.25
0.642
0.412
0.199
64.38
169. 00
150.00
2.51
8.33

324. 85
324. 85
281.42
                                                   (Continued)
              C-8

-------
I
I
I
I
a
i
i
i
i
i
i
i
i
i
i
f
i
i
i
Table C-2.  Particulate Experiments (Concluded)

Statistic
Frequency Distribution of
Calculated Minus Measured















By Percentiles
95.0
90.0
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
5.0
1.0
0.5
0.0
Experiment
1 x 1



148. 48
111.91
42.79
16.67
3.56
-7.99
-19.22
-24. 89
-31.89
-42. 44
-64. 44
-76. 44
-81. 44
-81.44
19 x 19



63.90
54.50
24.01
16.39
5.22
-7.42
-14. 01
-21.55
-26. 40
-36. 38
-40. 32
-65. 25
-79. 67
-79.67
31 x 31



35.64
30.70
22.37
8.45
1.32
-6.70
-13.31
-19. 95
-26.14
-37. 68
-47. 63
-59. 94
-72. 41
-72. 41
-



129. 57
95.09
54.80
37.31
14. 70
5.41
-7.89
-16.16
-23. 50
-32. 31
-36. 38
-59. 40
-66. 49
-66. 49
— _—



122.85
80.70
38.88
18.84
7.92
-5.79
-12.00
-18.09
-25.18
-35.13
-42. 40
-59. 40
-66. 49
-66. 49
                   C-9

-------
I

I
             the  same  averaging  area  of  19  x  19  km;  however,  a  31  x  31  km  area was  added
I           for  comparison  purposes.

•           C.I   EVALUATION USING  ANNUAL MEAN EMISSION  RATES AND  WIND  SPEEDS
                                                        3
                      The annual concentrations  in  yg/m  calculated with  the GHM using
fj           annual  mean  emissions  and wind speed for  each  of the  five  averaging pro-
—           cedures are  listed  in  Tables C-l and C-2  for  S02 and  particulates, respec-
*           tively.   The analysis  was limited to receptors in  areas in which detailed
ft           1  x  1 km  area source emission  data were available,  including  68 S(L receptors
             and  101 particulate receptors.  The  annual  mean  wind  speed for LaGuardia
jj           for  1969  is  5.1852  m/sec.   Statistics comparing  the calculations with
             measured  values are presented  in Tables C-l and  C-2.
™                    Of the  five  S02 experiments,  the  use of  the 19 x 19 km averaging
•           square  and the  Qt and  QF routine produce  the better results.  The large
             31 x  31 km averaging area tends to reduce the magnitude of the calculated
I           mean  emission values;  it therefore leads  to underprediction of the S0ซ
             concentrations.
•                    The results  of the particulate  experiments  show that the
•           19 x  19 km and  the  31  x 31  km  averaging areas produce the better calcu-
             lated concentrations.  The  remaining three  experiments, in addition to
I           having  large errors, overpredict the concentration.  The skill of these
             three procedures  is further reduced  when  the contribution from the point
•           sources are  added to the calculated  values.

I

I

I                                             C-10

I

-------
C.2  EVALUATION USING HOURLY ESTIMATES OF EMISSION RATES AND WIND SPEED
          The Gifford-Hanna model was used to calculate hourly SCL con-
centrations using emission rates multiplied by the K- values defined by
Equations A-4 and A-5 in Appendix A.
          Three-hourly wind speed and temperature observations for
LaGuardia Airport were used to evaluate K. and to define the wind speed.
Calculations were made for 10 stations located within the New York City
area (listed in Table 5 of the main body of the report) using each of
the five previously described procedures for estimating annual area source
emission values.  Verification statistics were calculated separately for
each of the five averaging procedures.
          Comparisons are presented both with and without the concentra-
tions from point sources, as calculated by SCIM using hourly varying
mixing layer height and atmospheric stability (see main body of report
for a description of this model), added to the Gifford-Hanna area source
calculations.
          The results of these calculations are shown in Tables C-3 through
C-22.  Tables C-3 through C-7 present results for 1-hour SCL concentrations
without point sources; Tables C-8 through C-12, for 24-hour concentrations
without point sources; Tables C-13 through C-17,  for 1-hour concentrations
with point sources; and Tables C-18 through C-22,  for 24-hour concentra-
tions with point sources.
          A few comments can be made on the calculations using both area
and point source emissions (Tables C-13 through C-22).

          •    Averaging procedure QJ generally produces calculations
               closest to measured values.

                                   C-ll

-------
o
oo
•3
o
I — I
oj
U
8
m


U
,0
n)
H










c
o
3
4->
LO











i— 1
r-H
<

m

m
00
04
fv.
(M

i-l

T-H
o
TH






Statistic

•^
t-*
8
i— i

m
t-.
•*
TH

ro
R
T-I

TH
oo
CO
TH

CM
O
Cx
TH

00
s
TH

TH
"tf
1C
T-H

8
to
TH

CTl
00
•*
TH

rO
TH
CM
TH

ra
CTI
00
CM
Number Cases
to
co
is
ro
M
tx
O
in
•*
TH
CTl
CO
tx
00
TH
TH
CM
IX
CM
TH
tO
m
ix
00
r-H
CM
00
CM
CTi
TH
to
in
CTl
CM
CM
tx
00
CTl
TH
•*
CTI
00
in
CTI
CM
00
CO
TH
CM
tx
Cx
8
ro
Mean Measured
ro
rO
3
CM
s
to
TH
CTl
TH
in
TH
O
tx
**
00
tx
tx
•*
TH
00
tx
IX
00
CM
CTI
00
CTI
TH
tx
TH
to
o
TH
a
o
CM
m
in
O
in
TH
00
ro
TH
•*
TH
CO
•*
00
in
yH
in
ro
o
00
TH
fx
TH
a\
•*
1/5
00
CO
00
CM
t^
^<
TH
3
CM
CM
fft
TH
•*
h
w
2
rt
S
0"
t/l
i
rt
O
2
i
8
A
S
TH
TH
CM
o
Oi
CM
TH
1
CTi
TH
CM
t^.
TH
TH
tn
CM
CM
TH
00
f^
CT>
ro
s
3
T-i
1
in
vo
o
ro
TH
1
•*
00
TH
TH
CM
in
TH
ya
vo
TH
CO
TH
TH
CM
TH
m
in
CM
e
1
9
tt)
2
00
o>
c?
TH
8
a
TH
00
M
CM
f~
TH
m
in
CM
CM
TH
CM
m
8
r-H
8
01
T-4
TH
TH
o
T-t
•*
TH
rJ
VO
•*
TH
ro
•*
CM
Ol
CTi
TH
8
00
(O
TH
t>
00
8
CM
Mean Absolute Error
TH
m
CM
CM
<ฃ>
1
00
^o
R
00
1
Ol
t^
s
T-t
1
s
CM
CO
00
1
•*
in
CO
o>
i
r<3
O
s
8
CO
ง
TH
1
TH
in
CM
CM
<0
$
o
TH
•*
TH
1
8
CTl
Ol
s
CD
TH
t^
1
t4
Id
0

•*
CM
(M
TH
00
•*
O
IQ
m
TH
t^
00
t^
t^
00
CM
CTl
ro
8
ro
t^
CO
in
CO
CM
58
CM
a
0)
BO
s
OS
Id
TH
o
o
TH
TH
o
8
TH
1
o
ro
CO
ro
O
O
CM
O

3
0
Linear Correlation
in
CM
CM
o
ง
TH
o
ro
3
o

in
|
O
ro
TH
in
o
S
to
o
to
CM
CM
O
in
ro
TH
o
8
0
to
CM
O
S
a1
00
8
3
tK

1
00
to
s
00
1
CTl
tx
TH
s
rH
1
S
CM
ro
00
•*
in
ro
CO
CTl
s
s
Is
s
ro
S
TH
1
TH
in
oo
CM
CM
to
S
o
T-I
Trf*
TH
8
CM
CTl
CTl
•*
CM
TH
m
to
Error at Max. Measurement
to
TH
TH
TH
tx
TH
CTl
00
TH
1
8
CM
1
S3
o
CM
CM
TH
1
tx
TH
TH
o
in
TH
s
o
ft
o
s
CM
Skewness of Error Dist.
CTl
CM
S
CO
oo
in

-------
a
o
*4_>
rt



rH
rt
O
OJ
a
O
 i
O
,0
rt
H










g
%
-t->
LO











i-H
i — 1
<
^O
CO

CO
00
CM
rx
CM

i—t

i-H
O
TH




o

Statistic

^
TH
R
TH

in
tx
^F
TH

ro
00
tx
TH

TH
00
ro
TH

CM
O
tx
rH

00
VO
r-
t— i

rH
•*
to
rH

O
t^
10
rH

CTi
00
TJ<
rH

ro
rH
tM
rH

ro
CTI
00
(M
Number Cases
U3
ro
ro
CM
&
3
rH
CTl
rO
00
TH
TH
CM
CM
TH
s
00
TH
CM
00
CM
CTI
TH
VO
1O
CTi
CM
0)
IX
00
TH
•*
CTi
00
in
CTI
CM
00
ro
CM
TH
CM
t^
tx
ro
O
ro
Mean Measured
s
CM
TH
Tf
CM
•*
8

in
rH
TH
1
ro
oo
O^
1
ro
ro
CM
TH
1
ro
TH
O
TH
1
TH
CM
K
CTl
TH
TH
t^
1
00

if>
ง
i-H
I
(H
9
ฃ>
i
CM
•*
ro
oo
i
s
s
TH
1
fx
o^
TH
ฐ?
m
m
00
CTl
i
CO
TH
O
00
[•X
ro
TH
to
CM
TH
TH
(O
ro
in
in
.

VO
O
8
CO
O
i
o
in
R
o"
ta
V
%
$
C/5
1
CO
_1
uT
p<
3
in
%
VO
ro
rH
in
in
CJ>
00
CO
CM
TH
in
TH
VO
TC
O
TH
TH
VO
CM
rH
ro
00
K
•*
rH
s
to
tx
TH
a
R
CM
8
tx
ct
CO
Oi
o>
CM
rH
tx
m
SP
TH
(A
CF*
V3
4J
a

CTi
m
rH
IX
O
rH
s
in
CO
00
ง
00
tx
TH
TH
CM
TH
TH
Oi
•*
m
s
TH
s
o
rH
tx
VO
s
ro
CM
tx
TH
ro
in
o>
s
rH
00
•*
CM
CM
CM
Max. Measurement
VO
ro
0
m
in
VO
i
CM
•*
$
00
1
CO-
00
s
rH
1
IX
CTl
VO
rH
00
1
in
m
in
00
CTi
8
in
in
vo
i
CO
rH
ro
CM
rH
rH
8
o
in
m
f
i?
CM
00
in
rH
R
CM
rH
CM
rH
1
3
O
CTi
VO
rH
Error at Max. Measurement
ro
ro
"?
m
m
rH
1
vo
m
TH
1
ro
O

CM
00
4
VO
VO
0
tx
•*
rH
CO
rH
tx
1
CM
rH
rH
1
CM
CTi
rH
VO
rH
rH
Skewness of Error Dist.
VO
tx
CM
ro
rH
ro
•*
in
in
in
VO
o
•*
CTl
8
•*
VO
CTl
ro
in
00
m
rH
tx
rH
rH
TH
ro
in
^J1
oo
m
oo
00
•*
in
Kurtosis of Error Dist.
                                                    C-13

-------
(U


(U
o

S
o
CO

rt
0)
ro

X
(J
I
(!)
6
in

U











1
3
CO










TH
<

m

ro

CM

O
fv
^-H
ro
00
m
CM
^<
ro
00
1O
o
o
00
•*
00
1
o
vH
•*
CM
O
t-l
1
S
IO
00
Cv
1
Ol
in
o>
ff>
=f
CO
CM
S
00
s
•#
s
TH
S
8
f
fc
in
CM
ซ
T-i
1
in
t-4
r~
CM
T-t
1
T-l
00
TC
00
00
t-i
i
(-<

^J<
O
t-H
•*
ro
in
t^
T-1
•*
CM
ro
t-i
in
10
tH
•*
t^
CTl
in
t— i
ro
in
ro
1O
CM
•*
Largest Positive Error
S
ro
•*
t-i
K
tv
00
in
o>
m
vo
tn
CM
ro
rH
m
m
ro
t^
Ol
Ol
CM
m
00
•*
TH
o
m
T*4
o
ro
t-H
00
•*
R
m
t-H
vo
tv
ro
ro
Ol
\0
oo
CM
CM
3
CM
t-H
TH
ro
>O
m
t-l
J
ro
S
O
IO
CM
ro
O
00
S
O
0\
00
t-H
o
•*
o\
CM
O
CM
•*
CM
O
>o
oo
CM
0

!?
o
Linear Correlation
10
•*
t-H
O
s
t-H
O
ro
S
O
S
O
o
l-x
ง
o
Ol
8
o
CM
S
O
i
O
K
CM
O
0
ro
TH
tH
O
O
TH
CM
O
Reduction of Variance
rH
cS
1— 1
s
in
t-i
tH
Ol
•*
O
g
•*
o
<ฃ>
m
in
O
Ol
t— i
•*
o
ro
ro
in
O
in
ง
T-i
S
•*
o
fป
00
00
0
r-,
CM
CM
ปH
w
(U
3
&
oo
tt
rt
0)
oT
1
w
•*
Ol
ro
ro
Y-l
in
in
00
ro
CM
t-H
in
t-H
S
8
t~i
TH
o
ง
Ol
in
TH
rv
O
TH
ซ
r^
in
ro
00
ง
00
IX
*H
r-l
(M
t— i
t-H
5
in
S
rH
S
o
TH
f-
>0
s
ro
CM
t^
TH
ro
in
rl
t-H
?
8
CM
Max. Measurement
S
a
*o
>O
O
oo
•*
00
1
o
TH
Tj<
CM
O
tH
S
ซ3
OO
tv
Ol
in
Ol
Ol
Ol
i
8
t^
Ol
10
oo
ro
tH
1
S
g
<ฃ>
1
ฃ
m
OJ
10
tH
1
in
tH
t^
•*
CM
tH
1
tH
00
S
00
TH
1
Error at Max. Measurement
ro
m
",'
(M
10
tH
00
•*
TH
1
00
m
TH
1
tx
Ol
o
g
=?
fx
o
tH
1
ซ
•*
N.
1
ง
TH
1

-------
*<;
IS?
a
o
o

>-
O
O











o
•rH
•s
00










i-H
<

CO

ro

fM

OJ

TH

1— 1

TH

ro

rH

O
O
4-1
W
•H
tS
•M
t/3

•*
TH
O
tx
TH

in
tx
Tf
TH

ro
00
tx
TH

TH
00
CO
TH

M
o
tx
TH

00
\o
tx
TH

TH
•*
^O
TH

8

ro
tx
00
TH
TH
CM
IX
CM
T-i
O
00
00
CO
TH
cs
8
TH
tx
TH
rO
•*
TH
1—1
TH
CM
^f
TH
in
in
0
01
TH
8
1
O
01
00
(M
01
CM
tx
TH
(M
CO
S
ro
n
2
1
CO
i
rt

(0
tx
CM
S
CO
0
CM
CM
CM
rx.
TH
TH
CO
tx
fx
o
CM
Mean Absolute Error
S
s
CM
10
in
m
15
tx
1
S
T-H
8
TH
1
TH
CM
TH
00
tx
m
CO
*
c>
00
CO
1
CO
tx
ง
*H
TH
1
s
s
CM
IO
CO
tx
CM
5
T-H
rH
(M
S
TH
TH
1
s
s
01
Largest Negative Error
CM
CO
TH
tx
CM
CM
Ch
10
tx
CM
CO
8
CO
O
in
CO
oo
O
m
oo

g
TH
TH
Tf
CM
CO
TH
tx
CM
CM
oo
TH
TH
Ol
o
TH
TH
ro

00
T-*
O
•*
S
O
CM
s
O
ID
00
CM
O
IO
CM
O
CO
*o
TH
o
s
ro
O
Ol
9
O
Linear Correlation
00
CM
CM
o
S
o
CO
S
o
IO
CO
O
o
tx
ง
O
o\
8
o
CM
ง
O

1
eu
aT
P.
,3
w
tx
tx
ti
TH
in
in
00
CO
CM
rH
in
TH
s
8
TH

o1
t/3
s
JS
4J*
&
Ji
(U
ซ
3
0
rH
tx
<ฃ>
oo
>o
CM
S
01
in
TH
tx
O
rH

1
in
in
tx
IX
s
O
O
TH
TH
CM
TH
00
tx
in
ro
*
Ol
CM
S
VO
CO
tx
S
TH
TH
X
s
CM

ro
8
•*
S
m
ro
•*
VO
Kurtosis of Error Dist.
                                              C-15

-------








ง
s

TH

in
t^.
3


5
in
^o
CM
O
CM
T-H
s
8
T-H
U3
•*
in
O
T-H
00
00
00

en
TH
ro
ซ
cS
ro
Mean Calculated
m
CM
00
CT\
T-H
s
2
•*
CTl
•*
T-H
T-H
ffl
8
T-H
s
•*
T-(
T-H
TH
ro
Tt*
T-H
55
VO
t^-
T-H
&
o
m
ro
Oป
O^
f^
ro
CM
in
CM
vo
T-H
s
CT>
T-H
CM
i
ฃ
2
rt
$
00
i
rt
o
TH
s
T-H
1— (
S
Tf
in
TH
Mean Absolute Error
o
TH
Oi
CM
S
ง
•*
oo
i
8
T-H
1
R
>o
00
tv.
1
S
%
a\
00
CO
S
r^
i
oo
CO
TH
1
O
TH
O)
CM
3
CM
00
o\
oo
TC
TH
1
S?
ro
TH
TH
ro
00
CM
t^
ro
TH
s
ฃ
w
ซ
3
CO
O
a\
IO
in
TH
ro
CM
TH
o\
ro
TH
oo
•*
00
t^
m
TH
i?
S
t~.
•*
TH
o
CM
.
ro
f^
CM
1)
f
rt
s
tH
h
M
iH
o
O
ID
TH

CM
O
4
s
o
T-H
o
0
s
o
TH
o
o
CO
0
o
m
T-i
9
S
4
TH
o
o
8
w
tt
rt

CM
ro
O
ง
CM
O
O^
oo
TH
d
X
CM
d
CM
Tf
CM
O
ซ
3
o
ซ
t^
CM
O
ro

O\
CM
TH
tx
in
oo
Tf
TH
Intercept, Least Squares
S
o
TH
t-N
*o
CO
10
o
Ol
o^
in
TH
K
O
*H
S
m
ro
00
ง
15
T-H
T-H
CM
T-H
T— 1
TP
31
T-4
s
0
TH
t^
<0
s
CO
CM
t^
TH
CO
m
o>
S
T-H
9
5!
CM
CM
Max. Measurement
0
TH
a
•*
^
g
CT>
•*
00
8
T-H
1
rt
00
r^
M<
CM
vs
ID
O>
S
S
\o
00
ro
TH
o
TH
&
3
CM
00
O\
oo
•*
TH
•*
00
ro
ro
TH
TH
1
ro
00
CM
t-.
ro
TH
Error at Max. Measurement
CM
o\
CM
i
CM
ซ
TH
•*
TH
TH
1
oo
in
TH
1
CM
to
s
o
m
ฐ,
&
TH
TH
TH
in
1
ro
TH
o
IO
00
ฐ,
$
o
Skewness of Error Dist.
r^
TH
TH
o\
o
f>
in
ro
t^
in
t^
CTl
t-^
S
ro
a\
00
ro
TH
O
m
CM
00
o\
r^
0\

-------
a

CM

tx
in
CM

CO
TH
CM

S
co
Number Cases
to
TH
ro
to
CM
TH
in
to
^
TH
IX
00
m
00
TH
Ol
00
m
CM
TH
TH
tx
•*
00
TH
8
a
TH
m
CO
Ol
CM
0]
(O
to
o
fM
Tj<
Oi
to
CM
01
M
to
in
TH
TH
M
O
00
CM
O
CO
Mean Measured
m
00
55
CM
00
TH
TH
9
in
TH
CM
tx
•
•*
ฃ
l>
•*
TH
ฃ
tx
00
CM
O
Ol
Ol
TH
•*
(M
S
Ol
00
Oi
in
to
TH
•*
m
CM
CO
3
CM
to
in
Mean Calculated
00
oo
tO
T-t
CM
ซ
01
3
iH
T-t
o\
LO
00
T— 1
T-H
m
i— 1
-*
i-H
*-H
^H
8
TH
00
CO
tx
CO
TH
TH
in
to
un
TH
TH
00
Ol
m
CO
8
TH
CM
CM
s
TH
00
TH
CO
tx
CM
CM
CO
14
W
Si
rt
&
co
a
S
s
1
4->
O
O
&
o\
VO
•*
CM
CO
CM
t-t
S
O
t^
TH
1
t~-
TH
•
TH
CM
TH
1
a
to
1
in
CM
ง
TH
1
CM
to
s
1-1
1
8
CM
CM
CM
O
in
tx
to
in
00
CO
*H
TH
TH
W
0\
m
CM
1
S
V
S
m
IO
S
TH
M
to
CM
TH
5
O
tx
TH
tx
TH
.
TH
CM
TH
o
TH
tx
tx
in
to
TH
vH
TH
TH
CM
CM
CO
TH
S
R
CM
S
CT>
to
TH
00
to
ซ
TH
CM
oo
00
to
CM
Mean Absolute Error
S
8
0
TH
tM
Ol
S
to
1
S
oo
m
^
TH
m
d\
CO
t
CO
CO
S
CO
I
R
o
K
m
i
$
CM
^
S
m
tx
o
*H
1
fv
ff\
O
to
o
TH
0\
TH
CM
tv
in
ffl
m
in
CM
Largest Negative Error
rH
O
01
to
TH
s
4
B
to
TH
r~
in
Ol
TH
1
8
TH
CO
to
00
to
TH
TH
S
f^
f-
a
CM
CM
CO
TH
10
oo
TH
01
to
Ol
CO
CM
CO
to
TH
o
CO
Ol
CO
TH
Largest Positive Error
R
00
to
CM
*H
O
tx
ro
tx
in
CM
5
S
•
Ol
TH
•*
Ol
in
00
TH
to
in
to
^
oo
to
Ol
m
Ol
VO
in
00
to
00
01
CO
CM
CO
00
CM
in
tx
TH
00
in
S
ra
,— i
$
9
to
TH
m
to
ง
&
8
1
oo
CM
O
s
4
&
to
TH
1
fx
in
oi
TH
CO
CO
O
00
CO
O
00
CM
O
s
TH
CO
IX
CO
CM
CO
o
8
*-l
tt
rt

-------
 X
CTi
U
1
CTi
U
O
CM

ง
CM

tx
m
CM

en
TH
CM

•*
>O
CO
Number Cases
o
TH
in
CM
S
<ฃ>
TH
TH
TH

CM
TH
CO

TH
a\
CM
•*
in
•*
TH
CM
CO
00
s
TH
CM
Ol
m
tv
TH
h
(U
f-l
I
V)
a
a
.
8
TH
1
t^
CO
00

TH
CM
8
in
in
TH
8
TH
O
TH
3
CO
CO
TH
I Mean Absolute Error
K
oo
S
TH
I
8
CO
CO
1
8
S
1
CO
CM
O
TH
•t
•*
m
m
•*
CO

in
m
00
fx
S
t>.
CO
TH
in
in
•*
CTl
CO
in
VO
TH
0
TH
Cx
J?
TH
s
tv.
s
oo
CTi
TH
s
&
s
Pi
s
B
ซ
S
o
S
4
S
o
in
CM
o
TH
TH
TH
1
N.
CO
O
•"*
fM
TH
N
CM
CM
00
00
O
ง
*?
00
TH
ซ?
1
X
CTl
in
m
O
TM
in
•*
O
CTl
S
O
O
1O
CO
o
00
tx
CO
O
CO
3
O
IX
TH
CO
O
in
S
o
o
s
o
•*
5J
o
•*
CM
in
o
Linear Correlation
CO
TH
CO
O
m
8
0
ro
&
O
O
CO
TH
0
CO
^<
TH
o
CTl
8
O
TH
o
TH
o
Vฃ>
CM
TH
o
i
o
m
M
CM
O
s
CM
O
Reduction of Variance
o
&
TH
to
s
CM
in
CO
ix
O
CTl
TH
tx
CM
s
tx
o
CO
3
O
m
TH
CTi
o
TH

rH
R
TH
•*
s
s
^<
VO
CM
S
IX
CTl
S
in
S
s
<0
o
tx
O
(O
tx
00
CM
TH

(M
T~t
S
O
CM
tx
O
CM
CM
CM
IM
O
CTl
m
(M
r-i
00
TH
s
tx
s
in
ฃ
10
s
o"
•
tt
s
t-l
o
w
•s
.a
M
o
ซ
                                                C-18

-------
ro

X
a
o
'3
•3
o
o
O

0)
I—I
-O
ni
H










a
.2
%
•4-J
CO











i-H
i-^
<

ro

ro

CM
(-^
CM

TH

TH

TH

ro

TH

0
Statistic

in
CTI

TH
o>
K
CTl
00
ro
VO
CM
TH
•*
TH
t^
00
CM
TH
in
oo
00
TH
TH
ro
CO
CM
TH
s
ง
T-H
Mean Absolute Error
s
o\
ro
ปH
TH
>o
TH
s
1
O>
ro
Si
ro
1
K
CM
m
ro
%
10
in
CO
ro
m
t-^
S
ง
•*
00
•*
s
CM
0\
ro
TH
1
t^
Ol
S
rs]
T-t
5
oi
CM
oo
TH
O
ง
t^
1

g
1-1
s
00
CTl
8
S
Largest Positive Error
s
s
tn
t-i
s
t^
ro
ง
CV
r^
•*
oo
CTl
00
TH
•*
CTl
CTl
•*
m
TH
t^
CM
CM
t*ป
CO
•*
in
t-~
in
O
in
CO
m
m
TH
o
<ฃ>
TH
TH
•*
TH
•*
TH
CM
CTl
t-~
CM
CM
•*
OO
0)
(2
tH
ro
O
"i
S
O
t^
CTl
4
•*
TH
0
ro
O
V
S
O
in
TH
o
a
CO
TH
ro
=?
ซ
TH
CM
1
CM
C-.
ฐ,
S
ฃ
ซ
rt
.3
TH
00
•*
O
TH
3
0
CTi
VD
CM
O
ง
ro
O
oo
rv
ro
O
ro
•*
CM
O
tx
TH
ro
O
in
S
o
0
8
0
S
•*
o
cU
in
O
Linear Correlation
TH
CO
C\]
O
ro
S
O
CO
tx
O
O
O
CO
TH
o
CO
•*
TH
o
CTl
in
O
o
TH
o
TH
o
(O
CM
TH
o
ง
o
in
CM
CM
0
s
CM
O
Reduction of Variance
CM
oo
•*
TH
in
tM
in
CM
CTi
ซ
VO
O
CM
S
T-I
*o
s
o
TH
o
in
o
ซ
t^

TH
CO
S
*-l
CO
0>
5
3-
V)
•M
S
CD
•ซ
4)
I
OO
S
oo
in
in
rf

00
S
5
CM
00
Intercept, Least Squares
r-ป
oo
TH
vo
TH
o
tx
TH
•*
s
•*
tx
•*
>o
CM
^
•*
tx
CTi
^O
•*
in
S
S
1O
o
TH
cS
.
S

-------








S
S

•ป
1
ilculat]
O
ฃ•
s
TH
CM
TH
rH
O
O
1 — 1
J3
rt
H














a
o
3
•M
V3











F— t
<

ro

fO

fM

rg

iH

l-l

1-H

ro

TH

O
Statistic

m
01
W3
(M

in
s

ro
00
CM

ro
ro
CM

TH
00
CM

in
00
CM

>o
ซ3
CM

00
10
CM

t-~
m
M

ro
1-1
tM

s
ro
Number Cases
.
00
m
00
TH
Ol
00
in
CM
TH
TH
t^.
3
T-t
8
S
1-4
in
ro
tTi
CM
CM
ro
CO
a
TH
CJl
ro
(M
CT>
CM
VO
m
TH
TH
(M
O
00
fM
O
ro
Mean Measured
01
TH
(M
ro
CM
s
oo
00
(M
TH
i-t
TH
l-H
ro
O
ro
t^
t--
m
i-H
ro
TH
00
TH
8
TH
TH
t-^
ro
00
s
O
00
m
S
00

oo
Ov
l-H
5
l-H
oo
•*
ซ
oo
in
Ol
8
^-i
a
l-H
^H
1-1
00
TH
O^
VO
•rH
l-H
ro
O
l-H
ro
m
00
in
CM
CM
l-H
ซ-H
ID
CM
iH
S
^f
TH
CM
Root- Mean- Square Error
S
T-t
to
t^
•*
in
3
rj<
t~-
i

g
00
g
ro
ซ
S
TH
00
l-H
in
m
00
5
o
ro
K
10
iH
S
s

TH
fc
00
CM
s
CM
CM
CM
CO
s
TH
00
in
•*
K
00
N
ro
TH
TH
oo
CM
CM
vH
K
in
ro
00
in
ro
CM
TH
CM
O
rH
Largest Positive Error
o
TH
CM
m
CM
CM
s
ฃ
TH
in
ro
TH
in
in
ซ
oo
oo
TH
•*
s
s
in
CM
VO
ro
<ฃ>
t^
ro
O>
in
in
in
0
i-H
CM
in
CM
CM
CM
ro
f-

t^
TH
•*
iH
0
60
ง
&
b
1
S
ฐ,
t-t
m
s
s
v
00
O
O
ro
ro
4
S
0
•*
in
TH
ro
CO
O
TH
CM
ฐ,
CM
00
9
in
in
9
S
S
1

s
O
TH
o
TH
o

O
Tf
O^
O
CM
ro
CTl
in
>O

(M
9
ro

-------
 tu



 o

 3
 o


 rt
 U
lO-
 rt
 O
 I
 a
 U

 a
 i—r
 ,ฃ>
 rt

 H










B












i— t


CO

co

CM

rj

TH

TH
Q
TH

ro

TH

0
Statistic

in
M

in
CM

CO
00
CM

ro
CO
CM

TH
oo
CM

in
00
CM

<ฃ>
10
CM

00
CM

in
CM

to
TH
CM

ro
Number Cases
TH
CO
CO
CM
in
CO
T-l
oo
in
oo
TH
CTl
00
in
CM
TH
TH
00
TH
CM
O
ffl
TH
in
CO
Cl
CM
CM
CO
CO
ง
CT>
CO
CM
CM
in
TH
TH
CM
O
oo
CM
O
CO
Mean Measured
s
TH
T-l
CM
CO
tx
00
00
CM
in
<ฃ>
TH
o
00
CM
TH
3
8
TH
in
8
CO
TH
CO
01
TH
00
CM
s
(O
CO
O
Mean Calculated
s
TH
in
00
TH
CM
TH
00
CTl
in
o
TH
CM
3
CM
CO
TH
CX
CM
CM
Ol
00
TH
s
CM
TH
00
CM
TH
I
t)
1
to
a
rt

-------
o
P-I
o


o
00

rt
rt

O
3
o
ffi

TH
•*
01
oo
m
o%
CM
00
ro
CM
TH
CM
fv
t-x
ro
O
ro
Mean Measured
3
X
CM
tO
o\
t^
CM
CM
s
ro
O
CM
"t
TH
s
TH
ง
KJ
•T— i
8
3
TH
s

in
TH
•*
TH
ro
CM
in
CO
CM
in
a\
CM
fc
C7>
in
CM
S
i
1
8
rt
g<
t/5
i
rt
0)
S
4L
1
s
m
TH
TH
rv
K
TH
•
ro
TH
TH
1
oo
TH
CTi
CM
••*
TH
CM
o>
w
10
(O
(O
S
TH
o
in
10
M3
CM
CJ\
00
ง
TH
S
CM
•*
TH
8
s
I
w
9
v
S
ro

CM
CM
ro
00
1
CM
0ป
oo
ซ3
oo
1
"*
CM
tx
CO
00
tx
•*
O
s
TH
1
00
CO
r^
m
TH
IQ
1
cS
CM
0\
rO
TH
1
TH
CM
in
c\
00
CM
TH
10
IO
1O
Largest Negative Error
S?
ffl
CT>
r^
CM
VO
VO
TH
f^
TH
(O
s
TH
TH
oo
00
CM
CM
>o
in
ti
t--
5
ro
K
^
t-~
CM
*O
TH
in
in
CM
S
IO
CM
in
ro
oo
in
10
TH
ro
•*
8
ro
T— t
s
o^
0\
tx
CM
Largest Positive Error
S
1O
m
01
00
CM
^O
3
TH
t-x
CO
.
00
ro
TH
CM
O
TH
s
T-4
3
a*
in
TH
00

01
TH
o
TH
s
d
in
1O
•*
o
Linear Correlation
vo
oo
CM
0

S
O
ro
oo
ro
O
TH
s
0
IO
TH
CM
! 0
TH
o
o
Ol
rji
in
TH
t-.
o
TH
(O
tv
s
oo
s
s
TH
vH
OJ
TH
o\
8
TH
o>
•*
in
S
TH
S
o
TH
t^
>o
3
CO
CM
tx
TH
ro
in
o>
s
TH
00
•*
CM
s
CM
Max.. Measurement
S
in
TH
to
r
00
CM
1
s
.
ro

^f
TH
1
o
TH
o
s
9
s
TH
1
TH
•*
TH
t
in
t*.
o
8
o
o
ft
CM
Skewness of Error Dist.
r^

TH
TH
TH

s
ro
s
CO
rv
o^
m
$
a
ป
•*
CM
CM
•*
•*
TH
00
Kurtosis of Error Dist.
                                                C-22

-------
 O
oo


.3
 o
 tu



 a)



 O
oo

 rt



I

Ol
rt
U
 a
O
U
rt
t-










ง
U
•M
oo












3
<

CO

ro
00
CM
r-v
CM

TH

TH

o
TH

co

TH

0
Statistic

•*
TH
O
t--.
TH

in
t^
•*
TH

ro
oo
t-~
TH

TH
00
rO
TH

CM
O
t^
TH

00
to
t^
TH

TH
S
TH

s
to
1-1

Ol
00
•*
TH

ro
TH
CM
TH

ro
Ol
OO
CM
Number Cases
ป
CO
CM
&
•*
TH
Ol
CO
oo
TH
TH
CM
CM
TH
tO
in
00
TH
CM
oo
CM
Ol
TH
tO
in
Ol
CM

t^.
00
Ol
TH
•*
01
oo
in
Ov
CM
00
ro
TH
CM
fc
to
O
ro
Mean Measured
tO
tO
to
TH
m
TH
o
T-i
s
X
TH
TH
•*
•*
CM
t-^
5
TH
in
t*~
o>
in
TH
t^
00
CO
O

00
to
3
CM
oo
ro
in
8
S
ro
in
TH
8
S
CM
Mean Calculated
Ol
CM
&
TH
t^
to
ro
•*
TH
s
o\
•*
TH
R
TH
CJ
TH
s
ro
TH
CM
m
tO
ro
TH
CO
in
CM
t^

N
a\
o\
m
CO
s
TH
CM
CM
rx
VO
ro
10
TH
f-~
00
a
CM
U
o
00
8
TH
IN.
t-H
S
T-l
S
1—1
0*
o
*-)
S
00
00
K
0!
cS
S

S
IX
TH
CM
m
TH
Ol
m
TH
CM
m
o
TH
TH
S
51
TH
Mean Absolute Error
ro
CM
K
•*
"?
CO
O
TH
o
t^
1
TH
ro
W
0>
\o
o\
TH
oo
ro
a\
CM
O>
CM
TH
t^
t^
t^
ro
in
8
TH
1
ro
CM
S
"?
TH
CO
in
S
TH
1
TH
CM
S
o
TH
1
a
CO
in
•*
TH
1
PJ
<1>
ฃ
ซ
DO

CM
t^
00
in
in
CM
a\
g
to
CM
ro
oo
a\

m
CM
00
CM
oo
o
ro
CM
m
10
g
5!
in
S
in
t^
a\
Largest Positive Error
CM
t^
1
K
S
TH
•^
LO
TH
i
TH
CM
CM
•*
TH
00
O
TH
g
S
•.
TH
o
TH
o
o
8
S
S
JS
t^
CM
•*
o
CM
TH
0
in
CM
CM
o
o\
S
o
ฐ


S
O
S
CM
O
t^
S
O
CM
3
O
Linear Correlation
CM
00
TH
O
S
o
o
TH
8
o
8
o
ro
ง
O
TH
cB
o
CO
ro
TH
o

ง
o
o
8
o
CM
TH
o
i
o
Reduction of Variance
oo
in
to
o
CO
•*
TH
O
oo
TH
CM
O
K
tn
TH
o
00
S
o
ro
CM
CO
O
ro
fc
O

S
in
0
TH
$
o
o>
m
m
o
•*
TH
rx
o
ฃ
3
g.
to
S
U
oT
S
1/5
f%
CM
X
TH
t^
CM
O
ro
TH
S
S
TH
to
^H
o\
TH
TH
K
tx
CTl
TH
TH
VO
CM
TH
•*
TH
CM
ง

0
TH
ง
CM
fv
>o
TH
CM
CM
{Q
S
TH
rC
CO
ro
TH
Intercept, Least Squares
S
TH
t^
to
g
S
CTl
CTl
m
TH
&
TH
S
m
CO
oo
8
tx
TH
TH
CM
TH
8
TH
$
$
TH
S
o
TH
tx
to
S
ro
CM
CN
TH
ro
m
H
TH
5
3
CM
CM
Max. Measurement
ro
CM
01
S3:
to
i
ro
O
TH
R
TH
CO
$
o\
to
a\
to
TH
"?
ro
Ol
&
01
1
TH
TH
in
in
to
i
CO
in
8
TH
1
CO
CM
Ol
f-
*
TH
CO
in
to
in
TH
1
Ol
•*
g
ฐt
t^
ro
TH
Error at Max. Measurement
00
CO
n
S
o
5
TH
i?
TH
CM
to
O
TH
CM
9
vH
"*
TH
1

to
in
f
00
CO
=?
ro
ro
TH
1
S
<=?
Skewness of Error Dist.
$
ro
TH
TH
to
TH
in
ro
to
to
TH
CM
rf
K
ro
00
ro
O
•*
ro
R
to

S
8
TH
00
00
•*
S
to
TH
in
•*
Kurtosis of Error Dist.
                                                      C-23

-------
 8
.2
o
a.
V
U


o
00


a

<
rt
U

^

o
O
tn
TH


O










o
ซ
4-1
in











3
<

m

m

CM

CM

TH

TH
o
TH

m

i— i

O
Statistic

Tf
TH
R
T-H

R
•*
TH

ro
00
ฃ

TH
00
ro
TH

CM
0
tv
TH

00

ro
tv
CO
CM
&
m
•*
TH
CTI
CO
tv
oo
T-H
T-H
CO
IV
CM
TH
o
m
CTl
CM
CM
tx
00
CTl
TH
•*
CTI
00
in
CTI
CM
00
ro
CM
TH
CM
tv
IV
ro
0
ro
Mean Measured
K)
*
TH
TH
00
CTl
ซ
>O
CM
ro
TH
<ฃ>
tv
oo
ro
CM
1O
CO
TH
TH

T-t
Mean Calculated
o
CM
8
CM
5
$
TH
CM
O
$
TH
00

ง
*O
TH
1
3
00
T-t
T-l
TH
1
TH
IN
8
O
3
in
in
•*
S
CM
ซ3
•*
in
tv
CTI
*t>
10
Largest Positive Error
S
00
ro
oo
tv
TH
in
S
m
TH
ro
>O
m
CM
CM
CM
in
o
8
CM
ro
S
in
TH
oo
•*
ง
CO
CM
a
BO
a
B,
8
S
S
o
S

S
o
8
o
8
o
TH
o
o
S
=?
CM
o
o
tv
CM
d
•*
TH
9
CM
CM
O
ta
•w
S

o
CV
•*
0
00
ro
(O
O
OO
S
0
o
S
g.
t/3
4J
S
-)
M
(U
I1
co
O
*H
^
tH
R
CO
TH
•*
in
00
in
TH
TH
TH
ฃ
TH
CO
CM
8
TH
a
CM
Tf
TH
CTl
00
5
TH
CO
O
TH
o
CO
TH
tv
TH
CM
CM
6
S
TH
S
5
TH
Intercept, Least Squares
s
TH
tv
V)
ง
S
CTl
CTl
m
TH
tv
O
TH
S
in
CO
oo
S
00
IX
TH
TH
CM
TH
TH
5
in
•*
CO
TH
S
o
TH
tv
ป
S
CO
CM
tv
TH
CO
in
TH
00
•*
CM
•*
CM
CM
Max. Measurement
to
•*
00
CO
m
<ฃ>
i
•*
TH
tv
1
S
S
CTl
1
rO
tv
!ฃ>
00
tv
tv
CTl
S
ffl

*H
1
Error at Max. Measurement
R
t
IV
10
O
tv
•*
TH
o
TH
TH
1
K
o
CM
•*
9
S
o
tv
00
f
ct
TH
1
in
<&
TH
3
TH
Skewness of Error Dist.
TH
o
o
TH
S
m
o
in

CTl
CO
00
^<
CO
S
in
CTl
TH
S
TH
CM
in
in
TH
tv
tv
S
in
Kurtosis of Error Dist.
                                                  C-24

-------
 ง
 • iH
 C/l
 .a
 g
 w
 o
 t/3



 1
 O
 CM
 0)
 O

 3
 o
 in
\0

 a
 o

 $
 i—i


 rt
 U




 1
 tu

 6
 U










a
o
•M
00











i— 1


CO

CO

CM

r\J

TH



TH

CO

t-H

O
Statistic

TH
R
TH

to
TH

ro
00
TH

TH
00
CO
TH

CM
O
TH

00
to
TH

TH
to
TH

o
to
TH

Oi
TH

ro
TH
(SI
TH

CM
Ol
oo
CM
Number Cases
to
CO
CO
CM
o
TH
Ol
ro
00
TH
TH
CM
CM
TH
to
to
00
TH
CM
00
CM
Ol
TH
to
to
a
CM
00
Oi
TH
Ol
00
to
Oi
CM
oo
ro
CJ
TH
CM
fc
ro
O
ro
Mean Measured
CM
ro
o
oo
CM
Ol
TH
oo
ro
00
to
TH
TH
tv
Ol
X
Ol
to
TH
CM
to
00
to
TH
TH
to
8
TH
to
to
00
CM
ro
TH
to
CM
to
CM
CTl
to
to
00
Mean Calculated
to
oo
CM
TH
TH
to
to
TH
5
s
TH
TH
o
ro
TH
CO
o
fe
TH
00
00
VO
TH
TH
TH
TH
CO
TH
s
ro
ro
ง
TH
to
to
00
CM
ro
8

-------
 w
 o
 CO
 0
 p-l
 S
 o
 CO

 rt
IS?
 rt
 u
 O


 6
 o
 a
 i—t
 ri
 H










1
rt
co











^


8
erf
in
i
S
TH
m
i
CM
CO
TH
CO
in
'"?
00
TH
TH
ro
O
CM
CM
1
CM
TH
00
a\
i
00
ro
in
TH
CM
ro
tt
*
8
S
m
CM
TH
in
S
TH
LO
CT,
in
^
to
TH

ง
CT,
CM
TH
TH
00
in
CM
TH
CM
00
in
TH
8
TH
TH
g
s
Mean Absolute Error
&
in
CO
i
oo
00
CM
CM
1
fe
0
TH
CTl
ro
00
CM
o
1
ro
TH
1
g
TH
CM
O
TH
1
&
m
CO
oo
CM
TH
00
TH
CTI
iH
O
TH
1
in
CM
CM
TH
TH
1
Largest Negative Error
ป
O
3
TH
o
oo
o
ro
TH
CM
in
CO
CO
CM
00
VO
8
CT.
CM
in
in
S
TH
ro
oo
TH
TH

-------
w

o
a
o

4->
a
'o

00
in
CO
TH
TH
tx
•*
oo
tH
s
s
TH
in
ro
CTl
co
CO
ro
ro
O
CO
•*
Ol
ro
CO
o\
CO
vo
in
T-H
tH
CO
o
00
s
ro
Mean Measured
ro
vo
oo
tx
CJ
ro
O
tx
tx
•*
Tf
tx
tH
CT>
oo
CM
T-H
TH
15
YH
8
vo
co
T-i
O\
VO
s
TH
CO
CO
vo
CTl
VO
m
in
TH
ง
TH
01
in
in
ro
3
8
VO
Mean Calculated
8
CO
S3
R
TH
TH
3
vo
ro
TH
tx
•*
ro
CM
TH
CO
ro
CM
a\
%
Ol
o
TH
CO
IX
8
TH
CM-
01
a\
ro
o
CM
CM
ro
CM
VO
>o
1
CM
m
h>
in
ro
Root-Mean-Square Error
tx
•*
$
00
Tf
iฃ)
5
TH
TH
TH
00
o^
ซ3
o\
s
00
1
S
tx
^O
in
ซ3
3
TH
1
8
s
co
ซ
TH
g
TH
s
•*
•*
TH
in
in
CM
0
ro
Mean Error
S
in
TH
^
ซ3
oo
in
m
<ฃ>
TH
TH
[X
ro
S
TH
CM
ฃ
S
ro
oo
S
5
TH
tx
TH
w
TH
ro
in
•*
TH
oo
TH
•*
TH

00
00
Ol
CO
in
in
a
t
•*
TH
VQ
tx
(M
1
S
s
in
S
TH
in
•*
O>
O
>o
3
TH
o
VQ
CJ
8
TH
S
m
oo
t
in
oo
3
TH
Largest Negative Error
s
i
tH
O
00
•*
TH
co

S
s
TH
f-l
0)
.s
.•a
8
PL,
tJ
0)
if
J
VO
CTl
ซ5
t>
•*
ca
S
IN
in
m

f;
ro
CO
CM
TH
ro
CM
00
TH
S
ro
CO
tH
TH
TH
CJ
tx
<ฃ>
in
TH
o
in
CO
CM
CM
CM
tx
co
ฐ,
&
1
.3
in
00
in
O
O
<ฃ>
TH
0
ro
CO
ro
O
a>
3
o

N
00
CM
tH
<ฃ>
tH
in
•*
CO
tH
CO
tH
CO
(O
CM
CO
01
m
vo
vo
tH

-------
ง

I

I

 rt

•3
 O

"rt

O
 o




CM


 .
CTi
TH


u

 
rt
-t-l
00

m
CTl
10
CM

m
rC
CM

CO
00
CM

CO
CO
CM

TH
00
CM

m
00
CM

tO
to
CM

00
to
CM

IN
in
CM

ro
TH
CM

•*
to
CO
Number Cases
10
TH
CO
CO
CM
TH
m
CO
rH
TH
f^
oo
in
00
TH
CTi
00
in
CM
TH
TH
tN
$
TH
CM
o
s
TH
in
CO
CTl
CM
CM
CO
CO
0
CM
rH
CTl
CO
CM
CTl
CM
tO
in
TH
TH
CM
ง
CM
0
CO
Mean Measured
CTi
00
CTl
in
TH
to
•*
in
o
TH
(M
CO
IN
CM
TH
s
3
OO
tO
5
T— 1
S
O
tO
TH
ro
to
CO
O
TH
51
IN
*o
CM
CO
CO
CO
O
CM
t^
CTl
S
T-i
%
t^
CO
CM
Mean Calculated
S
IN
•*
TH
•*
CTl
s
TH
in
00
CTl
TH
\o
ง
T-t
S
00
CTl
CTi
CO
to
CTl
CTl
OO
K
•*
TH
5
TH
to
CM
TH
m
C^
00
TH
h-
•*
TH
CO
TH
TH
in
ro
in
TH
s
w
a
u
$
to
i
S
a
S
4-1
o
o
ti
CM
t^
in
O
00
CO
I
lO
in
00
in
S
ฃ
CM
O
O
i;
TH
in
CO
CO
1
TH
t^
in
CM
TH
•*
00
CM
m
rH
S
CTl
00
CTi
m
VO
in
i
TH
00
•*
ซ3
1
S
a
u
S
cu
s
CTi
10
ro
O
TH
00
ro
rH
IN
IN
CTi
ฃ
R
TH
00
ro
O
in
c-~
s
ro
t^
CM
^O
^o
CM
TH
tv
t^
CO
oo
TH
00
•*
CO
CO
TH
ro
O
CTi
00
CTi
rH
00
TH
TH
Mean Absolute Error
S
ro
rH
CO
rH
1
CO
CTl
rH
CTl
CM
1
00
rH
CO
m
CO
00
CM
rH
CTl
CO
1
00
rH
S
CO
o
CM
CO
rH
CO
rH
1
in
to
TH
CTI
rH
rH
1
rH
CTl
ฃ
t-~
rH
00
ft
in
S
s

•M
1
CU
ซ
0)
i?
_)
CM
to
s
to
rH
•*
S
m
o
tN
^
to
oo
CM
tN
•*
oo
00
%
in
CM
oo
•*
00
tN
CTi
TH
tN
rH
in
O
rH
Tf
in
to
rH
TH
O
s
tr>
TH
8
ง
CO
00
ง
OO
4)
bO
a
rt
A
8
S
8
S
to
in
0
in
CO
0
R
o
S
o
8
O
T-l
*H
o
1-1
o^
9
S
o
TH
ro
ฐ,
S
9
PJ
•M
S
ra
_)
t-.
in
in
O
0
•*
CM
0
CO
tN
ro
0
to
ง
O
S
•*
o
t^
TH
ro
O
CTi
CM
in
0
to
S
o
CTi
CTl
TH
o
•*
CM
in
O
8
in
O
Linear Correlation
o
TH
CO
O
00
s
0
CTi
CO
TH
O
CTi
8
0
•*
to
rH
O
8
rH
O
O
00
CM
O
1
o
CTi
CO
O
O
m
CM
O
o
IN
CM
O
Reduction of Variance
CM
S
O
00
CM
rO
O
r^
CO
^H
O
r^
S
0
rH
00
t^.
0
CTl
in
rH
O
rH
S
TH
t^
g
O
oo
CM
•*
O
CO
CO
rH
TH
TH
s
T-t
(A
V
s
g.
oo
*
01
_)
aT
ง

-------
PJ
o
O.
3
o
i — i
rt
U
o
O
CM
 I

U
ri
H










ง
3
-I-)
to











i— r
<

ro

ro

CM

CM

TH

TH

TH

tO

TH

O
Statistic

in
Ol
to
CM

in
•*
CM

co
00
CM

ro
CO
C\J

TH
oo
CM

m
00
(M

tO
tO
CM

oo
to
CM

lx
l/)
(M

ro
TH
CM

3
CO
Number Cases
to
TH
ro
ro
CM
TH
in
ro
•*
T-t
[X
00
in
oo
TH
CTl
00
in
CM
TH
TH
lx
•*
oo
CM
o
•*
a\
TH
in
ro
O\
CM
CM
ro
ro
O
CM
•*
o\
ro
CM
0\
CM
to
in
TH
TH
CM
3
g
ro
Mean Measured
a
CTv
ro
TH
s
10
c\
g
s
TH
TH
oo
CT>
00
8
to
ro
tx
ro
$
TH
3
TH
ro
TH
to
in
to
oo
TH
CM
in
ro
in
TH
R
ro
M
TH
10
ro
O\
to
TH
Mean Calculated
ro
a\
•*

TH
*o
CTl
TH
Root-Mean-Square Error
CM
a\
ro
a\
i

TH
1
$
CM
o
i
Largest Negative Error
t^
CM
oo
o>
CM
CM
a\
CO
CO
CM
tx
CM
00
a
5
X
1— 1
8
o
(M
in
CTl
t^
IO
CM
TH
•*
TH
o
TH
CM
CM
TH
o^
TH
TH
o\
TH
01
CM
a
TH
ซ
TH
ซ5
00
CO
in
TH
Largest Positive Error
TH
tx
8
IO
TH
CM
00
tx
•*
m
s
^
>o
t^
TH
rx

o
3
=?
ffl
rO
O
ct;
CM
1
S
=?
ct
9
T-l
i-H
9
a
o
CM
o
i
o
Reduction of Variance
r-~.
to
TH
TH
ro
O
CO
O
TH
CO
•*
O
tx
oo
in
o
o\
CM
oo
O
ง
O
00
ffl
oป
0
CO
TH
oo
0
TH
TH
to
0
tx
CO
CO
TH
•*
CM
•*
TH
c/1

TH
ro
IX
in
l-~
TH
IX
in
oo
o
CM
TH
ro
•*
oo
CTl
s
f-.
to
CM
CM
in
oo
a\
TH
in
TH
ง
TH
tx
TH
to
Intercept, Least Squares
tx
oo
TH
10
TH
R
TH
•*
s
•*
tx
rf
to
CM
*
tx
a\
%
in
S
•*
tx
to
O
TH
tx
O
to
tx
00
CM
TH
to
TH
3
CO
TH
ro
TH
ro
to
CM
CO
Ol
g
to
TH
ffl
Max. Measurement
R
to
CM
TH
g
TH
ro
S
r^
ro
1
CM
ro
CO
ro
ro
CM
R
CM
CO
00
ง
in
i
Cft
TH
5
R
tx
to
CM
TH
s
>o
CM
CM
TH
1
in
m
?
TH
in
in
m
to
Error at Max. Measurement
TH
TH
CM
1
8
0
•*
TH
o
ro
O
TH
1
CM
ro
^
TH
VO
9
S
o
s
1-1
t^
m
TH
ff
TH
oo
to
9
Skewness of Error Dist.
T*
V- 1
TH
TH
3

oo
CM
ro
to
Ol
TH
o>
TH
o
00
to
CM
O
TH
CM
in
eci
to
s
tx
8
in
CM
CM
O
Kurtosis of Error Dist.
                                              C-29

-------






W
O
VI
C/l
1
a
0
t3
ฃ
6
'0
cซ
(U
•
1
O
•*
CM
•
TH
CM
U

TH
ro
ro

tx
(M
a.
in
tx
-4<
TH
TH
•*
o
IX
TH
TH
CM
a\
TH
a\
m
T-t
tx
to
Ol
O
8
o
TH
•*
rH
00
to
in
CM
tx
to
in
00
-*
Mean Calculated
s
&
*H
8
O
in
•*
to
%
oo
to
CM
tx
CM
g}
t-i
S
5
CM
s
1
(U
3
tJ1
OO
ง
(U
+-ป
o
o
A
R
$
ง
•*
o
O
w
s
TH
to

•*
to


S
TH
1
to
Tj<
00
CTi
in
i
CM
O^
to
00
CM
1
h
ID
•a
ri
M)
(U
z
ซ
V
w
2
s
J?
TH
TH
CM
•*
TH
Ol
CM
S
\O
in
to
CM
TH
oป
to
TH
to
f^
t^
CM
TH
•*
TH
TH
to
ง
CM
in
s
$
TH
iH
a
8
IX
ง
•*
10
to
ง
CM
in
o
TH
Largest Positive Error
K
TH
ro
CM
CM
CM
tx
CO
TH
in
o*
to
S
o
o
CM
tx
in
TH
o
CM
TH
CM
00
TH
CTl
tx
00
CTl
CM
CM
TH
a\
CM
CM
O
ro
.
ro
S
00
in
Intercept, Least Squares
fv
00
CM
TH
(O
TH
R
CM
TH
Tf
s
s
•*
(O
CM
3
•*
f.
Ol
*
S
S
o
tx
00
CM
TH
>o
TH
in
•*
CO
TH
CO
TH
to
o
TH
o>
Max. Measurement
%
S
ro
1
tx
in
CO
Oi
TH
1
CO
•*
CO
O
CO
TH
tx
CM
CM
CO
1
>o
o\
•*
TH
CM
1
O
tx
S
t
CM
in
Oi
ป
•*
$
8
CO
tx
IO
TH
1
CO
•*
s
in
i
CM
Ol
to
00
CM
1
Error at Max. Measurement
S
0
o
CO
O
TH
in
o
8
i-H
1
>o
^-H
^
5
o
g
T-t
1
8
O
CO
•*

-------
S
o
o
OH
0)
o
3
o
00
rt
a)
1*05
O
rt
O
CM
CM

O
rt
H










jj
C
00











a


CO

ro
oo
CM
tx
CM
tx
TH

TH
o






o
Statistic

Ol
CM

CM

ro
oo
CM

ro
ro
CM

TH
00
ro

in
oo
(M

rl

oo
CM

in
CM

ro
TH
(M

ro
Number Cases
TH
ro
in
TH
tx
oo
oo
TH
00
in
CM
TH
TH
tx
TH
CM
o
Ol
TH
m
ro
(M
ro
ro
8
%
CM
Ol
CM
ฃ
TH
TH
CM
o
oo
CM
O
ro
Mean Measured
8
TH
CM
8

00
iH
00
Ol
00
r2
TH
tx
ro
01
S
TH
ro
8
a
in
oo
8
ro
S
CM
TH
ro
O
in
rO
Mean Calculated
ro
TH
m
o!
S
00
ro
tx
O
00
CM
tx
3
00
*
m
in
CM
TH
ro
CM
CM
ro
00
ro
oo
TH
ฃ
s
TH
ro
O
O
in
TH
B'I
PL)

tx
00
CM
TH
iH
3
ro
ro
TH
ro
vo
CM
ro
52
VO
TH
Ol
MAX. Measurement
tx
CM
S
s
TH
rO
1
CM
Ol
m
TH
rO
CM
ro
ro
ro
ro
ป
ro
CM
CM
8
01
tx
t
Ol
f
rx
CM
Ol
s
ro
Ol
TH
TH
1
s
00
CM
CO
IX
CM
3
Error at Max. Measurement
in
'
8
o
oo
ro
CD
rO
O
TH
1
Ol
TH
o
o
9
8
9
00
TH
m
9
TH
TH
1
in
o
i
Skewness of Error Dist.
3
tx
s
O
ro
ro
10
Ol
iH
0
ro
O
5
CM
O
TH
CM
TH
o
in
S
m
S

s
o
Kurtosis of Error Dist.
                                                         C-31

-------
          t    Averaging procedures 1 x 1, 31 x 31, and Qฃ generally
               produce the highest linear correlation of  calculated
               values to measured values.
          0    Averaging procedures 19 x 19 and Qฃ generally produce
               the smallest RMS errors.
          t    Averaging procedures 19 x 19 and Qฃ generally produce
               the smallest mean absolute errors.
          The averaging procedure QS produced the better calculations
and has been used in the validation results reported in Section 3.2.
C.3  ANALYSIS OF GHM CALCULATIONS IN TERMS OF PARAMETERS OF THE VARIABLE
     EMISSION RATE ALGORITHM"
          The hourly area source emission rates were derived from annual
means on the basis of an algorithm which is a function of the hour of the
day and temperature.  An analysis was made of the hourly calculations
obtained with GHM using this algorithm and the five averaging procedures
by temperature, hour of the day, day of the week, and month of the year.
The mean and RMSE of calculated concentrations, without the addition of
concentrations from point sources, for all stations combined are shown
in Table C-23; a summary for calculations with point source concentrations
added is shown in Table C-24.   Both the measured and calculated values
exhibit a strong peak at 7 a.m. and a weaker peak at 7 p.m.   The amplitudes
of the cycles generally agree  except at 7 a.m. when the calculated con-
centrations are relatively too large.   There is no discernible variation
by day of the week.   Both measured and calculated values show an inverse
relationship to temperature, and the  amplitudes of the two distributions
                                  C-32

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
 I
 ri
 O
O
 o

U
 (U
 o

 o
 -
.3
 o
P-,
 +J
 s
 o
43
 ง
 5
F—I
 3

"rt
O
 o
U
T3
 u
•M
 o
 a)
PJ
 I
O
 rt
H
c
6
•H
•o
X
•rt
O-
•*<
X
o-
•H
X
CALCULATED RMSE 1
cc
u.
k—
en
C.
l^
ar
^
CALCULATE!} RMSC
1 CASES MEASURLH
in
3
*ฃ
or,
rg •-+
CO -O IT* J)
CO O
a i/> O r\j
^ in co o
i
in
Z *~
^ 0"
o c*

O *ol *O CD
•n coj >r -j-
rg ft rg r\j
f\i >o| rg f\i
ff"
c-
S ^
O fM
rs, rg
•4- in
-O ฃ
0* '.n
a- o
p*| ?• -*
rg
O fM
n") 0s
o o*
•c- rg
5S
Qs ฃฃ
rg IT
O r-
!n'Uo
r- O
rg rn
* oj fM O| rg C"
TT
2
•O
r\:
CT-
rg
„ฃ
ง CO
2 O
CM
o
O fM
CO ~*
r- -4"
*-*
rg.
ss
r<> O
rg m
fM fM
is
~* m
>r ^
rg —•
30 P-
co rg
O f
mr-
SS
J- 0
ฃrฐ

-
•a
rg
—

J2JS
J.
Ct> **"l
-to
o o
rg rg
rg ซr
r- O-
21
in co
f> CO
a- — .

r* rg
SX
cr in
— ซ rg
CO O
•4- O
rg rg
-r -4-
CM rg
rg
P.
a-
r-
-
X
i

CO
p-
m
rg
rg
rg
m
rg
rg
rg
eg
2
aj
0"
—

O
rg
(M
fM
CO
(M
CC

O
(T
-

rg
•a
rg
*
C1
rg
in
rg
O
O Z
O
in -•
rg a)
O in
i ฃ

O 0"
gj m
fM rg
O rป
a* j-
rg fM
^
f- O

i n

rg O
3* in

J3 -J-
-vj o
O ff
rg rg
ill O
3 at
>- I
rt O
~* in
co o-|O m
-O r-
>)• m
ir-

r,n
rg rg
O —
•4- m
CM rg
*^
in ao
O fM
fM fM
o a*
rJi

rg — ซ

O ป**
30 r-

fM m
rg rg
Cg CD
rg r\i
r^ m
X
I tt
K- U.
CO CO
*n o

'*'„;
rv.ro
rg <*•
•o in
;M rg
-• CO
o J
— m
ic>
— rg
r- f!
f\ O

rg CO
fM CO
r- O
-D O
rg in
~+ m
rg rg
< S




















MONTH
CO fM
ฃ5

"g fM

rg ru
SS
O CO
P- O
m ix
co r-
tO O
co in
CC CO
r- r-


(M P-
0* O
•4- rg
•c o^
o r-
o —
r^ m
m rj1
Z *
-1 U-

pn o
r^r>
ฃo


rg fsj
rg rg
*o>
pป rg
O -^
o a-
O1 CO
fM O

P- CO
^ o


;^:f
r- O
m m
•4- — •
rg ^M
ao co
m rg
^- in
X -J
(J -^
t X.
y 0'-0 O
r^m



r- o
22
CO rg
m o


tM (M
0 -•
.-M rg


^r-
SS


rg rv.
a- —
O rg
rg rg
co co m m m -4-
? 5
O C1*
CC P-
ป P-
rg fM
n .*M
in 
CO C1
0 2
N. 0
•ป rg
o —


rg rs. in CC
O ^
O 03
•M -^
a> ro
rg —4
mm
UJ
>• z.
^ ->
-4 m
fM fM
CO P-
P- CC

in <
h-
^ ^
~t O
-j <

ซ J
o o
O rg
rg rg
r- — •
O- o-
-4 rg

0 O
^ o
r— ao
rg fM
— i eg
rg  O
O HI
z o




















a.
X
LU

•4- CD
rg --

CO CO

3* %T

n- r-
o ->•
a* in
CM rg
O r-

gj o-
•O rg
a- o
rg g}
O CO

•o -a
J, g}
M m
0 O
-• o-
IM 0
 ,ro ^
?• *N) ^ &*
cogj
rg rg

O fM
P- 
•O O
r- r-
S5

*-
ฃ3

rg o
in o
CO rg
rg rg
rg rg
ป-
3> ซ
o o
rg rg
r- in
,ป• rg
•VI O
O c-

O CO
=5 f
tO 43
^ r\
•4- O-
rg O

o> -
o •ป
rg m
O rg
0 r-
rg o
kn rg
ป•* r-

O r-
j\ in
rg CM
ซ gf
1 1
ir- m
(V ift
IM O
'M r-
m rg
•- *
in *f-
CM CM
•O O
in r-
r~ o
AJ rg
0 —
ซr m
rg m
* *
in •a
i i
in IT
gf m
CO
CO
0>
ฐ

in
•0
CO
in
m
2
o
in
S
o
n
a
4-
0>
O
o
*
fM
A
*
O
o>
•0
rg
O
*4
CO
m
in
m
•0
iu
                                                     C-33

-------
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
I
I
I
I
I
I
I
 (U
 o
 a
 o
U
  C
O
 o
E
O
 (U
 s
 OJ

 3
 a)
5
13
 fi
 rt
 rt
U
o
 c
 o
 M

1

 I
 o
U
•o
 01
(M

O
X)
rt
H
as
fa'
CO.
X
X
X

u-
LJ

DL
>—
u.
i.
C
^
C/VLCULATFD Kr/SC
LJ
LLJ
_J
LJ
a
2.
t/1
UJ
0
^J
*t
a
5 ^
S

c
r\
c-
Z ^
f\
~


-
3!*T tf
; O1 r-
- -1 *
i
ft <
O —
ro vT
sj rf
- x* P-
•CM —
•O 4,
JJ -^
•ป_, •_}
r** CC
\
1 in 0" ซ JC
1 "- -c
' V C3
-M f
j ' J
~ซ
C LTv

4J fj>
-f\ :n -T
5
J-> 4)
— t,-^ fM
rJ cj* P—
Ifi;
rx:
4 — '
C*
$  r\j
J fป a
s -T ^
>;NT f<
3. -3, a
* fM —
rM •—
rM ซ—
rM r\

^- ro

.^


!n x
M -J-
JTซ cr
in IT,
-*

r-in
O f>
-- fM
CM CM
r* m
u u
3 u:
--,


CO — •
>l>c ซ
) ifj f\
i :r> r\
* o r
ro p-
^ 43

^d r~
CM r>

.0


71 oj
•O O 'CO — •>
-or-
-^ .^
o >n


fMCD
m P-
fM ro
f> ro
OC
2. OC
t- u.
-- fM
2S
c- irป


-if-
0-'
CM CM
O fM
si








O f
•J- r\
fM f\
i-l -M
fM O
a* o
Sฃ

a- o

CD ^B
'J f^1
^ CL
rx rs,
Ib f
r ir
•*• IT
-- cr
CT- c-

SS
^S



fM *M
> o -
-> c;
o c
•u 'j~
CM 
•*- fM
O 4)
in 4i

0* -J3


P- CD
CO O
o> *-ซ
-a- o
> D
-J 0
-ป ซr
m m
.n f\,
o 4;
me
p- —


o r-

c* o*
%r tt
7- 'j
•3- C?
r- P-
'J* fx.
•i, (71
rsj -\

.-M fM
3- 0
M -0
2 $


-0
rx p-
K* m
0 ซ
O CM)
4) r-
•-4 *^
> fM
0 O
o J
-* fM

CO —4
rn o
D. ^~
%/\ C
rsi fM
> O
Z C
















a
•c.
UJ

in cc
P- fs-
rvj —.
CM fx
4) O
O -C
fM o

(Nj ro

P- fX
aj O)

•O sn
in —4
cr to
rM --
rM PSJ
O 0*
!M J3

— o

•n fi
en —•
* t
j-
_i-
rM —
O* C
-- r\
—i ^>
m L"I
."M fM
^T X
-M rx.
^ซ

o rsj

n fM
"- J-
K C>
J fM
5* CD
3 fM
*J CC
"•"I

^ T

or trt
n in
rvj IM
1
rgfr.
fNt C
J- J
J- ^
n a.
^> en
Zj (M
^1 ^
I'M 0
r •
S "

— 1 CC
o r-
J 0
t- CM
-1 O
o a>

_, o>
2S


CO tf\
ปn P*
CD f>
CM CM
-i 4)
-j- m
-r sr
m 4;
1 1
^ tn
in
33
4)
'— 4
,.
(X'
'M
s
__,


_ซ
41
a>
P^
rs
O
AJ
ซr
?
43
ra
(V
s
•0
ni
•0
fNf
a>
in
in
0
UJ
                                                      C-34

-------
agree well.  Overall, the QS averaging procedure with point sources
added produced the best results for 20 of the possible 68 comparisons
of means and RMSE's in Tables C-23 and C-24.
                                  C-35

-------
                   Appendix D

FREQUENCY DISTRIBUTIONS FOR SCIM CALCULATIONS USING
        3 TO 96 HOUR SAMPLING INTERVALS

-------
                               Appendix D
            FREQUENCY DISTRIBUTIONS FOR SCIM CALCULATIONS USING
                    3 TO 96 HOUR SAMPLING INTERVALS
          Tables D-l through D-10 show the number of calculations,  means,
standard deviations, and frequency distributions of SCIM variable Q,  S,
and H calculations using proportionate stratified sampling for various
size sampling intervals.  Each table shows calculations for one of  10
sampling locations in New York City.  The sampling interval  was varied
from 1 to 32 3-hour periods or from 3 to 96 hours.
                                   D-l

-------
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
I
I
I
I
I
I
I
ff
o
•-
in

U
00
c
o
cc
g
n!
 4)
H
Q
•s
Cl r-

p-
rr
O1 oa t—
O> IH P-
m o| m
•J1 >ป•
__
•T in
-j r-
I
p-
rv
rv
•4- iri o
m in; rv
j- in


rr- ca
^
•J-

a
X cd rr
co -a —
o >r rv
rr, 4)


—" m
in r-
m 0*
•J- P-
(N p-
-i- -r

p-a
*O f*~
l/> Ct
X
•T
rv
p-
rv
rv
0
rv
%c
•tf
a

-j
•O rS C
rvj Cj rr
* 1 rr
1
-j- d P-
-J- cq rv
~O rvi rv
0 4 -C
rvi f| -v
•T •*] •ฃ
•j
.n rvj cr
rvi cq .j
O CT
00 •ฃ
-. &
~T <*.
-i C
\f\ f
•J- -4

ru rs
1—4 r*
o o-
r*" cc
rvj rv
•J- -J
rj- CD
rr\ &•
-I rr
— ' c
rvi rv


z
c

t-
<3


u.

c
a:
^
r-
•*

rr
rr
rv
-C
•ฃ.
•3
IT
rr
r~
^
^
2
•I

c
CT
tS) ^
UJ —
-J IT
Z
UJ
o
or
UJ
a.

•^ T

*— 21
^
32


ซo >o
rA p-l c* rvi1 p- ^~
P- P- rvi C7> r^ p-
in CP in o, ^ rv
ID p~ f*^ i7*
O o' rg o
^; in
O^ Qv
o O1 *o r*-1 0s f*"i
^ *-i[ f^ ^ r^ o
IT, m >f in iri o
%f 4"! ซ-l X IjO ป*•
f\J fSซ ^H

f\J sO


rvj p-1 cr> rv
(*^ 00 *Q r~*' \f} lO
^ ซ-~* co O' ^^ m
'^ CT* f^J GO OI r— r^, PM f-
r- %O' — < cOi >o -4"
^1"
1
O (7s1 ITi OO] ro — i
O col u^ r\J; r\i co
CT- ODj CNJ f-' •— i CO
•-^ t*~ ro (X1 co f^
ir\ f-— r- o*1 f** '-'
*j- nJ Oi r\t vO
CO CC *N -J-: O <3
m (T1 rซ ^ *r so
f\4 ^rf f^ fO
ru cd eg o
f\J H ^
O "4 0s m
co rv
in NT
rvi rv
O r- — < r\* -C •-"
CT4 (Nl CO ^Q tjT* \C
•-ป O ^^ rrป, IT !**•
in (^' o >^"i CT* >}"
•^- r\< | f^ CT*! ^ ^

co C| cr CT*' O rv
(•<•! cd -j- r— co r*^
rg ^i ca cc
oj rg( ^
0 J% >C X
.r -j-
ro sC
CT1 0s f*"* U"\! P^ IP
— • cd a^ r\j
U~\ O*\ O rO
•T ^
r\l ^
0 -
r^- r-
m r\
^ 0
-* *rt
co rv
rvi -v
o •ป•
r j x
*-ซ
CJv p—
-H -O
^ m

^o j*
rvi co
-3- xi
O P-( -H rv.
C?v O CO p-
— • ซCS CT rvj
.n CDi "" P-
-r ^
rvj rv






in o
3* O
CJ* Qv


rvj co
"*






0 0
•s\ o



•j- o
m >f

rv, —
rvi ^I
<3 0
'VI —
a- -i
in -j-
•o >t
J^ IT
O rv
m m
co j-
in •*







O C
o a
CO r-


O J-
o^ t*-
^  >j
h- m
o^ -^
r\j r\j


ro o
m •— *
^ o
o o


r*- fO
ni cu
no ^r
& o ปr ro
>ป• — > rvi P-
m — ซ m
rvj (vj! — —

co fv in C3^
r^ mi ซo rvi
o ^ m p-
CT- 01 CO -0
rvj co in rvj
rvj — i' —i —
i
CO O
-r o
cr O1  a* m
sj" O
r*) rv
0 -•
O U3
-3- CC
r^ rvj

rv. cc
O CL
•o rvj
-1- O
•* rj-
rr- rvi
CO O
rvi in
r*i Qv
rvj-
in f\
" "
m in1 P- >j-
— • -C' Li >3-
p- -r; rvi o rv
r\i ^H! ^-ป ^H


rvj OJ IT O
r- — ( >r ป
rvi co| in rv
""~
in m in p-
o in in -j
>D f\ p- col co in
m p-j rvi CO; L-V r\
ro ^ปป

rvi T
r^ m
IT ปO
rvi x
*^ t"*"
m rvj
m co
rvj — ซ
O P-
— •ฃ>
NJ- r^
rr< rvj







O O
O O
•c in


!rvl -^ -• -
1
m of p- m
ro *n j^ -o
— >o P- -r
co a-
rvi OD
rvi -i
rvi -~
-C rvj
-,-
CO (V *-4 O
o> rvi ro s*
in a-
ov r~
"VI 0"
rvi —







O O
C3 0
%}• m


O -0
rA r-
*O rvi
f-t ^H







0 0
o o
rv, —i


P- O
0 0
0^0,00
-t o o o
in a
o
"~*
-ป• a-
-" rvj
0 0

o> a-
rvi rvi
lf> >jy >JQ *Q
-* rป- r- f^-
^^ '

i
J-
co o^ r^ f*-
^^ CO *C *O
—^

•^ ^* ON ^
•— • (7* P4 f\J
m CM *j- ^
IT X
"^ r^-
-*

r\j ro
0 0
p- P-


•o o
^^ lT\! 'C rvj
m -• *O ro

^ cc
•— i

S cP
p- p-1


CT. o
cc rvi
p- rvi: o r"
co co! o in
O P- P- 

m p-

o oi ^* r^
o J o 

in rvt
^ rjป rvj pป-
in o; -T co

o PI
•— t







r- IT.








o ^ 
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I








,— 1
=tfc
G
0
•M
4->
a
00
O

*
g"
(U
2
g
a
H
bO
G
O
_]
8
<4H
•3
&
0)
JEj
DO
3
a.
00
MH
O
tt
(11
H

CM
Q
—
^


o
z
>— 1
_J
a.
i:
^
uo
,

















CM
m
<ป•
CM



•o
~H


fM







CO





•O





-1-




m





CM



->




o -r
P- CM

m
mm! m
lco in

(V
CNJ
—1




CM m
oo
-4-
fJN
O* CO uj
CM -O1 in
-< IT i CM
in CM
00 If.
CM *n

-t r-
-3- xi
~> P-
CT
oc
c
in
cr
•o
-4-
m> m co; CM
coir- a*
f_4



-r
>r
rg




m
CM CM
CM m
r-
co
00
in cj| in
00 -C CM
in in! o>
•o m
CO
CM -4"! O

X O
in
CM
mm CM
ฐ:ฐ *
J- CM! -ซ
•OJP- CO
m




p-
XI
•T




iฃ,
CM
P-




f-t
e'-
er




in
in
•j-
--4



CO
O
O
rg
l/)
Z
O
10
M
or
<
Q_
X
o
l_>
J-
rg m

m cr
in
P-
j-
00
o rg! m
in ol CM
cc cr
-0 0-
CM m


a-
X
o
in

•cm! CM
o -o
•O tM
IV
*
o mi ,-j
•C l/M IT
CM m1 p-

CD P-
0 .-M
o a-
.4- cr
>O CT
CM m

-> 0>
ai in
X CT
m .0
p- ro
CM m

•4- CT
o m
P- in
j. f-g
•O O
CM •ป•

Z
o
**•
t-
<
**
>
UJ
a

a
or
ซ
<—>
-1-
GO
m
CM
a-
CO
c
in
CM
CM
•1-
ป-<
xซ
p-
-X
rg
00
in
P-
v/S O
UJ J"
_i r-
Z
UJ
O
or
UJ
CL

0?
— M
1- I
3
CO
O CM
O CM
cr o
m •o'O — ' P- rg
in o co c>;p- co
co o'm inloo co
J- CP|CM CC CO -O
cj. ^>
^H ^H
o in
m rg
•^ ^^ V *W <~— ^J
—• -Hl^- <ป-|Oซ CO
*O >OJO CM. 00 O
-ป• ซr!-T m! o -o
O O ** ^ip- -4"
o ojco in,m CM
CM CMJ
1
r\j co
.n P-

ซr •ป•
if\ co O *^i^- oi CT* OJi ^O ^
J- J-|r- m]-< -1
fj- o o> -r;m in
CM o co in
CM CM
o-> co

in co
m CM

J- CO
<3 oo' oo m'm -o
•j" O*1 co •— * co m
rg inl CP O
in -r
P- p-l O> -rl oo P-
x> .n
^^ r-4
o> -^
ao Oj m CM


CO >T|CM in
in CM! r- mi -c o-
-o g3| cr o**' -4" -o
P- o. <— * crj in oj
en tn! o c>! -J1 >r
sT aoi co inj m CM
rg -^

1
1
1
p- colp- m o m
ox*—* oj •— < -o
in o>! ir. p-1 co co
-4- in
cr- P-
so cr1 o m
-J- 7-i 43 o
CM in co m:m CM
CM —•
p- •ฃ>

cr -o
O o O P-
m •ฃ>

CM on
cr o
o oigj x
•* OIP- ol m P-
3< ~*\ — co! xi •}•
rg co j a) in
rg ^
-^ cr
o ^^
p- P-
m --
in CM
-r cr
rg *+
o -<
i\ —>
*4 <0
in j-
JO 0

CO *O
m CM

O CM
<• gj'— < p-
—• cr
rg o
>i m
cr -a

in cr
* P-
OC -I"
o cr
g3 in
m CM

00 CO
p- m
U3 CTJ OO CM
^- —J (V — i
o o ,n m
m o a- gi
rg rg






in o
cr cr
cr cr









m rg







O O O O
in o
c- cr


0 0
ao P-


** m
in CM
-4 m
in rg.cr %fi a^ -o
cr m rg *f!rg o
m m'in in'in -o
^^ -o *-< co , -o -4-
rg -<


rg co'r-- p- in gj
P- rg|g3 P- CM O
CM r*!m O" — ' -o
cr injo P-.-O -4-
- -'- ,

cr ^
i
1
in cr o x
o *r:m <- P- o
P- CM O -O CM O
CM -o in -j"m in
-ซ -Dim CT!P- in
CM -4
in o
"
rg in

CO -O
o m cr p- 'M P-
CO f-ซ!cM P-,O O
r- -oij- crim P-
x in,o r-
•-4 f-t

o cr
-

p- rg
•a -r


m i-t
-j- cr gj Oi-o o
o x; in CM' CM -*
in CMICM o P- p-
cr in rg cr ; gj st
r-l *-^; ^>4

in in

•o cc


P- 00
O x'm rg X o
co CM! co •}• cr oj
f— mirg o ^- p-
aj in ^H co -o ^
1— t -^i r*






m o'o o -r CM
m in -^ o -f m
CM O -O '-''X -D
-^ rg m m
cr 'j\i-* cc
p- 1 -a ป*

in CM

-. p-
•c *


-r m o m
-c. -o cr -j CM o
m X[O r-<
ir- m
co ~r
i— ป -^

p- J-
CO CM
o ป
i~4

CO >f
m p-
>O -I


CM m1 m in
IT. NTi-o rjjm m
m -j(_. m
o m
cr cojrg m!-4- oe
^o >r 1 — ' a, i -o -r
f-4 ~*

•r o
j- x
co rg
_*

tn o


CM >}•
m -r'r- O>
N* F- -o -o -oj
m rg
CM CM!
1 i

-0 *
p- in
P- CO
r- o
m rg


CJ- CM
p- in
>00
o r 0
IT -"
00 0
$• in
p- o'm in
00
-r m


P- fM
P- <•
CM CM


>o o
cc ^ ^ m
p- -o -T m
— n
•f rg


CM m
•o ~*
co .0

-• p-
o —
O OCi <ฃ> C*
-r rg








3 0
in —>



CM — 1








m jt
0 _)



D-3

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
m
=tfc

 ง

ซ
4_)
to
O
 S
 iu
s
  irv ,-4 *j
co o' m >3
•-• o! **• CT
f\J f\
cr a
—i
I NT cr
in mi in P"*
O O, f\i r-
m NO
r- m
; in NT

-^ CO
i o mi
i rn o
m tf> oo ^ r* o
in ir*; r- rsj; o in
c> o o cc
eg rsi! .-i
f*^ mi m r*
sjQ CO O^ ^*
CT^ a> a* N
* in rrj
1
J NO CNI
O O
O CO
J CNI NT NOl NT —> OI (Nl 1C NO
> O NT
ซ NT m
m NT
f\J
00
sO
Is- -4 Is-
i- m< m
co mi IT
co ml NO
m NT
CT* in
0 Is-
r- in
r*~
r*-
sj-
a>
i o 0s! m<
> co cd co.
co mj o*
J ^
co i-[ NO
•T NO, Is-

J NT -T) in
ICQ NO
co NT* f*-
1 <"
--CO] Is-

NT oj cr>
—•mป m
X Is-

o o
O *C
O NT
CC LP
f** ,7*
m m
O *J-
m
ro

IT
(\
m
n
CC
p-
rv o >}•
-o 3 cr
^d m
-O Oj m
rn >ji CT
m
NO J m
— . u~t m
•T oj NO
X O l/> O
m, NT
LU ~-
-j -o
f—
Z| ul
a u
*^ CC
!J>. ,jj
 ir*.[ m Of CT* rn
IT\ IT*1 "^ C.H *A *d
a- ^
in O"
m h-, CT^ f\
QN f*-l f^ O*
mi NT
NO — 1

VT> IT> O CO U^ O
i/> f*- 33 fN* — * f-
O *4" O CO l^ rซ
rvjr^-
CO M O CJ1
in -T, m. ^
pg NC CT1 -H
O Is-
NT Is-
-H a
NO O| NO Ndl NO NT
NT mi' NO o
CO NO, — X
"f
—• Q in CM
Is- Is-! in -H
m. CT*
in mi
CO Is-

Is- r-| o — < oj pg
IT, i~-i Is- -ฃj
r~ co1 ^ -T
NT O O Is-
rg Psjj — <
x f-
NT CO
-H NO, mi rg
o-> m, m Is-
,j. ^! ,r m.
O> NO O O
co cd -H co
o m
m, a-
in m,

CO Is-

NO CM
O- mi
m a*
to mi
NT CO CNI Ot fN. in
O L.",! (NJ m
ซT 0>
Is- -<
r\j -T
NT 0>
NT CO
-si —
NT .""1 fNI -i"l
cc tri a* co
m, 10
o r-
-
r*g -
0 -Tj in aj
,*\j C7*
m m
,,
— m,
-• ON — ' NO) CO Is-
_! --| — < fj NT f-
o ^- F-* mj NT CT*
m. 0N -H cu
INf -—1 -—
j
CO CDJ NT <->
NO mv rn cc
NT o) m, u-i
x Is-! NT IT.
O NT
ฃ2



in o
CT* CT*
C'* CT*
O ~"
—i CU



00
1-1 a

m m

o m
O NT
CT> O
*O CO
mi ^*
in m



0 C
o c
co r-


o o* ซT m — c
mt c*g rg rg tn cr*
cr* -
m. tr
m. o

co cr
i — m
in m
i CT1 *J"' "^ f
f\j a
rg ^

' O* cc
oa c
): NT O


NT >O
cr* (7*
Is* 0s, t*- m
04 (Nl O NO — CVJ
c>* m CT* in co —
01 O
NO m
Is- J
I— m
— ,.
ซ— < i—
rซH -^
m r*-i r*- co
r\/ Pป-
CO CT1
NT m
o m
Is- NT
1 NO NC
m. oj' eg -^ -^ -H
o oj a* ir
— m
Is- C
co in
10 C\j
>o *r>r *r
m ^^ ^^ ^^
o m1 m m
•o m m. m
cc IT m u-ป


CO CC


CO CO
O •— " •— • <-H
•*• '-'j ^ -<
CC r-ป' — . — .
CT. f*i >^- f*.

CO IP

i in in
m m (v rNj
rg ^DI r— r*-
CNJ o , r*- h-
O r*; *o o
- 1
co *ol m -*, co co
_H CT* CT* P-
— —
in mi r- r- •-> -ป
O NT 35 in NT 0,
OJ OJ
CT* a-
co a
in NC
O IT
f*J NO
r-ซ ^H

— 1 (N/ NT NO
X NT

O CO O NC
O NT
—" Is-
m, oj oj --
•T a
NO CO
in in
O CT*
o co CT* m.
a in
r<"t O
-•-
-* r-
^ f^-
r<> t"^
— u-, O -O m o
m oi

m, NO
^ u~
NO O*
in m
— in
m, 01
mo
— NT
-O NC
NO NC
0 NT
m oj
(X OJ
NO Is-
in jo
!>- fNJ
m. o

in NT


NT mi
O Is-
t—t i-4

oc <•

m. IA m (NJ
r- O1 -T c-
O Is-
Oj -.
m, NT
rg in
-T -ป
o in
O NO
rg -^

^
r- rj
m o
** —
0 X
cc m
o r\
0 CO

1— t .—
O X
CT1 ^J
^O ro
-O CT*' NO CT*
o *oi m o
OJ —

OJ Is-
^-( 1—

ป-* o^
.n on NO co in CT*
r~ -Ti r- ml m. NC
3* NO
O -T
m. o,



•T 0
0 Is-
OJ-H



o ol o o
0 0
o ir
o q
>T m
cr o





o o
o c

^v f^| ^ซ —ซ
cr f—1 Is- r^'
1
Is- co' m. o*
in NT! *r Oj
in o Is- *n
CT* OI Is- M
co *o in in;



or-'——'
^3 r-l— m
r- M— CNII
CT> M Is- *O'



NT P-i CO O-
X> CO >T O
f-- c* o m.
-1 NO
O Is-
CT* *Oi NO in

— m
i
O m,
in cc CT* ^'
NT M Is- O^i
NO oj| — in
CT* H NO *T
I
o u"J in o-
NT m. oj cj
IT* NO! NO ปn
OJ 0


OJ NO
-H
•O in
i
i
r*- mi
(NJ •O('-> -O]
NT oj in a*

-------
I
I
I
I
I
I
I
I
I
1
I
I
I
I
I
I
I
I
I







0
=tt
ง
.d
Ifl
C/5
u
z
^
ง
. 0*
-j- o rg
cr cot co
o m f
mm o


rg

-4 in p-
oo co m
r- co .-4
& P" O
"4 cc r-
m rg •*>
rg

J- CO
p-
r- o -o
m in o
j- ^, rg
rg in m
mm in

0 rg
rg
j.
co rg r-
rg r-i in
f— >J- -J-
rg m rg
mm' o
m
•o co| p-
00 O, -C
rg P-, O
rg rgi (Ni
rg •"•*; m
m m

ฃ

•t -* rg
X 00 P-
-t J"
p- rg
rg r-
rg <•
m m

z
0
M
I—
<
>
UJ
c

u
-
gD
**
~-4
rg
.-4
O-
i/) p-
LU m
_i ซ
z.
UJ
u
a:
UJ
a.
z I
O 'J
t- X
3
-
O -0
a> .j-
m >-4
co .j-
_ o
rg -4

in in
p- r-
~* —4
cp cr
.—4 *-4
0 0
rg rg

in in
*j- m
P- CO
in er-
go cc
r- O
,— 1 —4

in r-
x cf*
O- rg
0 CC
p- m
0 0
r\j rg
in rg
-r in
P- 00
in in
•O P~
p- g;
^-4 i—i
in J- ปJ-
p- p-
— 1 ul
x -c
O P-
og ป-*

X P-
cr -i
o m
P- m
O if
rv, -^
P- rv
ff* fN.
— * ^
O r^-
r^ (— *
fO C* 00 CM
a* r- co m
u> N-;m rg
0 fvo fxj
oo -o;-^ m

i
ro rvj'rvi o
fV >O -^ i^O
(\f ^~ , l/^ 4^
Ul CC O ซJ
t.n r- co o
.-^ j^ m rn
•-^
i
•O O1 CO -4"
ro f>
0.-
f^ rn ' (^ I..O
co trvm r-
•— * IT* ro if\
m r^ ' f^ .^<




f*" OiCO f\
CNJ o^oo ro
*o •& i r^ f\
O"1 o'**l ^
f\j ^- ^ (j- ro


fO f*^ t O 0s
m rป jv .-*
^ >r
xC (\J
tr> o 'm o
•— >r m ,-4
O >o -j- ro
•—4
m ir>'f\i o
r"\ -o •—• ro
o tr — < -^
t/> CD' o O
-* -< (NJ *-<
O 'O ^ rO
•^
(^ ^.

O ^
** LTl •ฃ tT
f*l ^ NJJ CM
OT oi o  o ปy rxj
i
i
CNJ r\j m y
co O' >r •wD
0 >3-,Ov rr
O cu'^ r*-
O r>- r\j ,-*
cr tr.

'
^- f*-i


r^ C r- C"
O cu; ^^ r>j
^ O (N ^~
r- o m CM
30 L*. ^ fO

LTl >f

0 ^
f*> u~ J* c^
f— tT' sD — *
-4 O
Of 0
•wO Cฃ>| CD -C




Op ~*
PO h-
m r\j
<• rO

j^ 0s
^ O( P*- P~
0^ co|^ -*
cr O1 ~H rj
fi 17*
00 lT



ro ro
^- m






Q O
in o





O O
O O



*4" O -4" •ฃ>
o* sj- f ซ r—
*NJ f\l • fS* ซ— *


^ r*ir\i ru
•-^ i-^ r* r~t
m sf
j- m
o^ r*>. m o
m o r- -f
rsj oj

•—4 <~t

O ro .-ซ J-
O *-*, *Q *f
f\j CT* ปO O
o* ~t'r- ปo
>iJ m p" *ฃ
CNJ f\J

^^ ~^

C^ •— * O fvt
00 O' O^ "**
00 CT1 OC -^0
-^ -* CO O
p*- oj1 OC ^
r, rg

"^ '^

oo f"-, rvj pn
cc r\j o cr
i^* \c o in
ปf ^1 frt o*
>c CNJ &> in
CM (\j — ^ i— *


a- co 'i*- —i
•~~* ^ I'M g3
g; g;|-4 o
O (Ni
rg rg

cc o
cr 
o &
—> CO
^H


O C7*
m ปn
rg o
gj -i
m CP
^4


(J* rg
rg a
^- (N,
o oo
rg CT
_4

CP P"
•J CO
o m
CJ* OC
rg o^
^^

P- O
TV I-
iT\ P-
p- X
M O1
F-

-i- in
m in
O O
in in
(NJ CT^
^^


— O
X ซ
m -o
rg CP
,—4

^ ปo i c^ in
in rg ซj" m
fN. rgjrg o
j> T> m rg
-O !M|S> -O
rg Ml -^ ^-<

•ซ• o
m M
1} O
^* *^
o M jo '~g
rg aj ^^ m
p- rg ^ g3
rg M






0 0
0 0
•C in


•^ p-4






0 0
o o
gj- m


p- m
O cr
^ p—
rg J-
^^

m M
•-4 ^"
rg m
rg r^
m o-
t-t






O O
0 0
fN, -i


CO -J

TV m i m m
o — • ^ -*
co n,m m





O rsiirg r\i
CM rn ro m
^-< .X1 lA U*.
^0 sO 0 0
fv. ^ ซ^ 1^



* o -o -o
m -H P- p-[
gj g3i ^ >4*
gj m co co
r- gjiin. in



r
X gD. i/v in




-4 I*-'O —
rg OI-T -<
p- p- -> r-
a- r- >f M
P- in .n •*•





"^ CO \* - N^
-4 rg co x
O f-
O -"
rg g2
co r-
x -ol in m
j

x o

m CN,
x mjc* (N,
o g3.p- cc
o cc rg o
OO ^ 1 iT\ o
O CO -4 f-
x m







0 0
in -H



m m







•f JE
O -1



D-5

-------
I
I
I
I
I
I
i
i
i
i
i
i
i
i
i
i
i
i
i
00

O
 cu
2
 r
r~
cc
D r-i JC
ป-4 rj IT
1 "
•a N -ป
— ซ ir( N
•J- Cf< —
-T r"
CO IT
:--
z
c
—
t—
<
•—
>
LL
c

c
a.
ซ
ct
i/> a
LU U"
_l re
K*
z
LU
O
OL
LU
a.

•z i
O (J
>- I
2
X
^ rt
u t^ ff
i At />,
00 o O ro rg r\
•O ซ\ r- CO fj> IT
O Q gj cr< rg g;
-ซ cr
in •ป
^^ f-
•o <
1 — i cpi *j-  fo a1
cc x( o in, ^ in
CO COi C^ OL ^^ f-^
rg rgj f-4 r-1 *-^ in
*-• -i rg O o <
•O -p r- -r rg -H
"1
1


•r H a> -< r- -i-
oo CN' in ro; ,^i ^•
gj >rj ^- O rg ซ-i
o J rg ^- r- r-
^4 rg p- -^ cr (71
in -sfi r- IT
" 1
t| CsJ ^
gj CN rj- O gj >r
co q ro p- r— 0
CD rrj ro if\ -O r-
rg q in -^ •* CT-
— • Cfl rg CN ^ -J-
•o j-
f~ซ ^~
o- r-
go mi rg -(
ro ol oo rj
C CN *f •*• in rซj
ro ~T\ a-
CT* r\
•4- •ป•
•ป4 ^H

o -^ r- ซ—
o -< in co
-O •*

CM ~

co roi co oi >o gj
<• o| in co cu ro
^-4 ^r in rO| m c^

r- cc in co P- o
-" d CP- —
gj -t\ in •*
*T f*-
~"
%j- pJ p^ mi *c r\
co cs .4- -J- o rs
•a -Ji co o >f P-
o >r
-^ rv
in ^
~* ^
g; ao
r- a- o —
C" ro. -c X
in  -f
-o -ป
r- x
in r\
O ro, rg o
-. Cfl CO -J-
o- —
in ปr
in co
*r r-
rg —
rg r-
in OS -o ^
O UN o cr
-j- a-j r- rr
CO -H --> P-
in .ป•
ro O
-r cc
fx> crl -J- ro
rg —
go p-
O P-
in rr
CT 33 O coj ro g;
in act o H in r-
in rr






XI C
3ป C
o> a


-D -I"






rg —






o d o o
• • • •
in q o c
o 0-


oo r-





in ซ^ i ^ซ ^ i— u
CP ro o o co y
•T O-] ir\ r^ rg f
ro o o —> in x
in o co ire ro <\
*>4 (^4!
!
I
1
fw ^!j- ^-<
-o ^ og ^- j. — i
f*- co ^* -o o f^
O -O1 CO l/M m U>
< -* CO O, ^ fO
r-l r-l
•
t
a> rg. r- cr

r- P-
m i/v co o co a*
cr r-1, ^ 4f\ in r*
ro in 0s ซO| ^ (T*
^-* O*1 f*- u^ ro AJ
^ j
O >o| CM ซ0
C71 o — * PO
co •*•! co  o
— 'ft
1- ro


3- o! o g> -- co
o- rg, — c\4 rg go
go -c p- ro1 p- p*

* — " O -H| -^ O"
rg Oi co g> ^- rซ
H I
fM ^J-
cc —i r- O
'O -4- & f-1 m *r
ป/> -^ rซd M — < >o
f- ^' *Tป QL> (N* O
M O-
•— i

- j.
cr- in
p- in1 %ป• ro
1

rg ~o -T ro
rg rgi o P-
ro — 41 ro roi cf1 fs
co rgi o- J — 0s
r\J O
f-4 ปH
r- -c
p- -O ^- (S

C1 ro

-4 0
>r *j- a- OLJ p* *j-
ro r^l cr >O O- Cy
>ป• roi o o rg C*
•0 o. 00 ^> J- ^v
^^ —
p- go


O NO! 3> O
oo ซป•] ro q — i oo
O P- 10 -J-1 •ป• S3
rg ro -- •-! ro O
ro o co goi ^- ro
-"1






O O














o q o c
• w • ^ • •
o q o o o a
go in


* m


rg —



^J" ^^
ro r-
CO IT
0 .*•

p- p*
tn in
-ป• -J-
og ป-4i •-* ^-4


—i in
in in
o cc
ro ,}•
rg rt




in in
in in
CC X
kj- >!•
f— 1 <—t



O  -4
o coi^- -f
CC — 1 CO CCl
•O p-i gj gj|
rg ^' ^ -i
1
rg —4 in in
g} a>
gj in
ro p-
rg — i

g3 ro
in in
ao co
>ป• -r
*-4 ^4

ro -*
(V rg CT O
O ro
rg ro
rg -t


o in
ro o
in ซj-

co rg
rg CT.
1—4


— c in
>j- in
CO X

ro ao g3 -st
Pg ~4
rg —i
O t-
o in

o **
ป 0
0 OJ
\m -t
f" gj
rg —•

a< IT
-r rg
rt CO
-*• -c
rg -H

gj in.
3D rg
•T -a
•f -a
IN. — I







O o
m ^



pn t>
i—*

in —•
0- O
rg (Vi
ro 0s
•—<

ro ->
~* O
f-i rj

M4







in 2
• C
O -1



                                                D-6

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I







*H

a
0
•i-t
rt
00
U
fc

g"
CU
s
g
H
f
S
"rt
>
S
00
.3
"ft
1
00
•s
aj
H
a
1
H


i—
Uj
ฃ
fO
C\J

o; ^
0
z
or
C<1
LL
O

•
0
z
~
_l
5


(M
^






CO
Ji
i—
^

o
^
<






















*




n




CM




I
in Is- -j-
.00- o

• CM o in
—4 CM O

fM
-OCM cv
.-4 m r—
vD  -c

—. CM CJ-
; .—4
IO .-M
,r— f-
CO
o
|O Is- >C
of a-
o r-
CM CM


r- -c
0 Is-
IxT f-
(r< m
O "O
f— t
0
CM

CM

r-
IT
•C,
 f-
•O f—
-* -a
m f*
C -0
CM CM

ICM f
10 -o-
r*- CM
,0 r—
a a>
CM CM

m ^
.n O
in o
.n r-
o y
CM 
a

o
or:
1 ^
CM
O
•O
0

CM
(M
r--
in

!
•j* *3" •-< in fo o
rj> m -o co x sf

co m

xi" CM ซo fo
42 r— — 4 a> f-- o

m m -. in r- -o
O Is- • N!" CO CM CM
CM
O O
.n rv
CM -4"
*O *O ^ f*"> O f^1
^^ ^^|rn *T P>J rs.
>3" -sj" | O CD T*- X
CNJ f\J >O Nf , pg — t
•-^ O O ซ" ^ O* t*-
^O -^ O (Nj1— ^ ป-^
r- r> -H (7s o c^
r-i r- co *r -H a*
-j- m ^ r-
o GO vO ^r
—i '


x *r PO r-
•—4 r^
m CM


CM O
r\j o ^M ^ 'J" co
CO ^ m lT>,O O
O-.ro ^-Irg r^
X -J- *T O"h- O
rO f\j P- ^ (\j f\j
t—t t-*
\f\ (f<
O tA
m a<
O*1 CO! -O 'O1 fi 'J"
rn i— < ao LT\I ป-< L**.
co ^— *o O* ^ r^
O lA ^C CO CT* O
•^ O *C ^ . fsi r\<
r-ซ i-H

m o o o
O o IT o

^r -o
^, _ fn -, ^- cr
•O fj* *3" O *Oi O^ C
CP

CM
O
*
o
CM

^j-
•O
j.
*O
—4
CM
f—
in
O
in
•JD
•J-
•3-
•O
CM
•3-
-J-
00 (7-
LJ O
-LJ
l_>
LLJ
O,
?: x
2 —
ป- 1
^
—
f^ rsj -c ^
^-i •— ซ
— x>
0 h-
CM CM
r- ฃ'cr o
J3 *3-
o^ r-
o  in f-
O*- C?1 CO ^ 1 rO f\
ป-^ *^ CT1 ^i 20 ^^
r*- f*i >c "T I !Nj r\i
•^ •"•
O (X

Oi co n~i
— * (j^ IA tr> ' ir\ cr*
r*^ ^ • ^ o
r^ m


-f\ O
* *


0 0
a- o
CM CM


o o
ij^ & iA O O O
CT* ^ป Qfป C*' OO ^*





*o *c o r— CM f^
^— ^^ 0s ^
l-t l—t

mm

o r-


cj- m
m 0'vO 0 CMX
CT* m ' *o f^ ^^ CT*
•>^ O h* IT1 •*}• (Nj
•"^ *"*

^ ^ ^ _!1/> ^
r-- -o **" to o cr
-o ^- o (7s o a.
j^ O'O^ CO ^ ^
•*ฃ, r— 'i CO IA sf f*
^-* ^-*
I '
I '
*r> sj-
f\j m in m
LO i^-1 Is" r^ •— * *r
CO (N-O *-H'O *\l
Q* r— • -4* r— f\j c*
ซJ" O ^ iA;-J" f\i
•"* *~*
r\i -^

fO CO1 -O P0
O* *NJ t1*- O^ lA *T
O X rg o — * cc
O O r* oi>T ^
^" •— • CO ^O ->T t"\J
^-4 i—t



f\j o — * O -J- in
in r*~ t*~ o ' 0s r\j
sO O ^ CT1 1 f^\ O
m ซ-4] co ifปi^ r0
^^ *^
Pg i— (

ซr o

n ro
Jv ^\J f^. ^ ,— t *J
O 'O >o ^ in co
3* O co 0s
•j" •— <' f*- J^
"^ ~"*
^ (M

^ x
f< rj-
,*• CM

"l -ซ
O CT* "SJ O, "ฐ X
— ^~
•^ in o 7*
%C O ' fM CT^ f*1' O^
J^ -^ CO tn. -^ ng
i-^ f— <

O st




_, m m
"- ^. O 'O1 •— • <3
sO -5" -j" r\ji *j- m
fO ^; t-x ^ ^- O
•*r o JO o! ^ m
ซ—t .-^

CO O


rt <--
iX) O if) 0^
a* cc|^ co


ro r^-
CJ> D
O CO
xj r^lm ^' Ln o
^ 0



o o
0 0
>o m


CO sQ



0 0
xT f^l



0 O
O OjO O
sj- f^ <\( ^i






o CM'CM CM
Is- o>
cr o*
a- a-
o> o
T ^
ao ,r
CM in
CM -^

a> CM
^~ ^^
in, ro
ro CM
CM -•

CM -C
NO I— <
fM O
CM sT
(M — 4


CO -C
(Tl r-<
Is- O
fn j-
tM — •


in rป>
CM O
-i •ฃ>
r -r
t-4 (—4

CO -3-
in in
•3- Is-
— • o



CM O
^T -3"
W-* l-t

L^> sj-
CM *n
CO f-]
-4 O
t-l ^-4

c*^ O
CT- Is-
j> in
^ ^
-^


\r rt
CM <•
m m
—4 CO
*^


m o
^ r-

f\l t^
,ซ4



in i
O _)



D-7

-------





f-
(N)
=tt
C
o
• r*
4_>
a
in
CJ
>-
Z
if
QJ
S
S
(H

C
 co
(Nl INI
•0 (0
X
•ป
f\j co in
(Nl -p o-
iป •! •
o (NJ a
O US 0
INJ (Nl

P- O
-o m
(NJ pn
• 4
P- CP
INJ <
(NJ (\
m p-
m cc
o m
CO CO
— ' CC
(Nl (N

m m
INJ IT
in o
,—
(N
pr
r*-
(N
^
IT
(N
C
cc
•c
a
•o
—
(N
cr
e
i • * '
1 INJ rn| .c
3 INJ o| •ฃ
* r\j po

CP C
1 P- Ol
f

p-
r-
! 00 Ol n
jp-'ol J
J -• CPi C
- INI IN r-
J "
(Nj H cr
r- d -
J- Pi CP
3 ci i^ -c
4  (Nl J>
r (NI (NI
INI -a
1 P* *n
]•:-
i CP -i
=1- v
Pi INJ !\
s
r Z
: c

r i-
ซ <,

C
p-
p-
r'
^

P
fN
CP
-"
CP
ซ
vt
o
m
IT
0s
CP
c/1 CP
UJ (N
—1 •*
2
UJ
0
X
UJ
-L
ซ 3^i
u UJ, ca v:
n a — -

a.
d <
-i *-
K I
CO
— • •*
CO f\
•r -•
*-< rr\
CO >O
1
•*• CP1
p- -J-1
^f CP
o in
Cr> n!
CO O, P- -ฃ/'
"1
1
~r >r
fNJ (NJ
•o -c
~* •— •
-O •ฃ
CP CP]
1
1
o ml
-O IA
m -^
P- (NJ
m CP
O (Nl
CP O

OJ (Nl
in f—
ro r
O uj
p- p-v
• •
r^ crv
J- .-NJ
in ซ-•
O co
P- -T
P- (NJ
(<•' f^

t ,
•O m

O CP
in r^
(Nl in
in INJ
IN/ *J
fl (NJ
CO — i
-r >o
P- o
r~ CP
O f"
f*l fNI

(NJ CO
(ซ> >0
-0 CO
CP CP
CP O
INI fX

p- m
in p-
-J- ~r
r^ -o
O ซi
ro (NJ
-J CO
P- -J-
(Nl J-
X —
fTl >Or O 'NJ
CP UN P- J-
—i -4
1
co m
O P-
p- fNJ
rf- INI
J- m
CP ^\
m ^
CP J?
-d
ป a
sO CC,
CP J-
-. -
CP C
r" (Nj
1
1
1
P- O: -O O
in ~o
0 •ฃ
in co
O in
•& to! (<^ NC
*r a>
0 J-
O fN
x o~
(NJ CP
O (NJ
m r\;
CO CO
-J- CP
P- CC
(Nl iril P- r^
~r r-i 3 fNi
P- -i-j r" ^i
1
-o O
•O :NJ| ^j +j
(Nl Ifi CO (Nl
CP P-
—• un
— * -J p- ซj'. (^ -T
-t PH O 'nl O fNl
CP rq P- *ti f (Nl





in o
CP CP
CP CP











O O O O
•n 3
CP CP

O o
x r-


p- (A
a- CP
— ' -J-
-O CO
P-  m
in >o
J- (NJ
CO lซv

CO P*1
CP CP
1
P- fNJ| (Nl -*
co — *! cc in
O -O CP (Nl
•ป• m o -o
o r-j-4- (NI
""* 1
(NJ O:

P" O
m ซ-ซ o CP
CP (NJ INJ r--
o m CP o
CP ~Oim a ซr (N,


>ป• m o D
f<> CP m >r
in CM -o o
O fV INJ CP
O P-
—4

in f\


PO O; P- -T O — i
O P- J- ^> .n 7>
d CP
^) in
>C (Nl
"
p- (Nl
.n p-
CP (^
P- 0
CP -D —• O
-!• CO O C
CP ^3i in (^
s}- 03
XI .0
in (Nl m r^
-t j-| CP irv
O rJ^i o
o fv (JN r-| in rn
^H ^^
CO pg
O CP
-J d
1
,
|
^ in' -o '*-
ro ^ rn f\j
•J- f*1 *T 3*
d -o; * — ซj o CP
vD ^J
i-H *-4
ro CP
in ^j
CP -T
_5 (Nl
P- "O
^^ r^
0 -H
3* r- U*ป r\j


Pป r-i O -0
— IT. O -U
•N. Of (" ji
O -j| "<1 O
CP H -n ro

n m

jo in
r^, co> i ~ >^ , u • ปk.
CJ ^5' CP O| co d
O — i
p- m
f-t ^H





0 0
o a
.C IT

o rJ- pri
-< fNI
in >o
INJ — <
)
X IT
O (NJ
CP -"
^ J-
rx; -


if in
CO —
-o >r
ro PO
(NJ — *i


—• -T
rr> r"
CO CO
fNJ *T
(Nl — ซ
in in
>!• — i
O -T
ซj (^
M -.

PO CO
— . f^
JD pn
^^ fซN
tNI -(

in IT
vj- (Nl
rvj -•
ci -j-
(Nl -*

1
o o|
o -o
•4- •4-j
CP CP


CO CO
~H ^4
o o1
(NJ (Nl
~4 r-4


-o -o'
"> mi
p- P-|
pn roj
r^ ^4<

CO CO)
-4 _),
0 Oj
INJ (NJ[


•4- CO
INJ O
in PO
O X
l-t


CO CO
- •*•!
ฐ.^
INJ O
•^ .-4
CO CO
— o
o m
(N, CO
*-ป

CP CO
"** 'f'l
O -O!
(Nl O
-4 .H

0 00
o d
CP rซj
(Nl D
—

r" -o1 •ป p-
p- r~
tNJ PA
(M -*





i-l -^
(Nl -Oi
(NI r-
-





O O -^ JE
J? -^


0 -I


D-8

-------
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
1
 I
 I
 I
I
co
CM
•a
 a
  O
NO —I
CM If!
in m
ICM r-
— < CM

NT
r-
•*•
cc
CM

CM
O OJ t-
cn col P-
CN CM, >C
CM O -*
ON •~'; P~
i CM —i


o ^
-3- -0
in en
CM

_^
sT
CM
r- r-i — i
o in -o


r- r-
O

-r
00 CO C--
in en NJ-
,- rr-
r-< -T
r-* CM

O> CM
cc
CM
c\
fM
Nl-
O -O P-
-• cH -j-
.C CJ.
CO
** en CN,
—i CM! CM


CM CC
en en
CM

•3"
•r
CT^ 'i)1 CC
O CM CM
— 1 CM

t-H PN-
NT r-

CM NO
i— I CM
-* CM
P^l Nj-
CM *—
ON N}-
O -*
O ON
(—1 •-*

P- NT
Ntj en
CM 0
r~ in
o -j
- CM
0 D
ฃฃ
•J- — i
0 0
-x CM

2
0
t—
<

UJ

.">
CM
CM
j.
f^
•*•
CO
CM
^
fN-
ซr
CO
CM
CM
CM
^
fN_
•J-
CO
CM
CM
•O
*^
CC
l/> CM
UJ NO
~j en
Z
UJ
O
UJ
CL
•z r
C f-9

1- X
3
or <ฃ
<.— i
C.
T S*
CC
-
•ป• oo
Pg -H
*n oo
r"- MO1 •* -J — i
.r o> (M f- o> rซ-
CO -O.-O POV* -4
CM "O1^ ON, NO o
CM CJ>
PM
en en


co en

-i CO
co colo- -< in —i
co co;o t— i fN- p*
ซo ปo ON NO en CJN
in *n co o ^H co
NJ- -J-.CM CM rf
1
1
pป cn'o NJ-

en -<
•— • -NT1 ON O" -$• in
.r o> P- PM'cn en
•O P- O O CM P--
NO injO ONINJ O
CO NO' N}* CM


~H --

^ ฎ|*^ No,cn p~
p^. **
CM -JlCM — •
ฐQ GO O O* •"•* CM
-* 0 in m;r^ o
r^ co. *$• -4* ro o
-H \f\ , ro rvj
r\ป ^
r- ...
ซx •^

CM CM.X -X
-4 ^3- rsj rซv -o •-*
m (\i i ou i-* rg o
co -J1-^ co'o r^
^A o o O — • fVJ1 ซO CD
m u"* ' "*• r* m ^
O O1 r<^ rsj
f\J -4

•>*

ป-4 ^ ITV O1* fNJ IT\
*ซO ซ7*| O f— t** -C
tn o;c> ^lm ,r
^H ^
(^ Csl
m m
&* ro
o o
-j- m co NO
cncv-
0 0
NO 0-
NJ- in co •-*
NO ป— tj f*~ Nj"
in pf NJ- a-
-o .n
NT X
•"•"
O C7-
-T CO
-0 P-
•r N-P
JN CO
NT ON
in co
ON- P*
00-
^ PO
en CM

sj- o-
en p~
CO Nf
en in
P- Nf
If. CM
ซT 0
O *A
en^.

C -0
^ O*>
I—*





•-n Q
7- ON
O CJ*





CM m
m CJN
*^

•o o

m p^
CO L^
m cj-
t-H
NO O
co CM
r- NO
CM in
co ro r*^ o^
r^ CM





-.





O O O O
m o
0 0
o* &\ co r*"










CO ^
00 CM
^ rs
r— ซ— < o o* i m co
f*- GO r^ co m oc
CD r- K", OD m rvj
j\ co|r\j *-4 f^4





ro rvj ^o ^ in (\j
•— < co *f co - o m oc co
O 0"*
^O m



sO >J-
p- co -en
PM -" ~i
l




CM en p- <•
rn co o ^ ' *M r*~
O C'r?*- >-O CO ^
fNJ L*> | O •—• ' O> fO
r- m

en CM



i
t*~ <\*j CT* i^ O*1 tn
— * f\j ^ -T r^- m
m r-inj o *j- CT*
*o co.rsi (M ro *r
r~ ^ rn (^ *^


O r-


0 N!-

^ CO
NO in'Csj P^ en •— *
en in o* in ซ-^ CM
o in CNJ o o •ป•
P- Nj-.en (M!*-^








cu X|O —"'NT in
~* O ON en ** x
f— r~ -< -ซ
J3 NJ-

-
— < p-
fO fX


r\j m
(\i *r
ปH


O "•
'-* in CM NC IT NJ-
C7- in;en f- rn -•
j^1 in PM -^iPM in


CM 1C. CD NO

O1 O-
en xjcr- NJ-T— o
en NC|IO NO
O- 0
NJ in -< o — i in
O Nj-


oo o
O NO
L*^. CM
J* NO
-0 NT

CO O
in NO
en en
ro rซ*


CT- ^
O f**
ซปซ


m C*N
NT 'XI

O> ON
O- >O NO NO
*-( r-;p~ p*
-* o



o- r-
0 0



p- p-
-• o o o
oc in in in




CM CO




•* NT
r-- irvcM CM
CJN r- NO -o
-i O O O






-• -riNj- ,j-
NT PM.fN- P-
CL p*- in in
CM O



CM O1
NT CC
O O



r- -.
PM O
-r o o P-
PNJ -Jl^l o








in *o ^ Nf
en o j P- in
CM ->



in o-
en CO
O O



P- 0
CM 00
if. CJO NO
CM -<

m o
en *}•
en o
CM rl



P- P-
sC en
f —
CM —•


CM h"
-i 0

NT NT
CM P-
P* f\
O O



o o
PN- X
P- NO
o o


NO O
CO PP1 NO .ป•
CM oil"- -n
CM — 
-------
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
 I
I
 I
I
I
I









,_,
ro
=tfc
C
O
•rt
rt
O
^1
Z
g"
cu
s
H
CO
^
IH
60
t-H
CX,
E
rt
oo
MH
O

w
H
.
i
D
rt
H


H
u.
c\
r\

S J
IX
Z

^
f -0 ^J•
j CM o in
j CM m in
4 CM en


tM
es.

in CM <;
ft fO
CM a
-4 sC

oo r-
•O 'T
r- r-
en T fM
-o -4 co CT
3- -Q eM r*-
^ *
1*- o-
CO *M
^ o
*T O' <3 >3-
u> O, oo >fr
*}• >r r- co co r*-
in in vo in. cu vO
CT1 CT
ft —

0 r-
Ol -0 J-
00 (TV CM —*
; i


i CT^  *A ^""
1 tO ^ ^J f*~-
r~ o sC|inino^cMi-oin
11 in in c^
a CM r-] —
•^ CM f\

crป a
fM -4]
c-
CM
CO
>r
-3" m| 
3 CM O —<
i CM en
e-
-\
r— cnl co
o in ซป•
•f co( NT
ซ in in' in
3 -* O
•] CM CM
-n
SI
^-" CM 3-
en — i
~-
in o^ '~v
. J ,
i-sH cc
J en CM
•j CM en

o r
100
U-
•j CM CN
CM m,

•ซC e*-
j- a
O CO
< ~i en
-x i
CM en
in in
ฃe5

JCM C\
CM en


Z
a
1—
2

a

er
u
z z
ซ 4
"' L
-a
^
^
en

in
vj-
^


T
u
3>
in
in
a-
CJ
CO NO
LLJ CT
_< in
t—
Z
LU
O
LU
a.
z z
2 i.
>- 1
cc
a:
V-
-
^ r— ซr -c >o -i-
co -o: —• t~- en ~r
en cc

en ซjQ
co e~
-O ~T


csj r- in o
r*- col -r *o
in o> m m
in cc co -a-, o 01
rj- in
ft if
r- ml ro r\i
i
en f— -o en
*f f^

r-- c
s$- U"v CO f\*
\f\ •ฃ ft *Q. ft cf\
• •
• • • *
CO JD CO en; O -O
CNI enj -o coi — • — ป
t^ in
ft ^
r- *L.
"
m r\
a- •*• — -o
•f O -O O *O cu
a- -c
cc m
r- o o co
in o. —* f
.n CM, r* in
-J

•c —
O O
fJ- CO >}•
r-- cc, ft -C' CM r-
r>- >C

ft M o r-
X: Ol CM CV
CT in r*- mi eฐ e\
en ^H
CO CO
xC 0s
n?
C7- ~0
•-4 ft
f CT-
ซO 3-
T- 1—
m en
3- -c
sQ *O
-?
r~- CT*
r- Jof
CM -Mi X 0-
r- o^ * r-
f\j a*| r^ x
f*- iT\

CM f—
r- co
m r\

*Q tf>
in f\j
in in in IT,
co m
CO rr
r~- vn
r\i CM
iT* •*
r^ f\
co -^
cj- ^ en *n
r r~

en CO
1 — t fM C^ CO
csj -4- vn c- cc r^i
^* ^ซ
r\* c*
,*-*


fsj ^J"i 'O if*
cc >oi *r rg




^ tn rsj (NJ (*• f^
m r*-
O ^" J* ^
ir* cr1 in in co J-
in ^ x rs — >r
o IT! o H m 
in ojl CT* ^Oj *^ r\
•—< >H

r* m
•C ro
NO r^
• t


0 C-


•!• *
— i -r. in cq
en -H| o -T
• 1 • •
\o ooi r*- inj po %c
^o mi 0s M in r\
-H
r- o cr trs
>r o ^ ^

in c^
IT. O O — < CM —
r\j N-
\O *"%
w-4 —
fO f*
r\j i\j
if\ m
CO f\
%O rn

J- o
^O rnl ff1 f^l in rs
,-4 — 1

—i CO
"-C LTV
O O
O iT>
r- f\
^-ป — ^
^. o
r^ _
^^
o tv


^ f\
3- in
r- a-
b r-
^
JO 0-


<- co
u y-
r-- *
>r co
in e\
•* f
r~ en^ o m
^j rr*] t>- *.
t^ rcl-f t~
r* rnj o r^
*•"•*-*(•"•*






o a
o o
Nฃ in










0 C
o o
>r m




in cs






o a
o a
CM —




-*• Hf- iป-
-4" en e**i en
>n m in ir\
o in in m
Hi

en re-


en en

— ^T <• -3-
*-< CO CO CO
-





ej- in1 in in
m rg1 r- r~
-r CM -f -r
en en -o -O

j
o eni in in
-. ^f'o 0
r** ^ co a>
—t CO
*_ซ


CO O-
r- r—


•J- -t
-^ 3-, en -*
•* -H| O -4-
• ซj • •
•4" -C1 -T en


—i ca in o
CM o[ o ej'
C* en co •-*]
CM a> r^ o
_*

ft 0-


a- -<
-i ft a- ^
— u-\-i >r
prt ^i ^ fO
i— l



OJ iA O^ f\J
Q* |H4l 0> O*
—* ^ -^ ^
^r-o^



^ j*! ^- ^i
aj ซoj a^ *^
(Nj Oi -* -4-
1
P^ 3^ CO ft
—t 7- CM -4
e— ป 3< -r
en so in *n
"H











O C
n — t






in s
o _i





D-10

-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I











IO
en
=it
ri
o
tj
4-1
O
^

3"
a)
s
8
H
00
O
<4-i
!
rt
"o
tJ
UJ
H
^
0
Q
_Q
rt
H




l/>
t—
Z
rg
ro
sr
rg

2.
UJI NO
o:
z
t
a.
i.
PI

u_
O
<
O
z
_J

or
UJ
Z
o
z
_J
O-
i-
l/l

















pg
—4





=0








sr






rg





oo p- in
p- ml so
00 — f\
' pg NO' P-
in cc P-
—4—4 O>
0> .J-
^
gj rg NO
in ml 4-4
p- sr co
sr p-1 cr-
— 4 —4 3J


p- o-i n
•o co' m
n sr i o
— i '.n: rg
CP NC CP
i— i r\. in

rg
p- CP in
oo r~i sr
sr in pi
m p— sr
p- pi co
— " rgf sr
p~i m
p~i
o o' in
in ov o
rg in r-j
CO p- CP
—4 rg Ji
rg
sr P- in
pi p-i sr
p- Ntj n
m sc
O sT
ro — •
p- CT-
p- -^
0> 0-
p- p-
sr sr
NO SC
au co
p- r-
CD CO

rg NO
=o m
m in
rg co
PI *O
ON QLJ

in rg
cr NO
sr in
in in
NO C7*
•—4 -H
in rg
NT P-

sr m
00 p-
sr rg
r-4 —t
rg o1^
-o in
in r-
sr rg sr| tn sr
p- rg| co
—4 rg

in NC
NO NT
sr
— '
^
sr
in — NO
NO fN, ST
CO O O
—4 pi in

O- r-
O co
— ' NO
o r-
D p-
-< rg

co rg
PI NO
r*~ NC
(0 rg
ซ-4 Pt

ซo rg
rg o
rg 4-4

x cr-
-4 t\!

^
t—
NJ^
NO
O
a-
^
n
p-
sr
.M
NO
O
NO
NC
in
pg
ซo
o
r-.
L/I %o
UJ NO
f-
o
UJ
< a
>ปi
* Z X
UJ[ C O

o
e:
1 — -L
,.3
CC
2 Z|(-
u k-
i-O
CJN NO
•~* o
-4 -4
cc rg
rg p-
NO sr
r"l Lp
-H P-
in rg

-ซ 00

00 O
o x
C" -4
sf PI
—4 m
NO O
CO NO
-O 03
sr ^
m m

m cc
=0 ^
sr in
rg sr
CO P-
NO sr

rg cc
n p-
r~ ON
-^ in
O^ NO
NO fO
rj m
rg rg
pi in
rg o
NT -O
p- sr

a- sr
rg in
—4 n
0 0-
p- rg
S5 NJ-

o in
C -J-
NO in
p- rg
fM NC
p- sr

rg mi SO PI
rg in
*-* ^
in o
CO iM
t?\ —4
sr pi
01 ^
rg cc
NO PI
NO Nj
"" "*!
a- oo
in c*
3 ->

o> -<
—4 ^
o1* in
n co
E Nj"
NO O
NT 0s
J1 NC
•O -f
NO Nj-

0 P-
01 —4
o co
o mi co rv
ro p*>
sr rg
i—4 —4





in o
O CT1
Cf* O




LT> NT
•O -3-





0 0
.1 0
C* o




NO O
NT O
CO ON
sr —4
rg p-
rg -*
sr-
co in
in —
,-4 —
rg p-
rg —4

Nt NC
n 0s
r- o
in o-
o c
pi rg

rg a)
co in
CO -C
rg sr
n ^o
rg — .
PI in
P- 0
CO -4
—4 rg
in cc
fNj —4

NJ* PI
in NO
—4 NT
0 PN,
NO CT*
PN> •— *

p- CO
NO L'
O NC
x sr
LA O")
rg — 4

NT in
NT CC
NO -4
n NO
in co
rg —4

sr ct
in, sr
PI NO
.n su
rg —>

-4 NO
in —4
in NO
r- so
in cc
rg — <





O O
O 0
co r-




co O co rg
00 —4
sr 0s
p- > p- O>
sr o;in rg
*~* *~*

rj PI in rg
NO in1— * co
pi in !o NC
rg in!— NO
rn O p- pi
r-4 -4

i
0 O'~4 ^
kZ> CO O O^
ji -4 m ji
in NO'CO PI
in o . NO in
r-^ "H
i
!
p- o** fN- m
*g — 4 pi aj
in c^ ' ui ui
ft Nj" O^ P~
;sf 0
sr r\j
p- in

rn O co NO
sr in
l*-4

o- sr
-* in
-4 in
-4 sr
t— i


sr rg
O 3"
rg a:
CO J3
—4


O NO
O -0
co rg
sr in
pi O NO P1|— 4
""*

rsi i^v ro in
rO f\i •sj' rn
r* O^ in fM
^-ซ ^O Psi ปO
(<\ CC NQ (*>
— ซ



rvj ro| in m
iT >n o o
-- u-4 o >r
r— ir> r^- o
fO O' NC -4"
--< -^


C1 -C co -^
fO fX LT f^
,n x n x
rc\ •— ซ LP. ^-<
f**l i-T* i sO -si-
-<


r~4 f\J ปO ^
>O ^™" ^^ r*^
— * O %}• o>
rg ac| o f\i
•st" C7^
^-*

SO Nf


r\j %O a) r-
C~ •— * u3 *n
— * a* t\i r\.
rn ^P; .ฃ -^~
^<



O co >o rป
-^ j- ( — i in
O "NTJ 0s fNi
NC m
ro O*
— '





0 0
IT m
>c -r






o o
o 0,0 o










o o
*!D ^
n p-
•0 -0
—4


O CO
NO NO
CO *NJ
00 NC
—4


sr rg
O C"
rg x
CO NO
—4


P- O
NO — 4
-H m
ON NO
—4


-j in
O — 4
Cf C1
rg


r* O
C^ N?"
sr cr
O NO
rg






00
0 O
rg ^H




pg — 4


pg sr
-^ rg
rg —
pi pg



0 NO
P- NO
-C C7>
N}- —



sr o
sr rg
NO in
r-i rg


CO NO
sr P-
co CP
m -4



o rg
-H NO
CO J1
pi pg



p- pg
rg sr
a. o-
m -4



pi in
LI sr
NO rg
PI rg



-" NO
cr ji
J* rg



_ co
O rg
O rg
sr rg







0 0
ji ^





i
vn in

'S 2
i—4 i—4


!Nป- sr
rj rg
—4 ^
PJ P.I



rg pg
sr sr
CP OV
—4 1—4



sr sr
rg rg
-J -4
rg rg
j

in p-
p- r-
sr cf
-4 O



sr NO
rg p-
•^ O^
rg —4



p- p-:
rg p-
m o"
-0



rg p-
O rg
C1 -n
-4 -4



o P-
<*"i p-[
O- ON|
-H 0

I
1
O r>-
0 1^,
Jป ^ (
—4 O


1











D-n

-------
TECHNICAL REPORT DATA
•(Please read Instructions on the reverse before completing)
1. REPORT NO. • 2.
4. TITLE ANDSUBTITLE
Evaluation of the Multiple-Source Gaussian Plume
Dispersion Model - Phase I
7. AUTHOR(S)
Robert C. Koch
George E. Fisher
9. PERFORMING ORGANIZATION NAME AND ADDRESS
GEOMET, Incorporated
15 Firstfield Road
Gaithersburg, Maryland 20760
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Protection Agency
Research Triangle Park, North Carolina 27711
3. RECIPIENT'S ACCESSION NO.
5. REPORT DATE
April 1973 (issue date)
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
GEOMET Report Number EF-186
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-0281
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
Phase II Report published July 1975 as EPA Report No. EPA-650/4-75-018-b
 16. ABSTRACT

       Ten different ways of applying the Gaussian plume diffusion model to represent
  air quality in an urban area are compared.  The different techniques include
  different degrees of detail in representing spatial and temporal variations in
  emissions and in meteorological conditions.  The methods used to represent spatial
  and temporal variations are described.  It is-shown that some improvement results
  from the use of more detailed spatial and temporal variations.  It is suggested
  that greater improvements would result if more detailed measurements of emissions  .
  and meteorological conditions were available.
       The report places primary emphasis on the use of the Sampled Chronologica-1
  Input Model  (SCIM) as a computer program for the multiple-source Gaussian plume
  diffusion model; the characteristics of this program are described.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
Air Pollution Diurnal
Urban Areas Variation
Atmospheric Diffusion
Proving
Mathematical Models
Emission
Sequential Sampling
13. DISTRIBUTION STATEMENT
Release Unlimited
b.lDENTIFIERS/OPEN ENDED TERMS
Air Quality Model
19. SECURITY CLASS (Tins Report)
UNCLASSIFIED
20. SECURITY CLASS (This page)
UNCLASSIFIED
c. COSATI Field/Group
1302
0401
21. NO. OF PAGES
22. PRICE
EPA Form 2220-1 (9-73)

-------