United States
              Environmental Protection
              Agency
             Environmental Monitoring and Support EPA-600/4-79-070
             Laboratory           October 1979
             Research Triangle Park NC 27711
              Research and Development
*>EPA
Statistical
Analysis of the
Los Angeles
Catalyst
Study  Data

-------
                RESEARCH REPORTING SERIES

Research reports of the Office of Research and Development, U S Environmental
Protection Agency,  have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology  Elimination of traditional  grouping  was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are

      1.   Environmental Health Effects Research
      2   Environmental Protection Technology
      3   Ecological Research
      4   Environmental Monitoring
      5   Socioeconomic Environmental Studies
      6   Scientific and Technical Assessment Reports (STAR)
      7   Interagency Energy-Environment Research and Development
      8   "Special" Reports
      9.   Miscellaneous Reports

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 quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia  22161.

-------
        STATISTICAL ANALYSIS OF THE
      LOS ANGELES CATALYST STUDY DATA
                    by
             Johannes Ledolter
              George C.  Tiao
             Spencer B.  Graves
               Jian-tu Hsieh
             Gregory B.  Hudak

         Department of Statistics
         University of Wisconsin
         Madison,  Wisconsin 53706
         Contract No.  68-02-2261
             Harold B.  Sauls
             Project Officer

         Analysis and Reports Branch
 Environmental  Monitoring Systems Laboratory
       Environmental Protection Agency
Research Triangle Park, North Carolina 27711
         This study was conducted
            in cooperation with
        The University Of Wisconsin
         Madison, Wisconsin 53706
 Environmental Monitoring Systems Laboratory
     Office of Research and Development
   U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711

-------
                                DISCLAIMER

     This report has been reviewed by the Environmental Monitoring Systems
Laboratory, U.S. Environmental Protection Agency, and approved for
publication.  Approval does not signify that the contents necessarily
reflect the views and policies of  the U.S. Environmental Protection
Agency, nor does mention of trade  names or commercial products constitute
endorsement or  recommendation for  use.

-------
                                  FOREWORD
     Measurement and monitoring research efforts are designed to anticipate
potential environmental problems, to support regulatory actions by developing
an in-depth understanding of the nature and processes that impact health and
the ecology, to provide innovative means of monitoring compliance with regu-
lations and to evaluate the effectiveness of health and environmental pro-
tection efforts through the monitoring of long-term trends.  The Environmental
Monitoring Systems Laboratory, Research Triangle Park, North Carolina, has
the responsibility for:  assessment of environmental monitoring technology and
systems; implementation of agency-wide quality assurance programs for air
pollution measurement systems; and supplying technical support to other
groups  in the Agency including the Office of Air, Noise and Radiation, the
Office of Toxic Substances and the Office of Enforcement.

     To assist in determining and evaluating effects of the catalytic
converter upon roadway pollutants, the University of Wisconsin Statistics
Department was contracted to conduct time-series analyses, develop appro-
priate models, and apply other statistical techniques to the data collected
in the Los Angeles Catalyst Study.  This report documents their analyses
and findings.

                                            Thomas R. Mauser, Ph.D.
                                                   Di rector
                                  Environmental Monitoring Systems Laboratory
                                    Research Triangle Park, North Carolina
                                      i i i

-------
                                  ABSTRACT

     This research program was  initiated for the purpose of performing
statistical analyses on the data from the Los Angeles Catalyst Study.  The
objective of the program was to determine the effects of the  introduction of
the catalytic converter on the atmospheric concentration levels of a number of
air pol1utants .

     This report gives analyses of  the  CO, Pb,  SOj, 0 , NO, and NO- data
covering the period from June  197^  to November  19/7-  Models were designed
to evaluate the freeway contribution to CO and  Pb as a function of traffic,
windspeed and wind direction.   These models were used to assess both the time
trend  in the pollutant measurements and the pollution concentrations at points
near the freeway.   In addition, frequency distributions were determined for
ambient air quality data.

     This  report was  submitted  in  fulfillment of Contract  No. 68-02-2261 by
the University of Wisconsin under  the sponsorship of the U. S. Environmental
Protection Agency.  This  report covers  the period September 1977 to August 1978,
and work was completed as  of September  1978.
                                      IV

-------
CONTENTS
Disclaimer 	
Foreword 	
Abstract 	
Figures 	
Tables 	
1 . Introduction 	
2. Conclusions 	
3- Analysis of the CO Data 	
k. Analysis of the Pb Data 	
5. Analysis of SOg Data 	 	
6. Analysis of High Pollutant Concentrations 	
7- Frequency Distributions for Hourly and 4-Hourly Pollutant
Concentrations 	
8. Analysis of 0-, NO, and N02 Data 	 ,
Appendix 	
References 	 	 . ,
. . i i
. . i i i
. . iv
. . vi
. .VIM
. . 1
. . 3
. . 5
. . 16
. . 21
. . 2k

. • 29
. . 37
. . 40
. . k2

-------
                                   FIGURES
Number

1-1  Map of Los Angeles air monitoring  stations at San Diego Freeway .    44
3~1a Monthly means of 4-hr afternoon  (3  p.m. to 7 p.m.) CO
          concentrations, Site A	    45
3-lb Monthly 25th, 50th, 75th, and 90th  percentiles of 4-hr
          afternoon  (3 p.m. to 7  p.m.)  CO  concentrations, Site A  ...    45
3~2a Monthly means of 4-hr afternoon  (3  p.m. to 7 p-m.) CO
          concentrations, Site C	    46
3-2b Monthly 25th, 50th, 75th, and 90th  percentiles of 4-hr
          afternoon  (3 p.m. to 7  p.m.)  CO  concentrations, Site C  .  . .    46
3~3a Plot of 25th, 50th, and 75th percentiles of 4-hr afternoon
           (3 p.m. to 7 p.m.) CO concentrations, Site A	    47
3-3b Plot of 25th, 50th, and 75th percentiles of 4-hr afternoon
           (3 p.m. to 7 p.m.) CO concentrations, Site C	    48
3-4a Diurnal plots of CO at Site  C; winter-summer comparison (weekday)    49
3~4b Diurnal plots of CO at Site  C; winter-summer comparison (weekend)    49
3~5a Diurnal plots of CO at Site  A; winter-summer comparison (weekday)    50
3~5b Diurnal plots of CO at Site  A; winter-summer comparison (weekend)    50
3-6a Diurnal plot of traffic counts and  traffic speed, weekday 1976  .    51
3-6b Diurnal plot of traffic counts and  traffic speed, weekend 1976  .    51
3-6c Diurnal plot of traffic counts and  traffic speed, weekday 1977  •    51
3-6d Diurnal plot of traffic counts and  traffic speed, weekend 1977  •    51
3~7a Diurnal plot of traffic density  (count/speed) weekday 	    52
3~7b Diurnal plot of traffic density  (count/speed) weekend 	    52
3-8  Diurnal plots of WS_|_	    53
3-9  Diurnal plot of CO;  Sites A, B,  F, and C;  summer weekday  ....    54
3-10 Diurnal plot of CO;  Sites A, B,  F, and C;  summer weekend  ....    55
3-11 CO model 1977 summer weekday fit, Site C	    56
3-12 Plot of the ratio CO/ (ct+kTD) versus WSi;  Site F  .  .	    57
3-13 CO model 1977 summer weekday fit, Site F (median strip)  	    58
3-l4a Joint CO model  fit summer weekday, 1977,  Site A	    59
3-l4b Joint CO model  fit summer weekday, 1977,  Site B	    59
3-l4c Joint CO model  fit summer weekday, 1977,  Site C	    59
3-l4d Joint CO model  fit summer weekday, 1977,  Site F	    59
4-la Monthly means of 4-hr afternoon  Pb concentrations, Site A ....    60
4-lb Monthly means of 4-hr afternoon  Pb concentrations, Site B .   .  . .     60
4-lc Monthly means of 4-hr afternoon  Pb concentrations, Site C .   .  . .    61
4-ld Monthly means of 4~hr afternoon  Pb concentrations, Site D .   .  . .    61
4-2a Monthly means of afternoon (3 p.m. to 7 p.m.) windspeed 	    62
4-2b Monthly frequencies of afternoon  (3 p.m. to 7 p.m.)  winds
          from l45B-325°	     62
                                     v i

-------
Number                                                                    Page

4~3a Monthly means of 24-hr Pb concentrations, Site C	     63
4~3b Monthly means of 24-hr Pb concentrations, Site D	     63
4-4a Monthly means of 4-hr  (3 p.m. to 7 p.m.) Pb readings
          weekday-weekend comparison, Site C  	     64
4-4b Monthly means of 4-hr  (3 p.m. to 7 p.m.) Pb readings
          weekday-weekend comparison, Site D  	     64
4-5a Plot of Pb/TC vs TS	     65
4-5b Plot of Pb/TCxTS) vs WS|  	     65
5~la Monthly means of 4-hr hT-vol and membrane afternoon
          readings of S0=,; Site A	     66
5~lb Monthly means of 4-hr hi-vol afternoon readings of SO?; Site B   .     66
5~lc Monthly means of 4-hr hi-vol and membrane readings of SO?;
          Site C	7 ....     67
5~ld Monthly means of 4-hr hi-vol afternoon readings of SO?; Site D   .     67
5~2a Plot of quarterly 25th, 50th, and 75th percent!les, minimum
          and maximum of 4-hr hi-vol afternoon SO? concentrations
          at Site C	7	     68
5~2b Plot of quarterly 25th, 50th, and 75th percent!les, minimum
          and maximum of 4-hr membrane afternoon SO? concentrations
                                                                           68
                                                                           69
                                                                           69

                                                                           70

                                                                           70
                                                                           71
                                                                           71
                                                                           71
                                                                           72
                                                                           72
                                                                           72
                                                                           73
                                                                           73
                                                                           73
8-4a Diurnal plots of 0  (Monday-Thursday) at Sites A and C  	     74
8-4b Diurnal plots of NO (Monday-Thursday) at Sites A and C  	     75
8-4c Diurnal plots of NO  (Monday-Thursday) at Sites A and C 	     76
5~3a Monthly means of 4-hr hi-vol SO? (C-A) differences
5~3b Monthly means of 4-hr membrane SO? (C-A) differences
5-4a Plot of quarterly 25th, 50th, and 75th percent! les of
hi-vol afternoon SO? (C-A) differences 	
5-4b Plot of quarterly 25th, 50th, and 75th percent! les of
membrane afternoon SO? (C-A) differences . . . .
8-la Monthly means of 4-hr (3 p.m. to 7 p.m.) 0_ at Site A
8-lb Monthly means of 4-hr (3 p.m. to 7 p.m.) 0^ at Site C
8-lc Monthly means of 4-hr (3 p.m. to 7 p.m.) 0 (C-A) . .
8-2a Monthly means of 4-hr (3 p.m. to 7 p.m.) No at Site A
8-2b Monthly means of 4-hr (3 p.m. to 7 p.m.) NO at Site C
8-2c Monthly means of 4-hr (3 p.m. to 7 p.m.) NO (C-A) . .
8-3a Monthly means of 4-hr (3 p.m. to 7 p.m.) N02 at Site
8-3b Monthly means of 4-hr (3 p.m. to 7 p.m.) N02 at Site
8-^?c Monthlv means of 4-hr (3 D.m. tl 7 o.m.) N0» (C-a)


4-hr
4-hr






A 	
C 	

                                    VI I

-------
                                   TABLES
Number                                                                   Page

3-1  Observed and Predicted CO Averages  (Site  C)	     9
3-2  CO Trend Assessment at Site  C	    10
3~3  Observed and Predicted CO Averages  (Site  F)   	    13
3-4  Observed and Predicted CO Averages  (Sites A,  B,  F, and  C)   .  .  .  .    14
4-1  Observations for Summer 4-Hr Pb  Data	    18
4- la Observations for Summer 24-Hr  Pb  Data	    18
5-1  Observations for Summer 4-Hr S0j=  Hi Vol Data	    23
5~la Observations for Summer 4-Hr S0=. Membrane	    23
6-1  Description of Ancillary Data  Accompanying  the Upper  Five
          Percent of the CO Concentrations	    26
6-2  Description of Ancillary Data  Accompanying  the Upper  Five
          Percent of the Pb Concentrations	    27
6-3  Description of Ancillary Data  Accompanying  the Upper  Five
          Percent of the SOj (Hi-Vol)  Concentrations   	    28
7-la Estimates of  X, p, o? for Box-Cox Transformation:  Los Angeles
          Hourly CO Data,  Site C	    33
?-lb Estimates of u for Pareto Distribution with ZQ =  8.75:  Los
          Angeles Hourly CO Data, Site C	    33
7-2a Estimates of  A , p , a2 for Box-Cox Transformation:  Los Angeles
          4-Hr Afternoon S0j= Data (Hi-Vol), Site C	    34
7-2b Estimates of a for Pareto Distribution with z  as the 75th
          Percent!le:   Los Angeles  4-Hr Afternoon  SO?  Data
          (Hi-Vol) Site C	7	    34
7-3a Estimates of A, \i , a2 for Box-Cox Transformation:  Los Angeles
          4-Hr Afternoon S0j= Data (Membrane), Site C	    35
7~3b Estimates of a for Pareto Distribution with z  as the 75th
          Percent! le:   Los Angeles  4-Hr Afternoon  50-r  Data (Hi-Vol),
          Site C	7	    35
7-4a Estimates of A , \i , a2 for Box-Cox Transformation:  Los Angeles
          4-Hr Afternoon Pb Data,  S i te C	    36
7-4b Estimates of n for Pareto Distribution with zn as the 75th
          Overall Percent!le:   Los  Angeles 4-Hr Afternoon  Pb Data
          Site C	    36
8-1  Observations for Summer Across-the-Freeway Differences
          for Pollutants	    38
                                     VIII

-------
                                    SECTION  1

                                   INTRODUCTION
        This  report presents a statistical analysis of the  Los Angeles  Catalyst
   Study  (LACS) data collected between June  197^ and December 1977-

        The primary objective of the study was to determine the  impact  of
   catalyst equipped cars on the ambient air quality near freeways.  The cataly-
   tic converter was adopted by American automobile manufacturers  in order to
   meet federal and state emission standards.  It has been  used on new  cars
   since  the  1975 model year.  The converter was designed to reduce the emissions
   of carbon  monoxides  (CO) and hydrocarbons (HC) .  Since catalyst equipped
   cars are required to use unleaded gasoline, the catalytic converter  should
   have lead  to a reduction  in lead  (Pb) emissions.  However, as pointed out by
   Beltzer, Campion and Peterson (197*0 and  Bockian, Tsou,  Gibbons and  Reynolds
   (1977), the catalytic converter increased the emissions  of sulfuric  acid,
   which  is measured as sulfate ion  (SO,).

   DATA BASE

        The San Diego Freeway in Los Angeles, California  (Figure 1-1), was
   selected as the location for studying the environmental  impact of the cata-
   lytic  converter.  Four air monitoring sites, two on each side of the freeway,
   were established by the Environmental Protection Agency  (EPA).  The  sites
   were designated A, B, C, and D.
        Data on a number of pollutants  (including CO, Pb, S0j=, 0,, 0,, N0? , and
   NO) and on meteorological variables  (such as wind direction, windspeed,
   temperature, and relative humidity) were collected from June 197^ to December
   1977-  Traffic counts and traffic  speeds were measured on an hourly basis
   beginning September  1976.   In January  1977, a fifth measurement site (Site
   F) was added for CO  concentrations at  the freeway median strip.

  ,      The percentage  of catalyst equipped cars passing the measurement sites
 y_was a critical parameter and has been  studied by Parry, Meyer, and Rodes
""'; (1977) and Rodes and Evans  (1977)-   It was found that starting in September
   197^, the monthly  increase  rate for  registration of catalyst equipped cars
   was roughly .5 percent.  After adjusting for actual miles driven, it was
   estimated that the percentage of converter equipped cars passing the measure-
   ment sites is approximately 30 percent in January 1977 and more than ^tO
'  percent by the end of 1977-

-------
PROCEDURES

     This report presents the findings of statistical analyses on  (l) the
daily 4-hr SOr- readings from two types of sampling equipment;  (2)  the daily
4-hr afternoon and 24-hr Pb concentrations;  (3) the  hourly CO  measurements,
and (4) the hourly NO, N0?, and 0, measurements.  S0j= readings were collected
daily between 3 p.m. and / p.m. using a  hi-vol  (high volume) and a membrane
sampler.  An hi-vol sampler was also used to collect samples for Pb concen-
trations.

     An empirical mechanistic model was  designed  to  relate the CO  concentra-
tions at Sites A, B, C, and F to windspeed, wind  direction,  traffic count,
traffic speed, and the distance of the measurement site  from the center of
the freeway.  This model was used to extrapolate  the CO  concentrations to
other sites near the highway.  We also designed an empirical mechanistic
model to relate the 4-hr afternoon Pb  readings  at Site C to  windspeed, wind
direction,  traffic count, and traffic  speed.   This model  was used  to assess
the trend  in Pb emissions.

     Frequency distributions for the ambient air  quality data  were interpreted
by an alternative approach.  The distribution of  pollutant measurements are
usually  long-tailed and skewed  to the  right.   It  has been argued  in  the liter-
ature  (Larsen  [1969])  that the  lognormal distribution provides a good approxi-
mation.   In our analyses we used a more  flexible  class of distributions, which
are characterized by a power transformation of  the observation.  Specifically,
it transforms the pollutant concentrations to near normality and  includes
the lognormal as a special case.  We considered only high values of  the pol-
lutants and used the Pareto distribution to  represent the upper tail area of
the d i str i but ion.

-------
                                  SECTION 2

                                 CONCLUSIONS
PRINCIPAL FINDINGS

CO Findings

     CO emissions decreased from 1975 to 1976, but a slight increase occurred
in 1977 (traffic also  increased in 1977).

     The joint model relating the CO concentrations at stations next to the
freeway was used to predict CO concentrations for given values of the input
variables wind direction, windspeed, traffic count, and traffic speed.
The proposed model was capable of predicting the CO concentrations for
downwind as well as upwind sites.

Pb Findings

     Very little background Pb was present in the studied area; most of the
recorded Pb originated from automobile emissions on the freeway.  Pb concen-
trations decreased until the end of 1976.  In 1977, however, Pb concentrations
were substantially higher.  This observed increase in Pb was caused by the
additional northbound  traffic lane, which reduced afternoon traffic conges-
tions and increased the traffic speed.

     Afternoon Pb concentrations are higher on weekends than on weekdays.
This weekday-weekend difference was attributed to afternoon weekend traffic
speed being higher, resulting in increased Pb emissions.

SOjF Findings


     From June 197^ to May 1977 hi-vol and membrane SO? readings decreased
gradually at all four measurement stations.

     The available hi-vol S0j= data for June to November 1977 showed a
considerable increase.   On the other hand, membrane readings increased only
slightly during the same period.  A comparison of membrane and hi-vol SOg
concentrations showed that while seasonal patterns are similar, the hi-vol
data were consistently higher.

     From 1975 to 1977 the across-the-freeway differences  (C-A) for the
^f-hr membrane SOf readings increased slightly.  Hi-vol SQf  (C-A) differences
behaved differently; from 1975 to the spring of 1977, a significant reduction

-------
was observed.   In the summer and fall of 1977 the hi-vol SOg concentrations
were slightly higher than the averages of the previous year.


0 , NO, and  NCL Findings


     From  1976  to 1977, 0, concentrations decreased  slightly at  Sites A and C
From 1975  to 1977 significant  increases  in  the  NO and  N0_  concentrations at
Site C and  in the across-the-freeway  difference (C-A)  were noticed.

Analysis of  High Pollutant  Readings

     High  CO concentrations were usually  recorded during traffic congestions
(low traffic speed), but  high  Pb and  S0j=  readings usually  occurred at higher
than average traffic speeds.

     High  Pb readings are associated  with high  S0j=  readings and vice versa.

Frequency  Distributions for Pollutant Concentrations near  the  Freeway

     The hourly CO  and the  4-hr afternoon S0j= and Pb concentrations near the
freeway  (Site C) did not  follow a  lognormal distribution.  A square root
transformation  of the CO concentrations and a cube  root transformation of
the 4-hr afternoon  SOg readings brought the distributions  of the measure-
ments  to near normality.  For  the  4-hr afternoon Pb concentrations, no
transformation  appeared necessary, and the  readings themselves appeared
to be  distributed normally.

     The Pareto distribution gave  a good representation for the upper
tail area  of the distribution of the  pollutants studied.

-------
                                  SECTION 3

                           ANALYSIS OF THE CO DATA


     The CO data  consisted  of hourly observations at Sites A and C from
May  197A to November  1977, and at Sites B and F  (median strip) from February
1977 to November  1977.

MONTHLY MEANS AND PERCENT!LES

     As an  initial  step  in the analysis of the CO data, the afternoon readings
(3 p.m. - 7 p.m.)  were used to calculate monthly averages.  Extreme observa-
tions  in the data were eliminated according to the following procedure:  Let
z. be  an observation  for a particular month, z the average, and s the estimated
standard deviation of the z.'s.   If |z.-zl>3s, z. will be excluded and z
                           •           II
and  s  recalculated.   The process was repeated until all observations fell
within z±3s.

     The monthly  means for Sites A and C are given in Figures 3~la and 3~2a.
To see how the observations are distributed within each month, various
percentiles of the empirical frequency distribution can be studied; in
particular, the 25th, 50th, 75th, and 90th percentiles can be plotted as
shown  in Figures  3~lb and 3~2b for Sites A and C respectively.  The distance
between the 25th  percent!le and the 75th percent!le is called the midrange
and provides a robust measure for the variation  in the data.  The 50th
percent!le (or median) is a measure of the center of the distribution and
the 90th percentile describes the upper tail of  the distribution.

     From these figures we made the following observations:

  •  The level of  CO  at Site C was considerably  higher than the level at
     Site A.

  •  The behavior  of  CO was markedly seasonal; at both Site A and Site C the
     CO concentrations were highest during the winter months.

  •  During the summer months the CO level at Site A was very low since
     afternoon winds  blow predominantly across the freeway towards Site C.
     In the winter  the CO levels at Site A were  higher.  This may have
     been caused  by the changed meteorology; a larger percentage of
     winds blew across the freeway towards Site  A.

     To discern the yearly trend movements more  accurately, the strong seasonal
effects were blocked  out by plotting the 25th, 50th, and 75th percentiles for
each month separately.  The results are shown in Figures 3~3a and 3~3b.

-------
DIURNAL PLOTS OF CO

     In Figures 3-^a and 3-kb diurnal plots of CO for the summer months
(May-October) and the winter months  (November-April), weekdays  (Monday-
Thursday) and weekends  (Saturday, Sunday) are shown.  Note that in all
diurnal plots the value for time  (t) corresponds to the average of the hour
t to t+1, t = 0,1,...,23.  From these four diurnal curves we observed that
the behavior of CO changed from weekday  to weekend, and from summer to
winter.  The summer weekday curve had two peaks which corresponded roughly
to the morning and afternoon rush hour traffic.  The summer weekend curve
followed a different diurnal pattern, because of the changed weekend traffic
pattern.  The weekday pattern  in  the winter did not show a peak during the
morning rush hours.  This difference can be explained by the change
In windspeed and wind direction from summer to winter  (Tiao and Hillmer
[1977]).

     The diurnal plots  for Site A are given  in Figures 3~5a and 3~5b.
During most of the day  Site A  acted  as a background station.   In  the morning,
however, part of the rush-hour contribution was recorded at this  site because
the windspeed was  lower and frequently  in the direction of the ocean.
FACTORS  INFLUENCING  CO CONCENTRATIONS

     Traffic and meteorological conditions, particularly windspeed and wind
direction, were the  major  factors which  influenced CO concentrations.  When
studying  the effects of  the  catalytic converter, these variables must be
taken  into account to properly assess the trend movements of CO.  Hourly
observations on wind direction and windspeed were available for June 197^
and thereafter.  Reliable  traffic data were available from September 1976 to
December  1976, and September  1977 to November  1977-
TRAFFIC  DATA

     Hourly traffic counts and average traffic speeds were recorded for
every  lane on  the San Diego Freeway near the monitoring sites.  From this
information we calculated hourly total traffic counts and average speeds
(weighted by the number of cars).  Until January 1977 the freeway had four
lanes  in each  direction (northbound and southbound).  The lanes were separated
by a concrete  median strip.   In February 1977 a fifth northbound lane was
added.

     As  shown  in Figures 3~6a to 3~6d, hourly total traffic counts and
average  traffic speeds were plotted for weekdays and weekends  in 1976 and
1977.  Comparing Figures 3~6a and 3~6b, we noticed a difference between week-
day and  weekend traffic counts and traffic speeds.  Weekday traffic counts
were characterized by two peaks which occurred during the morning and after-
noon rush hours; traffic speed was also considerably reduced during rush
hours.   On weekends no morning peak was observed; traffic counts peaked
around noon and between 4 p.m. and 6 p.m.  Two factors explained this
difference in  peak times.  Traffic during the night hours was  heavier on

-------
weekends  than on weekdays, and on weekends  the  freeway was  less  congested.
The average  traffic  speed  rarely dipped  below 50 mph.

     A  slight  increase  in  traffic volume was noted  from  1976  to  1977-   A
comparison of  the  traffic  data for  the two  years showed  that  the additional
northbound  lane  helped  in  reducing  traffic  congestion.   The average weekday
traffic speed  for  the afternoon rush-hour period increased  significantly  from
1976 to 1977-  However,  the weekend  traffic speed remained  largely unaffected,

     Comparing the diurnal pattern of summer weekday CO  concentrations  with
that of weekday  traffic  counts, Tiao and Hillmer (1977)  observed that traffic
counts do not  reflect the  magnitude of CO during the afternoon rush hours.
They argued  that traffic congestion during  the  afternoon rush hour period
could be a possible  reason.  Thus,  instead of using traffic counts they
considered traffic density (TD), which is defined as the ratio of traffic
count to average speed.  The diurnal curves for the weekday and weekend
traffic density  in 1976  and 1977 are given  in Figures 3~7a and 3'7b.  Whereas
the weekend  traffic  density was largely unaffected by the additional north-
bound lane,  a  rather substantial decrease in traffic density was recognized
during  the weekday afternoon rush hours.

     When weekday  and weekend traffic densitites were compared with the
corresponding  summer CO  concentrations at Site  C, the diurnal patterns  looked
very similar.  CO  concentration appeared almost linearly related to the
traffic density.

     In the  winter,  the weekday CO averages did not exhibit a peak during the
morning rush hours.  The meteorological variables, wind direction and wind-
speed, must  be taken into  account in order  to explain this absence of a
morning peak because they  are important factors for diffusion and transport
of the freeway CO  contribution.
METEOROLOGICAL VARIABLES

     Data on windspeed and wind direction were collected on an hourly basis,
Wind direction (WD) was an important factor since  it controlled the trans-
port of CO.  The freeway  is situated 145°, therefore winds coming
from l45°-325e transported the freeway contribution towards Site C, while
Site A acted as a background station.  When winds were from 325°-l45° the
freeway contribution was  recorded at Site A.

     Windspeed (WS) is also an important variable since  it controls the
diffusion of CO as well as its transport.  The higher the windspeed, the
more CO will be diffused.

     Following Phadke, Tiao and Hillmer  (1977) and Tiao and Hillmer (1977),
the hourly wind vector was divided  into  two components, one of which was
perpendicular (WS ) and the other parallel (WS. .) to the freeway:

-------
               WS, = WS cos(WD-2359)

               WS  ( = WS sin(WD-235°)

     Positive directions of  these  two components of  the wind  vector are  shown
in Figure  1-1.

     The diurnal  patterns of WSj_ for the winter of  1975/76  and  the summer of
1976 are given  in  Figure 3-8.   Notice that  in  the  summer  the  perpendicular
wind component was higher than  in  the winter.  Also  notice  that  in the morn-
ing during the winter months  (including rush hours)  winds blew  away from Site
C towards  Site A.  This change  in  WSj_ could explain  the absence  of the CO
morning peak at  Site C during the  morning  rush hours in the winter.
EMPIRICAL MODELS FOR CO  CONCENTRATIONS  AT  SITES  A,  B,  C,  AND  F

     A new site, F, was  added  in  February  1977 at  the  median  of  the  freeway.
Since that time, hourly  data were collected  on CO  concentrations  at  the  four
locations, A, B, C, and  F-  The diurnal  curves of  CO for  the  summer  (May-
October) weekdays and weekends of 1977  are given in Figures 3~9  and  3-10.

     Empirical mechanistic models were  used  to relate  CO  concentrations  to
traffic and meteorological variables.   First, models for  CO at Sites C and
F were considered  individually.   Then a  joint model was proposed  which re-
lated CO concentrations  at all four  stations.
A MODEL  FOR  CO AT  SITE  C

     Tiao  and Hillmer  (1977) derived a model  relating  the diurnal  behavior of
CO at  Site C to  traffic density and the  perpendicular  wind component.   Their
model  is given by:

                               -b(WS. -co)2
               C0t = a  +  kTDte    -1-     + a^                     (3.1)


where  CO  , TD ,  and WS.   are the observed CO  concentration,  traffic  density,
and  perpendicular  wind—component at hour t respectively; a   is  the error
term;  a  is a parameter  measuring the background CO;  k  is a parameter propor-
tional to  emissions and

                          -b(WS t-co)2
                        e     •*-

is a diffusion factor  involving the perpendicular wind component  and two
parameters b and to.

     This  simple model, which  requires only four parameters  (a,  k, b,  co)  to
be estimated from  the data, was extensively verified on summer  and winter,

                                      8

-------
weekday and weekend CO averages from the LACS data base  (Tiao and Hillmer
[1977]).  The validity of this model may be illustrated by using the most
recent data on CO, traffic, and WS  .  Employing 1977 weekday summer CO
averages at Site C and the corresponding TD and WD  values, the parameters
the model were estimated as:
           WD  values,
                                                                             n
                     May  1977-October  1977 Weekdays   (M-R)

                     Parameter       Est imate    Standard Error
                        a
                        k
                        b
1.27
0.023
0.019
3-51
                                                  0.20
                                                  0.0014
                                                  0.0080
     The actual  CO  readings and the  initially predicted values for CO are
plotted  in  Figure 3-11 and listed  in Table 3-1.  Apart from an apparent
underpred ict ion  during the period  from the 8th to the 15th hours, the model
(3.1) produced a close agreement between actual and predicted values.

CO trend analysis at Site C

     The primary objective of this study was to assess the effect of the
catalytic converter on the CO concentrations.  An analysis on the CO data
was done in  terms of Model 3.1.  We  first investigated whether k, the para-
meter proportional  to the emissions, changed over time.
              TABLE 3-1.  OBSERVED AND PREDICTED CO AVERAGES

Hour
0
1
2
3
k
S
6
7
8
9
10
11
Observed
CO Average
2.36
1.67
1.38
1.18
1.24
1.97
3-43
5.82
6.80
6.63
6.56
6.26
Pred icted
CO Average
2.16
1.74
1.55
1.45
1.49
2.01
4.04
6.12
6.47
6.13
5.70
5.49
Hour
12
13
14
15
16
17
18
19
20
21
22
23
Observed
CO Average
5-79
5-47
5-91
6.76
7.00
7-36
6.36
5-03
3-85
3-51
3-27
3.06
Predicted
CO Average
5.23
5-05
5.42
6.65
7.29
8.01
7-33
5-51
3-89
3.61
3.46
2.85

*Averages for Site C; 1977 Summer Weekdays  (Model 3-0

     The entire time span was divided  into  the following seven periods:
(1) June 1974 through October 1974;  (2) November  1974 through April  1975;
(3) May 1975 through October 1975;  (4) November  1975 through April  1976;

-------
(5) May 1976 through October 1976;  (6) November 1976 through April 1977; and
(7) May 1977 through October 1977-  This division corresponded roughly to
the "summer" and "winter" periods.  To test for possible emission changes, a
different emission parameter k was allowed for each period.  However, the
remaining parameters a, b, and co were constrained to be the same  for all
seven periods.  Within each 6-month period the 24-hourly averages for weekday
CO and WS  were calculated for each of the three consecutive 2-month segments
This division was chosen to allow for possible changes  in  the meteorology
within each season.  Since continuous reliable traffic data for the entire
period under study were not available, the following procedure was adopted:
(l) for all the 2-month periods prior to January 1977,  the same average
weekday traffic density figures calculated from the available traffic data
for the period September 1976 through December 1976 were employed: and  (2)
from January/February  1977 onwards, traffic data from the  period  September
1977 through November  1977 were used.

     The CO, WS  and TD were then employed to calculate estimates of the
parameters a, b, to, k,, ..., k?.  A weighted least squares procedure was
used, which allowed for the possibility of variances differing from one
period to another.  The parameter estimates are given in Table 3~2:


                TABLE  3-2.  CO TREND ASSESSMENT AT SITE C


           Period        Parameter      Estimate       Std. Error



Summer
Winter
Summer
Wi nter
Summer
Wi nter
Summer



7<»
7V75
75
75/76
76
76/77
77
a
b
0)
k
k
k
k
k
k
k



i
1
0
L.
4
T
6
\J
1
1
0
2
0
0
0
0
0
0
0
.69
.018
.80
.0175
.0209
.0228
.0200
.0197
.0222
.0209
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
069
0021
13
00065
00070
00066
00082
00051
00091
00066

The estimates of k  to kr were comparable to those obtained by Tiao and
Hillmer  (1977)-  Table 3-2 shows that the estimate of k decreased between
winter  1974/75 and winter 1975/76, and also between summer 1975 and summer
1976.  This decrease can be associated with the increase in the number of
catalyst equipped cars.  The  increase in k from summer 1974 to summer 1975
was at first surprising, but may have been caused by the same TD data (de-
rived from the September-December 1976 period) having been used.  Specifi-
cally, a major change  in the speed limit (from 70 mph to 55 mph) occurred in
January  1975-  Since TD is the ratio of traffic counts to average speed, the
density  figures used for summer 1974 were inflated and could account for the
increase in k from summer 1974 to summer 1975-  Comparing the estimates
between  1976 and 1977, a slight increase was observed from summer 1976 to

                                      10

-------
summer  1977 and  from winter  1975/76  to winter  1976/77-,  The evidence,
however, was  not very  strong  (increase not significant at  .05  level).

A MODEL  FOR CO CONCENTRATIONS AT  SITE F

     Summer weekday data  for  1977 were used to design a CO model  for the
median  strip.  As a first approximation one might expect that  the CO concen-
trations were  linearly  related to the traffic densities;  CO = a + kTD.  Wind,
however, acts  to diffuse  CO emissions from automobiles.  A plot of the  ratio
COt/(a+KTD )  versus WSj_t  is given in Figure 3-12.  CO   TD , and WSj_t
(t=0, 1, ...,  23) are  hourly summer  weekday CO averages at Site F, TD,  and
the perpendicular component of the wind vector, respectively.  The least
squares  estimates of CO   = a + kTD  + a  were a and k.

     Figure 3"12 shows  that apart from three points (which corresponded to
the early morning period  2 a.m. to 5 a.m. and which are most likely influ-
enced by additional meteorological variables) the ratio decreased slowly
with increasing  WS.|.  Such a relationship can be approximated by the
diffusion factor
                     -Y|WS  |6
                    e

where 5  is chosen to be .5, and y is a parameter.

     These components  led to the  following model:

                                   -Y|WS.  I'5
                    C0t = (a+kTDt)e     ^   +at                   (3-2)


Summer  (May-October 1977) weekday CO averages at Site F and corresponding
weekday averages of WS| and TD were  used to estimate the parameters.   The
results are given below":

                    May 1977-October 1977 Weekdays (M-R)

                    Parameter     Estimate   Standard Error

                        a            2.17          0.30
                        k            0.047         0.0025
                        Y            0.16          0.021

The actual  CO readings and the predicted values are plotted in Figure 3~13
and listed in Table 3~3-  It can  be  concluded that Model 3-2 produces an
excel lent fit.
A JOINT MODEL FOR CO CONCENTRATIONS AT SITES A, B, C, AND F

     A joint model covering all four stations was designed  in addition to
the univariate models built for Sites C and F.
                                     11

-------
     As shown in Figure 1-1, Site A was 150 feet and Sites B and C were both
75 feet from the median strip F.  Any joint model covering different sites must
include the distance from the center of the freeway as an explanatory variable.
Since the wind vector controls the transport of the freeway contribution,
our model had to reflect that Sites A and B were on opposite sides of the
freeway from Site C.  Thus, d was chosen to be an indicator variable for the
distance such that d  = 2; d  =  1; d  = 0; d  = -1, where one unit corre-
sponded  to 75 feet.                h

     The proposed joint model took the form:
                                                      A

                                                         - ylliK   I ' ^
                                                            'it'
CO.  = ( ct+kTD	   —   	  e     -1-   +a.
  Jt   /      t               /TTflU  I                  ,              Jt
e J J


/1+Bldjl I
                                                                      (3.3)
where
  •  CO.  = the observed CO concentration for Site  j(A,B,F or C) at hour t

  «   TD  = the traffic density at hour t.

  «    d. = the distance of each  location from the  median strip as defined
        J   above.
  •  WS   = the perpendicular component of the wind vector at hour t.
sgn = sign function;  sgn(x)  =j_,  .,


a.  = error terms, and
                                      1  if x>0
                                          x<0
      a,k,n,3,w  and Y are  unknown parameters to be estimated from the data.
                                      12

-------
                TABLE 3-3.   OBSERVED  AND  PREDICTED  CO  AVERAGES-

Hour
0
1
2
3
4
5
6
7
8
9
10
11
Observed
CO Average
4.88
3-19
2.k]
1.87
2.20
4.68
10.36
14.50
13-73
11.58
9.80
8.46
Pred icted
CO Average
4.26
3-30
2.87
2.57
2.63
3-95
9.04
14.32
14.49
11.67
9.20
8.07
Hour
12
13
14
15
16
17
18
19
20
21
22
23
Observed
CO Average
7-56
7-25
8.06
10.38
11 .61
12.34
9-19
7.44
6.65
6.91
6.96
6.41
Predicted
CO Average
7.57
7-36
8.04
10.01
10.91
11 .82
10.91
8.71
6.72
6.93
7-00
6.04

            Averages  for  Site  F;  1977  summer weekdays  (Model 3.2)

     Model  3-3 was fitted  using  the  1977  summer weekday CO averages for the
 four sites  and the corresponding WS.  and  TD values.   The parameter estimates
 are given below:

                       Joint CO model  - Summer weekday  1977

                     Parameter      Estimate    Standard Error
                        a
                        k
                        n
                        3
                        03
                        Y
       1.58
       0.051
       0.088
       1.29
      -4.12
       0.17
 0.11
 0.0015
 0.014
 0.19
 0.32
 0.016
by:
     The correlation matrix between errors at Sites A, B, F, and C  is given
                         P=
1.00   .53   .28
      1.00   .58
            1.00
 •34
 • 32
 • 25
1 .00
Actual CO averages and predicted values are plotted  in Figures 3-l4a through
3-l4d and listed in Table 3-4.  Model 3-3  is quite parsimonious  in the para-
meters; apart from the covariances for the errors only six parameters have
to be estimated from the data.  It produced a satisfactory overall fit at
all four locations.
                                     13

-------
           TABLE 3-4.   OBSERVED AND  PREDICTED  CO AVERAGES'

Hour
0
1
2
3
it
5
6
7
8
9
10
1 1
12
13
14
15
16
17
18
19
20
21
22
23
Si te A
Averages
Obs. Pred.
2.01
1.73
1.50
1.33
1.34
2.00
2.3k
3-43
3-25
2.68
1.97
1.33
1 .01
0.86
0.76
0.73
0.59
0.49
0.64
0.97
1.52
1=91
2.31
2.43
1.73
1.66
1.63
1.57
1.57
1.72
2.30
2.92
2.66
1 .90
1.38
1.16
1 . 10
1.06
1 .04
1 .04
1.05
1 .10
1.19
1-33
1.49
1-74
1 .86
1 .92
Site B
Averages
Obs. Pred.
2.53
1 .94
1 .66
1.41
1.47
2.51
4.81
6.12
5.05
3-99
2.63
1.43
1.13
0.92
0.83
0.74
0.62
0.54
0.64
0.68
2.21
2.71
3-09
3.08
2.34
1.97
1 .82
1.70
1-71
2.22
4.20
6.27
5.82
3-89
2.31
1.52
1.31
1.18
1.14
1.16
1.21
1.39
1.73
2.11
2.35
2.87
3.08
2.95
Site F
Averages
Obs. Pred.
4.88
3-19
2.41
1.87
2.20
4.68
10.36
14.50
13-73
11 .58
9.80
8.46
7.56
7-25
8.06
10.38
11.61
12.34
9-19
7.44
6.65
6.91
6.96
6.41
3.87
2.83
2.35
2.03
2.11
3.54
9.04
14.73
14.92
11 .83
9-17
7.96
7-43
7.20
7.91
9-97
10.92
11.89
10.96
8.65
6.53
6.76
6.83
5-79
Site C
Averages
Obs. Pred,
2.36
1.67
1.39
1.18
1.24
1.97
3.43
5.82
6.80
6.63
6.56
6.26
5.79
5.47
5-91
6.76
7.00
7.36
6.36
5.03
3.85
3.51
3.27
3.06
2.25
1.98
1 .82
1.70
1.72
2.19
3-97
5.81
6.39
6.08
5.83
5.63
5.23
4.86
5-07
6.24
7.01
8.05
7-59
5.66
3-91
3.61
3-50
3.02

"Averages for Sites A,  B,  F,  and C;   1977 summer weekdays (Model  3.3)
                                    14

-------
Interpretation of the joint model

  *   If d=0  (median) Model 3.3 reduces to Model 3.2,  (the univariate model
      for CO  concentrations at Site F) .

  •   For any fixed d^O, the first factor in Model 3-3 corresponds to the
      univariate model for CO concentrations at Site C (Model 3-0-  Moreover,
      in the  joint model the background concentration a and the free-
      way contribution were diffused by the factor


                                                -Y | WS '5
                                               e

  •   For fixed WS ,  CO concentrations decreased with  increasing distance from
      the freeway.  This decrease was  less for the downwind sites.

  •   One advantage of the joint model (3.3) was it could be used to predict
      CO concentrations at any given distance from the freeway median.

 It is worth  noting that this model was different from other models considered
 in the literature, notably the HIWAY model (Zimmerman and Thompson [1975])-
The HIWAY model can  be used only to predict pollutant concentrations in the
downwind direction from the source.   In future research it would be worth-
while to compare the predictive accuracies of those models using the LACS
data  and data from other regions.
                                      15

-------
                                  SECTION 4

                           ANALYSIS OF THE Pb DATA
     Since catalyst equipped cars are required to use unleaded gasoline, the
catalytic converter should lead to a reduction in Pb emissions.  A statistical
analysis of ambient Pb concentrations was conducted  in order to test this
assumption.  The Pb data used consisted of 4-hr afternoon  (3 p.m. to 7 p.m.)
and 24-hr hi-vol Pb concentrations at Sites A, B, C, and D.

ANALYSIS OF MONTHLY MEANS

     Figures 4~la through 4~ld show the monthly averages of the 4-hr afternoon
readings, for Sites A, B, C, and D respectively.  Monthly averages of the
afternoon  (3 p.m.-7 p.m.) windspeed are shown  in Figure 4-2a.  Figure 4-2b
shows the monthly relative frequency of the afternoon wind from the direction
(l45°-325°), i.e., when the wind was blowing across the freeway towards Site
C.  The following observations were recorded:

  •  The levels of Pb at downwind Sites C and D were substantially higher
     than those at upwind Sites A and B.

  •  In the summer months when the afternoon winds were blowing predominantly
     across the freeway towards Site C  (Figure 4-2a and 4~2b), very little Pb
     was seen at Sites A and B.  This absence of Pb  indicated there was
     almost no background Pb in this area, and that most of the recorded Pb
     came from automobile emissions on the freeway.  In the winter months
     when  the proportion of.winds from l45°-325° was smaller, some of the
     freeway contribution was noted at Sites A and B.

  •  The 4-hr Pb concentrations at Sites C and D decreased gradually until
     the end of 1976-  The summer (May-October) means of the 4-hr afternoon
     readings are given in Table 4-1.  At Site C the decrease from summer
     1975 to summer 1976 was approximately 17 percent; at Site D the average
     annual decrease from 1975 to 1976 was 11 percent.

  •  Starting in February 1977, Pb concentrations increased sharply, especi-
     ally at Site C, which was closest to the freeway on the downwind side.
     Table 4-1 shows that from summer 1976 to summer 1977 Pb concentrations
     increased 70 percent.  The increase at Site D was approximately 40
     percent (less data were available for Site D since hi-vol sampling for
     Pb at this station was discontinued in June 1977)- The addition of a
     fifth lane (northbound) to the freeway explained this sudden increase.
     The opening of the extra lane increased afternoon traffic speed on lanes

                                     16

-------
     closest to Sites C and D  (Figure  1-1).  Laboratory studies  (Hirschler,
     et al.  [1957]) have shown that Pb emissions  increase with traffic  speed.

  •  The  increase  in Pb at Site D was  smaller than the  increase  at  Site  C.
     This  discrepancy may have occurred because most of the  larger  sized  Pb
     particles, which are emitted at higher speeds, settled  near  the freeway.

2^-hr Pb  averages

     The  monthly means of the 24-hr Pb data at Sites C and D are  plotted  in
Figures 4~3a and 4-3b.  The summer  (May-October) averages for these data are
listed  in  Table 4-la. Notice at both sites the 24-hr Pb concentrations
decreased  gradually until 1976.  From  1976 to 1977, because of the  change  in
traffic speed, the 24-hr Pb readings increased at both locations.   The
increase  observed  in the 24-hr Pb averages, however, was small compared  to
the  increase seen  in the 4-hr afternoon data, because the additional north-
bound lane affected the northbound  traffic speed mainly in the afternoon.

Comparison of weekday and weekend afternoon Pb

     To further explore the argument that higher speeds cause higher Pb
emissions, weekday and weekend Pb readings were compared.  Phadke,  Tiao and
Hillmer (1977) and Tiao and Hillmer (1977) reported that the afternoon Pb
readings  were about 60 percent higher on weekends than on weekdays, even
though on  weekends fewer cars passed the measurement stations.

     This  difference between weekday and weekend Pb readings was  attributed
to the afternoon traffic speed being higher on weekends than on weekdays,
resulting  in increased weekend Pb emissions.  A comparison of the traffic
data for  1976 and  1977 shows the average afternoon traffic speed  on weekdays
increased  from 37 mph in 1976 to 47 mph in 1977-  The speed of northbound
traffic closest to Site C increased even more; from 27 mph in 1976  to 47 mph
in 1977.   At the same time there was very little change in the weekend speed
between 1976 and 1977-  Therefore,  the speed difference between weekdays and
weekends  decreased in 1977 and thus the difference between weekday  and
weekend afternoon Pb should have been  less prominent.  This prominent reduc-
tion was  confirmed by Figure 4-4a (Site C) and Figure 4-4b (Site  D).

A simple model for the 4-hr afternoon Pb data at Site C

     Daily afternoon (3 p.m. to 7 p.m.) averages of traffic count,  traffic
speed, WS., and Pb from the periods August 27-December 27, 1976,  and September
l4-Novembe~r 30, 1977, were used to  build an empirical model for  the concentra-
tions.  Since reliable data on both traffic count and traffic speed were
available  for these two periods, there was no need to distinguish between
weekdays and weekends.

     The average of the 4-hr afternoon Pb readings for the 1976  period was
7-91  yg/m3.  For the 1977 data (after the additional northbound  lane was
opened)  the average rose to 10.87 yg/m3.  The proposed model helped to
determine whether this increase could be explained by changed meteorological
and/or traffic conditions.

                                     17

-------
             TABLE  4.1.   OBSERVATIONS  FOR  SUMMER  4-HR  Pb  DATA

4-Hr Pb Concentrations (3 p.m. to 7 p.m.)
Site Year(s)
1975
C 1976
1977
1974
n 1975
1976
1977
r 1976-1975
1977-1976
1975-1974
D 1976-1975
1977-1976
Mean
8.16
6.77
5.64
4.90
4.40
6.18
N/A
N/A
N/A
N/A
N/A
Standard
Devi at ion
2.78
2.25
2.14
1.29
1 .00
0.96
1.18
N/A
N/A
N/A
N/A
N/A
Number of
Observat ions
174
166
92
151
176
166
20
N/A
N/A
N/A
N/A
N/A
t-stat i st ic
N/A*
N/A
N/A
N/A
N/A
N/A
N/A
-5.05
16.14
-5.83
-4.71
7-63

 Not  Applicable




             TABLE 4.la.   OBSERVATIONS  FOR SUMMER 24-Hr  Pb  DATA

24-Hr Pb Concentrations
Site Year(s)
1974
1975
C 1976
1977
1974
n 1975
° 1976
1977
1975-1974
C 1976-1975
1977-1976
. 1975-1974
u 1976-1975
1977-1976
Mean
8.19
7-98
7.00
7.43
5.00
4.19
3.75
4.16
N/A
N/A
N/A
N/A
N/A
N/A
Standard
Devi at ion
1.83
1.48
1.60
1.19
1.02
1.01
0.73
0.70
N/A
N/A
N/A
N/A
N/A
N/A
Number of
Observations
143
173
166
74
46
57
57
63
N/A
N/A
N/A
N/A
N/A
N/A
t-stat ist ic
N/A*
N/A
N/A

N/A
N/A
N/A
N/A
-1.13
-5.86
2.07
-4.03
-2.67
3-14
'Not Applicable
                                     18

-------
     As noted earlier  (Hirshler, et al,  [1957]), an  increase  in traffic  speed
produces an  increase  in Pb emissions.  This relationship  is  illustrated  in
Figure 4~5a, where the ratio of afternoon Pb to the  total afternoon traffic
count  (TC)  is plotted against the average afternoon  traffic  speed.  Averages
of Pb/TC for successive traffic speed  intervals were plotted  instead of
individual observations.  This plot shows that the ratio  (Pb/TC)  increases
proportional to traffic speed, implying a relationship of the form:

                    Pb/TC = kTS or Pb = kTCxTS.                 (4.1)

     The wind affects Pb concentrations by means of  transport and diffusion.
These  influences are  shown in Figure 4-5b, where we  plotted Pb/(TCxTS)
against WS  which  is  to take Pb/(TCxTS) proportional  to the dispersion
factor
                                   -b(WS,-u>)
                                 e
                                            2
1                       (4.2)
where b and w are appropriate constants.

     Combining 4.1 and 4.2 the relationship can be written as

                                   -b(WS -u)2
                       Pb = kTCxTSe     ^                       (4.3)

     The parameters k = k. , for  1976 and k = k« for 1977 were introduced to
test whether a change occurred in the Pb emissions from 1976 to 1977-  After
having modeled the dependence of Pb on traffic and meteorological  variables,
it was possible to test whether  the parameter k, which was proportional to
the Pb emissions, changed over time.

     Since a larger contribution was expected from the northbound traffic
lane (because it was closest to  the receptor at Site C) , the effects of
the northbound and the southbound traffic were separated.   The next step
was to design the model:
where:     Pb  = the observed 4-hr afternoon Pb reading at Site C for day t,

       TC ,TC  = total afternoon northbound  (southbound) traffic count for
                 day t,

         N   S
       TS ,TS  = average afternoon northbound or southbound traffic speed
                 for day t,
         WS      average afternoon perpendicular component of the wind vector,

            a  = error term,

         k,,k9 = parameters proportional to  the Pb emissions; k = k, for the
                 1976 period and k = k2 for  the 1977 period,

                                      19

-------
             a = parameters measuring the relative contribution of the north-
                 bound and southbound lane,

           b,w = parameters in the dispersion factor  involving the perpendi-
                 cular wind component.

Using the set of data on Pb, traffic, and WS  described above, the parameters
in Model b.k were estimated by nonlinear lea's! squares.  The parameter
estimates and their standard errors are given below:

                    Parameter      Estimate  Standard Error

                         (1976)
                         (1977)
                       a
                       b
                       to
8.77
8.75
0.75
0.011
2.19
0.39
0.52
0.10
0.0035
0.56
Various diagnostic checks such as  residual plots failed  to  reveal  any apparent
inadequacy  in the model.  In particular, a plot of  residuals against traffic
speed showed no apparent relationship, thus  indicating that the model ade-
quately described the  relationship between Pb and traffic speed.

     Two  important observations were made from the  parameter estimates:

  •  The  estimate for  a  (.75) showed that the northbound lane, which is
     closest to the  receptor Site  C, had a much larger contribution to the
     Pb concentrations at this site than the southbound  traffic lane.  That
     the  heavier Pb  particles settle fairly quickly explained the  larger
     contribution of the northbound lane.

  •  The  estimated emission constants  kj and k£ for 1976 and 1977  did not
     differ significantly,  indicating  that the 37 percent increase in the
     means  from 1976 to  1977 was caused by the change  in traffic speed.
                                      20

-------
                                  SECTION 5

                              ANALYSIS OF SO^ DATA


     The catalytic converter was designed to reduce the levels of
hydrocarbon and carbon monoxide emissions.  Although, it successfully
reduces the emissions, the catalyst also converts the sulfur in gasoline
to sulfate ion  (SO-r).  This feature of the converter caused alarm since
an increase in ambient sulfate concentrations from catalyst equipped
automobiles could have harmful health effects.

     The SO? data analyzed consisted of daily 4-hr afternoon (3 p.m. to
7 p.m.) concentrations from both the hi-vol and membrane sampler.  The
difference in the 4-hr concentrations between Sites C and A reflected
the contribution from the freeway, since in the afternoon the wind blew
predominantly in a direction perpendicular to the freeway (Evans and
Rodes  [1977]).

     In analyzing the SO? data daily afternoon readings were used to calculate
monthly averages at the different sites.

MONTHLY SOjj AVERAGES FROM HI-VOL SAMPLERS

     The monthly SO? averages from the hi-vol sampler at Sites A, B, C, and
D are given by the solid lines in Figures 5~la through 5~ld.  It was noticed
that SO? decreased gradually at all four stations until May 1977

     Available data for June 1977 to November 1977 at Sites A and C showed a
considerable  increase in SO?.  No data were available to confirm this  increase
for Sites B and D.

     Monthly means at Sites C and D downwind from the freeway were slightly
higher than those for the background Sites A and B, indicating a contribution
to SO? from the freeway.  The contribution, however, was very small.


MONTHLY S0jj AVERAGES FROM MEMBRANE SAMPLERS

     The broken lines in Figures 5~la and 5~lc show the monthly membrane
sampler SO?  averages from June 1974 to November 1977-  A slight reduction
in membrane SO? was noticed at both Sites A and C until May 1977-  As on the
hi-vol  readings a slight increase occurred for the last part of the data
(June-November 1977)-

                                     21

-------
     A comparison of membrane and hi-vol SO? concentrations, showed that
although seasonal patterns were similar, the hi-vol data were consistently
higher than the membrane readings.


AN ADDITIONAL GRAPHICAL REPRESENTATION OF THE  SO?  DATA

     In Figures 5~la through 5~ld the monthly  averages of  the hi-vol and
membrane SO? are plotted.  These were useful for a preliminary  trend
assessment.  The minimum,  the 25th,  50th, and  75th percentiles, and the
maximum were also plotted  to describe the variations of  the  SO? samples.
Since every month (or quarter) did not have the same number  of  data points
box plots  (Tukey [1977], McGill, Tukey, and Larsen [1978]) were used to
incorporate this information into the plots.   The  box plots  are illustrated
in Figures 5~2a and 5~2b for both the hi-vol and membrane  SO? readings at
Site C.  The data are grouped  into 3-month periods corresponding  to the
seasons:  winter (December, January, February); spring  (March,  April  May);
summer  (June, July, August), and fall  (September,  October, November).  The
lower  line in each box corresponds to the 25th percentile  (lower  quartile);
the broken line gives the  median and the upper line gives  the 75th percentile
(upper quartile) of all observations for each  season.  The length of the
box measures the interquartile range and the width of the  box indicates the
group  size.  It  is chosen  proportional to the  square root of the  number of
observations in each group.  This choice was made  because most  measures of
variation  (e.g. the sample standard  deviation) are proportional to the
square  root of the group size.  Figures 5~2a and 5~2b show the  maximum and
the minimum SO?  readings within each season.


ACROSS-THE-FREEWAY DIFFERENCES

      It was pointed out earlier that the difference between  the 4-hr afternoon
readings at Sites C and A  can be represented as the contribution  from the
freeway.   Monthly means and quarterly box plots for the  (C-A)SO?  differences
between these sites are given  in Figures 5~3a  through 5~4b.  A  slight increase
was observed in  both monthly means and quarterly medians of  the membrane SO?
differences.  These observations were confirmed by the means of the summer
differences shown in Tables 5~1 and  5~'a.  The increase  in the  summer differ-
ences,  however, was not statistically significant.  Figure 5~4b indicates
that for the first year of the study little data on membrane SO? were
ava ilable.

     Hi-vol SO?  (C-A) differences behaved quite differently  from  the corre-
sponding membrane readings.  From 1975 to the  spring of  1977 a  significant
reduction  in the hi-vol SO? differences was observed.  At  the same time the
membrane differences  increased slightly.  The  hi-vol SO? differences for the
summer  and fall of 1977 were slightly higher than  the averages  for the
previous year, but were still significantly smaller than those  in 1975
(Table  5~1).  The discrepancy between the hi-vol and membrane readings raised
an  important question:  Which, if either, of the two measurement  methods was
giving  the correct pollutant reading?  One possible conjecture  was put
forward by Tiao and Hillmer  (1977)-  They argued (l) that  the 4-hr hi-vol
difference was caused by SO  reacting on the filter to form  SO?,  and  (2) that

                                     22

-------
this artifact formation also occurred on the membrane filter, but to a lesser
extent.  Although this theory explained the observed discrepancies fairly
well, further research is necessary to determine the precise nature of these
two fi1ters.
          TABLE 5-1•
OBSERVATIONS FOR SUMMER 4-HR SO^ HI  VOL DATA

Year(s)
1975
1976
1977
1976-1975
1977-1976
1977-1975
'V
Not Appl i cable
TABLE 5-la.
Mean
4.87
2.53
3-29
N/A
N/A
N/A

OBSERVATI
Standard
Deviat ion
3-98
3.86
5-49
N/A
N/A
N/A

ONS FOR SLIMMER
Number of t-statistic
Observat ions
164
156
87
N/A
N/A
N/A

4-HR SOjj MEMBRANE
N/A*
N/A
N/A
-5.34
1.26
-2.61



Year(s)
1975
1976
1977
1976-1975
1977-1976
1977-1975
Mean
.14
.67
.99
N/A
N/A
N/A
Standard
Deviat ion
4.53
2.04
2.15
N/A
N/A
N/A
Number of t-
Observat ions
46
150
134
N/A
N/A
N/A
stat istic
N/A
N/A
N/A
1 .12
1 .29
1.69
 Not Applicable
                                      23

-------
                                  SECTION 6

                  ANALYSIS OF HIGH POLLUTANT CONCENTRATIONS
     Ancillary variables such as traffic, meteorology, and the concentrations
of other pollutants which accompanied high concentrations of CO, Pb, and
SOj (hi-vol) at Site C were investigated.  The upper five percent of each of
the above three pollutants were identified.  They were augmented by the
corresponding concentrations of the remaining two pollutants, by meteoro-
logical variables  (wind direction, windspeed, relative humidity and temper-
ature), and by traffic count and traffic speed.  Earlier studies of this
kind using aerometric data from other regions can be found in Cleveland,
Kleiner, and Warner  (1976) and Tiao, Box, and Hamming (1974).

     Since the Pb and SOg data considered were daily 4-hr afternoon readings,
the accompanying meteorological and traffic variables were averaged over the
3 p.m. to 7 p.m. period.  For the hourly CO data the daily maxima were
cons idered.

     Several tables which present data for the upper five percent of the
highest daily CO readings are given.  Table 6-1 shows a dot diagram indicating
in which months the high values occurred.  Dot diagrams of the corresponding
Pb and SO? 4-hr afternoon readings  in terms of their percentiles and the
proportion of corresponding winds from the direction l45°-325° (blowing
across the freeway towards Site C) are given in Table 6-2.  Table 6-3 displays
the proportions of the corresponding measurements on windspeed, relative
humidity, temperature, traffic count and traffic speed which are above/below
their overall average for that particular hour of the day.
CO FINDINGS

     High concentrations of CO usually occurred during the fall and winter
period  (September-March)- When traffic speed was low and winds blew across
the freeway towards Site C with below average windspeeds, high concentrations
of CO were also recorded.
Pb FINDINGS

     High Pb readings were recorded in the summer of 1975 and  in  1977 when
fifth lane was added in the northbound direction.  The new lane reduced
                                     24

-------
traffic congestion  and  increased traffic speed.  High Pb  readings also
occur when traffic speed was above average and when winds  blew perpendicular
to the freeway at lower  than average speed.  Furthermore,  high Pb readings
were recorded when both  relative humidity and temperature  rose above average
High S0|= readings were associated with the upper five percent of the Pb
read ings.

SOjj FINDINGS

     High SQ=r readings occurred throughout the year, with  most of the high
SO^ concentrations recorded in 1975-  On days of high traffic speed and
across-the-freeway winds blowing towards Site C, high S0>  readings were
observed.  Furthermore,  high S0> values usually occurred on days with high
relative humidity.  A very large proportion of high S0j= readings was accom-
panied by high Pb concentrations.
                                      25

-------
time of occurrence
                      J	ftA
                            I  .      8.1
                            1/75
                         A«ft_
                                    :  •
                        1/76
                            1/77
                         11/77
correspond ing

4-hour afternoon     J*-<—^—L*-±SU—if i ti »t+i»i«i ti ti ^i*!*!—i_ftL

Pb                   0%          25%       50%.         75%        100%
   !• I   !• I   I* I « I   [•[•[•|«I»1«I»I>|8|
correspond ing

4-hour afterffoon
    •
    •
•  •
                     0%
     25%
•    •    •
• •  •    •
• i • i ii »i Si
                          50
            75%
100%
wi nd

d i rect ion
wi ndspeed
  l45°-325°

    77-6%


below average


    85.4%
             3250-145'

               22.4%



        above average



               14.6%
relat ive
humid i ty
temperature
traffic
count
traffic
speed
56.8%

48.8%
79-2%

95-8%

43.2%

51.2%
20.8%

4.2%

     TABLE  6-1.   DESCRIPTION  OF  ANCILLARY  DATA ACCOMPANYING  THE  UPPER

                 FIVE  PERCENT OF THE  CO  CONCENTRATIONS.
                                     26

-------
time of occurrence
                            1/75
                        1/76
                   • •      •••••   •
                   • •  i •• ••••••  •
                    1/77           H/77
correspond ing
daily max.
CO
                     0%
                251
          50%
 100%
correspond ing
4-hour afternoon
SO,
   I*I•I  1*1
                                   25%
                           50%

                      75%
100%
wi nd
d i rect ion
windspeed
  l45°-325°
    92.9%

below average

    67-4%
        7-1%

above average

     32.6%
relat i ve
humid i ty
temperature
traffic
count
traffic
speed
22.
30.
50.
33-
6%
0%
0%
3%
77.4%
70.0%
50.0%
66.7%
     TABLE 6-2.   DESCRIPTION  OF  ANCILLARY  DATA ACCOMPANYING THE UPPER
                 FIVE PERCENT OF THE  Pb  CONCENTRATIONS
                                     27

-------
time of occurrence
                              ••* i  •
                              *i   i  *
                         I  •  ••••••• •    I ••••• !•	&•	|
                         1/75          1/76         1/77          11/77
corresponding                            «    • I  •    •
                                              1,
daily max.           ,« ,   • .  S.M   I    1,
C°                   0?          25%        50%         75%        100?
corresponding                            •
4-hour afternoon             •    •    •••
SO
                     0%         25^         50^          75?       100?
wind                l45°-325°           325°-l45°
direction             9^.7?               5-3?

                  below average    above average

windspeed             45-9?             54.U

relative              13-5?             86.5?
humid i ty

temperature           41.7?             58.3?

traffic               71-4?             28.6?
count

traffic               28.6?             71.4?
speed

     TABLE 6-3-  DESCRIPTION OF ANCILLARY DATA ACCOMPANYING THE UPPER
                 FIVE PERCENT OF THE SO^ (HI-VOL) CONCENTRATIONS
                                     28

-------
                                  SECTION 7

  FREQUENCY DISTRIBUTIONS FO*R HOURLY AND 4-HOURLY POLLUTANT CONCENTRATIONS


CLASS OF POWER TRANSFORMATIONS  (MODEL 7.1)

     In Section 3 the monthly 25th, 50th, 75th, and 90th percentiles of the
hourly CO measurements were plotted.  Quarterly percentiles of the 4-hr
afternoon SO? concentrations were presented in Section 5.  The distributions
of the pollutant measurements appear to be long tailed and skewed to the
right.   Various models for such distributions have been put forward in the
literature, in particular the lognormal distribution, Aitchinson and Brown
(1957).

     There is substantial literature, Zimmer and Larsen (1965), Larsen
(1969,  1973), Michels (1971), and Kornreich (197*0, which indicates that the
lognormal distribution is generally useful in representing ambient air
quality data.  Theoretical reasons for believing so have been put forward by
Singpurwalla (1972), and Kahn (1973).  However, cases have been reported in
the literature, for example, by Kalpasanov and Kurchatova (1976), DeNevers
and Lee (1977), in which the data deviate significantly from the lognormal
hypothesis.

     In this section a more general class of distributions, which includes
the lognormal distribution as a special case, is considered.  This particular
class of distributions (which are sometimes called Box-Cox transformations,
due to the work by Box and Cox  [1964]) arises from power transformations.
Specifically, letting z. be the original observation, it is assumed that the
transformed variable x.,
                      1        _ \ _ i
                                                                      (7-1)
                         x. =


is normally distributed with mean y and variance a2.

     In Model 7-1, X is the transformation  parameter.   I f A = 1> the original z.
                                       z.A-1                                 '
is normally distributed; if A +0, 1im —^— = log z. is normally distributed,

i.e., z.  follows a lognormal distribution; if A = 1/2, the square root of
the original  observation is normally distributed.

     The introduction of the additional parameter A makes  it possible to
check whether the lognormal distribution provides a good overall representa-
tion of the frequency distribution or whether other power transformations


                                     29

-------
(such as square root or cube root) are more suitable.  The distribution of
the z.  can be written as:
     f(z.|X,y,a2) =
r
1

/27TCT
or
1
/2-rra
X ] ]


20 ^

1 s 1
2i 2a2
z.X-l
[' , , 1 2 i JT_ ,- -\ _L n

X

-Finn -7 -U 1 T for X — D

                                                                     (7.2)
Maximum likelihood estimates for the parameters X,  y, and a2  can  be derived
and the interested reader  is referred to the appendix  for technical details.

CO Analysis

     This approach is  illustrated for the hourly CO measurements  in Table 7~la,
for the 4-hr afternoon SO?  (hi-vol and membrane) measurements  and  for
the Pb data shown  in Tables 7~2a through 7~4a  (all for Site C) .

     The hourly CO concentrations were divided  into consecutive 3~month
periods to block out possible seasonal and/or  trend effects.   The  estimates
for X are given in the first column of Table 7-la.  Notice that X  varied
closely around  .50,  indicating that the square  roots of the hourly CO concen-
trations followed a  normal  distribution.  In columns 2 and 3 of Table 7-la.
estimates for y and  a2 are  given for the various seasons by constraining X
to be the same over  the entire time span.   In  this case the maximum likeli-
hood estimate of X was .50, further supporting  that the square root was the
appropriate transformation  to consider, and that the hourly CO data at Site
C did not follow a lognormal distribution.

SO, Analysis
      In  column  1 of Tables 7-2a and 7-3a estimates are given of X  for
 consecutive  6-month periods of the daily 4-hr afternoon hi-vol and membrane
 readings.  Compared with the values of X for CO  in Table 7;-la the  variability
 of  X  across  the periods was considerably greater for the SO? readings.
 Columns  2 and 3 present estimates of y and a2 for the various periods  by
constraining  X  to  be the same for all seasons.  The estimates were  .40 and
.35  for SO? hi-vol and SO? membrane  readings  respectively.  These estimate
indicated  the cube root transformed  the observations to near normality.
Pb Analysis

     The same analysis  is given for the Pb data at Site C.  No  transformation
was necessary for  the daily 4-hr afternoon Pb  readings  (overall estimate of X
was .85).  The readings themselves appeared to be distributed normally.
                                     30

-------
DISTRIBUTIONS FOR THE UPPER TAIL

     In studying air pollutant data, interest often centers on high readings.
Although the class of power transformations may provide a good oveA&tt
description of the frequency distribution, this class of distributions
does not necessarily always give a good fit.  The upper tail of the
distribution is of the main interest.

     Thus focus was placed on the high values of the ambient air quality
data and a distribution was proposed which provided an adequate description
of the upper tail probabilities.  In particular, the observations exceeding
some threshold value z  were assumed to follow a Pareto distribution
                                     rZ0 > °                   (7'3)
                                 and a > 1

Given z~, maximum likelihood estimates of the parameter a were easily derived
(details are given in the appendix).

     In deciding on the threshold value z ,  excessive values were avoided
since a sufficient number of observations have to be available for an effi-
cient estimation of the parameter a.

     For the hourly CO data at Site C, 8.75 ppm was chosen as the value of
ZQ.  This value was approximately the 95th percent! le of all hourly CO
measurements.  Again the parameter a was estimated separately for consecutive
3-month periods to block out possible seasonal and/or trend variation.

     The estimates of a are given in the first column of Table 7-lb.   The
second column compares the empirical frequency distribution with the theo-
retical Pareto distribution with parameter a.  Using the null hypothesis
which states that data are from a Pareto distribution, the statistic in
column 2 follows approximately a Chi-square distribution with eight degrees
of freedom (nine separate intervals were considered).  Comparing these
entries with the tabulated values from a Chi-square distribution with eight
degrees of freedom (xi(-05) = 15-5; x§(-°')  = 20.1) only one was particularly
large (for September T975 through November 1975.)

     In Tables 7~2b through l~kb estimates of ex and Chi-square goodness-of-
fit statistics are given for k-hr afternoon hi-vol and membrane SO? and Pb.
In order to obtain enough observations for estimation, the cutoff point z~ in
the Pareto distribution was chosen as the 75th percent! le of all 4-hr after-
noon data.  The agreement with the Pareto distribution was excellent in most
cases.

AN APPLICATION OF THE PARETO DISTRIBUTION TO DETERMINE TAIL  PROBABILITIES

     The Pareto distribution was used to assess the probability of exceeding
a certain standard.
                                     31

-------
     The first step was to  let z  be the air quality standard for  hourly
CO concentration, and the assumption that z  > z~ be the  threshold value
of the Pareto distribution.  The probability that z > z   then becomes:

               P(z > zs) =  P(z > z0)P(z > zjz > ZQ)                  (7.It)

The first factor  in Model 1.k was determined by the choice of the  threshold
value z_; the second term depended on the parameter in the Pareto
d i str i but ion.

     Specifically,
                                        z    1-a
                    P(z > zs[z > ZQ) =  ^                           (7-5)


As an  illustration we considered July, August, and  September  1977-   The
estimate of a was a = 8.30  for this period.   Furthermore,  P(z  >  8,75 ppm)
was  equal  to  .0509.

     Twenty ppm was assumed an important standard for  hourly  CO  concentra-
tions.  The probability of  exceeding 20 ppm  was  then

     P(z > 20 ppm) = P(x >  8.75 ppm)P(z > 20 ppm|z  > 8.75  ppm) = .0001219.
 We  let  x  be  the  number of  times  20  ppm was exceeded.   For  the  roughly
 2000  hourly  CO measurements which were collected over  a  period of  three
 months, the  model  shows:

                         P(x = 0) = 0.7837
                         P(x = 1) = 0.1910
                         P(x = 2) = 0.0233
                         P(x > 3) = 0.0020

 The expected number  of exceedances  was E(x) =  .2^37-

      These calculations are for  illustration only.   However, studies of
 this  kind can be useful  in the formulation and assessment  of air quality
 standards.
                                       32

-------
TABLE 7-la.  ESTIMATES OF A, y, a2 FOR BOX-COX TRANSFORMATION:
             LOS ANGELES HOURLY CO DATA, SITE C
Time Period
June, July, August (197*0
September, October, November (197*0
December, January, February (1975)
March, April, May (1975)
June, July, August (1974)
September, October, November (1975)
December, January, February (1976)
March, April, May (1976)
June, July, August (1976)
September, October, November (1976)
December, January, February (1977)
March, April , May (1977)
June, July, August (1977)
^
X
0.65
0.50
0.35
0.55
0.60
0.45
0.45
0.55
0.45
0.50
0.55
0.55
0.60
X = .50
y
1 .86
2.09
1.70
1.94
2.22
2.57
2.06
1.71
2.07
2.13
2.03
1.73
2.18
a2
1.56
1.85
3-70
1.36
1.23
1 .82
2.30
1.51
1.15
2.36
2.33
1.53
1 .42
TABLE 7-lb.
ESTIMATES OF a FOR PARETO DISTRIBUTION WITH
LOS ANGELES HOURLY CO DATA,  SITE C
ZQ= 8.75:
Time Period
June, July, August (1974)
September, October, November (1974)
December, January, February (1975)
March, April, May (1975)
June, July, August (1974)
September, October, November (1975)
December, January, February (1976)
March, April , May (1976)
June, July, August (1976)
September, October, November (1976)
December, January, February (1977)
March, April , May (1977)
June, July, August (1977)
a
10.33
6.29
4.84
8.69
10.18
5-53
6.16
7-35
10.77
5.23
5-84
8-35
8.30
2
*8
2.04
20.06
16.42
17-31
9-15
29.83
19-31
9.26
1-34
17-28
8.14
9.64
2.43
                                33

-------
TABLE 7-2a.  ESTIMATES OF X, y, o2 FOR BOX-COX TRANSFORMATION:
             LOS ANGELES 4-HR AFTERNOON SO^ DATA  (HI-VOL),
             SITE C
Time Period
November 1974 - April 1975
May 1975 October 1975
November 1975 - April 1976
May 1976 October 1976
November 1976 April 1977
May 1977 - September 1977
X
0.75
0.15
-0.10
-0.10
0.45
0.75
A = .40
y
4.81
6.21
5.04
5.19
3.83
5.41
S2
1.76
1.52
1.73
1-27
2.44
1.67
TABLE 7-2b.  ESTIMATES OF a FOR PARETO DISTRIBUTION WITH z  AS THE
             75th PERCENT! LE:  LOS ANGELES 4-HR AFTERNOON S0j= DATA
             (HI-VOL), SITE C
Time Period
November 1974 - April 1975
May 1975 - October 1975
November 1975 - April 1976
May 1976 - October 1976
November 1976 - April 1977
May 1977 - September 1977
a
7.40
4.54
4.78
4.54
4.84
6.41
4
13.20
7.07
8.82
16.64
9-89
6.31
                                34

-------
TABLE 7-3a.  ESTIMATES OF X, y, a2 FOR BOX-COX TRANSFORMATION:
             LOS ANGELES 4-HR AFTERNOON SOg DATA (MEMBRANE) ,
             SITE C.                      4
Time Period
November
May 1975
November
May 1976
November
May 1977
197^ - April 1975
- October 1975
1975 - April 1976
- October 1976
1976 - April 1977
- September 1977
A
-0.
-0.
0.
0.
0.
0.
70
15
25
35
30
55
X = .35
y
3
4
2
3
2
3
-94
.58
-49
.55
-52
.52
a
2.
1 .
1 .
2.
2.
2.
2
41
75
58
48
56
01
TABLE 7-3b. ESTIMATES OF a FOR PARETO DISTRIBUTION WITH z AS
75th PERCENT! LE: LOS ANGELES 4-HR AFTERNOON SO?
(HI-VOL) , SITE C
THE
DATA


Time Period
November 1974 - April 1975
May 1975 - October 1975
November 1975 - April 1976


May 1976 - October 1976
November 1976 - April 1977
May 1977 - September 1977
a
3.06
3-19
*§
1.73
5.88
too few data points
above overal 1 75th
percent i le
3.61
4.16
4.47
5.80
16.45
12.44








                               35

-------
TABLE 7-4a.   ESTIMATES OF X, y, a2 FOR BOX-COX TRANSFORMATION:
             LOS ANGELES 4-HR AFTERNOON Pb DATA, SITE C
Time Period
November 1975 April 1975
May 1975 - October 1975
November 1975 - April 1976
May 1976 - October 1976
November 1976 April 1977
May 1977 - September 1977
X
1 .00
0.50
1.15
0.00
0.85
1 .00
A = .85
y
5.00
5-78
4.30
4.76
6.21
7-90
. 52
4.50
4.04
2.81
2.78
6.31
2.84
TABLE 7-4b.  ESTIMATES OF a FOR PARETO DISTRIBUTION WITH z  AS THE
             75th OVERALL PERCENTILE:  LOS ANGELES 4-HR AFTERNOON
             Pb DATA, SITE C.
Time Period
November
May 1975
November
May 1976
November
May 1977
1974 - April 1975
- October 1975
1975 April 1976
- October 1976
1976 - April 1977
- September 1977
a
8.81
5.85
11 .00
10.88
8.41
6.04
2
X8
7.31
14.29
1.75
10.82
15.98
24.07
                                36

-------
                                  SECTION 8

                      ANALYSIS OF 0   NO, AND N02 DATA


     In the LACS program, chemi1uminescence instruments for monitoring
nitric oxide (NO) and nitrogen dioxide  (NO ) were added in January 1975 and
instruments for measuring ozone (0 ) were added in January 1976.  Hourly
observations for these three pollutants were made at Sites A and C.

     A preliminary analysis of the 4-hr afternoon (3 p.m.  to 7 p.m.)  0 ,  NO,
and N0_ measurements was conducted.  The afternoon period  was of particular
interest for trend assessment since during the afternoon hours the wind
usually blew roughly perpendicular to the freeway.  Thus the difference in
the readings at Sites C and A should have reflected the contribution  from
the freeway traffic.

MONTHLY MEANS OF 4-HR AFTERNOON 0   NO, AND NO


     The monthly averages of the 4-hr afternoon readings of 0.,, NO, and N0~
at Sites A and C as well as their across-the-freeway difference (C-A)  are
plotted in Figures 8-la through 8-3c.   From these plots we made the following
observat ions:

  •   3:  The level of 0-a_tjt he background site  (A) was considerably
     Jljjjher than 0, at Sfte C.  Since the NO emitted from the automobiles
     reacted very rapidly with the available 0, to form N0_, the background
     0, was essentially consumed before it reached Site C.

          0, was highly seasonal.   High concentrations occurred in the summer
     months because of increased solar  radiation and strong and persisent
     night and daytime inversions (Tiao, Box, and Hamming [1975]), and Tiao,
     Phadke, and Box [1976]).

          At both Sites A and C,  the concentrations of 0_ decreased from
     1976 to 1977;  a slight reduction also occurred in tne across-the-
     freeway difference.  This observation was confirmed by the sample means
     and standard deviations of (C-A) for the periods May-October  197& and
     May-October 1977 (Table 8-la).

  •  NO:  The NO concentration at Site  A practically vanished  during  the
     s~ummer months when winds blew from the ocean towards Site  C,  indicating
     there was very low NO background  in this area.  During the winter
     months we observed some NO at Site A since in  the winter  period  winds
     occasionally blew towards the sea.

                                      37

-------
     While Figure 8-2a through 8-2c exhibits very  little change  in the
background level, it does show a significant increase from  1975  to 1977
in the NO concentrations at Site C and  in the across-the-freeway differ-
ence (C-A).  Means and standard deviations for May-October  1975, 1976,
and 1977 are given in Table 8-1.
NO
	2_:  The monthly means at Site C were  considerably higher  than
those at Site A,  indicating a substantial contribution to NO  from
the freeway.  Although NO  is not considered a primary pollutant
from the automobile, it quickly formed  between Sites A and  C through
the reaction of NO with the available 0  .

     As  in the NO data, very few yearly changes were noticed in  NO,., at
the background station.  On the other hand, a significant increase from
1975 to  1977 in the concentrations at Site C and  in the across-the-
freeway  differences was observed.   The averages and standard deviations
of  (C-A) for the  period May-October are given in Table 8-1.
TABLE 8-1.  OBSERVATIONS FOR SUMMER ACROSS-THE-FREEWAY DIFFERENCES
                                 FOR POLLUTANTS     •   -\

Pol lutant
°3
Year(s)
1976
1977
Mean
-5-39
-6.29
Standard
Deviat ion
3.43
3.07
Number of
Observat ions
^93
676
t-stati
N/A
N/A
stic

     NO
     NO,
             1977-1976
N/A
N/A
1975
1976
1977
1976-1975
1977-1976
1977-1975
1975
1976
1977
1976-1975
1977-1976
1977-1975
0.273
0.314
0.435
N/A
N/A
N/A
0.0530
0.0623
0.0728
N/A
N/A
N/A
0.192
0.109
0.102
N/A
N/A
N/A
0.0241
0.0358
0.0308
N/A
N/A
N/A
N/A

560
576
668

N/A
N/A
N/A

596
556
688

N/A
N/A
N/A
-4.71

N/A
N/A
N/A

 4.44
20.21
18.86

N/A
N/A
N/A

 5.20
 5.60
12.69
                                38

-------
DIURNAL DIAGRAMS FOR 0 ,  NO, and NO

     Monthly means of the afternoon (3 p.m. to 7 p.m.) period (Figures 8-1 a
through 8-3c) were used to assess the overall  trend in 0_, NO, and N0?.
Summer (May-October), weekday (Monday-Thursday) 1977 diurnal  plots of these
three pollutants at Sites A and C are given in Figures 8-ka through 8-kc.
From these plots the following observations were made:

  •  Ozone at the background station peaks around 2 p.m. in the afternoon.
     As the wind transports ozone across the freeway  it is consumed by NO
     to form NO    This reaction is very fast and, thus, only low 0  concen-
     trations are recorded at Site C.

  •  NO concentrations at Site A are essentially zero from 10 a.m. through
     8 p.m. since the wind  is transporting the NO emissions from the auto-
     mobiles away from Site A towards Site C.   Since part of NO reacts with
     0 ,  NO at Site C peaks after the ozone is depleted.  There is appar-
     ently not enough 0, at the freeway to convert all of the NO automobile
     emissions to N0?.

  •  N0« concentrations at  Site C follow a pattern similar to that of
     ozone at Site A.  The  formation of N0« at Site C reaches its peak at
     the same time when 0,  peaks at Site A and gradually diminishes with the
     depletion of ozone.
                                      39

-------
                                 APPENDIX A

     MAXIMUM LIKELIHOOD ESTIMATES OF  X, y, AND a2 FOR  THE  DISTRIBUTIONS
                        INVOLVING POWER TRANSFORMATIONS

                                                        z.A-l
      It  is assumed that the  transformed variables  x. = — r - (l^i
-------
  MAXIMUM LIKELIHOOD ESTIMATE OF THE PARAMETER a  IN THE PARETO DISTRIBUTION

     Suppose we have n independent observations z.,...,z

from the Pareto distribution with density


               f(z.|a) =  -^z.-,   z. >z  >0                    (A.6)
                         Z0

The joint probability density of z},...,z  is


                            - 1  n n
     f(zrz2,...,zja) = hy^pl  n z.~a, z.>z0 for all l
-------
                                 REFERENCES

Aitchison, J.,  and J.S.C. Brown.  1957-  The Lognormal Distribution.  Cambridge
     University Press, New York, New York.

Beltzer, M., R. J. Campion, and W. L. Peterson.  1974.  Measurement of Vehicle
     Particulate Emissions.  Society of Automotive Engineers Paper 740286,
     February 1974.

Bockian, A. H., G. Tsou, D. Gibbons,  and R. Reynolds.  1977-  Sulfate Concen-
     trations at Two Los Angeles Freeways.  The Los Angeles Catalyst Study
     Symposium.  EPA-600/4-77-034, U.S. Environmental Protection Agency,
     Washington, D. C.,  p. 281.

Box, G.E.P., and D. R. Cox.  1964.  An Analysis of Transformations.  J. Royal
     Stat. Soc., Series B, 26:  211.

Cleveland, W. S.,  B. Kleiner, and J.  L. Warner.  1976.  Robust Statistical
     Methods and Photochemical  Air Pollution Data.  J. Air Pollut. Control
     Assoc., 26:36.

DeNevers, N., and  K. W.  Lee. 1977.  Extreme Values in TSP Distribution
     Functions. J. Air Pollut.  Control Assoc., 27:995.

Hirschler, D. A.,  L. F.  Gilbert, F.  W. Lamb, and L. M. Niebylski.  1957.
     Particulate Lead Compounds in Automobile Exhaust Gas.  Ind. Eng.
     Chem., 49:1131•

Kahn, H.   1973-  Distribution of Air Pollutants.  J. Air Pollut. Control
     Assoc., 23:973.

Kalpasanov, Y., and G. Kurchatova.  1976.  A Study of the Statistical
     Distribution of Chemical Pollutants  in Air.  J. Air Pollut. Control
     Assoc., 26:981

Kornreich, L. D. ed. 1974.  Proceedings of the Symposium on Statistical
     Aspects of Air Quality Data.  EPA-650/4-74-038, U.S. Environmental
     Protection Agency, Washington,  D.C.

Larsen, R.  I.   1969-  A New Mathematical Model of Air Pollutant  Concentration
     Averaging  Time and Frequency.  J. Air Pollut. Control Assoc., 19:24.

Larsen, R.  I.   1973-  An Air Quality Data Analysis System for  Interrelating
     Effects, Standards and Needed Source Reductions.  J. Air  Pollut.  Control
     Assoc., 23:993-


                                     42

-------
Michels, D. E.   1971-  Lognormal Distribution for  Plutonium  in  the  Outdoors.
     In:  Proceedings of Environmental Plutonium Symposium.   E.  B.  Fowler,
     R. W. Henderson, and M. F. Milligan, eds., Kept.  LA-4756,  Los  Alamos
     Scientific  Laboratory.

McGill, R., J. W. Tukey, and W. A. Larson.   1978.  Variations of  Box  Plots.
     The Am.  Statist.  32:12.

Parry,  E. P.,  R. A. Meyer, and C. E. Rodes.  1977-  Determination of  Percen-
     tage of Diesel Trucks and Catalyst Equipped Cars.   In:   The  Los  Angeles
     Catalyst Study Symposium, EPA-600/4-77-034, U.S.  Environmental Protection
     Agency,  Washington, D. C., p. 147.

Phadke, M. S., G. C. Tiao, and S. C. Hillmer.   1977-   Statistical Evaluation
     of the Environmental Impact of the Catalytic  Converter.  Paper presented
     at the 1977 Annual Meeting of the Air Pollution Control  Association,
     Toronto,  Canada, June 1977-

Rodes,  C. E.,  and G. F- Evans.  1977-  Summary of  LACS  Integrated Pollutant
     Data.  In:  The Los Angeles Catalyst Study Symposium, EPA-600/4-77-034,
     U. S. Environmental Protection Agency, Washington, D. C., p.301.

Singpurwalla,  N. D.  1972.  Extreme Values from a  Lognormal Law with Appli-
     cations to  Air Pollution Problems.  Technometrics, 14:703.

Tiao, G. C.,  G.E.P. Box, and W. J. Hamming.  1974.  A  Statistical Analysis of
     the Los Angeles Ambient Carbon Monoxide Data.  Technical Report No. 265
     Department  of Statistics, University of Wisconsin, Madison, Wisconsin.

Tiao, G. C.,  G.E.P. Box, and W. J. Hamming.  1975-  Analysis of the Los Angeles
     Photochemical Smog Data:  A Statistical Overview.  J. Air Pollut. Control
     Assoc.,  25:260.

Tiao, G. C.,  and S. C.  Hillmer.  1977-  Statistical Analysis of the Los Angeles
     Catalyst Study Data - Rationale and Findings.  In:  The  Los Angeles
     Catalyst Study Symposium.  EPA-600/4-77-034,  U.S. Environmental Protec-
     tion Agency, Washington, D. C.  p. 415.  Also in  Environ. Science and
     Technol., 12:820,  July 1978.

Tiao, G. C.,  M.  S. Phadke, and G.E.P. Box.   1976.  Some Empirical Models for
     the Los Angeles Photochemical Smog Data.  J.  Air  Pollut. Control Assoc.
     26:485-

Tukey,  J. W.   1977.  Exploratory Data Analysis.  Addison-Wesley,  Reading,
     Massachusetts.

Zimmer, C. E., and R. I. Larsen.  1965-  Calculating Air Quality  and  its
     Control.   J. Air Pollut. Control Assoc., 15:565-

Zimmerman, J.  R., and R. S. Thompson.  1975.  Users Guide  for HIWAY,  a  Highway
     Pollution Model.  USEPA, February 1975.

                                     43

-------
                                                   N
                                                            SiteD
Figure 1-1. Map of Los Angeles air monitoring stations at San Diego Freeway.

-------
                       1/75
                                          1/76
Figure 3-1a. Monthly means of 4-hour afternoon (3 p.m. to 7 p.m.) CO concentrations. Site A.
         15 r
                      1/75
                                          1/76
  Figure 3-1 b.  Monthly 25th, 50th, 75th, and 90th percentiles of 4-hour afternoon (3 p.m.
  to 7 p.m.) CO concentrations. Site A.

-------
Figure 3-2a.  Monthly means of 4-hour afternoon (3 p.m. to 7 p.m.) CO concentrations, Site C.
  Figure 3-2b. Monthly 25th, 50th, 75th, and 90th percentiles of 4-hour afternoon (3 p.m.
  to 7 p.m.) CO concentrations, Site C.

-------
          7.5
          5.0
           2.5
                           JAN
                                               I
                                                   FEB
                              I    I    I
                                                                         MAR
           2.5
                       I   y   I
                           APR
                                                    MAY
                         I
                                                                         JUN
          2.5
          7.5
          5.0
          2.5 •
                           JUL
                      Jl
                           OCT

                 74    75    76   77
                                                   AUG
          NOV

74    75   76    77
                                                                                 f
                                                                         SEP
       DEC

74    75   76   77
Figure 3-3a.  Plot of 25th, 50th, and 75th percentiles of 4-hour afternoon (3 p.m. to 7 p m.
CO concentrations. Site A.

-------
            12
                          JAN
                                                  FEB
                                                                         MAR
            12
                          APR
                                                    I    I
                                                 MAY
                                                                          JUN
            12
                          JUL
                                                  AUG
                                                                          SEP
            12
                          OCT

                   74   75   76   77
       NOV

74    75    76    77
        DEC

74    75    76   77
Figure 3-3b.  Plot of 25th, 50th, and 75th percentiles of 4-hour afternoon (3 p.m. to 7 p.m.)
CO concentrations, Site C.

-------
                          SUMMER 76

                        	WINTER 75/76
   12    15

HOURS
                              18    21   24
Figure 3-4a. Diurnal plots of CO at Site C;
winter - summer comparison  (weekday).
                                                                         • SUMMER 76

                                                                         - WINTER 75/76
                                                   0369
                                                                     12    15

                                                                   HOURS
                                                                                   21    24
                               Figure 3-4b.  Diurnal plots of CO at Site C;
                               winter - summer comparison  (weekend).

-------
                        - SUMMER 76

                        - WINTER 75/76
                                                                      • SUMMER 76

                                                                      • WINTER 75/76
  0     3    6    9    12    15   18    21   24
                                                 0369
Figure 3-5a. Diurnal plots of CO at Site A;
winter - summer comparison (weekday).
Figure 3-5b. Diurnal plots of CO at Site A;
winter  summer comparison (weekend).
                                          50

-------
 Figure 3-6a.  Diurnal plot of traffic counts
and traffic speed, weekday 1976.
Figure 3-6b.  Diurnal plot of traffic counts
and traffic speed, weekend 1976.
                                                                         COUNT

                                                                         SPEED
                    12    16   20    21
Figure 3-6c.  Diurnal plot of traffic counts
and traffic speed, weekday 1977.
Figure 3-6d.  Diurnal plot of traffic counts
and traffic speed, weekend  1977.
                                          51

-------
                                                           1976
                                                           1977
                      12      16     20     24
                                                                      12      16     20     24
Figure 3-7a.  Diurnal plot of traffic density
(count/speed) weekday.
Figure 3-7b.  Diurnal plot of traffic density
(count/speed) weekend.

-------
                                            —\
                    SUMMER 1976
                    WINTER 1975/76
0.
        0        3       6       9       12      15       18       21
                      Figure 3-8. Diurnal plots of WS..
                 _L
                                   53

-------
     16
     12
   I  8
                                    12    15    18     21     24
Figure 3-9.  Diurnal plot of CO; Sites A, B, F, and C; summer weekday.

-------
      16 r
      12
   t
                                    12
                                          15     18     21    24
Figure 3-10.  Diurnal plot of CO; Sites A, B, F, and C; summer weekend.
                              55

-------
   10
       X  PREDICTED

      —  OBSERVED
Q.
Q.
                                    12
15
18
21
24
               Figure 3-11. CO model 1977 summer weekday fit. Site C.

-------
    1.5
    1.0
o
h-

+


O
o
    0.5
                                           ws.
               Figure 3-12. Plot of the ratio CO/ («+ kTD) versus WS, ; Site F.

-------
 16
 14
 10
  0
 X  PREDICTED
—  OBSERVED
        0      36      9     12     15    18     21     24
gure 3-13.  CO model 1977 summer  weekday fit, Site F (median strip),
                           53

-------
                        X PREDICTED

                       	 OBSERVED
                       X PREDICTED

                      — OBSERVED
                                                                            XX*-*
Figure 3-14a. Joint CO model fit summer
weekday, 1977, Site A.
Figure 3-14b. Joint CO  model fit summer
weekday, 1977, Site B.
                         X PREDICTED

                        _ OBSERVED
                                                                    X PREDICTED

                                                                    — OBSERVED
                                                  0    3    6   9    12   15
Figure 3-14c.  Joint CO model fit summer       Figure 3-14d Joint CO model fit summer
weekday, 1977, Site C.                       weekday, 1977, Site F.
                                       59

-------
Figure 4-1a. Monthly means of 4-hour afternoon Pb concentrations. Site A.
Figure 4-1b. Monthly means of 4-hour afternoon Pb concentrations. Site B.
                               60

-------
 Figure 4-1c. Monthly means of 4-hour afternoon Pb concentrations. Site C.
Figure 4-1d. Monthly means of 4-hour afternoon Pb concentrations. Site D.
                                61

-------
    Figure 4-2a.  Monthly means of afternoon (3 p.m. to 7 p.m.) windspeed.
Figure 4-2b.  Monthly means of afternoon (3 p.m. to 7 p.m.) winds from 145° 325°
                                  62

-------
     Figure 4-3a. Monthly means of 24-hour Pb concentrations. Site C.
I
                                     1/76
     Figure 4-3b. Monthly means of 24-hour Pb concentrations, Site D.
                                  63

-------
                                                       WEEKEND

                                                       WEEKDAY
Figure 4-4a.  Monthly means of 4-hour (3 p.m. to 7 p.m.) Pb readings weekday  weekend
comparison.  Site C.
                                               	 WEEKDAY
Figure 4-4b. Monthly means of 4-hour (3 p.m. to 7 p.m.) Pb readings weekday - weekend
comparison. Site D.

-------
 *  2
 u
o
H


i
                       20                 40                 60



                            TRAFFIC SPEED TS, mph
                    Figure 4-5a. Plot of Pb/TC vs TS.
                                            •




                                                •
                •4-202


                                 WS^.mph





                 Figure 4-5b. Plot of Pb/TC x TS vs





                                  65

-------
      E
      •a
                                                    	 MEMBRANE
Figure  5-1a. Monthly means of 4»hour hi-vol and membrane afternoon readings of SOg; Site A.
                                                     	 MEMBRANE
     Figure 5-1b. Monthly means of 4-hour hi-vol afternoon readings of SOg; Site B.
                                        66

-------
                                                      	 MEMBRANE
                      1/75
                                         1/76
Figure 5-1c. Monthly means of 4-hour hi-vol and membrane readings of SOg; Site C
     I
                                                  	 MEMBRANE
                          1/75
                                            1/76
 Figure 5-1d.  Monthly means of 4-hour hi-vol afternoon readings of SO^;  Site D.
                                    67

-------
                      T
                                SU    FA    VUl    SP    SU    FA    Wi    SP    SU    FA


                                                 1976                  1977
 Figure 5-2a.  Plot of quarterly 25th, 50th, and 75th percentiles, minimum and maximum of 4-hour
 hi-vol afternoon SO5 concentrations at Site C.
                                     T
                       WI    SP   SU     FA   WI    SP    SU    FA    WI    SP    SU    FA

                             '975                  1976                  1977
Figure 5-2b. Plot of quarterly 25th, 50th, and 75th percentiles, minimum and maximum of 4-hour
membrane afternoon SO| concentrations at Site C.
                                            68

-------
1
                     1/75
                                       1/76
I
        Figure 5-3a.  Monthly means of 4-hour hi-vol SCN (C - A) differences.
                      1/75
                                        1/76
      Figure 5-3b. Monthly means of 4-hour membrane S0| (C - A) differences.

-------
                        SP    SU    FA    Wl    SP     SU    FA    Wl    SF    SL    F,

                          197C                   1976                  197"
Figure 5-4a.  Plot of quarterly 25th, 50th, and 75th percentiles of 4-hour hi-vol afternoon
SO  (C - A) differences.
                   Wl    SP    SU    FA    Wl    SP    SU    FA    Wl     SP    SU    FA
                          1975                   1976                   1977


Figure 5-4b. Plot of quarterly 25th, 50th, and 75th percentiles of 4-hour membrane
afternoon S0| (C - A) differences.
                                          70

-------
Figure 8-1a.  Monthly means of 4-hour (3 p.m. to 7 p.m.) 63 at Site A.
Figure 8-1b.  Monthly means of 4-hour (3 p.m to 7 p.m.) 03 at Site C.
Figure 8-1 c. Monthly means of 4-hour (3 p.m. to 7 p.m.) 63 (C A).
                            71

-------
Figure 8-2a.  Monthly means of 4-hour (3 p.m. to 7 p.m.) NO at Site A.
Figure 8-2b. Monthly means of 4-hour (3 p.m. to 7 p m.) NO at Site C.
Figure 8-2c. Monthly means of 4-hour (3 p.m. to 7 p.m.) NO (C - A).
                            72

-------
                                          A
 Figure 8-3a.  Monthly means of 4-hour (3 p.m. to 7 p.m.) NO2 at Site A.
Figure 8-3b. Monthly means of 4-hour (3 p.m. to 7 p.m.) NO2 at Site C.
Figure 8-3c.  Monthly means of 4-hour (3 p.m. to 7 p.m.) ISIO2 (C  A).
                              73

-------
0    36    9    12   15    18   21   24
                                              0    3    6    9   12   15   18   21   24
     Figure 8-4a. Diurnal plots of 03 (Monday - Thursday) at Sites A and C.
                                  74

-------
0    3    6    9   12    15   18   21   24
                                              0    3    6    9   12   15    18   21   24
    Figure 8-4b.  Diurnal plots of NO (Monday - Thursday) at Sites A and C.
                                    75

-------
0    3    6    9    12   15   18    21   24
                                              0    3    6   9   12    15   18   21   24
 Figure 8-4c.  Diurnal plots of N0£ (Monday - Thursday) at Sites A and C.
                                76

-------
                                    TECHNICAL REPORT DATA
                             (Please read Instructions on the reverse before completing)
 1. REPORT NO.
   EPA 600/4-79-070
                                                            3. RECIPIENT'S ACCESSI Of* NO.
 4. TITLE AND SUBTITLE

    STATISTICAL ANALYSIS  OF  THE  LOS ANGELES CATALYST
    STUDY DATA
              5. REPORT DATE
                        1979
              6. PERFORMING ORGANIZATION CODE
 7.AUTHOR(s) Johannes Ledolter,  George C. Tiao, Spencer  B.
    Graves, Jian-tu Hsieh,  Gregory B.  Hudak  (U. of  Wi.sc)
           R  Sai.U. FPA/FMSI,  RTF,  N,
              8. PERFORMING ORGANIZATION REPORT NO.
 9. PERFORMING ORGANIZATION NAME AND ADDRESS

    Department of Statistics
    University of Wisconsin
    Madison,  Wisconsin 53706
              10. PROGRAM ELEMENT NO.
              11. CONTRACT/GRANT NO.

                68-02-2261
 12. SPONSORING AGENCY NAME AND ADDRESS
    Environmental Monitoring  and Support Laboratory
    Office of Research and  Development
    U.S.  Environmental Protection Agency
    Research Triangle Park,  N.C. 27711
                                                            13. TYPE OF REPORT AND PERIOD COVERED
              14. SPONSORING AGENCY CODE

               EPA-600/08
 15. SUPPLEMENTARY NOTES
 16. ABSTRACT            •                                       ~~	—	•—
         This research was  initiated to perform  statistical  analyses of the
    data from the Los Angeles Catalyst Study.  The  objective is to determine
    the effects of the  introduction of the catalytic  converter upon the
    atmospheric concentration levels of a number  of air pollutants.

         This report  gives  an analysis of the  CO, Pb,  SO,,  0_, NO and HO^
    data covering the period from June 197^  to November T977-   Models are
    built to evaluate the freeway contribution to CO  and Pb as a function of
    traffic, windspeed and  wind direction.   These models are used to assess
    both the time trend  in  the pollutant measurements  and the pollution con-
    centrations at points near the freeway.  Furthermore frequency distributions
    for ambient air quality data near freeways are  discussed.

         This report  was  submitted in fulfillment of  U.S. E.P.A. Contract
    No. 68-02-2261 with  the University of Wisconsin.   It covers work done
    during the period September 1977 through August 1978.
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
                                              b.IDENTIFIERS/OPEN ENDEDTERMS
                           c.  COSATI Field/Group
  mobile  source pollution
  freeway pollution
  mobile  pollution  modeling
  CO model ing
  mobile  source monitoring
  freeway monitoring
  catalyst effects
 Los Angeles
 catalytic converter
 ozone
 lead
 sulfate
 carbon monoxide
  43F
  68A
 3. DISTRIBUTION STATEMENT
    Release to  public
19. SECURITY CLASS (This Report)
   Unclass if ied
                           21. NO. OF PAGES
77
                                              20. SECURITY CLASS (This pagej

                                                 Unclassified	
                                                                         22. PRICE
EPA Form 2220-1 (9-73)
                                            77

-------