EPA-R4-73-012a




October  1972
Environmental Monitoring  Series

-------
                                            EPA-R4-73-012a
                 EVALUATION
                      OF A
             DIFFUSION  MODEL
                       FOR
PHOTOCHEMICAL  SMOG  SIMULATION
                FINAL  REPORT
                         by
       A.Q. Eschenroeder,  J.R. Martinez, andR.A. Nordsieck
                General Research Corporation
                    P.O. Box 3587
               Santa Barbara, California 93105
                 Contract No. 68-02-0336
                Program Element No. 1A1009
             EPA' Project Officer: Ralph C .  Sklarew

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

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

                     October 1972

-------
This report has been reviewed by the Envirpnmental Protection Agency




'approved for publication.  Approval does not signify that the cpntents




necessarily reflect the views and policies, pf the Agency, npr dpe§




mention of trade names, or cpmmereial products constitute endprsenient




or recommendation for use.
                                  ii

-------
                                CONTENTS


SECTION     	   PAGE


            OVERVIEW AND SUMMARY OF THE WORK                           1
  I         STRUCTURE OF THE REPORT                                    9
  2         ANALYSIS OF CHEMICAL KINETICS IN SMOG CHAMBER
            EXPERIMENTS                                               11
            2.1   Introduction                                        11
            2.2   Chemical Mechanism                                  11
            2.3   Methodology of Adjustments and Evaluations           17
            2.4   Experimental Data Base                              18
            2.5   Results of Smog Chamber Validations                  20
            2.6   Adaptation of Chemical Model to Atmospheric
                  Modeling                                            43
  3         MODEL METHODOLOGY IMPROVEMENTS                            51
            3.1   Perspectives on Model Updating                      51
            3.2   Advective Air Trajectories                          51
            3.3   Vertical and Horizontal Eddy Diffusion
                  Coefficients                                        58
            3.4   Emission Flux Histories                             65
  4         TRANSVERSE DIFFUSION AND ITS EFFECT ON THE GRC
            MODEL                                                     66
            4.1   Lateral Diffusion Between Neighboring
                  Streamtubes                                         66
            4.2   Lateral Diffusion Effects in the Vicinity of
                  High-Flux Elevated Point Sources                    72
  5         VALIDATION STUDIES                                        79
            5.1   Selection of Days for Model Tests                   79
            5.2   Program Conversion                                  82
                                                                    iii

-------
 CONTENTS (Cont.)
 SECTION
             5.3   Atmospheric Validation Tests
             5.4   Techniques for Model Operation
             5.5   Sources of Uncertainty Due to Solar Radiation
                   and Particulate Reactions
    6        CONCLUDING REMARKS
APPENDIX A   A VIEW OF FUTURE PROBLEMS IN AIR POLLUTION MODELING
             REFERENCES
PAGE
 84
148

152
162
167
207
iv

-------
                               ILLUSTRATIONS
NO.   	  .PAGE



2.1   Experiment 329, Propylene/NO .   Plot of Propylene, NO, NO       23
                                  X                            Ł*


2.2   Experiment 329. Propylene/NO .   Plot of Ozone and PAN           24
                                  X


2.3   Experiment 253, Toluene/n-Butane/NO .   Plot of NO and NO.       27
                                         "X,                    &•


2.4   Experiment 253, Toluene/n-Butane/NO .   Plot of Ozone and PAN    28



2.5   Experiment 253, Toluene/n-Bu€ane/NO .   Plot of Toluene and

      n-Butane                           x                            29



2.6   Experiment 251, Toluene/n-Butane/NO .   Plot of NO and NO        30
                                         X                    Ł•


2.7   Experiment 271, Toluene/NO .  Plot of Toluene, NO, and N0?      32
                                X                              ^


2.8   Experiment 271, Toluene/NO .  Plot of Ozone  and PAN             33
                                X


2.9   Experiment 231, Dilute Auto Exhaust (Controlled Vehicle).

      Plot of NO and N02                                              38



2.10  Experiment 231, Dilute Auto Exhaust (Vehicle with Emission

      Controls).  Simulation of Reactive Hydrocarbon Decay            39



2.11  Experiment 222, Dilute Auto Exhaust (Uncontrolled Vehicle).

      Plot of NO and N02                                              41



2.12  Experiment 222, Dilute Auto Exhaust (Uncontrolled Vehicle).

      Plot of Ozone  and PAN                                           42



2.13  Experiment 336.  Dilution Model Compared with Measured

      Ethane Concentration                                            47



2.14  Experiment 336.  Effect of Dilution on Propylene Concentration.

      (Curves Computed using a Single set of Rate Constants)          48



3.1   Air Trajectory in the Los Angeles Basin                         53



3.2   Comparison of Ground Trajectory with Tetroon Trajectory         54



3.3   Comparison of Wind Speed Measurements at El Monte               56

-------
  ILLUSTRATIONS  (Cont.)



  NO.	 PAGE

  3.4   Comparison of Wind Direction Measurements at El Monte           57

  3.5   Vertical Diffusivity Versus Wind Speed                          60

  3.6   Vertical Diffusivity Versus Vertical Temperature Gradient       61

  3.7   Vertical Diffusivity Profiles                                   63

  3.8   Lateral Separation of Simultaneously Released Tetroon Pairs
       as a Function of Time Since Release                             65

  4.1   CO Flux Histories on Two Neighboring Trajectories               67

  4,2   Eddy Diffusivity Profile for Neighboring Streamtube Analysis    67

  4.3   CO Concentration Histories on Right-Hand (Commerce)
       Trajectory                                                      68

  4.4   CO Concentration Histories on Left-Hand (Synthetic)
       Trajectory                                                      68

  4.5   Superposed CO Flux Histories for 24 Trajectories                69

  4.6   Bounding CO Flux Histories for Parallel Trajectory Analysis     70

  4.7   CO Concentration Histories for Worst-Case Parallel Trajectories 70

 4.8   Superposed NO Flux Histories for 24 Trajectories                72

 4.9   Eddy Diffusivity Profile for Elevated Point Source/Lateral
       Diffusion Analysis                                              73

 4.10  Ground Concentration Effects of Elevated Point Sources:
       10 kg/hr/km2 Background Flux                                    75

 4.11  Ground Concentration Effects of Elevated Point Sources:
       15 kg/hr/km2 Background Flux                                    76

 4.12  Ground Concentration Effects of Elevated Point Sources:
       20 kg/hr/km2 Background Flux                                    77

 5.1   Air Quality and Meteorological Monitoring Network in the
       Los Angeles Basin                                               86
VI

-------
ILLUSTRATIONS  (Cont.)
NQ.        . .      .          . _ ., _	     ....     .               PAGE
5.2   Interpolated and Observed Oz&rie Concentration at a Monitor-
      ing Station in the West San Gabriel Valley on September 29,
      1969                                                           93

5.3   Inteirpdiated and observed Ozorife Concentration at a Monitor-
      ing Station in the West San Gabriel Valley on November 4,
      1969                                                           94

5.4   Observed Versus Commuted Carbon Monoxide Concentration         97

5.5   Observed Versus Computed Ozone Concentration                   98

5.6   September 115 1969 Trajectory Starting at Commerce at 0530    100

5.7   Trajectory No. 1—Computed and Observed CO Concentrations     101

5.8   Trajectory No. 1—Computed and Observed NO and N02
      Concentrations                                                101

5.9   Trajectory Nd. I—Computed and Observed Ozone Concentrations  101

5.10  September 11s 1969 Trajectdry Starting at Commerce at 0636    102

5.11  Trajectory No. 2—Computed and Observed CO Concentrations     103

5.12  Trajectory No. 2--Computed arid Observed NO and NO-
      Concentrations                                                103

5.13  Trajectory No. 2--Cdmputed and Observed Ozone Concentratidns  103

5; 14  September 11, 1969 Trajectory Starting in Oowntdwri Los
      Angeles at 0530                                               104

5.15  Trajectory Nd. 3—Computed arid Observed CO Concentrations     105

5;16  Trajectory No. 3--Cdmputed arid Observed NO and N0?
      Concentrations

5.17  Trajectory No. 3—•Cdniputed arid Observed Ozone Cdriceritratidns  105

5.18  September 11, 1969 Trajectory Starting in tiovritowh Los
      Angeles at 0630                                               l06

5il9  Trajeetdry Nd; 4—Gomputed arid Obsgifved CO Cdnceritratioii      107
                                                                     vii

-------
 ILLUSTRATIONS (Cont.)
 NO.
 5.20  Trajectory No.  4—Computed and Observed NO and N02
       Concentrations                                                  107

 5.21  Trajectory No.  4—Computed and Observed Ozone Concentrations   107

 5.22  September 29, 1969 Trajectory Starting at Commerce at 0530     108

 5.23  Trajectory No.  5—Computed and Observed CO Concentrations      109

 5.24  Trajectory No.  5—Computed and Observed NO and NO-
       Concentrations                                                  109

 5.25  Trajectory No.  5—Computed and Observed Ozone Concentrations   109

 5.26  September 29, 1969 Trajectory Starting at Commerce at 0630     110<

 5.27  Trajectory No.  6—Computed and Observed CO Concentrations      111

 5.28  Trajectory No.  6—Computed and Observed NO and NO-
       Concentrations                                                  111

 5.29  Trajectory No.  6—Computed and Observed Ozone Concentrations   111

 5..30  September 29, 1969 Trajectory Starting in Downtown Los
       Angeles  at 0530                                                112

 5.31  Trajectory No.  7—Computed and Observed CO Concentrations      113

 5.32  Trajectory No.  7—Computed and Observed NO and NO
       Concentrations                                                  113

 5.33  Trajecotry No.  7—Computed and Observed Ozone Concentrations   113

 5.34  September 29, 1969 Trajectory Starting Near the Coast at
       0230                                                            114

 5.35  Trajectory No.  8—Computed and Observed CO Concentrations      115

 5.36  Trajectory No.  8—Computed and Observed NO and NO
       Concentrations                                                  115

 5.37  Trajectory No.'  8—Computed and Observed Ozone Concentrations   115

 5.38  September 30, 1969 Trajectory Starting at Commerce at 0530     116
viii

-------
ILLUSTRATIONS (Cont.)
NO.                                                                 PAGE
5.39  Trajectory No. 9—Computed and Observed CO Concentrations      117

5.40  Trajectory No. 9—Computed and Observed NO and N02
      Concentrations                                                 117

5.41  Trajectory No. 9—Computed and Observed Ozone Concentrations   117

5.42  September 30, 1969 Trajectory Starting at Commerce at 0630     118

5.43  Trajectory No. 10—Computed and Observed CO Concentrations     119

5.44  Trajectory No. 10—Computed1 and Observed NO and NO-
      Concentrations                                                 119

5.45  Trajectory No. 10—Computed and Observed Ozone Concentrations  119

5.46  September 30, 1969 Trajectory Starting in Downtown Los
      Angeles at 0430                                                120

5.47  Trajectory No. 11-Computed and Observed CO Concentrations      121'

5.48  Trajectory No. 11—Computed and Observed NO and N02
      Concentrations                                                 121

5.49  Trajectory No. 11—Computed and Observed Ozone Concentrations  121

5.50  September 30, 1969 Trajectory Starting in Downtown Los
      Angeles at 0530                                                122

5.51  Trajectory No. 12—Computed and Observed CO Concentrations     123

5.52  Trajectory No. 12—Computed and Observed NO and N0?
      Concentrations                                                 123

5.53  Trajectory No. 12—Computed and Observed Ozone Concentrations  123

5.54  October 29, 1969 Trajectory Starting in"Downtown Los Angeles
      at 0530                                                        124

5.55  Trajectory No. 13—Computed and Observed CO Concentrations     125

5.56  Trajectory No. 13—Computed and Observed NO and N0?
      Concentrations                                                 125
                                                                     ix

-------
 ILLUSTRATIONS  (Cont.)
 N0._	_	....--   	  PAGE

 5,57  Trajectory No,  13—Computed  and  Observed  Ozone  Concentrations  125

 5.58  October  29,  1969 Trajectory  Starting at 0630 in Downtown
      Los  Angeles                                                     126

 5.59  Trajectory No,  14—Computed  and  Observed  CO  Concentrations     127

 5.60  Trajectory No.  14—Computed  and  Observed  NO  arid N02
      Concentrations                                                  127

 5,61  Trajectory No.  14—Computed  and  Observed  Ozone  Concentrations  127

 5.62  October  29,  1969 Trajectory  Starting at Commerce at 0630       128

 5,63  Trajeettiry No,  15—Computed  and  Observed  CO  Concentrations'     129

 5.64  Trajectory No.  15—Computed  and  Observed  NO  and NO,,
      Concentrations                                                  129

 5.65  Trajectory No.  15—Computed  and  Observed  Ozone  Concentrations  129

 5.66  October  29,  1969 Trajectory  Starting at El Monte at 0830       130

 5.67  Trajectory No.  16—Computed  and  Observed  CO  Concentrations     131

 5.68  Trajectory No.  16—Computed  and  Observed  NO  and NO- Concen-
      trations                                                       131

 5,69  Trajectory No.  16—Computed  and  Observed  Ozone  Concentrations  13l

 5*70  October  30,  1969 Trajectory  Starting at Pasadena at 0530       132

"5*71  Trajectory No,  17—Computed  arid  Observed  CO  Concentrations     133

 5.72  Trajectory No,  17—Computed  arid  Observed  NO  and ML
      Concentrations                                                  133

 5,73  Trajectory No.  17—Computed  and  Observed  Oztifle  Concentrations  133

 5.74  October  30i  1969 Trajectory  Starting at Ccmrniefce at 0630       134

 5.75  Trajectory No.  18—Comp'Uted  and  Observed  CO  Corieeritratioris     135

 5,76  Trajectory No.  18—Computed  and  Observed  NO  a'nd NO'
      Concentrations                                                  135

-------
ILLUSTRATIONS  (Cotit.)
N0,_   .     __                                                        PAGE
5,77  Trajectory No. IS—Computed  and Observed  Ozone  Concentrations  135

5.78  October 30, 1969 Trajectory  Starting at El Monte  at  0630       136

5.79  Trajectory No* 19—Computed  and Observed  ClO  Concentrations     137

5.80  Trajectory No. 19—Computed  and Observed  NO  Arid NO*
      Concentrations                                                 137

5,81  Trajectory No* 19—Computed  and Observed  Ozone  Concentrations  137

5,82  October 30, 1969 Trajectory  Starting in Downtown-Los Angeles
      at 0830                                                        138

5,83  Trajectory No, 20—Computed  and Observed  CO  Concentrations     1^9

5.8:4  Trajectory No, 20—^Computed  and Observed  NO  and NO A
      Concentrations                                                 139

5.85  Trajectory No. 20—Computed  and Observed  Ozone  Concentrations  139

5.86  November 4j 1969 Trajectory  Starting at Commerce  at  0530       140

5,87  Trajectory No, 21—Computed  and Observed  CO  Concentrations     141

5.88  Trajectory No* 21--Computed  and Observed  NO  and N0«
      Concentrations                                                 141

5,S9  Trajectory No. 21—Computed  and Observed  Ozone  Concentrations  141

5.90  November' 4, 1969 Trajectory  Starting at Commerce  at  0630       142

5.91  Trajectory 22-Computed and Observed Co Concentrations          143

5.92-  Trajectory No. 22—Computed  and Observed  NO  and NO,
      Concentrations                                                 143

5,93  Trajectory No-; 22^-Computed  and,Observed  Ozone
      Coneefifitfatlofis                                                 143

5.94  N'ovember 4, 196'9 Trajectory  Starting in Pasadena  at  0530       144

5J.-9S  Trajectory N§« 23--Cdmputed  aiid Observed  CO  Concentration's     145

-------
  ILLUSTRATIONS (Cont.)
  NO.
                                                                      PAGE
  5.96  Trajectory No.  23—Computed and Observed N09  Concentrations    145

  5.96  Trajectory No.  23—Computed and Observed NO and N02
        Concentrations                                                  1"

  5.97  Trajectory No.  23—Computed and Observed Ozone
        Concentrations

  5.98  November 4, 1969 Trajectory Starting in Downtown Los Angeles
        at 0530                                                        146

  5.99  Trajectory No.  24—Computed and Observed CO Concentrations     147

  5.100 Trajectory No.  24—Computed and Observed NO and N02
        Concentrations                                                  147

  5.101 Trajectory No.  24—Computed and Observed Ozone Concentrations  147

  5.102 NO,., Photolysis  Rate Constant,  k. ,  for September 11, 1969       154

  5.103 N02 Photolysis  Rate Constant,  k ,  for September 29, 1969       155

  5.104 N02 Photolysis  Rate Constant,  k_ ,  for September 30, 1969       156

  5.105 N02 Photolysis  Rate Constant,  k ,  for October 29, 1969         157

  5.106 N02 Photolysis  Rate Constant,  k ,  for October 30, 1969         158

  5.107 N02 Photolysis  Rate Constant,  k ,  for November 4, 1969         159

  A.I   (NO + N02)  - Concentration Ground  Level Huntington Park        170

  A.2   CO/NO  Ratios for Huntington Park  1968                         171
             X

  A.3a   Chemiluminescent Measurements  in New York - 1970               184

  A.3b   Chemiluminescent Measurements  in New York - 1970               184

  A.4    Quasiequilibrium Test  for  1969 Ground Data at El Monte-
        High  NO  Levels                                                  187

  A.5    Ozone  Inaccuracies  Needed  to Explain the Departures from
        Quasiequilibrium in Fig. A.4                                   188
xii

-------
ILLUSTRATIONS (Cont.)
NO.	  PAGE

A. 6   Quasiequilibrium in a Simulated.:Smog Chamber Experiment      189

A.7   LAPS Coordinate System                                       197

A.8   Cross Section of Depressed Six-Lane Freeway                  199

A.9   Wind-Oriented Coordinate System                              199

A.10  CO Concentration Profiles Normal to Roadway at Various
      Wind Aspect Angles                                           200

A.11  Ozone and Nitric Oxide in an Air Mass Moving Over a Roadway  200
                                                                    xiii

-------
xiv

-------
                                TABLES
                                                                    PAGE
2,1
2.2
2,3
2.4
2.5
2.6
Basic Kinetic Model
Experimental Data
Rate Constants Used in Propylene Simulations
Rate Constants for Toluene/n-Butane Simulation
Rate Constants for Toluene Experiment 271
Initial Hydrocarbon Concentrations (PPM) for Auto Exhaust
16
19
21
31
35

      Experiments                                                     36

2.7   Rate Constants Used in Simulation of Experiment 231 , Dilute
      Exhaust from a Vehicle with Emission Control                    40

2.8   Rate Constants Used for Simulating Experiment 222 , Dilute
      Exhaust from a Vehicle Without Emission Control                 44

2.9   Mole-Weighted Reactivity of Atmospheric and Smog Chamber
      Hydrocarbon Mixtures                                            49

4.1   "Worst Case" Bounding Error Fractions Due to Omission of
      Lateral Diffusion in Urban Modeling Assuming Zero Initial
      Concentration of Carbon Monoxide                                71

4.2   "Worst Case" Bounding Error Fractions Due to Omission of
      Lateral Diffusion in Urban Modeling, Assuming an Initial
      Concentration of Carbpn Monoxide of 10 ppm                      71

4.3   Maximum Fractional Contribution of an Elevated Point So.urce
      at a Ground Location Two Miles from the Plume Centerline        78

5.1   Directory of Air Quality and Meteorological Monitoring Sta-
      tions in the Los Angeles Basin                                  87

5.2   Trajectory Identification Table                                 89

5.3   Initial Concentrations Used in Atmospheric Simulations          90
                                                                      xv

-------
   TABLES  (Cont.)



   NO.	  PAGE

   5.4   Correlation Coefficients for CO and Ozone                        96

   5.5   Regression Equations for CO and Ozone                            99

   5.6   Rate Constants Used in Atmospheric Modeling Studies             149

   A.I   Concentrations (ppm) and Gas Phase Rate Constant Assumed
        for Comparative Analysis                                        178

   A.2   Upper Limit of (Surface Rate/Gas Phase Rate) Ratio              180

   A.3   Ford/New York Data (First 20 Minutes)                           185

   A.4   Air Quality Effects for 1974 Trajectory                         195
xvi

-------
OVERVIEW AND SUMMARY OF THE WORK
      For the past several years, General Research Corporation (GRC) has
been developing and refining a photochemical/diffusion model for the US
Environmental Protection Agency (EPA) and its predecessor agency.  The
application of this model is the prediction of air quality in terms of
pollutant emission .patterns and meteorological features of a particular
airshed.  Tracing from our first steps down to now, the efforts have
balanced the emphasis between fidelity in the air chemistry and realism
in the fluid dynamic transport.  Unlike earlier static models based on
superposition of plumes, ours is based on time-dependent processes.
Therefore, one significant improvement has been treatment of unsteady
diffusion and another has been the finite-difference formulation to allow
for atmospheric transformation processes.

      The work described in this report combines another round of model
improvements with a controlled evaluation first of chemistry alone, and
finally chemistry combined with diffusion.  The evaluation has been done
in parallel with two other contractors, Pacific Environmental Services
and Systems Applications, Inc., who are pursuing similar tests on- their
models, each of which is somewhat different from ours.  It should be
emphasized that these evaluation studies are carried out in parallel
with the only interaction between contractors being an exchange of
monthly progress letters and occasional informal meetings.

      Before summarizing our findings in a point-by-point narrative, it
is helpful to digress here and review briefly just what the GRC photo-
chemical/diffusion model does and how.  We take an initial state for an
airshed to be the spatial distributions of the concentrations of pollu-
tants of interest; e.g., the parts per million of carbon monoxide, ozone,
hydrocarbon, nitrogen dioxide, and nitric oxide.  We must also specify
the boundary conditions that control how the system evolves from its
initial state.  Strictly speaking, the boundary conditions are limited
to temporal and spatial emission source distributions of the various
pollutants just named.  In an indirect fashion, the kinematics of airflow

-------
including both advection and diffusion are the fundamental boundary condi-
tions for our model, since we follow an air mass-center as it is guided
by the winds from place to place around the air basin.   Where it goes will
influence what pollutant emissions it receives.

      We also trace the upward spread of pollutants in the air mass.as
they are introduced at the ground by emission sources.   The ongoing chemical
changes are simultaneously calculated.  By stratifying the air vertically
in the computation, we determine pollutant concentration as a function of
height so that our output takes the form of concentration profiles in the
air mass as functions of time-from-initial-state (or, equivalently, loca-
tion in the air basin).  To generate concentrations on a horizontal grid,
we need to compute many air mass trajectories and to interpolate concentra-
tions at prescribed time intervals.

      Returning to the summary of the present evaluation study, let us now
examine the findings of the chemical calculations.   The approach is to
formulate a functional list of reactions that describe phenomenologically
the main observable species in a laboratory smog chamber experiment.  With
an eye toward atmospheric application, we work to minimize the computing
load by collapsing some of the reaction chains into a single rate-controlling
step with overall stoichiometry specified.  Similarly, parallel reactions
involving analogous members of an organic species family are lumped into
a single composite step that involves a single class of generic reactants,
a composite rate constant, and a single class of generic products.  Our
adopted ground rules required:
      1.    Initial determination of chain stoichiometry (which there-
            after is held fixed),
      2.    Maintenance of reported rate constants within their measured
            intervals (except where reasons exist to believe otherwise),
            and
      3.    Adjustment of a minimum number of the unknown rate constants.

-------
      Following this procedure, we obtained simplified mechanisms for
fourteen chemical systems (dilute hydrocarbon/nitric oxide mixtures in
air) undergoing photooxidation in smog chamber experiments.  A single
prototype reaction mechanism involving twelve species and sixteen reac-
tion steps was used for all systems.  Different hydrocarbons were charac-
terized by different rate constants in the hydrocarbon oxidation reactions.
The ratios of rate constants followed ratios of the hydrocarbon reactivity
reported elsewhere.  This confirmed our earlier practice of scaling labor-
atory systems to uhe atmosphere according to these ratios.

      In evaluating the photochemical kinetic model, we had to compensate
for certain aspects that are peculiar to the smog chamber.  Surface-to-
volume ratios are much higher in the laboratory than in the atmosphere
and an appropriate reaction chain has to be included beginning with ozone
reacting with nitrogen dioxide and ending with nitric acid on the chamber
wall.  Another artificial feature of the smog chamber is the dilution of
the reaction sample by removal of sizable samples for analysis and replace-
ment with "clean" air.  In certain cases, this correction became so large
as to obscure the effects of some reactions; therefore, it is our recom-
mendation that all future smog chamber work utilize in situ measurements.
(The use of long-path infrared cells for many smog studies in the past
has demonstrated the feasibility of in situ measurements.)

      A striking example of the need for gas-solid reactions is found in
the analysis of dilute automobile exhaust in the smog chamber results.
Despite its small surface area compared with that of the walls, the fine
suspended particulate matter in the exhaust forced the addition of an
(N09 + particles)-reaction in order to account for nitrogen loss from the
pollutant fraction in the gas phase.  This was not necessary for the
synthetic hydrocarbon/NC
of particulate material.
synthetic hydrocarbon/NO  mixture experiments that were essentially free
                        X
      Moving from laboratory  experiments  to polluted atmospheres,  we
added several improvements and carried out  further cross-checks on the

-------
GRC photochemical diffusion model that was briefly outlined above.  In
preparation for the eventual coding of the air trajectory calculation,
we established the logic for objective determination of advection paths
of air mass-centers.  Certain intrinsic weaknesses in the data and in
the use of ground winds suggest that high degrees of refinement are unwar-
ranted.  Very poor agreement was found between wind directions and wind
speeds measured at two stations set up in the same area.  Likewise, very
poor agreement was noted between a computed ground track using station
data and the measured trajectory of a tetroon  flying with the wind at
a few hundred meters altitude.

      In the past, our vertical eddy diffusivity values were based on wind
speed, but a more detailed analysis of the data showed that vertical
temperature gradient was better than wind speed or Richardson number for
correlating measured diffusivities.  Based on these findings, we adopted
five vertical profiles of eddy diffusivity, each characterized by a range
of vertical temperature gradients.  Subsequent diffusion calculations for
nonreactive species required some downward adjustments of the diffusivity
values; however, they were still within the range of observational uncer-
tainty.  Horizontal diffusivities were obtained from radar measurements
of tetroons as reported in the literature.

      The pollution source program was updated by further automating the
generation of emission fluxes and by introducing the new numbers provided
by Systems Applications, Inc.  The"bearings and speeds for one-hour tra-
jectory segments are fed into this auxiliary program.  Given a geographical
and temporal starting point, the program generates a data deck that supplies
boundary conditions for the photochemical/diffusion model.  Most of the
numerous revisions given to us were incorporated in the pollutant emissions
calculations.

      A three-dimensional (vertical displacement, transverse displacement,
and streamwise displacement) time-dependent diffusion code was used to
—
 A neutral bouyancy balloon that floats with the air along levels approxi-
 mating constant ambient density.

-------
assess the importance of neglecting transverse horizontal diffusion in
the photochemical/diffusion model.  The 3-D code's coordinate frame fol-
lows the air, but Gaussian spread lateral to the air motion is considered,
for each time step as well as vertical diffusion using flux/gradient rela-
tionships.  Air parcels moving parallel to one another were assumed to
pass over emission fluxes differing widely from one another.  For the
worst case, errors between 27% and 39% were noted for CO-increments over
a five-hour period; however, the errors in CO-concentration after five
hours would scale down to only 10% to 20% because of the addition of
initial concentrations (usually 5 to 10 ppm) for high air pollution con-
ditions.  Another assessment of horizontal spread was made for an air
                                                           *
mass-center passing, at closest approach, one grid distance  away from
a stack emitting oxides of nitrogen.  A trajectory model would omit the
subsequent spread of the plume that, in reality, would raise ground con-
centration somewhere downwind.   Calculations showed that moderate values
of stack emission and distributed ground-based source emissions give a
maximum error of 5% due to omission of the stack plume contribution.

      The heart of our evaluation study is a series of tests of the simu-
lation model against real-world air quality measurements.  If there is any
return on~our investment in development and refinement efforts, it must
show up as successful predictions of contaminant concentrations in the
atmosphere.  Extensive groundwork in chemical and meteorological improve-
ment has been summarized above.  Our test design.will now be outlined
and the results will be summarized.

      Six days in the 1969 Los Angeles smog season (September-November)
were designated for the data base.  In addition to having two instrumental
trailers in operation measuring detailed aerometric data, on those six
 "Grid distance" refers to the cell size that specifies the spatial reso-
 lution of the emission-source inputs to the model.  At the present time,
 this distance is 2 miles.

-------
days airborne studies were conducted yielding numerous detailed temper-
ature profiles.  The profiles are needed to obtain the vertical eddy d±f-
fusivity of pollutants.   Ordinarily, only one or two soundings are avail-
able each day from instrumented balloons which telemeter information back
to a station.

      Three of the days  are designated "hands-off" days and the other
three, "hands-on" days.   The intent of the test design is to adjust para-
meters and to develop a set of operating rules for optimal model performance
based on "hands-on" data.  Then the test proceeds without any further model
manipulation for the "hands-off" days to see how well the predictions are
made.  Each test of the model involves taking measured initial contaminant
levels early in the morning (0230 to 0830 for our tests) and computing the
concentration histories through the morning and early afternoon hours to
exercise both the photochemical and meteorological parts of the model.
For "hands-on" operation, the diffusional parts of the model are examined
in  the absence of chemistry by checking against carbon monoxide measurements.
This has the combined advantage of being a nearly-inert tracer and of being
essentially all derived from widely distributed (vehicular) sources.

      Normalizing the diffusional parts of the simulation against carbon
monoxide highlights the chemical aspects in subsequent exercises of the
full model.  For example, the peaking of nitric oxide levels during the
morning commute-rush reflects first, the inability of the air to dilute
the material; and later, the effects of dispersion and transformation to
nitrogen dioxide that combine to cause a decay as they overcome the weaken-
ing emission sources as cars leave the road.  Without the advance tests
of diffusion, the interpretation of the nitric oxide tests would be ambi-
guous, at least, because of uncertainties in meteorological dispersion
superimposed on uncertainties in chemistry.

      Despite these seemingly difficult obstacles that must be overcome,
the test results were consistently more accurate than were the exercises

-------
of previous versions of the model.  The report presents the detailed data,
but the general findings are as follows:
      1.    The chemistry we used to model smog chamber experiments is
            applicable to the atmosphere if the hydrocarbon rate con-
                                           A
            stants are appropriately chosen  and the (NO,., + particulate)-
            rate is decreased.
      2.    The diffusion in the vertical is described well by using a
            set of profiles determined directly from inert gas disper-
            sion data found in the literature providing that diffusivity
            profiles for stable atmospheric conditions are uniformly
            decreased by factors of two or three.  This degree of uncer-
            tainty is not unusual for vertical diffusivities.
      3.    The source emission inputs give results consistent with the
            materials balances observed in the atmosphere with the con-
            spicuous exception of nitric oxide.  Its fluxes had to be
            reduced by 75% (as in tests of previous versions of the
            model ) in order to achieve observed loadings of NO 4- NO..
                                                       1            *•
            Carbon monoxide (and previously hydrocarbon ) require no
            adjustment.  No physical mechanisms have been specifically
            identified to explain this deficit; however, surface uptake
            of NO appears to be a strong possibility that must be in-
            vestigated in future field programs.
      4.    Curves of individual species fit the interpolated measure-
                 *A
            ments   rather well during the peaking phases; however, both
 A
  Automobile exhaust runs in the smog chamber were modeled by stratifying
  all the reactive hydrocarbons into two groups based on speed of oxida-
  tion rate.  Good atmospheric results were obtained using rates derived
  for the slower of the two groups of hydrocarbon.
A*
  Since the air parcel trajectories generally cut between monitoring
  stations rather than going right over them, we must use some inverse-
  distance-weighted averages to interpolate among the station measurements.

-------
            carbon monoxide and nitrogen dioxide are overpredicted at
            the ends of trajectories where their levels are relatively
            low.  Inadequate accounting for the mixing is likely respon-
            sible for these discrepancies.  Furthermore, in the case of
            nitrogen dioxide, heterogeneities can cause large errors
            since the chemistry in the model is treated homogeneously.
      5.    Ozone, one of the main pollutants, is relatively well pre-
            dicted in its time-phasing for net production and its level.
            This success is especially fortunate because ozone is the
            subject of much control planning activity and is a very
            sensitive indicator of validity for photochemical models.

      In an appendix to the report, we indicate two areas where future
 research is needed to build confidence in applying photochemical smog
 models.  One deals with gas-solid interactions and the other, with the
 interference of finite mixing rates with the reaction kinetics.  Gas-
 solid interactions include aerosol reactions and adsorption to urban
 surfaces.  These processes may well be responsible for the difficulty
 in  achieving a nitrogen balance in morning air samples.  Surface uptake
 of  pollutants will assume growing importance in the analysis of large-
 scale urban/rural air pollution.  The turbulence interference phenomenon
 occurs when chemistry proceeds rapidly compared with mixing.  Experimental
 evidence is analyzed to demonstrate that the magnitude of this effect
 could lead to significant modeling errors.  Both of these new problem
 areas must receive more attention in field studies before the modeling
methods can be further advanced to deal with them.

-------
1     STRUCTURE OF THE REPORT
      This report describes the rationale behind and the evaluation of
various advances in the existing GRC photochemical/diffusion model.  The
model's objectives and methodology have been discussed in detail in the
preceding Overview and Summary; therefore, we will emphasize the near-
term goals and activities without repeating the background discussions.

      Briefly, our immediate purposes are, first, to make changes that
should improve the chemistry and physics used in the GRC photochemical/
diffusion model and, second, to subject the updated model to controlled
evaluations using measured data.  A subsidiary objective is to convert
the program to an IBM system and create manuals to afford US Environmental
Protection Agency (EPA) personnel the opportunity to operate the model.
Improving the model's physical and chemical content necessitates some
refinement and some innovation.  Refinements in source inventories are
available from other recent work in the field.

      A brief preview of the remaining sections in this document is
given below.

      Section 2.  Chemical kinetic improvements are introduced in three
ways:  (1) updating input values to incorporate newly measured rate con-
stants, (2) adding or deleting reactions based on recent findings, and
(3) exercising the kinetics submodel for smog chamber conditions over a
wider span of systems and mixture ratios than that used previously.

      Section 3.  The meteorological innovations are based on better
choices of diffusion coefficients and systematized (but still manual)
wind field analysis.

      Section 4.  The introduction of chemical and meteorological improve-
ments is followed by controlled evaluations of modeling assumptions adopted
previously, i.e., 'the neglect of crosswind diffusion and the treatment of

-------
  large point  sources  in the framework of  a source model laid out on a two-
  mile grid.   Of  particular  concern  is the omission of  plumes from off-
  trajectory point  sources in inputs to the moving contrpl  volume.   These
  plumes are left out  if the mass  center of the  air parcel  never actually
  transects the grid square  containing the source.   This plume error and
  lateral diffusion error from area  source nonuniformities  are assessed
                                       *
  using the three dimensional LAPS code developed  at GRG.

        Section 5,  Model tests for  six days were  undertaken first for dif-
  fusion of CO and  subsequently for  major  species  undergoing diffusion com-
  bined with photochemistry.   Although the objective was four six-hour tra-
  jectories per day, some modifications  had  to be  incorporated in the form
  of tradeoffs between numbers and lengths  of trajectories.   (An auxiliary
  study qf transportation control  abatement  strategies  was  performed using
  three of the trajectories.   The  results  of this  task  are  reported in a
  separate volume.)
 *
 Appendix A,  See.  A.3.2.
10

-------
2     ANALYSIS OF CHEMICAL KINETI,GS IN SMQG CHAMBER EXPERIMENTS

2.i   INTRODUCTION
      The mathematical model of smog photochemistry described belgw is a
lumped-paramete.r model which is a key element of a larger model of polluted
atmospheres.  In developing and validating such a chemical model, the fol-
lowing objectives have guided our approach;
      1.     Reproduction of the essential features of smog chamber
            experiments
      2.     Capability to simulate experiments using a variety of hydro=
            carbons and hydrocarbon/NO  mixtures
                                      X
      3.     Ease of adaptation to atmospheric modeling
      4.     Retaining physical plausibility by using rate constant values
            which are in agreement with experimental data
      5.     Maintaining computational simplicity by using lumped
            parameters
      6.     Explicit consideration of surface reaction effects that may
            be peculiar to smog chambers

      In the following sections, we shall deal with the validation of the
model using smog chamber data.  First, we describe the kine.tie model itself
in some detail.  The methodology used for validation is reported next.
Then the smog chamber data and the results are described.  Finally, we
discuss criteria used in adapting the chemical model for simulating
atmospheric chemistry.

2,2   CHEMICAL MECHANISM

2.2.1  Reaction Steps
      The basic mechanism is composed of sixteen reactions.  Some of the
reactions are elementary and in gome cases a set of elementary reactions
                                                                       11

-------
 has been reduced to a single step, hence the lumped-parameter nature  of
 the model.  The reactions included in the model are shown below.

       Following the inorganic cycle
             hv + NO  -> NO + 0
                    1                implies                          (2.1)
                                     hv + N02 -> NO + 03
               + 0  + M -* 03 + M
             NO
  we have the hydrocarbon  oxidation  chain initiators
              0 + HC
              OH + HC + (b2)R02                                        ^2'4)

              0   + HC + (b)R0                                         (2.5)
 Reactions  (2.3)-(2.5) are lumped reactions which  represent  the oxidation
                                            9 — Ł\
 chains which have been postulated to occur.     HC   denotes a generic
 hydrocarbon and  RO   is an organic radical.  The b's   denote branching
 factors which account for the fact that  the oxidation chain produces  a
 multiplicity of radicals.  In reaction (2.4) we- have treated the radical
                                                         7 R
 attacking  HC  as  OH  because of its likely dominance.
 ^
  Because the reaction  0 + 02 + M ->• 0. + M  is known  to be  very  fast,
  the two reactions shown in (2.1) can be combined  into a  single  reaction
  whose net product is  NO + 0  , i.e.,  hv + N02 ->- NO + 0  .   This  is
  equivalent to assuming 0-atom quasistationarity.
12

-------
      The conversion of  NO  to  NO   occurs via  the  chain-carrying
reaction
            RO  + NO -»• NO  +  (y)OH                                   (2.6)
where  (y)  is a yield factor which represents  that  fraction  of  the  con-
version which returns  OH  to the system.  The  yield factor is less  than
one because all  R's  are not  H  .

      Note that reactions (2.4) and (2.6)  form  essentially a  closed  loop
early in the reaction and stability requirements  call for  by < 1 + a  ,
where  a  is a positive function  of other  rate  constants and  concentra-
tions which is small compared to  1   at early times  and gradually increases
throughout the reaction.  Imposition  of this constraint prevents  RO
runaway during the early part of  the  reaction.

      Chain-termination steps consist of the lumped  reaction

            R02 + N02 •* PAN*                                         (2.7)

and the elementary reactions

            OH + NO -> HONO                                           (2.8)
            OH + NO  -* HNO                                           (2.9)
      The photodissociation of  HONO  has been  suggested  as  a possible
                    7 8
source of OH-radical  '
            hv + HONO -> OH + NO                                     (2.10)
*
 PAN denotes peroxyacetylnitrate.
                                                                       13

-------
  Formation of  HONO  is assisted by the reaction

              NO + NO, + H?0 '-> 2HONO                                  (2.11)

  which is likely to proceed in two steps

              NO + NO  -> N203

              NO  + HO -> 2HONO

                                          4
  as suggested by Altshuller and Bufalini.   The  N^  reacts with   1^0
  so rapidly that the two reactions 'can be lufllped into reaction  (2.11).
  The late-time behavior of  0   and  NO   is best reproduced when the  fol-
  lowing reaction is included

              N02 + 03 + N03 + 02                                     (2.12)

        Nitrogen imbalances in smog chambers have prompted various investi-
  gators to attempt to track down the fate of the nitrogen compounds.   Gay
              Q
  and Bufalini  report that a large fraction of the nitrogen loss can be
  accounted for by nitrate formation on the walls of the chamber.  Follow-"
  ing their suggestions  as well as those of Dodge,   the reactions shown
  below  have been included in the mechanism.

                •3     9     O  fc                                         '(2.; 13)
                              2HN03                                  (2.15)

 It should be noted that up  to  how :no  :HNO   has been observed in the gas
                        9
 phase in smog chambers.   this prompted us t-6 compare fch'e -relative effi-
 ciency of reactions (2.9) and  (2.11)-.   In oaf simulations,  tfee result was
14

-------
that for 50% relative humidity  reaction  (2.15) is about six times faster
than reaction  (2i9), so oo.e would expect to find significantly more  HNO
ori the walls of the chamber than in the  gas phase.  As a final commentj
we note that the significance of reactions  (2.13)-(2.15) for atmospheric
modeling is not clear at this time,
      Aerosol was observed in the chamber experiments with dilute auto
     st.  The disappearance of  NO   from the system differed signifies
from that observed in the other experiments in which no aerosol was pre-
sent.  This promp
tion of the form
exhaust.  The disappearance of  NO   from the system differed significantly
                                    triments
sent.  This prompted the suggestion by Dodge   that a first-order reac-
            NO  + particulates -> products                           (2.16)
be added to the system  to account for the observed effects.  Inclusion of
this reaction has improved the simulation of  the auto-exhaust experiments.
Also, this reaction may be helpful in atmospheric modeling since aerosol
is observed in the real world.

      Previous versions of our kinetic model  also contained the reaction
            RO  + NO  -> PAN                                         (2.17)

                                         12
This reaction has been suggested by Hanst   as likely to be important in
the formation of PAN.  Such likelihood was supported'by our earlier model-
                                                                      •4),
                                                                      7,14,15
         13
ing work.    However, the increase of rate constants for reactions (2.4),
 (2.7) and  (2.8) which was required by recently available measurements
 caused reaction (2.17) to become unimportant in our model and  thus it has
been dropped.

      Table 2.1 shows a list of all the reactions in our current kinetic
model along with rate constants for which measurements exist.  The rate
 constants obtained by repeated trials from the simulations  are reported
 in the section on results of the adjustments and tests.
                                                                       15

-------
                                 TABLE 2.1

                           BASIC KINETIC MODEL
              Reaction
1.

la.
2.
3.
4.
5.
6.
7-
8.
9.

10.


11.
12.
13.
14.
15.
16.
hv + N00 -> NO + 0
i
o + q2 + M -> o3 + M
NO + 03 + N02 + 02
0 + HC -> (b1)R02
OH + HC ^ (b2)R02
0 + HC -> (b )RO
RO + NO -> NO + (y)OH
RO + NO ->• PAN
OH + NO+ HONO
OH + N00-> HNO_
2 3
hv + HONO ^ OH + NO
H.O
7
NO + NO + 2HONO
N°2 + °3 "" N°3 + °2
N03 + N02 -> N205
N205 -> N03 + N02
N2°5 + H2° "*" 2HN°3
NO + particulates ->- products
    Experimental Rate
    Constant Values^

      2.67(-l)min~1

               -2   -1
    1.32(-5)ppm  min

2..2(+l) to- 4.4(+l)ppm  min
                                           1.5(+3)ppm  min

                                           3.0(+3)ppm~ min~
                                      5(-2) to 1.25(-l)ppro 1min~1

                                           4.5(+3)ppm  min
                                                           7

                                          2.5(-3)ppm  min
                                                                     16

                                                                     17
                               14

                               15
                               18

                               19

                               18

                               20
  t
   The number in parentheses denotes the power of ten by which the coef-
   ficient must be multiplied, e.g.,  2.67(-l) = 2.67 x 10"1 .
  k
   Experimental values for these rate constants are often known for parti-
   cular hydrocarbons and will be reported in Sec. 2.5.
   For the validation process,  k^5  was converted to a pseudo-first-order
   rate constant by lumping water vapor content of air at 50% relative
   humidity into  k]^  since the smog chamber experiments were conducted
   at this level of humidity.   The resulting rate constant is 60.5  min"1.
16

-------
2.2.2  Quasistationarity Assumptions
      Several of the species included in our kinetic model can be assumed
to be in a quasistationary state with respect to the other species.
Apart from the computational advantages of this assumption, quasistation-
arity can be justified on physical grounds by examining the relative
rates of the various reactions involved.  Such a check has revealed that
it is likely that 0-atom,  RC>2 , OH , NO  , and  NO   are in a quasi-
stationary state.  0-atom quasistationarity can be justified on the basis
that the removal of 0-atom by the reaction  0 + 0  + M -> .0  + M  is known
to be very fast.  We have tested quasistationarity assumptions for  RO  ,
OH , NO  , and  NO   by solving parallel cases with and without station-
arity.  The results of the tests showed that assuming stationarity has a
negligible effect on the computed concentration of all the species.  A
similar test for  HONO  yielded negative results, thus  HONO  has been
retained as an active species.

2.3   METHODOLOGY OF ADJUSTMENTS AND EVALUATIONS
      Two kinds of adjustable parameters are available to us:  rate con-
stants and branching factors.  However, the number of free parameters is
limited by the fact that several elementary rate constants have been mea-
sured.  Our approach is to keep the values of the measured rate constants
within the range of experimental uncertainties.  The unknown values of
nonelementary, i.e., lumped, rate constants are then estimated from com-
parisons with analogous reactions, if they exist., and during the simula-
tion process itself.  Indeed, the object of the simulation is to deter-
mine the values of these unknown constants.  In cases where rate-control-
ling processes can be identified, the rate constants are sometimes avail-
able.  These are also confined to ranges of measurement wherever possible.
      The branching factors of the model are determined by the  NO  and
      ;cay prior to the  NO^  peak &
branching factor is estimated from
HC  decay prior to the  N07  peak and before the ozone buildup.  The
                                                                       17

-------
                   d[NQ]/dt                                           fy
              b2  ~  d[HC]/dt                                           (2

  In Eq.  (2.18),  it is  implied  that most  of  the   HE  decay is due tb the
  reaction  OH +  HC ->•  (b  )RO   .  This  is  justified by virtue df the fact
  that  reaction  (2.3) is  slow  compared to (2.4)  and that early in the reac-
  tion  the ozone  is essentially  zero and  thus  (2.5)  plays no role in  HC
  decay.   Since   NO -> NO   conversion  occurs via reaction (2.6),  the rela-
  tion  shown in  (2.18)  can be used because   R02   is in stationary state.

        Thus,  using the available data, we obtained linear least-squares
  fits  of the NO  and  HC  to estimate their  respective decay rates.   Then
  from  Eq.  (2.18),  we estimated  b   prior to  any adjustment of rate con-
  stants.   We then  chose  a value of  b^   from  among the various values
  obtained for each experiment set.  Subsequently,  we set  b..  = \>  .   The
  value of  b  was obtained by modeling  the late-time behavior of the  sys-
  tem.   Once set, the values of  the branching  factors remain constant
  throughout the  simulation.

  2.4    EXPERIMENTAL DATA BASE
        Experiments on  different mixtures of the following four systems  in
  air were  used in  the validation process:
       1.    Propylene/NO
                         X.
       2.    Toluene/n-Butane/NO
                                X
       3.    Toluene/NO
                       x
       4.    Dilute auto exhaust/NO
                                   x
 These  four groups  comprised a total of fourteen experiments which were
 used in testing  the kinetic model.   The number of experiments in each
 group  was four,  three, five,  and two, respectively.   Table 2.2  shows a
 detailed breakdown of  the  experimental data used in  our  simulations.  The
 experiment numbers shown on  the table are the same ones  used by EPA in
 their  laboratory procedures.
18

-------
                               TABLE 2.2

                           EXPERIMENTAL DATA
                          Initial Concentrations,
                                    ppm
                                                   Average Fractional
                                                     Dilution Rate
Group

Propylene

Toluene/
n-Butane


Toluene


Auto
Exhaust
Exp . No .
321
325
329
336
251
253
257
250
258
263
271
305
**
CUE 222 ,
***
CE 231
NO
1.23
.30
.29
1.14
1.11
.53
.27
1.17
.35
.54
.32
1.26
1.94
2.73
NO 2
.09
.04
.01
.04
.11
.11
.07
.08
.04
.06
.04
.06
.10
.23
HC
.275
.45
.24
.61
1.60/3.02
1.43/3.42
1.29/2.97
1.53
2.88
1.71
1.20
3,14
1.06/1.151"
.39/.201"
-4 -1 *
(x 10 min )
7.52
7.33
7,22
7.03
6.17
6.22
4.90
5.78
4.88
5.80
6.05
5.76
5.65
5.98
 **
ft**
  t
To obtain average volumetric dilution rate multiply by the chamber
volume (335 ft^) (see Sec. 2.6.1 for a detailed discussion of dilution
effects in smog chambers).
Exhaust from a vehicle without exhaust hydrocarbon and CO emission
controls.
Exhaust from a vehicle with exhaust hydrocarbon and CO emission controls,

The multiple hydrocarbon mixture has been aggregated into two classes:
high and low reactivity hydrocarbon.  The fraction shown here gives
the ppm of each class in the order high-reactivity/low-reactivity.
                                                                      19

-------
  2.5   RESULTS OF SMOG CHAMBER VALIDATIONS
        In this section, we discuss the results of the validation tests
  which were performed on our chemical model.   The goal of these validation
  tests was to simulate smog chamber experiments performed in EPA labora-
  tories.  The experimental data used for comparison with the model's
  results has been described in Sec.  2.4.  The outcome of the tests is pre-
  sented by means of graphs that are representative of the group of tests
  performed.  In addition, a table of the rate constants used in the vali-
  dation tests is presented for several experiments.  Finally, the branch-
  ing factors for the various simulations are also given.

  2.5.1  Propylene Experiments
        The simulation of the propylene experiments was carried out by-
  varying a single rate constant within the framework of a basic set of
  parameters.  Thus, after determining the branching factors and setting
  values for the other rate constants, it was  sufficient to adjust  k   (the
  OH attack on HC) to reproduce the experimental results obtained for the
  various propylene/NO  mixtures.   Two of the  experiments (nos.  321 and 329)
                      X4    -1   -1
  required  k,  = 6 x 10  ppm  min   .   Experiments number 32i> and 336
                       4                  4
  required  k  = 2 x 10   and  k,  = 3 x 10  ,  respectively.   Table 2.3
  shows the rate constants used in the four cases.  The branching factors
  were determined to be  b^ = b2 = 4   and  b  = 1 .   The yield factor for
  OH  in reaction (2.6)  is set to  0.25.   It should be noted that Stedman7
  measured  the  rate  constant  of the hydroxyl-propylene reaction  and
  obtained   k  =  2.5  x  10   ppm  min   .   (If this experimental value of  k,
             H                                                            4
  were  used  for all  the  simulations,  the  reaction would be slowed down in
  experiments 321, 329,  and 336, and  it would  be speeded up  in 325.)   Thus,
  our model values for   k4  agree well with the measured quantity,  which
  is a hundred  times greater  than a previous estimate of Westberg^  of
 k4 = 244 ppm  min   .  In our  previous  modeling work,  we had lowered
 this estimate to  80 ppm  min    .  To preserve the good agreement between
 data and simulation, we had  to raise kg   from  1500  pprn'^in"1   to   105 ,
 and  k?  from 6 to 600.   (It will be recalled that k    was  increased
20

-------
                                 TABLE 2.3

             RATE  CONSTANTS  USED IN PROPYLENE SIMULATIONS
Reaction Number


        1


        la


        2


        3
        4  (experiment  321 and 329)


        4:(experiment  325)


        4  (experiment  336)


        5


        6


        7


        8


        9


       10


       11


       12


       13


       14


       15


       16
  Rate Constant


  2.67(-l)min~1


1.32(-5)ppm~2min~1


    2.67(4-1)


    6.1(+3)**


    6.0(4-4)


    2.0(4-4)


    3.0(4-4)


   9.27(-3)**


    1.0(4-5)


    6.0(4-2)
 3.0(4-3)


1.0(-3)min





 5.0(-3)


 4.5(4-3)
             -1
   6.05(+l)min


     0 min'1
  Units in  ppm~  min    unless otherwise specified.
 *
  Measured rate constant from Ref. 2.
                                                                        21

-------
  from 10  to  1500  and  k    from  30  to  3000  in  order to  be consistent with
                      1 /  "1 ^
  experimental values.   '   )   The success of this  procedure is especially
  significant, since it shows  that  a workable  set  of rate constants is far
  from unique.   However,  this  is not surprising  in view of the highly non-
  linear character of  the  system.
                                                                   _3
        Referring  to Tables  2.1  and 2.3, we note that  k^ = 5 * 10    in
  Table 2.3 and  this is a  factor of 10  lower than  the lower bound of the
  experimental value shown on  Table 2.1.  This had to be done in order not
  to impair greatly the late-time behavior  of  N02  and  03 .  The same
  effect was  encountered  in modeling the other experiments with various
  types of hydrocarbons and HC/NO   mixtures.   In the  present model, reac-
  tion (2.12),   NO +  0  -> NO  + 0  ,  is rate-controlling late in the reac-
  tion.  Furthermore,  the  computed  concentrations  of N0_  and  0   are very
  sensitive to small changes in  k   .  Thus this  parameter is very impor-
  tant and the availability of a very  accurate value would be a boon for
  the modeler.

        Figures  2.1 and 2.2  illustrate  results obtained in the simulation
  of experiment  329.   Note that  the data are closely approximated by the
  model.   The ozone plot shown in Fig.  2.2  exhibits lower concentrations
  than are indicated by the  data late in the reaction.   This discrepancy
  is  due to dilution effects in the model and cannot be resolved by plausible
  adjustments in rate  constants.  Omission  of dilution  terms from the model
  results  in  ozone  concentrations which provide  a  better fit to the data.
  However, omitting dilution terms  also results  in markedly poorer fits
  for  the hydrocarbon and  NO  .   Thus it seems  that the inclusion of dilu-
  tion both benefits and hinders the results.  One possible explanation  of
  this apparent paradox is that the continuous-dilution approximation- used
 in the model is not accurate enough to simulate  those cases where the
 dilution rate becomes large  compared with the  chemical rate.   (See Sec.
 2.6.1 for a detailed  discussion of dilution effects in smog chambers.)
22

-------
n.oo
                                                                                  !n  'NO
                                                                                 •O  HC
                                                                                 A  NO-
                                                                     EXPERIMENTAL
                                                                     DATA
                 HC

                 NO
          4.00
                    e.oo
                              P. 00
                                       16.00      20.00      2H.OO
                                             TIME :CX10  M
 i          t~
28-00      38.00
              Figure  2.1.  Experiment 329,  Propylene/NO    Plot of Propylene,

                            NO,  N00

-------
  s
a
a:
o

§
c_>
 8
                  MODEL RESULTS
D OZONE


O PAN
                  EXPERIMENTAL  j  + OZONE

                  DATA        I  X PAN
  0.00      H.OO      8.00      12.00
                                        16.00      20.00      Ł4.00

                                             TIME  (X10  ']
                                      28.00     38.00      36.00
        Figure  2.2.   Experiment  329,  Propyjlene/N0x>   Plot of Ozone and PAN
  24

-------
2.5.2  Toluene/n-Butane Data
      The simulations of this binary hydrocarbon mixture were very success-
ful.  Most remarkable is the fact that we were able to reproduce accurately
all three experiments with a single set of parameters.  This contrasts
with the propylene tests where  k,  had to be. varied.  It should be noted
that two sets of hydrocarbon reactions were used in these simulations.
      Some remarks are in order about the rate constants for reactions
(2.3)-(2.5).  Published rate constant data were found for both toluene
                                     9 fi
and n-butane only for reaction  (2.3). '   For reaction  (2.4), a measured
                              f\ O T
rate constant could be located  '   only for n-butane.   The rate constant
for toluene in reaction (2.4) was estimated from the literature value of
k,  for propylene using the relative reactivity of the  two hydrocarbons.
This initial estimate of  k.  was then adjusted until agreement between
computed results and chamber data was obtained.

      No literature data for toluene or n-butane ozonolysis was found.
Nevertheless, the rate constants for the ozonolysis of  these hydrocarbons
are expected to be low, as indicated by the low values  given"'fbr aromatics
such as xylene.  For example, Niki,  in a scale of relative rate constants
(with propylene as unity), gives xylene ozonolysis an upper bound of 0.2
compared with relative rates ranging from 2 to 62 for the olefins.  Reac-
tivity scales are not useful for obtaining estimates for  k   since hydro-
carbon reactivity and  k   do not correlate well for other than olefinic
                                                * 6
compounds, as has been pointed out by Niki, et al.   For purposes of the
simulation, we chose some initial low value for the ozonolysis rate con-
stant and subsequently adjusted it to obtain good agreement with chamber
                                                         -4    -1   -1
measurements.  The resulting values remained low,  3 *  10   ppm  min
for toluene and  10   ppm  min    for n-butane.  These  values should be
compared with  9.27 x 10~  ppm" min    which is used for propylene.
      Based on relative reactivity considerations, the  OH + HC  rate
constants for toluene and n-butane were maintained at a 2:1 ratio with a
                                                                       25

-------
                                     4
 rate constant for toluene of  2 x 10   .  Although  this  rate  constant is
 higher than the one used in the pure toluene experiments,  this  result is
 not surprising inasmuch as we have synergistic effects  to  account  for.

       Figures 2.3-2.5 illustrate a simulation of experiment  253.   We can
 see that  NO  , both hydrocarbons, and  0   are reproduced very well
             X                            ~*
 indeed.  In contrast with the pure toluene runs (see  Fig.  2.8), the ozone
 buildup has both the correct time phasing and magnitude.   Similar  results
 were obtained for the other two experiments in the set.  However,  as
 shown in Fig. 2.6, the  NO   buildup occurs late for  experiment 251, al-
 though the shape of the curve is similar to that shown  in  the data.

       Table 2.4 shows a list of the parameters used for these cases.  The
 branching factors used are  b  = b  = 2  and  b  = 1  .  The  yield  factor
 for  OH  in reaction (2.6) was set to 0.5.

 2.5.3  Toluene Experiments
       The results of the simulations of the toluene experiments were
 fair.  The main problem encountered was that too much ozone  was produced
 in the simulation.  Moreover, it seems that no plausible adjustments qf
 the rate constants remedied the situation.  The reproduction of toluene
 and  NO  histories was generally good, but the  NO    tended  to  linger
 late in the reaction.   Figures 2.7 and 2.8 illustrate one  such  simulation,
 in this case for experiment 271.  Note that although  the   NO decay  is
 well modeled, the simulated  0   buildup is about 40 minutes too late.
 Since the interaction between  NO  and  0   is so strong,  efforts  aimed
 at speeding up the  0    buildup are bound to impair the modeling of   NO
 and vice  versa.   Finally,  typical  OH + toluene  rate constants used in
 these  simulations  were   10  ppm  min~  .   This compares with the range
       44                                      A
 2  x 10    to   6 x  10   used for propylene and with  2 x l(T   used in  the
 toluene/n-butane  case.
26

-------
                                                                   MODEL   JO  NO
                                                                   RESULTS  |A  :N02
                                                                EXPERIMENTAL I +   N0
                                                                DATA        ] *   NO,
:H-'QQ       fl-OQ      12 -ID       16-OT      ;2Q,'QO      2H-00      '28.^00      3g.:QQ
                                      TIME aro  !1i
   Figmre 2.3.   Experlmemt 253, Toluene/.n-Butane/NO .   Plot  of NO  and HO,
36 .,00

-------
NO
00
                                                 EXPERIMENTAL I + OZONE
                                                 DATA       I v PAN
                     g
                                                             T	T
                                                            16.00      KJ.OO     2H.QO
                                                                  TIME (X10  M
                                                                  28.QO
32-00
36.00
0.00       M.OD      8.00      12-00


      Figure 2.4.   Experiment  253, Toluene/n-Butane/NO .  Plot  of Ozone and  PAN

-------
                                                                      MODEL  In  TOLUENE

                                                                      RESULTS  Q  n-BUTANE
                                                                 EXPERIMENTAL  I *  'TOLUENE

                                                                 DATA         X  n-BUTANE
Q.DO
          M.QO
                    8.00
12.00
           1	T
16-00      20-00    , 2M.OO
     TIME (X1Q  ')
                                                                    28-00
                                                                              32.00
                                                           36.00
          Figure  2.5.   Experiment  253, Toluene/n-Butane/NO  .   Plot of Toluene

                         and n-Butane

-------
to
O
            0.00
                                                                                                  MODEL   I n   NO
                                                                                                  RESULTS j A   N02

                                                                                             EXPERIMENTAL! x   N0?
                                                                                             DATA
                                                                                  \ +   NO
H.QEt       §'.00      Ig'-OO      16.00      20-00     ,  2M-00       28.00'      ^
                                      TIME  TXIQ  M

    Figure 2.6.  Experiment 251,  Toluene/n,-Butane/NO .  Plot  of NO and  NO
                                                        x
                                                                                                              36.00

-------
Tolu
    .ene
n-Butane
                               TABLE 2.4
            RATE CONSTANTS FOR TOLUENE/n-BUTANE  SIMULATION
Reaction Number
       1
       la
       2
       3
       4
       5
       3
       4
       5
       6
       7
       8
       9
      10
      11
      12
      13
      14
      15
      16
Rate Constant Used in Validation
          2.67(-l)min~1
       1.32(-5)ppm  min
            2.67(+l)
           1.69(+2)**
            2.0(+4)
            3.0(-4)
            1.0(+4)
                                                   2.0(+2)
                                                 1.0(-3)min


                                                  5.0(-3)
                                                  4.5(+3)
                                                          -1
                                                 6.05(+1)min"
                                                    0 min
  Units are  ppm  min    unless otherwise specified.
  Measured values obtained from Refs. 2 and 6,  respectively.
                                                                       31

-------
to
NJ
                                                                                                     (D
                                                                                                     }  n
            NO

MODEL   I o  HC
RESULTS |
       ( A  N02
                                                                                          EXPERIMENTAL) "   ""2
                                                                                          DATA
        X   NO,
              i
            HI

            NO
0.00       M.OO       8.00      12.00      16.00       80.00      2M.OO
                                                TIME  (X10   ')
                                                                                      28.00
  38.00
36.00
                            Figure 2.7.  Experiment 271, Tolue.ne/N0  .  Plot of Toluene,  NO,  and N0

-------
                    EXPERIMENTAL  +  OZONE
                    DATA
Q. 00
         4.00
                                     16.00     GO.00      2M.OO
                                          TIME  (X10  >J
SB. 00
         38.00
                   36.00
          Figure 2.8.   Experiment 271,  Toluene/NO  .  Plot of  Ozone and PAN
                                                      X

-------
       The rate constant data used to simulate experiment 271 are  given
 in Table 2.5.   No additional data are shown for the other toluene experi-
 ments because the results obtained do not warrant it;  See Sec. 2.5.2  for
 a discussion of the rate constant data for toluene reactions.  The branch-
 ing factors used are-  b^ = b2 = 2  and  b^ = 1 .  The yield  factor for
 OH  is 0.5.

 2.5.4  Auto Exhaust Validations
       The experiments  with dilute auto exhaust introduced the additional
 complication of having to deal with a multiplicity of hydrocarbons.  Also>
 particulate matter was observed in these experiments, but was not observed
 in the previous experiments.  The auto exhaust data also exhibited
 NO ->- NO   conversion with relatively small amounts of hydrocarbon having
 reacted.  To account for  NO   disappearance in the presence of aerosol
                 11          X
 it was suggested   that reaction (2.16) be added to the model, the rate
 constant to be determined by adjustments during the simulation.   Because
 of the multiple hydrocarbon mixture, the branching factors were increased
 to reflect the increased length of the hydrocarbon oxidation chains with
 its concomitant increase in organic radicals.  Finally, since it  is not
 feasible to consider each hydrocarbon individually, the mixture was
 aggregated into three types:  nonreactive and low- and high-reactivity
 hydrocarbons.  Hydrocarbons considered nonreactive were ignored in the
 simulation.  Thus the modeling runs were conducted using the two  reactive
 hydrocarbon groups.  The initial hydrocarbon concentrations were  obtained
 from an analysis by Dodge   and are reproduced in Table 2.6.

       The initial rate constants for each group were obtained by  mole-
 weighted averages of the rate constants for typical members of each grbupi
 Thus for group I, we used n-butane and toluene to represent  the paraffins
 and aromatics, respectively.  For group II, ethylene and propylene were
 used.   Again,  these rate constants were adjusted during the simulation.
34

-------
                               TABLE  2.5

               RATE CONSTANTS FOR TOLUENE EXPERIMENT  271


                                                                       -•-
Reaction Number                        Rate  Constant Used  in  Validation
       1                                         2.67(-l)min  l


       la                                     1.32(-5)ppm~2min~1


       2                                           2.67(+l)

                                                          **
       3                                          1.69(+2)




       5                                           5.0(-4)




       7                                           2.0(+2)


       8                                           1.5(+3)


       9                                           3.0(+3)


      10


      11                                           5.0(-2)


      12                                           5.0(-3)


      13                                           4.5(+3)


      14


      15                                         6
      16                                            0  min
 *              — 1   — 1
  Units are  ppm  min    unless  otherwise  specified.
 %
  Measured constant obtained  from  Ref.  2.
                                                                      35

-------
                                 TABLE 2.6
    INITIAL HYDROCARBON CONCENTRATIONS (PPM) FOR AUTO EXHAUST EXPERIMENTS
Experiment
Number
Group
I. Low-Reactivity
II. High-Reactivity

Hydrocarbon Type
C. + paraffins
Aromatics (excluding benzene)
Ethylene
Olefins (excluding ethylene)
222
0.68
0.47
0.48
0.58
231
0.11
0,09
0.22
0.17
        It should be noted that rather large amounts of  CO  were present
  in the mixtures described above.   The data provided for these experiments
  show that for Exp. 222,  the concentration of  CO  was 53 ppm, and for
  Exp. 231, it was 12 ppm.   Dodge   has pointed out that  CO  may be partly
  responsible for the oxidation of   NO  and that this would explain, at
  least in part,  why the  NO ->• N0_   conversion occurs so rapidly even
  though very little hydrocarbon has reacted.   The  CO  effect could come
  about via the following  steps:
              OH + CO -> CO   + H
                                                                     (2.19)
              H + 0  + M -> HO  + M
             HO  + NO + OH + NO
             H0? + NO  -*• HNO  + o.
(2.20)

(2.21)

(2.22)
 Reaction  (2.21) is analogous to reaction (2.6)  and would be the step
 responsible for part of the  NO -> N02   conversion.  Our model does not
 include reactions (2.19)-(2.22), of  course,  and so if  CO  reactions are
 indeed significant, we will have to  compensate  for them by other means
36

-------
such as increasing branching factors and  k   .  We note, however, that
previous work with  CO   (Refs. 4, 5, 8, 22-24), appears to indicate that
CO  concentrations of the order of 100 ppm are required before  CO  can
be considered to play a  significant role in the oxidation of  NO  .  Thus
CO  is probably not important in experiment 231, but may be a significant
factor in experiment 222.

      Figure 2.9 shows the results for  NO  and  NO   obtained for the
controlled exhaust case, experiment 231.  No  ozone results are shown
because this experiment  produced very small amounts of ozone and  the
simulation behaved accordingly.  This is due  to the relatively large con-
centrations of  NO  that exist throughout the experiment.  As the graph
shows, the  NO  is well  modeled but the  NO   buildup in the model is
not as fast as the data  would indicate.  It is puzzling, however, that
the data appear to show  a rapid buildup of  NO.  even though  NO  decays
very little in the interval 0-80 min.  Nevertheless, we note that the
NO   achieves its correct magnitude late in the reaction.  The effect of
reaction (2.16) is to take  NO   out of the system to simulate the  NO
                              Z.                                       X
decay.  The  NO   plot shown in Fig. 2.9 illustrates that the  NO   removs
is well modeled.  Note that the correspondence between computed   NO   and
the (smoothed)  NO   data provided by EPA   (denoted by the asterisks)
                  X
is very close.  It should be noted, however,  that in this simulation 20%
of the difference between initial and final   NO   concentration is
                                               x,
accounted for by dilution.
      Figure 2.10 shows  a plot  of  the  computed  reactive hydrocarbon for
experiment 231.  The  intent here is not  to  reproduce  the hydrocarbon
data, but rather to show that relatively little hydrocarbon has reacted.
The experimental data show a final concentration of 0.3 ppm and the model
yields 0.26 ppm at 360 minutes.  As was  the case with NO   , dilution
accounts for 20% of the  concentration  change from  initial  to final value.
Thus the main characteristics of this  experiment (N0x disappearance,
hardly any ozone production, and a slow  reactive-hydrocarbon decay) are
reproduced well by the model.
                                                                      37

-------
to
CO
Q.QO
                                                                                          MODEL
                                                                                          RESULTS
                                                                                   D   NO
                                                                                   A   N0?
                                                                                   O   N0u
                                                                                      EXPERIMENTAL .)  x
                                                                                      DATA        I
                                                                                       NO
                                                                                       N02
                                                                                       NO,
M.QO
8-00
                              12-00
                                                         1 - T
                                                        16.00      20-00      2M-00
                                                              TIME  1X10  ')
88.00
36.00
                            Figure  2.9.  Experiment 231,  Dilute Auto  Exhaust  (Controlled Vehicle).
                                          Plot  of NO and N0n

-------
          0.6
       Ł
       IX
          0.4
          0.2
               FINAL EXPERIMENTAL
               HC CONCENTRATION
                                        I
                       80
160         240
   TIME, min
                                                           \.
                                                       320
Figure 2.10.  Experiment  231, Dilute Auto  Exhaust  (Vehicle  with Emission
              Controls).  Simulation of  Reactive Hydrocarbon Decay
       Table 2.7 contains the rate constants used in the simulation.  Note
 that  the ratio  k,  /k.   is 2.5, a plausible value in view of the relative
                  4a  4
 reactivity of the components of each hydrocarbon group.  Note also that
 the value of  k,   is the same as that used for n-butane in the binary
                4                            ,
 mixture experiments, whereas  k.  = 2.5 x 10   lies in the range used for
 propylene.  It is also  interesting to note that the value of  k    is only
 three times greater than the dilution "rate constant" of  5.98 x 10
 Finally, the branching  factors were increased considerably from previous
 values, with  b  = b  = 8  and  b  = 1  for both hydrocarbon groups.
 This  was to be expected in view of the increased length of the chains.
 The   OH  yield factor was set to  y = 1/8 .
       Figures 2.11 and 2.12 show the results obtained for experiment 222.
 It  can be seen that the time phasing of the  N02  peak is off, the peak
 occurring about 50 minutes late.  This of course affects the  concentra-
 tion-time curve for  NO   and for this reason, it has not been plotted.
                        x
                                                                       39

-------
                                 TABLE 2.7


            RATE CONSTANTS USED IN SIMULATION OF EXPERIMENT 231

            DILUTE EXHAUST FROM A VEHICLE WITH EMISSION CONTROL
                    Reaction Number
Rate Constant




Low-Reactivity j
Hydrocarbon


High- Reactivity
Hydrocarbon












1
la
2
3
4
5
3a
4a
5a
6
7
8
9
10
11
12
13
14
15
16
2.67(-l)min l
1.32(-5)ppm~2min~1
2
8
1
2
2
2
7
1
2
.67(41)
.39(41)
^0(44)
•8(-4)
.82(43)
.5(44)
•56(-3)
.0(45)
.0(42)
1.5(43)
3.0(43)
1.0
1
C-3)min~1
0(-3)
5.0(-3)
4.5(43)
1.4(
^Dmin'1
6, 05 (41) rain'1
2.0(-3)min~1
 *              -1   -1
  Units are  ppm  min    unless otherwise specified.
40

-------
                                                             MODEL RESULTS
                                                              EXPERIMENTAL
                                                              DATA
          Q  NO
          A  N02

          +  NO

          X  NO,
                                              1	T
o.oo      M.OO      s.QO      12.00     le.oo     eo.oo     SM.QO
                                         TIME  1X10  ')
S8.QO
          f-B-
3S.OO     36.00
Figure  2.11.   Experiment  222, Dilute Auto Exhaust  (Uncontrolled Vehicle).
                 Plot  of NO  and NO-
                                                                                  41

-------
  o
  cu
cc
a:

-------
However, the  NO   levels occurring late in the experiment agree well with
the measured values.  Furthermore, the  NO   peak has the correct magni-
tude.  Equally important is the reproduction of the ozone data.  Figure
2.12 shows that the time phasing and the magnitude approximate well the
experimental data.  Although it is not shown, the final concentration of
reactive hydrocarbon is 0.8.6 ppm compared with 0.7 ppm for the data.

      Table 2.8 contains all the rate constants used for simulating experi-
ment 222.  Note that  k  /k. = 2  and that  k.. .,  has the same value used
                       4a  4                 16
in experiment 231.  The values of the branching factors are  b.. = b  = 5
and  b- = b  = 10  for low- and high-reactivity hydrocarbons, respectively.
The branching factor  b   is i
yield factor is equal to 0.1.
The branching factor  b   is unity for both hydrocarbon groups.  The  OH
2.6   ADAPTATION OF CHEMICAL MODEL TO ATMOSPHERIC MODELING
      The parameters of the chemical system obtained from the smog chamber
simulation tests must be modified when we move from the smog chamber to
the atmosphere.  What one hopes to obtain from smog chamber experiments
is a qualitative agreement between laboratory and atmospheric observables.
Modeling the smog chamber experiments then gives us an indication that
the physical mechanism proposed for modeling these observables contains
the most important features of the highly complex phenomena which take
place in reality.  The evaluation process using smog chamber data also
gives us an understanding of the model's sensitivity to various parameters.

      Having obtained this information about the model using laboratory
data, the quantitative link with atmospheric observables must come from
attempts to mo-del the atmospheric processes themselves.  In order to do
this, we must identify those features of the chemical model which are most
likely to be heavily influenced by smog chamber conditions.  We must also
find out the degree of correlation which exists between the chamber and
atmospheric mixtures.
                                                                      43

-------
                                TABLE 2.8


             RATE CONSTANTS USED FOR SIMULATING EXPERIMENT 222

          DILUTE EXHAUST FROM A VEHICLE WITHOUT EMISSION CONTROL
                   Reaction Number
Rate Constant




Low-Reactivity
Hydrocarbon


High-Reactivity
Hydrocarbon












1
la
2
3
4
5
3a
4a
5a
6
7
8
9
10
11
12
13
14
15
16
2.67(-l)min
-1
1.32(-5)ppm~2min~1
2.67(4-1)
7.76(4-1)
1.5(4-4)
2.6C-4)
3.45(4-3)
3.0(4-4)
8.54C-3)
1.0(4-5)
2.0(4-2)
1.5(4-3)
3.0(4-3)
1.0(-3)min
1.0(-3)
5.0(-3)
4.5(4-3)











-1



1. 4(4-1) min"1
6. 05(4-1) min"1
2.0(-3)min~1
 *              -1   -1
  Units are  ppm  min    unless  otherwise specified.
44

-------
2.6.1  Wall Effects and Dilution Effects in Smog Chambers
      Two factors characteristic of smog chambers which are most likely
to influence the parameters as well as the nature of the chemical model
are dilution effects due to sampling in the smog chamber and wall effects.
Omitting dilution from the model has the effect of requiring unrealisti-
cally high values for some of the adjustable rate constants.  This is
especially significant with k.-adjustments needed to obtain satisfactory
simulations.

      Wall effects influence the nature of the model.  In our particular
case, three reactions, (2.13)-(2.15), have been added to try to account
for the nitrogen imbalance which is presumably due to  NO   reacting on
the chamber walls.  Whether these reactions will play any role under
atmospheric conditions is not known, but intuitively one would expect
them not to be significant.  Hopefully, the evaluation process under
atmospheric conditions will yield an answer to this question.

      The concerns we expressed earlier regarding the importance of wall
reactions as an  NO   sink were confirmed by comparison of absolute reac-
                   X.
tion rates throughout the simulation.  The reaction of  NO   with water
on the wall to form nitric acid dominated the  NO   removal as would be
                                          9      X
expected from prior experimental findings.   Specifically, this reaction
rate exceeded the gas phase production of nitric acid by about a factor
of 6.  In the course of checking out sensitivity of the system to the
chain breaking reactions, we individually varied reaction rate constants
in reactions (2.12), (2.13), (2.14), and (2.15).  A measure of the sen-
sitivity was the influence on nitrogen dioxide decay at late time.  The
greatest sensitivity of all was exhibited with respect to variation in
k..  .  A threefold increase in this rate constant, for example, resulted
in a fourfold decrease of end-point  N0?  concentration in the simulation.
The next reaction in the sequence, that between nitrogen dioxide and
nitrogen trioxide, had a lesser effect.  Perhaps counter to intuition,
an increase in  k-,  by a factor of three actually resulted in a slight
                                                                      45

-------
  (less  than 10%) increase in end-point nitrogen dioxide concentration.
  Rather large changes in the wall reaction rate  k    were introduced,
  but  these had virtually no influence on the system.
        Samples were periodically withdrawn during each smog chamber  experi-
  ment  to  analyze the composition of the gas.  "Clean" air replaced the
  sample in  each case.  In our evaluation process, we have found that this
  dilution can play a significant role in smog chamber experiments.   Thus
  dilution was accounted for in all of the simulations discussed previously.

        Given an average volumetric dilution rate for an experiment,  the
  first step was to convert this rate to a first-order "rate constant" by
  dividing by the chamber volume.  Then we subtract a factor due to dilu-
  tion  from  each chemical rate equation as shown below:
                 =R-6c                                          (2-23)
 where       c. = ith species concentration
              i
             R. = chemical rate for ith species
              i
              6 = average fractional dilution = average volumetric dilution
                  rate/chamber volume

       To test the validity of our approach for simulating dilution effects,
 we used the ethane concentration data in the propylene experiments, since
 ethane can be considered to be essentially unreactive.  The predicted and
 observed ethane concentrations for one such experiment, experiment 336,
 are shown in Fig.  2.13.   It can be seen that the two sets of data agree
 closely.  The maximum error is 5%.  Additional checks with other experi-
 ments produced similar results.
46

-------
                                            EXPERIMENTAL DATA
                                            COMPUTED
                                            CONCENTRATION
                    80
                              160        240
                                 TIME, min
320
          400
 Figure  2.13.   Experiment 336.   Dilution Model Compared with Measured
                Ethane Concentration
      The effects of dilution  on  the  reactive  species  can be seen in Fig.
2-. 14 which contains a plot  of  propylene  concentration  with and without
dilution;  The effect on  other species was  similar but the degree of
influence varies for each species.  Thus for nitric oxide, the dilution
effect is generally small due  to  the  rapid  decay of NO  by chemical
reaction, ±.e.,  R^  » 6c     in  Eq.  (2.23).  On the other hand,  for
ozone, dilution has been  observed to  play a large role late in the
reaction*
      Dilution can  also  account  for some of the discrepancy found in
the nitrogen balance  in  smog  chambers.   Our simulations indicate that
dilution effects can  account  for up to  25% of the nitrogen loss for those
experiments used in these  model  validations •.

      From the above  remarks »  it is clear that dilution can have a sig-
nificant impact oh  the whole  reaction.   Furthermore, because of system
                                                                       47

-------
          0.6
          0.4
          0.2
                                                 DILUTION
                                  WITH DILUTION
                      80
160        240
  TIME , min
                                                    320
400
Figure 2.14.   Experiment 336. Effect of Dilution on Propylene  Concentration
              (Curves Computed using a Single set of Rate Constants)
   nonlinearities,  simple scaling cannot be used to compensate for dilution
   effects.   Thus  caution must  be exercised when attributing changes  in
   species  concentration  to  chemical causes alone.   Therefore, it is  impera-
   tive that  dilution  data be included when experiments are reported  in  the
   literature.

         As a final comment, we note that the approach described above
   assumes  a  uniformly distributed dilution which implies a continuous
   sampling process.   This is,  of course, not the case, since sampling is
   done at  irregular intervals  and no dilution takes place between sampling
   points.  Whether approximating the discontinuous sampling process  by  a
   continuous  one  has  any significant effects on the simulation remains  to
   be  determined.
 48

-------
2.6.2  Smog Chamber vs Atmospheric Mixtures
      The reactivity of the mixture is one measure that can be used as
a guide to estimate the magnitude of the modifications in the rate con-
stants of the hydrocarbon reactions.  We have followed this approach in
              e\i
previous work.    Table 2.9 shows the mole-weighted reactivity of the
hydrocarbon mixtures in the atmosphere and in the smog chamber experi-
ments previously described.  The measure of reactivity is based on a
                                                25
hydrocarbon-consumption scale due to Altshuller.    The reactivity of
the Los Angeles atmosphere was obtained from our previous study of Los
                                  n r
Angeles atmospheric reaction data.
      It is obvious from the table that the dilute auto exhaust approxi-
mates the  reactivity of the Los Angeles atmosphere.  So we expect these
experiments to yield the most useful rate constant data for the model.
Our previous approach of using a  chemical model validated only for pro-
pylene and dividing its hydrocarbon rate constants by three for atmos-
pheric modeling purposes is, of course, a result of the reactivity
                               TABLE 2.9
                MOLE-WEIGHTED REACTIVITY OF ATMOSPHERIC
                 AND SMOG CHAMBER HYDROCARBON MIXTURES
                                                                      A
                  Mixture                                   Reactivity
Los Angeles Atmosphere                                         6+2
Propylene                                                       17
Toluene                                                          3
Toluene/n-Butane                                                1-6
Auto Exhaust—With  HC  and   CO   exhaust                        6.8
Auto Exhaust—Without  HC  and  CO  exhaust                      6
*                                       ,   25
 Based on a hydrocarbon  consumption  scale.
                                                                      49

-------
  relationship  shown  in  Table  2.9.   Finally, we  note  that  the reason for
  the  controlled vehicle exhaust  to  be  slightly  more  reactive than the
  uncontrolled  vehicle exhaust  is due to  the fact  that  the propylene con-
  tent of  the former  is  a larger  fraction of the total  hydrocarbon mix than
  is the case for the uncontrolled exhaust.
50

-------
3     MODEL METHODO-LOGY IMPROVEMENTS

3.1   PERSPECTIVES ON MODEL UPDATING
                27
      Initially,   our model treated a constant-thickness air layer with
a very compact lumped kinetic scheme.  Generalization of the model intro-
duced advection and more realistic chemistry, but ground-fixed coordinate
systems led to unacceptable numerical diffusion errors and the chemistry
                                                           O /
still needed more chain-termination steps.  Subsequent work   used scaling
                               r\ r
parameters from actual air data   for hydrocarbon reactivity adjustments
and for nitric oxide emission adjustments.  These adjustments have been
necessary in past work to convert theoretical estimates to the values
that better approximate what actually occurs in the atmosphere.  Further
model development saw the replacement of the Eulerian coordinate system
by the semi-Lagrangian system  now used.  Substantial improvements in computing
efficiency and accuracy were also made by introducing Fade approximant
numerical techniques.

      Still, some lingering questions remained unanswered.,  The present
work addresses these questions.  Chemical refinements were described in
the previous section.  These studies confirmed the validity of o.ur earlier
approaches in converting smog chamber findings to the atmosphere.  More-
over, the main structure of the model must necessarily deal with the
advection, diffusion, and the array of sources as they all interact in
air masses moving through regions where air pollution is to be predicted.
The following subsections deal individually with the current round of
improvements in modeling these phenomena.

3.2   ADVECTIVE AIR TRAJECTORIES
      The speed and direction of the air mass-center are determined, as
       O /
before,   by taking weighted averages of wind speeds and directions from
neighboring measurement stations.  It is useful to. examine the theoretical
basis for the weighting and its detailed application in the current phases
of our work.
                                                                       51

-------
      Reciprocal distance weighting of wind station readings is used as
a theoretical basis of interpolation.  The rationale for reciprocal-
distance over reciprocal-distance-square stems from the nature of fluid
dynamic singularities in plane flows.

      Because of the comparisons of horizontal and vertical scales in the
atmospheric boundary layer problem, the flow may be regarded more nearly
planar than three-dimensional.  Clearly, three-dimensional effects occur
at convergences and do generate vertical velocities, but horizontal
advection seems to be the most important character of the lower atmospheric
flow fields we are treating.  This being the case, it can be conceived
that any velocity field can be generated by the cumulative effect of plane
singularities such as sources, sinks, and vortex elements.  Classical
fluid dynamics shows that the dependence of the influence of each one of
these singularities upon distance is reciprocal in the distance from the
flow element.  Thus, if the flow at a field point is assumed to be the
superposition of flows described at neighboring points, it follows that
the relative weight given to each neighboring point should be proportional
to its reciprocal distance from the field point.

      The output of hand calculations of air trajectories takes the form
of schedules describing hourly magnitude and direction of successive
trajectory segments.  For each node point in the trajectory, air-quality
weighting factors are derived in addition to wind weighting factors.
These were applied to concentration data from the two or three nearest-
neighboring air monitoring stations (which do not generally correspond
to the neighboring wind measurement stations).   This gives an estimated
hourly history of air quality along the trajectory, thereby affording a
much larger data base than merely the end point concentrations.  The net
effect of the expanded base is to impose stricter standards on validation
compared with end point receptor-oriented calculations.
52

-------
      Figure 3.1 shows  two  typical  Los Angeles morning trajectories.

(Stations having hydrocarbon  data were chosen as origins for the trajec-

tory to minimize uncertainties  in initial conditions.)  In this case,

the Air Pollution Control District  (APCD) Downtown station has hourly

total hydrocarbon and methane concentrations.  The computed crossing

of trajectories may not have  occurred had there been data which would

allow the use of a shorter  interval size in the calculation.  The sampling

station located at the  City of  Commerce has gas chromatographic data

commencing in the early morning hours.  The validation runs for these

trajectories can be checked against interpolated air quality data from

nearby monitoring stations.


      Figure 3.2 illustrates  a  study comparing the surface wind trajectory

calculations with winds aloft tracked by an ESSA tetroon.  In this example,
    LEGEND:
    DOLA  - DOWNTOWN LOS  ANGELES
    COM
    ELM
    POMA
    WHTR
    WNTT
    RLA
    RVA
    VER
 COMMERCE
 EL MONTE
 POMONA
 WHITTIER
 WALNUT
 RANCHO LOS AMI 60S
-RIVERA
 VERNON
                                                        1330
POMA
                                                            WNTT
                                                  OCT. 29, 1969
           Figure 3.1.  Air Trajectory in the Los Angeles Basin
                                                                        53

-------
                         28
            •TETROON (MEASURED)
            •GROUND (CALCULATED)
MISH
      SEPTEMBER 30, 1969
                                                    BURBANK
                                                    COMMERCE
                                                    DOWNTOWN LOS ANGELES
                                                    HOLLYWOOD
                                               LACA —LA CANADA
                                                    LOS ANGELES INT'L AIRPORT
                                                    LENNOX
                                                    MISSION HILLS
                                                    PASADENA
                                                    VENICE
                                                    WEST LOS ANGELES
   Figure 3.2.   Comparison of Ground  Trajectory with Tetroon  Trajectory


the solid  line  is  the tetroon path and the  dashed  line is the calculated
surface  trajectory.   Short dashed arrows along the tetroon trajectory  show
one-hour segments  computed from wind station data.  Except for the  initial
segment, the computed surface winds have a  rightward heading from the
tetroon  path  (looking in the direction of tetroon  motion).  The result is
the surface trajectory heading north through Cahuenga Pass, over Burbank
and into the Verdugo Hills,  while the  tetroon travels due west in an
offshore direction until it  veers inland over Santa Monica late in  the
morning; therefore,  it appears that these differences between surface  and
                                                                      28
elevated winds  can be especially distinct.   Although Angell, et al.,
analyze  such disparities in  their paper, their comparisons exhibited a
higher degree of correspondence between paths than ours.  The difficulties
in neglecting height-dependence of advection are obvious to the modeler;
however, neither theoretical nor empirical  corrections are presently
available.  The means for handling this effect may come out of boundary
layer meteorological researches.
54

-------
      Even at a single location, variations in wind speeds and directions
appear to ciccur due to localized effects.  Figures 3.3 and 3.4 show plots
of Los Angeles Air Pollution Control District data versus Scott Research
Laboratories data.  Both sets of data were reportedly taken at El Monte,
California.  Because of topographic setting and instrumental differences,
the simultaneous readings show wide variances.  This may be due to hourly
averages of the Scott data versus instantaneous readings of the Los
Angeles APCD (LAAPCD) station.  In evaluating any air pollution simulation
model, this sort of deviation must be considered, as well as that described
in the preceding paragraph.  Confronted with these conflicts, we attempted
to resolve the differences by taking arithmetic means.

      An objective technique for approximating continuous wind fields
from discrete data will be needed in the near future.  In the course of
carrying out the hand calculations, we have developed the following
logical structure on which such a technique might be based:
      1.    Forward differencing is the zero order approximation; that
            is, each segment is laid out based on the hourly averaged
            wind that is interpolated from station data surrounding its
            origin-point.
      2.    Higher approximations will require interpolation of wind data
            to the midpoint of the zeroth order segment and then refining
            the magnitude and direction of the segment in the first
            approximation.
      3.    Reiteration of the segment calculation may be coded with
            either an iteration counter or a convergence criterion to
            terminate the succession of approximations for the segment.
            (Discreteness of wind data may prevent convergence if
            proximity tests intervene in the sequence and alter the
            selection of wind stations utilized for interpolation).
      4.    Beyond some maximum distance, wind stations must be rejected
            even if it means reducing the number of input stations.
                                                                       55

-------
    12,
    10
  E

  et
                                  •    •
                             I •    •
                            _L
I
I
                            468
                            SCOTT WIND SPEED DATA, mph
                      10
                      12
      Figure  3.3.   Comparison of Wind Speed Measurements at El Monte
                   (Mornings in 1969)
56

-------
   500.-
   400
CD

Ol
«=c



Q


z
o
t—I
I—
CJ
   300
Q


Q
   200
Q

C_5

Q-
   100
                  TOO
                              200
300
400
                      SCOTT WIND DIRECTION  DATA,  deg
500
 Figure 3.4.  Comparison of Wind Direction Measurements at El Monte

              (Mornings in 1969)
                                                                    57

-------
       5.    Closer than some minimum distance, a single station's wind
            vector should be used directly.
       6.    If all station distances from the region of interest exceed
            a certain threshold (input) for a certain numBer of segments
            (also input) the computation should be terminated with an
            exit flag.
       7.    Numerical experiments with directional weighting should be
            conducted for both trajectory generation and for inter-
            polated air quality at a trajectory point; for example, the
            air quality should be more sensitive to where it has been
            than to where it is going.  For improving the zero-order wind
            motion approximation,  direction and magnitude of a trajectory
            segment will be more sensitive to station readings near its
            future path than those it is moving away from.
       8.    Numerical experiments  should be conducted on the relative
                       -1         •          -2
            merits of r  -weighting versus r  -weighting.  This is done
            most directly by using a wind station as an unknown point.
            Then the three nearest neighbors (obeying the distance and
            barrier selection laws)  are weighted by each power and the
            merit of each is evaluated by some measure such as least sum
            of square residuals between predicted and measured value's i

 3.3    VERTICAL AND HORIZONTAL EDDY DIFFUSION COEFFICIENTS
       In the evaluation phase of developing the GRC photochemical/diffusion
 model, greater emphasis is placed  on the use of temperature profiles for
                                                                 0 "7
 the test days.  This led us to reappraise our earlier formulation   of the
 eddy diffusivity profile by incorporating more of the measured data that
 have been reported in the literature.
58

-------
      Up to now, the eddy diffusivity profile has been assumed to be
                             ft
trapezoidal from the "ground"  Up to the mixing depth (previously set at
                                        29
the inversion base).  Following Estoque>   the ramp portions of the profile
extends to a height of 50 meters and nonvanishing Values are set at the
bottom and top of the vertical mesh.  Assigning the flat portion of the
profile of value of  K   , the vertical diffusivity, we used the formula
                      Zoo
                       2
K^  * 8300 (u + 500) cm /seconds where  u  is the wind speed in cm/second.
  OO
The approximation was based on a fit of scattered data from various sources
                        77
cited in our 1969 paper.

                                                                         30
      A key element in the review of; this approach is the use of Hosier's
vertical diffusivity data (for a 0 to 90 m height interval) to determine
a useful method of calculation that can be based on available data.
Figures 3.5 and 3.6 show the vertical diffusivity plotted versus wind
speed and vertical temperature gradient.  The widely scattered values in
either Case are discouraging to the theorist who wishes to use finite
difference diffusion terms.  Note on Fig. 3.5 that the wind velocity
dependence previously chosen for  K   is inconsequential compared to the
                                   z
two-ofder-of-magnitude scatter of the data.  In contrast, an examination
of Fig. 3.6 reveals a relatively systematic dependence on vertical tem-
perature gradient.  The velocity formula plotted on Fig. 3.5 gives an
estimate of vertical diffusivity lying largely in the neutral stability
range.  This may be expected to approximate conditions .averaged over space
and time in the marine layer overlaid by an inversion layer.

      On the other hand, temperature gradient inputs are apparently far
more influential than wind speed inputs in determining vertical diffusivity
values.  As an improvement on our earlier approach, we have reconstructed
 Because of the assumption of uniform horizontal velocity profiles, we
 take OUE "ground" elevation to be something like mean rooftop height
 where the air sampler inlets were located on the Scott stations.  This
 is equivalent to setting  z - 0  at elevations an order of magnitude
 or ewo greater than roughness heights.
                                                                       59

-------
     10 V
CSJ
 o
     10
          I— I
          05

                                         + 500)
                                    I
         J
        0       100      200      300      400      500

                                WIND SPEED, cm/s
600      700
       Figure  3.5.  Vertical Diffusivity Versus  Wind  Speed (Ref.. 30)
 60

-------
     icr
               •  *•!
                                                                           CXI
                                                                           05
                                        100 < u < 700 cm/s

                                        CURVE FAIRED  TO DATA
               UNSTABLE

        VERY        \
        UNSTABLE-I*1*-}
      -2
       0                1
(AT/AZ), °c/ioo m
Figure 3.6.  Vertical Diffusivity Versus  Vertical Temperature Gradient
              (After Ref. 30)
                                                                        61

-------
 the diffusivity profiles to represent the broad stability categories
 delineated along the bottom of Fig. 3.6.  Dropping the dependence  on
 wind speed and introducing the dependence on temperature gradient, we
 might expect a larger dynamic range of diffusivity values throughout any
 given simulation run.  It is difficult to justify any more elaboration
 on the diffusivity calculation on the basis of the body of theory  that
 depends on fluxes and stresses.  The guesses necessary to arrive at a
 Richardson number or a Monin-Obukhoff length parameter cannot be seriously
 expected to yield better results than the correlation shown in Fig. 3.6
 (especially in view of the poor correlations with Richardson number cited
 by Hosier).

      Having a guide to the choice of the plateau value of vertical diffu"
 sivity, we now turn our attention to assigning a ground value for  diffu-
                                                                    -2
 sivity.  Surface shear stresses typically vary from 1 to 10 dynes-cm
                               *
 leading to friction velocity  u   values of 20 to 90 cm/second.  For
                                     *
 neutral stability conditions  K  = ku z  assuming that the vertical diffu-
                               z
 sivity approximately equals the eddy viscosity.  Typically, the coefficient
 k  takes on values about 0.4 if  u  - 50 cm/second.  Assuming the  "ground"
 plane to be at about 5 meters, we get  K(0)  ~ 10  cm /second.  Considering
 the larger roughness for urban areas as compared with the surfaces in
                                      32
 Table 3.1 on p. 72 of Pasquill's book,   this falls just above the range
 of  K  values listed there.  For very stable conditions as defined on
 Fig. 3.6, the value of  K(0)   is likely to be far less than 10  cm /second.

      Translating the ground values and the elevated values of vertical
 diffusivity (Fig. 3.6)  into profiles, we arrive at the data shown on
 Fig. 3.7.  The choice of 50 m for the knee height illustrates a nominal
 value.  Because of uncertainties, this height is taken to be the first
 mesh point above the ground in a five-point mesh.  For mesh ranges
 typically varying from 100 to 300 m, the knee will be 25. m to 75 m above
 the ground.  Details of K-profile shape at the top of the mes.h have little
62

-------
    2A
CI5
               2x10
                           KNEE HEIGHT A ± 50 m
                           ACCORDING TO ESTOQUE
                     29
                           IN MODEL CALCULATIONS A ^ 30
                           TO 70 m, DEPENDING ON MESH
                           INTERVAL SIZE
4x10
                         VERTICAL DIFFUSIVITY, cm^/s
                                    CO
                                    Ł
                                    in
                                                                        tvj
                                                                        oi
                                                                         I
2x10'
 IN URBAN, AREAS 0-F VERY TALL BUILDING HEIGHTS, ESTOQUE'S ESTIMATE MAY BŁ

TOO LOW; HOWEVER, THIS EXCEEDS ROUGHNESS HEIGHTS IN NEARLY ALL AREAS

OVER REGIONAL EXTENT.
 Figure 3.7-  Vertical Diffusivity Profiles (Fig. 3.6 Defines Stability
              Categories)
                                                                       63

-------
influence for the space and time scales of our modeling calculations;
therefore, for photochemical/diffusion validations the profiles are
taken to be constant all the way to the top.

      Reference to Fig. 3.6 shows how the vertical diffusivity values are
assigned.  Data on the graph are for 100 cm/second to 700 cm/second wind
speed.  As mentioned previously, a comparison of Figs. 3.5 and 3,6 shows
a tighter grouping of points on the temperature gradient plot than there
is on the velocity plot; hence, the use of this correlation in place of
the velocity formula.  Hybridization of these profiles may be a good
approach if the vertical mesh is taken to be higher than the inversion
base.
      Horizontal diffusivities,  K  ,  were needed for the transverse
diffusion tests of the model that are  reported elsewhere.  The choice of
                       /:   n
a typical value, 5 x 10  cm /s, was based on the Los Angeles tetroon data
                   33
of Angell and Pack.    One can infer a Fickian diffusion coefficient from
the horizontal separation of tetroons  flying a few hundred meters above
the ground.  The values grow with time so that some time scale charac-
terizing the problem must be selected.  It will be noted from Fig. 3.8
that the value of  K   selected represents a travel time of about one hour.
Figure 3.8 shows lateral separation of simultaneously released tetroon
pairs as a function of time since release for flights in morning (IFA)
and afternoon (IFF) at Idaho Falls, and flights at Atlantic City (ACY)
and Los Angeles (LAX) .  The heavy line represents the mean of these
weighted according to number of cases.  Also entered are lateral eddy
diffusivity (K ) isopleths based on Fickian theory, and dissipation
isopleths («) based on similarity theory (data reproduced from Angell
and Pack).33
64

-------
                               30        100
                             TRAVEL TIME, min
300
 Figure  3.8.   Lateral Separation of  Simultaneously  Released  Tetroon Pairs
              as  a  Function of  Time  Since Release (after Ref. 33)
3.4   EMISSION FLUX HISTORIES
      For the purposes of production computations, we produced a code
that traced air trajectories using a 2-mile interval grid for numerical
computations and source inventory compiled by Roberts, Roth, and Nelson,
The inputs are wind speeds and headings along line segments of a trajec-
tory.  The output is the flux of each pollutant into the air parcel for
the time/space-history traced out by the parcel.  This code, which was
developed by R. Nordsieck prior to the present phase of the work, produces
output on punched cards suitable for use as inputs to the GRC photochemical/
diffusion model.

      The primary revisions made in the emissions model during this study
involved- (1) the introduction of additional spatial and temporal variations
in freeway traffic emissions resulting from vehicle emission rate vfria-
tions with average speed, and (2) inclusion of a correction factor for
surface street emissions to account for the non-uniform temporal distri-
bution of vehicle cold-start emissions.
                                                                       65

-------
4     TRANSVERSE DIFFUSION AND ITS EFFECT ON THE GRC MODEL
      Two analyses have been conducted to assess the errors incurred  in
the model calculations due to the omission of transverse diffusion  in the
semi-Lagrangian formulation.  In the first, we have assessed  the  effects
of lateral exchange between adjacent stream tubes.  In the second analysis,
we have quantified the errors which can result when our Lagrangian  control
volume passes near an elevated point source, but fails to sweep over  it,
thus ignoring its contribution that spreads laterally into the control
volume.  Both of these investigations were performed using our L_ocal  Air
Pollution ^simulation (LAPS) code (see Appendix A, Sec. A.3.2) which
incorporates lateral diffusion but is limited to quasi-equilibrium  chemistry.

4.1   LATERAL DIFFUSION BETWEEN NEIGHBORING STREAMTUBES
      To examine the effect of lateral exchange between "parallel"  trajec-
tories, we located a relatively straight trajectory originating at  the
Commerce station and moving north to Burbank.  We then synthesized  a
                             *
parallel trajectory two miles  to the west by simply moving the start
point two miles west and assuming an identical wind history.  Figure  4.1
shows histories of CO surface flux obtained for these trajectories  from
our source emissions program.  (The fluxes are normalized with respect to
air density to yield the somewhat unusual units of meters per minute.)
Assuming a 200-m inversion height,  and the slightly unstable diffusivity
profile shown in Fig. 4.2, the LAPS model was used to simulate atmospheric
CO concentrations within two parallel area-source strips 3 km wide.
The lateral cell size was 500 meters, providing six cells over each source
strip.  Outside boundaries of the horizontal mesh were assumed fully
reflecting, simulating a semi-infinite area source on each side.  Initial
concentrations of CO were set to zero everywhere.  Figures 4.3 and  4.4
show the concentration histories obtained along two symmetrically located
 *
  Two miles was selected because the minimum resolution which can be
  expected in our model is set by the two-mile aggregation imposed by the
  source model.34
**
  The two-mile distance was approximated by 3 km in this test case for
  convenience in the use of metric units.
66

-------
                                                           RIGHT SIDE TRAJECTORY

                                                          • LEFT SIDE TRAJECTORY
                                         TIME, hours
       Figure  4.1.   CO  Flux Histories  on Two Neighboring  Trajectories
                        200
                                           INVERSION BASE
                      o
                      3.
                        100
                           VERTICAL EDDY DIFFUSIVITY, 104 cm2/s
Figure 4.2.   Eddy Diffusivity Profile for Neighboring Streamtube Analysis
                                                                                67

-------
             8r
          5  6
          o
          co
          Q-

          D-
          o
          o
          o   2
          i_i   ^
          C_>
          z
          o
                          RIGHT SIDE WITH

                        LATERAL DIFFUSION
                                           RIGHT SIDE,  NO EXCHANGE
FULL  REFLECTION AT OUTER  BOUNDARIES



          I	J	I
               012345

                                       TIME,  hours



 Figure 4.3.   CO Concentration  Histories  on Right-Hand (Commerce) Trajectory
            ^,   6
                                       LEFT SIDE,  NO EXCHANGE
                                                    LEFT SIDE WITH

                                                    LATERAL DIFFUSION
                                       FULL  REFLECTION AT OUTER BOUNDARIES
                                       TIME,  hours
Figure  4.4.   CO Concentration Histories  on Left-Hand  (Synthetic)  Trajectory
  68

-------
parallel paths 3.5 km (about 2 mi) apart with and without lateral diffusion
                                           /•   rt
(using a lateral eddy diffusivity of 5 x 10  era. /s) as discussed in Sec. 3.3.
The quantitative differences seem to be small.

      To get an idea of what a worst case might be like for the parallel
streamtube analysis, we superposed the CO flux histories from 24 test
trajectories to indicate the real-world bounds on CO flux histories in
the Los Angeles Basin (Fig. 4.5).  It was decided that the envelope shown
in Fig. 4.6, representing a constant ratio of 3:1 between maximum and
minimum, would provide a very acceptable bounding case for two trajectories
only two miles apart.  Using the simulation parameters described above,
the concentration histories shown in Fig. 4.7 were obtained for the worst
case comparison.
                                    10.00
                                   LOCAL TIME
     Figure  4.5.   Superposed  CO  Flux  Histories  for  24  Trajectories
                                                                      69

-------
8
e 6
"E
° 4
X
U_
o
0
Figure 4.6. Bo
-
___



_ J
•=<
L__ HIGH FLUX
1 | LOW FLUX
1 1
5 7 9 11 13
TIME, hours
unding CO Flux Histories for Parallel Trajectory Analysis
             FULL REFLECTION AT OUTER BOUNDARIES
                                     3         4
                                      TIME, hours
     Figure 4.7.   CO Concentration Histories  for Worst-Case Parallel
                    Trajectories
70

-------
      Table 4.1 summarizes worst case errors that may be encountered under

the assumption of zero initial concentration.  The fractional error will

be lower if the initial concentration is greater than zero.  In effect,
Table 4.1 shows the worst case errors for increments of concentration

rather than for concentration itself.  Table 4.2 shows the worst case
errors that could be found assuming an initial concentration of 10 ppm

of CO, which is a typical early-morning value.


                                TABLE 4.1

              "WORST CASE" BOUNDING ERROR FRACTIONS DUE TO
            OMISSION OF LATERAL DIFFUSION IN URBAN MODELING
         ASSUMING ZERO INITIAL CONCENTRATION OF CARBON MONOXIDE
Time
5 hr
8 hr
High Flux
0.28
0.36
Low Flux
0.39
0.44
                                TABLE 4.2

           "WORST  CASE" BOUNDING ERROR FRACTIONS DUE TO OMISSION
            OF LATERAL DIFFUSION IN URBAN MODELING, ASSUMING
           AN INITIAL CONCENTRATION OF CARBON MONOXIDE OF 10 PPM

                  Time        High Flux        Low Flux
                  5 hr            0.12            0.14

                  8 hr            0.17            0.19
                                                                       71

-------
4.2   LATERAL DIFFUSION EFFECTS IN THE VICINITY OF HIGH-FLUX
      POINT SOURCES
      A quantitative assessment of the pollutant contributions of high-
flux elevated point sources to neighboring trajectories was performed in
a parametric fashion, by determining the ranges of real-world surface
fluxes and point source fluxes encountered in the Los Angeles Basin and
examining the results of superposing various combinations.

      Power plant stacks were chosen as typical high flux elevated pollu-
tant sources, and hence, in order to set background fluxes, we needed to
establish the range of surface NO fluxes encountered in the Basin.  Pro-
ceeding as with the CO fluxes, Fig. 4.8 shows the NO flux histories
obtained for our 24 test trajectories.  The heavy lines indicate  the three
fluxes selected fr.om the range of NO fluxes encountered.  They are equiva-
lent to 10, 15 and 20 kg/hr/km .  The source model of Science Applications,
                                                         13.00     14.00
      Figure 4.8.  Superposed NO Flux Histories for 24 Trajectories
 72

-------
    34
Inc.   shows a distribution of power plant NO fluxes ranging up to
532 kg/hr (1172 Ib/hr) .  From this distribution, we chose nominal stack
NO fluxes of 60, 200, and 600 kg/hr as including roughly 33%, 67%, and
100% of the power plants on a relative occurrence basis.  An effective
point source height of 100 meters was used to include the height of an
average stack plus the additional plume rise due to jet momentum and
buoyancy effects.
      The meteorological conditions simulated for these cases included
an inversion base height of 183 meters with neutral (2 * 10  cm /s)
                                                       3   2
diffusivity below the inversion and very stable (5 x 10  cm /s) above.
The resulting profile of vertical eddy dif fusivities is shown in Fig. 4.9.
                                                    6   ?
As before, the lateral diffusivity was set at 5 x 10  cm /s.  A light wind
of 1 m/s  (about 2 knots) was assumed.
             o
             CŁ
             03
             O
             CO
                300,-
                200
                100
                                         INVERSION BASE
 Figure 4.9.
     0123
      VERTICAL EDDY DIFFUSIVITY, 104' cm2/s
Eddy Diffusivity Profile for Elevated Point Source/Lateral
Diffusion Analysis
                                                                       73

-------
      Since the object of these tests was to assess the effect of lateral
diffusivity on trajectories missing large point sources, and the basic
resolution of our source model is two miles, we have chosen to simulate
the average situation in which the stack is centered in a neighboring
two-mlle-wide path and we are reading concentrations two miles distant
at the center of another two-mile swath.  Accordingly, Figs. 4.10, 4.11,
and 4.12 show NO concentrations at the ground, two miles from the plume
centerline for each of the three background NO fluxes combined with stack
NO fluxes of 0, 60, 200, and 600 kg/hr.  Below these curves, Figs. 4.10,
4.11, and 4.12 plot the corresponding percent error incurred if lateral
diffusion were ignored and only the background flux is accounted for.
Table 4.3 summarizes the maximum fractional errors associated with each
combination of background and stack fluxes.  Even under the worst condi-
tions, these error percentages are much smaller than NO-flux uncertainties.
74

-------
      20
      15
  o
      10
0    5
a
o
                                                   STACK FLUX,  kg/hr
                                                            600
                                                            200
                                                            60
                                                         0
                                NO  SURFACE FLUX = 10 kg/hr/knT
                                RECEPTOR 2 mi FROM PLUME CENTERLINE
                                EFFECTIVE STACK HEIGHT = 100 m
                                WIND SPEED = 1 m/s
                                NEUTRAL STABILITY
                                                                       «6
                                                                       NJ
                                                                       4
  01
  LU
      2§ r-
      20
      15
      10
                                            STACK FLUX kg/hr
                                                 600
                 20
                                60
                             TIME, min
80
100
120
Figure 4.10.   Srouiid  Coneeiitratiori fiffeets Of  Elevated Po'int  Sources:"
                           2
               lO.kg/hE/km.  Background Flux
                                                                          75

-------
    .0-
    o_
         30
         25
         20
     LU

     O
     o
     o
     o
     o;
                                                       STACK FLUX, kg/hr §
NO SURFACE  FLUX   15 kg/hr/km"


RECEPTOR  2  mi FROM PLUME CENTERLINE

EFFECTIVE STACK HEIGHT = 100 m


WIND SPEED  =  1 m/s

NEUTRAL STABILITY
     o
     o;
     o
     o;
                                            STACK FLUX kg/hr


                                                  600
                                     60       80

                                 TIME, min
                    100
120
  Figure 4.11.   Ground Concentration Effects  of  Elevated Point  Sources:
                            2

                 15 kg/hr/km  Background Flux
76

-------
   a.
   O-
   o


   o
   o
   •z.
   o
   o
   QŁ
   CS
                                                       STACK FLUX, kg/hr

                                                               600

                                                               200
                                                               60
       NO SURFACE FLUX = 20 kg/hr/km^

       RECEPTOR 2 mi FROM PLUME  CENTERLINE

       EFFECTIVE STACK HEIGHT =  100 m

       WIND SPEED = 1  m/s

       NEUTRAL STABILITY
   cc.
   o
   O
   Of
        10
                STACK FLUX kg/hr

                     ,600
                  20
40       60       80

      TIME, min
120
Figure  4.12.   Ground  Concentration  Effects of Elevated Point  Sources:


               20 kg/hr/km2 Background Flux
                                                                          77

-------
                                 TABLE 4.3
           MAXIMUM FRACTIONAL CONTRIBUTION OF AN ELEVATED POINT
      SOURCE AT A GROUND LOCATION TWO MILES FROM THE PLUME CENTEPvLINE
                      Effective Stack Height =? 100 m
                      Neutral Stability,  2 knot wind


Average
Surface
Flux,
kg/hr/km2

10
15
20
Stack Flux, kg/hr
60
0.022
0.015
0.011
200
0.072
0.048
0.036
600
0.217
0.145
0,109
78

-------
5     VALIDATION STUDIES

5.1   SELECTION OF DA.YS FOR MODEL TESTS
      The culmination of the improvements in chemistry and meteorology is
a set of controlled retests of the GRC Photochemical/Diffusion model and
a conversion of the code to IBM 360-compatible form.  The selection of
days was originally predicated on a consistent basis with the design of
                                                          35
the Scott Research Laboratories Los Angeles Basin Program;   namely, that
morning air movements from Commerce to El Monte would be studied.  The
design philosophy was that primary pollutants (reactants) would be carefully
measured at Commerce with special emphasis on gas chromatographic resolution
of hydrocarbon samples.  Advection and reaction of the air mass would then
occur with the composition of the secondary pollutants (products) indicated
by (hopefully) downwind measurements taken at El Monte.  To implement this
philosophy, we conducted a systematic search for days having morning air
trajectories that nearly connected the two stations.  The original program
design did not consider the use of Los Angeles County Air Pollution Control
District station data because they are difficult to obtain and they are not
as detailed as the Scott Research Laboratories station data.

      According to the field program objectives, our rationale for a
requested selection of 9/4/69, 9/15/69, 9/27/69, 10/16/69, 10/29/69, and
11/4/69 was based on air movement calculations.  To get these dates, we
calculated trajectories that could be used to validate the transport/
diffusion, mo.dule (and ultimately the photochemical/diffusion model).  The
calculations used the 1969 Scott Research Laboratories data which were
collected at Commerce and El Monte in the Los Angeles Basin.  We performed
a search for trajectories which originated at Commerce in the morning and
arrived at El Monte later in the day.  The se'arch was accomplished using
a special trajectory-generation program designed to compute a wind trajec-
tory between two stations given wind speed and direction data at each
station.  The program calculates inverse distance-weighted averages of
                                                                      79

-------
wind speed and direction between the stations and uses these quantities
to generate the trajectory.  The output of the program consists of  the
following trajectory descriptors:
      1.    Geographical coordinates of the air parcel
      2.    Time, wind speed, and direction of the air parcel at
            every point in the trajectory
      3.    Distance of the air parcel from each of the two reference
            stations
A computer plot of the trajectory can be obtained on an optional basis.

      The program has the capability to detect and compensate for anomalies
in the data such ,as missing data points.  Moreover, the program warns the
user about the existence and nature of the data defects and of the  actions
taken to overcome the deficiencies.  This allows the analyst to assess
the reliability of a trajectory.

      We considered an air parcel to have "arrived" at El Monte if  it
passed within one mile of El Monte.  The search examined all the trajec-
tories originating at Commerce from 0600 to 0900 for the 72 days (Aug. 28
to Nov. 7) of data available at both Commerce and El Monte.  A total of
1368 possible trajectories over the 72 days were computer-analyzed.  Only
87 trajectories spread over 12 days were found to satisfy the miss-distance
criterion of one mile or less.  Additional evaluation of the wind and
aerometric data eliminated five of these days, thus leaving only seven
possible days with usable trajectories.  The seven days are listed below
in order of number of trajectories 'available on each day:
80

-------
Date
Oct. 1.6
Nov. 4
Sept. 4
Oct. 29
*
Sept. 6
Sept. 15
Sept. 27
No. of
Trajectories
18
9
9
8
6
3
3
Peak CO (ppm)
(Commerce)
7.9
25.2
12.2
18.4
10.2
10.0
5.0
                                                     Peak 03 (pphm)
                                                       (El Monte)
                                                          10.0
                                                          12.0
                                                          18.1
                                                          19.0
                                                          14.6
                                                           5.4
                                                          24.5
A
 Missing data
For Sept.  6, El Monte is missing NO ,  HC, and CO data.   For Commerce,  we
                                   X
found that Oct. 29 is missing the NO measurements for the interval 0600-
0730, which is precisely the interval when the trajectories originate.
The El Monte data could have been supplemented by LAAPCD Azusa data,
except for the hydrocarbon measurements.   Supplementing the Commerce  data
may not have been possible, since no LAAPCD station is  located nearby.

      The program objectives were then redirected and the original two-
station concept was discarded.  The newly adopted selection criteria
stressed availability of airborne temperature data and  high peak oxidant.
All but two of the requested dates were discarded in favor of the follow-
ing set of dates:  9/11/69, 9/29/69, 9/30/69, 10/29/69, 10/30/69, and
11/4/69.

      The chemical data for Sept, 29 are satisfactory and the point of
closest approach to El Monte is less than a mile.  However, the time  of
closest approach occurs around 1100.  Thus for a start  time of 0600,  we
have only a five-hour travel time.  The data for September 11 are excel-
lent, but for a 0600 Commerce start, the trajectory misses El Monte by
about 6 miles and would require interpolation to estimate the final con-
centrations.  The trajectories for Sept.  30 and Oct. 30 move in a westerly
direction from Commerce and would also require interpolation at the
destination point.

                                                                       81

-------
      The decision to obtain Los Angeles County Air Pollution  Control
 District data and to carry out validation studies for  these  days  forced
 changes in our fundamental approach.  Because of critical dependence  of
 model results on initial conditions, we sought trajectories  beginning at
                                *        .            .
 stations having hydrocarbon data  in addition to measurements  of  other-
 pollutants.  With one exception, the trajectories were source-oriented
 rather  than  receptor-oriented.  As mentioned previously, weighted averages
 of air  quality values were computed from station measurements  for each
 hourly  node  of each trajectory.

      Three  days were designated to allow adjustments  in parameters.
 These so-called "hands-on" days are 9/29/69, 9/30/69,  and 10/29/69.   The
 remaining three days are reserved for testing the fully adjusted  model
 without manipulation of coefficients.  These so-called "hands-off" days
 are 9/11/69, 10/30/69, and 11/4/69.  The hands-off days have lower peak
 oxidant  (generally) than the hands-on days.

 5.2  PROGRAM CONVERSION
      Conversion of the DIFKIN program for operation on the  IBM 360/50
 was completed early in the present phase of the work.  Specific tasks
 carried out  can be grouped into the following categories:
      1.     Elimination of nonstandard software which  is incompatible
             with the IBM 360/50.
      2.     Inclusion of FORTRAN software which is characteristic of
             the IBM 360/50 and cannot be used in our CDC 6400
      3.     Testing the accuracy of the solution to a  sample problem
             to determine the effect of th6, reduced precision of the
             IBM 360/50 (see Sec. 5.2.1)
      4.     Determining the differences in running time between CDC 6400
             and IBM 360/50 (see Sec. 5.2.2)
*
 The stations are Commerce, El Monte, Downtown, East San Gabriel Valley,
 and West San Gabriel Valley.
82

-------
      5.    Producing a punched deck of cards whose character code is that
            used by the IBM 360/50.  (Extended BCD for IBM 360/50 compared
            to BCD for CDC 6400.)
      6.    Eliminating diagnostic statements used when the program was
            being developed.
      7-    Using a capability of our computer center to generate a source
            deck with all the statement numbers sequentially ordered for
            maximum readability.

5.2.1  Precision Test
      A sample problem consisting of a polluted air parcel sweeping over
a heavily used freeway was run in single-precision mode on both the
CDC 6400 and the IBM 360/50 and the results compared.  For reference
purposes, we recall that the precision of the CDC 6400 and the IBM 360/50
is 14 and 7 decimal digits, respectively.  Thus we might expect some dif-
ference in the results produced by each machine.  Analyzing the answers,
we found that they agree to four significant figures and differ by at most
two units in the fifth (least) significant digit.  Thus the difference is
2 parts in 10,000 at worst.  The table below shows the frequency and magni"
tude of the divergence for one of the eleven species computed; the total
number of points sampled is 300.

            Units of Difference in
            Least Significant Digit         Number of Points
                      0                            65
                      1                           227
                      2                             8

      Regarding the conversion, two comments are in order.  The first is
that since the computational technique is unconditionally stable, no
problems are anticipated due to amplification of the roundoff error.  The
second is that for the major species, i.e., NO, N02, HC, 03, CO, the three
most significant figures are sufficient for comparison with experimental
data.
                                                                       83

-------
5.2.2  Timing Test
      The test problem ran for a considerably longer  time  in  the IBM 360/50
than in the CDC 6400.  For example, total running  time  for the  IBM 360/50
was 18.6 minutes to compute 30 minutes of real time,  for a real time/
computer time ratio of 1.6:1.  For the CDC 6400, the  figures  are 4.4 minutes
of computer time to compute 60 minutes of real time,  for a ratio of 14:1.
Thus the running time was increased by more than a factor  of  8.   Recent
improvements have greatly improved the speed ratios since  these timing
tests were conducted.  Consequently, the running time for  either system
should be reduced considerably to reflect current  practice.

5.3   ATMOSPHERIC VALIDATION TESTS

5.3.1  Introduction
      In this section, we report the results of validation tests carried
out for four trajectories on each of the following six  days:  Sept.  11,
29, 30, Oct. 29, 30, and Nov. 4, all in 1969.  Three  of the trajectories
also serve as baseline cases for the transportation control strategy
study which is described in a separate volume.

      September 29, 30, and Oct. 29 have been designated to be  so-called
"hands-on" days for validation purposes.  This means  that  parameter  adjust-
ments can be made in order to improve simulations  of  the measured concen-
tration histories of the various pollutants.  The  adjustable .'parameters
are the diffusion coefficients and the rate constant  of the reaction
OH + HC -> (b2) R02  , designated by  k, .  The experience gained  from working
with the hands-on days is then used to develop guidelines  for parameter
selection to simulate air quality for the remaining three  days,  Sept. 11,
Oct. 30, and Nov. 4.  The latter are thus designated  "hands-off"'days.
The guidelines and the inputs that are necessary to apply  them  are
described later in the report.
84

-------
      In addition to the test results, this section also contains descrip-
tions of the test procedures and criteria used in evaluating the model.
As a necessary adjunct to model evaluation, data base errors are examined
and illustrative examples given.  The test results are also assessed on
the basis of statistical correlations between predicted and observed
concentrations.

      A map of the Los Angeles Basin which contains the location of the
monitoring stations is included in Fig. 5.1; the legend for the abbrevia-
tions used on the map is found in Table 5.1.  A list of the trajectories
is shown in Table 5.2 and the initial concentrations used in the simula-
tions are in Table 5.3.

5.3.2  Approach and Criteria for Model Evaluation
      The approach taken for the hands-on cases was to use CO concentra-
tion histories to determine the diffusivity parameters for each trajectory.
The diffusivities thus obtained were then applied without change to the
same trajectory with the reactive species.  Because of spatial variability
of meteorological conditions, all the trajectories for a-single day could
not generally be described by the same diffusivity coefficients.  In
addition,, it was necessary to adjust the rate constant  k,  for the reac-
tion  OH + HC ->• (b.) RO   for some of the reactive cases.  Subsequently,
guidelines were worked out for determining the diffusivities and  k.  to
be used in the hands-off cases.  These guidelines appear in Sec. 5.4.

      Some remarks are in order about the nature of the test that is
applied to the computed concentrations in relation to observed concentra-
tions.  Let us begin by reviewing the simulation process:  the computed
results represent the history of the pollutants in an air parcel which
traverses a geographical area following a path determined by the local
hourly average wind speed and direction.
                                                                       85

-------
                                           J7W HOL
                                      102 W-
                                      WEST*
                                                 • 75W
                                                 DOLA
                                                   34WVERB  OCOM
                                                           116WPASA     m
                                                          •79WSGV    ^

                                                          lltW
                                                         1 ALMS   O ELM
                                                                                 97W
                                                                                 AZU
                                                                                                /61WONT
1QBW/
POMA
 r— /
                                                                    81W RVA
                                                                             106WWNTTI
                                                               • 22W
                                                               "RLA
                                                          112WCOMA
     LOS ANGELES COUNTY

AIR POLLUTION CONTROL DISTRICT


  METEOROLOGICAL NETWORK


       JANUARY 1972
                                                           72SC
                                                         A —101WLONB/
                                                        80W    -^     /
                                                        OOM    21W  /
                                                               LGB /
                                                                80SE  |~
                                                                 • 114W
                                                                   WHTR

                                                                 BgRVVKFI   "~— ~™ ""*"" Los Angeles Basin Limits
. Upper Santa Clara River
 Valley Basin Limits
 Antelope Valley Basin Limits

• County Boundary

 APCD Air Monitoring and
 Meteorological Stations

 APCD Meteorological Stations

 Cooperating Meteorological
 Stations

 Scott Research
 Laboratories Stations
     Figure  5.1.    Air Quality  and  Meteorological  Monitoring  Network in  the
                         Los Angeles  Basin
86

-------
                         TABLE 5.1
         DIRECTORY
        MONITORING
Abbreviation

    ALMS

    AZU

    BRT

    BUR

    BURK

    GOMA

    CPK

    DOLA

    DOM

    ELM

    ENC

    FOX

    HOL

    KFI

    LACA

    LANG

    LAX

    LENX

    LGB

    LONB

    MALC

    MDR
OF AIR QUALITY AND METEOROLOGICAL
STATIONS IN THE LOS ANGELES BASIN

               Location
 Alhambra

 Azusa - East San Gabriel Valley

 Brackett

 Hollywood-Burbank Airport

 Burbank - East San Fernando Valley

 Compton

 Canoga Park

 Downtown Los Angeles - LAAPGD Headquarters

 Dominguez

 El Monte

 Encino

 Gen. Wm. J. Fox Airfield

 Hollywood

 KFI Transmitter

 La Canada

 Lancaster

 Los Angeles International Airport

 Lennox

 Long Beach Airport

 Long Beach - South Coast

 Malibu

 Marina del Ray
                                                                 87

-------
 TABLE 5.1 (Cont.)




        Abbreviatirn




           MISH




           MWS




           NEW




           NP




           NTB




           ONT




           PASA




           PICO




           PMD




           POMA




           RB




           RESD




           RLA




           RVA




           SAU




           SM




           SP




           VEN




           VER




           WEST




           WHTR




           WNTT




           ZUM
                Location




 Mission Hills




 Mount Wilson




 Newhall




 Newport Beach




 Los  Alamitos Naval  Air Station




 Ontario International  Airport




 Pasadena - West  San Gabriel Valley




 Pico




 Palmdale Airport




 Pomona




 Redondo Beach




 Reseda




 Rancho  Los Amigos




 Rivera




 Saugus




 Santa Monica




 San  Pedro




 Venice




 Vernon




 West Los Angeles




 Whittier




Walnut




 Zuma Beach
88

-------
                             TABLE 5.2
                  TRAJECTORY IDENTIFICATION TABLE
Trajectory
Number
1
2
3
4
5
6
A
7
*
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
*
24
Date.
Sep 11
Sep 11
Sep 11
Sep 11
Sep 29
Sep 29
Sep 29
Sep 29
Sep 30
Sep 30
Sep 30
Sep 30
Oct 29
Oct 29
Oct 29
Oct 29
Oct 30
Oct 30
Get 30
Oct 30
Nov 4
Nov 4
Nov 4
Nov 4
Origin
Commerce
Commerce
Downtown Los Angeles
Downtown Los Angeles
Commerce
Commerce
Downtown Los Angeles
Near Coast
Commerce
Commerce
Downtown Los Angeles
Downtown Los Angeles
Downtown Los Angeles
Downtown Los Angeles
Commerce
El Monte
Pasadena
Commerce
El Monte
Downtown Los Angeles
Commerce
Commerce
Pasadena
Downtown Los Angeles
Start
Time
0530
0630
0530
0630
0530
0630
0530
0230
0530
0630
0430
0530
0530
0630
0630
0830
0530
0630
0630
0830
0530
0630
0530
0530
Closest
Terminal Station
Rancho Los Araigos
La Canada
West San Fernando
Valley
Mission Hills
Azusa
East San Gabriel
Valley
Walnut
Anaheim
La Canada
La Canada
La Canada
Burbank
Whittier
Walnut
Walnut
Azusa
Orange County
Long Beach
Orange County
Long Beach
El Monte
La Canada
La Canada
Mission Hills
Final
Time
1330
1330
1230
1330
1230
1330
1230
1230
1130
1130
1130
1030
1230
1330
1330
1330
1330
1030
1430
1230
1130
1230
1130
1430
Trajectories used in  transportation control strategy study.
                                                                        89

-------
                                TABLE 5.3
          INITIAL CONCENTRATIONS USED IN ATMOSPHERIC SIMULATIONS
Trajectory
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
CO, r
5.3
8
11
14
9
12
7
7
15.9
22
13
19
9
11
11
6
6
8
8.5
8
16
23.2
6
14
J-ilJ- L. J.CL-L. OULll
NO, pphm
25
17
10
14
39.9
43.5
16
2
67.5
70
35
38
5
25
30
12
21
48
23
13
36
72.6
3
35
-CllULclLXULlS
HC , pphm
60
60
25
25
67
77
40
18
117
129
45
65
30
30
24
30
30
63
27
25
100
130
25
40
NO , pphm
10
11.6
3
3
11.1
10.1
4
9
9.8
12
5
3
11
15
10
20
13
11
13.5
6
18
15.4
9
13
90

-------
      The test that is applied to the computed concentrations consists
of comparing the model's results with data which are obtained from air
monitoring stations.  This is a severe test inasmuch as an attempt is
being made to match computation and data in magnitude, time phasing, and
space.  Consideration of data base errors becomes important in the appli-
cation of such a test and these are discussed in the next section.

      For CO, the criterion of goodness of fit consists of matching the
observed concentrations with the computed results.  For the reactive
species, we focused on matching the computed ozone concentration with
the observations.  The close photochemical coupling among NO, NO , and
0- in many instances precluded a good match for all three species, since
a parameter adjustment made to improve the fit of one of the three would
degrade the match of the others.  (Such close coupling is obscured in
the averaged atmospheric samples used in the comparison process since
turbulence effects may interfere—see Appendix A.)  Since ozone is
generally considered to be the key indicator of photochemical smog, it
was felt that a good ozone fit at the possible expense of the others was
justified.

5.3.3  Data Base Errors
      As was mentioned above, the test of the model consists of comparing
the computed concentrations with observations along an air trajectory.
However, because of the sparseness of the air quality monitoring network,
the monitoring stations will not be on the path of the trajectory as a
general rule.  Hence, the data used for comparison must be obtained by
spatial interpolation (in pur case, we use inverse-distance weighting)
from those monitoring stations which are nearest to the trajectory's
nodes•  The model tests compare concentrations computed by the model
with concentrations which are interpolated from air monitoring data.
Therefore, the spatial interpolation itself can result in large errors
that bear no relationship to the validity of the model.  An estimate of
these errors can be obtained by performing the interpolation for a known
                                                                       91

-------
monitoring station and then comparing the actual and computed quantities.
Such an exercise was carried out using LAAPCD stations 1  (Downtown  Los
Angeles), 60 (East San Gabriel Valley), and 69  (East San  Fernando Valley)
for the measured data, and station 79 (West San Gahriel Valley)  as  the
point where the concentration is assumed unknown.  The results  for  ozone
showed that the relative error, i.e.,  true-computed|/true, ranged  from
6% to 63% with an average error of 35% for Sept. 29, 1969.  For  ozone on
Nov. 4, 1969, the error ranged from 0% to 33% with an average of 20%.
However, the absolute differences in concentration ranged from 0.3  to
25.8 pphm for Sept. 29 and from zero to 4.1 pphm on Nov.  4.  In  presenting
the results in graphical form as a comparison of model vs data it is the
absolute difference, rather than the relative error, which is immediately
apparent to the eye.  This is clear from Figs. 5.2 and 5.3, which show
the interpolated and actual concentrations for ozone on the two  days
mentioned previously.  The actual concentration is measured at station 79
(West San Gabriel Valley) and the stations used for interpolation are 1
(Downtown Los Angeles), 60 (East San Gabriel Valley), and 69 (East  San
Fernando Valley) .  Some sources of error in the interpolation are the
spatial inhomogeneities in source distribution, meteorological conditions,
and terrain.  All of these affect the production and flow of pollutants
in the area in question and thereby influence the pollutant levels.
Indeed, the appropriateness of assuming that the pollutant concentrations
at a point are representative of the levels in a region is the basic tenet
that comes under scrutiny when one considers the sources  of error in the
interpolation.  Mathematical models such as the one being evaluated in
this work may actually help to solve the problem of representativeness
by interpolating under constraints which take into consideration the
spatial inhomogeneities mentioned above.

      The interpolation discussed above was performed using inverse-
distance weighting.  The same calculation was performed using inverse-
distance-squared weighting and the two results were found to be  essentially
indistinguishable.   However,  it is unwarranted to draw general conclusions
on the basis of this limited test.
92

-------
   40
   30
                          OBSERVED AT STATION  79
                          (WEST SAN GABRIEL VALLEY)
LJJ
O

1
O
o
t-J
O
   20
   10
                                                /INTERPOLATED FROM STATIONS 1
                                                  (DOWNTOWN LOS ANGELES),
                                                  60 (EAST SAN GABRIEL VALLEY),
                                                  69 (EAST SAN FERNANDO  VALLEY)
   0600
0800
1000
1200
                                        TIME,  PST
 Figure 5.2.   Interpolated  and Observed Ozone Concentration  at a Monitoring
                Station in  the West San Gabriel Valley~on September 29,  1969
                                                                               93

-------
           20
                                  OBSERVED AT STATION 79
                                  (WEST SAN GABRIEL VALLEY)
                                       , ^INTERPOLATED FROM STATIONS 1
                                         (DOWNTOWN LOS ANGELES),
                                         60 (EAST SAN GABRIEL VALLEY),
                                         69 (EAST SAN FERNANDO VALLEY)
            0600
                          0800
                                         1000
                                    TIME, PST
                                                         1200
Figure 5.3.   Interpolated and Observed Ozone  Concentration at  a Monitoring
              Station in the West San Gabriel  Valley on November 4,  1969
 5.3.4  Assessment of Model Performance
       In this section we present  a general assessment  of the performance
 of  the model.  The results for  each trajectory are included in Sees. 5.3.5
 through 5.3.28, where each case is illustrated by a map  of the trajectory
 and three plots of computed and observed concentrations:   one for CO, one
 for NO and N02> and one for ozone.   It should be noted that results for
 reactive hydrocarbon are not shown.   This is due to the  paucity of hydro^
 carbon data available for comparison,  there being only three hydrocarbon^
 monitoring stations in Los Angeles County.
       One of the most important  factors  which affect the  performance 'of
 the model is the accuracy of the initial concentrations of  the three
 fundamental species NO, N02> and reactive hydrocarbon.-  The influence of
 initial  values is especially significant for a trajectory which starts
 around 0600 or later and runs for eight  hours or less.  In  such a case,
94

-------
the mass of the emitted pollutants Is considerably less than the initial
mass and as a consequence the computation is greatly influenced by the
initial values.  Most of the trajectories studied in this report fall
in this category.  One notable exception is trajectory number 8, which
starts at 0230 and has very low initial concentrations.  Availability of
hydrocarbon measurements constrains the selection of a starting location
since there are fewer hydrocarbon-monitoring stations than there are for
NO  or CO.  Hence we attempted to reduce the uncertainties in the initial
  X
conditions by starting trajectories at places of best-known hydrocarbon
levels.  In connection with the problem of uncertainty in the initial
values, we note that generally we seemed to obtain better results with
trajectories which started at Commerce than with those which began in
Downtown Los Angeles.  This could be attributed to the higher quality
of the Commerce data.

      The plots of ozone concentration shown in subsequent sections
illustrate that the computations matched the observations remarkably
well.  In order to achieve this, we had to settle for poorer fits for
NO and NO..  Usually, the model results for NO matched the data better
than the NO  predictions fit the NO, measurements.  (We recall that this
situation also prevailed, although to a smaller extent, in the simulation
of the smog chamber experiments.)  In several cases, early morning peaks
of NO were difficult to reproduce.  On the other hand, the NO  buildup
was generally accurate, but the decay was poorly reproduced, with the
N0? tending to linger at relatively high concentrations late in the day.
This behavior of NO- may be due to inadequacies in the kinetics or mixing
model for late-time NO  behavior.  Addition of the reaction of NO  with
particulates (reaction 2.16) Improved the late-time NO  decay, but the
improvement was small.  Increasing  k ,  is not- the answer to this problem,
because this Interferes with late-time  RO   control, with the result
being anomalously high concentrations of  R0? .  Additional research Is
needed to improve the late-time behavior of  N0?.
                                                                       95

-------
      In the simulation process, we had to reduce NO  fluxes consistently
 to 1/4 of the value estimated from source inventories.  This  adjustment
 is consistent with our previous work.   The necessity of scaling  down  the
 NO emissions arises from the fact that both the NO  balance and the  ozone
                                                  X
 production diverge greatly from the observed values when the  full NO flux
 is used;  Reducing the NO flux to 1/4 of its full value results in pre-
 dictions which fit the data much more accurately.  As discussed previously,
 these flux reductions may reflect atmospheric loss mechanisms  such as
 surface reactions that are as yet unidentified in any of the  field programs.

 Statistical Correlation of Computations and Observations
      One way of evaluating the aggregate performance of the model is
 to measure the correlation between computations and observations.  This
 contrasts with the case-by-case presentation of results contained in the
 next  several  sections.  By examining the relationship between  predictions
 and measurements in a highly aggregated form, it is easy to discern  trends
 in the model's performance.  These trends may be useful in obtaining
 correction factors to improve the predictions.  Naturally, the initial
 concentrations have been excluded from the statistical analysis since
 their inclusion would bias the correlations.
       Correlation coefficients for computed and observed concentrations
 were  obtained  for CO and ozone for the set of hands-on cases, the set of
 hands-off  cases, and for both sets together.  Table 5.4 shows the values
 of  the coefficients for each of the three groupings.  It is clear from

                                TABLE 5.4
                CORRELATION COEFFICIENTS FOR CO AND OZONE
        Species       Hands-on       Hands-off       Composite
         CO             0.90            0.63            0.80
         Ozone          0.94            0.88            0.92
96

-------
the table that  the  coefficient for the hands-on cases provides an upper
                            J!
bound of the expected  performance of the'model, and the coefficient  for
the hands-off cases, a lower  bound.   For ozone, the difference between
the upper and lower bounds  is small, thus indicating a good performance
in either situation.   The difference is greater for CO, however, indicating
a need for additional  adjustment in diffusivity parameters in the hands-off
cases.  It is interesting to  note that the lowest correlation coefficient
obtained here for CO  (in the  hands-off cases) matches the highest corre-
lation coefficient  obtained by other investigators using another CO
diffusion model which  employs a combination of Gaussian plume and box
  A 1  36
models.

      Figures 5.4 and  5.5 are scatter diagrams of observed vs predicted
concentrations  of CO and ozone, respectively.  The results of all 24
trajectories are contained  in these graphs.  The figures also include
plots of the least-squares  regression line.
               30.00 -i
               25.00 -
               20.00 -
           o
           o
           CO
           CD
           O
               15.00 -
               10.00 -
               5.00 -
               0.00
                  O.QO
                         5.00    10
                                .00    16.00    20.
                                COMPUTED CO tPPH)
                                            20.00    26-00    33.00
   Figure 5.4.  Observed  Versus Computed Carbon Monoxide Concentration.
                (Number of Points  =  149)
                                                                        97

-------
10-00 -
35-00 -
30.00 -
| es.00-
Q_
O
g 90.00 -
0
UJ
Of.
UJ
CO
oa 15.00 -
o
10.00 -

5.00 -





0
0 ° t
0 % ffl y
oo f o
ff1 a/*o
              I 00
        o-oo -f	r	i	1	:	1	1—'-	r~	-i	r
           0.00     5.00    10.00    15-00    20-00    25.00    30.00     36-00    40.00
                               COMPUTED OZONE (PPHMJ
       Figure  5.5.  Observed Versus  Computed Ozone Concentration.
                    (Number of Points =  151)
      It is apparent from Fig. 5.4 that for CO  the model tends to over-
estimate the low concentrations; these generally  occur  in the afternoon.
The CO peaks, on the other hand, are underpredicted.  For ozone, Fig. 5.5,
the model tends to underestimate the low concentrations but to be accurate
at medium levels.  High ozone concentrations  are  slightly overpredicted.

      The regression lines shown on Figs.  5.4 and 5.5 can be usec>t:o
correct the predictions and thus obtain a  better  estimate of the actual
concentration.  Table 5.5 shows the equations of  the regression lines for
CO and ozone, together with the standard error  of estimate.  If we ask
what the actual concentration,  y  , is likely to  be  given a predicted
concentration  y. , we can obtain an answer from the  regression equation.
The value of  a  , the standard error of estimate, provides a measure of
the tolerance which may be assigned to the corrected prediction.
98

-------
                       TABLE 5.5

         REGRESSION EQUATIONS FOR CO AND OZONE

             Regression Equation
                y = observed           Standard Error
Species         x = computed          of Estimate  (a)
 CO          y = 1.007'x - 0.346          3.663 ppm

 Ozone       y = 0.840x + 2.307          2.109 pphm
             2
    Cy'  - Y )  > y  = observed concentration
                 y  = concentration computed using regression
                      equation
                                                              99

-------
    5.3.5   Trajectory  No.  1,  September 11.  1969.  Starts at Commerce at 0530
           (Hands-Off)
          The  simulation of  CO agreed very  well with the data.  However, the
    results for  the  reactive  species  matched the  data poorly, with the  NO
    balance predicted  by the  model  being greater  than is shown by the data.
    Also,  Fig. 5.9 shows that the predicted ozone concentration, denoted by
    the solid  line,  is considerably higher  than is indicated by the data.
    This computation was done using the clear-day value of  k^ .  However,
    examination  of Fig. 5.102 (Sec. 5.5)  shows that for Sept. 11 the value
    of  k..   at Commerce is much lower than  the clear-day value.  Since the
    trajectory meanders around Commerce for most  of the day, it is possible
    that the low values of ozone may  be due to the low  k..  .  Substituting
    the Commerce k..   for  the clear-day  k..   yielded the dashed curve shown
    in Fig. 5.9.  The  lover   k..   produced much lower values of ozone, but
    the relatively small ozone peaks  shown  by the data were still not repro-
    duced  by the computation.   Nevertheless,  the  dramatic reduction in ozone
    obtained using .the new k1   is  indicative of  the large variations to
                          -'   J_
    which  the  predictions  are subject due to  the  uncertainty in the fundamental
                                                   3    —1   —1
    input   k^  .   Finally,  in  this case  k,  =  4 x  10  ppm  min

                                                      •LACA 108W          is
                                              r nn ~rc\ i  0830
                                              CAP 75W   .   ]030
                                                            79 WSGV
                                                            PASA new
Figure 5.6.   September 11, 1969 Trajectory Starting at Commerce at  0530  (No.  1)
   100

-------
                                                         	 MODEL RESULTS    Łj
                                                         O  INTERPOLATED     3
                                                            STATION DATA     t
    Figure 5.7-   Trajectory No.  1—Computed and Observed CO Concentrations
                                          MODEL RESULTS	—

                                        INTERPOLATED STATION - NO  O
Figure  5.8.   Trajectory No. 1—Computed and  Observed NO and NO  Concentrations
  Figure  5.9.  Trajectory No.  1—Computed  and Observed Ozone Concentrations
                                                                              101

-------
 5.3.6  Trajectory No.  2,  September 11,  1969, Starts^t^Conimerce at Q6.3Q
        (Hands-Off)
       Figure 5.11 shows that the early-morning CO buildup and decay are
 well reproduced by the model.   The low  CO concentrations in the afternoon
 are overestimated, however.   (There is  no data point at 1230 because it
 was missing from the observations.)  On the other hand, Figs. 5,12 and
 5.13 show that the predicted concentrations of the reactive species match
 the data very well.   The  computed  NO   balance is very good and the ozone
                                      X                     3-1   -1
 prediction is superior.   For this trajectory,   k,  = 4 * 10  ppm  min
LACA 108 W
    *   1330
                                                                        tx
                                                                        UQ
                      69 E SFV
                                         1230
                                          1130
                       60 E SSV
                                                 ELM
                                  BKTD
                          1  CAP  75  W
                            0930
                                      1030
                                     0630 COM
                          0830   0730     .RVA 81 W
                  LA INTERNATIONAL
                      AIRPORT
                                             KFID
    Figure 5.10.  September.il, 1969 Trajectory Starting at Commerce
                  at 0630  (No. 2)
102

-------
                                                           ___ MODEL RESULTS
                                                           O  INTERPOLATED
                                                              STATION DATA
                                                                       1430
     Figure  5.11.  Trajectory No. 2—Computed and Observed  CO Concentrations
                                                     MODEL RESULTS .
                                                  INTERPOLATED STATION - NO O

                                                               NO, A
Figure  5.12.   Trajectory No.  2—Computed and  Observed  NO and  NO  Concentrations
                             —— MODEL RESULTS

                              O  INTERPOLATED STATION DATA
  Figure 5.13.   Trajectory No.  2—Computed and  Observed Ozone  Concentrations
                                                                                  103

-------
  5.3.7  Trajectory No. 3, September 11, 1969, Starts in Downtown Los Angeles

         at 0530 (Hands-Off)


        Figure 5.15 shows that the simulation of CO matched the data well.


  However, the  NO   computation does not match the data, the peaks of NO
                  X                  i;

  and NO- being much lower than the observations shown in Fig. 5.16.  On


  the other hand, the ozone simulation is remarkably accurate, despite the


  poor quality of the  NO   predictions.  This case is one example where
                         X

  parameter adjustments made to improve the  NO   computations degrade


  the ozone and vice versa.  The value  of  k,   used in this case is

        3    -1   -1                        4
  4 x 10  ppm  min
                              1230
                    74 W SFV»


                       86 W ENC
    41 WBURK    .108 W LACA


1130 ^V69 E SFV


                        • 79  W SGV
                                                   0730

                                    27 W HOL«     V0630


                                                    0530

                                                1  DOLA
                                                        >89 COM
Figure 5.14.   September 11,  1969  Trajectory  Starting  in Downtown Los Angeles

              at 0530 (No.  3)
  104

-------
                            	MODEL RESULTS
                            O  INTERPOLATED STATION DATA
                                             0930
                                           TIME (PST)
                                                          1130
                                                                        1330
    Figure 5.15.   Trajectory No.  3—Computed and Observed CO Concentrations
               S 20
               S 15
                     MODEL RESULTS
                        INTERPOLATED - NO O
                        STATION DATA  .... A
g
                                             0930
                                           TIME (PST)
Figure  5.16.   Trajectory No.  3—Computed and Observed NO and N0_  Concentrations
  Figure 5.17-   Trajectory No.  3—Computed and Observed Ozone Concentrations
                                                                                  105

-------
  5.3.8  Trajectory No.  4,  September  11,  1969,  Starts in Downtown Los
         Angeles at 0630 (Haads-Off)
        The computed CO,  Fig.  5.19, agreed with the data to within 3 ppm,
  except for the last point which  is  about 5  ppm higher than the data.  As
  in the previous trajectory,  the  computed NO    balance is lower than indi-
                                              X
  cated by the data, but  this  time the  NO predicted by the model matches
  the data well.  Figure  5.21  shows that  the  computed ozone is generally
  lower than the measurements, but the  differences  are certainly within the
  error margin of interpolation and experimental uncertainty.   The value
                                                  3    -1   -1
  of  k   used in this simulation  was   k,  = 4 x 10   ppm  min
                        SAUD
                                 i MISH113W
                                                 LACA 108W
                                       69 E SFV
                                              1030
                                               0930   .79WSGV
                                               0830
                                             0730

                                                 • COM
BKTD
                                                           DKFI
Figure 5.18.  September 11, 1969 Trajectory Starting in Downtown  Los  Angeles
              at 0630 (No. 4)
 106

-------
                  20
               i  15
                        	 MODEL RESULTS

                         O  INTERPOLATED STATION DATA
                                             1030

                                            TIME (PST)
     Figure 5.19.   Trajectory  No. 4—Computed and Observed  CO Concentration
                      MODEL RESULTS  ' ' ' '
                        INTERPOLATED - NO
                        STATION DATA
Figure 5i20j   Trajectory No.  4—Computed  arid Observed  NO and  N0? Concehtratitins
               _  i's -
                            	 MODEL RESULTS

                             O  INTERPOLATED STATION DATA
    Figure  5*21,   Trajectory No1.  4—Computed  and Observed  Ozone  Concentrations


                                                                                    107

-------
   5.3.9  Trajectory No. 5, September 29, 1969, Starting at Commerce at
          0530 (Hands-On)
         This trajectory has the characteristic that it stays in the vicinity
   of Commerce most of the time from 0530 to 1030.  This gives us a greater
   degree of confidence in the data obtained by interpolation from the various
   stations along the way.

         The CO simulation shows a low peak, but the remaining concentrations
   are predicted accurately.  The simulation of the reactive species shows
   good agreement with the data, especially for the NO and ozone.  The NO
   buildup is accurately predicted, but the NO  decay fits the data poorly.
                                     3    -1   -1
   The value .of  k.  used was  8 x 10  ppm  min
                                           79 WSGV
                                                         1230,
                                                   1130/»ELM
                                                                   60 ESGV

                                          COM
0730*f»RVA
0830^0930
   •
  RLA
                                                          ®80 SE
                                                          DKF1
Figure 5.22.   September  29,  1969  Trajectory  Starting  at  Commerce  at  0530 (No.  5)
   108

-------
                             O   INTERPOLATED STATION DATA
                    Ol	
                    0530
                                 0730            0930
                                      TIME (PST)
     Figure  5.23.  Trajectory No. 5—Computed and Observed  CO Concentrations
                                            MODEL RESULTS 	
                                            INTERPOLATED STATION DATA  D NO
                                                            A NO,
Figure 5.24.   Trajectory No.  5—Computed and  Observed NO and N0? Concentrations
                             MODEL RESULTS 	
                             INTERPOLATED STATION DATA O
  Figure  5.25.  Trajectory No. 5—Computed and  Observed  Ozone  Concentrations
                                                                                  109

-------
 5.3.10  Trajectory No.  6, September 29, 1969, Starting  at  Commerce at
         0630 (Hands-On)               ~~	  --_          -
       This trajectory is similar to the previous one.   This  time,  however,
 the computed CO buildup and peak match the data closely.   The predicted
 CO decay is accurate until 1030.  The model overprediets  the 1130  and
 1230 CO concentrations  but at 1330 the approximation is good.
       The reactive species, Figs. 5,28 and 5.29, show a, behavior similar
 to  that  in trajectory no.  5:  the NO .and ozone curves are very  well
 modeled,  but  for N0~  only  the buildup is close to the data.   The value
                    ~      -i    	i
 of   k.   was  8  x 1CH  ppm  min
                                          79 WSGV*
                                                       BOESGVi
ZUM
                                          0830
                                                               1330
                                                                      DBKT
DOLA
     COM
VER«  \—  An 30
   O730t   XRVA
          1030
                                                                          50


                                                                          •=C
                                                0930   »80 SE WHTR
                                                      n KFI
                                            • 72 SC'LONB
     Figure 5.26.  September 29, 1969 Trajectory Starting at Commerce
                   at 0630  (No. 6)
110

-------
                                             MODEL RESULTS 	TTT-
                                             INTERPOLATED STATION DATA O
                                 0830           1030
                                      TIME (PST)
                                                            1230
     Figure 5.27.  Trajectory  No.  6—Computed  and Observed GO  Concentrations
                                                          q NO
                                      TIME (PST)
Figu.r-e 5,. 28.   Trajectory No.  6--Computed and Observed NO and N0_  Concentrations
    ..... ^ ^- -           .   _       -                                        2-
                                      TIME (PST)
                                                            1230
          5,.29;.   Trajectq.ry No.  6--Computed a,nd Observed O.zone Concentrations
                                                                                  111

-------
  5.3.11  Trajectory No,  71_Segjteinber_29J  1969,  Starting in Downtown Los
          Angeles at 0530 (Hands-on)
        The results obtained for  CO   are  shown in Fig. 5.31.  The reported
  0530 concentration at the downtown  station was 3 ppm.  Using this initial
  value yielded very low concentrations throughout the whole day.  In order
  to obtain a better approximation to the  data,  the initial value was
  adjusted to 7 ppm and this yielded  improved results.  However, the pre-
  dicted peak value of 12 ppm was still 4  ppm lower than the data indicate.
        Figure 5.32 shows that the predicted  NO   balance diverges con-
  siderably from that shown by the data.   In spite of this, the simulation
  of ozone was very accurate.   We might add that no adjustment of the initial
  concentrations was necessary for simulating the reactive species.  The
  value of  k,  was  5 x 10  ppm  min
                                                      60 ESGV
                                                         m
                                                 >ELM
                           0530f DOLA
                                                •80 SE
                                                                    75 EGA
                                                               • WNTT
                                                                         10
                                                                         Ki
                                                                         IX
Figure 5.30.   September 29,  1969 Trajectory Starting in Downtown Los Angeles
              at 0530 (No.  7)
 112

-------
 s '"
 i
                           MODEL RESULTS ——

                           INTERPOLATED STATION DATA O
               0730          0930

                    TIME (PST)
                           J	1	I	I
Figure 5.31.   Trajectory No. 1—Computed and

                Observed  CO Concentrations
                                                                                               MODEL RESULTS

                                                                                               INTERPOLATED STUTIOK DATA AHO,
                                                           Figure  5.32.  Trajectory  No.  7—Computed and

                                                                          Observed NO and NO   Concentrations
      Figure 5.33.   Trajectory No.  7—Computed and
                     Observed Ozone  Concentrations

-------
5.3.12  Trajectory No. 8, September 29, 19,69,  Starting Near  Coast  at 0230
        (Hands-on)
      This trajectory was reverse-calculated from  the Anaheim station at
1230.  The computation of pollution concentrations  commenced at  0230 near
the Lennox station of the LAAPCD.  As was the  case  with  the  previous tra-
jectory, we could not fit the  CO  data using  the  reported initial concen-
tration of 3 ppm.  Adjustment of the initial   CO   to 7 ppm produced' satis-
factory results from 0530 to 1130, as is shown in Fig. 5.35.
      As with  CO  , the initial values of  NO   and hydrocarbon were
 adjusted upward by the ratio 7/3.  However, the original initial values
 were very  low and  the adjustment did not cause large changes  in the simu-
 lation.  Figure 5.36 shows that the  NO  simulation does not  agree well
 with the data.  It should be noted that the  NO  data show a  second peak
 at  0930 which is suspect.  The computed  NO. , on the other hand, shows
 close agreement with the data after 0630.  For ozone, Fig. 5.37 shows
 that the simulation and the data are very closely correlated.  The value
                     4    -1   -1
 used for   k,  was  10  ppm  min
                                                  80 SE
fe;
                                                              1230
                                                              ANAHEIM
                                                    1030  1130
                                                          ORANGE CO.
                                                          AIRPORT
   Figure 5.34.   September  29,  1969  Trajectory Starting Near the Coast
                 at 0230 (No.  8)
114

-------
                         S ,0
                         s
                            0  O
                                                    MODEL RESULTS	
                                                    INTERPOLATED STATION DATA O
                                       0630          1030
                                            TINE (PST)
     Figure  5.35.   Trajectory No.  8—Computed and Observed CO  Concentrations
                                           MODEL RESULTS 	
                                           INTERPOLATED STATION DATA O NO
                                                           AND,
                            0230     0430
                                          0630    0830
                                            TIME (PST)
                                                         1030     1230
Figure 5.36.   Trajectory No.  8—Computed and Observed  NO and N0_ Concentrations
                                  MODEL RESULTS 	
                                  INTERPOLATED STATION DATA O
                                        SUNRISf
                            0230.     0430     0630 :    0830     1030      1230
                                             TIME, PST
    Figure 5.37.   Trajectory No.  8—Computed and Observed  Ozone Concentrations
                                                                                        115

-------
 5.3.13  Trajectory  No.  9,  September 30, 1969, Starting at Commerce at 0530
         (Hands-on)
       Figure 5.39 shows that the predicted  CO  fits the data well,  with
 the exception  of the  data  point at 0930.  In addition, the  computed  con-
 centration  at  0630  is low  by about 6 ppm.  Figure 5.40 illustrates that
 the predicted  NO   matches the data closely but that the  N02   is  con-
 siderably overestimated.   This situation is typical of the  unsatisfactory
 NO   balances  encountered  in the simulations.  The computed ozone  (Fig.
 5.41) is seen  to underestimate the early-morning concentrations and  then
 overshoot the  1130  value by 8 pphm.   We note that the relatively high
 ozone concentrations  indicated by the observations before 1030  are puzzling
 inasmuch as very little NO -> NO   conversion has taken place prior  to
                                 / ^   	-i    __ -I
 1030.  The  value of  k  is  10  ppm  min
                      SAUD
                                • MISH
                                        1130
                                    BUR*
                                     BURK«
   • LACA
r!030
 .0930    «PASA
                                     HOL«
                                WEST
                                         DOLA
                                              0730
                                                 0530 COM
                                 LAX<
                                    LENX
                                                                     BKTD
                                                       KFIO
     Figure 5.38.  September 30, 1969 Trajectory  Starting at Commerce
                   at 0530 (No. 9)
116

-------
      30r
      i\-      	MODEL RESULTS
              O  INTERPOLATED STATION DATA
      0530

Figure 5,39.
         TIME (PST)
                                 0930
                                               1130
Trajectory  No.  9—Computed  and
Observed CO Concentration
                     	MODEL RESULTS
                     O  INTERPOLATED STATION DATA
                                                                                     0730           0930
                                                                                          TIME (PST)
                                                                                                                1130
                                                                    Figure 5.40.   Trajectory No.  9—Computed  and
                                                                                    Observed NO  and N02  Concentrations
                                                       -Figure 5.41.   Trajectory  No. 9—Computed  and
                                                                       Observed Ozone Concentrations
                         TIME (PST)

-------
  5.3.14   Trajectory  No.  10,  September 30, 1969, Starting at Commerce  at
          0630  (Hands-on)                              \ ;
                                                      i
       This  trajectory  follows  approximately the same'path as trajectory
  No.  9.   However,  this  time  the early-morning values of  CO  are reproduced
  very well,  as  can be seen in Fig.  5.43,   As was the case with the previous
  trajectory, the model  is unable to follow the sharp drop in  CO  concen-
  tration  which  occurs- at 0930.   For  NO   it can be seen in Fig. 5.44  that
  the  peak is too high but the decay is fairly accurate.   The  NO   buildup
  is well  reproduced, but the  decay  is very poor.   For ozone, Fig. 5.45,  it
  can  be seen that  once  again  the low concentrations are underestimated.
  In contrast with  the previous  case,  the  model is  very accurate at the
                                                4     -1   -1
  higher ozone levels.   The value of  k.   is  10 ppm  min
                                                  1130
                                RESD
                                 •
                                         69 E SFV«
                                               09 30 \
                                        HOL 27 W«
          to
          LO
          tx
          fo
          I
                                                      LACA
79 W SGV
PASA
                                                 08303
                                                 DOLA<
                                                  0730' ^« 0630 COM
                                                      34 W
     Figure 5.42.   September 30, 1969 Trajectory Starting at Commerce
                   at 0630 (No. 10)
118

-------
                          MODEL RSSULTS	
                          INTERPOLATED STATION DATA O
                      TIME (PST)
Figure 5.43.
Trajectory  No. 10—Computed
and  Observed CO Concentrations
              	MODEL RESULTS
               O INTERPOLATED STATION DATA
                                                                                           MODEL RESULTS 	
                                                                                        INTERPOLATED STATION - NO O
                                                                                        DATA           „„ A
                                                                        TIME (PST)
                                                                Figure  5.44.   Trajectory No.  10—Computed and
                                                                                Observed  NO and N0?  Concentrations
                                                     Figure 5.45.   Trajectory  No. 10—Computed and
                                                                     Observed Ozone Concentrations
                      TIME (PST)

-------
  5.3.15  Trajectory No. 11, September 30, 1969, Starting in Downtown Los
          Angeles at 0430 (Hands-On)
        Figure 5.47 illustrates that the  CO  peak is about 6 ppm too low.
  This may indicate a deficiency in the strength of the emissions since  the
  diffusivities were set at the lowest value used for very stable atmos-
                              3   2
  pheric conditions,  2.5 * 10  cm /second .  The model is accurate from
  0730 to 0930, but does not reproduce the apparent peak after 0930.  The
  reactive species are well modeled.  This time the simulated  NO   balance
  is  excellent wi.th the exception of one very high observed concentration
  of  NO   at 1130.  The ozone is only slightly underestimated throughout
        X                                                    4-1-1
  the trajectory.  The value of  k,  used in this case is  10  ppm  min
                                  69  E SFV»
                          71 NWC
                                             1130  »LACA
                                                                           00
                                                         '79 U SGV
                                   0830
                                  LAX   0730
     0430 DOLA
0530 .  »COM
     VER
Figure 5.46.   ^^30,^1969 Trajectory Starting in Downtown Los Angeles
 120

-------
                                 	MODEL RESULTS
                                  O INTERPOLATED STATION DATA
                                              0830
                                            TIME (PST)
     Figure 5.47.   Trajectory No.  11—Computed  and Observed CO  Concentrations
                                            MODEL RESULTS 	
                                          INTERPOLATED STATION - NO O
                                          DATA           N02 A
Figure 5.48.   Trajectory No.  11—Computed  and Observed NO  and N0_  Concentrations

                                                                  o
                              —— MODEL RESULTS
                               O INTERPOLATED STATION OATA
  Figure  5.49.   Trajectory No.  11—Computed and  Observed Ozone Concentrations
                                                                                   121

-------
  5.3.16  Trajectory No. 12, September  30,  1969,  Starting in Downtown  Los
          Angeles at 0530  (Hands-On)
        The  CO  simulation shown  on  Fig.  5.51 indicates good agreement
  between prediction and observation, with the lower concentrations being
  overpredicted.  On Fig.  5.52, it  can  be  seen that the predicted  NO  con-
  centrations are low but  that the  NO   computations match the observations
  accurately.  The ozone,  Fig. 5.53,  is underestimated throughout most of
  the morning, but the error is small.   It  should be noted that the observed
  ozone concentrations which occur  from 0730  to 0930 are difficult to recon-
  cile with the high  NO   concentrations indicated by the data.   For  k. ,
                        4-1-1
  the value used was  10   ppm  min
                       SAUD
                           RESD
                                    BUR*
                                    69 E SFV

                                  HOL 27 W»
                                           •1030 »LACA
t0930
  0830
 0730
 0630
   0530
DOLA
     • COM
•79WSGV DASA
                                                        KFID
                  BKTQ
Figure 5.50.  September 30, 1969 Trajectory Starting  in Downtown Los Angeles
              at 0530 (No. 12)
 122

-------
                                                      — MODEL RESULTS
                                                      Q  INTERPOLATED STATION
                                                         DATA          :
                                       0730           0930
                                            TIME (PST)
    Figure 5.51.   Trajectory  No. 12—Computed  and Observed CO Concentrations
                                              MODEL RESULTS
                                            INTERPOLATED STATION - NO O

                                            DATA           NO, A
                                                                  1130
                                            TIME (PST)
Figure 5.52.  .Trajectory No.  12—Computed and Observed NO and NO- Concentrations
                                       0730            0930
                                            TIME (PST)
   Figure  5.53.   Trajectory No.  12—Computed and Observed Ozone Concentrations


                                                                                   123

-------
 5.3.17  Trajectory No.  13,  October 29,  1969,  Starting in Downtown Los
         Angeles at 0530 (Hands-On)
       Figure 5.55 shows that the simulated  CO  approximates the data
 closely.   However, from Fig.  5.56 we can see  that the  NO   simulation
 diverges  considerably from  the data.  Despite the low quality of the  NO
 simulation,  the computed ozone fits  the data closely.   This is, of course,
 the result of our emphasis  on obtaining accurate ozone predictions.  For
 this trajectoryj  the value  of  k  used was   4  x 10  ppm  min
                                                              '60 ESGV
                                                                            2
                                                       ELM
                                                                         D
                71  NWC
                                0530
                                     1 DOLA
,0930
Figure 5.54.  .October  29,  1969  Trajectory  Starting  in Downtown Los Angeles
              at 0530  (No.  13)
124

-------
                                                        —- MODEL RESULTS
                                                        O  INTERPOLATED STATION
                                                           DATA
                       0530
                                     0730          - 0930
                                          TIME (PST)
     Figure  5.55.   Trajectory No.  13 — Computed and Observed CO  Concentrations
                               MODEL RESULTS -
                            INTERPOLATED STATION NO O
                            DATA          N0 A
                                                  0930
                                             TIME (PST)
Figure 5.56.   Trajectory No.  13—Computed  and Observed  NO and NO  Concentrations

                                                                       o
                                                               1130
                                                                             1330
   Figure 5.57.
                               0930
                             TIME (PST)
Trajectory  No.  13—Computed and Observed  Ozone  Concentrations

                                                                125

-------
 5.3.18  Trajectory No.  14, October 29, 1969, Starting  in Downtown Los
         Angeles at 0630 (Hand's-On)
       From Fig. 5.59 it can be seen that the  CO   simulation fits the
 data fairly closely.  The  CO  peak produced by the model is low by 3 ppm.
 The early-afternoon concentrations of  CO  are overpredicted.
                                       is reproduced very  accurately.   On
                                                                   NO,,   peak
      Figure 5.60 shows that the  NO
the other hand, the simulated  ">->   v»j.v«_i.gwo  j-j-^m ^n^ uu>.u, !_,.!,_  '••^o
being low by about 4 pphm and occurring an hour  later than indicated by
the data.  Figure 5.61 shows that the modeling of ozone is remarkably
NO   diverges from  the  data,  the
 accurate.   For this trajectory, the value of  k.  used was
                                                             1fl4    -1 .  -1
                                                             10  ppm  mm
                            DOLA
                               0630
                                               • ELM
                                                                  DBKT
                                                                    •POMA
                                                    230  1330
                                                              i VINTt
                                                                          01
                                                                          !~H
                                                                          to
                                                                           fe:
  Figure 5.58.   October 29, 1969 Trajectory Starting  at  0630 in Downtown
                Los-Angeles (No. 14)
126

-------
IS3
—I
                18r     O
                                             MODEL RESULTS 	
                                             INTERPOLATED STATION DATA O
                0630     0730    0830
                                    0930     1030
                                      TIME (PST)
                                                 1130     1230     1330
             Figure 5.59.   Trajectory No.  14—Computed and
                             Observed CO Concentrations
                                                                                         0830           1030
                                                                                              TINE (PST)
                                                                         Figure  5.60.  Trajectory  No. 14—Computed and
                                                                                        Observed NO and N0_ Concentrations
                  Figure 5.61.   Trajectory No.  14—Computed and
                                  Observed  Ozone  Concentrations
                                                                                         0830            1030
                                                                                              TIME (PST)

-------
 5.3.19  Trajectory No.  15,  October  29,  1969,  Starting at Commerce at 0630
        (Hands-On)
      The   CO  simulation shown  in  Fig.  5.63  is  especially interesting be-
 cause the double peak shown by the  data is  reproduced by the model.  The
 maximum difference between  data  and simulation occurs near the end of the
 trajectory  and is approximately  2.5 ppm.  This could be due to greater
 dispersion  near the mountains.

      The photochemical simulations (Figs.  5.64  and  5.65)  show very good
 modeling of NO  and  0  .  The  simulated   N0?  peak is about 3 pphm too
 low and occurs one hour later than  is shown by the data.   Inaccuracies in
 the NO  emissions inventory may account for  this.   The computed end value
 of   NO   is 21 pphm compared to  about 7 pphm  for the data.   The high ter-
 minal values of  NO   appear to  be  a recurring problem in  atmospheric
 photochemical simulations.  It is likely that  this problem arises from
 departures  from expected photo-chemical equilibrium  conditions (see
 Appendix, Sec. A.2.2).  The value of  k.  used in this  trajectory was
 10   ppm  min
                                     • PASA
                                0730
                        DOLA«
                             0630
                           VER»COM
                             0930
                                            • ELM
                                                    >AZU
1330
         DBKT
            • POMA
     >WNTT
                to
                CO
      Figure  5.62.  October  29, 1969 Trajectory  Starting at Commerce
                   at 0630  (No. 15)
128

-------
Figure 5.63.   Trajectory No.  15—Computed
               and Observed CO
               Concentrations
                                               — 25
                                               S
                                               5 15
                                                      MODEL RESULTS 	•	
                                                      INTERPOLATED STATION DATA OHO
                                                                     1030
                                                                  TIME (PST)
Figure 5.64.   Trajectory No.  15—Compute
               and Observed  NO and N0?
               Concentrations
    Figure 5.65.  Trajectory No.  15—Computed and  Observed Ozone  Concentrations
                                                                              129

-------
5.3.20  Trajectory No. 16, October  29,- 1969,  Starting at El MQnte at 0830
      This trajectory shows  a  reversal in  direction of almost 180° at 1030,
This may be seen in Fig. 5.66.  The high value  obtained from the data at
1030 is mostly due to high concentrations  reported  at the downtown station
(1DOLA).  In view of the reversal in direction,  it  is likely that this
interpolated quantity is badly off  the mark.

      The high values shown  by the  data at the  end  of the trajectory could
not be reproduced by the simulation since  the trajectory enters some areas
with no sources at all.  Location of the sampling sites can induce large
deviations from the average  for the air mass.   The  ground value will con-
tinue to decrease due to diffusion  when source  strengths are small.

      Figure 5.68 shows that the simulation of   NO   is very accuratej  but
that for  N0?  the model produced very poor results.   As usual, the  ozone
simulation, shown in Fig. 5.69, fits the data closely.   The value used
for  k   was  10  ppm  min
                                          PASA           A^U
                                       1030><-^ 1130
                                                ELM_   ^X<1330     BKT

                                                                   POMA
0930,  ™°Xf23o"""'     °
                                                                      to
                                                                      to
                                                                      I
     Figure 5;66.  October 29, 1969 Trajectory  Starting at El Monte
                   at 0830 (No. 16)
130

-------
            M8DEL RESULTS-
            INTERPOLATED STATION DATA O
                                  1230
                         TIME (PST)
Figure 5.67.  Trajectory No.  16—Computed
                and Observed CO Concentrations
                                                                        0830
                                                                                                 MODEL RESULTS -^——
                                                                                               INTERPOLATED STATION - NO O
                                                                                               OAT*           „„. „.
                                                                                      1030           1230
                                                                                           TIKE (PST)
                                                                    Figure  5.68.
                Trajectory No.  16—Computed
                and Observed NO and  N02
                Concentrations
              — MODEL RESULTS
              O INTERPOLATED STATION DATA
    0830

Figure 5.69.
                                                                                          TIKE (PST)
                                                                                    Trajectory No.  16—Computed
                                                                                    and  Observed Ozone Concentrations

-------
 5.3.21  Trajectory No. 17, October 30. 1969.  Starting at Pasadena at 0530
         (Hands-Off)
       The trajectory was started at 0730 southeast of Pasadena because all
 of the  CO  data for Pasadena are missing  for this date.  As a result, the
 initial values used for the various species were  obtained by interpolation
 from neighboring stations.

       The results for  CO  show good reproduction of the buildup phase
 from 0730 to 0930.   The decay part of the  concentration curve has the
 right shape, but the concentrations are about 3 ppm too high.  The low
 points in mid-morning measurements could be due to horizontal intrusions
 of air directly from the ocean.
       The photochemical results show good  NO  and  0    simulations.  The
 NO   buildup is  well modeled, but the
                                       NO.   decay shows a very poor fit
                                        NO   interpolated  values may be
of the data.  The late morning dip of
due to the same dilution mechanisms suggested  above for the  CO  points.
The value used for  k.
                         was  10  ppm  min
                                      79
                                 0530 .PASADENA     • 60

                                    0630^
                                          • EL MONTE
                                                        DBKT
                                                 LA COUNTY (_f—'
                                                         /
                                     l» KVA               ^
                                    10930 WHTR  I—*	!
                                RLA« \    •   j LA HABRA
                        COMPTON •
                                                  ANAHEIM

                                                  1330
                                                    ORANGE COUNTY
                                                      AIRPORT
                                                                  to
                                                                  to
      Figure 5.70.   October 30, 1969 Trajectory  Starting at Pasadena
                    at 0530 (No. 17)
132

-------
UJ
u>
                 0730
                               0930           1130
                                    TIME (PST)
              Figure  5.71.   Trajectory No.  17—Computed
                              and Observed  CO Concentrations
                                                                                        -MODEL RESULTS
                                                                                     N02 A INTERPOLATED STATION DATA
                                                                                     NO O
                                                                                               0930           1130
                                                                                                     TIME (PST)
                                                                              Figure  5.72.
Trajectory  No. 17—Computed
and  Observed NO and N0?
Concentrations
                                                                                      INTERPOLATED STATION DATA
                                                                                      MODEL RESULTS
                                                                                               0930            1130
                                                                                                    TIME (PST)
                                                                              Figure 5.73.
Trajectory  No. 17—Computed
and  Observed Ozone
Concentrations

-------
 5.3.22   Trajectory No.  18,  October  30,  1969,  Starting at Commerce at 0630
         (Hands-Off)            	""	  	""" "
      This  trajectory  is  rather  short,  only  4 hours  long, and is interest-
 ing  because it  travels  toward  the coast early in-the morning.

      Figure 5.75 illustrates  the simulation  of   CO  ,  The time phasing
 and  shape of the curve  agree with the data, but  the  predicted  CO  peak
 is about 2.5 ppm lower  than the  data.   It  is  noted that there were not
 •available data  for 0930.

      It can be seen from Fig. 5.76  that the  behavior of  NO  is accurately
 simulated.   The NO   buildup  is also accurate,  but  the predicted peak is
 too  low.  This  trajectory is one of  the few examples  in which the computed
 ozone does  not  fit the  data.   However,  it  should be  noted that the ozone
 levels are  very low and that the maximum absolute difference is only 4
 pphm, although  the relative error is considerably greater.   The value
                     4    -1   -1
 used for k  was  10   ppm  min
                                                   '79 W SGV
                                                                       tvj
                                                        ELM            a
     Figure 5.74.  October 30, 1969 Trajectory Starting  at  Commerce
                   at 0630 (No, 18)
134

-------
                                                                 MODEL RESULTS •
Figure 5.75.   Trajectory No.  18—Computed
               and  Observed CO
               Concentrations
                                                               INTERPOLATED STATION - NO  O
                                                               DATA          „,,  A
Figure  5.76.  Trajectory No.  18—Compute-
               and  Observed NO and NO
               Concentrations
                                    — MODEL RESULTS
                                     O INTERPOLATED STATION
                                       DATA
    Figure 5.77.  Trajectory  No.  18—Computed and Observed Ozone  Concentrations
                                                                                  135

-------
  5.3.23  Trajectory No. 19, October 30, 1969, Starting at El Monte  at
         0630  (Hands-off)
       The results of the simulation for CO are shown in Fig.  5.79.  It
  can be  seen that the model overestimates CO throughout the day.  However,
  the maximum difference between model and data is about 2.5 ppm.

       As is generally the case with the reactive species, the NO and
  ozone predictions are very accurate, as is shown in Figs. 5.80 and 5.81,
  respectively.  The  NO   balance is poor after 1030, however.  The  N0?
  simulation exceeds the data after 1030.  The low  N02  data at 1130 and
  1230 are suspect inasmuch as  NO -> N0_  conversion continues during this
                                        4    -1    -1
  interval.  The magnitude of  k,  was 10  ppm   min
                                                  79 W SGV
                                                       ELM
                                                         0630
                                                       0730
                                                         80 SE
                                                           1430
      Figure  5.78.   October  30,  1969 Trajectory  Starting  at  El Monte
                    at  0630  (No.  19)
136

-------
                     -MODEL RESULTS
                   O  INTERPOLATED STATION DATA
                              1230
                      TIME (PST)
                                                                  1030            1230
                                                                       TIME (PST)
Figure  5.79.   Trajectory No.  19—Computed    Figure  5.8Q..  Trajectory No.  19—Computed
                and Observed CO                                and Observed NO and NO
                Concentrations                                  Concentrations
                                     	MODEL RESULTS

                                     O  INTERPOLATED STATION
                                        OATA
                                       1030
                                                     1230
                                             TIME (PST)
  Figure 5.81.   Trajectory  No. 19—Computed and Observed  Ozone Concentrations
                                                                                   137

-------
  5.3.24  Trajectory No.  20,  October 3.0,  1969,  Starting in Downtown Los
          Angeles at 0830 (Hands-off)
        The shape of the  CO simulation depicted on Fig. 5.83 does not match
  the data.  Nevertheless,  the maximum difference between model and data
  is 2 ppm.  The increase in CO concentration after 1030 is due tq the
  filling up of the air parcel, thus reducing the vertical concentration
  gradient and therefore  the  vertical  diffusion.
        For
NO  ,  the simulation results are inaccurate, as can be seen in
  X
  Fig.  5.84.   However,  in Fig.  5.85  the  simulated  ozone  shows  once again a
  very close  fit  to the data, despite  the  low quality of the  NO   simula-
  tion.   We note  that it is  difficult  to believe that so much  ozone could
  coexist with the  NO indicated by the data.   Finally, the  value of  k,
                 3-1-1                                       •
  used was 6 x 10   ppm   min
                                                                        to
                                                   1  DOLA
                                                      0830
                                                            COM
Figure 5.82.   October  30,  1969  Trajectory Starting in Downtown Los Angeles
              at 0830  (No.  20)
 138

-------
       10
       0830
                      .MPPEL RESULTS
                       INTERPOLATED STATION
                       DATA
                         I	i
                        1030
                     TIME (PST)
                                        1230
                                                              MODEL RESULTS
                                                           INTERPOLATED STATION  NO  O
                                                           U DATA            NQ  A
Figure fpg3,
Trajectory No. 20^-Computed
and  Observed CO
Concentrations
Figure 5.84.
                                                          1030
                                                        TIME (PST)
                                                                                          1230
Trajectory  No. 20—Computed
and Observed NO  and N0_
Concentrations
                            10
                             2 -
                             0830
                                       __ MODEL RESULTS
                                       O INTERPOLATED STATION
                                          DATA
                                             1030
                                           TIME (PST)
                                                              1230
  "Figure 5.85,  Trajectory No,  20--Computed and Observed Ozone  Concentrations
                                                                                    139

-------
 5.3.25  Trajectory No. 21, November 4, 1969, Starting at Commerce at
         0530 (Hands-off)
       This trajectory exhibits a reversal in direction at 0730 and this
 may introduce inaccuracies in the interpolation process used to calculate
 the concentrations along the path of the trajectory.

       As can be seen from Fig. 5.87, the early-morning CO buildup is
 underestimated by the model, with the relative error at the CO peak being
 about 21%.  The sharp decay and subsequent increase in concentration
 seen after 0930 are not properly simulated by the model because the CO
 emissions are increasing during this part of the trajectory.

       The simulation of  NO   illustrated in Fig. 5.88 shows that  NO
                            X
 is underpredicted but that  N0_  fits the data accurately.  The relatively
 high concentrations of ozone found in the data (Fig. 5.89) are suspect
 because the data show high concentrations of NO present throughout the
 trajectory.  Nevertheless, the ozone simulation is accurate, especially
 at the higher levels.   In this trajectory the value assigned to  k.   was
       3-1-1
 5 x 10  ppm   min

                                         PASA*             «AZU
                                                1130tr, «         BKTD
                                   1  DOLA   0930
                                  °83ii^^3ulo7^^30
                                              • RVA
      Figure  5.86.   November  4,  1969  Trajectory Starting at Commerce
                    at  053;0  (No.  21)
140

-------
                                         CO
             	MODEL RESULTS
              O  INTERPOLATED STATION
                 DATA

             j	L	,_
                    0730           0930
                         TIME (PST)
Figure 5.87.  Trajectory No.  21—Computed
                and  Observed CO Concentrations
                 	 MODEL RESULTS
                  O  INTERPOLATED STATION
                     DATA
                                                                                                 MODEL RESULTS -
                                                                                               INTERPOLATED STATION - NO  O
                                                                                               DATA
                                                                                                            N02 A
                                                                                       0730            0930
                                                                                            TIME (PST)
                                                                                                                    1130
                                                                      Figure  5.88.   Trajectory No.  21—Computed
                                                                                       and Observed NO and N02
                                                                                       Concentrations
     0530
                   0730            0'J30
                        TIME (PST)
                                                        Figure  5.89.   Trajectory  No. 21—Computed
                                                                         and Observed Ozone Concentrations

-------
 5.3.26  Trajectory No.  22, November 4, 1969, Starting at Commerce at



         0630 (Hands-off)


       This trajectory also exhibits a reversal in direction at 0730



 similar to that found in trajectory No. 21.  Thus the caveats mentioned



 previously regarding the accuracy of the interpolation also apply here.






       In contrast with the previous case, Fig. 5.91 shows that the CO



 buildup is reproduced accurately by the model.  However, the CO decay is



 greatly overestimated by the computation.  The CO emissions are high until



 0900, at which time they drop to about one-half of the 0900 value.






       Figure 5.92 shows that the initial NO level is very high and from



 Table 5.3 we can see that the initial hydrocarbon concentration is also



 very high (130 pphm).   This hydrocarbon-NO combination portends high  N0?



 levels and this is precisely what the simulation produces, as can be seen



 in Fig. 5.92.   The ozone simulation is generally accurate, however, although



 the ozone peak is overestimated by about 28%.   For this trajectory, we

                                 4    -1    -1
 used a value of  k,   equal to 10  ppm   min








                                     LACA                          §
                                                                 U3
                                                                 c^
                                                                 t-O

                                        1130                       a:

                                            PASA

                                           S-1030

                                                             D BKT
                                                QKFI
     Figure 5.90.  November 4, 1969 Trajectory  Starting  at  Commerce

                   at 0630 (No. 22)
142

-------
                0830           1030
                     TIME (PST)
Figure  5.91.  Trajectory 22—Computed and
               Observed CO Concentrations
                                                                                    MODEL RESULTS .
                                                                                  INTERPOLATED STATION - NO O
                                                                                  DATA
                                                                                               N02 A
                                                                               0830           1030
                                                                                    TIME (PST)
                                                              Figure  5.92.
Trajectory No. 22—Computed
and Observed NO  and N0~
Concentrations
                                                  Figure  5.93.   Trajectory No.  22—Computed  and
                                                                 Observed Ozone  Concentrations
                     TIME (PST)

-------
 5.3.27  Trajectory No. 23, November 4, 1969. Starting in. Pasadena at
         0530 (Hands-off)
       Figure 5.95 shows that the model underestimates CO concentrations
 throughout the trajectory.  In this case, the strength of the emissions
 was too low to produce the high values indicated by the data even under
 highly stable meteorological conditions.  Upon investigating the data, it
 became apparent that the high concentrations from 0630 to 0830 are due to
 the downtown and Commerce stations which, in view of the path of the
 trajectory, makes these interpolated concentrations suspect.  The sus-
 picion is heightened upon noting that at 0930, when the air parcel is again
 close to Pasadena, the predicted and observed concentrations are closely.
 matched.

       Figure.5.9o shows that the computed  NO   balance is very poor, with
                                              X
 the predicted NO and NO- considerably below their apparent observed values.
 As can be seen in Fig. 5.97, the computed ozone accurately reproduces the
 observations-,- the largest deviation occurring at 0930.   This is puzzling
 in view of the results obtained for CO at 0930, but with chemical
 processes at work, it is not surprising.  The value used for  k,   was
       3-1-1                                            4
 4 x 10  ppm   min
LACAt
• BURK
• HOL
	 -_ .^_ *
^ 	 " ~\ 1 DOLA
. ,1130 g
\1030 7
I PASA I
0930|^n530
loeV .ELM DBKT
0830*
-------
                                                   	 MODEL RESULTS

                                                    O  INTERPOLATED STATION
                                                       DATA
                                          TIME (PST)
    Figure 5.95.  Trajectory No.  23—Computed and Observed CO  Concentrations
 _. 40


 §

.jl»
 
-------
  5.3.28   Trajectory No.  24,  November 4, 1969, Starting in Downtown Los
          Angeles  at 0530 (Hands-off)
       The  simulation of CO  yielded the results shown in Fig.  5.99.   The
  early-morning  buildup of CO is accurately simulated.  Once again,  however,
  the  lower  concentrations-of CO are overpredicted by the model.
                                                NO   balance is  relatively
                                                  X
      In Fig. 5.100, it can be seen that the
accurate until 0930, although the NO peak is underestimated.   After 0930,
the computed  NO.  remains too high, while the NO  fits  the data correctly
For ozone, we see in Fig. 5.101 that the predicted ozone  buildup until
1230 is accurate, but after 1230 the model exceeds the  data by a maximum
                                          4    -1     -1
of 10 pphm.  The value of  k.  used was 10  ppm    min
                       DSAU
                         • NEWHALL
         ZUM
                 CANOGA
                 PARK
                       MISSION HILLS
                          1430
                      RESEDA
                        •
                     ENCINO*
L
HI
N
.ES
0 5 10 1
                                                     ,EL MONTE
                                                                  BKT
                                                                   Q
Figure 5.98.  November 4, 1969 Trajectory Starting  in  Downtown Los Angeles
              at 0530 (No. 24)
146

-------
                                    MODEL RESULTS 	
                                    INTERPOLATED STATION DAtt O
Figure 5.99.   Trajectory No.  24—Computed  and
                Observed  CO Concentrations
                                                                         o
                                                                         0430
                                                                                           MODEL RESULTS  _____
                                                                                           INTERPOLATED STATION O NO
                                                                                0630     0830     1030
                                                                                         TIME (PST)
                                                                                                     1230     1430
Figure 5.100.  Trajectory No.  24—Computed
                 and Observed  NO and NO  '
                 Concentrations
                                      D -INTERPOLATED STATION DATA
                                     	 MODEL RESULTS
                                  0430    • 0630     0830     1030
                                                  TIME (PST)
       Figure  5.101.   Trajectory No.  24—Computed  and Observed  Ozone  Concentrations

-------
  5.4   TECHNIQUES FOR MODEL OPERATION
       In  this section, we describe  several basic  features  of the atmospheric
  model to  assist the prospective model user.

  5.4.1  Kinetic Model for Atmospheric Simulation
       The chemical model used in atmospheric  simulation  is basic'ally the
  same one  used: to model the smog chamber experiment using a dilute auto
  exhaust mixture from a vehicle with exhaust hydrocarbon  and carbon
  monoxide  emission controls (experiment 231;  cf. pp.  37-40  and p. 43).
  This kinetic model was chosen for atmospheric applications because the
  smog's kinetics are not likely to be influenced by CO because of its concen-
  trations  in experiment 231 and in the atmosphere  (cf.  p.-36).

       The branching factors are features of the smog chamber model which
  have been retained on moving to atmospheric simulation.  Thus we have
  b  = b2 = 8  , and  b_ = 1 .  The OH yield factor,  y , remains equal to
  1/8.

       The kinetic model for the atmosphere differs from  the smog .chamber
  model mentioned above in that a. single hydrocarbon class is used rather
  than two.  The rate constants for the basic atmospheric model are shown
  in Table  5.6.  The constants for reactions 3 and 5 shown in Table 5.6 are
  rounded mole-weighted averages of the two reaction pairs (3,  3a)  and
  (5, 5a) shown in Table 2.7.  The first-order constant for  reaction 16
  shown in  Table 2.7 was halved in the process of parameter  adjustment
  required  for atmospheric simulation.  (See p. 95 for a discussion of
  the model's sensitivity to  k, ,.)
                              16

       The value of the rate constant of the first reaction depends on the
  intensity of ultraviolet light in the wavelength range 2900-3850 A.   The
 magnitude of  ^  is treated in the simulation as a function  of  solar
 zenith angle, and hence of time of day.   See p.  150 for an explanation
 of the derivation of  k^  and the sources of uncertainty associated with
 the value of this rate constant.
148

-------
                               TABLE 5.6
          RATE CONSTANTS USED IN ATMOSPHERIC MODELING STUDIES
                                                         *
            Reaction No.                    Rate Constant
                la                        1.32 (-5) ppnf2 min'1
                2                              2.67(+l)
                3                              2. 76 (+2)
                4                                 ***
                5                              4.0(-3)
                6                              1.0(4-5)
                7                              2.0(+2)
                8                              1.5 (+3)
                9                              3.0(4-3)
               10                                  t
               11                              1.0(-3)
               12                              5.0(-3)
               13                              4. 5 (+3)
               14                           1.4(4-1) min"1
               15                           6.05(4-1) min"1
               16                           1.0 (-3) min"1
  *             -1-1
   Units are ppm   min   unless otherwise specified.
 **
   This rate constant depends on sunlight intensity and in the simulation
   it is treated as a function of solar zenith angle, and hence  time  of
   day.
***
   The rate constant for this reaction is given in the section which
   describes each of the cases tested.
   This reaction is the photodissociation of HONO and its rate constant
   is obtained from the relation  k1Q = 3.75 * 10~3 k^.
                                                                     149

-------
       The rate constant for the photodissociation of HONO (reaction 10)
                                                          _3
 is obtained from  ^  using the equation  k1Q = 3.75 * 10   ^ .  This
 relation results from the assumption that the ratio  ^g^l  "^S a constant
 equal to the ratio of the quantities used in the smog chamber simulations.
 In atmospheric modeling, k1Q is thus a function of time.

       It became necessary to adjust the rate constant of reaction 4
 (OH + HC ->• (b?)R09) during the simulation process conducted under atmospheric
 conditions.  For this reason, the value of  k,   is given in the sections
 describing the results of the simulations.

 5.4.2  Chemical Inputs
       Two inputs are basic to the operation of the chemical model.  These
 are  k  , the rate constant of the reaction  hv + N02 •> NO + 0 , and
 i  . . the rate constant of  OH + HC ->- (b0)R00.
  H                                     /   Z

       The value of  k..  is of course a function of the intensity of ultra-
 violet light with wavelength in the range 2900-3850A.  The ultraviolet
 intensity depends on time of day, time of year, geographical location,
 and weather conditions.  In our simulations, we have used values of  k
 which correspond to clear-day conditions, i.e., no overcast.  Thus the
 magnitude of  k   used is an upper bound of the actual value.   This can
 be seen in Figs. 5.102-5.107 in Sec. 5.5.
       The derivation of the clear-day value of  k   is accomplished by
 means of the relationship between  k   and solar zenith angle given by
          •I Q                         i
 Leighton.    The solar zenith angle is a function of time of day, time
 of year, and geographical location.  Using solar ephemeris tables, a
 table of solar zenith angles is generated for the specific times and
 places of interest.   Given the solar zenith angle, the value of  k   is
 then determined.  In our model, we update  k   at 10-minute intervals,
 but this is arbitrary and can be changed to match the integration .step
 size.
150

-------
      The value of  k,  is determined by the concentrations of  NO, NO  ,
                     cl                                                •Ł•

and reactive hydrocarbon at the start of the trajectory or at sunrise,


whichever is later.  Thus  k,  is dependent on the initial mixture of


pollutants.  We recall that this is in agreement with findings in the


validation of the chemical model using smog chamber data for the more


reactive mixtures such as propylene/NO  and auto exhaust/NO .  During

                                      X              3     X 4    -1    -1
the atmospheric simulations,  k,  ranged from  4 x 10   to 10  ppm   min


The rules for determining  k,  will not be exact because of uncertainties


in the initial concentrations.  The variability of emissions from case


to case is another factor which affects the choice of  k,  .  Thus a plot


of  k,  on a graph with ordinate equal to  NO   concentration and abscissa
     T1                                       X

equal to reactive hydrocarbon concentration reveals some scatter in the


values of  k.  relative to the quantities  HC/NO   and  HC + NO  .  How-
            4                                   x              x

ever, some trends are apparent from such a plot and we have used these


to establish our guidelines.  In any event, we note that the range of


values of  k,  is not large, and that by far the most frequent values of

               3         4-1-1
k,  are  4 x 10   and  10  ppm  min    .  Based on these frequencies of


appearance, we can formulate a general rule:  if  HC/NO  >_ 1.7 , then
           "3    _ "1    1                      /     1    1

k, = 4 x 10  ppm  min   ; otherwise,  k, = 10  ppm  min    .  However, a


more involved, but more exact set of guidelines for selecting  k,  is as
follows :
1.
2.

3.



If HC/NO >_
X
If HC/NO <
X

If HC/NO <_
X
-1 . -1
ppm mm

1.7
1.7

1.2



, then
, then

or if



4 x
6 x

HC



103 <
103<

+ NO >
x



k, < 5 x
4 —
k. < 104
4 —

1.8 ppm



in3 -1 - -1
10 ppm mm
-1 . -1
ppm mm
4
, then k4 = 10


            If 1.4 < HC/NO  < 1.5  and  1.2 ppm < HC + NO  < 1.4 ppm  ,

                          X  3    -1    -1               X
            then  k, = 8 x 10  ppm   min
                                                                      151

-------
 5.4.3  Meteorological Inputs
       The three basic meteorological inputs are the maximum inversion base
 height, a table of inversion heights as functions of time, and the
 diffusivity parameters.   Establishing general guidelines for selecting
 these inputs is a difficult task;  this is especially true for the last
 two quantities.  The spatial and temporal variability of meteorological
 conditions impose a high degree of difficulty in trying to devise general
 rules.  Thus one would expect the  inversion height, for example, to
 attain different maximum levels at different locations.  Similarly, the
 changes of inversion height with time can be expected to show spatial
 variability.  Therefore, we must have information about inversion base
 height as a function of  time for each trajectory.  This information can
 be obtained from plots of inversion height isopleths for various times
 for an ^entire geographical region.  Given a description of the path of the
 trajectory, the required inversion base data are then obtained from these
 plots for the trajectory in question in a straightforward manner.

       Determining the diffusivity  parameters is somewhat more difficult.
 In this case, the required data consist of profiles of temperature as a
 function of height and time.  These data determine the stability classes
 and from these the diffusivity coefficients can be obtained using
 Fig. 3.7-  The time variation of the stability class determines the time
 dependence of the diffusivity:coefficients.   Results from our simulations
 indicate that a likely value of diffusivity in the very stable case is
         32                        32
 2.5 x 10  cm /a , rather than  5 x 10  cm /s  as shown in Fig.  3.7.  The
 other values of diffusivity shown  in Fig.  3.7 were used without change in
 the simulations.

 5.5   SOURCES OF UNCERTAINTY DUE TO SOLAR RADIATION AND PARTICULATE
       REACTIONS
       In the course of any serious evaluation process, any systematic
 validation-check of a simulation model demands these two things:
       1.     A quantitative characterization of each uncertainty
             entering the model
152

-------
      2.    A clear identification of the sensitivity of the model outputs
            to each uncertainty in the inputs
Many efforts in the past several years have been concentrated ou the
second aspect enumerated above.  Indeed, we and others have run parametric
analyses until the results are almost intuitive to any worker in the field.
Unanswered questions still surround the first requirement, however.  These
questions need not be a source of mystery because straightforward well-
planned investigations can be designed to get the answers.  No new dis-
coveries of laws of physics or chemistry will be necessary, but rather,
the commitment of resources that concentrate on finding out what is really
happening.  Suggestions for the future are embodied in Sec. 6; however,
for now, we must content ourselves with identifying and assessing the
uncertainties in both the input data and the validation data base.

      The primary process in the production of photochemical smog is the
photodissociation of nitrogen dioxide to form nitric oxide and atomic
oxygen.  Our rate constants for this reaction are derived in two ways.
In the first, we assume clear skies and deal with the atmospheric trans-
mission of radiation in the dissociation bands according to a zenith angle
derived from time of day, day of year, and location on the earth's surface.
                     f\ /•
In the second method,   Eppley ultraviolet detector readings are calibrated
to the clear-day curve of  k.  versus solar zenith angle.  The cosine
correction presupposes the preponderant contributions to be from direct
rather than scattered ultraviolet radiation. . As will be seen, this
leads to significant errors only at large zenith angles.

      Using these two methods, we have explored the uncertainty in  k
that occurs for predictions that assume clear day values.  Figures 5.102
through 5.107 show diurnal  k.  variations for each of the six data
days.

      It can be seen from the figures that the departures of the actual
k   from the theoretical clear-day values can be rather large.  As might
                                                                     153

-------
in

IO
    -.00
               SEP   11
                                  CLEAR DAY
       6.00     6.00     7.00
—1	1	1	1	1	1	1	1	1	1
 a.oo     9.00    10.on     11.00    12.00    n.oo    m.oo    15.00    is.oo     n.oo
                TIME tPST) (X1Q  2)
     Figure  5.102.   N02  Photolysis  Rate  Constant,  k ,  for September 11,  1969

-------
                             :MD -
                             .30 -
                             .25 -
                        tn

                        in
                             .10-
                                       SEP  29
                                            	1	1	
                               6.QD    6.00     7.DO     B.OO
—I	1	1	1	
 9.00     iQ.oo     a.oo    12.00
         TIME (PST) (X10  2)
                                                                                        13.00     IM.OO     15.00    16.00    17.00
01

Ln
                               Figure  5.103.   NO   Photolysis Rate Constant,  k  ,  for  September  29,  1969

-------
 .26 -
 -"-
 .10 -
 .OS -
-.00
           SEP  30
   6.00     6.00     7.00     8.00    9.00    10.BD     11.00    12.00    13.00     1M.DO    15.00    16.00    17.00
                                       10. HI     11.00    12..
                                        TIME (PST) (X1Q  2
 Figure  5.104.   N02 Photolysis  Rate  Constant,  k.. ,  for  September 30,
1969

-------
.60 -i
                         .00     9.00    10.00     11.00     12.00    13.00    14.00    15.00    16.00    17.00
  5.00
  Figure  5.105.   N0? Photolysis Rate Constant,  k.. ,  for October  29,
1969

-------
Ln
OO
                      to

                      g
                      ED
                      I
                      O.
                            6.00    6.00
                                                                                                  16.M    n.on
                            Figure  5.106.  N02 Photolysis Rate  Constant, k  ,  for October 30,  1969

-------
     .10 -
0)
z
o
       a.oo    B.oo
                                         10.00    1L.OO    12.00
                                          TIME (PST) (XIQ
                                                                                   16.00    17.00
      Figure 5.107.   N0?  Photolysis Rate Constant,  k , for November 4,  1969

-------
 be expected, there also exist marked differences in the value of  k^
 at different locations in the Los Angeles Basin, in this case Commerce
 and El Monte.  The ratio of (theoretical ^/actual k^ ranges from unity
 at a few points to 8/1 on September 11.  As a general rule, the theoretical
 k  is higher by a factor of 1.5 or 2 for Commerce and by about 1.3 for
 El Monte.

       To ascertain the magnitude of the uncertainty introduced, we ran
 parallel cases using the Commerce values of  k..   as well as the clear-day
 values for the trajectory starting near Pasadena at 0730 on October 30,
 1969 and ending west of Pomona at 1330.  This is by no means the worst
 case, as is easily ascertained from the Fig. 5.106.  The result of the
 test was that the concentrations of  NO   were perturbed very little, but
                                        X
 the peak ozone concentration was lowered from 13 pphm using the clear-
 day  k1  to 9 pphm using the Commerce  k  , a relative change of about
 31%.  Such a margin of error is significant and must be borne in mind in
 assessing the expected accuracy of the computed concentrations in view
 of the uncertainties in the inputs.  Interestingly enough, however, in
 this particular case (see Fig. 5.73), the simulation using the clear-day
 k.  is actually very close to the measured data and using the Commerce
 k..  would have degraded the simulation if no other rate constants were
 adjusted to compensate for the lower ozone values.  In closing, we note
 that another case has been tested using the data for September 11 since
 it shows the largest departures from theoretical values (see Sec. 5.3.5).

       Another source of uncertainty is the influence of  N0? + particulate
 reactions.   It will be recalled that such a lumped reaction, reaction
 (2.16), was introduced in the kinetic model in simulating smog chamber
 experiments involving auto exhaust in which the presence of aerosol was
 observed (see Sec.  2.5.4).  Introducing this reaction produced a better
 simulation  of the disappearance of  NO   from the system.
                                       x
160

-------
      The sensitivity of the model to changes in the rate constant of
this reaction was tested during the atmospheric validation runs.  The
effect of increasing  k-,  was to lower the  NCL  concentration and to
increase the ozone concentration late in the day.  The sensitivity of the
NO,,  peak was negligible compared to the sensitivity of ozone.  However,
the sensitivity of the end value of  N0~  almost matches that of  CL .
This interaction can be explained by realizing that in the kinetic model,
the reaction  N0_ + CL  is very strong at late times when  N0«  and  CL
are high; thus removing  NCL  will cause the  CL  to increase.  As an
example, we ran tests using the November 4, 1969 trajectory with
        -3    -1                   -3    -1
k, ft = 10   min    and  k., = 2 x 10   min   .  The peak ozone went from
25 to 32 pphm and the peak  N0_  underwent a negligible change.  The
concentration of  NO-  at the end of the trajectory was reduced from
21 to 16 pphm.
                                                                     161

-------
 6     CONCLUDING REMARKS
      These findings establish progress milestones in  the  understanding
 of  chemical kinetics and atmospheric dynamics underlying air  pollution
 models.

      Systematic investigations of laboratory smog chamber experiments
 have provided significant insights into the air chemistry  of  urban atmps=
 pheres.  Although the artifices of the apparatus prevent direct  applica-
 bility to the atmosphere, important qualitative features are  highlighted
 by  computer simulations of chamber studies.  Mixtures  of various hydrq-
 carbons with nitric oxide were modeled.  The hydrocarbons  included pro^
 pylene, toluene, toluene/n-butane, and auto exhaust from vehicles  with
 and without exhaust emission controls.  Added chain-breaking  reactions
 for smog chambers improved the predictions markedly.   These,  along with
 previously added    OH + HC  reactions, give consistent behavior for
 various initial mixtures placed in the chamber.  Incorporation of  recently
 reported rate data narrowed the options for adjustable parameters  (accord-
 ing to our ground rules, at least), but improved the validations overall.
 As  might be expected for a complex nonlinear system, the nonuniqueness
 of  the rate constants was discovered for validation of a given experiment;
 that is, an entire set of rate constants could be moved through  several
 orders of magnitude and preserve the same computed results.   Selection
 of  the best set was made using benchmark values that have  recently come
 out of the laboratory.  Rank order and proportionality of  hydrocarbon
 reactivity indices were shown to bear a direct relationship to the set
 of  rate constants for hydroxyl attack of the hydrocarbons.  This built
                                                       o -i
 further confidence in our previously adopted procedure  of reactivity
 scaling of smog chamber reaction rates to model atmospheric systems.

      Model methodology was substantially improved in  several areas.   The
 logic for air trajectory computation has been systematized by generaliza-
 tions derived from many hand calculations using actual data.  Consistency
 checks between tetroon trajectories and calculated ground  trajectories
162

-------
revealed some large sources of uncertainty that are not considered in air
pollution models.  The dominance of stratification over wind shear was
incorporated in changes in the calculation of vertical turbulent mixing
coefficients.  While previous formulations stressed wind speed dependence,
the newly adopted methods depend on potential temperature gradients.
The impermeable inversion base assumption was abandoned for upper boundary
conditions.  Vertical mesh intervals were extended well above the inver-
sion base with assignment of vertical diffusion coefficients controlling
upward mixing according to local stability conditions.  Thus the inver-
sion base was traced through the mesh by varying the diffusion coefficient
in time and space.  While the present work did not reexamine the source
inventories, many changes were made to assure direct comparison with other
model results.  Extensive improvements in the computer implementation of
source models permitted a high degree of responsiveness to the frequent
alterations.

      Turbulent diffusion transverse to the wind was assessed to evaluate
possible errors due to mass exchange between neighboring stream tubes.
As suspected previously, only minor perturbations are introduced by
lateral mixing perpendicular to the path of an air parcel.  The case of
side-by-side trajectories was tested using the GRC three-dimensional time-
dependent LAPS code which is especially adapted to treat conditions of
large transverse gradients in emission fluxes that typify localized large
sources.  Worst-case carbon monoxide area sources were tested using the
extreme values of fluxes determined from dozens of actual air trajectories
over the Los Angeles Basin.  The spread of power plant stack plumes from
off-trajectory sources into the control volume was investigated for a wide
range of parametric conditions of point emission and area emission strengths
To insure realism, the parametric values were selected from the actual
inventory statistics.  Even the largest percentage errors are not likely
to exceed other uncertainties due to model inputs.  Systematic examina-
tion of validation trajectories relative to these findings showed that
transverse diffusion errors are far smaller than other sources of uncer-
tainty in the model inputs.
                                                                      163

-------
       With the  chemical  and meteorological improvements,  the validation
 results  have been more gratifying  than ever before.   Consistency in
 assumptions for both  the kinetic rates and the diffusion  parameters can
 be maintained to model a wide variety  of  cases without ad hoc adjustments.
 Our abandonment of  the single-receptor validation criterion placed much
 more rigorous constraints  on the model tests than had been originally
 anticipated.  Each  hourly  trajectory node has an interpolated set of
 concentrations  based  on  neighboring station values.   Therefore,  we seek
 to match the shape  of each pollution history rather  than  just matching
 values at the end point.   Despite  these more severe  requirements, both
 diffusion and photochemical validations were remarkably successful.

       Some problems remain that are critical to  the  future success of
 simulation modeling.  It is likely that they stand alone  by now  as the
 main obstacles  to further  fidelity improvements.   Following this reason-
 ing,  one realizes the need to establish closer  coupling between  modelers
 and measurers than  has been the case in the past.  The continuity that
 is thus  assured will  build the scientific  foundations  needed to  attack
 air pollution abatement problems on a  rational basis.

       Finally,  of the myriad topics for additional research which are
 worthy of note, we  wish  to single out  a few which  should  improve future
 models when the problems posed have been  solved.   In what follows,  the
 order of discussion of the topics has  no  particular  significance.

       The first subject  that merits some  discussion  is  the need  for  a new
 measurement of  the  rate  constant of the reaction   NO +0  -> NO   +0  .
                                                     ^     -J    O    Ł,
 In our model, this  rate constant had to be  reduced by  a factor of 10 in
 order to. achieve accurate  reproduction  of smog chamber  experiments.   Since
 this  reaction appears to be rate-controlling at late times when   NO    and
 03 have reached high concentrations,  and since the  available measurements
 of this  rate constant are  rather old (1949  and 1957)  a  new measurement
 is warranted.
164

-------
      A second topic of interest is whether  HONO  exists in the atmos-
phere in significant amounts (around 1 pphm).  Should this not be the
case, the chain-breaking structure of the kinetic model will have to be
revised.  A corollary to this question is that if  HONO  is found in the
atmosphere, then the rate constant of the reaction  hv 4- HONO ^ OH + NO
needs to be determined.  There appears to be no measurement of this rate
constant at the present time.

      Thirdly, additional investigation is needed to ascertain the nature
of NO^-scavenging processes in the atmosphere.  The possibility that
heterogeneous reactions may play a significant role must be included in
such a study.  Such knowledge would help us to improve the late-time
behavior of  NO   in the simulations.
                                                                      165

-------
                               APPENDIX A
           A VIEW OF FUTURE PROBLEMS IN AIR POLLUTION MODELING
*
 This first appeared as  General Research Corporation TM-1631,  March 1972
 and was  subsequently published in Proceedings  of Summer Computer Simula-
 tion Conference,  Simulation Councils,  Inc.,  LaJolla,  Calif.,  June 1972,
 pp. 1013-1027-   Research reported in this document was  originated through
 independent efforts, not under a Government  contract or program.
                                                                     167

-------
 A.I   INTRODUCTION
       Vigorous  steps in abating air pollution  demand heavy  investments
 both  in  the public  and private sectors.   In  addition,  second-generation
 cleanup  measures are likely to add personal  inconveniences  to  the already
 growing  financial burdens.  Large dollar  outlays  and the need  for public
 support  demand  that decision-makers have  reliable means of  evaluating
 alternative abatement strategies.

       Mathematical  air quality models are quantitative tools that will
 play  a central  role in evaluating the environmental aspects of decisions.
 These decisions may take the form of regulations  aimed at rolling back
 existing pollution  or of minimizing the environmental  damage potential of
 future public works projects.  For abating existing sources, implementa-
 tion  planning must  show how control regions will  achieve ambient  air
 quality  standards within a specific number of  years.   This  requires
 predictions of  absolute levels of air pollution.  For  assessing impact
 of projected sources, a statement must be filed that demonstrates that
 these sources will not cause environmental damage.  This, too,  requires
 predictions of  absolute levels of air pollution.  These requirements
 impose stringent demands on the best air  quality  models presently
 available.

       For planning  long-range strategies  on a  national scale,  the objective
 is to choose rationally from among a field of  alternative abatement
 actions.  The selection criterion is built around maximum benefit/cost
 ratios.   Thus,  the  long-range considerations require measurements of
 alternatives on a relative scale.  Currently available air  quality models
 will  be  very useful in fulfilling this less stringent  requirement.

       Unfortunately, the time scale on implementation  plans and impact
 statements is much  shorter than that on national  planning goals.   It  is
 already  becoming evident, however, that the costs of really massive roll-
 back  strategies far outstrip our ability  to pay over the short time
168

-------
intervals required by recent statutes.  Consequently, we will be forced
to analyze less ambitious plans in order to define optimum steps toward
pollution abatement.  This realization may be many months or even years
away, but when its effects are felt, it will require the best products
of the modeler's art.

      This paper takes two directions in assessing these future needs:
first, it highlights some scientific problem areas that need immediate
attention; and, second,  it suggests some ways of adapting models for
abatement analysis applications.  Any approach to the unsolved phenomenology
problems inevitably leads to greater degrees of complexity in the model.
In direct opposition to this trend stands the need for sweeping simplifi-
cations in practical adaptations of the models.  More likely than not,
the air quality model will be but one of the many modules in any realistic
abatement simulation.  The conflict between pure and applied efforts can
be resolved only by a high degree of communication between the researchers
and the planners »-

      In the realm of phenomenology, we examine two potential sources of
error in predicting atmospheric reaction rates.  One involves gas-solid
interactions on urban surfaces and on aerosol particles.  The other arises
from turbulent fluctuations of reactant gas concentration.  Experimental
evidence of the problems is cited and research approaches to its solution
are outlined in each of the two cases.  Analytical corrections to the air
quality models are proposed.

      Systems- implementation schemes are then considered to define the
fidelity level that the models must achieve.  A hierarchy of different
versions emerges from the various objectives that are set forth.  For
some cases, the versions are already available, but for others, only a
sketch plan of the specifications can be given.
                                                                     169

-------
 A. 2   PHYSICAL INTERACTIONS WITH POLLUTANT REACTION  RATES

 A.2.1 Gas-Solid Interactions
       Urban surfaces near ground level and particle  surface's distributed
 through the mixing layer can serve as reaction sites  for  gas molecules
 impinging on them.  If these effects compete significantly with homogeneous
 reactions,  appropriate sink mechanisms must be introduced into atmdspheric
 models.  Indeed,  this has already been done for oxides  of nitrogen in our
                                             27
 earlier work.   Briefly summarizing, we noted   that  the observed buildups
 of  carbon monoxide and hydrocarbon during morning peak  traffic were well
 modeled by the values of emission fluxes and atmospheric  diffusion coef-
 ficients in the literature.   On the other hand, the sum of (NO -f NO^)
 was grossly overpredicted as shown in Fig. A.I (Curve A),   On the* graph,
 the symbol  "r" refers to the oxidation fate reduction (beldw that of pure'
 propylene)  and symbol "f" to the fraction of the inventory value  NO
                                                                      X
 emission flux  used.
             160


I
a
a
z~
O
1-

140

120

100
CURVE
A
B
C
D
r
1/2
1/3
1/2
1
f
1
1/4
1/4
1/4

          LU
          O
             80
          _  60
          CM
          O
          z
             40
             20
r-  CURVE PARAMETERS
                                 r = RATIO OF HYDROCARBON OXIDATION RATE
                                    TO PROPYLENE RATE
                                 f = NO EMISSION SCALE-DOWN RATIO
                        _L
             _L
_L
_L
_L
              0600      0800      1000      1200
                           TIME OF DAY, PST
                                   1400
   Figure A.I,   (NO + N02>  - Concentration Ground Level Buntingto'n Park
170

-------
      Due  to an apparent rapid removal of nitrogen oxides  from the gas
phase, the published values of emission rates had to be reduced by a factor
of four to offset  the losses.   Note that the first 2.5 hours  of buildup
are practically unaffected by the choice of reaction rates (over a factor
of three).  Although flux reduction is an ad hoc correction for these
results, it may not  be generally applicable to all types of surfaces or
to all types of days or even to all times during a given day.   Figure A.2
illustrates the difficulty very clearly.  It displays averages of the
CO/NO   mole ratios  for groups of 1968 data measured by Scott Research
     X       35
Laboratories.    It  is intended that  CO  be regarded as an inert tracer
so that variations between observed ratios and source ratios  reflect
loss of  NO  .   This removes uncertainties due to dilution and other
           X
interferences.   Types 1,  2,  and 3 denote high oxidant days and Type 0,
low oxidant days.  The factor of four (between source values  and air
         70
         60
         50
         40
      O  30
         20
         10
                         (CURVES FAIRED THROUGH DATA POINTS)
                             AVERAGE OF TYPE '1, 2, AND 3' DAYS (HIGH OXIDANT)
                           AVERAGE OF TYPE '0' DAY LOW OXIDANT
                   ALL SOURCES
               0800
                           0800
                                      1000
                                  TIME OF DAY, PST
                                                  1200
                                                             1400
           Figure A.2.  CO/NO  Ratios for Huntington Park 1968
                             X
                                                                       171

-------
 values)  shows up  clearly at the morning traffic  peak  for  high oxidant days.
 Low oxidant days  exhibit far lower discrepancies between  source ratios
 and ambient air ratios.

       These bits  of evidence show the nature of  the problem but they do
 not indicate a truly  general solution.  Indeed,  the deficit of gas  phase
 nitrogen oxides has been observed for years in laboratory photooxidation
                                               9
 experiments in smog chambers.  Gay and Bufalini  give an  excellent  review
 of these problems and demonstrate that analysis  of surface-adsorbed
 products greatly  improves  the nitrogen balance in these experiments.
 Substantial uptake of pollutants in soils has been demonstrated in
                     OT  OQ  OQ
 experimental  studies   ''  reported  in  the  literature.   From these results,
 ground absorption can be deduced or estimated.   They  average out  to the
                                          2
 following very approximate values (in mg/m «hr):  ~1  for  ozone,  ~8  for
 carbon monoxide and ~3  for nitrogen dioxide.  It is important to  note that
 the ozone value is for  atmospheric background concentration (-0,05  ppm)
 while the  CO  and  N0_  values are for tens or  hundreds  of ppm in  a 10-
                     37
 liter vessel.  Aldaz    assumes  that  the surface  reaction  rate must  be first-
                                                                      38
 order in the ozone concentration, but the Inman  and Ingersoll results
 show a linear concentration decay of  CO  suggesting  zero-order.  Before
 any of the  reported values can be used in models, the concentration (and
 possibly temperature) dependence of the uptake rates  must be determined.
 It is of interest, nevertheless, to compare these values  with the emission
 fluxes averaged over  the Los Angeles basin area.  For oxides of nitrogen
                                                 o
 (as NO^) the emissions  are approximately 10 mg/m • hr  (against 3 estimated
 for soil uptake)  and  for carbon monoxide it is approximately 100  (against
 8 estimated for soil  uptake).  Therefore, these  preliminary indications
 are consistent with our model adjustments and with the observed atmospheric
 C0/N0x  ratios; namely, that there could be significant perturbation of
 the N0x~balance,  but  only  minor effects on the CO-balance.   Surface
 uptake of ozone has been regarded as the major global sink for this gas;
 however, Ripperton and  Vukovich   indicate that  gas phase removal may .also
 be important even at  background conditions.  Some of  our  early calculations
 for a fully catalytic ground surface suggest that gas-phase reactions
172

-------
dominate the ozone balance in polluted air to within a few meters of the
ground.

      Modeling the uptake at the ground is a simple matter of adding
boundary conditions to the finite-difference approaches that simulate air
quality for distributed sources.  As indicated above, the lack of reliable
data is the main obstacle.  Algorithms are already developed for the flux
boundary condition at (or near) the ground level.  In general, this logic
is suitable only for zero-order surface reactions, but minor modifications
would generalize it to the rith-order uptake processes.  The modifications
would substitute  kc   for the constant flux now in the boundary conditions,
      Now turning to the gas-solid interactions at particulate surfaces,
                                             41
we begin with the evidence cited by Lundgren:
            On days of heavy smog, very hygroscopic, crystalline-
            like particles were found to comprise over half of
            the particulate dry weight in the 0.5-1.5 ym
            diameter size range.  These crystalline particles
            were analyzed by X-ray diffraction and identified
            as ammonium nitrate.
      It has been known for some time (see for example, Ref.  42)  that
Los Angeles aerosol has anomalously high fractions of nitrate when com-
                                                           3
pared with that from other cities.  Altshuller and Bufalini  cite smog
chamber experiments with auto exhaust emphasizing the rapidity of conversion
of nitrogen oxides to particulate nitrate.  They urge that further field
work be undertaken to determine atmospheric rates of conversion.
      Oxygen-atom reactions with particulate matter were discounted by
                     43
Leighton and Perkins.   They carried out a calculation of the mean displace-
ment during the gas-phase lifetime of an 0-atom.  For a high particle
               3
loading (1 mg/m ),  they then estimated the total volume of the spheres
                                                             -4
of influence of the particles.  It accounted for less than 10   of the
volume of the gas,  thereby forming a basis for neglecting the influence
on 0-atom chemistry.
                                                                     173

-------
       We can undertake a somewhat more detailed analysis of other pollutants
 by carrying out some calculations for simultaneous gas-phase and surface
 reactions in the neighborhood of a particle.   The physical picture is a
 steady state diffusion of a reactive species  toward the particle surface.
 For steady state, the governing equation is

             DV2n. + w± = 0                                         (A.I)

 where        D = Diffusion coefficient
             n. = Number density of ^th species
             w. = Net production rate of _ith species in the gas phase

 The form of the production term is predicated on a high level  of chemical
 activity for species  i .   This is based on a constant strong  source  s.
 opposed by a series of fast reverse reactions that consume species  i
 at a rate of  k n.   where  k   is a composite of gas-phase rate constants
                o            o
 (for reactions with  q  other species)  and  concentrations  given by
             k, = >  .0-  -  ^J^.n,                                 (A.2)
 where      6..  = 1     ;     i  =  i
             ij                  J
            «±j  - 0     ;     i  /  j

 so that for the gas  phase production term we  use

            Wi  = Si  "  kgni                                          (A<3)

       Outside of the range  of influence of  the  particle, we can calculate the
 homogeneous stationary state  number density of  species  i   by  setting
 w  = 0  giving

                  si
            n°°  ~ k                                                  (A. 4)
                  g

174

-------
Incorporating spherical symmetry and substituting Eq. A.3  in  Eq. A.I, we
get
             - k n. = -s
                                                                    (A. 5)
      The heterogeneous reaction enters the problem in the boundary
condition at the particle surface  (r = rn) where we assume a first-order
reaction balancing the influx of the jLth species,
dn
d
                        - "Vic
(A.6)
where  k   is the rate constant for the surface reaction.
        S         " "   "     --..---.              .....
      The homogeneous reaction is approached at an infinite distance
from the particle
            lim n. = n
                                                                    (A. 7)
which corresponds tQ the vanishing first and second derivatives on the
left=hand side of Eq. A.5.  Solving Eq. A.5 subject to Eqs. A.6 and A.7,
we can obtain by elementary methods the result
            n
                                           - exp
                                                         r -
                                                    (A. 8)
                                                                     175

-------
 which closely resembles the classical Debye equation for an ion in an
 electrolyte solution.  Equation A.8 expresses an exponential decay from
 the ambient level into the surface of the particle.  The e-folding distance
 is approximately the length a typical molecule travels before it reacts.
 At small fractions of the e-folding distance from the particle surface,
 the reciprocal r-dependence dominates.  The coefficient of the decay
 term brings in the influence of the surface reaction.

       It is instructive to examine  limiting cases of the surface concentra-
 tion by setting  r = r   in Eq. A.8.  For a very small diffusion coefficient
 (D) the molecules of species  i  are insulated from the surface by a thin
 film.  At the surface the concentration is zero because every molecule
 that penetrates the film is consumed at the surface on arrival.   This is
 the diffusion-controlled limit, and it results in a zero surface concentra-
 tion.  At high values of  k  ,  the  presence of the surface or the diffusion
                            o
 process has very little effect  and  we have approached the homogeneous
 reaction limit.  The free-molecule  limiting case where surface reaction
 rate is the controlling feature occurs with very small reactive  particles.
 Thus if  rQ « D/kg  and if  k   »  k D ,  this limit is approached.

       For cases of interest in  air  pollution, it is useful to compare the
 magnitudes of the terms in Eq.  A. 8.  Assume that the mean particle
 diameter is 0.5 ym.  The diffusion  coefficient is of the order of 0.1 cm2/s,
 and for active species in photochemical air pollution  k  ~ 0.004 s   .
                                                         o
 The least certain of all the constants is the surface reaction coefficient
 kg .   Kinetic theory sets  the  upper  limit  on   k   by the frequency of
 molecular collisions  T.  ,  on  a  unit area.
s
                                                                    (A.9)
176

-------
where       m  = Molecular mass of jLth species
             k = Boltzmann's constant
             T = Absolute temperature
Now  k   can be derived directly from Eq. A.9 by applying a collision
efficiency  n .   This is the fraction of collisions that result in surface
reaction.  The value of  n , a measure of chemical surface efficiency, is
determined by the surface coverage of reactant partners and by the activa-
tion energy required to make a reaction occur.  Combining Eqs. A.6 and A.9
with the concept of collision efficiency, we obtain
                                                                   (A. 10)
For nitric oxide at 300°K, this gives a value of ~(1.2 x 10 n) cm-s~ .
The measurements of Aldaz ahow values of  ri  for ozone on active surfaces
               -4
in excess of 10   (for the case of juniper).   At the time of this writing,
                                                                 -4
we do not have values available for oxides of nitrogen, but if 10   is
used, then  k   is of the order of unity.
             s
in Eq. A. 8:  (D/k r ) = 3 x 103, (k D)1'2/k  = 2 x 10 2 , and  /D/k  - 5
      Using these estimates, we get the following values for the parameters
                                                                         cm
This indicates that the second term (which is the fractional species de-
ficit in the neighborhood of the particle) is always small and that the
full value of concentration  n   is exposed to the surface.  Thus we
are in the regime that is dominated by surface- reactions even if collision
                                     _2
efficiencies range as high as  n - 10   .  This simplifies particle up-
take estimations considerably because Eq. A.9  can be used directly with
the ambient gas concentration at the surface (n   = n^) if the efficiency
factor is applied to the particle flux  T .

      Let us turn to the task of placing an upper limit on the role of
particles in competition with gas phase reactions in photochemical smog.
To obtain numerical comparisons between the rates, we must specify some
                                                                     177

-------
 gas phase reactions  and assign some concentration values.  Table A.I
 summarizes some typical conditions  on a smoggy day in Los Angeles.
                                TABLE  A.I
 CONCENTRATIONS  (ppm)  AND GAS  PHASE  RATE CONSTANT ASSUMED FOR COMPARATIVE
                                ANALYSIS
Species
Ozone
Nitric Oxide
Nitrogen Dioxide
Oxygen Atom
NO. Dissociation
Rate Constant
Early Time (~6-9 AM)
-3*
5.5 x 10
1 * 10"1
1 x lo'1
_QA
8 x 10
0.22 min"1
Late Time (-midday)
2 x iQ"1
-3*
2 x 10
5 x 10~2
-9*
5 x 10
0.30 min~
  Computed from stationary  state  relationships.
       The footnote  on  the  table  refers  to  the  reaction steps chosen to
 characterize  gas  phase rates.  These  elementary  processes are:
                       11
             hv + NO,
NO + 0
                         12
                     M
            NO +  03  -^ N°2 +  °2
(A.11)

(A.12)

(A.13)
 where  species  M   is  any  collision partner.   This  mechanism comprises
 the  fastest  gas phase reactions  that  influence -the species in Table A.I.
178

-------
For calculating the starred concentrations, we assumed stationarity for
0-atom giving
                 kllCN09
            c. = 1	—                                          (A. 14)
             °   k!2c02CM
and for ozone giving

                  kllCNO
            co
             U3
where  c  denotes ppm concentration of the subscript species.  Rate constants
                                                            —5    —2    -1
for reactions  (A. 12) and  (A. 13) were assumed to be 1.3'2 x 10   ppm   rain
          -1    -1                                                   44
and 40 ppm   min   consistent with some of our modeling calculations..
In a later section, we express caution about using Eq. A. 15 because of possi-
ble turbulence interference effects; however, for the present purposes
we will assume that we have properly averaged concentrations for use in
stationary state calculations.  Since all of the stoichiometric coefficients
in the reaction cycle are unity, stationarity permits us to calculate a
single gas phase reaction rate characterizing all of the transformations
in the cycle.

      The gas reaction rate for reactions (A. 12) and (A. 13) at early
                -2    -1    -1                                  -2    -1
time is 2.2 x 10   ppm   min  , and at late time it is  1.5 x 10   ppm
min   .  To get upper bounds on the surface rates using Eq. A. 9, we will
need to know the surface area per unit volume.  Following Leighton and
        43
Perkins,   we assume  r  = 0.25 ym  and a specific gravity of unity.  For
an aerosol loading of 200 yg/m  (which is typical for Los Angeles smoggy
                                            — fi   9       *^
days) this gives a surface area of 6.08 x 10   cm  per cm  ,  Using  T
for each species from Eq. A. 9, calculations can be made for the collision
rate of each species on the aerosol surfaces.  This gives the upper limit
for the surface reaction effect.  Ratios of surface to gas phase rates
                                                                     179

-------
 are shown for  this  limit  in Table  A.2.   If  the  ground level ozone flux
                      o 7
 measurements of  Aldaz   apply also to aerosol surfaces, efficiency factors
 in the range of  10    to 10~   reduce  all  of  the  numbers in Table A.2 to
 insignificant  levels.   On the other  hand, if  there  is a moderate degree
                                                     _2
 of surface activity on  the aerosol particles  (say 10   collision
 efficiency), we  see from  Table A.2 that  the morning  NO   levels could
                                                        2t
 be seriously affected.
                               TABLE A.2
           UPPER LIMIT OF  (SURFACE RATE/GAS PHASE RATE)  RATIO
Species
Ozone
Nitric Oxide
Nitrogen Dioxide

Oxygen Atom
Early Time (~6-9 AM)
8.5 x 10"1
2.0 x 10+1
1.6 x 10+1
-6
2.1 x 10
Late Time (midday)
4.5 x io+1
6.0 x lo"1
1.1 x 10+1
-6
2.0 x 10 °
 *       -                                    3
  For conditions  in Table A.I and for 200 ug/m  of 0.5-ym-diameter particles.
       This  possibility deserves serious consideration in view of  the
 atmospheric nitrogen balance results we cited above.  In some of  the  smog
 chamber  experiments reported by Gay and Bufalini,  the majority of  the
 oxides of nitrogen loaded initially show up in the solid phase after  only
 a few hours of  irradiation.  Finally, it should be noted from Table A.2
 that  our results agree with those of Leighton and Perkins43 for oxygen
 atoms; namely,  that gas-particle interactions are negligible by several
 orders of magnitudes even if oxygen atoms react at 100% efficiency  with
 the aerosol.  The influence of surface reactions on ozone concentration
 can become  moderate in the later stages of smog formation.
180

-------
      The critical factor in assessing the impact of these findings is
the surface reaction efficiency as well as the order of the reaction.
If these quantities were more accurately known, it would be possible to
tell which of the surface processes must be included.  Not covered in
the above discussion are the large families of organic radicals and
compounds that may also react with aerosol surfaces.  Certainly the
hydrocarbon reaction rates are much smaller than those discussed; there-
fore, heterogeneous reactions could contribute very significantly.  The
intermediates such as  RO, R02,  and HONO and the products like aldehydes
and alkyl nitrates must also be investigated.

      Having identified the significant processes, we can incorporate
them into the air quality model by adding sink terms to the continuity
equation for each species.  Such a term would depend on a rate constant,
the particulate level, and probably the species concentration raised to
some power greater than zero.  These reaction terms could be lumped in
with others.  If, for example, a high degree of correlation were found
between ozone and aerosol levels, the ozone terms could be augmented to
account for heterogeneous removal mechanisms.  There may also exist
product species which become detached from the particles.  If a surface
is coated with loosely bound B-molecules, an impact of an A-molecule on
the surface might be followed by ejection of an AB-molecule.  It is
clear that a great deal of research is yet to be done in this field.

      Before leaving the question of gas-solid interactions, we should
not fail to mention the possible effects of attenuation of incident sun-
light by the aerosol, particularly in the ultraviolet.  Since the photo-
dissociation primary processes are most sensitive to the ultraviolet input,
there may be a significant reduction of the reaction rates without a
commensurate reduction in total incident solar energy.  In one version
of our model, Eppley detector readings can be used directly to get the
               &
rate constants.   This automatically accounts for ultraviolet attenuation
down to ground level.  When such data are not available, any significant
A    '
 This calibration was based on a particular filter system having a 0.3-
 0.4 ym bandpass.
                                                                      181

-------
 reduction in rate constant must be accounted for by solving a  radiation
 transfer equation over the dissociation band of wavelengths.   To  the  best
 of our knowledge, this is not a part of existing atmospheric pollution
 models.  Of course, before extensive development of any new simulation
 logic is undertaken, the class of cases where the effect  is important
 must first be identified.  This done, it will be necessary to  delineate
 the regime with certain parametric criteria.

 A.2.2  Turbulent Fluctuation Interactions
      The second area of needed research is the influence of concentration
 inhomogeneities upon atmospheric reaction rate calculations.   As  an air
 parcel moves over an array of different emission sources, turbulence  folds
 in the gases of different composition.  The nonuniformity manifests itself
 as blobs that become less and less distinct due to the combined action
 of diffusion and reaction.  Attempts to describe such phenomena have  formed
 a body of theory for turbulent fields of scalar quantities.  Unfortunately,
 little supporting evidence in the form of actual observations  is  available
 for the testing of such theory.  We can, nevertheless, utilize some of
 the grosser aspects of the theory for model improvements if some  appro-
 priate experimentation is done.

      The specific problems arising due to fluctuating concentrations  can
 be clarified by considering the reaction between the pollutants   NO   and
 0  .  The reaction of ozone and nitric oxide in polluted air rapidly
 produces nitrogen dioxide and molecular oxygen.   Characteristic reaction
 times for part-per-hundred-million level concentrations of the reactants
 (0^ and NO) might range from several seconds to a minute or more.  This
 reaction is opposed by the photodissociation of NO  which has  a two or
 three minute characteristic reaction time in bright sunlight.  The result-
 ing quasi-equilibrium, therefore, can respond relatively rapidly  to
 changing reactant stoichiometry.  (Reactions (A.11), (A.12), and
 (A.13) in the previous section form the mechanism in question.)   If con-
 centration changes are induced by turbulent mixing, which may be
182

-------
characterized by the same time scales, there is an interaction between
the turbulent fluctuations and the mean reaction rates.

      Here is what is observed:  suppose we station ourselves at some
point and measure the atmospheric reactant concentrations with good time
resolution.  Typical records (see Fig. A.3) will show that there is
significant anticorrelation between  0.  and  NO  concentration.  For
example, one-minute  0_  and  NO  readings from the first 20 minutes of
this sample have a cross-correlation coefficient of -0.742 (see Table A. 3)
Consequently, if we calculate the mean reaction rate from
            dcNO    .
             dt  ~ ~kCNOC0,
                                              (A.16)
we find that the fluctuations introduce a correction into the rate
because
            CNO
   "NO
                                              (A.17)
and
                                                                    (A.18)
where bars denote means and primes, fluctuations.  Insertion of Eqs. A. 17
and A.18 in Eq. A.16, followed by averaging, gives
            dc
              NO
             dt
                   -k
CNOC0,
       1 +
(A.19)
Note that if there is vanishing correlation between the two concentrations,
the parenthetical factor is unity.  Negative correlation will clearly
reduce the reaction rate in Eq. A.19  (the amount depends on fluctuation intensity)
                                                                      183

-------
      0.4 P-
  8   °-1
       o-U
        0400
                                    NITRIC OXIDE
                                                                        00
                                                                        CXI

                     0410             0420

                        TIME OF DAY, EST
0430
    Figure A.3a.   Chemiluminescent  Measurements in New York - 1970
                                                                   (44)
    0.010  r-
    0.009  -
    0.008  -
 LU
 O
 z

 g  0.007
0.006  U

    0400
                         0410             0420

                            TIME OF DAY, EST
                                                        0430
    Figure A.3b.   Chemiluminescent Measurements in New York -  1970
184

-------
                                TABLE A.3




                 FOSD/NEW YORK DATA44 (FIRST 20 MINUTES)
Input Data
ID
Time After 0400 EST
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
Mean Values =
Correlation Coefficient =
X
NO ppm
0.150000
0.117000
0.052000
0.090000
0.057000
0.070000
0.063000
0.097000
0.122000
0.055000
0.080000
0.088000
0.058000
0.067000
0.060000
0.09SOOO
0.064000
0.078000
0.054000
0.080000
-0.742
Y
03 ppm
0.006150
0.006150
0.008120
0.007780
0.008240
0.007620
0.007940
0.007330
0.006420
0.007970
0.009140
0.007710
0.007080
0.007480
0.007980
0.007480
0.008480
0.007300
0.009070
0.007655

                                                                      185

-------
                      9 A
      In earlier work,   we tested the quasi-equilibrium of the  three-
reaction cycle given in reactions (A. 11),  (A.12)i and  (A.13).  It will
be. recalled that ozone quasistationarity requires that
                                                                    (A'20)

neglecting the turbulence effects on reaction  (A.13).  Based on  10-miriute
averaged concentrations, Fig. A. 4 shows thia systematic departure from
Eq. A. 20 in the direction of too high an apparent rate fbr reaction (A.13)
To suppress inaccuracies in small  NO  readings, we have omitted those
values less than (or equal to) one part per hundred million (pphm) .  One
explanation for the departure may be reactions besides (A. 11),  (A. 12) and
(A.13) playing a significant competitive role.  Mode'ling calculations for
these conditions have as yet failed to reveal such reactions;  Another
hypothesis is instrument inaccuracies at high ozone level.   Figure  A. 5
shows the results of back-calculating what the ozOhe meter (MAST) response
would have to be to bias the measurements as suggested in the data  in
Fig. A. 4.  To achieve quasiequilibriumj we wduld need an instrument that
would read 20 or 25 pp'hm when the actual value is 5 or 6.  This extent
of inaccuracy is unlikely especially since it has been reported that
MAST instrument readings are consistently below the true values.  This
correction is even in the wrong direction to support the plausibility of
a response curve passing through the data of Fig; A. 5.  Detailed back-
ground on data analysis and the calculations of rate constants is given
                          24
in ah earlier publication.

      Two possible explanations of the breakdown of  0,,/NO/NO,2   quasista-
tionarity will now be considered.  The first is the obvious possibility
of interference from competing reactions.  The second is the effect of
turbulence on the rate of reaction (A.13).  For these investigations, we
will make use of Our mathematical model of the chemical-kinetic aspects
of photochemical smog.
186

-------
     1.00
     0.75
     0.50
g  0.25


O
z
 O
 z
     0.00
 O



 O  -0.25
    -0.50
    -0.75
    -1.00
    -1.25
                               POINTS WITH[NO] > 1 pphm
                                                                            S:
                                                                            i
                    +    +    +
                       +
                 -H-
                                                LOCUS OF POINTS SATISFYING

                                                QUASI-EQUILBRIUM HYPOTHESIS
          +  .++
                          *+     +
                                                      +     +   .  +
                  1          1
                                     12        16

                                     OZONE, pphm
                                                       20        24        28
Figure A.4.  Quasiequilibrium Teat for  1969 Ground Data  at El Monte—

              High NO Levels
                                                                            187

-------
               0.00   2.50  5.00   7.50  10.00  12.50  15.00  17.50  20.00  22.50  25.00
                                MEASURED OZONE, pphm

  Figure A.5.  Ozone Inaccuracies Needed  to  Explain  the Departures from
               Quasiequilibrium in  Figure A.4
      To determine the effect  of  competing reactions, we used a 14-reaction,
                                                                        n I  ~| Q
10-species model of photochemical smog.   As reported in previous work,   '
the computed concentrations  simulated accurately the experimental work
                     45
of Altshuller, et al.    Using the results of this simulation, we computed
the ratio  k   [NO ]/k   [O.J[NO]   with the rate constants fixed at
            J-J-   ^   J_ J  J            "1*1
k.... = 0.4 min~l  and  k.. _ =  0.4 pphm  min   .  A plot of the logarithm
of the ratio versus ozone concentration  is shown in Fig. A.6.  It is
apparent that  the ratio  is different from unity and that the departure
increases with increasing ozone.   However, the maximum value of the logarithm
is only 0.04,  indicating that  the ratio  is very close to unity compared
with the atmospheric deviations shown previously.  It seems reasonable
to conclude from this that the interference from other reactions is
negligible and that  03, NO,   and  N02  closely approach quasiequilibrium
under the static conditions  that  prevail in a smog chamber.
188

-------

o
C)
X
o
z
o1
1
0
o

4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
~ 1
Q
c
•<;
N
+ + ** + ** 6
* * ^
-
-
,

i i i i i i i i i i i i
                     10   15  20   25   30  35   40  45  50   55   60
                                OZONE, pphm
Figure A.6.  Quasiequilibrium in a Simulated Smog Chamber Experiment
      In Fig. A. 6, it should also be noted  that  the  trend  away from
equilibrium is in a different direction from that  found  in the Los  Angeles
data (Fig. A«4).  This may be significant inasmuch as  it implies  an
excess of  N09  and a deficiency of  NO  and  0_  .  We are well aware
that this effect could be due to the model.  Nevertheless,  if  the effect
is accepted as real, this implies that what was  observed under atmospheric
conditions is probably not due to interfering reactions .   This is in
                                                   46
agreement with a statement by Schuck and Stephens,   who claim that
quasiequilibrium holds in the presence of competing hydrocarbon reactions
inasmuch as the rates for these reactions are very low.  However, they
go on to claim that quasiequilibrium holds  in the  Los  Angeles  atmosphere,
but offer no evidence to support their assertion.  Finally, we note that
the pair of reactions

                             °
            N0
               °
           '2    3
         NO + N0n
N03
2NOr
                                                                   189

-------
suggested by Leighton42 as likely candidates for interference, would pro-
duce results that contradict the data of Fig.  A.4.   In fact, these
reactions would perturb quasiequilibrium in the direction indicated in
Fig. A..6.

      The above arguments are based on the premise  that the smog chamber
experiment represents the principal chemical processes that really occur;
in the atmosphere.   Many more known (as well as unknown)  chemical inter-
actions take place in urban air and thus it is always possible that some
reactions may be interfering with the  0,/NO/NO   cycle in a manner which
produces the observed results.

      It is apparent from Eq. A.19 that turbulent fluctuations have the
effect of modifying the rate of reaction (A.13).  The apparent validity
of the quasi-equilibrium hypothesis is going to be  affected if we use
the product of time-averaged concentrations instead of the time average
of the product of concentrations.  This notion is in agreement with
Leighton's caveat about the effect on stationarity  of rapid changes in
concentration.

      Having identified the turbulence interference problem in air pollution
modeling, let us now consider some methods of attacking it.  To provide
further insight into the problem, experimental data like those shown in
Figs. A.3 and A.4 must be obtained simultaneously;  namely, time-resolved
concentrations of  0_, NO, NO-, temperature, and ultraviolet input.

      The monitoring station should be located at each of two types of
urban environment:   one in an area dominated by distributed transportation
emission sources and another generally downwind from intense nitric oxide
sources).  The first provides low fluctuation tests and the second, high
fluctuation tests.   The data obtained in such a program would provide
direct evidence as to whether the apparent deviations from quasiequilibrium
were due to fluctuation interferences.  Diagrams like Fig. A.4 could be
 190

-------
constructed both from time-averaged reaction rates and from rates derived
from time-averaged conceit rations.  Comparisons would test explicitly the
hypotheses advanced o: :va.

      Another phase of needed experimental research must be done in the
laboratory.  Its objective is v.h-.: construction of a theoretical approach
for inclusion of rate corrections in computer simulation models.  The method
of attack in the experiment must be designed to yield fundamental informa-
tion on turbulent flow reaction rates.  Thus it must incorporate a well-
understood (necessarily Sifaplt) reaction system in a controlled turbulent
mixing process such that react ion-time/mixing-time ratios are adjustable
and near unity.  This difficult set of specifications is being approached
in an experiment in progress at TRW Systems under the direction of
           47
R. G. Batt.    It involves the  N?0,  dissociation reaction in a free
shear layer in a low speed wind tunnel.  Optical probes are employed for
concentration measurements in addition to conventional aerodynamic instru-
mentation.  Other tunnel experiments for studying the effect may also be
conceived.  An expected end result of the experiment is a measure of steady
concentration fields and velocity fields as well as the ensemble of
turbulent statistical properties.

      Theoretical treatments based on both the field and lab findings must
be undertaken in parallel in order to give model modifications that account
for any mixing interferences with chemical reactions.  Theoretical inter-
pretation of the laboratory experiments will give a detailed treatment
of a reactive plume mixing into a background flow under steady conditions.
Coupled with the flow field through time-averaged velocity and concentra-
tion is a fluctuation budget equation like

            3fi
            ?r-     "                 ' e                        (A-21)
                                                                      191

-------
where       f  - c'2 ,  the mean square fluctuation of species  i
            D  = A diffusion coefficient for fluctuation transport
             D * Turbulence diffusivity
            E  = Dissipation of fluctuations

      This equation for total fluctuation content might incoporate the
statistical approaches of Bugnolo   and Corrsin      in integrated forms.
It holds for a reaction that is second-order in species  i .  As an
approximation, one might assume  D  ~ D .   The dissipation term consists
of two components, one due to diffusive smearing of fluctuations   and
another due to reactive destruction of fluctuations.    We can approximate
these by
               - 12    f  +                                        (A.22)
which expresses the two effects respectively with its two terms.  The
new symbols are defined as follows :
            D  = Molecular diffusivity
            A, = Microscale of concentration fluctuations
            k. = Second-border reaction rate constant for ith species

      Application of Eq. A. 21 to the laboratory results will show how
chemical dissipation of species fluctuations is influenced by chemistry.
This effect then feeds into the equations for mean concentrations through
the reaction rate terms (e.g., Eq. A. 19).  The validity of the theoretical
approach can be checked by computer modeling of the mean and fluctuating
concentrations in the experiment.  A key aspect of this validation procedure
is a knowledge of the transport and reaction coefficients.  Consequently,
it is essential to keep both the chemical system and the flow field as
simple as possible rather than to attempt to construct a physical scale
model of an urban airshed.
192

-------
      Applications of the research findings in air pollution simulation
models will require considerable simplification because inadequate know-
ledge of coefficients and complexities of geometry make an elegant treatment
inadvisable; however, the research is needed to tell us where the approxi-
mations are valid.  Quantitative implementation of rate corrections can
take the form of gross parameters like a plume gradient criterion to tell
when a correction is needed and a mixing delay time that is a function
of scale length and velocities.  The mixing delay time in turn can be
introduced into a correction factor that is formally applied in the
chemical rate portion of the simulation logic.

      All of the details of these formulations are yet to be worked out;
hence, it seems imperative that this be investigated in the immediate
future.  The large discrepancies observed between our notions of the
inorganic kinetics and the observations cannot be ignored.  Until their
sources are discovered and rectified in the simulation models, it seems
unwise to mount extensive validation efforts that repeatedly apply existing
approaches to more and more time-averaged air quality data.
                                                                     193

-------
A. 3   SYSTEMS IMPLEMENTATIONS OF PHOTOCHEMICAL/DIFFUSION MODELS

A.3..1  Few Runs/High Fidelity
      For evaluation studies of emission source contributions, the
influence of specific control measures can be determined by running  only
a  few simulations.  This approach yields the marginal decrease in air
quality ascribable to a particular emitter and, therefore, indicates what
marginal improvements might be bought by imposing controls.  Most frequently
this type of inquiry is served adequately by the simulation of relatively
few pollution scenarios.  Choices from an array of technological alterna-
tives for controls will be the result of these simulations; therefore,
the model should have high fidelity for this application.

      An example of the "few runs/high fidelity" mode of  air quality
simulation is our study of the influence of morning vehicle start emissions
                      13
on photochemical smog.    We sought the answer to the question "What
degradation of air quality is directly ascribable to motor vehicles
starting up in the morning?"  Despite many statements to  the contrary,
the answer is not merely the fractional emissions due to  cold starts.
Diffusion and chemistry introduce non-linearities which preclude simple
scale-up.  To investigate the air quality effects, we ran simulations with
and without starting emissions.

      The model was used with aerometric data from the Los Angeles basin
to study the buildup of air pollution as it is affected by starting
emissions.  The procedure relates meteorological factors, time/space
traffic distributions, and ultraviolet solar radiation with the photo-
chemical atmospheric mechanisms involved in air pollution.  (Averaging
over the daily activities of motor vehicles may not give  an adequate
description of the most severe conditions.)

      Table A.4 shows two of the main findings from the  simulation.
First, that air quality effects vary with pollutant and,  second, that
194

-------
                               TABLE A. 4
                AIR QUALITY EFFECTS FOR 1974 TRAJECTORY
(Ratios of concentration with cold-start to concentration without cold-
                                 start)
Time
1400 hours
1400 hours
Peak
Species
°3
CO
Spatially
Uniform Start
Distribution
1.039
1.024
1.125
Decentralized Start
Distribution
1.042
1.026
1.136
the geographical distribution of starts has no significant effect.  (The
density of morning starts for the nonuniform cases was assumed to be
three times as high at the outer edge of populated areas as it was at the
Federal building downtown varying linearly in a radial direction.)  The
numerical comparisons drawn from the table depend on the fidelity of the
model.  Some significance can be attached to the larger effect on peak
CO  than that on photochemical pollutants.  The combined action of reaction
and diffusion attenuates the start-effect for  0   and  N02 .   These pat-
terns emerged after only a very few simulations were carried out.

A.3.2  Moderate Number of Runs/Moderate Fidelity
      Studies of local problems over a wide range of conditions demand
more runs than the example described above.  On the other hand, the
reduction in scale from regional to local permits us to use a less detailed
physical and chemical formulation from a simplified version of the photo-
chemical/diffusion model.  Exemplifying this type of approach is the analysis
of air quality impact for a proposed high-capacity roadway.  The larger
number of cases arises from a multiplicity of factors derived from a many-
dimensioned parameter space.  For several miles of roadway near an urban
area, background source intensity and receptor sensitivity might both vary
widely.  Hourly traffic loadings change sharply from hour to hour and
                                                                      195

-------
meteorological conditions can exhibit large seasonal variations.   Every
year,  the emission characteristics of the vehicle population are  altered
by  the replacement units that have current control  systems.   Finally,  the
impact of the roadway can be fairly assessed only if we  simulate  the route
corridor with the facility and compare this with the alternative  of
traffic diversion over the surrounding network of surface  streets.   Indeed,
these  variations of parameters multiply into literally hundreds of
specific cases.

       An example of a reduced-fidelity model suitable for  this task  is
LAPS,  a code that we have recently developed and put into operation.
The key simplification leading to the efficiency features  of the  model
is  the choice of coordinate system.  The Lagrangian frame  of reference
is  chosen so that the downwind distance coordinate  is replaced with  time,
with air parcels traced through a streamline system.  This coordinate
system is illustrated in Fig. A.7.  Streamline curvature can be neglected
in  the local areas of freeways; hence, for the free flow above the
roadway, the air parcel representations are planar  control surfaces  moving
along  streamlines.  Each control surface that sweeps with  the wind along
the streamlines has superimposed upon it a spatial  grid.   This grid  con-
sists  of intervals in the vertical and crosswind directions.  Finite-
difference methods are used for the vertical diffusion differential
equations.  In particular, the Crank-Nicolson technique  is employed.  For
diffusion in the lateral direction, Gaussian dispersion  is employed.
Horizontal diffusion, therefore, is treated as an algebraic  correction
that runs concurrently with the finite-difference solutions  of the vertical
diffusion equations, permitting spatial variation of the vertical
diffusion coefficient.  An optional feature of LAPS is the inorganic
portion of the smog chemical mechanism involving the  NO/0 /NO,, cycle.
With two dimensional diffusion and limited chemistry, LAPS is considerably
faster running than our regional code DIFKIN.
—-          —
  Local Air Pollution Stimulator.
  DIFfusion and KINetics.
196

-------
                            •— Ax = uAt

                            MOVING CONTROL PLANE
                   Figure A.7-  LAPS Coordinate System
      Inputs to the program consist of wind direction, wind speed, vertical
and horizontal diffusion parameters and the whole set of descriptors that
characterize emission sources.  Output values of concentrations as func-
tions of time are obtainable by specifying what receptor locations are
of interest.  For example, in a particular environmental impact study,
the concent-ration on a nearby school playground might be desired.  If that
were the case, the model could give a history of concentration at that
location.

      Stagnant conditions are calculated by determining the dispersion
from the roadway under zero wind (but not zero diffusion).  This  is done
by centering the mesh on the roadway and computing the pollutant  spread
in a plane normal to the road centerline.
                                                                      197

-------
      A set of three typical cases serves to demonstrate the capability
of the GRC model to simulate the dispersion of nonreacting pollutants in
the vicinity of a freeway with various wind directions.   The freeway, a
six-lane depressed section shown in Fig.  A.8, was assumed to carry 100,000
vehicles/day.  Using the geometry shown in Fig. A,9, a wind speed of
1.3 mph was simulated at angles (6)  of 0°,  30°, and 90°  to the roadway.
The resulting  CO  distributions predicted by the model  are shown in
Fig. A.10.

      As an example of the speed of the LAPS model,  the  problem of computing
the  CO  concentration at 100 points in a vertical plane normal to a free-
way for 8 hours of real time required 3 minutes of central-processor time
on a Control Data 6400 computer.  The computation interval for this problem
was 0.1 minute; the output intervals were 1 minute for ground concentra-
tion profiles and 30 minutes for vertical data maps.

      To illustrate how the fidelity of the model captures the ozone-
depression effect near a freeway, we can examine the same input conditions
as above for a one mile per hour crosswind.   Background  levels are
chosen to be typical of the Los Angeles basin well into  a midsummer day.
Figure A.11 shows the reduction in ozone as the nitric oxide from the
vehicles mixes in and feeds the  NO  + 0«  -»• NO- + 0   reaction.   Downwind
of the roadway, ambient air dilutes  the emissions with air containing a
higher level of ozone.  Consequently,  the nitric oxide decreases back
down to a level near ambient.

      Since the chemistry is handled by a simple algorithm based on quasi-
stationary state, this technique is  suitable for running the many cases
we will encounter for future problems of  air quality impact evaluations.
Turbulent mixing effects of vehicle  wakes are treated internally by
aerodynamic formulas for the locally enhanced diffusion  coefficients.
Source geometries are not restricted to roadways.  Airports,  central power
stations and other source concentrations  can also be treated as inputs to
the LAPS code.
198

-------
Figure AiS.  Cross Section of Depressed  Six-Lane  Freeway
                    70 m SOURCE STRIP WIDTH
                                                         to
                                                         o
                                                         03
                                                         to
      Figure A.9.  Wind-Oriented Coordinate  System
                                                              199

-------
                .
               O

               I
                     9 = 0'
                             I— ROADWAY -—|
                                                    ZERO BACKGROUND
                                                    WIND SPEED = 1.3 mph
                                                    1971 VEHICLE MIX
                                                    TRAFFIC: 100,000 VEHICLES/DAY
                                                                      0= 90°
                                   0       50      100      150       200

                               NORMAL DISTANCE FROM ROADWAY CENTERLINE, m
   Figure A.10.   CO  Concentration  Profiles  Normal to Roadway  at  Various
                    Wind Aspect  Angles
                         100
                       I
                       g  10
                       z
                             EMISSIONS MODEL: .CALIFORNIA VEHICLE MIX IN 1971
                             ROADWAY DATA: EIGHT LANE FREEWAY CARRYING
                                         100,000 VEHICLES/DAY
                                                   1 mph WIND
Figure A.11.   Ozone and Nitric  Oxide in an Air Mass Moving Over  a Roadway
200

-------
 A-3.3  Many Runs/Low Fidelity
       The deeper we penetrate into  future problems of pollution abatement,
 the more respect we gain for  the  sometimes  subtle, but tightly coupled
 problems of economic externalities  and political realities.  Staggering
 social costs are now beginning  to appear on the horizon as the burden we
 must  bear to improve the environment.  The  job of standard-setting, which
 has nearly run  through  its first  round, only sets quantitative targets
 in  this  quest for better air  and  water quality.  Actions to achieve the
 goals  result in multiple feedbacks  of money and material, that may either
 damage or  benefit humanity depending on what value system we adopt.
                                                        C Q
      Models such as  the  Implementation Planning Program   have endeavored
 to relate  all of these  factors  for  the steps needed to improve air quality.
 Since the  air pollution  simulation  is only  a small part of an ensemble
 of logical functions, it must, of necessity, be highly simplified.  The
 sacrifice of fidelity in these applications is believed to be well justi-
 fied by the need for many runs.  For its scope (particulates and  SO
 from stationary  sources) and  for the limited availability of data, the
 IPP simulation did an admirable job  of  laying out the significant issues.
 Its weaknesses highlighted the specific  needs for better quantitative
 information, particularly in  the area of  damage assessment and the
 inclusion of mobile sources.

      Future models of this type will need chemistry in the air module
because nearly all pollutants of interest are reactive.   Refined cost figures
will be needed in the economic modules to reflect advances in control tech-
nology.  Aggregate damage indices may be incorporated in the form of
 summations or in integrals such as
where       D  . = damage index due to the i_th species acting on the jth
                  type of receptor
                                                                     201

-------
              p = population density of jth receptor
            $   = impact function for ith species on jth receptor
             13                       -              -
             c. = concentration of ith species
              x = location
              t = time
              a = area

This will be an objective index that goes beyond the mere question "Does
it or does it not exceed ambient standards?"  It can be tied to ambient
standards by normalizing the impact function.  Say, for example that the
impact of pollutant  i  on receptor  j  goes up like the nth power of  c
so that the expression
permits a comparative assessment of all species if  c,    is the ambient
standard.  Contributions can be collected by forming a total damage on
receptor  j  by summing all of the  D    over all  i.

      Now the use of this index in systems analysis requires a model
that spreads over all of the receptors, not just a few monitoring stations.
An important feature of future work will be reducing the three-dimensional,
time-dependent air pollution simulations to a usable size.  One key to
this reduction lies in a "black box" chemical model.  The black box functions
in a chemical sense like the Maxwell Demon operates in the kinetic sense;
i.e., while the Demon sorts out molecules in a certain energy range, the
black box converts  NO  molecules to  N0_  molecules in photochemical smog
according to the hydrocarbon decomposition products that are present.
The black box transfer function will need parameters depending on reactivity j
HC/NOx  ratio and, perhaps, temperature, humidity, arid aerosol levels.
The function, which must be obtained from curve fits of extensive kinetie-s
202

-------
simulations, will replace a dozen or more coupled differential equations.
If the Lagrangian fluid dynamic frame is retained in the systems model,
this simplification is essential, since many, many, simulations will be
required to supply values for the integrand of  D.   over an urban region.

      Future applications of this simplified air quality model will find
it coupled with economic input/output models, transportation network simu-
lations, land use models, and energy managment models (possibly all at
the same time).  Of course if this line of development is allowed to grow
unchecked, it invariably leads into the fantasy world of some systems
analysts who are unafraid to take on the universe.

      Pollution abatement strategies are limited more by human institutions
and resources than by technological advances.  Thus, an important class
of future problems in simulation will involve live participants in the
loop.  Invariably, actual decisions on incremental changes take directions
other than those selected by system optimization procedures.  This happens
because the latter procedures are incomplete with respect to variables
and constraints.  With people operating in a simulated abatement scenario,
the quantitative models are used as feedback generators, but they do not
control the action.  This application of gaming has been instituted in
the APEX and CITY exercises already under evaluation by the Environmental
Protection Agency.  They will serve as training devices as well as
testbeds for policy experiments.

      The usefulness of these games will depend heavily on the credibility
of the feedback models that tell the decision maker how much his decision
costs and what effects it will have in many sectors including that of
environmental quality.  Nevertheless, these models must simulate many
alternatives rapidly to enhance the value of the game even if it must be
at the expense of some fidelity.
                                                                     203

-------
 A. 4    SUMMATION
       Tight implementation schedules and tough regulations have been  laid
 out to abate pollution.  The urgency of policies already adopted has  led
 to hasty actions in some cases.  It is imperative that the research
 community concentrate on some of the ill-defined areas that may contain
 the key elements of understanding the consequences of air pollution
 control decisions.

       Mathematical simulation models at the very least provide a logical
 framework that highlights the unknowns.  At the most, they will serve as
 predictive tools in evaluating the impact of rulemaking and decision-making.
 As we  have moved into the science of air pollution simulation, we have
 discovered some serious deficiencies in present-day approaches.  The neglect
 of heterogeneous processes omits possibly the most important cleanup
 processes for oxides of nitrogen.  Unexplained shifts in HC/NO  ratios
                                                              2t
 are observed in morning air samples.  The largest discrepancies occur on
 the worst smog days.  Turbulent chemical kinetics are untouched in con-
 temporary simulation approaches.  These effects may well be responsible
 for apparent shifts several hundred percent away from quasiequilibrium
 states believed to govern the major pollutants., ozone and oxides of
 nitrogen in urban environments.

       This paper has assembled some suggestions for the attack on each
 problem.  These attacks are based on systematic gathering of observational
 evidence followed by careful data analysis feeding into refinements or
 corrections to the simulation models.  Unfortunately, many uses will be
 demanded of the models before these issues are faced.  Some of the expres-
 sions  derived in the preceding sections will serve at least as criteria
 for setting rough levels of confidence in existing models.  The urgency
 of the problems at hand must be used as a stimulant for the needed research
 rather than an excuse to overlook the deficiencies in our present under-
 standing of the problems.
204

-------
      With  the mature development of certain air quality simulation tech-
niques, a hierarchy of models will emerge embracing a wide range of fidelity
levels.  The large variety of applications anticipated places demands on
operating speed in some cases while other cases are characterized by needs
for precision in certain types of predictions.  The examples cited in the
closing section highlight these differences.

      Despite consultant advertising that touts "complete modeling capa-
bilities for air pollution simulation," we can still see some exciting
possibilities for new research.  This new work will fill significant
voids in the basic fiber of which the "complete" models are made.  But
simply building models is not enough.  They must be molded into useful
forms by creating special algorithms, by inventing approximations, by
designing input/output structures that meet the dynamic demands of air
quality management.
                                                                     205

-------
                               REFERENCES
1-    A.Q. Eschenroeder and J.R. Martinez, Concepts and Applications of
      Photochemical Smog Models, General Research Corporation TM-1516,
      June 1971 (to be published as an ACS Monograph in Advances in
      Chemistry).

2.    A,P. Altshuller and J.J. Bufalini, "Photochemical Aspects of Air
      Pollution:  A Review," Photochemistry and Photobiology, Vol. 4, 1964,
      pp. 97-146.

3.    A.P. Altshuller and J.J. Bufalini, "Photochemical Aspects of Air
      Pollution:  A Review," Environmental Science and Technology, Vol. 5,
      No. 5, January 1971, pp. 39-64.

4.    K. Westberg and N. Cohen, The Chemical Kinetics of Photochemical
      Smog as Analyzed by Computer, AIAA Third Fluid and Plasma Dynamics
      Conference Paper No. 70-753, Los Angeles, June 29-July 1, 1970.

5,    T.A. Hecht and J.H. Seinfeld, '^Development and Validation of a
      Generalized Mechanism for Photochemical Smog," Env. Science and
      Technology, Vol. 6, No. 1, January 1972., pp. 47-57.

6.    H. Niki, E.E. Daby and B. Weinstock, "Mechanisms of Smog Reactions",
      Advances in Chemistry, 1972  (in press).

7.    D.H. Stedman, E.D. Morris, Jr., E.E. Daby, H. Niki, and B. Weinstock,
      The Role of OH Radicals in Photochemical Smog, American Chemical
      Society Division of Water Air and Waste Chemistry, Chicago, Illinois,
      September 13-18, 1970.

8,    J.R. Holmes, A.D. Sanchez, and A.H. Bockian, Atmospheric Photochemistry:
      Some Factors Affecting the Conversion of NO to NO  , Pacific Con-
      ference^ on Chemistry and Spectroscopy, San Francisco, October 6-9,
      1970.

9.    B.W. Gay and J.J. Bufalini,  "Nitric Acid and the Nitrogen Balance
      of Irradiated Hydrocarbons in the Presence of Oxides of Nitrogen,"
      Environmental Science and Technology, Vol. 5, No. 5, May 1971,
      pp. 422-425.

10.   M. Dodge, private communication, May 26, 1972.

11.   M. Dodge, private communication, July 11, 1972.
                                                                     207

-------
12.   P.L. Hanst, "Mechanism of Peroxyacetylnitrate Formation", J. Air
      Pollution Control Association, Vol. 21, No. 5, May 1961, pp. 269-271.

13.   J.R. Martinez, R.A. Nordsieck, and A.Q. Eschenroeder, Morning
      Vehicle-Start Effects on Photochemical Smog, General Research
      Corporation CR-2-191, June 1971.

14.   F. Stuhl, private communication, Ford Motor Co., Scientific Research
      Staff, June 1, 1972.

15.   J. Anderson, private communication, University of Pittsburgh,  June
      7, 1972.

,16.   E.A. Sutton, "Chemistry of Electrons in Pure-Air Hypersonic Wakes,"
      AIAA Journal, Vol. 6, No. 10, October 1968, pp. 1873-1882.

17.   K. Schofield, "An Evaluation of Kinetic Rate Data for Reactions of
      Neutrals of Atmospheric Interest," Planetary and Space Sciences,
      Vol. 15, 1967, pp. 643-670.

18.   P.A. Leighton, Photochemistry of Air Pollution, New York Academic
      Press, 1961.

19.   G. Schott and N. Davidson, "Shock Waves in Chemical Kinetics:  The
      Decomposition of NO,, at High Temperatures," Journal of American

      Chemical Society, Vol. 80, p. 1841 (1958).

20.   S. Jaffe and H.W. Ford, "The Photolysis of Nitric Acid at 3660A and
      25°," Journal of Physical Chemistry, Vol. 71, p. 1832 (1967).

21.   W.R. Greiner, "Hydroxyl Radical Kinetics VI., Reactions with
      Alkanes in the Range 300°-500°K, J. Chem. Phys., 53, 1970, pp.
      1070-1076.

22.   J. Heicklen, K. Westberg, and N. Cohen, The Conversion of NO to NO

      in Polluted Atmospheres, Pennsylvania State University Center  for
      Air Environment Studies Publication 115-69, July 1969.

23.   K. Westberg, N. Cohen, K.W. Wilson, "Carbon Monoxide:  Its Role in
      Photochemical Smog Formation", Science, Vol. 171, March 12, 1971,
      pp. 1013-1015.

24.   A.Q. Eschenroeder and J.R. Martinez, Further Development of the
      Photochemical Smog Model for the Los Angeles Basin, General Research
      Corporation CR-1-191, March 1971.
                                                       j
25.   A.P. Altshuller, "An Evaluation of Techniques for the Determination
      of the Photochemical Reactivity of Organic Emissions", JAPCA,  Vol.
      16, No. 5, May 1966, pp. 257-260.
208

-------
26.   A.Q. Eschenroeder and J.R. Martinez, Analysis of Los Angeles Atmos-
      pheric Reaction Data from 1968 and 1969, General Research Corpora-
      tion CR-1-170, July 1970.

27.   A.Q. Eschenroeder and J.R. Martinez, Mathematical Modeling of Photo-
      chemical Smog. General Research Corporation IMR-1210, December 1969.
      Also a paper presented at the AIAA Eighth Aerospace Sciences Meet-
      ing, January 1970.

28.   J.K. Angell, D.H. Pack, L. Machta (R. Dickson, and W.H. Hoecker),
      "Three-Dimensional Air Trajectories Determined from Tetroon Flights
      in the Planetary Boundary Layer of the Los Angeles Basin", Journal
      of Applied Meteorology. Vol. 11, No. 3, April 1972, pp. 451-571.

29.   M.A. Estoque, "A Numerical Model of the Atmospheric Boundary Layer",
      Journal of Geophysical Research, Vol. 68, 1968, pp 1103-113.

30.   C.R. Hosier, "Vertical Diffusivity from Random Profiles", Journal
      of Geophysical Research, Vol. 74, No. 28, December 20, 1969, pp.
      7018.

31.   A.Q. Eschenroeder and J.R. Martinez, A Modeling Study to Characterize
      Photochemical Atmospheric Reactions to the Los Angeles Basin Area,
      General Research Corporation CR-1-152, November 1969, p. 18.

32.   F. Pasquill, Atmospheric Diffusion, Van Nostrand, London, 1962.

33.   J. Angell and D. Pack, "Mesoscale Diffusion Derived from Tetroon
      Flights", USAEC Meteorological Information Meeting, Atomic Energy
      of Canada Limited, Chalk River, Ontario, AECL-2787, September 1967.

34.   P. Roberts, P. Roth, and C. Nelson, Contaminant Emissions in the Los
      Angeles Basin—Their Sources, Rates, and Distributions, Systems
      Applications, Inc., Report 71 SA1-6, March 1971 (Appendix A).

35.   Final Report on Phase I, Atmospheric Reaction Studies in the Los
      Angeles Basin, Vols. I and II, Scott Research Laboratories, June
      30, 1969.

36.   F.L. Ludwig, A.E. Moon, W.B. Johnson, R.L. Mancuso, A Practical
      Multipurpose Urban Diffusion Model for Carbon Monoxide, Stanford
      Research Institute, September 1970.

37.   L. Aldaz, "Flux Measurements of Ozone Over Land and Water," Journal
      of Geophysical Research, Vol. 74, No. 28, December 20, 1969, pp.
      6943-6946.

38.   R.E. Inman, and R.B. Ingersoll, "Note on the Uptake of Carbon Mono-
      xide by Soil Fungi," Journal of the Air Pollution Control Association,
      Vol. 21, No. 10, October 1971, pp. 646-647.
                                                                     209

-------
39.   F.B. Abeles, L.E. Craker, I.E. Forrence, and G.R. Leather, "Fate of
      Air Pollutants:  Removal of Ethylene, Sulfur Dioxide arid Nitrogen
      Dioxide by Soil," Science, Vol. 173, No. 4000, September 3, 1971,
      pp. 914-916.

40.   L.A. Ripperton, and F.M. Vukovich, "Gas Phase Destruction of Trdpo-
      spheric Ozone," Journal of Geophysical Research, Vol. 76, No. 30,
      October 20, 1971, pp. 7328-7333.

41.   D.A. Lundgren, "Atmospheric Aerosol Composition and Concentration
      as a Function of Particle Size and of Time," Journal ~of the Air Pol-
      lution Control Association, Vol.  20, No. 9, September 1970, pp.
      603-608.

42.   J. Cholak, L.J. Schafer, D.W.  Yaeger, and R.A. Kehoe, "The Nature
      of the Suspended Matter," Chapter VIII in Air Pollution Foundation
      Report No. 9, An Aerometric Survey of the Los Angeles Basin,  August-
      November 1954.

43.   P.A. Leighton, and W.A. Perkins,  Photochemical Secondary Reactions
      in Urban Air, Air Pollution Foundation Report No. 24, August  1958.

44.   E. Daby, Ford Motor Company, private communication.   (Full results
      to be published in Journal of Air Pollution Control Association,
      Vol. 22, No. 4, April 1972.)

45.   A.P. Altshuller, S.L. Kopczynski, W.A. Lonneman, T.L. Becker  and R.
      Slater, "Chemical Aspects of the  Photo-oxidation of the Propylene-
      Nitrogen Oxide System," Environmental Science and Technology, Vol.  1,
      No. 11, November 1967, pp. 889-914.

46.   E.A. Schuck and E.R. Stephens, "Oxides of Nitrogen," Advances in
      Environmental Sciences, Vol. 1, New York,  John Wiley and Sons,
      1969, pp.  73-118.

47.   R..G. Batt, T. Kubota and J. Laufer, Experimental Investigations of
      the Effect of Shear-Flow Turbulence on a Chemical Reaction, AIAA
     ^Reacting Turbulent Flows Conference Paper,  San Diego, June 1970.

48.   D. Bugnolo, "Effects of a 'rMixing-in-Gradient'  on the Spectrum of
      the Electronic Density in a Turbulent Weakly Ionized Gas," Journal
      of Geophysical Research, Vol.  70, No. 15,  August 1965,  pp.  3725-3734.

49.   S. Corrsin, "Statistical Behavior of a Reacting Mixture in Isotropic
      Turbulence," Phys. Fluids, Vol. 1, No. 1,  January-February 1958  pp
      42-47.

50.   S. Corrsin, "The Reactant Concentration Spectrum in Turbulent Mixing
      with a First-Order Reaction," J.  Fluid Mech., 1961,  pp.  407-416.
210

-------
51.   S. Corrsin, "Further Generalization of Onsager's Cascade Model for
      Turbulent Spectra," Phys. Fluids, Vol. 7, No. 8, August 1964, pp.
      1156-1159.

52.   J. Hinze, Turbulence;  An Introduction to Its Mechanism and Theory,
      McGraw Hill, New York, 1959, p. 224.

53.   Air Quality Implementation Planning Program, TRW Systems Group
      SN11130, November 1970.
                                                                     211

-------
 BIBLIOGRAPHIC DATA
 SHEET
                     1. Report No.
-R4-73-012a
                                          3. Recipient's Accession No.
4. Title and Subtitle
    Evaluation of  a  Diffusion Model for  Photochemical
     Smog Simulation
                                          5. Report Date
                                              October 1972
                                                                    6.
 . Authorfs)
     A.Q.  Eschenroeder,  J.R.  Martinez,  R.A.  Nordsieck
                                          8. Performing Organization Kept.

                                            No-   CR-1-273
  Performing Organization Name and Address

     General Research  Corporation
     P.  0.  Box 3587
     Santa Barbara,  California  93105
                                           10. Project/Task/Work Unit No.
                                           11. Contract/Grant No.

                                                  68-02-0336
12. Sponsoring Organization Name and Address

     ENVIRONMENTAL  PROTECTION AGENCY

     Research Triangle Park, North  Carolina  27711
                                           13. Type of Report & Period
                                             Covered

                                                  Final
                                                                     14.
15. Supplementary Notes
16. Abstracts  Exten5ive jimprovements  have characterized this evaluation  of the GRC Photo-
 chemical/Diffusion model.   Despite the limitations  of smog chamber experimental data,
 they have served an essential  purpose toward  updating the kinetics portion of the
 model.  Consistency of rates  and  reactivities  is  now achievable using  recently measured
 coefficients  for a wide variety of systems.   Model  methodology revisions  have enhanced
 the realism of the advective  and  diffusive descriptions.  Previous assumptions regard-
 ing transverse (cross-streamline) horizontal  diffusion have been  confirmed by an ex-
 haustive series of parametric  tests.   Photochemical/diffusion validations were success-
 ful for trajectories occurring during four days  of  the 1969 smog  season in Los Angeles.
 Our measure of success is concentration-history  fidelity with a minimum of adjustments
 of diffusion  parameters.  (Chemical  coefficients  were scaled from the  smog chamber
 studies and held fixed for  the simulations carried  out to date).  Future  directions
 for air pollution model development  are discussed in detail in an appendix as informa-
 tion supporting the experimental  recommendations.
 17. Key Words and Document Analysis. 17a. Descriptor

   Air pollution
   Mathematical  models
   Photochemical  reactions
   Diffusion
   Advection
   Trajectories
   Reaction kinetics
   Smog
   Chambers
   Validity
 17b. Identifiers/Open-Ended Terms

   Diffusion  models
   Atmospheric modeling
 17c. COSAT1 Fjeld/Group   13B
 18. Availability Statement
     JTIS-35 IREV. 3-72)
                      Unlimited
19. Security Class (This
  Report)
     UNCLASSIFIED
                                                         20. Security Class (This
                                                            Page
                                                         	UNCLASSIFIED
                                                     [1- No. of Pages

                                                           230
                               G. P. O. 1973 —  746-772 / 4196. REGION NO. 4
                                                                               22. Ptice
                                                                               USCOMM-DC 14952-P

-------
  INSTRUCTIONS  FOR  COMPLETING  FORM  HTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
 Guidelines to Format Standards for Scientific and Technical Reports Prepared by or for the Federal Government,
 PB-180 600).

  1.  Report Number.  Each individually boupd report shall carry a unique alphanumeric designation selected by the performing
     organization or provided by the sponsoring organization.  Use uppercase letters and  Arabic numerals only.  Examples
     FASEB-NS-S7 and FAA-RD-68-09.

  2.  Leave blank.

  3.  Recipient's Accession Number.  Reserved for use by each report recipient.

  4  Title and  Subtitle. Title should indicate clearly and briefly the  subject coverage of the report, and be  displayed  promi-
     nently.  Set subtitle,  if used, in smaller type or otherwise subordinate it to main title.   When a report is prepared in more
     than one volume, repeat the primary title, add volume number and  include subtitle for the specific volume.

  5-  Report Date.  P!ach report shall carry a date indicating at least month and year.  Indicate the basis on which it was  selected
     (e.g., date of issue, date of approval, date of preparation.


  6-  Performing Organization Code.  Leave blank.

  7.  Author(s).  Give name(s) in conventional order  (e.g., John R. Doe, or  J.Robert Doe).  List author's affiliation if it differs
     from the performing organization.

  8.  Performing Organization Report Number.  Insert if performing organization wishes to  assign this number.

  9-  Performing Orgonization Name and Address.  Give name,  street, city, state, and zip code.   List no more than two levels of
     an organizational hierarchy.  Display the name  of the organization exactly  as  it should appear in Government indexes such
     as  USGRDR-I.

 10.  Project/Task/Work Unit Number.  Use the project, task and work  unit numbers under which the report was prepared.

 11.  Contract/Grant Number.  Insert contract or grant number under which report was prepared.

 12-  Sponsoring Agency Name and  Address.   Include zip code.

 13.  Type of Report and Period  Covered.  Indicate interim, final, etc., and,  if applicable,  dates covered.

 14-  Sponsoring Agency Code.  Leave blank.

 15.  Supplementary Notes.   Enter  information not included  elsewhere  but  useful, such  as:  Prepared in cooperation  with .
     Translation of   .   Presented at conference of  . .   To be published in . .    Supersedes . . .       Supplements

 3$.  Abstract.   Include a brief  (200 words or less) factual summary  of the  most significant information contained in the report.
     If 'he report contains a significant bibliography or literature survey, mention it here.

 17.  Kay Words and  Document Analysis,  (a).  Descriptors.  Select from the  Thesaurus of  Engineering  and Scientific Terms the
     proper authorized terms that identify the major concept  of the research  and  are sufficiently specific and precise to be used
     as index entries for cataloging.
     (b).  Identifiers and Open-Ended Terms.  Use identifiers for project names, code names, equipment designators, etc.  Use
     open-ended terms written in descriptor form for those subjects for which no descriptor exists.
     (c).  COSATI Field/Group.  Field and  Group assignments  are to be taken  from the  1965 COSATI Subject  Category  List.
     Since the  majority of  documents are mulndisciplinary in nature, the primary Field/Group  assignment(s) will be the specific
     discipline, area of human endeavor, or type of physical object.  The appHcation(s) will be  cross-referenced with secondary
     Field/Group assignments that will follow the primary posting(s).

 18.  Distribution Statement.  Denote releasability to the  public  or limitation for reasons  other  than security for  example  "Re-
     lease unlimited"  Cite any availability to the public, with  address and price.

 19 & 20. Security Classification.  Do  not submit classified reports to the National Technical

 21.  Number of Pages.  Insert the  total number of pages,  including this  one and unnumbered  pages, but excluding distribution
     list, if any.

 22.   Price.  Insert the price set by the National  Technical Information Service  or the Government Printing Office  if known


ORM  NTIS-33 (REV. 3-72)                                                                                    U SC OMM-DC V4SS 2-P?2

-------